Computers & Fluids 111 (2015) 114126

Contents lists available at ScienceDirect

Computers & Fluids

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

Review

The role of air modeling on the numerical investigation of coastal

dynamics and wave-structure interactions
Maria Gabriella Gaeta a,, Alberto Lamberti b
a
b

CIRI, University of Bologna, via Terracini 28, 40136 Bologna, Italy

DICAM, University of Bologna, viale Risorgimento 2, 40134 Bologna, Italy

a r t i c l e

i n f o

Article history:
Received 11 April 2013
Received in revised form 15 December 2014
Accepted 5 January 2015
Available online 14 January 2015
Keywords:
Two-phase model
Air compressibility
Breaking waves
Wave-induced loads

a b s t r a c t
The paper presents a numerical investigation on the role of air modeling in simulations related to coastal
dynamics. The implemented code, named COBRAS2, solves the FavreReynolds NavierStokes equations
for two-phase ows. The k model is adopted to dene the Reynolds stress; the polytropic expression is
chosen as the gas state equation to describe air compressibility; the Volume Of Fluid algorithm is implemented in order to track the interface. Simulations of dam-break wave and 1D water piston illustrate the
model validity and accuracy, where air inertia and compressibility play a signicant role in the reproduced dynamics. Wave breaking is analyzed in comparison with experimental data in order to focus
on the inuence of air ow in the wave propagation. Finally, air entrapment and compressibility are
investigated during the wave impact on deck and on vertical wall and the opportunity to solve the
implemented two-phase equations is discussed together with the aim to obtain accurate estimation of
wave-induced loads.
2015 Elsevier Ltd. All rights reserved.

Contents
1.
2.

3.

4.

5.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Description of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.
Two-phase modeling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.
Air compressibility treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.
Numerical resolution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.
Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.
Dam-break wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.
1D water piston. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Airwater dynamics during wave breaking and impact on structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
Wave breaking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2.
Wave-induced forces on structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction
Oceans and coastal regions are characterized by two-uid
dynamics, referring primarily to air and water interactions at the
Corresponding author.
E-mail address: g.gaeta@unibo.it (M.G. Gaeta).
http://dx.doi.org/10.1016/j.compuid.2015.01.001
0045-7930/ 2015 Elsevier Ltd. All rights reserved.

114
115
115
116
116
117
117
117
118
119
119
121
125
125

free-surface with the development of bubble spray and air pocket/plume entrainment in water. In general, these processes occur
as a result of wave propagation toward the shore or wave-structure
interactions.
Fig. 1 shows a schematic representation of the main two-uid
ows that characterize the coastal environment and involve both
air and water (phenomena related to atmospheric circulation,

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 1. A schematic representation of two-uid ows in coastal environment.

wind generation and soliduid interactions are ignored). Breaking of waves is one of the multiuid coastal processes that greatly
inuence hydromorphodynamics, i.e., turbulence generation,
energy dissipation, overtopping of maritime structures, run-up
and ooding. Following common analytical and numerical
approaches, wave breaking is often treated as a single-phase ow,
adopting the traditional kinematic and dynamic free-surface
boundary conditions.
This assumption is conventionally accepted and implemented
to numerically reproduce wave propagation and transformation
(Christensen [8]). Although, breaking waves are usually identied
as white waters, chaotic mixtures of air and water whose properties affect velocity and pressure eld at the vicinity of the freesurface.
Indeed, according to the Galvin [17] classication, different
breaker shapes commonly generate distinctive air entrainment
mechanisms. Spilling waves induce air engulfment via a surface
roller, with the development of bubbles close to the free-surface.
In the case of plunging waves, the entrapment of air pockets
occurs as an overturning jet develops and falls forward. Bubble
entrainment during wave breaking plays a key role in mass
and energy transport through the airwater interface (Melville
[34], Fhrbter [16]), thus revealing how a large portion of the
initial wave energy that is rst stored by the air and subsequently driven into the water, is dissipated via turbulence and mass
transfer.
Wave-induced loads on maritime structures signicantly vary
in their magnitudes and durations in relation to wave breakers
and their air content. In addition to the engulfment of bubbles with
typical dimensions measured in mere millimeters, the impact on a
vertical wall generally develops together with the inclusion of air
between the wall and wave front. Different authors (a review is
reported in Plumerault et al. [39]) have classied the impulsive
loads related to the breaker shape and the wave dynamics that
develop due to wall proximity.
According to a recent experimental investigation by Lugni et al.
[32] on vertical walls, the wave front interacts with the rigid
boundary before breaking and develops a pure ip-through impact
(a), or it can break as it meets the wall, thus entrapping a small (b)
0
or a large (b ) air bubble during impact. Finally, if the wave breaks
prior to wall impact, phase mixing occurs, which causes an irregular evolution of the wave front (c).
The entrapped air pocket, with dimensions near those of the
wave height, creates a cushioning effect as water approaches the
structure (Peregrine et al. [38]) and air compressibility induces a
signicant reduction in the impact pressure. However, if the air
bubble sizes are sufciently large, the enclosed bubble could
extend the pressure peak duration.
In general, the magnitude of impulsive pressures on vertical
walls increases with a reduction in the air pocket size (Bagnold

