Managing Invasive Native Scrublands for Improved Biodiversity Outcomes in Agricultural Landscapes

Full text

Managing Invasive Native Scrublands for Improved
Biodiversity Outcomes in Agricultural Landscapes
Final Scientific Report, June 2009
Veronica A.J. Doerr, Erik D. Doerr, Sue McIntyre, Jacqui Stol, Micah Davies,
Alex Drew, Garth Warren
CSIRO Sustainable Ecosystems
Darren Moore
CSIRO Information and Communication Technologies
Gary Howling
Department of Environment and Climate Change, New South Wales
FOR: Central West Catchment Management Authority

i
Please note:
This report was prepared as a result of work funded by the Central West Catchment
Management Authority (CMA). The views expressed here are those of the author(s) and do
not necessarily represent the views of the CMA or its employees.
The contents of this report do not necessarily represent the official policy of the Central West
CMA, the NSW Department of Environment, Climate Change and Water, the NSW Government
or any other agency or organisation.
While every effort has been made to ensure this information is correct, the Central West CMA
does not accept any responsibility for any errors, omissions or inaccuracies in this report.
Managers should take their own individual circumstances into account when making decisions,
and should seek extension advice where appropriate.
Also, clearing Invasive Native Scrub (INS) species requires permission under the Native
Vegetation Act 2003. Special provisions are available for landholders wishing to clear INS
under this Act.
Advice should be sought from your local CMA office before commencing any clearing
activity.

Contents
1. SCIENTIFIC REPORT .......................................................................................... 3
1.1 Summary................................................................................................................... 3
1.2 Introduction ............................................................................................................... 9
1.2.1 Invasive Native Scrub Management in Australia ................................................... 9
1.2.2 Mosaic Landscapes Around the World................................................................ 11
1.2.3 Mosaic Landscapes: more than a sum of their parts? ......................................... 12
1.2.4 INS Mosaics & Biodiversity: what do we already know?...................................... 14
1.3 Aims & Methods......................................................................................................17
1.3.1 Objectives ........................................................................................................... 17
1.3.2 Vegetation states, study landscapes & sampling sites ........................................ 18
1.3.3 Quantifying vegetation......................................................................................... 24
1.3.4 Bird surveys ........................................................................................................ 25
1.3.5 Assessing edge effects ....................................................................................... 26
1.3.6 GIS analyses of landscape proportions............................................................... 27
1.3.7 Statistical analyses.............................................................................................. 28
1.4 Results .................................................................................................................... 32
1.4.1 Descriptive Results: vegetation states................................................................. 32
1.4.2 Descriptive Results: birds in landscapes of the south-eastern Cobar Peneplain. 39
1.4.3 Statistical Results: effects of vegetation states on birds...................................... 42
1.4.4 Role of Open Woodland ...................................................................................... 48
1.4.5 Statistical Results: effects of landscapes on birds............................................... 50
1.4.6 Are edges special?.............................................................................................. 54
1.4.7 Inferring the effects of drought vs. land management ......................................... 55
1.5 Discussion............................................................................................................... 57
1.5.1 Ecological values of different vegetation states................................................... 57
1.5.2 Values of landscapes with different vegetation state proportions ........................ 59
1.5.3 INS mosaic landscapes: are they more than a sum of their parts? ..................... 60
1.5.4 Implications for best-practice management......................................................... 61
1.6 Future Research ..................................................................................................... 62
1.7 Short Report on Bioacoustic Sensor Networks ...................................................... 64
1.8 Acknowledgements................................................................................................. 69
1.9 Literature Cited .......................................................................................................69
1.10 Appendix A.............................................................................................................. 71
Cover photos:
Senna artemisoides, one of the less common INS shrubs, in full bloom. © Veronica Doerr
Ground cuckoo-shrikes depend on healthy native pastures & grasslands. © Damien Farine
This report should be cited as: Doerr, V. A. J., E. D. Doerr, S. McIntyre, G. Howling, J. Stol,
M. Davies, A. Drew, G. Warren and D. Moore. 2009. Managing Invasive Native Scrublands for
Improved Biodiversity Outcomes in Agricultural Landscapes. Report to the Central West
Catchment Management Authority, Dubbo.

3
1. SCIENTIFIC REPORT
1.1 Summary
Background on INS Management
Australia’s arid and semi-arid interior tends to be characterised by diverse native
ecosystems that sometimes include a dense mid-story layer, composed of shrub species or
of young trees, regenerating in particularly dense stands. While the phenomenon of dense
regeneration is a natural one which has always existed in the landscape, altered landscape
processes have resulted in increased dense woody vegetation, generally termed “Invasive
Native Scrub” (INS). Under the Native Vegetation Act 2003, an Assessment Methodology for
INS was developed on the basis that clearing INS in certain circumstances and under certain
conditions can improve or maintain environmental outcomes, not just production outcomes.
The stated aim of managing INS using the Assessment Methodology is to rehabilitate native
vegetation to create a mosaic of vegetation communities across the landscape. However,
there are still gaps in our understanding of what these mosaics should look like (which
vegetation states, how much of each, etc.) in order to support healthy native ecosystems.
Ecology of Mosaic Landscapes
The first step in developing an understanding of the ecological roles of vegetation
states and their mosaics in INS landscapes is to gain some general understanding of the
natural functioning of semi-arid heterogeneous ecosystems, or natural mosaic landscapes.
They tend to arise in arid and semi-arid parts of the world, where patchy dynamics are
created not just by different soil types or internal ecological processes, but also by external
environmental disturbances, including fire, unpredictable rainfall events (including pulses of
rainfall and periods of drought), and grazing. While there is still much to learn about the
vegetation dynamics of these systems, this general ecology suggests that animal species
living in these ecosystems should be specifically adapted to a patchy mosaic, but one that
shifts and changes across the landscape over time. Human activities, including agricultural
production, may intensify these disturbances or provide additional disturbances in the
landscape. Thus, management may need to ensure the general composition of the
landscape (proportions of different vegetation states) remains stable, even if any given site
undergoes transitions between vegetation states over time.
Ecology of INS Mosaics
Ecosystems that contain INS in Australia are classic mosaic landscapes, with key
disturbances stimulating changes between a wide variety of vegetation states. These states
include native perennial pastures, shrubland, and any of a wide variety of woodland types,
some with a shrubby mid-story varying greatly in density. The most comprehensive survey-
based study of biodiversity in INS ecosystems in Australia suggested that landscape
heterogeneity is important for maintaining biodiversity because some species were more
abundant at high levels of shrub cover, others were more abundant at low levels of shrub

4
cover, and some even preferred intermediate levels of shrub cover. However, it remains
unclear exactly how different distinct vegetation states contribute to supporting biodiversity,
and thus how much of each state should be present in a healthy landscape and in what
configuration, as well as how dynamic changes in vegetation states can best be managed to
ensure sustainability of native species.
Filling Knowledge Gaps: aims of this study
This project was designed to provide a preliminary investigation into remaining
questions about the value of INS landscape mosaics for biodiversity by focusing on the
compositional characteristics that mosaics should have to support native ecosystems.
Specifically, our objectives were to focus on collecting data on bird species in the south-
eastern Cobar Peneplain to:
1) Assess the relative contribution made by all vegetation states in the landscape
(including both open areas and INS areas) to supporting native species, quantified in
terms of avian species richness, diversity, and community composition
2) Determine whether landscapes with different proportions of the different vegetation
states support different numbers, diversity, or communities of native bird species
3) Evaluate whether any differences between landscapes are purely due to an additive
effect or whether some landscapes are more or less than a sum of their parts, and
thus support different numbers or diversity of species than would be expected simply
based on the proportions of different vegetation states
4) Discuss the consequences of our results for best-practice management of INS
Methods
Research was conducted in 18 study landscapes located in the south-eastern portion
of the INS region, north and west of the town of Tottenham, NSW. Each 500-ha study
landscape contained varying proportions of four different vegetation states:
 ANT; paddocks dedicated to agricultural production and containing few, if any, trees
 AWT; paddocks dedicated to agricultural production and containing scattered
paddock trees
 OSL; open shrubland or shrubby woodland, where clusters of trees and/or shrubs are
evident but are distinctly separated by open ground
 CSL; closed shrubland or shrubby woodland, where the tree and shrub cover is more
uniformly dense
All landscapes contained some areas of OSL, some areas of CSL, and areas of either
ANT or AWT, though a few landscapes had both agricultural vegetation states. Some
landscapes were dominated by one or a couple of vegetation states, while others had more
even representation of the different states. Depending on definitions used, AWT could be
considered open woodland, but a highly modified form, with overstory cover values well
below benchmark values and the ground layer significantly modified by cropping (current or
in the recent past) at most sites. Two separate smaller areas of open woodland (OWL) with
overstory characteristics much closer to benchmark values were also included in the study,

5
as this vegetation state was much less common than the other vegetation states in the
region and not found within any of the 18 primary study landscapes.
Within each vegetation state in each study landscape (and in the two OWL areas), data
on vegetation and birds present were collected at four sampling sites (a grand total of 236
sampling sites). Vegetation data were collected in April 2008, focusing on primary
characteristics (e.g., percent cover, height, and dominant species) of the ground, shrub, and
overstory layers. Bird surveys (point counts) were conducted to assess the presence and
relative abundance of different species in October 2007 and again in September 2008, with
two surveys conducted in consecutive weeks at each site in each year for a total of four
surveys at each sampling site (an overall total of 944 point counts).
Data from bird surveys were used to calculate species richness (number of different
species present) and the Shannon diversity index (a measure of evenness of distribution of
individuals among the species present, hereafter referred to as ‘diversity’). These were
calculated at different scales ranging from the individual sampling points up to the study
landscapes. We then analysed how richness, diversity, and species composition of the bird
community differed among vegetation states, across more quantitative variation in
vegetation, and in landscapes with different proportions of the different vegetation states.
We also assessed whether edges between scrub and agricultural vegetation states might
comprise an additional important resource making landscapes more than just a sum of their
vegetation states. To do this, we also performed point count surveys at edges to compare
with matched surveys in the interior of the constituent vegetation states and followed
individuals at edges to quantify their behaviour in the different vegetation states.
Results: value of vegetation states in the mosaic
All four of the primary vegetation states plus mallee (not considered INS) and small
areas of open woodland appeared to be ecologically valuable, and contributed to the overall
biodiversity of the region. Scrub vegetation states contained more species and were more
diverse than the agricultural vegetation states, but the agricultural vegetation states
supported different communities of species. Open woodland (OWL) was most similar to
agricultural areas with trees (AWT) in bird richness and diversity but did not support a unique
community. Thus, healthy mosaics should contain a mixture of agricultural areas with (AWT)
and without (ANT) an overstory of trees, open scrubland (OSL), and closed scrubland (CSL),
including some mallee and open woodland where they naturally occur.
While richness and diversity generally increased with increasing woody vegetation,
particularly shrub cover, there was some evidence that OSL may actually support more
species and be more diverse than CSL, and it may support a moderately different bird
community, especially where it combines patchy shrub cover with a grassed ground layer in
the spaces between shrubs. OSL, and to a lesser extent CSL, was most likely to support
species that are currently vulnerable or in decline, including species usually classified as
“declining woodland birds”, which were generally not present in OWL.

6
Lower richness and diversity in the agricultural vegetation states could possibly be due
to the loss of ground herbaceous cover within pastures (which was markedly low in this
study), and was not solely due to drought conditions in 2007. Declines in grassland birds
throughout Australia have been linked to changes in ground vegetation, though these
proposed links have not been fully tested. Thus, the contribution of agricultural vegetation
states could potentially be improved with pasture management aimed at retaining cover and
diversity of native grasses and forbs. Finally, AWT may be slightly more valuable to birds
than ANT because it was rarer and supported more rare bird species.
Results: how much of each
Landscapes with greater proportions of scrub vegetation states supported greater
numbers of bird species and greater avian diversity. However, landscapes with a majority of
agricultural vegetation states (>50%) supported a different community of bird species
compared to landscapes that contained a majority of scrub vegetation states. This was not
necessarily a threshold effect, and landscapes that had approximately 33-67% scrub
captured the majority of community variability in the study, with landscapes at the low end
generally supporting open country bird communities and landscapes at the high end
generally supporting scrub bird communities.
Most of the evidence suggested that landscapes in this study were not functioning as
more than a sum of their parts, and thus not supporting greater numbers of species or
greater diversity than would be expected based on the proportions of vegetation states they
contain. The richness and diversity of birds found in any given vegetation state did not
depend on the proportions of other vegetation states found in the landscape, or on other
unmeasured characteristics of landscapes. Edges between vegetation states did not support
more species or more individuals than the interior of adjacent scrub vegetation states. While
individuals at edges did use the scrub and agricultural vegetation states differently, the lack
of a positive edge effect suggests that these birds probably range widely over different
vegetation states rather than concentrate their activities specifically at edges.
This does not necessarily mean that the configuration (size of patches of each
vegetation state, connectivity between them, etc.) of landscapes has no influence on the
biodiversity of landscapes, as this study was not specifically designed to test for effects of
configuration. However, it does suggest that INS mosaic landscapes may not function in a
very different way from a bird perspective than other types of mixed-use landscapes.
Management recommendations
Our final project aim was to discuss the consequences of our results for best-practice
management of INS within the south-eastern Cobar Peneplain. We suggest several best-
practice management recommendations based on the current state of knowledge of bird
communities in landscapes containing INS:

7
1. Plan to achieve or maintain a mosaic-like mixture of vegetation states at the scale of
individual properties. Management may then occur on a paddock-by-paddock basis,
but according to an overall plan for achieving a mosaic at the property scale.
2. Mosaics should contain a variety of vegetation states, not just open pastures and
dense scrubby areas. Mosaics should include native perennial pastures both with
and without an open overstory of trees, scrub areas that have open grassy areas
between clumps of shrubs (“open” scrub), scrub areas that are dense or “closed”
(without clear open spaces in between trees and shrubs), as well as mallee and/or
open woodland if those states already exist in the area.
3. While no one vegetation state is most important for birds (having a mosaic mixture is
most important), there are two vegetation states that may particularly contribute to
supporting a diversity of native birds in the landscape and should thus be part of any
mosaic. These are:
a. Pasture paddocks with an open overstory of trees
b. Areas of open scrub (with clear separation between clumps of trees and
shrubs), particularly those that have a healthy grassy ground layer in the
areas between trees and shrubs
4. To maintain bird community diversity at the property scale, aim to achieve a
landscape with somewhere between 33% and 67% of the total area occupied of
various types of scrub (closed and open, dominated by various shrub and tree
species).
5. Landscapes dominated by scrub vegetation states support a greater diversity of bird
species, but landscapes dominated by agricultural vegetation states support a
different community of bird species. Thus, to maintain a diversity of native birds at
the regional scale, make sure that some landscapes are dominated (>50%) by scrub
vegetation states while others are dominated by agricultural vegetation states.
Additional recommendations based on other published studies:
1. The configuration of vegetation states may still be important but was not investigated
in this study. However, most of the bird species in this study have been included in
configuration research in other ecosystems. Thus, a precautionary approach would
involve following recommendations from other regions in Australia to plan the
configuration of vegetation states. At the moment these recommendations would
include:
a. ensure patches of any given vegetation state are at least 10ha in size
b. ensure patches are relatively round or square rather than linear (in other
words, avoid linear buffers unless they are connecting patches as in d. below)
c. maintain patches of scrub vegetation states that are separated from other
patches of scrub by no more than 1km
d. connect patches of scrub vegetation states that are separated by up to 1km
with either with a continuous corridor of scrub or with a paddock containing

8
scattered trees in which the trees are separated from each other by no more
than 100m
2. Managing pastures for greater cover of native perennial species using techniques
recommended for production purposes (such as controlling total grazing pressure,
grazing in phases separated by significant rest periods, destocking while native
plants are setting seed, etc.), is also likely to benefit native biodiversity. Note that
other research also suggests that managing for greater diversity of ground plants
(through reduced total grazing pressure and extended time intervals between
cropping) would also be beneficial. In this study, most agricultural vegetation states
had very low levels of ground cover and low plant diversity, which may have been
one reason why agricultural vegetation states supported fewer bird species and lower
levels of diversity.
3. Monitor the results of management on a regular basis, not just in terms of production
outputs but also in terms of inputs and changes in the composition, abundance and
diversity of native communities. Use the results of monitoring to adjust management
actions to ensure they are achieving the desired goals.
Future research
At the start of this project, key knowledge gaps recognised by the INS Advisory
Committee included: 1) the composition and configuration of INS mosaics that best support
biodiversity, 2) understanding shifts between vegetation states to ensure we know what
management techniques to use to create appropriate mosaics, and 3) understanding how
dynamic changes between vegetation states influence biodiversity. While this study helped
to fill the first of these gaps, it focused on birds and it would be wise to ensure the same
patterns apply to other taxonomic groups in the region. In terms of the second knowledge
gap, new research is adding to our understanding of transitions between vegetation states.
Yet these new insights have yet to be incorporated into our understanding of how to
successfully manage INS landscapes. The third knowledge gap is particularly important but
has not yet been addressed. Management to create and manage planned mosaics involves
major structural changes to landscapes, and we need to know how much change we can
accomplish at one time without negatively impacting native species. We propose four future
research projects that will address these remaining gaps and continue to improve the
understanding and management of INS landscape mosaics.

