Ecological response of an intermittently open lagoon to

For Review Purposes Only/Aux fins d'examen seulement

Ecological response of an intermittently open lagoon to

entrance opening regimes and catchment loads: a numerical

study

Jason D. Everett∗,a,b, Mark E. Bairdc, Iain M. Suthersa,b

aEvolution & Ecology Research Centre, School of Biological Earth and Environmental Sciences,

University of New South Wales, Sydney NSW 2052, AustraliabSydney Institute of Marine Science, Mosman NSW 2088, Australia

cCoastal and Regional Oceanography Laboratory, School of Mathematics and Statistics, University

of NSW, Sydney NSW 2052, Australia

Abstract

Ecological model behaviour of an intermittently closed and open lake or lagoon

(ICOLL) is analysed in order to assess the response to altered estuary forcings (catch-

ment loads, opening regimes and lake depth), initial conditions and model parame-

ters. Due to the low nutrient loads currently entering Smiths Lake NSW Australia,

changes in the opening regime do not exhibit a strong estuarine response, however

the system is sensitive to increased water depth and catchment loads. The water

depth of the lagoon is shown to drive a switch from a seagrass to an algal dominated

system at average depths greater than 7 m. Elevated loads (10×) of phosphorus

entering the estuary increase the productivity of phytoplankton which is limited by

phosphorus at lower loads. Moderate nutrient enrichment stimulates the produc-

tion of seagrass, which is generally nitrogen limited and catchment loads are able

to double without a decline in seagrass biomass. At elevated catchment loads, light

limitation of seagrass occurs due to the increased plankton biomass and total sus-

pended solids, and the system switches from seagrass dominated to algal dominated

at loads above 5 mg N m−2 d−1 and 0.09 mg P m−2 d−1. Increasing nutrient loads

∗Corresponding AuthorEmail address: [email protected] (Jason D. Everett)

Preprint submitted to Marine and Freshwater Research March 25, 2010


also increase the amplitude and decrease the frequency of predator-prey cycles. In

order to interpret the model output the sensitivity of different parameters and ini-

tial conditions are examined. The model showed non-linear response to variations

in some parameters with the feeding efficiency of small and large zooplankton the

most sensitive. For small perturbations in the initial conditions of dissolved inorganic

nitrogen, phytoplankton and zooplankton biomass, ensemble members initially di-

verge, before converging. Periods of lake opening accelerate trajectory convergence

with longer openings having the greatest effect. All simulations within the ensemble

converge to the same stable oscillation within 2 open/closed phases.

Key words: eutrophication, light attenuation, opening cycle, ICOLL, plankton,

seagrass, Australia, NSW, Smiths Lake

Introduction1

Eutrophication is a growing problem in lakes and rivers caused by the excessive2

inputs of nitrogen and phosphorus (Carpenter et al., 1998). Land clearing and urban-3

isation can lead to eutrophication in estuarine waters as a result of increased nutrient4

and sediment loads entering the estuary from the catchment (Harris, 2001). Biogeo-5

chemical modelling is an important tool which can be used to assess system change6

and assist with management of potentially impacted systems (Webster and Harris,7

2004). Nutrient exports from south-eastern Australian catchments into estuaries are8

generally considered small when compared to the Northern Hemisphere, resulting in9

oligotrophic conditions in most estuaries and coastal lagoons (Harris, 2001). Loads10

are highly variable due to differing levels of urbanisation and catchment flows. Total11

nitrogen (TN) and phosphorus (TP) loads from 24 catchments around Sydney have12

been shown to vary by three orders of magnitude with TN and TP increasing with13

population density (Harris, 2001).14

Intermittently Closed and Open Lakes or Lagoons (ICOLLs) are a common type15

of estuary in south-eastern Australia. Of the 134 estuaries in New South Wales16

2


(NSW), 67 (50%) are classified as intermittently open estuaries (Roy et al., 2001).17

ICOLLs are characterised by low freshwater inflow, leading to sand barriers (berms)18

forming across the entrance preventing mixing with the ocean. Increased residence19

time and lack of flushing makes ICOLLs particularly susceptible to pollution events20

such as increases in nutrient and sediment loads (Cloern, 2001; Lloret et al., 2008), yet21

there are far fewer studies of ICOLLs, than permanently open estuaries (Schallenberg22

et al., 2010). Following a rise in the water level, the sand barriers are intermittently23

breached. The timing and frequency of the entrance opening is related to factors24

such as the size of the catchment, rainfall, evaporation, the height of the berm25

and creek or river inputs (Roy et al., 2001). Additionally, marine currents, wave26

activity, weather patterns and the strength of the initial breakout will determine how27

long the estuary remains open to the sea. As each of these processes are variable28

and opening regimes are rarely predictable, ICOLLs seldom reach any long-term29

steady state (Roy et al., 2001). As an added complication, 72% of the estuaries30

in NSW are now artificially opened when they reach a predefined ’trigger’ height31

(DIPNR, 2004). Reasons for artificial opening include flood prevention strategies32

and flushing in order to minimise pollution and eutrophication. However, opening33

events have previously resulted in unintended ecological consequences (Schallenberg34

et al., 2010) such as de-oxygenation, nutrient enrichment and increased chlorophyll35

a concentrations (Twomey and Thompson, 2001; Gobler et al., 2005; Becker et al.,36

2009).37

Unlike permanently open estuaries which have regular tidal flushing, the opening38

regimes of ICOLLs are often unpredictable. There are many estuary-specific char-39

acteristics such as lake inflow, depth and catchment loads which will influence the40

ecology of the system after opening (Cloern, 2001). Understanding the interaction41

of physical and ecological processes within an ICOLL needs to be understood before42

decisions are made about altering the opening regimes. Biogeochemical modelling43

3


studies of ICOLLs, such as this one, allow the examination of the ecological impacts44

during altered opening regimes. The model configuration used in this study empha-45

sises the two different phases of the ICOLL cycle - open and closed - and the different46

hydrological processes which occur.47

The ecological model examined in this study was configured for Smiths Lake and48

was built using a combination of physical and physiological limits to the processes49

of nutrient uptake, light capture by phytoplankton and predator prey interactions50

(Everett et al., 2007). The model outputs were assessed against field data and shown51

to capture the important ecological dynamics of Smiths Lake. A cost function was52

used to compare the field and model data, normalised by the standard deviation53

of the field data, to assess the fit, as per the methods of Moll (2000). The fit for54

salinity (0.20) and dissolved inorganic nitrogen (DIN, 0.79) were considered very55

good (<1 standard deviation), while dissolved inorganic phosphorus (DIP, 1.49) and56

chlorophyll a (1.01) were considered good fits (1-2 standard deviations, Everett et al.,57

2007; Moll, 2000). The simulated DIN and DIP concentrations were consistent with58

observations for much of the open-closed cycle. At high lake levels, DIN and DIP are59

underestimated. The concentration of water column DIN and DIP remains relatively60

constant over the course of the model simulation. Both observations and model data61

show very low concentrations of water column DIP (Everett et al., 2007).62

Smiths Lake has a largely forested catchment with a small population and low63

levels of urbanisation (Webb McKeown & Associates Pty Ltd, 1998). Population64

pressure and increasing urbanisation will result in increased nutrient exports from65

the catchment. Changing rainfall patterns in south-eastern Australia (Hennessy66

et al., 2004; CSIRO and Bureau of Meteorology, 2007) will also alter the length and67

frequency of opening regimes and lake depths. As a result, there is a need for an68

analysis of the response of the state variables to changing estuary forcings such as69

nutrient loads and opening regimes.70

4


Simple forcings models have previously been used to simulate the response of71

coastal lagoons to changing nitrogen loads and water residence times (Webster and72

Harris, 2004). Their study focussed on the response of primary producers to increas-73

ing and decreasing nitrogen loads, and how the response is altered by changes in74

the flushing rate of the lagoon. The model showed that the highly episodic catch-75

ment nutrient loading regimes increases the vulnerability of the lagoon (Webster and76

Harris, 2004). When compared with steady loading regimes, episodic loads caused77

phytoplankton blooms at lower catchment loads.78

The behaviour of the ecological model can be used to explore the response of79

a broader range of ICOLLs to changing nutrient loads and opening regimes. The80

inorganic nitrogen load for Smiths Lake has been calculated as 0.5 mg DIN m−281

d−1 from field data (Everett, 2007) and the total nitrogen load calculated as 5 mg82

TN m−2 d−1 from modelled data (CSIRO, 2003). This value is low when compared83

with other eastern Australian estuaries due mainly to Smiths Lake’s largely forested84

catchment and small catchment to lake area ratio. Other estuaries in eastern Aus-85

tralia have much higher loads. Lake Macquarie (permanently open) and Narrabeen86

Lakes (ICOLL) for example, have an estimated 5% forested catchment and loads of87

70 mg DIN m−2 d−1. Turros and Coila Lakes (ICOLLs), which have catchments that88

are 40% forested have loads of 20 mg DIN m−2 d−1 (CSIRO, 2003).89

In this study, numerical experiments are used to examine the biological response90

of an estuarine model to changes in estuary forcings (opening regimes, catchment91

loads, lake depth), initial conditions and model parameters. The range of forcing92

values used in this study allow insights into the future behaviour of Smiths Lake93

under altered conditions, while also being used as a tool to understand other ICOLLs.94

