See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/251489306

Laboratory testing of mathematical models for high-concentration fluid mud

turbidity currents

Article  in  Ocean Engineering · January 2011

DOI: 10.1016/j.oceaneng.2010.10.020

CITATIONS READS

15 82

2 authors:

Mijanur R. Chowdhury Firat Y Testik

University of Toronto University of Texas at San Antonio
13 PUBLICATIONS   70 CITATIONS    72 PUBLICATIONS   612 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Structural Vulnerability of Coastal Bridges under Extreme Hurricane Conditions View project

All content following this page was uploaded by Firat Y Testik on 06 September 2018.

The user has requested enhancement of the downloaded file.

 Ocean Engineering ] (]]]]) ]]]–]]]

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Laboratory testing of mathematical models for high-concentration

fluid mud turbidity currents
M.R. Chowdhury, F.Y. Testik n
Civil Engineering Department, College of Engineering and Science, Clemson University, 110 Lowry Hall, Clemson, SC 29634-0911, USA

a r t i c l e in f o abstract

Article history: The propagation characteristics of fluid mud turbidity currents were investigated experimentally and
Received 19 June 2010 theoretically. Parameterizations for propagation phase transition times from slumping to self-similar and
Accepted 23 October 2010 self-similar to viscous phases are proposed. Predictive capabilities of different mathematical models that
Editor-in-Chief: A.I. Incecik
fall into three different modeling approaches (force-balance, box, shallow water) were evaluated for fluid
mud turbidity current propagation using our experimental observations. For the slumping and self-
Keywords: similar phases, the box and force-balance models showed superior predictive capabilities than the
Fluid mud one-layer shallow water models with deep ambient condition. Fluid mud turbidity currents have a non-
Turbidity current Newtonian rheology and their transition and propagation characteristics in the viscous phase differ vastly
Box model
from the Newtonian currents. We derived and presented a viscous force-balance expression for the
Shallow water model
propagation of a non-Newtonian power-law fluid current. We then experimentally evaluated the
Viscous spreading
Non-Newtonian fluid predictive capability of this force-balance and the viscous shallow water model by Di Federico et al.
(2006). Both models’ predictions are observed to be in notably good agreement with the experimental
data. The results of this study are expected to be useful for preliminary swift calculations of the fluid mud
turbidity current propagation characteristics as well as in deciding whether more detailed calculations at
the expense of complexity are required.
& 2010 Elsevier Ltd. All rights reserved.

1. Introduction importance for a number of engineering and scientific applications. In

Turbidity currents are the lateral flow of fluids, containing
suspended matter, into another fluid of different density
(Bonnecaze et al., 1993; Hosseini et al., 2006). They are a subclass
of gravity currents in which the driving force is caused by the density
difference between the intruding and the ambient fluid due to a
variety of reasons such as compositional or temperature differences
between the two fluids. Turbidity currents occur in a vast variety of
environmental, geological, and industrial situations caused by both
natural and man-made causes (see monographs by Simpson, 1997;
Ungarish, 2009; reviews by Kneller and Buckee, 2000; Middleton,
1993; Huppert, 2006; Meiburg and Kneller, 2010). These currents are
also an important mechanism of transportation of sediments in the
form of industrial or estuarial effluent (Bonnecaze et al., 1993),
distribution of sediments in deep and nearshore marine environment
(Mohrig and Marr, 2003; Seymour, 1986; Drago, 2002), formation of
the majority of sandstones in the geological record and abyssal plains
(Middleton, 1993), many of the world’s most important hydrocarbon
reservoirs (Kneller and Buckee, 2000), and the formation of submarine
canyons on continental shelves (Bonnecaze et al., 1993), among
many others. Therefore, turbidity current investigations are of great
n
Corresponding author. Tel.: + 1 864 656 0484; fax: + 1 864 656 2670.
E-mail address: ftestik@clemson.edu (F.Y. Testik).
particular, this study is motivated by the open-water pipeline dredged
material disposal operations. When dredged mud slurry, consisting of
cohesive particles such as silt and clay, is dumped in an open water
disposal site, a high concentration turbidity current of fluid mud
forms at the coastal seafloor (Neal et al., 1978; Teeter, 2000). There are
environmental concerns associated with the fluid mud turbidity
currents generated in open water dredge disposal operations as the
current may propagate several kilometers away from the source,
depending upon the bottom slope, the ambient flow field and the
released volume of slurry. For example, Teeter (2002) observed a fluid
mud gravity current propagation of approximately 3 km for the
pipeline discharge of approximately 5.2  105 m3 slurry. As the fluid
mud spreads, it may overrun everything in its path, killing benthos
such as clams and oysters (Nichols and Thompson, 1978). If there is a
high bed shear stress, entrainment of the underflow by the ambient
water can generate a turbid plume in the water (Teeter, 2000). This
turbid plume may pollute the water quality and block sunlight,
harming underwater flora and fauna.
The propagation dynamics of a gravity current is governed by
the buoyancy, inertia and viscous forces. At the earlier propagation
stage (henceforth, inertia–buoyancy stage), the viscous force is
negligible compared to the other two forces, and the driving
buoyancy force is balanced by the inertia force. As the current
propagates, it starts experiencing an increasing viscous force and
after some time, depending on the rheological properties of the

0029-8018/$ - see front matter & 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.oceaneng.2010.10.020

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
2 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]
current fluid, the viscous force may become comparable to the
inertia force. After this point, propagation dynamics is governed by
the viscous and buoyancy force balance (henceforth, viscous–
buoyancy stage). In the inertia–buoyancy stage, a gravity current
may exhibit two distinct propagation phases: slumping and self-
similar phases. The slumping phase, occurring near the source of the
dense fluid, is an adjustment phase in the formation of a fully
developed gravity current with characteristic features like nose,
head and body. The current front propagation speed in this phase is
approximately constant (Rottman and Simpson, 1983; Gladstone
et al., 1998; Huppert and Simpson, 1980). Though the flow in
slumping phase is mainly governed by the inertia–buoyancy
balance, the overlying ambient fluid is very dynamic in this phase,
affecting the propagation characteristics of the gravity current. After
the slumping phase, self-similar phase of the inertia–buoyancy
stage commences. Here, the front propagation velocity is no longer
constant, but decreases with distance. If viscous effects exhibit
greater importance before the slumping phase is completed, the
current may not exhibit the self-similar phase, but rather directly
transition into the viscous–buoyancy stage (Huppert and Simpson,
1980). Such a transition was observed in our experiments, the
results of which are presented later in this paper. Although not
observed in our experiments, in certain rare instances the current
may exhibit a viscous–buoyancy stage before the inertia–buoyancy
stage (see Maxworthy, 1983; e.g., current due to a variable inflow in
leak or volcanic eruption). Determination of the phase transition
times is essential for modeling the propagation of gravity currents
and is addressed in Section 3.
Cohesive particle mixtures such as fluid mud mixtures exhibit
profound non-Newtonian behavior (Whitehouse et al., 2000). Note
that in the case of liquefied/fluidized flows, especially during the
initial phase, cohesive particle mixtures may also exhibit pseudo-
Newtonian behavior (e.g., Lowe, 1982). However, our knowledge
on turbidity currents is mainly accumulated through investigations
with non-cohesive mixtures exhibiting Newtonian behavior
(e.g., Bonnecaze et al., 1993; Gladstone et al., 1998). Given the
rheological differences, cohesive and non-cohesive turbidity cur-
rent propagation characteristics are notably different. For example,
as we discuss later, it is highly likely that non-Newtonian rheology
causes fluid mud gravity currents to maintain a viscous–buoyancy
stage for a significant duration of their propagation, while most
non-cohesive turbidity currents may never experience the viscous–
buoyancy stage. One of the goals of this study is to identify and
elucidate these differences experimentally and also to provide a
high-quality dataset for use in studying cohesive turbidity currents.
A significant number of studies have elucidated a quantitative
prediction capability for the propagation characteristics of gravity
currents, mostly compositional gravity currents, such as current
front position and velocity, current height, and volume fraction of
particles (Von Karman, 1940; Benjamin, 1968; Hurzeler et al., 1996;
Kirwan et al., 1986; Bowen et al., 1984; Kuenen, 1952; Mulder et al.,
1998; Jiang and Leblond, 1993). These studies utilized various
mathematical modeling approaches characterized by different
levels of complexities, ranging from dimensional analysis to solving
complex Navier–Stokes equations. In addition to these mathema-
tical modeling approaches two-dimensional and three-dimensional
high-resolution numerical simulations of gravity currents have
been attempted (e.g., Blanchette et al., 2006; Cantero et al., 2007;
Haertel et al., 2000; Necker et al., 2002). In this study, we focus on
laboratory tests of three different widely used mathematical
modeling approaches that elucidate their predictive capabilities
for the propagation characteristics of fluid-mud turbidity currents
in the inertia–buoyancy (Section 4) and viscous–buoyancy stages
(Section 5). These approaches are force-balance models based on
dimensional arguments (Huppert, 1982; Didden and Maxworthy,
1982), box models based on simplifying assumptions on the current
Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
geometry (Huppert and Simpson, 1980; Dade and Huppert, 1995),
and shallow water models based on solutions of Navier–Stokes
equations (Bonnecaze et al., 1993; Harris et al., 2001; Di Federico
et al., 2006). Each model is briefly discussed in Sections 4 and 5.
Predictive models discussed above were developed for use in
mostly compositional gravity currents or turbidity currents com-
posed of non-cohesive sediments, and are not laboratory tested for
turbidity currents composed of cohesive sediments such as fluid
mud gravity currents. Turbidity currents have different propagation
characteristics compared to compositional gravity currents because
deposition (entrainment) of particles in a turbidity current reduces
(increases) the driving buoyancy force as the current propagates
(Harris et al., 2001; Bonnecaze et al., 1993). Moreover, cohesive
turbidity currents have different propagation characteristics com-
pared to their non-cohesive counter-parts, because of the differences
in their rheological and settling characteristics. Although different
modeling approaches have been laboratory tested for propagation
characteristics of compositional gravity currents and non-cohesive
turbidity currents (e.g., Bonnecaze et al., 1993; Hurzeler et al., 1996;
Gladstone et al., 1998), there has been no such systematical study
for cohesive turbidity currents. Therefore, one major goal of this
study involved evaluating the capabilities of these three well-
established mathematical modeling approaches used to predict
cohesive turbidity currents.
This manuscript is organized as follows. Experimental set-up,
measurement and data analysis methodologies, and the rheological
properties of the fluid mud mixtures are described in Section 2.
Phase transitions are experimentally modeled in Section 3. Model
predictions and experimental results for the propagation charac-
teristics of the fluid mud turbidity currents in the inertia–buoyancy
and viscous–buoyancy stages are presented in Sections 4 and 5,
respectively. Discussions and conclusions are given in Section 6.
2. Experimental methodology
2.1. Experimental set-up, data acquisition, and analysis
A lock-exchange tank was used to conduct the experiments.
Lock-exchange tanks are commonly used to investigate different
aspects of two-dimensional (2-D) gravity currents (e.g., Simpson,
1972; Simpson and Britter, 1979; Huppert and Simpson, 1980;
Rottman and Simpson, 1983; Bonnecaze et al., 1993; Fleischmann
et al., 1994; Gladstone et al., 1998; Rocca et al., 2008; Chowdhury
et al., 2009). A lock-exchange experiment involves instantaneous
release of a fixed volume of dense fluid into the less dense ambient
fluid. This dense fluid is initially separated from the ambient fluid
by the lock-gate. Once the gate is lifted, the dense fluid flows into
the main body of the tank, displacing the ambient fluid to form a
gravity current.
Our experiments were carried out in a rectangular Plexiglas
lock-exchange tank (430 cm length, 25 cm width, 50 cm height), a
simplified diagram of which is shown in Fig. 1. A removable vertical
aluminum plate located at a distance, x0 ¼24.5 cm, from one end of
the tank serves as the lock-gate. To avoid exchange of fluids between
the two reservoirs (i.e., water and fluid mud reservoirs) before the
experiments commence, the gate is sealed with petroleum jelly around
its edge. The tank is marked by vertical reference lines in every 10 cm
from the lock-gate to the end of the downstream section of the tank.
A mixing tank located next to the lock section is used to prepare the
fluid mud suspension. First, the water reservoir is filled with tap water
to a height of h0. Then, just before the experiment commenced the
fluid-mud reservoir was filled to the same level as the water reservoir.
For each experiment, fluid mud gravity current was generated by the
instantaneous release of the dense fluid mud by lifting the lock-
gate. As the current propagated, its progress was recorded by two
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]] 3

