An integrated model of clastic injectites and basin floor lobe complexes: implications for stratigraphic trap plays

Injectites sourced from base‐of‐slope and basin‐floor parent sandbodies are rarely reported in comparison to submarine slope channel systems. This study utilizes the well‐constrained palaeogeographic and stratigraphic context of three outcrop examples exposed in the Karoo Basin, South Africa, to examine the relationship between abrupt stratigraphic pinchouts in basin‐floor lobe complexes, and the presence, controls, and character of injectite architecture. Injectites in this palaeogeographic setting occur where there is: (i) sealing mudstone both above and below the parent sand to create initial overpressure; (ii) an abrupt pinchout of a basin‐floor lobe complex through steep confinement to promote compaction drive; (iii) clean, proximal sand beds aiding fluidization; and (iv) a sharp contact between parent sand and host lithology generating a source point for hydraulic fracture and resultant injection of sand. In all outcrop cases, dykes are orientated perpendicular to palaeoslope, and the injected sand propagated laterally beneath the parent sand, paralleling the base to extend beyond its pinchout. Understanding the mechanisms that determine and drive injection is important in improving the prediction of the location and character of clastic injectites in the subsurface. Here, we highlight the close association of basin‐floor stratigraphic traps and sub‐seismic clastic injectites, and present a model to explain the presence and morphology of injectites in these locations.

In sedimentary basins, lithology is the principle control on basin wide fluid migration (Bjørlykke, 1993;Jonk et al., 2005a), and in the absence of clastic injectites fractures and faults form the most efficient conduits for fluid flow (Chapman, 1987;Knipe et al., 1998;Aydin, 2000). However, clastic injectites create additional fluid flow pathways, and their impact depends on their timing and location (e.g. Hurst et al., 2003;Jonk, 2010;Ross et al., 2014). Net migration of fluids, including water and hydrocarbons, into an unconsolidated sandbody can provide the overpressure and trigger mechanism needed for sands to fluidize and inject (Vigorito & Hurst, 2010;Bureau et al., 2014). Post-injection, sandstone dykes and sills can act as fluid flow conduits for hydrocarbon leakage (Jonk, 2010) until cementation, at which point injectites become fluid flow baffles and barriers. Later, reactivation of clastic injectites as fluid flow conduits can occur through preferential brittle deformation of competent sandstones within a low-competence (majority mudstone) host rock (Jonk et al., 2005a).
For the first time, we present examples of injectites at outcrop where the palaeogeographic and stratigraphic context of the basin-floor parent sandstone lobe deposits are well constrained. We address the following objectives: (i) to document the architecture and character of injectites in basin-floor settings in terms of thickness and morphology in relation to parent sand, (ii) to investigate the association between the architecture and character of the basin-floor parent sandbody as a control on the location and orientation of injectites, (iii) to construct an integrated model of clastic injectites in basin-floor settings, (iv) to consider the role of basin-wide fluid flow pre-, syn-and post-injection, and (v) to discuss the association and implication for subsurface stratigraphic trap plays and the presence of injectites.

GEOLOGICAL SETTING
The Karoo Basin has long been interpreted as a retro-arc foreland basin that formed on the southern margin of the Gondwana palaeocontinent behind a magmatic arc and fold-and-thrust belt (Johnson, 1991;Visser & Praekelt, 1996;Catuneanu et al., 1998;Johnson et al., 2006). However, more recent studies suggest subsidence during the Permian was driven by mantle flow and foundering of basement blocks coupled to subduction of the palaeo-Pacific Plate to the south, pre-dating the Cape Orogeny (Tankard et al., 2009). The Ecca Group, a siliciclastic succession, was deposited in the southwestern Karoo Basin during the Permian . This part of the basin is subdivided into the Laingsburg and Tanqua depocentres (Fig. 1a), and this study focusses on three outcrop examples of exhumed clastic injectites hosted in deep water strata of the Ecca Group across these depocentres ( Fig. 1c and d).
Sand-prone Units C to G, which comprise the Fort Brown Formation (Fig. 1b), have been mapped over 2500 km 2 (van der Merwe et al., 2014), and are separated by regional mudstones interpreted to represent clastic input shutdown due to relative sea level rise Flint et al., 2011;Fig. 1b).

METHODOLOGY AND DATASET
Three outcrops were studied in detail; Bizansgat (Tanqua depocentre: injectites associated with Fan 3) (Figs 1-4), Zoutkloof and Slagtersfontein (Laingsburg depocentre: injectites associated with Unit C, Subunits C1 and C2) (Figs 1 and 5-7). Recognition criteria of injectites in the Karoo Basin include cross-cutting relationships, direct connection to overlying sandstones, preserved patterns on fracture surfaces of injectite margins, such as plumose patterns and parallel ridges, and blistered and mudstone clast-rich surfaces (c.f. Cobain et al., 2015). Field-based sedimentological and stratigraphic observations include logged vertical profiles, photo-panels, and dip and strike data of bedding and injectites. Physical correlation of individual beds and injectites between logs enabled the changing position of injectites with respect to host stratigraphy to be constrained from cm to km scale, which can be subtle.

