Why Are Woodland-Dependent Avian Insectivore Populations Vulnerable to Declines in Highly-Modified Landscapes? Anita Jean Cosgrove B Sc

Why are woodland-dependent avian insectivore populations vulnerable to declines in highly-modified landscapes? Anita Jean Cosgrove B Sc. Hons

A thesis submitted for the degree of Doctor of Philosophy at The University of Queensland in 2017 School of Earth and Environmental Sciences

i Abstract

Habitat loss, fragmentation, and degradation are the leading causes of faunal population declines. Substantial effort has been invested in identifying how these landscape changes are related to population parameters. However, to effectively mitigate declines, we must understand the underlying mechanisms impacted by landscape change. Woodland-dependent avian insectivores seem particularly vulnerable, and decline at rates disproportionately greater than other foraging guilds. This pattern suggests that reduced food availability in less-wooded landscapes may be a contributing factor. However, the evidence supporting this hypothesis is sparse.

I adopted a multidisciplinary approach to investigate the mechanisms driving population changes of an avian insectivore in highly-modified landscapes by integrating physiological indices with measures of prey availability and landscape structure. First, I presented a framework for conceptualising how the spatiotemporal scale of fragmentation influences the mechanisms impacting populations. I then used the eastern yellow robin (Eopsaltria australis) in the Brigalow Belt South bioregion of southern Queensland to test the importance of local and landscape factors on robin site occupancy, prey density, and robin condition. I also investigated whether an index of chronic stress could be used to predict changes to robin occurrence and identify potential threats to the population. I concluded this work by conducting a supplementary feeding experiment to determine whether food availability was causing the variation in robin condition.

The conceptual framework I developed contrasts the ways in which habitat fragmentation per se impacts faunal populations by restricting movement among habitat patches. When movement is restricted over small spatiotemporal scales, such as within a home range or on a daily basis, access to critical resources may be limited. Restrictions over intermediate scales can inhibit dispersal movements which reduces demographic exchange and rescue. Large-scale movement restrictions that prevent rare dispersal events may impede gene flow. I demonstrated through a global review of the habitat fragmentation literature, that limited resource access was the most understudied of these mechanisms, even though it can affect populations rapidly, while impeded gene flow received the most attention by far.

A pilot study conducted in the study region four years earlier found that robins were more physiologically stressed (had higher heterophil:lymphocyte ratios) in sites with less woodland in the surrounding landscape, but site occupancy was unrelated to woodland cover. Re-examining the same study system, I established that woodland cover was now the strongest predictor of

ii occupancy. While woodland cover at a local scale (within 500 m) was important for robin persistence, woodland cover at a broader scale relevant to dispersal movements (within 5 km) was not. While my stress index could not accurately predict the precise locations of future extirpations, patterns of stress indices (birds were more stressed where woodland cover was lower) did foreshadow a change to the pattern of occurrence. Extirpations from sites with low levels of woodland cover were balanced by colonisations of more-wooded sites.

Surprisingly, I found that arthropod prey density was greater at sites that had less surrounding woodland. This result was primarily driven by a large number of Formicidae (ants). However, Formicidae have low nutritional value and once they were excluded from the results, soil moisture, not woodland cover, became the primary determinant of arthropod density. I predicted that robin condition would be strongly associated with arthropod density, but this was not the case. Instead, robins had elevated stress levels at sites with apparently more-favourable foraging habitat. These findings confirm that prey density does not explain this species’ apparent vulnerability in less- wooded landscapes. However, there are other ways in which food availability could be affected in structurally-altered landscapes, such as through a reduced area from which to source prey. Providing supplementary food to robins across sites with varying degrees of woodland cover showed that condition did not generally improve after feeding, suggesting that the birds were not food limited; however, statistical power was low due to challenges in recapturing experimental birds.

The findings from my thesis have provided insights into why woodland-dependent avian insectivores are more vulnerable to declines in highly-modified agricultural landscapes. I identified that prey density was unrelated to the apparent vulnerability of eastern yellow robins in less-wooded landscapes. This conclusion runs counter to work in other systems, perhaps because my study region is located on more-fertile soil or has a shorter clearing history. Robins were more affected by woodland extent over small scales than landscape connectivity over broad scales, so conservation efforts focused on promoting brigalow regrowth alongside existing roadside strips of woodland may be beneficial even if landscape-scale restoration is impractical. I found some evidence that chronic stress indices can be used to predict changes to occurrence before the changes are apparent in presence/absence data. This result has important implications for applied conservation management, and future studies should examine how the nature, duration, and intensity of a stressor affect the predictive capacity of this stress index.

iii Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my research higher degree candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis.

iv Publications during candidature

Peer-reviewed papers

Publications included in this thesis

Cosgrove AJ, McWhorter TJ, Maron M. 2017. Using individual-condition measures to predict the long-term importance of habitat extent for population persistence. Conservation Biology DOI: 10.1111/cobi.12903 – incorporated as Chapter 3. There are several minor differences between the early online version of this article and the version that appears in this thesis. Three additional paragraphs have been added to Chapter 3 discussing the limitations of the work.

Contributor Statement of contribution Anita Cosgrove (Candidate) Conceptualised and designed the study (35%) Collected the data (100%) Analysed the data (100%) Interpreted the results (50%) Drafted the paper (100%) Edited the paper (55%) Todd McWhorter Edited the paper (15%) Martine Maron Conceptualised and designed the study (65%) Interpreted the results (50%) Edited the paper (30%)

v Contributions by others to the thesis

Five jointly prepared or submitted manuscripts form part of this thesis (Chapters 2-6). I undertook the majority of work for these articles, including generating ideas, data collection and analysis, interpretation, and drafting and revising the manuscripts. Martine Maron and Todd McWhorter contributed to the development of the ideas, provided guidance on study design and statistical methods, and assisted with editing written material. I originally stained all blood smears used in Chapters 5 and 6, while Todd McWhorter restained the bulk of these blood smears and permanently mounted a subset of the slides. Clive McAlpine provided helpful feedback on how to make Chapter 2 more publishable.

Statement of parts of the thesis submitted to qualify for the award of another degree

None

vi Acknowledgements

“The earth has its music for those who will listen” (Holmes 1955).

My journey to becoming a scientist has been life changing. There have been many challenges along the way and I am immensely grateful to everyone who provided guidance and support when I needed it most.

I will be forever grateful to Danielle Shanahan who presented a talk about the effects of habitat fragmentation on woodland birds while I was still employed in the insurance industry. She was the first person to open my eyes to the possibility that I could follow my dream to work on an issue close to my heart. It was intensely satisfying and remarkably fitting to undertake a PhD on this exact topic several years later.

I am deeply thankful to my primary supervisor, Martine Maron. This work would not have been possible without her leadership, guidance, and encouragement. Her passion for birds, ecology, and conservation was infectious and I feel blessed to have never once lost my enthusiasm for the topic. Her field assistance was invaluable and it was a joy to share that experience with someone who “gets it”. No matter how much literature I read and integrated, she always brought a different perspective to the table. Surprisingly, I learnt to enjoy writing which was due in large part to the detailed feedback she provided that allowed me to refine my skills. However, I think her greatest lesson was the importance of building a story – an invaluable skill for developing research questions, interpreting results, writing manuscripts, and presenting ideas. She has taught my detail- orientated brain to step back and look at the bigger picture – no mean feat!

I extend sincere thanks to my co-supervisor, Todd McWhorter, for investing his time and energy into my development and success. His physiological expertise was invaluable throughout the project, and he went above and beyond the call of duty by restaining and mounting the bulk of my blood smears. I am deeply appreciative that he and Andrea McWhorter welcomed me to their South Australian home and introduced me to their local woodlands. However, I am most grateful that he consistently modelled how to be a responsible scientist and was always so open and patient with my endless questions. I am a better scientist today because of him.

Special thanks to Warren and Lynelle Urquhart. They generously provided me with accommodation and access to field sites during my Moonie field work. Perhaps more importantly, they provided

vii companionship, shared a love of the bush, took me on scenic flights, and gave me a crash course on how to check if cows are pregnant. I will always treasure the time I spent with them.

In a similar vein, extended periods of solo field work were isolating experiences and I am truly thankful to everyone who made them less so. My Myall Park family included Lorraine, Dave, Janine, and John. At Goondiwindi, Raquel took me under her wing and very kindly invited me into her home for Christmas lunch. Their friendships brought me comfort and a sense of belonging.

I am grateful to all the landholders and managers who allowed me to work on their property: Tony and Carolyn Beattie; Len and Kate Cover; Robert, Joanne, and Nick Creagh; Rolf and Mary Crouch; Gavin and Helen Dales; Sean and Penny Grams; Laughlan and Fiona Hill; Warwick and Diane Hill; David and Penny Murphy; David and Meryl Newberry; Colin and Carol Piror, and their daughter Fiona; Geoff and Sue Reilly; Garnett and Heather Reis; Don and Deb Taylor; Stephen and Jodi Toohey; Steve Toot; Warren and Lynelle Urquhart; Bev Wallace; Robert and Eva Walton; Rowell and Debbie Walton; Gil at “Kerimbilla”; and Wally at “Tullamun”. Their woodland patches provided my study with the much needed middle ground between the large, intact protected areas and the narrow strips of roadside vegetation. Their generosity has improved this study immensely.

This project could not have proceeded without financial assistance from BirdLife Australia, Birds Queensland, the Wildlife Preservation Society of Queensland, the Australasian Society for the Study of Animal Behaviour, the School of Geography, Planning and Environmental Management at the University of Queensland, and an Australian Postgraduate Award from the Commonwealth Government. I thank these groups for their support.

I received assistance and training from many people. David Stewart and Will Goulding kindly gave me recordings of robin calls. Aaron Cosgrove and Anne-Pernelle Duc volunteered their time to assist me in the field. The GPEM professional staff helped with all my administrative and technical needs, particularly: Michael Tobe; Alan Victor; Linda Nothdurft; Joel Almond; Eros Bonilla; Suhan Chin; Genna McNicol; Lorraine Liang; Claire Lam; and Judy Nankiville. Jon Coleman, Brenda and Graham Smith, Stephen MacDonald, and David Edwards provided mist netting and bird-banding training. Nick Clark showed me how to blood sample birds, while Lucy Woolford taught me how to differentiate blood cells. David Drynan from the ABBBS provided advice about banding and dispersal movements, while James O’Connor and Glenn Ehmke from BirdLife Australia provided preliminary results from their State of Australia’s Birds 2015 report for the Brigalow Belt region.

viii I appreciated the support, discussions, and friendship from my colleagues in the Landscape Ecology and Conservation Group. Special mention to: Jeremy Simmonds; Will Goulding; Rowan Eisner; Zoe Stone; Andres Felipe Suarez Castro; Melissa Bruton; Emma Burgess; Leonie Seabrook; Christine Adams-Hosking; Clive McAlpine; James Watson; and Jonathan Rhodes. You all listened to my worries, shared your experiences, provided advice, and stimulated my thinking. My experience was richer because of you. I also received support from the coaches and peers at the Academic Writing Club. You taught me how small, regular steps can make a huge difference, not just in terms of writing, but throughout many aspects of life. I was able to face my fears and complete my PhD thanks to our daily “poms”.

Finally, deepest thanks to my husband, Aaron Cosgrove, for his love and support throughout my PhD. He helped me so much, including: visiting me in the field when I could not travel home; lending me his car and packing it to the brim; constructing feeding stations; and taking my check-in calls every day to make sure I was safe. He helped me empty pitfall traps and collect habitat data – you know it is true love when someone is willing to count twigs on the ground for you. He never once complained about my commitments or the drain on our finances. I am constantly amazed that he can readily tell others what my project is about and has developed a keen interest in the birds at home. Most importantly, he always believed in me. Thank you.

ix Keywords

Australian woodland birds, brigalow, condition indices, eastern yellow robin, Eopsaltria australis, food availability, habitat fragmentation, habitat loss, invertebrates, stress

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 050104, Landscape Ecology, 50% ANZSRC code: 060203, Ecological Physiology, 35% ANZSRC code: 050202, Conservation and Biodiversity, 15%

Fields of Research (FoR) Classification

FoR code: 0501, Ecological Applications, 50% FoR code: 0602, Ecology, 30% FoR code: 0606, Physiology, 20%

x Table of contents

Abstract ...... ii Declaration by author ...... iv Publications during candidature ...... v Publications included in this thesis ...... v Contributions by others to the thesis...... vi Statement of parts of the thesis submitted to qualify for the award of another degree ...... vi Acknowledgements ...... vii Keywords ...... x Australian and New Zealand Standard Research Classifications (ANZSRC) ...... x Fields of Research (FoR) Classification ...... x Table of contents ...... xi List of figures ...... xv List of tables ...... xviii List of plates ...... xxi

CHAPTER 1. Introduction ...... 1 1.1 Theory and concepts underlying the thesis ...... 2 1.1.1 The impacts of land-use change ...... 2 1.1.2 Australian woodland bird declines ...... 4 1.1.3 Effects of limited resource availability in highly-modified landscapes ...... 5 1.1.4 Using physiological measures to understand mechanisms causing population declines .. 7 1.2 Thesis structure and format ...... 9 1.3 Thesis objectives and research questions ...... 9 1.4 Thesis outline ...... 12 1.5 Study region – Brigalow Belt South bioregion ...... 13 1.5.1 Ecology of the Brigalow Belt South ...... 13 1.5.2 Historical land-use changes ...... 14 1.6 Study species – eastern yellow robin ...... 16 1.7 Suitability of study region and species ...... 18

CHAPTER 2. Habitat fragmentation and movement: linking scale and mechanism ...... 20 2.1 Abstract ...... 21 2.2 Introduction ...... 21 2.3 Conceptual framework ...... 23

xi 2.3.1 Mechanism and scale are linked ...... 23 2.3.2 Movement restrictions disrupt three key mechanisms ...... 23 2.3.2.1 Fine scale mechanism – limited resource access ...... 24 2.3.2.2 Intermediate scale mechanism – restricted demographic exchange ...... 25 2.3.2.3 Broad scale mechanism – impeded gene flow ...... 26 2.4 Literature review methods ...... 27 2.4.1 Literature search strategy and inclusion criteria ...... 27 2.4.2 Data collection and analysis ...... 28 2.5 Results ...... 29 2.6 Discussion ...... 33 2.6.1 Implications and limitations of the conceptual framework ...... 33 2.6.2 Distribution of research effort among mechanisms ...... 35 2.6.3 Geographic and taxonomic biases ...... 36 2.6.4 Conclusions ...... 37

CHAPTER 3: Using individual condition measures to predict the long-term importance of habitat extent for population persistence ...... 38 3.1 Abstract ...... 39 3.2 Introduction ...... 39 3.3 Methods ...... 41 3.3.1 Study region ...... 41 3.3.2 Study species ...... 41 3.3.3 Occupancy and individual condition data ...... 43 3.3.4 Explanatory variables ...... 43 3.3.5 Data analysis ...... 44 3.4 Results ...... 46 3.4.1 Predicting locations of extirpations from H:L ratios ...... 46 3.4.2 Predicting occupancy relationships from relationships between H:L ratios and habitat extent ...... 47 3.4.3 Testing the effect of landscape context at larger extents ...... 49 3.5 Discussion ...... 50 3.5.1 Scale of woodland effect ...... 53 3.5.2 Management implications...... 54

CHAPTER 4: Determinants of food availability for a ground-foraging insectivore differ when the nutritional value of prey is considered...... 55

xii 4.1 Abstract ...... 56 4.2 Introduction ...... 56 4.3 Methods ...... 59 4.3.1 Study area ...... 59 4.3.2 Study species ...... 60 4.3.3 Site selection and explanatory variables ...... 61 4.3.4 Arthropod pitfall trapping ...... 62 4.3.5 Statistical analyses ...... 63 4.4 Results ...... 65 4.4.1 All arthropod taxa combined ...... 67 4.4.2 All arthropod taxa excluding Formicidae ...... 70 4.4.3 Responses by individual arthropod taxa to woodland extent ...... 74 4.5 Discussion ...... 76 4.5.1 Arthropod responses to landscape structure ...... 76 4.5.2 Soil moisture and ground cover ...... 77 4.5.3 Prey accessibility and temporal fluctuations ...... 78 4.5.4 Limitations ...... 79 4.5.5 Conclusion ...... 79

CHAPTER 5 Individual condition in an Australian avian insectivore is related to foraging substrate characteristics ...... 81 5.1 Abstract ...... 82 5.2 Introduction ...... 82 5.3 Methods ...... 84 5.3.1 Study area ...... 84 5.3.2 Study species ...... 86 5.3.3 Condition indices ...... 86 5.3.4 Explanatory variables ...... 87 5.3.5 Statistical analysis ...... 89 5.4 Results ...... 91 5.5 Discussion ...... 95 5.5.1 Conclusion ...... 97

CHAPTER 6 Food restriction for an avian insectivore in a highly-modified landscape: a supplementary feeding experiment on the eastern yellow robin (Eopsaltria australis) ...... 99 6.1 Abstract ...... 100

xiii 6.2 Introduction ...... 100 6.3 Methods ...... 102 6.3.1 Study region ...... 102 6.3.2 Study species ...... 103 6.3.3 Habitat mapping ...... 103 6.3.4 Experimental design ...... 103 6.3.5 Experimental field methods ...... 104 6.3.6 Calculation of condition data ...... 105 6.3.7 Statistical analyses ...... 106 6.4 Results ...... 107 6.5 Discussion ...... 110 6.5.1 Conclusion ...... 113

CHAPTER 7 Synthesis ...... 114 7.1 Thesis overview ...... 115 7.2 Contributions to ecological and physiological knowledge ...... 119 7.2.1 Insectivorous birds and their prey ...... 121 7.2.2 Conservation physiology ...... 124 7.2.3 Methods for evaluating mechanisms impacting populations in structurally-altered landscapes ...... 126 7.3 Prognosis and management implications ...... 127 7.4 Limitations of the study and suggested future research ...... 131 7.5 Conclusion ...... 132

References ...... 134

Appendices ...... 164 Appendix A ...... 164 Appendix B ...... 197 Appendix C ...... 199 Appendix D ...... 204

xiv List of figures

Figure 1.1. A hierarchical conceptual model outlining the processes resulting from habitat removal, how these processes manifest as changes to landscape structure, and how changes to landscape structure are related to possible mechanistic causes of population declines. a (Ford et al. 2001); b (Watson 2011); c (Langlois et al. 2001); d (Connor & McCoy 1979); e (Richmond et al. 2011); f (Robinson et al. 1995); g (Robinson & Sherry 2012); h (Eyre et al. 2009); i (Karr & Freemark 1983); j (Keyghobadi 2007); k (Delaney et al. 2010)…………….…………………………………………..3

Figure 1.2. An overview of the thesis structure. Grey boxes denote chapters written as manuscripts for submission to peer-reviewed journals………………………………………………...…………10

Figure 1.3. In this thesis, I integrate information about food availability, landscape structure, and individual condition. The junctions addressed by each chapter of the thesis are indicated…...……11

Figure 1.4. The geographic range of the Brigalow Belt South bioregion which spans Queensland and New South Wales, Australia, together with woodland cover remaining in the study region for this thesis. Brigalow woodland is denoted with dark grey, while all other woodland is marked in light grey………………………………………………………………………………………...…..15

Figure 2.1. A conceptual framework outlining the mechanisms through which detrimental impacts can occur when faunal movements are disrupted by habitat fragmentation across a range of temporal and spatial scales. The absolute values of the scales of effect will vary amongst taxa depending on the dispersal ability, foraging range requirements, and generation time of the species……………………………………………………………………………………………….24

Figure 2.2. Cumulative number of articles published since 1994 from a random selection of articles that specifically investigated a mechanism (limited resource access, restricted demographic exchange, or impeded gene flow) impacted by habitat fragmentation…………………………...…30

Figure 2.3. Geographical distribution of studies included in this review using a Mollweide projection. Dotted lines represent the Tropics of Capricorn and Cancer, and the boundary of the subtropical regions. The mechanisms investigated were limited resource access (black triangles), restricted demographic exchange (white circles), or impeded gene flow (cross). Articles that focused on both demographic exchange and impeded gene flow are denoted with plus signs……..31

xv Figure 2.4. Proportion of articles from a random selection of articles published since 1994 that studied the effect of habitat fragmentation on an ecological mechanism categorised by the geographic zone where the study was conducted. “Temperate” includes the temperate zone plus higher latitudes. The total number of articles included in each zone is indicated above each bar….32

Figure 2.5. Proportion of articles from a random selection of articles published since 1994 that researched the effect of habitat fragmentation on an ecological mechanism divided by the taxa studied. The total number of articles examined for each taxon is noted at the top of each bar…..…32

Figure 3.2. Percentage of independently-explained variance for each independent variable from a model of eastern yellow robin extirpations. The independently-explained variance was non- significant for all variables………………………………………………………………...………..47

Figure 3.3. Percentage of independently-explained variance for each independent variable from models of eastern yellow robin presence. (a) shows the results obtained by Maron et al. (2012) based on surveys conducted in 2009-2010 (n = 42), while (b) shows the results obtained from the current study in 2013-2014 (n = 42). Black bars indicate the independently-explained variance was significant……………………………………………………………………………..…………….49

Figure 3.4. Percentage of independently-explained variance for each independent variable from models of eastern yellow robin presence. Panels show tests of whether the broader landscape context is better represented by (a) total woodland within a 5 km radius of each site (n = 60), or (b) accessible woodland within a 5 km radius of each site (n = 60). Black bars indicate the independently-explained variance was significant………………………………………………….50

Figure 4.1. Location of the study region in the Brigalow Belt South bioregion, Queensland, Australia. Black circles denote study sites………………………………………………………….60

xvi number of arthropods by the proportion of woodland within 500 m of each site, b) total number of arthropods by the number of woodland patches within 500 m of each site, and c) the number of arthropods excluding Formicidae by the soil moisture (%)……………………………….………..69

Figure 4.3. Results of negative binomial, quadratic, and loess models of the number of arthropods caught in relation to the proportion of area within a 500 m radius of each site (n = 25) that was covered by woodland. Only those taxa most frequently eaten by eastern yellow robins are reported. Hymenoptera graphs exclude Formicidae captures as they are reported separately……….……….75

Figure 5.1. Location of the study region in the Brigalow Belt South bioregion, Queensland, Australia. Black circles denote study sites………………………………………………...………..85

Figure 5.2. Scatterplots with 95% confidence intervals (grey shading) of the factors explaining the greatest variation in eastern yellow robin condition. Robust linear regressions of a) fourth-root transformed heterophil:lymphocyte ratios by the proportion of grass cover at each site, b) fourth- root transformed heterophil:lymphocyte ratios of eastern yellow robins by the number of pieces of fallen timber greater than 7 cm in diameter along 20 m transects, and c) body condition indices by ordinal date………………………………………………………………………………...………..93

Figure 6.1. Location of the study sites in the Brigalow Belt South bioregion, Queensland, Australia. Orange circles denote those sites where supplementary food was provided, blue circles denote those that were unfed control sites, and yellow circles indicate those sites where we were unable to recapture eastern yellow robins following the experimental treatment………………………..…102

Figure 6.2. Scatterplots with linear trend lines of the relationship between the proportion of woodland within 500 m of each site and the change in eastern yellow robin a) heterophil:lymphocyte ratio, b) leukocyte count, or c) body condition after an experimental treatment was applied. Robins either received supplementary food (hollow triangles) or acted as unfed controls (solid circles)………………………………………………………………………109

Figure D.1. (a) Receiver operating characteristic curve and (b) calibration plot for a model of eastern yellow robin occupancy that includes total accessible woodland within a 5 km radius of each site. Dotted diagonal lines represent (a) the accuracy of the model predicted by chance alone, and (b) expected result from perfect model calibration. Bars represent 95% confidence intervals for

xvii each bin. The numbers above each bar indicate the number of samples within each bin. The area under the curve (AUC) was 0.71………………………………………………………………..…204

List of tables

Table 3.1. Averaged coefficients, adjusted standard errors, confidence intervals, and summed

Akaike weights (Σωi) derived for each variable from models of eastern yellow robin extirpations……………….………………………………………………………………………....47

Table 3.2. Averaged coefficients, adjusted standard errors, confidence intervals, and summed

Akaike weights (Σωi) derived for each variable from models of eastern yellow robin occurrence...48

Table 4.1. A priori models developed to assess the habitat, landscape, and climatic factors associated with counts and dry-weights of arthropod prey targeted by eastern yellow robins. All models included northing as a fixed effect and log (trap effort) as an offset term………………….64

Table 4.2. Number of arthropods caught in pitfall traps in brigalow-dominated woodlands. Only those taxonomic groups likely to be eaten by eastern yellow robins were included in further analyses……………………………………………………………………………………..……….66

Table 4.3. Model selection results from a priori models examining the effects of environmental factors on the number of arthropods caught in pitfall traps at sites (n = 25) within brigalow- dominated woodlands. The table displays number of parameters (K), deviance (Dev), AICc 2 differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Deviance is defined as -2 × the maximised log-likelihood. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod abundance than the null model. Models with ωi < 0.01 are omitted from the table…………………………………………………………………..……….67

Table 4.4. Coefficients, standard errors, and 95% confidence intervals from the best approximating models identified through AICc model comparison techniques examining environmental variability in the number and weights of arthropods caught in pitfall traps (n = 25). Confidence intervals denoted in bold do not cross zero………………………………………………………...…………68

Table 4.5. Model selection results from a priori models examining the effects of environmental factors on dry weights of arthropods caught in pitfall traps at sites (n = 25) within brigalow-

xviii dominated woodlands. The table displays number of parameters (K), deviance (Dev), AICc 2 differences (ΔAICc), Akaike weights (ωi), and adjusted R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod dry weight than the null model. Models with ωi < 0.01 are omitted from the table……………………….…………70

Table 4.6. Model selection results from a priori models examining the effects of environmental factors on the number of arthropods (excluding Formicidae) caught in pitfall traps at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance 2 (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod abundance (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table………………………………………………………………………………………...……….71

Table 4.7. Model selection results from a priori models examining the effects of environmental factors on dry weights of arthropods (excluding Formicidae) caught in pitfall traps at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance 2 (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod dry weight (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table………………………………………………………………………………………...……….72

Table 4.8. Model selection results from a priori models examining the effects of environmental factors on the maximum number of arthropods (excluding Formicidae) caught in any one pitfall trap at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters 2 (K), deviance (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod abundance (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table……………………………………………………………………………....73

Table 4.9. Model selection results from a priori models examining the effects of environmental factors on the maximum dry weight of arthropods (excluding Formicidae) caught in any one pitfall trap at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R2 values. Models denoted in bold have substantial empirical support that they explain more of the

xix variation in arthropod dry weight (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table………………………..…………………………………………….74

Table 5.2. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin heterophil:lymphocyte ratios……….……92

Table 5.4. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin leukocyte counts………………..………..94

Table 5.5. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin body condition…………….……………..95

Table 6.2. 95% confidence intervals from Wilcoxon rank sum tests of changes to eastern yellow robin condition against experimental treatment (given supplementary food or unfed control)…...108

Table A.1. Details of the 150 articles included in the literature review (Chapter 2) by proximate process, geographic region, and taxon studied…………………………………………...………..164

Table B.1. Unpublished heterophil:lymphocyte ratios collected by Maron et al. (2012) from 51 eastern yellow robins near Meandarra and Moonie, Queensland, Australia………………………197

Table C.1. Results of model averaging testing the effectiveness of using H:L ratios to predict eastern yellow robin extirpations…………………………………………………………………..199

Table C.2. Results of model averaging testing the relationship between remnant woodland extent within 500 m of each study site and eastern yellow robin occurrence…………………….………201

Table C.3. Results of model averaging testing the relationship between the broader landscape context (total woodland extent within 5 km of each study site) and eastern yellow robin occurrence………………………………………………………………………………………….202

Table C.4. Results of model averaging testing the relationship between the broader landscape context (accessible woodland extent within 5 km of each study site) and eastern yellow robin occurrence……………………………………………………………………………….…………203

List of plates

Plate 1. An aerial view of the landscape near Moonie, Queensland. Most remaining woodland occurs in narrow strips alongside paddocks, crops, and roads. More-contiguous woodlands are crucial for facilitating dispersal and meeting minimum home range requirements………….………1

Plate 2. Farmers sometimes retain small islands of habitat within their properties, but if the gap crossing distance to the island is too great, few species will occupy the habitat patch……..………20

Plate 3. An example of brigalow woodland typically occupied by eastern yellow robins. The ground is pocketed with gilgai – shallow depressions in the cracking clay soil. Grass cover is sparse, but leaf litter, fallen timber, and wilga bushes are present………………………………...……………38

Plate 4. Moisture is of critical importance in this harsh environment. Gilgai depressions collect water after rain…………………………………………………………………………..…………..55

Plate 5. Sampling an eastern yellow robin. Metal leg bands are individually numbered to allow identification throughout the birds’ lifetime. A unique combination of colour bands are added to each leg to allow an observer to identify the bird from a distance. Image courtesy of Martine Maron……………………………………………………………………………………...………..81

xxi Plate 6. An eastern yellow robin visiting a supplementary feeding station stocked with mealworms. These robins usually catch prey from the ground, but became accustomed to using the raised trays………………………………………………………………………………………...……….99

Plate 7. An eastern yellow robin taking stock of the situation after being released back into his territory……………………………………………………………………………..……………..114

Plate 8. Sunset at Moonie, Queensland…………………………………………………………..133

xxii CHAPTER 1. Introduction

1 1.1 THEORY AND CONCEPTS UNDERLYING THE THESIS

1.1.1 The impacts of land-use change

Habitat loss, fragmentation, and degradation are driving biodiversity losses at an alarming rate (Cushman 2006; Vié et al. 2009; Hoffmann et al. 2010; Maxwell et al. 2016). Terrestrial vertebrate population abundances have fallen by 38% since 1970 and there is no suggestion that the rate of loss is slowing (WWF 2016). Forty per cent of all bird species are showing population declines (Vié et al. 2009). Habitat loss and degradation are placing stress on 93% of all threatened bird species with the bulk of the threat associated with land conversion for agriculture and timber extraction (Vié et al. 2009; Maxwell et al. 2016). These threats are likely to persist for the foreseeable future; although the rate of global deforestation has decreased since the 1990s, it still remains high (FAO 2010). Australia is currently ranked eighth in the world for deforestation, with over 5.8 million ha of forest lost from 2000-2012 (Hansen et al. 2013).

The effects of habitat destruction on populations can be considered to form a hierarchy (Figure 1.1). The removal of habitat can initiate four processes of landscape change: habitat loss; habitat fragmentation; habitat degradation; and alteration of the matrix (non-habitat land cover; Ford et al. 2001; Fischer & Lindenmayer 2007; Driscoll et al. 2013). Each of these processes alters structural elements within landscapes which may ultimately lead to population declines (Fahrig 2003; Ewers & Didham 2006; Driscoll et al. 2013). However, these processes of landscape change tell us nothing about the mechanisms operating directly on individuals. An organism does not die or fail to reproduce simply because the habitat patch it is in is too small and the next patch is a long distance away. Rather, patch size and isolation are associated with more-proximate factors that affect habitat suitability. For example, if the currently occupied patch has inadequate food resources, staying there may lead to starvation or an inability to successfully reproduce (Wingfield et al. 1998; Wiens et al. 2006; Boulton et al. 2008). However, travelling to a neighbouring patch to fulfil these needs may threaten survival, such as through an increased predation risk in the matrix (Van Vuren & Armitage 1994; Rodríguez et al. 2001; Bonte et al. 2012). By improving our understanding of how both the landscape elements and the proximate mechanisms contribute to declines, we can develop more robust and effective strategies for conserving species or restoring habitat.

A vast body of work has focused on identifying correlative relationships between habitat extent or configuration and the distribution or abundance of organisms (Fahrig 2003; Bennett et al. 2006; Ewers & Didham 2006; Mortelliti et al. 2010a). This bias towards focusing on patterns of

3 distribution or species richness when attempting to understand how fragmentation affects populations may have come about because of the relative ease with which surveys of fauna presence can be conducted. However, this focus on patterns provides us with a limited view of what is happening to individuals, populations, and communities in highly-modified landscapes (Johnson 2002; Fischer & Lindenmayer 2007). A more-balanced approach that takes into consideration the proximate causes of reduced population persistence is needed if we are to understand why declines occur (McGarigal & Cushman 2002; Watson 2011).

1.1.2 Australian woodland bird declines

One group of species that is sensitive to habitat loss and fragmentation is woodland- or forest- dependent birds (Bender et al. 1998; Villard et al. 1999; Mortelliti et al. 2010b; Laurance et al. 2011). In Australia, reports of woodland bird declines have become increasingly common over the past 20 years (Reid 1999; Mac Nally et al. 2009; Ford 2011; Rayner et al. 2014). While some studies provide evidence of historical declines (Saunders & Ingram 1995; Woinarski & Catterall 2004), others demonstrate ongoing losses (Mac Nally et al. 2009; Stevens & Watson 2013; Rayner et al. 2014). Taxa generally considered to be common and widespread are not immune to the problem, with declines observed for species such as grey shrike thrush (Colluricincla harmonica; Stevens & Watson 2013), grey fantail (Rhipidura albiscapa; Stevens & Watson 2013), striated thornbill (Acanthiza lineata; Szabo et al. 2011), and eastern yellow robin (Eopsaltria australis; Stevens & Watson 2013). The decline of common species is concerning as they could have disproportionately large potential flow-on effects to the structure and functioning of ecosystems (Gaston & Fuller 2008; Godet et al. 2015).

It has been challenging to identify the underlying drivers of population declines for woodland- dependent birds, partly because the observed changes have been inconsistent across species, habitat types, and regions (Barrett et al. 2003; Maron & Lill 2006; Watson 2011). Species most likely to be affected are small, sedentary, ground-foraging insectivores (Reid 1999; Watson 2011). Declines are most commonly reported from temperate eucalypt woodlands within agricultural regions of south eastern Australia and southern Western Australia (Saunders & Ingram 1995; Olsen et al. 2005; Watson 2011). Habitat loss, fragmentation, and degradation underlie most of the proposed mechanisms thought to be causing the population declines (Ford et al. 2001; Watson 2011; Maron et al. 2013); however, drought and climate change also seem to be contributing to the problem (Mac Nally et al. 2009; Stevens & Watson 2013; Rayner et al. 2014). Despite possible causal pathways being clearly articulated (Ford et al. 2001), relatively little progress has been made towards testing

4 the hypothesised mechanisms driving the declines (but see Cooper & Walters 2002; Stevens & Watson 2013).

1.1.3 Effects of limited resource availability in highly-modified landscapes

One promising hypothesised cause of avian declines is reduced resource availability in landscapes with a higher degree of structural alteration (Ford et al. 2001; Watson 2011). Here, I define a resource as any product which improves the growth or fitness of an individual as it becomes more abundant in the environment, and which, when used by an individual, becomes unavailable to others (Tilman 1982). There are several indications that suggest this mechanism may be important. Insectivores have been repeatedly identified as being particularly susceptible to declines in more- modified landscapes than other foraging guilds (Major et al. 2001; Watson 2011). This pattern is not limited to Australian woodlands, but is seen around the world (Gray et al. 2007; Bregman et al. 2014) including in the Americas (Kattan et al. 1994; Nebel et al. 2010; Laurance et al. 2011), Asia (Castelletta et al. 2000; Yong et al. 2011), and Africa (Sinclair et al. 2002). This association between population vulnerability and foraging guild suggests that prey access or availability is contributing to the problem (Razeng & Watson 2012; Paquette et al. 2014). Additionally, many of the woodland avifauna declines observed in Australia occur within agricultural landscapes (Saunders & Ingram 1995; Ford et al. 2001; Mac Nally et al. 2009) where the most productive woodlands are preferentially cleared to maximise agricultural output (Watson 2011). Thus, the remaining woodlands are more likely to be those that were originally on poor quality soils that are less productive and support fewer invertebrates (Watson 2011). Finally, altered resource access and availability can theoretically be brought about by multiple types of landscape change, including habitat loss, fragmentation, degradation, and changes to the matrix (Figure 1.1), so altered resource availability has the potential to be quite common.

There are many reasons why invertebrate prey could be less abundant or available to insectivores in agricultural landscapes. First, even if prey density was the same within small and large patches, random sampling effects would mean that small patches would contain fewer total prey than large patches (Connor & McCoy 1979), which could be important when patch size limits the home range area. Second, foraging across multiple habitat patches comes at a cost in terms of energetics (i.e. from travelling between patches) and any additional prey sourced would need to offset that cost (Hinsley 2000). Third, fragmented landscapes contain more edge habitat than contiguous landscapes (Forman & Godron 1986). Microclimate conditions differ between edge and interior habitat; edges are generally hotter, drier, windier, and lighter than core habitat (Matlack 1993; Chen

5 et al. 1999; Meyer et al. 2001). The desiccating conditions at edges may be unsuitable for some invertebrate taxa (Didham et al. 1998; Laurance et al. 2011). Fourth, edges are occupied by both woodland-dependent avifauna as well as habitat generalists that could increase the predation pressure on invertebrates (Barbaro et al. 2012). Fifth, agricultural practices undertaken in the matrix or within the patches themselves, such as pesticide use, increased run off, nutrient input, and grazing, can alter soil characteristics and invertebrate abundances (King & Hutchinson 1983; Yates et al. 2000; Watson 2011; Hallmann et al. 2014). Finally, invertebrate populations are themselves influenced by landscape structure, such as their ability to disperse to distant habitat patches (Laurance et al. 2011).

It is hypothesised that small habitat patches contain a lower prey density (i.e. prey abundance or weight per unit area) than large patches (Burke & Nol 1998; Zanette et al. 2000). However, this idea runs counter to many invertebrate studies that find mixed results of landscape structure on invertebrate abundance and richness with taxonomic groups showing idiosyncratic responses to habitat loss and fragmentation (Helle & Muona 1985; Van Wilgenburg et al. 2001; Laurance et al. 2011; Moreno et al. 2013). In reality, few studies have directly measured prey availability for avian insectivores and results have been mixed with prey showing either negative relationships (Burke & Nol 1998; Zanette et al. 2000; Luck 2003; Morrison et al. 2010) or no relationship (Şekercioğlu et al. 2002; Butcher et al. 2010) to measures of habitat loss or fragmentation. To compound the problem, some of these studies have limited replication at the level of habitat patch or landscape (Zanette et al. 2000; Şekercioğlu et al. 2002; Luck 2003).

Studies that investigate relationships between prey availability and landscape structure often assess invertebrate density based on all invertebrate taxonomic groups (Burke & Nol 1998; Zanette et al. 2000; Butcher et al. 2010; Morrison et al. 2010). However, drawing inferences from measures of total invertebrate density may be misleading if they do not accurately represent prey availability (Razeng & Watson 2012). Prey vary in nutritional value (Robel et al. 1995; Arnold et al. 2010; Razeng & Watson 2015) and this may influence birds to preferentially target certain taxa (Razeng & Watson 2015). Additionally, birds employ specific strategies, such as foraging in a certain strata, using particular capture methods, or feeding at a certain time of day (Higgins & Peter 2002; Antos & Bennett 2006; Moorman et al. 2007) and invertebrate sampling methods need to take these into consideration. Thus, it is unclear whether it is appropriate to treat all prey equally in food availability studies and greater understanding of the issue may be derived by taking some of these factors into consideration.

6 1.1.4 Using physiological measures to understand mechanisms causing population declines

Conservation physiology is a relatively young field of research that is quickly gaining traction with scientists (Lennox & Cooke 2014; Madliger et al. 2016). In essence, work within this field uses physiological responses to identify how environmental changes impact individuals, populations, and ecosystems (Wikelski & Cooke 2006; Cooke et al. 2013). This emerging discipline provides new opportunities for developing our knowledge about the processes driving deleterious effects on populations in highly-modified landscapes (Cooke et al. 2013). Although rarely stated, many studies of landscape change rely on the distribution or abundance of populations to have already declined (Ellis et al. 2012). In many population-level studies, individuals must have been lost from suboptimal localities for the researcher to detect the impact of landscape change (Maron et al. 2012). In contrast, physiological responses can show changes to individual condition before the effects are expressed at the population level (Janin et al. 2011; Ellis et al. 2012). By integrating physiological measures into studies of food availability and landscape structure, we have the opportunity to deepen our understanding of whether prey is limited in more-disturbed landscapes, how this variation affects individuals, and what flow-on effects impact the species at the population level.

If living in a more-modified landscape makes it harder to find food, is more taxing energetically, or involves traversing a dangerous or inhospitable matrix, it might be expected that individuals in these landscapes have higher physiological stress levels. Here, I define stress as the level of activation of the hypothalamus-pituitary-adrenal (HPA) axis system (Johnstone et al. 2012). The vertebrate HPA axis system is in a constant state of flux, but displays increased activation in response to a variety of stressors experienced by an individual (Costa & Sinervo 2004; Johnstone et al. 2012). Activation of this system results in an increased release of glucocorticoid hormones (glucocorticoid concentrations = GCs) and changes to the levels of differential white blood cells circulating in the blood (Costa & Sinervo 2004; Davis et al. 2008; Johnstone et al. 2012; Blas 2015). For example, in birds, higher heterophil:lymphocyte (H:L) ratios indicate higher long-term stress levels (Davis et al. 2008; Johnstone et al. 2012).

These types of changes can be used as metrics indicating the level of HPA activation at the time of sampling (Johnstone et al. 2012). GCs increase very rapidly in response to a stressor, in some cases within just 2-3 min (Romero & Reed 2005), and have a relatively short half-life of as little as 15 min (Sainio et al. 1988) suggesting they are suited to measures of short-term stressors. In contrast, changes to H:L ratios take approximately 30 min to several hours to take effect (Davis 2005; Davis

7 et al. 2008; Cīrule et al. 2012) and persist in the blood stream for longer (Davis et al. 2008). Thus H:L ratios may provide a more appropriate indicator of long-term stress than GCs (Goessling et al. 2015).

Stress measures show good potential for providing additional understanding on the mechanisms through which landscape change contributes to population declines (Ellis et al. 2012; Cooke et al. 2013). The effects of habitat loss and fragmentation on patch occupancy can be markedly different from the effects observed on physiological condition (Janin et al. 2011; Maron et al. 2012). For example, patch occupancy by common toads (Bufo bufo) in France was best predicted by habitat extent, whereas GCs and body condition were strongly associated with both habitat extent and fragmentation (Janin et al. 2011). Thus, the incorporation of physiological measures, such as chronic stress levels, into landscape ecology studies may deliver a more comprehensive picture of the study system and help provide new insights into the underlying mechanisms causing declines.

Food availability has a pronounced effect on stress levels within individuals (Maxwell et al. 1992; Clinchy et al. 2004; Hinam & St. Clair 2008; Barrett et al. 2015). Variation in food abundance was the best predictor of GCs in the common murre (Uria aalge; Kitaysky et al. 2007). Eurasian treecreeper (Certhia familiaris) chicks had higher GCs when raised in territories containing less invertebrate prey (Suorsa et al. 2003). H:L ratios were elevated in white leghorn chickens (Gallus gallus domesticus) while fasting (Davis et al. 2000). It is common for studies to report that individuals in poorer body condition had higher stress levels (Marra & Holberton 1998; Suorsa et al. 2004; Lobato et al. 2005); however, this alone does not provide strong support for the link between stress and food availability (Johnstone et al. 2012). Individuals that experience more stress may have compromised immune systems and be more susceptible to disease or inflammation (Johnstone et al. 2012), leading to poor body condition.

Increased stress levels, particularly those that are elevated for prolonged periods, can have indirect fitness costs (but see Bonier et al. 2004; Dantzer et al. 2014). For example, chronically increased H:L ratios in birds have been linked to a greater susceptibility to infection, slower growth rates, and reduced survival (Al-Murrani et al. 2002; Moreno et al. 2002; Kilgas et al. 2006). However, the effect of an ongoing, stress-inducing event, such as a chronic reduction in food availability, is likely to be apparent at a physiological level before demographic effects are realised (Ellis et al. 2012). Thus, patterns of stress levels could provide an early warning sign of a change to the population (Wikelski & Cooke 2006; Ellis et al. 2012).

8 1.2 THESIS STRUCTURE AND FORMAT

This thesis is built around five core chapters (Chapters 2, 3, 4, 5, and 6), each of which addresses one of the key research questions listed in the following section, followed by a synthesis chapter that draws the work together (Figure 1.2). The core chapters are written as stand-alone manuscripts, which are either in review or ready to be submitted to peer-reviewed journals. Thus, there is some overlap in the information presented in the introduction and method sections of several of the chapters. While the language within the core chapters refers to the manuscript co-authors as a collective, I completed the bulk of the work and outline the respective author contributions on the title page of each core chapter.

Each chapter has been written in accordance to the style guide relevant for each journal. However, I have made some minor modifications to improve the flow and consistency within the thesis. These changes include: spelling; capitalisation; punctuation; formatting; numbering of tables, figures, and appendices; heading style; and citations to other chapters within this thesis. Appendices and references for all chapters have been collated at the end of this thesis. Details of approvals for animal ethics and scientific permits have been removed from each chapter and collated below.

All research within this thesis was conducted under Animal Ethics Approval Certificates GPEM/540/12, GPEM/011/13, and GPEM/447/13 from the University of Queensland’s Native and Exotic Wildlife and Marine Animals Ethics Committee and Scientific Purposes Permits WISP12520913, TWB/05/2013, WITK12520813, WISP13082913, TWB/30/2013, WITK13082213, WISP13973214, and WITK13973314 from Queensland’s Department of Environment and Heritage Protection.

1.3 THESIS OBJECTIVES AND RESEARCH QUESTIONS

Using the eastern yellow robin (hereinafter “robin”) in the Brigalow Belt South bioregion of southern Queensland, Australia as a case study species, my primary objective was to improve our understanding of the mechanisms placing woodland-dependent avian insectivore populations at risk of declines in highly-modified landscapes and, specifically, whether prey density plays a role. A secondary objective was to further understanding of how measures of individual condition could predict future changes to population dynamics that would not be evident from a single snapshot of species occurrence. My final objective was to provide a way of conceptualising the impacts of habitat fragmentation that researchers could use to formulate their own research questions and

9 managers could use to improve conservation outcomes. I approached these objectives by focusing on the junctions between landscape structure, food abundance, and individual condition (Figure 1.3).

Figure 1.2. An overview of the thesis structure. Grey boxes denote chapters written as manuscripts for submission to peer-reviewed journals.

Figure 1.3. In this thesis, I integrate information about food availability, landscape structure, and individual condition. The junctions addressed by each chapter of the thesis are indicated.

The five key research questions addressed in this thesis are:

1. How do movement restrictions in fragmented landscapes negatively impact faunal populations, and are there biases in research effort allocated to each of the key mechanisms?

2. How does the association between site occupancy and landscape structure change over time, and how effective are individual condition measures at signalling future changes to population dynamics?

3. How is prey density related to woodland extent and how does this differ when nutritional value is taken into consideration?

4. How well do measures of prey density explain the variation in the physiological condition of robins, and how have patterns of condition changed over time?

5. How does experimentally increasing food availability for an avian insectivore impact individual condition in landscapes with varying degrees of woodland cover?

11 1.4 THESIS OUTLINE

Chapter 2: Habitat fragmentation and movement: linking scale and mechanism. A wide range of mechanisms are potentially impacted by habitat loss, fragmentation, and degradation (Ford et al. 2001). I develop a framework for conceptualising the mechanisms by which populations are affected when habitat fragmentation restricts faunal movements and how these relate to the spatiotemporal scale of fragmentation. I conduct a global systematic review of the literature to determine the relative research effort invested in understanding each mechanism and whether that effort varied across biogeographic zones or taxa. The research gaps identified by the review helped form the basis of the next four chapters.

Chapter 3: Using individual condition measures to predict the long-term importance of habitat extent for population persistence. This chapter is the first of four chapters based on a case study of the eastern yellow robin in the Brigalow Belt South bioregion of southern Queensland that has undergone heavy deforestation. A pilot study conducted in the region four years earlier found that robin chronic stress levels were elevated at sites that were surrounded by less woodland cover, but that site occupancy itself was unrelated to woodland cover. In this chapter, I test a prediction arising from the pilot study that sites previously containing robins with higher chronic stress levels would now show higher rates of extirpations. In addition, I test whether the previously-identified relationship between stress levels and woodland extent could predict current patterns of site occupancy, i.e. whether occupancy would now be positively related to woodland cover (within 500 m). I also refine the occupancy model developed in the pilot study to test both the effects of habitat extent and habitat connectivity at a broader landscape scale (within 5 km).

Chapter 4: Determinants of apparent food availability for a ground-foraging insectivore differ when the nutritional value of prey is considered. Chapter 3 assessed trends at a population level, whereas Chapter 4 delves into a mechanism operating at the individual level and starts to explore why robins were less likely to now occupy less-wooded landscapes. I identify habitat, landscape, and climatic factors related to prey availability and, in particular, whether prey density is lower at sites with less surrounding woodland. Instead of basing this investigation on total invertebrate density as is often done (Burke & Nol 1998; Zanette et al. 2000), I only consider prey taxa known to be consumed by robins. Drawing inferences based on total invertebrate density may be misleading given that birds show preferences towards certain prey groups (Naef-Daenzer et al. 2000; Hagar et al. 2007; Moorman et al. 2007; Razeng & Watson 2012). I also investigate how the factors associated with

12 prey density change when the least nutritious, but numerically dominant, prey group (Formicidae, i.e. ants) is excluded.

Chapter 5: Individual condition in an Australian avian insectivore is related to foraging substrate characteristics. While Chapter 4 explored the factors affecting the prey availability for robins in structurally-altered landscapes, Chapter 5 complements that work by assessing habitat, landscape, and climatic factors related to robin body condition. Importantly, I investigate how prey densities measured in Chapter 4 are associated with robin condition. I take the nutritional value of prey into consideration by including measures of prey density including and excluding Formicidae into my model sets. I also consider how the factors related to condition have changed since the previous study by Maron et al. (2012) and what these changes might mean for the population.

Chapter 6: Food restriction for an avian insectivore in a fragmented landscape: a supplementary feeding experiment. This chapter supplements the work conducted in Chapters 4 and 5 by experimentally establishing whether low prey availability is causing poor robin condition in less- wooded landscapes. I attempt to determine whether robins in landscapes with less woodland showed greater improvements in condition than those inhabiting more-wooded landscapes when they are given supplementary food. Such large-scale field-based experiments can be challenging, in part because they must be replicated at a spatial scale relevant to the experimental unit (Hurlbert 1984); in this case, landscapes that encompass a robin’s home range. However, such experiments are important because they allow stronger inferences about cause and effect to be drawn than simple observational studies (Eberhardt & Thomas 1991; Ewers et al. 2011). I discuss improvements that could be made to the sampling protocol that would enhance study outcomes in the future.

Chapter 7: Synthesis and conclusion. In the final chapter, I summarise the findings relating to each research question and discuss the contributions the thesis makes to ecological and physiological scientific knowledge. Finally, I review the limitations of my work and suggest avenues for future research.

1.5 STUDY REGION – BRIGALOW BELT SOUTH BIOREGION

1.5.1 Ecology of the Brigalow Belt South

I conducted the study in the Brigalow Belt South bioregion; one of 89 bioregions across Australia that has its own distinct climatic conditions, geology, landforms, vegetation communities, and fauna

13 (DSEWPC 2013). This bioregion encompasses approximately 27.2 million ha from just north of Rockhampton on the mid-Queensland coast down to Dubbo in central New South Wales (DSEWPC 2013). The study region covers up to 15,000 km2 of this bioregion in southern Queensland around Meandarra (27°19´S, 149°52É), Moonie (27°43´S, 150°21É), and Goondiwindi (28°33´S, 150°18É; Figure 1.4).

