Journal of Sedimentary Research, 2010, v.

80, 710–727
Research Article
DOI: 10.2110/jsr.2010.067

A TEST OF INITIATION OF SUBMARINE LEVEED CHANNELS BY DEPOSITION ALONE

JOEL C. ROWLAND,1 GEORGE E. HILLEY,2 AND ANDREA FILDANI3

1
Earth & Environmental Sciences Division, Los Alamos National Laboratory, MS-D401, Los Alamos, New Mexico 87545, U.S.A.
, 2Department of Geological and Environmental Sciences, Braun Hall, Stanford University, Stanford, California 94305, U.S.A.
3
Chevron Energy Technology Company, 6001 Bollinger Canyon Road, San Ramon, California 94583, U.S.A.
e-mail: jrowland@lanl.gov

ABSTRACT: Leveed submarine channels play a critical role in the transfer of sediment from the upper continental slopes to
intraslope basins and ultimately deeper marine settings. Despite a reasonable understanding of how these channels grow once
established, how such channels are initiated on previously unchannelized portions of the seafloor remains poorly understood. We
conducted a series of experiments to test whether leveed channels can start by deposition on a planar rigid bed. We
systematically varied the current density and outlet velocity to explore the relative influence of inertia and excess density on the
depositional dynamics of currents entering a basin and undergoing abrupt unconfinement. Under flow conditions ranging from
supercritical to subcritical (bulk Richardson numbers of 0.02 to 1.2) our experiments failed to produce deposits resembling or
exhibiting the potential to evolve into levees needed to create a self-formed channel. In the absence of excess density, a
submerged sediment-laden flow produced sharp-crested lateral deposits bounding the margins of the flow for approximately a
distance of two outlet widths down-basin. These lateral deposits terminated in a centerline deposit that greatly exceeded
marginal deposits in thickness. As excess density increased relative to the outlet velocity, the rate of lateral spreading of the flow
increased relative to the downstream propagation of the density current, transitioning from a narrow flow aligned with the channel
outlet to a broad radially expanding flow. Coincident with these changes in flow dynamics, the bounding lateral deposits extended
for shorter distances, had lower, more poorly defined crests that were increasingly wider in separation than the initial outlet, and
progressively became more oblong rather than linear. Based on our results and a review of previous experimental and numerical
models, we suggest that initiation of leveed channels from sediment-laden density currents traversing non-erodible beds is unlikely.
Partial confinement of these currents appears to be necessary to establish the hydrodynamic conditions needed for sediment
deposition along the margins of a density current that ultimately may create confining levees. We suggest that erosion into a
previously unchannelized substrate is the most likely source of this partial confinement.

INTRODUCTION
Submarine canyons and their down-dip extensions as channel systems
are the conduits through which terrigenous detritus is transferred into
deep sea (Shepard 1948). In a very broad sense, submarine channel
systems can be divided into two classes: (1) those with negative relief
relative to the surrounding seafloor (gullies and furrows in the sense of
Fedele and Garcia 2009), and (2) channels bounded by positive relief,
specifically wedge-shaped deposits that laterally bound channels and rise
above the surrounding seafloor, commonly referred to as levees (Fig. 1).
These levees are commonly built by deposition from turbidity currents
spilling out of the channel, and the intrinsic relationship between
throughgoing channels and constructional levees has been explored by
different authors (Piper et al. 1999; Skene et al. 2002; Straub and Mohrig
2008, among others). Overall, significant understanding of the processes
that control deposition and morphology onto established channel–levees
systems has been achieved (e.g., Piper and Normark 1983; Peakall et al.
2000; Deptuck et al. 2003; Pirmez and Imran 2003; Kane et al. 2008; Straub
et al. 2008); however, the processes that control the initiation of leveed
channels remain poorly understood. Considerable insight into development
of individual and regularly spaced systems of channels with negative relief
has been gained through physical experiments (e.g., Fedele and Garcia
2009), seafloor observations and numerical modeling (e.g., Fildani et al.
2006; Lamb et al. 2008), and stability analysis of sediment-laden density
currents (e.g., Izumi 2004; Hall et al. 2008). The knowledge gap on levee
initiation persists despite extensive documentation of these features in
seismic imaging of the seafloor, examination of outcrops, and decades of
investigations on the depositional responses of sediment-laden density
currents entering experimental basins (Table 1 and references therein).
The early development of leveed channels on previously unchannelized
portions of the seafloor has been interpreted in seismic imaging of the
Brunei slope (Straub and Mohring 2008), offshore of Angola (Gee et al.
2007), and in the Benin-major channel–levee system (Deptuck et al. 2003).
Interpretation of high-resolution seafloor imaging of the Amazon fan
suggests that channel development occurred rapidly across newly
deposited fan lobes since the Last Glacial Maximum (Jegou et al.
2008). Images of the Amazon channel–lobe systems shows master-leveed
channels feeding the lobes and small distributary channels across the
lobes, but image resolution is insufficient to show the transition from
channels to lobes or to decipher bed-scale architecture needed to make
process-based interpretations of channel development (Jegou et al. 2008).
In theories and models for the formation and spacing of submarine
channels with negative relief, the incipient channels develop in response to

Copyright E 2010, SEPM (Society for Sedimentary Geology) 1527-1404/10/080-710/$03.00

JSR INITIATION OF SUBMARINE LEVEED CHANNELS
instabilities within a much larger (relative to the channels) unconfined
current (e.g., Izumi 2004; Hall et al. 2008; Fedele and Garcia 2009; Straub
and Mohrig 2009). Other than submarine canyons and with the exception
of models suggesting the development of erosional channels in response
to cyclic steps (Fildani et al. 2006; Lamb et al. 2008; Normark et al. 2009)
few channels with only negative relief capture or route a significant
fraction of the current responsible for their formation. In contrast, leveed
channels are an important component of deep-water systems and major
conduits for the transport and routing of turbidity currents from upslope
source regions to deepwater depocenters (Shepard 1948; Menard 1955;
Normark 1970; McHargue 1991; Pirmez 1994; Posamentier and Kolla
2003; among others). Understanding the dynamical properties of
turbidity currents responsible for the development of leveed channels
(and the way levees start) is therefore critical to deciphering seismic and
geological records of these deposits in a mechanistic or process-based
context. For example, does the observation of leveed channels on regions
of the seafloor previously lacking leveed channels mark a change in the
character, frequency, magnitude, and composition of turbidity currents
sourced from upslope? Or alternatively, does development of leveed
channels arise inherently as an autocyclic response to seafloor deposition
with little regard to upslope controls? Additionally, how leveed channels
evolve across the seafloor (forward-stepping, back-stepping, or across
vast stretches quasi-synchronously) has dramatic implications for the
spatial distribution and stacking of sand bodies deposited in the channel
at its terminus.
Imran et al. (1998) argued that for leveed channels to develop, net
deposition along the margins of a sediment-laden flow must exceed that in
the core of the flow. This condition may be met under conditions that are
entirely depositional or are net-erosional in the flow center and
depositional on the margins (Imran et al. 1998). A common feature of
submarine leveed channels observed in seismic reflection panels is that the
‘‘wedge’’ of sediments constituting the levees sits conformably over flat
seismic reflectors of the pre-existing fan or slope surface while bottom-
most reflectors in the axis of the channel are incised below the base of the
levee wedge (Fig. 1). Skene et al. (2002) identified close numerical
relationships between the channels and related levees in a wide range of
geological environments and time span. Based on seismic data alone,
however, because of the low preservation and subsequent destruction of
the earliest morphologies, it is impossible to identify the early incision and
if it preceded, followed, or occurred coincident to levee development.
A number of numerical and physical experiments have produced
deposits with the basic condition of greater deposition along the margins
of a sediment-laden density current (Imran et al. 1998; Imran et al. 2002;
Bradford and Katopodes 1999a, 1999b; Metivier et al. 2005; Baas et al.
2004). However, to judge the applicability of experimental results to
understanding the mechanics controlling initiation of leveed channels, we
argue that experimental deposits should either produce clear self-
sustaining channels or exhibit depositional trends that would suggest a
progressive evolution of the experimental deposit toward a condition of
channelization. Under depositional conditions with no pre-existing
channel form, prior experiments fail to exhibit self-channelization, or
plausibility of channelization for three reasons: (1) the distance separating
lateral deposits is several times wider than the outlet channel and
increases in width away from the outlet offering little to no confinement
of the flows (e.g., Imran et al. 1998; Imran et al. 2002; Bradford and
Katopodes 1999a, 1999b); (2) the channels developed are a fraction of the
overall current width and depth (e.g., Yu et al. 2006); or (3) the levee
deposits extend only a short distance out from the channel outlet before
terminating in an axial deposit that extends the breadth of the nascent
channel and is several times greater in height than the lateral deposits,
effectively blocking further channel development (e.g., Imran et al. 1998;
Imran et al. 2002; Bradford and Katopodes 1999a, 1999b; Baas et al.
2004).
711
We present a series of experiments that systematically explore the
relationship between flow density and velocity (measured at the outlet)
and the geometry of deposits that accumulate adjacent to the channel
outlet to explore those conditions necessary to establish levee geometry
typically observed in submarine environments. We explored these
dynamical relationships with a non-erodible bed so that only deposition
and resuspension of sediment initially in suspension could occur. This
experimental setup was chosen for two reasons. First, the aggradational
nature of submarine levees clearly indicates that deposition is a key
process in formation of leveed channels. The role of erosional processes,
however, is less clear. Therefore, we consider the case of flows in which
channel formation may develop only by sediment deposition to determine
if these conditions can plausibly explain the observed deposits, or if other
conditions are requisite for the formation of these features. Second, in
fluvial settings (rivers and streams entering standing bodies of water)
observations and conceptual models (Bates 1953, Wright 1977),
numerical models (Edmonds and Slingerland 2007), and physical models
(Rowland 2007; Rowland et al. in press) all suggest that the
hydrodynamic conditions that arise when sediment-laden flows debouch
into standing bodies of water are sufficient to initiate levee formation
under depositional conditions alone with channel width similar to the
river-mouth outlet width. Shearing along the flow margins between the
outflowing river discharge and the still waters of the receiving body leads
to the turbulent transfer of momentum and sediment from the core of the
flow to the margins (Bates 1953; Rowland 2007; Rowland et al. 2009;
Rowland et al. in press) where the sediment deposits on the bed in
relatively quiescent conditions. Along the axis of the flow, however, high
streamwise velocities and bed shear stresses promote sediment entrain-
ment (Edmonds and Slingerland 2007; Rowland et al. in press), which
limits deposition (relative to the flow margins). The combination of these
processes leads to conditions sufficient to meet the criteria set forth by
Imran et al. (1998).
While depositional conditions alone appear sufficient to produce self-
sustaining leveed-channel systems in fluvial settings, there are several
reasons to expect the morphodynamics of dense underflows, and hence
the deposits they produce, to be different from surface water flows as they
enter clear water. First, the excess density of turbidity currents stabilizes
lateral fluid interfaces and suppresses shear-induced instabilities between
the current and ambient waters of the receiving basins. This suppression
of shearing and turbulence increases with increasing density contrast
across the fluid interface (Ellison and Turner 1959; Tennekes and Lumley
1972; Miles 1990). Second, excess density results in a lateral pressure
gradient that will drive a lateral collapse and spreading of an unconfined
density current. Finally, unlike a shallow fluvial discharge that is bounded
at its upper free surface by the atmosphere, turbidity currents are fully
submerged, resulting in the potential for shear and entrainment of
ambient fluids across the entire width of the flow (as opposed to only
along the lateral margins as is the case for surface flows entering clear
water). This shear and entrainment increase the volume of the current and
extract momentum through interfacial friction (Ellison and Turner 1959).
To test these hypotheses and examine the impact of changing flow
properties on morphodynamics near the channel outlet, we systematically
varied flow density relative to outlet velocity, which changes the relative
balance of momentum versus excess density, in sediment-laden density
flows undergoing sudden unconfinement onto a sloped basin bed.
EXPERIMENTAL SETUP
We introduced seven sediment-laden flows into a basin 3 m wide 3 8 m
long 3 0.6 m deep. A constant discharge was supplied to the basin via a
constant-head tank. The flow entered the basin through a rectangular
outlet 20 cm wide 3 2 cm high positioned at the end of an acrylic bed
3.7 m long 3 2.4 m wide that sloped 0.34u (0.006) away from the outlet
712 J.C. ROWLAND ET AL. JSR
JSR INITIATION OF SUBMARINE LEVEED CHANNELS 713

