Hindcast of the wave conditions along the west Iberian coast [600250]

Hindcast of the wave conditions along the west Iberian coast
L. Rusu, P. Pilar, C. Guedes Soares ⁎
Centre for Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Técnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal
article info abstract
Available online 15 April 2008 This paper describes the development of a wave prediction system for the west Iberian coast. The
implemented wave prediction system is based on two state-of-the-art spectral wave models, WAM for the
ocean area and SWAN for the nearshore. However, because of its extended geographical space the SWAN
model will include some generation effects in the coarse SWAN simulations, complemented by wavetransformation effects near the coast. The system was validated by means of extended hindcast runs in various
regions belonging to the continental Portuguese coastal environment, which were compared with buoy data,
focusing on the extreme energetic events and both direct comparisons and statistical results are presented.
© 2008 Elsevier B.V. All rights reserved.Keywords:
Wave modelsHindcastWAMSWANHIPOCAS
1. Introduction
Assessment and prediction of the wave conditions with numerical
models are of essential importance in the coastal areas. The coastal
regions are usually characterized by high traf fic of ships and therefore a
better knowledge concerning of wave conditions might preventaccidents or even ecological disasters. Furthermore, while wave data
has been collected by waverider buoys at various coastal locations there
are always appearing the need of knowledge of the wave conditions in
new coastal locations where development may be planned. In these
cases the use of numerical models to hindcast the conditions in those
locations is essential.
This has been the background of the HIPOCAS project “Hindcast of
Dynamic Processes of the Ocean and Coastal Areas of Europe ”(Guedes
Soares et al., 2002 ), which has produced a database of 44 years of
wind, sea-level and wave data along the European waters. The wave
hindcast was produced by the WAM model in its nested form (Gómez
and Carretero, 1997 ) and was run in the deep water mode ( Pilar et al.,
2008-this issue ). Therefore the results are appropriate for the
locations around the coasts where the bottom effects do not yet
start to modify the wave conditions, which depending on the local
bathymetry can be closer or further away from coast.
This work is a follow-up of that project and is also a demonstration
of its usefulness in one particular application. This work intends to
extend the wave simulations to shallow water where the bottom
effects start being important for the wave predictions. This is the case
in the entrance of many ports or in areas where coastal developments
are required. The strategy in the present work is to use the results of
the WAM runs performed in HIPOCAS to be used as boundaryconditions for a shallow water model that makes predictions closer to
coast and up to shallow water conditions.
The state of the art in medium and large scale wave modeling
today is the third generation wave models, which solve the spectral
action balance equation without prior assumption of the wave spectral
shape. The WAM model ( WAMDI Group, 1988 ) for generation area and
the SWAN model (acronym from Simulating Waves Nearshore), (Booijet al., 1999 ) for transformation area are two of them, probably the
most generalized and tested wave prediction models, which can beused both for hindcasting and forecasting purpose. This work uses
both models in nested mode in the predictions.
To perform wave hindcasting can also be one way to validate an
operational system. In this perspective, while the goal of the present work is
to perform reanalysis studies based o n numerical models for assessing the
nearshore conditions of continental Portugal, it is also one way of validatingan operational system based on those models. The analyses were focused
on the highest energy conditions as these are the extreme conditions that
are often critical for design or operation of ship or coastal structures and
they are the ones that are sometimes more dif ficult to model accurately.
The quality of numerical wave and sea-level hindcasts for offshore
and coastal areas depends on the quality and the accuracy of the
driving wind fields. Various studies ( Teixeira et al., 1995; Holthuijsen
et al., 1996 ) have shown that accuracy of the wind field has a large
impact on the predicted wave field. For the coastal simulations the
good quality of the boundary conditions (wave spectra) provided byCoastal Engineering 55 (2008) 906 –919
⁎Corresponding author.
E-mail address: c.guedes.soares@ist.utl.pt (C. Guedes Soares).
Table 1
Two-way WAM nested application for the North Atlantic Ocean
Grids North South East West Δx×Δy(°)
Coarse 1 70° 14° 20° −64° 2°×2°
Coarse 2 68° 20° 4° −50° 1°×1°
Medium 60° 24° 33° −0° 0.5×0.5°
0378-3839/$ –see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.coastaleng.2008.02.029
Contents lists available at ScienceDirect
Coastal Engineering
journal homepage: www.elsevier.com/locate/coastaleng

generation models, the bathymetry and the physics adopted are also
very important.
The reanalysis wind field for North Atlantic basin and eastern
European coasts, determined hourly with a resolution of 0.5° in
HIPOCAS ( Weisse and Feser, 2003 ) are used in this work. These wind
fields were produced by a regional model that was forced by the NCEP
reanalysis winds ( Tolman, 1998 ).
Next section will describe the implementation of the wave pre-
diction system, both in the ocean and the nearshore scales. Section 3
presents the results of the system validations in various locations of
the Portuguese Coast and Section 4 presents results of high resolution
applications of the system.2. Implementation of the wave prediction system
The general idea for following the wave generation and propaga-
tion from deep ocean to coastal environment is to nest into the ocean
Fig. 1. The geographical spaces of the WAM simulations.
