continental shelf research - university of · pdf filecalculations of the simpson–hunter...

Continental Shelf Research ] (]]]]) ]]]–]]]

Contents lists available at ScienceDirect

Continental Shelf Research

0278-43

doi:10.1

n Corr

E-m

Pleasenvir

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

Application of 3D hydrodynamic and particle tracking models for betterenvironmental management of finfish culture

Juan Moreno Navas n, Trevor C. Telfer, Lindsay G. Ross

Institute of Aquaculture, University of Stirling, FK94LA Stirling, UK

a r t i c l e i n f o

Article history:

Received 24 September 2009

Received in revised form

3 January 2011

Accepted 7 January 2011

Keywords:

3D hydrodynamic models

Fish-cage waste dispersion

Geographical Information Systems

Environmental management

43/$ - see front matter & 2011 Elsevier Ltd. A

016/j.csr.2011.01.001

esponding author. Tel.: +44 1786 767882; fa

ail address: [email protected] (J. Moreno

e cite this article as: Moreno Navonmental management of finfish cu

a b s t r a c t

Hydrographic conditions, and particularly current speeds, have a strong influence on the management

of fish cage culture. These hydrodynamic conditions can be used to predict particle movement within

the water column and the results used to optimise environmental conditions for effective site selection,

setting of environmental quality standards, waste dispersion, and potential disease transfer. To this

end, a 3D hydrodynamic model, MOHID, has been coupled to a particle tracking model to study the

effects of mean current speed, quiescent water periods and bulk water circulation in Mulroy Bay, Co.

Donegal Ireland, an Irish fjard (shallow fjordic system) important to the aquaculture industry. A

Lagangrian method simulated the instantaneous release of ‘‘particles’’ emulating discharge from finfish

cages to show the behaviour of waste in terms of water circulation and water exchange. The 3D spatial

models were used to identify areas of mixed and stratified water using a version of the Simpson–

Hunter criteria, and to use this in conjunction with models of current flow for appropriate site selection

for salmon aquaculture.

The modelled outcomes for stratification were in good agreement with the direct measurements of

water column stratification based on observed density profiles. Calculations of the Simpson–Hunter

tidal parameter indicated that most of Mulroy Bay was potentially stratified with a well mixed region

over the shallow channels where the water is faster flowing. The fjard was characterised by areas of

both very low and high mean current speeds, with some areas having long periods of quiescent water.

The residual current and the particle tracking animations created through the models revealed an

anticlockwise eddy that may influence waste dispersion and potential for disease transfer, among

salmon cages and which ensures that the retention time of waste substances from cages is extended.

The hydrodynamic model results were incorporated into the ArcViewTM GIS system for visualisation

and interrogation of results and to allow effective holistic environmental management and site

selection for fish cage aquaculture.

& 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Hydrographic conditions have an important influence onbiological and physical processes related to marine aquaculture,including settlement and transport of larval molluscs or fishectoparasites, rates of particle captured by suspension feeders,oxygen supply, influence on fish behaviour, growth and possiblyflesh quality of cultivated molluscs and fishes, and transport andaccumulation of soluble and particulate wastes released from thecultured organisms. Hydrographic measurement and use of thisdata in modelling water flow throughout coastal systems isbecoming an important management tool for marine aquaculture.

ll rights reserved.

x: +44 1786 472133.

Navas).

as, J., et al., Applicationlture. Continental Shelf Res

The production of finfish in cages causes a measurable impact onthe quality of nearby water and seabed sediments due to ammoniaexcretion, depletion of available dissolved oxygen and release offaecal material and uneaten feed (Beveridge, 2004). The most severeimpacts of marine fish cages have been associated with intensiveaquaculture operations in areas with inadequate water circulation.The use of a hydrodynamic modelling approach in aquacultureplanning regulation and monitoring was encouraged by Hendersonet al. (2001). Such hydrodynamic models have been used in differentsalmon culture studies focused on different environmental problemssuch as nutrient waste, pesticide dispersion, oxygen depletion anddispersion of ectoparasite larvae (Table 1).

The velocity profile of water currents depends on flow propertiessuch as the Reynolds number (ratio of inertia to viscosity), turbu-lence, acceleration, fluid properties and boundary characteristics.Fjordic systems are considered regions of restricted exchange (RREs),and are often areas associated with settlement of humans due to

of 3D hydrodynamic and particle tracking models for betterearch (2011), doi:10.1016/j.csr.2011.01.001

www.elsevier.com/locate/csr

dx.doi.org/10.1016/j.csr.2011.01.001

mailto:[email protected]

dx.doi.org/10.1016/j.csr.2011.01.001

dx.doi.org/10.1016/j.csr.2011.01.001

Table 1Summary of the use of hydrodynamic models in salmon culture in different countries and the environmental problems.

Author Country Software Dimensions Grid size (m) Target

Falconer and Hartnett (1993) Scotland DIVAST 2D 200 Nutrient, pesticide dispersion

Trites and Petrie (1995) Canada Unknown 2D 100 Oxygen depletion

Panchang et al. (1997) USA AWAST 2D 75 Waste dispersion

Dudley et al. (2000) USA AWAST 2D 30,75,150 Waste dispersion

Greenberg et al. (2005). Canada QUODDY 3D 500 Oxygen depletion

Murray and Gillibrand (2006) Scotland GF8 3D 100 Sea lice dispersion

Skogen et al. (2009) Norway NOWERCOM 3D 800 Eutrophication

Fig. 1. The location and bathymetry of the study area, Mulroy bay, off Ireland’s

north coast. The positions of the salmon cages are shown as white circles and the

two hydrological stations (1, 2).

J. Moreno Navas et al. / Continental Shelf Research ] (]]]]) ]]]–]]]2

fisheries, aquaculture, shipping and recreational activities. Theseactivities have the potential to enhance any nutrient loading whichmay in turn increase the risk of eutrophication (Tett et al., 2003).

Additional use of these water-based resources to expandexisting or develop new aquaculture production will increase thisrisk further if not properly managed. Monitoring of the environ-ment for this risk will only allow post hoc management afterimpact has occurred. Effective management requires predictivetools to model potential impacts and thus identify risks fromaquaculture development or to locate such activities in order tominimise their impacts and to adopt best practise for develop-ment and regulation.

The physical environment and vertical mixing within fjordicsystems has been extensively reviewed (Farmer and Freeland,1983;Freeland et al., 1980; Stigebrandt and Aure, 1989) along with wind/water interactions (Farmer, 1976; Farmer and Osborn, 1976;Dronkers and Zimmerman, 1982; Leth, 1995; Svendsen andThompson, 1978). It is accepted that fjordic systems may berepresented by a tripartite division within the water body: the nearsurface circulation zone, the intermediate zone between the surfacesand the sill, and the deep basins. In a wind-driven, three-layersystem, for example, the residual current is in the direction of thewind-stress in the surface layer, has mixed residual current direc-tions in the intermediate layer, and has a return current in the deepbasins.