relative to a horizontal plane (Fig. 2). The upstream end of the bed was
positioned 38 cm above the bottom of the basin. Thirty centimeters of
open water separated the side of the bed from the walls of the basin, and
the downstream end of the bed was 3.5 m from end of the basin. The
separation of the experimental bed from the solid walls of the basin
allowed the experimental density currents to flow off the bed and collect
in the bottom of the basin, which was intended to minimize the reflection
of the currents off the basin walls.
Two baseline runs were conducted in which the inflowing water was of
the same density as the waters of the receiving basin. In the first baseline
run, the water level in the basin was maintained at a constant elevation
equivalent to the top of the inflow channel (depth of 2 cm) by extracting
water from the basin via a pair of outflow siphons located at the
downstream end of the basin. This run served as our fluvial analog. In the
other baseline run and all subsequent runs, the inflow was submerged by
20 cm of standing water. During these runs, the basin water level was
allowed to rise (to prevent dilution of inflow waters stored in the sump
used to supply the head tank), resulting in an , 3% decrease in the head
difference between the feed tank and the outlet over the course of a run.
The flow rate into the basin was controlled by adjusting a valve located
just downstream of the head tank. We varied inflow rates between runs to
maintain constant ratio of shear velocity to sediment settling velocity (see
next section). In the absence of a flow meter in the system, we tested the
valve adjustments and measured the final flow rates by timing the rate of
basin infilling with water. The total duration of the experimental runs
varied from 15 minutes for the low density/high velocity inflows to
30 minutes for the high density/low velocity inflows (Table 1). Sediment
was fed into the density current via a standpipe tapped into the inflow line
downstream of the head tank (Fig. 2). The sediment feed rate was
controlled by a VibraScrewTM feeder.
Excess density of the inflow was provided by dissolving Cargill ‘‘top-
flo’’ evaporated salt in tap water. We measured the density of the inflow
and the receiving basin waters at the start and end of each experimental
run using a hydrometer. Due to the dense inflow sinking to the bottom of
the basin and experiencing limited mixing, the greatest observed increase
in basin water density (measured at the inflow point to the basin) was
1 kg/m3; this was observed in the experimental run with the maximum
inflow density (1120 kg/m3).
The simulated sediment used in the experiments consisted of ground
acrylic particles with a median size of 0.42 mm and a specific gravity of
1.2. The full distribution of settling velocities of the sediment in clear
water was measured using settling columns located at the U.S. Geological
Survey in Menlo Park, California (measurements presented in Figure 3).
Due to their size, the acrylic particles have relatively high settling
velocities but are still readily entrained and suspended in the experimental
flows. In all experimental runs, the presence of the plastic particles
contributed less than 0.1% of the excess density of the inflowing waters
relative to the basin waters.
All experimental runs were recorded by a digital video camera
positioned directly over the basin. A subset of the experiments was also
recorded from the side with a digital video camera enclosed in a clear
acrylic box that was submerged below the water surface. Flow velocities
within the density currents were estimated from the videos by tracking the
front location of yellow-green fluorescein dye that was periodically
injected into the currents. Larger, low-density (specific gravity 5 1.05)
particles were also introduced into flows in an attempt (largely
unsuccessful, except for the baseline/fluvial run, due to particle buoyancy)
to document flow velocities by tracking particle trajectories in the
recorded videos.
At the completion of each experimental run, the basin was slowly
drained to expose the deposited sediment for photographing and
topographic surveys. Complete cross sections of bed elevations were
recorded at downstream intervals of 2 cm using a laser line and digital
single-lens reflex (SLR) camera affixed to a rigid aluminum frame
spanning the basin. In this surveying system, a vertical laser sheet was
projected onto the bed, resulting in a line where the sheet intersects the
bed surface. The camera was located a fixed distance from the laser sheet
and at a fixed elevation above the basin and oriented at a 30u angle from
horizontal. At each survey cross section, an image of the laser line on the
flume bed was captured with the camera. On a perfectly flat bed, the laser
line would be straight and occupy a single row or set of rows (depending
on the line thickness) of pixels in the recorded digital image. In the
presence of bed topography the laser line becomes deflected up or down
within the digital image. By digitally extracting the laser line from
successive images of the bed, a digital elevation model of the deposit can
be constructed (see Rowland 2007 for full methodology). For these
experiments, this photogrammetric method had horizontal (in the cross-
stream direction) and vertical pixel resolutions of 0.38 and 0.43 mm,
respectively. Prior testing of this survey system determined that it could
accurately detect a minimum change in bed topography of 0.65 mm with
a 0.11 mm standard deviation.
We subtracted a baseline survey of a sediment-free bed from each of the
experimental bed morphologies to derive a map of deposit thicknesses.
Variations in bed deflections (largely due to sediment loading) between
individual surveys required that elevations of the baseline survey be
adjusted for each individual survey. These adjustments were made by
matching elevations of the baseline survey to portions of the deposit
surveys lacking deposition and interpolating elevations between these
fixed locations. The bed interpolations were adjusted until the average
difference between the deposit and baseline surveys, in the regions of no
deposition, was less than the pixel resolution of the survey images
(0.38 mm).
EXPERIMENTAL DESIGN AND SCALING
Experiments were designed with several simplifying assumptions to
explicitly test the relative influence of momentum versus excess density on
the morphodynamics of a current undergoing unconfinement. We
assumed that depositional levees are constructed from suspended
sediment delivered to the channel margins. We further assumed that the
coarsest fraction of the suspended load is responsible for initial levee
formation, and that the most of suspended sediment does not initially
deposit on the channel margins. To simulate this in the experiment, saline
flows provide the excess density, whereas the ground acrylic particles
simulate the coarse fraction whose deposition initiates the levees. As such,
we argue that sediment deposition initially has a minor impact on the
overall density and flow dynamics of the current. The combination of
feeding a fixed sediment flux into the inflow (as opposed to allowing the

FIG. 1.—Seismic sections of submarine channel levees. Note in Parts B and C that the depositional wedges of the levees appear to overlie a continuous (except were
incised by the channel) basal reflector. Also note that in all images the levee crests are located on the channel margins and show no evidence of inward (toward the
channel) stepping with levee aggradation. Image sources: A) Seismic section through Blue Channel and Yellow Channel of the late Pleistocene Amazon Fan (Shipboard
Scientific Party 1995). Each channel shows , 170 ms (136 m) of vertical aggradation. Modified from Mohrig and Buttles (2007), image courtesy of Mohrig and Buttles.
B) Seismic lines of two submarine channel cross sections constructed over slope detachment surface on the continental slope offshore Brunei. Modified from Straub and
Mohrig (2008), image courtesy of Straub. C) Seismic section of leveed submarine channel in a deep-water setting. Modified from Posamentier and Kolla (2003), image
courtesy of Posamentier.
714 J.C. ROWLAND ET AL. JSR

