1. INTRODUCTION
In the past, wetlands were often thought
of as wastelands but are now being acknowledged as important ecosystems
providing the biosphere with invaluable service functions [Costanza et al.,
1997]. The habitats and therefore functioning of the unique Lower River Murray
wetlands in South Australia are seriously under threat since subject to
cultural eutrophication, salinisation, invasion by exotic species or changed
hydrology. Only an early implementation of appropriate restoration concepts
promises a recovery and survival of these wetlands [Middleton, 1999].
The aim of
this study is to demonstrate how the use of dynamic wetland modelling
contributes to the development of robust management strategies for the
restoration of the Lower River Murray wetlands. The generic wetland ecosystem
model WETMOD was constructed based on ecosystem interactions between
macrophytes, phytoplankton, zooplankton, and dissolved inorganic nutrients in
the open water. Processes between these ecological entities prove to
be
fundamental to wetland ecosystems and are prime targets in the restoration of
degraded wetlands in the Lower River Murray region.
The model
WETMOD has been calibrated and validated by means of data from Lower River
Murray wetlands. Thus, data from the Pilby Creek wetland were used, which has
been restored by annual temporary drying since 1996, as well as from the four
degraded wetlands Lock 6, Sunnyside, Paiwalla and Reedy Creek. Each of these
four wetlands is affected by permanent inundation as a result of the river flow
regulation, but also by excess nutrients from agricultural drainage and high
abundance of common carp (Cyprinus carpio). A scenario analysis was run by WETMOD
to simulate restoration options for Reedy Creek and Lock 6 wetlands.
The
suitability of WETMOD for conducting scenario analysis forms a prerequisite for
its integration as a core unit model into the landscape model AQUALINK that is
currently under construction. The simulation of habitat conditions for each
single wetland before and after restoration scenarios will allow cumulative
assessments of landscape-wide restoration policies for the Lower Murray
floodplains.
2.
METHODS
2.1 Data
Sources
To run
specific simulations the following input data was taken from wetland specific
field data: water temperature, Secchi depth (measure of light penetration),
turbidity and initial values for phytoplankton, nitrate and phosphate. Input
data for solar radiation [Bowles et al., 1987] and water flow [Walker
and Hillman, 1982] was not
available from field data thus taken from the literature.
The input
data for Lock 6 and Pilby Creek wetlands were collected fortnightly in summer
and monthly in winter in 1997 [Marsh,
1997]. The input
data for Sunnyside and Paiwalla wetlands were collected fortnightly between
January and September, 1997 [Bartsch, 1997]. While the input data for water
temperature and turbidity of Reedy Creek wetland was collected fortnightly in
summer and monthly in winter in 2000 [Wen, unpbl data],
an estimated Secchi depth of 0.3m data was applied to the simulations
[Recknagel, pers com; Wen, pers com].
Simulation times corresponded with the time periods when samples were
collected.
2.2 Model
Design and Construction
WETMOD was
built using the dynamic simulation software STELLA v.6 which has been widely
applied in ecological modelling [Costanza
and Gottlieb, 1998]. Mass
balance models can be created through differential equations, which consider
source and sink relationships typical of ecological systems.
Figure 1.
Structural diagram of WETMOD
WETMOD was
developed based on the Patuxent Landscape Model (Pat_GEM) [Boumans et al.,
2000] and the lake ecosystem model SALMO [Recknagel and Benndorf, 1982]. The
current structure of WETMOD is diagrammatically represented in Figure 1. Figure 1 indicates that the model considers nutrient
concentrations, light levels, turbidity and water temperatures as limiting
factors for gross primary production of macrophytes and phytoplankton. Losses in biomass of the primary
producers are driven by respiration and mortality where phytoplankton is
additionally declined by sedimentation and zooplankton grazing. Growth of herb
ivory zooplankton is very much driven by water temperature and phytoplankton
biomass available for grazing. Zooplankton losses are simulated to occur
through mortality and predation. Nutrient concentrations in the open water
increase through loadings from surface runoff and the release of nutrients from
bottom sediments. Nutrient uptake by macrophytes and phytoplankton, nutrient
coprecipitation by soil particles during highly turbid events and nutrient
transport by out fluxes from wetlands are nutrient losses simulated by the
model.
2.3
WETMOD Calibration and Validation
Calibration
and validation of the wetland model was conducted for five Lower River Murray
wetlands representative for wetland categories distinguished for landscape
modelling. Wetland specific data were applied to simulate outputs comparable
with the measured data. While the generic model structure was maintained, only
6 site-specific constant parameters were calibrated to achieve a close fit to
the measured data. A range for each calibrated parameter was obtained once the
model was validated.
