Modelling the relative importance of internal and external nutrient loads on water column nutrient concentrations and phytoplankton biomass in a shallow polymictic lake

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

available at www.sciencedirect.com

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

Modelling the relative importance of internal and external nutrient loads on water column nutrient concentrations and phytoplankton biomass in a shallow polymictic lake夽 David F. Burger ∗ , David P. Hamilton, Conrad A. Pilditch Centre for Biodiversity and Ecology Research, University of Waikato, Private Bag 3105, Hamilton, New Zealand

a r t i c l e

i n f o

a b s t r a c t

Article history:

Lake Rotorua is a large (area 79 km2 ), shallow (mean depth 10.8 m), polymictic lake

Received 8 August 2006

in central North Island, New Zealand. The lake is eutrophic, with a mean external

Received in revised form

aerial load of 18.5 mg m−2 d−1 for total nitrogen and 1.2 mg m−2 d−1 for total phosphorus.

12 September 2007

Blooms of cyanobacteria and occasional anoxia of bottom waters occur during summer

Accepted 27 September 2007

(December–March). We used a vertically resolved water quality model, DYRESM–CAEDYM,

Published on line 3 December 2007

to examine the relative importance of internal and external nutrient inputs on water column nutrient concentrations and phytoplankton biomass, with particular emphasis on

Keywords:

cyanobacteria. External nutrient loads associated with nine major inflows to the lake and

Lake management

three additional inflows representing smaller geothermal and coldwater flows and residual

Nitrogen

flows, were represented as inputs to the model. Other forcing inputs to the model included

Phosphorus

local meteorological data, discharge from the only outflow, the Ohau Channel, and mea-

Cyanobacteria

sured rates of sediment nutrient release obtained from benthic chamber deployments which

Lake Rotorua

were used to prescribe ranges of sediment nutrient release that were simulated dynamically within the model. Profiles of water column nutrient concentrations, surface chlorophyll a concentrations and continuous temperature and dissolved oxygen measurements were used to validate the model. Simulated water column temperature and soluble reactive phosphorus (SRP) and ammonium (NH4 ) concentrations closely matched field measurements, and captured the timing and duration of stratification events as well as subsequent changes in bottom water nutrient concentrations. Surface water concentrations of chlorophyll a were also similar between simulated and observed data. Model simulations indicate that reductions in sediment nutrient fluxes would be more effective in reducing cyanobacterial biomass than similar proportional reductions in catchment fluxes, due to the coincidence of large sediment nutrient release events with high cyanobacterial biomass. This finding indicates that only a significant and prolonged reduction in external loads, which in turn reduces internal loads, will ultimately reduce cyanobacterial biomass in Lake Rotorua. © 2007 Elsevier B.V. All rights reserved.

夽

This paper has not been submitted elsewhere in identical or similar form. Corresponding author. Present address: WL Delft Hydraulics, P.O. Box 177, 2600 MH Delft, The Netherlands. Tel.: +31 15 285 85 85; fax: +31 15 285 85 82. E-mail address: [email protected] (D.F. Burger). 0304-3800/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2007.09.028 ∗

412

1.

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Introduction

Increased nutrient loads from anthropogenic sources have led to eutrophication of many aquatic systems worldwide (Wetzel, 1992). Attempts to reduce water column nutrient concentrations and mitigate phytoplankton blooms have generally focused on reductions of point sources from the catchment (Søndergaard et al., 2003). Even with severe constraints on point source loads to lakes, recovery from eutrophication has frequently been delayed due to high rates of internal recycling of nutrients between the sediments and overlying water column (Marsden, 1989; Søndergaard et al., 2003; Jeppesen et al., 2005). In eutrophic lakes in particular, fluxes of nitrogen (N) and phosphorus (P) from the sediments to the water column often represent an important source of nutrients for primary productivity (Forsberg, 1989; Søndergaard, 1989). Modelling assessments of lake eutrophication have often used mass balance equations (O.E.C.D., 1982), which provide a useful tool with which to assess the response of phytoplankton biomass to changes in catchment P loads. However, these models do not capture the complex ecological phenomena that occur in natural systems, such as phytoplankton succession, potential for nutrient co-limitation or temporal evolution of recovery (Arthonditsis and Brett, 2005a). Coupled hydrodynamic–ecological models are, therefore, often used to capture these phenomena and to link transport processes with biogeochemical cycles (Campos et al., 2001; Chan et al., 2002; Chen et al., 2002; Romero et al., 2004; Arthonditsis and Brett, 2005a). Applications of these models have frequently involved gross simplifications of nutrient cycling and regeneration associated with bottom sediments, as well as limited validation of sediment nutrient release parameters. Sediment nutrient release rates have been found to be highly sensitive in model simulations (e.g. Schladow and Hamilton, 1997), leading to both overestimates and underestimates of water column nutrient concentrations (e.g. Romero et al., 2004; Arthonditsis and Brett, 2005b). The use of acquired measurements and understanding of sediment nutrient fluxes in a coupled hydrodynamic–ecological model provides an opportunity to evaluate the relative contributions of external and internal nutrient loads in a lake ecosystem. In polymictic lakes in particular, there may be tight coupling between the bottom sediments and water column, and strong seasonal and interannual variability in nutrient availability and phytoplankton phenology. While many of these processes can be examined individually (e.g. Burger et al., 2007a), the use of an interdisciplinary model to capture this complexity may provide important insights into the eutrophication and management of eutrophic lakes. In shallow polymictic lakes, the subtle balance of mixing and stratification is especially relevant to the vertical distributions of nutrients and phytoplankton, and this type of system provides a severe test for inter-disciplinary models. Lake Rotorua, a eutrophic polymictic lake in the North Island, New Zealand, has shown little change in lake trophic status despite removal of treated wastewater inputs from the lake in 1991 (Rutherford et al., 1996). The removal of

wastewater inputs has had minimal impact on water column nutrient and chlorophyll a (chl-a) concentrations and large blooms of potentially toxic cyanobacteria have occurred repeatedly during summer (December–March). Measurements of total internal loads of soluble reactive phosphorus (SRP) and ammonium (NH4 ) contributed by the sediments may be comparable to external loads, but also show considerable variability depending on the frequency and duration of stratification events (Burger et al., 2007a). These loads may be far more accurately assessed with a model that captures the timing and duration of stratification events that play a critical role in the large changes on SRP and NH4 in bottom waters. We hypothesise that internal nutrient loads, derived from sediment nutrient release, may be at least as important as external nutrient loads in the dynamics of water column nutrient concentrations and phytoplankton biomass in Lake Rotorua. The primary objective of this study was to apply a coupled hydrodynamic–ecological model to simulate current external and internal loading rates, and then use the model to examine and quantify the effects of reducing external or internal load on lake water column nutrient concentrations, which in turn are integral to the growth and succession of phytoplankton in Lake Rotorua.