TABLE 1.— Summary of Prior and Current Experimental Conditions (all values in italics were estimated or calculated from data presented in
referenced work).

rc ra g9 uo Q
Experiment/Reference (kg/m3) (kg/m3) (m/s2) (m/s) (L/s) ho (m) Bo (m) Bo/ho Slope Frd Rio
Luthi (1981) 1007–1067.5 1000 0.07–0.66 0.23 3.5 0.05 0.3 6 5.0u 1.3–4 0.06–0.61
Imran et al. (1998), Numerical
simulations
Summary of Simulations 1033 1000 0.32 2.25–2.28* 4# 50# 12.5–17 0.1u–0.9u 2.2 0.209

Condtions for ‘‘Optimal 1033 1000 0.32 2.25–2.28* 4# 50# 12.5 0.1u–0.9u 2.2 0.209
Channelization’’
Bradford and Katopodes
(1999b), Numerical
simulations
Depositional Only 1002–1017 1000 0.016–0.16 0.5–1 1–2 10 5–10 0–5.7u 1.2–5.6 0.032–0.65

Imran et al. (2002), Physical 1033 1000 0.32 1.10–1.18 4.4–4.72 0.08 0.05 0.625 1.7u–5.7u 6.8–7.3 0.019–0.021
experiments

Parson et al. (2002) 1014–1058 1000 0.14–0.57 ? - . 0.22 , 4 ? ? ? 0 , 1 , 1

Baas et al. (2004) 1231–1578 1000 2.3–5.7 0.39–0.95 5.2–7.8 0.04–0.08 0.22 ? 3.7u–8.6u / 0 0.67–2.1 0.23–2.23

Yu et al. (2006) 1033 1000 0.32 0.05 0.01 0.002/0.01 0.02 10 4u . 1–0.88 1.29

Alexander et al. (2008) 1018 1000 0.18 0.06–0.18 1.58–1.74 0.084–0.121 0.3 1.7–3.6 0–20u / 0 0.47–1.16 0.74–4.5

This Study
Baseline/Fluvial (Fig 5 A, B) 998 998 – 0.305 1.22 0.02 0.2 10 0.34u 0.701 –

Baseline - Submerged 998 998 – 0.305 1.22 0.02 0.2 10 0.34u 0.701 –
(Fig 5 C, D)

Figure 5 E, F 1005 998 0.07 0.305 1.22 0.02 0.2 10 0.34u 8.2 0.02

Figure 5 G, H 1016 998 0.18 0.294 1.18 0.02 0.2 10 0.34u 4.9 0.04

Figure 5 I, J 1034 999 0.34 0.278 1.11 0.02 0.2 10 0.34u 3.4 0.09

Figure 5 K, L 1054 998 0.55 0.193 0.77 0.02 0.2 10 0.34u 1.8 0.30

Figure 5 M, N 1120 998 1.20 0.143 0.57 0.02 0.2 10 0.34u 0.9 1.17
JSR INITIATION OF SUBMARINE LEVEED CHANNELS 715

TABLE 1.— Continued.

Deposit
Re U* (m/s) Ws (m/s) Rp u*/Ws Morphology Ly/Bo Lx/Bo
35,000 0.016 0.0011 0.9 15.0 Broad fan-like NA NA

10,000,000* 0.16 * 5.8 3 1025–0.02 0.061–15.9 8–2784 * Ponded enclosed 3–28 28 - entire basin
nondepositional region
to sheetlike deposits
with greater lateral
deposition extending
the length of the basin
10,000,000* 0.16 * 5.8 3 1025–0.02 0.109–11.38 8–2784 * Deposits with greater 3–18 entire basin
lateral deposition
extending the length of
the basin

(5–20) 3 105 0.025–0.05 0.002–0.014 1.4–7.4 4–25 Short lateral deposits 1.6–3 1.6 for narrow deposit, no
both with and without centerline deposit for
significant axial rapidly divergent case
deposition
(8.8–9.4) 3 104 0.069–0.100 0.0037–0.0078 2.4–4.6 13–19 Lateral deposits , 8–18 , 26–32
surrounding elliptical
zone of no deposition
? ? , 0.0008–0.002 , 0.7–1.4 ? Switchig lobes, with NA NA
nascent depressions on
distal margins
24,000–35,000 0.03–0.07 0.013–0.025 1–14.5 1.5–29 Short lateral deposits 3 , 5
terminating and
significant axial
deposit
100 0.005 0.000017–0.0088 ? 0.6–294 Erosional non-leveed 0.75–1.5, though channel NA
channels near outlet, was a small fraction of
small leveed total current size
deposistional channels
in distal regions
4500–19200 0.004–0.013 0.0026 1.7 1.6–4.9 Elongate ridge aligned NA NA
witht the channel axis

6100 0.022 0.010 12 2.2 Lateral deposits 1.1–2.0 entire basin

extending the length of
the basin
6100 0.022 0.010 12 2.2 Short lateral deposits 1.2–1.4 2.3
terminating and
significant axial deposit
6118 0.022 0.010 12 2.2 Short lateral deposits 1.3–2.1 4
terminating and
significant axial deposit
5962 0.021 0.009 12 2.2 Short lateral deposits 1.3–2.9 4.75
terminating and
significant axial deposit
5738 0.020 0.009 11 2.3 Short lateral deposits 1.2–2.9 4.5
terminating and
significant axial deposit
4060 0.014 0.008 11 1.8 Short lateral deposits 1.2–2.9 2.7
terminating and
significant axial
deposit
3197 0.010 0.005 8 2.2 Broadly distributed 2.85–. 5.7 . 6
deposition with slightly
greater deposition on
the distal lateral
margins
716 J.C. ROWLAND ET AL. JSR

TABLE 1.— Continued.

rc 5 current density
ra 5 ambient fluid (basin) density
g9 5 reduced gravity (g) 5 [(rc 2 ra)/ra]g
uo 5 average velocity at outlet
ho 5 height of outlet
Bo 5 width of outlet
Frd 5 densimetric Froude number
Rio 5 Bulk Richardson number at outlet
Re 5 Reynolds number
u* 5 shear velocity
Ws 5 particle settling velocity
Rp 5 Reynolds particle number
Ly 5 lateral distance between crests of bounding lateral deposits
Lx 5 downstream distance to which lateral deposits extend before axial deposition exceeds lateral deposit thickness
1
non-densimetric Froude number
Notes on Analysis of Reported data
Luthi (1981) Assumed cf 5 0.005 to estimate u*, used CSF 5 0.75 and P 5 3.5 in relationships of Dietrich (1982) to estimate Ws
Imran (1998) * calculated using reported f (Ws/Uo) values and Ws estimated using the relationships of Deitrich (1982) with CSF 5 0.75 and P 5 3.5. # were baseline
values reported specific values of Bo or ho not reported when outlet aspect ratio was varied.
Bradford and Katopodes (1999b) Values of CSF 5 1 and P 5 6 were used to match the settling velocity given by the authors for one grain size in order to estimate
settling using Dietrich (1982) velocities for other grain sizes without listed settling velocities
Imran (2002) Values of CSF 5 0.75 and P 5 3.5 used relationships of Dietrich (1982) to estimate Ws
Parsons et al. (2002) Study reported g9 5 [(rc 2 ra)/rc]g so values were converted to be consistent with the definition of g9 listed above. Used CSF 5 0.75 and
P 5 3.5 in relationships of Dietrich (1982) to estimate Ws
Baas et al. (2004) All currents started at supercritical conditions but underwent hydrualic jumps in the channel section. Velocities reported are for current head,
velocities in body were higher. Slope values are for channel, unconfined basin had no slope.
Values for ho were estimated using reported discharges, head velocities and channel widths; Rio, Frd, and Re estimates were all derived from this estimate of ho. Values
of cf 5 0.005 and cf 5 0.01 were used for smooth and rough channel estimates u*, respectively.
Ws estimates assumed CSF 5 0.75 and P 5 3.5 in relationships of Dietrich (1982).
Yu et al. (2006) Inflow to the basin was through a 16 mm diameter pipe, ho and Bo are estimated in reference just downstream of inflow to tank. Flow into the tank was
supercritical but rapid entrainment of clear water lead to subcritical conditions over most of the basin; current thickness of 0.01m used by authors for estimated of Frd.
Assumed cf 5 0.01 to estimate u*.
Alexander et al. (2008) All inflows start as supercritical but undergo hydraulic jumps in the entrance channel, all values for current characteristics are for the body of
the current in the channel. Slope values are for channel, unconfined basin had no slope. Assumed cf 5 0.005 to estimate u*, used CSF 5 0.75 and P 5 3.5 in
relationships of Dietrich (1982) to estimate Ws

