Edge Effects in Fire-Prone Landscapes: Ecological Importance and Implications for Fauna

Home , Fauna (deity)

EDGE EFFECTS IN FIRE-PRONE LANDSCAPES: ECOLOGICAL IMPORTANCE AND IMPLICATIONS FOR FAUNA

KATE ANNA PARKINS

orcid.org/0000-0002-0882-638X

Submitted in total fulfilment of the requirements of the degree of

Doctor of Philosophy

June 2018

School of Ecosystem and Forest Sciences

Faculty of Science

University of Melbourne

ABSTRACT

The overarching aim of this thesis was to investigate the ecological importance of fire edges, focusing on the influence of fire-induced edge effects on fauna in forested landscapes. Edges are ubiquitous environmental features, occurring in a wide range of ecosystems and across multiple spatial scales. Edges have been extensively researched in some contexts, particularly agricultural and urban landscapes. Accordingly, much of our understanding about how edges influence animals comes from highly modified ecosystems. Fire is an agent of edge creation and a globally important driver of biome distribution and community composition, yet little is known about how fire edges affect ecological processes in flammable ecosystems.

In this thesis I review the literature on fire, fauna and edge effects to summarise current knowledge of faunal response to fire edges and identify knowledge gaps (Chapter 2). I developed a conceptual model for predicting edges effects in fire-prone landscapes, combining several drivers of faunal-fire responses. Fire-generated edge effects were found to differ from edges in modified systems, being temporally dynamic, spatially complex and characterised by the strength of the interaction between components of the disturbance regime and other biophysical factors.

In Chapter 3 I investigated the response of ground-dwelling mammals to burnt/unburnt edges created by prescribed burning. I used a space-for-time substitution design to explore how species use of fire edges changes over time as the burnt side of the edge regenerates. I found that understorey complexity was reduced on the burnt side of edges for the first two years after fire. Larger animals with generalist resource requirements were more active at burnt edges immediately after fire, whereas small mammals were generally less active on burnt edges for up to 3 years. Species were not following patterns of temporal change in vegetation structure, with high usage during times of reduced understorey complexity and low usage when complexity was high. This suggests that habitat change is not a good predictor of animal use at fire edges and that other important processes are likely occurring. For example, foxes and cats were using the burnt side of edges immediately after fire, which may have important implications for the long- term persistence of native fauna if changes in habitat structure at fire edges cause predation rates to increase.

In Chapter 4 I assessed the trade-off between deploying more detection units or extending the length of the sampling period on two frequently assessed variables in camera trapping studies –

iii species richness and detection probability. The trade-off between these two factors is expected to affect data quality, but there is little information about their relative influence. I examined the trade-off between increasing deployment time or increasing the number of detection units on species richness and detectability (Chapter 4). I found that that increasing the number of cameras deployed per site was an effective method for increasing the detection of ground-dwelling mammals. Multiple cameras and longer deployment times were necessary to detect a high proportion of species present. Increasing the number of cameras or increasing deployment length resulted in high overall detectability for the more detectable species, but multiple cameras were required to achieve high detectability in a reasonable time frame (<50 days) for less detectable species.

In Chapter 5 I investigated resource selection of a semi-arboreal mammal eight years after a major wildfire using GPS telemetry. Survival and persistence of animals after fire is largely driven by the abundance and distribution of remaining resources and the rate at which key habitat components regenerate or re-accumulate. I found that resource selection for the mountain brushtail possum (Trichosurus cunninghami) often depended on the sex of the animal and forest type, suggesting that considering spatial changes in resource availability and demographic class may be necessary to accurately determine patterns of resource selection after a major wildfire.

This thesis adds to the body of knowledge on the ecological importance of fire edges and their implications for fauna, while providing several important conceptual and methodological advances in the study of ecology. Edges are pervasive and important environmental features that require further attention. Mechanistic approaches based on the strength of habitat associations and resource availability may help to clarify the nature and strength of edge effects in fire-prone landscapes and improve predictive models. A better understanding of fire edges will enable land managers to integrate the needs of biodiversity in to future fire management planning.

DECLARATION

This is to certify that:

I. The thesis comprises only my original work towards the PhD except where indicated in the preface.

II. Due acknowledgement has been made in the text to all other material used.

III. The thesis is fewer than 100,000 words in length, exclusive of tables, maps, bibliographies and appendices.

Kate Parkins

June 2018

PREFACE

This thesis comprises four papers that present my PhD research (Chapters 2 - 5). The General Introduction (Chapter 1) outlines the key concepts and themes underlying the research. The Synthesis (Chapter 6) summarises key results, discusses implications for management and highlights areas for future research.

The data chapters have been prepared as stand-alone papers for publication in collaboration with co-authors. One of the chapters has been published (Chapter 2), and two are in preparation (Chapters 3 & 4). As such, this thesis does not contain a chapter describing the study area, study species and common methods. There is also some overlap in chapter content particularly with regards to descriptions of study areas. The pronoun ‘we’ is used instead of ‘I’ in recognition of the co-authors’ contributions. A comprehensive review of relevant literature is provided in Chapter 2.

Some assistance was provided by an electronics engineer (Kean Maizels) with the development and construction of the circuit boards used in the GPS collars as part of Chapter 5.

Field work was conducted under the National Parks Act (Research Permit Number 10007387). Faunal surveys were conducted with ethics approval from the University of Melbourne Animal Ethics Committee (Research Permit Numbers 1413324 and 1513673).

The chapters, co-authors and my contributions are as follows:

Thesis Manuscript title My contribution chapter

Chapter 2 Parkins K, York A, Di Stefano J. (2018) Edge effects in K Parkins 80% fire-prone landscapes: ecological importance and implications for fauna. Ecol Evol. 00:1-12. http://doi.org/10.1002/ece3.4076

Contributions: KP, JD and AY conceived the ideas and developed the structure. KP led the writing of the manuscript with editorial guidance from JD and AY.

Chapter 3 Parkins K, Scott A, Swan M, Sitters H, Di Stefano J, York K Parkins 70% A. (In prep). Habitat use at fire edges: Do animals follow temporal patterns of habitat change?

Contributions: KP led the fieldwork, with assistance from AS, MS, HS and JD. KP analysed the data with assistance from JD. KP wrote the manuscript with editorial assistance from JD, AY, HS, MS.

Chapter 4 Parkins K, Penman T, Di Stefano J, York A. (In prep). K Parkins 80% Increasing detectability in camera trap surveys: more cameras, more days, or both?

Contributions: KP, JD, AY conceived and designed the study. KP conducted the fieldwork, with assistance from AS, MS, HS and JD. KP analysed the data with assistance from TP and JD. KP wrote the manuscript with editorial assistance from JD, AY and TP.

Chapter 5 Parkins K, Maizels K, Di Stefano J, York A. The devil is in K Parkins 85% the detail: Forest type and sex influence post-fire resource selection in a semi-arboreal mammal.

Contributions: KP, JD, AY conceived and designed the study. KM assisted with the design and construction of the GPS devices. KP conducted the fieldwork, with assistance from JD and volunteers. KP analysed the data with assistance from JD and Bronwyn Hradsky. KP wrote the manuscript with editorial assistance from JD and AY.

vii

I contributed to the following manuscript during my candidature:

Fischer M, Parkins K, Maizels K. J, Sutherland D. R, Allan B. M, Coulson G, Di Stefano J. (2018) Biotelemetry marches on: A cost-effective GPS device for monitoring terrestrial wildlife. PLoS One, 13, e0199617

I also prepared a manuscript during my candidature relating to research conducted prior to my PhD:

Parkins K, Moloney P, Cheers G, MacHunter J. (In prep). Powerful owls in a peri-urban environment: testing habitat suitability models and detectability.

The theses of the two Masters students who worked on collaborative projects are:

Scott A. (2015). Edge effects in fire-prone landscapes: Influence of species traits and resource distribution on ground-dwelling mammal responses to fire edges. Masters thesis. University of Melbourne.

Langmaid K. (2017). Fire severity and vegetation diversity interact to influence range size in the mountain brushtail possum Trichosurus Cunninghami. Masters thesis. University of Melbourne.

Results from this research have been presented at the following conferences:

2015

Australian Mammal Society Conference, Alice Springs, Australia

2016

Ecological Society of Australia conference, Fremantle, Australia Victorian Biodiversity Conference, Melbourne, Australia SEFS Postgraduate student conference, Creswick, Australia

2017

Ecological Society of Australia conference, Hunter Valley, Australia

viii

STUDYING FIRE EDGES

The initial design for my thesis was for four separate studies (chapters) to address knowledge gaps associated with faunal responses to fire edges. The purpose of this section is to briefly outline some of the difficulties that arose during my research, resulting in only two of the four data chapters having a fire-edge focus.

Edges in modified systems are known to influence the movement of many species. I was interested to know if fire edges also affected animal movements. I aimed to design and build my own GPS collars to deploy on mountain brushtail possums, Trichosurus cunninghami. I designed a before-after-control-impact (BACI) study using prescribed fire as an experimental treatment, to examine how fire edges affect the movement of a semi-arboreal mammal. However, during a pilot study I discovered that the study species was very difficult to trap in the areas designated for burning, and therefore very unlikely to be caught in sufficient numbers in the limited number of edge locations identified as part of the initial BACI design. Due to the high levels of risk associated with this study, I ended up abandoning this project.

As part of a collaborative project undertaken during my candidature with a fellow PhD student, we customised off-the-shelf GPS devices and constructed them into GPS collars suitable for wallabies. This research was the precursor for converting the wallaby collars into three- dimensional trackers for mountain brushtail possums. GPS trackers traditionally provide two- dimensional movement data. However, collection of three-dimensional movement data will be important for many semi-arboreal species. In designing my own GPS collars for possums, I attempted to include a high sensitivity altimeter to record changes in air pressure, which could be converted into three-dimensional movement. However, in practice this proved very time consuming. This meant that when I encountered the problem identified above (coupled with tight time constraints associated with PhD candidature) I was unable to develop and implement another edge-related project.

The alternative was to explore three-dimensional movement patterns and resource selection of the mountain brushtail possum in an area burnt by high severity wildfire. Unfortunately, the three-dimensional data collected during this study had some unexplained errors, so I decided to exclude this from the thesis. Some preliminary examples of successful three-dimensional data collected during this study is provided in Appendix C.

Therefore, this thesis has two chapters which investigate fire edges (Chapters 2 & 3). The other two chapters are broadly relevant to ecologists. Given the majority of my focus over the last four years has been on edge effects, the general introduction and synthesis chapters (Chapters 1 & 6) are largely fire-edge focused.

ACKNOWLEDGMENTS

There are many people to thank for their generous support of this project over the last four years.

Firstly - my wonderful supervisors Alan York and Julian Di Stefano. Thank you for your expert guidance through this adventure. Thank you for the time, energy and dedication you put into this project. You have inspired and encouraged me from day one and I have learnt more in the last four years that I could have ever predicted. Alan - thank you for your experienced insights, your unwavering support and for reminding me to think about the bigger picture. Julian – thank you for the hours spent in the field with me, for your patience helping me with data analysis and for always being available to answer my questions.

Trent and Sandra Penman - you opened your home, your pantry (and your whisky cabinet) to me from my very first week as a new PhD student. Your generosity and warmth have gotten me through many of the ups and downs over the last few years. Trent- your door was always open to me, you helped me with every query and you never laughed at my silly analysis or coding questions (of which there were many). Thank you.

Bronwyn Hradsky – thank you for all your support as a friend and as a colleague. You reminded me to see the humour and beauty in the work we do, and inspired me to keep on keeping on when the possums were few and far between. Thank you for all your guidance and suggestions with study designs, GPS data wrangling, analysis and graphing.

Amy Scott and Kirsten Langmaid – Two masters students who collaborated on parts of this research. Thank you for your assistance with field work (I’m sorry about all the leeches). You made the long days and early starts in the field much more enjoyable.

The Fire Ecology and Biodiversity Group- especially Matthew Swan and Holly Sitters - from field work to lab group parties you guys have always been happy to help out. Thank you also to my wonderful friends and colleagues: particularly Manuela Fischer, Matthew Chick, Julio Najera, Garry Cheers, Melissa Huggins, Annalie Dorph, Sarah McColl-Gausden, Nadeeshani Karannagoda and Bess Hartcher. Thank you to Ben Vasic, John Sharp, Nick Lyons and Nathan Manders who volunteered to assist with fieldwork.

My family (Mum, Dad, Ben, Amelia, George and Arlo) - thank you for your loving encouragement, guidance, reassurance, wisdom and emotional (and financial) support over the

xi last four years. Thank you for making me laugh and reminding me that there is a big and fascinating world out there. Thank you for always telling me I could do anything I wanted to, if I just put my mind to it. Dad - thanks for enduring the coldest and wettest week of field work, and George - thank you for being the happiest and most enthusiastic field work volunteer I could have asked for. Ben - thank you for sticking with me through this crazy adventure. Thank you for riding the highs and lows with me, for always having my back, for understanding my long absences with field work and for your unwavering love and support.