The study region has a subtropical climate and averages about 600 mm of rainfall per year with wet summers and dry winters (Bureau of Meteorology 2015). Average annual rainfall increases as one moves south with Meandarra receiving 572 mm per annum, Meandarra 594 mm per annum, and Goondiwindi 619 mm per annum (Bureau of Meteorology 2015). There was similar rainfall in the 12 months prior to a pilot study conducted in August 2009 (Maron et al. 2012) that is referred to throughout this thesis and the commencement of field work for the current study in March 2013 (Bureau of Meteorology 2015).

The landscape consists of gently undulating plains formed on fertile cracking clay soils (Sattler & Williams 1999). These soils shrink and expand in response to drying and moisture to eventually form gilgai, a microrelief pattern of depressions approximately 1 m deep and several metres across (colloquially named “melonholes”) that form ephemeral pools after rain (Skerman 1953; McKenzie et al. 2004). Historically, the low plains were dominated by open brigalow (Acacia harpophylla) woodlands while the less-fertile rises were dominated by Casuarina and Eucalyptus open forests, especially belah (Casuarina cristata), poplar box (Eucalyptus populnea), molly box (Eucalyptus pilligaensis), ironbarks (Eucalyptus spp.), and cypress pine (Callitris glaucophylla) (Sattler & Williams 1999).

1.5.2 Historical land-use changes

European settlement began around the 1840s with low intensity wool production undertaken on the grasslands and grassy woodlands of the region as these were thought to have the highest grazing potential (Nix 1994). Over the next 40 years, minor clearing occurred in the eucalypt woodlands to source timber for housing and infrastructure (Seabrook et al. 2006). By the 1940s, agriculture had extended to cattle and cropping and it had become increasingly common to clear open eucalypt woodlands to support these endeavours (Seabrook et al. 2006). However, settlers had little success in clearing brigalow woodlands, even though they occurred on the most productive soils (Seabrook et al. 2006). Damaged brigalow suckers providing the root stock remains intact and, thus, re-sprouts to produce stands of dense regrowth (Skerman 1953; Johnson 1964).

15 A combination of heavy machinery and government policy ultimately drove the demise of brigalow woodlands. In the late 1940s, landholders started using chains strung between heavy machinery to uproot brigalow trees (Skerman 1953; Johnson 1964; Fensham & Fairfax 2003). Blade-ploughs were introduced and proved to be effective at controlling regrowth by severing the root stock (McDonald 1988; Fensham & Fairfax 2003; Seabrook et al. 2006). In 1962, the Queensland government introduced The Brigalow and Other Lands Development Act and provided incentives to encourage the clearing of brigalow to increase agricultural production (Fensham & Fairfax 2003; Seabrook et al. 2006). Although at least 10% of the brigalow woodlands on properties should have been retained as shade lines along fences, these were often lost in the intense fires used to burn cleared vegetation (McDonald 1988; Seabrook et al. 2006).

By the late 1990s, the intense and protracted clearing practices had taken their toll on the brigalow communities. The Queensland section of the Brigalow Belt bioregion was originally covered by approximately 6 million ha of brigalow woodland (Seabrook et al. 2006). While approximately 77% of the remnant brigalow woodland remained in 1940, this had fallen to about 63% by 1960, around 27% by 1980, and only 16% by 2000; far below the 30% threshold recommended for maintaining biodiversity and healthy ecosystems (McAlpine et al. 2002; Seabrook et al. 2006). As a result, the Queensland government introduced legislation aimed at halting broad-scale clearing of remnant vegetation and retaining areas of high value regrowth, although these changes have since been watered down (Evans 2016). Today, less than 10% of the original brigalow woodlands remain (DSEWPC 2012) and brigalow-dominated ecological communities are nationally listed as endangered (Commonwealth of Australia 2016).

1.6 STUDY SPECIES – EASTERN YELLOW ROBIN

Eastern yellow robins are small (approximately 20 g in body mass), endemic passerines found along the east coast of Australia (Higgins & Peter 2002). They have a patchy distribution north of Rockhampton, but are common and widespread south of that point along Queensland, New South Wales, ACT, and Victoria (Barrett et al. 2003). These woodland-dependent birds occupy a diverse range of forests and woodlands that generally have a tall shrub layer but sparse ground vegetation (Higgins & Peter 2002). Although they are often spotted on the edges of woodland, they tend not to venture into open habitat.

This sedentary species maintains a year-round home range of 10-12 ha in size in the eucalypt woodlands of New South Wales, although ranges compressed to 5-6 ha when neighbouring

16 territories were present (Higgins & Peter 2002; Debus 2006) and have been recorded as small as 0.8-2.0 ha (Marchant 1987). Initial estimates indicated that robins in the brigalow woodlands in the study region occupied home ranges of about 5-6 ha (Anita Cosgrove, unpublished data). Each territory is occupied by a pair or small family group and is actively defended against conspecific intrusion during the breeding season (Marchant 1987; Higgins & Peter 2002). Breeding usually occurs from July to January each year, but tends to taper off around October (Marchant 1984; Higgins & Peter 2002).

Eastern yellow robins are ground-foraging insectivores (Higgins & Peter 2002). They primarily eat invertebrates, but have also been recorded eating seeds, fungi, skinks, and grit (Higgins & Peter 2002). Arthropod prey taxa include Coleoptera, Formicidae, Diptera, Lepidoptera, Hemiptera, other Hymenoptera, Araneae, Orthoptera, Chilopoda, Diplopoda, and Phasmatodea (Higgins & Peter 2002; Razeng & Watson 2012). Robins adopt a sit-and-wait foraging strategy, perching on a tree trunk or low branch before pouncing on ground-dwelling prey (Ford et al. 1986; Higgins & Peter 2002). This species takes approximately 73-77% of food from the ground, 3-5% from foliage, 9- 11% from bark or tree trunks, and 5-7% from the air (Recher et al. 1985; Ford et al. 1986). They preferentially feed at sites that contain more leaf litter cover, more fallen logs, more canopy cover, and less ground vegetation, presumably because most of these attributes are associated with higher invertebrate abundances or detectability (Cousin 2007).

This species of robin has a Red List status of Least Concern (BirdLife International 2012). Population trends vary geographically and this variation is most apparent in studies of Atlas data (a citizen science program whereby the public can record surveys of avian abundance and occurrence). Trends observed between the two Atlas periods (1977-1981 and 1998-2002) generally showed increased incidence in southern Queensland and northern New South Wales, and decreases or no change in southern New South Wales and Victoria (Barrett et al. 2003). These general trends are similarly reflected in other studies (BirdLife Australia 2015a, 2015b), although historical declines have been recorded in the Queensland Brigalow Belt (Woinarski & Catterall 2004) and some declines have been noted in northern New South Wales (Reid 1999; Debus 2006; Stevens & Watson 2013).

Eastern yellow robins are sensitive to both habitat patch area and isolation in agricultural landscapes (Barrett et al. 1994; Watson et al. 2001; Watson et al. 2005). This pattern is typically attributed to more-modified landscapes having lower food availability or higher nest predation (Zanette 2000; Zanette et al. 2000). In northern New South Wales, robins tend to occupy woodland remnants with

17 lower perimeter to area ratios, i.e. patches with less edge habitat (Cousin 2007). In the study region, robin site occupancy was unrelated to landscape attributes, such as woodland extent and number of patches; however, indicators of chronic stress levels were elevated in less-wooded landscapes suggesting that these landscapes were suboptimal in some way (Maron et al. 2012).

1.7 SUITABILITY OF STUDY REGION AND SPECIES

The study region and species were chosen so that preliminary work conducted in the area by Maron et al. (2012) could be extended. However, there were several reasons why this region and species were well suited to the current study. First, little explanation has been put forth as to why population trends worsen towards the south of the species’ range. Watson (2011) highlighted the fact that most woodland bird declines seem to occur in temperate eucalypt woodlands, especially those with predominantly winter rainfall. He suggested that the avian declines were due to reduced invertebrate availability, both as a result of agricultural practices and the fact that these woodlands were generally retained on less-productive soils (Watson 2011). However, it is not insignificant that southern Queensland has a more-recent clearing history than New South Wales and Victoria due in part to the indestructible nature of brigalow during early settlement (ECC 1997; Seabrook et al. 2006). It is feasible that faunal relaxation is more advanced in the more-southern states and this is why reports of population declines become more evident towards the south. Additionally, the study region’s landscapes have relatively high structural connectivity due to a complex network of shade lines and travelling stock routes (interconnected grazing routes and reserves used by the pastoral industry to move stock on foot from one locality to another; Spooner et al. 2010; State of Queensland 2016a). Well-connected landscapes are thought to have longer relaxation times than those that are severely fragmented (Kuussaari et al. 2009). It is important to note that any effect of differential faunal relaxation would be confounded with soil productivity given that brigalow grows on a highly fertile clay base (Sattler & Williams 1999). However, focusing the study in Queensland provides a good contrast to the southern studies.

Second, the study region has been subject to heavy clearing; less than 15% of the study region is covered by remnant woodland, while less than 4% is remnant brigalow woodland (Anita Cosgrove, unpublished data). Several studies have found threshold effects, whereby habitat extent exerts a disproportionately large influence on the response variable of interest once extent falls below approximately 10-30% (Andrén 1994; Radford et al. 2005). Habitat extent within the study region falls below this theoretical threshold; thus, landscape effects should be emphasised in this region.

18 Third, Maron et al. (2012) identified that robins had higher chronic stress levels in landscapes with less woodland cover, but occupancy patterns showed no relationship to woodland extent. They hypothesised that this pattern of condition was an early indicator that extirpations would soon become evident in less-wooded landscapes (Maron et al. 2012). Thus, this system provided an excellent opportunity to test how effectively condition could predict upcoming changes to population dynamics. Additionally, the discrepancy between patterns of condition and occupancy in this system provides ideal conditions for using a conservation physiology approach to delve into the issue of reduced prey availability. This approach may not be possible in an area where population declines had advanced to the stage where the least-optimal sites were unoccupied.

Fourth, eastern yellow robins have many traits that could make them ideal for studying the potential effects of land-use change and prey reductions. They are sedentary rather than nomadic or migratory (Higgins & Peter 2002), making them highly dependent on local conditions. They are woodland-dependent and avoid open habitats (Higgins & Peter 2002). Landscape effects are more likely to be detected in a species that perceives the matrix as inhospitable rather than one that experiences the matrix along a gradient of suitability (Fischer et al. 2004; Watson et al. 2005; Ewers & Didham 2006). Similarly, this species’ strong sensitivity to landscape structure (Watson et al. 2005) should make it easier to detect landscape effects. Their foraging behaviours strongly target ground-dwelling prey (Higgins & Peter 2002) which reduces the need to adopt multiple invertebrate sampling techniques. Additionally, ground-dwelling prey are likely to be more strongly influenced by agricultural practices like grazing and run off (Barrett et al. 2007; Watson 2011) than, for example, canopy prey. Finally, they are still common and widespread, making them relatively easy to sample.

19 CHAPTER 2. Habitat fragmentation and movement: linking scale and mechanism

This chapter has been submitted to Diversity and Distributions (in review). Approximate percentage author contributions as follows: Anita Cosgrove – 77% Todd McWhorter – 4% Martine Maron – 19%

Plate 2. Farmers sometimes retain small islands of habitat within their properties, but if the gap crossing distance to the island is too great, few species will occupy the habitat patch.

20 2.1 ABSTRACT

The persistence of animal populations depends on individuals moving successfully around a landscape, but habitat fragmentation can hinder this by reducing functional connectivity. The proximate cause of population declines following habitat fragmentation is dependent on the spatial and temporal scale of movement restrictions. We present a conceptual framework highlighting the relationship between spatial and temporal scales, and three mechanisms through which detrimental impacts can occur when movement is disrupted in fragmented landscapes: limited resource access, restricted demographic exchange, and impeded gene flow. Each tends to be characterised by movement restriction at progressively larger spatial and temporal scales. We then review the literature to identify the proportion of studies conducted on each mechanism and whether biases existed in how often each was studied among different geographic zones or taxa. A random selection of 150 articles was classified by the mechanism, geographic region, and taxon studied in each article. We found that the overwhelming majority (74%) of articles investigated impeded gene flow, and only 15% and 11% explored restricted demographic exchange and limited resource access, respectively. Work on limited resource access was disproportionately low for particular taxonomic groups, such as reptiles and amphibians. Distinguishing which mechanisms are disrupted in a particular system is crucial because addressing each is likely to require a distinct conservation management response. We encourage greater focus on the more poorly-understood mechanisms of restricted demographic exchange and limited resource access, particularly in situations where they are likely to play an important role.

2.2 INTRODUCTION

The ability to move around landscapes is critical for the persistence of animal populations (Fahrig & Merriam 1985; Hanski 1991). Animals move for many reasons: to locate resources (Hinsley 2000), evade threats (Slagsvold et al. 2014), find mates (Dale 2001), disperse from natal areas (Szulkin & Sheldon 2008), avoid competition (Johnson & Gaines 1990), and cope with environmental stochasticity (Johnson & Gaines 1990). Movement is also integral to population- level processes, such as immigration, emigration, and gene flow (Bowne & Bowers 2004; Cushman 2006). Thus, barriers to movement have the potential to impede or prevent a wide range of mechanisms critical for the reproductive success or survival of individuals and the health and persistence of populations.

21 Habitat fragmentation reduces the structural connectivity of landscapes which can in turn reduce functional connectivity (Brooks 2003; Pe'er et al. 2011). Structural connectivity reflects the degree to which the physical structure of landscape elements allows individuals to move, while functional connectivity reflects the degree to which movements actually occur based on an interaction between landscape structure and the behavioural and biological characteristics of a species, such as movement ability, motivation, and decision-making processes (Wiens et al. 1997; Pe'er et al. 2011; Betts et al. 2015). Landscapes may provide variable degrees of connectivity for species depending on their affinity for habitat edges or their propensity to cross gaps (Pe'er et al. 2011).

Substantial work has been devoted to understanding how landscape fragmentation influences individuals and populations (Tischendorf & Fahrig 2000; Prugh et al. 2008; Fahrig 2017). While much of this work has focused on correlations between habitat configuration and the richness, distribution, or abundance of organisms (McGarigal & Cushman 2002; Prugh et al. 2008; Edge et al. 2017), an important subset has focused on the ecological mechanisms through which detrimental impacts can occur in fragmented landscapes (Banks et al. 2007). It is important to identify which specific mechanisms are impacted and why, so that targeted interventions can be developed. For instance, impeded gene flow in eastern collared lizards (Crotaphytus collaris collaris) inhabiting isolated rocky outcrops was initially addressed by translocating individuals (Templeton et al. 2001), but the underlying demographic issues were only rectified when fire suppression activities were modified to facilitate dispersal between outcrops (Brisson et al. 2003; Templeton et al. 2007).

In this paper, we classify the ecological mechanisms affected when habitat fragmentation restricts animal movements. First, we present a conceptual framework outlining the relationship between three core movement-related mechanisms (limited resource access, restricted demographic exchange, and impeded gene flow) and scale, both spatial and temporal. Second, we review the landscape ecology literature to explore the relative frequency of studies conducted on each of the three mechanisms, and to explore whether biases exist in the frequency that mechanisms were studied among geographic location or taxa. Finally, we outline the importance of distinguishing among these three types of mechanisms when identifying negative impacts of habitat fragmentation on movement, so that conservation management can be tailored appropriately.

22 2.3 CONCEPTUAL FRAMEWORK

2.3.1 Mechanism and scale are linked

Animal movements can be restricted at a range of spatial and temporal scales. Some movements are undertaken on a daily basis, some once-in-a-lifetime, and some only occur once in several generations. For example, unwillingness or inability to routinely cross roads may restrict an animal’s home range and influence their daily access to resources (Poessel et al. 2014), but may not impede once-in-a-lifetime dispersal events. Both the temporal and spatial scales at which movement restrictions operate relate to the mechanism affected, and ignoring this relationship could be costly in terms of conservation outcomes.

The consequences of small-scale movement restrictions are likely to occur over short time scales, whereas the consequences of large-scale restrictions may only be evident after many generations. For example, a gap between parts of a home range could quickly lead to poor condition of individuals if they continually expend extra energy navigating around it (Hinsley 2000; Smith et al. 2013) or death if gap crossings are undertaken in the presence of predators (Eccard et al., 2008). Population sizes can respond quickly to the health and survival of individuals (Desy & Batzli 1989). In contrast, inbreeding depression is likely to be associated with long-term movement impediments, generally requires quite severe levels of habitat isolation, and develops over multiple generations (Kenney et al. 2014).

2.3.2 Movement restrictions disrupt three key mechanisms

We consider three mechanisms through which impeded movements can have a direct, detrimental effect on animals: 1) limited resource access, 2) restricted demographic exchange, and 3) impeded gene flow (Figure 2.1). Habitat fragmentation can spatially segregate resources needed by an individual, reducing their ability to access the resources. Demographic exchange is restricted when individuals are not able to disperse to new localities or alternative populations to facilitate demographic processes and maintain sustainable demographic parameters within the metapopulation. This category includes aspects such as finding mates, balancing sex ratios, recolonising sites, and rescuing small, unstable populations. Finally, impeded gene flow occurs when the movement and integration of genetic material from one spatially discrete population to another is restricted.

For any given species, each of these mechanisms operates over characteristic spatial and temporal scales. However, the mechanisms are not mutually exclusive and can overlap in time and space. For example, if dispersal between two patches is prevented, an individual may also have trouble accessing sufficient resources if the occupied patch is depauperate. All three mechanisms do not need to be impeded for a species to go extinct (Harrisson et al. 2013). Individuals may be able to disperse successfully from a source population, but if they continually travel to a sink site harbouring insufficient resources, extinction may result (Pulliam 1988; Dunning et al. 1992). While the absolute spatiotemporal scales most relevant to each mechanism will vary among taxa depending on their dispersal abilities, foraging range requirements, and generation times, the relative scales applicable to each mechanism remain constant (Wiens 1989; Jackson & Fahrig 2012; Thornton & Fletcher Jr. 2014). Below, we consider each mechanism in turn.

2.3.2.1 Fine scale mechanism – limited resource access

Individuals must access a variety of resources on a regular basis. We define a resource as any product which improves the growth or fitness of an individual as it becomes more abundant in the environment, and which, when used by an individual, becomes unavailable to others (Tilman 1982). Food, water, and shelter from predators generally need to be sourced on a daily basis, while micronutrients and protected sites for reproduction are needed less frequently. If a single patch

24 cannot adequately meet the needs of an individual, that individual may utilise multiple patches (Dunning et al. 1992; Hinsley 2000). However, accessing spatially separated resources can carry a higher mortality risk, such as through predation or vehicle collisions, than accessing resources within continuous habitat (Lima & Dill 1990; Rodríguez et al. 2001; Cullen Jr. et al. 2016) and incurs increased energetic cost (Hinsley 2000), especially when individuals deliberately travel longer routes to avoid gaps (Desrochers & Hannon 1997; Smith et al. 2013).

Food restriction can have an almost immediate effect on the physiological condition of individuals through reduced body condition or increased stress levels (Kitaysky et al. 2007; MacCracken & Stebbings 2012). If insufficient food is sourced, reproductive output may be reduced (Hinam & St. Clair 2008) or the breeding process can be interrupted for the duration of the nutritional restriction (Wingfield et al. 1998).

2.3.2.2 Intermediate scale mechanism – restricted demographic exchange

Dispersal is a type of movement that tends to occur only once in an individual’s lifetime (e.g. dispersal from a natal territory) or regularly but infrequently (e.g. dispersal among breeding sites) (Greenwood & Harvey 1982). Habitat fragmentation can disrupt these movements (Cooper & Walters 2002; Coulon et al. 2010; Kang et al. 2016). Such disruption is likely to be triggered by isolation distances greater than those that restrict efficient daily movements for accessing resources.

While resource access impacts upon individual condition and fitness, dispersal primarily influences population dynamics. When individuals are unable to disperse freely this can prevent demographic rescue of populations by immigrants (Brown & Kodric-Brown 1977). Small populations are particularly susceptible to environmental and demographic stochasticity. Even quite low rates of immigration can buffer against these chance events, thereby generating rescue effects (Stacey & Taper 1992; Lecomte et al. 2004). Successful dispersal is also necessary to allow unoccupied patches to be recolonised (Holland & Bennett 2011), reduce kin competition (Hamilton & May 1977), maintain viable sex ratios (Dale 2001), and regulate population densities when intraspecific competition is high (Bowler & Benton 2005; Strevens & Bonsall 2011). There is a time lag between restriction of dispersal and observed demographic impacts. This delay is caused by several factors: individuals disperse infrequently, but multiple dispersals are needed before population impacts are noticeable; immigrants may not immediately participate in reproduction; or individuals may be reluctant to emigrate which may temporarily compensate for the lack of immigration.

25 Dispersal restrictions imposed by insufficient habitat connectivity can have dramatic impacts on populations. Experimentally manipulating the connectivity of common lizard (Lacerta vivipara) populations demonstrated that variance in population size was greater in treatments with unconnected patches; populations either went extinct quickly, or had a population boom followed by a sharp decline (Lecomte et al. 2004). Populations in connected treatments were more stable as dispersal provided a buffer against stochastic events (Lecomte et al. 2004). Thus, unimpeded dispersal movements are vital for long-term population persistence.

2.3.2.3 Broad scale mechanism – impeded gene flow

Effective genetic connectivity among populations may require less than 1% of the population to move within a generation (Mills & Allendorf 1996; Drees et al. 2011). The introduction of even a small number of genetically distinct individuals can facilitate genetic rescue (Amos & Balmford 2001; Vilà et al. 2003). For instance, a highly isolated population of adders (Vipera berus) contained only a few adult males (Madsen et al. 1999). Within seven years of temporarily translocating unrelated males into the population, there were eight times more males, recruitment had grown dramatically, and genetic variability had rapidly increased (Madsen et al. 1999). Severe reductions in habitat connectivity, however, can prevent even these movements (Delaney et al. 2010; Marsden et al. 2012; Roberts et al. 2013).

Small and isolated populations suffer from genetic drift (Holsinger & Weir 2009), which may lead to either a loss of genetic diversity or an increased frequency of harmful alleles (Amos & Balmford 2001; Frankham 2005). Reduced genetic diversity within populations can have deleterious effects such as lower reproductive success, inability to adapt to changing environments, and less resistance to disease (O'Brien et al. 1985; Reed & Frankham 2003; Aguilar et al. 2008). When individuals within a population are highly related, the risk of inbreeding depression increases (Reed & Frankham 2003). Inbreeding carries negative fitness consequences (Amos & Balmford 2001; Charlesworth & Willis 2009; Hedrick & Garcia-Dorado 2016), such as decreased survival, longevity, immune response, and hatching rates (Jiménez et al. 1994; Saccheri et al. 1998; Reid et al. 2003).

Habitat fragmentation impedes gene flow when movements are prevented for multiple generations (Amos & Balmford 2001; Storfer et al. 2007). For instance, sealed roads have increasingly segmented the distributions of timber rattlesnakes (Crotalus horridus) over the past seven to eight

26 generations (Clark et al. 2010), reducing genetic diversity and increasing population differentiation, with major roads having a greater impact than minor roads (Clark et al. 2010).

2.4 LITERATURE REVIEW METHODS

Although limited resource access, restricted demographic exchange, and impeded gene flow can all lead to local extinctions, the distribution of research effort to understand when and where each occurs is apparently not equal. Conservation solutions are developed based on knowledge of impairments within systems. A bias in knowledge development could potentially lead to important management strategies being overlooked. Biases in research effort already exist amongst taxa and geographic regions (Griffiths & Dos Santos 2012). Given that movement strategies are so variable across taxa and regions (Chesser 1998; Holyoak et al. 2008), the combination of these biases may actually hinder our ability to generalise findings. We undertook a systematic review of the existing habitat fragmentation literature to determine which of the three mechanisms were most studied, and whether particular mechanisms were more likely to be examined in particular geographic regions or for certain taxa.

2.4.1 Literature search strategy and inclusion criteria

We searched the ISI Web of Science platform on 13 March 2015 using terms that are associated with movement, resources, demography, or gene flow within fragmented landscapes. The search string was designed to deliver a broad and comprehensive list of articles particularly focused on the effects of habitat fragmentation on the mechanisms of interest. Specifically, a search of titles, abstracts, and keywords was conducted using the following terms: topic = (“habitat fragmentation” OR “habitat configuration”) AND ((resource* OR food OR forag* OR feed OR “home range”) OR (dispersal OR migration OR immigration OR emigration OR demograph*) OR (“gene flow” OR genetic*) OR (movement*)). The search was limited to articles published from 1994 onwards. The search was then refined to exclude all document types except “article”.

The titles and abstracts of articles were examined to determine whether each warranted inclusion in the literature review. Four criteria were used for inclusion. First, the article must have reported findings from an empirical study in a peer reviewed journal. Work of a purely theoretical nature, reviews, and meta-analyses were excluded; computer simulations were included provided they employed or were tested against an empirical dataset. Second, the focal taxon must have been an animal. Studies on plants were excluded because resource access does not apply to plant

27 movements. Third, the article must have investigated some aspect of habitat fragmentation per se, such as distance to neighbouring patch. Articles that only assessed the effects of habitat extent or patch size were excluded as these are more indicative of habitat loss. Finally, articles were only included if they explored a mechanism linking habitat fragmentation to a consequence for the focal taxon. Specifically, they must have investigated whether limited resource accessibility, restricted demographic exchange, or impeded gene flow (as per our definitions above) led to the observed fragmentation effects. Thus, studies that only identified relationships between species abundance or distribution and patch or landscape metrics were excluded from the review. Demographic consequences considered included effects on sex ratios, age structures, immigration, emigration, or colonisation. Whilst all the impeded gene flow studies we reviewed used genetic techniques, our focus was on the mechanism being investigated, not the approach adopted. Thus, it was quite possible for studies using genetic methods to be categorised under restricted demographic exchange.

Due to the large number of results from the initial keyword search (n = 4017), not all articles were subject to this filtering process. Rather, articles were assessed in a random order until we had a sample of 250 articles that met our inclusion criteria. The proportion of articles that examined each of the three mechanism categories did not vary by more than 1.8% between 200 and 250 articles. Thus, our results form a representative sample of relevant articles.

2.4.2 Data collection and analysis

The 250 articles selected for inclusion (a list of the data sources is found in Appendix A) were reviewed in full to extract the following information: type of mechanism examined (limited resource access, restricted demographic exchange, or impeded gene flow); geographic location of the study (tropics, subtropics as defined by Corlett (2013), and the temperate zone plus higher latitudes); and taxon examined (invertebrate, fish, amphibian, reptile, bird, or mammal) (Table A.1). Counts across categories may exceed 250 because some articles fell into multiple groupings. A chi- squared goodness of fit test was used to assess whether research was evenly distributed across the three mechanisms. Fisher’s exact tests were used to determine if the relative proportions of observed mechanisms were different than expected across geographic zones and across taxon. Analyses were conducted using R version 3.2.0 (R Core Development Team, 2015).

28 2.5 RESULTS

Our initial keyword search generated 4,017 results, 2,152 of which were passed through our filtering process in order to obtain 250 articles to analyse. Of the filtered articles, 228 (10.6%) were not an empirical study, 632 (29.4%) were not based on animal taxa, 684 (31.8%) did not study the effects of habitat fragmentation, and 358 (16.6%) did not investigate the effects of habitat fragmentation on one of the mechanisms of interest.

We found significant differences in research effort across the three mechanisms (Chi-square, χ2 = 201.17, df = 2, p < 0.001); 184 articles (73.6%) investigated impeded gene flow, 33 articles (13.2%) focused on restricted demographic exchange, and 24 articles (9.6%) addressed limited resource access. Nine articles studied the effects of habitat fragmentation on multiple mechanisms; eight considered both restricted demographic exchange and impeded gene flow, while one considered limited resource access and restricted demographic exchange. Over the 20-year period, the number of articles investigating impeded gene flow increased from zero per year to 27 per year, while research on limited resource access and restricted demographic exchange increased from zero to five per year (Figure 2.2).

Most work was conducted within the temperate zones or higher latitudes (204 or 75.3%) while 30 (11.1%) articles were from subtropical regions, and 37 (13.7%) articles were from the tropics (Figure 2.3). In all geographic zones, research on impeded gene flow constituted the greatest proportion of articles and the difference in relative frequency of articles on each mechanism was not statistically significant among regions (two-sided Fisher’s Exact Test p = 0.155; Figure 2.4).

The most frequently represented taxa in our sample were mammals (34.4%) and invertebrates (22.0%), with fish (14.8%), birds (12.8%), amphibians (11.2%), and reptiles (8.8%) receiving less attention (Figure 2.5). There were significant differences in the frequency with which each mechanism was studied within each taxon (two-sided Fisher’s Exact Test p < 0.001). Impeded gene flow was the mechanism most studied in all categories. Restricted demographic exchange was most studied in mammals (21 or 24.4% of all mammal articles). Work on invertebrates had the greatest proportion of articles investigating limited resource access (13 or 23.6% of all invertebrate articles), while there was no limited resource access research performed on amphibians, fish, or reptiles within our sample.

Figure 2.4. Proportion of articles from a random selection of articles published since 1994 that studied the effect of habitat fragmentation on an ecological mechanism categorised by the geographic zone where the study was conducted. “Temperate” includes the temperate zone plus higher latitudes. The total number of articles included in each zone is indicated above each bar.

Our conceptual framework emphasises how movements restricted by habitat fragmentation have implications at varying scales and levels of ecological organisation. The framework highlights that limited resource access, restricted demographic exchange, and impeded gene flow take effect over progressively larger spatial and temporal scales, and the consequences of each become evident after progressively longer time lags. Although all three mechanisms are likely to be important, we identified through our literature review that impeded gene flow has received the overwhelming majority of research attention.

Our framework can assist researchers to adopt a multi-scale perspective when considering the impacts of habitat fragmentation. By considering the link between time and space, researchers can more clearly disentangle potential fragmentation effects. For example, occupancy models for eastern yellow robins in Queensland, Australia indicate that habitat fragmentation (at a home range scale and at a dispersal scale) is not currently having a deleterious effect on occupancy (Chapter 3). However, genetic studies on eastern yellow robins in Victoria, Australia reveal that fragmentation is having some detrimental impacts on demography and gene flow (Harrison et al. 2012). This discrepancy could be attributed to several differences between the regions. Queensland has a more recent history of land clearing than Victoria (ECC 1997; Seabrook et al. 2006), plus the Queensland landscape appears to have a higher degree of structural connectivity than the Victorian landscape, possibly due to the presence of Travelling Stock Routes (Chapter 3; Lentini et al. 2011a; Harrison et al. 2012). By considering the impacts of movement restrictions over both time and space, researchers can more effectively tease apart the potential causes of change.

2.6.1 Implications and limitations of the conceptual framework

It is crucial to distinguish among the three mechanisms in the conceptual framework. Failing to consider multiple scales of movement restrictions potentially weakens conservation recommendations because each mechanism may require different conservation responses. This is especially true when multiple mechanisms are being affected. For example, limited resource access could be addressed by improving habitat quality, thereby increasing the carrying capacity of the habitat patch (Bonnot et al. 2013) and reducing the need for multiple patch use. Such a strategy would not be sufficient if the mechanism at work was restricted demographic exchange. Instead, this could be more effectively facilitated by adding corridors (linear strips of habitat) or stepping stones (small patches or isolated trees) to the matrix to help connect isolated patches (Fischer &

33 Lindenmayer 2002; Baum et al. 2004; Saura et al. 2014). These solutions could aid movements by making distant habitat patches more reachable and providing visible targets within the animal’s perceptual range.

In contrast, gene flow needs successful dispersal plus breeding but these do not always occur together (Lowe & Allendorf 2010; Peery et al. 2010). Dispersers often have lower fitness and reproductive success in new localities than philopatric individuals (Hansson et al. 2004) and this pattern has been attributed to several causes, including energetic costs incurred while dispersing (Coulon et al. 2010; Bonte et al. 2012). Increasing the permeability or quality of the matrix may help individuals travel more easily and supplement their food along the way (Zollner & Lima 2005; Driscoll et al. 2013) thereby increasing their chances of successfully reproducing. Alternatively, translocating genetically distinct individuals to inbred populations could provide rapid and substantial improvements, especially where it is not practical to increase landscape connectivity (Madsen et al. 1999; Pelletier et al. 2017).

Several frameworks have previously been developed for conceptualising the structural composition of landscapes (McIntyre & Barrett 1992; Wiens et al. 1993; Forman 1995; Watson 2002; Fischer et al. 2004; Manning et al. 2004). However, these frameworks were not extended to classify the mechanisms driving isolation effects on animal populations. Some authors have discussed how different types of movements are related to space and time (Roshier and Reid 2003). For example, Jeltsch et al. (2013) explored how different types of movements occurred over characteristic spatiotemporal scales and impacted ecosystems as a whole, such as through interspecific interactions and resource transportation. We have not focused on the relationship between movement type and scale, but rather take a narrower focus and instead consider how different movement restrictions influence mechanisms acting on species.

Driscoll et al. (2013) outlined how the quality of the matrix in fragmented landscapes influences proximate processes. Specifically, they discussed how features of the matrix influence species’ persistence in the landscape through subsidising resource availability, altering abiotic microclimates, and facilitating movements. For instance, woodland-dependent species may more readily travel between woodland patches interspersed with a pine plantation matrix than one that is more structurally dissimilar to the woodland (Driscoll et al. 2013). Our frame of reference aligns with that of Driscoll and colleagues (2013); we consider the role of functional connectivity on movement, which is heavily dependent on matrix quality. Thus, both frameworks are complementary for conceptualising the impacts of habitat fragmentation. In contrast to Driscoll et

34 al. (2013) we explore one of the three core effects of the matrix: how barriers to movement affect species. We also elucidate the relationships between the type of mechanism affected by fragmentation and the distance or timescale over which movements are restricted.

Many mechanisms other than those linked directly to physical isolation are important in fragmented landscapes, such as edge effects (Ries et al. 2004; Porensky & Young 2013). Although we do not specifically include these in our framework, they are nonetheless relevant. One component of edge effects is the inability to source sufficient resources, such as sheltered protection from predators (Bennett et al. 2013). Habitat gaps between shelter sites and feeding locations could pose substantial risks. These mechanisms are incorporated into our framework when they are driven by increased isolation, such as when habitat gaps make movement risky. However, these impeded mechanisms could also result from reduced habitat quality, such as when abiotic changes along edge habitat leads to changes in flora (Ries et al. 2004). It is important to distinguish whether habitat fragmentation per se, or related changes such as degradation, are impacting species because the conservation solutions differ.

2.6.2 Distribution of research effort among mechanisms

The most frequently-examined of the three mechanisms was impeded gene flow. It is unlikely that the heavy weighting towards genetic articles is an artefact of the keyword choices used for each mechanism because this imbalance was also evident in an informal scan of articles produced by a search of just “habitat fragmentation”. It is unsurprising that these numbers have increased in recent years given the greater availability and affordability of genetic technology (Storfer et al. 2010; Bolliger et al. 2014). However, given that molecular methods, such as assignment tests and pedigree analyses, can be used to infer dispersal movements (Manel et al. 2005; Holderegger & Wagner 2008), it is surprising that the frequency of articles investigating restricted demographic exchange have not similarly spiked over the past decade.

The key mechanisms we have identified in this paper can be considered to form an overlapping, non-nested hierarchy across scales. Each successive mechanism works at progressively larger scales (Wiens 1989). When habitat is fragmented, it takes time for the evidence of the negative effects to become apparent and the duration of this time lag is dependent upon where the affected mechanism sits in the hierarchy (Wiens 1989; Metzger et al. 2009). As the scale of fragmentation increases, impacts on mechanisms may accumulate. Systems that show signs of impeded gene flow may also be subject to impediments to demographic exchange and resource access. Thus, neglecting to take

35 the impacts on these lower-level mechanisms into account could compromise conservation efforts. Additionally, impacts can have synergistic effects on ecological systems (Zanette et al. 2003; Brown et al. 2016). It is currently unknown what interactive effects, if any, operate among the three mechanisms discussed here, but synergisms may partly explain why fragmentation effects are amplified at very low levels of habitat extent (Andrén 1994; Radford et al. 2005).

2.6.3 Geographic and taxonomic biases

The overall bias against limited resource access and restricted demographic exchange as a focus of research was disproportionately pronounced in particular taxa, notably reptiles, amphibians, and fish. It is difficult to generalise across taxa, but reptiles and amphibians in particular contain a greater proportion of less-vagile species than birds, large mammals, and flying insects (Mac Nally & Brown 2001; Allentoft & O’Brien 2010; Rivera-Ortíz et al. 2015). A considerable gap in knowledge about mechanisms operating at small to intermediate scales in generally less-mobile taxa could have important implications. For instance, landscape structure across small extents may have a greater impact on sedentary species than nomadic species (Maron et al. 2012). It will be important to gather a body of empirical data across a range of dispersal capabilities, range requirements, generation times, and environments (notably aquatic versus terrestrial) both within and between taxa. Without this, it would not be appropriate to generalise findings.

As has been often identified in literature reviews (Lawler et al. 2006; Bayard & Elphick 2010; Magle et al. 2012), the southern hemisphere and the tropical/subtropical regions were underrepresented in the literature we studied. However, the frequency that each mechanism was studied was consistent across geographic locations. The relative importance of different types of habitat fragmentation effects may vary between the northern temperate zones and lower latitudes. Many species undertake long-distance migrations to avoid harsh winters in the northern temperate region, while lower latitudes have a greater proportion of sedentary species (Newton & Dale 1996a, 1996b; Chesser 1998). This means there are important differences in dispersal ability, life history traits, and population demography in species between these regions as well as abiotic factors, such as climate and productivity, which may affect the relative influence of both habitat fragmentation and each of the three mechanisms on species’ responses (Newbold et al. 2013; Betts et al. 2014; Rivera-Ortíz et al. 2015; Keinath et al. 2017). Additionally, regions with a longer clearing history, such as the heavily studied areas in the northern temperate zone, are likely to be heavily impacted by impeded gene flow. Thus, geographic biases in research effort, especially when they occur in

36 conjunction with a bias towards a particular mechanism, may be hampering our ability to draw broad conclusions.

2.6.4 Conclusions

Habitat fragmentation poses a significant threat to species worldwide, but there is still uncertainty about the mechanisms through which population changes occur (Banks et al. 2007). More research targeting potential proximate causes is needed to help clarify our understanding, identify species at risk, and assess the benefits to targeted species of proposed conservation solutions. Our framework is designed to help stimulate thinking about the mechanisms potentially affected by habitat fragmentation, how the mechanisms relate to the spatial and temporal scale of movement restriction, the hierarchical nature of the mechanisms, and the consequences of having multiple mechanisms impeded simultaneously. By taking these factors into consideration, conservation managers will be well equipped to design targeted and effective interventions that address the core causes of species declines.

It has long been recognised that taxonomic and geographic biases in research effort have plagued ecology (Gaston & May 1992; Clark & May 2002). Here, our focus was not on the proportion of studies conducted in each taxon or geographic zone. Rather, we emphasised the increasing disparity in research effort amongst mechanisms and why this is concerning in light of the existing taxonomic and geographic biases. Given the global threat of land-use change to species and ecosystems, it is paramount that landscape ecologists derive a deeper understanding of the mechanisms through which habitat fragmentation drives ecological change. This could be achieved by finding innovative ways to approach studies of resource access and demography, such as with stable isotope tracers (Robb et al. 2011), drone technology (Whitehead et al. 2014), and radio frequency identification (RFID) systems (Bonter & Bridge 2011). Focusing research attention on mechanisms, rather than simply patterns of species occurrence and distribution, will garner new insights and progress our understanding of the threats posed by habitat fragmentation.

37 CHAPTER 3: Using individual condition measures to predict the long-term importance of habitat extent for population persistence

This chapter has been accepted for publication by Conservation Biology. Approximate percentage author contributions as follows: Anita Cosgrove – 73% Todd McWhorter – 3% Martine Maron – 24%

38 3.1 ABSTRACT

Habitat loss and fragmentation are causing widespread population declines, but identifying how and when to intervene remains challenging. Predicting where extirpations are likely to occur and implementing management actions before losses eventuate may be more cost-effective than trying to re-establish lost populations. Early indicators of pressure on populations could be used to make such predictions. In 2009/2010, the presence of eastern yellow robins (Eopsaltria australis) within 42 sites in a fragmented region of eastern Australia was found to be unrelated to woodland extent within 500 m of a site, but the robins’ heterophil:lymphocyte ratios (an indicator of chronic stress) were elevated in less-wooded sites. Such sites may be poorer quality with a higher probability of extirpation. Four years later, we resurveyed these 42 sites for robin presence to test whether 1) previously-measured H:L ratios could predict where extirpations would occur and 2) the previous pattern in H:L ratios was an early sign that woodland extent would emerge as a predictor of occupancy. We also surveyed robin occupancy at an additional 43 sites to determine whether current occupancy could be better predicted by landscape context at a larger scale, relevant to dispersal movements. We found no relationship between H:L ratios and extirpations; however, only four were observed. Woodland extent did emerge as a strong predictor of occupancy, consistent with the hypothesis that patterns of individual condition may reveal habitat relationships that become evident as local shifts in occupancy occur, but that are not uncovered by a snapshot of species distribution. Woodland extent at larger scales relevant to dispersal (5 km) was not related to occurrence. We recommend that conservation actions be focused on regenerating neighbourhoods of habitat large enough to support robin territories, rather than increasing connectivity within the landscape.

3.2 INTRODUCTION

Among the greatest conservation challenges is mitigating the threat to biodiversity posed by the severity and extent of habitat loss and fragmentation (Vié et al. 2009). Much work has been undertaken to identify relationships between declining populations and landscape change, why landscape structural change causes declines, and what actions can be taken to mitigate the observed declines (Turner 2005; Cushman 2006). However, the time lag between when land is cleared and when population declines become evident can complicate species management (Hylander & Ehrlén 2013). Acting early can mean the difference between species extinction or recovery (Martin et al. 2012).

39 Patterns of physiological measures can act as early warning signs of population changes, such as population declines (Wikelski & Cooke 2006). For instance, one week after an oil spill at the Galapagos Islands, corticosterone concentrations in marine iguanas (Amblyrhynchus cristatus) had risen dramatically (Wikelski et al. 2001). Within one year, population mortality had increased by approximately 30% (Wikelski et al. 2002). Stressors often affect the physiological health of individuals long before the consequences of the stress-inducing event or environment are apparent through reduced fitness or demographic parameters (Ellis et al. 2012).

Using physiological measures as a predictive tool could be advantageous in regions likely to be undergoing faunal relaxation (Ellis et al. 2012): the process whereby a system that has been subjected to increased environmental pressures loses individuals, populations, or species over time before settling into a new, more-depauperate equilibrium (Diamond 1972; Hylander & Ehrlén 2013). Researchers often identify relationships between species distributions and variables of interest, such as landscape structure, when exploring causes of population declines (Turner 2005; Cushman 2006). However, species distribution models are often built from data collected over a limited period of time (Guisan & Thuiller 2005). This snapshot of distribution can only reliably be used to make predictions when the data were collected from a system in equilibrium with its environment (Guisan & Thuiller 2005). This assumption is not met when populations are actively declining, such as during faunal relaxation and alternative means for predicting population declines may be useful.

Maron et al. (2012) proposed using heterophil:lymphocyte (H:L) ratios, an indicator of chronic stress (Davis et al. 2008), to predict future patterns of eastern yellow robin (Eopsaltria australis) persistence. The concentrations of different types of white blood cells in circulation change in response to stressors; in birds, heterophil concentrations increase while lymphocytes decrease (Johnstone et al. 2012). Individuals can have elevated stress levels in landscapes containing less habitat (Hinam & St. Clair 2008). This could be due to mechanisms such as food availability, predation pressure, or energetic costs. Maron et al. (2012) found a relationship between the extent of woodland in the surrounding landscape and robin H:L ratios, but not between woodland extent and site occupancy by robins. They hypothesised that sites containing robins with elevated H:L ratios were suboptimal, perhaps because home ranges may need to encompass multiple patches or be condensed into smaller patches. Further, they predicted that a relationship between occurrence and the extent of woodland in the surrounding landscape may emerge in the future. Four years later, we revisited the study area to test these predictions.

40 First, we tested the prediction that there would be more extirpations at sites where robins had relatively higher H:L ratios four years earlier (Maron et al. 2012). This aim focused on changes that occurred at the level of particular sites. Second, we tested whether the relationship they identified between H:L ratios and woodland extent was an early signal that woodland extent would start to influence patterns of site occupancy by robins. This aim focused on patterns at the population level – could patterns of H:L ratios predict changes to patterns of occupancy? Third, we refined the occupancy model and tested whether we could better explain robin occupancy by accounting for the broader landscape context (within 5 km) over which dispersal movements may play a role.

3.3 METHODS

3.3.1 Study region

The study was conducted around Meandarra (27°19´S, 149°52É), Moonie (27°43´S, 150°21É), and Goondiwindi (28°33´S, 150°18É) in the Brigalow Belt South bioregion of Queensland, Australia (Figure 3.1). The area has a subtropical climate with an average annual rainfall of approximately 600 mm, mostly falling in summer (Bureau of Meteorology 2015). The study was carried out in brigalow-dominated (Acacia harpophylla) woodlands on the low, fertile plains. On the less-fertile rises, forests are dominated by the tree species belah (Casuarina cristata), Eucalyptus spp. or white cypress pine (Callitris glaucophylla).

The study region has a relatively recent clearing history. Although settlement in the Brigalow Belt South bioregion began in the late 1800s, clearing woodland for agricultural purposes did not accelerate until the 1930s (Seabrook et al. 2006). However, damaged brigalow regenerates through suckering, so brigalow-dominant woodlands were only intensively cleared after blade ploughs were introduced in the 1960s (Seabrook et al. 2006). There has been little change to total woodland extent within the study region since 2009. Today, less than 15% of the study area is comprised of any type of remnant woodland, while less than 4% is remnant brigalow woodland. The majority of the remaining brigalow woodland occurs in narrow strips along roadsides and paddocks, with a few larger patches in protected areas.

3.3.2 Study species

Eastern yellow robins are sedentary, ground-foraging insectivores that live in pairs or small family groups (Higgins & Peter 2002). Each pair occupies a home range of about 5-6 ha which they defend

Figure 3.1. Location of the study sites. Solid circles indicate sites where eastern yellow robins were present while hollow diamonds represent sites where robins were not detected. Grey shading depicts the extent of all woodland, including remnant and regrowth. Grey shading on the inset denotes the Brigalow Belt South bioregion. against conspecifics during the breeding season (Higgins & Peter 2002). They are woodland- dependent birds that avoid open areas and cleared habitat (Higgins & Peter 2002). Based on work in southern Australian landscapes, they are thought to be sensitive to woodland patch area and isolation (Barrett et al. 1994; Watson et al. 2005). Although this species has a Red List status of Least Concern, population declines have been reported in some areas, especially sections of Victoria and New South Wales (Stevens & Watson 2013; BirdLife Australia 2015b). While Queensland populations currently appear stable, this may be due to their more-recent clearing history. Assessing early changes to population dynamics before declines are evident could provide valuable information.

42 3.3.3 Occupancy and individual condition data

Between August 2009 and September 2010, Maron et al. (2012) surveyed 42 sites for the presence of eastern yellow robins and recorded H:L ratios from peripheral blood smears for 51 birds across 27 sites. Blood for the smears was collected via brachial venipuncture and smears were later stained with Wright-Giemsa (Maron et al. 2012). Differential white blood cell counts were conducted by counting 100 leukocytes per sample at x400 magnification, and where possible, up to three counts were made per bird and averaged (Maron et al. 2012). The mean H:L ratios for each individual spanned a broad range (0.06 – 1.29) (Table B.1). From March 2013 to November 2014, we resurveyed these original sites plus an additional 43 sites for robin occupancy, using the methods of Maron et al. (2012). No individual birds were resampled for H:L ratios. All sites were 200 m x 50 m, occurred in brigalow-dominant woodland with a tree density of > 150 stems per ha, and were > 800 m apart to minimise spatial dependency. We conducted surveys by using call playback while walking a 200 m transect centred on the site. Robins are territorial and quickly respond to conspecific calls (Higgins & Peter 2002). Each survey was carried out for a minimum of 6 min in the early morning or late afternoon. Sites at which no response was detected were surveyed a second time at least one month after the first survey. Surveys were conducted year round; however, robins were noticeably more responsive to the call playback during the breeding season. Therefore, we ensured that all sites at which no robins were recorded were surveyed at least once during the peak of the breeding season (September-October). When a robin was seen or heard within a site at least once, the site was classed as having birds present. If both surveys elicited no response, the site was classified as unoccupied.

3.3.4 Explanatory variables

Following the methods of Maron et al. (2012), we collected habitat data from each site. Four 20 x 20 m quadrats were positioned randomly within each site and we counted the total number of shrubs within each quadrat. We also counted the number of wilga (Geijera parviflora) shrubs within each quadrat as we thought their high volume of litter fall may be important for robins. A 1 m2 microquadrat was positioned in each corner of each quadrat. A photograph was taken of each microquadrat from above and used to assess the proportion of live or standing dry grass cover since Maron et al. (2012) identified that grass cover was the most important predictor of robin site occupancy. We also searched each site for the presence of gilgai (classified as present/absent): depressions that form within the cracking clay soil when the ground is undisturbed (Russell 1973).

43 Gilgai may provide an indirect indicator of previous land management practices, such as levelling or thinning.

We calculated landscape variables from a habitat map based on the Biodiversity Status of Pre- clearing and Remnant Regional Ecosystems dataset and the Statewide Landcover and Trees Study Foliage Projective Cover 2011 dataset which have spatial resolutions of 5 ha and 25 m respectively (State of Queensland 2014a, 2014b). We classed woody vegetation in accordance with Maron et al. (2012) as any vegetation with a minimum foliage projective cover of 8% and a minimum patch size of 2 ha. We calculated patch width from the centre of the site as the shortest distance between two opposite edges.

We considered landscape metrics at two neighbourhood sizes: within a 500 m and within a 5 km radius of the centre of each transect. The 500 m radius neighbourhood matches that used by Maron et al. (2012) and represents an area encompassing a robin home range. Within a 500 m radius of each site, we measured the percentage of the neighbourhood area covered by remnant woody vegetation (remnant woodland within 500 m), the proportion of the neighbourhood area covered by all woodland, i.e. both remnant plus regrowth (total woodland within 500 m), and the number of habitat patches consisting of any type of woodland (no. patches). We calculated further metrics within a 5 km neighbourhood radius based on most recorded dispersal distances from the Australian Bird and Bat Banding Scheme (David Drynan, personal communication; Higgins & Peter 2002). Within this radius, we calculated the proportion of the neighbourhood area covered by all woodland (total woodland within 5 km) and the proportion of the neighbourhood area that consisted of accessible woodland (accessible woodland within 5 km). Accessible woodland was defined as any woodland accessible from the site with interpatch distances less than 75 m. This threshold was based on Doerr et al.’s (2011) finding of a 75 m average gap crossing distance for four species of woodland birds (including eastern yellow robins).

3.3.5 Data analysis

All analyses were performed in R version 3.2.0 using an information-theoretic approach (Burnham & Anderson 2002; R Core Development Team 2015). We checked for multicollinearity amongst all explanatory variables. Patch width was highly correlated with total woodland within 500 m (r = 0.93) and excluded from further analysis. There was another correlation in our third aim which we address later. We applied a fourth-root transformation to the H:L ratio data to improve normality. All explanatory variables were standardised to a mean of zero and a standard deviation of 0.5 to

44 allow comparison of parameter estimates (Gelman 2008). Some landscapes at the 5 km scale overlapped. We used spline correlograms from the ncf package to check for spatial autocorrelation in our response variables and in all the model residuals and found no causes for concern (Bjornstad 2015). We detected no overdispersion in our models.