9
1.2 Introduction
1.2.1 Invasive Native Scrub Management in Australia
Australia’s arid and semi-arid interior supports a range of diverse native ecosystems
that sometimes include a dense mid-story layer. This mid-story can be composed of shrub
species or of young trees, regenerating in particularly dense stands. The particular type of
vegetation found at any given site is thought to be the result of a wide variety of interacting
processes that cause change in the landscape, including fire, grazing by a suite of
herbivores, and unpredictable pulses of rainfall. At the eastern edges of these rangeland
ecosystems, agricultural production has been most intensive. As a result, the timing and
intensity of critical landscape processes have been altered, causing changes in the amount
of different types of vegetation in the landscape. In many areas, there has been an increase
in dense shrubs or densely regenerating trees (but see Fensham 2008 for an alternative
view), which are now considered an immediate problem for agricultural production. The
increase in dense vegetation may also be a problem for the overall health and biodiversity of
the native ecosystem which underpins agricultural production, thus posing a long-term threat
to agricultural sustainability.
While the phenomenon of patches of dense regeneration is a natural one which has
always existed in the landscape, the problem of altered landscape processes and the
resulting increase in dense woody vegetation, through both thickening and encroachment,
has been recognised as a significant issue by Government in the native vegetation
legislation. Trees and shrubs that regrow or establish in new areas particularly aggressively
have been termed “Invasive Native Scrub” (INS). INS is defined in current regulations as:
1. a species that invades plant communities where it has not been know to occur
previously or a species that regenerates densely following natural or artificial
disturbance; and
2. the invasion and/or dense regeneration of the species results in change of structure
and/or composition of a vegetation community; and
3. the species is within its natural geographic range.
Species that act as INS vary from “woody weeds” (e.g., turpentine and budda) which are a
particular feature of western areas, to dense white cypress pine and localised stands of
dense eucalypt regeneration (e.g., poplar box) in central regions.
Under the Native Vegetation Act 2003, an Assessment Methodology for INS was
developed on the basis that clearing INS in certain circumstances and under certain
conditions can improve or maintain environmental outcomes, not just short-term production
outcomes. The stated aim of managing INS using the Assessment Methodology is to
rehabilitate native vegetation to create a mosaic of vegetation states across the landscape
(Hassall & Associates et al. 2006). It is generally accepted that such a mosaic represents

10
the natural state of these ecosystems, and is what existed in these landscapes prior to
European settlement (Noble 1997). However, rather than managing the landscape to
recreate pre-European conditions, which are impossible to know precisely, the intent is to
manage the landscape as a mosaic of more open and more dense vegetation states, such
that both production and biodiversity are supported. Biodiversity must be defined not just as
the variety of species, but also as the variety of communities and ecological processes (such
as pollination, dispersal, nutrient and water flow, etc.) that are vital to ensuring the long-term
health and proper functioning of these systems1. Thus, understanding what is required in
terms of landscape mosaics to support biodiversity is no trivial task.
The details of the INS Assessment Methodology were designed to achieve these goals
based on the best available scientific evidence – on the requirements of native species, the
desirability of a heterogeneous landscape, and the effects of various landscape processes in
producing different densities of vegetation (Hassall & Associates et al. 2006). However,
there were still significant gaps in the necessary information. One key gap was related to the
requirements of native species, and in particular, what specific characteristics the
heterogeneous landscape should have. Existing evidence suggested that different animal
species prefer different densities of vegetation (Ayers et al. 2001), and thus it was clearly
desirable to have a variety of vegetation densities in any given landscape. Yet it was unclear
how much of a contribution different vegetation densities might make to the overall
biodiversity of the system, and thus whether particular vegetation “states”, when present in
greater proportions than others, result in greater diversity of species across the landscape as
a whole. It was also unclear whether the particular combination of different vegetation
densities might be important, such that the role of any given vegetation state depended on
the presence of other states in the landscape. This type of information could not only
improve the application of the Assessment Methodology, it is critical to helping landholders
plan a vegetation mosaic on their properties. This would facilitate a transition from a
regulatory framework, in which the focus is placed on paddock-scale decisions about
clearing INS, to a land management framework, in which state government and regional
Catchment Management Authorities (CMAs) provide information to assist landholders with
whole-of-property and regional planning and management.
The first step in developing an understanding of the ecological roles of each vegetation
type in INS landscapes, and what characteristics healthy diverse landscapes need to have,
is to gain some general understanding of the natural functioning of semi-arid heterogenous
ecosystems, or natural mosaic landscapes.
1 Note that the term ‘biodiversity’ is used to denote this meaning throughout the document. ‘Avian
diversity’ is used to refer specifically to Shannon’s diversity index calculated based on bird survey
data, and ‘avian richness’ is used to refer specifically to species richness calculated based on bird
survey data. Both of these are components of biodiversity, and thus factors that affect the avian
variables may affect biodiversity as a whole. Thus, ‘biodiversity’ is used when referring to concepts,
overall goals of management, and general conclusions, while the latter terms are used to specifically
refer to results of this study.

11
1.2.2 Mosaic Landscapes Around the World
While all landscapes are heterogeneous to some degree (Hobbs 1999), natural mosaic
landscapes are highly heterogeneous, with multiple communities and vegetation structures
coexisting in a patchy distribution at a range of spatial scales. They tend to arise in arid and
semi-arid parts of the world, where patchy dynamics are created not just by different soil
types or internal ecological processes, but also by external environmental disturbances
(Kleyer et al. 2007). Thus, they function in unique and different ways to other ecosystems.
In contrast, northern hemisphere temperate forest ecosystems are traditional
successional systems rather than mosaic systems, in which competition for light (an internal
process) causes grasslands to transition through different vegetation states and eventually
become mature forests. External disturbance such as fire simply resets this successional
process, but once it has been reset, the process will proceed in the absence of any particular
disturbance (Kleyer 2007). In contrast, African semi-arid savannah woodlands are natural
mosaic ecosystems in which fire, periodic grazing, and rainfall and drought events cause or
accelerate transitions between grasslands, dense thornveld, and treed savannahs (Wiegand
et al. 2006, Meyer et al. 2007). Patchy application of these disturbances in time and space
creates a patchy mosaic of different vegetation types in the landscape (Wiegand et al. 2005).
This may still be considered a successional process, but one in which transitions depend
more on external disturbances than on internal processes. Note that in the past, some
authors have suggested that these are state-and-transition dynamics rather than
successional dynamics (Westoby et al. 1989, Noble 1997). But the emerging view is that the
concept of successional dynamics should encompass both internally-driven and disturbance-
driven transitions between vegetation states, and that a state-and-transition model is often
an effective way to explore these dynamics in both traditional successional and mosaic
ecosystems (Wiegand et al. 2006, Meyer et al. 2007).
Landscapes can also become patchy in their distribution of vegetation types due to
human activities. For example, in many parts of the world, clearing and fragmentation of
continuous forests followed by the establishment of grazing pastures in cleared areas has
created an artificial landscape mosaic (Bennett et al. 2006). The term “mosaic landscape”
actually came into popular usage to refer to these types of artificial mosaics (Forman 1995),
though the concept was first developed and applied in natural systems (Watt 1947, Remmert
1991). Thinking of mixed production and natural landscapes as mosaics has been
particularly helpful in stimulating theories and research about appropriate land management,
as it is not possible or necessarily desirable to return most of these areas to continuous
natural vegetation. We must find an appropriate way to manage once continuous vegetation
types as mosaics, even if their natural ecological dynamics are not mosaic dynamics.
Human activities can also serve as an additional disturbance in natural mosaic
landscapes such as those containing INS plant species (Meyer et al. 2007, Kerle 2008,
Schlesinger et al. 2008), but such anthropogenic disturbances might have very different
consequences than in traditional successional landscapes. It might seem that managing for
both production and for the natural ecosystem should be easier in natural mosaic
landscapes, because the plants and animals should already be adapted to living in a patchy

12
environment. However, managing natural mosaic ecosystems may actually be more difficult,
because human disturbance can interact in a variety of ways with natural disturbance
processes, producing a wide variety of different outcomes from the same basic management
actions (Peters et al. 2006, Meyer et al. 2007). For example, the precise effects of livestock
grazing may depend on fire history, recent rainfall, and rainfall or drought in the immediate
future. Thus, effectively managing natural mosaic landscapes for both production and
biodiversity benefits may require particularly detailed knowledge of the complex ecological
dynamics of these ecosystems.
The lack of this knowledge, and the inappropriate use of information derived from
successional landscapes rather than natural mosaic landscapes, has resulted in
inappropriate management and widespread degradation of natural mosaic landscapes
around the world, largely due to homogenisation of natural mosaics and increases in the
denser vegetation types. In southern Africa, broadleaf thickets and dense thornveld are
becoming more dominant, replacing grassier savannah woodlands and grasslands, affecting
not only food production but also the diversity of native communities supported (Skowno and
Bond 2003, Wiegand et al. 2005, Meyer et al. 2007). In the northwestern U.S., western
juniper is growing more densely and replacing aspen forests (Belsky 1996). In the
southwestern U.S., shrubs like creosote and mesquite are replacing grasslands and native
pastures, eventually leading to functional desertification, at least from a primary production
perspective (Peters et al. 2006, Schlesinger et al. 2008). To manage these landscapes more
sustainably, we need to know the complex effects of human disturbances (i.e., our tools for
managing the landscape), but we also need to know what kind of mosaic best supports
sustainable native ecosystems (i.e., the characteristics of the landscapes we should be
aiming for in our management).
1.2.3 Mosaic Landscapes: more than a sum of their parts?
To restore and maintain healthy, functioning natural mosaic ecosystems, we need to
have an understanding of how the biodiversity of the native ecosystem depends on the
amount of each vegetation type in the landscape and on the spatial configuration of those
vegetation types. While this type of research has not yet been performed in Australia’s INS
ecosystems, information from research overseas and on artificial landscape mosaics in
Australia may raise some interesting questions and possibilities.
In particular, mosaic landscapes may have “emergent properties” – characteristics that
only apply to the landscape and not to lower levels of organisation like a vegetation patch –
and that those emergent properties significantly influence the number of species and the
types of ecological processes that the landscape successfully supports. Bennett et al.
(2006) suggested that there are four general types of landscape-level variables, or emergent
properties, that might be meaningful for the health of mosaic landscapes: extent of native
vegetation, composition (proportions of different vegetation types, number of patches, etc.),
configuration of each vegetation type (size and shape of patches, connections between
them), and geographic position or abiotic environmental variation.

13
Researching the effects of any of these types of variables on measures of ecological
health, such as species richness or diversity, requires quantification of both predictor and
response variables at the landscape scale. In other words, the mosaic itself must be treated
as the sampling unit, with species richness and other responses characterised for the whole
mosaic, not just for individual sampling sites within the mosaic (McGarigal and Cushman
2002, Bennett et al. 2006). This also means that whole landscape mosaics need to be
replicated, with a sufficient sample size to have the power to detect any effects of landscape-
level composition and configuration on the landscape’s ecological health.
Unfortunately, relatively few studies have collected data at the appropriate level and
with adequate sample sizes of replicated landscapes to test the effects of emergent
properties. Most of these studies have concentrated on artificial rather than natural mosaic
landscapes (Bennett et al. 2006). They have also tended to focus on simpler systems,
where the mosaic includes only a single native vegetation type, for example forests in a
mosaic of forests and pastures or crops (Pino et al. 2000). This ensures they can focus on
variables such as extent and configuration, rather than on potential interactions between
native vegetation types, which would require even larger sample sizes to investigate. For
example, in Australia, Radford and Bennett (2007) found that extent of woody vegetation was
an emergent landscape property positively related to the presence of woodland birds in the
artificial landscape mosaic of the eastern sheep-wheat belt. However, these types of
analyses have yet to be performed in natural mosaic landscapes such as those of the INS
region.
A special type of emergent property may arise due to an interactive effect of multiple
vegetation types. If some species require multiple vegetation types to meet their daily
needs, or to support different life history stages, they may only be present in landscapes that
have appropriate proportions of multiple vegetation types. Similarly, some species may be
particularly adapted to ecotones or edges, areas where multiple vegetation types intermingle,
and may only be present in landscapes that have sufficient ecotone habitat. Thus, species
richness in landscapes with these interactive emergent properties may be higher than
expected based on simply adding up the species supported by each individual vegetation
type – these landscapes may be more than a sum of their parts. Evidence in support of this
concept comes from the mosaic landscapes of the southern Kalahari, where yellow
mongoose prefer to nest under shrubs but forage in open grassy areas, so they specialise in
ecotones where shrublands and grasslands intermingle (Blaum et al. 2007). While this idea
has not yet been tested in Australian INS ecosystems, there is certainly evidence that many
Australian native species require multiple types of habitat for daily survival (Law and
Dickman 1998).
On the other hand, sometimes ecotones have a negative impact on overall species
richness or diversity as a result of edge effects, where the proximity of one vegetation type
decreases the value of the adjacent vegetation type (Ries et al. 2004). In Australia, noisy
miners (which are very common in the INS region) appear to prefer living at the edges of
woody and open habitats, and they behave aggressively toward other species, likely
producing a negative edge effect (Clarke and Oldland 2007). Thus, it is also possible that

14
interactive emergent properties can sometimes result in landscapes that are less than a sum
of their parts.
1.2.4 INS Mosaics & Biodiversity: what do we already know?
INS landscape mosaics quite clearly exhibit the dynamics and patterns identified for
other natural mosaic landscapes around the world. As highlighted in Noble (1997), a
combination of ecological data and historical records suggest that regions affected by INS
were historically areas consisting of a complex landscape mosaic of different vegetation
states, or habitat types. Differences in soil characteristics, climate, microtopography, and
disturbance history, mean that any given part of the landscape might have supported
grasslands, shrublands, or any of a wide variety of woodland types, some with a shrubby
mid-story varying greatly in density. In addition to shrubs, a number of native tree species
can regenerate very densely under certain conditions, thus also contributing to a complex
mid-story. Due to small-scale variations in soil, topography and disturbance, all of these
different vegetation states may once have had very small-scale, patchy distributions, with
multiple states in a landscape. Furthermore, these patchy distributions were almost certainly
shifting in space as disturbances continued to change local site characteristics. It was
probably in this kind of spatially variable and temporally dynamic environment that
communities of plants, invertebrates and animals developed.
The introduction of agriculture to this environment, including livestock production, has
apparently resulted in a homogenisation or polarisation of this spatial and temporal variability
(Figure 1). While some variability still exists, much of the finer-scale spatial and temporal
variation suggested by historical records has generally been lost (Noble 1997, Hassall &
Associates et al. 2006, Kerle 2008). The landscape mostly consists of relatively large
patches of uniform vegetation states: pasture (generally without shrubs but sometimes with
trees), which may be periodically cultivated, and wooded areas with relatively dense mid-
stories and little live plant cover in the ground layer. Vegetation states that are intermediate
to these are relatively rare. In addition, attempts are usually made to minimise temporal
variation, keeping a given paddock in the same vegetation state over a long period of time.
As a result, the ability of INS regions to support agricultural and livestock production,
particularly without heavy investment in inputs and intensive management, is thought to have
declined. An additional concern is that overall ecosystem health in these landscapes has
also declined as a result of this homogenisation, thus gradually eroding the resource base
upon which sustainable food production enterprises are thought to depend, particularly in the
long-term.

15
Historical variability Current variability
Figure 1. Diagrams representing historical high levels of spatial and temporal variability in
INS landscapes, and current lower levels. White areas represent native pastures/crops, dark
areas represent woodlands with a dense mid-story layer (usually considered INS), and grey
areas represent various states of moderate woody vegetation density, including open
woodland.
These declines raise a very fundamental question about landscape mosaics: how
much does the native ecosystem depend on spatial and temporal variability? How critical is
the mosaic itself to the long-term survival of native plants and animals? In other words, how
much can we homogenise the landscape (at the extreme reducing it down to just one
vegetation state), without compromising long-term sustainability of the ecosystem? There
are several possible answers to these questions. #1: It may be that most species in these
ecosystems are generalists, using a lack of specialisation to cope with unpredictable spatial
and temporal dynamics. If this is true, then it would be possible to maintain a healthy
landscape with just one vegetation state, as long as that state contained a diversity of plant
species. #2: Alternatively, most species may tend to specialise on particular vegetation
states, in which case multiple states are required for a healthy landscape, but a fine
patterning among them may not be necessary. #3: Finally, the majority of species might be
particularly adapted to the mosaic per se (to the interactive emergent properties discussed
above). In this case, species may require multiple types of habitat or an abundance of
habitat edges, and thus a finely patterned mosaic may be particularly important for overall
ecosystem health.
While these particular possibilities have not been directly investigated in Australian INS
ecosystems, there has been some work on the presence of native species relative to
vegetation characteristics that can begin to provide the answers. Most importantly, Ayers et
al. (2001) conducted a survey-based study of some elements of biodiversity in INS
ecosystems in Australia. While not all elements of biodiversity were addressed, and the
study was conducted in a limited number of geographically restricted study sites, this is the
most comprehensive study available of biodiversity in INS-affected ecosystems. They did
not divide the landscape into vegetation states, but rather analysed the influence of shrub
cover as a continuous variable on a range of measures of species diversity and landscape
function. All data were collected on a site scale (2 ha plots) in a single year (spring-summer
of 1999/2000) in three different regions. Surveys quantified the presence and abundance of
a diverse range of taxonomic groups, including plants, invertebrates, and various vertebrate

16
groups. While most broad taxonomic groups showed no general relationship to shrub cover,
individual species were identified that had different shrub cover preferences. Some species
were more abundant at high levels of shrub cover, others were more abundant at low levels
of shrub cover, and some even preferred intermediate levels of shrub cover. These results
were important in shaping an understanding of animals in INS ecosystems and informing
management. In particular, they suggested that possibility #1 above, that homogenisation of
the landscape in western New South Wales could still result in a healthy ecosystem, was
highly unlikely to be true. Instead, Ayers et al. (2001) provided strong empirical support that
a mosaic of vegetation states is required to maintain landscape health, partly because
different animal species require different vegetation states. In reviews of literature on INS
species and biodiversity, Hassall & Associates (1999) and (2006) also found observations
and anecdotal evidence to support this point, from Clipperton in an unpublished
undergraduate report to the University of New England, and from Cameron in an unpublished
document associated with the Western Plains Zoo, Dubbo.
However, when focusing on a breadth of taxa, sampling will necessarily be limited to
the site scale rather than the landscape scale. Thus, Ayers et al. (2001) did not have the
opportunity to explore the characteristics that landscape mosaics should have – what the
proportions of different vegetation states should be, and whether some species are
particularly adapted to the mosaic itself and may require a relatively fine patterning of
different vegetation states (possibility #3 above). Hassall & Associates (1999) also found
some observational data suggesting that scrub areas were more species rich than
agricultural areas or native grasslands, so healthy landscapes might have a greater
proportion of shrubby areas than open areas. Based on ecological information about
individual species such as grey kangaroos, they also proposed that some species might be
particularly adapted to the mosaic itself – to areas where two or more vegetation states exist
in close proximity – because different vegetation states are required for different daily
activities. As a result, intermediate levels of shrub cover and/or fine-scale patterning of
vegetation states in the landscape might be particularly important for biodiversity. However,
at the time there was no information in the literature to either support or refute these
hypotheses, nor was there any information on how dynamic changes between vegetation
states (i.e., a shifting mosaic as opposed to a static mosaic pattern) might influence
biodiversity.
Since the Hassall & Associates (1999) review and Ayers et al. (2001) analyses, there
have only been a few studies that have contributed to helping fill knowledge gaps on the
structure and degree of dynamism that INS mosaics might need to have to support native
species. Noble et al. (unpublished ms) also found greater diversity of bird species in
shrublands compared to pastures in sites in northwestern New South Wales and southern
Queensland, adding to the evidence that shrub habitats may be slightly more important
contributors to the overall mosaic. But they also found that there was major seasonal
variation in abundances, largely driven by rainfall, calling into question conclusions drawn
from just a single season of surveys. Modelling work by Bradstock et al. (2005) showed that
a mosaic of burning intervals (and thus vegetation states) is required to accommodate
different species, thus providing further evidence that vegetation mosaics are required for
supporting the full range of native species in these systems.