Funding for this research was generously provided by the Department of Environment, Land, Water and Planning, the Holsworth Wildlife Research Endowment and the University of Melbourne.

Kean Maizels, Blake Allan and Tristam Horn provided technical advice, assistance and guidance with the development and construction of GPS collars.

Finally, thank you Gillie Courtice and Leighton Llewellyn. I owe much of my love for stomping through the bush looking for critters to you both. You encouraged my love of nature from an early age, and for this I will be forever grateful.

xii

TABLE OF CONTENTS

ABSTRACT ...... III

DECLARATION ...... V

PREFACE ...... VI

STUDYING FIRE EDGES ...... IX

ACKNOWLEDGMENTS ...... XI

TABLE OF CONTENTS ...... XIII

LIST OF TABLES ...... XV

LIST OF FIGURES ...... XVII

1 GENERAL INTRODUCTION ...... 1

1.1 Edges in fire-prone landscapes...... 1

1.2 Research aims ...... 3

1.3 Thesis structure...... 4

2 EDGE EFFECTS IN FIRE-PRONE LANDSCAPES: ECOLOGICAL IMPORTANCE AND IMPLICATIONS FOR FAUNA ...... 7

2.1 Abstract ...... 7

2.2 Introduction ...... 8

2.3 Fire as an agent of edge creation ...... 9

2.4 Edge effects in fire-prone landscapes ...... 11

2.5 Predicting edge effects in flammable landscapes ...... 15

2.6 Conceptual model ...... 16

2.7 Future research ...... 21

3 HABITAT USE AT FIRE EDGES: DO ANIMALS FOLLOW TEMPORAL PATTERNS OF HABITAT CHANGE? ...... 24

xiii

3.1 Abstract ...... 24

3.2 Introduction ...... 25

3.3 Methods ...... 26

3.4 Results ...... 34

3.5 Discussion ...... 43

3.6 Conclusions ...... 48

4 INCREASING DETECTABILITY IN CAMERA TRAP SURVEYS: MORE CAMERAS, MORE DAYS, OR BOTH? ...... 50

4.1 Abstract ...... 50

4.2 Introduction ...... 51

4.3 Methods ...... 52

4.4 Results ...... 55

4.5 Discussion ...... 59

4.6 Conclusions ...... 62

5 THE DEVIL IS IN THE DETAIL: SEX AND FOREST TYPE INFLUENCE POST-FIRE RESOURCE SELECTION IN A SEMI-ARBOREAL MAMMAL .. 64

5.1 Abstract ...... 64

5.2 Introduction ...... 65

5.3 Methods ...... 66

5.4 Results ...... 75

5.5 Discussion ...... 82

5.6 Conclusions ...... 85

6 SYNTHESIS ...... 87

7 REFERENCE LIST ...... 94

8 APPENDICES ...... 117

xiv

LIST OF TABLES Table 3.1 Relationships between time since fire (TSF) and edge position (burnt/unburnt), and six habitat structure variables. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size), Akaike weight is the likelihood of the model being the best in the set; R2 values are the proportion of variance explained by the models. Models within 2 AICc units of the best model and the null models for each variable are presented. See Appendix A-2 for the complete set of models with estimates and 95 percent confidence intervals...... 36

Table 3.2 Relationships between time since fire (TSF) and edge position (burnt, unburnt), and native mammal species. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size) between the model and the best model, Akaike weight is the likelihood of the model being the best in the set; R2 is the proportion of variance explained by the model. We have presented models within 2 AICc units of the best model and the null models for each species. See Appendix A-3 for the complete set of models with estimates and 95 percent confidence intervals. We were unable to calculate R2 for agile antechinus using the mixed effects model, the single value for this species represents R2 from a generalised linear model after removing the random effect...... 39

Table 3.3 Relationships between time since fire (TSF) and edge position (burnt, unburnt), and non-native species. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size) between the model and the best model, Akaike weight is the likelihood of the model being the best in the set; R2 is the proportion of variance explained by the model. We have presented models within 2 AICc units of the best model and the null models for each species. . 41

Table 5.1 Burn severity categories...... 67

Table 5.2 Vegetation survey methods...... 71

xv associated with males represent contrasts with females. Akaike's information criterion (AICc) was used to rank models. Models within two units of the top-ranked model are shown with Akaike weights. Parameter estimates with 95 percent confidence intervals (CI) are displayed. Two measures of fit are included: marginal R2 (R2m) is the variance explained by fixed factors and conditional R2 (R2c) is the variance explained by both fixed and random factors...... 78

xvi

LIST OF FIGURES Figure 1.1 A fire edge between burnt and unburnt vegetation following a prescribed fire in the central highlands of Victoria, Australia. This image was taken five days post-burn...... 3

Figure 2.1 Different types of fire edges. A) Edges between burnt/unburnt, B) edges between differing fire severities (i.e. unburnt, moderate burn and high severity burn), C) edges at moisture gradients where the fuel becomes less flammable. Note- in image C) the edge-zone for this fire can be clearly seen, however there are several trees within the unburnt section that retain fire scars from a previous fire. These fire scars illustrate the temporal and spatial variability in edge locations (Images: A & B- DEWLP 2015; C-Parkins, 2015)...... 9

Figure 2.2 A conceptual model of the factors driving edge effects in fire-prone landscapes. The model considers the origin of edge creation, including biophysical factors and elements of the disturbance regime. Interactions between these factors influence edge architecture (edge size, shape and contrast), which influences edge dynamics (such as site permeability and landscape connectivity). Species traits such as the strength of habitat associations, diet specificity and mobility will also contribute to the dynamics occurring at fire edges. The unique interaction of all of these variables will influence how individual animals, species or communities respond to fire edges. The direction of arrows indicates the direction of influence...... 23

Figure 3.1 Location of the control and treatment sites in south-eastern Australia ...... 29

Figure 3.2 A fire edge between burnt and unburnt vegetation resulting from a prescribed burn in the Central Highlands of Victoria. This image was taken two weeks post-burn (Image: Parkins 2015)...... 31

Figure 3.3 A) edge locations in a treatment block burnt in 2013 B) Locations of each Elliott and camera trap, and vegetation survey in relation to an edge...... 32

Figure 3.4 Understorey complexity over time on burnt (grey) and unburnt (white) sides of fire edges, with 95 percent confidence intervals. Data represent model predictions from the TSF × edge interaction. Multiple sites were established in each TSF category (TSF 0 n=6, TSF 1-2 n=10, TSF 3 n=6, TSF 6-7 n=4, TSF 76 n=10)...... 35

xvii

Figure 3.5 Individual habitat variables and change over time on the burnt (grey) and unburnt (white) side of fire edges, with 95 percent confidence intervals. Data represent model predictions from the TSF × edge interaction...... 37

Figure 3.6 Native mammal activity over time at burnt (grey) and unburnt (white) sides of fire edges, with 95 percent confidence intervals. Data represent model predictions from the TSF × edge interaction. Activity Index is the proportion of days detected (number of days detected/number of total days per trapping method)...... 40

Figure 3.7 Invasive species activity over time at burnt (grey) and unburnt (white) sides of fire edges, with 95 percent confidence intervals. Data represent model predictions from the TSF × edge interaction. Activity Index is the proportion of days detected (number of days detected/number of total days per trapping method)...... 41

Figure 3.8 Selection for dense understorey vegetation (relativised electivity index) for native species at burnt (black) and unburnt (grey) edges, with 95 percent confidence intervals shown. TSF 76 represents an average of the burnt and unburnt (as there is no true fire edge at this time step) ...... 42

Figure 3.9 Selection for dense understorey vegetation (relativised electivity index) for invasive predator species at burnt (black) and unburnt (grey) edges, with 95 percent confidence intervals shown. TSF 76 represents an average of the burnt and unburnt (as there is no true fire edge at this time step)...... 43

Figure 4.1 Location of the ten sites established in the Powelltown-Noojee region of south- eastern Victoria, Australia. Six detection units (12 cameras) were established per site, along a 60 metre transect...... 54

Figure 4.2 Relative species richness for each number of camera pairs (1-6) deployed across the sample period (34 days) from the 1000 iterations. The dashed line represents the 17 days of real data. The lines represent mean values per camera pair, and the dotted lines represent 95 percent confidence intervals...... 56

Figure 4.3 Probability of detecting species at 80, 90 or 95 percent confidence, across 1-6 detection units. Lines represent the cumulative number of species for each number of camera pairs against the number of survey days, based on detectability and assuming all animals are present at a site. We have excluded the errors from these graphs to aid interpretation...... 57

xviii

Figure 4.4 Detection curves for the more commonly detected species in our study region. The number of days required to be 80 (red), 90 (yellow) and 95 (blue) percent confident that the species has been detected at a site if present, across the number of camera pairs deployed per site, with 95 percent confidence intervals...... 58

Figure 4.5 Detection curves for less commonly detected species in our study region. The number of days required to be 80 (red), 90 (yellow) and 95 (blue) percent confident that the species has been detected at a site if present, across the number of detection units deployed per site. Note- the values for agile antechinus and echidna from 1-3 detection units were very large numbers with high uncertainty, and have therefore been left off the graphs to aid interpretation...... 59

Figure 5.1 Area burnt by the 2009 wildfire north east of Melbourne, Australia, stratified by fire severity category. Possums were successfully trapped at three locations within the burn perimeter...... 68

Figure 5.2 Custom-made GPS circuit board, b) collar with GPS and VHF, c) Mountain brushtail possum with collar...... 69

Figure 5.3 Sampling protocol for each of the 825 × 15 m vegetation sampling locations...... 70

Figure 5.4 Decay class of large trees (>50 cm diameter at breast height). Tree form 1 - live healthy tree, no signs of decay; tree form 2 - live tree with slight decay (typically older trees beginning to senesce); tree form 3 - dead tree in the early stages of decay; tree form 4 - dead tree in the mid-stages of decay; and tree form 5 - highly decayed dead tree. Image modified from Banks et al. (2011b)...... 73

Figure 5.5 Interpolation map of wattle basal area (m2 / ha) at a dry forest site. Points indicate locations where vegetation surveys were conducted. Values were interpolated between points using inverse distance weighting...... 74

Figure 5.6 Bi-plot generated from an ordination procedure, demonstrating the difference between wet and dry forest types...... 76

Figure 5.7 Model predictions for the interaction between forest type and each habitat variable, by sex. Blue lines represent wet forest, orange lines represent dry forest and shaded areas are 95 percent confidence intervals...... 81

xix

Figure 5.8 Model predictions for midstorey connectivity by a) forest type and b) sex. Shaded areas are 95 percent confidence intervals...... 82

1 GENERAL INTRODUCTION

Fire is a major source of natural disturbance and a key driver of biome distribution and community composition worldwide (Pastro, Dickman & Letnic 2014). Integral to the functioning of some ecosystems, fire can help to maintain the diversity of many vegetation communities by promoting the regeneration of fire-responsive species (Auld & O'Connell 1991; Bell 1999). However, fire can also result in shifts in ecosystem states (Lindenmayer et al. 2011) and threaten the persistence of many flora and fauna species (Torre & Díaz 2004; Banks et al. 2011b). Future climate change scenarios predict an increase in wildfire activity, characterised by increases in fire extent, severity and frequency in many forested ecosystems (Bowman et al. 2009; Flannigan et al. 2009). Human intervention is changing aspects of the fire regime through fire exclusion and the purposeful application of prescribed fire (Bowman et al. 2011; Penman et al. 2011), altering the spatial and temporal occurrence of fire in ways that may affect biodiversity and ecosystem function (Bradstock, Keith & Auld 1995). With the increased prevalence of both planned and unplanned fire in many of the worlds’ terrestrial systems, a greater understanding of how fire is likely to affect ecological patterns and processes is of critical importance. However, predicting ecological responses to fire remains difficult. A better understanding of the mechanisms that underlie species responses to fire, interactions between fire and other processes that alter species distribution or population dynamics, and the influence of spatial and environmental context on fire responses will facilitate better management of flammable landscapes for biodiversity conservation (Driscoll et al. 2010; Griffiths & Brook 2014; Arnold, Murphy & Gibb 2017).

1.1 EDGES IN FIRE-PRONE LANDSCAPES In forested systems, interactions between biophysical properties and components of the disturbance regime result in complex configurations of habitat burnt at different severities and across different spatial and temporal scales. Variable terrain, differing vegetation types and weather conditions can influence fire behaviour and spread (Leonard, Bennett & Clarke 2014). Accordingly, fire events are often patchy, with different parts of the landscape being burnt at high, moderate or low severity, or remaining unburnt. Even large-scale, high-severity fires can produce a spatially diverse mosaic of different burn intensities (Bradstock 2008; Arthur, Catling & Reid 2012), and prescribed fire in relatively homogenous landscapes produce patchy outcomes (Penman et al. 2007). Different taxa exhibit different responses to fire, and it is assumed that fire- moderated patchiness and variability in a landscape will provide a range of habitats through time

1 and space that will benefit biota (Brockett, Biggs & Van Wilgen 2001; Burrows & Wardell- Johnson 2004; Bradstock et al. 2005).

