An integrated process‐based model of flutes and tool marks in deep‐water environments: Implications for palaeohydraulics, the Bouma sequence and hybrid event beds

Flutes and tool marks are commonly observed sedimentary structures on the bases of sandstones in deep‐water successions. These sole structures are universally used as palaeocurrent indicators but, in sharp contrast to most sedimentary structures, they are not used in palaeohydraulic reconstructions or to aid prediction of the spatial distribution of sediments. Since Kuenen's famous 1953 paper, flutes and tool marks in deep‐water systems have been linked to turbidity currents, as reflected in the standard Bouma sequence taught to generations of geologists. Yet, these structures present a series of unaddressed enigmas. Detailed field studies in the 1960s and early 1970s observed that flutes are typically associated with thicker, more proximal beds, whilst tools are generally prevalent in thinner, more distal, beds. Additionally, flutes and tool marks are rarely observed on the same surfaces, and flutes are seen to change downstream from larger wider parabolic to smaller narrower spindle‐shaped forms. No model has been proposed that explains these field‐based observations. This contribution undertakes a radical re‐examination of the formative flow conditions of flutes and tool marks, and demonstrates that they are the products of a wide range of sediment gravity flows, from turbulent flows, through transitional clay‐rich flows, to debris flows. Flutes are not solely the product of turbulent flows, but can continue to form in transitional flows. Grooves are shown to be formed by debris flows, slumps and slides, not turbidity currents, and in many cases the debris flows are linked to the debritic component of hybrid flows. Discontinuous tool marks, including skim (bounce) marks, prod marks and skip marks, are shown to be formed by transitional mud‐rich flows. Consequently, the observed spatial distribution of flutes and tool marks can be explained by a progressive increase in flow cohesivity downstream. This model of flutes and tool marks dovetails with models of hybrid flows that predict such a longitudinal increase in flow cohesivity. However, some deposits show grooves preferentially associated with Bouma TA beds, and these are likely formed by flows transforming from higher to lower cohesion, and are present in basins where hybrid beds are absent or rare. The recognition that sole structures may have no genetic link to the later overlying turbidity current deposits, and can be formed by a wide range of flow types, indicates that the existing pictorial description of the Bouma sequence is incorrect. A modified Bouma sequence is proposed here that addresses these points. In utilizing the advances in fluid dynamics since Kuenen's pioneering research, this study demonstrates that it is possible to use flutes and tool marks to interpret flow type at the point of formation, the nature of flow transformations, and the mechanics of the basal layer. These advances suggest that it is then possible to predict the nature of deposit type down‐dip. This new understanding, in combination with further testing in outcrop of the proposed relationships between sole marks and palaeohydraulics, opens up a wealth of possibilities for improving the understanding of deep‐water clastic environments, with implications for developing more complete facies models, assessing subaqueous geohazards and the resilience of seafloor infrastructure, and advancing our understanding of deep‐water sediments as archives of palaeoenvironmental change.


INTRODUCTION
The bases of sandstone beds in deep-water sedimentary successions are commonly ornamented with sole structures of inorganic origin that record the infilling of erosional bedforms generated in the underlying fine-grained substrate. Two categories of sole structures can be identified: scour marks such as flutes formed by turbulent scour; and tool marks formed by objects (tools) within the flow (D _ zuły nski & Sanders, 1962a;Collinson et al., 2006), which are subdivided further into continuous marks (for example, grooves and chevron marks) and discontinuous marks (for example, prod, bounce, skip and roll marks) (D _ zuły nski & Sanders, 1962a). The use of sole structures as indicators of palaeocurrent direction is longestablished (e.g. Hall, 1843), and every geoscience student is trained in their recognition and their value for palaeogeographic reconstruction. However, their wider utility for the interpretation of palaeohydraulic conditions and flow-substrate interactions is limited. This is in stark contrast to aggradational bedforms, such as ripples, dunes, upper-stage plane beds and antidunes, which have been used extensively to provide information concerning processes during deposition in addition to palaeocurrent information (e.g. Harms, 1969;Allen, 1984;Cartigny et al., 2014;Baas et al., 2016a). In large part, the focus on palaeocurrent information from tool marks and flutes reflects the lack of understanding of their formative conditions. Nonetheless, many important observations have been made concerning the distribution and association of flutes and tool marks. Exceptions exist, but flutes are typically associated with proximal locations, and tool marks with distal locations (e.g. Hsu, 1959;Craig & Walton, 1962;Walker, 1967;Lovell, 1969;Ricci Lucchi, 1969b;Slacza & Unrug, 1976;Remacha & Fern andez, 2003;Remacha et al., 2005;Collinson et al., 2006;Collinson & Mountney, 2019). Bed bases with both types are rare (e.g. Crowell, 1955;Wood & Smith, 1958;Sanders, 1965;Collinson et al., 2006), but where present commonly show cross-cutting relationships (Kuenen, 1957;D _ zuły nski & Sanders, 1962a;Enos, 1969a;Ricci Lucchi, 1969a;Draganits et al., 2008;Pyles & Jennette, 2009). Furthermore, although the Bouma sequence depicts flutes and grooves on the base of the T A division, both types are also found under T B and T C beds (e.g. Bouma, 1962;Pett & Walker, 1971;Crimes, 1973; Table 1). These field observations are supported across systems of different ages and tectonic settings, but have proven enigmatic. As such, no process explanations or synoptic models have been presented to explain why flutes and tool marks exhibit these general spatial and temporal variations, or why there are exceptions. A better understanding of the relationship between erosional bedforms and their overlying deposits has profound implications for our general understanding of deep-water systems. In particular, the use of deep-marine sedimentary successions as archives of palaeoenvironmental change, for reducing uncertainty in geohazard assessment, and for determining the resilience of seafloor infrastructure, requires improved understanding of the interactions between flows and substrate conditions, and the formation of erosional bedforms. The most accessible resource for these investigations is the sole structures preserved on the base of deep-water sandstones.
This paper aims to examine the formative mechanisms of flutes and the different types of tool marks. To achieve this aim, the present paper utilizes recent advances in the understanding of transitional flow processes between fully turbulent and laminar flow, and new data on seafloor substrates, to address the following objectives: (i) to discuss flow rheology at the time of sole structure formation, and the likely downdip deposit type; (ii) to use data from modern seafloor substrates to infer likely depths of erosion, nature of the ancient substrate and amount of substrate entrainment; (iii) to reassess the modern depiction of the Bouma sequence, which presents a genetic link between the basal erosive surface and the overlying deposit; (iv) to use the dimensions of grooves to infer flow properties and interpret objects that generated the tools; and (v) to discuss the location under a flow where flutes and tool marks form, which is widely accepted as being under the density current head. This wide range of objectives is integrated into a new synoptic process-orientated model that explains the distribution and association of scour marks and tool marks, which can be employed to transform the information that can be gained from detailed investigations of these sedimentary structures in all modern and ancient deep-water successions.

Classification of sole marks
Sole marks (Kuenen, 1957) are features identified on the base of beds (typically sandstones), formed by infilling of topography that was eroded into an underlying fine-grained substrate, generally cohesive mud. Sole marks can include organic forms such as burrows (Kuenen, 1957), but here the discussion is restricted to inorganic structures. They are typically subdivided into two categories: scour marks formed by turbulent  & Sanders, 1962a;Collinson et al., 2006). Scour marks include obstacle scours, longitudinal scours, mud ripples and gutter casts (e.g. D _ zuły nski & Walton, 1965;Allen, 1984;Collinson et al., 2006), but for brevity this has been restricted to the most common structureflutes (Enos, 1969a). Tool marks are further subdivided into continuous marks (grooves and chevron marks) and discontinuous marks (for example, prod, bounce, skip and roll marks) (D _ zuły nski & Sanders, 1962a). The term 'continuous mark' is not strictly true since grooves and chevrons do have terminations (e.g. Enos, 1969a). However, these structures are typically continuous on the scale of an individual outcrop. In contrast, discontinuous marks occur as individual or groups of structures centimetres to decimetres long (D _ zuły nski & Walton, 1965;Collinson et al., 2006).
Whilst these relationships are firmly embedded in the literature, it is interesting to note that in the quantitative data shown in Table 1 the relationship between Bouma divisions (and by inference downstream distance) and flutes and grooves is less clear. Grooves are more frequently observed in T A beds in Bouma (1962) and Crimes (1973) whilst flutes are more common in Pett & Walker (1971). Such variations suggest that measuring and aggregating Bouma divisions at a given location might be an imperfect surrogate for downstream position in a system, compared with observations from different longitudinal positions (e.g. Lovell, 1969;Slackza & Unrug, 1976;Remacha & Fern andez, 2003). Alternatively, these data may indicate that in some cases there can be a longitudinal variation from tool marks in proximal locations to scour marks downstream. What is clear is that there is a need for more quantitative documentation of the distribution of sole mark types in different settings. In particular, it is desirable to couple such quantitative data on the distribution of different sole structures to modern interpretations of sediment gravity flow processes, deposits and sub-environments.

STRUCTURE AND RATIONALE
To address the formative mechanisms, and thus utility, of flutes and tool marks, first a brief review is included of the fluid dynamics of mud-poor to mud-rich flows, in particular concentrating on transitional flows, between truly turbulent, and fully cohesive, laminar, flows. As will be demonstrated, the different types of flutes and tool marks can be linked to these differing flow types, and thus an understanding of these processes is critical when linking sole structures to flow dynamics. Sole structures are also dependent on the nature of the substrate, and thus the properties of modern seafloor substrates are examined next, and their applicability to older sediments considered. Once this key background on flow types and substrate properties is discussed, each of the sole structures is considered in turn, starting with flutes, then grooves and chevrons, and finally discontinuous tool marks. Lastly, a new process-based model of flutes and tool marks is proposed, and the implications of this model for the Bouma sequence, hybrid event bed models and for a number of other long-held paradigms within the field are examined.

THE FLUID DYNAMICS OF MUD-POOR TO MUD-RICH FLOWS
Recent years have witnessed a step-change in our understanding of flows that are transitional in their behaviour between turbulent and laminar states, due to the addition of mud in suspension (Wang & Plate, 1996;Baas & Best, 2002, 2009Baas et al., , 2011Baas et al., , 2016a. As an increasing quantity of clay is added to a flow, the particles begin to form flocs and longer chains because of electrostatic bonding, and eventually gel, which may significantly influence the rheology of the flow. The nature of such modification can be viewed as a competition between the factors that favour particle aggregation and gelling, notably clay concentration, clay type and water chemistry, and the forces of shear (as both mean shear and turbulence) that can break the bonds between clays. Thus, formation of such transitional flows is highly variable in both time and space as flows accelerate, decelerate, entrain and deposit sediment, and encounter changing water chemistries.
Despite this complexity, experimental studies (Baas & Best, 2002;Baas et al., , 2016a have shown that, as clays are added to a flow through direct substrate entrainment or abrasion of muddy clasts, a series of predictable and consistent changes occur that modify the mean velocity profile and turbulence structure of the flow (Fig. 1). In a clearwater flow moving over a flat and smooth surface, the flow develops a canonical turbulent boundary layer, with a logarithmic profile of horizontal velocity, and turbulence generation occurring in the zone of shear adjacent to the bed (Fig. 1A). A small zone of flow near the bed, the viscous sublayer, is dominated by viscosity and is often less than ca 1 mm in thickness for clearwater flows (Raudkivi, 1997;Bridge, 2003), with its thickness reducing at higher shear velocities. As clay is added to a flow, the first stage of turbulence modulation is characterized by an enhancement of turbulence near the bed when compared to the clearwater case, which appears linked to a thickening of the viscous sublayer (Wang & Plate, 1996;. Such sublayer thickening has also been shown in studies of drag reduction in the presence of fine-grained sediment (Gust & Walger, 1976;Gust, 1976;Best & Leeder, 1993;Li & Gust, 2000) as well as in studies of polymer flows where such sublayer growth has been well-documented (Ptasinski et al., 2001(Ptasinski et al., , 2003. A significant feature of this expanding viscous sublayer is that a zone of shear is established on its upper surface along which large-scale vortices, in the form of Kelvin-Helmholtz instabilities, are shed (Baas & Best, 2002;. This thus provides an additional source of turbulence compared to clearwater flows, and this regime has been termed a turbulence-enhanced transitional flow (TETF; Fig. 1; Baas et al., , 2016a. As more mud is added to the flow, near the bed the enhanced viscous sublayer and region of enhanced turbulence continue to grow, but in the outer flow, where fluid shear is less, the clays begin to form chains that eventually establish a region of undeforming flow, or plug flow, at the top of the flow. This lower transitional plug flow (LTPF; Fig. 1C) is characterized by turbulence enhancement near the bed but turbulence attenuation near the flow surface. At still greater mud concentrations, turbulence near the bed is unable to break the increasingly numerous and strong clay chains and hence turbulence near the bed begins to lessen, leading to a significant increase in the thickness of the viscous sublayer   (Fig. 1D). At the same time, the region of undeforming plug flow extends down from the flow surface towards the bed. This regime, where turbulence attenuation occurs both near the bed and within the outer flow, has been termed an upper transitional plug flow (UTPF; Fig. 1D). Lastly, as increasing clay concentration fosters longer chains, or perhaps gelling, the flow eventually adopts a profile where horizontal velocity is invariant throughout the flow depth, except for a thin zone of shear near the bed on which the flow  . The viscous sublayer (vsl) increases in thickness from ca 1 mm in turbulent flows as clay content increases, and shows a marked jump in thickness in the upper transitional plug flow (UTPF) regime; (F) to (J) transitional flows over a transverse bedform as a function of increasing clay concentration (top to bottom) showing the changing fluid dynamic features in the leeside of the bedform. Views are depicted parallel to flow (modified from Baas & Best, 2008). Flow is from left to right. See text for further details.
rides. This quasi-laminar plug flow (QLPF; Fig. 1E) possesses minimal turbulence throughout the flow depth and a thin near-basal shear zone with minor residual turbulence, overlying a thickened viscous sublayer . If the shear strength of the QLPF flow is great enough, it may be able to transport particles within the body of the plug flow with minimal displacement or rotation.
The transitional flow experiments described here obtained maximum volumetric clay concentrations of 16.6% and 19.2% kaolinite (Baas et al., , 2011 and 8.6% bentonite (Baas et al., 2016b), and thus the details of how flow structure develops at higher concentration remain unknown. The presence and importance of transitional flows in subaqueous density currents, and the presence of plug flow regions, has also been demonstrated recently in the laboratory experiments of Hermidas et al. (2018), who additionally noted the presence and importance of the free shear layer at the top of the current. Those experiments were run at slopes of 6.0 to 9.5°, with measurement durations of ca 1 min (40 to 100 s after the start of the experiments), and had maximum volumetric clay concentrations of 7% kaolinite (the clay formed 33% of the total sediment concentration, the rest consisting of sand, with or without silt). Estimates of the basal boundary layer, using viscosity values for the original mixtures measured ex situ by rheometer, predict laminar basal conditions for some flows (a plug flow, PF, in the classification of Hermidas et al., 2018). However, turbulence data demonstrate that there was considerable residual turbulence in the basal boundary layer (Hermidas et al., 2018, fig. 7), and that this turbulence is much higher than in the plug flow itself, thus consistent with the transitional plug flows of , although insufficient turbulence data are provided to ascribe the flows of Hermidas et al. (2018) to a specific transitional flow category of . Consequently, it is unclear from the work of Baas et al. ( , 2011Baas et al. ( , 2016b and Hermidas et al. (2018) whether flows with higher clay concentrations transform from a QLPF to a fully laminar plug flow (herein termed LPF) where there is no residual turbulence at the base of the flow, or whether flows retain a thickened viscous sublayer and an overlying plug, with an intervening shear layer.
This sequence of transitional flow regimes can be expected in a wide range of flows, but the precise conditions at which each flow stage is reached is a function of three principal factors: (i) the applied mean fluid shear that will act to break up the clay chains, such that greater clay concentrations are required to produce a given transitional flow regime at higher shear velocities (Baas & Best, 2002;Baas et al., , 2016a; (ii) the type of clay, or clays, present, such that clays that attain a higher viscosity and yield strength at lower volumetric concentrations (such as bentonite) will require a lower clay concentration to produce a given transitional flow regime at the same shear velocity (Baas et al., 2016b); and (iii) the degree of turbulence generated from other sources, such as grain and form roughness. For instance, flow over a gravel surface will generate additional turbulence that will tend to break up any growing clay chains. As such, flows over a rough bed surface will require a higher clay concentration to generate a given transitional flow regime at the same shear velocity . Form roughness, such as bedforms, or the topography on the top of debrites (e.g. Fonnesu et al., 2015), can also be expected to have the same effect. However, although the precise boundaries and phase space between these transitional flow regimes vary with applied fluid shear, additional sources of turbulence and clay type (and also water chemistry), these various transitional flows will eventually be generated. This has been demonstrated by experiments examining transitional flows moving over a fixed ripple bedform (Fig. 1F to J;Baas & Best, 2008) that show that TETFs produce enhanced turbulence, when compared to a turbulent flow, associated with the shear layer formed around the leeside separation zone. However, as more clay is added to the flow, turbulence becomes dampened both near the bed and in the outer flow, producing LTPF, and then UTPF ( Fig. 1G to I). Baas & Best (2008) distinguished two phases of flow within the UTPF regime for flow over fixed ripples, which they termed turbulence-attenuated transitional flow (TATF) ( Fig. 1H and I). In the first of these phases, turbulence is attenuated within the separation zone, but the length of the separation zone is similar to that under TF and TETF regimes. As clay content increases, a subsequent phase is reached where the length of the separation zone shortens, alongside a further attenuation of turbulence (Fig. 1I). Eventually, with the addition of more clay, a QLPF forms where flow in the leeside is stagnant with little or no turbulence in the bedform lee (Fig. 1J). The corollary of these changing transitional flow regimes for decelerating flows of mud and sand was investigated by Baas et al. (2011), who demonstrated that ripples increased in height and wavelength under both TETF and LTPF (Fig. 2). These flows possessed enhanced turbulence near the bed that was reasoned to augment turbulence generated from the leeside flow separation zone. The enhanced near-bed turbulence increased erosion at flow reattachment and provided a greater sediment flux downstream, thereby increasing ripple height and wavelength (Baas et al., 2011;Fig. 2). However, in these experiments, at clay concentrations in either the upper part of the LTPF regime or the lower part of the UTPF regime, turbulence at the bed decreasedin part as a result of the rapidly thickening viscous sublayerand led to a decrease in ripple height and wavelength (Fig. 2). As clay concentrations increased further in the LTPF and UTPF regimes, turbulence and bed sediment flux declined, leading to smaller bedforms and eventually a flat sediment bed.
Although the experiments of Baas et al. (2011) concerned aggradational bedforms, in the context of sole structure development they provide important insights into the patterns of turbulence and bed shear stress that may be expected over erosive bedforms generated in a mud bed. For instance, if a negative defect is formed in a mud bed, it can be speculated that bed erosion generated by flow separation over this defect is first enhanced within a TETF and lower LTPF, before decreasing and eventually ceasing under upper LTPF and UTPF regimes. The significance of these speculations is revisited later. Talling et al. (2012a) and Talling (2013) examined the properties of subaqueous debris flows, concentrating solely on the plug flow component, and presented an analysis of the potential influence of yield strength (Middleton & Hampton et al., 1973;Lowe, 1979) as a function of clay concentration (Fig. 3). On the basis of yield strength, flows were then subdivided into low (0.1 to 10 Pa, corresponding to 10 to 20% kaolin by volume), intermediate (10 to 100 Pa, 20 to 30% kaolin) and high-strength (100 to 1000 Pa, 30 to 40% kaolin) debris flows (Talling et al., 2012a;Talling, 2013). This analysis highlights the likely maximum clast size that can be transported by a flow (for the case of a kaolinite-dominated debris flow), illustrating how this size decreases with decreasing suspended clay concentration (and hence yield strength) and increasing clast density (Fig. 3). These relationships are critical in both determining the shear stress exerted on a cohesive bed by an overriding flow, and in determining how clasts can be transported within the body of the flow to act as tools that generate erosive structures in the underlying substrate. Because such models of subaqueous debris flows concentrate solely on the plug flow component, they are not directly comparable with previous transitional flow experiments (Baas et al., , 2011(Baas et al., , 2016b. However, given the typical clay concentrations, the intermediate-strength and high-strength debris flows, which are of interest in the subsequent discussion, likely compare to the quasi-laminar plug flows (QLPF) and potentially the fully laminar plug flows (LPF) described previously. This comparison is supported by the subaqueous sediment gravity flow experiments of Baker et al. (2017), who demonstrated a change in flow type at similar kaolinite concentrations to those of Talling et al. (2012a). The experiments of Baker et al. (2017) produced transitional flows with a dense, cohesive, lower layer (probably a QLPF / intermediate-strength debris flow; the experiments lacked turbulence data to confirm the former), at kaolinite concentrations of 22 to 25% by volume. Fully cohesive flows (likely a LPF / high-strength debris flow) were produced at volumetric concentrations of 27%. The experiments of Baker et al. (2017) also classify the low strength 'debris flows' of Talling et al. (2012a) and Talling (2013) as high-concentration turbidity currents; in shallow water settings, flows with such yield strengths are often referred to as fluid muds (Winterwerp & van Kesteren, 2004).

