A series of large-scale erosional scours are described from four modern deep-water canyon and/or channel systems along the northeast Atlantic continental margin. Regional-scale geophysical data indicate that most scours occur in zones of rapid flow expansion, such as canyon and/or channel termini and margins. High-resolution images of the scours cover ∼25 km2 at 2 × 2 m pixel size, and were obtained at depths of 4200–4900 m using Autosub6000, an autonomous underwater vehicle equipped with an EM2000 multibeam bathymetry system. Sedimentological and microfossil-based chronological data of scour fills and interscour areas were obtained via accurately located piston cores that targeted specific sites within imaged areas. These core data reveal a number of key findings. (1) Deep-water scours can be very long lived (>0.2 m.y.) and may undergo discrete phases of isolation, amalgamation, and infilling. (2) Deep-water scours can develop via a composite of cutting and filling events with periodicities of between tens of thousands and hundreds of thousands of years. (3) Immediately adjacent scours may have strikingly different sedimentological histories and do not necessarily evolve contemporaneously. (4) Scour infills are typically out of phase with sedimentation in intrascour areas, having thin sands internally and thick sands externally, or thick muds internally and thin muds externally. (5) Erosional hiatuses within scour fills may represent hundreds of thousands of years of time, and yet leave little visible record. Four distinct morphologies of scour are identified that range from 40 to 3170 m wide and 8 to 48 m deep: spoon shaped, heel shaped, crescent shaped, and oval shaped. Isolated scours are shown to coalesce laterally into broad regions of amalgamated scour that may be several kilometers across. The combined morphosedimentological data set is used to examine some of the putative formative mechanisms for scour genesis.


Large-scale erosional scours are a key component of many deep-water channels and fans, and have been intensively studied in both modern (e.g., Normark, 1970; Normark et al., 1979, 2009; Shor et al., 1990; Kenyon et al., 1995; Wynn et al., 2002a; Fildani et al., 2006) and ancient submarine environments (e.g., Vicente Bravo and Robles, 1995; Elliott, 2000a, 2000b; Lien et al., 2003; Macdonald et al., 2011). Despite these studies, little is known about the genesis of deep-water erosional scours; direct observations of flow over scours are lacking, and there is very little information on sedimentation within scours. Consequently, the evolution and longevity of deep-sea scours are largely unknown. Similarly, the morphological variation within deep-sea scours has not been examined in detail, mostly due to a lack of sufficient image resolution.

Here we address these limitations through morphosedimentological studies of a number of key deep-water scour sites along the northeast Atlantic continental margin. The detailed morphology of scours in these areas has been measured at unprecedented depths (4200–4900 m) using high-resolution multibeam bathymetry collected using a deep-diving autonomous underwater vehicle (AUV). Accurately located cores were taken within and adjacent to scours, and coccolith stratigraphy was utilized to provide age control, enabling comparative studies of intrascour and interscour deposition to be undertaken for the first time. Sedimentological data demonstrate that scours may show prodigious longevity (>0.2 m.y.) and that scour evolution in the deep sea may be complex and multiphase. To put these new observations into a broader context, we synthesize existing studies and develop a morphological classification of scour types.


Scour Morphology and Environment

Research on deep-water erosional scours began in the 1970s, pioneered by Bill Normark and colleagues during their studies of the Navy Fan, offshore California (e.g., Normark et al., 1979). These studies recognized giant flute-shaped depressions more than 500 m wide and 15 m deep on the modern seafloor. Around this time, much smaller examples, as deep as 1.5 m, were documented in Late Cretaceous submarine fan deposits of the Cerro Toro Formation (Magallanes Basin), southern Chile (Winn and Dott, 1979). Even at this embryonic stage of scour research, it was clear that there are major differences in the scale of features that can be recognized in modern and ancient settings.

Subsequent to these early findings, increasingly detailed deep-towed sidescan sonar and multibeam bathymetry images of the modern seafloor revealed that large-scale erosional scours are common features of many deep-water turbidite systems (for a listing of key examples, see Table 1). These studies have shown that large-scale scours are broadly restricted to margins or termini of submarine canyons and channels. Examples of scours imaged on channel levee backslopes, or between and/or along active channels, include the Navy Fan (Normark et al., 1979), Laurentian Fan (Shor et al., 1990), Stromboli Canyon (Kidd et al., 1998), Monterey Fan (Masson et al., 1995; Klaucke et al., 2004) and Redondo Fan (Normark et al., 2009). Scours in these examples are as much as 20 m deep and 400 m long. In some areas, stepped scours adjacent to channel bends are arranged in linear trains that extend for several kilometers, e.g., Monterey Canyon (Fildani and Normark, 2004; Fildani et al., 2006) and Eel Canyon (Lamb et al., 2008). Scours have also been encountered at the transition zone between canyons or channels and depositional fans and basins. Examples include the Valencia Channel mouth (Palanques et al., 1995; Morris et al., 1998), Rhône Neofan (Kenyon et al., 1995; Torres et al., 1997; Wynn et al., 2002a; Bonnel et al., 2005), Umnak Channel mouth (Kenyon and Millington, 1995), and Setúbal and Agadir Canyon mouths (Wynn et al., 2002a). These occur either as fields of isolated scours (250–1000 m wide, 500–1000 m long, 14–20 m deep) or as broad zones of amalgamated scour as wide as 3 km. In areas where slope canyons encounter a marked slope break at the base of slope, submarine plunge pools as much as 1.1 km wide and 75 m deep can occur, as along the U.S. continental slope (e.g., Lee et al., 2002).

Erosional scours documented in outcrop commonly have meter-scale geometries, such as in the Carboniferous Ross Sandstone of western Ireland (Chapin et al., 1994; Elliott, 2000a, 2000b; Lien et al., 2003), the Albian Black Flysch of northern Spain (Vicente Bravo and Robles, 1995), the Early Eocene Charo Canyon mouth of the Spanish Pyrenees (Millington and Clark, 1995), and the Eocene–Oligocene Annot Sandstone in the French Alps (Hilton, 1995; Morris and Normark, 2000). These outcrop-based studies helped document the nature of scoured substrates and infilling sediments, e.g., the multiepisodic scour and fill geometries visible in sandstones of the Albian Black Flysch (Vicente Bravo and Robles, 1995). However, precise paleoenvironmental interpretation of scours in outcrop can be challenging, as exemplified by observations of the well-studied Carboniferous Ross Sandstone. In this system, scours have been interpreted as occurring in a channel-lobe transition zone (Chapin et al., 1994; Macdonald et al., 2011), on channel flanks via single high-magnitude channel-initiating flows (Elliott, 2000a, 2000b), and in spillover lobes at the bends of sinuous channels (Lien et al., 2003).