Fan architecture
The depositional architecture of Fan 3 is well-constrained due to extensive outcrop study (e.g. Johnson et al., 2001;Pr elat et al., 2009;Jobe et al., 2012;Hofstra et al., 2015), and behind-outcrop research boreholes Luthi et al., 2006). Research borehole NB4 (Fig. 2) confirmed that Fans 1 and 2 are not present in this part of the study area Luthi et al., 2006). Fan 3 pinches out northward (down dip) from 65 m thick over 30 km (~2.2 m km À1 thinning rate) . Southward (oblique up dip) thinning is more abrupt, and Fan 3 thins to <2 m thick over a distance of 3 km (~22 m km À1 thinning rate) Oliveira et al., 2009). The fan has been interpreted as a basin-floor lobe complex comprising at least six sand-rich lobe deposits (Pr elat et al., 2009;Hofstra et al., 2017) and the most updip exposures at Ongeluks River are interpreted as channelized lobe deposits in a base-of-slope setting (Hofstra et al., 2015(Hofstra et al., , 2017. Beds at the southward pinchout of the lobe complex remain sand dominated, between 5 and 30 cm in thickness, and display some planar and ripple lamination. Across the Ongeluks River locality to the pinchout, the upper beds of Fan 3 remain thinner bedded than those below. Fan 4 also thins abruptly southward, although the mudstone between Fan 3 and 4 maintains a constant thickness (Oliveira et al., 2009). At the Ongeluks River locality (Fig. 2), Fan 3 is 65 m thick and is composed of clusters of sand-rich channel-fills, interpreted as base-of-slope channel complexes (Sullivan et al., 2000;Luthi et al., 2006) that incise lobe deposits (Hofstra et al., 2017). The channels are orientated dominantly towards the NE their Fig. 11) with variations to the N and E (Hodgetts et al., 2004). The palaeoslope feeding Fan 3 was NEfacing . The abrupt southeastward pinchout is interpreted to be due to lateral onlap, forming a sharp-based contact, onto a confining NE-SW-trending and NW-facing slope (Oliveira et al., 2009) in a proximal base-of-slope setting .

Injectites below Fan 3
Injectites exposed in the Bizansgat area of the Tanqua depocentre reported here occur in mudstones below Fan 3 ( Fig. 1b) in the most proximal exposures to the south of the outcrop belt (Fig. 2). The nature of the outcrop   and Tanqua depocentre, numbers 1-4 refer to Fans 1-4, whilst 5 refers to Unit 5, a 100 m thick channelized slope succession (Hodgson et al., 2011b). Location of injectites, studied in the present paper, denoted by asterisks. Ages from U-Pb zircon analysis of volcanic ashes (see Fildani et al., 2007;McKay et al., 2015) are displayed in boxes as Ma. (c) Tanqua depocentre study area. (d) Laingsburg depocentre study areas. means that the 3D geometry of the larger injectites exposed in the mudstone below Fan 3 can be constrained. Locally, a single main laterally extensive~1 m thick clastic sill steps up to the south and east to form a discordant relationship with the stratigraphy (Fig. 3a). Figure 4a and b shows the outcrop extent of the main stepped sill, which connects to at least three 0.4-0.6 m wide sub-vertical dykes that connect to the base of Fan 3 over a vertical distance of between 3 and 7 m. Steps on this sill are curvilinear along strike (Fig. 4), forming crescent-like geometries up to 200 m across and are no more than 1 m in vertical height. Propagating below the main sill are several thinner dykes (<0.2 m) that extend <6 m vertically, and bifurcate and taper out. Ridges that are orientated subhorizontally with the host strata ( Fig. 3c) mark the margins of these dykes. Margin structures on both the main stepped sill, and connecting dykes, include plumose patterns on fracture surfaces, parallel ridges, mudstone clastrich surfaces and planar surfaces ( Fig. 3b-f). The average strike of the steps is WNW-ESE, although there is a wide spread of orientations due to their curvilinear planform geometry (Fig. 4). Plumose features, observed on the margins of sills where they step through stratigraphy, form fan-like features with parallel striae down their centre and diverging striae away from the central axis (Fig. 4c). The direction of striae divergence is to the S, with a range from SW-SE. The dykes maintain a constant thickness at the scale of the outcrop, and are orientated N-S to NNE-SSW (Figs 3b and 4b).

Interpretation
All injectites studied in this area are close to the base of Fan 3 (Figs 3a and 4a), with sub-vertical dykes connecting Fan 3 with the large stepped sill. In the SE part of the outcrop, dykes directly connect the parent sand to the sill (Fig. 3), which supports local downward propagation (e.g. Von Brunn & Talbot, 1986;Rowe et al., 2002;Parize & Fri es, 2003;Le Heron & Etienne, 2005). The fine sand grain-size of the injectites is the same as Fan 3, and Fans 1 and 2 are not present in the underlying stratigraphy, which comprises several 100 0 s m of mudstone (King et al., 2009). Consequently, Fan 3 is interpreted as the parent sand for all the injectites.
The dykes are orientated approximately perpendicular to the NW-facing palaeoslope that confines Fan 3. Therefore the dyke orientation is hypothesized to relate to a gravitational stress regime. Although the injectites occur beneath the parent sand, the morphology of the curved steps and the orientation of structures on the injectite surfaces ( Fig. 4b) (plumose features indicate the propagation direction, Cobain et al., 2015) suggest that the main injectite sill stepped laterally outwards from its centre and cut up stratigraphy towards the south and east. The injectites, therefore, parallel the base of Fan 3 and continue beyond the depositional pinchout (Figs 3a and 4a). Net injection propagation direction was horizontal rather than vertical from the sharp-based sandbody with an abrupt upslope pinchout configuration in a lower slope to base-of-slope setting.