17
In terms of the more detailed structure these mosaics should have, Hodgkinson (2002)
discussed the need to understand links between landscape function and biological diversity.
He suggested that the spatial heterogeneity necessary in arid and semi-arid Australian
landscapes for retaining water and nutrients may be a good predictor of the type and scales
of heterogeneity required for plants and animals, but that this idea required further testing.
Briggs et al. (2007) analysed species richness and density of birds in the box/cypress pine
woodlands just southeast of the INS region as a function of remnant size and condition.
They found that, like most studies from more open woodlands further east, larger areas of
woodland supported more species. In a review of both published and unpublished literature
on white cypress pine woodlands throughout their range, Thompson and Eldridge (2005)
also found evidence that larger areas of pine woodlands supported a greater diversity of
species. This suggests that large blocks of different vegetation states might be preferable to
a fine-scale mosaic, at least for some species, but this needs to be verified specifically within
the region affected by INS, where cypress pine may function differently than it does in other
parts of its range.
Thus, while it is clear that the INS region should consist of a mosaic of different
vegetation states, significant gaps still remain in our understanding of what that mosaic
should look like and how rapidly mosaics might be able to change and still support healthy
native ecosystems. While the existing literature suggests that mosaics with more shrubby
areas than open areas should support a greater number of species, particularly if they
contain some intermediate vegetation states with only moderate shrub cover, these are
merely hypotheses that still require testing in the INS region. Furthermore, there is a small
amount of evidence that INS mosaics may function best when vegetation states exist in
relatively large blocks. Yet extensive work on landscape function has established the
important role of finer-scale patchiness in these landscapes, which may also be important to
biodiversity but has rarely been examined to date.
1.3 Aims & Methods
1.3.1 Objectives
This project was designed to provide a preliminary investigation into remaining questions
about the value for biodiversity of INS landscape mosaics in the south-eastern Cobar
Peneplain by focusing on the compositional characteristics that mosaics should have to
support native birds. Specifically, our objectives were to:
1) Assess the relative contribution made by all vegetation states in the landscape
(including both open areas and INS areas) to supporting native bird species,
quantified in terms of species richness, diversity, and community composition
2) Determine whether landscapes with different proportions of the different vegetation
states support different numbers, diversity, or communities of native bird species
3) Evaluate whether any differences between landscapes are purely due to an additive
effect or whether some landscapes are more or less than a sum of their parts, and

18
thus support different numbers or diversity of bird species than would be expected
simply based on the proportions of different vegetation states
4) Discuss the consequences of our results for best-practice management of INS
Birds are just one element of biodiversity but were the focus of this preliminary project
because they are one of the most tractable species groups in INS ecosystems as they are
the most commonly encountered and easily observed species. While they have successfully
been used as surrogates for all biodiversity in an ecosystem (Gregory et al. 2008), the
degree to which their needs and responses to the environment really do encompass those of
most other species is uncertain (Larsen et al. 2009). In other words, sometimes they appear
to be good indicators for many other aspects of biodiversity and sometimes they do not.
Thus, our objectives were to consider the species richness, diversity and community
composition of birds with the understanding that this informs us of the requirements of some
components of biodiversity in ecosystems containing INS. However, other species groups or
landscape processes may show different responses.
1.3.2 Vegetation states, study landscapes & sampling sites
The study was conducted in the south-eastern portion of the INS region in New South
Wales, from Tottenham in the east to Nymagee in the west, and from Quanda Nature
Reserve in the north to ~20km south of Bobadah in the south. This region was selected
because satellite imagery suggested that boundaries between vegetation states were more
distinct than they are further north and west due to higher levels of agricultural intensification.
Thus, it was more straightforward to select study landscapes that varied in proportions of
different vegetation states and to assess the value of different vegetation states for avian
richness and diversity.
To select specific study landscapes, we first needed to identify the most common
vegetation states in the study region. Using a combination of SPOT5 satellite imagery and
ground-truthing, we identified four primary vegetation states that differ largely based on
apparent density of woody vegetation (Figure 2). They were:
Agriculture with No Trees (ANT) – In this vegetation state, the ground layer is used for
agricultural production and there are either no trees or just a few widely scattered trees. The
ground layer is often perennial native pasture that is occasionally cropped and thus may be
in a pasture phase or a crop phase. Both phases are included in the same vegetation state
because cropping tends to influence the characteristics of the native pastures that return
after cropping, and we found very few native pastures that did not show evidence of
cropping.
Agriculture With Trees (AWT) – In this vegetation state, the ground layer is used for
agricultural production as above, but there are a number of trees throughout a paddock.
These areas may sometimes be referred to as open woodland, but they are highly modified
for production and much simpler than a true open woodland. For example, there may be
only one species of tree, like a kurrajong, specifically selected to remain in the pasture

19
because it provides an agricultural benefit. Overstory cover values were much lower than
benchmark values for open woodland, and cropping of the ground layer at most sites meant
that it was very low in diversity and often cover of native ground plants.
Open Scrubland (OSL) – Open Scrubland consists of open shrubland or shrubby woodland
(shrubs with a scattered overstory of trees), where clusters of trees and/or shrubs are
evident but are distinctly separated by open ground. The open areas should support a
grassy ground layer, but many do not due to long-term management history. Most areas of
OSL are classified as INS.
Closed Scrubland (CSL) – In contrast, Closed Scrubland consists of dense shrubland
and/or shrubby woodland, where the tree and shrub cover appears more uniform, without
obvious open areas. Dense regeneration of some tree species, like white cypress pine,
tends to fall into this vegetation state. Mallee is also a type of Closed Scrubland, though it is
not an INS community. Except for Mallee, most areas of CSL are dominated by INS species.
In addition to these four common vegetation states, open woodland (OWL) is
considered an important and desirable vegetation state in the region (Figure 2). Aside from
the highly modified AWT version, it was too uncommon for us to include it in detail in this
study. In this part of the INS region, open woodland with overstory cover relatively close to
benchmark values and a ground layer unmodified by cropping appears to mostly occur on
small patches of fluvial soils and is almost completely dominated by poplar box.

20
a) ANT b) AWT
c) OSL d) CSL
e) OWL
Figure 2. Examples of the main structural vegetation states currently found in the south-
eastern Cobar Peneplain: a) Agriculture No Trees (ANT), b) Agriculture With Trees (AWT),
c) Open Scrubland (OSL), d) Closed Scrubland (CSL) and e) Open Woodland (OWL). Note
that very sparse scattered trees may still be present in ANT, but not at the densities seen in
AWT, where canopy density may begin to approach that of an open woodland.

21
We selected 18 study landscapes of 500ha each that differed in the proportions of the
four primary vegetation states (Figure 3). The 500ha landscape size was chosen because it
was suggested that the majority of landholders in the area choose to manage their properties
as a series of blocks of approximately that size. Recognising that this is not necessarily a
biologically relevant landscape size, we attempted to select 500ha areas where the
proportions of vegetation states were similar both inside the 500ha as well as outside, thus
minimising the influence of the landscape size per se. Study landscapes were selected
based on their vegetation features only, not based on property boundaries, so most
landscapes were owned by multiple landholders.
The design could not be perfectly balanced because there was insufficient natural
variation in the region. Ideally, we would have selected landscapes that each contained all
four vegetation states, but in practice it was difficult to find ANT and AWT in the same
landscape. Landholders either chose to leave trees in their paddocks or they didn’t, rather
than having one paddock with trees and another without. Thus, only three study landscapes
contained all four vegetation states. Four study landscapes contained AWT, OSL and CSL,
and 11 study landscapes contained ANT, OSL and CSL. Based on qualitative evaluation of
satellite images, approximately half the landscapes appeared to contain >50% open areas
and half appeared to contain >50% scrub areas (considering both OSL and CSL).
Approximately half the landscapes contained more OSL than CSL, and half the landscapes
contained more CSL than OSL.
Blocks of a single vegetation state within landscapes were termed “habitat areas”.
Within each habitat area in each of the 18 landscapes we established four sampling sites
(n=236 sampling sites). These were randomly located, with the restrictions that they needed
to be at least 150m from another sampling point, they should be at least 100m from the edge
of the vegetation state, and they could not be adjacent to a source of water such as a farm
dam. There were some landscapes where sampling points occasionally needed to be closer
to the edge in scrub vegetation states because of the small size of the patch of scrub, and
one landscape where sites in scrub vegetation states needed to be closer to each other
because of the extremely small size of the scrub patches. These sampling sites were where
all vegetation and bird data were collected.
Despite its rarity, to get a sense of whether open woodland (with overstory cover
values closer to benchmark and a ground layer unmodified by cropping) is particularly
important to native birds in the region, we also included two open woodland habitat areas in
the study, each less than 200ha in size, and located four sampling sites within each one.
These were not considered as part of replicate landscapes – they were included simply to
investigate the value of the vegetation state itself.

22
a)
b)
Figure 3. a) Locations of the 18 500ha study landscapes, labelled with the numbers we
used to identify them. They are overlaid on SPOT5 satellite imagery and a cadastral layer,
showing block boundaries and roads. Locality names have been added where appropriate.
Tottenham is ~20km east of Landscape 21.
0 10km
N
Nyngan
Bobadah
Nymagee

23
b)
Figure 3. b) Satellite image of Landscape 6, showing the clear distinction between “habitat
areas” containing ANT, OSL and CSL. One side of the landscape boundary (in yellow) is
~2.2km long.

24
1.3.3 Quantifying vegetation
Vegetation states were identified based on qualitative differences in overall woody
vegetation density, but it was clear that these states should be defined quantitatively and
might differ in ways other than woody density. In addition, there could still be substantial
continuous variation in some vegetation characteristics, leading to variation within vegetation
states. Density of vegetation in the ground layer, mid-story or shrub layer, and overstory or
tree layer could vary. The height of the shrub and tree layers could be variable, as well as
the Broad Vegetation Type (BVT; sensu DEC 2006). Even within BVTs, the dominant shrub
or tree species at any given site could vary. Birds could potentially respond to these sources
of variation and thus we needed to measure them and take them into account in our
analyses. In other words, birds could be viewing the landscape as one with relatively
continuous variation in vegetation variables, rather than a series of vegetation states.
Thus, we assessed vegetation characteristics at all sampling sites. Vegetation surveys
were performed once at each site in April 2008 to represent a mid-point in time between the
two sets of bird surveys (see below), and consisted of all variables listed in Table 1. The
only variable likely to be different between the vegetation sampling and bird sampling times
was % Basal Area Cover – Ground, but relative differences across sites were similar within
all time periods, so the April 2008 sampling was considered sufficient for assessing relative
differences among landscapes. Note that the “shrub” layer in these surveys was a mid-story
layer which could be present or absent independent of other layers, and could consist not
just of shrub species but also of small densely regenerating trees. This broadly
corresponded to the vegetation layer considered problematic by landholders and in INS
legislation. The overstory layer corresponded to a mature tree layer, which could also be
present or absent independent of other layers, though in practice very few sites that
contained a mid-story layer lacked an overstory layer.
Ground layer variables were estimated visually for each of the four quadrants of a 20m
x 20m plot centred on the sampling site, then herbaceous plant cover estimates were
averaged over these four quadrants to yield the overall estimate for the site. Ground
herbaceous cover was based on the percentage of area occupied by the live basal area of all
plants rather than the area occupied by foliage as this is highly variable in response to
season and rainfall. Percent crown cover of the shrub layer was estimated using a bitterlich
gauge (Mueller-Dombois and Ellenberg 1974, Friedel and Chewings 1988). We also
quantified the percent projected foliage cover of the shrub layer. We did this by selecting an
average shrub of the dominant shrub species at the site and visually estimating its percent
foliage cover (including stems and branches as well as leaves) by looking directly up through
the branches and comparing to a set of printed standards. That value was then multiplied by
the percent crown cover of shrubs to yield the overall percent projected foliage cover of the
shrub layer at each site. Percent projected foliage cover of the overstory was estimated
every 10m along a 40m transect, laid in a random direction, by looking directly up through
the branches and comparing to a set of printed standards, then averaging the five estimates.
Heights of both shrub and overstory layers were estimated to the nearest 0.5m by measuring
the table height of these layers using a clinometer.

25
Table 1. Vegetation data collected at each sampling site.
Variable name Description
Ground Type evidence of cropping or not, based on presence of
crop seed heads, stubble or ploughed furrows
% Basal Area Cover – Ground Plants percent cover of basal areas of live grasses and forbs
% Crown Cover – Shrub Layer percent crown cover of shrubs and/or young trees
% Foliage Cover – Shrub Layer projected foliage cover (including leaves, stems and
branches) of shrubs and/or young trees
Height – Shrub Layer table height of shrub layer
% Foliage Cover - Overstory projected foliage cover (including leaves, stems and
branches) of the overstory of mature trees
Height - Overstory table height of overstory layer
Keith Class general ecosystem type or community type, based on
Keith (2004)
BVT Broad Vegetation Type (e.g., poplar box woodland),
nested within Keith Classes, based on DEC (2006)
Dominant 1, 2, 3 – Shrub Layer first, second and third most dominant species group
found in the shrub layer
Dominant 1, 2, 3 – Overstory Layer first, second and third most dominant species group
found in the overstory layer
Keith Class and BVT were determined by comparing species present in the site and
general soil characteristics and topography with published descriptions (Keith 2004, DEC
2006). We added an additional BVT, 20M, to indicate areas dominated by mallee that were
in the Western Peneplain Woodlands Keith Class. As noted in DEC (2006), these generally
occur in association with BVT 20 and are officially placed within that BVT. However, we
found they were fairly common in the study region and we wanted to be able to distinguish
them from the standard BVT 20, which is dominated by a mixture of grey box, white cypress
pine, and poplar box. The first and sometimes second and third dominant species group at a
site were estimated visually for both shrub and overstory layers. Species groups were either
single species that were common in the landscape (e.g., wilga or poplar box), closely related
groups of species (e.g., budda and other tree-like species of Eremophila), or simply other
trees or shrubs for the least common species in the landscape (e.g., Dwyer’s red gum or
belah). Finally a photo was also taken in a random direction at each site for later reference.
1.3.4 Bird surveys
Birds were surveyed at all sampling sites (a total of 236 in the 18 study landscapes and
the two additional open woodland habitat areas) during both October 2007 and September
2008. These months were selected to focus on the height of the breeding season. The
October surveys ensured we sampled birds that migrate to the area and breed relatively late
in spring, while the September surveys were conducted while temperatures were cooler and

26
more birds were active for longer during the morning. The two years also provided the
opportunity for an interesting comparison. The 2007 surveys were performed during drought
in the region – only two of the 12 months prior to our sampling period experienced above-
average rainfall according to Nyngan rainfall data from the Bureau of Meteorology. In
contrast, the 2008 surveys occurred after plentiful summer rains at the end of 2007, though
follow-up precipitation was limited, making it more an “average” year – five of the 12 months
prior to our sampling period experienced above-average rainfall according to Nyngan rainfall
data.
At each sampling site during each survey year, we performed two point counts to
assess the presence and relative abundance of birds. Thus, each sampling site was
surveyed a total of four times (so each vegetation state within each landscape was surveyed
16 times), yielding a total of 944 point counts. At each sampling site in each year, one count
was performed in the early morning and one in the late morning, each by different pairs of
observers and separated by about 5 days. All observers had extensive experience
identifying birds and were specifically trained in identification of local INS region birds by both
sight and sound. Point counts lasted 10 minutes, during which the presence of species was
noted within 50m of the sampling site (denoted ‘within 50m’ throughout the rest of this
report). In addition, we noted the presence of species both during the point count and in the
approach to sampling sites regardless of their distance from the sampling site, as long as
they were in the appropriate vegetation state (denoted ‘infinite distance’ throughout the rest
of this report). The number of individuals (abundance) of all birds seen or heard was also
noted within 50m of the sampling location during the 10 minute point counts.
1.3.5 Assessing edge effects
One way in which landscapes could function as more than a sum of their parts is if
ecotones, or edges between vegetation states, provide unique resources, either by
supporting more species or a greater abundance of species than the adjacent vegetation
states. Casual observations in the INS region suggest that boundaries between open areas
and woody areas may be particularly species rich. While this was not part of the original
project plan, we realised during the 2007 survey period that it would be valuable to collect
even a small amount of data to compare the richness of edge sites with that of their
component vegetation states. Thus, we conducted a few additional point counts at “edge
trio” sites, where one point count was performed in an agricultural vegetation state (either
ANT or AWT), one was performed in the adjacent scrubland vegetation state (either OSL or
CSL), and the third was performed at the edge of the two vegetation states. The points were
arranged linearly and were separated by 150m. Eight of these comparisons were made in
four different landscapes, all in 2007. If edges are particularly important, we might expect
the edge sites in these trios to have the greatest species richness and the greatest number
of individuals.
Finally, in 2007 we also conducted ‘behavioural follows’ at the edges between
vegetation states, to identify whether differential use of vegetation states for different
purposes could be at least partially responsible for any positive edge effects. We followed

