The Origin and 3D Architecture of a Km-Scale Deep-Water Scour-Fill: Example From the Skoorsteenberg Fm, Karoo Basin, South Africa

Scours, and scour fields, are common features on the modern seafloor of deep-marine systems, particularly downstream of submarine channels, and in channel-lobe-transition-zones. High-resolution images of the seafloor have improved the documentation of the large scale, coalescence, and distribution of these scours in deep-marine systems. However, their scale and high aspect ratio mean they can be challenging to identify in outcrop. Here, we document a large-scale, composite erosion surface from the exhumed deep-marine stratigraphy of Unit 5 from the Permian Karoo Basin succession in South Africa, which is interpreted to be present at the end of a submarine channel. This study utilizes 24 sedimentary logs, 2 cored boreholes, and extensive palaeocurrent and thickness data across a 126 km2 study area. Sedimentary facies analysis, thickness variations and correlation panels allowed identification of a lower heterolithic-dominated part (up to 70 m thick) and an upper sandstone-dominated part (10–40 m thick) separated by an extensive erosion surface. The lower part comprises heterolithics with abundant current and sinusoidal ripples, which due to palaeocurrents, thickness trends and adjacent depositional environments is interpreted as the aggradational lobe complex fringes. The base of the upper part comprises 2-3 medium-bedded sandstone beds interpreted as precursor lobes cut by a 3–4 km wide, 1–2 km long, and up to 28 m deep, high aspect ratio (1:100) composite scour surface. The abrupt change from heterolithics to thick-bedded sandstones marks the establishment of a new sediment delivery system, which may have been triggered by an updip channel avulsion. The composite scour and subsequent sandstone fill support a change from erosion- and bypass-dominated flows to depositional flows, which might reflect increasingly sand-rich flows as a new sediment route matured. This study provides a unique outcrop example with 3D stratigraphic control of the record of a new sediment conduit, and development and fill of a large-scale composite scour surface at a channel mouth transition zone, providing a rare insight into how scours imaged on seafloor data can be filled and preserved in the rock record.

Stratigraphically, the presence of scour-fills can provide important insights into the evolution of deep-water system as whole, as they may mark a change in slope gradient, a temporal change in the nature of the flows, or changes in sediment supply. Changes in slope gradient and loss of confinement of a turbidity current can result in rapid flow transformation and enhanced basal shearing, which results in scouring via a process called "flow relaxation" (Komar, 1971;Normark, 1987, 1991;García and Parker, 1993;Vicente Bravo and Robles, 1995;Wynn et al., 2002;Ito, 2008;Hofstra et al., 2015;Brooks et al., 2018a;Pohl et al., 2019). The depositional or erosional nature of flows either leads to infilling of the scour or further erosion where sediments are largely bypassed and deposited further downdip (Hofstra et al., 2015;Brooks et al., 2018a). Therefore, improved identification of scourfills, and their stratigraphic evolution, can contribute to improved understanding of source-to-sink approaches.
The deep-marine stratigraphy of Unit 5 from the Permian Karoo Basin succession in South Africa, provides a unique outcrop where a large composite erosion surface can be mapped with three dimensional constraints. The presence of the erosion surface marks a significant and abrupt change from an up to 70 m thick lower package of heterolithics to a 40 m thick package of amalgamated sandstones. Unit 5 palaeogeography has been constrained by past studies Hofstra et al., 2017;Johnson et al., 2001;Wild et al., 2009Wild et al., , 2005, and with the 3D outcrop control and research borehole data the following objectives are addressed: 1) to investigate the depositional environment of the thick basal package of heterolithics; 2) to document and establish the origin of the 3D erosion surface; and 3) to propose a palaeogeographic evolution of Unit 5 in the Skoorsteenberg area.