Flow Processes

Little is known about the underlying processes that form deep-water scours, or the longevity of these features. Most large-scale scours form in regions indicative of flow expansion (Mutti and Normark, 1987; Normark and Piper, 1991) induced by a lack of confinement or a change in slope. At these locations, turbidity currents are thought to frequently undergo a hydraulic jump, leading to locally increased turbulence and scouring of underlying sediments (Komar, 1971; Normark et al., 1979; Mutti and Normark, 1987; Lee et al., 2002). A linear scour array on the Monterey Fan has been interpreted in terms of repeated hydraulic jumps, so-called cyclic steps (Fildani et al., 2006), and this mechanism is thought to play a more generic role in deep-sea scour formation (e.g., Kostic and Parker, 2006). Information on scour longevity is limited, although Normark et al. (2009) estimated minimum ages for two scours in Redondo Canyon; one was directly dated (through an intrascour core) as being 950 yr old, and a second (based upon extrapolation of interscour accumulation rates to geophysical imaging of scour fill) was estimated as older than 4500 yr. However, these times largely indicate a period of stasis and mud infill, rather than active scour development.

Given the limited process data on large deep-water scours, some comparison has been made with much better studied small-scale erosional features such as flutes and tool marks. The flow processes of these small-scale erosional features have been recreated under laboratory conditions (Rücklin, 1938; Dzulynski and Walton, 1963; Dzulynski, 1965; Dzulynski and Simpson, 1966; Allen, 1969a, 1971, 1984) and studied in detail in outcrop (Eggenhuisen et al., 2010, 2011); however, these are typically several orders of magnitude smaller than scours documented on the modern seafloor. Allen (1971) linked the generation of small-scale erosional bedforms (i.e., flutes) to the presence of irregularities on the eroded bed, and Allen's (1971) defect theory related the size, shape, and structure of erosional bedforms (i.e., flutes) to the number and character of these irregularities, and the nature of the erosive flow. To test whether Allen's (1971) defect theory could be upscaled to large-scale modern scours, Shor et al. (1990) applied it to generation of a giant flute-shaped scour on the Laurentian Fan (100 m deep, 1000 m long), which was believed to have been generated by the 1929 Grand Banks turbidity current. Shor et al. (1990) performed calculations using both empirical parameters (from Allen, 1971) and measurements taken from the giant scour. However, both calculations produced durations of scour generation that were considered too short to realistically generate the observed scour; it was therefore assumed that the scour became unusually deep because erosion of seafloor surrounding the scour was hindered by a conglomerate veneer. It is critical that more recent work suggests that the scour may have existed prior to the 1929 event (Piper et al., 2007), and therefore it may have been the product of more than one flow. Allen's (1971) theory assumed that flute formation was the product of single flows (in fact, just the head portion of these flows), and thus may not be transferable if scours develop under successive flows. It therefore remains unclear whether flow processes operating on both small-scale and large-scale scours are comparable, and whether experimental studies of erosion can be upscaled to natural examples on the modern seafloor.


Our understanding of deep-water scours has traditionally been restricted by a lack of high-quality bathymetric data; most previous studies relied on low- to medium-resolution sidescan sonar or hull-mounted multibeam bathymetry. However, recent technological advances in deep-water research are now enabling significant improvements in resolution. For example, Normark et al. (2009) utilized an AUV to obtain unprecedented images of deep-water scours on the Redondo Fan, offshore California. They obtained high-resolution multibeam bathymetry and subbottom profiles at water depths of as much as 700 m by flying the AUV at a height of ∼70 m above seafloor. Their multibeam bathymetry data have 1.5 m lateral resolution and 0.3 m vertical accuracy, which is approaching outcrop-scale resolution (Normark et al., 2009). This pioneering study was a first step toward more detailed morphometric measurement of deep-sea scours, but still at relatively shallow depths. Here we utilized Autosub6000 (Huvenne et al., 2009) in order to image 4 examples of large-scale deep-water scours at water depths >4200 m. Autosub6000 is a newly developed AUV capable of operating close to the seafloor at water depths as great as 6000 m (Huvenne et al., 2009). The AUV was instrumented with a high-resolution multibeam bathymetry system; the resulting high-resolution images are combined with accurately targeted piston cores in order to provide information on erosion history and scour fill.


This study investigates erosional scours in four deep-water canyon and/or channel systems along the northeast Atlantic continental margin (Fig. 1). These are, from south to north, (1) Agadir Canyon mouth, (2) Horseshoe Valley, (3) Setúbal Canyon mouth, and (4) Whittard Channel margin. These locations display a variety of shallow- to deep-water sediment transport regimes, due to their different climatic settings (semiarid in the south to temperate and glacially influenced in the north) and varying proximity to the tectonically active Africa-Eurasia plate boundary (Weaver et al., 2000). Horseshoe Valley and Setúbal canyons exhibit active faulting related to the Azores-Gibralter Fracture Zone, and the headwaters of Agadir Canyon are affected by the seismically active High Atlas Mountains (Sébrier et al., 2006; Arzola et al., 2008; Duarte et al., 2010). In contrast, the Whittard Channel is in a passive margin setting. These differences link to significant variations in flow type; Agadir Canyon is dominated by infrequent, landslide-derived large-volume flows (>10 km3 of sediment, 1/10 k.y.) sourced from the Morocco Shelf (Wynn et al., 2002b), whereas the Whittard Channel is dominated by frequent (as many as 130/k.y.) smaller flows, mostly during glacial lowstands when fluvioglacial outwash supplied sediments directly to the head of Whittard Canyon (Toucanne et al., 2008). While these broad differences are known, the exact nature of flow types and magnitudes are poorly constrained for several of the examples (cf. North American margin; e.g., Piper and Normark, 2009).