27
individual birds of 37 different species for as long as possible, up to 15 minutes, and every
20 seconds we recorded their behaviour and their distance from the habitat edge – a
negative distance if they were in a scrub vegetation state and a positive distance if they were
in an agricultural vegetation state. Behaviours were classified as either alert, foraging,
preening, calling, flying, caring for young (feeding nestlings or fledglings or incubating),
resting, aggression (chasing or being chased) or unknown. We performed 120 of these
follows at all types of scrub/agriculture edges, between ANT and OSL, between ANT and
CSL, between AWT and OSL and between AWT and CSL, generating 2380 behavioural
observations. If birds that are living at edges between vegetation states are using the
different states for different purposes, we might expect the behaviours that we recorded to
differ between positive and negative distances from edge, possibly with more foraging
behaviours in the agricultural states and more resting or reproductive behaviours in the scrub
states.
1.3.6 GIS analyses of landscape proportions
For analyses of bird species richness and diversity in mosaic landscapes, we needed
to compare bird variables in landscapes with different compositions – different proportions of
the four primary vegetation states. To calculate these proportions, we used a GIS-based
approach. First, we developed a woody-nonwoody layer for the study region using SPOT5
remotely sensed imagery. Imagery for all our study landscapes was located on six different
SPOT5 tiles, with all imagery taken between late 2004 and early 2005. For all six tiles, we
used an object-oriented GIS (Definiens Developer) to utilise both the high spatial resolution
of the natural-colour data and the increased radiometric resolution of the multispectral data in
a single classification. This results in a woody-nonwoody layer that correctly classifies
relatively small areas of woody vegetation, such as most individual paddock trees.
We then developed a series of spatial layers that defined each vegetation state. A GIS
focal statistic ‘sum’ function was used to calculate the total number of woody vegetation cells
within 25m, 50m, 75m and 100m radius circular neighbourhoods. To calculate the density of
woody vegetation within each search area the output datasets were divided by the area of
their associated neighbourhood. The outputs were then multiplied by 100, generating for
each cell the percentage of vegetation within the user-defined neighbourhood. After which,
the output data were reclassified so that cells with the maximum percentage cover were
allocated a value of 100, while cells with the minimum percentage of vegetation cover were
allocated a value of 0. The final step involved identifying the threshold values within the
vegetation density layers for which each vegetation state could be best defined. ArcGIS and
the Python scripting language were used to implement each neighbourhood density
calculation as a spatial data layer. Automation of the process via Python was important as
some fine-tuning of the neighbourhood parameters was required before the optimal settings
for each vegetation state were identified. Table 2 lists the final neighbourhood sizes and
threshold values used to delineate each vegetation state.
The four primary vegetation states were then combined to form a single four-band
spatial layer of vegetation state. This layer was converted from a raster dataset to a feature

28
layer and intersected with polygon shapefiles delineating the study landscapes. The area of
each polygon of each vegetation state was then calculated and these values were
aggregated for each vegetation state within each landscape and divided by the total area of
landscapes to generate estimated proportions of vegetation states in each landscape. As
the object-oriented GIS occasionally classified areas of bare soil or depressions within
agricultural land or OSL as “woody” areas, this approach tended to overpredict the amount of
AWT in landscapes. Thus, we further corrected our estimates by reducing the amount of
AWT by 50% and allocating the additional area to either ANT or OSL, depending on whether
the overpredicted areas were contained within ANT or within OSL.
Table 2. Neighbourhood and threshold parameters used to map vegetation state.
Vegetation State Density Neighbourhood Threshold Value
ANT Circular neighbourhood with 50m radius <5
AWT Circular neighbourhood with 75m radius 2-20
OSL Circular neighbourhood with 25m radius 5-85
CSL Circular neighbourhood with 75m radius 50-100
1.3.7 Statistical analyses
Calculating richness, diversity & bird community composition: Bird survey data
were combined from the two years of the study, as analyses comparing data collected in
2007 vs. 2008 revealed no differences in responses to vegetation state (see 2.4.7 Inferring
effects of drought vs. land management). Point count data were used to calculate avian
species richness and a diversity index representing evenness, as well as to determine avian
community composition. These variables were calculated at three different hierarchically
structured spatial scales: sampling sites, vegetation states within a landscape (‘habitat
areas’), and landscapes. At each scale, abundance data (within 50m) from multiple point
counts (and multiple sampling sites in the cases of the two larger scales) were combined
additively. Species richness was calculated as the total number of different species
recorded, both strictly within the 50m point count radius (within 50m data) and also including
birds observed within the vegetation state but further from the sampling site than 50m
(infinite distance data). Community composition was determined based on the list (and
number of individuals) of species recorded within the 50m point count radii. Finally, diversity
was calculated from the within 50m data using the Shannon diversity index, H (Shannon
1948), hereafter referred to as ‘diversity’. This is a measure of evenness of abundance
across species that also takes into account the total number of species observed. It is
calculated as:
H = -Σ pi*ln(pi), for all i = 1 to S
where pi is the proportion of all individuals observed that belong to the ith species, and where
S is the total number of different species. Species richness measures included data for all

29
species, while diversity and community composition excluded woodswallows, as they forage
aerially in extremely large groups, making it difficult to assign them accurately to a given
sampling site and making their abundance within a 50m radius vary by orders of magnitude
due purely to the chance event of all or part of the local flock flying over at the time of
sampling.
Physical differences between vegetation states: We explored quantitative
differences between vegetation states, to investigate the particular vegetation features that
distinguished our vegetation states, and to identify potentially important sources of
continuous variation. We compared quantitative vegetation variables at the scale of
sampling sites using mixed-effects analysis of variance (ANOVA). This allowed us to model
the relationships between vegetation variables and vegetation states but also to include
landscape as a random factor, and thus control for lack of independence among data
collected at different sampling sites but within the same landscape. However, these mixed
effects models compare each vegetation state to all others combined, rather than performing
more traditional paired comparisons. Thus, we also performed standard ANOVA with
Tukey’s post-hoc paired comparisons, ignoring the lack of independence in the data, to
investigate how each vegetation state differed from each of the others. For qualitative
vegetation variables such as Broad Vegetation Types, we used summary statistics to
compare the different vegetation states.
Influence of vegetation states on birds: To analyse whether vegetation states
differed in bird species richness and diversity, we used mixed-effects ANOVAs, including
landscape as a random factor to control for lack of independence among data collected in
different vegetation states that were within the same landscape. We also performed
standard ANOVAs with Tukey’s post-hoc paired comparisons, ignoring the lack of
independence in the data, to investigate how each vegetation state differed from each of the
others. We analysed species richness based on both the infinite distance data and on data
collected within 50m of each sampling point. Both richness and diversity were calculated by
pooling data across all point counts performed in each habitat area.
We also examined whether different vegetation states tended to support different
communities of species using ordination analyses. Specifically, we used detrended
correspondence analysis (DCA) to analyse similarity in bird communities (based on species
composition and abundances) across habitat areas. The results are a graph in which habitat
areas are points and the two axes correspond to two types of underlying environmental
variation which is not measured or defined – it is presumed based on the similarities and
differences in birds among habitat areas (i.e., they are latent variables). Habitat areas that
appear closer together in the resulting diagrams are more similar to each other
compositionally than those that appear farther apart. We drew polygons around the habitat
areas for each vegetation state and visually evaluated how similar or different the
composition of vegetation states was based on the amount of overlap of polygons. Very little
overlap indicates that the compositions of bird communities are very different in different
vegetation states. We repeated DCA excluding species for which less than 20 individuals
were observed in total (across all landscapes and both years) to examine patterns for
common species only.

30
As an alternative approach, we analysed whether vegetation characteristics influenced
species richness (within 50m) and diversity in a more continuous fashion, without assuming
that the landscape is organised into distinct vegetation states. We used mixed-effects
ANOVA to model the presence or absence of trees, the presence or absence of shrubs,
ground type, and all quantitative vegetation variables (see Table 1) on richness and diversity
at all sampling sites (the scale at which vegetation variables were measured). The only
variable that we excluded was “% cover – scrub”, which was strongly correlated with “%
foliage cover – scrub”. We used a backwards stepwise procedure to create a final model,
sequentially eliminating variables with p > 0.05. The influence of qualitative variables (BVTs,
dominant scrub species and dominant tree species) on richness and diversity were modelled
using mixed-effects ANOVA but only for sampling sites within scrub vegetation states, as
these parameters were insufficiently variable in the other vegetation states.
Influence of landscapes on birds: Most importantly, we were interested in whether
landscape characteristics influenced bird species richness and diversity at both landscape
and sampling site scales. We used generalised linear models (GLMs) to analyse the
influence of the proportion of each vegetation state in each landscape on richness and
diversity of the landscape. We analysed species richness based on both the infinite distance
data and on data collected within 50m of each sampling point. Both richness and diversity
were calculated by pooling data across all point counts performed in each landscape. To
specifically assess whether the combination of vegetation states was important, rather than
just the amount of individual vegetation states, we also analysed the influence of habitat
diversity of each landscape, computed as The Shannon diversity index but using proportions
of each vegetation state in the landscape to represent each value of pi.
We also examined whether landscapes that differed in composition also differed in the
community composition of birds they supported. Again, we used DCA to arrange landscapes
in two dimensions (thought to represent two types of undefined environmental variation)
based on bird species composition (presence and abundances). We drew polygons around
landscapes with different vegetation compositions (e.g., with >50% agricultural vegetation
states vs. >50% scrub vegetation states) and visually evaluated how similar or different the
bird composition of landscapes was based on the amount of overlap of polygons. Very little
overlap indicates that the compositions of bird communities are very different in landscapes
dominated by different vegetation states.
Finally, to most directly assess whether landscapes were more than a sum of their
parts, we evaluated whether there were interactions between vegetation states. We did this
by analysing whether the richness and diversity of birds at sampling sites within a given
vegetation state were actually dependent on the proportions of the other vegetation states in
the landscape. We used mixed-effects ANOVAs, and controlled for landscape identity (and
thus ways in which study landscapes differed other than in the proportions of vegetation
states) as well as vegetation characteristics of the sampling site that were already shown to
influence richness and diversity in previous analyses.

31
Additional analyses – open woodland, effects of edges, and comparing 2007 and
2008: Simple summary statistics were used to explore the role of open woodland in this
system, qualitatively comparing vegetation data and bird survey data collected in the two
open woodland habitat areas to those data from the other vegetation states. Insufficient
sample size of open woodland sites prohibited the use of more definitive statistical analysis.
To compare the value of edge habitats versus the interior of adjacent vegetation states,
we used standard ANOVA with Tukey’s post-hoc paired comparisons to evaluate differences
between agricultural, edge and scrub point counts at “edge trio” sites. Response variables
were species richness (infinite distance) and total abundance (within 50m) at each sampling
site. To evaluate whether birds at edges were engaging in different behaviours in the two
different vegetation states, we used a contingency analysis to compare observed and
expected frequencies of different behaviours depending on whether the birds were in
agricultural versus scrub vegetation states.
As bird survey data were collected in two years, one a drought year and one with
approximately average conditions, we also compared data between the two years. This
helped us determine that it was appropriate to combine survey data from the two years for all
the other analyses, but also helped us explore whether or not any of our observations and
results might be attributable to drought conditions in 2007. We used paired t-tests to
determine whether bird species richness and total abundance at the scale of habitat areas
(vegetation states within a landscape) were greater in 2008 than in 2007. We included open
woodland vegetation states in these analyses. Pearson correlation coefficients were then
calculated to examine the relationships between species richness and total abundance of
habitat areas in 2007 and 2008. Finally, we calculated the difference between the species
richness and abundance measures in 2007 and 2008 (2008 values minus 2007 values), and
used ANOVA with Tukey’s post-hoc paired comparisons to examine whether those
differences were related to vegetation states. In other words, we examined whether
differences between the two survey periods were more pronounced for any particular
vegetation state. If the 2007 drought conditions were responsible for reducing species
richness and diversity in grassy areas in particular, there should be a greater positive
difference between 2007 and 2008 bird data in the agricultural vegetation states compared to
the scrub vegetation states.
All ANOVAs, GLMs, t-tests, contingency analyses and correlations were performed in
SYSTAT 10 (SPSS Inc., Chicago, Illinois, USA). Ordination analyses of community
composition were performed using CANOCO for Windows 4.5 (ter Braak and Smilauer
2002).

32
1.4 Results
1.4.1 Descriptive Results: vegetation states
Evidence of recent cropping was found in over half of all ANT sampling sites (30 out of
56). Cropping was less prevalent in AWT sites (8 out of 28 sites), and not evident at any of
the OSL or CSL sampling sites.
The basal area of ground herbaceous cover differed significantly between the different
vegetation states (ANOVA: F = 3.10, df = 3, 224, p = 0.028) but varied much less among
vegetation states than any of the other variables characterising vegetation structure. Pair-
wise comparisons showed that only ANT and CSL differed significantly in percent live plant
cover at the ground layer (p = 0.014), with CSL containing less plant cover at the ground
layer. However, ground herbaceous cover was very low overall so the difference between
ANT and CSL was small, with average ANT basal area cover only 4.9% compared to 2.1% in
CSL (Figure 4).
0
2
4
6
8
10
12
14
ANT AWT OSL CSL
Vegetation state
% Basal Area Cover - Ground
Figure 4. Mean + SD of the percentage of ground covered by the basal areas of ground
herbaceous plants estimated at sampling sites according to the vegetation state of the
sampling site.

33
The different vegetation states differed markedly and predictably in both percent crown
cover (ANOVA: F = 72.68, df = 3, 224, p < 0.0005) and percent foliage cover (ANOVA: F =
68.03, df = 3, 224, p < 0.0005) of the shrub layer (Figure 5). For both of these measures of
shrub cover, pairwise comparisons indicated that each vegetation state differed significantly
from every other state (p < 0.0005) except that ANT and AWT did not differ (p = 0.997)
a)
0
5
10
15
20
25
30
35
40
45
50
ANT AWT OSL CSL
Vegetation state
% Crown Cover - Shrubs
b)
0
5
10
15
20
25
30
ANT AWT OSL CSL
Vegetation State
% Foliage Cover - Shrub Layer
Figure 5. Mean + SD of a) the percentage of shrub cover and b) the percentage of shrub
foliage cover, according to the vegetation state of the sampling site.

34
The different vegetation states also differed in terms of the height of the scrub layer
(ANOVA: F = 133.66, df = 3, 224, p <0.0005), though perhaps not surprisingly since this
height was zero for almost all ANT sites (Figure 6). However, pairwise comparisons
indicated that all pairs of vegetation states differed significantly from each other (p < 0.02)
except for OSL and CSL which were similar (p = 0.991).
0
1
2
3
4
5
6
7
8
9
ANT AWT OSL CSL
Vegetation state
Height - Shrub Laye
r
Figure 6. Mean + SD of the height of the shrub layer according to the vegetation state of the
sampling site.
The different vegetation states differed significantly in percent foliage cover of the
overstory (ANOVA: F = 76.53, df = 3, 224, p < 0.0005), again unsurprisingly since canopy
cover was zero at all ANT sites (Figure 7). Pairwise comparisons revealed that each
vegetation state differed significantly from all others except for two pairs – the difference
between ANT and AWT was not quite statistically significant (p = 0.065), while AWT and
OSL were the most similar (p = 0.10).

35
0
5
10
15
20
25
30
ANT AWT OSL CSL
Vegetation state
% Foliage Cover - Overstory
Figure 7. Mean + SD of the projected foliage cover of the overstory layer according to the
vegetation state of the sampling site.
Canopy height differed significantly across vegetation sites (ANOVA: F = 206.16, df =
3, 224, p < 0.0005), but only because this height was zero at all ANT sites (Figure 8).
Pairwise comparisons found that ANT differed significantly from all other vegetation states in
terms of overstory height, but the other vegetation states did not differ significantly from each
other (p > 0.5).
0
2
4
6
8
10
12
14
16
18
20
ANT AWT OSL CSL
Vegetation state
Height - Overstory
Figure 8. Mean + SD of the height of the overstory layer according to the vegetation state of
the sampling site.

