At a large scale, subsurface imaging reveals the typical components of channel-levee systems, with channel floor deposits and overbank wedges. The most part of the channel axis has a transparent, highly-reflective facies, suggestive of coarse-grained channel fill deposits (Figures 7, 8). The only exception occurs at the channel mouth where the coarse-grained sediments of the channel-axis infill are buried under a homogeneous fine-grained deposit (TL) (Figure 8C).

The overbank deposits consist of a thick sedimentary wedge that tapers laterally and longitudinally. The left overbank wedge consists of relatively fine-grained sediments (TL) (Figure 8). At the base of the Acquarone high slope, a recent MTD can be recognized thanks to the rough seabed in the channel floor; according to its transparent, faintly reflective facies, it consists of a fine-grained MTD (Figures 8A,B).

In the right levee, the overbank deposits show a less homogeneous character, with varied acoustic facies. In the Bend area (Figure 7A), most of the units (I, II, III, VI and VII) are composed principally of well stratified fine-grained sediments (PL), while Units IV and V have a major coarse-grained component (PH). In the Laterally confined area (Figure 7B), there is a prevalence of well stratified fine-grained deposition (PL). At the end of the confined tract, in the Knick-points area, the left overbank deposits are composed by homogeneous fine-grained sediments (TL), while the right overbank by well-stratified fine-grained sediments (PL) (Figures 8A,B).

The seven seismic units recognized in the overbank wedges extend downslope, in the frontal splay and fan fringe elements. The frontal splay and the fan fringe are composed of different acoustic facies that vary longitudinally and laterally. The first 8 m (Units I, II, and III) of the Frontal splay area consist of relatively homogeneous fine-grained sediments (TL) (Figure 9). This low-amplitude package drapes more reflective deposits interpreted to be composed of relatively coarse-grained sediments (TH) of Units IV and V (Figure 9A). Their high reflectivity strongly attenuates the acoustic signal, making the recognition of the deeper units uncertain, which are thus characterized only on the basis of some minor reflections. In the fan fringe, the downslope continuation of Units IV, V, and VII present irregular high-amplitude reflectors interpreted as laterally discontinuous coarse-grained sediments, laterally the reflectors have a parallel configuration and slightly decrease in amplitude (PL) highlighting better organization and an increase in the fine content (Figure 9C).

Bends are crucial morphological elements that affect sediment distribution in the channel and overbank areas. In this case study, the origin of the bend is connected to the presence of the Acquarone high that deviates the channel pathway from a N-S to a NW-SE direction. The lateral confinement imposed by the Acquarone high led to a large amount of sediment discharge outside the bend on the eastern right levee (Figure 10).

The backslopes of channel levees are in general thought as sites of fine-grained turbidity currents and of predominantly mud and silt deposition (Mutti and Normark, 1987; Normark et al., 2002; Wynn and Stow, 2002; Hansen et al., 2015). In a current able to transport various grain sizes, the coarse sediment fraction is concentrated near the base of the current, while finer sediment fractions are more homogenously distributed over the height of the current (Kneller and Buckee, 2000) (Figure 13). Progressive levee construction increases flow confinement and, accordingly, causes reduction in the grain-size of the sediment that can overspill the channel. As a consequence, an overall fining-upward trend generally characterizes the overbank sequence (Hansen et al., 2015).

On the outer side of the bend of the Acquarone channel, despite the high relief of the levee crest (about 60 m), Units IV and V (Case I) are characterized by a high-reflectivity interpreted as deposits with a percentage of coarse-grained sediments (Figure 7A). There is a drastic change of facies between Unit IV and the overlying units belonging to Case II consisting in 8.5 m of relatively fine-grained deposit, as inferred from the low acoustic reflectivity (Figure 7A). The passage between PH and PL facies is sharp and abrupt, indicating that a sudden change in the proprieties of the flows occurred after the deposition of Unit IV. According to Baas et al. (2005), highly-turbulent flows keep the coarser particles more easily in suspension, owing to an upward directed component of turbulent velocity that overcomes the falling velocity of most particle sizes. This effect would eventually cause a decrease in the vertical segregation of the different grain-sizes compared to flows with weaker turbulence. In the study area, the high gradient of the confined tract (1.15°, Figure 4B) and the centrifugal force at the bend are consistent with an increase in the flow velocity, which can develop high turbulence. So, we suggest that, in Case I, whose parent flows have a high percentage of sand fraction, the increased turbulence enhanced the coarser fraction to reach the upper part of the flows and overspill. Therefore, the overbank deposit created by high-density turbidity currents has a higher sand to mud ratio than the product of low-density turbidity currents, which does not account for a significant coarse-grained fraction to be transported over the levee crest (Figure 13B).

