ARTICLE IN PRESS

Computers & Geosciences 34 (2008) 1319–1337

www.elsevier.com/locate/cageo

Sedflux 2.0: An advanced process-response model that generates

three-dimensional stratigraphy$
Eric W.H. Hutton, James P.M. Syvitski
Institute of Arctic and Alpine Research, University of Colorado at Boulder, 1560 20th Street, Campus Box 450, Boulder,
CO 80309-0450, USA
Received 8 September 2006

Abstract

Sedflux 2.0 is the newest version of the Sedflux basin-filling model. Sedflux 2.0 provides a framework within which
individual process-response models of disparate time and space resolutions communicate with one another to deliver
multigrain-sized sediment load across a continental margin. Version 2.0 introduces a series of new process models, and the
ability to operate in one of two modes to track the evolution of stratigraphy in either two or three dimensions. Additions to
the 2D mode include the addition of models that simulate (1) erosion and deposition of sediment along a riverbed,
(2) cross-shore transport due to ocean waves, and (3) turbidity currents and hyperpycnal flows. New processes in the 3D
mode include (1) river channel avulsion, (2) two-dimensional diffusion due to ocean storms, and (3) two-dimensional
flexure due to sediment loading. The spatial resolution of the architecture is typically 1–25 cm in the vertical and 10–100 m
in the horizontal when operating in 2D mode. In 3D mode, the horizontal resolution usually extends to kilometers. In
addition to fixed time steps (from days to hundreds of years), Sedflux 2.0 offers event-based time stepping as a way to
conduct long-term simulations while still modeling low-frequency but high-energy events.
r 2008 Elsevier Ltd. All rights reserved.

Keywords: Plumes; Turbidity currents; Flexure; Boundary-layer transport; Stratigraphy

1. Introduction and Tetzlaff and Priddy (2001)). The first set

The development of numerical models that
explore the evolution of continental margins is
comparatively recent. The modeling community has
developed three main types of models to simulate
the growth of continental margins (for a full
overview, see Paola (2000), Overeem et al. (2005),
$
Code available from server at http://www.iamg.org/
CGEditor/index.htm.
Corresponding author. Fax: +1 303 492 6288.
E-mail address: eric.hutton@colorado.edu (E.W.H. Hutton).
consists of individual process models intended to
model the effects of a single process on (usually) a
portion of a margin. For instance, Bonham-Carter
and Sutherland (1967) and later Morehead et al.
(2001) used river plume models to simulate the
progradation of a river delta. Harris and Wiberg
(2001) developed a detailed model to simulate the
transport of shelf sediments by wave and current
interactions (also, Wiberg and Smith, 1983; Li et al.,
1997; Reed et al., 1999). The modeling of con-
tinental slope processes has received somewhat less
attention. Imran et al. (2001) developed a numerical

0098-3004/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cageo.2008.02.013
 ARTICLE IN PRESS
1320 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337

Nomenclature L wave length (L)

m Bruun constant (dimensionless)
Symbol Description (units) q lithospheric point load (M)
/qS width-averaged long-term average water qs sediment flux (L3 T1)
discharge (L2 T1) r distance to load (L)
A river-type constant (dimensionless) s specific gravity of suspended sediment
a depth gradient of bottom-sediment shear (dimensionless)
strength S seabed gradient (dimensionless)
b shear strength of sediments on the sea- Y angular position of the river mouth
floor (dimensionless)
C volume concentration of sediment t time (T)
(dimensionless) T wave period (T)
C0 streambed sediment concentration t0 time when load is applied (T)
(dimensionless) u flow velocity (M T1)
Cd drag coefficient (.004) (dimensionless) u* flow shear velocity
cf stream drag coefficient (dimensionless) Uom near-bed wave velocity (L T2)
Cfs seafloor drag coefficient (dimensionless) ut threshold velocity for sediment transport
d grain size (L) (L T1)
D flexural rigidity of Earth’s crust w lithospheric deflection (L)
d0 near-bed wave orbital diameter (L) w0 isostatic deflection (L)
Ess efficiency of suspended sediment trans- ws sediment settling velocity (L T1)
port (dimensionless) x offshore distance normalized by near-
Ew water entrainment coefficient shore boundary position (dimensionless)
(dimensionless) X distribution of channel avulsion angles
Fd depositional sediment flux of flow (dimensionless)
(L T1) X horizontal position (L)
Fe erosion sediment flux of flow (L T1) Y horizontal position (L)
g acceleration due to gravity (L T2) Z seafloor elevation (L)
h water depth (L) a drag force ratio (dimensionless)
H wave height (L) b grain-size transport index (dimension-
hf flow thickness (L) less)
hb water depth of breaking waves (L) l flexural parameter (L)
hc wave closure depth (L) gb breaking wave index (dimensionless)
Hss height of significant storm wave for a Z bed elevation (L)
year (L) n diffusion coefficient for stream erosion
Is time fraction of ocean storms (dimen- (L2 T1)
sionless) n0 cross-shore current velocity (L T1)
k VON Karman constant (dimensionless) r fluid density (M L3)
k diffusion coefficient (L2 T1) rd density of the asthenosphere (M L3)
kc diffusion coefficient at near-shore bound- rs density of sediment (M L3)
ary (L2 T) rw density of water (M L3)

model to simulate the movement (but not failure) of
sediment down a continental slope as a debris flow.
Again, while not modeling the inception of turbidity
currents from sediment failure, Parker et al. (1986)
modeled the transport of sediment through turbid-
ity currents. This work then led to the modeling of
submarine fan formation through turbidity currents
(Imran et al., 1998).
Modeling of the entire continental margin has
received less attention, and has taken two routes.
The first route develops a small set of equations that
governs the evolution of the entire continental
margin (Ross et al., 1994; Steckler, 1999; Swenson
et al., 2000; Granjeon and Joseph, 1999). These
models do not attempt to simulate individual
processes that act to form stratigraphy. Instead,
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
they model the long-term result of complex inter-
actions between physical processes. The second
route links individual process models into a
component sequence-stratigraphic model (Martinez
and Harbaugh, 1993; Syvitski and Alcott, 1995;
Syvitski and Hutton, 2001; Ritchie et al., 1999). In
this way, the major processes that form stratigraphy
are able to interact with one another. Because of the
complexity of the component models, this route is
able to provide a detailed prediction of the evolving
stratigraphy but at a cost of increased computation
time and program complexity.
One such component sequence-stratigraphic
model is Sedflux 1.0c (Syvitski and Hutton, 2001).
It is a two-dimensional basin-filling model that
generates stratigraphy that varies vertically, and in a
single lateral dimension. The structure of Sedflux
1.0c provides an architecture within which a
collection of individual process models are able to
interact with one another, and with the sediment
deposits they produce. Not only does this model
produce stratigraphy, but it also tracks the geo-
technical properties of the deposits (grain size, bulk
density, porosity, etc.; Hutton and Syvitski, 2003).
Some of the process modules that contribute to
this stratigraphy are modeled using vertically
averaged variables that only vary horizontally. For
instance, the turbidity current model assumes
vertically averaged flow characteristics but allows
them to vary in the horizontal dimension. Other
processes, such as sediment plumes, are averaged
over a specified basin width. This width is user-
defined and is constant in time but can vary in
space. The suite of processes that Sedflux 1.0c
modeled consisted of (1) river-mouth dynamics,
(2) buoyant surface plumes, (3) hyperpycnal plumes,
(4) ocean storms, (5) slope instability, (6) turbidity
currents, (7) debris flows, (8) compaction, and
(9) flexure of the lithosphere.
Although a two-dimensional basin-filling model
may provide insights into how stratigraphy forms, it
contains a basic limitation in that it is two-
dimensional. In reality, sediment spreads laterally
rather than simply within a plane. Certainly, there
are situations where a two-dimensional model is
adequate, but, in general, this is not the case and so
a three-dimensional model is necessary.
Often one will use a two-dimensional model and
assume that sediment moves along some profile.
The profile need not be a straight line, only that
sediment transport follows this line. Some two-
dimensional models will introduce a width to
1321
approximate a third dimension (oftentimes, referred
to as 1.5-dimension models). For example, Skene
et al. (1997) adapt a one-dimensional model to
simulate a hyperpycnal flow that follows a channel
of non-constant width. In such cases though, the
deposit parameters only vary in a single dimension,
and the flow direction is also in a single dimension.
In reality, the geometry of a location may be such
that there is no single direction of flow and so
1.5-dimension models are not adequate. For in-
stance, ocean currents may be such that a surface
plume flows along bathymetric contours and so
deposits sediment along these contours. Once on the
ocean floor, gravity may become the major driver of
sediment transport rather than ocean currents, in
moving down-slope, perpendicular to bathymetric
contours.
Two-dimensional basin-filling models are often
based upon processes that are assumed to be one-
dimensional (for example, Steckler, 1999; Swenson
et al., 2000; Syvitski and Hutton, 2001). That is, the
model assumes that the variables fluctuate only in
one direction (the direction of transport). Again,
this may be a reasonable assumption in some cases
(the unidirectional flow of sediment down a narrow
submarine canyon, for instance) but in general it is
not. In fact, many transport processes are more
complicated and a two-dimensional transport model
will give a better approximation. A sediment gravity
flow, for instance, may create a mound-like
structure around which subsequent flows will have
to travel (McAdoo et al., 2000). Thus, although the
physics may be accurate, addition of more potential
flow directions gives rise to different results. For
instance, a two-dimensional basin-fill model may
produce an aggradational basin-floor fan because
successive flows attempt to run over previous flow
deposits (Skene et al., 1997; Pratson et al., 2001).
A three-dimensional basin-fill model can potentially
produce a basin-floor fan that builds laterally as
successive flows are allowed to move around
previous flow deposits (Bouma, 2000, 2001;
Richards et al., 1998). In this case, it is not the
physics of the higher-dimensional process that gives
rise to the different structure, rather it is the change
in flow direction (and, in fact, simply geometry).
Since the publication of the Sedflux 1.0c model
(Syvitski and Hutton, 2001), Sedflux has advanced to
address this issue. The version of Sedflux described
here (Sedflux 2.0) provides a modeling framework
that is able to produce and track the development of
three-dimensional structures. This three-dimensional
 ARTICLE IN PRESS
1322 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
basin-fill model is able to generate simulations that
can be compared with the three-dimensional data
that are now common within oil exploration. In
addition, it is able to produce output that can be
compared to two-dimensional data that are still
commonplace or even be run in a 2D mode. A
seismic section may lie along the believed direction of
flow, but in reality deposition may have occurred at
angles that are oblique to this direction. A three-
dimensional basin-fill model can provide insights into
how a three-dimensional feature is represented in a
two-dimensional section. Furthermore, a three-
dimensional seismic model coupled with a three-
dimensional basin-filling model will demonstrate
how a seismic line will portray these features.
The three-dimensional basin-filling model Sedflux
2.0 is an extension of the two-dimensional basin-
filling model Sedflux 1.0c (Syvitski and Hutton,
2001). As with the previous versions of Sedflux, the
basis of Sedflux 2.0 is an architecture that allows
various process modules to communicate with one
another to create either two- or three-dimensional
stratigraphy (depending upon which mode the user
chooses to run the model in). This architecture
allows for the straightforward addition of new
modules. New modules can be added either by
writing a ‘‘wrapper’’ that converts the new model’s
framework to that of Sedflux or, more simply, the
new model can be written using the application
programming interface (API) of Sedflux.
The process models present in the previous
Sedflux release are still present within the current
version. Although the principal change between the
two versions is the ability to track three-dimensional
architecture, a number of new processes were added
to the 2D mode. In addition, the 3D mode of
Sedflux 2.0 required addition of new modules that
were not necessary in two dimensions.
2. Development history
Since the publication of Syvitski and Hutton
(2001), Sedflux has advanced along three major
paths. The first is the addition of process models
that operate solely within the 2D mode of Sedflux.
The second is the expansion of the Sedflux
architecture to be able to track stratigraphy in three
dimensions and the addition of processes necessary
to do this. The third are additions that operate in
both the 2D and 3D mode. We describe only new or
extensively modified modules.
Sedflux 2.0 is a three-dimensional basin-filling
model that is able to predict stratigraphy that varies
laterally and vertically. The 3D mode of Sedflux
records sediment deposition in two lateral dimen-
sions while the 2D mode tracks sediment only in the
cross-shore direction. The individual processes that
operate within either mode of the Sedflux frame-
work are either one- or two-dimensional. These
processes produce sediment deposits that vary
laterally (in either one or two dimensions) but use
vertically averaged variables that only vary hor-
izontally. The absence of three-dimensional models
within Sedflux is not due to a limitation of its
architecture but rather because these models are not
yet incorporated.
Prior to a Sedflux simulation, the user specifies
the spatial and temporal resolution of the architec-
ture. The user defines the basin geometry at points
on an initial surface (or line, in the case of the 2D
mode) at a fixed resolution. Typically, this resolu-
tion is on the order of kilometers for 3D mode, and
tens of meters for 2D mode. The vertical resolution
defines the thickness over which sediment charac-
teristics are averaged (this ranges from centimeters
to meters). These resolutions only constrain the size
of the cells that Sedflux uses to track sediment
deposits. Each component module uses its own
resolutions, which are independent of the architec-
ture resolutions. For instance, Sedflux can run with
a daily time step, but because of stability con-
straints, the debris flow module typically runs with
time steps less than 1 s.
A series of rivers supply the basin with sediment
(the 2D mode is limited to a single river). Input files
specify the number of rivers, their location, and
hydrologic characteristics. These rivers transport
sediment composed of a user-defined number of
distinct grain types. Each river introduces its
suspended sediment through a surface plume or a
hyperpycnal flow, depending on the sediment-laden
density of the river water. The bedload component
of the river’s load is spread over a user-specified
distance.
3. Additions to Sedflux 3D
The transition of Sedflux to three dimensions
requires the addition of a channel avulsion model.
Because of the two-dimensional nature of Sedflux
1.0c, the river mouth was fixed to the land–ocean
interface along a profile. In three dimensions, the
river mouth can move laterally. The flexure and
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
diffusion modules were also improved to work
within the three-dimensional framework. Both
modules now solve their corresponding two-dimen-
sional governing equations.
3.1. Channel avulsion
River deltas distribute sediment laterally by
frequently changing the position of the river mouth
and the main distributary. Existing delta models
incorporate partially understood and complex
physics in an attempt to simulate realistically
the channel avulsions (Overeem et al., 2003;
Sun et al., 2002). However, the angular location of
the main distributary is more simply derived as a
stochastic Brownian process and has a Gaussian
distribution over short time scales, resulting in
localized deposition and lobe-like formations. Over
long time scales, the same process is uniform in
distribution, resulting in sediment spread over the
entire delta plain.
Using the distribution X for the change in an
angle after an avulsion (Dy), the angular position
After Avulsion
Θn+1
Xn
Sea Θn
Land
Hinge Point Before Avulsion
Fig. 1. Definition of terms for the channel avulsion model.
Fig. 2. Two delta simulations with initially straight coastlines and hinge points located halfway down the left side of each plot. Left plot
uses uniformly distributed channel angle jumps. Right plot uses normally distributed channel angle jumps. Colors represent deposit age
with blue being the oldest and red the newest.
1323
of the distributary channel after n+1 avulsions
is given by
Ynþ1 ¼ Yn þ X n (1)
where Yn is the current angle and Xn is the nth
realization of X (Fig. 1). In other words, the angular
position is just the sum of angular jumps that
are generated from the distribution X. Regardless of
the underlying physics, some probability distribu-
tion must represent this change in angle. The
precise distribution will not be known, but observa-
tions of large deltas suggest that the probability of
avulsing somewhere nearby is high, while the
probability of larger avulsions is low (Milliman
et al., 1987).
A simple numerical model illustrates the realistic-
looking deltas generated by such a stochastic
process (Fig. 2). This model assumes that an
avulsion happens every time step, the basin is flat-
bottomed, and the grid scale is such that one cell is
always filled by the river’s sediment with every time
step. The model randomly generates angles from the
distribution X, moves the mouth of the distributary
by these angles around the coastline, and fills empty
cells with sediment. The shoreline provides a
reflection boundary condition at Y ¼ 7901. Both
diagrams show a delta building into the sea. The left
figure picks Xn from a uniform distribution, and the
right figure uses a normal distribution (with a mean
of 0 and standard deviation of .051). Colors indicate
the time of deposition (cool colors are oldest, hot
colors are recent). A uniform distribution builds a
symmetric and radial delta while the normal
distribution creates a more lobe-like delta. These
 ARTICLE IN PRESS
1324 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
river-dominated delta morphologies would change
with the inclusion of waves, tides, and other
processes (Orton and Reading, 1993).
3.2. River plumes
The momentum-driven hypopycnal plume de-
scribed in Syvitski and Hutton (2001) is still the
basis for the river plume used in the 3D mode of
Sedflux 2.0. However, unlike implementations used
in previous versions of Sedflux, the plume deposit is
no longer averaged in the alongshore direction.
Instead, Sedflux 2.0 tracks the deposit due to a two-
dimensional surface jet that emanates from each
river mouth. In addition, the plume module
considers a set of open-coast plume scenarios able
to handle upwelling and downwelling against strong
and weak coastal currents.
When the plume enters the ocean and is met with
an alongshore current, it deflects in the direction of
the current. Chassaing et al. (1974) describes the
deflection of a jet’s centerline upon encountering a
crossflow. Modified for Coriolis force, the plume’s
centerline is
  0:37
x u0 y
¼ 1:53 þ 0:9 (2)
b0 v0 b0
where v0 is the ambient coastal current velocity, and
b0 and u0 are river-mouth width and velocity,
respectively. The model solves the same governing
equations described in Syvitski et al. (1998) but with
the coordinate system warped to reflect the curved
centerline. What was the cross-shore direction in the
original coordinate system becomes distance along
the plume’s (curved) centerline; alongshore distance
becomes distance perpendicular to the plume’s
centerline.
3.3. Bedload dumping
In the 3D mode of Sedflux 2.0, all of a river’s
suspended sediment is delivered to the ocean as a
hypopycnal plume. The river delivers the remaining
sediment as bedload in much the same way as that
of the 2D mode (Syvitski and Hutton, 2001).
Instead of being distributed evenly over a specified
distance, the river now spreads the sediment evenly
over a cone with a user-specified angle and radius.
This radius is meant to reflect the transport by both
river momentum and tides.
3.4. Two-dimensional diffusion of seafloor sediments
Sedflux1.0c modeled the resuspension and trans-
port of bottom sediments due to a host of effects as
a single diffusive process. Sedflux 2.0 extends this
model to include two-dimensional diffusion.
We assume that the amount of bottom sediments
that can be reworked by these processes (qs) is
proportional to bathymetric slope and a diffusion
coefficient, k:
 
qz ^ qz ^
qs ¼ kðt; z; DÞrz ¼ k iþ j (3)
qx qy
where the diffusion coefficient, k, is a function of
time (t), water depth (z), and grain size (D).
Although the model does not assume a specific
depth (or z) dependence of k, typically we assume an
exponential decline to reflect decreasing wave
energy with depth. The user specifies the rate at
which diffusion decreases with depth, and the
current ocean storm conditions scale the value of
k at the water surface.
This flux is interpreted as the maximum volume
of sediment that can be reworked under the local
conditions; not all of this sediment will be moved
down-slope however. For each grain size, the above
flux is scaled by a user-defined index (bi) between 0
and 1 that reflects the ability of this process to move
the ith grain size. Thus, the amount and direction of
transport of the ith grain size is
qsi ¼ bi qs (4)
This flux is drawn only from the resuspended
sediment (qs). Any sediment that may be left behind
(due to bio1) is mixed and redeposited.
Although geomorphological studies have mea-
sured diffusion coefficients, they do not directly
relate to the above k. Instead, they often represent
diffusion values integrated over entire margins,
or over long time scales. For spatial scales of
10–100 km and time scales of 105–106 years,
diffusion coefficients range from 102 to 103 m2 yr1.
3.5. Flexure of the lithosphere
Within the Sedflux 2.0 framework, a suite of
processes deposit and erode sediment causing the
load distribution over the region to evolve over the
simulation. Thus, the lithospheric load changes as
the model evolves. Depending upon how the load
distribution develops, this flexure can result in the
basin uplifting or subsiding (or both). The pattern
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
of subsidence in time and space largely determines
the gross geometry of time-bounded units because it
controls the rate at which space is created for
sedimentation (Angevine et al., 1990).
It is typical to model the deflection of Earth’s
crust due to loading by assuming that it behaves as a
beam (or a plate, in the case of two dimensions) on
an elastic foundation (for example, Lambeck, 1988;
Anderson, 1994; Syvitski and Hutton, 2001).
Lambeck (1988) solves this problem in two dimen-
sions. The solution shows the deflection of Earth’s
crust (w) to be a function of the non-dimensional
length lr as
ql
wðlrÞ ¼ KeiðlrÞ
2prd g
where Kei is the Kelvin function, q is the point load,
rd is the density of the asthenosphere, l is the
flexural parameter, and r is the distance from the
point load. We define the flexural parameter as
 1=4
D
l¼
rd g
where D is the flexural rigidity of Earth’s crust. The
Kelvin function behaves as an exponentially
damped sinusoid, and the flexural parameter acts
to control the width of the deflection. To solve the
Kei function numerically, we use the algorithm
developed by Amos (1986).
Eq. (5) predicts the deflection due to a point load;
it does not account for the additional weight of
material that fills the deflection. For instance,
a rise in sea level will cause a deflection that will
be filled with additional water, which in turn will
cause further deflection. If we know the density of
material that fills the deflection, the total deflection
is increased by a factor of ððrd Þ=ðrd  rw ÞÞ1:25
(Angevine et al., 1990). Using rd ¼ 3300 kg m3,
and rw ¼ 1030 kg m3, this amounts to an increase
in deflection by a factor of 1.6. Unfortunately, in
general, we do not know the density of the added
(or removed) material as it could be either air or
water. Thus, we solve for the total deflection by
iteratively solving Eq. (5) and updating the
load after successive iterations until the solution
converges.
Because the viscous asthenosphere has to flow out
of the way before the lithosphere can deflect,
there will be a time delay between the addition
of load and the lithosphere’s response. Eq. (7)
expresses this time delay as an exponential
1325
(Peltier, 1998):
wðtÞ ¼ w0 ð1  expðt=t0 ÞÞ (7)
where w0 is the equilibrium deflection as determined
by Eq. (5), t is the time since the load was applied,
and t0 is the response time associated with mantle
viscosity. In reality, response time is a function of
the viscosity of the underlying mantle (Peltier, 1998)
Typical relaxation times vary from about 1500 to
5000 years (Peltier, 1998; Huybrechts, 2002; Paulson
et al., 2005).
As an example of the flexure model, we estimate
the deflection of the lithosphere due to the weight
of the modern (last 5500 years) Mississippi delta
(Fig. 3; New Orleans is indicated by the vertical red
(5)
line). For this experiment, we estimated the sedi-
ment distribution and the timing of lobe switches
from the work of Fisk and McFarlan (1955). The
simulation began at the Last Glacial Maximum with
rising sea level and no sedimentation. A plane
dipping with a gradient of .0004 formed the initial
bathymetry (the grid resolution was 1 km). We
(6) include this initial period of rising sea level and no
sedimentation to include any residual loading
effects from the increased water load. Any sediment
deposited during this time would augment sub-
sidence rates by an amount directly related to the
density difference of sediment and water. However,
we assume that this signal is drowned by that of the
larger, more distributed water load. At 5500 years
BP, sea-level rise was stopped and sediment added
one lobe at a time. Fig. 3 shows the final deflections
due to the additional sediment load.
Fig. 4 shows the subsidence rate at New Orleans
throughout the model simulation. The dashed line
shows the response of the lithosphere due to
5
Deflection (m)
0
-5
-10
150
0 100
50 50 km)
Dis
tanc 100 nce (
e (k 0 Dista
m)
Fig. 3. Predicted lithospheric deflection due to sediment loading
of the Mississippi delta complex using the Sedflux flexure model.
Vertical red line indicates approximate location of New Orleans.
 ARTICLE IN PRESS
1326 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
0.6
Water and sediment
loading
Subsidence Rate (cm/y)
0.4
0.2
Water loading
0
0 2 4 6 8 10
Time (ky)
Fig. 4. Simulated subsidence rates of Mississippi delta at New
Orleans over the last 12 kyr. Dashed line shows subsidence rate
due only to increasing water loading caused by rising sea level.
Solid line includes subsidence caused by both water loading and
sediment loading. Note that both lines are coincident for the first
6000 years.
increased water loading from rising sea level.
Subsidence rates increase for 6 kyr until sea-level
rise stops. Because of the relaxation time (Eq. (7),
with t0 ¼ 2500 years), deflection continues even
though no new weight is added.
The solid line again shows subsidence rates
through the model run but now the effects from
both sediment and water are included. The lines
match one another for the first 6 kyr, but then
diverge with the addition of sediment. The sharp
jumps in subsidence rates are due to sudden shifts in
the location of sediment loading (due to lobe
switching). At the end of the simulation, we see
that the total subsidence rate is about .002 m yr1
and that 25% of that (.0005 m yr1) is due to water
loading that ended about 5400 years ago.
4. Additions to Sedflux 2D
When using sequence stratigraphy to analyze the
formation of a continental margin, it is critically
important to track the position of the shoreline.
Thus, it is also critically important to model the
processes that deposit and erode sediment near the
shore. This includes the deposition and erosion of
sediment within a river channel and the resuspen-
sion of shallow marine sediments.
The previous version of Sedflux models the
deposition and erosion of sediment along a river
using a geometric model based on an assumed
equilibrium profile of a river’s thalweg. This module
needed improvement because (1) it relies on know-
ing beforehand what the equilibrium profile for a
region is and (2) it is a steady-state model. In
general, the equilibrium shape of a river’s long-
itudinal profile depends on factors (for instance,
discharge, sediment load, and grain size) that vary
from region to region (Mackin, 1948) and so is
difficult to determine in advance. Also, a river does
not immediately attain its equilibrium shape but is
constantly changing in response to external forces
(Mackin, 1948; Leopold and Bull, 1979). A dynamic
model developed by Paola et al. (1992) replaces the
12 previous steady state.
Cross-shore processes driven by wave energy
govern the evolution of wave-dominated coastal
systems (Storms et al., 2002). Sedflux 1.0c modeled
this process using a diffusion-based model.
Although this is a common way of modeling coastal
evolution (Kaufman et al., 1992; Niedoroda et al.,
1995; Swenson et al., 2000; Schlager and Adams,
2001), it had two drawbacks that made it inade-
quate. Firstly, the evolving cross-shore profile
depended on an imposed (site-specific) diffusion
coefficient. Secondly, such models do not ade-
quately resolve the inner shelf. The work of Dean
(1991) and Bruun (1962) suggest that the cross-
shore profile is concave upward and follows a power
law. This is not the case for the diffusive model used
within Sedflux 1.0c or with those listed above.
A third cause of sediment deposition and erosion
in the near-shore environment is through surface
and subsurface plumes. For low river suspended
sediment concentrations Sedflux 1.0c implemented a
surface plume module (Syvitski et al., 1998) to
deliver suspended sediment to the shelf as ‘‘sediment
rain’’. For river concentrations greater than the
density of seawater, Sedflux 1.0c used a hyperpycnal
plume module (Mulder et al., 1998) to deliver the
sediment to the basin. Both of these models show
good agreement with field observations (Morehead
et al., 2001; Mulder et al., 1998) and use real-world
measures as input parameters. However, Kubo et al.
(2005) added a new hyperpycnal plume model to
Sedflux, giving the user a choice as to which model
to use.
4.1. Subaerial erosion and deposition by rivers
Sedflux 2.0 uses a model based on previous work
of Paola et al. (1992) to model the transport of
sediment along a river. This new module allows
both erosion and deposition along the stream profile
without having to define an equilibrium profile (as
was the case in previous Sedflux versions). Because
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
the diffusion equation forms the basis of the module
(Eq. (8)), the river’s equilibrium profile is linear
along its entire length (after setting time derivatives
to zero, the only remaining term is that of
topographic curvature). The diffusion coefficient
controls the rate at which the streambed moves
toward this equilibrium.
Accurate modeling of erosion and deposition on
the delta plain is critical for a number of reasons.
This process controls the position of the shoreline,
particularly during periods of rising sea level. For
instance, typical equilibrium river profiles over large
distances are often represented as an exponential or
a logarithmic function (Peckham, 1995). The
coefficients that define these curves may change
from river to river but their basic form will remain
the same—they are concave upward with shallow
and nearly linear slopes near the river mouth
(Peckham, 1995; Muto and Steel, 2000). Because
of this equilibrium geometry, a rapid sea-level fall
would result in an extreme amount of incision if the
stream profile were to instantaneously move to its
new steady-state position. In reality, it takes a
stream on the order of hundreds of years to reach
grade (Thorne and Swift, 1991). Thus, by reaching
grade immediately, a subsequent sea-level rise will
see larger than expected accommodation. However,
because the new model is based on the diffusion
equation, it limits the amount of erosion based upon
the profile’s current distance from equilibrium.
Thus, Sedflux 2.0 is capable of providing insight
into both the magnitude and rate of sea-level change
on stratigraphy.
While controlling the amount of accommodation,
the new model better controls the amount of
sediment transport to the marine basin. An equili-
brium model consistently provides an excess of
sediment to the basin during times of lowering sea
level. Because of the interplay between sediment
supply and accommodation, it is critical to model
them well.
To predict the erosion and deposition along a
river channel, the derivation of Paola et al. (1992)
does not assume beforehand that the diffusion
equation is the correct model; rather, they begin
from first principles with conservation of mass and
momentum. The final result is
qZ q2 Z
¼n 2 (8)
qt qx
where Z is the height of the bed, t is time, x is
position along the river channel, and n is the
1327
diffusion coefficient. The diffusion coefficient, n, is
expressed in terms of measurable quantities:
pffiffiffiffi
8hqiA cf
n (9)
C 0 ðs  1Þ
where /qS is the long-term average water dis-
charge, cf is a drag coefficient, C0 is the sediment
concentration of the bed, s is sediment specific
gravity, and A is a river-type dependent constant.
The value of A is user-defined and takes on one of
two values depending on the river type. For a
meandering river, A ¼ 1, while for a braided
river,A  ð=ð1 þ ÞÞ3=2 . The value e relates the shear
stress (t) in the center of a braided channel to the
critical shear stress (tc) needed for bank erosion
(t ¼ ð1 þ Þtc ). Measured values of e are typically
about 0.4 (Parker, 1978) for gravel bed rivers.
Sedflux 3D expands this erosion and deposition
routine to operate in a two-dimensional framework.
In this framework, the x dimension in Eq. (8) becomes
the along-channel distance. Elevation changes are
calculated in the same way, but are assumed to be
averaged over a user-defined channel-belt width.
Eq. (8) transports eroded river sediments into the
basin. Based on grain size, the sediment will move as
either bedload or suspended load. The process is
able to cause a change in both the amount of
sediment that the basin receives, as well as the way
that it receives it. For instance, during a period of
falling sea level (and channel incision), erosion
could conceivably cause the suspended sediment
concentration to increase enough to trigger a
hyperpycnal plume, where the normal mode of
dispersal would be a surface plume.
Fig. 5 shows an example of a Sedflux simulation
(operating in 3D mode) that demonstrates the use of
the new avulsion and channel erosion modules. For
this theoretical experiment, Sedflux ran for 1000
(model) years, using five distinct grain-size classes, a
grid resolution of 1 km, and a 1-year time step.
Throughout the model run, sediment supply was
held constant and the sediment distributed to the
basin through a surface plume. The snapshot of
basin topography in the left image of Fig. 5 is at a
low sea-level stand (25 m) and shows two incised
valleys in the delta topsets and several smaller
channelized features in the delta slope. The fall in
sea level caused marine sediments to be exposed,
which were then eroded by the new stream erosion
module. To simulate the incised valley, we specified
a decrease in the frequency of channel avulsion for
this period.
 ARTICLE IN PRESS
1328 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337

Sea Level (m)

0

-25
Time
Elevation (m)
0 0
-100 -100
-200 -200
-300 Incised -300
valleys

10 km 10 km
10 km 10 km

Fig. 5. Sedflux simulation that demonstrates new channel erosion model. (Left) As sea level falls, channel avulsion is reduced, which
causes increased erosion in regions noted with arrows. (Right) As sea level rises, plume sediments fill now submerged channels and river
sediment is deposited on delta plain.

in the bottom slices. In the top slice, two pockets of

coarser sediment separated by an undisturbed ridge
of finer sediment distinguish the two valley fills.
Depth (m)

0
-100 4.2. Cross-shore transport due to ocean storms
-200
z = -15 m
-300 z = -125 m
z = -195 m Sedflux 1.0c modeled the resuspension, transport,
D
2 and mixing of bottom sediments (wave energy,
ist 0 bottom currents, and bioturbation, for example) as
an 4 2
ce
(k 6 4 a single diffusive process. Sedflux now separately
m
) 8 6
ce (k )
m
models the resuspension and distribution of bottom
8 a n
10 10 Dist
sediments due to ocean-wave energy. The new
model is a composite of two models. The first is a
3 4 5 6 7 8
Grain size (φ)
modified version of the marine component of the
model presented by Swenson et al. (2005) to predict
Fig. 6. Sedflux internal stratigraphy as shown by depth slices at the reworking of the seafloor by offshore waves.
15, 125, and 195 m through deposited volume of sediment of The second is a diffusion-based model used only
channel incision experiment (Fig. 5).
within the near-shore.
For the purposes of this model, we define the
The right image of Fig. 5 shows the topography at near-shore as the region of depths less than the
the end of the simulation when sea level has risen closure depth, hc, as defined in Nicholls et al. (1998):
12 m. For rising sea level, the channel is once again  2
H ss
allowed to avulse over the entire delta. One can still hc ¼ 2:28H ss  6:85 (10)
see the flooded delta plain but plume sedimentation gT 2
has mostly filled in the incised valleys there. Base- where Hss is the height of the storm wave that is
level rise has also caused deposition on the new exceeded only 12 h each year and T its associated
delta plain. However, unlike plume sedimentation, period. This depth provides a boundary for the two
deposition on the delta plain was not as effective in wave resuspension models.
obscuring the incised valleys. For the outer shelf (depths greater than hc), we
Fig. 6 shows the internal stratigraphic architecture estimate the sediment flux, qs, at each position along
of the deposits this experiment generated. There are the profile as
three horizontal slices at elevations of 15, 125,  
16 r C fs ss U 3om U 2om qh
and 195 m and one dip profile at 0 km. The colors qs ¼ Is v0 þ (11)
correspond to average grain size along the slices. One 3p rs  r g ws 5ws qx
is able to distinguish the coarse near-shore sands where Cfs is a constant drag coefficient, ess is the
(yellow) in the top slice and more distal clays (blue) efficiency of suspended sediment transport, Is is the
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
time fraction (intermittency) of ocean storms, Uom is
the near-bed velocity due to waves, ws is settling
velocity, v0 is the velocity of cross-shore currents,
and h is the local water depth. The settling velocity
describes the rate at which grains settle from the
flow. For each grain size, Sedflux converts user-
defined removal rates to settling velocities (Bursik,
1995). Eq. (11) is from Coco (1999) and is the basis
for the marine component of the Swenson et al.
(2005) model. To accommodate a seafloor contain-
ing a range of grain sizes, we assume that the total
energy applied to the flow is distributed equally
among the multiple grain types. That is, an equal
fraction of energy is given to each sediment grain
size, which consumes the energy at its own rate
governed by its settling velocity.
In this equation, cross-shore currents and bathy-
metric slope (through the depth dependence of
shoaling waves) drive the flux of suspended sedi-
ment along the seafloor. However, the sediment flux
is also proportional to grain size (through settling
velocity), and water depth (through wave orbital
velocity). For each grain type, the user specifies a
grain size and settling velocity (although, in general,
settling velocity is a function of grain size, the user is
free to choose any settling velocity). The water
depth along the profile specifies the wave orbital
velocity through the equation for shoaling waves:
 
