CLLAMM Dynamic Habitat: Habitat mapping and dynamic modelling of species distributions

CLLAMM Dynamic Habitat: Habitat mapping and dynamic modelling of species distributions
With the collaboration of:
Research supported by:
Water for a Healthy Country
CLLAMM Dynamic Habitat:
Habitat mapping and
dynamic modelling of species
distributions
Sunil K. Sharma, Simon N. Benger, Milena B.
Fernandes, Ian T. Webster and Jason E. Tanner
June 2009
Water for a Healthy Country
CLLAMM Dynamic Habitat:
Habitat mapping and
dynamic modelling of species
distributions
Sunil K. Sharma1, Simon N. Benger2, Milena B.
Fernandes1, Ian T. Webster3 and Jason E. Tanner1,*
1
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
2
School of Geography, Population and Environmental Management Flinders
University, GPO Box 2100, Adelaide, SA 5001.
3
CSIRO Land & Water, GPO Box 1666, Canberra, ACT 2601.
*
Corresponding author: [email protected]
June 2009
Water for a Healthy Country Flagship Report series ISSN: 1835-095X
ISBN: 978 0 643 09780 3
The Water for a Healthy Country National Research Flagship is a research partnership
between CSIRO, state and Australian governments, private and public industry and other
research providers. The Flagship aims to achieve a tenfold increase in the economic, social
and environmental benefits from water by 2025.
The Australian Government, through the Collaboration Fund, provides $97M over seven
years to the National Research Flagships to further enhance collaboration between CSIRO,
Australian universities and other publicly funded research agencies, enabling the skills of the
wider research community to be applied to the major national challenges targeted by the
Flagships initiative.
© Commonwealth of Australia 2009 All rights reserved.
This work is copyright. Apart from any use as permitted under the Copyright Act 1968, no part
may be reproduced by any process without prior written permission from the Commonwealth.
Citation: Sharma, S. K., Benger, S. N., Fernandes, M. B., Webster, I. T. and Tanner, J. E.
(2009) The CLLAMM Dynamic Habitat: Habitat mapping and dynamic modelling of species
distributions. CSIRO: Water for a Healthy Country National Research Flagship and South
Australian Research and Development Institute (Aquatic Sciences), Adelaide.
DISCLAIMER
CSIRO and the authors advise that the information contained in this publication comprises
general statements based on scientific research. The reader is advised and needs to be
aware that such information may be incomplete or unable to be used in any specific situation.
No reliance or actions must therefore be made on that information without seeking prior
expert professional, scientific and technical advice. To the extent permitted by law, CSIRO
(including its employees, collaborators and consultants) excludes all liability to any person for
any consequences, including but not limited to all losses, damages, costs, expenses and any
other compensation, arising directly or indirectly from using this publication (in part or in
whole) and any information or material contained in it.
For more information about Water for a Healthy Country Flagship or the National Research
Flagship Initiative visit http://www.csiro.au
Printed in Adelaide
SARDI Aquatic Sciences Publication Number F2008/001062-1
SARDI Research Report Series Number 336
Author(s):
Reviewers:
Approved by:
Signed:
Date:
Distribution:
Circulation:
Sunil Sharma, Simon Benger, Milena Fernandes, Ian Webster and Jason
Tanner
Nathan Bott, Peter Fairweather, Rebecca Lester, Sasi Nayar, Bertram
Ostendorf, Brad Page, Daniel Rogers, Simon Goldsworthy, Keith Rowling,
Paul Van Ruth, Ian Overton and Kathryn Wiltshire
Qifeng Ye
27 July 2009
Public Domain
The CLLAMM Dynamic Habitat
i
Foreword
The environmental assets of the Coorong, Lower Lakes and Murray Mouth
(CLLAMM) region in South Australia are currently under threat as a result of ongoing
changes in the hydrological regime of the River Murray, at the end of the MurrayDarling Basin. While a number of initiatives are underway to halt or reverse this
environmental decline, rehabilitation efforts are hampered by the lack of knowledge
about the links between flows and ecological responses in the system.
The CLLAMM program is a collaborative research effort that aims to produce a
decision-support framework for environmental flow management for the CLLAMM
region. This involves research to understand the links between the key ecosystem
drivers for the region (such as water level and salinity) and key ecological processes
(generation of bird habitat, fish recruitment, etc). A second step involves the
development of tools to predict how ecological communities will respond to
manipulations of the “management levers” for environmental flows in the region.
These levers include flow releases from upstream reservoirs, the Lower Lakes
barrages, and the Upper South-East Drainage scheme, and dredging of the Murray
Mouth. The framework aims to evaluate the environmental trade-offs for different
scenarios of manipulation of management levers, as well as different future climate
scenarios for the Murray-Darling Basin.
One of the most challenging tasks in the development of the framework is predicting
the response of ecological communities to future changes in environmental
conditions in the CLLAMM region. The CLLAMMecology Research Cluster is a
partnership between CSIRO, the University of Adelaide, Flinders University and
SARDI Aquatic Sciences that is supported through CSIRO’s Flagship Collaboration
Fund. CLLAMMecology brings together a range in skills in theoretical and applied
ecology with the aim to produce a new generation of ecological response models for
the CLLAMM region.
This report is part of a series summarising the output from the CLLAMMecology
Research Cluster. Previous reports and additional information about the program can
be found at http://www.csiro.au/partnerships/CLLAMMecologyCluster.html
The CLLAMM Dynamic Habitat
ii
Table of Contents
Table of Contents.................................................................................................... iii
Acknowledgements ................................................................................................ vi
Executive Summary ............................................................................................... vii
1.
2.
Introduction ....................................................................................................... 1
Habitat Mapping of the Coorong and Surrounds........................................... 3
2.1.
Executive Summary ................................................................................................ 3
2.2.
Introduction.............................................................................................................. 4
2.3.
Reference sites for habitat mapping in the Coorong............................................... 5
2.4.
Habitat classification scheme.................................................................................. 6
2.4.1 Habitat zone............................................................................................................. 6
2.4.2
Habitat category.................................................................................................. 7
2.4.3
Landform ............................................................................................................. 7
2.4.4
Wetland system .................................................................................................. 7
2.4.5
Water regime ...................................................................................................... 8
2.4.6
Wetland type ....................................................................................................... 8
2.4.7
Habitat type....................................................................................................... 10
2.4.8
Salinity............................................................................................................... 10
2.4.9
Vegetation......................................................................................................... 11
2.4.10
Cover percent ............................................................................................... 11
2.4.11
Habitat condition........................................................................................... 11
2.4.12
Area .............................................................................................................. 12
2.5.
Methods................................................................................................................. 13
2.5.1
Available datasets............................................................................................. 13
2.5.2
Habitat digitization and interpretation ............................................................... 18
2.5.3
Imagery classification........................................................................................ 18
2.5.4
Map validation and accuracy testing................................................................. 19
2.6.
Results .................................................................................................................. 19
2.6.1
Habitat maps of the reference sites .................................................................. 19
2.6.2
Habitat classification ......................................................................................... 30
2.7.
Discussion ............................................................................................................. 32
2.8.
Summary and conclusions .................................................................................... 33
2.9.
References ............................................................................................................ 33
2.10. Appendices............................................................................................................ 36
Appendix 2.1: A brief description of landforms reported in the Coorong (From Seaman
2003)............................................................................................................................... 36
Appendix 2.2: Habitat maps for the reference sites in the Coorong............................... 37
Appendix 2.3: Some photos of the habitats in the Coorong. .......................................... 43
3.
Digital Elevation Model of the Coorong and Surrounds ............................. 45
3.1.
3.2.
3.3.
3.3.1
3.3.2
3.3.3
3.4.
3.4.1
3.4.2
3.5.
3.6.
3.7.
4.
Executive Summary .............................................................................................. 45
Introduction............................................................................................................ 46
Available datasets and methodology .................................................................... 48
Available topographic and bathymetric data for the region .............................. 48
Satellite Imagery ............................................................................................... 49
Methodology ..................................................................................................... 50
Results .................................................................................................................. 53
Generalized Additive Model implementation and prediction error analysis...... 53
Seamless bathymetry for the Coorong ............................................................. 58
Discussion ............................................................................................................. 60
Summary ............................................................................................................... 61
References ............................................................................................................ 62
Sediment mapping of the Coorong: Implications for habitat distributions64
4.1.
Executive Summary .............................................................................................. 64
The CLLAMM Dynamic Habitat
iii
4.2.
4.3.
4.3.1
4.3.2
4.3.3
4.3.4
4.4.
4.4.1
4.4.2
4.4.3
4.4.4
4.5.
4.6.
4.7.
Introduction............................................................................................................ 65
Methods................................................................................................................. 66
Study area......................................................................................................... 66
Sampling ........................................................................................................... 66
Modelling and mapping..................................................................................... 68
Goodness of fit of GAM and IDW ..................................................................... 72
Results .................................................................................................................. 72
Measured sediment attributes .......................................................................... 72
GAM for North Lagoon...................................................................................... 76
Spatial maps for the North Lagoon and the Coorong ....................................... 85
Goodness of fit of GAM and IDW models......................................................... 91
Discussion ............................................................................................................. 92
Summary and Conclusions ................................................................................... 93
References ............................................................................................................ 94
5. Mudflat Geomorphology and Availability at Varying Water Levels in the
Coorong .................................................................................................................. 99
5.1.
Executive Summary .............................................................................................. 99
5.2.
Introduction.......................................................................................................... 100
5.3.
Methods............................................................................................................... 102
5.3.1
Source Datasets ............................................................................................. 102
5.3.2
Interpolation Methods ..................................................................................... 103
5.3.3
Hypsometric Analyses .................................................................................... 104
5.3.4
Volumetric Analyses ....................................................................................... 104
5.3.5
Determination of Wind Effects on Water Level ............................................... 104
5.3.6
Methodological Limitations.............................................................................. 105
5.3.7
Linking to the Hydrodynamic Model................................................................ 105
5.4.
Results ................................................................................................................ 106
5.4.1
Surface Modelling ........................................................................................... 106
5.4.2
Mudflat Parameters......................................................................................... 107
5.4.3
Hypsometric Curves for Mudflats.................................................................... 109
5.4.4
Wind Effects on Water Level in the South Lagoon ......................................... 110
5.5.
Discussion ........................................................................................................... 111
5.6.
Summary, Conclusions & Management Implications.......................................... 114
5.7.
References .......................................................................................................... 115
5.8.
Appendices.......................................................................................................... 117
Appendix 5.1a:Surface Validation Results showing fit between predcited surface and
measured surface for each reference site. ................................................................... 117
Appendix 5.1b: Cross validation results for the thin plate spline models at each site.. 120
Appendix 5.2: High resolution surface models for each reference site. ....................... 121
Appendix 5.3: General mudflat slope parameters at each reference site .................... 124
Appendix 5.4: Hypsometric Curves for the Reference Sites (cummulative area in m2)126
6.
Spatial modelling of mudflat availability and fish habitat in the Coorong131
6.1.
6.2.
6.3.
6.3.1
6.3.2
6.4.
6.4.1
6.4.2
6.4.3
6.4.4
6.4.5
6.5.
6.6.1
6.6.2
6.6.3
Executive Summary ............................................................................................ 131
Introduction.......................................................................................................... 132
Methods............................................................................................................... 134
Datasets .......................................................................................................... 134
Spatial Model Development ............................................................................ 136
Results ................................................................................................................ 143
Spatial and temporal variation in water level and salinity in the Coorong ...... 143
Mudflat availability at the reference sites........................................................ 147
Modelling fish habitats using logistic regression............................................. 156
Validation of the models.................................................................................. 158
Habitat prediction for the key fish species ...................................................... 159
Discussion ........................................................................................................... 165
Water level and salinity variations in the Coorong.......................................... 165
Mudflat availability at the reference sites........................................................ 166
Habitat modelling and prediction for key fish species..................................... 167
The CLLAMM Dynamic Habitat
iv
6.6.4
Barrage outflow, water level and salinity in the Coorong................................ 168
6.6.
Summary, Conclusions & Management Implications.......................................... 169
6.7.
References .......................................................................................................... 170
6.8.
Appendices.......................................................................................................... 173
Appendix 6.1: Script used for modelling mudflat habitat. ............................................. 173
Appendix 6.2: Script used for modelling fish habitat. ................................................... 176
Appendix 6.4: Hourly salinity predictions for January and July of 1976, 1988, 1993 and
2005 at Barker Knoll, Noonameena and Salt Creek. ................................................... 184
Appendix 6.5: Mudflat availability on the day with maximum and minimum mean water
level for January and July of wet and dry years at Barker Knoll, Noonameena and Salt
Creek. ........................................................................................................................... 189
Appendix 6.6: Predicted salinities in July 1976, July 1988 and January 2005............. 201
Appendix 6.7: Barrage flow into the Coorong between 1960 and 2008 ...................... 202
The CLLAMM Dynamic Habitat
v
Acknowledgements
This research was supported by the CSIRO Flagship Collaboration Fund and
represents a collaboration between CSIRO, the University of Adelaide, Flinders
University and SARDI Aquatic Sciences.
We also acknowledge the contribution of several other funding agencies to the
CLLAMM program and the CLLAMMecology Research Cluster, including Land and
Water Australia, the Fisheries Research and Development Corporation, SA Water,
the Murray Darling Basin Commission’s (now the Murrray-Darling Basin Authority)
Living Murray program and the SA Murray-Darling Basin Natural Resources
Management Board. Other research partners include Geoscience Australia, the WA
Centre for Water Research, and the Flinders Research Centre for Coastal and
Catchment Environments. The objectives of this program have been endorsed by
the SA Department for Environment and Heritage, SA Department of Water, Land
and Biodiversity Conservation, SA Murray-Darling Basin NRM Board and MurrayDarling Basin Commission.
We are grateful to Russel Seaman and M. Miles (South Australian Department for
Environment and Heritage) for providing aerial photos, bathymetry and habitat maps
for the Coorong and regions, and Yuki Tunn (Map Land, DEH) for providing the
SPOT5 imagery. Thanks to Maylene Loo, Brian Deegan, Alec Rolston and Rebecca
Lester for sharing their knowledge about the Coorong. We would also like to thank
Pramod Aryal for help with writing model scripts.
At SARDI Aquatic Sciences, we wish to thank Bruce Miller-Smith, Mandee Theil,
Sonja Venema, Michelle Roberts, Gen Mount, Jason Nichols and Kathryn Wiltshire
for sample collection, preparation and analysis. Qifeng Ye, Craig Noell and David
Short are acknowledged for providing access to fish catch data and water quality
data. We also wish to thank Sabine Dittmann (Flinders University) for access to the
Mastersizer particle size analyser. Paul van Ruth, Kathryn Wiltshire, Simon
Goldsworthy, Keith Rowling, Nathan Bott, Brad Page and Sasi Nayer of SARDI,
Aquatic Sciences; Daniel Rogers and Bertram Ostendorf of Adelaide University,
Rebecca Lester and Peter Fairweather of Flinders University and Ian Overton of
CSIRO are thanked for reviewing components of this report and providing valuable
comments and suggestions.
The CLLAMM Dynamic Habitat
vi
Executive Summary
The Coorong is a 2-3 km wide lagoon running parallel to the coast for approximately
100 km from the Murray Mouth. It provides an important habitat for a wide range of
bird species, and in particular has been listed as a Ramsar wetland due to its
exceptional importance for migratory waterbirds. Over 85 species of water birds and
waders, and 68 fish species, have been reported in the region. In large part, the
regions uniqueness is due to the diversity of habitats located in a small area, in
particular the salinity gradient, which under natural conditions ranges from almost
fresh to estuarine through to hypersaline. Understanding the distribution, spatial
extent and connectivity of these habitats is crucial for successful conservation and
management of the region, particularly under the severely altered flow regimes
currently being experienced, which have resulted in lower than normal water levels
and extreme hypersalinity in the South Lagoon.
Regional habitat maps
As part of the Coorong, Lower Lakes and Murray Mouth Research Cluster
(CLLAMMecology), we developed comprehensive maps of a variety of habitat
attributes in the Coorong in the early to mid 2000’s. This mapping included
surrounding vegetation assemblages and wetlands, subtidal sediment
characteristics, and bathymetry. We used a variety of data sources, including preexisting mapping conducted by the South Australian Department for Environment
and Heritage, aerial photography, satellite imagery, in situ field surveys of benthic
habitats and sediments, and topographic surveys. In situ sampling focussed on the
12 reference sites selected for intensive study by the broader CLLAMMecology
research team, producing detailed habitat maps of an area ~2 km wide extending
across the lagoon from the eastern to the western shore. However, mapping efforts
based on existing data sources and remote sensing also produced maps of the entire
Coorong and its surrounds.
At a broad scale, a total of 5,755 ha was mapped using remote sensing, with about
half the mapped area in the lagoon (49.9%), 34.7% on the peninsula, 13% on the
mainland, and less than three percent on islands. The terrestrial and wetland habitat
categories covered 34.9 and 65.1 percent, respectively, of the mapped area. A total
of 19 landforms and 17 wetland types were identified in the reference sites. Twenty
four wetland habitats and 14 terrestrial habitat types were reported from these
reference sites. The Coorong and surrounding region were classified into ten broad
habitat categories including agricultural and pastoral land. The habitat maps
generated through the current study provide a reference for the available habitats in
the area, and baseline information for making future predictions on habitats based on
the major ecological drivers of the system: water level and salinity. They also
facilitate estimations on the likely occurrence of species or ecosystems in the region,
and can be used to help determine the likely ecological effects of management
decisions on water flow over the barrages.
Coorong Digital Elevation Model
A range of data sources were also combined with predictive modelling to develop an
accurate and comprehensive Digital Elevation Model (DEM) of the Coorong and
surrounds. While a relatively good data set on bathymetry is available from the
South Australian Water Corporation for the North Lagoon, and there is an existing
DEM for terrestrial areas of South Australia, which includes the area surrounding the
Coorong, there is limited bathymetric data available for the South Lagoon. The few
field data on depth in the South Lagoon were used to develop a predictive model of
bathymetry using remote sensing data, in order to generate a comprehensive
bathymetry, although the predictive ability of the model was poor for regions south of
Jack Point, where turbidity was high. This and the bathymetry data from the North
The CLLAMM Dynamic Habitat
vii
Lagoon, as well as the SA regional DEM, were then merged into a seamless DEM of
bathymetry and topography for the Coorong, allowing prediction of the extent of
different habitats based on depth at different water levels.
Benthic habitat maps
We also generated sediment and benthic habitat maps for the Coorong. Sediment
characteristics were obtained at a number of points at each of the sites from field
sampling, and modelling was then used to predict characteristics along the two
lagoons, based on distance from the Murray Mouth, the underwater topography,
distance to shore, and also salinity, amongst other variables. We found three main
depositional areas along the Coorong, where sediments are fine and organicallyenriched: (1) the middle channel of the lagoons, (2) the constriction between the
North and South Lagoons, and (3) the western (seaward) shore of the North Lagoon,
particularly south of Long Point.
Mudflats
The status of the Coorong wetlands as migratory bird habitat has been due primarily
to the opportunities they provide for large numbers of birds to feed in a highly
productive estuarine and lagoonal environment. A significant proportion of this
foraging occurs on the large tracts of mudflat found throughout the Coorong. The
productivity of the mudflats varies along the length of the Coorong, dependent
primarily on water quality (particularly salinity), nutrient inputs, sedimentary structure,
and the duration, frequency and extent of inundation. Resident macroinvertebrate
populations and aquatic vegetation such Ruppia species are both an indicator of
productivity in the mudflats and a food source for fish and birds.
High resolution topographic/bathymetric models for the 12 CLLAMMecology
reference sites, confirm the importance of the South Lagoon in terms of mudflat
habitat, as it contains some 61% of available mudflat, as measured in the reference
sites. All mudflats should be highly productive if the necessary physical, chemical
and biological conditions existed. Across all 12 reference sites the 0 m to 0.5 mAHD
elevation range is most significant as it contains approximately 43% of all available
mudflat area. The second most important elevation class is -0.5 m to 0 mAHD,
containing approximately 40% of total available mudflat area. This suggests that
manipulations of water level should be kept within this range. Ideally, the most
important elevation range is 0.2 m to 0.4 mAHD, as manipulations in this range
accomplish wetting and drying of the maximum area of mudflat. If water levels can
be maintained at close to optimal levels, then natural high-frequency, wind-driven
oscillations in water levels will inundate large areas of mudflat.
Dynamic habitat model
By combining the bathymetry data described above with the outputs of a CSIRO
hydrodynamic model detailing hourly water level and salinity along the Coorong for a
given flow scenario, we were able to develop dynamic habitat models for key bird
and fish species. The first spatial model predicts the location and extent of mudflats,
defined as soft sediment areas that are either emersed or covered by no more than
12 cm of water, where most waterbird foraging occurs. Mudflat availability varied
spatially and temporarily along the Coorong, and is influenced by tide, wind, rainfall
and evaporation, some of which are dependent on the distance from the Murray
Mouth and are affected by seasonal variation. The modelling of mudflats at different
water levels suggests that an average water level of 0.12 mAHD gives the maximum
average mudflat area in the three reference sites examined in detail (Barker Knoll,
Noonameena and Salt Creek), with the majority of the mudflats located on the
eastern shores. Out of seven key fish species examined, four (Yelloweye Mullet,
Smallmouth Hardyhead, Greenback Flounder and Tamar River Goby), demonstrated
a significant relationship with salinity. Among the three different salinity gradient
scenarios examined, a salinity range from 5 to 90 g/L along the Lagoon was found to
be the best in terms of the suitability of the entire Lagoon for these four key species,
The CLLAMM Dynamic Habitat
viii
as well as for supporting other important biological communities including both
macrophytes and infauna. The extent of habitat available for each of these species
depends on the salinity gradient present, which varies considerably over time.
Potential applications for the design of environmental flow strategies for the Coorong
This study has developed a set of tools that will help managers design better
intervention strategies for the rehabilitation of the Coorong region. These tools allow
managers for the first time to explicitly incorporate spatial considerations in the
design of environmental flow strategies for the Coorong, such as where mudflats will
be located under different management interventions. Similarly, it is now possible to
quantify the habitat that could be gained or lost under different management
intervention for a range of key wading bird and estuarine fish species. Another
important legacy of the project is to have collated and synthesised a large amount of
spatial information about the Coorong region, which can now be used for a variety of
other purposes in addition to the design of environmental flow strategies. Of
particular importance, is the finding that the South Lagoon appears to have
substantially more mudflat habitat than the North Lagoon, making it more important
that management startegies address the need to maintain such habitats in this region
in a productive state.
The CLLAMM Dynamic Habitat
ix
1. Introduction
The Coorong is a 2-3 km wide lagoon, separated from the Southern Ocean by a
barrier dune, and running parallel to the coast for approximately 100 km from the
Murray Mouth. Historically, it was permanently connected to both the freshwater
Lower Lakes and the ocean, however, in recent times these connections have been
threatened and even severed. In the 1940’s, a series of barrages were constructed
between the Lower Lakes and the Murray estuary and Coorong, in order to prevent
saltwater incursion into freshwaters used for human consumption and irrigation.
While freshwater continued to flow over the barrages, increasing abstraction
upstream and reduced rainfall has meant that these flows have declined
precipitously, and have frequently stopped, especially since the mid 1990’s (e.g.
Geddes 2005). As a consequence, less freshwater has been available to the
Coorong and Murray estuary. This reduction has resulted in closure of the Murray
Mouth, which is now only kept open by dredging, reduced water levels in the
Coorong, and increased salinity, to the point where the southern end of the Coorong
now has salinities in excess of six times those of seawater (e.g. Noell et al. 2009).
These physical changes in the Coorong have had major implications for the flora and
fauna that help make the Coorong a unique environmental asset. Historically, it has
provided an important habitat for a wide range of bird species, and it is listed as a
Ramsar wetland due to its exceptional importance for migratory waterbirds, which
were estimated to number in excess of 234,543 in 1982 (Wilson 2001). Over 85
species of water birds and waders, and 68 fish species, have been reported in the
region. These, in turn, rely on a rich assemblage of macro-invertebrates, as well as
extensive stands of macrophytes, to provide them with the abundant food resources
that make the region so attractive to them. However, increasing salinity has meant
that many of these food resources have declined greatly in abundance (e.g. Rolston
and Dittman 2009), and as a consequence, bird numbers have also declined (e.g.
Rogers and Paton 2009). Fish also are suffering from reduced food availability, as
well as directly from increased salinity, which is beyond their osmoregulatory ability to
cope with (e.g. Noell et al. 2009).
With limited freshwater resources available, it is necessary to optimise its use for
generating environmental flows and to understand what other management
interventions could be considered to maintain as much of the biological diversity and
ecological integrity of the Coorong as possible. As such, the Coorong Lower Lakes
and Murray Mouth research cluster (CLLAMMecology), has conducted an extensive
research project to examine the habitat characteristics, assemblages of key species
including birds, fish and macro-invertebrates, productivity and biogeochemical cycling
of the Coorong. The overall goal is to develop an integrated ecosystem model that
can be used to predict the future ecological state of the Coorong under a range of
defined flow scenarios.
In this report, we develop a dynamic GIS-based habitat model of the Coorong, that
can be used to predict the availability of habitats for key bird and fish species under
any pre-defined flow scenario. This model uses output from the CSIRO 1dimensional hydrodynamic model of the Coorong, which produces hourly outputs of
water level and salinity along the length of the two lagoons (Webster 2007). Both
models also utilise a seamless digital elevation model that we developed from a
number of data sources, and describe here. The model for birds predicts the extent
of mudflats available at different water depths in each of 12 reference sites that are
the subject of detailed study by the broader CLLAMMecology team, although only
results for 3 sites (Barker Knoll in the north, Noonameena mid-way down the North
Lagoon, and Salt Creek at the southern end of the South Lagoon) are presented for
brevity. It also builds upon more detailed mapping of topography at each of the
reference sites, as well as sediment mapping. The model for fish primarily links the
The CLLAMM Dynamic Habitat
1
probability of occurrence for a range of species to salinity, based on data collected by
Noell et al. (2009).
Chapter 2 of this report builds on existing work by the South Australian Department
for Environment and Heritage (Seaman 2003) to produce a broad-scale habitat map
of the area surrounding the Coorong. This exercise focuses on mapping different
vegetation and wetland types, using a combination of remote sensing data and field
verification. Chapter 3 develops a single seamless digital elevation model of the
region from a number of data sources, and uses predictive modelling to fill the
existing data gaps. This model underpins the dynamic habitat model, as a
knowledge of bathymetry is required to determine the extent of mudflat available at
different water levels. Chapter 4 then focuses on fine scale mapping of benthic and
near-shore habitats and sediment characteristics at each of the 12 reference sites.
Chapter 5 examines the geomorphology of the mudflats at these reference sites in
more detail, allowing centimetre scale prediction of water depth in the vertical scale
at different water levels. This level of accuracy is important, as most wading birds
have a very specific foraging niche with respect to water depth, with few foraging in
water > 20 cm deep (Rogers and Paton 2009). Finally, Chapter 6 presents the
dynamic habitat model, and details how habitat availability changes under different
conditions using a flow scenario based on actual water inputs to the Murray Darling
Basin over the last 112 years.
References
Geddes, M.C. (2005) The ecological health of the north and south lagoons of the
Coorong in July 2004. SARDI Aquatic Sciences and University of Adelaide.
Noell, C., Ye, Q., Short, D.A., Bucarter, L.B. and Wellman, N.R. (2009) Fish
assemblages of the Murray Mouth and Coorong region, South Australia, during an
extended drought period. CSIRO: Water for a Healthy Country National Research
Flagship and South Australian Research and Development Institute (Aquatic
Sciences), Adelaide.
Rogers, D.J. and Paton, D.C. (2008) Spatiotemporal Variation in the Waterbird
Communities of the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship, Canberra.
Rolston, A.N. and Dittmann, S. (2009) CLLAMMecology Invertebrate Key Species
Project: The distribution and abundance of macrobenthic invertebrates in the Murray
Mouth and Coorong Lagoons. CSIRO: Water for a Healthy Country National
Research Flagship, Canberra.
Seaman, R.L. (2003) Coorong and Lower Lakes habitat-mapping program. South
Australian Department for Environment and Heritage, Adelaide.
Webster, I.T. (2007) Hydrodynamic modelling of the Coorong. Water for a Healthy
Country National Research Flagship, CSIRO, Canberra.
Wilson, R. J. (2001) Wader surveys in the Coorong, South Australia in January and
February 2001. The Stilt, 40: 38-54.
The CLLAMM Dynamic Habitat
2
2. Habitat Mapping of the Coorong and Surrounds
Sunil K. Sharma1*, Jason E. Tanner1 and Simon N. Benger2
1
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
2
School of Geography, Population and Environmental Management, Flinders
University, GPO Box 2100, Adelaide, SA 5001
*corresponding author, Phone +61 (8) 8207 5448, Fax +61 (8) 8207 5448, E-mail:
[email protected]
2.1.
Executive Summary
The Coorong Lagoon and surrounding areas provide exceptional habitat for a wide
range of bird and fish species. Over 68 fish species and 85 species of water birds
and waders have been reported in the region. Unique biological diversity has
rendered the region as one of high conservation value, which has been recognised
through its listing as a wetland of national, as well as international, importance.
Understanding the distribution, spatial extent and connectivity of habitats has been
widely accepted as a crucial step for successful conservation and management of
biological diversity. Habitat mapping has become a popular tool for identification,
characterisation and visualisation of different habitats. Habitat maps are also very
useful as a monitoring tool for detecting change in habitats and evaluating human
impacts or the effects of climate change, and for assessing the ecological values of
the region.
The South Australian Department for Environment and Heritage (DEH) developed a
GIS based habitat map for the areas surrounding the Coorong and Lower Lakes in
2003, mainly focused on the state of the wetlands. This map does not include benthic
habitats along the Coorong or characterisation of physical environmental parameters.
In order to complement the DEH habitat map, the Dynamic Habitat Project under the
CLLAMMecology Research Cluster aimed to map the benthic habitats and areas
surrounding the Coorong. This habitat mapping exercise focussed primarily on the 12
representative sites across the Coorong selected for detailed study by the
CLLAMMecology Research cluster, and this information was extrapolated to map
habitats across the entire region through classification of LANDSAT5 Thematic
Mapper imagery.
A habitat classification scheme with 12 attributes encompassing a range of
geographical, physical, chemical and biological characteristics for each habitat was
used. Apart from drawing on and improving information derived from existing habitat
maps for the Coorong and Lower Lakes, the current habitat mapping program
incorporated a range of other data sources. These included remotely sensed imagery
such as satellite data and high resolution aerial photography, and field data collected
as part of the CLLAMM project.
A total of 5,755 ha was mapped, with about half the mapped area in the lagoon
(49.9%), 34.7% on the peninsula, 13% on the mainland, and less than three percent
on islands. The terrestrial and wetland habitat categories covered 34.9 and 65.1
percent, respectively, of the mapped area. A total of 19 landforms and 17 wetland
types were identified in the reference sites. Twenty four wetland habitats and 14
terrestrial habitat types were reported from these reference sites. The Coorong and
The CLLAMM Dynamic Habitat
3
surrounding region were classified into ten broad habitat categories including
agricultural and pastoral land.
The habitat maps generated through the current study provide a reference for the
available habitats in the area, and baseline information for making future predictions
on habitats based on the major ecological drivers of the system: water level and
salinity. They also facilitate estimations on the likely occurrence of species or
ecosystems in the region, and can be used to help determine the likely ecological
effects of management decisions on water flow over the barrages.
2.2.
Introduction
The Coorong Lagoon and surrounding areas provide exceptional habitat for a wide
range of bird and fish species. In particular, the lagoon has been an indispensable
habitat for many fish species for completing their life cycles as well as one of the
most visited destinations for migratory shorebirds in Australia (Edyvane 1999). A total
of 85 waterbirds species (Carpenter 1995) and 68 species of marine, estuarine and
freshwater fish (Department for Environment and Heritage 2000) have been reported
in the region. Unique biological diversity has rendered the region as one of high
conservation value, which has been recognised through its listing as a wetland of
national, as well as international, importance (Seaman 2003; Australian Nature
Conservation Agency 1996)
A strong spatial gradient of habitat types exists along the length of the Coorong,
defined by variations in physical environment and processes. Each habitat
constitutes a unique combination of environmental conditions and interactions, which
render the area suitable for habitation by different species and biological
communities. In a general sense, habitats are distinguished by their physical,
biological and chemical characteristics (Kostylev et al. 2001) and offer suitable areas
for a species or group of species. These habitat variables are often viewed as
environmental/ecological variables.
Understanding the distribution, spatial extent and connectivity of habitats has been
widely accepted as a crucial step for successful conservation and management of
biological diversity (Bowers 2008; Chong 2007). Habitat mapping has become a
popular tool for identification, characterisation and visualisation of different habitats
(Chust 2008; Lathrop 2006; Urbanski 2003). Habitat mapping aims to provide
baseline information about the physical and biological features present in an area,
and assists understanding of the interactions between them. Habitat maps are also
very useful as a monitoring tool for detecting change in these features and evaluating
human impacts or the effects of climate change, and for assessing the ecological
values of the region.
The South Australian Department for Environment and Heritage (DEH) developed a
GIS based habitat map for the areas surrounding the Coorong and Lower Lakes in
2003, mainly focused on the state of the wetlands, and mapped primarily through
aerial photo interpretation (Seaman 2003). The mapping defines the spatial
arrangement of primary vegetation assemblages and employs the standard DEH
classifications for South Australia. However, this map does not include benthic
habitats along the Coorong or characterisation of physical environmental parameters.
In order to complement the DEH habitat map, the Dynamic Habitat Project under the
CLLAMMecology Research Cluster aimed to map the benthic habitats and areas
surrounding the Coorong. The current habitat mapping exercise focussed primarily
on the 12 representative sites across the Coorong selected for detailed study by the
cluster (Figure 2.1), and this information was extrapolated to map habitats across the
entire region through classification of LANDSAT5 Thematic Mapper imagery.
The CLLAMM Dynamic Habitat
4
The habitat maps generated through the current study provide a reference for the
available habitats in the area, and baseline information for making future predictions
on habitats based on the major ecological drivers of the system: water level and
salinity. They also facilitate estimations on the likely occurrence of species or
ecosystems in the region, and can be used to help determine the likely ecological
effects of management decisions, such as how much water to release over the
barrages and when.
2.3.
Reference sites for habitat mapping in the Coorong
Twelve reference sites were identified between the Goolwa barrage in the North
Lagoon and Salt Creek in the South Lagoon, representing habitats across the system
(Figure 2.1). In addition to sites in the Goolwa and Mundoo channels, ten sites were
located along the main channel of the Coorong extending south from the Murrary
Mouth. The site at Goolwa Channel is placed between the Goolwa barrage and the
Murray Mouth. Another site was established in Mundoo Channel about one kilometre
west of the Mundoo barrage. Of the ten sites in the Coorong itself, six were in the
North Lagoon between the Murray Mouth and Noonameena. These were: Barker
Knoll, Ewe Island, Pelican Point, Mark Point, Long Point and Noonameena. The
remaining four sites were located in the South Lagoon between Parnka Point and
Salt Creek: Parnka Point, Villa dei Yumpa, Jack Point and Salt Creek. These sites
were selected by the CLLAMMecology research cluster as being representative of
the spatial gradient of habitats available in the lagoon, as well as suitable for
sampling for the range of projects being undertaken by the cluster.
Each reference site was defined as a band approximately 1.5 km wide (except for the
Mundoo Channel where the channel is narrow and bends sharply), centred on the
actual site selected by the CLLAMMecology research cluster. As well as examining
the aquatic habitats, adjacent terrestrial habitats were also mapped, with the sites
being delineated by a road or natural landscape element on the eastern boundary
and the Southern Ocean coast on the western boundary.
The CLLAMM Dynamic Habitat
5
Figure 2.1. Locations of the reference sites in the Coorong Lagoon: 1 = Goolwa; 2 = Mundoo;
3 = Barker Knoll; 4 = Ewe Island; 5 = Pelican Point; 6 = Mark Point; 7 = Long Point; 8 =
Noonameena (in the North Lagoon); 9 = Parnka Point; 10 = Villa dei Yumpa; 11 = Jack Point
and 12 = Salt Creek (in the South Lagoon). The Australian map (A) shows the state
boundaries and the study area in South Australia (SA) while the inset map (B) highlights the
study area shown in the main map.
2.4.
Habitat classification scheme
Previous habitat maps developed by DEH were based on 24 biological, physical and
environmental attributes (Seaman 2003). Although the habitat classification system
was very detailed and comprehensive, most of the attributes were missing for many
habitat units. To minimize the missing data in the database and also to incorporate
both wetlands and terrestrial habitats into the one habitat map, a habitat classification
scheme with 12 attributes, encompassing a range of geographical, physical,
chemical and biological characteristics for each habitat was used. These habitat
attributes are described briefly in the following section.
2.4.1 Habitat zone
The Coorong and surrounds include mainland, island, peninsula and lagoon systems,
and contain a huge diversity of wetland, terrestrial and benthic habitats within a small
geographical domain. The zoning of habitats specifies their location within these
landscapes. The mainland zone extends from the eastern shore of the lagoon inland.
The CLLAMM Dynamic Habitat
6
Those areas that are separated from the mainland and form island habitats, such as
Ewe Island and Hindmarsh Island, are found in the areas between the lakes and the
lagoon. The lagoon includes the main channel and the shores delineated by the high
water mark from the Murray Mouth to the South Lagoon, Goolwa Channel, Mundoo
Channel and creeks flowing into the channel. The peninsula zone is the landmass
separating the lagoon system and the Southern Ocean. The Murray Mouth divides
the peninsula into two with the northern portion from Goolwa to the Murray Mouth
being the Sir Richard Peninsula, while to the south the Younghusband Peninsula
extends from the Murray Mouth to the end of the South Lagoon.
2.4.2 Habitat category
Habitats in the mainland and peninsula zones belong to two broad categories:
terrestrial and wetland. These habitat categories are characterised by completely
different ecosystems with different biological components and physical processes. In
the Coorong, the prolonged drought of recent years may have rendered many areas
unsuitable for classification as wetland compared to wetter times. Although wetland
systems may include areas which are dry more often than wet (Aber 2007),
connectivity to the lagoon or another water source, and substrate type, were taken
into account when deciding whether an area should be categorised as a wetland or
not. All areas within the lagoon system were identified as wetland systems. The
ephemeral mudflats around the shores, and inland salt pans, were categorized as
wetland habitat. Wetlands were further subdivided into wetland types, as described
in section 2.4.4 below.
2.4.3 Landform
After identifying the habitat category, the habitats were classified into different
landforms through examination of the dominant pattern of the land surface.
Altogether 32 landforms were identified during the DEH habitat mapping of the
Coorong and Lower Lakes (Seaman 2003). However, the list included ‘lagoon’ and
‘island’ as types of landform, which we have already used for defining habitat zones.
To avoid repetition of the term, channel was redefined to designate the main body of
the lagoon under water. Similarly, mudflat was not used as a category of landform
rather we used this term to describe one of the wetland habitat types. A brief
description of the landforms reported in the Coorong by Seaman (2003) is given in
Appendix 2.1. In addition to these landforms, coastal swale and dune have been
introduced to designate the undulating area adjacent to the shores and open dune
without vegetative cover, respectively.
2.4.4 Wetland system
The wetland habitats were further classified into five major wetland systems based
on the Cowardin system (Cowardin et al. 1979) as adopted by Seaman (2003) for the
DEH habitat mapping. These wetland systems are: Marine, Estuarine, Riverine,
Lacustrine and Palustrine (Aber 2007) (Table 2.1).
The CLLAMM Dynamic Habitat
7
Table 2.1. Wetland system classification based on the Cowardin system (from Aber 2007).
Wetland System
Descriptions
Marine
Open ocean, continental shelf, including beaches, rocky shores, lagoons,
and shallow coral reefs. Normal marine salinity to hypersaline water
chemistry; minimal influence from rivers or estuaries. Where wave energy
is low, mangroves, mudflats or sabkhas may be present.
Estuarine
Deepwater tidal habitats with a range of fresh-brackish-marine water
chemistry and daily tidal cycles. Salt and brackish marshes, intertidal
mudflats, mangrove swamps, bays, sounds, and coastal rivers. Drowned
coasts, where supply of river sediment is insufficient to infill estuary basin.
Riverine
Freshwater, perennial streams comprised of the deepwater habitat
contained within a channel. This restrictive system excludes floodplains
adjacent to the channel as well as habitats with more than 0.5 g L-1
salinity.
Lacustrine
This system includes inland water bodies that are situated in topographic
depressions, lack emergent trees and shrubs, have less than 30%
vegetation cover, and occupy at least 8 ha. Includes lakes, larger ponds,
sloughs, lochs, bayous, etc.
Palustrine
All non-tidal wetlands that are substantially covered with emergent
vegetation - trees, shrubs, moss, etc. Most bogs, swamps, floodplains and
marshes fall in this system, which also includes small bodies of open water
(< 8 ha), as well as playas, mudflats and salt pans that may be devoid of
vegetation much of the time. Water chemistry is normally fresh but may
range to brackish and saline in semiarid and arid climates.
2.4.5 Water regime
In the wetland habitats, the water regime plays a key role in determining the
biological communities present, and also influences physical and chemical
processes. The DEH habitat mapping for the Coorong and Lower Lakes included
water regime and tidal class as separate wetland habitat attributes adopted from
Blackman et al. (1992) (Seaman 2003). The water regime characterises the
frequency of flooding, and the tidal classes indicate the frequency and influence of
oceanic tides upon the habitats (Blackman et al. 1992). The wetlands in the lagoons
are not subjected to tidal influences except for the areas around the Murray Mouth,
whereas the wetlands in the mainland and island zones are non-tidal and are
subjected to flooding events. Hence, both water regime and tidal class were
attributed under water regime in the habitat classification scheme used here. The
frequencies of flooding events were recorded as semi-permanent, intermittent and
seasonal. Within those areas subjected to tidal influences, subtidal signifies the
wetland is under water all the time while intertidal signifies wetland areas located
between the low and the high water mark.
2.4.6 Wetland type
The Directory of Important Wetlands in Australia (Environment Australia 2001) used
40 different wetland types to classify wetlands in Australia based on the Ramsar
Convention under Article 1.1. This wetland classification was accepted by the
Australia and New Zealand Environment and Conservation Council (ANZECC)
Wetland Network in 1994 and has not been changed since (Environment Australia
2001). Under this scheme, wetlands are first classified into three major categories:
marine and coastal, inland, and human made, with 12, 19 and 9 wetland types,
respectively. Seaman (2003) added three more wetland types to the marine and
coastal wetland category to include coastal dune shrubland, freshwater soak and
estuarine stream channel, and one to the inland category for freshwater/brackish
The CLLAMM Dynamic Habitat
8
mud or sand flats, during mapping of wetlands in the Coorong and Lower Lakes. All
these wetland types are listed in Table 2.2. These wetland types are identified by
water regime, salinity, substrate type and vegetation type. The Coorong and Lower
Lakes encompasses wetlands belonging to all three wetland categories. The
wetlands along the lagoon system itself are all classified as marine and coastal
wetlands.
Table 2.2. Wetland types for the Coorong and Lower Lakes adopted from Environment
Australia (2001) and Seaman (2003).
Inland
Human Made
B1 Permanent rivers and streams + waterfalls
C1 Water storage areas, reservoirs, barrages,
impoundment (>8ha)
B2 Seasonal irregular river and streams
C2 Ponds, farm, stock, tanks (<8ha)
B3 Inland deltas (permanent)
C3 Aquaculture
B4 Riverine floodplains
C4 Salt pans
B5 Permanent freshwater lakes (8ha, includes oxbow
lakes)
C5 Excavations, gravel pits, borrow pits, mining
B6 Seasonal/intermittent freshwater lakes (>8ha),
floodplain lakes
C6 Waste water treatment, settling ponds
B7 Permanent saline/brackish lakes
C7 Irrigated land, canals, ditches
B8 Seasonal/intermittent saline lakes
C8 Seasonally flooded arable land, farm land
B9 Permanent freshwater ponds (<8 ha), marshes and
swamps on inorganic soils, with emergent veg.
Waterlogged for at least most of the growing season.
Includes coves and open water enclosed with reeds
C9 Canals
B10 Seasonal/ intermittent freshwater ponds and marshes
on inorganic soils includes potholes, seasonally flooded
meadows, sedge marshes. Includes reed shorelines.
(Constructed/Artificial Wetlands?)
Marine and Coastal Zone Wetlands
B11 Permanent saline/brackish marshes
A1 Marine waters-permanent shallow waters less than six
metres deep at low tide, includes sea bays and straits
B12 Seasonal saline marshes
A2 Sub tidal aquatic beds, includes kelp beds,
seagrasses, tropical marine meadows
B13 Shrub swamps, shrub dominated freshwater marsh,
sedges and Gahnia sedgeland
A3 Coral reefs
B14 Freshwater swamp forest, seasonally flooded forest,
wooded swamps
A4 Rocky marine shores, includes rocky offshore islands,
sea cliffs. Rocky estuarine shores.
B15 Peatlands, forest, shrub or open bogs
A5 Sand, shingle or pebble beaches, includes sand bars,
spits, sandy islets
B16 Alpine
A6 Estuarine waters, permanent waters of estuaries and
estuarine systems of deltas
B17 Freshwater springs, rock pools
A7 Intertidal mud, sand or salt flats and algae.
B18 Geothermal wetlands
A8 Intertidal marshes, including saltmarshes, salt
meadows, saltings, raised salt marshes, tidal brackish and
freshwater marshes and vegetated shorelines.
B19 Inland Karst.
A9 Intertidal forested wetlands, includes mangrove
swamps, nipa swamps, tidal freshwater swamp forest
B20 Freshwater/brackish mud or sand flats.
A10 Brackish to saline lagoons and marshes with one or
more relatively narrow connections to the sea
A11 Freshwater lagoons and marshes in the coastal zone.
Reedbeds and vegetated bed sediments.
A12 Non tidal freshwater forested wetlands
A13 Coastal dune shrubland
A14 Freshwater soaks <.8ha within the coastal zone
A15 Estuarine stream channel
The CLLAMM Dynamic Habitat
9
2.4.7 Habitat type
The lagoon is comprised exclusively of wetland habitats, while the mainland and
peninsula zones contain a mix of terrestrial and wetland habitats. Land cover
(vegetation, cover density and land condition) distinguishes different habitat types in
the terrestrial system, whereas the substrate and /or submerged vegetation types
primarily define the habitat types in the wetland and benthic systems. A list of habitat
types used in this habitat mapping is given in Table 2.3. Patchy bare mud has been
used to identify a wetland habitat type in the shallower areas adjacent to the main
channel which contains sparse Polychaetes and Ruppia.
Table 2.3. Habitat types in the Coorong.
