The Swan-Canning Estuary in 2050 Baseline Emily Barnes Pritam Patil Shaokun Song Yunhan Wang Chris Whitwell Sarah Rui Zang Extreme Pablo Islas Arrioja Marley Butler Harrison Davis Zhenheng Fu Anthony Joseph Qing Zhang Jiakang Zhao Management Holly Child Xiaolin HuHan JiangSarah McCulloch Yiru Shi Aneesh Toolsee Patrick Zhang UWA Environmental Engineering Design 5552 Class (2018)
150
Embed
The Swan-Canning Estuary in 2050 - oceans.uwa.edu.au · The Swan-Canning Estuary in 2050 Baseline Emily Barnes Pritam Patil Shaokun Song Yunhan Wang Chris Whitwell Sarah Rui Zang
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Swan-Canning Estuary in 2050
Baseline Emily Barnes Pritam Patil Shaokun Song Yunhan Wang Chris Whitwell Sarah Rui Zang Extreme Pablo Islas Arrioja Marley Butler Harrison Davis Zhenheng Fu Anthony Joseph Qing Zhang Jiakang Zhao Management Holly Child
Xiaolin Hu
Han Jiang Sarah McCulloch Yiru Shi Aneesh Toolsee Patrick Zhang
UWA Environmental Engineering Design 5552 Class (2018)
i
ACKNOWLEDGEMENTS
This research was partially supported by RiverLab, a collaboration between
Woodside and the University of Western Australia.
We would like to thank the following people for their insight and time which all greatly
assisted in research for this project:
Mark Cugley - Rivers Estuaries Division, DBCA
Catherine Thomson – Department of Water and Environmental Regulation
Anas Ghadouani - CRC for Water Sensitive Cities
Peta Kelsey - Department of Water and Environmental Regulation
Kerry Trayler - Rivers Estuaries Division, DBCA
Greg Ryan - CRC for Water Sensitive Cities
Joanne Woodbridge - Eastern Metropolitan Regional Council
Karl Henning - Department of Water and Environmental Regulation
In particular, we would also like to thank Dr Matt Hipsey and Dr Peishing Huang (UWA) for access to and for assistance with the Swan Canning Estuarine Response Model who, without their help, this project would not have been possible. As well as Dr Gregory Ivey (UWA) for his vital guidance and leadership throughout this project.
ii
Table of Contents
Acknowledgements .................................................................................................................................. i
List of Figures ......................................................................................................................................... iv
List of Tables ........................................................................................................................................ viii
Executive Summary ................................................................................................................................ ix
and increasing storm surge events, are also expected to affect the Swan-Canning
system (Swan River Trust Technical Advisory Panel 2007). There is expected to be
an increase in intensity and frequency of extreme storm event driven by climate
change (Climate Council of Australia 2017), though due to limited reliable predictions
and data for the south west region of WA, this was not considered further in this study.
MSLR is occurring at an increasing rate (Pattiaratchi & Eliot 2005). A research study
and literature review conducted in 2017 by a group of Master of Professional
Engineering students at the University of Western Australia, found that MSLR will
increase the likelihood of coastal flooding events (ENVE5502 2017), and that flooding
was likely to increase in magnitude and frequency. Furthermore, summer flooding is
likely to increase alongside changes in the upper reaches of the system’s erosion and
sediment regimes, further increasing flooding (Green et al. 2007). The
summer/autumn floods will bring organic matter, sediments and nutrients into the
system during a time of year with an already greater chance of low dissolved oxygen
events (Green et al. 2007). The specific impact of this will depend on where rain falls
and the health of the catchments (Green et al. 2007).
2.5.2.3 Increasing Air Temperature
Mean global surface temperatures have increased by 0.76° C since 1850 (IPCC 2007)
and it was reported that the average surface water temperature has increased by
0.76°C since 1850 in the Perth region (Green et al. 2007). The possible effects of
increasing temperature on the system are complex and difficult to separate from other
phenomena, for example Australian regional trends in evaporation are not simply
reflective of the rises in temperature but potentially could be more representative of
changes in cloud cover or wind (Green et al. 2007). The influence of temperature on
DO is just one example of the need to include climate change driven temperature
changes for predicting the water quality of the system to 2050. The solubility of oxygen
decreases with increasing temperature and the metabolic rates of aquatic plants will
increase as temperatures rise, altering the BOD. As such, temperature within the
hydrodynamic model is subject to surface heating and cooling processes (Hipsey et
al. 2014). It is a parameter with significant influence on biochemical processes,
influencing the water quality of the system and therefore it is necessary to adapt this
parameter within the meteorological forcing of the model (see section 3.1.1) to reflect
potential changes driven by climate change.
2.6 THE SWAN-CANNING ESTUARY RESPONSE MODEL
The complexity of the SCE necessitates the use of advanced modelling techniques in
order to make predictions about the water quality in the system as a response to
climate change. Predicting the water quality in the estuary using a model requires that
both a hydrodynamic and biogeochemical processes, and their interdependence, are
captured. These capabilities, which are vital to understanding the drivers of water
15
quality in highly complex non-linear environments, can be found in the Swan-Canning
Estuary Response Model (SCERM).
The SCERM was initially developed as a tool to support decision makers combat
challenges to the systems health by providing a holistic view of the estuarine response
to both short and long-term stressors (Hipsey et al. 2016). In the past, the model has
simulated the time period from 2008 to 2012, with the results validated against real
measurements taken in the estuary. In order to meet the aims of this particular study,
the existing SCERM infrastructure was adapted so as to allow for estuarine conditions
to be forecast out to 2050. Specifically, this was achieved by predicting the model
forcing in 2050 and creating a set of boundary conditions that reflect this.
16
3 METHODS
3.1 DESCRIPTION OF SCERM
The Swan-Canning Estuary Response Model (SCERM) is a model of the Swan-
Canning Estuary developed to assist in understanding the drivers of water quality in a
highly complex non-linear environment. SCERM consists of a hydrodynamic driver,
Tuflow-FV, coupled with a biogeochemical model, AED2. The model domain uses a
finite volume mesh that extends from Fremantle eastwards as far as the Kent Street
Weir along the Canning, and up to the Great Northern Highway along the Swan (Figure
6).
Figure 6: A map of the domain and its corresponding boundary conditions and validation sites (Hipsey et al. 2014)
17
Furthermore, Figure 6 shows the locations where the boundary conditions are applied
in the model domain and the sites used for model validation. Hipsey et al. (2016)
describes the validation of the SCERM for these sites over the period from 2008 to
2012.
The SCERM is highly influenced by forcing provided by boundary conditions that can
be divided into tributary inflows, tidal forcing and meteorological data. Furthermore,
the model forcing also includes anthropogenic inputs, namely oxygenation plants.
Largely, input data is in hourly timesteps which TUFLOW interpolates onto the model
calculation timesteps, typically a few seconds. The raw data and methodology that is
used for creating these boundary conditions is described by Hipsey et al. (2016).
Figure 7 A map of the final output locations for the model
3.1.1 Boundary Conditions and Anthropogenic Inputs
3.1.1.1 Tributary Inflows
The model domain contains 8 tributaries captured as boundary conditions in the
model. These include Bayswater Drain, Bennet Brook, Helena River, Ellen Brook,
Susannah Brook, Canning River, Upper Swan (Walyunga), and Jane Brook. The
boundary conditions for each tributary consist of a flow rate and a host of water
properties associated with the flow as described in Appendix D. The ephemerality of
tributaries is captured by turning on and off the flow rate in the boundary condition
input file.
3.1.1.2 Tidal Forcing
The tidal forcing is read into the model as a change in water level relative to mean sea
level, and corresponding water quality parameters, at the point where the domain
18
intersects the open ocean. Tuflow, the hydrodynamics driver, then calculates the
corresponding flow caused by the tidal file depending on the existing water surface
elevations in the model relative to tidal height. Furthermore, density differences
between the estuary and the ocean also drive salt water intrusion into the estuary.
Tidal time series data, as input into boundary condition files, were obtained from the
Bureau of Meteorology (BOM 2018), from the Fremantle tidal gauge, as described in
Hipsey et al. (2016).
3.1.1.3 Meteorological Forcing
Meteorological forces play a significant impact in driving the hydrodynamic and
biogeochemical forces in the model, and as such, the model also requires these forces
to be captured in the model. Specifically, the model has two input boundary condition
files concerned with meteorology. The first contains the daily rainfall volume, and the
second contains hourly readings (at 10m above ground) of air surface temperature,
relative humidity, wind velocity and solar radiation. Meteorological time series data, as
input into boundary condition files, were obtained from the Bureau of Meteorology
(BOM 2018), from the South Perth Meteorological Station, as described in Hipsey et
al. (2016).
3.1.1.4 Artificial Oxygenation
Within the model domain exist oxygenation plants that operate by removing water from
the estuary, adding oxygen into it and then pumping it back into the river. From a
modelling perspective, this process is essentially captured by an inflow of oxygen rich
water at the location of the plant, and thus, its format closely resembles the tributary
inflows. Given the plant does not treat the water it oxygenates, the oxygenated water
input by the oxygenation plant boundary conditions has the same water quality
properties, with the exception of oxygen, of the surrounding water.
3.2 AIMS AND OUTCOMES
The aim of this study is to forecast the response of the water quality in the Swan-
Canning Estuary (SCE) to climate change, so as to inform decision making for the
management of the Swan-Avon catchment and its associated waterways. As such,
nine modelling scenarios have been investigated, including simulations for modelling
the influences of both climate change and management strategies and are
summarised in Table 1.
Specifically, the aim for the Reference Simulations is to model the SCE in both
contemporary conditions as well as those in the proposed forecast date, 2050. These
simulations will then form the comparative basis for other forecasting events and
simulations to be benchmarked against. Simulation 1, the Characteristic Baseline,
aims to capture the historical norm of the system by simulating mean estuarine
characteristics or the parameters that influence the water quality of the system.
Alternately, Simulation 2, the 2050 Baseline, aims to predict the most likely condition
19
of the SCE in 2050 by making assumptions about the expected conditions that impact
the system.
Specifically, these general aims for the Characteristic Baseline Simulation require:
• The hydrological ‘norm’ of the Swan-Canning Estuary using a Flow Duration
Curve
• The nutrient loading ‘norm’ for the SCE using Nutrient Duration Curves
• Determining if other meteorological/catchment conditions should be
accounted for
• Reflecting these conditions in the model inputs
Likewise, the aims for the 2050 Baseline Simulation include:
• Predictions of changes to estuary inflow in 2050
• Predictions of Mean Sea Level Rise in 2050
• Predictions of air surface temperature in 2050
• Reflecting these conditions in the model inputs
The extreme runs aim to capture a range of conditions the Swan-Canning estuary may
encounter as conditions change towards the forecast date of 2050. The low and high
extreme runs aim to provide the upper and lower bounds of conditions relating to water
quality in the Swan-Canning River, and results are observed in combination with the
“2050 baseline” simulation. This aims to identify how each key variable (sea level,
temperature, flow volume and flow concentration) impact the state of the Swan-
Canning system as they become more extreme.
The team also develop two specific extreme events; a drought and a flood. The context
of such runs is explained in detail under section 3.4. The runs aim to create scenarios
that are not necessarily realistic, but instead provide useful information in terms of
observing the system responses under extreme scenarios.
The aim of management runs is to model effectiveness of various management
strategies applied to the Swan-Canning estuary in the year 2050. The current
oxygenation run aims to determine the effectiveness of artificial oxygenation in 2050
based on current management strategy, Guildford and Caversham oxygenation at the
Upper Swan. An enhanced oxygenation run, aims to identify location for placing an
enhanced oxygenation based on the current oxygenation and to capture effectiveness
of oxygenation plant by adding an enhanced oxygenation plant includes two current
oxygenation plants. Alternatively, the aim of a nutrient reduction is to predict the
reduction load of nitrogen and phosphorus, then to determine the effectiveness of
reducing the nutrient concentrations for the year 2050.
The key water quality indicators for the system, as identified in Section 2.4, can then
be analysed for each of these simulations. This is achieved by graphically displaying
the model outputs for multiple location across the model domain.