gb pffiffiffiffiffiffiffi h 3=4
U om ðhÞ ¼ ghb (12)
2 hb
where gb (assumed to be 0.6) is the ratio of wave
height to water depth where the wave will break, hb
(Swenson et al., 2005). To limit the size of sediments
that a wave is able to mobilize, we use Komar’s
(1998) equation for the threshold of sediment
motion:
8  
ru2t < :21 d 0 1=2 for Dp:5 mm
d
¼   (13)
ðrs  rÞgd : :46p d 0 1=4 for D4:5 mm
d
where D is grain size, ut is near-bottom threshold
velocity, and d0 is orbital diameter of the wave
motion. Airy wave theory (Komar, 1998) gives a
relationship of d0 to wave height (H), period (T),
length (L), and water depth (h):
pd 0 pH
ut ¼ ¼ (14)
T T sinhð2ph=LÞ
In this equation we calculate the changing
characteristics of a wave as it travels over shallow-
1329
ing bathymetry. With these modifications, Sedflux
predictions of seafloor morphology are similar to
that of the Swenson model, but Sedflux is able to go
a step further by predicting the grain-size structure
of the seafloor sediments. For instance, lags of
coarser grained sediments form at shallower depths
where waves have moved the finer grained sedi-
ments to calmer waters.
Within the near-shore zone (depths less than hc),
sediment transport is more complex due to non-
linear waves and alongshore current. Thus, we
approximate the combined effects of these processes
using a diffusion equation with a non-constant
diffusion coefficient
dh
qs ¼ kc x1m (15)
dx
where x is offshore position normalized by the
position of the near-shore boundary. We use a
diffusion coefficient with this specific x-dependence
to ensure the equilibrium profile will be a Bruun
profile of the form hpxm (m ¼ 2/3). The constant kc
is chosen so that the sediment flux at the near-shore
boundary matches that of the Swenson equation at
the same water depth.
Based on the work of Kubo et al. (2005), we used
Sedflux to simulate the building of the Po River
delta over the last 8 kyr. Fig. 7 shows the results
from two of these experiments. The two simulations
ran under the same conditions, varying only wave
height. Both used a yearly time step, a horizontal
resolution of 400 m, and a vertical resolution of
20 cm. Kubo et al. (2005) provide the sea-level
history and Kettner and Syvitski (2005) the
sediment supply over this period. Surface plumes
distribute sediment into the ocean. Once deposited,
waves and bottom currents resuspend and transport
the sediment. Wave action is able to control the
slope of the delta as it builds out into the ocean. In
the first case (Fig. 7A), waves were limited to about
0.5 m in height and averaged about 0.4 m to reflect
present-day conditions (Kubo et al., 2006; Cavaleri
et al., 1997). For the second case (Fig. 7B), average
storm conditions produced waves of about 3.4 m.
Fig. 8 compares the final bathymetry of these two
runs with the modern-day bathymetry of the Po
River delta. The dashed curves are three represen-
tative profiles from the present-day Po delta, and
the solid curves are the final bathymetry from the
simulations in Fig. 7. The bathymetry generated
through high wave conditions matches well with a
profile taken far from the major sediment supply of
 ARTICLE IN PRESS
1330 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337

0 0
Water Depth (m)

10 10

20 20

30 30
0 20 40 60 0 20 40 60
Distance (km) Distance (km)

4 5 6 7 8 9
Grain Size (φ)

Fig. 7. Simulated grain-size profiles of modern-day Po River delta. (Left) Small ocean storms allow delta to build a steep shoreface, while
larger waves (right) considerably reduce delta slope.

0 45o 10’
5
Water Depth (m)

10
45o 00’
Large Waves c
Longitude

15 (modeled)
Small Waves b
20 (modeled)
44o 50’
25 a
b
30 c a

44o 40’
0 10 20 30 40 50 12o 10’ 12o 30’ 12o 50’
Cross-shore Distance (km) Latitude

Fig. 8. Comparison of simulated bathymetry with modern-day bathymetry of Po River delta. (A) Solid red and blue lines are simulated
bathymetry using small and large waves, respectively. Dashed lines are representative profiles from actual delta. (B) A map of modern-day
Po River delta with bathymetric contours at 1 m intervals. Thick solid lines correspond to dashed profiles presented in panel (A).

the current delta (dashed line (a)). The low wave call when a turbidity current or hyperpycnal flow is
condition simulation matches better with profiles triggered. This choice provides a means for compar-
that were taken nearer the present Po River outlet ison of basin evolution due to different implementa-
(dashed lines (b) and (c)). tions of the same process. In addition, the user may
decide on one model based on computation time
4.3. Turbidity current model, sakura (INFLO is computationally faster) or the appro-
priateness of one theory over the other. Sedflux
Sedflux1.0c used the one-dimensional steady-state triggers a turbidity current if the clay fraction of a
turbidity current model INFLO (Skene et al., 1997). seafloor failure is greater than a user-specified
Kubo et al. (2005) added the turbidity current amount (typically, 10%), and a hyperpycnal flow
model, called sakura, to Sedflux (Kubo et al. (2005) if the river density exceeds that of the ocean.
give a complete description of the model). The user The layer-averaged, three-equation model of
must specify which of the two models Sedflux will Parker et al. (1986) is the basis of the sakura model.
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337 1331

Accordingly, the three governing equations are and increases linearly with burial depth at a rate, a
qh q (Mulder et al., 1998). Hiscott (1994) provides an
þ ðuhf Þ ¼ E w u (16) empirical approximation for pz with respect to the
qt qx
Rouse number Z0:
q q  
ðuhf Þ þ ðu2 hf Þ log p 1=4 ffi 0:124 log Z0 þ 1:2 (23)
z 2
qt qx
ðrs  rw Þg q ðr  rw Þghf CS Z 0  ws =ðkun Þ (24)
¼  ðCh2f Þ þ s
2rw qx rw
where k is the von Karman constant (k ¼ 0.4) and
 C d ð1 þ aÞu2 (17) u* is the shear velocity of the flow.
As an example of the turbidity current model, we
q q
ðChf Þ þ ðuChf Þ ¼ F d þ F (18) present results of Kubo et al. (2005). The aim of this
qt qx study was to apply Sedflux 2.0 to a tank experiment
where u is the flow velocity, hf is the flow thickness, performed in the experimental earthscape (XES)
g is the acceleration due to gravity, Ew is the water facility of the St. Anthony Falls Laboratory. The
entrainment coefficient, rs and rw are the densities tank controlled basin subsidence, and sediment
of the sediment and ambient water respectively, C is input, making it possible to simulate turbidity
the volume concentration of sediment, S is the current sedimentation in a subsiding basin while
bottom slope gradient, Cd is the drag coefficient varying the rate of sediment supply. The experiment
(Cd ¼ .004), a is the ratio of the drag force at the was intended to emphasize the impacts of sub-
upper flow surface to that at the bed (a ¼ .43), and sidence on the evolution of a turbidite fan.
Fd and Fe are the flux of sediment deposition Although Sedflux usually operates with time steps
and erosion, respectively. Fukishima et al. (1985) of days to years, this study used a time step of 1 min
provide an empirical relationship between the to simulate the pulse-type flows that were as short as
entrainment coefficient and Richardson number: 2 min in the laboratory experiment. Sediment
:00153 concentrations of each grain-size fraction, flow
Ew ¼ (19) velocity, depth, and width of the injection pipe
:0204 þ Ri
were set to match that of the experimental setting.
where the Richardson number is Sedflux was then set to run in a basin of the size of
ðrs  rw Þghf C the XES tank (5 m  2 m), using only the turbidity
Ri ¼ (20) current model, sakura, as a means of distributing
rw u2
sediment. Fig. 9 plots grain size and sediment age of
This system of three equations contains five the final deposits of the Sedflux experiment. Overall,
unknowns (hf, u, C, Fd, and Fe) and thus two more the deposit thins with distance from the source.
equations are necessary to close the problem. These However, within the mini-basin the deposit briefly
final closure equations describe the erosion and thickens. Grain size shows a more monotonic
deposition rates using the flow’s ability to carry decrease with distance due to disparate settling
sediment. The flow will begin depositing sediment rates of the component grains.
once its concentration exceeds its capacity (Gf).
Likewise, the flow will erode sediment if its
5. Event-based time step
concentration is below capacity and its shear stress
is greater than that of the bed. The equations used
Primarily, high-energy events control the distri-
for the deposition and erosion rates are
( bution of sediment due to wave resuspension and
ws Cð2  1=pz Þ; for pz o:5 river transport (Storms, 2003; Paola et al., 1992).
Fd ¼ (21) That is, short-lived but high-energy ocean storms
0; for pz X:5
are able to move more sediment than is moved
during the long periods of low energy that separate
F e ¼ ððC d rf u2  b=Þ=ða  86; 400 sÞÞ (22)
these events. This is also the case for intermittent
where ws is particle settling velocity and rf is the but large river floods that become ‘morphology-
flow density. The constants a and b (a ¼ 3.5 N m2 forming’ events (Paola et al., 1992). Thus, to
and b ¼ 0.2 N m1) define the depth profile of the correctly model these processes, one must operate
bed shear strength. The shear strength at the bed is b at a time step that captures these high-energy events.