Fig. 2. a) The geographical spaces of the SWAN simulations. b) The first target area, the locations of the buoys.
Table 2
Computational grids for the SWAN simulations
Grids North South East West Δx×Δy(°) ng x×ngy=n p
Coarse 45° 35° −6° −11° 0.05°×0.1° 101×101=10201
Medium1 41.8° 39.8° −8.6° −9.6° 0.01°×0.02° 101×101=10201
Medium2 37.8° 36.3° −7.6° −9.8° 0.02°×0.02° 111×76=8436907 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

scale model higher resolution models that are able to better account
for the physical processes in the nearshore. For assessing the wave
conditions close to the Portuguese coasts two state-of-the-art spectral
wave models were nested. These are the WAM Cycle 4 wave genera-
tion model ( Günther et al., 1992 ) and SWAN version 40.41 ( Booij et al.,2004 ) for the coastal environment, which is used both for generation
and for wave transformation in the nearshore.
The basic scienti fic principles of these two models are similar in
the sense that they are third generation wave models based on in-tegrating the spectral action balance equation in all five dimensions
Fig. 4. First target area –energetic peak 1994/01/06hh18. a) signi ficant wave height fields and vectors; b) signi ficant swell height fields and vectors.
Fig. 3. a) Wind fields and vectors, 1994/01/06 h18. b) Signi ficant wave height fields (Janssen formulation results) —energetic peak 1994/01/06 h18.908 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

(time, geographic space and spectral space). In this equation, the
effects of wave generation, dissipation and nonlinear wave-wave in-
teractions are introduced by the source terms. The SWAN model con-
tains some additional formulations speci fic for shallow water.
However, the numerical techniques implemented are different in
the two models. SWAN (unlike WAM) uses an implicit upwind scheme
to propagate the wave action density, which has the great advantage
that the propagation time step is not limited by any numerical con-
dition since the scheme is unconditionally stable in the geographic
and spectral spaces.
In order to obtain the necessary high resolution of boundary
condition needed for the coastal model, two-way nesting techniques
(Gómez and Carretero, 1997 ) are employed in the WAM model. The
Table 1 describes the main features of the grid sets used for this
application; the grid spacing was noted with Δx×Δy(°). The three
domains of WAM simulations are illustrated in Fig. 1 and they are the
same as used in Pilar et al. (2008-this issue) .
The implementation of the WAM model was made for 24 direc-
tions and 25 frequencies bands and the energy balance equation wasintegrated with a time step of 300 s. The lowest resolved frequency is
0.0418 Hz.
The coupling between the generation and the transformation
models was made by nesting a large SWAN area into the WAM model,
as presented in Fig. 1 . This area covers the entire west Iberian coastand it is used as a general driver for the coastal simulations (see
Fig. 2 a). Spectra with 0.5° spatial resolution and 1 h time resolutions
provided by the WAM model are the boundary conditions of theSWAN model.
The SWAN model calibration was made on two areas: the first
coastal area covers the North of Portugal continental nearshore and
the second the South, as shown in Fig. 2 a. Because of the geographical
position of these areas, the wave climates are different. In the Northarea the predominant high wave conditions are from NW and W
directions while the South wave climate is in fluenced also from the W
and SW storms and the wave climate are characterized by lowervalues of the signi ficant wave height fields.
For the SWAN model calibration in the North area, data from four
buoys located in the computational domain were simultaneously
available. This permitted a cross validation not only in the two sides of
the target area but also both in deep and shallow water. The geo-
graphical locations of the four buoys used for calibration of the model
results are indicated in Fig. 2 b. The three buoys from de WAVEMOD
project ( Guedes Soares, 2000 ) are denoted by: B1 (8.92° W, 40.22° N),
B2 (9.07° W, 40.26° N) and B3 (9.24° W, 40.26° N) while the other one,B4 (9.088° W, 41.203° N), is maintained by Instituto Hidrográ fico. The
first three buoys were all located in the area of Figueira da Foz from
shallow to deep water (at 19.6 m, 72 m and 93 m), while the fourth
was positioned close to Leixões at 110 m.
Fig. 6. In situ comparison Buoy3 —SWAN, signi ficant wave height (between 03.12.1993– 28.02.1994).
Fig. 5. In situ comparison Buoy2 —SWAN, signi ficant wave height (between 03.12.1993– 28.02.1994).909 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

Fig. 9. In situ comparison Buoy3 —SWAN, mean direction (between 03.12.1993– 28.02.1994).
Fig. 8. In situ comparison Buoy3 —SWAN, mean period (between 03.12.1993 –28.02.1994).
Fig. 7. In situ comparison Buoy4 —SWAN, signi ficant wave height (between 01.12.1993– 28.02.1994).910 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

The characteristics for the corresponding computational grids
used in the SWAN simulations are given in Table 2 , where the grid
spacing was noted with Δx×Δy(°) and the number of the grid points
with np.
The SWAN simulations were performed in the non stationary
mode (either as pure non stationary or sequences of stationary simu-
lations) and all the three available numerical schemes were used.