Wetland Habitat types
Terrestrial Habitat types
Bare Mud
Dense Myoporum
Bare Mud with Polychaetes
Coastal Vegetation
Bare Mud with Ruppia
Open Coastal Vegetation
Bare Sand
Patchy Coastal Vegetation
Bare Sand and Grit
Dense Dune Vegetation
Bare Sand Mud
Dune Vegetation
Bare Sand Mud with Algae
Open Dune Vegetation
Bare Sand Mud with Ruppia
Patchy Dune Vegetation
Coastal Vegetation
Pasture
Dense Melaleuca
Patchy Pasture
Dense Samphire
Wet Grassland
Dense Sedge
Plantation
Dry Grassland
Bare Sand
Melaleuca
Unvegetated Sand Dune
Mud Flat
Open Coastal Vegetation
Open Samphire
Patchy Coastal Vegetation
Patchy Melaleuca
Patchy Bare Mud
Patchy Samphire
Rocky Outcrop
Samphire
Very Soft Mud
2.4.8 Salinity
The salinity of the Coorong is controlled by freshwater flow over the barrages and
from Salt Creek, and incursions of sea water from the Murray Mouth, as well as
evaporation rates and connectivity between the North and South Lagoons (Webster
2007). Due to extreme drought conditions, freshwater flow over the barrages has not
occurred since 2005. As a consequence, the salinity of the entire system has
The CLLAMM Dynamic Habitat
10
increased substantially. The salinity of the South Lagoon is currently above 120 g/L,
while salinity gradually decreases towards the Murray Mouth due to incursion of
marine water.
Salinity is considered to be one of the major ecological drivers in the Coorong, and
determines the quality of habitats for infauna and fish species. In the habitat
classification used here, salinity has been incorporated as a water quality parameter
that determines habitat condition in the subtidal and intertidal areas, as it impacts on
the distribution of infauna, Ruppia and birds.
2.4.9 Vegetation
Dominant floral species occurring within each habitat unit were reported. Plant
species present were identified from the habitat mapping database for the Coorong
and the Lower Lakes (Seaman 2003) and also from the Coastal Saltmarsh and
Mangrove Mapping database (Canty and Hille 2002). Dominant species were
recorded for each of the available habitat units. If the habitat units were outside of the
coverage area of the existing habitat mapping projects or information on flora or
fauna was missing, colour aerial photos from 2003 were interpreted in conjunction
with these habitat maps to assign flora to the habitat. For the benthic region of the
Lagoon, information from an underwater video survey was used (section 2.5.1).
2.4.10 Cover percent
Cover percent records the percentage of area covered by the vegetation. It gives a
relative indication of the abundance of the flora in a given habitat unit. Seaman
(2003) adopted the cover/abundance classes as prescribed by Heard and Chanon
(1997) for the Coorong and Lower Lakes DEH habitat mapping (Table 2.4). The
three lowest cover classes were semi-quantitative (few, sparse and plentiful), and so
we combined these into a single class of < 5% cover for consistency in the database.
Table 2.4. Cover or abundance classes for the Coorong habitat mapping (Modified from
Heard and Channon 1997).
Description
Limited cover, <5% area
Label
<5%
Any number of individuals, covering 5-25%
5-25%
Any number of individuals, covering 25-50%
25-50%
Any number of individuals, covering 50-75%
50-75%
Any number of individuals, covering more than 75%
>75%
2.4.11 Habitat condition
The habitat condition summarizes the overall environmental and biological status of
the habitat unit. It is subjectively assessed by taking into account the level of human
interference, ecological condition, invasive species and habitat connectivity. The
following six habitat condition categories were adopted from Seaman (2003) (Table
2.5).
The CLLAMM Dynamic Habitat
11
Table 2.5. Habitat condition classes adopted from Seaman (2003).
Habitat Condition
Description
Pristine
Pristine, or nearly so; no obvious signs of disturbance. Indigenous flora
dominant and abundant, 100 % ground cover. Structural diversity present
if applicable and microhabitats present. Surrounding ecosystems intact
with high connectivity. Habitat integrity is high. Reflects pre-European
vegetation or natural landscape feature.
Excellent
Vegetation structure intact, disturbance affecting individual species and
weeds are non- aggressive species limited to 5 - 20% coverage. Diverse
species, stable fauna habitat, structural diversity present, if applicable.
Habitat buffered by and linked to remnant vegetation with ecosystem
stability. Microhabitats present.
Very Good
Vegetation structure altered, indigenous and exotics together, 20-50%
weed invasion, obvious signs of disturbance (eg disturbance to
vegetation structure caused by repeated fires, the presence of some
more aggressive weeds, dieback and grazing). Core habitat areas exist
buffered by remnant vegetation. Obvious signs of use by fauna, areas of
structural diversity might exist with some microhabitats.
Good
Vegetation structure significantly altered by very obvious signs of multiple
disturbances. Retains basic vegetation structure or ability to regenerate it
(eg disturbance to vegetation structure caused by very frequent grazing).
Presence of aggressive weeds at high density (50 - 70%). Core habitat
areas exist that are buffered by scattered remnants. Species use of
habitats is likely to be opportunistic. Structural diversity limited to isolated
patches if at all, micro-habitat presence low.
Degraded
Basic vegetation structure severely impacted by disturbance. Scope for
regeneration but not to a state approaching good condition without
intensive management (eg disturbance to vegetation structure caused by
cropping, grazing or clearance; the presence of very aggressive weeds,
partial clearing, dieback and livestock grazing). Weed presence greater
than 70%. Habitats are impacted by disturbances and are not connected
with remnant buffers.
Completely degraded
The structure of the vegetation is no longer intact and the area is
completely or almost completely without native species. Habitats do not
exist, although areas might be used as opportunistic habitats or ‘stepping
stones’ to desirable habitat areas. Weed presence aggressive and
greater than 80%, monoculture can exist such as pasture.
Salinity level was considered for habitat condition in the intertidal and subtidal areas.
The following four habitat condition classes were used based on the cumulative
effect of the salinity level on macrophytes, infauna, fish and birds: very good (30-50
g/L), good (50-60 g/L), degraded (60-80 g/L) and completely degraded (>80 g/L)
(CLLAMM 2008). It should be recognized that the habitat conditions at most
reference sites would have been adversely and in some cases drastically affected by
recent years of drought and low water levels in the Coorong.
Bare sand in the intertidal area, sand bars or in closed depressions, and coastal or
dune vegetation with more than 50% cover were considered in good condition.
However, inland bare sand, open dune or coastal and open samphires were
considered in degraded condition. Exposed sand and patchy vegetation with
degraded land were categorised as being in a completely degraded condition.
2.4.12 Area
The area of each habitat unit was calculated in ArcGIS 9.3 (Environmental Systems
Research Institute 2008) and included in the habitat mapping database.
The CLLAMM Dynamic Habitat
12
2.5.
Methods
Apart from drawing on and improving information derived from existing habitat maps
for the Coorong and Lower Lakes (Seaman 2003; Department for Environment and
Heritage 2008), the current habitat mapping program incorporated a range of other
data sources. These included remotely sensed imagery such as satellite data and
high resolution aerial photography, and field data collected as part of the CLLAMM
project. Aerial photographs have been widely used for delineating both terrestrial and
benthic habitats and are usually complimented by ground verification (Barrett et al.
2001; Finkbeiner et al. 2001; Jordan et al. 2001). The current habitat mapping was
based on aerial photography taken in 2003, the most recent high resolution imagery
available for the region. The characterisation of each habitat unit and detailed
information on all of the attributes was obtained from a combination of data sources
including video survey, previous maps, expert knowledge of the area and field
observation. In addition to a detailed habitat mapping of the key sites, a general
habitat mapping of the entire region was undertaken using LANDSAT5 Thematic
Mapper imagery. The datasets and techniques used are described in the following
sections.
2.5.1 Available datasets
Habitat mapping for the Coorong and Lower Lakes from DEH
Habitat mapping of the Coorong and Lower Lakes (Figure 2.2.) was undertaken by
DEH in 2002-2003 in two stages (Seaman 2003). The habitats around the Lower
Lakes and creeks were classified in the first stage, which identified 518 habitat units
covering a total area of 24,400 hectares. The second stage included areas in the
Coorong National Park from the Murray Mouth to the southern Coorong National
Park near Kingston. One hundred and ninety five distinct habitat units were identified
covering 25,980 hectares in the second stage.
For the current study, we obtained the habitat map in the form of a shape file (ArcGIS
file format for vector data). The spatial coverage of the habitat map was determined
by collating and analysing existing spatial datasets at the 1:50,000 scale, and
interpretation of aerial photos at the 1:40,000 scale (Seaman 2003). We observed a
pronounced discrepancy between the DEH habitat maps and high resolution aerial
photos (0.5 m) from 2003 (Figure 2.3). These differences are probably due to a range
of factors including inconsistencies between data sources used to generate the DEH
maps, changes in habitat distribution since the original data was collected, and data
collection at different resolutions to that provided by the aerial photography.
Nonetheless, as described above, the DEH habitat map provided a wide range of
useful habitat information, which was extracted and augmented using other data
sources.
The CLLAMM Dynamic Habitat
13
Figure 2.2. Area covered by the
Coorong and Lower Lakes habitat
mapping project undertaken by
DEH.
Figure 2.3. Example of DEH habitat map from Ewe
Island on the peninsula side (red line) and the 2003
aerial photography, showing discrepancies between
the two.
Coastal Salt-marsh and Mangrove mapping
Mapping of coastal salt-marsh and mangrove was carried out to identify the
distribution, vegetation composition and ecological status of these habitat types
within South Australia (Department for Environment and Heritage 2008; Seaman
2003). A GIS layer containing the coastal salt marsh and mangrove map covering the
areas between the Murray Mouth and about five kilometres to the south of Salt Creek
was obtained from DEH (Figure 2.4). This layer had 1,860 mapped units, each
described by 23 attributes, including landform, estuarine and tidal class, vegetation
(cover), integrity, description, area and length. However, 69 of these mapping units
were outside the geographic scope of the project. This map also had very poor
matching to the aerial photos taken in 2003 (Figure 2.5), which is probably the result
of using a very high resolution for the photo or map interpretation, poor
orthorectification of the aerial photography used for the original mapping, and
changes in the distribution of some habitats. Nevertheless, the information on
vegetation type was highly useful for the current habitat mapping of the Coorong
region.
The CLLAMM Dynamic Habitat
14
Figure 2.4. Area covered by the
coastal salt-marsh and mangrove
mapping project in the Coorong
region.
Figure 2.5. Example of the coastal salt-marsh and
mangrove Map from Noonameena (red line) and the
2003 aerial photography, showing discrepancies
between the two.
Aerial photographs
Orthorectified aerial photographs are an important tool for habitat mapping (Kendall
et al. 2001). The most recent colour aerial photographs of the Coorong and region
were taken in 2003. DEH supplied geo-referenced and orthorectified aerial photos of
the Coorong and parts of the Lower Lakes to this project in the Enhanced
Compressed Wavelet (ECW) 3-band RGB format (Figure 2.6). These aerial photos
were used as the major tool for the current habitat mapping for the Coorong and
formed the spatial reference base for the project. The resolution of the aerial
photography is 0.5 m by 0.5 m, and ground features including terrestrial vegetation
are easily identifiable in these photos. Extensive mapping of reference points along
the length of the Coorong during fieldwork using differential GPS to an accuracy of
<40 cm confirmed the high quality of spatial representation within this dataset.
The CLLAMM Dynamic Habitat
15
Figure 2.6. Aerial photographs for the Coorong region taken in 2003.
Satellite imagery
LANDSAT 5 Thematic Mapper (TM) imagery (mosaic) covering the CLLAMM region
was purchased from MapLand, DEH. The LANDSAT5 mosaic was supplied by DEH
resampled to 25 m, orthorectified and histogram matched. The imagery was collected
by NASA’s LANDSAT5 mission equipped with the Thematic Mapper (TM)
multispectral sensor in 2004. The electromagnetic energy reflected from the earth’s
surface is captured by the sensors in seven different bands (1-7) at a resolution of 30
m. Each of these bands captures reflectance in a different range of wavelengths in
the electromagnetic spectrum, with different land cover types having different
reflectance signatures across the seven bands. For visualisation, a combination of
bands is displayed in red, green and blue colour for differentiating different land
covers, such as vegetation, rock and water. A false colour composite image of the
LANDSAT5 imagery is shown in Figure 2.7. The digital information contained within
the various bands of the multi-spectral imagery can be converted into land cover
information and maps through application of a range of well accepted digital image
classification techniques, as described in section 2.5.3 below.
The CLLAMM Dynamic Habitat
16
Figure 2.7. LANDSAT5 imagery (bands 2, 3 and 4) for the Coorong, Lower
Lakes and Murray Mouth.
Sediment sampling and underwater video survey
Sediment core samples were collected along three cross-lagoon transects from the
subtidal areas at each of the 12 reference sites used for the study. These samples
were analysed for sediment size fractions and organic matter content in the
laboratory. The physical composition of cores from each site in terms of general
sediment type and accompanying floral or faunal communities (if any) were noted to
aid in identifying habitat types.
Underwater video surveys were undertaken to complement the habitat type
information collected from the core samples. The video survey equipment comprised
a Morph Cam underwater video camera that was mounted on a small towable
platform that kept it at a 300 angle to the bottom. A boat was used to tow the
equipment and travelled along the same transects used for the core sampling, with
footage collected from all areas > ~0.4-0.5 m depth. While the video survey was
undertaken, the footage of the bottom surface was observed live on the screen and
any changes in the lagoon bed were recorded manually. The digital footage was
subjected to further analysis in the laboratory, which involved scoring of successive
image frames to record the presence or absence of underwater vegetation and other
notable characteristics. A GPS mounted on the boat was used to record the
geographic location for observed habitat transitions. An example of the video survey
footage from the North Lagoon is shown in Figure 2.8. The South Lagoon was too
shallow to use the boat and hence video surveys were not carried out beyond
Noonameena.
The CLLAMM Dynamic Habitat
17
A
B
Figure 2.8. Video footage from Mark Point (A) and Goolwa Channel (B) taken during the
underwater video survey.
Other data
Salinity data was acquired during fish sampling conducted at 10 locations between
Goolwa and Salt Creek in June 2008. The data were interpolated in ArcGIS using
Inverse Distance Weighting (IDW) (Environmental Systems Research Institute 2008)
to obtain salinity at the 12 representative sites. While, seasonal variations in salinity
were evident between summer and winter along the lagoon, they were not sufficiently
large to change our conclusions. In the North Lagoon slightly lower salinity (< 10 g/L)
was recorded in the winter whereas salinity difference could be as high as 30 g/L in
the South Lagoon between summer and winter.
2.5.2 Habitat digitization and interpretation
Habitat maps of the Coorong and surrounds were created by digitizing distinctive
units from the geo-referenced aerial photographs taken in 2003. Each of the habitat
units was visually interpreted based on standard aerial photo interpretation
techniques employing colour, appearance, texture, shape, etc. The minimum
mapping unit was not set beforehand and instead we sought to map all habitat types
identified within the reference sites. The digitization was performed at a scale of
1:1,000 to 1:1,500 in order to expedite the process without compromising the
accuracy of the maps. A finer scale of interpretation may enhance the map accuracy
at the expense of greater time for digitization (Kendall et al. 2001). Firstly, a
distinctive area with identical appearance in the photographs was digitized on screen
using ArcGIS Editor (Environmental Systems Research Institute 2008). Secondly, the
attributes as per the habitat classification scheme were assigned for each habitat
unit. The substrate data for the sub-tidal habitat in the benthic system were based on
the video survey data and their visual interpretation. The coastal salt-marsh and
mangrove map and DEH habitat map were used to derive information on vegetation
and cover/abundance. For those habitat units that matched the DEH habitat maps,
the habitat attributes were also collated from that dataset. A special case related to
the distributions of polychaete worm mounds in the sub-tidal areas, which were
interpreted through local knowledge of the areas.
2.5.3 Imagery classification
For selecting appropriate bands for classification of LANDSAT5 imagery, correlations
among the bands were analysed (Table 2.6). The thermal infrared band (band 6) is
not considered in this analysis as it is mostly used to determine surface
temperatures, and is not generally used for terrestrial vegetation classification (Sader
et al. 2001; Trisurat et al. 2000). High correlations can reduce the effectiveness of the
image classification process (Lillesand and Kiefer 2000). The correlation matrix
The CLLAMM Dynamic Habitat
18
showed that band 1 was highly correlated to bands 2 and 3, and moderately
correlated to bands 4, 5 and 7. Band 4 was relatively poorly correlated to bands 1
and 2 and moderately correlated to bands 5 and 7, whereas bands 5 and 7 were
relatively highly correlated. Bands 1, 4 and 7 were thus chosen for unsupervised
classification based on the criterion of minimum band correlation.
Table 2.6. Correlations between the individual bands of LANDSAT imagery (excluding
thermal band 6).
Layer
1
2
3
4
5
1
1
2
0.94
1
3
0.91
0.90
1
4
0.55
0.48
0.70
1
5
0.48
0.41
0.71
0.81
1
7
0.51
0.45
0.73
0.76
0.98
7
1
2.5.4 Map validation and accuracy testing
The habitat maps for the 12 reference sites produced by digitizing aerial photos were
printed out and taken to the sites for field verification. Those habitats that were not
interpreted accurately from the aerial photos were visited, identified and revised in
the habitat map database.
To assess the accuracy of the broader scale classified maps, 159 points, mostly on
the eastern shores of the Coorong and covering all subtidal, intertidal and terrestrial
habitat classes, were randomly selected and classified based on interpretation of the
aerial photos. The interpreted classes were validated during field observation and
assigned to the appropriate habitat classes for accurate assessment of the maps.
2.6.
Results
2.6.1 Habitat maps of the reference sites
The habitat maps of the 12 reference sites were analysed for each category of the
habitat classification scheme. The following sections summarise the findings from the
habitat classification in the Coorong.
Habitat zone
The distribution of habitat zones for all reference sites in the Coorong is provided in
Table 2.7. The reference sites were located between the mainland and the peninsula,
except for Mundoo Channel (between Hindmarsh and Ewe islands), Barker Knoll
(between the peninsula and Hindmarsh/Ewe islands) and Ewe Island (between the
peninsula and Ewe Island). All reference sites in the South Lagoon contained small
islands. 49.9% of the total area mapped was in the lagoon, 34.7% on the peninsula,
13% on the mainland and less than three percent was on islands.
The CLLAMM Dynamic Habitat
19
Table 2.7. Habitat zone distribution in the reference sites.
Habitat zone (Area, ha)
SN
Reference site
Island
Lagoon
Mainland
Peninsula
Total
129.66
44.74
116.52
290.93
1
Goolwa Channel
2
Mundoo Channel
10.33
37.94
17.93
3
Barker Knoll
22.55
97.29
15.69
4
Ewe Island
27.28
203.13
5
Pelican Point
182.90
6
Mark Point
7
66.20
70.28
205.81
223.96
454.37
27.79
315.43
526.13
236.91
24.09
250.60
511.60
Long Point
252.97
47.57
170.47
471.01
8
Noonameena
334.73
12.74
151.32
498.78
9
Parnka Point
64.95
450.65
110.37
625.97
10
Villa dei Yumpa
50.13
462.57
8.20
175.69
696.59
11
Jack Point
7.51
417.43
45.45
234.65
705.05
12
Salt Creek
17.32
452.42
53.48
179.60
702.81
Total Area
135.13
2872.91
748.34
1998.88
5755.26
2.35
49.92
13.00
34.73
100.00
Area Percent
Habitat category
The reference sites were classified into terrestrial and wetland habitat categories.
These habitat categories comprised 34.9 and 65.1%, respectively, of the mapped
area (Table 2.8). Approximately three quarters of the wetland category was located
within the lagoon system while the remaining quarter was located in the mainland,
peninsula and island habitat zones.
The CLLAMM Dynamic Habitat
20
Table 2.8. Habitat category distribution in the reference sites.
Habitat category (Area, ha)
SN
Reference site
Terrestrial
Wetland
1
Goolwa Channel
109.41
181.52
2
Mundoo Channel
5.02
61.19
3
Barker Knoll
73.46
132.35
4
Ewe Island
212.03
242.34
5
Pelican Point
305.52
220.60
6
Mark Point
265.97
245.63
7
Long Point
192.54
278.47
8
Noonameena
154.28
344.50
9
Parnka Point
71.14
554.83
10
Villa dei Yumpa
183.00
513.60
11
Jack Point
259.29
445.76
12
Salt Creek
176.28
526.53
Total Area
2007.93
3747.33
34.89
65.11
Area Percent
Landform
A total of 18 landforms were identified in the reference sites (Table 2.9). Dune,
channel and floodplain were major landforms constituting 43.7, 24.7 and 8.1% of the
total area, respectively. The dune landform without vegetation was dominant in the
peninsula, whereas channel was the predominant landform of the lagoon system.
The floodplains (including ephemeral mudflats) were found adjacent to the eastern
shore of the lagoon, primarily in the mainland or island habitat zones.
The CLLAMM Dynamic Habitat
21
Table 2.9. Landform distribution in the reference sites.
Reference sites (Area, ha)
SN
Landform
Goolwa
Channel
Mundoo
Channel
Barker
Knoll
1
Beach
14.92
2.70
14.85
2
Channel
126.33
31.58
87.65
3
Closed Depression
4
Coastal Swale
5
Ewe
Island
187.50
Pelican
Point
127.56
Mark
Point
Long
Point
Noonameena
Parnka
Point
Villa dei
Yumpa
15.97
4.04
32.01
49.93
66.27
210.47
220.72
302.72
36.38
420.49
1.42
1.57
1.10
4.73
1.42
11.02
8.20
4.24
2.21
Consolidated Dune
7.71
173.20
32.62
6.55
20.36
14.35
6
Cove
0.60
7
Dune
94.07
56.73
8
Flat
14.44
0.56
9
Floodplain
37.32
10
Rocky Outcrop
11
Rocky Shore
12
Salt Lake
13
Sand bar
14
Sandy Beach
15
Shoreline
16
Stream Channel
17
Undulating Plain
18
Vegetated Island
37.58
322.40
Salt
Creek
11.51
209.33
440.33
2514.12
10.96
7.46
47.86
87.41
13.24
62.04
28.85
418.89
0.60
38.83
271.80
235.85
1.11
36.06
26.68
15.63
23.64
145.87
130.84
25.78
7.93
16.58
4.74
71.14
163.24
87.26
141.72
47.83
29.88
4.12
28.33
59.05
463.63
26.57
3.46
56.13
10.47
147.43
28.21
203.87
203.87
32.61
32.61
89.91
3.67
0.88
22.73
0.06
1421.55
127.53
28.21
1.37
Total
8.63
4.81
0.07
26.28
Jack
Point
0.25
0.48
68.46
158.37
2.83
29.96
0.30
1.83
1.15
4.43
8.15
9.30
2.30
2.30
Wetland system
The areas in the wetland category were further classified into five major wetland
systems using the classification scheme of Seaman (2003) for wetland mapping of
the Coorong and Lower Lakes region. The distribution of these wetland systems
within the reference sites is given in Table 2.10. Estuarine wetland was the dominant
wetland system, comprising more than 80% of the total wetland area. Palustrine
wetlands constituted about one sixth of the area and were commonly located
adjoining the main channel. Less than two percent of the area along the coast was
classified as marine wetland. The areas classified as lacustrine and riverine wetlands
were negligible.
Table 2.10. Distribution of wetland system in the reference sites.
Wetland system (Area, Hectare)
SN
Reference site
Marine
Estuarine
Palustrine
11.59
140.85
28.60
0.48
38.66
21.69
0.84
99.15
29.79
1
Goolwa Channel
2
Mundoo Channel
3
Barker Knoll
4
Ewe Island
205.68
36.66
5
Pelican Point
194.94
24.32
6
Mark Point
245.63
7
Long Point
268.93
9.54
8
Noonameena
337.56
6.95
9
Parnka Point
21.36
155.57
377.90
10
Villa dei Yumpa
24.18
477.14
12.27
11
Jack Point
433.68
12.08
12
Salt Creek
473.06
51.65
3.41
Riverine
Lacustrine
1.35
1.83
Total Area
60.55
3070.84
611.45
1.35
3.14
Area Percent
1.62
81.95
16.32
0.04
0.08
Water regime
The wetlands in the reference sites were classified into five water regime categories
based on the water condition and the frequency of inundation. The distribution of
these water regime categories is given in Table 2.11. The lagoon system consisted of
the subtidal and intertidal water regimes, comprising about 67.2 and 9.7% of the
wetlands in the reference sites respectively. Some areas switched between these
categories, depending on the water level in the Coorong. About a quarter of the
wetland areas were seasonally inundated. The ocean coastlines were the classic
example of intertidal areas and were classified as tidal regions under water regime.
The CLLAMM Dynamic Habitat
23
Table 2.11. Distribution of water regimes in the Reference Sites.
Water regime (Area, Hectare)
SN
Reference site
Subtidal
Intertidal
Tidal
11.59
Semi Permanent
Seasonal
1
Goolwa Channel
126.33
3.33
2
Mundoo Channel
32.42
6.60
3
Barker Knoll
87.65
11.44
4
Ewe Island
180.22
23.74
5
Pelican Point
127.56
57.71
35.33
6
Mark Point
220.94
15.97
8.72
7
Long Point
220.72
32.25
25.50
8
Noonameena
302.72
32.97
8.82
9
Parnka Point
36.38
28.58
21.36
468.52
10
Villa dei Yumpa
420.49
42.09
24.18
26.84
11
Jack Point
322.40
95.03
12
Salt Creek
440.33
12.09
Total Area
2518.15
361.79
67.20
9.65
Area Percent
40.26
22.17
3.41
29.84
1.63
36.76
28.33
1.83
72.29
60.55
3.45
803.39
1.62
0.09
21.44
Wetland type
Of 44 wetland types mentioned in Seaman (2003), 17 were observed in the 12
reference sites. Although Seaman (2003) defined coastal dune shrubland as a new
marine and coastal wetland type, we classified it as terrestrial. The distribution of all
17 wetland types in these reference sites is presented in Table 2.12. Subtidal aquatic
bed (A2) was the dominant wetland type occupying about two-thirds of the wetland in
the study area. The second most commonly observed wetland type was seasonal
saline marshes (B12) followed by intertidal mud, sand or salt flats (A7). These two
wetland types covered more than 10 and 7 percent of the total wetland, respectively
(Figure 2.9).
The CLLAMM Dynamic Habitat
24
Table 2.12. Distribution of wetland types in the reference sites.
Wetland
type
A2
Reference site (Area, Hectare)
Goolwa
Channel
126.33
Mundoo
Channel
31.58
Barker
Knoll
87.65
A4
Ewe Island
180.22
Pelican
Point
127.56
15.63
A5
3.33
A7
11.59
A8
11.18
1.81
7.03
Mark
Point
Long
Point
210.47
220.72
Noonameena
302.72
32.61
0.16
0.63
7.28
22.73
15.81
32.25
32.01
1.72
11.71
12.03
8.72
7.33
8.00
9.53
0.07
A15
0.88
0.66
Villa dei
Yumpa
36.38
420.49
10.47
13.05
A9
Parnka
Point
Jack
Point
322.40
440.33
26.57
3.46
89.91
42.09
49.93
24.18
68.46
8.63
2.30
16.25
3.32
0.29
B1
B2
1.83
0.48
B8
B12
27.71
12.16
27.40
17.41
B13
24.32
17.54
1.78
204.57
4.81
174.04
19.73
68.80
1.35
B14
11.79
B15
B20
Salt
Creek
8.36
0.89
C4
NB: See Table 2.2 for definitions of wetland types.
1.43
0.29
0.17
70
60
AreaArea
(Percent)
(Percent)
50
40
30
20
10
0
A2
A4
A5 A7
A8
A9 A15 B1
B2
B8 B12 B13 B14 B15 B20 C4
Wetland Types
Figure 2.9. Total area covered by the wetland types in the reference sites (See Table 2.2 for
definitions of wetland types).
Habitat types
The habitat types in the wetland and terrestrial habitat categories are given in Figures
2.10 and 2.11, respectively. The first eight wetland habitat types (listed in Table 2.3)
were found in the subtidal areas within the main channel of the lagoon. Bare mud
was predominant in the deep channel throughout the lagoon, whereas the adjacent
shallower areas were characterised by the presence of polychaetes, fine filamentous
algae and Ruppia. Shallow areas between the channel and shores were interspersed
with polychaetes (0.1-1.5 m2) on the bare mud in the North Lagoon. Polychaetes
were mainly reported in the bare mud in shallower water at Mark Point.
Bare sand was commonly found in the intertidal areas and also in the salt pan/lake
areas. Samphire was found along both shorelines across the lagoon. However, the
cover density varied in different locations and was classified into four habitat types:
dense samphire (>75% cover), samphire (50-75% cover), patchy samphire (25-50%
cover) and open samphire (<25% cover). Coastal vegetation was reported mainly on
the floodplain areas on the eastern coast of the lagoon. In the open and patchy
coastal vegetation areas, the ground surface was mostly covered by small grasses.
Some grasslands without overstorey vegetation were reported from four reference
sites. Areas of dense and patchy Melaleuca were found in the Mundoo Channel site.
Extensive areas of sand dunes were exposed on the peninsula between the lagoon
and the Southern Ocean. However, the consolidated dunes were covered by dune
vegetation from dense to patchy cover densities in all reference sites except for
Mundoo Channel. A few patches of coastal vegetation were observed on the
mainland on the eastern coast of the lagoon. Pasture lands were reported from the
North Lagoon reference sites. The habitat maps and some photos are given in
Appendix 2.2 and 2.3, respectively.
The CLLAMM Dynamic Habitat
26
600
Very Soft Mud
Samphire
Rocky Outcrop
500
Patchy Samphire
Patchy Melaleuca
Patchy Coastal Vegetation
400
Patchy Bare Mud
Area in ha.
Open Samphire
Open Coastal Vegetation
300
Mud Flat
Melaleuca
Wet Grass Land
200
Dense Sedge
Dense Samphire
Dense Melaleuca
100
Coastal Vegetation
Bare Sand Mud with Ruppia
Bare Sand and Grit
un
do
o
M
G
oo
lw
a
C
Bare Sand Mud
ha
nn
el
C
ha
nn
el
Ba
rk
er
Kn
ol
l
Ew
e
Is
la
nd
Pe
lic
an
Po
in
t
M
ar
k
Po
in
t
Lo
ng
Po
in
N
t
oo
na
m
ee
na
Pa
rn
ka
Po
Vi
in
lla
t
de
iY
um
pa
Ja
ck
Po
in
t
Sa
lt
C
re
ek
Bare Sand Mud with Algae
0
Bare Sand
Bare Mud with Ruppia
Bare Mud with Polychaetes
Bare Mud
Reference Sites
Figure 2.10. Area and habitat types in the wetland category.
350
Sand Dune
Revegetated area
300
Planted Trees
Patchy Pasture
Area in ha.
250
Patchy Dune Vegetation
Patchy Coastal Vegetation
200
Pasture
Open Dune Vegetation
150
Open Coastal Vegetation
Grass Land
100
Dune Vegetation
Dense Myoporum
50
Dense Dune Vegetation
Coastal Vegetation
k
re
e
in
t
lt
C
Po
Sa
ck
um
pa
Ja
in
t
Po
a
Vi
lla
de
iY
na
rn
k
Pa
am
ee
oo
n
N
Lo
ng
Po
in
t
in
t
t
Po
k
Po
in
an
lic
Pe
M
ar
an
d
ll
Is
l
Ew
e
el
rK
no
rk
e
C
ha
nn
Ba
do
M
un
G
oo
lw
a
o
Ch
a
nn
el
0
Reference Sites
Figure 2.11. Area and habitat types in the terrestrial category.
Salinity
Salinity levels at the reference sites were derived through interpolating the salinity
data collected across the lagoon in June 2008 (Table 2.13). Between Goolwa and
Ewe Island, the salinity range was 30-40 g/L, increasing to 40-50 g/L between
Pelican Point and Mark Point. At Parnka Point, the barrier between the North and
South Lagoon, the salinity rose to more than 110 g/L. The salinity range remained
constant between Parnka Point and Jack Point at 110-120 g/L. In the area south of
Jack Point, the salinity level reached more than 120 g/L. It should be recognised that
the water levels in the Coorong are highest in June and salinity levels are
considerably higher in summer, particularly in the South Lagoon.
The CLLAMM Dynamic Habitat
27
Table 2.13. Salinity level at the reference sites in the Coorong in June 2008.
Salinity (g/L) (Area, Hectare)
Reference sites
30-40
Goolwa Channel
129.66
Mundoo Channel
36.13
Barker Knoll
88.31
Ewe Island
196.14
40-50
Pelican Point
127.56
Mark Point
220.94
Long Point
50-60
70-80
110-120
120-130
220.72
Noonameena
302.72
Parnka Point
36.38
Villa dei Yumpa
420.49
Jack Point
348.97
Salt Creek
443.78
Total Area
450.24
348.50
220.72
302.72
456.86
792.75
Area Percent
17.51
13.55
8.58
11.77
17.76
30.82
Vegetation
Samphire dominated by Sarcocornia and Halosarcia was commonly observed along
shorelines on both sides of the lagoon adjacent to the intertidal sand on the coastal
swale and floodplain landforms. In the mainland area, samphire or grasslands were
gradually replaced by coastal vegetation composed of Myoporum, Casuarina,
Acacia, saltbush and Melaleuca. In the peninsula, inland dune areas had dune
vegetation associations dominated by Olearia asillaris, Acacia longifolia and
Leucopogon parviflorus. A dense patch of mallee association of Eucalyptus
diversifolia and Melaleuca was found at the Salt Creek reference site.
In patchy or open vegetation, the ground cover mainly included introduced grasses
(less than 0.5 m tall) such as Paspalum (Paspalum distichum), Kikuyu (Pennisetum
clandestinum) and Couch (Cynadon dactylon). However, grasses composed of reed
species (Phragmites australis and Typha domingensis) were also reported at the
Goolwa Channel site.
In the subtidal areas, the mud sediments in shallower water were colonized by
Ruppia tuberosa at Noonameena, Parnka Point and Villa dei Yumpa. As this species
typically grows on ephemeral mudflats (Paton 2005) this suggests that these areas
are exposed for part of the year. Fine filamentous algae were observed at
Noonameena.
Cover
Table 2.14 presents the ground cover percent for the habitat types characterised by
vegetation in both wetland and terrestrial habitat categories. The highest area
(32.5%) was found under 25-50% cover. Only one percent of the areas had less than
5% cover.
The CLLAMM Dynamic Habitat
28
Table 2.14. Cover percentages for Reference Sites in the Coorong.
Cover percent (Area, ha)
Reference site
<5%
Goolwa Channel
1.84
Mundoo Channel
0.26
<25%
25-50%
50-75%
95.22
>75%
49.93
2.77
6.14
10.44
7.10
Barker Knoll
5.33
52.14
32.37
1.79
Ewe Island
0.74
18.92
182.66
8.36
185.43
56.93
13.74
Pelican Point
Mark Point
43.57
14.33
7.07
31.57
Long Point
11.59
95.48
38.67
17.74
Noonameena
7.44
77.86
5.57
19.07
Parnka Point
208.50
31.02
5.66
138.37
49.41
12.02
155.21
81.76
14.38
37.49
46.15
Villa dei Yumpa
Jack Point
6.21
7.51
Salt Creek
17.32
43.50
Total Area
23.79
330.95
743.88
533.39
177.57
Area Percent
1.04
14.45
32.47
23.28
7.75
Habitat condition
Although there were six habitat condition categories used by Seaman (2003), the first
two categories, pristine and extremely good, were not reported as this region has
experienced a high level of anthropogenic interference, invasion of introducedspecies
and high salinity levels. Habitat conditions in the 12 reference sites in the Coorong
are given in Table 2.15.
The CLLAMM Dynamic Habitat
29
Table 2.15. Habitat conditions in the Coorong.
Habitat conditions (Area, Hectare)
Reference site
Completely
Degraded
Very good
Good
Degraded
Goolwa Channel
131.03
43.91
115.09
Mundoo Channel
42.07
14.48
6.90
2.77
Barker Knoll
97.94
44.42
61.04
1.84
Ewe Island
204.86
191.02
18.92
39.57
Pelican Point
150.29
103.68
185.03
87.12
Mark Point
243.30
32.25
57.90
177.95
Long Point
28.21
267.73
115.61
59.37
Noonameena
42.25
11.17
445.19
Parnka Point
12.66
125.35
451.59
36.38
107.33
119.28
469.92
Villa dei Yumpa
Jack Point
5.65
158.96
168.94
371.23
Salt Creek
13.23
29.40
112.29
547.89
Total
971.48
1129.69
1857.79
1794.04
Area Percent
16.89
19.64
32.29
31.18
2.6.2 Habitat classification
Maximum likelihood classification of LANDSAT5 TM imagery
Figure 2.12 illustrates the habitat maps generated by applying maximum likelihood
classification to LANDSAT5 bands 1, 4 and 7. The blue colour represents areas
under water in the Lower Lakes and the subtidal areas in the Coorong, and is mostly
characterised by bare mud substrate. Intertidal sand areas were found around the
Murray Mouth and mostly on the eastern shore adjacent to the subtidal areas. Sand
dunes in the peninsula and the inland sand flats/built areas were classified into
separate classes. However, some misclassification between intertidal sand and
inland sand flat was unavoidable and these areas could not be reliably distinguished.
Both coastal and dune vegetation were classified as shrublands, whereas the dense
vegetation dominated by mallee or Myoporum was classified as woodland. Reed
grasses around Lake Alexandrina were also accurately represented in the map.
Pasture and grassland (small) were also classified into one class because of the
spectral signature overlap between them in the iso-cluster classification. Open lands
in exposed condition with no vegetation, and also the land not used for agriculture,
were categorised into one class in the maps.
The CLLAMM Dynamic Habitat
30
Figure 2.12. Habitat classification using LANDSAT5 bands 1, 4 and 7.
Validation of the classified map
The accuracy of the habitat classifications derived from the raw LANDSAT5 TM
bands was assessed using 159 validation points with known landuse/habitat. The
accuracies for each class as well as for the overall map are given in Table 2.16. The
classification derived from LANDSAT5 bands 1, 4 and 7 correctly classified 83% of
the validation points. The lowest accuracy was found for the inland sand flats. In
most cases these were classified as intertidal sand or coastal sand dune, which was
clearly due to their similar geomorphologic characteristics, leading to similarity in their
reflectance.
The CLLAMM Dynamic Habitat
31
Table 2.16. Accuracy assessment results for the classification maps.
Habitat Class
Bare mud (Subtidal Areas)
Bare Sand (Intertidal areas)
Pasture/Grass Land
Agriculture
Coastal Sand Dunes
Woodlands
Open Land
Reed Grass
Inland Sand Flats
Coastal/Dune Vegetation
Total
2.7.
Number of
validation
points
Number of
points
classified
correctly
Percent of
accuracy
17
21
19
7
25
22
20
2
10
16
159
16
17
12
7
23
21
18
2
3
13
132
94.12%
80.95%
63.16%
100.00%
92.00%
95.45%
90.00%
100.00%
30.00%
81.25%
83.01%
Discussion
The habitat mapping carried out for the current study involved both terrestrial and
wetland areas, including benthic habitats, in the 12 reference sites across the
Coorong and surrounding area. A specific habitat classification scheme with eight
physical, two biological and one chemical (water quality) parameters was used to
give a simple and clear delineation of unique habitats in the terrestrial and benthic
environment. All 24 attributes used in the DEH habitat maps for the wetlands in the
Coorong region could not be used because of the lack of detailed information for
each habitat unit, and a simpler classification scheme was therefore more
appropriate. However, the habitat maps generated from this study, at both the local
(reference site) scale and regional (CLLAMM) scale have been created within a GIS
database which allows future updates and also inclusion of additional parameters or
further detail on an existing attribute.
Among the wetland habitats, mudflats and samphire are the most significant from an
ecological perspective as they offer suitable foraging ground for wader species
(Edyvane 1999). These habitats were mostly reported along the eastern (landward)
shoreline of the Coorong, and are frequently inundated so offer suitable habitat for
macro-invertebrates, hence their value as bird habitat. For many shorebird species,
mudflats deeper than 12 cm at any point in time are generally not accessible
(Rogers, pers. comm) and cannot be used for foraging. Since there is no or limited
tidal influence along much of lagoon, longer term changes in water level and wind
action play a major role in inundating these areas. The maintenance of water levels
in the lagoon, particularly in the summer, requires inflow over the barrages. Lack of
such discharge for the past five years has resulted in increased exposure of
extensive mudflat areas, particularly in summer, which in turn are becoming
unsuitable for colonisation by macro-invertebrates as they dry out, and this is directly
affecting shorebirds through decreased availability of prey (CLLAMM 2008).
The lack of freshwater flow into the lagoon has also resulted in extreme hypersaline
conditions in the South Lagoon, creating unfavourable habitats for many organisms.
The dramatic rise in salinity has already had a profound impact on the ecosystems of
the region by apparently eliminating some important key species like Ruppia
megacarpa from the system, while others (e.g. R. tuberosa and Smallmouth
The CLLAMM Dynamic Habitat
32
Hardyhead) are now found only in the North Lagoon whereas they were previously
found in the South Lagoon (CLLAMM 2008).
The habitats in the broader Coorong and Lower Lakes region were classified by
using a combination of unsupervised isoclassification and supervised maximum
likelihood classification using three original bands 1, 4 and 7 of the LANDSAT5
imagery. The Coorong and surrounding region were classified into ten broad habitat
categories including agricultural and pastoral land.
The subtidal areas were perfectly differentiated by this classification, as found by
Talukdar (2004). However, inland sand flats and intertidal sand flats/bare sand were
often misclassified because of similar reflectance from bare sand surface on both
areas.
Due to the limitations of LANDSAT5 in capturing reflectance from the underwater
surface and also the moderate spatial resolution of 25 m, the terrestrial and wetland
habitats were not represented in great detail in these maps. The use of imagery with
high spatial and spectral resolution will enhance the ability to identify more variations
on the ground, enabling the extraction of habitat maps at finer scales.
2.8.
Summary and conclusions
This habitat mapping was accomplished at 12 reference sites, covering the Coorong
system from Goolwa Channel to Salt Creek. A great diversity of habitats was
reported, with 24 wetland habitats and 14 terrestrial habitats in these sites. Among
the wetland habitats, mudflats and samphire are vital for wader species However, the
habitats for fish and birds were subjected to extremely high salinity and
unprecedented low water levels primarily due to lack of fresh water flows through the
barrages, thus reducing their availability and quality.
The habitats in the Coorong are shaped and maintained by the fresh-water flow into
the system, which sustains the water level, as well as salinity gradient, supporting the
high biodiversity of the region. The relationships between the key bird and fish
species and their physical environment in terms of water level and salinity have been
explored by the CLLAMMecology Research Cluster. However, the biological
interrelationships are not yet sufficiently well known as to allow us to confidently
predict how changes in the abundance of one species will affect others. The
response of a species to changing habitat conditions together with the quantification
of the impacts of the species on other members of biological communities would be
very helpful in predicting the future states of ecosystems in the Coorong.
This mapping was done in a GIS platform by compiling all attributes for each habitat
type and the resultant GIS database allows to easily share and query information
about the habitat in the Coorong. In addition, new fields for habitat attributes could be
easily incorporated into the database to make it more comprehensive.
2.9.
References
Aber, J.S. (2007) Wetland definitions and classification. Retrieved 14/07/2008, from
http://academic.emporia.edu/aberjame/wetland/define/define.htm.
Australian Nature Conservation Agency (1996) A directory of important wetlands in
Australia, Second Edition. Canberra.
Barrett, N., Sanderson, J.C., Lawler, M.V., Halley, V. and Jordan, A. (2001) Mapping
of Inshore Marine Habitats in South-eastern Tasmania for Marine Protected Area
The CLLAMM Dynamic Habitat
33
Planning and Marine Management. Marine Research Laboratories - Tasmanian
Aquaculture and Fisheries Institute, University of Tasmania, Hobart, Tasmania.
Blackman, J.G., Spain, A.V. and Whiteley, L.A. (1992) Provisional handbook for the
classification and field assessment of Queensland wetlands and deep water habitats.
Queensland Department of Environment and Heritage, Townsville
Bowers, K. and Boutin, C. (2008) Evaluating the relationship between floristic quality
and measures of plant biodiversity along stream bank habitats. Ecological Indicators
8(5): 466-475.
Canty, D. and Hille, B. (2002) Coastal Saltmarsh and Mangrove mapping. South
Australian Department for Environment and Heritage, Adelaide.
Carpenter, G. (1995) Birds of the Lower Murray region of South Australia. The
Murray Mouth Biological Resource Assessment Workshop, Adelaide.
Chong, G.W. and Stohlgren, T.J. (2007) Species-area curves indicate the importance
of habitats' contributions to regional biodiversity. Ecological Indicators 7(2): 387-395.
Chust, G., Galparsoro, I., Borja, Á., Franco, J. and Uriarte, A. (2008) Coastal and
estuarine habitat mapping, using LIDAR height and intensity and multi-spectral
imagery. Estuarine, Coastal and Shelf Science 78(4): 633-643.
CLLAMM (2008) Response of the Coorong ecosystems to alternative Murray-Darling
flow scenarios. CLLAMMecology Research Cluster, Adelaide
Cowardin, L.M., Carter, V., Golet, F.C. and La Roe, E.T. (1979) Classification of
wetlands and deepwater habitats in the United States. U. S. Department of Interior,
Fish and Wildlife Service.
Department for Environment and Heritage (2000) Coorong, and Lakes Alexandrina
and Albert Ramsar Management Plan. South Australian Department for Environment
& Heritage, Adelaide.
Department for Environment and Heritage (2008) Coastal Management. Retrieved
13/08/2008, from http://www.environment.sa.gov.au/coasts/management.html.
Earth Sciences Research Institute (2001) ArcGIS Geostatistical Analysis: Powerful
exploration and data interpolation solutions. ESRI, Redlands, California.
Edyvane, K.S. (1999) Conserving Marine Biodiversity in South Australia, Part 2Identification of areas of high conservation value in South Australia. South Australia
Research and Development Institute, Aquatic Sciences, Adelaide.
Environment Australia (2001) A directory of important wetlands of Australia, Third
Edition. Environment Australia, Canberra.
Environmental Systems Research Institute (2008), ArcGIS 9.3 Desktop. ESRI,
Redlands, California.
Finkbeiner, M., Stevenson, B. and Seaman, R. (2001) Guidance for Benthic Habitat
Mapping: An Aerial Photographic Approach. U.S. NOAA Coastal Services Centre,
Charleston, SC.
The CLLAMM Dynamic Habitat
34
Heard, L. and Channon, B. (1997) Guide to a native vegetation survey using the
biological survey of South Australia methodology. Department of Housing and Urban
Development, South Australia, Adelaide.
Jordan, A.R., Lawler, M. and Halley, V. (2001) Estuarine Habitat Mapping in the
Derwent-Integrating Science and Management, NHT Final Report. Tasmanian
Aquaculture and Fisheries Institute, Hobart, Tasmania.
Kendall, M.S., Monaco, M.E., Buja, K.R., Christensen, J.D., Kruer, C.R., Finkbeiner,
M. and Warner, R.A. (2001) Methods used to map the benthic habitats of Puerto Rico
and the U. S. Virgin Islands. National Oceanic and Atmospheric Administration,
National Ocean Service, National Centres for Coastal Ocean Science Biogeography
Program, Silver Spring, MD.