5


Methods95

A 1-box ecological model of an intermittently open lagoon (Everett et al., 2007)96

is investigated. The sensitivity of the model to estuary forcings, initial values of the97

state variables and model parameters is examined. The sensitivity results presented98

here are for a 1-box configuration with an idealised lake height, flushing regime and99

solar radiation. A brief overview of the model is presented in the following section.100

For a more detailed description see Everett et al. (2007).101

Transport Model102

The transport model is reduced to a 1-box configuration for this study. The model103

is forced by evaporation (E ), rainfall (R) and tidal exchange (TE ). The equation for104

the change in concentration of a tracer is given by:105

dC

dt=CcrRI

V︸︷︷︸

Runoff

−CE

V︸︷︷︸

Evaporation

+CocTE

V︸︷︷︸

T idal Exchange

−C dV

dt

V︸︷︷︸

Dilution

(1)

where V is volume (m3), E is evaporation (m3 s−1) and is negative, Ccr is the106

concentration of creek flow into the model (mol m−3), RI is the flow of water off107

the catchment (m3 s−1), Coc is the concentration of the ocean (mol m−3), dVdt

is the108

change in volume over time (m3 s−1) and TE is the tidal exchange when open to the109

ocean (m3 s−1).110

Ecological Model111

The ecological model contains 17 state variables (Table 1). With the exception112

of state variables relating to phosphorus and total suspended solids, nitrogen is the113

basic currency of the state variables. The autotrophs (phytoplankton, epiphytes114

and seagrass) gain all their nutrients from the water column, with the exception of115

seagrass which is also able to draw nutrients from the sediment. Dissolved inorganic116

nitrogen, unflocculated phosphorus, and total suspended solids enter the system from117

6


the catchment. Total suspended solids and flocculated and unflocculated phosphorus118

sink out of the water column into the sediment. Depending on water column and119

sediment pore water concentrations, DIN and DIP are able to diffuse back into the120

water column from the sediment. Small and large zooplankton grow by feeding on121

small and large phytoplankton respectively.122

In the model, light is attenuated through the water column, epiphytes and sea-123

grass sequentially. The light model is adapted from Baird et al. (2003). The propor-124

tion of the downward solar radiation (mol photon m−2 s−1) available as photosyn-125

thetically available radiation (PAR) is assumed to be 43% (Fasham et al., 1990).126

The realised growth rate (µx) of each autotroph is determined by the minimum of127

the physical limit to the supply of nitrogen, phosphorus and light, and the maximum128

metabolic growth rate of an organism:129

µx = (µnitrogen, µphosphorus, µlight, µmax) (2)

The four autotrophs in the model have fixed internal stoichiometric requirements.130

Phytoplankton (small and large) and epiphytic microalgae adopt the Redfield ratio131

(C:N:P) of 106:16:1 (Redfield, 1958), while seagrass adopt a seagrass-specific ratio132

of 474:21:1 (Duarte, 1990). The light requirements of all the autotrophs are fixed at133

10 photons per carbon atom (Kirk, 1994).134

Growth rates of phytoplankton, epiphytic microalgae and seagrass135

Phytoplankton are divided into two size classes - small (PS) and large (PL).136

Phytoplankton cells obtain nutrients from the surrounding water column and absorb137

light as a function of the average available light in the water column and their138

absorption cross section. The maximum rate (s−1) at which a phytoplankton cell139

can obtain nutrients through diffusion from the water column is given by:140

µPS,N =ψPSDNDIN

mPS,N

(3)

7


where ψPS is the diffusion shape factor (m cell−1), DN is the molecular diffusivity141

of nitrogen (m2 s−1), and mPS,N is the nitrogen cell content of small phytoplankton142

cells (mol N cell−1). Phosphorus uptake is calculated in a similar manner. The143

diffusion shape factor (ψ) in this study is for a sphere and is calculated as 4πr,144

where r is the radius of the cell (m).145

The light-limited growth rate of small phytoplankton (and similarly for large146

phytoplankton) is given by:147

µPS,I =IAV aAPS

mPS,I

(4)

where IAV is the average light field in the water column (mol photon m−2 s−1),aA148

is the absorption cross-section of the cell (cell−1 m2), and mPS,I is the required moles149

of photons per phytoplankton cell (mol photon cell−1).150

In the model, seagrass grows on the benthos and the epiphytes grow on both the151

surface of the seagrass and the benthos. Both seagrass and epiphytes absorb water152

column nutrients through the same effective diffusive boundary layer which occurs153

along their surface. Nutrient uptake is divided between both epiphytes and seagrass154

by a ratio termed the seagrass uptake fraction (η). In the model, seagrass are able to155

extract nutrients from both the water column and the sediment. The zooplankton156

biomass is not modelled explicitly here, but is implied to be proportional to the157

aboveground biomass. Priority in nutrient uptake is given to the water column,158

however nutrient limitation in the seagrass only occurs when the combined sediment159

and water column nutrients do not meet demand.160

The nitrogen limited growth rate for epiphytes is given as:161

µEP,N =DN

BL(1 − η)DIN

1

EP(5)

where DN is molecular diffusivity of nitrogen, BL is the boundary layer thick-162

ness, η is the seagrass uptake fraction and DIN is the water column concentration163

8


of dissolved inorganic nitrogen. Phosphorus limited growth rate is calculated in a164

similar manner.165

When formulating seagrass dynamics, fluxes are calculated per m2. As a result,166

to convert the water column uptake to a growth rate requires division of the uptake167

rate by the biomass. The nitrogen (and similarly for phosphorus) limited uptake168

rate (s−1) for seagrass is the sum of the water column and the sediment uptake and169

is given as:170

µSG,N =DN

BLηDIN

1

SG︸︷︷︸

water column

+µmax

SG DINsedconc

kSG,N︸︷︷︸

sediment

(6)

where µmaxSG is the maximum growth rate of seagrass (s−1), DINsedconc is the171

sediment concentration of nitrogen and kSG,N is the half saturation constant for172

nitrogen uptake in seagrass.173

Light is available to the epiphytes after it passes through the water column, and174

to the seagrass after passing through the epiphytic layer. The light limited growth175

rate for epiphytes and seagrass respectively is given as:176

µEP,I = (Ibot − IbelowEP )16

1060

1

EP(7)

µSG,I = (IbelowEP − IbelowSG)21

4740

1

SG(8)

where IBOT is PAR at the bottom of the water column, IbelowEP and IbelowSG is177

the light below the epiphytes and seagrass layers respectively, 16:1060 and 21:4740178

is based on the Redfield and Duarte Ratio (Duarte, 1990; Redfield, 1958) and a 10:1179

photon:carbon ratio (Kirk, 1994).180

A linear mortality is applied to epiphytes (ζEP ) and a quadratic mortality is181

applied to seagrass (ζSG) as a closure term for the benthic autotrophs. A resorption182

coefficient (ωSG) is appended to the mortality term of the seagrass to represent the183

9


detainment of nutrients as carbon is lost through blade death (Hemminga et al.,184

1999). When seagrass die, the remaining nutrients after resorption breakdown into185

refractory detritus (RD) before breaking down further and being released into the186

water column as DIN and DIP .187

Zooplankton Grazing188

In the model small zooplankton graze on small phytoplankton as shown below189

(and similarly large zooplankton graze on large phytoplankton):190

µZS =µmax

ZS PS

KZS + PS(9)

where µmaxZS is the maximum growth rate of small zooplankton (s−1) and KZS is191

the half saturation constant of small zooplankton growth (mol N m−3). A quadratic192

mortality for zooplankton is selected on the assumption that the biomass of zoo-193

plankton will be proportional to their predator.194

Initial Conditions195

The model is run for a period of one closed and one open cycle to allow the196

model to reach a quasi steady-state before commencing the model simulations. The197

initial conditions of the spin-up are set from the ocean boundary conditions for water198

column state variables. The initial biomass of epiphytes and seagrass is based upon199

the literature (Duarte, 1990, 1999) and the mapped seagrass coverage in Smiths Lake200

(West et al., 1985). Initial conditions for sediment state variables are derived from201

Baird et al. (2003), Murray and Parslow (1997) and Smith and Heggie (2003). The202

values at the end of this ’spin-up’ are used as the initial conditions for the model203

(Table 1).204

Lake Boundary Conditions205

Nutrients (DIN and Punfloc) and TSSunfloc enter the lake through runoff from the206

catchment. Nutrients (DIN and DIP ) and plankton (PS, PL, ZS and ZL) enter207

10


from the ocean when the lake is open. Nutrient concentrations are interpolated from208

field sampling in the creeks (catchment) and at Seal Rocks Beach (oceanic). The209

boundary conditions for DIN and Punfloc from the catchment are 1.08×10−2 mol N210

m−3 and 9.03×10−5 mol P m−3 respectively. The boundary condition of TSSunfloc211

from the catchment is 0.1 kg TSS m−3 (Baird et al., 2003). Oceanic boundary212

conditions are found in Table 1.213

Idealised Configuration214

In this study, the transport model is used in an idealised single box configuration.215