36
There was considerable variation in the broad vegetation types (BVTs; Table 3)
represented by the four vegetation states (Figure 9). All sites within agricultural vegetation
states (ANT + AWT) belonged to BVT 1004 (cleared land), but scrub sites showed some
considerable variation. More than half of CSL sites were BVT 20M (the mallee form of
Western Peneplain woodland), while BVT 21 (poplar box woodland) was most common
among OSL sites. OSL sites were more variable with 6 different BVTs represented, while
CSL sites represented 4 different BVTs.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
ANT AWT OSL CSL
Vegetation state
20
21
22
23
20M
55
1004
Figure 9. Proportions of sampling sites within each vegetation state that were composed of
each of the BVTs in the region (see Table 3 for full description of BVTs).
Table 3. Broad Vegetation Types (BVTs) in the study region.
BVT# BVT (Broad Vegetation Type) name Keith Class
20 Grey Box/ White Cypress Pine/ Poplar Box/
Smooth-barked Coolibah (Red Box) on red earths
Western Peneplain Woodlands
21 Poplar Box woodland of the Cobar Peneplain Western Peneplain Woodlands
22 Derived native tall shrubland Western Peneplain Woodlands
23 Smooth-barked Coolibah (Red Box)/ White
Cypress Pine woodland of the Cobar Peneplain
Western Peneplain Woodlands
20M Mallee woodland of the Cobar Peneplain (but not
on rocky hills)
Western Peneplain Woodlands
55 Mallee woodland on rocky hills, mainly of the
Cobar Peneplain
Inland Rocky Hill Woodlands
1004 Cleared Cleared
Most agricultural sites lacked a shrub layer, but one ANT site and about a third of AWT
sites did have some shrubs. OSL showed the greatest diversity in dominant species in the

37
scrub layer (Figure 10). The most common dominant shrub species was wilga and this was
true in each vegetation state. Regenerating white cypress pine was the second most
common dominant “shrub” species and was almost as common a dominant as wilga at CSL
sites. Budda and turpentine were dominant species at some scrub sites and regenerating
mallee was the dominant species in the shrub layer at a number of CSL sites. Other shrub
species that were dominant at a small number of sites included a few species of Acacia and
punty bush. Other tree species that were dominant in the shrub layer at a few sites included
belah, poplar box, and rosewood.
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
ANT AWT OSL CSL
Vegetation state
wilga
budda
turpentine
other shrubs
cypress pine
mallee
other trees
none
Figure 10. Proportion of sampling sites in each vegetation state that had different dominant
species in the scrub layer.
Dominant overstory species varied between sites both within and across the different
vegetation states (Figure 11). Poplar box, red box, and white cypress pine were each fairly
common as dominant overstory species across vegetation types, though poplar box was
particular common as the dominant species at OSL sites. Kurrajong was the most common
dominant tree species at AWT site, but was never dominant at scrub sites. Mallee was the
dominant overstory species at over 60% of CSL sites and was also dominant at about 20%
of OSL sites. Other tree species that were dominant at a small number of sites included grey
box, Dwyer’s red gum, belah, Allocasuarina, and rosewood. At a few AWT and OSL sites,
mature wilgas were actually the dominant overstory species.

38
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
ANT AWT OSL CSL
Vegetation state
poplar box
red box
mallee
cypress pine
kurrajong
other trees
wilga
none
Figure 11. Proportion of sampling sites in each vegetation state that had different dominant
species in the overstory layer.
In summary, vegetation states primarily differed from each other in woody vegetation
density, represented by the percent cover and projected foliage cover of the scrub layer and
the projected foliage cover of the overstory layer. Thus, these quantitative assessments
provided support for the qualitative assignment of vegetation states at the start of the study.
While woody cover appeared to be negatively associated with ground herbaceous cover,
such that scrub states had lower levels of cover of live plants in the ground layer, the
difference was very small due to overall low levels of ground herbaceous cover.
There was still considerable variation within vegetation states in some parameters.
The scrub states OSL and CSL were variable in scrub and overstory cover, while ANT sites
and OSL sites were quite variable in amount of ground herbaceous plant cover. OSL sites
are particularly interesting. Given the high variability in both woody cover variables and
ground herbaceous cover, it would be logical to think these are correlated, and one simply
reflects variation in the other. However, within OSL sites, we found no significant
relationships between ground herbaceous cover and either foliage cover of the scrub layer,
foliage cover of the overstory, or total woody layer foliage cover (Regressions: all p>0.25).
Indeed, we observed several OSL sites where a healthy grass layer existed in the
“interspaces” between shrubs. While this probably depends on management of the ground
layer, it also probably depends to some extent on scale of patchiness and thus the size of
the interspaces to provide some freedom from competition. In addition, the OSL vegetation
state was the most variable in terms of composition, containing more BVTs and a greater
diversity of dominant shrub species. These variables may therefore need to be explored as
continuous sources of variation that could affect the richness and diversity of bird species.

39
1.4.2 Descriptive Results: birds in landscapes of the south-eastern Cobar
Peneplain
We observed only 10 agricultural specialists (defined as species found only in ANT and/or
AWT vegetation states) as compared with 42 scrub specialists (defined as species found
only in OSL, CSL, and/or OWL vegetation states). Most species (63) were observed in both
agricultural and scrub vegetation states (based on presence in habitat areas, not just within
50m point count radii; Table 4). The agricultural specialists that we did observe were present
at low abundances (an average of 1.6 individuals per habitat area) compared to relatively
high abundance of scrub specialists (an average of 7.7 individuals per habitat area). Table 4
provides information on a species-by-species basis, showing which vegetation states
species were observed in and how frequently they were observed.
These results are consistent with the fact that the majority of birds in the mosaic landscapes
of the south-eastern Cobar Peneplain could be classified as either canopy insectivores
(38.6% of species observed) or ground-foraging insectivores (34.2% of species observed).
Only 14.9% of species were ground-foraging granivores (seed-eaters). Canopy insectivores
would be expected to be scrub specialists, granivores would be expected to be agricultural
specialists (or possibly to prefer OSL), while ground-foraging insectivores might take
advantage of most vegetation states.
We determined the top 20 most abundant species (based on total number of individuals
observed within 50m point count radii), and the noisy miner was the most abundant by a
wide margin. Of these top 20 most abundant species, 12 were more commonly observed in
scrub habitats, 5 were more commonly observed in agricultural habitats, and 3 were equally
common in both.
We observed a number of species of special conservation concern. We observed one
species classified as “Vulnerable” on the EPBC Act List of Threatened Fauna, the superb
parrot, which was in both agricultural and scrub vegetation states, but most commonly at
AWT sites. We also observed 9 species that have been classified as “Vulnerable” on the
NSW Threatened Species List – 6 of these were more common in scrub habitats, 2 were
more common in agricultural habitats, and one was equally common in both. We observed
15 “declining species” as classified by Reid (1999), of which 12 were more commonly
observed in scrub habitats, one was more commonly observed in agricultural habitats, and
two were equally common in both. Note that these are species commonly referred to as the
“declining woodland birds”, yet in the INS region, they are primarily inhabitants of scrub
vegetation states. They were usually absent from open woodland and AWT sites (which
could be considered open woodlands highly modified for agriculture). We observed 23
different species that appear to be declining based on “Change maps” in the New Atlas of
Australian Birds (Barrett et al. 2003), and of these, 9 were more common in scrub habitat, 12
were more common in agricultural habitats, and 2 were equally common in both.

40
Table 4. All bird species observed at our study sites, with habitat specialists (species either
observed in agricultural vegetation states but never scrub states or in scrub vegetation states
but never in agricultural states) listed first, followed by habitat generalists (species observed
at both agricultural and scrub sites). The numbers indicate the number of landscapes in
which each species was observed in that vegetation state, although not all landscapes
contained all vegetation states. Numbers in parentheses in the column headings give the
number of landscapes containing each vegetation state.
Species ANT (14) AWT (7) OSL (18) CSL (18) OWL (2)
(a) habitat specialists
Brown Songlark * ‡ 5
Crimson Chat 1
Little Eagle * 1
Orange Chat ‡ 1
White-winged Fairywren * 1
Australian Pipit ‡ 8 1
Banded Lapwing ‡ 6 2
Black-shouldered Kite * 1 2
Spotted Harrier * ‡ 1 1
Stubble Quail * 1 1
Australian Wood Duck * 1
Black-faced Woodswallow * ‡ 1
Buff-rumped Thornbill 2
Shy Heathwren * 1
Striated Thornbill * 1
Variegated Fairy-wren 1
White-bellied Cuckoo-shrike * 1
White-eared Honeyeater * 1
White-necked Heron * ‡ 1
Zebra Finch * ‡ 1
Australian Owlet-nightjar * 2
Chestnut Quail-thrush + * 1
Gilbert's Whistler + * 2
Grey-fronted Honeyeater 1
Little Pied Cormorant * 1
Restless Flycatcher * # 1
Tawny Frogmouth * 1
Bar-shouldered Dove 10 9
Black Honeyeater * 3 1
Black-chinned Honeyeater + 1 1
Brown Goshawk * 4 1
Brown-headed Honeyeater 5 10
Crested Bellbird * # 3 8
Grey Fantail 9 12
Grey Shrike-thrush 13 15
Inland Thornbill 11 10 13
Mistletoebird 7 7
Peaceful Dove 3 4

41
Rainbow Bee-eater 2 6
Sacred Kingfisher * 1 1
Singing Honeyeater 3 1
Speckled Warbler + * # 1 1
Splendid Fairy-wren 10 6
Spotted Pardalote 5 2
Varied Sitella # 1 2
Western Gerygone 12 14
White-browed Babbler # 4 4
White-throated Gerygone 1 1
Yellow-plumed Honeyeater 3 10
Brown Treecreeper + # 5 3 1
Eastern Yellow Robin # 6 11 1
Laughing Kookaburra 3 2 1
(a) habitat generalists
Apostlebird 3 7 5 17 12 2
Australian Hobby 3 1 1
Australian Kestrel ‡ 8 3 1 1
Australian Magpie 14 7 18 18 2
Australian Raven 11 7 17 13 2
Australian Ringneck 5 11 7 18 14 2
Black Kite * 1 1
Black-eared Cuckoo 1 2 1
Black-faced Cuckoo-shrike 7 3 9 8
Blue Bonnet ‡ 6 14 7 17 8 2
Blue-faced Honeyeater 3 3 9 3 2
Brown Falcon ‡ 5 1 2 2
Chestnut-rumped Thornbill # 1 10 10
Cockatiel 4 14 7 16 14 2
Common Bronzewing 4 1 13 10 2
Common Starling 7 4 4 2
Crested Pigeon 12 11 7 18 13 2
Eastern Rosella 1 3 1
Emu # ‡ 6 1 2
Galah 9 14 7 15 17 2
Grey Butcherbird 8 6 18 18 2
Grey-crowned Babbler + # 14 4 4 17 12 2
Ground Cuckoo-shrike 1 2 1
Hooded Robin + # ‡ 1 1 1
Horsfield's Bronze-cuckoo 1 6 4
Jacky Winter ‡ # 1 2 9 9
Little Corella 1 1 1
Little Friarbird 2 1 11 9
Little Raven 16 13 7 17 12 2
Magpie-lark 13 7 14 14 2
Major Mitchell's Cockatoo + ‡ 6 4 3 6 1
Masked Woodswallow ‡ 5 2 9 11
Mulga Parrot 1 4 1
Noisy Friarbird 1 2 6 1

42
Noisy Miner 1 14 7 18 18 2
Olive-backed Oriole * 1 1 1
Pallid Cuckoo 1 1 2
Pied Butcherbird 11 6 17 14 2
Red-backed Kingfisher * ‡ 1 1 1
Red-capped Robin # 17 1 1 12 14
Red-rumped Parrot 7 12 7 14 8 1
Rufous Songlark ‡ 1 3 2 2
Rufous Whistler # 19 1 16 15
Southern Whiteface # 2 3 10 1
Spiny-cheeked Honeyeater 15 4 5 13 14
Spotted Bowerbird 3 8
Striated Pardalote 13 4 7 17 18
Striped Honeyeater 1 2 12 12
Superb Parrot ^ + 20 1 4 5 3 2
Tree Martin 2 1
Wedge-tailed Eagle ‡ 3 4 2
Weebill 2 1 14 18 1
Welcome Swallow 2 1
White-breasted Woodswallow 2 1 1
White-browed Woodswallow # ‡ 5 3 11 13
White-faced Heron * ‡ 1 1
White-plumed Honeyeater 8 1 2 6
White-winged Chough ‡ 3 1 10 7 2
White-winged Triller ‡ 3 1 8 6
Willie Wagtail 5 4 13 12
Yellow Thornbill 10 1 15 16
Yellow-rumped Thornbill 18 3 4 14 4
Yellow-throated Miner 2 1 3 1 1
^ Classified as Vulnerable on the EPBC Act List of Threatened Fauna
+ Classified as Vulnerable on the NSW Threatened Species List
* Species seen in 2008 but not in 2007
# Declining species as classified by Reid (1999)
‡ Species that were classified as “less common”, based on “Change maps” in the Birds
Australia New Atlas of Australian Birds (Barrett et al. 2003)
1 Rank of abundance. 1 = most abundant, 20 = 20th most abundant
1.4.3 Statistical Results: effects of vegetation states on birds
Regardless of whether we calculated species richness in habitat areas based on
aggregating the more liberal infinite distance data or the more conservative within 50m data
from the eight point counts performed in each habitat area, mixed effects ANOVAs revealed
that the vegetation state of habitat areas significantly influenced the number of species they
supported. Using the infinite distance data, only AWT sites had richness values that were
not significantly different from the mean across all vegetation states (AWT p = 0.189, all

43
other p < 0.01, model log likelihood = -200.1). Using the within 50m data, all vegetation
states had richness values that were significantly different from the global mean (all p <
0.026, model log likelihood = -171.4).
However in both models, the effect of landscape identity was not significant,
suggesting that vegetation states had similar influences on bird species, regardless of
differences in the amount or composition of those states among our study landscapes. Thus,
standard fixed effects ANOVAs were also appropriate for these data, and confirmed that the
vegetation state of habitat areas significantly influenced species richness (infinite distance
data: F = 12.75, df = 3, 53, p<0.001, R2 = 0.419; within 50m data: F = 34.58, df = 3, 53,
p<0.001, R2 = 0.662). Considering the infinite distance data, pairwise comparisons revealed
that ANT was significantly species poor compared to each of the scrub vegetation states
(both p < 0.001), while AWT was significantly less rich than OSL (p = 0.02), the most
species-rich state (Figure 12a). The two agricultural states and the two scrub states were
not significantly different from each other (both p > 0.4). Considering the within 50m data,
pairwise comparisons revealed that while the two scrub states were still not significantly
different from each other (p = 0.22), there was a trend for AWT to be more species rich than
ANT (p = 0.067). In addition, all other pairwise comparisons were significant (all p < 0.006),
with increasing numbers of species found in ANT, AWT and scrub vegetation states (Figure
12b).

44
a)
b)
Figure 12. Mean + SD of the number of species observed in habitat areas consisting of the
four primary vegetation states using a) the data collected at infinite distances from sampling
points and b) the data collected only within 50m of sampling points.
The same patterns were observed when comparing The Shannon diversity index
among vegetation states. In the case of the diversity data, the mixed effects model could
not be fit due to insufficient variability in diversity values within some landscapes. However,
standard fixed effects ANOVA revealed that vegetation states differed significantly in bird
species diversity (F = 25.58, df = 3, 53, p<0.001, R2 = 0.592). According to the pairwise
comparisons, ANT was significantly less diverse than all other vegetation states (all p <
0.004), while AWT was less diverse than OSL (p = 0.035). The two scrub states were not
0
10
20
30
40
50
60
ANT AWT OSL CSL
Vegetation state
Species Richness
0
5
10
15
20
25
30
35
ANT AWT OSL CSL
Vegetation state
Species Richness (<50m
)

45
significantly different from each other (p = 0.436). Thus, increasing diversity was found in
ANT, AWT and scrub vegetation states (Figure 13).
Figure 13. Mean + SD of species diversity (calculated using The Shannon diversity index) in
habitat areas consisting of the four primary vegetation states.
Habitat areas with different vegetation states also differed in the composition of the bird
communities they supported, based on DCA analyses (Figure 14a). All four vegetation
states supported different communities, though some OSL habitat areas were quite similar to
some CSL habitat areas. There was one noticeable outlier, habitat area #7, which was an
ANT habitat area that had very recently (within months) been cleared of scrub at the time of
the 2007 surveys. Unique species including hooded robin were detected in this habitat area
only during the 2007 surveys, not during 2008. It was surprising that AWT habitat areas
supported such a distinctly different community than ANT habitat areas, though Table 4 does
reveal a number of species that were present in one agricultural vegetation state but not the
other. In general, AWT sites were more likely to occasionally contain species normally found
in scrub vegetation states.
When we performed DCA on only the more common species in the system, we found
that the differences between OSL and CSL remained, but AWT habitat areas generally
represented a subset of the variation in community composition found across ANT habitat
areas (Figure 14b). In other words, OSL and CSL differed in the communities they
supported not just in terms of rare species but in terms of common species as well. In
contrast, AWT habitat areas contained unique communities mostly because they supported a
few rare species in these landscapes, such as the ground cuckoo-shrike, tree martin, and
both the black-eared and the pallid cuckoo. ANT always supported different communities
than the scrub vegetation states.
0
0.5
1
1.5
2
2.5
3
3.5
ANT AWT OSL CSL
Vegetation state
Species Diversity (<50m
)

46
a)
b)
Figure 14. Ordination diagrams from Detrended Correspondence Analysis. Habitat areas
that were more similar in terms of bird species composition (using presence and abundance
data) appear closer together. All habitat areas consisting of each vegetation state are
enclosed by coloured polygons, and greater overlap between polygons indicates greater
similarity in bird communities: a) all species observed, b) only common species, for which at
least 20 individuals were observed throughout the course of the study. Black=ANT habitat
areas, Purple=AWT habitat areas, Green=OSL habitat areas, Yellow=CSL habitat areas.

