The bases of active submarine channels are marked by large erosional features, such as knickpoints and plunge pools. However, their presence in ancient channel-fills has rarely been documented, so their importance in submarine channel morphodynamics requires investigation. Using seismic reflection data calibrated by wells from a buried submarine channel-fill, we document erosional features hundreds of metres long and tens of metres deep, here interpreted as knickpoints and a plunge pool, and provide a mechanistic process for their transfer into the stratigraphic record. Channel incision patterns are interpreted to record a transient uplift in an otherwise subsiding depocentre. Local structural complexities in the channel slope formed zones of preferential scouring. A switch to a depositional regime preserved the irregular channel base, inhibiting both the upstream migration of scours and smoothing of the channel base. The formation and preservation of these scours record the responses to salt tectonics and provide a unique snapshot of the formative processes of an ancient submarine channel. The presence of these exceptional basal scours indicates that headward erosion processes did not operate rapidly, challenging the paradigm that knickpoint migration controls channel evolution. Our results show that the primary erosion of the main channel surface and long-term channel evolution are both dominated by far more gradual processes.

Submarine channels are primary long-term conduits for the delivery of terrigenous and anthropogenic particulates across continental slopes to the deep ocean floor (Carter 1988; Hubbard et al. 2014; Kane and Clare 2019). Their longitudinal profile and stratigraphic record are interpreted to reflect variations in slope gradient and flow properties (Pirmez et al. 2000; Kneller 2003; Ferry et al. 2005; Covault et al. 2011). However, intrabasinal factors – such as tectonics (e.g. Gamberi and Marani 2007; Heiniö and Davies 2007; Georgiopoulou and Cartwright 2013; Stright et al. 2017; Mitchell et al. 2021), mass transport deposits (Tek et al. 2021; Allen et al. 2022) or autogenic channel processes (Sylvester and Covault 2016; Heijnen et al. 2020; Guiastrennec-Faugas et al. 2021) – can perturb the profiles of submarine channels and induce knickpoint development. The abrupt changes in channel floor gradient associated with knickpoints (Gardner 1983; Heijnen et al. 2020) can increase flow velocity and turbulence, leading to hydraulic jumps (Komar 1971), enhanced channel base scouring (Sumner et al. 2013; Dorrell et al. 2016) and, eventually, the formation of plunge pools (Gardner et al. 2020; Guiastrennec-Faugas et al. 2020, 2021). These erosional features produce an uneven channel floor topography in modern deep water environments (e.g. Bourget et al. 2011; Dalla Valle and Gamberi 2011; Gardner et al. 2020; Guiastrennec-Faugas et al. 2020, 2021; Mitchell et al. 2021), which provides a snapshot of the formative processes of submarine channels (e.g. Heijnen et al. 2020).

Geomorphic features associated with abrupt variations in channel gradient (i.e. knickpoints, plunge pools and crescentic bedforms) are considered to be highly dynamic and ephemeral (e.g. Heiniö and Davies 2007; Hage et al. 2018; Guiastrennec-Faugas et al. 2020, 2021; Heijnen et al. 2020; Chen et al. 2021; Tek et al. 2021). Time-lapse seafloor mapping of active modern systems with sandy surficial sediments has documented headward erosion rates greater than hundreds of metres per year (Guiastrennec-Faugas et al. 2020, 2021; Heijnen et al. 2020). Consequently, knickpoint migration has been argued to be the key driver of channel evolution in many submarine channel systems (Heijnen et al. 2020) rather than sediment redistribution by 3D flow fields related to channel curvature (e.g. Peakall et al. 2007; Peakall and Sumner 2015; Morris et al. 2022).

Through time, submarine channel profiles tend towards a smooth graded profile, which balances flow capacity and sedimentary load (e.g. Pirmez et al. 2000; Kneller 2003; Gerber et al. 2009; Covault et al. 2011; Pettinga and Jobe 2020). Consequently, the thicknesses of preserved submarine channel deposits are constant longitudinally over one to tens of kilometres (Di Celma et al. 2011; Macauley and Hubbard 2013; Jobe et al. 2020) and relatively smooth profiles are assumed in numerical models (e.g. McHargue et al. 2011; Sylvester et al. 2011). A rare example of preserved scours has been documented in a buried Quaternary submarine channel-fill; these scours are <10 m deep and 300 m to 1 km long (Snedden 2013). However, the detailed processes and implications of such remnant scours have not been examined.