115

[2], Hattori et al. [19]) and the highest impact force is commonly
observed when the plunging breaker entraps an air pocket, that
prolongs the duration of the peak pressures (Wood et al. [44]).
Applied mathematics and computer architectures offer the ability to develop inexpensive and rapid numerical investigations that
can provide detailed information on velocity, acceleration, pressure and turbulence. The majority of numerical investigations on
coastal dynamics and wave-structure interactions are based on
single-phase models (among others, Guanche et al. [18]). However,
it is widely recognized that the air phase is an important feature to
consider in studying these topics.
Due to large interfacial deformations, multi-uid codes typically adopt Eulerian or Lagrangian approaches to solve the governing
equations. In particular, among the multiphase Eulerian models,
the Level Set (LS) method is adopted to track the interface in the
two-phase incompressible Large Eddy Simulation (Lubin et al.
[31], Lubin and Glockner [30]) and the inviscid incompressible
(Colicchio et al. [12]) models, which were both recently implemented to investigate plunging breakers and impacting waves,
respectively. The most popular Lagrangian solver adopts the
Smoothed Particle Hydrodynamics (SPH) technique (Monaghan
[35]) and the newly updated (cSPH) by Colagrossi and Landrini
[11] to address violent wave impact.
Althogh Lubin et al. [31] demonstrated the inuence of air resistance in a numerical study on wave breaking, Colagrossi et al. [10]
reported on the numerical challenges and limitations of violent
sloshing ow simulation by adopting both the cSPH and LS models.
In particular, the absence of air compressibility leads to an overestimation of the pressure impact peaks, and the uctuations
induced by air bubble compression and expansion are not properly
computed.
The aim of the present paper is to investigate the role played by
compressible air modeling in the reproduction of breaking waves
and their interaction with structures. The implemented solver,
named COBRAS2, represents an extension of the two-dimensional
vertical (2DV) single-phase COBRAS0 solver, originally developed
by Cornell University (Lin and Liu [26]) and later implemented
by Lara et al. [23].
The mathematical formulation and the numerical resolution of
the two-phase governing equations are described in Section 2,
which focuses on air compressibility treatment. Section 3 illustrates selected benchmark tests chosen to validate the model, i.e.,
the results of simulations of a dam-break wave and a 1D water piston are compared with the experimental observations, and the
accuracy of the computed results is reported. Airwater interactions during breaking and at wave impact on a deck and a vertical
wall are discussed in Section 4, in which the inuence of air modeling is examined. Finally, some conclusions are provided in the
last section of the paper.
2. Description of the model
2.1. Two-phase modeling
The formulation of the two-uid equations and the most
appropriate closure laws follow the multiphase ow theory by
Drew and Passman [13], where the ensemble average of the
exact conservation equations is applied to each considered
phase.
Let uk x; t; a be the component indicator function or phase
index, which, for a given realization a of the ow, takes the value
of 1.0 if the phase k is present at the point x and time t, and the
value of 0.0 otherwise, assuring that:

uk x; t; a 1:0

116

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

The averaged equations implemented in the present model only

consider two uids: the liquid phase, with u 1:0, and the gaseous or second liquid phase, with u 0:0. For the partial-lled
cells, with 0:0 < u < 1:0, the following mixture quantities are
dened as:

qm uqf 1  uqa ; mm umf 1  uma ; pm upf 1  upa ;

um;i uuf ;i 1  uua;i ; T m;ij uT f ;ij 1  uT a;ij
2
where q is the density, m the kinematic viscosity, ui the i-th component of the velocity, p the pressure, T ij the stress tensor for m = mixture, f = uid and a = air cells . The adopted indices i and j, equal to 1
or 2, are referred to horizontal and vertical components in a 2DV
ow, and the Einstein notation is not adopted for the phase index u.
Only the ensemble average has been pointed out in order to
homogenize the phases and to dene the phase-related average
quantities such as the volume fraction. This assumption does not
necessarily imply that all the turbulent uctuations are averaged.
The Reynolds averaged governing equations would become
considerably more complex due to the additional correlation terms
involving the density uctuations, that are not properly known. As
Drew and Passman [13] also suggested, the most suitable method
to address this issue is the application to the mixture NavierStokes equations of the Favre average (hereafter denoted with a
double over-bar) to velocities and of the Reynolds average (hereafter denoted with a single over-bar) to density, pressure and other
quantities.
The interfacial forces between the air and water, constituting
the mixture terms in Drews theory, are supposed to be strong
enough to develop a uniform velocity at the interface, where the
phase changes. This assumption could be considered acceptable
for studies about large air bubbles entrapped into water and
admits to simply the equations and neglect unknown forces, difcult to model otherwise.
For a two-phase ow, the continuity equation is expressed by:

@ qm @qm um;i

0
@xi
@t

and the momentum equation leads to:


@qm um;i @qm um;i um;j
@p
@ 

 m qm g i
T m;ij  qm u0m;i u0m; j
@t
@xj
@xj
@xi
4
Expressions (3) and (4) constitute the FavreReynolds Averaged
NavierStokes (FRANS) equations written in terms of mixture
quantities and in a conservative form.
The last term in Eq. (4) represents the FavreReynolds stress
tensor, that needs to be modeled. Some authors (Sokolichin and
Eigenberger [40]) proved that if the second phase is present with
slight fraction, the standard k closure model could be implemented with mixture terms of the turbulent kinetic energy km
and dissipation m , and adopted to reproduce air turbulence.
Therefore, the formulation by Launder et al. [24] for the k
model, already validated by Lin and Liu [26] for coastal applications, is also used in the present model.

2.2. Air compressibility treatment

In COBRAS2, water is reasonably considered incompressible,
and air compressibility is accounted for: the relationship between
the gas density and pressure is therefore implemented in the
solver.
Among the possible options commonly adopted in hydraulics,
the polytropic state equation has been chosen as follows:

pa

pa;0

qa
qa;0

!c

where pa;0 and qa;0 are the reference gas pressure and density, pa is
the absolute gas pressure. The polytropic coefcient c usually
depends on the gas properties (Faltinsen et al. [14]): the value of
1.0 is chosen in isothermal conditions, and the value of 1.4 (hereafter adopted) is set in adiabatic conditions.
In the present model, the reference quantities pa;0 and qa;0 are
dened as air pressure and density in atmosphere, respectively
equal to 101 kPa and to 1.0 kg/m3.
Air compressibility has its only contribution to Eqs. (3) and (4)
in mixture and gas cells. Table 1 shows a scheme resuming the
relation between density, pressure and phase index for the three
different solvers, the compressible and incompressible two-phase
COBRAS2 and the single-phase COBRAS0.
For all the three models, constant density is set in water, as
1025 kg/m3 in sea, whereas the compressible and incompressible
solvers appears signicantly distinct at mixture and aerated cells.
Adding air compressibility, cell density depends on the phase index
and air pressure, according to the implemented polytropic law.
For the three solvers, the main difference in the density treatment is related to aerated cells: in COBRAS0, water is supposed
to move in a void (i.e., qa 0), whereas for the other two solvers,
air properties are accounted for.
2.3. Numerical resolution
The governing equations, with the nite difference scheme and
the staggered grid discretization, are solved using the two-step
projection algorithm (Chorin [7]) adopted to include air
compressibility.
The rst step consists of solving Eq. (4) in absence of the
pressure gradient and assuming incompressibility for the diffusive
;n1
density term. The intermediate velocity um;i
at the time step n 1

is obtained as:

u;n1
 unm;i
m;i

Dt



n
n n

1 @ qm um;j um;i
1 @  n
0n
 n
gi n
T m;ij  qnm u0n
m;i um;j
@xj
qm
qm @xj
6

where the index n stands for the current computational time.

