ARTICLE IN PRESS

Computers & Geosciences 34 (2008) 1284–1306

www.elsevier.com/locate/cageo

Development of a three-dimensional, regional, coupled wave,

current, and sediment-transport model$
John C. Warnera,, Christopher R. Sherwooda, Richard P. Signella, Courtney
K. Harrisb, Hernan G. Arangoc
a
US Geological Survey, Coastal and Marine Geology Program, Woods Hole, MA 02543, USA
b
Virginia Institute of Marine Science, Gloucester Point, VA 23062, USA
c
Institute for Marine and Coastal Sciences, Rutgers-The State University of New Jersey, New Brunswick, NJ 08901, USA
Received 5 September 2006

Abstract

We are developing a three-dimensional numerical model that implements algorithms for sediment transport and
evolution of bottom morphology in the coastal-circulation model Regional Ocean Modeling System (ROMS v3.0), and
provides a two-way link between ROMS and the wave model Simulating Waves in the Nearshore (SWAN) via the Model-
Coupling Toolkit. The coupled model is applicable for fluvial, estuarine, shelf, and nearshore (surfzone) environments.
Three-dimensional radiation-stress terms have been included in the momentum equations, along with effects of a surface
wave roller model. The sediment-transport algorithms are implemented for an unlimited number of user-defined non-
cohesive sediment classes. Each class has attributes of grain diameter, density, settling velocity, critical stress threshold for
erosion, and erodibility constant. Suspended-sediment transport in the water column is computed with the same
advection–diffusion algorithm used for all passive tracers and an additional algorithm for vertical settling that is not
limited by the CFL criterion. Erosion and deposition are based on flux formulations. A multi-level bed framework tracks
the distribution of every size class in each layer and stores bulk properties including layer thickness, porosity, and mass,
allowing computation of bed morphology and stratigraphy. Also tracked are bed-surface properties including active-layer
thickness, ripple geometry, and bed roughness. Bedload transport is calculated for mobile sediment classes in the top layer.
Bottom-boundary layer submodels parameterize wave–current interactions that enhance bottom stresses and thereby
facilitate sediment transport and increase bottom drag, creating a feedback to the circulation. The model is demonstrated
in a series of simple test cases and a realistic application in Massachusetts Bay.
r 2008 Elsevier Ltd. All rights reserved.

Keywords: Sediment transport; Nearshore modeling; Three-dimensional numerical model; Model coupling

1. Introduction
$
http://www.unidata.ucar.edu/software/netcdf/.
Corresponding author. Tel.: +1 508 457 2237;
1.1. Community modeling approach
fax: +1 508 457 2310.
E-mail addresses: jcwarner@usgs.gov (J.C. Warner),
csherwood@usgs.gov (C.R. Sherwood), rsignell@usgs.gov
Models for transport and long-term fate of
(R.P. Signell), ckharris@vims.edu (C.K. Harris), particles in coastal waters are essential for a variety
arango@marine.rutgers.edu (H.G. Arango). of applications related to commerce, defense, public

0098-3004/$ - see front matter r 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cageo.2008.02.012
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1285

health, and the quality of the marine environment. directives defined in an include file (cppdefs.h). This
There exists a need to develop a sediment-transport ensures that memory is allocated only as needed,
model that is freely available, well tested, widely and that only relevant computational algorithms are
accepted, and applicable to a variety of coastal compiled, creating a more efficient executable file.
settings. The ROMS community interacts through inter-
We are using a community approach to develop net, publications, and annual meetings. Revised
the model as a tool for both research and practical versions of model code, reports of bugs, and
applications. The need and value for this approach solutions to problems are posted on-line at the
was elucidated in a community sediment-transport ROMS website. Most model inputs and output files,
modeling workshop (Sherwood et al., 2002). A including those relevant for sediment-transport
community effort enables us to include a broad calculations, are written using the NetCDF data
range of processes and scales, more than would be architecture (http://www.unidata.ucar.edu/software/
feasible for individuals or small groups. We have netcdf/) in a format compliant with climate and
started with a model that is being used and forecast (CF) metadata conventions (http://www.
developed actively by a large research community. unidata.ucar.edu/software/netcdf/conventions.html).
We are incorporating proven methodologies from This allows users to capitalize on existing visualiza-
other models such as ECOMSed, EFDC, COHE- tion and processing tools that have been and are
RENS, and Delft3D. Scientists and engineers may continuing to be developed by various communities,
contribute to the model according to their expertize, and encourages documentation of model runs via
and users (including scientists from other disci- metadata embedded in input and output files. The
plines, students, resource managers, engineers, and code is written in modular Fortran90 and runs
operational personnel) may draw from well-tested, in serial mode or on multiple processors using
state-of-the-art algorithms. Incorporation of alter- either shared- or distributed-memory architectures
native parameterizations for similar processes (OpenMP or MPI). These characteristics made
allows us to compare them in identical frameworks. ROMS an ideal starting point for our development
Collaborative work on a community model helps of a community sediment-transport model.
identify key research and modeling issues, and
efficiently focus research efforts, minimizing dupli- 1.3. Objective
cation and preventing critical components from
being overlooked. Wide use and broad participation This paper describes the implementation in
in model development, along with extensive testing ROMS of a sediment-transport model, new bot-
and peer review, will produce a robust model that tom-boundary layer routines, a bed model to track
can serve the scientific community. morphology and stratigraphy, wave–current inter-
action, and coupling of ROMS to the surface wave
1.2. Regional oceanographic modeling system model Simulating WAves Nearshore (SWAN). The
(ROMS) coupled system is distributed as ROMS v3.0. Here
we provide background information about ROMS,
Our eventual goal is to produce a sediment model details of the new sediment algorithms, methods for
that may be coupled in a flexible way to any of a two-way coupling of ROMS to SWAN, and several
number of hydrodynamic modules. To reach this examples that demonstrate specific capabilities of
goal, we started with a specific model so we could the modeling system.
develop sediment-transport algorithms in the con- The model is continually evolving, and this
text of a completely functional framework. The description represents a snap-shot of current cap-
advancements we are making and the algorithms abilities and algorithms. Our plan is to eventually
that we are developing are linked integrally, for extract the sediment-transport components and
now, with the Regional Ocean Modeling System provide them as separate modules. Our long-term
(ROMS). ROMS is a numerical coastal ocean objectives are to expand the modeling system to
circulation model that includes several submodels include effects of cohesive sediments, couple with
that simulate, for example, sea ice, biological Boussinesq phase-resolving wave models, add sub-
processes, and sediment transport. For each appli- models for wave runup on the beach, and include
cation, different components of the model are submodels for detailed fluid mechanics and particle
included or excluded via C-preprocessor (cpp) interactions near the bed.
 ARTICLE IN PRESS
1286 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

2. Circulation and wave model framework 1 qp g

0¼  H zr (3)
r0 qs r0
2.1. Hydrodynamic model
with continuity as
ROMS is a three-dimensional, free surface, qZ qðH z uÞ qðH z vÞ qðH z OÞ
terrain-following numerical model that solves fi- þ þ þ ¼0 (4)
qt qx qy qs
nite-difference approximations of the Reynolds-
averaged Navier–Stokes (RANS) equations using and scalar transport:
the hydrostatic and Boussinesq assumptions (Chas- qðH z CÞ qðuH z CÞ qðvH z CÞ qðOH z CÞ
signet et al., 2000; Haidvogel et al., 2000) with a þ þ þ
qt qx qy qs
split-explicit time stepping algorithm (Shchepetkin  
q n y qC
and McWilliams, 2005; Haidvogel et al., 2007). It ¼ c0 w0  þ C source (5)
qs H z qs
uses a horizontal curvilinear Arakawa C grid and
vertical stretched terrain-following coordinates (see where u, v, and O are the mean components of
Section 2). ROMS has a flexible structure that velocity in the horizontal (x and y) and vertical (s)
allows choices for many of the model components, directions respectively; the vertical sigma coordinate
including options for advection schemes (second s ¼ (zZ)/D ranges from s ¼ 1 at the bottom to
order, third order, fourth order, and positive s ¼ 0 at the free surface; z is the vertical coordinate
definite), turbulence submodels, and boundary positive upwards with z ¼ 0 at mean sea level; Z is
conditions. It includes bottom- and surface-bound- the wave-averaged free-surface elevation; D is the
ary layer submodels, air-sea fluxes, surface drifters, total water depth D ¼ h+Z; h is the depth below
a nutrient-phytoplankton-zooplankton model, and mean sea level of the sea floor; Hz is the grid-cell
a fully developed adjoint model for computing thickness; f is the Coriolis parameter. An overbar
model inverses and data assimilation. Momentum, indicates a time average, and a prime (0 ) indicates a
scalar advection, and diffusive processes are repre- fluctuating turbulent quantity. Pressure is p;r and r0
sented using transport equations. The density field is are total and reference densities for seawater; g is
determined from an equation of state that accounts acceleration due to gravity; n and ny are molecular
for temperature, salinity, and suspended-sediment viscosity and diffusivity; C represents a tracer
concentrations. In this paper, the term constant quantity (for example, salt, temperature, and
refers to values that are time-invariant, and the term suspended-sediment); Csource are tracer source/sink
uniform refers to values that do not vary in space. terms; and a function r ¼ f(C) is required to close
The governing Eqs. (1)–(5) are presented in flux the density relation. These equations are closed by
form, in Cartesian horizontal coordinates and sigma parameterizing the Reynolds stresses and turbulent
vertical coordinates. For curvilinear grids, addi- tracer fluxes as
tional metric terms appear (Haidvogel et al., 2000)
qu qv
that are not shown here. A complete list of variables u0 w0 ¼ K M ; v0 w0 ¼ K M ,
is given in Table 1. The momentum equations are: qz qz
qr
qðH z uÞ qðuH z uÞ qðvH z uÞ qðOH z uÞ r0 w0 ¼ K H (6)
þ þ þ f qz
@t qx qy qs where KM is the eddy viscosity for momentum and
 
