Selection of JONSWAP Spectra Parameters during

Water-Depth and Sea-State Transitions

Juan Gabriel Rueda-Bayona, Ph.D.1; Andrés Guzmán, Ph.D., M.ASCE2; and Juan José Cabello Eras, Ph.D.3

Abstract: The design of marine structures requires the simulation of wave parameters that consider sea-state and water-depth transitions.
Proper selection of the model coefficients (e.g., alpha and gamma of the JONSWAP spectra) is then required, because of the wave-
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.

hydrodynamic nonlinearities during these ocean processes. Therefore, the model coefficient selection should be tested using a nonlinear anal-
ysis to assess the effect of the selected spectra coefficients over the modeled wave parameters. The present study performed a design of
experiment (DOE)-analysis of variance (ANOVA) and probability analysis to assess the effect of alpha and gamma parameters over the sig-
nificant wave height (Hs) and peak period (Tp) during sea-state and water-depth transitions. The DOE-ANOVA demonstrated for the mean
and extreme wave states of the study area that alpha and gamma parameters positively affect the Hs behavior in deep and intermediate waters.
Furthermore, the standardized effects of alpha and gamma over the Tp during extreme wave states suggest quadruplets of wave–wave inter-
actions. The joint and normal probability distributions of alpha and gamma for extreme and normal waves showed a Gaussian distribution,
allowing identification of specific alpha and gamma values for the JONSWAP spectra model. The selected alpha and gamma parameters
were then validated through the comparison of the modeled Hs (JONSWAP) against other local studies. Considering its relevance in design
strategies for offshore structures, this research contributed to the understanding of the nonlinear effects of alpha and gamma parameters over
the Hs and Tp during variations of water depth and wave states, easing the selection of the model coefficients. DOI: 10.1061/(ASCE)
WW.1943-5460.0000601. © 2020 American Society of Civil Engineers.
Author keywords: JONSWAP spectra; DOE-ANOVA; Probability; Waves; Numerical modeling.

Introduction modeler may understand the nonlinear effects of model coefficients

Scientific and engineering activities dedicated to coastal and off-
shore studies require in situ (measured) or modeled wave data to
understand local processes and to facilitate marine structures de-
sign. A well-modeled wave spectrum considers the effects of cli-
mate variability and water-depth reduction along the direction of
wave propagation; these effects may generate nonlinear processes,
such as quadruplets or wave shoaling (Holthuijsen 2010). A lack of
consideration of the nonlinear effects that generate the control pa-
rameters (e.g., coefficients) of numerical models may lead to over-
or underestimation of the behavior of wave parameters, such as sig-
nificant wave heights (Hs) and peak periods (Tp). As a result, the
structure’s design or research conclusions may be liable to error, af-
fecting the formulation of sustainable engineering solutions.
The generation of 2D–3D wave-modeled data requires a calibra-
tion and validation process, which can be optimized so that the
1 gamma) where the user may simulate the wave spectra for any site
Professor, Engineering Faculty, Civil Engineering, Water and Energy
(AyE) Research Group, Universidad Militar Nueva Granada, Bogotá:
Carrera 11 No. 101-80, Bogotá 1101111 (corresponding author). ORCID:
https://orcid.org/0000-0003-3806-2058. Email: juan.rueda@unimilitar.edu
.co; ruedabayona@gmail.com
2
Professor, Research Group for Structures and Geotechnics (GIEG), In-
stitute for Sustainable Development (IDS), Dept. of Civil and Environmen-
tal Engineering, Universidad del Norte, Km 5 via Puerto Colombia, Bloque
K, 8-33 K, Barranquilla, CO 081007. ORCID: https://orcid.org/0000-0003
-2472-1390. Email: faguzman@uninorte.edu.co
3
Professor, Energy Dept., Universidad de la Costa, Calle 58 # 55–66,
Barranquilla, CO 080002. Email: jcabello2@cuc.edu.co
Note. This manuscript was submitted on May 21, 2019; approved on
April 28, 2020; published online on July 17, 2020. Discussion period
open until December 17, 2020; separate discussions must be submitted
for individual papers. This paper is part of the Journal of Waterway,
Port, Coastal, and Ocean Engineering, © ASCE, ISSN 0733-950X.
over the wave parameters. In addition, a structural design that di-
rectly uses 1D wave models such as JONSWAP spectra may calcu-
late suitable structural and geometry parameters if the designer
knows the nonlinear interactions of the model parameters.
The JONSWAP spectra model (JONSWAP stands for JOint
North Sea WAve Project) is the most frequently used model for re-
search and design; it is considered universal for idealized
fetch-limited conditions and because of its applicability to variable
wind regimes in deep waters during storms and hurricanes
(Chakrabarti 2005). The advantage of the JONSWAP model is
that it is a sound representation of the swell spectra, particularly
in the high-frequency range (Lucas and Guedes Soares 2015). Be-
cause of these advantages, several consultants and research projects
use the JONSWAP model (Calini and Schober 2017; Cifuentes and
Kim 2017; Pascoal et al. 2017; Sun and Zhang 2017).
To consider the nonlinear wave processes, it is necessary to set
specific spectra parameters for the JONSWAP equation (alpha and
and weather conditions. If the user applies the JONSWAP spectra
with constant parameters for a broad study area, they may underes-
timate or overestimate the peak energy of the wave spectra. Thus,
simulating the spectral shape through models such as JONSWAP
must consider such changes on the weather and seafloor.
The shape of the JONSWAP spectra is dependent on parameters
such as the alpha coefficient (scale parameter) and the gamma co-
efficient (peak enhancement factor). The calculus of alpha and
gamma parameters depends on the dimensionless relations of the
peak frequency. However, it is difficult to set accurate values for
alpha and gamma because of their large dispersion (Holthuijsen
2010).
Several studies suggest consideration of constant values for
alpha or gamma (Boukhanovsky et al. 2007; Boukhanovsky and
Guedes Soares 2009; Ochi and Hubble 1976; Ortega et al. 2011).