Kostylev, V. E., Todd, B. J., Fader, G. B. J., Courtney, R. C., Cameron, G. D. M.,
Pickrill, R. A. and (2001) Benthic habitat mapping on the Scotian Shelf based on
multibeam bathymetry, surficial geology and sea floor photographs. Marine Ecology
Progress Series 219: 121-137.
Lathrop, R.G., Cole, M., Senyk, N. and Butman, B. (2006) Seafloor habitat mapping
of the New York Bight incorporating sidescan sonar data. Estuarine, Coastal and
Shelf Science 68(1-2): 221-230.
Lillesand, T.M. and Kiefer, R.W. (2000) Remote Sensing and Image Interpretation.
John Wiley & Sons Inc., New York.
Paton, D.C. (2005) 2005 winter monitoring of the southern Coorong. Project Report
to South Australian Department for Environment & Heritage, University of Adelaide,
Adelaide.
Sader, S.A., Hayes, D.J., Hepinstall, J.A., Coan, M. and Soza, C. (2001) Forest
change monitoring of a remote biosphere reserve. International Journal of Remote
Sensing 22: 1937-1950.
Seaman, R.L. (2003) Coorong and Lower Lakes habitat-mapping program. South
Australian Department for Environment and Heritage, Adelaide.
Talukdar, K. (2004) Extraction and classification of wetland features through fusion of
remote sensing images in the Okavango Delta, Botswana. International Society for
Photogrammetry and Remote Sensing, Estambul.
Trisurat, Y., Eiumnoh, A., Murai, S., Hussain, M.Z. and Shrestha, R.P. (2000)
Improvement of tropical vegetation mapping using a remote sensing technique: a
case of Khao Yai National Park, Thailand. International Journal of Remote Sensing
21: 2031-2042.
Urbanski, J.A. and Szymelfenig, M. (2003) GIS-based mapping of benthic habitats.
Estuarine, Coastal and Shelf Science 56(1): 99-109.
Webster, I. T. (2007) Hydrodynamic modelling of the Coorong. Water for a Healthy
Country National Research Flagship, CSIRO, Canberra.
The CLLAMM Dynamic Habitat
35
2.10. Appendices
Appendix 2.1: A brief description of landforms reported in the Coorong (From
Seaman 2003).
1. Beach: Short, low, very wide slope, gently or moderately inclined, built up or
eroded by waves, forming the shore of a lake or sea.
2. *Channel: Main body of the lagoon under water and other subtidal areas.
3. Closed depression: Landform element that stands below all points in the adjacent
terrain.
4. Consolidated dune/dune: Moderately inclined to very steep ridge or hillock built
up by the wind. This element may comprise dune crest and dune slope.
May also be consolidated due to the stabilising effects of vegetation.
5. Cove: Body of water, depth six metres or less bounded by land on three sides.
Water is connected permanently by a narrow or wide opening to a larger
water body.
6. Flat: A planar landform element that is neither a crest nor a depression and is
level or very gently inclined (<3% slope).
7. Floodplain: Alluvial plain characterised by frequent active erosion and
aggradations by channelled or over-bank stream flow. Unless otherwise
specified, frequently active is to mean that flow has an average
recurrence interval of 50 years or less.
8. Rock outcrop: Any exposed area of rock that is inferred to be continuous with
underlying bedrock on a large, very gently inclined or level landform.
9. Rocky shore: Shorelines adjacent to a water body having an aerial cover of
bedrock, stones and boulders alone or in combination with 75% or more
of the surface cover. The vegetative cover is less than 30%.
10. Salt lake: Lake containing a concentration of mineral salts, predominantly sodium
chloride in solution as well as magnesium and calcium sulphate.
11. Sand bar: Elongated, gently to moderately inclined low ridge containing coarse
grains, built up by water movement.
12. Sandy beach: Short, low, very wide slope, gently or moderately inclined, built up
or eroded by waves, forming the shore of a lake or sea. Composed of
coarse grains.
13. Shoreline: Extensive, low, very wide slope, gently or moderately inclined, built up
or eroded by waves, forming the shore of a lake or sea. Composed of a
combination of one or more of the following: coarse grain sands, mud flat,
rocky reef and rocky shore.
14. Stream channel: Linear, generally sinuous open depression, in parts eroded,
excavated, built up and aggraded by channel stream flow.
15. Undulating plain: Large very gently inclined or level landform of unspecified
geomorphological agent or mode of activity.
16. Vegetated island: Sediments built up over time through water movement forming
landform with low relief consolidated by stabilising effects of vegetation.
*Landform with modified description during this study.
The CLLAMM Dynamic Habitat
36
Appendix 2.2: Habitat maps for the reference sites in the Coorong.
Habitat types at Goolwa Channel
Habitat types at Mundoo Channel
The CLLAMM Dynamic Habitat
37
Habitat types at Barker Knoll
Habitat types at Ewe Island
The CLLAMM Dynamic Habitat
38
Habitat types at Pelican Point
Habitat types at Mark Point
The CLLAMM Dynamic Habitat
39
Habitat types at Long Point
Habitat types at Noonameena
The CLLAMM Dynamic Habitat
40
Habitat types at Parnka Point
Habitat types at Villa dei Yumpa
The CLLAMM Dynamic Habitat
41
Habitat types at Jack Point
Habitat types at Salt Creek
The CLLAMM Dynamic Habitat
42
Appendix 2.3: Some photos of the habitats in the Coorong.
Cleared Land (Salt Creek)
Mudflat (Villa dei Yumpa)
Samphire
Sedge bushes
Coastal Vegetation Myoporum and Tea Tree
Rocky Shore (Jack Point)
The CLLAMM Dynamic Habitat
43
Coastal vegetation: Myoporum, iceplant,
sedge, foxtails and salt bush
Open degraded area (Villa dei Yumpa)
Samphire and Rocky Outcrop (Noonemeena)
Rocky surface (Long Point)
Grassland taken up by Samphire
Large coastal vegetation
The CLLAMM Dynamic Habitat
44
3. Digital Elevation Model of the Coorong and
Surrounds
Sunil K. Sharma1*, Jason E. Tanner1 and Simon N. Benger2
1
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
2
School of Geography, Population and Environmental Management, Flinders
University, GPO Box 2100, Adelaide, SA 5001
*corresponding author, Phone +61 (8) 8207 5448, Fax +61 (8) 8207 5448, E-mail:
[email protected]
3.1.
Executive Summary
An accurate and comprehensive Digital Elevation Model (DEM) is considered a
primary dataset for developing habitat models of the Coorong and its surrounding
areas. A seamless merger of topographic and bathymetric data would provide a
dataset which would be helpful for the better understanding of the ecology of the
region. The Dynamic Habitat Project under the CLLAMMecology Research Cluster
aims to generate a seamless DEM for the Coorong and surrounding areas by
collating the existing topographic and bathymetric datasets. However, the
bathymetric datasets for the South Lagoon and the areas around the Murray Mouth
are not available. Although the Department of Water, Land and Biodiversity
Conservation (DWLBC) have recently used airborne LIDAR for collecting topographic
data around the South Lagoon, this did not include bathymetry and only became
available in late October 2008, and thus could not be used in this project. To achieve
the project’s aim of a seamless merged bathymetric and topographic dataset, firstly
we modelled the bathymetry for the Murray Mouth and the South Lagoon using
satellite data, survey data collected during fieldwork, and the existing bathymetric
data for the Coorong and; secondly, the derived DEMs from the bathymetric data
were integrated with the DEM of South Australia to derive a seamless DEM for the
region.
Generalized Additive Modelling, a non-parametric method, was used to model the
relationship between depth and reflectance signatures captured in LANDSAT5 and
SPOT5 imagery. For modelling bathymetry for the Murray Mouth, all ten bands
(seven from LANDSAT5 and three from SPOT5) were used as predictors, whereas
only LANDSAT5 (seven bands) was used for the South Lagoon due to the SPOT5
image for this region not overlapping areas with known bathymetry that could be
used to develop the model. Model selection was accomplished by applying both
forward and backward selection procedures, followed by a comparison of prediction
errors for a small subset of candidate models. For the Murray Mouth, a 4 variable
model with SPOT5 bands 2 and 3 and LANDSAT5 bands 3 and 5 was selected to
predict bathymetry. For the South Lagoon, a 6 variable model was selected with
LANDSAT5 bands 1-6. The predicted bathymetry for the Murray Mouth worked well
in capturing the trend along the main channel and in the near-shore areas. The South
Lagoon model gave good predictions between Parnka Point and Jack Point.
However, the depth for the southern quarter of the South Lagoon was
underestimated, which could be directly attributed to poor reflectance because of
turbid water.
A seamless DEM for the Coorong and surrounds was derived by merging the
bathymetry for the North Lagoon and Lakes, and the predicted bathymetry for the
The CLLAMM Dynamic Habitat
45
Murray Mouth and the South Lagoon with the DEM for South Australia. However, the
bathymetry models for the Murray Mouth and the South Lagoon were not perfect and
could be improved in a number of ways. In future, the availability of more
comprehensive ground survey data for these regions will provide an opportunity for
further improvement in the errors and uncertainty in these models. The seamless
DEM for the Coorong and surrounds will be very useful for modelling habitat
availability for birds, fish and macro-invertebrates at different water levels in the
Coorong and will help in making decisions for sustainable management of the region.
3.2.
Introduction
The Coorong, Lower Lakes and Murray Mouth region offers unique coastal habitats
for a wide variety of macro-invertebrate, fish and bird species (MDBC 2006; Geddes
2005; Geddes 2004) (Figure 3.1). The availability of mudflat habitats, in particular, is
dependent on periodic inundation caused by either fresh water flow through the
barrages or incursion of sea water through the Murray Mouth. An accurate and
comprehensive Digital Elevation Model (DEM) is considered a primary dataset for
developing habitat models of the Coorong and its surrounds (CLLAMM 2007). A high
resolution DEM of the Lagoon and the adjoining topography is required to model the
changes in habitat availability in terms of spatial distribution and extent at various
water levels. Specifically, the DEM will be highly useful for predicting the availability
of mudflat habitat for shore birds at different depths as a function of water levels.
The Coorong, Lower Lakes and Murray Mouth (CLLAMM) Research Cluster is taking
a holistic approach to exploring implications of major ecological drivers like water and
salinity levels at the ecosystem level, through understanding the interrelationships
between physical processes and the biological communities (CLLAMM 2007).
Understanding of the physical processes is a prerequisite for better ecological
management of the region, and requires a seamless merger of topographic and
bathymetric data (Gesch and Wilson 2001). One aim of the Dynamic Habitat Project
under the CLLAMMecology Research Cluster is to produce a seamless DEM for the
Coorong and surrounding areas by collating the existing topographic and bathymetric
datasets. The available datasets include the 3 second DEM for South Australia (SA)
and interpolated bathymetry for the Lower Lakes and the Northern Lagoon of the
Coorong. However, bathymetric datasets for the South Lagoon and the areas around
the Murray Mouth are not available. Shallow waters and numerous limestone reefs in
the South Lagoon, in particular, have prevented standard bathymetric data collection
using a boat with depth sounder. In late 2008, the South Australian Department of
Water, Land and Biodiversity Conservation (DWLBC) developed a DEM for the
Lower Lakes and the Coorong from airborne (topographic) LIDAR (Light Detection
and Ranging) data under the “Imagery Baseline Data Program for the SA NRM
Planning, Monitoring and Evaluation Project”. Although the DEM has high resolution
(2m by 2m), the dataset excluded all areas covered in water, as free surface water
interferes with the LIDAR signal (DWLBC 2009). Thus this data set could be used to
improve the terrestrial component of the DEM presented here, but was not used as it
only became available in early 2009, after the project was completed. To achieve the
project’s aim of a seamless merged bathymetric and topographic dataset, firstly we
modelled the bathymetry for the Murray Mouth and the South Lagoon using satellite
data, survey data collected during fieldwork, and the existing bathymetric data for the
Coorong and; secondly, the derived DEMs from the bathymetric data were integrated
with the DEM of SA to get a seamless DEM for the region.
Digital Elevation Models are generated from various sources of data including digital
photogrammetry, airborne or terrestrial laser scanning, aerial photographs and
satellite data (Buckley 2004). Although satellite data have found extensive
application in the study of terrestrial environments, limited research has been carried
out to explore the use of these data in aquatic environments (Nelson et al. 2003).
The CLLAMM Dynamic Habitat
46
Nonetheless, these data have been used for mapping topography underwater (Lafon
et al. 2002) by establishing a relationship between spatial reflectance and water
parameters including depth (Gordon and Brown 1973), turbidity, and bottom colour
(Lee et al. 1998; Maritorena 1996). In particular, Lafon et al. (1998) demonstrated a
strong relation between reflectance from the water surface and depth.
The availability of bathymetric data for most of the North Lagoon, and the satellite
imagery for the region, provided essential ingredients for modelling bathymetry for
the Murray Mouth and the South Lagoon. Therefore, we attempted to derive
bathymetric data for these areas to supplement the DEM for the region by modelling
the satellite reflectance data and the bathymetric data in the surrounding areas. A
popular non-parametric technique, Generalized Additive Modelling (GAM – Hastie
and Tibshirani 1990) was used for establishing a relationship between depth and
surface reflectance captured in the different bands of satellite imagery as explanatory
variables. Details of the available datasets and methodology for deriving the
bathymetry and the seamless integration of the DEMs are given in the following
sections.
Figure 3.1. Coorong (North Lagoon and South Lagoon), Lower Lakes (Lake Alexandrina and
Lake Albert) and Murray Mouth region with major locations and the barrages. The Australian
map (A) shows the state boundaries and the study area in South Australia (SA) while the
inset map (B) highlights the study area shown in the main map.
The CLLAMM Dynamic Habitat
47
3.3.
Available datasets and methodology
3.3.1 Available topographic and bathymetric data for the region
DEM for South Australia
A DEM for South Australia derived from NASA’s Shuttle Radar Topography Mission
(SRTM) project was obtained from the Spatial Information Services group (SIS) of
the Department of Primary Industries of South Australia (PIRSA). Each grid cell in the
model represents approximately 83 m2 (3 arc seconds of latitude/longitude), and the
vertical resolution or elevation for each cell is to the nearest one metre with reference
to the Australian Height Datum (AHD). This model contains many areas without
elevation data and the SIS of PIRSA are trying to improve the model by incorporating
elevation data obtained from other sources. While the dataset has reasonable
matching with the topographic data for the state (SIS 2005), a number of problems
exist. Water bodies are not represented by their true elevation, the coastlines are not
perfectly precise in the model (NASA 2005), and very flat areas tend to display the
influence of striping as an artefact of the radar altimetry. The DEM for the Coorong
and surrounding areas including the Lower Lakes, derived from the SRTM dataset, is
shown in Figure 3.2.
Figure 3.2. Digital Elevation Model derived from NASA’s STRM project for the Coorong and
surrounding areas.
The CLLAMM Dynamic Habitat
48
Bathymetry for the Lower Lakes and the Coorong
The South Australian Water Corporation (SA Water) collected depth data for the
Lower Lakes, including Lakes Alexandrina and Albert and the northern part of the
Coorong, in order to derive a bathymetry for the region. The data was provided as
point data at varying resolutions in AutoCAD DXF format. Metadata for the dataset is
not available. However, Miles (2006) had prepared a note about the data collection
procedure and accuracy of the bathymetric dataset. This note contains the following
information about the dataset:
•
The Lakes and the North Coorong were surveyed using an echo-sounder and
GPS mounted on a small boat in May 2004.
•
The spacing of transects was narrow around Hindmarsh Island and the
Coorong as compared to the spacing in the Lakes.
•
The eco-sounder was not used for surveying areas less than one metre deep.
•
To compliment the echo-sounder data in shallow areas, the shoreline around
the Lakes was delineated from aerial photographs and assigned the same
value as the average water level of the Lakes.
The echo-sounder data and the shoreline contour data were processed in a
Terramodel HDMS (Hydrographic Data Management System) and used to generate
an interpolated bathymetry with the geographic coordinate system in decimal
degrees of latitude and longitude, and vertical datum referenced to Australian Height
Datum (AHD), for the Lakes and the North Lagoon. Except for a few rectangular
blocks in Lake Albert that had a resolution of 5 m2, the Lakes had 100 m2 resolution.
The North Lagoon had a resolution of 25 m2 in the area between the Murray Mouth
and Mark Point, and 50 m2 between Mark Point and Parnka Point (Figure 3.1). The
Goolwa Channel had a resolution of 10 m2. The vertical accuracy of the original
depth data collected by the echo-sounder is believed to be +/- 10 cm (Miles 2006).
However, the accuracy of the boat position, the effects of motion on the depth and
position readings, and the effects of tidal variation on the interpolated bathymetry, are
not known.
Depth data for the Coorong
A dataset containing depth data along 51 transects with 486 points across the lagoon
was obtained from CSIRO Land and Water. This data was originally collected by SA
Water. Approximately half of these transects are in the South Lagoon with 241
points. The data collection procedure and accuracy of the dataset is not known.
However, this dataset has been used for hydrodynamic modelling in the South
Lagoon (Webster 2007).
3.3.2 Satellite Imagery
LANDSAT5 Imagery
LANDSAT5 imagery (mosaic) collected during 2004 using the Thematic Mapper (TM)
sensor, and covering the CLLAMM region, was purchased from the MapLand section
of the SA Department for Environment and Heritage (DEH). The imagery was
already georeferenced to the Geocentric Datum of Australia 1994 (GDA 94).
Electromagnetic energy reflected from the earth’s surface is captured by the sensor
in seven different bands at a resolution of 25 m2. Each band corresponds to a
specific range of wavelengths in the electromagnetic spectrum, and thus captures
different reflectance from the target surfaces. For visualisation, combinations of
bands are often used in Red, Green and Blue colours for differentiating land cover
types, whether vegetation, rock or water. Reflectance from the water surface is
sensed primarily in the visible spectrum (band 1 to band 3) of LANDSAT TM imagery.
In the near and mid infrared regions (bands 4, 5 and 7), water mostly absorbs
incident energy and thus provides minimal reflectance. Band 6 represents the
The CLLAMM Dynamic Habitat
49
thermal infrared, and is uniquely heat sensitive (Short 2008). LANDSAT data have
been used in numerous studies to derive information on a range of parameters in
aquatic and marine environments, including water quality (turbidity, chlorophyll, water
chemistry, etc), water depth, presence of submerged vegetation and algal blooms,
and others.
SPOT5 Imagery
Five pansharpened SPOT5 images encompassing the Coorong were obtained from
MapLand, DEH. These images had been georeferenced to the same coordinate
system as the LANDSAT5 imagery. The SPOT5 images were collected in 2004 by
the high resolution geometry (HRG) optical sensor mounted on the satellite.
Pansharpening is carried out to enhance the spatial resolution of the multispectral
images, without altering the spectral information, through blending the lower
resolution multispectral bands with a higher resolution panchromatic image over the
same area (Pohl and Van Genderen 1998). Because of the high resolution and wide
area coverage (up to 60 x 120 km), SPOT5 imagery is widely used in environmental
monitoring, oil and mineral exploration, and urban and water resources planning (SIC
2008).
3.3.3 Methodology
Input data preparation
The bathymetric data, and the reflectance values or spectral digital numbers (DNs)
from the satellite images for the areas on either side of the Murray Mouth and
Mundoo Channel, were used for DEM modelling for the Murray Mouth area. Firstly, a
geospatial framework with a consistent horizontal coordinate system was ensured by
re-projecting the bathymetry dataset referenced to GDA 94, in line with the other
datasets. The elevation data were already referenced to AHD. Secondly, the
LANDSAT5 and SPOT5 datasets were converted from the raster format into vector
format as point shape files, with the values for each pixel being allocated to a single
point at the centre of the pixel. The reflectance point values in the seven bands of
LANDSAT5 data were extracted for the region and joined together in ArcGIS. The
point spectral data for the SPOT5 bands were at 2.5 m spacing and the LANDSAT5
spectral point data had 25 m spacing. Hence, the nearest spectral point in the
SPOT5 data was joined to the LANDSAT5 data. Finally, the spectral point data were
joined to the closest depth data point in the bathymetry dataset. The final dataset had
6067 points with 10 reflectance values and corresponding depth. For modelling and
validation purposes, the dataset was randomly divided into a “training” dataset and a
“testing” or validation dataset comprising 80 and 20 percent of the original dataset,
respectively.
A separate model was developed for the South Lagoon using the reflectance of data
from the areas adjacent to Parnka Point, assuming that these areas would have
similar reflectance to the South Lagoon. The same procedure was used to derive the
dataset as in the case of the Murray Mouth. However, the reflectance data from the
SPOT5 imagery could not be used in the model because the areas used for the
model development and the South Lagoon were covered by two separate SPOT5
images from different dates that showed very different reflectance. The dataset for
the South Lagoon had 12977 points with depth data and seven predictor variables
derived from the spectral values from the seven bands of the LANDSAT5 imagery.
The data were again randomly divided into training (80%) and testing (20%)
datasets, with 10381 and 2596 points for model development and validation
purposes, respectively.
Modelling technique: Generalized Additive Model (GAM)
A Generalized Additive Model (GAM) fits a smooth non-parametric relationship
between the response variable and the explanatory variable(s) (Crawley 2002).
The CLLAMM Dynamic Habitat
50
Unlike parametric methods, where the shape of the curve fitted is defined a-priori, the
shape of the non-parametric function is defined entirely by the data (Lehmann 1998).
A general GAM formula for predicting a response variable y at location i with
predictive variables xij is given as Equation 3.1.
g ( yi ) = β 0 +
n
∑S
i =1, j =1
j
( xij )
(3.1)
where g is a link function, yi is the response variable, β0 is the intercept and Sj is a
spline smoother for the predictor variable xij.
The GAM model was applied to the data using the mgcv package in the R software
package (Wood and Augustin 2002), with the data being modelled as following a
Gaussian distribution with the identity link function. A spline smoother was used for
estimating the smooth relationship between the response and predictor variables.
The model applied a penalized regression spline to guarantee a smooth fit by
imposing a penalty to avoid overfitting (Wood and Augustin 2002). The smoothness
of each fit is described by the estimated degrees of freedom for each predictor
variable, which is chosen by minimizing the Generalized Cross Validation (GCV)
score (Wood 2008). The package provides a summary of the model, which includes
estimated degrees of freedom and the statistical significance of each predictor
variable, adjusted R2, deviance explained by the model, and the GCV values. Models
with higher values of adjusted R2 and deviance, and lower values of GCV, are
considered to provide a better fit to the data. Since there is no automatic model
selection function in the mgcv package, Wood and Augustin (2002) recommended
using backward selection by manually deleting variables with degrees of freedom
close to 1, a 95% confidence region that included zero everywhere in the parameter
space, or that reduce the GCV score.
The reflectance values from the LANDSAT5 and SPOT5 imagery were used as
explanatory variables for predicting bathymetry in the Murray Mouth region according
to Equation 3.2.
g(Depth) = β0 + S1(1s1) + S2(ls 2) + S3(ls 3) + S4(ls 4) + S5(ls 5) +
S6(ls 6) + S7(ls7) + S7(sp1) + S8(sp2) + S9(sp3)
(3.2)
where g is the identity link function, and ln and sp represent the reflectance values for
the respective bands in the LANDSAT5 and SPOT5 imageries.
For the South Lagoon bathymetry modelling, the spectral values from the seven
bands of the LANDSAT5 imagery were used as explanatory variables in the model.
Model selection and validation
Although Wood and Augustin (2002) recommended the backward selection method
for implementing GAM in the mgcv package, the forward selection approach has
frequently been used in ecology for generating a model with the least number of
significant variables which could explain almost the same variance as the model with
all predictors(Blanchet et al. 2008). However, traditional forward selection based
purely on adding variables until no additional variable is regarded as statistically
significant is well known for producing a model with an excessive number of predictor
variables. Hence, in implementing forward selection, we use the procedure
developed by Blanchet et al. (2008), whereby an additional stopping criterion is
specified. This criterion is that the adjusted R2 should not exceed that of the global
The CLLAMM Dynamic Habitat
51
model with all possible predictor variables. Both the forward and backward
approaches are applied to the data for conclusive model selection. These selection
methods are briefly described below.
Forward selection: Initially, a model with all possible predictor variables was fitted to
obtain the global adjusted R2, and to avoid inflation of the Type I error rate (Blanchet
et al. 2008). We then used this value as one of the stopping criteria during model
selection. Variable entry into the model was based on the F statistic, providing that
p<0.05 and that the variable did not have degrees of freedom close to 1 or a 95%
confidence region that included zero everywhere in the parameter space (Wood and
Augustin 2002). This was repeated in a stepwise fashion until either no additional
variables met the entry criteria, or the global adjusted R2 was reached.
Backward selection: Backward selection is implemented in the opposite fashion to
forward section. We started with the global model and then eliminated terms in a
stepwise fashion. The variable with the lowest F value was first eliminated, providing
it had a p>0.05. This was repeated in a stepwise fashion until no additional variables
met the exit criteria.
The prediction accuracy of the ‘best fit’ model derived through each variable selection
procedure was assessed by comparing the predicted and the measured values in the
validation dataset. The root mean square error, a commonly used measure of vertical
accuracy for United States Geological Survey (USGS) DEMs (Wechsler 2003) was
used to assess the prediction accuracy of the model (Equation 4.3). The model with
the highest prediction accuracy would have a low root mean square error. Based on
the prediction accuracy, and model comparison using analysis of deviance, the best
model was selected and applied to the data for predicting bathymetry in the Coorong.
N
Root Mean Square Error =
∑ (P − M )
i =1
i
i
2
(3.3)
where N is the number of measured data, P is the predicted value and M is the
measured data.
Seamless DEM for the Coorong and surroundings
A seamless DEM was created for the Coorong and the surrounding area by using
DEMs originating from the above three different sources. The bathymetric point data
for the Lakes and the North Lagoon were separated based on their resolution and
subsequently converted into a raster DEM using the point to raster conversion tool in
ArcGIS 9.3 (Environmental Systems Research Institute 2008). The resultant DEMs
with different resolutions were resampled to 25 m2 resolution using the bilinear
resampling method in ArcGIS 9.3. DEMs for the Lower Lakes and the North Lagoon
were then mosaiced into a new DEM. The DEMs for the Murray Mouth and the
South Lagoon derived from the GAMs were joined to the DEM for the North Lagoon
and the Lower Lakes to produce a complete DEM for the Coorong, Lower Lakes and
Murray Mouth (CLLAMM). For the merger of all DEMs, a procedure for developing a
seamless DEM from the topographic and bathymetric data for Tampa Bay in the USA
was followed (Gesch and Wilson 2001). The DEMs derived from the bathymetric
data were integrated into one raster with a 25 m2 grid cell resolution. A buffer zone of
600 m around the bathymetric DEM was extracted from the SRTM DEM and
converted from grid to point data with location attributes (XY) and elevation value (Z).
The bathymetric points along the shorelines above 0 mAHD were selected and
combined with the topographic data. This new dataset contained the elevation data
for the areas encompassing the interface between the shoreline and the land
surface, and ensured a seamless merger of the elevations from the topographic and
The CLLAMM Dynamic Habitat
52
bathymetric datasets. For a smooth merger of the topographic and bathymetry data,
a thin plate spline interpolation technique available in the Radial Basis Function of
Geostastistical Analyst in ArcGIS 9.3 was used (Environmental Systems Research
Institute 2008). Out of the 600 m buffer zone around the bathymetric DEM, a region
300 m out from the shoreline was cropped to avoid interpolation edge effects in the
final bathymetry and mosaic with the bathymetric DEM. Finally, a seamless DEM for
the region was derived by merging the resultant bathymetric DEM and SRTM DEM
using the Mosaic function in ArcGIS.
3.4.
Results
3.4.1 Generalized Additive Model implementation and prediction error
analysis
Murray Mouth
The global model had an adjusted R2 of 0.882. The model included 2 SPOT5 bands
and all 7 LANDSAT5 bands as predictor variables (Table 3.1). Forward selection
identified the “best” model that included 2 SPOT bands and all 7 LANDSAT5 bands,
although only SPOT bands 2 and 3 explained >2% of the variation in depth (Table
3.1). The four-variable model explained most of the variance (87.7%). Adding the fifth
and subsequent variables to the model resulted in very small increases in total
variance explained. The predictive powers of all these models were evaluated by
comparing the estimated and measured depths for the validation dataset. The best
predictive power was given by the models with 4-7 variables. Applying the principle of
parsimony, the model with four variables was chosen as the model had the least
number of variables and explained more than 99% of the variance explained by the
global model. This model was applied to predict the bathymetry of the Murray Mouth
region.
Table 3.1. Statistics associated with forward selection for depth prediction in the Murray
Mouth region.
Model
Name
Variable
added1
GCV2 score
Model 1
sp2
0.811
Model 2
sp3
Model 3
Variance
explained %
Root Mean
Square Error
0.68
68.4
0.87
0.400
0.84
84.5
0.60
ls1
0.362
0.86
86.0
0.58
Model 4
ls3
0.321
0.88
87.7
0.55
Model 5
ls5
0.314
0.88
88.0
0.55
Model 6
ls6
0.311
0.88
88.1
0.55
Model 7
ls7
0.310
0.88
88.2
0.55
Model 8
ls2
0.306
0.88
88.4
0.72
Model 9
ls4
0.306
0.88
88.4
0.70
Adjusted R2
1
The variable name in this column indicates the new variable added to the previous
model (e.g. model 1 is sp2, model 2 is sp2 + sp3, etc). sp2 = SPOT5 imagery band 2;
sp3 = SPOT5 imagery band 3; ls1 = LANDSAT5 band 1; ls2 = LANDSAT5 band 2; ls3 =
LANDSAT5 band 3; ls4 = LANDSAT5 band 4; ls5 = LANDSAT5 band 5; ls6 =
LANDSAT5 band 6; and ls7 = LANDSAT5 band 7. 2GCV = Generalized CrossValidation.
The relationships between depth and the explanatory variables for the four-variable
model are illustrated in Figure 3.3. The solid line is the smooth function of the
The CLLAMM Dynamic Habitat
53
explanatory variable, while the dashed lines indicate the 95% confidence region.
SPOT5 band 2 was monotonically related to depth, with higher values indicative of
greater depths. SPOT5 band 3 showed the opposite pattern. LANDSAT5 band 1
was negatively correlated to depth up to a value of 110, after which the relationship
became more erratic with a greater confidence interval. LANDSAT5 band 3 showed
a quadratic relationship with depth.
Figure 3.3. Response curves for predictor variables and depth for the final four-variable model
for the Murray Mouth region. The solid line is the smooth function of the explanatory variable,
while the dashed lines indicate the 95% confidence region.
South Lagoon
The global model with all seven bands of LANDSAT5 had an adjusted R2 of 0.757.
Both the forward and backward variable selection methods identified a “best” model
that included 6 LANDSAT5 bands, although only 3 of these explained >2% of the
total variation in depth (Table 3.2).
The CLLAMM Dynamic Habitat
54
Table 3.2. Model statistics for depth prediction in the South Lagoon.
Model
Name
Variable
added
GCV2
score
Adjusted
R2
Variance
explained %
Root Mean
Square Error
Model 1
ls3
0.259
0.68
67.6
0.50
Model 2
ls2
0.222
0.72
72.4
0.46
Model 3
ls6
0.205
0.74
74.4
0.44
Model 4
ls5
0.200
0.75
75.2
0.44
Model 5
ls1
0.197
0.76
75.6
0.43
Model 6
ls4
0.196
0.76
75.8
0.43
Model 7
ls7
0.195
0.76
75.8
0.43
1
The variable name in this column indicates the new variable added to the previous
model (e.g. model 1 is ls3, model 2 is ls3 + ls2, etc). ls1 = LANDSAT5 band 1; ls2 =
LANDSAT5 band 2; ls3 = LANDSAT5 band 3; ls4 = LANDSAT5 band 4; ls5 =
LANDSAT5 band 5; ls6 = LANDSAT5 band 6; and ls7 = LANDSAT5 band 7. 2GCV =
Generalized Cross-Validation.
Again, the predictive power of all models was evaluated by comparing the estimated
and measured depths for the validation dataset. Model 3 with three predictor variable
produced the lowest prediction errors in terms of the root mean square error, and
thus the model was chosen for predicting the bathymetry of the South Lagoon.
The relationship between depth and each of the predictor variables in the final 3
variable model is shown in Figure 3.4. LANDSAT5 bands 2 and 3 showed a
quadratic relationship with depth, and 6 had roughly monotonic influence.
Figure 3.4. Response curves for predictor variables and depth for the final six-variable model for the
South Lagoon. The solid line is the smooth function of the explanatory variable, while the dashed
lines indicate the 95% confidence region.
Figure 3.5 presents the relationship between predicted depth and measured depth in
the South Lagoon between Parnka Point and Salt Creek. The predicted depths were
very close to the measured depths between Parnka Point and Jack Point. However,
the depths were underestimated further south towards Salt Creek. The summary
statistics for the measured and predicted depths (Table 3.3) also demonstrate that
the modelled depths were underestimated relative to the measured depths. Given
that the model was developed using data from the North Lagoon near Parnka Point,
this result makes sense, as water quality is known to deteriorate moving south from
Parnka Point.
The CLLAMM Dynamic Habitat
55
4.0
Measured depth
3.0
Predicted depth
Depth (mAHD)
2.0
1.0
0.0
-1.0
-2.0
-3.0
-4.0
1
11
21
31
41
51
61
71
81
91 101 111 121 131 141 151 161 171 181 191
Depth locations between Parnka Point and Salt Creek
Figure 3.5. Predicted bathymetry for the South Lagoon compared to measured data.
Table 3.3. Summary statistics for the measured and associated predicted depths in the South
Lagoon.
Depth
Mean
Standard
Error
Median
Standard
Deviation
Minimum
Maximum
Count
Measured
-0.98
0.08
-1.11
1.11
-3.33
3.56
195
Predicted
-0.74
0.06
-1.1
0.87
-1.81
2.41
195
DEM for the Murray Mouth and the South Lagoon
Model 4 was used to predict the bathymetry of the Murray Mouth region. A DEM for
the Murray Mouth and surrounding region was then developed by converting the
bathymetric data into a raster in ArcGIS 9.3. The predicted bathymetry for the Murray
Mouth region was smoothed by applying a low pass filter using a 3X3 window (Figure
3.6), and then mosaiced into a DEM including the surrounding regions (Figure 3.7).
The predicted bathymetry worked well in capturing the trend along the main channel
and in the near-shore areas.
The CLLAMM Dynamic Habitat
56
Figure 3.6. Predicted bathymetry for the Murray Mouth region derived from the final 4 variable
model after smoothing by applying a low pass filter with a 3X3 window.
Figure 3.7. Mosaic of the bathymetry for the Murray Mouth and surroundings after
smoothing by applying a low pass filter with a 3X3 window.
Figure 3.8 shows the bathymetry for the South Lagoon derived from the final 3
variable model, after smoothing the surface by applying a low pass filter with a 3X3
window.
The CLLAMM Dynamic Habitat
57
Figure 3.8. Predicted bathymetry for South Lagoon derived from the final 3 variable model after
smoothing by applying a low pass filter with a 3X3 window.
3.4.2 Seamless bathymetry for the Coorong
Figure 3.9 presents an integrated bathymetry for the Coorong, Lower Lakes and
Murray Mouth. This bathymetry was finally merged with the SRTM DEM to develop a
seamless DEM for the CLLAMM region (Figure 3.10), by following the procedure
described by Gesch and Wilson (2001).
The CLLAMM Dynamic Habitat
58
Figure 3.9. Final bathymetry for the Coorong, Lower Lakes and Murray Mouth.
The CLLAMM Dynamic Habitat
59
Figure 3.10. Seamless DEM for the Coorong, Lower Lakes and Murray Mouth (CLLAMM) region.
3.5.
Discussion
The bathymetry modelling for the Murray Mouth and the South Lagoon using satellite
imagery produced a reasonable prediction of depths for both of these areas where
bathymetry was previously unknown. However, the depth for the southern quarter of
the South Lagoon was underestimated, which could be directly attributed to poor
reflectance because of less clear or turbid water. The water in the Murray Mouth
region tends to be clearer due to the tidal incursion of sea water and the water clarity
progressively diminishes towards the South Lagoon. The Salt Creek area at the far
end of the South Lagoon has virtually no tidal influence and the water is very turbid.
Water clarity was particularly important in the South Lagoon where visibility was very
low (<0.64m secchi depth at Parnka Point compared to 2 m in the North Lagoon and
Murray Mouth, Ye (2008), unpublished data). The influence of water clarity on
The CLLAMM Dynamic Habitat
60
reflectance is obvious in the higher accuracy of the depth predictions in the Murray
Mouth region compared to the South Lagoon.
The bathymetry models for the Murray Mouth and the South Lagoon were not perfect
and could be improved in a number of ways. Access to more comprehensive ground
survey data for these regions would enable improvement in the errors and
uncertainty in these models. Various techniques of handing errors and uncertainty
are described in many papers (Hengl et al. 2004; Wechsler 2003; Fisher 1998). One
commonly cited means of improving prediction would be to include a variable
accounting for water turbidity, which may enhance the predictive power of the model.
However, the turbidity data would need to align with the dates of satellite overpass,
and so while some turbidity data is presently available for the Coorong, it would be
difficult to use in this case.
The integration of the DEMs derived from different sources with different resolutions
and accuracy is a challenging task (Buckley 2004; Mitchell et al. 2002). Bathymetry
and topographic DEMs from three different sources with varying horizontal and
vertical resolution and accuracy were used for deriving a seamless DEM for the
Coorong and the surrounding areas. The interpolated bathymetric data for the Lakes
and the North Lagoon of the Coorong collected using an echo-sounder had different
spatial resolutions in different areas and the vertical accuracy was believed to be +/10 cm. By comparison, the modelled bathymetry for the Murray Mouth and the South
Lagoon had a resolution of 25 m2 and with mean absolute errors of 0.39 m and 0.33
m, respectively. Finally, the topographic DEM was derived through radar altimetry
from NASA's SRTM (Shuttle Radar Topography Mission) project with approximately
83 m2 resolution and one metre vertical accuracy. Nevertheless, a seamless DEM for
the region has been successfully derived through integrating these datasets.
However, the main issue remains the vertical resolution of the bathymetry compared
to the topographic DEM. Although a seamless DEM was derived by integrating these
two datasets, the elevation at the water-land interface may not be accurately
interpolated because of the large difference in the vertical resolution between the
datasets. Additional ground survey data on the land surface adjacent to the
shoreline to 1 cm vertical resolution would enhance the accuracy of the model, and it
is hoped that this will be achieved through future LIDAR altimetry efforts. LIDAR
technology has improved greatly in recent years, and modern experimental systems
are able to map bathymetry to at least 40 m depth with a vertical accuracy of 0.15 m
(Lohani, 2009). This technology could be the best option for acquiring a high
resolution bathymetry for the South Lagoon, particularly in areas where an ecosounder cannot be used, although it is fairly expensive and not currently widely
available.
The final seamless DEM was produced as a raster model with 25 m by 25 m
resolution, which enables users to employ the dataset in various applications
requiring elevation/bathymetry data. However, the vertical accuracy of the dataset
differs from those of the parent datasets and varies spatially according to the source
and the density of the bathymetric dataset.
3.6.
Summary
The reflectance values from the LANDSAT5 and SPOT5 bands were used to
generate a bathymetry for the Murray Mouth and the South Lagoon in order to
address the existing data gap and to generate a seamless bathymetry for the
Coorong and the region. The Generalized Additive Model predicted reasonable
bathymetry for the Murray Mouth and the South Lagoon. However, the depth for the
southern quarter of the South Lagoon was underestimated, which could be directly
attributed to poor reflectance because of less clear or turbid water. The accuracy of
the predicted bathymetry could be improved by applying a correction based on
ground survey data and also using water turbidity as a predictor, especially in the
The CLLAMM Dynamic Habitat
61
model for the South Lagoon. Finally, a seamless DEM for the Coorong and the
surrounding areas was derived by merging bathymetry and topographic DEMs from
three different sources with varying horizontal and vertical resolution and accuracy.
The final seamless DEM was produced as a raster model with 25 m by 25 m
resolution and would be very useful for various studies requiring elevation data.
3.7.
References
Blanchet, F.G., Legendre, P. and Borcard, D. (2008) Forward selection of
explanatory varibles. Ecology 89: 2623-2632.
Buckley, S.J. (2004) Integration, validation and point spacing optimisation of digital
elevation models. The Photogrammetric Record 19: 277-295.
CLLAMM (2007) The future of the Coorong, Lower Lakes and Murray Mouth
(CLLAMM). Retrieved 15/01/2008, from http://www.csiro.au/science/ps260.html.
Crawley, M.J. (2002) An introduction to data analysis using S-Plus. John Whiley and
Sons Ltd., Chichester.
Environmental Systems Research Institute (2008), ArcGIS 9.3 Desktop. ESRI,
Redlands, California.
Fisher, P. (1998) Improved modeling of elevation error with geostatistics.
GeoInformatica 2: 215-233.
Geddes, M.C. (2003) Survey to investigate the ecological health of the North and
South Lagoons of the Coorong, June/July 2003. SARDI Aquatic Sciences and
University of Adelaide, Adelaide.
Geddes, M.C. (2005) The ecological health of the north and south lagoons of the
Coorong in July 2004. SARDI Aquatic Sciences and University of Adelaide, Adelaide.
Gesch, D. and Wilson, R. (2001) Development of a seamless
bathymetric/topographic elevation model for Tampa Bay. Second Biennial Coastal
GeoTools Conference, Charleston.
Gordon, H.R. and Brown, O.B. (1973) Irradiance reflectivity of a flat ocean as a
function of its optical properties. Applied Optics 12: 1549-1551.
Hengl, T., Gruber, S. and Shrestha, D. P. (2004) Reduction of errors in digital terrain
parameters used in soil-landscape modelling. International Journal of Applied Earth
Observation and Geoinformation 5(2): 97-112.
Lafon, V., Froidefond, J.M., Lahet, F. and Castaing, P. (2002) SPOT shallow water
bathymetry of a moderately turbid tidal inlet based on field measurements. Remote
Sensing of Environment 81(1): 136-148.
Lee, Z., Carder, K.L., Mobley, C.D., Steward, R.G. and Patch, J.S. (1998)
Hyperspectral remote sensing for shallow waters: I. A semianalytical model. Applied
Optics 23: 181-190.
Lehmann, A. (1998) GIS modeling of submerged macrophyte distribution using
generalized additive models. Plant Ecology 139: 113-124.
The CLLAMM Dynamic Habitat
62
Maritorena, S. (1996) Remote sensing of water attenuation in coral reefs: a case
study in French Polynesia. International Journal of Remote Sensing 17: 155-166.
MDBC (2006) The Lower Lakes, Coorong and Murray Mouth asset environmental
management plan 2005/2006. Murray-Darling Basin Commission, Canberra.
Miles, M. (2006) Origin of bathymetric data for Lower Lakes, Goolwa and Coorong.
South Australian Department of Environment and Heritage, Adelaide
Mitchell, H. L., Fryer, J. G. and Pâquet, R. (2002) Integration and filtering of 3D
spatial data using a surface comparison approach. International Archives of the
Photogrammetry, Remote Sensing and Spatial Information System 34: 644-649.
NASA (2005) SRTM C-BAND DATA PRODUCTS. Retrieved 12/08/2008, from
http://www2.jpl.nasa.gov/srtm/cbanddataproducts.html.
Nelson, S. A., Soranno, P. A., Cheruvelil, K. S., Batzli, S. and Skole, D. L. (2003)
Regional assessment of lake water clarity using satellite remote sensing. Journal of
Limnology 62: 27-32.
Pohl, C. and Van Genderen, J.L. (1998) Multisensor image fusion in remote sensing:
concepts, methods and application. International Journal of Remote Sensing 99:
823-854.
Short, N. M. (2008) The Remote Sensing Tutorial. Retrieved 17/07/2008, from
http://rst.gsfc.nasa.gov/.
SIC (2008) SPOT-5 Satellite Imagery and Sensor Characteristics. Retrieved
12/07/2008, from http://www.satimagingcorp.com/satellite-sensors/geoeye-1.html.
SIS (2005) NASA Radar Shuttle 3 second DTM. Spatial Information Services, South
Australian Department of Primary Industries and Resources, Adelaide.
Webster, I. T. (2007) Hydrodynamic modelling of the Coorong. Water for a Healthy
Country National Research Flagship, CSIRO, Canberra.
Wechsler, S. P. (2003) Perceptions of digital elevation model uncertainty by DEM
users. The URISA Journal 15 (2): 58-64.
Wood, S. N. (2008) MT1007. Fisheries Assessment. Retrieved 15/02/2008, 2008,
from http://www.maths.bath.ac.uk/~sw283/SIP/fish.pdf.
Wood, S.N. and Augustin, N.H. (2002) GAMs with integrated model selection using
penalized regression splines and applications to environmental modelling. Ecological
Modelling 157(2-3): 157-177.
The CLLAMM Dynamic Habitat
63
4. Sediment mapping of the Coorong: Implications for
habitat distributions
Sunil K. Sharma1*, Milena B. Fernades1, Jason E. Tanner1 and Simon N. Benger2
1
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
2
School of Geography, Population and Environmental Management, Flinders
University, GPO Box 2100, Adelaide, SA 5001
*corresponding author, Phone +61 (8) 8207 5448, Fax +61 (8) 8207 5448, E-mail:
[email protected]
4.1.
Executive Summary
The Coorong, Murray Mouth and Lower Lakes region is renowned for its unique
biological diversity, providing habitat for more than 200 species of birds and fish. The
area is viewed as highly significant for its social, economic and ecological values at
local, national and international levels. The current low water level in the Lower
Lakes has been threatening the delivery of fresh water into the system, which is
necessary to maintain a salinity gradient and water levels conducive to ecological
health. The Coorong, Lower Lakes and Murray Mouth Ecology Research Cluster
(CLLAMMecology) aimed to develop a better understanding of the impact of the flow
regime on the functioning of the system for ecological sustainability of the region.
One of the objectives of the Dynamic Habitat Project within the CLLAMMecology
Research Cluster was to produce sediment maps to enhance our understanding of
benthic habitat distribution in the Coorong. The boundaries of sediment distribution
and biological communities are, however, subjective and sediment maps should be
seen as a first step in defining the distribution of habitats. In this work, we mapped a
series of sediment characteristics influencing benthic habitat distribution along the
Coorong, including sediment textural characteristics, as well as organic carbon and
nitrogen, and mineral composition.
Maps of the distribution of mean grain size, sorting, organic carbon and nitrogen
were developed for the estuary and North Lagoon using generalized additive models.
Sediment transport in the Coorong was assumed to be linked to tidal incursions at
the Murray Mouth, and also the quality and quantity of water flow through the
barrages. As a consequence, maps were produced taking into account distance from
the Murray Mouth, the underwater topography, distance to shore, and also salinity,
amongst other variables. As no detailed bathymetry data is available for the South
Lagoon, a simpler approach was used to map all parameters, including inorganic
carbon and gypsum, for the whole Coorong. In this case, inverse distance weighting
(IDW) was used, a mathematical method for surface fitting using the weighted
averages of nearby points.