To conduct model averaging, we used Akaike’s information criterion adjusted for small samples sizes (AICc) to assess the relative fit of a priori models and calculated averaged parameter estimates with the MuMIn package (Bartoń 2015). Averaging was performed across all models within the 95% confidence set and the relative importance of the predictors was assessed with summed Akaike weights (Σωi). The variance explained by each candidate model was assessed with Nagelkerke’s R2 value (Nagelkerke 1991). We used the hier.part package to perform hierarchical partitioning that identified the proportion of independently-explained variance that was attributable to each predictor (Chevan & Sutherland 1991).

For our first aim, we tested whether extirpations were more likely to occur where H:L ratios were higher using data from the 27 sites at which Maron et al. (2012) measured H:L ratios. We used a generalised linear model with a binomial error distribution based on the bias-reduction method developed by Firth (1993) using the brglm package (Kosmidis 2013). The bias-reduction method is recommended for use when the binomial response variable is strongly imbalanced—we had only four cases of extirpations. Using the same explanatory variables as Maron et al. (2012), we modelled extirpations as a function of H:L ratio, northing, grass cover, gilgai, no. shrubs, no. patches, and remnant woodland within 500 m.

Our second aim investigated whether the relationship between H:L ratios and habitat extent predicted future changes to patterns of occupancy. Since Maron et al. (2012) identified that, on average, H:L ratios tended to be higher in landscapes with less surrounding woodland, we tested whether the extent of remnant woodland was now an important predictor of occupancy. We tested this with a generalised linear model with a binomial error distribution applied to the current occupancy data from the 42 original sites. We modelled the effect of northing, grass cover, gilgai, no. shrubs, no. patches, and remnant woodland within 500 m on persistence (presence/absence).

Our final aim was to refine the occupancy model and determine whether woodland cover at a scale relevant to dispersal movements was an important predictor of occupancy. This aim was addressed using the full occupancy dataset of 85 sites; we randomly selected 70% of the sites for inclusion in model training (n = 60) and withheld 30% of our samples for model validation (n = 25). We refined

45 Maron et al.’s (2012) occupancy model in three ways to develop a more-complete picture of factors associated with robin occurrence. First, we replaced the total number of shrubs with number of wilga. The robins commonly foraged underneath wilga (Anita Cosgrove, personal observation), possibly because the additional leaf litter attracted invertebrates. Second, we replaced remnant woodland within 500 m with total woodland within 500 m. Although the two measures were highly correlated, the latter may have better represented how the birds experienced the landscape given they were observed in non-remnant woodland.

Third, we replaced number of patches with a measure of the broader landscape context at a scale likely to apply to robin dispersal movements. Initially, we included total woodland within 5 km. This variable was moderately correlated with northing (r = -0.5) so we removed northing from the model and instead included a random effect for study region (Meandarra/Moonie/Goondiwindi). Thus, we used a generalised linear mixed effects model to assess occupancy as a function of grass cover, gilgai, no. wilga, total woodland within 500 m, and total woodland within 5 km. However, connectivity could play an important role when considering the broader landscape context. As part of this final aim, we ran a second generalised linear mixed effects model that included ‘accessible’ woodland within 5 km (i.e. woodland that could presumably be accessed from the site given that interpatch distances were < 75 m) instead of total woodland within 5 km.

Of the above models, we selected the one with the lowest AIC and calculated a probability of occupancy for each sample within the validation dataset. With the aid of the PresenceAbsence package, we used the area under the receiver operating characteristic curve (AUC), the sensitivity, and the specificity to examine the model’s performance (Freeman & Moisen 2008). The threshold was adjusted to the prevalence in the training data (threshold = 0.648) as suggested by Lobo et al. (2008).

3.4 RESULTS

3.4.1 Predicting locations of extirpations from H:L ratios

Of the 27 sites where H:L ratios were originally obtained, we were unable to detect robins at only four sites, so robust evaluation of this prediction was not possible. Model averaging indicated that H:L ratios were not an important predictor of extirpations; in fact, no variable was important (Table 3.1; Table C.1; Figure 3.2). The global model explained 9% of the variance in occupancy.

46 Table 3.1. Averaged coefficients, adjusted standard errors, confidence intervals and summed

Akaike weights (Σωi) derived for each variable from models of eastern yellow robin extirpations (n = 27).

Variable Coefficient Adjusted SE Lower CI Upper CI Σωi H:L ratio -0.078 0.945 -1.931 1.775 0.20 Northing -0.675 1.130 -2.890 1.540 0.25 Grass cover -0.813 0.928 -2.631 1.005 0.28 Gilgai -1.078 1.636 -4.284 2.128 0.29 No. shrubs 0.941 1.208 -1.427 3.309 0.32 No. patches 0.649 1.102 -1.510 2.809 0.27 Remnant woodland within 500 m 1.016 1.122 -1.184 3.215 0.33 All 95% confidence intervals spanned zero.

Figure 3.2. Percentage of independently-explained variance for each independent variable from a model of eastern yellow robin extirpations (n = 27). The independently-explained variance was non-significant for all variables.

3.4.2 Predicting occupancy relationships from relationships between H:L ratios and habitat extent

Maron et al. (2012) detected robins at 28 of the 42 sites. We detected robins at 30 of the original sites with a shift in occupancy evident through colonisations at six sites and extirpations at four sites. The global model explained 32% of the variance in occupancy. The summed Akaike weights for remnant woodland within 500 m (positive effect) and grass cover (negative effect) were 0.93 and 0.86 respectively, and the confidence intervals of the estimate for remnant woodland within 500

47 m did not include zero (Table 3.2). In contrast to Maron et al. (2012), gilgai presence was no longer an important predictor of occupancy; it had a summed Akaike weight of 0.25 (Table 3.2; Table C.2; Figure 3.3a; Figure 3.3b). Hierarchical partitioning revealed that remnant woodland within 500 m and grass cover had the greatest independent influences on occupancy and each was responsible for 40% of the independently-explained variance (Figure 3.3b).

Table 3.2. Averaged coefficients, adjusted standard errors, confidence intervals and summed

Akaike weights (Σωi) derived for each variable from models of eastern yellow robin occurrence.

Variable Coefficient Adjusted SE Lower CI Upper CI Σωi Woodland extent within 500 m only (n = 42) Northing 1.258 0.959 -0.622 3.137 0.43 Grass cover -2.175 1.148 -4.425 0.074 0.86 Gilgai -0.815 1.203 -3.172 1.543 0.25 No. shrubs 1.505 1.228 -0.901 3.911 0.44 No. patches -0.075 0.837 -1.716 1.565 0.21 Remnant woodland within 500 m 2.545 1.231 0.132 4.958 0.93 Woodland extent within 5 km (n = 60) Grass cover -1.488 0.823 -3.101 0.125 0.76 Gilgai 0.842 0.863 -0.849 2.533 0.30 No. wilga 2.297 1.364 -0.377 4.971 0.72 Total woodland within 500 m 3.112 1.056 1.041 5.182 1.00 Total woodland within 5 km -1.818 1.085 -3.945 0.310 0.57 Accessible woodland extent within 5 km (n = 60) Grass cover -1.506 0.824 -3.120 0.108 0.77 Gilgai 0.786 0.871 -0.922 2.493 0.28 No. wilga 2.347 1.372 -0.342 5.037 0.73 Total woodland within 500 m 3.090 1.035 1.063 5.118 1.00 Accessible woodland within 5 km -1.781 1.037 -3.813 0.252 0.59 Coefficients with 95% confidence intervals not spanning zero are denoted in bold.

3.4.3 Testing the effect of landscape context at larger extents

Robins were found at 55 of the 85 sites and only occupied patches less than 75 m wide in one instance. Both the global model incorporating total woodland within 5 km and the global model including accessible woodland within 5 km each explained 39%. Regardless of whether we included total woodland within 5 km or accessible woodland within 5 km in our models, in each case, there was uncertainty over the best model (Table C.3; Table C.4). In both cases, total woodland within 500 m had a summed Akaike weight of 1.00 (Table 3.2). Total woodland within 5 km and accessible woodland within 5 km were not important and the confidence intervals of their estimates crossed zero (Table 3.2; Figure 3.4a; Figure 3.4b).

The global model incorporating accessible woodland within 5 km had the lower AICc (Table C.3; Table C.4), so validation proceeded on this model. It had reasonable discriminatory ability with an AUC of 0.71 (Figure D.1a). Sensitivity was 0.56 while specificity was 0.78. Confidence intervals were quite large due to the small sample size in the validation dataset (Figure D.1b).

3.5 DISCUSSION

Our results did not support the prediction that extirpation rates would be higher at sites where robins had higher H:L ratios four years earlier. It is unsurprising we were unable to detect a relationship between the original H:L ratio measurements and current occupancy given only four extirpations were detected since 2009/2010. Evidence that general patterns of stress measures could signal

50 upcoming changes to patterns of occupancy was inconclusive. Given there were slightly more colonisations than extirpations, the emergence of habitat extent as a predictor of occupancy was not driven solely by losses of robins from suboptimal less-wooded sites. However, as expected, woodland extent became the primary predictor of occupancy. At the time of our surveys, robin occupancy was strongly associated with woodland extent at a home range scale, but not across larger scales that may affect dispersal movements; the connectivity of the woodland across larger scales had little effect on occupancy. In line with the findings by Maron et al. (2012), all our occupancy models suggested that robins were less likely to inhabit sites with high levels of grass cover, probably because grass would provide cover for invertebrates and reduce the robins’ ability to detect prey (Maron & Lill 2005).

The H:L ratios of robins in this region were previously identified by Maron et al. (2012) as being negatively related to woodland extent within a 500 m radius. Maron and colleagues suggested that because of this, although occupancy was not yet associated with woodland extent, extirpations were more likely from sites in less-wooded neighbourhoods. We did not find evidence to support this prediction; we observed only four extirpations and these were not limited to sites that had previously contained robins with high H:L ratios. Wikelski et al. (2001) predicted heavy marine iguana mortalities from a population that was exposed to an oil spill based on a sudden spike in stress levels, and this outcome did eventuate (Wikelski et al. 2002). While physiological measures did provide an accurate early signal of upcoming localised mortalities in that case, the changes were in response to a sudden stressor with immediate impacts on individuals. In our study, the potential stressor (i.e. habitat loss) is chronic in nature. H:L ratios fluctuate within individuals in response to many factors, including age, reproductive status, and food availability (Johnstone et al. 2012). Given the potentially low repeatability of stress measures within individuals (Ochs & Dawson 2008), it seems inappropriate to use a single snapshot of stress levels to identify specific locations to be impacted by a chronic stressor, as opposed to identifying statistical patterns. Obtaining repeated measures from individuals over time and assessing differences in the rate of change within individuals may be a more-promising avenue of investigation.

We found conflicting evidence as to whether statistical patterns of physiological measures foreshadowed changes in patterns of occupancy. While stress measures are expected to fluctuate within individuals over time, at the population level, Maron et al. (2012) found there was a tendency for H:L ratios to be higher on average at sites with less surrounding woodland. While Maron et al. (2012) found no relationship between woodland extent and site occupancy, we found that woodland extent within 500 m had now become the strongest predictor of occupancy.

51 However, the result was driven by both extirpations at sites that generally had little surrounding woodland, and colonisations at sites with intermediate to high degrees of woodland cover. This pattern of extinctions and colonisations suggests that, generally speaking, less-wooded sites may indeed be suboptimal for robins in some way and the birds may be redistributing themselves within the region.

Other explanations are possible for the increased importance of habitat extent. Habitat change may not have been adequately captured by our chosen variables. For instance, small patches, such as those that occurred at our sites with very little surrounding woodland, may have generally been subject to different management strategies than large patches. Small patches also have more edge habitat, which can differ in temperature, wind, moisture, and floral communities, all of which may have subtle impacts on patch quality (Chen et al. 1999). Noisy Miners (Manorina melanocephala) favour edge habitat, aggressively exclude smaller birds from woodlands, and have increased in incidence across their range in recent years (Maron et al. 2013). Any expansion of their occurrence in the study region would likely come at the expense of robins.

Another possible explanation is that we have not accounted for natural fluctuations in occurrence and distribution due to seasonal and annual changes to resource availability and climate (Maron et al. 2005; Williams & Middleton 2008). While both this study and that of Maron et al. (2012) were conducted in years experiencing below-average rainfall, it is not clear how much these drier conditions affected resource distribution, and by extension, robin distribution. Resurveying each site across multiple and higher rainfall years would help provide a more-complete picture of whether the change we observed, although as predicted, was part of a longer-term trend.

An important limitation of this study is that we only used a single measure of condition. H:L ratios rise when the hypothalamus-pituitary-adrenal axis system is activated in response to a stressor or the immune system is fighting inflammation or infection (Davis et al. 2008; Weiss & Wardrop 2010). Ideally, multiple measures of condition would be used to differentiate between these causes. Thus, although H:L ratios are generally accepted to be a reliable indicator of chronic stress (Davis et al. 2008; Johnstone et al. 2012), it is possible that our results indicate that infection was relatively more prevalent within less-wooded landscapes. While this alternative explanation has implications for understanding why habitat extent had become increasingly important for occupancy, it still supports our conclusion that less-wooded sites appear to be suboptimal for robins.

52 Another limitation of this study was that extirpations are a coarse evaluation of population change and represent the final endpoint in a process that could have subtle and gradual effects over many years (Ellis et al. 2012). While occurrence data are relatively cheap and easy to collect (Joseph et al. 2006; Ellis et al. 2012; Casner et al. 2014), further assessments of the predictive power of physiological measures would benefit from incorporating a more-intermediate indicator of population change. Abundance data would fill this gap and allow researchers to assess changes before populations go locally extinct. Eastern yellow robins live in pairs or small family groups (Higgins & Peter 2002), so abundances at the site level would provide little information. However, evaluating changes in abundances over broader spatial scales, such as areas that could encompass several home ranges, could be valuable.

Relying on occurrence data as a proxy for habitat quality is not without problems because ecological traps (Dwernychuk & Boag 1972) can cause species to persist in poor quality habitat. Ecological traps occur when the cues used by animals to select optimal habitat no longer accurately reflect habitat quality so the animals preferentially select poor quality habitat (Schlaepfer et al. 2002). For instance, linear habitat corridors that are poor quality can help maintain local persistence, but can act as ecological traps and ultimately drain the metapopulation (Henein & Merriam 1990). Linear habitat elements were often used by robins in our study region, but it is unclear if these elements were good quality habitat or sinks. Comparing reproductive success in the linear elements to that in more expansive habitat patches would help clarify if these corridors were acting as ecological traps (Battin 2004).

3.5.1 Scale of woodland effect

Habitat extent and accessibility at a larger scale, likely to be relevant for robin dispersal movements, were not good predictors of robin occupancy. Instead, robins appear to be responding primarily to woodland extent at a scale relevant to movements within a home range. Habitat fragmentation disrupts multiple processes over progressively larger spatiotemporal scales (Lindenmayer & Fischer 2006). Movement restrictions within a home range can limit resource access, at intermediate scales demographic exchange is impeded, while at large scales gene flow is inhibited (Lindenmayer & Fischer 2006; Hinam & St. Clair 2008). Our results suggest the home range scale of effect is the first to become apparent.

Habitat extent at larger spatial scales could be increasingly important for occupancy over time as the effects from restricted demographic exchange and impeded gene flow become more prominent.

53 For example, woodlands in Victoria, Australia, have undergone heavy clearing over the past 160 years (ECC 1997). There, Harrison et al. (2012) found support that eastern yellow robin populations were male-biased in fragmented regions, but not in landscapes with continuous woodland. The cause of this bias is unclear; perhaps females preferentially chose mates in contiguous landscapes or were unable to disperse to isolated patches. While there was evidence of impediments to robin gene flow at fine scales, this effect was less supported at regional scales (Harrisson et al. 2012).

3.5.2 Management implications

Habitat loss is likely to be having a greater impact on robins within the study area than habitat fragmentation and robins generally did not occupy patches narrower than 75 m. Restoration actions could capitalise on the existing strips of woodland along roadsides by restoring narrow bands of brigalow inside fence lines, especially where linear patches intersect (Hall et al. 2016). Together, the resulting linear blocks appear to provide enough habitat to sustain robins, especially when there is more surrounding woodland. We also recommend that travelling stock routes be retained and protected (Lentini et al. 2011b). These contained enough habitat in many places to support robin territories and probably played an important role in reducing the impacts of habitat fragmentation.

Our study highlights the importance of repeatedly monitoring populations so important changes to dynamics can be identified before population declines are realised. Using multiple methods, such as condition and occurrence, to identify threats to populations can provide a more comprehensive assessment than occurrence alone. While we did not definitively demonstrate that patterns of physiological measures could generate meaningful predictions of population changes, this line of research warrants further investigation given its potential value to conservation.

54 CHAPTER 4: Determinants of food availability for a ground-foraging insectivore differ when the nutritional value of prey is considered

This chapter will be submitted to Emu. Approximate percentage author contributions as follows: Anita Cosgrove – 83% Todd McWhorter – 2% Martine Maron – 15%

Plate 4. Moisture is of critical importance in this harsh environment. Gilgai depressions collect water after rain.

55 4.1 ABSTRACT

Habitat loss and fragmentation are causing declines in the abundance and occurrence of Australian woodland-dependent birds. However, little progress has been made towards understanding the mechanisms driving these declines. Insectivores are particularly vulnerable compared to other foraging guilds, suggesting that reduced arthropod availability may be one such mechanism. However, it may be misleading to assess food availability based on total invertebrate abundance since prey vary in nutritional value and birds exhibit prey preferences. We conducted invertebrate pitfall trapping across 25 sites in eastern Australia to examine how eastern yellow robin (Eopsaltria australis) prey availability was related to characteristics of the habitat, landscape, and local climatic conditions, and how these relationships changed once the dominant, but less-nutritious, Formicidae (ants) were excluded. Robin populations are more vulnerable to declines at sites with less surrounding woodland, so we also assessed how the density of individual prey groups was related to woodland extent. When considering all arthropod taxa, prey abundance and weight were higher in landscapes with less woodland and fewer patches. However, this result was primarily driven by Formicidae responses. Once Formicidae were excluded, soil moisture and, to a lesser degree, grass cover showed the strongest associations with prey availability. Formicidae and Coleoptera were generally inversely related to woodland extent, while Orthoptera showed a positive association. In contrast to the hypothesis that smaller habitat patches contain less invertebrate prey per unit area than larger patches, our results suggest that this avian insectivore does not experience reduced food availability in less-wooded landscapes. However, those landscapes generally contained long, narrow patches which could mean higher energy costs for the birds while foraging or defending their territory. It is unclear whether the additional Formicidae inhabiting less-wooded landscapes could compensate for any potential shortfalls in energy requirements over the long-term, given their low nutritional value.

4.2 INTRODUCTION

Australian woodland birds are under threat. Habitat loss and fragmentation, combined with land-use intensification, have placed enormous pressure on woodland-dependent biota resulting in reduced abundance and occurrence (Ford et al. 2001; Barrett et al. 2007; Vié et al. 2009). Even previously common, widespread taxa are in decline (Szabo et al. 2011; Stevens & Watson 2013) which could have alarming knock-on effects on both the structure and function of ecosystems (Gaston & Fuller 2008).

56 Population changes have not been consistent within species or across regions (Barrett et al. 2003; Maron & Lill 2006), so it has been difficult to infer the underlying causes of the observed patterns. To date, declines have generally been most pronounced in the temperate eucalypt woodlands of southern Australia (Watson 2011). Habitat loss, fragmentation, or degradation form the basis of most of the mechanisms hypothesised to be driving the population declines (Ford et al. 2001; Watson 2011; Maron et al. 2013), although drought and climate change have also been suggested (Mac Nally et al. 2009; Rayner et al. 2014). Researchers have made little progress towards testing the proposed mechanisms (but see Cooper & Walters 2002; Stevens & Watson 2013) despite investing much effort into identifying spatial, temporal, and ecological patterns of avian distribution and abundance (Barrett et al. 2003; Maron et al. 2005; Robertson et al. 2013).

One common pattern is that insectivores appear to be particularly vulnerable to declines (Bregman et al. 2014). This pattern is not limited to the fragmented woodlands of Australia; it has also been observed in the rainforests of South America and Asia (Kattan et al. 1994; Castelletta et al. 2000; Laurance et al. 2011), the land-bridge islands of Malaysia (Yong et al. 2011), the savannahs of Tanzania (Sinclair et al. 2002), and across North America (Nebel et al. 2010). This propensity of avian insectivores to decline at rates disproportionately greater than that of other foraging guilds suggests that reduced prey availability or accessibility may be an underlying cause of declines (Razeng & Watson 2012; Paquette et al. 2014).

Several mechanisms could drive prey reductions in the fragmented woodlands of eastern Australia, many of which are embedded in an agricultural matrix. For instance, farming practices may influence arthropod populations by altering the availability of water and nutrients as well as the physical, chemical, and biological characteristics of the soil (Watson 2011). Additionally, pesticide use in agricultural regions can deplete invertebrate communities, including non-targeted species (Hallmann et al. 2014; Sanchez-Bayo & Goka 2014). Some invertebrate taxa may be less abundant in edge habitat due to the accompanying desiccating microclimate changes such as increased temperature, solar radiation, or wind speed (Didham et al. 1998; Laurance et al. 2011), or because both woodland-dependent birds and habitat generalists occupy edges, thereby increasing the predation pressure (Barbaro et al. 2012).

Invertebrate abundances in agricultural regions of Australia are not only influenced by present-day factors; historical conditions can also have a lasting impact. In the past, farmers preferentially cleared woodlands that grew on the most productive soils or that were located near a reliable water source (Fensham and Fairfax 2003; Lunt and Spooner 2005; Watson 2011; Simmonds et al. 2017).

57 This non-random pattern of clearing means that the woodlands that remain today are likely to be on less-productive soils and may actually be relatively marginal habitat for woodland-dependent species (Watson 2011). Additionally, land-use legacies play an integral role in shaping present-day communities (Lunt and Spooner 2005; Perring et al. 2016). For instance, a dense woody understory can form on land that was previously heavily grazed but now has little or no grazing (Lunt and Spooner 2005). This type of change to vegetation structure would almost certainly affect the composition of the associated invertebrate community (Bromham et al. 1999). Similarly, past fires have a major influence on future flora and fauna communities (Ross et al. 2002; Moretti et al. 2004; Lunt and Spooner 2005; Croft et al. 2010).

It is hypothesised that, for their size, smaller habitat patches contain less invertebrate prey than larger patches (Burke & Nol 1998; Zanette et al. 2000). However, this view is based on a limited number of studies (Burke & Nol 1998; Zanette et al. 2000), some of which were not well-replicated at the level of habitat patch (Zanette et al. 2000; Şekercioğlu et al. 2002). Studies of total invertebrate abundance or weight produce mixed results with positive (Burke & Nol 1998; Zanette et al. 2000), negative (Moreno et al. 2013), or no relationship (Şekercioğlu et al. 2002) identified between total invertebrate biomass and patch size.

However, total invertebrate biomass may not accurately represent the food available to the birds and drawing inferences from this type of measure may be misleading (Razeng & Watson 2012). Birds show preferences for particular prey taxa (Naef-Daenzer et al. 2000; Hagar et al. 2007; Moorman et al. 2007) which vary in nutritional value (Robel et al. 1995; Razeng & Watson 2015). Additionally, birds may exhibit specific foraging behaviours that target certain prey taxa, such as foraging within particular strata (Antos & Bennett 2006; Moorman et al. 2007). Furthermore, invertebrate taxonomic groups respond idiosyncratically to patch area (Didham 1997; Vasconcelos & Bruna 2012). Therefore, we could derive a better understanding of how insectivore food availability is related to habitat extent from examining invertebrate responses to the landscape in conjunction with prey preferences.

The eastern yellow robin (Eopsaltria australis) provides an excellent opportunity for examining relationships between food availability and landscape configuration. This is a small (approximately 20 g), sedentary, woodland-dependent insectivore that avoids open habitats (Higgins & Peter 2002). Although these robins are considered to be common and have a Red List status of Least Concern, their populations have declined in parts of Victoria and New South Wales (Debus 2006; Stevens & Watson 2013; BirdLife Australia 2015b). This species is sensitive to both patch area and patch

58 isolation in agricultural regions (Barrett et al. 1994; Watson et al. 2001; Watson et al. 2005). Eastern yellow robin populations are more vulnerable in areas with less surrounding woodland; they had elevated chronic stress levels in less-wooded landscapes and woodland extent became the strongest predictor of site occupancy within 4 years (Maron et al. 2012; Chapter 3). Zanette et al. (2000) used total invertebrate volume and dry weight to determine that smaller patches contained less prey biomass for eastern yellow robins than larger patches. However, this analysis was based on samples collected from just two small and two large patches (Zanette et al. 2000).

Here, our objective was to identify how eastern yellow robin prey availability varied with characteristics of the habitat, landscape, and local climatic conditions. If robin populations were more vulnerable to declines in landscapes with less woodland because there was less available food, we would have expected arthropod abundance to be positively correlated with woodland extent. Second, we assessed whether these correlates of arthropod abundance changed once the dominant, but less-nutritious, Formicidae (ants) were excluded. Out of the eight prey taxa most commonly eaten by Australian woodland-dependent birds, Formicidae contains the lowest proportion of crude fat and relatively low levels of crude protein and moisture (Razeng & Watson 2015). Finally, we determined how abundances within individual prey groups, especially those that have a higher nutritional value, were associated with woodland extent.

4.3 METHODS

4.3.1 Study area

Study sites were clustered around Meandarra (-27.32°S, 149.88°E), Moonie (-27.84°S, 150.12°E), and Goondiwindi (-28.40°S, 150.47°E) in the Brigalow Belt South bioregion of Queensland, Australia (Figure 4.1). The region experiences a subtropical climate with annual rainfall of about 600 mm, a large portion of which falls between December and February each year (Bureau of Meteorology 2015).

The study was conducted in woodlands dominated by brigalow (Acacia harpophylla) on the low, fertile plains in the region. Belah (Casuarina cristata), Eucalyptus spp., and wilga (Geijera parviflora) shrubs can also feature prominently in these woodlands. Although the region has undergone clearing for pasture and cropping since the late 1800s, the suckering nature of brigalow and its characteristically dense regrowth deterred farmers from extensively removing this habitat type (Seabrook et al. 2006). Brigalow deforestation only began in earnest during the 1960s

59 (Seabrook et al. 2006) and today, less than 4% of the study region is covered by remnant brigalow woodland which is currently listed as nationally endangered (Commonwealth of Australia 2016).

4.3.2 Study species

Eastern yellow robins are ground-foraging insectivores that employ a sit and wait foraging strategy, perching on low-hanging branches or tree trunks from which they pounce on ground-dwelling prey (Higgins and Peter 2002). Over 70% of all prey is taken from ground level (Recher et al. 1985; Ford et al. 1986). These robins primarily eat arthropods, but have occasionally been seen eating seeds, fungi, skinks, and grit (Higgins & Peter 2002). The most frequently reported prey in decreasing order of reporting frequency are: Coleoptera; Formicidae; Diptera; Lepidoptera; Hemiptera; other Hymenoptera; Araneae; and Orthoptera (Razeng & Watson 2012).

Figure 4.1. Location of the study region in the Brigalow Belt South bioregion, Queensland, Australia. Black circles denote study sites.

60 4.3.3 Site selection and explanatory variables

Twenty-seven 2 ha sites were selected from within the study region for sampling. Sites were only selected if they were accessible, occupied by eastern yellow robins, comprised of brigalow- dominant or -codominant woodland, and had tree density > 150 stems per ha to avoid highly degraded sites. All sites were located at least 800 m apart to reduce the risk of spatial autocorrelation.

Four 20 m x 20 m quadrats were positioned randomly within each site. We counted the number of trees (“no. trees”) and the number of shrubs over 30 cm high (“no. shrubs”) within each quadrat. In the four corners of each quadrat, we placed a 1 m2 microquadrat (16 microquadrats total per site) and assessed the proportion of the microquadrat that was covered by live or standing dry grass (“grass cover”). We ran a 20 m transect along the centre of each quadrat and counted the number of pieces of fallen timber greater than 7 cm in diameter that lay on or across the transect (“fallen wood”). At the 5 m, 10 m, and 15 m marks along the transect, we measured the depth of the leaf litter down to the soil surface to the nearest mm (“leaf litter depth”) and the percentage of volumetric soil water content (“soil moisture”; MPM160 Moisture Probe Meter). All habitat variables were averaged across each site before being included in the analyses.

Landscape variables were calculated from a map of total woodland extent in ArcGIS 10.2.1 (ESRI 2014). We created the map from the Statewide Landcover and Trees Study Foliage Projective Cover 2011 dataset which had a spatial resolution of 25 m (State of Queensland 2014b). Any type of woodland community with a minimum foliage projective cover of 8% and a minimum patch size of 2 ha was classified as woodland. We placed a buffer with a 500 m radius around the centre of each site and calculated the proportion of area within the buffer that was covered in woodland (“proportion of woodland”) and counted the number of woodland patches within each buffer (“no. patches”). The 500 m radius was selected because it would encompass the home range of the robins inhabiting each site (approximately 5-6 ha; Debus 2006), regardless of whether the site was positioned at the centre or edge of the home range. The area and width of each patch containing a site were also calculated, but these were highly correlated with the proportion of woodland (r > 0.90) so were excluded from further analyses.

61 4.3.4 Arthropod pitfall trapping

Arthropod pitfall traps were dug in along three of the four transects at each site from August 2013 to January 2015. Three traps were positioned along each transect at 5 m intervals (nine traps total per site). Traps consisted of two round, straight-sided, plastic containers approximately 12 cm in diameter and 10 cm deep placed inside each other with a removable lid. Using two containers allowed us to lift a trap out of the ground if needed without disturbing the soil. Traps were dug in so their lip was level with the surrounding soil. Lids were affixed to close the trap, then they were left for one week to help reduce digging-in effects on captures (Greenslade 1973).

After allowing the trap to settle for one week, we added approximately 2 cm of water with a drop of detergent. Additional water was added to the traps during the activation period as needed. Small pieces of wood found in the surrounding area were placed in the centre of the trap to prevent inadvertent vertebrate drownings. Each trap was opened each morning and closed each evening for a period of five consecutive days to ensure only diurnally-active arthropods were captured. Arthropods were removed from the traps each day and stored in 70% ethanol.

Although we aimed to keep each trap open for about 8-9 hr each day, this was not always possible due to logistical limitations. To account for differences in trap hours, we calculated trapping effort as the sum total for the site of the number of hours each trap was open and operational. Occasionally traps were left open overnight when they were inaccessible due to wet weather. Additionally, some traps were disturbed by wildlife (probably feral pigs, Sus scrofa) during the five day activation period and rendered inoperative. In both cases, we excluded all captures from the affected days from the analyses and excluded these days from the trapping effort.

Arthropods caught from each site were sorted by taxonomic group using a dissecting microscope (Leica EZ4HD) and dichotomous key (State of Queensland 2016b). Taxa were grouped in accordance with Razeng and Watson (2012, 2015); generally to order, although Formicidae were categorised separately from other Hymenoptera because ants formed the bulk of total captures and are ecologically distinct from bees and wasps (Cooper et al. 1990). Robins had difficulty eating mealworms > 40 mm long (Anita Cosgrove, personal observation), so we excluded all arthropods < 1 mm and > 40 mm. We counted the number of arthropods (“no. arthropods”) caught per taxonomic group per site then obtained a dry weight (accurate to the nearest 0.0001 g) after they had been dried in a 60 °C oven for 48 hr (“arthropod weight”).

62 We used the summed total counts or weights of arthropods per site in our analyses. However, using a total per site could potentially mask fine-grained effects at smaller scales. Resources are unlikely to be uniformly distributed across sites. Rather, prey could be clumped with robins preferentially foraging from microhabitats supporting higher prey densities (Cousin 2007). Thus, we also calculated the maximum number (“maximum number of arthropods”) and weight of arthropods (“maximum arthropod weight”) excluding Formicidae caught in any one trap at each site.

4.3.5 Statistical analyses

We used an information-theoretic approach to assess the relative support for alternative general and generalised linear models of arthropod numbers and weights, both including and excluding Formicidae, as well as for the maximum number and weight of arthropods excluding Formicidae caught in any one trap at each site (Burnham & Anderson 2002). The arthropod weight measures were natural logarithm transformed to improve normality. All explanatory variables were standardised to a mean of zero and a standard deviation of 0.5 with the arm package in R to allow comparison of parameter estimates (Gelman 2008; Gelman & Su 2015). We controlled for the north-south spread of sites by incorporating northing into every candidate model, including the null models. Variation in sampling effort was controlled by adding log (trap effort) as an offset in all models.

Candidate models were determined a priori to reflect alternative habitat, landscape, or climatic factors that may affect arthropod populations and the two response variables were examined separately using the same candidate explanatory variable sets (Table 4.1). Variables were grouped into five categories of interest, each of which could be related to arthropod biomass: ground cover (Bromham et al. 1999); vegetation structure (García et al. 2011); landscape structure (Burke & Nol 1998); short-term weather (Levings & Windsor 1984); and longer-term seasonal effects (Bell 1985). For each category, models of every possible combination of terms related to that category were included. We also generated a null model that included only northing and the offset term. This process generated a total of 16 candidate models to be used in each analysis of total arthropods, total arthropods excluding Formicidae, and maximum arthropods excluding Formicidae.

Generalised linear models of the number of arthropods were overdispersed when using Poisson distributions, so were instead modelled with negative binomial distributions and log links in the MASS package (Venables & Ripley 2002). Arthropod weights were modelled using linear regressions. We used Akaike’s information criterion corrected for small samples sizes (AICc) to

63 Table 4.1. A priori models developed to assess the habitat, landscape, and climatic factors associated with counts and dry-weights of arthropod prey targeted by eastern yellow robins. All models included northing as a fixed effect and log (trap effort) as an offset term. Category Models Ground cover Grass cover + Leaf litter depth + Fallen wood Grass cover + Leaf litter depth Grass cover + Fallen wood Leaf litter depth + Fallen wood Grass cover Leaf litter depth Fallen wood Vegetation structure No. trees + No. shrubs No. trees No. shrubs Landscape structure Proportion of woodland + No. patches Proportion of woodland No. patches Short-term weather Soil moisture Seasonal effects Ordinal date Null model Null determine the most parsimonious model from each candidate set of models, whereby the highest ranked models have the lowest AICc (Burnham & Anderson 2002).

Model comparisons were based on the differences in AICc values (ΔAICc); when ΔAICc values were > 2 units, the model with the lower AICc was considered to have substantially more empirical support (Burnham & Anderson 2002). Akaike weights (ωi) were used to determine the probability that a particular model was the best model given the observed data and the candidate model set (Burnham & Anderson 2002). All maximised log-likelihood and AICc calculations were performed in the AICcmodavg package (Mazerolle 2016). Goodness of fit for the number of arthropod models was assessed with Nagelkerke’s R2 in the fsmb package (Nagelkerke 1991; Nakazawa 2015), whereas the fit of arthropod weight models was examined using adjusted R2.

We fitted a series of regression models to the abundances of specific taxonomic groups to assess their relationships with the proportion of woodland. Sufficient data were available to examine the

64 effects on Araneae, Coleoptera, Diptera, Formicidae, Hemiptera, other Hymenoptera, and Orthoptera abundances. Northing and an offset term for log (trap effort) were included in each model. We compared negative binomial, quadratic, and loess regressions against the null models. The negative binomial, quadratic, and null generalised linear models were built with a negative binomial error distribution in the MASS package (Venables & Ripley 2002), while general additive models with a negative binomial distribution from the gam package (Hastie 2015) were used to create the loess models. We used AICc values to compare the performance of each model within each taxon and Nagelkerke’s R2 to evaluate goodness of fit (Nagelkerke 1991; Burnham & Anderson 2002; Nakazawa 2015).

Spline correlograms from the ncf package were used to assess spatial autocorrelation, both within the response variables and in the model residuals (Bjornstad 2015). We used a pairwise matrix of Pearson correlation coefficients to confirm that none of the explanatory variables had r > |0.50| and all variance inflation factors derived from the final models using the car package were < 3, confirming there was no multicollinearity (Zuur et al. 2010; Fox & Weisberg 2011). Residual scatter plots were examined to check the assumptions of normality and homoscedasticity.

All analyses were performed in the R statistical package (R Core Development Team 2015). Graphs were created with the ggplot2 package (Wickham 2009). Results are reported as mean ± standard error.

4.4 RESULTS

A total of 27,054 arthropods within the specified size range were caught across 27 sites (Table 4.2). Additionally, five vertebrates (all Scincidae) were trapped, along with multiple Acari, Collembola, Diplopoda, and Thysanoptera specimens, all of which were excluded for being over- or under-sized. We excluded Blattodea, Isopoda, Isoptera, Mantodea, Mecoptera, Pscoptera, Thysanura, and Scorpiones from further analyses because there is no evidence in the literature that eastern yellow robins consume these taxa (Higgins & Peter 2002). It is plausible that robins eat some of these taxa on rare occasions (particularly Blattodea, Isoptera, and Mantodea), but the prevalence of those three orders in the traps were so low that including them made little difference to the results. We included four unidentified larval individuals that were probably Lepidoptera or Coleoptera. Thus, the following groups were considered likely to be eaten by robins and included in the analyses: Araneae, Coleoptera, Diptera, Formicidae, Hemiptera, other Hymenoptera, Lepidoptera, Orthoptera, and unidentified larvae.

65 Table 4.2. Number of arthropods caught in pitfall traps in brigalow-dominated woodlands. Only those taxonomic groups likely to be eaten by eastern yellow robins were included in further analyses. Taxonomic group n % of total captures Included in further analyses Araneae 329 1.22 Yes Blattodeaa 2 0.01 No Coleoptera 76 0.28 Yes Diptera 805 2.98 Yes Formicidae 21,699 80.21 Yes Hemiptera 150 0.55 Yes Other Hymenoptera 84 0.31 Yes Isopodaa 3,842 14.20 No Isopteraa 9 0.03 No Lepidoptera 20 0.07 Yes Mantodeaa 3 0.01 No Mecopteraa 1 < 0.01 No Orthoptera 25 0.09 Yes Psocopteraa 1 < 0.01 No Scorpionesa 1 < 0.01 No Thysanuraa 3 0.01 No Unidentified larvae 4 0.01 Yes Total 27,054 a Excluded from further analyses because eastern yellow robins are not known to eat this taxon.

One site was excluded from all analyses because there was an exceptionally large number of arthropods caught (8,003 arthropods, primarily Formicidae, with a total weight of 2.4076 g) that exerted substantial leverage on the results. A second site was excluded because traps were inaccessible due to poor weather and left open each night for the duration of the trapping period. Excluding those sites and including only groups likely to be eaten by robins, the number of arthropods per site ranged from 53 to 2,844 (608 ± 128, n = 25) and total arthropod dry weights per site ranged from 0.1159 to 1.6842 g (0.4329 ± 0.0763 g, n = 25).

66 4.4.1 All arthropod taxa combined

There was strong support that the best candidate model for the total number of arthropods incorporated both the proportion of woodland and number of patches (Table 4.3). Model fit was good with an R2 of 0.71. Confidence intervals for both landscape metrics did not include zero (Table 4.4). Arthropod numbers were higher at sites with less surrounding woodland area and fewer woodland patches (Figure 4.2a; Figure 4.2b).

Similarly, there was strong evidence that the variance in total arthropod weight was best explained by the model of proportion of woodland and number of patches (Table 4.5). This model had moderate fit with an R2 of 0.39. The model of proportion of woodland also received strong support over the null model (Table 4.5). Again, the confidence intervals for these landscape-scale effects did not cross zero and both measures were negatively associated with arthropod weight (Table 4.4).

a 2 Model K Dev ΔAICc ωi R Northing + Proportion of woodland + No. patches 5 350.12 0.00 0.60 0.71 Northing + No. patches 4 357.58 4.29 0.07 0.48 Northing + Proportion of woodland 4 357.72 4.44 0.07 0.47 Northing + No. shrubs 4 358.16 4.88 0.05 0.46 Northing + Soil moisture 4 359.24 5.95 0.03 0.41 Northing + Grass cover + Leaf litter depth 5 356.38 6.25 0.03 0.52 Northing (null model) 3 362.50 6.37 0.02 0.27 Northing + Fallen wood 4 359.72 6.44 0.02 0.39 Northing + Grass cover 4 359.74 6.46 0.02 0.39 Northing + Grass cover + Fallen wood 5 356.84 6.72 0.02 0.51 Northing + Leaf litter depth 4 360.54 7.26 0.02 0.36 aThe lowest AICc value was 363.28.

67 Table 4.4. Coefficients, standard errors, and 95% confidence intervals from the best approximating models identified through AICc model comparison techniques examining environmental variability in the number and weights of arthropods caught in pitfall traps (n = 25). Confidence intervals denoted in bold do not cross zero. Model Coefficient SE Lower CI Upper CI Total arthropod numbers Proportion of woodland -0.7766 0.2788 -1.3442 -0.2103 Number of patches -0.7837 0.2563 -1.2490 -0.2368 Total arthropod weight Proportion of woodland -0.7709 0.2885 -1.3708 -0.1709 Number of patches -0.6123 0.2645 -1.1624 -0.0623 Arthropod numbers excl. Formicidae Soil moisture 0.9154 0.2745 0.3651 1.4662 Arthropod weight excl. Formicidae Grass cover -0.9504 0.4534 -1.8906 -0.0101 Maximum number of arthropods excl. Formicidae Grass cover -0.4717 0.2541 -0.9436 0.0095 Leaf litter depth -0.6506 0.2617 -1.1527 -0.1610 Soil moisture 0.7823 0.3241 0.1262 1.4435

68 a)

Figure 4.2. Scatterplots with 95% confidence intervals (grey shading) of the factors explaining the greatest variation in the abundance of arthropod captures. Negative binomial regressions of a) total number of arthropods by the proportion of woodland within 500 m of each site, b) total number of arthropods by the number of woodland patches within 500 m of each site, and c) the number of arthropods excluding Formicidae by the soil moisture (%).

69 Table 4.5. Model selection results from a priori models examining the effects of environmental factors on dry weights of arthropods caught in pitfall traps at sites (n = 25) within brigalow- dominated woodlands. The table displays number of parameters (K), deviance (Dev), AICc 2 differences (ΔAICc), Akaike weights (ωi), and adjusted R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod dry weight than the null model. Models with ωi < 0.01 are omitted from the table.

a 2 Model K Dev ΔAICc ωi R Northing + Proportion of woodland + No. patches 4 51.40 0.00 0.47 0.39 Northing + Proportion of woodland 3 57.08 2.53 0.13 0.28 Northing + No. patches 3 58.72 4.16 0.06 0.28 Northing + Fallen wood 3 58.72 4.17 0.06 0.27 Northing + Grass cover 3 58.98 4.42 0.05 0.25 Northing + No. shrubs 3 59.12 4.57 0.05 0.21 Northing + Grass cover + Fallen wood 4 55.98 4.59 0.05 0.29 Northing (null model) 2 62.14 4.73 0.04 0.21 Northing + Grass cover + Leaf litter depth 4 58.16 6.76 0.02 0.25 aThe lowest AICc value was 64.55.

4.4.2 All arthropod taxa excluding Formicidae

Soil moisture received strong support as the best predictor of the total number of arthropods once Formicidae were excluded (Table 4.6). Model fit was good with an R2 of 0.48. The 95% confidence intervals for soil moisture did not include zero and arthropod numbers excluding Formicidae were higher when soil was moister (Table 4.4; Figure 4.2c). Additionally, ordinal date received weak support over the null model with arthropod counts growing larger later in the year (Table 4.6).

There was uncertainty over the best model of total arthropod weights excluding Formicidae. The model of grass cover received the strongest weighting, but it performed only moderately better than the null model (Table 4.7). The model of grass cover had moderate fit with an R2 of 0.28. Grass cover had a negative association with arthropod weight excluding Formicidae (Table 4.4). The soil moisture model performed poorly (Table 4.7).

The most parsimonious models of the maximum number of arthropods excluding Formicidae included grass cover, leaf litter depth, and soil moisture (Table 4.8). Model fits were reasonable with the model containing grass cover and leaf litter depth having an R2 of 0.51, while the soil

70 moisture model had an R2 of 0.37. The confidence intervals for leaf litter depth and soil moisture did not cross zero; although, those for grass cover did include zero (Table 4.4). Both grass cover and leaf litter depth were negatively related to arthropod numbers, while soil moisture showed a positive relationship.

All of the chosen a priori models performed poorly when considering the maximum weight of arthropods excluding Formicidae (Table 4.9). Number of patches received negligibly more support than the null model (ΔAICc=0.14), but explained little of the variance in arthropod weights with an R2 of 0.20.

Table 4.6. Model selection results from a priori models examining the effects of environmental factors on the total number of arthropods (excluding Formicidae) caught in pitfall traps at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance 2 (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod abundance (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table.

a 2 Model K Dev ΔAICc ωi R Northing + Soil moisture 4 229.32 0.00 0.74 0.48 Northing + Ordinal date 4 235.00 5.67 0.04 0.22 Northing + No. patches 4 235.42 6.10 0.04 0.20 Northing + Leaf litter depth 4 235.84 6.52 0.03 0.17 Northing (null model) 3 238.86 6.67 0.03 < 0.01 Northing + Grass cover 4 236.02 6.69 0.03 0.16 Northing + Grass cover + Leaf litter depth 5 232.92 6.76 0.03 0.32 Northing + Grass cover + Leaf litter depth + Fallen wood 6 230.10 7.45 0.02 0.44 aThe lowest AICc value was 239.32.

71 Table 4.7. Model selection results from a priori models examining the effects of environmental factors on total dry weights of arthropods (excluding Formicidae) caught in pitfall traps at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance 2 (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod dry weight (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table.

a 2 Model K Dev ΔAICc ωi R Northing + Grass cover 4 68.78 0.00 0.19 0.28 Northing + Grass cover + Leaf litter depth 5 66.00 0.38 0.16 0.33 Northing + Proportion of woodland + No. patches 5 66.86 1.24 0.10 0.29 Northing + Proportion of woodland 4 70.34 1.57 0.09 0.23 Northing (null model) 3 73.32 1.69 0.08 0.20 Northing + No. patches 4 70.94 2.17 0.06 0.24 Northing + Leaf litter depth 4 71.22 2.45 0.06 0.24 Northing + Soil moisture 4 71.44 2.67 0.05 0.21 Northing + Grass cover + Leaf litter depth + Fallen wood 6 65.00 2.89 0.05 0.34 Northing + Grass cover + Fallen wood 5 68.76 3.15 0.04 0.25 Northing + No. trees 4 72.02 3.25 0.04 0.22 Northing + No. shrubs 4 73.08 4.30 0.02 0.19 Northing + Fallen wood 4 73.20 4.42 0.02 0.17 Northing + Ordinal date 4 73.20 4.43 0.02 0.17 aThe lowest AICc value was 78.77.

72 Table 4.8. Model selection results from a priori models examining the effects of environmental factors on the maximum number of arthropods (excluding Formicidae) caught in any one pitfall trap at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters 2 (K), deviance (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod abundance (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table.

a 2 Model K Dev ΔAICc ωi R Northing + Grass cover + Leaf litter depth 5 167.32 0.00 0.30 0.51 Northing + Leaf litter depth 4 171.18 0.69 0.21 0.38 Northing + Soil moisture 4 171.52 1.04 0.18 0.37 Northing + Grass cover + Leaf litter depth + Fallen wood 6 167.30 3.48 0.05 0.55 Northing + Grass cover 4 173.98 3.50 0.05 0.25 Northing + Leaf litter depth + Fallen wood 5 171.04 3.70 0.05 0.39 Northing + Fallen wood 4 174.62 4.14 0.04 0.22 Northing (null model) 3 177.72 4.38 0.03 0.04 Northing + Grass cover + Fallen wood 5 172.28 4.94 0.03 0.34 Northing + No. patches 4 176.32 5.84 0.02 0.12 aThe lowest AICc value was 180.49.

73 Table 4.9. Model selection results from a priori models examining the effects of environmental factors on the maximum dry weight of arthropods (excluding Formicidae) caught in any one pitfall trap at sites (n = 25) within brigalow-dominated woodlands. The table displays number of parameters (K), deviance (Dev), AICc differences (ΔAICc), Akaike weights (ωi), and Naglekerke R2 values. Models denoted in bold have substantial empirical support that they explain more of the variation in arthropod dry weight (excluding Formicidae) than the null model. Models with ωi < 0.01 are omitted from the table.

a 2 Model K Dev ΔAICc ωi R Northing + No. patches 4 57.42 0.00 0.16 0.20 Northing (null model) 3 60.42 0.14 0.14 0.17 Northing + Leaf litter depth 4 58.30 0.89 0.10 0.21 Northing + Grass cover 4 58.70 1.29 0.08 0.18 Northing + Proportion of woodland + No. patches 5 55.58 1.32 0.08 0.21 Northing + Soil moisture 4 58.86 1.45 0.08 0.20 Northing + Fallen wood 4 59.06 1.65 0.07 0.17 Northing + Proportion of woodland 4 59.26 1.84 0.06 0.16 Northing + Grass cover + Leaf litter depth 5 56.32 2.89 0.06 0.16 Northing + No. trees 4 60.30 2.06 0.04 0.12 Northing + Ordinal date 4 60.32 2.87 0.04 0.14 Northing + No. shrubs 4 60.38 2.91 0.04 0.13 Northing + Grass cover + Fallen wood 5 57.64 2.96 0.03 0.18 Northing + Leaf litter depth + Fallen wood 5 58.08 3.39 0.02 0.18 aThe lowest AICc value was 67.42.

4.4.3 Responses by individual arthropod taxa to woodland extent

Three of the taxonomic groups we examined were associated with the proportion of woodland. The number of Formicidae and Orthoptera were each best explained by the negative binomial regression models, while the variation in the number of Coleoptera was best explained by the loess model (Figure 4.3). The number of Formicidae exhibited an inverse relationship to the proportion of woodland while the number of Orthoptera was positively associated. The number of Coleoptera was highest at very low levels of woodland cover and fell as woodland extent increased. Coleoptera numbers increased somewhat at very high levels of woodland extent. For all other taxa (i.e. Araneae, Diptera, Hemiptera, and other Hymenoptera), the null model had the lowest AICc value.

75 4.5 DISCUSSION

When considering total measures for all arthropod taxa, robin prey availability was related to the extent and fragmentation of woodland within the surrounding landscape. Both arthropod abundance and weight were higher in landscapes with less woodland and less habitat fragmentation. However, after excluding the less-nutritious Formicidae, the picture changed dramatically – soil moisture became the dominant factor, with moister soils promoting greater prey availability. Grass cover also played a role whereby grassier sites harboured a lower biomass of arthropods. Similar results were obtained when considering the maximum number of arthropods excluding Formicidae. Traps surrounded by more grass cover and leaf litter caught fewer arthropods, while traps in moister localities caught greater numbers. Out of all the prey groups frequently eaten by robins, only Formicidae, Coleoptera, and Orthoptera showed a strong relationship between abundance and surrounding woodland extent, but this relationship varied in its direction.

4.5.1 Arthropod responses to landscape structure

Surprisingly, arthropod biomass was higher in areas with less surrounding woodland. This pattern was the opposite of that predicted if robin populations were more vulnerable in less-wooded landscapes due to reduced prey availability. However, this result was driven largely by the dominance of Formicidae in the samples and can be better understood by examining the samples in closer detail.