Zoutkloof; Laingsburg depocentre
Unit architecture Unit C of the Fort Brown Formation (Fig. 1b) has also been the focus of extensive study, and is subdivided into three subunits; C1, C2 and C3, each separated by a laterally extensive mudstone Flint et al., 2011;Hodgson et al., 2011a;van der Merwe et al., 2014). Extensive dip and strike outcrop control allow the distribution of sedimentary facies and architectural elements, and therefore depositional environments, to be constrained . Subunit C1 forms a 50 m thick lobe complex 8 km to the southeast (Fig. 5) where the overlying subunit C2 is thin-bedded and forms part of an external levee to a channel system van der Merwe et al., 2014). At the Zoutkloof locality, subunit C1 is sharp-based, thins from 2 m of amalgamated fine sandstone (Fig. 6c) to <12 cm thin bedded very fine sandstone over~1.5 km at the oblique up dip pinchout of the lobe complex (Fig. 6b). The confining palaeoslope at subunit C1 time, based on isopach thickness maps and palaeocurrents, was orientated N-S and E-facing Fig. 5). Locally, the base of C1 forms a sharp contact with the underlying mudstone, and the top surface is marked by the lower C mudstone that separates subunits C1 and C2  at a constant thickness of 0.9 m. This upper mudstone was used as a datum ( Fig. 6a and d).

Zoutkloof injectites
At Zoutkloof, injectites crop out over 1.7 km ( Fig. 6d-f) below subunit C1, in the upper 13 m of the 40 m thick regional mudstone that separates Units B and C (Brunt et al., 2013), at an abrupt, oblique lateral pinchout (Di Celma et al., 2011) (Figs 5 and 6d). At this locality, the main form of injection is stepped sills (Fig. 6f). Curved steps are no more than 2 m in vertical height and continue laterally for 10's m. Steps are closely spaced so that the sills are discordant with the host stratigraphy for more than 2-3 m. The majority of dyke margins exhibit ridges, both plumose and parallel (Cobain et al., 2015). Several sub-vertical dykes are observed to connect the base of subunit C1 with the stepped sills, the thickest is 1.5 m wide (between logs 7 and 8; Fig. 6a). Most other dykes are thinner (<0.3 m-thick) and connect with the base of subunit C1.
The steps and parallel ridges are primarily aligned E-W and the orientation of striae divergence of plumose patterns on the fracture surfaces is dominantly WSW (Fig. 6d). The dominant trend in dyke orientation measurements is NNW-SSE, approximately perpendicular to the orientation of the steps (Fig. 6). In the Zoutkloof area, all injectites are close to the base of subunit C1, at the NW margin of the sharp-based lobe complex, and vertical dykes connect large stepped sills with the base of subunit C1. Therefore, subunit C1 is interpreted to be the parent sand of the injectites. The main sills, fed by dykes sourced from the overlying parent  sand, abruptly step up stratigraphy to parallel the abrupt pinchout of the parent sand. Injection propagation is subparallel (WSW) to the unit pinchout direction and occurs where the base of parent sand has a sharp sand-to-mud contact. The orientation of the dykes is close to perpendicular to the slope-facing direction suggesting a causal relationship. The apparent propagation direction of subvertical dykes is downward but the ridges on the dyke margins suggest that propagation during injection was dominantly lateral (e.g. Kane, 2010).

Slagtersfontein; Laingsburg depocentre
Unit architecture C2 is the only subunit of Unit C present in the Slagtersfontein region of the depocentre. The tabular sandstones with intercalated hybrid beds support palaeogeographic and isopach maps that indicate the location to be at the edge of a lobe complex that thins abruptly to the south (  Hodgson et al., 2016), therefore this was chosen as a datum from which to hang the panel (Fig. 7a). Along the Slagtersfontein outcrop, subunit C2 is sand-prone, sharp-based, and thickens from 0 m at the western extent of the outcrop to >20 m thick downdip to the east over 1.5 km. Lower beds within subunit C2 are structureless and amalgamated sandstones, whereas the upper beds are thin bedded and laminated (Fig. 7b). Locally, the base of subunit C2 is erosional, and incises underlying mudstones to the east (e.g. Fig. 7b).