20
Simulation Name and
Description
Changes to 2008 Data
Simulation 1 Characteristic
Baseline an average
case scenario
• Nutrient Inflow concentrations scaled to
historical 50th percentile.
Simulation 2 2050 Baseline
2050 Most Likely
Conditions
• Nutrient Inflow concentrations scaled to
historical 50th percentile.
• Tributary Inflow decreased
• Mean Sea Level increased
• Air-surface temperature increased
Simulation 4 Low Extreme
An optimistic
scenario for 2050
• Nutrient inflow concentrations, inflow
volume, sea level rise and air surface
temperature scaled to historical 90th and
10th percentiles to represent favorable
conditions.
Simulation 5 High Extreme
A pessimistic
scenario for 2050
• Nutrient inflow concentrations, inflow
volume, sea level rise and air surface
temperature scaled to historical 90th and
10th percentiles to represent adverse
conditions.
Simulation 6 Summer Flood
Event
Unseasonable
inflow
• 2000 historical values for meteorological
data, tidal, inflow concentration and
volume
• 2050 mean (simulation 2) values for sea
level rise and surface air temperature.
Simulation 7 No Flow Inflow
Extended drought of
up to a year.
• 2050 mean (simulation 2) conditions
• No rainfall
• No inflow
Simulation 3 Current
Oxygenation
• Run Guildford and Caversham
oxygenation plant from January to
March 2050 at a base load of 30kg/hr of
O2 per hour each day.
Simulation 8 Enhanced
Oxygenation
• Current oxygenation (simulation 3)
conditions
• An enhance oxygenation added next to
the Nile Street monitoring site
Simulation 9 Nutrient Reduction • A reduction in nitrogen and phosphorus
loads into the system by 48%and 46%
respectively
Table 3: Simulations Overview
21
3.3 REFERENCE CONDITIONS
The aim of the first two simulations is to establish reference conditions, for both
contemporary and future catchment scenarios. The following sections outline the
motivations and methodology associated with developing each of the simulations.
Furthermore, Table 3 in the previous section, contains an overview of the reference
conditions simulations.
3.3.1 Catchment Conditions in 2008
Given the significant amounts of data required for the SCERM to accurately represent
the estuary, the decision was made to use a set of input data derived from real
measurements that represented normal estuary conditions. Specifically, the
hydrological conditions, in terms of total flow and temporal distribution, were the main
factors considered when representing normal estuary conditions. Furthermore, it was
decided that the year from which the characteristic baseline is formed should be
selected from the timeframe over which the SCERM had been validated, restricting
the possible choices from years in between 2008 and 2012 (Hipsey et al. 2014).
Figure 8: Flow Duration Curve for Walyunga monitoring station or “Upper Swan River”, as given in Figure 1. Total
recorded years 1976-2016. Data from the Department of Water (2018).
In order to determine which year of recorded data to use for our characteristic baseline,
a flow duration curve (FDC) was constructed for the Swan River. This curve ranks
years based on their annual flow volume, in an attempt to determine what is
considered to be a normal year, in this case one which had a 50% exceedance
22
probability. In layman’s terms, there would be an equal likelihood of any given year to
have more or less flow than the ‘normal’ year.
Figure 8 above shows a flow duration curve constructed from flow observations at the
Walyunga monitoring station, at the northern-extent of the model domain, for years
between 1976 and 2016. From Figure 7, it can be seen that the 2008 is the closest
year to the 50th percentile exceedance probability from those which have been
validated. It is worthwhile noting that Walyunga (also referred to as ‘Upper Swan’) was
chosen for the analysis given that it provides a dominant contribution to the estuaries
total inflow, as shown in Figure 9 below.
Furthermore, it was important to capture the typical temporal variability present in the
system. In the case of the Swan-Avon catchment, a normal year would contain dry
and warm summer months progressing to frontal rainfall events throughout the middle
of the year. Figure 10 below shows the hydrograph and hyetograph for the Swan-
Canning in 2008.
From this, it can be seen that the dry summer months lead to periods of low flow whilst
winter rainfall leads to increased streamflow, especially as catchment wetness
increases. Of particular interest is the first significant rainfall events that occur in April
provide an initial ‘flush’ of the estuary that may break stratification and decrease
nutrient concentrations.
Figure 9: Relative flow contributions to the SCE in 2008. Data from the Department of
Water (2018).
23
Given the large focus on nutrient loading in the SCE, understanding the distribution of
nutrient fluxes across each of the domain tributaries assists in defining which sub-
catchments and by extension, which tributaries, contribute to estuary nutrient loading
(Swan River Trust 2009).
Figure 11: Flow normalized average nutrient concentrations for each of the SCE tributaries in 2008. Flow and
nutrient concentration data from Department of Water (2018).
Figure 11 above shows the flow normalized nitrogen and phosphorous concentrations
for each of the SCE tributaries. From this, it can be seen that Ellenbrook contains the
highest inflow concentrations for both nitrogen and phosphorous, with the latter over
5 times the concentration of any other sub catchment.
Nitrogen concentration is considerably more uniform across the tributaries in 2008.
Nonetheless, the Ellenbrook inflow nitrogen concentration is still noticeably higher
(~1.5-2 times) than its counterparts.
Figure 10: Hydrograph (orange) and hyetograph (blue) for the Swan-Canning Estuary in 2008. Streamflow data from the Department of Water (2018) and rainfall data from DAFWA South Perth
Met. Station.
24
3.3.2 Simulation 1: The Characteristic Baseline
3.3.2.1 Motivation
The Characteristic Baseline Simulation is a vital component of this modelling study,
providing the comparative baseline for the further simulations. Analysis of possible and
current management strategies and the effect of climate change on the system
requires reference to some baseline, representing an average case contemporary
scenario for the main drivers of the system. The Characteristic Baseline aims to
provide a sense of perspective to forecasting simulations by contrasting them against
the SCE’s current state.
As discussed above, the year 2008 was selected due it is approximating a 50th
percentile of annual flow volume, whilst also showing a typical temporal flow
distribution. Nonetheless, although 2008 is a normal year hydrodynamically, it was not
necessarily normal in all conditions influencing water quality in the estuary, thus the
decision was made to alter 2008 from its historical conditions to represent a 50 th
percentile year across multiple factors. The result of adopting a characteristic baseline
as opposed to using a historical year is that forecasting simulation can then be
compared against a ‘historically-average’ baseline simulation as opposed to a single
year.
3.3.2.2 Nutrient Inflow
Nutrient concentrations (Nitrogen and Phosphorus) are considered a driver of water
quality of the system and historically have fluctuated significantly over recorded time.
The relationship between nutrient concentrations and inflows is not a novel concept
and has long been observed in the system (Henning & Kelsey 2015, Kelsey et al. 2011
& Jacowyna et al. 2000), and was further investigated in the preliminary work towards
establishing nutrient inflow values for a characteristic year. Using historical data
available did indeed confirm a strong relationship between inflow and nutrient
concentration for a number of the monitoring locations within the model domain. Given
this, as well as the clear limitations in taking an arbitrary nutrient quantity based on a
one year of average loads and dividing this load for all input locations across the model
domain, a duration curve technique was also applied for nutrient loads. An example of
such can be found in Figure 12.
25
Figure 12: Nutrient Concentration Duration Curve for Walyunga (Upper Swan). A FDC method was applied to
nutrient concentrations to create a normalised concentration by flow for each of the eight sites.
The analysis was conducted across each of the tributaries that included in the SCERM
as boundary conditions, as described in Section 3.1. The nutrient and flow data for
each tributary was sourced from Department of Water (DoW 2018) using the same
stations used to generate each of the tributary boundary conditions (Hipsey et al.
2016). Exceedance probability graphs for flow normalised concentrations were made
for each tributary so as to calculate the 50th percentile nutrient concentrations for each
inflow.
For each tributary, an annual mass flux for both nitrogen and phosphorous was
calculated and divided by the corresponding annual flow volume, resulting in a flow-
normalised nutrient concentration for each year. This was achieved by interpolating
fortnightly nutrient measurements onto daily timesteps, whereby they were multiplied
by daily flow volumes in order to obtain a daily mass flux, which are then summed over
the year. This process was repeated for each year with available data, and then
displayed graphically against their exceedance probability. Given these results, a
scaling factor was calculated for each tributary such that multiplying each
concentration reading for 2008 by the factor would change it to a 50th percentile. The
resultant nutrient duration curves for each of the estuary inflows can be found in
Appendix D, with Appendix E showing the 50th percentile scaling factors applied to
each of the boundary conditions.
26
Figure 13 shows the effect of scaling the flow normalized concentrations of nitrogen
and phosphorous from 2008 values to the 50th percentiles calculated as per the
above method. When comparing these results to Figure 7 (Section 3.2.1), Ellenbrook
is still the clearly the inflow with the highest concentrations of both nitrogen and
phosphorous. Furthermore, scaling the nitrogen concentrations to their 50th
percentile results in each of the remaining catchments showing almost uniform
nitrogen inflow concentrations.
3.3.2.3 Other Boundary Conditions and Anthropogenic Forces
Many of the boundary conditions were left unaltered from the actual conditions present
in 2008. Some model input parameters, such as mean sea level, were considered to
have much less annual temporal variability than flow and nutrients and as such, were
not changed.
Furthermore, given the interdependence of some boundary conditions (such as
meteorology and flowrate), many of the meteorological inputs could not be altered
without making complex and unreliable assumptions about the relationships between
interdependent parameters. For example, the interdependence of flow on rainfall is
not determined by a simple linear relationship (Hipsey et al. 2014, Kelsey et al. 2011,
Smith & Power 2014). Therefore, altering inflows from actual 2008 towards creating a
characteristic baseline year that reflect the contribution of an annual typical rainfall,
would not be feasible.
Although the Swan-Canning Estuary currently has 2 oxygenation plants that fall within
the model domain (as discussed in Section 3.1), neither of these plants were in
operation during 2008 and as such, are not present in the Characteristic Baseline
Simulation.
3.3.3 Simulation 2: 2050 Baseline (No Oxygenation)
3.3.3.1 Motivation
The 2050 Baseline Simulation is the created 2050 scenario to obtain a long-term
indication of the behaviour of system if no additional management strategies or
programs have been implemented from the Characteristic Baseline Simulation. In this
Figure 13: 50th Percentile flow normalized average nutrient concentrations for each of the SCE tributaries. Flow and nutrient concentration data from Department of Water (2018).
27
Simulation, a study was undertaken into how each of the models input parameters is
expected to change between the time of the baseline reference year (2008) and the
target year 2050 as a result of climate change. Specifically, the study investigated the
effects of mean sea level rise, tributary inflow, rainfall and air surface temperature.
Different methods of predicting conditions in 2050 were used for each parameter, as
discussed in the following sections.
Given that this Simulation aims to capture the ‘most likely’ scenario, the RCP 4.5
scenario was used to predict the effects of climate change on boundary conditions. As
many of the predictions for the model parameters have only been predicted on a global
scale, the creation of this Simulation often required the use of a simple regression
analysis to extract a trend from local data. This trend was then used to predict by
extrapolation the 2050 conditions, the results of which were then compared to the
global predictions.
Using the reference year (The Characteristic Baseline Simulation) as a template and
scaling the input parameters means that it is possible to assume that the yearlong time
series will be of the same form as the reference case where only the amplitude of each
variable may change. For example, the inflow resulted in a reduction in amplitude (see
the following section). Nonetheless, the non-linearity of the system means that the
model output may not necessarily directly reflect the change in the input boundary
conditions.