Thus, for the general driver use was made of the second-order upwind
scheme with third-order diffusion, so-called S&L and SORDUP which
is a second-order upwind scheme with second-order diffusion. The
first one was for the pure non stationary simulations and the second
for the sequences of stationary simulations. In the target area the first-
order scheme backward space, backward time (BSBT) was used forboth cases. This is imposed into the model by the CFL (Courant –
Friedrichs –Levy) condition ( Δt≤Δx/c).
As regards the physical processes, a compromise was made be-
tween the default parameterizations and the most appropriate ones.For each phenomenon or process there are usually in SWAN several
alternatives given by different formulations. The default parameter-
ization is usually the most common one and also is certainly the
simplest way when working with the wave model. However, the
possibilities opened by the SWAN model of using different formula-
tions for physical processes, as well as a wide range of values for some
coefficients the models, can lead to considerable improvements in
the results despite the fact that they require additional work forcalibrating.
Table 3
Statistical results for the entire wave field (first area)
Bmed Smed Bias RMSE SI r Buoy
JANSSEN Hs (m) 2.571 2.568 0.002 0.474 0.184 0.761 B1 (n =194)
Hs (m) 3.322 3.430 −0.108 0.561 0.169 0.901 B2 (n =67 6)
Hs (m) 3.312 3.659 −0.341 0.667 0.201 0.898
Tm (s) 8.347 10.17 −1.823 2.283 0.273 0.763 B3 (n =678)
Tp (s) 13.386 14.662 −1.276 2.551 0.191 0.532
Dir (°) 302.434 300.842 1.592 8.191 0.027 0.900Hs (m) 3.497 3.650 −0.153 0.659 0.188 0.873
Tm (s) 8.563 10.097 −1.534 2.091 0.244 0.713 B4 (n =459)
Tp (s) 13.225 14.542 −1.317 2.471 0.187 0.541
Dir (°) 305.725 295.672 10.053 14.965 0.049 0.819
KOMEN Hs (m) 2.571 2.785 −0.215 0.514 0.2 0.782 B1 (n =194)
Hs (m) 3.322 3.838 −0.516 0.749 0.225 0.914 B2 (n =67 6)
Hs (m) 3.312 4.090 −0.772 0.97 0.292 0.911
Tm (s) 8.347 6.540 1.807 2.206 0.264 0.648 B3 (n =678)
Tp (s) 13.386 14.617 −1.231 2.562 0.191 0.535
Dir (°) 302.434 299.482 2.953 9.206 0.03 0.898Hs (m) 3.497 4.155 −0.658 0.946 0.27 0.876
Tm (s) 8.563 6.698 1.865 2.36 0.276 0.505 B4 (n =459)
Tp (s) 13.225 14.510 −1.286 2.473 0.187 0.532
Dir (°) 305.725 292.626 13.099 18.792 0.061 0.779Table 4Statistical results for normal energetic conditions (H
Sb4.5 m)
Bmed Smed Bias RMSE SI r Buoy
JANSSEN Hs (m) 2.526 2.523 0.003 0.479 0.19 0.699 B1 (n =190)
Hs (m) 2.907 3.086 −0.179 0.541 0.186 0.804 B2 (n =570)
Hs (m) 2.926 3.306 −0.380 0.667 0.228 0.802
Tm (s) 8.113 9.892 −1.779 2.235 0.275 0.747 B3 (n =580)
Tp (s) 13.03 14.356 −1.326 2.638 0.202 0.470
Dir (°) 303.18 301.446 1.735 8.463 0.028 0.902Hs (m) 3.035 3.263 −0.227 0.632 0.208 0.754
Tm (s) 8.329 9.847 −1.518 2.047 0.246 0.715 B4 (n =379)
Tp (s) 12.847 14.21 −1.363 2.531 0.197 0.481
Dir (°) 306.121 295.952 10.169 15.318 0.05 0.816
KOMEN Hs (m) 2.526 2.735 −0.209 0.514 0.203 0.724 B1 (n =190)
Hs (m) 2.907 3.445 −0.538 0.757 0.260 0.826 B2 (n =570)
Hs (m) 2.926 3.690 −0.764 0.955 0.326 0.827
Tm (s) 8.113 6.374 1.739 2.162 0.266 0.625 B3 (n =580)
Tp (s) 13.03 14.305 −1.274 2.648 0.203 0.475
Dir (°) 303.18 300.159 3.021 9.291 0.031 0.900Hs (m) 3.035 3.707 −0.672 0.942 0.31 0.751
Tm (s) 8.329 6.612 1.717 2.263 0.272 0.502 B4 (n =379)
Tp (s) 12.847 14.179 −1.331 2.532 0.197 0.470
Dir (°) 306.121 292.943 13.178 19.074 0.062 0.775
Fig. 10. a) Wind fields and wind vectors, 2000/12/22 h00. b) Signi ficant wave height fields and wave vectors (Janssen formulation results) —energetic peak 2000/12/22 h00.911 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

For the wind growth parameterisation there are two formulations:
Komen et al. (1984) , which is the default used in the WAM Cycle 3
model and the Janssen (1991) used in the WAM Cycle 4 model. Two
different simulations were carried out with the linear growth term
activated. The bottom friction used the JONSWAP formulation, while
for the quadruplet wave-wave interactions the fully explicit com-
putations of the nonlinear transfer with DIA (Discrete Interaction
Approximation) per sweep were used.
