The value of weeds for wildlife and implications for management

by

Emma H. Carlos, B. Env. Sc. (Hons)

Submitted in fulfilment of the requirements for the degree of

Doctor of Philosophy

Deakin University

December, 2012

Preface

All research for this thesis had relevant ethics approval and government permits. Research in Chapter 2 was approved by the Deakin University Human Research Ethics Committee (STEC-16-2008-CARLOS). Research for Chapters 3 – 5 was conducted under permit from the Department of Sustainability and Environment (10004671) and procedures were approved by the Deakin University Animal Welfare Committee (A49- 2008).

This thesis is presented as a series of stand-alone research papers. Each data chapter (Chapters 2 – 5) has either been submitted for publication or has been written with the intention for submission to a specific journal. As required by each journal, data chapters are set out with their own Abstract, Introduction, Methods, Results and Discussion section. An effort has been made to make the style and format of the thesis sections consistent; however, some journal specific requirements have been retained. To minimise undue repetition all references have been consolidated at the end of the thesis.

I am the primary investigator and contributor to all aspects of this research but, as some data chapters are written with the intention for submission to a specific journal, the contribution of co-authors, Maria Gibson (Principle Supervisor) and Michael Weston (Associate Supervisor) is recognised. Both provided significant guidance on each data chapter and commented on drafts of each chapter. Thus, data chapters occasionally refer to ‘we’ in recognition of co-author contributions.

i Acknowledgements

I am grateful to the many people who have provided support, encouragement, assistance and advice to me throughout my PhD journey.

Firstly, thank you to my supervisors, Maria (Mary) Gibson and Mike Weston. Your knowledge, wisdom and unwavering support have been invaluable throughout my research process. Mary, your trust and belief in me has been endless and has encouraged me to challenge myself in ways I never anticipated. I have grown in confidence and ability because of your support and I sincerely thank you for this. Mike, I thank you for wholeheartedly going out of your way to help me in every aspect of this research. The passion and enthusiasm you communicate for research has been infectious and your humour has lightened many a dull day. Together, you both have made one amazing team and there is nothing more I could have hoped for.

My research would not have been possible without the support from the brilliant Parks Victoria staff at the You Yangs Regional Park and Point Cook Coastal Park. In particular: Lauren Kelly, for so cheerfully welcoming me to the You Yangs and for her friendly and approachable nature that put me at ease right from the word go; Bernie McCarrick at the Point Cook office for being an enthusiastic supporter of my research and making so much possible, and; David Flagg for being so accommodating, easy- going and for making such a significant contribution towards the spraying of ‘my’ boneseed at the You Yangs. Thank you also to David Roberts, John Argote, Mark Cullen, Russell Brooks and Alex Tomsic, and to Mike Helmen, Mick Smith and the rest of the team at Serendip Sanctuary. I am privileged to have had the opportunity to work and stay in these very special parks and it means a great deal to me that you were all so supportive.

ii Financial support was provided through the ANZ Holsworth Wildlife Research Endowment, Birds Australia (Victoria) and Deakin University Postgraduate Scholarship.

There are so many people within Deakin University that have helped me and I sincerely thank you all. In particular, I have always been provided with amazing technical support from Clorinda Schofield. We would all be lost without you Clorinda, thank you so much. I am also grateful to James Hill and Jess Bywater, for the technical support you provided; Robyn Cook, Clare Hake and more recently, Heather Andrews, thank you for your assistance with the endless paper chase; thanks are due to Kelly Miller, for her advice in the initial stages of my social research component, and to Desley Whisson, for her animal handling expertise and patience during the training process.

The enthusiastic people who volunteered for me during various stages of my project were amazing. I am so thankful to you all. A special thanks to Fanny Chevrier, Anthony Rendall, Erin Lennox, Sarah Ludlow, Kristin Scherlies, Debby Carlos and Jackson Carlos, who joined me on multiple trips for extended periods; your commitment and patience, particularly in the face of endless vegetation surveys, was admirable. I wouldn’t have managed the field work component without the help so willingly provided by so many; I am forever grateful.

During the analysis phase I was fortunate to have been provided with statistical advice from Gerry Quinn and Dale Nimmo. I especially thank you, Dale, for still having the courage to ask me how the stats were going, even when the answer was likely to result in a considerable loss of your valuable time. Thank you to those who have waded through drafts and provided constructive comments and editing throughout the writing process, with a special mention to Kristin Scherlies and Emmi Scherlies during the final write-up.

My thanks also go to my colleagues and fellow PhD journeyers, past and present, for the brainstorming, encouragement, advice, laughter and tears that have been shared. In particular: Nicole Schumann, I feel fortunate to have travelled through the PhD journey

iii with you from start to finish, and the brainstorming sessions we had helped not only my PhD, but many aspects of life; Evelyn Chia, your friendship means a lot to me, I’m glad you went against my advice and chose the desk next to the window; Christine Connelly and Tanya Pyk, your company and encouragement has always been well-timed and very much appreciated.

Finally, I thank my friends and family – Mum, Dad, Marlene, Jackson, Krissy, Heidi, Armstrong, Emmi and Riki. You not only provided support, encouragement and love, but were also some of my most reliable volunteers. I will always be grateful for your faith in me, and I love you all dearly.

iv Table of contents

Preface ...... i Acknowledgements...... ii Table of contents...... v List of figures ...... ix List of tables...... xi Definitions and abbreviations...... xiv Abstract ...... xv Chapter 1 General Introduction ...... 1 1.1 Invasive species and conservation: A global issue...... 2 1.2 The impacts of weeds on wildlife...... 2 1.2.1 The values of weeds for wildlife ...... 3 1.3 Impacts of weed management on wildlife...... 6 1.4 Thesis overview...... 10 1.4.1 Thesis objectives and structure...... 10 1.4.2 Study areas ...... 13 Chapter 2 Weeds and wildlife: Perceptions and practices of weed managers...... 18 2.1 Abstract...... 19 2.2 Introduction...... 20 2.3 Methods ...... 22 2.3.1 Statistical analysis ...... 24 2.4 Results...... 24 2.4.1 Observations of wildlife during weed management...... 25 2.4.2 Attitudes towards wildlife during weed management...... 25 2.4.3 Adaptive management for wildlife...... 26

v 2.4.4 The objectives and success of weed management...... 30 2.5 Discussion...... 31 2.5.1 Weed management projects: purpose, perceptions and practice...... 31 2.5.2 Weeding for wildlife ...... 33 2.6 Appendix A...... 36 Chapter 3 Vegetation structure, not weediness, influences wildlife habitat in a southeastern Australian woodland remnant ...... 42 3.1 Abstract...... 43 3.2 Introduction...... 44 3.3 Methods ...... 45 3.3.1 Study area...... 45 3.3.2 Study design ...... 45 3.3.3 Small mammals ...... 46 3.3.4 Birds ...... 47 3.3.5 Vegetation ...... 47 3.3.6 Variable selection ...... 48 3.3.7 Data analysis...... 48 3.4 Results...... 50 3.4.1 Vegetation ...... 50 3.4.2 Birds ...... 56 3.4.3 Small mammals ...... 61 3.5 Discussion...... 66 3.5.1 Wildlife habitat preferences ...... 66 3.5.2 Management implications ...... 68 3.6 Appendix A...... 70 Chapter 4 The influence of an emergent weed on wildlife in a southern Australian wetland . 72 4.1 Abstract...... 73

vi 4.2 Introduction...... 74 4.3 Methods ...... 76 4.3.1 Study area...... 76 4.3.2 Study design ...... 77 4.3.3 Variable selection and data analysis...... 79 4.4 Results...... 81 4.4.1 Vegetation ...... 81 4.4.2 Birds ...... 88 4.4.3 Small mammals ...... 89 4.5 Discussion...... 95 4.5.1 The influence of boxthorn on birds ...... 95 4.5.2 The influence of structure on species ...... 96 4.5.3 Facing the future...... 98 4.6 Appendix A...... 100 Chapter 5 ‘Wildlife-friendly’ weed management sustains birds and small mammals ...... 101 5.1 Abstract...... 102 5.2 Introduction...... 103 5.3 Methods ...... 104 5.3.1 Study area and design...... 104 5.3.2 Variable selection and data analysis...... 107 5.4 Results...... 110 5.4.1 You Yangs...... 110 5.4.2 Cheetham Wetlands...... 123 5.5 Discussion...... 124 5.5.1 Impacts of weed management on wildlife...... 125

vii Chapter 6 Synthesis...... 129 6.1 Overview...... 130 6.2 Major findings ...... 130 6.2.1 Consideration of wildlife by weed managers ...... 130 6.2.2 The influence of weeds on wildlife ...... 132 6.2.3 Weed management impacts on wildlife ...... 134 6.3 Management implications...... 136 6.4 Conclusion ...... 140 References ...... 141

viii List of figures

Figure 1.1 Typical adaptive management cycle...... 7 Figure 1.2 Structure and flow of thesis ...... 12 Figure 1.3 Woodland ecosystem of the You Yangs Regional Park, Victoria, Australia. Twenty-five study sites are shown ...... 15 Figure 1.4 Wetland ecosystem of Cheetham Wetlands, Victoria, Australia. Eighteen study sites are shown...... 16 Figure 1.5 Typical vegetation in the woodland ecosystem of the You Yangs Regional Park...... 17 Figure 1.6 Typical vegetation in the wetland ecosystem of Cheetham Wetlands...... 17 Figure 2.1 Mean scale scores (± 95% CI) of the ranked frequency that respondents (n = 67) observed native (circles) and non-native (triangles) wildlife using weeds of different characteristics and in different contexts...... 28 Figure 3.1 Habitat variables between site types at the You Yangs Regional Park...... 54 Figure A3.1 Cumulative bird species richness recorded over timed surveys (natural log + 1 transformed, ± 1 SE) in different site types in the You Yangs Regional Park...... 71 Figure 4.1 Habitat variables between site types at Cheetham Wetlands ...... 86 Figure A4.1 Cumulative bird species richness over timed surveys (natural log + 1 transformed, ± 1 SE) in different site types in Cheetham Wetlands...... 100

ix Figure 5.1 Mean standardised values (± 1 SE) of those habitat variables which were impacted by weed management, as shown by substantial support for models that included a treatment × time interaction term ...... 117 Figure 6.1 Basic adaptive management cycle with factors that require consideration with the inclusion of wildlife in weed management, as identified in the literature and through this research ...... 138

x List of tables

Table 2.1 Statistical comparisons between different weed categories/characteristics in relation to: the frequency respondents had been involved in management of these weeds; how important respondents believed management was for these weeds; and the frequency respondents had observed the use of these weeds by native and non-native wildlife ...... 27 Table 2.2 Weed management techniques used by managers in Victoria, Australia...... 29 Table 2.3 Beliefs cited by respondents (n = 62) to justify their view that weeds provide important habitat to wildlife...... 29 Table 2.4 Type of adjustments managers had made to weed management in order to better accommodate wildlife ...... 30 Table 3.1 Factor loadings for selected principal components derived from average vegetation life-form variables measured at the You Yangs Regional Park...... 52 Table 3.2 Factor loadings for selected principal components representing the vertical density of vegetation at the You Yangs Regional Park...... 53 Table 3.3 Seasonal flowering and fruiting vegetation between site types at the You Yangs Regional Park...... 55 Table 3.4 Bird species recorded at the You Yangs Regional Park ...... 57 Table 3.5 Comparison of house mouse capture rates, bird species richness and bird assemblages between site types at the You Yangs Regional Park .. 60 Table 3.6 Model selection results for birds and house mouse at the You Yangs Regional Park...... 63 Table 3.7 Relative influence of habitat variables on wildlife at the You Yangs Regional Park as derived from model averaging ...... 64

xi Table 4.1 Factor loadings of principal components derived from analyses of life-form variables measured in Cheetham Wetlands...... 84 Table 4.2 Factor loadings of principal components derived from analyses of vertical vegetation density scores (up to 180 cm tall) in Cheetham Wetlands...... 85 Table 4.3 Average seasonal richness and cover of flowering and fruiting plants in different site types at Cheetham Wetlands...... 87 Table 4.4 Bird species recorded in Cheetham Wetlands (all seasons)...... 90 Table 4.5 Comparison of bird species richness, bird assemblages and house mouse capture rates between site types over four seasons and when averaged over seasons at Cheetham Wetlands...... 92 Table 4.6 Model selection results for birds and house mouse at Cheetham Wetlands...... 93 Table 4.7 Model averaging results for birds and house mouse at Cheetham Wetlands...... 94 Table 5.1 Factor loadings for principal components derived from average life- form variables measured at the You Yangs Regional Park (YY) and Cheetham Wetlands (CW)...... 112 Table 5.2 Factor loadings for principal components representing the vertical density of vegetation (to 190 cm) at the You Yangs Regional Park (YY) and Cheetham Wetlands (CW)...... 113 Table 5.3 Model selection results for habitat variables and wildlife metrics in two ecosystems (You Yangs Regional Park, YY; Cheetham Wetlands, CW) where experimental weed management was conducted...... 114 Table 5.4 Bird species recorded in two ecosystems (You Yangs Regional Park, YY; Cheetham Wetlands, CW)...... 119

xii Table 5.5 Pair-wise comparisons of bird assemblages in impact, control and reference sites compared before (B), 1 – 18 days after (A1) or 42 – 60 days after (A2) weed management in the You Yangs Regional Park (YY) and Cheetham Wetlands (CW)...... 122

xiii Definitions and abbreviations

Weed – Throughout this thesis an invasive plant will be referred to as a weed. Weeds have been variously defined in the literature (Richardson et al. 2000; Grice and Martin 2006; Gibson 2010; Schlaepfer et al. 2011), but in this thesis the term will be used to define an unwanted plant that is characterised by its invasive nature and presumed detrimental effects. This has no connotation about the type of environment it invades. When the environment it invades is relevant, further specifications will be made, for example, agricultural weeds or, weeds in natural ecosystems. It is acknowledged that the term weed can be considered a colloquial term, or refers solely to plants that invade agricultural systems in some parts of the world; however, in this thesis the term is used as it is commonly used in Australia, where this research was conducted. Due to individual journal requirements, this term has briefly been redefined in each data chapter.

Wildlife – Throughout this thesis the term wildlife is used to refer to both vertebrate and invertebrate fauna. The term is used without suggestion of the origin or classification of biota; wildlife may be introduced or native (to Australia), threatened or common. When their origin or classification is important they are specified.

The ecosystems in which field research was conducted are often abbreviated as: YY – The You Yangs Regional Park, a woodland ecosystem. CW – Cheetham Wetlands, a wetland ecosystem.

At first mention of a species in a chapter both the common and scientific names are given. Common names are used in the text thereafter. Authorities are provided at the first mention of a plant species.

Bird nomenclature follows Christidis and Boles (2008).

xiv Abstract

Invasive species are one of the greatest threats to global biodiversity. However, while invasive, unwanted plants (commonly referred to as ‘weeds’) can detrimentally impact wildlife populations, they also can act as valuable resources for wildlife. In these situations, weed management may have unintended negative outcomes for such species. Weed managers require an understanding of the circumstances where weeds are valuable for wildlife and to identify appropriate weed management techniques that account for such relationships. This research examines the potential values of weeds for wildlife and the implications this has for weed management.

The perceptions and practices of weed managers to weed-wildlife interactions in Victoria, Australia, were evaluated using a questionnaire. Ninety percent of respondents were aware of wildlife using weeds and most (70%) had adjusted weed management programs to accommodate wildlife. However, there was no systematic approach to assessing wildlife before management, and rarely were wildlife monitored to determine project success. With much weed management aimed at biodiversity conservation and habitat improvement, the lack of wildlife assessment and monitoring during weed projects requires attention.

Weed-wildlife interactions and their management implications were examined in a Victorian 1) woodland ecosystem and 2) wetland ecosystem. Birds and small mammals in the woodland and wetland ecosystems displayed similar patterns: weeds supported some bird and small mammal species, and vegetation structure was the most important vegetation attribute in determining species occurrence and level of activity in weed habitat. When weed structure was similar to that of native vegetation, similar species occurred in both habitats. In contrast, when vegetation structure differed, weeds provided habitat for different species. This suggested the use of weed management

xv techniques that altered weed structure in these ecosystems could produce changes in the wildlife species exploiting these habitats.

‘Wildlife-friendly’ weed management (which involved retaining weed structure) was conducted experimentally in both ecosystems. Bird and small mammal occurrence and activity were not altered by weed management in either ecosystem, including the activity of bird species that favoured weeds.

This study showed that despite positive attitudes towards the value of weeds for wildlife and the management actions that take into account positive weed-wildlife interactions, weed managers lacked a systematic approach to assessing wildlife before management and when evaluating project success. This research also demonstrated that weeds provided habitat for some wildlife species in the woodland and wetland investigated, with weed structure being the most influential determinant of species occurrence. Incorporating this information into small-scale weed management programs led to successful wildlife-friendly management, at least in the short-term. Thus, consideration of wildlife before and after weed management could facilitate effective weed management that accounts for positive weed-wildlife interactions.

xvi Chapter 1

General introduction

Brown falcon Falco berigora perched on African boxthorn Lycium ferocissimum, Cheetham Wetlands, Victoria, Australia.

1 Chapter 1 – Introduction

1.1 Invasive species and conservation: A global issue

Worldwide, biological invasions are extensive in their number and variety (Vitousek 1997). Advances in technology and transport and an increasing interest in the use of exotic species has facilitated establishment of many species outside their natural range at a much faster rate than would occur naturally (Sindel 2000). While not all exotic species are dangerous to the environment in which they establish (Vitousek 1997), those that are harmful can threaten human health, economies, biodiversity and ecosystem integrity (Hobbs and Humphries 1995; Vitousek 1997). Such species have attracted global attention and are recognised as one of the greatest threats to biodiversity and ecosystems (Hobbs and Humphries 1995; Randall 1996; Williams and West 2000; Rodriguez 2006).

Weed species that invade natural ecosystems can compete with native plants for space, water and nutrients (Nelson and Wydoski 2008) and can, among other things, overrun native vegetation, changing nutrient cycling (D'Antonio and Meyerson 2002), the physical and chemical properties of soil (Neira et al. 2007), fire regimes (Vitousek 1997; D'Antonio and Meyerson 2002) and micro climates (Lindsay and French 2004). As a result, much time, effort and money has been, and continues to be, expended in attempts to reduce the detrimental impacts of these plants (D'Antonio and Meyerson 2002). However, the impacts of weeds on wildlife may not always be detrimental. This means that, in terms of wildlife, the effectiveness of weed management efforts may not be beneficial.

1.2 The impacts of weeds on wildlife

The impacts of weeds on wildlife have been studied less frequently compared with other aspects of weed ecology (French and Zubovic 1997; Adair and Groves 1998; Valentine et al. 2007; Cully Nordby et al. 2009). The efficacy of weed management in natural systems is unclear because of this knowledge gap. Given the global prevalence of invasive plants, the likelihood that interactions between weeds and wildlife are common place, is high (Goodenough 2010). When managing weeds with the intention of

2 Chapter 1 – Introduction conserving biodiversity, understanding such ‘weed-wildlife interactions’ is essential so that management can fulfil its stated aims.

The threats weeds pose to wildlife are many and varied, including: habitat modification which may be on a small (Lindsay and French 2006) or large scale (Stralberg et al. 2010); chemical toxicity of the weed species (Marshall et al. 2008); physical entanglement of wildlife (Priddel and Carlile 1995); harbouring of predators (Fulton and Ford 2001) or pest animals that compete for resources (Peter 2000); and, changing resources such as food availability (Lesica and Miles 2004).

These threats can have negative flow-on effects such as: behavioural changes, for example more energy-intensive foraging (Maron and Lill 2005); changes to parameters important in determining population viability, such as lowered reproductive success (Rodewald et al. 2010); and, overall decline in populations of desired species (Jellinek et al. 2004). Indeed, Schlaepfer et al. (2002) suggest there is a greater focus on documenting the negative effects of weeds compared to circumstances where they have no impact or have a positive influence. This thesis, therefore, focuses on weed-wildlife interactions with consideration for the potential values of weeds.

1.2.1 The values of weeds for wildlife Early research into the ecology of weed invasions gave the first indications that wildlife may, under some circumstances, benefit from weeds. Birds were quickly identified as significant contributors to the spread of weeds (Anonymous 1927); ingesting fruits and depositing their seeds elsewhere. Yet, this important aspect of weed invasion, the provision of resources to wildlife, was not recognised. More recently, the resources that weeds provide for wildlife have been considered as potentially valuable (Loyn and French 1991; Christopherson and Morrison 2004; Carlos and Gibson 2010).

Weeds can offer a wide range of feeding opportunities to other fauna. In Florida, for example, Haag et al. (1987) found that up to 50% of the winter diet of common

3 Chapter 1 – Introduction moorhens Gallinula chloropus consisted of two aquatic weeds: coontail Ceratophyllum dernersum L. and hydrilla Hydrilla verticillata (L.f.) Royle, while other common waterbirds in the area fed extensively on arthropods associated with the aquatic weed water hyacinth Echhornia crassipes (Mart.) Solms.

Vegetation structure provided by weeds also has proven beneficial for wildlife under some circumstances. This is understandable as structure can provide nesting sites (Nias 1986), shelter (Severns and Warren 2008), protection from predators (Sanderson and Kraehenbuehl 2006), and perching and roosting opportunities (Fisher and Goldney 1997). Nias (1986) demonstrated that the reproductive success of superb fairy-wrens Malurus cyaneus nesting in blackberry Rubus fruticosus L. patches was higher than those nesting in other vegetation. Nest concealment is an important factor contributing to clutch success (Nias 1986). Blackberries made nests less visible as they are thick and dense, and also have thorns that deter predators (Davies 1998).

Obviously, the literature demonstrates that weeds can be detrimental to wildlife but that, under certain conditions, there can be benefits for wildlife. In many cases, where a disadvantage or benefit has been identified in the literature, it is for a particular species, but we must consider all species within an environment. It is likely that weeds may benefit one or several native wildlife species but disadvantage one or more others (Goodenough 2010). Which species benefit, and which do not, must be understood for management to fulfil its objectives.

That weeds are able to provide resources to wildlife can be extremely important, especially in the light of native vegetation destruction and degradation, which is occurring worldwide (Fahrig 2001). In some cases, the resources provided by weeds replace that lost from such destruction and degradation, therefore the weeds effectively act as ecological substitutes for native vegetation. In California, weeds provided abundant resources for butterflies, particularly in terms of nectar, allowing for the maintenance of some populations that would otherwise have been lost due to the

4 Chapter 1 – Introduction destruction of their native habitat (Graves and Shapiro 2003). Similarly, in some regions of Australia, only one percent of native rainforest vegetation remains (Ekert and Bucher 1999). In these areas, two weeds, camphor laurel Cinnamomum camphor (L.) J.Presl. and large-leafed privet Ligustrum lucidum W.T.Aiton, occur in former agricultural land and can provide the only remaining resources to birds (Ekert and Bucher 1999).

Weeds can provide a particularly vital resource for some threatened species and the literature presents several examples of this. In riparian areas of California, there are large infestations of saltcedar Tamarix spp. which provide habitat for a variety of birds in the absence of remnant native vegetation (Sogge et al. 2008). One of these birds is the endangered southwestern willow flycatcher Empidonax traillii extimus, which frequently breeds in habitat that includes saltcedar (Shafroth et al. 2005). In Victoria, Australia, weeds, including gorse Ulex europaeus L. and spiny rush Juncus acutus L., provided the endangered eastern barred bandicoot Perameles gunnii with shelter for resting and protection from predators (Brown et al. 1991; Dufty 1994) and in South Australia, endangered southern brown bandicoots Isoodon obesulus obesulus are heavily dependent on blackberries for shelter (Sanderson and Kraehenbuehl 2006).

Another potential benefit weeds can provide is connectivity between fragmented habitats. Weeds have been suggested to act as stepping stones and corridors between rainforest patches in areas of northern Australia (Date et al. 1991; Ekert and Bucher 1999). The movement of animals (for example ‘gap-crossing’) can be prevented between patches of fragmented habitat when distances between patches are substantial or when suitable cover from predators does not exist (Arnold et al. 1993). When movements are impeded, gene flow slows or stops and, ultimately, genetic diversity of populations within fragments decreases (Coulon et al. 2004). Thus, weeds that connect fragmented patches, or act as stepping stones to reduce distances between patches, might be important in maintaining genetic diversity.

5 Chapter 1 – Introduction

In summary, weeds, under at least some circumstances, provide resources and can be important to the persistence of wildlife, suggesting that this needs to be taken into account in weed management. Not to do so may result in the decline and/or (at least local) extinction of fauna. However, it is not known if weed managers recognise the potential values of weeds for wildlife or whether they take this into account during management. If managers are to account for weed-wildlife interactions, they must first identify these interactions, the wildlife species affected and the weed attributes that influence these interactions. Understanding these can inform appropriate weed management for wildlife. Yet, it is rare that wildlife is considered when applying weed management programs (Lawrie 2002).

1.3 Impacts of weed management on wildlife

The cost, concerns and extent of weed invasion means there is a considerable volume of literature detailing the research into the best methods of weed control (Gosper 2004; Hulme 2006). Adaptive management (Figure 1.1) is considered a common approach to invasive species management and is said to be used extensively for weed management, because it is considered best practice (Shea et al. 2002).

Yet, the impact of weed management on non-target species and taxa, like wildlife, is rarely examined because it is not the priority of management (Valentine and Schwarzkopf 2008). Therefore the unintended effects of management on wildlife are relatively unknown. Considering the potential values of weeds, unintended effects could involve adverse changes to the availability and quality of food resources; perching and roosting sites; nesting structure; and, shelter and protection (see above).

6 OVERALL OBJECTIVES

SPECIFIC OBJECTIVES

PLANNING AND PREDICTION

MONITORING AND IMPLEMENTATION EVALUATION

Figure 1.1 Typical adaptive management cycle Chapter 1 – Introduction

Linz et al. (1996) demonstrated both positive and negative effects on birds upon removal of the weed cattail Typha spp. Cattail was managed with the application of glyphosate and had a positive effect on numbers of waterbirds that required open water for feeding and diving; but, numbers of marsh wrens Cistothorus palustris and red-winged Agelaius phoeniceus and yellow-headed Xanthocephalus xanthocephalus blackbirds, which require vegetated habitats for their life-cycle, were much lower in cattail managed areas (Linz et al. 1997). Thus, a mosaic of controlled and uncontrolled wetlands would be best to provide for a diversity of birds (Linz et al. 1999).

In North America, insects were introduced in an attempt to biologically control spotted knapweed Centaurea maculosa Lam. which infested semiarid environments (Ortega et al. 2004; Pearson and Fletcher 2008). The native deer mouse Peromyscus maniculatus then became dependent on these insects as a food source, and its population dramatically increased in size. Later, management of the weed reduced the number of insects, and consequently, halved the population of deer mice (Pearson and Fletcher 2008). While this reduction in a native faunal species was an indirect effect of the control of an invasive weed, it highlights the importance of considering the effects that different types of weed management can have on wildlife. Lawrie (2002) suggested a system to identify suitable weed management options to situations where wildlife inhabit weeds. These options included injecting herbicide into the stem of a weed so that a perch site remained for a bird, leaving removed weeds in piles to keep the structural complexity in a site, and using native plants to replace weeds (Lawrie 2002). However, scientific understanding of when to apply the principles outlined, or how effective they are, remains elusive and more work is needed.

Without considering the potential impacts weed management can have on wildlife, we effectively can be contributing to habitat loss if a weed has benefits to wildlife in the area. In these circumstances, when weeds benefit wildlife, gradual weed removal in conjunction with revegetation of natives has been recommended (Gosper and Vivian- Smith 2006; Carlos and Gibson 2010). At the policy level (for example, in Victoria,

8 Chapter 1 – Introduction

Australia; Environmental Weeds Working Group 2007) there seems to be little to guide weed managers as to the best weed management practices to use in these circumstances, particularly in terms of which native plant species to use for revegetation, or how to determine the most appropriate technique to use during ‘gradual’ weed removal. Selecting appropriate weed removal techniques is vital when using the gradual removal and revegetation approach because of the time lag that exists from planting and the point in time where natives have grown sufficiently to replace that which weeds provided (Gosper et al. 2006; Sogge et al. 2008). While this may be transitory, it has been suggested that it may take decades for weed removal and revegetation to restore native habitats (Nelson and Wydoski 2008).

Assessing the wildlife that use weeds before weed management commences can address questions relating to appropriate weed management when weed-wildlife interactions exist. Such assessments would highlight the aspects of weed vegetation that were most important for the wildlife using them. This could not only indicate the types of native vegetation that would be best for replacement, but which weed removal technique/s might be more sensitive to wildlife needs. Given that it is unknown whether managers take weed-wildlife interactions into account during management, it is equally unknown whether wildlife sensitive management approaches are considered or, if they are, how frequently they are used.

9 Chapter 1 – Introduction

1.4 Thesis overview

The overarching theme of this thesis centres around the potential for weed-wildlife interactions, the values which weeds may offer wildlife, and the implications this has for weed management practices. It first assesses the perceptions and practices of weed managers to gauge their awareness of weed-wildlife interactions and to examine whether management is adjusted in response to this. This thesis then goes on to assess the weed- wildlife interactions in two different ecosystems, identifying the most important attributes of these systems for wildlife. This information is then used to suggest management techniques that are sensitive to wildlife interactions with weeds. Finally, the effectiveness of these ‘wildlife-friendly’ techniques for wildlife is assessed. Therefore, this study has the potential to inform best-practice weed management such that it might account better for wildlife. Observational and experimental studies, as well as social research were involved and constitute a broad examination of the ecology and management surrounding the issue of weeds and wildlife.

1.4.1 Thesis objectives and structure This study had two main aims: to investigate weed-wildlife interactions considering the potential value of weeds for wildlife, and; to investigate the implications of weed management for wildlife. All chapters address both aims, but invariably some are focused on one aim more than the other. The order in which chapters are presented is shown in Figure 1.2.

Chapter 2 investigates whether weed managers have observed weeds providing for wildlife, whether they consider weeds are important for wildlife, whether they address this when implementing weed management, whether management objectives include wildlife, and examines management success relative to objectives. This chapter has been revised and is being considered for publication in the journal Society and Environment.

10 Chapter 1 – Introduction

Chapter 3 examines whether weeds support birds and small mammals and what attributes of weeds and other vegetation most influence these wildlife. This work is carried out in a woodland where the dominant weed reflects the native shrub structure.

Chapter 4 investigates similar concepts to those of chapter 3 but focuses on a wetland with a weedy shrub that is emergent to the low-growing native vegetation.

Chapter 5 experimentally investigates if weed management can be applied in a way that sustains the birds and small mammals that rely on weeds in these ecosystems.

Chapter 6 synthesises the preceding chapters by identifying key outcomes, their implications for weed management and directions for future research.

11 The potential values of weeds for wildlife and the potential impacts of weed management on wildlife. Chapter 1

The views, attitudes and Weed-wildlife interactions practices of weed managers and the aspects of weeds to the potential values of that are most important to weeds for wildlife. wildlife. Chapter 2 Chapters 3 and 4

The effectiveness of weed management approaches that account for wildlife. Chapter 5

Key outcomes and how they can be used to inform management. Chapter 6

Figure 1.2 Structure and flow of thesis. Chapter 1 – Introduction

1.4.2 Study areas Legislation, policy and classification relating to weeds are inconsistent between the eight States and Territories within Australia (Williams and West 2000). The research presented in this thesis is concentrated in the State of Victoria, Australia. Confining the study to one State meant that both social and field research was not influenced by jurisdictional differences in policy that occur throughout Australia. Field observations and experiments made throughout this research were focussed in 1) a woodland and 2) a wetland ecosystem, both located near Melbourne, Victoria (Figures 1.3 and 1.4).

The woodland occurs in the You Yangs Regional Park, a 2000 hectare reserve near Melbourne, Australia (37º 56’ S, 144º 25’ E). Sampling was conducted in 25 study sites in the woodland (Figure 1.3).

The You Yangs is recognised as having one of the country’s most dense populations of the invasive African shrub, boneseed Chrysanthemoides monilifera subsp. monilifera (DC.) T. Norl. (Roberts 2008). Boneseed is a ‘Weed of National Significance’ in Australia because of its ability to invade and spread causing significant economic and environmental damage (Brougham et al. 2006). It is thought to have been planted for erosion control, but has spread to more than 1300 ha (65%) of the park (Brougham et al. 2006; Roberts 2008). Boneseed now constitutes the lower and middlestorey vegetation in much of this area (Figure 1.5), out competing native species (Roberts 2008).

Throughout the woodland the canopy consists of a variety of Eucalyptus tree species, with sapling-sized black wattle Acacia mearnsii De Wild. and golden wattle Acacia pycnantha Benth. also common throughout the park, often growing in dense stands below the eucalypt canopy. Concerted weed management has meant that some areas of the woodland are not dominated by boneseed. These areas remain either shrub free or are dominated by the native shrubs snowy mint bush Prostanthera nivea A.Cunn. ex Benth. and drooping cassinia Cassinia arcuata R.Br. (Figure 1.5).

13 Chapter 1 – Introduction

The second study area, the wetland, lies within Point Cook Coastal Park along Port Phillip Bay, southern Victoria (37º 54’ S, 144º 48’ E). Cheetham Wetlands operated as a commercial saltworks until 1992, but now the shallow evaporation ponds and adjacent areas are managed for biodiversity (Antos et al. 2007). Sampling was conducted in 18 study sites throughout the wetland (Figure 1.4).

The dominant vegetation community, extending to varying degrees around the edges of most ponds, is coastal saltmarsh (Cook 1996; Figure 1.6). While this vegetation is dominated by native plant species, such as shrubby glasswort Tecticornia arbuscula (R.Br.) K.A. Shepherd & Paul G.Wilson, many weed species also occur.

African boxthorn Lycium ferocissimum Miers. is a particularly prominent weed in this ecosystem because it stands much higher than the native, low-growing saltmarsh vegetation (Figure 1.6). It is a clear emergent and adds another level of structural complexity to the habitat. Boxthorn is listed as a Weed of National Significance in Australia (Magnussen undated) and, as such, is listed as a high priority for control in Cheetham Wetlands (Smedley and Nance 2010). It is found in isolation or dense stands in many of the terrestrial areas of the wetland (Smedley and Nance 2010). It out- competes native vegetation, acts as shelter for pest animals and can impede access for people and vehicles (Muyt 2001).

14 Figure 1.3 Woodland ecosystem of the You Yangs Regional Park, Victoria, Australia. Twenty-five study sites are shown. Created by Emma Carlos. Figure 1.4 Wetland ecosystem of Cheetham Wetlands, Victoria, Australia. Eighteen study sites are shown. Created by Emma Carlos. Figure 1.6 Typical vegetation in the woodland ecosystem of the You Yangs Regional Park. Left: Boneseed dominated vegetation; Middle: Native vegetation without a shrub layer; Right: Native shrub vegetation dominated by the native shrub snowy mint bush.

Figure 1.5 Typical vegetation in the wetland ecosystem of Cheetham Wetlands. Left: Native saltmarsh vegetation dominated by shrubby glasswort; Right: Boxthorn standing emergent of low-growing native saltmarsh vegetation Chapter 2

Weeds and wildlife: Perceptions and practices of weed managers

Yellow-faced honeyeater Lichenostomus chrysops nest in African boneseed Chrysanthemoides monilifera subsp. monilifera.

18 Chapter 2 – Perceptions and practices of weed managers

2.1 Abstract

Weeds (plants that are invasive) have a wide range of negative impacts on biodiversity, yet can provide wildlife with key habitats and resources. Thus, weed removal has the potential to adversely affect wildlife, but whether this is considered during weed management is poorly known. To determine the extent of this we examined the perceptions of weed managers regarding wildlife and weed management in Victoria, Australia. We found 90% of managers were aware of wildlife using weeds and that most (70%) adjusted management programs to accommodate wildlife. Despite this, few (19%) had adopted the currently recommended response of combining gradual weed removal with revegetation. Incorporation of the monitoring of native vegetation into management programs was more common than incorporation of the monitoring of wildlife, which was rare. This highlighted the need for management to better respond to weed-wildlife relationships. If the improvement of wildlife habitat is included in the objectives of weed programs, as it should be, then wildlife should be incorporated into project monitoring. This would lead to a greater understanding of the role weeds and their management have in each situation and, ultimately, more informed decision making.

Keywords: Weeds; wildlife; weed managers; perceptions; habitat; monitoring.

19 Chapter 2 – Perceptions and practices of weed managers

2.2 Introduction

Weeds are unwanted plants that can have detrimental effects on the economy or environment (Richardson et al. 2000; Natural Resource Management Ministerial Council 2006). They invade agricultural, urban and natural systems (Richardson et al. 2000). They have a wide range of ecological effects, but best documented in natural systems are their negative effects on native plants, such as reduction of their abundance, diversity, recruitment, pollination and species survival (Randall 1996; Gibson 2010). Weeds also detrimentally affect wildlife (vertebrates and invertebrates) (Bailey et al. 2001; Fulton and Ford 2001; Jellinek et al. 2004; Valentine et al. 2007), important ecological processes such as fire regimes (Shafroth et al. 2005) and water flows (Griffin et al. 1989), and they can change soil properties (Neira et al. 2007). Thus, weeds are considered one of the most significant threats to biodiversity conservation worldwide (Hobbs and Humphries 1995; Rodriguez 2006; Bremner and Park 2007; Funk and Vitousek 2007) and extensive resources have been allocated to their management (Ewel and Putz 2004; Sinden et al. 2004; Pimentel et al. 2005).

Despite their negative effects, globally, there is growing evidence that weeds benefit wildlife under certain circumstances. Weeds may represent a food source (Lawrie 2002), provide habitat for breeding (Nias 1986), roosting and perching (Fisher and Goldney 1997), and refuge from predators (Brown et al. 1991; Sanderson and Kraehenbuehl 2006). Weeds become particularly important for wildlife when alternative native habitat is limited (Sutter et al. 1995; Graves and Shapiro 2003), especially for threatened species (Date et al. 1996; Sanderson and Kraehenbuehl 2006; Schmidt et al. 2009). For example, in areas of southern Australia, weeds characterise key habitat of the threatened southern brown bandicoot Isoodon obesulus obesulus (Schmidt et al. 2009). Management of these weeds would adversely impact the bandicoots (Schmidt et al. 2009), yet weed management is required by law in these areas (Sindel 2000). In California in the United States, weeds provide some butterflies with vital resources such that effective weed

20 Chapter 2 – Perceptions and practices of weed managers management would result in the disappearance of the butterflies (Graves and Shapiro 2003). Thus, managers face complex decisions as to when and how to manage weeds.

Covering more than 200 000 km2, the state of Victoria, Australia, represents a jurisdiction where weeds are a significant management issue. Total costs of weed management are among the highest in the continent (Commissioner for Environmental Sustainability 2008). The primary objective of managing weeds in the natural environment of Victoria is to maintain or promote indigenous biodiversity (Environmental Weeds Working Group 2007). Government guidelines developed to monitor biodiversity changes associated with weed invasion (Ainsworth et al. 2008) involve monitoring native vegetation but not wildlife. Wildlife are a prominent form of biodiversity and often attain iconic status among the general public (Martín-López et al. 2007). An understanding of how their needs are incorporated in weed management planning and implementation is therefore desirable. To assist with this, it is important to consider the types of weeds and methods of management that are undertaken so we can better understand and assess any interactions and impacts involving wildlife.

Practitioner and stakeholder perceptions of weeds and their management are available for parts of Europe, the Mediterranean and Australia (Bardsley and Edwards-Jones 2006; Bremner and Park 2007; King 2007; Garcia-Llorente et al. 2008; Andreu et al. 2009; Reid et al. 2009) but this has not involved wildlife.

We assess whether weed managers in Victoria consider wildlife during weed management programs by addressing the following five key questions. 1) Is weed management carried out for biodiversity, social or economic objectives? 2) Are weed managers aware of wildlife using weeds? 3) What are the attitudes of weed managers towards the idea that weeds potentially provide habitat for wildlife? 4) What are the management responses of weed managers in relation to wildlife during the implementation of management programs? 5) Are positive outcomes for wildlife incorporated into weed management objectives and are these outcomes measured when assessing project success?

21 Chapter 2 – Perceptions and practices of weed managers

2.3 Methods

We defined weed managers as organisational representatives who had some substantial role in the management of weeds in Victoria. Those surveyed worked in different regions of Victoria and many had been involved in weed management in multiple regions of this state. Most (72%) had worked on weed management in the Port Phillip and Westernport region that surrounds Melbourne. Weed managers from federal, state, and local governments, community groups, non-government organisations (NGO) and private companies were targeted with the expectation that they were likely to be guided by the objective of promoting indigenous biodiversity. These groups therefore were also expected to have greater experience in management of weeds in the natural environment (environmental weeds). Results from those who had experience in managing weeds in the agricultural industry (agricultural weeds) were not excluded from analysis, but farmers were not specifically targeted as biodiversity was not expected to be their primary focus.

Questionnaires (Appendix A) were sent to potential respondents in each group, and the snowball effect used to recruit further participants (Heckathorn 2002; Salganik and Heckathorn 2004; Walker and Brammer 2009). This means that response rates are unavailable, but respondents were well spread across the different target groups. Our overall sample size (n = 81 respondents) reflects that achieved of research that had similar target respondents (for example, Bardsley and Edwards-Jones 2006; Andreu et al. 2009). For individual survey questions, sample size varied as not every respondent answered every question.

The questionnaire was piloted and potential problems (for example ambiguities) rectified. The final questionnaire, with 13 pages and 29 questions, was distributed to participants (in 2009); with the option of participating in a random draw for a prize (~$A100). The questionnaire consisted of three sections (Appendix A). The first explored demographics and experience in weed management. The second section investigated the types of weeds managed, the management techniques employed, the

22 Chapter 2 – Perceptions and practices of weed managers objectives of weed management, and the success in attaining the objectives. The third section documented observations made of wildlife during weed management projects, attitudes towards weeds providing habitat for wildlife, and whether adjustments were made during weed management projects to accommodate wildlife. The questionnaire consisted of closed-ended (87%) and open-ended questions. Some (34%) closed-ended questions involved a five point Likert scale (from one [never] to five [always]) to examine observations, or a four point Likert scale (from one [very unimportant] to four [very important]) to examine attitudes.

A composite variable ‘experience’ was created by combining two ordinal variables (the length of time that a respondent had been involved in weed management [< 1 year, 1 – 5, 6 – 10, 11 – 15, 16 – 20, > 20 years] and the frequency that they were involved in weed management at the time [never, annually, monthly, weekly, daily]; after Manning and Munro 2006). These categories were changed to ranks (with the lowest numbers for categories representing least experience, for example, < 1 year and never). The ranks for each question were then summed to indicate overall experience. Higher scores indicated greater experience in weed management.

The term ‘weed’ was used broadly throughout the questionnaire, reflecting the Australian usage of the word which incorporates all contexts. When location or origin of the weed was important the context of the weed was specified as: agricultural weed (invasive and harmful in agricultural environments); environmental weed (invasive and harmful in natural environments); native weed (originating from Australia); exotic weed (originating from outside Australia). The terms native and non-native also were used to categorise wildlife, which describe their origin in relation to Australia.

23 Chapter 2 – Perceptions and practices of weed managers

2.3.1 Statistical analysis Standard non-parametric statistical analyses were used throughout to reflect the ordinal nature of the data (Quinn and Keough 2002) using SPSS (v. 17.0, SPSS Inc., Chicago). On occasion we present means and confidence intervals (± 95%) to clarify differences between variables, even though statistical tests were conducted on ranks or medians. Non-metric multi-dimensional scaling (MDS), analysis of similarities (ANOSIM) using Bray-Curtis resemblance matrices and similarity percentages (SIMPER) were performed using PRIMER 5 (Plymouth Marine Laboratory, UK) as appropriate (excluding outliers).

2.4 Results

Most respondents (n = 81) were from state government organisations (36%) while others were from contracting organisations (19%), NGO (14%), community groups (12%), local governments (11%) and private landholders (6%). Consultancy (1%) and federal government (1%) organisation types were excluded from analysis due to inadequate sample sizes. Most respondents were male (67%) and were aged 18 – 25 years (16%), 26 – 35 (22%), 36 – 45 (25%), 46 – 55 (25%), and > 55 years (12%). Experience varied between respondents (experience score 3 – 11, median = 7), and did not differ with organisation type (Kruskal-Wallis, Ȥ2 = 3.921, df = 5, p = 0.561).

Respondents had been involved more frequently in the management of environmental rather than agricultural weeds (Table 2.1), a product of our sampling. They had more frequently managed exotic rather than native weeds; shrubs were the most frequently managed weed type compared with trees, forbs, vines and grasses; and large weed infestations were managed less often than medium and small infestations (Table 2.1). Respondents had been involved more frequently in weed management projects where cutting and painting of weeds or manual control was conducted compared with other weed management techniques (Friedman, Ȥ2 = 271.956, df = 8, 80, p < 0.001; Table 2.2).

24 Chapter 2 – Perceptions and practices of weed managers

Environmental weeds were considered more important to manage than agricultural weeds, and exotic weeds more important to manage than native weeds (Table 2.1). Trees were considered less important to manage than shrubs, forbs, grasses and vines and there was no difference in the perceived importance of the management of infestations of different sizes (Table 2.1).

2.4.1 Observations of wildlife during weed management Most respondents had observed wildlife using weeds before (90%; n = 74) and after (73%; n = 65) weed management. The proportion of respondents observing the use of ZHHGVE\ZLOGOLIHGLGQRWGLIIHUEHWZHHQRUJDQLVDWLRQW\SH EHIRUH)ULHGPDQȤ  6.218, df S DIWHU)ULHGPDQȤ GI S  EXWWKRVHZKR noted wildlife using weeds were significantly more experienced than those who did not (Mann-Whitney, z = 99.000, df = 73, p = 0.010). Wildlife seen using weeds included birds (92%), reptiles (55%), mammals (48%), invertebrates (36%) and amphibians (22%) (n = 67 respondents). Weeds were used for sheltering/roosting (97%), movement pathways (81%), breeding (77%) and feeding (75%) (n = 67). Respondents observed non-native wildlife using exotic weeds more often than they used native weeds, a pattern not evident for native wildlife (Table 2.1; Figure 2.1). Both native and non-native wildlife were observed more frequently using environmental rather than agricultural weeds and in shrubs more frequently than trees, grasses, vines and forbs. There was no difference in the frequency that native wildlife were observed in different sized weed infestations; but non-native wildlife were observed using small weed infestations less often than medium or large infestations (Figure 2.1).

2.4.2 Attitudes towards wildlife during weed management When respondents indicated how important weeds were as habitat for wildlife, there were more ‘unknown’ responses for questions regarding non-native amphibians, reptiles and invertebrates, compared with other taxa (Friedman, Ȥ2 = 73.48, df = 9, p < 0.001; n = 73), so these responses were excluded from further analyses. There was a significant difference in the perceived importance of weeds as habitat for different wildlife groups

25 Chapter 2 – Perceptions and practices of weed managers

(Friedman, Ȥ2 = 89.381, df = 6, p < 0.001; n = 73). Most respondents considered weeds were important for native birds (78%, n = 69), non-native birds (83%, n = 68) and non- native mammals (78%, n = 66) compared with native mammals (60%, n = 64), reptiles (52%, n = 63), invertebrates (49%, n = 62) and amphibians (41%, n = 58). The reasons respondents believed that weeds were important habitat for wildlife fell into nine categories (n = 62 specific responses; Table 2.3).

2.4.3 Adaptive management for wildlife Most respondents (84%) considered adjusting weed management to accommodate wildlife as ‘important’ or ‘very important’ regardless of their experience (median scale score = 3 [important]; Kruskal-Wallis, Ȥ2 = 3.505, df = 3, p = 0.320; n = 73 excluding ‘unknown’ responses). Most respondents (71%; n = 73) indicated they adjusted weed management in some way to accommodate wildlife, especially the more experienced respondents (Mann-Whitney, z = 326.500, df = 72, p = 0.007). There was no difference in the number who adjusted weed management projects between organisation type (Friedman, Ȥ2 = 4.288, df = 5, 1, p = 0.509). A variety of adjustments were reported, for example, revegetation and changing the timing of management (Table 2.4). Some (38%) reported a combination of adjustments such as combining revegetation with the gradual removal of weeds. Of those who had observed wildlife using weeds, 19% cited this particular combination.

26 Table 2.1 Statistical comparisons between different weed categories/characteristics in relation to: the frequency respondents had been involved in management of these weeds; how important respondents believed management was for these weeds; and the frequency respondents had observed the use of these weeds by native and non-QDWLYHZLOGOLIH)ULHGPDQ Ȥ2) or Wilcoxon Signed Rank (z) non- parametric tests were used to compare responses between different weed categories/characteristics. Test statistics in bold indicate that responses were significantly different (p < 0.05) between weed categories/characteristics. The numbers of responses in each weed category/characteristic varied and are presented below the corresponding statistic in parentheses. They are in the same order that the weed categories/characteristics are listed in their associated column headings. When the number of responses was equal for each weed category/characteristic, this number is listed alone. Weed category/characteristic Environmental vs Exotic vs native Shrubs vs trees vs forbs vs Large vs medium vs agricultural (z) (z) vines vs JUDVVHV Ȥ2) VPDOOLQIHVWDWLRQV Ȥ2) Frequency weeds were -5.498 -5.402 34.374 56.520 managed (73; 78) (73; 78) (76) (77) Importance of management -5.811 -5.452 18.439 2.198 of weeds (76; 79) (76; 78) (76) (78) Observations of native -0.131 5.725 159.141 0.282 wildlife using weeds (62; 67) (65; 66) (67; 67; 66; 66; 65) (62; 65; 65) Observations of non-native -3.517 -0.886 109.194 9.361 wildlife using weeds (64; 67) (66; 67) (67; 67; 67; 65; 64) (63; 65; 65) Figure 2.1 Mean scale scores (± 95% CI) of the ranked frequency that respondents (n = 67) observed native (circles) and non-native (triangles) wildlife using weeds of different characteristics and in different contexts. Higher scale scores reflect higher frequencies of usage. Table 2.2 Weed management techniques used by managers in Victoria, Australia (n = 81), with mean (± 95% CI’s) scale scores (1 = never to 5 = always). Management technique Mean frequency Cut and painting 3.9 ± 0.2 Manual control 3.6 ± 0.2 Mechanical control 3.4 ± 0.2 Grazing 3.0 ± 0.2 Chemical control 2.9 ± 0.2 Biological control 2.3 ± 0.2 Shading/solarising 2.1 ± 0.2 Fire 1.8 ± 0.2 Prevention 1.7 ± 0.1

Table 2.3 Beliefs cited by respondents (n = 62) to justify their view that weeds provide important habitat to wildlife. Beliefs Percentage of respondents citing belief Alternative native habitat not available 32 Weed provides resource/s to wildlife 27 Species of wildlife using the weed is under threat 16 Overall threat status of the weed is low 8 Potential for replacement of weed with suitable native 8 species is low Other 7 Table 2.4 Type of adjustments managers had made to weed management in order to better accommodate wildlife (n = 93 responses from n = 50 respondents). Type of adjustment for weed management Percentage of respondents

Revegetation/replacement with native vegetation 34 Changing the timing of management (including avoiding 32 animal breeding seasons) Staging the removal of weeds 32 Retaining weed structure 26 Avoiding removal of important weeds 22 Herbicide selection or avoidance 20 Using an alternative weed management method 12 Pre-monitoring 6 Reducing noise 2

2.4.4 The objectives and success of weed management Weed management projects were conducted most often for environmental and biodiversity objectives (Friedman, Ȥ2 = 62.493, df = 2, p < 0.001; mean scores were: environmental/biodiversity, 3.6 ± 0.2; social, 2.4 ± 0.2; economic, 2.9 ± 0.3; n = 78). Objectives did not differ between organisation type, as shown by MDS and tested by ANOSIM (Global R = 0.162, p = 0.001, n = 75; R test statistics for pair-wise comparisons: -0.105 – 0.346) and SIMPER (average dissimilarity between groups: 16.97% – 20.76%). Although the ANOSIM result was statistically significant, the small Global R statistic suggests this difference was minimal (Clarke 1993). Respondents held different views in relation to the relative success of attaining different objectives (Friedman, Ȥ2 = 6.759, df = 2, p = 0.034; mean scores were: environmental/biodiversity, 3.3 ± 0.2; social, 3.1 ± 0.3; economic, 3.3 ± 0.3; n = 78). Views of the relative success of attaining these objectives were similar between Chapter 2 – Perceptions and practices of weed managers organisation types, as shown by MDS and tested by ANOSIM (Global R = 0.121, p = 0.023, n = 75; R test statistics for pair-wise comparisons: -0.079 – 0.403) and SIMPER (average dissimilarity between groups: 18.45% – 26.85%). Although the ANOSIM result again was statistically significant, the Global R statistic was so small that we do not regard it as a true measure of difference.

A variety of indicators were used to measure the success of weed management projects. These were categorised into biodiversity attributes (measuring native vegetation or wildlife populations), weed attributes (measuring number, cover or area of weeds), targeted programs (implementation of monitoring or follow up programs) and anthropogenic measures (amount of money or time spent on project). The frequency that different indicator categories were used to determine success varied (Friedman, Ȥ2 = 9.362, df = 3, p = 0.025; mean scores were: biodiversity attributes, 3.4 ± 0.2; weed attributes, 3.2 ± 0.2; targeted programs, 3.2 ± 0.2; anthropogenic measures, 2.8 ± 0.3; n = 75). This did not differ between organisation type (based on MDS and ANOSIM, Global R = 0.012, p = 0.396, n = 73). When considering only biodiversity attributes, wildlife presence was measured less frequently (mean: 2.6 ± 0.2) to gauge management success than was the presence of native plant species (3.7 ± 0.2; Friedman, Ȥ2 = 75.200, df = 3, p = 0.000).

2.5 Discussion

2.5.1 Weed management projects: purpose, perceptions and practice The promotion of biodiversity (including wildlife) through weed management is a worldwide phenomenon (Randall 1996; D'Antonio and Meyerson 2002; Reid et al. 2009). In our study, biodiversity conservation was a key objective of weed management in Victoria, regardless of the organisation conducting the weed management. While we had targeted our research towards managers that we expected to be guided by the objective of promoting indigenous biodiversity, the organisations we sampled varied in capacity and role with regard to weed management. For example, NGOs are often limited by funding (McNeely and Weatherly 1996) but have an abundance of labour via 31 Chapter 2 – Perceptions and practices of weed managers volunteers (Weston et al. 2003). Arguably, their strategies and project planning may be more limited than those of better funded entities (Curtis and Lockwood 2000). Yet we revealed similar objectives, perceptions of success and project evaluation across organisations of all types. This may result from the influence of widespread devolved government funding for weed management, associated with mandated project management standards as well as being guided by government strategies (for example, Natural Resource Management Ministerial Council 2006).

A very different result may have occurred had we captured the views of those more experienced in agricultural weed management. Given the effects of weeds in agricultural systems are widely referred to in terms of economic losses (van der Meulen et al. 2007), we would expect this to be reflected in their management objectives. Thus, the views of those more involved in agricultural environments are an important direction for future research.

That respondents had more frequently been involved in management of environmental weeds may go towards explaining the more frequent use of selective management methods (cutting and painting, manual control; Muyt 2001). These lower impact methods are more suitable in natural environments where non-target impacts on biodiversity are a concern (Muyt 2001), compared to more general chemical control. That wildlife is a potential non-target impact is highlighted by respondents who had adjusted their weed management projects to accommodate wildlife by better selecting or avoiding herbicides. Internationally, the use of lower impact methods are common (Andreu et al. 2009), but just how often considerations around wildlife influence choice of control method remains unknown.

A range of other factors also influence the choice of weed management methods. These include weed species, location and infestation size (Sindel 2000). Throughout the rest of Australia, herbicide is primarily used to manage Weeds of National Significance (WoNS), which threaten both agricultural and natural environments (Reid et al. 2009). In such cases, and particularly in an agricultural setting, more broad scale techniques are

32 Chapter 2 – Perceptions and practices of weed managers evidently more applicable. Given the large areas that can be managed with such techniques (Reid et al. 2009), it is unlikely that wildlife would remain completely unaffected; however, without their inclusion in project evaluation, this will go unnoticed.

Appropriate evaluation and monitoring of weed management projects is critical to refine programs and to permit adaptive management (Reid et al. 2009). It is now generally accepted that outcome-orientated evaluation is preferable to output-orientated evaluation (Downey 2011). In our study, managers more often used measures of biodiversity (outcome-orientated reporting) to assess the success of a project, yet output-orientated reporting (that included measuring weed attributes and implementation of targeted programs) was also used. Such a mix of reporting modes is to be expected given the need for transparency in project management; however, when managing weeds for biodiversity objectives, a greater emphasis on outcome orientated evaluation could also contribute to further understanding weed-wildlife relationships.

2.5.2 Weeding for wildlife Social research suggests that direct experience with a situation strongly influences intentions to engage in certain behaviours (Regan and Fazio 1977; Homer and Kahle 1988; Fulton et al. 1996). Our research demonstrated that more experienced weed managers frequently observed wildlife using weeds and made adjustments during weed management projects to accommodate wildlife. Additionally, some respondents based their attitudes regarding the importance of weeds for wildlife on whether alternative native habitat was available (32%) and/or whether the weed was providing resources to wildlife (27%); both of which require observation or information. This implies that the managers sampled here rely on their own observations and experience to guide their decisions regarding whether or not to adapt weed management for wildlife. Unfortunately, there are potential problems with this. Firstly, managers may inadvertently manage only for conspicuous wildlife, such as birds and mammals (which they most frequently observed). Secondly, the types and complexity of interactions

33 Chapter 2 – Perceptions and practices of weed managers between wildlife and weeds may be underappreciated. Understanding the nature of these interactions would better inform managers on the impacts that weed management could have. This represents a challenge to ecologists worldwide to better document and communicate what is and is not known about the use of weeds by wildlife.

Current biodiversity monitoring guidelines in Victoria include quantitative and systematic measurements of vegetation and its attributes (Ainsworth et al. 2008). A similar systematic approach to measuring wildlife attributes would better account for the presence of wildlife in weeds. Instead, our study suggests the presence or absence of alternative native habitat is often used by managers as the basis to decide whether or not adjustment to the weed management program is needed. This approach disregards the deleterious effects associated with displacement of wildlife (Wolff et al. 1998), assumes there is adequate habitat recognition, and is generally ‘assumption rich’. Given the experience-attitude-behaviour link and the global prevalence of weeds, it is likely that these types of assumptions are being made outside Victoria when a weed-wildlife interaction is recognised.

A good understanding of the impacts of weed management on wildlife is essential for managers to plan the most appropriate strategies. This has been highlighted in southwestern United States where management of saltcedar Tamarix spp. is largely influenced by native breeding birds that use the weed as habitat (Sogge et al. 2008). For example, the endangered southwestern willow flycatcher Empidonax traillii extimus is negatively impacted by biological control of saltcedar, which causes high levels of defoliation of the plant and renders it unsuitable for nesting. In these circumstances, smaller scale herbicide application or mechanical removal has been recommended (Sogge et al. 2008).

There is a clear need for more studies concerning weed control and its impacts on wildlife, including specific studies of techniques which may mitigate any impacts. Indeed, little is known about what wildlife may be particularly dependent upon weeds, and what life history attributes render them vulnerable or resilient to weed management.

34 Chapter 2 – Perceptions and practices of weed managers

Currently, a combination of revegetation and gradual weed removal in response to the presence of a weed-wildlife relationship is recommended (Gosper and Vivian-Smith 2006; Sogge et al. 2008; Carlos and Gibson 2010). This is to account for the time lag that occurs before revegetation sufficiently allows for replacement of weed resources. Time lags can be substantial; for example, five year old revegetation does not support the same richness or abundance of birds as adjacent weed vegetation (Carlos and Gibson 2010). Thus, weed management projects that include revegetation would need to be conducted over considerable time periods. A lack of long term funding and support could help to explain why, in our study, only 19% of managers had combined the two practices or, for example, why restoration occurs in only 29% of previously invaded sites in Spain (Andreu et al. 2009). Alternatively, if the experience-attitude-behaviour link applies, this practice may be infrequent merely due to the lack of proper assessment of potential weed-wildlife interactions.

The very process of weed management itself may reveal hitherto unknown species of wildlife relying on weeds as habitat, thus adaptive management with ongoing monitoring (Reid et al. 2009; Downey 2011), which includes wildlife, seems prudent. While funding for weed management is already limiting (Downey 2011), the suggestion to incorporate another element into the management process may not appeal. It is clear, however, that most managers are aware of wildlife in weeds and many are already taking steps to adjust their management, despite the finding that there is significantly less formal monitoring of wildlife compared to native vegetation. By conducting ‘fauna- inclusive’ monitoring, adaptive management could be better informed, avoiding any unintended deleterious outcomes. Alternatively, a re-assessment of the objectives of weed management may be required, recognising conservation of floral diversity as the objective of weed management, rather than the broader aim of biodiversity conservation.

35 Chapter 2 – Perceptions and practices of weed managers

2.6 Appendix A

Structure and content of questionnaire sent to respondents. Questions asked are numbered and respondents were asked to either: select the most appropriate option (indicated as (1) with options listed in italics); select as many options as apply (indicated as (#) with options listed in italics); select how often they had observed the listed items, from 1 – 5 or unknown on a Likert scale (indicated as (1 – 5) with items listed in italics); select how important they believed the listed items, from 1 – 4 or unknown on a Likert scale (indicated as (1 – 4) with items listed in italics); write a response (indicated as (_)).

General information 1. What is your age? (1) 18 – 25; 26 – 35; 36 – 45; 46 – 55; >55 2. Are you: (1) Male; Female 3. What type of organisation are you primarily involved with when managing weeds? (1) Local government; State government; Federal government; Community group; Contractor; Consultancy; Private land holder; Non-Government; Organisation; Other 4. What best describes your role/s in the above organisation? (#) Planner; Manager; Ground crew; Advisor; Volunteer; Other 5. How long have you been involved in weed management? (1) < 1 year; 1 – 5 years; 6 – 10 years; 11 – 15 years; 16 – 20 years; > 20 years 6. Overall, what scale best describes the area over which you are most often involved in weed management? (1) < 10 ha; 10 – 100 ha; 101 – 500 ha; Regional level; State level 7. How often are you involved in weed management in your current position? (1) Daily; Weekly; Monthly; Annually; Never

Section 1: Weed management

36 Chapter 2 – Perceptions and practices of weed managers

8. Where, across Victoria, have you been involved in managing weeds? (#; a map was provided)

Mallee; Wimmera; Glenelg Hopkins; North Central; Corangamite; Goulburn Broken; North East; East Gippsland; West Gippsland; Port Phillip & Western Port 9. In what vegetation types have you managed weeds? (#) Agricultural; Pastoral; Rainforest; Dry forest; Heathlands; Wet forest; Woodland; Alpine; Grasslands; Coastal; Wetland; Riparian; Other 10. Based on the weed management projects that you have been involved in, how often are the following types of weed management strategies conducted? (1 – 5) Manual control (by hand) Mechanical control (with aid of machinery) Chemical control Cut and Painting Shading/Solarising Release of biological control Fire Grazing Prevention Other 11. Based on the weed management projects you have been involved in, how often have the following weed categories been managed? (1 – 5) Agricultural weeds Environmental weeds Native weeds Exotic weeds Trees (e.g. pittosporum, pine) Shrubs (e.g. gorse, blackberry) Grasses (e.g. pasture grasses, serrated tussock) Vines (e.g. Japanese honeysuckle) Forbs (e.g. Paterson’s curse, dandelion) 37 Chapter 2 – Perceptions and practices of weed managers

Large infestations (covering >1 ha) Medium infestations (cover 0.25 – 1 ha) Small infestations (cover <0.25 ha) 12. Based on their negative impacts, from your accumulated knowledge or direct experience, how important do you think it is to manage the weed categories listed in question 11? (1 – 4) 13. Based on your accumulated knowledge or direct experience, indicate how often you think management is carried out for each of the reasons listed below. (1 – 5) To prevent the spread of a weed To eradicate a weed To help conserve biodiversity To improve growth of native vegetation To improve habitat quality for wildlife To improve ecosystem function To enable flow of water To help remove refuges for pest animals (e.g. foxes) To reduce fire fuel loads To decrease snake numbers To improve aesthetics To improve a view To be a good neighbour To improve recreation To have a social event To improve yields/economic values To meet requirements of the policy of your organisation To access available funding To project an image of environmental responsibility for your organisation or property To qualify for government accreditation (e.g. Land for Wildlife) or funding To provide activities for your staff or volunteers Other 38 Chapter 2 – Perceptions and practices of weed managers

14. Based on your accumulated knowledge or direct experience, indicate how often you think that projects with the objectives listed in question 13 have been successful? (1 – 5)

15. In your opinion, what makes a weed project successful? (_) 16. Based on the projects you have been involved in, how often is the success of weed management measured from the following? (1 – 5) From the presence of non weed plant species From the presence of specific animal species From a targeted monitoring project When the cover of the target weed has reached a pre-determined level From the success of the naturally regenerating native plants From the success of a revegetation programme When a follow up programme has been implemented When a certain amount of time has been spent in an area When the project funds have been used When the weed has been eradicated from an area When a pre-determined area of weeds have been removed or treated When a set number of plants have been removed or treated Photo point monitoring When the weed no longer looks like a problem Other 17. Based on the weed management projects you have been involved in, how often is ‘follow up’ work carried out? (1 – 5)

Section: Wildlife and weeds 18. Have you ever observed wildlife using weeds that have not had any management applied to them? (1) Yes; No

39 Chapter 2 – Perceptions and practices of weed managers

19. Have you ever observed wildlife using these weeds after weed management has been applied? For example, weeds that have been sprayed or piles of removed weeds? (1) Yes; No 20. What wildlife have you observed using the weeds you have been managing? (#) Mammals; Birds; Amphibians; Reptiles; Invertebrates; Feral (non-native) animals; Others. 21. How often have you seen native wildlife use the weed categories listed in question 11? (1 – 5) 22. How often have you seen non-native wildlife use the weed categories in question 11? (1 – 55) 23. What have you observed wildlife doing on or within weeds? (#) Feeding; Sheltering/roosting; Nesting; Moving through; Other. 24. Have you ever adjusted your weed management programmes because of wildlife? If yes, how and why? (_) 25. How important do you think weeds are as habitat for the following wildlife? (1 – 4) Native birds Native mammals Native amphibians Native reptiles Native invertebrates Non-native birds Non-native mammals Non-native amphibians Non-native reptiles Non-native invertebrates 26. What are your main reasons for deciding how important weeds are as habitat? (_) 27. Overall, how important do you think it is to adjust weed management to accommodate wildlife? (1 – 4) 28. What is your main reason/s for deciding how important it is to adjust weed management to accommodate wildlife? (_) 40 Chapter 2 – Perceptions and practices of weed managers

29. What have you observed after weed removal in terms of wildlife changes? (#) Loss of a native species; Loss of a non-native species; Gain of a native species; Gain of a non-native species; Increase in abundance of native species; Increase in abundance of non-native species; Decrease in abundance of native species; Decrease in abundance of non-native species; No change.

41 Chapter 3

Vegetation structure, not weediness, influences wildlife habitat in a southeastern Australian woodland remnant

African boneseed Chrysanthemoides monilifera subsp. monilifera in flower.

42 Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.1 Abstract

Invasive, unwanted plants (weeds) are recognised for their detrimental impacts, but their effects on wildlife vary. The assumption that weeds should be managed to improve wildlife habitat could lead to undesirable outcomes for any wildlife using them. To understand whether, and how, weeds provide habitat for wildlife, bird and small mammal assemblages were compared in weed and native dominated understorey vegetation in a woodland remnant in southeastern Australia. Structure of the vegetation, and not weed cover, influenced native bird assemblages. Different bird species were influenced by different structural aspects of the vegetation. The activity of the introduced house mouse Mus domesticus also was influenced by vegetation structure rather than weediness. The assumption that large-scale weed removal would quickly result in widespread improvements to wildlife habitat in this system is, therefore, questionable. By considering wildlife before weed management, wildlife that use weeds can be identified and weed management implemented in a way that accounts for these species.

Keywords: Weeds, wildlife, habitat, boneseed, birds, weed management, woodland management.

43 Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.2 Introduction

Invasive, unwanted plants (commonly referred to as weeds) are often managed in natural ecosystems to improve habitat for biodiversity, and this can attract substantial private and government investment (Sinden et al. 2004). Yet, the impacts of weeds and their management on wildlife (an important component of biodiversity) and their habitats are not well understood (Sax et al. 2005). The available studies have reported mixed effects of weeds on wildlife. For example, by changing micro climates weeds have altered invertebrate communities (Lindsay and French 2006); by providing fruit, weeds have enabled the persistence of omnivorous nest predators in usually fruit poor areas (Fulton and Ford 2001); weeds have facilitated movement of threatened birds by acting as stepping stones and corridors in fragmented landscapes (Date et al. 1996), and; provided habitat for native rodents (Christopherson and Morrison 2004), pest species (Peter 2000) and endangered mammals (Sanderson and Kraehenbuehl 2006).

The fact that weeds can potentially provide resources to wildlife is an important consideration in light of worldwide native vegetation loss and habitat destruction (Fahrig 2001). Weeds that provide for wildlife in place of native vegetation are, in effect, an ecological substitute. Such potential benefits of weeds need to be considered during weed management so that their management does not cause further loss of habitat or resources. Adjustments to weed management have been recommended when valuable weed-wildlife interactions are observed, such as replacement planting with native fruit bearing plants when weed fruits are an important seasonal resource to species (Gosper and Vivian-Smith 2009) or, leaving dead weeds in place when birds rely on weeds for perching (Lawrie 2002). Yet, if managers are to account for such weed-wildlife interactions, they first need to identify them, the wildlife species involved and the weed attributes that influence the interactions.

44 Chapter 3 – Vegetation structure, not weediness, influences wildlife

The aim of this study was to: 1) understand whether weeds are acting as habitat for wildlife; 2) determine if season influences the habitat value of weeds; 3) understand what characteristics of weedy vegetation most influence wildlife.

3.3 Methods

3.3.1 Study area The You Yangs Regional Park (YY) is a 2000 hectare woodland reserve near Melbourne, Australia (37º 56’ S, 144º 25’ E). It hosts one of the most dense infestations of the weedy shrub, African boneseed Chrysanthemoides monilifera subsp. monilifera (DC.) T. Norl. in Australia. Apparently planted in the YY for erosion control, it has spread to more than 1300 ha (65%) of the park (Brougham et al. 2006; Roberts 2008). As one of Australia’s ‘Weeds of National Significance’ and having severe adverse impacts on the vegetation throughout the park, it has been the subject of long-term and intensive management (Roberts 2008).

Snowy mint bush Prostanthera nivea A.Cunn. ex Benth. and drooping cassinia Cassinia arcuata R.Br. are two of the few native shrub species that remain dominant in small sections of the woodland, due to concerted management of boneseed. Sapling sized black wattle Acacia mearnsii De Wild. and golden wattle A. pycnantha Benth. are also common throughout the woodland, often growing in dense stands below a canopy of Eucalyptus species.

3.3.2 Study design Bird and small mammal assemblages were compared across 25 one hectare study sites. Sites were selected such that the shrub layer was either dominated by boneseed (‘boneseed sites’; n = 15), P. nivea (‘native shrub sites’; n= 5), or was largely without a shrub layer (‘open sites’; n = 5); these were referred to as ‘site types’. Sites were randomly selected within appropriate vegetation using a proportional stratified sampling

45 Chapter 3 – Vegetation structure, not weediness, influences wildlife approach (for example, the proportion of boneseed sites reflected the 65% of vegetation that was dominated by boneseed).

The average height of shrubs was greater in native shrub sites (mean ± SE, 115 ± 6 cm) than boneseed and open sites (73 ± 4 and 6 ± 2 cm, respectively). Ground cover species varied within and between sites, but the weedy grass annual veldtgrass Ehrharta longiflora Sm. was generally dominant in boneseed and native shrub sites (dominant in 40% of sites in each site type), while in open sites nodding saltbush Einadia nutans (R.Br.) A.J.Scott dominated (40% of sites). The dominant canopy species also varied between site types, with black wattle dominant in boneseed sites (73% of sites), golden wattle dominant in native shrub sites (60% of sites), and both golden wattle and yellow gum Eucalyptus leucoxylon F. Muell. dominating open sites (each species dominant in 40% of sites).

3.3.3 Small mammals Small mammals were captured live using Elliott traps (type A, 33 x 10 x 10 cm) and wire-mesh cage traps (44 x 20 x 17 cm). Traps were baited using a mixture of rolled oats, peanut butter, honey, linseed oil and vanilla essence; bedding was provided (hollofil), and traps covered with plastic (Tasker and Dickman 2002). Captured animals were identified, weighed, assessed for reproductive condition, and marked, before release.

A square trapping grid was set in the centre of each site with four lines of four Elliott traps, set ten metres apart and parallel to each other (total of 16). A cage trap was placed in each of the four corners of the grid (total of four). Each site was trapped for three consecutive nights, in the Australian summer, autumn, winter and spring in 2009/2010. Therefore, there was a total of 48 Elliott and 12 cage-trap nights per site, per season. Any trap failures were excluded from analysis to remove associated bias. Some sites with > 10% trap failures were excluded from some analyses.

46 Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.3.4 Birds A point-based survey was conducted in the centre of each site. All species seen within a 50 m radius of the survey point were recorded. Distance measurements from the central point were also made (using a laser rangefinder). Survey duration was set to 60 minutes and there was no influence of vegetation type on detectability (Appendix A).

Given that long survey times increase the risk of counting the same individual twice (Bibby et al. 1992), sampling was constrained so that additional individuals of a species were counted only when there was certainty that they were separate individuals. This restriction meant that counts of individuals were likely underestimated, leading to low counts, and hence small sample sizes. Low sample sizes did not allow for adjustments to account for differences in detection probability using distance measures (Buckland et al. 2001). Throughout this study counts of individuals were considered to represent an ‘activity index’ representing bird ‘activity’.

3.3.5 Vegetation At every Elliott trap location, a 1 x 1 m quadrat was established and cover of ‘life forms’ (shrubs, herbs, grass, lichen, bryophytes, bare ground, leaf litter and coarse woody debris) were estimated as percentages, as well as the proportion of living saplings (single stemmed trees of < 20 cm diameter at breast height [DBH], but > 1 m tall). Above each quadrat, the projective foliage cover of saplings and mature trees (single stemmed trees > 20 cm DBH and >1 m tall) was estimated collectively as a percentage, as was the health of this foliage (McNaught et al. 2006). Total percent cover of weeds (regardless of the strata layer) also was recorded in each quadrat. Vegetation data from all quadrats in a site was averaged for analysis.

The vertical density distribution of vegetation was measured by counting the number of touches (up to ten) of living or dead vegetation per 10 cm interval along a 180 cm vertical pole. This was measured in the four corners of each quadrat and averaged,

47 Chapter 3 – Vegetation structure, not weediness, influences wildlife within each 10 cm interval (Wilson and Paton 2004), for each trap location, giving a measure of vertical ‘vegetation density’.

In each site, seasonal flowering and fruiting vegetation was accounted for by identifying and counting the number of species that were flowering and/or fruiting (henceforth, the ‘richness’ of flowering/fruiting plants). The cover of the plants that were flowering/fruiting was also estimated, to give a measure of the total amount of vegetation offering flowers/fruits at a site.

3.3.6 Variable selection Small mammal activity was investigated using an average ‘capture rate’ which was derived from the average number of captures per site over the three trap nights.

Bird data was analysed using different metrics: species richness (the total number of species observed); species assemblages based on presence/absence data, and; the activity of species that were common. A species was defined as common if they were widespread (observed in ≥ 45% of sites) and highly active (with an activity index ≥ 40).

For the vegetation characteristics measured, life-form variables and vegetation density scores were separately reduced using Principal Components Analysis (PCA) in PASW Statistics 18.0 (2009). Where appropriate, data was transformed to reduce effects of outliers (Tables 3.1 and 3.2). Extracted components, along with the average cover of weeds per site were used to describe vegetation differences between sites and acted as predictors in analysis. Collectively they are referred to as ‘habitat variables’. Habitat variables were standardised for more direct comparisons.

3.3.7 Data analysis Birds and small mammals were assessed in relation to site type and habitat variables. Data analysis was conducted using R ver. 2.14.2 (R Development Core Team 2012) unless specified otherwise. Means ± standard errors are presented as appropriate.

48 Chapter 3 – Vegetation structure, not weediness, influences wildlife

Three methods were used to confirm that bird and small mammal responses were spatially independent. Bubble plots and variograms (using the geoR and gstat packages) were examined for bird richness, activity of common birds and small mammal capture rates, and provided little evidence of spatial autocorrelation. To ensure spatial independence of bird assemblages between sites, the RELATE routine was used in PRIMER v6 (2007), which compared two dissimilarity matrices; one containing the geographical locations of sites and the other containing the bird assemblage data of each site. A Mantel-type test was conducted comparing the two dissimilarity matrices, where if the rank correlation coefficient (Spearman’s ρ) was closer to one, among-sample relationships were similar between the matrices (Clarke and Gorley 2006) and, thus, spatially auto correlated. It was confirmed that bird species composition between sites was spatially independent (ρ = 0.108, p = 0.077).

The influence of site type on bird richness and small mammal capture rate was examined by comparing generalised linear models (glms) with and without (full and null model respectively) site type included as a parameter. The fit (log-likelihood: G2) of the full and null models were compared, with a significant G2 statistic showing that the full model was a better fit than the null model (Quinn and Keough 2002), and so site type explained some of the variation in the response variable. The effects of site type on bird assemblages were assessed using analysis of similarities (ANOSIM) and similarity percentage (SIMPER) in PRIMER v6. These analyses were conducted separately for each season as well as for data averaged across season.

Model selection and multimodel inference based on the information-theoretic approach (Burnham and Anderson 2002) was used to assess the influence of habitat variables on: bird richness, activity of common birds and small mammal capture rate. Models were based on all possible subsets of predictors (n = 31). For all glms assessment of residuals of each global model (model including all predictors) indicated that error structures were appropriate. Any models with influential observations were assessed with and without

49 Chapter 3 – Vegetation structure, not weediness, influences wildlife the influential observation/s. Any differences between these models were discussed, but only the model/s without the influential observations were presented.

To determine the best set of models, Akaike information criterion was used (corrected for small sample size, AICc; Burnham and Anderson 2002). For each model, an AICc value and the difference (Ui) between the AICc of each model and that of the most parsimonious model (model with lowest AICc), was calculated to facilitate model comparison. Models with Ui ≤ 2 had substantial support (Burnham and Anderson 2002). The variation explained (%) by the most parsimonious models was calculated. Akaike weights (wi) were also calculated for each model, and gave evidence as to whether a model was a good fit of the data (wi > 0.9 indicates good model fit). Summing the wi (to 0.95) provided a set within which there is 95% confidence that the best model exists (Burnham and Anderson 2002). When the number of models in this confidence set was high there was a high level of model uncertainty.

When model uncertainty was high and there was no single well-fitting model (wi < 0.9), model averaging was used to highlight the most important predictors for each response group (Burnham and Anderson 2002). This was based on averaged parameter estimates and summed wi across all models in which a predictor occurred. The higher the summed wi (maximum = 1), the more important the predictor variable. Calculations based on

AICc and model averaging analyses were performed using supplementary source code in R (M. Scroggie, unpubl.).

3.4 Results

3.4.1 Vegetation Two components were extracted from the life-form variables that accounted for 53% of the variation in the data (Table 3.1). The first component (canopy) described a gradient whereby higher component values represented increasing canopy complexity, in terms of more foliage cover that was also healthier, more living saplings and more leaf litter;

50 Chapter 3 – Vegetation structure, not weediness, influences wildlife this coincided with decreases in the cover of the shrubs (Table 3.1). The second component described a gradient whereby higher values represented increasing cover of ground layer vegetation (ground), where cover of bare ground decreased as the cover of grass and herbs increased (Table 3.1).

Two components were extracted from vegetation density scores that accounted for 58% of the variation in the data (Table 3.2). The first component described a gradient whereby increasing component scores represented increasing vegetation density between 51 – 140 cm (tall shrubs), while higher component values of the second component described a gradient of increasing vegetation density from 11 – 50 cm (low shrubs) (Table 3.2).

Boneseed sites had higher average cover of weeds compared with other site types. Open sites had higher canopy values and the lower values for low and tall shrubs than the other two site types, reflecting the lack of shrubby understorey for which open sites were selected. Native shrub sites tended to have higher values for tall shrubs (Figure 3.1).

The cover of fruiting vegetation was consistently highest in boneseed sites throughout all seasons but especially in winter, and when averaged over seasons (Table 3.3). Boneseed sites also tended to have higher richness of fruiting plants in winter; however, throughout the remaining seasons, richness of fruiting plants was similar between site types (Table 3.3). Differences in flowering plants between site types were less pronounced, except in spring when cover of flowering plants was higher in boneseed sites (Table 3.3).

51

Table 3.1 Factor loadings for selected principal components derived from average vegetation life-form variables measured at the You Yangs Regional Park. Components based on Principal Components Analysis (correlation matrix; Varimax rotation with a Kaiser Normalisation). Arcsine-transformed variables indicated ( ). Life-form variables (%) Canopy Ground Bare ground cover -0.238 -0.785 Leaf litter cover 0.631 -0.230 Coarse woody debris cover -0.300 0.055 Bryophyte cover a -0.052 -0.061 Lichen cover a -0.049 0.096 Herb cover a -0.398 0.744 Grass cover a -0.287 0.751 Shrub cover -0.707 0.114 Projected foliage cover 0.896 -0.043 Healthy foliage 0.888 -0.276 Living saplings 0.759 0.107 Variance (%) 38.240 14.620 Cumulative variance (%) 38.240 52.860

Table 3.2 Factor loadings for selected principal components representing the vertical density of vegetation at the You Yangs Regional Park. Components derived from the average number of vegetation pole touches in 10 cm intervals (to 180 cm). Components based on Principal Components Analysis (correlation matrix; Varimax rotation with S Kaiser Normalisation). Square root-transformed variables indicated ( ). Number of touches: Tall shrubs Low shrubs

0 – 10 cm 0.093 0.294 11 – 20 cm 0.055 0.921 21 – 30 cm 0.380 0.804 31 – 40 cm 0.494 0.785 41 – 50 cm 0.399 0.744 51 – 60 cm 0.827 0.216 61 – 70 cm 0.838 0.359 71 – 80 cm 0.319 0.859 81 – 90 cm 0.661 0.398 91 – 100 cm 0.718 0.511 101 – 110 cm S 0.890 0.130 111 – 120 cm S 0.574 0.532 121 – 130 cm S 0.605 0.153 131 – 140 cm S 0.561 0.243 141 – 150 cm S 0.022 0.187 151 – 160 cm S 0.378 0.106 161 – 170 cm S 0.072 -0.176 171 – 180 cm S -0.075 0.075 Variance (%) 45.026 12.806 Cumulative variance (%) 45.026 57.833

Figure 3.1 Habitat variables between site types at the You Yangs Regional Park. Mean standardised values (± 1 SE) of habitat variables between site types (z = boneseed sites, S = native shrub sites, ♦ = open sites). Habitat variables are average cover of weeds, and four components derived from Principal Components Analysis describing a: canopy complexity gradient, ground cover gradient, density gradient of tall shrubs and density gradient of low shrubs.

Table 3.3 Seasonal flowering and fruiting vegetation between site types at the You

Yangs Regional Park. Resource measured Site type Season Average Summer Autumn Winter Spring Average cover of Boneseed 25.3 36.2 11.3 22.4 23.8 fruiting plants (%) Native shrub 0.7 0.0 0.8 1.6 0.8 Open 0.1 0.1 0.5 0.0 0.2 Average richness of Boneseed 3.0 1.6 1.2 4.3 2.5 fruiting plants Native shrub 2.0 0.2 0.6 5.8 2.2 Open 0.8 0.2 0.4 2.2 0.9 Average cover of Boneseed 2.3 1.2 1.5 43.1 12.0 flowering plants (%) Native shrub 9.6 0.0 2.4 28.4 10.1 Open 0.2 0.0 0.0 2.8 0.8 Average richness of Boneseed 3.0 1.5 1.1 4.3 2.5 flowering plants Native shrub 2.0 0.4 0.8 5.8 2.3 Open 0.8 0.2 0.4 2.2 0.9

Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.4.2 Birds Forty-eight species were recorded. No species was present in every site, but nine species were widespread: recorded at > 45% of sites (Table 3.4). Of these widespread birds, only five were considered common (also having an activity index of ≥ 40; Table 3.4). While silvereyes Zosterops lateralis had an activity index of 46, the species was not investigated further due to a significant outlier, which, when excluded, reduced the activity index to < 40. Therefore, the common birds investigated further were the brown thornbill Acanthiza pusilla, superb fairy-wren Malurus cyaneus, eastern yellow robin Eopsaltria australis and New Holland honeyeater Phylidonyris novaehollandiae.

3.4.2.1 Influence of site type and season Site type did not substantially influence bird richness in summer, winter, spring or when data was averaged across season (Table 3.5). Bird richness tended to be lower in boneseed sites in autumn compared to the two other site types (boneseed sites, 2.3 ± 0.5 species; native shrub sites, 4.0 ± 0.3; and open sites, 4.0 ± 0.6), but this was not statistically significant (Table 3.5).

Bird assemblages were similar between boneseed and native shrub sites (Table 3.5). The assemblage of birds in open sites differed substantially to that of boneseed sites when data were averaged across seasons, in summer, and to a lesser extent, spring. Bird species that contributed to this dissimilarity changed with the seasons, with 11% being the greatest contribution made by any one species to the dissimilarity between the site types.

Differences in bird assemblages between open sites and native shrub sites were clearer in winter and spring (Table 3.5). The superb fairy-wren was more often present in native shrub sites compared to open sites (all seasons), contributing up to 14% to the dissimilarity between the two site types.

56

Table 3.4 Bird species recorded at the You Yangs Regional Park. The percentage of sites a bird species was observed for each site type and the total percentage of sites at which each species was observed are shown. Data presented is across all seasons. The total activity index is in parentheses. The four species indicated (z) were selected as the common species used in further analysis. Introduced species are indicated (*). Common and scientific names follow Christidis and Boles (2008). Common name Scientific name Site type Boneseed Native shrub Open All sites (n = 15) (n = 5) (n = 5) (n = 25) Brown thornbill z Acanthiza pusilla 100 100 60 92 (77) Superb fairy-wren z Malurus cyaneus 100 80 20 80 (66) Yellow thornbill Acanthiza nana 67 100 60 72 (37) Grey fantail Rhipidura albiscapa 73 60 60 68 (29) Eastern yellow robin z Eopsaltria australis 53 60 80 60 (50) Grey shrike-thrush Colluricincla harmonica 47 60 80 56 (24) Silvereye Zosterops lateralis 47 80 20 48 (46) New Holland honeyeater z Phylidonyris novaehollandiae 47 60 40 48 (40) Scarlet robin Petroica boodang 60 40 20 48 (24) Rufous whistler Pachycephala rufiventris 40 60 40 44 (20) Yellow-faced honeyeater Lichenostomus chrysops 47 40 0 36 (17) White-plumed honeyeater Lichenostomus penicillatus 27 0 60 28 (26) Red wattlebird Anthochaera carunculata 27 20 60 32 (31)

Table 3.4 Continued

Common name Scientific name Site type Boneseed Native shrub Open All sites Australian magpie Cracticus tibicen 20 60 20 28 (14) Willie wagtail Rhipidura leucophrys 13 40 40 24 (10) Little raven Corvus mellori 27 40 0 24 (9) White-winged chough Corcorax melanorhamphos 20 0 40 20 (32) Red-browed finch Neochmia temporalis 33 0 0 20 (24) White-throated treecreeper Cormobates leucophaea 13 20 40 20 (8) Crested shrike-tit Falcunculus frontatus 13 0 60 20 (7) Common blackbird* Turdus merula 13 40 0 16 (6) Eastern rosella Platycercus eximius 7 40 20 16 (5) Buff-rumped thornbill Acanthiza reguloides 20 0 0 12 (10) Yellow-rumped thornbill Acanthiza chrysorrhoa 7 20 20 12 (7) Golden whistler Pachycephala pectoralis 13 0 20 12 (5) Black-chinned honeyeater Melithreptus gularis 7 0 20 8 (7) European goldfinch* Carduelis carduelis 13 0 0 8 (4) Galah Eolophus roseicapillus 7 20 0 8 (4)

Horsefield' s bronze-cuckoo Chalcites basalis 7 20 0 8 (3) Sulphur-crested cockatoo Cacatua galerita 7 0 0 4 (4)

Table 3.4 Continued

Common name Scientific name Site type Boneseed Native shrub Open All sites Varied sittella Daphoenositta chrysoptera 0 0 20 4 (4) Australian wood duck Chenonetta jubata 7 0 0 4 (2) Black-faced cuckoo-shrike Coracina novaehollandiae 0 20 0 4 (2) Weebill Smicrornis brevirostris 0 0 20 4 (2) Brown-headed honeyeater Melithreptus brevirostris 7 0 0 4 (1) Common bronzewing Phaps chalcoptera 7 0 0 4 (1) Eastern spinebill Acanthorhynchus tenuirostris 7 0 0 4 (1) Fan-tailed cuckoo Cacomantis flabelliformis 0 0 20 4 (1) Flame robin Petroica phoenicea 7 0 0 4 (1) Laughing kookaburra Dacelo novaeguineae 0 20 0 4 (1) Restless flycatcher Myiagra inquieta 0 0 20 4 (1) Spotted pardalote Pardalotus punctatus 7 0 0 4 (1) Striated pardalote Pardalotus striatus 7 0 0 4 (1) Striated thornbill Acanthiza lineata 0 20 0 4 (1) White-browed scrub wren Sericornis frontalis 7 0 0 4 (1) Crimson rosella Platycercus elegans 0 40 0 8 (2) Little wattle bird Anthochaera chrysoptera 7 20 0 8 (2) White-winged triller Lalage sueurii 0 0 20 4 (1)

Table 3.5 Comparison of house mouse capture rates, bird species richness and bird assemblages between site types at the You Yangs Regional Park. Results are from glms that compare the fit of the full model, where site type is the predictor variable (df = 2), to that of the null model (G2). Error structures were based on a Poisson distribution (corrected for overdispersion), unless indicated (G Gaussian distribution). Results for ANOSIM are based on presence/absence data and presented as an R statistic for pair-wise tests between site types. Statistically significant results at p < 0.05 are indicated (*), and at p < 0.001 are indicated (*) in bold. Comparison Test statistic Summer Autumn Winter Spring All seasons

2 House mouse capture rates G 0.168 0.026*G 0.046*G 0.699* 0.556* Bird richness G2 0.836 5.423 6.031 4.981 2.650 Bird species assemblages Pair-wise tests Boneseed vs native shrub R 0.002 0.110 -0.110 -0.047 -0.041 Boneseed vs open R 0.573* 0.010 0.107 0.252* 0.472* Open vs native shrub R 0.290 0.226 0.380* 0.354* 0.260

Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.4.2.2 Influence of habitat variables Akaike weights were low for all models that addressed the influence of habitat variables on birds (Table 3.6). The most parsimonious models explained ≤ 3 % of the variation in bird richness; this contrasts to the four common bird species investigated, for which the most parsimonious models explained up to 42% of the variation (Table 3.6); however, the number of models in each 95% confidence set was high: bird species richness (20 models in set); brown thornbill (13); superb fairy-wren (17); eastern yellow robin (20); New Holland honeyeater (25). Such high model uncertainty meant model averaging was conducted.

Model averaging indicated that total weed cover did not influence any of the bird response variables, as indicated by the low summed wi values, relative to the other predictor variables (Table 3.7). Canopy was the only predictor that demonstrated substantial influence on any of the common birds (Table 3.7). Brown thornbills and superb fairy-wrens demonstrated negative responses to increases in canopy values (for superb fairy-wrens, this was only upon the removal of a substantial outlier). Eastern yellow robins showed the opposite response to canopy and New Holland honeyeaters were positively influenced by increases in low shrubs more than other habitat variables (Table 3.7).

3.4.3 Small mammals The introduced rodents, house mouse Mus domesticus and black rat Rattus rattus, were the only small mammal species captured (2065 trap nights). Black rats represented five percent of total captures and were not analysed further.

House mouse capture rates, on average, did not differ between boneseed and native shrub sites. Capture rates were lower in autumn, winter, spring and when averaged across season in open sites compared to boneseed and native shrub sites (Table 3.5).

61 Chapter 3 – Vegetation structure, not weediness, influences wildlife

Akaike weights were low for all models that addressed the influence of habitat variables on house mouse capture rates (Table 3.6). Despite some models explaining up to 93% of the variation in the data there was high model uncertainty; wi ≤ 0.3 and 19 models were in the 95% confidence set (Table 3.6). Consequent model averaging did not indicate that weeds were an influential variable on capture rates relative to other predictors; capture rates were positively influenced by increases in low shrubs (Table 3.7).

62

Table 3.6 Model selection results for birds and house mouse at the You Yangs Regional Park. Included are the parameters incorporated in the model, total number of parameters

(K), explained variation, AICc values, AICc differences (Ui), and Akaike weights (wi).

Models are listed in descending order of Ui. Only models with Ui ≤ 2 are listed or if the number of Ui ≤ 2 exceeded five, only the best five models are shown. Models were based on Gaussian (bird species richness, eastern yellow robin, superb fairy-wren and house mouse capture rates), Poisson and negative binomial (brown thornbill and New

Holland honeyeater respectively) distributions.

Response Predictor variables Variation K AICc Ui wi variables explained Bird species low shrubs 3% 2 131.0 0.000 0.124 richness tall shrubs 3% 2 131.0 0.007 0.123 ground 1% 2 131.4 0.421 0.100 weeds 1% 2 131.4 0.457 0.098 canopy 1% 2 131.6 0.609 0.091 Brown thornbill canopy + ground 42% 3 94.3 0.000 0.268 canopy 32% 2 95.2 0.877 0.173 canopy + weeds 37% 3 96.1 1.799 0.109 Eastern yellow canopy 21% 2 111.7 0.000 0.258 robin canopy + tall shrubs 25% 3 113.4 1.649 0.113 low shrubs 15% 2 113.6 1.915 0.099 New Holland low shrubs 18% 2 85.3 0.000 0.243 honeyeater ground + low shrubs 24% 3 86.3 1.030 0.145 Superb fairy- canopy 9% 2 59.3 0.000 0.279 wren canopy + ground 9% 3 60.3 1.010 0.168 weeds + canopy 9% 2 61.3 1.970 0.104 House mouse low shrubs 92% 2 -81.1 0.000 0.220 capture rates weeds + low shrubs 93% 3 -79.9 1.190 0.122 canopy + low shrubs 93% 3 -79.4 1.670 0.096

Table 3.7 Relative influence of habitat variables on wildlife at the You Yangs Regional Park as derived from model averaging. Model averaged data is summarised by parameter coefficients, their standard errors and sums of Akaike weights (wi). The most influential variables for wildlife have the highest sum of (wi). Highlighted in bold are those variables for which the 95% confidence interval does not include zero, indicating they are important variables in determining wildlife.

Response variable Predictor Coefficient Standard Sums of error wi Bird species richness weeds 0.164 0.464 0.320 canopy -0.030 0.370 0.263 ground -0.176 0.460 0.328 tall shrubs 0.184 0.425 0.343 low shrubs 0.181 0.435 0.337 Brown thornbill weeds -0.028 0.087 0.273 canopy -0.379 0.130 0.959 ground -0.114 0.117 0.531 tall shrubs 0.008 0.060 0.211 low shrubs 0.006 0.063 0.209 Eastern yellow robin weeds 0.026 0.244 0.207 canopy 0.792 0.473 0.756 ground -0.043 0.229 0.210 tall shrubs 0.098 0.290 0.267 low shrubs -0.234 0.408 0.364 New Holland honeyeater weeds 0.050 0.203 0.251 canopy 0.026 0.209 0.226 ground 0.162 0.261 0.390 tall shrubs 0.019 0.155 0.206 low shrubs 0.552 0.310 0.740

Table 3.7 Continued Response variable Predictor Coefficient Standard Sums of error wi Superb fairy-wren weeds 0.028 0.089 0.246 canopy -0.494 0.160 0.923 ground 0.071 0.114 0.365 tall shrubs 0.017 0.080 0.211 low shrubs 0.010 0.095 0.208 House mouse capture rates weeds 0.006 0.010 0.406 canopy -0.006 0.009 0.369 ground -0.002 0.007 0.254 tall shrubs 0.000 0.004 0.189 low shrubs 0.018 0.010 0.762

Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.5 Discussion

3.5.1 Wildlife habitat preferences Similarities in bird richness, bird assemblages and house mouse capture rates between boneseed and native shrub sites indicated that vegetation dominated by boneseed offered similar habitat for birds and mice as vegetation dominated by native shrubs. There were, however, differences in bird assemblages and house mouse activity between open sites and these two site types that were dominated by shrubs. As open sites were selected for their lack of shrub and weed cover, these differences indicate that bird and small mammals may respond to structural differences in vegetation at the YY, rather than weediness.

Modelling further suggested that structure influenced some species while weed cover did not. The brown thornbill and the superb fairy-wren both displayed a negative response to canopy. This reflected their dependence on lower storey vegetation, like shrubs (Recher et al. 1985; Barrett et al. 2008), which decreased in cover as canopy complexity increased. Indeed, the commonality of these bird species is probably explained by their preference for shrubs, which are widespread throughout the YY, primarily because boneseed occurs throughout 65% of the park (Brougham et al. 2006).

The pest rodent, house mouse, demonstrated an association with shrubs also. While this species is known to prefer dense, low growing vegetation structure (Mutze 1991), its presence is often linked to disturbance as well (Holland and Bennett 2007). Therefore, similar capture rates between site types could have been expected, which would reflect the long history of widespread disturbance in the YY (Prescott 1995). Alternatively, the notion that weeds harbour pest animals (Natural Resource Management Ministerial Council 2006) could lead to the expectation that this pest would show a considerable positive response to weed cover. Neither was the case in the YY, where mice were positively influenced by density of vegetation low to the ground, regardless of whether it was weed or native.

66 Chapter 3 – Vegetation structure, not weediness, influences wildlife

While the influence of vegetation structure on wildlife species is shown to be significant in many systems (Bennet 1993; Monamy and Fox 2005; Garden et al. 2007), seasonal variation in fruit and floral availability between weed and non-weed dominated vegetation has been shown to influenced bird richness in some areas (Gosper 2004). Despite seasonal differences in amounts of fruiting or flowering plants between the site types in the YY, there was little evidence that this influenced the wildlife measured; lower amounts of fruiting plants in boneseed sites over winter and higher amounts of flowering plants over spring did not correspond to changes in bird richness, assemblages or house mouse activity.

Despite having no obvious direct influence on the wildlife species measured, these results do not support the notion that weeds are benign for wildlife in the YY. Boneseed is likely having secondary or indirect impacts on some wildlife species. By suppressing the growth of native vegetation, boneseed has prevented the regeneration of overstorey and canopy species in the YY (Roberts 2008). The canopy gradient reflected this: canopy cover and health reduced as shrub cover increased. While this gradient was similar between sites that had native or weed shrub cover, the prevalence of boneseed throughout the YY would be expected to make a substantial contribution to the decrease in canopy complexity.

Canopy dependent species would be most influenced by the suppression of canopy regeneration (Roberts 2008), but other non-canopy dwelling species may still be negatively affected. In the YY the eastern yellow robin was common. Yet, as a forest interior species (Zanette et al. 2000) it may be subject to the impact boneseed is having on the canopy. While no substantial influence of weed cover on the activity of the robin, was found, it exhibited a positive association with canopy; as canopy complexity increased, so did the activity of the robin. A lack of recruitment of overstorey species means that senescing canopy vegetation cannot be replaced, and the cover and health of the canopy can only decrease. With further reductions in canopy complexity, modelling indicated the activity of the eastern yellow robin could follow a similar pattern of 67 Chapter 3 – Vegetation structure, not weediness, influences wildlife decline. As a species sensitive to disturbances like fragmentation (Zanette 2000) and urbanisation (Trollope et al. 2009), the potential for further declines in response to another widespread disturbance, like weed invasion, is of concern.

Complex interactions like these indirect effects are likely common throughout systems invaded by weeds. Elsewhere, weed invasions have been highlighted as ecological traps, causing changes to habitats so that organisms erroneously choose this changed habitat instead of nearby better quality habitat (Schlaepfer et al. 2002). For example, blackcaps Sylvia atricapilla consistently chose to nest in exotic vegetation that was structurally appropriate, but depredation in these systems was much higher than in nearby native vegetation and led to lower nest success (Remes 2003). Thus, equivalent use of weeds compared with similar native vegetation (in this study) could mask responses to weeds that may not be beneficial.

3.5.2 Management implications In this highly degraded, remnant woodland system, weeds generally provided equivalent habitat for wildlife to that of native vegetation with a similar structure. So, regardless of secondary impacts weeds may be having, conducting large-scale removal of weeds could be comparable to similar removal of native vegetation in the short-term.

When weeds provide wildlife with habitat, phased weed removal in conjunction with revegetation has been recommended (Gosper and Vivian-Smith 2006). This approach allows for wildlife to persist as weed resources are gradually replaced with natives. Yet, decisions regarding techniques for gradual removal or native plant species to use in revegetation, may be inappropriate without an understanding of the interactions between weeds and wildlife. For example, with this study demonstrating that structural aspects of the vegetation were most influential for some wildlife species, revegetation with native plant species that are less structurally complex than boneseed may not effectively sustain these species after boneseed removal.

68 Chapter 3 – Vegetation structure, not weediness, influences wildlife

With the outcomes of weed management on non-target organisms, like wildlife, rarely examined (Valentine and Schwarzkopf 2008), consideration of weed-wildlife interactions prior to weed management would inform weed management plans of how best to account for wildlife. When such projects are implemented the incorporation of wildlife measures into the monitoring process would reveal how wildlife respond to such weed management. Thus, allowing for the consideration of wildlife throughout the adaptive weed management process.

69 Chapter 3 – Vegetation structure, not weediness, influences wildlife

3.6 Appendix A

Two approaches were used to control for the possibility that site type could influence detectability and, thus, confound comparisons between site types. Firstly, because species accumulation curves can differ between habitats, a survey duration which represented an asymptote of species richness in all site types was set. Survey duration was set at 60 minutes because initial surveys conducted in the three site types (by determining cumulative bird richness in 5-minute intervals over 60 minutes) revealed an asymptote in all site types by this time (Figure A3.1). Repeated measures analysis of variance (ANOVA) confirmed there was neither an interaction between site type and time (F = 0.966, df = 22, 242, p < 0.081), nor a main effect of site type on bird richness (F = 1.265, df = 2, 22, p = 0.302); there was an expected positive effect of survey duration on bird richness (F = 36.44, df = 11, 242, p <0.001). Secondly, it was ensured that detectability did not change with distance in different site types. There was too few data on any species to model detectability with distance in the surveys; however, there was no evidence of an interaction between the distance birds were detected (in 10 m intervals up to 50 m) in the different site types (detections averaged within sites; 2-way ANOVA, F = 0.499, df = 8, 110, p < 0.855), nor a main effect of site type (F = 0.408, df = 2, 110, p < 0.666) suggesting detectability was equivalent across site types.

70

Figure A3.1 Cumulative bird species richness recorded over timed surveys (natural log + 1 transformed, ± 1 SE) in different site types in the You Yangs Regional Park. Site types were boneseed (top), native shrub (middle) and open (bottom). Survey time was set at 60 minutes reflecting the asymptote common to the three site types.

Chapter 4

The influence of an emergent weed on wildlife in a southern Australian wetland

Singing honeyeater Lichenostomus virescens commonly uses African boxthorn Lycium ferocissimum throughout Cheetham Wetlands, Victoria, Australia.

72 Chapter 4 – The influence of an emergent weed on wildlife

4.1 Abstract

Negative effects of invasive species on biodiversity are well documented, but positive interactions can occur. Considering global habitat loss, one such interaction that is especially important is the potential for weeds (plants that are invasive) to benefit wildlife. Weeds may be important to wildlife in the absence of native habitat, but also when they add structural complexity to native vegetation, for example, when they stand above surrounding vegetation as an ‘emergent’. To understand the influence of an emergent weed on wildlife, bird and small mammal assemblages were compared in vegetation with and without African boxthorn Lycium ferocissimum, an emergent weed in a coastal wetland in southeastern Australia. Structural complexity contributed by boxthorn was expected to lead to a change in wildlife presence and activity; however, wildlife responded inconsistently to the presence of boxthorn. For birds, species assemblage, but not richness, changed with the presence of boxthorn. The two most common native bird species appeared to be the drivers of these differences, and displayed different associations with the weed; one species preferred boxthorn while the other avoided it. These associations probably were due to the emergent nature of boxthorn, which may influence bird behaviour. In comparison, the only small mammal that was captured frequently was influenced most by vegetation structure at ground level, not weed cover. It is clear that vegetation structure is influential in this system, but also that emergent weeds can affect wildlife assemblages. Consequently, managing these types of weeds may cause further changes. This needs to be considered during weed management, along with whether biodiversity objectives would be met by causing such changes and, if not, how best to respond.

Key words: Invasive species; weed; biodiversity; habitat; wildlife; weed management; vegetation structure; African boxthorn Lycium ferocissimum.

73 Chapter 4 – The influence of an emergent weed on wildlife

4.2 Introduction

Invasive species threaten biodiversity and ecosystems worldwide (Hobbs and Humphries 1995). Plants that are unwanted in natural or human environments can be defined as weeds, and are often characterised by their invasive nature and the detrimental impacts they cause (Richardson et al. 2000; Natural Resource Management Ministerial Council 2006). In natural environments, these negative effects have been well-documented and are emphasised in the literature, which gives the impression that positive impacts are uncommon (Schlaepfer et al. 2011). However, weeds have been reported to provide ecosystem services, such as soil stabilisation and nutrient retention, act as catalysts for restoration of native vegetation and provide resources for wildlife (Schlaepfer et al. 2011). When positive interactions between wildlife and weeds have been demonstrated, adjustments to weed management have been recommended to avoid unintended adverse impacts on these species (Sogge et al. 2008); however, making adjustments to management is not a standard response to these situations, and those that are made are not necessarily appropriate. A general lack of understanding about the impacts weeds may have on wildlife and their habitat (Sax et al. 2005) is likely to contribute to this, as is the infrequent reporting of studies that find no ill effects of invasive species (Schlaepfer et al. 2011).

Rodriguez (2006) has suggested four broad circumstances in which positive interactions between invasive and native species are most likely. This includes when invasive species provide an escape from a competitor or predator, provide a limited resource, or increase habitat complexity. Weeds that produce fruit and occur in fruit poor vegetation have the potential to offer limited resources (for example to birds; Date et al. 1996), while prickly and thorny weeds may provide predator protection (for example for small birds and mammals; Nias 1986; Sanderson and Kraehenbuehl 2006). In many situations, weeds also have the potential to increase habitat complexity, particularly when they stand much taller than the vegetation in which they are established. Yet, it is believed that no

74 Chapter 4 – The influence of an emergent weed on wildlife quantitative investigations have been conducted on the influences of such ‘emergent’ weeds on wildlife.

Positive interactions between invasive and native species also are suggested to be more likely when invasive species replace the functional role of native species (Rodriguez 2006). For weeds, this can mean the provision of habitat for wildlife when native vegetation no longer exists. For example, in some riparian areas in the United States where native vegetation does not regenerate, invasions of saltcedar Tamarix spp. support sensitive birds such as yellow-billed cuckoos Coccyzus americanus and southwestern willow flycatchers Empidonax traillii extimus (Sogge et al. 2008). Given the global continuation of native vegetation removal and, consequently habitat loss (Fahrig 2001), this could arguably be the most important situation where weeds benefit wildlife. As such, the potential wildlife values of some weed species are recognised occasionally, but often this is only when they are established in areas where native vegetation is relatively absent (Magnussen undated).

In Australia, the weed African boxthorn Lycium ferocissimum Miers. is suggested to have the potential to fulfil the functional role of habitat for wildlife when there is no native alternative (Magnussen undated). Yet, it might also be expected that some native wildlife would benefit from weeds, like boxthorn, when they grow in areas where native vegetation persists. As a dense fruiting shrub, with branches terminating in sharp, sturdy thorns and growing up to three metres high, boxthorn potentially provides increased structural complexity to habitats (particularly where it grows as an emergent) and offers wildlife protection from predators and potential floral/fruit resources. If these benefits exist, managers may need to consider the implications of the removal of weeds such as boxthorn from areas where native vegetation remains, in addition to circumstances where the original vegetation has been removed.

The influence of emergent boxthorn on wildlife was examined by: 1) identifying whether boxthorn was acting as habitat for wildlife; 2) exploring which aspects of

75 Chapter 4 – The influence of an emergent weed on wildlife boxthorn were important for wildlife; and 3) determining if season influenced boxthorn use by wildlife.

4.3 Methods

4.3.1 Study area Cheetham Wetlands (CW) is a wetland environment covering 420 ha adjacent to Port Phillip Bay, Victoria, Australia (37º 54’ S, 144º 48’ E). The original wetland ecosystem was modified to operate as a commercial saltworks until 1992, but now the shallow evaporation ponds and adjacent areas have became important habitat for birds (Brett Lane and Associates 2009). The dominant vegetation community, extending to varying degrees around the edges of most ponds, is coastal saltmarsh (Cook 1996) characterised by shrubby glasswort Tecticornia arbuscula (R.Br.) K. A. Sheph. & Paul G. Wilson, beaded glasswort Sarcocornia quinqueflora (Bunge ex Ung. – Sternb.) A. J. Scott subsp. quinqueflora and other succulent herbs. Also present in the higher permanently dry areas, are coastal dune grassland and sedgeland communities. While these communities are all dominated by native plant species, boxthorn, along with many other weed species, also occurs.

Boxthorn is particularly prominent in this ecosystem because it stands much higher than the low growing saltmarsh vegetation; it is a clear emergent, adding another level of structural complexity to the habitat. Boxthorn is listed as a ‘Weed of National Significance in Australia’ (Magnussen undated) and, as such, is listed as a high priority for control in CW (Smedley and Nance 2010). It is found in isolation or dense stands in many of the terrestrial areas of the wetland (Smedley and Nance 2010). It out competes native vegetation, acts as shelter for pest animals and can impede access for people and vehicles (Muyt 2001; CRC for Australian Weed Management 2007).

76 Chapter 4 – The influence of an emergent weed on wildlife

4.3.2 Study design Eighteen sites were selected throughout the study area and compared bird and small mammal assemblages in boxthorn and non-boxthorn vegetation, reflecting other comparative studies on weed-wildlife impacts (Williams and West 2000). Sites were separate embankments that were bordered by wet or dry channels and/or evaporation ponds. Vegetation on the embankments either included boxthorn (‘boxthorn sites’; n = 13), or did not. Where boxthorn was absent, vegetation was dominated by shrubby glasswort, the typical saltmarsh shrub (‘saltmarsh sites’; n = 5). Embankments varied in size (0.30 – 0.75 ha) but sampling was standardised by using two randomly selected sampling areas (50 x 10 m) on each embankment. Sampling areas were located at least 50 m away from each other. In boxthorn sites, these two areas centred around one or more randomly selected boxthorn plants; in saltmarsh sites, these areas were randomly located. Surveys of bird and small mammal assemblages occurred in the two sampling areas in each site and, together, these data describe the site. Surveys were conducted in the Australian autumn, winter, spring (2009) and summer (2009/2010).

4.3.2.1 Small mammal activity Sixteen Elliott traps and four cage traps were deployed in two transects in each of the 18 sites. Within each of the sampling areas of each site, a transect of eight Elliot traps, spaced at five metre intervals, was deployed, with a cage trap at the first and fourth trap locations in each transect. In boxthorn sites, transects were set so the central boxthorn plant/s was/were in the centre of the transect. Thus, for each transect, two traps were located directly beneath the plant (within five metres of each other), and two traps were five, ten and 15 m from the central boxthorn plant, in opposite directions. In most cases, transects were straight, although there were bends in some to ensure each trap was the correct distance from any other boxthorn plants in the site.

Each site was trapped for three consecutive nights each season. Traps were baited using a mixture of rolled oats, peanut butter, honey, linseed oil and vanilla essence. Traps were covered with plastic and bedding (hollofil) was placed inside to protect captured animals 77 Chapter 4 – The influence of an emergent weed on wildlife from the elements (Tasker and Dickman 2002). From dawn, captured animals were identified and marked before they were released.

4.3.2.2 Bird assemblage and activity Point count surveys were conducted between dawn and four hours after sunrise on days without strong wind or persistent rain. The order in which sites were surveyed was stratified to ensure they were sampled at different times throughout the morning.

All bird species seen within each sampling area at all sites were recorded and individuals counted by EC (species flying above the sites were not recorded). Observations were made in the location that afforded a view of the whole sampling area, in each site. Distance measures were not taken because all detections were made within 32 m from the observation point (due to the small and linear nature of sampling areas). To ensure site type did not influence survey performance a standard survey time was set based on species accumulation (Appendix A). Thus, each sampling area was surveyed for 30 minutes (totalling 60 minutes a site). Given that longer survey times increase the risk of counting the same individual twice (Bibby et al. 1992), sampling was constrained so that additional individuals of a species were counted only when there was certainty that they were separate individuals. Therefore, counts of individuals are considered to represent a measure of species usage/activity rather than relative abundance, and is referred to as an ‘activity index’.

4.3.2.3 Vegetation structure and resources At every Elliott trap location, a 1 x 1 m quadrat was established to assess vegetation structure in terms of cover of vegetation ‘life-forms’ and vertical ‘vegetation density’. Percent vegetation cover of life-forms was estimated, which included shrubs, herbs, grass, tussock grass, sedges/rushes, lichen, and bryophytes, along with percent cover of bare ground, leaf litter and coarse woody debris (McNaught et al. 2006). As CW is treeless, coarse woody debris referred to dead branches, fallen or cut from shrubs. Total weed cover also was estimated, regardless of the stratum in which it existed. To evaluate

78 Chapter 4 – The influence of an emergent weed on wildlife the distribution of vertical vegetation density, the average number of touches of living or dead vegetation up a 180 cm vertical pole were counted (up to ten touches) in 10 cm increments.

In each site, seasonal flowering and fruiting vegetation were accounted for by identifying and counting the number of species that were flowering and/or fruiting. This is described as ‘richness’ of flowering and fruiting plants. The vegetative cover of the flowering/fruiting plants was also estimated, representing the total amount of vegetation offering flowers/fruits at a site.

4.3.3 Variable selection and data analysis 4.3.3.1 Predictor variables Vegetation data for life-forms and vegetation density were reduced separately using Principal Components Analysis (PCA) in PASW Statistics 18.0 (2009). Where appropriate, data was transformed to reduce effects of outliers (Tables 4.1 and 4.2). Extracted components, along with the average cover of weeds per site, were used to describe vegetation differences between sites and acted as predictors in analysis. Collectively they are referred to as ‘habitat variables’. Habitat variables were standardised for more direct comparisons.

4.3.3.2 Response variables Bird data from both sampling areas in a site was averaged (or pooled in the case of presence/absence data) so that each site is considered as a single datum. Different metrics were used to describe and quantify bird assemblages and activity: compositional data that consisted of presence/absence of each species, total bird richness (the number of different species), and the activity of individual species that were common. Species which were defined as common were ZLGHVSUHDG REVHUYHGLQ•RIVLWHV DQGKDGD KLJKDFWLYLW\LQGH[ • .

79 Chapter 4 – The influence of an emergent weed on wildlife

To describe small mammal activity, the number of captures in each trap was averaged over the three trap nights giving an average ‘capture rate’. Any trap failures were excluded from analysis. Rate of trap failure was similar (< 5%) between site types (F = 2.180, df = 1, 16, p = 0.159).

4.3.3.3 Statistical analyses Analysis was conducted in R ver. 2.14.2 (R Development Core Team 2012) unless otherwise specified. On occasion, means ± standard errors are presented to describe the variables.

Three different methods were used to confirm spatial independence between sites. Bubble plots and variograms (using the geoR and gstat packages; Zuur et al. 2009) were used and provided little evidence of spatial autocorrelation of bird richness, activity of common birds and small mammal capture rates between sites. To ensure spatial independence of bird composition between sites, the RELATE routine in PRIMER v6 was used (Clarke and Gorley 2006). It was confirmed that bird species composition EHWZHHQVLWHVZDVVSDWLDOO\LQGHSHQGHQW ȡ S  

Generalised linear models (glm) assessed the response of bird richness and small mammal capture rates to site type in the different seasons and, when averaged over season. To assess pooled differences between seasons glms were also used. The fit of full model (including the parameter site type) was compared to that of the null model.

Multivariate analyses of bird composition in the different seasons were conducted in PRIMER v6, using analysis of similarities (ANOSIM) and similarity percentage (SIMPER) to test for compositional differences between the site types.

Generalised linear mixed models (glmm) were used to assess whether small mammal capture rates at boxthorn sites differed with distance from the central boxthorn plant at each sampling area (using the MASS package). Distance from boxthorn was entered as a categorical predictor and represented placement of Elliott traps, either at zero (beneath

80 Chapter 4 – The influence of an emergent weed on wildlife the boxthorn plant) five, ten or 15 m from the centre of the boxthorn plant, and site was entered as a random factor (accounting for non-independence). The fit of the full (including the distance parameter) and reduced models were compared. The nature of the response was examined using the full model, with the zero distance category as the baseline to contrast the coefficients (± standard error) of the other distance categories.

Model selection and multimodel inference based on an information-theoretic approach (Burnham and Anderson 2002) was used to determine the influence of habitat variables on bird richness, activity of common birds and small mammal capture rates. Models were based on all possible subsets of predictors for each response variable (n = 31). A Gaussian distribution was assumed for all glms and assessment of residuals from each global model (the model including all predictors) indicated that error structures were appropriate. A set of calculations were used based on Akaike information criterion

(corrected for small sample size; AICc) to compare models (Burnham and Anderson

2002). A difference (Ui)of ” 2 between the AICc of each model and that of the best fitted model was considered to indicate that a model had considerable support and

Akaike weights (wi) of > 0.9 as evidence of a single most parsimonious model. A 95% confidence set was produced by summing the wi (to 0.95). When model uncertainty was high, model-averaging was used to highlight the most important predictors for each response variable. Summed wi (maximum = 1) was also used on averaged estimates to indicate predictor importance. Analysis based on AICc and model-averaging was performed using supplementary source code in R (M. Scroggie, unpubl. data).

4.4 Results

4.4.1 Vegetation 'XHWRLWVHPHUJHQWQDWXUHER[WKRUQZDVDSURPLQHQWIHDWXUHLQER[WKRUQVLWHV •FP taller than native shrubs); however, it was not necessarily the dominant shrub (23% of sites); this was shared with shrubby glasswort (53%) and seaberry saltbush Rhagodia candolleana Moq. (19%). The ground layer in both sites consisted of a mixture of

81 Chapter 4 – The influence of an emergent weed on wildlife grasses, such as Australian saltgrass Distichlis distichophylla (Labill.) Fassett and blue tussock-grass Poa poiformis (Labill.) Druce. and succulents, such as coast bonefruit Threlkeldia diffusa R. Br. and rounded noon flower Disphyma crassifolium (L.) L. Bolus.

Three components were extracted from the life-form variables that accounted for 66% of the variation in the data (Table 4.1). The first component described sites where high negative component values for shrub cover coincided with high positive values for bryophytes, sedges, rushes and herbs (Table 4.1). This represented a gradient of shrubby to increasingly exposed sites (exposed); the most exposed sites had diverse ground cover but provided little shrub cover. The second component represented a gradient of woody structural complexity (woody), with high positive component values for shrub cover, coarse woody debris and lichen, but negative values for grass and tussock cover (Table 4.1). The final component had high positive values for grass and tussock cover and high negative value for bare ground, representing a gradient of bare to grassy sites (grassy; Table 4.1).

Three components were extracted from vegetation density scores that accounted for 78% of the variation in the data (Table 4.2). The first component had higher values above 100 cm describing a gradient of increasing vertical density of tall vegetation (emergent). The second component had higher values from ground level and 40 cm high, describing increasing vertical vegetation density just above the ground (ground). The third component had high values from 41 – 100 cm describing increasingly dense vegetation at a middle height (shrubby; Table 4.2).

Average cover of weeds per site along with extracted components were standardised but due to correlations, two habitat variables were excluded from analysis: ground, which was correlated to two components extracted from the life-form data (woody, r = -0.508 and grassy, r = 0.588); and emergent, which was correlated with average weed cover (r = 0.768). The latter correlation was unsurprising given the presence of emergent

82 Chapter 4 – The influence of an emergent weed on wildlife boxthorn in CW. The remaining variables, exposed, woody, grassy, shrubby and average weed cover, were not strongly correlated (U” and were used as predictor variables for modelling.

Average cover of weeds reflected how the site types were selected; boxthorn sites clearly had higher weed cover than saltmarsh sites (Figure 4.1). There were no other clear differences in habitat variables between site types (Figure 4.1).

The cover and richness of fruiting vegetation consistently was highest in boxthorn sites throughout all seasons, but especially in autumn when saltmarsh sites did not provide any fruit resources (Table 4.3). Flowering vegetation also was greater in cover and richness in boxthorn sites, except in autumn when there was little difference in flowering vegetation between site types (Table 4.3).

83 Table 4.1 Factor loadings of principal components derived from analyses of life-form variables measured in Cheetham Wetlands. Components based on Principal Components Analysis (correlation matrix; Varimax rotation with Kaiser Normalisation). Where component names describe gradients: exposed, gradient of increasing exposure; woody, gradient of increasing woody cover; grassy, gradient of increasing grass cover. Where indicated, data was arcsine transformed (a) to reduce the effect of outliers. Life-form variables (%) Exposed Woody Grassy Bare ground cover a -0.079 -0.135 -0.962 Leaf litter cover 0.166 0.086 0.007 Coarse woody debris cover 0.260 0.806 -0.007 Bryophyte cover 0.827 0.195 -0.006 Lichen cover a -0.117 0.758 0.087 Herb cover 0.457 0.069 0.045 Sedge/rush cover 0.845 -0.060 -0.016 Tussock grass cover a 0.325 -0.716 0.407 Grass cover a -0.321 -0.429 0.707 Shrub cover -0.716 0.550 0.105 Variance (%) 27.075 23.144 15.633 Cumulative variance (%) 27.075 50.220 65.852 Table 4.2 Factor loadings of principal components derived from analyses of vertical vegetation density scores (up to 180 cm tall) in Cheetham Wetlands. Components based on Principal Components Analysis (correlation matrix; Varimax rotation with Kaiser Normalisation). Where component names describe gradients: emergent, gradient of increasing density of vegetation taller than 100 cm; ground, gradient of increasing vegetation density from ground level to 40 cm; shrubby, gradient of increasingly dense vegetation between 41 and 100 cm. Where indicated, data was square root transformed (s) to reduce the effect of outliers. Vegetation density between: Emergent Ground Shrubby 0–10cm 0.271 0.880 -0.104 11–20cm -0.035 0.973 0.099 21–30cm 0.008 0.918 0.280 31–40cm 0.083 0.632 0.680 41–50cm 0.168 0.106 0.921 51–60cm 0.338 0.200 0.873 61–70cm 0.482 0.433 0.650 71–80cm 0.480 0.467 0.459 81–90cm 0.529 0.563 0.384 91 – 100 cm 0.467 -0.068 0.734 101 – 110 cm 0.742 0.031 0.515 111 – 120 cm 0.751 0.456 0.142 121 – 130 cm s 0.860 -0.032 0.068 131 – 140 cm s 0.735 0.225 0.311 141 – 150 cm s 0.706 -0.150 0.304 151 – 160 cm s 0.622 0.200 0.178 161 – 170 cm s 0.786 0.142 0.198 Variance (%) 51.255 16.814 9.626 Cumulative variance (%) 51.255 68.069 77.696 Figure 4.1 Habitat variables between site types at Cheetham Wetlands. Mean standardised values (± 1 SE) of habitat variables between site types („ = boxthorn sites, n = 13; z = saltmarsh sites, n = 5). Habitat variables are average cover of weeds, and four components derived from Principal Components Analysis describing a: gradient of exposed sites, gradient of woody structural complexity, density gradient of shrubby vegetation and a gradient of grassy cover. Table 4.3 Average seasonal richness and cover of flowering and fruiting plants in different site types at Cheetham Wetlands. Resource Site type Season Average Summer Autumn Winter Spring Average cover of fruiting plants (%) Boxthorn 8.1 ± 1.5 7.9 ± 3.6 14.9 ± 4.0 5.6 ± 1.2 4.0 ± 1.9 Saltmarsh 0.2 ± 0.1 0.2 ± 0.2 0.0 ± 0.0 0.1 ± 0.1 0.4 ± 0.4 Average richness of fruiting plants Boxthorn 1.5 ± 0.1 1.6 ± 0.1 1.2 ± 0.1 1.2 ± 0.1 1.8 ± 0.3 Saltmarsh 0.2 ± 0.1 0.2 ± 0.2 0.0 ± 0.0 0.4 ± 0.3 0.2 ± 0.2 Average cover of flowering plants (%) Boxthorn 22.7 ± 2.9 26.4 ± 6.1 1.2 ± 0.9 41.9 ± 4.7 21.4 ± 2.9 Saltmarsh 10.5 ± 3.1 8.5 ± 7.7 2.2 ± 1.4 15.2 ± 8.6 16.0 ± 4.3 Average richness of flowering plants Boxthorn 4.7 ± 0.5 5.2 ± 0.7 0.4 ± 0.2 5.5 ± 0.5 7.5 ± 0.7 Saltmarsh 3.0 ± 0.7 2.6 ± 1.8 0.6 ± 0.4 3.4 ± 1.5 5.2 ± 1.5 Chapter 4 – The influence of an emergent weed on wildlife

4.4.2 Birds Twenty-nine bird species were recorded (Table 4.4). Richness varied within and between site types across the seasons (1 – 11 species). No species was present in every site, but some species were more common than others. The singing honeyeater Lichenostomus virescens was most common (78% of sites, total activity index of 66) and the white-fronted chat Epthianura albifrons also was common (56%, 47). The willie wagtail Rhipidura leucophrys was more widespread than the white-fronted chat, but had a much lower activity index (Table 4.4) so was not considered common. Thus, only singing honeyeaters and white-fronted chats were used in further analyses.

Generalised linear models suggested that site type influenced bird richness in autumn only, when saltmarsh sites had significantly higher richness of birds (Table 4.5; boxthorn, 2.8 ± 0.78; saltmarsh, 5.8 ± 2.6). Bird composition differed between site types within each season and when pooled across seasons (Table 4.5). The presence of singing honeyeaters in boxthorn sites made the highest contribution (11 – 27%) to dissimilarities in bird composition between site types in all seasons except spring. In spring, the presence of white-fronted chats in saltmarsh sites contributed most (23%) to the dissimilarity in composition.

Akaike weights were low (< 0.9) for all models that examined the influence of habitat variables on birds (Table 4.6). The number of models in the 95% confidence set was comparatively lower for white-fronted chats (9) than for bird richness (23) and singing honeyeaters (20). Variation explained by models ranged from 4 % (models for bird richness) to 70% (models for white-fronted chats; Table 4.6). Consequent model averaging suggested that no single habitat variable was particularly influential for bird richness (Table 4.7). In contrast, activity of white-fronted chats showed a substantial negative response to increases in weed cover. Model averaging for singing honeyeaters showed the reverse; their activity was positively influenced by increases in weed cover, but this response was more variable and not as strong as the negative response to weed cover displayed by white-fronted chats (Table 4.7). 88 Chapter 4 – The influence of an emergent weed on wildlife

4.4.3 Small mammals The house mouse Mus domesticus and black rat Rattus rattus were the only small mammal species captured (3379 trap nights). Black rats made up less than three percent of all captures so were not analysed further.

Capture rates (pooling site types) were significantly lower in summer compared to the other seasons (G2 = -30.349, p < 0.001; summer, 0.101 ± 0.060; autumn, 0.294 ± 0.093; winter, 0.223 ± 0.083; spring, 0.244 ± 0.084). When accounting for site types, house mouse capture rate was significantly higher in boxthorn sites in summer (Table 4.5; boxthorn, 0.353 ± 0.040; saltmarsh, 0.139 ± 0.050), but did not differ between site types in any other season or when averaged over season (Table 4.5).

House mouse capture rates differed with distance from boxthorn (G2 = 11.011, p = 0.012), but this difference was not a consistent decline in capture rates as distance from boxthorn increased (beneath, 0.280 ± 0.025; five, 0.180 ± 0.021; ten, 0.263 ± 0.028; 15, 0.236 ± 0.024).

Just five models were in the 95% confidence set that addressed the influence of habitat variables on house mouse capture rates. This indicated there was a good chance that one of these models best represented the data. One model stood out, explaining 70% of the variation in house mouse capture rates (Table 4.6). This model consisted of the grassy gradient and shrubby density gradient. Model averaging further suggested that capture rates of house mouse were positively influenced by the shrubby density gradient and the grassy cover gradient, with summed wi > 0.9 (Table 4.7).

89 Table 4.4 Bird species recorded in Cheetham Wetlands (all seasons). The percentage of sites at which each species was observed for each site and the total percentage is shown. In the final column, the total activity index for each species is given in parentheses. Species indicated (S) were most common and were used for further analysis. Names follow Christidis and Boles (2008), with introduced species indicated (*). Common name Scientific name Site type Total Boxthorn Saltmarsh (n = 13) (n = 5) (n = 18) Singing honeyeaterS Lichenostomus virescens 92 40 78 (66) Willie wagtail Rhipidura leucophrys 62 80 67 (15) White-fronted chatS Epthianura albifrons 38 100 56 (47) Common starling* Sturnus vulgaris 46 0 33 (23) Striated fieldwren Calamanthus fuliginosus 31 40 33 (12) White-browed scrub wren Sericornis frontalis 23 20 22 (15) Superb fairy-wren Malurus cyaneus 8 40 17 (17) House sparrow* Passer domesticus 8 40 17 (6) New Holland honeyeater Phylidonyris novaehollandiae 15 20 17 (4) Australian magpie Cracticus tibicen 23 0 17 (3) Chestnut-rumped heathwren Hylacola pyrrhopygia 0 40 11 (3) Little raven Corvus mellori 15 0 11 (3) Horsefield's bronze-cuckoo Chalcites basalis 15 0 11 (2) Table 4.4 Continued Common name Scientific name Site type Total Boxthorn Saltmarsh Little grassbird Megalurus gramineus 0 40 11 (2) White-plumed honeyeater Lichenostomus penicillatus 0 40 11 (2) Silvereye Zosterops lateralis 8 0 6 (11) Blue-winged parrots Neophema chrysostoma 8 0 6 (3) Red-capped plover Charadrius ruficapillus 0206 (2) Australasian pipit Anthus novaeseelandiae 0206 (1) Black-tailed native-hen Tribonyx ventralis 8 0 6 (1) Brown falcon Falco berigora 8 0 6 (1) Common blackbird* Turdus merula 8 0 6 (1) Eurasian skylark* Alauda arvensis 8 0 6 (1) Masked lapwing Vanellus miles 0206 (1) Purple swamphen Porphyrio porphyrio 0206 (1) Table 4.5 Comparison of bird species richness, bird assemblages and house mouse capture rates between site types over four seasons and when averaged over seasons at Cheetham Wetlands. Generalised linear models were used to test for differences between site types for both bird species richness and small mammal capture rates, with results tested against a t distribution for significance (df = 1, 16). To test for differences in bird assemblages between site types, ANOSIM was used based on presence/absence data. Statistically significant results (p < 0.05) are indicated in bold. Comparison Test statistic Summer Autumn Winter Spring All seasons House mouse t -3.201 -0.023 -0.954 -0.788 -1.589 Bird richness t 1.465 3.039 1.100 -0.194 1.518 Bird species assemblages Global R 0.612 0.373 0.489 0.577 0.459 Table 4.6 Model selection results for birds and house mouse at Cheetham Wetlands. Shown are the parameters incorporated in the model, total number of parameters (K), explained variation, AICc values, AICc differences (Ui), and Akaike weights (wi).

Models are listed in descending order according to Ui. Only models with Ui ”DUH listed or if the number of Ui ”H[FHHGHGILYHRQO\WKHEHVWILYHPRGHOVDUHVKRZQ

Response Models Explained K AICc Ui wi variable variation Bird species woody 13% 2 86.4 0.000 0.199 richness weed + woody 23% 3 87.5 1.140 0.113 weed 5% 2 87.9 1.570 0.091 exposed 4% 2 88.2 1.850 0.079 Singing weed 28% 2 85.5 0.000 0.196 honeyeater weed + exposed 38% 3 86.1 0.620 0.144 weed + woody 37% 3 86.4 0.972 0.121 weed + exposed + woody 47% 4 87.2 1.679 0.085 White-fronted weed 65% 2 39.5 0.000 0.328 chat weed + grassy 70% 3 40.6 1.048 0.194 House mouse shrubby + grassy 70% 3 -42.4 0.000 0.632 Table 4.7 Model averaging results for birds and house mouse at Cheetham Wetlands.

The most influential variables for wildlife have the highest sum of Akaike weights (wi). Highlighted in bold are those variables for which the 95% confidence interval does not include zero, indicating they are important variables in determining wildlife.

Response Predictor Coefficient Standard error Sums of wi Bird species richness weed -0.277 0.460 0.383 exposed -0.119 0.340 0.258 woody 0.510 0.521 0.554 shrubby 0.077 0.322 0.225 grassy 0.073 0.311 0.223 Singing honeyeater weed 0.980 0.546 0.770 exposed 0.340 0.461 0.485 woody 0.382 0.493 0.320 shrubby -0.046 0.273 0.213 grassy 0.094 0.319 0.189 White-fronted chat weed -1.417 0.398 0.999 exposed -0.023 0.144 0.146 woody 0.100 0.279 0.269 shrubby -0.052 0.245 0.188 grassy -0.084 0.179 0.316 House mouse weed 0.000 0.006 0.118 exposed -0.001 0.006 0.122 woody -0.002 0.007 0.154 shrubby 0.069 0.015 0.998 grassy 0.052 0.015 0.978 Chapter 4 – The influence of an emergent weed on wildlife

4.5 Discussion

Habitat loss and degradation resulting from vegetation clearance has been connected to the decline of wildlife, worldwide (Boakes et al. 2009). As vegetation removal continues, protection of existing habitat is crucial for species conservation (Boakes et al. 2009). The potential for weeds to provide such habitat should be considered. By definition, vegetation dominated by weeds is unlikely to meet our ideals of habitat, but, management plans and priorities may need to be assessed depending on wildlife responses to weeds.

4.5.1 The influence of boxthorn on birds This study revealed different bird responses to vegetation that incorporated emergent boxthorn. Throughout most of the year, this vegetation provided habitat for a similar number, but a different composition, of bird species compared to vegetation where boxthorn was absent.

Changes to both bird richness and composition also occurred throughout the seasons in the different site types. These changes often coincided with differences in fruiting and flowering vegetation: in autumn, bird richness in boxthorn sites was significantly lower than in saltmarsh sites, which was reflected by much lower floral availability in boxthorn sites. Bird composition between site types differed throughout all seasons, but spring was the only season when singing honeyeaters were not the main drivers of these differences and it was the only time of year when the flowering and fruiting vegetation was similar between site types. Such coincidental changes indicate that boxthorn may be providing food resources and their seasonal availability may influence bird assemblages.

The potential for fruiting or flowering weeds to cause shifts in bird assemblages has been recorded elsewhere (French and Zubovic 1997; Gosper 2004). Weeds that cause changes to nectar and/or fruit availability can influence nectarivores and fruigivores in particular (French et al. 2005). The effects of fruit and flowers on the birds in this study were only inferred through seasonal changes, and not directly assessed, so any changes

95 Chapter 4 – The influence of an emergent weed on wildlife to such resources cannot be strongly linked to differences seen in assemblages, which highlights a need for further research to clarify these parallels.

The shifts observed in bird assemblages between site types were driven primarily by singing honeyeaters and white-fronted chats. White-fronted chats avoided boxthorn sites, were more associated with saltmarsh sites and displayed a strong negative relationship with amount of weed cover. The response of singing honeyeaters was reversed. Their presence in boxthorn sites almost always was the major contributor to differences in bird assemblages between the site types and their relationship with weed cover was positive, despite variability in its strength. This indicates that for this species at least, boxthorn was important. When boxthorn previously was suggested to be of value, it was implied that this was only in circumstances where native vegetation was absent (CRC for Australian Weed Management 2007). This is not the case in CW, suggesting that boxthorn can have value to some species in situations where it grows amongst native vegetation as well. This could have weed management implications, but the negative response of white-fronted chats and general changes in bird assemblages highlight the need for clear management objectives and priorities.

4.5.2 The influence of structure on species Weed cover was the only habitat variable that clearly influenced the two bird responses that were measured. Considering that weed cover was correlated with the emergent vegetation gradient described, it would be expected that weed cover would, mostly, represent the cover of boxthorn. Therefore, the influence of emergent structure from weed cover cannot be separated, which indicates that it may be the taller standing characteristic of boxthorn that is influencing the bird assemblage, rather than the weed species per se. Therefore, the reverse responses of singing honeyeaters and white- fronted chats to weed cover could be explained by different responses to taller vegetation in the otherwise low lying saltmarsh vegetation.

96 Chapter 4 – The influence of an emergent weed on wildlife

A negative response to emergent vegetation by the white-fronted chat reflects the species’ habit of feeding and nesting near to the ground (Major 1991). Past research also has suggested that chats avoid nesting in boxthorn, as it lacks structure closer to the ground (Major 1991), thereby also supporting the negative response to the taller nature of boxthorn. The tall nature of boxthorn in CW also afforded high perching positions for predators like the brown falcon Falco berigora, which was incidentally seen sitting at the very top of boxthorn plants throughout CW. This may contribute to the avoidance of boxthorn by white-fronted chats. A similar perching behaviour was observed by singing honeyeaters, and elsewhere they have been known to increase in abundance where vegetation is augmented by medium and taller shrubs, because they provide more foraging opportunities (Luck et al. 1999). This indicates that singing honeyeaters may be exploiting a combination of increased food resources as well as the taller structural aspect offered by boxthorn. However, positive responses to weeds can be superficial in some situations, with other studies indicating the weed is actually an ecological trap (Remes 2003; Cully Nordby et al. 2009). More in-depth study would determine the exact nature of the interaction between singing honeyeaters and boxthorn, and identify just how important the weed is for the species.

In comparison, the house mouse was influenced by low level structure, as demonstrated by a strong positive response to increases in vegetation density of lower shrubs and cover of grasses. The importance of low, dense understorey vegetation for the house mouse has been indicated elsewhere for its ability to provide shelter (Mutze 1991). In CW, whether this shelter comes from weeds or native vegetation seems irrelevant; the species did not respond to weed cover; however, in summer, when capture rates were at the lowest in CW, mice were captured at a much higher rate in boxthorn sites. This suggests that in a time when mouse activity is normally low (Mutze 1991), boxthorn sites may provide better refuge than native vegetation sites. It is difficult to know exactly why this is the case, since modelling demonstrated such a strong influence of vegetation structure. It is possible that with a change of season, annual plants may have contributed

97 Chapter 4 – The influence of an emergent weed on wildlife to structural changes in the vegetation, or that food resources such as seeds, were higher in boxthorn sites, but this was not monitored. Regardless, house mice were still present in saltmarsh sites during summer (albeit at a much lower rate), and for the rest of the year native vegetation provided habitat for them just as well as vegetation with boxthorn. Furthermore, distance from boxthorn plants did not change this. Thus, controlling boxthorn, which is often suggested because of its ability to harbour mammalian pests (CRC for Australian Weed Management 2007), would not be expected to remove all house mouse habitat in CW.

4.5.3 Facing the future This research has highlighted that wildlife responses to boxthorn can vary. Some species favour boxthorn, others avoid it, and some have shown no preference either way. Overall, it means wildlife assemblages can be affected by boxthorn and, consequently, weed management may cause further changes to these assemblages. Thus, managers are faced with the questions of what changes will occur, whether biodiversity objectives will be met by causing those changes and, if not, how best to respond.

When a positive wildlife interaction with weeds is observed, managers need to know how the species would be affected by weed management, and what methods to use that are more sympathetic to wildlife. It has been recommended that gradual weed removal in conjunction with revegetation is the best approach to weed management when such interactions occur (Gosper and Vivian-Smith 2006; Carlos and Gibson 2010). However, in a system like CW, where boxthorn grows interspersed in the native vegetation, replacements may not be as important for more isolated plants compared to areas where dense monocultures exist. Furthermore, if a food resource, as well as structure, is provided, replacement vegetation may not fulfil all the aspects that the weeds provided.

If weeds are acting as a food resource, a further management issue arises: birds may be significant contributors to the spread of the weed (Brunner et al. 1976; Spennemann and Allen 2000; Stansbury 2001). In Australia, birds already have been implicated in the

98 Chapter 4 – The influence of an emergent weed on wildlife spread of many weeds, including boxthorn, because they feed on their fruits (Loyn and French 1991). Yet, it is unknown whether such food resources are critical dietary components, or how adaptable the species are that utilise them. Even if these fruits are found to be important, managers still will need to determine whether minimising the spread of weeds by birds or, conserving a potentially critical food resource, is their priority.

There are clearly many gaps in the understanding of exactly how weeds impact wildlife. Perhaps more important, is the need to clarify how weed management may further effect these interactions. Assessments of wildlife assemblages before and after weed management, along with appropriate controls, could permit a meta-analysis, potentially correlating life history traits of both wildlife and weeds with likely wildlife responses. Incorporating wildlife into the monitoring process that is currently recommended before, during and after weed management (Reid et al. 2009) would not only contribute to such an analysis, but also allow for adaptations in response to any adverse effects as they happen. With improvements to our understanding, management can only become better informed.

99 4.6 Appendix A

To account for differences in bird species accumulation in different habitats we set a survey duration which represented an asymptote of species richness in both site types. Survey time was set at 30 minutes because bird richness had reached an asymptote in both site types by this time (Figure A4.1).

Figure A4.1 Cumulative bird species richness over timed surveys (natural log + 1 transformed, ± 1 SE) in different site types in Cheetham Wetlands (CW): boxthorn sites (top) and saltmarsh sites (bottom). Chapter 5

‘Wildlife-friendly’ weed management sustains birds and

small mammals

Australian magpie Cracticus tibicen perched on African boxthorn Lycium ferocissimum in Cheetham Wetlands, Victoria, Australia.

101 Chapter 5 – Wildlife friendly weed management

5.1 Abstract

‘Wildlife-friendly’ weed management uses active measures to minimise or prevent negative impacts of weed management on local wildlife. Approaches include retention of weed structure, changing timing of management to avoid wildlife breeding seasons, or the gradual removal of weeds in conjunction with revegetation of native plants. The success of such approaches, although commonly used, has not been assessed. The short- term impacts of wildlife-friendly weed management which retained the original structural complexity offered by weeds were investigated. Birds and small mammal responses to management were assessed in two ecosystems in which these taxa were known to be influenced by the structural characteristics of weeds. Bird and small mammal occurrence and activity were not altered by wildlife-friendly weed management. Even the activity of bird species that favoured weeds remained similar after weed management. This study indicated that, in the short-term, weed management techniques that allowed for the retention of weed structure effectively sustained wildlife using weeds. Whether these trends would persist for the long-term remains to be determined.

Keywords: Weeds, wildlife, bird, weed management, wildlife-friendly, management impact.

102 Chapter 5 – Wildlife friendly weed management

5.2 Introduction

The effects of invasive, unwanted plants (commonly referred to as ‘weeds’) on wildlife vary. Weeds can benefit wildlife such as birds, mammals and invertebrates, for example, by providing food resources in the form of fruits (Ekert and Bucher 1999), seeds (Loyn and French 1991) or nectar (Graves and Shapiro 2003). In other circumstances weeds can have adverse impacts on wildlife, such as contributing to habitat modification (Lindsay and French 2006; Stralberg et al. 2010) and harbouring pest animals that compete for resources (Peter 2000).

When weeds are identified as beneficial for wildlife, ‘gradual’ removal of weeds in conjunction with revegetation (or regeneration) of native plants to eventually replace the weed resources, has been recommended (Gosper and Vivian-Smith 2006; Carlos and Gibson 2010). Some managers adopt these recommendations, while others use different methods, such as retaining weed structure by choosing to manage weeds with herbicide (Chapter 2). Such ‘wildlife-friendly’ weed management, which takes active measures to account for local wildlife that may be negatively impacted by management, is not uncommon (Chapter 2). Yet, the effectiveness of wildlife-friendly weed management in supporting wildlife has not been studied. In fact, it remains to be determined whether such ‘wildlife-friendly’ weed management may cause wildlife decline.

When faced with severe environmental change, species must respond rapidly (Tuomainen and Candolin 2011), and thus whether management is or is not appropriate would become apparent at short time-scales. This study investigated the short-term impacts of small-scale wildlife-friendly weed management on birds and small mammals. Weed management that retained weed structure was conducted in two ecosystems where vegetation structure was known to influence some bird and small mammal species (Chapters 3 and 4). It was anticipated that such management would assist in the retention of the wildlife species using weeds.

103 Chapter 5 – Wildlife friendly weed management

Specifically, the aim was to: 1) assess if weed management had an impact on birds and small mammals in the short-term; 2) identify the type and direction of any impacts.

5.3 Methods

5.3.1 Study area and design Two different ecosystems were used as case studies, a woodland (You Yangs Regional Park, YY) and a wetland (Cheetham Wetlands, CW), to investigate the impacts of weed management on wildlife (see Chapters 3 and 4). The design reflected a modified BARCI type design (before-after, reference-control-impact; Lake 2001). So, within each ecosystem there were three treatment types: ‘reference’ sites were dominated by native vegetation; ‘control’ sites by weed vegetation that was left unmanaged and ‘impact’ sites by weed vegetation that was managed. Birds and small mammal surveys were conducted in sites at different times: before weed management was implemented at impact sites (‘before’); 1 – 18 days after management (‘after-1’); 42 – 60 days after management (‘after-2’). This allowed for detection of acute wildlife responses to management during the period when weed leaves still appeared similar to pre-managed conditions, in ‘after- 1’. It also allowed for the detection of responses to managed weeds when leaves had wilted, browned, and the majority had fallen to the ground, in ‘after-2’. This design allowed for any observed changes in impact sites after management to be clearly attributed to management (before-after, control-impact), and the assessment of the direction and magnitude of such changes (before-after, reference-impact).

Management actions in the two ecosystems were tailored to reduce impacts on wildlife by retaining weed structure. In YY, management involved spraying glysophate on the dominant weedy shrub, African boneseed Chrysanthemoides monilifera subsp. monilifera (DC.) T. Norl. Consequently, boneseed died but much of its structure was retained. In CW, African boxthorn Lycium ferocissimum Miers., a shrubby weed, grows alone or in stands, amongst native vegetation, therefore management involved cutting down individual boxthorn plant/s in impact sites, and painting the remaining stump with

104 Chapter 5 – Wildlife friendly weed management glysophate. Cut boxthorn was left in situ, nearby, to continue to provide a structural resource for wildlife. The extent and shape of sites between ecosystems was different, to ensure suitable sampling units.

5.3.1.1 You Yangs Regional Park The YY is near Melbourne, Australia (37º 56’ S, 144º 25’ E) and boneseed is found through 65% of the park’s understorey (Roberts 2008). In the few remaining areas where boneseed is not dominant, open woodland or woodland with a native shrub understorey exists. Twenty-five one hectare sites were randomly selected. These were either dominated by boneseed (eight control sites and seven impact sites) or by native dominated vegetation that acted as reference sites. There were two types of reference sites, those with an open understorey (five open reference sites) and those with a native shrub understorey (five shrub reference sites). Sample sizes differed between treatments but were proportional to availability of suitable vegetation and were randomised within this vegetation. Sampling of wildlife occurred in the centre of the sites. In impact sites, herbicide was applied to boneseed in early March 2010.

5.3.1.2 Cheetham Wetlands Covering 420 ha, CW lies adjacent to Port Phillip Bay, near Melbourne, Australia (37º 54’ S, 144º 48’ E). The original wetland ecosystem was altered to operate as a commercial saltworks until 1992, and now is a protected area managed for conservation (Antos et al. 2007). Boxthorn is a prominent weed in this ecosystem, standing much taller than the low growing and treeless native saltmarsh that it grows amongst. Eighteen embankments throughout the wetland were selected as study sites. Vegetation in sites either included boxthorn (six control sites, seven impact sites), or did not (five reference sites that consisted of typical saltmarsh vegetation). Sampling was constrained within each site to two 10 x 50 m sampling areas, > 50 m away from each other. Sampling areas in control and impact sites centred around one or more randomly-selected

105 Chapter 5 – Wildlife friendly weed management boxthorn plants. In impact sites, management was applied to these central boxthorn plants in August 2010.

5.3.1.3 Birds Point counts were conducted on days without strong wind or persistent rain (White et al. 2005). All bird species seen were counted and additional individuals of a species were only counted when there was certainty they were separate individuals. Thus counts of individual species are considered to represent a measure of species usage/activity and are referred to as an ‘activity index’.

Survey durations were established separately in YY and CW (Chapters 3 and 4) so that bird detectability was not affected by the different vegetation types within the two ecosystems. In YY, birds seen within a 50 m radius of the count point were included and surveys were conducted for 60 minutes. Due to the small and linear nature of sampling areas in CW, surveys were conducted in the position that afforded the view of the whole sampling area and were conducted for 30 minutes each (total of 60 minutes per site). Data from sampling areas in each CW site were averaged so that each site was considered a datum.

5.3.1.4 Small mammals Sixteen Elliott and four cage traps were set in each site for three consecutive nights during each sampling period. Traps were baited using a mixture of rolled oats, peanut butter, honey, linseed oil and vanilla essence; bedding was provided (hollofil), and traps covered with plastic (Tasker and Dickman 2002). Captured animals were identified, weighed, assessed for reproductive condition, and marked before release.

The position of traps in each ecosystem differed due to the difference in site shape and size. In YY, traps were set in a grid in the centre of the site, with four lines of four Elliott traps, set ten metres apart and parallel to each other. A cage trap was set in each of the four corners of the grid. In CW, a transect of eight Elliott traps, deployed five metres

106 Chapter 5 – Wildlife friendly weed management apart, was set in each of the two sampling areas of each site, reflecting the linear nature of sites. Four cage traps were set in each site: at the end and middle of each transect. In control and impact sites, transects were set so the central boxthorn plant/s of each sampling area were in the middle of the transect with a cage trap and two Elliott traps beneath their foliage (approximately five metres apart).

5.3.1.5 Vegetation Vegetation was assessed in 1 x 1 m quadrats at the 16 Elliott trap locations, in each site. Total weed cover (%), regardless of strata, was estimated in each quadrat. Cover of bare ground, leaf litter, herbs, grass, sedges/rushes (CW only), lichen, bryophytes, shrubs, tussock grass (CW only), and coarse woody debris were also estimated. In CW, coarse woody debris referred to dead branches, fallen or cut from shrubs (including managed boxthorn), while in YY it referred to fallen timber from trees and shrubs. In YY, canopy cover, its relative health and the proportion of living saplings also was calculated for each quadrat. The vertical distribution of vegetation density was evaluated using the average number of touches of living or dead vegetation in 10 cm increments up a 180 cm vertical pole (up to ten touches per increment).

5.3.2 Variable selection and data analysis Due to the divergent survey methods required by the nature of the two ecosystems, data from YY and CW were analysed separately, but the same analytical processes were used. Different methods used between the ecosystems included different sizes and shapes of sites, wildlife survey techniques and weed management methods.

5.3.2.1 Vegetation variables For data from each site, Principal Components Analysis (PCA) was used to reduce 1) the number of vegetation cover variables and 2) the vertical vegetation density measurements. Some variables were square-root transformed to reduce the influence of outliers (YY vegetation density data only). These components as well as standardised

107 Chapter 5 – Wildlife friendly weed management average weed cover measures were used to describe the vegetation in the YY and CW. Henceforth these are referred to as ‘habitat variables’.

5.3.2.2 Wildlife variables Different measures were used to describe and quantify bird assemblages and activity: compositional data that consisted of presence/absence of each species, total bird richness (the number of different species), total bird activity (the total number of birds) and the activity of the most ‘common’ species. Species defined as common were widespread REVHUYHGLQ•RIVLWHV DQGKDGDKLJKDFWLYLW\LQGH[ • 

For small mammal activity, each three-day trapping period was reduced to an average capture rate per site. Any trap failures were excluded from analysis.

5.3.2.3 Statistical analyses To determine whether the vegetation and wildlife variables were influenced by weed management, or if time or treatment influenced the response variables, model selection and multimodel inference based on the information-theoretic approach (Burnham and Anderson 2002) was used to indicate which of five competing models best supported the data.

Generalised linear mixed models (glmm) were used in model comparisons. To account for non-independence over time, site was entered as a random factor in all models. Residuals indicated error structures applied to each model were appropriate. Models compared were: time (before, after-1, after-2), treatment (reference, control, impact), time + treatment, time + treatment + (time × treatment interaction), and a null model (which only included a random factor of site).

Models were compared using Akaike Information Criteria (using the lme4 and AICcmodavg packages in R ver. 2.14.2; R Development Core Team 2012). Models were corrected for small sample size (AICc) and a difference (Ui) between the AICc of each

108 Chapter 5 – Wildlife friendly weed management model and that of the best fitted model calculated. Models that had considerable support had a Ui ”, as did models with Akaike weights (wi) > 0.9 (Burnham and Anderson

2002). Multiple models with Ui ”RUwi < 0.9 indicated uncertainty as to whether one model was more likely than the other.

Substantial support for a model which included the time × treatment interaction term suggested that weed management had an impact on the response variable, while support for models which included time or treatment on their own indicated main effects of sampling time or treatment. Means and standard errors were used to clarify the nature of effects, such as acute or delayed impacts (differences between after-1 and after-2 in impact sites), and the direction of impacts (differences between impact and reference or control sites, after management).

To assess whether weed management influenced the composition of birds, analysis of similarities (ANOSIM) was conducted in PRIMER v6 (2007) and pair-wise tests were used to indicate whether weed management had an impact on bird composition. An R statistic close to zero indicated groups were barely separable (Clarke and Gorley 2006). Thus, if management had an effect on bird composition, after-1 and after-2 surveys in impact sites should be different from before surveys (with R statistics further from zero), but surveys in control or reference sites should be similar regardless of time (with R statistics close to zero).

Bubble plots and variograms (using the geoR and gstat packages) confirmed spatial independence of vegetation and univariate wildlife variables. Using the RELATE routine in PRIMER, bird species composition was confirmed as spatially independent EHWZHHQVLWHVLQERWKHFRV\VWHPV <<ȡ S &:ȡ 116, p = 0.019).

109 Chapter 5 – Wildlife friendly weed management

5.4 Results

5.4.1 You Yangs 5.4.1.1 Vegetation In YY two components explained 55% of the variation in the cover of different vegetation life-forms throughout the study period. The first component had high positive values for canopy cover and health, as well as proportion of living saplings and coarse woody debris, but high negative values for shrub cover (Table 5.1). This represented a gradient of vegetation with a complex and healthy canopy and a relatively open understory, to vegetation with little canopy complexity and high cover of shrubs; hereafter referred to as canopy. The second component had high positive values for grass and herb cover, but high negative values for cover of leaf litter (Table 5.1). This represented a gradient of vegetation with ground cover dominated by herbaceous plants at one end of the scale, and vegetation with ground mainly covered by leaf litter at the other; hereafter referred to as ground.

In YY two components were identified explaining 61% of the variation in the vertical distribution of vegetation density. The first component had high positive values for vegetation heights 51 – 130 cm, describing the density of tall shrubs (Table 5.2). The second component had high positive values of vegetation up to 50 cm above ground, describing the density of low shrubs (Table 5.2).

Model selection indicated wildlife-friendly boneseed management was successful in retaining structure while reducing weed cover. The weed cover model including the treatment × time interaction term had substantial support (Table 5.3), where there was lower weed cover in impact sites after management but similar weed cover throughout time in control and reference sites (Figure 5.1). For the other YY habitat variables, models that included time as a predictor and/or null models had the most support (Table 5.3).

110 Chapter 5 – Wildlife friendly weed management

5.4.1.2 Birds Forty species were detected over the sampling period. The most common species were the brown thornbill Acanthiza pusilla, followed by the yellow thornbill Acanthiza nana and superb fairy-wren Malurus cyaneus (Table 5.4).

Model selection indicated that boneseed management did not impact bird richness, total bird activity or activity of any common species as there was no support for any models that incorporated the treatment × time interaction term (Table 5.3). The null model for species richness had the most support, while for total bird activity the model with greatest support included both treatment and time (but not the interaction; Table 5.3). There was uncertainty for all models of common bird activity (Table 5.3), yet little evidence of an impact of weed management.

There was no difference in bird assemblages between treatments (global R = -0.116) or times (global R = -0.002). Bird assemblages were dissimilar in all treatment-time combinations (average dissimilarity > 60%), and pair-wise tests did not reveal any evidence of an effect of weed management in impact sites over time (Table 5.5).

5.4.1.3 Small mammals The house mouse Mus domesticus and the black rat Rattus rattus were the only small mammal species captured (3709 trap nights). Black rats accounted for < 3% of small mammal captures, so were not analysed further. Boneseed management seemed to have little effect on house mouse capture rates: model selection indicated support for the null model was very high (wi = 0.853; Table 5.3).

111 Table 5.1 Factor loadings for principal components derived from average life-form variables measured at the You Yangs Regional Park (YY) and Cheetham Wetlands (CW). Life-form variables (%) YY CW Canopy Ground Marshy Herbaceous Debris Tussock grass cover 0.262 -0.359 -0.456 Sedge/rush cover 0.819 -0.230 0.219 Bare ground cover -0.229 -0.651 -0.091 -0.041 -0.266 Leaf litter cover -0.408 -0.028 -0.101 0.032 0.038 Coarse woody debris cover 0.665 -0.423 0.196 0.139 0.845 Grass cover -0.238 0.799 -0.321 -0.106 -0.279 Herb cover -0.231 0.814 0.154 -0.577 0.643 Bryophyte cover -0.109 -0.054 0.884 0.210 0.040 Lichen cover -0.100 0.128 0.160 0.869 0.170 Shrub cover -0.830 0.045 -0.604 0.708 -0.122 Canopy cover 0.878 -0.091 Healthy foliage 0.884 -0.181 Living saplings 0.897 0.014 Variance (%) 38.855 16.072 26.367 19.639 13.737 Cumulative variance (%) 38.855 54.927 26.367 46.005 59.742 Table 5.2 Factor loadings for principal components representing the vertical density of vegetation (to 190 cm) at the You Yangs Regional Park (YY) and Cheetham Wetlands (CW). You Yangs data were square root transformed to reduce influence of outliers. Number of touches YY CW Tall shrubs Low shrubs Emergent Shrubby 0 – 10 cm -0.051 0.643 0.036 -0.029 11– 20 cm 0.159 0.927 -0.020 0.097 21 – 30 cm 0.453 0.807 -0.014 0.434 31 – 40 cm 0.562 0.700 0.108 0.751 41 – 50 cm 0.494 0.715 0.307 0.876 51 – 60 cm 0.669 0.520 0.240 0.915 61 – 70 cm 0.697 0.584 0.425 0.679 71 – 80 cm 0.521 0.701 0.422 0.477 81 – 90 cm 0.691 0.506 0.664 0.409 91 – 100 cm 0.639 0.556 0.598 0.629 101 – 110 cm 0.893 0.135 0.777 0.497 111 – 120 cm 0.782 0.249 0.877 0.171 121 – 130 cm 0.663 0.018 0.902 0.131 131 – 140 cm 0.308 -0.043 0.926 0.231 141 –- 150 cm -0.061 0.013 0.564 0.515 151 – 160 cm 0.194 0.110 0.407 0.392 161 – 170 cm 0.181 -0.105 0.800 0.181 171 – 180 cm -0.060 0.088 0.449 -0.002 181 – 190 cm -0.056 -0.107 Variance (%) 46.881 14.567 48.888 17.096 Cumulative variance (%) 46.881 61.448 48.888 65.984 Table 5.3 Model selection results for habitat variables and wildlife metrics in two ecosystems (You Yangs Regional Park, YY; Cheetham Wetlands, CW) where experimental weed management was conducted. For each response variable, mixed models were compared. Models included a random factor of site and different combinations of fixed factors: time + treatment + (time × treatment); time + treatment; time; treatment; or no fixed factor/s (‘null’). Presented are the ecosystems in which the study was undertaken, the response variable and their best model/s, total number of parameters in the model/s (K), Log- likelihood estimate/s (L-L), AICc value/s, AICc differences (Ui), and Akaike weight/s (wi). Models are listed in descending order according to Ui and only models with Ui ”DUHOLVWHG6RPHresponse variables required a logarithmic transformation

(L).

Ecosystem Response variable Model K L-L AICc Ui wi YY Weed Treatment × time 14 -20.766 76.532 0.000 1.000 Canopy Null 3 -26.32 58.978 0.000 0.734 Ground Null 3 -51.827 114.523 0.000 0.867 Tall shrubs Null 3 -71.547 149.433 0.000 0.510 Time 5 -69.725 150.319 0.887 0.327 Low shrubs Null 3 -58.064 122.467 0.000 0.473 Time 5 -55.808 122.485 0.018 0.469

Bird richness (L) Null 3 -16.017 38.372 0.000 0.691

Bird activity (L) Treatment + time 8 -13.008 44.199 0.000 0.731

Brown thornbill (L) Null 3 21.034 -35.729 0.000 0.584 Treatment 6 23.637 -34.039 1.69 0.251 Table 5.3 Continued

Ecosystem Response variable Model K L-L AICc Ui wi YY Yellow thornbill Time 4 -41.867 92.306 0.000 0.589 Null 2 -44.769 93.705 1.399 0.293 Superb fairy-wren Treatment + time 7 -51.085 117.843 0.000 0.572 Treatment 5 -53.843 118.555 0.712 0.400

House mouse (L) Null 3 29.285 -52.231 0.000 0.812 CW Weed Treatment × time 11 -21.394 71.074 0.000 1.000 Marshy Null 3 -44.945 96.370 0.000 0.478 Treatment 5 -42.717 96.685 0.315 0.408 Herbaceous Treatment 5 -27.096 65.442 0.000 0.424 Null 3 -29.933 66.347 0.905 0.270 Debris Treatment × time 11 -35.767 99.821 0.000 0.999 Emergent Treatment 5 -58.919 129.088 0.000 0.444 Treatment × time 11 -50.577 129.440 0.352 0.373 Shrubby Treatment 5 -41.534 94.317 0.000 0.442 Null 3 -44.221 94.922 0.604 0.326 Treatment × time 11 -19.364 67.014 1.572 0.193 Table 5.3 Continued

Ecosystem Response variable Model K L-L AICc Ui wi

CW Bird richness (L) Null 3 10.484 -14.488 0.000 0.634 Bird activity Null 2 -48.012 100.259 0.000 0.596 Time 4 -46.499 101.813 1.555 0.274 Singing honeyeater Treatment 4 -30.148 69.113 0.000 0.903 House mouse Null 3 -96.191 198.862 0.000 0.451 Time 5 -94.250 199.750 0.888 0.289 Type 5 -94.797 200.844 1.982 0.168 Figure 5.1 Mean standardised values (± 1 SE) of those habitat variables which were impacted by weed management, as shown by substantial support for models that included a treatment × time interaction term. Panel A shows average cover of weeds in the You Yangs Regional Park. Values are shown before (z), and after weed management (S, after-1; T, after-2) in different treatments: reference (open or shrubs), control (unmanaged weeds), and impact sites (managed weeds). Figure 5.1 Continued Panel B shows average cover of weeds in Cheetham Wetlands; panel C the average Debris cover (derived from Principal Components Analysis) in Cheetham Wetlands. Table 5.4 Bird species recorded in two ecosystems (You Yangs Regional Park, YY; Cheetham Wetlands, CW). Listed in order of most to least widespread (% of sites a species was recorded in the ecosystem). Their activity index before weed management (B), 1 – 18 days after management (A1), 42 – 60 days after management (A2), and total activity are listed. The most common species for each ecosystem are in bold and were used in further analysis. Introduced species are indicated (*). Common and scientific names follow Christidis and Boles (2008). Species Scientific name YY CW % BA1A2Total% BA1A2Total Brown thornbill Acanthiza pusilla 78 20 17 13 50 Yellow thornbill Acanthiza nana 53 7141940 Superb fairy-wren Malurus cyaneus 50 16 17 30 63 11 031 4 Scarlet robin Petroica boodang 43 10 10 6 26 Grey shrike-thrush Colluricincla harmonica 43 68620 Yellow-faced honeyeater Lichenostomus chrysops 40 3152240 Grey fantail Rhipidura albiscapa 40 86620 Eastern yellow robin Eopsaltria australis 39 3 7 17 27 New Holland honeyeater Phylidonyris novaehollandiae 33 15 14 26 55 Spotted pardalote Pardalotus punctatus 30 0 5 15 20 Red wattlebird Anthochaera carunculata 26 4121935 Silvereye Zosterops lateralis 25 5147 26 6 100 1 White-naped honeyeater Melithreptus lunatus 23 0 3 18 21 Red-browed finch Neochmia temporalis 21 48618 Table 5.4 Continued Species Scientific name YY CW % BA1A2Total % BA1A2Total Willie wagtail Rhipidura leucophrys 19 23 5 10 39 51 4 10 White-throated treecreeper Cormobates leucophaea 20 23 1 6 Golden whistler Pachycephala pectoralis 16 01 3 4 Buff-rumped thornbill Acanthiza reguloides 15 82 2 12 White-plumed honeyeater Lichenostomus penicillatus 14 3101124 6 01 0 1 Australian magpie Cracticus tibicen 14 14 2 7 Rufous whistler Pachycephala rufiventris 9 60 0 6 Black-faced cuckoo-shrike Coracina novaehollandiae 8 02 0 2 Eastern spinebill Acanthorhynchus tenuirostris 8 01 1 2 Little raven Corvus mellori 8 02 0 2 Striated pardalote Pardalotus striatus 8 20 0 2 White-winged chough Corcorax melanorhamphos 5 40 0 4 Brown-headed honeyeater Melithreptus brevirostris 5 40 1 5 Common blackbird* Turdus merula 4 00 1 1 11 00 2 2 Flame robin Petroica phoenicea 4 00 1 1 6 21 0 3 Black-chinned honeyeater Melithreptus gularis 4 01 0 1 Crested shrike-tit Falcunculus frontatus 4 11 0 2 Common bronzewing Phaps chalcoptera 4 10 0 1 Laughing kookaburra Dacelo novaeguineae 4 01 0 1 Table 5.4 Continued Species Scientific name YY CW % BA1A2Total% BA1A2Total Satin flycatcher Myiagra cyanoleuca 4 100 1 Striated thornbill Acanthiza lineata 4 001 1 Sulphur-crested cockatoo Cacatua galerita 4 100 1 Weebill Smicrornis brevirostris 4 100 1 Varied sittella Daphoenositta chrysoptera 3 56011 Galah Eolophus roseicapillus 1 003 3 Yellow-rumped thornbill Acanthiza chrysorrhoa 1 402 6 Singing honeyeater Lichenostomus virescens 83 15 21 21 57 Striated fieldwren Calamanthus fuliginosus 44 63615 White-fronted chat Epthianura albifrons 44 4 6 13 23 White-browed scrub wren Sericornis frontalis 22 66719 Australian raven Cracticus tibicen 17 004 4 Common starling* Sturnus vulgaris 11 011 2 European goldfinch* Carduelis carduelis 11 39012 House sparrow* Passer domesticus 11 201 3 Australasian pipit Anthus novaeseelandiae 6 101 2 Crested pigeon Ocyphaps lophotes 6 002 2 Red-capped plover Charadrius ruficapillus 6 010 1 Chapter 5 – Wildlife friendly weed management

Table 5.5 Pair-wise comparisons of bird assemblages in impact, control and reference sites compared before (B), 1 – 18 days after (A1) or 42 – 60 days after (A2) weed management in the You Yangs Regional Park (YY) and Cheetham Wetlands (CW). Pair-wise R statistics are based on multivariate ANOSIM, where values close to zero indicate similarity between groups, and closer to one indicate separation of groups. Ecosystem Treatment Time Pair-wise R statistic YY Impact B vs A1 0.089 B vs A2 0.217 A1 vs A2 -0.044 Control B vs A1 -0.084 B vs A2 0.025 A1 vs A2 -0.079 Reference (shrubs) B vs A1 -0.050 B vs A2 0.062 A1 vs A2 0.010 Reference (open) B vs A1 -0.098 B vs A2 -0.030 A1 vs A2 -0.114 CW Impact B vs A1 -0.071 B vs A2 0.032 A1 vs A2 -0.030 Control B vs A1 -0.088 B vs A2 -0.106 A1 vs A2 -0.044 Reference B vs A1 -0.160 B vs A2 -0.128 A1 vs A2 -0.109 Chapter 5 – Wildlife friendly weed management

5.4.2 Cheetham Wetlands 5.4.2.1 Vegetation In CW, three components were identified that accounted for 60% of the variation in cover of different vegetation life-forms. The first component had high positive values associated with more sedges/rushes and bryophytes but negative values associated with shrub cover (Table 5.1). This described a gradient of open vegetation with ground cover typical of a mesic environment, to vegetation that becomes shrubby; hereafter marshy. The second component had higher positive values associated with lichen and shrub cover and negative values associated with herbs and tussocks (Table 5.1). This described a gradient of shrubby vegetation to more open areas with various herbaceous ground covers; hereafter herbaceous. The final component described a gradient of vegetation where coarse woody debris and herb cover (positive values) gave way to tussocks and bare ground (negative values), hereafter debris (Table 5.1).

Two components were extracted accounting for 66% of the variation in the measurements of vertical vegetation density. The first described the density of emergent shrubs with high positive values 81 – 170 cm above ground. The second component described lower shrubby density, with high positive values 21 – 70 cm above ground (Table 5.2).

Model selection results indicated that boxthorn management successfully reduced weed cover while retaining much of the vegetation structure. Models for weed cover and debris that included a treatment × time interaction term had considerable support (Table 5.3). Weed cover was much lower in impact sites at both times after weed management compared with before management, but weed cover barely changed with time in reference or control sites, as expected (Figure 5.1). In impact sites, debris values were higher at both times after weed management compared with before management, but

123 Chapter 5 – Wildlife friendly weed management remained similar in reference and control sites across sampling times (Figure 5.1). There was model uncertainty for all other CW habitat variables.

5.4.2.2 Birds Fifteen bird species were recorded (Table 5.4). The singing honeyeater was the only common species. Models for bird richness and activity that included a treatment × time interaction term were not supported (Table 5.3). The model for singing honeyeater activity with only a treatment effect had considerable support. Investigation of means and standard errors indicated that activity of singing honeyeaters was consistently lowest in reference sites.

Bird assemblages did not differ with treatment (global R = -0.111) or time (global R = - 0.066), and pair-wise comparisons between treatment-time combinations provided little evidence of an effect of weed management on impact sites (Table 5.5).

5.4.2.3 Small mammals Again, the house mouse and black rat were the only small mammal species captured (2913 trap nights). Black rats were not analysed further due to the low number of captures (< 5%). Model selection indicated the model for house mouse capture rates that included the treatment × time interaction term was not supported, and that there was uncertainty between other models (Table 5.3).

5.5 Discussion

For wildlife species using weeds, a successful reduction in weeds could be analogous to habitat loss. As the loss of habitat is the largest threat to global biodiversity (Fahrig 2001), using sensitive weed management could reduce the risk of what effectively could be further habitat loss for wildlife. This research indicated that, in the short-term, weed management strategies that retained weed structure had little impact on wildlife that were dependent on such structure.

124 Chapter 5 – Wildlife friendly weed management

5.5.1 Impacts of weed management on wildlife The wildlife studied in these ecosystems did not respond to the weed management that effectively reduced weed cover. The lack of a response among wildlife is likely due to 1) retention of influential vegetation attributes such as structure, 2) the scale at which weed management was implemented and 3) the time-frame over which responses were monitored.

In the YY, spraying boneseed with herbicide meant that much of the structure provided by boneseed was retained; in the short-term, the foliage wilted and died, but was not completely lost (pers. obs.). Consistent tall shrubs and low shrubs density after management in all sites confirmed a lack of substantial structural changes after boneseed management. Boneseed management did not impact canopy either, the most influential habitat variable for brown thornbill and superb fairy-wrens (Chapter 3).

Unlike the YY, some native birds in CW were strongly influenced by weed cover (Chapter 4). The singing honeyeater favoured vegetation in CW where boxthorn was present (Chapter 4), yet was not negatively impacted by its management at the scale and in the manner implemented here. The consistency of singing honeyeater activity in impact sites, despite boxthorn management, suggests that retaining structure may have been useful in honeyeater persistence after management. This could reflect the management technique used, which involved leaving the cut boxthorn in situ specifically to provide structure. The increase in the debris gradient and consistency in shrubby and emergent density gradients in impact sites, confirmed that management was effective in increasing availability of cut boxthorn and maintaining vegetation structure after management.

Maintaining structural complexity of weed sites did not only support native birds; introduced rodents were supported also. Weeds are often suggested to provide for invasive animal species (Randall 1996) and weed management is expected to reduce such pests (for example, CRC for Australian Weed Management 2007); however, the 125 Chapter 5 – Wildlife friendly weed management invasive rodent, house mouse, did not respond to weed management in either CW or YY. Because this species was most influenced by lower structural complexity in these ecosystems (Chapters 3 and 4), conducting management in a way that retained the structural integrity of the vegetation after management also apparently assisted in accommodating this introduced species. The response of house mice to environmental change often differs to that of native small mammals in Australia (for example, Holland and Bennett 2007; Kelly et al. 2010). Therefore, it is difficult to relate the response to the retention of weed structure in the house mouse, to that of native small mammal species existing in similar weedy environments.

The fact that wildlife-friendly weed management may allow pest species to persist, could conflict with other biodiversity conservation objectives. However, in the case of the house mouse, capture rates in managed and non-managed weeds were no different to those in native reference sites with similar structure (Chapters 3 and 4). So, the species is likely to persist in native vegetation regardless of the type of management applied to weeds.

While the maintenance of habitat structure was important in maintaining wildlife after weed management, additional factors are likely to be involved. In CW in particular, the scale of boxthorn management that was conducted may be important in maintaining singing honeyeater activity. In native habitats, it is widely recognised that the probability of species survival drops abruptly after a threshold amount of habitat has been lost (Fahrig 2001). It is unlikely that such a threshold had been reached with the scale of management conducted in the wetland. The areas in which boxthorn was managed were small and, within each of these, only the central boxthorn was managed.

Conducting boxthorn management at this small scale potentially allowed species like the singing honeyeater to use living boxthorn plants that remained unmanaged and existed throughout impact sites. Management of boxthorn often involves widespread mechanical removal (Nicholls et al. 2007). This type of management covers much larger areas than 126 Chapter 5 – Wildlife friendly weed management that which was covered in this study, and could result in less availability of resources provided by non-managed boxthorn. It is possible that the small scale of management, as well as the structure provided by cut boxthorn, allowed the singing honeyeater to remain unaffected by management of boxthorn in this instance. Caution needs to be observed, however, when considering the application of these techniques in different systems or at larger spatial scales. If thresholds of weed management exist, then these are likely to be species dependant, as with habitat loss (Fahrig 2001).

The non-response of wildlife was consistent over the relatively short period sampled in both ecosystems. Some species have a longer delay before displaying a response to habitat loss (Fahrig 2001), thus, with more time, a negative response to management might be detected. Structural resources provided by dead weeds would be expected to last only a certain amount of time before they decay. Therefore, small-scale management without appropriate revegetation with natives may eventually be detrimental to wildlife that depended on these weed resources. The growth of native species to replace weed structure is essential. On the other hand, the time-frame of this study may not have detected recolonisation of managed sites by native animals; such recolonisation may take months or years (Nichols and Nichols 2003; Lindenmayer et al. 2009).

Even continued management at the same scale as in this current study, without further adjustments to management, could eventually negatively affect some species, particularly those that rely on weeds. For example, removing boxthorn plants in CW over a long period of time, even a small number at a time, could still have a substantial cumulative negative impact on the singing honeyeater. On its own, maintaining weed structure is only likely to sustain wildlife for a certain amount of time. In a wildlife- friendly weed management process, incorporation of revegetation is critical. Yet, allowing for revegetation to grow so that resources are replaced takes time (Nelson and Wydoski 2008; Carlos and Gibson 2010). Allowing for this time lag while conducting

127 Chapter 5 – Wildlife friendly weed management small scale wildlife-friendly management might, therefore, require greater time commitments to weed management than already required.

In different ecosystems, weed management has had diverse effects on wildlife. In some circumstances, weed reduction after management has resulted in negative impacts on birds, but only in particular seasons (Esler 1990). Elsewhere, the weed management technique itself has directly reduced lizard abundance (Valentine and Schwarzkopf 2008), while management had species-specific impacts in a coastal ecosystem invaded by bitou bush Chrysanthemoides monilifera subsp. rotundata (DC.) Norl. when broad- scale herbicide application reduced bitou bush fruit availability (Gosper 2004). In contrast, herbicide management of bitou bush did not impact ground-dwelling arthropods in the short-term (French and Buckley 2008). The high degree of situational- specificity in regard to the effect of weed management on wildlife means it is difficult to predict likely outcomes for wildlife in most areas where weed management is contemplated. Therefore the incorporation of wildlife assessments prior to management, and their inclusion in the monitoring process, is necessary to plan management approaches that are wildlife-friendly and to assess how such management will affect wildlife.

128 Chapter 6

Synthesis

Singing honeyeater Lichenostomus virescens nest in African boxthorn Lycium ferocissimum, Victoria, Australia.

129 Chapter 6 – Synthesis

6.1 Overview

This thesis considered the influences of weeds and their management on wildlife, with a view towards helping inform weed management in Australia. Under some circumstances, weeds can provide valuable resources to wildlife, particularly when weeds act as an ecological substitute for native vegetation (Graves and Shapiro 2003), provide for threatened wildlife (Sogge et al. 2008) or connect fragmented habitats (Date et al. 1991). Yet, the impacts of weed management on wildlife are rarely studied (Valentine et al. 2007).

Using social and ecological research methods, the interactions between weeds and wildlife and their management implications were examined. In particular, the often unacknowledged potential values of weeds for wildlife were considered. Here, the results of this thesis are summarised and implications of this work are discussed.

6.2 Major findings

6.2.1 Consideration of wildlife by weed managers Given the wide variety of interactions between weeds and wildlife, some of which could be considered ‘beneficial’ and some ‘detrimental’ to wildlife, management of weeds becomes more complex as questions regarding the impact of their removal are less clear cut. This is especially the case in situations where weed-wildlife interactions may be beneficial for wildlife. Chapter 2 considered the objectives, perceptions and practices of weed managers in relation to weed-wildlife interactions and, in particular, their potential importance as wildlife habitat. Given the wide variation evident in the literature in the ways weeds impact wildlife, weed manager’s observations and opinions regarding the potential value of weeds for wildlife were expected to vary similarly. This was true for the personal observations of managers – managers reported a variety of wildlife taxa using different weeds; but they had more or less consistent beliefs regarding the importance of weeds as wildlife habitat. Understandably, weeds were considered most important for those taxa which were observed using weeds most frequently. Managers 130 Chapter 6 – Synthesis reported that they adjusted weed management so that such species were not disadvantaged by weed removal. The more experienced managers observed wildlife using weeds more often and were more likely to have adjusted weed management. This indicated an ‘experience-behaviour’ link, which is a fundamental aspect of adaptive management; the most useful, and reportedly the most used, approach for management of invasive species (Shea et al. 2002).

The basis of the adaptive management approach is acceptance of a high level of uncertainty about the system that is being managed (Wilhere 2002). Both the literature and the managers surveyed here, confirm that there is much uncertainty regarding wildlife in weed management and that this is initially due to the varying responses of wildlife to weeds. The most basic adaptive management process normally addresses such uncertainty by following a planning, implementation and monitoring/evaluation cycle. Where, after initial implementation, performance is evaluated with adjustments made to plans for further management; however, this research revealed that wildlife was rarely monitored as part of weed management programs (Chapter 2), thus evaluation of how effective management was in terms of wildlife is relatively unknown. The lack of wildlife monitoring is a critical gap in the apparently commonly-used adaptive weed management process. This is especially the case considering 71% of weed managers indicated that they conducted management in a way that attempted to account for wildlife use of weeds.

The recommended response to positive wildlife interactions with weeds includes combined gradual removal of weeds and revegetation (Gosper and Vivian-Smith 2006; Carlos and Gibson 2010). In spite of this, only a low proportion of managers who had made adjustments to management to account for wildlife actually used this particular combination. Thus, clarifying the effectiveness of other approaches that weed managers use is important. Without inclusion of wildlife monitoring in weed management that aims to conserve biodiversity and improve wildlife habitat, the adaptive management

131 Chapter 6 – Synthesis process is incomplete. Therefore, the current weed management practices are in danger of not only being ineffectual, but risky for wildlife.

Further, Chapter 2 revealed that wildlife taxa that were most frequently observed using weeds were those that were typically conspicuous and active (for example birds). While the frequency of observation of such taxa may represent the actual amount these taxa use weeds, it also is likely to reflect the detectability of wildlife. Thus, managers may not be aware of wildlife using weeds that are less detectable or inconspicuous. Ideally, wildlife would be systematically assessed before management, to identify important weed- wildlife interactions and the attributes of weeds that are most important to these relationships, so appropriate management can be planned.

6.2.2 The influence of weeds on wildlife Chapters 3 and 4 assessed wildlife in vegetation dominated by weeds and native vegetation to understand whether weeds provided wildlife with habitat and, also, the aspects of these habitats that most influenced wildlife. In Chapter 3, bird richness in the woodland ecosystem of the You Yangs did not differ between vegetation dominated by boneseed Chrysanthemoides monilifera subsp. monilifera (DC.) T. Norl. and that dominated by native species. Similarly, in the wetland ecosystem of Cheetham Wetlands, bird richness did not differ between sites where the weed boxthorn Lycium ferocissimum Miers. was present, compared to where it was absent (Chapter 4).

Comparable richness of birds in weed and native vegetation indicated that weeds in the two ecosystems provided habitat for birds. Yet, in the wetland, bird composition differed between weed and native vegetation (Chapter 4) while, in the woodland, differences in bird composition were inconsistent between weed and native vegetation; woodland with a native shrub understorey had similar bird assemblages to that with a boneseed understorey, yet these assemblages were different to woodland that lacked a shrubby understorey (Chapter 3). These results suggest that weeds were the preferred habitat of some bird species and not others.

132 Chapter 6 – Synthesis

The differences in bird assemblages between the weed and native vegetation in each ecosystem reflected differences in vegetation structure in these ecosystems. The potential wildlife response to structural aspects of vegetation in the woodland and wetland was investigated in Chapters 3 and 4 by modelling the response of different bird metrics using structural characteristics of the vegetation and weed cover as predictors. Multi-model inference suggested that bird richness was not driven by any of these predictors in either the woodland or wetland, but the level of activity of some particularly common species was influenced by particular vegetation predictors. Different characteristics of woodland vegetation structure were important in explaining the activity of New Holland honeyeaters Phylidonyris novaehollandiae and superb fairy- wrens Malurus cyaneus (Chapter 3), while, in the wetland, weed cover influenced the activity of the singing honeyeater Lichenostomus virescens and white-fronted chat Epthianura albifrons (Chapter 4). With weed cover correlated with, and, thus, indistinguishable, from the density of tall (emergent) vegetation, the model selection process supported the notion that vegetation structure was driving the level of activity of some bird species in both ecosystems.

The only small mammal captured sufficiently often to allow for statistical analyses was the house mouse Mus domesticus (Chapters 3 and 4). As a generalist rodent pest, it did not respond specifically to weeds in either ecosystem, but only to vegetation structure close to the ground. This explained the similarities in capture rates between weed and native sites in the wetland and why, in the woodland, there were lower capture rates in sites that had an open understorey. Given that the response of house mice to environmental change is often very different to that of native small mammals in Australia (Holland and Bennett 2007; Kelly et al. 2010), it is difficult to use this species as an indicator of the response of native small mammals to weeds elsewhere. However, this research does suggest that weeds of a particular structure can provide refuge for this pest rodent in areas where there is little other structural complexity.

133 Chapter 6 – Synthesis

These results suggested that native bird species and pest rodents could use weeds as habitat because of the structure they provide. In the wetland, there was also some evidence that fruit and floral resources provided by weeds may play a role in bird responses (Chapter 4); changes in bird activity occasionally corresponded to changes in fruit/floral resources between native and weed vegetation. Elsewhere, birds have responded strongly to the nectar and/or fruit availability of weeds (French et al. 2005). In the wetland, birds were probably responding to a combination of changes to foraging resources as well as structure offered by weeds.

Overall, this research demonstrated that vegetation structure was important in driving activity levels of some species in each ecosystem. When weed structure was similar to that of native vegetation, weeds supported similar species. When the structure was different, weeds provided habitat for different species. Removal of weeds therefore has the potential to impact the species using them. When planning management of these weeds, such wildlife species are important to consider. The importance of vegetation structure, revealed in this study, can further be used to inform weed management techniques that might retain wildlife species using the weeds.

6.2.3 Weed management impacts on wildlife Chapter 5 examined whether weeds which were managed so that their structural characteristics were retained sustained the wildlife species that used them. The nature of the dominant weeds in the two ecosystems that were studied meant that different management techniques were required to account for weed structure. In the woodland, herbicide was applied to boneseed throughout impact sites, which killed plants but allowed their branches and stems to remain intact. In comparison, boxthorn in the wetland was cut and painted (with herbicide) and cut branches were piled up in situ, in an attempt to account for the tall nature of the plant when alive. These weed management methods were effective in substantially reducing the cover of weeds, while retaining much of the original structure of the vegetation (Chapter 5).

134 Chapter 6 – Synthesis

Birds and small mammals were monitored before and after weed management was applied but showed no change. This suggested that weed management had no impact on the species studied, despite their use of the weeds as habitat. This included the singing honeyeater, which had displayed a positive response to high weed cover before it was managed (Chapter 4). While this ‘wildlife-friendly’ technique was not compared with other techniques that do not attempt to account for wildlife, this research provided evidence that retaining weed structure during small-scale management could assist in retaining wildlife, at least in the short-term, in these ecosystems.

Impacts of weed management on wildlife in the longer term, and on a larger scale, remain unknown. It is reasonable to assume that managed weeds will decay and their structural integrity will decline over time. Therefore, this sort of management strategy might be best conducted in conjunction with a revegetation strategy so that structure is being replaced by native plants as managed weeds decay. This recommendation, while not new (Gosper and Vivian-Smith 2006; Carlos and Gibson 2010), is apparently infrequently adopted by managers (Chapter 2).

The assessment of wildlife before and after management that was conducted in this research was useful in indicating what species used weeds and how, and informed a weed management strategy that, in the short-term, successfully sustained these species. It also provided information for future improvements and adjustments. Using a similar approach that assesses wildlife before and after weed management may therefore be useful in the adaptive weed management process. With the high level of awareness of weed-wildlife interactions, positive attitudes towards the importance of weeds as potential habitat and current levels of implementation of management strategies designed to be ‘wildlife-friendly’, incorporating such assessments and monitoring of wildlife throughout the whole process would seem a simple and logical step. However, when considering wildlife in weed management there are a range of other factors that add further complexity to an already complex process.

135 Chapter 6 – Synthesis

6.3 Management implications

Findings from this research and the literature have demonstrated that effectively accounting for wildlife requirements throughout an adaptive weed management process must consider many other factors. In Figure 6.1 the most basic adaptive management cycle is shown with additional factors that would be useful considerations to enhance the inclusion of wildlife in weed management identified in each step.

Initially the broad overall objectives of weed management will dictate whether wildlife need to be considered during weed management. Many aspects will influence these objectives, such as regulatory requirements, but objectives are generally either economically, socially or environmentally directed (Chapter 2). Economic or social objectives for weed management are unlikely to include wildlife. It is when objectives are environmental and targeted towards biodiversity conservation, that consideration of wildlife is important. This research (Chapter 2) and that of others (Reid et al. 2009) indicated that the latter goal is frequent, so, assessment of wildlife as part of weed management is relevant. Clearly articulating whether wildlife conservation is an objective in weed management is imperative.

Setting specific wildlife objectives can clearly direct planning, implementation and monitoring for adaptive management. This, first, will require considering the wildlife that are using weeds. It is likely that there will always be advantaged and disadvantaged species in weed invasions and their management. For instance, in the wetland ecosystem of this study, white-fronted chats were a species that avoided weed habitat, yet singing honeyeaters responded positively to weeds (Chapter 4). The type of wildlife species that are (or are not) using the weeds and the legislative requirements associated with them also need to be considered. This will require wildlife assessments in weeds and nearby natives, not only because of the current lack of information, but because of the site and species specificity of weed-wildlife interactions.

136 Chapter 6 – Synthesis

Considering wildlife in the cycle of planning, implementation and evaluation of weed management will include considerations such as time-frames, availability of funding and the level of expertise available. These factors often are cited as constraints faced by weed managers (Chapter 2; Reid et al. 2009). Along with these, there are other little understood questions that require consideration, such as the most appropriate plant species for revegetation or the most meaningful aspects of wildlife to measure.

There are a number of areas throughout this process that would benefit from further research. Figure 6.1 shows many of these are in the ‘specific objectives’ and ‘monitoring and evaluation’ phases, where direct wildlife assessments are required. The centralisation of data collected by such assessments would mean managers could draw on the experience of others to facilitate their learning and, potentially, may underpin the generation of general principles regarding which weeds and which management are associated with desirable outcomes for wildlife. This also would help to reduce the influence of factors, such as availability of time and funds, which are often cited as barriers to monitoring (Reid et al. 2009).

137

Institutional Weed definition Stakeholders OVERALL OBJECTIVES requirements Priority weeds Economic/social Regulations/ SPECIFIC legislation OBJECTIVES OVERALL OBJECTIVES Environmental Feasibility PLANNING AND PREDICTION

Time-frame Are wildlife Biodiversity/ Invasive species included? conservation control

MONITORING AND IMPLEMENTATION EVALUATION Type of population response Novel/substitute Threatened wildlife Regulations/ Type of habitat SPECIFIC Priority legislation OBJECTIVES weeds provide Pest wildlife

Multiple species

Figure 6.1 Basic adaptive management Wildlife definition Wildlife assessment What wildlife are using weeds? cycle with factors that require consideration with the inclusion of wildlife in weed Time Advantaged/ management, as identified in the literature disadvantaged wildlife and through this research. Shaded boxes Continue to Gradient of indicate areas that require research. ‘planning and prediction’ responses

OVERALL OBJECTIVES Continue to Broader strategies ‘specific objectives’ Possible Population changes? outcomes SPECIFIC Wildlife OBJECTIVES friendly? Timing Follow up Type of evaluation PLANNING AND PREDICTION Environmental Habitat type factors PLANNING AND PREDICTION Available Funding MONITORING AND resources IMPLEMENTATION What is Indicators EVALUATION meaningful? Equipment, personnel Priority wildlife What to species measure? Centralisation Significance of weed of data MONITORING AND IMPLEMENTATION Weed species Statistical power EVALUATION

Re-infestation

Data analysis Control Methods Barriers Revegetation Wildlife friendly? Effect size Expertise Time-frame Toxicity Time-frame Short-term/ Funding Expertise Plant species long-term Direct disturbance to wildlife Time Whatisa (for example nesting) Wildlife resources Figure 6.1 Continued response? to replace

Chapter 6 – Synthesis

6.4 Conclusion

Throughout the world, weeds invade ecosystems and, in some circumstances, weeds offer alternative or new resources to wildlife. In this research, some native birds, as well as an introduced rodent, were influenced by the structural resources weeds provided. For some species, the weed structure was appropriate and provided them with habitat. Others did not benefit and were potentially disadvantaged by weeds. The decision regarding whether or not to adjust weed management to accommodate the species using weeds depends on the objectives of weed management and the priority placed on the wildlife in question. Wildlife-friendly weed management approaches take into account the factors relating to weeds that influence wildlife and use this to inform their management. This research has shown that such a strategy could be effective in retaining species that use weeds.

140 References

Sun-kissed dawn sky at the You Yangs Regional Park, Australia

141 References

Adair RJ, Groves RH (1998). Impact of environmental weeds on biodiversity: A review and development of methodology. Environment Australia, Canberra.

Ainsworth N, Adair R, Cheal D (2008). A method of monitoring biodiversity for changes associated with invasive plants. Department of Sustainability and Environment, Melbourne.

Andreu J, Vila M, Hulme P (2009). An assessment of stakeholder perceptions and management of noxious alien plants in Spain. Environmental Management 43: 1244 – 1255.

Anonymous (1927). Birds and prickly pear. The Emu 26: 203 – 206.

Antos MJ, Ehmke GC, Tzaros CL, Weston MA (2007). Unauthorised human use of an urban coastal wetland sanctuary: Current and future patterns. Landscape and Urban Planning 80(1 – 2): 173 – 183.

Arnold GW, Steven DE, Weeldenburg JR, Smith EA (1993). Influences of remnant size, spacing pattern and connectivity on population-boundaries and demography in euros Macropus-robustus living in a fragmented landscape Biological Conservation 64(3): 219 – 230.

Bailey J, Schweitzer J, Whitham T (2001). Salt cedar negatively affects biodiversity of aquatic macroinvertebrates. Wetlands 21(3): 442 – 447.

Bardsley D, Edwards-Jones G (2006). Stakeholders’ perceptions of the impacts of invasive exotic plant species in the Mediterranean region. GeoJournal 65: 199 – 210.

Barrett GW, Freudenberger D, Drew A, Stol J, Nicholls AO, Cawsey EM (2008). Colonisation of native tree and shrub plantings by woodland birds in an agricultural landscape. Wildlife Research 35(1): 19 – 32.

142 References

Bennet A (1993). Microhabitat use by the long-nosed potoroo, Potorous tridactylus, and other small mammals in remnant forest vegetation, south-western Victoria. Wildlife Research 20(3): 267 – 285.

Bibby CJ, Burgess ND, Hill DA (1992). Bird census techniques. Academic Press, London.

Boakes EH, Mace GM, McGowan PJK, Fuller RA (2009). Extreme contagion in global habitat clearance. Proceedings of the Royal Society B: Biological Sciences 277: 1081 – 1085.

Bremner A, Park K (2007). Public attitudes to the management of invasive non- native species in Scotland. Biological Conservation 139: 306 – 314.

Brett Lane and Associates (2009). Cheetham Wetlands operation manual- volume 1: Ecological values. Brett Lane and Associates, North Carlton.

Brougham KJ, Cherry H, Downey PO (2006). Boneseed management manual: Current management and control options for boneseed (Chrysanthemoides monilifera ssp. monilifera) in Australia. Department of Environment and Conservation NSW, Sydney.

Brown PR, Wallis RL, Simmons D, Adams R (1991). Weeds and wildlife. Plant Protection Quarterly 6(3): 150 – 153.

Brunner H, Harris RV, Amor RL (1976). A note on the dispersal of seeds of blackberry (Rubus procerus P. J. Muell.) by foxes and emus. Weed Research 16: 171 – 173.

Buckland ST, Anderson DR, Burnham KP, Laake JL, Borchers DL, Thomas L (2001). Introduction to distance sampling: Estimating abundance of biological populations. Oxford University Press, Melbourne.

Burnham KP, Anderson DR (2002). Model selection and multimodel inference: A practical information-theoretic approach. Second edn. Springer-Verlag, New York.

143 References

Carlos EH, Gibson M (2010). The habitat value of gorse Ulex europaeus L. and hawthorn Crataegus monogyna Jacq. for birds in Quarry Hills Bushland Park, Victoria. The Victorian Naturalist 127(4): 115 – 123.

Christidis L, Boles WE (2008). Systematics and taxonomy of Australian birds. CSIRO Publishing, Collingwood.

Christopherson K, Morrison ML (2004). Influence of yellow starthistle (Centaurea solstitialis) on small mammals in central California. Western North American Naturalist 64(2): 202 – 207.

Clarke KR (1993). Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18: 117 – 143.

Clarke KR, Gorley RN (2006). PRIMER v6: User manual/tutorial. PRIMER-E, Plymouth.

Commissioner for Environmental Sustainability (2008). State of the environment report- Victoria 2008. Commissioner for Environmental Sustainability, Melbourne.

Cook D (1996). Vegetation survey of Cheetham Saltworks. Holistic Ecology.

Coulon A, Cosson JF, Angibault JM, Cargnelutti B, Galan M, Morellet N, Petit E, Aulagnier S, Hewison AJM (2004). Landscape connectivity influences gene flow in a roe deer population inhabiting a fragmented landscape: An individual–based approach. Molecular Ecology 13(9): 2841 – 2850.

CRC for Australian Weed Management (2007). Weed management guide managing weeds for biodiversity: African boxthorn (Lycium ferocissimum).

Cully Nordby J, Cohen AN, Beissinger SR (2009). Effects of a habitat-altering invader on nesting sparrows: An ecological trap? Biological Invasions 11: 565 – 575.

Curtis A, Lockwood M (2000). Landcare and catchment management in Australia: Lessons for state-sponsored community participation. Society and Natural Resources 13(1): 61 – 73.

144 References

D'Antonio C, Meyerson LA (2002). Exotic plant species as problems and solutions in ecological restoration: A synthesis. Restoration Ecology 10(4): 703 – 713.

Date EM, Ford HA, Recher HF (1991). Frugivorous pigeons, stepping stones, and weeds in northern New South Wales. In: Saunders DA, Hobbs RJ (eds) Nature conservation 2: The role of corridors. Surrey Beatty and Sons, Armidale.

Date EM, Recher HF, Ford HA, Stewart DA (1996). The conservation and ecology of rainforest pigeons in northeastern New South Wales. Pacific Conservation Biology 2: 299 – 308.

Davies RJP (1998). Regeneration of blackberry-infested native vegetation. Plant Protection Quarterly 13(4): 189 – 195.

Downey PO (2011). Changing of the guard: Moving from a war on weeds to an outcome-orientated weed management system. Plant Protection Quarterly 26(3): 86 – 91.

Dufty A (1994). Population demography of the eastern barred bandicoot (Perameles gunnii) at Hamilton, Victoria. Wildlife Research 21: 44 – 57.

Ekert PA, Bucher DJ (1999). Winter use of large-leafed privet Ligustrum lucidum (Family: Oleaceae) by birds in suburban Lismore, New South Wales. Proceedings of the Linnean Society New South Wales 121: 29 – 38.

Environmental Weeds Working Group (2007). Guidelines and procedures for managing the environmental impacts of weeds on public land in Victoria 2007. Department of Sustainability and Environment, Melbourne.

Esler D (1990). Avian community responses to hydrilla invasion. Wilson Bulletin 102(3): 427 – 440.

Ewel JJ, Putz FE (2004). A place for alien species in ecosystem restoration. Frontiers in Ecology and the Environment 2(7): 354 – 360.

Fahrig L (2001). How much habitat is enough? Biological Conservation 100(1): 65 – 74.

Fisher AM, Goldney DC (1997). Use by birds of riparian vegetation in an extensively fragmented landscape. Pacific Conservation Biology 3: 275 – 288.

145 References

French K, Buckley S (2008). The effects of the herbicide metsulfuron-methyl on litter invertebrate communities in a coastal dune invaded by Chrysanthemoides monilifera spp. rotundata. Weed Research 48: 266 – 272.

French K, Major R, Hely K (2005). Use of native and exotic garden plants by suburban nectarivorous birds. Biological Conservation 121: 545 – 559.

French K, Zubovic A (1997). Effect of the weed Chrysanthemoides monilifera (bitou bush) on bird communities. Wildlife Research 24(6): 727 – 735.

Fulton DC, Manfredo MJ, Lipscomb J (1996). Wildlife value orientations: A conceptual and measurement approach. Human Dimensions of Wildlife 1(2): 24 – 47.

Fulton G, Ford H (2001). The pied currawong's role in avian nest predation: A predator removal experiment. Pacific Conservation Biology 7: 154 – 160.

Funk JL, Vitousek PM (2007). Resource-use efficiency and plant invasion in low- resource systems. Nature 446(7139): 1079 – 1081.

Garcia-Llorente M, Martin-Lopez B, Gonzalez J, Alcorlo P, Montes C (2008). Social perceptions of the impacts and benefits of invasive alien species: Implications for management. Biological Conservation 141: 2969 – 2983.

Garden JG, McAlpine CA, Possingham HP, Jones DN (2007). Habitat structure is more important than vegetation composition for local-level management of native terrestrial reptile and small mammal species living in urban remnants: A case study from Brisbane, Australia. Austral Ecology 32(6): 669 – 685.

Gibson M (2010). Weeds: A brief introduction. The Victorian Naturalist 27(4): 96 – 103.

Goodenough A (2010). Are the ecological impacts of alien species misrepresented? A review of the ‘native good, alien bad’ philosophy. Community Ecology 11(1): 13 – 21.

Gosper CR (2004). Consequences of weed invasion and control on plant-bird interactions and bird communities. PhD thesis, University of Woolongong.

146 References

Gosper CR, Vivian-Smith G (2006). Selecting replacements for invasive plants to support frugivores in highly modified sites: A case study focusing on Lantana camara. Ecological Management and Restoration 7(3): 197 – 203.

Gosper CR, Vivian-Smith G (2009). Approaches to selecting native plant replacements for fleshy-fruited invasive species. Restoration Ecology 17(2): 196 – 204.

Gosper CR, Whelan RJ, French K (2006). The effect of invasive plant management on the rate of removal of vertebrate-dispersed fruits. Plant Ecology 184(2): 351 – 363.

Graves SD, Shapiro AM (2003). Exotics as host plants of the California butterfly fauna. Biological Conservation 110(3): 413 – 433.

Grice AC, Martin TG (2006). Rangelands, weeds and biodiversity. Rangeland Journal 28(1): 1 – 2.

Griffin G, Stafford S, Morton G, Allan G, Masters K (1989). Status and implications of the invasion of Tamarisk (Tamarix aphylla) on the Finke River, Northern Territory, Australia. Journal of Environmental Management 29: 297 – 315.

Haag KH, Joyce JC, Hetrick WM, Jordan JC (1987). Predation on water hyacinth weevils and other aquatic insects by three wetland birds in Florida. The Florida Entomologist 70(4): 457 – 471.

Heckathorn DD (2002). Respondent-driven sampling II: Deriving valid population estimates from chain-referral samples of hidden populations Social Problems 49 (1): 11 – 34.

Hobbs RJ, Humphries SE (1995). An integrated approach to the ecology and management of plant invasions. Conservation Biology 9(4): 761 – 770.

Holland GJ, Bennett AF (2007). Occurrence of small mammals in a fragmented landscape: the role of vegetation heterogeneity. Wildlife Research 34: 387 – 397.

147 References

Homer PM, Kahle LR (1988). A structural equation test of the value-attitude- behaviour hierarchy. Journal of Personality and Social Psychology 54: 638 – 646.

Hulme PE (2006). Beyond control: wider implications for the management of biological invasions. Journal of Applied Ecology 43(5): 835 – 847.

Jellinek S, Driscoll D, Kirkpatrick J (2004). Environmental and vegetation variables have a greater influence than habitat fragmentation in structuring lizard communities in remnant urban bushland. Austral Ecology 29: 294 – 304.

Kelly LT, Nimmo DG, Spence-Bailey LM, Clarke MF, Bennett AF (2010). The short-term responses of small mammals to wildfire in semiarid mallee shrubland, Australia. Wildlife Research 37(4): 293 – 300.

King C (2007). Tackling weeds on private land initiative: Project evaluation report- August 2007. Department of Primary Industries, Melbourne.

Lake PS (2001). On the maturing of restoration: Linking ecological research and restoration. Ecological Management & Restoration 2(2): 110 – 115.

Lawrie S (2002). An audit system to account for birds and mammals utilising weeds species. In: Spafford J, Dodd J, Moore JJ (eds) 13th Australian weeds conference, September 8 – 13 2002. Weeds, threats now and forever? Plant Protection Society of Western Australia, pp 108 – 111.

Lesica P, Miles S (2004). Beavers indirectly enhance the growth of Russian olive and tamarisk along eastern Montana Rivers. Western North American Naturalist 64(1): 93 – 100.

Lindenmayer DB, MacGregor C, Wood JT, Cunningham RB, Crane M, Michael D, Montague-Drake R, Brown D, Fortescue M, Dexter N, Hudson M, Gill AM (2009). What factors influence rapid post-fire site re-occupancy? A case study of the endangered eastern bristlebird in eastern Australia. International Journal of Wildland Fire 18(1): 84 – 95.

148 References

Lindsay EA, French K (2006). The impact of the weed Chrysanthemoides monilifera ssp. rotundata on coastal leaf litter invertebrates. Biological Invasions 8: 177 – 192.

Lindsay EA, French K (2004). The impact of the herbicide glyphosate on leaf litter invertebrates within bitou bush, Chrysanthemoides monilifera ssp. rotundata, infestations. Pest Management Science 60(12): 1205 – 1212.

Linz GM, Bergman DL, Blixt DC, McMurl C (1997). Response of American coots and soras to herbicide-induced vegetation changes in wetlands. Journal of Field Ornithology 68(3): 450 – 457.

Linz GM, Bleier WJ, Overland JD, Homan HJ (1999). Response of invertebrates to glyphosate induced habitat alterations in wetlands. Wetlands 19: 220 – 227.

Linz GM, Blixt DC, Bergman DL, Bleier WJ (1996). Responses of red-winged blackbirds, yellow-headed blackbirds and marsh wrens to glyphosate-induced alterations in cattail density. Journal of Field Ornithology 67(1): 167 – 176.

Loyn R, French K (1991). Birds and environmental weeds in south eastern Australia. Plant Protection Quarterly 6(3): 137 – 148.

Luck GW, Possingham HP, Paton DC (1999). Bird responses at inherent and induced edges in the Murray Mallee, South Australia. 1. Differences in abundance and diversity. Emu 99(3): 157 – 169.

Magnussen C (undated). Weed management guide- African boxthorn. Queensland Department of Employment, Economic Development and Innovation, Licensed from the Commonwealth of Australia under a Creative Commons Attribution 3.0 Australia Licence.

Major R (1991). Breeding biology of the white-fronted chat Ephthianura albifrons in a saltmarsh near Melbourne. Emu 91: 236 – 249.

Manning M, Munro D (2006). The survey researcher’s SPSS cookbook. Pearson Education Australia, Frenchs Forest.

149 References

Maron M, Lill A (2005). The influence of livestock grazing and weed invasion on habitat use by birds in grassy woodland remnants. Biological Conservation 124(4): 439 – 450.

Marshall JM, Storer A, Leutscher B (2008). Comparative analysis of plant and ground dwelling arthropod communities in lacustrine dune areas with and without Centaurea biebersteinii (Asteraceae). The American Midland Naturalist 159(2): 261 – 274.

Martín-López B, Montes C, Benayas J (2007). The non-economic motives behind the willingness to pay for biodiversity conservation. Biological Conservation 139(1 – 2): 67 – 82.

McNaught I, Thackway R, Brown L, Parsons M (2006). A field manual for surveying and mapping nationally significant weeds. Bureau of Rural Sciences, Canberra.

McNeely J, Weatherly P (1996). Innovative funding to support biodiversity conservation. International Journal of Social Economics 23: 98 – 124.

Monamy V, Fox BJ (2005). Differential habitat use by a local population of subadult common dunnarts, Sminthopsis murina, following wildfire in coastal wet heath, New South Wales, Australia. Wildlife Research 32(7): 617 – 624.

Mutze G (1991). Mouse plagues in South Australia cereal-growing areas III. Changes in mouse abundance during plague and non-plague years, and the role of refugia. Wildlife Research 18(5): 593 – 603.

Muyt A (2001). Bush invaders of south-east Australia: A guide to the identification and control of environmental weeds found in south-east Australia. R.G. and F. J. Richardson, Meredith.

Natural Resource Management Ministerial Council (2006). Australian Weeds Strategy – A national strategy for weed management in Australia. Australian Government Department of the Environment and Water Resources, Canberra.

Neira C, Levin L, Grosholz E, Mendoza G (2007). Influence of Spartina growth stages on associated macrofaunal communities. Biological Invasions 9: 975 – 993.

150 References

Nelson SM, Wydoski R (2008). Riparian butterfly (Papilionoidea and Hesperioidea) assemblages associated with Tamarix-dominated, native vegetation-dominated and Tamarix removal sites along the Arkansas river, Colorado, U.S.A. Restoration Ecology 16(1): 168 – 179.

Nias R (1986). Nest site characteristics and reproductive success in the superb fairy- wren. Emu 86: 139 – 144.

Nicholls K, Lovett S, Price P (2007). Managing weeds in riparian areas. Land and Wool Australia, Canberra.

Nichols OG, Nichols FM (2003). Long-term trends in faunal recolonization after bauxite mining in the jarrah forest of southwestern Australia. Restoration Ecology 11(3): 261 – 272.

Ortega YK, Pearson DE, McKelvey KS (2004). Effects of biological control agents and exotic plant invasion on deer mouse populations. Ecological Applications 14(1): 241 – 253.

Pearson DE, Fletcher RJ (2008). Mitigating exotic impacts: restoring deer mouse populations elevated by an exotic food subsidy. Ecological Applications 18(2): 321 – 334.

Peter JM (2000). Birds and boxthorn. The Victorian Naturalist 117(2): 63 – 66.

Pimentel D, Zuniga R, Morrison D (2005). Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecological Economics 25: 273 – 288.

Prescott T (1995). The You Yangs Range. Yaugher Print, Belmont.

Priddel D, Carlile N (1995). Mortality of adult Gould's petrels Pterodroma leucoptera leucoptera at the nesting site on Cabbage Tree Island, New South Wales. Emu 95: 259 – 264.

Quinn G, Keough M (2002). Experimental design and data analysis for biologists. Cambridge University Press, Melbourne.

R Development Core Team (2012). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna.

151 References

Randall JM (1996). Weed control for the preservation of biological diversity. Weed Technology 10: 370 – 383.

Recher HF, Holmes RT, Schulz M, Shields J, Kavanagh R (1985). Foraging patterns of breeding birds in eucalypt forest and woodland of southeastern Australia. Australian Journal of Ecology 10: 399 – 419.

Regan D, Fazio R (1977). On the consistency between attitudes and behaviour: Look to the method of attitude formation. Journal of Experimental Social Psychology 13: 28 – 45.

Reid A, Morin L, Downey P, French K, Virtue J (2009). Does invasive plant management aid the restoration of natural ecosystems? Biological Conservation 142: 2342 – 2349.

Remes V (2003). Effects of exotic habitat on nesting success, territory density, and settlement patterns in the blackcap (Sylvia atricapilla). Conservation Biology 17(4): 1127 – 1133.

Richardson DM, Pyšek P, Rejmánek M, Barbour MG, Panetta FD, West CJ (2000). Naturalization and invasion of alien plants: concepts and definitions. Diversity and Distributions 6(2): 93 – 107.

Roberts D (2008). A history of boneseed control in the You Yangs Regional Park, Victoria. Plant Protection Quarterly 23(1): 51.

Rodewald A, Shustack D, Hitchcock L (2010). Exotic shrubs as ephemeral ecological traps for nesting birds. Biological Invasions 12(1): 33 – 39.

Rodriguez LF (2006). Can invasive species facilitate native species? Evidence of how, when, and why these impacts occur. Biological Invasions 8(4): 927 – 939.

Salganik M, Heckathorn D (2004). Sampling and estimation in hidden populations using respondent-driven sampling. Sociological Methodology 34: 193 – 239.

Sanderson KJ, Kraehenbuehl J (2006). Southern brown bandicoots Isoodon obesulus obesulus in Belair National Park. Australian Mammalogy 28: 147 – 152.

Sax DF, Kinlan BP, Smith KF (2005). A conceptual framework for comparing species assemblages in native and exotic habitats. Oikos 108(3): 457 – 464.

152 References

Schlaepfer M, Sax D, Olden J (2011). The potential conservation value of non-native species. Conservation Biology 25(3): 428 – 437.

Schlaepfer MA, Runge MC, Sherman PW (2002). Ecological and evolutionary traps. Trends in Ecology and Evolution 17(10): 474 – 480.

Schmidt B, Renowden C, Quin D (2009). Southern brown bandicoot strategic management plan for the former Koo Wee Rup swamp area. Ecology Australia, Fairfield.

Severns PM, Warren AD (2008). Selectively eliminating and conserving exotic plants to save an endangered butterfly from local extinction. Animal Conservation 11(6): 476 – 483.

Shafroth PB, Cleverly JR, Dudley TL, Taylor JP, Van Riper C, Weeks EP, Stuart JN (2005). Control of Tamarix in the western United States: Implications for water salvage, wildlife use, and riparian restoration. Environmental Management 35(3): 231 – 246.

Shea K, Possingham HP, Murchoch W, Roush R (2002). Active adaptive management in insect pest and weed control: intervention with a plan for learning. Ecological Applications 12(3): 927 – 936.

Sindel B (2000). Australian weed management systems. R.G. and R.F. Richardson, Melbourne.

Sinden J, Jones R, Hester R, Odom D, Kalisch C, James R, Cacho O (2004). The economic impact of weeds in Australia. CRC for Australian weed management, Adelaide.

Smedley K, Nance D (2010). Port Phillip Bay (western shoreline) Ramsar site pest plant and animal management plan. Practical Ecology, Preston.

Sogge MK, Sferra SJ, Paxton EH (2008). Tamarix as habitat for birds: Implications for riparian restoration in the southwestern United States. Restoration Ecology 16(1): 146 – 154.

Spennemann DHR, Allen LR (2000). The avian dispersal of olives Olea europaea: implications for Australia. Emu 100: 264 – 273.

153 References

Stansbury CD (2001). Dispersal of the environmental weed bridal creeper, Asparagus asparagoides, by silvereyes, Zosterops lateralis, in south-western Australia. Emu 101(1): 39 – 45.

Stralberg D, Toniolo V, Page GW, Stenzel LE (2010). Potential impacts of Spartina spread on shorebird populations in south San Francisco Bay. In: Ayres DR, Kerr DW, Ericson SD, Olofson PR (eds) Proceedings of the third international conference on invasive Spartina, 2004 Nov 8 – 10, San Francisco, CA, USA. San Francisco Estuary Invasive Spartina Project of the California State Coastal Conservancy, pp 175 – 184.

Sutter GC, Troupe T, Forbes M (1995). Abundance of Baird’s sparrows, Ammodramus bairdii, in native prairie and introduced vegetation. Ecoscience 2(4): 344 – 348.

Tasker EM, Dickman CR (2002). A review of Elliot trapping methods for small mammals in Australia. Australian Mammalogy 23: 77 – 87.

Trollope ST, White JG, Cooke R (2009). The response of ground and bark foraging insectivorous birds across an urban-forest gradient. Landscape and Urban Planning 93(2): 142 – 150.

Tuomainen U, Candolin U (2011). Behavioural responses to human-induced environmental change. Biological Reviews 86(3): 640 – 657.

Valentine L, Roberts B, Schwarzkopf L (2007). Mechanisms driving avoidance of non-native plants by lizards. Journal of Applied Ecology 44: 228 – 237.

Valentine L, Schwarzkopf L (2008). Effects of weed-management burning on reptile assemblages in Australian tropical savannas. Conservation Biology 23(1): 103 – 113. van der Meulen AW, Reeve IJ, Sindel BM (2007). Weed management on grazing properties: a survey of livestock producers in New South Wales and Victoria. Australian Journal of Experimental Agriculture 47: 1415 – 1421.

154 References

Vitousek PM, D'Antonio C, Loope LL, Rejmanek M, Westbrooks R (1997). Introduced species: a significant component of human-caused global change. New Zealand Journal of Ecology 21(1): 1 – 16.

Walker H, Brammer S (2009). Sustainable procurement in the United Kingdom public sector. Supply Chain Management: An International Journal 14(2): 128 – 137.

Weston M, Fendley M, Jewell R, Satchell M, Tzaros C (2003). Volunteers in bird conservation: Insights from the Australian Threatened Bird Network. Ecological Management and Restoration 4(3): 205 – 211.

White JG, Antos MJ, Fitzsimons JA, Palmer GC (2005). Non-uniform bird assemblages in urban environments: the influence of streetscape vegetation. Landscape and Urban Planning 71(2 – 4): 123 – 135.

Wilhere GF (2002). Adaptive management in habitat conservation plans. Conservation Biology 16(1): 20 – 29.

Williams JA, West CJ (2000). Environmental weeds in Australia and New Zealand: issues and approaches to management. Austral Ecology 25(5): 425 – 444.

Wilson D, Paton DC (2004). Habitat use by the southern emu-wren, Stipiturus malachurus (Aves: Maluridae) in South Australia, and evaluation of vegetation at a potential translocation site for S. m. intermedius. Emu 104(1): 37 – 43.

Wolff JO, Schauber EM, Daniel Edge W (1998). Effects of habitat loss and fragmentation on the behaviour and demography of gray-tailed voles. Conservation Biology 11(4): 945 – 956.

Zanette L (2000). Fragment size and the demography of an area-sensitive songbird. Journal of Animal Ecology 69: 458 – 470.

Zanette L, Doyle P, Tremont SM (2000). Food shortage in small fragments: evidence from an area-sensitive passerine. Ecology 81(6): 1654 – 1666.

Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM (2009). Mixed effects models and extensions in ecology with R. Springer Science Business Media, New York.

155