This Simulation does not include the two oxygenation plants (Guilford and Cavesham),
functioning within the bounds of the model as of 2018. To run a forecasting Simulation
for 2050 that does include the oxygenation plants, a baseline 2050 Simulation without
oxygenation would be necessary to determine within model time when the plants
would be theoretically in action. This would be to determine at what points in time the
conditions in the estuary are classified as anoxic, at the specific locations of the two
oxygenation plants. Simulation three will be ran with the two oxygenation plants turned
on as discussed in section 3.5.1
3.3.3.2 Estuary Inflow
To predict and model a most likely 2050 scenario, with expected patterns of climate
change included, it was necessary to obtain reasonable predictions for the declining
inflow into the Swan Estuary. Measured climate trends so far have indicated that the
non-linear relationship between streamflow and rainfall is known to have exaggerated
the impact of the patterns of declining inflow. For example, a 10-15% reduction in
rainfall during the period from 1975 to 2001 saw an approximate 50% decrease in
inflow into the reservoirs of Perth (Green et al. 2007). Due to the difficulties in obtaining
suitably statistically significant results for this non-linear relationship in order to make
predictions of the annual inflow for 2050, predictions for changing rainfall and potential
evaporation were not considered.
Though this study intended to focus on an average climate change scenario, namely
the RCP 4.5 pathway, the results found by Smith & Power (2014), as discussed in
28
section 2.2.3 were used as guidance for predicting inflows in the year 2050. The
resultant trend for predicting the reduction in inflows for southwest WA by Smith &
Power (2014) was obtained from a baseline range of years from 1911 to 2013.The
level of uncertainty in predicting these climate change trends should not be
overlooked. The majority of studies to date on the region have not incorporated the
effects of future warming as well as the increase in potential evaporation and the
response on streamflow. This is except for a study on the Stirling catchment by Berti
et al. (2004), which showed through sensitivity analysis, that when a 10% decline in
rainfall is coupled with an 11% increase in potential evaporation, a 40% decline in
streamflow is expected. Another study (Marillier et al. 2015) indicated that as the south
west region of WA has been reported as a region of the globe relatively susceptible to
climate change. As of such, this study found that the region showed a greater decline
in rainfall than interior and southern coastline regions (Marillier et al. 2015).
Henceforth, with all factors considered, assumptions will be held for the use of the
trend from Smith & Power for all consistently in all Simulations.
Altering flow in the boundary files from the baseline year, to then model a 2050 year,
required a scaling factor for each time step. This was done for each of the eight sites
as historically these eight sites consist of the majority of significant estuary inflow
within the model domain collectively, see section 3.3.1.
Baseline data set for this study only included the years ranging from 1976-2016. The
predicted mean trend (Figure 6, section 2.5.2) from Smith & Power (2014) was utilised
to predict from the 2008 annual inflow (GL) to that of 2050. The trend from Smith &
Power gave a 39% decrease in inflows from the 2008 to 2050. Though this modelling
was done under a high-end emission scenario, this predictive decline still fits within
that expected by global predictions.
Predictions obtained from Smith & Power (2014) trends are supported by global and
localised predictions from other such studies. For example, Green et al. (2007)
predicted a 22-55% decrease in streamflow to 2030, and a 45-75% decrease to 2070,
from a baseline ranging from 1925-1975. Therefore, the 39% decrease in inflows was
converted to individual scaling factors for each of the sites to alter the input boundary
condition files for the 2050 Baseline Simulation.
3.3.3.3 Mean Sea Level
The need to incorporate the influence of MSLR to a 2050 predicted Simulation is clear
when considering the strong tidal influence on the system. The rate of global MSLR
has experienced substantial growth from 0 mm/year to 0.013 mm/year, since the early
20th century (Church et al. 2013). For locations such as Perth this seemingly small
increase in height could translate to large increases in volume for the SCE. Particularly
for coastal locations, a local change of sea level over a relatively short time span could
be affected by tidal influences, storms and climatic variability. When scaling for longer
term patterns, climate change is considered a contributing factor (Church et al. 2013).
Thus, it was deemed necessary to confirm these global trends were applicable on a
29
localised scale. The International Panel on Climate Change (IPCC) predicted the
global mean sea level rising for Fremantle as one of the representative coastal
locations for the 5th Assessment Report (ICCP 2018), see Figure 14.
The trend from the 5th Assessment Report was used to provide an offset value to be
applied to the tidal time series data, as input into boundary condition files, to represent
the mean sea level rise from the baseline year (2008) to 2050. The trend for 2050 sea
level rise was selected from the RCP 4.5 prediction shown as the light blue line in
Figure 13, considering 1986-2005 as a baseline. For the mean sea level tidal data as
input to the model for this study, 2008 was chosen as the baseline year. Thus, the
value for the 2008 mean sea level rise, as predicted by modelling of the RCP 4.5 by
Church et al. (2015), was used as a baseline. The predicted 2050 value for mean sea
level under this scenario was also taken from this source and the increasing trend
based between these two values was then be applied to the 2008.
From Figure 13, a mean sea level of 0.04m was found for 2008 and of 0.23m for 2050
respectability. Thus, from the ICCP predictions for Fremantle, a MSLR between 2008
and the year 2050 can be estimated at a total of 0.19m. This value was used to
positively offset each input surface water elevation in the tidal boundary condition, to
then represent the change in mean sea level from the base case to the predicted most
likely climate change scenario in 2050.
Figure 14: An extract from the IPCC 5th Assessment Report of the observed and projected change in mean sea level for Fremantle, as one of the representative coastal locations (Church et al. 2015). The observed in situ relative mean sea level records from tide gauges (since 1970) are plotted in yellow, and the satellite record (since 1993) is provided as purple lines. The projected range from 21 CMIP5 RCP 4.5 scenario runs (90% uncertainty) is shown by the shaded region for the period 2006-2100, with the bold line showing the ensemble mean. Vertical bar at the right sides of each panel represent the ensemble mean and ensemble spread (5 to 95%) of the likely (medium confidence) sea level change at each respective location at the year 2100 inferred from RCP2.6 (dark blue), 4.5 (light blue), 6.0 (yellow) and 8.5 (red)
30
3.3.3.4 Air Temperature
The effect of climate change on air temperature can have substantial effects on the
water quality within the estuary, see Section 2.5.2. Specifically, the diffusion of oxygen
through the water surface is dependent on, among other things, the air surface
temperature. Average annual air temperatures for Perth have increased by
approximately 0.6°C in the last century to 1990 (Green et al. 2007). Though this trend
matched global predictions, it cannot be considered consistent in a prediction out to
2050 due to the influence of local effects (Green et al. 2007). Therefore, for this study
a simple linear regression analysis of data available was conducted to predict air
temperature in 2050. Data was taken from the longest recorded data set, Perth Airport
which maintained a daily temperature record from 1944 to current day. A 3-point
moving average was applied to the daily time step data, to replicate the average
duration for storm and heat wave events in the region, the results of which can be
seen in Figure 15.
Figure 15: The result of a 3-point moving average of daily average temperature and the resulting regression line. Data
taken from the Perth Airport Station (BOM 2018), the time given on the x axis is the number of days from the start of the
record to present day (1944-2018).
From this simple analysis and the trend found in Figure 15, daily average temperature
values predicted by the trend for 2008 were compared to those predicted for 2050. An
annual average for each was found and an offset representing the rise in average
annual temperature from 2008 to 2050 was established. Based on this method, the
projected warming in 2050 from 2008 was 1.04°C.
y = 7E-05x + 23.57R² = 0.0079
0
5
10
15
20
25
30
35
40
45
50
0 5000 10000 15000 20000 25000 30000
Tem
pe
ratu
re ℃
Days
3 Point Moving Average Temperature Time Series
3days moving average
regression line
31
To reinforce that this predicted rise was reasonable, other localised predictions were
assessed for comparison purposed. For a local prediction, we compared with the
results of Marillier et al. (2015). This prediction was made relative to a 1961-1990
baseline, which gave an estimated increase in temperatures of 1.0-1.3°C to 2050,
varying within different potential future climate regimes (wet, medium or a dry
scenario).
The result from the regression analysis were also compared to those found by Green
et al. (2007) who predicted air temperature risings for 2030 and 2070 based on a 1990
baseline year. The result of which was a projected mean warming in 2030 of 0.8 °C
and 1.4°C in 2070. Therefore, the predicted rise in air temperatures used as an offset
value for the 2050 Baseline Simulation was considered within range as to be expected
based on the collection of current work into the localised effect of climate change on
the system.
3.3.3.5 Nutrient Inflow
The factors that influence nutrient concentrations in the system are complex, with
rainfall and catchment land use being considered dominant. A strong correlation has
been shown between catchment rainfall and total nutrient load into the system,
whereby an increase in rainfall in the catchment mobilises more nutrients stored in the
catchment itself (Thompson 2017). Furthermore, changes to catchment land use
(typically driven by urbanisation, and by extension, population) are also expected to
affect the amount of nutrients entering the system. Ideally, the influence of these
factors would be considered when trying to predict nutrient inflows in a future scenario.
An analysis was conducted during the early stages of this study to consider the
possible influence of increasing population and changing land use in the catchment.
The SCERM does not contain any elements of catchment modelling, though work has
been done in the past of this nature (Kelsey et al. 2010). It was considered that the
possible extent of urban development could be partly factored in when considering the
effect of increasing population, as the dominating influence of this within the model
domain would be the increasing urbanisation within the coastal catchment region.
Ultimately, there was no significant relationship found between increasing population
and nutrient loads. Using predictions of the likely population increase for Perth, as
given by the Australian Bureau of Statistics a simple comparison showed the trend of
increasing population was of significantly smaller magnitudes to the trends of
decreasing flow. The influence of population on nutrient inflow was considered
negligible given the clearly dominant influence of flow. Furthermore, without
conducting extensive catchment modelling, attempts at quantifying the effect of land
use changes on the systems nutrient budget were unreliable. Given the complex non-
linear relationship between catchment land use, rainfall and nutrient influxes and the
use of catchment modelling outside the scope of this study, the nutrient concentrations
were not changed in the 2050 scenario.
32
3.4 EXTREME SIMULATIONS
3.4.1 The Runs from the Extreme conditions group
In previous sections the ‘mean’, or ‘most likely’ conditions of the Swan-Canning River
Estuary were discussed. From a design and management perspective, it is also
pertinent to determine the extreme extent of conditions that are likely to be
experienced, as well as some particularly poor but probable scenarios. In order to
predict the impacts of these extreme and poor conditions as well as the implications
of the changing climate, the study group decided on simulating four ‘Extreme’ condition
scenarios set in the year 2050. These are described in detail below.
3.4.1.1 2050 Low Extreme
The 2050 low extreme run aims to characterise an optimistic scenario by representing
a more favourable sector of the spectrum of environmental trends. The 2050 low
extreme run will aim to describe a “good year” in terms of certain conditional variables
relating to water quality in the Swan-Canning system in 2050. When making decisions
about each conditional variable the team applied the philosophy of the 2050 low
extreme run as “an optimistic approach to climate forecasting and annual variability,
resulting in a year that is unlikely but entirely possible”.
In order to create such a scenario, the team applied two main predictive tools to each
key variable. The first was to develop a local trend for each variable using real data.
Various sources were used and are explained in detail in the description of the
methodology of each key variable (section 3.3.0). The low and high extreme runs can
be thought of as a ‘one off’ forecast that predicts a trend beyond the timeframe of
available observed data. When applying regression analysis or trends to data we
adopted the same data sources and approach that was implemented to create the
“mean” 2050 condition (section 3.2.2). Following this trend prediction, statistical
analysis was used to determine an upper and lower bound based on the variation of
such variables. For each conditional variable with an adverse effect on the system the
team took the value from the lower quartile for the low extreme run, and the higher
quartile for high extreme run. The opposite quartiles were taken for beneficial
variables.
3.4.1.2 2050 High Extreme
The philosophy behind the 2050 high extreme run is the opposite of the 2050 low
extreme run. The 2050 high extreme run will aim to describe a “bad year” in terms of
certain conditional variables relating to water quality in the Swan-Canning system in
2050. It applies the philosophy of “a pessimistic approach to climate forecasting and
annual variability, resulting in a year that is unlikely but entirely possible”. The
approach reflects that of the 2050 low extreme run, using a regression and then
applying upper and lower bounds based on annual variation of data. However, in the
case of the 2050 high extreme run, the upper quartile of adverse variables are
considered.