1.1.

Study site

Lake Rotorua is a large (area 79 km2 ), shallow (mean depth 10.8 m) lake of volcanic origin in central North Island, New Zealand (Fig. 1). The lake catchment area is 425 km2 and is dominated by agriculture (48%) and plantation forestry (23%). There are nine major inflows (mean flow rate 0.22–2.75 m3 s−1 ), nine minor coldwater streams and eight minor geothermal streams (flow <0.22 m3 s−1 ). Collectively, these major and minor inflows, ungauged flow and rainfall are estimated to contribute a total external aerial load of 18.5 mg m−2 d−1 for total nitrogen (TN) and 1.2 mg m−2 d−1 for total phosphorus (TP) (see Burger et al., 2007a). Lake Rotorua has a water residence time of 1.5 years based on discharge from the only surface outflow, Ohau Channel. The lake is eutrophic (Rutherford et al., 1996) with annual mean concentrations of TP and TN of 0.055 and 0.814 mg L−1 , respectively (Burger et al., 2005). The phytoplankton community of the lake is dominated by cyanobacteria, Anabaena sp. and Microcystis sp., which exhibit co-limitation of N and P during summer months (Burger et al., 2007b). Until 1991, Lake Rotorua received wastewater from Rotorua city (population 60,000), which contributed aerial loads of up to 1.2 and 5.2 mg TN m−2 d−1 (White et al., 1992). The lake stratifies for periods of up to 4 weeks in summer, and stratification events are associated with rapid reductions in bottom water concentrations of dissolved oxygen (DO) and increases in NH4 and SRP (Burger et al., 2005). Sediment nutrient fluxes, based on benthic chamber measurements, are as high as 85 and 2200 mg NH4 m−2 d−1 (Burger et al., 2007a) and fluxes of SRP do not appear to be enhanced significantly by anoxic bottom waters. High sedimentation rates of organic matter are probably responsible for sustaining the high sediment nutrient fluxes (Burger, 2006).

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

2.

Methods

2.1.

Model description

DYRESM is a one-dimensional model which resolves around vertical distributions of temperature, salinity and density in lakes and reservoirs based on a dynamic Lagrangian layer structure, which simulates the lake as horizontally uniform layers that expand and contract in response to heat, mass and momentum exchanges (see Imberger and Patterson, 1981; Gal et al., 2003). DYRESM has been coupled to the ecological model CAEDYM, which simulates up to seven phytoplankton groups (from taxa to species), DO, and several species of organic and inorganic nitrogen, phosphorus and carbon, using a series of partial differential equations that are characterised by rate constants (Robson and Hamilton, 2004). These rate constants are defined by the user and vary in the model in response to other environmental variables (e.g. temperature, DO, etc.). The bottom sediments are included in the model as a permanent sink for particulate matter that settles out of the water column, with releases of dissolved nutrients from the sediments prescribed from overlying water column properties (e.g. temperature, DO, pH). A concise documentation of CAEDYM and its applications to lake and estuarine systems are given in Robson and Hamilton (2004) and Romero et al. (2004), respectively.

413

In the present study, the coupled DYRESM–CAEDYM model was run on an hourly time-step between 1 July 2001 and 30 June 2004, with daily input data for inflows, meteorology and one outflow and a daily output corresponding to midday. A geothermal heat flux of 12 MW was prescribed from the sediments to the water column to represent geothermal heat exchange observed in benthic chamber experiments (Burger, 2006), which slightly enhanced vertical convective circulation in the bottom waters and allowed better representation of water column temperatures. The geothermal heat flux model was identical to that used to model the hydrodynamics of other New Zealand lakes (e.g. Spigel et al., 2001; Hamilton et al., 2005).

2.2.

Meteorological data

Daily meteorological data required as input to the DYRESM–CAEDYM model were taken from Rotorua airport climate station adjacent to the lake (Fig. 1). Data included total daily rainfall and daily averages of hourly measurements of air temperature (◦ C), wind speed, short wave radiation and cloud cover which were used to derive longwave radiation inputs (Fig. 2). Daily averages of hourly wet and dry bulb temperature and atmospheric pressure were used to derive mean daily water vapour pressure input to the model (Antenucci and Imerito, 2002) (Fig. 2).

Fig. 1 – Map of Lake Rotorua and the nine dominant inflows and sub-catchments, nine minor coldwater inflows and eight minor geothermal inflows represented collectively in the model, lake sampling site (S1), Rotorua Airport climate station and Rotorua city.

414

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

between monthly measurements for the period July 2001–June 2004. The eight geothermal inputs and nine minor streams were represented as two individual inflows in the model, as they contributed only 0.3 and 1.2%, respectively, of total lake inflows (Table 1, Fig. 3). Measured data for these inflows were restricted to monthly values collected between July 1992 and October 1994. Accordingly, daily discharges over the current study period were estimated as a fixed proportion of the total discharge of all major inflows. Ungauged flows were estimated as the unknown term in a daily water balance equation that included the major and minor inflows, change in lake water storage based on hypsography and 7-day mean water level, to reduce short-term fluctuations from strong winds, rainfall, outflow from Ohau Channel and evaporation (Fig. 3). Water level records were taken from Mission Bay on the northeastern side of the lake with interruptions in the data, typically only a few days, filled by linear interpolation. Mean daily discharge through the Ohau Channel (Fig. 1) was derived from a continuous water level recorder on the outflow and stage height–flow relationships. Water loss due to evaporation was calculated from the daily average evaporative heat flux (Fischer et al., 1979) using wind speed from Rotorua airport climate station and daily mean water temperatures from a temperature logger (ODYSSEY, Dataflow Systems Ltd.) at 0.5 m depth at a central lake station (S1, Fig. 1).

Fig. 2 – Meteorological data used as input to the DYRESM model (July 2001–June 2004), including (A) air temperature (T), (B) short wave radiation (SW), (C) cloud cover (CC), (D) vapour pressure (eS ) and (E) wind speed at a reference height of 10 m above the lake (u10 ). All data were collected at the Rotorua Airport climate station and represent daily means calculated from hourly measurements.

2.3.