GEOLOGICAL SETTING
The Karoo Basin is bounded by the southern and western branches of the Cape Fold Belt ( Figure 1A), and was traditionally interpreted as a retroarc foreland basin that developed from the early Permian (e.g., (De Wit and Ransome, 1992;Veevers et al., 1994;Visser and Praekelt, 1996;Visser, 1997;López-Gamundí and Rossello, 1998). However, recent models relate Permian subsidence to longwavelength dynamic topography driven by the subducting palaeo-Pacific plate (Tankard et al., 2009), and no emergent Cape Fold Belt at the time of deep-water deposition (Blewett and Phillips, 2016). The Tanqua and Laingsburg depocentres make up the SW part of the Karoo Basin ( Figure 1A), and are filled by the Late Carboniferous to Early Jurassic Karoo Supergroup (>5 km thick). Within the Tanqua depocentre, this succession comprises the glacial Dwyka Group, overlain by the post-glacial deep-marine to shallow-marine Ecca Group, and the non-marine Beaufort Group ( Figure 1B). The Ecca Group is an approximately 1.4 km thick shallowing upward succession from deep-marine to fluvial settings (King et al., 2009;Flint et al., 2011). The 0.4 km thick Skoorsteenberg Formation is part of the Ecca Group and comprises four submarine fans (Fans 1-4) and an overlying succession termed Unit 5 (Bouma and Wickens, 1994;Johnson et al., 2001;Hodgson et al., 2006;Wild et al., 2009), which is the focus of this study. Several field studies (Bouma and Wickens, 1991;Johnson et al., 2001;Hodgson et al., 2006;Prélat et al., 2009;Kane et al., 2017;Hansen et al., 2019) and 11 research boreholes Spychala et al., 2017a;Hofstra et al., 2017) constrain the stratigraphic framework of the Skoorsteenberg Formation.
Originally, the distal (northern) area of Unit 5, at Skoorsteenberg, was recognised as Fan 5, and interpreted as a slope fan, and the southern, most proximal area, at Groot Hangklip, was referred to as Fan 6 (Wickens 1994;Basu and Bouma, 2000;Wach et al., 2000;Johnson et al., 2001;van der Werff and Johnson, 2003). Regional mapping of an overlying 12 m thick mudstone that correlated these sand-prone units led to the redefinition of Unit 5 (Wild et al., 2009).
In proximal (southern) areas of Unit 5 at Kleine Hangklip, stacked W-E and SW-NE orientated submarine slope channel complexes have been interpreted (Wild et al., 2005;Bell et al., 2020) that overlie the updip pinchout of Fans three and 4 (Hansen et al., 2019). In distal (northern) areas of Unit 5 submarine fan deposits have been mapped southeast of the study area at Blaukop, where sand-rich channel-fills incise into proximal lobes (Hofstra et al., 2017). This study focuses on the northern exposures of Unit 5 at Skoorsteenberg that are characterised by a thick lower part (∼70 m) of thin-bedded Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 2 sandstones and siltstone, and an upper part (∼40 m) of thickbedded sandstones. Previous interpretations of these outcrops include "interfan deposits" overlain by a slump scar-fill towards the top (Johnson et al., 2001) and as an axial channel conduit (22 m thick, 8 km wide) that diverges downdip into three distributary channels (van der Werff and Johnson, 2003). Overall, published studies point towards Unit 5 being a deep-water slope apron fed by multiple W-E and SW-NE submarine channel-levees feeding lobes , with an overall younging direction of conduits along slope to the NW.

DATA AND METHODS
This study is based on 24 measured outcrop sections and 2 behind outcrop cores (NS1 and NS2) located to the east of the outcrop area ( Figure 2A). These sections were logged at 1:50 scale (∼1 km cumulative thickness), recording grain size, sedimentary structures and bounding surfaces. Two cores and three outcrop logs cover the whole thickness of Unit 5, which is defined by underlying and overlying regional mudstones (Wild et al., 2009). Fifteen outcrop logs focus on the sandstone-prone upper part of Unit 5 ( Figure 2A).
For this study, Unit 5 is subdivided into a lower and upper part ( Figure 2B) using a distinctive concretion marker bed, which was walked out in order to observe the spatial and temporal distribution of overlying sedimentary facies. Photo panels and photogrammetric models of the outcrop created from Uncrewed Aerial Vehicle (UAV) imagery (Stratigraphy Group, University of Leeds, 2021), using Agisoft Metashape and LIME, were used to document and interpret stratigraphic surfaces and architectural elements. Quantitative analysis of the thickness variations (using the inverse distance weighted (IDW) interpolation method in ESRI ArcGIS) of the whole of Unit 5, and the lower and upper parts, was undertaken to determine regional changes. Table 1 summarizes the sedimentary facies scheme (Figure 3), determined by their lithology, sedimentary structures, bed thickness, contacts and geometries.