Here, we present a unique example of a buried submarine channel-fill that evolved above a dynamic stepped slope during the Oligocene–Miocene transition in the Campos Basin, offshore Brazil. The channel-fill preserves extreme longitudinal variations in basal erosion patterns (tens of metres deep, hundreds of metres long) formed above a cohesive substrate. Seismic mapping and well data analysis were carried out to: (1) assess channel incision and thickness patterns and their relationship with the underlying deposits; (2) examine the channel-fill sedimentology, architecture and stratigraphy; (3) propose mechanisms for the generation of local increased channel accommodation, ultimately recorded in exceptionally large and spatially variable channel thicknesses; and (4) explain the generation and preservation of the incisional basal channel surface from an evolutionary perspective. We propose that the interaction of sediment gravity flows with a seafloor topography deformed by salt controlled the development and preservation of channel incision patterns.

The Campos Basin, offshore Brazil (Fig. 1) evolved from a rift during the break-up of Gondwana (Jurassic–Early Cretaceous) to a marine passive margin from the Albian to the present day (Chang et al. 1992; Fetter 2009). We focus on a submarine channel-fill, informally named the Marlim Sul Channel (MSC; Fig. 1) (channel C2 of Casagrande et al. 2022), which incised sand-prone channel and lobe complexes deposited on a 40 km long stepped slope during the Oligocene–Miocene transition (the Marlim Unit of Casagrande et al. 2022). The stepped slope configuration was controlled by extensional salt tectonics and the associated fragmentation of the Albian–Cenomanian carbonate-prone interval (raft tectonics) that overlies Aptian evaporites (Casagrande et al. 2022; Fig. 1a). Through calibration with well data, it can be seen that the MSC-fill passes down-dip from mud-prone deposits with low root-mean-square (RMS) amplitude values to sand-prone deposits with moderate to high RMS amplitude values (Casagrande et al. 2022). Here, we investigate the sand-prone channel-fill sector.

We used two high-resolution 3D pre-stack time migrated seismic reflection datasets to document the basal surface and fill of the channel. The acquisition parameters were: (1) bin sizes of 6.25 m for cross-lines and 12.5 m for in-lines with 2 ms vertical sampling (approximate vertical resolution 18 m) and (2) bin sizes of 12.5 m for in-lines and cross-lines with 4 ms vertical sampling (approximate vertical resolution 22 m). Both volumes were converted to the depth domain. Seismic reflection data were processed to the zero-phase wavelet and are shown with SEG normal polarity. The Marlim Unit top is a trough with a negative reflection coefficient that indicates a decrease in acoustic impedance, whereas the base is a peak. Seismic interpretation was carried out using automatic and manual reflection tracking and was calibrated by 24 wells with basic wireline logs (gamma ray, density, neutron, sonic and resistivity logs). Eight wells (two cored) intersect the MSC-fill. The seismic geomorphology was interpreted in an RMS amplitude map. Real thicknesses were obtained from a map of the top and base seismic horizons using the higher resolution depth-converted seismic volume.

Longitudinal thickness variations of the channel-fill and the lobate features (LB and LC, Fig. 1d) were assessed by several measurements extracted from the thickness map and plotted graphically (Fig. 2). Within the channel-fill, the measurements were spaced every 400 m and were extracted at the thickest points of the channel cross-section. Twenty-six measurements were taken above the step; the same number were obtained in the lobate features. The 400 m distance was also used to plot the graph for LB and LC to match with the channel measurements and facilitate comparison between thicknesses. Thickness values in the lobes were extracted between 500 and 1000 m lateral to the MSC. The channel-fill stratigraphy was interpreted in a correlation panel along the MSC-fill based on well and 3D seismic reflection data. The panel is hung from a laminated mudstone package observed in all the wells.

The MSC is readily observed in maps and profiles from seismic reflection data. The sand-prone channel sector is divided into three segments (corridor, step and ramp) based on the thickness patterns extracted between the mapped top and basal surfaces (reported as average, minimum and maximum depths and the standard deviation (SD)), the morphology and the topographic configuration (Fig. 1d) (Casagrande et al. 2022, their fig. 11). Abrupt longitudinal changes in channel-fill thicknesses are observed at three sites above the step (T1, T2 and T3) where the channel incises two sand-prone lobate features (LB and LC) (Fig. 1c).

The channel has very low sinuosity and a constant width (300–350 m) in the corridor segment (Fig. 1d). The average thickness is 56 m (41–75 m, 8 m SD). The channel cuts partially preserved linear high-amplitude anomalies, which well calibration show to be sand-prone and 30 m thick. In the step segment, a salt-rooted synthetic normal fault, which cuts a carbonate raft, trends perpendicular to the channel course. The increased thickness and architectural changes across the fault support the interpretation of a break in slope between the corridor and step at the seafloor during the evolution of the Marlim Unit (Casagrande et al. 2022, fault marked in Fig. 1d and shown in map view and seismic profile in Fig. 3a, b). The average channel-fill thickness above the step is 52 m (26–86 m, 14 m SD). There are marked changes in the channel-fill thickness documented at bends T1, T2 and T3 (Figs 1d, 2), which share similar thickness maxima and downstream lengths, but different characteristics and relationships with the adjacent structural and stratigraphic elements. The channel width varies from 250 to 670 m and is wider at bends T1 and T2 (Fig. 1d). The channel is weakly sinuous across the step, apart from one pronounced bend at T3 (Fig. 1d). In the ramp segment, the distal part of the MSC has a constant width (350 m wide) and is thinner than on the step and corridor (average 18 m, 11–36 m, 7.3 SD) (Figs 1d, 2). The well–seismic calibration in the study area indicates that moderate to high seismic amplitudes relate to a sand-prone fill (Casagrande et al. 2022).

