Distribution and Structure of Arthropod Communities in Relation to Resource Patches and Spatial Scale in Dryland Woodland Ecosystems

Home , Agra (beetle), Camponotini, Duplex (moth), Mazia, Metallica (beetle), Rhytidoponera

Distribution and structure of arthropod communities in relation to resource patches and spatial scale in dryland woodland ecosystems

Alan B. C. Kwok

Evolution and Ecology Research Centre

School of Biological, Earth and Environmental Sciences University of New South Wales

Sydney, NSW, 2052

Australia

PhD thesis

March 2012

THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet Surname or Family name: Kwok

First name: Alan Other name/s: Bing Choong

Abbreviation for degree as given in the University calendar: PhD

School: Biological, Earth and Environmental Faculty: Science Sciences

Title: Distribution and structure of arthropod communities in relation to resource patches and spatial scale in dryland woodland ecosystems

Abstract 350 words maximum: (PLEASE TYPE) In dryland ecosystems, resources such as water, nutrients and habitat are concentrated into discrete patches. This resource concentration occurs at fine (e.g. around trees, grasses or logs) and broad (e.g. habitat remnants within an agricultural matrix) scales. Arthropods, which include insects, spiders, and a range of other invertebrates, provide a range of critical ecosystem functions in drylands. Arthropods may be particularly sensitive to changes in resource concentration given their small size and habitat requirements. Limited research, however, has examined how arthropods respond to changes in resource concentration across different spatial scales. This thesis examines how the concentration of resources affects the distribution and structure of arthropod communities at multiple spatial scales in south-eastern Australia. Chapter 1 provides an overview of resource patchiness in arid and semi-arid ecosystems, and describes how it is known to affect the biota. Chapters 2 to 4 investigate how the fine-scale distribution of resources (plants, and plant-associated patches) affects the distribution and composition of arthropod communities at local (plant-plant) scales. Specifically, chapter 2 examines how plant species and density affect the plant-resident arthropod fauna in a semi-encroached shrubland, and demonstrates that plant species is the overwhelming driver of arthropod diversity in these communities. Chapters 3 and 4 examine how the multilayered resource patch created by a dominant tree (mallee, Eucalyptus spp.) structures arthropod communities in mallee woodlands. At fine-scales, the canopy patch beneath mallee trees supports a distinct suite of arthropods compared to inter-tree areas. This is influenced by fire, which has taxon-specific effects. Chapters 5 and 6 investigate how the concentration and health of remnant patches at landscape scales affects arthropods. Chapter 5 evaluates the use of common landscape health indices as indicators of arthropod (spider, ant, and beetle) biodiversity, illustrating that these indices show only weak or inconsistent relationships with arthropod biodiversity. Chapter 6 uses a multi-scale approach to investigate the drivers of ant community structure in a fragmented grassy–box woodland. It indicates that finer-scale characteristics (particularly tree canopy cover and soil texture) drive ant communities within these landscapes. Chapter 7 summarises the findings and implications of the thesis, and proposes avenues for future research.

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

………………………………………………………… ……………………………………..…………… ……….…………………… … … ...…….… Signature Witness Date The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

Originality Statement

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project's design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed......

Date......

Copyright Statement ‘I hereby grant the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstract International (this is applicable to doctoral theses only). I have either used no substantial portions of copyright material in my thesis or I have obtained permission to use copyright material; where permission has not been granted I have applied/will apply for a partial restriction of the digital copy of my thesis or dissertation.'

Signed ……………………………………………...... Date ……………………………………………......

Authenticity Statement ‘I certify that the Library deposit digital copy is a direct equivalent of the final officially approved version of my thesis. No emendation of content has occurred and if there are any minor variations in formatting, they are the result of the conversion to digital format.’

Signed ……………………………………………......

Date ……………………………………………......

"If we and the rest of the backboned animals were to disappear overnight, the rest of the world would get on pretty well. But if invertebrates were to disappear, the land's ecosystems would collapse. The soil would lose its fertility. Many of the plants would no longer be pollinated. Lots of animals - amphibians, reptiles, birds, mammals - would have nothing to eat. These small creatures are within a few inches of our feet wherever we go on land, but often they are disregarded. We would do very well to remember them."

~ Sir David Attenborough ~

PREFACE

This dissertation consists of five stand-alone manuscripts (Chapters 2 to 6) that have been published, recently submitted for publication, or are intended for publication in peer-reviewed journals. Each chapter is self-contained and subsequently, there will be some repetition. A single reference list has been provided at the end of the dissertation to avoid unnecessary duplication.

This thesis is a compilation of my own work, with guidance from my supervisor

David Eldridge. All chapters were conceptualised either by myself or jointly with

David Eldridge. I conducted all data analyses and wrote and illustrated the manuscripts. Specific details for each chapter and the contribution of co-authors are detailed below.

Chapter 2: A. B. C Kwok and D. J. Eldridge. Shrub species, not spatial arrangement, structures arthropod communities in a shrub-encroached woodland

Study was conceptualised by Alan Kwok. David Eldridge provided guidance on the study design and structure of the manuscript in his role as academic supervisor. Submitted for publication in Insect Conservation and Diversity.

Chapter 3: A. B. C Kwok and D. J. Eldridge. Do trees modulate ground-dwelling arthropod communities in the mallee of south-eastern Australia?

Study was conceptualised by Alan Kwok. David Eldridge provided guidance on study design and the structure of the manuscript in his role as academic supervisor.

Chapter 4: A. B. C Kwok and D. J. Eldridge. Does fire affect the ground- dwelling arthropod community through changes to fine-scale resource patches?

Study was conceptualised by Alan Kwok. David Eldridge provided guidance on the study design and structure of the manuscript in his role as academic supervisor.

Chapter 5: A. B. C Kwok, D. J. Eldridge and I. Oliver. Do landscape health indices reflect arthropod biodiversity status in the eucalypt woodlands of eastern Australia? Austral Ecology, 36, 800 - 813.

Manuscript was conceptualised by Alan Kwok. David Eldridge assisted with parts of the statistical analyses and Ian Oliver established the project upon which the data were derived and contributed intellectually to the development of the manuscript.

Chapter 6: A. B. C Kwok, D. J. Eldridge and D. Freudenberger. Drivers of ant community structure in a grassy eucalypt woodland ecosystem in south-eastern Australia

Manuscript was conceptualised by Alan Kwok. David Eldridge provided guidance on the structure of the manuscript in his role as academic supervisor. David Freudenberger established the project upon which the data were derived and contributed intellectually to the development of the manuscript.

ACKNOWLEDGEMENTS

Firstly, a big thank you to David Eldridge for being a great supervisor, and for all the guidance in the past years. It has seemed like a long few years at times (don't forget Honours!), but we made it. Thanks for always having an 'open door', and for all the non-PhD chit chat. I think I'd have been much more stressed over the years if not for your generally jovial nature.

A massive thank you to Samantha Travers, who has been my think tank, field slave, stress ball, and "pick me up" for most of the PhD. I can't describe how much your overwhelmingly positive outlook and ceaseless encouragement helped me through. I may forget most things, but I won't forget all you've done for me!

Thanks to my co-authors - Ian Oliver and David Freudenberger for allowing me to delve into the depths of their invertebrate data and for helping in developing the manuscripts. I'm glad to be able to make use of all the effort that was put into collecting the data those years ago.

Also, my appreciation goes to all those who helped me beat these chapters into shape and/or provided feedback, including Yvonne Davila, Alison Foster, Aaron Greenville (including stats advice!), David Tongway and Samantha Travers. Special mention to Terry Koen, for being a biometrician who actually makes sense, and who helped me understand stats. To my field slaves - Yvonne Davila, George Madani, and Samantha Travers - your efforts are very much appreciated.

I would also like to express my gratitude to the owners of the properties upon which this work was conducted. In particular, all those at Australian Wildlife Conservancy's Scotia Sanctuary have been wonderful. Special thanks especially to Joe Stephens, Felicity L'Hotellier, Jennifer Cathcart, Tony Cathcart, Matt Hayward and all the others who helped along the way out at

iii

Scotia Sanctuary and made us feel so welcome. Flick, we are indebted to you for all your assistance and for always being so willing to help us out - I can never express how grateful I am for all you have done! Thank you also to Martin Westbrooke (Ballarat University) for information on Nanya and fire in mallee.

Of course, my family and friends also deserve my thanks. To my parents, who have always supported and encouraged me, thank you. To my brother, Steven, you may not have realised it, but you always helped to take my mind off it all and gave me sound advice when needed. I couldn't ask for more from my brother. To Nick Colman, for the countless chit chats over the years - they are always priceless.

A heartfelt 'thank you' to my close friends over the past few years who have put up with my rants, encouraged my cynicism (sometimes unknowingly) and who have always been willing to help out - Miguel Bedoya, Nick Colman, Yvonne Davila, Alison Foster, Aaron Greenville, George Madani, Samantha Travers, Marion Winkler, and Trevor Wilson. Cheers also to the SoBS Frisbee crew - for the weekly laughs, exercise, and much-needed distraction!

Thanks to all the BEES postgrads I’ve spent time with over the years – in particular Marie Attard, Alex James, Erin Roger, Dean Portelli, Celene Steinfeld, and Gilad Bino.

Finally, to Samantha Travers - again - because honestly, you are a major reason I'm writing this now. I can never thank you enough.... But I'll keep trying!

ABSTRACT

In dryland ecosystems, resources such as water, nutrients and habitat are concentrated into discrete patches. This resource concentration occurs at fine

(e.g. around trees, grasses or logs) and broad (e.g. habitat remnants within an agricultural matrix) scales. Arthropods, which include insects, spiders, and a range of other invertebrates, provide a range of critical ecosystem functions in drylands. Arthropods may be particularly sensitive to changes in resource concentration given their small size and habitat requirements. Limited research, however, has examined how arthropods respond to changes in resource concentration across different spatial scales. This thesis examines how the concentration of resources affects the distribution and structure of arthropod communities at multiple spatial scales in south-eastern Australia. Chapter 1 provides an overview of resource patchiness in arid and semi-arid ecosystems, and describes how it is known to affect the biota. Chapters 2 to 4 investigate how the fine-scale distribution of resources (plants, and plant-associated patches) affects the distribution and composition of arthropod communities at local (plant-plant) scales. Specifically, chapter 2 examines how plant species and density affect the plant-resident arthropod fauna in a semi-encroached shrubland, and demonstrates that plant species is the overwhelming driver of arthropod diversity in these communities. Chapters 3 and 4 examine how the multilayered resource patch created by a dominant tree (mallee, Eucalyptus spp.) structures arthropod communities in mallee woodlands. At fine scales, the canopy patch beneath mallee trees supports a distinct suite of arthropods compared to inter-tree areas. This is influenced by fire, which has taxon-specific effects. Chapters 5 and 6 investigate how the concentration and health of

remnant patches at landscape scales affects arthropods. Chapter 5 evaluates the use of common landscape health indices as indicators of arthropod (spider, ant, and beetle) biodiversity, illustrating that these indices show only weak or inconsistent relationships with arthropod biodiversity. Chapter 6 uses a multi- scale approach to investigate the drivers of ant community structure in a fragmented grassy–box woodland. It indicates that finer-scale characteristics

(particularly tree canopy cover and soil texture) drive ant communities within these landscapes. Chapter 7 summarises the findings and implications of the thesis, and proposes avenues for future research

LIST OF TABLES

Table 2.1: Mean (± SE) abundance of total arthropods, Psocoptera, Collembola and Psyllidae in relation to shrub species and shrub density...... 28

Table 2.2: Summary statistics for generalized linear model analyses on abundance of arthropods taxa sampled from Eremophila and Senna (shrub species, SS) and in low and high shrub density (density, D)...... 30

Table 2.3: Adjusted-R2 for the linear regressions between estimated plant biomass and the number of arthropods for Eremophila sturtii and Senna artemisioides at low and high densities...... 33

Table 3.1: Summary statistics for generalised linear model analyses on arthropod abundance in five patch types...... 60

Table 3.2: t-test results for pairwise comparisons of arthropod community composition across four treatments...... 62

Table 5.1: Total abundance and total, median and range of species and higher taxon (Family) richness by site for arthropod groups...... 116

Table 5.2: Descriptive statistics of all health variables used...... 116

Table 5.3: Spearman’s rank correlations (ρ) for tests of assemblage similarity between arthropod taxa and LFA or SCF (RELATE), and for subsets of LFA or SCF which show the strongest relationship to the arthropod taxa (BIO-ENV)...... 120

Table 5.4: BIO ENV results using habitat variables and LFA indices...... 123 vii

Table 6.1: Descriptive statistics for the continuous predictor variables used in this study...... 152

Table 6.3: DistLM results for species and functional groups composition using landscape context and within-patch habitat attributes...... 157

Table 6.4: Relative contribution of landscape context and within-patch variables to models explaining variation in ant species assemblage as calculated from marginal DistLM tests and multi-model inference...... 158

viii

LIST OF APPENDICES

Appendix 5.1: Soil surface features used to calculate soil surface condition.

...... 133

LIST OF FIGURES

Figure 2.1: Photographs of target shrub species: a) Senna artemisioides and b)

Eremophila sturtii...... 22

Figure 2.2: Photographs of density treatments, showing typical plant growing in a) low density and b) high density...... 24

Figure 2.3: Species accumulation curves for Senna artemisioides (dashed line) and Eremophila sturtii (solid line), using number of observed hemipteran species...... 27

Figure 2.4: Mean ± SE a) number of Hemiptera, b) number of Hemiptera species and c) effective number of Hemiptera species (sensu species diversity) averaged over two shrub subsamples per site for two shrub species (Senna artemisioides and Eremophila sturtii) across two shrub densities (low and high)...... 31

Figure 2.5: Non-metric multi-dimensional scaling ordination for hemipteran species composition in relation to a) shrubs growing in low density (open circles) and high density (closed diamonds), and b) Eremophila sturtii (open circles) and Senna artemisioides (closed circles)...... 31

Figure 2.6: Rarefaction curves for the number of Hemiptera species on Senna artemisioides (closed circles) and Eremophila sturtii (open circles)...... 33

Figure 3.1: Photos of typical study site within mallee community, showing a) resource patch under a tree canopy, b) woody debris on ground surface under canopy, c) inter-tree areas dominated by bare ground with scattered Triodia hummocks, and d) bare surface of open patch...... 53

Figure 3.2: Photos of artificially created patch types, showing a) litter only, b) litter + shade, and c) shade only patches...... 54

Figure 4.1: Photos of typical study site mallee community, showing a) Unburnt resource patch under canopy, b) unburnt open patch, c) burnt patch under canopy and d) burnt open patch...... 84

Figure 4.2: Mean ± S.E. abundance of a) ants, b) beetles, c) silverfish, d) cockroaches, e) isopods, f) spiders, g) jumping spiders and h) wasps across two patch types (canopy and open) and in two communities that differ in fire history (burnt and unburnt)...... 89

Figure 4.4: Mean ± S.E. number of spider morphospecies across two patch types (canopy and open) and in two communities that differ in fire history (burnt and unburnt)...... 93

Figure 5.2: Proportion of variance explained (Y-axis) by independent (white) and joint (black) components of three terrestrial indices of ecological integrity (structure–composition–function, SCF) (X-axis) indices on a) ants, b) beetles, and c) spiders as determined by hierarchical partitioning...... 119

Figure 6.2: Regression tree for soil clay content (%) and the three most abundant ant species: a) Iridomyrmex purpureus; b) I. septentrionalis; and c) Rhytidoponera metallica...... 160

xii

TABLE OF CONTENTS

Preface...... i Acknowledgements...... iii Abstract...... v List of tables...... vii List of appendices...... ix List of figures...... x

Chapter 1. General introduction...... 1

Thesis objectives...... 13

Chapter 2. Shrub species, not spatial arrangement, structures arthropod communities in a shrub-encroached woodland...... 16

Abstract...... 17 Introduction...... 18 Methods...... 21 Results...... 26 Discussion...... 34 Acknowledgements...... 42

Chapter 3. Do trees modulate ground-dwelling arthropod communities in the mallee of south-eastern Australia?...... 43

Abstract...... 44 Introduction...... 45 Methods...... 50 Results...... 57 Discussion...... 63 Acknowledgements...... 73

Chapter 4. Does fire affect the ground-dwelling arthropod community through changes to fine-scale resource patches?...... 74

Abstract...... 75 Introduction...... 76

Methods...... 81 Results...... 86 Discussion...... 94 Acknowledgements...... 101

Chapter 5. Do landscape health indices reflect arthropod biodiversity status in the eucalypt woodlands of eastern Australia?...... 102

Abstract...... 103 Introduction...... 104 Methods...... 108 Results...... 115 Discussion...... 124 Acknowledgements...... 132

Chapter 6. Drivers of ant community structure in a grassy eucalypt woodland ecosystem in south-eastern Australia...... 134

Abstract...... 135 Introduction...... 136 Methods...... 139 Results...... 150 Discussion...... 161 Acknowledgements...... 171

Chapter 7. General discussion...... 172

Key findings of this thesis...... 173 References...... 187

1. General Introduction

Chapter 1

General introduction

1. General Introduction

Resource limitation and heterogeneity in arid and semi-arid environments

Productivity in arid and semi-arid environments is constrained by the limited availability of water and nutrients in both time and space. Rainfall is generally low, with rainfall events that may be highly irregular throughout the year and spatially variable across a scale of tens to hundreds of kilometres. Additionally, levels of soil nutrients are relatively low, particularly of nitrogen and phosphorus.

Patterns of rainfall and nutrient availability drive the productivity pulses of the system and have a strong influence on all ecosystem processes as well as the biota (Noy Meir, 1973; Ludwig et al., 1997; Morton et al., 2011).

In arid and semi-arid environments, resources (water, nutrients, sediments, and habitat) are not distributed evenly in the landscape. Instead, they are concentrated in patches. A patch, in its simplest sense, is an area at any spatial scale that differs from its surroundings in its abiotic and biotic structure and composition (Pickett and Cadenasso, 1995). In arid ecosystems, the relatively resource-rich area under a tree or shrub is the most recognisable patch at fine spatial scales (1-10m2), as it contrasts with the surrounding, general matrix that is generally devoid of vegetation or occupied by annual grasses. Patches, however, occur at multiple spatial scales, ranging from patch–hillslope to catchment scales (Ludwig et al., 2004).

Resources are concentrated into patches at broad and fine spatial scales

At broader, landscape scales, resources are often concentrated into particular areas. While the definition of landscape is scale- and organism-dependent, when we refer to landscape scales this is typically from an anthropogenic 2

1. General Introduction

perspective (102-104 hectares, Lindenmayer and Fischer, 2006). The concentration of resources at these scales is often a natural phenomenon. For example, clay soils tend to be more nutrient-rich than sandy soils, retaining more water for longer periods of time. Similarly, a runon zone at the bottom of a hillslope captures water and sediments that are lost from the hill itself. The runon zone therefore becomes an area of higher resource concentration (e.g. bands of mulga vegetation, Tongway and Ludwig, 1990).

At finer spatial scales, resource patches usually form around the base of perennial vegetation or semi-permanent ground obstructions such as log mounds (Tongway et al., 1989; Ridolfi et al., 2008). These 'fertile islands' or

'islands of fertility' are a consequence of biotic and abiotic processes

(Schlesinger et al., 1990). Growing vegetation can increase water infiltration, concentrate soil nutrients, and reduce microclimatic extremes. Furthermore, water, sediments and organic matter accumulate around these obstructions as they are redistributed by wind and water processes, reinforcing the fertility of the patch (Aguiar and Sala, 1999). Fertile patches contrast markedly with the relatively bare, resource-poor matrix of soil away from perennial vegetation

(Kotliar and Wiens, 1990). Overall, the patches under the canopy have greater soil fertility, greater habitat heterogeneity and animal activity, and are more productive than the matrix (Tongway and Ludwig, 1997).

1. General Introduction

Resource patches are not static in time

The properties of any patch are a reflection of processes that are occurring at finer and broader scales than the patch in question (Kotliar and Wiens, 1990).

For example, the condition of a grove of trees is dependent upon the condition of individual trees within the grove (finer scale) as well as the hydrology of the matrix surrounding the grove (broader scale). The structure and condition of resource patches is not static in time, and changes at either finer or broader scales of patchiness affect the properties of the grove itself. How patchiness is defined is therefore scale-dependent, but it is also governed by processes occurring at multiple scales.

Anthropogenic practices play an important role in affecting landscape-scale resource concentration. Contiguous patches of native vegetation are subject to substantial modification by human practices, notably agricultural and urban expansion (Lindenmayer and Fischer, 2006). Such expansion alters the concentration of resources in the landscape, in particular habitat for fauna, through a series of predictable modifications to the landscape and the vegetation it supports (e.g. Forman, 1995; McIntyre and Hobbs, 1999). These processes include perforation, dissection, and fragmentation of the landscape, resulting in shrinkage and attrition of natural landscape patches (Forman,

1995). Generally, the landscape changes from being intact to variegated, and subsequently to fragmented with only remnant (relictual) patches remaining

(McIntyre and Hobbs, 1999). Over time the remaining habitat resources become concentrated into smaller, more isolated patches of vegetation (Lindenmayer and Fischer, 2006).

1. General Introduction

Resource concentration at finer spatial scales can be altered by natural and anthropogenic processes. Habitat fragmentation often degrades fine scale patchiness by altering disturbance regimes (e.g. grazing, Fischer et al., 2005; light, wind and nutrient regimes, Harper et al., 2005). Broad-scale processes such as fire also affect the distribution of resources at fine scales. Fire can destroy above–ground vegetation, homogenising the distribution of plant resources which affects broader scale resource function. Similarly, obstructions such as log mounds that are important for maintaining resource patchiness are destroyed by fire. These patch elements are therefore lost from the system.

Assessing ecosystem health by examining changes in patch type, quality and structure

A pervasive consequence of landscape modification (e.g. habitat fragmentation) is often the degradation of the ecological health of remnant patches. Although health is a value-laden concept, in resource-limited environments, a widely- used definition of ecosystem health is based on the ability of the system to capture and retain critical resources (Tongway and Ludwig, 1997; Ludwig et al.,

2004). Patches of vegetation capture resources, preventing their flow out of the system (Ludwig et al., 2004). The importance of monitoring changes to these patches as well as ecological health is well acknowledged (Tongway and

Ludwig, 2011). Examining how ecological health varies between remnants can give insights into how the distribution of resources affects the abiotic and biotic structure and composition of patches at multiple spatial scales.

1. General Introduction

Resource concentration affects the distribution and structure of animal communities

Resource concentration affects the distribution of animals through similar mechanisms across multiple spatial scapes. Essentially, the resource contrast between the patch (i.e. a tree and its associated resources, or a patch of remnant vegetation) and the surrounding matrix (i.e. open areas, inter-tree areas, or agricultural paddocks) determines the species that can persist in the different patch types. Logically, animals are expected to favour and inhabit areas where critical resources are plentiful or more easily available, or where their physiological and ecological requirements are met (e.g. 'healthier' trees or larger woodland remnants). These conditions may not be met by all patch types. Furthermore, the composition of species assemblages is likely to be driven by the resources and niches available within each different patch type, and the contrast among the patches of the same type.

Landscape-scale resource concentration affects the distribution of animals through a variety of processes. The most visible process is that of habitat loss.

What constitutes as habitat is species-specific, but in general, landscape modification results in loss of native vegetation and/or habitat for a wide range of species. It is well established that habitat loss is often coupled with the decline of species (Fahrig, 2003). Habitat loss rarely acts in isolation, and is often associated with processes such as habitat subdivision and degradation

(sensu fragmentation, Lindenmayer and Fischer, 2006). Fragmentation can impair animal movement, resulting in the separation and isolation of populations

(Thomas et al., 2001; Fischer and Lindenmayer, 2007). Changes to landscape

1. General Introduction

scale resource concentration may also affect animals by changing the composition, structure and distribution of fine-scale resource patches.

At fine scales, the concentration of resources into discrete patches affects the distribution of a range of animals. Typically, this is the contrast between the resource patch of a tree (or shrub), and the surrounding inter-tree (or inter- shrub) areas. For example, in arid savannas, bird and mammal activity is often concentrated in trees or in the area under and immediately surrounding the tree, which provides shelter and foraging resources (e.g. Dean et al., 1999; Eccard et al., 2006; Blaum et al., 2007aa). Landscape elements such as trees, shrubs, and log mounds have been widely proposed to be foci of vertebrate and invertebrate activity (e.g. Tongway et al., 1989; Ludwig et al., 2004). Most studies illustrating the effects of resource concentration, however, focus on a single or small number of species. There is surprisingly little information on the extent to which the concentration of resources at fine-scales can structure whole animal communities.

Our understanding of how landscape-scale resource concentration affects faunal communities is based around the responses of large animals (e.g. mammals, birds) to landscape change. Habitat fragmentation and landscape change can result in substantial changes to the abundance and composition of vertebrate communities in anthropogenic landscapes (e.g. Chiarello, 1999;

Watson et al., 2004). Importantly, in the context of biodiversity management, many species are unable to persist in small, isolated habitat fragments (e.g. birds, Watson et al., 2001). This research has guided conservation biology in

1. General Introduction

the past two decades. Consequently biodiversity conservation is heavily focussed on vertebrate taxa, with relatively little research conducted on other faunal groups (Clark and May, 2002).

Arthropods are ubiquitous components of terrestrial ecosystems

Invertebrates are the most abundant animals on earth, comprising nearly 99% of all animal species. Despite this, invertebrates are poorly valued and largely ignored by the public and scientists (Kellert, 1993; Nash, 2004). Invertebrates include arthropods such as insects, spiders and crustaceans. In the year 2000 there were 855,000 described arthropod species (May, 2000). Estimates of the total number of described and undescribed arthropod species range dramatically, from five to 30 million (see Odegaard, 2000; Hamilton et al., 2010).

They can be found in every ecosystem in the world, often in remarkable numbers. For example, in tropical rainforests the number of ants can reach over

8 million per hectare, constituting up to a third of the animal biomass (Hölldobler and Wilson, 1990).

Arthropods play a critical role in many ecosystem processes. They are involved in cycling nutrients from dead plant and animal material (e.g. Santos and

Whitford, 1981; Whitford et al., 1982), modifying soil properties, and pollination and seed dispersal. Ants and termites are widely known to influence the spatial distribution of soil nutrients, infiltration, and consequently plant growth (e.g. Lee and Wood, 1971; Sneva, 1979; Hobbs, 1985; Wagner, 1997). Arthropods can also have significant impacts on whole plant communities. For example, arthropod herbivores and/or phytophages can affect plant reproduction and

1. General Introduction

community structure (Agrawal et al., 2006; Samocha and Sternberg, 2010). In addition, arthropods are a key food source for many reptiles, birds, and mammals.

Due to their ubiquity and relationship with ecosystem function, arthropods have been widely proposed as indicators of ecological disturbance and the ecological health of a system. In terrestrial ecosystems, arthropods are commonly tested as indicators of sustainable forestry management (e.g. Langor and Spence,

2006; Lindenmayer et al., 2000), mining rehabilitation (e.g. Bisevac and Majer,

1999; Orabi et al., 2010; Majer et al., 2007), and other ecological disturbances

(e.g. Churchill, 1998; Read and Andersen, 2000). While advances have been made in developing arthropod bio-indicators, particularly with ants (Andersen and Majer, 2004), further validation is necessary to determine the widespread applicability of arthropods across a range of ecosystems.

Arthropods are affected by processes operating at multiple spatial scales

Consideration of spatial scale is critically important for small animals, which can be affected by resource patchiness at a hierarchy of scales (Kotliar and Wiens,

1990). This is analogous to the importance of scale to understanding patterns of resource concentration. For example, an ant may be affected by fine-scale ecological patterns (such as the cover of leaf litter near their nest site), as well as broader-scale factors which change over tens or hundreds of kilometres such as climate, topography, and soil texture (Johnson, 1992). Attempting to understand factors driving arthropod distribution necessitates consideration of ecological processes and patterns of resource concentration operating at

1. General Introduction

multiple spatial scales, ranging from fine patch scales to broad, landscape scales.

How does resource concentration, which is scale-specific, affect arthropod communities?

Much of the focus on how arthropods are affected by resource concentration has centred upon the response of plant-resident arthropods (e.g. herbivores).

This is not surprising given that an estimated one million species of phytophagous insect are dependent upon plants as a food source (Jander and

Howe, 2008). Many studies in agricultural systems illustrate how plant density, diversity, and neighbour identity affect arthropod activity and diversity (e.g.

Pimentel, 1961; Root, 1973; Mayse, 1978; Agrawal et al., 2006). These studies highlight that the spatial pattern and concentration of resources affects the behaviour of plant-resident arthropods. This has tangible effects on their distribution and subsequent patterns of richness and diversity, with potential broader scale implications for ecosystem functions that are dependent upon arthropods.

Ground-dwelling arthropods are also influenced by the fine-scale distribution of resources, particularly resources provided by plants. For example, many studies have demonstrated how leaf litter distribution and complexity affect arthropod community structure (e.g. spiders, Pearce et al., 2004; Batary et al., 2008; beetles, Koivula et al., 1999; Acari, Hansen and Coleman, 1998). In arid ecosystems, the structure of arthropod communities can also be affected by other small-scale variation such as the cover of lichens, with arthropod size-

1. General Introduction

distribution predictable from lichen dimensions (Shorrocks et al., 1991; Lalley et al., 2006). Fertile patches created by vegetation are presumed to have similar effects on arthropod communities, however there is limited quantitative evidence for this (except, for example Oliver et al., 2006; Barton et al., 2009).

Invertebrates may show responses similar to vertebrates in relation to landscape change and broad-scale resource concentration. However research specifically on invertebrates has lagged behind that for larger vertebrate animals. While some studies indicate clear effects of landscape-scale remnant characteristics (e.g. patch size, shape, isolation) on arthropod richness, community composition, and even genetic diversity (Abensperg-Traun et al.,

1996; Bolger et al., 2000; Bickel et al., 2006; Orrock et al., 2011), others have illustrated stronger relative effects of finer scale habitat characteristics (e.g. within-patch vegetation condition, Debuse et al., 2007; Poyry et al., 2009).

Overall, however, there is limited information assessing how resource concentration at broad and/or fine scales affects arthropod communities.

The study of invertebrates requires a multi-taxon and multi-ecosystem approach

The study of arthropods requires approaches from multi-taxonomic scales.

While species-based research provides the most detailed understanding of arthropod ecology, it is often expensive, time consuming, and requires considerable taxonomic expertise. Taxonomic expertise is a particular problem given the shortage of specialists worldwide, and detailed taxonomic information is simply not known for the majority of arthropod taxa (New, 1995; Dangerfield,

1. General Introduction

1997). Studies based on higher taxonomic levels (e.g. functional group, order, class, etc.), provide valuable insights into arthropod ecology, albeit with the loss of some resolution (Dangerfield, 1997; Majer et al., 2006). These studies can suggest how ecological processes or drivers broadly affect a range of arthropods, not merely isolated species. Thus, our understanding of arthropod ecology is enhanced by the use of higher taxon and species-based approaches.

This is particularly important for the management of biodiversity, which cannot always afford a single-species approach.

Research across multiple ecosystems and communities is necessary when attempting to generalise broad drivers of arthropod distribution. While the spatial turnover of arthropod species is high, when considered from a higher taxonomic scale, their distribution is ubiquitous. This means that it is possible that similar drivers of diversity or abundance may exist across a variety of ecosystems, particularly across ecosystems that share similar properties in terms of ecological functioning. An example is an arid ecosystem, which exhibits patchy resource distribution. By determining whether there are consistent drivers of faunal communities across ecosystems we can yield powerful insights and potentially allow the development of simple indicators or predictors of the biota.

1. General Introduction

Thesis objectives

The main objective of this thesis is to examine how the spatial distribution of resource patches affects arthropod communities at multiple spatial scales in several dryland woodland or shrubland ecosystems.

Chapter 1 provides an overview of how resources are concentrated at multiple spatial scales in dryland ecosystems, and how this is known to affect the arthropod biota.

Chapter 2 investigates how two different shrub species and different structural arrangements affect the distribution of shrub-resident arthropods within a semi- encroached woodland of western NSW. It describes a survey of arthropods, particularly Hemiptera, on two widespread, commonly co-occurring shrub species (Eremophila sturtii and Senna artemisioides), growing either as isolated plants or in high density patches.

Chapter 3 examines whether mallee trees modulate (structure) the ground- dwelling arthropod community through the creation of a unique subcanopy environment. Landscape modulator theory (Shachak et al., 2008) proposes that woody vegetation affects biotic communities through the creation and maintenance of resource patches. Focusing on broad taxonomic groups (order),

I contrast the arthropod community (ants, beetles, spiders, wasps, isopods, etc.) in the surface leaf litter directly under the tree canopy with open patches away from the canopy that are largely devoid of vegetation. Additionally, through creation and manipulation of artificial habitat plots, this chapter also investigates

1. General Introduction

the role of shade and/or leaf litter (i.e. structural complexity) in structuring the arthropod community at fine scales.

Chapter 4 further examines landscape modulation in mallee communities, specifically whether the strength of modulation is affected by ecological disturbance (fire). It compares the effects of landscape modulation (specifically the contrast between the arthropod community in the canopy and open patches) in an area that has not been burnt for at least 30 years, and in an area burnt four years prior to sampling. This compares the process of modulation between two different scenarios; modulation in an area with a relatively extensive resource patchiness, and one with a recently decayed and poorly-developed resource patchiness.

Chapter 5 examines the extent to which two widely used measures of ecosystem health are useful surrogates of ground-dwelling arthropod biodiversity in an extensive woodland community type in eastern Australia. It examines whether an index of biotic integrity, based upon structural, compositional and functional metrics or metrics used in Landscape Function

Analysis are able to explain variation in the abundance, richness, and diversity of ants, spiders, and beetles, as well as whether there are links between these predictors and arthropod species / functional group composition. This chapter has been published in Austral Ecology.

Chapter 6 assesses the relative importance of landscape context, within-patch habitat/vegetation characteristics and biotic interactions in structuring ant

1. General Introduction

communities in a fragmented grassy box woodland of south-eastern Australia. It investigates multiple scales of ant community drivers in order to determine whether finer scale variables are more effective than broad-scale variables in structuring species and functional group composition.

Chapter 7 presents the main findings and implications of the research in this thesis. It also discusses important areas for future research that are evident from the work in this thesis, as well as avenues for each specific theme.

2. Arthropods in shrublands

Chapter 2

Shrub species, not spatial arrangement, structures

arthropod communities in a shrub-encroached woodland

Alan B. C. Kwok and David J. Eldridge

Keywords: arthropod, encroachment, resource concentration, shrubs, hemiptera, plant density, psocoptera, collembola

2. Arthropods in shrublands

ABSTRACT

Plant-resident arthropods are closely tied to the composition, distribution and density of host and non-host species. In arid and semi-arid environments, shrub encroachment is a widespread form of vegetative change involving substantial structural and compositional shifts in vegetation. However, there has been very little research on the effects of this phenomenon on arthropod communities. We examined the role of shrub species and fine-scale plant density in structuring arthropod communities in a semi-encroached Australian woodland using two common and widespread shrub species, Turpentine (Eremophila sturtii) and

Silver Cassia (Senna artemisioides subsp. filifolia). We found five-times more arthropods (Psocoptera, Collembola and Hemiptera) on Eremophila compared with Senna, representing a significantly different arthropod community. Each shrub species also supported a unique species assemblage of hemipterans, even though the shrubs grow in close proximity (< 15 m). In contrast, we found limited effects of fine-scale plant density, with plants growing in low and high density supporting similar arthropod communities. Our data indicate that the shrubs in the semi-arid encroached woodlands within our study area support a range of arthropod taxa, and that shrub species is a more important driver of arthropod community structure than spatial arrangement (in terms of density).

As shrub encroachment alters plant community composition globally, it is imperative that research be conducted to improve our understanding of the broad-scale effects of encroachment and shrub management on plant-resident fauna.

2. Arthropods in shrublands

INTRODUCTION

The effects of habitat change on biotic communities are known to vary markedly with differences in spatial scale (Mazerolle and Villard, 1999; Driscoll, 2004;

Goslee and Sanderson, 2010). For example, changes in habitat structure at landscape scales can alter patterns of faunal distribution by changing immigration and emigration rates (sensu Equilibrium Model of Island

Biogeography, MacArthur and Wilson, 1967) and interaction networks

(Gonzalez et al., 2011). Changes at finer spatial scales, such as those around individual plants, can affect the abundance, diversity, and distribution of plant- resident herbivores such as arthropods, and in particular insects (Masumoto et al., 2000; Hodkinson et al., 2001; Sarin and Bergman, 2010). Conversely, insects can affect host-plant biomass (Flynn et al., 2006) as well as functions such as photosynthetic rate (Meyer and Whitlow, 1992).

Plant-resident arthropods are closely tied to the distribution of their hosts, and changes in the composition, density and spatial arrangement of host plants can alter the community structure of their residents (Schaffers et al., 2008; He et al.,

2010; Murakami and Hirao, 2010). One of the most influential hypotheses guiding the study of herbivore–plant interactions over the past few decades has been the Resource Concentration Hypothesis (Root, 1973). This hypothesis predicts that specialist herbivores will be more abundant and speciose in high density plant patches as these patches are easier to detect and have a higher concentration of resources than low density patches or isolated plants.

Furthermore, higher density patches provide larger areas of contiguous habitat, which may also increase faunal abundance and diversity (sensu Equilibrium

2. Arthropods in shrublands

Model of Island Biogeography). Support for this pattern has, however, been mixed (Bach, 1980, 1988; Cook and Holt, 2006). Indeed, recent developments in ecology have emphasised the importance of habitat heterogeneity, rather than patch size or isolation, in determining patterns of species diversity (Baldi,

2008). Thus, low density patches or isolated plants may provide greater heterogeneity in habitat and thus support a greater abundance and diversity of insects. However, such alternative hypotheses have not been tested widely in arthropod–plant systems.

In arid and semi-arid environments, shrub encroachment is a widespread form of vegetative change involving substantial structural shifts in vegetation (Van

Auken, 2000; Archer, 2010). Encroachment involves a staged shift from a perennial grass to a woody plant – dominated community (Browning et al.,

2008). The process commences with the incursion of isolated, often native, shrubs into open woodland or grassland, or increases in the density of existing scattered shrubs. Complete conversion to closed shrubland can occur in as little as 50 years (Archer, 2010), though this is affected by a range of factors including fire regimes and grazing (Van Auken, 2009). Such changes in the cover and composition of individual woody species can alter a range of ecosystem processes and functions such as nutrient dynamics and primary productivity (Eldridge et al., 2011). Encroachment also substantially alters resource heterogeneity, shifting the spatial distribution of nutrients from the scale of individual grasses to that of more widely-spaced shrub canopies

(Archer, 2010). Encroachment is also characterised by a decoupling of shrub

2. Arthropods in shrublands

biomass and density relationships, which can alter the structure of individual plants.

As shrub encroachment changes the composition and structure of the vegetation, it can have dramatic effects on faunal assemblages. Several studies have demonstrated that encroachment-driven changes in plant structure and composition are largely taxon-specific (e.g. Ayers et al., 2001; Blaum et al.,

2007b; Sirami et al., 2009; Eldridge et al., 2011). Woody plants are known to provide essential habitat for a range of plant-resident insects such as Hemiptera

(Carver et al., 1991) and Lepidoptera (Nielsen and Common, 1991). There is, however, little information regarding how changes to shrubland structure influence the distribution of arthropods (e.g. Marques et al., 2000; Reid and

Hochuli, 2007).

We examined three propositions about the effects of shrub species and fine- scale differences in shrub density and structure on plant-resident arthropods in a semi-encroached Australian woodland. Firstly, we expected that our two shrub species, Eremophila sturtii and Senna artemisioides subsp. Filifolia, would support different arthropod taxa, given their markedly different size and architecture, and the fact that many insects are commonly host-specific

(Peeters et al., 2001; Sugiura, 2011). Secondly, consistent with expectations under the Resource Concentration Hypothesis and the Equilibrium Model of

Island Biogeography, we predicted that shrubs growing in patches of high density or biomass, which are characteristic of encroached woodland, would support a greater richness and abundance of insect taxa than those growing as

2. Arthropods in shrublands

isolated individuals. Shrublands do not immediately attain high plant density, rather early stages of encroachment often support many isolated shrubs. The comparison between the two structural scenarios will provide us with insights into potential arthropod differences in relation to different stages of encroachment (Roques et al., 2001). Thirdly, we predicted that the abundance and diversity of arthropods would increase with increasing shrub biomass, consistent again with the Resource Concentration Hypothesis and Equilibrium

Model of Island Biogeography.