Geophysical Data

The majority of geophysical data presented here were collected during a research expedition on RRS James Cook in August 2008 (JC27). The main data set presented here is high-resolution multibeam bathymetry, collected using an EM2000 system housed within the Autosub6000 AUV. Autosub6000 was able to cover an area of ∼25 km2 within a 24 h mission, and the EM2000 system was able to image seafloor features with a pixel size of 2 × 2 m. Data were subsequently processed using the IFREMER (Institut Français de Recherche pour l'exploitation de la Mer) software suite Caraibes (http://www.ifremer.fr/fleet/equipements_sc/logiciels_embarques/caraibes/index.html). Additional data collected using hull-mounted multibeam bathymetry (EM120) provide information on the overall planform of the seafloor, although these data are generally of insufficient resolution to image the studied scours.

Sedimentological Data

A series of shallow piston cores was collected from each of the four work areas, with a maximum penetration of 6.5 m. Core sites were chosen once high-resolution multibeam bathymetry images were downloaded and visualized, and were selected to hit targets >50 m across (50 m is our estimated coring accuracy at water depths >4 km). RRS James Cook is equipped with a dynamic positioning system, while any potential offset created by drift of the corer was monitored using an ultra-short baseline acoustic positioning system located on the coring wire at depths as great as 2.5 km.

Visual core logging included sediment facies, color, and grain size. All logged deposits were interpreted for their mechanism of deposition (identified as turbidite, debrite, or hemipelagite) and used to infer a depositional subenvironment. Note that in this study we identify turbidite sands as being normally graded, well sorted, and deposited in an aggradational layer by layer fashion (with associated laminations and cross-laminations). Smooth green or gray-brown turbidite muds can be separated from pale brown hemipelagic muds due to the latter containing randomly dispersed foraminifera. Debrite deposits are identified as ungraded, poorly sorted, and clast rich, often containing evidence for mass flow, e.g., contorted remobilized sequences.

Dating Control

Microfossil-based dating of hemipelagic sediments in the studied cores was used to identify erosional hiatuses. Ratios of different coccolith species were identified and a combination of first and last appearance and overall abundance of dominant species was then used to develop a chronostratigraphy, tied into the oxygen isotope stratigraphy at specific oxygen isotope stages (OIS) (Weaver and Kuijpers, 1983; Weaver, 1994; Wynn et al., 2002b). Bioturbation of hemipelagic sediments and other potential errors means that ages are accurate to within ∼10%.


Agadir Canyon Mouth

Agadir Canyon extends northwestward 450 km from the Morocco Shelf (100–200 m water depth) to the eastern Agadir Basin (∼4500 m water depth) (Figs. 1 and 2A). The canyon is as wide as 30 km and acts as a conduit for large-volume siliciclastic flows (Wynn et al., 2002b; Frenz et al., 2009). Previous studies, utilizing medium-resolution (30 kHz) sidescan sonar, mapped a major zone of erosion in the canyon mouth with kilometer-scale scours focused immediately downstream of an intracanyon slope break (slope change of 0.2° to 0.04°; Wynn et al., 2002a). Both isolated and amalgamated scours were imaged in the erosion zone (see inset in Fig. 2A), but the relationship between the two end members was not clearly resolvable on existing data.

Our new high-resolution imagery (Fig. 2B) confirms the presence of both isolated and amalgamated scours, covering an area of ∼15 km2 in the canyon mouth. The deepest erosion is focused within isolated scours along the northern margin of the broad, flat canyon axis; these scours cut into the gently sloping northern margin of the canyon floor. Isolated scours are spoon shaped and elongated downslope, with U-shaped cross-sectional profiles that shallow and taper downstream. Maximum scour depths and sidewall slope angles are consistently located within the upstream 60% of the scours; the steepest slope angles (20°–50°) are largely confined to scour headwalls and sidewalls (see profiles A-A′ and B-B′; Fig. 2B). Scour dimensions are 150–600 m long, 40–225 m wide, and 8–20 m deep. Scours 1 and 2 (Fig. 2B) are the largest identified isolated scours. Scour 2 exhibits a rim opening and low sidewall slopes (0.5°–6.5°) along its southwest margin, where it borders a region of amalgamated scour. This amalgamated scour displays a broadly flat-bottomed morphology, but includes several erosional remnants within the scour floor and in cuspate rims at scour margins (see cross-section C-C′; Fig. 2B). The imaged area of amalgamated scour extends across >4 km2 and can be subdivided into smaller zones of amalgamation that are bound by high-standing topography. The headwall of the amalgamated scour comprises a series of cuspate scars, similar in apparent dimensions and morphology to the headwalls of adjacent isolated scours.

Three piston cores, JC27–09, JC27–12, and JC27–11, were obtained from within the imaged region along a south-southwest–north-northeast transect (Figs. 2B and 3). Core JC27–09 targeted the floor of the amalgamated scour, and recovered ∼4.0 m of sediment including a total of 13 turbidites. The youngest turbidites comprise thin (<5 cm), normally graded, very fine grained and well-sorted basal sands, with thin planar and cross-laminations (Tc and Td sequences; Bouma, 1962). Overlying mud caps are as thick as 65 cm, and intervening hemipelagites are present between each turbidite (dated as OIS 1–3; younger than 60 ka). In contrast, older deposits (below 1.8 m core depth and dated as OIS 4–5; 60–130 ka) display 5–10-cm-thick, normally graded, medium-grained, planar and cross-laminated basal sands (Tc and Td), with erosive bases and rip-up clasts; these sands are overlain by comparable thicknesses of turbidite mud. A thin clast-rich muddy debrite overlies an apparent erosional hiatus at 3.4 m that likely occurred around the OIS 5-6 boundary ca. 130 ka (as indicated by coccolith-based dating; Fig. 3). Immediately beneath this hiatus are the oldest hemipelagic sediments sampled in the core, which contain Pseudoemiliania lacunosa and are therefore older than OIS 13, or ca. 450 ka (Weaver and Kuijpers, 1983; Weaver, 1994). The hiatus therefore corresponds to the erosion of 320 k.y. of sedimentation and probably represents several meters of missing sediment (based upon regional hemipelagic sedimentation rates of 1–2 cm/k.y.; Wynn et al., 2002b; Frenz et al., 2009).

