The canyon-channel has a centerline that is 1,118 km long, commencing approximately 70 km from the canyon head in the Congo Estuary. The channel reaches a water depth of 5,000 m (Figure 3A). The system’s overall thalweg gradient (S) averages around 3.5 m/km (0.20°), and this gradient decreases from ∼8 m/km (0.46°) at the canyon head, to ∼2 m/km (0.11°) around the middle of the system (at ∼400 km), and remains relatively constant until the channel’s termination (Figure 3B).

In contrast to previous assessments of the entire Congo Canyon-Channel (Babonneau et al., 2002), where the canyon/channel depth decreased uniformly, the bankfull channel depth (H _B ) of the active channel shows an initial increase followed by a nearly-constant channel depth (∼100 m) across most of the system. Only towards the channel’s termination (from ∼840 km on, Figure 3C), is there a decrease in channel depth, coinciding with a similar decline in cross-sectional area (A) (Figure 3E). The bankfull width (B) maintains consistency at about 1,200 m throughout the entire system (Figure 3D). Consequently, the median aspect ratio (B/H _B ) remains around 12 for most of the system (Figure 3F), aligning with typical values observed in submarine systems (Konsoer et al., 2013; Shumaker et al., 2018; Lemay et al., 2020; Hasenhündl et al., 2023).

Sinuosity (SI) ranges between 1.5 and 2 over large spatial steps (step-length of 250 km), and shows peak values reaching up to 5 at small spatial steps (step-lengths from crossover to crossover, 10 km and 24.5 km, Figure 4A). The width-to-curvature ratio (B/R) reveals regular maxima of 2 (Figure 4B), while the dimensionless meander wavelength (L/B) mostly remains below 5 (Figure 4C), consistent with typical values in submarine channels (Lemay et al., 2020; Hasenhündl et al., 2023), with irregular values at the end of the system due to the almost rectilinear course.

An asymmetrical behavior in levee/wall crest elevation along the active channel is evident through the disparity in right and left levee/wall crest elevations normalized by the channel depth ((z _R - z _L )/H _B ; Figure 4D), diverging from earlier descriptions of symmetrical levees (Babonneau et al., 2002; Baudin et al., 2017). The average levee/wall slope in the 1994/98 dataset ranges at around ∼16° across large portions of the system and around 24° in the higher-resolution 2019 dataset. This discrepancy is expected due to enhanced levee/wall mapping precision with increased resolution. However, in both datasets, the trend in levee slope consistently declines from around ∼840 km towards the end of the channel (Figure 4E). The channel’s asymmetry is described by the absolute offset between the thalweg and the centerline (n _T ), normalized by bankfull width (B). Throughout the system, this thalweg offset to the centerline (n _T ) reaches up to about 30% of the bankfull width (B) (Figure 4F).

Three distinct reaches can be delineated based on variations in sinuosity (SI) and slight variations in other characteristics, such as width-to-curvature ratio (B/R) and dimensionless meander wavelength (L/B). These reaches are the upper reach (0–290 km), the middle reach (290–900 km) and the lower reach (900–1,118 km). In the upper reach, SI averages around 1.5 (Figure 4A), B/R consistently peaks below 2 (Figure 4B) and L/B varies between 2 and 6 (Figure 4C). In the middle reach, there are occasional SI peaks reaching 5, but mostly averaging around 2 (Figure 4A), accompanied by B/R consistently maxes out at 2 (Figure 4B) and L/B remains mostly below 5 (Figure 4C). In the lower reach, both SI and B/R significantly drop (Figures 4A,B) and L/B becomes clearly irregular (Figure 4C).

The correlation between levee/wall crest superelevation ((z _R - z _L )/H _B ) and width-to-curvature ratio (B/R), previously described for sweep and swing meanders, exhibits a stronger pattern at the bend apices in the upper reach (β ₁ =0.27–0.33, R ² =0.39–0.43, Supplementary Table S1), followed by those in the middle reach (β ₁ =0.11–0.19, R ² =0.40–0.53, Supplementary Table S1). In the lower reach, where there are fewer sharp bends (B/R is significantly smaller than in the middle and upper reaches), there is less clarity due to scatter at low B/R, lacking a distinct trend in superelevation (Figures 9A–F, R ² =0.00–0.10; Supplementary Table S1).