Although we did not sort catches to species level, many of the Formicidae were meat ants (Iridomyrmex purpureus) or Camponotus spp. (AntWiki 2016). Meat ant distribution is limited by dense shade and dense ground cover (Greenslade 1974). In the study region, patches with less surrounding woodland also tended to be narrow. Narrow patches were always bordered by roads, pasture, or crop fields, had abrupt edges with high exposure to sunlight, and tended to be grazed by livestock. Several of the narrow patches were quite disturbed with a relatively open canopy and bare ground. Thus, sites with less surrounding woodland likely provided more-favourable conditions for meat ants than more-contiguous woodland. Camponotus spp. show diverse responses to disturbances such as grazing (Hoffmann & Andersen 2003) so it is difficult to draw conclusions without knowing the species present in our samples. However, Bromham et al. (1999) examined pitfall captures of the Camponotus genus in the eucalypt woodlands of Victoria and found they were relatively more abundant in grazed woodlands than ungrazed woodlands.

76 While over 78% of our total captures were Formicidae, actual consumption of this taxon by robins is likely to be lower. There is doubt over whether Formicidae are a preferred prey item, or whether they are eaten opportunistically at times of peak physiological demand or to supplement the diet when more-nutritious food is difficult to find (Poulin & Lefebvre 1996; Razeng & Watson 2015). Formicidae were only noted in 24% of all foraging and dietary records for eastern yellow robins (Razeng & Watson 2012) and gut content analysis indicates that Hymenoptera (ants, bees, and wasps) formed just over 40% of the robins’ diet (Haylock 1985; Higgins & Peter 2002). The majority of our pitfall trapping was conducted between September and January when warmer temperatures would have promoted greater arthropod activity likely making more preferred prey groups easier to find (Saska et al. 2013). Prey preference is likely to play a role, so Formicidae probably formed a much lower proportion of the robins’ diet than that indicated by their overall abundance in the habitat. Thus, inferring how landscape configuration impacts robin food availability using total invertebrate abundances or biomass seems inappropriate in this study system, especially given that over 14% of captures were taxa not consumed by robins.

The only major prey type that was less abundant in less-wooded areas was Orthoptera, which makes up 4% of all dietary records for eastern yellow robins (Razeng & Watson 2012). This was unexpected, since this order tends to be less abundant in native vegetation compared to agricultural land (Attwood et al. 2008). Orthoptera may have been more abundant in more-wooded landscapes because those sites tended to have more grass cover – a key habitat characteristic for Orthopterans (Torrusio et al. 2002).

4.5.2 Soil moisture and ground cover

Once Formicidae were excluded, soil moisture explained the most variation in arthropod numbers with moister sites harbouring more arthropods. Variation in rainfall is one of the most important factors to drive changes to ground-dwelling arthropod populations (Levings & Windsor 1984; Frith & Frith 1990; Taylor 2008). Given that most rain falls during the warmest months in the study region, it is possible that the effects of soil moisture were somewhat confounded with those of temperature. Higher temperatures promote increased activity levels by arthropods and, by extension, trapability (Saska et al. 2013). While we did not explicitly assess the effect of temperature, we did find that arthropod numbers excluding Formicidae increased with the onset of spring and into summer.

77 While Formicidae abundances are often higher in habitats with sparse ground cover (Greenslade 1974; Lassau & Hochuli 2004), other ground-dwelling arthropod taxa can prefer areas with more complex ground covers (Lang 2000). For instance, Cousin (2007) found that Hemipterans and Coleopterans preferred microhabitats with more leaf litter and woody debris. Thus, it was initially surprising that, after excluding Formicidae, we found that arthropods were more abundant in microhabitats with less grass cover and shallower leaf litter. However, our sampling method has probably biased these results; arthropods can become more trappable when pitfall traps are surrounded by less complex ground cover (Greenslade 1964; Melbourne 1999; Lang 2000). It is recommended that alternative sampling methods be employed when assessing community differences across habitat types (Greenslade 1964; Melbourne 1999).

4.5.3 Prey accessibility and temporal fluctuations

Our results suggested that landscapes with less surrounding woodland did not negatively impact on prey availability per unit area for eastern yellow robins. The prey groups most frequently consumed by robins, Coleoptera and Formicidae, were actually more abundant in less-wooded landscapes. Even if prey with relatively higher nutritional values were depleted at times in these landscapes, it seems likely that Formicidae could supplement the diet to meet energy demands, at least over the short term. However, our study does not paint a full picture of the relationship between habitat loss and food availability.

First, temporal fluctuations in arthropod populations are highly responsive to changes in seasonal, short-term, and long-term climatic variations (White 2008). Avian energetic demands also vary throughout the year in response to a variety of factors, such as breeding (Nagy et al. 2007), moulting (Ben-Hamo et al. 2016), thermoregulation (Rezende & Bacigalupe 2015), and territory defence (Golabek et al. 2012). There need not be a continual mismatch between food availability and energy requirements for avian populations to be placed at risk. Rather, short-term asynchrony at critical times may be enough to impact population fitness (Visser et al. 2006; Williams & Middleton 2008; Maron et al. 2015). Our sampling regime was not extensive enough to identify peaks and troughs in prey availability across seasons and years (Bell 1985; Taylor 2008), and how these fluctuations were associated with times of high energy demand.

Second, the current study only addressed the matter of potential prey availability. It did not fully address the question of prey accessibility. One potential problem for robins inhabiting less-wooded landscapes is that of energetics – whether more energy is expended foraging throughout and

78 defending a strongly-elongated territory rather than one that is more circular in nature. In landscapes containing many linear elements, such as the study region, birds tend to aggregate at intersections where linear elements meet (Lack 1988; Lindenmayer et al. 2007; Hall et al. 2016). This pattern suggests that the linear features are inferior in some way to the adjoining intersections, perhaps because individuals must travel further to reach rich food patches or defend their territory (Lack 1988; Hall et al. 2016). Although Formicidae were more abundant in less-wooded areas, it is unclear whether their low nutritional value would make up for any shortfall in energy requirements over the longer term, potentially making these landscapes less than ideal for robins.

4.5.4 Limitations

An important limitation of this study is the potential bias from digging-in effects. We waited seven days before opening each trap to help ameliorate any digging-in effects (Greenslade 1973). Yet, our captures on the first day of trap activation almost always outnumbered the captures for days 2-5, suggesting either that digging-in effects were present or the invertebrates were initially attracted to the water or detergent. This almost certainly would have affected the absolute number of catches and perhaps led to particular taxa being overrepresented in the samples. However, we do not believe it impacted the relative number of catches across sites since each site was treated identically.

We broadly took prey preference and nutritional value into consideration when assessing prey density by only keeping traps open during daylight hours, only including those taxa known to be eaten by robins, and assessing models of prey density both including and excluding the less- nutritious Formicidae (Chapter 4). However, we did not comprehensively consider prey preference or nutritional value in a quantitative fashion. Adopting a more robust method, such as a weighting system, to account for diet composition and nutritional value could garner additional insights into the issue of food availability by providing results that more accurately reflect food availability from the perspective of the robin.

4.5.5 Conclusion

A key component of many studies attempting to understand why insectivorous woodland birds are declining is estimating relative or absolute food availability. While we only reported on prey groups likely to be eaten by eastern yellow robins, it was clear that different environmental factors were associated with the relative abundance of different orders or families of prey taxa. The nutritional value of these prey groups is likely to play an important role in understanding the complexities of

79 the problem (Razeng & Watson 2015). While we did not find evidence that prey availability was lower in less-wooded landscapes, this result was by no means conclusive and more work is needed in this area. Prey accessibility could pose a risk for eastern yellow robins in less-connected landscapes. Further developing our understanding of the mechanisms driving patterns of vulnerability exhibited by woodland bird populations in fragmented landscapes will be crucial for arresting further declines.

80 CHAPTER 5 Individual condition in an Australian avian insectivore is related to foraging substrate characteristics

This chapter will be submitted to Journal of Avian Biology. Approximate percentage author contributions as follows: Anita Cosgrove – 82% Todd McWhorter – 5% Martine Maron – 13%

81 5.1 ABSTRACT

Decreased availability of arthropod prey in modified landscapes is a potential mechanism causing woodland bird declines, since insectivore populations decline at rates disproportionately greater than other foraging guilds. Approaching the issue from a physiological perspective could prove useful since food availability affects individual condition before deleterious impacts on populations are realised. Eastern yellow robins (Eopsaltria australis) are woodland-dependent, ground-foraging insectivores sensitive to habitat loss and fragmentation. Previous work found that heterophil:lymphocyte ratios (H:L ratios; an indicator of chronic stress) were related to the extent of woodland within the surrounding landscape and this pattern pre-empted woodland extent becoming the most important predictor of occupancy within four years. Here, we examine 1) how well direct measures of arthropod availability explain the variation in robin H:L ratios and body condition and whether other environmental factors are important predictors of condition, and 2) how patterns of H:L ratios have changed over time and what these changes might signify. Across 25 sites in eastern Australia, we sampled robins for H:L ratios, collected indices of arthropod abundance via pitfall trapping, and assessed a range of habitat variables. We found that neither arthropod counts nor mass were related to H:L ratios. Rather, the variation in H:L ratios was best explained by characteristics of the robins’ foraging substrate. H:L ratios were higher at sites with less grass cover and more fallen timber, despite such sites being apparently more-favourable foraging habitat. We found no association between body condition and either arthropod density or foraging substrate. Although we found no evidence of a direct association between prey abundance and robin condition, we could not reject the possibility that reduced food availability was negatively impacting the birds. Sites with apparently more-favourable habitat could house more competitors, thus placing more pressure on resources. H:L ratios were not associated with woodland cover in the present study, probably because extirpations from less-wooded areas meant there were fewer robins to sample from these areas making the relationship increasingly difficult to detect.

5.2 INTRODUCTION

Globally, woodland bird populations are struggling to cope with a multitude of threats, particularly those brought about by habitat loss and land-use intensification (BirdLife International 2008; Vié et al. 2009). There is little doubt that habitat loss drives declines in population abundances and species diversity (Ewers & Didham 2006; Laurance et al. 2011). While much work has been done to identify patterns of population declines, more clarity is needed to elucidate the specific mechanisms through which the impacts of landscape modification are realised (Ford 2011).

One mechanism which may drive impacts is that of reduced food availability. Insectivores, particularly ground-foragers, appear to be more vulnerable to population declines in structurally- altered landscapes than other foraging guilds (Laurance et al. 2011; Stevens & Watson 2013; Bregman et al. 2014), suggesting that reduced arthropod availability or accessibility may be a key underlying cause (Razeng & Watson 2012). Prey reductions in modified landscapes could have deleterious effects on individual fitness via pathways that are more complex than simple starvation or smaller broods. For instance, birds that use multiple woodland patches separated by an inhospitable matrix to source sufficient resources incur greater energetic costs and predation risk (Hinsley 2000; Rodríguez et al. 2001). Even small prey shortfalls in modified landscapes could have synergistic effects with climate change to have pronounced impacts on individual birds, such as when hot, dry weather or drought further reduce prey abundance (Stevens & Watson 2013).

Invertebrate biomass per unit area could be lower in smaller or more-fragmented woodlands for numerous reasons. For example, these landscapes contain more edge habitat that is occupied by both woodland-dependent species and habitat generalists, increasing the predation pressure (Barbaro et al. 2012). Habitat edges are also drier which reduces their suitability for some prey taxa (Didham et al. 1998; Laurance et al. 2011). In agricultural landscapes, farming practices can alter the soil characteristics of the adjoining woodland as well as the availability of water and nutrients (Watson 2011). There is some support for the hypothesis that, relative to their size, small habitat patches contain less invertebrate prey per unit area than large patches (Burke & Nol 1998; Zanette et al. 2000), but the evidence for this is by no means conclusive (Şekercioğlu et al. 2002; Moreno et al. 2013).

Physiological measures could provide insight into the problem, since food availability affects individual condition long before effects become evident at the population level (Ellis et al. 2012; Deikumah et al. 2015). Increased food availability reduces stress levels within individuals (Davis et al. 2000; Kitaysky et al. 2007; Herring et al. 2011). Elevated stress levels, especially those sustained over long periods of time, can have indirect fitness costs, such as greater susceptibility to infection (Al-Murrani et al. 2002), slower growth rates (Moreno et al. 2002), and reduced survival (Kilgas et al. 2006; but see Bonier et al 2009; Dantzer et al. 2014; Milenkaya et al. 2015). Adopting a physiological approach can allow assessment of the impacts of anthropogenic landscape alterations that are still in the relatively early stages before population declines, occupancy rates, or range contractions are moderately advanced (Ellis et al. 2012).

83 One ground-foraging insectivore that shows deleterious responses to habitat clearing is the eastern yellow robin (Eopsaltria australis). This woodland-dependent species is sensitive to both patch area and patch isolation, particularly in agricultural landscapes (Barrett et al. 1994; Watson et al. 2001; Watson et al. 2005). In 2009, Maron et al. (2012) found that robins had higher chronic stress levels in less-wooded landscapes of southern Queensland. This physiological pattern at the individual level appeared to be an early sign that populations were undergoing change. By 2014, woodland extent within the surrounding landscape had indeed become the most important predictor of site occupancy (Chapter 3). However, the mechanism causing these physiological changes within the population is unknown. Zanette et al. (2000) identified that a lower biomass of invertebrate prey was available to eastern yellow robins in north eastern New South Wales in smaller woodland patches than larger patches. This result suggests that reduced food availability is a likely candidate. Yet, in Chapter 4 we found that prey abundance was inversely related to the extent of woodland within the surrounding landscape and this relationship was no longer apparent once Formicidae were excluded.

In this paper, we address two questions by examining relationships between eastern yellow robin condition and factors relating to the habitat, landscape, local climate and food availability. First, we ask: how well do direct measures of relative food availability explain variation in robin condition, and are other factors important predictors of condition? Second, we ask: what does the current pattern of chronic stress indicators tell us about the trajectory of robin populations in this region? We examine how physiological patterns have changed since Maron et al. (2012) and consider what those changes may signal.

5.3 METHODS

5.3.1 Study area

The study was conducted in the Queensland section of the Brigalow Belt South bioregion (Figure 5.1), around Meandarra (27°19´S, 149°52É), Moonie (27°43´S, 150°21É), and Goondiwindi (28°33´S, 150°18É), Australia. This area experiences a sub-tropical climate with cool, dry winters and hot, wet summers (Bureau of Meteorology 2015). The average annual rainfall for the region is approximately 600 mm (Bureau of Meteorology 2015). Birds were sampled from July 2013 to November 2014 and these two years had below-average rainfall (Bureau of Meteorology 2015).

Figure 5.1. Location of the study region in the Brigalow Belt South bioregion, Queensland, Australia. Black circles denote study sites.

Historically, the fertile clay plains of the region were dominated by brigalow (Acacia harpophylla) woodlands while the less-fertile rises were comprised of forests of belah (Casuarina cristata), Eucalyptus spp., or white cypress pine (Callitris glaucophylla). Intensive clearing of the brigalow- dominant woodlands began in earnest from the 1960s (Seabrook et al. 2006). Today, the region is dominated by pasture and agricultural crops. Most remaining brigalow occurs in narrow strips alongside pastures, crops, and roads, while a few larger patches are intact within protected areas. Brigalow-dominant and -codominant ecological communities are federally listed as endangered; less than 4% of the study region is currently covered by remnant brigalow woodland (Commonwealth of Australia 2016).

85 5.3.2 Study species

Eastern yellow robins are small (approximately 20 g), ground-foraging insectivores that live in pairs or small family groups (Higgins & Peter 2002). They are sedentary birds that maintain a home range of approximately 5-6 ha (Higgins & Peter 2002; Debus 2006). Breeding occurs from July to January with two to three clutches laid per year, but tends to taper off from October (Higgins & Peter 2002). The most commonly recorded prey are: Coleoptera; Formicidae; Diptera; Lepidoptera; Hemiptera; other Hymenoptera; Araneae; and Orthoptera (Razeng & Watson 2012). This woodland-dependent species occurs in all woodland types across the study region.

5.3.3 Condition indices

We selected 26 sites from which to sample robins. Criteria for site selection were that the site was accessible, it was in brigalow-dominant or -codominant woodland, it had a minimum tree density of at least 150 stems per ha, and it was at least 800 m from any other site to reduce spatial dependency.

Eastern yellow robins were caught in mist nets with the aid of call playback. Nets were monitored constantly so birds could be extracted immediately upon capture. A small blood sample (< 75 µl) was collected from each robin via brachial venipuncture. This blood was immediately used to make up to six smears on glass slides which were then air dried and fixed in 100% methanol for 2 mins. Each robin was banded, colour banded, weighed (accurate to the nearest 0.01 g), and measured. Eastern yellow robins are somewhat sexually dimorphic, so total head length (mm) could be used to determine sex, although this was not possible for birds in the middle of the range (Rogers et al. 1986). We observed most birds for several weeks following capture, so behavioural cues such as territorial calling or nesting behaviour could also be used to distinguish sex. We used the residuals from a linear regression of total body mass (g) against tarsus length (mm) for each robin as an index of body condition. This common method is not without criticism (Peig & Green 2009); however, it produces a measure that is easy to interpret.

Smears were originally stained with either Giemsa (Fluka 48900-100ML-F) or Wright-Giemsa (Sigma-Aldrich WG32-1L); however, stain quality was poor. All smears were later restained in an autostainer (Hematek 2000 Slide Stainer) stocked with Wright-Giemsa. We performed a differential white blood cell count at x1000 magnification throughout the monolayer on each smear until we reached 100 leukocytes. This count was repeated two to four times for each individual. We calculated the ratio of the number of heterophils to the number of lymphocytes (“H:L ratio”) for

86 each count, then averaged this for each bird. We also counted the number of lysed cells (both red and white lysed blood cells) and leukocytes within 10 monolayer fields at x400 magnification. Two to four counts were made per bird depending on the number of smears that were available. Lysed cells were averaged across all fields counted for an individual. The estimated total number of leukocytes per µl (“leukocyte count”) was calculated by averaging the number of leukocytes across 10 monolayer fields at x 400 magnification for an individual and multiplying by 200 (Fudge 2000; Walberg 2001). H:L ratios can change either as the hypothalamus-pituitary-adrenal axis is activated in response to a stressor or as the immune system responds to infection or inflammation making interpretation difficult (Davis et al. 2008; Weiss & Wardrop 2010). However, elevated total leukocyte counts generally indicate increased infection or inflammation (Davis et al. 2008; Clark et al. 2009; Wojczulanis-Jakubas et al. 2012). Thus total leukocyte counts could aid our interpretation of the results. This method for estimating leukocyte counts is quite coarse, so we used the interclass correlation coefficient (ICC) to assess the repeatability of our measures Lessells & Boag 1987). The ICC was 0.45 which is considered to be “fair” (Cicchetti 1994), so we were relatively confident that the data were representative of our samples.

5.3.4 Explanatory variables

At each site, we marked out four 20 m x 20 m quadrats from which to sample habitat variables. We counted the number of trees and the number of wilga (Geijera parviflora) shrubs within each quadrat. We chose to use just wilga shrubs as opposed to all shrubs because they tend to produce a substantial amount of litter fall which the robins foraged, and we often saw robins nesting or perching in them (Anita Cosgrove, personal observation). A 1 m x 1 m microquadrat was placed in each corner of each quadrat (16 microquadrats in total), within which we assessed the proportion of area that was covered in live or standing dry grass (“grass cover”). A 20 m transect was positioned up the centre of each quadrat. We counted the number of fallen pieces of timber > 7 cm in diameter that lay across the transect (“fallen timber”). At the 5 m, 10 m, and 15 m points along the transect, we measured the leaf litter depth down to the soil surface to the nearest 1 mm and the percentage of volumetric soil water content (“soil moisture”; MPM160 Moisture Probe Meter). We averaged the data across each site for each habitat variable.

Using ArcGIS 10.2.1 (ESRI 2014), we created a map of woodland cover within the region from the Statewide Landcover and Trees Study Foliage Projective Cover 2011 dataset which had a spatial resolution of 25 m (State of Queensland 2014b). We classed any woody vegetation with a minimum foliage projective cover of 8% and a minimum patch size of 2 ha as woodland. A buffer with a 500

87 m radius was placed around each site. Within each buffer, we calculated the proportion of area that was covered by woodland (“proportion of woodland”) and the number of woodland patches (“no. patches”). We used a 500 m radius as this would ensure our landscape metrics encompassed the robins’ territories, regardless of whether the site was positioned at the centre or edge of their territory.

We sampled arthropods at each site with wet pitfall traps. We positioned pitfall traps at the 5 m, 10 m, and 15 m marks along three of the four transects at each site (nine traps total per site). Traps were made from two round, plastic, straight-sided containers approximately 12 cm in diameter and 10 cm deep placed inside each other with a removable lid. Traps were dug in to ground level and left for one week to help avoid digging-in effects on the captures (Greenslade 1973). After the waiting period, we added approximately 2 cm of water with a drop of detergent to each trap and a small pile of wood designed to prevent inadvertent vertebrate drownings. For five consecutive days, the traps were opened each morning and closed each evening to ensure only diurnally-active arthropods were targeted. We removed captures on a daily basis and stored them in 70% ethanol.

We aimed to keep each trap open for a period of about 8-9 hrs each day, but this was not always possible due to logistical constraints. Therefore, we calculated the number of trapping hours conducted at each site to account for differences in trapping effort. Additionally, some traps were disturbed by animals or left open overnight when poor weather prevented access. In these cases, we excluded captures from the counts for those days and excluded those periods from the trapping hours.

Arthropods captured at each site were sorted by taxonomic group following that used by Razeng and Watson (2012; 2015). Most were sorted to Order, but Formicidae were separated from other Hymenoptera since ants formed the bulk of the captures and are ecologically distinct from bees and wasps (Cooper et al. 1990). We only included arthropods > 1 mm and < 40 mm in our counts, since robins had difficulty eating mealworms > 40 mm in length (Anita Cosgrove, personal observation). We included the following taxonomic groups in our arthropod counts based on dietary records (Higgins & Peter 2002; Razeng & Watson 2012): Araneae; Coleoptera; Diptera; Formicidae; Hemiptera; other Hymenoptera; Lepidoptera; Orthoptera; and unidentified larvae that were probably Coleoptera or Lepidoptera. We considered both the total number of arthropods caught per trapping hour and the number caught per hour excluding Formicidae in our analyses. It is unclear if Formicidae are preferred prey given their low nutritional value, or whether they are

88 opportunistically eaten when more-preferred prey are difficult to find (Poulin & Lefebvre 1996; Razeng & Watson 2015).

5.3.5 Statistical analysis

We conducted all analyses in the R statistical package, version 3.2.5 (R Core Development Team 2015). We used the lysed cell count as an indicator of smear quality since some cell types are more prone to lysing and this could influence the H:L ratio (Fudge 2000; Weiss & Wardrop 2010). A simple linear regression between H:L ratio and lysed cell count confirmed there was no significant relationship between the two and no correction was required. Similarly, sex was not related to any of the condition measures and was excluded from further analyses. H:L ratios respond to capture and handling stress (Johnstone et al. 2012). Sampling H:L ratios within 30 mins of capture should ensure H:L ratios are not elevated by capture stress (Davis 2005; Davis et al. 2008; Cīrule et al. 2012); however, this has not been explicitly tested in eastern yellow robins. Therefore, we performed a simple linear regression between H:L ratios and time since capture and found no relationship (n = 34, df = 32, p = 0.933), which confirmed that this was not an issue in our samples. Long periods of call playback could have elicited a stress response from the birds. Thus, we also ran a simple linear regression between H:L ratios and the time between the start of call playback and blood sampling and found no significant relationship (n = 34, df = 32, p = 0.086), suggesting that this was not a problem.

We assessed the relative support for alternative general linear regression models of H:L ratios, leukocyte counts, and body condition using an information-theoretic approach (Burnham & Anderson 2002). H:L ratios were fourth-root transformed to improve normality, while leukocyte counts were cube-root transformed. Body condition required no transformation. We controlled for the north-south spread of sites by adding northing as a fixed factor to all models, including the null models. We used a Pearson correlation matrix to assess pairwise relationships between all explanatory variables and identified those with r > |0.50|. Time of day was highly correlated with northing (r = 0.78) since birds were only caught after 1000 h at Meandarra, so time of day was removed from the analyses. Ordinal date was also correlated with northing (r = -0.71), but was left in the candidate model sets since the variance inflation factors produced for each model incorporating ordinal date were all < 3, indicating there were no multicollinearity issues (Zuur et al. 2010). We used the arm package to standardise all explanatory to a mean of zero and a standard deviation of 0.5 (Gelman 2008; Gelman & Su 2015).

89 A priori candidate models were designed to reflect alternative resource, habitat, landscape, or climatic factors that may affect robin condition and the three response variables were examined separately using the same candidate explanatory variable sets (Table 5.1). Variables were grouped into six categories of interest, each of which could be related to avian condition: food availability; foraging substrate; vegetation structure; landscape structure; short-term weather; and longer-term seasonal effects. For each category, models of every possible combination of terms related to that category were included. We also generated a null model that included only northing. This process generated a total of 18 candidate models to be used in each analysis of robin condition.

The most parsimonious model from each candidate model set was determined using Akaike’s information criterion corrected for small samples sizes (AICc) in the AICcmodavg package (Mazerolle 2016). We used the differences in AICc values (ΔAICc) to compare models and Akaike

Table 5.1. A priori models developed to assess the resource, habitat, landscape, and climatic factors associated with heterophil:lymphocyte ratios, leukocyte counts, and body condition indices of eastern yellow robins. All models included northing as a fixed effect. Category Models Food availability Arthropod numbers Arthropod numbers excl. Formicidae Ground cover Grass cover + Leaf litter depth + Fallen wood Grass cover + Leaf litter depth Grass cover + Fallen wood Leaf litter depth + Fallen wood Grass cover Leaf litter depth Fallen wood Vegetation structure No. trees + No. wilga No. trees No. wilga Landscape structure Proportion of woodland + No. patches Proportion of woodland No. patches Short-term weather Soil moisture Seasonal effects Ordinal date Null model Null

90 weights (ωi) were used to determine the probability that the model was the best model given the observed data and the candidate model set (Burnham & Anderson 2002). The goodness of fit for each model was assessed with adjusted R2 values.

We assessed spatial autocorrelation in both the response variables and the model residuals with spline correlograms generated from the ncf package (Bjornstad 2015). Variance inflation factors generated with the car package confirmed there was no multicollinearity in the final models (Fox & Weisberg 2011). We examined residual scatter plots to check the assumptions of normality and homoscedasticity. Results are reported as mean ± standard error. Graphs were created with the ggplot2 package (Wickham 2009).

5.4 RESULTS

We sampled a total of 34 robins across 27 sites. Time between capture and blood sampling ranged from 7-21 mins (13.4 ± 0.7 mins). One site was excluded from all analyses because the pitfall traps caught an extremely high number of Formicidae that exerted substantial leverage on the results. Additionally, one bird was excluded from the body condition analysis because the tarsus measurement was inaccurate. Untransformed H:L ratio values ranged from 0.0055 to 1.3508 (0.2219 ± 0.0442), untransformed leukocyte counts ranged from 1,050 to 7,800 leukocytes per µl (3,431 ± 257 leukocytes per µl), and body condition indices ranged from -2.2283 to 2.1761 (0.0440 ± 0.1680).

The most parsimonious model of H:L ratio included grass cover and fallen wood (Table 5.2). This model had a moderate fit with an adjusted R2 value of 0.34 and there was substantial support that it performed better than the null model (ΔAICc = 2.09). Grass cover was negatively related to H:L ratios, while fallen wood exhibited a positive relationship (Table 5.3; Figure 5.2a; Figure 5.2b).

All of the chosen a priori models performed poorly when considering robin leukocyte counts (Table 5.4). The null model had the highest ranking, but explained little of the variance in leukocyte counts. None of the birds exhibited signs of systemic infection or inflammation in their leukocyte counts that would warrant their exclusion from the study.

Ordinal date explained the most variation in body condition, although this model had weak support over the null model (ΔAICc = 1.26) and had poor fit (Table 5.5). Robins had marginally better body condition after October (Figure 5.2c).

91 Table 5.2. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin heterophil:lymphocyte ratios.

2 Model K LL AICc ΔAICc ωi R Northing + Grass cover + Fallen wood 5 18.57 -24.91 0.00 0.23 0.34 Northing + Soil moisture 4 16.50 -23.57 1.34 0.12 0.27 Northing + Fallen wood 4 16.30 -23.18 1.73 0.10 0.26 Northing (null model) 3 14.82 -22.82 2.09 0.08 0.22 Northing + Leaf litter depth 4 15.80 -22.16 2.74 0.06 0.24 Northing + Grass cover 4 15.78 -22.13 2.78 0.06 0.24 Northing + Grass cover + Fallen wood + Leaf 6 18.63 -22.02 2.89 0.05 0.31 litter depth Northing + Grass cover + Leaf litter depth 5 17.10 -21.98 2.93 0.05 0.27 Northing + Ordinal date 4 15.70 -21.98 2.93 0.05 0.24 Northing + No. arthropods excl Formicidae 4 15.46 -21.49 3.42 0.04 0.22 Northing + No. patches 4 15.08 -20.74 4.17 0.03 0.21 Northing + Fallen wood + Leaf litter depth 5 16.42 -20.62 4.28 0.03 0.24 Northing + Proportion of woodland 4 15.00 -20.58 4.33 0.03 0.20 The table displays number of parameters (K), maximised log-likelihood (LL), AICc differences 2 (ΔAICc), Akaike weights (ωi), and adjusted R values. Models denoted in bold have substantial empirical support that they explain more of the variation in heterophil:lymphocyte ratios than the null model. Heterophil:lymphocyte ratios were fourth-root transformed prior to analysis. Models with ωi < 0.02 are not displayed.

Table 5.3. Coefficients, standard errors and 95% confidence intervals from the best approximating models of eastern yellow robin heterophil:lymphocyte ratios and body condition as identified through AICc model comparison techniques. Model Coefficient SE Lower CI Upper CI Heterophil:lymphocyte ratios Grass cover -0.1159 0.0561 -0.2307 -0.0011 Fallen wood 0.1207 0.0523 0.0138 0.2276 Body condition Ordinal date 1.1269 0.5827 -0.0645 2.3185 Confidence intervals denoted in bold do not cross zero.

92 a)

Figure 5.2. Scatterplots with 95% confidence intervals (grey shading) of the factors explaining the greatest variation in eastern yellow robin condition. Robust linear regressions of a) fourth-root transformed heterophil:lymphocyte ratios by the proportion of grass cover at each site, b) fourth- root transformed heterophil:lymphocyte ratios of eastern yellow robins by the number of pieces of fallen timber greater than 7 cm in diameter along 20 m transects, and c) body condition indices by ordinal date. 93 Table 5.4. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin leukocyte counts.

2 Model K LL AICc ΔAICc ωi R Northing (null model) 3 -71.68 150.19 0.00 0.18 -0.02 Northing + Ordinal date 4 -70.69 150.81 0.62 0.13 0.01 Northing + Soil moisture 4 -71.08 151.58 1.39 0.09 -0.01 Northing + No. arthropods excl Formicidae 4 -71.15 151.72 1.53 0.08 -0.02 Northing + No. wilga 4 -71.23 151.90 1.71 0.07 -0.02 Northing + Proportion of woodland 4 -71.42 152.26 2.07 0.06 -0.04 Northing + No. patches 4 -71.50 152.42 2.23 0.06 -0.04 Northing + No. trees 4 -71.62 152.66 2.47 0.05 -0.05 Northing + No. arthropods 4 -71.65 152.74 2.55 0.05 -0.05 Northing + Grass cover 4 -71.66 152.75 2.56 0.05 -0.05 Northing + Fallen wood 4 -71.67 152.77 2.58 0.05 -0.05 Northing + Leaf litter depth 4 -71.68 152.79 2.60 0.05 -0.05 Northing + No. trees + No. wilga 5 -70.81 153.84 3.64 0.03 -0.03 The table displays number of parameters (K), maximised log-likelihood (LL), AICc differences 2 (ΔAICc), Akaike weights (ωi), and adjusted R values. Models denoted in bold have substantial empirical support that they explain more of the variation in heterophil:lymphocyte ratios than the null model. White blood cell counts were cube-root transformed prior to analysis. Models with ωi < 0.02 are not displayed.

94 Table 5.5. Model selection results from a priori models examining the effects of resource, habitat, landscape, and climatic factors on eastern yellow robin body condition.

2 Model K LL AICc ΔAICc ωi R Northing + Ordinal date 4 -41.94 93.35 0.00 0.22 0.05 Northing + Proportion of woodland 4 -42.29 94.05 0.70 0.15 0.03 Northing (null model) 3 -43.88 94.61 1.26 0.12 -0.03 Northing + No. wilga 4 -43.20 95.87 2.52 0.06 -0.02 Northing + Leaf litter depth 4 -43.22 95.92 2.57 0.06 -0.03 Northing + Grass cover 4 -43.29 96.07 2.72 0.06 -0.03 Northing + Fallen wood 4 -43.56 96.60 3.24 0.04 -0.05 Northing + Proportion of woodland + No. 5 -42.16 96.63 3.27 0.04 0.01 patches Northing + No. patches 4 -43.61 96.70 3.34 0.04 -0.05 Northing + No. arthropods 4 -43.83 97.14 3.79 0.03 -0.07 Northing + No. arthropods excl Formicidae 4 -43.85 97.18 3.83 0.03 -0.07 Northing + Soil moisture 4 -43.85 97.18 3.83 0.03 -0.07 Northing + No. trees 4 -43.87 97.22 3.86 0.03 -0.07 The table displays number of parameters (K), maximised log-likelihood (LL), AICc differences 2 (ΔAICc), Akaike weights (ωi), and adjusted R values. Models denoted in bold have substantial empirical support that they explain more of the variation in heterophil:lymphocyte ratios than the null model. Models with ωi < 0.02 are not displayed.

5.5 DISCUSSION

The results did not support our original hypothesis that relative food availability would explain the variation in chronic stress levels since there was not a direct link between arthropod abundance and H:L ratios. Instead, H:L ratios were best explained by characteristics of the robins’ foraging substrate, specifically grass cover and fallen wood. This association suggests that food availability or accessibility may play a more subtle or indirect role on chronic stress. Patterns of H:L ratios appeared to be a product of chronic stressors on the robins rather than immunity challenges since total leukocyte counts did not follow a similar pattern. Although body condition marginally improved later in the calendar year, it showed no relationship to food availability.

We found that robins exhibited elevated chronic stress levels at sites with habitat that would be expected to be more favourable for foraging: less grass and more fallen timber. Cousin (2007)

95 established that, at a fine scale, eastern yellow robins select foraging locations with deeper leaf litter, less ground vegetation, more logs, and a denser canopy, presumably because these attributes were associated with higher prey abundances and detectability. Thus, it is counterintuitive that sites consisting of apparently more-preferred foraging habitat contained robins with higher stress levels.

One possible explanation for our observed pattern is that sites characterised by more fallen timber and less grass cover attracted more conspecifics, resulting in a denser local population with closer and more-numerous neighbouring territories. The need for greater territory defence can elevate stress levels (Zuri et al. 1998). Additionally, home ranges are compressed when territories abut each other (Debus 2006) which would reduce the foraging range of the resident pair. While we did not measure robin population densities per se, we did observe that most of the robins with H:L ratios at the upper end of the range had one or more neighbouring territories in close proximity. So, even though the relative abundance of food was not lower at these sites, there is still a question over whether there was less food available to the robins at these sites. The effect of reduced food availability could be further exacerbated through exploitative competition if interspecifics were similarly attracted to these types of sites.

In contrast to Maron et al. (2012), we did not detect a relationship between H:L ratios and the extent of surrounding woodland. One explanation for this is that the strength of this association is likely to decrease over time as faunal relaxation progresses. With woodland extent becoming a more important predictor of site occupancy (Chapter 3), there will be fewer robins to sample from less- wooded sites. Thus, the signature pattern of elevated H:L ratios at these suboptimal localities may become increasingly difficult to detect. Alternatively, the signal may fluctuate over time, such as in response to season, climatic conditions, and breeding effort. Maron et al. (2012) sampled most of their robins during 2009 which was an especially dry year, at the end of one of the longest and most severe droughts on record in Australia (van Dijk et al. 2013; Bureau of Meteorology 2015). Robins could be expected to show elevated stress levels at that time, especially given that moisture drives arthropod abundance in this region (Chapter 4). However, these natural fluctuations do not appear to be the sole cause of the change in pattern since the range of H:L ratios measured by us was very similar to that recorded by Maron et al. (2012; Table B.1).

Leukocyte counts did not follow the same patterns as H:L ratios; total leukocytes were not related to any of our chosen variables. Since leukocyte counts were not higher at sites with less grass and more fallen timber, this suggests the H:L ratio responses were, in fact, stress-related.

96 We detected no relationship between body condition and arthropod abundances or foraging substrate, which is puzzling since studies frequently identify close associations between body condition and food availability (Schoech 1996; Strong & Sherry 2000; Brown & Sherry 2006). One limitation we faced was that our sample size was too small to control for the effects of sex, age, reproductive status, or moult which could have masked the effects of food availability. Alternatively, it is possible that all individuals were in poor condition if arthropod abundances were relatively lower than normal given the below-average rainfall. However, this explanation seems unlikely since we observed several successful breeding attempts.

An alternative explanation is that we did not adequately quantify food availability and consideration could be given to broadening our invertebrate sampling regime. Eastern yellow robins target ground-dwelling prey approximately 73-77% of the time, but they also take about 10-15% of their prey from tree trunks and branches (Recher et al. 1985; Ford et al. 1986; Higgins & Peter 2002). We could complement our pitfall captures with a method that targets arthropods on trunks and branches, such as an intercept trap or vacuum (Jäntti et al. 2001; Majer et al. 2003). This could be an especially important addition if the more-nutritious prey groups, such as Coleoptera, Araneae, and Hemiptera (Razeng & Watson 2015), form a higher proportion of total invertebrate captures from trunks and branches than they did on the ground.

5.5.1 Conclusion

Conservation physiology has the potential to provide critical insights into the mechanisms driving population declines (Ellis et al. 2012), such as those placing ground-foraging avian insectivores at risk. Although our direct measure of relative food abundance did not adequately explain the variance in robin stress levels, characteristics of the foraging strata were important for individual condition. Thus, we were unable to completely rule out the possibility that reduced access to food was limiting populations in these landscapes. More work incorporating both arthropod abundances and avian community effects may help clarify this issue.

The real advantage of using physiological measures to dig into the underlying mechanisms is that they are well suited to use on populations in the earlier stages of declines, or even before declines are evident (Wikelski et al. 2001; Wikelski et al. 2002). This is possible when mechanisms impact individual health, which in turn leads to changes in demographic parameters (Ellis et al. 2012). Tracking changes to the factors associated with avian stress levels over time may provide researchers and conservation managers with early warning signs of impending changes. While we

97 are hesitant to suggest that robin populations will soon decline in even apparently favourable habitat within the study region, the observed pattern of H:L ratios is concerning and bears close monitoring. Using the predictive capacity of physiological measures to take pre-emptive action could provide both substantial cost savings and a stronger chance for interventions to succeed rather than attempting to reverse extinctions after they occur.

98 CHAPTER 6 Food restriction for an avian insectivore in a highly-modified landscape: a supplementary feeding experiment on the eastern yellow robin (Eopsaltria australis)

This chapter will be submitted to The Wilson Journal of Ornithology. Approximate percentage author contributions as follows: Anita Cosgrove – 87% Todd McWhorter – 3% Martine Maron – 10%

Plate 6. An eastern yellow robin visiting a supplementary feeding station stocked with mealworms. These robins usually catch prey from the ground, but became accustomed to using the raised trays.

99 6.1 ABSTRACT

Woodland bird populations continue to be at risk from habitat loss, fragmentation, and degradation, with insectivores apparently most vulnerable. The mechanisms through which land-use change impacts insectivore populations are still unclear, but it is hypothesised that a reduction of arthropod prey availability in highly-modified landscapes could be driving the deleterious effects. To date, evidence for this hypothesis has been gathered from observational studies reporting correlations; however, experimental studies are needed to determine causation. Using the eastern yellow robin (Eopsaltria australis) as a case study species, we employed a before-after-control-impact (BACI) experimental framework to test whether individuals inhabiting less-wooded landscapes in southern Queensland were food restricted. Previous work found that robins in less-wooded landscapes had elevated chronic stress levels. Here, robins were allocated to one of two treatment groups: given supplementary food or unfed controls. We assessed how much stress levels and body condition improved following the treatment across a range of landscapes with varying degrees of woodland cover. An unexpectedly low sample size due to challenges in recapturing individuals (n = 10) precluded us from formally assessing the effects of both feeding treatment and woodland cover on condition. Nevertheless, individuals in the supplementary feeding treatment did not improve in condition, except for one case in a site which had a very low level of woodland cover. This result suggests that it may be worthwhile to direct research effort towards investigating whether food availability has a disproportionately high importance below a particular threshold of habitat extent. We also observed that some control birds showed higher stress levels following the experiment at sites where heavy machinery was used. Care should be taken to exclude such sites from future experiments.

6.2 INTRODUCTION

Woodland birds are bearing the brunt of continued land-use change with habitat loss, fragmentation, and degradation posing the greatest threats to population persistence (Ford et al. 2001; Barrett et al. 2007; Vié et al. 2009). Determining the processes through which landscape modifications impact populations has proved challenging because patterns of population change have been inconsistent within species and across regions (Barrett et al. 2003; Maron & Lill 2006). One common theme emerging from studies is that insectivores are particularly at risk because they consistently show greater responses to landscape change than other foraging guilds (Castelletta et al. 2000; Laurance et al. 2011; Bregman et al. 2014). This pattern suggests that food access or availability plays an

100 important role in driving the population changes observed in this subset of species (Razeng & Watson 2012; Paquette et al. 2014).

Most studies have investigated the issue of food availability in modified landscapes by identifying correlations among individual or population parameters, landscape measures, and some index of food abundance. For instance, the density and pairing success of male Ovenbirds (Seiurus aurocapillus) was higher in woodland patches with a larger core area, males were more likely to occupy territories in areas containing greater prey biomass, and prey biomass increased with woodland patch area (Burke & Nol 1998). The authors concluded that landscape structure was impacting the birds by altering food availability, which in turn affected the settlement preferences of dispersing females (Burke & Nol 1998). This type of approach can highlight patterns, but does not elucidate the causes of population change; experimental evidence is needed to determine causation (Field et al. 2012). In order to do this, food availability can be experimentally manipulated either through supplementing existing resources (Eikenaar et al. 2003; Schoech et al. 2008) or by reducing abundance or access to naturally available prey (Brown & Sherry 2006).

The eastern yellow robin (Eopsaltria australis) is an excellent candidate species for a food- manipulation experiment. Although quite common, this woodland-dependent insectivore is sensitive to both habitat patch size and isolation in agricultural regions (Barrett et al. 1994; Watson et al. 2001; Watson et al. 2005) and has been flagged as potentially at risk of declines (Reid 1999; Watson 2011). Robins had higher chronic stress levels in landscapes containing lower levels of woodland cover, suggesting that these areas provided suboptimal habitat (Maron et al. 2012). This pattern of individual condition was followed by reduced occupancy in less-wooded landscapes within four years (Chapter 3). Zanette et al. (2000) found that smaller woodland patches contained less prey biomass for eastern yellow robins per unit area than larger woodland patches. In contrast, more-recent work identified that prey was unexpectedly more abundant per unit area in less-wooded landscapes; however, once ants (which have a low nutritional value) were excluded, no association between prey abundance and woodland extent was detected (Chapter 4).

We hypothesised that eastern yellow robins were food restricted in less-wooded landscapes, and that this restriction was causing the observed pattern of individual condition. We aimed to test this using a before-after-control-impact (BACI) experimental design (Green 1993). BACI designs allow researchers to assess the impact of a treatment by comparing measures between control and treatment sites, both before and after the treatment is applied (Green 1993). Here, robins were allocated to one of two treatment groups: given supplementary food or unfed controls. We assessed

101 how much individual condition improved following the treatment across a range of sites with varying degrees of woodland cover.

6.3 METHODS

6.3.1 Study region

The study was conducted from July 2014 to January 2015 in the Brigalow Belt South bioregion near Meandarra (27°19´S, 149°52´E) and Goondiwindi (28°33´S, 150°18´E), Queensland, Australia (Figure 6.1). The region is subject to a sub-tropical climate with approximately 600 mm of rainfall per annum, much of which falls during summer (December-February; Bureau of Meteorology 2015). However, 2014 was a year of below-average rainfall with 531 mm falling in Meandarra and 445 mm in Goondiwindi (Bureau of Meteorology 2015).

102 The region has been progressively cleared for agriculture since the late 1800s. However, early attempts to clear brigalow (Acacia harpophylla) resulted in dense regrowth through suckering. It was only after the introduction of blade ploughs in the 1960s that farmers intensively deforested the brigalow woodlands to the extent that brigalow-dominant and -codominant ecological communities are now federally listed as endangered. Today, woodlands dominated by brigalow, belah (Casuarina cristata), Eucalyptus spp., and white cypress pine (Callitris glaucophylla) generally occur in narrow strips alongside pastures, crop fields, and roads with larger patches only persisting inside protected areas.

6.3.2 Study species

Eastern yellow robins are small (approximately 20 g) insectivores that inhabit a range of woodland types throughout eastern Australia (Higgins & Peter 2002). These are sedentary birds that often live in pairs or small family groups and maintain a home range of about 5-6 ha (Higgins & Peter 2002; Debus 2006). This is a ground-foraging species that perches on tree trunks or low-hanging branches from which they spot prey. Over 70% of prey is taken from the ground (Higgins & Peter 2002).

6.3.3 Habitat mapping

We created a map of total woodland within the region in ArcGIS 10.2.1 (ESRI 2014) from the Statewide Landcover and Trees Study Foliage Projective Cover 2011 dataset (State of Queensland 2014b) which had a spatial resolution of 25 m. Woodland was classed as any area of woody vegetation that had a minimum foliage projective cover of 8% and a minimum patch size of 2 ha. We placed a circular buffer with a 500 m radius around the centre of each site and calculated the proportion of the buffer area covered by woodland (“proportion of woodland”). The 500 m radius was used so it encapsulated the landscape used by the robins on a regular basis, regardless of whether the site was located at the centre or edge of their territory.

6.3.4 Experimental design

The experimental design involved capturing robins to obtain initial condition measures, then training the birds to accept food from feeding stations. Once we confirmed that the targeted individuals were taking the supplementary food, they were subjected to their experimental treatment (given supplementary food or unfed controls) for 14 days. The targeted birds were then recaptured to obtain a second round of condition measures.

103 Eastern yellow robins were sampled from sites that consisted of brigalow-dominant or -codominant woodland, had a minimum tree density of 150 stems per ha, and were at least 800 m apart to help reduce spatial dependency. Sites were paired based on the extent of surrounding woodland (proportion of woodland < or > 0.5). Both sites within the pair had either a low or high proportion of surrounding woodland, but one site would be allocated to the feeding treatment while the other would act as a control. Both sites were treated identically throughout the study, for example in terms of training intensity or access following wet weather. Sites were not allocated to treatments randomly. Rather, sites were assigned to treatments to ensure that both experimental treatment groups were balanced in terms of sites with a low or high proportion of surrounding woodland, sex, and whether targeted individuals immediately took supplementary food or required substantial training.

6.3.5 Experimental field methods

Call playback was used to attract individuals to mist nets that were constantly monitored so birds could be extracted immediately upon capture. A small blood sample (< 75 µl) was collected from each individual through brachial venipuncture and used to make up to six blood smears. Each smear was air dried before being fixed in 100% methanol for two minutes. Each bird was banded with a numbered leg band, colour banded to allow identification from a distance, weighed (to nearest 0.01 g), and measured (total head length and tarsus length to nearest mm) before release. Sex was determined using total head length (Rogers et al. 1986) and behavioural cues observed over the training period, such as territorial calling or nesting behaviour.

Feeding stations consisted of a straight-sided round plastic dish 24 cm in diameter and 4.5 cm deep affixed to a post approximately 1 m in height. Robins are primarily ground-foragers; however, trays needed to be raised so the contents were not consumed by feral pigs (Sus scrofa). A ring of white petroleum jelly was smeared around the post to deter ants. Two feeding stations were erected at each site close to a tree trunk or branch that would give the birds a perch from which to pounce. Mealworms (larval Tenebrio molitor) were used as supplementary food and replenished every day.

Robins undertook a period of training (up to 7 days) to ensure they would take mealworms from the feeding stations. Initially, 50 mealworms were added to each feeding station each day. We confirmed whether targeted individuals were taking the mealworms via direct observation or with a motion-activated camera (Scoutguard Sg550) that we circulated around the sites. If the targeted robins were reluctant to feed from the stations, we scattered mealworms on the ground beneath

104 feeding stations and played contact calls near the stations to entice the birds. If further intervention was needed, we tossed mealworms on the ground beneath the perching birds.

After confirming that the targeted birds were taking mealworms, we allocated the site to a feeding treatment (given supplementary food or unfed controls) for 14 days. In both cases, we visited the feeding stations every day for a similar duration; stations at fed sites were filled ad libitum each day while unfed stations remained empty. During the period the feeding treatment was applied, we checked that targeted individuals continued to take mealworms from the feeding stations via direct observations or with the motion-activated camera. Upon completion of the experimental treatment, we caught the targeted birds a second time to obtain a second blood sample, weight, and tarsus measure using the same methods as before. In most cases, it took multiple attempts to recapture the birds regardless of the feeding treatment to which they were allocated. We continued the experimental treatments throughout this period of recapture, up to an additional 14 days, to ensure the second condition sample remained reflective of the food available to the birds.

6.3.6 Calculation of condition data

We calculated three measures of robin condition: heterophil:lymphocyte ratios (H:L ratios), total leukocyte counts, and body condition. Bird blood contains five types of leukocytes (heterophils, lymphocytes, eosinophils, monocytes, and basophils) and the proportions of each in circulation change when individuals are exposed to stressors (Johnstone et al. 2012). Elevated H:L ratios indicate either that the hypothalamus-pituitary-adrenal axis is responding to a stressor, or that the immune system is responding to inflammation or infection (Davis et al. 2008; Weiss & Wardrop 2010). H:L ratios are commonly used as a comparative indicator of the relative degree to which individuals within a population are responding to long-term stressors; however, interpretation can be difficult due to their dual function (Johnstone et al. 2012). Total leukocyte counts generally rise when the immune system is activated, so can be used to help clarify the cause of observed patterns in H:L ratios (Davis et al. 2008; Clark et al. 2009; Wojczulanis-Jakubas et al. 2012).

Blood smears were originally manually stained with Wright-Giemsa in the field, but were later passed through an autostainer (Hematek 2000) using the same stain combination to improve stain quality. Two to four counts of all blood parameters were made for each bird, depending on the number of available smears. We counted both the total number of leukocytes and the number of lysed cells (both red and white lysed blood cells) within 10 monolayer fields at x400 magnification. Leukocyte counts were multiplied by 200 to give the estimated total number of leukocytes per µl

105 (Fudge 2000; Walberg 2001). Some types of leukocytes are more prone to lysing than others which could bias the H:L ratio (Fudge 2000; Weiss & Wardrop 2010); thus, we used the lysed cell count from each slide as an indicator of smear quality. We performed a differential leukocyte count of 100 leukocytes throughout the monolayer at x1000 magnification. H:L ratios were calculated by dividing the number of heterophils by the number of lymphocytes from each differential count. Both H:L ratios and total leukocyte counts were averaged for each time period (before and after the experimental treatment) for each robin. We calculated an index of body condition for each bird based on the residuals from a linear regression of weight against tarsus length.