Core JC27–12 targeted the floor of the deepest isolated scour, and recovered 1.0 m of sediment including 5 turbidites (Fig. 3). These turbidites typically display thin (<5 cm), fine- to medium-grained, planar and cross-laminated basal sands (Tb and Tc), with 5–15-cm-thick overlying mud caps. Coccolith dating of intervening hemipelagites reveals that the upper three turbidites were deposited during OIS 1–3 (younger than 60 ka). These turbidites are immediately underlain by a hemipelagite dated as OIS 7 (190–245 ka), indicating an erosional hiatus of ∼130 k.y. (Fig. 3). Two very large volume megaturbidites (>100 km3 of sediment) passed through Agadir Canyon between 60 and 130 ka (Wynn et al., 2002b; Frenz et al., 2009), and likely contributed to development of this hiatus. A further hiatus of >200 k.y. occurs above the lowermost turbidite in the core, as hemipelagite overlying this deposit contains P. lacunosa and is therefore older than OIS 13 (older than 450 ka).

Core JC27–11 sampled ∼3.0 m of sediment from an area of relatively smooth seafloor immediately northeast of the scoured area, and contains 6 texturally immature, sand-rich, poorly sorted turbidites as thick as 50 cm (Fig. 3). The thicker turbidite deposits exhibit erosive bases and contain wavy, planar, and ripple cross-laminations that are occasionally disturbed by convolute bedding. Coccolith dating reveals an absence of significant hiatuses in JC27–11 (Fig. 3), instead showing a relatively young and continuous sequence of sand-rich turbidites (all OIS 1–3; younger than 60 ka).

Visual analysis of the three cores provides compelling evidence for a correlative relationship between the upper three turbidites, based on turbidite mud color, relative stratigraphic position, and thickness and color of intervening hemipelagic intervals. Coccolith dating supports these correlations and indicates that the upper three turbidites were deposited in the past 60 k.y. (Fig. 3). However, the turbidite deposits show remarkable variation between cores, in terms of bed thickness, grain size, and sedimentary structures (Fig. 3). Cores JC27–09 and JC27–12 also show evidence for significant hiatuses beneath these turbidites, indicating phases of active erosion prior to 60 ka and removal of several meters of sediment.

Horseshoe Valley

The Horseshoe Valley is located offshore southwest Iberia, and is a broad conduit for sediments transported southwestward from the Lagos and Portimao Canyons to the Horseshoe Abyssal Plain (Fig. 4A) (Terrinha et al., 2009). Hull-mounted multibeam bathymetry data reveal a series of giant scours on the floor of the central fairway, on an overall slope of ∼0.5° (Terrinha et al., 2009; Duarte et al., 2010; Fig. 4A). The largest scours are as much as 5 km wide and 120 m deep, with long axes aligned parallel to slope. Seismic profiles of Duarte et al. (2010) suggest that scour locations are controlled by the underlying thrust fault morphology.

Our new high-resolution AUV data focus on a single large-scale erosional scour that is U shaped in cross section and is ∼3 km wide and 50 m deep (crescentic depression 1 of Duarte et al., 2010). The scour is oval in planform and, unlike the Agadir Canyon scours, is elongate along slope (i.e., perpendicular to downslope flow). The scour is at ∼4600 m water depth and displays average headwall slope angles of 30°, with maximum angles locally reaching 56° (Fig. 4B). Profiles of the headwall slope vary across the scour, ranging from smooth and constant to stepwise with quasi-terraced morphology (profiles A-A′ and C-C′; Fig. 4B). Two areas of morphologically distinct bedforms flank the scour: V-shaped chevrons to the west and lineations to the southeast and east (see insets, Fig. 4B). The chevrons are V-shaped positive relief features that are as much as 200 m across. Chevron limbs bound a hollow and flat-bottomed central region, and open out in a downstream direction. The lineations are negative relief features that may be fully isolated or amalgamated with other surrounding lineations; they are 40–80 m wide, 250–460 m long, and as deep as 3 m. All lineations >80 m in length appear to be amalgamated.

Two piston cores were recovered from the area imaged by high-resolution data (Figs. 4B and 5). Core JC27–24, recovered from outside the scour ∼700 m upslope of the scour headwall, contains 4.7 m of dominantly hemipelagic sediments interbedded with ∼20 thin turbidites. Turbidite deposits are 0.2–12 cm thick, and comprise thinly laminated fine sand bases overlain by structureless muds. In some cases the basal sand is absent. Coccolith ratios reveal that these deposits range in age from OIS 1–3 (younger than 60 ka) to OIS 8–12 (450 ka), and do not appear to be separated by significant hiatuses (Fig. 5). In contrast, core JC27–25*3, recovered from the scour floor, is dominated by seven thick turbidites (50–180 cm or more) with characteristics very different from those outside the scour. The turbidites display texturally mature, normally graded, fine- to medium-grained sand bases that are erosive, as much as 25 cm thick, and laminated or cross-laminated (Tb and Tc). Sand bases are overlain by as much as 1.5 m of structureless ungraded turbidite mud, and bounding hemipelagites are thin or absent (Fig. 5). Coccolith dating shows that the turbidite sequence cored inside the scour is relatively young, with ages restricted to OIS 1–4 (younger than 75 ka). Net accumulation rates during the past 75 k.y. are therefore more than 3 times higher inside the scour compared to the adjacent seafloor.

Setúbal Canyon Mouth

Setúbal Canyon is one of the largest canyons crossing the west Iberian margin, extending seaward from the continental shelf near Lisbon to the Tagus Abyssal Plain at 4840 m water depth (Fig. 6A) (Lastras et al., 2009). Erosional features have previously been documented in the lower canyon and canyon mouth, using medium-resolution (30 kHz) sidescan sonar (see inset, Fig. 6A). New high-resolution images within the canyon mouth reveal irregular crescentic scours that are elongate perpendicular to flow, with widely flaring limbs that point downstream (Fig. 6B). These scours are as much as 1.0 km in length and width and display width:length ratios of ∼1; amalgamated forms typically exhibit width:length ratios >>1. One large isolated scour reaches a maximum depth of 14 m and has a steep headwall with slope angles as high as 30°; it shallows downstream over a distance of several hundred meters and has a steep step that accounts for approximately half of the shallowing of the scour (profile A-A′; Figs. 6B, 6C). Two other scours are partly amalgamated, with headwall slope angles as high as 35° and a maximum depth of 22 m (profiles B-B′ and C-C′; Figs. 6B, 6C).