47
When we analysed continuous variation in vegetation characteristics rather than
vegetation states (at the scale of sampling sites rather than aggregating up to habitat areas),
we found that the coarse structural variables that separated vegetation states were still the
most significant predictors of both bird species richness (within 50m) and diversity. Mixed-
effects ANOVA with backward stepwise variable selection revealed that the presence of
trees (p = 0.001), the presence of scrub (p < 0.001), and the % foliage cover of scrub (p =
0.001; log likelihood of model = -601.95) were positively related to species richness (Figure
15a) at sampling sites. Furthermore, the random effect of landscape identity also explained
a significant proportion of the variation in richness (p = 0.029), suggesting that there were
other landscape-level variables that were important in addition to coarse structural
characteristics of sampling sites. The same overall result was obtained when we modelled
diversity instead of species richness (presence of trees p < 0.001; presence of scrub p =
0.002; % foliage cover of scrub p = 0.002; log likelihood of model = -173.86; Figure 15b),
though there was only a weak trend for the random effect of landscape identity to explain
variation in diversity values across sampling sites (p = 0.118).
a)
b)
Figure 15. Scatter plots showing the significantly positive relationships between % foliage
cover of scrub and a) bird species richness (within 50m) and b) diversity of sampling sites.
0
5
10
15
20
25
0 10203040506070
% Foliage - Shrub Layer
Species Richness (<50m)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
0 10203040506070
% Foliage- Shrub Layer
Species Diversity (<50m)

48
In addition to the strong influence of coarse vegetation structure on birds in our study
landscapes, there was one influence of vegetation composition on species richness within
scrub vegetation states. In a mixed-effects ANOVA controlling for differences among
landscapes, species richness differed significantly between sampling sites with different
dominant overstory species. Specifically, at sites in which mallee was the dominant species
group in the overstory, bird species richness was significantly reduced compared to average
richness across all sites (p = 0.023; Figure 16). However, much like ANT contained reduced
richness compared to other vegetation states but supported a different community, mallee
sites also supported a number of species found nowhere else, including the chestnut quail-
thrush, Gilbert’s whistler, and the grey-fronted honeyeater. Mixed-effects ANOVAs revealed
no significant effects of BVTs or dominant scrub species on bird richness or diversity (all p >
0.06 and all but one > 0.2).
Figure 16. Mean + SD of bird species richness (within 50m) of sampling sites in scrub
vegetation states according to the most dominant vegetation species in the overstory.
Mallee is significantly lower than the average across all species groups.
1.4.4 Role of Open Woodland
Open woodland (OWL) in a relatively unmodified state was not present in any of our 18
study landscapes, but we did collect data at a total of eight sampling sites in two OWL
habitat areas in the region. Given this relatively small number of sites, variability in basic
measures of vegetation structure was high, limiting the possibility for statistical comparisons
with other vegetation states. In addition, both of these OWL habitat areas were on fluvial
soils in the vicinity of creeks (dry at the time of surveys), which may to some extent confound
comparisons with sites in other vegetation states (none of which were on fluvial soils).
Nonetheless, we were able to make general comparisons by calculating mean vegetation
0
2
4
6
8
10
12
14
poplar
box grey box red box mallee callitris other
trees wilga
Dominant Overstory Species
Species Richness (<50m)

49
and bird variables at OWL sampling sites and assessing which of the other vegetation states
OWL was most similar to (Table 5).
Table 5. Mean values of vegetation variables, avian species richness and avian diversity in
open woodland (OWL) sites, compared to these same values in the other four vegetation
states.
Variable Mean  SD for OWL sites Relative to other vegetation states
%Basal Area - ground 16.1 ± 22.8 higher than all others
%Crown cover - shrubs 1.9 ± 2.2 slightly higher than AWT
%Foliage cover - shrubs 1.2 ± 1.4 slightly higher than AWT
Height - shrubs 2.5 ± 3.4 between AWT and OSL/CSL
%Foliage cover -
overstory
18.0 ± 10.0 similar to CSL
Height - overstory 17.1 ± 1.9 higher than all others
Bird species richness 5.1 ± 2.8 similar to AWT
Bird species diversity 1.10 ± 0.54 similar to AWT
The primary way in which OWL sites were clearly different from other vegetation states
was in terms of ground herbaceous cover. OWL sites had an average of over 16% basal
area cover of live plants, much higher than any of the other vegetation states we examined.
However, this difference was entirely due to two sites with ground herbaceous cover of 36%
and 65%, both of which were located adjacent to a creek bed (dry at the time of the study).
As we had deliberately avoided placing other sampling sites close to sources of water, it
would be reasonable to exclude these two sites as outliers. If we exclude them, the average
ground herbaceous cover at OWL sites drops to 4.5%, which is a little lower than the mean
observed at ANT sites and a little higher than at AWT sites.
Otherwise, OWL sites most closely resembled AWT sites, further suggesting that AWT
sites are highly modified versions of open woodland. There were few shrubs at OWL sites,
with both % cover and % foliage cover of shrubs averaging only slightly higher than values
measured at AWT sites. Four sites lacked a shrub layer altogether. Height of the shrub
layer averaged 2.5m, approximately double the average height at AWT sites and about half
that of scrub sites. Of the four OWL sites that had a shrub layer, wilga was the dominant
shrub at 3 sites and regenerating poplar box dominated the shrub layer at the fourth site.
Average canopy cover was almost identical to that at CSL sites (and much higher than
any of the other 3 vegetation states), but canopy height was higher than all vegetation states
including CSL. This may have been due to all OWL sites having the same dominant
overstory species, poplar box, while other vegetation states contained a range of dominant
species, as well as the presence of fluvial soils.
Average species richness and diversity of birds at the OWL sites were similar to values
recorded at AWT sites, suggesting that OWL in the INS region is not a particularly species-

50
rich vegetation state. Table 4 also reveals that OWL did not support a unique community of
species. Most birds found within OWL were generalists, also found in all other vegetation
states.
1.4.5 Statistical Results: effects of landscapes on birds
Analysing species richness using the conservative approach, based on birds found
only within 50m of each sampling site but aggregated across all sampling sites in a
landscape, GLMs revealed no significant influences of landscape composition on species
richness of landscapes (all p > 0.15). Bird species richness at the landscape scale was
unrelated to the proportions of individual vegetation states in landscapes as well as to the
habitat diversity (Shannon index applied to proportions of vegetation states, i.e., evenness of
proportions of all vegetation states in landscapes).
Analysing species richness using the more liberal approach, based on birds found at
infinite distances from each sampling site (though still within the vegetation state of the
sampling site) aggregated across all sites in a landscape, GLMs revealed that landscapes
with higher proportions of ANT were significantly more species poor (F = 4.593, df = 1,16, p
= 0.048, R2 = 0.174). This model was improved by analysing the proportion of both
agricultural vegetation states combined. Landscapes with greater proportions of agricultural
states were significantly more species poor (F = 5.660, df = 1, 16, p = 0.03, R2 = 0.215;
Figure 17a). There were also trends for landscapes with higher proportions of CSL and
greater habitat diversity to be more species rich (proportion of CSL: F = 4.025, df= 1, 16, p =
0.062, R2 = 0.151; habitat diversity: F = 3.481, df = 1, 16, p = 0.081, R2 = 0.127; Figures
17b&c).

51
a)
b)
c)
Figure 17. Relationships between bird species richness (infinite distance data) at the
landscape level and a) the proportion of agricultural vegetation states in the landscape, b)
the proportion of CSL in the landscape, and c) the habitat diversity of the landscape,
measured as evenness of distribution across the four primary vegetation states.
In contrast, bird diversity at the landscape scale was significantly influenced by the
proportion of all vegetation states in the landscape except OSL, where there was only a
trend. Bird diversity was lower in landscapes with more ANT (F = 6.735, df = 1, 16, p = 0.02,
R2 = 0.252) and more AWT (F = 7.771, df = 1, 16, p = 0.013, R2 = 0.285). Avian diversity
tended to be higher in landscapes with more OSL (F = 3.402, df = 1, 16, p = 0.084, R2 =
0
10
20
30
40
50
60
70
80
0 20406080100
% agricultural vegetation states in landscape
Species richness
0
10
20
30
40
50
60
70
80
0 102030405060
Species richness
% CSL in landscape
0
10
20
30
40
50
60
70
80
0.75 1 1.25 1.5
Species richness
Habitat diversity in landscape

52
0.124) and was significantly greater in landscapes with more CSL (F = 6.146, df = 1, 16, p =
0.025, R2 = 0.232). The best and simplest model analysed the proportion of both agricultural
vegetation states combined (the inverse of which also represented the influence of both
scrub vegetation states combined). Landscapes with greater proportions of agricultural
states were significantly less diverse (F = 12.307, df = 1, 16, p = 0.003, R2 = 0.399; Figure
18). Our measure of habitat diversity had no significant influence on bird diversity in
landscapes (F = 1.532, df = 1, 16, p = 0.234, R2 = 0.03).
Figure 18. Negative relationship between the proportion of agricultural vegetation states in a
landscape and bird species diversity within the landscape.
Landscapes also differed in the composition of the bird communities they supported, based
on DCA analyses. Landscapes with <50% scrub contained very different communities than
landscapes with >50% scrub (Figure 19a). The one exception was landscape #37, which
was unusual in having a relatively even amount of all four vegetation states, and had just
56% of its area occupied by agricultural vegetation states. We also performed an alternative
DCA, drawing polygons around landscapes with <33% scrub, 33-67% scrub, and >67%
scrub. There was more overlap between these composition classes (Figure19b). This result
suggests that landscapes with somewhere between 33 and 67% scrub have the potential to
capture a reasonable amount of bird community diversity in the region. However, there is no
threshold response below and above these values, so a range of landscapes with different
proportions of scrub may be important at a regional scale to maintain the full range of
different communities.
2
2.5
3
3.5
4
0 20406080100
% agricultural states
Species diversity

53
a)
b)
Figure 19. Ordination diagrams from Detrended Correspondence Analysis. Landscapes
that were more similar in terms of bird species composition (both presence and abundance
data) appear closer together. Landscapes containing different proportions of scrub
vegetation states are enclosed by coloured polygons, and greater overlap between polygons
indicates greater similarity in bird communities: a) Black=landscapes with <50% scrub,
Purple=landscapes with >50% scrub, b) Black=landscapes with <33% scrub,
Purple=landscapes with 33-67% scrub, and Green=landscapes with >67% scrub.
Finally, we were particularly interested in whether vegetation states were interacting to
support birds in the landscape, making mosaic landscapes more than just a sum of their
parts. When we tested this by analysing whether birds at sampling sites within a vegetation
state were influenced by the amount of other vegetation states in the landscape, we found no
significant relationships. Mixed-effects ANOVAs revealed that neither bird species richness
nor diversity at the sampling site scale were influenced by any of the landscape-level

54
vegetation state proportions for sites in ANT, AWT, OSL or CSL (all p > 0.05). Note that this
also means that the birds within any given vegetation state in a landscape were not
influenced by the proportion of that same vegetation state in the landscape, a measure of the
total area of habitat available.
1.4.6 Are edges special?
Within edge trios, the three site types (agriculture, edge, and scrub) differed
significantly in both species richness within 50m (ANOVA: F = 5.096, df = 2, 21, p = 0.016,
R2 = 0.327) and in total abundance (ANOVA: F = 6.035, df = 2, 21, p = 0.008, R2 = 0.365).
Pairwise comparisons revealed that edge sites were more species rich and contained more
individual birds than the adjacent agricultural vegetation states (p = 0.04 for richness and p =
0.02 for abundance), but they were not more species rich or more abundant than adjacent
scrub vegetation states (p = 0.96 for richness and p = 0.995 for abundance; Figure 20).
a)
b)
Figure 20. Mean + SD of a) number of bird species and b) total number of individuals
observed within 50m point counts designed to compare habitat edges (Edge) with the
adjacent agricultural (Ag) and scrub vegetation states.
0
2
4
6
8
10
12
Ag Edge Scrub
Vegetation state
Species Richnes
s
0
1
2
3
4
5
6
7
8
9
10
Ag Edge Scrub
Vegetation state
Avian Abundance (<50m
)

55
While there was no positive edge effect, birds found at edges used the agricultural and
scrub habitats for different purposes, though they still spent most of their time in the scrub
vegetation states. Overall, when corrected for the total number of behaviours observed in
each state, the frequencies of behaviours differed between the states (Contingency analysis:
X2 = 116.66, df = 8, p<0.001). Birds did forage more often in agricultural vegetation states as
predicted, but they also engaged in more aggressive behaviours (generally conspecific
chases) and were more frequently observed simply flying to a different location. In contrast,
in scrub vegetation states, birds were more likely to be alert but also to be engaged in higher
risk behaviours such as preening, resting, and caring for young (Table 6).
Table 6. Frequencies with which different behaviours were performed in agricultural habitat
versus scrub habitat by birds found at habitat edges.
Behaviour Frequency in
Agricultural Habitat
(n = 458)
Frequency in
Scrub Habitat
(n = 1922)
Frequency in
All Habitats
(n = 2380)
aggression 2.4% 0.3% 0.7%
alert 31.9% 45.7% 43.0%
calling 12.0% 10.6% 10.9%
caring for young 0.0% 1.5% 1.2%
flying 11.4% 7.2% 8.0%
foraging 37.8% 23.7% 26.4%
preening 0.7% 6.2% 5.1%
resting 0.0% 1.9% 1.6%
unknown 3.9% 3.0% 3.2%
Total 100% 100% 100%
1.4.7 Inferring the effects of drought vs. land management
We compared bird survey data in 2007, a drought year, with data collected in 2008, an
average year, to evaluate whether any of our general results and conclusions about
vegetation states and landscape composition could be strongly influenced by the 2007
drought conditions. This was a reasonable hypothesis, as a number of grassland specialist
species were not detected in 2007, but were expected to be in the landscape based on
general distribution and on the fact that Ayers et al. (2001) detected them just a few years
earlier, though slightly further west. Alternatively, this relative lack of grassland species in
the landscape could be the result of local land management practices that have reduced
ground herbaceous cover or altered landscape composition, widespread land management
practices that have led to the decline of many grassland birds across the continent, or other
causes of more widespread grassland bird declines.
More species were detected in habitat areas in 2008, based on species recorded
outside as well as inside the 50m point count radii, aggregated to the scale of habitat areas
(Paired t-test: t=-7.341, df=58, p<0.001). On average, five more species were detected in
each habitat area in 2008 compared to 2007 (mean + SD: 5.49 + 5.74). Twenty-nine

56
species were only detected in 2008, while eight species were only detected in 2007. Six of
the species only detected in 2008 could be considered grassland specialists (black-
shouldered kite, brown songlark, red-backed kingfisher, stubble quail, white-winged fairy-
wren, and zebra finch), and two of the species only detected in 2007 are grassland
specialists (crimson chat, orange chat).
However, abundance (aggregated to the scale of habitat areas) did not differ between
years (Paired t-test: t=-1.361, df=58, p=0.179). Furthermore, correlations between data
collected in 2007 and 2008 were high, suggesting that relative richness (Pearson correlation:
r=0.801, n=59, X2=58.00, df=1, p<0.001) and abundance (Pearson correlation: r=713, n=59,
X2=40.04, df=1, p<0.001) were broadly unchanged between the years. Sites that were
relatively species rich in 2007 were also relatively species rich in 2008 (Figure 21a&b).
a) b)
c) d)
Figure 21. Comparison of 2007 and 2008 bird survey data in habitat areas (n=59): a)
correlation between bird species richness observed in each habitat area in 2007 and in 2008,
b) correlation between total abundance of birds observed in each habitat area in 2007 and in
2008, c) Mean + SD of the difference in species richness between 2008 and 2007 by
vegetation state, and d) Mean + SD of the different in total abundance between 2008 and
2007 by vegetation state.
0
10
20
30
40
50
60
0 1020304050
Species richness in 2007
Species richness in 2008
0
20
40
60
80
100
120
140
0 20 40 60 80 100 120
Total abundance in 2007
Total abundance in 2008
0
2
4
6
8
10
12
14
1 2 3 4
ANT AWT
OS
L
CS
L
Difference in richness
-5
0
5
10
15
20
25
30
35
1 2 3 4
ANT
AWT OS
L
CS
L
Difference in abundance

57
Finally, vegetation state did not influence the difference between number of species in
a given habitat area in 2008 compared to 2007 (ANOVA: F=0.903, df=4,54, p=0.469,
R2=0.063; Figure 21c). Vegetation state was also unrelated to the difference in abundance
in habitat areas in 2008 compared to 2007 (ANOVA: F=1.142, df=4,54, p=0.347, R2=0.078),
partly due to large variation within vegetation states (Figure 21d). This suggests that while a
few grassland species may have returned to the system in low numbers in 2008, these
changes were so small that they were very unlikely to influence the broader patterns we
observed in bird differences among vegetation states and landscapes.
1.5 Discussion
1.5.1 Ecological values of different vegetation states
The first aim of the study was to assess the relative contribution made by all vegetation
states in the landscape (including open areas and INS areas) to supporting native species,
quantified in terms of avian species richness, diversity, and community composition. When
analysing differences between the four primary vegetation states found in our study
landscapes (ANT, AWT, OSL and CSL), both bird species richness and avian diversity were
greater in the scrub vegetation states than in the agricultural vegetation states, but scrub and
agricultural vegetation states supported different communities of species.
In terms of richness and diversity, there was a trend for AWT to contain more species
and be more diverse than ANT. OSL appeared to be more species rich and more diverse
than CSL, but substantial variation in these parameters within vegetation states meant that
any apparent differences were not significant. As the vegetation states primarily differed
from each other in terms of woody vegetation density, these results suggest that species
richness and diversity generally increased with the local density of woody vegetation (and
thus structural complexity) – a conclusion of many other studies in diverse ecosystems
around the world, but which needed to be confirmed for these particular study sites. This
conclusion was further supported by our analyses of vegetation characteristics as continuous
variables, in which the presence of trees, the presence of a scrub layer, and the % projected
foliage cover of the scrub layer were all significantly positively related to both species
richness and diversity. Though it was suggested that the presence of drought conditions in
2007 could have been partially responsible for the lower richness and diversity in less woody
agricultural vegetation states, comparisons to data collected in 2008 revealed that the
drought did not suppress richness and diversity in agricultural vegetation states any more
than it did in scrub vegetation states. Thus, the fact that areas with less woody vegetation
supported fewer bird species and lower levels of diversity remains a valid conclusion of this
study, though it could be a natural relationship or one caused by local loss of pasture
condition, ground layer management at a continental rather than local scale, and/or other
continental-scale influences.
However, analyses of community composition showed that each vegetation state
supported a somewhat different community of species, and that agricultural vegetation states
in particular supported unique species. Thus, greater structural complexity of vegetation