During a control simulation, the lake height increases linearly from 0.4 m to 2.0216

m Australian Height Datum (AHD) over 10 months which corresponds to average217

depths of 1.9 - 3.5 m (Fig. 1). After 10 months, the ICOLL opens to the ocean218

and drains for 48 hours to a lake height of 0.4 m AHD. A sinusoid tidal exchange219

is imposed for 2 months, with amplitude of 0.05 m. This cycle is repeated until a220

steady oscillation is reached (< 10 years).221

The depths and sinusoidal amplitude used in the model are chosen to be represen-222

tative of the average lake cycle and flushing rates of Smiths Lake observed over the223

last 10 years (Everett, 2007). Due to intermittent rainfall, there is a large variability224

in open/closed phases. On average the lake stays open for 62 days (from lake height225

data courtesy of Manly Hydraulics Laboratory). Depending on the height difference226

between Smiths Lake and the ocean, it takes 36-48 hours to drain and begin flushing227

with the ocean.228

A constant solar radiation is maintained throughout the simulations to prevent229

any seasonal effects influencing the sensitivity analysis. The solar radiation in the230

model is 95 W m−2 and is derived from the yearly mean of averaged 6 hourly, 2◦231

shortwave downward solar radiation, linearly interpolated in space to Smiths Lake232

(152.519E, 32.393S - data courtesy of the NOAA-CIRES Climate Diagnostic Centre).233

The control simulations in the following sections refer to a model run with no changes234

11


made to estuary forcings, initial conditions or parameters.235

Estuary Forcing236

Catchment Load : In order to examine the response of the model to varying catch-237

ment loads of dissolved inorganic nitrogen (DIN), dissolved inorganic phosphorus238

(DIP ) and total suspended solids (TSS), the loads are changed by 0.1×, 0.5×, 2.0×,239

10× and 100× the control values. This is undertaken in seven different combinations240

of DIN , DIP , TSS, DIN+DIP , DIN+TSS, DIP+TSS and DIN+DIP+TSS.241

The control values for DIN , DIP and TSS are 0.5 mg N m−2 d−1, 0.009 mg P m−2242

d−1 and 0.033 g TSS m−2 d−1. The inflow concentrations for DIN and DIP are243

determined from sampling of five creeks entering Smiths Lake. The value for TSS244

is based upon Baird et al. (2003).245

Opening Regimes : The current study varies the duration of the entire open/close246

cycle and the duration of the open phase to examine the effect on plankton and247

seagrass dynamics within Smiths Lake. The open/closed cycle duration is the total248

time between two successive closed phases (i.e. one complete open and closed cycle)249

and is undertaken at 12, 18, 24, 36, 48 and 60 months. The open duration is the250

total time the lake is open to the ocean and is undertaken by 7, 14, 30 and 60 days,251

making a total of 20 simulations (5 cycles × 4 open phases)252

Lake Depth: In order to examine the sensitivity of the model to lake depth,253

simulations are run with depths, when open to the ocean, of 0.4 m AHD (control254

simulation - representative of Smiths Lake), 2.5, 5 and 10 m AHD. These heights255

(AHD) correspond to average lake depths over the whole simulation of 2.5, 4.5 7.0256

and 12.0 m. During these simulations the same rate of lake rise is maintained, hence257

these simulations do not alter the catchment inflow.258

Initial Condition Sensitivity259

The initial conditions for DIN , PS, PL, ZS and ZL are randomly varied and260

the effect on the model is examined for three ensembles of ten members each. The261

12


three ensembles have open durations of two months, one month and two weeks. The262

initial conditions are varied based on a uniform random distribution with a mean263

of zero and a maximum of no more than the concentration of DIN (4×10−4 mol N264

m−3), which is the state variable with the lowest concentration. This formulation is265

used to prevent the initial conditions of state variables from taking negative values,266

while also keeping the total mass constant.267

Parameter Sensitivity268

A sensitivity analysis of all model parameters is undertaken by varying each269

parameter by 10%, 25% and 50% and analysing the change in each state variable270

(Table 2). Results for the 14 most sensitive parameters are presented here. The271

results from the 5th year of the simulation (by which time the model had reached272

a steady oscillation) are analysed for changes in the state variables, by comparison273

with the control simulation. Results of the parameter sensitivity are compared with274

the 11-box model of Smiths Lake Everett et al. (2007). The standardised sensitivity275

of each state variable to each parameter (Murray and Parslow, 1997) is calculated276

as:277

Normalised Sensitivity =V (1.1p) − V (0.9p)

V (p)0.2(10)

where V (1.1p) is the mean value of the state variable V when parameter p is278

increased by 10% and V (0.9p) is the mean value when the parameter is decreased279

by 10%. V (p) is the mean value of the state variable when there is no change in the280

parameter. If the normalised sensitivity is close to 1, the change in V is proportional281

to the change in p. If the sensitivity is close to 2, V is proportional to p2.282

13


Results283

Catchment Load Response284

Plankton Dynamics : Phosphorus is the limiting nutrient for the growth of small285

and large phytoplankton in the control simulations. Simulations run at 0.1× and286

2× the control phosphorus load result in only a small deviation from the values287

of PS and PL during the control simulations (Fig. 2A,B). There is no change in288

the overall shape of the oscillation. When the phosphorus load is increased by 10×289

and 100×, there is an increase in the frequency and amplitude of the oscillations290

(Fig. 2A,B). There is an associated increase in the yearly productivity of PS and291

PL from 19.3 and 10.8 mmol m−3 yr−1 at 2× DIP loads to 34.7 and 26.3 mmol292

m−3 yr−1 respectively at 10× DIP loads (Fig. 2C,D). The change in oscillation293

structure corresponds with a switch in the growth limitation of PS (Fig. 3A,C) and294

PL (Fig. 3B,D) from predominately phosphorus limited to nitrogen and maximum295

growth rate limited. In both 2× and 10× simulation (Fig. 3) PS and PL are out of296

phase, with the growth of PS limited by nitrogen when its biomass is increasing and297

by its maximum growth rate when the biomass is decreasing. The opposite is true298

of PL, which is limited by its maximum growth rate as its biomass is increasing and299

by nitrogen when its biomass is decreasing.300

Seagrass: Seagrass is nitrogen limited during the control simulations and the301

addition of 10× and 100× DIN load increases the seagrass biomass (Fig. 4A). These302

scenarios show no discernible difference in the light attenuation of phytoplankton and303

light availability for seagrass (Fig. 4C,D)304

The addition of 10× DIN +DIP and DIN +TSS load show little change in the305

seagrass light availability, however the addition of DIP allows the phytoplankton306

biomass, and therefore phytoplankton light attenuation, to increase (Fig. 4G). The307

addition of DIN allows the seagrass to obtain a higher biomass (Fig. 4E,I). The308

addition of DIN+DIP loads results in a small decrease in seagrass light availability309

14


(Fig. 4H). At higher loads the SG light availability oscillates, coinciding with the310

oscillation of phytoplankton biomass and light attenuation. The addition of 100×311

DIN + DIP and DIN + TSS loads (Fig. 4E,I) results in an increase in biomass312

during the shallower open phases when seagrass is maximum growth rate limited313

(not shown). There is a reduction in biomass during the closed phases (Fig. 4E),314

when seagrass becomes light limited (not shown). This is a result of decreased315

light availability to seagrass (Fig. 4H) due to an increase in the phytoplankton light316

attenuation (Fig. 4G).317

The simultaneous addition of DIN +DIP + TSS at 100× control loads results318

in a permanently lower seagrass biomass of 0.014 mol N m−2 (Fig. 4M) due to the319

reduction in light availability at the bottom of the water column (Fig. 4P).320

Opening Regimes321

There is an increasing DIN concentration in simulations with short open/close322

cycle durations (Fig. 5A). The highest mean DIN concentrations occur during the323

1 year cycle, and are consistent across all open phases. The highest mean DIP324

values occur at an open/closed cycle duration of 5 years and an open duration of 60325

days (Fig. 5B). PS and PL both increase in biomass as the open duration increases326

(Fig. 5C,D). The highest biomass of PS occurs during simulations with a long open327

duration (60 days) and long open/closed cycle duration (5 years), while the highest328

biomass of PL occurs during simulations with long open phases (30-60 days) and329

short open/closed cycle duration (1 year). The highest biomass of ZS and ZL330

occurs during simulations with long open phases (60 days) (Fig. 5E,F). The highest331

mean biomass of SG is 0.037 eutrophication N m−2 and occurs during the 5 year332

open/closed cycle (Fig. 5G). The duration of the open phase does not substantially333

affect the mean SG biomass.334

15


Lake Depth335

For simulations with an average lake depth of greater than 7 m, there is a re-336

duction of up to 97% in seagrass biomass when compared to the control simulations337

(Fig. 6H). At depths of 4.5 and 7 m, there is no discernible difference in SG biomass338

from the 2.5 m control simulations. In simulations with depths greater than 7 m,339