H z qp qZ q 0 0
n qu KH is the eddy diffusivity. Eddy viscosities and eddy
H zv ¼   H zg  uw 
r0 qx qx qs H z qs diffusivities are calculated using one of five options
qðH z S xx Þ qðH z S xy Þ qS px for turbulence-closure models in ROMS: (i) Brunt-
  þ (1) Väisälä frequency mixing in which mixing is based
qx qy qs
on the stability frequency; (ii) a user-provided
qðH z vÞ qðuH z vÞ qðvH z vÞ qðOH z vÞ analytical expression such as a constant or parabolic
þ þ þ þf shape; (iii) the K-profile parameterization (Large
qt qx qy @s
  et al., 1994), expanded to include both surface
H z qp qZ q 0 0
n qv
H zu ¼   H zg  vw  and bottom-boundary layers by Durski et al., 2004;
r0 qy qy qs H z qs
(iv) Mellor-Yamada level 2.5 (MY2.5) method
qðH z S yx Þ qðH z Syy Þ qSpy (Mellor and Yamada, 1982); and (v) the generic
  þ (2)
qx qy qs length-scale (GLS) method (Umlauf and Burchard,
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1287

Table 1
List of symbols

Symbol Description Dimensions

Ab Wave orbital excursion amplitude m

AR Wave roller area m2
C Tracer (temperature, salt, or suspended-sediment concentration) 1C, salinity, or kg m3
–
CdBF Bedform drag coefficient
Csource Tracer source/sink term C units ms1
D Total water depth m
D50 Median grain diameter m
Wave energy m3 s2
Es Erosion source term kg m2 s1
E0 Erosion rate for each sediment class kg m2 s1
FCC FCS FSS FSC Hyperbolic functions –
Hz Grid cell thickness m
KH Eddy diffusivity m2 s1
KM Eddy viscosity m2 s1
L Wave length m
N Wave action density m3 s1
Nbed Number of bed layers –
Qb Fraction of breaking waves –
Rz Wave roller shape function –
Sw Wave energy source/sink term m3 s2
Spx Spy Vertically varying vertical radiation stresses m2 s2
Sxx Sxy Syx Syy Vertically varying horizontal radiation stresses m2 s2
Sxx Sxy Syx Syy Vertically integrated horizontal radiation stresses m2 s2
T Near-bottom average wave period s
T* Ratio of twc/tce –
C Wave celerity m s1
cx Wave celerity x-direction m s1
cy Wave celerity y-direction m s1
cg Wave group celerity m s1
cy Wave celerity in directional (y) space m s1
cs Wave celerity in frequency (s) space m s1
c0 Turbulent concentration 1C, salinity, or kg m3
d0 Wave orbital diameter m
f Coriolis parameter s1
fw Wave friction factor –
g Gravity m s2
k Wave number ( ¼ 1/wave length) m1
kb Bottom roughness length m
kx Wave number in x-direction m1
ky Wave number in y-direction m1
k1 k2 Active layer thickness coefficients –
m Index for each sediment class –
p Pressure N m2
qbl Bedload transport rate kg m2 s1
qblx Bedload transport rate in x-direction kg m2 s1
qbly Bedload transport rate in y-direction kg m2 s1
qbl_slope Bedload slope factor –
s Vertical sigma coordinate –
s Specific gravity –
t Time s
u Velocity x-direction m s1
ub Bottom orbital velocity m s1
u Friction velocity m s1
uc Friction velocity due to currents m s1
uwc Friction velocity due to combined waves and currents m s1
u0 Turbulent velocity x-direction m s1
us Vertically varying stokes velocity x-direction m s1
 ARTICLE IN PRESS
1288 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

Table 1 (continued )

Symbol Description Dimensions

ū Depth-integrated velocity x-direction m s1

us depth-integrated stokes velocity x-direction m s1
V Velocity y-direction m s1
v0 Turbulent velocity y-direction m s1
vs Vertically varying stokes velocity y-direction m s1
v̄ Depth-integrated velocity y-direction m s1
vs Depth-integrated stokes velocity x-direction m s1
w0 Turbulent velocity s-direction m s1
ws Sediment settling velocity m s1
x Horizontal direction m
y Horizontal direction m
z Vertical elevation m
za Active layer thickness m
z0 Total bottom roughness length m
zoN Grain size bottom roughness m
zoST Sediment transport bottom roughness m
zoBF Bedform bottom roughness m
zoMIN Minimum bottom roughness m
zr Reference elevation for BBL m
F Non-dimensional bedload transport rate –
F̄ Bedload transport vector in direction of and direction perpendicular to currents –
FJ Bedload transport vector in direction of currents –
FJ1, FJ2 Bedload transport quantities in direction of currents –
F? Bedload transport vector in direction perpendicular to currents –
O Vertical velocity s-direction s1
a Roller parameter –
b Local bed slope –
g Wave height to water depth ratio –
gw Wave asymmetry factor –
g1 Linear drag coefficient –
g2 Quadratic drag coefficient –
dwbl Wave boundary layer height m
Z Wave averaged free surface elevation m
Zr Ripple wave height m
y Wave direction radians
ym Shields parameter (uses tm)
ysf Shields parameter (uses tsf) –
ysfJ Shields parameter (uses tsf in direction of currents) –
ysf? Shields parameter (uses tsf in direction perpendicular to currents) –
! Shields parameter vector in direction of and direction perpendicular to currents –
ysf
yc Critical shields parameter (uses tce) –
~
y Directional shields parameter –
k von Kármán’s constant (0.41) –
lr Ripple wave length m
r Density kg m3
r0 Reference density kg m3
rs Sediment density kg m3
rwater Water density kg m3
n Kinematic viscosity m2 s1
ny Tracer kinematic diffusivity m2 s1
s Wave frequency (relative to currents) s1
tbx Bottom stress x-direction m2 s2
tby Bottom stress y-direction m2 s2
tc Bottom stress due to currents alone m2 s2
tce Bottom critical erosion stress m2 s2
tm Mean bottom stress due to combined waves+currents m2 s2
tsx Surface stress x-direction m2 s2
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1289

Table 1 (continued )

Symbol Description Dimensions

tsy Surface stress y-direction m2 s2

tsf Total skin friction bottom stress, (maximum combined wave+current) m2 s2
tsfm Skin friction component due to form drag m2 s2
tw Bottom stress due to waves alone m2 s2
twc Combined bottom stress due to waves and currents m2 s2
f Sediment bed porosity –
fm Friction angle of sediment degrees

2003) as implemented by Warner et al. (2005) that where Rz vertically distributes the additional stress
also includes the option for surface fluxes of term due to the roller as an exponentially function
turbulence kinetic energy due to wave breaking. decaying with depth and g is the ratio of wave height
The wide choice in turbulence closures facilitates to water depth (g ¼ Hs/D), Hs is the significant wave
evaluation of the effects of turbulence parameter- height, k is the wavenumber (k ¼ 2p/L where L is
izations on model results (for example, see Wijese- wavelength), kx and ky are the wavenumber compo-
kera et al., 2003; Li et al., 2005). nents in the x and y directions and c is the wave-
We have modified ROMS to include physical propagation speed, computed as
processes that are important in nearshore regions by rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s g
adding radiation-stress terms to the momentum c¼ ¼ tanh kD (9)
equations based on Mellor (2003, 2005) where a k k
vertical coordinate transformation and phase aver- where s is the wave frequency (s ¼ 2p/T, where T is
aging are used to derive interacting current and wave period). The two options available for deter-
surface gravity wave equations. We neglect the mining the roller area (AR) are (1) to obtain AR
momentum transfer term that correlates wind- directly from the wave model or (2) compute AR
induced surface pressure fluctuations and wave based on a formulation from (Svendsen, 1984):
slope because methods to incorporate these pro- a
cesses are still being developed. The horizontal AR ¼ pffiffiffi H s LQb (10)
2
radiation-stress terms (on the rhs of Eqs. (1) and (2))
are where a is a parameter with value 0.06, and Qb is the
  fraction of breaking waves.
kx kx The vertical radiation-stress terms (last term on
S xx ¼ kE F CS F CC þ F CS F CC  F SS F CS
k2 the rhs of Eqs. (1) and (2)) are:
k x k x c2 
þ AR Rz F SS qE qðkDÞ
k L S px ¼ ðF CC  F SS Þ þ F CS ð1 þ sÞE
  2 qx qx
kx ky k x k y c2 
S xy ¼ Syx ¼ kE F F
CS CC þ AR Rz qðkDÞ
k 2 k L EF SS cothðkDÞ
  qx
ky ky 
Syy ¼ kE F F
CS CC þ F F
CS CC  F F
SS CS
F SS qE qðkDÞ
k2 S py ¼ ðF CC  F SS Þ þ F CS ð1 þ sÞE
2 qy qy
k y k y c2 
þ AR Rz (7) qðkDÞ
k L EF SS cothðkDÞ (11)
qy
where the terms in brackets are the traditional where the vertical structure functions in Eqs. (7) and
momentum flux terms due to the waves (Mellor, (11) are:
2003, 2005), and the last term is due to the surface
roller (Svendsen, 1984; Svendsen et al., 2002), with a sinhðkDð1 þ sÞÞ coshðkDð1 þ sÞÞ
F SS ¼ F CS ¼
vertical distribution expressed as sinh kD sinh kD
sinhðkDð1 þ sÞÞ coshðkDð1 þ sÞÞ
 4 F SC ¼ F CC ¼
2s cosh kD cosh kD
Rz ¼ 1  tanh (8)
g (12)
 ARTICLE IN PRESS
1290 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