H h0
z

x x0

Fig. 1. Schematic of the experimental set-up: 1—tap water, 2—fluid mud suspension behind the lock-gate, 3—mixing tank, 4—mixer, 5—pump and the piping, 6—lights, and
7—video cameras. Symbols: h0—fluid mud suspension depth before release; x0—length of the lock section; x—horizontal coordinate with the origin at the lock gate;
z—vertical coordinate with the origin at the tank bottom.

high-definition Sony camcorders with 30 frames per seconds from one Table 1
side of the Plexiglas tank. The field of view for each camcorder was Experimental conditions. Parameter values are given for the prepared fluid mud
suspensions before the release.
adjusted to cover the respective half of the channel length. Recorded
videos were later digitized using commercial software to obtain Exp. #a Cm (g/l)b Cv (%)c qc (g/cm3)d h0 (cm)e Rheological properties
30 images per second of recording. A systematic approach was used
to determine the front position and height information over time from m (Pa sn)f ng
the digitized images. First, optical calibration factors (total of 40
1 25 1.0 1.015 15 – –
calibration factors) between the vertical reference lines were calculated 2 50 1.9 1.03 15 0.0016 1.00
for use in converting current geometric characteristics from pixels to 3 75 2.8 1.045 15 0.0340 0.52
length units. The reason of calculating more than one calibration factor 4 100 3.7 1.06 15 0.0880 0.41
is simply to avoid any optical distortion errors that are associated with 5 150 5.4 1.09 15 0.3200 0.30
6 200 7 1.12 15 0.7800 0.24
such large field of views (E200 cm for each camera). Then, the
7 250 8.7 1.14 15 1.6500 0.19
experimental front positions were obtained by summing the distance 8 300 10.2 1.166 15 2.9000 0.17
of the front from the closest vertical reference line upstream of the front 9 350 11.8 1.19 15 4.6600 0.15
and the absolute position of the reference line. The distance of the front 10 400 13.2 1.214 15 7.0000 0.14
from the reference line is calculated by multiplying the total number of 11 450 14.7 1.24 15 10.000 0.124
12 350 11.8 1.19 10 4.6600 0.15
pixels between the front and the reference line and the respective
calibration factor for that position. Similarly, the current height is a
Exp. #—experiment number.
calculated by multiplying the number of pixels corresponding to the b
Cm—sediment mass/volume of water.
c
current height and the calibration factor for a given position. Cv—volume concentration.
d
The characteristics of gravity current propagation depend on the rc—density of the suspension.
e
h0—lock height.
ratio of the current height to the ambient fluid depth (hn =H). The f
m—consistency index.
terms ‘‘shallow ambient’’ and ‘‘deep ambient’’ are used extensively g
n—flow behavior index.
throughout the rest of this manuscript to refer to this ratio (shallow
ambient – 0:075 rðhn =HÞ o1 and deep ambient – hn =H o 0:075,
where hn is the current height and H the ambient fluid depth, see Measured density values are used to calculate the reduced gravity,
Huppert and Simpson, 1980). Many of the pipeline dredge disposal g0 , values of the fluid mud suspensions, a governing parameter for
operations are conducted in shallow ambient. For instance, in the gravity current propagation. Reduced gravity is defined in two
field study by Teeter (2002) for open water pipeline disposal different ways: (i) in terms of the suspension fluid density,
operations, the measured fluid mud turbidity current height is guc ¼ gðrc ra Þ=rc , and (ii) in terms of the ambient fluid density,
30 cm in 2 m water depth within 500 m from the discharge point. gua ¼ gðrc ra Þ=ra (where g is the gravitational acceleration). Both
This corresponds to shallow ambient conditions with hn/H value of definitions of reduced gravity are used in this study.
0.15. Therefore, our experimental conditions are chosen to yield The fluid mud suspension is generally considered to be a
shallow ambient conditions to mimic the field conditions. pseudo-plastic or shear thinning material (Winterwerp and
Kesteren, 2004) that can be expressed by Herschel–Bulkley
constitutive equation (Huang and Garcia, 1998) defined as
2.2. Fluid mud mixture  n1
@u @u
t ¼ ty þm  ð1Þ
@z @z
A total of 12 experiments with different concentrations of fluid
mud suspensions were carried out and tabulated (Table 1). Fluid Here t is the shear stress, ty is the yield stress, u is the x-velocity
mud suspensions were prepared by vigorously stirring cohesive component, @u=@z is the shear rate, n is the flow behavior index
Kaolinite clay particles (mean particle diameter, d¼0.7 mm, and which is a positive real number, and m is the consistency index
particle density, rp ¼2.62 g cm  3) with water in the mixing tank of the suspension. ty , m and n are determined experimentally.
until homogeneity of the suspensions was ensured. The densities of For fluid mud suspensions, the Herschel–Bulkley model can be
the prepared fluid mud suspensions, rc, were measured by an Anton simplified to the Ostwald power-law model given in Eq. (2), which
Paar DMA-35 densitometer that has an accuracy of 0.001 g cm  3. can be considered as an asymptotic case of the Herschel–Bulkley

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
4 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]

model with yield stress, ty ¼ 0:
 n1
@u @u
t ¼ m 
@z @z
Correlation of the power-law constants (m and n) to the volume
concentration (Cv ¼ Vs =ðVs þVw Þ, where Vs and Vw are the volumes
of suspended sediment and water in the prepared suspension,
respectively) for fluid mud suspensions was reported by Ng and
Mei (1994). Following the same procedure as Ng and Mei (1994),
we first measured rheological properties of fluid mud suspensions
with different concentrations and then correlated the calculated m
and n values of suspensions from the measured rheological proper-
ties to the Cv values of suspensions. This enabled us to predict the m
and n values of suspensions within our range of experimental
conditions, minimizing the effects of the measurement range
limitations of our viscometer. The rheological properties of the
prepared fluid mud suspensions were obtained using a BrookField
LVDVII Pro + viscometer with an enhanced UL adapter. Fig. 2a
shows typical apparent viscosity (ma) vs. shear rate and Fig. 2b
shows typical shear stress vs. shear rate measurements for fluid
mud suspensions of only three different concentrations for clarity
purposes. It is evident from Fig. 2a and b that shown fluid mud
suspensions exhibit non-Newtonian behavior as the viscosity
varies with shear rate. Similar non-Newtonian behavior is
observed/expected for all the experimental suspensions, except
the suspensions with the lowest two concentration values (Cv ¼1%
and 1.9%, Exps. #1 and 2 in Table 1) that exhibit Newtonian
behavior. Viscosity measurements are not available for the suspen-
sion used in Exp. #1 due to the limitations of our viscometer.
However, since the suspension used in Exp. #2 (with a higher
concentration value than the suspension used in Exp. #1) exhibits
Newtonian behavior it is evident that the suspension used in
Exp. #1 also exhibits Newtonian behavior. Note that lacking
viscosity measurement for the suspension used in Exp. #1 is of
no consequence as the gravity current formed in Exp. #1 did not
ð2Þ transition into the viscous–buoyancy stage. Ostwald’s power-law
constitutive equation (Eq. (2)) is fitted to the shear stress vs. shear
rate measurements (see solid lines in Fig. 2b) to determine the flow
behavior index n and consistency index m for seven different
prepared suspensions with different concentration values. Fitted
values of these rheological constants (symbols) for these suspen-
sions are shown in Fig. 2c and d. Here, m value increases and n value
decreases with the volume concentration increase. We developed
empirical parameterizations for m and n values as a function of
volume concentration for our fluid mud suspensions by fitting
curves (see solid lines in Fig. 2c and d). The developed parameter-
izations are in the same form as Ng and Mei’s and are as follows:
m ¼ 1  103 Cv3:43 and n ¼ 1:26Cv0:86 (here, m is in Pa sn). The
estimated m and n values using these parameterizations for
all of our experiments (except Exp. #1, and measured values for
Exp. #2) are tabulated in Table 1.
3. Propagation phases and transition times
As soon as the lock-gate was lifted, a fluid mud turbidity current
formed and began propagating in the 2-D experimental tank. Fig. 3
presents a photograph of this fluid mud turbidity current propa-
gating over a horizontal bottom. As is evident, the propagating
current forms a frontal zone with a distinct dividing line between
the intruding fluid mud current and ambient fresh water. The flow
structure in this frontal zone is rather complicated (see photo
visualization in Fig. 1 of Chowdhury et al., 2009) and is associated
with lobe-cleft patterns at the leading-edge and billows that form
above and behind the head of the current (Chowdhury et al., 2009;
Simpson, 1997). As the turbidity currents propagate, they undergo