Slagtersfontein injectites
Injectites exposed in the Slagtersfontein area are hosted within the regional mudstone separating Units B and C. The majority of injectites at the Slagtersfontein outcrop are 0.1-0.6 m thick sills that extend laterally for up to 500 m. Dykes (0.1-0.5 m thick) are common near the base of subunit C2, and are observed to connect to the base of Unit C (Fig. 7b and c). Injectites crop out over the entire exposure length of Unit C, and for a further kilometre up dip where Unit C is absent in the mudstone separating Units B and D (Fig. 7a). Injectites in the mudstone that separates Units B and C are most abundant close to, and directly connect with, Unit C where the base is erosive and has a sharp contact between the Unit C sandstone and the underlying mudstone. Injectite margins are mostly planar, although some parallel ridges are present on dykes. Some smaller injectites, mainly <0.2 m thick sills, occur close to the base of, and are directly connected to, Unit D (Fig. 7a). The outcrop character at Slagtersfontein only permitted collection of dyke orientation data, the mean of which is NW-SE (Fig. 7).

Interpretation
Injectites connect directly with subunit C2, therefore this is interpreted to be the parent sandstone for the main injectite network, with Unit D likely acting as a minor source (see Fig. 7a; direct connection of 2 small dykes between logs 9 and 10). The underlying Unit B is topped with several metres of thin bedded silty strata, which is consequently less likely to produce sandstone injectites; there is also an absence of any dykes emanating from this unit, in outcrop. The parent sand is at an abrupt sandprone pinchout of a lobe complex (subunit C2) where locally the base is in erosive contact with underlying mudstones. The majority of clastic injectites are sills that extend laterally beyond the parent sand towards the west in cross-section (Fig. 7a). Therefore, the net propagation direction of injected sand was to the west and south, with injectites exploiting pre-existing bedding plane weaknesses (Cobain et al., 2015). The orientation of the dykes are sub-parallel to the local NNE-facing palaeoslope, which suggests a causal relationship, such as a gravity-driven stress regime.

Comparison of study areas
Previous research in the Karoo Basin (Wickens, 1994;Wickens & Bouma, 2000;Hodgson et al., 2006;Oliveira et al., 2009;Pr elat et al., 2009;Di Celma et al., 2011;Flint et al., 2011;Brunt et al., 2013;van der Merwe et al., 2014) means that the palaeogeographic context of the parent sandbodies to the studied injectite networks is extremely well constrained. The style and extent of outcrop means that it has been possible to collect data and geometries of injectite networks to provide 3D constraints over several kilometres. The Fan 3 and Unit C study sites were deposited in basin-floor environments Di Celma et al., 2011;Brunt Fig. 6. Zoutkloof outcrop and injectites. (a) Correlation panel of logs taken along length of outcrop (see Fig. 6d for location). (b) Subunit C1 is a 10 cm thick very fine sandstone. (c) Subunit C1 is >2 m thick, massive, fine sandstone. (d) Map view of outcrop with Subunit C1 highlighted, injectites and log locations indicated. (e) Oblique view, Unmanned Aerial Vehicle (UAV) based, photograph of clastic dykes and sills at eastern end of Zoutkloof exposure (see Fig. 6d for viewing direction). Subunit C1 is highlighted. (f) UAV photograph of Zoutkloof area showing the bowl-like structure of the injectite complex at the eastern end of the exposure (see Fig. 6d for viewing direction). Rose diagrams depict directional data for patterns on a fracture surface and step and dyke orientations. Refer to Fig. 4 for rose diagram colours. et al., 2013). Injectites sourced from Fan 3 in the Tanqua area, and subunits C1 and C2 in the Laingsburg area, coincide with sites of abrupt basin-floor sandprone pinchout, with mudstone above and below. Additionally, the basal contact of the parent sand with the underlying mud is erosional and/or sharp where injection occurs. The injectites propagated laterally paralleling the base of the parent sandbody, and extend beyond the pinchout, and dykes are sub-parallel to the strike of the palaeoslope in all examples. Furthermore, the extensive previous research in the field area also helps to constrain where injectites are not present, meaning models are not biased towards outcrops that only show injectites. For example, detailed mapping and coring of the fringes of lobe complexes (Johnson et al., 2001;van der Werff & Johnson, 2003;Hodgson et al., 2006;Pr elat et al., 2009) has identified only rare isolated injectites associated with Fan 1 and Fan 4.

DISCUSSION Injectite emplacement in the Karoo Basin: mechanisms and controls
We have presented three examples of basin-floor lobe complex pinchouts that have been subject to post-depositional fluidization of the parent sandbody and clastic injection into the surrounding mudstone. Discussion on emplacement takes into account the common features observed across all outcrop examples described here, the well-constrained architecture and palaeogeography of each of the units, and the prerequisite conditions needed for clastic injection.

Conditions prior to injection
Typically, the same conditions observed to form overpressured and uncemented sand liable to fluidization in slope channel-fills are also met in these examples from basin-floor lobe complexes: (i) proximal deposits within the lobe complexes provide clean, fine to very fine sand (e.g. Marchand et al., 2015) that increases the likelihood of fluidization, and hence susceptibility for sediment transport (Richardson, 1971;Jolly & Lonergan, 2002); and (ii) the deep-marine environment and regional changes in clastic sediment supply allow for alternating sand-rich channel-fed lobe complexes encased by regional hemipelagic mudstone drapes that provide the seal required for overpressure to develop (Lorenz et al., 1991;Cosgrove, 2001;Jolly & Lonergan, 2002). These surrounding mudstones may also provide an additional source of pore fluids during the initial stages of compaction (Magara, 1981).