MAP DATA
Unit 5 has been subdivided into two parts using a distinctive concretion marker bed (5-12 cm thick) that is resistant to weathering, at a consistent stratigraphic level and was walked out across the outcrop area. The lower part is dominated by thin-bedded heterolithics, and the upper part by medium-to thick-bedded sandstone ( Figure 4). We present palaeocurrent and thickness data based on these two parts. The concretion marker bed is not identified in the NS1 and NS2 cores, which means the thickness of the lower and upper parts is poorly constrained.

Palaeocurrent Analysis
Four hundred and two palaeocurrent measurements were collected from current and climbing ripple lamination, groove marks, and orientation of margins of incision surfaces. The palaeocurrent data ( Figures 5B,C) have a narrow spread from N to NE, which is consistent with the overall depositional dip direction for the Skoorsteenberg Fm. (e.g. Hodgson et al., 2006;Prélat et al., 2009;Hansen et al., 2019). The lower thin-bedded part is dominated by current and climbing ripple laminations trending towards the NE (average 084°, n 207) ( Figure 5B), with the upper part showing more dispersed trends to the N to NE (average 074°, n 195) ( Figure 5C). Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932

Thickness Analysis
The Unit 5 isopach map shows eastward thinning from 120 m in the Skoorsteenberg area to 70 m at NS1 ( Figure 5A). The lower thin-bedded part is bounded by the basal mudstone below Unit 5 and the concretion marked bed at the top (Figure 4), and thickens to the NW from 33 to 68 m thick ( Figure 5B). The facies above the concretion marker bed change to medium-to thick-bedded, coarser grained sandstones, which are incised by a widespread erosion surface that can be correlated between field logs for kilometres ( Figure 4). Overlying this erosion surface is filled by medium-to thick-bedded sandstones that thicken westward up to 40 m ( Figure 5C).

ARCHITECTURE OF UNIT 5
Lower Part: Thin-Bedded Heterolithics The thin-bedded lower succession overlies the basal mudstone that separates Fan 4 and Unit 5 and is characterised by siltstone-and sandstone-prone heterolithics (FA2, FA3), dominated by sinusoidal, climbing, and current ripples ( Figures 4A, 6). Sinusoidal lamination is a form of highly aggradational climbing-ripple cross-lamination (Jopling and Walker, 1968), which indicate persistent high rates of sediment deposition. This suggests that sediment gravity flows were expanding and depositing rapidly, either due to a change in gradient or an abrupt change in topographic confinement (Allen, 1973;Kneller, 1995;Jobe et al., 2012). Individual beds are normally graded, and coarsening-or fining-upwards packages (<5 m thick) are identified, but thicker grain-size or thickness trends are not present. A more sandstone-prone heterolithic unit, up to 12 m thick, is present towards the top of the succession (Figures 4, 6).

Upper Part: Medium-to Thick-Bedded Sandstone
The upper section of the Unit 5 stratigraphy is constrained using the concretion marker bed as a basal datum and the capping regional mudstone at the top of Unit 5 (Figures 4, 6). The Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 4 concretion marker bed (5-12 cm thick) is identified by a distinctive brown-orange colour, is more resistant to weathering, and contrasts to the light grey to pale yellow of the surrounding stratigraphy ( Figure 4B). The bed is laterally continuous for kms and was walked out between outcrop logs.

