Version of Record: https://www.sciencedirect.

com/science/article/pii/S0307904X20304443
Manuscript_8ab2a4092fa59f4e6b8e33e1fd5b7767

xIntegrated modelling of sea-state forecasts for safe navigation and

operational management in ports: Application in the Mediterranean Sea

Christos Makris1,*, Yannis Androulidakis1,3, Theofanis Karambas1,4, Andreas

Papadimitriou2,4, Anastasios Metallinos2,4, Yiannis Kontos1, Vassilis Baltikas1, Michalis
Chondros2,4, Yannis Krestenitis1, Vicky Tsoukala2, Constantine Memos2,4

1
Laboratory of Maritime Engineering, School of Civil Engineering, Aristotle University of Thessaloniki,
54124, Thessaloniki, Greece, e-mails: [email protected]; [email protected];
[email protected]; [email protected]; [email protected]; [email protected]

2
Laboratory of Harbour Works, School of Civil Engineering, National Technical University of Athens,
Athens, Greece, e-mails: [email protected]; [email protected];
[email protected]; [email protected]; [email protected]

3
Department of Ocean Sciences, University of Miami/Rosenstiel School of Marine and Atmospheric
Science (RSMAS), Miami, FL 33149, USA. e-mail: [email protected]

4
Scientia Maris, 15772, Agias Elenis Str. 10, Zografos, Greece. e-mail: [email protected]

* Corresponding author: tel.: +30 2310 995708; e-mail: [email protected]

ABSTRACT

In this paper we present recent evolvements of three robust numerical models for the simulation of
the evolution of wave fields and hydrodynamic circulation in gulfs and coastal areas with large
harbours and significant urban port facilities. The models are integrated into a single software suite
for the development of a decision support tool to provide reliable forecasts of sea states prevailing
at selected important ports worldwide. The application of the integrated modelling platform is
designed to support approaching procedures of vessels to ports and it is based on co-operating,
high-resolution, hydrodynamic (ocean and wave) models that derive input data and boundary
conditions from global scale or regional, open sea and weather forecasts. The implementation of
short-term forecasts for sea conditions includes the development, validation, coupling, and
operational application of: i) the High Resolution Storm Surge (HiReSS) model for sea level
variations; ii) the 3rd generation spectral wave model called TELEMAC-based Operational Model
Addressing Wave Action Computation (TOMAWAC) for irregular wave propagation in offshore and
coastal areas; and iii) a high resolution phase-resolving wave model (WAVE-L) for port basins, based
on the hyperbolic mild-slope equations. This innovative product, designed for port-related end-
users, will improve the navigation safety at ports, optimize the berth occupancy, support the port

1
© 2020 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
pilotage operations, mooring and towage procedures, and may facilitate the port layout upgrade or
design. Hereby, pilot forecast implementations are presented concerning the Mediterranean Sea
and eight selected harbours in it.

KEYWORDS

Integrated modelling; Surface waves; Sea level elevation; Sea state forecast; Ports; Navigation safety

LIST OF ABBREVIATIONS

2-DH Two-Dimensional Horizontal

AUTh Aristotle University of Thessaloniki

AVANTI Access to Validated Nautical Information

CMEMS Copernicus Marine Environmental Monitoring Service

CNPs Certified Navigation Pathways

COSYNA Coastal Observing System for Northern and Arctic Seas

E Percent Error

ECMWF European Centre for Medium-Range Weather Forecasts

EI Error Index

ESA European Space Agency

FDM Finite Difference Method

GEBCO General Bathymetric Chart of the Oceans

GLOSS Global Sea-Level Observing System

GMT Greenwich Mean Time

HiReSS High Resolution Storm Surge model

HNHS Hellenic Navy Hydrographic Service

HRP Hit Rate of Percentiles

IMO International Maritime Organization

ISPRA Italian National Institute for Environmental Protection and Research

2
MoC Method of Characteristics

MSL Mean Sea Level

NOAA National Oceanic and Atmospheric Administration

RMSE Root-Mean-Square Error

SBE26 Sea-Bird Electronics Seagauge Wave & Tide Recorder model 26

SLP Sea Level Pressure

SOCIB Balearic Islands Coastal Ocean Observing and Forecasting System

SSH Sea Surface Height

SSI Storm Surge Index

SWAN Simulating WAves Nearshore

TOMAWAC TELEMAC-based Operational Model Addressing Wave Action Computation

UST Universal Standard Time

WAM WAve Modelling cycle 4 model

WAVE-L AUTh’s hyperbolic mild-slope equation wave model

WRF-ARW Weather Research and Forecasting model with the Advanced Research dynamic
solver

WS Willmott Skill score

1. INTRODUCTION
1.1 Background and incentive

Intercontinental ocean and short sea shipping are important levers for both freight transport and
passenger travel movement, significantly contributing to the economic and social development on a
global scale, e.g. accounting for almost 90% of world trade. Ports are vital links in the chain of
maritime transportations and have a decisive impact on their quality. The basic port operations with
regard to visiting vessels include successive discrete procedures such as vessel approach by selection
of Certified Navigation Pathways (CNPs), towage, selection of mooring location or berth position,
freighting, arrival and departure of passengers, and sailing from and to ports. The increase in
ship/vessel size and the related transport load exercise strong pressure on port authorities, which
often hardly adapt to the aforementioned developments on time. This leads to an increase in both
number and cost of maritime accidents, the majority of which involve collisions with other vessels,

3
ship grounding on the port seabed or on harbour structures. The quality and safety of maritime
transportations are therefore dependent on prevailing ocean and meteorological conditions in
relation to the port’s configuration at the time of approach.

Recent navigation safety reports show that 60% of marine accidents are human-induced [1]; the
majority of vessel approach accidents could probably be avoided by means of proper navigation
support tools [2]. Recent reports from the British Department for Transport [3], based on 920.000
movements of vessels in and out of 13 United Kingdom (UK) ports over the 2005-2009 period,
showed that one accident occurred per 1000 movements on average, while for commercial vessels
the frequency was higher, i.e. one accident per 240 movements. Most of the accidents occurred in
the vicinity of mooring sites or exact at berth positions (45%) and along CNPs inside the port basins
(40%). Almost one third (34%) of the port accidents were due to incompetent navigation. Moreover,
the certification of navigation paths in port areas by the European Space Agency (ESA) conforming to
the International Maritime Organization’s (IMO) e-Navigation strategy requires knowledge of
currently prevailing conditions including sea state and related environmental data.

We seek to investigate the impact of sea states, in terms of detailed maritime conditions’
forecasting, on the abovementioned processes and the development of an operational forecast
platform with a decision support tool for both port pilots and ship masters in several ports of the
Mediterranean Sea. This application is developed in the framework of the research project Accu-
Waves [4, 5] which will cover the implementation of operational marine forecast systems to support
navigation in major ports worldwide.

The background literature and relevant modelling applications generally covering met-ocean
forecasts is vast, yet the existing research work focusing on the development of sea-state forecasts
in port areas, based on coupling between state-of-the-art regional ocean simulations to significantly
high-resolution wave simulations inside the ports, is not replete. The 24/7 delivery of near-future
marine predictions to vessel operatives during their approach to ports through the user-friendly
operational platform maintained by MarineTraffic [6] is an innovative product that will significantly
increase navigation safety.

1.2 Literature review on previous research

The subsistent literature on operational sea-state forecasting in coastal areas concerns the short-
term predictions of primarily spectral wave characteristics and secondarily ocean circulation.
However, the usual spatial resolution of the implemented models is rather coarse for the detail
needed in the vicinity of harbour structures and inside port facilities. This has to do mostly with the
choice of phase-averaged, instead of phase-resolving, wave models and available processing
resources under time-efficient computations. Selected relevant research endeavours are concisely
presented below.

4
The Naval Research Laboratory created a wave hindcasting/forecasting system in support of the
Nearshore Canyon Experiment field program with the use of Simulating WAves Nearshore (SWAN)
model focusing on the coastal area in the Southern California Bight [7]. The stationary assumption
for 3rd generation spectral wave model computations used in this system is criticized and crude
bathymetric resolution issues are highlighted especially in areas near islands in the Bight, having
strong impact on the nearshore wave climate [7]. Solutions are examined in terms of computational
efficiency and expensiveness for operational use. The proposed system [7] provides poor predictions
of the swell occurrence timing and local sea growth/decay in terms of ocean boundary conditions
that influence the directional distribution of wave energy in coastal areas. Besides the latter, it is
also shown that resolution and accuracy issues of available local atmospheric forcing may drive
erroneous results of spectral wave model predictions.

A real-time nearshore wave, tide and current prediction system was demonstrated during the
Maritime Rapid Environmental Assessment 2004 Trial in Portuguese coastal waters for regional scale
daily forecasts [8]. Global scale forcing input from meteorological and oceanographic centers were
utilized. The attempt also focused on a limited beach experiment with Delft3D model to build a
coastal hydrodynamic modelling system, in order to predict nearshore wave-induced and longshore
currents near Pinheiro da Cruz. Once again, the coarseness of modelled wave forecasts had an ill
influence on nearshore current reproduction [8].

An evaluation of a high-resolution operational wave forecasting system in the Adriatic Sea was
performed with the use of SWAN model using wind inputs generated by ALADIN-8 operational
atmospheric model [9]. Real-time surface waves’ forecast was pursued in order to trace high seas
and extreme waves during Bora and Sirocco wind storms. The model prediction skill was rather high,
evaluated against in situ field measurements and altimeter observations, but referred to open sea
areas compared to the analysis of nearshore data given by us in the present paper. Comparisons at
five coastal stations showed that the forecast wave heights were underpredicted by an average of
30% reaching up to 50% errors. Higher-resolution wind forcing combined with realistic inland
orography could decrease the observed wave forecast bias.

To ensure marine operations’ safety and support, an ocean weather prediction system by three-way
coupled wave forecasts was developed in Prince William Sound of USA [10]. The authors
commented on the issue of wave forecast reliability that leads to uncertainty of planning, managing
and engineering operations. High-resolution, 36-hours daily forecasts of significant wave heights
showed an acceptable correlation with field data, but rather for open sea than nearshore areas.

Large-scale wave forecasts are operated by the European Centre for Medium-Range Weather
Forecasts (ECMWF) [11]. Changes to input and dissipation source terms of WAM (WAve Modelling)
Cycle 4 model were proposed in order to improve simulated wave data. Comparisons mainly
referred to ocean buoys and offshore altimeter wave data, albeit for limited areas in the forecast
domain not reaching the coastal fronts and port areas.

5
A novel paradigm in operational wave forecasting systems was presented in [12] for the National
Center for Environmental Prediction using WAVEWATCH-III numerical model. The system used a
mosaic of two-way nested grids in a single model implementation in order to provide suitable
resolution for multiple areas of gridded forecast products. Although a spectral partitioning algorithm
was introduced to separate individual sea states from the overall spectrum, thus providing
additional products for multiple sea states, the system is applied on rather large-scale regions and
coarse spatial resolutions.

The evaluation of a high-resolution wave forecasting system in port approaches was presented in
[13]. The objective was to test the SWAN model for real-time information about sea states around
Portuguese ports. Inclusion of white-capping offshore wave breaking and increase of spatial
resolution led to considerable enhancement of wave prediction accuracy in nearshore areas,
especially where fine bathymetry and wind data were available.

In all the aforementioned research efforts, much larger areas were investigated compared to our
case presented herein. The last in dynamic downscaling succession and finest grid resolution for 3rd
generation wave modelling in the aforementioned would be the starting level of our model
implementations, i.e. the coarsest resolution in our coastal wave model coupling effort. Moreover,
we hereby attempt to further use phase-resolving models in super fine resolution (Δx ≤ 2 m) around
ports and inside harbour basins. Finally, the coupled wave models are fed with reliable model
forecasts for sea level elevation and depth-averaged currents in coastal areas due to both
atmospheric forcing and astronomical tides in the context of a built-in simulation approach for
tidally influenced storm surges.

1.3 Review of sea-state forecast applications

The significance of the presented research and other existing sea-state forecast platforms and
applications is demonstrated by IMO's strategy under code e-Navigation [14]. The latter aims,
through digital analysis and data dissemination, to reduce the adverse effects of unpredictable
human factor on navigation, and thus increase competitiveness and safety of maritime transport
[15]. One of the electronic services sections required in Strategic Plan 2015-2019 refer to Local Port
Services including high-resolution meteorological and hydrographic data and information on
mooring positions [16]. For example, the Docking Assist system [17] is an effort to support ship
sailing operations in ports, according to vessel position with regard to harbour structures
surrounding it at any time and therefore contributes to improving port management efficiency. A
similar tool called Physical Oceanographic Real-Time System (PORTS®), developed by the USA’s
National Oceanic and Atmospheric Administration (NOAA), uses real-time data from US ports and
provides forecasts of sea state parameters for safe inland navigation [18]. A more general
application, named Access to Validated Nautical Information (AVANTI), was recently developed with
initiative of the International Harbour Masters' Association and the UK Hydrographic Office and
provides series of marine port information to its users online [19].