Inflow and outflow data

The nine major inflows were assigned as independent inputs into the model (Table 1, Fig. 3). Continuous flow data for daily input to the model were available for only one major inflow, Ngongotaha Stream, during the study period of July 2001–June 2004. Several years of historical flow data were available, however, for the Utuhina Stream (January 1991–April 1997), Waingaehe, Waiohewa and Waiowhiro streams (July 1992–June 1995). In addition, all major inflows were gauged monthly between July 1992 and June 2004. To estimate flows where data were not available, linear correlations were carried out between the Ngongotaha Stream discharge and that from each of the other major inflows except the Hamurana Stream, using all available data between 1991 and 2004 (mean Pearson correlation coefficient (R = 0.77). The Hamurana Stream, a large, predominantly spring-fed inflow, was assigned a daily flow based on linear interpolation

Fig. 3 – Mean daily (A) inflows and (B) outflows for Lake Rotorua, July 2001–June 2004. For (A), major represents the nine major inflows, ungauged is unmeasured inflows due to groundwater, ungauged catchment and storm water runoff and minor represents all 17 minor geothermal and coldwater inflows (see Table 1). For (B), outflow represents discharges from the Ohau Channel.

415

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Table 1 – Mean (July 2001–June 2004) discharge and total annual nutrient loads of soluble reactive phosphorus (SRP), total phosphorus (TP), ammonium (NH4 ), nitrate (NO3 ) and total nitrogen (TN), for Lake Rotorua inflows represented in the model Flow (m3 s−1 )

Inflow

SRP (t y−1 )

TP (t y−1 )

NO3 (t y−1 )

NH4 (t y−1 )

TN (t y−1 )

Awahou Hamurana Ngongotaha Puarenga Utuhina Waingaehe Waiohewa Waiowhiro Waiteti Minor coldwater Minor geothermal Residual flow

1.6 2.7 1.5 1.6 1.6 0.2 0.3 0.3 1.1 0.21 0.05 3.2

3.5 7.5 1.6 2.4 2.8 0.7 0.3 0.4 1.4 0.40 0.31 6.1

4.0 8.2 2.2 3.2 3.3 0.8 0.5 0.5 2.2 0.48 0.46 7.3

58.9 60.0 35.9 50.5 37.8 9.1 12.4 8.6 42.2 5.99 0.05 91.5

0.4 1.0 0.7 3.4 2.2 0.1 12.8 0.2 0.9 0.41 0.79 6.2

62.3 66.5 48.6 64.3 49.0 9.8 25.0 9.9 46.8 7.25 1.49 110.6

Total

14.4

27.5

33.1

413.0

29.2

501.5

Minor coldwater and minor geothermal represent nine and eight collective inflows, respectively, as shown in Fig. 1.

2.4.

Inflow water quality

Water quality variables for each of the model inflows included daily estimates of water temperature (◦ C), DO, SRP, TP, NH4 , nitrate (NO3 ) and TN concentrations (mg L−1 ), and pH. For the major inflows, values were estimated from linear interpolation between monthly measurements collected over the study period (Environment Bay of Plenty, unpublished data). Missing monthly measurements were replaced by monthly mean values calculated over the whole study period. For all streams except Awahou, water temperatures were estimated from daily mean air temperature based on correlations (R2 > 0.71) between Rotorua airport temperature and monthly stream temperature. For Awahou Stream, where water temperature was less closely correlated with air temperature (R2 = 0.38) and annual variability in water temperatures was <2 ◦ C, linear interpolation between monthly measurements was used to derive daily data. For all major inflows, DO concentrations were assumed to be at saturation and were estimated from the assigned daily water temperature (e.g. Hamilton and Schladow, 1997). For the minor coldwater and geothermal inflows, measurements of water temperature, DO and nutrient concentrations (TP, SRP, TN, NH4 and NO3 ), and pH were available only for the period July 1992–October 1994 (Environment Bay of Plenty, unpublished data). Based on these data, minor coldwater inflows had similar composition to the major inflows, and each variable was, therefore, approximated as a volumetric mean value for major inflows on each day of the modelling period. Daily values for each water quality variable for the residual inflow were also estimated by this method. Geothermal inflows had very different characteristics to coldwater inflows, with volumetric means showing high temperature (34.9 ◦ C), high nutrient concentration, particularly SRP and NH4 (0.187 and 0.479 mg L−1 , respectively) and low DO concentration (2.4 mg L−1 ). A constant value for each day was applied for each variable in the geothermal inflows, calculated as the volumetric mean of all eight geothermal inflows for the period July 1992–October 1994.

2.5.

Phytoplankton

The dynamics of two phytoplankton groups, represented by equivalent chl-a concentration, were simulated in the model; buoyant species representative of cyanobacteria and nonbuoyant species considered to be representative of other major taxa, mostly chlorophytes and diatoms. Phytoplankton parameters for these groups were assigned based on literature values (e.g. Holm and Armstrong, 1981; Grover et al., 1999; Wallace and Hamilton, 1999; Robson and Hamilton, 2004) or were calibrated within ranges given by literature values based on visual comparisons between model output and field measurements of chl-a (Table 2). Initial water column concentrations of chl-a for the two phytoplankton groups were proportioned according to cell counts (Burger, 2006). The effects of zooplankton grazing on phytoplankton were not simulated explicitly; however, their contribution to phytoplankton mortality was included in the model through a phytoplankton loss term that also included respiration.

2.6.

Sediment parameters

Sediment SRP and NH4 release rates and sediment oxygen demand inputs to CAEDYM were based on benthic chamber experimental measurements conducted in Lake Rotorua within the period of modelling (see Burger et al., 2007a). The in situ measurements included quantification of inorganic sediment nutrient release rates for four periods at three sites between February 2003 and January 2004 (Burger et al., 2007a). The range of release rates of SRP and NH4 obtained using the chambers were within independent estimates of sediment nutrient release based on changes in hypolimnion SRP and NH4 concentration during a stratification event (Burger et al., 2005). Maximum SRP and NH4 sediment release rates of 80 and 280 mg m−2 d−1 , respectively, were specified in the model with the daily fluxes predicted based on dissolved oxygen concentration and pH in the overlying waters.