Geographic location and parent sandstone architecture
Based on the outcrop positions of the observed injectites, and the existing palaeogeographic knowledge of the Karoo Basin (Wickens, 1994;Wickens & Bouma, 2000;Hodgson et al., 2006;Oliveira et al., 2009;Pr elat et al., 2009;Di Celma et al., 2011;Flint et al., 2011;Brunt et al., 2013;van der Merwe et al., 2014;Hofstra et al., 2017), the injectites are interpreted to be located at the abrupt pinch-out of sand-rich lobe complexes (Figs 3 and 5). At their abrupt updip pinchout, such as Bizansgat (Fan 3) and Zoutkloof (subunit C1) the parent sand is generally homogenous, well-sorted, and has a sharp contact with the underlying strata. The same configuration occurs in the abrupt lateral pinchout at Slagtersfontein (subunit C2). Clastic injectites occur stratigraphically beneath the parent sandstone, with net lateral propagation towards and beyond the margin of the parent sandstone lobe complex. In other examples, where injectites of seismic-scale are known to be sourced from lobe complexes (as observed in intra-slope lobes), the source point is the proximal lobe (complex) fringe (Yang & Kim, 2014;Spychala et al., 2015), or the lateral lobe margin pinchout (Monnier et al., 2014). In the latter case, the lobe reaches its highest point laterally. This suggests that an abrupt and sand-prone pinchout in the most elevated position on the lobe, which will typically occur in the proximal or lateral parts of lobes, is a preferential site for clastic injection processes.