2.4 WETMOD Scenario Analysis
In order to
test the suitability of the wetland model for decision support a scenario
analysis was applied to the highly degraded Lock 6 and Reedy Creek wetlands,
where data of the restored Pilby Creek wetland were used as a reference.
Scenarios for hypothetical restoration management included treatment of
agricultural drainage water for nutrient reduction, carp barriers and
drying-wetting cycles. Even though the simulated scenarios affected all state
variables in the wetland model, only simulation results for macrophytes,
phytoplankton, zooplankton and phosphate are presented in this paper.
Data in Table
1 summarise definitions of scenarios to be controlled by modified input data
according to habitat conditions at Pilby Creek. Turbidity was altered as an
indicator of carp activity, as WETMOD does not simulate carp population at this
stage.
Table 1.
Degrees of input changes controlling restoration scenarios for
two degraded wetlands.
|
|
Reedy
Creek |
Lock 6 |
Turbidity
|
- 25% |
- 20% |
|
Secchi
depth |
+ 25% |
+ 30% |
|
Phosphate
loadings |
- 19.3% |
¾ |
|
Nitrate
loadings |
- 19.3% |
¾ |
3. RESULTS
3.2
WETMOD Calibration Results
The wetland
model was firstly calibrated based on data collected from the Pilby Creek
wetland, used as a reference system, before it was applied and calibrated to
the remaining 4 wetlands. Of the 26 constant parameters implemented within the
wetland ecosystem model, 20 parameters were considered to be general and
remained constant during the calibration and validation process. Only 6
parameters were shown to be wetland specific, which were subject to calibration
for each wetland.
3.3
WETMOD Validation for Five Wetlands
WETMOD
predicted satisfactorily seasonal dynamics of phytoplankton and nutrients in
the open water for all four wetlands. On average the magnitudes were simulated
realistically over time, with observed trends in the decline and increase of
phytoplankton biomass and nutrient concentrations clearly simulated (Figure 2).
Predicted and measured trajectories met closely towards the end of the
simulation, which assumes good verification result in regards to seasonality.
3.3.1
Phosphate Concentration
Phosphate
concentrations were adequately simulated for each wetland by WETMOD. Timing and
magnitudes of peaks in phosphate concentration were slightly compromised in
most cases, with the timing delayed in simulations of the Paiwalla wetlands and
the magnitude of the large peak in mid-July underestimated for the Sunnyside
wetland (Figure 2a). Even though the phosphate peak as measured in the Reedy
Creek wetland in March was simulated correctly, the large peak in mid-July was
not predicted adequately. Simulation results for phosphate in the Lock 6 and
Pilby Creek wetlands were disappointing, with large overestimations in
September for the Lock 6 wetland. Also, the sharp decline in phosphate
concentration observed in the Pilby Creek wetland was not simulated adequately
by WETMOD, even though the declining trend in phosphate concentration was
simulated correctly (Figure 2a).
3.3.2
Phytoplankton Biomass
WETMOD predictions of phytoplankton
dynamics corresponded well with the measured data in most cases (Figure 2b),
where best results were achieved for the Lock 6 wetland. Simulations of algal
biomass for the Paiwalla and Sunnyside wetlands were slightly overestimated in
both timing and magnitude. The measured phytoplankton biomass for Reedy Creek
wetland was highly variable and a linear trend in seasonality was observed.
Generally, WETMOD was able to simulated these conditions reasonably well, with
the range in algal biomass magnitudes detected (Figure 2b). The measured data
for Pilby Creek phytoplankton biomass was more difficult to simulate. The algal
biomass trajectories for Pilby Creek wetland were closely predicted by WETMOD
between January and April, but were overestimated for the remaining time
period.
3.4
WETMOD Scenario Analysis Results
3.4.1 Scenario 1: Implementation of drying-wetting cycles
and carp exclusion to Lock 6 wetland
Improved water
quality as simulated according to Table 1 has not greatly affected the
phosphate concentrations, which is expected, however a sharp decline towards late
May was predicted (Figure 3a). This temporary decline corresponds well with the
increase in macrophyte biomass due to higher nutrient uptake and also a sharp
decrease in phytoplankton biomass, suggesting increased macrophyte competition
for nutrients.
(a) Phosphate (PO4-P)
Simulation
(b) Phytoplankton Biomass Simulation
Days
|
Days |
Sunnyside
Scenario 1:
Implementation of drying-wetting cycles and carp exclusion to Lock 6 wetland
(a)
Figure 3. Scenario analysis
using WETMOD for the restoration of Lock 6 (a) and Reedy Creek (b)
wetlands.