Concretion Marker Bed to Erosion Surface
The stratigraphy overlying the concretion marker bed consists of ∼5 m of siltstone-and sandstone-prone heterolithics (FA2 and FA3) above which two to three medium-to thick-bedded, sandstone beds are present (Figures 4, 6, 7B-D). These are truncated by an extensive erosion surface mantled by mudclasts ( Figure 7A), which can be correlated between the outcrop logs. Multiple smaller erosion surfaces merge onto the larger surface indicating its composite nature ( Figure 7D). To establish the shape and amount of erosion into the underlying stratigraphy, two measurement methods were employed ( Figure 8): 1) measuring the stratigraphic thickness between  (Sumner et al., 2012). Mudclast lags indicative of bypassing flows (Stevenson et al., 2015) Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 5 the concretion marker bed and the base of the erosion surface from the logs, which showed that net erosion is up to 28 m ( Figure 8A), and 2) mapping the erosion surface using photogrammetric models of the outcrop built from UAV imagery to provide 3D constraints on the shape ( Figures  8B-D), and the elevation change from inside to outside the cut to constrain erosion depth. The results of both methods showed that the area of maximum erosion is in the west of the study area, between logs SK03 and PK02 forming a deeper low aspect ratio heel of maximum erosion. The length of erosion is at least 1-2 km long in a downdip direction, and about 3-4 km wide in a strike section (Figure 8).

Erosion Surface to Top Unit 5
Above the laterally extensive erosion surface, and in the area of maximum erosion, the stratigraphy is characterised by a 40 m thick succession of amalgamated, structureless to parallel laminated, thick bedded sandstones (Figures 4, 6, 9A,B). In areas overlying less erosion, the succession is more stratified and characterised by ripple laminated medium-bedded sandstones (Figures 6, 9C).
The uppermost stratigraphy of Unit 5 comprises siltstoneprone, ripple laminated heterolithics, with rare sinusoidal laminations ( Figure 4). The heterolithics fine upwards to a 12-15 m thick, capping mudstone, indicating the termination of Unit 5.

NS1 and NS2
In both cores, the base of Unit 5 is defined by a several metres thick mudstone (Figure 10), with the top of the boreholes sited close to the top of Unit 5. In NS1, Unit 5 is ∼72 m thick, and consists of a basal ∼25 m thick heterolithic unit, overlain by a ∼20 m thick coarser grained, structureless to ripple laminated, medium to thick-bedded sandstone unit ( Figure 10). Another siltstone-prone heterolithic unit is overlain by a ∼15 m thick medium-to thick-bedded, structureless to ripple laminated sandstone package with mudclasts mantling erosion surfaces ( Figure 10). Unit 5 in NS2 is ∼91 m thick, with a lower ∼30 m siltstone-and sandstone-prone heterolithic package ( Figure 10). Above this a ∼25 m thick, very fine to finegrained, medium to thick-bedded sandstone package punctuates the succession, which is predominantly parallel and ripple laminated ( Figure 10). Small (<1 cm diameter) mudclasts at bed bases and truncation of beds mark erosion surfaces. Some Photo of log SK03 showing the concretion marker bed and the medium-bedded sandstone beds below the regional erosion surface. The sandstone fill of the large erosion surface can be seen as well as a small erosion surface towards the top of this fill.
Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 7 sandstones become argillaceous towards the bed tops, suggesting the presence of hybrid beds in this succession. Another heterolithic unit is overlain by a ∼20 m thick unit of mediumto thick-bedded structureless and climbing rippled sandstones ( Figure 10).
Fine-scale correlation of Unit 5 between the cores and outcrop logs is challenging in the absence of the concretion marker bed. Despite the 9 km distance between the cores, the two distinct sandstone packages may be correlated. However, their correlation with the Skoorsteenberg outcrops is uncertain. Nonetheless, the sedimentary facies observed in both cores, particularly the argillaceous sandstone beds interpreted as hybrid beds in NS2, suggest that these sandstones represent lobe complexes. Lobes have also been interpreted 7 km to the south of NS2 in the lower part of Unit 5 at Blaukop and core BK1 (Hofstra et al., 2017). These associations support the lower part of Unit 5 in the cores being lobe complexes, with more evidence for erosion in the upper sandstone package, although the facies support an interpretation of more lobe axes in a lobe complex. The thin-bedded heterolithics between share affinities (bed thickness, ripple laminated sandstones) to a similar