Nature of stratigraphic contact
Considering the geographic and stratigraphic distribution of the required unconsolidated sandstone and the surrounding fine grained sediments, injectites might be expected at all positions within lobe complexes. As long as sand remains unconsolidated, the surrounding hemipelagic mud may form a seal around the entire unit. The observation of preferential hydraulic fracture at a sharp sand-to-mud contact, with clean sands, however, favours the proximal area of lobe complexes at their base. In these situations, erosional relationships and/or steeper slopes promote a more abrupt onlap geometry and the formation of a sharp basal contact from where the injectites are sourced. Commonly, the upper part of lobe complexes are thin-bedded (e.g. Hodgson et al., 2006;Pr elat et al., 2009), and in such cases injectites are absent. In the presence of subtle confinement (Sixsmith et al., 2004), or in more distal settings (van der Werff & Johnson, 2003;Hodgson, 2009;Pr elat et al., 2009;Spychala et al., 2017), injectites are not observed. However, in a few cases where there is an abrupt sand-to-mud contact on top of a lobe complex, due to large-scale avulsion or sudden clastic input shutdown, injectites are observed (e.g. Subunit A5; Cobain et al., 2015). Where clastic material is finer and/or less well-sorted, clastic injection is not observed. What mechanism controls this preferential occurrence of injectites at the interface between clean sands and muds? A key attribute of clean sands is a tighter grain-size and shape distribution, and therefore higher permeability relative to less clean sands (Krumbein & Monk, 1942;Beard & Weyl, 1973). Transient changes in pressures related to variations in grain-size, and thus permeability, might be expected to influence the position of hydraulic fracturing. However, cyclic loading of sands in closed systems demonstrates that lower permeability sands exhibit higher transient pressures (e.g., Kelly et al., 2006). Consequently, variations in permeability do not appear to be the controlling mechanism. Furthermore, if aseismic, overpressure builds more gradually over geological time, and therefore the pressure at the sand-mud boundary  Jackson et al., 2011) (note that injectites in this setting may be more broad ranging), whereas this study reports examples from basin-floor lobe complexes. Injectites occur in areas where sand is steeply confined and/or proximal within the lobe complex, while palaeogeographic locations that are downdip exhibit subtle confinement or have less cleansand for fluidization and therefore do not produce injectites. may be similar at all points. In contrast, clean sands are more susceptible to fluidization (Richardson, 1971;Jolly & Lonergan, 2002), and consequently they may preferentially fill any hydraulic fractures that occur.
A substantial depth of burial prior to sand injection in the Karoo Basin examples examined herein consists of a number of lines of evidence, including the preservation of initial brittle, hydraulic patterns on fracture surfaces on the margins of injectites seen at the Zoutkloof and Bizansgat localities (Fig. 3d and e). These suggest that the muds were sufficiently hard to form and maintain these surface patterns; no evidence for later compaction of these surface patterns on dyke margins is observed (Cobain et al., 2015). Furthermore, the observed injectites show features (vertical distribution of particles within sills; lack of erosion) commensurate with high-concentration, laminar flow conditions, suggesting that the units were sufficiently far from the contemporaneous seabed that breakthrough and subsequent extrusion did not occur; such open-conduit conditions are linked to turbulent flow conditions (Cobain et al., 2015). However, despite the evidence given above that injection did not occur at very shallow depths, the fact that the injection occurred along extensional fractures, places a constraint on the depth at which they formed. The conditions for the formation of extensional fracture is that the differential stress should be less than 4 times the tensile strength (T) of the rock, (i.e. (r1Àr3) < 4T), (see e.g. Cosgrove, 2001). The differential stress increases with depths and extensional fractures can only form above the depth where (r1Àr3) = 4T. It is suggested that in the study area this depth was around several hundred metres. There is a notable absence of overlying slides and slumps, and the absence of growth strata above seabed folds and faults in the basin-fill (e.g. Hodgson et al., 2006;Di Celma et al., 2011;Flint et al., 2011;Jones et al., 2015) indicate it was largely tectonically quiescent. Therefore, fluidization and injection due to localized excess pore fluid pressures generated by depositional processes such as mass flows (Truswell, 1972;Jolly & Lonergan, 2002) and shallow seismicity (Obermeier, 1996;Lunina & Gladkov, 2015), in these outcrop examples, are considered unlikely trigger mechanisms.
Disequilibrium compaction is a major source of overpressure in sedimentary basins (Osborne & Swarbrick, 1997), however within a single body or unit, this overpressure will dissipate over geological time, and high overpressures can only be maintained in the shallow subsurface through high rates of sedimentation (Jonk, 2010). Therefore, disequilibrium compaction alone may not be an adequate source of overpressure to trigger clastic injectites. Overpressure due to fluid volume increase is associated with aquathermal expansion and clay dehydration, though these alone are considered too insignificant to generate high amounts of overpressure (Osborne & Swarbrick, 1997). Deep or regional seismicity has been commonly cited as a primary cause of sand intrusion, however, the energy required to fluidize and inject such quantities of sand in regionally extensive injectites likely exceeds that produced by earthquakes (Huuse et al., 2005;Duranti, 2007;Vigorito & Hurst, 2010). If such regional seismicity were a cause, then hydraulic fracturing, failure of encasing mudstone, and resultant injection would be expected across the entire lobe complex. Additionally, an absence of seismicity for a significant period would be needed in order to bury the sediments to depth and enable overpressure to build; consequently, a large- Fig. 10. Simplified map view illustrations of the orientation of parent sand pinchout and injectites at the three study sites, (a) Bizansgat (Fan 3), (b) Zoutkloof (subunit C1) and (c) Slagtersfontein (subunit C2). The yellow marks the parent sand, the grey is the underlying mudstone. The red lines are dykes, using mean orientation. The blue arrows show the mean direction for flow of the intrusions where recorded. Note that the dykes are sub-parallel to the pinchout of the sandbody (approximately perpendicular to the onlap slope) and that the dominant flow direction is at a high angle to the pinchout. scale change in tectonic regime would be required. Regional seismicity, therefore, is considered an unlikely trigger of injection for these deeper injectites (Duranti, 2007;Hurst et al., 2011).
Another mechanism for triggering injection in deepwater systems is the migration of fluids caused by lateral pressure transfer: the lateral transfer of fluids from deeper, overpressured parts of the basin along laterally extensive, inclined, porous units (Osborne & Swarbrick, 1997;Yardley & Swarbrick, 2000). The lower parts of the basin-fill are likely to experience enhanced overpressure as a result of compaction, and thus cause movement of fluid upwards towards the highest point. This form of fluid migration is most likely to be concentrated at the up dip margins of a unit (Cartwright, 2010), such as a lobe complex margin, where the abrupt pinchout architecture at the fringe of lobe complexes promotes fluid migration towards the edge (Monnier et al., 2014). The surrounding mud limits further fluid migration. Migration of fluids due to lateral pressure transfer operates in basins such as the Gulf of Mexico, where simple tilting causes a pressure gradient (Flemings et al., 2002;Gay et al., 2011). Lateral pressure transfer is interpreted to be the likely cause of post-Eocene intrusions along the margin of the San Joaquin Basin (Schwartz et al., 2003;Cartwright, 2010). In the San Joaquin Basin, the fluids that produce overpressure and cause lateral pressure transfer are not derived locally. Migrating hydrocarbons may also cause an increased pore pressure in sand units sealed by impermeable strata (Jolly & Lonergan, 2002). Consequently, increased overpressure of an unconsolidated sand body by compaction driven fluid expulsion, and fluid migration through lateral pressure transfer (water, oil, gas), is the preferred trigger mechanism responsible for clastic injection in the Karoo Basin (see also Cobain et al., 2015). The parent sand architecture in all examples promotes lateral fluid migration to the updip lobe complex margins. Larger-scale injectites have also been attributed to this kind of trigger (Huuse et al., 2005;Hurst et al., 2011;Løseth et al., 2013).