60 0.6

40 0.4
μa (cP)

τ (Pa)

Exp. #3
Exp. #4
20 Exp. #3
Exp. #5 0.2 Exp. #4
Exp. #5
Eq. 2
0 0
0 40 80 120 0 40 80 120
∂u/∂z (1/s) ∂u/∂z (1/s)

10 1
m (Pa sn)

5 0.5
n

0 0
0 5 10 15 0 5 10 15
Cv (%) Cv (%)

Fig. 2. Rheological properties of the fluid mud suspensions: (a) measured apparent viscosity, ma, as a function of shear rate, @u/@z, (b) shear stress, t, as a function of shear rate,
@u/@z, for three different concentrations of suspensions, (c) consistency index, m, as a function of volume concentration, Cv, (d) flow behavior index, n, as a function of volume
concentration, Cv. Symbols represent measurements and the legend identifies the corresponding experimental conditions given in Table 1. The solid line in (b) indicates the
fitted Power-law model, and solid lines in (c) and (d) represent the estimations by the developed empirical parameterizations for m and n values as a function of Cv.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]
Fig. 3. Propagation of the fluid mud gravity current in Exp. #5 after 9 s of releasing the fluid mud suspension. The current front position, xn, is at 102 cm from the lock-gate.
400
300
xn(cm)
200
100
0
0 25
Fig. 4. Front position of the fluid mud gravity currents, xn, as a function of elapsed propagation time, t. Symbols represent the data from the experiments tabulated in Table 1
and are defined in the legend.
three distinct phases (slumping, self-similar and viscous) as
discussed in the introduction. Because propagation characteristics
of each phase vastly differ, a different modeling approach is
required for each phase. Therefore, it is important to identify the
phase transition positions/times to implement the appropriate
modeling approach. In this section, the authors describe identifica-
tion of the phase transition times for 2-D fluid-mud turbidity
currents propagating over a horizontal bottom.
In Fig. 4, the experimentally measured front positions, xn (see
definition sketch in Fig. 1), of turbidity currents generated by the
release of the dense fluid mud with different initial concentrations
as a function of propagation time, t, are presented. As is evident, the
initial front propagation velocities of the suspensions with higher
initial concentrations are larger due to the larger driving gravita-
tional/buoyancy force. Henceforth, gravitational and buoyancy
force expressions are used interchangeably. However, all propaga-
tion curves have a similar initial trend, a steep constant-slope initial
portion indicating a constant front propagation velocity. These
steepest portions of the curves correspond to the slumping phases
of the respective experiments. It is difficult to discern the self-
similar and viscous phase transitions from this figure. In order to
determine the transitions to first self-similar and then viscous
phases, we used Rottman and Simpson’s procedure (1983) as
described below.
For saline gravity currents, Rottman and Simpson (1983)
showed that a log–log plot of the dimensionless front position
(Xn ¼xn/x0, where x0 is the lock length and h0 the lock height,
see Fig. 12pofffiffiffiffiffiffiffiffiffi
Rottman
ffi and Simpson, 1983) vs. dimensionless time
(T ¼ t=x0 = gua h0 ) collapses all the data to give constant slopes of 1,
2/3 and 1/3 for slumping, self-similar and viscous phases, respec-
tively. In Fig. 5, a log–log plot of Xn vs. T for the fluid mud gravity
currents from our experiments with different initial concentrations
Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
5
Exp. # 1
Exp. # 2
Exp. # 3
Exp. # 4
Exp. # 5
Exp. # 6
Exp. # 7
Exp. # 8
Exp. # 9
Exp. # 10
Exp. # 11
Exp. # 12
50 75 100 125
t (s)
is shown. In our non-dimensionalization and all our calculations
presented later in the paper, we used gua , initial value of gua behind
the lock, instead of the gua value at a given position. It is assumed
that the current density does not change significantly for our
experimental propagation distances and times due to low settling
velocities of clay particles and negligible entrainment of ambient
fresh water. To verify this assumption, we repeated Exp. #11 with
the highest initial suspension concentration and sampled small
amounts of fluid mud from the propagating current at 1, 2 and 3 m
from the lock-gate using sampling probes operating under vacuum.
The measured densities of those collected samples at different
positions showed negligible variation from the initial value,
verifying our assumption. Our experimental data (Fig. 5) is
characterized by three distinct slopes: E1 (more accurately
0.95) for slumping phase, 2/3 for self-similar phase, and a variable
slope value as a function of the n value of the suspension (see
Eqs. (15) and (17) in Section 5) for the viscous phase. As expected,
slope values for the slumping and self-similar phases conform to
Rottman and Simpson’s findings for compositional Newtonian
gravity currents. Please note that given the limited length of our
experimental tank only some of the generated turbidity currents
(Exps. #7–12 in Table 1) had sufficient propagation distance to
transition into a viscous propagation phase. After the slumping
phase, all of the currents (Exps. #1–11 in Table 1) exhibited the self-
similar phase, except for the current in Exp. #12 (see Table 1) that
transitioned directly into the viscous phase without transitioning
through the self-similar phase. Though the dimensionless propa-
gation time necessary for the current to transition into viscous
phase decreased due to an increased concentration, the dimension-
less transition time for the self-similar phase remained unaffected
(Fig. 5). It is important to note that while slumping and self-similar
propagation characteristics of fluid mud turbidity currents are
6 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]

Exp. # 1

Xn= xn/x0
Exp. # 2
Exp. # 3
5 Exp. # 4
Exp. # 5
Exp. # 6
Exp. # 7
Exp. # 8
Exp. # 9
Exp. # 10
Exp. # 11
Exp. # 12
0.5
1 10 100
T= t/ (x0 /(g h0)0.5)

20
Slope = 0.46
Slope =0.5
Slope = 0.37
Slope = 0.31
Xn= xn/x0

Slope = 0.32
Slope = 0.29
Exp. No. 7
Exp. No. 8
Exp. No. 9
Exp. No. 10
Exp. No. 11
Exp. No. 12
2
10 100
0.5
T= t/ (x0 /(g h0) )
Fig. 5. Log–log plots of dimensionless front position, Xn, vs. dimensionless time, T, for all experimental currents. Experimental data for all three propagation phases
(i.e., slumping, self-similar, and viscous) are shown in (a) and the part of the data corresponding to only the viscous phase for Exps. #7–12 are shown in (b). Symbols represent
the data from the experiments tabulated in Table 1 and are defined in the legend. Solid lines represent the slopes of the current position curves and the corresponding values
are given in the graphs.

similar to that of saline gravity currents, viscous propagation Table 2

characteristics of fluid mud turbidity currents and saline gravity The transition time and length for the slumping and viscous stage for experiments
in Table 1. The front positions xn at tn and tnn are determined from the
currents are quite different as evidenced by different slopes for the
experimental data.
propagation curves. This is because in the slumping and self-similar
phases inertia and buoyancy forces are the driving forces and Exp. tn (s) (from xn at tn (cm) tnn (from tnn (s) xn at tnn (cm)
the rheological differences in the fluids (i.e., non-Newtonian # Eq. (3)) (experimental) Eq. (4)) (experimental) (experimental)
versus Newtonian fluid behavior) are not of primary importance.
1 39.2 238.0 – – –
Conversely, because the buoyancy and the viscous forces are 2 27.8 236.0 401.4 – –
the driving forces in the viscous phase, rheological differences 3 22.8 241.0 134.3 – –
determine viscous propagation characteristics. 4 19.9 234.0 88.7 – –
5 16.4 248.0 52.2 – –
There are two phase transition times characterizing transitions:
6 14.3 250.0 35.4 – –
(i) from slumping phase to self-similar phase, tn, and (ii) from 7 12.9 255.0 25.7 26.0 363.6
slumping/self-similar phase to viscous phase, tnn. Determining 8 11.9 240.0 19.9 18.0 303.6
the value of dimensionless transition time to self-similar phase 9 11.1 249.0 16.1 16.0 298.5
pffiffiffiffiffiffiffiffiffiffiffiffiffi 10 10.5 215.5 13.8 12.0 249.7
½T  ¼ t  =ðx0 = gua,0 h0 Þ value is relatively straightforward using our 11 9.9 225.0 11.5 11.0 236.0
experimental data in Fig. 5. This experimental data from slumping 12 13.6 188.0 13.1 10.7 179.9
and self-similar phases collapses on two separate lines (shown as
solid lines in Fig. 5) and the Tn value corresponds to the T value at
the intersection point of these two lines (i.e., T¼24):
decreases as the initial concentration of the suspension increases.
t However, since the propagation velocity of higher concentration
T ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 24: ð3Þ
x0 = gua,0 h0 suspensions are higher, the transition length xn (xn ¼xn at tn), the
propagation distance till the end of slumping phase, is approximately
The transition time to self-similar phase, tn, for all of the experi- constant (E 215–255 cm) for all the experiments, except Exp. #12.
ments are given in Table 2. It can be seen from this table that tn This xn value corresponds to approximately 9–10 lock-lengths for our