33
The combination of the low extreme and high extreme runs aim to allow comparison
between the results of the “2050 low extreme, mean and 2050 high extreme” runs, in
a sense, creating a full range of possible scenarios for comparison.
3.4.1.3 2050 Summer Flood Event
The summer flood run aims to reproduce conditions that in the past caused major algal
blooms in the Swan-Canning system, but under 2050 scenarios. Two unseasonably
large rainfall events in January 2000 resulted in an estimated 270GL of freshwater
Year Mean Lower confidence bound Upper confidence bound
2008 0.79 0.70 0.88
2050 0.96 0.87 1.06
Offset 0.18 0.09 0.27
37
• Lastly, the International Panel on Climate Change compiled a special report on
climate emission scenarios (SRES) which predicted a global temperature
change by 2050 of 0.8-2.6 oC.
A consequence of long term forecasting (40+ years) is the amplification of uncertainty,
which can lead to a somewhat large range of possible values. To take this into account
in our simulations, the group has decided to use two ‘extreme’ scenarios to
supplement the mean scenario of a 1.04-degree increase. These two extreme
scenarios will cover what happens when the mean surface air temperature for 2050 is
in the upper bound of the baseline 2050 prediction, and another when the temperature
is at the lower bound of the predicted baseline 2050 temperature.
To do this, it was decided that we would take the previously predicted baseline mean
temperature for 2050 and we would alter this by selecting a mean that was 1.281
standard deviations above and below the baseline mean. This 1.281 standard
deviations provided us with a confidence interval of 80% which means 10% probability
of exceedance for the upper region and 90% probability of exceedance for the lower
region. Put simply, there is a 90% chance that the lower extreme case for mean
surface air temperature will occur in 2050, and there is a 10% chance that the high
extreme case will occur.
In order to reach a representative result data from the Perth airport site was analysed
due to its long running temperature measurements. An annual mean regression
analysis was then produced for this data. Using only the historical data (not the
predicted part of the regression) the team found the standard deviation of the annual
mean values from the previously calculated regression line, giving the standard
deviation from a moving mean. Using this value for standard deviation, we took the
2050 predicted mean temperature and added (or subtracted depending on which the
scenario) 1.281 times the standard deviation from the regression line. This gives us
two new values for the offset to be used as inputs for the model, the upper value which
in theory has a 10% chance of occurring and the lower value which in theory has a
90% chance of occurring.
Table 6: summary of results from temperature regression and variability analysis. The final predicted
mean, lower and upper bound off set values are shown in red.
3.4.2.3 Flow Volume
In order to predict the flow volumes for the 2050 high extreme and 2050 low extreme
scenarios we needed to quantify the change in flow conditions due to climate change
as well as specify what exactly a ‘best’ and ‘worst’ scenario would entail.
Percentage z z*SD
Maximum expected
Temperature 2050
Offset from (forecast)
2008 value
90 1.281 1.009172 27.1 2.0
50 0 0 26.0 1.0
10 -1.281 -1.00917 25.00 0.0
38
Following from 3.2.2, we use the same regression relationship found from Smith and
Power (2014). However, instead of using the mean annual flow in 2008 as a baseline
we adjusted this starting value to represent either a ‘2050 high extreme’ and ‘2050 low
extreme’ scenario. Upon analysis of the Water Corporation (2016) data on observed
inflow to Perth dams, a strong decline in the magnitude and variance of inflow was
found (Figure 18).
Figure 18: Historical inflow to Perth Dams (Water Corporation (2016).
In consideration of the strong trend, data from post 1975 was used to determine the
upper and lower bounds of the 2050 predicted inflow. By determining the recent (post
1975) standard deviation of inflow into Perth damns from (Smith & Power, 2014),
probability of exceedance plots used to determine the upper and lower bounds of the
10% and 90% exceedance values.
The upper value (10% probability of exceedance) is used as a starting point from which
the regression will continue in the ‘2050 low extreme’ scenario, likewise the lower
value (90% probability of exceedance) is used as a baseline for the ‘2050 high
extreme’ scenario. Applying such method means that we can simulate an extreme
year, but one that is still likely to occur and therefore would bear consideration in the
future. Each case holds a 10% probability of reaching an annual flow of that value, or
more extreme. Over the 42 years from the baseline year up until 2050 the probability
of annual flow reaching or exceeding either scenario’s critical value at least once is
98.8%, based on recent data. The results of the analysis are shown in Table 7 below.
39
Percentage z z*SD
Flow Volume at Walunga
(2050)
Scaling factor from 2008
(181 GL)
90 1.281 15.7 126.1 0.70
50 0 0 110.4 0.61
10
-
1.281 -15.7 94.7 0.52
Table 7: summary of results from flow volume regression and variability analysis. The final predicted
mean, lower and upper bound scaling values are shown in red.
In order to reach our final offset, amount the regression gradient is used to forecast
forward from each scenario specific starting point to 2050. This 2050 value is
compared to the mean 2008 value to reach a final scaling factor.
3.4.2.4 Nutrient Concentration
As described in section 3.2.2, many factors influencing potential trends in nutrient
concentration were considered. Anthropogenic factors such as population rise or
changes in land use were key considerations. Nutrient concentration in runoff form
catchments has also been shown to proportional to rainfall and flow volume (P Kelsey,
2010). Analysis of the influence of such variables on concentration data form the
catchments confirmed that the major influence was rainfall and inflow, and so the team
decided to neglect the influence of population rise and catchment use when setting
2050 high extreme and 2050 low extreme bounds for nutrient concentration. Instead,
the team used normalised concentrations by flow to determine bounds for 2050 high
extreme and 2050 low extreme.
The method used follows from section 3.2.2, where final selection of Nitrogen and
Phosphorous concentrations were based on the normalized concentrations by flow for
all years of the data. Nutrient data of each site location was sourced from the
Department of Water (2018).
We then applied 10% and 90% exceedance probabilities to the normalized data to
give upper and lower bounds for N and P at each site. Each site location/tributary was
scaled to the 10th and 90th percentile for both nitrogen and phosphorous by diving the
percentile by the 2008 concentration at each tributary. The result is a scaling factor for
10th and 90th percentiles that we apply to all nutrient concentrations in the 2008 model
year to achieve a final nutrient concentration with 10% and 90% probability of
exceedance. The normalized concentration curves can be found in the appendices, a
table summarising the nutrient concentration percentile for 2050 high extreme and
2050 low extreme in 2050 are shown below. Note the units in the following tables are
scaling factors, meaning the ratio of concentration of the 10th or 90th percentile over
the 2008 concentration of nitrogen and phosphorus. Mathematically this is represented
as 𝐶10𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑑𝑛𝑡𝑖𝑙𝑒
𝐶2008 𝑣𝑎𝑙𝑢𝑒 and
𝐶90𝑡ℎ 𝑝𝑒𝑟𝑐𝑒𝑑𝑛𝑡𝑖𝑙𝑒
𝐶2008 𝑣𝑎𝑙𝑢𝑒.
40
10%
exceedance
Bayswater
Drain
Bennet
Brook
Canning
River
Ellen
Brook
Helena
River
Jane
Brook
Susannah
Brook
Upper
Swan
N 0.852 0.572 0.584 0.692 0.573 0.349 0.423 0.445
P 0.81 0.783 0.763 0.526 0.634 0.356 0.690 0.404
90%
exceedance
Bayswater
Drain
Bennet
Brook
Canning
River
Ellen
Brook
Helena
River
Jane
Brook
Susannah
Brook
Upper
Swan
N 1.189 0.833 0.969 1.036 0.987 0.896 1.307 1.07
P 2.361 1.286 1.254 1.235 2.154 1.400 2.075 1.357
Tables 8: Scaling factors for both nitrogen and phosphorous applied to the 2008 concentration
data to achieve 10% and 90% probability of exceedance at all sites.
3.4.3 Key Parameters
3.4.3.1 2050 Low Extreme (sim 4)
Boundary Conditions:
Air Surface Temperature (OFFSET) -0.0
Sea Level Rise (OFFSET) +0.09m
Flow rate (SCALAR) Scaling factor of 0.697 to all inflow
Nutrient load (SCALAR) 90% scaling factor specific to each drain
3.4.3.2 2050 High Extreme (sim 5)
Boundary Conditions:
Air Surface Temperature (OFFSET) +2.01
Sea Level Rise (OFFSET) +0.27
Flow rate (SCALAR) Scaling factor of 0.523 to all inflow
Nutrient load (SCALAR) 10% scaling factor specific to each drain
3.4.3.3 2050 Summer Flood Event (sim 6)
Boundary Conditions:
Air Surface Temperature (OFFSET) +1.04
Sea Level Rise (OFFSET) +0.19
Flow rate Superimpose Dec 1st to March 31st 2000
Nutrient load Superimpose Dec 1st to March 31st 2000
41
3.4.3.4 “Mean Year 2050” No Inflow (sim 7)
Boundary Conditions:
Air Surface Temperature (OFFSET) +1.04
Sea Level Rise (OFFSET) +0.19
Flow rate (SCALAR) 0
Nutrient load (SCALAR) 0 0
3.5 MANAGEMENT SIMULATIONS
The methodologies for the management scenarios focus upon the inputs of the
SCERM model. By altering minimal inputs of the SCERM model, from the baseline
2050 simulation, analysis of these changes and identification of key drivers of change,
thus ensuring an effective and conclusive evaluation of the outputs.
3.5.1 Current Oxygenation Strategy
3.5.1.1 Motivation
The current oxygenation simulation is to demonstrate how the current oxygenation
management strategy will impart the Swan-Canning Estuary in the year 2050. This
simulation uses simulation 2: 2050 baseline as the reference simulation for modelling,
by adding the two implemented artificial oxygenation plants (Guildford and
Caversham) back to the model.
A study was undertaken to determine oxygen load and running hour of the current
oxygenation plant. The oxygen levels in the bottom layer of the Swan-Canning in
Upper Swan will fall below the bench line 4mg/L during the seventeen-day time period
between the 21st of January to 7th of February 2050 by predicting from simulation 2.
The decision has therefore been made to run the two artificial oxygenation plants at a
base oxygen load of 30kg per hour for all 24 hours in a day from 21st of January to 7th
of February 2050 to interpret the effect that these plants are likely to have on the
oxygen levels in the year 2050.
3.5.1.2 Method
Maintaining elevated levels of dissolved oxygen (DO) within estuarine environments
is essential to sustaining the health of the aquatic organisms within the ecosystem
(Bailey & Ahmadi 2014). An artificial oxygenation strategy was developed by the
Department of Parks and Wildlife in collaboration with Department of Water and the
Swan River Trust in response to increasing hypoxic and anoxic conditions within the
Swan-Canning Estuary caused by a number of environmental factors (DBCA
2015). At present, there are five oxygenation facilities located within in the Swan-
Canning Estuary that each cycle water from the Swan-Canning through the system
42
and supersaturate the estuarine water with dissolved oxygen to enhance oxygenation
in the system. This supersaturated oxygenated estuarine water is then released back
into the system via instruments that sit at the bottom of the Estuary to diffuse the
oxygen into the system effectively (DoW 2015). It is understood that hypoxia develops
when vertical stratification and warm water occur simultaneously and as temperatures
are expected to rise in the year 2050 we expect to see significantly higher levels of
hypoxia in the system (Stanley & Nixon 1992). The aim of the oxygenation plants is to
enhance the oxygen content throughout the stratified layers from the top layer to the
bottom layer (i.e. throughout the water column) in order to reduce the stressed areas.