METHODS

Study area

This study was conducted at the Australian Wildlife Conservancy’s Scotia

Sanctuary, which is located about 150 km south of Broken Hill, NSW, Australia

(33°43’S, 143°02’E). The climate is characterized by low and variable rainfall

(250mm mean annual rainfall), high evapotranspiration (~ 1500mm yr–1), hot summers (daily mean temperature: 30°C, daily maximum: 47.8°C, daily minimum: >15°C) and cool winters (daily mean: ≤17°C, daily maximum: 32.2°C, daily minimum: ≤6°C) (Australian Wildlife Conservancy, 2011).

Vegetation community and target plant species

This study was conducted within encroached Eremophila- and Senna- dominated communities. In the study area, encroachment is thought to have been triggered by soil disturbance resulting from widespread tree removal in many parts of original Casuarina pauper open woodlands in the early 1960s.

2. Arthropods in shrublands

This was likely for pasture improvement, however much of the study area has only been grazed lightly in the past 100 years (M. Westbrooke, pers. comm.).

The study area has not been grazed by domestic livestock since 1994. The former woodland areas in this study are now largely devoid of trees and are instead dominated by a wide variety of native shrubs of the genera Eremophila,

Dodonaea, Senna and Rhagodia. Two shrub species were investigated in this study: Eremophila sturtii R. Br. (Turpentine, hereafter Eremophila), and Senna artemisioides subsp. filifolia Gaudich, ex DC (Silver Cassia, hereafter Senna)

(Figure 2). Both shrub species are widely distributed in semi-arid eastern

Australia, occurring as either scattered individuals within woodland communities, or in extensive dense monocultures (Noble, 1997). Within the study area these two species are co-dominant, accounting for over 70% of all individual shrubs (A. Kwok, personal obs.).

a) b)

Figure 2.1: Photographs of target shrub species: a) Eremophila sturtii and b) Senna artemisioides

2. Arthropods in shrublands

Experimental design and shrub density

Ten sites were chosen within Eremophila–Senna shrublands that were formerly

Casuarina pauper woodlands. Individual sites occurred in the swales between low, east–west trending dunes. Sites were separated by distances of at least

500 m. At each site, we chose four sections at opposite ends of the swale within a plot about 300 m by 300 m in size. Within two of the four sections we chose a shrub growing in a patch considered to be at a low density. ‘Low density’ was defined as having fewer than two conspecific shrubs within a radius of 5 m (i.e. over an area of about 80 m2) (Figure 2.2a). Within the other two sections, we selected two shrubs growing in ‘high density’ patches, defined as 10–20 conspecific shrubs within an area of 80 m2 (Figure 2.2b). The biomass of each shrub was calculated using published allometric relationships relating shrub height to biomass (Harrington, 1979). We refer to this value herein as

"calculated biomass". Thus, in total, we sampled 80 shrubs (two species × two treatment configurations (densities) × two replicates × 10 sites).

Arboreal arthropods were sampled from all 80 shrubs by fogging each plant with a pyrethrum-based insecticide (Pyzap, 40 g L-1 Pythrethrins, 160 g L-1 piperonyl butoxide, Agserv-Ruddock Agriculture). Spraying was only conducted on wind- free days over a period of 1 week in May 2009. Thirty minutes after spraying, each shrub was shaken for 30 seconds to 1 minute, depending on the size of the shrub, to dislodge any arthropods. Arthropods were collected from a sheet placed below the shrub, and transferred to vials of ethanol for storage. Foliage dislodged from the shrubs during shaking was inspected separately for arthropods. These two sources of arthropods were pooled for our analyses.

2. Arthropods in shrublands

From all the taxa sampled, results were recorded only for Collembola,

Hemiptera (including a large number of Psyllidae) and Psocoptera. These taxa represented over 90% of the total number of arthropods sampled (unpublished data). Psocoptera were only identified at the ordinal level. Collembola were identified to the family level. Hemiptera were further identified to morphospecies and verified by a taxonomic expert (Gerrosimus Cassis, University of New

South Wales). Analyses of species richness, diversity and composition were restricted to the Hemiptera. Specimens are currently stored with the primary author at UNSW.

a) b)

Figure 2.2: Photographs of density treatments, showing typical plant growing in a) low density and b) high density.

Statistical analysis

For statistical analyses, the abundance of each arthropod taxon was based on the mean of the two replicates for each treatment configuration at each site.

Generalised Linear Models (GLM) were used to test for differences in arthropod

(Hemiptera, Psyllidae, Collembola and Psocoptera) abundance between shrub species (Eremophila, Senna), shrub density (low, high), and the interaction of 24

2. Arthropods in shrublands

these factors. While Psyllidae is a family of Hemiptera, it was treated separately in our analyses due to the presence and numerical dominance of only one

Psyllid species. GLMs were also used to test for differences between shrub density and shrub species in the species richness and diversity of Hemiptera.

For a diversity measure, we used a bias-controlled effective number of species

(Jost, 2006), which has been shown to be one of the least biased diversity estimates (Beck and Schwanghart, 2009). Bias-controlled effective number of species (hereafter ‘effective species diversity’) was calculated using the program SPADE (Chao and Shen, 2003). Each arthropod taxon was analysed in a separate model, using either a negative binomial (number of arthropods) or

Gaussian (number of hemipteran species, effective species diversity) distribution. The choice of distribution was based on visual analysis of the residual graphs (following Zuur et al., 2009). All GLM analyses were conducted in the program R (R Development Core Team, 2011) using the MASS package

(Venables and Ripley, 2002). Deviance explained by the model (also known as the Pseudo-R2) was calculated as per Zuur et al., (2009).

We used linear regression to examine the relationship between calculated plant biomass and the abundance of Psocoptera, Collembola and Hemiptera, for each shrub species and density treatment. Additionally, the abundance of

Psyllidae and calculated biomass was regressed for Eremophila only, as

Psyllids were found only on this shrub species. For Hemiptera, we examined the relationship between calculated plant biomass and the number of species and effective species diversity.

2. Arthropods in shrublands

Hemipteran species composition

A two-way Permutational Multivariate Analysis of Variance (PERMANOVA,

Anderson and Gorley, 2008) was used to test for differences in the hemipteran species composition between shrub species and shrub density, and the interaction of these factors. We used non-metric multi-dimensional scaling to display patterns of community composition within each shrub species and shrub configuration. These analyses were based on square-root transformed data and a zero-adjusted Bray–Curtis resemblance matrix (Clarke and Gorley, 2006).

The closer any given points are to each other in the resulting ordination, the greater their similarity in terms of species composition. The PRIMER (Version 6)

(Clarke and Gorley, 2006) + PERMANOVA (Anderson and Gorley, 2008) package was used for all multivariate analyses, including the creation of species accumulation and dominance curves. Analyses of species composition excluded rare species. For the purposes of our analyses, ‘rare’ species were defined as those that were recorded on less than five shrubs, or had less than five individuals sampled in total.

RESULTS

A total of 9,102 arthropods were collected in this study: 85% from Eremophila sturtii and 15% from Senna artemisioides. Fifty-one percent of individuals on

Eremophila were Psocoptera, 28% Hemiptera, 15% Psyllidae and 6%

Collembola. The fauna on Senna was dominated by Psocoptera (69%) with smaller numbers of Hemiptera (22%), Collembola (8%) and Psyllidae (1%). All

Collembola were from the family Sminthuridae. Species accumulation curves for

Hemiptera were near plateau for Eremophila, indicating that sampling effort was 26

2. Arthropods in shrublands

sufficient to record a high proportion of Hemiptera on this plant species (Figure

2.3). For Senna, however, the curve indicated that more sampling would reveal additional species.

Figure 2.3: Species accumulation curves for Senna artemisioides (dashed line) and Eremophila sturtii (solid line), using number of observed hemipteran species.

Effect of shrub density on the arthropod fauna

There was no effect of shrub density on the total abundance of arthropods

(Density (D): P >0.05), and this was the case for Eremophila and Senna (shrub species (SS) x D interaction: P >0.05; Table 2.1, 2.2). This was also the case for the number of Psocoptera, Psyllidae, and Collembola when analysed separately (Table 2.1, 2.2). The abundance of Hemiptera was significantly greater on low density shrubs, for both Senna and Eremophila (D: P < 0.05, SS

× D interaction P >0.05; Table 2.2, Figure 2.4a). The number of hemipteran species was higher on low density shrubs for Senna, but this was not

2. Arthropods in shrublands

statistically significant (P >0.05; Table 2.4; Figure 2.4b). The effective species diversity of hemipterans was 35% to 43% lower on high density Senna compared to all other density and shrub species groups (SS: P < 0.01; D: P <

0.01; SS × D interaction: P < 0.01; Table 2.2, Figure 2.4c).

The hemipteran species assemblage also differed markedly between high and low density shrubs (PERMANOVA Pseudo-F1,36 = 2.28, P = 0.02; Figure 2.5a), and this was the case for both Eremophila and Senna (SS × D interaction:

Pseudo-F1,36 = 1.73, P = 0.08).

Table 2.1: Mean (± SE) abundance of total arthropods, Psocoptera, Collembola and Psyllidae in relation to shrub species and shrub density. Different letters within a row indicate a significant difference at P < 0.01. L = low density, H = high density, E = Eremophila sturtii, S = Senna artemisioides. For Hemiptera (non-Psyllidae, see Figure 2.4.

2. Arthropods in shrublands

Effect of shrub species on the arthropod fauna

Arthropods were more abundant on Eremophila than Senna (SS: P < 0.001) and the trends were similar for all arthropod groups when analysed separately

(Table 2.1, 2.2). Across both shrub species, 39 hemipteran species from the sub-orders Auchenorrhyncha and Heteroptera were recorded across seven families: Cicadellidae (19 spp.), Delphacidae (8 spp.), Miridae (6 spp.),

Pentatomidae (3 spp), Tingidae (1 spp.), Lygaeidae (1 spp.), and Nabidae (1 spp.).

Thirteen species of Hemiptera were recorded only on Eremophila, 16 only on

Senna, and 10 were sampled on both shrub species. Of the 10 species sampled on both Eremophila and Senna, seven of these could be defined as rare on Senna by our definition (i.e. fewer than five individuals sampled or sampled on fewer than five shrubs in total). On Eremophila, the mean abundance of these seven species ranged from 0.4 ± 0.13 to 28.6 ± 4.63 individuals per shrub (24 to 1143 individuals in total), and were therefore not rare. The remaining three species common to both Eremophila and Senna were rare on both shrub species. Of the 16 species sampled only on Senna, 11 could be considered rare. This number was the same for those occurring only on

Eremophila.

2. Arthropods in shrublands

Residual Deviance deviance P

Total arthropods (0.61) Null 107.91 Shrub species 65.86 42.05 <0.001 Density 0.02 42.03 0.89 Shrub species × Density 0.005 42.02 0.94 Collembola (0.56) Null 85.83

Shrub species 40.12 45.71 <0.001 Density 0.42 45.29 0.52 Shrub species × Density 1.9 43.39 0.17 Psocoptera (0.38) Null 72.8

Shrub species 27.03 45.78 <0.001 Density 0.34 45.43 0.82 Shrub species × Density 0.06 45.37 0.81 Psyllidae (0.72) Null 135.38

Shrub species 96.62 96.62 <0.001 Density 0.559 38.199 0.14 Shrub species × Density 0.019 38.18 0.91 Hemiptera (0.73) Null 140.45

Shrub species 97.19 43.27 <0.001 Density 4.67 38.59 0.03 Shrub species × Density 0.01 38.58 0.91 Number of species (Hemiptera) (0.50) Null 39.06

Shrub species 15.8 23.26 <0.001 Density 1.48 21.78 0.22 Shrub species × Density 2.407 19.37 0.12 Effective species diversity (Hemiptera) (0.24) Null 46.02 Shrub species 2.75 43.27 0.004 Density 1.94 41.33 0.008 Shrub species × Density 6.47 34.85 0.014

Degrees of freedom (d.f.) for shrub species, density, and their interaction all 1, while residual d.f. are 38, 37, and 36 respectively. Total residual d.f. are 39. Pseudo R values of 1 indicate that all variation in the model was explained by the predictor variables.

2. Arthropods in shrublands

a) b)

2. Arthropods in shrublands

Eremophila supported a greater number of hemipteran species, irrespective of shrub density (4.78 ± 0.31 cf. 2.38 ± 0.3; SS: P < 0.001, Table 2.2, Figure 2.4b).

When shrub density was pooled and the abundance of individuals standardized

(rarefaction), hemipteran species richness was higher on Senna (n = 17) than

Eremophila (n = ~ 10: 95% confidence interval = 7 – 13; Figure 2.6). As noted above, hemipteran effective species diversity was dependent upon shrub species and density (SS × D interaction: P = < 0.01, Table 2.2, Figure 2.4c), and was significantly lower on high density Senna compared with all other treatments. The species assemblage of hemipterans differed markedly between shrub species (PERMANOVA Pseudo-F1,36 = 51.61, P = 0.0001), and as noted above this was the case regardless of shrub density (Figure 2.5b).

Relationships between calculated plant biomass and the arthropod fauna

For Eremophila at low densities, the number of individuals increased with calculated plant biomass for all arthropod taxa (Psocoptera, Hemiptera,

Psyllidae, and Collembola), as did the number of hemipteran species (R2 = 0.21 to 0.31; Table 2.3). At high plant density, relationships between calculated plant biomass and arthropod abundance were not statistically significant (Table 2.3).

In contrast, for Senna the number of Hemiptera (individuals and species) and effective species diversity decreased with calculated plant biomass, but only for plants growing at high density (R2 = 0.15 to 0.18; Table 2.3).

2. Arthropods in shrublands

Figure 2.6: Rarefaction curves for the number of Hemiptera species on Senna artemisioides (closed circles) and Eremophila sturtii (open circles).

Table 2.3: Adjusted-R2 for the linear regressions between estimated plant biomass and the number of arthropods for Eremophila sturtii and Senna artemisioides at low and high densities. Significant regressions are always positive for Eremophila but negative for Senna.

Senna artemisioides Eremophila sturtii

Density Low High Low High

Collembola n.s. n.s. 0.21* n.s.

Psocoptera n.s. n.s. 0.25* n.s.

Psyllidae n.s. n.s. 0.21* n.s.

Hemiptera n.s. 0.18* 0.31** n.s.

Number of species

(Hemiptera) n.s. 0.16* 0.23* n.s.

Species diversity

(Hemiptera) n.s. 0.16* n.s. n.s. n.s. = not significant at P < 0.05; * P < 0.05; ** P < 0.01.

2. Arthropods in shrublands

DISCUSSION

Our data indicates that the semi-arid encroached woodlands in our study area support a range of arthropod taxa, and shrub species is a more important driver of arthropod community structure than shrub spatial arrangement (in terms of density). The two shrub species in our system (Eremophila sturtii and Senna artemisioides subsp. filifolia) supported markedly different assemblages of arthropods, irrespective of whether they were growing in low or high density patches. This implies that plant-resident arthropods will benefit from encroachment primarily through broad-scale increases in the density of host plants, rather than changes in their spatial arrangement or fine-scale density.

Multiple shrub species enhance insect biodiversity

As we expected, Eremophila sturtii and Senna artemisioides supported clearly different hemipteran communities, even though the shrubs grew in close proximity (<15 m). Most hemipteran species were found exclusively or predominantly on only one shrub species. Indeed, our study indicates that the presence of multiple shrub species within a community likely increases community-scale arthropod richness and abundance. This is particularly the case in our system given that rarefaction curves indicated the presence of additional, unsampled species on Senna. While it is possible that additional sampling of Senna would reveal greater overlap between the hemipteran communities of the two shrub species, the general low number numbers on

Senna would suggest that such hemipteran species still show preference for

Eremophila.

2. Arthropods in shrublands

Few hemipteran species in our study could be considered generalists, that is, for our purposes, utilizing both Eremophila and Senna. No species was equally abundant on Eremophila and Senna – most of the hemipteran species that were sampled on both shrub species were abundant and readily sampled on

Eremophila, yet rarely recorded on Senna. In the most extreme case, one species of Cicadellidae was found in high numbers on every Eremophila sampled (~30 individuals per shrub), yet was found on Senna only as two isolated individuals. These results suggest that in our system, many of these species do appear to be abundant, yet genuinely restricted or ‘specialized’ to a particular shrub species. This pattern was consistent regardless of plant density.

Our results are consistent with the generally accepted relationship between plant and herbivore diversity (Strong et al., 1984; Attie et al., 2008). For phytophagous taxa (i.e. the Hemiptera) this result is not entirely surprising given it is well known they are often host-specific (e.g. Moir et al., 2010). The results are more surprising for non-herbivorous taxa (e.g. the Collembola and the

Psocoptera), since these groups are not thought to show strong host preference

(Broadhead and Thornton, 1955; Smithers, 1991), rather they tend to be affected by factors that could be expected to be influenced by plant density (e.g. relative humidity, Greenslade, 1991). The clear weight of abundance of these taxa on Eremophila indicates that their distribution within this system is heavily influenced by plant species composition. The presence of host-specific arthropod communities also suggests that associational effects (e.g. presence

2. Arthropods in shrublands

of species on Eremophila due to the nearby presence of Senna) are minimal in our system (e.g. Agrawal et al., 2006).

Shrub species affects arthropod richness

We found a greater richness of hemipteran species on Eremophila compared with Senna. While we did not identify the other taxa (Psocoptera and

Collembola) to the species level, it seems plausible that they follow a similar pattern. Such large differences in richness between the two shrub species were likely due, in part, to more individuals being sampled on Eremophila (as suggested by rarefaction analyses for Hemiptera). This does not, however, change the fact that the species richness of Hemiptera on any given Eremophila was significantly greater than on any given Senna.

A variety of species-specific plant characteristics, such as plant biomass and architecture, may explain why our shrub species had different levels of arthropod richness. Arthropod richness is often positively correlated with plant size or biomass (e.g. Masumoto et al., 2000; Hodkinson et al., 2001). This, however, was only partially supported by our results. Plant biomass did not differ significantly between our two shrub species (unpublished data), and relationships between abundance and calculated biomass were weak even within each shrub species.

Differences in arthropod richness between Senna and Eremophila could be due to distinct differences between the species in terms of plant and leaf architecture. Eremophila has many small leaves that are quite closely clustered

2. Arthropods in shrublands

together, and each leaf tends to be randomly oriented. In contrast, Senna has fewer, larger leaves that are more openly spaced and uniform in orientation.

Differences in plant architecture can exert strong effects on insect communities, both within and between plant species (Lawton, 1983; Reid and Hochuli, 2007).

For example, increased complexity on Eremophila may create more predator free spaces, resulting in reduced predation pressure on arthropods (e.g. Jeffries and Lawton, 1984; Obermaier et al., 2008), while also providing a greater number and diversity of niches and overall resources (e.g. flowers, Marques et al., 2000). Additionally, Eremophila leaves have a distinctly sticky, resinous surface. The surface texture of leaves has been shown to affect both the foraging behaviour and success of predatory arthropods (e.g. ladybirds,

Grevstad and Klepetka, 1992), and similar processes may occur on Eremophila.

Plant architecture also affects arthropods by influencing microclimate. For example, the temperature, relative humidity and light intensity around the surface of a leaf can vary dramatically between plant species and with architecture (e.g. Stoutjesdijk and Barkmann, 1992; Raghu et al., 2004). Thus, the difference in the density and texture of Eremophila and Senna leaves may create vastly different microclimates (e.g. Senna leaves may be drier and hotter because of their open nature). Microclimatic conditions can have a very strong influence on arthropod abundance and behaviour (e.g. Raghu et al., 2004), as well as mating success (Larsson, 1989; Guarneri et al., 2002). These conditions could also potentially be influenced by the density of plants (i.e. plants growing in isolation suffer greater insolation, leaf temperatures, etc.).

2. Arthropods in shrublands

Encroached shrublands are dynamic habitat for insects

In arid and semi-arid ecosystems, one of the dominant changes to vegetation in the past 200 years has, and continues to be, the broad-scale replacement of perennial grasses by trees and shrubs (Van Auken, 2000). Although an increasing body of research indicates variable effects of woody encroachment on arthropods (Ayers et al., 2001; Bestelmeyer, 2005; Blaum et al., 2009), there have been few attempts to quantify the biodiversity supported by these shrublands or elucidate mechanisms controlling these communities. We provide evidence that shrublands can support a range of arthropod taxa. Indeed, plant- resident arthropods are a key group of animals likely to benefit measurably from shrub encroachment because it increases the availability of habitat and resources specifically used by these animals, relative to scattered shrubs found in perennial grasslands or former woodlands. These resources will allow population increases of many arthropod species, and potentially result in changes to the distribution of some species.

The identity of the encroaching shrub species is likely to influence the composition of plant-resident fauna. In some areas, shrub encroachment can involve up to ten native shrub species (Ayers et al., 2001). These shrubs are likely to represent a diverse range of sizes and architectures and provide different seasonal resources, such as flowers. These differences are likely to affect which and how many arthropods are found within any given area, as discussed above. Thus, the biodiversity within an encroaching shrubland will be affected by the host-specificity of the fauna as well as the differences between encroaching shrub species in the resources they provide to arthropods. Indeed,

2. Arthropods in shrublands

it is possible that management actions that alter the composition of shrublands

(e.g. shrub ploughing, grazing; Daryanto and Eldridge, 2010) will affect patterns of arthropod biodiversity at broad spatial scales. Given the current, generally negative, attitude toward shrub encroachment, it is imperative that research be conducted to improve our understanding of the broad-scale effects of encroachment and shrub management on the plant-resident fauna.

Changes to ecological interactions within a shrubland may have broader scale ecological consequences for both arthropods and plants. Additional host plants can result in increased ecological opportunity for herbivores (Strong et al.,

1984) and potential diet-switching by predominantly monophagous species of arthropods (e.g. Agrawal et al., 2006). Diet-switching can have clear effects on insect fitness (Mody et al., 2007). More commonly, however, generalist herbivores or phytophagous species are beneficiaries when a plant invades a new area. Increases in sap-feeding activity by phytophagous animals may also have serious detrimental effects on plant growth, and these have only recently been broadly acknowledged (Zvereva et al., 2010). The extent to which arthropods benefit from encroachment, however, may depend upon plant- specific responses. For example, a plant may increase its production of anti- herbivore compounds to protect against generalist herbivores, but this may result in lower protection against specialist herbivores (e.g. Joshi and Vrieling,

2005). Knowledge of ecological interactions between arthropods and plants in encroached systems is severely lacking, and this is an area of research in thorough need of investigation.

2. Arthropods in shrublands

Shrub spatial arrangement has limited effects on the insect community

Contrary to our predictions, we found limited effects of plant density/plant spatial arrangement on the arthropod community in our study area. The abundance of three of the four arthropod groups was similar on plants growing at low and high densities. Hemiptera (non-Psyllidae) were the exception, with low density shrubs supporting a greater abundance and richness of individuals. Further, while we found that plants at low densities supported a compositionally distinct species assemblage of Hemiptera, this was largely due to subtle shifts in the relative abundance of each hemipteran species (data not shown). Our results demonstrate that, for Senna–Eremophila shrublands in our study area, differences in the spatial arrangement of resources at a scale of tens of metres have only weak effects on the arthropod fauna and community-scale diversity.

A number of mechanisms could potentially account for greater hemipteran abundance on low density shrubs. Isolated shrubs may be larger in size than those growing in close proximity, and may therefore be able to support more herbivores (e.g. Reznik, 1993 or the Plant Size per se Hypothesis, Lawton,

1983). In our system, however, there were no differences between the size of shrubs growing in low and high– density patches (unpublished data).

Alternatively, animals that are wind-drifting or dispersing may collect in greater numbers on isolated shrubs, compared with shrubs surrounded by neighbouring plants (e.g. ‘target area effect’, Gilpin and Diamond, 1976). Species-specific foraging and dispersal strategies or mechanisms can also determine patterns of distribution in relation to available resources (Bukovinszky et al., 2005; Cook and Holt, 2006).

2. Arthropods in shrublands

The effects of plant spatial arrangement on arthropod communities are highly variable. Arthropods have been widely used as a model system to test hypotheses relating to plant–herbivore interactions, such as the Equilibrium

Model of Island Biogeography (MacArthur and Wilson, 1967). This has spurred a myriad of hypotheses linking resource concentration to faunal communities

(Root, 1973; Agrawal et al., 2006). However these relationships vary markedly, with recent research illustrating a full spectrum of results. Inverse relationships between arthropod abundance or activity (e.g. oviposition, herbivory) and plant density have been reported widely in the literature for a range of insects (e.g.

Capman et al., 1990; Cook and Holt, 2006; Hamback et al., 2010). Indeed, many responses seem to be species-specific, even within a given functional group (e.g. parasitoids, Klapwijk and Lewis, 2011). The spatial scale used by researchers has been highlighted as one reason for this disparity (Grez and

Gonzalez, 1995). We focussed on the scale of individual plants where density varies over tens of metres, as this scale has largely been neglected in natural systems and specifically is non-existent in studies involving shrub encroachment (Ayers et al., 2001; Bestelmeyer, 2005). Our evidence suggests that at fine spatial scales, resource concentration has only minor affects on arthropods in Senna–Eremophila shrublands, and that plant species composition is the dominant driver of the distribution of many species in this system.

Conclusions

The results of our study indicate that shrub encroached landscapes support several arthropod groups, including Psocoptera, Hemiptera and Collembola. We

2. Arthropods in shrublands

have shown that in a simple system the spatial distribution of the arthropod community is driven by shrub species composition, and not local shrub density.

Shrub encroachment is likely to benefit many plant–resident arthropods simply through increased resource availability. Changes to the herbivorous and phytophagous fauna may have tangible consequences for both plants and the arthropods themselves. For example, herbivores may alter plant health and defensive responses, which may subsequently affect arthropod species and feeding guilds as it does in other systems. Global research is needed to establish if and how arthropod communities change across gradients of shrub encroachment, and whether this alters plant-animal interactions. This is especially important in areas where many shrub species are increasing in abundance simultaneously.

ACKNOWLEDGEMENTS

We thank Gerrosimus Cassis for identification of Hemiptera. The Australian

Wildlife Conservancy provided financial and logistical support. We also thank

Terry Koen and Aaron Greenville for statistical advice, and Matt Hayward

(AWC) who provided useful comments on an earlier draft.

3. Mallee trees as landscape modulators

Chapter 3

Do trees modulate ground-dwelling arthropod communities

in the mallee of south-eastern Australia?

Alan B. C. Kwok and David J. Eldridge

Keywords: arthropod, landscape modulator, fertile island, mallee, patch, isopod, spider

3. Mallee trees as landscape modulators

ABSTRACT

Landscape modulator theory proposes that organisms, such as woody vegetation, affect the surrounding biotic community through the creation and maintenance of multilayered resource patches. In south-eastern Australia, mallee trees (Eucalyptus spp.) create multilayered resource patches beneath their canopies through the provision of shade and accumulated leaf litter and woody debris ('litter'). Canopy patches differ markedly from the open, inter-tree areas that are predominantly bare soil with scattered hummock grasses. In this study we examine whether mallee trees modulate the ground-dwelling arthropod community, and the role of structural complexity in this modulation process. We compared the arthropod community active on the ground surface in naturally occurring ('canopy' and 'open') patches with artificially created patches, which had litter and/or shade added to the plot (litter addition only, shade addition only, litter + shade addition). Naturally occurring canopy patches supported a greater number of isopods, spiders and wasps than any other patch type. In contrast, ants increased in abundance with decreasing structural complexity (highest in open patch, lowest in canopy and litter + shade patch). In general there were no differences in the number of each arthropod taxon between the artificially created patches and the open patches. Similarly, the number of beetles, jumping spiders, and silverfish was similar among all patch types. Our study demonstrates that mallee trees are landscape modulators, structuring arthropod assemblages through the creation of distinct and complex resource patches, with four of the seven arthropod taxa clearly more abundant in either canopy or open patches. Resources patches created by vegetation are

3. Mallee trees as landscape modulators

a feature of arid ecosystems globally, and patterns similar to those we observed are likely to be evident in these ecosystems.

INTRODUCTION

Emerging evidence over the past few decades has demonstrated that organisms exert substantial non-trophic effects on both their physical environment and its associated biota (e.g. keystone species, Paine, 1969; ecosystem engineers, Jones et al., 1994). One of the more recent concepts in this field of non-trophic ecosystem effects is that of the landscape modulator

(Shachak et al., 2008). A landscape modulator is an organism that modifies landscape configuration at patch and patch-mosaic scales, with subsequent effects on biotic communities (Shachak et al., 2008).

A unique and important feature of landscape modulator theory is that resource patches created and maintained by the landscape modulator are multilayered, providing a variety of unique biophysical and ecological resources. These patches are fundamentally different from unmodulated patches due to the ways that the modulator exploits and controls resources and resource distribution.

This results in a different array of resources (e.g. physical, chemical) provided in modulated patches compared with unmodulated patches. Essentially, there is a contrast in resources between the patch and the matrix (Kotliar and Wiens,

1990). This resource contrast drives differences among the patches in the biotic community that they support (Agra and Ne'eman, 2009; Bennett et al., 2009), resulting in increasing community divergence with increasing resource contrast

(Shachak et al., 2008). Thus the distribution of these patches is proposed to

3. Mallee trees as landscape modulators

drive differences in resource availability and biotic communities at relatively fine spatial scales, but also at broader landscape scales (Shachak et al., 2008).

Woody species such as trees and shrubs are landscape modulators because they create multilayered patches that develop and decay over their lifetime in response to external abiotic pressures (Shachak et al., 2008). Woody species exploit and modulate environmental resources. For example, growing roots exploit soil resources in between trees (unmodulated patches), resulting in the horizontal redistribution of water and nutrients from these open areas to areas closer to the tree (i.e. the modulated patch). Additionally, vegetation modulates

(increases) infiltration through the positive effects of their roots on soil porosity in the area immediately under the plant (Eldridge and Freudenberger, 2005).

Over time, plant growth and biomass deposition increases the strength of modulation (Shachak et al., 2008). Thus, the modulated patch provides a greater variety and amount of resources (biomass, litter, soil water and nutrients) than unmodulated patches (sensu fertile islands, Garner and

Steinberger, 1989). These patches are a long-lived feature of the environment, and persist for many years (e.g. Tongway et al., 1989). A range of diverse patch types is predicted to support a more diverse biotic community at both patch- and landscape- scales (Shachak et al., 2008; Agra and Ne'eman, 2009).

A critical way in which woody species modulate patches is by depositing and accumulating leaf litter and fine- and coarse- woody debris (e.g. sticks, bark) around their bases. This material is a crucial component of many ecosystems. It is a source of organic material and nutrients for plants (Facelli and Pickett,

3. Mallee trees as landscape modulators

1991; Butterfield and Briggs, 2009), provides physical protection of the ground surface (thereby reducing erosion and providing shade), and is used as habitat and food for animals (Seastedt and Crossley, 1984; Hunter et al., 2003; Castro and Wise, 2009). In arid and semi-arid ecosystems where decomposition processes are retarded, this debris can remain on the soil surface in various stages of decay for many years (Moorhead and Reynolds, 1991). This spatial and temporal patchiness of resources is well acknowledged (Schlesinger et al.,

1996; Ludwig et al., 2005), and is a key mechanism that drives plant productivity and the distribution of fauna within these ecosystems (Ludwig et al.,

2004; Shelef and Groner, 2011). Currently, however, there is limited quantitative information on the effects of modulation and patchiness on biotic communities within a landscape modulator framework (except see Agra and Ne'eman, 2009;

Bennett et al., 2009). This is in spite of abundant literature on woody plants in arid and semi-arid ecosystems indicating that they can have far-reaching effects on a range of biota, particularly at fine spatial scales (e.g. Noble et al., 1996;

Facelli and Brock, 2000; Mazia et al., 2006).

Mallee (Eucalyptus spp.) is a widespread vegetation type that covers large areas of semi-arid Australia, often in association with Triodia spp. hummock grasses (Noble et al., 1980; Westbrooke et al., 1998). Mallee is an open shrubland or woodland community, with canopy cover stabilising between 20% to 30% in the absence of disturbance (Haslem et al., 2011). Mallee has been extensively cleared in south-eastern Australia (Callister and Westbrooke, 2006).

Mallee trees have a coppicing habit, with multiple stems arising from an underground lignotuber. In mallee communities, extensive pools of surface litter

3. Mallee trees as landscape modulators

(up to 1.5kg m-2) accumulate in the absence of wildfire (Bradstock and Gill,

1993), primarily around the base of trees. Inter-tree areas are predominantly unvegetated and lacking in litter and woody debris, but support scattered

Triodia hummocks to a maximum cover of generally less than 30%. Litter and woody debris from mallee trees, and the Triodia hummocks between the trees, provide the only permanent form of vertical biomass and habitat complexity on the ground surface. Thus, mallee trees create multilayered resource patches that are structurally and functionally unique within dune–mallee systems.

Ground-dwelling arthropods are a diverse and functionally important group of animals that inhabit mallee litter, and the surface and soil layers beneath it.

These animals include insects, spiders, and isopods, and these taxa include detritivores, predators, omnivores and herbivores. Many of these animals are important for releasing and recycling nutrients from plant litter (e.g. brown food webs, Bardgett et al., 2005; Bastow, 2011) and their activities influence soil properties such as infiltration (Colloff et al., 2010). Although our understanding of how leaf litter and woody debris affects arthropod communities has increased markedly in recent years (e.g. Donoso et al., 2010; Silveira et al., 2010), relatively little is known about how the distribution of litter and woody debris affects the composition of resident arthropod communities in arid and semi-arid systems.

Given the density of litter around mallee trees, arthropods are likely to play an important role in the functioning of mallee ecosystems. For example, mites have been shown to be an important component of the mallee biota (Noble et al.,

3. Mallee trees as landscape modulators

1996). They are likely to be important in the decomposition of buried leaf litter, as in other arid ecosystems (Santos and Whitford, 1981). Overall, however, little is known about the composition of arthropod communities in mallee ecosystems, or how they are affected by the spatial distribution of litter and woody debris. The effect of this variation on animal distribution and composition is likely to be substantial. This may consequently affect ecosystem function, given the importance of arthropods in providing a range of ecosystem services

(Lavelle et al., 2006).

In this study we examine the role of mallee trees as modulators of the ground- dwelling arthropod community. We have two major questions in this study.

Firstly, do mallee trees modulate the ground-dwelling arthropod community by supporting a unique composition of arthropods in the patch beneath the canopy compared with open, Triodia-dominated areas? Secondly, what is the role of patch structural complexity in structuring arthropod communities? To answer these questions we compare the ground-dwelling arthropod community in two naturally-occurring patch types ('canopy' and 'open'), and in three experimental plots in which we manipulated the structural complexity to separate the effect of the litter from the effect the canopy. Artificially-created patches were comprised of ‘litter’ (leaves and fine or coarse woody debris), artificial shade, or a combination of litter and shade.

Overall, we predicted that mallee trees would modulate the ground-dwelling arthropod community, and that the canopy patches would support a greater abundance and diversity of faunal groups than the open, Triodia-dominated

3. Mallee trees as landscape modulators

patches (e.g. Barton et al., 2009; Lindsay and Cunningham, 2009).

Furthermore, we predicted that arthropod activity (measured as abundance) and diversity would increase with increasing patch structural complexity (e.g.

Nitterus and Gunnarsson, 2006; Lindsay and Cunningham, 2009). Thus, we expected that the similarity in the arthropod community between patches would decline with increasing differences in the structural complexity of the patches

(from higher to lower complexity: canopy, litter + shade, litter only, shade only, open). Finally, we also predicted that the abundance of less mobile and organic matter-dependent detritivores, such as slaters and cockroaches (and their faunal predators), would increase with increasing structural complexity of the patch. In contrast, other more mobile taxa (ants) would exhibit the opposite pattern (e.g. Andersen, 2003).

METHODS

Study area

This study was conducted in January 2011 at the Australian Wildlife

Conservancy’s Scotia Sanctuary, which is located 150 km south of Broken Hill,

NSW, Australia (33°43’S, 143°02’E). The climate is characterised by low and variable rainfall (250mm: mean annual rainfall), high evapotranspiration

(~1500mm yr–1), hot summers (daily mean temperature 30°C, daily maximum

47.8°C, daily minimum >15°C) and cool winters (daily mean ≤17°C, daily maximum 32.2°C, daily minimum ≤6°C) (AWC, 2011).

3. Mallee trees as landscape modulators

Vegetation community

This study was conducted in dune–mallee communities. Dune–mallee communities within the study area are located on long, low (relief to 7 m) east– west trending sand dunes dominated by an overstorey of mallee trees

(Eucalyptus dumosa A. Cunn. ex J. Oxley and E. socialis F. Muell. ex Miq.) scattered between 5 m and 50 m apart (Figure 3.1a). Within inter-tree areas, scattered perennial hummock grasses (Triodia scariosa N.T. Burb, hereafter

Triodia) dominate the ground to low strata (Figure 3.1b). Shrub cover to 2m is sparse, with widely-spaced individuals of predominantly Senna artemisioides subsp. filifolia Randell and petiolaris Randell, and Acacia burkitii F. Muell. ex

Benth. At our study site, the soils are mainly calcareous, brownish and siliceous sands. Our specific study area had not been burnt by wildfire in approximately

30 years.

Sampling and plot design

We conducted this study at 10 sites within dune–mallee, with sites separated by distances of at least 500m. All sites were located on different dunes. At each site arthropods were sampled on the ground surface from within five patch types. Two of these were naturally occurring patches ('canopy' and 'open'), and three were artificially created to represent various levels of structural complexity. In total, 50 plots (5 patch types ×10 sites) were sampled. Naturally occurring and artificially created patch types were compared to determine how structural complexity influences arthropod assemblages in mallee communities.

3. Mallee trees as landscape modulators

Natural or artificially created patch types

Canopy patches were located in the area of woody debris on the ground surface around the base of mallee trees but within the canopy dripline (Figure

3.1a,b). Open patches were located in inter-tree areas dominated by Triodia, at least 10 m away from any tree canopy (Figure 3.1c,d). In this study area, canopy patches contain a relatively extensive litter bed (litter bed area: 23.6 ±

3.1m2, litter depth: 5.3 ± 0.7cm, litter load: 1.4 ± 0.1kg m-2, Figure 3.1b) which extends between 2.5 m and 3 m from the base of the tree (Samantha Travers, unpublished data, 2012). In contrast, open patches are naturally devoid of large amounts of litter and woody debris (< 0.020kg m-2, Jane Smith et al., unpublished data, 2012, Figure 3.1d). We consider canopy patches to be

'modulated', as they are created and maintained by the landscape modulator

(mallee trees). Open patches, in contrast, are relatively 'unmodulated', devoid of biomass and litter, and do not benefit from the growth of the mallee tree in terms of resource accumulation or deposition. We acknowledge that Triodia modulates its surroundings, though this is at a much finer spatial scale compared to mallee trees.

Three artificial patch types were created by manipulating i) litter and coarse- and fine- woody debris i.e. bark, and branches and ii) shade. Three experimental plots were created at each site: litter addition only (hereafter litter); shade addition only (hereafter shade); and litter and shade addition (hereafter litter + shade). Each plot was 2 m x 2 m, and was located at least 10 m from a tree canopy, within the open/Triodia areas. All patches, including the canopy and open patches, were separated by distances of at least 40 m.

3. Mallee trees as landscape modulators

a) b)

c) d)

Figure 3.1: Photos of typical study site within mallee community, showing a) resource patch under a tree canopy, b) litter and woody debris on ground surface under canopy, c) inter-tree areas dominated by bare ground with scattered Triodia hummocks, and d) bare surface of open patch.