Core JC27–39 sampled sediments from within the largest isolated scour (Figs. 6B and 7) and contains 50 cm of hemipelagite underlain by 3.35 m of chaotic facies, herein interpreted as debrite. The debrite comprises lithic clasts, mud clasts (of varying stiffness), sand lenses, lithic rich sands, and banded sands, which are inferred to represent remobilized lower canyon sediments. These canyon sediments include (1) thin-bedded turbidites, (2) canyon floor gravels (containing lithic clasts as large as 7.5 cm), and (3) coarse sands with rip-up clasts as much as 6.0 cm across. It seems likely that this debrite is the same as that identified by Arzola et al. (2008), that covers much of the lower canyon mouth area. The overlying uninterrupted hemipelagite suggests that no flows have passed through this location since the deposition of the debrite, which further indicates a shutdown of the system in the last few thousand years.

Whittard Channel Margin

The Whittard Canyon and Celtic Fan link the southern Irish Sea and English Channel paleoriver systems to the deep northwestern Bay of Biscay (Figs. 1 and 8A) (Droz et al., 1999; Zaragosi et al., 2000). Hull-mounted multibeam bathymetry data from the fan surface reveal the course of the main Whittard Channel, locally flanked by levees draped with fine-grained sediment waves (Fig. 8A). High-resolution AUV images across the western margin of the distal Whittard Channel reveal three distinct morphological features (Fig. 8B). These include (1) a portion of the active Whittard Channel with a smooth flat thalweg, (2) a heavily scoured channel margin, and (3) a pair of large-scale sediment waves in the overbank area of the active channel.

Four stages of erosional scour can be recognized in the overbank area: protoscour, isolated scour, early-stage amalgamated scour, and fully amalgamated scour. All of the scours have developed on lee slopes or in troughs of sediment waves. Protoscours are zones of shallow erosion that are as much as 100 m long and 40 m wide; they are shallow and flat-floored, with internal slope angles <10° and a maximum vertical relief of 8 m. Isolated scours are wider than they are long (at their maximum limits), and are as much as 890 m wide, 550 m long, and 30 m deep. They have a uniform heel shape that does not exhibit any obvious signs of coalescing from smaller features; internal slope angles are generally low (<10°), with steeper slopes (22°–50°) confined to the outer limits of the scour (profiles A-A′ and B-B′; Figs. 8B, 8C). One early-stage amalgamated scour has a distinctive scalloped headwall rim and low-relief interior hummocks that are characteristic of scour remnants following amalgamation (profile C-C′; Figs. 8B, 8C). This scour is as deep as 18 m, and is 750 m wide and 480 m long; internal slope angles range from 23° to 56°. The largest erosional feature within the imaged area is a late-stage amalgamated scour that extends for >2500 m in the across-slope direction and >1300 m downslope. Of all the scour types documented in this area, this region of amalgamated scour displays both the deepest level of scour and the steepest slopes (50 m and >62°, respectively). The scoured zone has an irregular outer rim and, unlike the other scours, a highly irregular floor (profile C-C′; Figs. 8B, 8C).

Two cores were collected from one of the scours and the adjacent updip sediment wave. Core JC27–63 sampled 2.3 m of sediment from a broad region of amalgamated scour in a wave trough (Figs. 8B and 9), and is dominated by a single 1.7-m-thick deposit. This deposit comprises a thin medium-grained basal sand, overlain by a thick ungraded structureless mud. The sand and mud layers are separated by a grain size break (Fig. 9). Other deposits sampled at the base of the core were interbedded organic-rich turbidites with erosive bases and thin mud caps. Core JC27–62 was recovered from a smooth wave crest located 1695 m to the north (just off the area covered by AUV bathymetry; Fig. 8). This core is dominated by thin-bedded turbidites composed of fine sand bases and thin mud caps; these levee-like deposits show an overall upward fining and thinning between 0.9 and 3.7 m (Fig. 9). The upper 0.9 m of the core comprises several thicker turbidites (as much as 7 cm thick), which are overlain by 53 cm of hemipelagite. There is no evidence for the thick mud deposit visible in core JC27–63. Insufficient hemipelagic sediments are present for coccolith dating, but comparison with other cores in the region suggests a Late Glacial turbidite succession overlain by Holocene hemipelagite (OIS 1, or younger than 15 ka).


Scour Morphologies and Sizes

The high-resolution images presented here provide detailed insight into the dimensions, morphology, and infill of erosional scours in a variety of deep-water environments. Based upon these morphologic data, we are able to identify the following categories of scour (Fig. 10).

Isolated Erosional Scours

Isolated erosional scours have a smooth and continuous outer rim with a regular internal morphology and a broadly symmetrical U-shaped across-slope profile. They are relatively flat bottomed, with steeper slopes of 20°–50° confined to scour margins. Their downslope profile is asymmetric, with a steep headwall and more gradual downslope opening. The examples of isolated scours presented here reveal four distinctly different types of scour shape and size.

Spoon-shaped scours display a regular elliptical shape in planform, and are elongated in the downslope direction (e.g., scours 1 and 2 in Fig. 2B). Spoon-shaped scours are the only type of scour that narrows; it is important that they close in the downslope direction (Fig. 10). Their elliptical planform produces a low width:length ratio of ∼0.4. Other examples of spoon-shaped scours include the Cerro Toro Formation of northern Chile (Winn and Dott, 1979; Jobe et al., 2009), Albian Black Flysch of northern Spain (Vicente Bravo and Robles, 1995), Ross Formation of Ireland (Elliott, 2000a, 2000b; Lien et al., 2003), and the modern Valencia Channel mouth in the western Mediterranean Sea (Palanques et al., 1995; Morris et al., 1998).

