E

Tt

1

2

M3

a4b5cQ16

7

a8

9

A10

R11

R12

A13

A14

15

K16

R17

C18

E19

320

U21

C22

H23

Z24

125

26

o27

l28

a29

t30

a31

m32

(33

t34

(a

0d

D P

RO

OF

ARTICLE IN PRESSG ModelCOMOD 5422 1–17

Ecological Modelling xxx (2009) xxx–xxx

Contents lists available at ScienceDirect

Ecological Modelling

journa l homepage: www.e lsev ier .com/ locate /eco lmodel

hree-dimensional modeling of the lower trophic levels inhe Ria de Aveiro (Portugal)

. Rodriguesa,∗, A. Oliveiraa,1, H. Queirogab,2, A.B. Fortunatoa,3, Y.J. Zhangc,4

Estuaries and Coastal Zones Division, National Civil Engineering Laboratory, Av. do Brasil, 101, 1700-066 Lisbon, PortugalBiology Department, University of Aveiro, Campus Universitário de Santiago, 3810-193 Aveiro, PortugalCenter for Coastal Margin Observation and Prediction, Oregon Health and Science University, 20000 NW Walker Road, Beaverton, OR 97006, United States

r t i c l e i n f o

rticle history:eceived 22 September 2008eceived in revised form 31 January 2009ccepted 6 February 2009vailable online xxx

eywords:ia de Aveirooastal LagoonCO-SELFED coupled modelnstructured gridshannels

a b s t r a c t

The water and the ecosystem dynamics of the Ria de Aveiro, a shallow, multi-branch lagoon locatedon the northwest coast of Portugal, are simulated using a new fully coupled 3D modeling system. Thismodel couples the hydrodynamic model semi-implicit Eulerian–Lagrangian finite-element (SELFE) andan ecological model extended from EcoSim 2.0 to represent zooplankton dynamics. The model appli-cation is based on an unstructured grid spatial discretization, which is particularly appropriate for thissystem given its complex geometry. The baroclinic circulation is calibrated and validated for differentenvironmental conditions, leading to velocity errors smaller than 5 cm/s across the lagoon. Ecologicalsimulations, focused on zooplankton dynamics represented by a site-specific formulation, are then pre-sented and compared against field data for two contrasting environmental conditions: Autumn 2000and Spring 2001. Results show that the fully coupled model is able to reproduce the dynamics of theecosystem in the Spring 2001, fitting the model results inside the range of data variation. During this

TE

ydrodynamicsooplankton dynamics

period zooplankton differences between data and model results are of about 0.005 mg C/l (60%), whileother ecological tracers’ differences are generally smaller than 20–30% along the several branches ofthe lagoon. In the Autumn 2000, the model tends to overestimate zooplankton by a factor of 10 and tounderestimate phytoplankton and ammonium, with discrepancies of about 0.1 mg C/l and 4.8 mol N/l,respectively. Factors like the ecological conditions imposed at the boundaries, the input parameters ofthe ecological model and the simplification of the ecosystem structure, since phytoplankton is the only

ered,

35

36

37

38

39

40

41

RR

ECprimary producer consid

. Introduction

The Ria de Aveiro is a shallow temperate coastal lagoon locatedn the Northwest (NW) coast of Portugal (4038′N, 845′W). Theagoon is about 45 km long and 10 km wide (Fig. 1) and has anverage depth of about 1 m (Dias et al., 2000). It is separated fromhe ocean by a sand spit, interrupted by an artificial tidal inlet

UN

CO

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

bout 20 m deep. The artificial inlet channel is connected to fourain branches, the Mira, Ílhavo, Espinheiro and S. Jacinto channels

Fig. 1), through which the main sources of freshwater flow intohe lagoon. The rivers Vouga and Antuã discharge to the Espinheiro

∗ Corresponding author. Tel.: +351 218443613; fax: +351 218443016.E-mail addresses: mfrodrigues@lnec.pt, marta.rodrigues@gmail.com

M. Rodrigues), aoliveira@lnec.pt (A. Oliveira), hqueiroga@bio.ua.pt (H. Queiroga),fortunato@lnec.pt (A.B. Fortunato), yinglong@stccmop.org (Y.J. Zhang).

1 Tel.: +351 218443631.2 Tel.: +351 234370787.3 Tel.: +351 218443425.4 Tel.: +1 503 748 1071.

42

43

44

45

46

47

48

49

304-3800/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.ecolmodel.2009.02.002

may explain the observed differences.© 2009 Elsevier B.V. All rights reserved.

channel, being the major sources of freshwater to the lagoon (Diaset al., 2000; Dias and Lopes, 2006). Smaller sources of freshwaterflow into the system through other channels, namely the Boco riverin the Ílhavo channel, the Caster River in the S. Jacinto channel andseveral small rivers in the Mira channel. Freshwater flows are poorlyknown due to a severe lack of data: Dias and Lopes (2006) referaverage annual flows of 50 and 5 m3/s, while Dias et al. (2000) referaverage annual flows of 29 and 2 m3/s, respectively, for the Vougaand the Antuã rivers; for the Vouga river, Vaz and Dias (2008) referan average annual flow of 31.45 m3/s, based on field measurementsin the Espinheiro channel from September 2003 to August 2004.Tides at the mouth of the lagoon are semi-diurnal, with a meantidal range of about 2 m (Dias et al., 2000). More detailed descrip-tions of the lagoon can be found in Dias et al. (2000) and referencesherein.

sional modeling of the lower trophic levels in the Ria de Aveiro

This lagoon plays an important ecological role, being the habi- 50

tat of several species of flora and fauna (Hermoso et al., 2001) that 51

are supported by the dynamics of the lagoon. In the lower trophic 52

levels, in particular, zooplankton is a very important biological com- 53

munity. Zooplankton is responsible for the secondary productivity 54

E

ARTICLE ING ModelECOMOD 5422 1–17

2 M. Rodrigues et al. / Ecological Mo

o55

o56

t57

a58

659

i60

61

t62

o63

d64

l65

i66

t67

t68

69

a70

d71

m72

i73

i74

i75

s76

77

A78

t79

n80

281

iQ282

(83

84

85

86

87

Q3 88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

results of an extension of the model EcoSim 2.0—ecological sim- 141

NC

OR

RE

CT

Fig. 1. Synthetic map of the Ria de Aveiro.

f the estuaries, which supports several vital functions for fishesr shrimps. The study of the factors that affect this community ishus fundamental for the management of the ecosystem (David etl., 2006). In the Ria de Aveiro, the copepoda community represents2% of the total zooplanktonic biomass, playing an important role

n the secondary production of the lagoon (Leandro, 2008).There are also several economic and social activities (e.g. indus-

ries, agriculture) in the Ria de Aveiro and it supports a populationf about 250 000 habitants (Ferreira et al., 2003). In the last fewecades, these activities have reduced the ecological quality of the

agoon (Lopes and Silva, 2006; Lopes et al., 2005). Some humannterventions, like the construction of a submarine outfall, reducedhe nutrient loads and improved its quality (Silva et al., 2000), buthere are still problems.

There is, thus, a need to develop strategies that contribute ton integrated management of the Ria de Aveiro, supported by aetailed and updated knowledge of the system and by tools andonitoring programs that improve this knowledge. In particular,

t is important to assess the impact of the human interventionsn this water system (e.g. outfall construction). Numerical models,ntegrated and validated with field data, are important tools forupporting management policies.

Most of the past ecological and water quality studies in the Ria deveiro were based on field data. The focus of these studies includehe oxygen consumption (e.g. Cunha et al., 1999), the variability of

U

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

utrients and chlorophyll a along the lagoon (e.g. Almeida et al.,005, 2007; Lopes et al., 2007), the effects of the mercury contam-

nation (Válega et al., 2008; Pato et al., 2008) and the zooplanktonLeandro et al., 2006a). The studies based on water quality and eco-

D P

RO

OF

PRESSdelling xxx (2009) xxx–xxx

logical models (e.g. Lopes et al., 2005, 2008; Saraiva, 2005; Trancosoet al., 2005; Lopes and Silva, 2006) are still scarce. The hydrodynam-ics of the lagoon has been characterized through both field data(e.g. Dias et al., 1999; Vaz and Dias, 2008) and numerical model-ing (e.g. Dias et al., 2003; Oliveira et al., 2006), but past numericalapplications were limited by insufficient horizontal or vertical res-olution. Indeed, among these studies, only one uses unstructuredgrids, which are fundamental to solve the complex geometry of thelagoon and the relevant spatial scales. However, this study, fromwhich was build the 3D model presented here, is based on a depth-averaged approach (Oliveira et al., 2006). An integrated analysisthat is able to tackle the adequate spatial scales is yet to be per-formed both to increase the knowledge of the system and to createthe basis for an operational forecast system to support the lagoon’smanagement.