current to entrain sediment from the bed of a channel upstream of the
discharge point) and the presence of a non-erodible bed restricts our
experiments to the exploration of deposition from currents carrying
sediments at concentrations below capacity conditions. With the non-
erodible bed, neither deposition from the current nor dilution of the
current by clear-water entrainment can be offset by sediment entrainment
from the bed, and an ‘‘equilibrium’’ current (in the sense of Parker 1982)
is not feasible.
The physical dimensions of modeled flow differ from natural flow, and so
experiments must be scaled to natural systems (Middleton and Wilcox
1994). Our model length scales were chosen to preserve a realistic aspect
ratio (width to depth; W/D; (Pirmez and Imran 2003)) for our density-
current outlet to preserve the relative influences of shear, water entrainment,
and friction on the dynamics of the current for the discharges, densities, and
sediment properties selected for the experiments.
For the aspect ratio selected in this experiment (W/D 5 10), we scaled
flow density and discharge using a version of the Froude number
modified for flows with excess density (densimetric Froude number
(Frd)):
.pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Frd ~U ½ðrc {ra Þ=ra gh, ð1Þ
where U 5 mean streamwise velocity, g 5 acceleration due to gravity,
h 5 flow thickness, and rc and ra are the densities of the current and the
ambient fluids, respectively. The densimetric Froude number provides a
measure of inertial forces relative to those produced by gravity acting on
the excess fluid density. Few actual observations of turbidity currents
exist on which to base Frd number similarity. Scaling from prior
experiments is mixed: some argue for prototype Frd numbers , 1 (e.g.,
Kane et al. 2008 and Straub et al. 2008) based on the work of Pirmez and
Imran (2003) on the Amazon channel and Klaucke et al. (1997), whereas
other experiments (Imran et al. 2002) and numerical treatments of
turbidity currents (e.g., Garcia and Parker 1989; Salaheldin et al. 2000;
Kostic and Parker 2006) prefer Frd . 1 based on outcrop and seafloor
observations taken to be evidence of hydraulic jumps (Komar 1971; Mutti
1977). Rather than designing our experiments around a constant value of
this poorly constrained parameter, we treat Frd as an independent
variable, and adjust the flow discharge and density to explore the impact
of changing Frd on flow morphodynamics. As per convention, we recast
Frd in terms of its equivalent bulk Richardson number, determined at the
outlet (Rio), defined simply as
1
Rio ~ ð2Þ
Fr2d
By varying the Rio, while holding all other variables as constant as
possible, we sought to explore the effect of changing Rio on the flow
dynamics and how these changes may impact leveed channel initiation.
While varying the Rio, we maintained a constant ratio of shear velocity
(u ) to sediment particle settling velocity (ws ) (Table 1, Fig. 4), and
maintained constant total volumes of both the density current and sediment
entering the basin. We fixed u =ws to achieve sediment suspension at the
outlet and to preserve similarity in sedimentation mechanics between
experimental runs, since experimental depositional morphologies appear
highly sensitive to the dominant settling velocity of the material in suspension
for a given flow regime (Imran et al. 1998; Bradford and Katopodes 1999a,
1999b; Imran et al. 2002; Rowland 2007; Rowland et al. in press).
JSR INITIATION OF SUBMARINE LEVEED CHANNELS
Our target u =ws value for all runs was set to 2.2 based on flow
conditions observed to generate levee forms in the baseline/fluvial run
(Fig. 4). Because ws depends on fluid density, as density of the inflow was
increased the settling velocity of the sediment in suspension decreased. To
maintain a constant u =ws value at the channel outlet to the basin
between runs, we decreased Uo, to adjust u proportionally to the
decrease in ws . We calculated u at the outlet as
pffiffiffiffiffi
u ~Uo cf ð3Þ
where cf is a friction factor set to 0.005 determined from prior
experiments conducted across the same experimental bed (Rowland 2007).
Next, we used the median settling velocity (under fresh-water conditions)
and density of the acrylic sediment to infer values of Powers roundness (P)
and the Corey Shape factor (csf); ws was then calculated for the different
fluid densities using empirical relationships determined by Dietrich
(1982). We began with experiments in which Uo and rc were adjusted to
achieve Rio . 1 (subcritical); following this experiment, the solution was
progressively diluted to reduce Rio. Based on the new fluid density, ws
was calculated and the inflow velocity was adjusted to maintain the u =ws
sediment suspension criterion. The feed rate on the sediment feeder was
also adjusted to match the flow discharge and maintain a constant total
sediment volume between experimental runs and ranged between 2.4 to
5 gm/s. This procedure was repeated a total of four times (Table 1).
FIG. 3.—Cumulative distribution of settling velocities for the acrylic sediment
used in experiments. Settling velocities were determined for particles in fresh water
at room temperature.
717
FIG. 2.—Side view of experimental basin and
setup (not to scale). Note that siphons were used
only in shallow-clear-water run to maintain a
constant basin level. In all other runs, the sump
served only to store excess saline waters needed
to maintain constant volume in head tank
feeding the basin.
We selected a bed slope (S) of 0.34u or 0.006 that falls within the range
reported for leveed channels along the Amazon submarine system (channel
system: 0.14–40u(Pirmez and Imran 2003); fan lobes: 0.11–0.18u (Jegou et al.
2008)). Given the range of slopes over which leveed channels are observed
on the seafloor, it seemed unlikely that slope is a critical parameter
controlling leveed initiation. The possible influence of slope on current
dynamics is further explored in the discussion section below.
RESULTS
Flow fields and deposition patterns for all seven experimental runs are
presented in Figure 5. The flow field is denoted in the overhead
photographs by the distribution of the yellow-green fluorescein dye,
and the red flow vectors were determined from the movement of the dye
front in successive video frames.
In the baseline/fluvial run (no excess density: ra 5 rc, Fig. 5A, B),
flow traveled approximately parallel to the outlet axis, exhibited limited
lateral divergence, and was bordered along its length by narrow shear
zones through which ambient basin fluids were entrained into the flow.
The lack of deposition observed along the flow centerline indicates that
velocities remained sufficient to prevent deposition in a zone approxi-
mately as wide as the channel outlet. Along the margins of the flow,
distinct levees with sharp crest lines developed with a separation distance
between crests of 1.1 to 2.0 outlet widths (Table 1).
The submerged baseline run produced a deposit whose spatial extent
was restricted relative to the baseline/fluvial run (Fig. 5C, D). The
migration rate of the dye front indicates that the submerged flow
decelerated rapidly; the deposition of the acrylic sediment adjacent to the
outlet supports this observation. We attribute this deceleration to the
entrainment of still basin waters into the flow along both the lateral and
upper margins of the flow, which results in rapid vertical expansion of the
discharging flow (Fig. 6A). The deposit from the baseline submerged flow
appeared similar to that of the baseline/fluvial flow within , 40 cm of the
outlet; however, at larger distances, deposition occurred across the
breadth of the flow, with peak deposition occurring along the flow axis
(Fig. 5B).
As Rio increased, so did the lateral spreading of currents entering the
basin. This lateral spreading was evidenced by dye-front migration and
the expanding geometry of the deposits created by the flow (Fig. 5E–N).
For example, the Rio 5 0.02 (Fig. 5E) dye-front and sedimentation
patterns indicate that distinct lateral shear zones bounded the sides of the
outlet and extended downstream , 40 cm (, 2 outlet widths). Beyond
this distance, the flow apparently collapsed and began to spread laterally.
Visual observations during the experiment indicated that following the
collapse the current thickness decreased, presumably due to the fact that
in those runs the rate of lateral spreading of the flow began to outpace the
rate at which ambient clear water was incorporated into the flow. For the
718 J.C. ROWLAND ET AL.
case in which Rio 5 1.17, this spreading produced currents that were
only a few millimeters thick and entrainment of clear water appeared to
be minimal (Fig. 6B).
Centerline Velocities
Streamwise flow velocities along the centerline of the seven experi-
mental flows are shown in Figure 7 normalized by Uo. Centerline
velocities decreased between 1.5 and 2 times more rapidly with distance
from the outlet for baseline submerged versus baseline/fluvial run. By
, 8 outlet widths, downstream velocities were 0.7Uo for the baseline/
fluvial run, while velocities for the submerged flow decreased to 0.4Uo at
an equivalent distance. Centerline velocities observed in the Rio 5 0.09
run were similar to the case of the submerged baseline run. In general,
there were no sudden changes in the current velocity suggestive of abrupt
hydraulic jumps. An exception may be the Rio 5 0.04 run, which showed
a drop from 0.9 to 0.6Uo between x/Bo 5 1.75 and 2.75 (where x is the
distance downstream of the outlet and Bo is the outlet width). This rapid
change is not associated with a corresponding deposit; the crest of the
Rio 5 0.04 deposit occurs at approximately x/Bo 5 5.5 (Fig. 5H), a
result consistent with the observations of Garcia (1993) and the recent
analysis of Huang et al. (2009).
For the Rio 5 1.17 flow, we observed rapid thinning of the flow while
it spread radially. This thinning and spreading is associated with the
observed downstream increase in centerline velocity. Given that the bed
slope is constant along its length, the downstream spreading of the flow,
increased centerline velocities, and decreasing flow thickness are most
plausibly explained by flow collapse.
Depositional Morphologies
Experimental flows with greater Rio produced laterally more extensive
and thinner deposits (Fig. 5D, F, H, J, L, and N). For all submerged
dense flows in which Rio , 1, the downstream distance to the center of
mass of the deposit decreased as Rio increased. The increase in the lateral
extent of the deposits is quantitatively reflected in the increase in the
second moment (the variance) of the deposits in the y (cross-stream)
direction (Fig. 5). The continued increase in deposit breadth of the
densest run (Rio 5 1.17) is clearly visible in Figure 5N, but unfortunately is
not reflected in the calculation of moments because the breadth of the
deposit exceeds the region of the bed surveyed for topography. For the
Rio 5 1.17 flow, the center of mass was located the furthest downstream of
all of the dense submerged flows. In all cases, the center of mass was located
at approximately the outlet center in the cross-stream direction.
JSR
FIG. 4.—Plot of outlet Richardson number
(Rio) versus the ratio of shear velocity (u ) to
sediment settling velocity (ws ) for physical and
numerical experiments of sediment-laden
density currents transitioning from con-
fined to unconfined conditions upon en-
tering basins. Assumptions and estimates
needed to derive Rio and/or u /ws values for
previous investigations are detailed in
Table 1. Gray band highlights the range of
u /ws values for which experimental levees
were observed to develop in fluvial analog
experiments (this study and Rowland 2007).
For supercritical flows, sharp-crested bounding lateral deposits develop
where lateral shear zones are observed along portions of the experimental
flows. As the Rio increases, these deposits extended for shorter distances,
had lower more poorly defined crests, were more spatially extensive, and
were increasingly oblong rather than linear (Fig. 5, Table 1).
Bed Forms
Sediment transport in experimental runs in which Rio , 1 (supercrit-
ical) appeared to undergo a transition from suspension to traction
transport at a downstream distance between x/Bo 5 3 to 5. Downstream
of this transition, we observed deposits exhibiting distinct bedforms
(Figure 5F, H, I, and J), and visual observations during the experiments
indicated that these bedforms migrated downstream. The form and
migration direction suggest that these features reflect flow conditions in
which Ri . 1 (subcritical); however, we have no direct measurements of
flow properties to support this inference. Such a flow transition might be
expected as the entrainment of basin water into the flow caused it to
decelerate and thicken downstream. Sediment transport in the
Rio 5 1.17 flow also appears to transition from suspended to tractive
transport downstream of the outlet. No distinct bedforms were observed
in this flow’s deposit.
DISCUSSION
The Formation of Leveed Channels over Previously Unchannelized Seafloor
Physical or numerical experiments that strive to model the processes of
levee initiation and development must, as a minimum, produce a channel
whose geometry is likely to evolve into a form that has a width similar to
initial outlet. If this were not the case, it is difficult to conceive of how
large leveed submarine channels develop across previously unchannelized
portions of the seafloor. Distributary-like channels developed on the
surface of submarine fans need not meet this criterion but nonetheless
ultimately require a larger feeder channel to have formed along previously
unchannelized portions of the seafloor.
Our physical experiments failed to produce deposits that might
eventually evolve into such large feeder channels when forces due to
density contrasts were even a small fraction of the inertial forces. In the
most inertially dominated flow (Rio 5 0.02), where levee-like deposits
developed downstream to , 2.5 x/Bo, the separation distance between
lateral deposits rapidly increased in width from 1 outlet width to , 2
outlet widths by a distance of 2.5 x/Bo. In addition, total deposition on
the centerline of the flow outpaced that on the levee crests by a factor
JSR INITIATION OF SUBMARINE LEVEED CHANNELS
. 1.5. If this pattern persisted, with continued deposition the channel
would be effectively blocked by sediment before levees were large enough to
confine the flow. To advance basinward, the developing channel would have
to erode through or deflect around this deposit. If such an evolutionary
scenario is correct, remnants of these incised high-relief deposits bounding
active channels should be a prominent feature observable in seismic
reflection panels of leveed submarine channel systems. Our review of
available seafloor and seismic data does not suggest that this is a
predominant mode of channel formation in natural systems.
As Rio increases, deposition becomes more evenly distributed across
the basin, and the potential for lateral confinement of the flows appears
even less likely. For our one subcritical (Rio 5 1.17) experimental run,
deposition along the lateral margins of the current did begin to outpace
deposition along the axis of the basin. The distance between the crests of
these deposits, however, ranged from 3 to 5 outlet widths – much wider
than could plausibly explain downstream extension of the channel. Over
the short durations of these experimental runs, we did not observe any
clear morphodynamic feedbacks that would eventually lead to a
narrowing of the deposit and/or increased confinement of the density
current. Should such deposits evolve into realistic channel geometries, a
progressive inward stepping of levee crests towards the channel would be
required. No such evolution, however, appears obvious in either seismic
records of submarine channel–levee systems from industry and oceano-
graphic seismic datasets (e.g., Posamentier and Kolla 2003; Fildani and
Normark 2004; Mohrig and Buttles 2007; Straub and Mohrig 2008; see
Fig. 1). Recent high-resolution imaging of the seafloor of the Amazon
Fan provides dimensions of master levee channels near the transition
between channel and depositional lobes (Jegou et al. 2008). These
channels substantially decrease in relief (50–60%) near the apex of the
depositional lobe (relative the fully channelized upslope reaches) but
channel widths increase by only a factor of 1.16 to 1.2, far less than that
observed in our experiments.
One attribute of natural systems not captured in our experiments is the
possible cumulative and interactive effect of deposition and erosion arising
from turbidity-current variability. Over the course of a single turbidity
current and between successive flow events, flow magnitudes, sediment
concentrations, and grain-size distributions could vary. Whether the
cumulative deposition and morphodynamic interactions from deposition
and possibly erosion due to a range of turbidity currents leads to channel
initiation is presently unknown. Experiments with successive flow events
that progressively waned, due to head-tank drainage, produced stacked
lobes but no leveed channels (Parsons et al. 2002). However, due to the lack
of observations of actual turbidity currents it is not possible to constrain
current variability in natural systems. Unfortunately, therefore, attempts to
systematically test the influence of this variability require exploring a
potentially infinite parameter space.
Previous Experimental and Numerical Studies
Previous experiments have explored self-channelization of turbidity
currents with suspended sediment traversing non-erodible beds using
both numerical and physical experiments (Table 1, Fig. 4). In work by
Imran et al. (2002), physical experiments with Rio 5 0.002 produced a
series of deposits similar in morphology to those seen in Figure 5F, H,
and J with u =ws ratios between 13 and 19. Numerical simulations
conducted by Imran et al. (1998), with Rio 5 0.209, explored a wider
range of parameters (settling velocities, Reynolds particle numbers, and
basin slopes) than the physical experiments, and classified the resulting
morphologies as ‘‘underdepositional,’’ ‘‘overdepositional,’’ and ‘‘optimum
channelization.’’ When compared to the results of Imran et al. (1998), our
experimental flows and deposits should fall between their optimum and
overdepositional, (‘‘ponded’’ regime) conditions. Similar to our deposits,
the deposits produced under optimum channelization conditions in these
719
previous numerical experiments displayed cross-stream spacing between
levee crests that was six times the width of the outlet.
In numerical simulations with an outlet Rio 5 0.65 and u =ws 5 4,
Bradford and Katopodes (1999b) explored the depositional morphologies
on both a flat and sloping non-erodible bed (Table 1). In the flat-bed
simulation, deposition was focused within a few outlet widths down-
stream of the outlet and the deposits were characterized by short levees
that terminated in a large mound of sediments along the channel
centerline. On a sloped nonerodible bed, deposition was more wide-
spread, but lateral deposition was minimal and deposit crests rapidly
diverged away from the turbidity-current outlet.
More recent laboratory experiments produced features described as
channel–levee systems linked to distal lobe or fan deposits (Baas et al.
2004). All of the currents were reported as supercritical, and the currents
were observed to undergo hydraulic jumps upon exiting the channel and
entering the basin. Precise estimates of current Ri were not possible due to
difficulty in determining the interface height of the currents; however,
based on approximate thicknesses, the bodies of the currents likely had Ri
numbers in the range of 0.20 (Table 1). The stratigraphy of the deposits
suggested that deposition along the inner margins of the levees occurred
during the waning stages of flow, resulting in a progressive narrowing of
the channels. A progressive narrowing of the overall deposit geometry
(relative to its width) with aggradation was taken as evidence for
increasing flow confinement in response to levee growth. These leveed
channels, however, were still , 3 times wider than the channel outlet and
terminated a short distance downstream (, 5 channel widths) in a lobe or
fan deposit with axial heights exceeding the levee crest heights by factors
ranging between , 1.3 and 1.6. Thus, even in these experiments, lobes
would have to be incised and their morphologic expression removed if
channels were to extend in this manner (e.g., Baas et al. 2004, their fig. 20).
Finally, the development of both net erosional and net depositional
channels formed by experimental turbidity currents has been recently
reported (Yu et al. 2006). In these experiments, flow was introduced into
a basin from a 16-mm-diameter pipe. The flows were reported to have
Ri , 1, although no velocities or thickness were reported to allow
determination flow regimes at the pipe outlet. Rapid mixing and
entrainment of basin waters at the pipe outlet, however, may have
caused Ri to exceed 1 a short distance downstream of the outlet, and the
authors reported estimated flow properties at this location (Table 1). Net-
erosional channels formed near the outlet to the basin, and no levee
development was observed along these channels. Net-depositional
channels formed in experiments where the volume of the inflowing
current was unable to fully cover the basin bed. As a result, the current
was observed to break into distinct flow ‘‘lanes.’’ Yu et al. (2006)
hypothesized that higher flow velocities under the center of the current
inhibited deposition relative to the margins, leading to the development of
leveed deposits. These leveed channels would appear to confine a small
fraction of the current passing over them, and so it is not clear how, or if,
these channels might develop into a single dominant channel that can
bypass the fan and deliver sediments farther into the basin.
The Role of Erosion in the Initiation of Channels
Our experimental results and those of others (Imran et al. 1998; Imran
et al. 2002; Bradford and Katopodes 1999b) have to date failed to explain
initiation of submarine leveed channels under conditions in which erosion
of the bed was not possible (Table 1, Fig. 4). Without some form of
bathymetric confinement, even flows with low excess densities expand
laterally and produce lateral deposits whose spacing is inappropriately
large (. 2 times the channel outlet) for extension of the channel
downstream. In contrast, production of levees whose spacing is fitted to
the channel requires that sediment remain in suspension in the high-
velocity core of flow, but rapidly deposits as the flow spreads laterally.
720 J.C. ROWLAND ET AL. JSR