and E ¼ gHs2/16 is the wave energy. The terms in where the group speed, cg, is
Eqs. (12) provide wave-induced stresses in the  
qs c 2kD
momentum equations that decay with depth. cg ¼ ¼ 1þ (18)
The momentum expressions derived by Mellor qk 2 sinhð2kDÞ
(2003, 2005) yield equations with mean (wave-phase The depth-integrated velocities are also expressed
averaged) velocities in a Lagrangian reference in a Lagrangian reference frame and are related to
frame. The Lagrangian and Eulerian reference the Eulerian reference frame by the depth-integrated
frames are related by the Stokes velocities us and Stokes velocities ūs and v̄s , with:
vs in the x and y directions, computed as
  kx E kx gAR ky E ky gAR
2kx cosh 2kDð1 þ sÞ DgAR ūs ¼ þ ; v̄s ¼ þ (19)
us ¼ Eþ , ckD k cL ckD k cL
c sinh 2kD L
  The Stokes velocities are subtracted from the
2ky cosh 2kDð1 þ sÞ DgAR Lagrangian velocities to maintain a consistent
vs ¼ Eþ (13)
c sinh 2kD L Eulerian reference frame in the model and for the
output.
where the last terms in the parentheses are the roller
contributions. Stokes velocities are subtracted from
Lagrangian velocities to maintain a consistent 2.2. Wave model
Eulerian reference frame for the entire model
dynamics. The modification of the momentum equations to
ROMS solves the equations with a mode-splitting include the effects of surface waves requires
technique (described in detail by Haidvogel et al., information on basic wave properties such as
2007) that requires depth-integrated momentum wave-energy, propagation direction, and wave-
equations. Including the radiation-stress terms, length. Other algorithms, such as the bottom-
these are: boundary modules and turbulence submodels may
qðDūÞ qðūDūÞ qðv̄DūÞ also require wave information such as wave period,
þ þ  fDv̄ bottom orbital velocity, and wave-energy dissipa-
qt qx qy
tion rate. These quantities are obtained from
qp qS xx qSxy SWAN (Booij et al., 1999). SWAN is a wave-
¼ D þ tsx  tbx   (14)
qx qx qy averaged model that solves transport equations for
wave action density N (energy density divided by
qðDv̄Þ qðv̄DūÞ qðv̄Dv̄Þ relative frequency):
þ þ þ fDū
qt qx qy qN qcx N qcy N qcs N qcy N S w
qp qS xy qS yy þ þ þ þ ¼ (20)
¼ D þ tsy  tby   (15) qt qx qy qs qy s
qy qx qy
where cx and cy are the propagation velocities in the
and continuity is x and y directions, s is the relative frequency, and y
is the wave direction. SWAN accounts for shoaling
qZ qðDūÞ qðDv̄Þ
þ þ ¼0 (16) and refraction through dependent variations in cx
qt qx qy and cy. The term S on the right-hand side is a
where the horizontal radiation-stress terms (Phillips, source/sink term representing effects of wind-wave
1969; Mellor, 2003, 2005) with roller contributions generation, wave breaking, bottom dissipation, and
based on Svendsen (1984) and Svendsen et al. (2002) nonlinear wave-wave interactions. SWAN also can
are: account for diffraction, partial transmission, and
  reflection. Specific formulations for wind input,
cg k x k x cg 1 k x k x c2 A R bottom stress, whitecapping, wave–wave interac-
Sxx ¼ E þ E  þ 2
c k 2 c 2 k L tions, etc. are described in detail in Booij et al.
2
cg k x k y k x k y c A R (2004). SWAN can be run separately and the output
S xy ¼ Syx ¼ E þ 2 used to force the hydrodynamic and sediment
c k2 k L
  routines (one-way coupling). Alternatively, SWAN
cg k y k y cg 1 ky ky c2 AR can be run concurrently with the circulation model
Syy ¼ E 2
þE  þ 2 (17)
c k c 2 k L with two-way coupling, whereby currents influence
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1291

the wave field and waves affect the circulation (see ROMS and SWAN must both use the same grid
Section 2.4). in our current implementation. The grid may be
curvilinear. SWAN depths must coincide with
2.3. Model domains ROMS r-points. ROMS has wetting and drying
capabilities. The algorithm identify cells with water
ROMS is discretized in horizontal dimensions depths less than a user-specified value, and prevents
with curvilinear orthogonal Arakawa C grid outward flux of water from those cells, a process
(Arakawa, 1966) with x(x-) and Z(y-) coordinates called cell flux blocking (Casulli and Cheng, 1992).
(Fig. 1). Grid-cell centers are termed r points and Flux of water onto cells is always permitted. The
are the locations of tracer concentrations, water same minimum depth can be specified in SWAN to
depth, and sea level. Velocities are computed on the exclude those points during wave computations.
grid-cell faces. The grid can be rectilinear with
constant or varying grid spacings, or curvilinear to 2.4. Model coupling
allow focusing of the mesh to specific areas, for
example to accommodate land-sea boundaries. The We used the Model-Coupling Toolkit (MCT;
model also allows land-sea masking to identify http://www-unix.mcs.anl.gov/mct/; Larson et al.,
regions of dry areas. The model uses a stretched 2005; Jacob et al., 2005) to couple ROMS with
vertical s-coordinate system, which are similar to SWAN (Warner et al., in press). MCT is an
sigma coordinates with additional flexibility: the open-source software library, distributed as a set
layers need not be a fixed percentage of the water of Fortran90 modules for constructing a coupled
column (see Haidvogel et al., 2000). Vertical model system from individual component models.
resolution can be adjusted to allow increased Each component model has its own grid and runs
resolution near the surface and bottom boundaries. on its own set of processors. The MCT provides
The bed model comprises a user-defined number of protocols for decomposition and allocation of
layers that extend vertically into the sea floor. See model grids among different processors, efficient
Section 3.2 for a detailed description. transfer of data fields between the different models,

Face
N
Layer
Zos

N N-1 Variable Grid Location point type

u u point
N-1 N-2 v point
v
w, KM, KH, k, w point
2
ψ, ε, P, B

2 1 tracer rho point

1 0
Zob

1 v
1

2 2 u u
Δη
Nbed-1

Nbed Nbed v
Δξ
Vertical grid section. Horizontal grid section.

Fig. 1. A vertical section of the ROMS grid showing water column and bed layers.
 ARTICLE IN PRESS
1292 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

and interpolation algorithms for the data fields that resuspension and transport, providing a feedback
are transferred. SWAN sends to ROMS arrays of from the sediment dynamics to the hydrodynamics.
wave height, wavelength, average wave periods at The bed layers are modified at each time step to
the surface and near the bottom, wave-propagation account for erosion and deposition (Fig. 2) and
direction, near-bottom orbital velocity, and wave-e- track stratigraphy. At the beginning of each time
nergy dissipation rate. ROMS provides to SWAN step, an active-layer thickness za is calculated based
arrays of water depth, sea-surface elevation, and on the relation of Harris and Wiberg (1997):
current velocity. Data exchange between SWAN
za ¼ max½k1 ðtsf  tce Þr0 ; 0 þ k2 D50 (21)
and ROMS occurs at user-defined synchronization
intervals. The frequency of data exchange depends where tsf is bottom skin-friction stress due to
on the application. If the exchanged fields fluctuate combined maximum wave and current interaction;
rapidly, more frequent synchronization is required. tce is the critical stress for erosion; and the overbar
However, data exchange increases run time, so indicates this is averaged over all sediment classes;
experience is required to determine the optimum D50 is the median grain diameter of surface
synchronization interval for each application. sediment; and k1 and k2 are empirical constants
(values of 0.007 and 6.0, respectively). The thickness
3. Sediment algorithms and implementation of the top bed layer has a minimum thickness
equivalent to za. If the top layer is thicker than za,
3.1. Sediment classes no action is required. If the top layer is less than za
thick, then the top layer thickness is increased by
The model is capable of representing an unlimited entraining sediment mass from deeper layers until
number of user-defined sediment classes. Each class the top layer thickness equals za. If sediment from
has fixed attributes of grain diameter, density, deeper than the second layer is mixed into the top
settling velocity, critical shear stress for erosion, layer, the bottom layer is split, enforcing the
and erodibility constant. These properties are used constant number of layers and conserving sediment
to determine bulk properties of each bed layer. Two mass.
classes of sediments (non-cohesive and cohesive) are Each sediment class can be transported by
included in the model framework, but the algo- suspended-load and/or bedload (described in Sec-
rithms governing cohesive sediment dynamics are tions 3.3 and 3.4). Suspended-load mass is ex-
still being developed and are not discussed here. changed vertically between the water column and
the top bed layer. Mass of each sediment class
3.2. Sediment bed available for transport is limited to the mass
available in the active layer. Bedload mass is
The sediment bed is represented by three-dimen- exchanged horizontally between the top layers of
sional arrays with a user-specified, constant number the bed. Mass of each sediment class available for
of layers beneath each horizontal model cell (Fig. 1). transport is limited to the mass available in the top
Each cell of each layer in the bed is initialized with a layer.
thickness, sediment-class distribution, porosity, and Suspended-sediment that is deposited, or bedload
age. The mass of each sediment class in each cell can that is transported into a computational cell, is
be determined from these values and the grain added to the top bed layer. If continuous deposition
density. The age property tracks the time that results in a top layer thicker than a user-defined
deposition last occurred in that layer. The bed threshold, a new layer is provided to begin
framework also includes two-dimensional arrays accumulation of depositing mass. The bottom two
that describe the evolving properties of the seabed, layers are then combined to conserve the number of
including bulk properties of the surface layer layers. After erosion and deposition have been
(active-layer thickness, mean grain diameter, mean calculated, the active-layer thickness is recalculated
density, mean settling velocity, mean critical stress and bed layers readjusted to accommodate it. This
for erosion) and descriptions of the subgrid-scale step mixes away any very thin layer (less than the
morphology (ripple height and wavelength). These active-layer thickness) of newly deposited material.
properties are used to estimate bed roughness in the Finally the surficial sediment characteristics, such as
bottom stress calculations. The bottom stresses are D50, ripple geometry, etc., are updated and made
then used by the sediment routines to determine available to the bottom stress calculations.
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1293

Active layer thickness (Harris and Wiberg, 1997).

za = k1(τsf - τce) + k2 D50

Erosion. (τsf > τce)

Entrain sediment from lower layers so that
surface layer is at least za thick. Split
bottom layer. Erode from surface layer.
For each sediment class
Active
erosion_flux =
layer
thickness
dt*Es*(1-poro)*frac*(τsf/τce -1),
MIN
ρs*(1-poro)*frac*za+dep_flux

Deposition.
Create new layer if
Active
deposition > user defined thickness.
layer Mix surface layer to be at least za thick.
thickness Combine bottom layer. For each
sediment class
∂C ∂C
= − ws
∂t ∂z

Fig. 2. Distribution of vertical layers in bed model. During erosion top layer thickness is increased to meet active layer thickness.
Deposition creates a new layer if timing and thickness criteria are met. Total number of layers must be constant, often requiring a merge or
splitting of bottom cells.