Our results suggest three main depositional areas along the Coorong, where
sediments are fine and organically-enriched: (1) the middle channel of the lagoons,
(2) the constriction between the North and South Lagoons known as Parnka Point,
and (3) the western (seaward) shores of the North Lagoon, particularly south of Long
Point. The sediment maps give the spatial distribution of sediment attributes and are
useful tools in helping to identify habitats for benthic fauna and foraging grounds for
migratory birds. This information will help managers to make informed decisions
about the allocation of limited resources for gaining the maximum benefit from the
system.
The CLLAMM Dynamic Habitat
64
4.2.
Introduction
The Coorong, Murray Mouth and Lower Lakes region offers a unique array of
habitats from freshwater, estuarine and marine to hypersaline, and provides an
interconnected ecosystem supporting more than 200 species of birds and fish,
including 20 species of migratory waders (CLLAMM 2007; EconSearch 2004;
Edyvane 1999; Carpenter 1995). The area is viewed as highly significant for its
social, economic and ecological values at local, national and international levels
(Department for Environment and Heritage 2007). The international significance of
the region is demonstrated through its listing as a Ramsar site, while its national
significance has been recognized by its designation as one of the six icon sites under
The Living Murray initiative (Seaman 2003). The 2008 low water level in the Lower
Lakes, attributed to the drought in the Murray River tributaries and the level of
extractions in the Murray Darling Basin, has interrupted the delivery of fresh water to
the system, which is necessary to maintain a salinity gradient and water levels
conducive to ecological health. At present, the water level in the Coorong is at an
historic low, and salinity has risen to extremely hypersaline in the southern part of the
lagoon, beyond the physiological limits of many species, which now seek refuge in
the northern part of the lagoon (CLLAMM 2008). The Coorong, Lower Lakes and
Murray Mouth Ecology Research Cluster (CLLAMMecology) aimed to develop a
better understanding of the impact of the water flow regime on the functioning of the
system, to provide management options for ecological sustainability of the region.
At the ecosystem level, physical, chemical and biological processes and the
interactions between them govern system function and sustainability (Mann 1991).
Physicochemical conditions are crucial determinants of the range of habitats
available for flora and fauna. In the Coorong, the presence of mudflats on the eastern
(landward) and western (seaward) shores and frequent inundation of these areas
provides good habitat for invertebrates, and hence offers substantial foraging
grounds for a large population of migratory birds (Edyvane 1999). Habitat
characterization in subtidal areas is less clearly defined, but key drivers are water
level, salinity, temperature and sediment characteristics (CLLAMM 2007 2008; Jones
1950).These physico-chemical variables drive the recruitment of larvae of sessile
fauna and the proliferation of aquatic vegetation (e.g. Ruppia spp.) (Heijs et al. 2000;
Snelgrove and Butman 1994), thus having a significant effect on the availability of
food resources for the wider system.
The protection of critical aquatic habitats requires a detailed definition of their spatial
distribution. One of the objectives of the Dynamic Habitat Project within the
CLLAMMecology Research Cluster was to produce sediment maps to enhance our
understanding of benthic habitat distribution in the Coorong. The mapping of benthic
habitats has been traditionally based on geophysical and sediment data, due to the
difficulties in acquiring biological data at the scale of the landscape (Kenny et al.
2003; Roff et al. 2003). The boundaries of sediment distribution and biological
communities are, however, subjective, and sediment maps should be seen as a first
step in defining the distribution of habitats (McBreen et al. 2008). In this work, we
mapped a series of sediment characteristics influencing benthic habitat distribution
along the Coorong, including sediment textural characteristics, as well as organic
carbon and nitrogen content, and mineral composition. Particle size distributions
were used as indicators of prevailing hydrodynamic conditions through the
delineation of depositional and erosional areas (Storlazzi and Field 2000). The
organic carbon and nitrogen content of sediments was investigated as a proxy for
organic matter accumulation, thus reflecting the availability and origin of food
resources for benthic detritivores (Ruttenberg and Goñi 1997). The effects of salinity
on sediment composition were assessed through the measurement of gypsum and
carbonate (inorganic carbon), which reflect changes in the saturation state of the
water column driven by evaporation and water circulation (Lazar et al. 1983).
The CLLAMM Dynamic Habitat
65
Maps of sediment distribution in the Coorong were produced by using surface
interpolation techniques (Méar et al. 2006; Leecaster 2003). The method used is
dependent upon the amount of data on sediments, its variability and trends, and also
the availability of other datasets that would potentially correlate with the variable of
interest. Maps of the distribution of mean grain size, sorting, organic carbon and
nitrogen were developed for the estuary and North Lagoon using generalized additive
models (GAM) (Wood 2003; Hastie and Tibshirani 1990). Sediment transport in the
Coorong was assumed to be linked to tidal incursions at the Murray Mouth, and also
the quality and quantity of water flow through the barrages (Webster 2005). As a
consequence, maps were produced taking into account distance from the Murray
Mouth, the underwater topography, distance to the eastern and the western shores,
and also salinity, amongst other variables. As no detailed bathymetric data are
available for the South Lagoon, a simpler approach was used to map all parameters,
including inorganic carbon and gypsum, for the whole Coorong. In this case, inverse
distance weighting (IDW) was used, a mathematical method for surface fitting using
weighted averages of nearby points (Environmental Systems Research Institute
2001).
These sediment maps give the spatial distribution of sediment attributes and are
useful tools for predicting the availability of habitat for benthic fauna and foraging
grounds for migratory birds. This information contributes to a better understanding of
the distribution of the areas of high ecological value for these species that will provide
managers with the ability to make informed decisions about the allocation of limited
resources for gaining the maximum benefit from the system.
4.3.
Methods
4.3.1 Study area
The Coorong is a large lagoon system on the southeast coast of South Australia, and
is geographically located between the Lower Lakes (Lake Alexandrina and Lake
Albert) and the Southern Ocean (Figure 4.1). Barrages were built at five locations,
Goolwa Channel, Mundoo Channel, Boundary Creek, Ewe Island and Tauwitchere to
prevent sea water entering into the Lakes. The lagoon begins at the Murray Mouth
and stretches southeast for over 100 kilometres, parallel to the ocean (CLLAMM
2007). The Younghusband Peninsula, a series of 1-3 kilometre wide sand dunes,
separates the lagoon from the ocean.
4.3.2 Sampling
Twelve sites were sampled along the length of the Coorong in October and
November 2006, and March 2007, with three cross-lagoon transects separated by
500 m per site (Figure 4.1). These transects were also surveyed using underwater
video (see Chapter 2), and the sites have been the subject of intensive investigation
by other groups in CLLAMMecology (Noell et al. 2009; Rogers and Paton 2009a;
Rogers and Paton 2009b; Rolston and Dittmann 2009). Typically, five sets of
sediment cores were collected at approximately even intervals along each transect
using 71 mm (internal diameter) PVC tubes capped with rubber bungs. The overlying
water in the tube was carefully discarded to minimise surface disturbance and the top
layer sliced. Samples for carbon, nitrogen and gypsum analysis (n = 178) comprised
the top 0-1 cm of the cores and were transferred into pre-combusted glass jars.
Samples for particle size analysis were obtained from additional cores (n = 178) and
comprised the top 0-4 cm of the cores. These samples were transferred into
aluminium trays. All samples were transported on ice and stored frozen at -20ºC.
The CLLAMM Dynamic Habitat
66
Figure 4.1. Locations of the reference sites in the Coorong Lagoon: 1 = Goolwa; 2 =
Mundoo; 3 = Barker Knoll; 4 = Ewe Island; 5 = Pelican Point; 6 = Mark Point; 7 = Long Point;
8 = Noonameena (in the North Lagoon); 9 = Parnka Point; 10 = Villa dei Yumpa; 11 = Jack
Point and 12 = Salt Creek (in the South Lagoon). The Australian map (A) shows the state
boundaries and the study area in South Australia (SA) while the inset map (B) highlights the
study area shown in the main map.
Particle size
Samples were oven-dried at 105oC for at least 16 h. Samples were dispersed with a
50 g L-1 sodium hexametaphosphate solution in an ultrasound bath for 15 min then
left to soak overnight, and sonicated again for 15 min (Bouyoucos, 1962; Buchanan,
1984). Dispersed samples were wet sieved to exclude particles >1 mm just before
analysis. Particles >1 mm typically comprised less than 3% of the total and were
discarded (Fernandes, unpublished results). Particle size analyses were performed
using a Malvern Mastersizer 2000 laser diffraction analyser, which has a practical
measuring range from 0.02 to 2000 μm. Particle size distributions were analysed with
the software package GRADISTAT, with the Folk and Ward graphical method used
to calculate grain size parameters (Blott and Pye 2001). This method is relatively
insensitive to samples with a large particle range in the tails of the distribution and
provides a robust tool to compare compositionally variable samples. Parameters
used to describe grain size distributions included the mean grain size, the spread of
sizes around the mean (sorting), the symmetry or preferential spread of the
The CLLAMM Dynamic Habitat
67
distribution to one side of the mean (skewness), and the degree of concentration of
the grains relative to the average (kurtosis).
Carbon and nitrogen
Samples were freeze-dried, sieved to remove large shell fragments >500 μm, and
homogenized with a mortar and pestle. These sieved samples were analysed for
carbon and nitrogen in a LECO TruSpec elemental analyser. Sub-samples for
organic carbon analysis were pre-treated with 1 N hydrochloric acid to remove
carbonates, rinsed with MilliQ water to remove hygroscopic salts and oven-dried at
50ºC for at least 16 h using a method modified from Fernandes and Krull (2008).
Organic carbon was calculated from concentrations measured in pre-treated and
untreated aliquots; inorganic carbon was calculated as the difference between total
carbon and organic carbon.
Gypsum
Samples were freeze-dried, sieved to remove large shell fragments >500 μm, and
homogenized with a mortar and pestle. Gypsum (CaSO4.2H2O) was determined
using the thermogravimetric method of Artieda et al. (2006), which has a
quantification limit of 2%. Approximately 10 g was sequentially dried at 70ºC and
90ºC, and gypsum determined as the loss in weight between these temperatures,
assuming a water loss of 14.95% for pure gypsum. Gypsum values <2% were
considered as 1% for the sake of spatial and statistical analysis.
Statistical analysis of measured data
Particle size parameters, as well as carbon, nitrogen and gypsum content of
sediments, were analysed with the software package STATISTICA (StatSoft, Tulsa,
OK). Principal-component analysis (PCA) was used as an exploratory technique to
identify broad sediment types according to particle size parameters. PCA was based
on standardized data in the correlation matrix, with principal components retained
when eigenvalues were >1. Sediments types were assigned according to the loading
of individual samples on the first two principal components. Analysis of variance
(one-way ANOVA) was used to determine the significance of observed differences (α
= 0.05) in carbon, nitrogen and gypsum content and C:N molar ratios between broad
sediment types (fine, intermediate and coarse), and location along the Coorong
(estuary, North and South Lagoons), both factors treated as fixed. Variables were
log-transformed when there was a need to improve normality as indicated by normal
probability plots. Tukey post-hoc tests were performed when significant differences
were detected. Linear regressions were used to assess whether and how a given
variable was related to other variables.
4.3.3 Modelling and mapping
Sediment modelling for the North Lagoon
Generalized additive modelling (GAM) was chosen to find the relationship between
the sediment characteristics measured and a range of potential predictor variables
(described below) for the northern part of the Coorong. A detailed description of the
model and the model selection procedures are given in Chapter 3. The GAM
including all variables for predicting sediment attributes is expressed as follows:
g(SA) =
β + s(Northing) + s(Easting) + s(Depth) + s(Slope) +
s(Aspect) + s(distance to the eastern shore) + s(distance
to the western shore) + s(distance to the nearest shore) +
s(distance from the Murray Mouth) + s(salinity)
(4.1)
where g is the identity link function, SA is the sediment attribute of interest, β is
the fitted intercept and s is a spline smoother for each predictor variable.
The CLLAMM Dynamic Habitat
68
The predictor variables were first analysed for inter-correlations. Table 4.1 presents
the correlation matrix for all ten variables. The geographical coordinates (Easting and
Northing), were highly negatively correlated (r = -0.98). Due to the NW-SE orientation
of the Lagoon, the distance to the Murray Mouth had a high positive (r = 0.97) and
negative (r = -98) correlation with the Easting and the Northing, respectively. Depth
had a very negligible correlation (r = <-0.08) with its two derivative variables (slope,
aspect) and with the other variables. The salinity gradient along the Lagoon, i.e. low
salinity (35 g/L) at the northern end (Goolwa Channel) and high salinity (~100 g/L) at
the southern end (Salt Creek), resulted in a high positive correlation (r = 0.97) with
the Easting and the distance from the Murray Mouth, and high negative correlation (r
= -0.98) with the Northing. The rest of the variables did not show strong correlation
with other variables, except for a negative correlation (r = -0.54) between the
distance to the eastern and the western shores. As we are using the GAM for
predictive modelling within the range of the predictor variables used to develop it,
collinearity in the predictor variables is not a problem, although it would be if we were
undertaking hypothesis testing (Quinn & Keough 2002)
Table 4.1. Correlation matrix for the predictor variables used in the GAM for modelling
sediment attributes in the Coorong.
Variable
1
2
3
4
5
6
7
8
9
1
1.00
2
-0.98
1.00
3
0.09
-0.10
1.00
4
-0.39
0.37
-0.08
1.00
5
0.19
-0.20
-0.03
-0.15
1.00
6
0.97
-0.98
0.07
-0.37
0.20
1.00
7
0.23
-0.27
0.26
-0.15
0.08
0.23
1.00
8
0.07
0.01
0.06
0.05
-0.16
0.04
-0.54
1.00
9
0.24
-0.20
-0.30
-0.37
0.07
0.21
0.05
0.11
1.00
10
0.97
-0.98
0.14
-0.37
0.17
0.97
0.33
0.07
0.23
10
1.00
Where 1 = Easting, 2 = Northing, 3 = Depth, 4 = Slope, 5 = Aspect, 6 = Distance from the
Murray Mouth, 7 = Distance to the eastern shore, 8 = Distance to the western shore, 9 =
Distance to the nearest shore and 10 = Salinity.
For modelling sediment attributes all ten predictor variables were used as long as
sufficient degrees of freedom were available. In the case of nitrogen, the input
dataset only had 67 sample points as some samples did not contain enough material
for analysis (either nitrogen was below the detection limit or the sample was limited in
size), and thus there were insufficient degrees of freedom available to fit all 10
variables. As Easting, Northing, Distance from Murray Mouth and Salinity were all
highly correlated (Table 4.1), only Salinity was retained (as the variable explaining
the highest % deviance in total nitrogen of this group). This left a 7 variable global
model, which was explored further.
GAM was implemented using the MGCV package in R-stat software (Wood and
Augustin 2002; Wood 2001). The final model was determined using forwards
stepwise selection, as proposed by Blanchet et al. (2008). This approach uses a
global statistic as a stopping criterion in order to find a model with the minimum
number of predictor variables that can explain almost the same variance as the
global model with all predictor variables. The global adjusted R2 was used as the
The CLLAMM Dynamic Habitat
69
stopping criterion for modelling sediment characteristics. For each variable, the
generalized cross validation (GCV) score was used to determine the optimum
degrees of freedom (smoothness of the fit) to use. Variables were then added to the
model on the basis of their F-statistic (after accounting for existing variables in the
model), until either no additional variables contributed significantly to the model fit, or
the global stopping criterion was reached. Analysis of variance (ANOVA) was also
performed to test for significant differences between the final model and the global
model.
Insufficient data were available to retain a subset for model validation purposes.
Hence all sample data were used for model development, and a predicted value was
then determined for each sample point. The goodness of fit of the final model, as well
as its precursors and the global model, was assessed by comparing the predicted
and the measured values for each sample point. A common measure of prediction
error, the root mean square error (Equation 4.2) was used to compare the prediction
power of the different models (Verfaillie et al. 2006; Antonopoulos et al. 2001) and
the model with the least error was applied to the data for the North Lagoon for
generating sediment attribute maps:
N
Root Mean Square Error =
∑ (P − M )
i =1
i
2
i
(4.2)
where N is the number of measured data, P is the predicted value for sample i
and M is the corresponding measured value.
Datasets for the generalized additive model
(a) Topographic variables
The underwater topography of the lagoon was considered to be a major factor
influencing sediment distribution in the Coorong. The lagoon is about 2-3 km wide
and is characterized by a deep channel in the middle and shallower areas towards
the eastern (landward) shore, whereas the western (seaward) shore is mostly steeply
sloping. An interpolated bathymetry (in vector format-point data) for the lakes and the
North Lagoon was obtained from the South Australian Water Corporation (SA Water).
The underlying data for this bathymetry were collected using an echo-sounder and
global positioning system (GPS) installed in a small boat in May 2004. The depth
data were referenced to the Australian Height Datum (AHD) and the accuracy of the
original dataset was assumed to be ± 10 cm (Miles 2006).
Slope and aspect also describe underwater topography and might have influenced
the distribution of sediment along the lagoon. Slope measures the steepness or
gradient of a surface in degrees between 0 (horizontal) and 90 (vertical), while aspect
is the compass direction of hill faces and is also measured in degrees between 0
(north) and 360 (north). These layers were derived from the digital elevation model
developed from the bathymetry data (see Chapter 3) using the Spatial Analyst tool in
ArcGIS (v 9.2) (Environmental Systems Research Institute 2008). The aspect layer in
degrees was lineralised by applying Equation 3 which also removed negative values,
as ArcGIS represents flat areas as having a slope of –1 rather than 0 (Menzel et al.
2006):
Aspect = (1-cos(aspect in degree)) + (1-sin(aspect in degree))
The CLLAMM Dynamic Habitat
(4.3)
70
(b) Geographical variables
Sediment distribution in estuaries is influenced by currents and the particle size of the
sediment itself. Coarse sediments require stronger currents for transportation and
settle more quickly than fine sediments (McLusky, 1981). Hence, fine sediments are
likely to be transported further from their source. In the Coorong, the major sources
of sediment include the Southern Ocean (via water flow through the Murray Mouth),
the Lower Lakes and River Murray (via water flow over the barrages) and the
adjacent shores via windblown transport and surface runoff. The western shore
adjoins active sand dunes on the Younghusband Peninsula, which are likely to be a
minor source of aeolian transported sediments. Thus, the distance to the Murray
Mouth, the eastern (landward) and western (coastal) shores, and the distance to the
nearest shore, were used in the model. These layers were derived by estimating the
distance to the nearest respective boundary (landward or coastal shore) from each
point in the bathymetry dataset in ArcGIS 9.3 (Environmental Systems Research
Institute 2008).
.
Geographic location has also been considered an important predictor variable for
explaining spatial variability in sediments (Dray et al. 2006). To take geographic
location into account, the spatial coordinates (easting and northing) were used as
explanatory variables in the model.
(c) Salinity
Salinity is one of the main ecological drivers of the Coorong and exerts a strong
influence on physical and biological processes (CLLAMM 2007). Hence, salinity (g/L)
was also included as a predictor variable in the model. Sampled data measured at 10
locations along the Coorong in June 2007 were used to derive a GIS-layer for salinity
using the IDW method (Burrough and McDonell 1998) in ArcGIS 9.3.
Sediment mapping for the North Lagoon
A dataset with all predictor variables was created for the entire North Lagoon. The
distance from the Murray Mouth, eastern shore, western shore and the nearest shore
were calculated for each point in the bathymetry data set, which consisted of
geographic co-ordinates (Easting and Northing) and depth. The raster datasets of
aspect, slope and salinity were converted into point format and joined to the
bathymetry.
The best model for each sediment attribute was applied to the entire dataset for the
North Lagoon and a predicted value for each data point obtained. The predicted
values were imported into GIS and joined with the respective geographical
coordinates. Finally spatial maps for each sediment attribute were derived by
interpolating the predicted data using the IDW method. All these operations were
performed in ArcGIS 9.3.
Sediment mapping for the Coorong as a whole
To produce sediment maps for the entire Coorong including both the North and the
South Lagoons, the IDW method available in the geostastical analyst extension of
ArcGIS 9.3 was used (Environmental Systems Research Institute 2008). The sample
size (Table 4.2) was not adequate to capture variability within the sites or across the
entire lagoon through the use of any of the standard geostatistical methods. Although
kriging is a commonly-used geostatistical method and has demonstrated a good
prediction capacity for sediment (Méar et al. 2006; Leecaster 2003) and soil
properties (Liu et al. 2006; Gotway et al. 1996), this method resulted in a radical
smoothing of the data and also could not depict the spatial variability of mean grain
size within individual sites or across the lagoon. However, the IDW method, an exact
interpolation, broadly captured the variability in the mean grain size at the regional
scale and had better goodness of fit than another exact interpolation method, the
radial basis function (thin-plate spline method) (Burrough and McDonell 1998).
The CLLAMM Dynamic Habitat
71
Hence, IDW was chosen to generate maps for all sediment parameters for the
Coorong.
Table 4.2. Sample size for sediment attributes in the North Lagoon and the entire Coorong.
Number of samples
Sediment Attributes
Coorong as a
whole1
North Lagoon only
Grain size and sorting
166
117
Organic and inorganic Carbon
144
112
Nitrogen
113
67
Gypsum
154
110
1
The total number analysed varied as a function of the amount of sediment available once
the samples were dried, and detection limits for the different tests.
IDW is a deterministic method and calculates the weighted average for each known
point in the neighbourhood as the inverse of the distance from the point to be
predicted (Longley et al. 2002). The number of neighbours, neighbourhood shape,
orientation and the form of the weighting function were tested to choose the best
parameters for interpolation. A model with a minimum of 5 and a maximum of 10
neighbours in a four-sectored neighbourhood oriented at 450 with an optimized power
value minimised the root mean square error. The model attempted to use the
specified number of data points (minimum 5 and maximum 10 neighbours) from
every sector. If five neighbours (the minimum) were not available within each sector,
the nearest points outside the sector were also used. The 450 orientation of the
neighbourhood captured a general trend in the data because of the NW-SE
orientation of the lagoon. The power value for distance determined the rate of weight
decay for the points in the neighbourhood. Higher power values rapidly reduced the
weights of points farther away from the prediction location and vice versa. The
optimum power value used in the model was chosen to give the prediction with
minimum root mean square error and was automatically determined by the software
(Environmental Systems Research Institute 2008).
4.3.4 Goodness of fit of GAM and IDW
A comparison of the goodness of fit of GAM and IDW was performed by evaluating
the root mean square errors derived from the difference between predicted and
measured values. Equation 2 was used for calculating the root mean square errors
for all sediment attributes for the GAM predictions while the values were directly
obtained from implementing the IDW model in the ArcGIS.
4.4.
Results
4.4.1 Measured sediment attributes
Particle size distribution
Sediments showed a wide range of particle size distributions, which were broadly
distributed into three main groups according to the grain size parameters of
individual samples (Figures 4.2 and 4.3):
(1) The finest sediments had high sorting and low kurtosis values (n = 36;
Figures 4.3a,c) resulting from a mixture of particles of different (albeit small)
The CLLAMM Dynamic Habitat
72
size classes. The sediments in this group consisted of muds and muddy
sands showing a unimodal or bimodal distribution, with either a fine mode
dominated by silts (~18μm) or a coarse mode dominated by fine sands
(~187μm) or both (Figure 4.2b). Forty-seven percent of these samples were
from deep sediments ≥1 m accumulating in the middle channel of the
lagoons, and 36% from intermediate depths of 0.5-1 m. The remaining few
samples in this group, or 17% of the total, came from shallow depths along
the western (seaward) shores mostly between Noonanema and Parnka
Point. These fine sediments were well represented throughout Parnka Point,
where 60% of samples fell into this group.
(2) Sediments with intermediate mean grain size and sorting values (n = 53;
Figure 4.3a) were predominant between Goolwa and Barker Knoll,
historically the estuarine part of the Coorong. Sixty-four percent of all
samples falling into this group came from this part of the Coorong (Figure
4.2a). These samples had low skewness and high kurtosis values (Figure
4.3b,c), indicative of a more homogenous group of particles when compared
to the fine sediments described above. The sediments in this group were
generally sands and muddy sands with a unimodal distribution centred on
fine sands (~209μm) (Figure 4.2b).
5
Estuary
North Lagoon
South Lagoon
4
16
intermediate
14
3
13
2
coarse
12
coarse
11
1
10
Volume (%)
Factor 2: 35.71%
intermediate
15
0
-1
9
8
7
6
-2
5
4
-3
-4
fine
3
fine
2
1
-5
-5
-4
-3
-2
-1
0
1
2
3
4
5
0
0.01
0.1
1
10
100
Factor 1: 52.04%
(b)
(a)
Figure 4.2. Principal Component Analysis (PCA) showing sediments according to mean
grain size, sorting, skewness and kurtosis (a), and the typical particle size distribution of
each group identified in the PCA (b).
The CLLAMM Dynamic Habitat
1000
Particle Size (µm)
73
7
Intermediate
Fine
Coarse
Other
6
Sorting ( m)
5
4
3
2
1
0
50
100 150 200 250 300 350 400 450
Mean (μm)
(a)
0.4
0.2
Skewness
0.0
-0.2
-0.4
-0.6
-0.8
0
50
100 150 200 250 300 350 400 450
Mean (μm)
(b)
3.4
3.2
3.0
2.8
2.6
2.4
Kurtosis
2.2
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0
50
100 150 200 250 300 350 400 450
Mean (μm)
(c)
Figure 4.3. Mean vs sorting (a), skewness (b) and kurtosis (c) values according to the
sediment groups identified in Figure 4.2. Samples labelled as ‘other’ were not clearly
separated by the principal components in Figure 4.2.
The CLLAMM Dynamic Habitat
74
(3) The coarsest sediments had low sorting, high skewness and intermediate
kurtosis values (n = 72; Figure 4.3) as a consequence of a negligible
contribution from fine size classes (Figure 4.2b). Sediments in this group
were classified as sands with a unimodal distribution centred on medium
sands (~253μm). The particle size distribution characteristic of this group
was the most common throughout the Coorong lagoons. These sediments
were particularly predominant along the eastern (landward) shores of the
North Lagoon, typically found at shallow depths <0.5 m (35%) and
intermediate depths of 0.5-1 m (26%). This sediment type was also
widespread throughout the South Lagoon.
Organic carbon and total nitrogen
The distribution of organic carbon and total nitrogen content of sediments broadly
followed particle size characteristics, with the highest values found in the fine
sediments and lowest in the coarse sediments described in the previous section
(Figure 4.4a). The results of ANOVA indicate a significant difference between these
sediment types for both organic carbon (F2,133=63.252, P<0.001) and total nitrogen
(F2,99=49.504, P<0.001). The slope of the linear regression between these variables
suggests C:N molar ratios between 10 (when the intercept is set to zero) and 11
(Figure 4a), suggesting an important planktonic or bacterial component to
sedimentary organic matter (Ruttenberg and Goni, 1997). Fine sediments had
significantly higher C:N ratios than the other sediment types (Figure 4b; ANOVA
F2,78=17.293, P<0.001). There were 162 samples analysed for organic carbon. Of
these, 11 samples showed inconsistently high organic carbon results when
compared to total carbon, leading to the calculation of negative inorganic carbon
values; these samples were excluded from the dataset.
14
6
12
4
10
3
C:N
Organic carbon (%)
5
Intermediate
Fine
Coarse
Other
2
6
OC = -0.2 + 9.1 * N
r2 = 0.91
1
0
0.0
8
0.1
0.2
0.3
0.4
0.5
0.6
4
0.7
Nitrogen (%)
2
Intermediate Fine
Coarse
(a)
(b)
Figure 4.4. Organic carbon vs nitrogen content of sediments according to the particle size
groups identified in Figures 4.2 and 4.3 (a), and C:N molar ratios for the same groups (b).
Values in (b) are the mean, boxes represent standard error and bars standard deviation.
Samples labelled as ‘other’ were either not analysed for particle size, or were not clearly
separated by the principal components in Figure 4.2.
The CLLAMM Dynamic Habitat
75
Inorganic carbon and gypsum
The results of ANOVA indicate a significance difference in inorganic carbon values
(F2,148= 83.132, P=<0.001) and gypsum (F2,154=31.609, P<0.001) between the
estuary, North and South Lagoons. Inorganic carbon values were lowest in the North
Lagoon and peaked in the South Lagoon, with intermediate values in the estuary
(Figure 4.5a). In the estuary, the highest inorganic carbon values were associated
with coarse sediments. Gypsum values also increased along the salinity gradient,
with the highest (and most variable) values recorded in the South Lagoon (Figure
5b). Gypsum was significantly enriched in fine sediments (ANOVA F2,139=14.969,
P<0.001).
18
7
16
6
14
12
10
4
Gypsum (%)
Inorganic carbon (%)
5
3
2
8
6
4
2
1
0
0
-1
-2
Estuary
North Lagoon
South Lagoon
-4
(a)
Estuary
North Lagoon
South Lagoon
(b)
Figure 4.5. The content of inorganic carbon (a) and gypsum (b) in sediments of the estuary
(n=55), North Lagoon (n=59) and South Lagoon (n=49). Values are reported as the mean, boxes
represent standard error and bars standard deviation.
4.4.2 GAM for North Lagoon
Mean grain size
The model with six-variables met the stopping criterion and stood as the final model
for mean grain size. Table 4.3 summarizes all the models tested, their variance
statistics, and prediction errors. The global model with all 10 predictor variables had
an adjusted R2 of 0.462 and a GCV score of 5874.8. The Northing and distance to
the eastern shore explained almost equal variability of 15% each. Depth was the third
most important variable and contributed 8.1% variability while the fourth variable,
distance to the western shore, explained slightly above 9%. The fifth and sixth
variables, slope and aspect contributed 6.4% and 3.4%, respectively. Forward
selection identified the “best” model as having six predictor variables (distance to the
eastern shore, depth, slope, northing, distance to the western shore and aspect).
This model had a lower cross validation score (GCV) and higher adjusted R2 of 0.474
than the global model. The analysis of variance did not show a significant difference
between the six-variable model and the global model. The prediction error measured
by the root mean square value improved with the addition of each new variable into
the model. The final model had the best prediction of mean grain size, as indicated
The CLLAMM Dynamic Habitat
76
by the lowest root mean square error, even lower than the global model. The sixvariable model was considered to be the best model for predicting mean grain size
distribution in the North Lagoon.
Square Error
Root Mean
explained (%)
Variance
R-Square
Adjusted
GCV3 score
added1
Variable
Model Name
Table 4.3. Forward selection models for mean grain size: summary of model parameters,
variance statistics and prediction error.
Model 1
east_dist
8004.7
0.116
15.0
85.32
Model 2
depth
7273.5
0.199
23.1
81.19
Model 3
northing
6391.2
0.329
38.6
72.54
Model 4
west_dist
5989.7
0.401
47.9
66.82
Model 5
slope
5760.6
0.451
54.3
62.55
Model 6
aspect
5781.1
0.474
57.7
60.18
Model 72
All
5874.8
0.462
57.0
62.71
1
The variable name in this column indicates the new variable added to the previous
model (e.g. model 1 is east_dist, model 2 is east_dist + depth, etc). east_dist = Distance
to the eastern shore, west_dist = Distance to the western shore, 2Global model with all
10 predictor variables; 3GCV = Generalized Cross-Validation.
The relationship between mean grain size and the explanatory variables in the sixvariable model is illustrated in Figure 4.6. Sediment becomes finer to about 500 m
distance to the eastern and western shores (Figure 6a and 6d) and with depth
(Figure 4.6b). The Northing had a complex influence on sediment grain size (Figure
4.6c). As slope increased to approximately 3.5 degrees, sediments became finer,
whereas further increases in the slope resulted in sediments becoming coarser
(Figure 4.6e). Aspect had no clear effect on grain size (Figure 4.6f).
The CLLAMM Dynamic Habitat
77
Figure 4.6. Response curves for predictor variables and mean grain size for the final six-variable model
for the North Lagoon. The solid line is the smooth function of the explanatory variable, while the dashed
lines indicate the 95% confidence region.
Sorting
A global model with all 10 predictor variables was implemented for sorting, a
measure of the spread of sizes around mean grain size. A five-variable model (depth,
distance to the eastern shore, easting, northing and slope) slightly exceeded the
adjusted R2 of 0.353 for the global model (Table 4.4). Depth alone explained about
30% of the variability, followed by distance to the eastern shore, which accounted for
nearly 6% of the variability. Northing explained about 5%, Easting ~ 3% and slope
<2% of the variabilityin mean grain size. This model also had the lowest GCV score
of all the models, and was considered the best model by the forward selection
method.
Analysis of variance indicated no significant difference between the final model and
the global model. Compared to its precursor models, the final model with five
predictor variables had the closest root mean square error of 0.773 to the global
model with error of 0.759. The best predictive power was found for the model with
five-predictor variables. Thus, this model was selected for predicting the distribution
of sorting values in the North Lagoon.
The CLLAMM Dynamic Habitat
78
Variance
explained %
Root Mean
Square Error
0.902
0.243
29.9
0.872
Model 2
east_dist
0.859
0.292
35.6
0.836
Model 3
easting
0.836
0.318
38.5
0.816
Model 4
northing
0.830
0.346
43.2
0.785
Model 5
slope
0.824
0.359
44.9
0.773
Model 62
All
0.870
0.353
46.9
0.759
Variable
added1
Adjusted
depth
GCV score3
Model 1
Model Name
R-Square
Table 4.4. Forward selection models for sorting: summary of model parameters, variance
statistics and prediction error.
1
The variable name in this column indicates the new variable added to the previous model
(e.g. model 1 is depth, model 2 is depth + east_dist, etc). east_dist = Distance to the eastern
shore; 2Global model with all 10 predictor variables; 3GCV = Generalized Cross-Validation.
The relationship between sorting and the explanatory variables in the final fivevariable model (Model 5) is illustrated in Figure 4.7. Sediments were increasingly
poorly sorted (high values) with depth (Figure 4.7a), distance to the eastern shore
(Figure 4.7b) and slope (Figure 4.7e). The sorting value also increased with both
Easting (moving from the west to east) and Northing (moving from the south to the
north), but the confidence intervals were large (Figures 4.7c,d).
The CLLAMM Dynamic Habitat
79
Figure 4.7. Relationship between the explanatory variables and sorting for the final five-variable model of
the North Lagoon. The solid line is the smooth function of the explanatory variable, while the dashed lines
indicate the 95% confidence region.
Organic carbon
A global model was run with all 10 predictor variables against sediment organic
carbon content. A summary of the model parameters, variance statistics and
prediction error is presented in Table 4.5. A three-variable model including depth,
slope and northing exceeded the global adjusted R2 of 0.529, making it the best
model found by the forward selection method. The first variable, depth, explained
much of the variability (44.7%) while the second and third variables, northing and
slope contributed 9.6% and 7.5% respectively. The rest of the variables explained
only 2.4% of the variability in total. This model also had a better GCV score than the
global model. The analysis of deviance did not show a significant difference between
these two models. The three-variable model identified by forward selection also had
the best goodness of fit due to the prediction error being very close to the model with
all predictor variables (Model 4). Thus, this model was selected for predicting organic
carbon distribution in the North Lagoon.
The CLLAMM Dynamic Habitat
80
Variance
explained %
Root Mean
Square Error
0.548
0.401
44.7
0.68
Model 2
northing
0.5063
0.477
54.3
0.62
Model 3
slope
0.483
0.533
61.8
0.56
Model 42
All
0.517
0.532
64.2
0.55
Adjusted
depth
GCV score3
Model 1
Model Name
R-Square
Variable added1
Table 4.5. Forward selection models for organic carbon: summary of model parameters,
variance statistics and prediction error.
1
The variable name in this column indicates the new variable added to the previous model
(e.g. model 1 is depth, model 2 is depth + slope, etc). 2Global model with all 10 predictor
variables; 3GCV = Generalized Cross-Validation.
The relationship between organic carbon and each of the predictor variables in the
final three-variable model is shown in Figure 4.8. Depth and slope show a cubic and
largely quadratic (except for large values where data are sparse) relationship with
organic carbon, respectively (Figures 4.8a,b). The highest organic carbon was
predicted at depths of about -2.9 m and the lowest at +0.17m (Figure 4.8a). The
northing shows a complex effect on organic carbon distribution in the North Lagoon
(Figure 4.8b).
Figure 4.8. Relationship between the explanatory variables and organic carbon for the model
with three variables in the North Lagoon. The solid line is the smooth function of the
explanatory variable, while the dashed lines indicate the 95% confidence region.
Total nitrogen
A global model with 7 variables was run against total nitrogen content. Table 4.6
presents a summary of the models tested with the forward selection method. Forward
selection identified the “best” model as having five-variables (depth, slope, distance
to the nearest shore, distance to the western shore, and distance to the eastern
shore). The first variable, depth, alone explained 44.7% of the variance and the
The CLLAMM Dynamic Habitat
81
subsequent three variables each added slightly more than 5%. The fifth variable
(distance to the eastern shore) explained a very low proportion of the variance (< 2
%). This model had the lowest GCV score and a higher adjusted R2 than the global
model. However, the difference between these two models was not statistically
significant.
Variance
explained %
Root Mean
Square Error
0.019
0.388
44.7
0.1584
Model 2
slope
0.0198
0.417
49.9
0.1576
Model 3
near_dist
0.0204
0.441
55.2
0.1569
Model 4
west_dist
0.02009
0.484
61.1
0.1623
Model 5
east_dist
0.01965
0.502
63.0
0.1623
Model 62
All
0.0219
0.497
64.5
0.1607
Variable
added1
Adjusted
depth
GCV score3
Model 1
Model Name
R-Square
Table 4.6. Forward selection models for nitrogen: summary of model parameters, variance
statistics and prediction error.
1
The variable name in this column indicates the new variable added to the previous model
(e.g. model 1 is depth, model 2 is depth + slope, etc). near_dist = Distance to the nearest
shore, west_shore = Distance to the western shore and east_dist = Distance to the eastern
shore; 2Global model with all 7 predictor variables; 3GCV = Generalized Cross-Validation.
Although the forward selection method found a five-variable model to be the best
model, the prediction error was slightly higher than the three-variable model with
depth, slope and distance to the nearest shore (Model 3). The single variable model
(Model 1) with depth as the only predictor variable also had lower prediction error
than the global model. Based on the lowest prediction error, the three-variable model
was selected for predicting nitrogen distribution in the North Lagoon.
The highest nitrogen content occurred at about -2.50 m, and lowest at -1.0 m (Figure
4.9a). Nitrogen content was highest when the slope was zero and lowest when it was
~ 2o (Figure 4.9b). High nitrogen content was also found at distances > 500 m from
the nearest shore.
The CLLAMM Dynamic Habitat
82
Figure 4.9. Relationship between three explanatory variables and nitrogen content in the North Lagoon.
The solid line is the smooth function of the explanatory variable, while the dashed lines indicate the 95%
confidence region.
Inorganic carbon
The global model with all 10 predictors had an adjusted R2 of 0.634. Forward
selection identified the “best” model as having four variables (northing, salinity,
distance to the nearest shore and distance to the western shore – see Table 4.7).
The northing and salinity explained 45 % and 13.9 % of the variability, respectively,
whereas the additional variables in Models 3 and 4 explained 7.1% and 4.3% of the
variability. This four-variable model had both a GCV score and adjusted R2 close to
that of the global model (Table 4.7). The analysis of deviance did not show a
significant difference between these two models.
Root Mean
Square Error
0.412
45
1.19
Model 2
salinity
1.41
0.528
58.9
1.03
Model 3
near_dist
1.35
0.579
66.0
0.93
Model 4
west_dist
1.21
0.627
70.3
0.87
Model 72
All
1.19
0.634
70.8
0.86
R-Square
1.63
Adjusted
northing
GCV score3
Model 1
Model Name
Variance
explained %
Variable added1
Table 4.7. Forward selection models for inorganic carbon: summary of model parameters,
variance statistics and prediction error.
1
The variable name in this column indicates the new variable added to the previous model
(e.g. model 1 is northing, model 2 is northing + salinity, etc). near_dist = Distance to the
nearest shore and west_shore = Distance to the western shore; 2Global model with all 10
predictor variables; 3GCV = Generalized Cross-Validation.
.
The CLLAMM Dynamic Habitat
83
Although Northing and salinity explained much of variability in the sample data, the
model with only these two variables had a prediction error higher than the fourvariable model, which had the lowest prediction error, close to the global model.
Hence, the four-variable model was chosen for predicting the distribution of inorganic
carbon in the North Lagoon.
Inorganic carbon content had a very complex relationship with both Northing and
salinity (Figures 4.10a,b), but decreased with distance to the nearest shore (Figure
4.10c). The inorganic carbon content gradually increased up to about 250 m from the
western shore and then decreased (Figure 4.10d)
Figure 4.10. Relationship between predictor variables and inorganic carbon content in the
North Lagoon. The solid line is the smooth function of the explanatory variable, while the
dashed lines indicate the 95% confidence region.
Gypsum
For gypsum distribution, the global model had an adjusted R2 of 0.195 and explained
only about 33 % of the deviance. As there is a very low correlation between the
predictors and gypsum distribution in the North Lagoon, a distribution map for
gypsum has not been generated. These results mainly reflect the low gypsum
content in the North Lagoon, mostly below the quantification limit of 2% for the
analytical method.
The CLLAMM Dynamic Habitat
84
4.4.3 Spatial maps for the North Lagoon and the Coorong
Particle size distribution
Sediment maps derived from GAM depicted general trends for mean grain size and
sorting distribution in the North Lagoon (Figures 4.11a,b). Fine sediments (< 100 µm;
see Figure 4.3) occupied the deep channel (below -2.0 mAHD) along the North
Lagoon and some areas on the western shore. Intermediate sediments (100 - 200
µm; see Figure 4.3) were mostly found in the areas adjacent to the deep channel and
predominated between Mark Point and Noonameena. Areas around the Murray
Mouth and adjacent to the eastern (landward) shores were mostly coarse sand (>
200 µm; see Figure 4.3) except for the north of the Murray Mouth.
Poor sorting values (> 3 µm) were mostly associated with fine and intermediate
sediments, while areas with coarse sand had better sorting (< 2 µm). Sediments on
the eastern (landward) shore were coarse and well sorted, whereas sediments
adjacent to the western (seaward) shore were fine to intermediate and poorly sorted.
Maps for mean grain size and sorting distribution for the entire Coorong generated by
the IDW method are presented in Figure 4.12. The distribution of both sediment
attributes is overly generalized and variability in the sample data is not reflected in
these maps. Intermediate sediments dominate overall, while coarse sediments are
found at the Murray Mouth region, areas between Pelican Point and Long Point and
to the south of Villa dei Yumpa (Figure 4.12a). In these maps, sediments have mostly
intermediate sorting (2-3 µm). The coarse sediments in the Murray Mouth, in the
small area to the north of Noonameena (on the eastern shore) and south of Villa dei
Yumpa, all have good sorting (< 2 µm). However, coarse sediments between Pelican
Point and Long Point have intermediate sorting values (2-3 µm) and a patch to the
north of Noonameena (on the western shore) and about 13 km to the north of Parnka
Point had poor sorting values (> 3 µm) (Figure 4.12b).
The CLLAMM Dynamic Habitat
85
Figure 4.11. Mean (μm) (a) and Sorting (μm) (b) distribution in the North Lagoon derived from the
generalized additive model. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker Knoll; EI
= Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena and PA
= Parnka Point. The sediment classes for the mean and sorting were approximately defined based
on the sediment groupings in Figure 4.2)
The CLLAMM Dynamic Habitat
86
Figure 4.12. Mean (μm) (a) and Sorting (μm) (b) distribution in the entire Coorong derived by the
inverse distance weighting method. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker
Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena;
PA = Parnka Point; VY = Villa dei Yumpa; JP = Jack Point and SC = Salt Creek. The sediment
classes for the mean and sorting were approximately defined based on the sediment groupings in
Figure 4.2)
Organic carbon and total nitrogen
The distribution of organic carbon and total nitrogen content of sediments are broadly
consistent with the sediment size distribution in the North Lagoon (Figure 4.13). High
values of organic carbon (> 2 %) and total nitrogen (> 0.3 %) were generally
associated with fine and intermediate sediments and low values with the coarse
sediments. The deep channel to the north of Ewe Island had mostly fine and
intermediate sediments, high organic carbon and total nitrogen contents. Coarse
sandy areas around the Murray Mouth had low organic carbon (< 1 %) but medium
total nitrogen content (0.11 - 0.3 %).
The CLLAMM Dynamic Habitat
87
Figure 4.13. Organic carbon % (a) and total nitrogen % (b) distribution in the North Lagoon derived
by the generalized additive model. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker
Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena
and PA = Parnka Point. The classes for the organic carbon and nitrogen were approximately
defined based on the sediment groupings in Figure 4.4).
The map generated by IDW for the whole Coorong predicts low organic carbon
content (<0.1 %) for both the North and South Lagoons, regardless of particle size
distribution. However, small patches of medium organic carbon content are evident at
some reference sites (Figure 4.14a). The map generated for total nitrogen content
predicts medium values (0.11 - 0.3 %) for the entire Coorong except for the areas
about 10 km to the north and 6 km to the south of Parnka Point (Figure 4.14b)
As in the particle size maps (Figure 4.12), the local influence of the sample data is
not clearly visible in organic carbon and total nitrogen maps (Figure 4.14), however, it
is seen in high resolution maps of the reference sites.
The CLLAMM Dynamic Habitat
88
Figure 4.14. Organic carbon (a) and total nitrogen content (b) distribution in the Coorong
derived by the inverse distance weighting method. GC = Goolwa Channel; MC = Mundoo
Channel; BK = Barker Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long
Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei Yumpa; JP = Jack Point and SC =
Salt Creek. The classes for the organic carbon and nitrogen were approximately defined based
on the sediment groupings in Figure 4.4).
Inorganic carbon and gypsum
High inorganic carbon content (> 3 %) was observed in the Murray Mouth region and
between Noonameena and Parnka Point, predominantly along the western shore.
Sediments with low inorganic carbon content (< 1%) were more likely to appear
between Ewe Island and Mark Point, whereas patches of low inorganic carbon also
existed at Goolwa Channel, Mundoo Channel, Ewe Island and to the north of
Noonameena. The area between Mark Point and Noonameena and to the north of
Parnka Point along the eastern shore showed medium inorganic carbon content (1 3 %) (Figure 4.15).
The CLLAMM Dynamic Habitat
89
Figure 4.15. Inorganic carbon % distribution in the North Lagoon derived by the generalized
additive model. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker Knoll; EI = Ewe
Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena and PA =
Parnka Point. The classes for the inorganic carbon were approximately defined based on the
sediment groupings in Figure 4.5).
In the maps of the whole Coorong, the South Lagoon is predicted to have high
inorganic carbon content while the North Lagoon shows medium inorganic carbon
between Parnka Point and Noonameena, around Ewe Island, Goolwa Channel and
Mundoo Channel. Sediments between Noonameena and Pelican Point and a patch
at Goolwa Channel had low inorganic carbon content (Figure 4.16a).
The prediction map for gypsum content in the whole Coorong is shown in Figure
4.16b. About half of the Coorong had high gypsum content (> 3 %) to the south of
Noonameena and the other half had medium (2.01 - 3 %) or low (< 2%) gypsum
content (Figure 4.16b).