Presumably, this rapid decrease in the ability of the flow to maintain
sediment in suspension is related to the abrupt change of flow velocity in
the cross-stream direction. If a narrow zone of relatively low velocities,
and hence low shear stresses at the bed, can be maintained at the flow
margins, deposition is localized there and further confines the flow.
Conversely, if the flow collapses radially, velocities diminish across a broad
radial zone, and may localize deposition at the front of the flow rather than
along its sides. The former condition was well documented in experiments of
sediment deposition onto levees of fully developed channels that confined
most of the turbidity current (Straub and Mohring 2008). Other
experiments have found that low-velocity zones can be maintained along
the channel margins if the total flow height does not exceed 1.3 times the
height of the bounding channel margins (Mohrig and Buttles 2007), whereas
flows with less confinement do not show a pronounced change in velocity
between channel and overbank regions. These observations, along with our
results, suggest that some form of turbidity-current confinement may be
needed to initiate the formation of levees.
A possible mechanism for this initial flow confinement may be the
erosion of a channel into pre-existing deposits. If erosional confinement
of turbidity currents is a prerequisite for levee initiation, then
understanding the mechanics and the timing of erosion is critical to
understanding the evolution of channelized submarine systems. For
example, is it possible for a current or succession of currents to erode the
seafloor just downslope of the point of flow unconfinement such that
erosion and levee development progressively advances across the seafloor
in a manner analogous to progradational river deltas? Or, do large
stretches of the seafloor first become incised with channels (e.g., Gee et al.
2007) and one or more of these channels subsequently ‘‘capture’’
previously unconfined currents and initiate levees along the length of
the existing incised channel? In an attempt to gain insight into these
questions we review previous experimental, numerical, and field studies of
erosional-channel systems to evaluate possible modes of levee initiation.
In experimental studies with erodible beds, limited bed erosion at the
channel outlet has been observed (Parsons et al. 2002; Baas et al. 2004). In
some cases, these erosional channels reach into the basin (Metivier et al.
2005; Yu et al. 2006). The subsequent development of depositional levees,
however, was either not observed (Yu et al. 2006) or ‘‘levees’’ consisted of
incised lobe deposits formed by tractive sediment transport (Metivier et
al. 2005). While these incised lobes may be similar in appearance to
depositional levees, it is unclear if these features are genetically similar to
submarine levees created by deposition of suspended sediment over-
spilling a channel margin. In numerical experiments (e.g., Bradford and
Katopodes 1999b), turbidity currents eroded a channel into the bed, but
no constructional levees developed. In contrast, in experiments where a
current was routed down an existing channel, levees began to develop
once the channel became partially filled and flow overspilled out of the
channel onto the surrounding basin floor (Bradford and Katopodes
1999b). More recent numerical simulations also show lateral deposition
and levee development arising from current overspilling a pre-existing
channel but no levee formation when currents transition from channel-
ized to unchannelized reaches (e.g., Huang et al. 2007).
Areas of flow expansion and bypass at channel outlets have been
described on the seafloor such as the scour fields found on the modern
San Lucas or Navy fans (Normark 1970; Normark et al. 1979). A wide
variety of erosional features at channel–lobe transitions (sensu Mutti
and Normark 1987) have been documented on the modern seafloor
(Wynn et al. 2002). Most of these features were discontinuous,
spread across a broad swath of the seafloor, and individually smaller in
breadth and depth than the source channel, whereas scours capable of
confining the flow were not observed adjacent to the channel outlet.
Alternatively, erosional scours might form at some distance from the
outlet and progressively confine the density flows as these features erode
headward.
Observations from the modern-day seafloor off the Monterey Fan
indicate that discontinuous erosional channels develop in response to
reconfinement of upslope-sourced flows that were diverted around and
through local topographic obstacles. One such channel exhibits levees at
the head of the channel, where flow was apparently reconfined, though
upslope the channel appears to be strictly erosional (fig. 8 of Fildani and
Normark 2004).
Fildani and Normark (2004) also report a line of scours located
downslope from the outside bend of a meander in the Monterey Canyon
(the Monterey East channel at the Shepard Meander). These features are
interpreted to have arisen as the flow was stripped off from the main
channel and underwent a series of cyclic hydraulic jumps. Numerical
modeling indicates that these net-erosional features migrate up-flow, and
may have helped focus subsequent flows (Fildani et al. 2006). Such a
scenario could cause the scours to coalesce, creating an erosional template
on which a new channel will evolve and depositional levees may be
constructed. Linear scour-shaped depressions have been observed in
other areas (Lamb et al. 2008; Normark et al. 2009; Heinio and Davies
2009) and, therefore, such a mechanism might supply the early erosional
phase needed for channel initiation in submarine environments. The
development of incisional channels followed by subsequent levee
development is also consistent with the pattern of channel evolution,
from simple linear incisions to fully leveed and highly sinuous channels,
observed offshore of Angola (e.g., Gee et al. 2007). Deptuck et al. (2003)
reported early incision caused by slope instability as a precursor for the
emplacement of the Benin-major channel–levee system.
While these seafloor observations provide some insight into possible
incisional dynamics that may help provide partial flow confinement
needed for levee initiation, they do not provide a definitive picture of
development of leveed channels across previously unchannelized sections
of the seafloor. Even with recent high-resolution imaging of the Amazon
Fan, Jegou et al. (2008) state that it is not possible to determine the
mechanism for channel formation across previously unchannelized
seafloor due to the inability to resolve depositional architecture at the
bed scale. Seismic reflectors within established levees, displayed at