Heel-shaped scours have outward-flaring limbs that originate at a central upslope location (Fig. 8B). The downslope termination of the scour develops via gradual shallowing across the scour width; scour limbs continue to flare out until this termination (Fig. 10). Heel-shaped scours are wider than they are long, resulting in width:length ratios of as much as 1.6. Other published examples of scours exhibiting a heel-shaped morphology occur on the Rhone Neofan off southern France (Kenyon et al., 1995; Torres et al., 1997; Wynn et al., 2002a; Bonnel et al., 2005) and on the Redondo Fan offshore California (Normark et al., 2009).

Crescentic scours have broadly lunate shapes with two downslope-pointing limbs (Fig. 6B). The downslope profile varies across the width of the scour, with more rapid downslope shallowing in the center of the scour compared to the limbs (Fig. 10). Unlike heel-shaped scours, the area between the terminations of the two limbs is positive relief. They are as wide as, or wider than, they are long, with a resultant width:length ratio of ∼1.3. Crescentic scours have previously been described from the canyon-basin transition zone off west Portugal (Wynn et al., 2002a) and the Valencia Channel mouth (Palanques et al., 1995; Morris et al., 1998).

Oval scours have an elliptical planform that is elongated in the across-slope direction (Fig. 10). The large oval scour imaged in this study (Fig. 4B) displays a more irregular rim than spoon-shaped scours; this may be due to a significantly steeper headwall resulting in small-scale retrogressive mass wasting. Oval scours can be very large; the example documented in the Horseshoe Valley is the widest and deepest isolated scour in the study area. Oval scours have also been documented in Eel Canyon, offshore California, where they were described as quasi-circular topographic depressions (Lamb et al., 2008).

The original controls on morphology of isolated scours remain poorly understood, especially as some examples show lateral variations in scour morphology within the same system, e.g., the Valencia Channel mouth (Palanques et al., 1995; Morris et al., 1998). In addition, as seen in this study, scours in comparable environments can look very different, e.g., the Agadir and Setúbal Canyon mouths (Figs. 2B and 6B). We speculate that a complex interplay of substrate character (e.g., sand/mud ratio, consolidation rate), seafloor morphology (e.g., slope angle, degree of channelization), flow character (e.g., volume, velocity, density), and flow frequency are important factors contributing to scour morphology and dimensions.

Amalgamated Erosional Scours

Regions of amalgamated scour develop via lateral coalescing of isolated scours; consequently, the overall size of amalgamated scours exceeds that of the isolated scours that form them. The morphology of amalgamated scours is defined both by the character and number of isolated scours that have been amalgamated. The upslope portions of amalgamated scour rims are cuspate (Figs. 2B and 8B), where each cusp is a relic of a former isolated feature that has since been incorporated into the amalgamated region. Erosional remnants of former isolated scour margins are commonly preserved on the floor of amalgamated scours, and take the form of irregular topography, hummocks, or elongate ridges of positive relief (Figs. 2B and 8C).

Of the types of isolated scours characterized here, spoon-shaped, heel-shaped, and crescentic scours all develop into broad regions of amalgamated scour (Figs. 2B, 6B, and 8B). In each case, the nature of the amalgamated region becomes highly irregular, although some key characteristics remain that allow the character of former isolated scours to be identified. In the case of spoon-shaped scours, the outer margins and inner remnant topography of the amalgamated region are aligned in the downslope direction, while the irregular upslope rim comprises a number of narrow, steep, and tightly rounded cusps (Fig. 2B). In comparison, the rims of heel-shaped or crescentic amalgamated scours comprise gently rounded cusps and maintain their widely flaring character (Figs. 6B and 8B). Amalgamated regions that grow via the coalescing of crescentic scours continue to shallow downslope more rapidly toward the center of the scour than at the margins, therefore retaining the overall crescentic shape.

It is notable that the oval isolated scour in the Horseshoe Valley is ∼3 km wide (Fig. 4B), wider than any region of amalgamated scour documented in this study. It is a fully isolated scour, with no evidence for amalgamation and no comparable isolated scours visible on the adjacent seafloor (Fig. 4B). This has been interpreted to result from structural control (Terrinha et al., 2009; Duarte et al., 2010). However, it is also possible that amalgamation is partly controlled by spacing of isolated scours, whereby this example has developed to a scale rarely achieved by isolated examples because it is located many kilometers away from adjacent scours. Overall, it appears that the point at which amalgamation occurs is controlled by the spacing, rate of lateral expansion, and longevity of original isolated scours.

Sedimentary Deposits within and Adjacent to Scours

Sedimentary deposits of the studied scours reveal some remarkable results. First, they demonstrate the longevity of some deep-sea scours, with those in Agadir being active over a period of at least 0.2 m.y. (since the oldest turbidites preserved in the scours are 190–245 ka) and possibly older than 450 ka (the age of the oldest hemipelagite). As a result of erosion, it is not possible to further determine whether the scour has been continuously active since before 450 ka, or whether it only initiated ca. 0.2 Ma and eroded down through >200 k.y. of sediment (see Fig. 3). This longevity is in sharp contrast to scours in river systems, which are typically active over periods of months to years (e.g., Lunt et al., 2004; Sambrook Smith et al., 2005; Hooke and Yorke, 2010). Our results therefore overturn the associated paradigm that states that scours are largely transient features. Scour deposits also indicate that erosion and deposition within some scours may be cyclical, with a periodicity of tens of thousands to hundreds of thousands of years (Fig. 3). This repeated cut and fill has fluctuations in the vertical on the order of several meters.

In all cases, the sedimentary infill of the scours is out of phase with sedimentation outside of the scours. In the Agadir example, sedimentation inside the scours over the past 60 k.y. is represented by a series of muddy turbidites with thin (<5 cm) sandy bases; in contrast, just outside the scoured area is a sequence of sand-rich turbidites as thick as 50 cm (Fig. 3). In the Horseshoe Valley example, the opposite occurs; recent sedimentation (younger than 75 ka) inside the scours is represented by very thick muddy turbidites with thin (<15 cm) sandy bases, whereas outside the scour there is little or no deposition (<12 cm of mud; Fig. 5). The character of deposits can also vary greatly between scour fills and associated interscour areas. For example, scours may be filled with thick (0.7–1.7 m) ungraded structureless muds (e.g., Agadir Canyon, Horseshoe Valley, and Whittard Channel), but can also contain thick debrites (e.g., Setúbal Canyon).