3.3. Suspended-sediment transport because it allows (1) reuse of the routines for
advection and diffusion of water-column tracers, (2)
Temperature, salinity, and sediment suspended in use of high-order numerical schemes for vertical
the water column are transported by solving the settling, and (3) formulation of the flux conditions
advection–diffusion equation (5). However for to ensure conservation of sediment in both bottom
suspended-sediment, an additional source/sink term sediments and the water column.
is added for vertical settling and exchange with the The vertical advection algorithm includes a piece-
bed as wise parabolic method (Colella and Woodward,
1984) and a weighted essentially non-oscillatory
qws;m C m
C source;m ¼  þ E s;m (22) (WENO) scheme (Liu et al., 1994). This method
qs integrates depositional flux over multiple grid cells,
where ws,m is the vertical-settling velocity (positive so it is not constrained by the CFL criterion. Zero-
upwards), Es,m is the erosion source (defined below), flux boundary conditions are imposed at the surface
and m equals one through the number of classes. and bottom in the vertical diffusion equation. The
The model solves each term of Eq. (5) indepen- source or sink term in the advection equation
dently, in the sequence: vertical settling, source/sink, represents the net of downward settling and upward
horizontal advection, vertical advection, vertical flux of eroded material and is only applied to the
diffusion, and finally horizontal diffusion. Separa- bottom computational cell. Erosional flux is para-
tion of these calculations has practical advantages meterized following Ariathurai and Arulanandan
 ARTICLE IN PRESS
1294 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

(1978) as v locations) and then interpolated to cell centers

tsf  tce;m (r points). The bedload transport vectors are
E s;m ¼ E 0;m ð1  fÞ ; when tsf 4tce;m partitioned into x and y components based on the
tce;m
magnitude of the bed shear stress as
(23)
tbx tby
where Es is the surface erosion mass flux qblx ¼ qbl ; qbly ¼ qbl (28)
tsf tsf
(kg m2 s1), E0 is a bed erodibility constant
(kg m2 s1), j is the porosity (volume of voids/
total volume) of the top bed layer, and m is an index 3.4.2. Soulsby and Damgaard
for each sediment class. The erosional flux for each The Soulsby and Damgaard (2005) formulae
sediment class is also limited by the availability of account for the combined effects of mean currents
that class in the top layer of the bed model. and asymmetrical waves on bedload flux. Their
formulations are based on numerical integration,
3.4. Bedload transport over a wave cycle, of the non-dimensional transport
equation
This version of ROMS implements two methods " ! #
! 0:5 ysf
for computing bedload transport: (1) the Meyer- F ¼ max A2 y ðysf  yc Þ ; 0 (29)
Peter Müeller (1948) formulation for unidirectional ysf
flow and (2) the formulae of Soulsby and Damgaard ! !
(2005) that accounts for combined effects of where F and ysf are vectors with components in
currents and waves. The formulae depend on the the direction of the mean current and in the
characteristics of individual sediment classes, in- direction perpendicular to the current, e.g.,
! ! !
cluding size D, density rs, specific density in water F ¼ ðFk ; F? Þ, ysf ¼ ðysf k ; ysf ? Þ,ysf ¼ j ysf j, yc is
s ¼ rs/r, and critical shear stress tc. Non-dimen- the critical Shields parameter (i.e. Eq. (26) with
sional transport rates F are calculated for each tce), and A2 ¼ 12 is a semi-empirical coefficient. The
sediment class and converted to dimensional bed- implementation of this method requires computa-
load transport rates qbl using tion of transport rates in the directions parallel and
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi perpendicular to the currents as
qbl ¼ F ðs  1ÞgD350 rs (24)
Fk ¼ max½Fk1 ; Fk2  (30)
These are horizontal vector quantities with
where
directions that correspond to the combined bed-
stress vectors. Fk1 ¼ A2 y0:5
m ðym  yc Þ (31)

3.4.1. Meyer-Peter Müeller Fk2 ¼ A2 ð0:9534 þ 0:1907 cos 2fÞy0:5

w ym
The Meyer-Peter Müeller (1948) formulation is þ A2 ð0:229gw y1:5
w cos fÞ (32)
1:5
F ¼ max½8ðysf  yc Þ ; 0 (25)
0:1907y2w
where F is the magnitude of the non-dimensional F? ¼ A2 ðym sin 2f þ 1:2gw yw sin fÞ
transport rate for each sediment class, ysf is the non- y3=2 1:5
w þ 1:5ym
dimensional Shields parameter for skin stress (33)
tsf where ym is the mean Shields parameter (i.e. Eq. (26)
ysf ¼ (26)
ðs  1ÞgD50 with tm) and tm is
yc ¼ 0.047 is the critical Shields parameter, andtsf is  1:5 !
tw
the magnitude of total skin-friction component of tm ¼ tc 1 þ 1:2 (34)
tw þ tc
bottom stress computed from
and tc is the bottom stress from the currents only, tw
tsf ¼ ðt2bx þ t2by Þ0:5 (27)
is the bottom stress from the waves only calculated
where tbx and tby are the skin-friction components in the bottom-boundary layer routines (see below).
of bed stress, from currents alone or the maximum The asymmetry factorgw is the ratio between
wave–current combined stress, in the x and y the amplitude of the second harmonic and the
directions. These are computed at cell faces (u and amplitude of the first harmonic of the oscillatory
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1295

wave stress. Following the suggestion of Soulsby implementation, bedload fluxes, erosion, and de-
and Damgaard (2005), we estimate the asymmetry position rates are multiplied by a scale factor. A
factor using Stokes second-order theory (e.g., scale factor with a value of one has no effect, and
Fredsøe and Deigaard, 1992) and constrain it to values greater than one accelerate the bed response.
be less than 0.2. The non-dimensional fluxes (Eqs. For bedload transport, the scale factor is multiplied
(30) and (33)) are rotated into x and y directions against the bedload transport rates. For suspended-
using the directions for mean current and waves and load transport, the scale factor multiplies the
dimensionalized with Eq. (24) to yield values for qblx exchange of sediment (erosive or depositional flux)
and qbly for each sediment class. at the bed-water interface. The magnitude of
sediment concentrations in the water column are
3.4.3. Bed slope not modified—just the exchange rate to and from
Computed bedload rates are modified to account the bed. For both bedload and suspended load,
for local bed slope following Lesser et al. (2004) sediment is limited in availability as described
with a bed slope term: previously, based on the true amount of sediment
tan jm mass (not multiplied by the scale factor). This
qbl_slope ¼ (35) morphological scale factor method works well for
ðtan jm  tan bÞ cos b
systems with unlimited sediment in the bed. How-
where the local bed slope b ¼ tan1 ðdz=dxa Þ is ever, it can generate extra sediment in systems with
evaluated for each direction of transport with a limited supplies of bed sediment. This occurs when
positive value of dz/dxa in the downslope direction, the amount of sediment to be eroded is limited by
and where jm is the friction angle of the sediment the amount available and application of the
(taken as 331). The bedload magnitudes are then morphological scale factor cannot remove the
multiplied by qbl_slope. scaled amount of sediment from the bed. Subse-
quent deposition does place a scaled amount of
3.4.4. Bedload numerics sediment on the bed thus creating new mass in the
Bedload fluxes are computed at grid-cell centers bed. Other approach (Lesser et al., 2004) is to limit
and limited by the availability of each sediment class the amount of sediment fluxed to the water column
in the top layer. Fluxes are translated to cell faces in these situations. This gives unrealistically low
using a simple upwind approach (e.g., Lesser et al., sediment concentrations, but conserves bed sedi-
2004): the bedload flux at each cell face is set to the ment.
bedload rate at the upwind cell center. Flux
differences are then used to determine changes of 3.6. Sediment density effects
sediment mass in the bed at each grid cell.
Effects of suspended sediment on the density field
3.5. Morphology are included with terms for the weight of each
sediment class in the equation of state for seawater
The bed model accounts for changes in sea floor density as
elevation resulting from convergence or divergence
in sediment fluxes. These morphological changes X
N sed
Cm
r ¼ rwater þ ðrs;m  rwater Þ. (36)
can have significant influence on flow and transport r
m¼1 s;m
when they are larger than a few percent of the water
depth. The morphological changes are accounted This enables the model to simulate processes
for by equating the bottom-boundary condition of where sediment density influences hydrodynamics,
the vertical velocity to the rate of change of such as density stratification and gravitationally
elevation of the sea floor. This method is completely driven flows.
mass conserving and retains tracer constancy
preservation. 3.7. Bottom stress calculations
A morphological scale factor is also provided to
allow an increased rate of morphological change, Reynolds stresses, production and dissipation of
which can be useful for simulating evolution over turbulent kinetic energy, and gradients in velocity
long time periods. Strategies for morphological and suspended-sediment concentrations vary over
updating are described by Roelvink (2006). In our short vertical distances, especially near the bed, and
 ARTICLE IN PRESS
1296 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