The CLLAMM Dynamic Habitat
90
Figure 4.16. Inorganic carbon % (a) and Gypsum % (b) distribution in the Coorong derived by
the inverse distance weighting method. GC = Goolwa Channel; MC = Mundoo Channel; BK =
Barker Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM =
Noonameena; PA = Parnka Point; VY = Villa dei Yumpa; JP = Jack Point and SC = Salt Creek.
The classes for the inorganic carbon and gypsum were approximately defined based on the
sediment groupings in Figure 4.5).
4.4.4 Goodness of fit of GAM and IDW models
The goodness of fit of the GAM and IDW models was assessed by comparing the
root mean square errors derived from the differences between predicted and
measured values. Based on this parameter, GAM performed better than IDW for
predicting mean grain size, sorting, organic carbon, nitrogen and inorganic carbon in
the North Lagoon (Table 4.8).
The CLLAMM Dynamic Habitat
91
Table 4.8. Comparison of root mean square values for the GAM and IDW models.
Root Mean Square
Error by GAM
Root Mean Square
Error by IDW
Mean grain size
62.6
85.93
Sorting
0.76
1.1
Organic carbon
0.56
0.96
Total nitrogen
0.158
0.197
Inorganic carbon
0.77
0.99
Parameters
4.5.
Discussion
Our results suggest three main depositional areas along the Coorong, where
sediments are fine and organically-enriched: (1) the middle channel of the lagoons;
(2) the constriction between the North and South lagoons known as Parnka Point;
and (3) the western (seaward) shores of the North Lagoon, particularly south of Long
Point. The sediments accumulating in these areas have higher C:N ratios, suggesting
that these fine sediments are partially derived from terrestrial runoff or riverine inputs
(Ruttenberg and Goñi 1997). Deposition is likely driven by reduced wave action and
minimal sediment resuspension as a consequence of depth (middle channel), low
flows (Parnka Point), or protection from prevailing winds by the dune system
(western shores) (Webster 2005). While at deeper sites this accumulation of organic
matter might lead to patchy anaerobic conditions detrimental to the survival of fauna
(e.g. invertebrates) (Fernandes and Tanner 2009), shallower depositional areas
might constitute important feeding grounds for higher trophic levels (e.g. fish and
birds) (Geddes and Francis 2008; Rossi 2003; Bachelet et al. 1996). The high
salinities recorded at sites south of Long Point, however, are likely to shift secondary
production towards microbes, negating the ecological potential of these fine-grained
sediments (Rizzo et al. 1996). Further north, many of the depositional areas would
also have limited value as feeding grounds for wading birds because depth would
push them out of the foraging depth range for these animals.
The composition and size of sediments in the estuarine part of the Coorong suggests
some accumulation of fines, albeit at much lower rates when compared to the
depositional areas described above. Strong flooding tidal currents potentially act to
transport coarse coastal sands into the mouth and estuary, and reduce the
deposition of fines (Webster 2005; 2006). The areas with the strongest erosional
character along the Coorong lagoons, where deposition is less likely to occur, are
typically found at shallower depths on landward shores. These sites are more
exposed to the prevailing south-southwesterly winds (Bone 1990), resulting in the
winnowing out of fine organic-rich sediment fractions by wind-driven wave action.
The impact of the salinity gradient on sediment composition is clearly reflected in the
spatial distribution of inorganic carbon and gypsum. While coarse calcareous sands
from coastal waters are likely the cause for slightly higher inorganic carbon values in
the estuary when compared to the North Lagoon (Li et al. 1996), the extremely high
inorganic carbon values in the South Lagoon are indications of the precipitation of
carbonate at high salinities (Ford 2007). This hypothesis is corroborated by the
precipitation of gypsum in the same area, a product typical of evaporitic basins
(Caumette et al. 1994). The precipitation of inorganic minerals in the salinity gradient
is potentially large enough to explain some of the discrepancies found in the analysis
The CLLAMM Dynamic Habitat
92
of organic carbon. A significant fraction of the precipitated inorganic minerals are
likely lost during the acid washing step of sample treatment, leading to unreasonably
high organic carbon results and negative inorganic carbon values.
Between the two methods used to map sediment attributes in the Coorong, GAM
demonstrated better goodness of fit based on the root mean square error when
compared to IDW. However, the aim of this study was to produce sediment maps
from the sample data, hence the best method would be the one which produces
better spatial prediction of sediment attributes. As a consequence, the sediment
maps derived from both methods were further evaluated to scrutinize the spatial
prediction ability of these two methods (Leecaster 2003; Isasks and Srivastava
1989). GAM depicted a clear pattern for particle size (mean and sorting), organic and
inorganic carbon, and total nitrogen distribution in line with the selected variables in
the final models. Depth (bathymetry) was one of the highly significant variables for
predicting all sediment attributes except for inorganic carbon. This finding is
consistent with the deposition of fine and organic-rich sediments along the deep
channel, leaving coarse organic-depleted sediments to accumulate on flat shores. In
contrast, inorganic carbon content was mostly a function of position along the salinity
gradient.
IDW maps were highly generalized and did not show a specific trend or pattern at the
local level despite huge variations in the sample data across the Lagoon. These
maps also suffered from the inherent problem of IDW “bull’s eye” features in the
maps (De Smith et al. 2006), which could be attributed to the sparse distribution of
samples. These features were highly visible in the maps when sample data had
extremely high or low values compared to other samples in the neighbourhood.
When other geostatistical methods including kriging, co-kriging and radial basis
functions were tested for the sample data, none were able to capture the variability in
the system. Adequate sampling density is required to produce a sensible map with a
high level of confidence using these methods. Leecaster (2003) suggested a
sampling density of 1.4 per square km for Santa Monica Bay to obtain a high
confidence level for the kriging method. As the Coorong offers a diverse
geomorphology as well as underwater topography both across and along the
Lagoon, the samples collected at the 12 sites were limited in number for using the
geostatistical methods including IDW. However, resource constraints meant that
sampling had to be targeted at the 12 reference sites, and did not allow a sampling
program designed to map the sediments of the whole Coorong. Due to the
complexity and variability in the system, a dedicated spatial sampling design based
on the prior analysis of spatial correlation among the variables is necessary to obtain
accurate sediment maps for the Coorong (Caeiro et al. 2003).
Although GAM has been widely used for studying non-linear species-environment
relationships parameters (Jensen et al. 2005; López-Moreno and Nogués-Bravo
2005; Denis et al. 2002), so far as we are aware, this method has not previously
been used to characterise the spatial distribution of sediment properties in a complex
environment. Application of GAM to the North Lagoon generated spatial maps for all
five sediment attributes, capturing the regional as well as local variability in the
sample data based on the non-linear relationship between the sediment attribute and
the predictor variables selected in the models. Forward selection with a stopping
criterion (Blanchet et al. 2008) was a straight-forward method to select a model with
the least number of predictor variables with almost the same predictive ability as the
global model.
4.6.
Summary and Conclusions
GAM was able to explore the significant relationships between the predictor variables
and the sediment characteristics in the North Lagoon with a small number of sample
data, and generated sediment maps depicting local variability both across and along
The CLLAMM Dynamic Habitat
93
the Lagoon. The variability explained by the selected models ranged between 45 and
70%, which could be expected to rise greatly with more samples and a sampling
design targeted at mapping the entire Lagoon, rather than relying on samples that
were collected to map smaller subsections of the Lagoon, as is done here. The
sediment maps derived from GAM were spatially consistent with the physical
distribution of various sediment attributes in the North Lagoon. However, the maps
for the entire Coorong were generated by using the inverse distance weighting
method and were greatly generalized without capturing specific patterns of sediment
deposition in the Lagoon.
Along the Coorong Lagoons, the deposition of fine organic-rich sediments occurs in
deeper waters of the middle channel, at Parnka Point, and along western (seaward)
shores, particularly between Noonameena and Parnka Point. The potential ecological
value of these shallow depositional areas as feeding grounds for higher trophic levels
is negated by the extremely high salinities recorded south of Long Point. The steep
salinity gradient between the estuary and the South Lagoon also changes the
composition of sediments, as indicated by intense precipitation of carbonates and
gypsum. The chemical changes in the water column driving mineral precipitation are
likely to have a strong influence on shifting the balance between autotrophs and
microbes at the base of the food chain.
4.7.
References
Antonopoulos, V., Papamichail, D. and M., K. (2001) Statistical and trend analysis of
water quality and quantity data for the Strymon River in Greece. Hydrology and Earth
System Sciences 5: 679-691.
Artieda, O., Herrero, J. and Drohan, P.J. (2006) Refinement of the differential water
loss method for gypsum determination in soils. Soil Science Society of America
Journal 70(6): 1932-1935.
Bachelet, G., deMontaudouin, X. and Dauvin, J.C. (1996) The quantitative
distribution of subtidal macrozoobenthic assemblages in arcachon bay in relation to
environmental factors: A multivariate analysis. Estuarine Coastal and Shelf Science
42(3): 371-391.
Blanchet, F.G., Legendre, P. and Borcard, D. (2008) Forward selection of
explanatory varibles. Ecology 89: 2623-2632.
Blott, S.J. and Pye, K. (2001) Gradistat: a grain size distribution and statistics
package for the analysis of unconsolidated sediments. Earth Surface Processes and
Landforms 26: 1237-1248.
Bone, Y. (1990) Sub-recent bryozoan-serpulid buildups in the Coorong lagoon, South
Australia. Australian Journal of Earth Sciences 37: 207-214.
Burrough, P.A. and McDonell, R.A. (1998) Principles of Geographical Information
Systems. Oxford University Press, Oxford.
Caeiro, S., Painbo, M., Goovaerts, P., Costa, H. and Sousa, S. (2003) Spatial
sampling design for sediment quality assessment in estuaries. Environmental
Modelling and Software 18: 853-859.
Carpenter, G. (1995) Birds of the Lower Murray region of South Australia. The
Murray Mouth Biological Resource Assessment Workshop, Adelaide.
The CLLAMM Dynamic Habitat
94
Caumette, P., Matheron, R., Raymond, N. and Relexans, J.-C. (1994) Microbial mats
in the hypersaline ponds of Mediterranean salterns (Salins-de-Giraud, France).
FEMS Microbiology Ecology 13: 273-286.
CLLAMM (2007). The future of the Coorong, Lower Lakes and Murray Mouth
(CLLAMM). Retrieved 15/01/2008, from http://www.csiro.au/science/ps260.html.
CLLAMM (2008) Response of the Coorong ecosystems to alternative Murray-Darling
flow scenarios. CLLAMMecology Research Cluster, Adelaide.
De Smith, M. J., Goodchild, M. F. and Longley, P. A. (2006) Geospatial Analysis: a
comprehensive guide to principles techniques and software tools. Matador, Leicester
Denis, V., Lejeune, J. and Robin, J. P. (2002) Spatio-temporal analysis of
commercial trawler data using General Additive models: patterns of Loliginid squid
abundance in the north-east Atlantic. Journal of Marine Sciences 59: 633-648.
Department for Environment and Heritage (2007) The Coorong: Natural Attractions.
Retrieved 15/01/2008, 2008, from
http://www.parks.sa.gov.au/coorong/natural/index.htm.
Department of Environment and Heritage (2007) The Coorong: Natural Attractions.
Retrieved 15/01/2008, 2008, from
http://www.parks.sa.gov.au/coorong/natural/index.htm.
Dray, S., Legendre, P. and Peres-Neto, P. R. (2006) Spatial modelling: a
comprehensive framework for principle coordinate analysis of neighbour matrices
(PCNM). Ecological Modelling 196: 483-493.
Earth Sciences Research Institute (2001) ArcGIS Geostatistical Analysis: Powerful
exploration and data interpolation solutions. ESRI, Redlands, California.
Earth Sciences Research Institute (2008) ArcGIS 9.2 Desktop, ESRI, Redlands,
California.
EconSearch (2004) Social, economic and environmental values and indicators for the
Lower Murray region, a report prepared for the Murray-Darling Basin Commission.
EconSearch and Australia Water Environments, Adelaide.
Edyvane, K. S. (1999) Conserving Marine Biodiversity in South Australia, Part 2Identification of areas of high conservation value in South Australia. South Australia
Research and Development Institute, Aquatic Sciences, Adelaide.
Fernandes, M. and Tanner, J. (2009) Hypersalinity and phosphorus availability : the
role of mineral precipitation in the Coorong lagoons of South Australia. Water for a
Healthy Country National Research Flagship and South Australian Research and
Development Institute (Aquatic Sciences), Adelaide.
Fernandes, M. and Krull, E. (2008) How does acid treatment to remove carbonates
affect the isotopic and elemental composition of soils and sediments? Environmental
Chemistry 5: 33-39.
Ford, P. W. (2007) Biogeochemistry of the Coorong. Review and identification of
future research requirements. Water for a Healthy Country National Research
Flagship. CSIRO Land & Water, Canberra.
The CLLAMM Dynamic Habitat
95
Geddes, M. C. and Francis, J. (2008) Trophic ecology pilot study in the River Murray
estuary at Pelican Point. Prepared for the Department for Environment and Heritage.
SARDI Aquatic Sciences Publication No. F2007/001193-1. SARDI Research Report
Series No. 274. South Australian Research and Development Institute (Aquatic
Sciences), Adelaide.
Gotway, C. A., Ferguson, R. B., Hergert, G. W. and Peterson, T. A. (1996)
Comparison of kriging and inverse distance methods for mapping soil parameters.
Soil Science Society of America Journal 60: 1237-1247.
Hastie, T. J. and Tibshirani, R. (1990) Generalized Additive Models. Chapman & Hall,
London.
Heijs, S. K., Azzoni, R., Giordani, G., Jonkers, H. M., Nizzoli, D., Viaroli, P. and van
Gemerden, H. (2000) Sulfide-induced release of phosphate from sediments of
coastal lagoons and the possible relation to the disappearance of Ruppia sp. Aquatic
Microbial Ecology 23(1): 85-95.
Isasks, E. H. and Srivastava, R. M. (1989) Applied geostatistics. Oxford University
Press, New York.
Jensen, O. P., Seppelt, R., Miller, T. J. and Bauer, L. J. (2005) Winter distribution of
blue crab Callinectes sapidus in Chesapeake Bay: application and cross-validation of
a two stage generalized additive model. Marine Ecology Progress Series 299: 239255.
Jones, N. S. (1950). Marine bottom communities. Biological Reviews 25: 283-313.
Kenny, A. J., Cato, I., Desprez, M., Fader, G., Schuttenhelm, R. T. E. and Side, J.
(2003) An overview of seabed-mapping technologies in the context of marine habitat
classification. ICES Journal of Marine Science 60(2): 411-418.
Lazar, B., Starinsky, A., Katz, A., Sass, E. and Benyaakov, S. (1983) The carbonate
system in hypersaline solutions - alkalinity and CaCO3 solubility of evaporated
seawater Limnology and Oceanography 28(5): 978-986.
Leecaster, M. (2003) Spatial analysis of grain-size in Santa Monica Bay. Marine
Environmental Research 56: 67-78.
Li, Q., McGowran, B., James, N. P., Bone, Y. and Cann, J. H. (1996) Mixed
Foraminiferal Biofacies on the Mesotrophic, Mid-Latitude Lacepede Shelf, South
Australia. Palaios 11(2): 176-191.
Liu, T. L., Juang, K. W. and Lee, D. Y. (2006) Interpolating soil properties using
kriging combined with categorical information of soil maps. Soil Science Society of
America Journal 70: 1200-1209.
Longley, P. A., Goodchild, M. F., Maguire, D. J. and Rhind, D. W. (2002) Geographic
Information Systems and Science. John Wiley & Sons, Ltd, Chichester
López-Moreno, J. I. and Nogués-Bravo, D. (2005). A generalized additive model for
the spatial distribution of snowpack in the Spanish Pyrenees. Hydrological Processes
19: 3167-3176.
McBreen, F., Wilson, J. G., Mackie, A. S. Y. and Aonghusa, C. N. (2008) Seabed
mapping in the southern Irish Sea: predicting benthic biological communities based
on sediment characteristics. Hydrobiologia 606: 93-103.
The CLLAMM Dynamic Habitat
96
Méar, Y., Poizot, E., Murat, A., Lesueur, P. and Thomas, M. (2006) Fine-grained
sediment spatial distribution on the basis of a geostatistical analysis: Example of the
eastern Bay of the Seine (France). Continental Shelf Research 26(19): 2335-2351.
Menzel, J. M., Ford, W. M., Edwards, J. W. and Ceperley, L. J. (2006) A habitat
model for the Virginia northern flying squirrel (Glaucomys sabrinus fuscus) in the
central Appalachian Mountains. Department of Agriculture, Forest Service,
Northeastern Research Station, Newtown Square, US.
Miles, M. (2006) Origin of bathymetric data for Lower Lakes, Goolwa and Coorong.
South Australian Department of Environment and Heritage, Adelaide.
Noell, C., Ye, Q., Short, D.A., Bucarter, L.B. and Wellman, N.R. (2009) Fish
assemblages of the Murray Mouth and Coorong region, South Australia, during an
extended drought period. CSIRO: Water for a Healthy Country National Research
Flagship and South Australian Research and Development Institute (Aquatic
Sciences), Adelaide.
Rizzo, W. M., Dailey, S. K., Lackey, G. J., Christian, R. R., Berry, B. E. and Wetzel,
R. L. (1996) A metabolism-based trophic index for comparing the ecological values of
shallow-water sediment habitats. Estuaries 19(2A): 247-256.
Roff, J. C., Taylor, M. E. and Laughren, J. (2003) Geophysical approaches to the
classification, delineation and monitoring of marine habitats and their communities.
Aquatic Conservation: Marine and Freshwater Ecosystems 13: 77-90.
Rogers, D.J. and Paton, D.C. (2009a) Spatiotemporal Variation in the Waterbird
Communities of the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship.
Rogers, D.J. and Paton, D.C. (2009b) Changes in the distribution & abundance of
Ruppia tuberosa in the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship.
Rolston, A.N. and Dittmann, S. (2009) CLLAMMecology Invertebrate Key Species
Project: The distribution and abundance of macrobenthic invertebrates in the Murray
Mouth and Coorong Lagoons. CSIRO: Water for a Healthy Country National
Research Flagship.
Rossi, F. (2003) Short-term response of deposit-feeders to an increase of the
nutritive value of the sediment through seasons in an intertidal mudflat (Western
Mediterranean, Italy). Journal of Experimental Marine Biology and Ecology 290(1):
1-17.
Ruttenberg, K. C. and Goñi, M. A. (1997) Phosphorus distribution, C:N:P ratios, and
d13COC in arctic, temperate, and tropical coastal sediments: tools for characterizing
bulk sedimentary organic matter. Marine Geology 139: 123-145.
Seaman, R. L. (2003) Coorong and Lower Lakes habitat-mapping program. South
Australian Department for Environment and Heritage, Adelaide.
Snelgrove, P. V. R. and Butman, C. A. (1994) Animal-sediment relationships
revisited: cause versus effect. Oceanography and Marine Biology: An Annual Review
32: 111-177.
The CLLAMM Dynamic Habitat
97
Storlazzi, C. D. and Field, M. E. (2000) Sediment distribution and transport along a
rocky, embayed coast: Monterey Peninsula and Carmel Bay, California. Marine
Geology 170: 289-316.
Verfaillie, E., Lancker, V. V. and Meirvenne, M. V. (2006). Multivariate geostatistics
for the predictive modelling of the surficial sand distribution in shelf seas. Continental
Shelf Research 26: 2454-2468.
Webster, I.T. (2005) An overview of the Hydrodynamics of the Coorong and Murray
Mouth. Water levels and salinity - key ecological drivers. CSIRO Land & Water,
Canberra.
Webster, I.T. (2006) Hydrodynamic modelling of the Coorong. Water for a Healthy
Country National Research Flagship. CSIRO Land & Water, Canberra
Wood, S.N. (2001) mgcv:GAMs and Generalized Ridge Regression for R. . 1(2), 2025.
Wood, S.N. (2003) Thin plate regresssion splines. Journal of the Royal Statistical
Society: Series B 65: 95-114.
Wood, S.N. and Augustin, N.H. (2002) GAMs with integrated model selection using
penalized regression splines and applications to environmental modelling. Ecological
Modelling 157(2-3): 157-177.
The CLLAMM Dynamic Habitat
98
5. Mudflat Geomorphology and Availability at Varying
Water Levels in the Coorong
Simon N. Benger1*, Jason E. Tanner2 and Sunil K. Sharma2
1
School of Geography, Population and Environmental Management, Flinders
University, GPO Box 2100, Adelaide, SA 5001
2
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
*corresponding author, Phone +61 (8) 8201 5994, Fax +61 (8) 8201 3521, E-mail:
[email protected]
5.1.
Executive Summary
The status of the Coorong wetlands as migratory bird habitat has been due primarily
to the opportunities they provide for large numbers of birds to feed in a highly
productive estuarine and lagoonal environment. A significant proportion of this
foraging occurs on the large tracts of mudflat found throughout the Coorong. The
productivity of the mudflats varies along the length of the Coorong, dependent
primarily on water quality (particularly salinity), nutrient inputs, sedimentary structure,
and the duration, frequency and extent of inundation. Resident macroinvertebrate
populations and aquatic vegetation such Ruppia species are both an indicator of
productivity in the mudflats and a food source for fish and birds.
This report presents high resolution topographic/bathymetric models for the 12
CLLAMMecology reference sites in the Coorong, as derived from survey data, along
with their geomorphological characteristics. Results confirm the importance of the
South Lagoon in terms of mudflat habitat, as it contains some 61% of available
mudflat, as measured in the reference sites. Mudflats throughout the Coorong are
generally likely to be geomorphically stable with mean mudflat slopes averaging
0.72%. Mudflat shapes are indicative of an accreting sedimentary environment. All
mudflats should be highly productive if the necessary physical, chemical and
biological conditions existed.
Across all 12 reference sites the 0 m to 0.5 mAHD elevation range is most significant
as it contains approximately 43% of all available mudflat area. The second most
important elevation class is -0.5 m to 0 mAHD, containing approximately 40% of total
available mudflat area. Hypsometric analysis shows that for the South Lagoon the
mudflat areas at elevations between 0m and 0.5 mAHD yield the greatest availability
of habitat and suggest that manipulations of water level should be kept within this
range. Ideally, the most important elevation range is 0.2 m to 0.4 mAHD, as
manipulations in this range accomplish wetting and drying of the maximum area of
mudflat, most of which is found in the South Lagoon. If mean water levels can be
maintained at close to optimal levels, then natural high-frequency, wind-driven
oscillations in water levels will inundate large areas of mudflat.
The CLLAMM Dynamic Habitat
99
5.2.
Introduction
The Coorong, part of the Ramsar-listed wetland of international significance located
at the terminus of the Murray-Darling River System, has been affected by muchreduced water levels in recent years. This is due to a combination of factors,
including reduced inputs from the Murray-Darling Baisn through the barrages at Lake
Alexandrina, reduced flows from the Upper South East Drainage (USED) scheme
and reduced connectivity between the South and North Lagoon due to sedimentary
deposition in the channel at Parnka Point (Figure 5.1).
The status of the Coorong wetlands as migratory bird habitat has been due largely to
the opportunities they provide for large numbers of birds to feed in a highlyproductive estuarine and lagoonal environment. A significant proportion of this
foraging occurs on the large tracts of mudflat found throughout the Coorong. These
mudflat areas are inundated periodically depending on water levels, which change at
a range of frequency cycles on an hourly, daily, seasonal, annual or decadal basis
depending on fluctuations in driving factors. The major drivers of water level are
regular tidal action (particularly in those areas closer to the Murray Mouth), extreme
tidal events such as king tides (which can push marine water far into the system),
freshwater inputs, evaporation (particularly in the summer months) and wind.
Figure 5.1. Locations of the reference sites in the Coorong (1-12).
The CLLAMM Dynamic Habitat
100
Mudflats can be of high ecological value because they support high levels of
productivity in both flora and fauna. They can be valuable as bird foraging areas and
as nursery and feeding areas for fish (Dyer et al. 2000). Many species of waders, in
particular, are dependent on coastal intertidal areas where they can feed on
macrobenthic invertebrates on exposed mudflats (Piersma et al. 1993). In the
Coorong, the productivity of the mudflats varies along its length, dependent primarily
on water quality (particularly salinity), nutrient inputs, sedimentary structure, and the
duration, frequency and extent of inundation. Resident macroinvertebrate
populations and aquatic vegetation such Ruppia species are both an indicator of
productivity in the mudflats and a source of prey for fish and birds. Rolston and
Dittmann (2009) present the results of an analysis of macroinvertebrate populations
throughout the Coorong.
As freshwater inputs to the Coorong system have reduced in recent years, this has
had a marked effect on the potential productivity of the mudflat areas. As water
levels have dropped, water quality has plummeted and salinity levels have risen.
Mudflat areas which were once subject to the more regular wetting and drying cycles
necessary to maintain their biological productivity (Boyes and Allen 2007) are
exposed for longer periods such that they dry-out permanently, or if they are
inundated, particularly in the South Lagoon, salinity levels are so high that
productivity levels would be similar to those found in salt lakes. Rolston and
Dittmann (2009) found that infaunal numbers declined rapidly when mudflats in the
Coorong were exposed as water levels drop.
In tidal mudflats, the mid-tidal region (halfway between the high and low water marks)
is usually the most important for infaunal communities in terms of species richness,
abundance and biomass, and productivity normally differs with respect to tidal
elevation and shore slope (Boyes and Allen 2007). In the Coorong, tidal mudflats are
found near the Murray Mouth, and other mudflat areas also experience substantial
changes in inundation on an hourly or daily basis due to wind-driven wave action and
wind-forced changes in water levels. Variations in water level and mudflat
morphology affect mudflat inundation and are critical to both the structural stability of
mudflats in the system and their productive potential.
In terms of the morphology of mudflats in the Coorong, topographical analysis allows
for prediction of habitat availability at varying water levels, and also provides an
indication of the stability of mudflat sediments, which in turn affects their productivity.
The cross-shore profiles of mudflats are dominated by tidal, wind and wave
processes and the contributions of these processes to total sediment transport (Kirby
2000). Erosional flats are dominated by wind-generated waves, which are
characterized by a concave-upwards profile. Tidally-dominated flats have a convex
upwards profile and are believed to accrete over time (Pritchard et al. 2001). The
productivity of mudflats can also have either a positive or negative influence on their
morphological stability. Biostabilisation may occur due to the presence and density
of microphytobenthos, algal mats, some species of worms and mussel mats, while
conversely biodestabilisation is the result of bioturbation caused by burrowing
bivalves, polychaetes and crustaceans (Uncles et al. 2003).
Restoring and maintaining mudflat productivity in the Coorong is a significant
conservation goal which will restore the now-flagging status of the Coorong as a
Wetland of International Significance, through providing viable foraging habitat for
birdlife. The Dynamic Habitat Program, under the CLLAMMecology Research
Cluster, seeks to quantify mudflat availability in the Coorong at varying water levels
along a range of reference sites located throughout the system. This has been
carried out through development of topographic models that can be analysed in a
GIS environment to predict spatial extent and availability of mudflats, when linked to
the varying water levels output from the existing hydrodynamic model for the
Coorong (Webster 2007). The fine-scale topographic models of the reference sites
form the basis of the dynamic habitat modelling presented in chapter 6 and can be
The CLLAMM Dynamic Habitat
101
run for various flow scenarios to predict potential bird habitat. The outputs of this
modelling will also be useful for incorporation into other models of bird foraging
behaviour developed by the Key Species Program within CLLAMMecology. In
addition, they will allow managers to make decisions on the operation of the various
management levers which affect water levels in the Coorong to achieve maximum
ecological benefit. In particular, where management agencies are operating in an
environment of restricted freshwater availability, there is a need to ensure that any
allocations are being put to maximum productive use to enhance the biological
viability of the system.
Compared with most other habitats mudflats have not been well researched and little
is known about the processes occurring in them (Kirby 2000; Dyer et al. 2000). This
study provides information on their physical potential as habitat and complements
work done in other parts of the CLLAMMecology Research Cluster to characterize
the biological attributes of mudflat areas throughout the Coorong. Compared to
sandy shores, mudflats have a greater complexity due to the behaviour of the
cohesive sediment and the roles of biological as well as physical processes (Kirby
2000). Characteristics of mudflat sediments throughout the Coorong are presented
in chapter 4.
5.3.
Methods
Twelve reference sites were utilised in this study, located along the length of the
Coorong: eight in the North Lagoon and four in the South Lagoon (Figure 5.1). For
the eight North Lagoon sites, detailed bathymetry was available from the South
Australian Water Corporation (SA Water), which was used as the basis of the
modelling carried out for this study. At the other sites located in the South Lagoon,
detailed surveying of mudflat morphology was carried out throughout 2007 and 2008
using a SOKIA SET5 30RK Surveying Instrument, complemented by Differential
Global Positioning System (DGPS) survey and at deeper water depths by kayakmounted sonar. At all sites, mudflat topography was interpolated from supplied and
field-collected data using radial basis functions in the Geostatistical Analyst extension
of ArcGIS (ver.9.3) (Environmental Systems Research Institute 2008). The source
datasets and processing and modelling methods used are described in the following
sections.
5.3.1 Source Datasets
North Lagoon Reference Sites
The South Australia Water Corporation (SA Water) collected depth data for the
Lower Lakes (Lakes Alexandrina and Albert) and the North Lagoon of the Coorong in
order to derive bathymetry for the region. The data were provided as point data at
varying resolutions in AutoCAD DXF format and the data collection procedures and
accuracy of the bathymetric dataset are described in chapter 3. The bathymetric data
for the North Lagoon had a point spacing of 25 m in the area between the Murray
Mouth and Mark Point and 50 m south to Parnka Point. The Goolwa Channel had a
point spacing of 10 m. The vertical accuracy of the original depth data collected by
the echo-sounder is ± 10 cm (Miles 2006). However, the accuracy of the boat
position, the effects of motion on the depth and position readings, and the effects of
tidal variation on the interpolated bathymetry are not known. The Ewe Island
reference site in the North Lagoon had only partial bathymetric data available due to
its proximity to the Murray Mouth, where no bathymetry was available. The
bathymetry for this site was modelled from LANDSAT5 spectral reflectance using the
methods described in chapter 3, and the finer resolution DEM for the site was
interpolated from the modelled data.
The CLLAMM Dynamic Habitat
102
South Lagoon Reference Sites
There was no suitable bathymetric data available for the South Lagoon, arising from
difficulties in accessing the Lagoon to carry out a bathymetric survey similar to the
North Lagoon. The shallow water depths of the South Lagoon, combined with
numerous outcropping limestone reef structures, make boat navigation difficult.
Consequently, the four reference sites: Parnka Point, Villa dei Yumpa, Jack Point
and Salt Creek, were manually surveyed using a combination of techniques. These
included a SOKIA SET5 30RK Surveying Instrument with an accuracy of ± 0.002 m,
a watercraft-mounted Garmin 400 Sonar with an accuracy ± 0.05 m and a Trimble
Pathfinder Pro XRS DGPS with a recorded horizontal site accuracy of ± 0.35m. For
those areas between the high-water shoreline, with the upper boundary usually
delineated by the presence of fringing vegetation, and shallow-water depths up to 0.5
m, a standard survey using tripod and prism arrangement was carried out recording
XYZ positional readings, which were later processed using ProLink 1.15 (Point
Software Inc. 2001). Trimble Pathfinder 4.00 (Trimble Navigation Ltd. 2007) software
was used to convert the survey data into ArcGIS (ver.9.3) (Environmental Systems
Research institute, 2008) shapefile point datasets. XY reference positioning of
survey station location and backsite locations was accomplished using Differential
GPS to an accuracy of ± 0.35m. Corrections to AHD were made using the closest
available reference data north of Parka Point.
Survey data from the Surveying Instrument were complimented by waterline data
collected during 2007 and 2008 at different times of the year, and hence varying
water levels, using the Trimble Pathfinder Pro XRS DGPS, which allowed the
measurements of mudflat slope and shape to be extrapolated over the 1km-wide
analysis area used for each site. For water depths greater than 0.5 m, a Garmin 400
Sonar mounted on a kayak, and calibrated using a depth measuring pole, was used
to collect depth measurements on transects across the full width of the Lagoon.
Depths were recorded as waypoint attributes in a GPS with a corrected horizontal
accuracy of <2m. At shallow depths of less than 0.5 m across the Lagoon and on the
far western shore, depths were measured manually using a measuring pole with an
accuracy of ± 0.005m. All water-based depth measurements were carried out early
in the morning on still days to avoid any influence of wind forced tilt of the water
surface and wave effects. GPS waypoint files were converted to XYZ positions in
metres and imported into ArcGIS (ver.9.3) (Environmental Systems Research
institute, 2008) as point shapefiles. This data was then corrected to AHD using tie-in
points with the ground survey data. The combined survey data yielded between 650900 XYZ points at each South Lagoon reference site, which was then interpolated to
produce DEMs for each site.
5.3.2 Interpolation Methods
For each reference site the topographic/bathymetric models (DEMs) were generated
using radial basis functions in the Geostatitistical Analyst extension in ArcGIS
(ver.9.3) (Environmental Systems Research Institute 2008). In all models thin-plate
splines were used, utilising a minimum of 10 neighbours and a maximum of 15 in a
standard search area. Four search sectors oriented at 45 degrees were used as this
captured the direction of major trends in the input data, due to the general NW-SE
orientation of the Lagoon. Thin-plate splines have been widely used in surface
modelling (Glenn et al. 2006) because they produce an excellent fit to the input data
being interpolated and can model scattered data points effectively without the need
for experimental data points (Boyd et al. 1999). While a number of other methods
are regularly used for surface interpolation from scattered point data, such as IDW,
B-splines and geostatistical methods such as kriging, they will in most cases produce
similar results dependent on the spatial distribution of the original data (Isaaks and
Srivastava 1989; Gooverts 1997). In this analysis, the thin plate spline models
produced the best fit and lowest Root Mean Square (RMS) error in the resultant
The CLLAMM Dynamic Habitat
103
surfaces. All models used an inverse multiquadratic kernel function for the
interpolation (Environmental Systems Research Institute 2008). The resultant
models were at 1 m horizontal resolution and 0.001 m vertical resolution, to allow for
accurate estimation of changes in habitat availability at different water levels. The 1
m horizontal resolution allows for mudflat areas to be calculated to the nearest metre,
although all results here are presented in hectares for ease of interpretation.
5.3.3 Hypsometric Analyses
Hypsometric characterisation of shoreline shape has been widely used in
geomorphological analysis of coasts (Kirby 2000). The hypsometric curve is
calculated as the cumulative area of shoreline available at ascending elevations. It
provides an indication of the nature of the surface and whether it is subject to
sedimentary accretion or erosion (Carter 1988). For the current study, the shape of
the hypsometric curve at each reference site also provides an indication of the rate of
change in habitat availability as water level in the Coorong increases or decreases.
In this analysis, hypsometry was calculated through a GIS routine implemented in
ArcGIS (ver.9.3) (Environmental Systems Research institute 2008) adapted from
scripts available in GT Spatial Tools and using the methods developed for tidal
mudflat modelling by Xander Bakker, Grontmij Nederland bv (Bakker pers. comm.).
Essentially the process involved extracting all DEM cells above a certain height (in
this case above -0.5 AHD, because this was approximately 0.5m below the minimum
observed water level in the South Lagoon), and converting respective AHD
increments to polygons. Polygonal areas can then be calculated for each grid cell
height and cumulative areas calculated. Plotted results then yield the hypsometric
curve for each site.
5.3.4 Volumetric Analyses
Calculations on the volumes of water required to manipulate water levels in the South
Lagoon were carried out using the 3D Analyst extension of ArcGIS (ver.9.3)
(Environmental Systems Research institute 2008). The bathymetric model used in
this process was generated from satellite and transect data and is described in detail
in chapter 3. All estimates are based on volumetric calculations relative to an input
reference plane, and as such assume an artificial horizontal water level. In reality,
any linearly flowing system over a large area will contain a natural gradient in the
water surface and this is evident in the water level output from the hydrodynamic
model (Webster 2007) for the Coorong. However, the surface gradient is highly
variable depending on flow conditions and time of the year. Water surface gradients
in the Coorong are also regularly affected by wind set-up conditions.
5.3.5 Determination of Wind Effects on Water Level
Wind is well known as a major driver of high frequency oscillations in water levels
and circulation over large water bodies (Shilo et al. 2007; Schwab and Beletsky
2003) and is considered a major ecological driver for the Coorong system
(Lamontagne et al. 2004). These effects have been quantified in the Lower Lakes
(Noye 1973) and in the North Lagoon of the Coorong (Noye and Walsh 1976) but not
in the South Lagoon where the majority of mudflats are present. Webster (2007)
stated that wind driven effects on water level in the Coorong are in the order of ± 5
cm. In the current study, wind effects on water level were measured in the South
Lagoon at Villa dei Yumpa where large mudflat areas are present. This area forms
the northernmost end of the main expanse of water which makes up the South
Lagoon. Measurements were taken over a series of individual time periods capturing
changes in water levels between zero wind speed and high wind speed events,
events which occur regularly in the Coorong. In particular, the sampling sought to
The CLLAMM Dynamic Habitat
104
capture the effects of the most common higher intensity southerly winds (Noye and
Walsh 1976) which blow parallel to the main axis of the South Lagoon.
An integrated water level, wind speed and wind direction data logger was not
available for the current study and therefore simple measurement methods were
applied. This involved the use of three graduated poles from which water levels could
be read, spaced at 100m intervals and varying depths in the Lagoon. Locations were
recorded using DGPS and depth correlated to AHD using the fine-scale bathymetric
model for Villa Dei Yumpa (appendix 5.2). Wind speed was measured using a digital
anenometer and wind direction was measured using handheld GPS. Measurements
were taken using three replicates at 30 minute intervals at each location. While this
provides only a partial insight into the effects of wind on water levels in the South
Lagoon, it does indicate the likely maximum changes possible and illustrates the
importance of these effects for wetting and drying of the mudflats over short
timescales.
5.3.6 Methodological Limitations
The North Lagoon bathymetry is an interpolated product, and because it was based
on echo-sounder measurements made in a boat which did not venture into depths of
less than 1 m, it is not ideal for the purposes of mudflat modelling. The bathymetry of
those areas between 1m depth and the shoreline is estimated linearly from the
waterline. However, the integrity of the final DEMs for the North Lagoon reference
sites was checked through a number of validation surveys using the Surveying
Instrument. As the South Lagoon sites were surveyed using the same instrument,
the models generated for these four reference sites provide a highly accurate
representation of mudflat shape.
The physical location and morphology of the mudflats of the Coorong is heavily
influenced by the underlying and exposed eroded Pleistocene limestone bedrock and
abandoned cliff structures (Bourman et al. 2000) which can be found throughout the
Lagoons. As such, some areas of the mudflats are not purely sedimentary structures
and therefore their geomorphic controls are not as strongly influenced by erosional
and depositional processes and sedimentary structure. The mudflat models
developed as part of the current study do not record the location of outcropping
limestone or tubeworm atolls within the model extents, which would obviously have
different properties and different habitat potential to sediments. However, these
areas are very limited in size within the analysis areas and are likely to have only a
small influence on mudflat area calculations.
5.3.7 Linking to the Hydrodynamic Model
A one-dimensional hydrodynamic model for the Coorong was developed by CSIRO
Land and Water (Webster 2007). This model is able to predict changes in water level
and salinity along an approximately 160 km linear transects running from the south of
the South Lagoon of the Coorong to just south of the Murray Mouth in the North
Lagoon (Figure 5.1). It is based on water level measurements at various locations
along the Coorong and tidal and barrage flow data, and wind data from the period
before the current disconnection between the Lower Lakes and the Coorong. Output
consists of modelled water levels at 1 km intervals at 1 hour time steps and the
model can be run for a range of possible flow scenarios.
A dynamic habitat model linking output from the hydrodynamic model to the finescale topographic/bathymetric models at each reference site was written in ArcGIS
(ver.9.3) (Environmental Systems Research institute 2008) ModelBuilder (chapter 6).
This allows modelled scenarios from the hydrodynamic model to be linked to the
mudflat models presented here and predict habitat availability for past, current or
predicted time periods. A full description of the dynamic habitat model and its
outputs is presented in chapter 6, and it is a tool that can be used by managers to
The CLLAMM Dynamic Habitat
105
understand how mudflat habitat availability has changed over time, and predict how it
will change in the future for any given flow scenario.
5.4.
Results
5.4.1 Surface Modelling
High resolution cell-based topographic models of shorelines, mudflats and
bathymetry were generated for each of the 12 reference sites using thin plate spline
models. Each of these covers a 1 km wide section of the Coorong and is positioned
perpendicular to the general linear axis of the Coorong, which generally follows the
ocean shoreline. Each model has a cell resolution of 1 m. The goodness of fit of the
spline surfaces was assessed through cross validation between the resultant
surfaces and the points from which they were generated. The validation plots
showing the relationship between the predicted and measured surface at each site
are shown in Appendix 5.1a and the cross validation results are presented in
Appendix 5.1b.
The quality of the resultant surfaces varies according to the nature of the input data,
particularly its spatial distribution, and the surface fitting characteristics of the thin
plate spline models. Larger numbers of input points were used in the North Lagoon
reference sites due to the availability of the interpolated bathymetric dataset for the
area, and at varying resolutions. The large numbers of points and their regular
spatial distribution also results in considerably reduced mean and RMS errors for the
North Lagoon, and therefore better surface fit, relative to the South Lagoon sites.
The twelve fine-scale topographic models depicting mudflats, shorelines and
bathymetry at each reference site are presented in Appendix 5.2 and an example is
shown in Figure 5.2. Each captures a very different morphology depending on the
location of islands within the site, shoreline shape and the physical processes acting
at each site. In geomorphic terms, the areas represented by the reference sites can
be subjected to various groupings dependent on where they occur in the system.
The most obvious differences occur between North and South Lagoon sites, with the
two Lagoons acting to some extent as separate systems although connected through
a narrow channel at Parnka Point. The North Lagoon is dominated by marine inputs
through the Murray Mouth and freshwater flows over the barrages. The South
Lagoon is affected by surface flows from the USED and groundwater flows, as well
as flow from the North Lagoon, and evaporation has a much greater influence on
water levels. The Mundoo Channel reference site is somewhat anomalous in the
system, occurring on a connecting channel from Lake Alexandrina, which is shaped
by channel flow and marine tidal variation.
In the North Lagoon, the reference sites from Pelican Point north are all dominated
by deeply- incised channels to depths of between -2.7 and -3.6 mAHD. The three
other North Lagoon sites, Mark Point, Long Point and Noonameena, do not contain
incised channels and have shallower maximum depths. This is most likely due to
increased sedimentation around the northern reaches of the North Lagoon,
influenced primarily by marine sediments from the areas around Murray Mouth.
In the South Lagoon, the bathymetry at Parnka Point, Villa dei Yumpa and Jack Point
is dominated by current-incised channels through which most south-north water flow
occurs in the system. Sedimentation is most likely greatest at the northern end of the
South Lagoon. Salt Creek is at the southern end of the system and is, on average,
the deepest part of the Lagoon, containing the largest water volume, but it is unlikely
to be subjected to significant flow effects. The dominant process at depth is more
likely wind induced gyres acting in a confined limestone-bounded basin. Areas that
function as salt lakes when disconnected from the Coorong at lower water levels,
such as those around Parnka Point, would also be subject to aeolian deflation when
The CLLAMM Dynamic Habitat
106
dry, in which strong winds scour sediments from the exposed basins (Cooke et al.
1993).
Figure 5.2. High Resolution Surface Model for Jack Point Reference Site – an example
5.4.2 Mudflat Parameters
General mudflat area parameters for each reference site at 0.5 m vertical intervals
within the range -0.5 m to 1.5 mAHD are provided in Table 5.2 below. Height
intervals of 0.5 m have been utilised for descriptive purposes only. The South
Lagoon dominates total mudflat availability in the Coorong, containing approximately
61% of total area, compared to approximately 39% in the North Lagoon, as
measured in the reference sites. Across all 12 reference sites, the 0 m to 0.5 mAHD
elevation range is most significant as it contains approximately 43% of all available
mudflat area. The second-most important elevation class is -0.5 m to 0 mAHD,
containing approximately 40% of total available mudflat area. Only approximately
17% of mudflat is available at elevations above 0.5 mAHD for the Coorong as a
whole. Villa dei Yumpa, Parnka Point and Jack Point are the most significant sites
across the Coorong in terms of mudflat area containing 22%, 18% and 15% of the
total respectively.
The general trend in the North Lagoon for the elevation classes in the -0.5 m to 0.5
mAHD range, is for increasing mudflat area from the south to the north up until
Pelican Point and then decreasing mudflat area in the remaining northern sites. The
limited and sometimes non-existent mudflat area available at elevations above 0.5
mAHD in the North Lagoon is due to the more stable mean water levels closer to the
Murray Mouth (less subject to seasonal and annual variation) and lower mean water
levels compared to the South Lagoon, resulting in less sedimentary deposition above
0.5 mAHD. The elevation class containing the majority of mudflat area for the North
Lagoon in terms of mudflat area is the -0.5 m to 0 m elevation range, which is
generally the area below the high water mark in tidally-influenced sites.
The CLLAMM Dynamic Habitat
107
The general trend for the South Lagoon is for increased mudflat availability moving
from south to north. The most important elevation class in terms of habitat
availability is likely to be the 0 m to 0.5 mAHD, as this is the water level range most
commonly experienced in recent years for the Coorong. Chapter 6 presents
historical water level data predicted from the hydodynamic model which is somewhat
higher under normal flow conditions. The 0 m to 0.5 mAHD range shows a trend of
increasing mudflat availability from south to north. It is also an important elevation
range for total mudflat area as it contains 49% of total mudflat across the four
reference sites of the South Lagoon. The relatively large surface areas in the range 0.5 m to 0.5 mAHD for Villa dei Yumpa reflect the influence of islands and a
peninsula dissecting the Lagoon (see Appendix 5.2) and the consequent large
mudflat area that has developed in low current conditions. The large areas of
mudflat at higher elevations at Parnka Point are primarily due to the presence of
large salt pan areas to the east which become connected to the Lagoon at higher
water levels. South-north water flow, strong winds from the south (Noye and Walsh,
1976), particularly during winter, and the constriction between the North and South
Lagoons at Parnka Point, would also be responsible for increased sedimentation in
the northern areas of the South Lagoon.
Of the total mudflat area available at the reference sites between -0.5 m and 1.5
mAHD in the South Lagoon, Salt Creek contains 11%, Jack Point 23%, Villa dei
Yumpa 36% and Parnka Point 28%. For the North Lagoon reference sites, the
mudflat area is fairly consistent across most sites, with Pelican Point containing the
largest mudflat area. This is due to the presence of a significant mid-channel sand
island at Pelican Point which is normally exposed only at low tide or when water
levels are low.
Table 5.2. General Mudflat Area Parameters at each Reference site (S-N).