With regard to the correlation between B/R and levee/wall slopes (α _R , α _L ), there are variations in the positioning of the maximum values within bends across different reaches. The highest wall slope asymmetry is displayed directly at the apices in the upper reach (Figure 9L, |β ₁ |=2.16–3.49, R ² =0.22–0.58; Supplementary Table S1). In contrast, the greatest asymmetry and outer levee slope occur 1 km downstream of the apices in the middle reach (Figure 9I), particularly in the higher resolution 2019 data (|β ₁ |=3.94–4.18, R ² =0.67–0.68, Supplementary Table S1). The results indicate that there is no significant difference between the values obtained directly at the apices and those obtained 1 km downstream in the lower reach. However it should be noted that the values for α _R and α _L are generally lower in the latter case (Figures 9G,H; Supplementary Table S1).

The thalweg offset to the centerline (n _T ) in the upper system reach does not show a pronounced trend (Figure 9Q, R, |β ₁ |=0.01–0.02, R ² =0.01–0.07; Supplementary Table S1). However, in the middle reach, particularly in the higher resolution 2019 data, the thalweg clearly moves 1 km downstream from the apices towards the outside of the bend (Figure 9O, β ₁ =−0.07, R ² =0.56; Supplementary Table S1). Despite the lower number of sharp bends in the lower reach, there is an observable correlation between curvature and thalweg offset 1 km downstream from the apices (Figure 9M, N, β ₁ =−0.05–0.13, R ² =0.35–0.47; Supplementary Table S1).

The identification of landslide-dams involves distinctive features such as increased thalweg elevation, a drop in thalweg gradient (S) followed by a peak downstream in S and decreasing levee/wall slopes (α _R , α _L ) upstream of the landslide. This identification of a set of morphometric fingerprints for landslide-dams through two independent characteristics (S and α _R -α _L ) allows for a systematic search for these features and can help assess their recurrence rate of these events. The newly identified landslide-dam at 40 km shows preservation qualities comparable to the more recent landslide at 95 km (Pope et al., 2022b), indicating a potential occurrence timeframe preceding the 1994 survey (approximately 5–20 years prior). The presence of additional potential older landslide-dams at 10 km and 190 km suggests that these events might be more frequent than previously assumed (Babonneau et al., 2004).

Typically, the appearance of meandering submarine channels has been described using sinuosity and individual cross-sectional shapes, drawing analogies with fluvial systems (Pirmez and Flood, 1995; Babonneau et al., 2010; Prerna and Mahender, 2020). The addition of information regarding the phase lag in thalweg position and levee slope offers insights into the sweep and swing dynamics of these meanders (Peakall et al., 2000; Straub et al., 2008; Sylvester et al., 2011; 2019). It was presumed that if the maximum thalweg movement and the steepest levee slope occurred at the B/R apex, then the maximum erosion would occur there, and the primary movement would be directed laterally outwards at the apex (Babonneau et al., 2010). Contrary to this assumption, the presence of a 1 km phase lag indicates that the point of maximal erosion of the levees and the thalweg is slightly downstream of the bend. This leads to an additional downstream component of its movement, contributing to a sweep and swing behavior akin to rivers with high curvature (Sylvester et al., 2019). While this behavior may not be directly observable in the repeated bathymetry due to the 100 m resolution of the 1998 data, it becomes apparent when examining the terraces (Figures 6D,F) and comparing the area upstream of the repeated survey (290–725 km, Figure 3). Despite the consistency in morphometric characteristics, the sinuosity increases in the older channel upslope. Essentially, as the channel becomes older, lateral spreading of the meanders intensifies until they encounter limitations imposed by the canyon further upstream.

Knickpoints, identifiable by thalweg gradient peaks (Pirmez and Flood, 1995; Babonneau et al., 2002; Heiniö and Davies, 2007; Heijnen et al., 2020; Guiastrennec-Faugas et al., 2021), can now be differentiated from steep drops caused by damming landslides due to unaffected levee/wall slopes near autogenic knickpoints. The suggested correlation between knickpoints and avulsions by Babonneau et al. (2002) lacks confirmation, with the most noticeable knickpoint located downstream of the youngest avulsion, which exhibits a migration rate of 15 km per 20 years (∼750 m/yr), contrasting with the timescales of avulsions (thousands of years) (Picot et al., 2019).

Sudden changes in morphometric characteristics (e.g., bankfull channel depth, sinuosity, width-to-curvature ratio, meander wavelength, levee slope), which are quite homogeneous over some sections, serve as robust indicators of avulsion, aligning well with previously documented avulsions (Picot et al., 2019). It appears that only the more recent avulsions (situated downstream of 744 km and therefore younger than ∼6 kyr (Picot et al., 2019)) are distinguishable in the morphometric characteristics and older avulsions (located upstream of 744 km and thus older than ∼11 kyr (Picot et al., 2019)) are no longer discernible from the current morphometric data.