6.3.7 Statistical analyses

A simple linear regression between lysed cell count and H:L ratio indicated that the two measures were unrelated and, thus, no correction for smear quality was needed. Similarly, no significant relationship was found between H:L ratio and time since capture indicating that capture and handling stress had not biased our stress measure.

BACI analyses can be performed by using the change in the measure of interest between the first and second time points as the response variable and treatment as the explanatory variable (Green 1993). In this study, we were particularly interested in the interaction between treatment and the proportion of woodland, specifically whether the condition of birds in less-wooded landscapes showed a greater response to supplementary feeding than those in more-wooded landscapes. However, we were able to obtain a smaller sample than planned and this precluded us from including proportion of woodland and its interaction with treatment in our analyses.

Instead, we simplified our analyses to test whether robins in the study region were food limited. We subtracted the first condition measure for each individual from the second to derive the change in condition for each bird (ΔH:L ratio, Δleukocyte count, and Δbody condition). Shapiro-Wilk tests (Shapiro & Wilk 1965) indicated that the change in condition measures were not normally distributed, so we used non-parametric two-sided Wilcoxon rank sum tests to identify whether each of the three change in condition measures varied by treatment. Levene tests from the MASS package indicated there was no heteroscedasticity (Venables & Ripley 2002). We calculated Cohen’s d to estimate the observed effect sizes for each condition measure (Cohen 1988). Generally, d values of 0.2, 0.5, and 0.8 are considered to be small, medium, and large effects respectively (Cohen 1988). Although we could not formally test the interaction between proportion of woodland and treatment, we created scatterplots superimposed with linear trend lines to visually

106 assess possible trends that could help inform similar future experiments. We also performed power analyses based on the observed means and standard deviations to determine what sample sizes would be needed to detect significant changes to H:L ratios and body condition should the experiment be repeated in future.

We checked for spatial autocorrelation in the response variables using spline correlograms from the ncf package (Bjornstad 2015). All analyses were performed in the R statistical package (R Core Development Team 2015). Means are presented ± standard error. Graphs were produced with the ggplot2 package (Wickham 2009).

6.4 RESULTS

We originally caught and sampled 23 eastern yellow robins across 17 sites. At two sites, the targeted birds never took mealworms from the feeding stations, so were excluded from the experiment. After the 14 day period of applying experimental treatments, we successfully recaptured only 10 of the targeted birds, each from separate sites. Five of these received supplementary food and five were unfed controls.

All blood samples were collected within 29 minutes of the birds being captured and so were not affected by capture and handling stress. All condition measures were within normal limits and consistent with previous work on this population of robins (Table 6.1; Table B.1; Chapter 5). No birds exhibited an excessively high leukocyte count, such as over 15,000 leukocytes per µl (Campbell 1995), which may have indicated that they were suffering from inflammation or illness and may have warranted their exclusion from the analyses.

Feeding treatment had no significant effect on ΔH:L ratio (Wilcoxon rank sum test: W = 11, n = 10, p = 0.84), Δleukocyte count (Wilcoxon rank sum test: W = 9, n = 10, p = 0.55), or Δbody condition (Wilcoxon rank sum test: W = 13, n = 10, p = 1.00). The 95% confidence intervals for all three condition measures covered a substantial range that included zero (Table 6.2). The effect sizes for ΔH:L ratio (d = 0.76) and Δbody condition (d = 0.70) were relatively large, while Δleukocyte count had a negligible effect size (d = 0.04).

The relationship between the change in condition and the proportion of woodland should have had a horizontal slope if controls were effective, either showing an intercept at zero, or perhaps showing a higher intercept if our continued presence at the sites had a detrimental impact on the birds.

107 Generally, there was a trend for ΔH:L ratios and Δleukocyte counts to rise at sites with more surrounding woodland (Figure 6.2a; Figure 6.2b) while the slope for Δbody condition was relatively flat (Figure 6.2c). In all cases, the intercepts were close to zero.

Power analyses indicated that future similar experiments on eastern yellow robins should target a minimum of 30 samples per group when testing changes to H:L ratios and 34 samples per group when testing changes to body condition.

Table 6.1. Summary statistics for eastern yellow robin condition measures taken before and after the experimental treatment. Δ indicates the change in the condition measure between the two time periods. Variable Range Mean ± SE Initial H:L ratioa 0.0099 - 1.3508 0.1908 ± 0.0851 Second H:L ratio 0.0211 - 0.5766 0.1443 ± 0.0542 ΔH:L ratio -1.1897 - 0.5148 -0.0889 ± 0.1354 Initial leukocyte count 2000 - 7800 3757 ± 423 Second leukocyte count 1800 - 8100 3400 ± 574 ΔLeukocyte count -2300 - 5850 275 ± 759 Initial body condition -2.2296 - 1.2017 -0.3432 ± 0.2628 Second body condition -1.5815 - 1.0634 0.0825 ± 0.2565 ΔBody condition -0.1884 - 1.9316 0.1649 ± 0.2411 a Heterophil:lymphocyte ratio

Table 6.2. 95% confidence intervals from Wilcoxon rank sum tests of changes to eastern yellow robin condition against experimental treatment (given supplementary food or unfed control). Condition measure Lower 95% CI Upper 95% CI ΔH:L ratioa -1.1565 0.1286 ΔLeukocyte count -3400 5850 ΔBody condition -0.3592 1.7748 a Heterophil:lymphocyte ratio

108 a)

Figure 6.2. Scatterplots with linear trend lines of the relationship between the proportion of woodland within 500 m of each site and the change in eastern yellow robin a) heterophil:lymphocyte ratio, b) leukocyte count, or c) body condition after an experimental treatment was applied. Robins either received supplementary food (hollow triangles) or acted as unfed controls (solid circles). The outlying robin that showed a dramatic improvement in condition following the provision of supplementary food is denoted with an arrow.

109 6.5 DISCUSSION

The results did not show the expected positive effect of food supplementation on condition, and therefore, did not support the hypothesis that robins were food limited in the study region. However, we were unable to rule out the possibility that they were food limited due to low power. Robins that received supplementary food showed little improvement in condition, except for one case in a site that had a very low level of woodland cover in the surrounding landscape. Although the overall effect sizes for the change in both H:L ratio and body condition were relatively large, the broad 95% confidence intervals indicated that the study had little power. While the change in leukocyte counts followed a similar pattern to that of the change in H:L ratio (Figure 6.2a; Figure 6.2b), the effect size for the change in leukocyte counts was negligible. This result suggests that the change in H:L ratio represented a stress response rather than an inflammatory response.

In most cases, robin condition changed very little in response to supplementary feeding. This finding was contrary to most other studies of supplementary feeding (Boutin 1990; Cucco & Malacarne 1997; Class & Moore 2013). It was apparent from the breadth of the confidence intervals that the low level of replication provided insufficient power to detect biologically-relevant effects. The confidence interval for ΔH:L ratio spanned almost the entire range of all previously measured eastern yellow robin H:L ratios (Table B.1; Chapter 5), suggesting that the study had low power. However, this does not explain why most fed birds showed a negligible improvement in condition.

One possible explanation is that the two week feeding period was too short to see substantial improvements in condition. Kitaysky et al. (2001) provided red-legged kittiwake (Rissa brevirostris) chicks with diets of differing quantity and quality and after 10 days found that body mass differed significantly between diet treatments – even with only five chicks per treatment. Maxwell et al. (1992) found that a two week period of food restriction significantly elevated H:L ratios in domestic turkeys (Meleagris gallopavo). These studies suggest that individual condition is highly responsive to relatively short-term dietary changes. We provided food ad libitum to ensure targeted robins had sufficient mealworm intake regardless of how much naturally available food was present and how many competitors were vying for that resource. Mealworms have a relatively high energy and protein content (Bell 1990), so they should have provided a nutritional boost if the birds were food restricted.

An alternative explanation is that the birds were not food limited at the commencement of the experiment. Herring et al. (2011) found that stress levels within white ibises (Eudocimus albus)

110 lowered in response to supplementary feeding, but only during years of low food availability. During food-rich years, stress levels were naturally low and showed little response to supplementary feeding (Herring et al. 2011). If a physiological indicator has a non-linear relationship to food availability, then the effects of food supplementation may only be apparent at very low initial levels of food availability. It seems likely that arthropod abundances in the study region were lower than usual because the abundance of robin prey were primarily associated with soil moisture (Chapter 4) and the study was conducted during a period of below-average rainfall. The range of H:L ratios observed during the current study was very similar to those measured in robins during a period of drought (Table B.1; Chapter 5). However, the mean initial H:L ratio in the current study was quite low, so perhaps our sample was not entirely representative of the larger population.

Differences in habitat quality can mask the effects of supplementary feeding. For example, northern goshawks (Accipiter gentilis) that received supplementary food appeared to have similar reproductive success as unfed control birds (Byholm & Kekkonen 2008). However, once territory quality was taken into account, it was apparent that supplementary feeding improved the reproductive success of goshawks occupying food-poor habitats, but made no difference to those in food-rich habitats (Byholm & Kekkonen 2008). A similar dynamic could have occurred here; we had hypothesised that that robin habitat quality varied with woodland extent. This issue emphasises the importance of accounting for habitat quality in food-supplementation experiments and it is unfortunate that we did not attain sufficient replication to do so. Recapturing more of the fed birds at sites with particularly low proportions of woodland cover would have provided important insight into the problem.

One fed bird exhibited a strong improvement in condition and may provide some insight as to what may be occurring in this system. Most fed birds inhabited sites with over 44% woodland cover in the surrounding landscape. Only one bird occupied a very low cover site with just 9% woodland cover, and it was this bird that showed a dramatic decrease in H:L ratio and increased body condition. Several authors have proposed that the effects of habitat loss and fragmentation on faunal populations respond to thresholds of habitat cover; once habitat extent falls below a particular threshold there is an abrupt shift in ecological response (Andrén 1994; Radford et al. 2005). Woodlands in more-intensively cleared agricultural landscapes may have higher exposure to the effects of agricultural practices than those in less-modified landscapes, and these could have synergistic, non-linear effects on invertebrate populations. Tillage and grazing alters hydrology (Yates et al. 2000; Strudley et al. 2008). Insecticides can decimate populations of non-target

111 invertebrate taxa (Hallmann et al. 2014; Sanchez-Bayo & Goka 2014). Nutrient cycles are altered by grazing and fertilisers (Duncan et al. 2008; Watson 2011). It is impossible to draw any conclusions or generalisations from a single data point, but this anomaly highlights that the threshold concept is worth exploring.

Generally, measures from control birds appeared unrelated to the proportion of surrounding woodland, despite showing moderate variation. At two control sites with high level of woodland cover H:L ratios slightly increased. These sites were subjected to considerable disturbance by large machinery during the experimental period, whereas the remaining three were not. One of the two disturbed sites was adjacent to a road that was resurfaced, while the other bordered a field that was repeatedly ploughed to remove brigalow regrowth. Noise and the presence of machinery can elevate stress indices (Dantzer et al. 2014), so these types of disturbances should be taken in to consideration in similar future experiments assessing changes to H:L ratios. Most control birds showed little change in condition, suggesting that our continued presence at each site did not unduly impact the results.

Valuable insights could be gained from replicating this study on a larger scale providing that solutions could be found to the key difficulties we encountered. First, some robins were extremely difficult to train; they avoided feeding stations and were averse to our presence. If this was due to wariness following their initial capture, one solution might be to introduce feeding stations to the site and commence training prior to the initial capture. However, care would need to be taken to ensure the quantity of supplementary food provided did not improve condition prior to the first measure. Additionally, an extra control group that did not receive training might be needed to show that the training, or the trainer’s presence, did not impact robin condition.

Second, most robins were exceedingly difficult to recapture at the conclusion of the experiment. We did so using a single mist net. This stage of the work may have benefited from having multiple nets erected within a site and monitored by one or two volunteers experienced in mist net extractions. We found that having food placed near the net was helpful on some occasions, so using baited, spring-loaded net traps may be a viable alternative trapping method. Another approach could be to incorporate more feeding by hand into the training so birds may be more willing to approach a person near the net, although that may not be feasible if training was conducted prior to the initial capture.

112 Although it is clear that a more effective recapture strategy would be needed to obtain higher replication, it may not be necessary to sample 30-34 birds per group (60-70 total birds) as indicated by the power analyses. If condition does show large changes below some threshold of woodland extent, then it may be worthwhile to collect a higher proportion of samples from sites with less than 50% woodland cover. Also, an unbalanced design could be adopted whereby fewer control birds are sampled than fed birds given that controls appeared to be effective here.

6.5.1 Conclusion

This study did not provide definitive answers as to whether eastern yellow robins were food limited in landscapes containing less woodland. The results indicated that supplementary feeding did not improve the birds’ condition. However, variation in habitat quality that could not be incorporated into the models due to the small sample size may have masked the true effects of feeding on condition. At very low levels of woodland cover, food availability may indeed be a key driver of robin condition.

Ecologists have directed some work towards understanding whether landscape configuration becomes critical for populations below some threshold of habitat extent; while much of the theoretical or simulated studies have supported the idea of a threshold, empirical research has produced mixed support (Fahrig 2002; Huggett 2005). Little has been done to identify the impact of a possible threshold on the ecological processes driving population changes. This study highlights that a focus on process, rather than simply patterns of species richness or distribution, could provide a much deeper understanding of whether thresholds exist and why they may have deleterious impacts on persistence.

113 CHAPTER 7 Synthesis

Plate 7. An eastern yellow robin taking stock of the situation after being released back into his territory.

114 7.1 THESIS OVERVIEW

In this thesis, I investigated a potential driver of avifauna loss from highly-modified landscapes: that prey density is reduced in less-wooded landscapes. Much previous work on this topic has focused on temperate systems with a long history of deforestation (Burke & Nol 1998; Zanette et al. 2000). Instead, I concentrated on a sub-tropical ecosystem that has only been intensively cleared for approximately 55 years (Seabrook et al. 2006). Atlas data from citizen science surveys indicate that woodland-dependent birds are not currently declining in the Brigalow Belt (BirdLife Australia 2015c), which is puzzling because of the extreme amount of habitat clearing in the region (Seabrook et al. 2006). I delved deeper into this conundrum by assessing changes to patterns of condition and occurrence for a case study species, and using these patterns of condition to identify potential threats to the population. However, it is not enough to recognise that population change is associated with landscape structure; rather, a deeper understanding of the underlying mechanisms causing the change is needed to avert and reverse biodiversity loss (Johnson 2002; McGarigal & Cushman 2002; Watson 2011). I explored these mechanisms, focusing particularly on whether there is lower relative prey density (relative abundance or weight of prey per unit area) for insectivores in less-wooded landscapes. I adopted a novel multidisciplinary approach to the issue of food availability in highly-modified landscapes by assessing prey availability in conjunction with landscape structure and avian physiology.

In this final chapter, I summarise the key findings from each of my research questions and discuss how my work has contributed to our understanding of avian ecology, conservation physiology, and landscape ecology. I outline what my findings suggest for the future of eastern yellow robins in the Brigalow Belt South bioregion and highlight management implications. I conclude by outlining the limitations of this study and suggested avenues for future research.

My main objectives in this thesis were to:

 conceptualise the mechanisms through which movement restrictions in fragmented landscapes can impact populations;

 understand how patterns of individual condition could be used to predict future changes to population dynamics; and

115  understand whether avian insectivores are more vulnerable to declines in more-modified landscapes because these areas have relatively lower prey density than more-intact landscapes.

These objectives were achieved by addressing five research questions.

Question 1 (addressed in Chapter 2): How do movement restrictions in fragmented landscapes negatively impact faunal populations, and are there biases in research effort allocated to each of the key mechanisms?

In Chapter 2, I presented a framework for conceptualising the mechanisms through which habitat fragmentation causes population declines. I focused only on those mechanisms impacted by movement restrictions that were a direct product of the structural isolation generated by habitat fragmentation per se (limited resource access, restricted demographic exchange, and impeded gene flow). I argued that these mechanisms are overlapping and hierarchical in nature, and that they operate over increasing spatial and temporal scales. Multiple mechanisms can be disrupted simultaneously and each requires distinct conservation solutions; thus, it is important that conservation managers consider movement restrictions across multiple scales in their system of interest.

The global literature review revealed that impeded gene flow had attracted the bulk of the research effort on habitat fragmentation, especially in more-recent years, while little work had been done to advance our understanding of limited resource access and restricted demographic exchange. I found that most studies had been conducted in the temperate zone, but the frequency that each mechanism was studied within each geographic zone was similar to the overall observed frequencies. When I analysed the research effort directed towards different taxonomic groups, it was clear that the focus on limited resource access was disproportionately low for generally less-vagile taxa, i.e. fish, amphibians, and reptiles. The overall research effort on birds was low compared to that of mammals or invertebrates. I concluded that this bias in research effort on the less-represented taxa prevents researchers from generalising conclusions because differences in dispersal capabilities, range requirements, and generation times could all influence the impacts of habitat fragmentation.

Question 2 (addressed in Chapter 3): How does the association between site occupancy and landscape structure change over time, and how effective are individual condition measures at signalling future changes to population dynamics?

116 In light of the research biases identified in Chapter 2, the remainder of the thesis concentrated on the issue of limited resource access for a case study species: a sedentary avian insectivore in the sub-tropical zone of Australia. This is a species that does not yet appear to be declining in my study region, but has been identified as vulnerable to declines in New South Wales (Debus 2006; Stevens & Watson 2013) and Victoria (Barrett et al. 2003; BirdLife Australia 2015b). Furthermore, an earlier study (Maron et al. 2012) identified that while surrounding woodland extent showed no relationship to robin occurrence, it had a strong negative association with stress levels. The authors predicted that less-wooded landscapes would have higher rates of extirpations and woodland extent would emerge to become an important predictor of occupancy (Maron et al. 2012).

In Chapter 3, I tested these predictions and determined that extirpations did not occur in places where robins had previously been highly stressed. As predicted, woodland extent was now strongly negatively related to occurrence; however, this was due to both extirpations within less-wooded landscapes and colonisations in landscapes with intermediate to high levels of woodland cover. Thus, landscapes with less woodland did indeed appear to be suboptimal in some way and robins appeared to be responding by redistributing themselves around the region. Importantly, patterns of stress indices showed some promise at predicting a change in occurrence patterns, which could prove useful for monitoring species at risk of declines.

I also refined the occupancy model to account for the extent and connectivity of woodland within the broader landscape (within 5 km of each site) that could influence dispersal movements. Woodland cover at this broader scale explained little of the variance in site occupancy and robins were primarily responding to landscape structure at a home range scale (within 500 m of each site). I suggested this may be due to the relative recency of habitat loss from the region, and that fragmentation over larger spatial scales may become more important in future.

Question 3 (addressed in Chapter 4): How is prey density related to woodland extent and how does this differ when nutritional value is taken into consideration?

Chapter 3 established that robin occurrence was increasingly more likely to be linked to the extent of remaining habitat at a home range scale. Therefore, in the remainder of the thesis, I focused attention on whether prey density was causing this potential vulnerability of the population to reduced habitat extent. In Chapter 4, I compared models of habitat, landscape, and climatic factors to identify the most important determinants of the abundance and weight per unit area of invertebrate prey likely to be eaten by robins. In contrast to Zanette et al. (2000), I found that

117 arthropods were more abundant per unit of trapping effort in less-wooded landscapes; however, this pattern was primarily driven by Formicidae (ants) which have a low nutritional value. Once I excluded Formicidae, woodland extent showed little association with prey density. Instead, prey had relatively higher densities at sites with greater soil moisture. The only taxa to show a negative association to woodland cover was Orthoptera (grasshoppers and crickets).

Thus, contrary to the hypothesis that prey density is reduced in more-modified landscapes, robins in less-wooded landscapes within the study region are not apparently exposed to lower prey densities within woodland habitat than those birds in more-wooded landscapes. However, it was clear that drawing inferences from total prey density did not adequately take into account the complex and idiosyncratic responses of prey taxa to landscape structure. I concluded that consideration should be given to prey preference and nutritional value in studies of food availability.

Question 4 (addressed in Chapter 5): How well do measures of prey density explain the variation in the physiological condition of robins, and how have patterns of condition changed over time?

After identifying factors associated with prey densities in Chapter 4, I then set out to determine how prey density was related to robin condition. Given that the original condition indices had been collected four years earlier (Maron et al. 2012), it would have been inappropriate to compare the original indices to current measures of prey density. Thus, while collecting the invertebrate pitfall trap data for Chapter 4, I sampled robins to obtain fresh condition indices. In Chapter 5, I compared models of robin condition that incorporated measures of habitat and landscape characteristics, local climate, and prey density. In contrast to Maron et al. (2012), woodland extent performed poorly at explaining the variance in physiological condition. This made sense given that in Chapter 3 I identified that woodland extent had become the most important predictor of occupancy. Fewer of the less-wooded sites were now occupied by robins, so it would be more difficult to detect a physiological signal that these sites were suboptimal.

Surprisingly, robin condition was unrelated to prey density, regardless of whether I used total abundance per unit area or excluded less-nutritious prey taxa. Thus, this exploration yielded no evidence to support the hypothesis that insectivores are more vulnerable to declines in more- modified landscapes due to reduced food availability. Instead, robins had higher stress levels at sites with apparently more-favourable foraging habitat, i.e. those sites with less grass cover and more fallen timber. This result raised the intriguing possibility that if reduced food availability influences robin stress, it is not through reduced prey densities. Instead, it may be because there is a smaller

118 area from which to source food in less-wooded landscapes. Furthermore, sites with more-favourable foraging habitat may attract more conspecifics, leading to compressed home range sizes. Thus, the prey available to the resident pair could be reduced due to their contracted foraging range. Of course, both these possibilities could be operating in my study system. For instance, food limitations in more-modified landscapes caused by a smaller woodland area in which to forage may explain the change in the pattern of occurrence, while smaller foraging ranges due to increased robin densities in more-favourable habitat could explain the change in the pattern of condition.

Question 5 (addressed in Chapter 6): How does experimentally increasing food availability for an avian insectivore impact individual condition in landscapes with varying degrees of woodland cover?

While the results from Chapters 4 and 5 suggested that prey density did not explain why less- wooded landscapes are apparently suboptimal for robins, causality can only be tested through experimental manipulation. Thus, I concluded my investigation by conducting a supplementary feeding experiment and assessing robin condition before and after applying the feeding treatment. I was unable to evaluate the change in condition across landscapes with varying degrees of woodland cover as originally planned due to low replication. Birds that received supplementary food generally did not show an improvement in condition, suggesting that robins in the region were not food limited. However, the study had low power which limited my ability to draw firm conclusions. This work highlighted the importance of taking habitat quality into consideration when conducting feeding experiments because natural variations in food availability can mask responses from feeding treatments.

One bird at a site with an exceptionally low level of woodland cover showed a dramatic improvement in condition after receiving supplementary food, which suggested that food availability played a role in that case. Although it is impossible to discern with the available data why that bird was food limited, it may be worthwhile to further explore the concept that ecological processes can have disproportionate effects on systems once habitat extent falls below a particular threshold.

7.2 CONTRIBUTIONS TO ECOLOGICAL AND PHYSIOLOGICAL KNOWLEDGE

In this thesis, I make a number of contributions to advancing our understanding of landscape ecology, conservation biology, and physiological ecology. Although I focused on a single case

119 study species, my findings have broader implications for research into the proximate causes of woodland bird declines. The main findings from this thesis are:

1. Little research effort has been dedicated to understanding the impacts of habitat fragmentation on resource availability compared to that given to its effects on demographic exchange and gene flow.

2. Eastern yellow robin site occupancy within the Brigalow Belt South appears to be in a state of flux; woodland extent at a home range scale has now become critically important for occurrence.

3. Patterns of H:L ratios show some potential for informing us about population changes currently underway that are not yet reflected in patterns of occurrence.

4. In the Brigalow Belt South bioregion, sites in less-wooded landscapes do not contain lower arthropod prey density for robins than landscapes with more woodland cover.

5. Prey density did not affect the condition of the insectivorous eastern yellow robin in the Brigalow Belt South bioregion.

The work conducted as part of this thesis provides a good example of two important practices that are uncommon in ecology. First, I replicated work that was conducted four years earlier on the same study system (Maron et al. 2012). The demand for novel results from funding bodies and publishers discourages the replication of ecological studies (Nakagawa & Parker 2015). However, replication of previous studies is a crucial step in scientific advancement that helps guard against spurious outcomes from a single study and improves our ability to generalise findings (Johnson 2002; Nakagawa & Parker 2015). The results obtained from the initial study revealed a snapshot of ecological patterns and suggested certain factors might be important influences on the study species (Maron et al. 2012). However, by repeating the study, I have been able to use the shifts in patterns of bird occurrence and condition after a four-year period to deepen our understanding of the forces at work in the system and gain additional clarity on potential threats. Changes to occupancy patterns revealed that the distribution of birds among sites is not stable (Chapter 3). Changes to patterns of individual condition hinted that the redistribution of robins within the region may be leading to reduced access to food due to either overcrowding or reduced home range sizes within more-

120 optimal sites (Chapter 5). Although I did not have data to test that hypothesised threat, the change in pattern is worth exploring.

Second, I formally tested predictions made in a publication four years earlier (Chapter 3; Maron et al. 2012). In published articles, authors typically propose reasons for their observed results and draw out their implications. However, it is relatively rare for researchers to then test whether the suggested implications eventuated (but see Cooper & Walters 2002; Wikelski et al. 2002; Bennett et al. 2012). Mac Nally and Bennett (1997) called on conservation biologists not simply to make a priori predictions, but to share these predictions prior to the commencement of the study. Notably, they provided an excellent example of how to do this (Mac Nally & Bennett 1997; Mac Nally et al. 2000). First, they explicitly outlined species and population traits that could make woodland birds vulnerable to extinctions and described how they proposed to test the effects of these factors (Mac Nally & Bennett 1997). They then carried out data collection and tested their predictions (Mac Nally et al. 2000). Maron et al. (2012) made explicit predictions about how eastern yellow robin populations in the Brigalow Belt South would change going forward, although they did not explicitly outline how these predictions would be tested. I later adopted the same methods used by Maron et al. (2012) to test their predictions and found mixed results; extirpation rates were not higher at particular sites that contained birds with elevated H:L ratios, but less-wooded sites were less likely to be occupied. To date, this type of approach has rarely been adopted, although it may become more common if pre-registration of studies gains wider acceptance.

7.2.1 Insectivorous birds and their prey

The key findings to come out of this thesis were that eastern yellow robins in the Brigalow Belt South are sensitive to low levels of woodland extent in the surrounding landscape, but this sensitivity does not appear to be a product of differences in relative prey density between sites. Since the original study conducted four years previously (Maron et al. 2012), robins had left less- wooded sites (presumably either through death or emigration) and colonised sites surrounded by more woodland (Chapter 3). This shift in occupancy had occurred to such an extent that woodland cover within 500 m had now become the strongest predictor of occurrence (Chapter 3). While robin condition was no longer related to woodland extent, this could be readily explained by the changes to occurrence – robins may well have left the least-favourable sites (Chapter 5), and so were no longer there to be sampled. When considering all prey taxa, I found that less-wooded landscapes actually yielded relatively more arthropods per unit of sampling effort than more-modified landscapes (Chapter 4). While this result was primarily driven by the abundance of the less-

121 preferred Formicidae, Coleopterans were also particularly abundant in landscapes with very low levels of woodland cover (Chapter 4). Robin condition was unrelated to prey density (Chapter 5) and providing supplementary food had little impact on robin condition, bar one exception (Chapter 6). The findings from each of these chapters consistently support my conclusions that robins are vulnerable to reduced woodland extent and that lowered prey density within the remaining woodland is not causing that vulnerability.

My findings run counter to some other studies of the ground-dwelling prey available to woodland- dependent avian insectivores in structurally-altered landscapes (Burke & Nol 1998; Zanette et al. 2000). Both Burke & Nol (1998) and Zanette et al. (2000) found that prey densities were lower in less-wooded landscapes. However, Şekercioğlu et al. (2002) found no relationship between prey density and woodland cover. These disparities emphasise that prey density is not universally lower in sites in less-wooded landscapes and it will be important to understand why this is the case. Some striking patterns stand out when comparing these studies. The work conducted by Burke & Nol (1998) and Zanette et al. (2000) were conducted in the temperate zone (in Ontario, Canada and New South Wales, Australia, respectively) in agricultural regions that have a long clearing history; in both cases, the land was intensively cleared by the 1800s (Harris 1987; Zanette et al. 2000). In contrast, the work by Şekercioğlu et al. (2002) was carried out in the tropical forests of Costa Rica that have undergone more-recent deforestation – mostly since the 1950s. My study was Australia’s subtropical zone where brigalow habitat has been intensively cleared since the 1960s (Chapter 1).

Time since clearing of habitat may be important for prey densities in more-modified landscapes. Although the relaxation of invertebrate fauna could be expected to be more advanced in those regions with longer clearing histories (Tilman et al. 1994; Kuussaari et al. 2009), it is unlikely that this is the primary cause of differences in prey density patterns between studies. Relaxation of invertebrate fauna should still be relatively advanced in regions cleared since the 1950s/1960s given that invertebrates have short generation times (Gonzalez & Chaneton 2002; Kuussaari et al. 2009; Krauss et al. 2010; Halley et al. 2016). Instead, landscapes with longer clearing histories may exhibit a different degree of habitat degradation to those with shorter clearing histories, either due to differing land-use legacies (Lunt & Spooner 2005; Perring et al. 2016) or perhaps because they have been exposed to deleterious agricultural practices for longer. Invertebrate densities can vary with microclimate (Didham et al. 1998), vegetation structure (Bromham et al. 1999), and soil structure (de Araújo et al. 2015), all of which can be altered by habitat degradation (Tongway et al. 2003; Lunt & Spooner 2005). Given that the impacts resulting from habitat loss, fragmentation, and

122 degradation are highly interrelated (Chapter 1), it is possible that patch-size effects on invertebrate fauna could be more apparent in regions with greater degradation.

The pattern that I outlined above whereby prey density seems to vary with habitat fragment size in temperate systems but not tropical/subtropical systems would suggest that temperate insectivores should be more sensitive to patch area than tropical insectivores. However, a meta-analysis has revealed the opposite to be the case: tropical insectivores show more sensitivity to patch size than their temperate counterparts (Bregman et al. 2014). The trend for higher sensitivity in tropical insectivores than temperate insectivores has been attributed to sharper contrasts between habitat and matrix structure in tropical regions and general differences in species traits, such as dispersal ability, reproductive output, and dietary specialisation (Bregman et al. 2014). However, none of these explanations shed any insight into why the relationship between prey density and patch area appears to vary across geographic zones, and how that relationship might be influencing the birds’ sensitivity to patch size.

Watson (2011) proposed the productivity hypothesis to explain avian insectivore declines in temperate eucalypt woodlands. He suggested that clearing for agricultural purposes was non- random and that farmers preferentially cleared woodlands on more-fertile soil. Thus, the woodlands remaining today are those growing on less-fertile soils (Watson 2011). He hypothesised that the remaining woodlands would support fewer invertebrates today than the historical woodlands because a) they were situated on less-productive soil, and b) agricultural practices had fundamentally altered nutrients and hydrology on the land (Watson 2011). Fewer invertebrates would then, in turn, support fewer insectivores (Watson 2011). The findings from Zanette et al. (2000) supported this hypothesis; they found fewer ground-dwelling invertebrates in smaller patches which would presumably be exposed to greater agricultural impacts and degradation than larger patches due to their higher perimeter to area ratio. The work by Zanette et al. (2000) was conducted in eucalypt woodlands that produce relatively little leaf litter (Sangha et al. 2006; March & Watson 2007; Watson 2011).

In contrast, my study was conducted within brigalow woodland which grows on highly-fertile clay. Undisturbed sites develop a deep, rich humus layer (Anita Cosgrove, personal observation) which would probably provide favourable conditions for litter-dwelling invertebrates (Lieberman & Dock 1982; Haskell 2000; Kazemi et al. 2009). Gilgai microrelief improves water capture and retention on the land (Kishné et al. 2014). Thus, according to the productivity hypothesis, my study region should support relatively high densities of invertebrates, potentially ameliorating food limitation for

123 insectivores. The fact that I found no relationship between prey density and woodland extent is consistent with this view of the study area, and potentially also with the productivity hypothesis.

However, at my disturbed sites, little leaf litter was present, grazing was common, and microrelief had been lost. Yet, I found no effect of leaf litter depth or gilgai presence on prey density, even after excluding Formicidae (Chapter 4). A key tenet of the productivity hypothesis is that agricultural practices are altering nutrient inputs, hydrology, and soil structure (Watson 2011). Grazing increases nutrients loads, changes floral communities, alters soil structure, and reduces the soil’s moisture holding capacity (Tongway et al. 2003; Watson 2011). Woodlands adjacent to crops may be exposed to increased chemical run off and spray drift (Yates & Hobbs 1997; Duncan et al. 2008). As mentioned above, it might be expected that these changes would be more pronounced in more- modified landscapes due to higher perimeter to area ratios. However, I did not find lower prey densities in less-wooded landscapes (Chapter 4). Thus, the extent to which my results were consistent with the productivity hypothesis was inconclusive.

7.2.2 Conservation physiology

Several authors have suggested that physiological measures could be used as a tool to predict upcoming population changes (Wikelski & Cooke 2006; Ellis et al. 2012; Deikumah et al. 2015; Maute et al. 2015). Physiological indices have been used to assess a species’ vulnerability to a threat (Deikumah et al. 2015; Legge et al. 2015; Maute et al. 2015). However, little progress has been made on evaluating whether predictions of population change made from physiological indices are accurate (but see Wikelski et al. 2002).

In this thesis, I make two novel contributions to our knowledge on the subject. First, I demonstrated that a snapshot of locations containing birds with elevated H:L ratios was not an accurate indicator of where extirpations would occur in my study system (Chapter 3). My findings run counter to those of Wikelski et al. (2002) who recorded increased stress levels in a population of marine iguanas exposed to an oil spill (Wikelski et al. 2001). The authors predicted that this spike in stress levels would precipitate high mortality in that population and their prediction was realised within one year (Wikelski et al. 2001; Wikelski et al. 2002). I attributed the difference in prediction accuracy between our studies to the different stressors facing our study organisms (Chapter 3). The oil spill produced a dramatic reduction in available food (Wikelski et al. 2001) which could be considered to be a sudden stressor likely to produce an extreme stress response. In contrast, habitat loss in my study region has been ongoing for many years (Seabrook et al. 2006). If food is restricted

124 in less-wooded landscapes, the restriction is likely to be less severe than that generated by the oil spill and would probably invoke a chronic, but milder stress response. My work highlights that the nature of the stressor must be taken into consideration when making predictions about populations from stress indices and further research is needed to understand how the type of stressor limits our ability to make predictions.

Second, I found some evidence that patterns of H:L ratios could provide an effective means of predicting an unfolding change to patterns of occurrence (Chapter 3). Originally, H:L ratios were more likely to be elevated in less-wooded landscapes, but occurrence was unrelated to woodland cover (Maron et al. 2012). Four years later, I found that woodland extent had become the strongest predictor of occurrence (Chapter 3). However, this change in the pattern of occupancy was not driven by robins leaving the specific sites where they had previously been more stressed. Rather, robins had left less-wooded sites in general and colonised more-wooded sites.

So, why was there a discrepancy between locations and patterns? Stress measures can be highly variable, even within individuals (Vleck et al. 2000; Ochs & Dawson 2008). Additionally, H:L ratios respond to many factors, such as time of day (Müller et al. 2011), age (Davis et al. 2000; Kulaszewicz et al. 2015), reproductive stage (Vleck et al. 2000; Wojczulanis-Jakubas et al. 2015), parasite load (Lobato et al. 2005; Wojczulanis-Jakubas et al. 2012), and food availability (Davis et al. 2000; Banbura et al. 2013). Maron et al. (2012) collected a single snapshot of H:L ratios. Given the potential for variability within individuals, it is implausible to suggest that the individuals most stressed at that point in time are always most stressed within the population, particularly since it had been four years since the birds were originally sampled. Rather, in this situation, it seems more appropriate to conclude that there was a tendency for robins within less-wooded sites to have higher stress levels, and it is this tendency that is most informative when assessing potential threats to the population.

To my knowledge, this is the first time that anyone has tested whether patterns of H:L ratios can effectively predict changes to patterns of occurrence or changes to factors affecting occurrence. This application could potentially aid conservation efforts. Although obtaining condition measures is often more time consuming and costly than assessing occupancy and distribution (Maron et al. 2012; Casner et al. 2014), incorporating condition measures into monitoring programs could allow conservation managers to intervene in at-risk populations sooner. The likely improvement in the effectiveness of conservation outcomes derives not from a lower monetary cost in detecting the

125 decline, but the greater cost-effectiveness of preventing extirpations rather than trying to reverse extinctions after the fact (Maron et al. 2012; Halley et al. 2016).

I also found that the factors most associated with H:L ratios changed since 2009, with robins in my study exhibiting higher stress levels at sites with apparently more-suitable habitat (Chapter 5). This change could be a product of the redistribution process suggested in Chapter 3. Robins may be leaving less-wooded sites and congregating in more-favourable habitat. Having more neighbouring territories can compress home ranges (Debus 2006). The birds may then show higher stress levels because they have access to less prey within a smaller range or need to more-vigorously defend their territory. A question exists over whether these apparently more-favourable sites can sustain robins at the current density. If not, then we may expect to see a gradual decline in these areas. Obviously, more work is needed to determine if there is any merit to this hypothesis. However, my work demonstrates the value in not simply predicting population changes based on patterns of H:L ratios, but using the change in pattern across multiple points in time to develop deeper insights into how problems within the population develop.

7.2.3 Methods for evaluating mechanisms impacting populations in structurally-altered landscapes

In this thesis, I highlighted research biases that have led to gaps in our understanding of the mechanisms through which movement restrictions in fragmented landscapes have impacted faunal populations (Chapter 2). Recent advances in molecular methods have resulted in an explosion of studies exploring the impact of habitat fragmentation on genetic diversity and differentiation. However, little advancement has been made in our understanding of the impacts on food availability and demographic exchange, despite new technologies, such as stable isotope tracers, drone technology, and radio frequency identification systems (Bonter & Bridge 2011; Robb et al. 2011; Whitehead et al. 2014). Nevertheless, our ability to investigate the less-explored mechanisms is not dependent on new technology. I designed an experiment to test food availability as a causal mechanism of robin vulnerability in more-modified landscapes (Chapter 6).

I did not attempt to answer the question of whether habitat loss or fragmentation per se were causing reductions in prey abundance. That type of question would require experimental manipulation of landscapes, either within mesocosms (Collins & Barrett 1997; Lecomte et al. 2004) or across large scales (McGarigal & Cushman 2002; Jenerette & Shen 2012). Studies such as the Biological Dynamics of Forest Fragments Project (Bierregaard Jr. et al. 1992) and the Stability of

126 Altered Forest Ecosystems Project (Ewers et al. 2011) demonstrate that it is possible to undertake such broad-scale manipulations (Wilson et al. 2016). However, these types of experiments can be logistically challenging and expensive, and opportunities to implement them are rare (Debinski & Holt 2000; Ewers et al. 2011; Jenerette & Shen 2012). Instead, I tested whether reduced prey density was causing poor condition in robins in less-wooded landscapes. This is a useful alternative to a landscape-scale manipulation and was only possible by adopting a multidisciplinary approach. While this approach does not directly elucidate how changes to landscape structure drives changes to an underlying mechanism, it does allow researchers to see how a mechanism causing declines varies across different landscapes (Jenerette & Shen 2012).

7.3 PROGNOSIS AND MANAGEMENT IMPLICATIONS

This study did not set out to test whether faunal relaxation is ongoing in the study region, and nor can it answer that question. However, the newly-evident importance of habitat extent at the home range scale (Chapter 3), even though habitat extent had changed little since the pilot study conducted by Maron et al. (2012), is consistent with the concept of faunal relaxation. When clearing first occurs, patches have higher occupancy rates than expected given their new size and/or isolation (Diamond 1972; Bierregaard Jr. et al. 1992; Helm et al. 2006). Over time, populations within smaller or more-isolated patches may thin out, shift to nearby suitable patches, or vanish as elevated stochastic extinction rates outpace lowered immigration or recolonisation rates (Bierregaard Jr. et al. 1992; Hylander & Ehrlén 2013). As smaller and more-isolated patches become increasingly less likely to be occupied, patch occupancy will become increasingly more related to habitat extent or isolation.

Interestingly, however, I recorded slightly more colonisations than extirpations (Chapter 3). This result is inconsistent with faunal relaxation (Hylander & Ehrlén 2013). While this would initially appear to suggest that the population had reached a state of dynamic equilibrium, this is not necessarily the case (Clinchy et al. 2002). For instance, transient colonisations can mask a long- term population decline (Clinchy et al. 2002). Most discussions of extinction debt focus on the impacts of low levels of structural connectivity, which prevents colonisations and promotes local extinctions (Hanski & Ovaskainen 2002; Ewers & Didham 2006; Hylander & Ehrlén 2013). Here, the study region has high structural connectivity which could slow down relaxation by facilitating movements between sites (Kuussaari et al. 2009). This would allow territories vacated by death or emigration to be quickly recolonised, even if they are sink sites. Extreme fragmentation is not an essential requirement for an extinction debt to exist (Kuussaari et al. 2009); reduced habitat extent

127 and quality can also drive delayed extinctions by altering vital rates and stochastic processes within populations (Hylander & Ehrlén 2013).

Regardless of whether relaxation is occurring in the region, it would seem the study system is currently in a state of flux given such a dramatic change in the importance of woodland extent for occupancy. This suggests the system is not in a state of equilibrium, so I cannot rule out the possibility that robins may still decline due to past habitat clearance. There are few systematic woodland bird surveys in the region which have evaluated changes to abundances or distribution (Woinarski & Catterall 2004; Woinarski et al. 2006). Draft results prepared for the State of Australia’s Birds 2015 for eastern yellow robins in the Brigalow Belt (including Brigalow Belt North and Brigalow Belt South bioregions) indicate that incidence rates have not declined between 1999 and 2013 (James O’Connor and Glenn Ehmke, personal communication). However, citizen science records are biased towards sites that are publically accessible and interesting to visit, such as protected areas (Isaac & Pocock 2015; Geldmann et al. 2016). Analysis of citizen science records would be less likely to detect occurrence changes in areas that received little survey effort, such as shade lines on private property. These less-expansive woodland strips may show evidence of declines sooner than large habitat patches given that patch area is a key driver of extinction risk (Kuussaari et al. 2009). Also, the data collected for the State of Australia’s Birds 2015 are from across a broad geographic range (BirdLife Australia 2015d) which could mask changes occurring in localised areas. Thus, the report may not comprehensively depict true population changes occurring within the study region.

It is particularly concerning that robins now have higher stress levels at apparently high-quality sites that contain less grass cover and more fallen timber (Chapter 5). Previous patterns of association in H:L ratios did pre-empt changes to patterns of occurrence with robin distribution shifting away from less-wooded sites (Chapter 3). Yet, the predictive capacity of patterns of physiological indices needs further testing. Thus, it would be premature for me to suggest that robin abundance or occurrence will soon decline within the sites with apparently high-quality foraging habitat. However, the situation bears close monitoring and provides an excellent opportunity to continue assessing how reliably patterns of H:L ratios can inform us of unfolding changes to population dynamics.

Habitat extent within the study region changed little between 2009 and 2014; however, the region is not safe from further deforestation. Vegetation management legislation and enforcement in Queensland was relaxed in 2013 (Evans 2016) which has led to an increase in land clearing rates

128 (State of Queensland 2016c). Additionally, coal seam gas operations have expanded considerably, with well clusters occurring to the east, north, and northwest of Meandarra, and north of Goondiwindi (State of Queensland 2015). Only one of the properties I studied contained a coal seam gas well, but several more landholders have been approached by natural resource companies (John Haken, personal communication). Coal seam gas operations have the potential to increase habitat loss and fragmentation, such as through the construction of access roads and infrastructure (Butt et al. 2013). Thus, these potential changes to legislation and resource management pose a threat to the structural connectivity of the woodland within the region.

I did not find an effect of landscape connectivity on robin occurrence (Chapter 3). I assessed separate models of robin occupancy that included either woodland extent within 5 km of each site or accessible woodland within 5 km, and assumed that the difference between these measures represented the effect of connectivity (Chapter 3). Accessible woodland explained a negligible amount of variation in occupancy over woodland extent within 5 km, suggesting that connectivity was not important for occupancy (Chapter 3). However, this method of assessing connectivity has not been used before and has not been validated, so caution must be used when drawing conclusions.

This result contradicts work from more-southern states, where studies more frequently find that eastern yellow robin populations are declining (Barrett et al. 2003; BirdLife Australia 2015b). In Victoria, there is evidence that dispersal restrictions in more-fragmented landscapes are negatively impacting demographic processes and reducing allelic diversity in eastern yellow robin populations (Harrisson et al. 2012). When comparing the study region map produced by Harrisson et al. (2012) to that produced for my region (Figure 1.4 on page 15), it is apparent that the Victorian landscape does not contain the same web of linear woodland elements as my study region. Thus, the lack of an effect of connectivity in my study area may be because shade lines, roadside vegetation, and travelling stock routes provided a complex network of habitat corridors (travelling stock routes are interconnected grazing routes and reserves used by the pastoral industry to move stock on foot from one locality to another; Spooner et al. 2010; State of Queensland 2016d). However, habitat fragmentation has greater deleterious impacts on demography and gene flow as the spatial and/or temporal scale of fragmentation increases (Chapter 2). The Victorian landscape has both a longer history of deforestation (ECC 1997; Seabrook et al. 2006) and less structural connectivity than the Brigalow Belt South bioregion. Thus these impacts are confounded, making it difficult to separate out the importance of the connectivity provided by linear networks. Thus, it is unclear whether the high connectivity within the Brigalow Belt South facilitates sustainable levels of dispersal, or

129 whether dispersal rates are suboptimal and this will only become apparent over time as populations become increasingly affected.

The Stock Route Network Management Bill 2016 was introduced to Queensland Parliament on 3 November 2016. Despite it becoming more common for livestock to be moved by road and rail, stock routes continue to play an important role today, especially during times of drought (State of Queensland 2016a, 2016d). Stock routes are currently administered under several pieces of legislation and this new Bill is designed to amalgamate these disjointed pieces into one Act (State of Queensland 2016d). The current management strategy for the stock route system includes actions to dispose of (i.e. remove from the network and either sell or designate an alternative use of the land) redundant sections of the stock route (State of Queensland 2016a), and the Bill includes a clause to facilitate this (State of Queensland 2016d). While some of the stock route is highly degraded, much of it has high ecological value as remnant open woodland (Davidson et al. 2005; Lindenmayer et al. 2010; Lentini et al. 2011a). It is hypothesised that stock routes act as habitat corridors and facilitate dispersal; however, this is yet to be tested (Lentini et al. 2011a). I recommend that practitioners and policymakers err on the side of caution and preserve the structural connectivity provided by the stock route until more information is available about whether a causal relationship exists between the low rates of robin declines in the study region and the high degree of structural connectivity.

I found that habitat extent emerged as an important factor for robin occurrence (Chapter 3). Clearly, it is critical that no further remnant brigalow be cleared. Brigalow-dominated ecological communities are nationally listed as endangered; however, this protection only extends to remnant brigalow and brigalow regrowth greater than 15 years old (Commonwealth of Australia 2016). Small patches of brigalow regrowth less than 2 ha in size may also be cleared for grazing, regardless of their age (State of Queensland 2013). However, allowing patches of brigalow to regenerate could provide a cheap and feasible supplement to the existing woodlands. Robins readily inhabit brigalow regrowth of all ages (Bowen 2009), plus regrowth has high ecological value for the broader avian community (Bowen et al. 2009) and reptiles (Bruton et al. 2013). Brigalow readily resprouts on fallow land (Skerman 1953; Johnson 1964). I was unable to include patch width within my occupancy models due to its high correlation with woodland extent; however, I did observe that robins rarely occupied patches less than 75 m wide in the study area (Chapter 3). While many roadside verges appeared to be too narrow to support robins, landholders could take advantage of these verges by permitting narrow strips of regrowth to regenerate alongside these corridors so the total width was greater than 75 m. These regenerated strips may be especially effective when located where linear woodland strips intersect (Hall et al. 2016). These intersections contain more-

130 diverse avifauna assemblages, perhaps because they are energetically cheaper to forage within or defend against conspecifics than an elongated patch (Lack 1988; Hall et al. 2016).

7.4 LIMITATIONS OF THE STUDY AND SUGGESTED FUTURE RESEARCH

My work throughout this thesis was based upon the earlier findings by Maron et al. (2012) that robins had relatively higher H:L ratios in less-wooded sites and a desire to understand what was driving that pattern. In the time since the earlier work by Maron and colleagues, patterns of robin condition had shifted so condition was no longer related to woodland extent (Chapter 5). This shift meant that I could not determine the cause of the earlier pattern of condition with my study design. This problem highlights an important issue – the potentially transient nature of condition patterns. Thus, the timing of follow-up studies needs careful consideration and is dependent on the research question. Future work aimed at examining the cause of condition patterns needs to be conducted very soon after measuring condition, if not at the time of measuring condition. However, studies aimed at detecting changes in condition or predicting the consequences of condition patterns can be conducted over longer temporal scales.

While I found that prey density was not reduced in more-modified landscapes, this does not mean that birds in these landscapes are not being impacted by reduced food availability. There are several other ways that food availability could be impacted (Connor & McCoy 1979; Hinsley 2000). Substantial work has been invested in identifying how invertebrate populations respond to edge habitat (Helle & Muona 1985; Major et al. 2003; Ewers 2008). Given that more-fragmented landscapes contain more edge habitat (Forman & Godron 1986) and microclimate conditions are more desiccating within edge habitat than core habitat (Didham 1997; Laurance et al. 2011), prey availability could be lower at habitat edges. Although some prey taxa do show negative responses to edges (Laurance et al. 2011), generally, there is not strong support for this hypothesis (Didham 1997; Didham et al. 1998).

Other sources of food restrictions are less well studied and each of these needs further exploration. First, even if prey abundance per unit area remained constant across all degrees of woodland cover, small patches would contain less prey than large patches (Connor & McCoy 1979). If home ranges are restricted to single patches, this would impact the total prey available to the individual. Second, energetics could impact the cost of acquiring sufficient food (Hinsley 2000). This could be important in situations where individuals range across multiple habitat patches, or when patches are exceptionally long and narrow (Hinsley 2000; Hall et al. 2016). Third, as I suggested in Chapter 5,

131 increased conspecific density can compress home range size (Debus 2006), thereby reducing the foraging area available to the individual. Finally, competition from the broader avian community could influence the individual’s access to food (Barbaro et al. 2012); thus, competitor density could be an important factor to take into consideration.

The use of physiological measures as predictive tools has immense potential in the field of conservation (Wikelski & Cooke 2006; Ellis et al. 2012; Maute et al. 2015). While I found some intriguing results in terms of how H:L ratios can and cannot be used to predict changes to occurrence in my study system, my work raises some important issues. Key questions worth exploring include: 1) how does the duration and intensity of the stressor affect the predictive capacity of H:L ratios; 2) over what timeframes can reasonable predictions about a population be made; and 3) can H:L ratios be used as predictive tools when mechanisms other than food restriction cause a stressor; for example, thermal stress resulting from climate change?