Sediment waves in the Bend area, in their early phase (Case I), are associated with parallel reflector with a high-reflectivity and have slow rates of up-current migration (Figure 5C, 13A). Anyway, later (Case II), the migration rate diminishes, the seismic response becomes less reflective and pure aggradational growth prevails (Figure 5C, 13B). Generally, up-current migrating sediment waves are associated with thicker and more abundant sand beds on their stoss side (Normark et al., 2002; Kostic, 2014). This general observation is consistent with a higher sand content in the overflow connected to Case I, further substantiating that they can be associated with high-density turbidity currents.

Generally, in a leveed-channel system, the upper part of turbidity currents continuously over-spills onto the overbank area and the flows within the channel are progressively depleted of fine-grained material (Peakall et al., 2000; Hansen et al., 2015). As a consequence, the thickness of overbank units is expected to decrease in a downstream direction. On the contrary, in the Laterally confined area of the right levee, Units II, III (Case II), and V (Case I) have higher thickness than upslope (Figure 10). This unusual and localized increase in sedimentation can have two different explanations:

1) A localized reduction in the channel depth value enhances the overspill of a thicker part of the flow. Different depositional features may reduce the channel relief, such as nested mounds that develop on the outside of a specific bend (i.e. the first sharp bend) (Peakall et al., 2000) or MTD coming from the surface of the Acquarone high. However, those feature cannot be recognized, as, if present, they are now buried under the more recent MTD (Figure 2).

In the case of the first hypothesis, the overspill of the lower, coarser-grained part of the flows should result in a high-amplitude seismic facies, similar to what observed in the high-density flows of Case II. This is not the case, as shown by Figure 7B, which highlights the low-amplitude parallel reflections of the overbank wedge. Therefore, we prefer the second hypothesis also because it is consistent with the spreading direction of the lobe-shaped depocenter in the Laterally confined area that is almost specular to the channel path at the bend. It can be hypothesized that the confining action of the Acquarone high could impose the thickening and the successive rebound of the upper portion of the flow outgoing from the bend (Figures 12B,D). In fact, a turbidity current can be reflected depending on the geometry of the bounding slope and the angle of incidence of the current (Patacci et al., 2015). This assemblage has commonly been detected in tectonically confined basins (Kneller et al., 1991; Kneller and McCaffrey, 1999; Amy et al., 2004; Tinterri et al., 2016, 2017; Tinterri and Piazza, 2019) resulting in the production of contained-reflected beds (sensu Pickering and Hiscott, 1985). Dalla Valle and Gamberi (2009) and Tek et al. (2021) provide other examples where changes of the overbank architecture are interpreted to be connected to flow reflection processes.

Laboratory experiments (Kneller et al., 1991; Kneller 1995; Haughton 1994; Amy et al., 2005; Patacci et al., 2015) and numerical models (Howlett et al., 2019) have shown that a reverse overflow forms as soon as a turbidity current impacts on a topographic high and a deceleration from supercritical to subcritical condition is accompanied by abrupt flow inflation. However, in order to produce a reflected overflow, the upper part of the flow should have sufficiently high momentum, so when the outbound velocity is too low at the point of incidence with the slope, turbidity-current reflection is not achieved (Patacci et al., 2015). In the case of the Acquarone channel, it is reasonable to argue that turbidity currents that spilled over the left side of the channel experienced complete reflection, being impeded in their travel by the Acquarone High (Figure 12B). However, the sedimentation patterns of Units I, IV, VI, and VII do not show this depocenter (Figure 12A). Probably the turbidity currents that generated those units had insufficient velocity due to their strongly depositional behaviour at the bend. In fact, as the current enters the bend and spills over onto the overbank, the in-channel velocity decreases significantly (Abd El-Gawad et al., 2012). Other flow proprieties, such as the volume of the low-density part of the flow and the height of the overspilling flow, could be involved in the occurrence of a reflected overflow, but they cannot be evaluated with the data presently available.