can be difficult to resolve with the vertical grid simple drag-law approach, we discuss ssw_bbl in
spacing used in regional-scale applications. ROMS detail.
provides algorithms to parameterize some of these The linear and quadratic drag-coefficient meth-
subgrid-scale processes in the water column and ods depend only on velocity components u and v in
in the bottom-boundary layer (BBL). Treatment the bottom grid cell and constant, spatially-uniform
of the BBL is important for the circulation coefficients g1 and g2 specified as input:
model solution because it determines the stress  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
exerted on the flow by the bottom, which enters the tbx ¼ g1 þ g2 u2 þ v2 u (38)
Reynolds-averaged Navier–Stokes equations as a  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
boundary conditions for momentum in the x and y tby ¼ g1 þ g2 u2 þ v2 v (39)
directions:
where g1 is the linear drag coefficient and g2 is the
qu qv quadratic drag coefficient. The user can choose
KM ¼ tbx ; K M ¼ tby (37)
qs qs between linear or quadratic drag by setting one of
Determination of the BBL is even more impor- these coefficients to zero. The bottom stresses
tant for the sediment-transport formulations be- computed from these formulae depend on the
cause bottom stress determines the transport rate elevation of u and v (computed at the vertical mid-
for bedload and the resuspension rate for suspended elevation of the bottom computational cell). There-
sediment, as discussed in Sections 3.3 and 3.4. fore, in this s-coordinate model, the same drag
ROMS implements either of two methods for coefficient will be imposed throughout the domain
representing BBL processes: (1) simple drag-coeffi- even though the vertical location of the velocity is
cient expressions or (2) more complex formulations different.
that represent the interactions of wave and currents The logarithmic formulation assumes that flow in
over a moveable bed. The drag-coefficient methods the BBL has the classic vertical logarithmic profile
implement formulae for linear bottom friction, defined by a shear velocity u* and bottom roughness
quadratic bottom friction, or a logarithmic profile. length z0 as
The other, more complex methods, implement some  
u z
of the many wave–current BBL models (e.g., juj ¼ ln (40)
k z0
Jonsson and Carlsen, 1976; Smith, 1977; Grant pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and Madsen, 1979; Madsen, 1994; Styles and Glenn, speed juj ¼ u2 þ v2 , friction velocity
wherepffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2000) and couple them with calculations of bottom u ¼ jtbx j þ jtby j, z is the elevation above the
roughness. ROMS offers three methods that imple- bottom (vertical mid-elevation point of the bottom
ment slightly different combinations of algorithms cell), k ¼ 0.41 is von Kármán’s constant, and z0 is a
for the wave–current interactions and moveable bed constant (but possibly spatially varying) bottom
roughness. The first method (sg_bbl) is based on the roughness length (m). Kinematic stresses are calcu-
wave–current algorithm and the ripple geometry lated as:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and moveable bed roughness of Styles and Glenn k 2 u u2 þ v 2
(2000, 2002). The second method (mb_bbl) uses tbx ¼ (41)
ln2 ðz=z0 Þ
efficient wave–current BBL computations devel-
oped by Soulsby (1995) in combination with pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
k2 v u2 þ v2
sediment and bedform roughness estimates of Grant tby ¼ (42)
and Madsen (1982), Nielsen (1986) and Li and ln2 ðz=z0 Þ
Amos (2001). These algorithms and an example of The advantage of this approach is that the velocity
their use on the Southern California continental and the vertical elevation of that velocity are used in
shelf are described by Blaas et al. (2005). The third the equation. Because the vertical elevation of the
method (ssw_bbl) implements either the wave–cur- velocity in the bottom computational cell will vary
rent BBL model of Madsen (1994) or that of Styles spatially and temporally, the inclusion of the eleva-
and Glenn (2000) along with moveable bed routines tion provides a more consistent formulation.
proposed by Wiberg and Harris (1994); Harris and More complex routines are required to simulate BBL
Wiberg (2001). The differences in approach among processes in the presence of waves and mobile sedi-
these routines are small, but they can produce ment. The short (order 10-s) oscillatory shear of wave-
significantly different results. After reviewing the induced motions in a thin (a few cm) wave-boundary
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1297

layer produces turbulence and generates large instanta- where d0 ¼ ubT/p is the wave-orbital dia-
neous shear stresses. The turbulence enhances momen- meter. When transport stage is below the
tum transfer, effectively increasing the coupling threshold for sediment transport (T* ¼ twc/
between the flow and the bottom and increasing the tceo1), ripple dimensions from the previous
frictional drag exerted on the mean flow, averaged over time step are retained.
many wave periods. The large instantaneous shear (2) Roughness lengths associated with grain rough-
stresses often dominate sediment resuspension and ness z0N, sediment transport z0ST, and bedform
enhance bedload transport. Sediment transport can roughness length (ripples) z0BF are estimated as
remold the bed into ripples and other bedforms, which z0N ¼ 2:5D50 =30 (44)
present roughness elements to the flow. Bedload
transport can also induce drag on the flow, because
momentum is transferred to particles as they are T
removed from the bed and accelerated by the flow. z0ST ¼ aD50 a1 (45)
1 þ a2 T 
Resuspended sediments can cause sediment-induced
stratification and, at high concentrations, change the
effective viscosity of the fluid. z0BF ¼ ar Z2r =lr (46)
The BBL parameterization implemented in ROMS where the sediment-transport coefficients are
requires inputs of velocities u and v at reference a ¼ 0.056, a1 ¼ 0.068, and a2 ¼ 0:0204
elevation zr, representative wave-orbital velocity lnð100D250 Þ þ 0:0709 lnð100D50 Þ (Wiberg and
amplitude ub, wave period T, and wave-propagation Rubin, 1989) with the bedform roughness D50
direction y (degrees, in nautical convention). The expressed in meters, and where ar is a coefficient
wave parameters may be the output of a wave model that may range from 0.3 to 3 (Soulsby, 1997).
such as SWAN or simpler calculations based on Grant and Madsen (1982) proposed ar ¼ 27.7/
specified surface wave parameters and should repre- 30 but we use as a default value ar ¼ 0.267
sent the full spectrum of motion near the bed (cf. suggested by Nielsen (1992). The roughness
Madsen, 1994; Wiberg and Sherwood, this issue). lengths are additive, so subsequent BBL calcula-
Additionally the BBL models require bottom sedi- tions use z0 ¼ max½z0N þ z0ST þ z0BF ; z0MIN ,
ment characteristics (median grain diameter D50, where z0MIN allows setting a lower limit on
mean sediment density rs, and representative settling bottom drag (default z0MIN ¼ 5e5 m).
velocity ws); these are based on the composition of (3) Initial estimates of (kinematic) bottom stresses
the uppermost active layer of the bed sediment based on pure currents tc ( ¼ tb) and pure waves
during the previous time step. Bed stresses associated tw (tb ¼ 0) are made as follows.
with mean current above the wave-boundary layer tb,
wave motions tw, and maximum vector sum of the ðu2 þ v2 Þk2
tc ¼ (47)
two twc from the previous time step are used as initial ln2 ðz=z0 Þ
estimates for the next time level.
and tw ¼ 0.5fwub2, where fw is the Madsen
The procedure for bottom-boundary layer calcu-
(1994) wave-friction factor, which depends on
lations in ssw_bbl is as follows:
the ratio of the wave-orbital excursion ampli-
tude to the bottom roughness length Ab/kb,
(1) Ripple height Zr and wavelength lr are calcu-
where Ab ¼ ubT/(2p) and kb ¼ 30z0:
lated using information from the previous time
8 9
step and the Malarkey and Davies (2003) >
> 0:3; Ab =kb p0:2 >
>
< =
implementation of the Wiberg and Harris fw ¼ expð8:82 þ 7:02ðAb =kÞ0:078 Þ; 0:2oAb =kb p100
>
> >
>
(1994) formulation, which is valid for wave- : expð7:30 þ 5:61ðAb =kÞ0:109 Þ; Ab =kb 4100 ;
dominated conditions. They approximate ripple
(48)
wavelength as 535D50 and ripple steepness as
"   2
Zr d0 (4) The pure currents and pure wave limits are used
¼ exp 0:095 ln
lr Zr as initial estimates for calculations towards
    consistent profiles for eddy viscosity and velo-
d0
þ0:442 ln  2:28 (43) city between z0 and zr, using either the model of
Zr Madsen (1994) or Styles and Glenn (2000). Both
 ARTICLE IN PRESS
1298 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

of these models assume eddy viscosity profiles and act as agents for sediment resuspension and
pffiffiffiffiffiffiffi
scaled by uwc ¼ twc in the wave-boundary bedload transport.
pffiffiffiffiffi
layer (WBL) and uc ¼ tb in the current
boundary layer, calculated as
( 4. Examples
kuwc z; zodwbl
KM ¼ (49)
kuc z; z4dwbl In this section we provide four examples that
where dwbl is the thickness of the WBL, which highlight the capabilities of the sediment-transport
scales as u*wcT/(2p). twc represents the maximum model. Example 1 demonstrates the ability of the
vector sum of wave- and current-induced stress, model produce a classic suspended-sediment profile
but the tb is influenced by the elevated eddy and illustrates the effect of varying vertical grid
viscosity in the WBL, and must be determined resolution and turbulence submodels. Example 2
through an iterative process. The shape and demonstrates the morphology component of the
elevation of the transition between these profiles model in a simulation of a lab experiment with a
and other details differ among the two models, migrating trench. Example 3 demonstrates the
but both the models of Madsen (1994) or Styles impact of dynamic coupling for wave–current
and Glenn (2000) return values for the hor- interactions at a tidal inlet. Example 4 is a realistic
izontal vectors tb, tw, and twc. The parameter tb application with complex bathymetry that demon-
is the mean (over many wave periods) stress strates transport and sorting of multiple sediment
used as the bottom-boundary condition in the classes.
momentum equations, and twc is the maximum
instantaneous stress exerted over the bottom by 4.1. Example 1: Steady uniform open-channel flow
representative waves and currents.
(5) When ripples are present, twc is a combination Example 1 exercises the models ability to simulate
of form drag, which does not directly contribute vertical profiles of suspended-sediment concentra-
to sediment transport, and skin friction, which tions (no bedload) with varying vertical grids and
does. The next step in the BBL calculations is to turbulence closures. The simulation represents
estimate the skin-friction component of twc suspended-sediment transport for steady horizon-
using the ripple dimensions and a bedform tally uniform flow in a straight rectangular channel,
drag-coefficient approach (Smith and McLean, modified slightly from Warner et al. (2005). The
1977; Wiberg and Nelson, 1992), as follows. channel is placed on a constant slope of 4  105 m/m
" and a depth-mean velocity of 1 m s1 is imposed at
 2 #1
Zr Zr both the upstream and downstream ends. The
tsfm ¼ twc 1 þ 0:5C dBF ln  1
lr k2 ðz0N þ z0ST Þ water-surface elevation is allowed to vary along
the length of the channel. Radiation conditions for
(50)
the free surface and 3D momentum at both ends
where CdBFE0.5 is a bedform drag coefficient allow waves to propagate out of the domain. An
for unseparated flow (Smith and McLean, unlimited supply of sediment is available in the bed.
1977). Additional details are listed in Table 2.
(6) Finally, because shear stress varies between Numerical simulations of vertical suspended-
ripple crests and troughs, an estimate of the sediment profiles can be sensitive to the number
maximum shear stress at the crests tsfm is and placement of vertical grid levels. As the number
calculated for use in sediment-transport algo- of vertical grid levels increases, the gradient of
rithms as suspended-sediment near the bed is better resolved.
  Effects of changing the number of vertical levels and
Zr
tsf ¼ tsfm 1 þ 8 (51) resolution are evaluated using the k–e turbulence
lr
closure. The number of vertical levels is varied with
In summary, the more advanced BBL rou- values of 10, 20, 40, and 80 evenly spaced cells, and
tines calculate current and wave-boundary layer 10 and 20 cells using the stretching parameters of
bottom stresses under the combined influence of ys ¼ 3, yb ¼ 1, and Tcline ¼ 0 (see Haidvogel et al.,
wave, currents, and mobile sediments. These 2000). Vertical profiles of suspended-sediment con-
stresses directly influence flow near the bottom verge with 40 or more evenly spaced cells (Fig. 3).
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1299