The present work aims at implementing a new, fully coupled,three-dimensional hydrodynamic and ecological model, ECO-SELFE(Rodrigues et al., 2008), in the Ria de Aveiro and validate it withfield data measured in the several branches of the lagoon underdifferent environmental conditions. This model allows for the rep-resentation of the hydrodynamic and the biological processes atthe relevant time and space scales, through the use of unstruc-tured discretizations of the domain. The ecological model allowsthe simulation of the cycles of carbon, nitrogen, phosphorous, silicaand iron, and includes a site-specific formulation for zooplankton,based on the field work of Leandro et al. (2006a,b) in the Ria deAveiro. This application of ECO-SELFE constitutes its first validationin a real system.

2. Methodology

A two-stage methodology is adopted in this study. First, thehydrodynamic model is calibrated with field data, in order to estab-lish the numerical conditions of the simulations (e.g. horizontal andvertical grids, time step). This approach optimizes the computa-tional time as the coupled model is twice more CPU demandingthan the hydrodynamic module alone. The implementation of the3D hydrodynamic model builds on the work of Oliveira et al. (2006).Secondly, the fully coupled 3D hydrodynamic and ecological modelis validated. This validation is performed for two contrasting envi-ronmental conditions, also allowing the analysis of the effects of theseasonal conditions on the ecosystem dynamics. For both stages,the simulation periods were defined based on the data available forthe validation of the coupled model. Thus, the structure of the paperreflects the methodology adopted: the description of the model andits set-up for both stages are presented in the following subsections.Section 3 presents and discusses the results for the hydrodynamicmodel assessment and the ECO-SELFE model application togetherwith a preliminary analysis on the importance of the environmentalfactors; the main conclusions and the directions for further researchare summarized in Section 4.

2.1. The 3D coupled numerical model ECO-SELFE

ECO-SELFE is a fully coupled three-dimensional hydrodynamicand ecological model. The hydrodynamic model, SELFE (semi-implicit Eulerian–Lagrangian finite-element; Zhang and Baptista,2008, serial version 1.5k2, available at http://www.stccmop.org/CORIE/modeling/selfe/), solves the three-dimensional shallow-waters equations and calculates the free-surface elevation and the3D water velocity, salinity and temperature. The ecological model

sional modeling of the lower trophic levels in the Ria de Aveiro

ulation (Bisset et al., 2004, available at http://www.myroms.org/) 142

to account for the simulation of several groups of zooplankton 143

(Rodrigues et al., 2008). The model includes the cycles of carbon (C), 144

nitrogen (N), phosphorus (P), silica (Si) and iron (Fe). Besides zoo- 145

TED

PR

OO

F

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx 3

ks of

p146

b147

n148

149

n150

t151

m152

n153

T154

i155

v156

c157

i158

159

t160

161

w162

e163

164

165

166

167

168

169

170

171

172

173

174

175

NC

OR

RE

C

Fig. 2. Sources and sin

lankton, the model can simulate several phytoplankton groups,acterioplankton, dissolved and fecal organic matter, inorganicutrients and dissolved inorganic carbon (DIC).

The integrated model is based on a finite-element/finite-volumeumerical scheme (Zhang and Baptista, 2008). For salinity andemperature, advection can be treated with Eulerian–Lagragian

ethods (ELM), an upwind or a total variation diminishing (TVD)umerical scheme, while for the ecological tracers upwind andVD methods are available (Zhang and Baptista, 2008). The domain

s discretized horizontally with unstructured triangular grids andertically with hybrid coordinates (partly terrain-following S-oordinates and partly Z-coordinates), allowing for high flexibilityn both vertical and horizontal grids.

The hydrodynamic and the ecological models are integratedhrough the transport equation:( )

U

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

∂C

∂t+ u

∂C

∂x+ v

∂C

∂y+ w

∂C

∂z= ∂

∂z

∂C

∂z+ Fc + C (1)

here C is a generic tracer, (u,v,w) is the velocity, is the verticalddy diffusivity, Fc is the horizontal diffusion and C are the sources

the ecological model.

and sinks calculated with the ecological model. A general overviewof the sources and sinks of the ecological tracers is presented inFig. 2.

Since the zooplankton state variables were added to the base for-mulation of EcoSim 2.0, the equations that describe the sources andsinks of zooplankton are presented here. Zooplankton is describedin terms of carbon (ZC), nitrogen (ZN) and phosphorous (ZP) contentaccording to

ZCl = z lZCl − ez lZCl − gz lZCl (2)

ZNl = FN

FCz lZCl − ez lZNl − gz lZNl (3)

ZPl = FP

FCz lZCl − ez lZPl − gz lZPl (4)

where the subscript l refers to each functional group of zooplankton,

sional modeling of the lower trophic levels in the Ria de Aveiro

z l is the zooplankton growth rate (days−1), ez l is the zooplankton 176

excretion rate (days−1), gz l is the zooplankton mortality rate due to 177

natural mortality and predation (days−1), and FC, FN and FP are the 178

quantity of food available for zooplankton (mmol m−3) expressed in 179

carbon, nitrogen and phosphorous, respectively. Zooplankton is not 180

CTE

D P

RO

OF

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

4 M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx

f the s

d181

aQ4182

g183

o184

e185

g186

d187

a188

p189

190

i191

e192

r193

m194

a195

m196

R197

a

198

199

200

201

202

203

204

205

206

207

208

209

NC

OR

RE

Fig. 3. Location o

escribed in terms of its content of silica and iron, which is a reason-ble assumption according to Vichi et al. (2006). The zooplanktonrowth rate is calculated based on Vichi et al. (2006), dependingn the water temperature among other factors. The model has sev-ral alternative equations to calculate the temperature-dependentrowth, based on the work of Leandro et al. (2006a,b) in the Riae Aveiro. The zooplankton loss terms (excretion and predation)re constant rates defined as a site-specific or as a functional grouparameter.

The zooplankton extension of EcoSim 2.0 leads to changesn the equations that represent the sources and sinks of othercological state variables. These state variables include the oneselative to the phytoplankton groups, labile dissolved organic

U

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

atter, fecal organic matter, ammonium, inorganic phosphorousnd iron, and DIC. More detailed descriptions of the coupledodel and of each of its components can also be found in

odrigues et al. (2008), Zhang and Baptista (2008) and Bisset etl. (2004).

ampling stations.

2.2. Model set-up

2.2.1. SELFE set-upHydrodynamic simulations were performed first to establish the

numerical conditions for the three-dimensional model. These sim-ulations were based on the 2D unstructured grid simulations ofOliveira et al. (2006). Since water level, current velocity and salinitydata along the Ria de Aveiro (Fig. 3) are available for June 1997 (Diaset al., 1999), the simulations were performed for 25 days (startingon June 1, 1997) to cover this period and allowing 2 days for spin-up. The horizontal domain was discretized with an unstructuredgrid of 21 268 nodes, covering the whole lagoon and extendingabout 10 km to the coastal zone (Fig. 4a). Some salinity data (e.g.

sional modeling of the lower trophic levels in the Ria de Aveiro

Almeida et al., 2007) suggest that tides propagate further than the 210

upper limit of the Caster river considered in the domain. However, 211

the unavailability of bathymetric data prevents the extension of 212

the domain further upstream. The spatial resolution varies from 213

1.5 km in the coastal area to 3 m in the narrow channels of the 214

D

F

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx 5

bathy

l215

v216

fi217

l218

w219

220

C221

t222

m223

M224

225

a226

w227

o228

T229

S230

h231

P232

b233

r234

t235

F236

t237

t238

S239

o240

t241

t242

t243

7244

a245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

TF

R

VACBM

RR

EC

TEFig. 4. Horizontal grid (a) and

agoon. The vertical domain was discretized in seven equally spacedertical S levels, since the lagoon is shallow (Fig. 4b). A drag coef-cient of 0.002 was assumed in the whole domain. Because the

agoon is mostly composed by very narrow channels, wind stressas neglected. The time step was set to 90 s.