416

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Table 2 – Parameter values used in CAEDYM Parameter

Unit

Value

References/remarks

Dissolved oxygen parameters Temperature multiplier for sediment oxygen demand Half-saturation constant for sediment oxygen demand Sediment oxygen demand

Dimensionless mg L−1 g m−2 d−1

1.08 0.4 2.8

Nitrogen parameters Denitrification rate coefficient Half-saturation constant for DO for denitrification Temperature multiplier for denitrification Nitrification rate coefficient Half-saturation constant for DO for nitrification Temperature multiplier for nitrification Maximum potential sediment ammonium release rate

d−1 mg L−1 Dimensionless d−1 mg L−1 Dimensionless g m−2 d−1

0.5 5 1.08 0.01 1 1.08 0.28

Phosphorus parameters Maximum potential sediment phosphate release rate

g m−2 d−1

0.08

Burger et al. (2007a)

Phytoplankton parameters Maximum potential growth rate at 20 ◦ C Irradiance parameter non-photoinhibited growth Photoinhibited saturation irradiance Half-saturation constant for phosphorus uptake Half-saturation constant for nitrogen uptake Minimum internal nitrogen concentration Maximum internal nitrogen concentration Maximum rate of nitrogen uptake Minimum internal phosphorus concentration Maximum internal phosphorus concentration Maximum rate of phosphorus uptake Temperature multiplier for growth limitation Standard temperature for growth Optimum temperature for growth Maximum temperature for growth Respiration rate coefficient Temperature multiplier for respiration Settling velocity

d−1 ␮E m−2 s−1 ␮E m−2 s−1 mg L−1 mg L−1 mg N (mg chl-a)−1 mg N (mg chl-a)−1 mg N (mg chl-a)−1 d−1 mg P (mg chl-a)−1 mg P (mg-chl-a)−1 mg P (mg chl-a)−1 d−1 Dimensionless ◦ C ◦ C ◦ C d−1 Dimensionless m d−1

Cyano, other 0.7, 1.8 120, 20 200, 10 0.006, 0.010 0.030, 0.045 2.5, 2.0 4.0, 4.5 1.5, 3.0 0.50, 0.25 2.0, 1.3 0.3, 1.0 1.08, 1.06 20, 12 28, 23 35, 30 0.05, 0.12 1.09, 1.05 4.32, 0.52

Robson and Hamilton (2004) Robson and Hamilton (2004) Wallace and Hamilton (1999) Holm and Armstrong (1981) Hamilton and Schladow (1997) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Hamilton and Schladow (1997) Hamilton and Schladow (1997) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Robson and Hamilton (2004) Romero et al. (2004)

Robson and Hamilton (2004) Burger (2006) Burger (2006)

Robson and Hamilton (2004)

Robson and Hamilton (2004) Burger et al. (2007a)

“Cyano” represents cyanobacteria and “other” represents a combined diatom and chlorophyte assemblage.

2.7.

Model validation

The model was calibrated against field data over a one-year period (commencing 1 July 2001) using water column comparisons of temperature and DO, and nutrient (TP, SRP, TN, NH4 and NO3 ) and chl-a concentrations. The final two years of the study period (commencing 1 July 2002) were used to validate the calibrated model. Field data were obtained from a variety of sources. Daily average water temperature profiles were derived from measurements taken at 5 min intervals from thermistors (ODYSSEY Ltd) deployed at 2 m depth intervals at a central lake station (S1, Fig. 1) between November 2002 and March 2004. Mean daily bottom (depth 19 m, hereafter referred to as bottom waters) DO for the period February 2003–March 2004 was measured with an in situ dissolved oxygen sensor (Greenspan Technology Ltd.) that logged data at 30 min intervals. Conductivity–temperature–depth (CTD) profiles (Seabird Electronics) taken approximately monthly at S1 (Fig. 1) between July 2001 and June 2004 provided vertical distributions of temperature, DO and photosynthetically available radiation (PAR, Licor Ltd.) (Gibbons-Davies, 2003; Scholes, 2004a, 2004b). Concentrations of chl-a, NH4 , NO3 , SRP, TP

and TN were analysed from samples collected from a depthintegrated sample of the surface-mixed layer (0–8 m) and from discrete bottom water (depth 19 m) samples, collected concurrently with CTD profiles.

2.8.

External versus internal loading

Two scenarios of nutrient loading were examined in model simulations. In the first scenario, the current total external nutrient load of 18.5 mg TN m−2 d−1 was reduced by 50% to equate to a 250 t y−1 reduction of TN, based on targets set for Lake Rotorua (see Rutherford et al., 1989). The aim of the proposed target reductions was to restore lake water quality to levels similar to the 1960s, before persistent summer phytoplankton blooms were observed in the lake. The external TP load of 1.2 mg m−2 d−1 was also reduced by 50% in the first management scenario. The second scenario involved a reduction of internal nutrient loads by 50%, to allow direct comparison with the external load reduction scenario. The maximum sediment NH4 fluxes were reduced from 280 to 140 mg m−2 d−1 and SRP fluxes from 80 to 40 mg m−2 d−1 in the simulations.

417

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Table 3 – Physical data inputs and parameters for DYRESM model Coefficient/variable

Unit

Value

Benthic boundary layer thickness Bulk aerodynamic momentum transport coefficient Critical wind speed Emissivity of water surface Lake latitude Background light extinction coefficient (Kd ) Mean albedo of water Minimum/maximum layer thickness Potential energy mixing efficiency Shear production efficiency Vertical mixing coefficient Wind stirring efficiency Geothermal heat flux

m

0.1 0.0013 4 0.96 −38 0.8 0.07 1.0/4.0 0.25 0.08 400 0.6 12

3.

Results

3.1.

Model simulations

m s−1 ◦

N m−1 m

mW

Simulations of water column temperature over the whole period (June 2001–July 2004) agreed well with measured temperature using the model parameters given in Table 3 (R > 0.99, Table 4). Variations between simulated and observed values were generally less than 1.5 ◦ C through the water column (Fig. 4). The timing and duration of stratification events were captured accurately over the two summer periods for which field thermistor data were available (Fig. 4), although simulations slightly underestimated the depth of the thermocline in January 2004. Over the simulation period, simulations of temperature (±1 standard deviation (S.D.)) at 1.5 m were 0.76 ± 0.58 ◦ C lower than corresponding observed values, with differences greater in winter (June–September, −1.02 ± 0.18 ◦ C) than in summer (December–March, −0.47 ± 0.70 ◦ C). Mean simulated temperature at depth 19 m was 0.32 ± 0.76 ◦ C lower than observed values over the entire simulation period, with differences also higher in summer than in winter (−0.01 ± 0.89 and −0.73 ± 0.23 ◦ C, respectively).

