Submarine channels are conduits for sediment delivery to continental margins, and channel deposits can be sandy components of the fill in tectonically active salt basins. Examples of salt-withdrawal basin fill commonly show successions of sandy channelized or sheet-like systems alternating with more mud-rich mass-transport complexes and hemipelagites. This alternation of depositional styles is controlled by subsidence and sediment-supply histories. Salt-basin fill comprising successions of largely uninterrupted meandering-channel deposition are less commonly recognized. This begs the questions: can sediment supply be large enough to overwhelm basin subsidence and result in a thick succession of channel deposits, and, if so, how would such a channel system evolve? Here, we use three-dimensional seismic-reflection data from a >1500 km2 region with salt-influenced topography in the Campos Basin, offshore Brazil, to evaluate the influence of salt diapirs on an Upper Cretaceous–Paleogene giant meandering submarine-channel system (channel elements >1 km wide; meander wavelengths several kilometers to >10 km). The large scale of the channels in the Campos Basin suggests that sediment discharge was large enough to sustain the meandering channel system in spite of large variability in subsidence across the region. We interpreted 22 channel centerlines to reconstruct the detailed kinematic evolution of this depositional system; this level of detail is akin to that of recent studies of meandering fluvial channels in time-lapse Landsat satellite images. The oldest channel elements are farther from salt diapirs than many of the younger ones; the centerlines of the older channel elements exhibit a correlation between curvature and migration rate, and a spatial delay between locations of peak curvature and maximum migration distance, similar to that observed in rivers. As many of the younger channel centerlines expanded toward nearby salt diapirs, their migration pattern switched to downstream translation as a result of partial confinement. Channel segments that docked against salt diapirs became less mobile, and, as a result, they do not show a correlation between curvature and migration rate. The channel migration pattern in the Campos Basin is different compared to that of a tectonically quiescent continental rise where meander evolution is unobstructed. This style of channelized basin filling is different from that of many existing examples of salt-withdrawal minibasins that are dominated by overall less-channelized deposits. This difference might be a result of the delivery of voluminous coarse sediment and high discharge of channel-forming turbidity currents to the Campos Basin from rivers draining actively uplifting coastal mountains of southeastern Brazil. Detailed kinematic analysis of such well-preserved channels can be used to reconstruct the impact of structural deformation on basin fill.

Submarine channels are conduits through which gravity flows transport terrigenous sediment across continental margins to deep-sea fans (Mutti and Normark, 1987; Piper and Normark, 2001). Sediment-gravity flows create and modify channels by erosion and deposition; these processes drive the lateral migration of the channel floor and, in its wake, the deposition of channel fill (Deptuck et al., 2003; Sylvester et al., 2011; Hubbard et al., 2014). Resultant sinuous channel patterns are common on the continental slope, for example offshore West Africa (Mayall et al., 2006; Deptuck et al., 2007; Hansen et al., 2017) and in the Gulf of Mexico (Sylvester et al., 2012), and on the continental rise (Flood and Damuth, 1987; Pirmez and Flood, 1995; Wynn et al., 2007, and references therein), where they resemble meandering rivers in map view (e.g., Damuth et al., 1983; Pirmez and Imran, 2003; Sylvester and Pirmez, 2017). Sinuosity is a key recognition criterion of submarine channels in three-dimensional (3-D) seismic-reflection and seafloor data sets (Normark et al., 1993), and, in the subsurface, high-amplitude seismic reflections defining sinuous channel deposits can indicate sand-dominated lithology (Kolla et al., 2007). Many examples of submarine channels meander: they show systematic lateral migration by bend growth, with downstream translation and cutoffs like in rivers (Abreu et al., 2003; Posamentier, 2003; Posamentier and Kolla, 2003; Kolla et al., 2007; Babonneau et al., 2010; Sylvester and Covault, 2016; Hansen et al., 2017; Covault et al., 2020).

Changes in submarine-channel migration and depositional patterns have been linked to variations in gradient (Clark et al., 1992; Pirmez et al., 2000). Tectonically active salt basins, such as offshore Angola and in the Gulf of Mexico, are characterized by topographic complexity and gradient variations (Prather et al., 1998; Beaubouef and Friedmann, 2000; Beaubouef et al., 2003; Gee and Gawthorpe, 2006; Oluboyo et al., 2014). In these settings, channels tend to be more deeply incised across the steep limbs of anticlines or basin margins, whereas sediment fairways broaden to form weakly confined channel and sheet-like deposits on relatively flat basin floors or topographic steps across the slope (Gee et al., 2007; Deptuck et al., 2012). In salt-withdrawal basins of the Gulf of Mexico, the interaction of subsidence and sediment supply creates depositional packages of sandy channelized, “bypassing” systems or more sheet-like, “ponded” systems alternating with muddier mass-transport complexes and hemipelagites (Prather et al., 1998; Winker and Booth, 2000; Sylvester et al., 2015). The stacking of these depositional packages within a basin can be used to reconstruct the deformation of salt-related structures, and there is opportunity for more detailed characterization of channel migration in response to deformation (Gee et al., 2007; Clark and Cartwright, 2009; Mayall et al., 2010).

In this contribution, we are motivated by the following questions. Can sediment supply be large enough to overwhelm significant basin subsidence and result in a thick succession of sinuous channel deposits, rather than the more commonly recognized successions of ponded sheets of sand or muddy hemipelagites and mass-transport complexes alternating with channel deposits? If so, how would such a long-lived channel system evolve? Would it exhibit evidence of meandering, such as systematic lateral migration, downstream translation, and cutoffs? Or would confinement by basin-bounding salt diapirs promote aggradation of channel bends, resulting in stacks of ribbon-like sand bodies (Peakall et al., 2000)? To answer these questions, we used 3-D seismic-reflection data from a >1500 km2 area of the deep-water Campos salt basin offshore Brazil to interpret the stratigraphic evolution of a giant submarine-channel system (Fig. 1). We interpreted paleochannel centerlines to reconstruct the migration of the channel system at a level of detail that approaches that of studies of meandering fluvial channels in time-lapse Landsat satellite images (e.g., Sylvester et al., 2019a, 2021), and paleochannel meander-belt evolution using crosscutting relationships and accretion patterns (e.g., Durkin et al., 2017, 2018). Results of this study can be applied to understanding the depositional response to salt deformation in other tectonically active basins where thick successions of meandering submarine-channel systems were deposited. We also aim to show that detailed kinematic analysis of channel migration, developed for meandering rivers (Sylvester et al., 2019a), is a useful tool for interpreting the evolution of sinuous submarine channels as well.

The Campos Basin is located along the southeastern continental margin of Brazil in the South Atlantic (Fig. 1). It is one of the most productive deep-water hydrocarbon basins in the world (Mohriak et al., 1990; Bruhn et al., 2003). The Campos Basin initiated during Late Jurassic breakup of Gondwana and opening of the South Atlantic (Guardado et al., 1989) and comprises Berriasian to early Aptian continental rift deposits overlain by middle Aptian salt (Davison, 2007; Karner and Gamboa, 2007; Winter et al., 2007), an early to middle Albian carbonate platform, and a late Albian to present succession of deep-water continental-margin deposits (Bruhn, 1998). The Aptian salt plays an important role in establishing the structural style of the Campos Basin; deformation was initiated by early Albian eastward basin tilting and subsequent gravity spreading as progradation occurred (Demercian et al., 1993; Davison et al., 2012; Quirk et al., 2012; De Gasperi and Catuneanu, 2014).

During the Late Cretaceous and Paleogene, small rivers drained high-relief coastal mountains of the Serra do Mar and delivered voluminous siliciclastic sediment to the Campos Basin (Bruhn and Walker, 1995) (Fig. 1). During the Late Cretaceous, the Trindade mantle plume promoted dynamic uplift of the Serra do Mar (Thompson et al., 1998; Cobbold et al., 2001; Meisling et al., 2001; Fetter et al., 2009; Quirk et al., 2013). Coastal rivers delivered sediment directly to submarine canyons, which had incised across an unstable, narrow shelf as sea level was high and rising (Fetter et al., 2009). Still under the influence of the Trindade mantle plume, basement faults were reactivated during the Paleogene, promoting coastal uplift and onshore drainage reorganization (Cobbold et al., 2001; Fetter et al., 2009). At the same time, long-term (i.e., 107 yr) sea level fell, and the Campos Basin margin prograded tens of kilometers into the South Atlantic, creating a wide shelf. Onshore, rivers developed along a series of small, high-altitude rift basins inland of the Serra do Mar. These rivers were progressively captured by headward erosion of the Paraíba do Sul River along the southwest-northeast trend of the Serra da Mantiqueira (Karner and Driscoll, 1999). This drainage reorganization accounts for the Neogene course of the Paraíba do Sul River, which is ∼600 km long, draining the coastal highlands of southeastern Brazil from the Serra do Mar to Campos dos Goytacazes in the northeast and forming a prominent delta (Cobbold et al., 2001).

We focus on a channel system probably as old as Late Cretaceous and as young as Oligocene in >2 km of water depth in a structural domain characterized by extensional and contractional salt stocks and walls formed during west-to-east translation of surrounding minibasins across the margin (Demercian et al., 1993; Mohriak et al., 2012) (Figs. 1 and 2). We constrained the age of the giant submarine-channel system of this study to Late Cretaceous–Paleogene based on published ages for horizons tied to our seismic-reflection data (Fetter et al., 2002; Fetter, 2009; Mohriak et al., 2009; Contreras, 2011). Fetter et al. (2009) analyzed Lower Cretaceous (Albian) to Miocene turbidites from 10 wells in the deep-water Campos Basin. The deposits showed a coarsening trend of average grain size from fine sand in the Albian and Cenomanian to medium sand in the Santonian to upper Eocene, with shifts to coarse sand in the Maastrichtian and upper Eocene. The coarser deposits are also relatively compositionally immature (i.e., lower quartz to feldspar ratio, with more volcanic rock fragments; Fetter et al., 2009). Fetter et al. (2009) interpreted the coarse, compositionally immature Upper Cretaceous to Paleogene deposits to be products of the short, mountainous rivers draining actively uplifting coastal source areas of southeastern Brazil; these rivers delivered sediment directly to submarine canyons across a narrow continental shelf, similar to the setting of tectonically active basins dominated by catastrophic flooding (Mutti et al., 1996) and hyperpycnal flows (Mulder and Syvitski, 1995). Moreover, long-duration hyperpycnal flows have been thought to promote meandering in submarine channels (Mulder et al., 2003). The Upper Cretaceous–Paleogene channel system of this study is one of many, which exhibit variable geometries and depositional trends, within a thick (<2500 m) Upper Cretaceous–Quaternary succession (Fig. 2). Paleoflow varied depending on local structural configuration and diapir orientation; Covault et al. (2020) documented north-to-south paleoflow for a Miocene channel system, and Ceyhan (2017) documented northwest-to-southeast, west-to-east, and north-to-south paleoflow for Pliocene–Pleistocene channel systems (Fig. 2).

We used a Kirchhoff pre-stack depth-migrated 3-D seismic-reflection volume with wavelengths of ∼50–100 m (vertical resolution ∼12.5–25 m) and a 25 m horizontal sampling rate. The seismic-reflection volume was provided by PGS Suporte Logístico e Serviços Ltda. Seismic-reflection data were processed to zero phase. We used the Paradigm SeisEarth interpretation and visualization product suite to map four regional horizons based on line-by-line continuity and terminations of relatively high-amplitude seismic reflections in cross section (Figs. 3 and 4) (Mitchum et al., 1977). These horizons cover an area ∼50 km long from west to east and ∼20 km wide from north to south. From base to top, these horizons are named 1, 2, 3, and 4 (Figs. 24; Files S1–S3 in the Supplemental Material1). We used root-mean-square (RMS) amplitude maps to identify channel deposits and reconstruct the evolution of channel centerlines (cf. De Ruig and Hubbard, 2006). We extracted RMS amplitude across mapped horizons and proportional slices between mapped horizons from 200 m depth windows; we tested smaller windows, but they did not resolve continuous channel geometries across the study area.

A common measure of a river channel is the bankfull discharge, which is the maximum amount of water that a channel can carry without overflowing its banks (Leopold and Wolman, 1960); channel shape and dimensions of meandering rivers are associated with bankfull flow (Wolman and Miller, 1960). However, in the submarine realm, channels are modified by sediment-gravity flows, which are mixtures of sediment and water in which the sediment component pulls interstitial water downslope under the action of gravity (Middleton and Hampton, 1973). Turbidity currents are a type of sediment-gravity flow and are primarily responsible for sculpting and driving the migration of meandering submarine channels (Bouma et al., 1985; Pirmez and Imran, 2003). Turbidity currents do not have a well-defined upper boundary because both velocity and sediment concentration decline upward gradually; therefore, it is less obvious what can be considered as bankfull.

On the seafloor, channel systems of large submarine fans are commonly composed of large erosional valleys with a relatively narrow, sinuous channel in the deepest location in the valley (i.e., the thalweg) (Fig. 5A). The depths of submarine-channel systems from thalweg to levee crest can be 10 to 100 times larger than bankfull depth in rivers, and levee-crest widths can be two to three times wider than for rivers (Pirmez and Imran, 2003; Konsoer et al., 2013; Shumaker et al., 2018). However, the planform characteristics of submarine channels and rivers are remarkably similar (Flood and Damuth, 1987; Pirmez and Imran, 2003); therefore, the effective width and depth of turbidity currents that form submarine channels are commonly narrower and shallower than the outer-levee-crest width and levee-thalweg depth (Pirmez and Imran, 2003). Moreover, both the velocity and concentration maxima of turbidity currents occur near the base of the flow (Sequeiros et al., 2010; Eggenhuisen and McCaffrey, 2012; Paull et al., 2018; Wang et al., 2020). This high-velocity core is likely to be thinner and narrower than the full channel or canyon, and it is the portion of the flow that contains most of the sediment load (Luchi et al., 2018) and sculpts the meandering planform of the channel (Pirmez and Imran, 2003). In the Zaire (Congo) Canyon, the largest observed turbidity currents were largely restricted to the bottom 40 m, a depth that corresponds to the part of the canyon that seems to have a characteristic width (Azpiroz-Zabala et al., 2017).

These characteristic channel geometries are commonly recognizable in the subsurface as sinuous, high-amplitude seismic reflections (Kolla et al., 2007), especially when they are abandoned, leaving thin accumulations of relatively coarse lag deposits, then passively filled by mud and, therefore, well preserved. The most recent channel on the proximal Bengal Fan has a well-defined meandering planform with a characteristic width that is much smaller than the distance between the outer-levee crests (Fig. 5A). The distance between the outer-levee crests has no simple and direct relation to the characteristic discharge of the channel-shaping flows, for reasons explained above (Pirmez and Imran, 2003). Similar morphologies develop in forward models of submarine channel-levee systems (Figs. 5B5C; Sylvester et al., 2011; Sylvester and Covault, 2016; Covault et al., 2016). The deposits within the narrow, sinuous channel located at the bottom of the system have been called “channel elements” (Mutti and Normark, 1987; Fildani et al., 2013; Hubbard et al., 2014), which migrate and stack over time to produce larger-scale, composite channel systems (e.g., Deptuck et al., 2003, 2007; Mayall et al., 2006; Hodgson et al., 2011; McHargue et al., 2011; Sylvester et al., 2011; Covault et al., 2016; Fig. 5C).

We interpreted 22 centerlines of “channel elements” sensuFildani et al. (2013) in the Campos Basin (Fig. 6). We analyzed these centerlines to reconstruct the detailed kinematic evolution of the depositional system (Figs. 6 and 7). We calculated migration distances between successive channel centerlines using a Python implementation of the dynamic time-warping algorithm, a technique that correlates each point along the first centerline to the closest point on the second centerline (Fig. 7; File S4 [footnote 1]; Sylvester et al., 2019a). We used the dtw function of the librosa library (McFee et al., 2015) and Euclidian distance as a distance measure between the centerlines. Dynamic time warping gives better estimates of the migration distance than simpler approaches like nearest-neighbor searches (Sylvester et al., 2019a) because it minimizes the sum of distances for all points along the centerlines, not just the point under consideration. As a result, the migration rate along a centerline varies relatively smoothly.

For comparison to other channel systems, we measured widths and meander half-wavelengths of channel elements in the Campos Basin (Fig. 8). We are most confident in width measurements of channel-element segments we mapped in horizons 2 and 3. We measured meander wavelengths by automatically identifying inflection points along the channel centerline (see Sylvester et al. [2013] and Sylvester and Pirmez [2017] for details). Figure 8 shows our measurements of channel elements in the Campos Basin plotted with those of other systems from the Bengal Fan (Kolla et al., 2012), the Dalia channel system offshore West Africa (Abreu et al., 2003), the Joshua Channel of the eastern Gulf of Mexico (Posamentier, 2003; Kramer et al., 2016), the Veracruz Fan of the western Gulf of Mexico (Winter, 2018), the Hikurangi Channel offshore New Zealand (McArthur and Tek, 2021), the Puchkirchen and basal Hall Formations of the Molasse Basin, Austria (Hubbard et al., 2009), and the La Jolla Channel offshore southern California, USA (Fildani et al., 2021). We replotted data collected by Pirmez (1994) for the Amazon Channel. We also plotted width and wavelength data for rivers (Leopold and Wolman, 1960; Howard and Hemberger, 1991; Sylvester et al., 2019a).

Regional Seismic-Reflection Horizons

As described above, we mapped four regional seismic-reflection horizons in the deep-water Campos Basin (Figs. 24), named, from base to top, horizons 1, 2, 3, and 4 (Fig. 2; Files S1–S3 [footnote 1]). Horizon 1 is a high-amplitude peak characterized by channel geometries and erosional truncation of underlying reflections. Horizon 2 is a peak that exhibits variable amplitude and geometry. That is to say, horizon 2 is continuous and approximately parallel to surrounding reflections where represented by low to moderate amplitude; however, it is characterized by channel geometries and erosional truncation of underlying reflections where represented by higher amplitude. Horizon 3 is a variable-amplitude peak that appears to be truncated by overlying high-amplitude channel reflections. To the south and east of our study area, horizon 3 separates more continuous, low- to moderate-amplitude reflections from overlying lower-amplitude faulted reflections. Horizons 1–3 are described in detail below in the Channel-System Architecture section. Horizon 4 is a regionally mappable, high-amplitude peak, which we mapped to tie our work to other studies in the region. We do not have well data to calibrate the seismic-reflection data; we attempted to tie published horizons to our study area based on reflection character and position. Our interpretations of horizon ages are based on studies that overlap or nearly overlap with our study area (Fetter et al., 2002; Fetter, 2009; Mohriak et al., 2009; Contreras, 2011). Our age interpretations of horizons 1–3 are tenuous, but they broadly put our depositional system in the context of the Campos Basin margin; horizons 1–3 could represent a depositional system somewhere within Upper Cretaceous to Oligocene strata. That is to say, the sequence between horizons 1–3 was deposited during the Late Cretaceous and/or Paleogene, when small rivers drained high-relief coastal mountains of the Serra do Mar and delivered voluminous siliciclastic sediment to the Campos Basin (Bruhn and Walker, 1995). The precise age of these deposits has no impact on our observations and interpretations of channel-system morphology and stratigraphy. Horizon 4 is a widely recognized marker for reference; it might be the Oligocene “blue marker” surface, which represents a carbonate depositional event separating Oligocene from Eocene reservoirs in the Barracuda oil field (Rangel et al., 2003).

Channel-System Architecture

Here, we describe the cross-sectional and map-view character of horizons 1–3 between five salt bodies (I–V) in more detail (Figs. 24 and 6; Files S1–S3 [footnote 1]). Horizon 1 is defined by the bases of high-amplitude, discontinuous, channel-shaped seismic reflections. In cross section 1–1′, the western margins of these channel-shaped reflections are commonly preserved, with the eastern margins eroded, and they predominantly stack from west to east (Figs. 2 and 4). Some of the margins of the channel reflections appear to transition upward into lower-amplitude, continuous reflections (cross section 1–1′ of Figs. 2 and 4). The high-amplitude, discontinuous reflections probably represent coarse-grained channel deposits, and the lower-amplitude, continuous reflections are finer-grained levee-overbank deposits based on seismic-facies models of deep-water depositional systems (Normark et al., 1993; Abreu et al., 2003; Deptuck et al., 2003). Channel-deposit reflections appear to climb and vertically stack up to horizon 2, which is an intermediate horizon defined by high-amplitude channel-shaped reflections transitioning into lower-amplitude, continuous reflections (Figs. 2 and 4). Cross section 1–1′ in Figures 2 and 4 shows that horizons 2 and 3 appear to erosionally truncate low-amplitude, continuous reflections in the western and central parts of the study area; horizons 2 and 3 appear to truncate high-amplitude channel-deposit reflections in the east. Horizon 3 is the top of the sequence, and it erosionally truncates underlying reflections (Fig. 4). Figure 4 cross section 1–1′ shows the channel system thins from west to east (see also Fig. 2). We discuss this pattern below in the context of the isochore map between horizons 1 and 3 (Fig. 3C).

Figures 6A6E show RMS amplitude maps between horizons 1 and 3. High values of RMS amplitude (dark gray to black parts of the maps) define sinuous ribbons, which correspond to the high-amplitude channel-deposit reflections in cross sections of Figures 2 and 4. We interpreted 22 channel elements (Fig. 6F), and we measured width and meander half-wavelength of segments of channel elements 15, 16, and 22, which we mapped in greater detail in horizons 2 and 3 (channel elements >1 km wide; meander wavelengths several kilometers to >10 km) (Figs. 6 and 8). Reconstructed centerlines of individual channel elements are plotted in Figures 6 (right column) and 7 and Files S4 and S5 (footnote 1). These channel elements stack in an ordered pattern to form a giant meandering channel system. This system exhibits systematic lateral migration, with successive channel elements in close proximity to, but usually slightly downstream of, the preceding channel element. Channel migration is dominated by downstream translation; this pattern is also present in the overlying channel systems of the Campos Basin (Fig. 2; Covault et al., 2020). Horizon 1 is the base of this large meandering channel system. Horizon 2 is an intermediate horizon that is located above a meander cutoff in the central-eastern part of the study area (channel element 16; Fig. 6). Horizon 3 is the top of this channel system, which was locally eroded by a younger channel system that is not the focus of this study (Figs. 2 and 4).

Basin subsidence and salt diapir growth were variable in time and space during channel-system evolution. This interpretation is based on the isochore map between horizons 1 and 3, which shows a significant decrease in thickness from west to east, along the path of the channel system, across three well-defined depocenters (Fig. 3C; see also cross section 1–1′ in Figs. 2 and 4). This variation in thickness reflects differential basin subsidence; higher subsidence rates in the western area promoted the accumulation of a thick, aggradational channel and levee-overbank sequence, whereas lower subsidence rates in the east promoted a thinner succession of laterally offset, meandering channel deposits (Fig. 4). That said, Figures 2 and 4 show erosion by overlying channel systems across horizon 3; therefore, part of the west-to-east thinning could be the result of erosion. Even with the west-to-east thickness decrease between horizons 1 and 3, we were able to reconstruct the individual channel elements (Fig. 6). This was a surprise; we expected the spatial variability in subsidence to (1) inhibit our reconstructions of all channel elements, especially in the east where the succession is thinner, and (2) promote alternating channelized and more sheet-like depositional architecture, especially in the thicker, expanded western side. The persistence of large channel elements suggests that the sediment discharge was large enough to maintain the meandering channel system in spite of significant basin subsidence in the western part of the area.

The influence of salt diapirism is reflected in submarine-channel evolution. At the scale of the entire channel belt, the overall west-east orientation and predominance of downstream channel translation probably result from confinement by the surrounding salt diapirs (Figs. 3, 6, and 7). In the east, channels appear to run into salt body V, where they turn and exit the study area to the southeast. Based on the observation that some of our reconstructed channel centerlines pass directly over the salt diapirs, we conclude that at least salt bodies I, II, and V had less positive relief during the times that correspond to horizons 1 and 3 than at present. The missing channel segments were eroded following channel formation as a result of the continued rise of the diapirs. We discuss the detailed submarine-channel evolution in the next section.

Channel Kinematic Evolution

In addition to reconstructing the channel system, we have also analyzed the migration patterns between consecutive channel elements (Fig. 7F). Using dynamic time warping, we estimated migration vectors (thinner red lines in Figs. 7A7E) so that they are approximately perpendicular to the initial centerline (black) and there are no significant gaps between the corresponding points on the next centerline (red). We selected five time steps between channel centerlines to illustrate the influence of salt diapirs on channel migration (Figs. 7A7E); the migration vectors for all 21 time steps are shown in Files S4 and S5 (footnote 1). We also plotted the dimensionless curvature of the initial centerline (mean channel width of 1000 m divided by local radius of curvature; blue area under the curve) and the migration distance to the next centerline (green area under the curve) as functions of along-channel distance. Points of no migration are marked by red dots on both the centerline maps and the migration-distance plots for comparison. In a few cases, we linked inflection points (i.e., points of zero curvature) to the corresponding points of no migration to show the similarity between the two series when accounting for a phase lag (dashed black lines). A similar phase lag has been predicted and documented in rivers (Howard and Knutson, 1984; Furbish, 1988; Sylvester et al., 2019a). The location of maximum migration of a river meander does not coincide with the bend apex; instead, it is located downstream from the maximum curvature, at a distance approximately two to four times the channel width (Sylvester et al., 2019a). In freely meandering rivers, migration rate can be predicted using a simple model that is based on the weighted sum of upstream curvatures (Howard and Knutson, 1984; Furbish, 1988; Sylvester et al., 2019a). Significant departures from this prediction exist in locations where rivers are confined by less-erodible banks, which inhibit migration (Güneralp and Rhoads, 2011; Sylvester et al., 2019b). Files S4 and S5 (footnote 1) show the 21 time steps between the 22 channel centerlines, including plots of dimensionless curvature and migration distance.

Initially, channel centerlines were far from salt bodies and, like in rivers, there was a remarkable similarity between the along-channel variation of migration distance and curvature, with a phase lag of about two to seven times channel width (Fig. 7A). As the channel system evolved, the limbs of some bends docked against salt diapirs. Subsequently, migration was negligible for these bends, and, as a result, the relationship between curvature and migration is less clear. This is similar to river segments that are affected by banks resistant to erosion. Bends 1 and 2 underwent significant expansion before getting confined by the structures to the south and north (Figs. 6 and 7A–7B). The apex of bend 3 got close to the southern structural high early on and, as a result, shows significant translation during time steps 1–14. The downstream limb of bend 3 was immobilized by a promontory of salt structure III (Fig. 7B). Continued translation of the upstream limb of bend 3 moved it closer to the immobile downstream limb, effectively squeezing the bend until it was cut off at time step 15 (Figs. 6 and 7B–7C). Time steps before and after the cutoff are shown in Figures 7B and 7C (see also Fig. 6 and Files S4 and S5 [footnote 1]). By time step 16, bends 1, 2, and 4 expanded to the locations of salt diapirs and largely became immobile (Figs. 7B7C). An exception is the formation of bend 3a at centerline 16 following the cutoff of bend 3 (Fig. 7C). Free of the salt body III obstruction, the smaller, tighter bend 3a rapidly translated downstream (∼500 m per time step; Figs. 7C7D) until time step 20, when another cutoff occurred (Fig. 6 and Files S4 and S5 [footnote 1]). Bend 3a also shows a well-defined relationship between migration rate and curvature when a consistent ∼2–3 km lag is considered (Figs. 7C7D). Another migration pattern emerged following the cutoff at time step 15: bend 4 reversed direction from expanding toward salt body IV to migrating away from it (Figs. 7C7D). This is not predicted by a simple kinematic model, and it might reflect spatial changes in subsidence related to movement of salt bodies IV and V. That is to say, initially, during time steps 1–15 (centerlines 1–16), uplift above and adjacent to salt body V might have created a paleotopographic low to the north of the diapir; as a result, bend 4 grew by meander expansion into the accommodation (Figs. 7A7B). Then, accelerated uplift of salt body IV shifted the depocenter to the south and bend 4 became smaller by meander contraction into the new low (Figs. 7C7D).

Similar to the cutoff at time step 15, the cutoff at time step 20 seems to be related to the translation of bend 3a and concomitant tightening of bend 4, with its downstream limb immobilized against salt (Fig. 7D). We interpreted only two additional centerlines following the cutoff at time step 20 (Fig. 7E). These last channel elements are in close proximity and appear to roughly follow a path of steepest descent between salt bodies and into deeper water to the southeast beyond the study area (Fig. 7E). Consequently, there is no clear relationship between migration distance and curvature for these channel centerlines.

In summary, in addition to autogenic meander development and channel-system aggradation, channel evolution can be explained by a combination of (1) salt bodies serving as topographic barriers to meander migration, and (2) salt deformation promoting local uplift, tilting, and basin subsidence, which directed channel pathways into low topography. Here, in the Campos Basin, resultant patterns of channelization diverge from those expected across a smooth, unperturbed seafloor, such as a tectonically quiescent continental rise. We discuss this below.

How Big Are the Channels?

The meandering channel system we have studied likely formed during the Late Cretaceous and/or Paleogene, when coarse sediment was shed off high-standing coastal mountains of southeastern Brazil. The widths of the individual channel elements reflect the characteristic discharge of the channel-forming turbidity currents, like fluvial channel dimensions are related to river discharge (Leopold and Maddock, 1953; Konsoer et al., 2013). To put the scale of these channel elements in context, we measured widths and meander wavelengths of some of the largest meandering channels on the seafloor and channel elements of the Miocene Dalia channel system, which is representative of hydrocarbon reservoirs in the subsurface offshore West Africa (Fig. 8; Abreu et al., 2003). Some of the largest rivers in the world deliver sediment to seafloor channels, which, in turn, disperse it across some of the largest submarine fans in the world, including the Amazon and Bengal Fans. The Dalia system received sediment from West Africa, prior to the Pliocene onset of the modern Zaire (Congo) River and submarine canyon-fan system (Ferry et al., 2005). The seafloor channels of the largest submarine fans in the world exceed the size of the Dalia channel elements (Fig. 8), which are similar in size to other subsurface channel systems in West Africa (e.g., McHargue et al., 2011). However, the Upper Cretaceous–Paleogene channel elements we studied in the Campos Basin are wider and their meanders are larger than those of many of the reaches of the Amazon and Bengal submarine channels and some of the largest river bends on Earth (Fig. 8). The size of the Campos Basin channel elements described here attests to the extraordinary sediment discharges of the turbidity currents that shaped them.

Similar scales of channel systems have been documented in other settings in which rapidly uplifting landscapes deliver voluminous terrigenous sediment to continental margins (Milliman and Syvitski, 1992) (Figs. 8 and 9). For example, offshore New Zealand, the Hikurangi Channel is a large (several kilometers wide), long (∼1500 km) sinuous submarine-channel system (Lewis, 1994; McArthur and Tek, 2021) receiving voluminous sediment from small rivers draining rapidly uplifting mountains since the late Miocene (Adams, 1980; Ota et al., 1996; Hales and Roering, 2009) (Figs. 8 and 9). In the subsurface, a large (3–5 km wide), sinuous submarine-channel system in the Molasse foreland basin, Austria (Upper Puchkirchen Formation; De Ruig and Hubbard, 2006), developed in response to an increase in early Miocene sediment supply from the tectonically active Alps (Kuhlemann, 2007; Sharman et al., 2018) (Figs. 8 and 9). More recently, hydrocarbon exploration efforts in the deep-water western Gulf of Mexico have uncovered a series of Miocene channel systems similar in scale to the Campos Basin channel system of this study (Winter et al., 2017; Winter, 2018; Clark et al., 2019; Sickmann and Snedden, 2021). Cenozoic uplift of the North American Cordillera and accelerated uplift of southern Mexico during the late Miocene (Witt et al., 2012; Molina-Garza et al., 2015; Villagómez and Pindell, 2021) increased sediment supply to the western Gulf of Mexico and promoted the development of this large deep-water depositional system, called the Veracruz Fan (Hessler et al., 2018). All that said, there are numerous tectonically active regions drained by small, steep rivers that do not supply sediment to giant meandering channels. For example, offshore southern California, USA, the La Jolla submarine canyon-channel system receives longshore drift–transported sediment (Covault et al., 2007; Normark et al., 2009) sourced from short rivers draining coastal highlands and the Peninsular Ranges (Inman and Jenkins, 1999; Warrick and Farnsworth, 2009a, b). However, coastal uplift and catchment-integrated denudation rates are much lower than in the previously cited examples (Kern and Rockwell, 1992; Covault et al., 2011) in which giant meandering submarine channels formed (Fig. 9). The formation of submarine canyon-channel systems in general, and giant meandering channel systems specifically, is related to factors in addition to tectonic activity in regions of small, steep rivers (e.g., Milliman and Syvitski, 1992). For example, along the tectonically active West Coast of the contiguous United States, the formation of submarine canyon-channel systems is related to greater yields of durable terrestrial clasts from rivers to the coast, with no significant correlation between canyon occurrence and the slope or width of the continental shelf (Smith et al., 2017). Determining the controls on canyon-channel formation, although beyond the scope of our study, has continued to be an active research topic since the last century (e.g., Shepard, 1936, 1981; Bernhardt and Schwanghart, 2021).

How Do Submarine Channels Fill Salt Basins?

The distribution and character of depositional elements in salt basins have been related to variations in gradient resulting from deformation. Channels can be deeply incised across steep salt-basin margins or the open slope and transition to more sheet-like deposits in flat basin floors or steps across the slope (Pirmez et al., 2000; Gee et al., 2007; Deptuck et al., 2012; Prather et al., 2012). Over time, irregular salt-basin topography can be healed by deposition or renewed by subsidence. This tug-of-war between sedimentation and subsidence creates basin fill comprising alternating depositional packages of channelized, “bypassing” systems with more sheet-like, “ponded” systems (Winker and Booth, 2000; Sylvester et al., 2015). The Brazos-Trinity salt-withdrawal basin 4 in the northwestern Gulf of Mexico is a commonly cited example of typical minibasin fill (Figs. 10A10C; Beaubouef et al., 2003; Mallarino et al., 2006; Pirmez et al., 2012; Sylvester et al., 2015). The upper ∼20 m of the shallow subsurface shows multiple weakly confined channel elements that feed submarine lobes (Fig. 10B; Beaubouef et al., 2003). A depositional strike–oriented seismic-reflection profile of the succession deposited since Oxygen Isotope Stage (OIS) 6 (<191 ka; Pirmez et al., 2012; Sylvester et al., 2015) shows a lack of channelization in the distal part of the basin (Fig. 10C). Three mud-dominated condensed sections with draping geometries bound two stratigraphic units that consist of turbidites and mass-transport deposits (Fig. 10C). Brazos-Trinity basin 4 is an example of a salt-withdrawal minibasin with a relatively large proportion of sandy channelized and lobe architectural elements; other minibasins can have much larger proportions of muddy mass-transport deposits and hemipelagites (e.g., the Fuji Basin consists of only ∼5% sandy channelized turbidites; Madof et al., 2009). The seismic-stratigraphic character of these salt-withdrawal minibasins is more layered than that of our study area in the Campos Basin.

In the Campos Basin, we analyzed a large meandering channel system across a region with salt-influenced topography (Figs. 10D10F). An isochore map of this system exhibits significant west-to-east variation, which indicates differential subsidence related to salt deformation (Fig. 3C). However, the overall character of depositional elements, through a sequence that is locally nearly 1 km thick, does not vary much through time. Moreover, the channel elements are very large, much larger than those described in other salt basins, such as the Brazos-Trinity basin 4 (Fig. 10). There is no evidence of sheet-like, lobe architectural elements interstratified with the giant channel deposits. However, it is possible that very thin lobes are not resolved in our seismic-reflection data. For example, Prélat et al. (2010) compiled dimensions of lobes from a range of deep-water depositional systems; individual lobes range from several to tens of meters thick, and numerous lobes can stack to form lobe complexes, in some cases >100 m thick (Prélat et al., 2010). Here, in the Campos Basin, relatively continuous, out-of-channel seismic reflections surrounding the giant channel fill are likely to be levee-overbank deposits, not lobes (Normark et al., 1993).

Salt deformation plays an important role in channel-system evolution in the Campos Basin (Fig. 7). Although the kinematics of freely meandering segments of the system are similar to the migration patterns observed in meandering rivers, the overall west-east trend, the dominant downstream translation, and the cutoff events that follow bend tightening seem to be the result of confinement by salt structures, which served as barriers that prevented channels from freely wandering across the seafloor (Fig. 7). Instead of confining the channel system and promoting pure aggradation resulting in a vertical stack of ribbon-like channel deposits, salt structures promoted downstream translation of channel bends. In addition to signaling the presence of seafloor relief by underlying salt bodies guiding channel evolution, changes in channel migration can also be used to reconstruct the deformation of salt-related structures in the Campos Basin. For example, the migration pattern of bend 4, initially expanding and then contracting, might be related to changes in the relative uplift of salt bodies IV and V (Figs. 6 and 7).