Table 2 equation k-e closure, the original Mellor and

Model parameters for test case 1—open-channel flow Yamada (1982) level 2.5 closure (MY25) and an
Model parameter Variable Value
analytical expression (ANA) obtained by stipulating
a parabolic shape to the eddy viscosity profile:
Length, width, depth Xsize, Esize, 10 000, 100, 10 m  z
depth K M ¼ ku z 1  (52)
Number of grid spacings Lm, Mm, Nm 100, 10, 20 D
(+variable) where u* is the friction velocity, z is the distance
Bottom roughness Zob 0.0053 m above the bed, and D is the water depth (10 m). The
Time step dt 30 s
friction velocity is calculated by substituting the log
Simulation steps Ntimes 5000
Settle velocity ws 1.0 mm s1 law (u(z)/u* ¼ 1/k ln(z/z0) ) into the depth-mean
Erosion rate E0 5  105 kg m2 s1 flow equation and integrating over the entire depth
Critical stresses tce 0.05 N m2 of flow, yielding
Porosity j 0.90
Bed slope S0 4  105
kū
u ¼ (53)
Inflow/outflow boundary ū 1 m s1 lnðD=z0 Þ  1 þ z0 =D
condition
where ū is the depth-mean velocity, and parameters
in Table 2 produce u* ¼ 0.0625 m s1. The eddy
diffusivity in the model is determined from the
Comparison of vertical resolutions turbulent Prandtl number (Pr; ratio of eddy
0
viscosity/eddy diffusivity) and, for neutrally stable
10 even flow, Pr ¼ 0.39/0.49 ¼ 0.80 (Kantha and Clayson,
20 even
1994; Warner et al., 2005). Therefore the algebraic
-2 eddy diffusivity is KH ¼ KM/0.80 (Table 3).
40 even
The slope of the free surface should equal the
80 even
bottom slope in steady uniform open-channel flow,
Channel depth (m)

10 stretched producing a momentum balance of

-4 20 stretched
qZ
t=r ¼ u2 ¼ g D (54)
qx
-6 This theoretical balance is (0.0625 m s1)2 ¼
(9.81 m s2)(0.00004)(10 m). The modeled balance
depends on the calculated bottom shear stress and,
therefore, on the details of the turbulence closure.
-8 The free-surface slopes generated using ANA and
k–e are very nearly equal to 4  105 so Eq. (54)
holds with approximately the values calculated
-10 above. The slope calculated using the MY25 is
0.2 0.3 0.4 0.5 closer to 3  105, which balances u* of
Sediment concentration (kg m-3) 0.0544 m2 s2. The reduced shear near the bed with
MY25 is apparent in the velocity profiles and results
Fig. 3. Effect of vertical resolution on computed profiles of in less mixing and lower sediment concentrations
sediment concentration. Increasing the number of evenly
distributed cells shows convergence to a steady profile. Simula-
(Fig. 4). This behavior is a result of the wall
tion with 20 levels using increased boundary layer resolution is function used in the original MY25 closure, but
consistent with simulation using 80 evenly distributed layers. results consistent with ANA and k-e can be
obtained using the alternative wall function pro-
The same profile can be obtained with 20 levels posed by Blumberg et al. (1992); see Warner et al.
when the vertical stretching parameters are used to (2005).
increase resolution near the bed (Fig. 3).
Simulations with different turbulence closures 4.2. Example 2: Trench migration
produce significantly different profiles of velocity,
eddy diffusivity, and sediment concentration This example tests the sediment-transport compo-
(Fig. 4). Results were obtained with the two- nents of bedload (Meyer-Peter Müeller formulation),
 ARTICLE IN PRESS
1300 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

0 0 0

Channel depth (m) -2 -2 -2

analytical
-4 k−ε -4 -4
MY25

-6 -6 -6

-8 -8 -8

-10 -10 -10

0 0.5 1 0 0.04 0.08 0.1 0.2 0.3 0.4
Velocity (ms-1) Eddy diffusivity (m2s-1) Suspended sediment
concentration (kgm-3)

Fig. 4. Vertical profiles of velocity, eddy diffusivity, and sediment concentration for three of turbulence closure options of analytical
parabolic expression, k–e, and MY25. Simulations used 20 vertical stretched levels.

Table 3 flow (left to the right in Fig. 5) as the upstream end

Results for open-channel flow test case is filled, and the downstream end eroded.
qZ/qx u* (m2 s2)
Computed and observed velocity and suspended-
sediment profiles at five locations, and the initial
Calculated 4.00e5 0.0625 and final bed elevations are compared in Fig. 5. The
ANA 4.21e5 0.0643 velocity profiles are in good agreement with the
k–e 3.98e5 0.0626
measurements. The deceleration of near-bottom
MY25 3.00e5 0.0544
flow in the middle of the trench is captured well.
Suspended-sediment profiles at the upstream and
downstream ends match the observations but, in the
suspended-sediment load, and morphologic evolu- trench, calculated concentrations of suspended-
tion by simulating the laboratory experiment of van sediment deviate slightly from the observed profiles.
Rijn (1987), also described in van Rijn (1993). The At the end of the simulation, the modeled trench has
experimental and model setup (Table 4) consist of migrated as far as the observed trench, but has
flow along a 30-m straight channel with a vertical not filled as much (compare red and blue lines in
depression (trench) incised in the mobile sand bed Fig. 5).
(Fig. 5). The bed material is well-sorted fine sand
(D50 ¼ 140 mm). Flow is steady with a depth-mean 4.3. Example 3: Tidal inlet wave– current coupled
velocity of 0.51 m s1. In the model simulations, system
flow and suspended-sediment are allowed to reach
steady state before morphologic evolution is in- This example demonstrates wave–current cou-
itiated, there is an unlimited supply of available pling and includes wave-induced forcing by the
sediment, and the k-e turbulence-closure model is radiation-stress terms. The domain is a rectangular
used. basin 15 km in width and 14 km long, with a
As the flow travels into the deeper water of the uniform initial depth of 4 m (Table 5). The northern,
trench, flow velocity and bottom stress decrease. western, and eastern edges are open with radiation
Sediment begins to settle out of suspension, and boundary conditions. Along the center of the
bedload transport converges, so sediment accumu- domain is a wall with a centered 2 km wide inlet.
lates at the upstream end of the trench. At the The model is forced by oscillating the water level on
downstream end of the trench, erosion occurs as the northern edge with a tidal amplitude of 1 m.
depth decreases, flow accelerates, and transport Waves are also imposed on the northern edge with a
diverges. The trench migrates in the direction of the wave height of 1 m, directed to the south with a
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1301

period of 10 s. The model is run with two config- inlet currents (Fig. 6). At the peak of the ebb tide
urations: (1) one-way coupled with wave informa- the combined wave–current bottom stresses are
tion passed to the circulation model and (2) two- maximum near the location of maximum currents.
way fully coupled. The model hydrodynamics were Bathymetric evolution produces a flood and a larger
simulated for a period of 2 days with a morphologic ebb shoal. By contrast, the two-way coupled model
scale factor of 10, simulating a 20-day period. results show greatly increased wave heights in front
In the one-way coupled system, wave heights of the inlet as the approaching wave interacts with
evolve to a steady state, decreasing southward an opposing current. The increased wave heights
toward the inlet and showing no effect from the create combined bottom stresses that are greater
than the one-way coupled system, and the peak
Table 4 bottom stresses are located near the maximum wave
Model parameters for test case 2—migrating trench heights. The morphology evolves a stronger ebb
shoal due to the higher stresses and the shoal is
Model parameter Variable Value displaced slightly further seaward.
Length, width, Xsize, 30, 0.5, 0.39 m
depth Esize, 4.4. Example 4: Evolution of surficial sediment
depth distribution in Massachusetts Bay
Number of grid Lm, Mm, 300, 4, 20
spacings Nm
Bottom roughness Zob 0.000833 m Here we highlight the sediment-transport cap-
Time step dt 0.05 s abilities of the model and its ability to simulate the
Simulation steps Ntimes 30 000 initial (no morphology), transport of a mixed grain size bed and the
12000 with morphology evolution of the sea floor sediment grain size
Morphology factor morph_fac 0 for initial, 90 for morph
distribution. A detailed description is provided in
Settle velocity ws 11.0 mm s1
Erosion rate E0 0.35  102 kg m2 s1 Warner et al. (in press). In this simulation the
Critical stresses tcd, tce 0.11 N m2 sediment bed was initialized with 10 vertical levels
Porosity j 0.40 with the top 6 layers at 0.01 m thick and the bottom
Bed slope S0 4  104 4 at 0.10 m thick. All layers had a porosity of 0.50
Inflow/outflow ū 0.51 m s1
and initial spatially-uniform distributions of sedi-
boundary
condition ments with 7 equal fractions of grain size ranging
from 7 phi (fine silt) to 1 phi (coarse sand) with

measured velocity modelled velocity

measured sediment concentration modelled sediment concentration
measured final bed elevation modelled final bed elevation
0