© ASCE 04020038-1 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 The literature reported the use of constant values to assess the
JONSWAP spectra: the Donelan and Scott spectra (Sanil Kumar
and Ashok Kumar 2008), the estimation of the wave-crest distribu-
tion of nonlinear waves in shallow waters (Wang 2014), and the
analysis of the bottom slope effect over the nonlinear transforma-
tions of irregular waves (Dong et al. 2014). Ji et al. (2018) proposed
an empirical formula to estimate the maximum wave setup through
a coupled wave-current model. That study set a constant gamma of
3.3 for the control parameters of the SWAN model (Deltares
2014a). Liu et al. (2007) generated irregular waves via JONSWAP
spectra to estimate the wave-current forces acting on a pile group
on the base foundation of a bridge; the research reported a constant
value of 2.8 for the gamma parameter.
There is evidence in the literature of different gamma values in
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
studies focused on the analysis of the wave spectra (Mackay 2011,
2016; Montazeri et al. 2016; Sakhare and Deo 2009; Wijaya and
Van Groesen 2016; Zanaganeh et al. 2009). For instance, Myrhaug
(2018) used distributions to provide probabilistic properties of the
spectra wave steepness and also varied the gamma parameter of the
JONSWAP spectra to generate the wave data required for the anal-
ysis. Calini and Schober (2017) characterized the JONSWAP rogue
waves through inverse spectra data. The study used a range of
0.008 < alpha < 0.02, and 2 < gamma < 8 to generate the random
wave data.
The literature also reports the utilization of DOE-ANOVA for
nonlinear analysis and its application in diverse research topics.
Using DOE-ANOVA, Power et al. (2019) identified the factors
that generate changes in run-up levels; Derschum et al. (2018) stud-
ied the effect of debris loads in building codes such as FEMA
P-646 (FEMA and NOAA 2012) and ASCE 7-16 (ASCE 2017);
Hanley et al. (2014) assessed soft engineering alternatives for
coastal protection (e.g., sand nourishment); and Fragasso et al.
(2019) validated the coefficient selection of the Yeoh hyperelastic
material model. Furthermore, Young and Scully (2018) evaluated
vessel performance during low and high wave conditions through
ANOVA to assess the sheltering effectiveness of three coastal
inlet structures.
The specialized literature established the necessity of testing the
alpha and gamma ability when considering the nonlinear effects in
the modeled wave spectra during wave-state and water-depth tran-
sitions. The assumptions adopted in the literature—such as a cons-
tant gamma parameter of JONSWAP spectra used for the boundary
conditions of the numerical models (e.g., SWAN) or related to 1D
wave parameters modeling—could have limited the expected re-
sults. As a result, it is necessary to assess whether having JONS-
WAP parameters in situ is essential, because the spectra differ at
every location for the different oscillations of climate variability
(hours, days, and months). Furthermore, the energy scale coeffi-
cient (alpha) is a must-have variable during sea conditions because
this parameter controls the wave frequency and modulates the spec-
tral shape from the sea-to-swell state.
In order to select the model coefficients of the JONSWAP spec-
tra considering the nonlinear interactions generated during water-
depth and sea-state transitions, this study proposes the following
methodology: (1) hydrodynamic modeling, (2) wave climate anal-
ysis, (3) generation of coefficients, (4) DOE–ANOVA and proba-
bility analysis, and (5) validation. The hydrodynamic modeling
generated wave-time series to feed the GA for the generation of
the alpha and gamma data sets. The DOE-ANOVA then assessed
the effect of gamma and alpha parameters over Hs and Tp. The
probability analysis generated normal, log-normal, and joint prob-
ability distributions to select alpha and gamma parameters for spe-
cific sites according to the wave-state and the water-depth
transitions. Finally, the validation compared the calculated Hs
© ASCE
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
using the selected alpha and gamma coefficients against Hs data re-
ported by secondary studies.
Methodology
To analyze the nonlinear effects of JONSWAP parameters, this
study followed the following stages: (1) hydrodynamic modeling,
(2) wave climate analysis, (3) generation of coefficients, (4)
DOE-ANOVA and probability analysis, and (5) validation. This re-
search implemented a 3D wave-hydrodynamic model Delft3D
(Deltares 2014a, b) to simulate wave parameters, followed by the
wave climate characterization and definition of local sea states.
Subsequently, in order to generate the coefficients (alpha and
gamma), a genetic algorithm (GA) model was fed with the modeled
Hs and Tp by Delft3D to generate the alpha and gamma parameters
using the JONSWAP spectra equation [Eq. (1)]. The alpha and
gamma coefficients generated by the GA were then assessed
through the DOE-ANOVA and probability analysis to determine
the representative alpha and gamma parameters for specific places,
considering the sea-state and water-depth transitions. Finally, it was
validated through the Eq. (2) (Hasselmann et al. 1973) using the se-
lected alpha and gamma parameters to generate a local 1D wave
spectra. Accordingly, Eq. (3) (Holthuijsen 2010) was used to calcu-
late the Hs, which is validated against secondary reported wave
data from other studies:

∞
−4
S(ω)JONSWAP = αg 2 ω−5 e−1.25(ω/ωp ) γ δ dω (1)
0
where

−((ω−ωp )2 /2σ 2(a,b) ω2p ) σ a=0.07 when, ω ≤ ωp
δ=e & σ (a,b) =
σ a=0.09 when, ω > ωp
α = energy scale coefficient; g = Earth’s gravity (9.80665 m/s2);
ω = frequency (rad/s); ωp = peak frequency (rad/s); γδ = peak en-
hancement factor; γ = peak enhancement coefficient; δ = describes
the width of the region near the peak; and σ(a,b) = shape parameters.

∞
−4
S(ω)JONSWAP = αg 2 ω−5 e−1.25(ω/ωm ) γ δ dω (2)
0
where
ωm = 2π(g/U )(gx/U 2 )−0.33
x = fetch length (25 km); and U = wind speed (5 m/s).
√
Hs = 4.004 m0 (3)
where m0 = zero-order spectral moment.
The Delft3D model utilized in stage (1) of the methodology (hy-
drodynamic modeling) can solve currents and waves through
double-way communication between a hydrodynamic module
and the SWAN wave model (Deltares 2014a); the wave model sol-
ves the action–balance equation (Booij et al. 1999). The two-way
communication mode of Delft3D considers the interactions be-
tween waves and currents, decreasing the probability of over- or
underestimation in modeled waves, water levels, and currents
(Elhakeem et al. 2015; Escobar 2011; Garcia et al. 2015; McCombs
et al. 2014; Rueda-Bayona et al. 2019). The hydrodynamic module
of Delft3D solves the Navier–Stokes equations by considering
the Boussinesq hypothesis. The model uses a flexible mesh and
applies the finite difference method for solving the RANS
(Reynolds-averaged Navier–Stokes) equations. We selected the
turbulence k-ϵ model (Uittenbogaard et al. 1992) and the ADI
04020038-2 J. Waterway, Port, Coastal, Ocean Eng.
 (a) (b)

Fig. 1. (a) Grid model for the study area; and (b) location of virtual points, wave buoy (DIMAR), and water level station (Casa Pilotos). Projected
coordinates in Magna-Sirgas Bogota Zone; units in meters.
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.

Table 1. Locations of wave buoy and observation points

Latitude Longitude Water depth
ID (N) (W) (m)
DIMAR wave buoy 11.161000 74.681000 150
Water level station (Casa 11.097769 −74.852839 0
pilotos)
Point A 11.167972 74.677314 105
Point B 11.147295 74.691496 28

Table 2. Time period of water level and significant wave-height data for
the model calibration Fig. 2. Information flow scheme between the Delft 3D Flow and
SWAN wave model.