In addition to the marked differences between intra and interscour sedimentation style and associated thicknesses, there is evidence for large-scale variations in the infill of adjacent scours. In Agadir, the deepest isolated scour displays two significant hiatuses, the youngest of which occurs between OIS 3 and 7, representing at least 130 k.y., while for much of this period, the adjacent amalgamated scour shows deposition of a conformable sequence of mud-dominated turbidites (Fig. 3).

Insights into Scour Genesis and Evolution

High-resolution multibeam bathymetry data integrated with cores provide new insights into scour genesis and evolution. A key observation is that isolated scours merge laterally through time into larger areas of amalgamated scour (Figs. 2, 6, and 8), with the morphology of the original isolated scours often preserved as a series of scour rims on the headwall of the amalgamated scour. This may be a key process in maintaining depositional relief in these settings, as shown by Whittard and Agadir Canyons, where amalgamation of isolated scours is seen on the edge and center of the channel, respectively. Microfossil-based dating suggests that development of isolated scours, lateral amalgamation, and eventual infilling in the Agadir system may take tens of thousands to hundreds of thousands of years, probably due to the low event frequency, and in particular the low frequency of large-volume events (e.g., one major flow every 10 k.y. in Agadir; Wynn et al., 2002b).

While it is clear that scours grow and amalgamate with time, the genesis of the initial isolated scours is more difficult to discern. Cores that contain conformable sequences of scour fill (e.g., Horseshoe Valley) clearly do not penetrate the basal scour surface. In areas where a potential basal surface is recovered (e.g., the lower hiatus in the isolated scour core from Agadir Canyon), a lack of seismic profiles of sufficient resolution means that the possibility of deeper (uncored) hiatuses cannot be excluded. Nevertheless, the recognition of repeated cut-and-fill cycles over tens of thousands to hundreds of thousands of years within the scours, along with the observed facies variations, enables an assessment to be made of flow processes responsible for these large-scale erosional periods. Of key importance is the synchronous existence of infilling amalgamated scours and actively eroding isolated scours adjacent to one another in Agadir Canyon mouth during the late Quaternary (60–130 ka; Figs. 2 and 3). There are several potential explanations for this observation: (1) lateral migration in the primary zone of erosion; (2) hydraulic jump formation; and (3) spatial variations in bed shear stress related to flow-topography interaction. Each process is considered in the following.

Migration of the Primary Zone of Erosion

Well-defined isolated scours are present on the higher gradient flanks to the north of the main channelized zone (Fig. 2). This may suggest that the initiation of erosion has been more recent on the northern flank, with insufficient time for scour to amalgamate and for the channel floor transverse gradients to be reduced, and that consequently the primary zone of erosion has migrated laterally. The higher gradients on the channel flank may also act to focus erosion in this position. However, given that the zone of flow expansion in Agadir is so broad, it is difficult to see how small-scale variations in the primary zone of erosion can lead to such marked changes in scour infill between adjacent scours.

Hydraulic Jumps

Deep-sea scours have long been linked to the formation of hydraulic jumps (Mutti and Normark, 1987; Normark and Piper, 1991), given their positions in zones of flow expansion and with rapid changes in basal slopes. Numerical and physical modeling of hydraulic jumps have reproduced scour formation (Kostic and Parker, 2006; Postma et al., 2009), and in certain cases scour trains can form (cyclic steps), each associated with a hydraulic jump (Taki and Parker, 2005; Kostic and Parker, 2006). The cyclic step concept has subsequently been utilized to interpret scour trains in deep-sea environments (Fildani et al., 2006). These two models of hydraulic jumps are very different. In the former, hydraulic jumps are not triggered by small-scale basal topography, but by a marked break in slope and associated flow spreading, while in the case of cyclic steps, hydraulic jumps are in phase with the topography, and flow depths are thought to be of the same magnitude as the scour depth (Fildani et al., 2006; Duarte et al., 2010).

Hydraulic jumps may be responsible for the initiation of scours in the present examples and have been suggested for the Horseshoe Valley scours (Duarte et al., 2010); however, constraints on process are very limited. For the case of Agadir Canyon, some evidence is available to examine the potential for hydraulic jumps. The erosional surfaces in these scours may be related to the two large-volume turbidity currents that occurred ca. 60 ka and between 60 and 130 ka (Wynn et al., 2002b; Frenz et al., 2009). These flows had sediment volumes of 110 km3 and 230 km3, respectively (Frenz et al., 2009), and therefore likely had flow depths on the order of many tens of meters, possibly >100 m. Scours depths range between 8 and 20 m, and therefore it is unclear how this relatively small scale bed topography would trigger individual hydraulic jumps. If individually triggered hydraulic jumps did occur, then they might preferentially form in deeper scours with steeper headwalls while not in shallower, lower gradient scours. In the absence of such individual topographically triggered hydraulic jumps, it is unclear how a broader hydraulic jump zone would lead to erosion in one scour and concurrent sedimentation in another, as observed in the mouth of Agadir Canyon. Perhaps some of the better understood flow-topography relationships discussed in the following may also be applicable to hydraulic jumps.

Spatial Variations in Bed Shear Stress Related to Flow-Topography Interaction

Erosional features such as backward-facing steps and flutes are known to feature flow separation and flow reattachment within the scour, with the position of the reattachment point and the associated zone of high bed shear stress related to flow variables such as velocity, step height, and the slope of the headwall (Allen, 1968, 1969b, 1984). Given this, local variations in scour topography may lead to differences in erosion and deposition. In the case of Agadir Canyon, the isolated scour marked by erosion (60–130 ka) has a steep high headwall while the amalgamated scour has more subdued topography (Fig. 2). This might lead to differences in the basal shear stress exerted on the floor of the scours (see Allen, 1968), possibly accounting for the observed differences.

Scour Genesis: Summary

The origin of the scours cannot be specifically determined; however, the sedimentary fill of the scours provides evidence for periodic large-scale scouring in some cases. The marked differences in sedimentary infill between adjacent scours in the mouth of Agadir Canyon are difficult to interpret in terms of spatial variations in turbidity currents. Instead, they suggest that different scour types may interact with the overlying flows in contrasting ways, leading to juxtaposition of erosion and deposition in adjacent scours. Given the likely flow depths in Agadir, it is not clear that bed topography is sufficient to trigger localized hydraulic jumps over specific scours. Consequently, it may be the case that flow-topographic interactions may account for the observed local variations in scour erosion and deposition. Note that other factors such as underlying structure, e.g., sediment waves or thrust faults, may also locally influence location of individual scours (e.g., oval scour in Horseshoe Valley; Terrinha et al., 2009; Duarte et al., 2010).