These oxygenation facilities are located in areas within the Swan-Canning that have
been identified by the Department of Parks and Wildlife to be impacted by significant
levels of anoxia, there are two oxygenation plants located in the Upper Swan, in the
suburbs of Caversham and Guilford and there are three oxygenation plants located in
the Canning River, the Bacon, Camsell and Nicholson oxygenation facilities (Figure
19). The oxygenation management strategy began in 1998 in the Canning River
system where two artificial oxygenation plants were installed to combat the anoxic
conditions in the system. The systems were monitored to determine their effectiveness
in the Estuary and it was found that oxygen levels were enhanced within a 5km radius
of the plants (DBCA 2015). The strategy was extended to the Upper Swan Estuary
where another two artificial oxygenation plants were installed in 2008 to combat the
anoxic conditions in the Upper Swan caused by the high nutrient load and poor
flushing abilities that are present in the upper reaches of the Estuary. These systems
are understood to increase the oxygen levels in the Estuary within a 10km impact zone
of the plants, 10km upstream and downstream of these facilities. A third oxygenation
plant was installed upstream from the Kent Street Weir in the Canning River in 2014
as there was further need for oxygenation as anoxic conditions were becoming more
common (Hipsey et. al. 2014). Sampling and previous model simulations have shown
that oxygenation plants in the Upper Swan improve oxygen conditions in 39-92% of
the 10km target zones (Department of Water, 2015).
43
Figure 19. The locations of the Guilford, Caversham, Bacon, Camsell and Nicholson artificial oxygenation
facilities in the Swan-Canning Estuary (DWER 2015).
The purpose of this model run is to demonstrate how the current oxygenation
management strategy will impact the Swan-Canning Estuary in the year 2050. To
allow for an effective interpretation of the impact that is had on the system by these
oxygenation facilities it is assumed that the only management strategy undertaken for
the Swan-Canning in 2050 is the oxygenation strategy that has been in place since
2014. The model does not extend past the man-made barrier at Kent street weir and
therefore the area of the canning river where the oxygenation plants are located is not
included in the model. Only the two oxygenation plants that are located in the Upper
Swan are in included in this analysis.
The time period input concentrations of these artificial oxygenation facilities where
used by Hipsey et al (2014) when determining the benefit of the facilities in 2014. This
time series concentration data was added to the baseline 2050 model inputs
(simulation two). The inputs of these artificial oxygenation facilities were deciphered
for the purpose of a study prepared by the Swan River Trust that modelled in the
oxygen dynamics of the Upper Swan Estuary and Canning Pool to create an optimal
oxygenation strategy for the system (Hipsey et. al. 2014). The oxygenation plant input
rates were calculated based on oxygenation consumption data logs that are monitored
as part of the oxygenation management program run by the Western Australian
Government. The specifications of each plant flow rate were incorporated into the
model inputs, and it was configured so that the supersaturated oxygenated flow would
44
be released over the bottom 1m of the centre cell across the appropriate locations in
the estuary (Hipsey et. al. 2014). This approach was validated using oxygen and
salinity data profiles of the area for the dates at which they were simulated (2008 and
2010) and was found to be an effective with a mean absolute error of 25% for predicted
oxygen values over the simulated years. We can therefore be confident of our results
for oxygenation in the Swan-Canning Estuary for 2050.
The artificial oxygenation facilities located in the Upper Swan Estuary each release a
base load of 30kg/hr of O2 and a heightened load of 60kg/hr (Hipsey et. al. 2014). The
facilities begin releasing oxygen into the system when oxygen levels fall below 4mg/L,
as levels of poor oxygen conditions within the Swan-Canning Estuary can be defined
as any value less than 4mg/L (
Table 9)
Table 9. Oxygenation classifications for the Swan-Canning Estuary (Hipsey et. al 2014).
The model does not allow for these artificial oxygenation facilities to be systematically
turned on and off when the oxygen levels in the system fall below 4mg/L. It is for this
reason that the results of simulation two were inspected to decipher an appropriate
time to “turn on” the artificial oxygenation facilities in the Upper Swan Estuary. Due to
time constraints it was only feasible to model the first three months of year 2050 to
allow for analysis. Figure 20 shows the results from Simulation Two at Middle Swan
Bridge. Using Figure 20 we have predicted that the oxygen levels in the bottom layer
of the Swan-Canning in this area will fall below the bench line 4mg/L during the
seventeen-day time period between the 21st of January to the 7th of February 2050.
The decision has therefore been made to “turn on” the artificial oxygenation facilities
during this period (with an estimated base loaf of 30kg/hr) to interpret the effect that
Classification Concentration
Low Oxygen < 4mg/L
Hypoxia < 2mg/L
Anoxia 0 mg/L
45
these facilities are likely to have on the oxygen levels in this area in the year 2050. As
can also be seen in Figure 20 there seems to be frequent occasions where the bottom
oxygen levels fall below the benchmark 4mg/L, it was decided that the artificial
oxygenation facilities would not be “turned on” during this time to allow for defined
analysis. Through our analysis we were able to determine the effectiveness that the
artificial oxygenation facilities will have over time by isolating a time of activation and
analysing the residual results from the input. The facilities were “turn on” permanently
and maintained a constant inflow during this period to combat the low oxygen levels
in the system at the time.
Figure 20 Oxygen levels at the Middle Swan Bridge from January until March. As can be seen from the figure,
levels of oxygen fall below 4mg/L between January 21st and February 7th.
3.5.2 Enhanced Oxygenation Strategy
3.5.2.1 Motivation
Oxygen levels in areas of the Upper Swan are still likely to fall below 4mg/L in 2050
regardless of the two artificial oxygenation plants in the Upper-Swan Estuary from the
results obtained for simulation 3. An enhanced oxygenation simulation was decided to
run for assessing the benefit that an enhanced oxygenation strategy is likely to have
on the Swan-Canning Estuary in 2050.
One additional artificial oxygenation plant was decided to add next to the Nile Street
monitoring site. All initial conditions and operation conditions of this additional
46
artificial oxygenation plant were consistent with the conditions for simulation 3, this
will allow for an accurate interpretation of the impact that an additional artificial
oxygenation plant is likely to have on the oxygen levels within the Swan-Canning
Estuary for the year 2050
3.5.2.2 Method
The results from simulation three, section 4.3.1, were analysed and it was concluded
that regardless of the two artificial oxygenation systems installed the Upper Swan
estuary oxygen levels in the areas of the Upper Swan are still likely to fall below 4mg/L
in 2050, Figure 21. It was decided that a simulation will be run of an enhanced
oxygenation strategy, allowing for an assessment of the benefit that an enhanced
oxygenation strategy is likely to have on the Swan-Canning Estuary in 2050.
Figure 21 Dissolved oxygen levels at Nile street
The oxygen outputs of simulation three were analysed to determine the most
appropriate location of the additional oxygenation plant(s). It is understood that high
salinity and high levels of stratification in the Lower Swan Area (from Blackwall reach
to Heathcote) will negatively impact the ability of the oxygen to mix throughout the
water column (Hipsey at. al. 2014). It is also understood that the inundation of
terrestrial vegetation and sediments into the system is increased in the Upper-Swan
Estuary, the decay of this vegetation and introduction of sedimentation is likely to lead
to a dramatic decrease in dissolved oxygen (Kneis, Forster & Bronstert 2009). It is for
47
these reasons that the decision was made to keep the artificial oxygenation strategy
within the Upper-Swan area and as these artificial oxygenation systems are estimated
to have an impact within the 10km impact zone of the system only one additional
artificial oxygenation system was added. It is important to note however, that the
impact zone is highly dependent on the location and seasonal changes in the system
and this will likely affect the impact of the new oxygen plant (Hipsey et al, 2014). This
oxygenation facility was “added” just next to the Nile Street monitoring site to ensure
that the result analysis will present the maximum impact on the area.
Figure 22 Location of the artificial oxygenation systems, including the simulated “additional oxygenation system”,
in the Swan-Canning Estuary along with the location of the seven study areas.
All initial conditions were consistent with the initial conditions for simulation three,
aside from the addition of the single artificial oxygenation system at Nile Street. The
inputs for the artificial oxygenation system are the same for the oxygen plant flow rate
as quantified by Hipsey et. al. (2014) and where combined with the estuarine water
specifications at Nile Street to simulate the addition of the third oxygenation facility. It
is assumed, as it was for simulation three, that the oxygenation facility is producing a
base load of 30kg/hr of dissolved oxygen and it will also be run for the same duration,
24 hours a day from January 21st until February 7th, 2050. The simplistic approach
will allow for an accurate interpretation of the impact that an additional artificial
oxygenation facility is likely to have on the oxygen levels within the Swan-Canning
Estuary for the year 2050.
48
3.5.3 Nutrient Reduction
3.5.3.1 Motivation
In this simulation, a study was undertaken on how each of the models input parameters
expected to change between the time of baseline reference year 2008 and the target
year 2050 as a result of nutrient change. Specifically, the study investigates the effects
of changes in Nitrogen and Phosphorus levels. It was planned to reduce Nitrogen and
Phosphorus levels by 48% and 46% respectively. This reduction target is based on
projections of the Streamflow Quality Affecting Rivers and Estuaries (SQUARE) Model
and the framework for the Nutrient Offset Contributions Scheme for the Swan-
Canning catchment (Kelsey et. al. 2010; BDA Group 2008).
The simulation does not include oxygenation plants. To run a forecasting simulation
that has the combined effects of both oxygenation plant and nutrient reduction, which
is very likely to be applied in reality, the 2050 forecasting simulation of nutrient
reduction without oxygenation plant is necessary. Furthermore, through a forecasting
simulation that has the effects of both oxygenation plant and nutrient reduction, how
the oxygenation plant affects the nutrients level could be investigated.
3.5.3.2 Method
Changes in the nutrients levels of the Swan-Canning Estuary will impact the growth
rates of various algae species thus impacting the overall health of the system (De
Roach 2006; Turner et. al. 2006). With nitrogen and phosphorous values increasing in
the Swan-Canning Estuary through the effects of an increasing population and climatic
responses the Swan River Trust along with the Government of Western Australia and
the Department of Water (DoW) have outlined a plan to reduce the Nitrogen and
Phosphorus levels in the Swan-Canning Estuary by 48% and 46% respectively
(Kelsey et. al. 2010). This reduction target is based on projections of the Streamflow
Quality Affecting Rivers and Estuaries (SQUARE) Model and the framework for the
Nutrient Offset Contributions Scheme for the Swan- Canning catchment (Kelsey et. al.
2010; BDA Group 2008).
The Nutrient Offset Contributions Scheme was developed for the Swan River Trust by
the BDA Group in 2008. The Scheme investigates the financial, legislative and
implementation constraints that surround offsetting residual loads of nutrients from
new developments into the Swan-Canning Estuary in an attempt to improve water of
quality of the system. A reduction in nutrient inflow into the system is proposed by
ensuring that all new land developments within the Swan-Canning Catchment area
manage the runoff caused by the development by making offset contributions
equivalent to the residual loads contributed by the developer. It is proposed that these
contributions be collected into a ‘Nutrient Management Fund’, run by the Swan River
Trust that is used to mitigate and manage nutrient runoff caused by new land
developments to deliver equivalent nutrient reductions (BDA Group 2008).
The predictive SQUARE Model has allowed for the ability to determine the potential
reduction in nutrients in thirty sub-catchments across the Swan-Canning Estuary thus
49
ensuring that realistic and effective nutrient targets are proposed (Kelsey et. al. 2010).
The model was able to identify the sub-catchments with the highest nutrient loads thus
allowing for the effective implementation of management methods in the sub-
catchments of high nutrient contributions and the ensuring that sub-catchments with
acceptable water quality be maintained. The SQUARE Model also allows for the
differentiation between rural and urban sub-catchments as the nutrient contributions
from rural and urban areas is likely to be very different and this allows for the focus of
the management methods to altered for urban and rural areas (Swan River Trust
2009). The nutrient reduction targets are variant across each of the thirty sub-
catchments as can be seen below in Figure 23. The maximum proposed Nitrogen
reduction the Swan-Canning Estuary is 1.0mg/L and which equates to a total reduction
in nitrogen across the system of 48%. The maximum proposed Phosphorus reduction
in the Swan-Canning Estuary is 0.1mg/L which equates to a reduction in phosphorus
across the system of 46%. It is important to note that these reduction targets are higher
than the Estuarine threshold targets proposed by Anders and Schroeder (2003) as
they consider the financial, legislative and implementation constraints associated with
any proposed management methodologies.
50
Figure 23 Total Nitrogen and Total Phosphorus target concentrations for the thirty sub-catchments within the
Swan-Canning Estuary (Kelsey et. al. 2010).