Different computational time steps in the non stationary mode
were tested to find the better computational ef ficiency and also a good
numerical accuracy. In the non stationary mode a 20 min time stepwas chosen (for the case of the pure non stationary simulations).
When both the boundary conditions and the wind were quasi
stationary for periods longer than 6 h, sequences of stationary simu-lations were performed with a 3 h time step. The number of itera-
tions was set from 1 (which is the default value), to 4. In this way the
numerical accuracy was increased until the model passes to another
time step. Every time when passing from one set of simulations to
another one the hot files were used.
The implementation of the SWAN model was made for 36 direc-
tions and 30 frequencies logarithmically spaced from 0.05 Hz to 0.6 Hzat intervals of Δf/f=0.1. It is assumed that there are no currents, which
implies that refraction is due only to spatial variations of water depths.
3. Validation of the wave prediction system3.1. Validation in the North of Portugal
The analysis on the Northern part of Portugal was focused on a
three-month period, between 1 December 1993 and 28 February 1994
which was one of the most energetic periods encountered in the west
Portuguese nearshore in the last decades ( Rusu et al., 2004 ). According
toPaillard et al. (2000) , the most remarkable feature of the wave
climate during this period was high level of long periodic swell.
Three major storms, on 6th and 10th January and 4th February,
produced signi ficant wave heights around 8 m (registered at the buoys
B2 and B3, Figueira da Foz area). In all three cases the high waves were aresult of a combination of high swell and wind seas. In the following
figures, the wind conditions and wave field simulations for the energetic
peak registered on 6th January 1994 are presented. Fig. 3 a presents wind
fields and vectors in the coarse area and can be observed the low wind
velocity near the Portugal coast. The signi ficant wave height fields for
the same area and time are illustrated in Fig. 3 b.
Fig. 4 a and b show the signi ficant wave and swell height fields and
vectors simulated in the first target area. The dark circles mark the
location where the largest signi ficant wave and swell heights were
identi fied. In case of the signi ficant wave height fields, the maximum
simulated value was 9.18 m, very close to the maximum value of thesignificant swell height, 8.8 m. It is clear that the two fields are very
similar and the predominant component of the high waves is swell,which conforms to the observation made by Paillard et al. (2000) .
The comparison with the buoy measurements was made for the
following periods in which data was available: Buoy 1 (01.12.1993 –
26.
12.1993); Buoy 2 (03.12.1993 –28.02.1994); Buoy 3 (03.12.1993 –
28.02.1994) and Buoy 4 (01.12.1993 –28.02.1994). For the first two
buoys, only the signi ficant wave heights were available while for the
other two directional data were available. Some direct comparisons
between SWAN simulations (for the two formulations) and the buoys
are given in Figs. 5 –8for the signi ficant wave heights, Figs. 9 and 10 for
the peak periods and the mean directions. The discontinuities in thecurves describing the buoy data refl ect some gaps that were en-
countered in the measured data field.
Statistics were computed for signi ficant wave height, mean wave
period, peak wave period and mean wave direction. Those statistics
are the average values for measurements ( B
med) and simulations
(Smed), the bias, root mean square error (RMSE), scatter index (SI) and
Fig. 11. Second target area —energetic peak 2000/12/22 h06. a) signi ficant wave height fields and wave vectors, b) signi ficant swell height fields and swell vectors.Table 5
Statistical results for high energetic conditions (H S≥4.5 m)
Bmed Smed Bias RMSE SI r Buoy
JANSSEN Hs (m) 4.7 4.713 0.013 0.091 0.019 0.917 B1 (n =4)
Hs (m) 5.555 5.279 0.277 0.658 0.118 0.832 B2 (n =1 06)
Hs (m) 5.639 5.746 −0.107 0.67 0.119 0.812
Tm (s) 9.729 11.812 −2.083 2.547 0.262 0.601 B3 (n =98)
Tp (s) 15.489 16.474 −0.985 1.955 0.126 0.444
Dir (°) 298.018 297.272 0.746 6.353 0.021 0.922Hs (m) 5.684 5.485 0.199 0.772 0.136 0.725Tm (s) 9.673 11.278 −1.606 2.285 0.236 0.522 B4 (n =80)
Tp (s) 15.014 16.113 −1.1 2.167 0.144 0.491
Dir (°) 303.85 294.349 9.501 13.166 0.043 0.849
KOMEN Hs (m) 4.7 5.185 −0.486 0.493 0.105 0.914 B1 (n =4)
Hs (m) 5.555 5.954 −0.399 0.706 0.127 0.854 B2 (n =1 06)
Hs (m) 5.639 6.456 −0.817 1.053 0.187 0.833
Tm (s) 9.729 7.523 2.206 2.448 0.252 0.519 B3 (n =98)
Tp (s) 15.489 16.464 −0.976 1.978 0.128 0.436
Dir (°) 298.018 295.47 2.548 8.681 0.029 0.912Hs (m) 5.684 6.277 −0.593 0.962 0.169 0.729
Tm (s) 9.673 7.109 2.563 2.773 0.287 0.484 B4 (n =80)
Tp (s) 15.014 16.083 −1.069 2.172 0.145 0.485
Dir (°) 303.85 291.123 12.727 17.392 0.057 0.817912 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

Fig. 14. In situ comparison Faro buoy —SWAN, mean wave direction December 2000.
Fig. 13. In situ comparison Faro buoy —SWAN, mean wave period December 2000.
Fig. 12. In situ comparison Faro buoy —SWAN, signi ficant wave height December 2000.913 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

Pearson's Correlation Coef ficient ( r). If Xirepresent the measured
values, Yithe simulated values and nthe number of observations, the
mentioned statistics can be de fined with the relationships:
Bmed¼X˜¼Pn
i¼1Xi
nSmed¼Y˜¼Pn
i¼1Yi
nBias¼Pn
i¼1Xi/C0Yi ðȚ
nð1Ț
RMSE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pn
i¼1Xi/C0Yi ðȚ2
nvuuut
SI¼RMSE
X˜r¼Pn
i¼1Xi/C0X˜/C16/C17
Yi/C0Y˜/C16/C17
Pn
i¼1Xi/C0X˜/C16/C172Pn
i¼1Yi/C0Y˜/C16/C172 ! 1
2
ð2Ț
The values of these statistical parameters are given in Table 3 .
The Janssen formulation provides better results for the indicators
of Bias, RMSE and SI of signi ficant wave height at all four buoys (see
Table 3 ). Also it is clear that both formulations overestimated the
significant wave height (exception for B1, where the Bias was positive
for the Janssen formulation, but nearly zero). In general, wave peak
period and mean wave direction are represented with similar accuracy
for both formulations. The results of wave mean periods are very
different; systematically the Janssen formulation overestimated this
parameter and the Komen formulation underestimated it. In term ofstatistical results (RMSE and r), some improvement is shown when
using the Janssen formulation.
The response of the system based on numerical models might be
different when passing from normal wave conditions to highly ener-getic conditions. It is actually well known the tendency of most
generation wave models of under evaluating the peak of the highest
storms. For this reason, the same average values, bias, RMSE, scatterindex and correlation coef ficient were computed separately for nor-
mal wave conditions and storm conditions. The limit between thecommon wave climate and the highly energetic conditions was con-
sidered to be the signi ficant wave height of 4.5 m. Thus, Table 4
presents the results of the statistical parameters computed for allcases in which the signi ficant wave height is lower than 4.5 m, while
Table 5 presents the results of the same statistical parameters for the
cases when the signi ficant wave height is greater or equal to this limit.
The results presented in the Tables 4 and 5 show that for normal
energetic conditions the statistical results are more or less equal to the
statistical results for the entire wave field. The case of highly energetic
conditions brings some modi fications relative to the values of bias: for
the Janssen parameterization, the model underestimated the values of
the signi ficant wave height for de buoys B1, B2 and B4 (the value of
bias is positive) and for the buoy B4 the model overestimated thevalues of the signi ficant wave height (the value of bias is negative) but
the absolute value of bias is small (0.107); for the Komen parameter-ization the model systematically overestimated the signi ficant wave
height (the values of bias are negative).
The comparison of the RMSE values calculated for highly energetic
conditions and the entire wave field using the Komen parameteriza-
tion shows that these conditions produced an increase of the RMSE
value in deep water (for de buoys B3 and B4) and a decrease in shallow
water (for buoy B1 and B2) for the signi ficant wave height. For the
Janssen parameterization the tendency of increasing the RMSE valuewas maintained except for the case of buoy B1. In regard to mean wave
period, peak wave period and mean wave direction (there are avail-
able data only at the buoys B3 and B4) the RMSE values calculated for
Fig. 15. a) The geographical spaces of the medium and high resolution areas. b) Signi ficant wave height fields, wave (black arrows) and wind (white arrows), vectors, energetic peak
2001/01/02 h00.Table 6
Statistical results for the entire wave field (second area)
n=494 Bmed Smed Bias RMSE SI r
JAN Hs (m) 1.695 1.723 −0.029 0.393 0.232 0.931
Tm (s) 5.731 5.613 0.118 1.274 0.222 0.723Dir (°) 231.00 239.55 −8.546 27.219 0.118 0.744
KOM Hs (m) 1.695 2.147 −0.452 0.673 0.397 0.921
Tm (s) 5.731 3.933 1.798 2.037 0.356 0.675
Dir (°) 231.00 237.923 −6.915 30.300 0.131 0.697914 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

highly energetic conditions increase for the mean period and decrease
for peak period and mean direction compared with the results ob-
tained for the entire wave field, for both formulations.
The SI parameter calculated for high energetic conditions de-
creases and for normal energetic conditions it increases, compared
with the SI values for the entire wave field. This pattern is maintained
for all wave parameters and for both formulations of the SWAN model.The behavior of the correlation coef ficient in various conditions is not
so clear, mainly for high energetic conditions. These cases were alsoconsidered in Rusu et al. (2005a) .