Next, in the second stage of the calculation, the Poisson Pressure
Equation (PPE) is solved taking into account additional terms
induced by air compressibility from Eq. (3) as:

@
1 pn1
m
Dt
@xj qnm @xi

@u;n1
m;i
@xj

n
unm;i @ qnm
1 qn1
m  qm
n
n
D
t
qm
qm @xi

Table 1
Density in water, mixture and air cells for the compressible COBRAS2, incompressible COBRAS2 and COBRAS0 solvers.
Flow model

Compressible COBRAS2

Incompressible COBRAS2

COBRAS0

Water cell u 1:0

q = qf constant
q = qm u; p Eqs. (2) and (5)
q = qa f p Eq. (5)

q = qf constant
q = qm u Eq. (2)
q = qa small but nite

q = qf constant
q = uqf Eq. (2)
q=0

Mixture cell 0:0 < u < 1:0

Air cell u 0:0

117

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

The last two terms at the right-hand side of Eq. (7) are dened
as the unsteady density term and the density gradient term. Using
the denitions of the phase index u and of the mixture density qm ,
Eq. (7) can be split into an incompressible liquid part, related to the
intermediate velocity divergence, and a compressible gas part, as:

@
1 pn1
m
Dt
@xj qnm @xi

@u;n1
m;i
@xj

1  u

qn1
 qna
a

qnm

Dt

unm;i

@ qna
@xi

8
The chosen equation of state is used to simplify the right-hand
side terms between the brackets in Eq. (8).
Indeed, linearizing Eq. (5):
n
n1  pn
qn1
qa pn1
a
a qa pa pa
a

dqa
j n
dpa qa

the nal expression for the PPE leads to:

!
@u;n1
@
1 pn1
1  un
m;i
m

n
@xj qm @xi
@xj
Dt qnm
"
#
i
q0 1 h n1
@ qna
1=c1
n
n
n
pa  pa pa
um;i
1=c
@xi
patm c

Dt

10

where the contribution of air compressibility to the pressure

gradient amplies the diagonal terms of the PPE matrix.
Eq. (10) is iteratively solved using the Implicit Cholesky
Conjugate Gradient algorithm.
The computation of the mixture density qm at partial lled cells
with the nite difference scheme is a crucial point, as serious errors
might be generated due to the inconsistent average method
(Jaluria and Torrance [21]) and for large density variations (order
of 103 as in the present study), the harmonic approximation, hereafter adopted, is recommended to treat air and water ows.
2.4. Boundary conditions
Traditional boundary conditions at the interface are commonly
applied in coastal models, but they hardly compute a disrupted
airwater interface since the mean free-surface is not clearly
dened during turbulent dynamics. For this reason, Brocchini and
Peregrine [4,5] proposed a new dedicated formulation that
accounts for the multiphase mixing ow occurring at a splashing
airwater interface.
In COBRAS0, the traditional kinematic and dynamic conditions
and the zero-gradient assumptions for both k and  values are
imposed at the free-surface, also under the assumption of no turbulence exchanges between air and water. Such an approximation
can be considered correct in case of smooth interface, where air
bubbles are present for short times at the free-surface and the mixing layer is negligible.
In COBRAS2, on the other hand, the governing equations are
solved for the entire computational domain. Therefore, pressure,
velocity, turbulence and energy dissipation are also calculated in
air, without interfacial boundary conditions.
In general, the top boundary is imposed open in the present
study, in order to minimize sort of roof effects on the ows, and
at the other boundaries, specic conditions are implemented. In
case of coastal simulations, the approach suggested by Torres-Freyermuth et al. [42] is implemented to generate waves at the left
boundary of the domain. At solid cells, reproducing impermeable
structures, no-slip or free-slip conditions could be chosen for velocities, and Neumann condition is imposed to pressures.
In addition, the model presents the particular feature to simulate ltration, not investigated here, and solves the Forccheimers

law to evaluate the seepage ow through permeable structures

(Clementi et al. [9]).
3. Model validation
In this section, two benchmark tests are performed in order to
point out the model validation. The inuence of air modeling in
simulations is shown and the numerical accuracy of the results is
discussed. The rst described test reproduces the dam-break wave
propagation, and the treatment of air compressibility is investigated in the second subsection, in which a 1D water piston is
reproduced.
3.1. Dam-break wave
Dam-break wave propagation over a dry bed is simulated using
the proposed two-phase solver. This test is commonly considered
as a benchmark case used to evaluate model stability and accuracy
in reproducing the sharp interface, where air and water signicantly interact in chaotic patterns.
Two documented laboratory congurations are studied: experiments carried out by Martin and Moyce [33], in which the dambreak wave propagates over a horizontal dry bed, and tests by
Aureli et al. [1], in which a 1/10 ramp is included in the ume to
investigate the wave run-up on the beach.
A water reservoir, with dimensions L0 wide and H0 high, is
placed at the left boundary of the numerical domain, and an
instantaneous gate opening is reproduced at t 0. Dimensionless
quantities are dened as following:

s t g=H0

X x=L0

Z z=L0

11

where t; x and z are the time, the wave front position and the water
column level measured at the left boundary.
The tests conducted by Martin and Moyce [33] reproduced two
initial conditions, which are characterized by the ratio H0=L0 equal
to 1.0 (left box in Fig. 3) and 2.0 (right box in Fig. 3), respectively. A
square computational region with dimensions equal to 4L0 is
implemented. Open ow is admitted through the top boundary
of the domain, and no-slip conditions are imposed at the other
boundaries.
The numerical convergence is investigated using three different
grid sizes. In Fig. 2, the wave front evolution X for the rst simulated case is plotted by varying the square cell dimensions
Dx=L0 0:05; 0:1; 0:125.
In comparison with the grid of Dx=L0 0:1, the choice of coarser cells leads to a reduction of 4.5% in the maximum value of X,
whereas a lower dependence on the mesh size is observed with
the smaller grid (increase of 1.3%). Therefore, uniform square cells
with dimensions of 0:1L0 are chosen for the tests because they represent a better compromise between the computational costs and
numerical accuracy.
Using a 2.8 GHz Intel Core i7 machine, a run of the test typically
completes in approximately 1 min with COBRAS0, and the time
needed for COBRAS2 increases by a factor of 3 (by a factor of 2 if
air compressibility is neglected).
In Fig. 3, the wave front position X is plotted against time s for
the two dam-break tests, and a comparison between the laboratory
data (dots) and the numerical results from COBRAS2 (solid line)
and COBRAS0 (dotted line) is shown.
The computed results satisfactorily agree with the measurements, and both solvers simulate the process in a proper manner.
In particular, during the rst instants of the wave propagation
(s < 0:5), the two models comparably reproduce the falling of
the water column because a limited role of air motion is observed
at this initial stage.

118

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 4. Time evolution of the water level g at two wave gauges during the dambreak propagation: COBRAS2 (solid line) and the laboratory data (dots) by Aureli
et al. [1].
Fig. 2. Grid analysis on the time evolution of the dam-break wave front
X : Dx=L0 0:05; 0:1; 0:125.

Then, as the wave quickly advances, COBRAS2 displays a higher

accuracy in the simulations in comparison to COBRAS0, thus
revealing a stronger inuence of air resistance in the calculation
during the propagation stage. Indeed, the water acceleration is reasonably reduced by the air inertia, leading to a better agreement
between the computations and experiments.
Moreover, the run-up of the dam-break wave over a ramp
(Aureli et al. [1]) is well reproduced by the model. Fig. 4 shows
the water elevation at two gauges, as modeled by COBRAS2, in
comparison with the experimental measurements. In particular,
the last section (G2) located on the ramp, is cyclically ooded,
proving that the present solver is able to correctly replicate both
wet and dry conditions.
3.2. 1D water piston
A 1D piston is simulated in order to investigate the correct
reproduction of air compressibility. The test set-up reproduces
the computations by Wemmenhove et al. [43], which are used to
validate the current results. A 1  10 m2-domain is lled with
two phases. Initially at rest, a water column of dimensions
1  4 m2 is xed at a distance of 4 m from the bottom, and air is
enclosed in the space below the water. The numerical domain of
the test is reported in Fig. 5, in the right-hand panels. The top
boundary is set open, and no-slip conditions are otherwise
imposed. The initial air density qa t 0 1:0 kg/m3 is set equal
to the atmospheric value, and the pressure eld in both the phases
is computed according to Eq. (10).
COBRAS0 and the incompressible and compressible COBRAS2
models are used to simulate the piston motion. The computations

clearly give different results because the numerical reproduction of

the process is strongly related to air modeling.
The water piston operation is described as follows. When
released, the volume of water, initially at rest, accelerates
downward due to gravity, pushing down the entrapped air that
compresses and expands until equilibrium is reached. With
COBRAS0, the entrapped air is treated as a void, and thus water
moves rapidly downward, lling the void space and reaching the
bottom of the domain at time t  0:9 s.
If running the test with COBRAS2 solvers, the air is initially at
rest under the weight of the water column. Neglecting compressibility, the air is modeled with constant properties (as incompressible); therefore, the water piston remains xed in time and the
hydrostatic pressure distribution characterizes both the air and
water phases.
The realistic dynamics of the piston are reproduced only with
the compressible COBRAS2 model. Fig. 5 shows the computed pressure proles (left-hand panels) and the air density (middle panels)
at four different time instants (also reported in Fig. 6). The corresponding position of the water column is drawn in blue color at
the right-hand panels in the same gure. As water moves downward due to gravity (a), the entrapped air volume is reduced and
the consequent air density increase according to Eq. (5), until it
reaches its maximum value (b) of 1.72 kg/m3.
During the following time instants, the air spreads (c), inducing
a reduction of the corresponding density and the upward water
motion. Finally, equilibrium is achieved (d): the stationary pressure and density at the bottom of the air column are equal to
39 kPa and 1.39 kg/m3 respectively.
Fig. 6 shows the time oscillation of the water column yw (bottom level), resulting from the simulation using the compressible
solver. Compared with Wemmenhove et al. [43] and with

Fig. 3. Time evolution of the dam-break wave front X: COBRAS2 (solid line), COBRAS0 (dotted line) and the laboratory data (dots) for the two tests H0=L0 1:0 (left-hand
panel) and 2.0 (right-hand panel) by Martin and Moyce [33].

119

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 7. Spatial evolution of the wave elevations gmax ; gmean and gmin : COBRAS2 (line)
and the laboratory data (dots) by Ting and Kirby [41].

4. Airwater dynamics during wave breaking and impact on

structures
4.1. Wave breaking

Fig. 5. Prole of the pressure p (left-hand panels), the air density qa (middle) and
the correspondent position of the 1D water piston (in blue map, right-hand panels)
at four different time instants: t = 0.15 s (a), 1.05 s (b), 1.85 s (c) and 4.85 s (d).

Fig. 6. Time evolution of the lower boundary position yw for the 1D water piston.

analytical results from Eq. (5), the air compression and expansion
are correctly simulated with decreasing amplitude in the level uctuations. When pressure of the entrapped air volume reaches a balance under the water column weight, the piston motion stops at
time t 6:0 s, and the nal air volume is equal to 2.86 m3, comparable to the value of 2.8 m3, that Wemmenhove et al. [43] found.

The numerical analysis on wave breaking is carried out to

reproduce the laboratory experiments performed by Ting and Kirby [41], which are commonly considered valid benchmark tests for
coastal computations.
A regular plunging wave with height H 0:125 m and period
T 5 s propagates over an impermeable 1/35 ramp. The numerical
domain is 30 m long and 0.7 m high and is discretized by
2503  72 cells, with a non uniform spatial length
Dx 0:010:03 m and a uniform height Dy 0:01 m. The reference system follows the one adopted in the experiments: the coordinate x starts where the water depth d is equal to 0.38 m (at the
wavemaker, d d0 0:4 m). The wave is generated from the left
boundary, and the temporal evolution of the free-surface oscillations and velocities are here imposed. The top boundary is set
open, and no-slip conditions are applied at the bottom and right
boundaries.
In Fig. 7, the spatial evolution of the computed water elevation
(solid lines) is compared with the laboratory data (dots); the
phase-averages of the maximum gmax , the mean gmean and the
minimum gmin values are reported. With the current two-phase
model, the matching between the measurements and the computations presents a high degree of accuracy, and both the wave crest
and the trough are satisfactorily predicted at the breaking point
and along the ramp. Table 2 reports the comparison between the
solvers and the laboratory data at the section xb , i.e., at the breaking point.
Compared with the single-phase model, COBRAS2 appears to
better reproduce the breaking location xb and the maximum water

Table 2
Wave characteristics at the breaking point for Ting and Kirby [41] test: the maximum
water elevation gb;max , the breaker position xb and the wave period T b .

Laboratory data
COBRAS0
COBRAS2

gb;max (m)

xb (m)

T b (s)

0.1608
0.1456
0.1615

7.79
7.52
7.85

0.950T
0.935T
0.958T

120

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 8. Phase-averaged free-surface elevation g=d at the breaking point.

Fig. 9. Phase-averaged velocities at the breaking point: horizontal u=C (top) and vertical w=C (bottom) components at y  g=d 0:24 (-), 0.5 (- -), 0.76 ( - ), 0.92 ( ).

elevation gb;max . The higher accuracy is induced by the absence of

numerical approximations at the free-surface and also by the air
resistance modeling during the wave propagation.
In fact, as Lin and Liu [27] also reported in their numerical
study, COBRAS0 slightly underestimates the breaker amplitude,
and the simulated wave breaks earlier than in the observations.
The reproduction of a lower wave height with COBRAS0 is possibly
related to the absence of air modeling and the imposed conditions
on velocity and pressure at the free-surface. The turbulence closure
law also plays a role in this wave height reduction as Losada et al.
[29] have discussed.
The phase-averages of the free-surface (Fig. 8), the velocity
components (Fig. 9) and the turbulent kinetic energy (Fig. 10) are

reported at the breaking point: the laboratory data (left-hand panels) and results by COBRAS0 (middle panels) and COBRAS2 (righthand panels) are compared.
In the gures, the normalized values are shown, in which g is
the wave elevation (eviscerated by the mean still water elevation
g), d is the water depth at the beginning of the ramp, T is the wave
period, u and w are the horizontal and vertical components of the
velocity, C is the wave celerity, k is the kinetic turbulent energy.
Velocities in Fig. 9 and turbulence in Fig. 10 are evaluated at different vertical sections y  g=d 0:24, 0.5 (- -), 0.76 ( - ),
0.92 ( ) with y the vertical coordinate.
The water oscillation is accurately reproduced by COBRAS2,
both in elevation and frequency, whereas an underestimation of

121

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 10. Phase-averaged turbulent kinetic energy k=C 2 at the breaking point at y  g=d 0:24 (-), 0.5 (- -), 0.76 ( - ), 0.92 ( ).

the velocity still affects both the solvers. Using COBRAS2, only a
rough improvement in the horizontal velocity is observed, and a
reduction in the turbulent kinetic energy is reported.

Table 3
Discretization analysis for French [15] test: the computed peak wave pressures pw;max
by COBRAS2 on the deck at X 0:0; 0:25 and 0:4 and for the three used grids.
Grid cells

pw;max (kPa) with COBRAS2

X 0:0

X 0:25

X 0:4

700  30
1400  60
2000  60

0.213
0.235
0.238

0.158
0.162
0.165

0.097
0.105
0.106

4.2. Wave-induced forces on structures

Loads on maritime structures are strictly correlated, in magnitude and duration, with the air content that characterizes the
impacting wave front. Two-phase simulations are therefore performed with COBRAS2 to estimate the inuence of air modeling
on the computation of the wave-induced forces on horizontal
decks and on vertical walls.
In the rst case, the experiment carried out by French [15] is
numerically replicated, and the uplift loads induced by a solitary
wave are evaluated on a at emerged deck. The force F, calculated
as the pressure integration on the deck of length L 1:5 m, is normalized by the weight F s of the generated wave volume above the
deck clearance c 0:07 m.
Fig. 11 shows the numerical domain of the test, with dimensions of 14  0.6 m2; the initial water level d at the wavemaker
is equal to 0.4 m. The simulated wave has amplitude equal to
0.1 m, and is generated imposing at the left boundary the kinematic conditions characterizing the solitary wave (Lee et al. [25]). The
top and the right boundaries are set open, allowing air and water to
escape out, respectively.
Grid convergence is assessed by comparing the peak pressure at
three points on the deck: X x=L 0:0, 0.2 and 0.4, where X 0 at
the front section of the deck. Three grids are subsequently investigated: 700  30 cells (uniform square dimension of 2 cm),
1400  60 cells (uniform square dimension of 1 cm) and
2000  120 cells (x-variable dimension with ner square grid
equal to 0.5 cm close the deck).
Table 3 reports the results of the three tested discretizations for
the COBRAS2 runs. Compared with the 1400  60 grid, the application of coarser cells leads to reductions in the peak pressure values
of 9.4%, 2.5% and 7.6% at the three selected points, respectively.

Fig. 11. Numerical set-up for the simulation of the wave impact on the deck by
French [15].

Additionally, lower grid dependence results from the adoption of

smaller cells, with increases in the peak pressure of 1.2%, 1.8%
and 0.95%, respectively.
Using COBRAS2, one computation with a 1400  60 grid
required approximately 5 h (4 h for incompressible COBRAS2 and
2 h for COBRAS0) on a 2.8 GHz Intel Core i7 machine; an increment
of a factor 2 in the grid resolution implies a running time increase
by a factor of 4. The intermediate grid resolution (1400  60) is
therefore considered the better choice in terms of computational
costs and numerical accuracy and is used for the current
simulation.
The computed loads on the deck are compared with the measurements taken by French [15]. In Fig. 12, the normalized force
F  , evaluated as F=F s , is plotted against the normalized time s equal

Fig. 12. Normalized force F  on the horizontal deck: COBRAS0 (dotted line),
COBRAS2 (solid lines) and the laboratory data (dots) by French [15].

122

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

p
to t g=d, where t; g and d are the time, the gravity acceleration
and the water depth, respectively. A positive sign is adopted for
the uplift force values.
The two solvers deliver different results. The presence of air
inertia is critical to correctly reproduce the wave propagation,
whereas air ow modeling inuences the reproduction of the loads
on the deck.
Good agreement with the laboratory data is observed for
COBRAS2, both in terms of the magnitude and the temporal evolution of the forces. The extreme values of the uplift and the downlift
loads are overestimated with COBRAS0, in which air is treated as a
void and the cushioning effect that occurs at impact is not reproduced. In addition, the break up of entrapped bubble unrealistically
induces numerical instabilities in COBRAS0, resulting in both highfrequency disturbances on the force signals and abnormal peaks in
magnitude.
The inuence of air compressibility on the estimation of waveinduced loads is also investigated. The most signicant difference
consists of a decrease in the downlift force magnitude (at time
5 < s < 6:5), but no signicant variations in the uplift loads are
evident.
Fig. 13 shows the map of the wave pressure pw on the deck surface X (abscissa) and through the time s (ordinate). Furthermore,
the gure shows that the impact on the deck begins at s 2.

Fig. 13. Map of the wave pressure pw along the deck X (abscissa) in time s
(ordinate).

Positive pressures are observed for X < 0:5 and for several time
instants 2 < s < 4. The contribution to the uplift load appears
to be given by the wave impacting the seaward side of the deck,
and sub-atmospheric pressures, responsible of downlift force (the
deck is not inundated), affect the leeward side.
Laboratory observations (Oumeraci et al. [36], Lugni et al. [32])
and recent numerical analyses (Plumerault et al. [39]) clearly show
that the impact pressures on vertical walls are strictly dependent
on the breaker shape and on the air engulfment in water. Therefore, representation of the correct breaker and its evolution on
the wall with two-phase models are crucial for realistic estimation
of the wave-induced loads on structures.
Fig. 14 (top) illustrates the numerical set-up of the simulations
with a on the wave impact on a vertical wall. The initial water
depth d0 at the wavemaker is set equal to 0.71 m, and a solitary
wave, with an amplitude of 0.21 m, is generated at the left boundary. The wave propagates over a 1/15 ramp and hits a vertical wall
of height A, equal to 2 m, located at the right-hand side of the channel. The chosen grid contains 1118  180 cells, and outputs are
saved with a frequency rate of 1000 Hz. This discretization is considered a good compromise between the computational costs (a
run with COBRAS2 takes approximately 12 h) and the spatial and
temporal reproduction of the process, as also explained in Lamberti
et al. [22].
To study the development of two breaker shapes at the impact,
the wall position is modied along the ramp for the two
simulations.
With reference to the nomenclature of Lugni et al. [32], Fig. 14
(bottom panels) shows the computed free-surface proles at
impact for the pure ip-through wave (a) and for the plunging
0
breaker characterized by a large (b ) pocket of entrapped air and
no air/water mixing. Broken waves are not investigated in this
work because the large amount of bubbles gives a strong tridimensionality to the impact dynamics that 2DV model cannot
properly reproduce.
In the following gures, the time t is displayed relative to the
wave impact time t0, and the wave pressure pw is depurated by
the hydrostatic pressure.
In Fig. 15, the velocity (top) and pressure (bottom) in the water
by the COBRAS2 results are reported at eight signicant time
instants for the impact (a). The black horizontal lines represent
the initial water level. No high and violent pressures develop on
the structure because the wave hits the wall before it breaks. In
this case, the air entrapped between the wave front and the structure is necessarily expelled in the upward direction, and not
through the water.

Fig. 14. Numerical set-up (top panel) and water prole on the wall (bottom panel) for the ip-through wave (a) and the plunging wave with a large air pocket (b ).

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

123

Fig. 15. Wave impact (a): snapshots of the water velocity (top panels) and pressure (bottom panels) by COBRAS2 at eight time instants.

Fig. 16. Wave impact (a): time series of the wave pressures along the wall at
y 0:71 m by COBRAS2 (-) and COBRAS0 (- -).

Fig. 17. Wave impact (a): time evolution of the wave pressure proles along the
wall by COBRAS2.

In agreement with the experimental evidence by Lugni et al.

[32], it is possible to distinguish among three stages that characterize the computed ow evolution along the wall. The wave
approaches the wall at t  t0 6 0:04 s, with a horizontal velocity
at the crest approximately equal to 2.0 m/s; next, in the wave
focusing stage, the wave crest and trough move toward each other
0:04 < t  t0 6 0:0 s with a comparable speed of 2.5 m/s, until
impact occurs and a vertical jet develops with a rising velocity of
approximately 10 m/s 0 < t  t0 6 0:15 s. As also experimentally
observed, the computed free-surface develops a backward

concavity during the wave rising 0:02 < t  t0 6 0:1 s, and the
jet seems to leave the wall. Furthermore, the inuence of air modeling in the dynamics appears to be signicant only when the ow
reverses (t  t0 > 0:6 s) and a closed cavity induces bubble entrapment t  t0  1:1 s.
Fig. 16 reports the time series of the computed wave pressures
at y 0:71 m (initial water level) from COBRAS0 and COBRAS2.
The typical church roof shape of the pressure signal characterizing the wave impact on a wall is evident. A peak pressure of
approximately 6 kPa rapidly develops in the water approaching
the wall and propagates with a celerity of the same order of magnitude as that of the wave trough. Fig. 17 displays the vertical distribution of the dynamic pressure on the wall at eight time
instants. The position of the maximum value nearly corresponds
to the initial water level d, whereas the increasing pressures from
the lower to upper water levels is indicative of a sloshing motion
up the wall.
The two solvers provide similar results in terms of magnitude
and time increase until reversal ow develops. The COBRAS2 model is able to reproduce the slow decay of the wave pressure, whereas the signals obtained from COBRAS0 present high frequency
instabilities, likely induced by the implemented boundary conditions at the airwater interface.
0
Different dynamics occur for the impact case b . The plunging
breaker encloses a large air pocket, due to a more pronounced
wave overturning.
As shown in Fig. 18, COBRAS2 properly solves the evolution of
the entrapped bubble in all stages of its existence, i.e., bubble
development when the breaker hits the wall 0 < t  t0 6 0:02 s,
the rise of bubble 0:02 < t  t0 6 0:10 s, the change in the
bubble shape when the wave is deected in a vertical jet
0:1 < t  t0 6 0:14 s, and its eventual escape across the freesurface t  t0 > 0:14 s.
In the panels of the gure, arrows show the velocity vectors in
the water during impact. The rising jet reaches a velocity of nearly
5 m/s upward, and the bubble rises on the wall with a maximum
speed of 1 m/s, roughly persisting in the water also when the wave
is deected backward.
Details of the impact are plotted with a higher temporal resolution in Fig. 19. Five snapshots in the upper panels show the wave
pressure during the impact, and the pressure time series at different wall positions (y = 0.72, 0.78, 0.82, 0.85, 0.89 m) are displayed
in the bottom panel.

124

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

Fig. 18. Wave impact (b ): velocity eld by COBRAS2 at nine time instants after the impact.

Fig. 19. Wave impact (b ): wave pressure by COBRAS2 at ve time instants after the impact (top panels) and time series (bottom panel) on the wall at y = 0.72, 0.78, 0.82, 0.85
and 0.89 m.

The pressure increases as the water approaches the solid

boundary, and an entrapped air cavity develops from the overturning of the wave crest advancing toward the wall. This bubble compresses and expands according to the polytropic law, moving
upward due to its buoyancy, and fragmenting and escaping

through the free-surface. In particular, at point D, the maximum

values of pressure on the wall are observed during the initial and
violent compression of the entrapped bubble. After nearly 1 ms following the impact, the air expansion and compression lead respectively to the pressure decrease and rapid increment.

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

125

Finally, the impact process is found to be dependent on whether

the air is numerically considered; in fact, the water ow propagates with a typical celerity related to the air-induced resistance.
Wave breaking could occur earlier or later depending on whether
air is accounted for, and in the case of numerical simulations of
wave-induced loads on structures, the strong correlation between
the impact pressure and the breaker shape should be accurately
modeled to avoid numerical instabilities and errors in the force
estimation.

non-negligible. This development provides the opportunity to

extend the numerical analysis to complex processes, i.e., windgenerated waves (among others, Longo et al. [28]), and investigate
the role of wind in wave run-up and overtopping.
In particular, the current solver contains the original feature
used to address ows through of permeable structures. The air
water-porous media interactions can be properly modeled to
investigate the inuence of sediment sizes and the distribution of
gravel-sand beaches on ooding hazards.

5. Conclusions

References

This paper presents a novel two-phase model, implemented with

the inclusion of compressible gas, and an investigation of the role of
air modeling in numerical simulations related to coastal issues was
carried out. The governing equations for two-phase ows were
derived using the FavreReynolds average and appropriate model
closures. Validation cases and numerical applications were
described to demonstrate the accuracy and features of the solver.
The inuence of air modeling was found to be non-negligible in
coastal studies because the resistance induced by the air during
wave propagation and the reproduction of entrapped bubbles into
the water led to more realistic processes. Indeed, an improvement
in the representation of the ow dynamics was shown.
In addition, a comparison with the single-phase solver was discussed. Breaker height and position, which largely inuence the
wave-structure interactions, are reproduced with higher accuracy
using the two-phase model. As the water propagates in the air
rather than in a void (as in the case of COBRAS0), wave breaking
generally occurs later due to the greater effort of the water to push
away the viscous air as modeled with COBRAS2.
Due to the correlation between the breaker shape and the
wave-induced loads, the processes that develop during the wave
impacts on structures take advantage of the air modeling inclusion,
i.e., the reproduction of bubble entrapment and cushioning
processes.
However, one of the main drawbacks of COBRAS2 is the computational cost. In general, the running time increases by a factor of 4
(3 if excluding air compressibility) compared with the same singlephase solver run. This aspect is not negligible when numerical analyses are chosen for coastal studies. The air modeling is of great
importance, and the level of accuracy to be reached in the results
must be carefully evaluated before adopting a solver. To overcome
such limitations, a future work will propose a single-phase model
for water waves, in which novel appropriate free-surface boundary
conditions will take into account the multiphase mixing layer
(Brocchini [3]) that devops at the weakly splashing airwater
interface. With this approach, more reasonable CPU times might
be possible.
The other crucial aspect consists of the model bi-dimensionality, which might inuence the numerical results, especially during wave breaking and impact.
The 3D effects can be expected in terms of generally reduced
wave loads (Bullock et al. [6]), air leakage that might occur in the
sideways directions, and turbulence eld that might develop along
the waterfront. But Peregrine [37] reported that wave pressures
could be enhanced by such aspects as wave focusing and the shape
of the incident waves at the wall.
Nevertheless, a natural extension of the model will be the
development of the third dimension. Higuera et al. [20] recently
began working on this issue, adopting the OpenFoam code, and
the effective development of 3D eddies after wave impact on a wall
is described.
Finally, the model is demonstrated as a promising tool for
addressing complete coastal dynamics, in which air modeling is

[1] Aureli F, Maione U, Mignosa P, Tomirotti M. Unsteady ows due to

dambreaking. Part I. Numerical modelling and comparison with literature
experimental data (in italian). LAcqua 1998;4:2942.
[2] Bagnold RA. Interim report on wave-pressure research. J Inst Civil Eng
1939;12:20126.
[3] Brocchini M. Free surface boundary conditions at a bubbly/weakly-splashing
airwater interface. Phys Fluids 2002;14:183440.
[4] Brocchini M, Peregrine DH. The dynamics of strong turbulence at free surfaces.
Part 2. Free-surface boundary conditions. J Fluid Mech 2001;449:25590.
[5] Brocchini M, Peregrine DH. The dynamics of turbulent free surfaces. Part 1.
Description. J Fluid Mech 2001;449:22554.
[6] Bullock GN, Obhrai C, Peregrine DH, Bredmose H. Violent breaking wave
impacts. Part 1: Results from large-scale regular wave tests on vertical and
sloping walls. Coast Eng 2007;54(8):60217.
[7] Chorin AJ. Numerical solution of the NavierStokes equations. Math Comput
1968;22:74562.
[8] Christensen ED. Large eddy simulation of spilling and plunging breakers. Coast
Eng 2006;53:46385.
[9] Clementi E, Gaeta MG, Lamberti A. Filtration through low crested structures in
2d: experimental and numerical investigations. In: Proc Coastal Structures
2007; 2007. p. 92738.
[10] Colagrossi A, Colicchio G, Lugni C, Brocchini M. A study of violent sloshing
wave impacts using an improved sph method. J Hydraul Res 2010;48:94104.
[11] Colagrossi A, Landrini M. Numerical simulation of interfacial ows by
smoothed particle hydrodynamics. J Comput Phys 2003;191(2):44875.
[12] Colicchio G, Landrini M, Chaplin J. Level-set computations of free surface
rotational ows. J Fluids Eng Trans ASME 2005;127(6):111121.
[13] Drew DA, Passman S. Theory of multicomponent uids applied mathematical
sciences, vol. 135. Springer; 1999.
[14] Faltinsen OM, Landrini M, Greco M. Slamming in marine applications. J Eng
Math 2004;48:187217.
[15] French JA. Wave uplift pressures on horizontal platforms. In: Proc Civil Eng in
the oceans conference; 1970. p. 187202.
[16] Fhrbter A. Air entrainment and energy dissipation in breakers, In: Proc 12th
Int Conf Coast Eng; 1970. p. 39198.
[17] Galvin CJ. Breaker type classication on three laboratory beaches. J Geophys
Res 1968;73:36519.
[18] Guanche R, Losada IJ, Lara JL. Numerical analysis of wave loads for coastal
structure stability. Coast Eng 2009;56:54358.
[19] Hattori M, Arami A, Yui T. Wave impact pressure on vertical walls under
breaking waves of various types. Coast Eng 1994;22:79114.
[20] Higuera P, Lara JL, Losada IJ. Simulating coastal engineering processes with
openfoam. Coast Eng 2013;71:11934.
[21] Jaluria Y, Torrance KE. Computational heat transfer. Hemisphere Pub; 2002.
[22] Lamberti A, Martinelli L, Gaeta MG, Tirindelli M, Alderson J. Experimental
spatial correlation of wave loads on deck front. J Hydraul Res
2011;49(sup1):8190.
[23] Lara JL, Garcia N, Losada IJ. Rans modelling applied to random wave interaction
with submerged permeable structures. Coast Eng 2006;53:395417.
[24] Launder BE, Morse A, Rodi W, Spalding D. Prediction of free shear ows: a
comparison of the performance of six turbulence models. In: Free shear ow,
NASA conference; 1972. p. 361426.
[25] Lee JJ, Skjelbreia JE, Raichlen F. Measurements of velocities in solitary waves. J
Waterway Port Coast Ocean Eng 1982;108(2):20018.
[26] Lin P, Liu PLF. A numerical study of breaking waves in the surf zone. J Fluid
Mech 1998;359:23964.
[27] Lin P, Liu PLF. Turbulence transport, vorticity dynamics, and solute mixing
under plunging breaking waves in surf zone. J Geophys Res Oceans
1998;103(C8).
[28] Longo S, Chiapponi L, Clavero M, Mkel T, Liang D. Study of the turbulence in
the air-side and water-side boundary layers in experimental laboratory wind
induced surface waves. Coast Eng 2012;69:6781.
[29] Losada IJ, Lara JL, Christensen ED, Garcia N. Modelling of velocity and
turbulence elds around and within low-crested rubble-mound breakwaters.
Coast Eng 2005;52:887913.
[30] Lubin P, Glockner S. Numerical simulations of breaking solitary waves. In: Proc
33th int conf coast eng; 2012.
[31] Lubin P, Vincent S, Abadie S, Caltagirone JP. Three-dimensional large eddy
simulation of air entrainment under plunging breaking waves. Coast Eng
2006;53:63155.

126

M.G. Gaeta, A. Lamberti / Computers & Fluids 111 (2015) 114126

[32] Lugni C, Brocchini M, Faltinsen OM. Wave impact loads: the role of the ipthrough. Phys Fluids 2006;18(12):122101. 17 pp.
[33] Martin JC, Moyce W. An experimental study of the collapse of liquid columns
on a rigid horizontal plane. Philos Trans R Soc London, Ser A 1952;244:31224.
[34] Melville W. The role of surface-wave breaking in airsea interaction. Annu Rev
Fluid Mech 1996;28:279321.
[35] Monaghan J. Smoothed particle hydrodynamics smoothed particle
hydrodynamics smooth particle hydrodynamics. Annu Rev Astron Astrophys
1992;30:54374.
[36] Oumeraci H, Bruce T, Klammer P, Easson W. Piv-measurements of breaking
wave kinematics and impact loading of caisson breakwaters. In: Proc 4th Int
Conf Port Eng Dev Countries; 1995. p. 2394410.
[37] Peregrine DH. Water wave impact on walls. Ann Rev Fluid Mech
2003;35:2343.
[38] Peregrine DH, Bredmose H, Bullock G, Hunt A, Obhrai C. Water wave impact on
walls and the role of air. In: Proc 30th Int Conf Coast Eng; 2006. p. 4494506.

[39] Plumerault LR, Astruc D, Villedieu P, Maron P. A numerical model for aeratedwater wave breaking. Int J Numer Methods Fluids 2012;69:185171.
[40] Sokolichin A, Eigenberger G. Applicability of the standard k-turbulence model
to the dynamic simulation of bubble columns. Part i. Detailed numerical
simulations. Chem Eng Sci 1999;54(1314):227384.
[41] Ting F, Kirby J. Observation of undertow and turbulence in a laboratory surf
zone. Coast Eng 1994;24(12):5180.
[42] Torres-Freyermuth A, Losada IJ, Lara JL. Modelling of surf zone processes on a
natural beach using rans equations. J Geophys Res Oceans 2007;112. http://
dx.doi.org/10.1029/2006JC004050.
[43] Wemmenhove R, Loots G, Veldman A. Numerical simulation of hydrodynamic
wave loading by a compressible two-phase model. In: European Conference on
computational uid dynamics; 2006.
[44] Wood DJ, Peregrine DH, Bruce T. Wave impact on a wall using pressureimpulse theory. i: Trapped air. J Waterway Port Coast Ocean Eng 2000;126.