6
Other online modelling applications, services and platforms of mid-resolution met-ocean forecasts
related to coastal areas, comprising but not necessarily focusing on ports, can be indicatively divided
to the following, among many in the field:

• On a global scale, there are fully versatile, freely provided, and widely used initiatives of open data
for ocean parameters, such as the European Union's Earth Observation Programme, Copernicus [20],
and NOAA’s Center for Operational Oceanographic Products and Services, which is the authoritative
US source for accurate, reliable, and timely tides, water levels, currents, and other coastal
oceanographic and meteorological information [18].

• On a regional scale, e.g. focusing on the Mediterranean and other European Seas, there are several
local forecast systems and services, such as WaveForUs [21], Poseidon [22], KASSANDRA Storm
Surge System [23], Balearic Islands Coastal Ocean Observing and Forecasting System (SOCIB) [24],
Coastal Observing System for Northern and Arctic Seas (COSYNA) [25], etc., usually covering ocean
forecast needs on a national level.

The bulk of the aforementioned applications could significantly contribute to the rapidly expanding
field of e-Navigation. However, these applications do not provide high-resolution sea level and wave
climate forecasts at port scale, i.e. sea-state conditions adjacent to harbour protection structures
inside port basins, rendering this discrepancy the main objective to tackle herein.

1.4 Scope of research

Within the Accu-Waves project [4] we currently develop a tool to provide fine resolution forecast
data on prevailing sea states in the vicinity of ports and inside harbour basins [5]. The goal is to
support navigation procedures of vessels’ approach to ports. The basic decision support tool is
developed in the form of a single software suite with inter-connected sub-codes, which is able to
provide a reliable 3-day prognostic depiction of maritime conditions in ports and their surrounding
areas with a 3-hour step. Related sea conditions refer to the following parameters:

a) weather data: wind intensity/direction and atmospheric pressure at sea level;

b) sea level elevation due to meteorological conditions and tidal effects, and respective ocean
currents’ intensity/direction;
c) wind-induced wave characteristics (significant wave height, spectral peak period and main
direction of wind waves propagation);
d) swell characteristics (height, mean period and main direction of long oceanic waves);
e) seiches characteristics in ports.

In this paper, we present recent developments of three robust numerical models for the integrated
simulation of sea level variations, wave propagation and transformation in gulfs and coastal areas
with port facilities. Model H (HiReSS) [26, 27, 28, 29, 30, 31] is a 2-DH hydrodynamic (storm surge)
model for the simulation of barotropic circulation and sea level variations, based on the depth-
averaged shallow water equations (Subsection 2.1.1). Model A (TOMAWAC) [32, 33, 34, 35, 36] is a
7
3rd generation spectral model that simulates wind-induced irregular offshore wave fields on a
triangular finite element mesh covering areas of a few dozens of Km2 across port approaches (see
Subsection 2.1.2). Model B (WAVE-L) [37, 38, 39] is based on the hyperbolic mild-slope equation and
it simulates the transformation of complex wave fields in harbours and coastal areas in the vicinity of
ports with varying bathymetries (Subsection 2.1.3).

Main goal of this study is to develop new features of models H and B and formulate a model A setup
in order to render them fully operational as an integrated ensemble of simulation software for high-
resolution forecasts of sea level variations, depth-averaged currents, spectral and regular wave
fields, in areas around and inside port basins. A second goal is to calibrate and validate all models
against sea level and wave data from in situ observations by tide gauges of either available national
hydrographic services [40] or our own field measurement operations, and from laboratory
experiments of wave propagation and transformation at flume scales [41, 42]. Moreover, the
numerical algorithms of the hydrodynamic and wave models are fitted in an integrated modelling
system suite for automated operational forecasting of wave characteristics, surge-induced and tidal
sea levels in and around significant ports of the Mediterranean Sea [5]. Expansion of application to
50 important ports worldwide is ongoing and set as an ultimate research goal. The final product is
consisted by maps and high-resolution datasets of sea levels and wave disturbance characteristics
for 3-day forecasts useful for navigation around and inside ports.

2. METHODOLOGY

The operational forecast implementation is based on the following steps:

a) Adaptation of three hydrodynamic numerical models;

b) Calibration, testing, and integration of the models into a single suite;
c) Implementation of the above suite to 8 ports in the Mediterranean Sea;
d) Adaption of the forecast application in the operational MarineTraffic platform [96] providing
wind, wave, sea level and current data at 3-hour intervals for 3-day forecasts.

Model B tackles high-resolution wave propagation and transformation inside harbour areas, while
model A provides wave boundary conditions to model B, with model H contributing input data of sea
levels and mean currents to both models A and B.

2.1 Numerical models

2.1.1 Model H: HiReSS

High Resolution Storm Surge (HiReSS) numerical model is built on a Fortran code, developed in the
Laboratory of Maritime Engineering in Aristotle University of Thessaloniki [27, 43]. Model H
simulates the 2-DH barotropic mode of hydrodynamic circulation in large water bodies, enclosed
seas, gulfs and coastal areas over a rather shallow continental shelf, based on the shallow water

8
equations [44, 45]. Thus, model H can predict the elevation of sea level (Sea Surface Height; SSH)
and the depth-integrated sea currents (induced by atmospheric forcing; wind and pressure) [46].
HiReSS can take into account the combinatory effects of several processes, such as the inverse
barometer (response of sea level to atmospheric pressure gradient of large barometric systems);
shear stresses of wind applied on the air-water interface; geostrophic Coriolis forces on large water
masses; astronomical tides; ocean bottom friction; turbulence of horizontal vortices through the
eddy viscosity concept; impacts of the wave-induced mean flows (Stokes drift) on the wind-driven
currents in open seas; additional sea surface set-up caused by wave breaking in nearshore coastal
zones. Therefore, model H can reproduce the variations of sea level (i.e. positive or negative surges),
ranging from decimeters to a few meters and lasting from several hours to a few days. The main
advance of the HiReSS model is the tidal signal parameterization in the Navier-Stokes equations. The
extended continuity and momentum equations, in order to account for meteorologically driven and
tidally affected circulation can be written as:

+ + =0 (1)

√
+U +V − V+Z = − +E + + !" #
− C% !" #
(2)

√
+U +V + U+Z = − +E + + !" #
− C% !" #
(3)

Z = 0.9g − 0.7g +,-.

, Z = 0.9g − 0.7g +,-.
(4)

where ζ is the free surface elevation in the continuity equation (Eq. 1) yielding values of SSH; h=d+ζ is
the total water depth of the sea and d is the local still water depth; U and V are the depth-integrated
horizontal velocity components along the x and y axes of an ortho-regular staggered Cartesian grid
of the Arakawa-C type for the Finite Difference Method (FDM); t is the time; fC is the Coriolis
coefficient; g is the acceleration of gravity; PA is the atmospheric Sea Level Pressure (SLP); ρo is the
average seawater density; Cb is the bottom friction coefficient following a logarithmic wall-law on
the bed boundary layer [26, 47] and Cs is the air-water drag coefficient [26, 48]; Wx and Wy are the
wind velocity components at 10 m above Mean Sea Level (MSL); Εh is the horizontal eddy viscosity
coefficient for internal friction forces is based on a Smagorinsky-type of approach [26, 29].

The boundary conditions of SSH or current velocities on the unique open boundary (i.e. the Gibraltar
strait) of the Mediterranean domain are implemented via a Dirichlet-type approach (a priori known
values of parameters), e.g. ζ=ζtide(t), from a simplistic static model for astronomical tides (see below)
or automated forecasts from established databases [20]. The boundary conditions of SSH or current
velocities on the coastal solid boundaries are based on the approximation of irregular flow and are
of the Von Neumann type (a priori known boundary-adjacent variable gradient) [49]. The chosen

9
numerical scheme of integration is an explicit “leap-frog” algorithm with small time step dt ≈ 30 sec,
in order to keep a sufficiently low Courant number.

Model H also takes into account the effects of astronomical tides on barotropic circulation (Zx and Zy
terms of Eqs. 2, 3, and 4) through a static model parameterization [50], following a formulation that
combines the equilibrium tidal potential with the self-attraction/loading effect under specific
coefficient parameterizations [51, 52]. Tide forecasts are based on the solution of harmonic equation
ζtide on all grid cells with discrete longitudes and latitudes, concerning both semi-diurnal and diurnal
tidal range signals [46]. Forecast of the tidal oscillation signal is based on the solution of a harmonic
equation, which immediately applies the tidal balance budget (or potential of astronomical tide
generation) by using the numerical series of Schureman [53] with equations of partial tidal
harmonics:

ζ = k o sin2 ( ϕ ) cos ( σ t t + x + 2λ ) , v = 2 
 tide,2 
ζ tide = ∑ ζ tide,v ( λ, ϕ,t ) ζ tide,1 = k o sin ( 2ϕ ) cos ( σ t t + x + λ ) , v = 1  (5)
 
( )
v

ζ tide,0 = k o 3 sin ( ϕ ) cos ( σ t t + x ) , v = 0 

2

Specifically, the second one (ν=2) corresponds to the semi-diurnal tidal signal (e.g. M2 principal lunar
mode), the third one (ν=1) to the diurnal tidal signal (e.g. P1 principal solar mode) and the fourth
one (ν=0) to the long-period species of tidal signals; t is the Universal Standard Time (UST), λ and φ
the geographic spherical coordinates, ζtide is the oceanic total tide amplitude, ko is the magnitude of
the partial tide, σt is the identical cyclic frequency of the partial tidal amplitude, and χ is the
astronomical argument for every partial tidal budget (half range between high and low waters) in
relation to midnight zero at the Greenwich meridian (00:00 Greenwich Mean Time, GMT). The major
constants and values of all the basic components of tidal modes are presented in Table 1. The mean
longitudes of the sun and moon and the lunar perigee, ho, so, at Greenwich midnight are also given
by:

ho = 279.69668 + 36000.768930485Td + 3.03 ⋅ 10−4 Td2

so = 270.434358 + 481267.88314137Td − 0.001133Td2 + 1.9 ⋅ 10−6 Td3 (6)
Td = ( 27392.500528 + 1.0000000356D ) 36525
D = day + 365 ⋅ ( yr − 1975 ) + int ( yr − 1973 ) 4

where day is the number of day per year (day=1 on January 1st), yr is the year of study from 1975
and on, and int[ ] is the sign for the integer function of a number.

2.1.2 Model A: TOMAWAC

TELEMAC-based Operational Model Addressing Wave Action Computation (TOMAWAC) [32] is an

open-source code for a 3rd generation, phase-averaged, directional, spectral wave model, developed
by Électricité De France R&D’s Laboratoire National d’Hydraulique et Environnement [33, 34]. It
simulates the evolution in space and time of the spectrum of sea surface elevation in waters of any

10
depth. The numerical calculations are executed by the Finite Elements Method over an unstructured
mesh. It is noted that the size ratio of the largest to the smallest computational element can exceed
a value of 100, so TOMAWAC is also suitable for both offshore and nearshore applications, where
high computational resolution is required. The model captures processes of wind wave generation
and propagation; energy dissipation due to white-capping, bottom friction, wave shoaling and
depth-limited breaking; non-linear triad and quadruple wave–wave interactions, wave-structure
interaction (diffraction), wave-current interaction [54]. The main equation used by TOMAWAC for
the evolution of the directional spectrum of the wave action density Ν is as follows, written in the
form of a radiative transfer equation [55]:

12 1!45 2# 1!65 2# 17859 2: 1785; 2:

13
+ 14
+ 16
+ + = <7=4 , =6 , >, ?, @: (7)
189 18;

EF 8;
+ D6 , =5 > = 1H 14 − = 14 , =5 J = 1H 16 − = 16
1G 1H 1I 1G 1H 1I
>5 = AB =4 /= + D4 , ?5 = (8)
8

where k=2π/L is the wavenumber, L is the wave length, U=(Ux,Uy) with U=|U| is the transport rate
velocity and its components through geographic and spectral space (x, y and σ, d respectively)
derived using the linear wave theory, Cg is the relative (or intrinsic) group celerity of waves (as
observed in a moving frame of reference), C=σ/k the wave celerity, σ the cyclic frequency, and Q is
an ensemble of source terms accounting for the generation and dissipation of waves due to all
aforementioned wave processes [56]. Further details are described analytically in TOMAWAC user
manual [57].

The numerical solution is done on a planar 2-D computational domain (for water areas) built upon a
triangular mesh. The mesh-based discretization technique inherently allows for variability of cell
sizes particularly facilitating resolution refinement in areas with complex geometries (gulfs, straits,
bays, port approaches, coastlines, etc.) and rapidly diverging bathymetries. The user-defined density
of spatial discretization points is always chosen to match the spatial scale of variation of the
bathymetry, as the forcing input of wind fields is rather crude in terms of resolution. TOMAWAC’s
solver deals with a transport (convection-type) equation (Eqs. 7, 8) with source terms, that leads to
the calculation of four-variable functions on a 4-D mesh in spherical or Cartesian coordinates, angle
and frequency domain (φ, λ, θ, f) corresponding to a transport vector. TOMAWAC’s solver uses a
fractional step method, viz. the two steps of convection and source term integration are completed
in a successive way. The convection step is treated by the Method of Characteristics (MoC; piecewise
ray method). In case diffraction is not taken into account and the water depth is constant over time,
the characteristics have to be traced only once, at the beginning of the computation. This makes the
propagation scheme very fast and efficient. The source term integration is carried out through a
semi-implicit scheme [58].

2.1.3 Model B: WAVE-L

11
The WAVE-L model [37, 39] is based on the 2-DH, depth-integrated, harmonic, hyperbolic
formulation of the mild-slope equation for wave propagation [59]. It is a phase-resolving wave
model and simulates regular or quasi-regular wave propagation in coastal waters of mildly sloping
beds [60]; it resolves processes of wave-current interaction, wave shoaling, refraction, diffraction,
wave reflection at solid boundaries, energy dissipation due to bottom friction, and depth-induced
wave breaking in a combinatory way [38, 61]. The numerical solution of the equations is based on an
explicit scheme applied on a grid staggered between the cell values of surface elevation and mean
velocities. Along the open sea and lateral boundaries sponge layers are used.

The basic equations for mass and quantity of motion conservation can be derived by replacing both
pressure and velocity distributions that correspond to linear theory (for small amplitude waves) in
the linearized Navier-Stokes equations (valid for periodic wave propagation from deep to shallow
waters). Thus, for numerical simulations of wave transformation in 2-DH (depth-averaged)
formulation, the continuity and momentum equations can be written as:

K ! L "# ! L "#
+ + =0 (9)

7M K: NK "
L
+" − " MOP !Q"#
=v L
+v L
− f% σUU (10)

7M K: NK "
L
+" − " MOP !Q"#
=v L
+v L
− f% σVU (11)

where η is the wave-induced free-surface elevation, Uw and Vw are the depth-integrated horizontal
(wave-driven) velocity components along the x and y axes, respectively, c is the wave (phase)
celerity, σ=2π/Τ is the wave angular frequency, fb is the normalized bed friction coefficient, vh is the
horizontal eddy viscosity coefficient. The above equations result from the replacement of pressure
and velocity distributions, from linear theory (for rather short waves), in the linearized Navier-Stokes
equations, and thus provide the ability to describe transmission of simple harmonic undulations
(monochromatic waves) at any depth. Extra terms for energy dissipation are further added and
specialized modifications of the wave generation mechanism can also account for quasi-irregular
wave propagation inside the computational domain.

Depth-limited wave breaking in shoaling areas, shallow waters, and even above submerged
structures or alongshore bars, can be modeled in both monochromatic and pseudo-spectral waves’
simulations using the eddy viscosity concept for Reynolds stresses that are expressed via a
coefficient νh in the r.h.s. of the momentum equations (Eqs. 10, 11) [62], where D defines the energy
dissipation due to wave breaking [63], and coefficient Qb can be derived, based on the Rayleigh
distribution assumption for wave trains in nearshore areas, from the equations:

X /Y
v = 2h (12)

12
 abc fgh `
D = Q % fP ρgH_
`
, where = for irregular waves (13)
[ de bc fh

m. `nF
!op − opq3 #
k
D = ij or D= for monochromatic waves (14)
[ l

where fs is the mean spectral frequency (fs=1/Tm, Tm mean spectral wave period), Hm is the maximum
wave height with Hm=γd, γ is the wave breaking parameter (γ≈0.55-1.0) for spectral waves, H is the
regular wave height, T is the monochromatic wave period, overbar denotes mean values, E=ρgH2/8
is the wave energy, Est corresponds to Hst=0.42Hb, Hb=γd is the regular breaking wave height, Qb is
the percentage of breaking waves at a particular depth d, and Hrms is the root-mean-square wave
height Hrms=2(<2η2>)1/2 and brackets <∙> denote time-averaged values. It is inferred that for the total
prevalence of breaking waves Qb=1 whereas for non-breaking waves Hrms<<Hm, i.e. Qb<<1. This
modelling approach simulates breaking of random waves in complex bathymetries, conforming to
the requirements of operational pseudo-random wave forecasts with model B.

Energy dissipation due to bottom friction is modeled using the linearized (normalized by local depth
d) terms in the r.h.s. of the momentum equations (Eqs. 10, 11) in x- and y-directions of the Cartesian
horizontal plane. The linearized bottom friction coefficient fb is a function of the wave-induced
velocity and the wave friction coefficient fw, following the relationship:

f% σ = f rUÙ + VÙ ⁄d
` U
(15)

In model B, the wave generation can be simulated on any longitudinal and lateral boundary
simultaneously, with corresponding expansion of the peripheral sponge layers by an exponential
damping factor of the wave energy content [64], DF(x) as:

DF!x# = expy72a /z
−2 { /z : ln ~•, 0 < > ≤ >q (16)

where xs is the width of the sponge layer, b=1+rs+exp(-1/rs), where rs=10/Ns and Ns is the number of
grid points inside the sponge layer. This way, we are now capable to spatially restrict the
computational field in areas adjacent to harbours and thus reduce demand of computational time
and resources.

Monochromatic waves are generated along any boundary line by a sinusoidal excitation equation:

… Ž3
‚ƒ∗ = 2 ` sin ˆ7@ − @‰ : cosŒ• Ž4 (17)

where tf=sin(φ)∙x/c and φ is the angle of incident wave direction. Moreover, the new version of
model B is modified to simulate multi-directional, quasi-irregular waves (multiple frequency waves
generated on the boundary, yet propagating with single group celerity). The generation and
propagation of spectral waves may furthermore account for several different angles and directions
simultaneously, practically following a modelling approach [65] that provides the directional

13
spreading function D(f,θ) by the Fourier series representation for the wrapped normal spreading
function [41], as:

e“h
D!f, θ# = + ∑—
e˜ exp ’− ” cos•n!θ − θ_ #– (18)
`• • `

where N is the number of terms in the series, θm is the mean wave direction (0° or 45° in our test
cases), and σm the directional spreading parameter (either 10° or 30° in our tests).

Partial and full reflection of incipient waves from harbour structures are modeled based on an
updated version of the Karambas and Bowers [66] modelling approach of an extra dissipation term
in the r.h.s. of the momentum equations (Eqs. 10, 11) inserting a turbulent eddy viscosity coefficient
νγ:

1 œ 1 œ 1 I 1 I
… š› + 16 , … š› + 16 for the x- and y-components (19)
14 14

where νγ is calculated via a system of complex equations (based on a complex wave number K) of the
friction coefficient fs, thus rendering them iteratively solvable for given values of the reflection
coefficient CR from literature.

If we assume that for a typical distance of 2Sw (Sw: structure width) the water depth in front of a
solid boundary (breakwater, seawall, quay, etc.) is constant, and that the area of implementation of
the coefficient νγ lies inside 0 ≤ x ≤ 2Sw with a full reflection applied exactly at x=2Sw, for an incident
wave with amplitude, ai, and reflected amplitude, ar, it follows [61]:

‚ = •ƒ ž ƒ!G3a84# + •Ÿ ž ƒ!G3 84#

(20)

ˆ ` − š› ˆ¡ ` = • ` ¡ ` (21)

where σ2=gk∙tanh(kd) (k=2π/L being the wave length), and Κ is transformed into a complex number
that satisfies the relation of Eq. 22. An analytical expression for Cr=|ai/ar| can be finally derived as
follows:

£
7 ¥ ¦§¨£©ª :a7 a¥ ¦§¨£©ª :
AŸ = ¢£¤ « (22)
7 ¥ ¦§¨£©ª : 7 a¥ ¦§¨£©ª :
¤

If the Cr value is known, the aforementioned system of equations can be solved with an
approximation method and thus yield the value for the coefficient νγ. The reflection coefficient Cr for
rubble mound breakwaters could be estimated from classic empirical relations in past literature, e.g.
[67].

WAVE-L is a stable and robust model for the detailed simulation of wave propagation, within any
port configuration, based on a Fortran code developed in the Laboratory of Maritime Engineering in
Aristotle University of Thessaloniki (AUTh), rendering it an efficient tool to consider alternative
scenarios of wave penetration in ports and wave agitation near coastal structures in harbours and

14
marinas. The numerical solution of model B’s equations is based on an established, explicit,
validated scheme of FDM, which has been traditionally applied in numerical integration of long wave
equations on a staggered orthogonal grid [68, 69].

2.1.4 Model integration setup

The main steps taken to ensure an efficient set up of the operational simulations and the models’
interactions are:

a) The general mode of model H in the Accu-Waves framework (global mode) is forced using
boundary and initial configuration data (if needed for ports in open sea areas) from the
Copernicus Marine Environmental Monitoring Service (CMEMS) [20, 70, 71], and global-scale
atmospheric input from NOAA [72]. The presented application for the Mediterranean Sea is
forced by higher resolution (~10 km) meteorological simulations based on the Weather
Research and Forecasting model with the Advanced Research dynamic solver (WRF-ARW,
version 3.2.0) developed in AUTh [28, 73, 74, 75, 76].
b) Model A is executed using spatial and temporal results of SSH and mean current speeds (U, V),
in conjunction with data from the CMEMS database [77, 78], concerning open-sea significant
wave height (Hs), spectral peak wave period (Tp), and mean wave direction (ap). The
atmospheric forcing is the same as for model H. Information on the horizontal current velocity
components and sea surface elevations obtained from model H are interpolated to model A’s
mesh to model the effect of wave-current interaction.
c) Results obtained from model A in terms of Hs, Tp and ap are extracted along the wave
generation line of model B. It should be noted that the values of the aforementioned variables
are extracted at coordinates that coincide with the cell centers of model B boundaries, to
ensure compatibility and smooth interaction between the two models. Definitive bathymetry
for model B execution is obtained by adding model H’s SSH to d.

In general, the three aforementioned numerical models tackle different needs in terms of area
coverage and accuracy; schematics of the sequence of model interaction are provided in Figure 1.
The refined computational domain of model B is geographically nested to the coarser one by model
A. The nesting techniques for model H output used as input in models A and B is based on a classic
Wiener–Kolmogorov prediction algorithm for spatial analysis such as the linear or nonlinear Kriging
method (depending on spatial complexity of the domain), i.e. by minimizing the variance of the error
estimate ε of a random variable Z (representing any scalar maritime parameter, as Hs or SSH) at a
central point of interpolation x0 as follows:

¬!>m # = -®!>m # − -!>m # = ∑2

ƒ˜ ¯ƒ !>m # × ±!>ƒ # − -!>m # (23)

where i=1,N is the number of assimilated/interpolated grid points and wi are the individual spatial
weights based on each point location compared to x0. For the interpolation of vectoral parameters
(e.g. currents U, V), the Nearest Neighbor algorithm is implemented.

15
2.2 Ports and maritime areas of application

The implementation of the operational integrated model suite concerns eight very important ports
of the Mediterranean Sea, i.e. three Greek, two Spanish, and one Italian, Israelian, and Moroccan
commercial harbours (Figure 2). Nevertheless, the final outcome of the Accu-Waves initiative [4, 5]
will be to build a sea-state forecast application which is intended for 25 coastal regions globally,
covering areas with complex bathymetries and diverse coastlines. These areas will contain in total 50
port facilities with high traffic load and significant commercial interest, serving international shipping
operations. Bathymetric information in the relevant sea areas of the selected port sites were
obtained through local services, as the Hellenic Navy Hydrographic Service (HNHS) [40], the
Navionics platform [79], and the General Bathymetric Chart of the Oceans (GEBCO) database [80].

In general, model H is applied to much larger water bodies than the vicinity of a single port in order
to capture large scale meteorological processes present over such broader areas (Table 2). Herein
we present pilot implementations of model H in the Mediterranean Sea with 1/20o (almost 5 km)
spatial resolution (Figure 2). Results focus on the harbours of Algeciras and Barcelona in Spain,
Genova in Italy, Haifa in Israel, Thessaloniki, Patra and Piraeus in Greece, and Tanger-Med in
Morocco to provide crucial information of sea level predictions locally and input to models A and B
in these areas.

Regarding model A (Table 2), ad hoc delineations of the sea area are performed, where
environmental input data are sought from globally established met-ocean data sources (see §2.3).
Overall expanse of interest around a port is defined as the water surface of a circular area centered
at the port with a 3 to 45 km radius. Model A is one-way coupled to the coarser model H, covering
maritime areas around ports of typically ≤2500 Km2. For the case of swell dominated sea states, a
modification of model A’s spectral spreading parameter s is used to accurately take into account the
long wave propagation [81].

Pilot implementation of model B is carried out in all Mediterranean ports (only the three largest
Greek harbours shown herein for the sake of brevity). The high resolution of model B is based on
Kriging interpolations of the finest available resolution bathymetric depth charts [40, 79]. Model B is
one-way coupled to the coarser model A for input boundary conditions; the port area of typically
≤10 Km2 is integrated in larger domains. It also receives input of local changes in bathymetry from
model H [31, 39]. Model B’s resolution corresponds to a very fine discretization step of Δx ≤ 2 m
inside the port basin, in order to appropriately resolve and describe waves with length L ≥ 10 m.

3. VALIDATION OF NUMERICAL MODELS

The evaluation of the models’ performance is conducted by comparisons of simulation results with
available field observations and experimental data. The validation parameters and skill scores refer
to sea levels (model H) and characteristic wave heights (models A and B).
16
3.1 Field observations and experimental data

We performed field observations in the Thessaloniki port during autumn and winter of 2019 with the
use of a Sea-Bird Electronics measuring instrument, namely SBE26 Seagauge Wave & Tide Recorder
[82] (Figure 3a). SBE26 combines nonvolatile flash-type memory with stable time-base and a high-
frequency quartz hydrostatic pressure sensor to provide real-time data of wave and tide recordings
of high accuracy. For tide and water level monitoring, the pressure sensor output was integrated to
average out wave action with a fine time-resolution of one minute. The wave features records were
based on burst sampling, with one sample per 0.25 s, i.e. giving a burst duration of almost 8.6 min
for 2048 samples in every hour of the day. All calculations were performed with the use of a modular
software package Seasoft for Waves [83], which provides all pre-deployment planning,
communication setup and uploading of pressure data from the SBE26, separation of the uploaded
data into separate wave and tide files, removal of local atmospheric barometric pressure from tide
data, statistical analysis of spectral wave parameters, etc. [84]. The needed local, real-time,
(atmospheric) barometric pressure data were provided by the local weather station of AUTh’s
Department of Meteorology and Climatology via their Meteorological Observations online service
[73]. The wave burst data were processed to compute wave statistics, results from auto-spectrum
analysis, statistics from surface wave zero-crossing analysis, thus producing the observed timeseries
of spectral wave characteristics, and the relevant Fast Fourier Transform coefficients.

Figure 3 presents depictions of the measuring equipment together with the chosen locations of in
situ observations performed in the Thessaloniki port basin. Two sets of field data were retrieved:

a) for tidal and sea level measurements during 30/09/2019 – 14/11/2019,

b) for tidal, sea level and wave measurements during 12/12/2019 – 28/12/2019.

The water depth at the point of measurements was around 10 m for the first set of field recordings
and around 12 m for the second set of in situ measurements. These types of instrumentation can
produce data only for progressive waves. Thus, for our case of a nearshore station and in order to
avoid reflections from vertical structures and waterfronts in the port basin, we chose the location of
SBE26 gauge immersion to be in front of a wave-dissipating beach-type formation with sloping
bathymetry, i.e. ahead of the end (roundhead) of Pier 6 in Thessaloniki port (Figure 3b).

Additional tide-gauge measurements of hourly- and daily-averaged SSH values were collected from
available data sources. The Global Sea-Level Observing System (GLOSS) and Italian National Institute
for Environmental Protection and Research (ISPRA) databases [85, 86] provided sea level
information for three Italian stations (Genova, Trieste and Venice; Figure 2). Observational data from
four Greek stations located in port basins (Alexandroupolis, Chios, Lefkada, and Thessaloniki; Figure
2) were also obtained by HNHS [40]. Both datasets were retrieved in order to validate model H
simulations. The recording periods correspond to 1995-2005 and 2012-2015. Values of SSH field
observations were derived from measured data after subtraction of the MSL, determined by a post-
processing heuristic technique of a high-pass filter with a cut-off frequency of 1/30 days. This was
done in order to exclude long-term monthly oscillations of the sea level induced by steric effects [87]

17
in the Mediterranean; these types of effects cannot be simulated by the 2-DH barotropic model H.
Moreover, referring to long-term hindcasting storm surge climate runs (see Subsection 3.2), model H
did not include computation of the astronomical tides. For these cases the SSH tide-gauge data (at 4
Greek stations from 1995 to 2005) were de-tided using the T-Tide software [88].

The evaluation of model B’s performance was conducted by comparisons of simulation results with
experimental data of both regular and irregular wave propagation and diffraction around semi-
infinite breakwaters and through breakwater gaps [42, 89]. We also numerically reproduce an
elliptical shoal experiment [41], concerning irregular incident waves, and therefore also test a
directional pseudo-spectral wave generator for mild-slope equation wave models.

3.2 Model H validation

The HiReSS model has been applied in the past on a number of regions comprising large water
bodies and marginal seas [45], and further calibrated and thoroughly validated via comparisons of
hindcast modelling results against in situ observations for either short periods with intense storm
conditions [28, 31] or large periods (> 15 yr; for extreme events of SSH) in the Mediterranean,
Aegean and Ionian Seas to investigate the impact of climatic changes on the coast [26, 29, 30].

The daily-averaged simulated SSH time series (in hindcasting mode) provided by model H agrees well
with the respective tide-gauge measurements during 2012 in the ports of Alexandroupolis,
Thessaloniki, Genova and Venice (Figure 4). For further quantitative evaluation of model H’s
performance the root-mean-square error (RMSE) is presented on the graphs, followed by the classic
Pearson product-moment correlation coefficient r. Moreover, the Willmott Skill (WS) score or Index
of Agreement [90] is also calculated for SSH. The higher the WS and r (with ≤1 as a limit), the better
match is reached between simulated values of SSH and observations from tide gauges. High
correlation coefficient values and prediction skills were achieved, e.g. r=0.52-0.78 and WS=0.68-
0.82. Acceptable errors are also tracked reaching down to 12.8% of the occurred SSH maxima. It
should be noted that the forecast skill of the storm surge model highly depends on the accuracy and
resolution of the atmospheric forcing input [29, 30, 31]. A general ability of model H to capture high
seas is confirmed. The latter is further corroborated by the results of Figure 5, which portrays
comparisons of SSH hindcasts with model H in Venice and Trieste, an area with extremely strong
storm surge events [26], from 2013 to 2015. Good agreement is achieved between the modelled
results and in situ observations of combined weather- and tide-induced sea level variations. Small
discrepancies are found for large values of SSH, i.e. model H slightly overestimates the total sea
level.

In order to further investigate the aforementioned discrepancy and thus evaluate model H’s ability
to capture characteristic values of SSH extremes, which are crucial to navigation and berthing in
ports, we used the Storm Surge Index (SSI) [87] for comparisons of model and field data. SSI is
calculated as the average of the three highest independent storm surge maxima per year; 5-day
separated events were only considered referring to the estimated maximum duration of a

18
Mediterranean storm [91]. Figure 6 presents the evaluation of model H, based on comparisons of
long-term hindcasting model results against in situ observations by tide gauges during 1995-2005 in
four Greek ports (Figure 2). Several statistical measures for inter-annual maxima of tidal-free SSH are
compared, i.e. absolute SSHmax, SSI, and 99th percentile of SSH, together with the probability of
occurrence of locally observed SSHmax. The comparisons support the generally efficient performance
of model H, ranging from plausible to very good, with rather small errors for the 10-yr averages of
SSH extremes. A tendency to sufficiently estimate the statistically significant high values rather than
the absolute maxima, e.g. the 99th percentile in all stations, was also noticed. Errors range from
miniscule, e.g. 4‰ in Thessaloniki and Chios ports for the 99th SSH quantile to 13% in
Alexandroupolis port for the 10-yr extreme SSH. The probability of SSH local maxima occurrence is
also well estimated by model H, indicating that it can capture the frequency of appearance of peaks
in sea level elevation time series at coastal areas and ports [92, 93].

As extra measures of comparison for model H’s skill to reproduce sea levels, the corresponding
Percent Error (E) and Error Index (EI), based on SSI values, were also calculated as:

 SSImod + SSI obs 

(
E(%) = 100 ⋅ SSImod − SSIobs ) 
 2

 (24)
(
EI = SSI mod − SSI obs ) (
σ SSI
2
mod
+ σ SSI
2
obs )2
where the overbar denotes temporal averages, as derived from modelled (mod) and observed (obs)
data, and σSSI is the standard deviation of the inter-annual variability of SSI time series for each
station. E and EI turn positive when there is model overestimation of the sea level elevation over the
field values. Furthermore, we calculated the Hit-Rate-of-Percentiles (HRP) index [94] for the entire
10-year SSH time series, in order to examine if bias correction is necessary for modelled storm surge
results. HRP-index is calculated as the sum of the categorical fractions, i.e. differences between the
sorted (from 1st to 99th) percentiles of simulated and observed values of SSH, compared to an
allowed deviation. The latter was taken equal to half the average standard deviation of the modelled
and in situ SSH time series, i.e. (σSSHmod+σSSHobs)/2 that is 5-10% of average observed SSHmax. Model H
was found to score higher than the limit of 0.95 in average and reached the value of 0.98 at the
Thessaloniki port station. Therefore, bias correction was not considered necessary for forecasts of
SSH.

Table 3 presents collective evaluation by statistical skill scores of model H for the long-term climatic
hindcast results against tide-gauge observations of SSH maxima during 1995-2005 in four Greek
ports. The largest simulation error (yet acceptable, EI ≈ -0.561) is traced in Alexandroupolis station
where, however, the HRP-index is quite high denoting good reproduction of the statistical
distributions of SSH time series. In Thessaloniki port, the skill scores are high with low errors (E <
7%). Overall skill factors at the aforementioned Greek ports reveal a high average correlation (r =
0.841) with low RMSE = 0.026 m, and appreciable WS = 0.712.

19
3.3 Wave models A and B validation

The TOMAWAC wave model has been used extensively for wave propagation and coastal
engineering studies for over 20 years and thorough validations of the model with experimental and
field data has been conducted [33, 34, 35, 95]. For a detailed overview of the validation procedure
for the TOMAWAC wave model the reader should refer to the above. In this paper, we focus on
fundamental validation only for models H and B, which were created by members of our research
team. Nonetheless, we present validation of operational runs of models A and B in Subsection 3.4.

Model B is based on the WAVE-L code which has also been applied at various coastal areas in the
past, especially for the design of real-life harbour projects. Extensive validation via comparisons of
modelling results against experimental data [96] referred to monochromatic wave propagation and
refraction over an elliptic shoal on a 1:50 plane sloping seabed [61].

Hereby, an elliptical shoal experimental setup with a directional spectral wave generator [41] is also
numerically reproduced with model B as a test of quasi-irregular propagation over an uneven
bottom. The numerical wave flume is 35 m wide and 29 m long with a constant water depth of 0.46
m. The elliptical shoal has axes of 10.5% - 11.3% of the flume dimensions with its tip at 0.3048 m
from the seabed. The wave period T of the incident waves (Tp for spectral waves) is 1.3 sec and
representative wave height H (Hs for irregular waves) is 2.54 cm. Figure 7 presents satisfactory
comparisons of model results against experimental data of both implementations, in terms of
normalized wave heights H/Ho, where Ho is the offshore wave height. Tests refer to several
experimental cases U3, B3 and N3 (non-breaking spectral waves), and M2 (non-breaking
monochromatic waves) of [41], at various different transects of the shoal. Pearson correlations are
quite high (r > 0.9), with rather low RMSE (ranging from 0.01 to 0.254 for H/Ho), giving a 5‰
minimum to a 12% maximum deviation from experimentally available maximum H/Ho ratio. In
general, the comparisons show a good agreement between the model results and the experimental
for narrow-banded directional spectra. Figure 9 depicts model B results for the normalized wave
height H/Ho and the wave-induced free surface elevation η, reproduced for the said experiment,
referring to unidirectional non-breaking spectral waves over the shoal (test case U3); 2-DH
distribution patterns of Hs and η is plausible compared to their experimental equivalents.

Evaluation of model B’s performance is also conducted by comparisons of simulation results with
experimental data of multi-directional irregular wave diffraction around semi-infinite breakwaters
and through breakwater gaps [42, 89]. The incident significant wave height for the case of multi- and
unidirectional irregular waves is Hs = 0.055-0.0775 and 0.0775 m, respectively; the peak spectral
period is Tp = 1.30 s. Figure 8 presents comparisons of model B results against experimental data in
terms of diffraction coefficient KD for quasi-irregular with relevant spreading parameter s = 19 or ∞
and two different initial angles of propagation. Results correspond to cross-sectional distributions of
KD at a distance Y = 3L from the breakwaters. Comparisons of model results against experimental
data are proven to be satisfactory, excelling especially for the case of unidirectional quasi-spectral

20
wave propagation in the vicinity of port structures. Figure 10 shows plausible representations,
compared to relative experimental data, for the 2-D maps of calculated H/Ho for the case of a
breakwater gap width B = 3.92 m (B/L = 2, L corresponding to peak period for quasi-irregular waves),
under oblique (45°) and transverse (90°) wave attack. Instant depictions of η, after steady state is
reached, are also given to represent the diffraction phenomenon and consequent ripple formation.

3.4 Operational forecast validation

In order to evaluate all models’ performance in operational mode, we performed further validation
of HiReSS, TOMAWAC and WAVE-L forecast simulations against field data in Thessaloniki port (Figure
2).

Figure 11 presents comparisons of model H operational forecast results against in situ observations
of SSH (m) performed by our team during autumn and winter of 2019 in Thessaloniki port (see
Subsection 3.1; Figure 3). The investigated cases refer to both hourly (tide signal) and daily-averaged
(surge signal) data mainly for an entire month (October 2019) and complementarily for a week
during December 2019. Statistical measures were also computed for quantitative validation of
model H runs (WS, RMSE, r, and HRP-index). Acceptable to very high correlation coefficient values
and prediction skill were also achieved for the operational mode of model H, e.g. r = 0.50 - 0.87, WS
= 0.80 - 0.93, HRP-index = 0.9 - 1.0 (Figure 11). The errors in SSH present an upper limit of 7 cm being
lower than 20% of the occurred SSH maxima. The latter are slightly over estimated by the model
under certain weather conditions referring to the well-protected northern Thermaikos bay area. It
can be deduced that the chosen spatial spacing of the computational domain provide efficient
oceanographic predictions locally in coastal areas. We also note model H’s ability to capture the
local astronomical tide pattern in coastal areas, as the reproduced semi-diurnal tide signal is
confirmed by reliable in situ data. The above are corroborated both in autumn and winter time
periods. The weather-driven component of the surge-induced SSH forecasts can also be considered
good, since modelled results remain close to in situ observations even on the daily-averaged signal
(heuristically diminishing the tidal effects on the SSH signal).

Figure 12 presents scatter plot comparisons of integrated model A and B operational forecast results
for irregular and pseudo-spectral significant wave height Hs (m) against field data derived from the
conducted in situ observations with the SBE26 wave gauge in Thessaloniki port (Figure 3).
Comparisons refer to a 3-day forecast period of December 22nd – 24th 2019. The calculated statistical
measures are acceptable for the specific operational forecast case, particularly in very nearshore
(rather shallow) waters, as they read: WS = 0.533, RMSE = 8.6 cm (corresponding to 18.86% of the
maximum Hs,obs for the 3-day forecast), r = 0.317, and HRP-index = 0.94. Based on the latter, the
statistical distributions of percentiles for the modelled values are found to be close to the
corresponding probabilistic properties of the observed Hs time series. For the evaluation of modelled
spectral wave periods, we note that the irregular wave period that corresponds to Hs, i.e. Ts, reaches
an RMSE up to 0.81 sec, corresponding to only 13.83% of the maximum Ts,obs for the 3-day forecast.

21
Therefore, the integrated A and B models’ implementation is found to have a plausible forecast
performance skill, compared to operational wave predictions with 3rd generation spectral models by
other researchers, e.g. [7, 8, 9, 10, 11, 12, 13]. Nevertheless, discrepancies between model results
and field data still persist. The expected moderate performance of the used wave models in
operational mode within coastal waters is highly influenced by two factors: a) the accuracy of input
data [20] at the boundaries, and b) the fact that for small wave height and period (e.g. Hs = 0.3 m
and Ts = 3 s), the deviations of model results relative to measurements are enhanced. This, however,
will not practically affect the quality of our final product, since only larger waves (e.g. over 0.5 m)
would be of interest for vessel pilotage. For these kinds of datasets, Copernicus [20] behaves well;
preliminary comparisons of them against Poseidon [22] wave-buoy data in the central Aegean have
confirmed this during the initial met-ocean input data evaluation phase (not shown for the sake of
brevity).However, the integrated model A and B (i.e. refined resolution model B, forced on the open
boundary with numerical data by model A) simulations of nearshore wave characteristics seem to
improve compared to the case of just applying the coarser model A near and inside ports. Hence, the
integrated model A and B implementation in the vicinity of harbour structures within port areas
seems capable of taking into account the fine and abrupt topographical changes in coastal areas and
better reproduce phenomena as reflections from structures, diffraction and grouping of waves, etc.

The discrepancies between model results and field data are prominent, verifying the practical
problems of operational forecast wave models in the existing literature to perform very well
especially in nearshore areas. The major cause of the latter would probably be the coarse resolution
of both meteorological forcing input and computational mesh, as well as the reliability of available
global- or regional-scale boundary conditions in coastal areas. In addition, the refined resolution
model B seems able to operationally reproduce mid- to shallow-water effects near and inside port
basins.

4. APPLICATION OF INTEGRATED MODELS

Pilot implementations of the integrated sea-state forecast modelling system are presented in the
Mediterranean Sea with a focus on several important ports in it (Figure 2). In this Section we provide
operational sea level and wave modelling output for characteristic Greek ports. The portrayed
results concern maps and graphs of meteorologically and tidally induced SSH, depth-averaged
hydrodynamic circulation and significant wave heights at steady state depictions of the 3-hourly
time intervals for a 3-day forecast. This leads to the production of 24 representations of tide/surge
and wave impacts for each daily implementation of the integrated forecast model.

4.1 Operational mode of model H

The operational function of model H in short-term, sea-level prediction mode has been also tested
by a comparison with another available forecast product from the well-validated operational service

22
of Copernicus platform [20], in the entire Mediterranean Sea. Figure 13 presents model H
operational forecasts of the sea surface elevation due to storm surge (SSH) in the Mediterranean
basin, driven by a very low barometric system (cyclone with a minimum SLP = 990 hPa) over the
Italian peninsula and the Adriatic Sea (Figure 13c); the respective barometric chart is provided [73] is
also presented in Figure 13a. The HiReSS simulations are compared with Copernicus modelled fields
(Figure 13b). A comparable plausible depiction by the two modelling systems is obvious. The low
barometric pressure (“bad” weather) influences the Adriatic and the North Aegean Seas producing
SSH > 30 cm. The two modelling systems’ results agree over the entire Mediterranean Sea [28, 44].

It is also shown that the effects of the atmospheric pressure systems are strongly correlated to the
SSH variability over the entire basin. Model H forecast results of simulated SSH fields and sea current
speed in the entire Mediterranean Sea basin are presented in Figure 14, in conjunction with
atmospheric weather conditions (SLP and winds). Negative surges, i.e. sea surface below MSL, are
apparent in the Adriatic and Aegean Seas, naturally due to the presence of a large barometric high
(“good” weather) and northerly winds, driving waters southward away from the coast. On the
contrary, large values of positive SSH are shown in the Gulf of Gabes and near the Gibraltar straits,
which are influenced by the cyclonic motions of a low barometric system over the northwestern
African coast, characterized by easterly winds. A secondary weaker low-pressure system over the
eastern Mediterranean coast (~1008 hPa; Figure 14a) also increased the sea level below the
cyclone's core (inverse barometer effect) over the northeastern Levantine Sea (¬15 cm; Figure 14b).
Westward and northward currents related to the atmospheric cyclonic wind circulation (Figure 14a)
were also well reproduced by model H over the western Mediterranean basin (2° E - 5° W; Figure
14b) and the eastern Levantine basin (~35° E), respectively, contributing on the storm surge
formation over these two areas. These examples confirm that model H may acceptably reproduce
the effects of the inverse barometer effect in tandem with the wind-driven barotropic circulation on
the sea level variations.

Figure 15 presents characteristic SSH time series produced by model H forecasts in seven ports of
the Mediterranean basin. Sea levels in Greek ports and Genova range from zero (“dead calm” sea
state in Patra port) down to -20 cm, whereas storm-induced sea surface elevation is evident in
Algeciras and Barcelona up to 30 cm. As expected, the SLP time series graph shows reversed
patterns, corroborating the influence of the inverse barometer effect in these areas. Semi-diurnal
undulating configurations are also clearly seen in the surge- and tide-driven SSH timeseries,
revealing characteristic tidal elevation patterns in the Mediterranean Sea. Therefore, model H is
found to successfully incorporate the tidal effects even if astronomic tidal ranges are considered to
be insignificant in the specific region.

Conclusively, model H seems able to provide the necessary automated forecasts of sea levels
(ranging at times from -0.80 m to +1 m) and currents in a characteristic large aquatic body
(Mediterranean Sea) with refined information in diverse coastal areas near and inside ports.

23
4.2 Operational mode of wave models A and B

The results of pilot simulations with wave models A and B model mainly concern charts of mesh and
gridded fields (or a combination of the two) producing maps of wave heights and propagation
direction for spectral waves in coastal areas and quasi-regular waves inside ports. The presented
maps comprise both models A and B domains, with the highest feasible spatial resolution (i.e.
discretization step of Δx = 2 m) in port waters near the harbour structures, and Δx = 25 – 500 m in
deeper waters, e.g. from narrow waterways and CNPs to offshore areas, respectively. Refinement of
resolution is chosen in terms of the computational load, given the available resources and time for
operational sea-state forecasting. Pilot results focus on the three largest commercial transit ports of
Greece.

Figure 16 depicts model A’s forecasts of Hs inside and in the vicinity of the Piraeus port (Greece) for
strong winds blowing from the southern sector. Wave penetration and agitation is minimal on the
interior port facilities; Hs quickly reduces from 2.60 m in approach areas to values lower than 0.50 m
in protected mooring canals and berthing locations. Sheltering effects leeward of insular formations
and breakwaters, together with refraction of spectral waves in port waterways and canals, are
simulated with acceptable accuracy by model A. However, for a detailed wave pattern portrayal by a
phase-resolving model, Figure 17 presents plotted results concerning simulated fields of gridded Hs
data (as narrow-banded spectral wave fields) and relevant free-surface elevation, depicting steady-
state conditions of an extreme case (incident Ho = 2 m) of southern seas in the Thessaloniki port
basin. The protection offered by the sub-aerial breakwater is obvious, as transmitted Hs in the
leeward side of the structure decreases by diffraction, reaching hardly up to 1/4 of the offshore
wave height on the open boundary. Reflection patterns of the free surface elevation for
multidirectional quasi-irregular waves are also visible in the port basin for the case of incident harsh
sea states from the southern-southwestern sector. The resolution is very high in the case of model B
and wave estimation accuracy is further ensured by the application of peripheral sponge layers and
wave reflection techniques that eliminate unphysical reflections on lateral boundaries and
structures; this discrepancy remains for phase-resolving wave models applied in harbour areas [97].

Figure 18 presents the spatial distribution of an extreme case of Hs (>3.5 m) by integrated

simulations with models A and B, inside and in the vicinity of an open-sea port, i.e. the new Patra
port (southern Ionian Sea; Figure 2), for strong winds blowing from the northwestern-northern
sector. The synthesis of results from both wave models is achieved by a technique of densification of
the mesh inside the port ensuring compatibility of the two model’s mesh/grid. This way one can
overall benefit from each model’s distinct advantages and special features. Finally, Figure 19
presents a typical steady-state condition of a 3-hour sea state corresponding to light winds by the
SW-S sector in a physically protected port, i.e. Thessaloniki in the semi-enclosed Thermaikos Gulf at
the northwestern coast of the Aegean Sea (Figure 2). Results were produced by real-time, 3-day,
operational forecasts during the period December 22nd – 24th 2019. The 2-DH spatial distribution of
significant wave height Hs (m) is provided by integrated simulation with models A and B (fed by

24
model H output for sea level). Consequently, the integrated model results are considered to form a
novel complete body of information on sea states and weather conditions able to support pilotage
guidance and, in general, safe approaching procedures of ship vessels to port and harbour basins.