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]] 7

experimental set-up and conforms to observations of previous studies current as follows:

for compositional Newtonian gravity currents that reported xn values
of 5–10 lock-lengths (see Meiburg and Kneller, 2010). For very high
concentration suspensions, the current may make a transition to
viscous phase even before the slumping phase has been completed
and the self-similar phase may be absent (see Exp. #12 in Fig. 5). This
is the reason why the slumping length, xn, in Exp. #12 is much less
than that in our other experiments.
The transition time to viscous phase, tnn, is determined both
experimentally and theoretically and is presented in Table 2. The
experimental tnn value is calculated by determining the dimensionless
time value that the experimental data starts deviating from the solid
trend lines in Fig. 5. The theoretical tnn is calculated using the
expression provided by Di Federico et al. (2006) for a non-Newtonian
gravity current. The simplified form of this theoretical expression for a
fixed volume release as in our experimental case is given in Eq. (4):
!1=ðn þ 6Þ
r3c q0 n þ 3
t  ¼ c1 ð4Þ
m3 guc2n
Here, q0 is the volume released per unit width. This expression is
based on dimensional considerations when the viscous force becomes
comparable to the inertia force acting on a gravity current. Therefore,
we embedded an empirical constant coefficient, c1, term in Eq. (4).
Using the experimental tnn values, we estimated the value of c1 as 6.
Experimental and calculated tnn values using Eq. (4) are presented in
Table 2.
xn ¼ KI ðguc q0 Þ1=3 t 2=3 : ð7Þ
Here, KI is a constant of proportionality and its value should be
obtained experimentally. The front position predictions by Eq. (7) and
the experimental observations are presented in Fig. 6. Here, different
symbols represent the experimental data for fluid mud suspensions
with different initial concentrations and solid lines represent the
predictions for the corresponding suspension. As is evident, the
predictions from this simple model agree well with the experimental
observations, except for the earlier times of the current propagation.
The discrepancy in the earlier stages of the propagation is due to the
omission of ambient water dynamics in the model. As soon as the gate
is lifted in a lock-exchange set-up, an initial counter-flow of ambient
water occurs above the forming fluid-mud current which then hits the
end-wall and is reflected back, forming a bore (Rottman and Simpson,
1983). This bore affects the current until its energy is dissipated
(i.e., till the end of slumping phase). Clearly, the simple expression
in Eq. (7) is incapable of modeling this complex phenomenon.
The best fit value of KI for all of our experiments is calculated to be
1.14, except for Exps. #11 and 12. Average R2 value for Exps. #1–10
with KI ¼1.14 is 0.974 (see Table 3) and the average R2 value decreases
to 0.876 for the present value of KI ¼1 that represents an exact balance
between the inertia and buoyancy forces. KI value being larger than
this preset value is partly due to the ambient water dynamics in the

400
4. Inertia–buoyancy stage
300 Exp. # 1
Here we describe the first propagation stage of the fluid mud
xn (cm)

Exp. # 3
gravity currents in which the inertia and buoyancy forces are the 200
Exp. # 5
driving forces. This section details the laboratory tests of five Exp. # 7
discrete mathematical models to evaluate their predictive cap- Exp. # 9
100
abilities for characterizing the fluid mud turbidity currents in the Exp. # 11
inertia–buoyancy stage. Tested models are at different levels of Eq. 7
complexities. These models, from the simplest to the most complex 0
0 25 50 75 100
are the force-balance model, box models for the compositional and t (s)
suspension currents, and shallow water models for the composi-
tional and suspension currents. All five are based upon the Fig. 6. Comparison of the current front position predictions by the force-balance
model (solid lines, Eq. (7)) with the experimental data. Every alternate experiment is
assumption that the current is fully inviscid and the flow is
shown in the figure for clarity purposes. Symbols represent the data from the
governed by purely inertia–buoyancy balance. Therefore, we experiments tabulated in Table 1 and are defined in the legend.
used our laboratory’s experimental data for t otnn (i.e., discarding
the viscous phase data for Exps. #7–12) in this evaluation. To
ensure the integrity of our manuscript, each model is briefly
Table 3
introduced, and appropriate references are provided for the Coefficient of correlation, R2, values for model predictions in the inertia–buoyancy
detailed derivations. stage of the experiments in Table 1.

Exp. # FBa CBb SBc CSWd SSWe

4.1. Force-balance model (Fr¼ 0.8) (Fr¼ 1.19)

This simple single-equation model for the inertia–buoyancy 1 0.98 0.96 0.96 0.95 0.93
2 0.99 0.93 0.95 0.99 0.948
stage is based upon the balance between the buoyancy and the 3 0.98 0.98 0.98 0.98 0.95
inertia forces (see Huppert, 1982 for details). From dimensional 4 0.98 0.92 0.94 0.97 0.94
grounds, the order of magnitudes of the inertia, Fi, and buoyancy, 5 0.97 0.97 0.97 0.98 0.98
Fg, forces for the two-dimensional spreading of a gravity current 6 0.98 0.86 0.88 0.98 0.94
7 0.97 0.93 0.94 0.96 0.92
generated from a fixed initial volume can be estimated as follows:
8 0.98 0.99 0.94 0.97 0.91
rc guc q20 w 9 0.95 0.99 0.94 0.95 0.92
Fg  , ð5Þ 10 0.96 0.99 0.95 0.93 0.92
x2n 11 0.88 0.99 0.99 0.80 0.60
12 0.60 0.99 0.99 0.80 0.65
rc q0 xn w
Fi  , ð6Þ
t2 a
FB—force balance model (with KI ¼1.14).
b
CB—compositional box model.
where w is the width of the current, and  implies an order of c
SB—suspension box model (with Cs ¼ 1).
magnitude relationship. Equating Eqs. (5) and (6), one can obtain d
CSW—compositional shallow water model.
the expression for the front position of a two-dimensional gravity e
SSW—suspension shallow water model.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
8 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]

vicinity of the gate as described above and partly due to viscous
effects. Though initially small, the viscous effects steadily accumulate
as the current propagates (see Section 3).
4.2. Box models
Box models assume that the current evolves in a series of equal-
area rectangles with uniform properties at any instant of time
(Huppert, 1998). Two different types of box models are used to
model the inertia–buoyancy dominated propagation of composi-
tional and suspension gravity currents. Although fluid mud gravity
currents are suspension-driven currents, given the very small
settling velocity of the cohesive sediment and the experimental
propagation length, the box model for compositional gravity
currents is appropriate for our analysis as discussed later in
Section 6. To ensure completeness, both models and their predic-
tions, first for compositional currents and then for suspension
currents, are presented in this section.
The box model solution for a compositional gravity current was
obtained by combining two governing conditions (see the detailed
derivation by Huppert and Simpson, 1980): (i) conservation of volume
pffiffiffiffiffiffiffiffiffi
(i.e., q0 ¼ x0 h0 ¼ xn hn ); and (ii) Froude number, Fr ¼ Un = guhn
(where Un is the current front velocity), condition at the current
head. There are two different Fr conditions for shallow and deep
ambient fluid; for our purposes, the shallow ambient Fr condition is
more applicable. The resulting parameterization for the front position
of a gravity current from the lock-gate is
 6=7
7 used hindered settling velocity (Wh) which is suitable for cohesive
xn ¼ ðgu3 q0 H2 Þ1=6 t : ð8Þ
12 a
In Fig. 7, the front position predictions expressed in Eq. (8) and
the experimental observations are presented. Here we see that
for most part of the current propagation, the predictions of the
compositional box model closely agree with the experimental data
whereas at the later propagation times, the compositional box
model over-predicts the current front position. This over-predic-
tion is because the gravity current begins to experience increasing
viscous affects while the box model is constructed with the inviscid
flow assumption. At these later propagation times, the current flow
is not purely dominated by the inertia–buoyancy balance; rather a
relatively small magnitude viscous force retards the current
propagation. We hasten to note, however, that unlike force-balance
expression (Eq. (7)), the compositional box model solution has
no adjustable parameter. Therefore, without any adjustable
500
400
300 Exp. # 1
xn (cm)
parameters, the compositional box model agrees well with the
experimental observations for much of the current propagation.
In addition to the two governing conditions stated above, the
box model for the suspension gravity currents is also characterized
by a third governing condition that incorporates the evolution of
particle concentration due to deposition as the gravity current
propagates (Gladstone and Woods, 2000). In our analysis, we
employ the box model solution for 2-D suspension currents
proposed by Dade and Huppert (1995). Omitting the details and
derivations (see Dade and Huppert, 1995), the dimensionless front
position, Xn, for a suspension current in a shallow ambient is
parameterized as follows:
 1=7
Q0 D 2
Xn  fs ðss TÞ ð9Þ
6 ss
Here,
!7=13 !
D2=7 b q0
ss ¼ , Q0 ¼ 2
6:9Cs Q0
6=7 h0
is the dimensionless
pffiffiffiffiffiffiffiffiffiffiffiffi
ffi 2-D volume behind the lock-gate, b
(Ws = gua,0 h0 ) is the dimensionless settling number, and
D ð ¼ H=h0 Þ is the dimensionless ambient water depth. Cs is a free
parameter that is equal to unity for a compositional gravity current
and varies for a suspension current based on a number of factors
such as initial concentration and size of particles (Dade and
Huppert, 1995). The expression for fs ðss TÞ is found in Appendix B
of that article. It should be noted that in our calculations of Eq. (9)
for fluid mud suspension currents investigated in this study, we
sediments and concentrated suspensions as suggested by
Winterwerp and Kesteren (2004) and Coussot (1997) instead of
the Stokes settling velocity (Ws ¼ ðrp rw Þgd2 =18mw ; mw and rw are
the dynamic viscosity and density of water, respectively) that is
used by Dade and Huppert (1995) to model the experimental
observations of non-cohesive suspension current of Bonnecaze
et al. (1993). The hindered settling velocity can be expressed as
(Winterwerp and Kesteren, 2004):
 p
kCv
Wh ¼ Ws 1 : ð10Þ
100
Here k and p are empirical constants. We used k ¼1 and p ¼5.1
for estimating hindered settling velocity of the clay particles since
they have been found to agree well with the experimental
observations for cohesive sediments (see page 19 of Coussat,
1997). A comparison of the front position predictions by the box
model solution for suspension currents and our experimental
observations for fluid mud suspension currents is given in Fig. 8.
Interestingly, the best fit value for the free parameter Cs with
500
400