Reference Stull (1988) Imberger and Patterson (1981) Measured Patten et al. (1975) Spigel et al. (1986) Spigel et al. (1986)

The parameters for DO, N and P used in the ecological sub-model are listed in Table 2. Model simulations tended to underestimate DO in the surface waters (depth integrated 0–8 m) (Fig. 5), with a mean difference of −1.05 ± 1.20 mg L−1 (R value > 0.6, Table 4). At 19 m, concentrations of DO were over-estimated by the model (mean −0.69 ± 1.21 mg L−1 ), particularly during stratification events when DO measured in bottom waters declined to zero. Simulations of SRP and TP captured not only the observed increases in hypolimnion P during stratification, but also the inter-annual variability observed over the study period (Fig. 6, R > 0.78, Table 4). Mean differences in SRP and TP concentration between model simulations and monthly field measurements were within 0.001 ± 0.009 and 0.002 ± 0.015 mg L−1 , respectively, in both surface and bottom waters throughout the study period. Bottom water NH4 concentrations were similar between the model simulation and field measurements over the whole period (mean difference 0.028 ± 0.064 mg L−1 , Table 4), although the model underestimated the maximum summer concentration of NH4 in both 2003 and 2004 (Fig. 7B). Simulations of NH4 in the surface waters were also well reproduced by the model (Fig. 7A, mean difference 0.006 ± 0.034 mg L−1 ), except for the elevated con-

Table 4 – Statistical comparison between model simulations and monthly field measurements of surface (depth integrated, 0–8 m) and bottom (depth 19 m) temperature and concentrations of dissolved oxygen (DO), soluble reactive phosphorus (SRP), total phosphorus (TP), ammonium (NH4 ), nitrate (NO3 ), total nitrogen (TN) and chlorophyll a (chl-a) over the calibration (Cal., July 2001–June 2002) and validation (Val., July 2002–June 2004) periods Surface-mixed layer RMSE

◦

Temperature ( C) DO (mg L−1 ) SRP (mg L−1 ) TP (mg L−1 ) NH4 (mg L−1 ) NO3 (mg L−1 ) TN (mg L−1 ) chl-a (␮g L−1 )

Bottom waters R

Cal.

Val.

0.97 1.23 0.008 0.008 0.024 0.005 0.106 8.5

0.86 1.74 0.007 0.011 0.039 0.013 0.232 13.1

Cal. 0.99*** 0.89*** 0.44 0.85*** 0.40 −0.29 0.38 0.45

RMSE Val. 0.99*** 0.55*** 0.29 0.40* −0.01 0.17 0.41* 0.29

R

Cal.

Val.

0.88 1.38 0.010 0.011 0.024 0.023 0.137

0.67 1.38 0.008 0.016 0.085 0.015 0.296

Cal. 0.99*** 0.89*** 0.72*** 0.82*** 0.83*** −0.21 0.40

Val. 0.99*** 0.91*** 0.91*** 0.76*** 0.96*** 0.38* 0.59**

RMSE represents root mean square error and R the correlation coefficient, with significant correlations indicated by *p < 0.05, **p < 0.01 and ***p < 0.001. Field data to validate chl-a concentrations in bottom waters were not available.

418

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Fig. 4 – Temperature (◦ C) in Lake Rotorua derived from (A) daily means of loggers deployed at 2 m depth intervals, and (B) daily output from model simulations, for the period November 2002–April 2004.

centrations in observed NH4 in autumn of the two years (April–May). Concentrations of NO3 were underestimated in surface waters (mean difference 0.005 ± 0.017 mg L−1 , Fig. 7C) and overestimated in bottom waters (mean difference 0.001 ± 0.018 mg L−1 , Table 4, Fig. 7D). Model simulations of TN showed reduced variability compared with field measurements, and were much lower than the field data in both the

Fig. 5 – (A) Surface (depth 1.5 m) and (B) bottom (depth 19 m) dissolved oxygen concentrations in Lake Rotorua derived from model simulations, and monthly or daily observed values, for the period July 2001–June 2004.

Fig. 6 – Comparison between Lake Rotorua model simulations (line) and monthly field measurements (points) for soluble reactive phosphorus (SRP) concentrations in (A) surface waters (depth 0–8 m) and (B) bottom waters (depth 19 m), and total phosphorus (TP) concentrations in (C) surface waters and (D) bottom waters, from July 2001–June 2004. Note surface and bottom water concentrations are on different scales.

surface (mean difference 0.144 ± 0.132 mg L−1 , Fig. 7E) and bottom (mean difference 0.1472 ± 0.183 mg L−1 Fig. 7F) waters. In model simulations, concentrations of DO in the bottom waters during stratification events did not appear to strongly influence the release of dissolved nutrients from the sediments. For example, in summer 2001–2002, when the lake was stratified on three occasions for periods of 11–20 days, increases in SRP and NH4 concentrations occurred even when DO concentrations remained greater than 5 mg L−1 at 19 m. The onset of stratification and subsequent bottom water anoxia were characterised by calm conditions, with mean daily wind speeds of up to 4.7 m s−1 observed on the day prior to each of the three events in 2001–2002 but mean daily wind speeds during stratification of only 3.3 m s−1 . Total chl-a concentrations in surface waters were reproduced relatively well by model simulations, which captured both the seasonal and inter-annual variability in observed data (Fig. 8). For the simulation period (July 2001–June 2004), observed mean monthly concentrations of total chla (±S.D.) in surface waters (0–8 m) were 30.9 ± 8.9 ␮g L−1 in summer (December–March) and 17.5 ± 9.2 ␮g L−1 in winter (June–September). Corresponding daily mean chl-a concentra-

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

419

Fig. 8 – (A) Comparison between model simulations (line) and monthly field measurements (points) for chlorophyll a (chl-a) in the surface waters (depth 0–8 m) and (B) model simulations of cyanobacteria (Cyano) and diatoms plus chlorophytes (other), expressed as chlorophyll a concentration, in the lake surface waters, from July 2001–June 2004.

Fig. 7 – Comparison between Lake Rotorua model simulations (line) and monthly field measurements (points) for ammonium (NH4 ) concentrations in (A) surface waters (depth 0–8 m) and (B) bottom waters (depth 19 m), nitrate concentration (NO3 ) in (C) surface waters and (D) bottom waters, and total nitrogen concentration (TN) in (E) surface waters and (F) bottom waters, from July 2001–June 2004. Note surface and bottom water concentrations are on different scales.

tions simulated by the model were 25.8 ± 6.1 ␮g L−1 in summer and 17.4 ± 3.1 ␮g L−1 in winter. In the model simulations, 66% of summer phytoplankton biomass (i.e. represented by chl-a) was contributed by cyanobacteria and 34% by diatoms and other non-buoyant phytoplankton species (Fig. 8). In winter, cyanobacteria contributed 60% and other species 40% of total chl-a concentration. During summer, cyanobacterial biomass increased in the surface waters, particularly during stratification events,

with a mean daily difference in concentration between the surface and bottom waters of 32%. Concentrations of cyanobacteria were up to 67% higher in the surface waters on some days (2 February 2002), and reached values of up to 43 ␮g chl-a L−1 (30 January 2004, Fig. 8B). Increases in cyanobacterial biomass in model simulations occurred at the onset of stratification, and before dissolved nutrient concentrations released from the sediments during stratification were circulated through the water column. During stratification, concentrations of diatoms and other species in the surface waters generally decreased, while cyanobacteria concentrations continued to increase, at rates of up to 4.9 ␮g chl-a L−1 d−1 (18 January 2003), declining only during partial-mixing events such as that observed on 7 February 2003. Cyanobacteria densities declined with breakdown in stratification, and the high concentrations of SRP and NH4 declined rapidly in association with increases in diatom densities. Maximum rates of increase of diatoms and other species after stratification were 2.9 ␮g chl-a L−1 d−1 (2 February 2002).