Substrate controls on erosion
In addition to understanding flow dynamics, it is important to consider the role of the seafloor substrate in the formation of flutes and tool marks. The nature of the seafloor substrate not only governs the threshold at which an overriding flow will erode the bed, but is also an important control on the location, extent, depth and morphology of the erosional features that form once this threshold is exceeded. Grain size is the primary control on the erosion threshold for non-cohesive sediment (i.e. coarse silt, sand and gravel), and hence the prediction of erosion in granular media is generally straightforward (Soulsby & Fig. 3. Estimation of the maximum clast size that can be supported by the yield strength (matrix strength) of a mud-rich fluid, and by buoyancy, for increasing kaolin concentrations. Modified from Talling et al. (2012a) and Talling (2013). Whitehouse, 1997). Some notable exceptions exist, such as where biological factors inhibit sediment transport (e.g. Parsons et al., 2016) or where the substrate is composed of calcareous or biogenic grains. In the latter case, density corrections are required to the classic Shields approach (e.g. Miller & Komar, 1977;Oehmig, 1993). However, most of the world's oceans are floored with cohesive muddy substrates (Dutkiewicz et al., 2015), with the cohesive component composed of clay and non-sortable silt (≤10 lm;McCave et al., 1995), and this is the substrate in which scour and tool marks are most commonly found (Allen, 1984). So what effect does a cohesive substrate have on erosion at the base of a flow?
The specific case of deep-water cohesive sediments As cohesive sediment accumulates at the seafloor, it starts to consolidate under self-weight burial, which leads to a linear increase in bulk density and undrained shear strength with depth, known as normal consolidation (Skempton, 1954). The effect of this consolidation may serve to depth-limit erosion (Parchure & Mehta, 1985;Winterwerp & van Kesteren, 2004), although studies of modern deep-water sediments have revealed several deviations from a simple normally-consolidated profile that are detailed below.

Fluid-like benthic boundary layer
The first exception to the trend of strength linearly increasing with depth is found at the seawater-seafloor interface, which is typically composed of unconsolidated aggregates (Boudreau & Jorgensen, 2001) and in some cases may be treated more as a fluid than a sediment (Winterwerp & van Kesteren, 2004), because of high water content (>>50% of the mass is water, thus water content is >>100% relative to the dry mass), very low undrained shear strengths (<<1 kPa) and intense bioturbation (Baudet & Ho, 2004;Colliat et al., 2011;Hill et al., 2011;Kuo & Bolton, 2013). Whereas this interfacial 'benthic boundary' layer is often lost or disturbed by piston coring or in situ geotechnical testing, shallow box coring of modern deepwater seafloor sediments commonly reveals a thin, centimetres-thick, layer of highly mixed fluid-like mud overlying a more competent mud that has begun to consolidate (Figs 4C and 5). This seafloor layer of low shear strength can be easily eroded under even relatively low bed shear stresses (for example, Fig. 6A points 'a' and 'b', which transition very rapidly into mass erosion).

Shallow strengthening
The second exception is based on in situ shear strength measurements, which indicate that deep-water sediment is often much stronger within a zone a few tens of centimetres to approximately a metre below the seafloor than would be expected from normal consolidation alone, and sometimes by an order of magnitude (Fig. 4). None of the sites shown in Fig. 4 have undergone any loading other than that experienced by progressive accumulation of sediment, nor is there any variation in lithology. Hence these cohesive sediments are apparently overconsolidated, or more correctly phrased, they have a high yield stress ratio (vertical yield stress/effective overburden pressure) (Burland, 1990). This enhanced strength is then lost at depth (more than tens of centimetres to ca 1 to 2 m below seafloor; Fig. 4), where a normally consolidated trend is restored.
The exact reason for the zone of greater yield stress ratio is unclear, but it has been linked to Fig. 4. Plots of undrained shear strength against depth for cohesive sediments from a range of modern deep-water locations -(A) to (D) -worldwide (WD = water depth). Grey filled polygon indicates expected shear strength for each site assuming normal consolidation during burial (defined as virgin consolidation by Skempton, 1954). All sites -(A) to (D) -feature apparently over-consolidated sediments in the top 1 m, despite the lack of significant post-depositional loading. biological influences related to food and chemical dependencies that may explain the depthlimitation. Possible explanations include: (i) sediment strength mediation by sulphate-reducing bacteria that are abundant in the top 1 to 2 m of most of the world's ocean substrates (Parkes et al., 2000); (ii) stabilizing effects of bioturbation by organisms such as polychaete worms (Colliat et al., 2011;Kuo & Bolton, 2013), and (iii) particle-bonding effects by EPS secreted by organisms such as diatoms (Ehlers et al., 2005). Regardless of the cause, this enhanced strength will provide a much higher resistance to erosion than normally consolidated sediment and may strongly depth-limit and otherwise control the nature of erosion (Fig. 6). This shallow strengthening of muds may explain why some powerful sediment density flows only erode localized scours or grooves, but do not cause widespread erosion (e.g. Amy & Talling, 2006;Talling et al., 2013a). Where some sediment density flows do succeed in 'breaking through' this strengthened layer, they may erode large volumes of cohesive sediment, potentially manifested as abundant intraclasts of substrate, transform to transitional flows or debris flows, and deposit hybrid event beds (e.g. Haughton et al., 2003Haughton et al., , 2009Talling et al., 2004), whereas other slightly less powerful flows entrain little sediment and remain as lower-density turbidity currents.
Exposure of previously-buried sediment at the seafloor Truly over-consolidated sediment can also be found at, or close to, the seafloor where erosion or uplift have exposed previously buried sediment (Burland, 1990). Experiments using clearwater flows have found that over-consolidated (remoulded shear strength, c u , >200 kPa), cohesive (and also lithified or weakly cemented) sediment effectively inhibits the types of erosion observed in lower-strength clays (Annandale, 1995;Fig. 6). Erodibility in such materials may instead be controlled by localized weaknesses and imperfections within the sediment mass, such as discontinuities, joints and bedding surfaces (Annandale, 1995), or where the sediment bed is homogeneous, erosion may be controlled largely by sediment abrasion (Yin et al., 2016). Biological modification of the substrate The final exception relates to the influence of benthic and microbial organisms, which are abundant in cohesive sediment within approximately the top metre below the seafloor worldwide (Parkes et al., 2000;Murray et al., 2002). The interactions of these organisms with the seafloor substrate and shallow subsurface sediment can significantly modify their geomechanical properties ( Table 2). The magnitude, type and depth-extent of such modifications are strongly controlled by physico-chemical factors that often strengthen the substrate (Murray et al., 2002), and thus aid the formation of flute and tool marks. For instance, internal burrow pressures of up to 40 kPa have been reported for some benthic organisms that exceed the typical shear strength of deep-water seafloor sediment (Murray et al., 2002;Fig. 4). The effects of such bioturbationinduced pressure lead to compaction and strengthening in cohesive sediment. Other exceptions undoubtedly exist, but the intention here is to highlight that the mechanical behaviour of modern deep-water cohesive sediments can be spatially and temporally complicated, and can exert a strong control on when, where and how erosion occurs.

Seafloor substrates over geological time
A key question is whether the modern seafloor is a good analogue for sediment over geological time, and thus whether these variations with depth are typical. The level and type of bioturbation in deep-sea substrates experienced a major change during the Great Ordovician Biodiversification event (Orr, 2001;M angano et al., 2016;Buatois & M angano, 2018). Since the Ordovician, the diversity of deep-sea trace fossils has fluctuatedoften related to large-scale changes in ocean circulation and oxygenation, such as basinscale anoxic eventsand some ichnotaxa, such as Zoophycos and Ophiomorpha, have changed their environmental range (e.g. Cummings & Hodgson, 2011a;Uchman & Wetzel et al., 2011). However, such changes appear unlikely to have dramatically altered the influence of these fauna on sediment properties. The successful application of ichnofacies and ichnofabric models that integrate modern and ancient traces to diagnose deep-sea environments (Heard & Pickering, 2008; Meadows & Tait (1989);  Enhanced compaction Internal burrow pressures result in localized/differential compaction Hammond (1970); Elder & Hunter (1980); Trevor (1978); Murray et al. (2002) Enhanced adhesion or interparticle-bonding Formation of biologically-induced flocs, biofilms, or inter-particle bonding by EPS Fleming & Richards (1982); Denny (1989); Bromley (1996); Meadows et al. (1990); Reynolds & Gorsline (1992) Armouring of sediment surface Winnowing brings finer sediment to the surface, which is removed by currents, leaving an armouring of coarser sediments Singer & Anderson (1984) Loss of anisotropy/ heterogeneity Bioturbation mixes sediment vertically and laterally Winston & Anderson (1971); Gingras et al. (2008) Lateral variations in substrate strength Spatially variable density of benthic colonization results in localized differences in magnitude of modification Murray et al. (2002) Enhanced bed roughness Seafloor expression of burrows provides (biogenic) roughness at the sediment-flow interface Meadows & Meadows (1991); Bromley (1996); Davies (1982); Poulos (2001) Reworking of cohesive sediments into faecal pellets Cohesive sediment excreted as bonded pellets by invertebrates such as polychaetes that line burrows (e.g.

Ophiomorpha)
Moore ( Cummings & Hodgson, 2011b;Callow et al., 2014;Heard et al., 2014;Knaust et al., 2014;Buatois & M angano, 2018; Rodr ıguez-Tovar & Hern andez-Molina, 2018) further suggests that bioturbation has not changed fundamentally, otherwise this approach would not work. Interestingly, work in shallow marine successions has argued that the mixed layer due to bioturbation, presently approximately the upper 10 cm, may have increased in thickness slowly through the Cambrian, only reaching modern conditions in the late Silurian (Tarhan et al., 2015;Tarhan, 2018a). In marked contrast, a progressive decrease in near-bed substrate strength over the whole Phanerozoic has recently been inferred based on a decline in the number of studies reporting flutes and tool marks as a function of geological time period (Tarhan, 2018b). Tarhan (2018b), however, does not consider potential observational bias, and the process arguments supporting a link to substrates do not consider many of the processes discussed above.
Considering the evolution of trace fossils, and allowing for the postulated changes in mixed layer depth in shallow marine conditions, herein it is concluded that modern deep-sea seafloors are likely a good analogue for deep-sea substrates since at least the late Silurian, albeit there is a need for further research.

FLUTES
Flute casts or flutes (after Crowell, 1955) are erosive features that widen downstream from a point, the 'nose', abruptly deepening downstream before gradually decreasing in depth towards their downstream end (Figs 7 and 8). They are formed by the erosion of a cohesive substrate that is subsequently infilled by sand, or in some cases gravel (Winn & Dott, 1977;Jobe et al., 2010), and thus are observed as sole marks on the base of beds. Flutes generally range in length from several centimetres to ca 0.50 m (Allen, 1971a), with widths of 0.01 to 0.20 m and depths of a few centimetres to 0.1 m (Collinson et al., 2006). However, flutes that are metres long, up to 1.0 m wide, and 1.5 m deep are known (Winn & Dott, 1977Jobe et al., 2010). Even larger 'flutes', metres to hundreds of metres long, have been observed on the upper surface of beds where they are typically referred to as megaflutes (e.g. Elliott, 2000;Macdonald et al., 2011a,b;Hofstra et al., 2015), but the analysis herein is restricted to sole marks. Allen (1971a) introduced a summary figure for flute morphology (Fig. 7A), later referred to as the 'ideal flute' (Allen, 1984), which in addition to the basic features described above, also exhibits lateral furrows and a median ridge (Fig. 8B). However, flutes can also consist of simple smooth forms that lack a median ridge and lateral furrows (Fig. 8B). Flutes exhibit a great variation in planform shape, from parabolictransitional examples (Figs 7B and 8A) to long, thin spindles (Figs 7B and 8C) and asymmetrical, and comet-shaped, forms (see Fig. 7B). Parabolic-transitional flutes are relatively rare (Fig. 8A), with parabolic flutes far more common (Figs 7B and 8B), the latter form representing the 'ideal flutes'. Parabolic forms range in size from a few centimetres to >0.50 m long, and have length to width ratios between 1 and 4 (Allen, 1971a). In contrast, spindle-shaped flutes are fairly common, 0.05 to 0.15 m long, typically lack median ridges and lateral structures, and are much longer than they are wide (Allen, 1971a(Allen, , 1984. Comet-shaped flute marks are rare and tend to be smaller still, typically a few centimetres in length and rarely more than 0.1 m long, and they have sinuous edges in the downstream direction (Allen, 1971a(Allen, , 1984. Polygonal flutes in mud beds were described by Allen (1971a;reproduced in Collinson et al., 2006 andMountney, 2019); however, the present re-analysis of the examples listed therein fails to identify clear examples, perhaps reflected in the absence of this form in later summaries (Allen, 1984). The present paper thus concludes that polygonal forms do not occur in muds, although such forms are well-known from cave scallops where dissolution processes dominate (Allen, 1971a;Richardson & Carling, 2005). Flutes can be found covering entire bedding planes (conjugate), in clusters or as individual marks (isolate), and typically many different types of flute can exist on the same bed (Allen, 1971a(Allen, , 1984Pett & Walker, 1971).