Degraded wetland simulations before implementation of management
scenarios; implementation of managed
scenarios outcomes.
Scenario 2: Treatment
of agricultural drainage and carp exclusion to Reedy Creek wetland (b)
Days
Zooplankton biomass
increased to abundances greater than under degraded conditions, which
strengthens grazing pressure on phytoplankton. Both increased abundance of
macrophytes and zooplankton indicate the potential recovery of biodiversity.
Water turbidity appeared to be the key driving variable for scenario 1 in order
to consecutively stimulate macrophyte growth, zooplankton abundance and inhibit
phytoplankton growth.
3.4.2 Scenario 2: Treatment of agricultural drainage and
carp exclusion to Reedy Creek wetland
Phosphate
concentrations in Reedy Creek wetland were lowered by approximately 50 %
(Figure 3b) with a reduction of nutrient loadings by 20% and an improvement of
light penetration as simulated in scenario 2 according to Table 1. The
reduction of turbidity considerably stimulated the growth of macrophytes and
consequently abundance of zooplankton in the open water. As a result of both
enhanced competition by plants and grazing by zooplankton, phytoplankton
biomass decreased to approximately half of its magnitudes observed under
degraded conditions (Figure 3b).
4. DISCUSSION
During this study a
generic wetland ecosystem model was developed considering four ecological
entities fundamental to wetland dynamics in the open water: macrophytes,
phytoplankton, zooplankton, and nutrients. Simulations of five qualitatively
different wetlands were performed. The model validation was satisfactorily
achieved, with measured state trajectories realistically predicted by WETMOD.
Many
ecosystem models are designed specifically to be used as decision making tools
for ecosystem management [e.g.
Hamilton and Schladow, 1997]. Such
models allow scenario analysis for testing management options and predicting
their effectiveness. The scenario analyses for Reedy Creek and Lock 6 wetlands,
have demonstrated that WETMOD can accordingly be utilised as a decision making
tool. As results of the scenario analyses have indicated, that external
nutrient loadings and water turbidity are key control variables to be explored
for the restoration of the two wetlands.
WETMOD will
be further developed towards a more complex wetland ecosystem model, simulating
additional processes relevant to the wetland dynamics such as interactions with
bottom sediments. In the future, WETMOD aims to become core of a landscape
model AQUALINK for the Lower River Murray. It is designed to cumulatively
assess restoration concepts for Lower Murray wetlands at landscape scale.
5. REFERENCES
Bartsch, D. L., Impact of Irrigation Drainage on Sunnyside Wetland: A Comparative Limnological Study, Honours thesis. University of Adelaide, Adelaide, 1997.
Bowles, B. A.,
I. J. Powling, and B. F. Burns, Effects on Water Quality of Artificial Aeration
and Desertification of Tarago Reservoir, Victoria. Murray-Darling Basin
Ministerial Council. pp. 104. State Pollution Control Commission, Sydney, 1987.
Boumans, R. M., F. Villa, R. Costanza, A. Voinov, H. Voinov and T. Maxwell, Non spatial calibrations of a General Unit Model for Ecosystem Simulations. Special Issue for Ecological Modelling, to cover the International Conferences on Applications of Machine Learning. Adelaide, Australia, Nov 27 to Dec 1, 2001.
Costanza, R., and S. Gottlieb, Modelling
Ecological and Economic Systems with STELLA: Part II, Ecological Modelling, 112, 81-84,
1998.
Costanza,
R., R. d'Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, S. Naeem, K.
Limburg, J. Paruelo, R. V. O'Neill, R. Raskin, P. Sutton and M. van der Belt,
The Value of the World's Ecosystem Services and Natural Capital, Nature, 387,
253-260, 1997.
Hamilton,
D.P. and G.S. Schladow, Prediction of water quality in lakes and reservoirs.
Part 1 - Model description, Ecological Modelling, 96, 91-110, 1997.
Marsh, F., A comparative study of the impacts of carp on phytoplankton and water quality in two Lower River Murray wetlands. Honours Thesis, University of Adelaide, Adelaide, 1997.
Middleton,
B., Wetland Restoration, Flood Pulsing and Disturbance Dynamics John Wiley
& Sons, Inc., New York, 1999.
Recknagel,
F. and J. Benndorf, Validation of Ecological Simulation Model
"SALMO", Hydrobiologia, 67, 113-125, 1982.
Walker, K. F. and T. J.
Hillman, Phosphorous and Nitrogen Loads in Waters Associated with the River
Murray near Albury-Wodonga, and their Effects on Phytoplankton Populations, Australian
Journal of Marine and Freshwater Research, 33, 223-243, 1982.