NHESSNatural Hazards and Earth System SciencesNHESSNat. Hazards Earth Syst. Sci.1684-9981Copernicus PublicationsGöttingen, Germany10.5194/nhess-16-1499-2016High-resolution wave and hydrodynamics modelling in coastal areas:
operational applications for coastal planning, decision support and
assessmentSamarasAchilleas G.achilleas.samaras@unibo.ithttps://orcid.org/0000-0002-0201-3695GaetaMaria GabriellaMiquelAdrià MorenoArchettiRenatahttps://orcid.org/0000-0003-2331-6342CIRI – EC, Fluid Dynamics Unit, University of Bologna, Via del Lazzaretto 15/5, Bologna, 40131, ItalyDepartment of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Viale Risorgimento 2, Bologna, 40136, ItalyAchilleas G. Samaras (achilleas.samaras@unibo.it)1July20161661499151824February20164March201620May201623May2016This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://nhess.copernicus.org/articles/16/1499/2016/nhess-16-1499-2016.htmlThe full text article is available as a PDF file from https://nhess.copernicus.org/articles/16/1499/2016/nhess-16-1499-2016.pdf
Numerical modelling has become an essential component of today's coastal
planning, decision support and risk assessment. High-resolution modelling
offers an extensive range of capabilities regarding simulated conditions,
works and practices and provides with a wide array of data regarding
nearshore wave dynamics and hydrodynamics.
In the present work, the open-source TELEMAC suite and the commercial
software MIKE21 are applied to selected coastal areas of South Italy.
Applications follow a scenario-based approach in order to study
representative wave conditions in the coastal field; the models' results are
intercompared in order to test both their performance and capabilities and
are further evaluated on the basis of their operational use for coastal
planning and design. A multiparametric approach for the rapid assessment of
wave conditions in coastal areas is also presented and implemented in areas
of the same region. The overall approach is deemed to provide useful insights
on the tested models and the use of numerical models – in general – in the
above context, especially considering that the design of harbours, coastal
protection works and management practices in the coastal zone is based on
scenario-based approaches as well.
Introduction
Accurate predictions of waves, currents and sea level
variations in coastal areas are essential for a wide range of research and
operational applications, as they govern inundation, sediment and pollutant
transport, coastal morphology evolution and interactions with structures.
Accordingly, numerical models that can serve the above purposes have become
the main tool for researchers, engineers and policymakers around the world
involved in coastal planning, risk management and monitoring activities.
Following the above considerations, the development of reliable modelling
systems or methods that can scale down from the ocean to the coastal scale
has emerged as a need in today's research. Reliable information on the
hydrodynamics of the zone defined as nearshore, in particular, can serve a
key role in coastal planning and hazard mitigation, as relevant processes at
that scale differ significantly from those described in larger-scale
oceanographic models. It is self-evident that, in the above context, the
capabilities and limitations of such systems and methods – apart from their
structure – would depend on those of the numerical models they comprise.
A series of model coupling and nesting techniques, as well as entire
methodological frameworks, has been proposed and applied in various research attempts for the
development of modelling systems with the aforementioned characteristics.
Among the early works on the subject, one can indicatively refer to Ozer et
al. (2000), who proposed a coupling module for tides, surges and waves,
applying it, however, to relatively low-resolution simulations for the North
Sea. Regarding more recent and complete attempts, one can refer to the work
of Warner et al. (2010), who developed the Coupled
Ocean–Atmosphere–Wave–Sediment Transport system (COAWST); Ge et
al. (2013), who developed the FVCOM system to simulate multi-scale dynamics
at the East China Sea shelf and the Changjiang Estuary; and Barnard et
al. (2014), who developed a modelling system for predicting storm impact on
high-energy coasts (CoSMoS).
In contrast, integrated systems comprising atmosphere, ocean and coastal
models do present a number of challenges for their users regarding both data
interoperability and downscaling/nesting techniques, while they also demand
significant computational expense in order to arrive to high-resolution
simulations near coasts. Furthermore, for a series of activities in
coastal/marine planning (e.g. identification of wave energy sites, see
Reikard, 2009; Bozzi et al., 2014), vulnerability/risk assessment (e.g.
Stockdon et al., 2012; Idier et al., 2013; Archetti et al., 2016) and coastal
protection measures/infrastructure design (e.g. van Duin et al., 2004;
Burcharth et al., 2014; Karambas, 2014; Karambas and Samaras, 2014), either
only parts of local hydrodynamics information are required (mainly wave
properties to drive nearshore models) or the respective approaches are based
on the study of frequent/extreme condition scenarios. Accordingly, a number
of methods have been developed in order to estimate coastal wave properties
from offshore information or larger-scale simulations. One can refer to the
early work of O'Reilly and Guza (1993), who proposed wave energy
transformation coefficients based on the comparison of two spectral wave
models' results or more recent ones using nesting and data assimilation
schemes (Bertotti and Cavaleri, 2012; Rusu and Soares, 2014) and
machine-learning techniques (Camus et al., 2011; Plant and Holland,
2011b, a). A work that stands out in recent literature is that of Long et
al. (2014), who proposed a probabilistic method based on model scenarios for
constructing wave time series at inshore locations.