The promotion of the patch mosaic burning paradigm in fire-management, where fire is used to create a mosaic of patches representative of a range of fire histories (Parr & Andersen 2006), coupled with the natural variability of fire in forested landscapes has resulted in a very patch- focused view of flammable landscapes. For example, many studies have compared the occurrence or abundance of fauna in burnt or unburnt patches (Banks et al. 2011a; Pastro, Dickman & Letnic 2014; Chia et al. 2015; Swan et al. 2016), patches of differing fire severity (Diffendorfer et al. 2012; Doumas & Koprowski 2013; Lindenmayer et al. 2013; Robinson et al. 2014; Chia et al. 2015; Berry et al. 2017), or in patches of different fire age classes (Catling, Coops & Burt 2001; Van der Ree & Loyn 2002; Smucker, Hutto & Steele 2005; Covert-Bratland, Block & Theimer 2006; Haslem et al. 2012). While much has been gained from this approach, classification of fire-prone landscapes into discrete patches can have important implications for how we view and manage flammable ecosystems. In particular, the ecological importance of the interface between patches in fire-prone landscapes remains poorly understood.

Edges have been extensively researched in agricultural and urban landscapes (e.g. Murcia 1995; Driscoll & Donovan 2004; Villasenor et al. 2015) and much of our understanding about how edges affect animals comes from these highly modified ecosystems. Considerably less is known about edge effects in natural systems where disturbances such as fire are agents of edge creation. A fire edge is defined as a boundary, interface or transition zone generated by fire, between areas of differing structural characteristics (Figure 1.1). Fire has the potential to generate multiple, often overlapping edges in space and time, through individual fire events and the broader fire regime. Fire edges can result from wildfire, prescribed burning, or from fire management processes (i.e. construction of fuel breaks). Fire edges can form at both small scales (e.g. intra- burn edges from a single fire) and at large scales (e.g. inter-burn edges between different fire events) and can influence habitat connectivity at both the site and landscape scale.

Several conceptual frameworks have been established to predict species response to edges in modified systems (Murcia 1995; Cadenasso et al. 2003b; Ries et al. 2004; Ries & Sisk 2004). However, these models are expected to perform poorly in the context of fire edges for several reasons. Fire edges are temporally dynamic, spatially complex and are characterised by the strength of the interaction between components of the disturbance regime and other biophysical factors. Unlike edges in modified landscapes, that are often maintained in a relatively stable state,

2 we expect fire edges to change over time as the vegetation regenerates, with the strength of fire- induced edge effects predicted to reduce over time as the burnt-edge regenerates. With the increasing prevalence of both planned and unplanned fire events worldwide, there is a need to increase our understanding of faunal responses to fire. A focus on edge effects in flammable landscapes will be an important step in improving our understanding of the processes driving species fire responses.

Figure 1.1 A fire edge between burnt and unburnt vegetation following a prescribed fire in the central highlands of Victoria, Australia. This image was taken five days post-burn.

1.2 RESEARCH AIMS The overarching aim of this thesis was to investigate how fire edges influence habitat use and resource selection for fauna in fire affected landscapes. This research is important for several reasons. Firstly, edges are pervasive, ubiquitous environmental features that have the potential to influence the distribution of fauna over broad temporal and spatial scales. Secondly, edge effects have been poorly studied in flammable landscapes, and their associated effects on biota are largely unknown. Finally, under climate change predictions wildfires are expected to increase in extent, frequency and severity (Cary et al. 2012), and management agencies are increasing the use of prescribed fire to combat this (Penman et al. 2011). Therefore, understanding the role of edges

3 in flammable systems and species resource selection in fire-affected landscapes will better enable managers to integrate biodiversity conservation into fire management planning.

A fire-induced change in structure may represent a disruption to the continuity of resources for some species and fire edges are therefore expected to influence animal activity patterns, particularly for species reliant on vegetation for cover, foraging or nesting. The ground-dwelling mammal fauna of south-eastern Australian forest were chosen as a focal group for this research, as many species are reliant on understorey habitat and are expected to be affected by the creation of fire edges in the landscape. They also perform a range of important ecological functions including seed and fungal spore dispersal, pollination, nutrient cycling and providing prey for other vertebrates (Dickman & Steeves 2004).

1.3 THESIS STRUCTURE To facilitate a ‘chapters as papers’ structure, this thesis does not contain an initial chapter outlining the study site, study species and common methodologies. New methods and relevant aspects of the study region are described sequentially in each chapter. This results in some degree of overlap, particularly regarding the description of the study region. In addition, several chapters have been written in preparation for publication and the pronoun ‘we’ is used instead of ‘I’ to recognise contributions from co-authors (see Table page vii).

The focus of each chapter is as follows:

Chapter 2: Edge effects in fire-prone landscapes: ecological importance and implications for fauna.

In this chapter I review the literature on fire, fauna, and edge effects to summarise current knowledge of faunal responses to fire edges, and identify knowledge gaps. I developed a conceptual model to predict faunal responses to fire edges, combining several drivers of faunal- fire responses, and outline key areas for further research.

This chapter has been published in Ecology and Evolution.

Parkins, K., York, A. & Di Stefano, J. (2018) Edge effects in fire‐prone landscapes: Ecological importance and implications for fauna. Ecol Evol, 00:1-12. http://doi.org/10.1002/ece3.4076

Chapter 3: Habitat use at fire edges: Do animals follow temporal patterns of habitat change?

In response to the knowledge gaps outlined in Chapter 2, this chapter focuses on quantifying how fire edges influence the activity of ground-dwelling mammals and how these patterns change over time as the burnt area regenerates. I used a space-for-time substitution sampling design, selecting a series of burnt/unburnt edges in areas treated by prescribed fire between 0-7 years previously. Treatment sites were compared to a series of long unburnt control locations. I used remote wildlife cameras and live mammal trapping to detect species. I used a resource selection index to assess the congruence between patterns in animal activity and understorey complexity over time, on both sides of fire edges.

The key predictions were

1. Species with behaviours strongly linked to ground-level vegetation complexity would reduce their activity at fire edges until the vegetation had regenerated to pre-burn levels. 2. Species reliant on complex understorey vegetation would follow the patterns of temporal change observed in the regenerating vegetation.

Chapter 4: Increasing detectability in camera trap surveys: more cameras, more days, or both?

As part of collecting data to answer several questions about edge effects in fire-prone landscapes, we established a series of control locations (Chapter 3). This provided an opportunity to develop a methodological chapter assessing the trade-off between deploying more cameras and extending the length of the sampling period on two frequently assessed response variables in camera trapping studies – species richness and detection probability.

The key predictions were:

1. Survey length could be considerably reduced by increasing the number of cameras deployed per site. 2. Increasing the number of cameras (rather than deployment time) would be an effective method for increasing species detectability in camera trapping studies.

Chapter 5: The devil is in the detail: sex and forest type influence post-fire resource selection in a semi- arboreal mammal

In this chapter I investigated resource selection of a semi-arboreal mammal (mountain brushtail possum) in burnt forest eight years after a major wildfire. I used GPS telemetry to collect movement data from the study species in two different forest types (wet and dry) that had been

5 severely burnt during the 2009 Black Saturday wildfire in south-eastern Victoria. I analysed data for males and females and tested if resource selection was influenced by both forest type and sex.

The key prediction was:

1. Post-fire resource selection would depend on both forest type and sex

Chapter 6: Synthesis

The aim of this chapter was to summarise the key results, describe links between the other chapters where possible, place the research in a broader context and discuss implications for management.

2 EDGE EFFECTS IN FIRE-PRONE LANDSCAPES: ECOLOGICAL IMPORTANCE AND IMPLICATIONS FOR FAUNA

2.1 ABSTRACT Edges are ecologically important environmental features and have been well researched in agricultural and urban landscapes. However, little work has been conducted in flammable ecosystems where spatially and temporally dynamic fire edges are expected to influence important processes, such as recolonisation of burnt areas and landscape connectivity. We review the literature on fire, fauna, and edge effects to summarise current knowledge of faunal responses to fire edges and identify knowledge gaps. We then develop a conceptual model to predict faunal responses to fire edges and present an agenda for future research. Faunal abundance at fire edges changes over time, but patterns depend on species traits and resource availability. Responses are also influenced by edge architecture (e.g., size and shape), site and landscape context, and spatial scale. However, data are limited and the influence of fire edges on both local abundance and regional distributions of fauna is largely unknown. In our conceptual model, biophysical properties interact with the fire regime (e.g., patchiness, frequency) to influence edge architecture. Edge architecture and species traits influence edge permeability, which is linked to important processes such as movement, resource selection, and species interactions. Predicting the effect of fire edges on fauna is challenging, but important for biodiversity conservation in flammable landscapes. Our conceptual model combines several drivers of faunal fire responses (biophysical properties, regime attributes, species traits) and will therefore lead to improved predictions. Future research is needed to understand fire as an agent of edge creation; the spatio-temporal flux of fire edges across landscapes; and the effect of fire edges on faunal movement, resource selection, and biotic interactions. To aid the incorporation of new data into our predictive framework, our model has been designed as a Bayesian Network, a statistical tool capable of analysing complex environmental relationships, dealing with data gaps, and generating testable hypotheses.

2.2 INTRODUCTION Edges, barriers, boundaries or ecotones are ubiquitous environmental phenomena, occurring in a wide range of ecosystems and across multiple spatial scales (Cadenasso et al. 2003a). Edges can occur both within and between landcover types, and refer to the interface or transition zone between areas of differing structural characteristics. Edges can be dynamic or static, and frequently exhibit greater heterogeneity than adjacent areas (Peters et al. 2006; Yarrow & Salthe 2008). Edges are ecologically important because they influence a wide range of patterns and processes (Ries et al. 2004). The resulting ecological changes at edges are collectively known as edge effects.

Edge effects result from both abiotic (e.g. radiation, moisture, temperature) and biotic (e.g. species interactions) sub-processes that interact to generate environments with different structural attributes and species assemblages compared to other parts of the landscape (Murcia 1995; Craig et al. 2015). A key aspect of edges is their capacity to influence the flow of energy and materials. They have been described as ecological analogues to cellular membranes (Harper et al. 2005), being semi-permeable boundaries that allow certain materials or organisms to flow freely while restricting or prohibiting the movement of others (Laurance, Didham & Power 2001). Understanding this aspect of edges is critical as movement processes influence the fate of individuals and the structure and function of ecosystems (Fahrig 2007; Nathan et al. 2008).

Edge effects have been extensively researched in some contexts, particularly in highly modified agricultural and urban landscapes and have been the subject of several reviews (see Murcia 1995; Cadenasso et al. 2003a; Ries et al. 2004; Harper et al. 2005). While significant progress has been made in the study of edges, several aspects remain poorly understood. For example, fire is an agent of edge creation and a globally important driver of biome distribution and community composition (Bond & Keeley 2005; Pastro, Dickman & Letnic 2011), yet little is known about how fire edges affect ecological processes in flammable ecosystems. Here, we outline this knowledge gap, focusing on the potential influence of fire edges on animals. We then describe a conceptual model for predicting the edge response of animals in flammable landscapes and present an agenda for future research.

We define a fire edge as an interface or transition zone generated by fire, resulting in a boundary between areas of differing structural characteristics. This may refer to edges between burnt and

8 unburnt vegetation, between different burn severities, or adjacent areas burnt at different times or frequencies. Several types of fire edges are shown in Figure 2.1.

2.3 FIRE AS AN AGENT OF EDGE CREATION Earth is intrinsically flammable (Bowman et al. 2009), with wildfires predicted to increase in extent and severity as a result of climate change (Flannigan et al. 2009). In response to this prescribed fire is increasingly being used as a management tool globally (Penman et al. 2011; Stephens et al. 2012; Fernandes et al. 2013). Given this likely increase in fire activity, a better understanding of fire edges with respect to their ecological function and implications for animals is important. To date no studies have quantified the impact of fire on architectural properties of edges such as shape and contrast, or how the characteristics of fire edges change in space and time.

Edges are created by fire through the consumption of fuel and the factors that cause fires to extinguish. In forested landscapes there are many factors that interact to influence fire behaviour and spread (Bradstock et al. 2010) and determine the location and architecture of fire edges. Explicit consideration of the factors that cause fires to extinguish will be central to understanding the ecological dynamics occurring at fire edges, as different methods of

9 extinguishment are likely to result in different edge characteristics (e.g. size, shape, contrast), influencing species’ edge response.

FUEL CONTINUITY

Variability in fuel structure and distribution is common in natural systems, and continuity of surface or near-surface fuels is important for the spread of fire in most landscapes (Catchpole 2002). Fires often extinguish in areas where the continuity of plant material is not sufficient to sustain burning, creating burnt/unburnt edges. Fuel discontinuities include areas of topographical change (i.e. cliff lines, rocky outcrops) often characterised by a sharp reduction in the amount of available fuel. They may also be created through the construction of fuel breaks or roads.