1332 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
0.1
Water Depth (m)
0.3
0.5
0.7
0.9
0 1 2 3
Distance (m)
500 1500
Deposition Time (min)
Fig. 9. Sedflux predictions of a subsiding laboratory mini-basin experiment (taken from Kubo et al., 2005). For these experiments,
sediment is introduced to basin through turbidity current model, sakura.
A daily time step may be able to adequately
resolve individual storms and so correctly model the
various storms that cause large movements of
sediment. However, such a small time step is not
practical when conducting simulations covering
geologic time scales. The solution is to introduce a
time step that varies in length depending upon the
magnitude of energy applied to the system (Storms,
2003).
When using a variable time step, these processes
will require a time scale over which the process is
morphology forming. The ratio of the amount of
time during stormy conditions to the total length of
a time step defines the time scale (nTs/Tt). Here, a
‘stormy’ day is any day that is energetic enough to
cause significant morphologic changes. For the case
of the distribution of sediment within a river, this
may be the number of days that the river is bank-
full. For the resuspension of shallow marine
sediments, this may be the number of days that
ocean waves are large enough to result in significant
bed shear stress.
This procedure assumes that the long-lasting calm
periods move an insignificant amount of sediment
compared to the short-lived storms. This may not
always be the situation however. In such a case, the
model will simulate the calm periods as well. Here, a
second time scale is necessary, which is defined as
the ratio of the length of a single stormy period to
the length of the time step (mDts/Tt). The energy
conditions are constant over each of these time
steps. By also modeling the calm periods the
number of time steps increases linearly with the
ARTICLE IN PRESS
0.1
0.3
0.5
0.7
0.9
4 5 0 1 2 3 4 5
Distance (m)
2500 3 4 5 6
Grain Size (φ)
number of stormy periods (n/m). Even by modeling
both the calm and stormy events, the computational
saving is significant. Say the average length of a
large storm is a week, each year contains approxi-
mately one of these storms, and we wish to use a
100-year time step, then there will be 200 time steps
within the larger 100-year time step (100 storms
separated by 100 calm periods). On the other hand,
if we reduced the time step to a week to match the
length of the storms, over a 100-year period there
would be over 5000 time steps. Thus, only modeling
the storms and averaging over the calm periods
represents a speed increase of 25 times.
Fig. 10A shows a year’s worth of simulated
discharge data for the Eel River in Northern
California. Because the Eel is a ‘flashy’ river, this
technique works well since every year only a few
large flood events carry most of the sediment.
The dashed line plots the discharge for every
day of the year, while the solid line plots only the
events that Sedlux will model. Not every river
exhibits the flashy behavior of the Eel. For instance,
Fig. 10B shows a similar time series but this time for
the Klinaklini River. In this case, to account for
90% of the total sediment, one must consider at
least 150 discharge events (compared to 15 for
the Eel).
For wave action, we see a similar result (Fig. 11A
and B). We have conducted two simulations that
together demonstrate the effect that time averaging
has on wave resuspension. Both simulations ran
under the same conditions except the left simulation
(A) used an event-based time step that modeled the
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337 1333

1000
6000
800
Discharge (m3/s) 600
4000

400
2000
200

0 0
0 100 200 300 400 0 100 200 300 400
Time (days) Time (days)

Fig. 10. Time series of simulated sediment load from Eel River (A) and Kliniklini River (B). Solid lines represent events that carry 90% of
sediment for the year, while dashed lines plot daily values. For Eel River, the largest 15 events carry 90% of sediment as compared to
150 events for Kliniklini.

0 0
Water Depth (m)

10 10

20 20

30 30

0 10 20 30 40 0 10 20 30 40

0 0
Water Depth (m)

10 10

20 20
16 22 28 34 16 22 28 34
Distance (km) Distance (km)

3 4 5 6 7 8 9
Grain Size (φ)

Fig. 11. Experiment to demonstrate affect of ocean-storm averaging and variability on shape and internal structure of a delta. Upper-left
simulation (A) used an event-based time step, while upper-right (B) used yearly averaged ocean-storm conditions. Bottom two simulations
use ocean-storm distributions with high variability (C) and low variability (D). Mean of distributions is the same in both cases. Both
simulations result in similarly shaped deltas but with different internal architecture.