7.5 CONCLUSION

The persistence of woodland-dependent avifauna is under threat due to the loss, fragmentation, and degradation of habitat (Olsen et al. 2005; Vié et al. 2009). To effectively preserve these diverse and important communities, conservation decision-makers need greater understanding of the mechanisms through which landscape change is causing declines (McGarigal & Cushman 2002; Watson 2011). Although reduced food availability in more-modified landscapes is thought to be negatively impacting avian insectivores (Burke & Nol 1998; Zanette et al. 2000; Watson 2011), this hypothesis has not been comprehensively tested. In this thesis, I demonstrated that prey density was not reduced in less-wooded landscapes and was unrelated to avian condition. However, avian distribution appeared to be in a state of flux with less-wooded landscapes becoming less populated. These findings contribute to our understanding of the drivers of population change in highly- modified agricultural landscapes. Additionally, I confirmed that patterns of stress indices show some potential for providing information about population threats and effectively signal changes that were underway in a population, but not yet evident through occurrence records. I suggested practical recommendations that could be implemented by land holders and policy makers to help conserve woodland biota in the Brigalow Belt South. My work provides a foundation for future empirical research to investigate alternative ways that food availability could be influencing insectivore populations in agricultural landscapes.

132

Plate 8. Sunset at Moonie, Queensland.

133 References

Aguilar A, Jessup DA, Estes J, Garza JC. 2008. The distribution of nuclear genetic variation and historical demography of sea otters. Animal Conservation 11:35-45. Al-Murrani WK, Al-Rawi IK, Raof NM. 2002. Genetic resistance to Salmonella typhimurium in two lines of chickens selected as resistant and sensitive on the basis of heterophil/lymphocyte ratio. British Poultry Science 43:501-507. Allentoft M, O’Brien J. 2010. Global amphibian declines, loss of genetic diversity and fitness: a review. Diversity 2:47-71. Amos W, Balmford A. 2001. When does conservation genetics matter? Heredity 87:257-265. Andrén H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71:355-366. Antos MJ, Bennett AF. 2006. Foraging ecology of ground-feeding woodland birds in temperate woodlands of southern Australia. Emu 106:29-40. AntWiki. 2016. Australian ants: identification keys. AntWiki. Available from http://www.antwiki.org/wiki/Australian_ants#Identification_Keys. Arnold KE, Ramsay SL, Henderson L, Larcombe SD. 2010. Seasonal variation in diet quality: antioxidants, invertebrates and blue tits Cyanistes caeruleus. Biological Journal of the Linnean Society 99:708-717. Attwood SJ, Maron M, House APN, Zammit C. 2008. Do arthropod assemblages display globally consistent responses to intensified agricultural land use and management? Global Ecology and Biogeography 17:585-599. Banbura J, Skwarska J, Banbura M, Gladalski M, Holysz M, Kalinski A, Markowski M, Wawrzyniak J, Zielinski P. 2013. Spatial and temporal variation in heterophil-to- lymphocyte ratios of nestling passerine birds: comparison of blue tits and great tits. PLoS ONE 8:e74226. Banks SC, Piggott MP, Stow AJ, Taylor AC. 2007. Sex and sociality in a disconnected world: a review of the impacts of habitat fragmentation on animal social interactions. Canadian Journal of Zoology 85:1065-1079. Barbaro L, Brockerhoff EG, Giffard B, van Halder I. 2012. Edge and area effects on avian assemblages and insectivory in fragmented native forests. Landscape Ecology 27:1451- 1463. Barrett G, Silcocks A, Barry S, Cunningham R, Poulter R. 2003. The new atlas of Australian birds. Royal Australasian Ornithologists Union, Hawthorne East, Australia.

134 Barrett GW, Ford HA, Recher HF. 1994. Conservation of woodland birds in a fragmented rural landscape. Pacific Conservation Biology 1:245-256. Barrett GW, Silcocks AF, Cunningham R, Oliver DL, Weston MA, Baker J. 2007. Comparison of atlas data to determine the conservation status of bird species in New South Wales, with an emphasis on woodland-dependent species. Australian Zoologist 34:37-77. Barrett RT, Erikstad KE, Sandvik H, Myksvoll M, Jenni-Eiermann S, Kristensen DL, Moum T, Reiertsen TK, Vikebø F. 2015. The stress hormone corticosterone in a marine top predator reflects short-term changes in food availability. Ecology and Evolution 5:1306-1317. Bartoń K. 2015. MuMIn: multi-model inference. R package version 1.15.1. Available from https://cran.r-project.org/package=MuMIn. Battin J. 2004. When good animals love bad habitats: ecological traps and the conservation of animal populations. Conservation Biology 18:1482-1491. Baum KA, Haynes KJ, Dillemuth FP, Cronin JT. 2004. The matrix enhances the effectiveness of corridors and stepping stones. Ecology 85:2671-2676. Bayard TS, Elphick CS. 2010. How area sensitivity in birds is studied. Conservation Biology 24:938-947. Bell GP. 1990. Birds and mammals on an insect diet: a primer on diet composition analysis in relation to ecological energetics. Pages 416-422 in Morrison ML, Ralph CJ, Verner J, and Jehl Jr. JR, editors. Studies in avian biology no. 13. Avian foraging: theory, methodology, and applications. Cooper Ornithological Society, Los Angeles, USA. Bell HL. 1985. Seasonal variation and the effects of drought on the abundance of arthropods in savanna woodland on the Northern Tablelands of New South Wales. Australian Journal of Ecology 10:207-221. Ben-Hamo M, Burns DJ, Bauchinger U, Mukherjee S, Embar K, Pinshow B. 2016. Behavioural responses during feather replacement in house sparrows. Journal of Avian Biology 47:103- 108. Bender DJ, Contreras TA, Fahrig L. 1998. Habitat loss and population decline: a meta-analysis of the patch size effect. Ecology 79:517-533. Bennett AF, Radford JQ, Haslem A. 2006. Properties of land mosaics: implications for nature conservation in agricultural environments. Biological Conservation 133:250-264. Bennett VA, Doerr VAJ, Doerr ED, Manning AD, Lindenmayer DB, Yoon H-J. 2012. Habitat selection and post-release movement of reintroduced brown treecreeper individuals in restored temperate woodland. PLoS ONE 7:e50612.

135 Bennett VA, Doerr VAJ, Doerr ED, Manning AD, Lindenmayer DB, Yoon H-J. 2013. Causes of reintroduction failure of the brown treecreeper: implications for ecosystem restoration. Austral Ecology 38:700-712. Betts MG, Fahrig L, Hadley AS, Halstead KE, Bowman J, Robinson WD, Wiens JA, Lindenmayer DB. 2014. A species-centered approach for uncovering generalities in organism responses to habitat loss and fragmentation. Ecography 37:517-527. Betts MG, Gutzwiller KJ, Smith MJ, Robinson WD, Hadley AS. 2015. Improving inferences about functional connectivity from animal translocation experiments. Landscape Ecology 30:585- 593. Bierregaard Jr. RO, Lovejoy TE, Kapos V, dos Santos AA, Hutchings RW. 1992. The biological dynamics of tropical rainforest fragments. Bioscience 42:859-866. BirdLife Australia. 2015a. The state of Australia's birds 2015: east coast. BirdLife Australia, Carlton, Australia. BirdLife Australia. 2015b. The state of Australia's birds 2015: south-eastern mainland. BirdLife Australia, Carlton, Australia. BirdLife Australia. 2015c. The state of Australia's birds 2015: headline trends for terrestrial birds. BirdLife Australia, Carlton, Australia. BirdLife Australia. 2015d. Measuring the state of Australia’s terrestrial birds: how Australian Bird Indices (ABIs) are made. BirdLife Australia, Carlton, Australia. BirdLife International. 2008. State of the world's birds: indicators for our changing world. BirdLife International, Cambridge, UK. BirdLife International. 2012. Eopsaltria australis. The IUCN Red List of Threatened Species 2012. International Union for Conservation of Nature, Cambridge, UK. Available from http://www.iucnredlist.org/details/22704849/0. Bjornstad ON. 2015. ncf: spatial nonparametric covariance functions. Available from http://cran.r- project.org/package=ncf. Blas J. 2015. Stress in birds. Pages 769-810 in Scanes CG, editor. Sturkie's avian physiology. Academic Press, London, UK. Bolliger J, Lander T, Balkenhol N. 2014. Landscape genetics since 2003: status, challenges and future directions. Landscape Ecology 29:361-366. Bonier F, Martin PR, Moore IT, Wingfield JC. 2009. Do baseline glucocorticoids predict fitness? Trends in Ecology & Evolution 24:634-642. Bonnot TW, Thompson III FR, Millspaugh JJ, Jones-Farrand DT. 2013. Landscape-based population viability models demonstrate importance of strategic conservation planning for birds. Biological Conservation 165:104-114.

136 Bonte D, et al. 2012. Costs of dispersal. Biological Reviews 87:290-312. Bonter DN, Bridge ES. 2011. Applications of radio frequency identification (RFID) in ornithological research: a review. Journal of Field Ornithology 82:1-10. Boulton RL, Richard Y, Armstrong DP. 2008. Influence of food availability, predator density and forest fragmentation on nest survival of New Zealand robins. Biological Conservation 141:580-589. Boutin S. 1990. Food supplementation experiments with terrestrial vertebrates: patterns, problems, and the future. Canadian Journal of Zoology 68:203-220. Bowen M. 2009. Quantifying the ecological values of brigalow regrowth for woodland birds: a hierarchical landscape approach. Ph.D. thesis. The University of Queensland, St Lucia, Australia. Bowen ME, McAlpine CA, Seabrook LM, House APN, Smith GC. 2009. The age and amount of regrowth forest in fragmented brigalow landscapes are both important for woodland dependent birds. Biological Conservation 142:3051-3059. Bowler DE, Benton TG. 2005. Causes and consequences of animal dispersal strategies: relating individual behaviour to spatial dynamics. Biological Reviews 80:205-225. Bowne DR, Bowers MA. 2004. Interpatch movements in spatially structured populations: a literature review. Landscape Ecology 19:1-20. Bregman TP, Şekercioğlu ÇH, Tobias JA. 2014. Global patterns and predictors of bird species responses to forest fragmentation: implications for ecosystem function and conservation. Biological Conservation 169:372-383. Brisson JA, Strasburg JL, Templeton AR. 2003. Impact of fire management on the ecology of collared lizard (Crotaphytus collaris) populations living on the Ozark Plateau. Animal Conservation 6:247-254. Bromham L, Cardillo M, Bennett AF, Elgar MA. 1999. Effects of stock grazing on the ground invertebrate fauna of woodland remnants. Austral Ecology 24:199-207. Brooks CP. 2003. A scalar analysis of landscape connectivity. Oikos 102:433-439. Brown CJ, Harborne AR, Paris CB, Mumby PJ. 2016. Uniting paradigms of connectivity in marine ecology. Ecology 97:2447-2457. Brown DR, Sherry TW. 2006. Food supply controls the body condition of a migrant bird wintering in the tropics. Oecologia 149:22-32. Brown JH, Kodric-Brown A. 1977. Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58:445-449. Bruton MJ, McAlpine CA, Maron M. 2013. Regrowth woodlands are valuable habitat for reptile communities. Biological Conservation 165:95-103.

137 Bureau of Meteorology. 2015. Climate data online. Bureau of Meteorology, Melbourne, Australia. Available from http://www.bom.gov.au/climate/data/. Burke DM, Nol E. 1998. Influence of food abundance, nest-site habitat, and forest fragmentation on breeding ovenbirds. Auk 115:96-104. Burnham KP, Anderson DR. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer-Verlag New York, Inc., New York, USA. Butcher JA, Morrison ML, Ransom Jr. D, Slack RD, Wilkins RN. 2010. Evidence of a minimum patch size threshold of reproductive success in an endangered songbird. Journal of Wildlife Management 74:133-139. Butt N, Beyer HL, Bennett JR, Biggs D, Maggini R, Mills M, Renwick AR, Seabrook LM, Possingham HP. 2013. Biodiversity Risks from Fossil Fuel Extraction. Science 342:425- 426. Byholm P, Kekkonen M. 2008. Food regulates reproduction differently in different habitats: experimental evidence in the goshawk. Ecology 89:1696-1702. Campbell TW. 1995. Avian hematology and cytology. Iowa State University Press, Ames, USA. Casner KL, Forister ML, Ram K, Shapiro AM. 2014. The utility of repeated presence data as a surrogate for counts: a case study using butterflies. Journal of Insect Conservation 18:13-27. Castelletta M, Sodhi NS, Subaraj R. 2000. Heavy extinctions of forest avifauna in Singapore: lessons for biodiversity conservation in Southeast Asia. Conservation Biology 14:1870- 1880. Charlesworth D, Willis JH. 2009. The genetics of inbreeding depression. Nature Reviews Genetics 10:783-796. Chen J, Saunders SC, Crow TR, Naiman RJ, Brosofske KD, Mroz GD, Brookshire BL, Franklin JF. 1999. Microclimate in forest ecosystem and landscape ecology: variations in local climate can be used to monitor and compare the effects of different management regimes. Bioscience 49:288-297. Chesser RT. 1998. Further perspectives on the breeding distribution of migratory birds: South American austral migrant flycatchers. Journal of Animal Ecology 67:69-77. Chevan A, Sutherland M. 1991. Hierarchical partitioning. American Statistician 45:90-96. Cicchetti, DV. 1994. Guidelines, criteria, and rules of thumb for evaluating normed and standardized assessment instruments in psychology. Psychological Assessment 6:284-290. Cīrule D, Krama T, Vrublevska J, Rantala MJ, Krams I. 2012. A rapid effect of handling on counts of white blood cells in a wintering passerine bird: a more practical measure of stress? Journal of Ornithology 153:161-166. Clark JA, May RM. 2002. Taxonomic bias in conservation research. Science 297:191-192.

138 Clark P, Boardman W, Raidal S. 2009. Atlas of clinical avian hematology. Wiley-Blackwell, Oxford, UK. Clark RW, Brown WS, Stechert R, Zamudio KR. 2010. Roads, interrupted dispersal, and genetic diversity in timber rattlesnakes. Conservation Biology 24:1059-1069. Class AM, Moore IT. 2013. Effects of food supplementation on a tropical bird. Oecologia 173:355- 362. Clinchy M, Haydon DT, Smith AT. 2002. Pattern does not equal process: what does patch occupancy really tell us about metapopulation dynamics? American Naturalist 159:351-362. Clinchy M, Zanette L, Boonstra R, Wingfield JC, Smith JNM. 2004. Balancing food and predator pressure induces chronic stress in songbirds. Proceedings of the Royal Society B 271:2473- 2479. Cohen J. 1988. Statistical power analysis for the behavioral sciences. Lawrence Erlbaum Associates, Hillsdale, USA. Collins RJ, Barrett GW. 1997. Effects of habitat fragmentation on meadow vole (Microtus pennsylvanicus) population dynamics in experiment landscape patches. Landscape Ecology 12:63-76. Commonwealth of Australia. 2016. EPBC Act list of threatened ecological communities. Commonwealth of Australia, Department of the Environment. Available from http://www.environment.gov.au/cgi-bin/sprat/public/publiclookupcommunities.pl. Connor EF, McCoy ED. 1979. Statistics and biology of the species-area relationship. American Naturalist 113:791-833. Cooke SJ, Sack L, Franklin CE, Farrell AP, Beardall J, Wikelski M, Chown SL. 2013. What is conservation physiology? Perspectives on an increasingly integrated and essential science. Conservation Physiology 1:cot001. Cooper CB, Walters JR. 2002. Experimental evidence of disrupted dispersal causing decline of an Australian passerine in fragmented habitat. Conservation Biology 16:471-478. Cooper RJ, Martinat PJ, Whitmore RC. 1990. Dietary similarity among insectivorous birds: influence of taxonomic verses ecological categorization of prey. Pages 104-109 in Morrison ML Ralph CJ, Verner J, and Jehl Jr. JR, editors. Studies in avian biology no. 13. Avian foraging: theory, methodology, and applications. Cooper Ornithological Society, Los Angeles, USA. Corlett RT. 2013. Where are the subtropics? Biotropica 45:273-275. Costa DP, Sinervo B. 2004. Field physiology: physiological insights from animals in nature. Annual Review of Physiology 66:209-238.

139 Coulon A, Fitzpatrick JW, Bowman R, Lovette IJ. 2010. Effects of habitat fragmentation on effective dispersal of Florida scrub-jays. Conservation Biology 24:1080-1088. Cousin JA. 2007. Habitat selection by ground-foraging robins in eastern Australia: scale dependence and hierarchical selection. Ph.D. thesis. University of New England, Armidale, Australia. Croft P, Reid N, Hunter JT. 2010. Experimental burning changes the quality of fallen timber as habitat for vertebrate and invertebrate fauna: implications for fire management. Wildlife Research 37:574-581. Cucco M, Malacarne G. 1997. The effect of supplemental food on time budget and body condition in the Black Redstart Phoenicurus ochruros. Ardea 85:211-221. Cullen Jr. L, Stanton JC, Lima F, Uezu A, Perilli MLL, Akçakaya HR. 2016. Implications of fine- grained habitat fragmentation and road mortality for jaguar conservation in the Atlantic Forest, Brazil. PLoS ONE 11:e0167372. Cushman SA. 2006. Effects of habitat loss and fragmentation on amphibians: a review and prospectus. Biological Conservation 128:231-240. Dale S. 2001. Female-biased dispersal, low female recruitment, unpaired males, and the extinction of small and isolated bird populations. Oikos 92:344-356. Dantzer B, Fletcher QE, Boonstra R, Sheriff MJ. 2014. Measures of physiological stress: a transparent or opaque window into the status, management and conservation of species? Conservation Physiology 2:cou023. Davidson I, Scammell A, O'Shannassy P, Mullins M, Learmonth S. 2005. Travelling stock reserves: refuges for stock and biodiversity? Ecological Management & Restoration 6:5-15. Davis AK. 2005. Effect of handling time and repeated sampling on avian white blood cell counts. Journal of Field Ornithology 76:334-338. Davis AK, Maney DL, Maerz JC. 2008. The use of leukocyte profiles to measure stress in vertebrates: a review for ecologists. Functional Ecology 22:760-772. Davis GS, Anderson KE, Carroll AS. 2000. The effects of long-term caging and molt of single comb white leghorn hens on herterophil to lymphocyte ratios, corticosterone and thyroid hormones. Poultry Science 79:514-518. de Araújo ASF, Eisenhauer N, Nunes L, Leite LFC, Cesarz S. 2015. Soil surface-active fauna in degraded and restored lands of northeast Brazil. Land Degradation & Development 26:1-8. Debinski DM, Holt RD. 2000. A survey and overview of habitat fragmentation experiments. Conservation Biology 14:342-355. Debus SJS. 2006. Breeding and population parameters of robins in a woodland remnant in northern New South Wales, Australia. Emu 106:147-156.

140 Deikumah JP, McAlpine CA, Maron M. 2015. Matrix intensification affects body and physiological condition of tropical forest-dependent passerines. PLoS ONE 10:e0128521. Delaney KS, Riley SPD, Fisher RN. 2010. A rapid, strong, and convergent genetic response to urban habitat fragmentation in four divergent and widespread vertebrates. PLoS ONE 5:e12767. Desrochers A, Hannon SJ. 1997. Gap crossing decisions by forest songbirds during the post- fledging period. Conservation Biology 11:1204-1210. Desy EA, Batzli GO. 1989. Effects of food availability and predation on prairie vole demography: a field experiment. Ecology 70:411-421. Diamond JM. 1972. Biogeographic kinetics: estimation of relaxation times for avifaunas of southwest Pacific islands. Proceedings of the National Academy of Sciences USA 69:3199- 3203. Didham RK. 1997. An overview of invertebrate responses to forest fragmentation. Pages 302-320 in Watt AD, Stork NE, and Hunter MD, editors. Forests and Insects. Chapman & Hall, London, UK. Didham RK, Hammond PM, Lawton JH, Eggleton P, Stork NE. 1998. Beetle species responses to tropical forest fragmentation. Ecological Monographs 68:295-323. Doerr VAJ, Doerr ED, Davies MJ. 2011. Dispersal behaviour of brown treecreepers predicts functional connectivity for several other woodland birds. Emu 111:71-83. Drees C, Hüfner S, Matern A, Nève G, Assmann T. 2011. Repeated sampling detects gene flow in a flightless ground beetle in a fragmented landscape. Hereditas 148:36-45. Driscoll DA, Banks SC, Barton PS, Lindenmayer DB, Smith AL. 2013. Conceptual domain of the matrix in fragmented landscapes. Trends in Ecology & Evolution 28:605-613. DSEWPC. 2012. Brigalow (Acacia harpophylla dominant and co-dominant). Department of Sustainability, Environment, Water, Population and Communities, Canberra, Australia. Available from http://www.environment.gov.au/biodiversity/threatened/communities/ brigalow.html. DSEWPC. 2013. National Reserve System. Department of Sustainability, Environment, Water, Population and Communities, Canberra, Australia. Available from http://www.environment.gov.au/parks/nrs/index.html. Duncan DH, Dorrough J, White M, Moxham C. 2008. Blowing in the wind? Nutrient enrichment of remnant woodlands in an agricultural landscape. Landscape Ecology 23:107-119. Dunning JB, Danielson BJ, Pulliam HR. 1992. Ecological processes that affect populations in complex landscapes. Oikos 65:169-175.

141 Dwernychuk LW, Boag DA. 1972. Ducks nesting in association with gulls - an ecological trap? Canadian Journal of Zoology 50:559-563. Eberhardt LL, Thomas JM. 1991. Designing environmental field studies. Ecological Monographs 61:53-73. ECC. 1997. Box-ironbark forests and woodlands investigation. Resources and issues report. Environment Conservation Council, Melbourne, Australia. Eccard JA, Pusenius J, Sundell J, Halle S, Ylönen H. 2008. Foraging patterns of voles at heterogeneous avian and uniform mustelid predation risk. Oecologia 157:725-734. Edge CB, Fortin MJ, Jackson DA, Lawrie D, Stanfield L, Shrestha N. 2017. Habitat alteration and habitat fragmentation differentially affect beta diversity of stream fish communities. Landscape Ecology 32:647-662. Eikenaar C, Berg ML, Komdeur J. 2003. Experimental evidence for the influence of food availability on incubation attendance and hatching asynchrony in the Australian reed warbler Acrocephalus australis. Journal of Avian Biology 34:419-427. Ellis RD, McWhorter TJ, Maron M. 2012. Integrating landscape ecology and conservation physiology. Landscape Ecology 27:1-12. ESRI. 2014. ArcGIS desktop version 10.2.1. Environmental Systems Research Institute, Redlands, USA. Available from www.esri.com. Evans MC. 2016. Deforestation in Australia: drivers, trends and policy responses. Pacific Conservation Biology 22:130-150. Ewers RM. 2008. Spatio-temporal variation in mortality rates of Mecodema spp. (Coleoptera: Carabidae) across a forest-grassland edge in New Zealand. Insect Conservation and Diversity 1:40-47. Ewers RM, Didham RK. 2006. Confounding factors in the detection of species responses to habitat fragmentation. Biological Reviews 81:117-142. Ewers RM, et al. 2011. A large-scale forest fragmentation experiment: the Stability of Altered Forest Ecosystems Project. Philosophical Transactions of the Royal Society B 366:3292- 3302. Eyre TJ, Maron M, Mathieson MT, Haseler M. 2009. Impacts of grazing, selective logging and hyper-aggressors on diurnal bird fauna in intact forest landscapes of the Brigalow Belt, Queensland. Austral Ecology 34:705-716. Fahrig L. 2002. Effect of habitat fragmentation on the extinction threshold: a synthesis. Ecological Applications 12:346-353. Fahrig L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics 34:487-515.

142 Fahrig L. 2017. Ecological responses to habitat fragmentation per se. Annual Reviews of Ecology, Evolution and Systematics 48, DOI: 10.1146/annurev-ecolsys-110316-022612. Fahrig L, Merriam G. 1985. Habitat patch connectivity and population survival. Ecology 66:1762- 1768. FAO. 2010. Global forest resources assessment 2010: main report. Food and Agriculture Organization of the United Nations, Rome, Italy. Fensham RJ, Fairfax RJ. 2003. A land management history for central Queensland, Australia as determined from land-holder questionnaire and aerial photography. Journal of Environmental Management 68:409-420. Field A, Miles J, Field Z. 2012. Discovering statistics using R. SAGE Publications Ltd, London, UK. Firth D. 1993. Bias reduction of maximum likelihood estimates. Biometrika 80:27-38. Fischer J, Lindenmayer DB. 2002. The conservation value of paddock trees for birds in a variegated landscape in southern New South Wales. 2. Paddock trees as stepping stones. Biodiversity and Conservation 11:833-849. Fischer J, Lindenmayer DB. 2007. Landscape modification and habitat fragmentation: a synthesis. Global Ecology and Biogeography 16:265-280. Fischer J, Lindenmayer DB, Fazey I. 2004. Appreciating ecological complexity: habitat contours as a conceptual landscape model. Conservation Biology 18:1245-1253. Ford HA. 2011. The causes of decline of birds of eucalypt woodlands: advances in our knowledge over the last 10 years. Emu 111:1-9. Ford HA, Barrett GW, Saunders DA, Recher HF. 2001. Why have birds in the woodlands of southern Australia declined? Biological Conservation 97:71-88. Ford HA, Noske S, Bridges L. 1986. Foraging of birds in eucalypt woodland in north-eastern New South Wales. Emu 86:168-179. Forman RTT. 1995. Land mosaics. The ecology of landscapes and regions. Cambridge University Press, Cambridge, UK. Forman RTT, Godron M. 1986. Landscape ecology. John Wiley & Sons, New York, USA. Fox J, Weisberg S. 2011. An R companion to applied regression, second edition. Sage, Thousand Oaks, USA. Frankham R. 2005. Genetics and extinction. Biological Conservation 126:131-140. Freeman EA, Moisen G. 2008. PresenceAbsence: an R package for presence absence analysis. Journal of Statistical Software 23:1-31. Frith D, Frith C. 1990. Seasonality of litter invertebrate populations in an Australian upland tropical rainforest. Biotropica 22:181-190.

143 Fudge AM, editor. 2000. Laboratory medicine: avian and exotic pets. W.B. Saunders Company, Philadelphia, USA. García RR, García U, Osoro K, Celaya R. 2011. Ground-dwelling arthropod assemblages of partially improved heathlands according to the species of grazer and grazing regime. European Journal of Entomology 108:107-115. Gaston KJ, Fuller RA. 2008. Commonness, population depletion and conservation biology. Trends in Ecology & Evolution 23:14-19. Gaston KJ, May RM. 1992. Taxonomy of taxonomists. Nature 356:281-282. Geldmann J, Heilmann-Clausen J, Holm TE, Levinsky I, Markussen B, Olsen K, Rahbek C, Tøttrup AP. 2016. What determines spatial bias in citizen science? Exploring four recording schemes with different proficiency requirements. Diversity and Distributions 22:1139-1149. Gelman A. 2008. Scaling regression inputs by dividing by two standard deviations. Statistics in Medicine 27:2865-2873. Gelman A, Su Y-S. 2015. arm: data analysis using regression and multilevel/hierarchical models. R package version 1.8-6. Available from https://CRAN.R-project.org/package=arm. Godet L, Gaüzere P, Jiguet F, Devictor V. 2015. Dissociating several forms of commonness in birds sheds new light on biotic homogenization. Global Ecology and Biogeography 24:416-426. Goessling JM, Kennedy H, Mendonça MT, Wilson AE. 2015. A meta-analysis of plasma corticosterone and heterophil:lymphocyte ratios - is there conservation of physiological stress responses over time? Functional Ecology 29:1189-1196. Golabek KA, Ridley AR, Radford AN. 2012. Food availability affects strength of seasonal territorial behaviour in a cooperatively breeding bird. Animal Behaviour 83:613-619. Gonzalez A, Chaneton EJ. 2002. Heterotroph species extinction, abundance and biomass dynamics in an experimentally fragmented microecosystem. Journal of Animal Ecology 71:594-602. Gray MA, Baldauf SL, Mayhew PJ, Hill JK. 2007. The response of avian feeding guilds to tropical forest disturbance. Conservation Biology 21:133-141. Green RH. 1993. Application of repeated measures designs in environmental impact and monitoring studies. Australian Journal of Ecology 18:81-98. Greenslade PJM. 1964. Pitfall trapping as a method for studying populations of Carabidae (Coleoptera). Journal of Animal Ecology 33:301-310. Greenslade PJM. 1973. Sampling ants with pitfall traps: digging-in effects. Insectes Sociaux 20:343-353. Greenslade PJM. 1974. Distribution of two forms of the meat ant, Iridomyrmex purpureus (Hymenoptera: Formicidae), in parts of South Australia. Australian Journal of Zoology 22:489-504.

144 Greenwood PJ, Harvey PH. 1982. The natal and breeding dispersal of birds. Annual Review of Ecology and Systematics 13:1-21. Griffiths RA, Dos Santos M. 2012. Trends in conservation biology: Progress or procrastination in a new millennium? Biological Conservation 153:153-158. Guisan A, Thuiller W. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993-1009. Hagar JC, Dugger KM, Starkey EE. 2007. Arthropod prey of Wilson's warblers in the understory of douglas-fir forests. Wilson Journal of Ornithology 119:533-546. Hall M, Nimmo D, Bennett AF. 2016. At the crossroads: does the configuration of roadside vegetation affect woodland bird communities in rural landscapes? PLoS ONE 11:e0155219. Halley JM, Monokrousos N, Mazaris AD, Newmark WD, Vokou D. 2016. Dynamics of extinction debt across five taxonomic groups. Nature Communications 7:12283. Hallmann CA, Foppen RPB, van Turnhout CAM, de Kroon H, Jongejans E. 2014. Declines in insectivorous birds are associated with high neonicotinoid concentrations. Nature 511:341- 343. Hanski I, Ovaskainen O. 2002. Extinction debt at extinction threshold. Conservation Biology 16:666-673. Hamilton WD, May RM. 1977. Dispersal in stable habitats. Nature 269:578-581. Hansen MC, et al. 2013. High-resolution global maps of 21st-century forest cover change. Science 342:850-853. Hanski I. 1991. Single-species metapopulation dynamics: concepts, models and observations. Biological Journal of the Linnean Society 42:17-38. Hansson B, Bensch S, Hasselquist D. 2004. Lifetime fitness of short- and long-distance dispersing great reed warblers. Evolution 58:2546-2557. Harris RC, editor. 1987. Historical atlas of Canada: the land transformed, 1800-1891. University of Toronto Press, Toronto, Canada. Harrisson KA, Pavlova A, Amos JN, Takeuchi N, Lill A, Radford JQ, Sunnucks P. 2012. Fine-scale effects of habitat loss and fragmentation despite large-scale gene flow for some regionally declining woodland bird species. Landscape Ecology 27:813-827. Harrisson KA, Pavlova A, Amos JN, Takeuchi N, Lill A, Radford JQ, Sunnucks P. 2013. Disrupted fine-scale population processes in fragmented landscapes despite large-scale genetic connectivity for a widespread and common cooperative breeder: the superb fairy-wren (Malurus cyaneus). Journal of Animal Ecology 82:322-333. Haskell DG. 2000. Effects of forest roads on macroinvertebrate soil fauna of the Southern Appalachian mountains. Conservation Biology 14:57-63.

145 Hastie T. 2015. gam: generalized additive models. R package version 1.12. Available from https://CRAN.R-project.org/package=gam. Haylock J. 1985. The use of time and energy by the brown thornbill and yellow robin in a temperate wet forest in winter. Honours thesis. Monash University, Melbourne, Australia. Hedrick PW, Garcia-Dorado A. 2016. Understanding inbreeding depression, purging, and genetic rescue. Trends in Ecology & Evolution 31:940-952. Helle P, Muona J. 1985. Invertebrate numbers in edges between clear-fellings and mature forests in northern Finland. Silva Fennica 19:281-294. Helm A, Hanski I, Pärtel M. 2006. Slow response of plant species richness to habitat loss and fragmentation. Ecology Letters 9:72-77. Henein K, Merriam G. 1990. The elements of connectivity where corridor quality is variable. Landscape Ecology 4:157-170. Herring G, Cook MI, Gawlik DE, Call EM. 2011. Food availability is expressed through physiological stress indicators in nestling white ibis: a food supplementation experiment. Functional Ecology 25:682-690. Higgins PJ, Peter JM, editors. 2002. Handbook of Australian, New Zealand and Antarctic birds. Volume 6: pardalotes to shrike-thrushes. Oxford University Press, Melbourne, Australia. Hinam HL, St. Clair CC. 2008. High levels of habitat loss and fragmentation limit reproductive success by reducing home range size and provisioning rates of northern saw-whet owls. Biological Conservation 141:524-535. Hinsley SA. 2000. The costs of multiple patch use by birds. Landscape Ecology 15:765-775. Hoffmann M, et al. 2010. The impact of conservation on the status of the world's vertebrates. Science 330:1503-1509. Hoffmann BD, Andersen AN. 2003. Responses of ants to disturbance in Australia, with particular reference to functional groups. Austral Ecology 28:444-464. Holderegger R, Wagner HH. 2008. Landscape genetics. Bioscience 58:199-207. Holland GJ, Bennett AF. 2011. Recolonization of forest fragments by a native rodent following experimental 'extinctions'. Austral Ecology 36:521-529. Holmes RV. 1955. Fireside fancies. Edward Brothers, Inc., Ann Arbor, USA. Holsinger KE, Weir BS. 2009. Genetics in geographically structured populations: defining,

estimating and interpreting FST. Nature Reviews Genetics 10:639-650. Holyoak M, Casagrandi R, Nathan R, Revilla E, Spiegel O. 2008. Trends and missing parts in the study of movement ecology. Proceedings of the National Academy of Sciences USA 105:19060-19065.

146 Huggett AJ. 2005. The concept and utility of 'ecological thresholds' in biodiversity conservation. Biological Conservation 124:301-310. Hurlbert SH. 1984. Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54:187-211. Hylander K, Ehrlén J. 2013. The mechanisms causing extinction debts. Trends in Ecology & Evolution 28:341-346. Isaac NJB, Pocock MJO. 2015. Bias and information in biological records. Biological Journal of the Linnean Society 115:522-531. Jackson HB, Fahrig L. 2012. What size is a biologically relevant landscape? Landscape Ecology 27:929-941. Janin A, Léna J-P, Joly P. 2011. Beyond occurrence: body condition and stress hormone as integrative indicators of habitat availability and fragmentation in the common toad. Biological Conservation 144:1008-1016. Jäntti A, Aho T, Hakkarainen H, Kuitunen M, Suhonen J. 2001. Prey depletion by the foraging of the Eurasian treecreeper, Certhia familiaris, on tree-trunk arthropods. Oecologia 128:488- 491. Jeltsch F, et al. 2013. Integrating movement ecology with biodiversity research - exploring new avenues to address spatiotemporal biodiversity dynamics. Movement Ecology 1:6. Jenerette GD, Shen WJ. 2012. Experimental landscape ecology. Landscape Ecology 27:1237-1248. Jiménez JA, Hughes KA, Alaks G, Graham L, Lacy RC. 1994. An experimental study of inbreeding depression in a natural environment. Science 266:271-273. Johnson DH. 2002. The importance of replication in wildlife research. Journal of Wildlife Management 66:919-932. Johnson ML, Gaines MS. 1990. Evolution of dispersal: theoretical models and empirical tests using birds and mammals. Annual Review of Ecology and Systematics 21:449-480. Johnson RW. 1964. Ecology and control of brigalow in Queensland. Queensland Department of Primary Industries, Brisbane, Australia. Johnstone CP, Reina RD, Lill A. 2012. Interpreting indices of physiological stress in free-living vertebrates. Journal of Comparative Physiology B 182:861-879. Joseph LN, Field SA, Wilcox C, Possingham HP. 2006. Presence-absence versus abundance data for monitoring threatened species. Conservation Biology 20:1679-1687. Kang W, Minor ES, Woo D, Lee D, Park CR. 2016. Forest mammal roadkills as related to habitat connectivity in protected areas. Biodiversity and Conservation 25:2673-2686. Karr JR, Freemark KE. 1983. Habitat selection and environmental gradients: dynamics in the "stable" tropics. Ecology 64:1481-1494.

147 Kattan GH, Alvarez-López H, Giraldo M. 1994. Forest fragmentation and bird extinctions: San Antonio eighty years later. Conservation Biology 8:138-146. Kazemi F, Beecham S, Gibbs J, Clay R. 2009. Factors affecting terrestrial invertebrate diversity in bioretention basins in an Australian urban environment. Landscape and Urban Planning 92:304-313. Keinath DA, Doak DF, Hodges KE, Prugh LR, Fagan W, Sekercioglu CH, Buchart SHM, Kauffman M. 2017. A global analysis of traits predicting species sensitivity to habitat fragmentation. Global Ecology and Biogeography 26:115-127. Kenney J, Allendorf FW, McDougal C, Smith JLD. 2014. How much gene flow is needed to avoid inbreeding depression in wild tiger populations? Proceedings of the Royal Society B 281:20133337. Keyghobadi N. 2007. The genetic implications of habitat fragmentation for animals. Canadian Journal of Zoology 85:1049-1064. Kilgas P, Tilgar V, Mänd R. 2006. Hematological health state indices predict local survival in a small passerine bird, the great tit (Parus major). Physiological and Biochemical Zoology 79:565-572. King KL, Hutchinson KJ. 1983. The effects of sheep grazing on invertebrate numbers and biomass in unfertilized natural pastures of the New England Tablelands (NSW). Australian Journal of Ecology 8:245-255. Kishné AS, Morgan CLS, Neely HL. 2014. How much surface water can gilgai microtopography capture? Journal of Hydrology 513:256-261. Kitaysky AS, Kitaiskaia EV, Wingfield JC, Piatt JF. 2001. Dietary restriction causes chronic elevation of corticosterone and enhances stress response in red-legged kittiwake chicks. Journal of Comparative Physiology B 171:701-709. Kitaysky AS, Piatt JF, Wingfield JC. 2007. Stress hormones link food availability and population processes in seabirds. Marine Ecology Progress Series 352:245-258. Kosmidis I. 2013. brglm: bias reduction in binomial-response generalized linear models. R package version 0.5-9. Available from www.cran.r-project.org/package=brglm. Krauss J, et al. 2010. Habitat fragmentation causes immediate and time-delayed biodiversity loss at different trophic levels. Ecology Letters 13:597-605. Kulaszewicz I, Wojczulanis-Jakubas K, Jakubas D. 2015. Variation of the Savi's warbler (Locustella luscinioides) leucocyte profiles and body condition in relation to age, sex and moult. Annales Zoologici Fennici 52:325-338. Kuussaari M, et al. 2009. Extinction debt: a challenge for biodiversity conservation. Trends in Ecology & Evolution 24:564-571.

148 Lack PC. 1988. Hedge intersections and breeding bird distribution in farmland. Bird Study 35:133- 136. Lang A. 2000. The pitfalls of pitfalls: a comparison of pitfall trap catches and absolute density estimates of epigeal invertebrate predators in arable land. Journal of Pest Science 73:99-106. Langlois JP, Fahrig L, Merriam G, Artsob H. 2001. Landscape structure influences continental distribution of hantavirus in deer mice. Landscape Ecology 16:255-266. Lassau SA, Hochuli DF. 2004. Effects of habitat complexity on ant assemblages. Ecography 27:157-164. Laurance WF, et al. 2011. The fate of Amazonian forest fragments: a 32-year investigation. Biological Conservation 144:56-67. Lawler JJ, et al. 2006. Conservation science: a 20-year report card. Frontiers in Ecology and the Environment 4:473-480. Lecomte J, Boudjemadi K, Sarrazin F, Cally K, Clobert J. 2004. Connectivity and homogenisation of population sizes: an experimental approach in Lacerta vivipara. Journal of Animal Ecology 73:179-189. Legge S, Garnett S, Maute K, Heathcote J, Murphy S, Woinarski JCZ, Astheimer L. 2015. A landscape-scale, applied fire management experiment promotes recovery of a population of the threatened Gouldian finch, Erythrura gouldiae, in Australia's tropical savannas. PLoS ONE 10:e0137997. Lennox R, Cooke SJ. 2014. State of the interface between conservation and physiology: a bibliometric analysis. Conservation Physiology 2:cou003. Lentini PE, Fischer J, Gibbons P, Lindenmayer DB, Martin TG. 2011a. Australia’s Stock Route Network: 1. A review of its values and implications for future management. Ecological Management & Restoration 12:119-127. Lentini PE, Fischer J, Gibbons P, Hanspach J, Martin TG. 2011b. Value of large-scale linear networks for bird conservation: a case study from travelling stock routes, Australia. Agriculture Ecosystems & Environment 141:302-309. Lessells CM, Boag PT. 1987. Unrepeatable repeatabilities: a common mistake. Auk 104:116-121. Levings SC, Windsor DM. 1984. Litter moisture content as a determinant of litter arthropod distribution and abundance during the dry season on Barro Colorado Island, Panama. Biotropica 16:125-131. Lieberman S, Dock CF. 1982. Analysis of the leaf litter arthropod fauna of a lowland evergreen forest site (La Selva, Costa Rica). Revista de Biologia Tropical 30:27-37. Lima SL, Dill LM. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68:619-640.

149 Lindenmayer DB, Cunningham R, Crane M, Michael D, Montague-Drake R. 2007. Farmland bird responses to intersecting replanted areas. Landscape Ecology 22:1555-1562. Lindenmayer DB, Cunningham RB, Crane M, Montague-Drake R, Michael D. 2010. The importance of temperate woodland in travelling stock reserves for vertebrate biodiversity conservation. Ecological Management & Restoration 11:27-30. Lindenmayer DB, Fischer J. 2006. Habitat fragmentation and landscape change: an ecological and conservation synthesis. Island Press, Washington, USA. Lobato E, Moreno J, Merino S, Sanz JJ, Arriero E. 2005. Haematological variables are good predictors of recruitment in nestling pied flycatchers (Ficeduld hypoleuca). Ecoscience 12:27-34. Lobo JM, Jiménez-Valverde A, Real R. 2008. AUC: a misleading measure of the performance of predictive distribution models. Global Ecology and Biogeography 17:145-151. Lowe WH, Allendorf FW. 2010. What can genetics tell us about population connectivity? Molecular Ecology 19:3038-3051. Luck GW. 2003. Differences in the reproductive success and survival of the rufous treecreeper (Climacteris rufa) between a fragmented and unfragmented landscape. Biological Conservation 109:1-14. Lunt ID, Spooner PG. 2005. Using historical ecology to understand patterns of biodiversity in fragmented agricultural landscapes. Journal of Biogeography 32:1859-1873. Mac Nally R, Bennett AF. 1997. Species-specific predictions of the impact of habitat fragmentation: local extinction of birds in the box-ironbark forests of central Victoria, Australia. Biological Conservation 82:147-155. Mac Nally R, Bennett AF, Horrocks G. 2000. Forecasting the impacts of habitat fragmentation. Evaluation of species-specific predictions of the impact of habitat fragmentation on birds in the box-ironbark forests of central Victoria, Australia. Biological Conservation 95:7-29. Mac Nally R, Bennett AF, Thomson JR, Radford JQ, Unmack G, Horrocks G, Vesk PA. 2009. Collapse of an avifauna: climate change appears to exacerbate habitat loss and degradation. Diversity and Distributions 15:720-730. Mac Nally R, Brown GW. 2001. Reptiles and habitat fragmentation in the box-ironbark forests of central Victoria, Australia: predictions, compositional change and faunal nestedness. Oecologia 128:116-125. MacCracken JG, Stebbings JL. 2012. Test of a body condition index with amphibians. Journal of Herpetology 46:346-350. Madliger CL, et al. 2016. Success stories and emerging themes in conservation physiology. Conservation Physiology 4:cov057.

150 Madsen T, Shine R, Olsson M, Wittzell H. 1999. Restoration of an inbred adder population. Nature 402:34-35. Magle SB, Hunt VM, Vernon M, Crooks KR. 2012. Urban wildlife research: past, present, and future. Biological Conservation 155:23-32. Majer JD, Recher HF, Graham R, Gupta R. 2003. Trunk invertebrate faunas of Western Australian forests and woodlands: influence of tree species and season. Austral Ecology 28:629-641. Major RE, Christie FJ, Gowing G. 2001. Influence of remnant and landscape attributes on Australian woodland bird communities. Biological Conservation 102:47-66. Major RE, Christie FJ, Gowing G, Cassis G, Reid CAM. 2003. The effect of habitat configuration on arboreal insects in fragmented woodlands of south-eastern Australia. Biological Conservation 113:35-48. Manel S, Gaggiotti OE, Waples RS. 2005. Assignment methods: matching biological questions with appropriate techniques. Trends in Ecology & Evolution 20:136-142. Manning AD, Lindenmayer DB, Nix HA. 2004. Continua and Umwelt: novel perspectives on viewing landscapes. Oikos 104:621-628. March WA, Watson DM. 2007. Parasites boost productivity: effects of mistletoe on litterfall dynamics in a temperate Australian forest. Oecologia 154:339-347. Marchant S. 1984. Nest-records of the eastern yellow robin Eopsaltia australis. Emu 84:167-174. Marchant S. 1987. Territorialism and co-operative breeding of the eastern yellow robin Eopsaltria australis. Corella 11:6-14. Maron M, Goulding W, Ellis RD, Mohd-Taib F-S. 2012. Distribution and individual condition reveal a hierarchy of habitat suitability for an area-sensitive passerine. Biodiversity and Conservation 21:2509-2523. Maron M, Grey MJ, Catterall CP, Major RE, Oliver DL, Clarke MF, Loyn RH, Mac Nally R, Davidson I, Thomson JR. 2013. Avifaunal disarray due to a single despotic species. Diversity and Distributions 19:1468-1479. Maron M, Lill A. 2005. The influence of livestock grazing and weed invasion on habitat use by birds in grassy woodland remnants. Biological Conservation 124:439-450. Maron M, Lill A. 2006. Intraspecific variation in detection of bird-habitat relationships: declining birds in southern Australian woodlands. Pacific Conservation Biology 12:301-312. Maron M, Lill A, Watson DM, Mac Nally R. 2005. Temporal variation in bird assemblages: how representative is a one-year snapshot? Austral Ecology 30:383-394. Maron M, McAlpine CA, Watson JEM, Maxwell S, Barnard P. 2015. Climate-induced resource bottlenecks exacerbate species vulnerability: a review. Diversity and Distributions 21:731- 743.

151 Marra PP, Holberton RL. 1998. Corticosterone levels as indicators of habitat quality: effects of habitat segregation in a migratory bird during the non-breeding season. Oecologia 116:284- 292. Marsden CD, et al. 2012. Spatial and temporal patterns of neutral and adaptive genetic variation in the endangered African wild dog (Lycaon pictus). Molecular Ecology 21:1379-1393. Martin TG, Nally S, Burbidge AA, Arnall S, Garnett ST, Hayward MW, Lumsden LF, Menkhorst P, McDonald-Madden E, Possingham HP. 2012. Acting fast helps avoid extinction. Conservation Letters 5:274-280. Matlack GR. 1993. Microenvironment variation within and among forest edge sites in the eastern United States. Biological Conservation 66:185-194. Maute K, French K, Legge S, Astheimer L, Garnett S. 2015. Condition index monitoring supports conservation priorities for the protection of threatened grass-finch populations. Conservation Physiology 3:cov025. Maxwell MH, Hocking PM, Robertson GW. 1992. Differential leucocyte responses to various degrees of food restriction in broilers, turkeys and ducks. British Poultry Science 33:177- 187. Maxwell SL, Fuller RA, Brooks TM, Watson JEM. 2016. The ravages of guns, nets and bulldozers. Nature 536:143-145. Mazerolle MJ. 2016. AICcmodavg: model selection and multimodel inference based on (Q)AIC(c). R package version 2.0-4. Available from http://CRAN.R-project.org/package=AICcmodavg. McAlpine C, Sutcliffe T, Taylor K. 2002. One hundred and fifty years of landscape change for two subregions of the Southern Brigalow: patterns and management implications. Pages 27-41 in Franks AJ, Playford J, and Shapcott A, editors. Landscape health of Queensland. The Royal Society of Queensland, St Lucia, Australia. McDonald L. 1988. Cattle country: the beef cattle Industry in Central Queensland 1850s - 1980s. Boolarong Publications, Brisbane, Australia. McGarigal K, Cushman SA. 2002. Comparative evaluation of experimental approaches to the study of habitat fragmentation effects. Ecological Applications 12:335-345. McIntyre S, Barrett GW. 1992. Habitat variegation, an alternative to fragmentation. Conservation Biology 6:146-147. McKenzie N, Jacquier D, Isbell R, Brown K. 2004. Australian soils and landscapes: an illustrated compendium. CSIRO Publishing, Collingwood, Australia. Melbourne BA. 1999. Bias in the effect of habitat structure on pitfall traps: an experimental evaluation. Australian Journal of Ecology 24:228-239.

152 Metzger JP, Martensen AC, Dixo M, Bernacci LC, Ribeiro MC, Godoy Teixeira AM, Pardini R. 2009. Time-lag in biological responses to landscape changes in a highly dynamic Atlantic forest region. Biological Conservation 142:1166-1177. Meyer CL, Sisk TD, Covington WW. 2001. Microclimatic changes induced by ecological restoration of ponderosa pine forests in northern Arizona. Restoration Ecology 9:443-452. Milenkaya O, Catlin DH, Legge S, Walters JR. 2015. Body condition indices predict reproductive success but not survival in a sedentary, tropical bird. PLoS ONE 10:e0136582. Mills LS, Allendorf FW. 1996. The one-migrant-per-generation rule in conservation and management. Conservation Biology 10:1509-1518. Moorman CE, Bowen LT, Kilgo JC, Sorenson CE, Hanula JL, Horn S, Ulyshen MD. 2007. Seasonal diets of insectivorous birds using canopy gaps in a bottomland forest. Journal of Field Ornithology 78:11-20. Moreno J, Merino S, Martínez J, Sanz JJ, Arriero E. 2002. Heterophil/lymphocyte ratios and heat- shock protein levels are related to growth in nestling birds. Ecoscience 9:434-439. Moreno ML, Fernández MG, Molina SI, Valladares G. 2013. The role of small woodland remnants on ground dwelling insect conservation in Chaco Serrano, Central Argentina. Journal of Insect Science 13:40. Moretti M, Obrist MK, Duelli P. 2004. Arthropod biodiversity after forest fires: winners and losers in the winter fire regime of the southern Alps. Ecography 27:173-186. Morrison EB, Lindell CA, Holl KD, Zahawi RA. 2010. Patch size effects on avian foraging behaviour: implications for tropical forest restoration design. Journal of Applied Ecology 47:130-138. Mortelliti A, Amori G, Capizzi D, Rondinini C, Boitani L. 2010a. Experimental design and taxonomic scope of fragmentation studies on European mammals: current status and future priorities. Mammal Review 40:125-154. Mortelliti A, Fagiani S, Battisti C, Capizzi D, Boitani L. 2010b. Independent effects of habitat loss, habitat fragmentation and structural connectivity on forest-dependent birds. Diversity and Distributions 16:941-951. Müller C, Jenni-Eiermann S, Jenni L. 2011. Heterophils/lymphocytes-ratio and circulating corticosterone do not indicate the same stress imposed on Eurasian kestrel nestlings. Functional Ecology 25:566-576. Naef-Daenzer L, Naef-Daenzer B, Nager RG. 2000. Prey selection and foraging performance of breeding great tits Parus major in relation to food availability. Journal of Avian Biology 31:206-214.