PS and PL more limited by their maximum growth rate as opposed to phospho-340

rus limitation, when compared to simulations with depths of 2.5 and 4.5 m (Fig. 7341

Col 1,2). At average depths of greater than 7 m, there is a more sudden shift in342

seagrass and epiphytic growth limitation from phosphorus and nitrogen respectively,343

to light limitation (Fig. 7 Col 3,4). This results in a reduced growth rate for sea-344

grass (Fig. 8F) which in turn allows more nutrients to be available for increased345

phytoplankton growth (Fig. 8A,B).346

Initial Conditions347

An ensemble of 10 members with different initial conditions of each of the state348

variables shows large standard deviations during the first year (Fig. 9A). However,349

the mean standard deviation of state variables (DIN , DIP , PS, PS, ZS and ZL)350

is reduced almost 10 fold between the first (9.6×10−5) and the second closed phase351

(1.4×10−6). The 2 month long open phase acts to accelerate the trajectory conver-352

gence of the model. The model is subsequently run for simulations with shorter open353

phases (1 month and 2 weeks). These simulations take longer to converge but by the354

end of the second open phase the mean standard deviation is 2.4×10−6 and 3.7×10−5355

respectively (Fig. 9B,C). Convergence within two phases illustrates the importance356

of steady states in the model behaviour and provides confidence that using the 5th357

year is a meaningful comparison of these states.358


The sensitivity of the parameters is investigated using a normalised sensitivity360

derived from a power law relationship. The feeding efficiency of ZS (βZS) and ZL361

16


(βZL) and the mortality of ZS (ζZS) and EP (ζEP ) are the most sensitive parameters362

in this configuration of the model (Table 3). DIP , phytoplankton, zooplankton and363

epiphytes are relatively sensitive to the radii of PS and PL when compared to364

the sensitivity of many other parameters in the model. The nine most sensitive365

parameters are investigated further by running the same simulations with changes366

in the value of the parameters of ± 25 and 50% (Table 4).367

The non-linearity of the model is illustrated by changes in the normalised sensitiv-368

ity as the percentage change of the parameter is increased. State variable-parameter369

combinations with changes of more than 0.1 between simulations are considered to370

have different sensitivities at different magnitudes of permutation and are highlighted371

in Table 4. The parameters which exhibited the greatest change in normalised sen-372

sitivity when changed by 25 and 50% are rPL, ζEP , βZS and βZL. The control373

parameter values for rPL and ζEP , are 1 µm and 0.05 d−1 respectively (Table 2) with374

a range (±50%) of 0.5-1.5 µm and 0.25-0.75 d−1 respectively. The control parameter375

values for βZS and βZL are 0.31 and 0.34 respectively (Table 2) with a range (±50%)376

of 15.5-46.5 and 17-51 respectively. The model responds in a relatively linear manner377

to the other parameters tested (Table 2).378

Discussion379

In this paper we have undertaken numerical experiments with changes in estuarine380

forcings, parameter values and initial conditions from an assessed configuration of381

Smiths Lake, Australia. It is not possible to co-vary all these factors simultaneously.382

As an example we have not simultaneously co-varied opening regimes and catchment383

loads. By examining each estuarine response in isolation, we gain less complicated,384

yet revealing, insights into the behaviour of intermittently open estuaries, while not385

deviating significantly from the assessed model.386

17


Catchment Load Response387

Plankton Dynamics : The responses of plankton communities and biomass to388

eutrophication in coastal lagoons are not well documented (Rissik et al., 2009). Phy-389

toplankton and zooplankton biomass in intermittently open estuaries is generally low390

during times of little freshwater influx and therefore small nutrient loads (Grange391

and Allanson, 1995; Froneman et al., 2000). Plankton biomass was higher in estuar-392

ies following freshwater pulses and their accompanying nutrient loads (Wooldridge,393

1999; Froneman et al., 2000). Similar responses are seen in this study with ele-394

vated loads (10×) of phosphorus entering the estuary increasing the productivity395

of phytoplankton which is limited by phosphorus at lower loads. This increase in396

productivity allows the biomass of PS and PL to turn over more quickly and hence397

decrease the period of oscillation and increase the amplitude (Fig. 2). The high398

availability of non-limiting DIN and light, combined with a high growth rate com-399

pared with other state variables allows phytoplankton to quickly take advantage of400

any additional phosphorus which becomes available in the water column. Similarly401

Edwards and Brindley (1996) found that under increasing nutrient loads in an NPZ402

model produced stable predator-prey (limit) cycles of greater amplitude.403

This model produced strong predator-prey cycles that are typical of numerical404

models (Edwards and Brindley, 1996). At high loads (10× and 100×), the oscillations405

of PS and PL are out of phase. PS is nitrogen limited when the biomass is increasing406

and maximum growth rate limited when it is decreasing, while PL is the opposite.407

Even though the maximum biomass increases throughout the closed phase, the switch408

in growth limitation always occurs at the same place on the curve rather than at409

the same water column DIN:DIP ratio. The switch between nitrogen and maximum410

growth rate limitation occurs at the same time in alternate cycles of PS and PL.411

That is, when PS changes from nitrogen limited to maximum growth rate limited,412

PL changes from maximum growth rate limited to nitrogen limited. This dynamic413

18


is driven by PL, which is able to draw down the most DIN from the water column414

(as a result of its higher biomass), resulting in PS becoming nitrogen limited. The415

switch in PS growth limitation to nitrogen limitation corresponds with the lowest416

point of DIN concentration (not shown).417

Seagrass: The modelled seagrass biomass is relatively insensitive to changes in418

the catchment loads of up to 2×. Up to these loads the biomass of seagrass stays419

relatively constant and is not influenced by changes in light availability. A 10×420

increase in catchment loads of DIN (even in the presence of 10× increase in DIP421

and/or TSS) doubled the seagrass biomass to 0.11 mol N m−2 at the end of the five422

year simulation when compared to the 2× simulation (0.05 mol N m−2). Under these423

conditions there is only a small increase in light attenuation by the phytoplankton424

and non-phytoplankton component of the water column. These simulations illustrate425

the importance of light for seagrass growth.426

The addition of 100× the limiting nutrient for seagrass (nitrogen) allowed seagrass427

to attain a higher biomass, however the addition of 100× loads of phosphorus (the428

limiting nutrient for phytoplankton), led to a bloom in phytoplankton which resulted429

in reduced light availability to the seagrass (Fig. 4). The same occurred when the430

levels of TSS are increased 100×, which also increased the light attenuation of the431

water column.432

In both these light-limiting scenarios the seagrass is able to recover during the433

open phases when the lake level is low and flushing occurs with cleaner oceanic434

water. In contrast, the addition of DIN +DIP +TSS at 100× levels resulted in the435

collapse of the seagrass to 0.014 mol N m−2, with very little recovery during the 2436

months that the lake is open to the ocean. In this scenario, the light attenuation of437

the water column (phytoplankton and non-phytoplankton components) is increased438

beyond what could be flushed out of the estuary during the open phase, and the439

estuary switches to an alternative state (Scheffer and Carpenter, 2003).440

19


Benthic macrophytes, such as seagrass, are extremely important in the function-441

ing of coastal lagoons. Benthic macrophytes are able to intercept nutrients from the442

water column, which is often sufficient to maintain reduced nutrient concentrations in443

the overlying water column (Grall and Chauvaud, 2002). As the macrophytes become444

stressed from increased nutrient and sediment loads, long-term retention of organic445

matter will decline, and be replaced by algal blooms and unattached ephemeral al-446

gae (McGlathery et al., 2007; Duarte, 1995; Valiela et al., 1997). Shallow lagoons,447

such as Smiths Lake, are at greater risk as there is a smaller volume of water into448

which to absorb the nutrient loads, but also, the photic zone generally reaches to the449

benthos, and so the macrophytes macrophytes play a larger role in controlling the450

overlying water column than in deeper systems (McGlathery et al., 2007). Seagrass is451

especially important during closed phases of ICOLL’s, when phytoplankton blooms452

cannot be flushed out. If seagrass dominates the primary production of the lagoon,453

and nutrients will be bound by plant tissue, rather than by phytoplankton, slowing454

the nutrient release back into the water column (McGlathery et al., 2007).455

Moderate nutrient enrichment stimulates the production of the benthic macro-456

phytes in this study. This converges to the well-documented response at higher457

loads where benthic macrophytes reduce their distribution and production within458

the ecosystem through light limitation (Borum and Sand-Jensen, 1996; Smith and459

Schindler, 2009). Due to the shallow depths of Smiths Lake, and the low catchment460

loads (even at 10×), light limitation only becomes growth limiting for SG after 180461

days of a 1 year closed phase. The reduction in biomass is small, and is less than462

the increase during the first half of the cycle. Hence, the overall change in seagrass463

biomass is positive, even in cycles with light limitation.464

Opening Regimes465

In Smiths Lake, PS and PL are generally growth limited by phosphorus, and466

SG is growth limited by nitrogen (Everett et al., 2007). Changes in the nutrient467