The relief of levees above the channel floor commonly diminishes in a downslope direction (Hiscott et al., 1997; Posameinter and Kolla, 2003; Hansen et al., 2015). In our case study, along the flat segments (flat I-II, Figure 4C) downslope from knick-points, the relief of the levees crest increase (respectively about 7 and 3 m for the right levee and 5 and 4 m for the left one) vs. the upslope levee tract (Figure 4C). This increase in relief could be due to a localized increase in the erosional behaviour of turbiditic flows in the channel floor or to an increase in the growth rate of levee crests. The latter hypothesis is here preferred because the flat I segment on the levee crest profiles coincides with the upslope apex of a depositional field of bedforms (Figure 12A).

In the Knick-points area, the units that fall into Case I show a marked depocenter highlighting increased capability of flows to overtop the levee crest (Figure 12A). A reduction in flow velocity is to be expected when a flow passes through a knick-point (Heijnen et al., 2020). Considering the inverse relationship between flow velocity (U) and flow height (h) marked by the Froude number Fr = U ( ρ − ρw ρg ) h , where ρ = density of the flow, ρw = density of water and g = acceleration of gravity, flow thickening is connected to flow deceleration.

In the Acquarone channel, slopes are always one order of magnitude above the 0.1° threshold gradient for Froude supercritical flow deposits (Postma et al., 2014). This is why it is reasonable to assume that, irrespective of their nature, all the turbidity currents reached the supercritical conditions and, passing the knick-point, experienced a hydraulic jump. We suggest that in the flat I segment (Figure 4C), where the slope of the channel axis is halved, the turbidity currents rapidly slow down and go through a hydraulic jump, whose consequence is the thickening of the flow and the entrainment of water that reduce the density of the current (Figure 14) (Komar, 1971; Postma et al., 2009).

Recent experiments by Ono et al. (2021) were carried out on a mixture of sediment (including fine-sand to gravel), simulating flows equivalent to the high-density turbidity currents according to the definition of Lowe (1982). The results of Ono el al. (2021) show that after the occurrence of a large hydraulic jump, where the flow depth increased eight time respect the initial value, smaller ones, where the flow was only two times thicker than the initial value, occurs. We have interpreted that the flow behaviour of Case I in the KP1 could thus be connected to a drastic increase in flow thickness. This process can account for a long-lasting overspill and for the formation of the depocenter associated with the Knick-points area (Figure 12A). Anyway, only Unit VII show a depocenter occurring after the KP2. Accordingly with the data of Ono et al. (2021), we interpreted the absence of a depocenter after KP2 for Unit V and IV to be connected to a possibly smaller hydraulic jump whose effect are potentially not related to a long-lasting overspill. The smaller size of the jump can be connected to a slow-down of the velocity of the flow following the first hydraulic jump and to the milder slope difference between the second ramp (1, 2°) and the second flat (0, 9°) that could have not provided sufficient deceleration to the flow.

Similarly, Case II, interpreted as associated to low-density turbidity currents, does not show any increase in sedimentation in the Knick-points area (Figure 12B). This occurrence is here explained as also resulting from a smaller hydraulic jump. In the crossing of a topographic step, the height of the jump has been described as the ratio of Froude number before (Fr_pre in Figure 14) and after the jump (Fr_post in Figure 14). The higher the difference between the Froude number values (∆Fr in Figure 14), before and after the jump, the higher the size of the hydraulic jump (Cartigny et al., 2014).