Longitudinal distribution of flutes
A number of observations have been made concerning the variation in flute occurrence and morphology with downstream distance, and as a function of variables such as bed thickness and Bouma division, that in turn vary downstream. Flutes at the base of T A beds were observed to be wide and 'bulbous' [defined as very broad at the upstream end, sensu D _ zuły nski & Walton (1965), and used in this sense throughout the present paper; note that Allen (1984) defined 'bulbous' flutes differently, as having narrow deep heads], whilst those below T B and T C beds were narrow with pointed noses (Pett & Walker, 1971). Larger flutes have been observed on the base of thicker beds (Sestini & Curcio, 1965;Middleton, 1970;Tanaka, 1970;Allen, 1984), with Pett & Walker (1971) showing a clear relationship between flute width and bed thickness, but not between flute depth and bed thickness, suggesting that there may be a substrate control on flute depth. Such measurements assume that no later loading of flutes has occurred (e.g. Kelling & Walton, 1957). These field observations suggest that flutes become narrower, smaller, and have more pointed noses, with downstream distance. Furthermore, these relationships have been used to imply that larger flutes associated with thicker, more proximal sands, were formed by faster, thicker and longer-lived currents, and correspondingly, that smaller more distal flutes on thinner beds were formed by slower, thinner, shorter-lived currents (Allen, 1984). In addition to these field observations, Allen (1971a) modelled the distribution of flute marks with downstream, distance, based on defect theory, and predicted that flutes would become smaller downstream, and that they would change from whole bed surfaces covered in flutes (conjugate) to isolated flutes.

Experiments
The earliest experiments that were conceived to understand the development of flutes are those of Fuchs, [1895;see Wetzel (2006) for an English translation of some key parts] who used sand  Allen (1971aAllen ( , 1984. and plaster-of-Paris to succeed in reproducing a range of 'bulges' on the bases of beds. Mutti et al. (2009) argued that Fuchs (1895) successfully reproduced small flutes experimentally, although none of these 'bulges' show much similarity with the planform and cross-sectional form of flutes (D _ zuły nski & Walton, 1963). Later work by R€ ucklin (1938) is more widely credited as forming the first flutes (D _ zuły nski & Walton, 1963; Allen, 1984), albeit that the similarities are limited (Allen, 1971a) and these experiments were not in mud beds; the bed composition was 5.8% clay <10 lm and 94.2% coarse grains consisting of quartz (20 to 80 lm) and mica (50 to 200 lm). Subsequent work examined flute formation in weak mud beds with flows composed of plaster-of-Paris (D _ zuły nski & Walton, 1963; D _ zuły nski, 1965), but it is not clear if these features resulted directly from erosion, or from deformation and loading (Allen, 1971a). In contrast, the work of Allen (1968Allen ( , 1969Allen ( , 1971a on flows over weak and higher-strength mud beds did demonstrate the formation of flutes, with conjugate forms in weak beds, and individual flutes in higher-strength modelling clay. The key breakthroughs in understanding flute formation have been derived principally from the seminal work of Allen (1971aAllen ( , 1973Allen ( , 1975 who employed clearwater flows to dissolve beds composed of plaster-of-Paris. Whilst the processes are different, i.e. dissolution versus abrasiondriven erosion in muds, the experiments produced analogous forms, and enabled the formative processes to be studied in detail. However, the focus on studies using plaster-of-Paris has meant that the understanding of substrate controls on flute initiation and development remains limited. The nature of formative flows for flutes Allen (1968Allen ( , 1971a demonstrated that flutes are associated with turbulent flows that produce flow separation. The nature of flow separation changes as a function of flute morphology, from prominent horizontal rollers (with rotation in the downstream direction) within parabolic flutes (Figs 7 and 9), to a pair of rotating vortices (with rotation transverse to the main flow) within narrow flutes ( Fig. 9) (Allen, 1971a). The initiation of flutes has been associated with the presence of defects in the bed, which can be produced by hollows formed by bioturbation, steps in bed height, or inhomogeneities within the substrate (Allen, 1984). Alternatively, flows over some very weak mud substrates can form their own defects that can then develop into flutes (Allen, 1969). Knowledge of flute development on stronger planar beds is largely lacking, since Allen's (1971a) experiments with modelling clay all used initial defects. Yin et al. (2016) also used modelling clay but with planar beds, and found that, whereas defects developed and grew with time, the resulting erosional structures were more analogous to bedrock erosion features in rivers (Richardson & Carling, 2005), rather than flutes in muds. However, the results of Yin et al. (2016) suggest that defects in planar mud beds can form purely from abrasion by a sediment-laden flow. These previous experiments on clay beds, although limited in number and in terms of substrate measurements (with the exception of Yin et al., 2016), also suggest that there is likely a strong substrate control on flute formation. Weak beds (for example, Fig. 5) may be unable to maintain the relatively steep slopes associated with flutes, notably at the nose, whilst beds that are too strong enable undercutting of the margins and the production of structures more analogous of bedrock rivers (Yin et al., 2016).
For a given substrate, the evolution of bed defects is dependent on initial defect size and the properties of the flow, with Allen (1971a) referring to these as 'unstable' forms leading to the development of parabolic flutes (Type I) and 'stable' forms producing spindle-shaped flutes (Type II) (Fig. 9). Allen (1971a) related this developmental divergence to the nature of separated flow within flutes, and in particular to the transition of the viscous sublayer upstream of the flute, to a turbulent flow, via the sublayer rolling up into a series of vortices. Based on theory derived from experiments on unstable laminar shear layers (Sato, 1956(Sato, , 1959 and experimental data on flutes (Allen, 1971a), a criterion for the critical defect length, X crit , was introduced, based on the downstream distance for transition to a turbulent flow (Allen, 1971a(Allen, , 1984: where d is flow depth, U is mean downstream velocity and m is kinematic viscosity. Consequently, when the downstream length of the initial bed defect, X, is greater than the critical defect length, X crit , turbulent flow can directly act on the bed defect, and flutes will exhibit 'unstable' behaviour (Type I, Fig. 9), whereas if X < X crit then flutes are 'stable' (Type II, Fig. 9). Allen (1971a) argued that, as flows travel downstream, flow velocity declines because of progressive sedimentation, and also assumed that the flow depth would either decrease downstream (Allen, 1984) or that flow depth was unlikely to increase downstream (Allen, 1971a). However, whereas this may not necessarily be true, the product of velocity and depth will decline, unless the flow is undergoing autosuspension (Pantin, 1979;Parker et al., 1986). Analysis of Eq. 1, using the criterion that the product of velocity and depth decreases downstream, therefore predicts that the critical defect length, X crit , increases with distance downstream. Assuming that there is no downstream variation in initial defect size, or substrate properties, this in turn predicts that parabolic forms are more prevalent upstream, and that spindle-shaped flutes are more likely downstream, as seen in field observations where larger bulbous flutes are observed below T A beds, and smaller, narrower and more pointed flutes are observed below Bouma T B and T C beds (Pett & Walker, 1971).

Downstream distribution of flutes: a paradox
The prediction that, as flows decelerate, flutes decrease in size whilst changing morphologically until flows are no longer able to erode the substrate, suggests that there should be a lack of erosive bedforms downstream of flutes. Paradoxically, however, erosive bedforms in the form of tool marks are generally plentiful in more distal locations (e.g. Walker, 1967;Lovell, 1969;Slackza & Unrug, 1976), whereas flutes are typically preferentially associated with more proximal locations (see 'Distribution and association Fig. 9. Schematic evolution of flutes from an initial bed defect, showing stable and unstable developmental paths: V is the time-averaged areal mean erosion rate, a parameter that changes over time, t, and was measured in the experiments from repeated profiles; X is the initial length of the defect. In addition to the bed streamlines shown in planform for the final stable and unstable forms, the flow fields are shown longitudinally and in cross-section. These patterns were derived from experiments using clear water flows over plaster-of-Paris beds. Based on Allen (1971a). of scour marks and tool marks'). Given that flows have decelerated to a point where they cannot erode the bed, then it is paradoxical that they are able to support grains large enough (order millimetres to centimetres diameter) to form the range of tool marks typically observed in these distal environments (see Tool marks section). The pioneering work of Allen (1971aAllen ( , 1973Allen ( , 1975Allen ( , 1984 involved consideration of clearwater flows, and analogies with cohesionless turbidity currents. The present work assesses what the effects of cohesive transitional flows would be on flute dynamics and morphology, and whether these flows offer an answer to the apparently paradoxical distribution of flutes and tool marks.

Transitional flows and flute dynamics
As clay is added to an initial cohesionless turbulent flow, the flow is modified forming a turbulence-enhanced transitional flow (TETF; see 'The fluid dynamics of mud-poor to mud-rich flows' above), which occurred at kaolin clay concentrations as low as 0.046% in the experiments of Baas & Best (2008). As demonstrated in work on ripples, TETF is associated with enhancement of turbulence within the separation zone as a result of growth of the internal shear layer ( Fig. 1; Baas & Best, 2008). The influence of this enhanced turbulence is most notable at the flow re-attachment point. If such TETFs erode into a cohesive bed, this may lead to enhanced erosion and increased maximum depth of flutes, as compared to turbulent flows, potentially producing wider and more bulbous flutes. With increasing clay content, turbulence-attenuated transitional flows (TATF) develop ( Fig. 1; Baas & Best, 2008), marked by an initial decline in turbulence in the shear layer of the separation zone generated behind the leading edge of the flute, relative to turbulent flows. However, Baas & Best (2008) noted no corresponding decline in the length of the separation zone. Such turbulence attenuation likely leads to shallower flutes. Further increases in clay content within the TATF region ( Fig. 1) will lead to additional declines in turbulence in the shear layer, and a progressive decrease in the length of the flow separation zone, likely leading to smaller, thinner flutes. At some point, the cohesive strength of the flow will destroy the flow separation zone entirely (Baas & Best, 2008), with the likely demise of further flute development at this point. Transitional flows over mobile beds showed a rapid decline in bedform height and wavelength, at some point between the upper part of the lower transitional plug flow regime and the lower part of the upper transitional plug flow regime (Fig. 2). These morphological changes occur in response to decreasing turbulence in the flow separation zone, and potentially also because of the rapid increase in the thickness of the viscous sublayer (Fig. 1). By analogy, flutes may also cease to actively form around the transition between LTPF and UTPF conditions. This reanalysis of flutes demonstrates that flutes will continue to form, and indeed may be enhanced, under transitional flows. Thus the key conclusion of Allen (1968Allen ( , 1971a) that flutes are the product of turbulent flows does not strictly hold, because they can also be formed under transitional flows, provided bed shear stress exceeds the critical erosion threshold.
If the criterion of Allen (1971a) for separating the development of initial bed defects (Eq. 1) is re-examined for transitional flows, herein it is noted that transitional flows should have an effect through increases in the dynamic viscosity, which can increase by an order of magnitude, or more, in transitional flows . Changes in clay content may therefore produce the same downstream effects as changes in velocity. As flows become more transitional, and dynamic viscosity increases, the critical defect size, X crit , should increase and thus flutes are more likely to become 'stable' and change from parabolic flutes to spindle-shaped flutes.
Examination of flow separation dynamics, and of the stability criterion for flute types, demonstrates that transitional flows likely influence flute evolution and morphology. Increases in turbulence in TETFs and LTPFs are postulated to lead to the development of wide, bulbous flutes. Further increases in suspended sediment concentration in the upper part of LTPFs or the lower part of UTPFs likely lead to progressive turbulence dampening and thus decreased flute sizes, more stable spindle-shaped flutes, and ultimately a loss of flute production or growth entirely.
In terms of understanding the downstream distribution of sole marks, the key question is how flows transform with downstream distance. Many turbidity currents have been postulated to gradually transform downstream from non-cohesive to cohesive through the erosion and ingestion of mud from the seafloor, and through the increasing importance of clay cohesion relative to turbulence generation (Haughton et al., 2003Talling et al., 2004). In such cases, the increasing cohesion of the flow would work in tandem with the decreasing product of flow velocity and depth to encourage a transition to smaller spindle-shaped flutes, and ultimately to a lack of flute development. This offers a potential solution to the paradox of how erosive tool marks can be found downstream of flutes, but only if these tool marks are associated with more cohesive currents such as transitional flows or debris flows. The origin and development of tool marks are thus examined next.

TOOL MARKS
As noted earlier, tool marks can be subdivided into continuous and discontinuous forms; the former consist of grooves and chevron marks, and the latter of prod, bounce, skip and roll marks. These structures are considered in turn, starting with the continuous forms. However, first the source of the tools is reviewed.

The nature of tools
The nature of the tools forming tool marks has been unclear in most past studies, because the tool is found rarely at the end of marks such as grooves. Examples include mudstone clasts, pieces of wood, pebbles, bones and shells (D _ zuły nski & Radomski, 1955; D _ zuły nski & Sla z czka, 1958; Glaessner, 1958;McBride, 1962;D _ zuły nski & Walton, 1965;Enos, 1969a;D _ zuły nski, 1996). A number of authors have concluded that mudstone clasts are the most likely tools (D _ zuły nski & Radomski, 1955;Wood & Smith, 1958;D _ zuły nski & Walton, 1965;Middleton & Hampton, 1973, although Kuenen (1957) argued that mudstone clasts were improbable as tools since they would undergo rapid rounding through abrasion and therefore would not produce grooves with internal striations. Kuenen (1957) instead suggested that stones or shells pulled by 'a sail of seaweed' would enable clasts to be dragged along rather than rolling along (see also D _ zuły nski & Sla z czka, 1958). The potential for abrasion of mudstone clasts is examined later, after the nature of individual tool marks has been described, and potential formative mechanisms considered.
An alternative approach to identifying the nature of the formative tools is to consider the availability of tools in deep-water clastic environments. Extrabasinal pebbles are typically restricted to high-gradient, tectonically active systems (Hsu, 1959;Winn & Dott, 1977;Leszczy nski, 1989;Jobe et al., 2010). Plant fragments, many of which are too small or fragile to act as tools, are thought to be preferentially associated with hyperpycnal currents (Zavala et al., 2012;Deville et al., 2015;Zavala & Arcuri, 2016) or the collapse of shelf-edge deltas (Hodgson, 2009), and may be concentrated towards the lower energy parts of the flow, i.e. the top and back of the flow (Haughton et al., 2003;Kneller & McCaffrey, 2003;Hodgson, 2009). Consequently, plant fragments are less likely to be in direct contact with the bed. Furthermore, tool marks are widely reported from Palaeozoic deep-water strata prior to the advent of plants that were greater than a few centimetres in size, and before the development of significant internal structure, in the Devonian (Kenrick et al., 2012), and therefore before plant fragments as likely tool makers (e.g. Craig & Walton, 1962;Enos, 1969a;Parkash & Middleton, 1970;Clayton, 1994;Haines et al., 2001).
Additional sources of tools, such as bones and shells, appear to be unusual and relatively rare. In contrast, mudstone clasts are ubiquitous in deep-water clastic systems across a wide range of environments, including broad sediment bypass zones such as channel-lobe transitions, channel lag deposits, and in the deposits of debris flows and hybrid events (Mutti & Nilsen, 1981;Johansson & Stow et al., 1995;Haughton et al., 2003Haughton et al., , 2009Posamentier & Kolla, 2003;Stevenson et al., 2015;Brooks et al., 2018a). These mudstone clast-rich deposits extend from proximal areas on the slope, through submarine channels, all the way to the fringes of basin-floor lobes (Posamentier & Kolla, 2003;Talling et al., 2004;Luthi et al., 2006;Hodgson, 2009;Talling, 2013;Stevenson et al., 2015). Many of these mudstone clasts are intra-basinal as a result of the erosion of seafloor muds, but mudstone clasts incorporated into debris flows can be very far-travelled (Talling, 2013;Stevenson et al., 2015). As highlighted earlier, the generation of such intraclasts may be favoured by the presence and erosive break-up of a near-surface over-consolidated layer.