20


availability of DIN and DIP drive changes in the biomass of the seagrass and468

phytoplankton respectively. Due to the availability of DIN in the water column469

throughout the closed phases, a longer closed phase results in an increased biomass of470

SG (Fig. 5). The ocean provides an additional source of DIP to supplement the low471

DIP in the lake. Therefore, the longer the lake is open to the ocean, the higher the472

mean biomass of phosphorus-limited phytoplankton. Similar responses of increased473

phytoplankton biomass during opening events have occurred in Australia (Twomey474

and Thompson, 2001) and the USA (Gobler et al., 2005), and were attributed to475

oceanic nutrient addition and sediment nutrient release respectively. The longer the476

cycle duration is, the longer the lake must stay open to the ocean to maintain the477

same biomass of DIP , PS and EP (Fig. 5B,C,G). A reduction in the length of the478

open phase, for a given cycle duration, results in a reduced average biomass of DIP ,479

PS and EP . Within this study however, these increases in biomass do not increase480

the phytoplankton biomass enough to significantly decrease the light availability to481

the SG. As a result SG maintains a large biomass at all opening regimes. This482

is mainly due to the severe phosphorus limitation of phytoplankton which prevents483

significant blooms, and the low TSS loads due to the highly forested catchment.484

At the present loads entering Smiths Lake there is no decline in SG biomass485

or blooms or PS or PL as a result of changes in the lake cycle duration or open486

phase duration. Schallenberg et al. (2010) highlights that the relationship between487

opening regimes and eutrophication depends on multiple factors specific to each la-488

goon. Longer residence times generally favour phytoplankton dominance (McGlath-489

ery et al., 2007; Valiela et al., 1997), however due to the nutrient limitation within490

Smiths Lake, increased residence time does not have as large an impact as is expected.491

Contrary to initial expectations, the biomass of SG is not decreased by an increased492

residence time and an accumulation of nutrients (Fig. 5). Because phytoplankton493

and SG are limited by two different nutrients, changes in their biomass rarely occur494

21


simultaneously. The opening phases are driving DIP availability through oceanic495

enrichment, and DIN accumulation during the closed phases is driving DIN avail-496

ability. If changes in the stoichiometry of the catchment loads were to occur (not497

considered in this study), it is expected that residence time will have a much more498

substantial influence.499

Lake Depth500

Water depth is shown to be important as it drives a switch from a seagrass501

dominated system to an algal dominated system at average depths greater than 7 m502

(Fig. 6). As would be expected due to the reduced light penetration at these increased503

depths, the biomass of seagrass is reduced as the system becomes dominated by504

plankton. The increased frequency of oscillations of phytoplankton and zooplankton505

in simulations with greater depths is a result of their increased growth rate due506

to the availability of nutrients that occurs with the decline of seagrass biomass.507

Interestingly, the epiphytes also reduce in biomass becoming light limited at depths508

greater than 7 m (Fig. 7). This is despite the fact that they require less light than509

seagrass per carbon, and the same light as phytoplankton (Redfield, 1958; Kirk,510

1994; Duarte, 1990). Phytoplankton do not become light limited as they have first511

access to the light and absorb light at the average light availability through the entire512

water column, whereas epiphytes are at the bottom of the water column. Sensitivity513

analysis of depth changes of this extent (>7 m) are an important consideration,514

allowing the model formulation to be applicable for use in other estuaries.515

Assimilative Capacity of Smiths Lake516

Based upon the results of the catchment load scenarios, Smiths Lake can assim-517

ilate more than the current nutrient and sediment loads without large changes in518

the ecosystem. If it is assumed that any catchment load increases will equally affect519

DIN , DIP and TSS, the catchment loads are able to double without either a detri-520

mental effect to seagrass, or forcing a switch to an alternative plankton-dominated521

22


system. At loads of 10×, there is a change in oscillation frequency and size; however522

there is a further increase in seagrass biomass suggesting the increase in plankton523

biomass is not fundamentally changing the ecosystem. At loads of 100× however,524

there is a definite switch to a plankton dominated system, and large declines in sea-525

grass biomass. There is also an associated build-up in water column nutrients. As526

nutrients are entering the lake in the same DIN:DIP ratios, the plankton remains527

growth limited by phosphorus, however the seagrass becomes light limited. Smiths528

Lake’s assimilative capacity is in part determined by its small catchment area to lake529

area ratio of 2:1 (Haines et al., 2006), its predominately forested catchment (Webb530

McKeown & Associates Pty Ltd, 1998) and its large biomass of seagrass which is531

likely to play an important role in intercepting water column nutrients and enhanc-532

ing the lagoons resistance to eutrophication (McQuatters-Gollop et al., 2009). If533

Smiths Lake loads increase to levels similar to more urbanised estuaries (10-100×),534

simulations suggest that the system would switch to an alternate state, from seagrass535

dominated to microalgal dominated. As a result residence time and opening/closing536

cycles would play a greater role in determining ecosystem state.537


In order to interpret the model output, it is important to also consider the sensi-539

tivity of different parameters. A sensitivity analysis provides a measure of the sensi-540

tivity of changes in model parameters with respect to changes in the state variables541

of interest (Jorgensen and Bendoricchio, 2001). Sensitivities are usually examined542

at a few different levels which are decided upon based on prior knowledge about543

the parameter. If a parameter or forcing is found to be sensitive, more effort can544

be put into making sure the parameter is accurate. The parameters to which the545

model state variables are most sensitive are the feeding efficiency of ZS (βZS) and546

ZL (βZL) and the mortality of ZS (ζZS) and EP (ζEP ). For the majority of state547

variable-parameter combinations, the normalised sensitivity for the state variables548

23


is small (< 1) up to a ± 50% change in parameter values. The non-linearity of549

βZS and βZL at parameter changes of ± 25 and 50% is shown by a large change550

in the normalised sensitivity to phytoplankton biomass (Table 4). The normalised551

sensitivity for βZS increased from -1.28 to -1.78 for PS and βZL increased from -1.15552

to -1.35 for PL. There is a relatively large degree of uncertainty surrounding the553

parameter values for feeding efficiency (βZS = 0.31 and βZL = 0.34), however, the554

values chosen in this study are extracted from a compilation of 27 field studies on 33555

different species (Hansen et al., 1997). The range of values for βZS and βZL in those556

studies is 0.10-0.42 and 0.23-0.45 respectively. By varying these model parameters557

by up to ± 50%, the sensitivity for the entire range of possible values from Hansen558

et al. (1997) is captured. Within the study there was one exception with a value of559

0.10 which fell below the minimum value for βZS. This value could be considered an560

outlier, given its distance of more than two standard deviations from the mean.561

The predators of zooplankton are not represented explicitly in this model but their562

dynamics are approximated by the closure term. The closure terms can significantly563

affect the dynamics of an ecological model (Edwards and Yool, 2000; Steele and Hen-564

derson, 1992). In this study a quadratic mortality is implemented which represents565

grazing from a predator whose biomass is proportional to that of the zooplankton566

(Edwards and Yool, 2000). Nevertheless, the quadratic form of model closure does567

not necessarily increase stability or reduce oscillatory behaviour (Edwards and Yool,568

2000). Therefore it is not surprising that the quadratic mortality coefficients of ZS569

and ZL are the most sensitive parameters in the model (For further discussion of the570

relative costs and benefits of different closure terms, refer to Steele and Henderson571

(1992) and Edwards and Yool (2000)).572

The sensitivity of ζZS and ζZL is greatly reduced in the simulations presented in573

this paper when compared to their sensitivity in the 11-box model of Smiths Lake574

(Everett et al., 2007). The difference is that the current study utilises a 1-box model575

24


with idealised physical configuration (idealised inflow and incoming solar radiation).576

Due to the quadratic formulation of the mortality term any values for zooplankton577

biomass in the 11-box configuration which are above the mean value, would have578

been magnified, leading it to appear more sensitive than in the 1-box configuration.579

Therefore, the sensitivity of ζZS and ζZL may be underestimated in this and other580

studies which smooth out the patchiness of zooplankton biomass.581

Zooplankton has been shown to be the most poorly simulated state variable in582

aquatic biogeochemical models (Arhonditsis and Brett, 2004). One suggested reason583

for this is the need to convert observed zooplankton data from units of individuals584

per volume to units of carbon or nitrogen per volume. A second possible reason is585

that entire zooplankton life-histories need to be fully simulated to achieve accurate586

representation (Arhonditsis and Brett, 2004). A possible shortcoming of this model587

is that it contains both these potential sources of error. Of the 27 field studies588

mentioned above (Hansen et al., 1997), the feeding efficiencies of only two studies589

are a direct measure of nitrogen efficiency, the modelled variable in this study. In all590

other cases, the values are given in units of volume, dry weight or carbon content591

(Hansen et al., 1997). Also, in this study zooplankton growth is exponential, which is592

characteristic of micro-zooplankton, rather than age structured, as is characteristic593

of macro-zooplankton (Baird and Suthers, 2007). Therefore, life histories are not594

explicitly modelled in this study, apart from the two different size classes - small and595

large zooplankton.596

Summary597

A previously developed and assessed model of Smiths Lake, Australia, is used598

to investigate the ecosystem response to changes in estuarine forcings and biologi-599

cal parameters for a broader range of idealised ICOLLs. There is little change in600

estuarine response for small changes in depth and catchment loads due to the rela-601

tively small loads and low light attenuation under current Smiths Lake conditions.602