In summary, the average channel thickness data indicate that the channel-fill is thicker above the corridor and lower gradient step relative to the higher gradient ramp (Fig. 2). There are three prominent areas of anomalously thick channel-fills (T1, T2 and T3), which are examined in more detail in the following text.

Relationship between the Marlim Sul Channel and older deposits

The break in slope at the step favoured the formation of the LB and LC distributary channel and lobe complexes (Casagrande et al. 2022). Low sinuosity high-amplitude seismic features in the corridor connect with LB and LC at the entrance of the step and are interpreted as feeder channels (Fig. 1d, white arrows). The MSC incised the corridor and step deposits between the feeder channels and the lobe complexes (Fig. 1c). Seismic mapping indicates truncation of the reflections below the Marlim Unit, which is supported by the thinner average lobe thicknesses of LB (c. 20 m) and LC (c. 60 m) compared with the thicknesses of T2 (Figs 2, 4). These observations support the complete removal of LB and LC beneath the MSC (Fig. 1c) and deep incision below the lobes. A few metres of cores in some wells indicate that the rocks underlying the lobes and the channel-fill are laminated and bioturbated mudstones (Fig. 5e).

Calibration with well logs in the cored wells and the well log patterns in several other wells show that the stratigraphy underlying the channel-fill and the lobes has a higher gamma ray response and density than the sand-prone deposits of LB and LC and the channel-fill itself (Fig. 6b) and a density and neutron logs cross-over pattern (density log to the right and neutron to the left) suggesting non-reservoir facies. These mudstones form most of the deposits that were eroded by channel incision. There is a broad trend of decreasing channel thickness along the channel from the corridor to the ramp (Fig. 2). The magnitude of this decrease is 0.11° along the channel path (25 km) or 0.17° if we consider a straight measured distance of 17 km (Fig. 2). T1–T3 punctuate this trend (Fig. 2). Notably, there is no marked change in channel thickness across the step-to-ramp transition.

Anomalously thick areas (T1–T3)

The three anomalously thick channel-fill sites (T1–T3; Figs 1, 2, 4) are all located on ‘clockwise’ orientated meanders on the step and they have differing relationships to the faults and adjacent deposits (LB and LC).

T1 is 1 km long and 350 m wide and initially thickens downstream from 45 to 75 m over 350 m (average basal slope gradient c. 5°; Fig. 4b1) and then thins from 75 to 45 m over 650 m (average basal slope gradient 2.6°). The thickness patterns support T1 as a longitudinally asymmetrical feature (Fig. 4b1). At T1, the channel around the bend is on average 150 m wider (maximum width 550 m) than the corridor up-dip and also the next bend down-dip. T1 also coincides with the break in slope at the up-dip edge of the step (Figs 3, 4a, b1). The synthetic normal fault associated with the slope break is interpreted to have influenced the formation of T1 and minor outboard faults do not show any evidence of movement within the Marlim Unit (Figs 3, 4b2). Similar increases in thickness are observed down-dip of the break in slope in the adjacent lobate features LB and LC (Fig. 4a), suggesting that the lobes and channel thicknesses record the same local accommodation control.

T2 is in the central part of the step, where the MSC thickens from 40 to 86 m over a downstream distance of 1 km (c. 2.6° average gradient), then abruptly thins downstream to 32 m over 400 m (Figs 2, 4c1), forming an average counter slope of c. 8°. T2 also coincides with a wider (c. 650 m) and more sinuous channel segment compared with the immediately downstream channel bend that turns anticlockwise (c. 250 m) (Fig. 4a). The zone of thickest deposits is 1.4 km long and a maximum of 350 m wide (Fig. 4c1). Highly irregular channel edges are observed at the contact with LC (Fig. 4a, black arrow). On the outer bend, T2 is partially bounded by a NW–SE-trending normal fault (Figs 3c, 4a, c2). At depth, the outer bend fault is salt-rooted and coincides with the edges of an underlying carbonate raft (Fig. 3c) and is mappable to the base of T2. The MSC reflections within T2 are rotated and thicken towards this fault (Fig. 4c2). The rotation and thickening of the reflections towards the salt-rooted fault suggests that it was active during the evolution of the MSC and induced a local depression during channel incision. The geometry of the outer bend supports the view that the fault formed a synthetic listric fault segment as it propagated and widened towards the contemporaneous seafloor, as has been documented in exhumed systems (e.g. Hodgson and Haughton 2004; Baudouy et al. 2021).