Scour Fill

When currents are insufficient to cause net erosion of scours, then the scour fill appears to be a function of flow type. When flows are carrying significant amounts of coarse-grained sediments, they appear to have sufficient energy to largely keep the scours free of sediment (thin sandy turbidites), while interscour areas may exhibit large-scale deposition and the development of thick sands (e.g., Agadir Canyon; Fig. 3). Mechanistically, scours likely exhibit local flow acceleration and turbulence enhancement. In contrast, when flows are carrying large volumes of mud but little sand, then scours are preferentially filled compared with interscour areas (e.g., Horseshoe Valley; Fig. 5). We suggest that the thick muds observed in some scour fills are not the products of the tails of dilute turbidity currents, but are instead the products of higher concentration mud-rich basal flow components (Talling et al., 2004; Amy et al., 2005; Baas et al., 2009). This is further supported by the presence of associated thin clean sands at the base of the thick muds, giving a bipartite distribution characteristic of many hybrid flows (e.g., Talling et al., 2004).

Scour Abandonment

Scour abandonment and infilling may occur in response to a general system shutdown, e.g., during sea-level highstand, represented by Holocene hemipelagic drape in Setúbal Canyon mouth and Whittard Channel scours (Figs. 7 and 9). Scours may also be partially infilled by debrites, possibly from canyon and/or channel margin failures, e.g., the debrite fill in Setúbal Canyon mouth scour (Fig. 7). Muddy scour fills may therefore be generated by both allocyclic factors, e.g., changing and/or reducing sediment supply, and autocyclic factors, e.g., canyon thalweg migration or canyon margin failure.

Morphologic Features Associated with Scours

The V-shaped chevrons imaged alongside a giant oval scour in Horseshoe Valley (Fig. 4B) morphologically resemble erosional chevrons described from the seafloor beyond the Setúbal Canyon mouth, offshore west Iberia (Wynn et al., 2002a). However, the limbs of chevrons imaged in Horseshoe Valley are positive relief features, indicating that they are depositional in origin. They are therefore comparable to depositional chevrons, also as much as 200 m across, reported from beyond the mouth of Valencia Channel, where they are thought to be composed of coarse sand-sized sediments moving over a muddy substrate (Palanques et al., 1995; Morris et al., 1998).

Erosional lineations were also imaged adjacent to the oval scour in Horseshoe Valley (Fig. 4B). These features closely resemble longitudinal streaks identified by Wynn et al. (2002a) and Morris et al. (1998) from modern canyon and/or channel mouth environments. However, those documented here are significantly smaller and more closely spaced. Isolated erosional lineations are as much as 80 m long, while lineations that exceed 80 m in length are coalesced with adjacent features.


This work demonstrates that deep-water erosional scours can be dynamic long-lived features with multiple cut-and-fill stages. In addition, they may exhibit marked variability, both between adjacent scours and between intrascour and interscour areas. Consequently, scours add much spatial variability into deposits, with implications for existing simple architecture models of submarine channels and canyons (Mayall et al., 2006; Wynn et al., 2007). Furthermore, the recognition of multiple cut-and-fill stages in large (km wide) scours will make them very difficult to recognize in outcrop or core, where they are likely to be recognized as the primary conduits, rather than as components of much larger scour surfaces.

The complexity observed herein also indicates that simple approaches such as estimating scour longevity based on measurements of interchannel sedimentation and geophysically derived depths (e.g., Normark et al., 2009) may not always be appropriate. This same complexity also limits application of processes determined from smaller, better-studied erosional features such as flutes. It is clear that, unlike flutes, both erosion and infill may take place over multiple flows, and that these features, despite occasional morphological similarities, are not simply upscaled versions of flutes. In this respect the widely used term megaflute is potentially misleading. Flutes are cut and filled by single flows, while scours are features that persist on the seafloor during multiple flows.


In this study the detailed morphologies of deep-water scours have been combined with sedimentological and chronological data to provide new insights into scour morphology, sedimentology, and genesis, as follows.

1. Deep-water scours can be very long-lived features (>0.2 m.y.). This overturns the paradigm that scours are merely transient features.

2. Scours can show multiple cut-and-fill cycles with periodicities of tens of thousands to hundreds of thousands of years.

3. The infill of scours is shown to be typically out of phase with interchannel deposits, with thin sands internally and thick sands externally, or thick muds internally and thin muds externally.

4. Adjacent scours can have markedly different sedimentary infills.

5. Different scour types may interact with overlying flows in contrasting ways, leading to this juxtaposition of erosion and deposition in adjacent scours.

6. Isolated scours documented in this study are associated with canyon and/or channel termini and margins, and display four different morphologies: spoon shaped, heel shaped, crescent shaped, and oval shaped.

7. Isolated scours may coalesce into broad areas of amalgamated scour; evidence for the presence of isolated scours is often preserved within the region of amalgamation as a series of scour rims on the scour headwall, or as remnants on the scour floor.

Taken together, these points demonstrate both the range of different types of scour and the associated high spatial variability in sedimentary deposits within these depositional settings.

We thank the Captain, officers, and crew of RRS James Cook for their assistance during data collection on JC027. We particularly thank Pete Stevenson and Maarten Furlong for their assistance with autonomous underwater vehicle mapping. We also thank Esther Sumner for providing graphic logs of JC027 cores, reviewers Joris Eggenhuisen and Francisco Javier Hernández-Molina, and editors David Piper and Dennis Harry, whose comments helped to strengthen the manuscript. This project was funded as part of the Natural Environment Research Council (NERC) Oceans 2025 Programme. Heather Macdonald was funded by a NERC CASE (Council for Advancement and Support of Education) Studentship (NER/S/A/2006/14147) at the University of Leeds. The CASE sponsor was the Marine Geoscience Research Group at the National Oceanography Centre (NOC), Southampton.