This poses an interesting question, which requires further exploration: how enduring are the identified morphometric fingerprints (Table 1) over time? Additional future time-lapse bathymetric surveys could provide insights into the persistence of these fingerprints. Coupled with measurements of turbidity current activities, this approach might offer future perspectives on comprehending flow activity and channel-modifying processes in other systems solely through bathymetric mapping.

Until now, the identification of landslide-dams relied on a few repeated surveys for qualitative assessments (Biscara et al., 2013; Pope et al., 2022b; Talling et al., 2022), or examinations of landslide scars or deposits (Babonneau et al., 2002; Smith et al., 2005; Mountjoy et al., 2009; Pope et al., 2022b). These methods are time-consuming and subject to individual judgement and interpretation. The identification of the previously unidentified landslide-dam at 40 km suggests that it is possible to systematically identify landslide-dams using single bathymetric surveys alone.

Comparison of bathymetric data at different resolutions shows that the higher resolution of 15 m in 2019 provides more accurate cross-sectional characteristics, particularly for levee/wall slopes (Figure 4E; Figure 9G–L; Supplementary Table S1) and thalweg position (Figure 4F; Figure 9M–R; Supplementary Table S1). However, lower resolution data (100 m in 1994/98) still reveals discernible trends and changes in morphometric characteristics, although with lower absolute values (e.g., mean levee/wall slope in the 1994/98 data is ∼16° and in the 2019 data ∼24°). Longitudinal characteristics (e.g., sinuosity, curvature or wavelength) are less affected by data resolution when a centerline is used as a reference axis, providing advantages over thalweg-based analyses (Shumaker et al., 2018; Bührig et al., 2022). This approach not only measures thalweg movement but also serves as a more independent reference axis as thalweg determination is highly dependent on resolution.

This makes the approach suitable for systems with low-resolution bathymetric data (Bührig et al., 2022). While absolute values at the cross-sectional level are resolution-dependent (Hasenhündl et al., 2023), trends and changes in characteristics provide valuable insights. The revealed fingerprints (Table 1) facilitate a systematic and efficient search for channel-modifying features within bathymetry data, without relying on visual inspection.

The examination of the bankfull active channel and identified channel-modifying features allowed for the categorization of the Congo Canyon-Channel into three distinct morphodynamic reaches, as shown in Figure 10. It is noteworthy how constant the bankfull width (B) of the active channel remains throughout the Congo submarine system’s extent (Figure 3D), while other morphometric characteristics vary across distinct reaches (Figure 3) (Hasenhündl et al., 2023). The upper reach (0–290 km) of the system within the incised canyon is characterized by higher thalweg gradients (S) (Figure 3B) and a marginally reduced cross-sectional area (A). Presuming, bankfull discharge, this likely reflects the higher driving forces for turbidity currents due to the steeper thalweg gradient (Konsoer et al., 2013). Evidently, these turbidity currents appear as erosive flows (Babonneau et al., 2004; Talling et al., 2022) typified by a dense frontal cell (Azpiroz-Zabala et al., 2017; Simmons et al., 2020; Pope et al., 2022a). Consequently, these flows erode the thalweg and the outer levees at the bend apices (Figure 9L), resulting in oversteepened active channel walls inside the canyon (wall slopes) and contributing to the triggering landslides and slumps from the canyon margins (Figure 10D). Landslides may be more common than previously thought (Babonneau et al., 2004). They have a significant impact on the active channel morphology within the canyon by damming the active channel, creating meander cut-offs and terraces (Pope et al., 2022b).

The steep drop downstream of a landslide dam (i.e., the knickpoint) can migrate upstream when subsequent turbidity currents travel over it once the area behind the blockage has been infilled. The migrating drop cuts into the backfill, creating terraces and re-excavating the infill material, similar to other knickpoints (Pope et al., 2022b; Talling et al., 2022). These channel-damming landslides likely occur on timescales of decades within the Congo Canyon-Channel, as evident from two similar-sized, well-preserved landslides at 40 km and 95 km. The frequency of such events in other canyon systems will likely depend on sediment delivery, tectonic regime, and local substrate characteristics. These blockages can store substantial quantities of sediment and other material. For instance, the landslide-dam at 40 km retains ∼170 ± 40 Mt of sediment (Pope et al., 2022b), equivalent to several years of sediment transported by the Congo River, which has an annual sediment flux of ∼43 Mt (Milliman and Farnsworth, 2011; Pope et al., 2022b). Since these landslides leave fingerprints in the morphometric characteristics over decades, it is improbable that the current erosive flows can completely mobilize the accumulated material. Consequently, these landslides have a lasting impact on the morphology, akin to the impacts observed in high alpine rivers and landslides (Korup, 2006). Even long runout flows observed in January and March 2020, did not excavate most of the infill upstream of the 95 km landslide-dam (Talling et al., 2022). Hence, a portion of the trapped material remains permanently deposited in newly formed terraces through the cutting of a new channel, while other parts are temporarily stored and stepwise transported through upstream migrating steep drops, ultimately creating a new channel (Smith et al., 2005; Pope et al., 2022b).