3.2.

Internal versus external loading

Mean daily sediment nutrient release rates computed as output from the model were 28.1 mg m−2 d−1 for NH4 and 8 mg m−2 d−1 for SRP. These internal nutrient loads were greatest during summer (December–March), with mean daily fluxes of 40.2 mg NH4 and 11.5 mg SRP m−2 d−1 . Mean daily loads of external nutrients used as input to the model were 18.5 mg TN and 1.2 mg TP m−2 d−1 over the corresponding period. Concentrations of SRP in surface waters showed the greatest reduction in model simulations when internal nutrient fluxes were reduced by 50% compared with the equivalent percentage reduction from external sources (Fig. 9). Mean daily summer (December–March) SRP concentrations in surface waters over the simulation period declined by 31% with the internal load reduction, compared with a 6% increase with external load reduction (Table 5). Ammonium concen-

420

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Fig. 10 – Comparisons between model simulations of current lake status (control), 50% external nutrient load reduction (external) and 50% reduction in sediment nutrient fluxes (internal) on (A) total chl-a and (B) cyanobacteria (Cyano) in the surface waters (depth 0–8 m), expressed as chlorophyll a concentration, from July 2001–June 2004.

centration declined by 33% in the surface-mixed layer with internal load reductions, but increased by 1% with external load reduction.

4. Fig. 9 – Comparisons between model simulations of current lake status (control), 50% external nutrient load reduction (external) and 50% reduction in sediment nutrient fluxes (internal) for concentrations of soluble reactive phosphorus (SRP) in (A) surface waters (depth 0–8 m) and (B) bottom waters (depth 19 m), and concentration of ammonium in (C) surface waters and (D) bottom waters, from July 2001–June 2004.

trations behaved similarly to SRP, with a large decrease in surface waters with internal load reduction (mean 48%) and an increase with external load reduction (13%) (Table 5). In bottom waters, concentrations of SRP and NH4 decreased by 36 and 48%, respectively, with internal load reduction over summer (Table 5). Mean daily concentrations of NO3 decreased by 88% in bottom waters in summer simulations of external load reduction (Table 5). Surface water concentrations of cyanobacteria also showed the greatest decrease in model simulations when internal nutrient fluxes were reduced (Fig. 10). A decline of up to 53% from the control simulation biomass was observed on some days in summer (e.g. 18 February 2002), with the largest decreases generally occurring during stratification events (Fig. 10). Mean daily cyanobacterial biomass in surface waters in summer declined by 14% in simulations of internal load reduction, compared with a 2% increase in simulations with external load reduction (Table 5). Mean daily total chl-a con-

Discussion

The model simulations of water quality in Lake Rotorua provide high frequency resolution of complex dynamics and rapid transitions of nutrients and phytoplankton in the water column of a polymictic lake, including the timing and duration of stratification events and concomitant increases in SRP and NH4 concentrations in bottom waters. The model simulations also captured the inter-annual variability associated with the duration over which stratification events persisted, which resulted in considerable differences in bottom water nutrient concentrations. However, statistical comparisons of model simulations with monthly field measurements indicate that there is a considerable amount of unexplained variability, particularly for surface water concentrations of NH4 and NO3 , although errors are within the range reported elsewhere (e.g. Arthonditsis and Brett, 2005a,b) and the larger relative errors for these nutrients are attributable to some extent to their relatively low concentrations in surface waters. Overall, simulations of nitrogen species concentrations did not replicate measured values as well as for phosphorus, though the magnitude of changes in concentrations of ammonium associated with stratification events was reproduced well by the model. It is notable that mass ratios of TN:TP were lower (5:1) for internal loads than for external loads (15.2:1), so that with greater internal loading, there may be increased likelihood of dominance by cyanobacteria (Smith, 1983). A previous application of an earlier version of the DYRESM model to Lake Rotorua (Rutherford et al., 1996) was not able to

421

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Table 5 – Daily mean (±S.D.) surface depth integrated (0–8 m) and bottom (depth 19 m) concentrations of dissolved oxygen (DO), nutrients (SRP, TP, NH4 , NO3 , TN) and total chlorophyll a (T chl-a) for summer (December–March) between July 2001 and June 2004, at current levels and those predicted to result from a 50% reduction in external and internal nutrient loading Lake depth