T3 is located on a tight bend in the MSC, 1.5 km up-dip from the step-ramp slope break, and thickens downstream over 1 km from 45 to 75 m (average gradient c. 1.5°; Fig. 4a, d1), then thins from 75 to 40 m over 530 m (average gradient c. 4°). The maximum width and length are 280 m and 1.6 km, respectively. The area of thickening is located at the centre line of the channel, not towards the outer bend. Overall, the bend displays similar upstream and downstream widths (c. 400 m). T3 is intersected by the same fault system present at T2 (Fig. 4a). The faults are parallel to the main channel orientation and have a complex configuration that forms a graben, with T3 parallel to one of the bounding faults of the graben (Fig. 4d2). Thickening of the reflections above the Marlim Unit in the central low block of the graben (Fig. 4d2, pink arrows) and the lack of morphological changes in the bend across the faults suggest post-MSC fault movements. However, the reflections within T3 are rotated and thicken towards the outer bend, with onlapping reflections above T3 (Fig. 4d2). We interpret that the rotation and thickening of the reflections are due to a listric fault segment that parallels the outer bend of T3 (Fig. 4d2).

Channel-fill sedimentology and stratigraphy

The sedimentology of the MSC-fill is constrained by cores from two wells (W2 and W8), which indicate a predominant fill of moderately sorted, amalgamated, structureless, fine-grained sandstones (Fig. 5) with subordinate medium- and rare coarse-grained sandstones. Well-sorted fine-grained sandstones with plane-parallel lamination are secondary (Fig. 5). No coarse-grained or mud-clast-rich layers are observed within the channel-fill. The well log motifs from gamma ray and density/neutron logs are mainly blocky, which is compatible with the homogeneous character of the sandstones (Figs 5, 6).

Structureless sandstones record rapid deposition from turbulent high-density flows (Lowe 1982). Traction in less dense flows interpreted from plane-parallel laminated sandstones is a subordinate process. The paucity of coarse-grained residual lags in the cores suggests depletive flows during channel filling.

In the upper part of the channel-fill, a unit characterized by the abrupt intercalation of sandstone beds with bioturbated laminated mudstones (Figs 5, 6), named the upper mud-prone package (UMP), is identified in all the wells above the step, suggesting longitudinal continuity (Fig. 6). The UMP culminates with a 3–5 m thick mudstone package (Figs 5, 6). As a result of its fine-grained siliciclastic composition and crude lamination, the mudstones are interpreted as dilute turbidity current deposits (Talling et al. 2012), possibly with a minor contribution of background sedimentation. The UMP is interpreted to record a period of sand starvation. Abrupt changes from sandstone to mudstone facies in a channel-fill have been related to channel deactivation (Mutti and Normark 1987). The impedance contrast between oil-saturated sandstones and the UMP permits seismic detection of the fine-grained package. Seismically, the top of the UMP coincides with a change in seismic polarity (a black peak indicates an increase in acoustic impedance due to a relatively higher density), which can be tracked in the thickest parts of the channel-fill in the step and corridor (Fig. 6a, white arrows).

The seismic profile along the channel-fill shows high-amplitude semi-parallel seismic reflections. In T2 and T3, well correlations support the view that the basal reflections likely pinch-out against the channel base, suggesting that deposition was contained by the incisional surface (Fig. 6a). There is greater reflection continuity towards the top of the channel-fill, supported by well log correlations.

The longitudinal continuity permits the MSC-fill to be subdivided into a lower sand-prone package, overlain by a fine-grained unit, the lower mud-prone package (up to 4 m thick; Fig. 6b). The lower sand-prone package and lower mud-prone package are interpreted to be the oldest preserved MSC-fill, with the lower mud-prone package recording a regional reduction in the supply of sand. An overlying intermediate package consists of two sand-prone units intercalated with an interbedded (mudstone and sandstone) unit (Fig. 6b). Above the UMP is the upper sand-prone package. Based on the correlation of the lower sand-prone package and the intermediate package, we interpret that the upper sand-prone package is the only package recorded in the MSC-fill above the ramp (Fig. 6b).

The recognition of mud-prone and sand-prone stratigraphic packages supports waxing to waning sediment supply cycles during the fill of the MSC. The increased continuity of the stratigraphic packages suggests an aggradational pattern to the infill. Together, the evidence from the sedimentology and stratigraphy of the MSC supports passive filling without modification of the topographic irregularities at the base of the channel after infilling commenced.

Incisional phase

The MSC architecture reveals a progressive increase in incision depth with distance up-dip (Fig. 7) and three anomalously deep areas of incision that punctuate this trend (Fig. 6). This pattern is in contrast with that predicted by equilibrium slope concepts in above-grade slopes, where deeper incision is expected across slope convexities (e.g. the step-to-ramp break) and shallower incision or aggradation (e.g. channel–levee systems) are expected above low-gradient areas, such as a step (e.g. Pirmez et al. 2000; Ferry et al. 2005; Deptuck et al. 2012; Shumaker et al. 2018). Here, we assess why this channel does not fit the standard model and examine the mechanics that caused the observed channel evolution.

There is a progressive longitudinal decrease in channel thickness, and thus incision, across the corridor, step and ramp. The channel thickness, and consequently depth of incision, is greatest up-dip in the corridor (56 m) and step (52 m) and progressively decreases by 0.17° in a downslope direction towards the ramp (18 m). This suggests a primarily longitudinal deformation of the seafloor and also that the forcing mechanism operated on a considerably longer length scale than the scale of the observed ramps and steps. Furthermore, the channel is observed to incise below the adjacent lobes and reaches 50–60 m depth outside T1–T3. Given that channel widths are c.300–670 m, the width to depth ratios are c. 5–11, which represent low-aspect ratios for primary submarine channels (e.g. Konsoer et al. 2013; Jobe et al. 2016; Shumaker et al. 2018), although they are similar to other heavily incised systems (e.g. Allen et al. 2022).

The combination of a progressive longitudinal change in incision, its continuity across ramps and steps, incision to well below the adjacent lobes and the very narrow nature of the filled conduit all imply that there was pronounced uplift to drive channel incision. Furthermore, this uplift was greatest in the up-dip parts, decreasing down-dip, to primarily record a basinward tilt (Fig. 7). The uplift led to an increase in channel gradient (Fig. 2), resulting in enhanced incision. This relationship has been documented from the morphometric analysis of submarine channels (e.g. Shumaker et al. 2018). However, here, the incision is also linked to the total uplift and, because this increases up-dip, the maximum incision is higher in proximal locations. Despite the pronounced uplift, the ‘stream power’ (erosive power) of the submarine channel flows was sufficient to keep up with deformation in the case of an antecedent channel or to regrade the slope in the case of channel propagation. Slope regrading by channel-forming processes, such as knickpoints and deep incision, has been demonstrated in other channel systems evolving above mobile slopes (e.g. Mayall et al. 2010; Kane et al. 2012; Jolly et al. 2017; Pizzi et al. 2023).

The exact cause of this deformation is challenging to determine in the seismic reflection data. However, the study area is affected by salt tectonics and the associated synsedimentary slope deformation and tilting have been proposed to control the evolution of the Marlim Unit (Casagrande et al. 2022). Locally, the MSC is affected by salt-related faults, but the progressive nature of this tilting across the corridor, step and ramp indicates that the uplift is related to movement on the underlying structures at a greater wavelength and, therefore, likely to be from outside the study area.

The channel shows greater sinuosity on the step than in the corridor or ramp, suggesting that, despite the overall longitudinal tilting, the subtle slope differences between these domains were sufficient to drive a morphological change in the channel. The channel may have initiated and propagated through the area during or after the period of uplift, or it may have already formed prior to the uplift as an antecedent system. The lack of obvious growth in the bend amplitude during this incisional phase in bends T1–T3, which are characterized by near-vertical channel thickness accumulation, suggests that the channel was present prior to the uplift.

The channel base is characterized by three exceptionally deep erosional areas (T1–T3), each c. 1 km long, with >20 m relief (up to c. 40 m) and average gradients of 1.5–5° on the up-dip scour surface and 2.6–8° on the down-dip scour surface. These gradients are one to two orders of magnitude greater than the gradient calculated from the average longitudinal variation in channel thickness (0.17°). Although the original seafloor slope is hard to ascertain, the present seafloor slope in the study area of 1.5° may provide a broad guide and suggests that these up-dip scour gradients are likely to represent major increases in the original seafloor slope.

Controls on basal scour development

We concentrate here on the T1–T3, kilometre-scale scours. However, the rugosity along this erosional surface is very high. The spatial association of T1 and the high thickness patterns in the lobes (LB and LC) immediately downstream of the fault-controlled slope break (Figs 3b, 4a) suggests a structural control in accommodation. The fault controlling the slope break is therefore interpreted to have been active during the deposition of the lobes and the evolution of the channel. The incisional feature is up to c. 30 m deep, with a clear concavity, and extends down-dip from the slope break for c. 700 m and up-dip for c. 300 m (Fig. 4b1). The concave morphology of the scour, and its position close to the fault at the slope break, suggest that this is a channelized plunge pool (cf. Mitchell 2006; Gardner et al. 2020; Guiastrennec-Faugas et al. 2020, 2021). Such plunge pools are observed to be fairly short longitudinally (c. 50 m) as a function of channel width W (c. 0.33W) in recently deposited sandy substrates (Guiastrennec-Faugas et al. 2021). However, they can be longer (500–2000 m) and cover a wide range with respect to width (c. 0.16–2W) in more consolidated substrates, as might be expected in the example herein, given the magnitude of basal erosion (Mitchell 2006; Gardner et al. 2020).

The width of T1 is c. 350 m, so a plunge pool of c. 1 km length fits within the observed trends, albeit one of the most elongated relative to channel width (maximum 400 m) documented to date. Plunge pool depths in these stronger substrates can be up to 200 m for slope gradients of c. 3° (Mitchell 2006) or 100 m for slopes of c. 10° (Gardner et al. 2020), with the former gradients more like the present day seafloor gradient in the study area (1.5°). Such channelized plunge pools have been linked to the formation of hydraulic jumps (Mitchell 2006; Sumner et al. 2013; Dorrell et al. 2016; Gardner et al. 2020; Guiastrennec-Faugas et al. 2020, 2021). The fact that the scour is observed to extend up-dip beyond the slope break is consistent with observations at unconfined slope breaks where plunge pools extend into the slope (e.g. Lee et al. 2002; Hodgson et al. 2022).

The deepest area of incision in the T2 is a thick area, up to 40 m greater than the average depth of the incision, is around the bend apex across the central part of the channel, extending up-dip on the inner channel bend (Fig. 4c1). We interpret that the fault system at the outer bend of T2 (i.e. the salt-rooted fault that bifurcates) led to the formation of a depression in the channel (Fig. 3c). The magnitude of the depression, ultimately reflected in the anomalous thickness patterns of T2 and the rotation of the reflections (Fig. 4), suggests that the fault system was active during channel development, with rates of synsedimentary faulting in balance with channel-forming processes. The steep gradient at the upstream end of T2 records a knickpoint. This knickpoint may have undergone headward erosion around the bend, forming the present day scoured surface. Channelized flows traversing knickpoints experience increased flow velocities that drive substrate entrainment (Sumner et al. 2013; Dorrell et al. 2016), which explains the deep scouring at T2. The channel widens to c. 650 m here, partly reflecting the fact that the bend apices are wider than the inflections (Palm et al. 2021), but also suggesting that the presence of the listric fault, and the associated scouring, broadened the channel.

The erosional feature at T3 starts from the up-dip bend inflection point and extends around the bend and past the apex. The maximum incision is >30 m above the average channel incision depths and is typically along the channel centreline, with the deepest area extending across at least half of the width of the channel (Fig. 4d1). The rotation and thickening of the MSC reflections towards the outer bend is evidence for synsedimentary activity of the listric fault segment during channel incision, forming a localized depression at the seafloor. As with T2, the up-dip part of the depression formed a knickpoint that may have undergone headward erosion up to the bend inflection.

T1, T2 and T3 represent local increases in channel depth during the incisional phase of the MSC. They are explained by the interaction of the flows with a complex slope affected by salt-related faults, perpendicular (T1) or parallel (T2, T3) to the main flow direction. The formation of fault-related knickpoints and the associated scours (T2 and T3) and plunge pool (T1) are interpreted to have been preserved prior to large-scale knickpoint migration and smoothing of the channel base. These are rare examples of filled geomorphic features, which have been mostly documented in modern channels (Bourget et al. 2011; Dalla Valle and Gamberi 2011; Gardner et al. 2020; Guiastrennec-Faugas et al. 2020, 2021; Heijnen et al. 2020), with only a few examples of structurally controlled knickpoints preserved in buried channel systems (Adeogba et al. 2005; Stright et al. 2017). The scale of these reported knickpoints is similar to the features presented here. Nevertheless, the basal scours in the MSC are much larger than the scours previously documented for submarine channels (maximum depth of 9 m for scour bends; Snedden 2013). This study provides a comprehensive 3D understanding of these features and explains their preservation.

The deformation associated with the development of large scours in the two bends, and the plunge pool at T1, was sufficiently rapid to lead to large-scale erosion, but slow enough to allow channel-forming processes to not be so disturbed that channel avulsion or crevasse splay development occurred (e.g. Casagrande et al. 2022). This suggests that the timescales of deformation and channel development are comparable. It is unknown whether fault-induced deformation in extensional salt-driven domains is typically in the same range as channel-forming processes, or whether this is an exceptional example.

Given that the basal surfaces of submarine channels record long-term erosion and/or sediment bypass (e.g. Mutti and Normark 1987; Deptuck et al. 2007; Hubbard et al. 2014; Hodgson et al. 2016) as channels tend to reach equilibrium, the formation of a smooth basal surface is expected (Sylvester et al. 2012; Hubbard et al. 2014). However, the preservation of kilometre-scale scours and a highly rugose basal surface suggests that the process of channel base smoothing was interrupted relatively soon after the formation of these features. The up-dip knickpoint migration of the scours in T2 and T3 was limited, suggesting that the cessation of erosion was relatively rapid or that the erosion rate declined sufficiently that the further migration of knickpoints was reduced prior to the full cessation of erosion. Because the preserved geomorphic features scale with the basal channel surface, we interpret that the incision surface is the product of a single channel element (sensu Sprague et al. 2005; McHargue et al. 2011) and not a composite surface (i.e. not a channel complex). However, the channel-fill, with sand- and mud-prone packages that reflect variations in the sediment supply, represents multiple waxing and waning cycles in sediment supply, pointing to the fill being a composite body comprising several channel storeys (sensu Sprague et al. 2005).

So, what caused the preservation of this snapshot of a basal surface prior to later smoothing? The cessation of erosion in this channel could have been related to several processes, including a rapid switch from uplift to subsidence, a change in flow properties leading to a change from erosion to deposition, or avulsion and complete or gradual channel abandonment. Avulsion can be dismissed here as the architecture of the MSC-fill records a passive style of deposition, much of it characterized by thick structureless sands indicative of the passage of high-density turbidity currents. Consequently, there is clear evidence that the channel stayed open and active long after the burial of the basal surface.

A return to subsidence after transient uplift would lead to the channel base moving towards a different longitudinal slope equilibrium marked by deposition to reduce the thalweg gradient. The incoming flows would have less erosive power and capacity to transport their load, thereby ‘fossilizing’ the thalweg morphology and producing the facies and stacking patterns interpreted in the MSC-fill. The interpreted transient and rapid salt-related uplift phase, with a basinward tilt, could have occurred during a longer period of overall subsidence – for example, due to a change in the mode of diapir growth (e.g. Jackson and Hudec 2017) and/or a change in tilt direction due to the application of a non-uniform load (e.g. Duffy et al. 2021).

The length scale of the basinward tilting is much longer than the length of the corridor step-ramp. The transient uplift is therefore a far-field effect of salt tectonics beyond the study area and formed in the context of the overall lateral tilting observed across a broader area (Casagrande et al. 2022) in this extensional salt domain. Furthermore, we interpret that the formation of several deep scours at the channel base is a function of synsedimentary faulting driven by underlying salt movement. Thus, a transient switch to the uplift and reactivation of faults as a near-field response to far-field salt tectonics would be consistent with the dominance of salt tectonics in the evolution of this system. In the Campos Basin, the extensional reactivation of salt-rooted salt structures has been observed in the study area during the Oligocene and Miocene (e.g. Casagrande et al. 2022). Regional studies document salt tectonic activity during the Paleogene and Neogene as a response of transpression and transtension in basement structures (e.g. Fetter 2009) and during the Paleogene due to the continuation of regional thin-skinned extension (do Amarante et al. 2021).

Alternatively, a change in flow character could lead to the preservation of the channel base morphology. A reduction in flow magnitude, sediment concentration or a change in sediment calibre to one with less fines (less efficient), or some combination of these, would result in flows that are more depositional, raising the theoretical equilibrium profile (e.g. Kneller 2003). Thus, at a given point in the system, there may be a change from flows being dominantly erosional to dominantly depositional. This change might have occurred through a relatively gradual change in boundary conditions linked to extra-basinal allocyclic controls, such as sea-level variations (e.g. Deptuck et al. 2003), but is enough to pass a threshold from dominantly erosional to dominantly depositional (or non-erosional) flow behaviour. This scenario is likely to be supported by the channel cutting down to a harder, less erodible substrate, as already discussed, and the mantling of the channel base by material that was harder to entrain, such as gravel or mud clasts. Initial changes in the flow conditions may have acted to reduce erosion of the plunge pool and knickpoints – for instance, by a change from supercritical to subcritical flow regimes. Further changes in the flow conditions would then progressively decrease the erosion rate before passing the threshold for deposition.

In summary, the preservation of the undulating channel base morphology, and the lack of channel smoothing, could be linked to a return to subsidence after transient uplift, a change in flow properties or a combination of both these factors.