Depositional Environment of Lower Heterolithics-Prone Part
The heterolithic succession in the lower part of Unit 5 (70 m thick) has an abundance of sinusoidal, climbing and current ripples but no major coarsening-or fining-upwards trends. Thick accumulations of thinbedded heterolithics in deep-water settings either occur in external levees adjacent to submarine channels, as internal levees and terrace deposits within large-scale erosion surfaces, or at lateral or distal lobe fringes and basin plain settings (Walker, 1975;Normark and Piper, 1991;Skene et al., 2002;Deptuck et al., 2003;Kane et al., 2007;Kane and Hodgson, 2011;Hansen et al., 2015;Spychala et al., 2017b). Sinusoidal ripples have previously been described in the Karoo Basin, in external and internal levees and aggradational lobe fringe deposits in the Fort Brown Formation in the Laingsburg depocentre (Kane and Hodgson, 2011;Morris et al., 2014a;Spychala et al., 2017b). Previous work has constrained the palaeogeographic context of the study area where there is a downdip architectural change from submarine channel complexes 25 km south of the study area (e.g. Wild et al., 2005;Bell et al., 2020) to lobe-dominated deposits mapped southwest of Skoorsteenberg (Hofstra et al., 2017).
Thick accumulations of heterolithics, or thin-bedded turbidites, in external levee successions have been observed from many outcrops, modern seafloor studies and in the subsurface (Clemenceau et al., 2000;Kane et al., 2007;  Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 9 Babonneau et al., 2010;Kane and Hodgson, 2011;Maier et al., 2012Maier et al., , 2013Paull et al., 2013;Morris et al., 2014a;Hansen et al., 2015). Typical characteristics include underlying frontal lobes, thinning away from an adjacent submarine channel, and an overall finingand thinning upwards trend attributed to levee growth and increased flow confinement allowing only the upper, fine-grained parts of turbidity currents to overspill and deposit sediments (Buffington, 1952;Manley et al., 1997;Skene et al., 2002;Deptuck et al., 2003;Kane et al., 2007;Kane and Hodgson, 2011;Nakajima and Kneller, 2013;Hansen et al., 2015). In the study area, the heterolithics do not show fining-upwards trends, and whilst thinning and palaeocurrent trends can be seen towards the NE (Figure 5), no contemporaneous submarine channel system is identified to account for flow stripping and overspilling of turbidity currents. Furthermore, the heterolithic package directly overlies the capping mudstone of the underlying Fan 4 system with no thicker sandstone beds that could be interpreted as frontal lobe present, thus making an external levee origin unlikely. An internal levee or terrace deposit interpretation is not supported due to the absence of a confining erosion surface, and the consistent palaeocurrent directions.
Lobe fringes are also composed of packages of heterolithics but require certain conditions to accumulate packages of up to 70 m thick. Aggradational lobe fringes documented from the Laingsburg depocentre in the Karoo Basin, were pinned in one location by the presence of intrabasinal slopes (Spychala et al., 2017b). Aggradational onlaps form in weakly confined basins where the bounding slope angles are less than 1° (Smith, 2004;Smith and Joseph, 2004;Spychala et al., 2017b). The effects of subtle topography on sedimentary facies and depositional architectures in deep-water settings has been widely documented (Hansen et al., 2019;Pyles, 2008;Smith, 2004;Spychala et al., 2017b). The sedimentary structures in the lower part of Unit 5 indicate that the very fine-grained sandstones, sandy siltstones and siltstones with climbing and sinusoidal ripples were rapidly deposited from density stratified turbidity currents with high rates of suspended sediment load fallout. The lack of hybrid event beds within this succession supports these heterolithics being deposited in lateral rather than Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 frontal lobe fringe settings (Hansen et al., 2019;Spychala et al., 2017a). However, the thickness of the heterolithic package is greatest in the west (Figures 5, 6), whereas if a SE-facing intrabasinal slope was present to pin the lobe fringe setting, a thinning trend would be predicted. The underlying upper Fan 4 deposits are also thickest in the Skoorsteenberg area (Spychala et al., 2017a), which initially might have formed a subtle high after deposition of the mudstone between Fan 4 and Unit 5. However, that Fan 4 and Unit 5 are thickest in the same location suggests increased subsidence rates may have affected this area during sedimentation allowing a greater thickness of thin beds to accumulate. Palaeocurrents towards the N and NE ( Figure 5) indicate that turbidity currents were largely sourced from the south with the NE trend indicating that they were likely following the main downslope gradient at the time of deposition.