Variable

Units −1

Current

External

Internal

Surface

DO SRP TP NH4 NO3 TN T chl-a

mg L mg L−1 mg L−1 mg L−1 mg L−1 mg L−1 ␮g chl-a L−1

7.5 0.012 0.046 0.027 0.003 0.314 25.7

± ± ± ± ± ± ±

0.9 0.009 0.011 0.026 0.004 0.034 6.1

7.5 0.013 0.046 0.032 0.000 0.313 25.9

± ± ± ± ± ± ±

0.9 0.012 0.014 0.035 0.001 0.044 6.4

7.4 0.008 0.031 0.014 0.003 0.205 17.2

± ± ± ± ± ± ±

0.8 0.004 0.005 0.012 0.004 0.018 3.4

Bottom

DO SRP TP NH4 NO3 TN T chl-a

mg L−1 mg L−1 mg L−1 mg L−1 mg L−1 mg L−1 ␮g chl-a L−1

6.1 0.029 0.057 0.074 0.015 0.343 17.7

± ± ± ± ± ± ±

1.9 0.027 0.023 0.082 0.021 0.061 7.0

6.1 0.028 0.055 0.082 0.002 0.332 17.7

± ± ± ± ± ± ±

1.9 0.027 0.024 0.088 0.004 0.071 7.0

6.2 0.019 0.037 0.038 0.015 0.220 11.9

± ± ± ± ± ± ±

1.8 0.015 0.011 0.041 0.021 0.034 4.4

reproduce the duration and strength of thermal stratification as precisely as in this study. The reasons for the improvement in the current model application may be related to more accurate alignment of time of model output with the time when samples were collected, prescription of a geothermal heat flux from the bottom sediments, and the feedback of phytoplankton biomass to the light attenuation coefficient, as the earlier model application did not include a water quality module. Model simulations of water column temperatures were generally found to be slightly cooler than field observations. Gal et al. (2003) suggest that the DYRESM model is highly sensitive to long wave radiation inputs, which in the present study were prescribed only by daily average cloud cover. Model simulations suggest that development of cyanobacterial blooms coincide with the onset of stratification when integrated water column concentrations of cyanobacteria are high (>30 ␮g L−1 ). cyanobacterial biomass generally continued to increase over the duration of stratification in the model simulations, implying that there were sufficient nutrients to support net growth despite isolation of the surface waters from the lake sediments, which are the dominant source of nutrients to the water column during summer months. This observation suggests that interactions amongst light limitation and mixing depth may be important in regulating biomass of cyanobacteria in Lake Rotorua and supports the findings of phytoplankton bioassays (Burger et al., 2007b), which demonstrated that both light and nutrients (both N and P) limited phytoplankton biomass and influenced its community assemblage, depending on location and time of year. Parameters used to characterise the two groups of phytoplankton represented in the model were prescribed using literature values and represent only ‘average values’ over the assemblage of species represented by the two phytoplankton groups. For example, the parameters used to specify phytoplankton growth rates, nutrient uptake rates, nutrient storage, light responses and settling rates may be expected to vary between species and lakes, as well as with time for individual species (e.g. Kirk, 1994; Reynolds, 1997). In Lake Rotorua, cyanobacteria populations were dominated at various times by Anabaena planktonica or Microcystis aeruginosa, and these

species responded differently to N and P in enrichment bioassays (Burger et al., 2007b). The maximum sediment nutrient release rates specified in the model represent measured values derived from experiments conducted in the lake within the simulation period (Burger et al., 2007a), and have provided critical data for refining the accuracy of simulations. Comparisons of simulations of external and internal nutrient load reductions clearly indicate that sediment nutrient fluxes play a major role in determining water column nutrient concentrations in Lake Rotorua under present loading rates and trophic status. For nutrient concentrations in the surface-mixed layer over the summer period, a 50% reduction in sediment nutrient release rates was at least 30% more effective in reducing SRP and TP concentrations, and at least 35% more effective in reducing NH4 and TN concentrations, than the 50% reduction in external loads. Increases in summer SRP (6%) and NH4 (13%) concentrations in surface waters observed in simulations of external nutrient load reduction suggest that a reduction in external NO3 concentrations may enhance rates of sediment nutrient release, due to a decline in NO3 concentration adjacent to the lake sediments derived from plunging coldwater inflows to the lake bottom, and subsequent change in redox potential. A large reduction in NO3 concentrations may deplete electron acceptors that are important for SRP adsorption in lakebed sediments, and subsequently enhance rates of sediment SRP release to the overlying water column (Kleeberg and Kozerski, 1997). However, it is uncertain whether high concentrations of NO3 derived from external sources reach the bottom waters of Lake Rotorua, or whether this is a limitation of the current model. An increase (2%) in cyanobacterial biomass observed in simulations of external load reduction may also be attributed to the concurrent increase in dissolved nutrients observed in model runs. Reductions in summer total chl-a concentration and cyanobacterial biomass (33 and 14%, respectively) with a 50% reduction in internal nutrient load were substantially larger than for the same reduction in external load. This result suggests that sediment nutrient fluxes are currently more

422

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

important in influencing the high observed nutrient concentrations and cyanobacterial biomass in surface waters than external nutrient loads, due to coincidence of stratification and major nutrient releases with elevated cyanobacterial densities during summer. Due to the polymictic nature of the lake, stratification events are relatively short-lived and nutrients released from the lake sediments are quickly mixed into surface waters following mixing. During summer months (December–March), internal N and P loads calculated by the model represented 71 and 91%, respectively, of the total nutrient load to the lake from all sources. Not only are the total loads contributed by the sediments much greater than the equivalent external loads, but N is released in the form of NH4 which is more readily assimilated by the phytoplankton compared with NO3 , which is the dominant form (82%) of N associated with external loads. Further, the timing of maximum internal nutrient loading over summer coincides with the period of elevated temperature and high irradiance that increase phytoplankton growth rates during this period. Our model simulations demonstrate how internal loads exert a critical role in lake water column nutrient concentrations and cyanobacterial biomass in shallow, eutrophic lakes, which offsets direct responses to external load reductions and will delay responses to management actions that restrict external loads. In systems such as Lake Rotorua, only a significant and prolonged reduction in external loads will eventually reduce internal nutrient recycling. The high external nutrient loads to the lake, coupled with a moderate lake surface area to volume ratio and a water residence time of 1.5 years, suggest that the lake will remain eutrophic for some time and respond only slowly to external load reductions. The application of a inter-disciplinary water quality model to polymictic lakes is valuable in quantifying the high variability of internal nutrient loads and the influence of these loads on phytoplankton dynamics; extrapolations from experimental methods are unlikely to capture the scales of these dynamics. While such models have relatively large data requirements for input and calibration, they provide an important management tool for assisting direction in lake restoration efforts.

Acknowledgements We thank Paul Dell, John McIntosh, Paul Scholes, Glenn Ellery, Matt Bloxham and John Gibbons-Davies (Environment Bay of Plenty) for provision of the lake and inflow water quality data used to validate the model. Tora Uraoka and Jose Beya´ (The University of Waikato) and Dr Bob Spigel and Dr Kit Rutherford (NIWA) provided assistance with aspects of the modelling and data set-up.

references

Antenucci, J., Imerito, A., 2002. The CWR DYnamic Reservoir Simulation Model: DYRESM Science Manual. Centre for Water Research, University of Western Australia, Nedlands. Arthonditsis, G.B., Brett, M.T., 2005a. Eutrophication model for Lake Washington (USA). Part I. Model description and sensitivity analysis. Ecol. Model. 187, 140–178.