153 Nagelkerke NJD. 1991. A note on a general definition of the coefficient of determination. Biometrika 78:691-692. Nagy LR, Stanculescu D, Holmes RT. 2007. Mass loss by breeding female songbirds: food supplementation supports energetic stress hypothesis in black-throated blue warblers. Condor 109:304-311. Nakagawa S, Parker TH. 2015. Replicating research in ecology and evolution: feasibility, incentives, and the cost-benefit conundrum. BMC Biology 13:88. Nakazawa M. 2015. fmsb: functions for medical statistics book with some demographic data. R package version 0.5.2. Available from https://cran.r- project.org/web/packages/fmsb/index.html. Nebel S, Mills A, McCracken JD, Taylor PD. 2010. Declines of aerial insectivores in North America follow a geographic gradient. Avian Conservation and Ecology 5:1. Newbold T, Scharlemann JPW, Butchart SHM, Şekercioğlu ÇH, Alkemade R, Booth H, Purves DW. 2013. Ecological traits affect the response of tropical forest bird species to land-use intensity. Proceedings of the Royal Society B 280:20122131. Newton I, Dale L. 1996a. Relationship between migration and latitude among west European birds. Journal of Animal Ecology 65:137-146. Newton I, Dale LC. 1996b. Bird migration at different latitudes in eastern North America. Auk 113:626-635. Nix H. 1994. The brigalow. Pages 198-233 in Dovers S, editor. Australian environmental history: essays and cases. Oxford University Press, Melbourne, Australia. O'Brien SJ, et al. 1985. Genetic basis for species vulnerability in the cheetah. Science 227:1428- 1434. Ochs CL, Dawson RD. 2008. Patterns of variation in leucocyte counts of female tree swallows, Tachycineta bicolor: Repeatability over time and relationships with condition and costs of reproduction. Comparative Biochemistry and Physiology A 150:326-331. Olsen P, Weston M, Tzaros C, Silcocks A. 2005. The state of Australia's birds 2005: woodlands and birds. Wingspan 15:S1-S32. Paquette SR, Pelletier F, Garant D, Bélisle M. 2014. Severe recent decrease of adult body mass in a declining insectivorous bird population. Proceedings of the Royal Society B 281:20140649. Pe'er G, Henle K, Dislich C, Frank K. 2011. Breaking functional connectivity into components: a novel approach using an individual-based model, and first outcomes. PLoS ONE 6:e22355. Peery MZ, et al. 2010. Genetic analyses of historic and modern marbled murrelets suggest decoupling of migration and gene flow after habitat fragmentation. Proceedings of the Royal Society B 277:697-706.

154 Peig J, Green AJ. 2009. New perspectives for estimating body condition from mass/length data: the scaled mass index as an alternative method. Oikos 118:1883-1891. Perring MP, De Frenne P, Baeten L, Maes SL, Depauw L, Blondeel H, Carón MM, Verheyen K. 2016. Global environmental change effects on ecosystems: the importance of land-use legacies. Global Change Biology 22:1361-1371. Pelletier A, Obbard ME, Harnden M, McConnell S, Howe EJ, Burrows FG, White BN, Kyle CJ. 2017. Determining causes of genetic isolation in a large carnivore (Ursus americanus) population to direct contemporary conservation measures. PLoS ONE 12: e0172319. Poessel SA, Burdett CL, Boydston EE, Lyren LM, Alonso RS, Fisher RN, Crooks KR. 2014. Roads influence movement and home ranges of a fragmentation-sensitive carnivore, the bobcat, in an urban landscape. Biological Conservation 180:224-232. Porensky LM, Young TP. 2013. Edge-effect interactions in fragmented and patchy landscapes. Conservation Biology 27:509-519. Poulin B, Lefebvre G. 1996. Dietary relationships of migrant and resident birds from a humid forest in central Panama. Auk 113:277-287. Prugh LR, Hodges KE, Sinclair ARE, Brashares JS. 2008. Effect of habitat area and isolation on fragmented animal populations. Proceedings of the National Academy of Sciences USA 105:20770-20775. Pulliam HR. 1988. Sources, sinks, and population regulation. American Naturalist 132:652-661. R Core Development Team. 2015. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Radford JQ, Bennett AF, Cheers GJ. 2005. Landscape-level thresholds of habitat cover for woodland-dependent birds. Biological Conservation 124:317-337. Rayner L, Lindenmayer DB, Gibbons P, Manning AD. 2014. Evaluating empirical evidence for decline in temperate woodland birds: a nationally threatened assemblage of species. Biological Conservation 171:145-155. Razeng E, Watson DM. 2012. What do declining woodland birds eat? A synthesis of dietary records. Emu 112:149-156. Razeng E, Watson DM. 2015. Nutritional composition of the preferred prey of insectivorous birds: popularity reflects quality. Journal of Avian Biology 46:89-96. Recher HF, Holmes RT, Schulz M, Shields J, Kavanagh R. 1985. Foraging patterns of breeding birds in eucalypt forest and woodland of southeastern Australia. Australian Journal of Ecology 10:399-419. Reed DH, Frankham R. 2003. Correlation between fitness and genetic diversity. Conservation Biology 17:230-237.

155 Reid JM, Arcese P, Keller LF. 2003. Inbreeding depresses immune response in song sparrows (Melospiza melodia): direct and inter-generational effects. Proceedings of the Royal Society B 270:2151-2157. Reid JRW. 1999. Threatened and declining birds in the New South Wales sheep-wheat belt: I. Diagnosis, characteristics and management. Consultancy report to NSW National Parks and Wildlife Service. CSIRO Wildlife and Ecology, Canberra, Australia. Rezende EL, Bacigalupe LD. 2015. Thermoregulation in endotherms: physiological principles and ecological consequences. Journal of Comparative Physiology B 185:709-727. Richmond S, Nol E, Burke D. 2011. Avian nest success, mammalian nest predator abundance, and invertebrate prey availability in a fragmented landscape. Canadian Journal of Zoology 89:517-528. Ries L, Fletcher Jr. RJ, Battin J, Sisk TD. 2004. Ecological responses to habitat edges: mechanisms, models, and variability explained. Annual Review of Ecology Evolution and Systematics 35:491-522. Rivera-Ortíz FA, Aguilar R, Arizmendi MDC, Quesada M, Oyama K. 2015. Habitat fragmentation and genetic variability of tetrapod populations. Animal Conservation 18:249-258. Robb GN, McDonald RA, Inger R, Reynolds SJ, Newton J, McGill RAR, Chamberlain DE, Harrison TJE, Bearhop S. 2011. Using stable-isotope analysis as a technique for determining consumption of supplementary foods by individual birds. Condor 113:475-482. Robel RJ, Press BM, Henning BL, Johnson KW, Blocker HD, Kemp KE. 1995. Nutrient and energetic characteristics of sweepnet-collected invertebrates. Journal of Field Ornithology 66:44-53. Roberts JH, Angermeier PL, Hallerman EM. 2013. Distance, dams and drift: what structures populations of an endangered, benthic stream fish? Freshwater Biology 58:2050-2064. Robertson O, Maron M, Buckley Y, McAlpine C. 2013. Incidence of competitors and landscape structure as predictors of woodland-dependent birds. Landscape Ecology 28:1975-1987. Robinson SK, Thompson III FR, Donovan TM, Whitehead DR, Faaborg J. 1995. Regional forest fragmentation and the nesting success of migratory birds. Science 267:1987-1990. Robinson WD, Sherry TW. 2012. Mechanisms of avian population decline and species loss in tropical forest fragments. Journal of Ornithology 153:S141-S152. Rodríguez A, Andrén H, Jansson G. 2001. Habitat-mediated predation risk and decision making of small birds at forest edges. Oikos 95:383-396. Rogers K, Rogers A, Rogers D. 1986. Bander's aid: a guide to ageing and sexing bush birds. A. Rogers, St Andrews, Australia.

156 Romero LM, Reed JM. 2005. Collecting baseline corticosterone samples in the field: is under 3 min good enough? Comparative Biochemistry and Physiology A 140:73-79. Roshier DA, Reid JRW. 2003. On animal distributions in dynamic landscapes. Ecography 26:539- 544. Ross KA, Fox BJ, Fox MD. 2002. Changes to plant species richness in forest fragments: fragment age, disturbance and fire history may be as important as area. Journal of Biogeography 29:749-765. Russell JS. 1973. Effects of land smoothing on gilgaied brigalow soils on certain soil properties and plant growth. Tropical Grasslands 7:209-218. Saccheri I, Kuussaari M, Kankare M, Vikman P, Fortelius W, Hanski I. 1998. Inbreeding and extinction in a butterfly metapopulation. Nature 392:491-494. Sainio E-L, Lehtola T, Roininen P. 1988. Radioimmunassay of total and free corticosterone in rat plasma: measurement of the effect of different doses of corticosterone. Steroids 51:609-622. Sanchez-Bayo F, Goka K. 2014. Pesticide residues and bees - a risk assessment. PLoS ONE 9:e94482. Sangha KK, Jalota RK, Midmore DJ. 2006. Litter production, decomposition and nutrient release in cleared and uncleared pasture systems of central Queensland, Australia. Journal of Tropical Ecology 22:177-189. Saska P, van der Werf W, Hemerik L, Luff ML, Hatten TD, Honek A. 2013. Temperature effects on pitfall catches of epigeal arthropods: a model and method for bias correction. Journal of Applied Ecology 50:181-189. Sattler P, Williams R, editors. 1999. The conservation status of Queensland's bioregional ecosystems. Environmental Protection Agency, Brisbane, Australia. Saunders D, Ingram J. 1995. Birds of southwestern Australia: an atlas of changes in distribution and abundance of the Wheatbelt fauna. Surrey Beatty & Sons Pty Ltd, Chipping Norton, Australia. Saura S, Bodin Ö, Fortin M-J. 2014. Stepping stones are crucial for species' long-distance dispersal and range expansion through habitat networks. Journal of Applied Ecology 51:171-182. Schlaepfer MA, Runge MC, Sherman PW. 2002. Ecological and evolutionary traps. Trends in Ecology & Evolution 17:474-480. Schoech SJ. 1996. The effect of supplemental food on body condition and the timing of reproduction in a cooperative breeder, the Florida scrub-jay. Condor 98:234-244. Schoech SJ, Bridge ES, Boughton RK, Reynolds SJ, Atwell JW, Bowman R. 2008. Food supplementation: a tool to increase reproductive output. A case study in the threatened Florida scrub-jay. Biological Conservation 141:162-173.

157 Seabrook L, McAlpine C, Fensham R. 2006. Cattle, crops and clearing: regional drivers of landscape change in the Brigalow Belt, Queensland, Australia, 1840-2004. Landscape and Urban Planning 78:373-385. Şekercioğlu ÇH, Ehrlich PR, Daily GC, Aygen D, Goehring D, Sandi RF. 2002. Disappearance of insectivorous birds from tropical forest fragments. Proceedings of the National Academy of Sciences USA 99:263-267. Shapiro SS, Wilk MB. 1965. An analysis of variance test for normality (complete samples). Biometrika 52:591-611. Simmonds J, van Rensburg B, Maron M. 2017. Non-random patterns of vegetation clearing and potential biases in studies of habitat area effects. Landscape Ecology 32:729-743. Sinclair ARE, Mduma SAR, Arcese P. 2002. Protected areas as biodiversity benchmarks for human impact: agriculture and the Serengeti avifauna. Proceedings of the Royal Society B 269:2401-2405. Skerman PJ. 1953. The brigalow country and its importance to Queensland. Journal of the Australian Institute of Agricultural Science 19:167-176. Slagsvold T, Hušek J, Whittington JD, Wiebe KL. 2014. Antipredator behavior: escape flights on a landscape slope. Behavioral Ecology 25:378-385. Smith MJ, Forbes GJ, Betts MG. 2013. Landscape configuration influences gap-crossing decisions of northern flying squirrel (Glaucomys sabrinus). Biological Conservation 168:176-183. Spooner PG, Firman M, Yalmambirra. 2010. Origins of Travellling Stock Routes. 1. Connections to Indigenous traditional pathways. The Rangeland Journal 32:329-339. Stacey PB, Taper M. 1992. Environmental variation and the persistence of small populations. Ecological Applications 2:18-29. State of Queensland. 2013. Managing category C regrowth vegetation: a self-assessable vegetation clearing code. State of Queensland (Department of Natural Resources and Mines), Brisbane, Australia. State of Queensland. 2014a. Biodiversity status of pre-clearing and remnant regional ecosystems - south east Qld. State of Queensland (Department of Science, Information Technology, Innovation and the Arts), Brisbane, Australia. Available from http://dds.information.qld.gov.au/dds/. State of Queensland. 2014b. Foliage projective cover 2011. State of Queensland (Department of Science, Information Technology, Innovation and the Arts), Brisbane, Australia. Available from http://dds.information.qld.gov.au/dds/.

158 State of Queensland. 2015. Queensland globe: coal seam gas globe. State of Queensland, Brisbane, Australia. Available from https://www.business.qld.gov.au/business/support-tools- grants/services/mapping-data-imagery/queensland-globe/coal-seam-gas. State of Queensland. 2016a. Queensland stock route network management strategy 2014-19. State of Queensland (Department of Natural Resources and Mines), Brisbane, Australia. State of Queensland. 2016b. Key to orders of common larger adult insects. State of Queensland (Queensland Museum), South Brisbane, Australia. Available from http://www.southbank.qm.qld.gov.au/Find+out+about/Animals+of+Queensland/Insects/Wh at+insect+is+that/Insect+identifier/Key+to+Orders+of+Common+Larger+Adult+Insects#.V 2IBORKndfc. State of Queensland. 2016c. Land cover change in Queensland 2014-15: a Statewide Landcover and Trees Study (SLATS) report. State of Queensland (Department of Science Information Technology and Innovation), Brisbane, Australia. State of Queensland. 2016d. Stock Route Network Management Bill 2016: explanatory notes. State of Queensland, Brisbane, Australia. Stevens HC, Watson DM. 2013. Reduced rainfall explains avian declines in an unfragmented landscape: incremental steps toward an empty forest? Emu 113:112-121. Storfer A, et al. 2007. Putting the 'landscape' in landscape genetics. Heredity 98:128-142. Storfer A, Murphy MA, Spear SF, Holderegger R, Waits LP. 2010. Landscape genetics: where are we now? Molecular Ecology 19:3496-3514. Strevens CMJ, Bonsall MB. 2011. Density-dependent population dynamics and dispersal in heterogeneous metapopulations. Journal of Animal Ecology 80:282-293. Strong AM, Sherry TW. 2000. Habitat-specific effects of food abundance on the condition of ovenbirds wintering in Jamaica. Journal of Animal Ecology 69:883-895. Strudley MW, Green TR, Ascough II JC. 2008. Tillage effects on soil hydraulic properties in space and time: state of the science. Soil and Tillage Research 99:4-48. Suorsa P, Helle H, Koivunen V, Huhta E, Nikula A, Hakkarainen H. 2004. Effects of forest patch size on physiological stress and immunocompetence in an area-sensitive passerine, the Eurasian treecreeper (Certhia familiaris): an experiment. Proceedings of the Royal Society B 271:435-440. Suorsa P, Huhta E, Nikula A, Nikinmaa M, Jantti A, Helle H, Hakkarainen H. 2003. Forest management is associated with physiological stress in an old-growth forest passerine. Proceedings of the Royal Society B 270:963-969. Szabo JK, Vesk PA, Baxter PWJ, Possingham HP. 2011. Paying the extinction debt: woodland birds in the Mount Lofty Ranges, South Australia. Emu 111:59-70.

159 Szulkin M, Sheldon BC. 2008. Dispersal as a means of inbreeding avoidance in a wild bird population. Proceedings of the Royal Society B 275:703-711. Taylor SG. 2008. Leaf litter invertebrate assemblages in box-ironbark forest: composition, size and seasonal variation in biomass. Victorian Naturalist 125:19-27. Templeton AR, Neuwald JL, Brazeal H, Robertson RJ. 2007. Restoring demographic processes in translocated populations: the case of collared lizards in the Missouri Ozarks using prescribed forest fires. Israel Journal of Ecology & Evolution 53:179-196. Templeton AR, Robertson RJ, Brisson J, Strasburg J. 2001. Disrupting evolutionary processes: the effect of habitat fragmentation on collared lizards in the Missouri Ozarks. Proceedings of the National Academy of Sciences USA 98:5426-5432. Thornton DH, Fletcher Jr. RJ. 2014. Body size and spatial scales in avian response to landscapes: a meta-analysis. Ecography 37:454-463. Tilman D. 1982. Resource competition and community structure. Princeton University Press, Princeton, USA. Tilman D, May RM, Lehman CL, Nowak MA. 1994. Habitat destruction and the extinction debt. Nature 371:65-66. Tischendorf L, Fahrig L. 2000. On the usage and measurement of landscape connectivity. Oikos 90:7-19. Tongway DJ, Sparrow AD, Friedel MH. 2003. Degradation and recovery processes in arid grazing lands of central Australia. Part 1: soil and land resources. Journal of Arid Environments 55:301-326. Torrusio S, Cigliano MM, de Wysiecki ML. 2002. Grasshopper (Orthoptera : Acridoidea) and plant community relationships in the Argentine pampas. Journal of Biogeography 29:221-229. Turner MG. 2005. Landscape ecology: what is the state of the science? Annual Review of Ecology Evolution and Systematics 36:319-344. van Dijk AIJM, Beck HE, Crosbie RS, de Jeu RAM, Liu YY, Podger GM, Timbal B, Viney NR. 2013. The Millennium Drought in southeast Australia (2001-2009): natural and human causes and implications for water resources, ecosystems, economy, and society. Water Resources Research 49:1040-1057. Van Vuren D, Armitage KB. 1994. Survival of dispersing and philopatric yellow-bellied marmots: what is the cost of dispersal? Oikos 69:179-181. Van Wilgenburg SL, Mazerolle DF, Hobson KA. 2001. Patterns of arthropod abundance, vegetation, and microclimate at boreal forest edge and interior in two landscapes: implications for forest birds. Ecoscience 8:454-461.

160 Vasconcelos HL, Bruna EM. 2012. Arthropod responses to the experimental isolation of Amazonian forest fragments. Zoologia 29:515-530. Venables WN, Ripley BD. 2002. Modern applied statistics with S. Fourth edition. Springer, New York, USA. Vié J-C, Hilton-Taylor C, Stuart SN, editors. 2009. Wildlife in a changing world - an analysis of the 2008 IUCN Red List of Threatened Species. IUCN, Gland, Switzerland. Vilà C, et al. 2003. Rescue of a severely bottlenecked wolf (Canis lupus) population by a single immigrant. Proceedings of the Royal Society B 270:91-97. Villard M-A, Trzcinski MK, Merriam G. 1999. Fragmentation effects on forest birds: relative influence of woodland cover and configuration on landscape occupancy. Conservation Biology 13:774-783. Visser ME, Holleman LJM, Gienapp P. 2006. Shifts in caterpillar biomass phenology due to climate change and its impact on the breeding biology of an insectivorous bird. Oecologia 147:164-172. Vleck CM, Vertalino N, Vleck D, Bucher TL. 2000. Stress, corticosterone, and heterophil to lymphocyte ratios in free-living adelie penguins. Condor 102:392-400. Walberg J. 2001. White blood cell counting techniques in birds. Seminars in Avian and Exotic Pet Medicine 10:72-76. Watson DM. 2002. A conceptual framework for studying species composition in fragments, islands and other patchy ecosystems. Journal of Biogeography 29:823-834. Watson DM. 2011. A productivity-based explanation for woodland bird declines: poorer soils yield less food. Emu 111:10-18. Watson J, Freudenberger D, 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 JEM, Whittaker RJ, Freudenberger D. 2005. Bird community responses to habitat fragmentation: how consistent are they across landscapes? Journal of Biogeography 32:1353-1370. Weiss DJ, Wardrop KJ, editors. 2010. Schalm's veterinary hematology. Wiley-Blackwell, Ames, USA. White TCR. 2008. The role of food, weather and climate in limiting the abundance of animals. Biological Reviews 83:227-248.

161 Whitehead K, Hugenholtz CH, Stephen M, Brown O, LeClair A, Tamminga A, Barchyn TE, Moorman B, Eaton B. 2014. Remote sensing of the environment with small unmanned aircraft systems (UASs), part 2: scientific and commercial applications. Journal of Unmanned Vehicle Systems 2:86-102. Wickham H. 2009. ggplot2: elegant graphics for data analysis. Springer-Verlag, New York, USA. Wiens JA. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397. Wiens JA, Schooley RL, Weeks Jr. RD. 1997. Patchy landscapes and animal movements: do beetles percolate? Oikos 78:257-264. Wiens JA, Stenseth NC, Van Horne B, Ims RA. 1993. Ecological mechanisms and landscape ecology. Oikos 66:369-380. Wiens JD, Noon BR, Reynolds RT. 2006. Post-fledging survival of northern goshawks: the importance of prey abundance, weather, and dispersal. Ecological Applications 16:406-418. Wikelski M, Cooke SJ. 2006. Conservation physiology. Trends in Ecology & Evolution 21:38-46. Wikelski M, Romero LM, Snell HL. 2001. Marine iguanas oiled in the Galapagos. Science 292:437-438. Wikelski M, Wong V, Chevalier B, Rattenborg N, Snell HL. 2002. Marine iguanas die from trace oil pollution. Nature 417:607-608. Williams SE, Middleton J. 2008. Climatic seasonality, resource bottlenecks, and abundance of rainforest birds: implications for global climate change. Diversity and Distributions 14:69- 77. Wilson MC, et al. 2016. Habitat fragmentation and biodiversity conservation: key findings and future challenges. Landscape Ecology 31:219-227. Wingfield JC, Maney DL, Breuner CW, Jacobs JD, Lynn S, Ramenofsky M, Richardson RD. 1998. Ecological bases of hormone-behavior interactions: the "emergency life history stage". American Zoologist 38:191-206. Woinarski JCZ, Catterall CP. 2004. Historical changes in the bird fauna at Coomooboolaroo, northeastern Australia, from the early years of pastoral settlement (1873) to 1999. Biological Conservation 116:379-401. Woinarski JCZ, McCosker JC, Gordon G, Lawrie B, James C, Augusteyn J, Slater L, Danvers T. 2006. Monitoring change in the vertebrate fauna of central Queensland, Australia, over a period of broad-scale vegetation clearance, 1973-2002. Wildlife Research 33:263-274. Wojczulanis-Jakubas K, Jakubas D, Chastel O, Kulaszewicz I. 2015. A big storm in a small body: seasonal changes in body mass, hormone concentrations and leukocyte profile in the little auk (Alle alle). Polar Biology 38:1203-1212.

162 Wojczulanis-Jakubas K, Jakubas D, Czujkowska A, Kulaszewicz I, Kruszewicz AG. 2012. Blood parasite infestation and the leukocyte profiles in adult and immature reed warblers (Acrocephalus scirpaceus) and sedge warblers (Acrocephalus schoenobaenus) during autumn migration. Annales Zoologici Fennici 49:341-349. WWF. 2016. Living Planet Report 2016. Risk and resilence in a new era. WWF International, Gland, Switzerland. Yates CJ, Hobbs RJ. 1997. Temperate eucalypt woodlands: a review of their status, processes threatening their persistence and techniques for restoration. Australian Journal of Botany 45:949-973. Yates CJ, Norton DA, Hobbs RJ. 2000. Grazing effects on plant cover, soil and microclimate in fragmented woodlands in south-western Australia: implications for restoration. Austral Ecology 25:36-47. Yong DL, Qie L, Sodhi NS, Koh LP, Peh KS-H, Lee TM, Lim HC, Lim SL-H. 2011. Do insectivorous bird communities decline on land-bridge forest islands in Peninsular Malaysia? Journal of Tropical Ecology 27:1-14. Zanette L. 2000. Fragment size and the demography of an area-sensitive songbird. Journal of Animal Ecology 69:458-470. Zanette L, Doyle P, Trémont SM. 2000. Food shortage in small fragments: evidence from an area- sensitive passerine. Ecology 81:1654-1666. Zanette L, Smith JNM, van Oort H, Clinchy M. 2003. Synergistic effects of food and predators on annual reproductive success in song sparrows. Proceedings of the Royal Society B 270:799- 803. Zollner PA, Lima SL. 2005. Behavioral tradeoffs when dispersing across a patchy landscape. Oikos 108:219-230. Zuri I, Gottreich A, Terkel J. 1998. Social stress in neighboring and encountering blind mole-rats (Spalax ehrenbergi). Physiology & Behavior 64:611-620. Zuur AF, Ieno EN, Elphick CS. 2010. A protocol for data exploration to avoid common statistical problems. Methods in Ecology and Evolution 1:3-14.

163 Appendices

APPENDIX A

Table A.1. Details of the 250 articles included in the literature review (Chapter 2) by proximate process, geographic region, and taxon studied.

Citation Proximate process Geographic region Taxon Aars et al. 1995 Demography Temperate Mammal Abbas et al. 2010 Gene flow Subtropical and temperate Fish Abbott & Curtis 2010 Gene flow Temperate Amphibian Alda et al. 2011 Gene flow Temperate Bird Alò & Turner 2005 Gene flow Temperate Fish Amos et al. 2012 Gene flow Temperate Bird Anderson et al. 2004 Gene flow Temperate Amphibian Andreassen et al. 1998 Demography Temperate Mammal Angelone et al. 2011 Gene flow Temperate Amphibian Aponte et al. 2003 Demography Tropical Reptile Arbeláez-Cortés et al. 2014 Gene flow Tropical Bird Arens et al. 2007 Gene flow Temperate Amphibian Arroyo-Rodríguez et al. 2013 Demography Tropical Mammal Arruda et al. 2011 Gene flow Tropical Amphibian Assmann & Janssen 1999 Gene flow Temperate Invertebrate Atickem et al. 2013 Gene flow Tropical Mammal

164 Citation Proximate process Geographic region Taxon Ayllon et al. 2006 Gene flow Temperate Fish Banaszek et al. 2011 Gene flow Temperate Mammal Bancroft & Turchin 2003 Resources and demography Temperate Invertebrate Banks et al. 2005a Demography and gene flow Temperate Mammal Banks & Lindenmayer 2014 Demography Temperate Mammal Banks et al. 2005b Demography Temperate Mammal Bartákova et al. 2013 Gene flow Tropical and subtropical Fish Bech et al. 2009 Gene flow Temperate Bird Beebee 2007 Gene flow Temperate Invertebrate Bei et al. 2014 Gene flow Subtropical Bird Bender & Fahrig 2005 Demography Temperate Mammal BenDor et al. 2009 Demography Temperate Reptile Bennett et al. 2009 Demography Temperate Reptile Berens et al. 2014 Resources Temperate Bird Berkman et al. 2013 Gene flow Temperate Bird Berry et al. 2005 Gene flow Temperate Reptile Bidlack & Cook 2001 Gene flow Temperate Mammal Blackburn et al. 2011 Resources Temperate Invertebrate Blair et al. 2013 Gene flow Subtropical Reptile Blanco-Fontao et al. 2013 Resources Temperate Bird Boff et al. 2014 Gene flow Temperate Invertebrate Boissin et al. 2011 Gene flow Temperate Fish

165 Citation Proximate process Geographic region Taxon Bouza et al. 1999 Gene flow Temperate Fish Bowers and Dooley Jr. 1999 Demography Temperate Mammal Bowne et al. 2006 Demography Temperate Reptile Braunisch et al. 2010 Gene flow Temperate Bird Breyne et al. 2014 Gene flow Temperate Mammal Brisson et al. 2003 Demography Temperate Reptile Brückmann et al. 2011 Resources Temperate Invertebrate Bucher & Entling 2011 Resources Temperate Invertebrate Bull et al. 2011 Gene flow Temperate Mammal Burton et al. 2002 Gene flow Temperate Mammal Caplins et al. 2014 Gene flow Temperate Invertebrate Cerame et al. 2014 Gene flow Subtropical and temperate Bird Clark et al. 1999 Gene flow Subtropical Reptile Cooper et al. 2002 Demography Temperate Bird Coudrain et al. 2013 Resources Temperate Invertebrate Coulon et al. 2004 Gene flow Temperate Mammal Coulon et al. 2010 Gene flow Subtropical Bird Cronin 2007 Demography Temperate Invertebrate Croteau et al. 2007 Gene flow Temperate Bird Cureton II et al. 2014 Gene flow Temperate Reptile Dawkins et al. 2010 Gene flow Subtropical Invertebrate DeHaan et al. 2011 Gene flow Temperate Fish

166 Citation Proximate process Geographic region Taxon Dehais et al. 2010 Gene flow Temperate Fish Delaney et al. 2010 Gene flow Temperate Reptile and bird Delattre et al. 2013 Demography Temperate Invertebrate DiLeo et al. 2010 Gene flow Temperate Reptile Dixo et al. 2009 Gene flow Tropical Amphibian Dixon et al. 2006 Demography Temperate Mammal Drees et al. 2011 Gene flow Temperate Invertebrate Dutta et al. 2013 Demography and gene flow Tropical Mammal Emel & Storfer 2015 Gene flow Temperate Amphibian Engelhardt et al. 2008 Gene flow Temperate Invertebrate Epps et al. 2007 Gene flow Temperate Mammal Faulks et al. 2011 Gene flow Temperate Fish Fickel et al. 2012 Gene flow Temperate Mammal Fietz et al. 2014 Gene flow Temperate Mammal Finger et al. 2009 Gene flow Temperate Invertebrate Finnegan et al. 2012 Gene flow Temperate Mammal Fitzpatrick et al. 2014 Gene flow Temperate Fish Fowler et al. 2000 Gene flow Subtropical Mammal Frankham et al. 2014 Gene flow Temperate Mammal Fuentes-Montemayor et al. 2013 Resources Temperate Mammal Fünfstück et al. 2014 Gene flow Tropical Mammal Gabrielsen et al. 2013 Gene flow Temperate Amphibian

167 Citation Proximate process Geographic region Taxon Galbusera et al. 2004 Gene flow Tropical Bird Gavin et al. 1999 Gene flow Temperate Mammal Gerlach & Musolf 2000 Gene flow Temperate Mammal Gibbs 1998 Gene flow Temperate Amphibian González-Wangüemert & Vergara-Chen 2014 Gene flow Temperate Fish Goodman et al. 2001 Gene flow Temperate Mammal Gortat et al. 2010 Gene flow Temperate Mammal Gula et al. 2009 Gene flow Temperate Mammal Guschanski et al. 2007 Gene flow Tropical Mammal Haag et al. 2010 Demography and gene flow Tropical and subtropical Mammal Haapakoski et al. 2013 Resources Temperate Mammal Habel et al. 2013 Gene flow Temperate Invertebrate Hale et al. 2013 Gene flow Temperate Amphibian Hamer et al. 2006 Resources Temperate Bird Hames et al. 2001 Demography Temperate Bird Hänfling et al. 2004 Gene flow Temperate Fish Hänfling & Weetman 2006 Gene flow Temperate Fish Hapeman et al. 2011 Gene flow Temperate Mammal Harrisson et al. 2014 Demography and gene flow Temperate Bird Hels & Nachman 2002 Demography Temperate Amphibian Hill et al. 2006 Gene flow Temperate Invertebrate Hitchings & Beebee 1998 Gene flow Temperate Amphibian

168 Citation Proximate process Geographic region Taxon Hoehn et al. 2013 Gene flow Temperate Reptile Hokit et al. 20110 Gene flow Subtropical Reptile Holland & Bennett 2011 Demography Temperate Mammal Horreo et al. 2011 Gene flow Temperate Fish Hovel & Wahle 2010 Demography Temperate Invertebrate Ims & Andreassen 1999 Demography Temperate Mammal Iwai & Shoda-Kagaya 2012 Gene flow Subtropical Amphibian Janecka et al. 2014 Gene flow Subtropical Mammal Johansson et al. 2007 Gene flow Temperate Amphibian Johansson et al. 2005 Gene flow Temperate Amphibian Johnson et al. 2004 Gene flow Temperate Bird Jordan et al. 2009 Gene flow Temperate Amphibian Kamler et al. 2007 Resources Temperate Mammal Kappes et al. 2009 Gene flow Temperate Invertebrate Kawakami et al. 2008 Gene flow Temperate Bird Kawamura 2005 Gene flow Temperate Fish Keller et al. 2005 Gene flow Temperate Invertebrate Keller & Largiadèr 2003 Gene flow Temperate Invertebrate Kenney et al. 2014 Gene flow Subtropical Mammal Keyghobadi et al. 2005 Gene flow Temperate Invertebrate Kitanishi et al. 2012 Gene flow Temperate Fish Klapwijk & Lewis 2011 Demography Temperate Invertebrate

169 Citation Proximate process Geographic region Taxon Klapwijk & Lewis 2012 Resources Temperate Invertebrate Kluger et al. 2011 Gene flow Temperate Invertebrate Kossenko & Kaygorodova 2007 Resources Temperate Bird Kozfkay et al. 2011 Gene flow Temperate Fish Kraaijeveld-Smit et la. 2005 Gene flow Temperate Amphibian Kraaijeveld-Smit et al. 2002 Gene flow Temperate Mammal Kuehn et al. 2007 Gene flow Temperate Mammal Lancaster et al. 2011 Gene flow Temperate Mammal Landry & Lapointe 1999 Gene flow Temperate Mammal Lane-deGraaf et al. 2014 Demography and gene flow Tropical Mammal Laurence et al. 2013 Gene flow Temperate Mammal Lee et al. 2012 Gene flow Temperate Mammal Lee et al. 2010 Gene flow Subtropical Mammal Li et al. 2006 Gene flow Temperate Mammal Liu et al. 2013 Gene flow Subtropical and temperate Invertebrate Ljungberg et al. 2013 Resources Temperate Invertebrate Luquet et al. 2011 Gene flow Temperate Amphibian Machkour-M'Rabet et al. 2012 Gene flow Tropical Invertebrate Macreadie et al. 2012 Resources Temperate Invertebrate Mäki-Petäys et al. 2005 Demography and gene flow Temperate Invertebrate Mapelli et al. 2012 Gene flow Temperate Mammal Marsden et al. 2012 Demography and gene flow Tropical and subtropical Mammal

170 Citation Proximate process Geographic region Taxon Marsh et al. 2004 Demography Temperate Amphibian Marshall et al. 2009 Gene flow Temperate Reptile Matthysen et al. 2001 Demography Temperate Bird Mattila et al. 2012 Gene flow Temperate Invertebrate Mergey et al. 2012 Gene flow Temperate Mammal Mikulíček & Pišút 2012 Gene flow Temperate Amphibian Milko et al. 2012 Gene flow Temperate Invertebrate Miller et al. 2013 Gene flow Tropical Bird Millions & Swanson 2007 Gene flow Temperate Mammal Milton et al. 2009 Gene flow Tropical Mammal Monaghan et al. 2001 Gene flow Temperate Invertebrate Monaghan et al. 2002 Gene flow Temperate Invertebrate Morales et al. 2011 Gene flow Tropical Fish Morita & Yokota 2002 Demography Temperate Fish Mortelliti & Boitani 2008 Resources Temperate Mammal Mouret et al. 2011 Gene flow Temperate Reptile Nair et al. 2012 Gene flow Tropical Amphibian Neuwald 2010 Gene flow Temperate Mammal Ni et al. 2014 Resources Tropical Mammal Niedziałkowska et al. 2012 Demography and gene flow Temperate Mammal Nielsen et al. 1997 Gene flow Temperate Fish Noël et al. 2007 Gene flow Temperate Amphibian

171 Citation Proximate process Geographic region Taxon Nunziata et al. 2013 Gene flow Temperate Amphibian Ordóñez-Gómez et al. 2015 Resources Tropical Mammal Osborne & Williams 2001 Resources Temperate Invertebrate Paolucci et al. 2012 Demography Tropical Invertebrate Pavlacky Jr. et al. 2012 Demography Subtropical Bird Pavlova et al. 2012 Gene flow Temperate Bird Peacock & Dochtermann 2012 Gene flow Temperate Fish Pereoglou et al. 2013 Gene flow Temperate Mammal Pernetta et al. 2011 Gene flow Temperate Reptile Pittman et al. 2011 Gene flow Temperate Reptile Polic et al. 2014 Resources Temperate Invertebrate Provencher & Riechert 1994 Resources Temperate Invertebrate Quéméré et al. 2010 Gene flow Tropical Mammal Ramirez & Haakonsen 1999 Gene flow Temperate Invertebrate Rantalainen et al. 2004 Resources Temperate Invertebrate Reding et al. 2010 Gene flow Tropical Bird Reid et al. 2008 Gene flow Temperate Fish Remón et al. 2013 Gene flow Temperate Reptile Richmond et al. 2009 Gene flow Subtropical Reptile Roberts et al. 2013 Gene flow Temperate Fish Roffler et al. 2014 Gene flow Temperate Mammal Ruell et al. 2012 Gene flow Temperate Mammal

172 Citation Proximate process Geographic region Taxon Safner et al. 2011 Gene flow Temperate Amphibian Sandlund et al. 2014 Gene flow Temperate Fish Sato et al. 2010 Gene flow Temperate Fish Sato & Harada 2008 Gene flow Temperate Fish Sato et al. 2014 Gene flow Temperate Mammal Schiffer et al. 2007 Gene flow Tropical Invertebrate Schmuki et al. 2006 Gene flow Temperate Invertebrate Schrey et al. 2012 Gene flow Subtropical Reptile Schwalm et al. 2014 Gene flow Temperate Mammal Seppä & Laurila 1999 Gene flow Temperate Amphibian Serieys et al. 2015 Gene flow Temperate Mammal Shanahan et al. 2011 Gene flow Subtropical Bird Shepherd & Whittington 2006 Resources Temperate Mammal Sielezniew & Rutkowski 2012 Gene flow Temperate Invertebrate Slack et al. 2010 Gene flow Temperate Fish Smith & Batzli 2006 Demography Temperate Mammal Sommer 2003 Gene flow Tropical Mammal Souza et al. 2002 Gene flow Subtropical Mammal Stephens et al. 2013 Gene flow Temperate Mammal Sterling et al. 2012 Gene flow Temperate Fish Stevens & Hogg 2003 Gene flow Temperate Invertebrate Stoner et al. 2013 Demography Temperate Mammal

173 Citation Proximate process Geographic region Taxon Tawatao et al. 2011 Gene flow Tropical Invertebrate Taylor et al. 2011 Gene flow Tropical, subtropical and temperate Mammal Thomé et al. 2010 Gene flow Tropical and subtropical Amphibian Thomé et al. 2014 Gene flow Tropical and subtropical Amphibian Turner et al. 2006 Gene flow Temperate Fish Uimaniemi et al. 2003 Gene flow Temperate Bird van Apeldoorn et al. 1994 Demography Temperate Mammal Van Houdt et al. 2005 Gene flow Temperate Fish Vangestel et al. 2010 Resources Temperate Bird Vucetich et al. 2001 Gene flow Temperate Mammal Walker et al. 2008 Demography Temperate Mammal Wang et al. 2007 Gene flow Temperate Fish Watanabe et al. 2008 Gene flow Temperate Invertebrate Watanabe & Omura 2007 Gene flow Temperate Invertebrate Weyer et al. 2014 Gene flow Temperate Reptile White et al. 2014 Gene flow Temperate Bird Whiteley et al. 2013 Gene flow Temperate Fish Whiteley et al. 2006 Gene flow Temperate Fish Williams et al. 2003 Gene flow Temperate Invertebrate Wilmer & Wilcox 2007 Gene flow Subtropical Invertebrate With et al. 2002 Resources Temperate Invertebrate Wozney et al. 2011 Gene flow Temperate Fish

174 Citation Proximate process Geographic region Taxon Wu et al. 2010 Gene flow Temperate Mammal Wynne et al. 2003 Gene flow Temperate Invertebrate Yamazaki et al. 2014 Gene flow Temperate Fish Yamazaki et al. 2011 Gene flow Temperate Fish Yannic et al. 2014 Gene flow Temperate Mammal Yumnam et al. 2014 Gene flow Tropical Mammal Zaccara et al. 2014 Gene flow Temperate Fish Zamudio & Wieczorek 2007 Gene flow Temperate Amphibian Zavodna et al. 2005 Gene flow Tropical Invertebrate Zeigler et al. 2013 Demography Tropical Mammal Zhou et al. 2011 Gene flow Subtropical Invertebrate Zhu et al. 2011 Gene flow Subtropical Mammal Zuberogoitia et al. 2013 Gene flow Temperate Mammal

175 References for articles included in the literature review (Chapter 2)

Aars J, Andreassen HP, Ims RA. 1995. Root voles - litter sex-ratio variation in fragmented habitat. Journal of Animal Ecology 64:459-472. Abbas K, Zhou XY, Li Y, Gao ZX, Wang WM. 2010. Microsatellite diversity and population genetic structure of yellowcheek, Elopichthys bambusa (Cyprinidae) in the Yangtze River. Biochemical Systematics and Ecology 38:806-812. Abbott CL, Curtis JMR. 2010. Initial investigation of population and conservation genetics in the coastal giant salamander, Dicamptodon tenebrosus, using nuclear microsatellite markers. Northwest Science 84:46-51. Alda F, Sastre P, De La Cruz-Cardiel PJ, Doadrio I. 2011. Population genetics of the endangered Cantabrian capercaillie in northern Spain. Animal Conservation 14:249-260. Alò D, Turner TF. 2005. Effects of habitat fragmentation on effective population size in the endangered Rio Grande silvery minnow. Conservation Biology 19:1138-1148. Amos JN, Bennett AF, Mac Nally R, Newell G, Pavlova A, Radford JQ, Thomson JR, White M, Sunnucks P. 2012. Predicting landscape-genetic consequences of habitat loss, fragmentation and mobility for multiple species of woodland birds. PLoS ONE 7:e30888. Andersen LW, Fog K, Damgaard C. 2004. Habitat fragmentation causes bottlenecks and inbreeding in the European tree frog (Hyla arborea). Proceedings of the Royal Society B 271:1293- 1302. Andreassen HP, Hertzberg K, Ims RA. 1998. Space-use responses to habitat fragmentation and connectivity in the root vole Microtus oeconomus. Ecology 79:1223-1235. Angelone S, Kienast F, Holderegger R. 2011. Where movement happens: scale-dependent landscape effects on genetic differentiation in the European tree frog. Ecography 34:714- 722. Aponte C, Barret GR, Terborgh J. 2003. Consequences of habitat fragmentation on age structure and life history in a tortoise population. Biotropica 35:550-555. Arbeláez-Cortés E, Roldán-Piña D, Navarro-Sigüenza AG. 2014. Multilocus phylogeography and morphology give insights into the recent evolution of a Mexican endemic songbird: Vireo hypochryseus. Journal of Avian Biology 45:253-263. Arens P, van der Sluis T, van't Westende WPC, Vosman B, Vos CC, Smulders MJM. 2007. Genetic population differentiation and connectivity among fragmented moor frog (Rana arvalis) populations in The Netherlands. Landscape Ecology 22:1489-1500.

176 Arroyo-Rodríguez V, González-Perez IM, Garmendia A, Solà M, Estrada A. 2013. The relative impact of forest patch and landscape attributes on black howler monkey populations in the fragmented Lacandona rainforest, Mexico. Landscape Ecology 28:1717-1727. Arruda MP, Morielle-Versute E, Silva A, Schneider MPC, Gonçalves EC. 2011. Contemporary gene flow and weak genetic structuring in rococo toad (Rhinella schneideri) populations in habitats fragmented by agricultural activities. Amphibia-Reptilia 32:399-411. Assmann T, Janssen J. 1999. The effects of habitat changes on the endangered ground beetle Carabus nitens (Coleoptera: Carabidae). Journal of Insect Conservation 3:107-116. Atickem A, Rueness EK, Loe LE, Serbezov D, Bekele A, Stenseth NC. 2013. Population genetic structure and connectivity in the endangered Ethiopian mountain nyala (Tragelaphus buxtoni): recommending dispersal corridors for future conservation. Conservation Genetics 14:427-438. Ayllon F, Moran P, Garcia-Vazquez E. 2006. Maintenance of a small anadromous subpopulation of brown trout (Salmo trutta L.) by straying. Freshwater Biology 51:351-358. Banaszek A, Jadwiszczak KA, Ziomek J. 2011. Genetic variability and differentiation in the Polish common hamster (Cricetus cricetus L.): genetic consequences of agricultural habitat fragmentation. Mammalian Biology 76:665-671. Bancroft JS, Turchin P. 2003. An experimental test of fragmentation and loss of habitat with Oryzaephilus surinamensis. Ecology 84:1756-1767. Banks SC, Finlayson GR, Lawson SJ, Lindenmayer DB, Paetkau D, Ward SJ, Taylor AC. 2005a. The effects of habitat fragmentation due to forestry plantation establishment on the demography and genetic variation of a marsupial carnivore, Antechinus agilis. Biological Conservation 122:581-597. Banks SC, Lindenmayer DB. 2014. Inbreeding avoidance, patch isolation and matrix permeability influence dispersal and settlement choices by male agile antechinus in a fragmented landscape. Journal of Animal Ecology 83:515-524. Banks SC, Ward SJ, Lindenmayer DB, Finlayson GR, Lawson SJ, Taylor AC. 2005b. The effects of habitat fragmentation on the social kin structure and mating system of the agile antechinus, Antechinus agilis. Molecular Ecology 14:1789-1801. Bartákova V, Reichard M, Janko K, Polačik M, Blažek R, Reichwald K, Cellerino A, Bryja J. 2013. Strong population genetic structuring in an annual fish, Nothobranchius furzeri, suggests multiple savannah refugia in southern Mozambique. BMC Evolutionary Biology 13:196. Bech N, Boissier J, Drovetski S, Novoa C. 2009. Population genetic structure of rock ptarmigan in the 'sky islands' of French Pyrenees: implications for conservation. Animal Conservation 12:138-146.

177 Beebee TJC. 2007. Population structure and its implications for conservation of the great silver beetle Hydrophilus piceus in Britain. Freshwater Biology 52:2101-2111. Bei YJ, Chen WC, Sun BH, Li JH, Lai JL, Meng SQ. 2014. Population structure of the endangered Hume's pheasant (Syrmaticus humiae) inferred from a partial sequence of the mitochondrial DNA control region. Biochemical Systematics and Ecology 57:69-77. Bender DJ, Fahrig L. 2005. Matrix structure obscures the relationship between interpatch movement and patch size and isolation. Ecology 86:1023-1033. BenDor T, Westervelt J, Aurambout JP, Meyer W. 2009. Simulating population variation and movement within fragmented landscapes: an application to the gopher tortoise (Gopherus polyphemus). Ecological Modelling 220:867-878. Bennett AM, Keevil M, Litzgus JD. 2009. Demographic differences among populations of northern map turtles (Graptemys geographica) in intact and fragmented sites. Canadian Journal of Zoology 87:1147-1157. Berens DG, Chama L, Albrecht J, Farwig N. 2014. High conservation value of forest fragments for plant and frugivore communities in a fragmented forest landscape in South Africa. Biotropica 46:350-356. Berkman LK, Nielsen CK, Roy CL, Heist EJ. 2013. Resistance is futile: effects of landscape features on gene flow of the northern bobwhite. Conservation Genetics 14:323-332. Berry O, Tocher MD, Gleeson DM, Sarre SD. 2005. Effect of vegetation matrix on animal dispersal: genetic evidence from a study of endangered skinks. Conservation Biology 19:855-864. Bidlack AL, Cook JA. 2001. Reduced genetic variation in insular northern flying squirrels (Glaucomys sabrinus) along the North Pacific Coast. Animal Conservation 4:283-290. Blackburn HB, Hobbs NT, Detling JK. 2011. Nonlinear responses to food availability shape effects of habitat fragmentation on consumers. Ecology 92:98-107. Blair C, Jiménez Arcos VH, Mendez de la Cruz FR, Murphy RW. 2013. Landscape genetics of leaf- toed geckos in the tropical dry forest of Northern Mexico. PLoS ONE 8:e57433. Blanco-Fontao B, Sandercock BK, Obeso JR, McNew LB, Quevedo M. 2013. Effects of sexual dimorphism and landscape composition on the trophic behavior of greater prairie-chicken. PLoS ONE 8:e79986. Boff S, Soro A, Paxton RJ, Alves-dos-Santos I. 2014. Island isolation reduces genetic diversity and connectivity but does not significantly elevate diploid male production in a neotropical orchid bee. Conservation Genetics 15:1123-1135.

178 Boissin E, Hoareau TB, Berrebi P. 2011. Effects of current and historic habitat fragmentation on the genetic structure of the sand goby Pomatoschistus minutus (Osteichthys, Gobiidae). Biological Journal of the Linnean Society 102:175-198. Bouza C, Arias J, Castro J, Sanchez L, Martinez P. 1999. Genetic structure of brown trout, Salmo trutta L., at the southern limit of the distribution range of the anadromous form. Molecular Ecology 8:1991-2001. Bowers MA, Dooley Jr. JL. 1999. A controlled, hierarchical study of habitat fragmentation: responses at the individual, patch, and landscape scale. Landscape Ecology 14:381-389. Bowne DR, Bowers MA, Hines JE. 2006. Connectivity in an agricultural landscape as reflected by interpond movements of a freshwater turtle. Conservation Biology 20:780-791. Braunisch V, Segelbacher G, Hirzel AH. 2010. Modelling functional landscape connectivity from genetic population structure: a new spatially explicit approach. Molecular Ecology 19:3664- 3678. Breyne P, Mergeay J, Casaer J. 2014. Roe deer population structure in a highly fragmented landscape. European Journal of Wildlife Research 60:909-917. Brisson JA, Strasburg JL, Templeton AR. 2003. Impact of fire management on the ecology of collared lizard (Crotaphytus collaris) populations living on the Ozark Plateau. Animal Conservation 6:247-254. Brückmann SV, Krauss J, van Achterberg C, Steffan-Dewenter I. 2011. The impact of habitat fragmentation on trophic interactions of the monophagous butterfly Polyommatus coridon. Journal of Insect Conservation 15:707-714. Bucher R, Entling MH. 2011. Contrasting effects of habitat fragmentation, population density, and prey availability on body condition of two orb-weaving spiders. Ecological Entomology 36:680-685. Bull RAS, Cushman SA, Mace R, Chilton T, Kendall KC, Landguth EL, Schwartz MK, McKelvey K, Allendorf FW, Luikart G. 2011. Why replication is important in landscape genetics: American black bear in the Rocky Mountains. Molecular Ecology 20:1092-1107. Burton C, Krebs CJ, Taylor EB. 2002. Population genetic structure of the cyclic snowshoe hare (Lepus americanus) in southwestern Yukon, Canada. Molecular Ecology 11:1689-1701. Caplins SA, Gilbert KJ, Ciotir C, Roland J, Matter SF, Keyghobadi N. 2014. Landscape structure and the genetic effects of a population collapse. Proceedings of the Royal Society B 281:20141798. Cerame B, Cox JA, Brumfield RT, Tucker JW, Taylor SS. 2014. Adaptation to ephemeral habitat may overcome natural barriers and severe habitat fragmentation in a fire-dependent species, the Bachman's sparrow (Peucaea aestivalis). PLoS ONE 9:e105782.