Exp. # 3
Exp. # 1
Exp. # 5 300
xn (cm)

200 Exp. # 3
Exp. # 7 Exp. # 5
200 Exp. # 7
Exp. # 9
100 Exp. # 9
Exp. # 11 100 Exp. # 11
Eq. 8 Eq. 9
0 0
0 25 50 75 100 0 25 50 75 100
t (s) t (s)

Fig. 7. Comparison of the current front position predictions by the compositional Fig. 8. Comparison of the current front position predictions by the suspension box
box model (solid lines, Eq. (8)) with the experimental data. Every alternate model (solid lines, Eq. (9)) with the experimental data. Every alternate experiment is
experiment is shown in the figure for clarity purposes. Symbols represent the data shown in the figure for clarity purposes. Symbols represent the data from the
from the experiments tabulated in Table 1 and are defined in the legend. experiments tabulated in Table 1 and are defined in the legend.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]] 9

coefficient of correlation values, R2 4 0.92, is 1 for all the experi-

ments. This observation confirms our earlier statement that fluid 600
mud gravity currents behave like compositional gravity currents
for small propagation times. Since the best Cs fit value is unity for 500
suspension-box model solution, the prediction curves of the model
400
are similar to that of compositional counterparts.

xn (cm)
Exp. # 1
300 Exp. # 3
Exp. # 5
4.3. Shallow water models 200 Exp. # 7
Exp. # 9
Shallow water models for gravity current propagation are based 100 Exp. # 11
on the Navier–Stokes solutions with the basic assumption that Eq. 11
0
vertical accelerations are negligible, so that the pressure field is
0 25 50 75 100
purely hydrostatic. Unlike box models, shallow water models
t (s)
provide horizontal variations of current properties (such as height)
(Kneller and Buckee, 2000). Depending upon the relative depth
of the ambient fluid (h/H, see Huppert and Simpson, 1980), either
one-layer or two-layer shallow water model formulations 400
are employed (see Meiburg and Kneller, 2010; Ungarish, 2009).
One-layer shallow water formulations neglect the motion of the
overlying fluid; hence, they are more applicable for a gravity 300 Exp. # 1

xn (cm)
current with a deep ambient (h/Ho0.075). Conversely, two-layer Exp. # 3

shallow water formulations account for also the dynamics of the 200
Exp. # 5

overlying ambient fluid layer, and are more applicable for a gravity
current with a shallow ambient (0.075rh/Hr1). For the propaga-
tion of the compositional gravity currents in the inertia–buoyancy
stage, a similarity solution for the one-layer shallow water model
exists. However, a closed analytical solution for the two-layer
model is unavailable and numerical solution is required. Nor is a
similarity solution available for the one- or two-layer shallow
water model for suspension gravity currents where deposition is
used to incorporate a progressive density change into the current.
Based upon the reasoning in Section 4.2, compositional shallow
water models are adequate for the analysis of our experimental
observations. To ensure completeness, both shallow water models
and their predictions, first for compositional currents and then for
suspension currents, are presented in this section. Although the
two-layer models are more applicable for our experimental con-
ditions, one-layer shallow water models are employed in our
analysis for simplicity.
We first discuss one-layer compositional shallow water model
and its predictions. As detailed derivations and formulations of the
one-layer model are found in Bonnecaze et al. (1993), we provide
only a brief description here. Similarity solution for the dimension-
less front position, Xn, is expressed as
 1=3
27Fr 2 Q0
Xn ¼ T 2=3 :
122Fr2
Given that one-layer shallow water model assumes a deep
ambient, deep water Fr condition (Fr¼1.19) is typically used in
Eq. (11). In our calculations we first employed the deep water Fr
determined from experimental observations by Huppert and
Simpson (1980) and then we employed the actual Fr measurements
that correspond to the shallow ambient condition in our
experiments.
The front position predictions by Eq. (11) with Fr¼1.19 and
experimental observations are presented in Fig. 9a. Here, the model
predictions significantly over-predict the experimental observa-
tions, mainly due to the use of deep water Fr condition. If an Fr value
of 0.8 is used, which is a more representative value for our shallow
ambient experiments, the model predictions improve significantly
(see Fig. 9b). The discrepancy observed here between the model
predictions and the experimental data at the initial propagation
times corresponds to the slumping phase of propagation in which
Exp. # 7
Exp. # 9
100 Exp. # 11
Eq. 11
0
0 25 50 75 100
t (s)
Fig. 9. Comparison of the current front position predictions by the compositional
shallow water model (solid lines, Eq. (11)) using two different Fr conditions ((a)
Fr¼ 1.19 and (b) Fr ¼0.8) with the experimental data. Every alternate experiment is
shown in the figure for clarity purposes. Symbols represent the data from the
experiments tabulated in Table 1 and are defined in the legend.
the overlying water is very dynamic. The one-layer shallow water
model does not consider this phenomenon.
Because there is no exact solution for one- or two-layer shallow
water models of suspension gravity currents, shallow water models
for suspension currents are solved numerically (e.g., Bonnecaze
et al., 1993). Hogg et al. (2000) and Harris et al. (2001) in two
separate communications provided asymptotic solutions for
shallow water models of suspension gravity currents. These
solutions are obtained for the case when the settling velocity
of the particles is much lesser than the initial velocity of propaga-
ð11Þ
tion of these currents (Hogg et al., 2000). This is clearly the case for
our experiments with cohesive sediments as discussed earlier.
Since the intention of asymptotic solution by Harris et al. (2001) is
to capture the flow behavior throughout the current’s entire
propagation within the inertia–buoyancy stage, rather than using
Hogg’s solution (with propagation time limits), we employed
Harris et al.’s solution in our analysis (see Eq. 3.69 of Harris
et al., 2001). See Harris et al. (2001) to reference these formulations
of the asymptotic solution in their entirety. The predictions by
Harris et al.’s solution and our experimental observations are
presented in Fig. 10, which shows a close agreement between
the predictions and the overall experimental data trend. As in
the case of one-layer compositional shallow water model predic-
tions, the discrepancy between the predictions and the data
are attributable to the use of the deep water Fr condition. The
use of a more representative shallow ambient Fr condition is quite
complex in Harris et al.’s solution and is out of the scope of
this study.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
10 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]

4.4. Inter-model comparisons
The models are compared based upon the goodness of predic-
tions by calculating the coefficient of correlation, R2, between the
experimental data and the model predictions. Tabulations of the R2
values for the predictions of the force-balance model, the composi-
tional and suspension box models, the compositional (with Fr¼0.8)
and suspension shallow water models (with Fr¼1.19) are listed in
Table 3. Predictions of the described five models and the select
experimental data (Exps. #1, 3, 5, 7, 9, 11) are presented in a single
figure (Fig. 11) for comparison. Fig. 11 is plotted as dimensionless
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
front position (Xn ¼ xn =x0 ) vs. dimensionless time (T ¼ t= ðh0 =gua Þ)
to collapse predictions for all the experiments of a given model on a
single line.
As can be seen from Table 3, R2 values are more than 0.9 for the
force-balance and box models for all experiments, except Exps. #11
and 12 for the force-balance model and Exp. #6 for the box models.
The R2 values for compositional and suspension box models are
close. This is expected as the best fit value of the free parameter Cs
in the suspension box model is 1. In calculating R2 values, we
400
300
xn (cm)
considered only the portion of the current in the inertia–buoyancy
stage and discarded the data from viscous–buoyancy stage. Please
note that only in Exps. #7–12 the current showed an apparent
transition to viscous stage (see Fig. 5a and b). The R2 values for
Exp. #6 (with Cv ¼7%) for box models are smaller than the R2 values
for other experiments, because the current experiences more
viscous effects than in other experiments. These more pronounced
effects are due to the current propagating farther in the inertia–
buoyancy stage and closer in the viscous–buoyancy stage due to
higher viscosity associated with higher concentration currents.
Since the viscous–buoyancy stage is not fully discernible in the log–
log plot of dimensionless front position vs. dimensionless time (see
Fig. 5), we could not discard the experimental data in Exp. #6 that
likely to have originated from initial part of the viscous–buoyancy
stage and we expect that this is the main reason for the lower R2
values. On the other hand, the R2 values for Exps. #7–12 with the
highest concentrations among our experiments are high. This is
because we could identify the viscous transition position and used
only the experimental data from inertia–buoyancy stage, discard-
ing the data from viscous–buoyancy stage. The force-balance
expression shows quite high R2 values for Exps. #1–10 (even for
Exp. #6), possibly due to an empirical adjustment parameter in the
force-balance expression.
As can be seen from Fig. 9a, predictions of shallow water model with
deep ambient Fr condition (Fr¼1.19) significantly overestimate the
experimental observations. Therefore, R2 values for these predictions
are not included in Table 3. Conversely, predictions of the shallow water

Exp. # 1 model with the adjusted shallow ambient Fr condition (Fr¼0.8) and
200 the suspension shallow water model with deep ambient Fr condition
Exp. # 3
Exp. # 5
are in good agreement with the experimental observations and
Exp. # 7
exhibit higher R2 values (see Table 3). Of the shallow water models
100
Exp. # 9
with different input conditions, the compositional shallow water
model with Fr¼ 0.8 provides the most accurate predictions. This
Exp. # 11
0 predictive capability indicates that, given the appropriate Fr value
0 25 50 75 100 input into the model, the compositional shallow water model is the
t (s) appropriate modeling selection. However, because determining the
correct Fr value requires using either direct measurement or modeling
Fig. 10. Comparison of the current front position predictions by the suspension using a different approach (e.g., use of a box model), implementing a
shallow water model (solid lines) using deep ambient Fr conditions (i.e., Fr ¼1.19)
with the experimental data. Every alternate experiment is shown in the figure for
compositional shallow water model may be impractical for shallow
clarity purposes. Symbols represent the data from the experiments tabulated in ambient. The suspension shallow water model with Fr¼ 1.19 exhibits
Table 1 and are defined in the legend. R2 values close to those of compositional shallow water model with

30

20
Xn= xn/ x0

FB - Eq. 7
10
CB - Eq. 8
SB -Eq. 9
CSW -Eq. 11 with Fr = 0.8
SSW

0
0 50 100
0.5
T = t / (h0 /g )

Fig. 11. Inter-model comparisons. Open circles represent experimental data from Exps. #1, 3, 5, 7, 9, 11 and lines (see legend) represent the model predictions. Data from every
alternate experiment is shown in the figure (represented by the same symbol) for clarity purposes. In the legend, FB—force balance model; CB—compositional box model;
SB—suspension box model; CSW—compositional shallow water model (with Fr ¼0.8); SSW—suspension shallow water model.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]
Fr¼0.8. Therefore, the effect of the Froude number condition is more
pronounced in compositional shallow water models than in suspension
shallow water models. This discrepancy is perhaps attributable to the
incorporation of the particle deposition in suspension shallow water
models. Deposition reduces the current propagation velocity. As seen in
Fig. 9a, the use of Fr¼1.19 in compositional shallow water model
provides an over-prediction of the experimental propagation distance.
Sediment deposition effects in suspension shallow water models
compensate for this over-prediction to some degree. We expect better
R2 values if appropriate Fr values are used in the suspension shallow
water model; however, the use of this relatively cumbersome model
(see Harris et al., 2001) is out of the scope of this work.
Among the models considered with different complexities, the
simplest force-balance model provided the best predictive capability
for Exps. #1–10. The experimental adjustment parameter, KI, that
gives the best fit to the experimental data is constant (¼1.14) for the
range of fluid-mud concentration values tested, except for Exps. #11
(best fit KI ¼1 with R2 value of 0.95) and 12 (best fit KI ¼0.93 with R2
value of 0.92). This constancy is especially important for using the
simplistic force–balance approach in quick estimations. Both com-
positional and suspension box models provided accurate estima-
tions with the latter providing slightly more accurate estimations.
Compositional and suspension shallow water models, the most
complex models among all those considered in this study, provided
the least accurate predictions. This is mainly because single-layer
shallow water models that are appropriate for deep ambient are
used for predicting the experimental front propagation distances
that correspond to shallow ambient experimental conditions. We
expect that the use of two-layer shallow water models will
significantly improve the predictions albeit at the expense of
simplicity. Both force-balance and compositional box model solu-
tions are obtained for compositional gravity currents such as a saline
solution where there is no particle settling. The good prediction of
both models for the fluid mud gravity current would encourage
premature conclusions that fluid mud turbidity current can be
considered as a compositional gravity current for short propagation
distances/times as in our experiments.
5. Viscous–buoyancy stage
This section elucidates the second propagation stage of the fluid
mud gravity currents in which the viscous and buoyancy forces are
the driving forces. As discussed earlier, fluid mud turbidity currents
with sufficiently high levels of concentration (see Table 1) have a
non-Newtonian rheology and their viscous propagation character-
istics vastly differ from the Newtonian gravity currents. Therefore,
using mathematical models developed for viscous propagation of
Newtonian gravity currents are inadequate for modeling viscous
propagation of fluid mud gravity currents. To our knowledge, there
are no viscous force-balance and compositional and suspension
box models for non-Newtonian gravity currents. A force-balance
expression for the front position of the viscous non-Newtonian
gravity current is derived in Section 5.1. There is a recent self-
similarity solution (Di Federico et al., 2006) for a one-layer
compositional shallow water model for viscous propagation of a
non-Newtonian gravity current that has not been laboratory tested
yet. These two available models, non-Newtonian force-balance
(Section 5.1) and compositional shallow water (Section 5.2), were
laboratory tested to evaluate their predictive capabilities. In six of
our experiments (Exps. #7–12 in Table 1), fluid mud gravity
currents propagated in the viscous–buoyancy stage for a significant
time period. Experimental data from these two experiments for
t 4tnn (i.e., discarding the data corresponding to inertia–buoyancy
stage) are used in our evaluation.
Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
11
5.1. Force-balance model
A force-balance expression for the front position in the viscous–
buoyancy stage of a non-Newtonian gravity current was obtained
by equating the order of magnitudes of buoyancy and viscous
forces acting on the current. In this derivation, we followed a
similar procedure by Didden and Maxworthy (1982) by considering
the non-Newtonian rheological properties of fluid mud suspen-
sions expressed by Ostwald power-law constitutive equation
(Eq. (2)) as follows.
The volume continuity requirement for a 2-D gravity current
gives the following expression (Didden and Maxworthy, 1982;
Huppert, 1982):
V ¼ hn ðtÞxn ðtÞ ¼ qfb t a ð12Þ
Here, V is the volume released per unit width from the source,
hn is the representative height of the current of length xn, and qfb
( 40) and a ( Z0) are constants. The case of a ¼0 corresponds to the
release of a fixed volume of fluid, while the case of a ¼1 corresponds
to a constant inflow rate. For a ¼0, qfb represents the initial volume
per unit width (i.e., qfb ¼ q0 ¼ V).
Using Eq. (12), the order of magnitude of the buoyancy force
acting on a 2-D gravity current can be expressed as (Huppert, 1982)
rc guc q2fb t2a w
Fg  rc guc h2n w  : ð13Þ
x2n
Note that Eq. (13) is obtained for the general case of a time-
dependent source condition, and for the specific case of constant
volume release (i.e., a ¼0) Eq. (13) simplifies to Eq. (5).
The order of magnitude of the viscous force acting on a 2-D
non-Newtonian gravity current, based on the Ostwald power-law
constitutive relationship (Eq. (2)), can be obtained as follows:
 n
Un nða þ 1Þ
Fv  m xn w  mxn 2n þ 1 qn
fb t w: ð14Þ
hn
Equating Eqs. (13) and (14), one can obtain the following force-
balance relationship for the current front position:
 1=ð2n þ 3Þ
ð2 þ nÞ=ð2n þ 3Þ rc guc
xn ¼ Kv qfb t ðaðn þ 2Þ þ nÞ=ð2n þ 3Þ : ð15Þ
m
Here Kv is a free parameter included based upon dimensional
considerations and it needs to be determined experimentally.
Calculating the model predictions for the viscous spreading
characteristics of the gravity currents in lock-exchange experi-
ments is cumbersome and requires a careful consideration of the
problem (see also the discussion in page 257 of Ungarish, 2009).
Although all of our experiments were carried out by releasing a
fixed volume of dense fluid mud (i.e., a ¼0), the direct implementa-
tion of the respective expressions of the force-balance (Eq. (15))
and the shallow water models (presented in Section 5.2, Eq. (17)
later) for a ¼0 is inadequate. Upon release of the fluid mud
suspension, the initial propagation of the current in the inertia–
buoyancy stage is considered to be inviscid. When the viscous
transition occurs after a significant propagation distance, the
current has already been well-established with a characteristic
height and velocity. Though the viscous transition occurs over a
finite propagation distance, for modeling simplicity it must be
assumed that the transition occurs at a well-defined spatial
position. When the current head passes the viscous transition
position, most of its body is still in the inviscid region. For viscous
propagation modeling purposes, the portion of the current in the
inviscid region can be replaced with a source of fluid mud located
at the viscous transition position (i.e., xn at tnn). A crude, but
an essential assumption, in our analysis is that this source supplies
a constant flux (V_ ¼ U  h , where Unn and hnn are the representa-
tive velocity and height of the laboratory gravity current at
12 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]

the transition position, respectively) fluid mud (i.e., a ¼1). Table 4

Characteristics of this hypothetical viscous gravity current origi- Proportionality constant, Kv, and coefficient of correlation, R2, values for viscous
non-Newtonian force-balance and shallow water models in the viscous–buoyancy
nating from the source are representative of the characteristics of
propagation stage of the experiments in Table 1.
the laboratory generated fluid mud gravity current in the viscous–
buoyancy stage. Considering the transition position as the origin, Exp. # Force-balance model Shallow water model
the total released volume of fluid mud per unit width, V, at a
particular time, t, from the source can be calculated as Kv R2 Kv R2