25


Larger load increases, as would be applicable in other estuaries, exhibit a switch in603

estuarine condition resulting in a decline in seagrass biomass and large changes in604

predator-prey cycles.605

Acknowledgements606

This work comprises part of JDE’s PhD and was funded by Australian Research607

Council (ARC) Linkage grant (LP0349257). MEB was supported by ARC Discovery608

Project DP0557618. The authors wish to thank Gerard Tuckerman and Great Lakes609

Council for their support. We also thank Peter Scanes, Geoff Coade, Ed Czobik610

and the Department of Environment and Climate Change (NSW) for processing611

the nutrient samples. The Manly Hydraulics Laboratory provided lake height data,612

the Department of Water and Energy provided bathymetry data and NOAA-CIRES613

Climate Diagnostic Center provided the solar radiation data. The authors would614

also like to thank the following colleagues for assistance in collecting the field data:615

A. Everett, E. Gale, S. Moore, K. Wright, L. Carson, T. Mullaney, R. Piola and M.616

Taylor.617

References618

Arhonditsis, G. B. and Brett, M. T. (2004). Evaluation of the current state of619

mechanistic aquatic biogeochemical modelling. Marine Ecology Progress Series620

271, 13–26621

Baird, M. E. and Suthers, I. M. (2007). A size-resolved pelagic ecosystem model.622

Ecological Modelling 203, 185–203623

Baird, M. E., Walker, S. J., Wallace, B. B., Webster, I. T., and Parslow, J. S. (2003).624

The use of mechanistic descriptions of algal growth and zooplankton grazing in625

an estuarine eutrophication model. Estuarine, Coastal and Shelf Science 56(3-4),626

685–695627

26


Becker, A., Laurenson, L. J., and Bishop, K. (2009). Artificial mouth opening fos-628

ters anoxic conditions that kill small estuarine fish. Estuarine, Coastal and Shelf629

Science 82(4), 566 – 572. doi:10.1016/j.ecss.2009.02.016630

Borum, J. and Sand-Jensen, K. (1996). Is total primary production in shallow coastal631

marine waters stimulated by nitrogen loading? Oikos 76(2), 406–410632

Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N.,633

and Smith, V. H. (1998). Nonpoint pollution of surface waters with phosphorus634

and nitrogen. Ecological Applications 8(3), 559–568635

Cloern, J. (2001). Our evolving conceptual model of the coastal eutrophication636

problem. Marine Ecology Progress Series 210, 223–253637

CSIRO (2003). Simple Estuarine Response Model II638

http://www.per.marine.csiro.au/serm2/index.htm Accessed: 12th March 2007639

CSIRO and Bureau of Meteorology (2007). Climate change in Australia: Technical640

report 2007. Tech. rep., CSIRO641

DIPNR (2004). Discussion Paper: ICOLL Management Guidelines. Tech. rep.,642

Department of Infrastructure, Planning and Natural Resources, N.S.W.643

Duarte, C. (1990). Seagrass nutrient content. Marine Ecology Progress Series 67(2),644

201 – 207645

Duarte, C. (1995). Submerged aquatic vegetation in relation to different nutrient646

regimes. Ophelia 41, 87–112647

Duarte, C. (1999). Seagrass ecology at the turn of the millennium: challenges for648

the new century. Aquatic Botany 65(1-4), 7–20649

Edwards, A. and Yool, A. (2000). The role of higher predation in plankton population650

models. Journal of Plankton Research 22(6), 1085–1112651

27


Edwards, A. J. and Brindley, J. (1996). Oscillatory behaviour in a three-component652

plankton population model. Dynamics and Stability of Systems 11, 347–370653

Everett, J. D. (2007). Biogeochemical Dynamics of an Intermittently Open Estuary:654

A Field and Modelling Study. Ph.D. thesis, University of New South Wales, Sydney655

Everett, J. D., Baird, M. E., and Suthers, I. M. (2007). Nutrient and plankton656

dynamics in an intermittently closed/open lagoon, Smiths Lake, south-eastern657

Australia: An ecological model. Estuarine, Coastal and Shelf Science 72, 690–702658

Fasham, M., Ducklow, H., and McKelvie, S. (1990). A nitrogen-based model of659

plankton dynamics in the oceanic mixed layer. Journal of Marine Research 48(3),660

591–639661

Froneman, P. W., Pakhomov, E. A., Perissinotto, R., and McQuaid, C. D. (2000).662

Zooplankton structure and grazing in the atlantic sector of the southern ocean663

in late austral summer 1993 - part 2. biochemical zonation. Deep-Sea Research I664

47(9), 1687–1702665

Gobler, C. J., Cullison, L. A., Koch, F., Harder, T., and Krause, J. (2005). Influence666

of freshwater flow, ocean exchange and seasonal cycles on phytoplankton - nutrient667

dynamics in a temporarily open estuary. Estuarine Coastal and Shelf Science 65,668

275–288669

Grall, J. and Chauvaud, L. (2002). Marine eutrophication and benthos: the need670

for new approaches and concepts. Global Change Biology 8, 813–830. doi:671

10.1046/j.1365-2486.2002.00519.x672

Grange, N. and Allanson, B. (1995). The influence of freshwater inflow on the nature,673

amount and distribution of seston in estuaries of the eastern cape, south africa.674

Estuarine, Coastal and Shelf Science 40, 403–420675

28


Haines, P., Tomlinson, R., and Thom, B. (2006). Morphometric assessment of inter-676

mittently open/closed coastal lagoons in New South Wales, Australia. Estuarine,677

Coastal and Shelf Science 67(1-2), 321–332678

Hansen, P., Bjørnsen, P., and Hansen, B. (1997). Zooplankton grazing and growth:679

Scaling within the 2-2,000 µm body size range. Limnology and Oceanography680

42(4), 687–704681

Harris, G. (2001). Biogeochemistry of nitrogen and phosphorus in Australian catch-682

ments, rivers and estuaries: effects of land use and flow regulation and comparisons683

with global patterns. Marine & Freshwater Research 52(1), 139 – 149684

Hemminga, M., Marba, N., and Stapel, J. (1999). Leaf nutrient resorption, leaf685

lifespan and the retention of nutrients in seagrass systems. Aquatic Botany 65,686

141–158687

Hennessy, K., Page, C., McInnes, K., Jones, R., Bathols, J., Collins, D., and Jones,688

D. (2004). Climate Change in New South Wales. Part 1: Past climate variability689

and projected changes in average climate. Tech. rep., New South Wales Greenhouse690

Office691

Jorgensen, S. and Bendoricchio, G. (2001). Fundamentals of Ecological Modelling692

(Elsevier Science Ltd, Oxford, UK), 3rd edn.693

Kirk, J. (1994). Light and photosynthesis in aquatic ecosystems. 509 pp. (Cambridge694

University Press), 2nd edn.695

Lloret, J., Marın, A., and Marın-Guirao, L. (2008). Is coastal lagoon eutrophication696

likely to be aggravated by global climate change? Estuarine, Coastal and Shelf697

Science 78(2), 403–412698

29


McGlathery, K., Sundback, K., and Anderson, I. (2007). Eutrophication in shallow699

coastal bays and lagoons: The role of plants in the coastal filter. Marine Ecology700

Progress Series 348, 1–18701

McQuatters-Gollop, A., Gilbert, A., Mee, L., Vermaat, J., Artioli, Y., Humborg, C.,702

and Wulff, F. (2009). How well do ecosystem indicators communicate the effects of703

anthropogenic eutrophication? Estuarine Coastal and Shelf Science 82, 583–596.704

doi:10.1016/j.ecss.2009.02.017705

Moll, A. (2000). Assessment of three-dimensional physical-biological ecoham1 simu-706

lations by quantified validation for the north sea with ices and ersem data. ICES707

Journal of Marine Science 57(4), 1060–1068708

Murray, A. G. and Parslow, J. S. (1997). Port Phillip Bay integrated model: Final709

report. Tech. Rep. Technical Report 44, CSIRO Division of Marine Resources,710

Canberra711

Redfield, A. (1958). The biological control of chemical factors in the environment.712

American Scientist 46, 205–221713

Rissik, D., Ho Shon, E., Newell, B., Baird, M., and Suthers, I. (2009). Plankton714

dynamics due to rainfall, eutrophication, dilution, grazing and assimilation in an715

urbanized coastal lagoon. Estuarine, Coastal and Shelf Science 84, 99–107. doi:716

10.1016/j.ecss.2009.06.009717

Roy, P., Williams, R., Jones, A., Yassini, I., Gibbs, P., Coates, B., West, R., Scanes,718

P., Hudson, J., and Nichol, S. (2001). Structure and function of south-east Aus-719

tralian estuaries. Estuarine, Coastal and Shelf Science 53(3), 351–384720

Schallenberg, M., Larned, S., Hayward, S., and Arbuckle, C. (2010). Contrasting721

effects of managed opening regimes on water quality in two intermittently closed722