179 Clark AM, Bowen BW, Branch LC. 1999. Effects of natural habitat fragmentation on an endemic scrub lizard (Sceloporus woodi): an historical perspective based on a mitochondrial DNA gene genealogy. Molecular Ecology 8:1093-1104. Cooper CB, Walters JR, Priddy J. 2002. Landscape patterns and dispersal success: simulated population dynamics in the brown treecreeper. Ecological Applications 12:1576-1587. Coudrain V, Herzog F, Entling MH. 2013. Effects of habitat fragmentation on abundance, larval food and parasitism of a spider-hunting wasp. PLoS ONE 8:e59286. Coulon A, Cosson JF, Angibault JM, Cargnelutti B, Galan M, Morellet N, Petit E, Aulagnier S, Hewison AJM. 2004. Landscape connectivity influences gene flow in a roe deer population inhabiting a fragmented landscape: an individual-based approach. Molecular Ecology 13:2841-2850. Coulon A, Fitzpatrick JW, Bowman R, Lovette IJ. 2010. Effects of habitat fragmentation on effective dispersal of Florida scrub-jays. Conservation Biology 24:1080-1088. Cronin JT. 2007. From population sources to sieves: the matrix alters host-parasitoid source-sink structure. Ecology 88:2966-2976. Croteau EK, Lougheed SC, Krannitz PG, Mahony NA, Walker BL, Boag PT. 2007. Genetic population structure of the sagebrush Brewer's sparrow, Spizella breweri breweri, in a fragmented landscape at the northern range periphery. Conservation Genetics 8:1453-1463. Cureton II JC, Janis M, Lutterschmidt WI, Randle CP, Ruthven III DC, Deaton R. 2014. Effects of urbanization on genetic diversity, gene flow, and population structure in the ornate box turtle (Terrapene ornata). Amphibia-Reptilia 35:87-97. Dawkins KL, Furse JM, Wild CH, Hughes JM. 2010. Distribution and population genetics of the threatened freshwater crayfish genus Tenuibranchiurus (Decapoda : Parastacidae). Marine and Freshwater Research 61:1048-1055. DeHaan PW, Bernall SR, DosSantos JM, Lockard LL, Ardren WR. 2011. Use of genetic markers to aid in re-establishing migratory connectivity in a fragmented metapopulation of bull trout (Salvelinus confluentus). Canadian Journal of Fisheries and Aquatic Sciences 68:1952-1969. Dehais C, Eudeline R, Berrebi P, Argillier C. 2010. Microgeographic genetic isolation in chub (Cyprinidae: Squalius cephalus) population of the Durance River: estimating fragmentation by dams. Ecology of Freshwater Fish 19:267-278. Delaney KS, Riley SPD, Fisher RN. 2010. A rapid, strong, and convergent genetic response to urban habitat fragmentation in four divergent and widespread vertebrates. PLoS ONE 5:e12767. Delattre T, Baguette M, Burel F, Stevens VM, Quénol H, Vernon P. 2013. Interactive effects of landscape and weather on dispersal. Oikos 122:1576-1585.

180 DiLeo MF, Row JR, Lougheed SC. 2010. Discordant patterns of population structure for two co- distributed snake species across a fragmented Ontario landscape. Diversity and Distributions 16:571-581. Dixo M, Metzger JP, Morgante JS, Zamudio KR. 2009. Habitat fragmentation reduces genetic diversity and connectivity among toad populations in the Brazilian Atlantic Coastal Forest. Biological Conservation 142:1560-1569. Dixon JD, Oli MK, Wooten MC, Eason TH, McCown JW, Paetkau D. 2006. Effectiveness of a regional corridor in connecting two Florida black bear populations. Conservation Biology 20:155-162. Drees C, Hüfner S, Matern A, Nève G, Assmann T. 2011. Repeated sampling detects gene flow in a flightless ground beetle in a fragmented landscape. Hereditas 148:36-45. Dutta T, Sharma S, Maldonado JE, Wood TC, Panwar HS, Seidensticker J. 2013. Fine-scale population genetic structure in a wide-ranging carnivore, the leopard (Panthera pardus fusca) in central India. Diversity and Distributions 19:760-771. Emel SL, Storfer A. 2015. Landscape genetics and genetic structure of the southern torrent salamander, Rhyacotriton variegatus. Conservation Genetics 16:209-221. Engelhardt CHM, Pauls SU, Haase P. 2008. Population genetic structure of the caddisfly Rhyacophila pubescens, Pictet 1834, north of the Alps. Fundamental and Applied Limnology 173:165-176. Epps CW, Wehausen JD, Bleich VC, Torres SG, Brashares JS. 2007. Optimizing dispersal and corridor models using landscape genetics. Journal of Applied Ecology 44:714-724. Faulks LK, Gilligan DM, Beheregaray LB. 2011. The role of anthropogenic vs. natural in-stream structures in determining connectivity and genetic diversity in an endangered freshwater fish, Macquarie perch (Macquaria australasica). Evolutionary Applications 4:589-601. Fickel J, Bubliy OA, Stache A, Noventa T, Jirsa A, Heurich M. 2012. Crossing the border? Structure of the red deer (Cervus elaphus) population from the Bavarian-Bohemian forest ecosystem. Mammalian Biology 77:211-220. Fietz J, Tomiuk J, Loeschcke V, Weis-Dootz T, Segelbacher G. 2014. Genetic consequences of forest fragmentation for a highly specialized arboreal mammal - the edible dormouse. PLoS ONE 9:e88092. Finger A, Schmitt T, Zachos FE, Meyer M, Assmann T, Habel JC. 2009. The genetic status of the violet copper Lycaena helle - a relict of the cold past in times of global warming. Ecography 32:382-390.

181 Finnegan LA, Wilson PJ, Price GN, Lowe SJ, Patterson BR, Fortin M-J, Murray DL. 2012. The complimentary role of genetic and ecological data in understanding population structure: a case study using moose (Alces alces). European Journal of Wildlife Research 58:415-423. Fitzpatrick SW, Crockett H, Funk WC. 2014. Water availability strongly impacts population genetic patterns of an imperiled Great Plains endemic fish. Conservation Genetics 15:771-788. Fowler EV, Houlden BA, Hoeben P, Timms P. 2000. Genetic diversity and gene flow among southeastern Queensland koalas (Phascolarctos cinereus). Molecular Ecology 9:155-164. Frankham GJ, Handasyde KA, Norton M, Murray A, Eldridge MDB. 2014. Molecular detection of intra-population structure in a threatened potoroid, Potorous tridactylus: conservation management and sampling implications. Conservation Genetics 15:547-560. Fuentes-Montemayor E, Goulson D, Cavin L, Wallace JM, Park KJ. 2013. Fragmented woodlands in agricultural landscapes: the influence of woodland character and landscape context on bats and their insect prey. Agriculture Ecosystems & Environment 172:6-15. Fünfstück T, et al. 2014. The genetic population structure of wild western lowland gorillas (Gorilla gorilla gorilla) living in continuous rain forest. American Journal of Primatology 76:868- 878. Gabrielsen CG, Kovach AI, Babbitt KJ, McDowell WH. 2013. Limited effects of suburbanization on the genetic structure of an abundant vernal pool-breeding amphibian. Conservation Genetics 14:1083-1097. Galbusera P, Githiru M, Lens L, Matthysen E. 2004. Genetic equilibrium despite habitat fragmentation in an Afrotropical bird. Molecular Ecology 13:1409-1421. Gavin TA, Sherman PW, Yensen E, May B. 1999. Population genetic structure of the Northern Idaho ground squirrel (Spermophilus brunneus brunneus). Journal of Mammalogy 80:156- 168. Gerlach G, Musolf K. 2000. Fragmentation of landscape as a cause for genetic subdivision in bank voles. Conservation Biology 14:1066-1074. Gibbs JP. 1998. Genetic structure of redback salamander Plethodon cinereus populations in continuous and fragmented forests. Biological Conservation 86:77-81. González-Wangüemert M, Vergara-Chen C. 2014. Environmental variables, habitat discontinuity and life history shaping the genetic structure of Pomatoschistus marmoratus. Helgoland Marine Research 68:357-371. Goodman SJ, Tamate HB, Wilson R, Nagata J, Tatsuzawa S, Swanson GM, Pemberton JM, McCullough DR. 2001. Bottlenecks, drift and differentiation: the population structure and demographic history of sika deer (Cervus nippon) in the Japanese archipelago. Molecular Ecology 10:1357-1370.

182 Gortat T, Gryczyńska-Siemiątkowska A, Rutkowski R, Kozakiewicz A, Mikoszewski A, Kozakiewicz M. 2010. Landscape pattern and genetic structure of a yellow-necked mouse Apodemus flavicollis population in north-eastern Poland. Acta Theriologica 55:109-121. Gula R, Hausknecht R, Kuehn R. 2009. Evidence of wolf dispersal in anthropogenic habitats of the Polish Carpathian Mountains. Biodiversity and Conservation 18:2173-2184. Guschanski K, Olivieri G, Funk SM, Radespiel U. 2007. MtDNA reveals strong genetic differentiation among geographically isolated populations of the golden brown mouse lemur, Microcebus ravelobensis. Conservation Genetics 8:809-821. Haag T, et al. 2010. The effect of habitat fragmentation on the genetic structure of a top predator: loss of diversity and high differentiation among remnant populations of Atlantic Forest jaguars (Panthera onca). Molecular Ecology 19:4906-4921. Haapakoski M, Sundell J, Ylönen H. 2013. Mammalian predator-prey interaction in a fragmented landscape: weasels and voles. Oecologia 173:1227-1235. Habel JC, Husemann M, Schmitt T, Dapporto L, Rödder D, Vandewoestijne S. 2013. A forest butterfly in Sahara Desert oases: isolation does not matter. Journal of Heredity 104:234-247. Hale JM, Heard GW, Smith KL, Parris KM, Austin JJ, Kearney M, Melville J. 2013. Structure and fragmentation of growling grass frog metapopulations. Conservation Genetics 14:313-322. Hamer TL, Flather CH, Noon BR. 2006. Factors associated with grassland bird species richness: the relative roles of grassland area, landscape structure, and prey. Landscape Ecology 21:569- 583. Hames RS, Rosenberg KV, Lowe JD, Dhondt AA. 2001. Site reoccupation in fragmented landscapes: testing predictions of metapopulation theory. Journal of Animal Ecology 70:182-190. Hänfling B, Durka W, Brandl R. 2004. Impact of habitat fragmentation on genetic population structure of roach, Rutilus rutilus, in a riparian ecosystem. Conservation Genetics 5:247- 257. Hänfling B, Weetman D. 2006. Concordant genetic estimators of migration reveal anthropogenically enhanced source-sink population structure in the River Sculpin, Cottus gobio. Genetics 173:1487-1501. Hapeman P, Latch EK, Fike JA, Rhodes OE, Kilpatrick CW. 2011. Landscape genetics of fishers (Martes pennanti) in the northeast: dispersal barriers and historical influences. Journal of Heredity 102:251-259. Harrisson KA, Pavlova A, Amos JN, Radford JQ, Sunnucks P. 2014. Does reduced mobility through fragmented landscapes explain patch extinction patterns for three honeyeaters? Journal of Animal Ecology 83:616-627.

183 Hels T, Nachman G. 2002. Simulating viability of a spadefoot toad Pelobates fuscus metapopulation in a landscape fragmented by a road. Ecography 25:730-744. Hill JK, Hughes CL, Dytham C, Searle JB. 2006. Genetic diversity in butterflies: interactive effects of habitat fragmentation and climate-driven range expansion. Biology Letters 2:152-154. Hitchings SP, Beebee TJC. 1998. Loss of genetic diversity and fitness in common toad (Bufo bufo) populations isolated by inimical habitat. Journal of Evolutionary Biology 11:269-283. Hoehn M, Dimond W, Osborne W, Sarre SD. 2013. Genetic analysis reveals the costs of peri-urban development for the endangered grassland earless dragon. Conservation Genetics 14:1269- 1278. Hokit DG, Ascunce M, Ernst J, Branch LC, Clark AM. 2010. Ecological metrics predict connectivity better than geographic distance. Conservation Genetics 11:149-159. Holland GJ, Bennett AF. 2011. Recolonization of forest fragments by a native rodent following experimental 'extinctions'. Austral Ecology 36:521-529. Horreo JL, Martinez JL, Ayllon F, Pola IG, Monteoliva JA, Héland M, Garcia-Vazquez E. 2011. Impact of habitat fragmentation on the genetics of populations in dendritic landscapes. Freshwater Biology 56:2567-2579. Hovel KA, Wahle RA. 2010. Effects of habitat patchiness on American lobster movement across a gradient of predation risk and shelter competition. Ecology 91:1993-2002. Ims RA, Andreassen HP. 1999. Effects of experimental habitat fragmentation and connectivity on root vole demography. Journal of Animal Ecology 68:839-852. Iwai N, Shoda-Kagaya E. 2012. Population structure of an endangered frog (Babina subaspera) endemic to the Amami Islands: possible impacts of invasive predators on gene flow. Conservation Genetics 13:717-725. Janecka JE, Tewes ME, Laack L, Caso A, Grassman LI, Honeycutt RL. 2014. Loss of genetic diversity among ocelots in the United States during the 20th century linked to human induced population reductions. PLoS ONE 9:e89384. Johansson M, Primmer CR, Merilä J. 2007. Does habitat fragmentation reduce fitness and adaptability? A case study of the common frog (Rana temporaria). Molecular Ecology 16:2693-2700. Johansson M, Primmer CR, Sahlsten J, Merilä J. 2005. The influence of landscape structure on occurrence, abundance and genetic diversity of the common frog, Rana temporaria. Global Change Biology 11:1664-1679. Johnson JA, Bellinger MR, Toepfer JE, Dunn P. 2004. Temporal changes in allele frequencies and low effective population size in greater prairie-chickens. Molecular Ecology 13:2617-2630.

184 Jordan MA, Morris DA, Gibson SE. 2009. The influence of historical landscape change on genetic variation and population structure of a terrestrial salamander (Plethodon cinereus). Conservation Genetics 10:1647-1658. Kamler JF, Ballard WB, Wallace MC, Gipson PS. 2007. Diets of swift foxes (Vulpes velox) in continuous and fragmented prairie in northwestern Texas. Southwestern Naturalist 52:504- 510. Kappes H, Jordaens K, Van Houtte N, Hendrickx F, Maelfait JP, Lens L, Backeljau T. 2009. A land snail's view of a fragmented landscape. Biological Journal of the Linnean Society 98:839- 850. Kawakami K, Harada S, Suzuki T, Higuchi H. 2008. Genetic and morphological differences among populations of the Bonin Islands white-eye in Japan. Zoological Science 25:882-887. Kawamura K. 2005. Low genetic variation and inbreeding depression in small isolated populations of the Japanese rosy bitterling, Rhodeus ocellatus kurumeus. Zoological Science 22:517- 524. Keller I, Excoffier L, Largiadèr CR. 2005. Estimation of effective population size and detection of a recent population decline coinciding with habitat fragmentation in a ground beetle. Journal of Evolutionary Biology 18:90-100. Keller I, Largiadèr CR. 2003. Recent habitat fragmentation caused by major roads leads to reduction of gene flow and loss of genetic variability in ground beetles. Proceedings of the Royal Society B 270:417-423. Kenney J, Allendorf FW, McDougal C, Smith JLD. 2014. How much gene flow is needed to avoid inbreeding depression in wild tiger populations? Proceedings of the Royal Society B 281:20133337. Keyghobadi N, Roland J, Strobeck C. 2005. Genetic differentiation and gene flow among populations of the alpine butterfly, Parnassius smintheus, vary with landscape connectivity. Molecular Ecology 14:1897-1909. Kitanishi S, Yamamoto T, Edo K, Higashi S. 2012. Influences of habitat fragmentation by damming on the genetic structure of masu salmon populations in Hokkaido, Japan. Conservation Genetics 13:1017-1026. Klapwijk MJ, Lewis OT. 2011. Spatial ecology of multiple parasitoids of a patchily-distributed host: implications for species coexistence. Ecological Entomology 36:212-220. Klapwijk MJ, Lewis OT. 2012. Host-parasitoid dynamics in a fragmented landscape: holly trees, holly leaf miners and their parasitoids. Basic and Applied Ecology 13:94-105.

185 Kluger EC, Berlocher SH, Tooker JF, Hanks LM. 2011. Consequences of habitat fragmentation for the prairie-endemic weevil Haplorhynchites aeneus. Environmental Entomology 40:1388- 1396. Kossenko SM, Kaygorodova EY. 2007. Reproduction of the middle spotted woodpecker Dendrocopos medius in the Nerussa-Desna woodland, SW Russia, with particular reference to habitat fragmentation, weather conditions and food supply. Ardea 95:177-189. Kozfkay CC, Campbell MR, Meyer KA, Schill DJ. 2011. Influences of habitat and hybridization on the genetic structure of redband trout in the Upper Snake River Basin, Idaho. Transactions of the American Fisheries Society 140:282-295. Kraaijeveld-Smit FJL, Beebee TJC, Griffiths RA, Moore RD, Schley L. 2005. Low gene flow but high genetic diversity in the threatened Mallorcan midwife toad Alytes muletensis. Molecular Ecology 14:3307-3315. Kraaijeveld-Smit FJL, Lindenmayer DB, Taylor AC. 2002. Dispersal patterns and population structure in a small marsupial, Antechinus agilis, from two forests analysed using microsatellite markers. Australian Journal of Zoology 50:325-338. Kuehn R, Hindenlang KE, Holzgang O, Senn J, Stoeckl B, Sperisen C. 2007. Genetic effect of transportation infrastructure on roe deer populations (Capreolus capreolus). Journal of Heredity 98:13-22. Lancaster ML, Taylor AC, Cooper SJB, Carthew SM. 2011. Limited ecological connectivity of an arboreal marsupial across a forest/plantation landscape despite apparent resilience to fragmentation. Molecular Ecology 20:2258-2271. Landry P-A, Lapointe F-J. 1999. The genetic heterogeneity of deer mouse populations (Peromyscus maniculatus) in an insular landscape. Researches on Population Ecology 41:263-268. Lane-deGraaf KE, Fuentes A, Hollocher H. 2014. Landscape genetics reveal fine-scale boundaries in island populations of Indonesian long-tailed macaques. Landscape Ecology 29:1505- 1519. Laurence S, Smith MJ, Schulte-Hostedde AI. 2013. Effects of structural connectivity on fine scale population genetic structure of muskrat, Ondatra zibethicus. Ecology and Evolution 3:3524- 3535. Lee JS, Ruell EW, Boydston EE, Lyren LM, Alonso RS, Troyer JL, Crooks KR, VandeWoude S. 2012. Gene flow and pathogen transmission among bobcats (Lynx rufus) in a fragmented urban landscape. Molecular Ecology 21:1617-1631. Lee KE, Seddon JM, Corley SW, Ellis WAH, Johnston SD, de Villiers DL, Preece HJ, Carrick FN. 2010. Genetic variation and structuring in the threatened koala populations of southeast Queensland. Conservation Genetics 11:2091-2103.

186 Li Z, Xia L, Li Y, Yang Q, Liang M. 2006. Mitochondrial DNA variation and population structure of the yarkand hare Lepus yarkandensis. Acta Theriologica 51:243-253. Liu M, Zhang J, Chen Y, Compton SG, Chen X-Y. 2013. Contrasting genetic responses to population fragmentation in a coevolving fig and fig wasp across a mainland-island archipelago. Molecular Ecology 22:4384-4396. Ljungberg P, Hasper TB, Nilsson PA, Persson A. 2013. Effects of small-scale habitat fragmentation on predator-prey interactions in a temperate sea grass system. Marine Biology 160:667-675. Luquet E, Lena J-P, David P, Joly P, Lengagne T, Perrin N, Plénet S. 2011. Consequences of genetic erosion on fitness and phenotypic plasticity in European tree frog populations (Hyla arborea). Journal of Evolutionary Biology 24:99-110. Machkour-M'Rabet S, Hénaut Y, Calmé S, Legal L. 2012. When landscape modification is advantageous for protected species. The case of a synanthropic tarantula, Brachypelma vagans. Journal of Insect Conservation 16:479-488. Macreadie PI, Geraldi NR, Peterson CH. 2012. Preference for feeding at habitat edges declines among juvenile blue crabs as oyster reef patchiness increases and predation risk grows. Marine Ecology Progress Series 466:145-153. Mäki-Petäys H, Zakharov A, Viljakainen L, Corander J, Pamilo P. 2005. Genetic changes associated to declining populations of Formica ants in fragmented forest landscape. Molecular Ecology 14:733-742. Mapelli FJ, Mora MS, Mirol PM, Kittlein MJ. 2012. Population structure and landscape genetics in the endangered subterranean rodent Ctenomys porteousi. Conservation Genetics 13:165-181. Marsden CD, et al. 2012. Spatial and temporal patterns of neutral and adaptive genetic variation in the endangered African wild dog (Lycaon pictus). Molecular Ecology 21:1379-1393. Marsh DM, Thakur KA, Bulka KC, Clarke LB. 2004. Dispersal and colonization through open fields by a terrestrial, woodland salamander. Ecology 85:3396-3405. Marshall JCJ, Kingsbury BA, Minchella DJ. 2009. Microsatellite variation, population structure, and bottlenecks in the threatened copperbelly water snake. Conservation Genetics 10:465- 476. Matthysen E. 1999. Nuthatches (Sitta europaea : Aves) in forest fragments: demography of a patchy population. Oecologia 119:501-509. Matthysen E, Adriaensen F, Dhondt AA. 2001. Local recruitment of great and blue tits (Parus major, P-caeruleus) in relation to study plot size and degree of isolation. Ecography 24:33- 42.

187 Mattila ALK, Duplouy A, Kirjokangas M, Lehtonen R, Rastas P, Hanski I. 2012. High genetic load in an old isolated butterfly population. Proceedings of the National Academy of Sciences USA 109:E2496-E2505. Mergey M, Larroque J, Ruette S, Vandel JM, Helder R, Queney G, Devillard S. 2012. Linking habitat characteristics with genetic diversity of the European pine marten (Martes martes) in France. European Journal of Wildlife Research 58:909-922. Mikulíček P, Pišút P. 2012. Genetic structure of the marsh frog (Pelophylax ridibundus) populations in urban landscape. European Journal of Wildlife Research 58:833-845. Milko LV, Haddad NM, Lance SL. 2012. Dispersal via stream corridors structures populations of the endangered St. Francis' satyr butterfly (Neonympha mitchellii francisci). Journal of Insect Conservation 16:263-273. Miller MP, Bianchi CA, Mullins TD, Haig SM. 2013. Associations between forest fragmentation patterns and genetic structure in Pfrimer's parakeet (Pyrrhura pfrimeri), an endangered endemic to central Brazil's dry forests. Conservation Genetics 14:333-343. Millions DG, Swanson BJ. 2007. Impact of natural and artificial barriers to dispersal on the population structure of bobcats. Journal of Wildlife Management 71:96-102. Milton K, Lozier JD, Lacey EA. 2009. Genetic structure of an isolated population of mantled howler monkeys (Alouatta palliata) on Barro Colorado Island, Panama. Conservation Genetics 10:347-358. Monaghan MT, Spaak P, Robinson CT, Ward JV. 2001. Genetic differentiation of Baetis alpinus Pictet (Ephemeroptera : Baetidae) in fragmented alpine streams. Heredity 86:395-403. Monaghan MT, Spaak P, Robinson CT, Ward JV. 2002. Population genetic structure of 3 alpine stream insects: influences of gene flow, demographics, and habitat fragmentation. Journal of the North American Benthological Society 21:114-131. Morales P, Vila I, Poulin E. 2011. Genetic structure in remnant populations of an endangered cyprinodontid fish, Orestias ascotanensis, endemic to the Ascotan salt pan of the Altiplano. Conservation Genetics 12:1639-1643. Morita K, Yokota A. 2002. Population viability of stream-resident salmonids after habitat fragmentation: a case study with white-spotted charr (Salvelinus leucomaenis) by an individual based model. Ecological Modelling 155:85-94. Mortelliti A, Boitani L. 2008. Interaction of food resources and landscape structure in determining the probability of patch use by carnivores in fragmented landscapes. Landscape Ecology 23:285-298.

188 Mouret V, Guillaumet A, Cheylan M, Pottier G, Ferchaud A-L, Crochet P-A. 2011. The legacy of ice ages in mountain species: post-glacial colonization of mountain tops rather than current range fragmentation determines mitochondrial genetic diversity in an endemic Pyrenean rock lizard. Journal of Biogeography 38:1717-1731. Nair A, Gopalan SV, George S, Kumar KS, Shikano T, Merilä J. 2012. Genetic variation and differentiation in Indirana beddomii frogs endemic to the Western Ghats biodiversity hotspot. Conservation Genetics 13:1459-1467. Neuwald JL. 2010. Population isolation exacerbates conservation genetic concerns in the endangered Amargosa vole, Microtus californicus scirpensis. Biological Conservation 143:2028-2038. Ni Q-Y, Huang B, Liang Z-L, Wang X-W, Jiang X-L. 2014. Dietary variability in the western black crested gibbon (Nomascus concolor) inhabiting an isolated and disturbed forest fragment in southern Yunnan, China. American Journal of Primatology 76:217-229. Niedziałkowska M, Fontaine MC, Jędrzejewska B. 2012. Factors shaping gene flow in red deer (Cervus elaphus) in seminatural landscapes of central Europe. Canadian Journal of Zoology 90:150-162. Nielsen JL, Carpanzano C, Fountain MC, Gan CA. 1997. Mitochondrial DNA and nuclear microsatellite diversity in hatchery and wild Oncorhynchus mykiss from freshwater habitats in southern California. Transactions of the American Fisheries Society 126:397-417. Noël S, Ouellet M, Galois P, Lapointe F-J. 2007. Impact of urban fragmentation on the genetic structure of the eastern red-backed salamander. Conservation Genetics 8:599-606. Nunziata SO, Lannoo MJ, Robb JR, Karns DR, Lance SL, Richter SC. 2013. Population and conservation genetics of crawfish frogs, Lithobates areolatus, at their northeastern range limit. Journal of Herpetology 47:361-368. Ordóñez-Gómez JD, Arroyo-Rodríguez V, Nicasio-Arzeta S, Cristóbal-Azkarate J. 2015. Which is the appropriate scale to assess the impact of landscape spatial configuration on the diet and behavior of spider monkeys? American Journal of Primatology 77:56-65. Osborne JL, Williams IH. 2001. Site constancy of bumble bees in an experimentally patchy habitat. Agriculture Ecosystems & Environment 83:129-141. Paolucci LN, Solar RRC, Sobrinho TG, Sperber CF, Schoereder JH. 2012. How does small-scale fragmentation affect litter-dwelling ants? The role of isolation. Biodiversity and Conservation 21:3095-3105. Pavlacky Jr. DC, Possingham HP, Lowe AJ, Prentis PJ, Green DJ, Goldizen AW. 2012. Anthropogenic landscape change promotes asymmetric dispersal and limits regional patch occupancy in a spatially structured bird population. Journal of Animal Ecology 81:940-952.

189 Pavlova A, Amos JN, Goretskaia MI, Beme IR, Buchanan KL, Takeuchi N, Radford JQ, Sunnucks P. 2012. Genes and song: genetic and social connections in fragmented habitat in a woodland bird with limited dispersal. Ecology 93:1717-1727. Peacock MM, Dochtermann NA. 2012. Evolutionary potential but not extinction risk of Lahontan cutthroat trout (Oncorhynchus clarkii henshawi) is associated with stream characteristics. Canadian Journal of Fisheries and Aquatic Sciences 69:615-626. Pereoglou F, Lindenmayer DB, Macgregor C, Ford F, Wood J, Banks SC. 2013. Landscape genetics of an early successional specialist in a disturbance-prone environment. Molecular Ecology 22:1267-1281. Pernetta AP, Allen JA, Beebee TJC, Reading CJ. 2011. Fine-scale population genetic structure and sex-biased dispersal in the smooth snake (Coronella austriaca) in southern England. Heredity 107:231-238. Pittman SE, King TL, Faurby S, Dorcas ME. 2011. Demographic and genetic status of an isolated population of bog turtles (Glyptemys muhlenbergii): implications for managing small populations of long-lived animals. Conservation Genetics 12:1589-1601. Polic D, Fiedler K, Nell C, Grill A. 2014. Mobility of ringlet butterflies in high-elevation alpine grassland: effects of habitat barriers, resources and age. Journal of Insect Conservation 18:1153-1161. Provencher L, Riechert SE. 1994. Model and field-test of prey control effects by spider assemblages. Environmental Entomology 23:1-17. Quéméré E, Louis Jr. EE, Ribéron A, Chikhi L, Crouau-Roy B. 2010. Non-invasive conservation genetics of the critically endangered golden-crowned sifaka (Propithecus tattersalli): high diversity and significant genetic differentiation over a small range. Conservation Genetics 11:675-687. Ramirez MG, Haakonsen KE. 1999. Gene flow among habitat patches on a fragmented landscape in the spider Argiope trifasciata (Araneae : Araneidae). Heredity 83:580-585. Rantalainen M-L, Kontiola L, Haimi J, Fritze H, Setälä H. 2004. Influence of resource quality on the composition of soil decomposer community in fragmented and continuous habitat. Soil Biology & Biochemistry 36:1983-1996. Reding DM, Freed LA, Cann RL, Fleischer RC. 2010. Spatial and temporal patterns of genetic diversity in an endangered Hawaiian honeycreeper, the Hawaii akepa (Loxops coccineus coccineus). Conservation Genetics 11:225-240. Reid SM, Wilson CC, Mandrak NE, Carl LM. 2008. Population structure and genetic diversity of black redhorse (Moxostoma duquesnei) in a highly fragmented watershed. Conservation Genetics 9:531-546.

190 Remón N, Galán P, Vila M, Arribas O, Naveira H. 2013. Causes and evolutionary consequences of population subdivision of an Iberian mountain lizard, Iberolacerta monticola. PLoS ONE 8:e66034. Richmond JQ, Reid DT, Ashton KG, Zamudio KR. 2009. Delayed genetic effects of habitat fragmentation on the ecologically specialized Florida sand skink (Plestiodon reynoldsi). Conservation Genetics 10:1281-1297. Roberts JH, Angermeier PL, Hallerman EM. 2013. Distance, dams and drift: what structures populations of an endangered, benthic stream fish? Freshwater Biology 58:2050-2064. Roffler GH, Talbot SL, Luikart G, Sage GK, Pilgrim KL, Adams LG, Schwartz MK. 2014. Lack of sex-biased dispersal promotes fine-scale genetic structure in alpine ungulates. Conservation Genetics 15:837-851. Ruell EW, et al. 2012. Urban habitat fragmentation and genetic population structure of bobcats in coastal southern California. American Midland Naturalist 168:265-280. Safner T, Miaud C, Gaggiotti O, Decout S, Rioux D, Zundel S, Manel S. 2011. Combining demography and genetic analysis to assess the population structure of an amphibian in a human-dominated landscape. Conservation Genetics 12:161-173. Sandlund OT, Karlsson S, Thorstad EB, Berg OK, Kent MP, Norum ICJ, Hindar K. 2014. Spatial and temporal genetic structure of a river-resident Atlantic salmon (Salmo salar) after millennia of isolation. Ecology and Evolution 4:1538-1554. Sato JJ, Kawakami T, Tasaka Y, Tamenishi M, Yamaguchi Y. 2014. A few decades of habitat fragmentation has reduced population genetic diversity: a case study of landscape genetics of the large Japanese field mouse, Apodemus speciosus. Mammal Study 39:1-10. Sato T, Demise T, Kubota H, Nagoshi M, Watanabe K. 2010. Hybridization, isolation, and low genetic diversity of kirikuchi char, the southernmost populations of the genus Salvelinus. Transactions of the American Fisheries Society 139:1758-1774. Sato T, Harada Y. 2008. Loss of genetic variation and effective population size of kirikuchi charr: implications for the management of small, isolated salmonid populations. Animal Conservation 11:153-159. Schiffer M, Kennington WJ, Hoffmann AA, Blacket MJ. 2007. Lack of genetic structure among ecologically adapted populations of an Australian rainforest Drosophila species as indicated by microsatellite markers and mitochondrial DNA sequences. Molecular Ecology 16:1687- 1700. Schmuki C, Vorburger C, Runciman D, Maceachern S, Sunnucks P. 2006. When log-dwellers meet loggers: impacts of forest fragmentation on two endemic log-dwelling beetles in southeastern Australia. Molecular Ecology 15:1481-1492.

191 Schrey AW, Ashton KG, Heath S, Mushinsky HR, McCoy ED. 2012. Range-wide genetic analysis of the threatened bluetail mole skink identifies similar genetic structure with sympatric lizards. Journal of Herpetology 46:241-247. Schwalm D, Waits LP, Ballard WB. 2014. Little fox on the prairie: genetic structure and diversity throughout the distribution of a grassland carnivore in the United States. Conservation Genetics 15:1503-1514. Seppä P, Laurila A. 1999. Genetic structure of island populations of the anurans Rana temporaria and Bufo bufo. Heredity 82:309-317. Serieys LEK, Lea A, Pollinger JP, Riley SPD, Wayne RK. 2015. Disease and freeways drive genetic change in urban bobcat populations. Evolutionary Applications 8:75-92. Shanahan DF, Possingham HP, Riginos C. 2011. Models based on individual level movement predict spatial patterns of genetic relatedness for two Australian forest birds. Landscape Ecology 26:137-148. Shepherd B, Whittington J. 2006. Response of wolves to corridor restoration and human use management. Ecology and Society 11:1. Sielezniew M, Rutkowski R. 2012. Population isolation rather than ecological variation explains the genetic structure of endangered myrmecophilous butterfly Phengaris (=Maculinea) arion. Journal of Insect Conservation 16:39-50. Slack WT, Sumners JA, Rooney AP, Taylor CM. 2010. Conservation genetics of the threatened Bayou darter (Percidae: Etheostoma rubrum) in the Bayou Pierre system of southwestern Mississippi. Copeia 1:176-180. Smith JE, Batzli GO. 2006. Dispersal and mortality of prairie voles (Microtus ochrogaster) in fragmented landscapes: a field experiment. Oikos 112:209-217. Sommer S. 2003. Effects of habitat fragmentation and changes of dispersal behaviour after a recent population decline on the genetic variability of noncoding and coding DNA of a monogamous Malagasy rodent. Molecular Ecology 12:2845-2851. Souza FL, Cunha AF, Oliveira MA, Pereira GAG, Pinheiro HP, dos Reis SF. 2002. Partitioning of molecular variation at local spatial scales in the vulnerable neotropical freshwater turtle, Hydromedusa maximiliani (Testudines, Chelidae): implications for the conservation of aquatic organisms in natural hierarchical systems. Biological Conservation 104:119-126. Stephens HC, Schmuki C, Burridge CP, O'Reilly-Wapstra JM. 2013. Habitat fragmentation in forests affects relatedness and spatial genetic structure of a native rodent, Rattus lutreolus. Austral Ecology 38:568-580.

192 Sterling KA, Reed DH, Noonan BP, Warren Jr. ML. 2012. Genetic effects of habitat fragmentation and population isolation on Etheostoma raneyi (Percidae). Conservation Genetics 13:859- 872. Stevens MI, Hogg ID. 2003. Long-term isolation and recent range expansion from glacial refugia revealed for the endemic springtail Gomphiocephalus hodgsoni from Victoria Land, Antarctica. Molecular Ecology 12:2357-2369. Stoner DC, Wolfe ML, Mecham C, Mecham MB, Durham SL, Choate DM. 2013. Dispersal behaviour of a polygynous carnivore: do cougars Puma concolor follow source-sink predictions? Wildlife Biology 19:289-301. Tawatao N, Harper NE, Mohamed M, Khen CV, Searle JB, Hill JK. 2011. Impacts of forest fragmentation on the genetic diversity and population structure of Pachycondyla obscurans in Sabah, Malaysian Borneo. Asian Myrmecology 4:59-68. Taylor AC, Walker FM, Goldingay RL, Ball T, van der Ree R. 2011. Degree of landscape fragmentation influences genetic isolation among populations of a gliding mammal. PLoS ONE 6:e26651. Thomé MTC, Zamudio KR, Giovanelli JGR, Haddad CFB, Baldissera Jr. FA, Alexandrino J. 2010. Phylogeography of endemic toads and post-Pliocene persistence of the Brazilian Atlantic Forest. Molecular Phylogenetics and Evolution 55:1018-1031. Thomé MTC, Zamudio KR, Haddad CFB, Alexandrino J. 2014. Barriers, rather than refugia, underlie the origin of diversity in toads endemic to the Brazilian Atlantic Forest. Molecular Ecology 23:6152-6164. Turner TF, Osborne MJ, Moyer GR, Benavides MA, Alò D. 2006. Life history and environmental variation interact to determine effective population to census size ratio. Proceedings of the Royal Society B 273:3065-3073. Uimaniemi L, Orell M, Kvist L, Jokimäki J, Lumme J. 2003. Genetic variation of the Siberian tit Parus cinctus populations at the regional level: a mitochondrial sequence analysis. Ecography 26:98-106. van Apeldoorn RC, Celada C, Nieuwenhuizen W. 1994. Distribution and dynamics of the red squirrel (Sciurus vulgaris L.) in a landscape with fragmented habitat. Landscape Ecology 9:227-235. Van Houdt JKJ, Pinceel J, Flamand M-C, Briquet M, Dupont E, Volckaert FAM, Baret PV. 2005. Migration barriers protect indigenous brown trout (Salmo trutta) populations from introgression with stocked hatchery fish. Conservation Genetics 6:175-191.

193 Vangestel C, Braeckman BP, Matheve H, Lens L. 2010. Constraints on home range behaviour affect nutritional condition in urban house sparrows (Passer domesticus). Biological Journal of the Linnean Society 101:41-50. Vucetich LM, Vucetich JA, Joshi CP, Waite TA, Peterson RO. 2001. Genetic (RAPD) diversity in Peromyscus maniculatus populations in a naturally fragmented landscape. Molecular Ecology 10:35-40. Walker FM, Sunnucks P, Taylor AC. 2008. Evidence for habitat fragmentation altering within- population processes in wombats. Molecular Ecology 17:1674-1684. Wang C, Yu X, Tong J. 2007. Microsatellite diversity and population genetic structure of redfin culter (Culter erythropterus) in fragmented lakes of the Yangtze River. Hydrobiologia 586:321-329. Watanabe K, Monaghan MT, Omura T. 2008. Longitudinal patterns of genetic diversity and larval density of the riverine caddisfly Hydropsyche orientalis (Trichoptera). Aquatic Sciences 70:377-387. Watanabe K, Omura T. 2007. Relationship between reservoir size and genetic differentiation of the stream caddisfly Stenopsyche Marmorata. Biological Conservation 136:203-211. Weyer J, Jorgensen D, Schmitt T, Maxwell TJ, Anderson CD. 2014. Lack of detectable genetic differentiation between den populations of the prairie rattlesnake (Crotalus viridis) in a fragmented landscape. Canadian Journal of Zoology 92:837-846. White NE, Bunce M, Mawson PR, Dawson R, Saunders DA, Allentoft ME. 2014. Identifying conservation units after large-scale land clearing: a spatio-temporal molecular survey of endangered white-tailed black cockatoos (Calyptorhynchus spp.). Diversity and Distributions 20:1208-1220. Whiteley AR, Coombs JA, Hudy M, Robinson Z, Colton AR, Nislow KH, Letcher BH. 2013. Fragmentation and patch size shape genetic structure of brook trout populations. Canadian Journal of Fisheries and Aquatic Sciences 70:678-688. Whiteley AR, Spruell P, Rieman BE, Allendorf FW. 2006. Fine-scale genetic structure of bull trout at the southern limit of their distribution. Transactions of the American Fisheries Society 135:1238-1253. Williams BL, Brawn JD, Paige KN. 2003. Landscape scale genetic effects of habitat fragmentation on a high gene flow species: Speyeria idalia (Nymphalidae). Molecular Ecology 12:11-20. Wilmer JW, Wilcox C. 2007. Fine scale patterns of migration and gene flow in the endangered mound spring snail, Fonscochlea accepta (Mollusca : Hydrobiidae) in arid Australia. Conservation Genetics 8:617-628.

194 With KA, Pavuk DM, Worchuck JL, Oates RK, Fisher JL. 2002. Threshold effects of landscape structure on biological control in agroecosystems. Ecological Applications 12:52-65. Wozney KM, Haxton TJ, Kjartanson S, Wilson CC. 2011. Genetic assessment of lake sturgeon (Acipenser fulvescens) population structure in the Ottawa River. Environmental Biology of Fishes 90:183-195. Wu Y, Xia L, Zhang Q, Lei F, Yang Q. 2010. Genetic diversity in the male-specific SRY gene of Lepus yarkandensis. Chinese Science Bulletin 55:834-840. Wynne IR, Loxdale HD, Brookes CP, Woiwod IP. 2003. Genetic structure of fragmented November moth (Lepidoptera : Geometridae) populations in farmland. Biological Journal of the Linnean Society 78:467-477. Yamazaki Y, Nakamura T, Sasaki M, Nakano S, Nishio M. 2014. Decreasing genetic diversity in wild and captive populations of endangered Itasenpara bittering (Acheilognathus longipinnis) in the Himi region, central Japan, and recommendations for conservation. Conservation Genetics 15:921-932. Yamazaki Y, Yamano A, Oura K. 2011. Recent microscale disturbance of gene flow in threatened fluvial lamprey, Lethenteron sp N, living in a paddy water system. Conservation Genetics 12:1373-1377. Yannic G, Pellissier L, Le Corre M, Dussault C, Bernatchez L, Cote SD. 2014. Temporally dynamic habitat suitability predicts genetic relatedness among caribou. Proceedings of the Royal Society B 281:20140502. Yumnam B, Jhala YV, Qureshi Q, Maldonado JE, Gopal R, Saini S, Srinivas Y, Fleischer RC. 2014. Prioritizing tiger conservation through landscape genetics and habitat linkages. PLoS ONE 9:e111207.

Zaccara S, Antognazza CM, Buonerba L, Britton R, Crosa G. 2014. Human-mediated contact zone between endemic and invasive Barbus species (Osteichthyes: Cyprinidae) in a regulated lowland river: genetic inferences and conservation implications. Italian Journal of Zoology 81:571-583. Zamudio KR, Wieczorek AM. 2007. Fine-scale spatial genetic structure and dispersal among spotted salamander (Ambystoma maculatum) breeding populations. Molecular Ecology 16:257-274. Zavodna M, Arens P, Van Dijk PJ, Partomihardjo T, Vosman B, Van Damme JMM. 2005. Pollinating fig wasps: genetic consequences of island recolonization. Journal of Evolutionary Biology 18:1234-1243.

195 Zeigler SL, De Vleeschouwer KM, Raboy BE. 2013. Assessing extinction risk in small metapopulations of golden-headed lion tamarins (Leontopithecus chrysomelas) in Bahia State, Brazil. Biotropica 45:528-535. Zhou Y-B, Yang M-X, Yihua W-l, Liu G-m, Wang H-y, Wei J-G, Jiang Q-W. 2011. Effect of habitat fragmentation on the schistosome-transmitting snail Oncomelania hupensis in a mountainous area of China. Transactions of the Royal Society of Tropical Medicine and Hygiene 105:189-196. Zhu L, Zhang S, Gu X, Wei F. 2011. Significant genetic boundaries and spatial dynamics of giant pandas occupying fragmented habitat across southwest China. Molecular Ecology 20:1122- 1132. Zuberogoitia I, Zalewska H, Zabala J, Zalewski A. 2013. The impact of river fragmentation on the population persistence of native and alien mink: an ecological trap for the endangered European mink. Biodiversity and Conservation 22:169-186.

196 APPENDIX B

Table B.1. Unpublished heterophil:lymphocyte ratios collected by Maron et al. (2012) from 51 eastern yellow robins near Meandarra and Moonie, Queensland, Australia. Individual H:L Ratio 1 0.2326 2 0.8741 3 0.4275 4 0.9574 5 0.3459 6 0.6939 7 0.2563 8 0.4552 9 0.1121 10 0.0896 11 0.1034 12 0.4843 13 0.3228 14 0.0716 15 0.1630 16 0.7241 17 0.9200 18 0.2938 19 1.2927 20 0.1099 21 0.1674 22 0.1928 23 0.5519 24 0.4854 25 0.0586 26 0.1813 27 0.9278 28 0.1054 29 0.1626 30 0.0757

197 Individual H:L Ratio 31 0.1852 32 0.4179 33 0.4204 34 0.2070 35 0.1356 36 0.2200 37 0.3422 38 0.1527 39 0.5081 40 0.1233 41 0.6091 42 0.2123 43 0.1909 44 0.1271 45 0.1535 46 0.1082 47 0.1262 48 0.1373 49 0.8867 50 0.6683 51 0.1804

198 APPENDIX C

Table C.1. Results of model averaging testing the effectiveness of using H:L ratios to predict eastern yellow robin extirpations *.

2 Model K LL AICc Δ ωi R Null 1 -16.37 34.83 0.00 0.07 - No. shrubs 2 -15.65 35.54 0.71 0.05 -0.18 Remnant woodland within 500 m 2 -15.65 35.55 0.72 0.05 -0.18 Gilgai 2 -15.68 35.62 0.78 0.05 -0.16 Northing 2 -15.88 36.01 1.17 0.04 -0.20 No. patches 2 -16.07 36.39 1.56 0.03 -0.22 Grass cover + no. shrubs 3 -15.18 36.88 2.05 0.03 -0.18 Grass cover 2 -16.34 36.94 2.11 0.02 -0.21 Gilgai + remnant woodland within 500 m 3 -15.24 36.99 2.16 0.02 -0.20 No. patches + remnant woodland within 500 m 3 -15.29 37.08 2.25 0.02 -0.23 H:L ratio 2 -16.42 37.09 2.25 0.02 -0.25 No. shrubs + remnant woodland within 500 m 3 -15.33 37.16 2.33 0.02 -0.22 Grass cover + remnant woodland within 500 m 3 -15.34 37.19 2.36 0.02 -0.20 Gilgai + northing 3 -15.39 37.29 2.46 0.02 -0.22 No. patches + no. shrubs 3 -15.40 37.31 2.48 0.02 -0.24 Northing + no. shrubs 3 -15.41 37.33 2.50 0.02 -0.24 No. patches + northing 3 -15.42 37.34 2.51 0.02 -0.24 Gilgai + no. patches 3 -15.51 37.54 2.71 0.02 -0.23 Gilgai + grass cover 3 -15.52 37.56 2.72 0.02 -0.20

199 2 Model K LL AICc Δ ωi R Gilgai + no. shrubs 3 -15.56 37.64 2.81 0.02 -0.23 Northing + remnant woodland within 500 m 3 -15.66 37.83 3.00 0.02 -0.25 Remnant woodland within 500 m + H:L ratio 3 -15.68 37.87 3.04 0.02 -0.26 Gilgai + H:L ratio 3 -15.71 37.93 3.10 0.01 -0.25 No. shrubs + H:L ratio 3 -15.72 37.95 3.12 0.01 -0.27 Grass cover + no. shrubs + remnant woodland within 500 m 4 -14.58 38.04 3.21 0.01 -0.19 Gilgai + grass cover + remnant woodland within 500 m 4 -14.66 38.18 3.35 0.01 -0.18 Grass cover + northing 3 -15.92 38.35 3.51 0.01 -0.26 Northing + H:L ratio 3 -15.94 38.39 3.56 0.01 -0.29 Grass cover + no. patches 3 -16.10 38.70 3.87 0.01 -0.27 Grass cover + no. patches + remnant woodland within 500 m 4 -14.92 38.70 3.87 0.01 -0.22 No. patches + H:L ratio 3 -16.13 38.77 3.93 0.01 -0.31

2 2 * Abbreviations: number of parameters (K), maximised log likelihood (LL), delta AICc value (Δ), Akaike weight (ωi), and Nagelkerke R values (R ). Model averaging was based on the 95% confidence set of models. Only those models with a Δ < 4 are displayed.

200 Table C.2. Results of model averaging testing the relationship between remnant woodland extent within 500 m of each study site and eastern yellow robin occurrence *.

2 Model K LL AICc Δ ωi R Grass cover + remnant woodland within 500 m 3 -19.93 46.49 0.00 0.13 0.24 Grass cover + northing + no. shrubs + remnant woodland within 500 m 5 -17.55 46.77 0.28 0.11 0.34 Grass cover + no. shrubs + remnant woodland within 500 m 4 -18.85 46.78 0.29 0.11 0.27 Grass cover + northing + remnant woodland within 500 m 4 -19.03 47.13 0.64 0.09 0.29 Gilgai + grass cover + remnant woodland within 500 m 4 -19.60 48.29 1.79 0.05 0.24 Remnant woodland within 500 m 2 -22.14 48.59 2.10 0.04 0.13 Gilgai + grass cover + northing + no. shrubs + remnant woodland within 500 m 6 -17.18 48.77 2.27 0.04 0.34 Gilgai + grass cover + no. shrubs + remnant woodland within 500 m 5 -18.60 48.87 2.38 0.04 0.27 Grass cover + no. patches + remnant woodland within 500 m 4 -19.90 48.89 2.39 0.04 0.24 Gilgai + grass cover + northing + remnant woodland within 500 m 5 -18.62 48.90 2.40 0.04 0.29 Grass cover + no. patches + no. shrubs + remnant woodland within 500 m 5 -18.83 49.33 2.84 0.03 0.27 Grass cover + no. patches + northing + no. shrubs + remnant woodland within 500 m 6 -17.50 49.40 2.91 0.03 0.34 Northing + remnant woodland within 500 m 3 -21.53 49.69 3.19 0.03 0.16 Grass cover + no. patches + northing + remnant woodland within 500 m 5 -19.02 49.72 3.22 0.03 0.29

201 Table C.3. Results of model averaging testing the relationship between the broader landscape context (total woodland extent within 5 km of each study site) and eastern yellow robin occurrence *.

2 Model K LL AICc Δ ωi R Grass cover + no. wilga + total woodland within 500 m + total woodland within 5 km 6 -25.02 63.63 0.00 0.20 0.40 Grass cover + no. wilga + total woodland within 500 m 5 -26.46 64.03 0.40 0.16 0.35 Grass cover + gilgai + no. wilga + total woodland within 500 m 6 -25.75 65.08 1.44 0.10 0.37 No. wilga + total woodland within 500 m + total woodland within 5 km 5 -27.01 65.14 1.50 0.09 0.34 Grass cover + gilgai + no. wilga + total woodland within 500 m + total woodland within 5 km 7 -24.52 65.19 1.56 0.09 0.42 Grass cover + total woodland within 500 m 4 -28.34 65.40 1.77 0.08 0.33 Grass cover + total woodland within 500 m + total woodland within 5 km 5 -27.29 65.70 2.07 0.07 0.37 Gilgai + no. wilga + total woodland within 500 m + total woodland within 5 km 6 -26.37 66.32 2.68 0.05 0.37 Total woodland within 500 m + total woodland within 5 km 4 -29.04 66.80 3.17 0.04 0.31 Grass cover + gilgai + total woodland within 500 m 5 -28.11 67.32 3.69 0.03 0.34 No. wilga + total woodland within 500 m 4 -29.30 67.33 3.69 0.03 0.29 Total woodland within 500 m 3 -30.54 67.51 3.88 0.03 0.26

202 Table C.4. Results of model averaging testing the relationship between the broader landscape context (accessible woodland extent within 5 km of each study site) and eastern yellow robin occurrence *.

2 Model K LL AICc Δ ωi R Grass cover + no. wilga + total woodland within 500 m + accessible woodland within 5 km 6 -24.84 63.27 0.00 0.23 0.41 Grass cover + no. wilga + total woodland within 500 m 5 -26.46 64.03 0.76 0.15 0.35 No. wilga + total woodland within 500 m + accessible woodland within 5 km 5 -26.94 64.99 1.72 0.10 0.34 Grass cover + gilgai + no. wilga + total woodland within 500 m + accessible woodland within 7 -24.46 65.07 1.80 0.09 0.42 5 km Grass cover + gilgai + no. wilga + total woodland within 500 m 6 -25.75 65.08 1.81 0.09 0.37 Grass cover + total woodland within 500 m 4 -28.34 65.40 2.13 0.08 0.33 Grass cover + total woodland within 500 m + accessible woodland within 5 km 5 -27.24 65.60 2.33 0.07 0.37 Gilgai + no. wilga + total woodland within 500 m + accessible woodland within 5 km 6 -26.46 66.51 3.24 0.04 0.36 Total woodland within 500 m + accessible woodland within 5 km 4 -29.05 66.84 3.57 0.04 0.31

203 APPENDIX D

204