Origin of the Erosion Surface
The prominent large-scale erosion surface that widens and shallows downdip above the heterolithic succession is within the upper sandstone-prone part of Unit 5. In deep-water settings, large scale, high aspect ratio erosional surfaces of this geometry are likely scours that vary in dimensions from 10s of metres to multiple kilometres in width and length and cm to 10s of metres in depth (Ito et al., 2014;Hofstra et al., 2015). An alternative interpretation of a high aspect ratio channels would be subparallel sided and not shallow downdip so prominently. Large scour surfaces can form in the headwall areas of slide scars (e.g. Pickering and Hilton, 1998;Lee et al., 2004;Moscardelli et al., 2006;Dakin et al., 2013). Alternatively, scours are commonly concentrated in channel-lobe-transition-zones (CLTZs) (Hofstra et al., 2015;Brooks et al., 2018a), in channel mouth settings (Carvajal et al., 2017;Droz et al., 2020;Maier et al., 2020;Pohl et al., 2020Pohl et al., , 2019, or have a multi-event origin. Large-scale erosion surfaces formed by submarine landslides are associated with downdip Mass Transport Deposits (MTDs), and have been documented in slope settings in several subsurface examples (Moscardelli et al., 2006), modern seafloor datasets (Gamberi et al., 2011;Macdonald et al., 2011), and in some outcrop examples (Pickering and Hilton, 1998;Dakin et al., 2013;Brooks et al., 2018b). The erosion surface within Unit 5 has previously been interpreted as a slump scar (Johnson et al., 2001). In the translational domain, slump scar surfaces are the basal shear surface and are overlain by debrites or slumped sediments related to the initial sediment failure. In Unit 5, the erosion surface is infilled by turbidites, which if a slump origin is advocated points to the surface being in the proximal evacuation zone. The scale of the erosion surface described here would imply a large volume mass failure, and the absence of any slumped sediment or debrite above the erosion surface or downdip makes a slump scar origin unlikely.
High-resolution bathymetric data from modern deep-water systems have revealed extensive scouring in channel mouth settings, where the confining channel surface widens and shallows (Carvajal et al., 2017;Droz et al., 2020;Maier et al., 2020), rather than forming a discrete CLTZ between well-defined channels and well-defined lobes. Scouring of channel margins is shown to be extensive especially in areas with higher slope gradients (>1°) (Carvajal et al., 2017). Coalescence of scour surfaces is likely a major driver for channel avulsion, inception and propagation resulting in further turbidity current confinement (Droz et al., 2020). In the La Jolla channel, these scours form laterally extensive erosion surfaces that can extend for kilometres downdip of the channel mouth (Maier et al., 2020). The scale and subtle relief of the scours reported from channel mouths is similar to the erosion surface seen in Unit 5. However, these scours have been shown to occur adjacent to, or within, channels. Although submarine channel complexes have been reported from updip areas, there is no evidence for a channel at this stratigraphic level in Unit 5 around Skoorsteenberg. If it is a channel mouth setting, then the channel did not propagate further into the basin. Scouring is commonly reported from CLTZs where turbidity currents loose confinement resulting in rapid flow deformation and enhanced basal shearing of the turbidity current (Komar, 1971;Normark, 1987, 1991;García and Parker, 1993;Vicente Bravo and Robles, 1995;Wynn et al., 2002;Ito, 2008;Hofstra et al., 2015;Brooks et al., 2018a) via a process referred to as "flow relaxation" (Pohl et al., 2019). Interpreted exhumed CLTZs are characterized by scour-fills, and thin and discontinuous structureless and structured sandstones dominated by ripple and climbing ripple lamination (García and Parker, 1993;Hofstra et al., 2015;Brooks et al., 2018a) that might be the remnants of sediment waves (Hofstra et al., 2018). Scours in CLTZs have been shown to vary in depth and dimensions, and outcrop studies from the Karoo Basin suggest that they can form individual small-scale scours, or large-scale composite scours interpreted to represent prolonged periods of weakly confined sediment bypass (Hofstra et al., 2015;Brooks et al., 2018a). The 3D exposure of the erosion surface within Unit 5 indicates a 3-4 km wide, 1-2 km long, and up to 28 m deep surface. The scale of the surface is large compared to other outcrop studies, and suggests that this is a composite scour surface in a channel mouth transition zone that originated from bypassing flows that deposited sediment further downdip, with the main scour-fill infilled by subsequent flows.