An integrated model of injectites in basinfloor lobes
Synthesizing the observations discussed previously enables an integrated model of injectites in basin-floor lobes to be proposed. Injectites are observed to form preferentially at the updip margins of basin-floor lobe complexes (Bizansgat Fan 3 and Zoutkloof subunit C1) and on lateral margins where the pinchout is abrupt and sandprone (Slagtersfontein subunit C2) (Fig. 8). This geographic distribution is linked to the nature of the triggering mechanisms. The presence of patterns on fracture surfaces, the absence of significant compaction of these structures, and the evidence for confined laminar flow, suggest that these injectites formed at substantial depths, but the extensional nature of fracturing indicates a maximum depth of no more than a few hundred metres.
Consequently, disequilibrium compaction and lateral pressure transfer are the likely trigger mechanisms, and in the case of a lobe complex deposited above a basinal slope, these mechanisms will lead to updip fluid migration. Furthermore, in a tilted sandbody the confining lithostatic pressure will also decrease updip. Therefore, hydraulic fracturing will predominantly occur at the updip margin where fluid migration and the lowest confining pressures combine. Within the proximal lobe complex, injectites are shown to occur at pinchouts (Figs 8 and 9); these areas both concentrate fluid-flow from lateral transfer and provide sharp boundaries at their basal surfaces between clean sands and the underlying mudstones. We argue that initiation of hydraulic fracturing is favoured at the bases of these pinchouts because these clean sands are the most susceptible to fluidization (Richardson, 1971;Jolly & Lonergan, 2002) and therefore will preferentially infill any hydraulic fractures that occur. Theoretically, hydraulic fracturing might be expected to occur on the upper surface of the most updip point, as shown in some examples (Cobain et al., 2015), but in many cases proximal parts of lobes exhibit a transition towards lower permeability facies (e.g., thinner bedded siltstones and sandstones) at their tops (Fig. 8;Pr elat et al., 2009). The distal parts of basinfloor lobes are not favoured sites for injection as a consequence of their down-dip position, and their more heterogeneous, mud-rich, facies including thin-bedded silts and sands, and hybrid beds (Fig. 8;Hodgson, 2009;Pr elat et al., 2009;Marchand et al., 2015;Spychala et al., 2017). Whilst the physical linkage between sills and the parent sands suggests that the initial hydraulic fracturing and injection can be downwards, the increasing lithostatic pressure below the parent sands will encourage lateral propagation with sands able to step beyond the lobe complex margins (Figs 8 and 9). This is supported by the direction of injection flow being at a high angle to the orientation of sand pinchout (Fig. 10).
The dykes at all three study sites are aligned sub-parallel to the strike of the palaeoslope (Fig. 10), which suggests that a controlling factor in injectite morphology is the orientation of the slope onto which the lobes onlap. Tensile features would preferentially develop perpendicular to slope facing direction in a gravitational stress field, leading to a narrow range of dyke orientations after injection was triggered. This would provide the necessary anisotropy for the documented preferred direction. In contrast, several studies have found limited to no relationship between injectite orientation and palaeoslope (Hiscott, 1979;Rowe et al., 2002;Diggs, 2007;Jackson, 2007;V etel & Cartwright, 2010;Bain & Hubbard, 2016;Palladino et al., 2016), and ascribe measured orientations to later tectonic controls (e.g. Diggs, 2007;V etel & Cartwright, 2010;Palladino et al., 2016), or in association with submarine channel orientation (e.g. Jackson, 2007) and/ or the emplacement direction of mass transport emplacement (Hiscott, 1979;Rowe et al., 2002). However, here we demonstrate that for injectites sourced from lobe complexes in tectonically quiescent basins, palaeoslope can be a controlling factor on injectite orientations.

Stages of fluid flow associated with injectites
Understanding fluid flow through time in sedimentary basin-fills is essential when considering aquifers and hydrocarbon reservoirs. In large-scale cases, injectites can promote basin-wide fluid flow and offer vertical and lateral permeable networks through low permeability successions (Huuse et al., 2005;Vigorito et al., 2008;Jonk, 2010;Hurst et al., 2011). Four main elements of basinwide fluid flow are identified (Jonk et al., 2005a): (i) gravity-driven, downward flow of meteoric water (Bjørlykke, 1993), (ii) compaction of sediments through burial causes fluids to be expulsed and flow upwards (Osborne & Swarbrick, 1997), (iii) upward flow of fluids through overpressure (Osborne & Swarbrick, 1997), and (iv) upward migration of hydrocarbons due to buoyancy (Bonham, 1980). Clastic injectites are associated with basinal fluid flow at several stages; pre-injection, during the process of clastic injection, post-injection and pre-cementation, and post-cementation (Fig. 9).

Pre-clastic injection
The migration of fluids as a trigger for clastic injectites through lateral pressure transfer has already been discussed; a schematic representation of the processes is shown in Fig. 9b.

During injection
During clastic injection, grains are suspended and transported down a pressure gradient, by fluids moving from the overpressured parent unit towards the tip of the propagating hydraulic fracture, a source of relatively lower pressure (Cosgrove, 2001). The flow regime during injection can be turbulent (Hubbard et al., 2007;Scott et al., 2009;Hurst et al., 2011) or laminar (Duranti, 2007;Cobain et al., 2015) (Fig. 9c).

Post-injection, pre-cementation
In previous studies, petroleum inclusions in late diagenetic cementation phases, and multiple cementation phases, indicate that injectites can act as long-lived fluid flow conduits (Jonk et al., 2005b(Jonk et al., , c, 2007Ross et al., 2014). Injectites can act as fluid flow conduits up to depths of approximately 1 km (Jonk et al., 2005a;Jonk, 2010) prior to cementation. However, thicker sandstones (i.e. 20-30 m) can remain uncemented up to depths of 1.5-2 km burial, for example those within the Tertiary of the Northern North Sea (Lonergan et al., 2000;Duranti et al., 2002). Additionally, many of the large-scale injectite networks in the Tertiary of the North Sea have maintained excellent reservoir properties (Hurst & Cartwright, 2007) and outcrop examples such as the Panoche Giant Injection Complex have been shown through fluid inclusion analysis to have maintained migration of fluid for almost 2 Ma post injection (Minisini & Schwartz, 2007;Hurst et al., 2011). Besides acting as fluid migration pathways, clastic injectites can connect otherwise separate reservoirs, and form traps when injected solely into-or capped by-impermeable strata (Frey-Mart ınez et al., 2007;Schwab et al., 2015).