58
leads to greater species richness not simply through addition of species. There is
compositional turnover as well, highlighting the fact that greater structural complexity is not
always better. AWT appeared to be particularly important, despite the fact that is was not
the most structurally complex vegetation state, as it was different in composition compared to
ANT because it supported several species that were relatively rare in the region. Such rarity
could simply have been due to the fact that AWT sites were much less prevalent in the
landscape than ANT sites. Thus, while the agricultural vegetation states were not as rich or
diverse as scrub vegetation states, they were still ecologically valuable, possibly AWT in
particular.
There were two possible exceptions to these general trends. First, analyses of
vegetation composition revealed that scrub sites with mallee species dominant in the
overstory were less species rich than scrub sites dominated by other species. However, just
as the agricultural vegetation states were lower in richness but supported a different
community, mallee sites also supported a number of unique species. Thus, where it
naturally occurs in the landscape, mallee may also contribute to the ecological values of the
region. The second exception was the fact that OSL appeared to have equal value in terms
of richness and diversity as CSL, despite the fact that it was characterised by lower woody
vegetation density. Combined with the moderate difference in composition between these
two scrub vegetation states, the fact that OSL contained many species currently classified as
vulnerable or declining (including many “declining woodland birds”), and the fact that OSL
sites were more diverse in terms of shrub species and BVTs, this suggests that OSL may
play a particularly important role in supporting both plant and bird components of biodiversity
in mosaic landscapes of the south-eastern Cobar Peneplain.
Open woodland is often discussed as a particularly valuable vegetation state in the
region. The value of open woodland that has not been highly modified by agriculture could
only be investigated in this study through qualitative comparisons with the other vegetation
states due to low availability of open woodland sites for sampling. Our results suggest that
open woodland is most similar to AWT (which may be considered a highly modified version
of open woodland) in its contribution to regional avian richness and diversity. Unlike AWT,
there were few signs that open woodland supported unique species. As it may naturally
occur only on particular soil types, it may have only a localised influence on biodiversity.
In summary, all four of our primary vegetation states plus mallee and small areas of
open woodland appeared to be ecologically valuable, and contributed to the overall diversity
and richness of birds in the region. Thus, healthy mosaics should contain a mixture of ANT,
AWT, OSL and CSL, including some mallee and open woodland where they naturally occur.
While agricultural vegetation states were valuable more for supporting a small number of
different species rather than for their richness or diversity, this could be due to the loss of
ground herbaceous cover within pastures. Thus, the contribution of agricultural vegetation
states could potentially be improved with pasture management aimed at retaining cover and
diversity of native grasses and forbs, though there could still be non-local influences limiting
the degree of potential improvement. There was some evidence that OSL may be
particularly valuable, especially where it combines patchy shrub cover with a grassed ground
layer in the spaces between shrubs. Thus, the value of OSL could also be improved with

59
management aimed at retaining cover and diversity of grasses and forbs. Finally, AWT may
be slightly more valuable than ANT and is currently underrepresented in the landscape.
1.5.2 Values of landscapes with different vegetation state proportions
The second aim of the study was to determine whether landscapes with different
proportions of the different vegetation states supported different numbers, diversity, or
communities of native bird species. Indeed, as expected based on the results of the
vegetation state analyses, landscapes with greater proportions of the less rich and less
diverse agricultural vegetation states were themselves less species rich and less diverse, but
they supported different communities. This was particularly evident in the analyses of avian
diversity at the landscape scale. Given that different vegetation states supported different
communities of species, it would be reasonable to expect that landscapes that had relatively
equal proportions of all vegetation states (i.e., that had a high habitat diversity score, based
on applying Shannon’s index to proportions of vegetation states in a landscape) would be the
most diverse in terms of bird species. There was a trend for landscapes with high habitat
diversity scores to be more species rich, though diversity at the landscape scale was not
related to the diversity of habitats in the landscape. Instead, avian diversity at the landscape
scale was positively related to the proportion of scrub vegetation states in the landscape and,
conversely, negatively related to the proportion of agricultural vegetation states in the
landscape.
Community composition analyses again suggested that the more species poor
landscapes dominated by agricultural vegetation states supported a different community than
the more species rich and more diverse scrub-dominated landscapes. Landscapes with
<50% scrub and landscapes with >50% scrub were almost completely separated in DCA
analyses. However, this was not necessarily a threshold effect, as landscapes with <33%
scrub, 33-67% scrub, and >67% scrub also had different communities, though the separation
was less distinct. These results suggest that landscapes with different levels of scrub
support different communities, but that maintaining a landscape with intermediate levels of
scrub, very approximately between 33% and 67%, should support a diverse range of
communities in that landscape. To support diverse bird communities across the region
rather than within each local landscape or property, there needs to be a balance between
landscapes dominated by (>50%) scrub and landscapes dominated by agricultural
vegetation states.
In summary, landscapes with a greater proportion of scrub vegetation states supported
greater numbers of bird species and greater avian diversity. However, landscapes with a
greater proportion of agricultural vegetation states supported a different community of bird
species. To balance the benefits from these different components of biodiversity within a
property, it may be best to maintain mosaic landscapes that have approximately 33-67%
scrub (OSL or CSL or preferably both). Regionally, it might be important to maintain some
landscapes that are dominated by agricultural vegetation states as well as landscapes that
are dominated by scrub vegetation states.

60
1.5.3 INS mosaic landscapes: are they more than a sum of their parts?
The third aim of the study was to evaluate whether these differences between
landscapes are purely due to an additive effect or whether some landscapes are more (or
less) than a sum of their parts. If landscapes are more than a sum of their parts, they will
support greater numbers or diversity of species than would be expected simply based on the
proportions of different vegetation states in the landscapes. Evidence in this study is mixed,
but generally suggests that this type of emergent property does not exist in INS
landscapes—they are largely just a sum of their constituent vegetation states.
Positive evidence in favour of this emergent property came from one of the mixed-
effects ANOVAs. In these analyses, the units of analysis were either habitat areas or
sampling sites, both of which were nested within landscapes. Modelling landscape identity
as a random effect controlled for that lack of independence in the units of analysis and the
significance of the random effect indicated whether there were consistent differences of
some sort among landscapes that influenced the response variables (bird species richness
and diversity). Landscape identity did have a significant influence on species richness when
modelling vegetation characteristics directly, rather than modelling vegetation states,
suggesting that some characteristic of landscapes per se was important, in addition to the
woody vegetation density of their individual sampling sites.
However, there was more evidence against the presence of a more-than-sum-of-parts
emergent property than there was in favour of it. All of the other mixed-effects ANOVAs
showed no significant effects of landscape identity. In addition, relationships between
proportions of vegetation states in landscapes and species richness and diversity appeared
to be generally linear rather than nonlinear (see Figures 17 & 18). A nonlinear effect might
be expected if a particular combination of vegetation states was having more than an
additive effect on birds in these landscapes. Most importantly, bird species richness and
diversity within any given vegetation state were unaffected by the proportion of other
vegetation states in the landscape, suggesting that birds were responding directly to
individual vegetation states rather than particular combinations of states. Landscapes could
also be more than a sum of their parts if edges between vegetation states created new and
different, more species-rich habitats. But edges were no more rich nor did they support
greater abundances of birds than the interiors of their constituent scrub vegetation states.
Birds living at these edges did appear to use agricultural states and scrub states for different
purposes, which would help to explain a positive edge effect if we had found one. In the
absence of a positive edge effect, these differences simply suggest that more generalist
species use different vegetation states differently in the landscape, but they probably range
fairly widely rather than concentrate their activities specifically at edges in order to do so.
In summary, there is very little evidence that any landscapes in this study were
functioning as more than a sum of their parts, supporting greater numbers of species or
greater diversity than would be expected simply based on the proportions of vegetation
states they contain. This does not necessarily mean that the configuration (size of patches
of each vegetation state, connectivity between them, etc.) of landscapes has no influence on

61
the biodiversity of those landscapes, as this study was not specifically designed to test for
effects of configuration nor did it examine all taxa in the ecosystem. However, it does
suggest that INS mosaic landscapes may not function in a very different way than other
types of mixed-use landscapes. Thus, until further research reveals more about the
functioning of INS mosaics, best-practice recommendations for planning landscape
configuration or other landscape-level properties (aside from vegetation state proportions)
can be derived from studies in other Australian ecosystems.
1.5.4 Implications for best-practice management
Our final project aim was to discuss the consequences of our results for best-practice
management of INS within the south-eastern Cobar Peneplain. We suggest several best-
practice management recommendations based on the current state of knowledge of bird
communities in landscapes containing INS:
1. Plan to achieve or maintain a mosaic-like mixture of vegetation states at the scale of
individual properties. Management may then occur on a paddock-by-paddock basis,
but according to an overall plan for achieving a mosaic at the property scale.
2. Mosaics should contain a variety of vegetation states, not just open pastures and
dense scrubby areas. Mosaics should include native perennial pastures both with
and without an open overstory of trees, scrub areas that have open grassy areas
between clumps of shrubs (“open” scrub), scrub areas that are dense or “closed”
(without clear open spaces in between trees and shrubs), as well as mallee and/or
open woodland if those states already exist in the area.
3. While no one vegetation state is most important for birds (having a mosaic mixture is
most important), there are two vegetation states that may particularly contribute to
supporting a diversity of native birds in the landscape and should thus be part of any
mosaic. These are:
a. Pasture paddocks with an open overstory of trees
b. Areas of open scrub (with clear separation between clumps of trees and
shrubs), particularly those that have a healthy grassy ground layer in the
areas between trees and shrubs
4. To maintain bird community diversity at the property scale, aim to achieve a
landscape with somewhere between 33% and 67% of the total area occupied of
various types of scrub (closed and open, dominated by various shrub and tree
species).
5. Landscapes dominated by scrub vegetation states support a greater diversity of bird
species, but landscapes dominated by agricultural vegetation states support a
different community of bird species. Thus, to maintain a diversity of native birds at
the regional scale, make sure that some landscapes are dominated (>50%) by scrub
vegetation states while others are dominated by agricultural vegetation states.

62
Additional recommendations based on other published studies:
1. The configuration of vegetation states may still be important but was not investigated
in this study. However, most of the bird species in this study have also been studied
as part of configuration research in other ecosystems. Thus, a precautionary
approach would involve following recommendations from other regions in Australia to
plan the configuration of vegetation states. At the moment these recommendations
would include:
a. ensure patches of any given vegetation state are at least 10ha in size
b. ensure patches are relatively round or square rather than linear (in other
words, avoid linear buffers unless they are connecting patches as in d. below)
c. maintain patches of scrub vegetation states that are separated from other
patches of scrub by no more than 1km
d. connect patches of scrub vegetation states that are separated by up to 1km
with either with a continuous corridor of scrub or with a paddock containing
scattered trees in which the trees are separated from each other by no more
than 100m
2. Managing pastures for greater cover of native perennial species using techniques
recommended for production purposes (such as controlling total grazing pressure,
grazing in phases separated by significant rest periods, destocking while native
plants are setting seed, etc.), is also likely to benefit native biodiversity. Note that
other research also suggests that managing for greater diversity of ground plants
(through reduced total grazing pressure and extended time intervals between
cropping) would also be beneficial. In this study, most agricultural vegetation states
had very low levels of ground cover and low plant diversity, which may have been
one reason why agricultural vegetation states supported fewer bird species and lower
levels of diversity.
3. Monitor the results of management on a regular basis, not just in terms of production
outputs but also in terms of inputs and changes in the composition, abundance and
diversity of native communities. Use the results of monitoring to adjust management
actions to ensure they are achieving the desired goals.
1.6 Future Research
At the start of this project, key knowledge gaps recognised by the INS Advisory
Committee included: 1) the composition and configuration of INS mosaics that best support
biodiversity, 2) understanding shifts between vegetation states to ensure we know what
management techniques to use to create appropriate mosaics, and 3) understanding how
dynamic changes between vegetation states influence biodiversity. While this study helped
to fill the first of these gaps, it focused on birds and it would be wise to ensure the same
patterns apply to other taxonomic groups in the region. In terms of the second knowledge
gap, new research is adding to our understanding of transitions between vegetation states.
Yet these new insights have yet to be incorporated into our understanding of how to

63
successfully manage INS landscapes. The third knowledge gap is particularly important but
has not yet been addressed. Management to create and manage planned mosaics involves
major structural changes to landscapes, and we need to know how much change we can
accomplish at one time without negatively impacting native species. We propose four future
research projects that will address these remaining gaps and continue to improve the
understanding and management of INS landscape mosaics.

64
1.7 Short Report on Bioacoustic Sensor Networks
Background
Most land managers would benefit from regular biodiversity monitoring to assess patterns
across seasons or assess changes over time in response to management. However,
monitoring is labour-intensive and thus too expensive to implement extensively. Automated
monitoring approaches involving sensor networks could provide a cost-effective alternative.
Sensor networks are rapidly becoming a critical tool for large-scale monitoring and research
because they are designed to collect data in situations where traditional labour-intensive data-
collection methods are prohibitively expensive or logistically infeasible. Sensors have been
effectively used to collect abiotic data (such as temperature, humidity, etc.) for some time.
Methods for collecting biotic data such as measures of species richness and diversity are still
in development, but could prove to be a relatively inexpensive way of collecting large amounts
of data about ecosystem health.
Project Objectives
Within the context of the broader work on avian richness and diversity in INS landscape
mosaics, we were also able to trial the use of bioacoustic sensors as a surrogate for avian
diversity surveys. In collaboration with CSIRO Information and Communication Technologies
(ICT), we used bioacoustic sensors to record sound at a number of point count sites at the
same time that point counts were being performed. Our primary objective was then to develop
a simple, automated way of characterising the overall sound profile in each of the recordings
that would be correlated with avian species richness or diversity as measured by the on-
ground point count teams. During deployment of the sensors, ICT also aimed to gather
additional information to help them continue to develop the networking capability of these
sensors, which allows managers to control the sensors and download data remotely.
Methods
We attempted to collect recordings at half of the point counts performed in 2008 using
bioacoustic sensors developed by ICT. Sensors were networked to laptop computers in the
field which configured sensors to specific locations and controlled start times of recordings
(Figure 23). All recordings were 10 minutes long, the same length of time as point counts
were performed at the same sites. To minimise extraneous noise in the recordings, sensors
were placed 20m away from the centre of point count sites, where the point count team stood
and recorded all birds seen or heard.

65
a) b)
Figure 23. Bioacoustic sensors in INS landscapes: a) a deployed sensor with antenna, wind
filter on a microphone, electronics housing, and cable to connect to a 12 volt battery, and b)
configuring a sensor with a laptop computer in the field.
Recordings occasionally failed or were abandoned due to high wind speeds, which we knew
would create too much masking noise in the recording. Nonetheless, some recordings still had
significant wind noise (above a subjectively-determined threshold intensity) or other acoustic
interference. We eliminated these recordings from the study, yielding a total of 95 analysable
recordings, distributed across all five vegetation states, including open woodland.
We experimented with a wide variety of methods to simply and automatically characterise
sound profiles from these recordings. We evaluated each method by comparing the results
with species richness data from a test set of 11 recordings, selected because the point counts
performed at the same time suggested these sites ranged from low to high species richness.
We also evaluated an existing approach developed by Sueur et al. (2008) – the only published
approach to characterising sound profiles for biodiversity purposes. We modified the protocol
slightly from the original. As did Sueur et al. (2008), we divided each recording into a series of
time bands and frequency bands, calculated the acoustic energy within each of the resulting
cells, then used those energy values to calculate the acoustic equivalent of The Shannon
diversity index, the Acoustic Entropy Index. Unlike Sueur et al. (2008), we used biological
data to determine the width of both time and frequency bands. We did this by identifying the
20 most abundant species in point counts during 2008, then using stock recordings of these
species to calculate the average call length and frequency range. Average call length was 1.5
seconds, but most species give repeated calls, so we used a time band width of 5 seconds
(yielding a total of 120 time bands for a 10 minute recording). Average frequency range was

66
~4 kHz and our recordings covered a range of 8 kHz. However, the calls of most species were
concentrated in terms of acoustic energy in just 1/2 to 1/3 of their total frequency range. Thus
we used two possible frequency band widths, 2 kHz and 1.3 kHz (yielding a total of either 4 or
6 frequency bands per recording). We then analysed whether the resulting Acoustic Entropy
Indices for all analysable recordings were accurate predictors of either species richness
(infinite distance data) or diversity (within 50m) based on point counts at the sampling site.
Results
None of our experimental approaches were correlated with species richness using our test
data set. And even though Sueur et al. (2008) successfully showed that their measure of
acoustic entropy correlated with species richness using a simulated dataset, we found that
neither of our Acoustic Entropy Indices were related to either species richness or diversity
using real data collected at the same time as recordings (all p > 0.2). In fact, Acoustic Entropy
Indices never explained more than 1% of the variation in bird species richness or diversity
values (Figure 24).