Stratigraphic Evolution of Unit 5 at Skoorsteenberg
The stratigraphic evolution of Unit 5 at Skoorsteenberg (Figure 11) is based on our preferred interpretation of the  Figure 6), with location indicated in (A). An erosion surface present higher up the stratigraphy is also highlighted.
Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 depositional environment of the lower heterolithic part and the origin of the erosion surface. The basal heterolithics are interpreted as the aggradational fringes of multiple stacked lobe complexes identified towards the E and SE (Hofstra et al., 2017), with lobe complexes also interpreted in cores NS1 and NS2. The aggradational lobe complex fringes are interpreted to have formed in an area that underwent preferential subsidence during sedimentation as the isopach thicks of Unit 5 and upper Fan 4 coincide, rather than representing the infill of pre-existing topography ( Figure 11A). Two to three ∼0.5 m thick fine-grained sandstone beds are present above the package of heterolithics (Figures 4, 6, 7) across the entire outcrop areas unless cut out by the overlying erosion surface. Palaeocurrent trends are similar to the heterolithics package, i.e., towards the N and NE ( Figures  5B,C). These sandstone beds appear abruptly without any coarsening-and thickening-upwards signature observed in the underlying heterolithics (Figures 4, 6). Hence, the abrupt appearance of these coarser and thicker sandstone beds below a thicker coarse-grained sandstone package mark the initiation of increased sediment supply to the area. A simple basinward progradation of the system would appear as a more gradual change, especially in distal settings of the basin described here. Similar deposits have been identified in the ancient deep-marine basin-floor successions of the Windermere Supergroup in Canada, where they have been interpreted as avulsion splays (Terlaky et al., 2016). However, these avulsion splays contain an abundance of fine-grained matrix and mudstone clasts, likely due to being the first flows that breach the levee updip and thus entraining mud-prone substrate (Terlaky et al., 2016). Mud-clast rich sandstone beds interpreted as crevasse splays (or "crevasse lobes") were also observed in cores taken as part of IODP leg 155 in the Gulf of Mexico . Similar fine-grained sandstones with abundant sinusoidal laminae and climbing ripples that have a mounded geometry were interpreted as frontal splays (or frontal lobes) in the Fort Brown Formation in the Karoo Basin (Morris et al., 2014b). The sandstone beds with some climbing-ripple and parallel lamination observed here are clean. This character and their abrupt appearance suggests that these sandstones either are 1) frontal lobes recording the establishment of a new slope conduit, or 2) avulsion splays where redirection of flows from an existing conduit eroded a sand-rich substrate. Establishment of a new slope conduit would follow the overall pattern in Unit 5 of submarine channels and lobes moving NW over time ( Figure 11B). Slope submarine channel avulsions occur via a range of mechanisms (Jobe et al., 2020), including levee collapse (Brunt et al., 2013;Ortiz-Karpf et al., 2015), overspill and flow-stripping (Piper and Normark, 1983;Fildani et al., 2006), and/or channel aggradation (Kolla, 2007;Armitage et al., 2012), and drivers such as climate cyclicity (Picot et al., 2019) In more distal settings, an autogenic mechanism invoked is a downstream gradient decrease during lobe aggradation to a point where the channel will start to aggrade forcing it to migrate and/or avulse to find a new higher gradient downstream pathway (e.g., Groenenberg et al., 2010;Prélat et al., 2010) (Figure 11B).
Above these medium-bedded sandstones, the erosion surface incised up to 28 m into the substrate ( Figure 11C) and was likely sculpted and widened by successive bypassing flows. The size of the erosion surface is similar in size to composite scour surfaces reported from modern seafloor datasets (Carvajal et al., 2017;Droz et al., 2020;Maier et al., 2020) and comparable to the largest reported from exhumed settings (Hofstra et al., 2015).
The subsequent filling of the erosion surface indicates that the flows transitioned from dominantly erosional and bypassing to dominantly depositional. It is not possible to resolve whether this is due to internal or external factors, or a combination of factors controlling the nature of the flows. Internal factors may include the flows becoming more sand prone and less efficient over time (Al Ja'Aidi et al., 2004;Heerema et al., 2020), as the feeder conduit matured, or that the new downstream pathway gradient decreased due to upstream erosion and downstream deposition resulting in aggradation . External factors may include a transient period of decreased flow magnitude due to changes in sediment supply, for example caused by eustatic and climatic fluctuations. The sandstones that fill the erosion surface are thick-bedded, amalgamated, structureless to parallel laminated with no evidence for hybrid-bed prone facies. Furthermore, the lack of fine-grained heterolithics or bed tops suggest that the flows may have been stripped and finer grain-sizes deposited downdip, making these lobes more similar in character to intraslope lobes than basin-floor lobes (Spychala et al., 2015;Brooks et al., 2018c). Small-scale erosion surfaces towards the top of the sandstone-prone part of the succession are interpreted as either distributary channels where lags are present, or scour-fills, and are linked to a final phase of basinward progradation of the system ( Figure 11E). FIGURE 10 | Correlation between cores NS1 and NS2 with interpreted sedimentary facies indicated. The regional extent of the two sandstone packages is unknown.

CONCLUSION
This study describes a unique outcrop in Unit 5 of the Karoo Basin, South Africa, where a large (2 long × 4 wide km) and high aspect ratio (28 m deep) erosion surface can be mapped with three dimensional constraints. The scale of the high aspect ratio erosion surface, and the geometry that widens and shallows downdip supports interpretation of a composite scour surface. The scour surface marks a significant and abrupt change from a lower package of heterolithics to an upper package of amalgamated sandstones, which indicates a change in sediment supply to the area, reflecting either establishment of a new slope conduit, or an updip avulsion event. The underlying thick package of heterolithics is interpreted as aggradationally stacked lobe complex fringes that were deposited in an area of increased subsidence. Below the large scour surface multiple thin to mediumbedded sandstone beds are present, which are interpreted as frontal lobes before large, bypass dominated flows cut the substrate to form the scour surface. The upper sandstone-prone package is interpreted as lobe deposits that infill the scour surface and show a change from erosional and bypassing flows to depositional flows. Whilst large-scale scours are commonly observed on modern seafloor data, their preservation in outcrop is rare and provides a unique Frontiers in Earth Science | www.frontiersin.org September 2021 | Volume 9 | Article 737932 14 opportunity into how the presence of scours and scour-fills can provide important insights into the source-to-sink configuration of deep-water systems.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
DH, RH, LH, and AP coordinated the work. The main data collection was done by RH with the help of LG and DL. All authors discussed the results. LH wrote the manuscript, with support from DH, RH, LG, DL, AP, and RW.

FUNDING
This research was funded by Equinor ASA.