Groove casts
Groove casts, also referred to as groove marks or grooves, were first named by Shrock (1948), and subdivided into drag marks and slide marks by Kuenen & Sanders (1956) and Kuenen (1957), referring to those features observed below greywackes (i.e. muddy sands) and those formed from slumping, respectively. However, later research found that it is difficult to differentiate these two types (D _ zuły nski & Sla z czka, 1958; Bouma, 1962). The term 'drag mark' has since been used more generally to refer to grooves in deep-water systems and other environments, as well as glacial striae, features formed by drifting grounded ice and boulders on playa floors (Allen, 1984).
Large numbers of grooves can cover entire surfaces (Figs 10 and 11), where they may show a range of sizes, or grooves can be present as isolated examples (See section on Discontinuous tool marks). Where present in groups, they are typically parallel or sub-parallel to one another (Kuenen, 1957;Allen, 1984;Collinson et al., 2006). However, grooves may also show crosscutting relationships with angles of up to 90°, although typically <40°( Fig. 10B; Walton, 1965;Enos, 1969a;Ricci Lucchi, 1969a). Groove casts have been seen to commence at an 'irregular bulge' (D _ zuły nski & Sla z czka, 1958) representing the counterpart of the original irregular depression, or from chevron marks ( Fig. 13A and B), whilst terminations can consist of either: (i) a tapering of the groove to meet the original substrate surface; (ii) a rounded end, sometimes with an associated small mud ridge in the downstream direction; or (iii) an abrupt, twisted end (D _ zuły nski & Sanders, 1962a;D _ zuły nski & Walton, 1965). Terminations are, however, very rarely seen, with Enos (1969a) reporting just 10 terminations across >1500 beds. Even when terminations are present, most lack their formative tools. Key unaddressed questions concern how the tools are transported away from the ends of their grooves, and ultimately where these tools are deposited.
Mapping of grooves beneath individual event beds suggests that grooves may cover lengths and areas far in excess of those identified from individual outcrops, as shown in Fig. 14. In these examples from the basin plain deposits of the Miocene Marnoso-arenacea Formation in the Italian Apennines (see Table 3 for context) grooves are present for distances in excess of 40 km and over areas up to ca 300 km 2 . It is unknown whether individual grooves are continuous for these distances or whether these consist of a succession of individual isolated grooves. The highly parallel nature of groove casts, their large longitudinal and areal extent, and the frequent occurrence of internal parallel striae (Fig. 10E), all suggest that the tools were dragged in a single position (without rotating), at a constant height (no bouncing), through a substrate that was sufficiently strong that it could not be deformed by fluid stresses or flow back into the eroded space once the groove had been cut. The occasional spiralling of internal laminae and twisted termination suggest that in these rare cases the clasts are able to rotate, albeit relatively slowly with respect to their downstream movement in the case of the internal striae. Crowell (1955) claimed that R€ ucklin (1938) had produced groove marks, but the feature produced has little in common with groove marks (D _ zuły nski & Walton, 1965), and furthermore these experiments were not in mud beds (5.8% clay <10 lm; 94.2% silt and sand). The very first work to produce grooves was thus Kuenen (1957, see plate 1D) who produced grooves ('slide marks' of Kuenen, 1957) from experimental slumps, using a 2 cm thick sandy cover sliding over a clay layer at inclinations from a few degrees to 10 to 20°. Subsequently, Ten Haaf (1959; reported in D _ zuły nski & Sanders, 1962a) studied erosive marks caused by snowballs catapulted over a surface of fresh snow, and concluded that groove marks were linked to flows with great current velocity, interpreted as the product of turbidity currents. Later experiments using plaster-of-Paris for the currents and kaolin clay or gelatine for the substrate were undertaken to examine sole structures formed under 'artificial turbidity currents' (D _ zuły nski & Walton, 1963(D _ zuły nski & Walton, , 1965D _ zuły nski, 1965;  , 1966). A variety of tools (fish bones, hardened mud or plaster-of-Paris fragments; numbers and sizes unknown) were placed at the base of a short ramp and on the clay floor of the tank, and plaster-of-Paris currents were then released down the ramp. The experiments succeeded in making short individual grooves, including one that showed internal laminae that spiralled longitudinally (D _ zuły nski, 1965). However, sub-parallel groups of grooves that characterize outcrop examples were not reproduced. The grooves were also associated with a range of other tool marks including prod, bounce and skip marks (D _ zuły nski & Walton, 1963(D _ zuły nski & Walton, , 1965D _ zuły nski, 1965D _ zuły nski & Simpson, 1966). This is in contrast to outcrop examples where these features are commonly separated in space or time, implying that the optimal conditions for groove formation were not achieved in these experimental studies.

Experiments
A key question concerning the experiments of D _ zuły nski and co-workers is how representative the flows that formed these grooves and associated tool marks are of turbidity currents, even allowing for the scale of the experiments (Peakall et al., 1996). Relatively few details of the experiments were given, but basic flow parameters can be estimated. Densities and viscosities were not measured, but the proportions of water to plaster-of-Paris were 3 : 2 or 2 : 1 (D _ zuły nski & Walton, 1965), 3 : 1 in the case of the experiments on tool marks (D _ zuły nski & Walton, 1963), or 2 : 1 / 3 : 1966). No details of the mixtures used in D _ zuły nski (1965) were provided. Whereas plaster-of-Paris is quite a variable material, assuming a typical bulk density of 785 kg m À3 and an absence of changes in volume as a result of the dissolution of the plaster-of-Paris and initial hydration of the calcium sulphate hemihydrate minerals (note that these volume changes are small; Jørgensen & Posner, 1959), gives flow densities of ca 1520 kg m À3 , ca 1390 kg m À3 and ca 1260 kg m À3 for the 3 : 2, 2 : 1 and 3 : 1 mixtures, respectively. Viscosities at the time of mixing can be estimated at ca 1.0 to 2.5 Pa s À1 (Murakami & Hanada, 1956), about the same as runny honey (Yanniotis et al., 2006), albeit plaster-of-Paris increases in viscosity rapidly after just a few minutes, if there were any delays in the experiments (Murakami & Hanada, 1956). The yield strength of the plaster-of-Paris flows used in these experiments is harder to estimate, although plaster-of-Paris does exhibit yield strength at high concentrations (Rees, 1983). In summary, the experiments of D _ zuły nski & Walton (1963,1965) likely had viscosities equivalent to kaolin suspensions with approximately 20% by volume concentration (cf. Talling, 2013, fig.  9A), had densities representative of kaolin suspensions with volumes of 15 to >30% (cf. Fig. 3) and likely had some yield strength. These flows consequently had densities largely in the intermediate-strength debris flow field (Fig. 3), had viscosities equivalent to the lower boundary of the intermediate-strength debris flow field (Talling, 2013) and had a yield strength likely in the broad range for the low strength debris flows (0.1 to 10.0 Pa) of Talling (2013), or the lower and upper transitional plug flows of Baas et al. ( , 2011. In turn, flow rheology is also dependent on applied stress and thus velocity (Baas et al., , 2011Talling, 2013). Whereas velocities are unknown, and thus the exact rheology cannot be specified, it is clear that the experiments are more representative of transitional plug flows, or intermediate-strength debris flows, than the turbidity currents that these workers compared them to.
A beautiful example of apparently well-defined parallel grooves has been observed in an experimental subaqueous debris flow composed of kaolinite-water slurries with approximately 40% by weight kaolinite, where the coherent head had broken off from the body of the flow because of hydroplaning ( Fig. 15; Hampton, 1970;Middleton & Hampton et al., 1973). Presumably, the grooves were formed by: clasts that were larger than the thickness of the basal water layer beneath the hydroplaning block; at the back of the broken-off debris flow component; or as fluid dissipated underneath the flow as it came to rest. Small clasts of kaolinite are observed behind the flow, and likely formed the tools. Kastens (1984) noted the similarities between the grooves shown in these experiments and those observed on the modern seafloor (Fig. 12). However, it should be noted that there was no initial substrate in these experiments (Hampton, 1970), unlike Kuenen's (1957) experiments with slumps, and therefore the grooves were cutting into a deposit formed by the passing current.

Chevron marks
Chevron marks (Dunbar & Rodgers, 1957) consist of a series of open and continuous V-shaped, or U-shaped, ridges that are aligned in a given direction (Fig. 13). The chevrons have been shown to close in the downstream direction (Craig & Walton, 1962;D _ zuły nski & Sanders, 1962a). Chevrons comprise a continuum of forms from uninterrupted chevrons (V-shaped or U-shaped ridges), when the whole form is present, through cut chevrons consisting of Vshaped forms that are cut down the middle, to interrupted chevrons, with ridges and furrows either side of a clear groove mark (Fig. 16; Craig & Walton, 1962;D _ zuły nski & Sanders, 1962a). Chevron marks have been observed to occasionally transition downstream between these different forms: from uninterrupted, to cut, or interrupted chevrons (Craig & Walton, 1962;D _ zuły nski & Sanders, 1962b), and from interrupted to uninterrupted chevrons (D _ zuły nski & Sanders, 1962a). Allen (1984) suggested that these transitions represent the concave-up trajectory of the clast as it gets closer to (and/or cuts) the bed, and then moves away again. However, as with grooves, transitions are unusual and they are typically constant in form where observed, coming under the 'continuous' class of tool marks. The different forms are consequently associated with different positions of the tool relative to the bed, with the tool cutting into the bed (interrupted chevrons, cut chevrons) or presumably at a constant height above the bed (uninterrupted chevrons). Chevrons are typically a few millimetres wide, with greater widths typically associated with interrupted chevrons (Craig & Walton, 1962). In longitudinal cross-section, the downstream end of the chevron ridge is steepest and is folded over on itself ( Fig. 16; D _ zuły nski & Sanders, 1962a; Allen, 1984). The chevrons appear to form by fluid stressing of weak ductile muds, and thus are partly a function of the bed substrate properties (D _ zuły nski & Walton, 1965). The fluid stressing itself is believed to be caused by wakes that form around the moving tool, and have been likened to the wakes that form behind ships (Craig & Walton, 1962;D _ zuły nski & Sanders, 1962a;Allen, 1984). In some cases, transitions occur from uninterrupted chevrons to regular groove marks (Kuenen, 1957; Fig. 13A and B), indicating that the formative tool was moving downward through the flow, and then into contact with the bed. The loss of chevrons when the tool makes contact with the bed potentially implies an abrupt increase in substrate strength. However, abrupt changes in the cohesive strength of the seafloor may be relatively unusual, albeit that biological controls and/or oxygenation may create these. Alternatively, such transitions may suggest that the fluid dynamics around the tool itself alter the strength of wakes impacting the substrate. Here, the potential for variations in the strength of wakes is explored by further considering ship wakes.
Ships form interrupted chevrons because the vessel itself cuts the bow wave, in the same way as the interrupted chevrons are assumed to form. In contrast, the bow wave of a tool above the surface of a bed may be able to propagate downward forming uninterrupted chevrons. For ships, the magnitude of the transverse waves forming the wake is a strong function of the length-based Froude number, F L = U/√(gL), where U is streamwise velocity, g is acceleration due to gravity and  L is the length of the ship's waterline (e.g. Parnell & Kofoed-Hansen, 2001;Soomere, 2007). A socalled 'hump speed' occurs when F L is ca 0.56, producing increased wave energy. This can be further exacerbated if the vessel is in shallow water, as characterized by the depth-averaged Froude number F h = U/√(gh), where h is the water depth. As F h increases, wave heights increase, and if the critical value of F h coincides with the 'hump speed' very large waves can be generated, which is a problem for some fast ferries (Parnell & Kofoed-Hansen, 2001). It is not clear how far such analogies can be taken with respect to a fully submerged tool with wakes rather than waves, but it does suggest that different regimes may exist that could lead to major changes in the size and strength of the wakes generated around a moving tool. Consequently, the absence of chevrons in most grooves may suggest that the uppermost part of the substrate was too consolidated, or the particle velocity and orientation were suboptimal, for the generation of sufficiently strong wakes capable of deforming the substrate.

The nature of formative flows for grooves and chevrons
This section reviews the possible mechanisms for the formation of grooves and chevrons, in terms of the evidence from groove orientations and cross-cutting relationships, and flow type  (low-density and high-density turbidity currents, granular flows, liquefied/fluidized flows and debris flows; the latter equivalent to quasi-laminar plug flows and laminar plug flows, described earlier) (Fig. 17), in the light of progress in understanding the dynamics of sediment gravity flows.