To determine the effectiveness of reducing the nutrient concentrations in the Swan-
Canning Estuary the model will be run for the year 2050 with an overall nutrient
reduction. The Nitrogen and Phosphorus concentrations will be reduced by 48% and
46% respectively for each of the seven regions in respect to the concentration values
noted in the 2008 baseline year, as per the prescribed reductions presented by the
DoW in 2010. That is, we have assumed for this model output that all tributaries in the
Swan-Canning Estuary have had a reduction in Nitrogen and Phosphorus load into
the system by 48% and 46% respectively by the year 2050. A single, overall
percentage reduction for all of the tributaries in the Swan-Canning Estuary was chosen
for simplicity, as the purpose of this assessment is to evaluate the effect that a
51
reduction in nutrients is likely to have on the state of the Swan-Canning Estuary and
a total reduction of nutrients will present an effective result for this purpose. The load
targets have been derived using the climate sequence for the period 1997 to 2006 and
the targets will be different if deduced from a different time period as load is dependent
on local rainfall and inflow patterns (Swan River Trust 2009). The results from this
model output are simplistic approximations of the effects that the proposed nutrient
reduction removal strategy, from the Swan Canning Water Quality Improvement Plan,
will have on the Swan-Canning Estuary by the year 2050.
52
4 RESULTS AND ANALYSIS
4.1 REFERENCE CONDITIONS
Figure 24: Simulation 1 (the Characteristic Baseline) vs Simulation 2 (2050 Baseline) Dissolved Oxygen (DO) Plots. The top plot (a) shows the values for the sites in the surface layer of the domain. Similarly, the bottom plot (b) shows the oxygen concentrations in the benthic cells (bottom layer). From left to right the x-axis shows the various monitoring sites across the river, Blackwall Reach (BLA), Armstrong Spit (ARM), Heathcote (HEA), Nile St (NIL), St John of God Hospital (STJ), Success Hill (SUC) and Middle Swan Bridge (MSB), from west to east of the model domain. The red line shows the hypoxic threshold of 2mg/L. The errors bars show the maximum and minimum values for DO concentrations for each site.
53
Figure 25: Simulation 1 (The Characteristic Baseline) vs Simulation 2 (the 2050 Baseline) time series plots over one year at MSB (Middle Swan Bridge), in Upper Swan. The top plot (a) shows the dissolved oxygen time series and the bottom (b) shows the salinity time series plot. The red line shows the hypoxic threshold value of 2mg/L.
54
Figure 26: Simulation 1 (the Characteristic Baseline) vs Simulation 2 (the 2050 Baseline) salinity plots. The top plot (a) shows the values for the sites in the surface layer of the domain. Similarly, the bottom plot (b) shows the salinity in the benthic cells (bottom layer). From left to right the x-axis shows the various monitoring sites across the river, Blackwall Reach (BLA), Armstrong Spit (ARM), Heathcote (HEA), Nile St (NIL), St John of God Hospital (STJ), Success Hill (SUC) and Middle Swan Bridge (MSB), from west to east of the model domain. The errors bars show the maximum and minimum values for salinity (psu) for each site.
55
Figure 24 shows comparative results of the DO concentrations between the
Characteristic Baseline Simulation and the 2050 Baseline Simulation, for each site
along the river. From left to right the x-axis shows the various monitoring sites across
the river, Blackwall Reach (BLA), Armstrong Spit (ARM), Heathcote (HEA), Nile St
(NIL), St John of God Hospital (STJ), Success Hill (SUC) and Middle Swan Bridge
(MSB), from west to east of the model domain, and are shown in Figure 7
There was an overall decrease in median and Inter Quartile Range (IQR), or the likely
range of DO concentration values for 2050 for the benthic cells (Figure 24b) at each
site. This was also the case for the surface cells (Figure 24a), though to a lesser extent,
with the percentage differences between the two medians no greater than 3% (MSB)
for the surface values. This is in comparison to the DO benthic concentrations for the
2050 Baseline Simulation differing from the median values of the Characteristic
Baseline Simulation by up to approximately 35% (STJ). This significant difference
between the two Simulations suggests that typically values of DO concentrations can
be expected to be reduced by the year 2050.
Also, from the results in Figure 15b, the site MSB (Middle Swan Bridge, Upper Swan),
the eastern extent of the estuary and model bounds, showed the largest change in
IQR from the Characteristic Baseline Simulation to the 2050 Baseline Simulation. This
site showed an increase in the likely range of DO concentrations in the 2050 Baseline
of 47% from the range of DO values in the Characteristic Baseline.
Patterns across the complete time series results for DO (see Appendix E for the plots
for all 7 locations in the system), do not deviate substantially between the two
Simulations, at least for the majority of the locations along the river. The four locations
at the greatest distance from the coast all experienced low oxygen (hypoxic), during
both of the simulations.
To builds on the trends identified for MSB (Upper Swan) in Figure 24, unlike other
locations, MSB experienced significant differences between the DO results of the two
Simulations. As can be seen in Figure 25a, a maximum difference in concentration of
5.5mg/L occurred in June for this site. This was during a prolonged hypoxic event at
this location from the beginning of April until the end of May, where the benthic DO
concentrations fell below 2mg/L for the 2050 Baseline Simulation. Importantly, the
Characteristic Baseline Simulation did not mimic this two month long low oxygen
event, discussed in section 5.1.
Another significant point of interest from Figure 25a, is the hypoxic event during
February (this event occurred in both Simulations). The lowest hypoxic DO
concentration occurs at this time, MSB, falling close to anoxic levels for the 2050
Baseline Simulation, though both Simulations experienced hypoxic levels at this time
period. The February low oxygen event also aligns with a similar timed significant
stratification event, with salinity differences between surface and benthic regions
averaging around 6 psu for the duration, as shown in the bottom plot, Figure 26b. This
56
was the strongest stratification event, meaning the event showing the greatest
difference in salinity between the top and benthic cells, for both Simulations.
During the 2050 Baseline Simulation, other locations, further downstream, also
experienced deviations from the patterns of the Characteristic Baseline, though not to
the magnitude of that shown at MSB. The results of both STJ and SUC showed shorter
stints of low oxygen events, where the DO concentrations for the benthic cells fell
below hypoxic levels for the Characteristic Baseline STJ in September, but not in the
2050 Baseline, see Appendix E.
Additionally, the change in salinity, for each site, between the two Simulations is shown
in Figure 26. The median and IQR of salinity values (psu) were consistently larger for
the 2050 Baseline than the Characteristic Baseline, which was exaggerated for the
benthic cells. The results of Figure 26b show a trend of increasing salinity, particularly
for the Upper Swan, or the eastern extent of the Estuary. The complete set of time
series plots comparing salinity between the two Simulations can be found in Appendix
E. In general, there appeared to be no notable differences in the duration of
stratification events between the two Simulations for the different sites across the river,
stratification events were present at approximately the same time and for similar
durations. The intensity of stratification, or lack of mixing, in the river was clearly
exacerbated slightly for those sites further up river (STJ, SUC and MSB, Appendix E
for the full time series plot). In general, stratification was considered to occur when
there was significant separation in salinity values (psu) between the benthic and
surface layers, at the same point in time, as discussed in section 2.4.
A clear example of stratification can be seen in Figure 25b, for the period from the start
of April to the end of May there was a clear difference in salinity trends between the
two simulations. The Characteristic Baseline shows no distinctive separation between
the salinity surface and benthic values during this time, whereas there was a maximum
difference in salinity of approximately 10 psu for the two layers of the 2050 Baseline
Simulation. This clear stratification event at this location, coincides with the time of
year of the prolonged anoxic conditions identified for the 2050 Baseline Simulation, in
Figure 25a.
57
4.2 EXTREME SIMULATIONS
4.2.1 2050 Low and High Extreme (sim 4 and 5)
4.2.1.1 Salinity
Salinity results display a strong trend of increased vertical stratification as the model
scenario becomes more extreme from the low, mean to the high extreme run (sims
4,2,5). This is most prominent in the Upper Swan Region but is also seen throughout
the lower to middle reaches. Figure 27 shows the vertical stratification at Middle Swan
Bridge from early April through to July, displaying strong stratification for high extreme
(sim 5), but very minor stratification for low extreme (sim 4). Similarly, the level of
vertical stratification at Nile street (Figure 28) in the ‘low extreme’ simulation is much
smaller than the stratification seen in the ‘high extreme’ simulation.
Comparison of the box plot diagrams for surface and bottom waters (Figure 30, Figure
31) helps to explain the increase in stratification seen for the ‘high extreme’ run. The
‘high extreme’ run has a much smaller variability in the surface profile than the bottom
profile for the upstream regions. What this indicates is that the surface layer remains
relatively fresh but the bottom layer, even upstream, is fluctuating between relatively
saline and fresh depending on the time of year. Of the four parameters within the
model that differ between runs (sea level, flow volume, temperature, nutrient
concentration), there are two parameters in the model that are most likely to influence
the contrast in vertical stratification seen; sea level rise and flow volume reduction.
Figure 29 displays the results for salinity in the lower Swan at Blackwall Reach. Very
minor difference between the low, mean and high extreme scenarios are observed,
and this provides insight as to the cause for the increased stratification shown in the
middle to upper reaches of the Swan. The Swan River has been shown to typically
fluctuate between a pattern of becoming a highly vertically stratified salt wedge estuary
and a slightly vertically stratified estuary, where salinity increases towards the sea but
varies little with depth (Thomson, 2001). The latter often occurs when flow is low and
the tidal or wind forces dominate the hydro dynamics. Hence, as expected we see
higher stratification in the upper reaches at times of high flow in the winter months.
What is counter intuitive, is that we see higher vertical stratification for the ‘high
extreme’ run, which has lower inflow.
A suggestion for such observation is that the flow levels are high enough in all runs to
create a clear salt wedge at times of high flow, shown by the freshwater floating on the
surface in the upper reaches for all simulations. The flow has not been reduced to a
point the prevents the formation of the salt wedge and the development of two distinct
layers. The surface salinity profile for the Upper Swan (Figure 27) region shows the
salinity is equal for all runs at times of high flow in (April), but coming out of this into
May and June the high extreme run salinity rises much faster than the low extreme
run. The bottom profiles differ, where the ‘high extreme’ run has higher salinity
throughout the entirety of the run causing increased vertical stratification to the fresh
surface water.
58
As we observe what happens further downstream over the months of April to August,
for example Nile Street (Figure 28), the bottom profile is similar, but the surface profile
now differs between runs. What this suggests is that the lower inflow in the ‘high
extreme’ run allows the salt wedge to creep further upstream during the winter months.
A key point is that the reduction in flow volume is not sufficient to reduce the effect of
the salt wedge and vertical stratification, but instead effects the location in the river
where it may have its greatest effect. The differences in sea level rise are also likely
to have contributed to such changes in hydrodynamics, pushing the salt wedge
upstream with an increased volume of saline water. The implications of this are
significant, and such effects are explored further in ‘no inflow’ simulation, where the
inflow is reduced to zero.
Figure 27: Salinity (psu) results for upper Swan River. All three simulation scenarios are displayed, shown by the
lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
59
Figure 28: Salinty (psu) for middle Swan River. All three simulation scenarios are displayed, shown by the lower
extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
Figure 29: Salinty (psu) for lower Swan River. All three simulation scenarios are displayed, shown by the lower
extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
60
Figure 30: Salinty (psu) box plots for surface. All three simulation scenarios are displayed, shown by the lower
extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
Figure 31: Salinty (psu) box plots for bottom. All three simulation scenarios are displayed, shown by the lower
extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
61
4.2.1.2 Dissolved Oxygen
Dissolved Oxygen results for the low, mean and high extreme show similar trends in
the lower to middle upper reaches of the Swan but begin to show contrasting results
as one progresses further upstream. In the upper reaches we see more extreme low
DO events for the bottom water profile of the ‘high extreme’ run, and a greater number
of events where the DO falls below the hypoxic threshold (Figure 32, Figure 33). The
differentiation between scenarios in low DO events is seen more prominently in the
autumn and winter months for the upper region (Figure 32), and in the spring/summer
months for the middle swan (Figure 33). Most notably, at St John Hospital (Figure 33),
from September to December we see seven separate occurrences of DO falling below
the hypoxic threshold in the ‘high extreme run’, and only two events for the ‘low
extreme’ run.