For the litter, and litter + shade treatments, leaf, stick, and bark material was collected from an equivalent sized area (4 m2) from under the canopy of a mallee tree at least 50 m away from any other plot. This material was spread evenly over the surface of the plot, to achieve an even cover of litter (Figure

3.2a,b). A fence constructed of coarse wire (5mm high, horizontal gap width of approximately 75mm) was placed around all artificially created plots to prevent the loss of litter from the plot. Litter was defaunated upon plot construction by heavy spraying of the debris using a pyrethrum-based insecticide (Pyzap, 40 g

L-1 Pythrethrins, 160 g L-1 piperonyl butoxide, Agserv-Ruddock Agriculture). For

3. Mallee trees as landscape modulators

the shade and litter + shade treatments, the effect of tree canopy shade was simulated by using a green shade cloth. To create these ‘shade tents’, two 0.5 m steel poles were placed directly opposite each other, and the cloth draped over the top of the poles onto an eyelet. The four corners of the cloth were then pegged into the ground to form a ‘tent’. This held the cloth roughly 0.45 m above the ground in the centre, and approximately 0.1 m off the ground at the corners (Figure 3.2b,c). The shade cloth used was designed to reduce UV penetration by 50%. On days where the ambient temperature was 40 - 45°C, and the ground surface 55 - 60°C, this shade cloth reduced the temperature on the ground surface by up to 15°C - conditions that are similar to those under mallee canopies (A. Kwok, personal observation). All plots were established in

January 2010, one year prior to sampling, and were checked once every 4 months between construction and sampling to ensure that all structures were intact. The levels of leaf litter within plots did not appear to vary over the course of the study.

a) b)

Figure 3.2: Photos of artificially created patch types, showing a) litter only, b) litter + shade, and c) shade only patches.

3. Mallee trees as landscape modulators

Arthropod sampling

Ground-dwelling arthropods were sampled in each plot using pitfall traps in

January 2011, one year after the plots were established. These traps were plastic cups (70mm diameter by 70mm deep) filled with a small amount of propylene glycol and buried below the litter and woody debris layer and flush with the soil. Five cups were placed in each plot: one near each corner (but at least 0.35 m from the edge of the plot and one in the centre of the plot. Traps were left open for seven consecutive nights. At the end of sampling the five traps for each plot were combined, and are not distinguished further. We acknowledge that pitfall sampling favours mobile and active fauna, and that their captures can be affected by habitat structure (e.g. Melbourne, 1999;

Driscoll, 2010). This may have resulted in an under-representation of the litter fauna and/or over-representation of some fauna, however, we did this to ensure standardised methods were used across all treatments. Arthropod samples were sorted to order. Within the Hymenoptera, ants and wasps were analysed separately. Similarly, in the case of spiders we analysed jumping spiders

(Salticidae) separately. For simplicity, we refer to each of these arthropod groups as a separate taxon. Spiders were further sorted to morphospecies for assemblage composition.

Statistical analyses

Generalised Linear Models (GLM) were used to test for differences among treatments (canopy, open, litter, shade, litter + shade) in the total number of arthropods, non-ant arthropods, each arthropod taxon (beetles, isopods, spiders, jumping spiders, silverfish, or wasps) and detritivores (isopods,

3. Mallee trees as landscape modulators

cockroaches, and silverfish). Each taxon was analysed in a separate model. All models used a negative binomial distribution, with the exception of beetles

(which used a Gaussian distribution). The choice of distribution was based on visual analysis of the residual graphs (following Zuur et al., 2009). All GLM analyses were conducted in the program R (R Development Core Team, 2011) using the MASS package (Venables and Ripley, 2002). Significant main effects were further tested using pairwise post hoc Tukey's tests.

A one-way Permutational Multivariate Analysis of Variance (PERMANOVA,

Anderson and Gorley, 2008) was used to test for differences in the arthropod community composition among the five patch types. The arthropod community included all broad arthropod orders (described above). Significant main effects were further tested using pairwise t-tests. Canonical Analysis of Principal

Coordinates was used to display patterns of community composition within each treatment. The closer any given points are to each other in the resulting ordination, the greater their similarity in terms of species composition. To examine the magnitude of the differences in community composition among treatments, we compared dissimilarity values based on Similarity of

Percentages (SIMPER). These values were based on a Bray–Curtis similarity matrix, with increasing dissimilarity values representing increasing dissimilarity between two treatments. All multivariate analyses were conducted within the

PRIMER (Clarke and Gorley, 2006) + PERMANOVA (Anderson and Gorley,

2008) statistical package.

3. Mallee trees as landscape modulators

Indicator Species Analysis (ISA) within the program PC-Ord (McCune and

Mefford, 1999) was used to determine the nature of taxon- and species-specific responses to patch type. Indicator Species Analysis calculates an Indicator

Value (IndVal), which aims to quantify the concentration of a species or taxon to any given group (e.g. treatment) (Dufrene and Legendre, 1997). The statistical significance of this procedure is tested using a randomisation procedure (n =

1000).

RESULTS

Effect of patch type on arthropod abundance

Patch type had variable effects on the number of arthropods. The total number of arthropods was significantly greater in open patches than in those patches containing both litter and shade (i.e. canopy, litter + shade patches) (Table 3.1,

Figure 3.3a). This was largely due to the high abundance of ants in each of the open samples. Accordingly, ant abundance increased with decreasing patch complexity (Table 3.1, Figure 3.3b). Ant abundance was also greater in the shade patches than in litter + shade patches (Table 3.1, Figure 3.3b).

In contrast to ants, several taxa were markedly more abundant in canopy or litter patches. The total number of non-ant arthropods was greater in the canopy patch than any of the other patches, although only differences with the shade and open patches were statistically significant (Table 3.1, Figure 3.3c). When these data were analysed separately for each taxon, isopods, spiders and wasps were on average two to eight-times more abundant in the canopy than all other patches (except for litter, in the case of wasps) (Table 3.1, Figure

3. Mallee trees as landscape modulators

3.3d,e,f). Furthermore, the abundance was similar across all other treatments of each of these taxa. Beetles generally responded positively to the addition of litter, and had a higher abundance in both litter and litter + shade patches compared to the shade and open patches (Table 3.1, Figure 3.3g). The number of silverfish, jumping spiders and the total number of detritivores (silverfish, cockroaches, and isopods combined) was similar across all patches (Table 3.1,

Figure 3.3h-j), though there was some support for the notion of a positive relationship between abundance and patch complexity the total number of detritivores (Table 3.1, Figure 3.3i,j).

Indicator Species Analysis showed that many of these taxa had moderate, though highly statistically significant, associations with canopy patches

(isopods, IndVal = 32.0%, P = 0.02; spiders, IndVal = 29.0%, P < 0.001; wasps,

IndVal = 41.5%, P = 0.01). A high abundance of ants, by contrast, was more indicative of open patches (IndVal = 32.8%, P = 0.02).

Effect of patch type on arthropod community structure

There were significant differences in arthropod community composition among patch types (PERMANOVA: Pseudo-F = 2.07, P < 0.001; Figure 3.4), primarily between the naturally occurring open or canopy patches and the manipulated patches (litter only, litter + shade, shade only) (Table 3.2). Additionally, the arthropod community in the canopy patch was substantially different to that in the open patch, driven by more isopods and wasps in the former patch (see above; Table 3.2, Figure 3.4). SIMPER analyses revealed that differences among treatments in arthropod community structure resulted from subtle

3. Mallee trees as landscape modulators

changes in the densities of each taxon. The only consistent pattern was that wasp abundance accounted for an average of 18% of the difference between the canopy and every other patch type. Patches did not increase in dissimilarity along the gradient of patch complexity. The dissimilarity among patch types ranged from 20.17 (shade vs. open) to 25.58 (canopy vs. shade).

Effect of patch type on spider species assemblages

Most spider species were found in low numbers at only one or two sites, and this prevented formal analyses of species composition. However, Indicator

Species Analysis of the most commonly sampled spider morphospecies suggests that preferences exist for different patch types. The abundance of

Zodariidae species 1 was significantly higher in the canopy than in any other patch type (n = 22, IndVal = 54.1%, P < 0.001), as was Lycosidae species 1 (n

= 20, IndVal = 33.0%, P = 0.02). Zodariidae species 2 was most commonly sampled species in the litter + shade patch (n = 57, IndVal = 33.3%, P = 0.03), and was found in every patch except the canopy.

3. Mallee trees as landscape modulators

Table 3.1: Summary statistics for generalised linear model analyses on arthropod abundance in five patch types. For all models, patch d.f. = 4, null residual d.f. = 49, and patch residual d.f. = 45. Results of pairwise comparisons for significant analyses are shown in Figure 3.3.

Deviance Residual deviance P

Total arthropods Null 73.606 Patch 20.843 52.764 <0.001

Total non-ant arthropods Null 70.477 Patch 21.654 48.823 <0.001

Total detritivores Null 58.649 Patch 5.6648 52.984 0.230

Ants Null 80.584 Patch 28.743 51.841 <0.001

Beetles Null 69.504 Patch 21.510 47.993 <0.001

Isopods Null 39.497 Patch 18.249 21.247 <0.001

Spiders Null 69.011 Patch 20.238 48.773 <0.001

Jumping spiders Null 60.608 Patch 4.761 55.847 0.310

Silverfish Null 59.916 Patch 2.138 57.779 0.710

Wasps Null 69.392 Patch 15.160 54.232 < 0.01

3. Mallee trees as landscape modulators

Figure 3.3: Mean ± S.E. number of a) total arthropods, b) ants, c) total non-ants, d) isopods, e) spiders, f) wasps, g) beetles, h) silverfish, i) jumping spiders and j) detritivores across two natural patches (canopy, open) and three manipulated patches (litter, litter + shade, shade). Different letters indicate a significant difference using Analysis of Deviance and Tukey's test for pairwise comparisons at P < 0.05. Treatments: C: Canopy, LS: Litter + shade, LIT: Litter only, SH: Shade only, OP: Open patches. Note different scale of Y-axis for each graph. S.E. is the standard error of the mean.

3. Mallee trees as landscape modulators

Table 3.2: t-test results for pairwise comparisons of arthropod community composition across four treatments.

Canopy Litter + shade Litter Shade Open Canopy - - - - - Litter + shade 1.65 * - - - - Litter 1.28 0.86 - - - Shade 1.53 * 1.20 1.08 - - Open 2.00 ** 1.67* 1.78 * 0.78 -

* : statistically significant at P < 0.05 ** : statistically significant at P < 0.01

Figure 3.4: Canonical Analysis of Principal Coordinates of arthropod community composition across two natural patches (canopy, open) and three manipulated patches (litter, litter + shade, shade). Vectors indicate Pearson moment correlation for each arthropod taxon, pointing to increases in the abundance of that taxon. The closer any given points are to each other, the greater their similarity in terms of community composition.

3. Mallee trees as landscape modulators

DISCUSSION

In this study we found that natural 'modulated' patches beneath the canopy of mallee trees supported a distinctly different ground-dwelling arthropod community than the inter-tree (open, or 'unmodulated') patches that are dominated by bare soil with perennial grass hummocks. The arthropod community in the canopy patch is characterised by higher numbers of isopods, wasps, and spiders, while open patches tended to be characterised by many more ants. Thus, these taxa appear to be patch sensitive, responding to the resource contrast between the two patch types (Kotliar and Wiens, 1990). Few taxa showed differences in abundance across artificially created patches, with the exception of beetles. Silverfish were the only taxon to show no differences between any patch types. Overall, these results are consistent with our predictions that canopy patches support a different arthropod community than open patches and artificially created patches containing leaf litter, shade, or both elements. Our study demonstrates that in a semi-arid mallee woodland, the dominant woody species (mallee trees, Eucalyptus spp.) are landscape modulators (sensu Shachak et al., 2008), structuring arthropod assemblages through the creation of distinct patches.

Differences in the arthropod community between the canopy and open patches most likely result from the unique local environmental conditions created by the landscape modulators (mallee trees). For example, mallee canopies capture considerable rainfall, resulting in the stem funnelling water towards the base of the tree (Nulsen et al., 1986). Soil moisture is therefore substantially higher in the area immediately around stems, and decreases with distance from the

3. Mallee trees as landscape modulators

stems (Nulsen et al., 1986). This, together with shade from the canopy and structural complexity on the ground surface (e.g. sticks, leaves, logs, bark) creates and reinforces a fertile island resource patch (e.g. Whitford et al., 1997), providing unique conditions and resources for arthropods within mallee communities.

Mallee trees are the primary landscape modulator in these ecosystems. Open areas are largely unmodulated by mallee trees in terms of the provision of litter and woody debris, resource accumulation, and microclimate amelioration. The soil in between trees is primarily exploited for nutrients and water, with little modulation (e.g. root growth, resource capture) to increase the levels of these resources. While scattered perennial grass hummocks dominate patches between mallee trees, they are compact, and shed relatively little leaf litter.

They are likely to modulate the soil directly beneath them. They do not, however, seem to modulate these open areas in the same way or to the same extent that mallee trees modulate the area immediately beneath their canopy, though no data currently exists to verify this. Thus, it seems that open patches are not affected by perennial vegetation in the same way as patches underneath mallee trees, and in this study they seem to support a lower diversity of arthropod taxa as a consequence. However, it should be noted that the perennial grass hummocks that exist within open patches represent an important component of these areas, but they were not specifically sampled for arthropods in this study. It is therefore possible that the hummocks themselves support a relatively rich and diverse arthropod fauna, different to that of the open areas and the patches beneath mallee. Irrespective of this, landscape

3. Mallee trees as landscape modulators

modulator theory (Shachak et al., 2008) provides a framework within which to view patch creation by woody species, and further predict the consequences of these patches on biotic communities as a whole.

Isopods, wasps, and spiders are more abundant in canopy patches

In our study, several arthropod taxa, particularly isopods, were clearly more abundant in canopy patches than all other patch types. Isopods, in particular, are highly concentrated beneath the canopy. All isopods are moisture dependent to some extent, burrowing to depths of 1 m to reach soil that is moist enough for survival (Shachak and Yair, 1984; Jones et al., 2006). This microclimate is more likely to exist under the trees, where the canopy increases soil moisture by funnelling rainfall (see above) and, together with litter and woody debris, reduces evaporation rate and temperature extremes (Zhang and

Zak, 1995; Shumway, 2000; Jones et al., 2006).

Many isopod species are also highly photonegative (Warburg, 1968), and are therefore likely to benefit from structural complexity that provides shelter (e.g. litter, canopy). In mallee, the shade of a tree can reduce surface temperatures from 65 ˚C to 50 ˚C, modulating the local ecological conditions by creating a cooler and moister environment in the canopy patch (A. Kwok, personal observation). Shade provided by macrohabitat elements, such as the canopy of perennial vegetation, directly affects the distribution of arthropods (e.g. Pearson and Lederhouse, 1987; Nakamura et al., 2009; Carpintero et al., 2011). For example, in a subtropical rainforest, the addition of shade to habitat patches increased the colonisation of rainforest-associated arthropods (Nakamura et al.,

3. Mallee trees as landscape modulators

2009). In arid shrublands, beetles move between shade and open patches in order to thermoregulate, and the preference for a particular patch type is driven by the branch and canopy density of shrubs in the landscape (Shelef and

Groner, 2011).

We also observed a greater abundance of spiders in the canopy patch. In addition to a cooler, more stable microclimate, litter provides spiders with greater structural complexity, which gives more nesting opportunities for silken retreats, a range of refugia for shelter, and complex environments for hunting

(Uetz, 1979; Bultman and Uetz, 1982; Castro and Wise, 2009). Species of arthropod that move within soil spaces or burrow in the soil may also benefit from increased soil porosity, which is greater under trees (Oliver et al., 2006).

The data we have for spider species suggests that they also exhibit clear preferences for particular patch types.

The canopy patch provides a wide variety of foraging resources for arthropods of multiple trophic levels. Leaf litter is a key source of organic material for microinvertebrates (e.g. Collembola, Acari, Psocoptera; Santos and Whitford,

1981; Kampichler and Bruckner, 2009) and macroinvertebrates (e.g. isopods,

Bastow, 2011; Vos et al., 2011). Several studies have reported a greater abundance of below and above–ground arthropod taxa under the canopy than in the open patches in arid ecosystems (e.g. Noble et al., 1996; Doblas-Miranda et al., 2009; Liu et al., 2011), as well as ants in agricultural areas (Oliver et al.,

2006). We suggest that mallee trees, through the input of leaf litter, support animals of multiple trophic levels including decomposers (mites, isopods,

3. Mallee trees as landscape modulators

cockroaches), omnivores (ants), and predators (ants, spiders, wasps). This food processing chain concentrates animal diversity spatially and temporally (e.g.

Bastow, 2011) and the activity of these animals contributes to the island of fertility created and maintained by the trees.

Ants are more abundant in open patches

We found twice as many ants in the open patches than the most structurally complex (canopy, litter + shade) patches. Other studies, however, have shown that ant species richness and abundance is substantially higher under trees than adjacent open habitats, with each patch also supporting a unique species assemblage (e.g. Andrew et al., 2000; Andersen et al., 2006; Oliver et al.,

2006). This has been attributed to greater leaf litter under the canopy, and a greater concentration of soil nutrients. In our study site, leaf litter is virtually restricted to the canopy patch, yet we still found a greater abundance of ants in open patches. Higher ant abundance in the open may be due to the thermophilic habit of many ant species (Hölldobler and Wilson, 1990), and also because more complex habitats impede locomotion and therefore reduce the energy efficiency at the scale of an individual ant (e.g. Kaspari and Weiser,

1999).

It is also possible that behaviourally dominant ants (e.g. the Dominant

Dolichoderinae, Andersen, 1995) favour, and therefore become most abundant in the open patches. For example, ant community structure and abundance can vary within a canopy patch, with dominant species monopolising the space closest to the base of trees, and subdominant species occurring further out

3. Mallee trees as landscape modulators

towards the canopy edges (e.g. Carpintero et al., 2011). In mallee communities, the canopy patch is likely to support a greater number of cryptic species due to the presence of leaf litter (Andersen, 1995), while hot, open patches may support a greater abundance, or at least activity of dominant taxa (e.g. Lindsay and Cunningham, 2009).

The composition of ground-dwelling arthropods can be heavily influenced by the sampling methods employed. Pitfall traps are known to favour mobile and active animals, and may be influenced by habitat structure (Melbourne, 1999). This can result in an over-representation of mobile fauna in samples, and an under- representation of cryptic fauna. This may partially explain the dominance of ants in open patches the present study, in which treatment plots varied substantially in habitat structure. It is interesting to note, however, that several other mobile taxa (e.g. spiders) were more abundant in the canopy patch. This suggests the typical bias exerted by pitfall traps does not always manifest. This also highlights that a comprehensive sampling regime using multiple methods (e.g. litter extraction, pitfall traps, etc.) should be employed when sampling ground- dwelling arthropods.

Artificially created structural complexity had variable effects on arthropods

We failed to detect a strong, predictable effect of our artificially created patches on the arthropod community. We had expected detritivores to increase with increasing structural complexity on the ground surface (i.e. with the addition of litter and woody debris and/or shade) and ants to show the opposite trend,

3. Mallee trees as landscape modulators

consistent with past research that illustrates how habitat complexity affects faunal assemblages (e.g. Tews et al., 2004; Nitterus and Gunnarsson, 2006).

Indeed, many experimental studies using litter and/or shade addition have confirmed the importance of these elements to arthropods, using experimental plots ranging from 0.1 m2 to 5.3 m2 (e.g. Bultman and Uetz, 1982; Mazia et al.,

2006; Castro and Wise, 2009). We used relatively large experimental plots in this study (4 m2), but we failed to detect any effects of structural complexity on the majority of the arthropod fauna, apart from differences between naturally occurring canopy and open patches. There were instances of differences in arthropod abundance between particular litter treatments. For example, beetles and Zodariidae species 2 attained significantly higher abundance in litter + shade treatments than open or shade only treatments. This was not, however, apparent for any other taxa. While we quantified structural complexity coarsely

(e.g. presence or absence of litter), structural complexity can also vary at finer scales. For example, the abundance of each species utilising the litter may be affected by the structure of individual leaves (e.g. curly vs straight, Bultman and

Uetz, 1982).

Our results indicate that small, artificially created patches do not mimic the canopy patch of mallee trees. This would suggest that small, naturally occurring patches of litter that do exist within open areas are unlikely to support a community similar to that of the canopy patch. This is likely the case because litter patches that exist in the open are not modulated by the tree (see above).

Additionally, they may support fewer arthropods as they are physically smaller than natural canopy patches, and therefore unable to sustain larger populations.

3. Mallee trees as landscape modulators

In our study area, patches of litter in open areas are generally less than 4 m2, substantially smaller in size and extent than natural litter patches under the trees. In the Chihuahuan desert the number of arthropods has been found to be positively correlated with the amount of surface litter (e.g. Santos et al., 1978), though this is not always the case in other ecosystems (Nakamura et al., 2009).

A temporal factor may also influence the arthropod community in isolated litter patches. Our artificial plots were established for one year prior to sampling. This may not be long enough for animals to disperse and increase in number within these small, isolated litter patches. This may be especially the case for poor dispersers and colonisers such as isopods (Paris, 1965). Our litter plots were also physically disturbed, with fresh and decayed material mixed to form one layer. Natural canopy patches have several distinct layers graduating from well decomposed fragments on the bottom, to relatively fresh litter on the surface.

The lack of layering in small, isolated litter patches may influence habitat suitability, particularly for moisture-dependent species such as isopods.

Spatially patchy distribution of biota in arid ecosystems

Resource patchiness can develop through landscape modulation, abiotic redistribution of resources, or ecosystem engineering by biotic species

(Shachak et al., 2008). Regardless of the method of patch formation, the effects of perennial vegetation on ecosystem function and soil properties are well established. For example, trees and shrubs are important for capturing sediments, organic material and water that is redistributed across the ground surface, preventing their loss from the ecosystem (Ellis et al., 2006). Growing

3. Mallee trees as landscape modulators

vegetation also influences soil porosity and increases water infiltration and soil- nutrient distribution (Titus et al., 2002). Vegetation thus concentrates resources spatially and temporally. This patchiness is reinforced over time through feedback loops that involve abiotic (e.g. wind and water redistribution of sediments) and biotic (e.g. increased animal activity within patch, nutrient inputs from fauna and flora, etc.) processes (e.g. Lightfoot and Whitford, 1987; Lavelle et al., 2006; Butterfield and Briggs, 2009).

Increasing evidence indicates that the concentration of resources affects the composition of entire biotic assemblages. In many ecosystems, resource patches created by vegetation influence the distribution of arthropods, including ants (Andrew et al., 2000), beetles (Koivula et al., 1999), spiders (Castro and

Wise, 2009), and invertebrates in general (Lemperiere and Marage, 2010). This knowledge has been extended to illustrate that patches created by woody species structure the composition and distribution of entire animal assemblages

(e.g. beetles, Barton et al., 2009; soil arthropods, Liu et al., 2011). The concentration of resources around perennial vegetation also has strong effects on plant communities, for example through facilitative and/or competitive mechanisms (e.g. Facelli and Brock, 2000). Thus, the distribution of woody species and the resource patches they create has ecosystem-wide effects on fauna and flora.

Conservation implications

Our study illustrates the impact of common landscape elements on arthropod distribution and community structure. It emphasises the need to sample the full

3. Mallee trees as landscape modulators

range of landscape elements (e.g. canopy, open patches) when attempting to document arthropod biodiversity, particularly in ecosystems where resources are patchily distributed, or limiting (e.g. semi-arid and arid ecosystems). Our results indicate that the loss or reduction of landscape elements may have unexpected, indeed, unknown ramifications for arthropod biodiversity. For example, fire generally destroys the above–ground component of multilayered resource patches, and this component may take years to attain the levels necessary for the persistence or colonisation of each taxa. Thus, frequent fire may result in the loss of patch-dependent taxa. Further research is needed to determine how fire affects taxa which depend on plant-created patches.

Conclusions

In this study we have shown how the resource patch created by a dominant tree affects the distribution of arthropod communities in a semi-arid mallee woodland. This effect is evident at a broad taxonomic scale (order) and for broad arthropod groups. The creation of resource patches by woody vegetation is a characteristic of resource limited ecosystems. The factors responsible for concentrating fauna into the resource patch – the canopy patch providing shade, ameliorated microclimate, and a broader range of habitat and foraging niches – are universal properties of perennial vegetation in arid ecosystems. In these ecosystems, woody species are likely to act as landscape modulators, affecting the community composition of ground-dwelling arthropods at broad and fine taxonomic scales.

3. Mallee trees as landscape modulators

ACKNOWLEDGEMENTS

We thank the Australian Wildlife Conservancy for logistical support, and

Samantha Travers and George Madani for field assistance. Samantha Travers and Yvonne Davila provided constructive comments on the research presented here.

4. Fire and arthropods

Chapter 4

Does fire affect the ground-dwelling arthropod community through changes to fine-scale resource patches?

Alan B. C. Kwok and David J. Eldridge

Keywords: arthropod, landscape modulator, fertile island, fire, mallee, patch, isopod, spider

4. Fire and arthropods

ABSTRACT

In arid and semi-arid ecosystems, resources are concentrated into patches at fine and broad spatial scales. In south-eastern Australia, mallee trees

(Eucalyptus spp.) create multilayered resource patches beneath their canopies by providing shade and accumulating woody and non-woody debris ('litter').

These patches are distinctly different from open, inter-tree areas which support scattered hummock grasses. Each patch type supports unique ground-dwelling arthropod communities. The purpose of this study was to examine whether fire affects the strength of modulation of the arthropod community. Using pitfall traps, we sampled ground-dwelling arthropods under the canopy of mallee trees, and in open areas largely devoid of vegetation. Under the canopy, traps were placed underneath leaf litter. Sampling was conducted in an area not burnt in 30 years ("unburnt") and an area burnt 4 years prior to sampling ("burnt").

These areas contrast in the development of the canopy and its associated resource patch, with the area burnt four years ago having a substantially smaller resource patch than the unburnt area. Five taxa (cockroaches, isopods, spiders, jumping spiders and wasps) were more abundant under the canopy than in the open patch, in both burnt and unburnt communities. Spider species richness also showed this pattern. Ants showed the opposite pattern, and were more abundant in open patches irrespective of fire. Beetles and silverfish were equally abundant in the canopy and open patches in both burnt and unburnt communities. Irrespective of patch type, silverfish, wasps, and isopods were more abundant in the unburnt than the burnt community. Ants were the only taxa more abundant in the burnt community. Burnt and unburnt communities supported a similar abundance of beetles, cockroaches and spiders. Overall,

4. Fire and arthropods

there were clear differences in arthropod community composition between patch types and fire history. These results demonstrate that many arthropod taxa are affected by the presence of the canopy and the resource (litter, shade) patch it creates on the ground surface. Furthermore, this is the case in an area recently burnt by fire where the resource patch is relatively undeveloped. Some taxa, however, showed clear associations with the canopy patch, and were in substantially lower numbers in the burnt community, suggesting a dependence upon older, more developed patches of vegetation at both fine and broad scales.

INTRODUCTION

The spatial patchiness of resources is a key characteristic of semi-arid and arid ecosystems worldwide (Aguiar and Sala, 1999; Ridolfi et al., 2008). In these ecosystems, resources such as water and critical soil nutrients (e.g. phosphorus, nitrogen) are concentrated under and around perennial vegetation

(e.g. trees, shrubs; Schade and Hobbie, 2005; Eldridge et al., 2011) or any object that forms a semi-permanent obstruction in the landscape (e.g. log mounds; Tongway et al., 1989). Over time these obstructions or patches become ‘islands of fertility’ that have higher levels of nutrients and water, as well as different soil properties (such as enhanced soil carbon, Schlesinger et al., 1996; Butterfield and Briggs, 2009) and ecological conditions (e.g. shade and litter regimes) compared with the surrounding matrix. Fertile patches are widely acknowledged to affect the distribution of plants, with the sub-canopy environment supporting a different assemblage of species compared with open inter-tree areas (i.e. facilitation, Callaway and Walker, 1997). Similarly, fertile

4. Fire and arthropods

patches become foci for animal activity (e.g. Dean et al., 1999). Over time the activity of these plants and animals further reinforces the fertility of the patch through nutrient inputs.

Landscape modulators are species that regulate critical resources (e.g. habitat, microclimate, nutrients, etc.) through the creation of multilayered resource

('modulated') patches (Shachak et al., 2008). Woody plants are landscape modulators as they affect nutrient and water distribution and ecological conditions in their immediate vicinity, thereby creating distinct landscape patches. Broadly, modulated patches are akin to ‘fertile islands’ (Garner and

Steinberger, 1989). They contain higher levels of soil nutrients, greater structural complexity (e.g. standing vegetation, litter, stick, bark) and a different microclimate compared to unmodulated patches. Unmodulated patches represent the surrounding open, or poorly vegetated matrix. Landscape modulator theory proposes that the resource contrast between these patches drives differences in the biotic communities utilising each patch type. A growing body of evidence indicates the importance of woody vegetation as modulators of plant (Agra and Ne'eman, 2009), fungal (Bennett et al., 2009), and arthropod communities (Chapter 3, this thesis).

A key tenet of landscape modulator theory is the biodiversity cycling hypothesis, which focuses on the growth and decay of landscape patches. This hypothesis predicts that resource contrast (and therefore the strength of landscape modulation) is lowest following any process that destroys a resource patch (e.g. plant mortality, land clearing, wildfire; Shachak et al., 2008). In the absence of

4. Fire and arthropods

disturbance, resource contrast between modulated and unmodulated patches increases over time (for example due to the growth of the tree canopy and litter accumulation), and is paralleled by increasing divergence of the biotic communities resident in each patch type. Ecological disturbance, such as fire, may destroy the modulator, and the modulated patch. This is predicted to reset the stage of patch development to "time zero", with little contrast between modulated and unmodulated patches (Shachak et al., 2008). New growth by woody plants is needed for the development of patches once again. Thus, differences in the composition of the biotic assemblages are predicted to be lowest immediately following disturbance and greatest at community maturity.

Disturbance is consequently a crucial process that modifies multilayered resource patches as well as the strength of modulation and its effects on the biota. It is also critical to the persistence of modulator-free patches at landscape scales. Thus, disturbance events are thought to play a major role in determining both local- and landscape-scale patterns of biodiversity (Shachak et al., 2008).

Mallee is a widespread vegetation community found across much of semi-arid eastern Australia. Mallee communities are dominated by multi-stemmed

Eucalyptus spp. which grow as large shrubs or small trees. Mallee often occurs in association with spinifex hummock grasses (Triodia spp.) that are scattered in the areas between trees (dune–mallee communities, hereafter referred to as mallee). Mallee trees create multilayered resource patches, comprising the biomass of the trees as well as a distinct ground and soil environment. The area beneath the canopy is characterised by high levels of litter as well as coarse and fine woody debris (~1.5kg m-2; Bradstock and Gill, 1993), while open, inter–

4. Fire and arthropods

patch areas between trees are largely devoid of this material (< 0.02kg m-2,

Jane Smith et al., unpublished data, 2012). Mallee trees are landscape modulators, as the patches below the canopy structure the distribution of ground-dwelling arthropod assemblages (Chapter 3, this thesis), supporting a greater abundance of several arthropod taxa (spiders, isopods, wasps) than open patches.

In Australia, fire is the dominant non-anthropogenic form of ecosystem disturbance (Morton et al., 2011), and this is the case in mallee and dune– mallee communities. Triodia spp. in particular are highly flammable, and these ecosystems have co-evolved with fire (Griffin, 1991; Allan and Southgate, 2002;

Greenville et al., 2009). During a fire, all of the above ground vegetation can be destroyed, but most adult mallee trees survive as they possess underground lignotubers from which stems re-sprout following the event (Noble, 1982). At landscape scales, entire vegetation communities and their fauna are intrinsically tied to fire (e.g. ants, Andersen, 1991; Andersen et al., 2007a; vertebrates,

Friend, 1993; Letnic et al., 2005; plants, Noble, 1982; Watson et al., 2009).

At finer scales (i.e. areas immediately around trees), fire affects landscape modulation by burning the mallee tree and destroying the resource patch of woody debris. The loss of these resources, together with changes to the ecological conditions of the patch (e.g. loss of shade), presumably weakens the modulating effect of the tree. Furthermore, the woody debris changes composition substantially in the decades following fire. For example, the proportion of large, intact leaves declines and is replaced by smaller, fragments

4. Fire and arthropods

of leaves, and the proportion of sticks in the litter bed increases, along with the total mass of litter and woody debris (Samantha Travers, unpublished data,

2012). These changes to the resource patch have the potential to affect resource patch-dependent taxa, as well as those taxa that are more typically associated with open patches. To our knowledge, however, there is little information on how fine-scale patch structure and landscape modulation is affected by fire, and how this influences the within and among-patch abundance and distribution of modulator-dependent fauna such as arthropods.

This study has two aims. Firstly, we examine the role of mallee trees as landscape modulators of the ground-dwelling arthropod community in the mallee–spinifex dunefields of eastern Australia. Secondly, we examine the role of a major ecological disturbance (fire) in mediating landscape modulation. We address these aims by comparing modulation of the ground-dwelling arthropod community at two points of the biodiversity cycling hypothesis (Shachak et al.,

2008): patch "maturity" (30 years post-fire, hereafter “unburnt”), and shortly following patch decay, (four years post-fire, hereafter “burnt”). We acknowledge that 30 years post-fire may not strictly be maturity within a mallee community.

This is, however, roughly the time at which these communities attain peak flammability (Bradstock, 1990), and when many aspects of the modulated and unmodulated patches stabilise (e.g. litter, canopy, Triodia and ground cover;

Haslem et al., 2011).

We have three hypotheses. Firstly, in the unburnt community, canopy patches will support a different composition of ground-active arthropods compared with

4. Fire and arthropods

open, unmodulated patches (i.e. there is modulation of arthropod community composition by mallee trees). Based on previous research (Chapter 3, this thesis) this will be driven by a greater abundance of modulator-associated taxa

(i.e. isopods, wasps, and spiders) in the canopy patches than the open patches, while taxa associated with open patches (ants) will show the opposite pattern.

Modulator-independent taxa (beetles and silverfish) will be in similar abundance in canopy and open patches. Secondly, due to the recent decay of the multilayered resource patch, there will be no differences in arthropod community composition between the canopy and open patches in the burnt community (sensu biodiversity cycling hypothesis). Lastly, we expect there will be differences in the composition of the arthropod assemblage of burnt and unburnt communities, with a lower abundance of modulator-associated taxa and a greater abundance of taxa associated with open patches in the burnt community.

METHODS

Study area

This study was conducted in January 2011 at the Australian Wildlife

Conservancy’s Scotia Sanctuary, which is located 150 km south of Broken Hill,

NSW, Australia (33°43’S, 143°02’E). The climate is characterised by low and variable rainfall (250mm: mean annual rainfall), high evapotranspiration

(~1500mm yr–1), hot summers (daily mean temperature: 30°C, daily maximum:

47.8°C, daily minimum: >15°C) and cool winters (daily mean: ≤17°C, daily maximum: 32.2°C, daily minimum: ≤6°C) (Australian Wildlife Conservancy,

2011).

4. Fire and arthropods

Vegetation community

This study was conducted in dune–mallee woodland. Dune–mallee communities within the study area are located on long, low (relief to 7 m) east- trending sandy dunes dominated by an overstorey of mallee trees (Eucalyptus dumosa A. Cunn. ex J. Oxley and E. socialis F. Muell. ex Miq.) scattered between 5 m and 50 m apart. Within inter-tree areas, scattered perennial hummock grasses (Triodia scariosa N.T. Burb) dominate the ground to low strata. The projected foliage cover of mallee trees and leaf litter in dune–mallee communities varies depending upon when the community was last burnt.

Canopy cover generally increases rapidly to between 20% and 30% until roughly 30 years post-fire, after which time it stabilises (Haslem et al., 2011).

Similarly, leaf litter and woody debris stabilises at roughly 30% approximately

25 years after fire. At our study site, the soils are mainly calcareous, brownish and siliceous sands.

We conducted this study in two areas that differed in their fire history. The burnt community was located in an area of mallee that was burnt by wildfire approximately four years prior to sampling. The unburnt community was located in an area of mallee that had not been burnt by fire in approximately 30 years.

These two areas were separated by a distance of 10 km, though they are both part of a large, contiguous patch of mallee vegetation. In both communities shrub cover to 2 m is sparse, with widely-spaced individuals of predominantly

Senna artemisioides subsp. filifolia Randell and petiolaris Randell, and Acacia burkittii F. Muell. ex Benth.

4. Fire and arthropods

The size and extent of the resource patch differed between the burnt and the unburnt community. In our study area, a typical mallee canopy is seven times larger at 30 years post-fire compared to four years post-fire (~145 m2 vs. 23 m2,

Samantha Travers, unpublished data, 2012; Figure 4.1). Furthermore, the litter bed is deeper in the unburnt community than the burnt community (~ 55mm at

30 years cf. 15mm at 4 years), with a greater volume (1.4 ± 0.1kg m-2 at 30 years cf. 0.9 ± 0.1kg m-2 at 4 years). The litter bed also extends further from the base of the tree (0.7 m at 4 years cf. 2.6 m at 30 years; Samantha Travers, unpublished data, 2012; Figure 4.1).

Sampling design

We conducted this study at 16 sites within mallee-dominant dunes. Eight sites were located in the burnt community, and eight were located in the unburnt community. All sites were separated by at least 500 m. At each site, two areas approximately 250 m2 were established approximately 100 m apart. Within each area arthropods were sampled from within two naturally occurring patch types: canopy, and open (Figure 4.1). Canopy patches were located in the area of litter and woody debris on the ground surface, around the base but within the dripline of mallee trees. Open patches were located in inter-tree areas dominated by T. scariosa, at least 10 m away from any tree canopy. Whilst open areas are not strictly patches in the typical sense of an area concentrating resources (sensu

Ludwig et al., 2004), we refer to these areas of the landscape as patches for brevity. Open patches are naturally devoid of litter and woody debris. We consider canopy patches to be 'modulated', as they are created and maintained by the landscape modulator (mallee trees). Open patches, in contrast, are

4. Fire and arthropods

relatively 'unmodulated'. We acknowledge that T. scariosa modulates its surroundings, though this is at a much finer spatial scale compared to mallee trees. In total we sampled 64 patches (2 patch types × 2 replicates × 8 sites × 2 fire histories).

a) b)

c) d)

Figure 4.1: Photos of typical study site mallee community, showing a) Unburnt resource patch under canopy, b) unburnt open patch, c) burnt patch under canopy and d) burnt open patch

Arthropod sampling

We sampled ground arthropods in each plot using pitfall traps. These traps were plastic cups (70mm diameter by 70mm deep) filled with a small amount of

4. Fire and arthropods

propylene glycol and buried under the litter and woody debris layer and flush with the ground surface. Five cups were placed in each plot. In both the burnt and the unburnt community, traps were placed within the leaf litter layer, and always within the dripline of the canopy. Traps were thus constrained by the visible spatial extent of the litter and woody debris patch. Traps were left open for seven consecutive nights. We acknowledge that pitfall sampling favours mobile and active fauna and may have resulted in an under-representation of the litter fauna, but we did this to ensure standardised methods were used across all treatments (see Chapter 3). Arthropod samples were sorted to order, but in the case of spiders, we analysed jumping spiders (Salticidae) separately.

Statistical analyses

For statistical analyses, to calculate the abundance we took the mean of the two replicates for each treatment configuration at each site. We used generalised linear models (GLM) to test for differences in the total number of ants, beetles, isopods, spiders, jumping spiders, silverfish, wasps, or the number of spider morphospecies between patch types (canopy, open), fire histories (burnt, unburnt), and their interaction. Each taxon was analysed in a separate model using a negative binomial distribution. The choice of distribution was based on visual analysis of the residual graphs (following Zuur et al., 2009). All GLM analyses were conducted in the program R (R Development Core Team, 2011) using the MASS package (Venables and Ripley, 2002).

We used a two-way Permutational Multivariate Analysis of Variance

(PERMANOVA, Anderson and Gorley, 2008) to test for differences in the

4. Fire and arthropods

arthropod community composition between the patch types, fire histories, and the interaction of these factors. The arthropod community included all broad arthropod orders (described above). Significant main effects were further tested using pairwise t-tests. We used Canonical Analysis of Principal Coordinates

(CAP) to display patterns of community composition within each treatment. To examine the magnitude of the differences in community composition among patch types and fire history, we compared dissimilarity values based on

Similarity of Percentages (SIMPER). These values were based on a Bray–

Curtis similarity matrix, with increasing dissimilarity values representing increasing dissimilarity among two treatments. All multivariate analyses were conducted within the PRIMER (Clarke and Gorley, 2006) + PERMANOVA

(Anderson and Gorley, 2008) statistical package.

RESULTS

Effect of patch type on arthropod abundance

Ants were the only taxa to be consistently more abundant in the open patch than the canopy patch, and this was consistent for both burnt and unburnt communities (Patch (P): P < 0.001; P × Fire history (FH) interaction: P >0.05;

Table 4.1; Figure 4.2a). Beetles and silverfish were equally abundant in the canopy and open patches in both burnt and unburnt communities (P: P >0.05; P

× FH interaction: P >0.05; Table 4.1; Figure 4.1b,c).