The widely used tidal front model proposed by Simpson andHunter (1974) identifies well mixed and stratified regions sepa-rated by a boundary or front in which various combinations oftidal velocity and water depth produce mixed and stratifiedregions under constant heat flux. The utility of the tidal frontmodel is that the buoyancy inputs are spatially uniform. As aresult, the tides are the principal energy source for vertical mixingand the tidal front plot is a valuable indicator of the spatialdistribution of tidal mixing. The potential energy of the watercolumn is given by integrating the gravitational potential energyover the depth. The energy required to mix the water column thusincreases with increasing stratification (Lee et al., 2005), with thepotential energy contributors being heating or cooling across thewater surface, the tides and the winds (Simpson and Bowers,1981). The majority of world salmon production occurs in fjordicenvironments exhibiting high stratification and the assessment ofsuch areas is best achieved using a 3D hydrodynamic model.

Geographic information systems (GIS) are used extensively forspatial modelling and management of environmental resources.They have been used for the management of fish farming in manycountries (Nath et al. 2000; Meaden and Kapetsky, 2001; Kapetskyand Aguilar-Manjarrez, 2007). Although they have considerablestrength for combining spatial analysis and process modelling forthe development of ‘‘what-if’’ decision support scenarios, they lackthe abilities to undertake detailed hydrodynamic modelling in threedimensions for estimation of the dispersal of fish farm wastes(Panchang et al., 1997). However, outputs from dispersion models,such as that developed in this study, may be incorporated into GISfor decision support modelling for aquatic resources management

Please cite this article as: Moreno Navas, J., et al., Applicationenvironmental management of finfish culture. Continental Shelf Res

and robust fish farm site selection (Perez et al., 2002; Corner et al.,2006).

The objectives of this study was to apply a 3D hydrodynamicand particle tracking model, MOHID, to predict water circulationand to map the main hydrological parameters that influencesalmonid cage culture in a complex fjardic system with substan-tial aquaculture production (Mulroy Bay, Co Donegal, Ireland) Theincorporation of the model outcome into GIS allowed furtherinvestigation of its potential for initial site selection for moreeffective management and regulation of marine fish cageaquaculture.

2. Study area

Mulroy Bay is a fjardic inlet (Fig. 1), a glacially derivedembayment in low lying land situated on the northern coast ofCo Donegal, Ireland (55 15’N, 7 45’W). Mulroy Bay can be dividedinto four main areas: the Outer Bay, Northwater, Broadwater andthe Channel. The latter is about 100–150 m wide and has threedistinct sections; the First, Second and Third Narrows. Severalareas within the Bay are licensed for Atlantic salmon farming witha production of approximately 800–900 tonnes per annum overthe last five to ten years.


dx.doi.org/10.1016/j.csr.2011.01.001

J. Moreno Navas et al. / Continental Shelf Research ] (]]]]) ]]]–]]] 3

Maximum depths within the Bay relative to chart datum are47 m in Northwater and 40 m in Broadwater, being shallowernearer the narrows and deeper in the basins (Fig. 1). Approxi-mately 62% of the total surface area has a depth of 0–10 m, while28% is between 10 and 20 m and 10% is between 20 and 47 m.

Wind speed data from 2005 indicated a mean value of 6.8 m/s (atMalin Head), predominantly from a south-westerly direction (180–2701), though on occasion wind directions were highly variable. Thetidal range varies from 3.2 to 4.2 m (neap to spring) at the Bar at themouth of the Outer Bay to 1.2–1.6 m in Broadwater. The areas offlooding and drying are not extensive as they are limited by a smalltidal range and steep shores. The tidal stream was very strong in thethree Narrows sections with a maximum current speed of 2.5 m/sand there is some delay in turnover of water from the inner to outerbays illustrated by a 143 min difference between high water at themouth and Broadwater (Parkes, 1958) mainly due to friction and thechannel geometry.

The general flow circulation is characterised by several eddies,and a clearly wind driven scenario. The system can be consideredas three layers, with a residual current in the direction of the windstress in the surface layer, a return current in the deep parts of thebay and mixed residual current directions in the mid-water layer.

3. Field sampling and data collection

Hydrographic measurements were taken at two stations withinMulroy Bay (Fig. 1) using two Valeport BFM 308 Direct Recordingcurrent metres deployed at 2 m below the surface and 2 m abovethe seabed. Deployment was for 16 days from the 8 to 24 February2005 during which current speed (m/s), direction (degrees magN)and water pressure (dB) were recorded at 20 min intervals over a60 s averaging period. For comparison of the quiescent period, only,an additional data set was used which had been collected between13 and 28 September 2000 at three different stations within the areausing identical sampling protocols.

A series of surveys were carried out in the study area duringspring and summer 2007 and winter 2008. Salinity and temperature

Fig. 2. Distribution of Hunter–Simpson stratification criteria in Mulroy Bay, showing

the mixed areas as light grey (values o1), stratified as dark grey (values 42) and the

sampling stations (A–H) used.


profiles were taken at 1 m depth intervals at 7 stations (Fig. 2) inorder to characterize and compare differences in the study area.Temperature readings were taken using a WTW portable oxygenmetre. Salinity was measured using an Oxi 197, LF 196 conductivitymetre and probe. The density was calculated as a function oftemperature, salinity and pressure (Fofonoff and Millard, 1983).The differences between surface and bottom densities values werecalculated.

4. The three-dimensional (3D) water modelling system,MOHID

The hydrodynamic model MOHID (Modelo Hidrodinamico)used in this work was developed by MARETEC (Marine andEnvironmental Technology Research Centre) at Instituto SuperiorTecnico, Technical University of Lisbon. The model solves theequations of a three-dimensional flow for incompressible fluidsand an equation of state relating density to salinity and tempera-ture (Martins et al., 1998, 2001; Santos, 1995). The MOHID modelhas been applied to several coastal and estuarine areas and it hasshown its ability to simulate complex features of the flows. It hasbeen used, for example, in coastal circulation, nutrient loads andresidence time models in several places around the world (seehttp://www.mohid.com/Publications/JP.asp).