Mudflat Area (ha)
Reference Site
% Total
Mudflat
-0.5 to 0
mAHD
0.0 to 0.5
mAHD
0.5 to 1.0
mAHD
1.0 to 1.5
mAHD
Total -0.5 to
0.5 mAHD
Salt Creek
20.01
27.47
25.97
9.44
82.89
6.86
Jack Point
71.05
94.53
15.44
0.41
181.43
15.01
148.97
100.43
11.42
1.27
262.09
21.68
Parnka Point
13.76
136.73
34.28
22.23
207.01
17.13
Noonameena
32.47
17.70
20.11
0.12
70.40
5.82
Long Point
18.87
23.19
14.51
0.00
56.57
4.68
Mark Point
23.98
17.51
10.26
0.00
51.75
4.28
Pelican Point
39.57
46.38
12.15
0.00
98.09
8.12
Ewe Island
12.55
8.61
8.82
0.18
30.17
2.50
Barker Knoll
31.08
21.06
17.48
1.61
71.23
5.89
Mundoo Channel
56.54
12.97
0.00
0.00
69.51
5.75
Goolwa Channel
12.89
6.66
7.55
0.52
27.60
2.28
481.75
513.23
177.98
35.79
1208.75
100
39.86
42.46
14.72
2.96
100
Villa de Yumpa
Total
% Total Mudflat
Mudflat slopes at each reference site in the same four elevation classes are provided
in Figure 5.3. The 0.5 mAHD elevation increments used are for descriptive purposes
The CLLAMM Dynamic Habitat
108
only. The slope of mudflats provides an indication of their stability, with low slope
mudflats more stable and more biologically active (Dyer et al., 2000) as larger areas
are subject to wetting and drying action as water levels increase and decrease due to
tides, wind and wave action. Steeper slopes in mudflats are indicative of erosional
surfaces and are likely to be less stable and less biologically active (Dyer et al.,
2000). Across the 12 reference sites the highest mean mudflat slopes are present at
Goolwa in the 0 m to 0.5 mAHD range (1.64%) and Villa dei Yumpa at the 1.0 m to
1.5 mAHD range (1.43%). Mean mudflat slopes are generally quite low throughout
the Coorong and average 0.72% across all reference sites. Generally, the North
Lagoon reference sites had slightly lower mean mudflat slopes (0.67%) compared to
the South Lagoon (0.82%).
2.50
Salt Creek
Jack Point
Slope in degrees
2.00
Villa dei Yumpa
Parnka Point
1.50
Noonameena
Long Point
1.00
Mark Point
Pelican Point
0.50
Ewe Island
Barker
0.00
‐0.5 to 0.0m
0.0 to 0.5m
0.5 to 1.0m
Elevation class (AHD)
1.0 to 1.5m
Mundoo
Goolwa
Figure 5.3. Mudflat Slopes at each Reference Site.
At the four reference sites in the South Lagoon mean mudflat slopes are very low at
less than 0.82%, indicative of stable mudflats in a relatively low energy environment.
Across all four sites the lowest slopes are found generally in the 0 m to 0.5 mAHD
range. Highest slopes are found at elevations above 0.5 mAHD, except for Salt
Creek where the highest slopes occur at depths below -0.5 mAHD. The highest
maximum slopes are found at Parnka Point, which is indicative of the
geomorphically-constrained deeply-incised channel running through this section
connecting the North and South Lagoons of the Coorong.
5.4.3 Hypsometric Curves for Mudflats
The hypsometric curves for the mudflats at each reference site provide an indication
of the stability of the mudflats in terms of accretion or erosion and also provide an
indication of where the greatest benefits will be achieved in maximizing habitat
availability by increasing water level. The shape of the hysometric curves may,
however, be influenced by the quality of the input data, which is an important
consideration for the North Lagoon reference sites, in particular at higher elevations,
due to the interpolation process used (Milles 2006). The hypsometric curve for Jack
Point is shown in Figure 5.4 and the hypsometric curves for all reference sites are
shown in Appendix 5.4. Hypsometry of the North Lagoon mudflats is markedly more
linear than that of the South Lagoon. Only Mundoo Channel in the North Lagoon
The CLLAMM Dynamic Habitat
109
exhibits a typical marine shoreline hypsometry, with steep slopes at lower elevations
followed by lower slopes at higher elevations (Carter 1988). Pelican Point, in
particular, exhibits rapid gains in cumulative mudflat area between 0.25 m and 0.3
mAHD, which is probably the exposure threshold of the large mid-channel sand
island.
The hypsometric curves at all four reference sites in the South Lagoon show a rapid
increase in cumulative mudflat area above 0 mAHD. In general terms, this increase
is greatest in the 0 m and 0.5 mAHD range. Parnka Point and Jack Point display
somewhat typical hypsometric curves for shorelines, with a characteristic sinusoidal
shape (Carter 1988), with areas of concavity in the curve at elevations below 0
mAHD and areas of convexity above 0 mAHD. Salt Creek shows modest, nearlinear increases in mudflat availability above -0.5 mAHD, while Villa dei Yumpa
shows a rapid linear increase between -0.5 m and 0.5 mAHD and minimal gains
above 0.5 mAHD. Relatively small accumulations in mudflat area occur in the -0.5 to
0 mAHD range at all sites. With Villa dei Yumpa being the most important site in the
Coorong in terms of total available mudflat area, the hyposmetric curve shows that 1
m of variation in water level between -0.5 m and 0.5 mAHD achieves a large and
significant 250 ha gain in cumulative mudflat area.
Figure 5.4. Hypsometric curve for Jack Point Reference Site – an example.
5.4.4 Wind Effects on Water Level in the South Lagoon
The results of the measurements of wind speed in relation to water level carried out
at Villa dei Yumpa in the South Lagoon provide some indication of the likely influence
of wind in wetting and drying mudflats over short timescales, and these are
presented in Figure 5.5. While only a portion of the data collected was directly
comparable due to variations in wind direction, they do show that changes in water
level can be in the order of 0.8m between 0 and 16.4 m/s. This is considerably
greater than the 0.1m suggested in Webster (2007).
The CLLAMM Dynamic Habitat
110
Wind Speed (m/s)
0
2
4
6
8
10
12
14
16
18
0.8
0.7
Water Level (m AHD)
0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
Figure 5.5. Wind effects on water level at Villa dei Yumpa (wind direction 328-342 deg.).
5.5.
Discussion
The hydrodynamic model for the Coorong (Webster 2007) shows that for typical flow
scenarios there is usually some gradient present in the water surface along the
length of the Lagoon, dependent on flows to and from the system. These gradients
would need to be taken into account for determination of changes in habitat
availability at different water levels. However, some generalisations can be made
about the area of mudflats in relation to AHD. Only some 17% of total available
mudflat exists at elevations above 0.5 mAHD. A rise in water level of 0.5 m above 0
mAHD results in a decrease of nearly 43% in mudflat availability. In terms of mudflat
area alone, water level manipulations within this elevation range will return the
greatest habitat benefit, but this does not take into account the effects of poorer
water quality on habitat at lower water levels. Higher salinities, in particular, have a
marked impact on productivity (Rolston and Dittmann 2009).
Results show clearly that the South Lagoon is considerably more important than the
North Lagoon in terms of mudflat availability, as it contains the majority (61%) of all
available mudflat at the reference sites. This occurs even though only four reference
sites in the South Lagoon were included in the analysis compared to eight in the
North Lagoon. Thus, if these sites are typical of the Coorong, then it is likely that the
South Lagoon contains substantially more than two thirds of the available mudflat in
the Coorong, although a full fine-scale bathymetry for the full Coorong would be
required to confirm this. The management implications are that water level
manipulations in the South Lagoon are far more important in terms of increasing
available mudflat habitat. Although they are currently less productive biologically due
to lower water levels and high salinity levels (Rolston and Dittmann 2009), increased
water availability in the South Lagoon could provide considerable ecological benefits,
provided water quality improved sufficiently for macroinvertebrates and macrophytes
to colonise the mudflats that would become available.
The mudflats of the Coorong are exposed to a unique range of physical and
biological processes depending on where they occur in the system. While the current
study quantifies the availability of mudflats at various water levels and can be linked
to the system-wide hydrological model, it does not account for the varying levels of
productivity at the various reference sites. So, while habitat may be available at a
The CLLAMM Dynamic Habitat
111
given water level, it may not necessarily be biologically productive or suitable as
high-quality foraging habitat. This would depend on a complex range of interactions
at each reference site.
The exact positioning of the location of the 1km-wide analysis polygon used for each
reference site will influence available habitat at any one site due to variations in the
shoreline and the area of mudflat captured within the polygon at each reference site.
However, results do provide an indication of the mudflat morphology at various points
along the Coorong and also permit comparison between the different systems of the
North and South Lagoons.
The lagoon-scale bathymetric modelling described in chapter 3 makes it possible to
calculate the quantity of water required to achieve significant ecological benefits in
the South Lagoon. However, this would need to be delivered in such a way as to
flush existing low quality water from the system. Previous discussions on making
environmental flows available to the Coorong focus on delivery over the Barrages
(MDBC 2003; 2004), but as this delivers water to the South Lagoon through a
constricted channel at Parnka Point, little flushing can occur. The volume of the
South Lagoon at 0 mAHD is estimated at 98 GL from the bathymetric model of
chapter 3. To achieve a 0.5 m rise in water level above 0 mAHD in the South
Lagoon it would be necessary to provide an estimated 44 GL to the system,
assuming that all input water was retained in the South Lagoon and did not flow into
the North Lagoon. However, this would only dilute the existing hypersaline water,
dependent on the salinity of input flows, and thus substantially greater flows would be
required to both increase water levels and flush out existing low-quality water.
Transferral of hypersaline water through the North Lagoon would also be likely to
have significant ecological impacts.
Low summertime water levels in the South Lagoon in recent years, down to
approximately 0 mAHD, result in much larger areas of mudflat being exposed
compared to typical water levels above 0.5 mAHD. However, this occurs in
combination with hypersaline conditions and drying of exposed sediments and
consequently viable, biologically-active mudflats are reduced in availability. Even
with good water quality, available mudflat habitat does not necessarily equate to
high-quality mudflat. Rather, quality will also depend on how high above water the
mudflat is or what depth of water covers it, and how long it has been exposed. The
best quality mudflats, in terms of habitat for macroinvertebrates and thus foraging
grounds for birds, are likely to be those that are covered by a shallow layer of water,
and/or that have only recently become exposed (Dyer 1998). Mudflats that have
been well above water level for extended periods (weeks to months), are likely to
have few macroinvertebrates (Rolston and Dittmann 2009), and will thus also be very
poor foraging grounds for birds. Similarly, while mudflats > 10-20 cm deep may have
a high abundance of macroinvertebrates (depending on species) and also provide
suitable habitat for Ruppia and various fish species, they will not be available to
wading birds (Rogers and Paton 2009).
Dyer et al. (2000) found that most significant differences in flora and fauna in the
upper zones of mudflats were due to variations in sediment grain size and that the
main driving variables are tidal range, exposure to waves and mudflat slope. Mudflat
slopes across the four reference sites in the South Lagoon show an increase at
elevations above 0.5 mAHD (Appendix 3), indicating that they are likely to be less
productive (Dyer 1998) than mudflat areas below 0.5 mAHD. However, low overall
slopes of less than 0.72% across all reference sites indicate that all mudflat areas are
likely to be highly stable and relatively productive (Dyer et al. 2000), if the other
appropriate physical, chemical and biological conditions are present.
Geomorphically-stable mudflats with limited sedimentary processes occurring are
characteristic of sheltered estuarine or lagoonal sites (Boyes and Allen 2007). Small
geomorphic elements within the surface shape of mudflats are also likely to affect
biological productivity (Anibal et al. 2006). In particular, alternating meso-topographic
The CLLAMM Dynamic Habitat
112
concave and convex features may contain varying macroinvertebrate populations
despite no differences in physico-chemical characteristics of the sediments (Anibal et
al. 2006). Such meso-topographic features are present in mudflats throughout the
Coorong, but are not captured in the resolution of the topographic/bathymetric
models presented here.
The generally-linear hypsometry of the North Lagoon contrasts with that of the South
Lagoon. Results indicate that between -0.5 m and 1.0 mAHD mainly linear gains in
cumulative mudflat area occur. This suggests that there is no critical elevation
threshold at which water levels should be maintained in the North Lagoon, in terms of
maximising mudflat availability. In the current conditions during disconnection from
the Lower Lakes, marine inputs maintain mean water levels at little higher than 0
mAHD.
The hypsometry of the South Lagoon reference sites, which contain the majority of
total mudflat area, suggests that the greatest changes in cumulative mudflat area
occur in the range 0.2 m to 0.4 mAHD. This is important because it suggests that
artificially varying the water level within this range will achieve the maximum possible
benefits in terms of wetting and drying the maximum area of mudflat. Based on the
South Lagoon volumetric model presented in chapter 3, the manipulations in this
range would involve injecting some 18 GL of water into the system, ignoring any
water loss into the North Lagoon.
An important aspect of the physical characteristics of mudflats which determine their
geomorphic form and behaviour is the grain size distribution. Grain size plays an
important role in mudflat dynamics (Uncles et al. 2003), both in the way in the
mudflats will respond to erosional and depositional processes and in terms of
biological productivity. Grain size also affects drainage and water temperature which
act as drivers of mudflat dynamics (Le Hir et al. 2000) and influences the distribution
of fauna and flora (Boyes and Allen 2007; Dyer et al. 2000,). The morphological
characteristics of the Coorong mudflats should therefore be considered in the context
of their grain size characteristics. Grain size structure and spatial distribution
determined from sampling conducted as part of the Dynamic Habitat project is
presented in chapter 4.
The evolution of the mudflat often involves a winnowing process by which coarser
grained material, which is mainly the skeletal remains of intertidal macrofauna, is
driven upshore by waves and often lies above the typical high water mark (Kirby,
2000). This is consistent with conditions observed in the Coorong (chapter 3). The
winnowed mud fraction is reworked into deposition areas, resulting in an
accumulation of finer sediments at depth (chapter 3). Deposition on flats leads to
thin, discontinuous, normally-consolidated veneers, which provide host sediments for
a restricted invertebrate population (Kirby 2000). This is compounded in the Coorong
by high levels of salinity, particularly in the South Lagoon, which have reduced the
quantity and diversity of macroinvertebrates in these habitats (Rolston and Dittmann
2009). Highly saline environments therefore provide only limited foraging for waders,
with Artemia spp. (brine shrimp) usually being the only food source available (Masero
et al. 1999).
Wind-driven changes in water levels in the Coorong can have a significant effect on
the area of mudflat inundated over short time periods (Rogers and Paton 2009).
Water level variations due to wind are commonly observed in large water bodies
(Thompson 1983). Measurements made during extreme wind events at Villa dei
Yumpa in the South Lagoon, showed a change in water level of approximately 0.8m
in less than 24 hours between windspeeds of 0 and 16.4 m/s, with a wind direction
between 228 and 342 degrees (Southerly winds) which achieve a near maximum
wind set up along the South Lagoon. Such changes can occur in a matter of hours,
as afternoon winds develop and increase in intensity, or as frontal systems move
through. Wind is therefore likely to be the primary high frequency factor affecting
inundation of mudflats, and in turn mudflat availability and possibly productivity. In
The CLLAMM Dynamic Habitat
113
addition, in the North Lagoon and estuarine area, there is a tidal movement that
propagates through the Murray Mouth. Daily tidal signals are obvious as far south as
Mark Point, although the tidal cycle is reduced to ~ 10 cm compared to ~ 80 cm at
Barker Knoll, and only the spring/neap tidal cycle penetrates to Long Point. In
combination, tidal activity and wind forcing can change inshore water levels at Mark
Point by up to 60 cm over periods of days to weeks (Webster 2007).
5.6.
Summary, Conclusions & Management Implications
Mudflats form critical habitat for many species in the Coorong, in particular wading
birds. Maximising the productivity of mudflats, through maintaining water levels and
water quality is essential for ensuring the Ramsar status of the Coorong by providing
coastal foraging habitat required by migratory birds. Healthy mudflats support large
and diverse macroinvertebrate populations, are important as fish breeding areas and
form substrate for submerged aquatic vegetation such as Ruppia species. Restoring
the productivity of the Coorong mudflats is an important conservation goal for the
region.
The characterisation of mudflat morphology undertaken in this study allows for
detailed estimates of mudflat habitat availability at different water levels. This study
confirms the importance of the South Lagoon in terms of mudflat habitat, as it
contains some 61% of available mudflat as measured in the reference sites. The
results show conclusively that for the South Lagoon the mudflat areas at elevations
between 0 m and 0.5 mAHD yield the greatest availability of habitat and suggest that
manipulations of water level should be kept within this range. Ideally, the most
important elevation range is 0.2 m to 0.4 mAHD, because manipulations in this range
accomplish wetting and drying of the maximum area of mudflat, most of which is
found in the South Lagoon, as measured in the reference sites. If water levels can
be maintained at close to optimal levels, then natural, high-frequency wind-driven
oscillations in water levels will inundate large areas of mudflat. Mudflats throughout
the Coorong are generally likely to be geomorphically stable with mean mudflat
slopes averaging 0.72%. Mudflat morphology is thus indicative of an accreting
sedimentary environment (Pritchard et al. 2002) and all mudflats should be highly
productive if the necessary physical, chemical and biological conditions existed.
Currently, however, mudflats throughout the Coorong are in poor condition in terms
of biological productivity, apart from some areas near the Murray Mouth (Rolston and
Dittmann 2009) and this is indicative of the degraded state of the system (Phillips and
Muller 2006).
By linking the GIS models presented here to the hydrodynamic model of Webster
(2007), as detailed in section 6.3.2, it is possible to predict how mudflat availability
will change along the length of the Coorong under any given flow scenario. This is
modelled for a number of scenarios in chapter 6. This will allow managers to
compare different potential scenarios for their influence on mudflat availability, and
take this important habitat into account when making decisions on the quantity,
timing and duration of flows to be released into the system. Water levels have direct
implications for biological functioning and may also affect productivity at higher
trophic levels (Boyes and Allen 2007). While water levels above 0.5 mAHD provide a
diminishing return in terms of increases in mudflat habitat availability, the flushing
effects of freshwater inputs into the system at these water levels would most likely be
far more beneficial in terms of increased ecological viability, mainly due to the effects
on water quality and greater productivity at lower salinities.
The CLLAMM Dynamic Habitat
114
5.7.
References
Anibal, J., Rocha, C. and Sprung, M. (2006) Mudflat surface morphology as a
structuring agent of algae and associated macroepifauna communities: A case study
in the Ria Formosa, Journal Sea Research 57: 36-46.
Bourman, R.P., Murray-Wallace, C.V., Belperio, A.P. and Harvey, N. (2000) Rapid
coastal geomorphic change in the river murray estuary of Australia, Marine Geology
170: 141-168
Boyd, S.K., Ronsky, J.L., Lichti, D.D., Salkauskas, D. and Chapman, M.A. (1999)
Joint surface modeling with thin-plate splines, Journal of Biochemical Engineering 12:
525-532.
Boyes, S.J. and Allen, J.H. (2007) Topographic monitoring of a middle estuary
mudflat, Humber Estuary, UK – Anthropogenic impacts and natural variation, Marine
Pollution Bulletin 55: 543-554.
Carter, R.W.G. (1988) Coastal Environments: An Introduction to the Physical,
Ecological and Cultural Systems of Coastlines. Academic Press, New York, 617p.
Cooke, R.U., Warren, A. and Goudie, A.S. (1993) Desert Geomorphology. London:
UCL Press
Dyer, K.R. (1998) The typology of intertidal mudflats. In: Black, K.S., Paterson, D.M.
and Cramp, A. (eds) Sedimentary Processes in the Intertidal Zone. Geological
Society, London, Special Publications 139: 11-24.
Dyer, K.R., Christie, M.C., and Wright, E.W. (2000) The classification of intertidal
mudflats, Continental Shelf Research 20: 1039-1060.
Environmental Systems Research Institute (2008), ArcGIS 9.3 Desktop. ESRI,
Redlands, California.
Glenn, N.F., Streutker, D.R., Chadwick, D.J., Thackray, G.D. and Dorsch, S.J. (2006)
Analysis of LiDAR-derived topographic information for characterizing and
differentiating landslide morphology and activity, Geomorphology 73: 131-148.
Gooverts, P. (1997) Geostatistics for Natural Resources Evaluation, Oxford
University Press, New York.
Isaaks, E.H. and Srivastava, B.M. (1989) An Introduction to Applied Geostatistics,
Oxford University Press, New York.
Kirby, R. (2000) Practical implications of tidal flat shape. Continental Shelf Research
20: 1061–1077.
Lamontagne, S., McEwan, K., Webster, I., Ford, P., Leaney, F. and Walker, G.
(2004) Coorong, Lower Lakes and Murray Mouth. Knowledge gaps and knowledge
needs for delivering better ecological outcomes. Water for a Healthy Country
National Research Flagship CSIRO, Canberra.
Le Hir, P., Roberts,W., Cazaillet, O., Christie, M., Bassoullet, P. and Bacher, C.
(2000) Characterization of intertidal flat hydrodynamics, Continental Shelf Research
20: 1433–1459.
Masero, J.A., Pérez-González, M., Basadre, M., and Otero-Saavedra, M. (1999)
Food supply for waders (Aves: Charadrii) in an estuarine area in the Bay of Cádiz
(SW Iberian Peninsula), Acta Oecologia 20: 429-434.
MDBC (2003) Description of hydrological modelling for a demonstration of the
recovery of 500GL for the Murray Mouth, Coorong and Lower Lakes, draft report
prepared by Julianne Martin, Murray Darling Basin Commission, Canberra.
The CLLAMM Dynamic Habitat
115
MDBC (2004) The Barrages Release of 2003, a report prepared by the MurrayDarling Basin Commission, Department of Water, Land and Biodiversity
Conservation and the South Australian Water Corporation.
Miles, M. (2006) Origin of bathymetric data for Lower Lakes, Goolwa and Coorong.
South Australian Department of Environment and Heritage, Adelaide.
Noye, B.J. (1973) The response of lake levels to an unsteady wind stress. Australian
Journal of the Mathematics Society Bulletin 8: 422-33.
Noye, B.J., and Walsh, P.J. (1976) Wind-induced water level oscillations in shallow
lagoons, Australian Journal of Marine and Freshwater Research 27: 417-30.
Phillips, B. and Muller, K. (2006) Ecological Character of the Coorong, Lakes
Alexandrina and Albert Wetland of International Importance, South Australian
Department for Environment and Heritage, Adelaide.
Piersma, T., De Goeij, P. and Tulp, I. (1993) An evaluation of intertidal feeding
habitats from a shorebird perspective: towards relevant comparisons between
temperate and tropical mudflats. Netherlands Journal of Sea Research 31: 503–512.
Point Software Inc. (2001) ProLink 1.15, Point Software and Sytsems, Seigen,
Germany.
Pritchard, D., Hogg, A. J. and Roberts, W. (2002) Morphological modelling of
intertidal mudflats: the role of cross-shore tidal currents, Continental Shelf Research
22: 1887-1895.
Rogers, D.J. and Paton, D.C. (2009) Spatiotemporal Variation in the Waterbird
Communities of the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship.
Rolston, A.N. and Dittmann, S. (2009) CLLAMMecology Invertebrate Key Species
Project: The distribution and abundance of macrobenthic invertebrates in the Murray
Mouth and Coorong Lagoons. CSIRO: Water for a Healthy Country National
Research Flagship.
Schwab, D. J., and Beletsky, D. (2003) Relative effects of wind stress curl,
topography, and stratification on large-scale circulation in Lake Michigan, Journal of
Geophysical Research 108(C2): 3044.
Shilo, E., Y. Ashkenazy, A. Rimmer, S. Assouline, P. Katsafados, and Y. Mahrer
(2007) Effect of wind variability on topographic waves: Lake Kinneret case, Journal of
Geophysical Research 112: C12024.
Thompson, R.O.R.Y. (1983) Set-up of Sydney Harbour by waves, wind and
atmospheric pressure, Australian Journal of Marine and Freshwater Research 34:
97-103.
Trimble Navigation Ltd. (2007) Trimble Pathfinder 4.00, Sunnyvale, California.
Uncles, R.J., Bale, A.J., Brinsley, M.D., Frickers, P.E., Harris, C., Lewis, R.E., Pope,
N.D., Staff, F.J., Stephens, J.A., Turley, C.M. and Widdows, J. (2003) Intertidal
mudflat properties, currents and sediment erosion in the partially mixed Tamar
Estuary, Ocean Dynamics 53: 239-251.
Webster, I.H. (2007) A Hydrodynamic Model for the Coorong, CSIRO Land and
Water, Canberra.
The CLLAMM Dynamic Habitat
116
5.8.
Appendices
Appendix 5.1a:Surface Validation Results showing fit between predcited
surface and measured surface for each reference site.
Salt Creek
Jack Point
Villa dei Yumpa
The CLLAMM Dynamic Habitat
117
Parnka Point
Noonameena
Long Point
Mark Point
The CLLAMM Dynamic Habitat
118
Pelican Point
Ewe Island
Barker Knoll
Mundoo Channel
The CLLAMM Dynamic Habitat
119
Goolwa Channel
Appendix 5.1b: Cross validation results for the thin plate spline models at
each site.
Reference Site
Model
Number of
Validation
Points
Mean Error
RMS Error
Regression
Function*
Salt Creek
704
0.0005
0.2548
0.962x + 0.021
Jack Point
602
-0.0014
0.1978
0.960x + 0.007
Villa dei Yumpa
628
-0.0063
0.2081
0.948x + 0.008
Parnka Point
868
-0.0120
0.4426
0.708x + 0.119
Noonameena
920
0.0003
0.0352
1.000x + 0.000
Long Point
713
0.0003
0.0642
0.999x - 0.004
Mark Point
2718
0.0001
0.0371
1.000x + 0.000
Pelican Point
1977
0.0001
0.0439
1.000x + 0.000
Ewe Island
13987
-0.0003
0.0701
1.000x + 0.000
Barker Knoll
3594
0.0008
0.1571
1.000x + 0.001
Mundoo Channel
9321
0.0001
0.0581
0.999x + 0.000
Goolwa Channel
7769
0.0001
0.0431
1.000x + 0.000
*Relationship between predicted (y) and actual (x) depths.
The CLLAMM Dynamic Habitat
120
Appendix 5.2: High resolution surface models for each reference site.
Note differences in vertical scale between sites.
Salt Creek
Jack Point
Villa dei Yumpa
Parnka Point
The CLLAMM Dynamic Habitat
121
Noonameena
Long Point
Mark Point
Pelican Point
The CLLAMM Dynamic Habitat
122
Ewe Island
Barker Knoll
Mundoo Channel
Goolwa Channel
The CLLAMM Dynamic Habitat
123
Appendix 5.3: General mudflat slope parameters at each reference site
Salt Creek
AHD
Slope in percent
2
Area (m )
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
200106
0.00
14.23
14.23
1.16
1.08
0.0 to 0.5m
274669
0.00
14.26
14.26
0.88
0.84
0.5m to 1.0m
259725
0.00
8.46
8.46
0.76
0.65
1.0m to 1.5m
94444
0.01
8.00
8.00
0.62
0.50
Jack Point
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
710500
0.00
6.79
6.79
0.50
0.47
0.0 to 0.5m
945271
0.00
8.47
8.47
0.38
0.47
0.5m to 1.0m
154373
0.00
7.44
7.44
1.27
0.62
1.0m to 1.5m
4150
0.00
2.65
2.65
0.38
0.44
Villa dei
Yumpa
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
1489699
0.00
3.50
3.50
0.21
0.22
0.0 to 0.5m
1004309
0.00
8.92
8.92
0.38
0.38
0.5m to 1.0m
114184
0.00
12.13
12.13
1.33
0.98
1.0m to 1.5m
12675
0.00
13.07
13.07
1.43
2.56
Parnka Point
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
137646
0.00
60.60
60.60
1.15
1.43
0.0 to 0.5m
1367330
0.00
75.01
75.01
0.31
0.69
0.5m to 1.0m
342823
0.00
55.95
55.95
1.22
1.28
1.0m to 1.5m
222285
0.00
30.77
30.77
1.14
1.04
Noonameena
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
324747
0.00
0.97
0.97
0.42
0.27
0.0 to 0.5m
177023
0.07
1.27
1.20
0.68
0.15
0.5m to 1.0m
201053
0.01
1.26
1.25
0.56
0.20
1.0m to 1.5m
1221
0.00
0.52
0.52
0.20
0.15
The CLLAMM Dynamic Habitat
124
Long Point
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
188667
0.10
5.59
5.48
0.77
0.49
0.0 to 0.5m
231888
0.18
5.37
5.19
0.67
0.41
0.5m to 1.0m
145134
0.03
4.41
4.38
0.53
0.31
1.0m to 1.5m
12
0.49
0.53
0.04
0.50
0.01
Mark Point
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
239840
0.00
2.73
2.73
0.67
0.46
0.0 to 0.5m
175088
0.06
3.04
2.98
0.86
0.47
0.5m to 1.0m
102604
0.00
3.05
3.05
0.64
0.45
1.0m to 1.5m
0
0.00
0.00
0.00
0.00
0.00
Pelican Point
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
395667
0.00
4.63
4.63
0.57
0.62
0.0 to 0.5m
463754
0.00
3.97
3.97
0.39
0.48
0.5m to 1.0m
121506
0.00
1.97
1.97
0.41
0.39
1.0m to 1.5m
0
0.00
0.00
0.00
0.00
0.00
Ewe Island
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
310847
0.00
18.10
18.10
0.59
0.74
0.0 to 0.5m
210577
0.00
4.66
4.66
0.71
0.74
0.5m to 1.0m
174780
0.00
4.53
4.53
0.84
0.75
1.0m to 1.5m
16071
0.00
4.29
4.29
0.65
0.87
Barker
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
125549
0.00
26.29
26.29
1.57
2.48
0.0 to 0.5m
86103
0.01
23.42
23.41
1.91
2.45
0.5m to 1.0m
88209
0.00
20.06
20.06
1.55
1.74
1.0m to 1.5m
1815
0.01
8.46
8.46
1.11
1.16
The CLLAMM Dynamic Habitat
125
Mundoo
Channel
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
565357
0.00
10.47
10.47
0.41
0.95
0.0 to 0.5m
129737
0.00
7.24
7.24
0.23
0.31
0.5m to 1.0m
0
0.00
0.00
0.00
0.00
0.00
1.0m to 1.5m
0
0.00
0.00
0.00
0.00
0.00
Goolwa
Channel
AHD
Area (m2)
MIN
MAX
RANGE
MEAN
SD
-0.5 to 0.0m
128855
0.01
15.13
15.12
0.85
1.01
0.0 to 0.5m
66556
0.01
14.76
14.74
1.65
1.07
0.5m to 1.0m
75453
0.00
8.11
8.11
1.28
0.76
1.0m to 1.5m
5179
0.00
2.63
2.62
0.27
0.30
Appendix 5.4: Hypsometric Curves for the Reference Sites (cummulative
area in m2)
Goolwa Channel
The CLLAMM Dynamic Habitat
126
Mundoo Channel
Barker Knoll
Ewe Island
The CLLAMM Dynamic Habitat
127
Pelican Point
Mark Point
Long Point
The CLLAMM Dynamic Habitat
128
Noonameena
Parnka Point
Villa dei Yumpa
The CLLAMM Dynamic Habitat
129
Jack Point
Salt Creek
The CLLAMM Dynamic Habitat
130
6. Spatial modelling of mudflat availability and fish
habitat in the Coorong
Sunil K. Sharma1*, Simon N. Benger2, Jason E. Tanner1 and Ian T. Webster3
1
SARDI Aquatic Sciences, PO Box 120, Henley Beach, SA 5022.
2
School of Geography, Population and Environmental Management, Flinders
University, GPO Box 2100, Adelaide, SA 5001
3
CSIRO Land & Water, GPO Box 1666, Canberra, ACT 2601
*corresponding author, Phone +61 (8) 8207 5448, Fax +61 (8) 8207 5448, E-mail:
[email protected]
6.1.
Executive Summary
Water level and salinity are the key ecological drivers in the Coorong. Water level
determines the availability of mudflat habitats, which constitute vital habitat for many
species of waterbirds, supporting highly productive foraging grounds. Salinity has
long been acknowledged as the most significant water quality parameter directly
influencing the distribution of all biological communities in the Coorong including fish,
macrophytes and micro-invertebrates and indirectly influencing waterbird numbers
through limiting food resources.
This report presents spatial models of mudflat and fish habitat to predict the
availability of mudflats for the waterbirds and habitats for key fish species under a
range of hydrodynamic scenarios using water level and salinity simulations. The first
spatial model predicts the location and extent of mudflats, defined as soft sediment
areas that are either immersed or covered by no more than 12 cm of water, where
most waterbird foraging occurs. The second spatial model predicts occurrence
probabilities of key fish species throughout the Coorong, based on the results of
logistic regression modelling using observed presence from an extensive sampling
effort carried out between 2006-08. Both models use water level data from the
hydrodynamic model generated for the current climate (Scenario A) prescribed by the
CSIRO Sustainable Yields project, while the second model also uses salinity data
from the hydrodynamic model. The spatial models run continuously for the series of
simulated water level and salinity data produced by the hydrodynamic model, and
subsequently maps are generated for each scenario.
Three representative sites from the set of 12 used by CLLAMMecology, Barker Knoll,
Noonameena and Salt Creek, were analysed for maximum and minimum water
levels in two wet (1976 and 1993) and two dry (1988 and 2005) years, selected
based on salinity levels in the South Lagoon. Mudflat availability varied spatially and
temporarily along the Coorong, and is influenced by the primary factors of tide, wind,
rainfall and evaporation, some of which are dependent on the distance from the
Murray Mouth and are affected by seasonal variation. Underwater topography was
also found to be significant for mudflat availability. The modelling of mudflats at
different water levels suggests that an average water level of 0.12 mAHD (Australian
Height Datum) gives the maximum average mudflat area in the three reference sites,
with the majority of the mudflats located on the eastern shores.
Out of seven key fish species, Yelloweye Mullet (Aldrichetta forsteri), Smallmouth
Hardyhead (Atherinosoma microstoma), Greenback Flounder (Rhombosolea
tapirina) and Tamar River Goby (Afurcagobius tamarensis), demonstrated a
significant relationship with salinity. Among the three different salinity gradient
The CLLAMM Dynamic Habitat
131
scenarios examined, a salinity range from 5 to 90 g/L along the Lagoon was found to
be the best in terms of the suitability of the entire Lagoon for these four key species,
as well as for supporting other important biological communities including both
macrophytes and infauna.
While this study suggests an appropriate water level and salinity gradient for
maximizing available mudflat areas, as well as the “health” of biological communities
in the Coorong, it is unable to provide advice on the amount and timing of barrage
outflow to achieve them. Additional research and modelling considering all physical
and hydrological factors may be able to provide an answer to the question: how
much and how frequently does freshwater need to be released through the barrages
to maintain an optimum water level and salinity gradient in the Coorong? Mixing
mechanisms, groundwater inputs and local precipitation inputs for the Coorong
remain unknown. However, from a management perspective, managers can use the
spatial models developed through this study as management tools to instantly assess
mudflat availability and the habitat suitability for these species for specified flow
scenarios, and to assist informed decisions on barrage outflows, once the Lower
Lakes recharge.
6.2.
Introduction
The Coorong lagoonal system forms part of one of the largest estuaries in Australia
and has been recognized nationally and internationally for its ecological, social and
economic significance (Seaman 2003; Edyvane 1999). The ongoing drought
conditions in the Murray Darling Basin have led to a situation where virtually no
freshwater flows have entered the Lagoon since 2003 (CLLAMM 2008). The ecology
of the Lagoon and surrounding region has been negatively affected by rising salinity
and reduced water levels, particularly in the South Lagoon (Geddes 2005a; 2003).
Consequently, ecosystem degradation in the Coorong is threatening many iconic bird
and fish species. For example, salinities in the South Lagoon have risen to become
hypersaline (100-160 g/L) in recent years, and key plant species like Ruppia
tuberosa (Tuberous tassel) can no longer survive and have become restricted to the
North Lagoon (CLLAMM 2008).
Before the recent ecological degradation, the Coorong supported a diverse array of
interconnected habitats driven by salinity and water level. Lamontagne et al. (2004)
acknowledged salinity and water level as the key ecological drivers of the system.
Based on the salinity gradient, the habitats in the Coorong were differentiated into
freshwater in the areas around the Murray Mouth, estuarine at the upper end and
marine at the lower end of the North Lagoon and in the South Lagoon. Specific
biological communities of macrobenthos, fish and phytoplankton colonised these
habitats, depending on their salinity tolerance.
In the Coorong, water level and periodic inundation also play a key role in defining
spatial extent for fish habitats as well as availability of mudflats. These mudflats
constitute highly productive feeding grounds for wading birds and attract large
numbers of local and intercontinental species (Rogers and Paton 2009a; Wilson
2001), resulting in the regions designation as a Ramsar listed wetland. As a
consequnece, Australia is obligated to maintain the area in the state that it was in at
the time of declaration. However, Wilson (2001) reported a drastic reduction in the
number of waterbirds in the Coorong since early 1980s. In the past three decades,
the largest flock of birds (234,543) counted in 1982 while 48,425 birds were counted
in 2001. The number of Sharp-tailed Sandpipers, Red-necked Stints, Curlew
Sandpipers, Red-necked Avocets and Red-capped plovers declined sharply over the
period (Wilson 2001).
Restoration and conservation of all these habitats are considered highly significant
for the biological diversity and ecological sustainability of the region. Freshwater
The CLLAMM Dynamic Habitat
132
inputs through the barrages that separate the Coorong and the Lower Lakes, and
through the Upper South East Drainage (USED) scheme, are essential to maintain
water level and salinity gradients as well as frequent inundation of mudflats.
Precipitation and evaporation influence water level and salinity in the Coorong
(Webster 2007), whereas tide and wind affect mudflat inundation and vary
temporarily and spatially along the lagoon (see Chapter 5). The mudflats around the
Murray Mouth are influenced by daily tidal routines whereas large tidal events such
as king tides may inundate mudflats in areas beyond the Murray Mouth region.
Spatio-temporal variation in all these factors impacts on salinity and water levels
along the lagoon. In response to fluctuations in salinity and water levels, the available
habitats are subjected to change, which ultimately affects the composition and
productivity of biological communities.
The restoration of the diverse ecosystems of the Coorong is of utmost importance to
maintain its status as a Ramsar Wetland of International Significance as well as its
iconic site status under the Living Murray Initiative. It has been acknowledged that a
detailed understanding of the ecology of the Coorong is necessary to prescribe a
decision-support framework for an effective intervention, which would aim to restore
the vital lagoonal ecosystem in the Coorong. The necessity of ecological knowledge
at the local level and, in particular, understanding of the relationships between
species and their environment, have been widely acknowledged by many ecologists
as key to the successful conservation and management of ecosystems (Aber 2007;
Carter et al. 2006; Gibson et al. 2004).
The accelerated degradation of ecosystems and biodiversity in the Coorong require a
detailed understanding of the species-environment relationships for iconic bird and
fish species to predict their habitats at varying salinity and water levels. Rogers and
Paton (2009a) and Noell et al. (2009) conducted a detail on the waterbirds and fish
species in the Coorong under the Key Species Project of the CLLAMMecology
research cluster, respectively, and provide a summary of the important species
present. The modelling of relationships enables us to predict habitat availability under
changed climatic conditions. The Dynamic Habitat Project under the Coorong, Lower
Lakes and Murray Mouth Ecology (CLLAMMecology) Research Cluster aims to
develop spatial models to predict the habitats for waterbirds and key fish species
under varying hydrodynamic scenarios. The first model predicts mudflat availability
for waterbirds by quantifying available mudflat areas up to 12 cm water depth, where
the majority of waterbird foraging occurs (Rogers, pers. comm.; Wildlife Habitat
Management Institute 2000). The accessibility of mudflat for waterbirds is likely to be
determined by their size. As the average size of the major waterbird species does not
exceed 21 cm (for Sharp-tailed Sandpiper), the shallower mudflat areas are generally
available for foraging. However, in a recent study, Rogers and Paton (2009a) found
that the foraging behaviour of waterbirds is influenced by the water depth and
foraging grounds are likely to be inaccessible below 20 cm depth in the Coorong. The
second model predicts occurrence probability for key fish species including four
commercially important species: Black Bream (Acanthopagrus butcheri), Greenback
Flounder (Rhombosolea tapirina), Yelloweye Mullet (Aldrichetta forsteri) and
Mulloway (Argyrosomus hololepidotus), two common small-bodied estuarine fish:
Smallmouth Hardyhead (Atherinosoma microstoma) and Tamar River Goby
(Afurcagobius tamarensis) and one species of conservation significance: Congolii
(Pseudaphrites urvillii), based on salinity levels along the lagoon. Both models use
water level data from the hydrodynamic model developed for the Coorong (Webster
2007) generated for the current climate (Scenario A) as prescribed by the CSIRO
Sustainable Yields project (Chiew et al. 2008), and the second model also uses
salinity data generated from the hydrodynamic model.
The first model compliments the study on mudflat geomorphology and availability at
different water levels in the Coorong described in Chapter 5. Chapter 5 discusses the
geomorphological characteristics of mudflats at the 12 CLLAMMecology reference
sites and presents an analysis of mudflat availability at varying water depths in these
The CLLAMM Dynamic Habitat
133
sites. The model in the current study integrates digital elevation models (DEMs)
described in the previous chapter and hourly water level data from Webster’s (2007)
model to predict foraging ground up to 12 cm water depth; a part of the mudflat
specifically utilized by waterbirds in the Coorong. The model generates mudflat
habitat maps for hourly water level data showing the availability of mudflats in the
eastern (landward), western (seaward) sides of the Coorong and in the channel
areas. The second model predicts the occurrence probability of the key fish species
in relation to the salinity levels in the Coorong.
Both habitat models are developed within a GIS framework in ArcGIS (ver. 9.3) and
iteratively apply hourly water level and salinity data generated by the hydrodynamic
model. These models are not truly dynamic process models as they neither simulate
the physical process for predicting habitat change nor use outputs of previous
iterations as input to the subsequent iteration (Environmental Systems Research
Institute 2006). However, these models use hourly data on water level and salinity
from the hydrodynamic model and deliver hourly habitat maps depicting the dynamic
influence of these ecological factors on habitat availability in the Lagoon.
6.3.
Methods
6.3.1 Datasets
Bathymetry for the reference sites
The bathymetry generated for studying mudflat geomorphology and availability over
the 12 Coorong reference sites (see Chapter 5 in Appendix 5.2) was used in these
models. Depth data for the North Lagoon were collected by the South Australia
Water Corporation (SA Water) by using an echo-sounder and GPS mounted on a
boat. However, such a survey was not feasible in the South Lagoon due to the
shallow water depth (Miles 2006). Hence, depths at four reference sites in the South
Lagoon were measured manually using a combination of techniques (see Chapter 5).
The bathymetry for all reference sites was derived by interpolating the depth data
using Radial Basis Functions available in the Geostatistical Analyst extension in
ArcGIS (ver. 9.3) (Environmental Systems Research Institute 2001). The bathymetry
represented the topography at these reference sites in metres (m) in relation to
Australian Height Datum (AHD) with 1 m horizontal resolution and 0.001 m vertical
resolution. The process of bathymetry development for the reference sites in the
Coorong is described in Chapter 5 and the resultant models for the 12 reference sites
are also illustrated there.
Hydrodynamic model: water level and salinity data
Webster (2007) developed a hydrodynamic model which simulates water levels and
salinity within the Coorong as these respond to barrage flows, USED flows, sea level
changes, wind, evaporation, precipitation and exchange through the Murray Mouth.
This one-dimensional model outputs time series of simulated water level and salinity
for each of 102 and 14 cells, respectively, along the centreline of the Coorong
between the Murray Mouth and the southern end of the Lagoon. Water level is output
at one km intervals, whereas salinity is output at intervals of 5 to 10 km (Figure 6.1).
The CLLAMM Dynamic Habitat
134
Figure 6.1. Locations of water level and salinity data points generated through the
hydrodynamic model in relation to the CLLAMMecology reference sites. (GC = Goolwa
Channel; MC = Mundoo Channel; EI = Ewe Island; BK = Barker Knoll; PP = Pelican Point; MP
= Mark Point; LP = Long Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei
Yumpa; JP = Jack Point and SC = Salt Creek.
The hydrodynamic model was run using barrage flows derived for a single climatic
and development scenario from the CSIRO Sustainable Yields (SY) Project. These
flows were made available by the Murray-Darling Basin Authority (MDBA) and are
based on SY simulations that have been modified by the Victorian 2030 climate
approach. Synthetic time series of flows were constructed by analysing the daily time
series of climatic data for the period 1891-2008 in combination with an inflow model
run using the current state of agricultural development and specifically current water
management rules. This analysis is being used to assess the current and likely future
water availability in the Murray-Darling Basin (Chiew et al. 2008). The scenario we
use develops a flow time series using the historical climate sequence and inflows
adjusted to the current level of development. This represents the baseline scenario.
For habitat modelling of the Coorong, we chose to predict habitat availability for the
baseline scenario (Scenario A). The hydrodynamic model simulates water level and
salinity for 117 years (the period between 1891 and 2008). Salinity in the South
Lagoon was found to be a good indicator of very wet and dry years as this part of the
Coorong better reflected the influence of consecutive wet and dry conditions
producing ‘very low’ or ‘very high’ salinities compared to the North Lagoon. Two ‘very
wet’ and ‘very dry’ years in the past decades were selected for a comparison of
habitat availability under these conditions. Based on the simulated daily average
salinities for 1891-2008 under scenario A, we identified 1976 and 1993 as wet years;
and 1988 and 2005 as dry years. The hydrodynamic model was run for these years
using scenario A. Water level and salinity data were linked to the geographical
coordinates of the respective points.
The CLLAMM Dynamic Habitat
135
Other GIS layers
The site boundary up to the high water mark level was digitised from the 2003
orthorectified airphoto to specify the area for analysis. The high water mark signifies
the edges on the eastern and western shores of the lagoon. In order to specify the
mudflats on either side of the shores, masking layers were created for the eastern
and the western shores and applied in the model.
Fish Catch Data
The fish habitat modelling described in section 6.3.2 used fish catch data and salinity
as the only predictor variables. These data were collected along the Coorong
between October 2006 and July 2008. A detailed description of the sampling
methods used is described in Noell et al. (2009).
6.3.2 Spatial Model Development
Modelling mudflat availability
A spatial model for predicting mudflat availability under various flow scenarios was
developed by integrating the water level output from the hydrodynamic model with
the fine resolution bathymetry. The model was initially developed in ModelBuilder
within the ArcGIS platform utilizing functionality and tools available in ArcToolBox
including spatial analyst, data management, conversion and analysis tools. Once the
model was run successfully and delivered desirable outputs, the model was exported
as a python script. The script was modified to incorporate looping ability so that
hourly water level from the hydrodynamic model could be used to generate maps for
mudflat availability at the same interval. The model development processes for both
models are described in the following sections.
Model for single run developed in ArcGIS Model Builder platform
The fine resolution bathymetry (1x1 m), water level data from the hydrodynamic
model and a boundary layer covering up to the high water mark level were integrated
into ArcGIS ModelBuilder to predict the spatio-temporal availability of mudflat
habitats in the Coorong. The entire model is shown in Figure 6.2 and the four parts of
the model are illustrated in Figures 6.3-6.6. The model is composed of six model
parameters (blue ovals) and uses 12 processes (yellow rectangles) and generates 12
outputs (green ovals) including 9 intermediate and 3 final outputs.