Post-cementation
When cemented, injectites become fluid flow barriers, preventing any further migration of basinal fluids. However, cemented injectites also have the potential to act as conduits, through structural deformation in the form of fractures focussed on the competent sands within lowcompetence mudrock host lithology (Jonk et al., 2005a) (Fig. 9d). Understanding the timing of deformation phases helps to determine if clastic injectites will be reactivated as fluid flow conduits.

Implications for hydrocarbon extraction
Is there an association of stratigraphic traps and clastic injectites?
Each outcrop locality presented herein is an example of a basin-floor lobe complex that has been subject to clastic injection at its abrupt proximal (Bizansgat, Zoutkloof) or lateral (Slagtersfontein) pinchout. In each case, injectites are fed from the sharp sand-to-mud contact that marks the base of a lobe complex, they then parallel the base of the depositional body, stepping upwards and outwards (e.g. Figs 3a and 7a), ultimately projecting beyond the limit of the lobe complex. The clastic injectites produced are of sub-seismic scale.
Sandy lobe complexes such as those described have been a prime target for hydrocarbon exploration as stratigraphic traps (e.g., Halbouty, 1966;Walker, 1978;Brown et al., 1995;Gardiner, 2006;Stoker et al., 2006;Nagatomo & Archer, 2015). In particular, proximal turbidites on the basin floor as they provide clean sands that pinch out abruptly, providing an optimal trap configuration. We have shown that these sands are prone to injection, particularly on a sub-seismic scale. In addition, dykes can have a strong preferential orientation at abrupt pinchout of lobe complexes against confining slopes, and that injection flow will be towards, and beyond, sand pinchout. This helps to constrain the architecture and prediction of injectite networks at stratigraphic traps on the basin-floor. The presence of clastic injectites at stratigraphic traps can be beneficial; they can provide connection between otherwise separated sand units, allowing flow of hydrocarbons through impermeable mudrocks, and balancing pressure differences across reservoir complexes. However, the complicated geometry of injectites and their potential to connect otherwise separate sand bodies needs to be taken into consideration when building reservoir models and when using outcrops as analogues for geological and petrophysical model development.

Are basin-floor lobe injectites under-reported?
The relative lack of documented examples of injectites associated with lobe complexes compared to submarine slope channel-fills may simply be due to less of these systems being drilled and therefore a data bias. However, this disparity is also likely a reflection of scale. Parent sands of the injectites described here are volumetrically larger than many slope channel-fills, but comprise much thinner lobe complexes. Therefore, as observed in the Karoo Basin outcrops, thinner injectites can be expected as a product of remobilization in comparison to slope channel-fills, thus being sub-seismic scale and frequently unrecognized or poorly documented on many seismic data sets (e.g. Shepherd et al., 1990). Another factor contributing to the lack of recognition in subsurface data is the style of injection; Karoo injectites are primarily laterally extensive sills. These would be hard to identify in reflection seismic data, and misinterpretation as primary deposits rather than remobilized units in core is possible.

CONCLUSIONS
The majority of injectites are reported as being sourced from channel-fills or intraslope lobes in submarine slope settings, and have been rarely documented in base-ofslope and basin-floor environments. The three outcrop examples of clastic injectites presented here are associated with basin-floor environments, and specifically occur at the abrupt pinchouts of basin-floor lobe complexes. Architecture and bed-scale similarities across the injectite parent sand have led to the development of a model to help predict likely areas and orientations of clastic injectites in a deep marine system. Injectites occur where sand is: (i) confined and pinches out abruptly, (ii) proximal within the lobe complex, and (iii) exhibits sharp contacts with underlying and/or overlying mudstone. In contrast, palaeogeographic locations that exhibit subtle to no confinement have less clean-sand for fluidization, and heterolithic stratigraphic boundaries do not result in injectites. Clastic injectites, even those of a sub-seismic scale, provide the potential to rearrange fluid flow pathways within deep-water successions. Injectites, such as those in the Karoo Basin, can extend laterally for several kilometres, and beyond the stratigraphic pinchout, yet are too thin to be resolved in seismic data. However they may connect otherwise separate bodies of sand or reservoirs, offering highly permeable networks through impermeable successions. The association of clastic injectites and stratigraphic traps can be beneficial in subsurface plays. This is because they provide connection between otherwise separate sand units, allowing flow of hydrocarbons through impermeable mudstones, and balancing pressure differences across reservoirs. In the Karoo Basin, we see clastic injection and therefore the potential for fluid flow in basin floor settings, where, up until now, injectites and associated fluid flow have dominantly been associated with channelized slope environments.