30


and open coastal lakes. Estuarine, Coastal and Shelf Science 86(4), 587 – 597.723

doi:DOI: 10.1016/j.ecss.2009.11.001724

Scheffer, M. and Carpenter, S. (2003). Catastrophic regime shifts in ecosystems:725

linking theory to observation. Trends in Ecology & Evolution 18, 648–656. doi:726

10.1016/j.tree.2003.09.002727

Smith, C. and Heggie, D. (2003). Benthic Nutrient Fluxes in Smiths Lake, NSW.728

Tech. Rep. Record 2003/16, Geoscience Australia729

Smith, V. H. and Schindler, D. W. (2009). Eutrophication science: where do730

we go from here? Trends in Ecology & Evolution 24(4), 201–207. doi:731

10.1016/j.tree.2008.11.009732

Steele, J. and Henderson, E. (1992). The role of predation in plankton models.733

Journal of Plankton Research 14, 157–172734

Twomey, L. and Thompson, P. (2001). Nutrient limitation of phytoplankton in a735

seasonally open bar-built estuary: Wilson Inlet, Western Australia. Journal of736

Phycology 37, 16–29737

Valiela, I., McClelland, J., Hauxwell, J., Behr, P. J., Hersh, D., and Foreman, K.738

(1997). Macroalgal blooms in shallow estuaries: Controls and ecophysiological and739

ecosystem consequences. Limnology and Oceanography 42(5), 1105–1118740

Webb McKeown & Associates Pty Ltd (1998). Smiths Lake Estuary Process Study.741

Tech. rep., Prepared for Great Lakes Council, N.S.W. Australia742

Webster, I. T. and Harris, G. P. (2004). Anthropogenic impacts on the ecosystems743

of coastal lagoons: modelling fundamental biogeochemical processes and manage-744

ment implications. Marine & Freshwater Research 55(1), 67–78745

31


West, R. J., Thorogood, C., Walford, T., and Williams, R. J. (1985). An estuar-746

ine inventory for New South Wales, Australia (New South Wales, Department of747

Agriculture, Sydney)748

Wooldridge, T. (1999). Estuarine zooplankton community structure and dynam-749

ics. In B. Allanson and D. Baird, eds., Estuaries of South Africa, pp. 141–166750

(Cambridge University Press)751

32


Table 1: The 17 state variables used in the ecological model. The initial conditions and ocean boundary conditions and their symbol and

units are also shown.

State Variable Symbol Initial Conditions Ocean Boundary Condition Units

Dissolved Inorganic Nitrogen DIN 7.3 × 10−4 9.5 × 10−4 mol N m−3

Dissolved Inorganic Phosphorus DIP 5.1 × 10−6 1.0 × 10−4 mol P m−3

Small Phytoplankton PS 1.1 × 10−4 1.6 × 10−4 mol N m−3

Large Phytoplankton PL 3.3 × 10−4 2.1 × 10−4 mol N m−3

Small Zooplankton ZS 9.8 × 10−4 1.5 × 10−4 mol N m−3

Large Zooplankton ZL 1.1 × 10−3 1.5 × 10−4 mol N m−3

Epiphytic and Benthic Microalgae EP 8.4 × 10−5 0 mol N m−2

Seagrass SG 6.0 × 10−2 0 mol N m−2

Refractory Detritus RD 21 × 10−2 0 mol N m−2

Sediment Dissolved Inorganic Nitrogen DINsed 3.3 × 10−2 0 mol N m−2

Unflocculated Phosphorus Punfloc 2.7 × 10−7 0 mol P m−3

Flocculated Phosphorus Pfloc 2.0 × 10−8 0 mol P m−3

Sediment Dissolved Inorganic Phosphorus DIPsed 1.8 × 10−3 0 mol P m−2

Unflocculated Sediment Phosphorus Psed unfloc 3.1 0 mol P m−2

Flocculated Sediment Phosphorus Psed floc 5.6 0 mol P m−2

Unflocculated Total Suspended Solids TSSunfloc 5.2 × 10−3 0 kg TSS m−3

Flocculated Total Suspended Solids TSSfloc 2.2 × 10−5 0 kg TSS m−3

33


Table 2: Parameter values for the ecological model. The ranges tested in the sensitivity analysis are shown in column 2 and 3. N/T =

Not Tested.

Parameter Value 10% Range 50% Range

Cell radius of PS rPS = 2.5 × 10−6 m 2.25-2.75 × 10−6 1.25-3.75 × 10−6

Cell radius of PL rPL = 1.0 × 10−5 m 0.9-1.1 × 10−6 0.5-1.5 × 10−6

Cell radius of ZS rZS = 4.0 × 10−6 m 3.6-4.4 × 10−6 N/T

Cell radius of ZL rZL = 1.0 × 10−3 m 0.9-1.1 × 10−3 N/T

Cell radius EP rEP = 5.0 × 10−6 m 4.5-5.5 × 10−6 N/T

Mortality rate of ZS ζZS = 94.78 (mol N m−3)−1 d−1 85.30-104.26 47.39-142.17

Mortality rate of ZL ζZL = 18.03 (mol N m−3)−1 d−1 16.23-19.83 9.02-27.05

Mortality rate of EP ζEP = 0.05 d−1 0.045-0.055 0.025-0.075

Mortality rate of SG ζSG = 4.22 (mol N m−2)−1 d−1 (Duarte, 1990) 3.80-4.64 2.11-6.33

Rate of RD Breakdown RD = 0.1 d−1 (Murray and Parslow, 1997) 0.09-0.11 N/T

Feeding efficiency of ZS βZS = 0.31 (Hansen et al., 1997) 0.28-0.34 0.155-0.465

Feeding efficiency of ZL βZL = 0.34 (Hansen et al., 1997) 0.31-0.37 0.17-0.51

Sinking rate of Pfloc w Pfloc = 5 d−1 (Baird et al., 2003) 4.5-5.5 N/T

Sinking rate of Punflocc w Punfloc = 5 d−1 (Baird et al., 2003) 4.5-5.5 N/T

34


Table 3: Normalised sensitivity when parameters are varied by 10%. The rows represent the

parameters changed, and the columns the state variables examined. The values in bold are the

parameters with a normalised sensitivity greater than |0.4|.

DIN DIP PS PL ZS ZL EP SG

rPS -0.04 -0.43 0.68 -0.52 0.64 -0.62 -0.44 -0.01

rPL -0.04 -0.52 -0.59 0.57 -0.58 0.71 -0.53 -0.01

rZS 0 0 0 0 0 0 0 0

rZL 0 0 0 0 0 0 0 0

rEP 0 0 0 0 0 0 0 0

ζZS -0.01 -0.15 1.05 -0.14 0.01 -0.22 -0.15 0

ζZL -0.01 -0.17 -0.2 0.52 -0.18 0.06 -0.17 -0.02

ζEP 0 0 0 0 0 0 -1.01 0

ζSG 0 0 0 0 0 0 0 -0.41

rRD 0 0 0 0 0 0 0 0.4

sedxch -0.05 0.03 0.04 0.04 0.03 0.04 0.03 -0.01

βZS 0.01 0.18 -1.28 0.18 0 0.25 0.18 -0.01

βZL 0.02 0.24 0.24 -1.11 0.26 0.09 0.25 0.02

w Pfloc 0 0 0 0 0 0 0 0

w Punfloc 0.01 0 -0.01 -0.01 -0.01 -0.01 0 0

35


Table 4: Table of normalised sensitivity at the 25 (and 50% - in brackets) levels for the 8 most sensitive parameters from Table 3. Bold

represents the parameters whose sensitivity changed by more than 0.1 from 10 to 25% or 25 to 50%.

DIN DIP PS PL ZS ZL EP SG

rPS 0.04 (-0.04) -0.48 (-0.50) 0.65 (0.63) -0.53 (-0.50) 0.63 (0.62) -0.58 (-0.58) -0.49 (-0.51) -0.01 (0)

rPL -0.04 (-0.39) -0.47 (-0.48) -0.57 (-0.58) 0.62 (0.61) -0.54 (-0.56) 0.61 (0.65) -0.47 (-0.48) -0.01 (-0.01)

ζZS -0.01 (-0.01) -0.15 (-0.18) 1.03 (1.02) -0.12 (-0.11) 0 (0) -0.22 (-0.21) -0.15 (-0.18) 0 (-0.01)

ζZL -0.01 (0) -0.14 (-0.09) -0.14 (-0.11) 0.48 (0.49) -0.16 (-0.11) 0.03 (-0.05) -0.14 (-0.09) -0.02 (-0.02)

ζEP 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) -1.07 (-1.33) 0 (0)

ζSG 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) -0.42 (-0.47)

βZS 0.01 (0.03) 0.18 (0.3) -1.28 (-1.78) 0.14 (0.25) 0 (0.09) 0.27 (0.22) 0.18 (0.30) -0.01 (0.01)

βZL 0.02 (0.02) 0.25 (0.30) 0.25 (0.31) -1.15 (-1.35) 0.27 (0.33) 0.10 (0.11) 0.25 (0.30) 0.03 (0.02)

36


0 50 100 150 200 250 300 3501.8

2

2.2

2.4

2.6

2.8

3

3.2

3.4

3.6

Open Phase

Days

Ave

rage

Dep

th (

m)

1.8

1.9

2

Av.