Groove orientations
As noted earlier, grooves are typically parallel to one another, but they can also show cross-cutting relationships. There has been much debate concerning the interpretation of cross-cutting groove marks, with interpretations as the product of a single flow (e.g. Kuenen & Ten Haaf, 1958;Enos, 1969a;Allen, 1971b) or multiple flows (e.g. Crowell, 1958;Mulder et al., 2002). Multiple flows should exhibit two or more maxima in terms of the distribution of crossing groove directions, and a consistent relationship between the age of the mark and orientation. However, these relationships are not typically observed (Enos, 1969a;Allen, 1984). A key consideration here is that in contrast to flutes, which form over a period of time, grooves are thought to be cut by a tool near-instantaneously, so individual marks can reflect small-scale changes in current direction, rather than time-averaged properties (Allen, 1971b; see later discussion). Explanations for cross-cutting relationships from single flows include: (i) variations in turbulent flow related to the growth and decay of lobes (and clefts) at the head of a turbidity current, which are associated with secondary flows (Kuenen & Ten Haaf, 1958;Allen, 1971b); (ii) flow divergence in an expanding current (Potter & Pettijohn, 1963); (iii) a 'meandering' migration of the flow over time (Walker, 1970); (iv) variations in flow direction between split debrite blocks in a transforming flow (Felix & Peakall, 2006;Draganits et al., 2008); and (v) rotation of blocks in the flow that are much larger than the grooves (Draganits et al., 2008). Rotation of blocks in the flow can explain even the largest angular differences (90°) between grooves (Draganits et al., 2008). Other ideas discussed by Ricci Lucchi (1969a), including Coriolis force, flows in the ambient fluid, irregularities on the bed, and transverse slopes, were all considered untenable by Allen (1971b).
Low-density turbidity currents (Fig. 17A) The formation of grooves was first linked to turbidity currents by Kuenen (1953; see also Kuenen & Sanders, 1956) although the density of these turbidity currents was not inferred. Turbulent, low-density turbidity currents as agents for the formation of grooves were postulated by the catapulting snowball experiments of Ten Haaf (1959), in the experiments of D _ zuły nski and coworkers (D _ zuły nski & Walton, 1963;D _ zuły nski, 1996 and through consideration of suspended sediment within the head (Allen, 1971b). However, the experiments of D _ zuły nski and co-workers have been shown here to be more comparable to transitional plug flows and intermediate-strength debris flows, and in any case only succeeded in generating isolated grooves over short distances. Similarly, snowballs only generate straight grooves in very soft substrates, by momentum alone, and leave the tool at the end of the groove (Ten Haaf, 1959). Key questions are whether low-density turbulent turbidity currents can: (i) transport groups of particles in near-parallel straight lines; (ii) transport particles that are partially 'submerged' and erode into a substrate, at a constant depth, particularly where this substrate has sufficient strength to avoid fluid stressing; (iii) keep a particle in a fixed position with respect to the bed surface (i.e. without rotation), thus maintaining grooves of constant width and form; (iv) hold particles at constant heights above the bed, as required for the formation of chevrons; and (v) preserve the grooves and chevrons in a pristine form. Video analysis of cobbles in bedload-rivers shows that sliding of particles is typically limited to events of less than one grain diameter in length, rolling events consist of short sub-parallel straight segments a few grain diameters in length, and that particles are dispersed laterally within the flow relatively quickly (Drake et al., 1988;Seizilles et al., 2014). Overpassing of gravels across much finer sands is likely more applicable to transport of tools over muds, and might lead to clasts travelling in near-parallel straight lines. However, analysis of gravel overpassing shows that clasts typically roll, and sometimes bounce, rather than move across the bed in a fixed orientation (Allen, 1983), such as via sliding. Particles above the bed in a turbulent flow typically move either as saltation load with characteristic ballistic profiles, or in suspension, where particles move within the flow (Bagnold, 1973;Francis, 1973;Lee & Hsu, 1994). In both cases, particles would not be expected to maintain a constant height above the bed, as is postulated to occur in the formation of chevron marks. It is also unclear why forms associated with turbulent flow, such as flutes (see earlier) do not form. Perhaps the substrate is too firm for the applied turbulent bed shear stresses to cause erosion (cf. Fig. 6A), and thus flutes do not form. A similar argument might explain why grooves and chevrons are typically preserved pristinely, apparently unmodified by turbulent flow.
Consequently, the different observations combined (Fig. 17A) indicate that low-concentration turbulent currents are highly unlikely to produce and preserve groups of parallel to sub-parallel grooves or chevron marks. Nonetheless, it may be possible for isolated short grooves to form from the movement of individual particles overpassing a deformable bed, in a manner similar to the snowball effect, albeit the grooves may not be as regular if particles roll, and the tool would be expected to be present at the end [see for instance fig High-density turbidity currents (HDTCs) with traction carpet (Fig. 17B) Many of the postulated mechanisms in the literature for groove formation involve high-concentration layers at the base of turbidity currents. D _ zuły nski & Sanders (1962a) suggested turbulent traction carpets that restrict the impact of largescale turbulent eddies from the main body of the flow. Dense concentrations of near-bed sand, approaching laminar conditions, were: (i) inferred for some thin, closely spaced groove marks by D _ zuły nski & Walton (1965); (ii) invoked in the form of a thin high-viscosity fluidized sheet at the base of turbidity currents by Hsu (1959; see also Sanders, 1965); and (iii) proposed as having formed in flows exhibiting turbulence suppression because of high near-bed concentrations of suspended particles (Ricci Lucchi, 1995). Draganits et al. (2008) also suggested that the basal layer of the flow was laminar, and might be formed by the head of concentrated density flows (sensu Mulder & Alexander, 2001). These high-concentration layers are associated with flows that are interpreted in current classifications as high-density turbidity currents (Kneller & Branney, 1995;Mulder & Alexander, 2001;Talling et al., 2012a). Such flows form massive or graded Bouma T A deposits, possibly with inversely graded layers at their base (T B-3 of Talling et al., 2012a), as a result of incremental deposition under high-concentration turbulence-damped conditions (Sohn, 1997;Talling et al., 2012a) or velocity fluctuations on the scale of seconds (Cartigny et al., 2013). The key controlling difference is the sediment fall-out rate, with lower rates associated with inversely graded layers, and higher rates with massive or normally graded T A deposits (Sumner et al., 2008).
The highest-concentration basal layers have been called 'traction carpets' (D _ zuły nski & Sanders, 1962a; Lowe, 1982) or laminar shear layers (Vrolijk & Southard, 1997;Sumner et al., 2008) and are thought to be at most a few centimetres thick (Hiscott, 1994;Sohn, 1997;Talling et al., 2012a), although the experiments of Sumner et al. (2008) only produced layers <5 mm thick. Inverse grading, where present, occurs as a result of larger particles moving away from the bed, through a geometrical mechanism of larger particles moving over smaller ones, and kinetic sieving as smaller particles migrate downward (Sohn, 1997;Dasgupta & Manna, 2011). Dispersive pressure is not an important process in traction carpets as implied in earlier work (Sohn, 1997;Dasgupta & Mann, 2011). High-speed imaging of large particles close to the bed in the experiments of Postma et al. (1988) also reveals that particles do not remain at fixed heights within the flow, but rather move vertically and rotate within the flow. Experimental work has also shown that clasts higher in the flow preferentially glide along the top of these high-concentration basal layers (traction carpets) rather than settle through them (Postma et al., 1988), explaining discontinuous mudstone clast layers in discrete horizons within Bouma T A beds, at bed amalgamations, or dispersed within the flow (Hiscott et al., 1997;Talling et al., 2012a).
Given these processes, it does not appear that larger clasts would be incorporated and maintained within traction carpets and therefore be dragged along the bed and form grooves (Fig. 17B). In many cases, the clasts, as shown by groove dimensions, are considerably larger [typically a few tens of millimetres, and rarely >1 m in diameter (Allen, 1984)] than the thickness of the traction carpets. In fact, typical grain sizes within these layers are less than a tenth of the thickness of the traction carpet (Sohn, 1997). Furthermore, larger particles preferentially move away from the bed if initially incorporated into a traction carpet that is forming, or glide along the upper surface of the traction carpet rather than sinking in once a high-concentration layer has formed. Both mechanisms also militate against the formation of repeated chevrons, where a tool needs to be held at a constant height above the bed, or partly within the bed. The dense medium of a traction carpet that extends to the bed would also appear incapable of enabling the formation and propagation of the 'bow' waves around clasts that are thought to form chevrons.
High-density turbidity currents (HDTCs) with high-concentration basal layer (Fig. 17C) High-concentration basal layers can also be formed in HDTCs in the absence of traction carpets (e.g. Lowe, 1982;Baker et al., 2017), either during initial tractional sedimentation prior to the development of traction carpets (Lowe, 1982) or during bypass of the HDTC (Baker et al., 2017). Failure of the sediment bed in canyon systems has also been postulated to lead to the downstream formation of a high-concentration basal layer beneath a turbidity current (Paull et al., 2018). Tractional sedimentation in sand-rich HDTCs is typically dominated by upper-stage plane beds and dune-like bedforms, associated with a turbulent flow regime, prior to increasing sediment concentrations near the base leading to the development of traction carpets (Lowe, 1982). In a supercritical regime, high-concentration basal layers have been postulated to develop over cyclic steps (Hughes Clarke, 2016;Paull et al., 2018). This initial turbulence-driven tractional regime is highly unlikely to be able to maintain larger clasts in fixed positions and at a constant height within the flow, and thus form grooves, for the reasons discussed above for 'low-density turbidity currents'. In the postulated high-concentration basal layers of Monterey Canyon, associated with upslope migrating bedforms, large (ca 0.45 m diameter) spherical and cuboid instrumented 'artificial-clasts' are observed to rotate within the flow, and are thought to be rafted in, or at the upper interface of, the dense layer (Paull et al., 2018). These 'artificial-clasts' further suggest that clasts are not dragged in a fixed position at the base of high concentration basal layers. In the same campaign, an 800 kg tripod was moved several kilometres down canyon. Whilst the movement of the tripod demonstrates the power of such flows, the authors do not consider the density or dimensions of the structure to be representative of a natural clast (6000 kg m À3 ; 2.5 tall, 1.5 m long legs and a large basal cross-sectional area, see fig. 4b of Paull et al., 2018) and once tipped over it is likely to be hydrodynamically stable. Furthermore, the tripod was observed to stop and start during a flow event ( fig. 4c of Paull et al., 2018) indicating that it was not fixed in place in the flow. Thus, the available evidence suggests that these postulated high-concentration basal portions of turbidity currents are not capable of holding clasts in a fixed position. For coarser gravel-rich HDTCs, few tractional structures are formed, and traction carpets are thought to dominate (Lowe, 1982).
Most current knowledge of HDTCs is based on their depositional characteristics, in the form of bedforms and traction carpets, as discussed above. However, HDTCs may also exhibit a bypass phase, as can be observed in laboratory experiments (Baker et al., 2017). Little is known about the structure of HDTCs during bypass, with wide variation in estimated flow concentrations for what constitutes HDTCs: between 5% and 9% (Mulder & Alexander, 2001), >10% (Talling et al., 2012a), >20 to 30% (Lowe, 1982) or from ca 7 to 45% (Kuenen, 1966;Middleton, 1967). Density stratification of gravity currents is also known to be important, and thus these estimates of flow concentration are likely not bulk concentrations but instead reflect basal conditions (e.g. Peakall et al., 2000;Peakall & Sumner, 2015). During sediment bypass, HDTCs may have lower basal sediment concentrations (less pronounced stratification), with these only increasing as flows decelerate and sediment falls out from suspension rapidly (Peakall & Sumner, 2015). However, herein the possibility is considered that flows may bypass with basal concentrations at which hindered settling and dispersive pressure become important (e.g. Lowe, 1982). This would dampen turbulence and potentially lead to near-bed turbulence being extinguished through reduction of mixing (Cantero et al., 2012(Cantero et al., , 2014, and/or near-bed turbulence suppression through the transitional behaviour of clays present within the flow (Baas & Best, 2002;Baas et al., , 2011Baas et al., , 2016b. This, in turn, may lead to laminar basal layers. However, larger particles would not be expected to remain at a constant height within a hindered settling zone, with large particles either falling through the layer, or moving away from the bed if dispersive pressure is important enough (Fig. 17C). If turbulence is extinguished entirely, flows may undergo rapid sedimentation with little if any tractional component (Cantero et al., 2012). Consequently, larger particles are unlikely to remain at fixed heights within the flow and be dragged through a substrate to form grooves (Fig. 17C). Granular flows (Fig. 17D) Grain or granular flows are sediment gravity flows composed of cohesionless grains maintained by dispersive pressure induced by grain-to-grain collisions (Bagnold, 1956;Lowe, 1976a). Groove marks formed by granular flows have been interpreted from pyroclastic flows, with Pittari & Cas (2004) interpreting the formative flow as a highly concentrated granular flow that was capable of keeping clasts in a fixed position, and noting that the flows had 35 to 40% fine-grained ash. Similarly, Sparks et al. (1997) argued for a dense concentrated granular avalanche, and assumed this had a fine-grained component. Grooves have also been recognized in other pyroclastic flow deposits (Cole et al., 2002;Sparks et al., 2002), albeit Sparks et al. (2002) attributed groove formation to a turbulent flow component at the head.
Grain flows, particularly polymodal sand and gravel flows, are able to form in deep-water environments (e.g. Middleton, 1970;Middleton & Hampton, 1973Lowe, 1976a;Iverson et al., 1997;Henstra et al., 2016), but they may be restricted to comparatively steep slopes (more than a few degrees; Lowe, 1976a). Furthermore, in subaqueous deep-water environments, the fine-grained silt-clay component, if more than a few percent, is likely coupled to the water phase producing a debris flow (Lowe, 1976a;Iverson, 1997;Pittari & Cas, 2004), and so the prevalence of granular flows will be restricted. It is unclear whether subaqueous granular flows without a significant fine-grained component are able to maintain clasts in a fixed position without clast rotation (Fig. 17D). This might be possible only if the grains in the basal part of the flow lock together and the flow then glides downslope by inertia (pers. comm., George Postma). However, grooves beneath granular flows have not been reported in deep-water systems, suggesting that this is unlikely.
Fluidized, liquefied and nearly-liquefied flows ( Fig. 17E and F) Fluidized flows with an overriding gravity current (Sanders, 1965), or thin fluidized traction carpets as discussed earlier (Hsu, 1959;Sanders, 1965), have been suggested as mechanisms for the formation of grooves. Truly liquefied flows are produced where pore pressure equals the weight of the grains, leading to the grains temporarily losing contact with one another and floating within the surrounding fluid (Lowe, 1976b). Re-sedimentation then occurs from the base upward as grains settle through the fluid. Consequently, subaqueous liquefied flows, even of coarse silts and sands, are unlikely to move more than a kilometre, since resettling takes place relatively quickly (Lowe, 1976b). In contrast, truly fluidized flows have an external source of fluid that enables the upward velocity of water to match, or exceed, the settling velocity of the grains (Lowe, 1976b). Fluidized flows are therefore likely restricted to very thin flows generated from the tops of liquefied flows, and thus are considered to be unimportant in deepwater settings (Lowe, 1976b). In both cases, flows have no strength and cannot hold a tool in place and drag it through a substrate; consequently, such flows will not be associated with groove or chevron formation (Fig. 17E).
However, in some cases, usage of the terms 'liquefied' and 'fluidized' flows has altered from these definitions, leading to potential confusion. Liquefied flows (Talling et al., 2013a) and liquefied debris flows (Talling et al., 2012a) were defined where: "it is unknown whether all, or a significant part of the sediment weight is borne by excess pore fluid pressure" (Talling et al., 2013a); and "where excess pore pressures primarily support the grains" (Talling et al., 2012a), respectively. Both definitions extend liquefied flows to those with pore fluid pressures below where liquefaction occurs, thus reflecting different flow processes. Analogies are made with subaerial debris flow experiments where pore pressures in excess of the total normal stress have been recorded, implying that these experimental flows were sometimes liquefied (Iverson, 1997;Major & Iverson, 1999;Iverson et al., 2010). However, the measurements in those experiments were local, recorded by sensors at the base of the flow (Major & Iverson, 1999). Typical pore pressures within such subaerial debris flows were well below the liquefaction limit, balancing about 80% or more of the total normal stress (Major & Iverson, 1999), and they were described as 'nearly liquefied' (Iverson, 1997;Major & Iverson, 1999;Iverson et al., 2010). Driving forces for these elevated pore pressures are contractive shearing, where sediment undergoes rapid contraction as a result of shear from an overlying flow, which can lead to a very rapid rise of pore pressure (Iverson et al., 2000;Iverson, 2005) and sediment consolidation (Iverson, 1997); note that contractive shearing may be a mechanism for Bouma T A formation (see Appendix S1). In the case of sediment consolidation, it will be progressively hindered (rather than monotonic; Iverson, 1997) as pore pressures rise, and thus the process of pore pressure increase from this mechanism is self-regulating, with compaction unable to drive the flow towards true liquefaction. The elevated pore pressures in the body of these flows enhance flow mobility and keep shear strength very low (Major & Iverson, 1999;Iverson et al., 2010). Consequently, tools are unlikely to be held in fixed positions at the base of these flows whilst dragged through a substrate, and grooves and chevrons should not form below the body of such flows (Fig. 17F). In contrast, the fronts of these nearly-liquefied flows in subaerial environments do not exhibit elevated pore pressures as they are relatively dry and thus have higher shear strengths (Iverson, 1997;Major & Iverson, 1999;Iverson et al., 2010). Whilst the pore pressure distribution in the fronts of subaqueous debris flows is unknown, these flows will be wet, and thus herein it is postulated that shear strength will not be as high; therefore grooves and chevrons are much less likely to be formed at the flow front in subaqueous flows (Fig. 17F).
Fluidized subaqueous density flows have been claimed to be long-lived based on experiments and field studies. These studies are critiqued in Appendix S1, which concludes that the experiments (Ilstad et al., 2004a;Breien et al., 2010) are neither fluidized nor liquefied, and the F5 facies of Mutti and co-workers (Mutti, 1992;Tinterri et al., 2003;Mutti et al., 2009) is not formed by fully fluidized flow. Consequently, fluidized subaqueous flows are not considered further as a mechanism for groove formation. Thus a return to the definitions of liquefied and fluidized flows are recommended here, as envisaged by Lowe (1976b), reflecting the mechanisms of liquefaction and fluidization themselves. The elevated pore pressures recognized in experimental subaerial debris flows (Iverson, 1997;Major & Iverson, 1999) may be typical of some subaqueous debris flows; the 'liquefied flows' and 'liquefied debris flows' of Talling et al. (2012aTalling et al. ( , 2013a, are herein renamed 'nearly-liquefied flows' (Iverson, 1997). However, the presence of clasts that have been transported over long distances in most debris flows in deepwater systems indicates significant yield strength (and thus lower pore pressures). These latter examples are discussed below in the 'Debris flows and hybrid events' section.
Debris flows and hybrid events (Fig. 17G, H and I) Processes of groove formation. Based on outcrop studies, only Draganits et al. (2008) and Pyles & Jennette (2009) have suggested that grooves on the base of sand beds might be the product of debris flows (equivalent to quasilaminar plug flows and laminar plug flows; see earlier), and in the former case this was one of two suggestions (see section on High-density turbidity currents with traction carpet). Pyles & Jennette (2009) inferred that grooves in the Carboniferous-aged Ross Formation, Ireland, were formed by shale clasts being dragged by a laminar flow, interpreted as a debris flow, and noted that the clasts in the overlying debrite scaled with the width of the grooves. However, no other process arguments were provided to substantiate this interpretation. Additionally, the 'slide marks' of Kuenen & Sanders (1956) and Kuenen (1957), a sub-classification of grooves (see introduction to the Groove casts section), were interpreted as formed from slumping, although debris flows were not specifically considered. It is argued here that debris flows, and the debritic flow components of hybrid events, can account for the observed attributes of grooves, in addition to those formed from slumps and slides. In contrast to the other mechanisms discussed above, debris flows have been shown to form multiple parallel to sub-parallel grooves in a natural seafloor example ( Fig. 12; Kastens, 1984). Such flows have cohesive strength, are loaded with clasts, and can exhibit laminar conditions, thus enabling clasts to be held in position at the base of the flow. However, in order for clasts to erode the substrate there needs to be a slip condition at the base of the flowthat is the clasts need to be moving relative to the substrate. A slip condition can occur in one of two ways. If the plug flow extends all the way to the flow base (a laminar plug flow; Fig. 17G), then there can be a slip condition at the base (for example, Fig. 18). Alternatively, the flow may exhibit a shear-layer that separates the plug flow region from the basal viscous sub-layer (quasi-laminar plug flow), as shown experimentally in  and in the re-analysis herein of data from Hermidas et al. (2018). In this QLPF case, then although at the base of the flow the velocity goes to zero, there is still slip between the base of the plug flow, where the clasts are attached, and the bed (Fig. 17H and I). In both cases, clasts (tools) at the base of the plug zone are moving faster than the substrate and therefore can be dragged through a substrate in straight lines at a fixed depth, thus forming grooves (Fig. 17H).
Subaerial debris flows are known to transport particles adjacent to the bed in their heads where there is a slip condition, followed by basal particles moving vertically and laterally away from the bed (Johnson et al., 2012;Fig. 18). The migration of these particles would explain both cross-cutting grooves (i.e. the distribution of groove orientations around a single maximum direction; see Groove orientations section), and also the absence of particles at the ends of grooves ( Fig. 17G and H). Particles are simply uplifted into the main debris flow and transported down system away from the site of groove formation (Fig. 18). Lobes and clefts, which have been postulated as a mechanism for cross-cutting grooves in turbidity currents, may occur, but they have not been observed in subaqueous debris flows in the laboratory (Sohn, 2000). A cohesive flow also explains the lack of particle rotation observed in most groove casts, since particles are held in place by the cohesive strength of the flow (Fig. 17G to I). Debris flows are known to concentrate larger particles towards the front of the flow (Iverson, 1997;Gray & Kokelaar, 2010;Johnson et al., 2012), and these outsized clasts may therefore be the primary tools for groove formation, explaining the limited cross-cutting of grooves ( Fig. 17G and H). In subaerial debris flows, large clasts also accumulate at lateral margins, but these are rapidly deposited so are less likely to form grooves (Gray & Kokelaar, 2010). Similar parallel longitudinal grooves have been observed in experiments with subaqueous debris flows where hydroplaning at the front provides a slip-component ( Fig. 15; Hampton, 1970;Middleton & Hampton et al., 1973). Grooves are also observed in other natural flows where the tools are supported by flows with cohesive strength, as shown by the giant grooves (kilometres to tens of kilometres long) at the base of large-scale mass transport deposits (MTDs) observed in three-dimensional seismic reflection data (e.g. Posamentier & Kolla, 2003;Gee et al., 2005;Ortiz-Karpf et al., 2017;Soutter et al., 2018). The longitudinal continuity of these MTD grooves raises the question as to what the longitudinal extent of individual grooves is in deep-water clastic systems (see later discussion). Analogies can also be made with Note that the groove is being cut in a downstream direction, but that the base of the flow behind the head is moving more slowly than the front speed, therefore in a moving-frame of reference as shown here the groove appears to be cut upstream. Flow dynamics modified from Johnson et al. (2012). glacial striae on rock (Allen, 1984), where the tools are 'welded' onto the bottom of an ice sheet, to the formation of mega-scale glacial lineations by fractured ice at the base of ice-sheets (Clark, 1993;Piasecka et al., 2018), and to a lesser extent the grounding of ice in sediments in oceans (Reimnitz et al., 1977;Vogt et al., 1994;Piasecka et al., 2018), where the weight of ice maintains the tool on the substrate. A further groove-like feature, referred to as 'glide tracks', is formed by large outrunner blocks in front of debris flows (Prior et al., 1984;Nissen et al., 1999). However, unlike grooves formed beneath the parent flow, these typically exhibit depth changes along the track at distances on the order of the glide track width, reflecting the lack of stability of the blocks (Kuijpers et al., 2001;De Blasio et al., 2006). This is in marked contrast to the uniformity in depth observed in grooves in outcrop.
Chevron tool marks are also more readily explained by QLPF debris flows, because these allow a particle to be held at a fixed height above, or partly within, the substrate, thus enabling a series of repeated bow waves to form around the particle, and the formation of the chevrons (Fig. 17I). In contrast, as noted earlier, it is hard to envisage a mechanism for maintaining a particle at a fixed height above the bed for either turbulent low-density turbidity currents or high-density turbidity currents with a highconcentration basal layer. The generation of chevrons would, however, suggest that the basal layer is sufficiently fluidal that bow waves are able to propagate through it to the substrate surface (Fig. 17I). The type of chevron (uninterrupted, cut or interrupted, Fig. 16) may in turn reflectand in deposits, predictthe thickness of this fluidal layer, and thus the depth from the bed to the base of the plug layer, relative to the size of the cutting clasts.
Nature of debris flows forming grooves. Taking groove widths in the range 10 to 100 mm (see earlier discussion), and assuming that the clasts that created them are of the same diameter (Kastens, 1984;Fig. 12) rather than asperities of much larger particles, the minimum cohesive strength of the debris flow required to support the clasts can be calculated. Using the approach of Talling et al. (2012a), based in part on Hampton (1975), suggests that grains of 10 to 100 mm diameter and of typical densities (based on mud densities 0 to 10 m below the seafloor taken from Flemings et al. (2006)) can be transported by low-strength (1 to 10 Pa) to intermediate-strength (10 to 100 Pa) debris flows. The boundary between low and intermediate densities is at an approximate diameter of 20 mm, and thus most of these clast sizes would require intermediate-strength debris flows (Fig. 3). These strengths equate to volumetric kaolinite concentrations of between >13 to 30%, although more cohesive clays, such as bentonite, would produce the same strengths at considerably lower volumetric concentrations (Marr et al., 2001;Baas et al., 2016b;Baker et al., 2017). These intermediate-strength debris flows are mobile enough to traverse low-gradient slopes and reach fan fringes, and to produce relatively thin (<2 m thick) deposits (Schwab et al., 1996;Talling et al., 2004Talling et al., , 2010Talling et al., , 2012aDucassou et al., 2013). Hydroplaning of debris flows, where ambient fluid is injected beneath the head (Hampton, 1970;Mohrig et al., 1998), may occur in these intermediate-strength flows (cf. Baker et al., 2017) but preferentially occurs for high-strength debris flows (Ilstad et al., 2004b;Talling et al., 2012a), albeit that flow velocity is also a controlling factor in hydroplaning.
Where flows undergo hydroplaning, tools will groove the bed if larger than the thickness of the basal water layer beneath the debris flow, or grooves may be formed by the debris flow immediately behind the hydroplaning head. The likely rarity of hydroplaning in these intermediate-strength debris flows suggests, however, that outsized clasts towards the front of the flow may be responsible for the formation of grooves, albeit flow in the body of the current may also be able to generate grooves.
Hybrid events and groove formation. The debritic component of clast-rich hybrid event beds is typically associated with intermediate-strength cohesive debris flow (Talling, 2013), and these can reach the very distal portions of submarine lobes and basin plains (Talling et al., 2004;Hodgson, 2009). Hybrid event beds, including clastrich types, are common in these distal locations (Haughton et al., 2003Talling, 2013), with hybrid event beds accounting for >31% (Fonnesu et al., 2018) and >83% (Spychala et al., 2017a), respectively, of total thickness in the basin plain and lobe frontal fringes. Some flows may erode and generate clasts a considerable distance updip, as shown by exotic mud clasts (Haughton et al., 2003Talling et al., 2007aTalling et al., , 2012b. In combination with the absence of the debritic component of hybrid event beds in proximal areas, this observation of extensive clast transport suggests that hybrid event beds are capable of bypassing a debritic component over large longitudinal distances; for instance, tens of kilometres in the case of the Marnoso-arenacea ( Fig. 14; Tal ling et al., 2012b;Talling, 2013). Other hybrid event beds likely source mudstone clasts more distally (Hodgson, 2009;Kane et al., 2017;Fonnesu et al., 2018). However, given that flows must transition from being primarily erosive, thus generating mudstone clasts, to primarily depositional, and that debris flows deposit en masse, bypass of the clast-rich component likely also occurs.
Some models of hybrid events assume that the debritic component travels across an underlying sand (e.g. Haughton et al., 2009) or above a sand-rich 'high density flow' (Fonnesu et al., 2016), in which case it is unlikely that the debritic component can groove the seafloor through this layer. However, such models are largely based on studies of depositional hybrid events. Models of hybrid events transforming from an initial debris flow do show a separate bypassing debritic component (Haughton et al., 2003, fig. 11C). Furthermore, the evidence for largescale bypass of some hybrid events and the absence of deposits in proximal areas, suggests that the hybrid events were not depositional at all points and thus travelling over sand beds. Similarly, the evidence for active erosion of the seafloor, demonstrates that the hybrid event cannot be travelling over a pre-existing sand-layer. Furthermore, the process arguments presented herein suggest that a sand-rich high-density flow as envisaged by Fonnesu et al. (2016) has insufficient cohesive and frictional strength to hold tools in a rigid position and thus groove the bed. Instead, these process mechanisms strongly suggest that the debritic component must be a separate component in touch with the bed during the bypass phase. These observations and arguments are in keeping with those of Talling (2013) who tackled the question of why the debritic component almost always ends up above the deposits of a high density turbidity current deposit, and is capped by the deposit of a low density turbidity current. Talling (2013) concluded that this is due to the longitudinal structure of the flow, with the basal sand likely deposited by a forerunning HDTC that moves faster than the debris flow component, which in turn moves faster than the LDTC (Fig. 19).
Intermediate-strength debris flows, not associated with hybrid event beds, can also occur in distal areas, albeit more rarely (e.g. Ducassou et al., 2013;Spychala et al., 2017b), and these bypassing flows should also act to form grooves up-dip. Therefore, the presence of grooves and chevrons indicates that they were cut by a clastrich intermediate-strength debris flow (QLPF to LPF) and, given their prevalence, in many cases by a debritic component of a hybrid event.
Grooves and hybrid event beds: implications for hybrid event bed processes Some studies show that hybrid event beds often lack grooves at their base (e.g. Haughton et al., 2009;Jackson et al., 2009;Talling et al., 2012a;Grundv ag et al., 2014), although in some cases grooves may be observed up-dip where outcrop allows such correlation (Figs 11A and 14A;Fonnesu et al., 2018). However, grooves have been observed on the bases of some hybrid event beds (HEBs) containing both clast-rich and clast-less debrite units (Figs 11B and 14A;Talling et al., 2004Talling et al., , 2012aPatacci et al., 2014;Southern et al., 2015;Fonnesu et al., 2016Fonnesu et al., , 2018, albeit it should be noted that debrites in HEBs can show significant spatial variations ( Fig. 14B; Fonnesu et al., 2015). A more detailed assessment of the presence or absence of grooves at the base of hybrid event beds is not possible at present, as grooves are often reported for entire outcrop sections, usually as palaeocurrent indicators, rather than specifically linked to hybrid beds (Spychala et al., 2015(Spychala et al., , 2017bMalkowski et al., 2017Malkowski et al., , 2018, or tool marks are treated as a single category, and thus grooves are not specified (e.g. Hodgson, 2009). The present paper concentrates on those examples where grooves are present. Examples of grooved hybrid event beds have been interpreted following the standard paradigm that views them as the product of erosion under the head of a turbulent turbidity current, potentially with a dense stratified basal layer (Fonnesu et al., 2016;Fig. 19A). However, as argued previously, low-density and high-density turbidity currents are unable to explain groove formation, and the grooves themselves indicate erosion by a dominantly bypassing debris flow component. The grooved surfaces are, in turn, overlain by clean sand, followed by a debritic interval and finally more sand (Talling et al., 2004;Patacci et al., 2014;Fonnesu et al., 2016Fonnesu et al., , 2018, to produce the typical tripartite hybrid event bed signature (Haughton et al., 2003Talling et al., 2004;Fig. 19B).
The presence of grooves up-dip from the deposits of hybrid event beds (for example, Figs 11A and 14, Beds 3 and 5) may signify erosion by a clast-rich debritic component, with or without a forerunning turbidity current, followed by deposition from a later turbiditic component (Fig. 20A). Beds 1, 3 and 6 ( Fig. 14) show examples of clast-less debrites running to the limits of the outcrop suggesting that if there is a clast-rich unit towards the front of the flow then it is beyond the outcrop limits. This model has analogies with the model of hybrid event beds where a debris flow erodes or bypasses material up-dip, before undergoing flow transformation and successive deposition of sand from a Fig. 19. Models of hybrid-bed generation. (A) Forerunning turbidity current that cuts grooves, followed by a multi-layered flow, with high-density flow (HDF) overlain by a plug flow, and in turn a low-density flow (LDF). Modified from Fonnesu et al. (2016). (B) Standard model of hybrid-bed formation with deposition of sand from a forerunning turbidity current, followed by deposition of a clast-rich debrite, 'L' and 'T' represent laminar and turbulent flow, respectively. The resulting deposit consists of sand at the base, an overlying clast-rich debrite, and finer-grained deposits (silts or sands) at the top, to give a hybrid-bed. Modified from Haughton et al. (2009). (C) Debris flow with either a forerunning turbidity current depositing sand, or sand separating and settling at a late stage from the laminar plug. Modified from Talling (2013). forerunning turbidity current, followed by debritic deposition, to form the hybrid bed down-dip (Talling et al., 2004). However, in these models (Talling et al., 2004), the trailing turbidity current that forms the capping component of the tripartite hybrid event bed, and that might be expected to be deposited on the grooved surfaces, is low-density, forming thin beds characterized by T CDE divisions. Consequently, this model cannot explain the presence of high-density turbidity current deposits overlying grooved surfaces, up-dip of hybrid event beds (for example, Figs 11A and 14; Fonnesu et al., 2018). The dominant model of hybrid event bed formation, where flows gradually transform down-dip from non-cohesive to cohesive, would have the same issue ( Fig. 19B; Haughton et al., 2003Haughton et al., , 2009Talling et al., 2004); assuming that a debritic component does interact with the substrate at some point, and therefore forms grooves, the deposits overlying such erosive surfaces would be expected to be composed of low density turbidites.
In cases where grooves are present at the base of hybrid event beds, the longitudinal flow transformation model from turbulent turbidity current at the front, through a following debris flow and then a dilute turbidity current ( Fig. 19B; Talling et al., 2004;Haughton et al., 2009), is inapplicable because, as argued earlier, the basal surface is the product of a debris flow, as also is the division above the basal sand. Three models can be postulated to explain the observations. Firstly, erosion by the head of a turbidity current may lead to increasing mud and mud-clast content and local transformation into a debris flow (e.g. Talling et al., 2004;Kane et al., 2017), which cuts the basal grooves (Fig. 20B). This debris flow may then be followed by a turbidity current component and then the debritic component, as in the model of Haughton et al. (2009; Fig. 20B). Secondly, the flow may be broken longitudinally into a series of debritic and turbiditic components, reflecting retrogressive failure up-dip (Piper et al., 1999;Brooks et al., 2018b), or the debris flow component may split into a series of discrete blocks during flow transformation (Felix & Peakall, 2006;Felix et al., 2009). In this case, the first debritic pulse cuts the grooves, followed by deposition from the subsequent turbidity current and debritic components (Fig. 20C). However, such processes would be expected to form multiple stacked debrites, in contrast with vertical sequences of hybrid event beds (Haughton et al., 2003Talling et al., 2004;Fig. 19B).
Thirdly, the deposits may reflect a bypassing debris flow that cuts the grooves and then starts to become vertically stratified, with sand deposition from the debris flow followed by deposition of the debrite component (Fig. 20D). The influence of vertical stratification in hybrid event beds has previously been suggested from both laboratory (Baas et al., , 2011Sumner et al., 2009) and field studies ( Fig. 19C; Talling et al., 2004Talling et al., , 2010Talling et al., , 2012a. However, sand separating from a mudsand debris flow is physically simpler (and has been modelled experimentally), than from a debritic phase with larger mud clasts. In the latter case, the mudstone clasts would need to be less dense than the flow, whereas the sand would have to be denser than the flow and thus able to settle. However, as flow density must be high for the larger mudstone clasts to remain supported then yield strength would be expected to be significant. Given this, the sand would have to settle through a dense, possibly high strength, material. Alternatively, if a clast-rich debrite is at the front of the flow, followed up-dip by a clast-less debrite, sand may separate from this component. It remains unclear how feasible such a mechanism is. In summary, with the possible exception of the case where the front of the turbidity current develops into a debris flow through erosion (Fig. 20B), all of the models suffer from limitations. Flow transformation from an initial debris flow (Fig. 20A) does not predict high-concentration turbiditic deposits above grooves. Longitudinal flow segregation with multiple debritic components (Fig. 20C) predicts too many debrites in the hybrid event bed. Finally, in the case of the vertical segregation model (Fig. 20D), there are issues concerning how sand segregates from a mixture of mud and large mud clasts, or whether there are longitudinal variations from clastrich to clast-less debrite. Flows forming hybrid event beds therefore potentially have more complex longitudinal and temporal changes than have hitherto been postulated, and these account for observations such as high-density turbidity current deposits overlying grooved surfaces up-dip of hybrid beds.
An alternative to the different models of hybrid event beds is that the flow forming the grooved surface is entirely separate from the flow forming the overlying deposits (Fig. 20E). In this case, an initial debris flow, slide or slump cuts a grooved surface on the seafloor after the flow bypasses down-dip. Given that debris flows typically either deposit en masse or not at all, the grooves may be left in pristine form on the seafloor as the flow bypasses, as seen in Fig. 12 (Kastens, 1984), or may be covered by a thin layer of unconsolidated mud from minor flow transformation of the top of the debris flow, as well as any subsequent hemipelagic deposition. A subsequent turbidity current may 'ingest' any unconsolidated mud (of low strength, see Seafloor substrates section) and deposit onto the grooved surface.
Herein it is also noted that flutes can be present at the base of some hybrid event beds (e.g. Patacci et al., 2014), often in beds without clastrich components (Talling et al., 2004(Talling et al., , 2012aFonnesu et al., 2018), or on the same surfaces as grooves and other tool marks (Fonnesu et al., 2016). This indicates that turbulent or transitional flows also eroded the basal erosion surface, and in the case of mixed groove-flute assemblages that flow evolution took place, or a later turbulent or transitional flow event eroded grooves from an earlier event. The presence of flutes in the absence of grooves may suggest the occurrence of a forerunning turbidity current, or a cohesive transitional flow. Kuenen (1957) argued against mudstone clasts being the primary source of tools on the grounds that they would undergo rapid rounding through abrasion. Mudstone clasts do indeed abrade on timescales of tens of minutes to hours depending on applied shear stress and composition (Smith, 1972). Such abrasion is observed in turbidity currents, with transported clasts typically showing rounded to sub-rounded clasts (Johansson & Stow et al., 1995). Similarly, hybrid event bed debrites can show rounded to angular clasts, again likely reflecting transport distance (e.g. Davis et al., 2009;Hodgson, 2009). Grooves show both smooth curved internal surfaces and surfaces marked by parallel striae. Consequently, these surfaces likely reflect the nature of clasts that cut them, with longer-travelled sub-rounded clasts likely cutting the smooth surfaces. Striae are likely formed by clast asperities perhaps from recently eroded clasts, although potentially they may be the product of armoured mudstone clasts in some cases. The nature of the groove morphology may, in part, be a reflection of the time period of formation; although the true length of grooves is unknown because of outcrop extent, they are known to be up to tens of metres long. Current velocities for deep-water flows are poorly known (Talling et al., 2013b;Peakall & Sumner, 2015). However, Talling et al. (2012a) calculated that a 1 m thick, intermediate-strength, kaolin-rich debris flow on a slope of 0.1°(typical of mid to lower fans; Pirmez & Imran, 2003) would require a velocity of ca 0.5 to 1.0 m s À1 . Such flow velocities are in-line with velocities measured or estimated for distal turbidity currents (Klaucke et al., 1997;Pirmez & Imran, 2003;Vangriesheim et al., 2009;Stevenson et al., 2014;Peakall & Sumner, 2015). Assuming a velocity of 1 m s À1 , the clast travel time over the length of grooves observed in outcrop equates to seconds to tens of seconds, which is too short a period to result in significant abrasion (e.g. Smith, 1972). Moreover, the mudstone clasts are likely more indurated than the substrate into which they are cutting.