The box plot profiles for the surface and bottom (Figure 36, Figure 37) show the
surface profiles for each run are similar throughout the length of the Swan system.
However, the bottom profile dissolved oxygen in the middle to upper reaches shows
much greater variation between runs, with the ‘high extreme’ run displaying a lower
mean DO and greater variation in dissolved oxygen.
The greatest low DO events occur in phase with the greatest level of stratification. As
has been outlined by Thomson (2001), it is typical to see low DO conditions beneath
highly vertically stratified regions. The differences seen between the low and high
extreme scenarios may in part be explained by this, as regions of more intense
stratification show more extreme low DO events. The difference in temperature offset
between the low and high extreme runs is also likely to contribute to difference in DO
we observe, as the capacity for water to retain oxygen increases with temperature.
Warmer water temperatures also result in increased sediment oxygen demand (Swan
River Trust, 2007).
62
Figure 32: Dissolved Oxygen (mg/L) for upper Swan River. All three simulation scenarios are displayed, shown
by the lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
Figure 33: Dissolved Oxygen (mg/L) for upper/ middle Swan River. All three simulation scenarios are displayed,
shown by the lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
63
Figure 34: Dissolved Oxygen (mg/L) for middle Swan River. All three simulation scenarios are displayed, shown
by the lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
Figure 35: Dissolved Oxygen (mg/L) for lower Swan River. All three simulation scenarios are displayed, shown by
the lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
64
Figure 36: Dissolved oxygen box plots for surface. All three simulation scenarios are displayed, shown by the
lower extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
Figure 37: Dissolved oxygen box plots for bottom. All three simulation scenarios are displayed, shown by the lower
extreme (sim 4), mean (sim 2) and upper extreme (sim 5).
65
4.2.2 Year 2050 Summer Flood Event (sim 6)
Figure 38: Observed flow data form the upper Swan used in sim 6.
The observed 2000 flood flow data indicated that the event started around the 22nd of
January, reached a peak on the 25th which was sustained for 4 days until the 29th,
after which it gradually dissipated and returned to normal levels on the around the 25th
of February. The scale of such an event was unprecedented given the
characteristically dry summers of the region.
0
50
100
150
200
250
300
Fow
Vo
lum
e m
3/s
Flow Volume: Upper Swan
66
4.2.2.1 Salinity
Figure 39. The salinity results indicate a rapid flushing of the system and decrease in stratification until the end of
February as conditions begin to return to normal. This pattern is common to simulated sites along the Swan
Canning
Algal species require the absence of highly saline waters for rapid growth. For many
freshwater algal groups such as Microcystis the critical value is around the 10 psu
level (Atkins et al., 2001). A simple check of the simulated results in Figure 39 indicates
that conditions at STJ are well within the required levels for an algal bloom. Cross
referencing with observed values from the same time period affirms that this process
has been suitably reproduced in the simulation. It is interesting to note that after the
flow stops in late February it takes approximately one month for salinity to return to
pre-flood levels following a rapid intrusion of the saltwater wedge.
67
Figure 40. The observed measurements of salinity strongly match the rapid flushing and slow recovery of saline
waters seen in the modelled scenario (Figure 39) (DWER 2018).
4.2.2.2 Total Nutrients
Excessive concentrations of phosphorous is the most common cause of eutrophication
in freshwater lakes, reservoirs, streams, and in the headwaters of estuarine systems
(Correll, 1999; Schindler, Carpenter, Chapra, Hecky, & Orihel, 2016). In the ocean,
nitrogen is generally considered to be the limiting nutrient controlling primary
production (Correll, 1999). Estuaries are a somewhat of a transition zone in regard to
salinity and algal group dominance. In these conditions both excessive nitrogen and
phosphorous can both potentially contribute to algal blooms (Correll, 1999). However
for algal blooms caused by the resulting flow of intense rainfall events form in
predominantly fresh highly flushed conditions. In these cases fresh water algal groups
(such as Microcystis) are more susceptible to rapid growth and formation of bloom
conditions. As such it may be more pertinent to consider total pohospurious as the the
limiting nutrient for scenarios such as the one modelled here.
68
Figure 41. Total Phosphorous for STJ did not reach excessive values that would suggest a eutrophied system
nor would it suggest a bloom is particularly likely.
Comparing to observed data for the same timescale in 2000 (Figure 42) suggests that
levels of phosphorous predicted in the simulation were not uncharacteristic for the
system previous to the flood. A mild peak in surface concentration for STJ was
observed that matches the peak of the flow data (Figure 38), however this spike
returns to normal levels almost half a month before flow rates return to typical values.
This may suggest the system has been flushed to such a degree that initially large
amounts of nutrients have flown out to sea. However, the same flow rates and
concentrations caused a bloom in 2000. It would be intuitive to come to the conclusion
that the sharp and premature decline of phosphorous in the system was caused by
the sequestration of the nutrient by a large bloom of algae. Indeed, after the flow (and
accompanying nutrients) returns to a minimal level the subsequent increase in total
phosphorous on the bottom of the system (dotted line Figure 41) could represent the
detritus sequestered phosphorous of a falling biomass of algae. If this was the case,
we would expect to see a rapid decline in oxygen near to the riverbed as the algal
biomass settles and is anaerobically decomposed. Such a decrease is observed in the
simulation and is displayed in Figure 45, however, this could alternatively indicate a
return of the salt wedge and associated stratified conditions, or a return to pre-flood
quasi-steady conditions (hypoxic conditions were present immediately prior to the
flood).
69
Figure 42. Observed data for total phosphorous taken at varied depths (DWER 2018)
Figure 43.Total Chlorophyll-a counterintuitively decreases as conditions for algal growth improves, and
decreases further post flood event.
70
Figure 44. Low levels of TCHLA were simulated relative to observed data taken in the same period (DWER 2018)
If the model had indeed simulated a large algal bloom a sure indicator of this would be
an increased level of chlorophyll-a (a depth-averaged measurement used to represent
the concentration of algal biomass in the water column). There is a small increase in
total chlorophyll-a after the initial flushing of the system (Figure 43), however this
increase appears small in proportion to the levels of conditions that contribute to algal
bloom risk. Indeed, observed data (although discontinuous) shows a much higher
variability (between 4 and 100 ug/L) and increased level of chlorophyll-a (average of
16 ug/L) for the same period (Figure 44). In explaining this, there exists the possibility
that the model wasn’t able to capture the extent of biogeochemical processes in this
case. These processes are still under development, and particularly for some algal
groups there is a low correlation between predicted and observed values (SCERM
2016). Additionally, in order to save on simulation time, the model used in this study
utilised a limited biogeochemical module, as such plankton group results are highly
speculative. For the resolution required in this case (approximately 1 month)
reproduction of real world results may be variable. This may explain some
discrepancies between results, such as causal factors and algal biomass.
Modelling limitations aside, the quantitative and temporal sensitives that must be
achieved for an algal bloom to occur are highly variable in real world and simulated
conditions. The dynamics in river-estuary conditions are subject to a range of complex
couplings and forcings, not just water quality. Tidal coupling, day night cycling,
meteorological conditions and a variety of local factors all must ‘align’ for a rapid
growth of algae to be possible. The ability to capture the sensitivities and interactions
of these processes are beyond current understanding and as such it is difficult to
71
reproduce single events consistently. As such it is recommended that only the results
of the physical parameters (dissolved oxygen and salinity) be considered to be
representative here.
4.2.2.3 Dissolved Oxygen
Figure 45. Rapid decline in bottom dissolved oxygen (dotted line) may seem to indicate anaerobic decomposition
of algal biomass. However, it is more likely that this is due to salinity stratification or other external forcings as
only a small amount of biomass (Tchla) was simulated.
4.2.3 “Mean Year 2050” No Inflow (sim 7)
4.2.3.1 Salinity
Lack of inflow from rivers and drains, as well as excessive evaporation of fresh water
throughout the estuary can result in a highly saline system. We use the Nile Street and
St John Hospital results to make specific comparisons to the mean 2050 run, as they
display similar salinity levels from January to April, but then strongly contrasting levels.
The comparison between the two simulations in Figure 46 clearly demonstrates the
contrast between a riverine flushed estuary and one that does not experience any
fresh water incursion. In the 2008 baseline simulation (blue line in Figure 46) heavy
spring rainfall in April and normal winter rainfall can be seen causing rapid decreases
in salinity through flushing. However, any such flushing is absent in the no inflow
simulation (shown in Figure 46 as the red time series). Other sites show the same
pattern and be viewed under appendices.
72
Figure 46: Nile Street salinity time series (middle Swan River). Note the initial dramatic increase in simulated
salinity (red) approaching 2008 levels (blue) by Febuary despite an artifically lower initial salinity due to modellin g
error.
Figure 47: Salinity time series for St John Hospital in the middle Swan. As the initial 2008 salinity is near the
erroneous initial 20 psu of the simulation the St John Hospital provides the only accurate indication of e arly
salinity rise and deviation due to spring rainfall.
Another key point in the salinity results is the difference in stratification between runs.
Figure 46 and Figure 47 show extreme vertical stratification in sim 1 (2008 mean
73
baseline), most notably from May to August shown by the difference between the top
and bottom salinity levels (dotted and full blue lines). In comparison, the no inflow 2050
run shows enhanced salinity but to no levels of vertical stratification. These differences
in hydrodynamics have significant impacts on the biology and other water quality
parameters of the system, most notably dissolved oxygen (DO) patterns (discussed in
the following section).
Figure 46 confirms the increase in salinity from January to April is similar for both runs,
as we would expect, as the mean 2050 run replicates the rainfall pattern of 2008 (dry
summer) are similar to the 2050 no inflow conditions. In other words, the effect of the
2050 predicted mean sea level rise temperature increase is being isolated here, where
the significant increase in volume of saline water to the system. This suggests that the
dramatic deviation in salinity in April is due to seasonal inflow alone. An extension of
a dry summer and a delay of spring rain is shown to increase the duration of
excessively saline waters but does not significantly increase the severity during this
period or throughout winter. However, an autumn drought could increase salinity prior
to an evaporative intense summer which would predispose the system to becoming
hyper saline.
Without seasonal flushing, salinity throughout the system does not experience the
dramatic rise and fall of a standard estuary. The result is a fairly constant raised salinity
level throughout the entirety of the system, as presented by Figure 48 showing a much
smaller variation in salinity when compared to the 2008 baseline year, acting almost
like a “sidearm” to the ocean.
Figure 48: Box plot of bottom salinity throughout the extent of the Swan River for the “mean 2050” run (blue) and
the “no inflow 2050” run (red).
74
In the summer months, when the solar irradiance and temperature is high, the salinity
slowly increases due to the associated evaporation. Fresh water loss from evaporation
leads to more pronounced increases in salinity towards the riverine extents (as the
river surface becomes large relative to the water depth). Generally upstream regions
are fresher due to catchment inflow and their distance from the saline waters nearer
to the ocean inlet. However, the Nile Street and St John Hospital sites in the Middle
Swan reach hyper saline levels and exceed the marine salinity level of 35.5 psu in the
summer (Figure 46 and Figure 47). The Blackwall Reach, Armstrong Spit and
Heathcote sites also experience the same forcing towards hyper salinity, but, as they
are within the tidal influence of the estuary proper, are exposed to the now relatively
fresh ocean water.
The perpetual input of marine water with the tidal estuary mixing provides a soft cap
on the Lower Swan’s ability to become hyper saline. In other words, the Lower Swan
is strongly buffered against any rise in salinity above marine levels despite significantly
detrimental conditions. It is possible that this cap can be breached somewhat as in all
three Lower Swan sites there is a significant rise above this level in the summer
conditions at the tail end of the simulation. One explanation for this is that a critical
point is reached around November wherein the buffering effect of mixing sea water is
overcome by evaporative forcing, thereby allowing increasing salinity.