FUEL FLAMMABILITY

Gradients of moisture availability govern both the accumulation of fuel and its ability to burn (Bradstock 2010). Topographic locations such as gullies, shaded aspects or riparian zones are commonly more productive than other drier parts of a landscape yet these areas often remain unburnt (or burnt at lower intensities). This is commonly due to higher fuel moisture, and/or because they support intrinsically less flammable vegetation, facilitating edge creation nearby. For example, patches of Chenopod Mallee that are less flammable than surrounding Triodia Mallee in south eastern Australia (Haslem et al. 2011), or patches of evergreen (Afromontane) forest surrounded by highly flammable fynbos shrubland in South Africa (Van Wilgen, Higgins & Bellstedt 1990).

WEATHER

Severe ambient weather conditions (i.e. strong winds, high temperatures, low humidity) contribute to the ease of ignition, rate of spread, pattern of fire intensity, and the location and architecture of fire edges. The area burnt at differing severities (or remaining unburnt) will be strongly influenced by these factors (Bradstock 2010). For example, highly flammable areas may remain unburnt due to a sudden wind change re-directing the fire-front.

MANAGEMENT ACTIONS

Fire edges can also result from human generated discontinuities in fuel caused by prescribed burning, construction of fuel breaks or roads, or from suppression activities during fire. Edges at fuel discontinuities (i.e. fuel breaks), and those resulting from suppression activities are likely to

10 be higher contrast than edges resulting from natural processes, and more similar to edges in agricultural and urban landscapes.

2.4 EDGE EFFECTS IN FIRE-PRONE LANDSCAPES Fire-generated edge effects are likely to differ from other edge types in several ways. They are temporally dynamic, spatially complex and are characterised by the strength of the interaction between components of the disturbance regime and other biophysical factors. While edge effects and faunal-fire responses have been well studied independently, how animals respond to fire edges remains poorly understood. Here we review the current state of knowledge and highlight several aspects of existing edge literature and fire research generally relevant to understanding interactions between animals, fire and edges in flammable landscapes.

We conducted a literature search for papers that specifically investigated faunal responses to fire edges. We focused our search on studies that defined an edge zone or included distance from edge in the analysis. We searched for papers in the Web of Science, using the following search criteria: TOPIC: (“fire edge effects” OR “fire edge” OR “burn edge” OR edge) AND (fire OR wildfire) AND (fauna OR animals), and sorted the results by relevance, not excluding any years. We also included relevant papers cited in reference lists. Here we summarise the key findings.

FIRE EDGES AND EDGE EFFECTS ARE TEMPORALLY DYNAMIC

Edges in modified systems (e.g. between pasture and forest) are often maintained at a relatively stable state, but this is not the case for fire edges. Edges resulting from fire are in a constant state of flux due to (a) post-fire regeneration, and (b) the occurrence of new fires. Fire edges are ephemeral, dynamic parts of fire-affected landscapes, where an edge zone may be impermeable to some species immediately after fire, but highly permeable at a certain point in time post-fire. However, it is not the time per se that is important in determining faunal succession (Monamy & Fox 2000), but the pattern of resource regeneration and re-accumulation.

Some animals, including birds, small mammals and reptiles, avoid or minimise time spent at hard edges (high contrast) in modified landscapes (Goosem 2001; Lehtinen, Ramanamanjato & Raveloarison 2003; Laurance, Goosem & Laurance 2009; Wilson et al. 2010). Although data are lacking, similar responses may occur at fire edges immediately post-fire, where the contrast between the burnt and unburnt side of the edge is often high. Unlike edges in modified landscapes, we would expect these effects to reduce over time as vegetation regenerates and the fire-edge softens. Some small mammal species avoid fire edges for 4-5 years post-fire, until shrub

11 cover and seed production reached sufficient levels to sustain viable populations in these areas (Borchert & Borchert 2013). In contrast, reduced understory cover post-fire can enable predators to hunt more effectively (Conner, Castleberry & Derrick 2011; Leahy et al. 2016; Hradsky et al. 2017), resulting in the increased prevalence of some predators at fire edges briefly post-fire, while the contrast between burnt/unburnt remains high.

The temporal extent of a fire edge effect will also be influenced by species specific factors. Early successional post-fire habitats are characterised by vigorous growth of understorey vegetation and are often dominated by ground-foraging herbivorous mammal species. As the cover of woody plants increases over time granivorous species and those that feed on invertebrates move into the regenerating landscape (Torre & Díaz 2004). A study in chaparral shrublands (USA) at a high-contrast burnt/unburnt edge along the perimeter of a wildfire found edge response differed between species with differing resource requirements. A habitat generalist (pinyon mouse, Peromyscus truei) was found to occupy burnt edge and unburnt habitat in similar abundance within one year of a high intensity wildfire. In contrast, a late seral specialist (Californian mouse, Peromyscus californicus) was more prevalent in unburnt habitat, took 4-5 years to occupy the edge zone and was only detected in high abundance in burnt vegetation nine years after fire (Borchert & Borchert 2013). Burnt chaparral produces an abundance of seeds immediately after fire, likely driving the high occupancy of the burnt site by granivorous species, or those with generalist habitat and/or diet requirements. However, unburnt chaparral provides better protection from predators, especially early post-fire. This asymmetry in food and habitat resources across burnt/unburnt edges creates conditions in which some species may increase their use of fire edges (Sitters et al. 2014), with access to abundant food on the burnt side and protection from predators on the unburnt side.

FIRE EDGES AND EDGE EFFECTS ARE SPATIALLY VARIABLE

Fire edges form at multiple spatial scales, occurring as external perimeters to a single fire event (external burn edge), within the boundary of a single fire event (internal burn edges), or as temporally overlapping edges between multiple fires. Fire events are often patchy, with different parts of the landscape being burnt at high, moderate or low severity, or remaining unburnt. Large-scale, high-severity fires can produce a spatially diverse mosaic of different burn intensities (Bradstock 2008; Arthur, Catling & Reid 2012), and prescribed fire in relatively homogenous landscapes produce patchy outcomes (Penman et al. 2007). Many fire-management strategies have been designed to deliberately increase variability through the use of dynamic fire mosaics across space and time (Bradstock et al. 2005; Parr & Andersen 2006).

Edge effects resulting from fires are expected to be shorter lived than edges in highly modified landscapes due to the dynamic nature of post-fire regeneration. However, repeated, unpredictable disturbances like fire can produce a mosaic of patches at different successional stages, resulting in multiple, overlapping edges that may substantially influence species distributions and community structure (Wiens 1976).

In modified landscapes the magnitude and extent of edge effects increase at locations where several edges are present (Fletcher 2005). In fire-prone landscapes, multiple fires result in an intricate network of fire edges, each with a unique trajectory of temporal change. How an animal responds to a fire edge may be a function of the edge itself, and/or the context of that edge within the broader landscape. For example, northern spotted owls (Strix occidentalis caurina) were attracted to hard edges caused by severe fire and salvage logging when these were small patches within a larger area burnt at low severity (Comfort et al. 2016). However, for this species the influence of both hard and soft edges on habitat selection depended on spatial scale. With the exception of attraction to hard edges at a small spatial scale, northern spotted owls were generally attracted to soft edges and avoided hard edges (Comfort et al. 2016). Soft edges were characterised by an intact canopy and regenerating shrub layer, resulting in structurally complex forest promoting prey availability and facilitating hunting under a closed canopy. Hard edges were characterised by high severity burnt patches, some of which had been salvage logged, adjacent to forest burnt at low severity. These areas likely supported fewer prey, and resulted in a sub-optimal hunting environment.

Studies at edges in fragmented landscapes have shown species to be more strongly associated with local habitat resources rather than edge structure (Schultz, Franco & Crone 2012; Villasenor et al. 2015), and this is likely to be similar for many species at fire edges. Small mammal abundance both close to and distant from a fire edge was influenced by site specific factors such as the presence of riparian zones, topography and shrub species composition (Diffendorfer et al. 2012). It is therefore important to consider fire edges within the context of other edges in the surrounding landscape (Cochrane & Laurance 2002).

RESPONSES TO FIRE EDGES ARE SPECIES SPECIFIC

Species-specific traits and resource requirements are major drivers of post-fire recovery and recolonisation, and are likely to play a key role in how certain species respond to fire edges. Species with preferences for disturbed and/or open conditions dominate recently burnt sites,

13 whereas species with preference for older vegetation are more prevalent at unburnt sites (Diffendorfer et al. 2012).

Small species with limited mobility may find edges difficult to cross until the vegetal components vital to their survival regenerate to sufficient levels (Santos, Bros & Mino 2009). Fire-induced edge effects can last for several years for species strongly associated with microhabitat structure. For example, burnt sites were characterised by reduced gastropod species richness for up to four years after fire (Santos, Bros & Mino 2009), and reduced beetle community composition for up to five years (Elia et al. 2016). In contrast, larger, highly mobile animals such as the Canada lynx (Lynx canadensis) use fire edges during the first year post-fire because of the contrasting vegetation characteristics provided by the burnt/unburnt edge (Vanbianchi, Murphy & Hodges 2017). Similarly, Californian spotted owls (Strix occidentalis occidentalis) have a higher probability of selecting fire edges than contiguous habitat, with survival and reproductive rates higher in areas containing edge habitat (Eyes, Roberts & Johnson 2017).

In situ survival and ex situ colonisation may drive post-fire recovery and edge response. Distance from fire edge had little or no effect on small mammal abundance in either chaparral shrubland or tall wet forest (Banks et al. 2011a; Diffendorfer et al. 2012), nor did it affect species richness or abundance of several cockroach species in foothills forest (Arnold, Murphy & Gibb 2017). However, post-fire recovery of some litter detritivore species were limited by distance from burn edge, with this variable an important determinant of post-fire assemblages up to six years after fire (Arnold, Murphy & Gibb 2017). Species richness of birds in semi-arid Mallee shrublands was also higher at sites closer to unburnt vegetation and at sites containing unburnt patches, suggesting that colonisation from ex situ populations was an important process for the recovery of avifauna post-fire in this system (Watson et al. 2012). These studies suggest that in some cases surviving individuals are driving population recovery in burnt areas while in others distance- from-edge, edge permeability and subsequent recolonisation from unburnt areas is important.

Responses to fire edges are likely to be species, not taxa, specific. For example, bark-probing woodpecker species such as black-backed woodpeckers (Picoides arcticus) and hairy woodpeckers (Picoides villosus) are attracted to recently burned areas because of increases in wood-boring beetles (Vierling, Lentile & Nielsen-Pincus 2008), with higher reproductive success at edges than deep in burnt forest (Nappi & Drapeau 2009). In contrast, other species of woodpeckers preferentially nest further from the edge of burnt patches as a predator avoidance strategy (Vierling, Lentile & Nielsen-Pincus 2008).

RESPONSES TO FIRE EDGES INVOLVE COMPLEX INTERACTIONS

Edges created by fire never occur in isolation from other environmental patterns and processes. Accordingly, fire edges are inherently complex due to these interactions with other variables. Edge effects may be exacerbated, diminished or masked by the interaction of other factors, making it difficult to determine if animals are responding independently to a fire edge effect or to other processes. Much of the existing fire edge data has been collected at pre-existing edges (e.g. between forest and highly modified agricultural land) that were subsequently affected by fire (see Figueiredo & dos Santos Fernandez 2004; Pires et al. 2005; Mendes-Oliveira et al. 2012). In these studies fire-induced edge effects were confounded by the presence of hard, anthropogenically modified edges, making it difficult to determine whether animals were responding to pre-existing forest/farmland edges or burnt/unburnt edges. However, for some species it may be this unique interaction between fire and other processes that enables them to utilise edge habitat. For example, the location of greater prairie-chicken (Tympanuchus cupido) lek sites (areas for communal courtship display) were influenced by an interaction between patch edge and fire, with leks positioned near patch edges at recently burnt sites (Hovick et al. 2015). Furthermore, processes such as pyric herbivory may interact with fire edge effects to influence the spatio-temporal dynamics of fire edges. Herbivores are often attracted to the flush of vegetation growth post-fire because of enhanced forage quality and increased productivity in burnt areas (Wilsey 1996; Eby et al. 2014). However, post-fire grazing may intensify or prolong fire-induced edge effects, changing the nature of species’ interactions and influencing species’ responses to these edges.

2.5 PREDICTING EDGE EFFECTS IN FLAMMABLE LANDSCAPES Current models predicting the response of fauna to edges are predominantly based on the distribution and quality of resources in adjoining habitats (see Cadenasso et al. 2003b; Ries et al. 2004). While these models have proven effective in predicting edge effects across a range of environments, they are unlikely to perform well in flammable ecosystems for two reasons.