Five taxa were more abundant in the canopy patch than the open patch, in both burnt (B) and unburnt (UB) communities: cockroaches; isopods, spiders;

4. Fire and arthropods

jumping spiders and wasps (P: P < 0.05; P × FH interaction: P >0.05; Table 4.1,

Figure 4.2d-h).

Table 4.1: Summary statistics for generalised linear model analyses on abundance of eight arthropod taxa and spider species richness in canopy and open patches in burnt and unburnt communities. For all models, patch d.f. = 1, fire history d.f. = 1, patch × fire history d.f. = 1 residual d.f. = 28.

Deviance Residual deviance P Ants Null 79.967 Patch 24.981 54.985 <0.001 Fire history 20.788 34.197 <0.001 Patch x Fire history 0.634 33.562 0.426

Beetles Null 37.738 Patch 3.254 34.484 0.071 Fire history 0.397 34.087 0.528 Patch x Fire history 1.327 32.759 0.249

Cockroaches Null 24.804 Patch 6.916 17.888 0.019 Fire history 2.226 15.661 0.465 Patch x Fire history 0.233 15.428 0.628

Isopods Null 34.585 Patch 7.096 27.489 <0.001 Fire history 9.553 17.936 <0.001 Patch x Fire history 0.269 17.667 0.604

4. Fire and arthropods

Silverfish Null 65.392 Patch 0.055 65.337 0.815 Fire history 31.943 33.394 <0.001 Patch x Fire history 3.045 30.349 0.081

Spiders Null 45.905 Patch 17.823 28.081 <0.001 Fire history 1.774 26.307 0.183 Patch x Fire history 0.337 25.970 0.562

Jumping spiders Null 33.142 Patch 7.587 25.555 <0.001 Fire history 0.668 24.887 0.413 Patch x Fire history 0.069 24.819 0.793

Number of spider species Null 40.738 Patch 13.247 27.491 0.002 Fire history 3.252 24.240 0.431 Patch x Fire history 0.185 24.055 0.667

Wasps Null 55.276 Patch 17.748 37.527 <0.001 Fire history 7.225 30.302 <0.001 Patch x Fire history 0.367 29.934 0.544

4. Fire and arthropods

Effect of fire on arthropod abundance

Three taxa were more abundant in the unburnt community than the burnt community: silverfish (UB: 3.13 ± 0.53 cf. B: 0.47 ± 0.12; P < 0.001), isopods

(UB: 0.81 ± 0.37 cf. B: 0.03 ± 0.03; P < 0.001) and wasps (UB: 4.22 ± 1.03 cf.

B: 2.00 ± 0.40; P < 0.001) (Table 4.1, Figure 4.2c,e,h). In each case this was independent of patch type (P × FH interaction: P >0.05, Table 4.1, Figure

4.2c,e,h).

Burnt and unburnt communities supported a similar abundance of beetles (B:

4.16 ± 0.89 cf. UB: 3.53 ± 0.55), cockroaches (B: 0.56 ± 0.12 cf. UB: 0.38 ±

0.13), spiders (B: 10.06 ± 1.14 cf. UB: 11.59 ± 0.83) and jumping spiders (B:

1.28 ± 0.24 cf. UB: 1.59 ± 0.41), irrespective of patch type (P: P >0.05; P x FH interaction: P >0.05) (Table 4.1, Figure 4.2b, d, f, g). Ants were the only taxa more abundant in the burnt community than the unburnt community (B: 670 ±

189 cf. UB: 245.30 ± 33.33; B: P < 0.001).

Effect of patch type and fire on arthropod community composition

We found significant differences in arthropod community composition between canopy and open patches (Pseudo-F = 16.72, P < 0.001, Figure 4.3). This was consistent in the burnt and unburnt communities (P × FH interaction: Pseudo-F

= 1.37, P = 0.23, Figure 4.3), although the difference was stronger in the unburnt treatment (Canopy × Open pairwise t = 3.24, P < 0.001) compared with the burnt treatment (Canopy × Open pairwise t = 2.38, P = 0.01). Overall,

PERMANOVA analyses also indicated differences in the arthropod community composition between the burnt and the unburnt community (Pseudo-F = 8.99, P

4. Fire and arthropods

< 0.001, Figure 4.3). Furthermore, patch type explained 31% of the total variation explained by the model, while fire history explained 25% (total variation explained by all factors = 43%).

Similarity of Percentages (SIMPER) analyses indicated that differences among patch types in arthropod community composition were mainly due to small changes in the abundance of each taxa. The dissimilarity among the four patch types (burnt canopy, burnt open, unburnt canopy, unburnt open) ranged in value from 20.13 to 37.31. The greatest dissimilarity in community composition was between the burnt open and unburnt canopy patches, which represent the least structurally complex to the most structurally complex patches. Sixty-one percent of the difference between these two patch types was driven by three- times as many ants in the burnt open patch, while 8% of the difference was due to four-times as many wasps in the unburnt canopy patch. The most similar patch types were the burnt canopy and the unburnt open. SIMPER analyses indicate that ants accounted for 54.5 ± 3.8 % of the difference between any two patch types, however this is likely because ants were the most numerically dominant taxa. Canonical Analysis of Principal Coordinates confirmed that most taxa showed differences in abundance among patch types and in relation to fire

(Figure 4.3).

4. Fire and arthropods

Figure 4.3: Canonical Analysis of Principal Coordinates of arthropod community composition across two patches (canopy, open) and in two communities that differ in fire history (unburnt, and burnt). Vectors indicate Pearson moment correlation for each arthropod taxon, pointing to increases in the abundance of that taxon.

Spider morphospecies assemblages

In addition to spider abundance, the number of spider species was also greater in canopy than in open patches, and this was consistent for the burnt and unburnt community (P: P < 0.05; FH: P >0.05; P × FH interaction: P >0.05;

Table 4.1, Figure 4.4). Overall, the burnt and unburnt communities had a similar number of spider species (B: 7.44 ± 0.77 cf. UB: 9.13 ± 0.67; P >0.05; Table

4.1).

Most spider species were found in low numbers at only one or two sites, and this prevented formal analyses of species composition. Based on analyses of

4. Fire and arthropods

the most abundant species, however, there was evidence of preference for patch types, and effects of fire. For example, the most abundant taxon was a species of Zodariidae, which was found predominantly in the unburnt community (B: n = 6 cf. UB: n = 34). Furthermore, within the unburnt community this species was found only in the open patch (n = 34). A second species of

Zodariidae was roughly three times more abundant in the unburnt community (n

= 24 cf. n = 7), and in both unburnt and burnt communities was only found in the canopy patch. A species of Lycosidae was the third most common species, and this species appeared equally abundant in burnt and unburnt communities (B: n

= 10 cf. UB: n =11), but in each community the majority of individuals were found in the canopy patch (n = 7 and n = 9 for the burnt and unburnt community respectively).

Figure 4.4: Mean ± S.E. number of spider morphospecies across two patch types (canopy and open) and in two communities that differ in fire history (burnt and unburnt).

4. Fire and arthropods

DISCUSSION

In this study we found that mallee trees affect the structure and composition of the ground-dwelling arthropod community, and that fire plays a key role in this process. This study has three main outcomes. Firstly, at fine-scales, mallee trees create distinct multilayered patches that support a different arthropod community than open inter-tree patches. Secondly, for several taxa modulation occurred in both the recently burnt and unburnt communities. Thirdly, burnt and unburnt communities supported different arthropod communities, with some taxa virtually absent from the burnt community. Thus, we found that each arthropod taxon has a different level of dependence on the resource patch, and therefore fire. Our study indicates that in a mallee woodland, the structure of arthropod communities is influenced by the fine-scale distribution of resource patches, and this is intrinsically tied to a broad-scale ecological disturbance

(fire).

In this study we used space-for-time substitution to investigate whether modulation is affected by fire. Thus, it is important to note that we only sampled one burnt area and one unburnt area, meaning that fire history is pseudoreplicated ( Hurlburt, 1984). Consequently it is possible that the differences we observed in the arthropod community between burnt and unburnt areas were due to intrinsic differences between the areas that pre- existed the fire, rather than the fire itself. As fire is known to destroy fine-scale patchiness (Haslem et al., 2011), and indeed because of the effects of mallee trees elucidated previously in this thesis (Chapter 3), we strongly suspect that differences we observed were a consequence of the fire event. Furthermore,

4. Fire and arthropods

the "burnt" and "unburnt" communities were only 10 km apart, and were part of a large, contiguous area of mallee vegetation, likely supporting similar arthropod communities and subject to the same patch creation, destruction and colonisation processes irrespective of fire. However, we acknowledge the limitations of our sampling design, and this should be taken into consideration when interpreting our results and explanations.

Mallee trees modulate the ground-dwelling arthropod fauna

In this study, mallee trees affected the arthropod community in both burnt and unburnt areas through their multilayered resource patch. This indicates that even a relatively poorly developed resource patch, such as that present in a recently burnt area (discussed below), can affect the distribution of arthropods at relatively fine spatial scales. This was evident at the ordinal and, in the case of spiders, species level. In arid and semi-arid ecosystems, fertile patches around perennial vegetation are known to support greater animal activity (e.g.

Dean et al., 1999) and a different arthropod assemblage than the relatively infertile areas between trees (e.g. Noble et al., 1996; Andrew et al., 2000;

Andersen et al., 2006). Our study indicates that resource patches created by trees can structure entire biotic assemblages at an ordinal, and likely species level.

Differences in the arthropod community between canopy and open patches are likely the result of the unique resources and environmental conditions provided by the mallee patch within this system. The patch beneath the canopy, created by perennial vegetation, is cooler and moister than open, unvegetated patches

4. Fire and arthropods

(Weltzin and Coughenour, 1990; Moro et al., 1997; Shumway, 2000), creating a less extreme, less variable microclimate. These conditions are favoured by many arthropods (e.g. Pearson and Lederhouse, 1987; Chikoski et al., 2006;

Nakamura et al., 2009). In addition, litter and woody debris in canopy patches provides an abundance of food and shelter for decomposers, including microarthropods (e.g. mites) and isopods (Santos et al., 1978; Bastow, 2011), and consequently a range of predators (e.g. spiders).

Modulation of the arthropod assemblage in the burnt community was unexpected given the small extent of the resource patch. As noted earlier, the canopy patch is substantially larger in the unburnt than the burnt community.

Despite this, our data suggest that the canopy patch in the burnt community supports higher populations of some taxa, and overall a richer arthropod community than open patches. This may be the case as the burnt canopy patch still contrasts markedly with the resource-deprived open areas, and provides sufficient resources and habitat to sustain a greater number of arthropods.

The spatial concentration of resources is known to affect the distribution of a variety of biota at fine patch scales. For example, ant species composition varies between canopy patches and unvegetated open areas in semi-arid ecosystems (e.g. Oliver et al., 2006; Carpintero et al., 2011). Similarly, patches created by vegetation are known to support unique microbial (Smith et al.,

1994), fungal (Bennett et al., 2009), and floral (e.g. Agra and Ne'eman, 2009) communities. Thus, in resource-limited environments, perennial vegetation influences the distribution of resources, which affects the distribution and

4. Fire and arthropods

composition of a broad range of taxa. Perennial vegetation can therefore be considered to be landscape modulators of the biota in these ecosystems (sensu

Shachak et al., 2008).

Arthropod community composition differs between burnt and unburnt areas

In this study we found that arthropod community composition differed between the burnt and unburnt communities. The unburnt community had fewer ants but significantly more wasps, isopods and silverfish than the burnt community. The abundance of spiders and beetles showed no difference between burnt and unburnt areas. Most arthropods are generally regarded to be fire resilient (e.g.

Andersen et al., 2005; Vasconcelos et al., 2008), with declines in abundance immediately following fire often quickly reversed (Abbott et al., 2003). These patterns, however, can be quite variable. For example, in spinifex (Triodia spp.) grasslands, spider abundance has been reported as being both stable

(Langlands et al., in press) and unstable (Langlands et al., 2006) in the first ten years following fire. Similarly, ant abundance also shows variable responses to fire (e.g. Andersen and Muller, 2000).

Isopods and silverfish were the taxa most affected by fire. Isopods, in particular, were virtually absent from burnt habitats. Reduced abundance of these taxa in burnt communities may be attributed to the loss of soil organic matter in the upper layers of soil following fire (Noble et al., 1990), and the loss of the canopy and surface litter, resulting in higher surface temperatures and lower moisture levels. These characteristics are crucial as both taxa are dependent on moisture

4. Fire and arthropods

and litter for a suitable microclimate and food (silverfish, Peters and Campbell,

1991; isopods, Paris, 1965; Shachak et al., 2008). Our results are consistent with studies indicating a decline in silverfish following fire (Abbott, 1984;

Andersen and Muller, 2000). Similarly, a decline in isopods following fire has been well documented in a variety of ecosystems, including savanna (Andersen and Muller, 2000), tall forests (Pitzalis et al., 2005; Fattorini, 2010), and alpine habitats (Moretti et al., 2004). Our results suggest that in mallee the relationship of isopods to fire is likely through changes to the multilayered resource patch.

As the development of the resource patch is a slow, cumulative process that takes many years (Samantha Travers, unpublished data, 2012), this may render isopods less resilient to fire than other arthropods.

The biodiversity cycling hypothesis in mallee woodlands

The biodiversity cycling hypothesis predicts that increasing resource contrast between patch types results in increasing divergence in the biotic communities they support (Shachak et al., 2008). We had mixed support for our hypothesis that modulation of the arthropod community would be non-existent in the burnt mallee community. When the entire arthropod community was considered, there were clear differences between patch types in both communities, despite the recent loss of the multilayered patch in the burnt community. This result indicates two things. Firstly, there may be substantial development of the modulated patch in the first few years after a fire, at least enough to sustain higher arthropod populations than open areas away from the trees. In this case, sampling needs to occur sooner after a fire in order to capture the absence of the modulated patch. Secondly, it is clear that arthropods are affected by

4. Fire and arthropods

changes that occur at fine and broad scales. Modulation, however, acts specifically on each taxon. For some taxa (e.g. spiders), there was no evidence that modulation was weaker in the burnt community, while for cockroaches modulation appeared stronger in the burnt community.

There was also evidence that modulation may be stronger in the unburnt community for some taxa. For example, the difference in the abundance between the canopy and open patches of beetles or wasps was greater in the unburnt community. Similarly, isopods showed preference for the canopy patch in both communities, and a seemingly strong dependence upon it. This suggests that the resource contrast between the two patch types may increase over time in the absence of fire, with subsequent effects on some arthropod taxa, in line with the biodiversity cycling hypothesis. At broader scales, we found a lower relative abundance of taxa preferring modulated patches (isopods) and higher abundance of taxa preferring unmodulated patches (ants) in the burnt community. It is therefore possible that arthropod communities may change predictably in relation to disturbance and patch decay in our system. However, data on arthropod communities along a chronosequence of time-since-fire does not currently exist.

The biota of mallee communities are intrinsically linked to fire. We provide evidence suggesting that the responses of some biota to fire are the consequence of changes to the availability of resource patches at finer scales

(e.g. around trees). Various other studies have also suggested that the development of fine-scale resource patches is important for mallee biota. For

4. Fire and arthropods

example, malleefowl (Leiopoa ocellata) are dependent upon litter for nest construction (Frith, 1962), and it takes several decades for enough litter to accumulate after a fire to allow optimal malleefowl breeding densities

(Benshemesh, 1989). Similarly, the litter-dependent Boulenger's skink (Morethia boulengeri) is most likely to occur 100 years post-fire (Haslem et al., 2011).

These examples suggest that biodiversity in mallee communities do cycle in response to disturbance and changes to resource patchiness. In a similar way, plant community composition is related to fire regimes (e.g. Bradstock and

Cohn, 2002). Increasing evidence indicates an intricate relationship between the biota and fire in mallee ecosystems, particularly in terms of resource patches (Haslem et al., 2011; Kelly et al., 2011).

Conservation implications

Mallee communities are used for a variety of purposes such as biodiversity conservation, agriculture and pastoralism. Patterns of post-fire vegetation structure are directly relevant to many of these activities. The value of long- unburnt stands of vegetation for biodiversity are becoming apparent (e.g.

Haslem et al., 2011; Kelly et al., 2011). These stands, however, have the potential to fuel larger fires, which are incompatible with many anthropogenic activities. Consequently fire management often centres around fire suppression and prescribed burning of vegetation (Sandell et al., 2006). While we currently have only a limited understanding of arthropod community structure in mallee environments, it seems likely that fire plays an integral role in determining local- and landscape-scale biodiversity. Management that seeks to prevent the longer-term development of the mallee resource patch is likely to affect ground-

100

4. Fire and arthropods

dwelling arthropod communities by favouring species that prefer open habitat or those who are not reliant upon the resources created by the patch.

Fire is a crucial component of biodiversity management worldwide (Bowman et al., 2009; Driscoll et al., 2010). Critical to conservation efforts is a knowledge of species responses to fire and an understanding of how the spatial and temporal arrangement of fires affect the biota (Driscoll et al., 2010). Animals may be affected by fire through its effects on fine-scale resources patches, however this is not known for the majority of taxa. Furthermore, crucial to fire management and conservation is the distinction between wildfires and prescribed burning.

These fires do not necessarily have the same effects on the abiotic and biotic environment, and may send biotic communities along different trajectories.

Determining when this occurs is an important step toward balancing fire management for multiple land uses.

ACKNOWLEDGEMENTS

We thank the Australian Wildlife Conservancy for logistical support. Samantha

Travers and George Madani for field assistance. Samantha Travers and

Yvonne Davila provided constructive comments on the research presented here.

101

5. Arthropods and landscape health

Chapter 5

Do landscape health indices reflect arthropod biodiversity

status in the eucalypt woodlands of eastern Australia?

Alan B. C. Kwok, David J. Eldridge, and Ian Oliver

Published in Austral Ecology, 36, 800 - 813

Keywords: arthropod, biodiversity assessment, biotic integrity, ecosystem health, indicator, landscape function analysis

102

5. Arthropods and landscape health

ABSTRACT

Ecosystem or landscape health indices are important tools for land managers.

In particular, strong predictable relationships between ecosystem or landscape health indices and biotic diversity are often generalised, though seldom validated. Here we use data from a semi-arid eastern Australian woodland to examine the relationships between arthropod community structure and two sets of landscape health indicators; Landscape Function Analysis (LFA), and a

Terrestrial Index of Ecological Integrity based on common vegetation metrics

(structure, composition, and function; SCF). Hierarchical partitioning revealed that the ability of LFA or SCF to account for variation in arthropod richness was low, with the variable of importance taxon-dependent. Similarly, multivariate analyses indicated relatively weak and inconsistent relationships between LFA and SCF indices, and arthropod assemblage structure. Results obtained for additional habitat attributes commonly used in terrestrial vegetation monitoring were similar. Our study indicates that strong predictable relationships are rarely apparent, particularly for arthropods. This indicates that these indices have limited use as surrogates of arthropod biodiversity. These results are contrary to the past literature, highlighting the need for additional research and the development of a conceptual and empirical framework linking health indices and arthropod biodiversity. This is necessary to further the theoretical and practical application of these measurements in environmental management.

103

5. Arthropods and landscape health

INTRODUCTION

Climate change, overgrazing and overexploitation of natural resources are recognized as major threats to biodiversity worldwide (e.g. Thomas et al., 2004;

Ehrlich and Pringle, 2008). These threats have resulted in predictions of biodiversity decline at unprecedented levels (Pimm and Raven, 2000).

Consequently, the conservation of biodiversity has become a crucial part of environmental management across all forms of land tenure from private landholdings to national governance. There are still considerable gaps, however, in our understanding of how human practices and anthropogenic disturbances affect basic ecological processes, and consequently, biodiversity

(e.g. Stutchbury, 2007). Crucial to the management and conservation of biodiversity are rapid, cost-effective and scientifically-defensible methods for assessing how anthropogenic practices affect ecosystem health and condition, and the implications for biodiversity (Western, 1992).

A number of approaches and methodologies have been developed to assess landscape health. These include direct measurements of the biophysical environment, which are designed to reflect overall landscape health (e.g. vegetation structure and diversity, Gibbons et al., 2008), and measures of ecosystem function (Kienast et al., 2009; Rowe et al., 2009). In particular, indices of biotic integrity (Karr, 1991) have been used extensively to monitor change in ecosystem condition due to anthropogenic influences, particularly in aquatic systems (e.g. Karr, 1991; Andreasen et al., 2001; Klemm et al., 2003;

Martinez-Crego et al., 2010). Terrestrial indices of ecological integrity (TIEI) have also gained increasing popularity based on similar principles. Ecosystem

104

5. Arthropods and landscape health

characteristics such as the composition of the biota (including exotic species), the structure of the habitat (e.g. patch characteristics), and measures of ecosystem function (e.g. decomposition, erosion, infiltration etc.) are key inclusions in any TIEI (Noss, 1990; Andreasen et al., 2001; Oliver, 2002). These indices are also useful for broad terrestrial vegetation (Noss, 1990; Gibbons and

Freudenberger, 2006; Oliver et al., 2007; Liira and Kohv, 2010) and biodiversity monitoring (Failing and Gregory, 2003; O'Conner and Kuyler, 2009). Indices based purely on the biotic attributes (e.g. animal populations) have also been useful as they can be highly correlated with ecosystem condition (arthropods,

Karr and Kimberling, 2003; birds, Bryce, 2006).

Substantial developments have also been made over the past two decades in methodologies to assess the health and condition of semi-arid and arid environments (‘rangelands’ e.g. Pyke et al., 2002; Herrick et al., 2005; Watson et al., 2007). While rangeland ‘health’ (or condition) has traditionally been viewed in the context of pastoralism, i.e. healthier landscapes produce more plant biomass for livestock grazing (Wilson et al., 1984), there has been increased acceptance that functional integrity is a more appropriate way to view the health of rangelands. Functional integrity has been defined in many ways, but broadly it is the ability of landscapes to capture, retain and use critical resources such as water and nutrients (Ludwig et al., 2004). This concept is also closely tied to the ability of landscapes to resist stress (stability or resistance; Holling, 1986) or recover from stress (resilience, sensu Mageau et al., 1995), both of which are related to resource retention and production.

105

5. Arthropods and landscape health

An increasingly popular field-based monitoring procedure that is used to assess the functional integrity of rangeland ecosystems worldwide is Landscape

Function Analysis (LFA; Ludwig et al., 2004). Landscape Function Analysis has been widely adopted in a range of environments including Australasia (Ludwig and Tongway, 1993; Watson et al., 2007), the Iberian Peninsula (Maestre and

Cortina, 2004) and the Middle East (e.g. Ata Rezaei et al., 2006). The procedure comprises a suite of measurements that quantify the spatial arrangement and characteristics of resource patches in the landscape. These patches (fertile islands, sensu Garner and Steinberger, 1989) comprise grass tussocks, logs, shrub and tree hummocks that are known to capture resources such as seed, water, nutrients and organic matter, and are sites of maximum resource retention, productivity and biotic diversity. Their characteristics therefore reflect the functional integrity, or more broadly, the health of a site.

The increasingly widespread adoption of ecosystem health monitoring has stimulated research on the relationships between health and biodiversity, as well as the development of indicators and surrogates. While a positive relationship between ecosystem health and biodiversity is theoretically possible, it has not been confirmed nor quantified widely. The LFA methodology was never intended as an index of the biodiversity value of landscapes. Although only limited research to date has identified reasonable relationships between

LFA and indicators of biodiversity (e.g. Ludwig et al. 1999; 2004), the procedure is used often by a range of agencies to monitor the functional integrity of the landscape and give a broad indication of potential biodiversity value, particularly in Australia (e.g. www.lmdcma.com.au). Similarly, despite the fact that

106

5. Arthropods and landscape health

structure–composition–function indices are highly regarded as potential biodiversity surrogates (Oliver et al., 2007), their relationship to biodiversity has not been validated in rangeland systems, nor has the use of a TIEI. Indeed, a predictable empirical relationship between the two components (health and biodiversity) needs to be established prior to the widespread development or use of a surrogate or indicator (Smyth et al., 2009).

Here we describe a study that examines the relationships between arthropod community structure and two sets of landscape health indicators using data from a semi-arid eastern Australian woodland. We used two sets of landscape health indices; LFA and a terrestrial index of ecological integrity based on common structure, composition, and function metrics. We investigated arthropods because they are the dominant group of invertebrates in terrestrial ecosystems, are relatively abundant and species-rich in semi-arid environments, and are recognised as useful in monitoring biological systems

(Kremen, 1992; Andersen et al., 2004). Our research described here does not seek to operationalise these indices as surrogates or indicators of biodiversity, but rather is broadly aimed at improving our understanding of the relationships between landscape health and ground-active arthropod assemblages in rangeland systems. To achieve this aim we ask two specific questions: 1) how much variation in richness and composition of arthropod assemblages can be accounted for by indices of landscape health; and 2) are single habitat attributes able to account for greater variation in arthropod richness and composition than landscape health multimetrics? Overall, our study aims to improve the

107

5. Arthropods and landscape health

management of semi-arid rangelands by providing quantitative insights into patterns of arthropod biodiversity and how they relate to landscape health.

METHODS

Study area

Data used in this study were collected from the north-west floodplains of New

South Wales within the Darling Riverine Plains ecoregion (Thackway and

Creswell, 1995), across an area spanning approximately 2500 km2. The study area was is described by the 1:50,000 topographic maps “Burren Junction” and

“Pilliga” (148o30’–149o00’ E and 30o00’–30o30’ S). The climate is semi-arid, rainfall is summer-dominant and averages 475mm per year. Average temperatures are 20 – 35oC in January and 4 – 17oC in July. All sampling was conducted in February 2001.

The geology of the area is predominantly Quaternary alluvium comprising major and minor functional and non-functional plains and low-lying drainage depressions. Low dunes and coarse-textured elevated rises with relief to 5 m are superimposed on the plains. The dominant soils on the plains are coarsely cracking grey and brown clays (Grey and Brown Vertisols, Isbell, 1996). The low rises and dunes are dominated by hard-setting, red and brown duplex loams with coarse textured topsoils that grade to finer texture at depth (Red and

Brown Kandosols, Isbell, 1996). Native vegetation on the plains is dominated by an overstorey of coolibah (Eucalyptus coolibah), black box (E. largiflorens) and myall (Acacia pendula), grading to bimble box (E. populnea) and white cypress

108

5. Arthropods and landscape health

pine (Callitris glaucophylla) on the low rises and dunes. Large areas of native grassland, both natural and derived, are also a feature of the landscape.

Data were generated within 43 woodland remnant patches varying in size, shape, isolation and surrounding land-use, representing the diversity of remnant types that exist in the area. Thirty-four remnants were on private land and were grazed frequently by domestic stock, and nine were on travelling stock routes and grazed less frequently. The bimble box community was selected for the study because it provided a range of sites of varying condition at the landscape scale (i.e. severely degraded patches represented by a few trees, to large expanses of intact native vegetation). Bimble box is the dominant tree species over large areas of eastern Australia and the community has been extensively modified in the past by overgrazing, fire, clearing, weed proliferation and invasion of exotic plant and animal pests. Today it continues to be threatened by clearing and overgrazing.

Within each remnant a 50 m fixed transect formed the basis for all arthropod and landscape health sampling, and was located in an area considered representative of the remnant. Transects were placed centrally within small (<

10 ha) remnants or at least 100 m from the remnant edge. Transects were orientated from highest to lowest elevation along the maximum slope (generally

< 1%), consistent with LFA methodology (Tongway, 1995).

Landscape health assessment 1: Landscape Function Analysis

Landscape Function Analysis categorises two ecosystem components: 1) landscape organisation and 2) soil surface condition (SSC; Tongway, 1995) at

109

5. Arthropods and landscape health

relatively fine (<104 m2 scales). These two components reflect the ability of the landscape to capture and retain resources (i.e. the functional integrity). We acknowledge that ‘health’ is a highly value-laden and context-dependent concept (Wilson et al., 1984; Tongway and Ludwig, 1997). We use the term

‘health’ as a synonym for the functional integrity of the site; thus, the higher the score for the LFA indices, the greater the health of the landscape. These assessments were made at each site as detailed below.

Landscape organisation

Along the 50 m transect we recorded the number of persistent elements on the soil surface such as perennial grass butts, shrub and tree hummocks and logs that intersected the transect. These elements trap resources (water, soil) moved by overland flow (water movement across the landscapes), and are critical for moderating the effect of runoff water on plant growth and soil stability. They are therefore critical for the long-term functioning of arid and semi-arid systems

(Ludwig et al., 2005). Landscape Function Analysis involves the measurement of three site–level attributes; 1) the number of obstructions to overland flow per unit length of on-ground transect; 2) the average width of these obstructions along the transect; and 3) the average distance between obstructions (fetch).

Together these measurements characterise the functional integrity of a site

(sensu Ludwig and Tongway, 2000).

Soil surface condition

Within each site we assessed the morphology of the soil surface (soil surface condition) within ten 0.5 m2 quadrats. This involved the assessment of 13

110

5. Arthropods and landscape health

surface features (Appendix 5.1). These measurements were used to derive three indices of soil surface condition: 1) stability, a measure of how the soil withstands erosive forces or reforms after erosion; 2) infiltration, which indicates the extent to which rainfall infiltrates into the soil, and 3) nutrients, which provides a measure of how efficiently organic material is recycled in the soil

(Tongway, 1995).

Landscape health assessment 2: terrestrial index of ecological integrity

In a 20m x 50m plot bisected by the 50m transect, we also derived indices of health, which reflected aspects of landscape structure, composition and function

(SCF, sensu Noss, 1990). These indices were derived from individual site–level measurements, which were combined into a single terrestrial index of ecological integrity (Andreasen et al., 2001). Structure was based on the cover of trees, shrubs, perennial and annual grasses, forbs, bare ground, cryptogams, litter, logs, as well as the number of patches (measured using LFA). Composition was based on tree, shrub and groundstorey plant species richness, percentage of plants that were perennial and native, and the degree of woody plant regeneration. Function was based on attributes that provide insights into how a site maintains key ecosystem processes such as the degree of mistletoe infestation, canopy dieback, extent of tree hollows, the extent of erosion and the cover of perennial grass butts, soil organic matter, and soil texture. These 24 different attributes are recognised as important elements of biodiversity assessment (Oliver, 2002; Gibbons and Freudenberger, 2006). Collectively, we refer to this set of indices as SCF.

111

5. Arthropods and landscape health

For each site, each of the 24 attributes was measured and assigned a particular score depending on its perceived effect upon structure, composition or function, with a higher score equating with a healthier landscape. Thus, for example, the range of tree cover values encountered in the bimble box community (range: 0 to 49%) was divided into five classes and scored accordingly: 1 = 0 – 2%, 2 =

2.1 – 5%, 3 = 5.1 – 10%, 4 = 10.1 – 25% and 5 = > 25%). The values of each attribute were placed into classes rather than being treated as nominal values.

The use of classes is both valid and standard procedure when creating indices of biotic integrity or for the terrestrial index of ecological integrity (Andreasen et al., 2001; Gibbons and Freudenberger, 2006). At the level of the site, the scores for each attribute were then summed. For example, for composition, we summed up the values received for tree, shrub and groundstorey plant species richness, percentage of perennial and native plants and woody regeneration.

This total value was then divided by the maximum possible score for that index to give the final index score for the site (sensu Karr, 1991). Overall therefore, we had nine indices associated with two sets of landscape health measures; 1)

LFA (six indices), and 2) SCF (three indices).

Arthropod sampling

All arthropods were sampled using pitfall traps (250ml polycarbonate screw cap containers 100mm high, 67mm diameter). Ten traps were located in a two by five grid pattern (traps 10m apart) centred along the 50m transect. Traps were buried with the lip of the container flush with the ground-surface and contained

125ml of monoethylene glycol to kill and preserve the specimens. All traps were active for 11 days. Upon removal, containers were capped and transported to

112

5. Arthropods and landscape health

the laboratory where the monoethylene glycol was decanted off (using a 0.5mm sieve) and replaced with 70% ethanol. Specimens were sorted under a dissecting microscope to major groups by biodiversity technicians. All ants, beetles and spiders recovered from the traps were further identified to species or morphospecies by specialists at the Australian Museum.

Statistical analyses

LFA and SCF indices and variation in arthropod abundance, species richness and effective species diversity

We used both univariate and multivariate analyses to examine trends in the arthropod community metrics in relation to LFA and SCF. Species richness was calculated as the total number of ant, beetle, or spider morphospecies recorded from each survey site. For a diversity measure, we used a bias-controlled effective number of species (Jost 2006), which has been shown to be one of the least biased diversity estimates (Beck and Schwanghart, 2009). Bias-controlled effective number of species (hereafter ‘effective species diversity’) was calculated using the program SPADE (Chao and Shen, 2003). We used hierarchical partitioning (HP) (Chevan and Sutherland, 1991) to determine the independent and joint capacity of each landscape health variable (or habitat attribute) to explain variation in arthropod abundance, species richness or effective species diversity. This technique gives a measure of how much variation can be attributed solely to that of a variable, compared to that which is due to the presence of other variables in the model (Mac Nally, 2002).

113

5. Arthropods and landscape health

We also used HP to determine the independent capacity of four specifically selected LFA variables to explain variation in arthropod abundance, species richness, and effective species diversity when an additional eight habitat attributes were included in the analyses. For LFA, variables which showed moderate to strong statistically significant co-correlations (α >0.5, P < 0.001) were removed from these analyses (two variables). In doing so we retained the maximum amount of variables in the model whilst minimising the amount of shared variation that could be explained by multiple variables. Thus, the four

LFA variables selected for these analyses were the number of obstructions, width of obstructions, stability, and infiltration. Habitat attributes were raw values that are also used to construct structure, composition, and function (above), and represent an alternate method to a multimetric. The habitat attributes selected were cover of forbs and herbs (%), tree canopy cover (%), litter cover (%), soil organic matter (%), proportion of plants as perennials, cover of low shrubs (%), cover of annual grasses (%) and the proportion of plant species as perennials.

These were selected following correlation analysis which included LFA variables, with highly correlated variables (>0.7) excluded from analyses. For

HP analyses, some non-normal variables were appropriately transformed prior to analyses (log, or square-root). We evaluated models based on the R2 goodness of fit statistic, and the statistical significance of independent effects was calculated using a randomisation test with 1000 iterations (Mac Nally,

2002). All HP analyses were conducted using the hier.part package (Walsh and

Mac Nally, 2003) within the R statistical program (R Development Core Team,

2011).

114

5. Arthropods and landscape health

Relationships between LFA and SCF and species composition

Multivariate methods were used to analyse relationships between the species assemblage structure of ants, beetles and spiders, and LFA, SCF, or a combined LFA/habitat attributes variable set. All multivariate analyses were conducted in PRIMER v6 (Clarke and Gorley, 2006). Biotic matrices were created using Bray–Curtis similarity index on log (x + 1) transformed data.

Environmental variable matrices were created using a Euclidean distance similarity index based on log (x + 1) data, with all variables normalised (Clarke and Gorley, 2006).

Several tests were used to examine the relationship between LFA, SCF and habitat attributes and arthropod species assemblages. RELATE, a multivariate equivalent of a Mantel test was use to examine the correlation between matrices based on LFA or SCF and arthropod species assemblages (Clarke and Gorley, 2006). Further, the BEST (specifically BIO-ENV) analysis was used to identify a subset of factors (LFA or SCF indices and/or habitat attributes) which are the ‘optimal’ match to a second resemblance matrix (i.e. a species matrix), based on the strength of a Spearman rank correlation coefficient. A global permutation test (9999 permutations) was used to assess the statistical significance of the obtained correlations.

RESULTS

We sampled a total of 60,765 arthropods from 457 species across three arthropod groups (Table 5.1). Ants were the most abundant (96% of all specimens). Stability was the most variable LFA index, while for Structure–

115

5. Arthropods and landscape health

Composition–Function (SCF) the most variable was composition, though all indices showed substantial variation (Table 5.2).

Table 5.1: Total abundance and total, median and range of species and higher taxon (Family) richness by site for arthropod groups. a Genera-level richness; b Excluding ants.

Species richness Higher taxon richness Abundance

Total Median Range Total Median Range

Ants a 119 29 13-39 38 16 9-21 57 880

Beetles 173 12 4-27 26 7.5 4 – 15 1180

Spiders 165 17 9-29 24 10 6-13 1705

Total 457 40b 60 765

Table 5.2: Descriptive statistics of all health variables used.

Variable Mean Median Range SD

Fine-scale

Stability % 60.6 61.1 43.8 – 80.1 6.7

Infiltration % 30.1 28.6 20.7 – 48.9 5.9

Nutrients % 24.0 24.1 17.0 – 39.1 4.4

Obstructions/10 m 7.1 7.3 0.33 – 15.4 3.5

Obstruction width m 2.6 1.7 0.04 – 7.5 2.1

Fetch length m 1.8 1.1 0.55 – 15.0 2.1

Coarse-scale

Structure % 56.6 55.6 44.4 – 66.7 5.3

Composition % 74.4 75.0 58.3 – 91.7 8.4

Function % 58.4 56.7 46.7 – 76.7 6.9

116

5. Arthropods and landscape health

LFA, SCF and arthropod abundance, species richness, and effective species diversity

Hierarchical partitioning indicated that the independent ability of LFA indices to account for variation in the arthropod community varied with both arthropod taxon (ants, beetles, spiders) and diversity measure (abundance, species richness, effective species diversity) (Figure 5.1). In only one case was a single variable (infiltration in relation to effective diversity of spider species) able to independently explain >10% of variation in the arthropod community. Despite this, several variables were statistically significant, varying among taxa.

Infiltration was statistically associated with increases in abundance and effective diversity of spiders (Figure 5.1c). The number of obstructions was significantly associated with increases in ant abundance, but decreases in ant effective diversity. Additionally, nutrients and stability were associated significantly with ant abundance.

Hierarchical partitioning analyses for the three SCF indices were similar, though the index that independently explained the greatest amount of variation was generally consistent within a taxonomic group (Figure 5.2). For ants, structure had the greatest independent explanatory power for abundance, species richness, and effective species diversity. Structure also independently accounted for the most variation in spider species richness and effective species diversity, while composition was most important for spider abundance.

The total amount of variation explained by the three SCF indices was low for all taxa (generally <10% for all measures), with the highest amount of variation explained being 13.7% for the effective species diversity of spiders and 12.3%

117

5. Arthropods and landscape health

for ant abundance (Figure 5.2a). In only one case was a variable statistically significant in explaining variation in the model (structure and ant abundance).

Figure 5.1: Proportion of variance explained by independent (white) and joint (black) components of six landscape function analysis indices on (a) ants, (b) beetles and (c) spiders as determined by hierarchical partitioning. Z-score indicates significant effect as determined by randomization tests (P < 0.05). The higher the Z-score, the stronger the relationship. Positive or negative relationships , for significant results, are shown by + or - symbols, respectively. Variables: Fet, mean fetch length; Inf, infiltration; Nut, nutrients; Obs, mean number of obstructions; ObW, mean width of obstructions; Stb, stability. Negative joint effects indicate that variable is acting as a suppressor of other variables.

118

5. Arthropods and landscape health

Figure 5.2: Proportion of variance explained (Y-axis) by independent (white) and joint (black) components of three terrestrial indices of ecological integrity (structure– composition–function, SCF) (X-axis) indices on a) ants, b) beetles, and c) spiders as determined by hierarchical partitioning. Symbols are as in Figure 5.1.

LFA, SCF and arthropod species assemblages

BIO-ENV analyses confirmed that LFA and SCF were generally weak in explaining patterns of the arthropod assemblage composition (Table 5.3).

Subsets of LFA indices selected were not statistically significant, with the exception of a relationship with spiders (ρ = 0.275, P < 0.01; Table 5.3).

Variables commonly included in LFA subsets were stability (four cases) and obstruction width (three cases). Weak but statistically significant relationships (P

< 0.05) were evident between subsets of SCF and ant functional groups (ρ =

0.226), beetles (ρ = 0.178) and spiders (ρ = 0.191) (Table 5.3).

119

5. Arthropods and landscape health

Taxa BIO-ENV (ρ) LFA subset BIO-ENV (ρ) SCF subset

All taxa 0.180 Obs, ObW, Stb 0.133 All

Ants 0.103 Stb, Ntr 0.120 Str, F

Ants# 0.233 ObW, Stb, Ntr 0.226* All

Beetles 0.079 Fet 0.178* C, F

Spiders 0.275* ObW, Stb 0.191* Str, C

Note: Indices: * Statistically significant at P < 0.05. # Functional groups only.LFA: Obs = Mean number of obstructions; ObW = Mean width of obstructions; Fet = Mean fetch length; Stb = Stability; Inf = Infiltration; Ntr = Nutrients; and SCF: Str = Structure; C = Composition; F = Function.

LFA, SCF, habitat variables and arthropod abundance, species richness, and effective species diversity

The HP analyses that included LFA indices together with eight habitat variables revealed that habitat variables were as important as LFA indices for independently explaining variation in arthropod abundance, species richness and effective species diversity (Figure 5.3). However, no variable consistently explained substantial variation across arthropod taxa, and no single variable was statistically significant in more than two (of a possible nine) cases. No variable independently explained >10% of the variation in arthropod species richness, abundance or effective species diversity. A substantial amount of the

120

5. Arthropods and landscape health

variation explained by habitat attributes was related to the joint effects of variables (Figure 5.3). Furthermore, the independent contributions of habitat variables were often offset by negative joint contributions. The total amount of variation explained by all variables together varied little (~ 19% and 29%), the exception being for the number of spiders (40%) and the species richness of beetles (8%).

LFA, SCF, habitat variables and arthropod species assemblages

Four-variable BIO-ENV subsets, which included both habitat variables and LFA indices, explained more variation in biotic assemblages than subsets that only included LFA or SCF (Table 5.4). In the case of ants and beetles, a single habitat variable (tree canopy cover) had a stronger correlation than analyses using LFA variables. For every taxon, with the exception of ant functional groups, tree canopy cover was the single most important variable, and was included in every pair of variables and subset of four strongest variables (Table

5.4). Tree canopy cover and stability were selected as the best pair of variables for ants and spiders. For beetles, tree canopy cover and the cover of forbs and herbs were the best pair of variables. For ant functional groups, litter cover was the single most important variable, and with stability, formed the strongest pair.

The strongest four-variable subset generally included the strongest pair of variables, but resulted in slight increases in explanatory power with the exception of beetles (Table 5.4).

121

5. Arthropods and landscape health

Figure 5.3: Proportion of variance explained (Y-axis) by independent (white) and joint (black) components of four LFA indices and seven habitat attributes (X-axis) on a) ants, b) beetles, and c) spiders as determined by hierarchical partitioning. All symbols and abbreviations for LFA are as in Figure 5.1. Habitat attribute variables: Org, soil organic matter; Lit, litter cover; FH, cover of forbs and herbs; AGr, cover of annual grasses; LS, cover of low shrubs; TC, cover of tree canopy; Per, proportion of plant species that are perennial.

122

5. Arthropods and landscape health

Table 5.4: BIO ENV results using habitat variables and LFA indices, showing correlation between community structure and specified variable/s

Rank Taxon ρ Ants Single variables analysed 1 Tree cover (%) 0.184** 2 Number of obstructions per10 m 0.086 3 Stability 0.08 4 Perennial grass cover (%) 0.066 5 Litter cover (%) 0.05 Variables analysed in pairs 1 Tree cover (%) + stability 0.216** Four variables analysed 1 Tree cover (%) + cover of perennial grasses (%) + number of 0.244** obstructions per 10m + Stability Ant functional groups Single variables analysed 1 Litter cover (%) 0.145* 2 Tree cover (%) 0.101 3 Number of obstructions per 10m 0.084 4 Stability 0.068 5 Obstruction width 0.052 Variables analysed in pairs 1 Litter cover (%) + stability 0.192* Four variables analysed 1 Tree cover (%) + Litter cover (%) + number of obstructions per 0.259** 10m + Stability Beetles Single variables analysed 1 Tree cover (%) 0.198* 2 Proportion of plants as perennials 0.11 3 Cover of forbs and herbs (%) 0.089 4 Cover of annual grasses (%) 0.043 5 Litter cover (%) 0.04 Variables analysed in pairs 1 Tree cover (%) + forbs/herb cover (%) 0.21** Four variables analysed 1 Tree cover (%) + forb/herb cover (%) + proportion of plants as 0.199** perennials + annual grass cover (%) Spiders Single variables analysed 1 Tree cover (%) 0.219** 2 Width of obstructions 0.192** 3 Proportion of plants as perennials 0.154** 4 Stability 0.149* 5 Soil organic matter (%) 0.132* Variables analysed in pairs 1 Tree cover (%) + stability 0.301** Four variables analysed 1 Tree cover (%) + soil organic matter (%) + Obstruction width + 0.379 stability ** * P < 0.05; ** P < 0.001. LFA, Landscape Function Analysis

123

5. Arthropods and landscape health

DISCUSSION

Recognition of the unprecedented loss of biodiversity due to anthropogenic disturbances has spurred the development and use of surrogates and indicators for biodiversity monitoring (e.g. Churchill, 1998; Andersen and Majer, 2004).

Currently there is a strong emphasis on indicators and surrogates of biotic diversity for regional environmental management (e.g. Muñoz-Erickson et al.,

2007; Smyth et al., 2009). For many taxa, however, including ants (Whitford et al., 1998; Andersen et al., 2004), beetles (Pearce and Venier, 2006) and spiders (Pearce and Venier, 2006), their reliability as indicators of biotic integrity in relation to rangeland and forestry management has been called into question.

Our study shows that there may not be a predictable and unequivocal relationship between a potential surrogate and biodiversity despite the fact that the relationship is theoretically sound and supported empirically by earlier research (Ludwig et al., 1999; 2004). Such results reinforce our view that surrogates should not be used widely without a clear knowledge of how environmental and ecological factors drive patterns of biodiversity in a given ecosystem.

Arthropod biodiversity in relation to LFA and SCF

Few studies have explicitly examined potential relationships between landscape health and biodiversity using LFA. Ludwig et al. (1999) showed that with increasing distances from a watering point there were increases in the density and size of perennial vegetation patches (obstructions), as well as plant and grasshopper diversity. Ludwig et al. (2004) also showed how changes in plant diversity, plant production and the activity of medium-sized mammals varied in

124

5. Arthropods and landscape health

relation to the functional integrity of sites at scales ranging from that of individual plants to whole catchments. Similarly, Dawes-Gromadzki (2005) demonstrated a positive relationship between landscape condition (assessed as the number of permanent plant patches) and the abundance of macro- arthropods, though her study did not use LFA explicitly. The results of our study are therefore somewhat at odds with previous research which has used LFA, and highlights the inherent variability in relationships between invertebrate communities and measures of landscape function. While there appears to be strong congruence between the LFA attributes and patterns of biotic diversity under some circumstances, we currently do not know why these relationships are so inconsistent. This lack of understanding likely stems from inadequate knowledge of the life history of invertebrate taxa (Redak, 2000) and of relationships among invertebrates, soil type, disturbance history and resource retention (Ludwig et al., 2004). Indeed, despite the utility of LFA in rangeland monitoring (e.g. Watson et al., 2007) and its potential promise as a biodiversity indicator, key questions remain regarding the exact nature of this relationship

(Ludwig et al., 2004). These questions concern the relationships between resource availability and distribution and species persistence, in particular the spatial and temporal conditions under which resources (e.g. patch obstructions) become important for the fauna, and how and when this is modified by disturbance such as fire and grazing (Ludwig et al., 2004).

Globally there is a wealth of research linking landscape condition (using indices of integrity) to biodiversity values (e.g. Karr and Kimberling, 2003; Bryce, 2006;

Diffendorfer et al., 2007). Although indices of ecological integrity are regarded

125

5. Arthropods and landscape health

as potential surrogates for biodiversity in forest and woodland systems (e.g.

SCF, Oliver et al., 2007), such relationships have not been validated in Australia

(Fisher and Kutt, 2007). We found very few strong links between SCF and arthropod biodiversity, with these variables unable to explain significant variation in arthropod abundance, species richness, or species diversity. All elements of SCF were however related to the community structure of at least one taxon (ants, beetles, or spiders). It is important to note, however, that SCF explained less variation in community structure than many simpler habitat attributes. Contemporary biodiversity management is focussed on species assemblages and the structure of biotic communities, rather than species richness or species diversity per se (Jennings et al., 2008), For this reason we would consider SCF to be of only limited use in biodiversity monitoring as simpler, stronger, and more reliable predictors are available for predicting arthropod diversity and community structure.

The use of multimetrics has been criticised as they do not always accurately reflect variation in their individual components (Suter, 1993; Andreasen et al.,

2001). Of specific concern is the loss of information due to the reduction of multiple measurements into a smaller number of indices (e.g. raw habitat attributes into SCF). Here, a low value from one variable can be ‘compensated’ by increases in several of the other variables (Suter, 1993; McCarthy et al.,

2004), and the exact reason for a given multimetric value is not known. In our case, the final multimetrics of structure–composition–function may not capture sufficient variation in fine-scale factors, which may be important for arthropods.

This could contribute to the apparent poor relationships between the predictor

126

5. Arthropods and landscape health

variable and the measure of diversity. There are several ways to alleviate this problem, including a multiplicative approach (whereby compensation between variables is not linear; McCarthy et al., 2004) and the use of ‘raw’ values

(Andreasen et al., 2001). In our study, using raw values (‘habitat attributes’) did reveal a few notably important factors, but this was completely taxon- dependent.

Arthropod biodiversity in relation to additional habitat attributes and LFA

We also investigated whether individual habitat attributes were more closely aligned with arthropod diversity than composite (multiple) metrics. We identified several habitat attributes which were as important as LFA indices in explaining variation in arthropod abundance and richness, and these were primarily measures of vegetative cover (e.g. forb, litter and shrub cover). Tree canopy cover was the variable most strongly related to arthropod assemblage structure, and was important for all taxa (Table 5.4). Thus, while tree cover is not necessarily a strong predictor of arthropod abundance or diversity, it may be a useful predictor of arthropod community structure. Overall, however, attributes generally had a low explanatory power, indicating few strong, independent relationships between biodiversity and condition. These results are similar to those reported both for LFA and SCF.

Structure and composition of vegetation are considered reliable indicators of animal species diversity (e.g. habitat heterogeneity, see review by Tews et al.,

2004). Hence, measurement and monitoring of these attributes are an integral part of research (Hughes et al., 2000; Stoner and Joern, 2004; Ferrier and

127

5. Arthropods and landscape health

Guisan, 2006). They are also widely used by conservation managers (e.g. www.environment.nsw.gov.au), particularly to assess vegetation condition and habitat quality for vertebrates. However, there remains a paucity of data relating habitat attributes to arthropod diversity (Tews et al., 2004). Our results support previous research showing that, although somewhat inconsistent, measures of vegetative cover and structure are related strongly to arthropod diversity (e.g.

Gardner et al., 1995; Debuse et al., 2007; Sanders et al., 2008). In particular, other studies have also recognized the importance of trees in structuring arthropod communities (Lassau and Hochuli, 2004; Lassau et al., 2005; Oliver et al., 2006; Barton et al., 2009; Gollan et al., 2009). Though responses were taxon-specific, it still may be possible to predict broad community structure based on tree cover, particularly within a single vegetation community, and/or where there is little extrinsic environmental variation. However where a large gradient in environmental variation exists across the study area (e.g. in temperature, soil type, rainfall), changes to plant cover and even species composition may simply be a reflection of these underlying gradients (e.g.

Thompson and Eldridge, 2005; Tzialla et al., 2006). In these cases a more effective method would be to use the factor that is driving tree cover as a direct surrogate.

Many authors have stressed the importance of soil characteristics such as hardness (Bestelmeyer and Wiens, 2001; Gollan et al., 2009), erodibility (Schell and Lockwood, 1997) and organic matter content (Lavelle et al., 2006) as drivers of arthropod community structure. In our study, soil stability was an important predictor in analyses of arthropod assemblage structure, though

128

5. Arthropods and landscape health

mainly in combination with tree cover. Soil organic matter was also important for spiders. Given that the drivers of community structure varied widely across taxa, any study investigating patterns of arthropod distribution in relation to environmental factors should encompass a wide variety of soil and vegetation attributes. Many of the potential drivers identified in this study are part of broad environmental monitoring programmes (McCarthy et al., 2004; Gibbons and

Freudenberger, 2006).

Arthropod communities are structured by fine- and broad-scale environmental variation

Effective indicators must be ecologically relevant at a scale appropriate to the target taxa, either directly or indirectly (Zurlini and Girardin, 2008). Arthropod activity can be influenced by environmental variation occurring over fine

(centimetre) to broad (kilometre) scales (e.g. beetles, McIntyre, 1997; Barton et al., 2009; insects in general, Major et al., 2003; Lassau and Hochuli, 2004).

While individual components of LFA such as the number of resource patches are ecologically relevant to fauna and flora across multiple spatial scales

(Ludwig et al., 2004), LFA typically reflects ecosystem function at the patch– hillslope to catchment scales (e.g. hundreds to thousands of metres). Thus, LFA is expected to be more closely aligned with the diversity of a variety of animals

(Ludwig et al., 2004) as well as cryptogamic crusts and vascular plants (David

Tongway, personal communication, 2010). However while arthropods may be responding to variation in landscape function at the patch–hillslope scale, it is possible that finer-scale environmental variation, not directly captured by LFA, plays a greater role in structuring these communities. This is analogous to

129

5. Arthropods and landscape health

within–patch characteristics such as vegetation structure being more important in structuring faunal communities than broader, landscape context characteristics such as patch size or connectivity (e.g. Debuse et al., 2007;

Poyry et al., 2009). There is, therefore, a unique scaling issue when using LFA to predict arthropod communities. Theoretically, either LFA data need to be down-scaled to consider variation at the scale of individual arthropod movements (e.g. tens of metres) and/or arthropod data need to be up-scaled, such as placing a large array of pitfall traps across entire landscapes, to harmonize these two scales (David Tongway, personal communication 2010).

Currently we lack a standard and empirically-tested method for scaling data within a framework of functional integrity that is appropriate for small animals.

Without this, generalizations cannot be made among similar studies.

As arthropods respond to such a wide variety of environmental variables across multiple spatial scales, elucidating broad drivers of community structure has been a significant research challenge. Drivers can be taxon-specific (e.g. litter structure is particularly important for spiders, Bultman and Uetz, 1984), but there are broad factors which affect a range of taxa (e.g. habitat heterogeneity,

Tews et al., 2004; climate, Andrew and Hughes, 2005). In general, our research highlights the general lack of alignment of LFA and similar measures with most taxa. Many of the factors that drive arthropod communities are specific measures or surrogates of arthropod habitat or the microclimate associated with that habitat (e.g. fine-scale habitat availability, Mazia et al., 2006; Nitterus and

Gunnarsson, 2006; disturbance history, Hoffmann and Andersen, 2003). This was confirmed by our study, with all arthropod communities being structured

130

5. Arthropods and landscape health

primarily by tree cover, yet affected to a lesser degree by different factors such as litter cover and the number of obstructions (ants), plant composition (beetles) and obstruction width (spiders). Leaf litter is known to structure spider communities and their prey (Bultman and Uetz, 1984). Similarly, the selection of nesting sites by ants is often influenced by soil texture (Bestelmeyer and Wiens,

2001). Landscape Function Analysis indices, however, are constructed specifically to measure the ability of the landscape to capture abiotic resources, that is, ecosystem function (Ludwig et al., 2004). This, together with the scale mis-match described above, suggests that LFA indices do not directly measure the components of ecosystems that exert the strongest influence over arthropods. This is a fundamental part of LFA which will always limit the applicability of LFA as a surrogate of arthropod biodiversity.

Conclusions

The relationships between broader biodiversity values and landscape health deserve greater attention, particularly in environments where the management or conservation of biodiversity is paramount. While the accuracy of LFA (and individual soil surface condition attributes) is continually being validated

(McIntyre et al., 2003; Maestre and Puche, 2009), its potential use in understanding patterns of biodiversity is not. The situation is similar for any index of ecological integrity based on structure–composition–function metrics, given their widespread applicability in vegetation monitoring. We found only weak relationships between arthropod biodiversity and LFA and SCF. This could be due to the way multimetrics are derived, although simple habitat attributes showed similar patterns. As arthropods are potentially influenced by

131

5. Arthropods and landscape health

fine- to broad-scale environmental variation (centimetres to kilometres), it is possible that LFA and SCF are not reflecting variation at the scale that is driving these communities. A conceptual framework specifically linking landscape health (and associated indices), arthropods and spatial scale is necessary to provide a standard platform upon which to base empirical predictions and testing.

ACKNOWLEDGEMENTS

Many staff from the former NSW Department of Land and Water Conservation were involved in the collection of data upon which this study is based. We thank

Sarah Coulson, Dale Collins, John Lemon, Warren Martin, Chris Nadolny and

Peter Serov. Lance Wilkie from the Australian Museum co-ordinated the species-level identification of arthropod taxa. All animal sampling and surveys were approved under NSW Agriculture Animal Research Authority No. 95/044, and NSW NPWS Scientific Investigation License No. A2868. Two anonymous reviewers provided constructive feedback on an earlier draft. Finally we acknowledge the significant contribution made by David Tongway and John

Ludwig (CSIRO, Australia) to our understanding of landscape processes, monitoring and health assessment in arid and semi-arid environments.

132

5. Arthropods and landscape health

Appendix 5.1: Soil surface features used to calculate soil surface condition (SSC) indices

Soil surface feature Interpretation Surface roughness Defined as the vertical distance between the lowest and highest points in the quadrat, and is a measure of the potential for retention of rainfall on the surface. Crust resistance Assesses the degree to which the soil crust, in this case biological or physical, can be mechanically disturbed to yield sediment. It is measured on a dry soil. Crust brokenness Defined as the percentage of the surface covered with cracks and relates to the capacity of the surface to disintegrate and erode. In addition, the extent to which the surface cracks may be indicative of potential microsites for seeds to settle. Crust stability Defined as the degree to which surface soil aggregates maintain their stability when wetted and is measured using the Emerson slake test (Tongway and Hindley 2004). Erosion type and severity Assesses erosional features such as rilling, sheeting, scalding, terracettes and pedestals. Deposited materials Assesses the nature and quantity of materials deposited on the surface from upslope sources. Foliage cover of vascular Measures the capacity of the vegetation to plants intercept raindrops. Cover of cryptogams As for vascular plants. (biological soil crusts)

Basal cover of perennial Indicates examine the effect of cover on overland plants and other flow processes. permanent cover components Soil texture Assesses texture using the bolus technique (McDonald et al., 1990) on surface soils. Litter cover and Three attributes are assessed to provide an decomposition indication of the decomposition of plant material: i) the origin and ii) cover of litter, and iii) the degree of litter incorporation into the soil.

133

6. Multi-scale drivers of ant communities

Chapter 6

Drivers of ant community structure in a grassy eucalypt

woodland ecosystem in south-eastern Australia

Alan B. C. Kwok, David J. Eldridge and David Freudenberger

Keywords: fragmentation, ant assemblages, diversity, functional group, patch quality

134

6. Multi-scale drivers of ant communities

ABSTRACT

Anthropogenic-induced landscape-scale change can have significant effects on the structure of biotic communities. Increasing evidence, however, highlights the importance of ecological drivers that operate at finer spatial scales. We investigated the relative importance of landscape context (patch size, isolation), within-patch characteristics (e.g. vegetation cover, soil texture) and ant–ant

(biotic) interactions in structuring ant communities in patches of eucalypt woodland in an agricultural region of south-eastern Australia. We sampled ant assemblages at 35 woodland remnants varying in size, isolation, and within- patch characteristics. Variation in ant community structure in relation to environmental variables was predominantly in terms of species or functional group composition, not abundance or species diversity measures. For ant species assemblages, the environmental predictor of highest relative importance was soil texture (% clay content, followed by tree canopy cover). For ant functional groups, tree canopy cover was the variable of highest importance, followed by distance to the nearest remnant 10 ha or more in size, and soil texture. The abundance of the dominant ant functional group was a generally weak predictor of ant abundance, species richness, and effective species diversity. Our results suggest that within-patch attributes of soil texture and tree canopy cover are particularly important drivers of ant communities in the eucalypt woodlands of south-eastern Australia. Conversely, landscape change or context attributes (remnant size, isolation) and biotic interactions are relatively weak drivers. Thus, attributes that are typically important for vertebrate biodiversity management (e.g. remnant size, isolation) are not necessarily of greatest importance for invertebrates such as ants. Careful

135

6. Multi-scale drivers of ant communities

attention needs to be directed to ecological characteristics that exist at finer patch–scales when managing or surveying invertebrate biodiversity.

INTRODUCTION

Anthropogenic disturbance is well recognised as a significant driver of changes in biotic communities (Heller and Zavaleta, 2009). It is well-established that disturbances such as habitat fragmentation, patch condition decline and reduced habitat connectivity have negative and cascading effects on many measures of biodiversity (e.g. Fischer and Lindenmayer, 2007). Landscape context (e.g. habitat isolation, size of habitat fragments), in particular, can play a crucial role in affecting species and entire biotic communities across a range of taxa (Fahrig, 2003; Lindenmayer et al., 2008). Research to date has focused predominantly on the more charismatic and visible fauna, particularly birds, and to a lesser extent, mammals (Fahrig, 2003; Prevedello and Vieira, 2010), as well as keystone or umbrella species (e.g. Murphy and Noon, 1992;

Lindenmayer et al., 2008). Increasingly, however, greater attention is being devoted to understanding how more abundant and widespread taxa such as invertebrates are affected by processes of ecological change at a range of scales (Crist, 2009).

Ants as a group are involved in a range of key ecological processes (Whitford,

1996). They are the dominant arthropods in nearly all terrestrial ecosystems, including heavily altered environments. Ants play an important role in key ecosystem services including pedogenesis, nutrient cycling (Wagner and Jones,

2004; Wardle et al., 2011), water redistribution (James et al., 2008), seed

136

6. Multi-scale drivers of ant communities

dispersal, and germination. They are also important predators and consumers

(e.g. Folgarait, 1998). Loss of ant species richness or changes to ant community composition may have consequences for ecosystem function across multiple trophic levels (Andersen and Patel, 1994; Andersen, 1997; Sanders and van Veen, 2011). Furthermore, ants are known to respond to a broad range of environmental disturbances (e.g. mining, Andersen and Sparling, 1997; grazing, Hoffmann, 2000; urban development, Sanford et al., 2009), including anthropogenic-induced landscape change (Crist, 2009). Due to their ecological importance and ease of sampling, ants have widely been proposed as indicators of ecological change and are now used regularly in ecosystem monitoring (Andersen and Majer, 2004; Underwood and Fisher, 2006), although there remains debate surrounding their efficacy as indicators of disturbance

(Whitford et al., 1999; Andersen and Majer, 2004).

Generally, research on the processes that drive ant community dynamics has focused on drivers at three spatial scales, operating from broad to fine scales: changes to landscape context (e.g. size of habitat patch, isolation and connectivity), patch–scale changes to habitat or resources (e.g. in the condition and/or structure of vegetation within a remnant), and biotic interactions within a patch (e.g. species, functional group relationships). While several studies indicate that landscape context is the primary attribute structuring invertebrate communities (e.g. Gove et al., 2009) there is increasing evidence that finer- scale habitat characteristics or processes are of similar or greater importance

(Mazerolle and Villard, 1999; Hoffmann, 2010). These include characteristics of within-patch habitat, such as vegetation structure, plant species composition

137

6. Multi-scale drivers of ant communities

(Mazerolle and Villard, 1999; Stoner and Joern, 2004; Poyry et al., 2009), and soil properties (e.g. texture, clay content: Bestelmeyer and Wiens, 2001; Crist,

2009).

There is also a growing body of literature suggesting that biotic interactions (e.g. inter-specific and inter-functional group interactions) can also have an important influence in structuring ant communities (Andersen and Patel, 1994; Arnan et al., 2011). This is particularly the case in Australia where there can be strong interactions between species and genera (functional group classification scheme, Andersen, 1995). Associations and interactions between functional groups can be particularly important in influencing ant species richness and diversity (Andersen, 1995; Arnan et al., 2011).

While many studies have established a strong case for the importance of different drivers of ant communities and their associated spatial scale in isolation, there have been few attempts to examine the independent effects of both biotic and environmental drivers across multiple spatial scales (although see Debuse et al., 2007; Spiesman and Cumming, 2008; Narendra et al., 2011).

Determining the relative importance of a large number of drivers, operating at different spatial scales, is important for conserving and managing land for invertebrates, and will provide valuable insights into mechanisms structuring invertebrate communities in general.

Our study aimed to determine the relative importance of a range of scale- dependent drivers on ant communities in the eucalypt woodlands of south-

138

6. Multi-scale drivers of ant communities

eastern Australia. We focused on ant communities because they are ubiquitous and important components of biotic communities, and there is evidence as noted above that they respond to a broad range of scale-dependent environmental changes. We specifically hypothesised that soil type and patch- scale attributes such as vegetation structure would be relatively more important than landscape context (patch size and connectivity), or biotic interactions, in explaining variation in ant community abundance, species richness, diversity and composition. We also hypothesised that patch scale variation in habitat structure would have the greatest effect on ant community composition. These hypotheses are consistent with substantial research on ant-landscape ecology in the past decade (Mazerolle and Villard, 1999; Bestelmeyer and Wiens, 2001;

Stoner and Joern, 2004; Crist, 2009; Poyry et al., 2009) and recent global syntheses of the responses of ant communities to ecological disturbance

(Hoffmann, 2010; Hoffmann and James, 2011).

METHODS

We tested these hypotheses by examining changes in ant community structure across 35 sites ranging from homogenised agricultural fields to intact remnant woodlands, asking whether the patterns we observed are species- or functional- group specific.

Study area

The study was conducted in the Riverina region of south-eastern New South

Wales, Australia (35o 50' S 146o 00' E) centred on the village of Savernake. This region is dominated by intensive dryland and irrigated annual crops and non- 139

6. Multi-scale drivers of ant communities

native pastures. The climate is temperate, with a long-term average annual rainfall of 497mm (Bureau of Meteorology, 2012). Rainfall is slightly winter- dominant, with 18% more rain falling between April and September compared with the six summer months. Summer temperatures average 30oC and winter temperatures 16oC (Bureau of Meteorology, 2012).

The geology of the general area is Quaternary alluvium, with outcrops of Upper

Ordovician material and thinly covered bedrock on the upper slopes and crests of hills and ridges (Beattie, 1972). The Quaternary deposits comprise extensive, level to locally depressed plains of clays and clay loams, traversed by discontinuous low ridges associated with the levees of prior streams.

Superimposed on the plains are three smaller units: i) sand ridges and lunettes up to 10 m high dominated by well-sorted loams and sandy loams; ii) rocky outcrops of shallow, stony loams on exposed granites or meta-sediments to 20 m in relief, and iii) shallow depressions and lowlands of cracking grey clays

(Butler, 1958; Kent et al., 2002).

The study area occurred within the widely cleared grassy eucalypt woodlands in the eastern part of the wheat–sheep belt in New South Wales (Moore, 1970).

Remnant vegetation on the fine-textured soils was dominated by grey box

(Eucalyptus microcarpa Maiden), yellow box (E. melliodora Schauer.), buloke

(Allocasuarina leuhmannii L.A.S. Johnson) and white cypress pine (Callitris glaucophylla J. Thomps. & L.A.S. Johnson). These soils also supported a diverse but sparse shrubby understorey of Acacia, Eutaxia, Dodonaea and

Maireana spp. The groundstorey comprised a moderate cover of native

140

6. Multi-scale drivers of ant communities

perennial grasses (e.g. Austrodanthonia and Austrostipa spp.) and introduced annual grasses (e.g. barley grass Hordeum leporinum and wild oats Avena spp.). Coarse-textured soils were dominated by grey box, yellow box, white cypress pine and Blakely’s red gum (E. blakelyi Maiden). The understorey and groundcover were dominated by Acacia, Hibbertia, Eragrostis and Austrostipa spp. (Eldridge and Freudenberger, 2005).

Site selection

We chose 35 sites that encompassed the wide range of remnant woodland types found within the region. This included remnants of various shapes and sizes, vegetative condition and land uses (e.g. native woodlands harvested for timber, grazing reserves, roadside reserves, and cultivated paddocks with single remnant trees). All sites were subject to grazing by kangaroos. Livestock grazing was variable, with some sites continuously grazed, some occasionally grazed during a year, and others that had not been grazed for decades. Most of the sites were on plains (fine-textured soils; n = 24), with the remainder on low sandy ridges (coarse-textured, n = 11). The dominant soils on the plains were hard, alkaline, red duplex (red brown earths Dr 2.3, Northcote, 1966) or Red

Chromosols (Isbell, 1996). The surface soils of the sand ridges were generally loose, greyish-brown, neutral red earths (Gn2, Northcote, 1966) or Kandosols

(Isbell, 1996), gradually becoming finer (higher clay content) with depth

(Northcote, 1966).

141

6. Multi-scale drivers of ant communities

Assessing landscape-scale characteristics

The area (hectares, hereafter 'patch area') and dimensions of each woodland remnant were measured within a Geographic Information System (GIS) that held a contemporary SPOT panchromatic satellite image. Site isolation

(hereafter 'TenHa') was measured in the GIS as the distance (in kilometres, edge to edge) to the nearest woodland patch greater than ten hectares in size.

Assessing within-patch characteristics

Patch-scale or within-patch habitat characteristics were measured using a variety of methods. The functional integrity of each remnant or site (i.e. ability to retain critical resources at the scale of tens to hundreds of metres) was assessed using the Landscape Function Analysis method first described by

Tongway (1995) (and updated by Tongway and Ludwig, 2011). This method included the measurement of permanent and semi-permanent ground layer obstructions (e.g. perennial grass tussocks, shrubs, and logs) that were able to capture redistributed water and sediments. The total number of these obstructions, their width and distance apart (fetch) were measured at each site along a single 25m transect randomly positioned but orientated upslope. For the purposes of statistical analyses, however, only the total number of obstructions per 25m was used, based on evidence that arthropods are influenced more by the abundance of obstructions than their width or separation (Kwok et al.,

2011).

Detailed measurements were also made of soil surface features. Along each

25m transect, the soil surface was first divided into a number of readily

142

6. Multi-scale drivers of ant communities

identifiable microsites which were later amalgamated into three categories: i) bare or biological soil crust, ii) litter-covered soil and iii) grassy patches. Bare- crusted microsites were devoid of vascular plants, and ranged from bare soil surfaces to soil colonised almost entirely by cryptogamic crusts (lichen, moss or liverworts) or cyanobacteria and algae (indicated by the organic staining). Litter surfaces were defined as those with >50% cover of dead or detached plant material. Grass microsites were generally characterised by perennial grass tussocks, but occasionally had a variable, but sparse cover of annual and perennial forbs. The projected foliage cover of grass was generally >70%.

Attributes of the soil surface were used to derive an index of stability for each site which relates to the extent to which the soil resists erosion. Near each transect seven surface features within each bare-crust, litter and grass patch were measured within five 0.5m2 quadrats per site, following Tongway (1995) and as described in Eldridge et al. (2006). An index of soil stability was calculated using this soil surface assessment data using the method described by Tongway and Ludwig (2011). Soil texture was also derived from this procedure, and converted into a semi-quantitative predictor variable based on clay content.

Vegetation structure and habitat complexity was measured using the method of

Catling and Burnt (1997), as modified by Watson et al. (2001). An index of habitat complexity was derived from visual assessment of tree cover, shrub cover, groundstorey cover and the cover of litter. Each of these layers of habitat was rated on a score from 1 to 4, representing classes of projected cover for

143

6. Multi-scale drivers of ant communities

that habitat layer (e.g. for tree, shrub, and groundstorey cover a score of 1 = 0–

10% projected cover, 2 = 11–20%, 3 = 20–50%, 4 = >50%). For leaf litter, a score of 1 = 0–10%, 2 = 10–40%, 3 = 40–70%, 4 = >70%. These factors, as a score of habitat complexity, have been shown to be correlated with bird diversity

(Watson et al., 2001) and ant diversity (Lassau and Hochuli, 2004).

The diversity of vascular plants was also measured at each site. All tree and shrub species were identified from within a 50 m2 area and their cover measured. A 20m2 quadrat was established within the larger quadrat, within which all ground layer species were identified and cover measured. The dominance of forbs and grasses was determined from four 1m2 quadrats within the 20m2 quadrat and their projected cover assigned a score as for all other vegetative layers (see above). An active timed search radiating out from the

50m2 area was conducted to locate additional plant species, with the time limit scaled according to the size of the remnant. All plant surveys were conducted during spring.

Ant sampling

Ants were sampled using a pitfall array used for concurrent small mammal and reptile trapping. Each array (one per site) consisted of four 20L buckets (300mm diameter) buried level with the soil surface. The buckets were placed in a ‘T’ formation with 25 m of fine drift netting along each axis as described in detail by

Catling and Burt (1997). Traps were active for three nights and all sampling was conducted in March 2000. Invertebrates were removed from the traps using a hand vacuum device, and were placed into jars of ethanol for preservation. Ants

144

6. Multi-scale drivers of ant communities

were sorted to morphospecies and identified to species where possible by Alan

Andersen and Tony Hertog (CSIRO Australia). Ants were classified to functional group as per Andersen (1995).

Selection of predictor variables

Initially two sets of environmental predictors were chosen that were expected to influence ant community structure: landscape scale characteristics and within- patch habitat characteristics. Within each of these sets, if two predictors showed moderate to strong correlations (α >0.65, P < 0.05), only one was included in the models. The variable selected was the one which showed the lowest correlation with other variables in the dataset. In addition, for functional group analyses, a third set of biological predictors (essentially biotic interactions) was derived for the analysis.

For landscape change/context characteristics the area of the remnant (ha, hereafter ‘patch area’) and distance (km) to the nearest woodland patch greater than 10 ha in extent (TenHa) were chosen for inclusion in models as these predictors have been previously shown to have a strong influence on ant species assemblages (Debuse et al., 2007; Crist, 2009).

For within-patch attributes, three measurements of vegetation or habitat complexity were chosen. Two of these, the projected cover of the tree canopy and surface leaf litter, are commonly included in measures of habitat complexity

(see above) and vegetation surveys (e.g. McCarthy et al., 2004). A greater variety of vegetation attributes were recorded, but many of these were highly

145

6. Multi-scale drivers of ant communities

auto-correlated, leaving only the two selected attributes. Furthermore, the habitat complexity score derived from these attributes was highly correlated (α =

0.73) with tree canopy cover, so only tree canopy cover was selected for analyses, based on principles of parsimony. The other predictor variable was an index representing the vascular plant community. This index was based on the number of species and percent of exotic species per site and was generated using a Principal Components Analysis (PCA). The first PCA axis was used.

This axis represented the gradient from –1 to +1 between the number of total and exotic species respectively, and represented 86% of variation in the plant community species richness.

For analyses of drivers of ant functional group (sensu Andersen, 1995), a biological predictor group was included in analyses. Here, the abundance of the dominant taxon was used to predict the abundance or richness of the subordinate functional group. For Subordinate Camponotini and Generalised

Myrmicinae, this predictor was the number of Dominant Dolichoderinae. For

Opportunists, this predictor was the number of Generalised Myrmicinae. The use of these derived predictors is based on known competitive and dominance interactions between the functional groups (Andersen, 1995; Arnan et al.,

2011).

Statistical analyses

Both univariate and multivariate procedures were used to examine relationships among environmental predictors and ant community structure. Species richness was calculated as the total number of ant species recorded from each survey

146

6. Multi-scale drivers of ant communities

site. For a diversity measure, the bias-controlled effective number of species was used (Jost, 2006). This diversity measure, hereafter termed ‘effective species diversity’ has been shown to be one of the least biased diversity estimates (Beck and Schwanghart, 2009). Effective species diversity was calculated using the program SPADE (Chao and Shen, 2003).

Hierarchical partitioning (HP) (Chevan and Sutherland, 1991) was used to determine the independent and joint capacity of environmental predictors to explain variation in ant abundance, species richness or effective species diversity of both the entire ant community and also within each functional group.

This technique gives a measure of how much variation can be attributed solely to that of a variable, compared to that which is due to the presence of other variables in the model (Mac Nally, 2002). For HP analyses, some non-normal variables were appropriately transformed prior to analyses (log, or square-root).

Models were evaluated based on the R2 goodness of fit statistic, and the statistical significance of independent effects was calculated using a randomisation test with 1000 iterations (Mac Nally, 2002). Due to the problems associated with using more than 9 variables in HP (Olea et al., 2010), grazing was not used in analyses of abundance, species richness or effective species diversity as it does not appear to strongly relate to these aspects of community structure (Hoffmann, 2010). All HP analyses were conducted using the hier.part package (Walsh and Mac Nally, 2003) in the R statistical program (R

Development Core Team, 2011).

147

6. Multi-scale drivers of ant communities

Species or functional group composition and environmental predictors

DistLM was used to model the relationship between ant species or functional group composition and predictor variables. Functional group composition refers to analyses conducted where functional groups were the response variables, rather than species. DistLM is akin to a multivariate multiple regression, using a distance-based redundancy analysis approach (Anderson and Gorley, 2008).

This procedure calculates the relationship between a multivariate matrix (in this case a Bray–Curtis resemblance matrix of species composition, or Modified

Gower matrix in the case of functional group composition) and an individual predictor when alone in a model (marginal test), as well as the independent contribution of a predictor when others are included in the model (conditional test). The significance of marginal tests was assessed using permutation tests

(n = 9999). For conditional tests we used the "BEST" selection procedure to generate the number of models required by multi-model inference and averaging (Anderson and Gorley, 2008). The total number of models required for nine variables was 511. We used Canonical Analysis of Principal coordinates (CAP) and non-metric multi-dimensional scaling (nMDS) to display these results. All multivariate analyses were conducted in the PRIMER (Version

6) (Clarke & Gorley, 2006) + PERMANOVA (Anderson and Gorley, 2008) package.

Multimodel inference and model averaging of predictor variables

To gain a more robust insight into drivers of the ant community, multivariate models (species and functional group composition) were subjected to multimodel inference and model averaging. This procedure is intended to determine

148

6. Multi-scale drivers of ant communities

the relative importance of each predictor variable to the response variable over a series of models, rather than a single model. This process is detailed by

Burnham and Anderson (2002), but is briefly as follows. For species or functional group composition, a set of alternative models was constructed from all linear combinations of the predictor variables, using DistLM. A modified

Akaike Information Criterion (AICc) for each model was calculated as an indicator of model parsimony. Using the AICc, the AICc difference was calculated. This was the difference in AICc values between the selected and most parsimonious model, and this value allows ranking of the models. Using the AICc difference, Akaike weights were calculated for each model to describe the weight of evidence for any given model relative to all other models in the set, that is, the overall rank of the model across all models generated (Burnham and Anderson, 2002). Following this, the relative importance of each variable was calculated. This value was calculated for each variable separately by summing the Akaike weights across all models in which the variable appears

(Burnham and Anderson, 2002). The larger the value of the summed Akaike weight, the more important that variable was relative to the others in the set.

DistLM and multimodel inference are complementary sets of analyses. However multimodel inference calculates a value of importance for each variable based on a series of models, rather than a single model approach that DistLM uses.

Thus, when attempting to determine the relative importance of a specific variable toward community structure, greater weight should be placed upon the results from multimodel inference.

149

6. Multi-scale drivers of ant communities

We further analysed the relationship between the variable of highest relative importance to species and functional group composition. For species composition, this variable was a continuous predictor, and so Indicator Species

Analysis (ISA) was used within the program PC-Ord (McCune and Mefford,

1999) to determine the nature of species-specific responses. ISA calculates an

Indicator Value (IndVal), which quantifies the concentration of a species abundance to the predictor variable (e.g. soil type) (Dufrene and Legendre,

1997). The statistical significance was tested using a randomisation procedure

(n = 1000). Additionally, regression trees (De'ath and Fabricius, 2000) were used to visualise the responses of the three most numerically dominant ant species to soil texture.

For functional group composition, the variable of highest relative importance was a semi-quantitative variable, and so we used Permutational Analysis of

Variance (PERMANOVA, Anderson and Gorley, 2008) to test for differences in arthropod community composition among the classes of the predictor variable.

Significant main effects were further tested using pairwise t-tests, and displayed in the ordinations described above. PERMANOVA analyses were conducted in the PRIMER (Version 6) (Clarke & Gorley, 2006) + PERMANOVA (Anderson and Gorley, 2008) package.

RESULTS

There was substantial variation in patch area, and to a lesser extent distance to the nearest ten hectare remnant (Table 6.1). Patch-scale soil stability and number of obstructions showed moderate variation across the study area (Table

150

6. Multi-scale drivers of ant communities

6.1). There was, however, sufficient variation to examine how these attributes may influence ant community structure.

A total of 145 ant species (spp.) from 32 genera were recorded. Camponotus

(24 spp.), Melophorus (16 spp.), Monomorium (16 spp.) and Iridomyrmex (11 spp.) were the most taxon-rich genera recorded. Species accumulation curves approached plateau, indicating that a substantial proportion of species present had been sampled (data not shown).

Iridomyrmex septentrionalis comprised 24% of all individuals caught.

Rhytidoponera metallica and I. purpureus contributed a further 14% and 12%, respectively. Thirty–six species were represented by a single individual, and in total 68 species were represented by five or fewer individuals. Rhytidoponera metallica was also the most widespread species, found at 32 (91% of) sites.

Forty–six species were found only at a single site, and 99 species were found at five or fewer sites.

Ant abundance, species richness, and effective species diversity and environmental variables

Hierarchical partitioning analyses revealed that few variables were able to explain significant amounts of variation in ant abundance, species richness, and effective species diversity (Table 6.2). The total abundance of ants was negatively related to tree canopy cover, as well as patch area. Similarly, the total number of ant species significantly decreased with increasing patch area

(Table 6.2).

151

6. Multi-scale drivers of ant communities

Table 6.1: Descriptive statistics for the continuous predictor variables used in this study. SD = Standard deviation

Mean Median Range SD

Landscape context characteristics Patch area (ha) 58.40 16.10 0.1 – 363.7 94 Distance to nearest remnant 0.76 0.40 0.01 – 3.5 0.85 (TenHa)

Within-patch characteristics Soil stability 60.34 61.76 34.31 – 8.51 77.21 Number of obstructions 2.79 1.33 0 – 14 3.70 Total no. of vascular plant species 39.46 39 15 – 70 16.17 No. exotic vascular plant species 13.11 12 6 – 23 3.80

152

Table 6.2: Results of hierarchical partitioning analyses between landscape context, habitat characteristics and biotic predictors and ant abundance, richness, and effective species diversity for all ants, and within ant functional groups. Z = Z-score. I (%) is the independent contribution of variable to explained variance only, as compared with total variance explained by all variables in the model. Dominant Subordinate Generalised

All ants Dolichoderinae Camponotini Myrmicinae Opportunists I (%) Z I (%) Z I (%) Z I (%) Z I (%) Z Abundance Patch area (ha) 28.79 1.99 (-) 25.41 13.70 25.59 11.97 TenHa 5.40 5.48 6.65 5.91 10.30 Tree canopy cover 37.50 3.27 (-) 18.26 31.15 3.78 (-) 11.37 12.01 Litter cover 6.73 2.98 2.28 11.33 3.80 Plant community 13.01 7.19 7.94 13.57 17.08 Number of obstructions 2.38 23.99 1.35 9.32 29.66 1.73^(-) Soil stability 2.45 10.51 7.02 6.71 1.91 Soil clay content 3.75 6.18 17.42 1.69^(+) 7.60 0.33 Dominant taxa predictor n.a. n.a. 12.49 8.61 12.93 Total variation explained 35.75 27.00 48.70 27.80 35.50 Species richness Patch area (ha) 34.84 2.51 (-) 7.42 19.05 19.29 12.52 TenHa 5.48 14.79 2.88 7.03 23.34 Tree canopy cover 16.77 24.33 10.06 14.09 1.80 Litter cover 8.24 5.53 4.32 7.93 1.24 Plant community 15.78 5.31 9.34 28.08 4.85 Number of obstructions 10.70 2.07 1.37 17.76 6.40 Soil stability 7.15 9.94 21.79 1.04 6.36 Soil clay content 1.04 30.61 13.68 0.86 36.51

153

Dominant taxa predictor n.a. n.a. 17.50 3.91 6.98 Total variation explained 33.30 24.92 35.32 19.50 24.58 Effective species diversity Patch area (ha) 3.93 8.56 2.27 16.15 10.17 TenHa 42.71 2.62 1.22 2.16 4.20 Tree canopy cover 17.64 25.99 12.01 1.51 1.03 Litter cover 3.70 2.49 9.21 7.28 4.90 Plant community 2.59 7.74 5.06 30.43 5.08 Number of obstructions 1.95 41.60 2.92 (-) 34.02 2.03^ (-) 12.06 11.26 2.28(+) Soil stability 2.47 1.41 16.09 20.04 3.87 Soil clay content 25.01 9.59 7.71 2.00 42.40 Dominant taxa predictor n.a. n.a. 12.42 8.38 17.07 Total variation explained 9.00 33.87 28.67 12.68 27.91

Notes: Z-score only presented where statistically significant (Z ≥ 1.65). For significant results, direction of relationship between predictor and response variable (+ or -) is given. TenHa = distance to nearest ten hectare remnant Plant community represented the number of exotic species (higher values) and total vascular richness (lower values). For Subordinate Camponotini and Generalised Myrmicinae, the dominant taxa predictor was the number of Dominant Dolichoderinae. For Opportunists the predictor taxa was the number of Generalised Myrmicinae. ^ Indicates variable was statistically significant only where dominant taxa predictor was absent in analyses, though variance explained is from model including dominant taxa.

154

6. Multi-scale drivers of ant communities

Relationships among environmental predictors and functional groups were similarly variable and dependent upon the biodiversity measure in question

(Table 6.2). Only the number of obstructions was consistently important; more obstructions in the landscape resulted in a greater number of Opportunists, but fewer Dominant Dolichoderinae and Subordinate Camponotini. The number of

Subordinate Camponotini also decreased with increasing tree canopy cover, and increased with soil clay content.

No variable independently explained more than 14% of the total variation in ant abundance, species richness, or effective species diversity (Table 6.2). The total amount of variation explained by all variables together ranged from 9%

(total effective species diversity) to 49% (number of Subordinate Camponotini).

Ant species assemblages and environmental variables

In all DistLM analyses (species and functional groups), increasing the number of variables in the model increased the amount of variation in the species composition explained, but generally reduced model parsimony (reduced AICc weighting, Table 6.3).

Seven variables explained significant variation in ant species composition when included alone in a model (marginal DistLM, P ≤ 0.05; Table 6.4). The two statistical methods (DistLM and model averaging) yielded similar results, both highlighting that tree canopy cover, soil clay content, and soil stability were the best predictors of ant species assemblages (Table 6.3, 6.4). Ant species composition was best predicted by a model including tree canopy cover and soil

155

6. Multi-scale drivers of ant communities

clay content, though this was not necessarily the most parsimonious model

(Table 6.3). Overall, multi-model inference indicated that the variables that have the greatest effect on ant species composition are soil clay content, followed by tree canopy cover and soil stability (Table 6.4).

Many species showed only subtle, and not necessarily strong, statistical increases or decreases in relation to soil clay content (Figure 6.1a). For example, Camponotus sp. nr. consobrinus was more abundant at sites of high clay content, while Iridomyrmex sp. A and sp. D showed the opposite pattern

(Figure 6.1a). Of the 145 species recorded in this study, Indicator Species

Analysis showed that four species were statistically associated with particular levels of soil clay content. Pheidole sp. A (P = 0.05, Indicator Value (IndVal) =

49.1%, I. sp. D (P = 0.05, IndVal = 49.1%), and I. sp. B (P = 0.05, IndVal =

51.8%) were found predominantly at sites characterised by sand to sandy loam soils i.e. percentage clay ~15%. Pheidole sp. F was found mainly at sites of moderate (45%) clay content (P = 0.05, IndVal = 46.9%). Additionally, the three most common species illustrate the variation in relation to soil clay content. For

I. purpureus, abundance increased with increasing clay content, particularly at sites with >60% clay (Figure 6.2a). For I. septentrionalis, abundance was clearly highest at mid levels of clay content (~31–40%) (Figure 6.2b). Rhytidoponera metallica was generally more abundant above 22% soil clay content (Figure

6.2c).

156

6. Multi-scale drivers of ant communities

Table 6.3: DistLM results for species and functional groups composition using landscape context and within-patch habitat attributes. The best model for each number of variables is shown. All results are based on 511 models. Variable abbreviations: Can, Tree canopy cover (%); Clay, soil clay content (%); Stb, soil stability; Obs, number of obstructions; Plnt, plant community score; Area, patch area; Litt, cover of leaf litter; Grz, grazing (present/absent); TenHa, distance to nearest 10 ha remnant. RSS = residual sum of squares.

No. Variables selected AICc AIC AICc R2 RSS Variables weight rank

Species composition 1 Can 275.32 0.006 24 0.10 80544 2 Can, Clay 273.86 0.006 1 0.20 72143 3 Can, Clay, Stb 274.00 0.011 2 0.25 67322 4 Can, Clay, Stb, Obs 274.98 0.007 10 0.29 64031 5 Can, Clay, Stb, Obs, Plnt 276.24 0.0038 59 0.32 61045 6 Can, Clay, Stb, Obs, Plnt, Area 277.96 0.0016 217 0.35 58600 7 Can, Clay, Stb, Obs, Plnt, Area, Litt 279.85 0.0006 403 0.37 56139 8 Can, Clay, Stb, Obs, Plnt, Area, Litt, Grz 282.13 0.0002 496 0.40 53972 9 Can, Clay, Stb, Obs, Plnt, Area, Litt, Grz, TenHa 284.62 5.8x10-5 511 0.42 51740 Functional groups composition 1 Can -93.508 0.016 1 0.14 1.91 2 Can, Clay -93.446 0.016 2 0.20 1.78 3 Can, Clay, TenHa -92.619 0.010 5 0.24 1.69 4 Can, Clay, TenHa, Area -92.155 0.008 14 0.29 1.58 5 Can, Clay, TenHa, Area, Stb -91.854 0.007 19 0.34 1.46 6 Can, Clay, TenHa, Stb, Obs, Plnt -90.965 0.005 58 0.39 1.37 7 Can, TenHa, Area, Obs, Stb, Plnt, Litt -88.896 0.002 195 0.41 1.31 8 Can, TenHa, Area, Obs, Stb, Plnt, Litt, Clay -85.503 2.9x10-4 436 0.42 1.30 9 Can, TenHa, Area, Obs, Stb, Plnt, Litt, Clay, Grz -89.829 4.7x10-5 509 0.42 1.28

157

6. Multi-scale drivers of ant communities

Table 6.4: Relative contribution of each landscape context and within-patch variables to models explaining variation in ant species assemblage as calculated from marginal DistLM tests and multi-model inference. SS = sums of squares. TenHa = remnant ten hectares or greater in size. Multimodel inference is based on a series of models, while marginal DistLM results are calculated only when that variable is in the model.

Multi-model inference Marginal DistLM test

AICc rank Summed AICc weights SS Variation Pseudo- P

(trace) explained F Species composition 1 Soil clay content 0.543 8793.2 0.10 3.59 <0.001 2 Tree canopy cover 0.475 9191.3 0.10 3.77 <0.001 3 Soil stability 0.454 8307 0.09 3.37 <0.001 4 Number of obstructions 0.370 4215.9 0.05 1.63 0.050 5 Patch area (ha) 0.364 6634.9 0.07 2.63 0.001 6 Cover of leaf litter 0.330 5032.6 0.06 2.00 0.015 7 Plant community 0.324 7415.9 0.08 2.97 <0.001 8 Distance to TenHa 0.307 3819.9 0.04 1.47 0.091 9 Grazing 0.201 2401.8 0.03 0.91 0.547 Functional group composition 1 Tree canopy cover 0.668 0.322 0.14 5.41 0.001 2 Distance to TenHa 0.464 0.076 0.03 1.15 0.335 3 Soil clay content 0.394 0.156 0.07 2.40 0.069 4 Number of obstructions 0.392 0.147 0.06 2.25 0.081 5 Plant community 0.384 0.184 0.08 2.87 0.029 6 Patch area (ha) 0.375 0.214 0.10 3.39 0.017 7 Soil stability 0.372 0.151 0.07 2.37 0.076 8 Cover of leaf litter 0.372 0.045 0.02 0.66 0.622 9 Grazing 0.230 0.005 <0.01 0.07 0.944

158

6. Multi-scale drivers of ant communities

Figure 6.1: Ordination for ant a) species assemblage in relation to soil texture, using canonical analysis of principal coordinates and b) functional group assemblage in relation to tree canopy cover, using non-metric multidimensional scaling. Straight vectors represent increasing Pearson correlation for specific species or functional group in relation to site. For species ordination, only correlations for species >0.5 are shown. Vectors with arrowheads indicate increasing Pearson correlation for the second and third highest ranking environmental variables, in relation to site (abbreviations as per 6.4). Functional group abbreviations: DD: Dominant Dolichoderinae, SC: Subordinate Camponotini, GM: Generalised Myrmicinae, OPP: Opportunists. Variable names as per Table 6.2.

159

6. Multi-scale drivers of ant communities

Figure 6.2: Regression tree for soil clay content (%) and the three most abundant ant species: a) Iridomyrmex purpureus; b) I. septentrionalis; and c) Rhytidoponera metallica. The values at the root of each split indicate the value of soil clay content (%) for the specified node. The values at the leaf of the node are the mean abundance for the split. The vertical lengths are proportional to the variation explained by each split.

Functional group assemblages and environmental variables

Three variables explained significant variation in functional group composition when placed in a model alone (marginal DistLM, Table 6.4). The strongest and most parsimonious predictive model of functional group composition contained only tree canopy cover, though the addition of clay content had similar properties (Table 6.3). Across all models generated, tree canopy cover was the variable of highest relative importance in driving functional group composition,

160

6. Multi-scale drivers of ant communities

followed by distance to the nearest remnant exceeding 10 ha in area (TenHa;

Table 6.4).

Functional group composition differed significantly between sites of low (0-10%) and high (21-50%) tree canopy cover (PERMANOVA Pseudo-F2,34 = 3.14, P

(perm) = 0.006, Pairwise t = 2.18, P = 0.002; Figure 6.2b). These differences were driven primarily by five-times more Subordinate Camponotini (64 ± 16 mean ± standard error; vs. 11 ± 4) and four-times more Dominant

Dolichoderinae (238 ± 97 vs. 66 ± 20) at sites of low canopy cover (Figure

6.1b). Slight differences between functional group composition of medium and high canopy cover sites (Pairwise t = 1.60, P = 0.05) were the result of four- times more Generalised Myrmicinae in sites of moderate (11 to 20%) cover (82

± 44 vs. 23 ± 16 individuals), and nearly double the number of Dominant

Dolichoderinae (131 ± 34 vs. 72 ± 21). No differences existed between the composition of functional groups in sites of low and medium tree canopy cover

(Pairwise t =1.35, P = 0.26).

DISCUSSION

This study found that there is no single driver of ant community structure in an extensive area of grassy eucalypt woodland in south-eastern Australia. Overall, the ranking of environmental predictors in terms of relative importance to structuring the ant community was measure- and functional group-dependent.

This was particularly the case for abundance, species richness, and species diversity. Drivers of species and functional group composition were generally clearer, with soil texture (clay content) and tree canopy cover of primary

161

6. Multi-scale drivers of ant communities

importance, and to a lesser extent, distance of nearest remnant and the number of surface obstructions such as grass tussocks and fallen timber.

Within-patch versus landscape scale drivers of ant communities

Overall we found few consistent or significant predictors of ant abundance, species richness, or effective species diversity, regardless of the scale at which the predictor was based. This gave mixed support for the hypothesis that soil type and patch scale attributes such as vegetation structure would be stronger drivers of ant abundance, species richness, and diversity than landscape scale variables. For the total number of ants, a patch scale (tree canopy cover) attribute was of highest importance, followed by a landscape scale (patch area) variable. The number of obstructions was an important driver of effective species diversity for several functional groups and on two out of three occasions there was greater diversity when there were fewer obstructions (i.e. sites with more bare ground). These results corroborate research indicating that ecological disturbance results in changes to species composition, but not necessarily species richness or abundance (e.g. Holsten et al., 1995; Hoffmann,

2010; Hoffmann and James, 2011). This highlights the need to focus on assemblage composition when studying drivers of biotic communities, rather than on the abundance of animals or number of species within the community.

Using multimodel averaging, we had moderate support for our hypothesis that within-patch variation in habitat would be the strongest drivers of ant community composition. For species and functional groups, soil texture and tree canopy cover were the highest or amongst the highest predictors of composition.

162

6. Multi-scale drivers of ant communities

Additionally, the number of obstructions was relatively important for both species and functional group composition. An increasing body of evidence indicates a higher relative importance of vegetation structure and within-patch characteristics in structuring ant assemblages (Mazerolle and Villard, 1999;

Prugh et al., 2008). Key drivers of ant communities include tree canopy cover

(or conversely cover of bare ground, see below), soil texture, plant biomass, and habitat complexity (Abensperg-Traun et al., 1996; Debuse et al., 2007; Gibb and Parr, 2010). However, some studies have shown strong effects of landscape context on ants (e.g. Ås, 1999; Gibb and Hochuli, 2002; Sobrinho and Schoereder, 2007), or have demonstrated the importance of factors operating at multiple spatial scales (e.g. Spiesman and Cumming, 2008). We found only one landscape scale driver to be relatively important on one occasion, and only for functional group composition. Variable drivers of community structure are also evident for many other arthropod taxa, although patch-scale characteristics are usually the most dominant (e.g. spiders, Batary et al., 2008; butterflies, Poyry et al., 2009; invertebrates in general, Dauber et al., 2005; Thornton et al., 2011).

Soil texture and tree canopy cover drive ant assemblages

Soil texture was the primary driver of ant species composition. Variation in soil clay content resulted in subtle changes to the abundance of most species, rather than large, obvious changes to the community per se. Thus, only two

Iridomyrmex species and two Pheidole species showed detectable responses to soil clay content. Soil texture controls the distribution of ants through a variety of mechanisms. Clay content affects physiological tolerances and reproductive

163

6. Multi-scale drivers of ant communities

success, influencing the spatial distribution of species over relatively fine scales

(Johnson, 1992; 2000). Texture also controls soil water capacity and retention, and this influences ant nest building activity, as well as the characteristics and structural integrity of the nests (Whitford et al., 1999; Johnson, 2001). Our results are therefore consistent with a broad range of literature indicating the importance of soil texture in structuring ant communities (e.g. Kirkham and

Fisser, 1972; Whitford et al., 1999; Bestelmeyer and Wiens, 2001).

Tree canopy cover (or the lack thereof) was also a relatively important driver of ant communities. This was both in terms of composition and abundance, species richness, and species diversity. Three of the four major functional groups (Subordinate Camponotini, Dominant Dolichoderinae, and Opportunists) were more abundant where there was less tree canopy cover. A considerable body of ecological theory predicts a positive relationship between animal abundance and diversity and tree canopy cover or habitat complexity (the latter two variables were correlated in this study) (e.g. Andersen, 1986; Tews et al.,

2004). Contrary to theory, however, ant abundance and species richness often show strong negative relationships with tree cover and habitat complexity (e.g.

Lassau and Hochuli, 2004; Dauber et al., 2005). Most ants are thermophilic

(Hölldobler and Wilson, 1990), and therefore attain their highest productivity and richness in hot, open environments (Andersen, 2003). This may explain greater abundance and species richness at sites of lower shade and tree canopy cover in this study. Habitat complexity can also structure ant communities at finer scales by mediating species interactions (e.g. Gibb, 2005; Luque and Lopez,

2007). Our results support observations that many genera, and consequently

164

6. Multi-scale drivers of ant communities

functional groups (e.g. Dominant Dolichoderinae, Hot Climate Specialists,

Subordinate Camponotini), are more abundant in hot, open environments where the vegetation structure is simplified (Andersen, 1997; Vanderwoude et al.,

1997; Lassau and Hochuli, 2004; House et al., 2006).

Soil texture and tree canopy showed independent effects on the ant community.

At broader spatial scales soil texture often controls patterns of vegetation, resulting in marked shifts in vegetation structure and composition (e.g. Williams et al., 1996). Such changes to broad habitat type are known to affect ant communities (Bestelmeyer and Wiens, 2001; Andersen et al., 2007b). At our sites, variation in soil texture (sandy/sandy loam to clay) was associated with slight changes to the species composition of the overstorey and shrubby understorey vegetation. There were, however, no statistically significant relationships between soil texture and tree canopy cover, though sites characterised by fine-textured clay soils tended to be more open (David

Eldridge, unpublished data, 2012). Thus, within sites of a similar soil texture, there were still changes to ant community structure in relation to tree canopy cover. This suggests that the ant community is driven at a slightly broader scale by soil texture, and at finer scales by tree canopy cover.

It is important to recognise that a strong relationship exists between within- patch vegetation characteristics and broader scale landscape context.

Vegetation cover and landscape context, for example, are intrinsically linked

(Lindenmayer et al., 2008). We found a moderate positive correlation (α = 0.4), between tree canopy cover and patch area. It could be suggested, therefore,

165

6. Multi-scale drivers of ant communities

that tree canopy cover is, at least, a partial reflection of changes that have occurred to the patch at the landscape scale. Thus the effect of remnant size on the ant community could be mediated through tree canopy area, given that larger remnants tend to have greater tree cover (e.g. Gibbons and Boak, 2002).

We found that patch area and tree canopy cover were relatively important drivers of total ant abundance and species richness, all of which were weak, negative relationships. Furthermore, tree canopy cover was a strong driver of species and functional group composition. The statistical techniques we used are able to separate the independent effects of a given variable from the joint effects of that variable when considered with others. These analyses indicate that for total species richness, patch area appears to be the stronger driver of the two variables. For the total abundance of ants, as well as species and functional group composition, however, tree canopy cover has a far stronger effect.

Patch size and isolation are weak predictors of ant community structure

In our study, landscape-scale variables were generally poor predictors of ant community composition. The main exception was distance to the nearest ten hectare remnant, which was relatively important for functional group composition. This, however, appeared to be only a weak negative relationship with the Generalised Myrmicinae. For some functional groups, landscape context variables had a similar importance to patch-scale variables in predicting simple measures of community structure (abundance, species richness, effective species diversity), but in these cases few variables were clearly dominant. Landscape context is predicted to control species abundance and

166

6. Multi-scale drivers of ant communities

richness for various reasons, many of which are explained by theory of island biogeography (e.g. species–area relationships, distance to nearest habitat remnants, etc., MacArthur and Wilson, 1967; Losos and Ricklefs, 2010).

Remnant size is predicted to be an important driver of animal communities in fragmented landscapes. Specifically, remnant size interacts with the distance of the remnant to sources of dispersal (i.e. other remnants), with a predicted negative relationship between immigration and distance to nearest remnant

(e.g. MacArthur and Wilson, 1967). These concepts are readily applicable for larger animals in fragmented habitats (e.g. Fahrig, 2003). Smaller animals such as invertebrates, however, are more likely to respond to large changes in habitat structure scaled according to their behaviours and movements, which may explain why remnant size was of minor relative importance in this study.

Contemporary theories of landscape change are broadly applied across taxa, with little implicit consideration of scale-specific issues. This is one of the reasons why broad scale predictors such as patch area do not consistently predict arthropod responses and are not immediately applicable to the study of invertebrates (e.g. Abensperg-Traun and Smith, 1999; Driscoll and Weir, 2005;

Debuse et al., 2007). While predictions like those made from Island

Biogeography Theory can be used successfully to gain insights into invertebrate ecology, there is a need to develop specific frameworks that have been scaled appropriately to the target taxa.

167

6. Multi-scale drivers of ant communities

Biotic interactions do not predict ant community structure

In this study, biotic interactions were weak between the majority of functional groups. Analysis of these interactions was based on the use of the abundance of dominant functional groups (generally the Dominant Dolichoderinae) as predictors of the abundance or richness of other functional groups. These predictors, therefore, were much weaker than vegetation characteristics in explaining variation in ant communities. This agrees with recent research indicating that broad habitat type is a more important driver of ant communities than biotic interactions (Narendra et al., 2011). A wealth of research, however, indicates that ant community structure can be controlled by biotic interactions between functional groups, at least at fine scales and among individual species

(e.g. Gibb, 2005). These interactions may be direct (e.g. the Dominant

Dolichoderinae are widely known to suppress the Subordinate Camponotini,

Andersen and Patel, 1994; Andersen, 1997), or indirect (e.g. Arnan et al.,

2011). It should be noted, however, that our analyses of biotic interactions were based at the functional group level. It is quite possible that strong interactions are occurring between species of different functional groups, and these would not necessarily be detected without additional, detailed analyses (e.g. Debuse et al., 2007). While there is little doubt that biotic interactions between ant functional groups can be important determinants of community structure, further research is necessary to determine under what environmental and habitat conditions these interactions predominate.

168

6. Multi-scale drivers of ant communities

Research and conservation implications

The functional group classification scheme for Australian ants (Andersen, 1995) is proposed to be least useful in arid and semi-arid landscapes (Hoffmann and

Andersen, 2003). This is because these ecosystems have a relatively open, simplified vegetation structure. Disturbance to these ecosystems does not necessarily result in large, site-level changes to habitat structure, which are the changes that typically affect ant communities. This minimises the effects of the disturbance on the fauna. We found, however, that there appear to be detectable functional group patterns in relation to the environmental predictors that we used. While the utility of the functional group scheme may be reduced in xeric ecosystems, it can still illustrate potentially meaningful ecological responses within the community.

There is an increasing demand for accurate indicators of arthropod biodiversity values for management, restoration and conservation purposes (e.g. Karr and

Kimberling, 2003). Ants have been proposed as a highly useful and relevant taxon for this purpose because of their abundance, ease of sampling, and response to a broad range of ecological disturbance (Andersen et al., 2002;

Andersen and Majer, 2004; Sanford et al., 2009). Our study indicates that ants may be useful as indicators of ecological disturbance, but there are several caveats. Firstly, it is imperative that any studies of arthropod communities, in relation to ecological change or variation, include assessments of species and/or functional group composition. Secondly, careful attention must be paid to the ecological change that is of concern, and to how it could actually affect ants.

In our study we investigated many variables potentially driving ant communities,

169

6. Multi-scale drivers of ant communities

but of these only a select few showed meaningful and statistically significant relationships with the ant community. If these issues are taken into consideration ants may be useful as a reflection of ecological change.

One of the pressing conservation concerns apparent from this study is the rarity of many ant species. Many of the ant species in our study area were found at a low number of sites, and/or in relatively low abundance. Conservation or land management for the benefit of such species is difficult, and efforts need to be highly site-specific and targeted. However a lack of ecological information on the majority of these species hampers any sort of management for their benefit.

Further study is required to determine the exact rarity of these species.

Conservation in agricultural landscapes needs to manage multiple land uses to maximise biodiversity. These land uses often contrast in terms of their vegetation structure, such as an agricultural paddock versus a remnant woodland set aside for nature conservation. It would be ideal if we were able to reliably predict the composition of a species assemblage based on simple environmental predictors, such as tree canopy cover, or soil clay content. For ants, however, individual species often vary subtly in relation to ecological variation. Thus, rather than there being, for example, a distinct suite of species at sites of low or high tree canopy cover, there are subtle changes to many species and subsequently the overall composition of the community. How these dynamics can be quantified, monitored, and managed appropriately within a conservation context is a significant emerging research challenge.

170

6. Multi-scale drivers of ant communities

ACKNOWLEDGEMENTS

We gratefully acknowledge all those involved in the Savernake and Native Dog

(SAND) Farmscapes project. We would also like to acknowledge Alan Andersen and Tony Hertog, who identified the ants.

171

7. General discussion

Chapter 7

General discussion

172

7. General discussion

The objective of this thesis was to examine how the spatial distribution of resource patches affected arthropod communities at multiple spatial scales in dryland woodland ecosystems. Specifically, I investigated how resource concentration affected arthropod community composition and structure at the scale of individual trees or shrubs, and at broad, landscape-scales in fragmented landscapes. Study sites ranged from encroached shrublands, to fragmented open grassy woodlands to low, open woodlands. A key characteristic of each of these ecosystems was the spatially patchy distribution of resources due to the clumping of perennial vegetation and their associated patches. In each ecosystem the structure and composition of the arthropod community was heavily influenced by the distribution and nature of these resource patches. This was clearly evident at fine spatial scales, where trees or shrubs formed distinct patches that affected the biota at ordinal and species levels.

Key findings of this thesis

Chapter 2 described a field-based study that investigated how plant density and species structured plant-resident arthropods in an encroached shrubland ecosystem co-dominated by two shrub species. It revealed that shrub species is the dominant driver of arthropod distribution in this eastern Australian shrubland. Eremophila sturtii supported roughly five-times more arthropods

(Collembola, Psocoptera, and Hemiptera) than Senna artemisioides subsp. filifolia. Importantly, however, each shrub species supported its own unique hemipteran species assemblage. In contrast, there were virtually no differences

173

7. General discussion

between the arthropod communities found on shrubs growing in low and high– density patches. In the context of shrub encroachment, these findings highlight that changes to the composition of vegetation could have dramatic consequences for plant-resident fauna.

Chapters 3 and 4 investigated the role of mallee (Eucalyptus spp.), a dominant tree, in structuring arthropod communities at fine and broad–spatial scales in south–eastern Australia. These Chapters demonstrated that modulation of the arthropod community occurs through the creation and maintenance of a multilayered resource patch beneath the tree canopy. Such multi-layered patches are distinct from the inter-tree patches that are dominated by bare soil with scattered perennial grasses. Specifically, Chapter 3 demonstrated clear differences between the arthropod communities supported by either canopy or open patches. Several taxa (isopods, spiders, and wasps) displayed strong associations with the canopy patch, while ants were more abundant in open patches. However, experimental manipulation of habitat complexity, through the provision of shade and/or woody debris, did not noticeably affect the arthropod community.

Chapter 4 further developed the theme of landscape modulation in mallee ecosystems, specifically examining how it is affected by disturbance (fire). The main outcome of this Chapter is that modulation of the arthropod community was evident as quickly as four years after a fire. Thus, many taxa (e.g. spiders, cockroaches, and wasps) were strongly associated with the canopy patch both

174

7. General discussion

in an area burnt four years ago and one burnt 30 years ago. A particularly interesting result from this study is the dependence that isopods appear to have on the canopy patch, and also on habitat that has not been burnt for many years. Chapters 3 and 4 also highlighted the intrinsic links between resource patchiness, ecological disturbance, and the composition of the biota, at fine- and broad-spatial and taxonomic scales.

Moving to a broader, landscape-scale focus on resource concentration, Chapter

5 evaluated whether two common measures of landscape health could be used as surrogates for arthropod (ant, spider, and beetle) biodiversity in a semi-arid woodland. Neither indices derived using Landscape Function Analysis (fine- scale) nor those encompassed in a terrestrial index of ecological/biotic integrity

(broad-scale) showed strong, predictable relationships with arthropod abundance, species richness, or diversity. The indices of landscape health likely need to be scaled down for use with invertebrates. However, this necessitates specific methodology tailored to arthropod ecology. Detailed information about arthropod ecology, necessary to develop this methodology, does not currently exist.

Continuing the theme of broad, multi-scale drivers of arthropod community structure, Chapter 6 focused on identifying drivers of ant assemblages in a fragmented grassy–box woodland in south-eastern Australia. We investigated the relative importance of landscape context, within-patch habitat characteristics

(vegetation structure, litter cover, etc.), and ant–ant interactions in structuring

175

7. General discussion

ant species and functional group assemblages. Generally, within-patch characteristics, particularly tree canopy cover and soil texture, had the strongest influence on ant species and functional groups. Landscape context characteristics (size of patch and distance to nearest remnant) were relatively poorer predictors, as were biotic interactions between functional groups. This work demonstrated that ant communities are driven predominantly by changes that occur to habitat at finer spatial scales. It also illustrated how relationships between biotic communities and the abiotic environment vary depending on which component of the biotic community is investigated, and how this must be considered when studying drivers of biotic communities.

Issues relevant to ecological studies in dryland ecosystems

Arthropod communities exhibit substantial temporal variability, which is a significant issue for ecological research, particularly in resource-limited environments. All Chapters in this thesis were 'snap shot' studies, designed to elucidate the processes structuring arthropod communities at one particular point in time. In Australian semi-arid and arid ecosystems, the availability of resources is unpredictable and can dramatically increase and decrease over short (e.g. season to season) and long (e.g. annual, decadal) temporal scales, particularly in response to patterns of rainfall (Morton et al., 2011). This may influence how resource patchiness affects the biota. For example, Chapter 2 was conducted in early winter of a relatively rainfall-poor year. Sampling in warmer months, or after large rainfall events, could yield potentially different results due to flowering or enhanced plant growth. Chapters 3 and 4 were conducted following a year of remarkably high rainfall (2010), and while this

176

7. General discussion

would be expected to have resulted in greater arthropod abundances, in some parts of the study area arthropod abundances appeared to be lower (A. Kwok, personal observations). The response of arthropods to long-term temporal variation in resource availability is unknown for arid ecosystems.

Regional scale variability in arthropod communities is also an important issue to consider in studies of arthropod ecology. Spatial variability and turnover of communities across broader scales (e.g. beta diversity), for example across ecosystems, landscapes, and bioregions was not accounted for in this study.

Broad scale variability restricts wider generalisation of results found. In particular, for studies based at broader spatial scales (e.g. Chapters 5 and 6), patterns observed in relation to landscape health or resource concentration may be quite different to other fragmented landscapes, where the processes responsible for (and consequences of) fragmentation are likely to be different.

Similarly, patterns of host-specificity amongst phytophagous species complicates quantification of beta diversity as little is known of the ecology of these taxa (Odegaard, 2006).

Chapter 4 used a space-for-time substitution approach to investigate the effect of fire on the arthropod community. It is well acknowledged that this approach may result in erroneous conclusions being made (see Walker et al., 2010). This approach is, however, widely used in studies examining the effects of fire (e.g.

Potts et al., 2003; Haslem et al., 2011; Kelly et al., 2011). In the case of Chapter

4, the "burnt" and "unburnt" communities were 10 km apart, but were part of a

177

7. General discussion

large, contiguous area of mallee vegetation, likely supporting similar arthropod communities and subject to the same patch creation, destruction and colonisation processes irrespective of fire. For this reason, we have confidence in our conclusions regarding arthropods, landscape modulation, and fire.

Patchy begets patchy: arthropods are distributed in relation to resources

The concentration of physical resources into discrete patches is a universal characteristic of dryland ecosystems. It is only recently, however, that information has detailed how this affects biotic communities, not simply species activity. The results of this thesis indicate that arthropods are critically dependent upon resources provided by perennial vegetation, whether this is plant biomass (e.g. Hemiptera), and/or the ambient conditions or habitat created by the plants. Indeed, for some arthropod orders (e.g. isopods) the conditions necessary for survival are likely only found under the tree canopy or other forms of vegetation. Similar ecological filters are likely placed on species, as shown in this thesis and by other studies on ants (e.g. Oliver et al., 2006; Carpintero et al., 2011). I therefore suspect that resource patches associated with perennial vegetation structure arthropod communities across a variety of dryland ecosystems, particularly at finer spatial scales. In dryland woodland ecosystems, each landscape element provides a different variety of resources and creates different ecological conditions, which have direct effects on arthropods, and likely other faunal communities.

178

7. General discussion

The concentration of arthropods into resource patches is likely to be important for ecosystem function. In this study, resource patches supported a range of detritivores (e.g. isopods, silverfish). These animals are important for releasing nutrients stored in leaf litter and affecting decomposition processes (e.g. Arriaga and Maya, 2007; Bastow, 2011). Additionally, many predatory and omnivorous taxa are concentrated into resource patches. It is also likely that soil microbial and microarthropod communities are substantially larger in canopy patches

(e.g. Noble et al., 1996). Thus, over time, the activity of all of these taxa reinforces the fertility of the patch through positive feedback mechanisms.

Where patches concentrate resources, resources concentrate the arthropod community, forming a distinct, self-sustaining ecosystem. Given their abundance, arthropods are likely a critical component of this system.

Implications of this research

There are several important implications of the work embodied in this thesis.

Firstly, elements of perennial vegetation are crucial to maintaining biodiversity at multiple spatial scales. The distribution of perennial vegetation influenced the resident arthropod community in every ecosystem studied in this thesis.

Perennial vegetation directly and indirectly supports a broad range of arthropod taxa. In parallel, open areas between vegetation also support particular suites of arthropods. This is important for biodiversity management as disturbance (e.g. fire, grazing) has the potential to remove or degrade landscape elements, altering community composition. Thus, appropriate management of landscape elements and disturbances is crucial for maintaining desired biodiversity values,

179

7. General discussion

as shown recently by studies in dryland ecosystems (Lumsden and Bennett,

2005; Barton et al., 2009; Manning et al., 2009).

The second implication of the work in this thesis is that each dryland vegetation community (mallee, shrubland, box woodland), supports a unique arthropod assemblage. While it is difficult to determine the exact overlap of species between vegetation communities, many taxa attain greater abundance in a specific community because of the composition of the vegetation (Tews et al.,

2004). For example, shrublands are dominated by taxa directly dependent upon the plant biomass (Hemiptera, Psocoptera, Collembola). Woodlands, in contrast, are likely to support a mixture of plant-resident fauna and ground- active fauna (ants, spiders, beetles) due to the greater overall structural complexity of the vegetation. Additionally, many of these taxa may be restricted to particular communities due to fine-scale processes (e.g. isopods to moister microhabitats, Hemiptera to a particular host species).

The final implication of this research is that broader, landscape-scale characteristics may not be as important as fine-scale environmental characteristics in influencing arthropod biodiversity. Traditionally, biodiversity management has focussed on broad-scale ecological processes (e.g. Schwartz,

1999) and the creation of large conservation reserves (Lindenmayer et al.,

2002). Increasingly though, as shown in Chapters 5 and 6, large scale ecological processes may not be driving arthropod biodiversity unless they directly or indirectly affect finer scale aspects of habitat (such as the canopy cover provided by trees within a remnant). This emphasises the need to

180

7. General discussion

consider fine-scale landscape variation and resource concentration when managing for arthropod biodiversity.

Specific avenues of future research

Taxonomic resolution is a critical issue in any scientific study of biodiversity.

This is particularly the case with invertebrates, where a large numbers of species remain unidentified and undescribed (New, 1995). In this thesis each chapter focussed on a different taxonomic level, ranging from species (Chapters

2, 5 and 6) to order (or broad taxonomic group only) (Chapters 3 and 4). Both species and broader taxon analyses provide complementary information. For example, in mallee woodlands there are processes that affect entire arthropod orders in different ways (Chapters 3 and 4). These patterns are not always evident when focussing on fine taxonomic (e.g. species) scales. However, focusing on broad taxonomic patterns can obscure species-specific patterns.

This was also hinted at in Chapters 3 and 4 for spider morphospecies. Clearly, identification of specimens to the species or morphospecies level is ideal in all studies of arthropod ecology, as this permits investigations across a range of taxonomic scales. Detailed taxonomic analyses would have been beneficial in this thesis, and would have shed additional light onto processes affecting the distribution of arthropods in arid woodlands.

Shrub encroachment is a global issue in dryland ecosystems. Despite this, virtually nothing is known about how it affects plant-resident arthropods. The two most pressing issues for arthropod-encroachment research are how

181

7. General discussion

encroachment affects arthropod communities (plant-resident and ground- dwelling), and whether changes to arthropod-plant interactions are occurring within encroached landscapes. Our study was limited as we only used two shrub species in one shrubland community. Plant composition in encroached landscapes is variable, and encroachment may involve up to ten coexisting shrub species (Ayers et al., 2001), all varying in structural complexity and relatedness. While studies have highlighted the host-specificity of Hemiptera

(Moir et al., 2010), further research is necessary to determine the extent to which it occurs in shrub-encroached landscapes.

Whether fertile patches universally support unique arthropod communities in dryland ecosystems remains largely unknown. It seems likely that they do, given that the processes responsible for patch creation are universal (Garner and Steinberger, 1989), as are the abiotic consequences. Studies investigating the role of fertile or modulated patches in structuring arthropod communities in other dryland ecosystems are necessary in order to determine the universality of the modulation phenomenon. Furthermore, while arthropod responses to fire are often researched (e.g. Andersen, 1991; Blanche et al., 2001), the extent to which this is driven by fire-induced shifts in resource patchiness is poorly understood.

Mallee vegetation is intrinsically linked to fire, and it was beyond the scope of this study to comprehensively determine how fire affects landscape modulation and patch-dependent or patch-sensitive fauna. In the mallee of eastern

182

7. General discussion

Australia, a spectrum of sites of varying time since fire exists, ranging from a few days to decades or centuries (Haslem et al., 2011). Emerging evidence indicates how substantially the resource patch in these mallee communities changes over long periods of time following fire (Haslem et al., 2011; Samantha

Travers, unpublished data. 2012), and how this affects the persistence of some vertebrates (Kelly et al., 2011). How fire and associated changes to resource patches affect the ground-dwelling arthropod community over long time scales remains largely unknown, though we suspect these effects to be substantial.

Establishing such relationships is crucial to effective conservation of arthropod biodiversity.

Indices derived from landscape (or ecosystem) function analysis have been proposed to show links to faunal biodiversity (Ludwig et al., 1999; Ludwig et al.,

2004). While the research outlined in Chapters 5 and 6 did not reveal clear relationships between health indices and measures of arthropod biodiversity in agricultural landscapes, the links among components of landscape health and biodiversity were apparent at finer spatial scales. In the mallee woodlands I found that perennial obstructions (trees and their associated patches), on their own, are important elements for arthropod biodiversity, allowing the persistence of some arthropod taxa. Intuitively this should translate into broader-scale relationships among biodiversity and these measures of landscape health (such as the number of obstructions). This was not strictly the case in the other eucalypt woodlands I examined, however the sampling methodology may have obscured potential patterns (see below). In such ecosystems, however, tree cover was an important predictor of arthropod communities. The role of plant-

183

7. General discussion

created resource patches across multiple spatial scales has not been examined widely within single ecosystems. Research that focuses on the resources provided by perennial vegetation at multiple spatial scales is necessary to gain insight into why patches drive biotic communities at fine-scales, but do not translate into clear patterns at broader scales.

Toward a holistic understanding of arthropod ecology

A holistic understanding of arthropod ecology requires consideration of processes occurring at multiple spatial scales. In this thesis, arthropods exhibited patterns in relation to the fine- and broad-scale patterns of perennial vegetation in the landscape. This pattern was much stronger at finer scales, where the arthropod community can be broadly predicted within a patch associated with a tree or shrub, relative to adjacent unvegetated areas. Yet the quantity of these fine-scale elements (fertile patches) was a poor predictor of arthropod biodiversity at broader, landscape scales. While inferential studies such as Chapters 5 and 6 are important for gaining insight into multi-scale drivers of biodiversity, they do not resolve the detail necessary to determine what is affecting arthropods at finer scales. Typically, they encapsulate fine- scale variation in the form of the cover of leaf litter, though the scale at which this variation is measured may be too coarse to yield meaningful insights. Our knowledge of arthropod ecology can be significantly enhanced through the consideration of multiple spatial scales or finer details within a given scale (i.e. litter composition).

184

7. General discussion

A key issue that emerges from my results is the value of a comprehensive understanding of arthropod ecology within a single system. While studies across ecosystems (e.g. this thesis) are crucial for increasing our understanding of universal processes important to arthropods, they are single elements in a multi-element system. There is a need to determine fine- and broad-scale processes (as above) within a single community. For example, while I elucidated clear effects of resource patches on arthropods in a mallee system, we are unaware of how this translates into broader scale patterns of biodiversity in this community. The sampling designs used in Chapters 5 and 6 focussed on broader scale patterns, and may not be able to capture processes operating at predominantly finer spatial scales (e.g. modulation and the importance of fine scale vegetation elements, as in Chapters 3 and 4). A hierarchical nested sampling design would have been more appropriate, and indeed is necessary to determine whether the importance of patches and vegetative elements at finer scales manifests in changes to faunal communities at broader, landscape scales.

Perhaps the most important knowledge gap highlighted by this thesis is the need to further quantify the role that arthropods play in the functioning of dryland ecosystems. Clearly, dryland ecosystems support a range of biotic components, including detritivores, omnivores, predators, herbivores, and phytophages. While I elucidated some of the patterns affecting the distribution of these taxa in eastern Australia, very little is known about the ecosystem services and function that these taxa provide. Furthermore, we know little of the interactions among taxa, which may have substantial ecological consequences,

185

7. General discussion

for example affecting nutrient cycling (e.g. Lawrence and Wise, 2000) and trophic and species-level community structure (Megias et al., 2011). While it is well acknowledged that arthropods play a central role in a variety of ecological processes (e.g. Whitford, 2000; Lavelle et al., 2006), in many dryland ecosystems quantitative information is glaringly absent. This is certainly the case in Australia, despite 70% of the continent being arid or semi-arid. Indeed, while it is highlighted that there is a pressing need for data regarding arthropod ecology (e.g. Morton et al., 2011), this does not generate the research attention it deserves. Arthropods have, and will continue to be integral components of terrestrial ecosystems. A greater understanding of their lives will enrich our appreciation of not only these fascinating creatures, but also the world within which they live.

186

REFERENCES

Abbott, I. (1984) Changes in the abundance and activity of certain soil and litter fauna in the Jarrah forest of Western Australia after a moderate intensity fire. Australian Journal of Soil Research, 22, 463-469.

Abbott, I., Burbidge, T., Strehlow, K., Mellican, A., and Wills, A. (2003) Logging and burning impacts on cockroaches, crickets and grasshoppers, and spiders in Jarrah forest, Western Australia. Forest Ecology and Management, 174, 383- 399.

Abensperg-Traun, M., Smith, G.T., Arnold, G.W., and Steven, D.E. (1996) The effects of habitat fragmentation and livestock-grazing on animal communities in remnants of gimlet Eucalyptus salubris woodland in the Western Australian wheatbelt.1. Arthropods. Journal of Applied Ecology, 33, 1281-1301.

Abensperg-Traun, M. and Smith, G.T. (1999) How small is too small for small animals? Four terrestrial arthropod species in different-sized remnant woodlands in agricultural Western Australia. Biodiversity and Conservation, 8, 709-726.

Agra, H. and Ne'eman, G. (2009) Woody species as landscape modulators: their effect on the herbaceous plants in a Mediterranean maquis. Plant Ecology, 205, 165-177.

Agrawal, A.A., Lau, J.A., and Hamback, P.A. (2006) Community heterogeneity and the evolution of interactions between plants and insect herbivores. Quarterly Review of Biology, 81, 349-376.

Aguiar, M.R. and Sala, O.E. (1999) Patch structure, dynamics and implications for the functioning of arid ecosystems. Trends in Ecology & Evolution, 14, 273- 277.

Allan, G.E. and Southgate, R.I. (2002) Fire regimes in the spinifex landscapes of Australia. Flammable Australia: The fire regimes and biodiversity of a continent (ed. by R.A. Bradstock, J.E. Williams & A.M. Gill), pp. 145-176. Cambridge University Press, Cambridge.

Andersen, A.N. (1986) Diversity, seasonality and community organization of ants at adjacent heath and woodland sites in southeastern Australia. Australian Journal of Zoology, 34, 53-64.

Andersen, A.N. (1991) Responses of ground-foraging ant communities to 3 experimental fire regimes in a savanna forest of tropical Australia. Biotropica, 23, 575-585.

Andersen, A.N. and Patel, A.D. (1994) Meat ants as dominant members of Australian ant communities - An experimental test of their influence on the foraging success and foraging abundance of other species. Oecologia, 98, 15- 24.

187

Andersen, A.N. (1995) A classification of Australian ant communities, based on functional groups which parallel plant life-forms in relation to stress and disturbance. Journal of Biogeography, 22, 15-29.

Andersen, A.N. (1997) Functional groups and patterns of organization in North American ant communities: A comparison with Australia. Journal of Biogeography, 24, 433-460.

Andersen, A.N. and Sparling, G.P. (1997) Ants as indicators of restoration success: Relationship with soil microbial biomass in the Australian seasonal tropics. Restoration Ecology, 5, 109-114.

Andersen, A.N. and Muller, W.J. (2000) Arthropod responses to experimental fire regimes in an Australian tropical savannah: ordinal-level analysis. Austral Ecology, 25, 199-209.

Andersen, A.N., Hoffmann, B.D., Muller, W.J., and Griffiths, A.D. (2002) Using ants as bioindicators in land management: simplifying assessment of ant community responses. Journal of Applied Ecology, 39, 8-17.

Andersen, A.N. (2003) Ant biodiversity in arid Australia: productivity, species richness and community organisation. Records of the South Australian Museum Monograph Series, 7, 79-92.

Andersen, A.N., Fisher, A., Hoffmann, B.D., Read, J.L., and Richards, R. (2004) Use of terrestrial invertebrates for biodiversity monitoring in Australian rangelands, with particular reference to ants. Austral Ecology, 29, 87-92.

Andersen, A.N. and Majer, J.D. (2004) Ants show the way Down Under: invertebrates as bioindicators in land management. Frontiers in Ecology and the Environment, 2, 291-298.

Andersen, A.N., Cook, G.D., Corbett, L.K., Douglas, M.M., Eager, R.W., Russell-Smith, J., Setterfield, S.A., Williams, R.J., and Woinarski, J.C.Z. (2005) Fire frequency and biodiversity conservation in Australian tropical savannas: implications from the Kapalga fire experiment. Austral Ecology, 30, 155-167.

Andersen, A.N., Hertog, T., and Woinarski, J.C.Z. (2006) Long-term fire exclusion and ant community structure in an Australian tropical savanna: congruence with vegetation succession. Journal of Biogeography, 33, 823-832.

Andersen, A.N., Parr, C.L., Lowe, L.M., and Muller, W.J. (2007a) Contrasting fire-related resilience of ecologically dominant ants in tropical savannas of northern Australia. Diversity and Distributions, 13, 438-446.

Andersen, A.N., Van Ingen, L.T., and Campos, R.I. (2007b) Contrasting rainforest and savanna ant faunas in monsoonal northern Australia: a rainforest patch in a tropical savanna landscape. Australian Journal of Zoology, 55, 363- 369.

188

Anderson, M.J. and Gorley, R.N. (2008) PERMANOVA+ for PRIMER: Guide to software and statistical methods. Primer-E: Plymouth.

Andreasen, J.K., O'Neill, R.V., Noss, R., and Slosser, N.C. (2001) Considerations for the development of a terrestrial index of ecological integrity. Ecological Indicators, 1, 21-35.

Andrew, N., Rodgerson, L., and York, A. (2000) Frequent fuel-reduction burning: the role of logs and associated leaf litter in the conservation of ant biodiversity. Austral Ecology, 25, 99-107.

Andrew, N.R. and Hughes, L. (2005) Diversity and assemblage structure of phytophagous Hemiptera along a latitudinal gradient: predicting the potential impacts of climate change. Global Ecology and Biogeography, 14, 249-262.

Archer, S.R. (2010) Rangeland conservation and shrub encroachment: new perspectives on an old problem. Wild rangelands: conserving wildlife while maintaining livestock in semi-arid ecosystems (ed. by J.T. du Toit, R. Kock & J.C. Deutsch), pp. 53-97. Blackwell Publishing Ltd., UK.

Arnan, X., Gaucherel, C., and Andersen, A.N. (2011) Dominance and species co-occurrence in highly diverse ant communities: a test of the interstitial hypothesis and discovery of a three-tiered competition cascade. Oecologia, 166, 783-794.

Arriaga, L. and Maya, Y. (2007) Spatial variability in decomposition rates in a desert scrub of Northwestern Mexico. Plant Ecology, 189, 213-225.

Ås, S. (1999) Invasion of matrix species in small habitat patches. Conservation Ecology, 3, 1-5.

Ata Rezaei, S., Arzani, H., and Tongway, D. (2006) Assessing rangeland capability in Iran using landscape function indices based on soil surface attributes. Journal of Arid Environments, 65, 460-473.

Attie, M., Bourgoin, T., Veslot, J., and Soulier-Perkins, A. (2008) Patterns of trophic relationships between planthoppers (Hemiptera : Fulgoromorpha) and their host plants on the Mascarene Islands. Journal of Natural History, 42, 1591- 1638.

Australian Wildlife Conservancy (2011) Scotia's Climate, http://www.australianwildlife.org/AWC-Sanctuaries/Scotia-Sanctuary/Scotia- Climate.aspx. Last accessed 16/2/12.

Ayers, D., Melville, G., Bean, J., Beckers, D., Ellis, M., Mazzer, T., and Freudenberger, D., eds. (2001) Woody weeds, biodiversity and landscape function in western New South Wales. WEST 2000, Dubbo, NSW.

189

Bach, C.E. (1980) Effect of plant density and diversity on the population dynamics of a specialist herbivore, the stripped cucumber beetle, Acalymma vittata (Fab.). Ecology, 61, 1515-1530.

Bach, C.E. (1988) Effects of host plant patch size on herbivore density - underlying mechanisms. Ecology, 69, 1103-1117.

Baldi, A. (2008) Habitat heterogeneity overrides the species-area relationship. Journal of Biogeography, 35, 675-681.

Bardgett, R.D., Usher, M.B., and Anderson, J.M. (2005) Biological diversity and function in soils. Cambridge University Press, Cambridge.

Barton, P.S., Manning, A.D., Gibb, H., Lindenmayer, D.B., and Cunningham, S.A. (2009) Conserving ground-dwelling beetles in an endangered woodland community: Multi-scale habitat effects on assemblage diversity. Biological Conservation, 142, 1701-1709.

Bastow, J.L. (2011) Facilitation and predation structure a grassland detrital food web: the responses of soil nematodes to isopod processing of litter. Journal of Animal Ecology, 80, 947-957.

Batary, P., Baldi, A., Samu, F., Szuts, T., and Erdos, S. (2008) Are spiders reacting to local or landscape scale effects in Hungarian pastures? Biological Conservation, 141, 2062-2070.

Beattie, J.A. (1972) Groundsurfaces of the Wagga Wagga Region, New South Wales Soil Publication No. 28. CSIRO, Melbourne.

Beck, J. and Schwanghart, W. (2009) Comparing measures of species diversity from incomplete inventories: an update. Methods in Ecology and Evolution, 1, 38-44.

Bennett, L.T., Kasel, S., and Tibbits, J. (2009) Woodland trees modulate soil resources and conserve fungal diversity in fragmented landscapes. Soil Biology & Biochemistry, 41, 2162-2169.

Benshemesh, J. (1989) Management of the malleefowl - with regard to fire. The mallee lands: a conservation perspective (ed. by J.C. Noble, P.J. Joss & G.K. Jones), pp. 206-211. CSIRO Publishing, East Melbourne.

Bestelmeyer, B.T. and Wiens, J.A. (2001) Ant biodiversity in semiarid landscape mosaics: The consequences of grazing vs. natural heterogeneity. Ecological Applications, 11, 1123-1140.

Bestelmeyer, B.T. (2005) Does desertification diminish biodiversity? Enhancement of ant diversity by shrub invasion in south-western USA. Diversity and Distributions, 11, 45-55.

190

Bickel, T.O., Bruhl, C.A., Gadau, J.R., Hölldobler, B., and Linsenmair, K. (2006) Influence of habitat fragmentation on the genetic variability in leaf litter ant populations in tropical rainforests of Sabah, Borneo. Biodiversity and Conservation, 15, 157-175.

Bisevac, L. and Majer, J.D. (1999) Comparative study of ant communities of rehabilitated mineral sand mines and heathland, Western Australia. Restoration Ecology, 7, 117-126.

Blanche, K.R., Andersen, A.N., and Ludwig, J.A. (2001) Rainfall-contingent detection of fire impacts: Responses of beetles to experimental fire regimes. Ecological Applications, 11, 86-96.

Blaum, N., Rossmanith, E., Fleissner, G., and Jeltsch, F. (2007a) The conflicting importance of shrubby landscape structures for the reproductive success of the yellow mongoose (Cynictis penicillata). Journal of Mammalogy, 88, 194-200.

Blaum, N., Rossmanith, E., Popp, A., and Jeltsch, F. (2007b) Shrub encroachment affects mammalian carnivore abundance and species richness in semiarid rangelands. Acta Oecologica-International Journal of Ecology, 31, 86- 92.

Blaum, N., Seymour, C., Rossmanith, E., Schwager, M., and Jeltsch, F. (2009) Changes in arthropod diversity along a land use driven gradient of shrub cover in savanna rangelands: identification of suitable indicators. Biodiversity and Conservation, 18, 1187-1199.

Bolger, D.T., Suarez, A.V., Crooks, K.R., Morrison, S.A., and Case, T.J. (2000) Arthropods in urban habitat fragments in southern California: Area, age, and edge effects. Ecological Applications, 10, 1230-1248.

Bowman, D.M.J.S., Balch, J.K., and Artaxo, P. (2009) Fire in the Earth system. Science, 324, 481-484.

Bradstock, R.A. (1990) Relationships between fire regimes, plant species and fuels in mallee communities. The mallee lands: a conservation perspective (ed. by J.C. Noble, P.J. Joss & G.K. Jones). CSIRO, Melbourne.

Bradstock, R.A. and Gill, A.M. (1993) Fire in semiarid, mallee shrublands - size of flames from discrete fuel arrays and their role in the spread of fire. International Journal of Wildland Fire, 3, 3-12.

Bradstock, R.A. and Cohn, J.S. (2002) Fire regimes and biodiversity in semi- arid mallee ecosystems. Flammable Australia: the fire regimes and biodiversity of a continent (ed. by R.A. Bradstock, J.E. Williams & A.M. Gill), pp. 238-258. Cambridge University Press, Cambridge.

Broadhead, E. and Thornton, I.W.B. (1955) An ecological study of three closely related psocid species. Oikos, 6, 1-50.

191

Browning, D.M., Archer, S.R., Asner, G.P., McClaran, M.P., and Wessman, C.A. (2008) Woody plants in grasslands: Post-encroachment stand dynamics. Ecological Applications, 18, 928-944.

Bryce, S.A. (2006) Development of a bird integrity index: Measuring avian response to disturbance in the Blue Mountains of Oregon. Environmental Management, 38, 470-486.

Bukovinszky, T., Potting, R.P.J., Clough, Y., van Lenteren, J.C., and Vet, L.E.M. (2005) The role of pre- and post-alighting detection mechanisms in the responses to patch size by specialist herbivores. Oikos, 109, 435-446.

Bultman, T.L. and Uetz, G.W. (1982) Abundance and community structure of forest floor spiders following litter manipulation. Oecologia, 55, 34-41.

Bultman, T.L. and Uetz, G.W. (1984) Effect of structure and nutritional quality of litter on abundances of litter-dwelling arthropods. American Midland Naturalist, 111, 165-172.

Bureau of Meteorology (2012) Climatic averages Australia.

Burnham, K.P. and Anderson, D.R. (2002) Model selection and multimodel inference: a practical information-theoretic approach. 2nd edition edn. Springer- Verlag, New York.

Butler, B.E. (1958) Depositional systems of the Riverine Plain in relation to soils. Soil Publication No. 10. CSIRO, Melbourne.

Butterfield, B.J. and Briggs, J.M. (2009) Patch dynamics of soil biotic feedbacks in the Sonoran Desert. Journal of Arid Environments, 73, 96-102.

Callaway, R.M. and Walker, L.R. (1997) Competition and facilitation: A synthetic approach to interactions in plant communities. Ecology, 78, 1958-1965.

Callister, K. and Westbrooke, M. (2006) Pre-European distribution, structure and composition of semi-arid woodlands of the Victorian Mallee. Proceedings of the Royal Society of Victoria, 118, 245-258.

Capman, W.C., Batzli, G.O., and Simms, L.E. (1990) Responses of the common sooty wing skipper to patches of host plants. Ecology, 71, 1430-1440.

Carpintero, S., Reyes-Lopez, J., and Luque, G.M. (2011) Ant community structure under Retama sphaerocarpa shrubs in a semi-arid environment. Entomological Science, 14, 147-153.

Carver, M., Gross, G.F., and Woodward, T.E. (1991) Hemiptera. The insects of Australia: A textbook for students and researchers. 2nd edition. (ed. by CSIRO). Melbourne University Press, Carlton, Australia.

192

Castro, A. and Wise, D.H. (2009) Influence of fine woody debris on spider diversity and community structure in forest leaf litter. Biodiversity and Conservation, 18, 3705-3731.

Catling, P.C. and Burt, R.J. (1997) Studies of the ground-dwelling mammals of eucalypt forests in north-eastern New South Wales: The species, their abundance and distribution. Wildlife Research, 24, 1-19.

Chao, A. and Shen, T.J. (2003) Program SPADE (Species Prediction And Diversity Estimation). Program and user's guide published at http://chao.stat.nthu.edu.tw.

Chevan, A. and Sutherland, M. (1991) Hierarchical partitioning. The American Statistician, 45, 90-96.

Chiarello, A.G. (1999) Effects of fragmentation of the Atlantic forest on mammal communities in south-eastern Brazil. Biological Conservation, 89, 71-82.

Chikoski, J.M., Ferguson, S.H., and Meyer, L. (2006) Effects of water addition on soil arthropods and soil characteristics in a precipitation-limited environment. Acta Oecologica-International Journal of Ecology, 30, 203-211.

Churchill, T.B. (1998) Spiders as ecological indicators in the Australian tropics: family distribution patterns along rainfall and grazing gradients. In Proceedings of the 17th European Colloquium of Arachnology (ed. by S. P.A.), pp. 325-330. British Arachnological Society, Edinburgh.

Clark, J.A. and May, R.M. (2002) Taxonomic bias in conservation research. Science, 297, 191-192.

Clarke, K.R. and Gorley, R.N. (2006) Primer v6: User manual/tutorial. PRIMER- E, Plymouth, UK.

Colloff, M.J., Pullen, K.R., and Cunningham, S.A. (2010) Restoration of an ecosystem function to revegetation communities: the role of invertebrate macropores in enhancing soil water infiltration. Restoration Ecology, 18, 65-72.

Cook, W.M. and Holt, R.D. (2006) Influence of multiple factors on insect colonization of heterogeneous landscapes: A review and case study with periodical cicadas (Homoptera : Cicadidae). Annals of the Entomological Society of America, 99, 809-820.

Crist, T.O. (2009) Biodiversity, species interactions, and functional roles in ants (Hymenoptera: Formicidae) in fragmentated landscapes: a review. Myrmecological News, 12, 3-13.

Dangerfield, J.M. (1997) Abundance and diversity of soil macrofauna in northern Botswana. Journal of Tropical Ecology, 13, 527-538.

193

Daryanto, S. and Eldridge, D.J. (2010) Plant and soil surface responses to a combination of shrub removal and grazing in a shrub-encroached woodland. Journal of Environmental Management, 91, 2639-2648.

Dauber, J., Purtauf, T., Allspach, A., Frisch, J., Voigtlander, K., and Wolters, V. (2005) Local vs. landscape controls on diversity: a test using surface-dwelling soil macroinvertebrates of differing mobility. Global Ecology and Biogeography, 14, 213-221.

Dawes-Gromadzki, T.Z. (2005) Bugs beneath the surface: the functional significance of soil macroinvertebrates to landscape health in Australia’s tropical savannas. Insect Science, 12, 307-312.

De'ath, G. and Fabricius, C. (2000) Classification and regression trees: a powerful yet simple technique for ecological data analysis. Ecology, 81, 3178- 3192.

Dean, W.R.J., Milton, S.J., and Jeltsch, F. (1999) Large trees, fertile islands, and birds in arid savanna. Journal of Arid Environments, 41, 61-78.

Debuse, V.J., King, J., and House, A.P.N. (2007) Effect of fragmentation, habitat loss and within-patch habitat characteristics on ant assemblages in semi-arid woodlands of eastern Australia. Landscape Ecology, 22, 731-745.

Diffendorfer, J.E., Fleming, G.M., Duggan, J.M., Chapman, R.E., Rahn, M.E., Mitrovich, M.J., and Fisher, R.N. (2007) Developing terrestrial, multi-taxon indices of biological integrity: An example from coastal sage scrub. Biological Conservation, 140, 130-141.

Doblas-Miranda, E., Sanchez-Pinero, F., and Gonzalez-Megias, A. (2009) Different microhabitats affect soil macroinvertebrate assemblages in a Mediterranean arid ecosystem. Applied Soil Ecology, 41, 329-335.

Donoso, D.A., Johnston, M.K., and Kaspari, M. (2010) Trees as templates for tropical litter arthropod diversity. Oecologia, 164, 201-211.

Driscoll, D.A. (2004) Extinction and outbreaks accompany fragmentation of a reptile community. Ecological Applications, 14, 220-240.

Driscoll, D.A. and Weir, T. (2005) Beetle responses to habitat fragmentation depend on ecological traits, habitat condition, and remnant size. Conservation Biology, 19, 182-194.

Driscoll, D.A. (2010) Few beetle species can be detected with 95% confidence using pitfall traps. Austral Ecology, 35, 13-23.

Driscoll, D.A., Lindenmayer, D.B., Bennett, A.F., Bode, M., Bradstock, R.A., Cary, G.J., Clarke, M.F., Dexter, N., Fensham, R., Friend, G., Gill, M., James, S., Kay, G., Keith, D.A., MacGregor, C., Russell-Smith, J., Salt, D., Watson, J.E.M., Williams, R.J., and York, A. (2010) Fire management for biodiversity

194

conservation: Key research questions and our capacity to answer them. Biological Conservation, 143, 1928-1939.

Dufrene, M. and Legendre, P. (1997) Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs, 67, 345-366.

Eccard, J.A., Dean, W.R.J., Wichmann, M.C., Huttunen, S.M., Eskelinen, E., Moloney, K.A., and Jeltsch, F. (2006) Use of large Acacia trees by the cavity dwelling Black-tailed Tree Rat in the southern Kalahari. Journal of Arid Environments, 64, 604-615.

Ehrlich, P.R. and Pringle, R.M. (2008) Where does biodiversity go from here? A grim business-as-usual forecast and a hopeful portfolio of partial solutions. Proceedings of the National Academy of Sciences of the United States of America, 105, 11579-11586.

Eldridge, D.J. and Freudenberger, D. (2005) Ecosystem wicks: Woodland trees enhance water infiltration in a fragmented agricultural landscape in eastern Australia. Austral Ecology, 30, 336-347.

Eldridge, D.J., Freudenberger, D., and Koen, T.B. (2006) Diversity and abundance of biological soil crust taxa in relation to fine and coarse-scale disturbances in a grassy eucalypt woodland in eastern Australia. Plant and Soil, 281, 255-268.

Eldridge, D.J., Bowker, M.A., Maestre, F.T., Roger, E., Reynolds, J.F., and Whitford, W.G. (2011) Impacts of shrub encroachment on ecosystem structure and functioning: towards a global synthesis. Ecology Letters, 14, 709-722.

Ellis, T.W., Leguedois, S., Hairsine, P.B., and Tongway, D.J. (2006) Capture of overland flow by a tree belt on a pastured hillslope in south-eastern Australia. Australian Journal of Soil Research, 44, 117-125.

Facelli, J.M. and Pickett, S.T.A. (1991) Plant litter - its dynamics and effects on plant community structure. Botanical Review, 57, 1-32.

Facelli, J.M. and Brock, D.J. (2000) Patch dynamics in arid lands: localized effects of Acacia papyrocarpa on soils and vegetation of open woodlands of south Australia. Ecography, 23, 479-491.

Fahrig, L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics, 34, 487-515.

Failing, L. and Gregory, R. (2003) Ten common mistakes in designing biodiversity indicators for forest policy. Journal of Environmental Management, 68, 121-132.

195

Fattorini, S. (2010) Effects of fire on tenebrionid communities of a Pinus pinea plantation: a case study in a Mediterranean site. Biodiversity and Conservation, 19, 1237-1250.

Ferrier, S. and Guisan, A. (2006) Spatial modelling of biodiversity at the community level. Journal of Applied Ecology, 43, 393-404.

Fischer, J., Fazey, I., Briese, R., and Lindenmayer, D.B. (2005) Making the matrix matter: challenges in Australian grazing landscapes. Biodiversity and Conservation, 14, 561-578.

Fischer, J. and Lindenmayer, D.B. (2007) Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biogeography, 16, 265-280.

Fisher, A. and Kutt, A. (2007) Biodiversity and land condition in tropical savanna rangelands: technical report, Tropical Savannas CRC, Darwin.

Flynn, D.F.B., Sudderth, E.A., and Bazzaz, F.A. (2006) Effects of aphid herbivory on biomass and leaf-level physiology of Solanum dulcamara under elevated temperature and CO2. Environmental and Experimental Botany, 56, 10-18.

Folgarait, P.J. (1998) Ant biodiversity and its relationship to ecosystem functioning: a review. Biodiversity and Conservation, 7, 1221-1244.

Forman, R.T. (1995) Land mosaics: the ecology of landscapes and regions. Cambridge University Press, New York.

Friend, G.R. (1993) Impact of fire on small vertebrates in mallee woodlands and heathlands of temperate Australia - a review. Biological Conservation, 65, 99- 114.

Frith, H.J. (1962) The mallee fowl, the bird that builds an incubator. Angus and Robertson, Sydney.

Gardner, S.M., Cabido, M.R., Valladares, G.R., and Diaz, S. (1995) The influence of habitat structure on arthropod diversity in Argentine semiarid Chaco forest. Journal of Vegetation Science, 6, 349-356.

Garner, W. and Steinberger, Y. (1989) A proposed mechanism for the formation of ‘Fertile Islands’ in the desert ecosystem. Journal of Arid Environments, 16, 257-262.

Gibb, H. and Hochuli, D.F. (2002) Habitat fragmentation in an urban environment: large and small fragments support different arthropod assemblages. Biological Conservation, 106, 91-100.

Gibb, H. (2005) The effect of a dominant ant, Iridomyrmex purpureus, on resource use by ant assemblages depends on microhabitat and resource type. Austral Ecology, 30, 856-867.

196

Gibb, H. and Parr, C.L. (2010) How does habitat complexity affect ant foraging success? A test using functional measures on three continents. Oecologia, 164, 1061-1073.

Gibbons, P. and Boak, M. (2002) The value of paddock trees for regional conservation in an agricultural landscape. Ecological Management & Restoration, 3, 205-210.

Gibbons, P. and Freudenberger, D. (2006) An overview of methods used to assess vegetation condition at the scale of the site. Ecological Management and Restoration, 7, S10-17.

Gibbons, P., Briggs, S.V., Ayers, D.A., Doyle, S., Seddon, J., McElhinny, C., Jones, N., Sims, R., and Doody, J.S. (2008) Rapidly quantifying reference conditions in modified landscapes. Biological Conservation, 141, 2483-2493.

Gilpin, M.E. and Diamond, J.M. (1976) Calculations of immigration and extinction curves from the species-area distance relation. Proceedings of the National Academy of Sciences of the United States of America, 73, 4130-4134.

Gollan, J.R., Ashcroft, M.B., Cassis, G., Donnelly, A.P., and Lassau, S.A. (2009) Testing common habitat-based surrogates of invertebrate diversity in a semi-arid rangeland. Biodiversity and Conservation, 18, 1147-1159.

Gonzalez, A., Rayfield, B., and Lindo, Z. (2011) The disentangled bank: How loss of habitat fragments and disassembles ecological networks. American Journal of Botany, 98, 503-516.

Goslee, S.C. and Sanderson, M.A. (2010) Landscape context and plant community composition in grazed agricultural systems of the Northeastern United States. Landscape Ecology, 25, 1029-1039.

Gove, A.D., Fitzpatrick, M.C., Majer, J.D., and Dunn, R.R. (2009) Dispersal traits linked to range size through range location, not dispersal ability, in Western Australian angiosperms. Global Ecology and Biogeography, 18, 596- 606.

Greenslade, P. (1991) Collembola. The insects of Australia. A textbook for students and research workers. 2nd edition. (ed. by CSIRO). Melbourne University Press, Carlton, Australia.

Greenville, A.C., Dickman, C.R., Wardle, G.M., and Letnic, M. (2009) The fire history of an arid grassland: the influence of antecedent rainfall and ENSO. Journal of Wildland Fire, 18, 631-639.

Grevstad, F.S. and Klepetka, B.W. (1992) The influence of plant architecture on the foraging efficiencies of a suite of ladybird beetles feeding on aphids. Oecologia, 92, 399-404.

197

Grez, A.A. and Gonzalez, R.H. (1995) Resource concentration hypothesis - effect of host-plant patch size on density of herbivorous insects. Oecologia, 103, 471-474.

Griffin, G.F. (1991) Characteristics of three spinifex alliances in central Australia. Journal of Vegetation Science, 1, 435-444.

Guarneri, A.A., Lazzari, C., Diotaiuti, L., and Lorenzo, M.G. (2002) The effect of relative humidity on the behaviour and development of Triatoma brasiliensis. Physiological Entomology, 27, 142-147.

Hamback, P.A., Bjorkman, M., and Hopkins, R.J. (2010) Patch size effects are more important than genetic diversity for plant-herbivore interactions in Brassica crops. Ecological Entomology, 35, 299-306.

Hamilton, A.J., Basset, Y., Benke, K.K., Grimbacher, P.S., Miller, S.E., Novotny, V., Samuelson, G.A., Stork, N.E., Weiblen, G.D., and Yen, J.D.L. (2010) Quantifying Uncertainty in Estimation of Tropical Arthropod Species Richness. American Naturalist, 176, 90-95.

Hansen, R.A. and Coleman, D.C. (1998) Litter complexity and composition are determinants of the diversity and species composition of oribatid mites (Acari : Oribatida) in litterbags. Applied Soil Ecology, 9, 17-23.

Harper, K.A., Macdonald, S.E., Burton, P.J., Chen, J.Q., Brosofske, K.D., Saunders, S.C., Euskirchen, E.S., Roberts, D., Jaiteh, M.S., and Esseen, P.A. (2005) Edge influence on forest structure and composition in fragmented landscapes. Conservation Biology, 19, 768-782.

Harrington, G. (1979) Estimation of above-ground biomass of trees and shrubs in a Eucalyptus populnea F. Muell woodland by regression of mass on trunk diameter and plant height. Australian Journal of Botany, 27, 135-143.

Haslem, A., Kelly, L.T., Nimmo, D.G., Watson, S.J., Kenny, S.A., Taylor, R.S., Avitabile, S.C., Callister, K.E., Spence-Bailey, L.M., Clarke, M.F., and Bennett, A.F. (2011) Habitat or fuel? Implications of long-term, post-fire dynamics for the development of key resources for fauna and fire. Journal of Applied Ecology, 48, 247-256.

He, Y.F., D'Odorico, P., De Wekker, S.F.J., Fuentes, J.D., and Litvak, M. (2010) On the impact of shrub encroachment on microclimate conditions in the northern Chihuahuan desert. Journal of Geophysical Research-Atmospheres, 115.

Heller, N.E. and Zavaleta, E.S. (2009) Biodiversity management in the face of climate change: A review of 22 years of recommendations. Biological Conservation, 142, 14-32.

Herrick, J.E., Van Zee, J.W., Havstad, K.M., Burkett, L.M., and Whitford, W.G. (2005) Monitoring manual for grassland, shrubland and savanna ecosystems.

198

USDA-ARS Jornada Experimental Range, Las Cruces, New Mexico. Distributed by University of Arizona Press

Hobbs, R.J. (1985) Harvester ant foraging and plant-species distribution in annual grassland. Oecologia, 67, 519-523.

Hodkinson, I.D., Bird, J.M., Hill, J.K., and Baxter, R. (2001) Host plant growth characteristics as determinants of abundance and phenology in jumping plant- lice on downy willow. Ecological Entomology, 26, 376-387.

Hoffmann, B.D. (2000) Changes in ant species composition and community organisation along grazing gradients in semi-arid rangelands of the Northern Territory. Rangeland Journal, 22, 171-189.

Hoffmann, B.D. and Andersen, A.N. (2003) Responses of ants to disturbance in Australia, with particular reference to functional groups. Austral Ecology, 28, 444-464.

Hoffmann, B.D. (2010) Using ants for rangeland monitoring: Global patterns in the responses of ant communities to grazing. Ecological Indicators, 10, 105- 111.

Hoffmann, B.D. and James, C.D. (2011) Using ants to manage sustainable grazing: Dynamics of ant faunas along sheep grazing gradients conform to four global patterns. Austral Ecology, 36, 698-708.

Hölldobler, B. and Wilson, E.O. (1990) The Ants. Harvard University Press, Cambridge, USA.

Holling, C.S. (1986) Resilience of ecosystems: Local surprise and global change. Development of the Biosphere (ed. by W.C. Clark & R.E. Munn), pp. 292-317. Cambridge University Press, Cambridge.

Holsten, E.H., Werner, R.A., and Develice, R.L. (1995) Effects of a spruce beetle (Coleoptera: Scolytidae) outbreak and fire on Lutz spruce in Alaska. Environmental Entomology, 24, 1539-1547.

House, A., Burwell, C., and Brown, S. (2006) Ant assemblages in Brigalow regrowth vegetation. Ecological Management and Restoration, 7, 136-140.

Hughes, J.B., Daily, G.C., and Ehrlich, P.R. (2000) Conservation of insect diversity: a habitat approach. Conservation Biology, 14, 1788-1797.

Hunter, M.D., Adl, S., Pringle, C.M., and Coleman, D.C. (2003) Relative effects of macroinvertebrates and habitat on the chemistry of litter during decomposition. Pedobiologia, 47, 101-115.

Hurlburt, S.H. (1984) Pseudoreplication and the design of ecological field experiments. Ecological Monographs, 54, 187-211.

199

Isbell, R.F. (1996) The Australian Soil Classification. CSIRO Publishing, Melbourne.

James, A.I., Eldridge, D.J., Koen, T.B., and Whitford, W.G. (2008) Landscape position moderates how ant nests affect hydrology and soil chemistry across a Chihuahuan Desert watershed. Landscape Ecology, 23, 961-975.

Jander, G. and Howe, G. (2008) Plant interactions with arthropod herbivores: State of the field. Plant Physiology, 146, 801-803.

Jeffries, M.J. and Lawton, J.H. (1984) Enemy free space and the structure of ecological communities. Biological Journal of the Linnean Society, 23, 269-286.

Jennings, M.D., Hoekstra, J., Higgins, J., and Boucher, T. (2008) A comparative measure of biodiversity based on species composition. Biodiversity and Conservation, 17, 833-840.

Johnson, R.A. (1992) Soil texture as an influence on the distribution of the desert seed-harvester ants Pogonomyrmex rugosus and Messor pergandei. Oecologia, 89, 118-124.

Johnson, R.A. (2000) Habitat segregation based on soil texture and body size in the seed-harvester ants Pogonomyrmex rugosus and P. barbatus. Ecological Entomology, 25, 403-412.

Johnson, R.A. (2001) Biogeography and community structure of North American seed-harvester ants. Annual Review of Entomology, 46, 1-29.

Jones, C.G., Lawton, J.H., and Shachak, M. (1994) Organisms as ecosystem engineers. Oikos, 69, 373-386.

Jones, C.G., Gutierrez, J.L., Groffman, P.M., and Shachak, M. (2006) Linking ecosystem engineers to soil processes: a framework using the Jenny State Factor Equation. European Journal of Soil Biology, 42, S39-S53.

Joshi, J. and Vrieling, K. (2005) The enemy release and EICA hypothesis revisited: incorporating the fundamental difference between specialist and generalist herbivores. Ecology Letters, 8, 704-714.

Jost, L. (2006) Entropy and diversity. Oikos, 113, 363-375.

Kampichler, C. and Bruckner, A. (2009) The role of microarthropods in terrestrial decomposition: a meta-analysis of 40 years of litterbag studies. Biological Reviews, 84, 375-389.

Karr, J.R. (1991) Biological integrity - a long-neglected aspect of water-resource management. Ecological Applications, 1, 66-84.

Karr, J.R. and Kimberling, D.N. (2003) A terrestrial arthropod index of biological integrity for shrub-steppe landscapes. Northwest Science, 77, 202-213.

200

Kaspari, M. and Weiser, M.D. (1999) The size-grain hypothesis and interspecific scaling in ants. Functional Ecology, 13, 530-538.

Kellert, S.R. (1993) Values and perceptions of invertebrates. Conservation Biology, 7, 845-855.

Kelly, L.T., Nimmo, D.G., Spence-Bailey, L.M., Haslem, A., Watson, S.J., Clarke, M.F., and Bennett, A.F. (2011) Influence of fire history on small mammal distributions: insights from a 100-year post-fire chronosequence. Diversity and Distributions, 17, 462-473.

Kent, K., Earl, G., Mullins, B., Lunt, I., and Webster, R. (2002) Native vegetation guide for the Riverina: Notes for land managers in its management and revegetation. Johnstone Centre, Charles Sturt University, Albury.

Kienast, F., Bolliger, J., Potschin, M., de Groot, R.S., Verburg, P.H., Heller, I., Wascher, D., and Haines-Young, R. (2009) Assessing landscape functions with broad-scale environmental data: Insights gained from a prototype development for Europe. Environmental Management, 44, 1099-1120.

Kirkham, D.R. and Fisser, H.G. (1972) Rangeland relations and harvester ants in Northcentral Wyoming. Journal of Range Management, 25, 55-60.

Klapwijk, M.J. and Lewis, O.T. (2011) Spatial ecology of multiple parasitoids of a patchily-distributed host: implications for species coexistence. Ecological Entomology, 36, 212-220.

Klemm, D.J., Blocksom, K.A., Fulk, F.A., Herlihy, A.T., Hughes, R.M., Kaufmann, P.R., Peck, D.V., Stoddard, J.L., Thoeny, W.T., Griffith, M.B., and Davis, W.S. (2003) Development and evaluation of a Macroinvertebrate Biotic Integrity Index (MBII) for regionally assessing Mid-Atlantic Highlands streams. Environmental Management, 31, 656-669.

Koivula, M., Punttila, P., Haila, Y., and Niemela, J. (1999) Leaf litter and the small-scale distribution of carabid beetles (Coleoptera, Carabidae) in the boreal forest. Ecography, 22, 424-435.

Kotliar, N.B. and Wiens, J.A. (1990) Multiple scales of patchiness and patch structure - a hierarchical framework for the study of heterogeneity. Oikos, 59, 253-260.

Kremen, C. (1992) Assessing the indicator properties of species assemblages for natural areas monitoring. Ecological Applications, 2, 203-217.

Kwok, A.B.C., Eldridge, D.J., and Oliver, I. (2011) Do landscape health indices reflect arthropod biodiversity status in the eucalypt woodlands of eastern Australia? Austral Ecology, 36, 800-813.

201

Lalley, J.S., Viles, H.A., Henschel, J.R., and Lalley, V. (2006) Lichen-dominated soil crusts as arthropod habitat in warm deserts. Journal of Arid Environments, 67, 579-593.

Langlands, P.R., Brennan, K.E.C., and Pearson, D.J. (2006) Spiders, spinifex, rainfall and fire: Long-term changes in an arid spider assemblage. Journal of Arid Environments, 67, 36-59.

Langlands, P.R., Brennan, K.E.C., and Ward, B. (in press) Is the reassembly of an arid spider assemblage following fire deterministic? Austral Ecology.

Langor, D.W. and Spence, J.R. (2006) Arthropods as ecological indicators of sustainability in Canadian forests. Forestry Chronicle, 82, 344-350.

Larsson, F.K. (1989) Insect mating patterns explained by microclimatic variables. Journal of Thermal Biology, 14, 155-157.

Lassau, S.A. and Hochuli, D.F. (2004) Effects of habitat complexity on ant assemblages. Ecography, 27, 157-164.

Lassau, S.A., Hochuli, D.F., Cassis, G., and Reid, C.A.M. (2005) Effects of habitat complexity on forest beetle diversity: do functional groups respond consistently? Diversity and Distributions, 11, 73-82.

Lavelle, P., Decaens, T., Aubert, M., Barot, S., Blouin, M., Bureau, F., Margerie, P., Mora, P., and Rossi, J.P. (2006) Soil invertebrates and ecosystem services. European Journal of Soil Biology, 42, S3-S15.

Lawrence, K.L. and Wise, D.H. (2000) Spider predation on forest-floor Collembola and evidence for indirect effects on decomposition. Pedobiologia, 44, 33-39.

Lawton, J.H. (1983) Plant architecture and the diversity of phytophagous insects. Annual Review of Entomology, 28, 23-39.

Lee, K.E. and Wood, T.G. (1971) Physical and chemical effects on soils of some australian termites, and their pedological significance. Pedobiologia, 11, 376-&.

Lemperiere, G. and Marage, D. (2010) The influence of forest management and habitat on insect communities associated with dead wood: a case study in forests of the southern French Alps. Insect Conservation and Diversity, 3, 236- 245.

Letnic, M., Tamayo, B., and Dickman, C.R. (2005) The responses of mammals to La Nina (El Nino Southern Oscillation)-associated rainfall, predation, and wildfire in central Australia. Journal of Mammalogy, 86, 689-703.

Lightfoot, D.C. and Whitford, W.G. (1987) Variation in insect densities on desert creosotebush - is nitrogen a factor. Ecology, 68, 547-557.

202

Liira, J. and Kohv, K. (2010) Stand characteristics and biodiversity indicators along the productivity gradient in boreal forests: Defining a critical set of indicators for the monitoring of habitat nature quality. Plant Biosystems, 144, 211-220.

Lindenmayer, B.D. and Fischer, J. (2006) Habitat fragmentation and landscape change. CSIRO Publishing, Collingwood, VIC.

Lindenmayer, D., Hobbs, R.J., Montague-Drake, R., Alexandra, J., Bennett, A., Burgman, M., Cale, P., Calhoun, A., Cramer, V., Cullen, P., Driscoll, D., Fahrig, L., Fischer, J., Franklin, J., Haila, Y., Hunter, M., Gibbons, P., Lake, S., Luck, G., MacGregor, C., McIntyre, S., Mac Nally, R., Manning, A., Miller, J., Mooney, H., Noss, R., Possingham, H., Saunders, D., Schmiegelow, F., Scott, M., Simberloff, D., Sisk, T., Tabor, G., Walker, B., Wiens, J., Woinarski, J., and Zavaleta, E. (2008) A checklist for ecological management of landscapes for conservation. Ecology Letters, 11, 78-91.

Lindenmayer, D.B., Margules, C.R., and Botkin, D.B. (2000) Indicators of biodiversity for ecologically sustainable forset management. Conservation Biology, 14, 941-950.

Lindenmayer, D.B., Manning, A.D., Smith, P.L., Possingham, H.P., Fischer, J., Oliver, I., and McCarthy, M.A. (2002) The focal-species approach and landscape restoration: a critique. Conservation Biology, 16, 338-345.

Lindsay, E.A. and Cunningham, S.A. (2009) Livestock grazing exclusion and microhabitat variation affect invertebrates and litter decomposition rates in woodland remnants. Forest Ecology and Management, 258, 178-187.

Liu, R.T., Zhao, H.L., Zhao, X.Y., and Drake, S. (2011) Facilitative effects of shrubs in shifting sand on soil macro-faunal community in Horgin Sand Land of Inner Mongolia, Northern China. European Journal of Soil Biology, 47, 316-321.

Losos, J.B. and Ricklefs, R.E. (2010) The theory of island biogeography revisited. Princeton University Press, New Jersey.

Ludwig, J.A. and Tongway, D.J. (1993) Monitoring the condition of Australian arid lands. In: D.H. McKenzie linked plant-soil indicators. Ecological Indicators (ed. by D.H. McKenzie, D.E. Hyatt & V.J. McDonald), pp. 765-772. Elsevier Applied Science, New York.

Ludwig, J.A., Tongway, D.J., Freudenberger, D.O., Noble, J.C., and Hodgkinson, K.C. (1997) Landscape ecology - function and management. Principles from Australia's rangelands. CSIRO Publishing, Melbourne, Australia.

Ludwig, J.A., Eager, R.W., Williams, R.J., and Lowe, L.M. (1999) Declines in vegetation patches, plant diversity, and grasshopper diversity near cattle watering-points in Victoria River District, northern Australia. Rangeland Journal, 21, 135-149.

203

Ludwig, J.A. and Tongway, D.J. (2000) Viewing rangelands as landscape systems. Rangeland Desertification: Advances in Vegetation Sciences (ed. by O. Arnalds & S. Archer), pp. 39-52. Kluwer Academic Publishers, Dordrecht, The Netherlands.

Ludwig, J.A., Tongway, D.J., Bastin, G.N., and James, C.D. (2004) Monitoring ecological indicators of rangeland functional integrity and their relation to biodiversity at local to regional scales. Austral Ecology, 29, 108-120.

Ludwig, J.A., Wilcox, B.P., Breshears, D.D., Tongway, D.J., and Imeson, A.C. (2005) Vegetation patches and runoff-erosion as interacting ecohydrological processes in semiarid landscapes. Ecology, 86, 288-297.

Lumsden, L.F. and Bennett, A.F. (2005) Scattered trees in rural landscapes: foraging habitat for insectivorous bats in south-eastern Australia. Biological Conservation, 122, 205-222.

Luque, G.M. and Lopez, J.R. (2007) Effect of experimental small-scale spatial heterogeneity on resource use of a Mediterranean ground-ant community. Acta Oecologica-International Journal of Ecology, 32, 42-49.

Mac Nally, R. (2002) Multiple regression and inference in ecology and conservation biology: further comments on identifying important predictor variables. Biodiversity and Conservation, 11, 1397-1401.

MacArthur, R.H. and Wilson, E.O. (1967) The theory of island biogeography. Princeton University Press, Princeton, N.J.

Maestre, F.T. and Cortina, J. (2004) Insights into ecosystem composition and function in a sequence of degraded semiarid steppes. Restoration Ecology, 12, 494-502.

Maestre, F.T. and Puche, M.D. (2009) Indices based on surface indicators predict soil functioning in Mediterranean semi-arid steppes. Applied Soil Ecology, 41, 342-350.

Mageau, M.T., Costanza, R., and Ulanowicz, R.E. (1995) The development and initial testing of a quantitative assessment of ecosystem health. Ecosystem Health, 1, 201-213.

Majer, J.D., Recher, H.F., Graham, R., and Gupta, R. (2006) Trunk invertebrate faunas of Western Australian forests and woodlands: Seeking causes of patterns along a west-east gradient. Austral Ecology, 31, 503-511.

Majer, J.D., Brennan, K.E.C., and Moir, M.L. (2007) Invertebrates and the restoration of a forest ecosystem: 30 years of research following bauxite mining in Western Australia. Restoration Ecology, 15.

204

Major, R.E., Christie, F.J., Gowing, G., Cassis, G., and Reid, C.A.M. (2003) The effect of habitat configuration on arboreal insects in fragmented woodlands of south-eastern Australia. Biological Conservation, 113, 35-48.

Manning, A.D., Gibbons, P., and Lindenmayer, D.B. (2009) Scattered trees: a complementary strategy for facilitating adaptive responses to climate change in modified landscapes? Journal of Applied Ecology, 46, 915-919.

Marques, E.S.D., Price, P.W., and Cobb, N.S. (2000) Resource abundance and insect herbivore diversity on woody fabaceous desert plants. Environmental Entomology, 29, 696-703.

Martinez-Crego, B., Alcoverro, T., and Romero, J. (2010) Biotic indices for assessing the status of coastal waters: a review of strengths and weaknesses. Journal of Environmental Monitoring, 12, 1013-1028.

Masumoto, T., Sunahara, T., and Suzuki, N. (2000) Effects of nonhost and host plants on insect herbivory covarying with plant size in the cruciferous plant Turritis glabra. Population Ecology, 42, 145-152.

May, R.M. (2000) The dimensions of life on Earth. Nature and human society: the quest for a sustainable world (ed. by P.H. Raven & T. Williams). National Academy, Washington, DC.

Mayse, M. (1978) Effects of spacing between rows on soya bean arthropod populations. Journal of Animal Ecology, 1.

Mazerolle, M.J. and Villard, M.A. (1999) Patch characteristics and landscape context as predictors of species presence and abundance: A review. Ecoscience, 6, 117-124.

Mazia, C.N., Chaneton, E.J., and Kitzberger, T. (2006) Small-scale habitat use and assemblage structure of ground-dwelling beetles in a Patagonian shrub steppe. Journal of Arid Environments, 67, 177-194.

McCarthy, M.A., Parris, K.M., Van Der Ree, R., McDonnell, M.J., Burgman, M.A., Williams, N.S.G., McLean, N., Harper, M.J., Meyer, R., Hahs, A., and Coates, T. (2004) The habitat hectares approach to vegetation assessment: An evaluation and suggestions for improvement. Ecological Management and Restoration, 5, 24-27.

McCune, B. and Mefford, M.J. (1999) PC-ORD. Multivariate Analysis of Ecological data, Version 4. MjM Software Design, Gleneden Beach, Oregeon, USA.

McDonald, R.C., Isbell, R.F., Speight, J.G., Walker, J., and Hopkins, M.S. (1990) Australian soil and land survey. Field handbook. Second edition. Inkata Press, Melbourne.

205

McIntyre, N.E. (1997) Scale-dependent habitat selection by the darkling beetle Eleodes hispilabris (Coleoptera: Tenebrionidae). American Midland Naturalist, 138, 230-235.

McIntyre, S. and Hobbs, R. (1999) A framework for conceptualizing human effects on landscapes and its relevance to management and research models. Conservation Biology, 13.

McIntyre, S., Tongway, D., and Lambeck, R. (2003) Improved vegetation planning for rural landscapes. CTC27 Project report by CSIRO Sustainable Ecosystems (Canberra and Brisbane) and Greening Australia WA Inc. to Land & Water Australia, Canberra.

Megias, A.G., Sanchez-Pinero, F., and Hodar, J.A. (2011) Trophic interactions in an arid ecosystem: From decomposers to top-predators. Journal of Arid Environments, 75, 1333-1341.

Melbourne, B.A. (1999) Bias in the effect of habitat structure on pitfall traps: An experimental evaluation. Australian Journal of Ecology, 24, 228-239.

Meyer, G.A. and Whitlow, T.H. (1992) Effects of leaf and sap feeding insects on photosynthetic rates of goldenrod. Oecologia, 92, 480-489.

Mody, K., Unsicker, S.B., and Linsenmair, K.E. (2007) Fitness related diet- mixing by intraspecific host-plant-switching of specialist insect herbivores. Ecology, 88, 1012-1020.

Moir, M.L., Brennan, K.E.C., Majer, J.D., Koch, J.M., and Fletcher, M.J. (2010) Plant species redundancy and the restoration of faunal habitat: lessons from plant-dwelling bugs. Restoration Ecology, 18, 136-147.

Moore, R.M. (1970) South-eastern temperate woodlands and grasslands. Australian Grasslands (ed. by R. Moore), pp. 169-190. Australian National University Press, Canberra.

Moorhead, D.L. and Reynolds, J.F. (1991) A general-model of litter decomposition in the northern Chihuahuan desert. Ecological Modelling, 56, 197-219.

Moretti, M., Obrist, M.K., and Duelli, P. (2004) Arthropod biodiversity after forest fires: winners and losers in the winter fire regime of the southern Alps. Ecography, 27, 173-186.

Moro, M.J., Pugnaire, F.I., Haase, P., and Puigdefabregas, J. (1997) Effect of the canopy of Retama sphaerocarpa on its understorey in a semiarid environment. Functional Ecology, 11, 425-431.

Morton, S.R., Smith, D.M.S., Dickman, C.R., Dunkerley, D.L., Friedel, M.H., McAllister, R.R.J., Reid, J.R.W., Roshier, D.A., Smith, M.A., Walsh, F.J.,

206

Wardle, G.M., Watson, I.W., and Westoby, M. (2011) A fresh framework for the ecology of arid Australia. Journal of Arid Environments, 75, 313-329.

Muñoz-Erickson, T.A., Aguilar-González, B., and Sisk, T.D. (2007) Linking ecosystem health indicators and collaborative management: a systematic framework to evaluate ecological and social outcomes. Ecology and Society, 12.

Murakami, M. and Hirao, T. (2010) Nestedness of insect assemblages on small Bahamian islands: importance of spatial processes. Insect Conservation and Diversity, 3, 229-235.

Murphy, D.D. and Noon, B.R. (1992) Integrating scientific methods with habitat conservation planning - reserve design for Spotted Owls. Ecological Applications, 2, 3-17.

Nakamura, A., Catterall, C.P., Burwell, C.J., Kitching, R.L., and House, A.P.N. (2009) Effects of shading and mulch depth on the colonisation of habitat patches by arthropods of rainforest soil and litter. Insect Conservation and Diversity, 2, 221-231.

Narendra, A., Gibb, H., and Ali, T.M. (2011) Structure of ant assemblages in Western Ghats, India: role of habitat, disturbance and introduced species. Insect Conservation and Diversity, 4, 132-141.

Nash, S. (2004) Desperately seeking charisma: Improving the status of invertebrates. Bioscience, 54, 487-494.

New, T.R. (1995) An introduction to invertebrate conservation biology. Oxford University Press, New York.

Nielsen, E.S. and Common, I.F.B. (1991) Lepidoptera. The insects of Australia: A textbook for students and researchers. 2nd edition. (ed. by CSIRO). Melbourne University Press, Carlton, Australia.

Nitterus, K. and Gunnarsson, B. (2006) Effect of microhabitat complexity on the local distribution of arthropods in clear-cuts. Environmental Entomology, 35, 1324-1333.

Noble, J.C., Smith, A.W., and Leslie, H.W. (1980) Fire in the Mallee Shrublands of Western New South Wales. The Australian Rangeland Journal, 2, 104-114.

Noble, J.C. (1982) The significance of fire in the biology and evolutionary ecology of mallee Eucalyptus populations. Evolution of the flora and fauna of arid Australia (ed. by W.R. Barker & P.J.M. Greenslade), pp. 153-159. Peacock Publications, Adelaide.

Noble, J.C., Kaliszweski, M., and Whitford, W.G. (1990) Soil acari from burnt and unburnt mallee habitats in western New South Wales. The mallee lands: a

207

conservation perspective (ed. by J.C. Noble, P.J. Joss & G.K. Jones), pp. 224- 226. CSIRO, Melbourne.

Noble, J.C., Whitford, W.G., and Kaliszweski, M. (1996) Soil and litter microarthropod populations from two contrasting ecosystems in semi-arid eastern Australia. Journal of Arid Environments, 32, 329-346.

Noble, J.C. (1997) The delicate and noxious scrub : CSIRO studies on native trees and shrub proliferation in the semi-arid woodlands of Eastern Australia. CSIRO Publishing.

Northcote, K.H. (1966) Atlas of Australian Soils, explanatory data for Sheet 3, Sydney-Canberra-Bourke-Armidale Area. CSIRO, Melbourne.

Noss, R.F. (1990) Indicators for monitoring biodiversity - a hierarchical approach. Conservation Biology, 4, 355-364.

Noy Meir, I. (1973) Desert ecosystems: environment and producers. Annual Review of Ecology and Systematics, 5, 195-214.

Nulsen, R.A., Bligh, K.J., Baxter, I.N., Solin, E.J., and Imrie, D.H. (1986) The fate of rainfall in a mallee and heath vegetated catchment in southern western Australia. Australian Journal of Ecology, 11, 361-371.

O'Conner, T.G. and Kuyler, P. (2009) Impact of land use on the biodiversity integrity of the moist sub-biome of the grassland biome, South Africa. Journal of Environmental Management, 90, 384-395.

Obermaier, E., Heisswolf, A., Poethke, H.J., Randlkofer, B., and Meiners, T. (2008) Plant architecture and vegetation structure: Two ways for insect herbivores to escape parasitism. European Journal of Entomology, 105, 233- 240.

Odegaard, F. (2000) How many species of arthropods? Erwin's estimate revised. Biological Journal of the Linnean Society, 71, 583-597.

Odegaard, F. (2006) Host specificity, alpha- and beta-diversity of phytophagous beetles in two tropical forests in Panama. Biodiversity and Conservation, 15, 83- 105.

Olea, P.P., Mateo-Tomas, P., and de Frutos, A. (2010) Estimating and Modelling Bias of the Hierarchical Partitioning Public-Domain Software: Implications in Environmental Management and Conservation. Plos One, 5.

Oliver, I. (2002) An expert panel-based approach to the assessment of vegetation condition within the context of biodiversity conservation stage 1: The identification of condition indicators. Ecological Indicators, 2, 223-237.

208

Oliver, I., Pearce, S., Greenslade, P.J.M., and Britton, D.R. (2006) Contribution of paddock trees to the conservation of terrestrial invertebrate biodiversity within grazed native pastures. Austral Ecology, 31, 1-12.

Oliver, I., Jones, H., and Schmoldt, D.L. (2007) Expert panel assessment of attributes for natural variability benchmarks for biodiversity. Austral Ecology, 32, 453-475.

Orabi, G., Moir, M.L., and Majer, J.D. (2010) Assessing the success of mine restoration using Hemiptera as indicators. Australian Journal of Zoology, 58, 243-249.

Orrock, J.L., Curler, G.R., Danielson, B.J., and Coyle, D.R. (2011) Large-scale experimental landscapes reveal distinctive effects of patch shape and connectivity on arthropod communities. Landscape Ecology, 26, 1361-1372.

Paine, R.T. (1969) A note on trophic complexity and community stability. American Naturalist, 103, 91-&.

Paris, O.H. (1965) Vagility of P32-labeled isopods in grassland. Ecology, 46, 635-648.

Pearce, J.L., Venier, L.A., Eccles, G., Pedlar, J., and McKenney, D. (2004) Influence of habitat and microhabitat on epigeal spider (Araneae) assemblages in four stand types. Biodiversity and Conservation, 13, 1305-1334.

Pearce, J.L. and Venier, L.A. (2006) The use of ground beetles (Coleoptera: Carabidae) and spiders (Araneae) as bioindicators of sustainable forest management: A review. Ecological Indicators, 6, 780-793.

Pearson, D.L. and Lederhouse, R.C. (1987) Thermal ecology and the structure of an assemblage of adult tiger beetle species (Cicindelidae). Oikos, 50, 247- 255.

Peeters, P.J., Read, J., and Sanson, G.D. (2001) Variation in the guild composition of herbivorous insect assemblages among co-occuring plant species. Austral Ecology, 26, 385-399.

Peters, W.L. and Campbell, I.C. (1991) Thysanura. The insects of Australia: A textbook for students and researchers. 2nd edition. (ed. by CSIRO). Melbourne University Press, Carlton, Australia.

Pickett, S.T.A. and Cadenasso, M.L. (1995) Landscape ecology: spatial heterogeneity in ecological systems. Science, 269, 331-334.

Pimentel, S. (1961) The infleunce of plant spatial patterns on insect populations. Annuals of the Entomological Society of America, 54, 61-69.

Pimm, S.L. and Raven, P. (2000) Biodiversity - Extinction by numbers. Nature, 403, 843-845.

209

Pitzalis, M., Fattorini, S., Trucchi, E., and Bologna, M.A. (2005) Comparative analysis of species diversity of Isopoda Oniscidea and Collembola communities in burnt and unburnt habitats in central Italy. Italian Journal of Zoology, 72, 127- 140.

Potts, S.G., Vulliamy, B., Dafni, A., Ne'eman, G., O'Toole, C., Roberts, S., and Willmer, P. (2003) Response of plant-pollinator communities to fire: changes in diversity, abundance and floral reward structure. Oikos, 101, 103-112.

Poyry, J., Paukkunen, J., Heliola, J., and Kuussaari, M. (2009) Relative contributions of local and regional factors to species richness and total density of butterflies and moths in semi-natural grasslands. Oecologia, 160, 577-587.

Prevedello, J.A. and Vieira, M.V. (2010) Does the type of matrix matter? A quantitative review of the evidence. Biodiversity and Conservation, 19, 1205- 1223.

Prugh, L.R., Hodges, K.E., Sinclair, A.R.E., and Brashares, J.S. (2008) Effect of habitat area and isolation on fragmented animal populations. Proceedings of the National Academy of Sciences of the United States of America, 105, 20770- 20775.

Pyke, D.A., Herrick, J.E., Shaver, P., and Pellant, M. (2002) Rangeland health attributes and indicators for qualitative assessment. Journal of Range Management, 55, 584-597.

R Development Core Team (2011) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/.

Raghu, S., Drew, R.A.I., and Clarke, A.R. (2004) Influence of host plant structure and microclimate on the abundance and behavior of a tephritid fly. Journal of Insect Behavior, 17, 179-190.

Read, J.L. and Andersen, A.N. (2000) The value of ants as early warning bioindicators: responses to pulsed cattle grazing at an Australian arid zone locality. Journal of Arid Environments, 45, 231-251.

Redak, R.A. (2000) Arthropods and multispecies habitat conservation plans: are we missing something? Environmental Management, 26 (S1), S97-S107.

Reid, A.M. and Hochuli, D.F. (2007) Grassland invertebrate assemblages in managed landscapes: Effect of host plant and microhabitat architecture. Austral Ecology, 32, 708-718.

Reznik, S.Y. (1993) Influence of target plant density on herbivorous insect oviposition choice: Ambrosia artemisiifolia L. (Asteraceae) and Zygogramma suturalis F. (Coleoptera: Chrysomelidae) Biocontrol Science and Technology, 3, 105-113.

210

Ridolfi, L., Laio, F., and D'Odorico, P. (2008) Fertility island formation and evolution in dryland ecosystems. Ecology and Society, 13.

Root, R.B. (1973) Organization of a plant-arthropod association in simple and diverse habitats - fauna of collards (Brassica oleracea). Ecological Monographs, 43, 95-120.

Roques, K.G., O'Connor, T.G., and Watkinson, A.R. (2001) Dynamics of shrub encroachment in an African savanna: relative influences of fire, herbivory, rainfall and density dependence. Journal of Applied Ecology, 38, 268-280.

Rowe, D.K., Parkyn, S., Quinn, J., Collier, K., Hatton, C., Joy, M.K., Maxted, J., and Moore, S. (2009) A rapid method to score stream reaches based on the overall performance of their main ecological functions. Environmental Management, 43, 1287-1300.

Samocha, Y. and Sternberg, M. (2010) Herbivory by sucking mirid bugs can reduce nectar production in Asphodelus aestivus Brot. Arthropod-Plant Interactions, 4, 153-158.

Sandell, P., Tolhurst, K., Dalton, J., Scott, B., and Smith, M. (2006) Fire management prescriptions for the Victorian mallee parks. Proceedings of the Royal Society of Victoria, 118, 395-412.

Sanders, D., Nickel, H., Grützner, T., and Platner, C. (2008) Habitat structure mediates top-down effects of spiders and ants on herbivores. Basic and Applied Ecology, 9, 152-160.

Sanders, D. and van Veen, F.J.F. (2011) Ecosystem engineering and predation: the multi-trophic impact of two ant species. Journal of Animal Ecology, 80, 569- 576.

Sanford, M.P., Manley, P.N., and Murphy, D.D. (2009) Effects of urban development on ant communities: implications for ecosystem services and management. Conservation Biology, 23, 131-141.

Santos, P.F. and Whitford, W.G. (1981) The effects of microarthropods on litter decomposition in a Chihuahuan desert ecosystem. Ecology, 62, 654-663.

Santos, P.F.E., DePree, E., and Whitford, W.G. (1978) Spatial distribution of litter and microarthropods in a Chihuahuan desert ecosystem. Journal of Arid Environments, 1, 41-48.

Sarin, C. and Bergman, K.O. (2010) Habitat utilisation of burnet moths (Zygaena spp.) in southern Sweden: a multi-scale and multi-stage perspective. Insect Conservation and Diversity, 3, 180-193.

Schade, J.D. and Hobbie, S.E. (2005) Spatial and temporal variation in islands of fertility in the Sonoran Desert. Biogeochemistry, 73, 541-553.

211

Schaffers, A.P., Raemakers, I.P., Sýkora, K.V., and Ter Braak, C.J.F. (2008) Arthropod assemblages are best predicted by plant species composition. Ecology, 89, 782-794.

Schell, S.P. and Lockwood, J.A. (1997) Spatial analysis of ecological factors related to rangeland grasshopper (Orthoptera : Acrididae) outbreaks in Wyoming. Environmental Entomology, 26, 1343-1353.

Schlesinger, W.H., Reynolds, J.F., Cunningham, G.L., Huenneke, L.F., Jarrell, W.M., Virginia, R.A., and Whitford, W.G. (1990) Biological feedbacks in global desertification. Science, 247, 1043-1048.

Schlesinger, W.H., Raikes, J.A., Hartley, A.E., and Cross, A.E. (1996) On the spatial pattern of soil nutrients in desert ecosystems. Ecology, 77, 364-374.

Schwartz, M.W. (1999) Choosing the appropriate scale of reserves for conservation. Annual Review of Ecology and Systematics, 30, 83-108.

Seastedt, T.R. and Crossley, D.A. (1984) The influence of arthropods on ecosystems. BioScience, 34, 157-161.

Shachak, M. and Yair, A. (1984) Population dynamics and the role of Hemilepistus reaumuri in a desert ecosystem. Symposium of Zoological Society of London, 53, 295-314.

Shachak, M., Boeken, B., Groner, E., Kadmon, R., Lubin, Y., Meron, E., Ne'Eman, G., Perevolotsky, A., Shkedy, Y., and Ungar, E.D. (2008) Woody species as landscape modulators and their effect on biodiversity patterns. Bioscience, 58, 209-221.

Shelef, O. and Groner, E. (2011) Linking landscape and species: Effect of shrubs on patch preference of beetles in arid and semi-arid ecosystems. Journal of Arid Environments, 75, 960-967.

Shorrocks, B., Marsters, J., Ward, I., and Evennett, P.J. (1991) The fractal dimension of lichens and the distribution of arthropod body lengths. Functional Ecology, 5, 457-460.

Shumway, S.W. (2000) Facilitative effects of a sand dune shrub on species growing beneath the shrub canopy. Oecologia, 124, 138-148.

Silveira, J.M., Barlow, J., Louzada, J., and Moutinho, P. (2010) Factors affecting the abundance of leaf-litter arthropods in unburned and thrice-burned seasonally-dry Amazonian forests. Plos One, 5.

Sirami, C., Seymour, C., Midgley, G., and Barnard, P. (2009) The impact of shrub encroachment on savanna bird diversity from local to regional scale. Diversity and Distributions, 15, 948-957.

212

Smith, J.L., Halvorson, J.J., and Bolton, H. (1994) Spatial relationships of soil microbial biomass and c and n mineralization in a semiarid shrub-steppe ecosystem. Soil Biology & Biochemistry, 26, 1151-1159.

Smithers, C.N. (1991) Psocoptera. The insects of Australia: A textbook for students and researchers. 2nd edition. (ed. by CSIRO). Melbourne University Press, Carlton, Australia.

Smyth, A.K., Brandle, R., Chewings, V., Read, J., Brook, A., and Fleming, M. (2009) A framework for assessing regional biodiversity condition under changing environments of the arid Australian rangelands. Rangeland Journal, 31, 87-101.

Sneva, F.A. (1979) The western harvester ants: their density and hill size in relation to herbaceous productivity and big sagebrush cover. Journal of Rangeland Management, 32, 46-47.

Sobrinho, T.G. and Schoereder, J.H. (2007) Edge and shape effects on ant (Hymenoptera : Formicidae) species richness and composition in forest fragments. Biodiversity and Conservation, 16, 1459-1470.

Spiesman, B.J. and Cumming, G.S. (2008) Communities in context: the influences of multiscale environmental variation on local ant community structure. Landscape Ecology, 23, 313-325.

Stoner, K.J.L. and Joern, A. (2004) Landscape vs. local habitat scale influences to insect communities from tallgrass prairie remnants. Ecological Applications, 14, 1306-1320.

Stoutjesdijk, P. and Barkmann, J.J. (1992) Microclimate, vegetation and fauna. Cambridge University Press, New York.

Strong, D.R., Lawton, J.H., and Southwood, R. (1984) Insects on plants: community patterns and mechanisms. Harvard University Press, Cambridge (MA).

Stutchbury, B.J.M. (2007) The effects of habitat fragmentation on animals: Gaps in our knowledge and new approaches. Canadian Journal of Zoology, 85, 1015- 1016.

Sugiura, S. (2011) Structure and dynamics of the parasitoid community shared by two herbivore species on different host plants. Arthropod-Plant Interactions, 5, 29-38.

Suter, G.W. (1993) A critique of ecosystem health concepts and indexes. Environmental Toxicology and Chemistry, 12, 1533-1539.

Tews, J., Brose, U., Grimm, V., Tielborger, K., Wichmann, M.C., Schwager, M., and Jeltsch, F. (2004) Animal species diversity driven by habitat

213

heterogeneity/diversity: the importance of keystone structures. Journal of Biogeography, 31, 79-92.

Thackway, R. and Creswell, I.D., eds. (1995) An interim biogeographic regionalisation of Australia: a framework for establishing the national system of reserves. Australian Nature Conservation Agency, Canberra.

Thomas, C.D., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., de Siqueira, M.F., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega- Huerta, M.A., Peterson, A.T., Phillips, O.L., and Williams, S.E. (2004) Extinction risk from climate change. Nature, 427, 145-148.

Thomas, J.A., Bourn, N.A.D., Clarke, R.T., Stewart, K.E., Simcox, D.J., Pearman, G.S., Curtis, R., and Goodger, B. (2001) The quality and isolation of habitat patches both determine where butterflies persist in fragmented landscapes. Proceedings of the Royal Society of London Series B-Biological Sciences, 268, 1791-1796.

Thompson, W.A. and Eldridge, D.J. (2005) Plant cover and composition in relation to density of Callitris glaucophylla (white cypress pine) along a rainfall gradient in eastern Australia. Australian Journal of Botany, 53, 545-554.

Thornton, D.H., Branch, L.C., and Sunquist, M.E. (2011) The influence of landscape, patch, and within-patch factors on species presence and abundance: a review of focal patch studies. Landscape Ecology, 26, 7-18.

Titus, J.H., Nowak, R.S., and Smith, S.D. (2002) Soil resource heterogeneity in the Mojave Desert. Journal of Arid Environments, 52, 269-292.

Tongway, D. (1995) Monitoring soil productive potential. Environmental Monitoring and Assessment, 37, 303-318.

Tongway, D.J., Ludwig, J.A., and Whitford, W.G. (1989) Mulga log mounds - fertile patches in the semi-arid woodlands of eastern Australia. Australian Journal of Ecology, 14, 263-268.

Tongway, D.J. and Ludwig, J.A. (1990) Vegetation and soil patterning in semiarid mulga lands of eastern Australia. Australian Journal of Ecology, 15, 23-34.

Tongway, D.J. and Ludwig, J.A. (1997) The nature of landscape dysfunction in the rangelands. Landscape ecology - function and management. Principles from Australia's rangelands. (ed. by J.A. Ludwig, D.J. Tongway, D.O. Freudenberger, J.C. Noble & K.C. Hodgkinson), pp. 49-61. CSIRO, Melbourne.

Tongway, D.J. and Ludwig, J.A. (2011) Restoring degraded landscapes: putting principles into practice. Island Press, Washington.

214

Tzialla, C.E., Veresoglou, D.S., Papakosta, D., and Mamolos, A.P. (2006) Changes in soil characteristics and plant species composition along a moisture gradient in a Mediterranean pasture. Journal of Environmental Management, 80, 90-98.

Uetz, G.W. (1979) Influence of variation in litter habitats on spider communities. Oecologia, 40, 29-42.

Underwood, E.C. and Fisher, B.L. (2006) The role of ants in conservation monitoring: If, when, and how. Biological Conservation, 132, 166-182.

Van Auken, O.W. (2000) Shrub invasions of North American semiarid grasslands. Annual Review of Ecology and Systematics, 31, 197-215.

Van Auken, O.W. (2009) Causes and consequences of woody plant encroachment into western North American grasslands. Journal of Environmental Management, 90, 2931-2942.

Vanderwoude, C., Andersen, A.N., and House, P.N. (1997) Community organisation, biogeography and seasonality of ants in an open forest of south- eastern Queensland. Australian Journal of Zoology, 45, 523-537.

Vasconcelos, H.L., Leite, M.F., Vilhena, J.M.S., Lima, A.P., and Magnusson, W.E. (2008) Ant diversity in an Amazonian savanna: Relationship with vegetation structure, disturbance by fire, and dominant ants. Austral Ecology, 33, 221-231.

Venables, W.N. and Ripley, B.S. (2002) Modern Applied Statistics with S. Fourth edition. Springer, New York.

Vos, V.C.A., van Ruijven, J., Berg, M.P., Peeters, E., and Berendse, F. (2011) Macro-detritivore identity drives leaf litter diversity effects. Oikos, 120, 1092- 1098.

Wagner, D. (1997) The influence of ant nests on Acacia seed production, herbivory and soil nutrients. Journal of Ecology, 85, 83-93.

Wagner, D. and Jones, J.B. (2004) The contribution of harvester ant nests, Pogonomyrmex rugosus (Hymenoptera, Formicidae), to soil nutrient stocks and microbial biomass in the Mojave Desert.

Walker, L.R., Wardle, D.A., Bardgett, R.D., and Clarkson, B.D. (2010) The use of chronosequences in studies of ecological succession and soil development. Journal of Ecology, 98, 725-736.

Walsh, C. and Mac Nally, R. (2003) The hier.part package. Hierarchical Partitioning. R project for statistical computing. URL: http://cran.r-project.org/.

Warburg, M.R. (1968) Behavioral adaptations of terrestrial isopods. American Zoologist, 8, 545-559.

215

Wardle, D.A., Hyodo, F., Bardgett, R.D., Yeates, G.W., and Nilsson, M.C. (2011) Long-term aboveground and belowground consequences of red wood ant exclusion in boreal forest. Ecology, 92, 645-656.

Watson, I.W., Novelly, P.E., and Thomas, P.W.E. (2007) Monitoring changes in pastoral rangelands - the Western Australian Rangeland Monitoring System (WARMS). Rangeland Journal, 29, 191-205.

Watson, J., Freudenberger, D., and Paull, D. (2001) An assessment of the focal-species approach for conserving birds in variegated landscapes in southeastern Australia. Conservation Biology, 15, 1364-1373.

Watson, J.E.M., Whittaker, R.J., and Dawson, T.P. (2004) Avifaunal responses to habitat fragmentation in the threatened littoral forests of south-eastern Madagascar. Journal of Biogeography, 31, 1791-1807.

Watson, P.J., Bradstock, R.A., and Morris, C. (2009) Fire frequency influences composition and structure of the shrub layer in an Australian subcoastal temperate grassy woodland. Austral Ecology, 34, 218-232.

Weltzin, J.F. and Coughenour, M.B. (1990) Savanna tree influence on understory vegetation and soil nutrients in northwestern Kenya. Journal of Vegetation Science, 1, 325-334.

Westbrooke, M., Miller, J.D., and Kerr, M.K.C. (1998) The vegetation of the Scotia 1: 100 000 map sheet, western New South Wales. Cunninghamia, 5, 665-684.

Western, D. (1992) The biodiversity crisis - a challenge for biology. Oikos, 63, 29-38.

Whitford, W.G., Steinberger, Y., and Ettershank, G. (1982) Contributions of subterranean termites to the economy of Chihuahuan desert ecosystems. Oecologia, 55, 298-302.

Whitford, W.G. (1996) The importance of the biodiversity of soil biota in arid ecosystems. Biodiversity and Conservation, 5, 185-195.

Whitford, W.G., Anderson, J., and Rice, P.M. (1997) Stemflow contribution to the 'fertile island' effect in creosotebush, Larrea tridentata. Journal of Arid Environments, 35, 451-457.

Whitford, W.G., De Soyza, A.G., Van Zee, J.W., Herrick, J.E., and Havstad, K.M. (1998) Vegetation, soil, and animal indicators of rangeland health. Environmental Monitoring and Assessment, 51, 179-200.

Whitford, W.G., Van Zee, J., Nash, M.S., Smith, W.E., and Herrick, J.E. (1999) Ants as indicators of exposure to environmental stressors in North American desert grasslands. Environmental Monitoring and Assessment, 54, 143-171.

216

Whitford, W.G. (2000) Keystone arthropods as webmasters in desert ecosystems. Invertebrates as webmasters in ecosystems (ed. by D.C. Coleman & P.F. Hendrix), pp. 25-41. CABI Publishing, New York.

Williams, R.J., Duff, G.A., Bowman, D., and Cook, G.D. (1996) Variation in the composition and structure of tropical savannas as a function of rainfall and soil texture along a large-scale climatic gradient in the Northern Territory, Australia. Journal of Biogeography, 23, 747-756.

Wilson, A.D., Tongway, D.J., Graetz, R.D., and Young, M.D. (1984) Range and inventory monitoring. Management of Australia's rangelands (ed. by G.N. Harrington, A.D. Wilson & M.D. Young), pp. 113-128. CSIRO, Melbourne.

Zhang, Q.H. and Zak, J.C. (1995) Effects of gap size on litter decomposition and microbial activity in a subtropical forest. Ecology, 76, 2196-2204.

Zurlini, G. and Girardin, P. (2008) Introduction to the special issue on "Ecological indicators at multiple scales". Ecological Indicators, 8, 781-782.

Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., and Smith, G.M. (2009) Mixed effects models and extensions in ecology with R Springer, NY.

Zvereva, E.L., Lanta, V., and Kozlov, M.V. (2010) Effects of sap-feeding insect herbivores on growth and reproduction of woody plants: a meta-analysis of experimental studies. Oecologia, 163, 949-960.

217