Data Time period (dd-mm-yyyy-hh) Time resolution

Water level 01-01-2012-00 to 02-04-2012-11 1h
(Casa Pilotos)
Hs- (DIMAR buoy) 21-02-2009-05 to 25-03-2009-01 3h
scheme (alternating direction implicit) to solve the temporal
components.
Wave-Hydrodynamic Model Setup
The procedure for applying the GA model when in situ wave spec-
tra are not available comprised a calibration of the Delf3D and
SWAN models to feed the GA model with Hs and Tp records.
As a result, we implemented the hydrodynamic and wave models
close to the Magdalena River delta in the north of Colombia. In
the study area (Fig. 1), the available in situ data is scarce because
the equipment deployment is restricted due to navigation security
restrictions (Restrepo et al. 2016). Then, we solved the scarce
wave in situ data at the study area by implementing the Delft 3D
model to obtain wave-time series at different sea states and water
depths.
For this research, the Delft 3D Flow module has sigma coordi-
nates (10 layers) with flexible mesh and cell size in a range between
50 km and 500 m [Fig. 1(a)]; the utilized bathymetry data for the
hydrodynamic model are from the ETOPO1 database (NOAA
2018a). For the transport boundary conditions of Deflt3D, we
used the met-ocean thermohaline data are from the World Ocean
Atlas, WOA2013 (Locarnini et al. 2013; Zweng et al. 2013) and
the wind and heat fluxes are from the NCEP–North American Re-
gional Reanalysis: NARR database (NOAA 2016). Also, the water
level data for the open boundary conditions were retrieved from the
numerical results of the GRENOBLE model (Le Provost et al.
1994).
The peak period and significant wave-height data used for the
SWAN model boundary conditions come from the WaveWatch
III model database (NOAA 2018b); the wave data have a spatial
resolution of 0.17° and is at a virtual point of 11.5 N and 74.5
W. The wave model has a grid with a cell size of 500 m, with a sim-
ilar extension of the external grid cells of the Delft3D model.
Table 1 indicates the locations [Fig. 1(b)] of a buoy that belongs
to the Maritime General Direction (DIMAR, in Spanish) of Colom-
bia that measured wave data, and the location of a water level sta-
tion known as Casa Pilotos; the data of these locations will be
utilized for the model calibration. The locations Point A and
Point B represent the monitoring points of the numerical model
where the wave data will be numerically generated and required
for the (1) stages, (2) wave climate analysis, (3) generation of co-
efficients, and (4) DOE-ANOVA and probability analysis. Table 2
shows the resolution and time period of the in situ water level and
significant wave height data for the model calibration (Fig. 1,
Table 1).
We depicted the information flow scheme among data and
Delft3D and SWAN models in Fig. 2. The interaction between
models is in a two-way mode. Deflt3D solves the hydrodynamics
and provides currents, wind, tides, and the bed-level evolution to
the SWAN model. Thereafter, SWAN solves wave hydrodynamics
and gives wave parameters to the Delft3D model. The two-way
communication between the Delft3D and SWAN models allowed
proper simulating of the physical processes.
The calibration of the Delft3D model showed that the correla-
tion between in situ data (Casa Pilotos) and modeled water levels
resulted in a determination coefficient of 0.79, with p-value ≪ 0.05
and 95% confidence [Fig. 3(a)]. The calibration of the significant

© ASCE 04020038-3 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 (a)
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
(b)
Fig. 3. Model calibration results: (a) water levels; and (b) significant wave height.
wave heights resulted in a determination coefficient of 0.70 with
p-value ≪ 0.05 and 95% confidence [Fig. 3(b)].
The design of experiments (DOE) is a planned strategy of mea-
surements or observations (runs) that evaluates independent vari-
ables (factors) to obtain a response from dependent variables.
The objective of DOE is to optimize processes or analyze the be-
havior of variables. For its part, the analysis of variance
(ANOVA) verifies the hypothesis that the means of two or more
populations are equal; ANOVA sets two hypotheses to test the sig-
nificance of one or more factors; the means of the response variable
are compared according to the different levels of the factors. If all
the means of the levels of the factors (means of the population) are
equal, then the hypothesis is null, while the alternative hypothesis
shows that at least one mean is different. In a two-way ANOVA or
factorial ANOVA, two independent variables are affecting the de-
pendent variable. The factors selected comprise two or three levels
(low, middle, and high), which can be the minimum, mean, and
maximum values of a dataset of one independent variable. The
ANOVA then determines the main effect and the interaction effects
of the factors (independent variables) over the responses (depen-
dent variable). Three assumptions must be considered applying
the ANOVA: normality, independence, and homoscedasticity
(Montgomery 2017).
For the factorial ANOVA or factorial design, the levels of the
factors can be denoted with a negative or positive sign. For in-
stance, Factor A can have three levels: low (A−), middle (A) and
high (A+). The interaction between A with another factor (B) can
be written as AB. To analyze the effects of the factor over the re-
sponse, the Pareto chart and the main effects plot allow visualiza-
tion of the interactions among the independent and dependent
variables. The Pareto chart represents the absolute values of the
© ASCE
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
standardized effects from the minimum to the highest effect and
plots a vertical line over the horizontal bars to show which effects
are statistically significant. The main effects plot shows the re-
sponse of each dependent variable to each factor level (low, middle,
and high) through a plot line. Hence, the probability analysis allows
identification of the representative value of a population and the re-
lation between the wave parameters through probability distribu-
tion curves and joint probabilities graphs. The probability
analysis considers the limited central theorem to adjust the data dis-
tribution into a normal distribution.
The DOE-ANOVA analysis performed in this study targets the un-
derstanding of the modeled significant wave heights (Hs) and peak pe-
riods (Tp) behavior when varying the two main parameters (alpha and
gamma) of JONSWAP spectra model. To do so, we must identify rep-
resentative normal and extreme sea states of the study area, to classify
and group Hs and Tp datasets for the DOE-ANOVA analysis. In the
section Wave Climate Analysis, seven sea states were classified;
then, we extracted the modeled Hs and Tp hourly data of each sea
state (Table 3). Also, to generate alpha and gamma parameters of
the seven classified sea states, we applied a genetic algorithm model
seen in the Generation of Coefficients section. As a result, the Hs,
Tp, alpha, and gamma datasets were organized for points A and B
(Fig. 1, Table 1) to perform the DOE-ANOVA.
Considering that the seven Hs and Tp hourly dataset are contin-
uous samples for the DOE-ANOVA, and the data (population)
have a different order of magnitude, this study used the maximum
value technique for normalizing, which involved dividing the sam-
ples of each of the seven datasets by its maximum value. As a re-
sult, the maximum value technique allows limiting the maximum
range up to 1 and the minimum range to the minimal value of
each dataset. The normalized datasets enable to set the level factor
04020038-4 J. Waterway, Port, Coastal, Ocean Eng.
 of the DOE-ANOVA, where the lowest factor is the minimal nor-
malized sample value, and the highest factor is the maximum nor-
malized sample value equal to 1.
Results
The methodology applied to analyze the nonlinear effects of the
JONSWAP spectra parameters over the modeled significant wave
heights (Hs) and peak periods (Tp) followed the recommended
Table 3. Wave event classification for the DOE-ANOVA analysis
Start date End date Max
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
strategy. First, we implemented the wave-hydrodynamic model to
generate Hs and Tp time series. Second, we defined the sea states
of the study area through a wave climate analysis using cumulative
probability curves. We then applied the GA model to generate
the alpha and gamma parameters. Finally, we performed the
DOE-ANOVA and joint probability analysis to define the specific
alpha and gamma values for the study area.
Wave Climate Analysis
The study used the fourth quartile of probability (95%) of the wave
data to select the wave states for the DOE-ANOVA analysis. This
wave data comes from the WaveWatch III virtual point previously
Tp
mentioned (11.5 N and 74.5 W) to graph the cumulative probability