67
a)
b)
Figure 24. Relationships between acoustic entropy and bird diversity data: a) entropy
calculated with 4 frequency bands versus bird species richness, b) entropy calculated with 4
frequency bands versus avian diversity. Results for entropy calculated with 6 frequency bands
were similar.
Conclusions
We have not yet found a simple way to quantify the overall sound profile of a bioacoustic
recording at a given site that is correlated with or can predict avian richness or diversity
measures collected on the ground at that site. In other words, we do not yet have a substitute
for labour-intensive on-ground ecological monitoring. While Sueur et al.’s method was
correlated with the species richness of data they simulated using studio-quality recordings of
multiple taxonomic groups, our test of these methods shows that they are not successful when
compared with real survey data. We suspect this relates to particular characteristics of their
simulated recordings which may not match most real-world situations. We are currently
exploring these and other potential explanations for the discrepancy between our results and
those of Sueur et al. (2008).
0
5
10
15
20
25
30
0.4 0.6 0.8 1
Acoustic Entropy Index
Species Richness
0
0.5
1
1.5
2
2.5
0.4 0.5 0.6 0.7 0.8 0.9 1
Acoustic Entropy Index
Diversity (within 50m
)

68
The most promising of the approaches we experimented with attempted to use algorithms to
automatically identify different segments of sound (for example, a single weebill call or one
syllable of a pied butcherbird call), then count the number of unique segments (which might
correlate with species richness) as well as the number of times each unique segment appears
in each recording (which might correlate with abundance within species and allow calculation
of a diversity index). At the moment, the software tends to be overly sensitive to small
amounts of noise, so it treats small abiotic noises in the environment as unique segments, and
it tends to divide each individual bird call into too many different segments. However, there is
hope that with more technological development, this automated approach could deliver
measures of richness and diversity straight from sensors in the field to the computers of
managers (via the mobile phone network or other communications infrastructure).
The primary alternative would be to design species recognition software that would recognise
the specific calls of all (or a majority of) species in the landscape. This would have the
advantage of providing data not just on richness and diversity but also on composition.
However, while the technology does exist and such software programs have been designed
elsewhere, they require a great deal of initial investment in time and labour to develop –
generally more than would be available for technological development in natural resource
management. In addition, there are no guarantees this approach would work in Australia,
where geographic variation in calls might mean that new software would have to be developed
for each location at which managers might wish to put a sensor network, and large numbers of
local recordings would need to be available for training the software. In addition, the relative
simplicity of Australian bird calls and the prevalence of birds that mimic the calls of other
species could severely limit the accuracy of species recognition software.
Outcomes
New sensor network technology could eventually help organisations such as Catchment
Management Authorities collect data on avian diversity and site health remotely, thus
permitting significant savings on labour costs and/or significant increases in the spatial extent
of data collection. Pre-processed data could be delivered direct to desktops so that
managers could receive information on site quality directly, without having to analyse or
interpret the data themselves. However, while the technology to collect information on
components of biodiversity is now available, we do not yet have a way to reliably process
that information, aside from the labour-intensive method of having someone listen to all the
recordings to count bird species. Nonetheless, with continued investment in technological
development, and an emphasis on testing new technologies against real data rather than
simulated data, we should be able to develop more cost-effective automated approaches to
monitoring biodiversity in the near future.

69
1.8 Acknowledgements
This project would not have been possible without the generous support of the Central
West Catchment Management Authority (CWCMA), in partnership with the Western
Catchment Management Authority, and the Invasive Native Scrub Advisory Committee, all of
whom provided advice and guidance along the way. Most of all, the project would not have
been possible without the support of the many private landholders who kindly allowed us
access to their properties and often took considerable time to show us around and discuss
the issues, including Paul & Chris Brooks, John Chamberlain, Laura Douglas from NSW
National Parks, Michael Dutschke, Kevin & Margaret Easey, Joe & Gabrielle Holmes, Max
Holmes, Shane and Kathleen Klante, Terry Klante, Craig & Fiona Larkings, Geoff Martin and
family, Andrew & Megan Moseley, Max Tremain, Les & Colleen Tumpey, Michael Wass &
family, Peter Weston & family, and Wayne Whillock & family.
Data collection involved long, often hot and fly-filled days, and we sincerely thank Kelly
Barr, Chris Davey, Damien Farine, Stuart Harris, Patricia Jones, Alex McLaughlin, Les
Overs, Ingrid Pollet, Vicki Saint, and Jessica (Kika) Tarsi for working so hard with us.
Additional advice and guidance were kindly provided by Dani Ayers and Julian Seddon
(NSW Department of Environment and Climate Change), Kerry Bridle and Andre Zerger
(CSIRO Sustainable Ecosystems), Les Overs (CSIRO Information & Communication
Technologies) and Ray Thompson (CWCMA). Finally, a special thanks to the people of
Tottenham NSW who helped us feel welcome in the region, particularly Meryl & Ian Boothby
of Meadow View Farmstay, and everyone at the Happy Valley Caravan Park, the Tottenham
Hotel, the Tottenham Bowling Club, Gem of the West Café (best burgers around!) and
Tottenham Rural Trading Pty Ltd.
1.9 Literature Cited
Ayers, D., G. Melville, J. Bean, D. Beckers, M. Ellis, T. Mazzer & D. Freudenberger (2001)
Woody Weeds, Biodiversity and Landscape Function in Western NSW. West 2000,
Dubbo.
Barrett, G., A. Silcocks, S. Barry, S. Cunningham & R. Poulter (2003) The New Altas of
Australian Birds. Royal Australasian Ornithologists Union, Hawthorn East, Victoria.
Belsky, A. J. (1996) Viewpoint: Western juniper expansion: is it a threat to arid northwestern
ecosystems? Journal of Range Management, 49, 53-59.
Bennett, A. F., J. Q. Radford & A. Haslem (2006) Properties of land mosaics: Implications for
nature conservation in agricultural environments. Biological Conservation, 133, 250-
264.
Blaum, N., E. Rossmanith, G. Fleissner & F. Jeltsch (2007) The conflicting importance of
shrubby landscape structures for the reproductive success of the yellow mongoose
(Cynictis penicillata). Journal of Mammalogy, 88, 194-200.
Bradstock, R. A., M. Bedward, A. M. Gill & J. S. Cohn (2005) Which mosaic? A landscape
ecological approach for evaluating interactions between fire regimes, habitat and
animals. Wildlife Research, 32, 409-423.
Briggs, S. V., J. A. Seddon & S. J. Doyle (2007) Structures of bird communities in woodland
remnants in central New South Wales, Australia. Australian Journal of Zoology, 55, 29-
40.

70
Clarke, M. F. & J. M. Oldland (2007) Penetration of remnant edges by noisy miners
(Manorina melanocephala) and implications for habitat restoration. Wildlife Research,
34, 253-261.
DEC (2006) Reconstructed and extant distribution of native vegetation in the Central West
Catchment. NSW Department of Environment and Conservation, Dubbo.
Fensham, R. J. (2008) Leichhardt's maps: 100 years of change in vegetation structure in
inland Queensland. Journal of Biogeography, 35, 141-156.
Forman, R. T. T. (1995) Land mosaics: the ecology of landscapes and regions. Cambridge
University Press, Cambridge.
Friedel, M. H. & V. H. Chewings (1988) Comparison of crown cover estimates for woody
vegetation in arid rangelands. Australian Journal of Ecology, 13, 463-468.
Gregory, R. D., P. Vorisek, D. G. Noble, A. Van Strien, A. Klvanova, M. Eaton, A. W. G.
Meyling, A. Joys, R. P. B. Foppen & I. J. Burfield (2008) The generation and use of bird
population indicators in Europe. Bird Conservation International, 18, S223-S244.
Hassall & Associates (1999) Woody weeds and biodiversity: a review of previous and current
work. Dubbo, New South Wales.
Hassall & Associates, S. Briggs & P. Norman (2006) Documenting the Science Behind the
Invasive Native Scrub Tool. Central West Catchment Management Authority, Dubbo.
Hobbs, R. J. (1999) Restoring the Health and Wealth of Ecosystems. CSIRO, Perth.
Hodgkinson, K. C. (2002) Fire regimes in Acacia wooded landscapes: effects on functional
processes and biological diversity. in Flammable Australia: the fire regimes of a
continent (Bradstock, R. A., J. E. Williams & A. M. Gill, eds.), Cambridge University
Press, Cambridge. pp. 259-277.
Keith, D. A. (2004) Ocean shores to desert dunes: the native vegetation of New South Wales
and the ACT. NSW Department of Environment and Conservation, Sydney.
Kerle, A. (2008) Managing rangeland vegetation with fire: a literature review and
recommendations. Western Catchment Management Authority.
Kleyer, M. (2007) Mosaic cycles and conservation management. Basic and Applied Ecology,
8, 293-294.
Kleyer, M., R. Biedermann, K. Henle, E. Obermaier, H.-J. Poethke, P. Poschlod, B.
Schroder, J. Settele & D. Vetterlein (2007) Mosaic cycles in agricultural landscapes of
Northwest Europe. Basic and Applied Ecology, 8, 295-309.
Larsen, F. W., J. Bladt & C. Rahbek (2009) Indicator taxa revisited: useful for conservation
planning? Diversity and Distributions, 15, 70-79.
Law, B. S. & C. R. Dickman (1998) The use of habitat mosaics by terrestrial vertebrate
fauna: implications for conservation and management. Biodiversity and Conservation,
7, 323-333.
McGarigal, K. & S. A. Cushman (2002) Comparative evaluation of experimental approaches
to the study of habitat fragmentation effects. Ecological Applications, 12, 335-345.
Meyer, K. M., K. Wiegand, D. Ward & A. Moustakas (2007) The rhythm of savanna patch
dynamics. Journal of Ecology, 95, 1306-1315.
Mueller-Dombois, D. & H. Ellenberg (1974) Aims and methods of vegetation ecology. Wiley
and Sons, New York.
Noble, J. C. (1997) The delicate and noxious scrub. CSIRO Wildlife and Ecology, Canberra.
Peters, D. P. C., B. T. Bestelmeyer, J. E. Herrick, E. L. Fredrickson, H. C. Monger & K. M.
Havstad (2006) Disentangling complex landscapes: New insights into arid and semiarid
system dynamics. BioScience, 56, 491-501.
Pino, J., F. Roda, J. Ribas & X. Pons (2000) Landscape structure and bird species richness:
implications for conservation in rural areas between natural parks. Landscape and
Urban Planning, 49, 35-48.
Radford, J. Q. & A. F. Bennett (2007) The relative importance of landscape properties for
woodland birds in agricultural environments. Journal of Applied Ecology, 44, 737-747.
Reid, J. R. W. (1999) Threatened and declining birds in the New South Wales sheep-wheat
belt: I. Diagnosis, characteristics and management. CSIRO Wildlife and Ecology,
Canberra.
Remmert, H., eds. (1991) The mosaic-cycle concept of ecosystems. Springer, Berlin.

71
Ries, L., R. J. Fletcher, J. Battin & T. D. Sisk (2004) Ecological responses to habitat edges:
mechanisms, models, and variability explained. Annual Review of Ecology Evolution
and Systematics, 35, 491-522.
Schlesinger, W. H., J. F. Reynolds, G. L. Cunningham, L. F. Huenneke, W. M. Jarrell, R. A.
Virginia & W. G. Whitford (2008) Biological feedbacks in global desertification. Science,
247, 1043-1048.
Shannon, C. E. (1948) A mathematical theory of communication. Bell System Technical
Journal, 27, 379-423.
Skowno, A. L. & W. J. Bond (2003) Bird community composition in an actively managed
savanna reserve, importance of vegetation structure and vegetation composition.
Biodiversity and Conservation, 12, 2279-2294.
Sueur, J., S. Pavoine, O. Hamerlynck & S. Duvail (2008) Rapid acoustic survey for
biodiversity appraisal. PLoS ONE, 3, e4065. doi:10.1371/journal.pone.0004065.
Thompson, W. A. & D. J. Eldridge (2005) White cypress pine (Callitris glaucophylla): a
review of its roles in landscape and ecological processes in eastern Australia.
Australian Journal of Botany, 53, 555-570.
Watt, A. S. (1947) Pattern and process in the plant community. Journal of Ecology, 35, 1-22.
Westoby, M., B. Walker & I. Noymeir (1989) Opportunistic management for rangelands not at
equilibrium Journal of Range Management, 42, 266-274.
Wiegand, K., D. Saitz & D. Ward (2006) A patch-dynamics approach to savanna dynamics
and woody plant encroachment - Insights from an arid savanna. Perspectives in Plant
Ecology Evolution and Systematics, 7, 229-242.
Wiegand, K., D. Ward & D. Saltz (2005) Multi-scale patterns and bush encroachment in an
and savanna with a shallow soil layer. Journal of Vegetation Science, 16, 311-320.
1.10 Appendix A
Scientific names of species or species groups referred to in this report.
Common name Scientific name
- Bird species of INS landscapes
apostlebird Struthidea cinerea
Australian hobby Falco longipennis
Australian kestrel Falco cenchroides
Australian magpie Gymnorhina tibicen
Australian owlet-nightjar Aegotheles cristatus
Australian pipit Anthus australis
Australian raven Corvus coronoides
Australian ringneck Barnardius zonarius
Australian wood duck Chenonetta jubata
banded lapwing Vanellus tricolor
bar-shouldered dove Geopelia humeralis
black honeyeater Certhionyx niger
black kite Milvus migrans
black-chinned honeyeater Melithreptus gularis
black-eared cuckoo Chalcites osculans
black-faced cuckoo-shrike Coracina novaehollandiae
black-faced woodswallow Artamus cinereus
black-shouldered kite Elanus axillaris
blue bonnet Northiella haematogaster
blue-faced honeyeater Entomyzon cyanotis
brown falcon Falco berigora

72
Common name Scientific name
brown goshawk Accipiter fasciatus
brown songlark Cincloramphus cruralis
brown treecreeper Climacteris picumnus
brown-headed honeyeater Melithreptus brevirostris
buff-rumped thornbill Acanthiza reguloides
chestnut quail-thrush Cinclosoma castanotus
chestnut-rumped thornbill Acanthiza uropygialis
cockatiel Nymphicus hollandicus
common bronzewing Phaps chalcoptera
common starling Sturnus vulgaris
crested bellbird Oreoica gutturalis
crested pigeon Ocyphaps lophotes
crimson chat Ephthianura tricolor
eastern rosella Platycercus adscitus
eastern yellow robin Eopsaltria australis
emu Dromaius novaehollandiae
galah Eolophus roseicapillus
Gilbert's whistler Pachycephala inornata
grey butcherbird Cracticus torquatus
grey fantail Rhipidura albiscapa
grey shrike-thrush Colluricincla harmonica
grey-crowned babbler Pomatostomus temporalis
grey-fronted honeyeater Lichenostomus plumulus
ground cuckoo-shrike Coracina maxima
hooded robin Melanodryas cucullata
Horsfield's bronze-cuckoo Chalcites basalis
inland thornbill Acanthiza apicalis
jacky winter Microeca fascinans
laughing kookaburra Dacelo novaeguineae
little corella Cacatua sanguinea
little eagle Hieraaetus morphnoides
little friarbird Philemon citreogularis
little pied cormorant Phalacrocorax melanoleucos
little raven Corvus mellori
magpie-lark Grallina cyanoleuca
Major Mitchell's cockatoo Cacatua leadbeateri
masked woodswallow Artamus personatus
mistletoebird Dicaeum hirundinaceum
mulga parrot Psephotus varius
noisy friarbird Philemon corniculatus
noisy miner Manorina melanocephala
olive-backed oriole Oriolus sagittatus
orange chat Epthianura aurifrons
pallid cuckoo Cuculus pallidus
peaceful dove Geopelia placida
pied butcherbird Cracticus nigrogularis
rainbow bee-eater Merops ornatus
red-backed kingfisher Todiramphus pyrrhopygia
red-capped robin Petroica goodenovii
red-rumped parrot Psephotus haematonotus
restless flycatcher Myiagra inquieta
rufous songlark Cincloramphus mathewsi

73
Common name Scientific name
rufous whistler Pachycephala rufiventris
sacred kingfisher Todiramphus sanctus
shy heathwren Calamanthus cautus
singing honeyeater Lichenostomus virescens
southern whiteface Aphelocephala leucopsis
speckled warbler Pyrrholaemus sagittatus
spiny-cheeked honeyeater Acanthagenys rufogularis
splendid fairy-wren Malurus splendens
spotted bowerbird Chlamydera maculata
spotted harrier Circus assimilis
spotted pardalote Pardalotus punctatus
striated pardalote Pardalotus striatus
striated thornbill Acanthiza lineata
striped honeyeater Plectorhyncha lanceolata
stubble quail Coturnix pectoralis
superb parrot Polytelis swainsonii
tawny frogmouth Podargus strigoides
tree martin Petrochelidon nigricans
varied sittella Daphoenositta chrysoptera
variegated fairy-wren Malurus lamberti
wedge-tailed eagle Aquila audax
weebill Smicrornis brevirostris
welcome swallow Hirundo neoxena
western gerygone Gerygone fusca
white-bellied cuckoo-shrike Coracina papuensis
white-breasted woodswallow Artamus leucorynchus
white-browed babbler Pomatostomus superciliosus
white-browed woodswallow Artamus superciliosus
white-eared honeyeater Lichenostomus leucotis
white-faced heron Egretta novaehollandiae
white-necked heron Ardea pacifica
white-plumed honeyeater Lichenostomus penicillatus
white-throated gerygone Gerygone olivacea
white-winged chough Corcorax melanorhamphos
white-winged fairy-wren Malurus leucopterus
white-winged triller Lalage tricolor
willie wagtail Rhipidura leucophrys
yellow thornbill Acanthiza nana
yellow-plumed honeyeater Lichenostomus ornatus
yellow-rumped thornbill Acanthiza chrysorrhoa
yellow-throated miner Manorina flavigula
zebra finch Taeniopygia guttata
- Plant species of INS landscapes
allocasuarina Allocasuarina sp.
belah Casuarina cristata
budda Eremophila mitchellii
cypress pine Callitris sp.
dwyer's red gum Eucalyptus dwyeri
grey box Eucalyptus mircocarpa
kurrajong Brachychiton populneus
mallee Eucalyptus spp.
poplar box Eucalyptus populnea

74
Common name Scientific name
rosewood Alyectron oleifolius
punty bush Senna spp.
smooth-barked coolibah (red box) Eucalyptus intertexta
turpentine Eremophila sturtii
white cypress pine Callitris glaucophylla
wilga Geijera parviflora
- Other species mentioned
aspen Populus spp.
creosote Larrea tridentata
grey kangaroo Macropus giganteus/fuliginosus
mesquite Prosopis spp.
yellow mongoose Cynictis penicillata

75