largest ocean storms and averaged over calm structure of the delta. The simulation shown in
periods. The right simulation (B) used yearly Fig. 11C and D used ocean-wave distributions with
averaged wave conditions. Larger storms produce large (C) and small (D) variability. Although the
a larger subaqueous delta, and more variable final bathymetry is similar in both cases, the internal
internal architecture than average conditions. structure generated by the highly variable storm
As with flood events, the variability of storm environment produced bedding patterns with greater
events within the year also has an affect on the variability.
 ARTICLE IN PRESS
1334 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
6. Program notes and structure
In this section we give a brief description of the
main input and output files for Sedflux. A more
detailed user manual is included with the Sedflux
distribution (available via an anonymous ftp from
iamg.org).
6.1. Input files
There are four principal input files required for a
Sedflux simulation. These are ASCII files that
describe the computational environment, boundary
conditions, sediment properties, and process-speci-
fic variables. The user specifies the name of the file
that describes the computational environment on
the command line. This file then references all
subsequent input and output files.
6.1.1. Computational environment
This file describes global parameters for a Sedflux
simulation as well as the location of the other input
files. The file consists of at least two groups. The
first defines the spatial resolution, and the location
of the sediment and bathymetry files. Subsequent
groups define the time step, duration, and process
definition files for periods within the simulation.
A simulation consists of at least one such period.
6.1.2. Bathymetry
The bathymetry file is a comma-delimited text file
that describes the starting elevations of the nodes
used in the Sedflux grid. In 2D mode, this file
contains two columns of data: horizontal position
and elevation. If necessary, Sedflux will linearly
interpolate between points to create an array with
the specified horizontal resolution. In 3D mode,
bathymetry is provided as a matrix with each
element specifying the elevation of a particular
node in the model grid. The horizontal resolutions
define the positions of the nodes.
6.1.3. Sediment
The sediment tracked in Sedflux is composed of a
number of distinct grain types. The user defines
these grain types in a sediment file. This file consists
of a series of groups that define various grain type
specific parameters. The parameters that define a
grain type are grain size, grain density, saturated
density, void ratio in closest compacted state,
diffusion index (b in Eq. (4)), removal rate, and a
compaction coefficient.
6.1.4. Process definitions
The process definition file specifies which pro-
cesses are active and sets any necessary parameters
for the processes. This file consists of a series of
groups (one for each process) that contain para-
meter/value pairs that define the process-specific
constants.
6.2. Output files
There are two types of output files for Sedflux,
both of which consist of binary data. The first is a
three-dimensional grid of sediment property data
for each cell of sediment within the simulation. The
second records seafloor properties over the entire
simulation grid.
6.2.1. Property file
Sedflux is able to track a large number of
sediment properties throughout a simulation
(Hutton and Syvitski, 2003). A property file records
a specified property for all of the sediment cells of a
simulation. For large simulations, these files contain
a large amount of data and so they are written in a
compressed format. Appendix A (available via an
anonymous ftp from iamg.org) provides a matlab
m-file that is able to read these output files.
6.2.2. Measuring station file
Sedflux enables the user to set up a series of
‘tripods’ on the simulation grid. Each tripod
measures a user-specified property at regular inter-
vals. Bathymetric slope, water depth, mean grain
size, and percent clay are examples of such proper-
ties. Again, a matlab m-file used for reading these
files is provided in Appendix A.
7. Application programming interface
The Sedflux project is aimed toward helping two
distinct sets of users. The first consists of those
whose interest is in conducting numerical experi-
ments. The second group (which oftentimes over-
laps with the first) consists of those that write
numerical models. Thus far, this paper has de-
scribed the interface that a model user sees.
However, Sedflux also provides a set of libraries
that contain an API aimed toward the model
developer.
The Sedflux API consists of functions that allow
developers to interact with the Sedflux modeling
environment without having to worry about specific
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
implementation details. The benefit of the interface
is that it provides the developer with a wide range of
functions to use and so reduce the amount of
duplicate code written. The interface eliminates
dependencies of a developer’s program with the
underlying implementation of the Sedflux architec-
ture. Thus, the programmer need not worry that the
implementation will change since the interface will
not. The programmer can program to an interface,
rather than to an implementation.
The Sedflux architecture consists of many objects
for developers to access. Sediment cells hold
packages of sediment that can be compacted, added
to (or subtracted from) one another, queried for
geotechnical properties, or stacked on top of one
another to form columns of sediment. These
columns of sediment combine with one another to
form cubes of sediment. The Sedflux framework
contains many other objects that describe, for
instance, types of grains, sediment distributions,
waves, and rivers. Through the API, the user can
access and manipulate these structures without
having to worry about how the structures are
actually built.
8. Summary
Sedflux 2.0 provides the earth science community
with a new version of a stratigraphic simulation
model. Based on the architecture of previous
versions, Sedflux is now able to track the evolution
of stratigraphy in three dimensions. However, it can
still be run in a two-dimensional mode that
functions much like previous versions.
Although it may seem as though the three-
dimensional mode might render a two-dimensional
mode obsolete, this is not the case. Since the
publication of Sedflux 1.0c, Sedflux has seen
advances in both versions. The 2D version is faster,
requires less memory, is easier to set up, and
simulates a wider range of processes. Apart from
simulating regions that are nearly two-dimensional,
the 2D modeling environment is ideal for obtaining
a rough estimate of inputs that will later be used
within a larger three-dimensional run. In addition, it
is a testing environment for new models that will
eventually progress to become modules that are run
within the 3D environment.
Sedflux 1.0c contained many of the important
processes that distribute sediment within the marine
environment. However, it lacked a shelf sediment
transport model, and a fluvial deposition and
1335
erosion subroutine. Both of these processes are
now present within Sedflux 2.0. In addition, Sedflux
now contains two hyperpycnal plume models for the
user to choose from. The Sedflux environment
provided an ideal place to test the new model
(sakura) by providing a system to run many
simulations with a large range of boundary condi-
tions and to explore the feedbacks both between
previous flows as well as other processes.
Many processes are presently not part of the new
three-dimensional Sedflux architecture. For exam-
ple, although Sedflux is able to simulate turbidity
currents and debris flows along a profile, it lacks
equivalent models that are able to travel over a two-
dimensional grid. However, the structure of Sedflux
allows these additions to be made and researchers
are invited to contribute to this community model-
ing effort.
The Sedflux API provides a means for modelers
to access the underlying architecture. Researchers
are now able to communicate with Sedflux without
having to worry about the details of the implemen-
tation of the architecture. This makes it easier to
either write wrappers around existing models so that
they can communicate with Sedflux or to write
models from scratch within the Sedflux framework.
In addition, the modeler need not duplicate the
work of others by rewriting commonly used pieces
of code such as grid interpolation, or matrix solvers.
All of this code is open source and so we encourage
other developers to contribute to this growing
library. As the library grows, the amount of
duplicate work done in the community will decline,
and the quality of the library will increase, which
will benefit everyone.
Appendix A. Supplementary materials
Supplementary data associated with this article
can be found in the online version at doi:10.1016/
j.cageo.2008.02.013.
References
Amos, D.E., 1986. ALGORITHM 644: a portable package for
Bessel functions of a complex argument and nonnegative
order. ACM Transactions on Mathematical Software 12 (3),
265–273.
Anderson, R.S., 1994. Evolution of the Santa Cruz Mountain,
California, through tectonic growth and geomorphic decay.
Journal of Geophysical Research 99 (B10), 20,161–20,179.
 ARTICLE IN PRESS
1336 E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
Angevine, C.L., Heller, P., Paola, C., 1990. Quantitative
Sedimentary Basin Modeling, Continuing Education Course
Note Series, vol. 32. The American Association of Petroleum
Geologists, Tulsa, OK, 133pp.
Bouma, A.H., 2000. Coarse-grained and fine-grained turbidite
systems as end member models: applicability and dangers.
Marine and Petroleum Geology 17, 137–143.
Bouma, A.H., 2001. Fine-grained submarine fans as possible
recorders of long- and short-term climatic changes. Global
and Planetary Change 28, 85–91.
Bonham-Carter, G.F., Sutherland, A.J., 1967. Diffusion and
settling of sediments at river mouths: a computer simulation
model. Transactions of the Gulf Coast Association of
Geological Societies 17, 326–338.
Bruun, P., 1962. Sea level rise as a cause of shore erosion. Journal
of the Waterways and Harbours Division, American Society
of Civil Engineers (ASCE) 88, 117–130.
Bursik, M.I., 1995. Theory of sedimentation of suspended
particles from fluvial plumes. Sedimentology 42, 831–838.
Cavaleri, L., Curiotto, S., Mazzoldi, A., Pavanati, M., 1997.
Long term directional wave recording in the Northern
Adriatic Sea. Nuovo Cimento 20C, 103–110.
Coco, G., 1999. Mechanics for the formation of rhythmic