(dd/mm/yyyy (dd/mm/yyyy Hs associated

Sea state hh:mm) hh:mm) (m) (s) curve for extreme waves (Rueda Bayona 2015, 2017). After that,
the wave height of 95% of the cumulative probability of extreme
Event 1 25/01/2001 0:00 03/03/2001 19:00 2.73 6.49 waves was considered by applying the peak over threshold
(extreme)
(POT) technique (Holthuijsen 2010), which allows the extraction
Event 2 11/03/2001 20:00 19/03/2001 15:00 2.17 6.28
(extreme) of extreme wave records in time series. In order to identify the
Event 3 17/06/2001 23:00 02/07/2001 14:00 2.27 6.63 joint distribution of wave heights and directions, we did a wave
(extreme) rose—a frequency analysis of wave vectors with an oceanographic
Event 4 09/09/2001 20:00 15/09/2001 12:00 1.28 6.14 convention where north is equal to 0° derived from a 90° rotation of
(normal) the Cartesian coordinates.
Event 5 12/10/2001 23:00 20/10/2001 15:00 1.22 6.02 The mean and extreme wave conditions identified by the cumu-
(normal) lative probability of significant wave height (Hs) [Fig. 4(a)] indi-
Event 6 07/11/2001 9:00 15/11/2001 12:00 1.52 5.95 cated a value of 1.78 and 2.87 m for the second and fourth
(normal)
quartiles. We selected the fourth quartile (95%) of Hs (2.87 m) as
Event 7 10/12/2001 20:00 17/12/2001 14:00 1.46 5.59
(normal)
the base for the extreme or maximum wave heights. Then, we ex-
tracted all the Hs records over the 2.87 m considered extremes to

(a) (b)

(c) (d)

Fig. 4. Wave climate analysis of the virtual buoy at 11.5 N and 74.5 W of the WaveWatch III database: (a) cumulative probability curve; (b) extreme
cumulative probability curve; (c) wave rise; and (d) POT applied to WaveWatch III virtual point for 30-year three-hourly records.

© ASCE 04020038-5 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 generate a new cumulative probability curve [Fig. 4(b)]. We then
selected the value of 3.59 m as the base (black line) for the POT
(Holthuijsen 2010) [Fig. 4(d)] for the fourth black quartile of the
new probability curve [Fig. 4(b)]. The subsequent wave direction
statistics showed waves from 45° (northeast) [Fig. 4(c)].
The POT of the WaveWatch III virtual point identified that,
from the 30 years of wave data at the study area, 2001 was the
year with the highest Hs (4.68 m) [Fig. 4(d)]. We then did a POT
for each virtual point of 2001 depicted in Fig. 1(b); the results
showed that three wave events exceeded the threshold of 2.12 m
of wave height (Fig. 5).
The wave time series of 2001 showed wave growth and dissipa-
tion, evidencing a sea-state development during extreme and nor-
mal conditions (Fig. 5). Consequently, we identified and
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
We then evaluated the GA model efficiency and effectiveness
for this study through a performance test for determining the min-
imum number of chromosomes for solving the cost function and
the minimum time for the calculations. According to the test results
(Fig. 6), 100 chromosomes reached a correlation of 0.9923 between
the target and artificial Hs result; after 100 chromosomes, the cor-
relation increased.
Furthermore, the test for Ep and Hs indicated that the correlation
remained at about 0.80. The computational time for 100 chromo-
somes was 24.96 s, 53.88 s for 200 chromosomes, and 735.05 s
for 400 chromosomes. According to the results, this research
used 100 chromosomes to solve the cost function efficiently; the
population-size test was evidence of the efficiency of the GA to
find a robust solution inside a feasible region with manageable

analyzed seven wave events through the DOE-ANOVA (Table 3).
Generation of Coefficients
This research generated alpha and gamma data sets through a val-
idated GA model (Rueda Bayona 2017; Rueda-Bayona et al. 2020)
fed with modeled Hs and Tp time series generated by the calibrated
Delft3D model (Deltares 2014a, b). The GA model generates, mu-
tates, and crosses populations of alpha and gamma datasets (chro-
mosomes), solves the JONSWAP spectra, and calculates Hs
through the zero-order moment till the targeted hourly Hs value
is reached. The GA model selects the optimal hourly alpha and
gamma, considering the minimum difference between the calcu-
lated Hs through JONSWAP spectra and the targeted Hs generated
by the Delft3D model. More details of the functioning and structure
of the GA model may be found in Rueda Bayona (2017) and
Rueda-Bayona et al. (2020).
computational time. With 100 chromosomes, the GA found the
alpha and gamma values for the time series (2001) of the Points
A and B. Fig. 7 shows the numerical solution for Point A, where
the residuals remain near to 0.05 m; some peaks were close to
0.4 m did not significantly affect the total correlation coefficient
of 0.9923. Given the scatter of the alpha and gamma, it was neces-
sary to apply a probabilistic analysis.
Fig. 7(a) depicts the comparison of the targeted Hs (in situ) gen-
erated by the Delft3D model and the modeled Hs (artificial) by the
GA model. In Fig. 7(b) is evidenced that the GA model solves the
JONSWAP spectra and selects successfully the calculated Hs using
the zero-order moment for each hourly time interval, considering
the minimum difference (residual) between the in situ and artificial
Hs. Figs. 7(c and d) show the utilized alpha and gamma values by
the GA model for each hourly time interval. Then, it is observed
that alpha and gamma did not depict a defined behavior along the
time [Figs. 7(c and d)], evidencing that these two parameters

Fig. 5. Peak over threshold for 2001 at Points A and B.

Fig. 6. Population-size test for determining the efficient number of chromosomes. According to the GA, the variables analyzed are peak energy (Ep),
Hs objective (Hs-obj), and artificial (Hs art). (Data from Rueda Bayona 2017.)

© ASCE 04020038-6 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 (a)

(b)

(c)
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.

(d)

Fig. 7. Numerical results of the GA for Point A with hourly time resolution: (a) Hs target versus artificial; (b) residual ΔHs; (c) alpha coefficients; and
(d) gamma coefficients.