V ¼ h U  t ð16Þ 7 0.93 0.965 0.9287 0.962

8 0.96 0.9 0.933 0.896
Equating Eqs. (12) and (16) for a ¼1, one can show that 9 0.8 0.98 0.936 0.86
10 1 0.987 0.941 0.975
qfb ¼ h U  . Once qfb is calculated, current front position with 11 0.98 0.966 0.9433 0.952
respect to the viscous transition position can be determined by 12 0.8 0.98 0.936 0.944
using Eq. (15). Simplified representation of the inviscid portion of
the current as a source at the transition position has limitations.
The viscous transition position cannot be considered to be origin
(i.e., location of the source) for the full duration of the viscous source approach are in good agreement with the experimental data
current propagation. When the entire current is in the viscous– until V¼q0 and deviations of the predictions from the experimental
buoyancy stage, the source location should be shifted. Hence, in observations increase as the current propagates. We can mitigate
mathematical modeling of the viscous propagation of the current, these deviations by shifting the source location as the entire
when the released fluid mud volume from the source, V, becomes current volume enters the viscous region. Fig. 12b presents typical
equal to the initial lock volume, q0, the source location should be comparison of model predictions by shifting the source location
shifted to the respective front position at that particular time. and the experimental data. The time at which the source location is
A comparison of the force balance model predictions and the shifted (i.e., when V ¼q0) is indicated by an arrow in the graph. The
experimental observations corresponding to the viscous–buoyancy velocity Unn at the shifted position can now be determined from the
stage for current front positions in select experiments for clarity model prediction as the time required to travel the distance
purposes are presented in Fig. 12. The estimated m and n values between the new and previous source locations are known.
(see Table 1) are employed in these calculations. In Fig. 12a, the The best fit Kv values are calculated and tabulated in Table 4.
source is considered to be stationary at the viscous transition
position throughout the experiment. As is evident in Fig. 12a, the 5.2. Shallow water model
predictions of the force-balance expression with the stationary
Recently, Di Federico et al. (2006) provided a self-similarity
solution for a viscous one-layer shallow water model of a
450 non-Newtonian gravity current. In that solution, the rheological
properties of the intruding non-Newtonian fluid are represented by
the Ostwald power-law constitutive relationship. We provide a
300 brief solution of their elegant derivation of the front position as
xn (cm)

 n=ð2n þ 3Þ !ðaðn þ 2Þ þ nÞ=ðnð2n þ 3ÞÞ

n 1=2 rc guc q1=2
sw
xn ¼ Zn qsw t ðaðn þ 2Þ þ nÞ=ð2n þ 3Þ :
Exp. #7 2n þ 1 m
Exp. #9
150 Exp. #10 ð17Þ
Exp. #12 Here, ZN is the similarity variable at the nose of the current and it
Eq. 15
needs to be determined by numerically integrating Eqs. (15) and
0 (16) of Di Federico et al. (2006) for a a0 (the interested reader is
0 20 40 60 80 100 120 kindly referred to Di Federico et al., 2006), and qsw is a constant
t (s) related to volume released per unit width, V, expressed as
 a=n
ð2n þ aÞ=2n a rc guc
V ¼ qsw t : ð18Þ
m
400
Using Eqs. (12) and (18), one can manipulate Di Federico’s front
position expression (Eq. (17)) into the same functional form of the
300 force-balance expression in Eq. (15) as follows:
V=q0
xn (cm)

 n=ð2n þ 3Þ  1=ð2n þ 3Þ
n ð2 þ nÞ=ð2n þ 3Þ rc guc
200 xn ¼ Zn qfb t ðaðn þ 2Þ þ nÞ=ð2n þ 3Þ :
2n þ 1 m
ð19Þ
100 Exp. #10
Eq. 15 In Eq. (19), the term Zn ðn=ð2n þ1ÞÞn=ð2n þ 3Þ is constant for a
particular experiment since it depends only upon the value of n.
0
0 20 40 Hence, a comparison of Eqs. (15) and (19) shows that the only
t (s) difference between the force-balance and shallow water model
expressions are the proportionality constants: empirically deter-
Fig. 12. Comparison of the predictions by the viscous force-balance model (solid mined Kv (henceforth, empirical Kv) for the force-balance model
lines) with the experimental data (symbols, see the legend) for the front position of and the analytical expression Zn ðn=ð2n þ1ÞÞn=ð2n þ 3Þ (henceforth,
the fluid mud gravity currents in the viscous–buoyancy propagation stage. (a)
Predictions using a fixed source location (xn at tnn, shown with thick vertical solid
analytical Kv) for the shallow water model.
lines) for the entire experimental viscous propagation, (b) predictions for Exp. #10 Based on the same mathematical modeling considerations
using a shifting source location (shown in the graph) as V becomes equal to q0. described in Section 5.1, Eq. (19) is implemented to predict the