Following Martins et al. (2001), the Cartesian coordinateframework equations are as follows:

@u1

@x1þ@u2

@x2þ@u3

@x3¼ 0 ð1Þ

@u1

@tþ@ðuju1Þ

@xj¼ fu2�g

rZr0

@Z@x1�

1

r0

@pS

@x1�

g

r0

Z Z

z

@ru@x1

dx3þ@

@xjAj@u1

@xj

� �ð2Þ

@u2

@tþ@ðuju2Þ

@xj¼�fu1�g

rZr0

@Z@x2�

1

r0

@pS

@x2�

g

r0

Z Z

z

@ru@x2

dx3þ@

@xjAj@u2

@xj

� �ð3Þ

@p

@x3¼�rg ð4Þ

where (ui) are the velocity components xi, Z is the free surfaceelevation, Ai turbulent viscosity coefficients, f the Coriolis para-meter, ps the atmospheric pressure g the gravitational accelera-tion, r the density and r0 the density anomaly as the depth meandensity minus the density at particular height. The density iscalculated as a function of salinity and temperature by theequation of state (Leendertsse and Liu, 1978).

A bathymetry model was obtained by digitising AdmiraltyChart (SNC 2699) to produce an initial GIS vector data layerfollowed by interpolation to a 50 m grid resolution. An Arakawa Cgrid was used for spatial discretization (Arakawa and Lamb,1977). The MOHID model allows several options for verticaldiscretization: Cartesian coordinates, sigma coordinates or ageneric vertical coordinate. In this study a sigma coordinate waschosen with 5 vertical layers. The temporal discretization iscarried out by means of a semi implicit ADI (Alternate DirectionImplicit) algorithm, introduced by Peaceman and Racford in 1955(Fletcher, 1991). The model was run for specific dates, 15 dayscovering a lunar tidal cycle, this being the minimum time periodsuggested by the Scottish Environmental Protection Agency(SEPA, 2005). In fact, the model was run over a total of 20 daysin total, from which 15 days were extracted for the calibrationand validations stages.


http://www.mohid.com/Publications/JP.asp

dx.doi.org/10.1016/j.csr.2011.01.001


Bottom stress was parameterized using a quadratic law, bycalculating the bottom drag coefficient, Cd, from the expression:

Cd¼K

ln ðzþzb0Þ=zb

0

� � !2

ð5Þ

where K is the Von Karman constant, z height above the bed andzb

0 is the bottom roughness length.Vertical eddy viscosity/diffusivity was determined with a

turbulence closure model selected from those available in theGeneral Ocean Turbulence Model GOTM (Burchard et al., 1999).

In this application a 3D model forced with both tide and windwas implemented. Water level imposed at the boundary wastaken from the FES2004 global tide solution (Lyard et al., 2006),which gave heights of tidal constituents. This model solution wasused because no data was available from nearby tide gauges.Wind was assumed to be uniform in the modelled area and equalto that measured at the closest meteorological station. Theparameters used in the model calculations are summarised inTable 2.

The residual current ue was estimated by averaging thehorizontal flow u over the modelled time period T, (spring–neapcycle) using the equation:

ue ¼ 1=T

Z T

0udt ð6Þ

A particle tracking model was coupled to the hydrodynamicmodel to describe the movement of the passive tracers. Theparticle tracking assumed that the velocity of each particle (up)can be split into a large scale organised flow, characterised by amean velocity (uM), provided by the model, and a smaller scalerandom fluctuation (uF) so that uP¼uM+uF. The particle trackingmodel used the equation:

@xi

@t¼ uiðxi,tÞ ð7Þ

where ui is the mean velocity and xi the particle position, thisequation is solved using an explicit method:

xtþDti ¼ xt

iþDtuti ð8Þ

Random movement was calculated following the procedureof Allen (1982) which considers the diffusive process acting on theseparticle to be represented by random walks. In this study everyparticle is considered as a water parcel whose paths are modelled astheir move through the water. The random displacement wascalculated using the mixing length and the standard deviation ofturbulent velocity provided from the hydrodynamic model.

For the validation process the relative mean absolute error(RMAE) and the Index of Agreement (IoAd) were computed overthe full 15 day period using Eqs. (9) and (10), respectively. TheRMAE has been used by several authors to the evaluate numerical

Table 2Parameters used in the model calculations.

Physical parameter Numerical value

Time step 2 s

Grid mesh 50 m

Horizontal cells (x,y) 193, 244

Vertical coordinate Sigma

Vertical layers 5

Horizontal eddy viscosity 0.407738 m2 s�1

Vertical eddy viscosity 0.001 m2 s�1

Drag coefficient 0.03

Wind rugosity coefficient 0.0025

River discharge No

Temperature 7 1C

Salinity 32.6 psu


model results (Fernandes et al., 2001; Sousa and Dias, 2007;Sutherland et al., 2004) and is given by

RMAE¼Qm�Qcj jh i

Qmj jh ið9Þ

where Qm and Qc are the measured and computed velocity vector,respectively.

The preliminary classification of RMAE ranges suggestedby Walstra et al. (2001) is shown in Table 3.

The index of agreement measure (IoAd) (Dawson et al., 2007;Wilmott,1981) is given by

IoAd¼ 1�

Pni ¼ 1 Qm�Qcð Þ

2

Pni ¼ 1 Qc�Qi

�� þ Qm�Qi

�� 2ð10Þ

where Qm and Qc is the measured and computed velocity vector,respectively, and Qi is the measured mean velocity vector. Thebest results are when IoAd close to 1.

Stratification was quantified by determining the potentialenergy anomaly, j as the energy required to completely mixthe water column, following Simpson et al. (1978) and Simpsonand Bowers (1981) and using the equation:

j¼ g

H

Z 0

�Hr��rðzÞ� �

zdz ð11Þ

where the z coordinate is depth vertically upwards from the seasurface, g is the gravitational acceleration, r is the mean densityand H the water column height.

The Simpson–Hunter stratification parameter (Simpson andHunter, 1974) is given by

S¼ log10h

9U39

" #ð12Þ

where S is the stratification parameter, h the water depth, and U isthe magnitude of the instantaneous tidal stream velocity over onetidal cycle. The stratification parameter was calculated with h asthe mean water depth for each cell and U as a mean tidal velocitymodelled for each grid position. Although the Simpson–Hunterstratification parameter may be used, it is assumed that thestratification was only caused by thermal heat as there is littlefresh water inflow relative to the bay volume from two smallstreams at Kindrum and Milford, confirmed by the relativelysmall change in overall salinity between mouth and heads ofthe different bays (Telfer and Robinson, 2003)

A value of S at 1.5 indicates the presence of a front whereasvalues of So1 indicate well mixed regions and S42 wellstratified areas (Perry et al., 1983). To verify the existence of wellmixed and stratified zones the surface stratification was calcu-lated, defined as the difference between the surface and thebottom densities following Perry et al. (1983) and Muelbertet al. (1994) .The energy required to mix the water columnincreases with increasing stratification; in mixed waters itis o10 Jm�3, in frontal waters 10–20, J m�3 and in stratifiedwaters 420 J m�3 (Lee et al., 2005).