-0.1
depth in channel (m)

-0.2

-0.3

-0.4
initial bed elevation

-0.5

12 14 16 18 20 22
distance along channel (m)

Fig. 5. Migrating trench test case showing initial (black line), final measured (cyan), and final modeled (red) bed elevations. Vertical
profiles of measured and modeled suspended-sediment concentration and velocity are compared.
 ARTICLE IN PRESS
1302 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

Table 5 critical shear stresses ranging from 0.022 to 0.27 Pa

Model parameters for test case 3—tidal inlet respectively. The simulation had a tidal forcing with
Model parameter Variable Value
7 main constituents along the open boundary. After
a spin-up time period, forcing with realistic wind
Length, width, depth Xsize, Esize, 15,000, 14,000, 4.0 m stress and wave fields were imposed to simulate an
depth 8-day December 1992 storm. This storm was
Number of grid Lm, Mm, Nm 75, 70, 10
repeated 10 times with an intervening 1-day period
spacings
Bottom roughness zob 0.015 m to allow sediment to settle out of the water column.
Time step dt 10 s The model time step was 30 seconds and hourly
Simulation steps Ntimes 17280 steps (2 day) results were saved.
Morphology factor morph_fac 10 ( ¼ 20 day scaled The repeating storm simulation generated realis-
simulation)
tic patterns of bottom stress, sediment resuspension,
Settle velocity ws 11.0 mm s1
Erosion rate E0 5  103 kg m2 s1 and bathymetric change (Fig. 7). The instantaneous
Critical stresses tcd, tce 0.10 N m2 bottom stress (maximum combined wave/current)
Porosity j 0.50 at the peak of storm activity is greatest in the
Bed thickness bed_thick 10.0 m shallow water along the coastline and on the crest of
Northern edge tide A, Tt 1.0 m, 12 h
Stellwagen Bank, coinciding with locations of
Northern edge wave Hsig 2m
height increased bottom orbital velocities from the wave
Northern edge wave T 10 s model (Fig. 7). Stresses are lower in Stellwagen
period Basin where the near-bottom wave currents are
Northern edge wave y From 01 attenuated in the deep water, and in Cape Cod
direction
Bay because the Cape shelters the Bay from waves
from the northeast. Along the western shore of

Hsig (m) Max combined bottom Bed thickness (m)

stress (Pa)
1 4 12
One−way Coupling
ocean)

0.8
3
Distance (km)

10
0.8 11
0.6
2
0.4
(waves

5 0.6 10
1
0.2

0 0 0 9

1 4 12

0.8
3
Distance (km)

10
Full coupling

0.8 11
0.6
2
5 0.4
0.6 10
1
0.2

0 0 0 9
0 5 10 15 0 5 10 15 0 5 10 15
Distance (km) Distance (km) Distance (km)

Fig. 6. Tidal inlet test case. Comparison of significant wave height, bottom stress, and bed thickness for a one-way coupled simulation
(wave parameters sent to ocean model) to a fully coupled simulation (wave parameters to ocean model and ocean data to wave model). See
text for full data transfer description. For the fully coupled system, wave heights show effect of currents, maximum bottom stress in
enhanced due to increased wave heights, and bed thickness develops a stronger flood shoal than the one-way coupled system.
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1303

Fig. 7. Results from model simulation of Massachusetts Bay for the evolution of a mixed sediment bed in response to a sequence of 10
idealized northeast storms, modeled after the December 1992 storm with winds from 501. Panels show instantaneous wave–current bottom
stress at peak of storm (A), and change in bathymetry (B) and final mean surficial grain size (C) following the 10-storm sequence. The
observed surficial grain size distribution (D) qualitatively matches the evolved sediment texture.

Massachusetts Bay (in the region opposite of coarser in regions of high stress and finer in areas of
Stellwagen Basin) the combined wave and current low stress. The surficial sediment texture qualitatively
stress is high but then decreases between Plymouth resembles the observed distribution (Fig. 7d; Poppe et
and Barnstable. The instantaneous suspended-sedi- al., 2003). The crest of Stellwagen Bank, the outer
ment concentrations during the storm (not shown) Cape, and the western shore of Massachusetts Bay
are greatest in the regions of highest stress and are north of Plymouth have been winnowed to sizes of
lowest in Cape Cod Bay and in Stellwagen Basin. At 2–3j and sediment deposition in Cape Cod Bay and
the end of the simulation, net erosion has occurred Stellwagen Basin has produced a surface of 5–6j
along the crest of Stellwagen Bank, along the material. The material east of Stellwagen Bank is
western shore of Massachusetts Bay (in the region slightly finer in the model, possibly because the Gulf
of high stress), and along the outer arm of Cape of Maine coastal current which could transport the 4
Cod, reaching maximum values of 0.02 m (Fig. 7b). and 5j material further to the south is not simulated.
Deposition occurs in Stellwagen Basin immediately
west of Stellwagen Bank, and in Cape Cod Bay. 5. Future work
The surficial grain size distribution after 10 storms
(Fig. 7C) has approached a steady-state, with Future model improvements will be implemented
continued small changes that do not significantly alter using an open-source, community development ap-
the pattern described here. The sediment texture is proach. We plan to investigate alternative approaches
 ARTICLE IN PRESS
1304 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

to wave-induced circulation in the nearshore, includ- EuroSTRATAFORM project. We thank the re-
ing the vortex-force representation of McWilliams et viewers for their comments and the developers of
al. (2004). We also are investigating other model- ROMS for open access to their code.
coupling methodologies, such as the Earth System
Modeling Framework (ESMF, http://www.esmf.u-
car.edu/). Algorithms to represent cohesive sediment References
behavior and biodiffusive mixing are under develop-
ment. Several aspects of the BBL calculations will be Arakawa, A., 1966. Computational design for long-term numer-
ical integration of the equations of fluid motion: Two-
improved by including effects of sediment-induced
dimensional incompressible flow. Part I. Journal of Compu-
stratification, time-dependent ripple evolution, and tational Physics 1, 119–143.
representation of bedforms that develop under Ariathurai, C.R., Arulanandan, K., 1978. Erosion rates of
combinations of waves and currents. These processes cohesive soils. Journal of Hydraulics Division 104 (2),
and others are currently being investigated. 279–282.
Blaas, M., Dong, C., Marchesiello, P., McWilliams, J.C.,
Stolzenbach, K.D., 2005. Sediment-transport modeling on
6. Conclusions Southern Californian Shelves: a ROMS case study. Con-
tinental Shelf Research 27, 832–853.
We are developing a coupled wave–current-sedi- Blumberg, A.F., Galperin, B., O’Connor, D.J., 1992. Modeling
ment transport-morphodynamic oceanographic cir- vertical structure of open-channel flows. Journal of Hydraulic
culation model applicable to studies in rivers, lakes, Engineering, ASCE 118 (H8), 1119–1134.
Booij, N., Ris, R.C., Holthuijsen, L.H., 1999. A third-generation
estuaries, coastal environments, and the coastal wave model for coastal regions, Part I, Model description and
ocean. The ocean circulation model ROMS has been validation. Journal of Geophysical Research 104 (C4),
coupled using the MCT to the surface wave model 7649–7666.
SWAN. We have incorporated nearshore radiation- Booij, N., Haagsma, I.J.G., Holthuijsen, L.H., Kieftenburg,
stress terms and a surface roller model to account for A.T.M.M., Ris, R.C., van der Westhuysen, A.J., Zijlema, M.,
2004. SWAN Cycle III version 40.41 User Manual. Delft
surfzone (nearshore) processes. The sediment-trans- University of Technology.
port algorithms have been implemented to transport Casulli, V., Cheng, R., 1992. Semi-implicit finite difference
an unlimited number of user-defined sediment classes. methods for three-dimensional shallow water flow. Interna-
Suspended-sediment transport is computed with the tional Journal for Numerical Methods in Fluids 15, 629–648.
advection–diffusion equation and a vertical-settling Chassignet, E.P., Arango, H.G., Dietrich, D., Ezer, T., Ghil, M.,
Haidvogel, D.B., Ma, C.-C., Mehra, A., Paiva, A.M., Sirkes,
algorithm not restricted by the CFL criteria. Erosion Z., 2000. DAMEE-NAB: the base experiments. Dynamics of
and deposition algorithms control mass balance Atmospheres and Oceans 32, 155–183.
between suspended sediment in the water column Colella, P., Woodward, P., 1984. The piecewise parabolic method
and the evolution of a multi-level bed framework. (PPM) for gas-dynamical simulations. Journal of Computa-
tional Physics 54, 174–201.
The bed tracks the mass, fractions, thickness, and
Durski, S.M., Glenn, S.M., Haidvogel, D.B., 2004. Vertical
surface properties of the sediment and thus allows mixing schemes in the coastal ocean: comparison of the level
computations of morphological behavior and strati- 2.5 Mellor-Yamada scheme with an enhanced version of the
graphy. Bedload transport occurs in the top layer for K profile parameterization. Journal of Geophysical Research
all the sediment classes. Additional features are being 109, C01015.
implemented to account for such conditions as mixed Fredsøe, J., Deigaard, R., 1992. Mechanics of coastal sediment
transport. In: Liu, P.P.L.-F. (Ed.), Advanced Series on Ocean
grain sizes, cohesive behavior, and bed biodiffusivity. Engineering, Vol. 3. World Scientific, Singapore, p. 392.
The model in its current state (ROMS 3.0) is Grant, W.D., Madsen, O.S., 1979. Combined wave and current
available online and model development is being interaction with a rough bottom. Journal of Geophysical
conducted as an open-source community effort. We Research 84 (C4), 1797–1808.
hope for feedback that will add features and make Grant, W.D., Madsen, O.S., 1982. Movable bed roughness in
unsteady oscillatory flow. Journal Geophysical Research 87
the model more robust as the community of users (C1), 469–481.
and developers grows. Haidvogel, D.B., Arango, H.G., Hedstrom, K., Beckmann, A.,
Malanotte-Rizzoli, P., Shchepetkin, A.F., 2000. Model
Acknowledgments evaluation experiments in the North Atlantic Basin: Simula-
tions in nonlinear terrain-following coordinates. Dynamics of
Atmospheres and Oceans 32, 239–281.
This work has been supported by the US Haidvogel, D.B., Arango, H.G., Budgell, W.P., Cornuelle, B.D.,
Geological Survey Coastal and Marine Geology Curchitser, E., Di Lorenzo, E., Fennel, K., Geyer, W.R.,
Program and by the Office of Naval Research Hermann, A.J., Lanerolle, L., Levin, J., McWilliams, J.C.,
 ARTICLE IN PRESS
J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306 1305