Morphology and abrasion of mudstone clasts
For cases where the extent of individual grooves is not limited by the timescale between initial impingement of a clast on a bed, and being uplifted vertically at the flow front (Johnson et al., 2012;Fig. 20. Models for groove formation by flows forming hybrid beds. (A) Model for grooves found up-dip of a hybrid-bed deposit. A bypassing debris flow, with or without a forerunning turbidity current, cuts a grooved surface, and a later turbidite is deposited on top of the grooved surface. (B) to (E) Models for hybrid beds with grooves at the base. (B) A flow with a debris flow component and a forerunning turbidity current. The head of the turbidity current erodes unconsolidated mud, and mud clasts, undergoing transformation into a debris flow, producing longitudinal segregation from frontal debris flow, through turbidity current, and back into a second debris flow component. (C) Longitudinal flow segregation, with multiple debrite components from initial conditions (for example, periodic retrogressive failure) or from separation and break-up of an initially single debris flow (e.g. Felix et al., 2009). The first debris flow cuts the grooves and is then followed successively by turbiditic and debritic components. (D) A single debris flow, with the frontal part cutting the grooves, followed by later separation and settling of sand from a laminar clast-rich plug flow (see Fig. 19C). (E) An initial debris flow cuts a grooved surface and bypasses down-dip. Given that debris flows deposit en masse, then the grooves may be left in pristine form on the sediment surface, or may be covered by a thin layer of unconsolidated mud from minor flow transformation of the top of the debris flow, and any subsequent hemipelagic deposition. If a turbidity current is generated prior to a thicker consolidated mud developing it may 'ingest' any unconsolidated mud, and then at some point deposit directly onto the grooved surface. In this case, one flow cuts the erosive surface and an entirely separate flow accounts for the deposit.  mudstone clasts (Smith, 1972). Nonetheless, individual grooves are likely to be shorter than the tens of kilometres observed under large MTDs, likely reflecting that in MTDs they are cut by much larger tools (Gee et al., 2005;Ortiz-Karpf et al., 2017;Soutter et al., 2018) where abrasion relative to clast size will be less important, and in at least some cases by stronger tools (e.g. Soutter et al., 2018).  1958) are considered to be the most useful for palaeocurrent analyses since they are asymmetrical, with a longer shallower upstream slope, and a shorter steeper downstream slope; the shallower slope may be ornamented with longitudinal striae (D _ zuły nski et al., 1959;Lanteaume et al., 1967;Allen, 1984). Bounce marks (Figs 21 and 22;Wood & Smith, 1958), also called skim marks (Allen, 1984), are symmetrical to slightly asymmetrical and typically tens to hundreds of millimetres long (>3 m in exceptional cases), <50 mm wide, less than a few millimetres deep and may contain parallel internal striae (Lanteaume et al., 1967;Allen, 1984). Skip marks (Fig. 22;D _ zuły nski et al., 1959) are a series of discontinuous tool marks, typically at a similar spacing, that are produced by a single tool. The morphology of each mark may be almost identical, or variable, but similar enough to be recognizable as being formed by the same tool (D _ zuły nski & Walton, 1965). Skip marks can include tumble marks ( Fig. 22; Allen, 1984), formed from a tool somersaulting, such as fish vertebrae and angular mud clasts (D _ zuły nski & Walton, 1965;Allen, 1984). Skip marks can also consist of a series of bounce marks that can sometimes become almost continuous, approaching the appearance of grooves (Collinson et al., 2006). Lastly, roll marks ( Fig. 22; D _ zuły nski & Sla z czka, 1958) are made by cylindrical objects (for example, fish vertebrae, ammonite and straight orthocone shells) that enable the tool to roll over the bed (D _ zuły nski & Sanders, 1962a;D _ zuły nski & Walton, 1965;Bates, 1974). Discontinuous tool marks can be superimposed on, and consequently can be younger than, both flutes Fig. 22. Styles of discontinuous tool marks as seen in cross-section (x-y) and planform (x-z). Modified from Allen (1984). Striations may occur on skim and prod marks, as illustrated. and grooves (Fig. 21B;D _ zuły nski & Sanders, 1962a;D _ zuły nski & Walton, 1965;Ricci Lucchi, 1995). Given their small size and erosion depths (typically millimetres to tens of millimetres) (D _ zuły nski & Walton, 1965;Collinson et al., 2006) relative to flutes and grooves, any evidence of their formation prior to flutes and grooves may be lost, although where erosion by later forms is limited, evidence for discontinuous tool marks that formed earlier can be found (Fig. 21C).