5. CONCLUSIONS

In Accu-Waves project [4, 5] a decision support tool is being developed to provide fine resolution
forecasts on prevailing sea states at selected important ports of the Mediterranean basin (expanded
to worldwide implementation in the future). The application will support approaching procedures of
vessels to ports. In this paper, we present recent developments of a co-operating, high-resolution,
hydrodynamic and wave model suite that derives data as boundary conditions from global-scale
open-sea forecasts. Verification and application of hydrodynamic circulation, spectral and phase-
resolving wave models for offshore, coastal areas and port basins were presented. Model coupling,
nesting and integration are also attempted towards the materialization of a cloud-based operational
forecast platform that will provide wind, wave, sea level and current data for a 3-day forecast at 3-
hourly intervals in and around ports with global commercial interest and high transportation loads.

In the framework of an integrated tool for short-term marine weather and sea-state prognoses in
broader areas around and inside harbours, a robust operational forecast model for storm surges was
built. HiReSS (model H) was validated by comparisons of simulation output against sea level
observations from tide gauges located in ports of the Mediterranean Sea. It satisfactorily simulates
the sea level variations inside harbour areas, also roughly estimating the mean sea currents there.
Case studies in eight Mediterranean ports were presented. Results are also considered as crucial
input (local bathymetric changes) in irregular wave simulations with the integrated TOMAWAC and
WAVE-L models (A and B) sequence.

Furthermore, WAVE-L (model B) was validated by comparisons of model output against

experimental data by classic laboratory physical simulations for both regular and spectral waves. It
adequately simulates the wave propagation in nearshore areas over uneven bottoms and specifically
the wave penetration inside harbours, around and on the leeward side of breakwaters,
incorporating wave-structure interaction and plausible diffraction modelling. The presented case
studies in characteristic Greek ports corroborate that. Conclusively, the integrated models are
proved to satisfactorily simulate the sea surface elevation and wave characteristics in coastal areas
over uneven beds and specifically inside pots, properly incorporating the wave-structure interaction.

Model H results address significant needs of port authorities, ship pilots and navigators towards
battling problems of vessel impact on the harbour bed during mooring, towage and berth
operations, according to high-resolution and short-term sea-state forecasting. The embedded A/B
models’ results address significant needs such as safe spatial and temporal planning of navigation
towards and inside ports, port operations to and from mooring sites, while facilitating the ship-pilot
and port-navigator consultation. This will allow more efficient management of the navigation and
towage services. Indeed, the procedure to certify navigation paths in ports by ESA requires

25
knowledge of operational conditions, including sea state and related environmental data. The safety
issue is underlined also in the e-Navigation strategy by the IMO, where the aim is to analyze and
provide quality marine weather data for limiting the environmentally driven human error in
navigation.

Various tests of the combined hydrodynamic and wave models are ongoing and further pursued to
check the robustness in specific port configurations and tackle a number of communication and
performance problems among the models. Future research includes further site-specific validation
of the operational phase of the models’ performance, by conducting and utilizing further available in
situ measurements in ports [20, 72]. Moreover, certain methodologies should also be investigated in
the future for possible port downtime identification and prediction, due to severe sea-state
conditions, in-port seiching, and overtopping of harbour structures.

The new integrated modelling application seeks to address significant needs such as berth positions
assignment according to auxiliary high-resolution, short-term weather and sea-state forecasts. This
will significantly improve ports’ arsenal towards reducing maritime accidents during these
procedures, and thus assist on reduction of port downtime risk by delays in harbour serviceability
and/or malfunction of port facilities, in the framework of digital support tools’ update for Ports
Safety Management Systems. All the above attest that the project's results constitute a very useful
set of information that combine meteorological forecasts and hydrodynamic simulations of different
scales with management data for port navigation and anchorage procedures.

ACKNOWLEDGEMENTS

This research has been co-financed by the European Union and Greek national funds through the
Operational Program “Competitiveness, Entrepreneurship and Innovation”, under the call
“RESEARCH – CREATE – INNOVATE”; Project Name: ACCU-WAVES [4]; Project Code: T1EDK-05111.
The authors are obliged to MarineTraffic [6] for providing Copernicus [20] and NOAA [72] input data,
and to Assoc. Prof. Y. Pytharoulis from AUTh’s Department of Meteorology and Climatology for
providing regional weather predictions and local weather observations [73]. Initial configuration of
HiReSS sea level forecasts was built within WaveForUs project [21], funded by the national action
“COOPERATION 2011: Partnerships of Production and Research Institutions in Focused Research and
Technology Sectors” in the framework of the operational program “Competitiveness and
Entrepreneurship (NSRF2007-2013)”. The tide-gauge data of sea level variations, used for HiReSS
model validation, were provided by the HNHS [40].

REFERENCES

[1] IMO, Report of the Maritime Safety Committee at its eighty-fifth session, 2008.

26
[2] M. Porretta, D. Jimenez-Banos, M. Crisci, G. Solari, A. Fiumara, A. GNSS evolution for maritime an
incremental approach, Inside GNSS, 2016.

[3] Marine Accident Investigation Branch, Department for Transport, UK, Safety Digest 2005-2009,
Southampton

[4] Accu-Waves. A decision support tool for navigation management in ports. http://accuwaves.eu/
(accessed 12/10/2019)

[5] C. Memos, C. Makris, A. Metallinos, T. Karambas, D. Zissis, M. Chondros, G. Spiliopoulos, M.

Emmanouilidou, A. Papadimitriou, V. Baltikas, Y. Kontos, G. Klonaris, Y. Androulidakis, V. Tsoukala,
Accu-Waves: A decision support tool for navigation safety in ports, 2019.

[6] MarineTraffic, https://www.marinetraffic.com/ (accessed on 28/04/2020)

[7] Rogers W.E.; Kaihatu J.M.; Hsu L.; Jensen R.E.; Dykes J.D.; Holland K.T.; Forecasting and
hindcasting waves with the SWAN model in the Southern California Bight. Coastal Engineering, 54(1),
pp.1-15 (2007). doi:10.1016/j.coastaleng.2006.06.011

[8] Allard, R., Dykes, J., Hsu, Y.L., Kaihatu, J. and Conley, D., 2008. A real-time nearshore wave and
current prediction system. Journal of Marine Systems, 69(1-2), pp.37-58.

[9] Dykes, J.D., Wang, D.W. and Book, J.W., 2009. An evaluation of a high-resolution operational
wave forecasting system in the Adriatic Sea. Journal of Marine Systems, 78, pp.S255-S271.

[10] Singhal, G., Panchang, V.G. and Lillibridge, J.L., 2010. Reliability assessment for operational wave
forecasting system in Prince William Sound, Alaska. Journal of waterway, port, coastal, and ocean
engineering, 136(6), pp.337-349.

[11] Bidlot, J.R., 2012, June. Present status of wave forecasting at ECMWF. In Proceeding from the
ECMWF workshop on ocean waves (pp. 25-27).

[12] Chawla, A., Tolman, H.L., Gerald, V., Spindler, D., Spindler, T., Alves, J.H.G., Cao, D., Hanson, J.L.
and Devaliere, E.M., 2013. A multigrid wave forecasting model: A new paradigm in operational wave
forecasting. Weather and Forecasting, 28(4), pp.1057-1078.

[13] Rusu, L. and Soares, C.G., 2013. Evaluation of a high-resolution wave forecasting system for the
approaches to ports. Ocean Engineering, 58, pp.224-238.

[14] Baldauf, M. and Hong, S.B., 2016. Improving and Assessing the Impact of e-Navigation
applications. International Journal of e-Navigation and Maritime Economy, 4, pp.1-12.

[15] Weintrit, A. and Neumann, T., 2013. Marine Navigation and Safety of Sea Transportation: STCW,
Maritime Education and Training (MET), Human Resources and Crew Manning, Maritime Policy,
Logistics and Economic Matters. CRC Press.

27
[16] IMO (2014). e-navigation Strategy Implementation Plan.

[17] Improved Port Efficiency And Safety Using A Novel Wireless Network And Differential Global
Navigation Satellite System Providing Enhanced Vessel Navigation, http://www.dockingassist.eu/
(accessed 23/04/2020)

[18] PORTS® Physical Oceanographic Real-Time System,

https://tidesandcurrents.noaa.gov/ports.html, (accessed 23/04/2020)

[19] New AVANTI port information software, https://safety4sea.com/new-avanti-port-information-

software/ (accessed 23/04/2020)

[20] COPERNICUS MARINE ENVIRONMENT MONITORING SERVICE, Providing PRODUCTS and

SERVICES for all marine applications, http://marine.copernicus.eu/ (accessed 12/10/2019)

[21] WaveForUs: A Pilot System For The Development And Delivery Of Daily Wave And Circulation
Forecasts For Public And Emergency Use In The Thermaikos Gulf, http://wave4us.web.auth.gr/
(accessed 23/04/2020)

[22] Poseidon System: Monitoring, Forecasting and Information System for the Greek Seas,
http://www.poseidon.hcmr.gr/ (accessed 23/04/2020)

[23] KASSANDRA Storm Surge Modelling System, http://kassandra.ve.ismar.cnr.it:8080/kassandra

(accessed 23/04/2020)

[24] SOCIB – WMOP, http://thredds.socib.es/ (accessed 23/04/2020)

[25] COSYNA - Coastal Observing System for Northern and Arctic Seas,
https://www.hzg.de/institutes_platforms/cosyna/index.php.en (accessed 23/04/2020)

[26] Y.S. Androulidakis, K.D. Kombiadou, C. V. Makris, V.N. Baltikas, Y.N. Krestenitis, Storm surges in
the Mediterranean Sea: Variability and trends under future climatic conditions, Dyn. Atmos. Ocean.
(2015). doi:10.1016/j.dynatmoce.2015.06.001.

[27] Y.N. Krestenitis, Y.S. Androulidakis, Y.N. Kontos, G. Georgakopoulos, Coastal inundation in the
north-eastern mediterranean coastal zone due to storm surge events, J. Coast. Conserv. 15 (2011)
353–368. doi:10.1007/s11852-010-0090-7.

[28] Y. Krestenitis, I. Pytharoulis, T.S. Karacostas, Y. Androulidakis, C. Makris, K. Kombiadou, I.

Tegoulias, V. Baltikas, S. Kotsopoulos, S. Kartsios, Severe Weather Events and Sea Level Variability
Over the Mediterranean Sea: The WaveForUs Operational Platform, in: 2017. doi:10.1007/978-3-
319-35095-0_9.

28
[29] C.V. Makris, Y.S. Androulidakis, Y.N. Krestenitis, K. Kombiadou, V.N. Baltikas, Numerical
Modelling of Storm Surges in the Mediterranean Sea under Climate Change, in: 36th IAHR World
Congr., 2015.

[30] C. Makris, P. Galiatsatou, K. Tolika, C. Anagnostopoulou, K. Kombiadou, P. Prinos, K. Velikou, Z.

Kapelonis, E. Tragou, Y. Androulidakis, G. Athanassoulis, C. Vagenas, I. Tegoulias, V. Baltikas, Y.
Krestenitis, T. Gerostathis, K. Belibassakis, E. Rusu, Climate change effects on the marine
characteristics of the Aegean and Ionian Seas, Ocean Dyn. (2016). doi:10.1007/s10236-016-1008-1.

[31] C. Makris, Y. Androulidakis, V. Baltikas, Y. Kontos, T. Karambas, Y. Krestenitis, HiReSS: Storm

surge simulation model for the operational forecasting of sea level elevation and currents in marine
areas with harbor works, in: Proc. 1st DMPCO Conference, 2019.

[32] Open TELEMAC-MASCARET. The mathematically superior suite of solvers. TOMAWAC - Wave
propagation in coastal areas, http://www.opentelemac.org/index.php/presentation?id=20 (accessed
12/10/2019)

[33] M. Benoit, F. Marcos, F. Becq, Development of a third generation shallow-water wave model
with unstructured spatial meshing, in: Proc. Coast. Eng. Conf., 1996.

[34] M. Benoit, F. Marcos, F. Becq, TOMAWAC: A prediction model for offshore and nearshore storm
waves, in: Proceedings, Congr. Int. Assoc. Hydraul. Res. IAHR, 1997.

[35] J.M. Hervouet, TELEMAC, a hydroinformatic system, Houille Blanche. (1999).

doi:10.1051/lhb/1999029.

[36] R. Issa, D. Rougé, M. Benoit, D. Violeau, A. Joly, Modelling algae transport in coastal areas with a
shallow water equation model including wave effects, J. Hydro-Environment Res. (2010).
doi:10.1016/j.jher.2009.10.004.

[37] T.V. Karambas, S. Christopoulos, I. Avgeris, HARBOUR_L: Integrated mathematical model for the
design of harbor works, in: 5th Pan-Hellenic Conf. Harb. Work., 2010.

[38] T. V. Karambas, A.G. Samaras, An integrated numerical model for the design of coastal
protection structures, J. Mar. Sci. Eng. (2017). doi:10.3390/jmse5040050.

[39] C. Makris, T. Karambas, V. Baltikas, Y. Kontos, A. Metallinos, M. Chondros, A. Papadimitriou, V.

Tsoukala, C. Memos, WAVE-L: An integrated numerical model for wave propagation forecasting in
harbor areas, in: Proc. 1st DMPCO Conference, 2019.

[40] Hellenic Navy Hydrographic Service (HNHS), https://www.hnhs.gr/en/ (accessed 12/10/2019)

[41] C.L. Vincent, M.J. Briggs, Refraction—diffraction of irregular waves over a mound, J. Waterw.
Port, Coast. Ocean Eng. (1989). doi:10.1061/(ASCE)0733-950X(1989)115:2(269).

29
[42] Y.X. Yu, S.X. Liu, Y.S. Li, O.W.H. Wai, Refraction and diffraction of random waves through
breakwater, Ocean Eng. (2000). doi:10.1016/S0029-8018(99)00005-0.

[43] H. de Vries, M. Breton, T. de Mulder, Y. Krestenitis, J. Ozer, R. Proctor, K. Ruddick, J.C. Salomon,
A. Voorrips, A comparison of 2D storm surge models applied to three shallow European seas,
Environ. Softw. (1995). doi:10.1016/0266-9838(95)00003-4.

[44] Y.N. Krestenitis, Y. Androulidakis, K. Kombiadou, C. Makris, V. Baltikas, G. Kalantzi, Operational

Oceanographic Forecasts in the Thermaikos Gulf: The WaveForUs Project, in: Proc. 12th Int. Conf.
Prot. Restor. Environ., 2014.

[45] Υ. Krestenitis, Κ. Kombiadou, Y.S. Androulidakis, C. Makris, V. Baltikas, C. Skoulikaris, Y.N.

Kontos, G. Kalantzi, Operational Oceanographic Platform in Thermaikos Gulf (Greece): Forecasting
and Emergency Alert System for Public Use, in: Proc. 36th IAHR World Congress, 2015.

[46] Krestenitis Y, Androulidakis Y, Kombiadou K, Makris C, Baltikas V, Operational Forecast System

of Storm Tides in the Aegean Sea (Greece), in: Proc. 2015 ASLO Aquat. Sci. Meet., 2015.

[47] X.H. Wang, Tide-induced sediment resuspension and the bottom boundary layer in an idealized
estuary with a muddy bed, J. Phys. Oceanogr. (2002). doi:10.1175/1520-
0485(2002)032<3113:TISRAT>2.0.CO;2.

[48] S.D. Smith, E.G. Banke, Variation of the sea surface drag coefficient with wind speed, Q. J. R.
Meteorol. Soc. (1975). doi:10.1002/qj.49710142920.

[49] P.J. Roache, Computational fluid dynamics, Hermosa, Albuquerque, 1972.

[50] E.W. Schwiderski, On charting global ocean tides, Rev. Geophys. (1980).
doi:10.1029/RG018i001p00243.

[51] K. Sakamoto, H. Tsujino, H. Nakano, M. Hirabara, G. Yamanaka, A practical scheme to introduce

explicit tidal forcing into an OGCM, Ocean Sci. (2013). doi:10.5194/os-9-1089-2013.

[52] K. Matsumoto, T. Takanezawa, M. Ooe, Ocean tide models developed by assimilating

TOPEX/POSEIDON altimeter data into hydrodynamical model: A global model and a regional model
around Japan, J. Oceanogr. (2000). doi:10.1023/A:1011157212596.

[53] P. Schureman, Manual of harmonic analysis and prediction of tides, 1958.

doi:10.5962/bhl.title.38116.

[54] C.C. Mei, The applied dynamics of ocean surface waves., (1983).

[55] H.L. Tolman, A Third-Generation Model for Wind Waves on Slowly Varying, Unsteady, and
Inhomogeneous Depths and Currents, J. Phys. Oceanogr. (1991). doi:10.1175/1520-
0485(1991)021<0782:atgmfw>2.0.co;2.

30
[56] G.J. Komen, L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann, P.A.E.M. Janssen, Dynamics
and Modelling of Ocean Waves, Cambridge University Press, Cambridge, 1994.
doi:10.1017/CBO9780511628955.

[57] TOMAWAC User Manual, http://wiki.opentelemac.org/doku.php?id=user_manual_tomawac

(accessed 12/10/2019)

[58] K. Hasselmann, S. Hasselmann, E. Bauer, P.A.E.M. Janssen, G.J. Komen, L. Bertotti, P. Lionello, A.
Guillaume, V.C. Cardone, J.A. Greenwood, M. Reistad, L. Zambresky, J.A. Ewing, The WAM model - a
third generation ocean wave prediction model., J. Phys. Ocean. (1988).

[59] G.J.M. Copeland, A practical alternative to the “mild-slope” wave equation, Coast. Eng. (1985).
doi:10.1016/0378-3839(85)90002-X.

[60] A. Watanabe, K. Maruyama, Numerical Modeling of Nearshore Wave Field Under Combined
Refraction, Diffraction and Breaking, Coast. Eng. Japan 29. (1986) 19–39.
doi:10.1080/05785634.1986.11924425.

[61] S. Christopoulos, I. Avgeris, T. V. Karambas, HMAR-HARBOURL: An integrated numerical model

for harbour layout design, in: Proc. Int. Offshore Polar Eng. Conf., 2012.

[62] J.A. Battjes, J.P.F.M. Janssen, Energy Loss and Set-Up Due To Breaking of Random Waves., Proc.
Coast. Eng. Conf. 1 (1979) 569–587.

[63] W. Rattanapitikon, R. Karunchintadit, T. Shibayama, Irregular wave height transformation using

representative wave approach, Coast. Eng. J. (2003). doi:10.1142/S0578563403000865.

[64] J. Larsen, H. Dancy, Open Boundaries in Short Wave Simulations - A New Approach,
Regelungstechnik RT. (1983). doi:10.1016/0378-3839(83)90022-4

[65] C. Lee, K.D. Suh, Internal Generation of Waves for Time-Dependent Mild-Slope Equations, Coast.
Eng. (1998). doi:10.1016/S0378-3839(98)00012-X

[66] T.V. Karambas, B.C. Bowers, Representation Of Partial Wave Reflection And Transmission For
Rubble Mound Coastal Structures, WIT Trans. Ecol. Environ. 18 (1996). doi:10.2495/HY960421.

[67] P. Bruun, Design and Construction of Mounds for Breakwaters and Coastal Protection; Elsevier:
Amsterdam, The Netherland, 1985.

[68] C.G. Koutitas, Numerical solution of the complete equations for nearly horizontal flows, Adv.
Water Resour. 1 (1978) 213–217. doi:10.1016/0309-1708(78)90005-2.

[69] T.V. Karambas, C.D. Memos, Boussinesq model for weakly nonlinear fully dispersive water
waves. J. Waterw. Port Coast. Ocean Eng., 2009. doi:10.1061/(ASCE)0733-950X(2009)135:5(187)

31
[70] MEDITERRANEAN SEA PHYSICS ANALYSIS AND FORECAST, Metadata provided by CMEMS, E.U.
Copernicus Marine Service Information,
https://doi.org/10.25423/CMCC/MEDSEA_ANALYSIS_FORECAST_PHY_006_013_EAS5 (accessed on
28/04/2020)

[71] Clementi, E., Pistoia, J., Escudier, R., Delrosso, D., Drudi, M., Grandi, A.,Lecci R., Cretí S., Ciliberti
S., Coppini G., Masina S., Pinardi, N., 2019. Mediterranean Sea Analysis and Forecast (CMEMS MED-
Currents, EAS5 system) [Data set]. Copernicus Monitoring Environment Marine Service (CMEMS).

[72] National Oceanic and Atmospheric Administration, NATIONAL CENTERS FOR ENVIRONMENTAL
INFORMATION, Model Data, Global Forecast System (GFS), https://www.ncdc.noaa.gov/data-
access/model-data/model-datasets/global-forcast-system-gfs (accessed 12/10/2019)

[73] AUTh, Department of Meteorology and Climatology, Meteorological Observations online

service, https://meteo.geo.auth.gr/en/meteo-obs (accessed 23/04/2020)

[74] Tegoulias, I., Kartsios, S., Pytharoulis, I., Kotsopoulos, S., & Karacostas, T. S., 2017. The Influence
of WRF Parameterisation Schemes on High Resolution Simulations Over Greece. In Perspectives on
Atmospheric Sciences. Springer, pp. 3-8.

[75] Pytharoulis, I., Tegoulias, I., Kotsopoulos, S., Bampzelis, D., Kartsios, S., Zanis, P. et al., 2014.
High-resolution WRF hindcasts over central Greece: Characteristics of simulated convective activity
and model evaluation. In Proc.: 15th Annual WRF Users' Workshop, Boulder, CO, pp. 23-27.

[76] Pytharoulis, I., Karacostas, T., Tegoulias, I., Kotsopoulos, S., Bampzelis, D., 2015. Predictability of
Intense Weather Events Over Northern Greece. In Proc.: 95th American Meteorological Society
Annual Meeting. Phoenix, AZ, USA, 895.

[77] MEDITERRANEAN SEA WAVES ANALYSIS AND FORECAST, Metadata provided by CMEMS, E.U.
Copernicus Marine Service Information,
https://doi.org/10.25423/CMCC/MEDSEA_ANALYSIS_FORECAST_WAV_006_017 (accessed on
28/04/2020)

[78] Korres, G., Ravdas, M., & Zacharioudaki, A., 2019. Mediterranean Sea Waves Analysis and
Forecast (CMEMS MED-Waves) [Data set]. Copernicus Monitoring Environment Marine Service
(CMEMS). https://doi.org/10.25423/CMCC/MEDSEA_ANALYSIS_FORECAST_WAV_006_017

[79] Navionics, https://webapp.navionics.com/ (accessed 12/10/2019)

[80] General Bathymetric Chart of the Oceans (GEBCO), www.gebco.net (accessed 12/10/2019)

[81] Y. Goda, Y. Suzuki, Computation of refraction and diffraction of sea waves with Mitsuyasu’s
directional spectrum, 1975.

32
[82] Sea-Bird Scientific, CTDs, Moored, SBE 26plus Seagauge Wave & Tide Recorder
https://www.seabird.com/moored/sbe-26plus-seagauge-wave-tide-recorder-
discontinued/family?productCategoryId=54627473779 (accessed 23/04/2020)

[83] Sea-Bird Scientific, Software, https://www.seabird.com/software (accessed 23/04/2020)

[84] Sea-bird Scientific, SBE 26plus Seagauge Wave & Tide Recorder, User Manual, 2015

[85] Global Sea Level Observing System (GLOSS), https://www.gloss-sealevel.org/fast-mode-data-

delivery (accessed 12/10/2019)

[86] Italian National Institute for Environmental Protection and Research (ISPRA)
http://www.isprambiente.gov.it/pre_mare/coastal_system/maps/mc_wiz.html (accessed
12/10/2019)

[87] D. Conte, P. Lionello, Characteristics of large positive and negative surges in the Mediterranean
Sea and their attenuation in future climate scenarios, Glob. Planet. Change. (2013).
doi:10.1016/j.gloplacha.2013.09.006.

[88] R. Pawlowicz, B. Beardsley, S. Lentz, Classical tidal harmonic analysis including error estimates in
MATLAB using T TIDE, Comput Geosci (2002). doi: 10.1016/S0098-3004(02)00013-4

[89] Y.S. Li, S.X. Liu, Y.X. Yu, G.Z. Lai, Numerical modelling of multi-directional irregular waves
through breakwaters, Appl. Math. Model. (2000). doi:10.1016/S0307-904X(00)00003-2.

[90] C.J. Willmott, S.M. Robeson, K. Matsuura, A refined index of model performance. Int. J. Climatol.
(2012). doi:10.1002/joc.2419

[91] P. Galiatsatou, C. Makris, P. Prinos, D. Kokkinos, Nonstationary joint probability analysis of

extreme marine variables to assess design water levels at the shoreline in a changing climate, Nat.
Hazards. (2019). doi:10.1007/s11069-019-03645-w.

[92] Y. Galiatsatou, P.; Makris, C.; Kokkinos, D.; Prinos, P.; Krestenitis, Climate change effects on
extreme total water levels of the Greek coastal zone, in: 1st Int. Conf. Des. Manag. PORT, Coast.
OFFSHORE Work., 2019.

[93] C. Makris, P. Galiatsatou, Y. Androulidakis, K. Kombiadou, V. Baltikas, Y. Krestenitis, P. Prinos,

Climate change impacts on the coastal sea level extremes of the east-central Mediterranean Sea, in:
XIV PRE Conference, 2018.

[94] R. Schoetter, P. Hoffmann, D. Rechid, K.H. Schlünzen, Evaluation and bias correction of regional
climate model results using model evaluation measures, J. Appl. Meteorol. Climatol. (2012).
doi:10.1175/JAMC-D-11-0161.1.

[95] T. Awk, TOMAWAC v7p3 Validation Manual, 2018.

33
[96] J.C.W. Berkhoff, N. Booy, A.C. Radder, Verification of numerical wave propagation models for
simple harmonic linear water waves, Coast. Eng. (1982). doi:10.1016/0378-3839(82)90022-9.

[97] V. Karambas, T.; Makris, C.; Baltikas, 2-DH Post-Boussinesq Modeling of Nonlinear Wave
Propagation and Transformation in Nearshore Zones and Inside Ports, Proc. Coast. Struct. 2019 Conf.
(2019).

TABLES

Table 1. Values of parameters and constants of the major tidal modes (harmonics of the eigenmodes
of tidal oscillations) in model H.

Table 2. Basic parameterizations and main attributes of the wave and storm surge models (A, B and
H).

Table 3. Validation of model H long-term “climatic” hindcast results against in situ observations of
SSH maxima (m) by tide gauges during 1995-2005 in 4 Greek ports. Statistical errors are provided for
quantitative validation together with collective values for the Willmott Sill score WS, RMSE, Pearson
correlation coefficient r and HRP-index.

FIGURES

Figure 1. Workflow diagram between high-resolution numerical models H, A and B forced by

global/regional scale meteorological and oceanographic databases, and producing output for a Port
Safety Management System. SLP: sea level pressure; Wx, Wy: wind zonal and meridional
components; SSH: sea surface height; Hs: significant wave height; Tp: peak spectral period; as: mean
spectral wave direction; H: regular wave height; T: monochromatic wave period; a: regular wave
propagation direction.

Figure 2. Bathymetry chart of the Mediterranean Sea (including Black Sea; excluding Atlantic Ocean)
referring to the computational domain (Level I) of model H from the GEBCO database with the
locations of eight selected ports (black dots). Additional Mediterranean stations used for evaluation
reasons are marked with a cross.

Figure 3. Depiction of a) the SBE26 measuring instrument and its base-holder for field recordings of
tides, sea levels, and wave features, b) the locations of in situ observations performed in the
Thessaloniki port basin: approximately 22° 54.621’ longitude, 40° 37.888’ latitude.

34
Figure 4. Comparisons of model H hindcast results against in situ observations of SSH (m) by tide
gauges, during 2012 in Thessaloniki, Alexandroupolis, Genova and Venice ports [40, 85]. The
following statistical measures are provided for quantitative validation; WS: Willmott Sill score; RMSE:
root-mean-square error; r: Pearson correlation.

Figure 5. Comparisons of model H hindcast results against in situ observations of SSH (m) by tide
gauges [86], during 2013-2015 in Trieste and Venice ports. The following statistical measures are
provided for quantitative validation; WS: Willmott Sill score; RMSE: root-mean-square error; r:
Pearson correlation.

Figure 6. Comparisons of model H long-term hindcasting model results against in situ observations
by tide gauges during 1995-2005 in 4 Greek ports [40]. From top to bottom graph, the storm-
induced sea level maxima are represented by the statistical measures: a) SSHmax, b) SSI, c) 99th
percentile of SSH (m), and d) probability of occurrence of locally observed SSHmax.

Figure 7. Comparisons of model B results against experimental data [41], in terms of normalized
wave height H/Ho, for spectral waves propagating over a circular shoal corresponding to
experimental test cases a) U3, b) B3, c) M2 and d) N3 in [41].

Figure 8. Comparisons of model B results against experimental data [42, 89], in terms of diffraction
coefficient KD for experimental data of: a) unidirectional quasi-irregular waves with spreading
parameter s=∞ and inimal angle of propagamon θο=90°, b) multi-directional quasi-irregular waves
with s=19 and θο=45°, and c) unidirectional quasi-irregular waves with spreading parameter s=∞ and
initial angle of propagation θο=45°.

Figure 9. Depiction of model B results for: a) normalized wave height H/Ho and b) wave-induced free
surface elevation η, reproducing experimental data for unidirectional non-breaking spectral waves
propagating over a circular shoal corresponding to experimental test case U3 in [41].

Figure 10. Depiction of model B results for: a) the wave-induced free surface elevation η and b) the
normalized wave height H/Ho, reproducing experimental data [42, 89] for uni-directional irregular
waves propagating and diffracted through a breakwater gap corresponding to a spreading
parameter s=∞ and inimal angle of propagamon θο=45° (upper graphs) and θο=90° (lower graphs).

Figure 11. Comparisons of model H operational forecast results (mod) against in situ observations
(obs) of SSH (m) performed by our team during Autumn and Winter of 2019 in Thessaloniki port (see
Figure 3). Cases: a) hourly data and b) daily-averaged data for an entire month (October 2019); c)
hourly data and d) daily-averaged data for an entire week (during December 2019). The following
statistical measures are also provided in a table for quantitative validation; WS: Willmott Sill score;
RMSE: root-mean-square error; r: Pearson correlation coefficient; HRP-index: Hit-Rate-of-Percentiles
index.

35
Figure 12. Scatter plot comparisons of operational forecast results (mod) of significant wave height
Hs (m) by a) model A and b) model B (forced on the boundary by model A output), against field data
(obs) derived from the conducted in situ observations with the SBE26 wave gauge in Thessaloniki
port (Figure 3). Comparisons refer to a 3-day period of December 22nd – 24th 2019. The following
statistical measures are also provided in a table for quantitative validation; WS: Willmott Sill score;
RMSE: root-mean-square error; r: Pearson correlation coefficient; HRP-index: Hit-Rate-of-Percentiles
index.

Figure 13. Comparisons of operational forecasts for the spatial distribution of the free surface
elevation due to storm surge SSH (m) in the Mediterranean Sea: a) Real weather barometric chart
map created with data provided by [73]; b) Storm surge-induced sea level SSH (m) forecast by the
Copernicus (ex MyOcean) platform [20]; c) Storm surge-induced sea level SSH (m) forecast with
model H (HiReSS model used in [4] and [21]). Time of implementation: February 1st 2015, UTC 11:00.

Figure 14. a) Atmospheric weather conditions chart over southern Europe, where contours and color
bar refer to SLP (hPa) and vectors to wind field speed (m/sec); b) Model H forecast results chart of
the entire Mediterranean Sea basin, where contours and color bar refer to SSH (m) and vectors to
sea current velocities (m/sec). Time of implementation: April 21st 2019, UTC 00:00.

Figure 15. Graphs of a) model H results about surge- and tide-induced SSH (m) and b) SLP (hPa) in
seven characteristic ports of the Mediterranean basin (Algeciras, Barcelona, Genova, Haifa, Patra,
Piraeus, Thessaloniki ports; Figure 2). Forecasts refer to a 3-day period of April 19th - 22nd 2019.

Figure 16. Depiction of spatial distribution of significant wave height Hs (m) in operational forecast
mode by simulation with model A inside and in the vicinity of the Piraeus port (Greece, west-central
coast of the Aegean Sea; Figure 2) for strong winds by the South sector (lower graph depicts a zoom-
in version).

Figure 17. Results charts in the Thessaloniki Port basin (Greece, northwestern coast of the Aegean
Sea; Figure 2) for a) wave-induced free surface elevation, and b) significant wave height Hs (m), in
operational forecast mode by simulation with model B. Multi-directional, quasi-irregular high wave
fields are shown approaching from the Southern-Southwestern sector referring to a steady-state
representation of a selected 3-hourly sea state from a high seas period in May 2019; Peripheral
sponge layer and eliminated reflections on lateral boundaries are also shown.

Figure 18. Operational forecast results for spatial distribution of significant wave height Hs (m) by
integrated simulation with models A and B inside and in the vicinity of the new Patra port (Greece,
southeastern coast of the Ionian Sea; Figure 2) for strong winds by the NW-W sector.

Figure 19. Operational forecast results for the 2-DH spatial distribution of significant wave height Hs
(m) by integrated simulation with models A and B in the Thessaloniki port (Greece, northwestern
coast of the Aegean Sea; Figure 2). Implementation refers to real-time operational forecasts for a 3-
day period of December 22nd – 24th 2019; light winds by the SW-S sector.
36
 TABLES

Table 1. Values of parameters and constants of the major tidal modes (harmonics of the eigenmodes of
tidal oscillations) in model H.

Symbol of
Tidal Oscillation Type of
Harmonic ko (m) σt (10-4/sec) Τ (hrs) χ (°)
Eigen-mode Tide
Constituent
M2 Principal lunar 0.242334 1.10519 12.42 2ho-2so
S2 Principal solar 0.112841 1.45444 12.00 0 Semi-diurnal
N2 Elliptical lunar 0.046398 1.37880 12.66 2ho-3so+po ν=2
K2 Declination luni-solar 0.030714 1.45842 11.97 2ho
K1 Declination luni-solar 0.141565 0.72921 23.93 ho+90
O1 Principal lunar 0.100514 0.67598 25.82 ho-2so-90 Diurnal
P1 Principal solar 0.046843 0.72523 24.07 ho-90 ν=1
Q1 Elliptical lunar 0.019256 0.64959 26.87 ho-3so+po-90
Mf Fortnightly lunar 0.041742 0.053234 327.86 2so
Long Period
Mm Monthly lunar 0.022026 0.026392 661.30 so-po
ν=0
Ssa Semiannual solar 0.019446 0.003982 4382.89 2ho
Table 2. Basic parameterizations and main attributes of the wave and storm surge models (A, B and H).

Attribute / Parameterization Implementation characteristics

Shallow water equation for
Hydrodynamic model HiReSS (Model H)
barotropic circulation and storm surge
Regional Sea Mediterranean basin
Forcing / Driving field NOAA and AUTh [72, 73] SLP, Wx, Wy products, 0.1°×0.1°, 3-hour
Spatial resolution / Domain 1/20° × 1/20° Enclosed basin
Bathymetry GEBCO [80] Staggered grid Arakawa-C type
Simulation time span 3 days (72 hrs) with 3-hrs output
Integration time step / Output 30 sec 3-hour
Free-surface / Bottom friction Smith and Banke [48] Wang [47]
Boundary / Nesting technique Dirichlet / Open sea boundary with estimated free surface
Eddy viscosity treatment Boussinesq hypothesis: Smagorinsky model for horizontal eddies
Tidal component approach Static model by Schwiderski [50]
Wave model TOMAWAC (Model A) 3rd generation phase-averaged spectral
Thermaikos, Patraikos, Saronikos Gulfs, Gibraltar & Haifa Bay, Gibraltar Strait,
Gulfs, Bays, Local Seas
Iberian & Ligurian Sea
Forcing / Driving field NOAA and AUTh [72, 73, 74] Wx, Wy products, 0.1°×0.1°, 3-hour
Initial / Boundary Copernicus MEDSEA ANALYSIS FORECAST
CMEMS [20]
Conditions WAV-006-017 products, 0.042°×0.042°, 1-hour
Spatial resolution / Domain Varying: 50 – 500 m Semi-enclosed gulfs
Bathymetry Hydrographic Services [40, 79] Finite Elements [57]
Frequency range 0.04 – 1 Hz 0.056 – 1 Hz
Integration time step / Output 10 min 3-hour
Integration scheme MoC for propagation – Semi-Implicit for source terms [58]
Hyperbolic mild-slope equation phase-
Wave model WAVE-L (Model B)
resolving
Ports Algeciras, Barcelona, Genova, Haifa, Patra, Piraeus, Tanger Med, Thessaloniki
Driving field Model A results Hs, Tp, ap
Spatial resolution / Domain Fixed: 2 m Port approaches and harbour basins
Bathymetry Hydrographic Services [40, 79] Staggered orthogonal grid [68, 69]
Integration time step / Output 0.1 sec 3-hour
Wave breaking model Eddy viscosity Battjes and Janssen [62]
Boundary conditions Sponge layer Partial/Full reflection from structures [66]
Wave generation Quasi-irregular waves Lee and Suh [65]
Table 3. Validation of model H long-term “climatic” hindcast results against in situ observations of SSH
maxima (m) by tide gauges during 1995-2005 in 4 Greek ports. Statistical errors are provided for
quantitative validation together with collective values for the Willmott Sill score WS, RMSE, Pearson
correlation coefficient r and HRP-index.

Port Site Mean Ei (%) Mean EIi HRP-index

Thessaloniki -6.29% -0.206 0.98

Chios 8.28% 0.281 0.94

Alexandroupolis -17.20% -0.561 0.98

Lefkada -4.17% -0.133 0.95

AVERAGE -0.05 -0.15 0.96

Overall skill factors at 4 Greek ports

Pearson Correlation, r 0.841

Root-mean-square error, RMSE (m) 0.026

Willmott Skill score, WS 0.712