Firstly, these models do not explicitly consider the process of edge creation and we argue that edge effects cannot be effectively understood in isolation from the processes that generate them. Understanding the process of creation is of particular importance for edges created by fire where temporal changes in edge architecture are expected to be influenced by fire regime attributes and both static (e.g. topography) and dynamic (e.g. climate) biophysical properties. Fire affects both the horizontal and vertical composition of vegetation through flame height and radiant heat, resulting in three dimensional edge effects (depth, length and height). Understanding how a disturbance interacts with biophysical factors to influence the physical characteristics of an edge

15 will be essential for predicting edge permeability, species’ edge response and how edge effects might change in time and space.

Secondly, while current predictive models suggest that edge response will differ between mobile and sessile organisms (Ries et al. 2004), we argue that consideration of traits that go beyond mobility will be important predictors of species’ edge response in flammable systems. Faunal traits influence species susceptibility to environmental change and predispose some species to decline at a greater rate than others in the face of adverse environmental shifts (Pimm, Jones & Diamond 1988; Webb et al. 2010). Quantifying interactions between traits and extrinsic factors can improve the capacity to predict species responses to threatening processes (Murray et al. 2011). Understanding how traits interact with edge characteristics in a changing landscape, and how this influences movement, biotic interactions and access to resources for different species may enhance the predictive capacity of edge effects models.

2.6 CONCEPTUAL MODEL In response to the issues identified above, we have developed a conceptual model (Figure 2.2) for predicting how fire is likely to shape the physical properties of an edge and influence species’ edge responses. While this model was designed to predict the responses of fauna to fire edges, it could also be applied to other disturbance contexts, as components of the disturbance regime can be modified to suit any edge creation process that results in a changed landscape (i.e. naturally or anthropogenically occurring). For the purposes of this paper we discuss the components of the model and their interactions using fire as the disturbance process.

BIOPHYSICAL PROPERTIES

In fire-prone landscapes the location of unburnt patches often occurs in a non-random manner, largely due to variations in topography, climate and soils, and their influence on fire behaviour. Topographic locations with higher fuel moisture may experience lower fire severity and intensity, and have a lower probability of burning than adjacent drier topographies (Penman et al. 2007; Bradstock 2010; Collins et al. 2012). Topography is therefore an important biophysical feature influencing where fire edges occur.

At large spatial scales, long-term climatic fluctuations are correlated with the probability of fire ignition and spread, whereas at a smaller spatial scales local weather conditions (particularly temperature and wind speed) can influence fire behaviour (Alexander et al. 2006), and therefore the position and physical characteristics of fire edges. Post-fire rainfall contributes to the rate of

16 vegetation recovery, which will determine temporal changes in edge contrast. Further, variations in soil moisture and nutrient levels contribute to heterogeneous patterns of vegetation, that in turn influence edge location and architecture.

DISTURBANCE REGIME

Fire regimes incorporate the effects of discrete fire events with the cumulative effect of multiple fires over time, and are characterised by spatially variable patterns in fire type, severity, spatial extent, patchiness, frequency and seasonality (Gill 1975). Fire regimes generate a spatially and temporally shifting pattern of patches (Parr & Andersen 2006) and their effect on animals is usually considered in this context (Taylor et al. 2012; Griffiths, Garnett & Brook 2015; Kelly et al. 2015). However, patches have edges, and edge characteristics are likely to influence both fine- scale movements and the broader distribution of many species.

Fire extent refers to the overall size of a fire and is predominately determined by biophysical properties. Extent is correlated with perimeter length, which defines an important component of fire-related edge habitat.

Fire severity is principally influenced by weather, but also by topography and fuel load (Bradstock et al. 2010). Fire extent and severity interact to generate a particular configuration of burnt, partially-burnt and unburnt areas, collectively referred to as patchiness. Internal patchiness will influence the spatial pattern (size, shape and contrast) of intra-fire edges, but patchiness can also be conceptualised at a landscape scale as different fires burn and extinguish through time. Many fire management strategies aim to increase landscape variability by creating temporally and spatially dynamic fire mosaics, often referred to as patch mosaic burning (Bradstock et al. 2005; Parr & Andersen 2006; Di Stefano et al. 2013). However, the influence of patch mosaic burning on fire edges and how these might affect animal species and communities has not yet been considered as part of this paradigm.

Fire frequency (related to time since fire and inter-fire interval) influences the temporal and spatial flux of fire edges. Fire frequency can affect the physical properties of edges (e.g. contrast) due to its interaction with climate, particularly post-fire rainfall, which contributes to edge regeneration.

Seasonality of fire influences edge architecture due to its effect on fire severity and patchiness, as well as the rate at which plants regenerate post-fire. Unplanned fires more commonly occur in the driest months due to the seasonal growth and curing of fuel. Ease of ignition and flame transfer are also increased by high temperatures and low humidity common during summer (Bradstock 2010). However, seasonality can also be affected by prescribed burning activities that generally occur in different seasons to wildfires. Prescribed burning is usually undertaken early or late in the dry season when weather conditions are generally milder, more stable, with adequate fuel moisture to result in low-intensity fire (Penman et al. 2011).

EDGE ARCHITECTURE

The architecture of an edge refers to the physical characteristics of an edge zone, including size (volume) or depth-of-influence, shape and contrast. Architecturally different edges are predicted to have equally divergent edge effects. Edges resulting from fire are likely to be compositionally diverse due to inherent variability in fire behaviour in different landscapes and under different climatic conditions.

The volume of an edge (length, width and height) is expected to affect the willingness of animals to cross it. Edge volume may alter foraging success and exposure to predation at small spatial scales, and metapopulation dynamics at large spatial scales (Nams 2011).

Edge shape will likely influence permeability, with tortuous edges being more permeable than straight ones (Fagan, Fortin & Soykan 2003). For example, meadow voles (Microtus pennsylvanicus) crossed concave edges twice as often as straight or convex edges (Nams 2012). Edge shape can also either concentrate or disperse animals, depending on species specific edge responses. In modified agricultural and urban systems, species that are attracted to edges are more likely to collect in convexities and disperse from concavities, while the opposite is largely true for animals that avoid edges (Nams 2011).

Edge contrast refers to the differences in structure and composition between adjoining parts of the landscape (i.e. between different vegetation growth stages) and is a key element influencing the movement of material, energy and organisms (Villaseñor et al. 2014). In flammable landscapes, edge contrast is strongly influenced by topography, climate and fire severity. Fire edges are likely to have high contrast immediately post-fire (Fig. 1c), and this contrast is expected to decrease over time as burnt parts of the landscape regenerate. However, different vegetation types have different regenerating capacities (i.e. resprouters compared to seeders) and differing levels of resilience to fire (i.e. forests are likely to support more vegetation structure post-fire

18 than grasslands). However, animals themselves can also influence edge contrast. Many grazing animals are attracted to recently burnt areas (Meers & Adams 2003; Savadogo, Sawadogo & Tiveau 2007) and intense post-fire grazing can maintain the divergence in structure between burnt and unburnt areas. Topography can also play a role in determining edge contrast. For example, in the temperate regions of the southern hemisphere edges on north-facing slopes are likely to be characterised by stronger edge contrast than those on south-facing slopes due to increased radiation and lower moisture, resulting in increased flammability.

SPECIES TRAITS

Species’ edge responses are usually attributed to physical architecture and biotic dynamics, but are also likely to be influenced by a species’ ability to perceive boundaries (Baguette & Van Dyck 2007), as well as a series of morphological, behavioural and life-history traits. How species respond to fire edges will be a function of their mobility (Ries et al. 2004), but also habitat and diet requirements, adaptability to disturbance, and susceptibility to other processes such as competition and predation.

Highly mobile organisms are more likely to survive edge creation compared to sessile species. Mobility and body size have an allometric relationship with metabolic rate, energy use and physical ability, with larger animals generally requiring bigger home ranges than smaller animals (Lehman, Rajaonson & Day 2006). Larger body size might require foraging over large areas, increasing the chances of encountering more edge habitat, however larger animals may be more able to cross edges and exploit adjacent habitat than smaller ones (Lees & Peres 2009).

Diet specificity and habitat associations will also influence species’ edge responses. Species with more specialised requirements have been shown to avoid fire-affected areas until important resources re-accumulate (Borchert & Borchert 2013). In contrast, species with generalist food or habitat requirements are predicted to fare better in a newly disturbed environment due to their ability to exploit a broader range of resources.

Fire edges may also influence the thermal landscape, potentially exacerbating fire edge effects for some species. In fragmented landscapes patch edges frequently experience higher average temperatures and larger thermal variability than that of patch interiors (Tuff, Tuff & Davies 2016). Reduced vegetation cover after fire may cause species with narrow temperature thresholds to avoid the burnt side of edges for the first few years after fire. However, some species (such as

19 ectotherms) may benefit from the contrast occurring at fire edges by ‘shuttling’ (Dreisig 1984) across burnt and unburnt edges to regulate body temperature. Understanding species thermal sensitivities and temperature thresholds will improve our ability to predict species responses to edges created by fire.

EDGE DYNAMICS

PERMEABILITY/CONNECTIVITY

Edges are often characterised by the rate at which they facilitate or impede movement of resources and organisms, processes that are strongly influenced by edge architecture. Edge permeability - the ease with which animals, materials and/or energy cross a boundary - influences species activity patterns, influencing species movement at or near edges (Nams 2012). A hard edge is a boundary that individuals may find difficult to cross, whereas a soft edge will be reasonably permeable to most animals. The degree to which materials, energy or organisms can flow across an edge has been largely attributed to vegetation structure (Cadenasso et al. 2003b), however characteristics of the animals themselves (e.g. resource requirements or physical traits) can also influence boundary permeability, predominantly through changing rates of predation and competition.

Edge permeability is a site-specific concept and its expression in time and space influences connectivity, determining the capacity of populations to move across landscapes. Edge permeability and landscape connectivity are the result of interactions between biophysical properties, components of a disturbance regime and the physical architecture of edges. Landscape connectivity is thought to depend on how an organism perceives and responds to landscape structures at various spatial scales (Bélisle 2005). The ability of animals to cross fire edges, access available refuges and recolonise burnt landscapes will be influenced by both small- scale permeability and the functional connectivity of the wider landscape. Unburnt refuges may sustain source populations that can recolonise burnt landscapes (Robinson et al. 2013), however recolonisation rates may be influenced by the permeability of an edge for the species in question.

ECOLOGICAL FLOWS, ACCESS TO RESOURCES AND SPECIES INTERACTIONS

The key dynamics commonly affected by edges are ecological flows, access to resources and species interactions. Edges can amplify, diminish or reflect ecological flows (Ries et al. 2004) and the rate at which these dynamics are affected is largely a function of edge permeability. Changes in processes occurring at edges can result in increased or reduced access to resources for some

20 species, thereby changing the nature of species interactions. Changes to resource availability at edges can influence interspecific competition and alter community composition (Youngentob et al. 2012). For example, predators are known to exploit recently burnt areas (McGregor et al. 2014; Hradsky et al. 2017), as reduced cover increases access to structurally complex habitats and therefore better hunting opportunities (Doherty, Davis & van Etten 2015). Edges are known to increase predation risk for many species, particularly for birds (Fisher & Wiebe 2006; Vetter, Rücker & Storch 2013), and small mammals (Kingston & Morris 2000; Šálek et al. 2010; Hof, Snellenberg & Bright 2012). Accordingly, animals living near edges may alter their behaviour to compensate for high predation risk, such as decreasing their use of high-exposure locations or reducing visually conspicuous behaviours (Anderson & Boutin 2002).

SPECIES EDGE RESPONSE

The culmination of all the factors listed above result in a species’ edge response, which is commonly described as being positive, neutral or negative (Ries et al. 2004). Edge response can be considered at the community, species or individual level. Multi-species edge responses are often reported using a measure of community composition such as species richness. Single species responses are usually measured as a change in occupancy, abundance or behaviour, and may be partitioned further into sex or age-specific effects. Both single and multi-species edge responses occur along a continuum of spatial and temporal scales.

2.7 FUTURE RESEARCH Our model provides a conceptual understanding of edge creation in flammable landscapes and associated implications for fauna. However, few data quantifying these processes exist providing several avenues for future research.

1. Understanding fire as an agent of edge creation The agent of edge creation can strongly influence ecological patterns and processes, but few edge-related studies have considered how the process of creation influences edge effects. Understanding how fires interact with biophysical properties to create edges is an important first step in future fire-edge research.

2. Modelling the spatio-temporal flux of fire edges in flammable landscapes Edges in highly modified systems are often maintained at a relatively stable state, whereas edges in natural systems are dynamic, changing both spatially and temporally. In flammable landscapes, modelling the effect of fire cycles and plant regeneration rates on the distribution, abundance

21 and architecture of edges will be an important precursor to understanding fire-induced edge effects more broadly, particularly at landscape scales.

Fire edges need to be studied at multiple spatial and temporal scales. Understanding how permeability at single edges interact to influence landscape-scale structural and functional connectivity will be important for the conservation of biodiversity in fire-prone systems, particularly when considering landscapes that contain complex and varied fire histories.