Discontinuous tool marks
The nature of formative flows for discontinuous tool marks Roll marks require a flow where the particles are not periodically lifted from the bed by buoyant or turbulent forces, and thus flow concentration is likely comparatively low and turbulence limited. Furthermore, roll marks will be hindered by high viscosity and the development of thickened viscous boundary layers. Consequently roll mark formation will be favoured by relatively fluidal flows with comparatively low viscosity, low concentration and limited turbulence; these likely include weak turbulent flows (TF) and lower-concentration transitional flows (TETF and LTPF). Similarly, skip marks can be formed by tumbling particles that are in close contact to the bed, but the tools are either more angular than those involved in roll marks, or experience sufficient lift forces to periodically lose contact with the bed. These tumble marks may also be favoured by weak turbulent flows, or transitional flows (TETF and LTPF). However, a greater spacing between tumble marks implies that a significant buoyant force is present that supports the particle, and thus an association with higher-concentration transitional flows (for example, UTPF).
The asymmetrical morphology of prod marks suggests that tools may exhibit ballistic trajectories like saltating grains, and thus approach the bed at comparatively low angles, before rebounding from the bed and being lifted up at a higher angle (Bagnold, 1973;Francis, 1973;Lee & Hsu, 1994). Experiments with large saltating particles, up to 6 mm diameter, showed a narrow range of incidence angles (10°to 35°) and a take-off angle range of 21°to 87°, with a mean of ca 65° (Ancey et al., 2002), in keeping with qualitative observations of prod marks. Saltation is normally linked to turbulent flows (Pilotti & Menduni, 1997); but Francis (1973) demonstrated experimentally that in higher-viscosity flows composed of glycerinewater mixtures saltation could also occur under laminar conditions, suggesting that for clay-rich flows, transport under transitional flow conditions (Baas & Best, 2002) would also be possible. Laminar or transitional clay-rich flows would also act to provide a buoyant force aiding transport of larger tools.
The presence of fine striae on the upstream slope of some prod marks suggests that the incident grain was not spinning when it impacted the bed. This is surprising since grains typically rotate during saltation, driven by bed collisions, grain-to-grain collisions and velocity gradients across the particle (Best, 1998). Larger particles, like those observed to form tool marks, are known to rotate more slowly than smaller particles, for instance about four or five rotations per second for ca 4.8 mm diameter particles in water (Francis, 1973;Best, 1998) in contrast to ca 40 rotations per second for ca 1.4 mm diameter particles in water (Lee & Hsu, 1994;Best, 1998). Two mechanisms, in addition to increased particle size, may act to reduce or eliminate rotation of tools: increased viscosity (Best, 1998) and reduction in the velocity gradient. Transitional flows possess enhanced viscosities (potentially an order of magnitude or more increase relative to a clearwater flow), and also exhibit greatly reduced shear immediately adjacent to the bed, notably for lower and upper transitional flows, and quasi-laminar plug flow (Baas et al., , 2016b. This basal zone of low shear, representing a thickened viscous sublayer, was ca 6 mm thick in the experiments of , representing ca 4 to 5% of flow thickness. Although how the thickness of this low shear zone scales with flow thickness for larger flows is unknown, for flows that are metres to tens of metres deep the thickness of this basal zone might be expected to be of the order of centimetres to 10 cm. Given that particle saltation heights are typically about two to four times the grain diameter for rotating particles in liquids (Francis, 1973;Fernandez Luque & Van Beek, 1976;Krecic & Hanes, 1997), and potentially as low as a third of this in the absence of rotation (Krecic & Hanes, 1997), saltation trajectories may well be expected to be restricted to this basal zone of low shear if occurring in transitional flows. Consequently, the presence of this basal zone of low shear, in combination with enhanced viscosities, suggests that the formative flows in prod marks exhibiting striae are transitional flows, most likely upper transitional plug flows (see later). An additional control on particle rotation may be the interaction of toolsparticularly for mudstone clastswith a cohesive bed, leading to markedly inelastic collisions. The loss of kinetic energy associated with inelastic collisions might result in a corresponding reduction in imparted angular momentum (Wiberg & Smith, 1985) and thus rotation rates. A reduction or cessation of particle rotation in turn affects particle trajectories through reduction of lift associated with the Magnus effect (Rubinow & Keller, 1961), leading to much shorter saltation hop lengths and lower trajectories (Krecic & Hanes, 1997). Moreover, the reduction or absence of rotation further reduces turbulence generation towards the base of the flow, because rotating grains have been shown to generate additional turbulence (Best, 1998).
In sharp contrast to prod marks, bounce marks are formed by particles that graze the bed with a concave-up trajectory (Allen, 1984). This concave trajectory is atypical of grains in a bedload layer, which typically roll, slide or saltate (e.g. Lee & Hsu, 1994), and suggests that the tool is largely supported by the flow, presumably by the buoyant force. However, the tool is not fully supported, in contrast to particles transported in a well-developed plug flow, such as an intermediate-strength debris flow. Skip marks consisting of repeated bounce marks further argue for a significant buoyant force, since the morphologies do not fit with saltating tools in terms of their longitudinal symmetry and length to width ratio. The observation that some skip marks consist of bounce marks that are almost continuous and start to look like grooves, suggests that in these cases the particles are further supported by the flow, and close to becoming intermediatestrength debris flows that fully support the tools (see Groove casts section). Given the processes identified herein, 'bounce marks' is a remarkably poor choice of term for these features, since unlike prod marks the particles do not really bounce, but rather skim the surface. Consequently the term 'skim marks' used by Allen (1984) is recommended here, because it reflects the key process.
As noted earlier, the change in the morphology of flute marks and the ultimate cessation of their formation is interpreted to be caused by a change from turbulent flow to transitional flow, with a corresponding reduction of the size, and eventual elimination, of the flow separation zones at the flow-bed interface integral to flute formation (Baas & Best, 2008; Flutes section). Discontinuous tool marks do not appear to form simultaneously with flutes, although they can overprint them. Furthermore, as observed here, some types of tool mark, such as skim (bounce) marks, provide evidence for a substantial buoyant force that enables particles to graze the bed in gentle arcs. Herein, discontinuous marks are interpreted to be typically the product of transitional flows. If the flow is either fully turbulent, or in the turbulence-enhanced transitional flow (TETF), or lower transitional plug flow (LTPF) regimes, flutes are likely to form (see Flutes section), assuming that substrate conditions enable mass erosion from the bed. In such flow regimes, substantial bed turbulence is present, encouraging scour, and the mud clasts responsible for discontinuous tool marks may not be supported within the flow. Cessation of flute development likely occurs in the upper part of the lower transitional plug flow regime, or the lower part of the upper transitional plug flow regime (UTPF), because of a loss of turbulence (see Flutes section), at a stage where increasing concentration and viscosity may enable support for tools within the flow. At the other end of the spectrum, if the flow is a plug flow of sufficient cohesive strength to form an intermediate-strength debris flow (upper part of the quasi-laminar plug flow regime, QLPF, or a fully laminar plug flow, LPF), tools will be held rigidly in place within the flow and form grooves. Between these two end members, transitional flows of progressively increasing strength can support clasts, with: (i) prod marks envisaged as forming in the upper part of the UTPF regime (e.g. Baas & Best, 2008); (ii) prod marks with striae at their upstream end likely reflecting stronger flows (uppermost part of the UTPF regime); and (iii) skim marks forming in stronger transitional flows (uppermost part of the UTPF and lower part of the QLPF regimes; ) that possess sufficient density to provide significant buoyant force (Fig. 23). Skip marks dominated by relatively short marks will be the product of an upper UTPF regime, whereas skip marks consisting of repeated longer skim marks are likely formed in the lower QLPF regime. Roll marks are most likely the product of low concentration, low viscosity flows with limited turbulence (TF, TETF and LTPF), whereas skip marks that involve tumbling may reflect either: (i) low concentration, low viscosity conditions; or (ii) if the spacing between tumble marks is greater, a higher buoyancy associated with transitional flows such as UTPF. Both forms require relatively planar beds to pass over, are unlikely to be able to form over surfaces composed of flutes, and cannot form under higher concentration flows such as upper QLPFs (Fig. 23). The influence of tool properties on the nature of discontinuous tool marks The nature of the tools themselves, and the availability of tools, also helps to determine the character of tool marks. However, the tools carried by a flow are at least partially controlled by the fluid dynamics, which will limit the maximum size and density of the particles. Particle shape is a more independent parameter, although as noted earlier (see Morphology and abrasion of mudstone clasts) it is in part a function of travel distance for tools such as mudstone clasts. The shape of particles will affect the nature of the tool marks and in some cases, notably where fossils are the tools, they can produce very characteristic tool marks (e.g. , 1960D _ zuły nski & Walton, 1965;Howe, 1999). Similarly, the presence of fine striae in some prod marks and skim marks, implies particles with sharp asperities. Particle shape is also known to affect saltation with trajectories becoming longer and lower, and thus flatter, with decreasing sphericity (Williams, 1964;Rice, 1991), thus potentially affecting contact angles with the substrate.

DISCUSSION
Distribution and association of scour marks and tool marks: a process explanation The present analysis of formative mechanisms for the range of different sole structures enables an explanation for their observed spatial and temporal distributions. As discussed earlier, flutes are typically associated with thicker sandstones in proximal locations, whilst tool marks are frequently associated with thinner, more distal sandstones, and flutes and tool marks are typically found on different bedding planes at a given point spatially. Furthermore, large bulbous flutes are typically found up-dip from small spindle-shaped flutes (Pett & Walker, 1971). To summarize the key process mechanisms proposed herein, large bulbous flutes are likely formed by turbulent and turbulence-enhanced transitional flows, whereas spindle-shaped flutes are associated with stronger transitional flows (LTPF and lower UTPF). Discontinuous tool marks may be formed by a variety of flow types, from laminar to turbulent. However skim marks are associated with flows with significant buoyant force that enables the tools to gently graze the bed in low curving arcs. Similarly, prod marks with striae at their up-dip ends indicate flows that are comparatively viscous with thickened basal sublayers. Finally, grooves and chevrons can only be formed by debris flows exhibiting a slip condition that have the cohesive strength needed to maintain tools, primarily mudstone clasts, in fixed positions at a given height within the flow that are dragged through the substrate. In the case of chevrons, this requires a fluidal layer at the base of the debris flow plug, suggesting a quasi-laminar plug flow. Observed examples of longitudinal variation from large bulbous flutes, to spindle-shaped flutes, to discontinuous and continuous tool marks therefore suggest increasing flow cohesion downstream, and enables a process-orientated model to be proposed (Fig. 24A).