Part A
Part B
Part C
Part D
Figure 6.2. Overall model for predicting mudflat availability in the Coorong. Each part is
presented in detail in Figures 6.3-6.6. Dark blue ovals represent input data or layer, light blue oval
represents parameter input, yellow rectangles represent processes and green ovals represent
model intermediate and final outputs.
The water level data from the hydrodynamic model contained hourly time series data
for 102 points distributed linearly along the Lagoon. These data points were first
assigned geographic coordinates and a field name was given to each column of
The CLLAMM Dynamic Habitat
136
water level data. The water level data from the hydrodynamic model were
interpolated using the inverse distance weighted (IDW) interpolation in the Spatial
Analyst extension in ArcGIS (ver 9.3) (Environmental Systems Research Institute
2009). IDW applies an inverse power weighting function to the distance of the
measured points. Nearby points have higher weights and influence than points
located at greater distance (De Smith et al. 2006). The goodness of fit of the models
generated by IDW and other methods were compared using the Geostatistical
Analyst extension in ArcGIS. Although Radial Basis Functions are considered to be
the best method for topographical surface interpolation with dense sampling data
(Environmental Systems Research Institute 2001), in this case the IDW produced the
lowest root mean square error. Kriging produced the highest root mean square error
of all three methods. Hence, the IDW was used for interpolating water level data at
specified times to generate a two dimensional water level surface of 1 x 1 m grid cells
(Water Level Raster). The area of interest for raster interpolation was specified by
setting the extent to the boundary layer in the model properties. The water level
surface (Water Level Raster) was applied to the DEM of the site (DEM site) to derive
an output (Water Level Mask 01) with values 0 and 1 using Map Algebra (1). In the
output raster, a value of 1 signifies a ‘true’ condition and represents the area less
than or equal to the water level while the 0 value indicates a ‘false’ condition and
represents the area above the water level.
Figure 6.3. Mudflat habitat model part (A).
A raster mask was required to delineate the area at the specified water level for
identifying the mudflat area used by waterbirds. In the model part B (Figure 6.4), the
output of the model part A (Water Level Mask 01) was assigned 1 to 0 and 0 to
NoData (Reclassify 1) to generate a raster mask with 0 value (Water Level Mask 1).
In the Map Algebra (2), the water level raster (Water Level Raster) and raster mask
(Water Level Mask 1) were applied together with water depth up to 12 cm from the
water level to the DEM of the site. The water depth up to 12 cm was used to extract
the mudflat area utilized by waterbirds in this case. However, this value could be
changed to suit the various depths and mudflat areas utilized by different species of
waterbirds. In the output raster (Subtidal Mudflat area 01), a value of 1 indicates the
area up to 12 cm water depth and 0 signifies the subtidal area deeper than 12 cm
water depth. Reclassify (2) was applied to Subtidal Mudflat area 01 to assign both
The CLLAMM Dynamic Habitat
137
values 0 and 1 to NoData (Exposed area) for identifying the exposed mudflat area
above the water level whereas reclassify (3) excluded the mudflat area with 1 value
and the rest to NoData (Mudflat area). The outputs of both reclassify (2) and (3) are
shown in the model part C (Figure 6.5).
The first output of the model, exposed area between the water level and the high
water mark boundary, was obtained by employing an ‘extraction by mask’ operation.
This operation uses the high water mark boundary layer as an input layer and the
raster specifying the area below water level as a mask (Exposed area). Similarly,
another ‘extraction by mask’ operation was applied to the DEM of the site by using
the Mudflat area raster to obtain the DEM for the mudflat area up to 12 cm water
depth from the water level (Mudflat DEM). The DEM for the mudflat area (Mudflat
DEM) changed from floating point to integer data type (DEM integer) and would be
the input for the model part D (Figure 6.6).
DEM for Site
Water Level Raster
Figure 6.4. Mudflat habitat model part (B).
DEM
SiteSite
DEMforfor
Reclassify
Reclassify
(2)
2
Reclassify 3
Water Level Raster
Figure 6.5. Mudflat habitat model part (C).
The CLLAMM Dynamic Habitat
138
In the model part D (Figure 6.6), the mudflat areas were segregated into the eastern
shore (landward) and the western shore (seaward) by applying the extracting mask
for the east shore and mask for the west shore, respectively. These outputs were the
final outputs from this model and they show the mudflat availability at 1 cm vertical
resolution up to 12 cm depth on both shores of each reference site.
WL Depth
Figure 6.6. Mudflat habitat model part (D).
Figure 6.7 provides the model interface for running the model where the model
parameters, input files and the location and names of the three final outputs are
specified. The input files are comprised of four feature shape files and one DEM for
any given reference site. The water level dataset supplies the water level data at 102
points and these data are interpolated to generate a water level surface. Since the
model predicts the mudflat availability for water level at one point in time, it requires
specifying the water level data point from the attribute table of water level dataset to
be used for running the model. The high water mark boundary, and masks for the
east and west shores, are used to generate three output files: exposed mudflat area
between the water level and the high water mark boundary and available mudflats on
the eastern and the western shores.
The CLLAMM Dynamic Habitat
139
Figure 6.7. Interface for implementing the model in ArcGIS.
Dynamic model for predicting mudflat availability
The above model was developed in the ModelBuilder platform in ArcGIS and runs for
one sequence of water level data at one point in time. The model requires specifying
the water level data to be used for modelling each time. Since the hydrodynamic
model simulates water level data every hour for the specified flow and physical
scenario, a model was written in geo-processing python script for automating the
execution using a series of water level data in the dataset. The model terminates only
at the end of the water level data in the dataset and generates three outputs for each
water level time step. The python script for the model is given in Appendix 6.1.
For assessing the spatial variability in the habitats along the Coorong, three sites,
one close to the Murray Mouth (Barker Knoll), a second near the middle of the North
Lagoon (Noonameena) and the third near the end of the South Lagoon (Salt Creek)
were chosen from the 12 reference sites used by CLLAMMecology. Seasonal
variations were taken into account by analysing habitat availability in January and
July, representing two extreme seasonal conditions in summer and winter,
respectively. The daily averages for minimum and maximum water levels were used
to cover the full spectrum of water level variations in these two months. The days
with the minimum and maximum daily average water levels were chosen for both wet
and dry years. The model was implemented for the simulated hourly water level data
between 6:00 AM to 8:00 PM and analysed for temporal variation in mudflat
availability at the three sites. This time period was chosen to approximate maximum
daylight (summer) hours when foraging by waterbirds would be likely to occur.
Modelling fish habitat
Spatial model for predicting fish habitat
Logistic regression was applied within a GIS framework for predicting the probability
of occurrence of key fish species in the Coorong. This method has been used for
spatial modelling of habitat suitability (Gross et al. 2002; Shriner et al. 2002) and
environmental management (Mathew et al. 2007; Xie et al. 2005; Álvarez-Arbesú and
Felicísimo 2002). Logistic regression is a general linear regression method and is
developed to explore the relationship between a discrete response variable
The CLLAMM Dynamic Habitat
140
(especially binary form data such as presence and absence) and independent
variables (Hosmer et al. 1989). A general formula for a logistic regression is similar to
that of a linear regression and is shown as equation 6.1.
Figure 6.8. Predicting fish habitat based on fish occurrence probability using logistic regression.
Yf = b0f + b1f x1f + b2f x2f + …….. + bnf xnf
(6.1)
Where Yf is the linear value for species “f”, b0f is the coefficient for the intercept, b1f,
b2f, to bnf are variable coefficients and x1f, x2f to xnf are the independent variables.
The species occurrence probability is predicted based on the linear value (Yf) derived
from equation 6.1 for the associated significant predictor variables and the formula is
given as equation 6.2.
Probability of Occurrence = (1 / (1 + exp (-Yf )))
(6.2)
The probability values range between 0 and 1. A value close to 1 represents higher
occurrence probability.
A spatial model was developed to predict the probability of occurrence of fish species
based on salinity levels using logistic regression in ArcGIS (ver. 9.3) (Figure 6.8). In
addition to salinity, other water quality variables including water temperature, pH,
dissolved oxygen and secchi depth may also be used in the model. However, salinity
is the only variable generated by the hydrodynamic model, and thus is the only
variable against which predictions can be made under the flow scenario examined
here. Both of the outputs from the hydrodynamic model: salinity and water level, and
also the full Digital Elevation Model (DEM) for the Coorong (see Chapter 3) were
used in the model.
The salinity level data was interpolated for the Coorong using the inverse distance
weighted (IDW) method. The coefficients derived from the logistic regression for the
intercept and salinity specific to each fish species were used in equation 6.1 to obtain
linear prediction value for the species. Subsequently, the linear value for the species
was transformed to a probability value using equation 6.2 in the single output map
algebra function.
The CLLAMM Dynamic Habitat
141
The water level data obtained from the hydrodynamic model were also interpolated
and integrated with the DEM of the Coorong to derive the water surface for the
Lagoon at the specified water level. The resultant water surface map is a binary map,
where 1 signifies the area under water and 0 signifies the area above the specified
water level. Finally, the output for the occurrence probability was combined with the
water level surface. The final map illustrated the probability of occurrence of fish
species based on the salinity level and water level generated by the hydrodynamic
model.
Figure 6.9 provides the model interface for implementing the logistic model for
predicting the occurrence probability of fish species in the Coorong. The input files
included salinity data, water level data, the DEM and the Lagoon boundary layer. The
coefficients specific to individual fish species derived from the logistic regression
must be entered to calculate the occurrence probability. The probability map and the
water surface layer are integrated into the final output specifying the habitat with
varying probability for specified salinity and water levels.
Figure 6.9. Interface for implementing the model in ArcGIS.
Dynamic model for predicting fish habitat
A spatial model for predicting fish habitat at varying salinity and water levels was
developed in the ModelBuilder platform in ArcGIS 9.3 and runs for one sequence of
water level and salinity data at one point in time. The model requires specification of
the salinity and water level data to be used for modelling each run. Since the
hydrodynamic model simulates salinity and water level data every hour for the
specified flow and physical scenario, a model was written in geo-processing python
script for automating the execution using a time-series of salinity and water level data
The CLLAMM Dynamic Habitat
142
available from the hydrodynamic model. The dynamic fish habitat model terminates
only at the end of salinity and water level data in the dataset and generates an
occurrence probability map for each hourly time step. The python script for the model
is given in Appendix 6.2.
Fish catch data (number) were converted into binary data (1 for presence and 0 for
absence) for the key fish species in the Coorong. Out of 94 data samples, about 70%
of the data (65 samples) were randomly selected for modelling the relationship
between the species and salinity whereas the remaining data (29 samples) were
used for validating the relationship in SPSS 16.0 (SPSS Inc). For classification,
cases with a predicted value above a cut-off value of 0.5 were classified as ‘present’
and below 0.5 were classified as ‘absent’.
The predictive power of the models was evaluated by using the validation dataset,
and analysed for accurate classification of the data into presence and absence
against true presence and absence. The models were also analysed for Receiver
Operator Characteristic (ROC) curves in SPSS 16.0 (SPSS Inc.). The ROC curve is
widely used as a tool for measuring the accuracy of a model by evaluating the model
performance in classifying a variable into two possible outcomes (Gönen 2006). The
curve is a graphical representation of the probability of a true positive against a false
positive at the specified cut-off value. The sensitivity is used to represent the
probability of correctly classified data (true positive) and is plotted on the y-axis. The
specificity implies the probability of the correct classification of absence data and is
plotted on the x-axis of the ROC curve. The area under the curve represents the
probability estimated by the model that a randomly chosen presence case will
exceed the probability prediction for a randomly chosen absence case. Thus, the
ROC curve is considered as an important tool for measuring of the accuracy of the
model. A large area under the curve implies a high level of accuracy in the model.
The average monthly salinity representing a low, medium and high range, as derived
from the hydrodynamic model, were chosen for predicting the probability of
occurrence of the key fish species in the Coorong. The monthly average salinity was
low in July 1976, medium in July 1988 and high in January 2005.
6.4.
Results
6.4.1 Spatial and temporal variation in water level and salinity in the
Coorong
The monthly average water levels were higher in July than in January for 1976, 1988,
and 2005, but not for 1993 (Figure 6.10). Relatively high rainfall and low evaporation
during the winter months contributed to high water levels in July. The monthly
average water level reached a maximum along the Lagoon and steadily rose
between Tauwitchere Channel (0.77 mAHD) and Salt Creek (0.87 mAHD) in July
1976. Water levels were reduced along the Lagoon during January, particularly in dry
years. Less rainfall (<22 mm at Meningie) and rising evaporation in the summer
months (above 225 mm for the lakes during December and January) resulted in
significant loss of water, which was complimented by lower or no freshwater inputs
over the barrages, resulting in drastically lower water levels in the Lagoon. The
monthly average water levels were reduced more markedly in the South Lagoon than
in the North Lagoon with average differences in water levels of 0.27 mAHD and 0.17
mAHD in January 1988 and 2005, respectively.
The CLLAMM Dynamic Habitat
143
Water level (mAHD)
1
0.8
0.6
0.4
0.2
0
-0.2
1
11
21
31
41
51
61
71
81
91
101
Locations (North to South)
Jan-1976
Jul-1976
Jan-1988
Jul-1988
Jan-1993
Jul-1993
Jan-2005
Jul-2005
Figure 6.10. Monthly average water level along the Lagoon in the summer (January) and
winter (July) months for the wettest years (1976 and 1993) and driest years (1988 and 2005)
in the past four decades. The hydrodynamic model does not predict the water level for Parnka
Point (location 55).
The three selected reference sites also followed the same general pattern for the
monthly average water levels. The water levels were consistently lower in January
than July in 1976, 1988 and 2005. However, higher water levels were predicted for
January than July in 1993 at Noonameena and Salt Creek (Figure 6.11 and 6.12). At
all three sites, the average water levels exceeded 0.51 and 0.57 mAHD in July 1976
and 1988, respectively. Salt Creek had the lowest average water level of -0.13
mAHD in January 1988 and of -0.09 mAHD in January 2005 (Figure 6.12). At Barker
Knoll, the average water level was 0.17 mAHD in January, while it reached above
0.57 mAHD in July 1988. The hydrodynamic model predicts hourly as well as daily
variations in the water level along the Coorong. Generally, water levels subsided
during the night and rose to maximum levels in the afternoon, primarily due to wind
effects. The daily variation in water levels at Barker Knoll, Noonameena and Salt
Creek for January and July of 1976, 1988, 1993 and 2005 are presented in Appendix
6.3.
The CLLAMM Dynamic Habitat
144
Average water level (mAHD)
0.8
Jan. 1976
Jul. 1976
Jan. 1993
Jul. 1993
0.6
0.4
0.2
0
Barker1 Knoll
Noonameena
Salt Creek
Reference sites
Figure 6.11. Monthly average water level at three reference sites for January and July in the
wettest years (1976 and 1993).
Average water level (mAHD)
1
0.8
Jan. 1988
Jan. 2005
Jul. 1988
Jul. 2005
0.6
0.4
0.2
0
-0.2
Barker1 Knoll
Noonameena
Salt Creek
Reference sites
Figure 6.12. Monthly average water level at three reference sites for January and July in
1988 and 2005.
Higher water levels generally coincided with lower salinity and vice versa. For
example, the lowest average salinity of 18.2 g/L was predicted for July 1976 when
the monthly average water level was at its highest (0.8 mAHD) along the Lagoon.
Similarly, the highest average salinity (85.5 g/L) was predicted at the lowest average
water level (0.18 mAHD) in January 2005 (Figure 6.13). The Lagoon had the lowest
salinity range of ~ 2 – 48 g/L in July 1976 and reached the highest range of ~41 –
125 g/L in January 2005. Predicted salinities in the North Lagoon were consistently
lower than in the South Lagoon.
The CLLAMM Dynamic Habitat
145
140
Salinity (g/L)
120
100
80
60
40
20
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Locations (North to South Lagoon)
Jan-1976
Jul-1976
Jan-1988
Jul-1988
Jan-1993
Jul-1993
Jan-2005
Jul-2005
Figure 6.13. Monthly average salinity levels along the Lagoon in the summer (January) and
winter (July) months for the wettest years (1976 and 1993) and driest years (1988 and 2005)
in the past four decades.
The average salinity was predicted to be below 15 g/L in the northern part of the
Lagoon and did not exceed 50 g/L in the South Lagoon in 1976 and 1993 (Figure
6.14). The average salinity was above 40 g/L at Barker Knoll, 50 g/L at Noonameena
and 89 g/L in Salt Creek in January 1988 (Figure 6.15). The salinity further increased
and reached approximately 125 g/L at Salt Creek in January 2005. Raised water
levels in July 1988 and July 2005 had some influence on salinity with levels reduced
to almost half those in January. The daily variation in salinity for these three sites for
January and July of 1976, 1988, 1993 and 2005 is presented in Appendix 6.4.
Average salinity (g/L)
60
Jan. 1976
Jul. 1976
Jan. 1993
Jul. 1993
40
20
0
Barker1 Knoll
Noonameena
Salt Creek
Reference sites
Figure 6.14. Monthly average salinity (g/L) at three reference sites for January and July in
1976 and 1993.
The CLLAMM Dynamic Habitat
146
Average salinity (g/L)
160
Jan. 1988
Jan. 2005
Jul. 1988
Jul. 2005
120
80
40
0
Barker1 Knoll
Noonameena
Salt Creek
Reference sites
Figure 6.15. Monthly average salinity (g/L) at three reference sites for January and July in
1988 and 2005.
6.4.2 Mudflat availability at the reference sites
Of the three reference sites modelled for this study, the mudflat habitats were
available primarily on the eastern and western shores of Noonameena and Salt
Creek (Figure 6.16 and 6.17) while some mudflats also appeared within the channel
at Barker Knoll (Figure 6.18). The majority of mudflat areas were present on the
eastern shore at Barker Knoll and Salt Creek whereas both shores had similar
mudflat areas at Noonameena. These maps present spatial as well as temporal
variations in the amount of mudflat areas available to waterbirds with changes in
water levels. A detailed analysis of the distribution of mudflats at the maximum and
minimum water levels in the winter (July) and summer (January) months of the wet
and dry years between 6:00 to 20:00 is presented below.
The CLLAMM Dynamic Habitat
147
Figure 6.16. Map of mudflat availability generated by the spatial model at Noonameena on
22 July 1988 at 12:00 PM: an example.
Figure 6.17. Map of mudflat availability generated by the spatial model at Salt Creek on 9th
July 1993 at 7:00 AM: an example.
The CLLAMM Dynamic Habitat
148
Figure 6.18. Map of mudflat availability generated by the spatial model at Barker Knoll on 8th January
1988 at 9:00 AM: an example.
Mudflat availability at different water levels
At Barker Knoll, the maximum mudflat habitat (around 3.2 ha) became available at a
mean water level of -0.066 mAHD with around 2.4 ha located on the eastern shore.
The mudflat declined sharply beyond 0.90 mAHD and did not occur above 1.24
mAHD, as the edge of the shores had been reached (Figure 6.19). When the mean
water level was 0.99 mAHD at Noonameena, the mudflat area reached its maximum
(around 6.4 ha) with about 2.9 ha and 3.5 ha on the eastern and the western shores,
respectively. Below 0.8 mAHD, the mudflat areas did not change much (around 4.4
ha) and were almost equally distributed on both shores. The mudflat areas were
reduced dramatically with an increase in the mean water level above 1.0 mAHD and
no mudflats were found above 1.4 mAHD at Noonameena (Figure 20). At Salt Creek,
the maximum mudflat area of around 6.2 ha occurred at mean water level between
0.26 to 0.29 mAHD. The western shores offered less mudflat areas above 0.29
mAHD, while the mudflat habitat area increased on the eastern shore above 0.29
mAHD, except for water levels between 0.54 and 0.89 mAHD. Above 0.93 mAHD,
the mudflat area gained on the eastern shore was more than the loss of habitat on
the western shore resulting in a net increase in the mudflat areas above 0.93 mAHD
at Salt Creek (Figure 6.21).
The CLLAMM Dynamic Habitat
149
Mudflat area (ha)
4.00
3.00
2.00
1.00
0.00
-0.30 -0.14 -0.07
0.00
0.09
0.18
0.20
0.26
0.31
0.34
0.56
0.72
0.87
1.17
Mean water level (mAHD)
Eastern shore
Western shore
Channel
Total
Figure 6.19. Mudflat availability on the eastern shore, western shore, channel and the total
areas at different water levels at Barker Knoll.
Mudflat area (ha)
8.00
6.00
4.00
2.00
0.00
-0.09 -0.05
0.00
0.01
0.02
0.28
0.35
0.47
0.54
0.69
0.95
1.00
1.10
1.28
Mean water level (mAHD)
Eastern shore
Western shore
Total
Figure 6.20. Mudflat availability on the eastern shore, western shore and the total areas at
different water levels at Noonameena.
Mudflat area (ha)
7.00
5.00
3.00
1.00
-1.00
-0.15 -0.15 -0.11 -0.10
0.26
0.27
0.28
0.41
0.41
0.52
0.84
0.93
1.03
1.08
Mean water level (mAHD)
Eastern shore
Western shore
Total
Figure 6.21. Mudflat availability on the eastern shore, western shore and the total areas at
different water levels at Salt Creek.
The CLLAMM Dynamic Habitat
150
Spatial and temporal variation of mudflat area
The mean water levels and corresponding mudflat areas available between 6:00 to
20:00 are presented in Figures 6.22 to 6.29 for the day on which the average water
level reached the minimum or maximum in each of January and July of the two dry
and wet years at the selected reference sites. The availability of mudflat on the
eastern and western shores, and also in the channel in the case of Barker Knoll, for
these water levels is illustrated in Appendix 6.5.
Barker Knoll, being closest to the Murray Mouth, experienced high variability in water
levels during the day in both summer and winter months, which greatly affected
mudflat availability. On the day the water level reached its maximum in January for
the two wet years (22nd January 1976), the water level at Barker Knoll followed a
typical diurnal pattern due to tidal influence. The water level dropped initially to 0.19
mAHD and then rose to 0.30 mAHD at 13:00. In the afternoon, the water level
reduced gradually to the lowest level of 0.05 mAHD at 17:00 and again increased to
0.21 mAHD. The mudflat availability did not follow exactly the same pattern and
variability. However, the mudflat areas generally expanded with reducing water
levels. Thus, the maximum areas of mudflat (around 2.5 ha) were observed at the
mean water level, 0.05 mAHD in the late afternoon (at 17:00), and the minimum
areas (around 1.89 ha) occurred in the morning (at 7:00) at 0.40 mAHD.
Although the water level dropped gradually from around 0.8 to 0.62 mAHD,
throughout the day, there was not much change in mudflat area (around 4.1 to 4.4
ha) at Noonameena. On the same day, Salt Creek had the highest mudflat area of
around 6.0 ha while the mean water level was almost static at 0.41 mAHD (Figure
6.22).
On the day the water level reached its minimum in January in the wet years (31st
January 1976), the water level at Barker Knoll decreased from -0.017 mAHD at 6:00
to the lowest level of -0.30 mAHD at 11:00 and gradually increased to 0.17 mAHD at
20:00. The mudflat area increased to around 3.2 ha at -0.066 mAHD in the late
afternoon (at 17:00) at Barker Knoll (Figure 6.23). At Noonameena, the water level
slightly reduced from -0.02 mAHD at 6:00 to -0.09 mAHD at 20:00 while the mudflat
area increased by 0.2 ha over the period from 4.3 ha to 4.5 ha. The water level was
almost static at around 0.27 mAHD throughout the day at Salt Creek, with an
available mudflat area of around 6.17 ha.
Mudflat area (ha)
0.8
6.00
0.6
4.00
0.4
2.00
0.00
6:00
0.2
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
1
8.00
0
20:00
Hours (22 Jan. 1976)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.22. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 22 Jan. 1976. The average water level was at its maximum for January in the two wet
years on this day.
The CLLAMM Dynamic Habitat
151
Mudflat area (ha)
0.4
6.00
0.2
4.00
0
2.00
-0.2
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
8.00
-0.4
20:00
Hours (31 Jan. 1976)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.23. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 31 Jan. 1976. The average water level was at its minimum for January in the two wet
years on this day.
The average water level reached a July maximum for the two wet years on 21st July
1976 in the Lagoon. The water level increased gradually from 0.56 mAHD at 6:00 to
1.24 mAHD at 15:00 and subsequently dropped to 0.84 mAHD at 20:00 at Barker
Knoll. Although there was a continuous increase in water level, the mudflat area
peaked at 2.27 ha at a water level of 0.59 mAHD (at 8:00). Further rises in the water
level above 0.65 mAHD after 9:00 reduced the mudflat habitat gradually up to a
water level of 0.90 mAHD. Water levels above 1.17 mAHD covered almost all the
mudflat, and no mudflat became available between 14:00 to 16:00. As the water
receded, the water level dropped below 0.90 mAHD, thus resulting in about 2.0 ha of
mudflat at Barker Knoll (Figure 6.24). At Noonameena, the water level stayed above
1.1 mAHD throughout the day. At the minimum water level (1.1 mAHD) between 8:00
to 11:00, the maximum mudflat area available was around 0.86 ha. At water levels
above 1.19 mAHD, all mudflats were submerged under greater than 12 cm of water
and not accessible for foraging by the majority of wading birds for the rest of the day.
At Salt Creek, the water level increased by around 0.60 mAHD in July 1976
compared to January 1976, to around 1.0 mAHD and fluctuated very little for the
whole day. However, there was a slight change in mudflat availability ranging
between 5.3 and 5.8 ha.
The minimum July water level for the two wet years occurred on 9th July 1993. The
water levels were steady at Noonameena and Salt Creek, but varied at Barker Knoll.
At Barker Knoll, the water level was -0.056 at 6:00 and dropped to -0.12 at 8:00
before rising to 0.37 mAHD at 16:00, and then decreasing to 0.012 mAHD at 20:00.
The availability of mudflat was a maximum at around 3.2 ha at a water level of -0.056
mAHD and slightly decreased with decreasing water level. Water levels above -0.056
mAHD also decreased mudflat availability and reached a minimum of 1.95 ha at the
maximum water level of 0.37 mAHD at 16:00. At Noonameena, the water level
fluctuated between -0.009 and 0.082 mAHD, with only a slight impact on mudflat
availability (4.33 and 4.36 ha). The water level at Salt Creek was around 0.28 mAHD,
about 0.7 mAHD below the water level on 21st July 1976. However, mudflat
availability increased by about 0.8 ha and reached around 6.16 ha (Figure 6.25).
The CLLAMM Dynamic Habitat
152
Mudflat area (ha)
1.6
6.00
1.2
4.00
0.8
2.00
0.4
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
8.00
0
20:00
Hours (21 Jul. 1976)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.24. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 21 Jul. 1976. The average water level was at its maximum for July in the two wet
years on this day.
0.4
6.00
0.2
4.00
0
2.00
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
Mudflat area (ha)
8.00
-0.2
20:00
Hours (9 Jul. 1993)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.25. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 9 Jul. 1993. The average water level was at its minimum for July in the two wet years
on this day.
In January of the dry years, the Lagoon experienced the maximum and minimum
water levels on the 17th and 8th January 1988, respectively. On the 17th, at Barker
Knoll there was slight reduction in the water level in the morning from 0.31 mAHD at
6:00 to 0.17 mAHD at 14:00 and it then gradually rose to a maximum of around 0.36
mAHD at 20:00. Mudflat availability was a maximum of 2.1 ha at the minimum water
level of 0.17 mAHD and a minimum of 1.96 ha at the maximum water level of 0.36
mAHD. At Noonameena, the water level reached a maximum of 0.39 mAHD at 7:00
and gradually dropped to minimum of 0.28 mAHD at 20:00. Similar to Barker Knoll,
the minimum and maximum mudflats of 4.12 ha and 4.09 ha were available at
maximum (0.39 mAHD) and minimum (0.28 mAHD) water levels, respectively. At Salt
Creek, the water level remained at 0.15 mAHD and did not change throughout the
day. Likewise, the mudflat area was stable at around 3.15 ha (Figure 6.26).
The CLLAMM Dynamic Habitat
153
When the water level was at a minimum on 8th January 1988, it progressively
decreased from a maximum of 0.163 mAHD at 6:00 to the local minimum of -0.09
mAHD at 11:00 and rose to 0.029 mAHD at 15:00 before dropping to the lowest level
of -0.17 mAHD at 19:00 at Barker Knoll. Mudflat availability was 2.11 and 2.68 ha at
the maximum and minimum water levels, respectively. However, the maximum
mudflat availability of 3.19 ha was reached at water levels around -0.05 mAHD at
midday (12:00). At Noonameena, water levels varied between -0.02 mAHD at 6:00
and 0.02 mAHD at 18:00. These water level changes had very little influence on the
mudflat availability, which ranged between 4.34 ha at the maximum water level and
4.36 ha at the minimum water level at 6:00. Water levels ranged between -0.10
mAHD and -0.13 mAHD at Salt Creek with the maximum mudflat availability of 3.53
ha at the lowest water level of -0.10 mAHD in the late morning to afternoon between
11:00 and 16:00 (Figure 6.27).
0.6
0.4
4.00
0.2
2.00
0
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
Mudflat area (ha)
6.00
-0.2
20:00
Hours (17 Jan. 1988)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.26. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 17 Jan. 1988. The average water level was maximum for January in the two dry
years on this day.
Mudflat area (ha)
0.2
0.1
4.00
0
2.00
-0.1
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
6.00
-0.2
20:00
Hours (8 Jan. 1988)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.27. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 8 Jan. 1988. The average water level was minimum for January in the two dry years
on this day.
The CLLAMM Dynamic Habitat
154
In July of the dry years, the Lagoon had maximum and minimum average water
levels on the 22nd and 8th July 1988, respectively. On 22nd July, at Barker Knoll, there
was a slight reduction in the water level in the morning from 0.65 mAHD at 6:00 to
0.63 mAHD at 8:00 which gradually rose to maximum water level around 0.92 mAHD
at 16:00. The mudflat availability fluctuated between 2.05 ha and 2.26 ha with the
changing water level. The maximum available mudflat area of 2.26 ha occurred at the
minimum water level around 0.63 mAHD in the morning. At Noonameena, the
maximum water level of 1.04 mAHD was reached in the morning at 7:00 and it
gradually dropped to 0.95 mAHD at 16:00 and did not change thereafter. The
available mudflats at the maximum and minimum water levels were 4.34 ha and 5.85
ha, respectively. However, the maximum mudflat area of 6.39 ha was available at a
water level around 0.99 mAHD at noon (12:00). At Salt Creek, the water level was a
maximum 0.99 mAHD at 7:00 and progressively dropped to around 0.84 mAHD at
20:00 on the day. The mudflat area was a minimum of 5.27 ha at the maximum water
level and a maximum of 5.35 ha at the minimum water level (Figure 6.28).
8.00
1.2
6.00
0.9
4.00
0.6
2.00
0.3
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
Mudflat area (ha)
On 8th July 1988, the water level was at the minimum. At Barker Knoll, the water level
only changed slightly during the day between 0.31 mAHD (10:00) and 0.38 mAHD
(6:00). The mudflat area reached a maximum of 2.034 ha at the minimum water level
and minimum mudflat area of 1.92 ha occurred at the maximum water level. At
Noonameena, the water level gradually increased from 0.45 mAHD at 6:00 to 0.54
mAHD at 19:00. However, there was little variation in mudflat availability, which only
ranged between 4.07 ha and 4.09 ha. At Salt Creek, the water level ranged between
0.49 mAHD at 6:00 and 0.54 mAHD at 16:00. Unlike Barker Knoll and Noonameena,
the maximum mudflat area of 5.82 ha coincided with the maximum water level of
0.54 mAHD (Figure 6.29).
0
20:00
Hours (22 Jul. 1988)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.28. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 22 Jul. 1988. The average water level was at its maximum for July in the two dry
years on this day.
The CLLAMM Dynamic Habitat
155
0.6
6.00
0.4
4.00
0.2
2.00
0.00
6:00
8:00
10:00
12:00
14:00
16:00
18:00
Water level (mAHD)
Mudflat area (ha)
8.00
0
20:00
Hours (8 Jul. 1988)
Mudflat - Barker Knoll
Mudflat - Noonameena
Mudflat - Salt Creek
Water Level - Barker Knoll
Water Level - Noonameena
Water Level - Salt Creek
Figure 6.29. Mudflat availability and water levels between 6:00 - 20:00 at three reference
sites on 8 Jul. 1988. The average water level was at its minimum for July in the two dry years
on this day.
6.4.3 Modelling fish habitats using logistic regression
The key fish species of the Lagoon were modelled against salinity as the only
predictor variable using logistic regression. A summary of the model parameters for
all seven key species is given in Table 6.1. Salinity was only a significant predictor of
the presence of four species: Yelloweye Mullet, Smallmouth Hardyhead, Greenback
Founder and Tamar River Goby. Salinity was not a significant predictor for the other
three species, Black Bream, Congolii and Mulloway, although this could be due to the
rare occurrence of these species in the data. In the entire 94 sampling attempts,
each of these species were reported less than seven times.
The CLLAMM Dynamic Habitat
156
Table 6.1. Model coefficients and parameters for seven key species in the Coorong.
Variables
β
S.E.
df
Sig.
-0.135
0.038
1
0.000
9.854
2.624
1
0.000
0.054
.025
1
0.035
-1.226
1.078
1
0.255
-0.078
0.019
1
0.000
4.829
1.055
1
0.000
-0.081
0.033
1
0.013
2.906
1.342
1
0.030
-0.051
0.039
1
0.194
-.172
1.593
1
0.914
Salinity
-0.027
.021
1
0.213
Constant
-0.789
1.022
1
0.440
-0.083
.054
1
0.124
1.022
2.018
1
0.613
Yelloweye Mullet
Salinity
Constant
Smallmouth Hardyhead
Salinity
Constant
Greenback Founder
Salinity
Constant
Tamar River Goby
Salinity
Constant
Black Bream
Salinity
Constant
Congolii
Mulloway
Salinity
Constant
β = Estimated coefficient; S.E. = Standard Error of estimates; df = degree of freedom; Sig. =
Significance value and values > 0.05 are bold.
The prediction of presence and absence cases at the 0.5 cut off value for the
selected (training) dataset for the four species is summarised in Table 6.2. The
overall prediction accuracy was found to be above 70% for all four species. The
model for Yelloweye Mullet accurately classified 46 out of 47 presences and 17 out of
18 absences, and had the highest overall prediction accuracy of 96.7 %. For
Smallmouth Hardyhead, the model correctly classified 50 out of 51 presences.
However, the overall prediction accuracy of the model was only 76.9% as the model
failed to classify any of the absence data (14) accurately. The model for Greenback
Flounder correctly classified 37 out of 40 presences, but only 8 of the 25 absences,
again suggesting a patchy distribution or that an important constraining variable is
absent from the model. Although the model for Tamar River Goby had a prediction
accuracy of 72.3%, the model failed to classify 13 out of 20 presences accurately.
The CLLAMM Dynamic Habitat
157
Table 6.2. Classification summary for the selected dataset by the models.
Yelloweye Mullet
1
47
0
18
Classification by the
model
1-1
0-0
1-0
0-1
46
17
1
1
Smallmouth Hardyhead
51
14
50
0
1
14
76.9
Greenback Flounder
40
25
37
17
3
8
83.1
Tamar River Goby
20
45
7
40
13
5
72.3
Fish species
Original data
Overall prediction
accuracy %
96.9
The cut-off value was 0.50. 1 = species presence; 0 = species absence;1-1= true presence;
0-0= true absence; 1-0 = False-negative and 0-1 = False-positive.
6.4.4 Validation of the models
The accuracy of the models was assessed by using the component of the data set
aside for validation. The models predicted occurrence with > 96% success for
Yelloweye Mullet, > 89% for Greenback Flounder, and > 79% for Smallmouth
Hardyhead and Tamar River Goby. The slightly lower prediction accuracies of the
models for Smallmouth Hardyhead and Tamar River Goby were due to the inability of
the model to classify 6 absences for the former and 6 presences for the latter. The
classification summary and prediction accuracies of the models for the four species
are shown in Table 6.3.
Table 6.3. Classification summary for the unselected (validation) dataset by the models.
Yelloweye Mullet
Original
data
1
0
20
9
Classification by the
model
1-1
0-0
1-0
0-1
19
9
1
0
Smallmouth Hardyhead
23
6
23
0
0
6
79.3
Greenback Flounder
17
12
17
9
0
3
89.7
Tamar River Goby
6
23
0
23
6
0
79.3
Fish species
Overall prediction
accuracy %
96.6
The cut-off value was 0.50. 1 = species presence; 0 = species absence; 1-0 = False-negative
and 0-1 = False-positive.
The ROC curves, which illustrate the accuracy of models, are given in Figure 6.30.
The model for Yelloweye Mullet had the highest area (0.955) followed by Greenback
Flounder with 0.867, and Smallmouth Hardyhead and Tamar River Goby, both with
0.741. These values correspond to 95.5%, 86.7% and 74.1% accuracy, respectively
(Table 6.4).
The CLLAMM Dynamic Habitat
158
(a). Yelloweye Mullet
(b). Smallmouth Hardyhead
(c). Greenback Flounder
(d). Tamar River Goby
Figure 6.30. Receiver Operator Characteristics (ROC) curve for four species.
Table 6.4. Statistical parameters for Receiver Operator Characteristic (ROC) curves for the
logistic models of four fish species.
Species
Yelloweye Mullet
Area Std. Errora
Asymptotic
Sig.b
Asymptotic 95% Confidence
Interval
Lower Bound Upper Bound
0.955
0.031
0.000
0.895
1.016
Smallmouth Hardyhead 0.741
0.050
0.001
0.644
0.838
Greenback Flounder
0.867
0.042
0.000
0.786
0.949
Tamar River Goby
0.741
0.052
0.000
0.640
0.843
6.4.5 Habitat prediction for the key fish species
The predicted occurrence along the Coorong of the four key fish species significantly
related to salinity is presented in Figures 6.31 to 6.34. In July 1976, the Lagoon had
an estuarine condition (with salinity < 35 g/L) up to Parnka Point and was marine
(salinity between 35 to 47 g/L) in the South Lagoon (Appendix F). The models
predicted a very high probability (> 75%) occurrence of Yelloweye Mullet and
Greenback Flounder in the entire Lagoon (Figures 6.31 and 6.32). Smallmouth
Hardyhead had a very high probability of occurrence in the South Lagoon between
The CLLAMM Dynamic Habitat
159
Jack Point and Salt Creek and a high probability to the south of Noonameena down
to Jack Point. The species had a moderate probability of occurrence in the North
Lagoon, except around Ewe Island (Figure 6.33). Tamar River Goby had a very high
probability of occurrence up to midway between Noonameena and Parnka Point, a
high probability (50 – 75%) up to Villa dei Yumpa and a moderate probability (25 –
50%) to the south of Villa dei Yumpa (Figure 6.34).
In July 1988, the salinity levels were elevated by about three times in the North
Lagoon and two and half times in the South Lagoon relative to the salinity level in
July 1976 (Appendix 6.6). The salinity ranged between 5 g/L around the Murray
Mouth and 89 g/L at Salt Creek in the South Lagoon. High salinity levels impacted on
the occurrence probability of the key fish species along the Lagoon. Very high
occurrence probabilities of Yelloweye Mullet and Greenback Founder were restricted
to north of Villa dei Yumpa and Parnka Point in the South Lagoon, respectively, with
a low probability of occurrence to the south of Jack Point due to high salinity levels.
Tamar River Goby was also impacted negatively, with a very high probability of
occurrence limited to north of Long Point. In contrast, the rising salinity level favoured
Smallmouth Hardyhead and the area with a very high probability included the entire
North and South Lagoon.
Salinity rose to extremely high levels in January 2005 with a minimum of 42 g/L
around the Murray Mouth and a maximum of 124 g/L at Salt Creek in the South
Lagoon (Appendix 6.6). High salinity levels in the North Lagoon favoured a very high
probability of occurrence for Smallmouth Hardyhead throughout the entire Coorong
except for the area around the Murray Mouth, which had a high probability. Tamar
River Goby was restricted to a moderate probability of occurrence in the North
Lagoon, and low south of Pelican Point. Yelloweye Mullet and Greenback Founder
both had a low probability of occurrence from about 10 kilometres south of
Noonameena including the South Lagoon. However, Yelloweye Mullet had a very
high probability of occurrence north of Noonameena whereas Greenback Flounder
had a very high occurrence probability between Goolwa Channel and Ewe Island and
a high occurrence probability to the south of Long Point. Table 6.5 summarises the
probability of occurrence for the key fish species and their relationship with salinity.
The CLLAMM Dynamic Habitat
160
Figure 6.31. Habitat prediction for Yelloweye Mullet in July 1976, July 1988 and January 2005. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker
Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei Yumpa, JP = Jack
Point and SC = Salt Creek.
The CLLAMM Dynamic Habitat
161
Figure 6.32. Habitat prediction for Greenback Flounder in July 1976, July 1988 and January 2005. GC = Goolwa Channel; MC = Mundoo Channel; BK =
Barker Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei Yumpa, JP =
Jack Point and SC = Salt Creek.
The CLLAMM Dynamic Habitat
162
Figure 6.33. Habitat prediction for Smallmouth Hardyhead in July 1976, July 1988 and January 2005. GC = Goolwa Channel; MC = Mundoo Channel; BK =
Barker Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei Yumpa, JP =
Jack Point and SC = Salt Creek.
The CLLAMM Dynamic Habitat
163
Figure 6.34. Habitat prediction for Tamar River Goby in July 1976, July 1988 and January 2005. GC = Goolwa Channel; MC = Mundoo Channel; BK = Barker
Knoll; EI = Ewe Island; PP = Pelican Point; MP = Mark Point; LP = Long Point; NM = Noonameena; PA = Parnka Point; VY = Villa dei Yumpa, JP = Jack Point
and SC = Salt Creek.
The CLLAMM Dynamic Habitat
164
Table 6.5. Salinity level and occurrence probability for the four key fish species.
Key Fish Species
Yelloweye Mullet
Smallmouth Hardyhead
Greenback Founder
Tamar River Goby
6.5.
Occurrence Probability/ Salinity (g/L)
< 25%
> 80
> 75
> 49
25 – 50%
72 - 80
< 22
61 - 75
35 - 49
50 – 75%
64 - 72
22 - 43
47 - 61
22 - 35
> 75%
< 64
> 43
< 47
< 22
Discussion
6.6.1 Water level and salinity variations in the Coorong
Water level and salinity are the key ecological drivers in the Coorong, and influence
the overall distribution of biological communities and the state of the ecosystem as a
whole. Changes in water level influence the availability of mudflat habitats for
waterbirds (Rogers and Paton 2009a) and their main food source by impacting on
distribution of Ruppia tuberosa (Rogers and Paton 2009b) and macroinvertebrates.
Salinity has been recognized as the most critical factor for the ecological
sustainability of the Coorong, impacting directly on the aquatic biological
communities (fish, vegetation and macroinvertebrates) and indirectly on the
waterbirds through influencing their primary food source (R. tuberosa) (Rogers and
Paton 2009b; Rolston and Dittmann 2009).
Variations in water level are primarily attributed to the timing and volume of
freshwater input over the barrages, Upper South East Drainage (USED) flow, and the
opening of the Murray Mouth (Webster 2005, 2007). The seasonal variation in water
levels is mainly caused by rainfall and evaporation, while wind and tide have a higher
frequency temporal effect. In the summer, the Coorong and surrounding region
receive low rainfall (average <22 mm per month at Meningie) with very high
evaporation (average > 225 mm per month for the Lakes) compared to the winter
season rainfall (average >40 mm per month at Meningie) and evaporation (average <
60 mm per month for the Lakes). High rainfall with less evaporation in July is likely to
favour higher water levels than in January. On a shorter time-scale, the tide has
diurnal or semi-diurnal effects in areas around the Murray Mouth and up to 15 km
from the Mouth in the North Lagoon, whereas wind influences water levels both in the
North and South Lagoon (Webster 2007). Although the volume of the water inputs,
season (time), rainfall, evaporation, tide and wind all act together to determine water
level at any point in time, the seasonal variation in the water levels is due primarily to
rainfall and evaporation in no/low flow situations, resulting in high and low water
levels in winter and summer, respectively
The monthly barrage flow data used for the scenario modelling (Appendix 6.7)
showed large fluctuations in the quantity of water released into the Coorong in the
past decades. Among the four selected years, the predicted average water level was
higher in July than in January except for 1993. High water levels in January 1993
were attributed to almost five times more water volume (2199 GL) released into the
Coorong than in July. Although the amount of water had a large impact on the water
level, it was not possible to find a fixed relationship between the volume of water and
the water level in the Coorong. For example, the model predicted a higher water level
with about 73 GL released in July 2005 than in July 1993 with 433 GL of water
released. Therefore, other hydrodynamic factors present in the Lagoon such as water
flow out from the Lagoon, underground leakage, etc. could also play a significant role
in determining the water levels in the Coorong.
The water levels vary longitudinally in the Lagoon due to the main hydrological
drivers; water volume, timing, Murray Mouth opening and environmental variables
such as rainfall, precipitation and evaporation. The average water level in the South
The CLLAMM Dynamic Habitat
165
Lagoon was predicted to be as high as 8 cm above the water level in the North
Lagoon during barrage flows. However, in the summer without barrage flows, the
water level in the Lagoon is entirely influenced by sea level and the Murray Mouth
opening and drops below 0 mAHD, causing a significant reduction in the water flow
at Parnka Point and isolating the South Lagoon. Eventually, further loss of water due
to evaporation lowers the water level in the South Lagoon below that of the North
Lagoon (Webster 2007). The Coorong did not receive barrage flows in January 1988
or 2005, resulting in average water levels of around 0.12 mAHD in the North Lagoon
and around -0.10 mAHD in the South Lagoon.
Salinity levels in the Coorong are influenced by a number of factors including; salty
water inputs through the Murray Mouth and USED, their transport and mixing, and
evaporation. Freshwater inputs through the barrages tend to raise the water level,
compensating for evaporative water loss and eventually reducing salinity levels along
the Lagoon if flows are large enough (Webster 2007). For example, about 30 to 40
GL of waterflow over the barrages restored estuarine conditions in areas up to 15 km
from the Murray Mouth for about 20 days (Geddes 2005b).
During barrage flows in the four selected years, the salinity level was maintained
below ~ 20 g/L in the Murray Mouth and in some areas in the North Lagoon, and it
did not exceed 56 g/L in Salt Creek at the end of the South Lagoon. However, zero
flows in January 1988 and 2005 doubled salinity levels along the Lagoon, reaching
~40 g/L around the Murray Mouth and ~ 126 g/L at Salt creek.
6.6.2 Mudflat availability at the reference sites
For this analysis, we designated mudflats up to 12 cm water depth as suitable for
foraging by the majority of waterbird species (Rogers pers. comm., Wildlife Habitat
Management Institute 2000). Analysis of mudflat availability was performed for the
days with maximum and minimum water levels in both wet and dry years. Water level
ranges between -0.30 to 1.25 mAHD, -0.09 to 1.30 mAHD and -0.15 to 1.09 mAHD
were modelled for Barker Knoll, Noonameena and Salt Creek, respectively, although
there was not a simple relationship between mudflat availability and water level.