3. Understanding the effect of fire edges on edge dynamics Ecological flows, resource selection and species interactions are predicted to be influenced by edge architecture. However, the interaction between edge architecture and edge dynamics has not yet been studied in flammable ecosystems. Better knowledge of these relationships will aid our understanding of the underlying mechanisms that drive species and community edge responses in fire-prone landscapes.

CONCEPTUAL MODELS AS BAYESIAN NETWORKS: ENHANCING FUTURE RESEARCH

To direct future research efforts our conceptual model has been designed as a Bayesian Network (BN). A BN is a statistical framework capable of analysing complex environmental relationships between a range of variables (Penman, Price & Bradstock 2012). BNs have strong predictive power where empirical datasets are incomplete, and can be used to predict species responses in cases where potential drivers are correlated. In the absence of a complete dataset, a BN will allow the sensitivity of the output to be tested against different factors by inserting a range of possible values into each node. This will identify the most influential nodes, generating hypotheses and highlighting focal points for further research. Further, the conceptual model can become a numerical model as empirical data are acquired, strengthening predictive outputs over time.

In our conceptual model we propose that biophysical properties can interact with components of the fire regime to influence edge architecture. Edge architecture and species traits influence edge permeability, which affect important processes (i.e. movement, resource selection and species interactions) that influence both species and community-level edge responses. Conversion of the conceptual model into a BN will enable these conceptual advances to be effectively combined with new data as fire-edge research is conducted. The capacity of the BN to deal with data gaps and to be used as a hypothesis generation tool will be particularly useful given the current paucity of information about faunal responses to fire edges.

3 HABITAT USE AT FIRE EDGES: DO ANIMALS FOLLOW TEMPORAL PATTERNS OF HABITAT CHANGE?

3.1 ABSTRACT Edges are ecologically important environmental features that have been well researched in agricultural and urban landscapes. However, edges remain poorly understood in flammable ecosystems where spatially and temporally dynamic fire edges are expected to influence activity patterns, particularly for animals reliant on vegetation for cover, foraging or nesting. We used a space-for-time substitution sampling design and selected a series of 26 treatment sites burnt by prescribed fire, where time since fire ranged from 0-7 years. Ten long-unburnt sites acted as controls. At each treatment site we identified a burnt/unburnt edge and used camera traps and Elliott traps to survey ground-dwelling mammals. Habitat structure was measured at all 36 sites. We used general and generalised linear mixed models to determine the response of both habitat and animals to time since fire on both burnt and unburnt sides of edges. In addition, we used a resource selection index to assess the congruence between changes in understorey complexity and animal activity identified in the first set of analyses. Understorey complexity followed a hump-shaped trend over time on the burnt side of the edge, remaining constant on the unburnt side. Larger animals with generalised resource requirements were more active at burnt edges immediately after fire. Despite some activity on the burnt side of edges immediately after fire, small mammals were generally less active on burnt edges for up to three years. Activity patterns of native species did not follow observed changes in vegetation structure on the burnt side of edges, with high usage during times of reduced understorey complexity, and low usage when complexity was high. Habitat change on the burnt side of fire edges may therefore not be a good predictor of native animal use. In general, patterns on the unburnt sides of edges were similar over time. Foxes and cats were also using the burnt edge more than expected immediately after fire. Therefore, integrated predator and fire management are recommended to improve biodiversity conservation in flammable ecosystems.

3.2 INTRODUCTION Fire influences the distribution and abundance of fauna by altering the availability of resources in time and space (Fox 1982; Bond & Keeley 2005; Pastro, Dickman & Letnic 2011). Understanding how fire-induced changes in resource availability affect biota is critical for conservation in fire-prone regions as it helps managers predict the outcomes of different approaches to management. Many fire management strategies aim to generate heterogeneous fire patterns, often referred to as patch mosaic burning (Bradstock et al. 2005; Parr & Andersen 2006). This has led to a patch-focused view of flammable landscapes, with many studies comparing the occurrence or abundance of fauna in burnt and unburnt patches (Banks et al. 2011a; Pastro, Dickman & Letnic 2014; Chia et al. 2015; Swan et al. 2016). While much has been gained from this approach, it does not explicitly consider the effect that fire edges (either between or within patches) might have on species abundance or movement over local or landscape scales (Parkins, York & Di Stefano 2018).

Edges are ecologically important because they influence a wide range of patterns and processes (Ries et al. 2004), with the resultant ecological changes collectively known as edge effects. Edge effects can result from changes in abiotic (e.g. radiation, moisture, temperature) and biotic (e.g. species interactions) sub-processes that interact to generate environments with different structural attributes and species assemblages compared to other parts of the landscape (Ries et al. 2004; Craig et al. 2015). Edges have been extensively researched in agricultural and urban landscapes (e.g. Murcia 1995; Driscoll & Donovan 2004; Villasenor et al. 2015) and much of our understanding about how edges affect animals comes from these highly modified ecosystems. Considerably less is known about edge effects in natural systems where disturbances such as fire are agents of edge creation.

Edges in agricultural or urban systems are often characterised by sharp transitions between adjacent habitats (e.g. between forest and farmland). This abrupt change in vegetation structure affects movement, behaviour and interspecific relationships of many species (Goosem 2001; Fahrig 2003; Ries et al. 2004; Wilson et al. 2010; da Rosa et al. 2018). Although data are lacking, similar responses are expected to occur at fire edges immediately post-fire, where the contrast between the burnt and unburnt sides of the edge is high. However, these effects are expected to reduce over time as vegetation regenerates and the edge contrast reduces (Parkins, York & Di Stefano 2018).

A fire edge is an interface or transition zone generated by fire, resulting in a boundary between areas of differing structural characteristics(Parkins, York & Di Stefano 2018). Fire edges can be between burnt and unburnt vegetation, between different burn severities or areas burnt at different frequencies. For some animals, a fire-induced change in structure represents a disruption to the continuity of resources and is therefore expected to influence activity patterns, particularly for animals reliant on vegetation for cover, foraging or nesting. Although not explicitly studied, fire edge effects are embedded in much of the existing research seeking to explain how fire affects animals. For example, unburnt patches in fire-affected landscapes are often regarded as refuges for animals, potentially enabling species to survive a fire event, persist after fire, and facilitate recolonisation of burnt habitat (Robinson et al. 2014). However, recolonisation from unburnt refugia requires individuals to cross a fire edge, making the spatial and temporal dynamics of edges an important aspect of refuge theory.

Our primary goal was to quantify the response of ground-dwelling mammals to fire edges ranging from 0 – 7 years since fire. We also sought to determine whether animal activity mirrored temporal changes in habitat structure. We expected reduced understorey complexity on the burnt side of edges immediately after fire, with the contrast between burnt and unburnt decreasing over time as vegetation on the burnt side regenerated. We predicted that species with behaviours strongly linked to understorey habitat (see table S1, supplementary material) would reduce their activity at burnt edges until the vegetation had regenerated. In comparison, we did not expect habitat change at fire edges to affect species with generalist habitat and diet requirements. We also predicted that two introduced predators, feral cat, Felis catus, and red fox, Vulpes vulpes, would increase their activity on the burnt side of edges while understorey complexity remained low. Cats and foxes are known to exploit recently burnt sites (McGregor et al. 2014; Leahy et al. 2016; McGregor et al. 2016; Hradsky et al. 2017) as reduced cover in burnt areas increases access to adjacent unburnt habitats, providing enhanced hunting opportunities (Doherty, Davis & van Etten 2015).

3.3 METHODS

STUDY REGION

Our study area was in the Central Highlands of Victoria, a topographically variable landscape of tall eucalypt forest in southeast Australia (~37°20’ - 37°55’S and 145°30’ - 146°20’E). Elevation ranges from 200 – 1500 metres and the climate is temperate with cool wet winters and warm dry summers. Mean annual rainfall is 1110 mm and average annual temperatures range from 7.2 26

(min.) to 18.5oC (max.). All sites were located in Eucalypt forest dominated by mixed overstorey species including Mountain Ash (Eucalyptus regnans), Messmate Stringybark (Eucalyptus obliqua), Mountain Grey Gum (Eucalyptus cypellocarpa), with a diverse open understorey of shrubs, ferns, grasses and herbs.

Large wildfires (>100 000 ha) have occurred in this region over the last 100 years, most recently in 1983 and 2009. In an attempt to reduce the impact of severe wildfires on human life and property, prescribed fire is increasingly applied during the cooler, wetter months of spring and autumn. Prescribed fires in these forests are relatively small (several hundred hectares), and usually result in a low-severity burn (Penman et al. 2007).

SAMPLING DESIGN AND SITE SELECTION

We used a space-for-time substitution sampling design; selecting twenty-six treatment sites and ten control sites (Figure 3.1). Treatment sites were burnt by prescribed fire in the seven years prior to the study. Sites were identified in ArcMap 10 (ESRI 2011), using a spatial layer representing time since fire. Potential edge locations were identified by randomly positioning points in gully systems, as fire edges often exist where drier midslope vegetation transitions into wetter gully vegetation. Edges were ground-truthed in the field and sites established where edges occurred (Figure 3.2). Suitable edges were classified as those with at least 50 metres of unburnt vegetation adjacent to at least 50 metres of burnt vegetation, at sites unaffected by other factors such as logging, bulldozer lines, fuel breaks, roads or houses. Sites were randomly positioned within burnt areas, at a minimum distance of 400 metres from other sites. Within each burn block we established one to three sites. At each site a 60-metre transect was established perpendicular to the edge, with 30 metres extending into burnt forest, and 30 metres into unburnt vegetation (Figure 3.3). Sites were positioned to include a mix of north and south facing slopes.

Ten control sites were established in areas unaffected by fire or logging since 1939, and positioned within the same forest types and with similar topographical variation as treatment sites. We systematically positioned the control sites within 0-300 metres of gullies, as this was the range observed at treatment sites. Control sites were used as a long-unburnt reference state and to account for any position-on-slope effect, as the distribution of some species is influenced by proximity to gullies (e.g. Claridge et al. 2008).

MAMMAL SURVEYS

We used camera trapping and live trapping to detect ground-dwelling mammals. Surveys were conducted once, between January and June 2015. Pairs of unbaited Reconyx HC cameras were deployed for 20 days at 2 m, 10 m and 30 m from the edge into both the burnt and unburnt forest, resulting in 12 cameras per site (Figure 3.3B). Each pair of cameras was positioned parallel to the edge, but facing in opposite directions (i.e. on opposing sides of a tree). Cameras were fastened to a tree at a height of 30 cm. A patch of vegetation (100 × 100 cm) was cleared at a distance of 1.5 metres in front of the camera, to allow the motion sensor to work effectively and to aid identification of species. We used two camera models, HC500 and HC600, and randomly allocated model types to camera stations.

Reconyx cameras use a passive infrared motion detector and infrared flash to capture images of passing animals. We used the maximum sensitivity setting and programmed cameras to record images continuously while movement was detected, with five photographs per trigger event. Fauna captured in photos were identified to species level, using a reference guide (Menkhorst & Knight 2004) and a library of images acquired from other studies to aid species identification. Species identification was completed by two researchers with similar levels of experience. For each photo we recorded the species, date and time of day.

We used 18 aluminium Elliott traps (9 × 10 × 33 cm) per site, deployed over four consecutive nights to capture small mammals. Traps were placed along a transect at intervals of 2 m, 10 m, and 30 m either side of the edge zone (into burnt and unburnt vegetation) (Figure 3.3), and in locations considered likely to be encountered by animals, such as adjacent to logs or runways of dense vegetation. Three traps per position on the transect were deployed, a minimum of 3 m apart. To provide food for captured animals, traps were baited with peanut butter, rolled oats and golden syrup. A handful of unbleached cotton was added to provide insulation, and plastic bags were placed over traps to protect animals from rain. Trapped animals were identified, weighed, and sexed. Each animal was marked uniquely on the base of the tail to detect recaptures within the trapping period, and then released. To minimise disturbance during camera trapping, Elliott traps were deployed at the conclusion of the camera trapping session.

Figure 3.1 Location of the control and treatment sites in south-eastern Australia

HABITAT STRUCTURE ASSESSMENTS

We measured habitat structure variables that were both potentially impacted by fire and important for a range of ground-dwelling species (Catling, Burt & Forrester 2000; Fox, Taylor & Thompson 2003; McElhinny et al. 2006). Vegetation structure was quantified using the point

29 intercept method (Jonasson 1983) along 30 m transects running parallel to the edge at 2 m, 10 m and 30 m (Figure 3.3). Measurements were recorded at 15 evenly-spaced points along each sampling transect.

At each sampling point, vegetation was classified as fern (ground fern, tree fern, bracken fern), shrub (small, large), tree (small, large), grass, herb or sedge and recorded as present or absent in each of five vertical strata (0-20 cm; 21-50 cm; 51-100 cm; 101-200 cm; 201-300 cm). We divided the summed presences by the total number of samples to estimate the proportional cover for each variable on both the burnt and unburnt side of the edge. Canopy cover, litter depth and coarse woody debris (CWD) cover was also recorded at the same sampling points. Canopy cover was recorded as percentage cover using a densitometer, litter depth was measured using a ruler, and values for both were averaged across sampling points. Coarse woody debris cover was quantified by measuring the diameter of each log (>10 cm diameter and >50 cm in length) that intersected the 30 m transect. We converted log diameter to log volume per transect (Van Wagner 1968).

We reduced the number of vegetation variables by excluding any that had low variation and/or were present at less than 50 percent of sites. Six variables were selected for analysis: litter depth (cm), coarse woody debris (volume), small shrubs (0-50 cm), bracken fern Pteridium esculentum (0- 100 cm), Wire Grass Tetrarrhena juncea (0-100 cm), and dead material (0-100 cm). At each site, and for each side of the edge, we created an understorey complexity score by summing standardised (divided by maximum) values of these six variables.

Figure 3.2 A fire edge between burnt and unburnt vegetation resulting from a prescribed burn in the Central Highlands of Victoria. This image was taken two weeks post-burn (Image: Parkins 2015).

Figure 3.3 A) edge locations in a treatment block burnt in 2013 B) Locations of each Elliott and camera trap, and vegetation survey in relation to an edge.

During preliminary analysis we found that distance from gully had little influence on the response variables, so we excluded this from the final analysis. In addition, animal data from the individual distance-from-edge positions were sparse so we pooled data from the three positions on either side of the edge (i.e. 2 m, 10 m and 30 m in burnt and unburnt, respectively). Vegetation data were also pooled, so the two data sets could be compared.

We conducted our analysis in three stages. In stage 1 we used linear mixed models (LMMs) to assess the response of individual structural variables (litter depth, coarse woody debris, small shrubs, Bracken Fern, Wire Grass, and dead material) and the understory complexity score (UCS) to two categorical predictor variables, edge position (burnt, unburnt) and time since fire (0, 1-2, 3, 6-7 and 76 years since fire). Mixed models were necessary to account for spatial nesting, as sites were nested within burn blocks and data from each half of the transect (burnt, unburnt) was nested within site. We selected the most parsimonious random effect structure by running a global model containing all fixed effects and their interactions (Zuur et al. 2009). We used Akaike’s Information Criterion corrected for small samples size (AICc) to compare levels of support for models containing site, transect and transect nested within site, as alternative random effects. Inclusion of transect resulted in the lowest AICc for all models and was subsequently used as the random effect for the remainder of the analysis.

For each structural response variable, a candidate set of models was built containing singular, additive and interactive combinations of edge position and time since fire (TSF). Models were compared using an information-theoretic approach where those with lower AICc values are considered the most parsimonious, and Akaike weights are used to indicate the relative likelihood that a model was the best in the set (Burnham and Anderson 2002). Models were constructed in R using the package lme4 (Bates, Maechler & Bolker 2011). Marginal (fixed effects) and conditional (fixed plus random effects) R2 was calculated in the package MuMIn (Barton 2013) and used as a measure of model fit (Nakagawa & Schielzeth 2013).

In stage 2 we used generalised linear mixed models (GLMMs) with binomial errors to assess the response of animal activity to edge position and TSF. The response variable was a species activity index – calculated as the proportion of days a species was detected (i.e. presence [1] or absence [0] per species per day, summed over the total number of days per trapping method). Elliott trap data were used to create small mammal indices (agile antechinus Antechinus agilis, and bush rat Rattus fuscipes), and camera data were used for all other species. Our activity index (generated in the same manner for both trapping techniques) does not differentiate between a single detection of many individuals or frequent detection of one individual, and therefore

33 represents total species activity. We used an activity index in preference to presence/absence because it provided higher resolution information (Geary et al. 2018). In addition, all statistical models built using the activity index ran, while several models built using presence-absence data failed to converge.

In stage 3 we assessed the congruence between animal activity and understorey complexity over time on both sides of the edge. To achieve this we compared habitat use (the activity index) to resource availability (the understorey complexity index) using a resource selection index. Using the understorey complexity and species activity models that included the time since fire and edge interaction, we ran the bootMer function in the lme4 package (Wood & Scheipl 2014; Bates et al. 2015) to generate 1000 iterations of model predictions for each combination of TSF category and edge position. We standardised the outputs (by dividing by the maximum) to generate comparable data sets representing habitat use and availability for each species over time, on both the burnt and unburnt side of the edge.

We used Vanderploeg and Scavia’s Relativised Electivity Index (Vanderploeg & Scavia 1979) to combine use and availability data into a resource selection index, reflecting the degree to which species use of habitat mirrored the availability of understorey vegetation. The index is calculated as E = [W-(1/n)]/[W+(1/n)], where n is the number of resource types and W = (ui/ai)/∑ ui/ai, where u is resource use and a is resource availability for resource type i. We calculated the index separately for the burnt and unburnt sides of the edge, so in our calculations n = 5, the number of TSF categories. The index ranges between -1 and +1, with negative and positive values indicating use is less than or greater than expected based on availability, respectively, and 0 indicating use in proportion to availability. For each combination of TSF category and edge position we used the median of the 1000 iterations as a measure of central tendency and the 2.5th and 97.5th percentiles as lower and upper 95 percent confidence limits.

3.4 RESULTS

HABITAT STRUCTURE AT FIRE EDGES

Understorey complexity was best explained by the interaction between TSF and edge position, with an Akaike weight of 1. This model was 22.6 AICc units better than the next best model, with a marginal R2 of 0.46 and a conditional R2 of 0.69. Understorey complexity followed a hump-shaped trend over time on the burnt side of the edge but remained constant on the unburnt side (Figure 3.4). Understorey complexity was lower at burnt edges compared to unburnt edges 0-2 years after fire. By three years post-fire this pattern was reversed, with 34 complexity higher at burnt edges. Six to seven years after fire, understorey complexity was similar on both sides of the edge.

The interaction between TSF and edge position proved the best model for four of the six structural variables, including: litter depth, dead material (0-100 cm), wire grass (0-100 cm) and bracken fern (0-100 cm) (Table 3.1). All individual structural variables were reduced on the burnt edge immediately post-fire (Figure 3.5).

Table 3.1 Relationships between time since fire (TSF) and edge position (burnt/unburnt), and six habitat structure variables. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size), Akaike weight is the likelihood of the model being the best in the set; R2 values are the proportion of variance explained by the models. Models within 2 AICc units of the best model and the null models for each variable are presented. See Appendix A-2 for the complete set of models with estimates and 95 percent confidence intervals.

Akaike Marginal Conditional Response Variable Model AICC weight R2 R2

Dead material 0-100cm TSF × EDGE 0.00 0.995 0.58 0.72

NULL 35.03 0.000 0.00 0.43

Litter TSF × EDGE 0.00 0.962 0.65 0.78

NULL 50.40 0.000 0.00 0.40

Shrub 0-50cm NULL 0.00 0.465 0.00 0.60

TSF 1.93 0.177 0.15 0.60

Wire Grass 0-100cm TSF × EDGE 0.00 1.000 0.33 0.78

NULL 20.79 0.000 0.00 0.42

Bracken fern 0-100cm TSF × EDGE 0.00 0.462 0.25 0.55

NULL 0.90 0.295 0.00 0.29

Coarse woody debris NULL 0.00 0.359 0.00 0.05

TSF 0.18 0.328 0.12 0.12

ANIMAL ACTIVITY AT FIRE EDGES

A total of 2592 nights of live trapping resulted in the capture of 55 bush rats, 51 agile antechinus and 15 dusky antechinus (Antechinus mimetes). Due to low numbers the dusky antechinus was excluded from analysis. From 8640 trap nights, camera traps detected 11 species, with five species recorded at a sufficient number of sites for modelling (see Appendix A-4). These included three native species (swamp wallaby, Wallabia bicolor, mountain brushtail possum, Trichosurus cunninghami and common wombat, Vombatus ursinus) and two non-native species (red fox, Vulpes vulpes, and feral cat, Felis catus).

The activity of swamp wallabies, mountain brushtail possums and agile antechinus was best predicted by an interaction between TSF and edge position (Table 3.2). Wombat activity was best predicted by TSF alone and bush rat activity was best predicted by edge position alone. The model containing TSF and edge position in an additive combination was the best model for predicting fox activity, while cats did not respond to TSF or edge (Table 3.3). Antechinus and bush rat activity on the burnt side of edges was higher immediately after fire compared to the next time step (Figure 3.6a and b respectively), however this was not statistically detectable for bush rats. Swamp wallabies were more active on the burnt edge for the first two years, but had similar activity levels on both sides of the edge from three years post-fire (Figure 3.6c). Mountain brushtail possums were more active on the burnt side of the edge, but only immediately after a burn (Figure 3.6d). Wombats had a hump shaped response to time since fire. Fox activity decreased with time since fire (Figure 3.7b).

SELECTION AT FIRE EDGES

Selection for dense understorey vegetation at fire edges varied over time and between species (Figure 3.8). The response of all native species was similar, but patterns were more distinct for antechinus, possums and swamp wallabies. These species used the burnt side of edges more than expected based on habitat availability immediately after fire. Mountain brushtail possums used both sides of the edge more than expected immediately post-fire (Figure 3.8d). In contrast, agile antechinus used the burnt edge less than expected from 1-7 years after fire (Figure 3.8a). Both species of non-native predators used the burnt side of edges more than expected based on habitat availability immediately after fire (Figure 3.9). Several species used the unburnt side of edges less than expected based on habitat availability. These were: wombats immediately after fire, bush rats 1-2 years, swamp wallabies and agile antechinus 6-7 years, and mountain brushtail

38 possum 76 years after fire (Figure 3.8). Foxes used the unburnt side of edges more 1-2 years after fire, but then less in the two oldest TSF categories (Figure 3.9).

Conditional Species Model AICc Akaike weight Marginal R2 R2 Swamp wallaby TSF × EDGE 0.00 0.875 0.04 0.13

NULL 20.79 0.000 0.00 0.09

Mountain Brushtail TSF ×EDGE 0.00 0.607 0.10 0.35 Possum

NULL 9.23 0.006 0.00 0.31

Wombat TSF 0.00 0.338 0.04 0.13

NULL 0.43 0.237 0.00 0.13

TSF + EDGE 1.05 0.200 0.04 0.13

EDGE 1.20 0.186 0.00 0.13

Agile Antechinus TSF × EDGE 0.00 0.824 0.33 -

NULL 5.84 0.045 0.11 0.21

Bush Rat EDGE 0.00 0.757 0.05 0.15

NULL 9.84 0.006 0.00 0.04

Akaike Species Model AIC Marginal R2 Conditional R2 Weight

Cat NULL 0.00 0.558 0.00 0.27

EDGE 1.02 0.336 0.01 0.28

Fox TSF + EDGE 0.00 0.714 0.23 0.37

NULL 7.56 0.016 0.00 0.25

3.5 DISCUSSION Wildfires are predicted to increase in extent and severity in the flammable forested ecosystems in Australia as a result of climate change (Flannigan et al. 2009). To combat this, prescribed burning is increasingly being used as a fire management tool (Penman et al. 2011). Given this likely increase in fire activity, a better understanding of fire edges and their implications for fauna is important. We show that low-intensity prescribed fire reduced understorey complexity at burnt edges for two years. Despite some activity immediately after fire, small mammals were generally less active on the burnt side of edges for at least three years. In contrast, larger-bodied species were more active on the burnt side of edges immediately after fire, with patterns of high activity decreasing over time. In general, animals did not follow patterns of vegetation change over time. Most native mammals, and two non-native predators, cats and foxes, used the burnt side of edges more than expected for the first two years after fire, while understorey complexity was low. However, three years after fire, use of the burnt edge decreased for all native mammals despite the peak in understorey complexity during this time. Our results show that animal activity

43 at fire edges is not just driven by habitat structure, and that other processes such as predation are likely to be important.

HABITAT CHANGE AT FIRE EDGES

Understorey complexity was reduced on the burnt side of edges for two years following fire. During this time, the contrast between burnt and unburnt vegetation was high. However, three years after fire understorey complexity on the burnt side was greater than the unburnt side, and greater than long unburnt control sites. Bracken fern cover returned slightly faster than other components, reaching peak cover three years after fire. Litter and wire grass increased steadily over time at burnt edges, reaching peak availability 6-7 years after fire. Dead material and Wire Grass remained higher on the burnt side of edges 6-7 years after fire, while all other variables were similar on both sides, suggesting the availability of resources was comparable across fire edges by this time.

The rate at which vegetation regenerates after fire is strongly dependent on the regenerative strategies of species (Pausas 1999; Broncano & Retana 2004) (i.e. re-sprouters versus re-seeders, or re-accumulation over time in the case of litter and CWD). However, a range of other biophysical factors also influence the speed of regeneration, including post-fire weather (particularly rainfall) and vegetation type (Keeley, Fotheringham & Baer‐Keeley 2005; Plavsic 2014). Post-fire regeneration can be stimulated by nutrient increases from ash and by stronger light availability due to a more open understorey which promotes an initial flush of vegetation growth (Torres et al. 2017).