Channel migration patterns in the Campos Basin are different compared to those of a smooth, unperturbed seafloor, such as a tectonically quiescent continental rise. Figure 11 compares the depositional pattern of the freely meandering Bengal channel system to the channel system in the Campos Basin. A time slice through the Bengal channel system in the shallow subsurface shows meandering expansion, downstream translation, and cutoffs across a broad valley (Fig. 11A). This pattern is similar to models of freely meandering channel systems, like that shown in Figure 11B, created using meanderpy, a Python module that implements a simple numerical model of meandering (Howard and Knutson, 1984; Sylvester, 2018). This relatively simple model produces realistic patterns of long-term meandering. In the Campos Basin, only the earliest channel forms were far enough away from salt bodies to freely meander, and, consequently, they exhibit a relationship between migration distance and curvature with a lag of about two to four times channel width (Fig. 7A). Subsequent channel evolution in the Campos Basin was dominated by downstream translation between salt structures, which appear to have inhibited meander growth by expansion (Fig. 11C). This is similar to the migration pattern of rivers that are confined by banks resistant to erosion; Figure 11D shows the evolution of meanders interacting with the bank of the Mamoré River in the Amazon Basin, Brazil, since 1986. Figure 11D is a Landsat image downloaded from the U.S. Geological Survey Earth Explorer site (; accessed January 2021). The red-to-blue colors of the preserved channel deposits indicate the bar type of the deposits resulting from the channel migration (Sylvester et al., 2021); red indicates relatively sandy point bars, and blue indicates counter-point bars that are probably finer grained and are produced when the phase lag between curvature and migration rate results in significant downstream translation (Sylvester et al., 2021). Channel migration is dominated by downstream translation of bends and the deposition of counter-point bar deposits along the less-erodible bank of the Mamoré River valley. A close-up view of one of these bends (Fig. 11F) shows that it had a very similar evolution, with similar meander geometries, to bend 3 in the Campos Basin (Fig. 11E): (1) translation along an erosion-resistant margin; (2) blockage of the downstream meander limb by a promontory; (3) tightening of the bend as the upstream limb translates downstream; and (4) cutoff of the bend and formation of a new, relatively high-curvature bend that starts to translate downstream. This type of meander-bend evolution seems to be common along the edges of incised valleys of Amazonian rivers and results in high curvatures being associated with limited or no migration; that is to say, it is the reason behind some of the most pronounced outliers in the relationship between curvature and migration rate (Sylvester et al., 2019b). The same evolutionary pattern seems to be present in submarine channels of the Campos Basin, but the confinement is provided by salt-related structural highs.

Considering the importance of submarine channels as conduits for sediment delivery to the deep sea and their role in tectonically active salt-basin filling, we were motivated to better understand how they evolved in a region with salt-influenced topography in the deep-water Campos Basin. Specifically, we investigated whether submarine-channel deposits can be the primary depositional elements in salt basins where sediment supply overwhelms subsidence and, if so, whether the stratigraphic evolution reflects processes of meandering, such as systematic lateral migration, downstream translation, and cutoff of bends. We used 3-D seismic-reflection data to interpret paleochannel migration at a level of detail that approaches that of studies of meandering rivers in time-lapse satellite imagery. Salt-basin fill is composed of a continuous succession of some of the largest meandering channel elements ever recorded; this suggests that sediment discharge was large enough to sustain the meandering channel system in spite of large variability in subsidence across the region. We documented early migration of this channel system to be like that of a freely meandering river. However, as the submarine-channel system expanded toward nearby salt diapirs, its migration pattern fundamentally changed. With nowhere to expand further, the submarine channel translated downstream to produce a pattern characteristic in rivers that are confined by banks resistant to erosion. The channel migration pattern in the Campos Basin is different compared to that of a tectonically quiescent continental rise where meander evolution is unobstructed by salt diapirs. Moreover, this style of channelized basin filling is different from that of many existing examples of salt-withdrawal minibasins that are dominated by overall less-channelized deposits. Changes in channel migration can be used to understand the past deformation of salt diapirs and dispersal of sediment and pollutants across continental margins.

We thank PGS Suporte Logístico e Serviços Ltda for access to seismic-reflection data. We are grateful to Emerson for access to Paradigm SeisEarth interpretation and visualization software. We thank the sponsors of the University of Texas at Austin Bureau of Economic Geology (UT-Austin BEG) Quantitative Clastics Laboratory (QCL) ( We are grateful to UT-Austin BEG Applied Geodynamics Laboratory (AGL) Principal Investigator Mike Hudec and former post-doctoral researcher Dan Carruthers, who initiated collaborative QCL-AGL work and provided context of the salt tectonics and seismic stratigraphy of the Campos Basin. Gillian Apps and Frank Peel of the AGL also provided valuable insights into submarine channels in the Campos Basin. We thank Tim Lawton for geologic context of southern Mexico. An anonymous reviewer, Vittorio Maselli, Mark Deptuck, Michal Janocko, Andrea Fildani, and Geosphere Science Editor David Fastovsky provided thorough reviews that improved the manuscript.

1Supplemental Material. File S1: Base channel system horizon in comma-separated values (CSV) format. File S2: Intermediate channel system horizon in CSV format. File S3: Top channel system horizon in CSV format. File S4: Time steps between channel centerlines in Encapsulated PostScript (EPS) format. File S5: Graphics Interchange Format (GIF) animation of channel migration. Please visit to access the supplemental material, and contact with any questions.
Science Editor: David E. Fastovsky
Associate Editor: Andrea Fildani
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.