The quiescent period was given by the percentage incidence ofcurrent speeds within the range 0–3 cm/s (SEPA, 2005), and was

Table 3Classifications error for RMAE (after Walstra et al., 2001).

Classification RMAE

Excellent o0.2

Good 0.2–0.4

Reasonable 0.4–0.7

Poor 0.7–1

Bad 41


dx.doi.org/10.1016/j.csr.2011.01.001


obtained from the modelled numerical values. To obtain theresidence time, a modification of the approach of Braunschweiget al. (2003) was used, in which the six fish productioncage-blocks were considered as boxes either 100�150 m or100�200 m in area. These boxes were used as the release pointsfor Lagrangian tracers with every particle being considered as apassive volume of water. The movement of these tracers through-out the fjard system was examined with 1000 particles beingused to simulate one cage; a group of 6 cages being representedby 6000 particles. It was considered that 1000 particles weresufficiently representative rather than the 20,000 used by Allen(1982) as this avoided excessive computational effort. The GOTMwas used as a 3D model while the particle tracking model was 2D.The values of the constants used are summarised in Table 2.

The time period simulated was based on the worst case,represented by a day with neap tides and the initial modelrunning in the flood tide period without wind. The hydrodynamicmodel results were incorporated into the Arc View GIS system inorder to provide an easy-to-use graphical user interface forvisualisation, interrogation of results and as an input to a furtherspatial modelling project.

Fig. 4. Examples of salinity vs. bottom-depth plot from the stations H and E

illustrated the difference in salinity between zones in winter season.
5. Results
5.1. Field data

Mean seasonal temperature ranged from 6 1C in February2008, to 12 1C in April 2007 to 16 1C in August 2007. In thesummer (16 1C) and winter (6 1C) seasons there was completethermal vertical mixing with the same values of temperature inthe water column in the whole study area. The density profile(Fig. 3) shows that during the spring period, April 2007, severalstations showed stratified patterns (stations C, D, G, H) mainlydue to differences between surface and bottom values of 1.6 1Cmaximum in Kindrum and 1 1C in the Milford area. No variationwas observed at the sampling stations at Glinsk and Millstone(stations A and B, respectively) in the Narrows.

The salinity during the summer and winter seasons wascompletely vertically mixed with the same values throughoutthe water column. Seasonal salinity values fluctuated from30.9 psu in January to 34.0 psu in August. The values at severalstations over the whole area were very similar for summer andwinter periods (sampled in February 2008 and August 2007) both

Fig. 3. Density variation with depth from all stations for April 2007.


at the surface and the seabed with little variation occurringbetween stations. Salinity depth profiles obtained in February2008 (Fig. 4) indicated that there were different values in thedifferent areas of the fjard, 30.9–31.0 psu in Kindrum area,possibly caused by rain and small rivers draining into the area,while elsewhere within Broadwater values were close to 32.6 psu.

In spring time the density differences between the surface andthe bottom at the Narrows stations A and B and Station D were lessthan 0.056 kg m�3, whereas at Broadwater and Kindrum stations(H, G, E, C) there was more stratification with values greater than0.110 kgm�3, the maximum difference of 0.295 kgm�3 being in theKindrum area. The potential energy anomaly at the narrows station(Station A, B, C) was almost zero, between 0 and 0.95 J m�3, while inBroadwater, Moross channel and Kindrum differences were inter-mediate between the maximum in Kindrum zone (13.47 J m�3) andthe minimum in Moross channel (1.038 J m�3).

5.2. Modelled results

In this study, the 3D hydrodynamic model was based on a 15day lunar tidal cycle with forcing by a real wind data set. Amathematical sigma coordinate was used with 5 vertical layers;the first layer was at the bottom and the fifth at the surface. Thesensitivity analysis showed that the variation in the eddy viscos-ity did not have a large influence on the modelling outcomes,while the drag coefficient has the highest relative sensitivity.Model validation was based on hydrodynamic measurementsmade in Mulroy Bay between 15 and 24/02/05 (9 days in total).

For sea surface elevation (Fig. 5), the calculated values ofIoAd were close to 1 and low values of RMAE revealed a goodagreement between the prediction of the model and the observa-tions (Fig. 5 and Table 4) according to Walstra et al. (2001). Thevalues of the observed and modelled mean current velocity(Fig. 5), at station 1, were similar with differences of 0.015 m/sat the surface and 0.035 m/s near the seabed. At station 2 thedifferences in mean modelled current velocity was higher, being0.084 m/s at the surface and 0.088 m/s near the seabed betweenobserved and modelled values.


dx.doi.org/10.1016/j.csr.2011.01.001

Fig. 5. Time series of the elevation, current direction and current intensity in Mulroy Bay measured (dotted and dark line) and modelled (grey line) at the two stations A

and B. Values of RMAE and IoAd are given in Table 4.

Table 4RMAE, and IoAd for the eastward (U) and northward (V) velocities, current

intensity at the superficial (s) and the bottom (b) and the tidal surface elevation

for the stations.

Parameter Station 1 Station 2

RMAE IoAd RMAE IoAd

Us 0.008 0.971 0.3 0.960

Vs 0.54 0.341 0.05 0.980

Ub 0.05 0.976 0.39 0.929

Vb 0.28 0.620 0.18 0.937

Current intensity s 0.903 0.903

Current intensity b 0.896 0.762

Elevation 0.025 0.950 0.057 0.927

Fig. 6. Distribution of modelled mean current speed in Mulroy bay. The data

shown is the current speed (m/s) in layer 4 which approximates cage positions in

the water column.


Current directions were almost identical at both stations(Fig. 5), showing that the model can predict the current directionaccurately. At station 1 the model agreed well with the eastwardvelocity measurements (Table 4) in the surface and bottom layers,with IoAd values close to 1 and low values of RMAE (Table 4)which were within the excellent category according to Walstraet al. (2001).

The northward velocity results were more complicated. TheRMAE values were 0.28 and 0.54 for bottom and surface levels,respectively, with IoAd values varying from 0.34 to 0.62 (Table 4).The model did not predict this velocity component in this stationaccurately although it did provide very good approximation ofcurrent speed and current direction. For this reason the north-ward velocity influences may be low. At station 2 the modelagreed well with the northward and eastward velocities mea-sured at the surface and the bottom, with IoAd values close to1 and excellent and good values of RMAE (Table 4).