Figure 49: The simulated salinity reaches a plateau at near marine levels, then, at the end of the simulation, is
seen exceeding this cap.
75
Hyper salinity and low variability of salinity are typical properties are common to so
called inverse estuaries, which have been observed in other areas of low flow and high
evaporation. In these systems, the upstream extents become increasingly dense and
hyper-saline. As this dense water sinks towards the river bed and is pulled by gravity
towards the ocean while the relatively less dense water moves upstream close to the
surface, becoming increasingly saline itself through evaporation.
While the salinity upstream is still lower than the near ocean salinity, the change of
conditions are much more dramatic in the upstream extents and we are confident that
given a longer period of simulation the upstream salinity will continue to rise
dramatically, becoming hypersaline and surpassing the downstream salinity (Schettini,
Valle-Levinson, & Truccolo, 2017) and (Lavın, Godınez, & Alvarez, 1998). Such
conditions would represent an inverse estuary.
4.2.3.2 Dissolved Oxygen
The no inflow run shows significantly fewer and less extreme periods of low DO events
(Figure 50 and Figure 51). In sim 1 (2008 baseline), the months of May to August
display periodic states of extremely low DO levels below the hypoxic threshold, most
notably in the bottom section of the water. This can be explained in part by
understanding the salinity data. Extremely low levels of DO in the Swan River have
been shown to correlate with areas of water underneath severe vertical stratification
(Thomson, 2001). The Swan-Canning system (in its typical state) shows strong
stratification where the different densities of freshwater inflow and saline ocean water
interact. In the case where inflow is completely stopped (sim 7), the extent of vertical
stratification is lessened, and the result is that the system displays less extreme low
DO events.
76
Figure 50: Success Hill DO levels. Note the severe low DO levels (<2mg/L) in sim 1 bottom in February and from
April to July.
Figure 51: Nile Street DO levels. Note the severe low DO levels in sim 1 bottom.
77
Overall, dissolved oxygen shows lower variability throughout the year when compared
to the mean 2050 run. The mean 2050 run shows a seasonal peak in DO throughout
winter, whereas the no inflow run tends to show a lower seasonal rise in DO. The no
inflow run shows very similar trends throughout the summer months as the 2050 mean
simulation.
Although the extremely low DO events are shown to be milder and less frequent, the
implications for the water quality of the Swan may not be entirely positive. It is possible
a drying climate will reduce extreme low DO events such as major fish kills, but the
implications for the overall health of the system may be contrasting. The lack of a
seasonal rise in DO over the winter months may have significant effects on the state
of Swan River. Dissolved oxygen is a key concern for aquatic ecosystem health and
water quality. Temperature increases due to climate change will increase the capacity
of the water to hold oxygen, however, this will also increase the amount of bacterial
respiration and rate of decomposition. Increased stratification due to the saline wedge,
and sediment oxygen demand are also likely to cause lower than average dissolved
oxygen. Potential for increased frequency of extreme low dissolved oxygen events is
also apparent as flushing is reduced.
4.2.3.3 Phytoplankton
Phytoplankton concentration up stream of Heathcote show significantly lower levels of
dinoflagellates in the no inflow run (Figure 52). This can be explained by the high
salinity, reducing the ability for the species to reproduce. It was surprising to see that
in the lower Swan, such as Blackwall Reach, the 2050 no inflow run showed very
similar rise and falls to the 2050 mean run (Figure 53).
78
Figure 52: Success Hill Dyno concentrations.
Figure 53: Blackwall Reach dyno concentrations.
79
4.3 MANAGEMENT SIMULATIONS
4.3.1 Oxygenation (sim 3 and 8)
Results are presented in the form of oxygenation concentration plots over time for
each of the seven monitored locations. As the simulations were run from January to
March 2050, oxygen concentrations at the seven locations were plotted for this time
period. Box plots were constructed from the results obtained from the simulations to
decipher the median and quartiles of the oxygen concentrations for the two simulations
at the seven locations. The results from simulation three and simulation eight were
compared to results obtained from simulation two (no oxygenation, baseline 2050
simulation) to evaluate the effect that the artificial oxygen strategies had on the oxygen
concentrations in the Swan-Canning Estuary.
Simulation Description
2 · Using 2050 baseline date;
· Switching off oxygenation plant in the model.
3 · Using 2050 baseline data
· Switching on oxygenation plant by using 2017 oxygenation data.
8 · Using 2050 baseline data
· Addition of an extra oxygenation plant at Nile St river using similar
operating conditions as the two other oxygenation plants.
Table 10 Description of all oxygenation Simulations
80
Figure 54. Oxygen concentrations recorded at Nile Street for Simulation 2, Simulation 3 and Simulation 8.
Figure 55 Oxygen concentrations recorded at St John Hospital for Simulation 2, Simulation 3 and Simulation 8.
81
Figure 56 Oxygen concentrations recorded at Success Hill for Simulation 2, Simulation 3 and Simulation 8.
Figure 57 Oxygen concentrations recorded at Middle Swan Bridge for Simulation 2, Simulation 3 and Simulation
8.
82
Figure 58 Box plot representing the concentration of oxygen recorded at the surface water layer over January
2050 to March 2050 for the seven study areas.
Figure 59. Box plot representing the concentration of oxygen recorded at the bottom water layer over January
2050 to March 2050 for the seven study areas.
83
4.3.1.1 Current Oxygenation Strategies
The impact of the current artificial oxygenation strategy for the year 2050 is analysed
by comparing the oxygenation concentration results from simulation 3 and simulation
2.
It can be seen in Figure 58 that under the implementation of an artificial oxygenation
strategy at Middle Swan Bridge the minimum level of oxygen that can be attained by
the river is above 4mg/L. However, even though there is an approximate increase of
1-1.2 mg/L of oxygen in the bottom layer, the mean and minimum level of oxygen that
the Middle Swan Bridge area can attain is still less than the required 4 mg/L for a
‘healthy’ Swan-Canning system. This can be seen in Figure 59
It can be observed from the Success Hill plot Figure 56 that switching on the oxygen
plants will approximately increase the oxygen level by 1 mg/L at both the bottom layers
and upper layers for what region of time. However, enough oxygen is not being
pumped into the river to increase the levels above 4 mg/L for the bottom layer. The
upper layer is showing satisfactory results since the minimum oxygen level that can
be attained by the river is above 4 mg/L with the oxygenation plants switched on.
A slight increase of oxygen level can be observed on the bottom layers of St John river
however, the levels of increase are not enough to bring up the oxygen content in the
river above 4 mg/L (Figure 55) Negligible changes in the oxygen level is observed as
from Nile St., Heathcote, Armstrong Spit and Blackwall reach.
Lower reaches of the SCE (Blackwall reach to Heathcote) are unaffected by the
artificial oxygenation systems as oxygen levels display minimal to no increase, as seen
in Appendix K.
4.3.1.2 Enhanced Oxygenation
It can be observed in the plots that the introduction of a new oxygenation plant slightly
improves oxygen conditions at Nile St river. It was expected that the oxygen conditions
further downstream would be improved. However, the plant had no influence on the
Middle Swan Bridge, Success, St John, Heathcote, Armstrong and Blackwall Reach
rivers. The new oxygenation plant had minimal effect on the oxygen content at the
bottom layers in the river. The oxygenation plant was unsuccessful in pumping enough
oxygen to bring up the level above 4 mg/L.
A similar pattern is observed at the surface layers compared to bottom layers. The
new oxygenation plant does not affect the Middle Swan Bridge, Success, St John,
Heathcote, Armstrong and Blackwall Reach rivers. However, the plant manages to
bring the level of oxygen on the surface layer above 4 mg/L at Nile St. There is no
significant change in oxygen concentration after the introduction of an additional
artificial oxygenation facility. There is also no significant decrease in the duration of
the low oxygen events that are present in 2050.
84
4.3.1.3 Limitations
The shortened run periods for Simulation 3 and Simulation 8 resulted in discrepancies
in the spin up time between Simulation 2. This has resulted in the model displaying a
strong gradient in salinity for the initial conditions along the salt wedge in Simulation
2. This gradient was not held constant for Simulation 3 and 8. The lower Swan is too
fresh in the initial conditions for Simulation 3 and Simulation 8 and the time taken for
the salt wedge to move upstream has taken substantially longer. Examining the salinity
plots seen in Figure 60 it is evident that the difference in salinity levels between
Simulation 2 and Simulation 3 become more apparent as it moves further upstream.
Slightly more saline water will act to impede the oxygen diffusion limiting the dispersion
of oxygen across it target zone. The inconsistencies in time period of analysis between
Simulations 2 and Simulations 3 and 8 also mean that the analysis of the box plots is
not completely reliable as range for different time periods and a different number of
data points.
Another limitation is that the model cannot be run reactive to changes in oxygen it is
unable to represent the true nature of the oxygen plants that operate in the upper
Swan River. This limitation prevents the oxygenation management strategies being
accurately depicted with the model and hence unable to be accurately analysed.
Figure 60 Discrepancies in salinity as a result of inconsistent run-up time.
85
4.3.1.4 Cost Benefit Analysis for oxygenation
A cost-benefit analysis is applied in order to understand the performance of the
additional oxygenation plant. Hipsey (2014) undertook an economic assessment for
the two existing oxygenation plants at the upper stream region. The cos-benefit
analysis should be consistent for all the oxygenation plants, including the two existing
plants and the additional plant, hence the cost-benefit analysis for the additional
oxygenation plant will follow the methodology which Hipsey (2014) applied for the two
existing oxygenation plants. In this cost-benefit analysis, we only focus on
performance of plant, the costs such as plant construction cost, indirect costs
associated with operation are not considered.
The operational cost for a plant 𝜃𝑝𝑙𝑎𝑛𝑡 is defined by calculating the electricity cost,
𝜃𝑝𝑜𝑤𝑒𝑟 and the oxygen input cost, 𝜃𝑜𝑥𝑦𝑔𝑒𝑛 (Hipsey et. al., 2014).
𝜃𝑝𝑙𝑎𝑛𝑡 = 𝜃𝑝𝑜𝑤𝑒𝑟 + 𝜃𝑜𝑥𝑦𝑔𝑒𝑛 (1)
The electricity cost:
𝜃𝑝𝑜𝑤𝑒𝑟 = ∑ 𝐸𝑗к𝑡𝑝𝑗,𝑡𝑡 (2)
Where 𝐸𝑗 is the energy units consumed (kWh) for the jth plant, к𝑡 is a binary flag (0,1)
indicating if the plant is operational or not at any time t, 𝑝𝑗,𝑡 is the price of ($ kWh-1).
The energy price is varied between peak and off-peak times.
Hipsey et. al. (2014) assumed the energy consumption of 44kWh for Guildford and 65
kWh for Caversham. He also mentioned that the off-peak time is from 10 pm to 8 am,
and the price of energy is 0.278, 0.100 $ kWh-1 for peak time and off-peak time
respectively. For the additional oxygenation plant, we assume the energy consumption
is the mean energy consumption of two existing oxygenation plants, and rounded to
the nearest integer, which is 55 kWh. Peak time of 12 hr and off peak of 10 hr are
determined by the off-peak time period. The additional oxygenation plant runs 24
Table 11. Flow Normalised Nutrient Concentrations (mg/L) for nitrogen and phosphorous. The table contains values for the reference year (2008), and the 50th percentile of recorded years. See Appendix C for corresponding duration curves.𝑁𝑖𝑡2008 is the normalised by flow nitrogen concentration in 2008, 𝑁𝑖𝑡50𝑡ℎ is the 50th percentile value for normalised nitrogen concentration found from the Duration Curve, 𝑆𝑐𝑎𝑙𝑒𝑁𝑖𝑡 is the scalar found from the normalised nitrogen concentration in 2008 and the 50Th percentile concentration for each site. 𝑃ℎ𝑜𝑠2008, 𝑃ℎ𝑜𝑠50𝑡ℎ and 𝑆𝑐𝑎𝑙𝑒𝑃ℎ𝑜𝑠 area the equivalent for phosphorus concentrations.