topography in the nearshore region. Ph.D. Dissertation,
University. of Plymouth, UK, 149pp.
Chassaing, P., George, J., Claria, A., Sananes, F., 1974. Physical
characteristics of subsonic jets in a cross-stream. Journal of
Fluid Mechanics 62, 41–64.
Dean, R.G., 1991. Equilibrium beach profiles—characteristics
and applications. Journal of Coastal Research 7 (1), 53–84.
Fisk, H.N., McFarlan Jr., E., 1955. Late Quaternary deltaic
deposits of the Mississippi River. Geological Society of
America (Special Paper) 62, 279–302.
Fukishima, Y., Parker, G., Pantin, H.M., 1985. Prediction of
ignitive turbidity currents in Scripps Submarine Canyon.
Marine Geology 67, 55–81.
Granjeon, D., Joseph, P., 1999. Concepts and applications of a 3-
D multiple lithology diffusive model in stratigraphic model-
ing. In: Harbough, J.W., Watney, W.L., Rankey, E.C.,
Slingerland, R., Goldstein, R.H., Franseen, E.K. (Eds.),
Numerical Experiments in Stratigraphy; Recent Advances in
Stratigraphic and Sedimentologic Computer Simulations, vol.
62. Society for Sedimentary Geology (SEPM) Special Pub-
lication, pp. 197–210.
Harris, C.K., Wiberg, P.L., 2001. A two-dimensional, time-
dependent model of suspended sediment transport and bed
reworking for continental shelves. Computers & Geosciences
27, 675–690.
Hiscott, R.N., 1994. Loss of capacity, not competence, as the
fundamental process governing deposition from turbidity
currents. Journal of Sedimentary Research A 64, 209–214.
Hutton, E.W.H., Syvitski, J.P.M., 2003. Advances in the
numerical modeling of sediment failure during the develop-
ment of a continental margin. Marine Geology 203 (3–4),
367–380.
Huybrechts, P., 2002. Sea-level changes at the LGM from ice-
tsunamic reconstructions of the Greenland and Antarctic ice
sheets during the glacial cycles. Quaternary Science Reviews
21, 203–231.
Imran, J., Parker, G., Katopodes, N.D., 1998. A numerical model
of the channel inception on submarine fans. Journal of
Geophysical Research 103 (C1), 1219–1238.
Imran, J., Harff, P., Parker, G., 2001. A numerical model of
submarine debris flow with graphical user interface. Compu-
ters & Geosciences 27, 717–729.
Kaufman, P., Grotzinger, J.P., McCormick, D.S., 1992. Depth-
dependent diffusion algorithm for simulation of sedimenta-
tion in shallow marine depositional systems. In: Franseen,
E.K., Watney, W.L., Kendall, C.G.S., Ross, W. (Eds.),
Sedimentary Modeling: Computer Simulations and Methods
for Improved Parameter Definition, vol. 233. Kansas
Geological Survey Bulletin, pp. 489–508.
Kettner, A.J., Syvitski, J.P.M., 2005. Predicting discharge and
sediment flux to the Po river, Italy since the Late Glacial
Maximum. International Association of Sedimentologists
(IAS) Special Issue.
Komar, P.D., 1998. Beach Processes and Sedimentation, second
ed. Prentice-Hall, Englewood Cliffs, NJ, 429pp.
Kubo, Y., Syvitski, J.P.M., Hutton, E.W.H., Paola, C., 2005.
Advance and application of the stratigraphic simulation
model 2D-SedFlux: from tank experiment to geological scale
simulation. Sedimentary Geology 178, 187–195.
Kubo, Y., Syvitski, J.P.M., Hutton, E.W.H., Kettner, A.J., 2006.
Inverse modeling of post Last Glacial Maximum transgressive
sedimentation using 2D-SedFlux: application to the northern
Adriatic Sea. Marine Geology 234, 233–243.
Lambeck, K., 1988. Geophysical Geodesy, The Slow Deforma-
tions of the Earth, first ed. Clarendon Press, Oxford, 718pp.
Leopold, L.B., Bull, W.B., 1979. Base level, aggradation, and
grade. Proceedings of the American Philosophical Society
123, 168–202.
Li, M.Z., Amos, C.L., Heffler, D.E., 1997. Boundary layer
dynamics and sediment transport under storm and non-storm
conditions on the Scotian shelf. Marine Geology 141,
157–181.
Mackin, J.H., 1948. Concept of the graded river. Geological
Society of America Bulletin 59, 463–512.
Martinez, P.A., Harbaugh, J.W., 1993. Simulating Nearshore
Environments, first ed. Pergamon Press, New York, NY,
265pp.
McAdoo, B.G., Pratson, L.F., Orange, D.L., 2000. Submarine
landslide geomorphology, US continental slope. Marine
Geology 189, 103–136.
Milliman, J.D., Qin, Y.S., Ren-Meie, E., Saito, Y., 1987. Man’s
influence on the erosion and transport of sediment by Asian
rivers; the Yellow River (Huange) example. Journal of
Geology 95 (6), 751–762.
Morehead, M.D., Syvitski, J.P.M., Hutton, E.W.H., 2001. The
link between abrupt climate change and basin stratigraphy: a
numerical approach. Global and Planetary Change 28,
107–127.
Mulder, T., Syvitski, J.P.M., Skene, K.I., 1998. Modeling of
erosion and deposition by turbidity currents generated at
river mouths. Journal of Sedimentary Research 68 (1),
124–137.
Muto, T., Steel, R.J., 2000. The accommodation concept in
sequence stratigraphy: some dimensional problems and
possible redefinition. Sedimentary Geology 130, 1–10.
Nicholls, R.J., Birkemeier, W.A., Lee, G., 1998. Evaluation of
depth of closure using data from Duck, NC, USA. Marine
Geology 148, 179–201.
Niedoroda, A.W., Reed, C.W., Swift, D.J.P., Aroto, H.,
Hoyanagi, K., 1995. Modeling shore-normal large-scale
coastal evolution. Marine Geology 126, 181–199.
 ARTICLE IN PRESS
E.W.H. Hutton, J.P.M. Syvitski / Computers & Geosciences 34 (2008) 1319–1337
Orton, G.J., Reading, H.G., 1993. Variability of deltaic processes
in terms of sediment supply, with particular emphasis on
grain size. Sedimentology 40, 475–512.
Overeem, I., Veldkamp, A., Tebbens, L., Kroonenberg, S.B.,
2003. Modelling of Holocene stratigraphy and depocentre
migration of the Volga delta due to Caspian Sea-level change.
Sedimentary Geology 159 (3), 159–175.
Overeem, I., Syvitski, J.P.M., Hutton, E.W.H., 2005. Three-
dimensional numerical modeling of deltas. In: Bhattacharya,
J.P., Giosan, L. (Eds.), River Deltas: Concepts, Models and
Examples, vol. 83. SEPM (Society for Sedimentary Geology)
Special Publication, pp. 13–30.
Paola, C., 2000. Quantitative models of sedimentary basin filling.
Sedimentology 47, 121–178.
Paola, C., Heller, P.L., Angevine, C.L., 1992. The large-scale
dynamics of grain-size variation in alluvial basins, 1: theory.
Basin Research 4 (2), 73–90.
Parker, G., 1978. Self-formed straight rivers with equilibrium
banks and mobile beds. Part 2. The gravel river. Journal of
Fluid Mechanics 89, 127–146.
Parker, G., Fukushima, Y., Pantin, H.M., 1986. Self-accelerating
turbidity currents. Journal of Fluid Mechanics 171, 145–181.
Paulson, A., Zhong, A., Wahr, J., 2005. Modelling post-glacial
rebound with lateral viscosity variations. Geophysics Journal
International 163, 357–371.
Peckham, S.D., 1995. Self-similarity in the three-dimensional
geometry and dynamics of large river basins. Ph.D. Disserta-
tion, University of Colorado.
Peltier, W.R., 1998. Postglacial variation in the level of the sea:
implications for climate dynamics and solid-earth geophysics.
Reviews of Geophyics 36, 603–689.
Pratson, L.P., Imran, J., Hutton, E.W.H., Parker, G., Syvitski,
J.P.M., 2001. BANG1D: a one-dimensional, Lagrangian
model of subaqueous turbid surges. Computers & Geos-
ciences 27, 701–716.
Reed, C.W., Niedoroda, A.W., Swift, D.J.P., 1999. Modeling
sediment entrainment and transport processes limited by bed
armoring. Marine Geology 154, 143–154.
Richards, M., Bowman, M., Reading, H., 1998. Submarine-fan
systems I: characterization and stratigraphic prediction.
Marine and Petroleum Geology 15, 689–717.
Ritchie, B.D., Hardy, S., Gawthorpe, R.L., 1999. Three-
dimensional numerical modeling of coarse grained clastic
deposition in sedimentary basins. Journal of Geophysical
Research 104 (B8), 17,759–17,780.
Ross, W.C., Halliwell, B.A., May, J.A., Watts, D.E., Syvitski,
J.P.M., 1994. Slope readjustment: a new model for the
development of submarine fans and aprons. Geology (22),
511–514.
Schlager, W., Adams, E.W., 2001. Model for the sigmoidal
curvature of submarine slopes. Geology 29 (10), 883–886.
Skene, K.I., Mulder, T., Syvitski, J.P.M., 1997. INFLO1:
a model predicting the behaviour of turbidity currents
1337
generated at river mouths. Computers & Geosciences 23,
975–991.
Steckler, M.S., 1999. High resolution sequence stratigraphic
modeling: 1. The interplay of sedimentation, erosion and
subsidence. In: Harbaugh, J., Watney, L., Rankey, G.,
Slingerland, R., Goldstein, R., Franseen, E. (Eds.), Numerical
Experiments in Stratigraphy, vol. 62. Society for Sedimentary
Geology (SEPM) Memoir, pp. 139–149.
Storms, J.E.A., 2003. Event-based stratigraphic simulation of
wave-dominated shallow-marine environments. Marine Geol-
ogy 199, 83–100.
Storms, J.E.A., Weltje, G.J., Van Dijke, J.J., Geel, C.R.,
Kroonenberg, S.B., 2002. Process-response modeling of
wave-dominated coastal systems: simulating evolution and
stratigraphy on geological timescales. Journal of Sedimentary
Research 72, 226–239.
Sun, T., Paola, C., Parker, G., Meakin, P., 2002. Fluvial fan-
deltas: linking channel processes with large-scale morphody-
namics. Water Resources Research 38 (8), 1151.
Swenson, J.B., Voller, V.R., Paola, C., Parker, G., Marr, J.G.,
2000. Fluvial-deltaic sedimentation: a generalized Stefan
problem. European Journal of Applied Mathematics 11,
433–452.
Swenson, J.B., Paola, C., Pratson, L., Voller, V.R., Murray,
A.B., 2005. Fluvial and marine controls on combined
subaerial and subaqueous delta progradation: morphody-
namic modeling of compound-clinoform development. Jour-
nal of Geophysical Research 110, F02013.
Syvitski, J.P.M., Alcott, J.M., 1995. DELTA6: numerical
simulation of basin sedimentation affected by slope failure
and debris flow runout. In: Proceedings of the Pierre Beghin
International Workshop on Rapid Gravitational Mass Move-
ments, Grenoble, France, pp. 305–312.
Syvitski, J.P.M., Hutton, E.W.H., 2001. 2D SEDFLUX 1.0C: an
advanced process-response numerical model for the fill of
marine sedimentary basins. Computers & Geosciences 27 (6),
731–754.
Syvitski, J.P.M., Nicholson, M., Skene, K., Morehead, M.D.,
1998. PLUME1.1: deposition of sediment from a fluvial
plume. Computers & Geosciences 24 (2), 159–171.
Tetzlaff, D.M., Priddy, G., 2001. Sedimentary process modeling:
from academia to industry. In: Merriam, D.F., Davis, J.C.
(Eds.), Geological Modeling and Simulation: Sedimentary
Systems. Kluwer Academic/Plenum Publishers, New York,
NY, pp. 45–70.
Thorne, J.A., Swift, D.J.P., 1991. Sedimentation on continental
margins, II: application of the regime concept. In: Swift,
D.J.P., Oertel, G.F., Tillman, R.W., Thorne, J.A. (Eds.),
Shelf Sand and Sandstone Bodies: International Association
of Sedimentologists, special publication 14, pp. 33–58.
Wiberg, P.L., Smith, J.D., 1983. A comparison of field data and
theoretical models for wave–current interactions at the bed on
the continental shelf. Continental Shelf Research 2, 147–162.