Miller, A.J., Moore, A.M., Powell, T.M., Shchepetkin, A.F., Mellor, G.L., Yamada, T., 1982. Development of a turbulence
Sherwood, C.R., Signell, R.P., Warner, J.C., Wilkin, J., 2007. closure model for geophysical fluid problems. Reviews of
Regional Ocean forecasting in terrain-following coordinates: Geophysics and Space Physics 20, 851–875.
model formulation and skill assessment. Journal of Computa- Meyer-Peter, E., Müeller, R., 1948. Formulas for bedload
tional Physics. transport. In: Report on the 2nd Meeting International
Harris, C.K., Wiberg, P.L., 1997. Approaches to quantifying Association Hydraulic Structure Research. Stockholm, Swe-
long-term continental shelf sediment transport with an den, pp. 39–64.
example from the northern California STRESS mid-shelf Nielsen, P., 1986. Suspended sediment concentrations under
site. Continental Shelf Research 17, 1389–1418. waves. Coastal Engineering 10, 23–31.
Harris, C.K., Wiberg, P.L., 2001. A two-dimensional, time- Nielsen, P., 1992. Coastal Bottom Boundary Layers and
dependent model of suspended sediment transport and bed Sediment Transport. Advanced Series on Ocean Engineering,
reworking for continental shelves. Computers & Geosciences vol. 4. World Scientific, Singapore, 324pp.
27, 675–690. Phillips, O.M., 1969. The Dynamics of the Upper Ocean.
Jacob, R., Larson, J., Ong, E., 2005. M x N Communication and Cambridge Press, Cambridge.
parallel interpolation in community climate system model Poppe, L.J., Paskevich, V.F., Williams, S.J., Hastings, M.E.,
version 3 using the model coupling toolkit. International Kelly, J.T., Belknap, D.F., Ward, L.G., FitzGerald, D.M.,
Journal of High Performance Computing Applications 19 (3), Larsen, P.F., 2003. Surficial sediment data from the Gulf of
293–307. Maine, Georges Bank, and vicinity: A GIS Compilation. US
Jonsson, I.C., Carlsen, N.A., 1976. Experimental and theoretical Geological Survey Open-File Report 03-001, /http://pubs.
investigations in an oscillatory boundary layer. Journal of usgs.gov/of/2003/of03-001/index.htmS.
Hydraulic Research, ASCE 14 (1), 45–60. Roelvink, J.A., 2006. Coastal morphodynamic evolution techni-
Kantha, L.H., Clayson, C.A., 1994. An improved mixed layer ques. Coastal Engineering 53, 277–287.
model for geophysical applications. Journal of Geophysical Shchepetkin, A.F., McWilliams, J.C., 2005. The regional ocean
Research 99 (C12), 25,235–25,266. modeling system (ROMS): a split-explicit, free-surface,
Large, W.G., McWilliams, J.C., Doney, S.C., 1994. A review and topography-following-coordinates ocean model. Ocean Mod-
model with a nonlocal boundary layer parameterization. elling 9, 347–404.
Reviews of Geophysics 32, 363–403. Sherwood, C.R., Harris, C.K., Geyer, W.R., Butman, B., 2002.
Larson, J., Jacob, R., Ong, E., 2005. The model coupling toolkit: Toward a community coastal sediment-transport modeling
a new fortran90 toolkit for building multiphysics parallel system: Report of the Second Workshop. EOS, Transactions,
coupled models. International Journal of High Performance American Geophysical Union 83 (51).
Computing Applications 8 (19), 277–292. Smith, J.D., 1977. Modeling of sediment transport on continental
Lesser, G.R., Roelvink, J.A., van Kester, J.A.T.M., Stelling, shelves. In: Goldberg, E.D., McCave, I.N., O’Brien, J.J.,
G.S., 2004. Development and validation of a three-dimen- Steele, J.H. (Eds.), The Sea, Vol. 6. Wiley-Interscience, New
sional morphological model. Coastal Engineering 51, York, pp. 539–577.
883–915. Smith, J.D., McLean, S.R., 1977. Spatially averaged flow over a
Li, M.Z., Amos, C.L., 2001. SEDTRANS96: the upgraded and wavy surface. Journal of Geophysical Research 82 (12),
better calibrated sediment-transport model for continental 1735–1746.
shelves. Computers & Geosciences 27, 619–645. Soulsby, R.L., 1995. Bed shear-stresses due to combined waves
Li, M., Zhong, L., Boicourt, B., 2005. Simulation of Chesapeake and currents. In: Stive, M.J.F. (Ed.), Advances in Coastal
Bay Estuary: sensitivity to turbulence mixing parameteriza- Morphodynamics: An Overview of the G8-Coastal Morpho-
tions and comparison with hydrographic observations. dynamics Project, Co-Sponsored by the Commission of
Journal of Geophysical Research 110, C12004. The European Communities Directorate General XII
Liu, X.-D., Osher, S., Chan, T., 1994. Weighted essentially non- pp. 4.20–4.23.
oscillatory schemes. Journal of Computational Physics 115, Soulsby, R.L., 1997. Dynamics of Marine Sands. Thomas
200–212. Telford, London, 249pp.
Madsen, O.S., 1994. Spectral wave–current bottom boundary Soulsby, R.L., Damgaard, J.S., 2005. Bedload sediment transport
layer flows. In: Coastal Engineering 1994. Proceedings of the in coastal waters. Coastal Engineering 52 (8), 673–689.
24th International Conference on Coastal Engineering Styles, R., Glenn, S.M., 2000. Modeling stratified wave and
Research Council, Kobe, Japan, pp. 384–398. current bottom boundary layers on the continental shelf.
Malarkey, J., Davies, A.G., 2003. A non-iterative procedure for Journal of Geophysical Research 105 (C10), 24,119–24,139.
the Wiberg and Harris (1994) oscillatory sand ripple Styles, R., Glenn, S.M., 2002. Modeling bottom roughness in the
predictor. Journal of Coastal Research 19 (3), 738–739. presence of wave-generated ripples. Journal of Geophysical
McWilliams, J.C., Restrepo, J.M., Lane, E.M., 2004. An Research 107 (C8), 24/1–24/15.
asymptotic theory for the interaction of waves and Svendsen, I.A., 1984. Wave heights and set-up in a surf zone.
currents in coastal waters. Journal of Fluid Mechanics 511, Coastal Engineering 8, 303–329.
135–178. Svendsen, I.A., Haas, K., Zhao, Q., 2002. Quasi-3D nearshore
Mellor, G.L., 2003. The three-dimensional current and surface circulation model SHORECIRC, User’s Manual, Draft
wave equations. Journal of Physical Oceanography 33, Report, Center for Applied Coastal Research, Department
1978–1989. of Civil Engineering, University of Delaware, Newark.
Mellor, G.L., 2005. Some consequences of the three-dimensional Umlauf, L., Burchard, H., 2003. A generic length-scale equation
currents and surface wave equations. Journal of Physical for geophysical turbulence models. Journal of Marine
Oceanography 35, 2291–2298. Research 61, 235–265.
 ARTICLE IN PRESS
1306 J.C. Warner et al. / Computers & Geosciences 34 (2008) 1284–1306

van Rijn, L.C., 1987. Mathematical modelling of morphological Wiberg, P.L., Nelson, J.M., 1992. Unidirectional flow over
processes in the case of suspended sediment transport. Delft asymmetric and symmetric ripples. Journal of Geophysical
Technical University, Delft Hydraulics Communication No. Research 97 (C8), 12,745–12,761.
382, Delft, The Netherlands. Wiberg, P.L., Rubin, D.M., 1989. Bed roughness produced by
van Rijn, L.C., 1993. Principles of Sediment Transport in Rivers, saltating sediment. Journal of Geophysical Research 94 (C4),
Estuaries, and Coastal Seas. Aqua Publications–I11, Am- 5011–5016.
sterdam, The Netherlands, 614pp. Wiberg, P.L., Sherwood, C.R. (this issue). Calculating wave-
Warner, J.C., Sherwood, C.R., Arango, H.G., Signell, R.P., 2005. generated bottom orbital velocity from surface wave para-
Performance of four turbulence closure models implemented meters. Computers & Geosciences, doi:10.1016/j.cageo.
using a generic length scale method. Ocean Modelling 8, 2008.02.010.
81–113. Wijesekera, H.W., Allen, J.S., Newberger, P.A., 2003. Modeling
Warner, J.C., Perlin, N., Skyllingstad, E. (in press). Using the study of turbulent mixing over the continental shelf:
model coupling toolkit to couple earth system models. comparison of turbulent closure schemes. Journal of Geo-
Environmental Modelling and Software. physical Research 108 (C3), 3103.
Wiberg, P.L., Harris, C.K., 1994. Ripple geometry in wave-
dominated environments. Journal of Geophysical Research
99 (C1), 775–789.