Observations of knickpoints and associated scours on the order of metres to >10 m in modern submarine channel systems and canyons have revealed that they can migrate upstream at rates of up to hundreds of metres per year (Guiastrennec-Faugas et al. 2020, 2021; Heijnen et al. 2020). This dynamism, and their impact on vertical and lateral erosion, have been used to argue that knickpoints control channel evolution (Heijnen et al. 2020). However, these observations of rapid knickpoint migration are in sandy, largely cohesionless, sediments within active channels.

Here, we show an example of a channel cut into consolidated and cohesive mud-prone substrates. Although the true timescales for the erosion of this example channel are unknown, the fault-induced deformation occurs on timescales sufficiently slow that they do not induce channel avulsion or other major channel planform changes (e.g. Holbrook and Schumm 1999; Peakall et al. 2000). Furthermore, the knickpoints generated at T1–T3 have not had time to dissipate, while simultaneously smoothing the basal surface, something that takes place on decadal timescales in modern examples with cohesionless substrates (Guiastrennec-Faugas et al. 2020; Heijnen et al. 2020). Consequently, knickpoint migration in consolidated, cohesive, mud-prone substrates is likely to be associated with much slower migration rates than in modern sandy examples (see also Allen et al. 2022). This suggests that rapid knickpoint migration in weak substrates may not be representative of erosion during channel inception or long-term controls on channel evolution, but rather marks a periodic overprint whose long-term stratigraphic record may be limited and primarily recorded in the channel-fill.

The limited retreat of basal knickpoints observed herein suggests, in the generic case of submarine channels in consolidated mud-prone substrates (or equivalent high-strength substrates), that the long-term evolution of the basal geomorphic surface of a channel element is not primarily controlled by knickpoint erosion. Instead, channel evolution and bend development are probably controlled by the 3D flow fields of traversing currents (e.g. Giorgio Serchi et al. 2011; Dorrell et al. 2013; Sumner et al. 2014), which, in turn, are related to channel curvature (e.g. Peakall and Sumner 2015; Morris et al. 2022). However, knickpoint migration probably has a key role in the basinward reworking of sediments during aggradation of the conduit (Heijnen et al. 2020; Guiastrennec-Faugas et al. 2021).

The MSC is a rare example of a highly rugose submarine channel base preserved in an ancient system influenced by salt tectonics. The channel base contains geomorphic features considered to be ephemeral in modern systems and is dominated by three areas with anomalous (tens of metres) channel thicknesses. Channel incision patterns across the stepped profile are interpreted to reflect a salt-controlled transient uplift and basinward tilting of the slope. This scenario is controlled by a combination of far-field salt tectonics (a change to a basinward tilt) and near-field salt tectonics (the activation of listric faults), which is a scenario that is probably applicable to other evaporite-dominated basins worldwide. The singular preservation of these erosional features results from a combination of factors (salt tectonics, substrate strength and/or flow and sediment supply) that induced changes in flow behaviour from erosional to depositional and inhibited headward knickpoint migration and channel base smoothing. The exceptional nature of this salt-generated basal scouring preserved in this ‘salty snapshot’ challenges the present paradigm that knickpoint migration dominates channel processes and evolution. Our results show that the primary erosion of the main channel surface and the long-term evolution of the channel are much more gradual.

This study is part of the first author's PhD project, which is supported and funded by Petrobras S.A., Brazil. The authors thank Petrobras and the Brazilian National Agency of Petroleum, Natural Gas and Biofuels (ANP) for permission to use seismic and well data. The authors also thank Mary Ford as editor and reviewer, and Zane Jobe as reviewer for their insightful comments and helpful suggestions to improve the manuscript.

JC: conceptualization (lead), data curation (lead), formal analysis (lead), funding acquisition (supporting), investigation (lead), methodology (lead), project administration (lead), resources (lead), software (lead), supervision (supporting), validation (lead), visualization (lead), writing – original draft (lead), writing – review and editing (lead); DMH: conceptualization (supporting), data curation (supporting), formal analysis (supporting), funding acquisition (lead), investigation (supporting), methodology (supporting), project administration (supporting), resources (supporting), software (supporting), supervision (lead), validation (supporting), visualization (supporting), writing – original draft (supporting), writing – review and editing (supporting); JP: conceptualization (supporting), data curation (supporting), formal analysis (supporting), funding acquisition (supporting), investigation (supporting), methodology (supporting), project administration (supporting), resources (supporting), software (supporting), supervision (supporting), validation (supporting), visualization (supporting), writing – original draft (supporting), writing – review and editing (supporting).

This work was funded by Petrobras.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The data that support the findings of this study are the property of the Brazilian National Agency of Petroleum, Natural Gas and Biofuels (ANP). Restrictions apply to the availability of these data, which were used under a research licence agreement between Petrobras and ANP. For more information about how to access geological and geophysical data from Brazilian basins, contact helpdesk@anp.gov.br

This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License (http://creativecommons.org/licenses/by/4.0/)