The average measured current speeds for the 4th layer wereused within the model as this gave the best estimation of the cageenvironment approximating the depth of the middle and bottom


of the cages where the majority of the waste originates (Corneret al., 2006) The mean current speed modelled in this layer isgiven in Fig. 6, from which two regions can be differentiated:(1) the Narrows, where modelled mean current speeds rangedfrom 0.2 m/s to more than 1 m/s, and (2) Broadwater, where themodelled mean current speeds ranged between 0.03 m/s atKindrum and 0.2 m/s within the Morross channel.


dx.doi.org/10.1016/j.csr.2011.01.001


The quiescent period, defined as percentage of the modelledperiod where current speeds were less than 0.03 m/s (SEPA,2005), were given for the first layer (bottom), as this providedan approximation of the seabed environment (Fig. 7). The quies-cent period in this layer ranged from 10% to 80% of the tidalperiod. Areas of Kindrum and Milford were considered mostlyquiescent, in contrast to the well flushed and only minimallyquiescent areas within the Narrows and Moross Channel (definedaccording to criteria set by SEPA, 2005). The modelled andmeasured quiescent periods at the two stations at the surfaceand bottom were very similar at about 1% (Table 5). Comparingthe modelled result with the hydrographic data set from 2000 butwith the same time period (15 days) that includes the locations ofthree fish cages, it can be seen that in Broadwater differences areoverall less than 10%, but with 25% at Kindrum and 10% atMoross.

The modelled residual currents (Fig. 8) show a circulationstructure with an anticlockwise eddy in the Millstone area wherethe maximum salmon production is located. At the Narrows the 3Dhydrodynamic model showed the same patterns in the top, mediumand bottom layers which are similar due to the low depth.

Fig. 7. Distribution of the modelled percentage of quiescent period in Mulroy Bay.

The data shown is the quiescent period in layer 1 which approximates the bottom

environment.

Table 5Quiescent period at the surface (S) and the bottom (B) for stations 1 and 2, and

three cage sites, showing the data set used.

Station Data set Quiescent period

Measured S Measured B Modelled S Modelled B

Station 1 Feb 2005 1% 1% 1% 1%

Station 2 Feb 2005 1% Nr 1% 1%

Broadwater Sep 2000 10% 25% 17% 20%

Kindrum Sep 2000 95% 99% 98% 75%

Moross Sep 2000 16% 30% 10% 20%

nr¼not measured due to current metre failure.

Fig. 8. Distribution of residual current in the central sections of Mulroy Bay. A

very clear circulation structure can be indentified with anticlockwise and clock-

wise eddies in different areas of the narrows. In Millstone area an anticlockwise

eddy (black square) may affect cage water exchange.


To illustrate the mixing of the volume of water (simulatingeffluents) from the cages, the particles in different cages positionswere colour-coded. The particles were then advected by themodel flow field for one day (17/2/2005). The results are pre-sented as an animated dispersion model (Animation 1 and Fig. 9).The animation clearly shows the current eddy, at certain times ofthe tide, in the Millstone area and mixing water from the Glinsk(colour red) and Millstone (colour yellow). The water from Glinskis eventually flushed away to the Broadwater area, while inMillstone part is flushed to Broadwater and the rest remains tofeed the eddy. Particles originating from fish farms in the innerbays (Moross in green and Broadwater in blue) exhibited smallchanges at the beginning of the model run but by the end thevolume of particles are dispersed so that the Broadwater particlesend up in the mouth and in Moross Channel while MorossChannel particles area are spread throughout the channel.

The distribution of the modelled Simpson–Hunter criteria can beseparated into two areas (Fig. 2). The narrows, Glinsk, Millstone andthe northern part of the Moross Channel are considered well mixed,with values less than 1 for the Simpson–Hunter criteria while theremainder of Mulroy Bay can be considered as stratified withmodelled Simpson–Hunter criteria values of greater than 2.

6. Discussion

Numerical circulation models can provide a practical solutionto the problem of coastal mixing (Wildish et al., 2004) and waterexchange in Regions of Restricted Exchange that host importantindustries, such as aquaculture, and can improve resolution ofthe combined effects of tidal and wind-driven forcing as well asreflecting complex topography and intertidal drying zones(Hargrave, 2003). 3D modelling schemes require intensive com-puter resources and may suffer from computational instabilityproblems, but the models can provide complete spatio-temporalinformation on water currents for the entire computationaldomain. 2D models will provide vertical and depth averagedcomponents of velocity (speed and direction) but this informationis insufficient in some cases, particularly when the differencesbetween the surface and the bottom velocities may differ


dx.doi.org/10.1016/j.csr.2011.01.001

Fig. 9. An example of the mixing of the volume of water from four cage sites. The initial position of the four cages is shown by the boxes (A) each of which has a different

grey scale. The subsequent dispersion pattern is illustrated in B, C and D.


considerably. In addition, the 2D approach does not providedifferential information for near seabed environments such asthe extended quiescent period. Panchang et al. (1997) suggestthat the vertical variations in current speed are likely to affect thedispersion of waste and resuspension of settled wastes and a 3Dapproach would therefore be more appropriate for modellingsuch wastes.

The study area on which the present work is based is charac-terised by two main hydrographic areas that could host aquaculturesites; a highly energetic and well flushed part of the fjard in thenarrows and a low energy, poorly flushed part in the Broadwater,Kindrum and Millford areas. The results indicate that the areafollows a seasonal cycle that contains three contrasting regimes. Inthe summer and winter the systems show complete thermal verticalmixing while in spring the inner bay system is stratified but themore hydrodynamic Narrows area is always well mixed. Similarhydrological behaviour was encountered in a Scottish fjord (Rippethet al., 1995). The measured values for temperature and salinity arein good agreement with those directly measured in Mulroy Bay byother authors (C-Mar, 2000; Nunn, 1996;Telfer and Robinson, 2003).The temperature is highest (16–17 1C) and the oxygen concentrationis lowest, close to 7 mg/l, in summer time.

The model has been forced by tides and winds. C-Mar (2000)found a direct relationship between wind forcing and the con-centration of dissolved oxygen in the water and the increase ofwater exchange with the ocean. The numerical model couldprovide important insight by showing how the fjard wouldrespond to a considerable increase in the temperature, changingwind direction and intensity scenarios and dissolved oxygenevolution, so providing the possibility of modelling climate-related ‘‘what-if’’ scenarios in advance.


Cross (1993) suggested that net cage locations must considercirculation dynamics, including the evaluation of back eddy andmass transport contribution to waste dispersion for a site. Brooksand Churchill (1991) noticed that finer grid resolution (less than100 m) is needed for the characterisation of circulation in coastalareas similar to Mulroy Bay, and Panchang et al. (1997) suggestedthe use of a fine (75 m) grid size in areas with aquacultureactivities. The latter modelled the presence of an eddy whichinhibited the exit of salmon waste. Hargrave et al. (1995) notethat most studies in aquaculture have shown that the local extentof altered benthic community structure and biomass is limited toless than 50 m from the edge of the cages and for this reason themodel was parameterized to this horizontal resolution.

Residual current velocity and the developed animation showthe presence of an eddy in the largest fish production area inMillstone which may affect environmental quality by retainingrather than dispersing waste. This is the first animated example ofhow a physical circulation structure could affect different aqua-culture sites. It is clearly important to consider the inter-relation-ship among them and to be able to identify such areas forenvironmental management because nutrient, waste dispersionand potential disease transfers from the cages may affect theother sites. This kind of model could provide important informa-tion for the optimisation of fish cage locations.

The general flow circulation in Mulroy Bay is characterised byseveral eddies, which are clearly wind driven.The tidal streams arevery strong, particularly in the Narrows which have the lowestquiescent values and maximum mean current speeds. Hargraveet al. (1995) noticed that benthic variables which are correlated withorganic matter sedimentation can be used to scale the degree oforganic enrichment and also suggested that biological processes are


dx.doi.org/10.1016/j.csr.2011.01.001


not always sufficient to limit organic matter accumulation especiallyin areas where hydrographic conditions and/or low current speedsresult in low rates of oxygen supply to the sediment surface. A 3Dscheme could provide much improved spatial and temporal infor-mation about the conditions in the seabed.

This study was the first to assess water column stratificationand quiescent periods in a marine area that hosts salmonaquaculture production. The modelled periods of quiescent waterand mean current speed were confirmed as reasonably accurateby comparison with measurements of hydrographic conditions atdifferent locations within Mulroy Bay, and even comparinghydrographic data sets from different time periods that includethe locations of the three fish cages sites. It is important to notethat although the times modelled were almost the same 15 daytidal lunar cycle, the different wind conditions could affect theresult. Telfer and Robinson (2003) showed that the areas with thelowest values of oxygen in deep water coincide with areas withthe highest values of quiescent water (Kindrum and Milford).C-Mar (2000) reported high values of sediment oxygen demand inthe area, and the quiescent period may be a good modelledindicator of potential oxygen depletion in the seabed.

In strongly stratified marine systems, dissolved material canbe effectively trapped in the upper or lower parts of the watercolumn. In an aquaculture context, stratification can be animportant factor in the dispersal of organic matter from certainfarms in the inner portions of fjords (Wildish et al., 2004). A slightthermal stratification was observed in the spring period duringthis study in Northwater and Broadwater, (C-Mar, 2000; Minchin,1981) and in the southern region at Milford. Although there areno large rivers draining into the bay to significantly affect salinity,a shallow halocline can develop in parts of Northwater andBroadwater where water from land runoff lies on the surfaceduring calmer weather. Salinity depth profiles obtained in January1999 (C-Mar, 2000) indicated that there was saline stratificationin the southern region of Broadwater, probably caused by landrunoff and streams which drain into this area.

The modelled Simpson–Hunter tidal stratification parameterindicated that most of Mulroy Bay was potentially stratified inspring time with well mixed areas in the shallow narrows. Thevalues are in good agreement with the direct measurements ofwater column stratification based on observed density profiles. Interms of potential energy, where the sigma value is low andpotential energy is close to zero the Simpson–Hunter criterion isbelow 2. Stations with highest density differences and potentialenergy have the highest values of the Simpson–Hunter criteria withvalues more than 2.

Simpson and Hunter (1974) used the surface current ampli-tude at springs, whereas in this study the mean surface currentspeed of a lunar tidal cycle was used. This provided a better cagesite approximation but produced lower values of current speedwhich could lead to higher values of the Simpson–Hunter criteria.Several stratification parameters have been defined by otherauthors but Hunter and Sharp (1983) noticed that, at a given site,all could be approximately related to each other. Lu et al. (2001)used only four tidal constituents: M2, S2, K1 and O1, and foundthat the resulting modelled stratification distributions did notchange significantly in comparison to results modelled by Pingreeand Griffiths (1980) who used only one constituent, M2. In thepresent study, both elevation and current data were generatedand included more tidal constituents from the FES2004 model,M2, S2, N2, K2, K1, O1, Q1, P1 and M4. No significant changes ofstratification parameter distribution were expected.

The Simpson–Hunter stratification criterion was selected forits simplicity and ease of calculation. It was used to model theworst possible scenario where the stratification was only causedby thermal heat. The area has no influences from river discharge


affecting the stratification and other important energy sources forthe water mixing of the water column, such as wind mixing andtidal stirring, are not taken into account. A strong wind blowingfor many hours and high turbulence from the bottom can producea mixed layer and reduce the density differences.

The results from the hydrodynamic model have be incorpo-rated into GIS to provide an adaptable graphical user interface for2D, 3D and temporal visualisation, for interrogation of results andas an input to other aquaculture spatial models or decisionsupport systems. This offers the possibility of combining the datawith layers of spatial information about economic and socialaspects, communications and security of the study area todevelop an integrated ecological approach to aquaculture activ-ities. As described by Henderson et al. (2001) the main potentialand recommendation for using modelling in aquaculture activ-ities is as an indicator of environmental change, as a strongdescriptor of physical processes, as a tool for best practices indevelopment and regulations, as a cost effective alternative toextensive field’s studies, and to provide fast predictions ofpotential impacts for different aquaculture scenarios. The hydro-graphy assessed in this study may be crucial in any decisionmaking process in order to avoid future environmental problems.In our study we predict areas with low hydrodynamic regime thatmay be affected by high environmental impacts such as highwaste accumulation and methane gas production. In addition, thehydrodynamic and water quality models could be useful in theassessment of the mixing zone and in designating allowable zonesof effect for nutrient and chemical discharges. However, anymodelling process may give false or inaccurate predictions andthus there are risks in applying modelling approaches to anydecision making process, where complex environmental pro-cesses are oversimplified (Henderson et al., 2001). Althoughprocess of 3D circulation modelling is expensive; the benefits ofsuch a decision support tool which is well tuned for aquaculturedevelopment are also considerable (Andrefout et al., 2006). Thereare also considerable opportunities to link with other kinds ofmodels and the benefits for aquaculture accruing from such acomprehensive spatial decision support tool are high.

In summary, this work has shown that the use of a 3DHydrodynamic Model coupled to a Particle Tracking Model canprovide spatially explicit information on the hydrodynamic con-ditions governing the water dynamics and the transport and fateof pollutants in the near and far field marine cage environment.The use of GIS, integrated with the hydrodynamic model forplanning, regulation, monitoring and site selection for cageaquaculture should be encouraged.

Appendix A. Supporting information

Supplementary data associated with this article can be foundin the online version at doi:10.1016/j.csr.2011.01.001.

References

Allen, C.M., 1982. Numerical simulation of contaminant dispersion in estuaryflows. Proceedings of the Royal Society 381, 179–194.

Arakawa, A., Lamb, V., 1977. Computational design of the basic dynamicalprocesses of the UCLA General Circulation Model. Methods of ComputationalPhysics 17, 174–264.

Andrefout, S., Ouillon, S., Brinkman, R., Falter, J., Douillet, P., Wolk, F., Smith, R.,Garen, P., Martinez, E., Laurent, V., Lo, C., Remoissenet, G., Scourzic, B., Gilbert,A., Deleersnijder, E., Steinberg, C., Choukroun, S., Buestel, D., 2006. Review ofsolutions for 3D hydrodynamic modelling applied to aquaculture in SouthPacific atoll lagoons. Marine Pollution Bulletin 52, 1138–1155.

Beveridge, M., 2004. Cage Aquaculture, Third Edition Blackwell Publishing.


doi:10.1016/j.csr.2011.01.001

dx.doi.org/10.1016/j.csr.2011.01.001


Braunschweig, F., Martins, F., Chambel, P., Neves, R., 2003. A methodology toestimate renewal time scales in estuaries; the Tagus estuary case. OceanDynamics 53, 137–145.

Brooks, D., Churchill, L., 1991. Experiment with a terrain-following hydrodynamicmodel for Cobscook Bay in the Gulf of Maine, Sparlding, ed, Tampa,pp. 215–225.

Burchard, H., Bolding, K., Villarreal, M.R., 1999. GOTM, a General Ocean TurbulenceModel. Theory, implementation and test cases. European commission, EUR18745EN.

C-Mar, 2000. The environmental management of Mulroy Bay in relation toaquaculture production. Queens University of Belfast, Centre for MarineResource Management.

Corner, R.A., Brooker, A.J., Telfer, T.C., Ross, L.G., 2006. A fully integrated GIS-basedmodel of particulate waste distribution from marine fish-cage sites. Aqua-culture 258, 299–311.

Cross, S.F., 1993. Oceanographic characteristics of netcage culture sites consideredoptimal for minimizing environmental impacts in coastal British Columbia,Ministry of Agriculture, Fisheries and Food.

Dawson, C.W., Abrahart, R.J., See, L.M., 2007. HydroTest: a web-based toolbox ofevaluation metrics for the standardised assessment of hydrological forecasts.Environmental Modelling and Software 22, 1034–1054.

Dronkers, J., Zimmerman, J.T.F., 1982. Some principles of mixing in tidal lagoons.Oceanologica Acta 4, 107–117.

Dudley, R.W., Panchang, V.G., Newell, C.R., 2000. Application of a comprehensivemodelling strategy for the management of net-pen aquaculture waste trans-port. Aquaculture 187, 319–349.

Falconer, R.A., Hartnett, M., 1993. Mathematical modelling of flow, pesticide andnutrient transport for fish farm planning and management. Ocean and Shore-line Management 19, 37–57.

Farmer, D.M., 1976. The influence of wind on the surface layer of a stratified inlet:part I. Analysis. Journal of Physical Oceanography 6, 941–952.

Farmer, D.M., Freeland, H.J., 1983. The physical oceanography of fjords. Progress inOceanography 12, 147–219.

Farmer, D.M., Osborn, T.R., 1976. The influence of wind on the surface layer of astratified inlet: part I. Observations. Journal of Physical Oceanography 6,931–940.

Fernandes, E.H.L., Dyer, K.R., Niencheski, L.F.H., 2001. TELEMAC-2D calibration andvalidation to the hydrodynamics of the Patos Lagoon (Brazil). Journal ofCoastal Research 34, 470–488.

Fletcher, C.A.J., 1991. Computational Techniques for Fluid Dynamics, 2nd EditionSpringer Verlag, New York.

Fofonoff, N.P., Millard, R.C., 1983. Algorithms for computation of fundamentalproperties of seawater. UNESCO. Technical Paper in Marine Science, 44 pp.

Freeland, H.J., Farmer, D.M., Levings, C.D., 1980. Fjord Oceanography. PlenumPress, New York.

Greenberg, A., Shore, A., Page, H., Dowd, M., 2005. A finite element circulationmodel for embayments with drying intertidal areas and its applications to theQuoddy region of the Bay of Fundy. Ocean Modelling 10, 211–231.

Hargrave, B.T., 2003. A scientific review of the potential environmental effects ofaquaculture in aquatic ecosystems. vol. I. Far field environmental effects ofmarine finfish aquaculture. Fisheries and Oceans Canada, p. 2450.

Hargrave, B.T., Phillips, L.I., Doucette, M.J., White, T.G., Milligan, D.J., Wildish, D.J.,1995. Biochemical observations to assess benthic impacts of organic enrichmentfrom marine aquaculture in the Western Isles, region of the Bay of Fundy, 1994,Canadian Technical Report of Fisheries and Aquatic Science, p. 2062.

Henderson, A., Gamito, S., Karakassis, I., Pederson, P., Smaal, A., 2001. Use ofhydrodynamic and benthic models for managing environmental impacts ofmarine aquaculture. Journal of Applied Ichthyology 17, 163–172.

Hunter, J.R. and Sharp, G.P., 1983. Physics and fish populations: Shelf sea frontsand fisheries, FAO Fisheries Report, vol. 2, FAO, Rome, San Jose, Costa Rica, 18–29 April 1983, p. 291.

Kapetsky, J.M., Aguilar-Manjarrez, J., 2007. Geographic information systems,remote sensing and mapping for the development and management of marineaquaculture. FAO, Rome, p. 458.

Lee, O., Nash, D.M., Danilowicz, B.S., 2005. Small scale spatio temporal variabilityin Ichthyoplnakton and zooplankton distribution in relation to tidal mixingfront in the Irish Sea. ICES Journal of Marine Science 62, 1021–1036.

Leendertsse, J., Liu, S., 1978. A three-dimensional turbulent energy model for non-homogeneous estuaries and coastal systems. In: Nihoul, J.C.J. (Ed.), Hydro-dynamics of Estuaries and Fjords. Elsevier, Amsterdam, pp. 387–405.

Leth, O.K., 1995. A study on the effect of local wind on the dynamics of the upperlayer in the inner part of Malagen. In: Skjoldal, H.R., Hopkins, C., Erikstad, K.E.,Leinaas, H.P. (Eds.), Ecology of Fjords and Coastal Waters. Elsevier Science, pp.185–194.

Lu, Y., Keith, R.T., Wright, D.G., 2001. Tidal currents and mixing in the Gulf of StLawrence: an appliction of the incremental approach to data assimilation.Canadian Bulletin of Fisheries and Aquatic Sciences 58, 723–735.

Lyard, F., Lefevre, F., Letellier, T., Francis, O., 2006. Modelling the global oceantides: modern insights from FES2004. Ocean Dynamics 56, 394–415.


Martins, F., Leitao, P.C., Silva, A., Neves, R.J., 2001. 3D modelling in the Sado estuaryusing a new generic vertical discretization approach. Oceanologica Acta 24,51–62.

Martins, F.A., Neves, R.J., Leit~ao, P.C., 1998. A three-dimensional hydrodynamicmodel with generic vertical coordinate. In: Babovic, V., Larsen, L.C. (Eds.),Proceedings of Hydroinformatics, 98(2). Balkerna, Rotterdam, Copenhague,Denmark, pp. 1403–1410.

Meaden, G.J., Kapetsky, J.M., 2001. Geographical information systems and remotesensing in inland fisheries and aquaculture. FAO Fisheries Technical Paper.FAO Rome, pp. 318.

Minchin, D., 1981. The scallop Pecten maximus in Mulroy Bay. Fisheries Bulletin,Dublin, 21 pp.

Muelbert, H.J., Lewis, M.R., Kelley, D.E., 1994. The important of small-scaleturbulance in the feeding of herring larvae. Journal of Plankton Research 16,927–944.

Nath, S.S., Bolte, J.P., Ross, L.G., Aguilar-Manjarrez, J., 2000. Application ofgeographical information systems (GIS) for spatial decision support in aqua-culture. Aquacultural Engineering 23, 233–278.

Nunn, J.D., 1996. The marine Mollusca of Ireland, 2. Mulroy Bay, Co. Donegal.Bulletin of the Irish Biogeographical Society 19, 15–138.

Murray, A.G., Gillibrand, P.A., 2006. Modelling salmon lice dispersal in LochTorridon, Scotland. Marine Pollution Bulletin 53, 128–135.

Panchang, V., Cheng, G., Newell, C., 1997. Modelling hydrodynamics and aqua-culture waste transport in coastal Maine. Estuaries 20, 14–41.

Parkes, H., 1958. A general survey of the marine algae of Mulroy Bay, Co Donegal.The Irish Naturalist Journal 12, 277–283.

Perez, O.M., Telfer, T.C., Beveridge, M.C.M., Ross, L.G., 2002. Geographical Informa-tion Systems (GIS) as a simple tool to aid modelling of particulate wastedistribution at marine fish cage sites. Estuarine Coastal and Shelf Science 54,761–768.

Perry, R.I., Dilke, B.R., Parsons, T.R., 1983. Tidal mixing and summer planktondistributions in Hecate Strait, British Columbia. Canadian Bulletin of Fisheriesand Aquatic Sciences 40, 871–887.

Pingree, R.D., Griffiths, D.K., 1980. A numerical model of the M2 tide in the Gulf ofSt Lawrence. Oceanologica Acta 3, 221–225.

Rippeth, T.P., Midgley, R.P., Simpson., J.H., 1995. The seasonal cycle of stratificationin a Scottish fjord. In: Skjoldal, H.R., Hopkins, C., Erikstad, K.E., Leinaas, H.P.(Eds.), Ecology of Fjords and Coastal Waters. Elsevier Science, pp. 85–92.

Santos, A.P., 1995. Modelo hidrodinamico de circulacao oceaanica e estuarina.Ph.D. Thesis, IST Lisbon (in portuguese).

SEPA, 2005. Hydrographic data requirements for application to discharge frommarine cage fish farms. Attachment VIII. In: Regulation and Monitoring ofMarine Cage Fish Farming in Scotland—a Procedures Manual, SEPA, Stirling,Scotland.

Simpson, J.H., Allen, C.M., Morris, M.C., 1978. Fronts on the continental shelf.Journal of Geophysical Research 83, 4607–4614.

Simpson, J.H., Bowers, D., 1981. Models of stratification and frontal movement inthe shelf seas. Deep-Sea Research 28, 727–738.

Simpson, J.H., Hunter, J.R., 1974. Fronts in the Irish Sea. Nature 250, 404–406.Skogen, M.D., Eknes, M., Asplin, L.C., SandvikA, D., 2009. Modelling the environ-

mental effects of fish farming in a Norwegian fjord. Aquaculture 298, 70–75.Sousa, M.C., Dias, J.M., 2007. Hydrodynamic model calibration for a mesotidal

lagoon: the case of Ria de Aveiro (Portugal). Journal of Coastal Research, SI 50(Proceedings of the ninth International Coastal Symposium), Gold Coast,Australia, pp. 1075–1080.

Stigebrandt, A., Aure, J., 1989. Vertical mixing in basin waters of fjords. Journal ofPhysical Oceanography 19, 917–926.

Sutherland, J., Peet, A.H., Soulsby, R.L., 2004. Evaluating the performance ofmorphological models. Coastal Engineering 51, 917–939.

Svendsen, H., Thompson, O.R.Y., 1978. Wind driven circulation in a fjord. AmericanMeteorological Society 6, 703–712.

Telfer, T., Robinson, K., 2003. Environmental quality and carrying capacity foraquaculture in Mulroy Bay Co. Donegal. Marine Environment and HealthSeries, No. 9. Marine Institute. Ireland.

Tett, P., Gilpin, L., Svendsen, H., Carina, P., Larson, E.U., Kratzer, S., Fouilland, E.,Janzen, C., Lee, J., Grenz, C., Newton, A., Ferreira, J.G., Fernandes, T., Scory, S.,2003. Eutrophication and some European waters of restricted exchange.Continental Shelf Research 23, 1635–1671.

Trites, R.W., Petrie, L., 1995. Physical oceanographic features of Letang Inletincluding evaluation and results from a numerical model. Canadian TechnicalReport of Hydrography and Ocean Sciences, pp. 163.

Walstra, D.J.R., van Rijn L.C., Blogg, H., van Ormondt M., 2001. Evaluation of ahydrodynamic area model based on the Coast3D Data at Teignmouth 1999. In:Proceedings of Coastal Dynamics 2001 Conference, Lund, pp. D4.1–D4.4.

Wildish, D.J., Dowd, M., Sutherland, T.F., Levings, C.D., 2004. A scientific review ofthe potential environmental effects of aquaculture in aquatic ecosystems.Volume 3. Near field organic enrichment from finfish aquaculture. CanadianTechnical Report of Fisheries and Aquatic Science, p. 2450.

Wilmott, C.J., 1981. On the validation of models. Physical Geography 2, 184–194.


dx.doi.org/10.1016/j.csr.2011.01.001

continental shelf research - university of · pdf filecalculations of the simpson–hunter...

Documents