The availability of mudflat areas accessible to waterbirds is governed by the
underwater topography. For a given water level, relatively flat areas provide more
available foraging area than steeply sloping areas (see Chapter 5). A study of the
morphology of mudflats in the Coorong discussed the mudflat slopes (%) for all
references sites (see Chapter 5), however, the slopes of the eastern and western
shores were not differentiated. In the current study, mudflat areas are estimated as
the sum of all areas available on the eastern and western shores, and in the channel
in the case of Barker Knoll. Because of differences in the topography (slope), these
areas offer varying amounts of mudflat for specified water levels. At Barker Knoll, the
maximum area of mudflat (3.19 ha) was available at -0.05 mAHD with more than
75% of mudflat located on the eastern shore, implying the eastern shore is relatively
flat compared to the western shore. Similarly the maximum mudflat area (6.17 ha)
occurred at 0.27 mAHD at Salt Creek with more than 60% found on the eastern
shore. However, the western shore has a lesser slope at Noonameena and thus
contributed about 55% of the maximum mudflat area (6.38 ha). It should be noted
here that the hydrodynamic model only gives water level along the centreline of the
Lagoon, and we assume that this is representative of the entire width. In reality, wind
is likely to bank water along one shore, resulting in changing water levels and mudflat
availability across the Lagoon.
Temporal variation in mudflat area was evident at Barker Knoll due to diurnal or
semi-diurnal tidal influences on the water level, whereas Noonameena and Salt
Creek had very little change in mudflat area over the period of a day as water level
varied little at this time-scale.
The CLLAMM Dynamic Habitat
166
Although the topography of the Lagoon is a major factor for determining accessible
mudflat areas in the Coorong, the water level is the variable that managers can
manipulate to maximize mudflat availability. As the water levels vary longitudinally in
the Lagoon, it is not possible to maximize mudflat area at all locations at the same
time. Based on the three selected references sites, the highest average mudflat area
was observed at an average water level of ~0.12 mAHD on the day with minimum
water level in January 1976. The highest mudflat area of 6.17 ha occurred at Salt
Creek at 0.26 mAHD followed by Noonameena with 4.43 ha at -0.06 mAHD and the
lowest area of 2.64 ha at Barker Knoll at -0.11 mAHD.
6.6.3 Habitat modelling and prediction for key fish species
Salinty has been recognized as the most significant ecological variable for the
distribution of biological communities including key fish species in the Coorong (Noell
et al. 2009; Geddes 1987). The occurrence of the key fish species was modelled
against salinity using logistic regression. The model found a significant relationship
between salinity and the occurrence of four key species: Yelloweye Mullet,
Smallmouth Hardyhead, Greenback Flounder and Tamar River Goby in the Coorong.
However, the model for Smallmouth Hardyhead failed to classify any of the absent
data accurately, suggesting that an important environmental variable constraining the
presence of this species was absent from the model, or that distribution within
suitable habitat is patchy. Smallmouth Hardyhead was found to be the most salt
tolerant species, and was collected from the South Lagoon at 149‰ total dissolved
solids (TDS) (~ 130 g/L) in 1984 (Geddes 1987) and up to 133.5 g/L in December
2006 (Noell et al. 2009). According to the species-salinity relationship established by
the model, this species is highly likely to occur above 43 g/L to the maximum salinity
level (125 g/L) predicted for January 2005. In contrast to Smallmouth Hardyhead,
other key species preferred low salinity. Geddes (1987) reported Yelloweye Mullet,
Black Bream and Congolii from the South Lagoon at salinities below 55‰TDS. For
Yelloweye Mullet, salinities below 64 g/L offer a high likelihood of occurrence in the
Coorong. Habitat modelling of Yelloweye Mullet using Non Parametric Multiplicative
Regression (NPMR) (McCune 2006) also demonstrated salinity as the major
predictor variable for this species along with water temperature, and it is likely to
occur with consistent probability of occurrence of > 90 % at salinity < 64 g/L (Sharma
et al. 2009). Greenback Flounder was reported in the list of commonly found fish
species in the North Lagoon during 1984, when the salinity ranged from 5 to 65 g/L
(Geddes 1987). However, the model found Greenback Flounder to be less salinity
tolerant than Yelloweye Mullet and predicted a high likelihood of occurrence at
salinities below 47 g/L.
Another key fish species of the Coorong, Tamar River Goby, which has an estuarine
life cycle (Noell et al. 2009), was not reported in the study by Geddes (1987).
However, it was collected in 26 samples out of 94 during sampling undertaken in
2006-08. This species had the lowest salinity tolerance of the four key species and is
only highly likely to occur at salinities below 22 g/L.
The prediction maps for these key species were generated by applying the
regression coefficients in a model in a GIS platform. The probability of occurrence
was estimated based on the species-salinity relationship. The model consistently
made accurate predictions of where species would occur, although it was slightly less
accurate at predicting where they would be absent. Salinities between < 10 to 50 g/L
favoured the low salinity tolerant species; Tamar River Goby, Greenback Flounder
and Yelloweye Mullet. However, Smallmouth Hardyhead had a low probability (< 25
%) of occurrence in the North Lagoon where salinity was below 30 g/L. In January
2005, habitat suitability for Smallmouth Hardyhead increased, with an elevated
salinity range between 40 to 125 g/L, at the expense of other key species. As salinity
in the North Lagoon increased above 40 g/L, Greenback Flounder was restricted to
the north of Pelican Point in the North Lagoon while Tamar River Goby had a low
The CLLAMM Dynamic Habitat
167
probability of occurrence (< 25 %) in most of the Lagoon except to the north of
Pelican Point where probability of occurrence ranged between 25 to 50 %. However,
salinity in July 1988 ranged around 5 to 90 g/L and offered suitable habitat for low
salinity tolerant key species (Tamar River Goby) to the north of Long Point and
Greenback Flounder had high probability to the north of Parnka in the North Lagoon.
Yelloweye Mullet, with slightly higher salinity tolerance to Greenback Flounder, had a
high probability of occurrence north of Villa dei Yumpa, while the high salinity tolerant
species, Smallmouth Hardyhead, had a high probability of occurrence in the salinity
range 43 to 89 g/L, between 9 km north of Parnka Point in the North Lagoon to the
southern end of the South Lagoon.
6.6.4 Barrage outflow, water level and salinity in the Coorong
The quantity, frequency and duration of the freshwater outflow over the barrages
greatly influence salinity levels and the ecological health of the Coorong (Geddes
1987; 2003; 2005a; 2005b) while also influencing the water level in both lagoons
(Webster 2005, 2007). Managers are interested to know how much and how
frequently freshwater should be released into the Coorong to maintain the optimum
water level which maximizes mudflat habitat for the waterbirds and the best range of
salinity for supporting maximum diversity of biological communities. The analysis of
mudflat availability at three representative sites and the modelling of the key fish
species-salinity relationships suggests that an average water level of 0.12 mAHD is
required to secure maximum mudflat area whereas a salinity range of 5 to 90 g/L, as
occurred in January 1988, provides the most suitable salinity conditions for these key
species. At this time, the North Lagoon had an estuarine condition (< 30 g/L) around
the Murray Mouth south to Noonameena, and a marine condition (30 to 50 g/L) to the
north of Parnka Point. In the South Lagoon, the salinity ranged from hypersaline with
50 to 65 g/L to the north of Villa dei Yumpa, to highly hypersaline with 65 to 80 g/L to
the north of Jack Point and extremely hypersaline with 80 to 90 g/L to the south of
Jack Point. Maintaining an estuarine condition in the North Lagoon also offers a
suitable habitat for a wide range of micro-benthic species (Geddes 1987; Rolston and
Dittmann 2009) and the macrophyte Ruppia megacarpa (Geddes 1987). Ruppia
megacarpa is believed to have become extinct from the system in 1990s due to
elevated salinities in the Coorong (Nicol 2005). A salinity range of 50 to 90 g/L in the
South Lagoon provides suitable habitat for high salinity tolerant fish species
(Smallmouth Hardyhead), macro-invertebrates such as Capitella capitata,
Australonereis spp., Simplisetia and insect larvae (Rolston and Dittmann 2009), and
the macrophyte Ruppia tuberosa (Paton 2005).
An investigation of the barrage outflows used in the hydrodynamic modelling for
predicting water level and salinity was not helpful for prescribing an appropriate
volume and frequency for water inputs into the Coorong. However, Geddes (1987)
noted that volumes above 1000 GL released into the Coorong for three consecutive
months between August to October 1983 eventually reduced the salinity range from
40 to 130‰TDS to 25 to 60‰TDS in the North Lagoon and about 70‰TDS in the
South Lagoon. The measured salinity range of ~ 40 to 105 g/L in the Coorong in
October 2008 was similar to the salinity condition prior to barrage outflows in 1983.
This suggests that initially we require a similar volume of water for a few months to
substantially reduce the salinity in the system, although this is unimaginable under
current conditions in the Murray Darling Basin. Once the salinity is in the appropriate
range, barrage outflows would be required to mitigate the impact on salinity of sea
water incursion and to compensate for the evaporative loss of water, particularly in
summer months. Management of the Coorong is a very complex issue and requires
further research and modelling taking into account the current situation in the Lagoon
and the dynamics of environmental as well as physical factors like opening of the
Murray Mouth, sea water incursions, etc.
The CLLAMM Dynamic Habitat
168
6.6.
Summary, Conclusions & Management Implications
A spatial approach to modelling mudflat availability and key fish species habitats in
the Coorong is used to understand and visualise the spatio-temporal variations in
these habitats for a given hydrodynamic situation. The spatial models developed as
part of this study run continuously for several series of simulated water level and
salinity data and subsequently maps are generated for each scenario.
The mudflat habitat was taken to constitute the mudflat area up to 12 cm depth below
water level, representing accessible foraging ground for most species of waterbirds,
and having a high abundance of macro-invertebrates under moderate salinities. The
areas close to the Murray Mouth (Barker Knoll) are often influenced by diurnal or
semi-diurnal tidal effects resulting in short-term temporal variation as well as frequent
inundation of mudflats, supporting a high diversity and density of macroinvertebrates. However, the temporal variation at the other two sites at Noonameena
and Salt Creek was not as evident as it was at Barker Knoll unless there was a
significant change in the water level due to strong wind or water released through the
barrages or the Upper South East Drainage scheme.
Along the Lagoon, the availability of mudflat is determined by the water level and the
underlying topography. Flatter areas tend to offer more mudflat at a given water level.
The eastern shore is flatter than the western shore in the Coorong. For this reason,
the eastern shore offered more mudflat than the western shore among the three
selected sites. The topography of the Lagoon is relatively stable over short periods of
time, apart from areas directly adjacent to the Murray Mouth. However, managers
could manipulate or control the water level through opening or shutting of barrages in
order to maximize availability of mudflat in the Coorong. An analysis of mudflat
availability at different water levels suggests that an average water level of 0.12
mAHD gives the maximum average mudflat area in the three reference sites.
The first spatial model generates maps for mudflat areas in the eastern and western
shores and the channel (if any) and also the exposed mudflat areas above the
waterline. The maps depict the mudflat areas up to 12 cm depth at 1 cm resolution.
However, waterbirds are specific in their prey and are specialized in their use of
mudflats under different water depths (Australian Online Coastal Information 2009).
The 1 cm depth resolution maps allow a detailed analysis of mudflats at different
depths, and would be very useful to understand the relationship between different
waterbirds and their requirement of mudflat at particular depths. Importantly,
although model results for only three selected sites are presented here, the model
can be run for any given flow scenario over any time period for any of the 12
reference sites along the Coorong. The model could also be modified in the future to
take into account the time since drying/wetting, which may be important for some
macrophytes and macroinvertebrates.
The second spatial model used species-specific relationships between the fish
distribution and salinity to predict the likelihood of occurrence of fish under different
salinity regimes. Out of seven key fish species, Yelloweye Mullet, Smallmouth
Hardyhead, Greenback Flounder and Tamar River Goby demonstrated a significant
relationship with salinity levels. Among the three different salinity gradients
examined, the salinity range from 5 to 90 g/L along the Lagoon was found to give the
greatest representation of these four key species, and is also known to support other
important biological communities including both macrophytes and infauna.
Although this study was able to suggest appropriate water levels and salinity
gradients to maximize mudflat areas and biological communities in the Coorong, we
were not able to define the volume and frequency of water to be released through the
barrages or Upper South East drainage scheme. This requires further research and
hydrodynamic analysis considering all physical and hydrological factors.
The CLLAMM Dynamic Habitat
169
It is important for managers to understand the influence of both water level and
salinity on mudflat habitats as well as the aquatic habitats for fish, macrophyte and
infauna. The spatial models developed for this study allow managers to readily
quantify these habitats for specified flow scenarios, and support informed decisions
on the amount and frequency of barrage outflows once the Lower Lakes are
recharged and excess water is available for the Coorong.
6.7.
References
Aber, J.S. (2007) Wetland definitions and classification. Available at
http://academic.emporia.edu/aberjame/wetland/define/define.htm [accessed
14/07/2008].
Álvarez-Arbesú, R and Felicísimo, A.M. (2002) GIS and logistic regression as tools
for environmental management: a coastal cliff vegetation model in Northern Spain.
In: Management Information Systems 2002: GIS and Remote Sensing. (Ed. CAaP
Brebbia, P.) WIT Press, Southampton.
Australian Online Coastal Information (2009) Shorebird counts. Available at
http://www.ozcoasts.org.au/indicators/shorebird_counts.jsp [accessed 0/03/2009].
Carter, G.M., Stolen, E.D. and Breininger, D.R. (2006) A rapid approach to modeling
species-habitat relationship. Biological Conservation 127: 237-44.
Chiew, F.H.S., Teng, J., Kirono, D., Frost, A. J., Bathols, J.M., Vaze, J., Viney, N.R.,
Young, W. J., Hennessy, K.J. and Cai, W.J. (2008) Climate data for hydrologic
scenario modelling across the Murray-Darling Basin. A report to the Australian
Government from the CSIRO, Murray-Darling Basin Sustainable Yields Project.
CSIRO, Canberra.
CLLAMM (2008) Response of the Coorong ecosystems to alternative Murray-Darling
flow scenarios. CLLAMMecology Research Cluster, Adelaide.
De Smith, M.J., Goodchild, M.F. and Longley, P.A. (2006) Geospatial Analysis: a
comprehensive guide to principles techniques and software tools. Matador, Leicester.
Environmental Systems Research Institute (2001) ArcGIS Geostatistical Analysis:
Powerful exploration and data interpolation solutions. ESRI, Redlands, California.
Environmental Systems Research Institute (2006) Advanced modelling through
simulations.
http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=An_overview_of_a
dvance_modeling_through_simulations.
Environmental Systems Research Institute (2009) ArcGIS 9.3 Desktop. ESRI,
Redlands, California.
Edyvane, K.S. (1999) Conserving Marine Biodiversity in South Australia, Part 2Identification of areas of high conservation value in South Australia. South Australia
Research and Development Institute, Aquatic Sciences, Adelaide.
Geddes, M.C. (1987) Changes in salinity and in the distribution of macrophytes,
macrobenthos and fish in the Coorong Lagoons, South Australia, following a period
of River Murray flow. Transactions of the Royal Society of South Australia 111: 173181.
The CLLAMM Dynamic Habitat
170
Geddes, M.C. (2003) Survey to investgate the ecological health of the North and
South Lagoons of the Coorong, June/July 2003. SARDI Aquatic Sciences and
University of Adelaide, Adelaide.
Geddes, M.C. (2005a) The ecological health of the north and south lagoons of the
Coorong in July 2004. SARDI Aquatic Sciences and University of Adelaide, Adelaide.
Geddes, M.C. (2005b). Ecological outcomes from the samll barrage outlfow of
August 2004. Final report. Prepared for the Department of Water, Land and
Biodiversity Conservation. South Australia Research and Development Institute
(Aquaric Sciences), Adelaide.
Gibson, L.A., Wilson, B.A., Cahill, D.M. and Hill, J. (2004) Spatial prediction of rufous
bristlebird habitat in a coastal heathland: a GIS-based approach. Journal of Applied
Ecology 41: 213-23.
Gönen, M. (2006) Receiver Operating Characeristic (ROC) curves. In the proceeding
of SUGI 31, San Francisco.
Gross, J.E., Kneeland, M.C., Reed, D.F. and Reich, R.M. (2002) GIS-based habitat
models for mountain goats. Journal of Mammalogy 83(1): 218-28.
Hosmer, D.W., Lemeshow, S., Hosmer, D.W., and Lemeshow, S. (1989) Applied
Logistic Regression. John Wiley & Sons Inc., New York.
Lamontagne, S., McEwan, K., Webster, I., Ford, P., Leaney, F. and Walker, G.
(2004) Coorong, Lower Lakes and Murray Mouth. Knowledge gaps and knowledge
needs for delivering better ecological outcomes. Water for a Healthy Country
National Research Flagship CSIRO, Canberra.
Mathew, J, Jha, V.K. and Rawat, G.S. (2007) Application of binary logistic regression
analysis and its validation for landslide susceptibility mapping in part of Garhwal
Himalaya, India, International Journal of Remote Sensing 28(10): 2257-75.
McCune, B. (2006) Non-parameric habitat models with automatic interactions.
Journal of Vegetation Science 17: 819-30.
Miles, M. (2006) Origin of bathymetric data for Lower Lakes, Goolwa and Coorong.
South Australian Department of Environment and Heritage, Adelaide.
Nicol, J. (2005) The ecology of Ruppia spp. in South Australia, with reference to the
Coorong: A literature review. SARDI, Aquatic Sciences, Adelaide.
Noell, C., Ye, Q., Short, D.A., Bucarter, L.B. and Wellman, N.R. (2009) Fish
assemblages of the Murray Mouth and Coorong region, South Australia, during an
extended drought period. CSIRO: Water for a Healthy Country National Research
Flagship and South Australian Research and Development Institute (Aquatic
Sciences), Adelaide.
Paton, D.C. (2005) 2005 winter monitoring of the southern Coorong. Project Report
to South Australian Department for Environment & Heritage, University of Adelaide,
Adelaide.
Rogers, D.J. and Paton, D.C. (2009a) Spatiotemporal Variation in the Waterbird
Communities of the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship.
The CLLAMM Dynamic Habitat
171
Rogers, D.J. and Paton, D.C. (2009b) Changes in the distribution and abundance of
Ruppia tuberosa in the Coorong. CSIRO: Water for a Healthy Country National
Research Flagship.
Rolston, A.N. and Dittmann, S. (2009) CLLAMMecology Invertebrate Key Species
Project: The distribution and abundance of macrobenthic invertebrates in the Murray
Mouth and Coorong Lagoons. CSIRO: Water for a Healthy Country National
Research Flagship.
Seaman, R.L. (2003) Coorong and Lower Lakes habitat-mapping program. South
Australian Department for Environment and Heritage, Adelaide,
http://www.environment.sa.gov.au/biodiversity/pdfs/wetlands/coorong/coorong_lower
_lakes_habitat_report.pdf.
Sharma, SK, Tanner, J and Ye, Q (2009) Habitat model for Yelloweye Mullet
(Aldrichetta forsteri) in the Coorong using Non-Parametric Multiplicative Regression.
(unpublished article).
Shriner, S.A., Simons, T.R. and Farnsworth, G.L. (2002) A GIS-based habitat model
for Wood Thrush, Hylocichla mustelina, in Great Smoky Mountains National Park. In:
Predicting species occurrences: Issues of accuracy and scale. (Eds. MJ Scott, PJ
Heglund & MMLe al.).
Webster, I.T. (2005) An overview of the Hydrodynamics of the Coorong and Murray
Mouth. Water levels and salinity - key ecological drivers. CSIRO Land and Water,
Canberra.
Webster, I.T. (2007) Hydrodynamic modelling of the Coorong. Water for a Healthy
Country National Research Flagship, CSIRO, Canberra.
Wildlife Habitat Management Institute (2000) Shorebirds. Madison, ftp://ftpfc.sc.egov.usda.gov/WHMI/WEB/pdf/SHOREbirds1.pdf.
Wilson, R. J. (2001) Wader surveys in the Coorong, South Australia in January and
February 2001. The Stilt 40: 38-54
Xie, C., Huang, B., Claramunt, C. and Chandramouli, M. (2005) Spatial Logistic
Regression and GIS to Model Rural-Urban Land Conversion. Second International
Colloquium on the Behavioural Foundations of Integrated Land-use and
Transportation Models: Frameworks, Models and Applications, Toronto, Canada.
The CLLAMM Dynamic Habitat
172
6.8.
Appendices
Appendix 6.1: Script used for modelling mudflat habitat.
# bird_habitat.py
# Created on: Fri Feb 27 2009 10:24:35 AM
# (generated by ArcGIS/ModelBuilder)
# Usage: bird_habitat_finer <DEM_for_the_site>
<Specify_the_mean_water_level_for_the_site> <Site_boundary_to_high_water_mark_level>
<Change_the_mean_water_level_in_the_expression> <West_shore_line_mask>
<East_shore_line_mask> <Available_mudflat_area>
<Area_Exposed_between_high_water_mark_and_the_shoreline>
<Mudflat_area_in_eastern_shore> <Perimeter_of_the_eastern_shore>
<Mudflat_area_in_western_shore> <Perimeter_of_the_western_shore>
# --------------------------------------------------------------------------# Import system modules
import sys, string, os, arcgisscripting, math
# Create the Geoprocessor object
gp = arcgisscripting.create()
# Check out any necessary licenses
gp.CheckOutExtension("spatial")
# Load required toolboxes...
gp.AddToolbox("C:/Program Files/ArcGIS/ArcToolbox/Toolboxes/Spatial Analyst Tools.tbx")
gp.AddToolbox("C:/Program Files/ArcGIS/ArcToolbox/Toolboxes/Conversion Tools.tbx")
gp.AddToolbox("C:/Program Files/ArcGIS/ArcToolbox/Toolboxes/Data Management
Tools.tbx")
gp.AddToolbox("C:/Program Files/ArcGIS/ArcToolbox/Toolboxes/Analysis Tools.tbx")
# Set the Geoprocessing environment...
gp.outputCoordinateSystem =
"PROJCS['GDA_1994_MGA_Zone_54',GEOGCS['GCS_GDA_1994',DATUM['D_GDA_1994',
SPHEROID['GRS_1980',6378137.0,298.257222101]],PRIMEM['Greenwich',0.0],UNIT['Degre
e',0.0174532925199433]],PROJECTION['Transverse_Mercator'],PARAMETER['False_Eastin
g',500000.0],PARAMETER['False_Northing',10000000.0],PARAMETER['Central_Meridian',14
1.0],PARAMETER['Scale_Factor',0.9996],PARAMETER['Latitude_Of_Origin',0.0],UNIT['Meter
',1.0]]"
gp.extent =
"Q:\CLLAMM_PROJECTS\HABITAT_MODELLING\BIRD_HABITAT\\ewe_bdry.shp"
# Local variables...
water_level =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\water_level_data.dbf"
DEM_site =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\ewe_dem"
bdry_site =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\ewe_bdry.shp"
west_coast =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\west_coast.shp"
east_coast =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\east_coast.shp"
bdry_high_water_mark =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\bdry_high_water_mar
k.shp"
modelPath =
"Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\OUTPUT_BIRD_HABI
TAT\\"
# setting workspace
gp.workspace = modelPath
try:
# Get list of fields for looping
The CLLAMM Dynamic Habitat
173
fields1 = gp.ListFields
("Q:\\CLLAMM_PROJECTS\\HABITAT_MODELLING\\BIRD_HABITAT\\water_level_data.dbf"
, "T*")
fields1.reset()
# Get the first field and start loop
field1 = fields1.Next()
while field1:
wLevel = field1.Name
print "Water level in the list is: ", str(wLevel)
#output 1: A raster from water level data
wl_IDW = modelPath + wLevel + "_IDW"
#output 2: DEM for the subtidal_area extracted by using average water level value
wl_Site = modelPath + wLevel+ "_site"
#output 3: DEM for the subtidal_area extracted by using average water level value
wl_DEM = modelPath + wLevel+ "_dem"
#output 4: Generating mask for the subtidal area
wl_Mask = modelPath + wLevel+ "_mask"
#output 5: Total mudflat area in the site both up to 12 cm (0) and below 12 cm (1) depth
wl_totalArea = modelPath + wLevel+ "_totalArea"
#output 6: Available mudflat area in the site with less than 12 cm depth
wl_MF = modelPath + wLevel + "_MF"
#output 7: DEM for the Available mudflat area
wl_MFDEM = modelPath + wLevel+ "_MFDEM"
#output 8: Available mudflat area in the east coast
wl_MFEast = modelPath + wLevel +"_MFEast"
#output 9: Available mudflat area in the east coast
wl_DEPTH = modelPath + wLevel +"_DEPTH"
#output 10: Available mudflat area in the east coast
wl_DEPTHint = modelPath + wLevel +"_DEPTHint"
#output 11: Available mudflat area in the wast coast
wl_MFWest = modelPath + wLevel + "_MFWest"
#output 12: Available mudflat area in the wast coast
wl_MFExp = modelPath + wLevel + "_MFExp"
#output 13: Available mudflat area in the wast coast
wl_AreaExp1 = modelPath + wLevel + "_AreaExp1"
#output 14: Available mudflat area in the wast coast
wl_AreaExp = modelPath + wLevel + "_AreaExp"
#Generating raster from water level data
gp.Idw_sa(water_level, wLevel, wl_IDW, "1", "2", "VARIABLE 10", "")
gp.AddMessage("Successful" + gp.GetMessages())
print "Successful", gp.GetMessages()
#Process: Extract water level rater with the site mask
gp.ExtractByMask_sa(wl_IDW, DEM_site, wl_Site)
#Raster properties: Getting the mean water level from the interpolated water level raster
mean = str(gp.GetRasterProperties(wl_Site,"MEAN","0"))
print "Mean is: ", mean
InWhereClause = "value <= "+ mean +""
#Process: Extracting the bathymetry below the mean water level
gp.ExtractByAttributes_sa(DEM_site, InWhereClause, wl_DEM)
print gp.GetMessages()
#Generating a mask for the areas (subtidal) defined by the water level
InExpression1 = wl_Mask + " = ( " + wl_DEM +" - "+ wl_DEM +")"
gp.MultiOutputMapAlgebra_sa(InExpression1)
print gp.GetMessages()
#Applying 12 cm depth condition to the Mudflat bathymetry (wl_DEM) to areas
#below 12 cm depth as ture (1) and areas upto 12 cm depth as false (0).
InExpression2 = wl_totalArea+ " = ( " + wl_DEM + " <= ("+wl_Mask+" + "+ mean +" 0.12))"
gp.MultiOutputMapAlgebra_sa(InExpression2)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
# Process: Reclassify Mudflat area as 1 and the rest of area as Nodata
gp.Reclassify_sa(wl_totalArea, "VALUE", "0 1;0 1 NODATA", wl_MF, "DATA")
The CLLAMM Dynamic Habitat
174
# Process: Extracting bathymetry for the Mudflat area
gp.ExtractByMask_sa(DEM_site, wl_MF, wl_MFDEM)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
#Mudflat depth in cm from the water level down below derived from substracting
#the mean water level. It means water depth at water level is 0 cm.
InExpression3 = wl_DEPTH+ " = ( "+ mean +" - " + wl_MFDEM + " )"
gp.MultiOutputMapAlgebra_sa(InExpression3)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
#Process: Converting Mudflat bathymetry to integer value.
InExpression4 = wl_DEPTHint + " = Int(100 * " + wl_DEPTH + ")"
gp.MultiOutputMapAlgebra_sa(InExpression4)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
# Process: Extracting avialable Mudflat area in the eastern shore
gp.ExtractByMask_sa(wl_DEPTHint, east_coast, wl_MFEast)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
# Process: Extracting avialable Mudflat area in the western shore
gp.ExtractByMask_sa(wl_DEPTHint, west_coast, wl_MFWest)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
# Process: Reclassifying total area (areas below water level) as NODATA and
#the rest of the areas, previously NODATA to 1.
gp.Reclassify_sa(wl_totalArea, "VALUE", "0 NODATA;0 1 NODATA;NODATA 1",
wl_MFExp, "DATA")
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
# Process: Extracting areas exposed between the water level and the high water mark
level
gp.ExtractByMask_sa(wl_MFExp, bdry_high_water_mark, wl_AreaExp1)
gp.AddMessage("Successful" + gp.GetMessages())
print gp.GetMessages()
#Process: Cleanign the exposed area by using bathymetry dataset
gp.ExtractByMask_sa(wl_AreaExp1, DEM_site, wl_AreaExp)
#Delete intermediate layers
gp.AddMessage ("Cleaning up ......." + gp.GetMessages())
gp.delete(wl_IDW)
gp.delete(wl_Site)
gp.delete(wl_Mask)
gp.delete(wl_DEM)
gp.delete(wl_MFDEM)
gp.delete(wl_MF)
gp.delete(wl_totalArea)
gp.delete(wl_DEPTH)
gp.delete(wl_DEPTHint)
gp.delete(wl_MFExp)
gp.delete(wl_AreaExp1)
print gp.GetMessages()
# looping
field1 = fields1.Next()
except:
print gp.GetMessage(1)
The CLLAMM Dynamic Habitat
175
Appendix 6.2: Script used for modelling fish habitat.
# --------------------------------------------------------------------------# Created on: Tue Apr 08 2008 10:47:39 AM
# (generated by ArcGIS/ModelBuilder)
# --------------------------------------------------------------------------# Import system modules
import sys, string, os, arcgisscripting, math
# Create the Geoprocessor object
gp = arcgisscripting.create()
# Check out any necessary licenses
gp.CheckOutExtension("spatial")
# Load required toolboxes...
gp.AddToolbox("C:/Program Files/ArcGIS/ArcToolbox/Toolboxes/Spatial Analyst Tools.tbx")
# Set the Geoprocessing environment...
gp.outputCoordinateSystem =
"PROJCS['GDA_1994_MGA_Zone_54',GEOGCS['GCS_GDA_1994',DATUM['D_GDA_1994',
SPHEROID['GRS_1980',6378137.0,298.257222101]],PRIMEM['Greenwich',0.0],UNIT['Degre
e',0.0174532925199433]],PROJECTION['Transverse_Mercator'],PARAMETER['False_Eastin
g',500000.0],PARAMETER['False_Northing',10000000.0],PARAMETER['Central_Meridian',14
1.0],PARAMETER['Scale_Factor',0.9996],PARAMETER['Latitude_Of_Origin',0.0],UNIT['Meter
',1.0]]"
#gp.CellSize = "25"
# Local variables...
salinity_data = "Q:\\DATA_FOR_SCRIPT\\salinity4_test.dbf"
water_level = "Q:\\DATA_FOR_SCRIPT\\water_level4.dbf"
dem_10 = "Q:\\DATA_FOR_SCRIPT\\dem_10"
coorong_bdry = "Q:\\DATA_FOR_SCRIPT\\test10_bdry.shp"
#Output_variance_of_prediction_raster = ""
modelPath = "Q:\\DATA_FOR_SCRIPT\\"
# Setting workspace
gp.workspace = modelPath
try:
# Get list of fields for looping
fields1 = gp.ListFields ("Q:\\DATA_FOR_SCRIPT\\water_level4.dbf", "T*")
fields1.reset()
# Get the first field and start loop
field1 = fields1.Next()
fields2 = gp.ListFields ("Q:\\DATA_FOR_SCRIPT\\salinity4_test.dbf", "SAA*")
fields2.reset()
field2 = fields2.Next()
while field1:
The CLLAMM Dynamic Habitat
176
wLevel = field1.Name
print "Water level in the list is: ", str(wLevel)
# output
wl_output = modelPath + wLevel + "_IDW33"
wl_dem = modelPath + wLevel + "_dem"
# provide a default value if unspecified
InExpression1 = wl_dem + " = (" + dem_10 + " < " + wl_output + ")"
#gp.AddMessage("Working on + wLevel; interpolating by using IDW ....") +
gp.GetMessage())
#print "IDW failed ", gp.GetMessages()
# Process: IDW...
#print "I am here."
gp.Idw_sa(water_level, wLevel, wl_output, "25", "2", "VARIABLE 10", "")
gp.AddMessage("Successful" + gp.GetMessages())
print "Successful", gp.GetMessages()
# Process: Single Output Map Algebra...
gp.MultiOutputMapAlgebra_sa(InExpression1)
# Delete intermediate layers
#
gp.AddMessage ("Cleaning up .......")
gp.Delete (wl_output)
#
while field2:
saLevel = field2.Name
print "Salinity Level in the list is: ", str(saLevel)
# output names
sa_output = modelPath + saLevel + "_IDW1"
sa_log1 = modelPath + saLevel + "_log1"
sa_log2 = modelPath + saLevel + "_log2"
sa_log3 = modelPath + saLevel + "_log3"
YM_pred = modelPath + saLevel + "_YM1"
# Give a name for the fish habitat raster output
YM_habi = modelPath + saLevel + "_habitat"
# provide coeffecient values in InExpression2 for a fish species
InExpression2 = sa_log1 + " = 4.574 - (0.074 * " + sa_output + ")"
# Do not need to change the values
InExpression3 = sa_log2 + " = ( "+ sa_log1 +" - (2 * " + sa_log1 + "))"
InExpression4 = sa_log3 + " = exp(" + sa_log2 + ")"
InExpression5 = YM_pred + " = (1 /(1 + (" + sa_log3 + ")))"
InExpression6 = YM_habi + " = (" + wl_dem + " * " + YM_pred + ")"
#gp.AddMessage("Working on + wLevel; interpolating by using IDW ....") +
gp.GetMessage())
# Process: IDW...
gp.Idw_sa(salinity_data, saLevel, sa_output, "25", "2", "VARIABLE 10", "")
gp.AddMessage("Successful" + gp.GetMessages())
print "Successful", gp.GetMessages()
# Process: Single Output Map Algebra...
gp.MultiOutputMapAlgebra_sa(InExpression2)
gp.MultiOutputMapAlgebra_sa(InExpression3)
gp.MultiOutputMapAlgebra_sa(InExpression4)
gp.MultiOutputMapAlgebra_sa(InExpression5)
gp.MultiOutputMapAlgebra_sa(InExpression6)
field2 = fields2.Next()
# Delete intermediate layers
gp.AddMessage ("Cleaning up .......")
gp.Delete (sa_output)
gp.Delete (sa_log1)
gp.Delete (sa_log2)
gp.Delete (sa_log3)
The CLLAMM Dynamic Habitat
177
gp.Delete (YM_pred)
gp.Delete (wl_dem)
# looping
field1 = fields1.Next()
except:
gp.AddMessage(gp.GetMessage(1))
print gp.GetMessage(1)
The CLLAMM Dynamic Habitat
178
Appendix 6.3: Hourly water level predictions for January and July of 1976,
1988, 1993 and 2005 at Barker Knoll, Noonameena and Salt Creek.
1. Barker Knoll 1976
1.4
Water level (AHDm)
1.2
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
January
Day of Month
July
2. Noonameena 1976
1.4
Water level (AHDm)
1.2
1
0.8
0.6
0.4
0.2
0
1
2 3 4 5 6 7 8
January
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
179
3. Salt Creek 1976
1
Water level (AHDm)
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2 3 4 5 6
January
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
Day of Month
4. Barker Knoll 1993
1.2
Water level (AHDm)
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
January
July
Day of Month
5. Noonameena 1993
1
Water level (AHDm)
0.8
0.6
0.4
0.2
0
-0.2
1 2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
180
6. Salt Creek 1993
0.8
Water level (AHDm)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2 3 4 5 6
January
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
Day of Month
7. Barker Knoll 1988
1.2
Water level (AHDm)
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
1 2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
Day of Month
8. Noonameena 1988
1.4
Water level (AHDm)
1.2
1
0.8
0.6
0.4
0.2
0
-0.2
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
January
July
The CLLAMM Dynamic Habitat
Day of Month
181
9. Salt Creek 1988
1.4
Water level (AHDm)
1.2
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
January
July
Day of Month
10. Barker Knoll 2005
0.8
Water level (AHDm)
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
1 2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
Day of Month
11. Noonameena 2005
0.7
Water level (AHDm)
0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
1 2 3
January
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
182
12. Salt Creek 2005
0.7
Water level (AHDm)
0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
1 2 3
January
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
183
Appendix 6.4: Hourly salinity predictions for January and July of 1976, 1988,
1993 and 2005 at Barker Knoll, Noonameena and Salt Creek.
1. Barker Knoll 1976
40
35
Salinity (g/L)
30
25
20
15
10
5
0
1
2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
2. Noonameena 1976
16
14
Salinity (g/L)
12
10
8
6
4
2
0
1
2 3 4 5 6
January
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
184
3. Salt Creek 1976
50
45
Salinity (g/L)
40
35
30
25
20
15
10
5
0
1
2 3 4 5 6
January
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
4. Barker Knoll 1993
35
30
Salinity (g/L)
25
20
15
10
5
0
1
2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
5. Noonameena 1993
20
18
Salinity (g/L)
16
14
12
10
8
6
4
2
0
1
2 3 4
January
5 6 7 8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
185
6. Salt Creek 1993
50
49
Salinity (psu)
48
47
46
45
44
43
1
2 3 4
5 6 7 8
January
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
7. Barker Knoll 1988
50
45
Salinity (g/L)
40
35
30
25
20
15
10
5
0
1
2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
8. Noonameena 1988
60
Salinity (g/L)
50
40
30
20
10
0
1
2 3 4 5 6
January
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
186
9. Salt Creek 1988
100
Salinity (g/L)
95
90
85
80
75
70
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
January
Day of Month
July
10. Barker Knoll 2005
60
Salinity (g/L)
50
40
30
20
10
0
1
2 3 4 5
January
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Day of Month
July
11. Noonameena 2005
80
70
Salinity (g/L)
60
50
40
30
20
10
0
1
2 3 4
January
5 6 7 8
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
187
12. Salt Creek 2005
160
140
Salinity (g/L)
120
100
80
60
40
20
0
1 2
January
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
July
The CLLAMM Dynamic Habitat
Day of Month
188
Appendix 6.5: Mudflat availability on the day with maximum and minimum
mean water level for January and July of wet and dry years at Barker Knoll,
Noonameena and Salt Creek.
1. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached maximum level at Barker Knoll (22 January 1976).
2
1.75
0.5
1.5
0.4
1.25
0.3
1
0.75
0.2
0.5
0.1
Mudflat area (ha)
Mean water level (mAHD)
0.6
0.25
0
0
6:00
7:00
8:00
9:00
10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jan. 1976)
Mean water level
East Coast
West Coast
Channel
1.4
1.75
1.2
1.5
1
1.25
0.8
1
0.6
0.75
0.4
0.5
0.2
0.25
0
6:00
Mudflat area (ha)
Mean water level (mAHD)
2. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached maximum level at Barker Knoll (21 July 1993).
0
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (21 Jul. 1993)
Mean water level
East Coast
West Coast
Channel
The CLLAMM Dynamic Habitat
189
0.2
3
0.1
2.5
0
2
-0.1
1.5
-0.2
1
-0.3
0.5
-0.4
Mudflat area (ha)
Mean water level (mAHD)
3. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached minimum level at Barker Knoll (31 January 1976).
0
6:00
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (31 Jan. 1976)
Mean water level
East Coast
West Coast
Channel
0.4
3
0.3
2.5
0.2
2
0.1
1.5
0
1
-0.1
Mudflat area (ha)
Mean water level (mAHD)
4. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached minimum level at Barker Knoll (9 July 1993).
0.5
-0.2
0
6:00
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (9 Jul. 1993)
Mean water level
East Coast
West Coast
Channel
The CLLAMM Dynamic Habitat
190
5. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached maximum level at Barker Knoll (17 January 1988).
1.75
1.5
0.3
1.25
1
0.2
0.75
0.5
0.1
Mudflat area (ha)
Mean water level (mAHD)
0.4
0.25
0
0
6:00
7:00
8:00
9:00
10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (17 Jan. 1988)
Mean water level
East Coast
West Coast
Channel
6. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached maximum level at Barker Knoll (22 July 1988).
1.75
1.5
0.8
1.25
0.6
1
0.4
0.75
0.5
0.2
0.25
0
6:00
Mudflat area (ha)
Mean water level (mAHD)
1
0
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jul. 1988)
Mean water level
East Coast
West Coast
Channel
The CLLAMM Dynamic Habitat
191
7. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached minimum level at Barker Knoll (8 January 1988).
3
2.5
0.1
2
0
1.5
1
-0.1
Mudflat area (ha)
Mean water level (mAHD)
0.2
0.5
-0.2
0
6:00
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jan. 1988)
Mean water level
East Coast
West Coast
Channel
8. Mudflat availability on the eastern and western shores and the channel for the day
the mean water level reached minimum level at Barker Knoll (8 July 1988).
1.75
1.5
0.3
1.25
1
0.2
0.75
0.5
0.1
Mudflat area (ha)
Mean water level (mAHD)
0.4
0.25
0
6:00
0
7:00
8:00
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jul. 1988)
Mean water level
East Coast
West Coast
Channel
The CLLAMM Dynamic Habitat
192
9. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Noonameena (22 January 1976).
2.3
0.8
2
0.6
1.7
0.4
1.4
Mean water level
0.2
6:00
7:00
8:00
East Coast
Mudflat area (ha)
Mean water level (mAHD)
1
West Coast
1.1
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jan. 1976)
10. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Noonameena (21 July 1976).
1.4
1.3
0.55
0.45
1.2
0.35
1.1
Mean water level
East Coast
1
West Coast
0.9
6:00
7:00
8:00
0.25
Mudflat area (ha)
Mean water level (mAHD)
0.65
0.15
0.05
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (21 Jul. 1976)
The CLLAMM Dynamic Habitat
193
11. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Noonameena (31 January 1976).
2.45
2.35
-0.039
2.25
-0.069
Mean water level
2.15
East Coast
Mudflat area (ha)
Mean water level (mAHD)
-0.009
West Coast
-0.099
2.05
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (31 Jan. 1976)
12. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Noonameena (9 July 1993).
2.3
0.065
2.2
2.1
0.025
Mean water level
East Coast
2
Mudflat area (ha)
Mean water level (mAHD)
2.4
1.9
West Coast
-0.015
6:00
7:00 8:00
1.8
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (9 Jul. 1993)
The CLLAMM Dynamic Habitat
194
13. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Noonameena (17 January 1988).
2.1
2.08
0.35
2.06
0.3
2.04
0.25
Mudflat area (ha)
Mean water level (mAHD)
0.4
2.02
Mean water level
East Coast
West Coast
0.2
2
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (17 Jan. 1988)
14. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Noonameena (22 July 1988).
3.75
3.25
1
2.75
0.95
2.25
Mean water level
0.9
6:00
7:00
East Coast
Mudflat area (ha)
Mean water level (mAHD)
1.05
West Coast
1.75
8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jul. 1988)
The CLLAMM Dynamic Habitat
195
15. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Noonameena (8 January 1988).
2.35
2.3
0.01997
2.25
0.01497
2.2
2.15
0.00997
2.1
0.00497
Mean water level
East Coast
West Coast
Mudflat area (ha)
Mean water level (mAHD)
0.02497
2.05
-0.00003
2
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jan. 1988)
0.6
2.2
0.55
2.1
0.5
2
0.45
1.9
Mean water level
0.4
6:00
7:00
8:00
East Coast
Mudflat area (ha)
Mean water level (mAHD)
16. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Noonameena (8 July 1988).
West Coast
1.8
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jul. 1988)
The CLLAMM Dynamic Habitat
196
17. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Salt Creek (22 January 1976).
4.6
4.1
0.415
3.6
3.1
0.41
2.6
2.1
0.405
Mean water level
East Coast
West Coast
Mudflat area (ha)
Mean water level (mAHD)
0.42
1.6
0.4
1.1
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jan. 1976)
18. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Salt Creek (21 July 1976).
5.5
4.5
1.06
3.5
Mean water level
1.03
West Coast
1
6:00
2.5
East Coast
7:00
8:00
Mudflat area (ha)
Mean water level (mAHD)
1.09
1.5
0.5
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (21 Jul. 1976)
The CLLAMM Dynamic Habitat
197
19. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Salt Creek (31 January 1976).
4.2
0.265
3.7
0.26
0.255
3.2
Mean water level
0.25
East Coast
West Coast
0.245
0.24
6:00
2.7
Mudflat area (ha)
Mean water level (mAHD)
0.27
7:00
8:00
2.2
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (31 Jan. 1976)
0.29
4.1
0.28
3.6
0.27
3.1
Mean water level
East Coast
0.26
0.25
6:00
West Coast
2.6
7:00
8:00
Mudflat area (ha)
Mean water level (mAHD)
20. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Salt Creek (9 July 1993).
2.1
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (9 Jul. 1993)
The CLLAMM Dynamic Habitat
198
21. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Salt Creek (17 January 1988).
2.2
2
1.8
-0.15
1.6
1.4
-0.155
Mean water level
East Coast
West Coast
Mudflat area (ha)
Mean water level (mAHD)
-0.145
1.2
-0.16
1
6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (17 Jan. 1988)
22. Mudflat availability on the eastern and western shores for the day the mean water
level reached maximum level at Salt Creek (22 July 1988).
4.7
4.2
1
3.7
3.2
0.95
2.7
0.9
2.2
Mean water level
0.85
0.8
6:00
East Coast
1.7
West Coast
1.2
7:00
8:00
Mudflat area (ha)
Mean water level (mAHD)
1.05
0.7
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (22 Jul. 1988)
The CLLAMM Dynamic Habitat
199
23. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Salt Creek (8 January 1988).
2.5
-0.1
2
-0.11
-0.12
Mean water level
1.5
East Coast
-0.13
-0.14
6:00
Mudflat area (ha)
Mean water level (mAHD)
-0.09
West Coast
7:00
8:00
1
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jan. 1988)
24. Mudflat availability on the eastern and western shores for the day the mean water
level reached minimum level at Salt Creek (8 July1988).
4.3
3.8
0.55
3.3
0.5
2.8
Mean water level
East Coast
0.45
West Coast
0.4
6:00
7:00
8:00
2.3
Mudflat area (ha)
Mean water level (mAHD)
0.6
1.8
1.3
9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00
Hours (8 Jul. 1988)
The CLLAMM Dynamic Habitat
200
GC = Goolwa Channel
MC = Mundoo Channel
BK = Barker Knoll
EI = Ewe Island
PP = Pelican Point
MP = Mark Point
LP = Long Point
NM = Noonameena
PA = Parnka Point
VY = Villa dei Yumpa
JP = Jack Point
SC = Salt Creek.
Appendix 6.6: Predicted salinities in July 1976, July 1988 and January 2005.
The CLLAMM Dynamic Habitat
201
Appendix 6.7: Barrage flow into the Coorong between 1960 and 2008
(Source: Webster, I. T., CSIRO).
Barrage flow (1000 Gl/month)
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
Years
Monthly barrage flow into the Coorong between 1960 and 2006
Mean monthly barrage flow (1000 Gl)
2
1.5
1
0.5
0
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
Years
Mean monthly barrage flow into the Coorong between 1960 and 2006
The CLLAMM Dynamic Habitat
202
The CLLAMM Dynamic Habitat
203
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement