ARTICLE IN PRESS

Continental Shelf Research 30 (2010) 692–706

Contents lists available at ScienceDirect

Continental Shelf Research

0278-43

doi:10.1

� Corr

E-m

journal homepage: www.elsevier.com/locate/csr

The tidal and wind induced hydrodynamics of the composite system AdriaticSea/Lagoon of Venice

Tomas Lovato a,�, Alexey Androsov b, Dmitry Romanenkov c, Angelo Rubino a

a Department of Environmental Sciences, University of Venice, Dorsoduro 2137, Venice 30123, Italyb Alfred Wegener Institute for Polar and Marine Studies, Am Handelshafen 12, Bremerhaven 27570, Germanyc P.P. Shirshov Institute of Oceanology, Pervaya Liniya 30, St. Petersburg 199053, Russia

a r t i c l e i n f o

Article history:

Received 15 December 2008

Received in revised form

17 December 2009

Accepted 6 January 2010Available online 18 January 2010

Topical issue on "Tides in Marginal Seas"

(in memory of Professor Alexei V.

Nekrasov). The issue is published with

support of the North Pacific Marine Science

Organization (PICES).

processes within the lagoon. Firstly, a model calibration was performed by comparing the results of the

model, forced using tides and ECMWF atmospheric pressure and wind fields, with observations

collected for a set of 33 mareographic stations uniformly distributed in the Adriatic Sea and in the

Lagoon of Venice. A second numerical experiment was then carried out by considering only the tidal

forcing. Through a comparison between the results obtained in the two experiments it was possible to

assess the reliability of the estimated parameter through the composite forcing. Model results were

Keywords:

Hydrodynamic model

Tide

Air–sea interaction

Near shore currents

Adriatic Sea

Lagoon of Venice

43/$ - see front matter & 2010 Elsevier Ltd. A

016/j.csr.2010.01.005

esponding author. Tel.: +39 0412348911; fax

ail address: lovato@unive.it (T. Lovato).

a b s t r a c t

A shallow water hydrostatic 2D hydrodynamic numerical model, based on the boundary conforming

coordinate system, was used to simulate aspects of both general and small scale oceanic features

occurring in the composite system constituted by the Adriatic Sea and the Lagoon of Venice (Italy),

under the influence of tide and realistic atmospheric forcing. Due to a specific technique for the

treatment of movable lateral boundaries, the model is able to simulate efficiently dry up and flooding

then verified by comparing simulated amplitude and phase of each tidal constituent as well as tidal

velocities simulated at the inlets of the lagoon and in the Northern Adriatic Sea with the corresponding

observed values. The model accurately reproduces the observed harmonics: mean amplitude

differences rarely exceed 1 cm, while phase errors are commonly confined below 151. Semidiurnal

and diurnal currents were correctly reproduced in the northern basin and a good agreement was

obtained with measurements carried out at the lagoon inlets. On this basis, the outcomes of the

hydrodynamic model were analyzed in order to investigate: (i) small-scale coastal circulation features

observed at the interface between the adjoining basins, which consist often of vortical dipoles

connected with the tidal flow of Adriatic water entering and leaving the Lagoon of Venice and with

along-shore current fields connected with specific wind patterns; (ii) residual oscillations, which are

often connected to meteorological forcing over the basin. In particular, it emerges that small-scale

vortical features generated near the lagoon inlet can be efficiently transported toward the open sea,

thus contributing to the water exchange between the two marine regions, and a realistic representation

of observed residual oscillations in the area would require a very detailed knowledge of atmospheric as

well as remote oceanic forcing.

& 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Coastal lagoons, among other transitional systems, are verysensitive to the hydrodynamic processes of the neighbouringmarine regions, which are principally governed by the tidalregimes and the weather conditions. The resulting circulation,which can be profoundly influenced by water injections at the seashore, can also induce and/or sustain a large spatial and temporalvariability of physical, chemical and biological characteristics

ll rights reserved.

: +39 0412348964.

(Kjerfve, 1994; Ward, 1998). Therefore, a deep understanding ofthe hydrodynamic processes that govern the fluxes of energy andmatter not only within coastal lagoons, but also in the regions oftransition toward the open sea, constitutes a strong prerequisitefor the integrated management of the coastal ecosystems(see, e.g., Zaldivar et al., 2003; Guyondet and Koutitonsky, 2006).

The hydrodynamics of the Adriatic Sea shelf and its coastalsystems were thoroughly investigated using both statisticaland deterministic models (Hendershott and Speranza, 1971;Malanotte Rizzoli and Bergamasco, 1983; Orlic et al., 1994;Malacic et al., 2000; Cushman-Roisin and Naimie, 2002; Zavatar-elli et al. 2002; Janekovic et al., 2003; Petaccia et al., 2006;Lionello et al., 2006; Pullen et al., 2007). As most of these studies

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 693

focused on the basin scale circulation and its variability, only asmall emphasis was given to processes involving the integratedhydrodynamics of the open sea and its neighbouring coastal areas(see, e.g., Bergamasco et al., 1998; Bellafiore et al., 2008). Often inthe past, open sea and coastal basins were examined as separatedsystems. In the case of the Lagoon of Venice, the boundariesbetween these two systems were usually located near the lagooninlets (Defina, 2000; Casulli and Zanolli, 2002; Umgiesser et al.,2004) and, in general, the interior lagoon area was completelyneglected when modelling the general circulation of the AdriaticSea (Cushman-Roisin and Naimie, 2002; Janekovic and Kuzmic,2005; Martin et al., 2006). So far, different aspects of the coastalhydrodynamics, which may be influenced of even directly arise bythe presence of small-scale bathymetric features and/or localcoastal currents at the lagoon inlets are still unclear.

Recently, fresh hydrodynamic data were collected from boththe Adriatic Sea and the Lagoon of Venice. These data providednew insights on the long-term tidal and non-tidal characteristicsof the local coastal currents (Gacic et al., 2004; Book et al., 2005),on the surface circulation and its spatial and temporal variabilityalong the northern Adriatic coast (Kovacevic et al., 2004;Chavanne et al., 2007), and on the interpretation of sea surfacepatterns of both the Adriatic Sea and the Lagoon of Venice (see e.g.Janekovic and Kuzmic, 2005; Ferla et al., 2007). Furthermore,recent bathymetric charts for the Lagoon of Venice were compiledby the Venice Water Authority on the basis of in situ measure-ments collected in the period 2000–2004. These data represent agood basis for the calibration and verification of a new hydro-dynamic model of the composite system Lagoon of Venice–NorthAdriatic basin, which, on such basis, can be used for investigatingthe local mesoscale and small-scale phenomena which areexpected to be markedly influenced by the complex morphologyand bathymetry profiles of the lagoon area and the nearby marineregion.

In fact, the numerical model provides a means of isolating theinfluence of tidal forcing on the generation of specific featuresassociated to the water exchange between the lagoon and theopen sea, like, e.g., small-scale tidally induced eddies. Previousinvestigations (Imasato, 1983; Geyer and Signell, 1990, 1992)demonstrate that eddies generated in tidal flows and typicallylocalized nearby headlands, inlets, or shoals, are inherentlytransient and their spatial structure may vary considerablythrough the tidal cycle. The propagation of local transitionaleddies has thus important implications to dispersion processessince their development may strongly enhance the spatial andtemporal variability of biogeochemical characteristics of theLagoon of Venice, as well as, sedimentation rates and/or pollutantspreading.

In order to achieve these goals, a 2D boundary-conforming,finite-difference hydrodynamic numerical model was applied tothe composite system constituted by the Adriatic Sea and theLagoon of Venice. The newest experimental datasets were used toproduce a realistic bathymetry, especially in the Lagoon of Veniceand its inlets. The model was calibrated for the whole computa-tional domain through a comparison between simulated andobserved amplitude and phase of the principal constituents of thetide. After the calibration, an analysis of the robustness andaccuracy of the model was carried out for the lagoon and the opensea, using the harmonic data from tide gauges and currentmetermeasurements. The hydrodynamic model was finally addressed tothe reproduction of observed tidally and atmospherically inducedsea water levels and small-scale circulation features characteriz-ing the northern area of the composite system and the zonebetween the adjoining basins.

The paper is organized as follows: in the next section theobservational data sets and the relevant characteristics of the

performed numerical experiments and model set up aredescribed. In Section 3 the results of the calibration andverification experiments are presented, with a particular attentionto the reproduction of the spatial distribution of the tidalcomponents and tidal currents. In Section 4, the outcomes ofthe model are discussed and simulated water levels and small-scale flow features are compared to previous experimentalevidence. The work is summarized in the final section.

2. Methods

2.1. Formulation of the problem and numerical method

The original numerical hydrodynamic model proposed byAndrosov et al. (1995, 2002) was here implemented in a two-dimensional, barotropic version, which includes a specific schemefor the treatment of movable lateral boundaries (see, e.g., Balzano,1998). In a 2D domain, the initial boundary-value problem in theCartesian system of coordinates for the vertically averagedequations of motion and continuity is described by

@v

@tþv � rvþgrz¼ f k� v�

CD

Hvjvjþ

1

HrðKV HrvÞþ

~F

rwH�

1

rw

rp;

@zdtþr � ðHvÞ ¼ 0 ð1Þ

where v=(u,v) is the fluid averaged velocity, z the sea surfacedisplacement, h the undisturbed water depth, H=h+z the totalwater depth, r¼ @

@x ;@@y

� �the gradient operator, f the Coriolis

parameter, k the unit vector in the vertical direction, CD thebottom friction coefficient, ra and rw the air and water density,respectively, p the atmospheric pressure, and KV the eddyviscosity coefficient. The tangential wind stress, ~F is expressedthrough the resistance quadratic law (Hellerman, 1967)described by

~F ¼ CDWraUjUj; ð2Þ

where CDW is the wind drag coefficient and U the wind velocityvector. Note that the wind drag coefficient was consideredconstant in this investigation.

The parameterization of the bottom stress still represents amajor issue when modelling the hydrodynamic circulation inshallow water embayment (see e.g. Umgiesser et al., 2004;D’Alpaos and Defina, 2007). In the present study, the bottomfriction coefficient CD was computed according to Blumberg andMellor (1987) and has the following form:

CD ¼1

klnðH=z0Þ

� ��2

; ð3Þ

where z0 the bed roughness length and k the von Karmanconstant. In agreement with the cited work, a minimum value forthe bottom friction coefficient was applied in order to avoid thevanishing of the bottom drag effect when the water depth is verylarge (see Table 1).

The boundary-value problem is transformed into boundary-fitted coordinates mapping the physical domain into a computa-tional rectangle where the sea level and/or current at the openboundary are specified as function of time. At the otherboundaries, which are mappings of the shores, we set no-slipconditions. In particular, the open boundary problem is solvedthorough the reduced boundary-value formulation described inAndrosov et al. (1995): when it is not possible to prescribe bothvelocity and elevation values at the open boundaries, only the sea

ARTICLE IN PRESS

Table 1Parameters of the hydrodynamic model.

Parameter Symbol Value Reference

Time step Dt 100 s

Grid spacing Dx, Dy 12 Km–50 m

Dry up threshold H* 0.1 m Model default

Bottom roughness length z0 0.005 m Calibrated

Minimum bottom friction coefficient value CDmin 2.5�10�3 Blumberg and Mellor (1987)

Horizontal viscosity coefficient KV 0.5 m2 s�1 Model default

Wind drag coefficient CDW 2.5�10�3 Orlic et al. (1994)

Water density rW 1025 kg m�3 Orlic et al. (1994)

Air density ra 1.247 kg m�3 Orlic et al. (1994)

SOR convergence criteria – 1�10�5 m Model default

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706694

level is prescribed, regardless of the boundary regime, andradiation conditions are applied to the water outflow.

The equations, in the form of contravariant fluxes (seeRomanenkov et al., 2001), are solved using the so-calledpressure-correction method (Van Kan, 1986). At each time step,the velocity field due to the advective transport is determined bysplitting in coordinate directions. Advection is approximated by ascheme with a switch allowing for the use of an upstream schemeof the first, second, or third order. Then, for the advection field andthe explicit level representation, the momentum equations areintegrated by splitting in coordinate directions and preliminaryvelocities are found. Afterward, the level is determined from theboundary-value problem for a two-dimensional elliptic equationwhich is obtained from a linear combination of the continuityequation and of the divergence of the momentum equation. Oncethe water level has been computed, the final velocity componentsare determined from the momentum equations. The scheme hassecond-order accuracy in both space and time, with a time steplimited only by advection terms in the equation of motion. For adetailed description of the complete formulation of the boundary-value problem and its solution method the reader is referred tothe work of Androsov et al. (2002).

2.2. Study area and model set up

The physical domain considered in this work extends from thenorthern Ionian Sea to the Lagoon of Venice (Fig. 1). The AdriaticSea is a semi-enclosed elongated basin with an extension ofapproximately 700 km along the major axis in the northwest–southeast direction. This regional sea is characterized by abathymetric profile which decreases southward, from thenorthern shallow area (average bottom depth of about 35 m) tothe wide depression of the Jabuka Pit (more than 1200 m waterdeep). The Lagoon of Venice (Fig. 2), which has a surface of about500 km2 and an average depth of approximately 1 m, is located inthe north-western part of the Adriatic Sea and represents themajor wetland connected to the basin. The lagoon is characterizedby a complex network of channels with depths spanning fromcirca 2 m to circa 20 m which cross large areas of very shallowtidal flats. The water exchange between the Northern Adriatic Seaand the lagoon takes place at three narrow inlets (500–1000 mwide) which are situated along the lagoon eastern boundary.

Sea elevation patterns in the Adriatic Sea are mainly driven bythe tidal regime and by the atmospheric forcing that significantlyinfluences the system circulation, especially during the winterperiod (Artegiani et al., 1997a, b). The mean amplitude of the sealevel excursion in the Adriatic Sea significantly increases from theOtranto straight (�0.5 m) toward the northern section, withvalues up to 1 m in the Lagoon of Venice.

The numerical model was applied to the integrated domain bymeans of a curvilinear boundary-conforming grid, composed by287�363 nodes (Figs. 1 and 2). The mesh was refined along thethree lagoon inlets and the major lagoon navigation channels. Infact, these features are recognized to play a primary role in thepropagation of tidal waves inside the lagoon (Umgiesser, 1997).The resulting cells have side lengths varying between approxi-mately �50 m, in the most refined areas inside the lagoon, andapproximately 12 km in the southern part of the domain, incorrespondence of the southern open boundary.

The bottom geometry was obtained through the aggregation ofthe US Navy unclassified 1/601 bathymetric data base DBDB1 forthe Adriatic Sea and the hydrographical chart of the Lagoon ofVenice (scale 1:5000) edited by the Venice Water Authoritythrough its concessionary Consorzio Venezia Nuova (MAV, 2005).The bottom discretization was realized by bilinear interpolation ofthe depth data on the numerical grid. The minimum depth for thetidal flats was set to 0.2 m, namely twice the magnitude of thewet/dry threshold value, in order to avoid numerical instabilities(see e.g., Bates et al., 2005).

The external forcing considered in the simulations were seawater elevation gradient, wind stress, and atmospheric pressuregradient. Other forcings, such as, e.g., temperature and salinitygradients, were not included since it was shown in a number ofstudies that the tidal and atmospherically induced circulation ofthe composite system can be adequately described in thebarotropic case (see e.g., Janekovic et al., 2003; Umgiesser et al.,2004; Chavanne et al., 2007; Bellafiore et al., 2008).

According to the existing literature (Tsimplis et al., 1995;Cushman-Roisin and Naimie, 2002; Janekovic and Kuzmic, 2005),the sea water level at the open boundary (OB) was first assessedthrough the superimposition of the seven major tidal harmonics,grouped on the basis of their characteristic frequencies in asemidiurnal group (M2, S2, N2, and K2) and a diurnal group (K1, O1,and P1). The tidal components thus assessed were then optimizedas explained in Section 2.4. The OB of the system was located�250 Km south of the Otranto Strait in order to reduce thenumerical disturbances introduced by the artificial imposition ofthe boundary conditions. The tidal data, derived from the OSU-TPXO tidal inverse model (Egbert and Erofeeva, 2002), werelinearly interpolated at each node of the boundary line. Data fromthe OSU dataset for the Mediterranean Sea were used for the K1,O1, M2, and S2 constituents and data from the OSU global modelwere used for P1, K2, and N2 (see, e.g., Martin et al., 2006).

The meteorological data used to drive the hydrodynamicmodel were the wind components at 10 m height and the meansea level pressure obtained from the operational archive of theECMWF atmospheric model surface analysis (Persson, 2003). Theatmospheric fields used in the present work were generated dailyat times 00, 06, 12, 18 GMT and were provided on a regular

ARTICLE IN PRESS

Fig. 1. Bathymetric chart of the Adriatic Sea and curvilinear boundary-fitted grid used for the integrated domain. Note that grid resolution in the plot is a fifth of the

original one. Mareographic stations (circle) and the position of ADCP surveys (diamonds) are also indicated. Stations with the asterisk mark were used in the calibration.

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 695

latitude-longitude grid of 0.51. The meteorological data wereinterpolated, in order to provide spatial and temporal inputsconsistent with the model requirements: at each time step theatmospheric fields were applied to the nodes of the computa-tional grid using a bivariate method (Akima, 1984) and by linearlyreconstructing the values within the time interval of the inputdatasets.

Table 1 summarizes the values of the parameters selected forthe numerical simulations. The sea elevation time series, storedwith an hourly time step, were analyzed using the least squaresmethod (Foreman, 1977) for a period of approximately 9 monthsafter the spin up of the simulation (2000 h). In fact, the leastsquares method applied for the extraction of tidal harmonicsrequires a minimum length of the data sample of approximatelyhalf a year in order to reliably determine all tide constituentsprescribed at the open boundary. All numerical simulationscarried out with the external meteorological forcing refer to thecomplete year 2004.

2.3. Observational dataset

The observational data used in this study for the calibrationand validation of the model are divided in two groups: the tidegauge data and the current meter data. The data belonging to thefirst group were obtained from 70 stations located along theAdriatic coast and inside the Lagoon of Venice, the data belongingto the second group from 13 current meters located in thenorthern Adriatic and from 3 long-term ADCP surveys carried outat the three inlets of the lagoon (Figs. 1 and 2).

The harmonic constants for the seven most pronounced tidalconstituents (M2, S2, N2, K2, K1, O1, P1) at the coastal tide gauges werecompiled on the basis of the work of Janekovic and Kuzmic (2005)and Ferla et al. (2007). For the stations of Rovinj, Bakar, Mali Losinj,Zadar, Split, Vis, and Dubrovnik the data were provided by theHydrographic Institute of the Republic of Croatia (HIR, 2004). Inaddition, the harmonic constants for the tide gauge of the CNR Tower(st. 70 in Fig. 1), located in the middle of the Gulf of Venice, wereestimated from the hourly time series provided by the MareographicCentre of the Venice Municipality.

The set of depth averaged tidal velocity data was compiledintegrating the current meter data reported by Cushman-Roisin andNaimie (2002), stations 7–16 (Fig. 1), and the measurementsconducted along the Italian coast near Senigallia by Book et al.(2005), stations 4–6 (Fig. 1). Both data sets included the semi-majoraxis amplitude and the ellipse inclination of the two most energeticconstituents, M2 and K1. From the long term current measurementscarried out during the period 2001–2002 at the three inlets of thelagoon (Gacic et al., 2004), stations 1–3 (Fig. 2), amplitudes andphases of all the seven principal tidal constituents of the velocity canbe retrieved, thus allowing for a quantitative validation of the modelresults at the adjoining region of the domain.

2.4. Model calibration

The numerical solution of the initial boundary-value problemwas realized through the estimation of the optimal values for thebottom roughness length (z0) and the tidal component valuesalong the open boundary (zOB).

ARTICLE IN PRESS

Fig. 2. Bathymetric chart of the Lagoon of Venice and curvilinear boundary-fitted grid used for the integrated domain. Note that grid resolution in the picture is one third of

the original one. Mareographic stations (circle) and the position of ADCP surveys (diamonds) are also indicated. Stations with the asterisk mark were used in the

calibration.

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706696

A number of numerical experiments were carried out, in order totune the model, using both tidal and meteorological forcing for a yearlong period. Tidal harmonic values for a set of 33 mareographicstations homogeneously distributed within the composite domain(they are marked in Figs. 1 and 2) were used for the comparison withthe corresponding values obtained from the least square fit analysis ofthe simulated sea water levels. The objective of the calibration was todetermine the values of the two parameters (z0 and zOB) whichminimize, for each harmonic component, the function

Lðz0; zOBÞ ¼X

i ¼ 1;ns

½Sobsi �Ssim

i �2; ð4Þ

where ns is the number of mareographic stations and Sobs and Ssim theobserved and simulated values.

A second numerical experiment was performed without theinclusion of the meteorological forcing in order to evaluate therelevance of non-tidal signals (which are known to occur asconsequence of intense meteo-climatic events in the Adriatic basin,see Vilibic, 2006) on the estimation of the tidal harmonics. In fact, thetraditional least squares minimization, which is highly sensitive to thepresence of such non-tidal components in the input signal, wouldlead to an over-fit of the non-tidal components in an attempt tominimize the total residual error when the analyzed time series aretoo short or noisy (Leffler and Jay, 2009). Such a comparison wascarried out also to provide a further verification of the approachadopted here for the calibration of the hydrodynamic model.

3. Results

3.1. Tidal harmonics calibration

The results of the calibration experiment are reported in Table 2.The differences between observed and simulated values for the

amplitudes and phases of the tidal constituents are summarized usingfive statistical parameters: mean of absolute differences (|M|),standard deviation of differences (STD), root mean squared error(RMSE), correlation (R), and the average of simulated values (Msim).

The tidally induced water levels were accurately reproduced bythe model (Table 2) and the results obtained for the whole domainappear to be in good agreement with the tide gauge dataset, theaverage correlation of both amplitude and phase being 0.98. TheRMSE values generated for the differences of the semidiurnal groupcomponents is between 0.4 and 1.1 cm for the amplitude andbetween 91 and 161 for the phase. Diurnal group constituentspresented RMSE values with a reduced range of variability, spanningin this case from 0.6 to 0.9 cm and from 41 to 81 for amplitude andphase, respectively. The standard deviation of the simulatedamplitudes exceeded 1 cm only in the case of the lunar semidiurnalconstituent M2 (the most energetic component of the tide in theregion), while the STD of the phase was circa 91 for the semidiurnalgroup of constituents and circa 71 for the diurnal group.

The results of the model run carried out with the mere tidalforcing are reported in Table 3. In this case, the correlationbetween the model output and the observations of bothamplitudes and phases was similar for all harmonicconstituents, with no significant changes in the RMSE and STDfor the more energetic constituents of the tide, namely M2, S2, andK1. Small deviations were produced for the phase of the minortidal harmonics, in particular for N2 and K2, with differencesaround 21.

3.2. Model Verification

The outcomes of the tidal harmonic analysis performed on thewater levels simulated for the full set of mareographic stationsare presented in Table 4, grouped respectively for the Adriatic Sea

ARTICLE IN PRESS

Table 2Results of model calibration for the semidiurnal and diurnal constituents at the stations depicted in Figs. 1 and 2, obtained using the tidal and meteorological forcing.

M2 S2 N2 K2 K1 O1 P1

Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha

Msim 18.0 253.2 11.1 241.2 3.2 245.8 3.7 235.3 14.9 83.8 5.6 71.6 4.2 80.5

|M| 0.7 6.5 0.7 13.8 0.3 7.5 0.5 14.0 0.8 3.8 0.9 3.6 0.5 6.1

STD 1.2 9.1 0.7 9.7 0.3 11.2 0.6 11.4 0.9 6.3 0.6 4.4 0.4 8.7

RMSE 1.1 9.2 0.8 16.1 0.4 11.1 0.6 16.6 0.9 6.3 1.0 4.6 0.6 8.6

R 0.99 0.99 0.99 0.99 0.96 0.98 0.95 0.98 0.99 0.96 0.96 0.98 0.97 0.95

Differences between simulated and observed amplitudes and phases are summarized through the mean of absolute differences (|M|), standard deviation of differences

(STD), root mean square error (RMSE), correlation (R). Averages of simulated values of amplitude and phase (Msim) are also reported. Amplitudes are in cm and phases are in

degrees.

Table 3Results of model calibration for the semidiurnal and diurnal constituents at the stations depicted in Figs. 1 and 2, obtained using only the tidal forcing.

M2 S2 N2 K2 K1 O1 P1

Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha

Msim 18.2 252.6 11.0 241.4 3.2 243.6 3.8 239.2 14.6 83.5 5.0 71.0 4.2 80.9

|M| 0.9 6.6 0.5 13.8 0.3 7.3 0.5 10.8 1.0 3.6 0.4 3.7 0.5 5.2

STD 1.2 9.1 0.7 9.8 0.3 10.9 0.6 11.6 0.8 6.2 0.5 4.1 0.4 8.2

RMSE 1.2 9.4 0.7 16.0 0.4 10.8 0.6 14.1 1.0 6.1 0.6 4.5 0.6 8.1

R 0.99 0.99 0.99 0.99 0.96 0.98 0.95 0.98 0.99 0.96 0.96 0.98 0.97 0.95

Results are presented as in Table 2.

Table 4Results of model verification for the comprehensive set of 70 mareographic stations available for the Adriatic Sea (a) and the Lagoon of Venice (b), obtained using the tidal

and meteorological forcing.

M2 S2 N2 K2 K1 O1 P1

Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha Amp Pha

aMsim 11.5 171.5 7.3 166.8 2.1 175.2 2.3 160.6 10.2 65.1 4.1 54.4 2.8 56.2

|M| 1.6 14.0 0.9 22.4 0.4 16.8 0.4 24.8 0.9 5.9 0.8 5.6 0.7 9.6

STD 1.8 17.6 1.0 17.5 0.5 23.5 0.5 17.4 1.1 7.8 0.6 7.7 0.3 10.0

RMSE 1.9 21.0 1.2 27.6 0.5 23.8 0.5 28.6 1.2 8.0 1.0 7.6 0.8 12.0

R 0.96 0.95 0.96 0.93 0.92 0.93 0.92 0.91 0.99 0.90 0.97 0.79 0.98 0.78

bMsim 22.6 302.2 14.0 294.0 4.0 297.0 4.7 288.4 18.2 94.9 6.7 82.2 5.2 96.0

|M| 0.4 3.3 0.5 9.7 0.2 4.5 0.6 7.6 0.7 2.3 0.6 3.5 0.4 4.0

STD 0.5 4.3 0.4 4.5 0.1 4.8 0.4 4.9 0.5 2.5 0.4 3.1 0.4 3.3

RMSE 0.5 4.3 0.6 10.6 0.3 5.6 0.6 8.9 0.8 3.0 0.7 3.9 0.5 4.9

R 0.97 0.99 0.97 0.99 0.94 0.99 0.86 0.99 0.95 0.98 0.77 0.97 0.72 0.99

Results are presented as in Table 2.

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 697

(a) and the Lagoon of Venice (b). The model was driven by bothtidal and meteorological forcings.

The mean correlation for the amplitudes and for the phases ofall the considered harmonics was 0.96 and 0.89, respectively, inthe Adriatic group of stations. Small differences are noticeable incomparison to the results obtained in the calibration experimentfor the mean absolute error of the amplitudes, which exceeded1 cm only for the M2 component of the tide. The mean values ofthe phase showed differences of nearly 21 for the diurnal group ofconstituents, while the deviations for each component of thesemidiurnal group was between 81 and 141. A similar behaviourcan be noted in the computation of the STD and RMSE statistics.

The means of absolute differences for the group of tide gaugeslocated inside the Lagoon of Venice were remarkably small foreach tidal constituent, with average values of 0.5 cm for the

amplitude and 41 for the phase. Furthermore, the standarddeviation and the RMSE values were in the same range of thoseobtained in the calibration experiment (see Table 2). Minorconstituents of the tide, namely K2 and P1, were characterized by alower agreement between simulated and observed amplitude,which lead to a decrease in the average correlation in comparisonwith the calibration results. On the other hand, the correlationachieved in the simulation of the phases for both tidal constituentgroups was between 0.97 and 0.99.

The simulated amplitude of the semidiurnal M2 tide variesfrom 6 cm in the Otranto straight up to �25 cm in northern end ofthe Adriatic Sea, with a marked minimum near the border ofmiddle and northern parts of the basin. The amplitude of thediurnal K1 tide smoothly increases from the south toward thenorth and it ranges between 2 and 20 cm. The amplitude

ARTICLE IN PRESS

Fig. 3. Comparison between observed and simulated amplitudes and phases. Upper panels (a–b): group of semidiurnal constituents; lower panels (c–d): group of diurnal

constituents.

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706698

distributions for other harmonics (their average values are shownin Table 4) show qualitatively the same pattern of the principalconstituents, depending on the group (semidiurnal or diurnal)they belong to. A straightforward view of the model to data fit forthe harmonics of semidiurnal and diurnal groups of the wholedomain is showed in Fig. 3. The amplitudes (Fig. 3a and c) of bothgroups of constituents correctly line up along the equality line,with a lower accuracy for the diurnal component O1, whichappears to be slightly biased. The phase of the diurnalcomponents (Fig. 3d) shows a tight correspondence to the idealfit line, while for the semidiurnal data (Fig. 3b) the dispersion ismore pronounced.

The simulated current velocities were analyzed using thesame least square method applied to the water level time series,in order to recognize the contribution of each tidal constituent tothe current variability. In particular, for each component,we compared simulated and observed major and minor axiscomponents and inclination of the tidal ellipse. In Fig. 4 acomparison between the model and observational sets for thenorthern Adriatic area (stations 4–16 in Fig. 1) is depicted. Agood agreement between observed and simulated semi-major

axis can be noted for both M2 and K1 velocities, the ab-solute mean differences being 0.8 and 0.6 cm/s, respectively.A pronounced scatter in the tidal ellipse inclination appears,instead, for both tidal components, with absolute differencesup to 601.

The degree of realism of the tidal dynamics simulated at thelagoon inlets was assessed though an accurate comparisonbetween simulated velocities and ADCP velocities collectedduring 3 long term surveys carried out by Gacic et al. (2004). InTable 5 the tidal velocity components, phase, and inclination ofthe tidal ellipse are reported, computed by mean of a harmonicanalysis of the simulated velocity time series in the three inlets(station 1–3 in Fig. 2). The agreement of the simulated values withthe corresponding observed values is remarkably good, althoughthe simulated major axis component at the station of Chioggiawere commonly overestimated. A good agreement was alsoachieved between simulated and observed phases for eachconstituent of the tidal velocity, with minor differences in thevalues of S2, K2, and O1 harmonics. As one can clearly see, thevalues obtained for the ellipse inclination indicate a strongpolarisation of tidal currents along the principal axis of each inlet.

ARTICLE IN PRESS

Fig. 4. Simulated vs. observed semi-major axis velocity and inclination of M2 and K1 tidal components of the velocity at stations 4–16 located in the North Adriatic Sea

(see Fig. 1).

Table 5Tidal ellipse parameters for the seven major tidal constituents from observed and

simulated data at stations 1–3 (Fig. 2).

St. 1 (Chioggia) St. 2 (Malamocco) St. 3 (Lido)

Obs. Sim. Obs. Sim. Obs. Sim.

M2

U-major (cm/s) 48.1 52.2 75.7 71.1 66.5 68.3

U-minor (cm/s) �0.1 �0.1 �0.1 0.0 �0.3 0.8

Phase (deg.) 14.6 13.4 23.1 26.4 36.1 36.9

Inclination (deg.) – 120.8 – 12.6 – 33.5

S2

U-major (cm/s) 28.3 33.7 44.6 46.1 37.9 43.5

U-minor (cm/s) 0.0 �0.1 �0.1 0.0 �0.3 0.5

Phase (deg.) 25.2 14.3 34.2 26.5 45.5 37.2

Inclination (deg.) – 120.8 – 12.5 – 33.5

N2

U-major (cm/s) 7.4 8.8 12.1 12.1 11.2 11.5

U-minor (cm/s) 0.0 0.0 0.0 0.0 0.2 0.1

Phase (deg.) 14.9 16.5 24.4 28.4 35.6 38.1

Inclination (deg.) – 120.8 – 12.5 – 33.5

K2

U-major (cm/s) 8.4 11.0 14.5 18.3 12.3 19.1

U-minor (cm/s) 0.0 0.0 0.0 0.0 0.0 0.2

Phase (deg.) 20.6 23.2 27.4 36.0 40.0 45.9

Inclination (deg.) – 120.8 – 12.5 – 33.4

K1

U-major (cm/s) 17.8 20.9 28.0 31.7 27.3 30.2

U-minor (cm/s) 0.0 0.0 0.1 0.0 0.0 0.2

Phase (deg.) 167.0 169.9 169.6 176.8 179.8 182.5

Inclination – 120.7 – 12.5 – 33.6

O1

U-major (cm/s) 4.9 7.3 6.9 10.3 7.0 10.9

U-minor (cm/s) 0.0 0.0 0.1 0.0 0.4 0.1

Phase (deg.) 166.1 150.6 163.8 154.8 178.7 160.2

Inclination (deg.) – 120.6 – 12.3 – 33.8

P1

U-major (cm/s) 6.4 6.3 9.1 8.7 8.8 9.1

U-minor (cm/s) �0.1 0.0 0.1 0.0 0.1 0.0

Phase (deg.) 173.5 168.6 170.8 172.6 180.1 179.3

Inclination (deg.) – 120.6 – 12.2 – 33.6

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 699

4. Discussion

The application of both tidal and meteorological forcingsallows us to reproduce correctly the tidal characteristics of theintegrated domain constituted by the Adriatic Sea and the Lagoon

of Venice. A comparison between the results obtained in thecalibration and in the pure tidal experiments was carried out toevaluate the influence of wind stress and atmospheric pressuregradient on the computation of the tidal harmonics, derived fromthe simulated sea water level time series. In semi-enclosed basinslike the Adriatic Sea, the wind forcing may generate freeoscillations with frequencies close to those of the tidal constitu-ents (see, e.g., Vilibic, 2006), which, consequently, may lead tomisleading results in the computation of the tidal harmonics. Thedifferences emerged in this comparison, however, were small.This ensures, a posteriori, the reliability of our model calibration(which was performed including the atmospheric forcing) and therobustness of the adopted method used to compute the tidalharmonic constituents. Moreover, the calibrated bottom rough-ness length coefficient (see Table 1) was of the same magnitude ofvalues reported in previous studies of the northern Adriatic basin(see, e.g., Zavatarelli and Pinardi, 2003).

The agreement between observed and simulated amplitudesand phases of the tidal harmonics for the whole basin wasgenerally good. Note, however, that the tidal characteristics of theAdriatic Sea were reproduced less satisfactorily (see Table 4a),owing to the lower accuracy obtained for the stations locatedalong the eastern Adriatic coast and, secondly, in the approachesto the open boundary. As reported by Janekovic and Kuzmic(2005), errors in the model results can be related both to a coarseresolution used to represent complex morphological structureslike, e.g., those existing in the eastern Adriatic coast, and even tothe uncertainty in the data. A misfit between observations andmodel outputs can also be related to the non-tidal circulation ofthe Mediterranean Sea propagating toward the Adriatic basin(Tsimplis et al., 1995). This aspect, especially manifested in thebaroclinic gradient emerging when the Modified LevantineIntermediate Water enters the Adriatic Sea through the OtrantoStrait and recirculates within the southern and central Adriaticbasins (Socal et al., 2009), has not been considered here and, veryprobably, it could improve the reproduction of the simulated seawater levels if considered in the model setup. Nevertheless, theaccuracy obtained in our simulations was satisfactory andcomparable with those reported in previous studies carriedout through the application of different hydrodynamic models(Cushman-Roisin and Naimie, 2002; Umgiesser et al., 2004;Janekovic and Kuzmic, 2005; Bellafiore et al., 2008). Moreover,the calibrated model reproduced accurately the behaviour ofthe most energetic constituents of the tide in the Adriatic Sea(Fig. 5), namely M2 and K1. The wave nature of tidal movements in

ARTICLE IN PRESS

Fig. 5. Co-tidal charts of the simulated M2 (left) and K1 (right) tidal constituents for the Adriatic Sea. Amplitudes are in cm and phases are in degrees.

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706700

Adriatic Sea has been shown through an analytical model of theM2 tide by Hendershott and Speranza (1971) as a solution of theproblem on long wave reflection in the end of elongated basinallowing for Coriolis force (the so-called Taylor’s problem). Theoccurrence of a semidiurnal tidal amphidrome in the Adriatic Seahas been explained by sufficiently weak energy losses in northernpart of the basin. As regards the diurnal tide, the extensionof the Adriatic Sea appears to be larger than a quarter of diurnaltidal wave length and, thus, the formation of apparentamphidrome within the basin does not occur. These featuresof tidal wave dynamics have found a confirmation in thesubsequent works and in the present study. The co-tidal chartsderived from the model clearly show the evolution of thesemidiurnal tide, resembling a progressive Kelvin wave, whichcyclonically rotates around an amphidromic point positioned onthe Ancona–Sibenik line (Mosetti, 1987), and the nature of thediurnal tide, associated to a topographic wave progressing cross-wise from the eastern to the western coast (Cushman-Roisin et al.,2001).

Beside these well-known characteristics, the simulated spatialdistribution of the K1 amplitude and phase is characterized by thepresence of an amplitude minimum located in the approaches toVieste (st. 4 in Fig. 1) and a degenerated amphidromic point nearOtranto (st. 1 in Fig. 1), whereas, in the southern Adriatic basin,M2 semidiurnal component shows higher amplitudes than thosereported in the work of Cushman-Roisin and Naimie (2002). Onthe other hand, both M2 and K1 co-tidal charts are in rather goodagreement with those presented by Janekovic and Kuzmic (2005),who used a refined data assimilation procedure to simulate thetidally induced circulation in the Adriatic Sea. Note that ouranalysis refers to a much larger number of tide gauge observa-tions than in previous works (see Figs. 1 and 2). In addition aclassification of the tide type is usually carried out through theestimation of the tidal factor F (F ¼ ðAK1

þAO1Þ=ðAM2

þAS2Þ, where

AX is the respective amplitude value) which allows one toestimate the relative importance of individual diurnal andsemidiurnal tidal constituents. Calculation of F reveals (notshown) that the mixed semidiurnal type prevails in the Adriatic

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 701

Sea, although the tide maintains a diurnal character in proximityof the semidiurnal amphidromic point.

The comparison between simulated and observed depthaveraged tidal velocities showed that the model reliably repro-duces the current dynamics in the integrated domain. Differencesobtained for the northern Adriatic set of stations were comparableto those reported in earlier works (Cushman-Roisin and Naimie,2002; Janekovic et al., 2003; Bellafiore et al., 2008). Simulatedtidal ellipses correspond only partially to observed ones. This is aproblem which seems to be common to all numerical modelsimplemented for the simulation of the North Adriatic basin. Thecauses of such disagreement are possibly in the presence of localbathymetric features that cannot be completely captured even bya very high resolution numerical model (see e.g., Cushman-Roisinand Naimie, 2002; Janekovic and Kuzmic, 2005). An additionalsource of uncertainty in the reproduction of coastal currentsregards the absence of baroclinic effects in the simulatedcirculation. For instance, the fresh water discharges of the Poand other Italian rivers are known to remarkably modulate thespatial pattern and the intensity of the stratification during bothsummer and winter and, thus, to contribute to shape the currentfields along the western Adriatic coast. Such fields appear to benoticeably influenced by the seasonal alternation of regions of

Fig. 6. Simulated (continuous lines) and observed (crosses) water level referring to 6 sta

and 2 for their locations).

enhanced and reduced vertical stability (see Socal et al., 2009).However, the good precision achieved in the simulation of thetidal currents at the inlets of the Lagoon of Venice (see Table 5)confirms the validity of the selected modelling approach: thisresult would have been difficult to achieve in a disjoint simulationof the two domains.

4.1. Observed water levels and small-scale circulation patterns

Results presented in the previous section showed that themodel efficiently reproduces the temporal and spatial evolution oftwo relevant variables, namely the tidal elevation and the depthaveraged currents.

On this basis, the calibrated hydrodynamic model wasaddressed to the reproduction of observed tidally and atmo-spherically induced sea water levels characterizing the northernarea of the composite system and to the investigation of thesmall-scale circulation features occurring in the zone between theadjoining basins. Measured and simulated sea water levels werecompared, in order to provide an estimation of the performance ofthe model in reproducing the sea surface evolution at short timescales. Measurements come from the hourly sampling of two

tions located in the North Adriatic Sea and inside the Lagoon of Venice (see Figs. 1

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706702

independent monitoring networks: the RMN system of theNational Agency for the Environmental Protection and the surveysof the Mareographic Centre of the Venice Municipality. In Fig. 6the hourly water level time series simulated and measured at 6tidal stations within the northern Adriatic Sea and the Lagoon ofVenice during the late summer 2004 (see their location in Figs. 1and 2), are depicted. The model correctly reproduces observedphase and amplitude, thus preserving the characteristics oftypical mixed tidal regime encountered in the northern Adriaticarea (see, e.g., Polli, 1960).

As far as the near shore circulation patterns are concerned, thesimulated velocity fields in the approaches to the central inlet ofthe Lagoon of Venice were compared to surface velocity fieldsobserved in the area by means of a high frequency radar having aspatial resolution of �750 m (Kovacevic et al., 2004). In particularwe focused on a descriptive comparison between observed andsimulated velocity patterns for a Sirocco and Bora wind episodes,whose average wind speed were �5 and �15 m/s, respectively.Besides the discrepancies which may be due to other processeslike, e.g., stratification, it must be noted that a precise, quantita-

Fig. 7. Vector maps of current (thin line) as observed by Kovacevic et al. (2004) (pane

Sirocco wind and during an episode of strong Bora wind. The prevailing wind direction

tive comparison between observations and simulated depth-averaged velocities emerging in the presence of the combinedtidal and wind forcing cannot be obtained in principle. However, aqualitative comparison is still possible: it shows that in both windscenarios considered, the model was able to capture differentaspects of the observed velocity patterns, both in intensity anddirection (Fig. 7).

The good agreement between observations and simulationssuggests that also the dynamics in the very approaches to theinlet, produced by our numerical model, correspond to robustfeatures occurring during Bora and Sirocco episodes. During thesimulated Bora episode, for instance, mean current velocities ofabout 15 cm/s occurred, which contributed to produce a peculiarlocal circulation, characterized by a noticeable southward dis-placement of the water plume exiting from the lagoon inlet(Fig. 7d). This, on its turn, led to the formation of a strong, ratherpersistent eddy located between the inlet dike and the shoreline.The simulated Sirocco episode, instead, was characterized bysubstantially lower mean water velocities (�5 cm/s) and, conse-quently, to a less pronounced modifications of the mean local flow

ls a and b) and as reproduced by the model (panels c and d) during an episode of

is also indicated (bold line).

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 703

pattern (Fig. 7c). Moreover, vortical structures on the sea side ofthe inlets similar to those observed by Kovacevic et al. (2004)using the HF radar are clearly visible in the simulated fields (inthis region the model horizontal mean spatial resolution is circa250 m) during both events.

Numerical experiments carried out under idealized conditionsby D’Alpaos and Defina (2007) indicate that the tidally inducedflow is responsible for the formation of local transitional eddies incorrespondence of the inlet mouths. As follows from the theoryand also from experimental data, outflows from a sharp channelend may form a dipole of two large counter-rotating eddies byadvection of vorticity created at the boundaries (Wells and vanHeijst, 2003). If the dipole structure is sufficiently long-lived, itcan self-propagate away from the channel or may be involved byinflow backward into the channel when the flow is periodic. Thedipoles propagation away from the source region was found tosignificantly enhance the efficiency of tidal exchange between abay and a sea connected by a narrow channel. The prevision of its

Fig. 8. Flow velocity distributions simulated at the inlet near Chioggia (see Fig. 2) durin

below 1 m/s. The correspondent tidal phase is indicated in the upper-left box.

occurrence was based on the relatively simple criteriaW/UTo0.13, where U is the tidal velocity of the channel jet,T the tidal period, and W the width of the channel. The presentmodel was thus used to investigate the vortex genesis andevolution at the southern inlet near the town of Chioggia(see Fig. 2) throughout a tidal cycle. Two vortices were simulated,which originated outside the lagoon during ebb (see Fig. 8a, blackarrows), while a sharp current plume transported water from thelagoon toward the open sea (see Fig. 8a, white arrow). Then,during the maximum tidal inflow, a smaller vortex was generatedinside the lagoon, in the southern area of the inlet (see Fig. 8b,black arrow). For typical values W, U, and T (550 m, 0.5 m/s, and�12 h, correspondingly) in Chioggia inlet, the criterion proposedby Wells and van Heijst (2003) is satisfied. Accordingly, despitethe fact that bed friction can lead to a noticeable decay of vorticitynear the outlet, the dipole vortical structure generated during theprevious tidal phase is still visible on the outer part of the inletand it tends to escape the inlet region and to translate toward the

g ebb tide, (a) and flood tide (b). During the simulation period, the wind speed was

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706704

open sea (see Fig. 8b, white arrows). It is anticipated that thischaracteristic of the tidally induced circulation at the interface ofthe adjoining basins, which can exert a profound influence, e.g., onlocal sedimentation and nutrient segregation, represents a keyprocess in the study of biogeochemical state variables of theintegrated system Lagoon of Venice–North Adriatic basin.

4.2. Residual sea level in the northern Adriatic basin

According to the early studies of Finizio et al. (1972) and Godinand Trotti (1975), the concomitant presence over the North Adriaticbasin of long-lasting low air-pressure and strong Sirocco winds is ableto induce basin-wide oscillations, or seiches, whose characteristicfrequencies are found to be close to those of the major tidalconstituents, being between 21 and 23 h and around 11 h. Particularfeatures of spatial and temporal evolution of the Adriatic non-tidaloscillations were discussed in a number of subsequent papers(Cerovecki et al., 1997; Vilibic et al., 1998; Tomasin and Pirazzoli,1999; Pasaric and Orlic, 2001; Sobey, 2003). More recently, throughthe analysis of long term series of residual observed sea levels for twostations located in the eastern Adriatic coast, Vilibic (2006) found theperiods of the first and of the second modes of the Adriatic seiches tobe 21.5 and 11 h, respectively.

A statistical procedure based on the wavelet transform(Daubechies, 1990) is here adopted to investigate the spectralcomposition of both simulated and observed residual sea levels. Awell known limit of the Fourier transform is that it only providestime averaged results as the signal is assumed to be stationary.Conversely, the application of the wavelet analysis allows to

Fig. 9. The wavelet power spectrum of observed (a) and simulated (b) residual water

normalized by their variance). The shaded contour is composed of regular intervals sp

within the 95% confidence interval for a red-noise process with a lag-1 coefficient of

references to colour in this figure legend, the reader is referred to the web version of

decompose a one dimensional time series into a two dimensionaltime-frequency image and thus it provides a synthetic represen-tation of the variable structure of the signal and the changingcontribution of its components (for further details see Torrenceand Compo, 1998). The main advantage is then represented bythe possibility of localising events of interest at their exacttemporal location when dealing with non-stationary processes(see, e.g., Drago and Boxall, 2002; Zanchettin et al., 2009).Residual levels simulated for a model point located near theCNR Tower (st. 70 in Fig. 1) were computed by subtracting fromthe sea water levels -simulated using both tidal and meteorolo-gical forcing- the corresponding sea water levels simulated usingonly tidal forcing (see Section 2.4 for further details). The sameprocedure was used to compute the residual sea water levelsmeasured at the CNR Tower by the Mareographic Centre of theVenice Municipality.

Fig. 9 displays the wavelet power spectrum of the hourly valuesof both observed and simulated residual levels, obtained using aMorlet wavelet (data were normalized by their variance). The redcontour encloses regions that lie in the 95% confidence interval for ared-noise process with a lag-1 coefficient of 0.7. During autumn andwinter seasons, when atmospheric perturbations are more intense(see Artegiani et al, 1997a), the variance close to the period of thefirst mode is significantly above the confidence interval for bothtime series. On the other hand, the characteristic period of thesecond mode cannot be clearly recognized in the power spectrumobtained from observations, while the one derived from thenumerical experiment reveals significant regions on this band incorrespondence of the winter period. Compared to the analysis ofthe observed data, the wavelet spectrum of the simulated residual

levels at the station 70 (CNR Tower) obtained using a Morlet wavelet (data are

aced with a normalized variance equal to 2. The red line bounds regions that lie

0.7. The cone of influence is also indicated (blue lines). (For interpretation of the

this article.)

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706 705

levels lack of those components with periods lower than 6 h and,thus, a larger weight is assigned to the variance of the spectrumassociated to the first and to the second modes of the Adriatic basin.As one can see in Fig. 9b, significant regions exist that largely extendaround the diurnal period. This fact may be explained by the effectof sea level variations induced by local meteorological phenomenalike land-sea breezes or by the effect of residual tides. We areconvinced that our meteorological forcing, characterized by a coarseresolution both in space and time (see Section 2.2), is not adequateto generate diurnal wind-induced variability. In fact, only a very highresolution meteorological model can be able to partly capture land-sea breeze induced dynamics (see, e.g., Pullen et al. (2003)).Probably, these signals can be explained as manifestations ofresidual spring-neap tidal oscillations possibly interacting withmeteorologically induced sea level oscillations at diurnal andsemidiurnal time scales.

5. Conclusion

In this paper, the implementation of a boundary conformingcoordinates, finite difference, 2D hydrodynamic model for thecomposite domain constituted by the Adriatic Sea and the Lagoonof Venice was presented. Numerical simulations were carried outunder realistic conditions, including tidal and meteorological forcings.A wide ensemble of recent experimental datasets was used toreproduce the bottom geometry of the physical system and calibratethe model. For validation, the model results were compared to tidegauges derived harmonic constants collected in stations located in thewhole domain and currentmeter data.

Simulation results for the diurnal and semidiurnal tidal constitu-ents were satisfactory, with mean errors in amplitude and phaserarely exceeding 1 cm and 151 for the Adriatic Sea, and 0.5 cm and 91for the Lagoon of Venice, respectively. Furthermore, the depthaveraged velocities agreed quite good with the velocities measuredby mean of a current meter, especially in the case of the velocitiesrecorded at the interface between the lagoon and the open sea.

Finally, the model was addressed to the reproduction of theresidual water levels collected in the northern part of the systemand to the investigation small-scale circulation patterns observedin the marine region near the Lagoon of Venice. The comparisonbetween simulated and observed velocity fields in front of theVenetian littoral evidenced a reliable reproduction of the coastalcirculation pattern under the influence of the tide and mostimportant winds (Bora and Sirocco) characterized by the forma-tion of vortical structures near the inlets. The analysis of thesimulated and observed sea water levels clearly indicated thecapability of the model to capture atmospheric and residual tidecontributions to the hydrological fields in the composite domain.The elucidation of these features requires a deeper investigationof the interactions between the tide and the meteorologicallyinduced oscillations, as the residual water levels cannot beconsidered completely detided and, thus, the contribution ofresidual tidal effects may become relevant in the analysis ofobserved as well as simulated sea water levels.

The results achieved through the numerical simulationsperformed with the barotropic model allowed to consistentlydescribe the tidal and atmospherically induced hydrodynamicfeatures of the composite system Adriatic Sea–Lagoon of Venice.In particular, the model realistically reproduced the local circula-tion patterns which characterize the exchanges between thelagoon and the open sea and, thus, it can be used to simulatesuccessfully the transport and dispersion processes whichinfluences the spatial and temporal distribution of nutrientsand/or pollutants, which play a crucial role in the simulation ofbiogeochemical processes in coastal areas.

References

Akima, H., 1984. On estimating partial derivatives for bivariate interpolation ofscattered data. Rocky Mountain Journal of Mathematics 14 (1), 41–52.

Androsov, A.A., Klevanny, K.A., Salusti, E., Voltzinger, N.E., 1995. Open boundaryconditions for horizontal 2-D curvilinear-grid long-wave dynamics of a strait.Advances in Water Research 18, 267–276.

Androsov, A.A., Kagan, B.A., Romanenkov, D.A., Voltzinger, N.E., 2002. Numericalmodelling of barotropic tidal dynamics in the strait of Messina. Advances inWater Research 25, 401–415.

Artegiani, A., Bregant, D., Pascini, E., Pinardi, N., Raicich, F., Russo, A., 1997a. TheAdriatic Sea general circulation. Part I: air–sea interactions and water massstructure. Journal of Physical Oceanography 27, 1492–1514.

Artegiani, A., Bregant, D., Pascini, E., Pinardi, N., Raicich, F., Russo, A., 1997b. TheAdriatic Sea general circulation. Part II: baroclinic circulation structure. Journalof Physical Oceanography 27, 1515–1532.

Balzano, A., 1998. Evaluation of methods for numerical simulation of wetting anddrying in shallow water flow models. Coastal Engineering 34, 83–107.

Bates, P.D., Lane, S.N., Ferguson, R.I., 2005. Computational Fluid Dynamics:Applications in Environmental Hydraulics. John Wiley & Sons 531pp.

Bellafiore, D., Umgiesser, G., Cucco, A., 2008. Modeling the water exchangesbetween the Venice Lagoon and the Adriatic Sea. Ocean Dynamics 58,397–413.

Bergamasco, A., Carniel, S., Pastres, R., Pecenik, G., 1998. A unified approach to themodelling of the Venice Lagoon–Adriatic Sea ecosystem. Estuarine, Coastal andShelf Science 46 (4), 483–492.

Blumberg, A.F., Mellor, G.L., 1987. A description of a three-dimensional coastalocean circulation model. In: Three Dimensional Coastal Ocean Models, Ed.Heaps, AGU, pp. 1–16.

Book, J.W., Perkins, H.T., Cavaleri, L., Doyle, J.D., Pullen, J.D., 2005. ADCPobservations of the western Adriatic slope current during winter 2001.Progress in Oceanography 66, 270–286.

Casulli, V., Zanolli, P., 2002. Semi-implicit numerical modelling of nonhydrostaticfree-surface flows for environmental problems. Mathematical and ComputerModelling 36, 1131–1149.

Cerovecki, I., Orlic, M., Hendershott, M.C., 1997. Adriatic seiche decay and energyloss to the Mediterranean. Deep-Sea Research I 44, 2007–2029.

Chavanne, C., Janekovic, I., Flament, P., Poulain, P.-M., Kuzmic, M., Gurgel, K.W.,2007. Tidal currents in the northwestern Adriatic: high-frequency radioobservations and numerical model predictions. Journal of GeophysicalResearch 112, C03S21, doi:10.1029/2006JC003523.

Cushman-Roisin, B., Naimie, C.E., 2002. A 3D finite-element model of the Adriatictides. Journal of Marine Systems 37, 279–297.

Cushman-Roisin, B., Gacic, M., Poulain, P.M., Artegiani, A., 2001. PhysicalOceanography of the Adriatic Sea. Kluver Academic Publishers 303 pp.

D’Alpaos, L., Defina, A., 2007. Mathematical modeling of tidal hydrodynamics inshallow lagoons: a review of open issues and applications to the Venicelagoon. Computational Geosciences 33 (4), 476–496.

Daubechies, I., 1990. The wavelet transform, time-frequency localization andsignal analysis. IEEE Transactions on Information Theory 36, 961–1005.

Defina, A., 2000. Two-dimensional shallow flow equations for partially dry areas.Water Resources Research 36 (11), 3251–3264.

Drago, A.F., Boxall, S.R., 2002. Use of the wavelet transform on hydro-meteorological data. Physics and Chemistry of the Earth 27, 1387–1399.

Egbert, G.D., Erofeeva, S.Y., 2002. Efficient inverse modeling of barotropicocean tides. Journal of Atmospheric and Oceanic Technology 19 (2),183–204.

Ferla, M., Cordella, M., Michielli, L., Rusconi, A., 2007. Long-term variations on sealevel and tidal regime in the lagoon of Venice. Estuarine, Coastal and ShelfScience 75 (1–2), 214–222.

Finizio, C., Palmieri, S., Riccucci, A., 1972. A numerical model of the Adriatic for theprediction of high tides at Venice. Quarterly Journal of Royal MeteorologicalSociety 98, 86–104.

Foreman, M.G.G., 1977. Manual for tidal heights analysis and prediction. PacificMarine Science Report, 77–10, 66pp.

Gacic, M., Mancero Mosqueda, I., Kovacevic, V., Mazzoldi, A., Cardin, V., Arena, F.,Gelsi, G., 2004. Temporal variations of water flow between the Venetian lagoonand the open sea. Journal of Marine Systems 51 (1–4), 33–47.

Godin, G., Trotti, L., 1975. Trieste-Water Levels 1952–1971: A Study of the Tide,Mean Level, and Seiche Activity, vol. 28. Miscellaneous Special Publication24 pp.