Arthonditsis, G.B., Brett, M.T., 2005b. Eutrophication model for Lake Washington (USA): part II, model calibration and system dynamics analysis. Ecol. Model. 187, 179–200. Burger, D.F., 2006. Dynamics of internal nutrient loading in a eutrophic, polymictic lake (Lake Rotorua, New Zealand). Unpublished Ph.D. Thesis, The University of Waikato, Hamilton. Burger, D.F., Hamilton, D.P., Pilditch, C.P., Gibbs, M.M., 2005. Sediment phosphorus release during stratification in polymictic Lake Rotorua, New Zealand. Verhandlung Internationale Vereingung de Limnologie 29 (2), 811–814. Burger, D.F., Hamilton, D.P., Pilditch, C.A., Gibbs, M.M., 2007a. Benthic nutrient fluxes in a eutrophic, polymictic lake. Hydrobiologia 584, 13–25. Burger, D.F., Hamilton, D.P., Hall, J.A., Ryan, E.F., 2007b. Phytoplankton nutrient limitation in a polymictic eutrophic lake: community versus species–specific responses. Fundam. Appl. Limnol.-Arch. Hydrobiol. 169 (1), 57–68. Campos, H., Hamilton, D.P., Villalobos, L., Imberger, J., Javam, A., 2001. A modelling assessment of potential for eutrophication ˜ of Lake Rinihue. Chil. Arch. Hydrobiol. 151, 101–125. Chan, T.U., Hamilton, D.P., Robson, B.J., Hodges, B.R., Dallimore, C., 2002. Impacts of hydrological changes on phytoplankton succession in the Swan River, Western Australia. Estuaries 25, 1406–1415. Chen, C., Ji, R., Schwab, D.J., et al., 2002. A model study of the coupled biological and physical dynamics in Lake Michigan. Ecol. Model. 152, 145–168. Fischer, H.B., List, E.G., Koh, R.C.Y., Imberger, J., Brooks, N.H., 1979. Mixing in Inland and Coastal Waters. Academic Press, San Diego, CA. Forsberg, C., 1989. Importance of sediments in understanding nutrient cycling in lakes. Hydrobiologia 176–177, 263–277. Gal, G., Imberger, J., Zohary, T., Antenucci, J.P., Anis, A., Rosenberg, T., 2003. Simulating the thermal dynamics of Lake Kinneret. Ecol. Model. 162, 69–86. Gibbons-Davies, J., 2003, Rotorua Lakes water quality 2002. Environment Bay of Plenty Environmental Report 2003/02, Whakatane. Grover, J.P., Sterner, R.W., Robinson, J.L., 1999. Algal growth in warm temperate reservoirs: nutrient-dependent kinetics of individual taxa and seasonal patterns of dominance. Arch. Hydrobiol. 145, 1–23. Hamilton, D.P., Schladow, S.G., 1997. Prediction of water quality in lakes and reservoirs. Part I. Model description. Ecol. Model. 96, 91–110. Hamilton, D.P., McBride, C., Uruoka, T., 2005. Lake Rotoiti Fieldwork and Modelling to Support Considerations of Ohau Channel Diversion from Lake Rotoiti. The University of Waikato, Hamilton. Holm, N.P., Armstrong, D.E., 1981. Role of nutrient limitation and competition in controlling the populations of Asterionella formosa and Microcystis aeruginosa in semicontinuous culture. Limnol. Oceanogr. 26, 622–634. Imberger, J., Patterson, J.C., 1981. A dynamic reservoir simulation model—DYRESM: 5. In: Fischer, H.B. (Ed.), Transport Models for Inland and Coastal Waters. Academic Press, New York, pp. 310–361. Jeppesen, E., Søndergaard, M., Jensen, J.P., et al., 2005. Lake responses to reduced nutrient loading—an analysis of contemporary long-term data from 35 case studies. Freshwater Biol. 50, 1747–1771. Kirk, J.T., 1994. Light and Photosynthesis in Aquatic Ecosystems. Cambridge University Press, 509 pp. Kleeberg, A., Kozerski, H.P., 1997. Phosphorus release in Lake Grosser Muggelsee and its implications for lake restoration. Hydrobiologia 342 (343), 9–26.

e c o l o g i c a l m o d e l l i n g 2 1 1 ( 2 0 0 8 ) 411–423

Marsden, M.W., 1989. Lake restoration by reducing external phosphorus loading: the influence of sediment phosphorus release. Freshwater Biol. 21, 139–162. O.E.C.D., 1982. Eutrophication of waters. Monitoring, assessment and control. O.E.C.D. Publication No. 42077, Paris. Patten, B.C., Egloff, D.A., Richardson, T.H., 1975. Total Ecosystem Model for a Cove in Lake Texoma, System Analysis and Simulation in Ecology. Academic Press, New York, pp. 206–423. Reynolds, C.S., 1997. Vegetative Processes in the Pelagic: A Model for Ecosystem Theory. Ecology Institute, Germany, 203 pp. Robson, B.J., Hamilton, D.P., 2004. Three-dimensional modelling of a Microcystis bloom event in the Swan River estuary. West. Aust. Ecol. Model. 174 (1–2), 203–222. Romero, J.R., Antenucci, J.P., Imberger, J., 2004. One- and three-dimensional biogeochemical simulations of two differing reservoirs. Ecol. Model. 174, 143–160. Rutherford, J.C., Dumnov, S.M., Ross, A.H., 1996. Predictions of phosphorus in Lake Rotorua following load reductions. N. Z. J. Mar. Freshwater Res. 30, 383–396. Rutherford, J.C., Pridmore, R.D., White, E., 1989. Management of phosphorus and nitrogen inputs to Lake Rotorua, New Zealand. J. Water Res. Plan. Manage. 115–133,431–439. Schladow, S.G., Hamilton, D.P., 1997. Effect of major flow diversion on sediment nutrient release in a stratified reservoir. Mar. Freshwater Res. 46, 189–195. Scholes, P., 2004a. Rotorua Lakes Water Quality 2003. Environment Bay of Plenty Environmental Publication 2004/02, Whakatane. Scholes, P., 2004b. Rotorua Lakes Water Quality 2004. Environment Bay of Plenty Environmental Publication 2004/17, Whakatane.

423

Smith, V.H., 1983. Low nitrogen to phosphorus ratios favor dominance by blue-green algae in lake phytoplankton. Science 221, 669–671. Spigel, R., Howard-Williams, C., James, M., Gibbs, M.M., 2001. A coupled hydrodynamic-ecosystem study of Lake Taupo: A preliminary model. Report to Environment Waikato. National Institute for Water and Atmospheric Research Client Report CHCO1/52, Christchurch. Søndergaard, M., 1989. Phosphorus release from a hypertrophic lake sediment: experiments with intact sediment cores in a continuous flow system. Arch. Hydrobiol. 116, 45–59. Søndergaard, M., Jensen, J.P., Jeppesen, E., 2003. Role of sediment and internal loading of phosphorus in shallow lakes. Hydrobiologia 506–509, 135–145. Spigel, R.H., Imberger, J., Rayner, K.N., 1986. Modelling the diurnal mixed layer. Limnol. Oceanogr. 31, 533–556. Stull, R.B., 1988. An Introduction to Boundary Layer Meteorology. Kluwer Academic Publishers, Dordrecht. Wallace, B.B., Hamilton, D.P., 1999. The effect of variations in irradiance on buoyancy regulation in Microcystis aeruginosa. Limnol. Oceanogr. 44, 273–281. Wetzel, R.G., 1992. Clean water: a fading resource. Hydrobiologia 243 (244), 21–30. White, E.W., Payne, G.W., Pickmere, S., 1992. The restoration of Lake Rotorua—comment on progress. Taupo Research Laboratory, File Report No. 138, DSIR Marine and Freshwater, Taupo.