The present work follows the rationale described above, comparing two
modelling suites in the representation of nearshore dynamics and proposing a
multiparametric scenario-based approach for the rapid assessment of wave
conditions in coastal zones. Nevertheless, this work also served as the
background study for the development of a modelling system coupling
atmosphere, ocean and coastal dynamics, as described in Gaeta et al. (2016).
Wave modelling within the TELEMAC and MIKE21 suites is performed using
TOMAWAC and MIKE21-SW respectively. TOMAWAC and MIKE21-SW are characterized
as third-generation spectral wave models, as they do not require any
parameterization on either the spectral or the directional distribution of
power (or action density). The physical processes modelled comprise
(a) energy source/dissipation processes (wind-driven interactions with
atmosphere, dissipation through wave breaking/whitecapping/wave-blocking due
to strong opposing currents, bottom friction-induced dissipation);
(b) non-linear energy transfer conservative processes (resonant quadruplet
interactions, triad interactions); and (c) wave propagation-related processes
(wave propagation due to the wave group/current velocity,
depth-/current-induced refraction, shoaling, interactions with unsteady
currents). The models compute the evolution of wave action density N by
solving the action balance equation (Booij et al., 1999):
∂N∂t+∇x,y[cg+UN]+∂∂σcσN+∂∂θcθN=Stotσ,
where N=Eσ, E being the variance density and
σ the relative angular frequency, cg is the
intrinsic group velocity vector, U is the ambient current,
cσ, cθ are the propagation velocities in spectral space
(σ, θ) and Stot is the source/sink term that
represents all physical processes which generate, dissipate or redistribute
energy. Broken down to its components, Stot can be written as
Stot=Sin+Swc+Snl4+Sbf+Sbr+Snl3,
where Sin represents the energy transfer from wind to waves,
Swc the dissipation of energy due to whitecapping, Snl4
the nonlinear transfer of energy due to quadruplet (four-wave) interactions,
Sbf the dissipation due to bottom friction, Sbr the
dissipation due to wave breaking and Snl3 the nonlinear transfer of
energy due to triad (three-wave) interactions. TOMAWAC and MIKE-SW
parameterize similarly the above processes; TOMAWAC, however, does offer more
options regarding the available approaches/models to be used for most of
them. Therefore, and regarding the processes of interest for the model
intercomparison as presented in Sect. 3.1.3, the respective common
approaches/models applied in this work are (i) the Battjes and Janssen (1978)
model for bathymetric breaking; (ii) the model of Hasselmann et al. (1973)
for bottom friction dissipation using a constant friction coefficient;
(iii) the Komen et al. (1984) and Janssen (1991) dissipation model for
whitecapping; and (iv) the LTA (Lumped Triad Approximation) model of
Eldeberky and Battjes (1983) for triad interactions (the SPB model of Becq
(1998) – available only in TOMAWAC – is also tested). As for diffraction,
its effect is simulated using the phase-decoupled approach proposed by
Holthuijsen et al. (2003), based on the revised version of the Mild Slope
Equation model of Berkhoff (1972) proposed by Porter (2003). Both models
solve the governing equation by means of finite element-type methods to
discretize geographical and spectral space, while the geographical domain is
discretized by unstructured triangular meshes. Finally, and regarding
specifically the coupling with 2-D hydrodynamics, it should be noted that the
models compute and provide as output the four components of the radiation
stress tensor, Sxx, Syy, Sxy and Syx, evaluated by
Sxx=∫∫E2[2ncosθ2+2n-1]dσdθ,Syy=∫∫E2[2nsinθ2+2n-1]dσdθ,Sxy=Syx=∫∫Ensinθcosθdσdθ,
as well as the respective wave-induced forces along the x and y axes
(i.e. Fx and Fy), evaluated by integrating the radiation stresses over
the water depth.
Hydrodynamics modelling within the TELEMAC and MIKE21 suites is performed
using TELEMAC-2D and MIKE21-HD respectively. The models solve the 2-D
shallow water equations (also referred to as Saint-Venant equations; see
Hervouet, 2007), derived by integrating the Reynolds-averaged Navier–Stokes
equations over the flow depth. Adopting the formulation of TELEMAC-2D for
Cartesian coordinates, the equations of continuity and momentum along the x
and y axes can be written as Eqs. (6), (7) and (8) respectively:
∂h∂t+u⋅∇h+hdivu=Sh,∂u∂t+u⋅∇u=-g∂ζ∂x+Sx+1hdivhvt∇u,∂v∂t+u⋅∇v=-g∂ζ∂y+Sy+1hdivhvt∇v,
where h is the water depth, u, v are the velocity components and
u the velocity vector, g is the gravitational acceleration, vt
is the momentum diffusion coefficient, ζ is the free surface elevation,
Sh is a term representing sources/sinks of fluid and Sx,
Sy are terms representing sources/sinks of momentum within the domain
(i.e. wind, Coriolis force, bottom friction, radiation stresses/forces from
wave models). These primitive equations are solved by means of finite
element/volume methods, while the geographical domain is discretized by
unstructured triangular meshes. As also mentioned previously for TOMAWAC
and MIKE-SW, and, although TELEMAC-2D and MIKE21-HD have a lot of
similarities, TELEMAC-2D does offer more parameterization options regarding
the definition of physical and numerical parameters. In the present work, the
use of the hydrodynamics models is focused on the representation of
wave-generated currents a task achieved through their direct coupling –
through radiation stresses – with the respective spectral wave models within
the TELEMAC and MIKE21 suites (see Eqs. (3)–(8) in the previous).
Multiparametric approach for the rapid assessment of nearshore wave
conditions
The methodology followed in the present work for the rapid assessment of
nearshore wave conditions (within the framework set in the previous; see
Sect. 1) comprises a number of steps aiming to establish an efficient and
computationally reasonable approach for operational use. The approach is
scenario-based; thus its first step consists in defining a number of scenarios representing wave conditions
in the wider area of interest. This is done by performing a spectral analysis
of sea surface elevation records from nearshore/offshore buoys in order to
produce a data set of three aggregated wave parameters, namely the
significant wave height Hs, the peak period Tp and
the mean wave direction Dirm. Next, data set parameters are
further divided into a number of classes each, forming by aggregation the
sets of Hs-Tp-Dirm, henceforth
referred to as “scenarios”. These scenarios are afterwards used (in
sequence) as boundary conditions for the wave model runs, resulting in an
extensive data set of model results for the entire computational domain,
stored in ASCII files that are properly named on the basis of the input wave
scenarios. These files form the high-resolution wave conditions database
along with a query algorithm, serving as the “bridge” between
coarser-resolution operational models and the aforementioned produced data
set. The query algorithm is responsible for (a) identifying the boundary wave
conditions given by the coarser-resolution model (as sets of
Hs-Tp-Dirm) and (b) scanning the
data set for the ASCII file corresponding to the specific wave conditions and
retrieving it. In the case that no data set file matches exactly the set of
defined wave parameters, the algorithm will additionally (c) define the upper
and lower classes' boundaries for all three parameters (i.e. Hs,
Tp, Dirm) on the basis of their original query
values, scan the data set and retrieve the respective ASCII files,
(d) implement a trilinear interpolation in the three-dimensional
Hs-Tp-Dirm space (according to
Bourke, 1999; Kreyszig, 2010) for each node of the computational mesh and
finally (e) store the derived parameter values in a new query-tailored ASCII
file. The latter will represent the nearshore wave conditions for the
query-defined set of wave parameters.
It should be noted that the division to a large number of parameter classes
at the first steps of this approach will lead to a large number of scenarios
and, consequently, a large number of runs to be performed by the coastal wave
model, with the respective effect on computational cost. However, this will
accordingly lead to a higher accuracy of the trilinear interpolation method
as well, considering that its intrinsic error becomes lower with the increase
in scenario discretization. Given that – in the framework of an operational
system – response speed is of the essence, the combination of the specific
interpolation method with an adequately high number of defined scenarios is
deemed to deliver the best performance overall due to its simplicity and
implementation speed.
TELEMAC and MIKE21 have been extensively used over the years in research,
operational and engineering design applications in maritime/coastal
hydraulics; for MIKE21 this use leans significantly towards the last two
categories, it being one of the most widespread commercial suites for
relevant applications. Their models have been separately evaluated and
validated for several case studies. Regarding TELEMAC, exemplary reference
can be made to the work of Brière et al. (2007) on assessing its performance
for a hydrodynamic case study; Brown and Davies (2009), Luo et al. (2013) and
Villaret et al. (2013) on coupled wave/hydrodynamics–sediment
transport/morphological modelling; Sauvaget et al. (2000) on the modelling of
tidal currents; and Jia et al. (2015) on wave–current interactions in a
river- and wave-dominant estuary. Regarding MIKE21, respective literature
review would include the work of Siegle et al. (2007) and Ranasinghe et
al. (2010) on coupled wave/hydrodynamics–sediment transport/morphological
modelling, Babu et al. (2005) on the modelling of tide-driven currents, Kong
(2014) on the impact of tidal waves on storm surge and Aboobacker et
al. (2009) and ArıGüner et al. (2013) on wave modelling. However, and
given the fact that regarding system architecture and modelling components
TELEMAC and MIKE21 have a lot of similarities (see also Sect. 2.1),
literature has to show limited references on their comparative evaluation.
The rationale behind the model intercomparison presented in the following
derives from the general framework within which this work is carried out,
that is the use of high-resolution wave and hydrodynamics models for (a) the
development and application of a multiparametric approach for the rapid
assessment of wave conditions at inshore locations (presented in Sects. 2.2
and 3.2) and (b) the development of a modelling system coupling atmosphere,
ocean and coastal dynamics (presented in Gaeta et al., 2016). Accordingly,
the TELEMAC and MIKE21 suites are compared in fundamental wave–hydrodynamics
modelling applications, aiming to
test models' performance and the representation of the various processes
governing wave propagation and wave-induced nearshore hydrodynamics. The
comparison is performed for both single wave events and time series or random
waves, representative of typical applications for coastal planning, decision
support and assessment. Apart from a coastal stretch near the city and
harbour of Brindisi, applications (using only TOMAWAC) are also performed for
the area around the city of Bari, including its harbour. Specifically
regarding the latter – and given the inherent limitations posed by the
inclusion of diffraction in phase-averaged models – it should be noted that
the objective was solely to test the extent to which spectral models like
TOMAWAC could be used to capture diffraction effects near harbour entrances
(when the detailed agitation inside the harbour is not of interest), without
the need to resort to separate time-demanding applications using
phase-resolving models. The intercomparison also retains a strong
user-oriented component, presenting examples of how models perform under
typical coastal application scenarios and how basic physical processes affect
the computed parameters of interest.
Study areas and mesh generation
The first of the two study areas is located northwest of the city of Brindisi
(South Italy), comprising Torre Guaceto, a marine protected area and state
natural reserve of significant importance. The selected
rectangular outline of the domain for the model applications measures about
21 km in the longshore and 7.5 km in the cross-shore direction; Fig. 1a shows
the wider study area and the aforementioned outline. The second study area
comprises the coastal area around the city and harbour of Bari (South Italy);
the outline of the computational domain in this case measures about
16.5 km in the longshore and 8.5 km in the cross-shore
direction (see Fig. 1b).
Satellite images of the wider areas, outlines of the computational
domains, meshes, bathymetries, linear transects and points for results'
analysis for the Brindisi–Torre Guaceto (a, c) and Bari
(b, d) case studies (background images from Google Earth, 2016;
privately processed).
As mentioned in Sect. 2.1, both the TELEMAC and MIKE21 modelling suites
discretize the computational domain by unstructured triangular meshes. Mesh
generation for TELEMAC applications was done using Blue Kenue, a data
preparation, analysis and visualization tool for hydraulic modellers
developed by the National Research Council of Canada; the respective work for
MIKE21 was done using MIKE Zero, the DHI tool for managing MIKE projects.
The bathymetric and shoreline data used in this work resulted from the
digitization of nautical charts acquired from the Italian National
Hydrographic Military Service (“Istituto Idrografico della Marina
Militare”). For the case study of Brindisi–Torre Guaceto the triangular
mesh was created defining two density zones (20 m edge length below the
-10 m isoline and 250 m for the rest of the field), resulting in a
mesh consisting of 55 340 nodes forming 109 124 elements. It should be
noted that the mesh was first created in Blue Kenue and afterwards properly
transformed to MIKE Zero format, maintaining the exact same nodes and
connections in order to exclude mesh-dependent divergences in the model runs.
Figure 1c shows the mesh and bathymetry of the computational domain, along
with the three linear transects and six points for which model results will
be intercompared (see Sect. 3.1.3). For the case study of Bari, three density
zones were defined arriving to the finest discretization of 10 m edge length
in order to represent harbour structures, 250 m being the lowest
discretization moving offshore. The resulting mesh consists of 25 202 nodes
forming 46 144 elements; Fig. 1d shows the mesh and bathymetry of the
computational domain, along with the linear transect and three points used
for results' analysis (see Sect. 3.1.3).
a TELEMAC-only run (see Sections 2.1 and 3.1.1). bHs is significant wave height, Tm is mean
wave period, Dirm is mean wave direction.
c Current speed/direction are intercompared only at PTc1, PTc2 and PTc3. d Stand-alone TOMAWAC runs (see Sects. 2.1 and 3.1.1).
Application set-up for model intercomparison
Table 1 presents a detailed overview of all model runs; the table is divided
in two parts, the top one referring to the Brindisi–Torre Guaceto
applications and the bottom one to the Bari applications (see also Fig. 1).
Runs for the Brindisi–Torre Guaceto case study refer to coupled wave and
hydrodynamics models applications, that is coupled TOMAWAC–TELEMAC-2D and
MIKE21-SW–MIKE21-HD runs for the TELEMAC and MIKE21 suites respectively.
Runs for the Bari case study refer to stand-alone TOMAWAC applications, in the
framework of the conceptual approach as presented in Sect. 3.1.1. Every model
run is assigned a different codename, henceforth used for its identification,
with each line of Table 1 defining the forcing used (i.e. single wave events
or time series of random waves); the processes included in the wave models'
set-up (see Table 2 and Sect. 2.1); the transects along which or the points at
which results are intercompared; the parameters included in the comparison;
and, finally, a reference to the figure(s) presenting the specific results in
Sect. 4.
Definition of the processes included in TELEMAC and MIKE21 spectral
wave models' set-up (see Table 1).
a Processes applied only to TELEMAC runs as Triads
(SPB) are available only in TOMAWAC (see Sects. 2.1 and 3.1.1).
b Processes applied to stand-alone TOMAWAC runs (see Sects. 2.1
and 3.1.1).
Characteristics of the wave events (WE1, WE2) and time series (TS1,
TS2) used as forcings for TELEMAC and MIKE21 runs (see also Table 1).
The forcings were selected to represent a wide range of conditions regarding
the wave climate in the areas of interest. The two single wave events
selected, henceforth denoted as WE1 and WE2, represent the 50- and 2-year
return period waves as resulted from the analysis of Regione Puglia (2009).
The two 12 h time series selected, henceforth denoted as TS1 and TS2, were
identified after analysis of wave data from the buoy of Monopoli (lat/long:
40∘58.5′ N, 17∘22.6′ E; depth: 90 m), part of
the Italian wave metric network RON (“Rete Ondametrica Nazionale”;
Corsini et al., 2006). All their characteristics are presented in Fig. 2.
The processes included in the wave models' set-up are presented in Sect. 2.1.
It should be highlighted that each of these common processes (also presented
in Table 2) was included in the set-up of TOMAWAC and MIKE21-SW using the same
parameterizations. Energy transfer from wind to waves (term Sin in
Eq. 2) and nonlinear energy transfer due to quadruplet (four-wave)
interactions (term Snl4 in Eq. 2) were not included, as their
effects on spectral evolution would have been insignificant for the model
intercomparison as it has been set-up on the basis of the rationale presented
in Sect. 3.1.1 (i.e. focus on the nearshore, dictating the relatively small
size of the computational domain in the cross-shore direction).
Considering that model results presented over the entire computational domain
(as 2-D fields of the respective parameters) would pose significant
challenges to the perceptibility of any intercomparison attempt (between both
different modelling suites and different processes), it was deemed preferable
to compare model results along linear transects from the offshore
computational boundary to the shoreline (for WE1 and WE2) or at
specific points (for TS1 and TS2). For the Brindisi–Torre Guaceto case
study, transects TRc1, TRc2 and TRc3 were defined in order to capture areas
of different/representative bathymetry profiles alongshore; the pairs of
points PTc1–PTc4, PTc2–PTc5 and PTc3–PTc6 were defined at specific
locations of the aforementioned transects respectively. The first point of
each of the previous pairs was selected to fall within the breaker zone and
second one before the breaker line; given that – regarding the hydrodynamics
– the objective was to compare wave-generated currents, the hydrodynamics
models' results were analysed only at points PTc1, PTc2 and PTc3. The
locations of the points were decided to not change between runs for different
forcings, in order to facilitate the comprehensibility of the presented
results. For the Bari case study, the objective being to test the diffraction
algorithm's performance in spectral wave models (see also Sect. 3.1.1), one
transect was defined (TRh1) and three points along it: one at the vicinity of
the outer breakwater tip (PTh1), one right at the middle of the harbour's
entrance (PTh2) and one inside the harbour close to the entrance (PTh3). All
transects, points and bathymetric profiles are presented in Fig. 1c and d.
Multiparametric approach for the rapid assessment of wave
conditions
The multiparametric approach presented in this work was applied to three
areas of interest in South Italy: the areas around the cities/ports of
Brindisi and Bari, as well as the Gulf of Taranto (see Fig. 3). Accordingly,
the scenarios representing wave conditions in the wider area were defined
based on the analysis of data from the buoys of Monopoli (lat/long:
40∘58.5′ N, 17∘22.6′ E; depth: 90 m; see Fig. 3)
and Crotone (lat/long: 39∘01.4′ N, 17∘13.2′ E; depth:
95 m; see Fig. 3), covering the period from 1 January 1989 to
31 December 2012. For each buoy data set, wave parameters were further
divided into a number of classes each – according to Table 3 – forming by
aggregation the scenarios to be used for the wave model runs (i.e. sets of
Hs-Tp-Dirm). Figure 4a and b show
the frequencies of occurrence of the scenarios' Hs-Tp
and Hs-Dirm pairs respectively for the
Monopoli data set; Fig. 4c and d show the respective frequencies for Crotone.
It should be noted that all directions follow the nautical direction
convention; negative values were used in Fig. 4b for representation issues,
as gaps in certain direction ranges (i.e. corresponding to what would be
seaward wave origins) were omitted.
Simulations were performed using MIKE-SW, the spectral wave model of the
MIKE21 suite (see description in Sect. 2.1). Mesh generation was done using
MIKE Zero (Fig. 3 shows the modelling domains' outlines); the overall set-up
methodology is described in Sect. 3.1, including the processes of energy
dissipation due to bathymetric breaking and bottom friction. The previously
defined scenarios were used – in sequence – as boundary conditions for the
model runs; the scenarios resulted from the Monopoli data set were used in
the Brindisi and Bari runs, while the ones from the Crotone data set in the
Gulf of Taranto runs. Model results created three extensive data sets (one
for each study area), stored in properly named ASCII files, as described in
Sect. 2.2. The performance of the developed query algorithm, also described
in Sect. 2.2, was tested for a series of exemplary cases before its
operational implementation.
Computational domain outlines for the three areas in South Italy
where the proposed multiparametric approach was applied and locations of the
Monopoli and Crotone buoys; the grid lines and points represent the
WAVEWATCH III rectilinear grid (background image from Google Earth
(2016); privately
processed).
In the framework of the Research Project “TESSA” (Development of
Technologies for the Situational Sea Awareness), the specific multiparametric
approach was applied using WAVEWATCH III (Tolman, 2009) as the
coarser-resolution model that would feed sets of Hs-Tp-Dirm to the query algorithm in order to retrieve/create
the nearshore wave conditions file (based on MIKE-SW results); the model's
rectilinear grid is presented in Fig. 3.
Class properties applied to the wave parameter data sets for
scenarios' definition.
Frequencies of occurrence of the scenarios'
Hs-Tp and Hs-Dirm pairs
for the Monopoli data set (a and b respectively) and the
Crotone data set (c and d respectively).
Comparison of TELEMAC and MIKE21 results (Hs-Tm-Dirm) along transects (a) TRc1 and
(b) TRc2, for the Brindisi–Torre Guaceto case study (forcing WE1;
*= Tc14).
(a) Comparison of TELEMAC and MIKE21 results
(Hs-Tm-Dirm) along transect TRc3 for
the Brindisi–Torre Guaceto case study (forcing WE1; *= Tc14);
(b) effect of different processes on Hs for TELEMAC
(top) and MIKE21 (bottom).
Comparison of TELEMAC and MIKE21 results (Hs-Tm-Dirm) along transects (a) TRc1 and
(b) TRc2, for the Brindisi–Torre Guaceto case study (forcing WE2;
*= Tc24).
(a) Comparison of TELEMAC and MIKE21 results (Hs-Tm-Dirm) along transect TRc3 for the
Brindisi–Torre Guaceto case study (forcing WE2; *= Tc24);
(b) effect of different processes on Hs for TELEMAC
(top) and MIKE21 (bottom).
Comparison of TELEMAC and MIKE21 results (Hs) at points
within (PTc1, PTc2, PTc3) and outside of the breaker zone (PTc4, PTc5, PTc6)
for the Brindisi–Torre Guaceto case study for runs (a) Tc31,
(b) Tc32 and (c) Tc33/Tc34*.
Comparison of TELEMAC and MIKE21 results (Curr. speed/direction) at
points within the breaker zone (PTc1, PTc2, PTc3) for the Brindisi–Torre
Guaceto case study for runs (a) Tc31, (b) Tc32 and
(c) Tc33/Tc34*.
Comparison of TELEMAC and MIKE21 results (Hs) at points
within (PTc1, PTc2, PTc3) and outside of the breaker zone (PTc4, PTc5, PTc6)
for the Brindisi–Torre Guaceto case study for runs (a) Tc41,
(b) Tc42 and (c) Tc43/Tc44*.
Comparison of TELEMAC and MIKE21 results (Curr. speed/direction) at
points within the breaker zone (PTc1, PTc2, PTc3) for the Brindisi–Torre
Guaceto case study for runs (a) Tc41, (b) Tc42 and
(c) Tc43/Tc44*.
Comparison of TOMAWAC results for the Bari case study:
(a, d) along transect TRh1 for forcings WE1 and WE2 respectively;
(b, c) and (e, f) as wave fields at the harbour of Bari for
forcings WE1 and WE2 respectively.
Comparison of TOMAWAC results (Hs-Tm-Dirm) for the Bari case study at points PTh1, PTh2 and PTh3
for forcings (a) TS1 and (b) TS2.
Results and discussion
As described in Sect. 3.1 and presented in Tables 1 and 2, model
intercomparison regards the Brindisi–Torre Guaceto case study. Figures 5
and 6 show the comparison of TELEMAC and MIKE21 results
(Hs-Tm-Dirm) along transects TRc1,
TRc2 and TRc3 for forcing WE1, as well as the effect of different processes
on Hs along the specific transects, separately for each modelling
suite; Figs. 7 and 8 show the respective results for forcing WE2. The overall
agreement between model results is good, and all parameters are very close
for the majority of runs for both forcings, with a general observation being
that TELEMAC constantly produces slightly higher values of Hs and
lower values of Tm than MIKE21. The extensive set of runs tested
allows for a more detailed analysis of the models' performance, as presented
in the following. For runs Tc11 and Tc21, including the processes of breaking
and bottom friction dissipation, Hs values are practically
overlapping along most part of all three transects, with the exception of the
divergences observed at the vicinity of the breaker line (more noticeable for
the relatively mild slope TRc1 rather than TRc2 and TRc3); Tm and
Dirm show small divergences as well, mostly noticeable after
breaking for the steeper slope profiles of TRc2 and TRc3 and for the
higher-wave forcing WE1 (i.e. Tc11 run). The inclusion of the process of
energy dissipation due to whitecapping in runs Tc12 and Tc22 results in a
small decrease of Hs overall, which is more clearly noticeable in
Figs. 6b and 8b presenting such effects separately for TELEMAC and MIKE21;
changes in Tm and Dirm are barely noticeable between
Tc11–Tc12 and Tc21–Tc22 runs. The additional inclusion of the non-linear
triad interactions in runs Tc13 and Tc23 leads to the most noticeable
discrepancies between model results (again, more noticeable for the
relatively mild slope TRc1 rather than TRc2 and TRc3), which is limited (as
expected) to the shallow water sections of the studied profiles/transects
where the specific process's effect becomes significant. Although both suites
use the LTA model of Eldeberky and Battjes (1983), the inclusion of triads
seems to have a rather small effect on MIKE21 Hs results (slight
decrease of wave height and shift of the breaker line seaward), with the
effect on the wave energy spectrum, however, becoming more evident when
comparing Tm values. In contrast, TELEMAC runs result in
higher Hs values right before breaking and quite lower
Tm values inshore. Dirm results show small
divergences for both modelling suites. Additionally to runs Tc13 and Tc23,
the effect of triad interactions was also tested using the SPB model of Becq
(1998), available as an alternative option only in TOMAWAC; the test was
included as Tc14 in the set of runs, and its results are represented as
dotted lines in all figures (noted accordingly). Following Becq-Girard et
al. (1999) remarks on the validity range of the LTA model, Tc14 results show
indeed a quite different representation of the process by TOMAWAC, with
milder evolution of the wave energy onshore and smaller changes to all
parameter values than Tc13 produced (see Figs. 6b and 8b in particular).
Figures 9 and 10 show the comparison of TELEMAC and MIKE21 results
(Hs and Curr. speed/direction respectively), for the time series
forcing TS1; Figs. 11 and 12 show the respective results for forcing TS2.
Significant wave height values are compared at points along transects TRc1,
TRc2 and TRc3 (see Fig. 1), three of them within the breaker zone (PTc1,
PTc2, PTc3) and three outside of it (PTc4, PTc5, PTc6); the wave-generated
currents' speed and direction are compared only at points PTc1, PTc2 and PTc3
(see also Sect. 3.1.3). Regarding Hs, the comparison between
results at pairs PTc1–PTc4, PTc2–PTc5 and PTc3–PTc6 highlights the effect
different processes have on model results for propagating waves towards the
nearshore and how including/omitting them may become significant (or
insignificant) for various operational, planning and engineering design
applications in coastal areas. TELEMAC and MIKE21 results at points PTc4,
PTc5 and PTc6 are close and in-phase for all processes, with higher
discrepancies observed for the higher-wave forcing TS2. At points PTc1, PTc2
and PTc3, the conclusions drawn from the analysis of the wave events' results
in the previous can be clearly identified here as well, with the most
significant alterations in the different suites' results observed again for
the runs where triad interactions were included in the modelled processes
(i.e. Tc33/Tc34 and Tc43/Tc44); it should be also noted that the higher-wave
forcing TS2 leads to smaller variations of Hs than TS1 overall,
thus minimizing the effect of the different approach for triads modelling in
run Tc44 too. Regarding the wave-generated currents, TELEMAC and MIKE21
results are in relatively good agreement for all runs considering the order
of magnitude of the resulting current speeds, as well as the sensitivity of
current directions within the breaking zone. Figure 10 shows that for runs
Tc31 and Tc32 results are very close with the exception of the period up to
hour 4 at PTc1, where TELEMAC shows current speeds close to zero (with the
respective effect on current direction). As noted in the previous, the
introduction of triad interactions results in a more significant effect when
modelled with TELEMAC, although the SPB model does lead to smoother results
regarding both speed and direction (run Tc34). Point PTc3 shows larger
divergences than points PTc1 and PTc2 that attributed to the combination of
its location in the computational domain and the significant shift in the
forcing's direction after hour 6 (see Fig. 2). Figure 12 shows that at points
PTc1 and PTc2 results are in good agreement for both TELEMAC and MIKE21,
following the remark regarding the small Hs variations observed
in the breaking zone for TS2 (see Fig. 11). At PTc3 TELEMAC results are
similar to the MIKE21 ones between hours 3 and 7 but significantly higher at
the beginning and the end of the simulated time series.
Regarding the Bari case study, it should be stated again (as in Sect. 3.1.1)
that the objective of its inclusion in this work was solely to test the
extent to which spectral models could be used to capture diffraction effects
near harbour entrances in the framework of operational approaches like the
one presented in Sect. 3.2. This was done while keeping in mind the inherent
limitations posed by the inclusion of diffraction in phase-averaged models,
as well as the fact that a detailed study of harbour agitation would require
the use of a phase-resolving model. Figure 13 shows TOMAWAC results (with and
without the inclusion of diffraction) along transect TRh1 and as wave fields
at the area of the harbour, for forcings WE1 (Fig. 13a, b and c respectively)
and WE2 (Fig. 13d, e and f respectively). Results show noticeable differences
in wave characteristics around the breakwater's tip and near the harbour
entrance, while the model also manages to capture the diffusion of the wave
height inside the harbour area; larger effects are observed for the
higher-wave forcing WE1. Figure 14 shows TOMAWAC results (with and without
the inclusion of diffraction) at points PTh1, PTh2 and PTh3 for forcings TS1
and TS2 (Fig. 14a and b respectively). Differences are noticeable for all
parameters, being relatively more significant at points PTh2–PTh3 and for
the higher-wave forcing TS2.
Finally, the multiparametric approach presented in this work was successfully
implemented in the framework of the Italian Flagship Research Project TESSA,
using WAVEWATCH III (Tolman, 2009) to feed
sets of offshore wave characteristics to the query algorithm in order to
provide nearshore wave conditions from the created database of MIKE-SW results. Its performance was
tested for a series of different wave conditions for the three areas of
interest (i.e. Brindisi, Bari and Gulf of Taranto; see Fig. 3) and the
algorithm managed to deliver results in a fast and seamless way at all times.
Conclusions
This work presents the comparison of the TELEMAC and MIKE21
modelling suites in fundamental wave and hydrodynamics applications for the
representation of nearshore dynamics and proposes a multiparametric
scenario-based approach for the rapid assessment of wave conditions in
coastal zones that aims to serve as an operational tool for coastal planning,
decision support and assessment. The study areas for the presented
applications are all located in South Italy and comprise the coastal area
around the city/port of Brindisi, the coastal area around the city/port of
Bari and the Gulf of Taranto. For the first one, TELEMAC and MIKE21 are
intercompared for a series of application set-ups aiming to test the models'
performance and the representation of the various processes governing wave
propagation and wave-induced nearshore hydrodynamics. For the study area of
Bari (including its harbour), the spectral wave model of TELEMAC (i.e.
TOMAWAC) is applied with and without the inclusion of the representation of
the processes of diffraction in order to test the extent to which similar
models could be used to capture diffraction effects near harbour entrances,
in the framework of operational approaches like the one presented in this
work when the detailed agitation inside the harbour is not of interest.
TELEMAC and MIKE21 results are compared on the basis of wave/current
characteristics, along linear transects from the offshore to the nearshore
and at specific points inside/outside the breaker zone and near the entrance
of the harbour for the study area of Bari. Analysis shows an overall
satisfactory agreement between the two modelling suites and is deemed to
provide useful insights on both their individual capabilities and their
comparative evaluation. The specific tasks also served as the background
study for the development of a modelling system based on a multiple-nesting
approach, coupling atmosphere, ocean and coastal dynamics (described in Gaeta
et al., 2016), while it also retains a strong user-oriented component,
showing examples of how models perform under typical coastal application
scenarios and how basic physical processes affect the computed parameters of
interest. The proposed multiparametric approach is presented in detail,
consisting of the definition of a number of wave scenarios on the basis of
field measurements, a data set of wave model results using these scenarios as
boundary conditions and a query algorithm based on the trilinear
interpolation that bridges coarser-resolution operational models and the
aforementioned data set in order to provide query-tailored fields of
nearshore wave dynamics. The implementation of the specific approach as part
of an operational chain for all three study areas in South Italy in the
framework of the Italian Flagship Project TESSA supports the rationale behind
this study, while setting the basis for future work on the same path.
Acknowledgements
This work was performed and funded in the framework of the Italian Flagship
Project “TESSA – Development of Technologies for the Situational Sea
Awareness” supported by the PON01_02823/2 “Ricerca &
Competitività 2007–2013” program of the Italian Ministry for Education,
University and Research.
The authors would like to thank Andrea Pedroncini from DHI Italia and Antonio
Bonaduce from the Euro-Mediterranean Center on Climate Change for their
valuable help on various modelling aspects of the MIKE21 suite and on
assisting with the operational implementation of the multiparametric approach
in the framework of the TESSA Project respectively. The authors would also
like to thank the Handling Editor Ivan Federico and the two referees for
their constructive comments and suggestions.
Edited by: I. Federico
Reviewed by: C. Koutitas and one anonymous referee