Geyer, W.R., Signell, R.P., 1990. Measurements of tidal flow around a head-landwith a shipboard acoustic Doppler current profiler. Journal of GeophysicalResearch 95, 3189–3197.

Geyer, W.R., Signell, R.P., 1992. A reassessment of tidal dispersion. Estuaries 15 (2),97–108.

Guyondet, T., Koutitonsky, V.G., 2006. Tidal and residual circulations in coupledrestricted and leaky lagoons. Estuarine, Coastal and Shelf Science 77 (3),396–408.

Hellerman, S., 1967. An updated estimate of the wind stress on the world ocean.Monthly Weather Review 95 (9), 607–626.

Hendershott, M.C., Speranza, A., 1971. Co-oscillating tides in long, narrow bays:the Taylor problem revisited. Deep-Sea Research 18, 959–980.

Hydrographic Institute of the Republic of Croatia (HIR), 2004. Tides Tables, AdriaticSea–East Coast in the period 2003–2007. Tiskara HHI, 5 volumes, 50 pp.

ARTICLE IN PRESS

T. Lovato et al. / Continental Shelf Research 30 (2010) 692–706706

Imasato, N., 1983. What is tide-induced residual current? Journal of PhysicalOceanography 13, 1307–1317.

Janekovic, I., Kuzmic, M., 2005. Numerical simulation of the Adriatic Sea principaltidal constituents. Annales Geophysicae 23, 3207–3218.

Janekovic, I., Bobanovic, J., Kuzmic, M., 2003. The Adriatic Sea M2 and K1 tides by 3Dmodel and data assimilation. Estuarine, Coastal and Shelf Science 57, 873–885.

Kjerfve, B., 1994. Coastal Lagoon Processes. Elsevier, Amsterdam 577 pp.Kovacevic, V., Gacic, M., Mancero Mosquera, I., Mazzoldi, A., Marinetti, S., 2004. HF

radar observations in the northern Adriatic: surface current field in front of theVenetian Lagoon. Journal of Marine Systems 51, 95–122.

Leffler, K.E., Jay, D.A., 2009. Enhancing tidal harmonic analysis: robust (hybrid L1/L2) solutions. Continental Shelf Research 15, 78–88.

Lionello, P., Sanna, A., Elvini, E., Mufato, R., 2006. A data assimilation procedure foroperational prediction of storm surge in the northern Adriatic Sea. ContinentalShelf Research 26, 539–553.

Magistrato alle Acque di Venezia (MAV), 2005. Carta digitale della Laguna diVenezia in scala 1:5000. Consorzio Venezia Nuova—Servizio Informativo.

Malacic, V., Viezzoli, D., Cushman-Roisin, B., 2000. Tidal dynamics in the northernAdriatic Sea. Journal of Geophysical Research 105 (C11), 26,265–26,268.

Malanotte Rizzoli, P., Bergamasco, A., 1983. The dynamics of the coastal region ofthe Northen Adriatic Sea. Journal of Physical Oceanography 13, 1105–1130.

Martin, P.J., Book, J.W., Doyle, J.D., 2006. Simulation of the northern Adriaticcirculation during winter 2003. Journal of Geophysical Research 111, C03S12,doi:10.1029/2006JC003511.

Mosetti, R., 1987. Determination of the current structure of the M2 tidalcomponent in the northern Adriatic by applying the rotary analysis to theTaylor problem. Bollettino di Oceanografia Teorica e Applicata 4, 165–172.

Orlic, M., Kuzmic, M., Pasaric, Z., 1994. Response of the Adriatic Sea to the bora andsirocco forcing. Continental Shelf Research 14 (1), 91–116.

Pasaric, M., Orlic, M., 2001. Long-term meteorological preconditioning of the NorthAdriatic coastal floods. Continental Shelf Research 21, 263–278.

Persson, A., 2003. User Guide to ECMWF forecast products. ECMWF MeteorologicalBulletin M3.2, pp. 123.

Petaccia, S., Serravall, R., Pellicano, F., 2006. Improved method of sea levelforecasting at Venice (Northern Adriatic Sea). Communication in NonlinearScience and Numerical Simulation 11, 281–296.

Polli, S., 1960. La Propagazione delle Maree nell’Adriatico. Atti del Convegnodell’Associazione Geofisica Italiana, Roma 1959, 11pp.

Pullen, J., Doyle, J.D., Hodur, R., Ogston, A., Book, J.W., Perkins, H., Signell, R., 2003.Coupled ocean-atmosphere nested modeling of the Adriatic Sea during winterand spring 2001. Journal of Geophysical Research 108, C10, doi:10.1029/2003JC001780.

Pullen, J., Doyle, J.D., Haack, T., Dorman, C., Signell, R.P., Lee, C.M., 2007. Bora eventvariability and the role of air–sea feedback. Journal of Geophysical Research112, C03S18, doi:10.1029/2006JC003726.

Romanenkov, D.A., Androsov, A.A., Voltzinger, N.E., 2001. Comparison of forms ofthe viscous shallow-water equations in the boundary-fitted coordinates.Ocean Modelling 3, 193–216.

Sobey, R.J., 2003. Seiche modes of elongated natural basins. Coastal EngineeringJournal 45 (3), 421–438.

Socal, G., Acri, F., Bastianini, M., Aubry, F.B., Bianchi, F., Cassin, D., Coppola, J., DeLazzari, A., Bandelj, V., Cossarini, G., Solidoro, C., 2009. Hydrological andbiogeochemical features of the Northern Adriatic Sea in the period 2003–2006.Marine Ecology—an Evolutionary Perspective 29, 449–468.

Tomasin, A., Pirazzoli, P.A., 1999. The seiches in the Adriatic Sea, Atti IstitutoVeneto di Scienze, Lettere, Arti, Venezia 157, 299–316.

Torrence, C., Compo, G.P., 1998. A practical guide to wavelet analysis. Bulletin ofthe American Meteorological Society 79 (1), 61–78.

Tsimplis, M.N., Proctor, R., Flather, R.A., 1995. A two-dimensional tidal model forthe Mediterranean Sea. Journal of Geophysical Research 100, 223–239.

Umgiesser, G., 1997. Modeling the Venice Lagoon. International Journal of SaltLake Research 6, 175–199.

Umgiesser, G., Melaku Canu, D., Cucco, A., Solidoro, C., 2004. A finite elementmodel for the Venice Lagoon. Development, set up, calibration and validation.Journal of Marine Systems 51, 123–145.

Van Kan, J., 1986. A second-order accurate pressure-correction scheme for viscousincompressible flow. SIAM Journal on Scientific and Statistical Computing 7(3), 870–891.

Vilibic, I., Leder, N., Smircic, A., 1998. Forced and free response of the Adriatic Sealevel. Nuovo Cimento della Societa Italiana di Fisica C4, 439–451.

Vilibic, I., 2006. The role of the fundamental seiche in the Adriatic coastal floods.Continental Shelf Research 26, 206–216.

Ward, J.V., 1998. Riverine landscapes: biodiversity patterns, disturbance regimes,and aquatic conservation. Biological Conservation 83 (3), 269–278.

Wells, M.G., van Heijst, G.J.F., 2003. A model of tidal flushing of an estuary bydipole formation. Dynamics of Atmospheres and Oceans 37 (3), 223–244.

Zaldivar, J.M., Cattaneo, E., Plus, M., Murray, C.N., Giordani, G., Viaroli, P., 2003.Long-term simulation of main biogeochemical events in a coastal lagoon:Sacca Di Goro (Northern Adriatic Coast, Italy). Continental Shelf Research 23,1847–1875.

Zanchettin, D., Rubino, A., Traverso, P., Tomasino, M., 2009. Teleconnections forceinterannual-to-decadal tidal variability in the Lagoon of Venice (northernAdriatic). Journal of Geophysical Research 114, D07106, doi:10.1029/2008JD011485.

Zavatarelli, M., Pinardi, N., Kourafalou, V.H., Maggiore, A., 2002. Diagnostic andprognostic model studies of the Adriatic Sea general circulation: seasonalvariability. Journal of Geophysical Research 107 (C1), 3004, doi:10.1029/2000JC000210.

Zavatarelli, M., Pinardi, N., 2003. The Adriatic Sea modelling system: a nestedapproach. Annales Geophysicae 21, 345–364.