Five river boundaries were considered: Vouga, Antuã, Boco,aster and Mira. At the ocean boundary the model was forced byidal elevations. Eleven tidal constituents, taken from the regional

odel of Fortunato et al. (2002), were used (Z0, M2, S2, N2, O1, K1,4, MN4, MS4, MSF and M6).At the river boundaries, flow measurements are not available

t all branches for the period of simulation. Thus, these flowsere estimated based on historical data, the basins areas, mete-

rological data and previous modeling works on the Ria de Aveiro.he flows estimations were based on the time series available atNIRH (Sistema Nacional de Informacão dos Recursos Hídricos;ttp://www.snirh.pt) for the Ponte de Águeda (1935–1990) and theonte Minhoteira (1979–1989) stations, and the proportion of theasins areas determined by Saraiva (2005). With this approach, theiverine flows in Vouga, Antuã, Caster, Boco and Mira were, respec-ively, of 13 m3 s−1, 4.3 m3 s−1, 3.5 m3 s−1, 1.9 m3 s−1 and 1.9 m3 s−1.or the same period of time, Vaz (2007) set a flow of 7 m3 s−1 inhe Vouga river based on a calibration procedure. An analysis ofhe river flows and the precipitation during the period for whichNIRH data are available, suggests that for the precipitation regimebserved during June 1997 the Vouga river flow could be smaller

UN

CO

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

han 13 m3 s−1. Preliminary simulations using the 13 m3 s−1 andhe 7 m3 s−1 flows in the Vouga river were performed. Based onhe results of these simulations, the present study used a flow ofm3 s−1 in the Vouga river and reduced the other river’s flowsccordingly. Thus, the river flows at the boundaries were set as:

able 1low, salinity and temperature considered at the river boundaries for Autumn 2000 and S

iver Autumn 2000

Flow (m3/s) Salinity Temperature (C)

ouga 3.92 0 19.2ntuã 1.29 0 19.9aster 1.06 33.1 22.6oco 0.56 0 20.4ira 0.56 0 22.4

PR

OO

metry (b) of the Ria de Aveiro.

7 m3 s−1 in the Vouga, 2.3 m3 s−1 in the Antuã, 1.9 m3 s−1 in theCaster, 1 m3 s−1 in the Boco and 1 m3 s−1 in the Mira. The flows usedin the calibration of the hydrodynamic model were then assumed ascharacteristics of the environmental conditions, namely the precip-itation regimes, of June 1997. Thus, for other periods of simulationwhere river flows are not available at SNIRH, the flows were esti-mated based on a ratio between the precipitation in the period ofsimulation and the precipitation in June 1997 and the flows deter-mined previously for June 1997. The precipitation data were alsoobtained from SNIRH (station Gafanha da Nazaré).

Initial conditions of salinity were set to spatially decrease gradu-ally from 36 in the open ocean boundary to 0 in the river boundaries.Salinity was set constant at all boundaries and equal to 36 in theocean boundary and to 0 in the river boundaries.

2.2.2. ECO-SELFE set-upECO-SELFE simulations were performed for two different

periods: a period denoted as “Autumn 2000”, lasting from05/September/2000 to 04/October/2000, and a period denoted as“Spring 2001”, lasting from 01/March/2001 to 04/April/2001, andallowing 2 days for spin-up. These periods were chosen based on thechemical and ecological data available, in particular zooplanktondata, and in order to evaluate the influence of different environ-mental conditions. In the Ria de Aveiro there is a lack of zooplanktondata and one of the few studies where this data were collectedalong all the branches of the lagoon (Fig. 3) is the project Mod-

sional modeling of the lower trophic levels in the Ria de Aveiro

elRia (Universidade de Aveiro, 2003; Almeida et al., 2005; Saraiva, 271

2005). Data of salinity, temperature, chlorophyll a and nutrients 272

concentrations are also available from the ModelRia project. To 273

complement this data and validate the scalar transport of the 274

tracers along the lagoon, salinity and temperature data from the 275

pring 2001.

Spring 2001

Flow (m3/s) Salinity Temperature (C)

31.33 0 14.010.29 0 15.48.5 0 18.34.48 0 17.04.48 0 15.5

UN

CO

RR

EC

TED

PR

OO

F

Please cite this article in press as: Rodrigues, M., et al., Three-dimensional modeling of the lower trophic levels in the Ria de Aveiro(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

6 M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx

Table 2Ecological tracers input parameters.

Parameter Value References

PhytoplanktonHalf-saturation for NO3 uptake (mmol NO3 m−3) 0.824 Bisset et al. (2004)Half-saturation for NH4 uptake (mmol NH4 m−3) 0.141 Bisset et al. (2004)Half-saturation for SiO uptake (mmol SiO m−3) 1.824 Bisset et al. (2004)Half-saturation for PO4 uptake (mmol PO4 m−3) 0.0515 Bisset et al. (2004)Maximum phytoplankton 24 h growth rate (d−1) 0.4 Lalli and Parsons (1997)Base temperature for exponential growth (C) 27 Bisset et al. (2004)Phytoplankton exponential temperature factor (C−1) 0.0633 Bisset et al. (2004)Nitrate uptake inhibition for NH4 (mol−1) 1.28 Bisset et al. (2004)Maximum phytoplankton C:N ratio (mol C/mol N) 14 Bisset et al. (2004)Balanced phytoplankton C:N ratio (mol C/mol N) 6.625 Bisset et al. (2004)Absolute minimum phytop.C:N ratio (mol C/mol N) 5.500 Bisset et al. (2004)Maximum phytoplankton C:Si ratio (mol C/mol Si) 5.521 Bisset et al. (2004)Balanced phytoplankton C:Si ratio (mol C/mol Si) 5.521 Bisset et al. (2004)Absolute minimum phytop. C:Si ratio (mol C/mol Si) 4.5831 Bisset et al. (2004)Maximum phytoplankton C:P ratio (mol C/mol P) 106.0 Bisset et al. (2004)Balanced phytoplankton C:P ratio (mol C/mol P) 106.0 Bisset et al. (2004)Absolute minimum phytop. C:P ratio (mol C/mol P) 88.0 Bisset et al. (2004)Maximum quantum yield (mol C/mol quanta) 0.0833 Bisset et al. (2004)Compensation light level (mol quanta) 10.0 Bisset et al. (2004)Light level for photoinhibition (mol quanta) 10 000.0 Bisset et al. (2004)Maximum lighted limited C:Chl ratio 60.0 Bisset et al. (2004)Rate of change in light limited C:Chl ratio 0.12 Bisset et al. (2004)Minimum lighted limited C:Chl ratio 25.0 Bisset et al. (2004)Rate of change in nutrient limited C:Chl ratio 12.2 Bisset et al. (2004)Minimum nutrient limited C:Chl ratio ((g C/g Chl)−1) 60.0 Bisset et al. (2004)Rate of change in package effect ((g C/g Chl)−1) 0.01429 Bisset et al. (2004)Maximum package effect ((g C/g Chl)−1) 0.05 Bisset et al. (2004)Fraction of DOM released by phytoplankton 0.3333 Bisset et al. (2004)Fraction of fecal matter released by phytoplankton 0.3333 Bisset et al. (2004)Fraction of inorganic matter released by phytoplankton 0.3333 Bisset et al. (2004)Phytoplankton excretion rate (d−1) 0.005 Bisset et al. (2004)Phytoplankton natural mortality rate (d−1) 0.0025 Bisset et al. (2004)Refuge population (mmol C/m−3) 0.02 Bisset et al. (2004)Half-saturation for DOP uptake (mmol DOP m−3) 0.00001 Bisset et al. (2004)C:P ratio where DOP uptake begins (mol C/mol DOP) 500.0 Bisset et al. (2004)Half-saturation for DON uptake (mmol DON m−3) 0.00001 Bisset et al. (2004)C:P ratio where DON uptake begins (mol C/mol DON) 500.0 Bisset et al. (2004)Half-saturation constant DOC uptake (mmol DOC m−3) 130.0 Bisset et al. (2004)Maximum 24 h bacterial growth rate (d−1) 2.0 Bisset et al. (2004)

BacterioplanktonBase temperature for exponential growth (C) 27 Bisset et al. (2004)Bacteria exponential temperature factor (C−1) 0.092 Bisset et al. (2004)C:N ratio of bacteria (mol C/mol N) 5.0 Bisset et al. (2004)C:P ratio of bacteria (mol C/mol P) 60.0 Bisset et al. (2004)Fraction of DOM released by bacterioplankton 0.4583 Bisset et al. (2004)Fraction of fecal matter released by bacterioplankton 0.0834 Bisset et al. (2004)Fraction of inorganic matter released by bacterioplankton 0.4583 Bisset et al. (2004)Bacterial gross growth carbon efficiency 0.3 Bisset et al. (2004)Maximum nitrification rate (d−1) 0.4 Bisset et al. (2004)Half-saturation for nitrification (mmol NH4 m−3) 0.1 Bisset et al. (2004)