Downstream of the upper canyon reach (290–900 km), morphometric characteristics remain uniform except for sinuosity (SI), which initially jumps to its highest values and then decreases down-system (Figure 4A). This is remarkable because different sections of varying ages, separated by avulsions (Picot et al., 2016), display similar thalweg gradient (S) (Figure 3B), channel depth (H _B ) (Figure 3C), cross-sectional area (A) (Figure 3E), width-to-curvature ratio (B/R) (Figure 4B) and dimensionless meander wavelength (L/B) (Figure 4C). We interpret this uniformity as an equilibrium state for turbidity currents, completely contained within the channel. There are no obvious signs of excessive overspill (Figures 6D,F). This equilibrium state, free from the constraint of the canyon, facilitates the free formation of meanders. As sections of the channel become older, meanders become more pronounced (higher SI upstream, Figure 4A) (Kastens and Shor, 1986; Babonneau et al., 2010), resulting in sweep (downstream movement) and swing (outward movement) meanders, which are typical features observed in models (Peakall et al., 2000; Sylvester et al., 2011) and high curvature rivers (Sylvester et al., 2019) (Figure 10C). In contrast to the canyon upper reach, there is little evidence that sediment and material are trapped locally. Instead it is continuously deposited by the meandering movement on the inner margins (point bars) (Babonneau et al., 2010) and eroded from the outer margins or bypassed by the turbidity current throughout the entire reach. No evidence is apparent in the bathymetric data of side-wall collapses in the middle reach comparable with those observed in seismic cross-sections from other systems (Lewis et al., 1998).

In the youngest part of the channel, located further downstream (900–1,117.5 km), all characteristics change except for thalweg gradients (S) and bankfull width (B) (Figure 3). These changes support the views on the age and development of submarine channels (Konsoer et al., 2013; Shumaker et al., 2018; Jobe et al., 2020; Hasenhündl et al., 2023). Submarine channels begin relatively straight with a fixed channel width (Figure 3D) while their channel depth (H _B ) gradually increases (Figure 3C) due to continuous turbidity current overspill, as indicated by the sediment waves. (Figures 7D,F; Figure 10B). Consequently, the slope of the levees develops due to levee aggradation (Figure 4E). Once a certain H _B and age are reached, turbidity currents are confined by the channel and lateral movements begin. This leads to the development of meanders, as observed upstream in sinuosity (SI) (Figure 4A) and width-to-curvature ratio (B/R) (Figure 4B).

In this straighter reach, probable upstream migrating knickpoints become the common channel-modifying features (Figure 10B), observed for the first time at this water depth (ca. 4,600 m). These knickpoints may form at the transition between channel and lobe (Heijnen et al., 2020) and exhibit upstream migration at a rate comparable to other submarine channels (a few kilometers per decade). This rate is much faster than the 0.001–1 m/year in rivers (Guiastrennec-Faugas et al., 2020; Heijnen et al., 2020). However, no knickpoints have been identified upstream in the middle reach (290–900 km), where swing and sweep of meander evolution dominates. This is in contrast to other systems, such as Bute Inlet or Knight Inlet (Gales et al., 2019; Heijnen et al., 2020), where knickpoints have been observed. In those systems, knickpoints appear to be able to migrate throughout the entire channel reach. It is therefore surprising that these features fail to do similar in the Congo Canyon-Channel, although local disturbances have led to their generation in the upper reach (e.g., landslides, (Pope et al., 2022b)). In the lower reach, their upstream migration leads to stepwise material transport along the channel bed (Heijnen et al., 2022), accompanied by deposition due to levee overspill and on the lobe.

The morphodynamic reaches identified in this study are characterized by common channel-modifying features. The results show that sediment and material transport downslope within a canyon-channel system occurs at staggered intervals and can differ across various morphodynamic reaches. This enhanced understanding of the Congo submarine system’s morphodynamics, which is likely applicable to similar systems, can help better position seabed infrastructure such as cables. Identifying specific morphodynamic reaches is critical to prevent damage to infrastructure, as processes such as channel-damming landslides or knickpoints can cause significant local erosion that could undermine cables (Talling et al., 2022).