3.2. Validation in the South of Portugal
The simulations on the coastal area of the South of Portugal were
made in the month December 2000 and in the period 1st February2001 until 10th March 2001. These were compared with the available
data from a buoy maintained by Instituto Hidrogra fico, located at 93 m
depth and geographical position: 36.905°N and 7.898°W (near toFaro). Because the high wind velocity (around 23 m/s) and the low
atmospheric pressure present in the SE of Portugal on the 22nd
December at 00 h (see Fig. 10 a), the high signi ficant wave height wave
fields were simulated near to target area (see Fig. 10 b).
Six hours later, the signi ficant wave height fields ( Fig. 11 a) re-
gistered maximum values near to the south coast. These high waveheight fields are the result of a combination of local wind seas and
swell generated by a distant storm. Fig. 11 b presents the signi ficant
swell height fields and swell vectors for the same area at same timethan in Fig. 11 a. The difference existing between the two fields and
between the wave and swell vectors can clearly be seen. Direct com-
parisons between SWAN simulations (for the two formulations) and
Faro buoy are given in Figs. 12 –14.
More information about these simulations can be encountered in
Rusu et al. (2005b) . The results of the statistical errors for the entire
wave field are given in Table 6 .
The statistical errors calculated for Hs maintain the trend found in
the statistical results listed in Tables 3 and 4 for the buoys located in
deep water. Some differences appear for the mean wave period andmean wave direction parameters. Both model formulations under-
estimated the mean wave period, but for Janssen formulation a clear
improvement of the statistical parameters (Bias, RMSE and SI) is noted.
For mean wave direction the result of simulations are similar for both
model formulations and the values of RMSE, SI and r are worse than the
previous values listed in the Table 3 .
In general, it can be seen that the simulations made with both
formulations are in good agreement with measurements. Nevertheless,
the statistical analyses showed that the Janssen formulation provided
better results and in consequence this formulation was chosen for the
future simulations.
4. Medium and high resolution simulations
A further development of the reanalysis studies is to develop and
calibrate in the Portuguese nearshore an effective wave prediction
system based on numerical models. This system should cover any area
and would be focussed especially in approaches to the important ports
of Portugal. For this purpose various medium and high resolution
simulations were made along of the Portuguese coast. In Fig. 15 are
pres
ented the areas where simulations have been made. For all these
simulations, the Janssen formulation was used with the previous
physical and computational parameterizations of SWAN model.
In the present application, the new areas are centred on the central
part of continental Portugal, on the entrance to the ports of Sines,
Lisbon and Peniche as illustrated in Fig. 15 a. Both medium areas ( first
with yellow and second with red lines) cover the port of Lisbon, and in
Fig. 16. The WAM model simulations, signi ficant wave height fields and wave vectors —energetic peak 2001/01/02 h00.Table 7
Computational grids for the SWAN simulations
Grids Xaxis (°) Yaxis (°) Δx×Δy(°) ng x×ngy=np
Medium1 1.24° 1.5° 0.02°×0.02° 63×76=4788
Medium2 1.5° 1.8° 0.02°×0.02° 76×91=6916Small1 0.5° 0.5° 0.005°×0.005° 101×101=10201Small2 1° 1° 0.01°×0.01° 101×101=10201915 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

every medium area is nested high resolution area (yellow high reso-
lution area is centred on the entrance to the port of Sines and other to
the port of Peniche)(see Table 7 ).
Thefirst simulations were performed on the areas marked with
dashed yellow lines, in the period from September 2000 until mid
March 2001. This time interval is characterized by two periods: the
first three months were normal from the energetic point of view,
while the last period was highly energetic. In the classi fication of the
first twenty five highest wave events registered at the Sines buoy
(located in the area of interest) in the last 25 years, presented in Rusuet al. (2004) , three of these storms took place in the three-month
period between December 2000 until the end of February 2001.
Fig. 15 b presents the results of simulations of an important storm
registered in this area in the beginning of January 2001. The simulatedmaximum signi ficant wave heights approach the value of 9 m while
the swell component is about 8.5 m. The wind pattern is from NorthWest (which is the most common in the Iberian nearshore) with
maximum wind velocity fields about the 16 m/s. The Fig. 16 shows the
significant wave height fields simulated with the WAM model in the
North Atlantic Ocean for the energetic peak registered on 2nd January
Fig. 18. a) Wind fields and wind vectors, 1981/12/29 h00, b) Signi ficant wave height fields and wave vectors —energetic peak 1981/12/29 h03.
Fig. 17. Medium and high resolution area, energetic peak 2001/01/02 h00. a) Medium area for Lisbon –Sines nearshore, bathymetric map and wave vectors, b) High resolution area,
significant wave height fields and wave vectors.916 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

2001. It can be seen clearly that the entire west Iberian coast is affected
by a big storm with high signi ficant wave height.
Fig. 17 a presents the bathymetric map of medium resolution that
covers the areas of Lisbon and Sines nearshore and includes the wave
vectors for the energetic peak registered on 2nd January 2001. Also in
thisfigure the location of the Sines buoy ( −8.9289 W, 37.9211 N) is
marked with a yellow triangle, corresponding at 97 m average depth.More information about these simulations can be found in Rusu et al.
(2005c) .
The second simulations were performed in the areas marked with
dashed red lines, in the period from 1st December 1981 until 28thFebruary 1982. On 29 December 1981, at 00UTC, the maximum wind
velocity was 22.1 m/s and the low pressure center is located in front of
Lisbon area ( Fig. 18 a). In these conditions high signi ficant wave height
fields were generated at South of Lisbon ( Fig. 18 b) and the Cascais
buoy registered signi ficant wave height values above 8 m on 29
December 1981. The buoy position marked with yellow triangle inFig. 19 ai s( −9.31° W, 38.59° N), with corresponding 23 m average
depth. The entire central part of Portugal coast was affected by thisstorm; also the Sines buoy registered signi ficant wave heights of more
than 8 m.The Fig. 19 a, b show the medium and high resolution areas. The
refraction effect on the wave, near to the coast, can be very well
observed in these pictures. More information about these simulations
can be found in Guedes Soares et al. (2004) .
Direct comparisons of the signi ficant wave height between the
Sines buoy and the results of the SWAN simulations for the entire
analysed period (1st of September 2000 and 10th of March 2001) are
shown in Fig. 20 .Table 8 summarises, in terms of statistics, the quality
of SWAN model simulations on the Sines area.
A comparison of the signi ficant wave heights registered by the
Cascais buoy and the same wave parameter simulated by the SWAN
Table 8
Statistical results for the entire wave field in the period 1st of September 2000 –10th of
March 2001
n=1 5 1 5 Bmed Smed Bias RMSE SI r
Hs (m) 2.397 2.591 −0.194 0.502 0.209 0.934
Tm (s) 7.179 8.00 −0.82 1.50 0.208 0.804
Tp (s) 12.022 12.605 −0.58 2.04 0.169 0.693
Dir (°) 296.686 290.12 6.57 13.16 0.044 0.760
Fig. 20. Direct comparison SWAN —Sines Buoy, signi ficant wave height, between 01.09.2000 –10.03.2001.
Fig. 19. Medium and high resolution area, energetic peak 1981/12/29 h00. a) Medium area for Lisbon –Peniche nearshore, bathymetric map and wave vectors. b) High resolution area,
significant wave height fields and wave vectors.917 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

model is shown in the Fig. 21 . The statistical results for the signi ficant
wave heights and mean wave period are listed in the Table 9 .
The direct comparisons at the Sines and Cascais buoys and the
statistical parameters show a good correspondence for the results
provided by the numerical wave models. There are good correlation
coefficients and RMS errors for the signi ficant wave heights and mean
periods. In terms of mean wave direction (only for the Sines buoy) theresults are in general good, with a very good result of scatter index.
Generally, the tendency to overestimate the signi ficant wave height
end mean period was maintained (negative bias), except some peaksthat were underestimated.
5. Conclusions
The present work is integrated into the objective of developing a
system based on numerical models able to assess the wave conditions
in the Portuguese nearshore. The outputs of this system are especially
focussed on the strategic areas from an economical point of view.
In validating the system, both the direct comparisons with the
buoy time series and the statistical parameters show a good cor-
respondence between the results provided by the numerical wave
models and buoy data. The results are acceptable in terms of sig-
nificant wave heights and periods and very good in terms of wave
direction (especially at the west Iberian coast).
A better resolution in the geographical space, complemented with
a higher quality of the bathymetry should lead also to the improve-ment of the results as concerns the signi ficant wave height and the
periods. Looking at the results it is obvious that the good estimationsof the wave directions are due to the higher resolution in the di-
rectional space (10°) as well as due to the fact that the wave pattern in
the Portuguese nearshore is most of the time regular (the incoming
waves are usually from North/West). In addition, it is possible that due
to the fact that the in fluence of the refraction is very strong in the
coastal environment that the errors induced by the wind fields and the
large scale model, are being attenuated by this effect.
The extended hindcasts performed in the coastal environment of
Portugal show that the wave conditions can be assessed here with a
good approximation. The new developments of the SWAN model asregards the increase of the computational ef ficiency, but also the
increase of the model performances in both directions (offshore andnearshore) offer new possibilities for calibrating better numerical
schemes. Thus the fact that the SWAN model allows quasi oceanic
scales permitted the implementation of the global SWAN area covering
all the coastal environment of continental Portugal (Figs. 2a and 15 a).
Some new capacities of SWAN, such as accounting for refl ections
and phase-decoupled diffraction, usually associated with the fine
areas are very effective tools for improving the results of the high-resolution simulations.
Acknowledgements
This work is a follow-up of the HIPOCAS project —Hindcast of
Dynamic Processes of the Ocean and Coastal Areas of Europe ( www.
mar.ist.utl.pt/hipocas ) which was partially funded by the European
Union, within the programme Environment and Sustainable Devel-
opment (EVK2-1999-00248), and uses wind and wave data produced
in that project.
The work of the first author has been funded by Fundação para
a Ciência e Tecnologia (Portuguese Foundation for Science and
Technology) under grant SFRH/BD/13176/2003.
References
Booij, N., Ris, R.C., Holthuijsen, L.H., 1999. A third-generation wave model for coastal
regions, Part I, Model description and validation. J. Geophys. Res. 104, 7649 –7666.