Dep

. (m

)Open Phase

Days

Figure 1: Plot of the average lake depth (m) of the idealised model formulation against time

(days). The same cycle is repeated each year. The average depth of the lake is taken from

lake depth measurements and is different to the Australian Height Datum (AHD), which

is the reference datum for altitude measurement in Australia. The inserted box shows the

sinusoidal oscillation which occurs during the open phase.

37


0

0.5

1

1.5

2x 10

−3

A) Biomass of PS

mol

N m

−3

0

0.5

1

1.5

2x 10

−3

B) Biomass of PL

mol

N m

−3

0 100 200 300 400 500 600 7000

0.5

1

1.5x 10

−6

C) Productivity of PS

mol

N m

−3 d

−1

Days0 100 200 300 400 500 600 700

0

0.5

1

1.5x 10

−6

D) Productivity of PL

mol

N m

−3 d

−1

Days

2x 10x Control

Figure 2: Biomass of A) PS, B) PL (mol N m−3), productivity of C) PS, D) PL (mol N

m−3 d−1) in response to changes to the phosphorus load entering the estuary. The lines

represent the changes to the phosphorus load of 2×, 10× and the Control). The grey

shaded areas represent the time the lake is open to the ocean.

38


0

0.5

1

1.5

2x 10

−3 PS Biomass and Growth Limitation

A) 2× DIP Catchment Loads

mol

N m

−3

0

0.5

1

1.5

2x 10

−3PL Biomass and Growth Limitation

B) 2× DIP Catchment Loads

mol

N m

−3

0 100 200 300 400 500 600 7000

0.5

1

1.5

2x 10

−3

C) 10× DIP Catchment Loads

mol

N m

−3

Days0 100 200 300 400 500 600 700

0

0.5

1

1.5

2x 10

−3

D) 10× DIP Catchment Loads

mol

N m

−3

Days

Nitrogen Limited Phosphorus Limited Maximum GR Limited

Figure 3: Growth limitation of PS (A, C) and PL (B, D) in response to changes to the

phosphorus load entering the estuary. In plots A/B (2× increase in DIP Catchment Loads)

and C/D (10× increase in DIP Catchment Loads), the line shows the biomass and the

colour of the line represents the process that limits growth (nitrogen uptake, phosphorus

uptake, light capture or maximum metabolic rate). The grey shaded areas represent the

time the lake is open to the ocean.

39


0

0.2

Seagrass Biomass (mol N m−2)

A)

DIN

0

1

2

Non−Phytoplankton Light Attenuation (m−1)

B)

0

1

2

3

Phytoplankton Light Attenuation (m−1)

C)

0

1

2x 10

−4

Light Availability for Seagrass (mol photons m−2 s−1)

D)

0

0.2

E)

DIN

+D

IP

0

1

2F)

0

1

2

3G)

0

1

2x 10

−4

H)

0

0.2

I)

DIN

+T

SS

0

1

2J)

0

1

2

3K)

0

1

2x 10

−4

L)

0 200 400 6000

0.2

Days

M)

DIN

+D

IP+

TSS

0 200 400 6000

1

2

Days

N)

0 200 400 6000

1

2

3

Days

O)

0 200 400 6000

1

2x 10

−4

Days

P)

0.1x 2x 10x 100x Control

Figure 4: Sensitivity of seagrass to changing catchment loads. Columns give seagrass

biomass (mol N m−2), water column light attenuation (m−1), phytoplankton light attenu-

ation (m−1) and seagrass light availability (mol photon m−2 s−1). Rows show catchment

load simulations of DIN , DIN + DIP , DIN + TSS and DIN + DIP + TSS. The grey

shaded areas represent the time the lake is open to the ocean.

40


121824364860

A) DIN (mol N m−3)

Cyc

le D

urat

ion

(mth

s)

2

4

6

x 10−4 B) DIP (mol N m−3)

3.2

3.43.63.8

4x 10

−6

121824364860

C) PS (mol N m−3)

Cyc

le D

urat

ion

(mth

s)

1.3

1.4

1.5

1.6

x 10−4 D) PL (mol N m−3)

1

1.5

2

x 10−4

121824364860

E) ZS (mol N m−3)

Cyc

le D

urat

ion

(mth

s)

6

7

8x 10

−4 F) ZL (mol N m−3)

5

5.5

6

x 10−4

7 14 30 60121824364860

G) EP (mol N m−2)

Open Duration (d)

Cyc

le D

urat

ion

(mth

s)

4.24.44.64.855.2

x 10−5

7 14 30 60

H) SG (mol N m−2)

Open Duration (d)

0.04750.0480.04850.0490.0495

Figure 5: The contours represent the mean biomass of the state variables at a lake opening

height of 3.5 m. The mean value is calculated from the final closed phase. The white

crosses show the location of the mean simulation values used to create the contours

41


0

1

2x 10

−3 A) DIN

mol

N m

−3

0

1

2x 10

−5B) DIP

mol

P m

−3

0

0.5

1x 10

−3 C) PS

mol

N m

−3

0

1

2x 10

−3D) PL

mol

N m

−3

0

1

2x 10

−3 E) ZS

mol

N m

−3

0

1

2x 10

−3F) ZL

mol

N m

−3

0 100 200 300 400 500 600 7000

1x 10

−4 G) EP

mol

N m

−2

Days0 100 200 300 400 500 600 700

0

0.05H) SG

mol

N m

−2

Days

2.5 m (control) 4.5 m 7.0 m 12.0 m

Figure 6: The biomass of DIN , DIP , PS, PL, ZS, ZL, EP and SG for simulations as

the lake depth is changed. The lines represent the changes in estuary depth from 2.5, 4.5,

7 and 12 m. The grey shaded areas represent the time the lake is open to the ocean.

42


0

0.5

1x 10

−3

PS (mol N m−3)

2.5

m

0

1

2x 10

−3

PL (mol N m−3)

0

1x 10

−4

EP (mol N m−2)

0

0.05

SG (mol N m−2)

0

0.5

1x 10

−3

4.5

m

0

1

2x 10

−3

0

1x 10

−4

0

0.05

0

0.5

1x 10

−3

7.0

m

0

1

2x 10

−3

0

1x 10

−4

0

0.05

0 200 400 6000

0.5

1x 10

−3

Days

12 m

0 200 400 6000

1

2x 10

−3

Days0 200 400 600

0

1x 10

−4

Days0 200 400 600

0

0.05

Days

Nitrogen Limited Phosphorus Limited Light Limited Maximum Growth Rate Limited

Figure 7: Biomass of PS, PL, EP and SG (columns 1-4 respectively) coloured by process

limiting growth rate. Depth simulations of 2.5, 4.5, 7 and 12 m are plotted across the rows.

The line shows the biomass and the colour represents the growth limitation options of

nitrogen, phosphorus, light and maximum metabolic rate. The grey shaded areas represent

the time the lake is open to the ocean.

43


0

0.5

1x 10

−8 A) PS

mol

N m

−3 s

−1

0

1

2

x 10−9B) PL

mol

N m

−3 s

−1

0

1

2

x 10−9 C) ZS

mol

N m

−3 s

−1

0

0.5

1x 10

−9D) ZL

mol

N m

−3 s

−1

0 100 200 300 400 500 600 7000

2

4

6x 10

−11 E) EP

mol

N m

−2 s

−1

Days0 100 200 300 400 500 600 700

0

0.5

1

1.5

2x 10

−8F) SG

mol

N m

−2 s

−1

Days

2.5 m (control) 4.5 m 7.0 m 12.0 m

Figure 8: The realised growth rate for PS, PL, ZS, ZL, EP and SG as the lake depth is

changed. The colours represent the changes in estuary depth from 2.5, 4.5, 7 and 12 m.

The grey shaded areas represent the time the lake is open to the ocean.

44


0

0.5

1

1.5x 10

−3

Sta

ndar

d D

evia

tion A) 2 Month Open Phase

DINDIPPSPLZSZL

0

0.5

1

1.5x 10

−3

Sta

ndar

d D

evia

tion B) 1 Month Open Phase

DINDIPPSPLZSZL

0 100 200 300 400 500 600 700 800 900 10000

0.5

1

1.5x 10

−3

Sta

ndar

d D

evia

tion

Days

C) 2 Week Open Phase

DINDIPPSPLZSZL

Figure 9: The standard deviation of state variable biomass for 10 ensemble members for

the first 3 years of the simulation. The plots represent A) 2 week open phase, B) 1 month

open phase and C) 2 month open phase. The lines represent the 6 state variables of DIN ,

DIP , PS, PL, ZS and ZL. The grey shaded areas represent the time the lake is open to

the ocean.

45

Documents

Ecological response of an intermittently open lagoon to