Fig. 5.—Still frames from overhead video (A, C, E, G, I, K, and M) and image plots of the resulting experimental deposits (B, D, F, H, J, L, and N) for each of the
seven experimental runs. In the still frames, portions of the flow field are highlighted by the presence of yellow-green dye and red arrows represent velocity vectors
determined by tracking the progression of the dye front (or large, neutrally buoyant particles in the case of A) between successive video frames. The lengths of the vectors
are proportional to the distance traveled by the dye in a 1 second time interval. The grid squares on the flume bed are 20 cm 3 20 cm. In A, the vectors along the margins
of the flow are not to scale (larger) for visualization; prior experiments on similar shallow flows indicate that these velocities, at a maximum, are approximately 0.05 the
local centerline velocity (Rowland et al. 2009). Deposit thickness depicted in B, D, F, H, J, L, and N are scaled by the height of outlet into the basin (2 cm). All distances, x
and y, are scaled by the outlet width (20 cm). Grid lines in the plots of deposition are offset from the grid lines in the still images by 5 cm. Also shown in B, D, F, H, J, L,
and N are the locations of the deposit centroid (black dot) with the standard deviations (square root of the eigenvalues of the second moment) depicted by the solid black
lines and aligned with the principal direction of the second moments. The moments calculated for the Rio 5 1.17 run N are not directly comparable to the prior runs
because much of the sediment deposition occurred beyond the limited of the topographic survey and, as such, are not included in the calculations.
JSR INITIATION OF SUBMARINE LEVEED CHANNELS 721
722 J.C. ROWLAND ET AL. JSR
JSR INITIATION OF SUBMARINE LEVEED CHANNELS 723
724 J.C. ROWLAND ET AL.
resolutions higher than are presently available, would be needed to better
constrain processes responsible for initiation of leveed channels.
Specifically, if submarine leveed channels progressively and incrementally
advance downslope, analogous to a fluvial channel prograding into open
water, individual levee beds would accrete onto the levees in a downslope
direction, resulting in seismic reflectors subparallel to, and truncated
against, the underlying basal layer. Alternatively, if levee formation
occurs contemporaneously across a long stretch of incised seafloor,
seismic reflectors should be continuous and nearly parallel to the
underlying basal layer in the downslope direction.
The Effect of Clear Water Entrainment and Basin Slope on
Current Dynamics
In all of our experimental runs the density currents underwent dramatic
adjustments upon entering the basin and losing lateral confinement. By
systematically varying Rio our goal was to explore if there was a balance
of initial inertial and gravity forces that would evolve into a set of flow
dynamics, within the basin, capable of producing levee deposits. By only
varying outlet conditions, the question arises as to whether there is some
change in basin slope that would have been more suitably matched to our
imposed outlet conditions such that the dramatic changes in velocity and
sediment transport dynamics we observed would have been avoided.
In theory it is possible for the driving and resisting forces in a density
current to be balanced so that an equilibrium or steady flow condition is
achieved (Ellison and Turner 1959). The Fr number at this equilibrium or
‘‘normal’’ condition can be expressed as (Kostic and Parker 2006)
rffiffiffiffiffiffiffi
S
Frn ~ , ð4Þ
CF
where CF is the total friction exerted on the current both from contact
JSR
with the bed and from the entrainment of ambient basin fluids across
current interfaces, calculated as:
 