Booij, N., Haagsma, I.J.G., Holthuijsen, L.H., Kieftenburg, A.T.M.M., Ris, R.C., van der
Westhuysen, A.J., Zijlema, M., 2004. User Manual for SWAN, Version 40.41. Delft
University of Technology, Delft, the Netherlands. 115 pp.
Gómez, M., Carretero, J.C., 1997. A two-way nesting procedure for the WAM model:
application to the Spanish Coast. J. Offshore Mech. Arct. Eng. 119, 20 –24.
Guedes Soares, C., 2000. Probabilistic based models for coastal studies. Coastal Eng.
40, 279– 283.
Guedes Soares, C., Weisse, R., Carretero, J.C., Alvarez, E., 2002. A 40 years Hindcast of
Wind, Sea Level and Waves in European Waters, Proceedings of the 21st
International Conference on Offshore Mechanics and Arctic Engineering ,A S M E ,
Paper OMAE2002-SR28604.
Guedes Soares, C., Rusu, L., Pilar, P., 2004. Wave hindcast in the coastal environment
of Portugal, (in Portuguese). In: Guedes So ares, C., Gonçalves de Brito, V. (Eds.),
As Actividades Marítimas e a Engenharia. Edições Salamandra, Lisbon, Portugal,pp. 73– 82.
Günther, H., Hasselmann, S., Janssen, P.A.E.M., 1992. The WAM Model Cycle-4.0 User
Manual, Technical Report 4, Deutsches Klimarechenzentrum Hamburg.
Holthuijsen, L.H., Booji, N., Bertotti, L., 1996. The propagation of wind errors through
ocean wave hindcasts. J. Offshore Mech. Arct. Eng. 118, 184 –189.
Janssen, P.A.E.M., 1991. Quasi-linear theory of wind-wave generation applied to wave
forecasting. J. Phys. Oceanogr. 21, 1631 –1642.
Komen, G.J., Hasselmann, S., Hasselmann, K., 1984. On the existence of a fully developed
windsea spectrum. J. Phys. Oceanogr. 14, 1271 –1285.
Paillard, M., Prevosto, M., Barstow, S., Guedes Soares, C., 2000. Field measurements of
coastal waves and currents in Portugal and Greece. Coastal Eng. 40, 285 –296.Table 9
Statistical results for the entire wave field in the period 1st of December 1981– 28th of
February 1981
n=1 35 Bmed Smed Bias RMSE SI r
Hs (m) 2.120 2.403 −0.283 0.608 0.287 0.908
Tm (s) 7.704 8.900 −1.195 1.801 0.234 0.860
Fig. 21. Direct comparison SWAN —Cascais Buoy, signi ficant wave height, between 01.12.1981– 28.02.1982.918 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919

Pilar, P., Guedes Soares, C., Carretero, J.C., 2008. 44-year wave hindcast for the North
East Atlantic European Coast. Coastal Eng. 55, 861 –871 (this issue). doi: 10.1016/j.
coastaleng.2008.02.027 .
Rusu, E., Ventura Soares, C., Pinto, J.P., Silva, R., 2004. Extreme events and wave forecast
in the Iberian Nearshore. In: McKee Smith, J. (Ed.), Coastal Eng. 2004. World
Scienti fic Pub. Co., pp. 727 –739.
Rusu, L., Pilar, P., Guedes Soares, C., 2005a. Hindcasts of the wave conditions in approaches
to ports of the North of Portugal. Proceedings Fifth International Symposium on OceanWave Measurement and Analysis (WAVES 2005), Madrid, Spain, p. 9.
Rusu, L., Pilar, P., Guedes Soares, C., 2005b. Wave hindcast in the southern part of the
Portuguese continental nearshore. (in Portuguese). 4as Jornadas de Engenharia
Costeira e Portuária, Angra do Heroísmo, Portugal, p. 10.
Rusu, L., Pilar, P., Guedes Soares, C., 2005c. Reanalysis of the wave conditions in the
approaches to the Portuguese port of Sines. In: Guedes Soares, C., Garbatov, Y.,Fonseca, N. (Eds.), Maritime Transportation and Exploitation of Ocean and Coastal
Resources. Taylor and Francis, pp. 1137 –1142.Teixeira, J.C., Abreu, M.P., Guedes Soares, C., 1995. Uncertainty of ocean wave hindcasts
due to wind modelling. J. Offshore Mech. Arct. Eng. 117, 294 –297.
Tolman, H.L., 1998. Validation of NCEP's ocean winds for the use in wind wave models.
Global Atmos. Ocean. Syst. 6, 243 –268.
WAMDI Group: Hasselmann, S., Hasselmann, K., Bauer, E., Janssen, P.A.E.M., Komen, G. J.,
Bertotti, L., Lionello, P., Guillaume, A., Cardone, V.C., Greenwood, J. A., Reistad, M.,
Zambresky, L., and Ewing, J. A., 1988. The WAM Model – A Third Generation Ocean
Wave Prediction Model, J. Phys. Ocean., vol. 18, 177 p.
Weisse, R., Feser, F., 2003. Evaluation of a method to reduce uncertainty in wind
hindcasts performed with regional atmosphere models. Coastal Eng. 48, 211 –225.919 L. Rusu et al. / Coastal Engineering 55 (2008) 906 –919