The lower part of bipartite turbidity currents consist of a relatively dense basal layer sustained by the combined effect of hindered settling, matrix strength and dispersive pressure controlled by sediment concentration (that may exceed 20–30% according to Lowe, 1982). In contrast, a dilute, turbulent low-density turbidity current is primarily sustained by turbulence which is largely independent from particle concentration (Lowe, 1982). The flow deceleration related to the hydraulic jump, in high-density turbidity currents, causes the en-masse deposition of the coarsest particles of the basal layer (Postma and Cartigny, 2014; Postma et al., 2014; Hage et al., 2018; Slootman and Cartigny, 2020) and the concentration of the basal dense layer is further reduced by the water entrainment effect (Figure 14A). As a consequence, we can expect that high-density turbidity currents experience a drastic slowdown after KP1, due to the marked decrease in sediment concentration, producing a strong hydraulic jump with the consequent increase in sediment deposition on the levee (Figure 14A). While low-density turbidity currents, whose principal transport mechanism is the turbulence of the fine-grained sediments, are expected to experience a milder velocity reduction after KP1 producing a weak hydraulic jump (Figure 14B).

In summary, we suggest that the pattern of sediment distribution created by Case I in the Knick-points area, consisting in a localized increase in the overbank thickness, is the result of high-density turbidity currents experiencing a strong hydraulic jump (Figure 12C); while the sedimentation pattern of Case II does not generate a depocenter in the same area since their former flows (low-density turbidity currents) are interpreted as producing smaller and weaker hydraulic jumps (Figure 12D). We have interpreted the knick-points, inferred by tectonic structures transverse to the channel course whose topography is not completely healed, led to the occurrence of hydraulic jumps whose amplitude, that can be connected to the density stratification of the flow, may or not have a depositional expression on the overbank area.

The thickness distribution of the seven units shows that the two different types of flow behave differently also under the rapid flow expansion regime ensuing at the end of the channel confinement. Units falling in Case I present a thick and elongate lobe with very high reflectivity (Figure 12A). This kind of deposition is interpreted as resulting from high-density turbidity currents that, exiting the channel confinement, rapidly lose competence, deposit the coarsest material and create a thick frontal lobe with small-scale coarse grained sediment waves in front of the channel mouth (Figure 6). Moreover, the direction of the crests is indicative of another important behaviour of the lower dense portion of the turbidity current. Where the relief of the bulge in the frontal splay becomes important (Figures 4A, 9B), the sediment wave crests divert by about 30° eastward from their initial WWN-EES orientation (Figure 6). Similar results can also be observed in field studies and flume experiments related to flow deflection against lateral or oblique confining slopes (Amy et al., 2004; Tinterri et al., 2017; Piazza and Tinterri, 2020). The mud portion is completely absent, as shown by the transparent high reflectivity of the frontal splay (Figure 9A), indicating that the upper dilute part of the turbidity currents largely bypassed this area or were eroded by successive flows.

Units belonging to Case II do not present any thick concave upward accumulation, but, rather, deposition on the frontal splay results in a low-reflectivity package, which drapes the underlying morphology (Figure 12B). Sedimentation is higher towards the SW, in the fan fringe, than adjacent to the channel mouth where much of the sediment bypasses and is carried further downslope in the Gioia channel. The loss of confinement at the end of the channel is generally associated with an increase of turbulence in turbidity currents. The increase in turbulence at the mouth of submarine channels has been connected both to the occurrence of hydraulic jump (Mutti and Normark, 1987; Mutti and Normark, 1991; Mutti et al., 2003; Baas et al., 2004) and to flow relaxation mechanism (Pohl et al., 2019). In turbidity currents with a substantial mud component, as low-density turbidity currents, increase in turbulence, which, preventing sediment deposition, contributes to the flow run-out (Mutti and Normark, 1987; Mutti and Normark, 1991; Mutti et al., 2003). Moreover, the high value of the slope (1.2°, Figure 4B) and the convergence created by the topography in the frontal splay caused turbidity currents to accelerate in the principal flow direction, while laterally, in the fan fringe, the gentler slope caused a little bypass and a greater amount of fine-grained deposition. In fact, low-density turbidity currents react to the end of confinement and to the morphology of the slope in the same way as the upper dilute part of high-density turbidity currents.