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]
450
300
xn (cm)
Exp. #7
150 Exp. #9
Exp. #10
Exp. #12
Eq. 19
0 only the slumping and self-similar phases. In a single experiment
0 20 40 60 80 100 120
t (s)
400
300
V= q 0
xn (cm)
200
100 Exp. #10
Eq. 19
0
0 20 40
t (s)
Fig. 13. Comparison of the predictions by the viscous compositional shallow water
model (solid lines) with the experimental data (symbols, see the legend) for the front
position of the fluid mud gravity currents in the viscous–buoyancy propagation stage.
(a) Predictions using a fixed source location (xn at tnn, shown with thick vertical solid
lines) for the entire experimental viscous propagation, (b) predictions for Exp. #10
using a shifting source location (shown in the graph) as V becomes equal to q0.
viscous spreading of fluid mud gravity current. Since the problem is
now modeled as a ¼1, Zn is determined by solving Eq. (15) of
Di Federico et al. (2006) by fourth-order Runge–Kutta method for a
given experimental condition. The analytical Kv values (tabulated
in Table 4) are very close to the empirical Kv values in the force-
balance expression. Shallow water model predictions, which are in
good agreement with the experimental data are presented in
Fig. 13. Comparing force-balance and shallow water model pre-
dictions, the R2 values are very similar, with slightly better values
for the force-balance model predictions. This is simply due to the
fact that force-balance expression has a free parameter which is
fitted using the experimental data (i.e., empirical Kv) while the
shallow water model lacks this free parameter. It is important to
note that although both force-balance and shallow water models
are intended for compositional gravity currents, their predictions
are in good agreement with the experimental data for the fluid mud
gravity currents.
6. Discussions and conclusions
In this study, propagation characteristics of fluid mud suspen-
sion gravity currents were investigated experimentally and
theoretically. Fluid mud suspension currents with sufficiently high
concentration values demonstrate non-Newtonian rheology. Given
the rheological differences, their propagation characteristics
are vastly different from the propagation characteristics of the
Newtonian suspension (e.g., non-cohesive suspension currents)
and compositional currents (e.g., saline currents) that have been
Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
13
studied in detail. To elucidate the propagation characteristics of
fluid mud suspension gravity currents, laboratory experiments of
fluid mud gravity current propagation were conducted in a lock-
exchange set-up with different initial fluid mud suspension
concentrations. Experimental observations were compared with
the predictions of available mathematical models with different
levels of complexities to assess the model predictive capabilities
and the simplifying assumptions embedded in the models.
As the experimental fluid mud currents propagate, some
currents exhibited all slumping, self-similar, and viscous phases.
Conversely, other currents, lacking the propagation distance in the
experimental tank to transition into the viscous phase, exhibited
(Exp. #12) we observed that the current did not experience the self-
similar phase, directly transitioning from the slumping phase into
the viscous phase. Here, with the same volume released for both, a
fluid mud gravity current with higher concentration transitioned
into the viscous phase earlier than a fluid mud current with a lower
concentration. If the concentration of the fluid mud suspension is
such that it yields to tnn otn, a bypass of the self-similarity phase
may be expected. In each phase, propagation characteristics of the
current differ and should be modeled using a relevant model for the
particular phase. Therefore, phase transition times/propagation
distances were identified to determine both the propagation phase
of the current and the corresponding mathematical model to use.
We obtained a new expression for the phase transition time, tn,
for the slumping – self-similar phase transition (see Eq. (3)) and
modified the phase transition time, tnn, expression for the self-
similar – viscous phase transition (see Eq. (4)) proposed by
Di Federico et al. (2006) based on our experimental observations.
Predictions of these expressions showed good agreement with our
observations.
Three different modeling approaches were considered, from the
simplest to the most complex: force-balance, box, and shallow
water models. In the inertia–buoyancy stage (i.e., slumping and
self-similar phases), inertia and buoyancy forces are the governing
forces acting on the current and the viscous forces can be neglected.
Therefore, the rheological differences in the non-Newtonian and
Newtonian suspensions play a minimal role in the current propa-
gation, yielding similar propagation characteristics for the two in
the inertia–buoyancy stage. Five different mathematical models
proposed for the propagation of Newtonian compositional and
suspension currents were tested for the propagation of the fluid
mud suspension currents using our laboratory experimental
observations for the inertia–buoyancy stage. Comparisons of the
model predictions with the experimental data and inter-compar-
isons of the models showed that the compositional and suspension
box models and the force-balance model had superior predictive
capabilities than the mathematically more advanced composi-
tional and suspension shallow water models. The force-balance
model and compositional and suspension box models had approxi-
mately similar predictive capabilities (see Table 4). Of the three
models, the simplicity of the force-balance model calculations and
the absence of empirical adjustments in the compositional box
model calculations make them particularly good choices for initial
calculations. The predictions of one-layer compositional shallow
water model significantly overestimate the experimental observa-
tions when deep ambient Fr condition is used. However, the
predictions radically improved with the use of a representative
Fr condition for the shallow ambient in our experiments.
In the viscous–buoyancy stage, rheological differences between
Newtonian and non-Newtonian suspensions play a major role, and
viscous mathematical models for non-Newtonian suspensions are
the ones to use for predicting the propagation characteristics.
However, to our knowledge, only analytical solution of composi-
tional shallow water model (Di Federico et al., 2006) is available for
14 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]]
modeling viscous propagation characteristics of non-Newtonian
gravity currents in the considered mathematical modeling
approaches. In this paper, we derived and presented a force-
balance expression for the propagation of a non-Newtonian
power-law fluid and will present a non-Newtonian compositional
viscous box model in a separate paper. The calculations of viscous
model predictions for a current initially propagating in the inertia–
buoyancy stage and transitioning into the viscous–buoyancy stage
are not straightforward and additional upstream flow boundary
considerations/assumptions are developed. Compared with our
experimental observations, the predictions of both viscous force-
balance and compositional shallow water models showed that the
predictions of both models are in notably good agreement with the
experimental data. Inter-model comparisons indicate that predic-
tive accuracy of each model is similar. However, the great
simplicity of force-balance model calculations over compositional
shallow-water model calculations makes the force-balance model
a good choice for swift calculations.
Another important finding of this study is that fluid mud
suspension currents can be modeled as compositional gravity
currents for shorter propagation distances/times. Very low settling
velocities of the clay particles and the tendency of the fluid mud
suspensions to remain in a fluid-like state due to particle cohe-
siveness permit this compositional gravity current modeling. For
example, a clay particle in a gravity current of 10 cm thickness
takes tens of hours to settle. Note that our experiments lasted less
than 1–2 min. Therefore, models for compositional gravity currents
can be used for modeling fluid mud suspension currents for short
propagation distances/times. The research results of this study are
expected to be of use for preliminary swift calculations of the fluid
mud suspension current propagation characteristics and to decide
whether more detailed calculations (e.g., CFD based numerical
simulations or two-layer shallow water models) at the expense of
complexity are required.
Acknowledgments
This research was supported by the funds provided by the USACE
Grant W912HZ-09-C-0068 to the second author. The first author is a
graduate student under the guidance of the second author.
References
Benjamin, T., 1968. Gravity currents and related phenomena. Journal of Fluid
Mechanics 31, 209–248.
Blanchette, F., Piche, V., Meiburg, E., Strauss, M., 2006. Evaluation of a simplified
approach for simulating gravity currents over slopes of varying angles.
Computers and Fluids 35, 492–500.
Bonnecaze, R., Huppert, H., Lister, J., 1993. Particle-driven gravity currents. Journal of
Fluid Mechanics 250, 339.
Bowen, A., Normark, W., Piper, D., 1984. Modelling of turbidity currents on Navy
submarine fan, California continental borderland. Sedimentology 31, 169–185.
Cantero, M., Balachandar, S., Garcia, M., 2007. High-resolution simulations of
cylindrical density currents. Journal of Fluid Mechanics 590, 437–469.
Chowdhury, M.R., Testik, F.Y., Khan, A.A., 2009. Three-dimensional flow structure at
the frontal zone of a gravity-driven fluid mud flow. Journal of the Visualization
12 (4), 287.
Coussot, P., 1997. Mudflow Rheology and Dynamics. International Association for
Hydraulic Research Monograph Series, Balkema, Rotterdam, p. 255.
Dade, W., Huppert, H., 1995. A box model for non-entraining, suspension-driven
gravity surges on horizontal surfaces. Sedimentology 42, 453–472.
Didden, N., Maxworthy, T., 1982. The viscous spreading of plane and axisymmetric
gravity currents. Journal of Fluid Mechanics 121, 27–42.
Di Federico, V., Malavasi, S., Cintoli, S., 2006. Viscous spreading of non-Newtonian
gravity currents on a plane. Meccanica 41, 207–217.
Drago, M., 2002. A coupled debris flow-turbidity current model. Ocean Engineering
29, 1769–1780.
Fleischmann, C.M., Pagni, P.J., Williamson, R.B., 1994. Salt water modeling of fire
compartment gravity currents. Fire Safety Science. In: Proceedings of the Fourth
International Symposium, Ottawa, June 13–17. International Association for
Fire Safety Science, Gaithersburg, pp. 253–264.
Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
Gladstone, C., Phillips, J., Sparks, R., 1998. Experiments on bidisperse, constant-
volume gravity currents: propagation and sediment deposition. Sedimentology
45, 833–844.
Gladstone, C., Woods, A., 2000. On the application of box models to particle-driven
gravity currents. Journal of Fluid Mechanics 416, 187–195.
Haertel, C., Meiburg, E., Necker, F., 2000. Analysis and direct numerical simulation of
the flow at a gravity–current head. Part 1. Flow topology and front speed for slip
and no-slip boundaries. Journal of Fluid Mechanics 418, 189–212.
Harris, T., Hogg, A., Huppert, H., 2001. A mathematical framework for the analysis
of particle-driven gravity currents. Proceedings of the Royal Society of London –
A – Mathematical, Physical and Engineering Sciences 457, 1241–1272.
Hogg, A., Ungarish, M., Huppert, H., 2000. Particle-driven gravity currents:
asymptotic and box model solutions. European Journal of Mechanics B – Fluids 19,
139–165.
Hosseini, S., Shamsai, A., Ataie-Ashtiani, B., 2006. Synchronous measurements of the
velocity and concentration in low density turbidity currents using an Acoustic
Doppler Velocimeter. Flow Measurement and Instrumentation 17, 59–68.
Huang, X., Garcia, M., 1998. A Herschel–Bulkley model for mud flow down a slope.
Journal of Fluid Mechanics 374, 305–333.
Huppert, H.E., Simpson, J.E., 1980. The slumping of gravity currents. Journal of Fluid
Mechanics 99, 785–799.
Huppert, H.E., 1982. The propagation of two-dimensional and axisymmetric
viscous gravity currents over a rigid horizontal surface. Journal of Fluid
Mechanics 121, 43–58.
Huppert, H., 1998. Quantitative modelling of granular suspension flows. Philoso-
phical Transactions: Mathematical, Physical and Engineering Sciences 356,
2471–2496.
Huppert, H., 2006. Gravity currents: a personal perspective. Journal of Fluid
Mechanics 554, 299–322.
Hurzeler, B., Imberger, J., Ivey, G., 1996. Dynamics of turbidity current with reversing
buoyancy. ASCE Journal of Hydraulic Engineering 122, 230–236.
Jiang, L., Leblond, P., 1993. Numerical modelling of an underwater Bingham plastic
mudslide and the waves which it generates. Journal of Geophysical Research
98, 10303–10317.
Kirwan, A., Doyle, L., Bowles, W., Brooks, G., 1986. Time-dependent hydrodynamic
models of turbidity currents analyzed with data from the Grand Banks and
Orleansville events. Journal of Sedimentary Petrology 56, 379–386.
Kneller, B., Buckee, C., 2000. The structure and fluid mechanics of turbidity currents:
a review of some recent studies and their geological implications. Sedimentology
47, 62–94.
Kuenen, P., 1952. Estimated size of the Grand Banks turbidity current. American
Journal of Science 250, 874–884.
Lowe, D., 1982. Sediment-gravity flows, II: depositional models with special
reference to the deposits of high-density turbidity current. Journal of Sedimen-
tary Petrology 52, 279–297.
Maxworthy, T., 1983. Gravity currents with variable inflow. Journal of Fluid
Mechanics 128, 247–257.
Meiburg, E., Kneller, B., 2010. Turbidity currents and their deposits. Annual Review
of Fluid Mechanics 42, 135–156.
Middleton, G., 1993. Sediment deposition from turbidity currents. Annual Review of
Earth and Planetary Sciences 21, 89–114.
Mohrig, D., Marr, J., 2003. Constraining the efficiency of turbidity current generation
from submarine debris flows and slides using laboratory experiments. Marine
and Petroleum Geology 20, 883–899.
Mulder, T., Syvitski, J., Skene, K., 1998. Modeling of erosion and deposition by
turbidity currents generated at river mouths. Journal of Sedimentary Research
68, 124–137.
Neal, R., Henry, G., Greene, S., 1978. Evaluation of the submerged discharge
of dredged material slurry during pipeline dredge operations. TR D-78-44,
US Army Engineers Waterways Experiment Station, Vicksburg, Mississippi,
176pp.
Necker, F., Haertel, C., Kleiser, L., Meiburg, E., 2002. High-resolution simulations of
particle-driven gravity currents. International Journal of Multiphase Flow
28, 279–300.
Ng, C., Mei, C., 1994. Roll waves on a shallow layer of mud modelled as a power-law
fluid. Journal of Fluid Mechanics 263, 151–183.
Nichols, M., Thompson, G., 1978. A field study of fluid mud dredged material: its
physical nature and dispersal. Technical Report D-78-40, US Army Engineers
Waterways Experiment Station, Vicksburg, Mississippi, p. 91.
Rocca, M.L., Adduce, C., Sciortino, G., Pinzon, A., 2008. Experimental and numerical
simulation of three-dimensional gravity currents on smooth and rough bottom.
Physics of Fluids 20, 106603.
Rottman, J., Simpson, J., 1983. Gravity currents produced by instantaneous releases
of a heavy fluid in a rectangular channel. Journal of Fluid Mechanics 135,
95–110.
Seymour, R.J., 1986. Nearshore auto-suspending turbidity flows. Ocean Engineering
13 (5), 435–447.
Simpson, J., 1972. Effects of the lower boundary on the head of a gravity current.
Journal of Fluid Mechanics 53, 759–768.
Simpson, J., 1997. Gravity Currents: in the Environment and the Laboratory.
Cambridge University Press, p. 244.
Simpson, J., Britter, R., 1979. The dynamics of the head of a gravity current advancing
over a horizontal surface. Journal of Fluid Mechanics 1994, 477–495.
Teeter, A., 2002. Sediment dispersion near dredge pipeline discharge in Laguna
Madre, Texas. Technical Note DOER-N-16, US Army Engineer Waterways
Experiment Station, Vicksburg, MS.
 M.R. Chowdhury, F.Y. Testik / Ocean Engineering ] (]]]]) ]]]–]]] 15

Teeter, A., 2000. Underflow spreading from an open-water pipeline disposal. Technical
Note TN-DOER-N7, US Army Engineer Waterways Experiment Station, Vicksburg, MS.
Ungarish, M., 2009. An Introduction to Gravity Currents and Intrustions. Chapman &
Hall/CRC Press, Boca Raton, FL, p. 489.
Von Karman, T., 1940. The engineer grapples with nonlinear problems. Bulletin of
the American Mathematical Society 46, 615–683.
Whitehouse, R., Soulsby, R., Roberts, W., Mitchener, H., 2000. Dynamics of
Estuarine Muds: A Manual for Practical Applications. Thomas Telford, London,
p. 209.
Winterwerp, J., Kesteren, W.V., 2004. Introduction to the Physics of Cohesive
Sediment in the Marine Environment. Elsevier, Amsterdam, p. 466.

Please cite this article as: Chowdhury, M.R., Testik, F.Y., Laboratory testing of mathematical models for high-concentration fluid mud
turbidity currents. Ocean Engineering (2010), doi:10.1016/j.oceaneng.2010.10.020
View publication stats