distributed nonlinearly, which makes it challenging to select a rep- Table 4. Standardized effects of alpha and gamma (factors) over Hs and Tp
resentative alpha and gamma value for the seven sea states. (responses) behavior
Sea state Experiment 1 (Point A) Experiment 2 (Point B)
DOE-ANOVA and Probability Analysis Event 1 No standardized effect. Gamma = Hs (+)
(extreme)
The results of the previous section showed that JONSWAP spectra Event 2 No standardized effect. Gamma = Hs (+)
parameters defined a scattered distribution along the time, which (extreme)
makes it difficult to select proper and specific alpha and gamma val- Event 3 Alpha = Tp (−), Gamma = Hs (+)
ues for the study area. Accordingly, we applied the DOE-ANOVA (extreme) Hs-Tp (−)
to analyze nonlinear effects and transformation of the waves, using Event 4 Alpha = Hs (+) Gamma = Hs (+), Tp (+)/Alpha
the alpha and gamma parameters generated by the GA model, and (normal) = Hs-Hs (+), Tp-Tp (−)
the Hs and Tp created by the Delft3D model. Event 5 Alpha = Hs (+) Alpha = Tp-Tp (−), Hs-Hs
Then, this research designed two experiments according to the (normal) (+)/Gamma = Tp (+), Hs (+)
seven sea states classified earlier for Point A and for Point B, re- Event 6 Alpha = Hs (+) Gamma = Hs (+)
(normal)
spectively, where Experiment 1 gathers seven datasets of Point
Event 7 Alpha = Hs (+)/Gamma Alpha = Tp-Tp (+),
A, and Experiment 2 comprises seven datasets related to the sea (normal) = Hs-Hs (−)
states defined in Table 4. Before the DOE-ANOVA analysis, we
verified that the 14 datasets satisfied the principles of normality, in- Note: Signs (+) and (−) show a positive and negative standardized effect.
dependence, and homoscedasticity. Each dataset was analyzed in- Hs-Hs and Tp-Tp represent the second-order interaction of factors.
dependently in order to prevent statistical time dependence
among the 14 datasets. an effect that could be related to the high nonlinearity of the
The identified factors in Table 4 presented p-value ≪ 0.05 of wave process; the horizontal axes of Figs. 9(b and d) indicated
significance with 95% probability confidence. As a result, the the factor levels values which were 0.39 (low) and 1 (high) for
DOE-ANOVA showed that alpha positively dominates the Hs be- Hs for Tp in both cases.
havior at Point A (Table 4), and gamma dominates the Hs behavior Phillips (1960) and Hasselmann (1962, 1963a, b) produced pi-
at Point B. During extreme events, there was no statistical evidence oneering evidence of the nonlinear interactions between waves.
that alpha or gamma control the variance of Hs and Tp at They explained that, during nonlinear interactions, wave compo-
Point A. Contrary to Point A, the gamma parameter at Point B nents were in resonance, allowing energy exchange and driving
showed a significant effect over Hs variance. Considering the the wave spectra to redistribute the energy along the frequency do-
DOE-ANOVA results in Table 4, we used the three maximum ex- main. During quadruplet wave–wave interactions, there is an en-
treme events (Fig. 5) to generate the Pareto charts and main effects ergy transfer from the peak of wave spectra to the low-frequency
plots (Figs. 8 and 9). Fig. 8 revealed that, during extreme events, region, which represents the spectra evolution. From the peak of
alpha and gamma coefficients are directly proportional to Hs and the spectra to the region of higher frequencies, energy is transferred
inversely proportional to Tp at both points (A and B). In normal and dissipated because of white-capping (Holthuijsen 2010).
conditions, alpha positively affects Hs behavior at Point A. At It is possible to observe that, during the extreme wave events for
Point B, gamma positively affected the Hs for three of four normal deep waters (Table 4, Events 1, 2 and 3, Point A), the absence of a
events; the horizontal axes of Figs. 8(b and d) showed the experi- standardized effect indicated high nonlinear wave interactions that
mental factor levels for Hs and Tp. could be generated by quadruplet wave–wave interactions. Thus,
During normal conditions (Fig. 9), alpha and gamma are di- because of the high-order interactions previously mentioned,
rectly proportional to Hs at Points A and B. However, standardized there was no statistical evidence that alpha or gamma parameters
effects for Tp revealed a higher-order effect of alpha and gamma may dominate or control the wave spectra. Event 3 for Point A
through BB interaction [Fig. 9(a)] and AB interaction [Fig. 9(c)], showed a Tp (−) and Hs-Tp (−) standardized effect for the alpha

© ASCE 04020038-7 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 (a) (b)
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.

(c) (d)

Fig. 8. DOE-ANOVA analysis: (a) pareto chart of standardized effects for alpha at Point A during State 3 (extreme); (b) main effects plot for alpha at
Point A during State 3 (extreme); (c) pareto chart of standardized effects for gamma at Point B during State 1 (extreme); and (d) main effects plot for
gamma at Point B during State 1 (extreme).

(a) (b)

(c) (d)

Fig. 9. DOE-ANOVA analysis: (a) pareto chart of standardized effects for alpha at Point A during State 6 (normal); (b) main effects plots for alpha at
Point A during State 6 (normal); (c) pareto chart of standardized effects for gamma at Point B during State 6 (normal); and (d) main effects plots for
gamma at Point B during State 6 (normal).

© ASCE 04020038-8 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
(a)
(c)
Fig. 10. Joint probability: (a) Hs-Tp at Point A; (b) Hs-Tp at Point B; (c) alpha-gamma at Point A; and (d) alpha-gamma at Point B.
parameter, which shows that alpha has a dominant inverse effect
over the wave period. Furthermore, the standardized effects suggest
that frequencies redistribute along the spectra because of spectra
evolution and white-capping dissipation.
We developed joint probabilities for wave and spectra pa-
rameters [Figs. 10(a and b)] to verify whether the water depth
could affect the Hs and Tp distribution. For the Hs-Tp joint
probability, Point A showed two cores that represent waves
with high (Hs = 1.4 m, Tp = 1.4 s) and low (Hs = 0.7 m, Tp =
4.5 s) frequencies. In addition, one of the two probability
cores seen at Point A vanished at Point B. The previous behav-
ior of joint probabilities displayed wave shoaling because of the
water-depth reduction from 102 m at Point A to 28 m at Point
B. When a wave propagates to shallow water, the frequency dis-
persion decreases and phase speed approaches the group veloc-
ity. As a result, the phase speed becomes less dependent on
frequency. Wave shoaling limits the linear wave theory applica-
tion because the phase speed and group velocity approach zero
at the waterline, resulting in a wave amplitude that tends to in-
finity (Holthuijsen 2010).
The wave shoaling effect is evident in Figs. 10(b and c) because
the probability density of alpha is higher at Point B than Point
A. Therefore, an alpha coefficient of 0.01 and a gamma coefficient
of 1 is suggested. Considering the joint probabilities in Fig. 10, it is
possible to argue that water-depth reduction generated a probability
core concentration for Hs-Tp and alpha-gamma; thus, wave shoal-
ing reduced the spectra energy spreading. The propagation of
waves straight to the coast is associated with refraction at the
study area; thus, the waves from Point A (deep-water) to Point B
(shallow water) change in direction and reduce the wave height
© ASCE
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
(b)
(d)
before the breaking zone. The control volume variation of the
water column drives waves to propagate obliquely to the coast to
retain mass conservation. This change in wave direction associates
a phase-speed increment and a wave-frequency redistribution
[Fig. 10(b)].
We selected Events 1 and 6 as extreme and normal wave events,
respectively (Fig. 11), in order to estimate the alpha and gamma
probabilities through a probability distribution. Fig. 11 shows the
behavior of spectra parameters in the probability domain where
normal distributions have symmetric and asymptotic tails, indicat-
ing a proper distribution of data according to the central limit the-
orem; to apply the ANOVA, the dataset modeled by the GA must
satisfy the assumptions of normality, homoscedasticity, and inde-
pendence (Montgomery 2017). The normal distribution is then ev-
idence that the samples generated by the GA have a short statistical
distance between mean and median for their population. The stat-
istical closeness of central values in normal distributions indicates
that the second quartile (50%) may represent the population of data.
We generated the probability distributions of the alpha and
gamma coefficients (Fig. 11) and selected the second quartile of
probability (50%) because it represents the statistical median of
the data; the second quartile implies that most of the alpha and
gamma values are close to the central value. As a result, Table 5
shows the representative values of alpha and gamma for the
study area.
The second quartiles showed that the JONSWAP spectra pa-
rameters vary during extreme and normal conditions. For Point
A (100 m)—classified as deep water according to the water-depth
wavelength ratio—the alpha parameter remains constant with re-
spect to Point B at intermediate depth (28 m). In addition, the
04020038-9 J. Waterway, Port, Coastal, Ocean Eng.
 (a) (b) (c)
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.

(d) (e) (f)

(g) (h)

Fig. 11. Probability distribution for alpha and gamma coefficients (dots) in normal and extreme wave events, modeled by GA: (a) gamma for Point A
(extreme); (b) gamma for Point B (extreme); (c) alpha for Point A (extreme); (d) alpha for Point B (extreme); (e) gamma for Point A (normal); (f)
gamma for Point B (normal); (g) alpha for Point A (normal); and (h) alpha for Point B (normal).

Table 5. Probability results for the second quartile (50%) of alpha and
gamma distributions
Alpha Gamma

Sea state Point A Point B Point A Point B

Extreme 0.02 0.02 7.00 5.48
Normal 0.01 0.01 5.00 5.45

gamma parameter for Point A was higher than gamma for

Point B. Therefore, the gamma coefficient must increase to simu-
late high waves during extreme events. We suggest identifying the
wave states to apply the probability analysis for individual data
sets. Hence, the alpha and gamma values selection will be correct.
During normal wave conditions, the alpha parameter remains
constant for Points A and B but lower in respect to the extreme
sea state; during normal conditions, the gamma coefficient of Point Fig. 12. JONSWAP spectra for Points A and B during normal and ex-
treme wave states.
B was higher in respect to Point A (Fig. 12). The alpha parameter

© ASCE 04020038-10 J. Waterway, Port, Coastal, Ocean Eng.

J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038

 reduction in normal conditions related to extreme conditions is due to
water-depth reduction (Table 5). When waves propagate to the coast,
the wave spectra concentrate the energy frequency distribution, re-
sulting in uniform waves before they arrive at the breaking zone. Fur-
thermore, the gamma coefficient increased at Point B because of the
decrease of spectra spreading, leading to an increment of the gamma
value at Point B. As a result, Table 5 shows the alpha and gamma
values in the study area, which considers the water-depth and the
wave-event transitions. To validate the selected alpha and gamma
parameters after the probability analysis, we calculated the
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
Fig. 13. Data pixel (dash-line polygon) of the WaveWatch III database
nearest to Point A (triangle symbol) and Point B (circle symbol)
(Table 5).
(a)
(b)
Fig. 14. Wave climate analysis of the data pixel of the WaveWatch III database nearby to Point A and Point B (Fig. 13): (a) POT applied for three-
hourly records of the 2001 year; (b) cumulative probability curve; and (c) extreme cumulative probability curve.
© ASCE 04020038-11
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
JONSWAP spectra [Eq. (2)] using the values for Point A during nor-
mal and extreme sea states (Table 5). Fig. 12 shows the wave spectra
and the calculated Hs [Eq. (3)] for the normal and extreme condi-
tions. Because in the study area there is not sufficient in situ informa-
tion to validate the selected alpha and gamma values shown
previously (Table 5 and Fig. 12), we had to analyze the wave data
reported in the literature and open databases.
The calculated Hs for normal and extreme sea states seen in
Fig. 12 were similar to the reported Hs values in the literature.
Devis-Morales et al. (2017) plotted the monthly maximum signif-
icant wave height values in the Caribbean Sea calculated from
January 1979 to December 2014 by ERA-Interim reanalysis.
Fig. 6 of that study showed the maximum monthly Hs values at
point A and point B were between 3.5 and 4 m. In addition,
Mesa García (2010) developed a methodology to correct modeled
wave time series for regions with scarce in situ data, utilizing da-
tabases such as WaveWatch III, TOPEX/POSEIDON, and
JASON-1. That research generated cumulative probability curves
for the Caribbean Sea at several locations near the Point A and
Point B of this study. The reported Hs values of that study indi-
cated for 50% of probability values (mean value) about 1.91 m
and values of 4.25 m related to the 99% of probability
(extreme value).
To complement the validation of the selected alpha and gamma
values (Table 5) for the study area, this study extracted three-hourly
wave reanalysis data for the 2001 year (Fig. 13). As seen in Fig. 13,
the pixel encloses to Point A and Point B; then, we performed a
wave climate analysis to the extracted data pixel to validate the cal-
culated Hs of Point A and B, shown in Fig. 12.
Fig. 14 presents the wave climate analysis to the extracted data
pixel (Fig. 13), where Fig. 14(a) depicts three-hourly Hs values of
(c)
J. Waterway, Port, Coastal, Ocean Eng.
 the 2001 year. The black horizontal line in Fig. 14(a) represents the
POT applied to the 2001 year, which denoted the maximum Hs as-
sociated with the fourth quartile (95%) of the extreme cumulative
probability curve in Fig. 14(c).
The cumulative probability curve in Fig. 14(b) showed a Hs of
2.06 m, which represents a normal sea-state condition for the 2001
year. In addition, the maximum Hs of the extreme cumulative prob-
ability curve was associated with the extreme sea-state conditions
for the location, with a value of 4.48 m. As a result, the 1D wave
spectra modeled through Eq. (2) (Fig. 12) using the specific
alpha and gamma parameters for Point A and Point B (Table 5)
was successfully validated against the extracted wave reanalysis
data (Figs. 13 and 14).
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
Conclusions
This research applied the GA model when in situ raw wave data is
not available; it analyzed the nonlinear effects of JONSWAP pa-
rameters over the Hs and Tp during water-depth and sea-state
transitions. This study recommends the following steps consider-
ing the wave data limitation: (1) hydrodynamic modeling; (2)
wave climate analysis; (3) generation of coefficients; (4)
DOE-ANOVA and probability analysis; and (5) validation. The
DOE-ANOVA analysis facilitated the identification of the alpha
and gamma effect over Hs and Tp. The gamma coefficient influ-
enced Hs almost linearly at intermediate depths (Point B) and
the alpha coefficient had a linear influence on Hs at deep water
depths (Point A). The research analyzed nonlinear wave–wave in-
teractions during extreme wave events known as quadruplets. In
addition, the water-depth reduction concentrated the spectra en-
ergy according to the joint probability of Hs-Tp and alpha-
gamma, suggesting a shoaling effect over the wave spectra. As
a result, the wave period modulated the energy distribution
along the frequency spectra domain.
The standardized effects of alpha and gamma coefficients during
extreme events had an influence directly proportional to Hs and in-
versely proportional to Tp. Furthermore, during normal sea states,
alpha positively affected Hs at Points A and B. The normal and log-
normal probability distributions for the extreme and normal waves
allowed selection of the JONSWAP spectra coefficients for deep
and intermediate water depths during extreme and normal condi-
tions. This study developed a methodology to show the validation
of the selected alpha and gamma JONSWAP spectra parameters
through the comparison of calculated Hs against other studies
and local databases, indicating that the applied methodology al-
lowed proper identification of the spectra parameters. As a result,
this research may be considered a reference to select model coeffi-
cients of parametric equations such as the JONSWAP spectra.
The heuristic methodology presented herein represents accurate
alpha and gamma parameters for JONSWAP spectra to determine
Hs and Tp for the study area. Nevertheless, the methodology should
be evaluated in other domains to improve confidence in its
applications.
Data Availability Statement
Some or all data, models, or code generated or used during the
study are available from the corresponding author by request.
The items available are: MDF and grid file of Delft3D model, Mat-
lab code for plotting Fig. 11, and sf6, sfx, and sgp files for plotting
Figs. 7 and 8 in Statgraphics software.
© ASCE 04020038-12
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
Acknowledgments
Authors thank Universidad Militar Nueva Granada, Universidad
del Norte, and Universidad de la Costa for the financial support
through the research project INV-ING-2985.
References
ASCE. 2017. Minimum design loads and associated criteria for buildings
and other structures. ASCE/SEI 7-16. Reston, VA: ASCE.
Booij, N., R. C. Ris, and L. H. Holthuijsen. 1999. “A third-generation wave
model for coastal regions: 1. Model description and validation.”
J. Geophys. Res.: Oceans 104 (C4): 7649–7666. https://doi.org/10
.1029/98JC02622.
Boukhanovsky, A. V., and C. Guedes Soares. 2009. “Modelling of multi-
peaked directional wave spectra.” Appl. Ocean Res. 31 (2): 132–141.
https://doi.org/10.1016/j.apor.2009.06.001.
Boukhanovsky, A. V., L. J. Lopatoukhin, and C. Guedes Soares. 2007.
“Spectral wave climate of the North Sea.” Appl. Ocean Res. 29 (3):
146–154. https://doi.org/10.1016/j.apor.2007.08.004.
Calini, A., and C. M. Schober. 2017. “Characterizing JONSWAP rogue
waves and their statistics via inverse spectral data.” Wave Motion 71:
5–17. https://doi.org/10.1016/j.wavemoti.2016.06.007.
Chakrabarti, S. 2005. Handbook of offshore engineering. Amsterdam,
Netherlands: Elsevier.
Cifuentes, C., and M. H. Kim. 2017. “Hydrodynamic response of a cage
system under waves and currents using a morison-force model.”
Ocean Eng. 141: 283–294. https://doi.org/10.1016/j.oceaneng.2017.06
.055.
Deltares. 2014a. Delft3D-WAVE. Simulation of short-crested waves with
SWAN—User manual. Delft, Netherlands: Deltares.
Deltares. 2014b. Delft3D-FLOW. Simulation of multi-dimensional hydro-
dynamic flows and transport phenomena, including sediments—User
manual. Delft, Netherlands: Deltares.
Derschum, C., I. Nistor, J. Stolle, and N. Goseberg. 2018. “Debris impact
under extreme hydrodynamic conditions part 1: Hydrodynamics and
impact geometry.” Coastal Eng. 141: 24–35. https://doi.org/10.1016/j
.coastaleng.2018.08.016.
Devis-Morales, A., R. A. Montoya-Sánchez, G. Bernal, and A. F. Osorio.
2017. “Assessment of extreme wind and waves in the Colombian
Caribbean Sea for offshore applications.” Appl. Ocean Res. 69: 10–
26. https://doi.org/10.1016/j.apor.2017.09.012.
Dong, G., H. Chen, and Y. Ma. 2014. “Parameterization of nonlinear
shallow water waves over sloping bottoms.” Coastal Eng. 94: 23–32.
https://doi.org/10.1016/j.coastaleng.2014.08.012.
Elhakeem, A., W. Elshorbagy, and T. Bleninger. 2015. “Long-term hydro-
dynamic modeling of the Arabian Gulf.” Mar. Pollut. Bull. 94 (1–2):
19–36. https://doi.org/10.1016/j.marpolbul.2015.03.020.
Escobar, C. A. 2011. “Relevancia de procesos costeros en la hidrodinámica
del Golfo de Urabá (Caribe colombiano).” Bull. Mar. Coastal Res.
40 (2): 327–346.
FEMA and NOAA (Federal Emergency Management Agency and National
Oceanic and Atmospheric Administration). 2012. FEMA P-646:
Guidelines for design of structures for vertical evacuation from tsuna-
mis. Redwood City, CA: Applied Technology Council.
Fragasso, J., L. Moro, L. M. Lye, and B. W. T. Quinton. 2019.
“Characterization of resilient mounts for marine diesel engines:
Prediction of static response via nonlinear analysis and response surface
methodology.” Ocean Eng. 171: 14–24. https://doi.org/10.1016/j
.oceaneng.2018.10.051.
Garcia, M., I. Ramirez, M. Verlaan, and J. Castillo. 2015. “Application of a
three-dimensional hydrodynamic model for San Quintin Bay, B.C.,
Mexico. Validation and calibration using OpenDA.” J. Comput. Appl.
Math. 273: 428–437. https://doi.org/10.1016/j.cam.2014.05.003.
Hanley, M. E., et al. 2014. “Shifting sands? Coastal protection by sand
banks, beaches and dunes.” Coastal Eng. 87: 136–146. https://doi.org
/10.1016/j.coastaleng.2013.10.020.
J. Waterway, Port, Coastal, Ocean Eng.
 Hasselmann, K. 1962. “On the non-linear energy transfer in a gravity-wave
spectrum Part 1. General theory.” J. Fluid Mech. 12 (4): 481–500.
https://doi.org/10.1017/S0022112062000373.
Hasselmann, K. 1963a. “On the non-linear energy transfer in a gravity
wave spectrum Part 2. Conservation theorems; wave-particle analogy;
irreversibility.” J. Fluid Mech. 15 (2): 273–281. https://doi.org/10
.1017/S0022112063000239.
Hasselmann, K. 1963b. “On the non-linear energy transfer in a gravity-
wave spectrum. Part 3. Evaluation of the energy flux and swell-sea in-
teraction for a Neumann spectrum.” J. Fluid Mech. 15 (3): 385–398.
https://doi.org/10.1017/S002211206300032X.
Hasselmann, K., et al. 1973. “Measurements of wind-wave growth and
swell decay during the Joint North Sea Wave Project (JONSWAP).”
Ergänzungsheft zur Deutschen Hydrographischen Zeitschrift
8 (12): 93.
Downloaded from ascelibrary.org by University of Exeter on 07/17/20. Copyright ASCE. For personal use only; all rights reserved.
Holthuijsen, L. H. 2010. Waves in oceanic and coastal waters. Cambridge,
UK: Cambridge University Press.
Ji, C., Q. Zhang, and Y. Wu. 2018. “An empirical formula for maximum
wave setup based on a coupled wave-current model.” Ocean Eng.
147: 215–226. https://doi.org/10.1016/j.oceaneng.2017.10.021.
Le Provost, C., M. L. Genco, F. Lyard, P. Vincent, and P. Canceil. 1994.
“Spectroscopy of the world ocean tides from a finite element hydrody-
namic model.” J. Geophys. Res. 99 (C12): 24777–24797. https://doi.org
/10.1029/94JC01381.
Liu, S., Y. Li, and G. Li. 2007. “Wave current forces on the pile group of
base foundation for the East Sea Bridge, China.” J. Hydrodyn. 19 (6):
661–670. https://doi.org/10.1016/S1001-6058(08)60001-3.
Locarnini, R. A., et al. 2013. World ocean atlas 2013, Volume 1:
Temperature. NOAA Atlas NESDIS 73. Silver Spring, MD: U.S.
Department of Commerce.
Lucas, C., and C. Guedes Soares. 2015. “Bivariate distributions of signifi-
cant wave height and mean wave period of combined sea states.” Ocean
Eng. 106: 341–353. https://doi.org/10.1016/j.oceaneng.2015.07.010.
Mackay, E. B. L. 2011. “Modelling and description of omnidirectional
wave spectra.” In Proc., European Wave and Tidal Energy.
Southampton, UK: University of Southampton.
Mackay, E. B. L. 2016. “A unified model for unimodal and bimodal ocean
wave spectra.” Int. J. Mar. Energy 15: 17–40. https://doi.org/10.1016/j
.ijome.2016.04.015.
McCombs, M. P., R. P. Mulligan, and L. Boegman. 2014. “Offshore wind
farm impacts on surface waves and circulation in Eastern Lake
Ontario.” Coastal Eng. 93: 32–39. https://doi.org/10.1016/j.coastaleng
.2014.08.001.
Mesa García, J. C. 2010. “Metodología para el reanálisis de series de oleaje
para el Caribe Colombiano.” M.Sc. thesis, Facultad de Minas,
Universidad Nacional de Colombia.
Montazeri, N., U. D. Nielsen, and J. Juncher Jensen. 2016. “Estimation of
wind sea and swell using shipboard measurements—A refined paramet-
ric modelling approach.” Appl. Ocean Res. 54: 73–86. https://doi.org/10
.1016/j.apor.2015.11.004.
Montgomery, D. C. 2017. Design and analysis of experiments. Hoboken,
NJ: John Wiley & Sons.
Myrhaug, D. 2018. “Some probabilistic properties of deep water wave
steepness.” Oceanologia 60 (2): 187–192. https://doi.org/10.1016/j
.oceano.2017.10.003.
NOAA (National Oceanic and Atmospheric Administration). 2016. “NCEP
North American Regional Reanalysis: NARR.” Accessed July 4, 2020.
https://www.esrl.noaa.gov/psd/data/gridded/data.narr.html.
NOAA (National Oceanic and Atmospheric Administration). 2018a.
“ETOPO1 Global Relief Model.” ETOPO1 Global Relief Model.
Accessed July 20, 2018. https://www.ngdc.noaa.gov/mgg/global/.
NOAA (National Oceanic and Atmospheric Administration). 2018b.
“NOAA WAVEWATCH III® CFSR Reanalysis Hindcasts.” NOAA
WAVEWATCH III. Accessed July 20, 2018. https://polar.ncep.noaa
.gov/waves/CFSR_hindcast.shtml.
Ochi, M. K., and E. N. Hubble. 1976. “Six-parameter wave spectra.” In
Proc., 15th Int. Conf. on Coastal Engineering, 301–328. Reston, VA:
ASCE.
© ASCE 04020038-13
J. Waterway, Port, Coastal, Ocean Eng., 2020, 146(6): 04020038
Ortega, S., A. F. Osorio, P. Agudelo-Restrepo, and J. I. Velez. 2011.
“Methodology for estimating wave power potential in places with
scarce instrumentation in the Caribbean Sea.” In OCEANS 2011
IEEE, 1–5. Santander, Spain: IEEE.
Pascoal, R., L. P. Perera, and C. Guedes Soares. 2017. “Estimation of direc-
tional sea spectra from ship motions in sea trials.” Ocean Eng. 132:
126–137. https://doi.org/10.1016/j.oceaneng.2017.01.020.
Phillips, O. M. 1960. “On the dynamics of unsteady gravity waves of finite
amplitude Part 1. The elementary interactions.” J. Fluid Mech. 9 (2):
193–217. https://doi.org/10.1017/S0022112060001043.
Power, H. E., B. Gharabaghi, H. Bonakdari, B. Robertson, A. L. Atkinson,
and T. E. Baldock. 2019. “Prediction of wave runup on beaches using
gene-expression programming and empirical relationships.” Coastal
Eng. 144: 47–61. https://doi.org/10.1016/j.coastaleng.2018.10.006.
Restrepo, J. C., K. Schrottke, C. Traini, J. C. Ortíz, A. Orejarena, L. Otero,
A. Higgins, and L. Marriaga. 2016. “Sediment transport and geomor-
phological change in a high-discharge tropical delta (Magdalena
River, Colombia): Insights from a period of intense change and
human intervention (1990–2010).” J. Coastal Res. 32 (3): 575–589.
https://doi.org/10.2112/JCOASTRES-D-14-00263.1.
Rueda Bayona, J. G. 2015. “Caracterización hidromecánica de plataformas
marinas en aguas intermedias sometidas a cargas de oleaje y corriente
mediante modelación numérica.” Master thesis, Facultad de Minas,
Universidad Nacional de Colombia.
Rueda Bayona, J. G. 2017. “Identificación de la influencia de las varia-
ciones convectivas en la generación de cargas transitorias y su efecto
hidromecánico en las estructuras Offshore.” Ph.D. thesis, Civil and
Environmental Engineering Dept., Universidad del Norte.
Rueda-Bayona, J. G., A. Guzmán, and R. Silva. 2020. “Genetic algorithms
to determine JONSWAP spectra parameters.” Ocean Dyn. 70: 561–571.
https://doi.org/10.1007/s10236-019-01341-8.
Rueda-Bayona, J. G., A. F. Osorio-Arias, A. Guzmán, and G.
Rivillas-Ospina. 2019. “Alternative method to determine extreme hy-
drodynamic forces with data limitations for offshore engineering.”
J. Waterw. Port Coastal Ocean Eng. 145 (2): 05018010. https://doi
.org/10.1061/(ASCE)WW.1943-5460.0000499.
Sakhare, S., and M. C. Deo. 2009. “Derivation of wave spectrum using data
driven methods.” Mar. Struct. 22: 594–609. https://doi.org/10.1016/j
.marstruc.2008.12.004.
Sanil Kumar, V., and K. Ashok Kumar. 2008. “Spectral characteristics of
high shallow water waves.” Ocean Eng. 35 (8): 900–911. https://doi
.org/10.1016/j.oceaneng.2008.01.016.
Sun, Y., and X. Zhang. 2017. “A second order analytical solution of fo-
cused wave group interacting with a vertical wall.” Int. J. Nav.
Archit. Ocean Eng. 9 (2): 160–176. https://doi.org/10.1016/j.ijnaoe
.2016.09.002.
Uittenbogaard, R. E., J. A. T. M. van Kester, and G. S. Stelling. 1992.
Implementation of three turbulence models in 3D-TRISULA for rectan-
gular grids. Delft, Netherlands: Delft Hydraulics.
Wang, Y. 2014. “Calculating crest statistics of shallow water nonlinear
waves based on standard spectra and measured data at the Poseidon
platform.” Ocean Eng. 87: 16–24. https://doi.org/10.1016/j.oceaneng
.2014.05.012.
Wijaya, A. P., and E. Van Groesen. 2016. “Determination of the significant
wave height from shadowing in synthetic radar images.” Ocean Eng.
114: 204–2015. https://doi.org/10.1016/j.oceaneng.2016.01.011.
Young, D. L., and B. M. Scully. 2018. “Assessing structure sheltering via
statistical analysis of AIS data.” J. Waterw. Port Coastal Ocean Eng.
144 (3): 04018002. https://doi.org/10.1061/(ASCE)WW.1943-5460
.0000445.
Zanaganeh, M., S. J. Mousavi, and A. F. Etemad Shahidi. 2009. “A hybrid
genetic algorithm–adaptive network-based fuzzy inference system in
prediction of wave parameters.” Eng. Appl. Artif. Intell. 22 (8): 1194–
1202. https://doi.org/10.1016/j.engappai.2009.04.009.
Zweng, M. M., et al. 2013. World ocean atlas 2013, Volume 2: Salinity.
NOAA Atlas NESDIS 74. Silver Spring, MD: U.S. Department of
Commerce.
J. Waterway, Port, Coastal, Ocean Eng.