Fecal organic matterFecal regeneration temperature base (C) 27 Bisset et al. (2004)Fecal regeneration exponential temperature factor (C−1) 0.092 Bisset et al. (2004)Fecal carbon regeneration rate (d−1) 0.1 Bisset et al. (2004)Fecal nitrogen regeneration rate (d−1) 0.1 Bisset et al. (2004)Fecal silica regeneration rate (d−1) 0.13 Bisset et al. (2004)Fecal phosphorous regeneration rate (d−1) 0.1 Bisset et al. (2004)

ZooplanktonFraction of DOM released by zooplankton 0.25 Saraiva (2005)Fraction of fecal matter released by zooplankton 0.5 SetFraction of inorganic matter released by zooplankton 0.25 Saraiva (2005)Availability of prey to predator 0.75 SetCapture efficiency of zooplankton 1 Vichi et al. (2006)Half-saturation for total food ingestion (mmol C m−3) 1.042 Vichi et al. (2006)Assimilation efficiency of zooplankton’s predators 0.5 SetZooplankton excretion rate 0.15 Arhonditsis et al. (2000)Zooplankton mortality rate (d−1) 0.15 Arhonditsis et al. (2000)

D

ARTICLE ING ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Mo

Table 3Boundary conditions of the ecological tracers at the ocean boundary.

Parameter Autumn/2000 Spring/2001

Zooplankton C (mg C/l) 0.017 0.008Zooplankton N (mg N/l) 0.0020 0.0008Zooplankton P (mg P/l) 0.0003 0.0002Phytoplankton C (mg C/l) 0.15 0.10Phytoplankton N (mg N/l) 0.025 0.018Phytoplankton P (mg P/l) 0.0037 0.0025Phytoplankton Si (mg Si/l) 0.06 0.04Chlorophyll a (g C/l) 2.43 1.68Bacterioplankton C (mg C/l) 0.01 0.01Bacterioplankton N (mg N/l) 0.002 0.002Bacterioplankton P (mg P/l) 0.0004 0.0004DOC (mg C/l) 24.35 45.28DON (mg N/l) 0.050 0.050DOP (mg P/l) 0.0332 0.0034Fecal organic C (mg C/l) 7.31 13.58Fecal organic N (mg N/l) 0.015 0.015Fecal organic P (mg P/l) 0.0099 0.0009Fecal organic Si (mg Si/l) 0.03 0.03NH4

+ (mol N/l) 4.80 3.38N −

PSD

C276

t277

278

t279

a280

p281

a282

a283

284

w285

s286

b287

b288

a289

t290

o291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

TB

P

ZZZPPPPCBBBDDDFFFFNNPSD

O3 (mol N/l) 8.13 32.65O4

3− (mol P/l) 2.50 0.88iO2 (mol Si/l) 23.90 34.70IC (mg C/l) 24.00 24.00

aracterizacão Sinóptica da Ria de Aveiro (CSRA) project (Fig. 3) forhe 2000 and 2001 years were also considered.

The numerical conditions used were the ones that resulted fromhe calibration of the hydrodynamic model, except for the bound-ry conditions that were changed to correspond to the simulatederiods. The river flows (Table 1) were determined based on thepproach described above, since, as for June 1997, data were notvailable.

The temperature and the salinity boundary conditions (Table 1)ere first determined as the average of the CSRA field data mea-

ured in each of the two periods at the stations closest to each

UN

CO

RR

EC

TE

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

oundary. However, as the stations are not located exactly at eachoundary, a sensitivity analysis on the temperature was performedt the boundaries, starting with the values measured at the sta-ions. A backward compatibilization was then performed basedn the differences between the results measured and computed

able 4oundary conditions of the ecological tracers at the river boundaries.

arameter Autumn/2000

Vouga Antuã Caster Boco

ooplankton C (mg C/l) 0.014 0.0140 0.003 0.004ooplankton N (mg N/l) 0.0015 0.0011 0.0003 0.0004ooplankton P (mg P/l) 0.0003 0.0002 0.0001 0.0001hytoplankton C (mg C/l) 0.35 0.34 0.23 0.24hytoplankton N (mg N/l) 0.061 0.059 0.0410 0.041hytoplankton P (mg P/l) 0.0090 0.0087 0.0059 0.0059hytoplankton Si (mg Si/l) 0.15 0.14 0.10 0.10hlorophyll a (g C/l) 5.80 5.60 3.80 3.93acterioplankton C (mg C/l) 0.01 0.01 0.01 0.01acterioplankton N (mg N/l) 0.002 0.002 0.002 0.002acterioplankton P (mg P/l) 0.0004 0.0004 0.0004 0.0004OC (mg C/l) 6.72 14.12 10.75 58.18ON (mg N/l) 0.190 0.230 0.135 0.305OP (mg P/l) 0.1262 0.1525 0.0896 0.2024ecal organic C (mg C/l) 2.02 4.24 3.23 3.35ecal organic N (mg N/l) 0.057 0.069 00.0 0.091ecal organic P (mg P/l) 0.0378 0.0459 0.060 0.0608ecal organic Si (mg Si/l) 0.11 0.14 0.08 0.18H4

+ (mol N/l) 4.25 4.35 6.05 7.27O3

−(mol N/l) 102.70 22.65 6.90 20.18O4

3− (mol P/l) 1.70 2.32 2.43 2.90iO2 (mol Si/l) 174.33 31.35 35.35 34.13IC (mg C/l) 24.00 24.00 24.00 24.00

PR

OO

F

PRESSdelling xxx (2009) xxx–xxx 7

at each station where data are available. Initial conditions wereset based on the average values on the CSRA field data for eachperiod.

A total of 23 ecological tracers were considered: one group ofzooplankton (as C, N and P), one group of phytoplankton, namelydiatoms (as C, N, P, Si and chlorophyll a), bacterioplankton (as C, Nand P), labile dissolved organic matter (as C, N and P), fecal organicmatter (as C, N, P and Si), NH4

+ (which also includes NO2−), NO3

−,PO4

3−, SiO2 and DIC. These were chosen based on the available dataand also as a compromise between the number of tracers to sim-ulate and the computational time needed. The iron cycle was notconsidered in this simulation, since iron data are not available forthe simulated periods. The input parameters of the ecological modelare listed in Table 2.

For the ecological tracers where field data were available, theinitial and boundary conditions were set based on the ModelRiadata. These tracers include: carbon zooplankton, carbon phyto-plankton, chlorophyll a, DOC, NH4

+, NO3−, PO4

3−, SiO2. Since nodata was available at the initial time of the simulations, for Autumn2000 these values were determined as an average between thefield data of June 2000 and September 2000, while for Spring 2001the average values of December 2000 and March 2001 were used.The phytoplankton values were determined based on the ratio tochlorophyll a proposed by Portela (1996): 60 mg C/mg chlorophylla. For the variables where data are not available, the initial andboundary conditions were estimated based on analytical initial con-ditions of the EcoSim 2.0 model or other modeling applicationsin the Ria de Aveiro (e.g. Saraiva, 2005). Zooplankton as N and Pwas estimated based on the carbon zooplankton measures. Phy-toplankton as N, P and Si were based on the carbon to nutrientsratios of the analytical initial conditions of EcoSim 2.0 (availableat http://www.myroms.org). For the bacterioplankton and the DICvariables, the analytical conditions of EcoSim 2.0 were also used,and fecal organic silica was assumed equal to fecal organic nitro-gen according to these analytical conditions. The dissolved organic

sional modeling of the lower trophic levels in the Ria de Aveiro

nitrogen (DON) was estimated based on the values presented in the 327

Ria de Aveiro application of Saraiva (2005). The dissolved organic 328

phosphorous (DOP) was determined based on the inorganic N/P 329

ratio of the measured values in the ModelRia project and the partic- 330

ulate organic matter was considered as 30% of the dissolved organic 331

Spring/2001

Mira Vouga Antuã Caster Boco Mira

0.007 0.014 0.0140 0.009 0.011 0.0140.0008 0.0015 0.001 0.00150 0.001 0.00150.0002 0.0003 0.0002 0.0002 0.0002 0.00030.45 0.03 0.16 0.14 0.14 0.170.079 0.006 0.028 0.024 0.025 0.0290.0115 0.0009 0.0040 0.0034 0.0037 0.00430.19 0.01 0.07 0.06 0.06 0.077.48 0.58 2.63 2.28 2.35 2.780.01 0.01 0.01 0.01 0.01 0.010.002 0.002 0.002 0.002 0.002 0.0020.0004 0.0004 0.0004 0.0004 0.0004 0.0004

17.65 8.18 11.12 37.35 23.48 21.300.150 0.190 0.230 0.135 0.305 0.1500.0995 0.0127 0.0152 0.0090 0.0202 0.00995.30 2.45 3.34 11.21 7.04 6.430.045 0.057 0.069 00.0 0.091 0.0450.0298 0.0037 0.0047 0.0028 0.0059 0.00280.09 0.11 0.14 0.08 0.18 0.094.65 2.60 9.53 5.25 9.13 8.13

111 92.85 95.93 83.18 66.18 192.952.40 1.05 1.01 1.10 2.30 3.48

45.55 20.88 38.50 49.48 117.70 95.0824.00 24.00 24.00 24.00 24.00 24.00

INE

8 cal Mo

m332

(333

i334

d335

F(

ARTICLEG ModelCOMOD 5422 1–17

M. Rodrigues et al. / Ecologi

UN

CO

RR

EC

TE

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

ater, which was estimated based on the application of Saraiva2005). The boundary conditions are listed on Tables 3 and 4. Thenitial conditions were set as spatially varying based on the valuesetermined for each boundary.

ig. 5. Comparison between data (Dias et al., 1999) and SELFE results for water level in Jue) Varela, (f) Muranzel, (g) Costa Nova and (h) Vagueira.

PRESSdelling xxx (2009) xxx–xxx

D P

RO

OF

sional modeling of the lower trophic levels in the Ria de Aveiro

For the calculation of the spectral irradiation, the atmospheric 336

parameters were considered constant in each of the two periods 337

simulated and were estimated based on the values measured at a 338

meteorological station located in the University of Aveiro (Table 5). 339

ne/1997: (a) Friopesca, (b) Vista Alegre, (c) Espinheiro/Cais do Bico, (d) Miradouro,

UN

CO

RR

EC

TED

PR

OO

F

Please cite this article in press as: Rodrigues, M., et al., Three-dimensional modeling of the lower trophic levels in the Ria de Aveiro(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx 9

Fig. 6. Comparison between data (Dias et al., 1999) and SELFE results for velocity in June/1997: (a) Friopesca, (b) Vista Alegre, (c) Espinheiro/Cais do Bico, (d) Miradouro, (e)Varela, (f) Muranzel, (g) Costa Nova and (h) Vagueira.

UN

CO

RR

EC

TED

PR

OO

F

Please cite this article in press as: Rodrigues, M., et al., Three-dimensional modeling of the lower trophic levels in the Ria de Aveiro(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

10 M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx

Fig. 7. Comparison between data (Dias et al., 1999) and SELFE results for salinity in June/1997: (a) Friopesca, (b) Vista Alegre, (c) Espinheiro/Cais do Bico, (d) Miradouro, (e)Varela, (f) Muranzel, (g) Costa Nova and (h) Vagueira.

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx 11

Fig. 8. Influence of the river flow in the Mira channel:

Table 5Atmospheric conditions considered for the calculation of the spectral irradiation.

Parameter Autumn 2000 Spring 2001

Wind 6.7 m/s, N 5.2 m/s, SWAir temperature 18.5 C 12.5 CARC

3340

3341

342

S343

m344

345

346

347

348

349

350

351

352

353

354

355

356

able for a detailed spatial characterization. Recent studies also 357

F

tmospheric pressure 1021.0 mbar 1014.75 mbarelative humidity 85.6% 85.4%loud cover 0.2 0.5

. Results and discussion

.1. SELFE simulations assessment

UN

CO

RR

EC

TED

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

These simulations were performed for the June 1997 period.imulations to calibrate the horizontal and vertical grids’ refine-ent, time step and the drag coefficient were done previously. The

ig. 9. Comparison between data (CSRA data, 13 h time series) and ECO-SELFE results for

PR

OO

F

(a) Costa Nova station and (b) Vagueira station.

following validation analysis is based on the best results from thecalibration procedure.

Water levels phase are represented by the model along thebranches of the lagoon (Fig. 5). In terms of amplitude, the model rep-resents the data with average errors smaller than 10–15 cm (Fig. 5).In a few stations (e.g. Vista Alegre, Fig. 5b; Vagueira, Fig. 5h) themodel tends to underestimate by 25–50 cm the elevations at lowtide. These differences may be due several factors. The boundaryconditions, which were taken from a regional model, have someerrors as they do not include all frequencies. The drag coefficientmay also be a source of errors, since a constant value in the wholedomain is considered. However, adequate information is unavail-

sional modeling of the lower trophic levels in the Ria de Aveiro

suggest that bathymetric changes in the inlet channel affect the 358

response of the M2 constituent (Araújo et al., 2008). Since the 359

bathymetry that is used in the model combines data from 1987 360

and 2004 and is not from the June 1997 period this may also con- 361

saliniity in Spring 2001: (a) Vagueira, (b) Varela, (c) Vista Alegre and (d) Vouga.

INE

1 cal Mo

t362

a363

i364

g365

i366

t367

s368

r369

b370

m371

372

e373

e374

e375

A376

377

378

379

380

381

382

383

384

385

386

FmV

ARTICLEG ModelCOMOD 5422 1–17

2 M. Rodrigues et al. / Ecologi

ribute for the differences observed. There are also several wettingnd drying areas in the lagoon poorly represented in the grid, due tots shallowness, in particular in the Mira channel. An increase in therid resolution in this channel could also lead to an improvementn the water levels results. Tests done in the Mira channel showedhat refining the grid resolution by splitting the elements by fourignificantly improved the water levels representation. However,efining the grid considerably increases the CPU times, which maye limiting when running the coupled hydrodynamic-ecologicalodel.

UN

CO

RR

EC

TE

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

Velocities phase differences between model and data are gen-rally smaller than 5–10 min along the lagoon (Fig. 6). Magnituderrors are generally smaller than 5 cm/s (Fig. 6). The largest differ-nces between the model results and field data occur in the Vistalegre (Ílhavo channel) and Varela (S. Jacinto channel) stations, but

ig. 10. Comparison between data (CSRA data, point measurements along the four branchain branches: (a) Autumn 2000 and (b) Spring 2001. Mira channel: M1, M7, M13 and M113 stations. S. Jacinto channel: O1, O7, O13 and C2 stations.

PRESSdelling xxx (2009) xxx–xxx

these are generally smaller that 10–15 cm/s. These differences maybe due to the bathymetry data that are used, which, as mentionedbefore, are not from June 1997.

Salinities (Fig. 7) are represented by the model for most of thestations with average differences smaller than 5. At the Varelastation (upper northern position of the S. Jacinto) the model under-estimates salinity by 5–10, which may be due to the uncertaintyassociated with the river flows at the Caster river boundary. Sincea salinity of zero was imposed here, the proximity of this stationto the nearby boundary suggests that the grid should be extended

D P

RO

OF

sional modeling of the lower trophic levels in the Ria de Aveiro

further north, to the limit of tidal propagation. However, the lack 387

of bathymetric data prevents the extension of the domain. The 388

largest salinities errors, of about 15, were observed in the Mira chan- 389

nel (Costa Nova station, Fig. 7h). These errors may be due to the 390

river flow boundary conditions, which present an important uncer- 391

es of the lagoon) and ECO-SELFE results for salinity and temperature along the four6 stations. Ílhavo channel: I1, I7 and I13 stations. Espinheiro channel: V1, V5, V9 and

O

ARTICLE ING ModelECOMOD 5422 1–17

M. Rodrigues et al. / Ecological Mo

Table 6Relative deviation (R) between data and model results for salinity and temperatureduring Autumn 2000 and Spring 2001. Data variation range and mean value (inparentheses) are presented.

Station Salinity Temperature (C)

Data R Data R

Autumn 2000Vagueira 16.2–30.1 (22.2) 2.2 18.3–20.6 (19.8) 1Vista Alegre – – –Vouga – – –Varela 31.5–34.3(33.1) 6.6 18.2–20.1 (19.2) 0.8

Spring 2001Vagueira 0.4–10.7 (2.4) 1.7 15.3–17.3 (16.6) 1.2Vista Alegre 0.4–11.5 (3.9) 1.2 14.3–16.2 (15.2) 1.2VV

t392

u393

t394

t395

t396

(397

f398

f399

T400

c401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

and Ci is the tracer predicted value. 420

Fma

ouga 0.1–19.4 (6.6) 3.3 13.5–14.7 (13.1) 0.2arela 0.5–6.1 (3.0) 1.7 15.8–18.6 (17.2) 1.1

ainty (Dias and Lopes, 2006). Sensitivity simulations performedsing 5 m3 s−1 (Vaz, 2007) and 3 m3 s−1 (Saraiva, 2005) show thathere is a significant influence of the river flow in the salinity alonghis channel. The use of a larger flow reduces the salinity errors athe Costa Nova station, but increases them at the Vagueira stationFig. 8). This behavior suggests the existence of an extra source of

UN

CO

RR

EC

TED

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

reshwater between these two stations during this period, with sur-ace or groundwater origin, that was not considered in the model.he analysis of satellite images suggests that agricultural dischargehannels may be one of the sources of this freshwater.

ig. 11. Average values of the ecological tracers in the Ria de Aveiro in (a) Autumn 2000odel results and the field data. The range of variation of each parameter is defined in th

s the minimum value measured in the period. All the values were standardized by the av

OF

PRESSdelling xxx (2009) xxx–xxx 13

Globally, the results suggest that the model is able to reproducethe hydrodynamics along all the branches of the Ria de Aveiro. Themagnitude of the observed differences in water levels, velocitiesand salinities is similar or smaller than those achieved in previousapplications in the lagoon for the same period (e.g. Vaz, 2007). Inparticular, the velocity field, which is fundamental for a good rep-resentation of the scalar transport within the ecological model, issimulated by the model with errors of only 5 cm/s.

3.2. ECO-SELFE validation

Results for salinity and temperature of ECO-SELFE simulationsare compared with the CSRA data, while ecological tracer’s resultsare compared with the ModelRia data. Results are presented onlyfor the ecological tracers for which data are available: zooplankton,phytoplankton (estimated from chlorophyll a), chlorophyll a, NH4

+,NO3

−, PO43−, SiO2 and DOC. The relative deviation (R) between

the model results and the data was calculated as

R = 1N

N∑i=1

|Xi − Ci| (5)

where N is the total number of observations, Xi is the observed value

PR

sional modeling of the lower trophic levels in the Ria de Aveiro

3.2.1. Salinity and temperature 421

Salinity variations during the tidal cycle are represented by the 422

model with amplitude errors ranging between 2 and 10 in the 423

(a1) with zooplankton detail (a2) and (b) Spring 2001: comparison between thee upper limit by the maximum value measured in the period and in the lower limiterage value of the field data.

INE

1 cal Mo

A424

p425

t426

s427

d428

i429

t430

s431

r432

t433

d434

1435

u436

m437

d438

c439

440

s441

r442

t443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

F(

ARTICLEG ModelCOMOD 5422 1–17

4 M. Rodrigues et al. / Ecologi

utumn 2000 (Table 6) and 1–4 in the Spring 2001 (Table 6), whilehase errors are negligible (Fig. 9, for Spring 2001). In both periodshe horizontal variation of salinity in the lagoon is globally repre-ented by the model with differences between the model and theata of about 1–5 (Fig. 10). Exceptions are the upstream stations

n the S. Jacinto and Espinheiro channels in the Autumn 2000 (sta-ions C2 and V13, Fig. 10a), with errors of about 10. In the presentimulation, a very large value of salinity was used for the Casteriver boundary for Autumn 2000 based on the field data for the sta-ion C2, but some differences between the coupled model and theata remain. As observed for the SELFE calibration simulation (June997), the salinity data from C2 suggest that salinity propagatespstream of the computational domain. The model also overesti-ates salinity at the M7 and, to a smaller extent, the M1 stations

uring Spring 2001. Errors in the riverine boundary conditions may

UN

CO

RR

EC

TE

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

ause the observed differences.Temperature is represented by the model with errors generally

maller than 1.5 C in both periods (Table 6 and Fig. 10). The modelepresentation of both salinity and temperature horizontal varia-ions along the branches of the lagoon, which is within the range

ig. 12. Comparison between data and ECO-SELFE results for the ecological tracers in ModECO-SELFE (TC)—model average values in a tidal cycle).

PRESSdelling xxx (2009) xxx–xxx

of data variance measured, confirms the ability of the model torepresent the scalar transport in the Ria de Aveiro.

3.2.2. Ecological tracersEcological tracers’ model results (averaged over the whole Ria

de Aveiro and for all ModelRia stations) are first compared with theaverage values of the data measured in the ModelRia stations andtheir range of variation, providing a general overview of the modelbehavior. In the Autumn 2000 period, although the magnitude ofthe model results for most ecological tracers is roughly within therange of the field data, some tracers are outside this range (Fig. 11a).Zooplankton, in particular, is overestimated by the model in thisperiod. In Spring 2001 the model results fit within the range ofvariation of the field data for all the ecological tracers evaluated,showing that the model is able to represent these tracers in the Ria

D P

RO

OF

sional modeling of the lower trophic levels in the Ria de Aveiro

de Aveiro during this period (Fig. 11b). 458

Table 7 summarizes the relative deviation between the observed 459

and the predicted values. These deviations were calculated consid- 460

ering the model results in the instant of the observations and also 461

considering the average of the model results during a tidal cycle 462

elRia stations (MR1, MR2, MR3, MR5, MR6 and MR7) for the period of Autumn 2000

INE

cal Mo

(463

r464

e465

466

m467

2468

a469

a470

i471

f472

A473

f474

r475

d476

(477

s478

t479

t480

i481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

F(

ARTICLEG ModelCOMOD 5422 1–17

M. Rodrigues et al. / Ecologi

neap tide/spring tide). The later approach reduces the uncertaintyelative to the exact time of the measurements and some phaserrors that derive from the hydrodynamic model.

A more detailed analysis shows that zooplankton is overesti-ated by the model in all ModelRia stations during the Autumn

000 (Fig. 12). The differences between the model and the datare of 0.08–0.1 mg C/l, which corresponds to an overestimation byfactor of 10 when compared with the data. Although the error

s significant, the agreement between model and data comparesavorably with other published modeling studies in the Ria deveiro for the same period, where data were overestimated by a

actor of 20–100 (Saraiva, 2005). In the Spring 2001 the modelepresents zooplankton concentrations in the lagoon with smallerifferences of only 0.005 mg C/l, when considering the tidal cycleTable 7). During this period the model is also able to reproduce the

UN

CO

RR

EC

TED

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

patial variation of zooplankton, representing the larger concentra-ions observed in the Mira channel (MR7 station, Fig. 13). Althoughhe model represents the mean concentrations of zooplankton,t fails to predict the amplitude of variation of zooplankton in

ig. 13. Comparison between data and ECO-SELFE results for the ecological tracers in ModECO-SELFE (TC)—model average values in a tidal cycle).

PRESSdelling xxx (2009) xxx–xxx 15

Mira channel, where the largest concentrations of zooplankton areobserved during flood, which might be a punctual situation with amarine origin. This difference may derive from the lack of data atthe marine boundary to impose adequate boundary conditions.

As observed for zooplankton, phytoplankton is better repre-sented by the model in the Spring 2001 than in the Autumn 2000(Table 7). Considering the tidal cycle, these differences are of about0.06 mg C/l (30%) in Spring 2001 and 0.1 mg C/l (60%) in Autumn2000 when phytoplankton is underestimated by the model. Thelarger differences are observed in the MR3 station in both periods(Figs. 12 and 13) and in MR7 in Autumn 2000 (Fig. 12). The resultsobserved for phytoplankton are similar to those achieved for chloro-phyll a, which is predicted by the model with average deviations ofabout 1.5 g C/l (Table 7, Figs. 12 and 13).

For DOC the model tends to overestimate the data by about

PR

OO

F

sional modeling of the lower trophic levels in the Ria de Aveiro

10–12 mg C/l in the Autumn period (Table 7). The larger differences 497

being observed in the station located near the Antuã river (MR3 498

station, Fig. 12). In Spring 2001 the deviations are smaller, about 499

40% relative to the average data (Table 7). In this period the largest 500

elRia stations (MR1, MR2, MR3, MR5, MR6 and MR7) for the period of Autumn 2000

E

ARTICLE IN PRESSG ModelECOMOD 5422 1–17

16 M. Rodrigues et al. / Ecological Modelling xxx (2009) xxx–xxx

Table 7Relative deviation (R) between data and model results for ecological tracers duringAutumn 2000 and Spring 2001. Data variation range and mean value (in parentheses)are presented. Relative deviations are calculated considering the MR1, MR2, MR3,MR5, MR6 and MR7 stations. R (TC) refers to the relative deviations calculated withthe model average values in a tidal cycle.

Ecological tracer Data R R (TC)

Autumn 2000Zooplankton (mg C/l) 0.01–0.12 (0.006) 0.08 0.10Phytoplankton (mg C/l) 0.10–0.34 (0.18) 0.15 0.10Chlorophyll a (g C/l) 1.7–5.6 (3.1) 1.9 1.1DOC (mg C/l) 7.6–12.7 (10.3) 10.5 12.5NH4

+ (mol N/l) 4.5–9.6 (6.1) 4.8 4.8NO3

− (mol N/l) 7.4–18.5 (11.6) 6.8 6.4PO4

3− (mol P/l) 2.9–4.2 (3.5) 0.8 0.8SiO2 (mol Si/l) – – –

Spring 2001Zooplankton (mg C/l) 0.001–0.04 (0.009) 0.009 0.005Phytoplankton (mg C/l) 0.06–0.33 (0.20) 0.07 0.06Chlorophyll a (g C/l) 0.9–5.5 (3.7) 1.4 1.4DOC (mg C/l) 6.4–75.3 (28.1) 11.3 10.4NH4

+ (mol N/l) <2–8.8 (4.2) 1.2 0.8NPS

d501

u502

S503

t504

t505

506

t507

2508

o509

4510

d511

a512

e513

a514

b515

t516

D517

w518

r519

w520

P521

r522

c523

f524

a525

526

d527

p528

t529

t530

f531

m532

t533

o534

w535

t536

s537

i538

p539

t540

q541

t542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

there is a lack of data. 581

NC

OR

RE

CT

O3− (mol N/l) 5.0–135.5 (61.3) 27.7 29.1

O43− (mol P/l) <1.0–3.4 (1.5) 0.2 0.01

iO2 (mol Si/l) 10.0–113.5 (50.4) 19.0 14.6

ifferences are observed in the MR5 station, where the model isnable to reproduce the amplitude of variation of the data (Fig. 13).ince there is a lack of data to establish the boundaries concentra-ions during the periods simulated, this uncertainty may contributeo the observed errors.

Regarding nutrients, the model is also able to represent theseracers with smaller differences in Spring 2001 than in Autumn000. NH4

+ tends to be underestimated by the model through-ut the lagoon during Autumn 2000. The errors are of about.8 mol N/l, which correspond to a difference of about 80% to theata (Table 7). The underestimation of NH4

+ as been observed innother model application to the lagoon in the same period (Lopest al., 2008). In the Spring 2001, the differences between modelnd data are smaller, of about 20% (Table 7). PO4

3− is representedy the model with deviations smaller than 0.85 mol P/l (25%) inhe Autumn 2000 and than 0.02 mol P/l (15%) in the Spring 2001.uring the Spring 2001 these differences are very low, of about 1%,hen considering the results for the tidal cycle. The model also

epresents well the seasonal variation of some variables, like NO3−,

hich is larger in the Spring 2001 than in the Autumn 2000, andO4

3−, which is larger in the Autumn 2000 (Figs. 12 and 13). SiO2 isepresented in the Spring 2001 with differences of about 30%, whichorresponds to an error of about 15–19 mol Si/l. The observed dif-erences may be due to several factors, including the uncertaintyssociated with the boundary conditions, as mentioned above.

The larger errors observed in the Autumn 2000 may alsoerive from the phytoplankton–zooplankton dynamics. During thiseriod, the phytoplankton biomass increases in the beginning ofhe simulations, which leads to a growth of zooplankton, and theno reduction in phytoplankton. Since the zooplankton’s growth isueled by food availability, the overestimation of phytoplankton

ay contribute to the increase of zooplankton and to the reduc-ion of NH4

+ predicted by the model. The following zooplanktonverestimation leads to the model phytoplankton underestimationhen compared with data. The phytoplankton overestimation in

he beginning of the simulations may derive from the modeledtructure of the ecosystem, since only one primary producer groups considered (phytoplankton). Indeed, another important primary

U

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

roducer (macroalgae) may compete for the available resources,hereby reducing the growth of the phytoplankton and, conse-uently, the available food for zooplankton. However, increasinghe complexity of the ecosystem representation would also increase

D P

RO

OF

Fig. 14. Influence of the phytoplankton temperature-dependent maximum growthrate in the phytoplankton concentration at the MR7 station (Spring 2001).

the uncertainty associated with the parameterization. Additionally,data are unavailable for other primary producers. Several parame-ters considered in the establishment of the ecological model are alsosources of errors and uncertainty, since site-specific information fortheir establishment is unavailable in the Ria de Aveiro. Sensitivityanalysis performed on the ecological model parameters (Rodrigueset al., 2008, 2009) showed the relative influence of these param-eters in the final results of the model. Among these parameters,the phytoplankton temperature-dependent growth rate was oneof the most relevant. Since the zooplankton growth depends onthe quantity of food available, this parameter will also affect zoo-plankton concentration. The parameters related to food ingestionby zooplankton and its excretion and mortality rates also affect theresults significantly. Several tests in the Ria de Aveiro showed theinfluence of these parameters in a real system (Fig. 14) and the needto obtain more site-specific data.

4. Conclusions

The reliability of a new fully coupled three-dimensional,unstructured grid, hydrodynamic and ecological model (ECO-SELFE) was demonstrated in an application to the Ria de Aveiro,using a site-specific formulation for zooplankton. The applicationof the coupled model was its first application in a real system andallowed the model validation with a reasonably adequate set of fielddata. The application also allowed the evaluation of the differentenvironmental conditions in the dynamics of the lagoon.

First, simulations using the hydrodynamic model (SELFE) alonewere done, for efficiency. These simulations showed the good per-formance of the model in the reproduction of the water levels,velocities and salinities in the Ria de Aveiro. In particular, thevelocity field, which is fundamental for the correct simulation ofthe transport processes, is represented by the model with errorssmaller than 5 cm/s. These simulations also showed the importanceof using accurate boundary conditions, namely for the freshwaterinputs in the lagoon. Although Ria de Aveiro is a very impor-tant estuarine system, with ecological and economical values, riverflows data are scarce. This question needs to be addressed in futurestudies in order to establish adequate methodologies to determineor estimate the boundary conditions for freshwater inputs when

sional modeling of the lower trophic levels in the Ria de Aveiro

The simulations performed with ECO-SELFE for the Autumn 582

2000 and Spring 2001 periods showed that globally the model 583

reproduces the ecological dynamics along the branches of the Ria 584

during the Spring, fitting the model results inside the range of 585

D

INE

cal Mo

d586

t587

p588

e589

d590

a591

p592

593

c594

fi595

p596

i597

w598

A599

a600

t601

m602

603

u604

l605

o606

a607

UQ5608

609

A610

611

E612

e613

m614

c615

616

e617

m618

R619

A620

621

622

A623

624

A625

626

A627

628

629

630

B631

632

C633

634

635

D636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

RE

CTE

ARTICLEG ModelCOMOD 5422 1–17

M. Rodrigues et al. / Ecologi

ata variation. In the Autumn period, the model tends to overes-imate the zooplankton concentrations by a factor of 10. In thiseriod, phytoplankton and ammonium concentrations are under-stimated. Although some differences between the model and theata remain, the results achieved here compare similarly or favor-bly with other studies in the same periods (Saraiva, 2005), inarticular for zooplankton.

Possible explanations for these differences include the boundaryonditions used and the parameterizations considered. The simpli-cation of the ecosystem structure, namely in terms of the primaryroducers, may also lead to the observed differences. The availabil-

ty of more ecological data, namely with longer temporal coverage,ill also be useful to perform more specific validations of the model.dditional exploitation of the model for different scenarios willllow a more detailed study of the environmental factors affectinghe zooplankton dynamics, contributing for the lagoon manage-

ent.Due to the complexity of the model and the spatial resolution

sed, CPU times are a limiting factor. Therefore, for complex simu-ations and longer simulation periods the use of the parallel versionf the model, which was recently developed, is essential to achievecceptable computational times.

ncited reference

Pina (2001).

cknowledgements

This work was partially funded by the Laboratório Nacional dengenharia Civil, project “Hidrodinâmica e Transporte em EstuáriosLagunas”, and by the Luso-American Foundation for Develop-ent, project “Towards a nowcast-forecast system for estuarine and

oastal water quality”.The authors would like to thank Prof. Vanda Brotas for reviewing

arly versions of this manuscript and Prof. Maria Dolores Manso foraking available the meteorological data used in the simulations.

eferences

lmeida, M.A., Cunha, M.A., Alcântara, F., 2005. Relationship of bacterioplankton pro-duction with primary production and respiration in a shallow estuarine system(Ria de Aveiro, NW Portugal). Microbiol. Res. 160, 315–328.

lmeida, M.A., Cunha, M.A., Dias, J.M., 2007. Bacterial productivity distribution dur-ing a rainy year in an estuarine system. Microb. Ecol. 53, 208–220.

raújo, I.B., Dias, J.M., Pugh, D.T., 2008. Model simulations of tidal changes in a coastallagoon, the Ria de Aveiro (Portugal). Cont. Shelf Res. 28, 1010–1025.

rhonditsis, G., Tsirtsis, G., Angelidis, M.O., Karydis, M., 2000. Quantification of theeffects of nonpoint nutrient sources to coastal marine eutrophication: appli-cations to a semi-enclosed gulf in the Mediterranean Sea. Ecol. Model. 129,209–227.

isset, W.P., Debra, S., Dye, D., 2004. Ecological Simulation (EcoSim) 2.0 TechnicalDescripton. Florida Environmental Research Institute, FL, USA.

UN

CO

R

Please cite this article in press as: Rodrigues, M., et al., Three-dimen(Portugal). Ecol. Model. (2009), doi:10.1016/j.ecolmodel.2009.02.002

unha, M.A., Almeida, M.A., Alcântara, F., 1999. Compartments of oxygen consup-tion in a tidal mesotrophic estuary (Ria de Aveiro, Portugal). Acta Oecol. 20,227–235.

avid, V., Sautour, B., Galois, R., Chardy, P., 2006. The paradox high zooplanktonbiomass—low vegetal particulate organic matter in high turbidity zones: whatway for energy transfer? J. Exp. Mar. Biol. Ecol. 333, 202–218.

PR

OO

F

PRESSdelling xxx (2009) xxx–xxx 17

Dias, J.M., Lopes, J.F., 2006. Implementation and assessment of hydrodynamic, saltand heat transport models: the case of Ria de Aveiro lagoon (Portugal). Environ.Model. Softw. 21, 1–15.

Dias, J.M., Lopes, J.F., Dekeyser, I., 1999. Hydrological characterisation of Ria de Aveiro,Portugal, in early summer. Oceanol. Acta 22, 473–485.

Dias, J.M., Lopes, J.F., Dekeyser, I., 2000. Tidal propagation in Ria de Aveiro lagoon,Portugal. Phys. Chem. Earth 25, 369–374.

Ferreira, J.G., Simas, T., Nobre, A., Silva, M.C., Shifferegger, K., Lencart-Silva, J., 2003.Identification of sensitive areas and vulnerable zones in transitional and coastalportuguese systems. INAG, Lisbon, Portugal, 151 pp.

Fortunato, A.B., Pinto, L.L., Oliveira, A., Ferreira, J.S., 2002. Tidally generated shelfwaves off the western Iberian coast. Cont. Shelf Res. 22 (14), 1935–1950.

Hermoso, J., Martins, F., Coelho, C., 2001. Classificacão e Gestão da Área de PaisagemProtegida da Foz do Cáster no âmbito do Programa LIFE ESGIRA-MARIA. In: VIIConferência Nacional sobre a Qualidade do Ambiente, Portugal.

Lalli, C.M., Parsons, T.R., 1997. Biological Oceanography: An Introduction. The OpenUniversity, 314 pp.

Leandro, S.M., 2008. Environmental forcing of an estuarine gradient of zooplanktonabundance and production. PhD Thesis, Universidade de Aveiro, Aveiro, Portugal.

Leandro, S.M., Queiroga, H., Rodríguez-Grana, L., Tiselius, P., 2006a. Temperature-dependent development and somatic growth in two allopatric populations ofAcartia clausi (Copepoda:Calanoida). Mar. Ecol. Progr. Ser. 322, 189–197.

Leandro, S.M., Tiselius, P., Queiroga, H., 2006b. Growth and development of nau-plii and copepodites of the estuarine copepod Acartia tonsa from southernEurope (Ria de Aveiro, Portugal) under saturating food conditions. Mar. Biol.150, 121–129.

Lopes, C.B., Lillebo, A.I., Dias, J.M., Pereira, E., Vale, C., Duarte, A.C., 2007. Nutrientdynamics and seasonal succession of phytoplankton assemblages in a SouthernEuropean Estuary: Ria de Aveiro, Portugal. Estuar. Coast. Shelf Sci. 71, 480–490.

Lopes, J.F., Silva, C., 2006. Temporal and spatial distribution of the dissolved oxygenin the Ria de Aveiro lagoon. Ecol. Model. 197, 67–88.

Lopes, J.F., Dias, J.M., Cardoso, A.C., Silva, C.I.V., 2005. The water quality of the Riade Aveiro lagoon, Portugal: from the observations to the implementation of anumerical model. Mar. Environ. Res. 60, 594–628.

Lopes, J.F., Silva, C.I., Cardoso, A.C., 2008. Validation of a water quality model for theRia de Aveiro lagoon, Portugal. Environ. Modell. Softw. 23, 479–494.

Oliveira, A., Fortunato, A.B., Dias, J.M., 2006. Numerical modeling of the Aveiro inletdynamics. In: Smith, J.M. (Ed.), 30th International Conference on Coastal Engi-neering Book. World Scientific Publishing, USA, pp. 3282–3294.

Pina, P., 2001. An integrated approach to study the Tagus estuary water quality.M.Sc. Thesis, Instituto Superior Técnico, Universidade Técnica de Lisboa, Lisboa,Portugal.

Portela, L.I., 1996. Modelacão matemática de processos hidrodinâmicos e de qual-idade da água no estuário do Tejo. PhD Thesis, Instituto Superior Técnico,Universidade Técnica de Lisboa, Lisboa, Portugal.

Rodrigues, M., Oliveira, A., Costa, M., Fortunato, A.B., Zhang, Y., 2009. Sensitivityanalysis of an ecological model applied to the Ria de Aveiro. Proceedings of the10th International Coastal Symposium. J. Coastal Res. SI56, 5.

Rodrigues, M., Oliveira, A., Queiroga, H., Zhang, Y.J., Fortunato, A.B., Baptista, A.M.,2008. Integrating a circulation model and an ecological model to simulate thedynamics of zooplankton. In: Spaulding, M. (Ed.), 10th Estuarine and CoastalModeling Conference Book. ASCE, USA, pp. 427–446.

Saraiva, S., 2005. Modelacão ecológica da Ria de Aveiro: o papel das macroalgas.M.Sc. Thesis, Instituto Superior Técnico, Universidade Técnica de Lisboa, Lisboa,Portugal.

Silva, J.F., Duck, R.W., Oliveira, F., Rodrigues, M., 2000. Utilizacão da água e sanea-mento na área do Baixo Vouga: impactos na Ria de Aveiro. In: Actas do IICongresso Ibérico de Planeamento e Gestão de Água, Porto, Portugal.

Trancoso, A.R., Saraiva, S., Fernandes, L., Pina, P., Leitão, P., Neves, R., 2005. Modellingmacroalgae using a 3D hydrodynamic-ecological model in a shallow, temperateestuary. Ecol. Model. 187, 232–246.

Universidade de Aveiro – Centro das Zonas Costeiras e do Mar, Instituto SuperiorTécnico – Centro de Ambiente e Tecnologias Marítimos and Hidromod, 2003.ModelRia – Modelacão da qualidade da água na laguna de da Ria de Aveiro. FinalReport, POCTI/CTA/10219/98, Aveiro, Portugal.

sional modeling of the lower trophic levels in the Ria de Aveiro

Vaz, N., Dias, J.M., 2008. Hydrographic characterization of an estuarine tidal channel. 703

J. Marine Syst. 70, 168–181. 704

Vaz, N., 2007. Study of heat and salt transport processes in the Espinheiro Channel 705

(Ria de Aveiro). PhD Thesis, Universidade de Aveiro, Aveiro, Portugal. 706

Zhang, Y.-L., Baptista, A.M., 2008. SELFE: a semi-implicit Eulerian–Lagrangian finite- 707

element model for cross-scale ocean circulation. Ocean Model. 21, 71–96. 708