Stratigraphic hierarchy and three‐dimensional evolution of an exhumed submarine slope channel system

Submarine slope channel systems have complicated three‐dimensional geometries and facies distributions, which are challenging to resolve using subsurface data. Outcrop analogues can provide sub‐seismic‐scale detail, although most exhumed systems only afford two‐dimensional constraints on the depositional architecture. A rare example of an accessible fine‐grained slope channel complex set situated in a tectonically quiescent basin that offers seismic‐scale, down‐dip and across‐strike exposures is the Klein Hangklip area, Tanqua‐Karoo Basin, South Africa. This study investigates the three‐dimensional architecture of this channel complex set to characterise the stratigraphic evolution of a submarine channel‐fill and the implications this has for both sediment transport to the deep‐oceans and reservoir quality distribution. Correlated sedimentary logs and mapping of key surfaces across a 3 km2 area reveal that: (i) the oldest channel elements in channel complexes infill relatively deep channel cuts and have low aspect‐ratios. Later channel elements are bound by comparatively flat erosion surfaces and have high aspect‐ratios; (ii) facies changes across depositional strike are consistent and predictable; conversely, facies change in successive down depositional dip positions indicating longitudinal variability in depositional processes; (iii) stratigraphic architecture is consistent and predictable at seismic‐scale both down‐dip and across‐strike in three‐dimensions; (iv) channel‐base‐deposits exhibit spatial heterogeneity on one to hundreds of metres length‐scales, which can inhibit accurate recognition and interpretations drawn from one‐dimensional or limited two‐dimensional datasets; and (v) channel‐base‐deposit character is linked to sediment bypass magnitude and longevity, which suggests that time‐partitioning is biased towards conduit excavation and maintenance rather than the fill‐phase. The data provide insights into the stratigraphic evolution and architecture of slope channel‐fills on fine‐grained continental margins and can be utilised to improve predictions derived from lower resolution and one‐dimensional well data.

Channel-cuts and their fills are commonly time transgressive (e.g. McHargue et al., 2011;Sylvester et al., 2011;Hubbard et al., 2014;Hodgson et al., 2016), formed by numerous energetic flows that excavated the channel. The preserved expression of flows that bypassed sediment further down-dip is composite erosion surfaces and associated heterogeneous channelbase-deposits. In the current study the term channel-base-deposit is used instead of the commonly used channel-base-drape (e.g. Barton et al., 2010) because a drape infers low-energy or background depositional processes, whereas channel base facies commonly indicate repeated cycles of erosion, entrainment and deposition by high-energy flows (e.g. Mutti & Normark, 1987). The nature of channel-base-deposits is commonly used to infer the characteristics of their parent flows, and can be useful in predicting the presence or absence of sandstone down-dip (Walker, 1975;Mutti & Normark, 1987;Barton et al., 2010;Hubbard et al., 2014;Stevenson et al., 2015;Li et al., 2016).
This study documents an exhumed Permian slope channel complex set (Unit 5, Skoorsteenberg Formation) that crops out in the Tanqua depocentre, Karoo Basin, South Africa ( Fig. 1A and B). A series of depositional strike, depositional dip, and oblique oriented cliff-faces permit documentation of the lateral, longitudinal, and vertical architecture of channel-fills in a mud-rich, fine-grained system (Fig. 1C). The objectives of this study are: (i) to elucidate the stratigraphic evolution of the channel complex set; (ii) to investigate the down-dip and acrossstrike architectural and facies variability within the channel complex set; (iii) to document the facies and distribution of channel-base-deposits; and (iv) to discuss the implications for reservoir connectivity and interpretation of subsurface data.

GEOLOGICAL SETTING
The Karoo Basin developed during the Permian due to subsidence induced by mantle flow processes associated with subduction of the Palaeo-Pacific plate, before transitioning to a retroarc foreland basin related to an adjacent fold and thrust belt, from approximately 250 Ma (Cape Fold Belt;De Wit & Ransome, 1992;Veevers et al., 1994;Visser & Praekelt, 1996;Viglietti et al., 2017). The deposits of the Karoo Supergroup, comprising the glacial Dwyka Group, the marine Ecca Group and the non-marine Beaufort Group, record basin deepening followed by shallowing through the late Carboniferous to the Triassic ( Fig. 1D; Smith, 1990;Bouma & Wickens, 1991;Johnson et al., 1996;Hodgson et al., 2006). This study concerns the Ecca Group, which records a progradational, shallowing-upward marine succession ( Fig. 1D; Smith, 1990;Bouma & Wickens, 1994;Hodgson et al., 2006).
The Ecca Group in the Tanqua depocentre, located in the south-west of the Karoo Basin ( Fig. 1), is subdivided into the Lower and Upper Ecca Group: The Lower Ecca Group consists of the relatively sand-starved basin-floor Prince Albert and Whitehill formations (Visser, 1994;Boulesteix et al., 2019); the Upper Ecca Group includes the sand-starved basin-floor Tierberg Formation (Visser, 1994;Boulesteix et al., 2019), the basin-floor to base of slope Skoorsteenberg Formation and slope to shallow-marine   P a n e l 1 P a n e l 2 P a n e l 3 P a n e l 4 P a n e l 5 Waterford Formation ( Fig. 1D; Bouma & Wickens, 1991;Bouma & Wickens, 1994;Johnson et al., 2001;Wild et al., 2005;Hodgson et al., 2006;Wild et al., 2009;Poyatos-Mor e et al., 2016).
The Skoorsteenberg Formation is 450 m thick in the Tanqua Depocentre and comprises five sandstone-prone units. The lower four units are interpreted as a progradational succession of submarine fans (Fans 1 to 4), overlain by a fifth unit (Unit 5) interpreted as a base of slope to lower-slope system (Bouma & Wickens, 1991, 1994Johnson et al., 2001;Wild et al., 2005;Hodgson et al., 2006;Pr elat et al., 2009;Hansen et al., 2019). Each fan is interpreted as a lowstand systems tract and is overlain by regionally correlated fine-grained packages interpreted as the combined transgressive and highstand systems tracts (Goldhammer et al., 2000;Johnson et al., 2001;Flint et al., 2011).
This study focusses on Unit 5 in the Klein Hangklip study area located in the south of the Tanqua depocentre ( Fig. 1B and C). Here, the preserved stratigraphy is 55 m thick, although Unit 5 is ca 100 m thick regionally. Unit 5 at Klein Hangklip is interpreted to represent a submarine slope environment consisting of intraslope lobes, channel-fills and lateral channel splay deposits (Wild et al., 2005). The channelfills have an internal hierarchy and are interpreted as a series of channel complexes (Wild et al., 2005). A regional fine-grained unit separates Unit 5 from the underlying Fan 4, which is subdivided into sandstone-rich Upper and Lower Fan 4 (Hodgson et al., 2006;Spychala et al., 2017;Hansen et al., 2019).

METHODOLOGY AND DATA SET
This study utilises 21 sedimentary logs measured at 1 : 20 scale (Fig. 2). Logs were used to construct seven correlation panels; three oriented down depositional-dip and four oriented along depositional-strike (Figs 1C and 2). Correlations were made in the field by mapping key packages and surfaces. In addition, aerial and unmanned aerial vehicle (UAV) photographs were used to further support correlations in areas difficult to access and were used to guide and supplement geometric interpretations observed in the field. Data collected include lithology, bed thickness and palaeocurrents (n = 107) measured from ripple cross-lamination, wood-fragment long-axis orientation and channel incision surfaces.

FACIES AND CHANNEL HIERARCHY
The outcrops at Klein Hangklip have been interpreted as submarine channel-fills containing turbidites (Wild et al., 2005). Six lithofacies were identified and are summarised in Table 1. The lithofacies are grouped into a channel-fill facies association.

Lithofacies description and interpretation
F1: Siltstone Description. Fine-grained to coarse-grained siltstones form packages or caps to individual beds (Fig. 3A). Fine-grained siltstone packages appear homogenous at outcrop, though on a micro-scale siltstones in the Tanqua depocentre typically consist of 0.1 to 1.0 mm scale beds (Boulesteix et al., 2019). Typically, coarse-grained siltstones are well-bedded ( Fig. 3A), have bed thicknesses of 0.01 to 0.07 m, and are frequently rippled. Claystones were not observed in the study area.

F2: Laminated sandstone
Description. Very fine-grained to fine-grained sandstone beds 0.05 to 2.5 m thick with alternating finer and coarser 0.5 to 1.0 mm thick laminae which are bed-parallel to sub-horizontal (Fig. 3B). Beds are commonly sharp-topped or amalgamated, but rare examples grade to siltstone. Commonly, laminated sandstones form an upper division to a sandstone bed, which has a lower, structureless sandstone division. Organic material, including leaf or wood-fragments, is commonly preserved parallel to laminae.
Interpretation. Sandstones with parallel lamination can be deposited from the repeated formation and collapse of near-bed layers termed 'traction carpets' in high-concentration flows (Dzulynski & Sanders, 1962;Hiscott & Middleton, 1979;Lowe, 1982;Sohn, 1997;Sumner et al., 2008); or by the Mudstone chips may be present locally Rapid settling from a high concentration flow (e.g. Lowe, 1982) F5a F5b Cross-bedded sandstone 0.3-1.0 Lower finegrained sandstone Cross-bedded sandstones with foresets that are tens of centimetres high. Foresets can be clast rich, or clast poor Deposition and reworking from fast-moving, longlived, low-concentration turbidity currents (Allen, 1970a(Allen, , 1982Baas et al., 2004;Sumner et al., 2012;Talling et al., 2012) migration of low-amplitude bed-waves in lowconcentration flows (Southard, 1991;Best & Bridge, 1992). It is challenging to distinguish which depositional process forms a given parallel laminated deposit at outcrop (e.g. Talling et al., 2012). However, parallel laminated sandstones at Klein Hangklip are frequently associated with thinner-bedded deposits in channel off-axis and margin positions, where turbidity currents were likely relatively low-concentration (Altinakar et al., 1996;Hansen et al., 2015;Jobe et al., 2017), and therefore migration of low-amplitude bed-waves is preferred. Interpretation. Ripple laminated sandstones are deposited from fully turbulent flows, or parts of flows, which are low concentration, have relatively low rates of fallout, and can rework the bed (e.g. Walker, 1965;Allen, 1982;Southard, 1991;Baas, 1994Baas, , 1999. Thick packages of rippled sandstone with supercritical angles of climb are interpreted to be deposited by sustained, relatively dilute, flows in which the depositional rate was in excess of that of bedform migration (e.g. Sorby, 1908;Allen, 1970bAllen, , 1991Baas et al., 2000;Jobe et al., 2012).

F4: Structureless sandstone
Description. Very fine-grained to fine-grained sandstone beds 0.02 to 6.7 m thick, which form packages up to 10 m thick. Beds are apparently structureless and ungraded to weakly graded ( Fig. 3D and E). Sandstone beds are frequently dewatered, which is typical of sandstones in the basin (e.g. Hodgson et al., 2006). Beds often form either: (i) highly amalgamated packages in which bedding surfaces are challenging to distinguish (Facies F4b; observed along bed tops, or rarely observed within the bed itself. Mudstone clasts are commonly observed on bed tops, or isolated within the bed.
Interpretation. Apparently structureless sandstones are deposited incrementally from high-concentration, near-bed, parts of flows in which high depositional rates inhibit the development of tractional bedforms (Kuenen, 1966;Lowe, 1982;Kneller & Branney, 1995;Baas et al., 2004;Leclair & Arnott, 2005;Sumner et al., 2008). Rapid near-bed aggradation is interpreted to create excess pore pressures, resulting in dewatering (Lowe, 1975). An alternative mode of deposition of is from the en masse freezing of sandy debris flows (e.g. Shanmugam, 1996). However, structureless sandstones in the field area commonly grade laterally into laminated sandstones and do not pinch-out abruptly, which are suggestive of spatial changes in flow behaviour, as opposed to freezing of the flow (Talling et al., , 2013. As such, sandstones here are interpreted as the product of high-concentration near-bed layers of turbidity currents. F5: Cross-bedded sandstone Description. Very fine-grained to fine-grained cross-bedded sandstones with foresets that reach tens of centimetres in height are rarely observed in the study area (Fig. 3F). Foresets may contain abundant mudstone-clasts aligned parallel to laminae, though more typically clasts are absent.
Interpretation. The development of dune-scale cross-bedded sands is suppressed at high nearbed concentrations and sediment fallout rates (Lowe, 1988;Baas et al., 2011). Dune-scale 30 cm cross-bedding in deep-water systems is therefore interpreted to form from long-lived, fast-moving, relatively dilute flows that had low rates of sediment fallout (Walker, 1965;Allen, 1970a;Sumner et al., 2012).

Lens cap
F6: Mud-clast-rich sandstone Description. Mudstone-clast-rich sandstones are structureless. They also contain an abundance of mudstone clasts distributed throughout the bed or package ( Fig. 3E and G).

Facies associations
One major facies association is identified in the study area. This is a channel-fill facies association, which is described below. FA1: Channel-fill Channel-fills can exhibit a wide-range of fillstyles, which can be partially preserved, symmetrical or asymmetrical; and can be filled with different proportions of sandstone, mudstone or debrites. Such variability can occur at the individual channel element, complex, and complex set scale (e.g. Pickering & Corregidor, 2005;Mayall et al., 2006;Pyles et al., 2010;Moody et al., 2012;Bayliss & Pickering, 2015;Zhang et al., 2015;Li et al., 2016). Despite this, a generalised facies association for the Klein Hangklip channel-fill is proposed. Divergence from these descriptions across different channel elements, complexes and the channel complex set are described in the results.
Where symmetrical, a given channel-fill typically exhibits a gull-wing-like geometry. The channel axis is located at the thickest point of the channel-fill, and predominantly consists of comparatively thick-bedded, amalgamated F4a, F4b and F6 (Table 1; Figs 3 to 5). Channel offaxis positions overlie the steep erosional cut of the channel, with the fill being stratigraphically thinner, and comprise comparatively thinnerbedded F4, F4b and F2; with subordinate F6 (Figs 3 to 5). Channel margin positions overlie the low-gradient, upper parts of the channel-cut, corresponding to the upper and outer parts of a gull-wing geometry. Channel margin deposits are comparatively thin and are laterally extensive (Fig. 4). Channel margin facies typically consist of comparatively thin-bedded F2 and F4a, with localised F1 and F3 (Figs 3 to 5).

Channel hierarchy
Submarine slope channel-fills are hierarchically organised ( Fig. 5; e.g. Sprague et al., 2002Sprague et al., , 2005Di Celma et al., 2011;McHargue et al., 2011;Moody et al., 2012;Macauley & Hubbard, 2013;Stright et al., 2014;Li et al., 2016). The hierarchy used in the current study is based on Sprague et al. (2002Sprague et al. ( , 2005, though note the substitution of 'channel fill' for 'channel element'; from the smallest to the largest scale, the hierarchy consists of: (i) beds/facies that share similar lithologies; (ii) storeys, which consist of beds filling an individual erosion surface or scour (Friend et al., 1979); (iii) channel elements, which represent a single cycle of cutand-fill (Fig. 5), and can contain numerous storeys; (iv) channel complexes formed from two or more nested channel elements (Fig. 5); and (v) channel complex sets formed by two or more stacked channel complexes (Fig. 5). This hierarchical scheme has been applied to the Unit 5 succession based on scale and stacking patterns; and by correlation of logs and mapping of bounding surfaces both in the field and using aerial photography.  . Schematic summary of the hierarchical scheme applied to the Klein Hangklip channel-fill. 'Channel complex set' is the largest hierarchical level observed and consists of two channel complexes, each comprising four channel elements. Sub-environment geometries and facies distributions are schematically illustrated based on previous work (Mutti, 1977;Campion et al., 2000;Sullivan et al., 2000;Eschard et al., 2003;Beaubouef, 2004;Macauley & Hubbard, 2013).

Architectural element descriptions
The architectural elements that make up the Klein Hangklip channel complex set are described below. The channel complex set is interpreted to consist of two channel complexes: stratigraphically, 'KHKC' and 'KHKD', each of which is made up of four channel elements.
Channel element 1: KHKC The exposed outcrop of KHKC has a lenticular geometry (Fig. 4), representing approximately one half of the original channel complex from its left-hand channel margin to the channel-axis (relative to palaeoflow; Fig. 6). KHKC incises into the underlying siltstones and locally incises Upper Fan 4 (Fig. 6). The maximum observed incision is 20.5 m. The base of KHKC is rarely exposed, though the channel-base-deposit is observed 100 m to the south of Log 19 where KHKC incises F3 facies of Upper Fan 4 (Fig. 7A). KHKC is at least 890 m wide, has a maximum thickness of 32.9 m at Log 3 ( Fig. 2; Panel 5), and thins and fines northward ( Fig. 6; Panels 5 to 7). The minimum thickness of KHKC is 9.5 m at the channel margin position of Log 16 ( Fig. 2; Panel 7), although KHKC visibly continues to thin for 125 m to the north of that position. No external lev ees belonging to KHKC are identified, although their existence cannot be ruled out. If lev ees are present, they are likely to be siltstone-prone, and weather recessively, resulting in exposures being covered by vegetation and scree. Four channel elements are identified in KHKC; these are (KHK) C1 to (KHK) C4. KHKC1: C1 is lenticular in geometry with a maximum thickness of 10.2 m at its axis (Log 20), and is 5.4 m thick in the channel margin at Log 18 (Figs 4 and 6). C1 continues to thin northward; however, it becomes increasingly challenging to discern individual channel elements. The base of C1 is rarely exposed, except south of Log 19 (Figs 1C and 7A), where it comprises: (i) a basal incision surface lined by mudstone clasts which cuts into underlying sandstones; overlain by (ii) 0.2 m of thin-bedded coarse siltstone; and (iii) 1.1 m of lenticular thin-bedded sandstones, with rare localised mudstone clasts and mediumbedded sandstones that lie on a subtle erosion surface into thin-bedded sandstones (Fig. 7B). The upper surface of the channel-base-deposit is mudstone-clast-rich and is immediately overlain by KHKC1. The channel-base-deposit is poorly exposed at Log 3, where a 0.1 m thick (minimum) mudstone-clast-rich siltstone lag is observed below the basal F4 of the channel-fill. The channel axis (for example, Log 3; Fig. 2) of C1 is composed of amalgamated F6 with mudstone-clast-rich bed tops at the base, overlain by F4b (Fig. 4). F6 beds decrease in number and mudstone clasts reduce in size stratigraphically upward (Figs 4 and 6). The channel margin of C1 primarily comprises F4a, although F2 and F3 are observed locally (Figs 4 and 6).
KHKC2: C2 is 16.8 m thick at its channel axis at (Log 3; Fig. 2; Panel 4), thinning to 6.2 m to the north-east at Log 18, and 3.5 m to the south at Log 13 (Fig. 8). The base of C2 incises into C1 and is marked by a channel-base-deposit that has variable facies. In the westernmost logged sections (Logs 20 and 21), the channel-base-deposit of C2 is a basal mudstone-clast-rich layer overlain by a 0.1 to 0.2 m thick fine siltstone with localised mudstone clasts. One kilometre down-dip to the east (Log 3) the base of C2 is characterised by amalgamated, mudstone-clast-rich sandstones that are locally eroded through a thin siltstone bed (Fig. 8). At Logs 4, 5 and 13, the channel-basedeposit of C2 is expressed as a clast-rich composite channel-base-deposit that is up to 1.5 m thick with multiple composite scour surfaces. The channel-base-deposit is composed of siltstone with decimetre-scale rounded and angular sandstone clasts, remobilised bedded sandstones up to up to 1 m in length, and abundant mudstone clasts that range from centimetres to a metre in length. The deposit is also locally expressed as a 0.3 m thick clast-rich siltstone with thin beds of discontinuous clay-rich sandstones, or as an amalgamation surface with the underlying sandstones of C1, mantled by mudstone clasts. The fill of C2 varies spatially, both laterally from channel axis to channel margin positions, and in different panels down depositional dip (Figs 6 and 8). Typically, the channel axis comprises F4b, with F6 locally observed in the lower 2 m (Figs 6 and 8). Up to 4.5 m of F3 is observed in the channel axis of Logs 20 and 21 ( Fig. 6; Panel 4). Bedding becomes less ambiguous as amalgamation decreases towards off-axis and channel margin positions (Fig. 4); this is accompanied by a decrease in bed thickness, and an increase in the proportion of F4a and F2 (Figs 4 and 6).
KHKC3: The basal erosion surface and channelbase-deposit of C3 are relatively flat-lying and have been identified at outcrop through an abrupt stratigraphic change in facies from thickbedded to relatively thin-bedded sandstones (Fig. 4). C3 has a tabular geometry relative to those observed in C1 and C2 (Figs 4 and 6). The fill of C3 has a maximum thickness of 8 m at Log 3, which is located in the axis of KHKC (Figs 2 and 4), but is typically 3 to 4 m thick elsewhere (Fig. 2). The channel-base-deposit of C3 is laterally variable, comprising a 5 to 20 cm siltstone at Logs 4, 5, 16, 18 and 20 (Fig. 2), a clast-rich siltstone at Logs 1, 2 and 21 (Fig. 2), and amalgamated thin beds at Logs 3, 12 and 15 (Fig. 2). The fill of C3 predominantly consists of F4a and F4b (for example, Log 3; Table 1; Figs 4 and 6). Channel axis and off-axis positions contain F6, which is typically located near the base of the channel element (Figs 4 and 6). Bed thickness and degree of amalgamation decrease into channel margin positions, which comprise F2 and F4a, with rare F6 overlying the channel-basedeposit (Figs 4 and 6). In southern channel margin localities (Logs 4 and 5) the fill of C3 is dominated by F2, with localised F4a (Figs 8 and 9).
KHKC4: C4 is the uppermost channel element of the KHKC channel complex and has a relatively tabular geometry (Figs 6 and 8). The total thickness of C4 is unknown as the upper portion is poorly-exposed and therefore thicknesses should be considered a minimum value. C4 is thickest in the channel axis at Log 17 (7.8 m; Fig. 2), the measured thicknesses at other localities are typically 3 to 5 m (Figs 6 and 8). The C4 erosion surface and channel-base-deposit are rarely exposed, and they have a variable character where they can be observed. The channelbase-deposit is typically a thin siltstone (<10 cm, locally up to 30 cm; Fig. 7E) that may contain mudstone clasts, although the base is also locally expressed as an amalgamation surface. The facies of the C4 channel element fill is variable, particularly from channel axis to channel margin. In axial positions, C4 comprises F4a with local F4b and F6 (Figs 4 and 6). Off-axis positions consist of F2 and F4a, with F4b and F6 locally present down-dip (Figs 6 and 8).
Channel margin positions are predominantly composed of F2, with localised F4a (Figs 4 and 6). The south-eastern most position of C4, Log 5, has 2 m of F1 at the base and is overlain by 1 m of F3 (Fig. 9).
Channel element 2: KHKD The uppermost sandstones of KHKC are overlain by approximately 2 m (minimum) of siltstone at Log 5, which is incised by the erosion surface of the overlying channel-fill, KHKD. The KHKD channel complex is exposed in a 2.7 km wide (minimum) outcrop oriented oblique to strike, (Panel 1; Figs 2 and 9) and is not exposed on the north face of Klein Hangklip. KHKD is thickest (39.0 m) at Log 7 (Fig. 6) and thins into the east and west channel margins to 6.2 m (Log 10) and 4.5 m thick (Log 6), respectively (Figs 2 and 8). The western composite-cut of KHKD is steeper than the eastern channel-cut (Figs 8 and 9). The margin of the channel complex pinches out approximately 40 m to the west of Log 6 (Figs 8 and 9). The pinchout of the eastern margin is not observed due to the oblique nature of the outcrop, although the thinning and change of facies to predominantly laminated sandstone suggests that it is located south of Logs 9 and 10 (see Fig. 10). The channel axis of KHKD incises a maximum of approximately 22.5 m into the underlying fill of KHKC at Log 7 (Figs 8 and 9). Similar to KHKC, no external lev ees are identified, although they are possibly present but not exposed. Four channel elements have been identified in KHKD (D1 to D4): KHKD1: D1 is the oldest channel element in the KHKD channel complex. D1 is mapped over 2.7 km in Panel 1 and is best exposed on the southern face of the outcrop (Figs 8 and 9). The basal erosion surface and channel-base-deposit of KHKD1 form the base of the channel element and are well-exposed across the outcrop (Figs 9 and 11A). In channel margin positions, the channel-base-deposit is predominantly composed of at least 11 cm of F1 with rare mudstone clasts and lenticular very fine sandstone beds. In channel off-axis positions, the channelbase-deposit is 5 to 30 cm thick and consists of F1 and thin F4a, which are discontinuous and locally slumped (Fig. 11B). In channel axis positions, the channel-base-deposit is composite, up to 2.5 m thick, and comprises: amalgamated F6; remobilised clasts that are commonly sheared [ Fig. 11C; F1, which is commonly amalgamated; F7 (Fig. 11C); and F4b, which is locally amalgamated with underlying KHKC sandstones ( Fig. 11D)]. KHKD1 is up to 17.0 m thick in the channel axis at Log 7 (Fig. 2) and thins laterally to 2.5 m and 2.2 m at its western and eastern channel margins at Logs 6 and 10, respectively (Panel 1; Figs 2 and 8). In proximal positions, D1 thins northward from 9.5 m at Log 20, to 6.3 m at Log 21 over 70 m (Fig. 6). The proportion of F4a decreases, which is accompanied by an increase in F2 (Fig. 6). On the southern outcrop face, the channel-base-deposit of D1 is 1.5 m thick and composite at Log 7 (Fig. 11C), becoming thinner and more silt-prone to both the east and west at Logs 9 and 6, respectively ( Fig. 2; Panel 1). The facies distribution of D1 is asymmetrical (Figs 8  and 9). Towards the west (Logs 5 and 6; Fig. 2), D1 predominantly consists of F2 (Figs 8 and 9), with abundant laminae-parallel plant fragments. The proportion of F4b and amalgamated F6 increase into the thickest and most axial parts of D1 at Logs 7 and 14, adjacent to the steep western channel-cut (Fig. 9). Thin (<10 cm) layers of F1 wedge out from the steep western channelcut ( Fig 11B). F1 beds are incised by overlying deposits that contain abundant clasts of F1 towards the axial positions ( Fig. 11B and C). The eastern channel-cut, from Log 9 to Log 8, has a relatively shallower gradient (Fig. 8). The channel-fill exhibits a gradual decrease in F4b and F6 away from the channel axis, and a concurrent proportional increase in F2 to the east of Log 7 (Fig. 8). Stratigraphically within D1 the proportions of F4b and F6 decrease upward with a concurrent increase in F2 (Figs 8 and 9). Locally, F5 is observed at the base of D1 at Log 20 ( Fig. 3F) and Log 8.
KHKD2: D2 has a subtle unconformable contact with D1 and can be traced on aerial photographs between Logs 13 and 14 (Fig. 9). D2 has a lenticular geometry, thickening from east of Log 13 towards Log 7 (Fig. 9), but is poorly exposed to the east where the cliffs are lichen-covered. The base of D2 is characterised by a sharp contact between thin-bedded F2 of D1, to thick-bedded F2 (Fig. 7D) and 1.2 m of F4b of D2, respectively. The upper-fill of D2 consists of 3.3 m of bedded (0.02 to 0.3 m thick) F6 and also contains localised 0.1 to 0.5 m thick scour-fills consisting of centimetre-scale beds of F4a and thinbedded F6.
KHKD3: D3 has a relatively tabular geometry, with thicknesses between 3.9 m and 1.2 m (Fig. 9). The channel-base-deposit of D3 is 0.1 to 0.3 m thick, flat-lying, laterally continuous, and consists predominantly of F1, which is locally clast-rich. Localised amalgamation of thinbedded F4a is also observed. However, adjacent to the steeper western channel cut of KHKD, the channel-base-deposit of D3 is expressed as an up to 0.3 m thick composite deposit consisting of mudstone-clasts and sandstone-clasts and argillaceous sandstones. Laterally to the east, over 2 to 3 m, this channel-base-deposit is preserved as a sandstone amalgamation surface. In all cases the channel-base-deposit overlies a sandstone bed mantled with mudstone clasts. At Logs 13 and 7, D3 consists of F4b and F6 (Figs 8  and 9). The proportions of F4b and F6 decrease to the east and west with a concurrent increase in the proportion of F2 (Figs 8 and 9).
KHKD4: D4 is the uppermost channel element identified in the KHKD channel complex. The thickness of D4 is variable, from 13.0 m at Log 7 to 1.4 m thick at Log 10 (Fig. 2); these are minimum recorded thicknesses due to modern erosion. The basal surface of D4 is relatively flatlying, suggesting that the channel element has an overall tabular geometry (Figs 8 and 9). The channel-base-deposit of D4 is typically characterised by fine-grained, thin-bedded F4a, which incise into and amalgamate with the underlying sandstones of D3 ( Fig. 7G and H). Locally, the channel-base-deposit comprises 10 cm of F1, which is clast-rich and overlain by centimetre-thick to decimetre-thick beds of F6. Where KHKD is thickest, the fill of D4 predominantly comprises F6b at the base, and localised F2 in the upper 2 to 3 m (Figs 8 and 9). Where KHKD thins to the east, D4 consists of F4a and F4b, with localised F2 (Figs 8  and 9). Conversely, to the west D4 predominantly comprises F2 at Log 5 (Figs 8 and 9).

Palaeocurrents
Unit 5 at Klein Hangklip is interpreted to have north-eastward oriented palaeoflow (Wild et al., 2005). Palaeoflow directions of the Hangklip channel-fills are constrained by ripple crosslaminations, channel-cut orientations, outcrop constraints, and preferential orientation of wood-fragment long axes. KHKC Ripple cross-lamination measurements from Logs 1, 12 and 20 through KHKC suggest northward palaeoflow, whereas measurements from Logs 4 and 21 suggest southward and eastward palaeoflow, respectively (Fig. 10A). Long-axes of wood-fragments in Logs 4 and 15 are oriented east-west (Fig. 10A). The northward thinning of KHKC towards Logs 15, 16 and 18 (Fig. 6) suggests that the axis of the channel was situated south of those localities (Fig. 10A). The absence of KHKC in south-eastern Logs 10 and 11 indicates that the channel was located to the north of these positions. Channel margin facies observed in channel elements C3 and C4 at Logs 4 and 5 suggest that the axis of these channel elements was positioned to the north (Figs 9 and 10A).
These data suggest that KHKC was oriented north-east/south-west, with a north-eastward palaeoflow consistent with regional palaeocurrent observations (e.g. Hansen et al., 2019), and is essentially straight at the scale of the outcrop (Fig. 10A). Variable ripple cross-lamination trends are interpreted to represent flows that over-spilled the channel axis, and which may have been deflected off confining slopes (e.g. Kane et al., 2010).

KHKD
The western margin of KHKD is well-constrained at Log 13, where the strike of the channel-cut is oriented north-east/south-west (Figs 9 and 10B). Long-axes of wood-fragments show preferred orientations to the east at Logs 13 and 20, and north-east at Log 7, respectively (Fig. 10B). At Log 11, the preferred wood-fragment orientation is east/south-east. The presence of KHKD at Logs 20 and 21, and eastward of Log 5, but absence at Logs 3 and 4, suggest that KHKD curved around these positions to the south (Figs 6 and 10B). The absence of KHKD in localities on the north face of Klein Hangklip suggests that the channel lay to the south (Figs 6 and 10B). Exposure of KHKD is continuous along the eastwest oriented southern face, suggesting that the orientation of KHKD is parallel to sub-parallel to the outcrop (see Panel 1; Figs 2 and 10B). Log 10 contains a greater proportion of channel margin facies relative to Logs 9 and 11, which are both thicker, and show an increase in channel axis facies (F4b and F6) in their respective directions (Panel 1; Fig. 2B), suggesting that the channel margin is located to the south of these positions (Fig. 10B). The channel-cut and wood-fragment orientation at Logs 7 and 13 suggests that KHKD was sinuous at the scale of the outcrop (Fig. 10B). This is further supported by facies asymmetry at the western margin (e.g. Pyles et al., 2010), and facies transitions between Logs 9, 10 and 11, which suggest that the channel axis curved around to the north of Log 10.

Channel architecture interpretations
Type-1: Low aspect ratio channel elements KHKC1 and KHKC2, and KHKD1 and KHKD2, are the two lowermost channel elements in their respective channel complexes (Figs 6 and 8).
Each channel element is thickest in the channel complex axis, thinning towards the margin of the channel complex, and incised into the underlying stratigraphy by 5.5 to 22.5 m (Figs 4  and 9). Basal erosion surfaces have gull-wing geometries, and are typically mantled by mudstone clasts in the channel axis, which decrease in abundance towards the channel margins (Figs 6,7 and 11A to D). At the channel axes, channel-base-deposits are typically composite, forming packages up to 2.5 m thick, and consist of clast-rich siltstone, discontinuous argillaceous sandstones and locally remobilised beds (Fig. 11C). In off-axis positions, channel-basedeposits are more silt-rich with thin sandstone beds, and localised metre-scale slide blocks (Fig. 11B); locally the channel-base-deposit is preserved as a mudstone clast-rich amalgamation surface (Fig. 11D). At channel margin positions, the channel-base-deposit is characteristically represented by 0.5 to 0.3 m of siltstone. The channel axis of low aspect ratio channel elements is characterised by amalgamated F4b with localised F6 near the base, and F2 is sometimes observed in the upper few metres (Figs 6  and 8). Off-axis positions typically consist of F4a and F4b with increasing proportions of F2 towards the channel margin; F6 is rarely observed (Figs 6 and 8). The channel margins consist of F4a and F2 (Figs 6 and 8).
Type-2: High aspect ratio channel elements KHKC3 and KHKC4, and KHKD3 and KHKD4 constitute the uppermost channel elements in their respective channel complexes (Figs 6 and  8). The channel elements are tabular in geometry and have comparatively high aspect ratios with relatively consistent thicknesses (typically 2 to 6 m), but locally have thicknesses between 1.2 m and 13 m (Figs 6 and 8). Basal erosion surfaces are typically tabular to subhorizontal and commonly mantled by mudstone clasts (Figs 7 and 9). The associated channel-base-deposit is typically 0.05 to 0.3 m of siltstone with localised mudstone clasts, though it may be present as an amalgamation surface marked by mudstone clasts. Channel axis deposits of high aspect ratio channel elements consist of bedded F4a and uncommon F4b, with localised F6 and F2 at the base and top, respectively (Figs 4 and 6). Off-axis positions comprise comparatively thinner-bedded F4a and F2 (Figs 4 and 6). Channel margin positions are characterised by a proportional increase in F2 (Figs 4 and 6). F1 and F3 are observed in KHKC4 in south-eastern localities (Fig. 6).
Channel element architecture KHKC and KHKD are each composed of four channel elements; two low aspect ratio channel elements overlain by two high aspect ratio channel elements. Both channel complexes are high sand-to-gross, the minimum recorded value is 88.5% in KHKC at Log 15. All other logged sections have sand-to-gross values in excess of 90%.
KHKC. Palaeocurrent directions, outcrop geometries and facies transitions suggest that KHKC is relatively symmetrical and straight at the scale of the outcrop (Fig. 10A). The channel axis of each channel element in KHKC is situated in the channel complex axis (Fig. 6) suggesting aggradationally stacked channel elements with relatively symmetrical facies distribution. Channel margin positions are dominated by bedded F2 and F4a, with localised F1 and F3 (Figs 4  and 6). Off-axis positions are characterised by F2 and F4a, infrequent F4b and rare F6 (Figs 4 and  6). Channel axis positions are dominated by F6 and F4b in the lower, low aspect ratio channel elements, but comprise bedded F4a and F4b with localised F6, and an upward proportional increase in F2 (Figs 4 and 6).
KHKD. Palaeocurrents and outcrop geometries suggest that KHKD was sinuous at the scale of the outcrop (Fig. 10B). The channel-complex geometry is asymmetrical, with a steeper western channel cut and shallower eastern channel cut (Figs 8 and 9). Each channel element is thickest and contains respectively higher proportions of F6 and F4b in positions in the channel axis immediately east of the western channel cut (Figs 8 and 9). Facies transitions to channel margins dominated by F2 are gradual to the east, but comparatively abrupt to the west against the steeper channel cut (Figs 8 and 9). The asymmetrical channel element facies distribution suggests that the channel axes of successive channel elements were concentrated in the outer bank of the channel complex bend, and stacked aggradationally (Figs 8 and 9; see also Navarro et al., 2007;Jobe et al., 2010;Labourdette & Bez, 2010;Pyles et al., 2010;Li et al., 2016). These relationships are strongest in low aspect ratio channel elements, D1 and D2 (Figs 8  and 9). The upper, high aspect ratio channel elements D3 and D4 exhibit more-laterally extensive deposits of F4a and F4b; however, F6 is only identified in the channel complex axis (Figs 8 and 9).

Distribution of channel-base-deposits
Low aspect ratio channel element channel-basedeposits typically have a relatively steep angular contact to underlying strata and exhibit facies asymmetry (for example, Figs 9 and 11). Channel axis positions are characterised by relatively thick composite channel-base-deposits (up to 2.5 m), comprising clast-rich siltstones, remobilised clasts of F4 and F6 up to 1.5 m in length, mudstone rafts and amalgamated packages of F6 (Fig. 11C). Locally, channel-base-deposits are not composite, and are identified only as clast-rich amalgamation surfaces with the sandstones of the underlying channel element (Figs 7C and 11C). Typically, off-axis positions exhibit a slightly thinner deposit and contain localised slumped beds that were shed from the channel erosion surface (Fig. 11B). Channel margin positions are characterised by <30 cm thick, locally clast-rich, siltstone deposits that typically overlie a mudstone-clast-rich bed top. KHKD2 differs from other low aspect ratio channel elements because it lacks a preserved channel-base-deposit (Fig. 7C). Instead, the subtle erosion surface is marked by a sharp, sub-parallel, facies transition from thinbedded F4a to F4b (Fig. 7C).
High aspect ratio channel element channelbase-deposits are relatively flat-lying, are characterised by siltstones up to 40 cm thick, and are less heterogeneous compared to those of lowaspect ratio channel elements (Fig. 7E to H). Channel-base-deposits in channel axis positions frequently contain millimetre-scale to centimetre-scale clasts of siltstone and commonly overlie a mudstone-clast-rich bed top. Locally, the channel-base-deposit is marked by a clast-rich amalgamation surface. Channel off-axis and margin positions are comparatively clast-poor, although local beds of siltstone starved ripples are observed. The KHKD4 channel-base-deposit comprises 5 to 10 cm of thin-bedded coarsesiltstone to very fine-sandstone beds, and locally a <10 cm siltstone, which are subtly incised by overlying beds of the channel-fill ( Fig. 7G and   H). In contrast, the KHKD3 channel-base-deposit is locally composite adjacent to the steep KHKD channel-cut surface.

Stratigraphic evolution
Incision and the resultant development of a composite channel complex-set erosion surface is the first recorded phase of channel evolution (Figs 12A and 13). In common with other documented examples, the development of the surface is likely to have been time-transgressive (Sylvester et al., 2011;Hodgson et al., 2016;Englert et al., 2020). However, the record of the development of the surface, and its formative processes, is obscured due to both the erosion of stratigraphy deposited during excavation of the surface, and the limited exposure of the surface and its channel-base-deposit. The 1.5 m thick channel-basedeposit mantling the channel complex set erosion surface ( Fig. 7A and B) is interpreted to have been deposited from relatively dilute parts of multiple flows, which were primarily confined in the channel axis (e.g. Hubbard et al., 2014). These deposits suggest that the erosion surface acted as a long-lived conduit for the bypass of sediment into the deeper basin ( Fig. 13; e.g. Hubbard et al., 2014;Stevenson et al., 2015;Englert et al., 2020). The nature of channel-base-deposits is used as a proxy for the energy, and number of, bypassing, partially bypassing, and depositional flows; and the relative durations of the complete sedimentbypass, bypass-dominated, and depositional phases (sensu Stevenson et al., 2015). This approach assumes that the preserved deposit reflects the time-averaged flow-processes during deposition, and that any material that was deposited is preserved.
The fill of KHKC records the transition from complete sediment bypass to a depositional zone (sensu Stevenson et al., 2015;Figs 12B and 13), relative to the channel axis. Aggradational phases are interpreted to be comparatively short-lived due to decreased evidence of erosion and reworking (see also Englert et al., 2020). The first stage of this aggradation is recorded by the strongly channelised fill of C1 (Fig. 12B). A subsequent phase of incision and filling is marked by channel element C2 (Fig. 12C), indicating an increase in flow-energy, sediment bypass, and degradation of the slope (Fig. 13). Flows causing degradation of the slope were not necessarily larger (e.g. Sylvester et al., 2011), but were more erosive than flows resulting in aggradation. The C2 channelbase-deposit is highly composite, suggesting that there was a long-lived bypass-dominated zone before aggradation of the C2 fill (Fig. 13). The individual fills of C1 and C2 record a vertical decrease in evidence for sediment bypass up stratigraphy (Fig. 13), from F6 and F4b at the base to F2 and F4a at the tops (Figs 4 and 6).
The upper-fill of C2 was partially incised during the formation of the higher aspect-ratio channel element C3, indicating an increase in flow energy, erosion, and sediment bypass (Figs 12D and 13). The C3 fill was also partially incised as part of the development of channel element C4. The thin siltstone-rich channelbase-deposits, shallow depth of incision and less-composite nature suggest that erosion and bypass was less-pronounced and protracted, probably due to less-erosive flows, compared to the incisions related to C1 and C2 (Fig. 13). The fills of C3 and C4 are similar and suggest a temporal decrease in flow energy (Figs 12D and 13).
Following the fill of the KHKC channel complex, approximately 2 m of siltstone was deposited, representing a prolonged hiatus of sand supply to the area. The period of silt-prone deposition was ended by incision of the KHKD channel-cut (Fig. 12E) indicating an increase in flow energy and sediment bypass ( Fig. 13; e.g. Kneller, 2003); possibly driven by eustatic sealevel fall (e.g. Flint et al., 2011) and the overall progradation of the system (e.g. Hodgson et al., 2006). The highly composite channel-basedeposit includes depositional and erosional features at the base of the D1 channel axis, suggesting that the channel axis was a long-lived sediment bypass zone ( Fig. 13; e.g. Hubbard et al., 2014;Stevenson et al., 2015).
A gradual transition to net depositional flows resulted in the fill of the KHKD channel complex (Figs 12F and 13). KHKD1 records an upward decrease in in F6 and bed amalgamation, supporting a temporal reduction in sediment bypass (Figs 12F and 13). Facies asymmetry, with F6 and F4b being more common adjacent to the steep western channel-cut (Fig. 12F), suggests that the highest energy flow components were located near the base of the outer-bank (secondary or helical flow; see Keevil et al., 2006;Imran et al., 2007;Peakall et al., 2007;Peakall & Sumner, 2015). Erosion of F1 lenses from the channel-cut in towards the channel axis ( Fig. 11B and C) suggests that flows were highly energetic in the thalweg of the channel axis (Reimchen et al., 2016), and decelerated against the outer channel-cut (e.g. Keevil et al., 2006).
The incision related to the fill of D2 (Fig. 7C) is interpreted to represent erosion by bypassing flows. The channel-base-deposit related to D2 is sharp and non-composite, suggesting a shortlived bypass-dominated phase, or that subsequent flows eroded the initial channel-basedeposit (Fig. 13). The D2 fill records aggradation of the channel fill by lower energy, depositional flows (Fig. 13). However, the upper fill of D2 contains scours, which were infilled by thin beds. This may be associated with a gradual increase in flow energy, and record the transition from depositional, to partially bypassing, to fully bypassing flows, which resulted in the incision of the D3 channel-cut (Figs 12G and 13; see also package 2 of Pyles et al., 2010). The upper part of D2 is weakly incised, considered to reflect a decrease in flow energy and sediment bypass compared to the incision of underlying channel elements (Fig. 13). The lack of incision and composite channel-base-deposits suggests  Fig. 13. Schematic illustration of channel evolution and flow behaviour in the channel axis of respective channel complexes. Highly composite erosion surfaces and channel-base-deposits suggest the excavation of channel cuts and bypass phases record long periods of time. However, most of the thickness of the channel-fill is represented by aggradational sandstones developed under strongly depositional flows. This suggests that time-partitioning in submarine channels may be biased towards individual surfaces, which occupy a small portion of the channel thickness. that the amount, and subsequent duration, of sediment bypass was limited (Fig. 13). The channel-base-deposit is overlain by high aspect ratio channel element D3 (Fig. 12H), indicating that flows became deposition-dominated (Fig. 13). The fill of the D3 channel element was incised as part of the development of channel element D4 (Figs 12H and 13). The incision surface of D4 is relatively flat-lying compared to low aspect ratio channel elements, suggesting limited phases of bypassing and partially bypassing flows (Fig. 13). However, the channelbase-deposit is locally composite in the area adjacent to the steep channel-cut of KHKD, which is located on the outer bend of the channel. This suggests that higher-energy parts of flows were concentrated at the outer bend, and that bypass and substrate remobilisation was more efficient in these locations. The subsequent aggradation of the D4 channel element reflects the transition to depositional flows (Fig. 13).

Three-dimensional channel architecture
The KHK outcrop permits quasi-3D facies distribution to be recorded in a series of depositional dip-oriented and strike-oriented panels (Figs 6  and 8). The KHKC channel complex shows no substantial down depositional-dip changes in geometry or sandstone to gross. However, subtle changes in facies are recorded in dip-oriented sections, from F4 to F2 to F4, and varying thickness and distribution of F6 (Figs 4 and 6). The KHKC channel complex is interpreted to be relatively straight at the scale of the outcrop. Channel complex KHKD, which is inferred to be sinuous in planform, has a steeper western (outer-bank) channel cut, and displays a more pronounced facies variation across depositional-strike with a greater proportion of F4b and F6 compared to the eastern side of the channel cut. These observations are made on a 0.1 to 1.0 km scale, in predominantly sandstone-filled channels; similar distributions have been noted in other channel systems such as the Beacon Channel (Pyles et al., 2010). However, at a system-scale (i.e. over tens to hundreds of kilometres), channel-fills may show major longitudinal facies variability in response to the dominance of different flow processes. In proximal areas, slumps and debris flows are more abundant, whereas in distal parts of the system deposits of high-density and lowdensity turbidites are more commonly observed (De Ruig & Hubbard, 2006;Malkowski et al., 2018). Therefore, at the scale of the depositional system, flow processes, and the methods of sediment mobilisation in up-dip areas are likely to have the strongest influence on heterogeneity. Channel sinuosity, and its effect on the inherited flow processes, is likely to determine local facies distribution and channel element geometries over longitudinal profiles of 0.1 to 1.0 km (e.g. De Ruig & Hubbard, 2006;Pyles et al., 2010;Reimchen et al., 2016;Malkowski et al., 2018).

Controls on channel element architecture
Each channel complex contains two low aspect ratio channel elements overlain by two high aspect ratio channel elements. Low aspect ratio channel elements have lenticular geometries and facies distributions, with bed thicknesses that decrease from channel axis to channel margin. Highly erosive flows are interpreted to have incised the basal channel-cuts ( Fig. 13; Pyles et al., 2010;McHargue et al., 2011;Fildani et al., 2013;Hubbard et al., 2014;Hodgson et al., 2016). Subsequently flows became progressively confined, enhancing flow efficiency (sensu Mutti, 1992), increasing bypass and erosion, and further entrenching the channel (e.g. Hodgson et al., 2016). Development of thick, composite channelbase-deposits (Figs 7A, 11B and 11C) is interpreted to represent a prolonged sediment bypassdominated period in the channel axis through the early stages of waning sediment supply ( Fig. 13; e.g. Hubbard et al., 2014). This early stage of channel aggradation was likely characterised by numerous flows which were bypassing, partially bypassing and depositional. The gradual transition to net depositional flows is preserved in the channel-base-deposit where deposits of earlier depositional flows were eroded and remobilised by bypassing or partially bypassing flows (see also Vendettuoli et al., 2019). Remobilised deposits and clasts of beds ( Fig. 11B and C) suggest that successive flows were transitional between depositional, and fully bypassing, remobilising beds ( Fig. 13; e.g. Stevenson et al., 2015). This is possibly linked to smaller cycles of waxing and waning energy superimposed on the overall trend (e.g. Vendettuoli et al., 2019). KHKD2 differs from other low aspect ratio channel elements as it lacks a composite channel-base-deposit (Fig. 7C). The subtle, amalgamated KHKD2 channel-basedeposit surface is interpreted to represent either: (i) a relatively rapid transition from near-complete sediment bypass to deposition-dominated flows, inhibiting the development of a composite channel-base-deposit (Fig. 13); or (ii) incision A gradual upward increase in the number of net depositional flows resulted in initial aggradation and deposition of amalgamated F6 at the base of the channel complex axis (Fig. 13). Packages of F6 overlain by F4 at the base of channel-fills are interpreted to record the transition from bypassing, to partially bypassing, to depositional flows. During aggradation, higher-concentration flows, or parts of flows, were more strongly influenced by channel topography, and were contained in the channel axis where they rapidly deposited thick packages of F4b (e.g. Hubbard et al., 2014). Local changes in intra-channel topography and gradient may have resulted in subtle down-dip and across-strike variations in depositional facies caused by flow acceleration or deceleration, and entrainment of substrate. Dilute, low-concentration flow components were able to surmount channel topography to the channel margins, depositing thinner bedded F4a and F2 ( Fig. 12; e.g. Campion et al., 2000;Jobe et al., 2017). Disparity in deposit thickness resulted in more rapid fill of accommodation in the channel axis relative to the channel margin (see also : Hubbard et al., 2014).

Subsurface implications
Seismic reflection data do not have the vertical resolution to show complicated geometric relationships and small-scale features such as channel-base-deposits ( Fig. 14; e.g. Alpak et al., 2013;Morris et al., 2016). Therefore, well-data are crucial in identifying these features in the subsurface in order to build realistic geological models. The Klein Hangklip seismic-scale channel-fill has a high sand percentage, typically in excess of 90%. In such a channel-fill, the heterogeneities controlling reservoir performance are likely to be related to sandstone facies and depositional processes (e.g. Hirst et al., 2002;Lien et al., 2006;Zhang et al., 2015;Porten et al., 2016;Bell et al., 2018), and the distribution and character of channel-base-deposits (Larue & Hovadik, 2006;Funk et al., 2012;Alpak et al., 2013;Jackson et al., 2019).
Core logging and image log/dip-meter analysis are common ways to identify and interpret channel-base-deposits and erosion surfaces (e.g. Barton et al., 2010;Morris et al., 2016). This study reveals that these methods may prove challenging as: 1 Highly composite channel-base-deposits are locally observed as clast-rich amalgamation surfaces, not suggestive of substantial sediment bypass (Figs 11 and 14). Thus, small changes in well placement could result in dramatically different core interpretations (Fig. 14).
2 Siltstone bypass channel-base-deposits may be challenging to differentiate from abandonment or low-energy depositional drapes without identification of clasts or a mudstone-clast-rich basal surface (Fig. 14), which may influence the prediction of reservoir sandstones down-dip.
3 Some erosion surfaces are characterised by subtle amalgamation surfaces and facies changes, which may not be interpreted as a major bounding surface (Figs 7,11 and 14).
4 High aspect ratio channel elements have relatively flat-lying erosion surfaces, which may make dip-meter identification challenging (Fig. 14).

CONCLUSIONS
The Klein Hangklip outcrop of Unit 5 of the Skoorsteenberg Formation permits high-resolution analysis of the architecture, facies and stacking patterns of a submarine slope channel complex set, composed of two stacked channel complexes. The lower channel-complex, which is 32.9 m thick and at least 890 m wide, is straight at the scale of the outcrop; the upper channel-complex is 39.0 m thick, is sinuous, and exhibits architectural and facies asymmetry. Each channel-complex consists of four channelelements, which are grouped into: 1 Lower channel-elements which are: (i) lowaspect ratio; (ii) incise up to 20 m into underlying stratigraphy; and (iii) exhibit strong facies trends from abundant amalgamated structureless sandstone and mudstone-clast-rich facies in channel-axis positions to bedded structureless sandstone and laminated sandstones in channelmargin positions.
2 Upper-channel elements are: (i) high-aspect ratio; (ii) do not incise deeply into underlying stratigraphy; and (iii) have less-contrasting channel-axis to channel-margin facies transitions.
In both channel complexes systematic and predictable facies changes are observed from comparatively thick-bedded amalgamated sandstones deposited from energetic flows in channel-axis positions, to bedded and laminated sandstones in channel-margin positions. Conversely, positions in comparative down-depositional dip positions (for example, from an updip channel axis position to a down-dip channel axis position) show subtle, but non-systematic heterogeneity in sandstone facies.
Channel-base-deposits mapped in depositional-dip and strike sections exhibit spatial and temporal heterogeneity at metres to hundreds of metres length scales, which could inhibit accurate characterisation in subsurface and limitedoutcrop studies. The depth of incision and composite nature of channel-base-deposits are used as proxies for the existence of long-lived bypass conduits. This analysis suggests that time-partitioning during channel evolution is strongly biased towards excavation and maintenance of these conduits, whereas aggradation of the relatively thick-bedded sandstone-fills was comparatively short-lived.