1
CF ~cf zew 2{ Ri : ð5Þ
2
We calculated ew following Parker et al. (1986):
0:00153
ew ðRiÞ~ , ð6Þ
0:0204zRi
setting Ri 5 Rio. For our subcritical run (Rio 5 1.17) we obtain
Frn 5 0.94 versus the outlet Frd 5 0.92. Setting Frn equal to the
right-hand side of Equation 1 we obtain a normal velocity
(Un) 5 0.145 m/s versus the 0.143 m/s estimated at the outlet to the
basin. This analysis suggests that the outlet conditions for this run were
almost identical to the theoretical equilibrium conditions for the basin
slope and current dynamics (Equation 5). At most, the current would
have needed to be accelerated by 1.4% to achieve the predicted normal
flow condition. In reality, we observe a . 40% increase in flow velocity
prior to the onset of continuous current deceleration down-basin. We
suggest that this dramatic deviation from predicted normal behavior
arises from the lateral freedom of an unconfined current to both spread
and collapse. Inherent in the assumption of normal flow conditions is a
condition of steady (unchanging) dynamics. As long as a lateral
pressure gradient in the unconfined current is sufficient to overcome
resistance to spreading (due to friction with the bed and the ambient
basin fluid) the current will spread, thin, and be unsteady regardless of
the imposed basin slope.
In our supercritical runs, the basin slope would need to be increased
from 0.047 to 6 if these density currents were to meet a normal condition
without undergoing adjustments in velocity down-basin of the outlet.
JSR INITIATION OF SUBMARINE LEVEED CHANNELS
Such an analysis, however, neglects the most dramatic changes that these
currents undergo: the transfer of momentum due to interfacial friction as
the flow entrains ambient fluid and as a result becomes diluted as it
increases in height (Parker et al. 1986). Application of Equations 5 and 6
suggests that the interfacial friction related to this entrainment is 3 to 20
times larger than the friction exerted by the bed alone. Additionally, by
40 cm downstream of the outlet, this entrainment would cause the current
to increase in thickness, and be diluted proportionally, by 40 to 145%. In
currents where the source of excess density is supplied by the suspension
of sediment (turbidity currents) it may be possible for entrainment of
sediment from the bed to offset the effects of entrainment of ambient fluid
to produce a self-accelerating current (Parker et al. 1986). Regardless of
the basin slope, these supercritical currents could avoid the observed
rapid deceleration (Fig. 7), the loss to sediment transport capacity and
competence, and the massive deposition of sediment along the flow axis
(Fig. 5) only by eroding the bed to provide the excess density necessary to
offset the impact that entrainment of ambient fluid has on the dynamics
of these flows.
CONCLUSIONS
We introduced a series of sediment-laden density currents with varying
Rio into clear water. These flows traversed a non-erodible bed and did not
725
FIG. 6.—Side views of: A) the submerged
clear-water flow (no density contrast between
current and basin waters, and B) the initially
subcritical (Rio 5 1.17) current. Each grid
square is 20 cm by 20 cm, the height of the outlet
into the basin is 2 cm. Arrows depict flow
direction and are not scaled to the velocities.
produce depositional morphologies that achieved or appeared to have the
potential to achieve self-confinement. Our results suggest that the
dynamics of initiation of leveed channels is likely different than those
documented in fluvial settings. Flows with low initial Rio produced
bounding lateral deposits that extended only a few outlet widths
downstream and the sedimentation rate on these lateral deposits was
greatly exceeded by the deposition along the axis of the flow. As density
was increased (Rio increased), bounding lateral deposits became smaller,
were poorly fitted to the outlet, and provided little or no confinement of
the current. Based on our results and a review of prior experimental work,
we suggest that partial confinement of a sediment-laden density current
may be a prerequisite to levee formation. Under the range of outlet
conditions explored in our experiments, depositional currents on a non-
erodible bed do not appear to be able to provide self-confinement, and
incision into a pre-existing deposit may be necessary to develop leveed
submarine channels. To the extent that the dynamics of the laboratory-
modeled saline flows and geometry of the experiment are analogous to
natural systems, it seems unlikely that these flows produce deposits that
ultimately confine such currents. However, it is always possible that the
flow conditions simulated in the laboratory serve as poor analogies to
those found in natural systems, and as such, better sea-floor imaging,
observations of real turbidity currents at the appropriate spatial and
temporal scales, and experiments that explore the effects of an erodible
726 J.C. ROWLAND ET AL.
bed, a range of basin slopes, and currents whose density changes
substantially with sediment concentration may lend further insight into
the processes of levee initiation.
ACKNOWLEDGMENTS
Clair Le Gall provided critical assistance in the execution of these
experiments and analysis of experimental videos. Support for this research
was provided by Chevron Energy Technology Company, and the Terman
Fellowship to GEH. Discussions with Tim McHargue, Jake Covault, and
other members of the Clastic Stratigraphy Team at Chevron helped the early
phases of this project. We thank Federico Falcini for comments and
suggestions that greatly improved the manuscript. We also thank Steven
Reneau for his review of the manuscript. Experiments were conducted at the
UC Berkeley Richmond Field Station; support for these facilities and
experimental supplies were provided by the American Chemical Society
Petroleum Research Fund and the National Center for Earth Surface
Dynamics (NCED). Reviews by Michael Lamb, John Grotzinger, Ian Kane,
Kyle Straub, Morgan Sullivan, David Piper, John Southard, and two
anonymous reviewers resulted in major improvements to this manuscript.
REFERENCES
ALEXANDER, J., MCLELLAND, S.J., GRAY, T.E., VINCENT, C.E., LEEDER, M.R., AND
ELLETT, S., 2008, Laboratory sustained turbidity currents form elongate ridges at
channel mouths: Sedimentology, v. 55, p. 845–868, doi:10.1111/j.1365-
3091.2007.00923.x.
BAAS, J.H., VAN KESTEREN, W., AND POSTMA, G., 2004, Deposits of depletive high-
density turbidity currents: a flume analogue of bed geometry, structure and texture:
Sedimentology, v. 51, p. 1053–1088, doi: 10.1111/j.1365-3091.2004.00660.x.
BATES, C.C., 1953, Rational theory of delta formation: American Association of
Petroleum Geologists, Bulletin, v. 37, p. 2119–2162.
BRADFORD, S.F., AND KATOPODES, N.D., 1999a, Hydrodynamics of turbid underflows. I:
Formulation and numerical analysis: Journal of Hydraulic Engineering, v. 125, p.
1006–1015.
BRADFORD, S.F., AND KATOPODES, N.D., 1999b, Hydrodynamics of turbid underflows.
II: Aggradation, avulsion, and channelization: Journal of Hydraulic Engineering,
v. 125, p. 1016–1028.
DEPTUCK, M.E., STEFFENS, G.S., BARTON, M., AND PIRMEZ, C., 2003, Architecture and
evolution of upper fan channel-belts on the Niger Delta slope and in the Arabian Sea:
Marine and Petroleum Geology, v. 20, p. 649–676.
JSR
FIG. 7.—Centerline velocities normalized by
the initial outlet velocity (Uo) for each of the
seven experimental runs. The downstream dis-
tances (x) are shown normalized by the outlet
width (Bo). ‘‘Sub-base’’ and ‘‘base-fluvial’’ refers
to the submerged and unsubmerged baseline
flows with no excess density
(rc 5 ra), respectively.
DIETRICH, W.E., 1982, Settling velocity of natural particles: Water Resources Research,
v. 18, p. 1615–1626.
EDMONDS, D.A., AND SLINGERLAND, R.L., 2007, Mechanics of river mouth bar
formation: Implications for the morphodynamics of delta distributary networks:
Journal of Geophysical Research-Earth Surface, v. 112, no. F02034, doi:10.1029/
2006JF000574.
ELLISON, T.H., AND TURNER, J.S., 1959, Turbulent entrainment in stratified flows:
Journal of Fluid Mechanics, v. 6, p. 423–448.
FEDELE, J.J., AND GARCIA, M.H., 2009, Laboratory experiments on the formation of
subaqueous depositional gullies by turbidity currents: Marine Geology, v. 258, p.
48–59.
FILDANI, A., AND NORMARK, W.R., 2004, Late Quaternary evolution of channel and lobe
complexes of Monterey Fan: Marine Geology, v. 206, p. 199–223.
FILDANI, A., NORMARK, W.R., KOSTIC, S., AND PARKER, G., 2006, Channel formation by
flow stripping: large-scale scour features along the Monterey East Channel and their
relation to sediment waves: Sedimentology, v. 53, p. 1265–1287.
GARCIA, M., 1993, Hydraulic jumps in sediment-driven bottom currents: Journal of
Hydraulic Engineering, v. 119, p. 1094–1117.
GARCIA, M., AND PARKER, G., 1989, Experiments on hydraulic jumps in turbidity
currents near a canyon–fan transition: Science, v. 245, p. 393–396.
GEE, M.J.R., GAWTHORPE, R.L., BAKKE, K., AND FRIEDMANN, S.J., 2007, Seismic
geomorphology and evolution of submarine channels from the Angolan continental