Increasing flow cohesion downstream
This progressive increase in cohesion is in agreement with the standard model for the formation of hybrid event beds (Haughton et al., 2003Talling et al., 2004) that proposes flows gradually transform downstream from non-cohesive to cohesion-dominated. In this model, as flows ingest mud, and decelerate as slopes decline, cohesive forces begin to progressively dominate over turbulent forces. These rheological changes would lead to a successive decline in flute size and their eventual disappearance, through the development of more transitional flows and discontinuous tool marks, and lastly the formation of grooves and chevrons as flows form debritic components that are likely associated with hybrid event beds (Fig. 24A). This longitudinal relationship between discontinuous tool marks, and grooves and chevrons, has not been demonstrated in the field, because tool marks have not typically been subdivided. However, the presence of hybrid event beds at the distal fringes of submarine lobes and basin plains (Talling et al., 2004;Hodgson, 2009;Spychala et al., 2017a) suggests that, in these cases, grooves and chevrons are preferentially found in distal locations. However, it should be noted that transformations can start and finish anywhere along the transport profile shown in Fig. 24.
Decreasing flow cohesion downstream Whereas the postulated model of longitudinal change from low to high cohesion explains the typical field observations of flutes and tool marks as summarized in the literature, some field observations predict the opposite transition from grooves in T A beds to flutes in T C beds (Table 1; Fig. 24. A process-orientated conceptual model for the longitudinal distribution of flutes and tool marks. (A) The distribution of flutes and tool marks is shown for a flow that is increasing in cohesion with longitudinal distance, as hypothesized for instance for many hybrid event beds. (B) The distribution of flutes and tool marks for flows that decrease in cohesion with distance; note that the order of the sole structures with distance is reversed relative to (A). Note that transformations can start and finish anywhere along the transport path, that flows may also vary temporally at a point, and that flutes and tool marks will vary with substrate conditions (see main text for details). TF = turbulent flow, TETF = turbulence-enhanced transitional flow, LTPF = lower transitional plug flow, UTPF = upper transitional plug flow, QLPF = quasi-laminar plug flow, LPF = laminar plug flow (see Fig. 1 and accompanying text for more detail on transitional flow types). Bouma, 1962;Crimes, 1973). These observations are in keeping with flows that start as high-concentration debris flows (or slumps and slides) and progressively dilute downstream, reducing flow cohesion. In such cases where initial highcohesion flows transform to lower-cohesion flows (e.g. Piper et al., 1999;Talling et al., 2004;Felix et al., 2009), the distribution of flutes and tool marks should be reversed relative to the hybrid event bed model of increasing flow cohesion, with grooves up-dip, then discontinuous tool marks, and finally flutes, assuming that the flow remains erosive throughout (Fig. 24B). Such a scenario is supported by observations of the Annot Sandstone, Pe€ ıra Cava, France, where grooves are present but hybrid beds and mass transport deposits are very rare, and decrease as a proportion of beds downstream (Table 3; Cunha et al., 2017). Some slumps with coeval overlying sands ('welded slump-graded sand couplets') do occur in Pe€ ıra Cava, and are interpreted as the products of partial flow transformation from high to low-concentration (Stanley, 1982). Similarly, extensive grooves are observed in the Zumaia section of the Basque Basin, Spain (Tables 1 and 3; Crimes, 1973) yet hybrid beds are not reported in that time interval (Table 3). Analysis of the distribution of recurrence intervals suggests that the basin-plain beds at Zumaia were sourced from large-scale disintegrating slope collapses (Clare et al., 2014(Clare et al., , 2015. Furthermore, a large, transversely sourced slump, down dip from Zumaia, exhibits grooves at its base (Crimes, 1976). Both the recurrence intervals and the observed slump with basal grooves suggest that, in this basin, the grooves were likely formed from flows transforming from high to low-cohesion flows. Again, it should be appreciated that Fig. 24 is schematic, and flows may initiate with different flow properties and may not transform entirely.
Other sole type distributions Some debris flows may travel very large distances without undergoing significant flow transformation (Ducassou et al., 2013), and therefore may be able to form grooves at any point. Similarly, turbulent flows that do not undergo flow transformation to more cohesive flows may just form flutes, with velocity and flow depth controlling the change from larger to smaller flutes without a viscosity change, as envisaged by Allen (1971a). Potentially, other scenarios are also possible. The present paper postulates that some sediment gravity flows may first transform from high to low-cohesion through a range of dilution mechanisms (e.g. Talling et al., 2002;Felix & Peakall, 2006;Felix et al., 2009), prior to the flow decelerating, with viscous forces becoming more important towards the end of the flow (e.g. Talling et al., 2004). The accompanying changes in flow cohesion would be expected to result in changes from grooves to sole structures associated with lower cohesion (Fig. 24A), and then a switch back to sole structures related to increasing cohesion (Fig. 24B), with the point of lowest flow cohesion determining the range of the more fluidal sole structures. It should also be noted that, whereas the present discussion only considers longitudinal changes in flow properties and thus sole structures, there will also be changes laterally, depending on how flow structure changes from on-axis to off-axis.

Temporal changes in sole structures
The prevalence of flutes or tool marks on a given surface suggests that in many flows, at a given point spatially, the erosive phase of the flow only comprises a single flow type, and thus the major changes in flow type are longitudinal. However, flutes can pre-date or post-date tool marks ( Fig. 21B and C), indicating that in some examples there is also a temporal variation in the nature of flow types within a given event. This assertion assumes that the erosive surface formed from a single event. Flutes followed by tool marks suggests an increased cohesive flow strength over time, which may be associated with a forerunning more turbulent flow phase and a slower, more viscous, later flow component (for example,Figs 19B and 20A), or potentially with increased seafloor erosion (see Fig. 4 and discussion in the 'Seafloor substrates' section). Flutes postdating tool marks indicates that flows have become more turbulent, suggesting that a more dilute turbulent flow phase followed a faster, higher-concentration, flow phase. Amy et al. (2005a) demonstrated experimentally that for stratified gravity currents, both of these scenarios are possible, and that the variation in the vertical distribution of viscosity controls whether the lower viscosity, more turbulent, layer either outruns or lags the higher viscosity layer. These temporal changes in flow properties have been postulated previously as explanations for flutes cutting tool marks or vice versa (D _ zuły nski & Sanders, 1962a;Draganits et al., 2008;Pyles & Jennette, 2009).

Implications for the Bouma sequence
The formation of grooves by dominantly bypassing debritic flow components, and the successive development of grooves and then flutes (or vice versa), demonstrate that the erosive surface and overlying deposits can be produced by different types of current. Thus many sole structures do not have a genetic link to the overlying turbidity current deposit, as encapsulated in the present pictorial Bouma sequence. It is noted here that Bouma (1962) and early workers (D _ zuły nski & Walton, 1965;Harms & Fahnestock et al., 1965;Walker, 1965Walker, , 1967 only considered the five internal divisions of the sequence, and did not incorporate a basal erosive surface. However, later workers added the erosive surface to the base of the Bouma T A division in summary diagrams of the Bouma sequence (Blatt et al., 1972), and then explicitly linked this surface to sole marks, including tool marks (Middleton & Hampton, 1973, to yield the standard pictorial Bouma sequence we know today and that has been almost universally adopted (e.g. Bridge & Demicco, 2008;Leeder, 2011;Talling et al., 2012a;Boggs, 2014;Pickering & Hiscott, 2016;Collinson & Mountney, 2019;Fig. 25).
The recognition of erosion by one phase of the flow, for instance grooved surfaces formed by debris flows, and deposition by a subsequent phase of the flow, producing turbidity current deposits, also implies that the temporal gap between erosion and deposition can be considerable. Whilst a time gap between the erosive surface and the underlying deposit is implicit in the Bouma sequence, this time gap has been assumed to be very short for heterolithic and unconfined settings, based on the pioneering work of Kuenen (1957) who stated: "the conclusion is inevitable that flute casts and drag marks [grooves] result from the same current that deposited the covering bed a moment later". For bypass surfaces, such as channel bases and scours, no genetic linkage between erosive surface and overlying surface is typically implied (e.g. Stevenson et al., 2015), albeit the dominant sand-on-sand surfaces in axial locations (e.g. Hubbard et al., 2014) in these environments may limit the frequency of sole structures. Herein, the authors challenge the belief for unconfined and heterolithic settings that there is a genetic linkage between erosive surface and overlying deposit, and argue that the time gap in the Bouma sequence can be orders of magnitude greater than that previously envisaged.
The erosively grooved surface, if present, can be overlain by the full range of sandy Bouma sub-divisions (T A to T C ) (Table 1; Fig. 11; Bouma, 1962;Pett & Walker, 1971;Crimes, 1973), again illustrating that there can be a temporal disconnect between the erosive and Fig. 25. Evolution of the classic Bouma sequence in pictorial form. Bouma (1962) initially defined five divisions. Blatt et al. (1972) added an erosive base to the A-division a decade later, and Middleton & Hampton (1973 then explicitly linked the erosive base on the A-division to flutes and tool marks. The combination of the Blatt et al. (1972) and Middleton & Hampton (1973 figures gives us the present form of the Bouma sequence and, in many cases, this explicitly links grooves, as well as flutes and other tool marks, to the base of the A-division (e.g. Collinson et al., 2006, as pictured here). depositional phases of the same event. A grooved erosive surface, therefore, is neither a part of the sedimentological record of a waning turbidity current, nor the classical Bouma sequence. Here, a new pictorial version of the Bouma sequence is suggested (Fig. 26), that returns to the original Bouma (1962) sequence as a record of waning turbidity currents, and recognizes that the basal components of the Bouma sequence at a given longitudinal or lateral position can be deposited on an erosive surface that may record waxing turbidity currents, and processes other than turbidity currents. These additional processes recorded by the erosive surface may include debritic flow components, and transformation of flows between debritic and turbiditic components where flutes and groove marks are superimposed.
There remains debate as to whether the erosive surface and the overlying deposit are the product of the same flow, albeit one that may have multiple rheological components. As discussed earlier, most workers suggest that cross-cutting groove marks likely represent the product of single flows (e.g. Kuenen & Ten Haaf, 1958;Allen, 1971b), yet others have argued for multiple flows (Crowell, 1958;Mulder et al., 2002). The present study suggests that debritic flow components are able to erode grooved surfaces whilst bypassing almost all sediment, as demonstrated by a modern example from the eastern Mediterranean ( Fig. 12; Kastens, 1984). Furthermore, there is no record of a mudstone clast lag or coarse granules (particles >1 mm) immediately overlying these surfaces (Talling et al., 2007c;Talling, 2013) and only occasional examples of clasts deposited at the ends of grooves. Consequently, this suggests that intermediate-strength debris flows can bypass the surface entirely, thus allowing successive debris flows to produce cross-cutting marks. Such intermediate-strength debris flows may exhibit limited shear mixing and consequently only generate small-scale turbidity currents (Talling Fig. 26. Revised Bouma sequence in pictorial form, highlighting the time gap between the basal surface and the basal sand-rich division, which can either be the Bouma A, B or C division. The nature of the erosive surface provides information on the flow that formed the surface and subsequently bypassed down-dip. Grooves indicate erosion by a debritic flow component, and therefore a debrite will be located down-dip unless flow transformation has occurred. Flutes indicate that a turbulent flow, or a weaker transitional flow (TETF, LTPF or lower UTPF), formed the surface and a turbidite will be located down-dip, unless flow transformation has subsequently occurred. For simplicity, discontinuous tool marks are not shown. However, prod marks are likely linked to weaker transitional flows, and skim marks, and prod marks with upstream striae, to stronger transitional flows (see text for discussion). There is evidence from some examples that the basal surface may even represent a separate flow event to the overlying turbidite and thus there is no genetic linkage (see text for details). Talling, 2013). A key question is whether the tails of these dilute flows might be expected to lead to deposition of turbiditic sands or just to muds on top of the grooves. The pristine grooves observed on the modern seafloor immediately up-dip of a debris flow deposit ( Fig. 12; Kastens, 1984) show that some grooves can be preserved without any overlying deposit from the flow. It thus raises the clear possibility that the erosive surface and overlying deposits within the Bouma sequence may not be formed by a single flow event.
Formation of flutes and tool marks under the head?
It has long been argued that flutes and tool marks form during erosion by the head of the turbidity current and that deposition takes place in the body (e.g. Kuenen & Ten Haaf, 1958;Middleton, 1967;Allen, 1971b;Cantero et al., 2008). The present study has shown that some tool marks, such as grooves, form under debris flows. Additionally, the presence of superimposed grooves and flutes, formed by debritic and turbulent turbidity current components of the same flow, show that sole marks cannot always form under the head of a flow. Where grooves precede flutes (Kuenen, 1957;D _ zuły nski & Sanders, 1962a;Enos, 1969a;Ricci Lucchi, 1969a;Draganits et al., 2008), there must be erosion in some part of the turbiditic body of the flow, and for the opposite case of flutes preceding grooves, the debris flow component must form grooves under the body of the flow (D _ zuły nski & Sanders, 1962a;Ricci Lucchi, 1969a), unless grooves and flutes are formed by separate flows. In systems with long flow run-outs, there may also be a substantial temporal gap between the erosion of grooves and flutes or vice versa, given that the lengths of clast-rich and clast-less debritic components can be up to several tens of kilometres ( Fig. 14; Amy et al., 2005b;Amy & Talling, 2006;Talling et al., 2012b). The hypothesis that flutes and tool marks can only form in the head is also at odds with the evidence that many turbidity currents bypass for much of their duration, with variations between bypass and erosion (Stevenson et al., 2015). Similar arguments can be made based on the superimposition of discontinuous tool marks on flutes (Fig. 21B) and vice versa (Fig. 21C), representing a change from turbulent flow to transitional flow in the former case, and the opposite in the latter case (Fig. 23). In both cases, this transition between turbulent and transitional flow is unlikely to happen within the spatial extent of the head. It would therefore appear that scour and tool marks are not limited to the head of the flow, and could instead form for a far greater proportion of some flows.

Implications for palaeocurrent measurements
The present synthesis illustrates that, when taking palaeocurrent measurements, it is important to note the type of sole structures measured, because this can provide a host of other information that can aid interpretation of flow properties and enhance prediction. However, there may also be a link between palaeocurrents and sole structure type, at least in areas where interaction with topography is important. Kneller & McCaffrey (1999) pointed out that flow density, and density stratification, influence the nature of topographic interaction. Herein, it has been demonstrated that different sole structures are related to flows that have different density, stratification and rheology. Consequently, it might be expected that at a given point, different flow types, and therefore sole structure type, may give different palaeocurrent measurements when flows interact with topography. Beautiful examples of this are shown from the Marnoso-arenacea, Italy, where grooves show an enhanced variability relative to flutes, which is interpreted to be a result of topographic interaction (Muzzi Magalhaes & Tinterri, 2010, fig. 20;Tinterri & Muzzi Magalhaes, 2011;Bell et al., 2018). However, in other examples of interaction with topography, flutes and grooves show similar palaeocurrents (Tinterri et al., 2016;Cunha et al., 2017). Such variations might reflect the nature of topography, and incident angles, as well as the aforementioned flow properties (cf. Kneller & McCaffrey, 1999).

CONCLUSIONS
This paper presents a radical re-examination of the formative flow conditions of flutes and tool marks formed in deep-water environments, and demonstrates that flutes are not solely the product of turbulent flows, but can also be formed by transitional flows. The paper also shows that discontinuous tool markssuch as skim marks and prod marks with up-dip striaeare the product of more cohesive transitional flows than flutes. Although, since the pioneering work of Kuenen (1953), grooves and chevron marks have been assumed almost universally to have been formed by turbidity currents, herein it is proposed that they are formed by debris flows, as well as slumps and slides as Kuenen (1957) recognized. Chevron marks further indicate that there must be a fluidal layer at the base of the debris flow, as seen in quasi-laminar plug flows (sensu . The cross-cutting nature of some flutes and tool marks indicates that flows can also undergo transitions in flow type at a given spatial location. Flutes and tool marks are thus the product of a range of sediment gravity flow types, but in most cases they are not the product of low-density or high-density noncohesive turbidity currents, as envisaged in past literature. This fluid dynamic linkage is used herein to propose the first synoptic model that explains the observed longitudinal distribution of flute type, and different tool mark types, in terms of progressive changes in cohesion of flows down-dip, with flows transforming from turbulent non-cohesive flows, through transitional flows, to debris flows, or vice-versa. This model also provides a means for more detailed analysis of the relationships between sole marks and palaeohydraulic conditions in outcrop. The recognition that grooves and chevrons are dominantly the product of debris flows (and also slumps and slides) demonstrates that existing pictorial descriptions of the Bouma sequence incorrectly assume a genetic link between the basal erosive surface and the overlying deposit. This work introduces a new pictorial version of the Bouma sequence that incorporates this insight and illustrates that the erosive surface can represent significant sediment bypass. In addition, it is shown that the formation of flutes and tool marks is not restricted to the head of gravity currents. It is also evident that substrate characteristics are crucial for sole structures, yet remain poorly understood. This study shows that modern seafloor substrates exhibit a narrow (<1 m thick) zone of shallow strengtheningup to an order of magnitude stronger than predicted by consolidationin the top few decimetres to approximately 2 m. This variation in shear strength with depth may lead to rapid flow bulking if erosion breaks through this layer, and account for the bimodality in flow transformation.
Although knowledge of aggradational bedforms has increased through decades of flow measurement and experimentation, almost no work has been undertaken on flutes and tool marks since the pioneering work of the 1950s to the early 1970s, thus restricting their utility to palaeocurrent indicators. In the interim, our knowledge of the fluid dynamics of sediment gravity currents, and the nature of the shallow seafloor substrate, have advanced enormously. Herein, it is demonstrated that it is possible to use flutes and tool marks to interpret: (i) flow type at deposition; (ii) the nature of flow transformation; and (iii) the nature of the basal layer within debris flows where chevron marks are present. This new understanding suggests that it is then possible to predict the nature of deposit type down-dip. The present study demonstrates that there is much information to be gleaned from a greater understanding of these sole structures, and that there is much more to be learnt from refocusing on these under-utilized sedimentary structures.