margin: Journal of Sedimentary Research, v. 77, p. 433–446.
HALL, B., MEIBURG, E., AND KNELLER, B., 2008, Channel formation by turbidity
currents: Navier–Stokes-based linear stability analysis: Journal of Fluid Mechanics,
v. 605, p. 185–210.
HEINIO, P., AND DAVIES, R.J., 2009, Trails of depressions and sediment waves along
submarine channels on the continental margin of Espirito Santo Basin, Brazil:
Geological Society of America, Bulletin, v. 121, p. 698–711, doi: 10.1130/B26190.1.
HUANG, H., IMRAN, J., AND PIRMEZ, C., 2007, Numerical modeling of poorly sorted
depositional turbidity currents: Journal of Geophysical Research, v. 112, C01014,
doi:10.1029/2006JC003778.
HUANG, H., IMRAN, J., PIRMEZ, C., ZHANG, Q., AND CHEN, G., 2009, The critical
densimetric Froude number of subaqueous gravity currents can be non-unity or non-
existent: Journal of Sedimentary Research, v. 79, p. 479–485, doi: 10.2110/jsr.2009.048.
IMRAN, J., PARKER, G., AND KATOPODES, N., 1998, A numerical model of channel
inception on submarine fans: Journal of Geophysical Research-Oceans, v. 103, p.
1219–1238.
IMRAN, J., PARKER, G., AND HARFF, P., 2002, Experiments on incipient channelization of
submarine fans: Journal of Hydraulic Research, v. 40, p. 21–32.
IZUMI, N., 2004, The formation of submarine gullies by turbidity currents: Journal of
Geophysical Research, v. 109, doi:10.1029/2003JC001998.
JEGOU, I., SAVOYE, B., PIRMEZ, C., AND DROZ, L., 2008, Channel-mouth lobe complex of
the recent Amazon Fan: The missing piece: Marine Geology, v. 252, p. 62–77,
doi:10.1016/j.mergeo.2008.03.004.
JSR INITIATION OF SUBMARINE LEVEED CHANNELS
KANE, I.A., MCCAFFREY, W.D., AND PEAKALL, J., 2008, Controls on sinuosity evolution
within submarine channels: Geology, v. 36, p. 287–290, doi: 10.1130/G24588A.1.
KLAUCKE, I., HESSE, R., AND RYAN, W.B.F., 1997, Flow parameters of turbidity currents
in a low-sinuosity giant deep-sea channel: Sedimentology, v. 44, p. 1093–1102, doi:
10.1111/j.1365-3091.1997.tb00449.x.
KOMAR, P.D., 1971, Hydraulic jumps in turbidity currents: Geological Society of
America, Bulletin, v. 82, p. 1477–1488.
KOSTIC, K., AND PARKER, G., 2006, The Response of turbidity currents to a canyon–fan
transition: Internal hydraulic jumps and depositional signatures: Journal of Hydraulic
Research, v. 44, p. 631–653.
LAMB, M.P., PARSONS, J.D., MULLENBACH, B.L., FINLAYSON, D.P., ORANGE, D.L., AND
NITTROUER, C.A., 2008, Evidence for superelevation, channel incision, and formation
of cyclic steps by turbidity currents in Eel Canyon, California: Geological Society of
America, Bulletin, v. 120, p. 463–475.
LUTHI, S., 1981, Experiments on non-channelized turbidity currents and their deposits:
Marine Geology, v. 40, p. M59–M68.
MCHARGUE, T.R., 1991, Seismic facies, processes, and evolution of Miocene inner fan
channels, Indus Submarine Fan, in Weimer, P., and Link, M., eds., Seismic Facies and
Sedimentary Processes of Submarine Fans and Turbidite Systems: New York:
Springer-Verlag, p. 402–413.
MENARD, H.W., 1955, Deep-sea channels, topography, and sedimentation: American
Association of Petroleum Geologists, Bulletin, v. 39, p. 236–255.
METIVIER, F., LAJEUNESSE, E., AND CACAS, M.C., 2005, Submarine canyons in the
bathtub: Journal of Sedimentary Research, v. 75, p. 6–11, doi: 10.2110/jst.2005.002.
MIDDLETON, G.V., AND WILCOCK, P.R., 1994. Mechanics in the Earth and Environ-
mental Sciences: New York, Cambridge University Press, 459 p.
MILES, J.W., 1990, Richardson’s number revisited, in List, E.J., and Jirka, G.H., eds.,
Stratified Flows: New York, American Society of Civil Engineers, p. 1–7.
MOHRIG, D., AND BUTTLES, J., 2007, Deep turbidity currents in shallow channels:
Geology, v. 35, p. 155–158, doi: 10.1130/G22716A.1.
MUTTI, E., 1977, Distinctive thin-bedded turbidite facies and related depositional
environments in the Eocene Hecho Group (South-central Pyrenees, Spain):
Sedimentology, v. 24, p. 107–131.
MUTTI, E., AND NORMARK, W.R., 1987, Comparing examples of modern and ancient
turbidite systems: problems and concepts, in Leggett, J.K., and Zuffa, G.G., eds.,
Marine Clastic Sedimentology: Concepts and Case Studies: London, Graham &
Trotman, p. 1–38.
NORMARK, W.R., 1970, Growth patterns of deep-sea fans: American Association of
Petroleum Geologists, Bulletin, v. 54, p. 2170–2195.
NORMARK, W.R., PIPER, D.J.W., AND HESS, G.R., 1979, Distributary channels, sand
lobes, and mesotopography of navy submarine fan, California Borderland, with
applications to ancient fan sediments: Sedimentology, v. 26, p. 749–774.
NORMARK, W.R., PAULL, C.K., CARESS, D.W., USSLER, W, III, AND SLITER, R., 2009,
Fine-scale relief related to Late Holocene channel shifting within the floor of the
upper Redondo Fan, offshore Southern California: Sedimentology, v. 56, p.
1670–1689, doi:10.1111/j.1365-3091.2009.01052.x.
PARKER, G., 1982, Conditions for the ignition of catastrophically erosive turbidity
currents: Marine Geology, v. 46, p. 307–327.
PARKER, G., FUKUSHIMA, Y., AND PANTIN, H.M., 1986, Self-accelerating turbidity
currents: Journal of Fluid Mechanics, v. 171, p. 145–181.
PARSONS, J.D., SCHWELLER, W.J., STELTING, C.W., SOUTHARD, J.B., LYONS, W.J., AND
GROTZINGER, J.P., 2002, A preliminary experimental study of turbidite fan deposits:
Journal of Sedimentary Research, v. 72, p. 619–628.
PEAKALL, J., MCCAFFREY, B., AND KNELLER, B., 2000, A process model for the evolution,
morphology, and architecture of sinuous submarine channels: Journal of Sedimentary
Research, v. 70, p. 434–448.
727
PIPER, D.J., AND NORMARK, W.R., 1983, Turbidite depositional patterns and flow
characteristics, Navy Submarine Fan, California Borderland: Sedimentology, v. 30, p.
681–694.
PIPER, D.J.W., HISCOTT, R.N., AND NORMARK, W.R., 1999, Outcrop-scale acoustic facies
analysis and latest Quaternary development of Hueneme and Dume submarine fans,
offshore California: Sedimentology, v. 46, p. 47–78.
PIRMEZ, C., 1994, Growth of a submarine meandering channel–levee system on the
Amazon Fan [unpublished Ph.D. thesis]: Columbia University, New York, 587 p.
PIRMEZ, C., AND IMRAN, J., 2003, Reconstruction of turbidity currents in Amazon
Channel: Marine and Petroleum Geology, v. 20, p. 823–849.
POSAMENTIER, H.W., AND KOLLA, V.E., 2003, Seismic geomorphology and stratigraphy
of depositional elements in deep-water settings: Journal of Sedimentary Research,
v. 73, p. 367–388.
ROWLAND, J.C., 2007. Tie Channels [Ph.D. thesis]: Berkeley, California, University of
California, Berkeley, 176 p.
ROWLAND, J.C., STACEY, M.T., AND DIETRICH, W.E., 2009, Turbulent characteristics of a
shallow wall-bounded plane jet: experimental implications for river mouth hydrody-
namics: Journal of Fluid Mechanics, v. 627, p. 423–449, doi:10.1017/
S0022112009006107.
ROWLAND, J.C., DIETRICH, W.E., AND STACEY, M.T., in press, Morphodynamics of
subaqueous levee formation: insights into river mouth morphologies arising from
experiments: Journal of Geophysical Research-Earth Surface.
SALAHELDIN, T.M., IMRAN, J., CHAUDHRY, M.H., AND REED, C., 2000, Role of fine-
grained sediment in turbidity current flow dynamics and resulting deposits: Marine
Geology, v. 171, p. 21–38.
SHEPARD, F.P., 1948. Submarine Geology: New York, Harper and Brothers, 338 p.
SHIPBOARD SCIENTIFIC PARTY, 1995, Site 937, in Flood, R.D., Piper, D.J.W., and
Klaus, A., et al. Proceedings of the Ocean Drilling Program, Initial reports,
Volume 155: College Station, Texas, Ocean Drilling Program, p. 383–408.
SKENE, K.I., PIPER, D.J.W., AND HILL, P.S., 2002, Quantitative analysis of variations in
depositional sequence thickness from submarine channel levees: Sedimentology, v. 49,
p. 1411–1430.
STRAUB, K.M., AND MOHRIG, D., 2008, Quantifying the morphology and growth of
levees in aggrading submarine channels: Journal of Geophysical Research-Earth
Surface, v. 113, F03012, doi: 10.1029/2007JF000896.
STRAUB, K.M., AND MOHRIG, D., 2009, Constructional canyons built by sheet-like
turbidity currents: observations from offshore Brunei Darussalam: Journal of
Sedimentary Research, v. 79, p. 24–93, doi: 10.2110/jsr.2009.006.
STRAUB, K.M., MOHRIG, D., MCELROY, B., BUTTLES, J., AND PIRMEZ, C., 2008,
Interactions between turbidity currents and topography in aggrading sinuous
submarine channels: A laboratory study: Geological Society of America, Bulletin,
v. 120, p. 368–385, doi: 10.1130/B25983.1.
TENNEKES, H., AND LUMLEY, J.L., 1972, A First Course in Turbulence: Cambridge,
Massachusetts, The MIT Press, 300 p.
WRIGHT, L.D., 1977, Sediment transport and deposition at river mouths—synthesis:
Geological Society of America, Bulletin, v. 88, p. 857–868.
WYNN, R.B., KENYON, N.H., MASSON, D.G., STOW, D.A.V., AND WEAVER, P.P.E., 2002,
Characterization and recognition of deep-water channel–lobe transition zones:
American Association of Petroleum Geologists, Bulletin, v. 86, p. 1441–1462.
YU, B., CANTELLI, A., MARR, J., PIRMEZ, C., O’BYRNE, C., AND PARKER, G., 2006,
Experiments on self-channelized subaqueous fans emplaced by turbidity currents and
dilute mudflows: Journal of Sedimentary Research, v. 76, p. 889–902, doi: 10.2110/
jsr.2006.069.
Received 2 September 2009; accepted 30 March 2010.