RESPONSE OF NATIVE FAUNA TO FIRE EDGES

We predicted that species with behaviours strongly linked to understorey habitat (see Appendix A-1) would reduce their activity at burnt edges until the vegetation had regenerated. While our results generally support this prediction, there were a few exceptions. In particular, we observed some activity of both bush rats and agile antechinus at burnt edges immediately after fire, when the contrast between burnt/unburnt was at its highest and vegetation cover was at its lowest. Bush rats and agile antechinus have been shown to survive fire events (Banks et al. 2011a; Swan et al. 2016), and this activity may be due to residual populations remaining in the burnt landscape. However, surviving a fire does not guarantee survival in the post-fire environment which may be characterised by a sharp reduction in food and shelter, changes in competition and increased predation risk (Sutherland & Dickman 1999). A combination of these factors is potentially driving the reduced activity observed in small mammals from one year after fire.

Species use of fire edges is predicted to be a function of their mobility, habitat and diet requirements, adaptability to disturbance and susceptibility to other processes such as competition and predation (Parkins, York & Di Stefano 2018). As such, we expected species with different resource requirements and life history strategies to respond to fire edges in different ways. Species with generalist habitat requirements have been shown to occupy burnt, edge and unburnt areas in similar abundance within one year of a high severity wildfire, while species with specialist resource requirements took several years to occupy edge zones and were only detected in high abundance in burnt vegetation nine years after fire (Diffendorfer et al. 2012; Borchert & Borchert 2013).

Bush rats are late seral specialists, preferring areas with high log volume and a dense midstorey, and avoiding areas associated with frequent burning (Spencer et al. 2005). They have high home range fidelity and are not known to readily increase or shift their home range into unburnt areas following fire (Thompson et al. 1989; Sutherland & Dickman 1999; Fordyce et al. 2015; Swan et al. 2016). In contrast, agile antechinus are able to exploit a broader range of habitat components than the bush rat and individuals may shift their home ranges after fire (Swan et al. 2016). Agile antechinus were responding to both time since fire and edge, whereas for bush rats edge was the best model. Despite these differences in fire-response, both species of small mammal demonstrated some activity on the burnt side of edges immediately after fire, followed by reduced activity, which remained low for several years after fire.

We predicted that species with generalist resource requirements would exhibit similar levels of activity on both side of the edge, regardless of the high contrast in vegetation in the first few years after fire. This prediction was supported for wombats, who did not respond to edge. However, swamp wallabies and mountain brushtail possums increased their use of the burnt side of edges immediately after fire, with high activity continuing for the swamp wallaby for two years. Early successional post-fire habitats are characterised by vigorous growth of understorey vegetation and are often dominated by ground-foraging herbivorous mammal species (Torre & Díaz 2004). Herbivores are often attracted to the flush of vegetation post-fire because of enhanced forage quality and increased productivity in burnt areas (Wilsey 1996; Eby et al. 2014). Fire-stimulated plant growth often has higher protein and increased levels of potassium, nitrogen and magnesium compared to older vegetation (Van de Vijver, Poot & Prins 1999). Hypogeous fungi is also often stimulated by fire (Claridge, Trappe & Hansen 2009), a food source for both swamp wallabies and possums (Claridge & Lindenmayer 1998; Claridge, Trappe & Claridge

2001). The high activity of these species at burnt edges in the first few years after fire is likely being driven by the availability of abundant, high quality food on the regenerating side of the edge.

RESPONSE OF INVASIVE PREDATORS TO FIRE EDGES

Predators strongly influence the structure and function of many ecological communities (Ritchie & Johnson 2009) and are known to exploit recently burnt areas (McGregor et al. 2014; Leahy et al. 2016; McGregor et al. 2016; Hradsky et al. 2017). We therefore predicted that cats and foxes would increase their activity at burnt edges after fire, while the contrast between burnt and unburnt remained high. Our results show that foxes were more active at recently burnt edges immediately after fire. However, cats did not respond to either time since fire or edge.

Changes in habitat structure can affect predator-prey interactions by influencing perceptions of risk, the relative abundance of predators and prey, and the rate at which predators can find and kill prey (Gorini et al. 2012). A prescribed fire in south-eastern Australia resulted in an 80 percent reduction in understorey cover, coupled with a five-fold increase in the occurrence of invasive predators within the first three months after fire (Hradsky et al. 2017). We suggest that the brief increase in fox activity on the burnt side of edges is due to a combination of reduced understorey complexity, ease of movement and increased access to prey.

DO SPECIES FOLLOW TEMPORAL PATTERNS OF RESOURCE CHANGE?

In general, animals were not following temporal patterns of resource availability. Most native species were found to be using the burnt side of edges more than expected based on the availability of resources immediately after fire. Use of the burnt edge during this time is likely due to residual populations of small mammals remaining in burnt areas, and increased exploration of the burnt edge by some of the larger species. For medium-sized ground-dwelling species like the swamp wallaby, the flush of highly nutritious vegetation immediately post-fire is likely driving their use of recently burnt edges. However, three years after fire, use of burnt edges was less than expected for all species despite the peak in understorey habitat. One possible explanation may be that although the understorey vegetation is dense during this time, it may not have constituted a good food resource, as it was primarily CWD, bracken fern and wire grass driving the increased complexity during this time. The habitat accommodation model (Fox 1982) suggests that species will re-enter fire affected areas when their resource requirements are met during the post-fire succession of vegetation. The model has not previously been applied to fire edges, but suggests that species with a requirement for dense understorey structure should peak in abundance on the

46 burnt side of the edge three years after fire, due to the high availability of complex understorey vegetation. However, we show that animals were not following patterns of vegetation change at fire edges. These results were unexpected, particularly for the swamp wallaby, as previous research has shown this species to have high abundance in areas with dense vegetation (Di Stefano et al. 2007; Williamson, Doherty & Di Stefano 2012). These results suggest that species’ edge response may not be solely related to habitat structure and that other important processes are occurring at edge zones.

Both foxes and cats were using the burnt side of edges more than expected immediately after fire. Both species are known to exploit recently burnt and open habitats adjacent to dense vegetation (McGregor et al. 2014; Hradsky et al. 2017) with reduced cover increasing access to structurally complex habitats, and providing better hunting opportunities (Doherty, Davis & van Etten 2015). Foxes have been shown to change their diet in response to fire, with higher consumption of larger mammals pre-fire, and a shift to medium-sized mammals post-fire (Hradsky et al. 2017). Fox selection of fire edges was higher on the burnt side immediately after fire, followed by selection for unburnt edges 1-2 years post-fire. This suggests that foxes may be using unburnt vegetation as a conduit for hunting in burnt areas in the first few years after burning.

Edge effects may also extend beyond the burnt side of the edge and influence the unburnt habitat. For some species, use of the unburnt side of edges did not remain constant despite the lack of variation in understorey complexity on this side of the edge. For example, use of both the burnt and unburnt sides of edges declined 1-2 years after fire for the bush rat. Agile antechinus and swamp wallaby use of the unburnt side was also reduced 6-7 years after fire. However, it is important to note that these data were collected close to fire edges and over a small spatial scale. Therefore, these patterns may not be representative of patterns occurring further away from fire edges.

MANAGEMENT IMPLICATIONS

The results from this study provide several important insights for management of fire-prone, forested landscapes in Australia. Patchy ‘pyrodiverse’ burning is often thought to benefit biodiversity (Parr & Andersen 2006), and unburned patches within larger burns can provide ecological refuges for native fauna (Robinson et al. 2013). However, fire-moderated changes in vegetation structure may have negative consequences for fauna if they increase predation

47 pressure at fire edges. Small unburnt refuges will have a high edge-to-area ratio and targeted use of burnt/unburnt edges by predators may diminish the protective value of small unburned patches of vegetation for native fauna. Therefore, integrated predator and fire management is recommended to improve biodiversity conservation in flammable ecosystems.

Fire managers often utilise pre-existing edges as the boundaries for future burns. Multiple prescribed burns adjacent to each other may result in large parts of a landscape that are affected by fire-induced edge effects. Land managers should therefore carefully consider the spatial dynamics of fire edges and the implications this may have in limiting activity patterns and ability to recolonise burnt areas for the first few years following a burn, particularly for small ground- dwelling mammals. Careful planning of future prescribed fires is important, and attempts to reduce the amount of young (<3 years since fire) edges in a landscape is encouraged.

Habitat structure is often used as surrogates for animal biodiversity in flammable systems (Haslem et al. 2011). However, our results show that animal use of fire edges did not follow patterns of habitat availability, suggesting that a measure of habitat complexity after fire may not be a good predictor of animal occurrence. Future fire management must therefore acknowledge the potential for fire to affect fauna and vegetation differently.

3.6 CONCLUSIONS We show that understorey complexity was reduced on the burnt side of fire edges for the first two years after fire. Ground-dwelling mammals varied in their response to fire edges. Larger animals with generalised resource requirements were more active at burnt edges immediately after fire. Despite some activity on the burnt side of edges immediately after fire, small mammals were generally less active on burnt edges for up to three years. Our results show that species were not following patterns of temporal change in vegetation structure, with high usage during times of reduced understorey complexity and low usage when complexity was high. This suggests that habitat change is not a good predictor of animal use at fire edges, and that other important processes are likely occurring. In particular, foxes and cats were using the burnt side of edges immediately after fire, which may have important implications for the long-term persistence of native fauna if changes in habitat structure at fire edges cause predation rates to increase. Fire-prone landscapes have traditionally been thought of in terms of habitat classes such as burnt/unburnt, core area/matrix, with little consideration given to the boundaries between these patches. This study demonstrates that fire edges are ecologically important features of fire-affected landscapes. Their capacity to influence both native and invasive species

48 suggests that consideration of fire edge effects in future fire management planning will contribute to the conservation of biodiversity in flammable systems.

ACKNOWLEDGEMENTS

Thank you to Amy Scott, Matthew Swan, Holly Sitters, Julio Najera, Kirsten Langmaid and John Sharp for helping with fieldwork; Trent Penman and Graham Hepworth for providing statistical advice; and local land management agencies for facilitating access to many sites. Funding was provided by the Victorian Government Department of Environment, Land, Water and Planning, and the Holsworth Wildlife Research Endowment. This research was conducted with ethics approval from the University of Melbourne Animal Ethics Committee (AEC 1413324).

4 INCREASING DETECTABILITY IN CAMERA TRAP SURVEYS: MORE CAMERAS, MORE DAYS, OR BOTH?

4.1 ABSTRACT Camera trapping is an increasingly popular method for surveying wildlife. Two important considerations in the design of camera surveys are (a) the number of detection units (cameras or camera-groups) per site and (b) deployment time. The trade-off between these two factors will likely affect data quality, but there is little information about their relative influence on species detectability. In this study, we investigate the trade-off between deploying more detection units or extending the length of the sampling period, on two frequently assessed variables in camera trapping studies – species richness and detection probability. We used a bootstrapping procedure to iteratively sample each possible combination of detection units (1-6 camera pairs) and camera days (1-34) to investigate the effect of both factors on species richness and detection probability. We show that increasing the number of cameras deployed per site is an effective way of increasing species richness and probability of detection. Multiple detection units in combination with longer deployment times were necessary to detect a high proportion of species present. Increasing detection units or deployment time (or both) resulted in high overall detection probability for the more detectible species (e.g. wombats, swamp wallabies), but multiple detection units were always needed to achieve high detection probability in a reasonable timeframe (<50 days) for less detectible species (e.g. agile antechinus, mountain brushtail possums). Camera trapping is a widely used survey technique and these findings will help inform decisions about the number of detection units and deployment time required for future camera trapping surveys.

4.2 INTRODUCTION Camera trapping is an increasingly used survey technique that employs remote photography to record animals as they pass in front of a camera (Rowcliffe & Carbone 2008; O'Connell, Nichols & Karanth 2010). In the last twenty years camera traps have superseded many other survey techniques due to their ease of use, applicability across a broad range of study species, capacity to operate in a range of environments, and record data under harsh weather conditions. Recent advances in battery technology and micro-storage capabilities mean that camera traps can now be deployed for long time periods and store large volumes of data. After the initial purchase there is also relatively less financial output associated with their continued use when compared to other methods (O'Connell, Nichols & Karanth 2010). For small ground-dwelling mammals, camera trapping is considerably cheaper and logistically easier than live trapping (De Bondi et al. 2010). In addition, it is superior from an ethical position as animals do not need to be physically detained or handled, resulting in significantly less interference to normal behaviour and considerably reducing the risk of injury or death.

Camera traps have been used for a broad range of applications. These include monitoring single species (Linkie et al. 2013; Meek & Vernes 2016), broad-scale biodiversity surveys (Tobler et al. 2008; O'Brien et al. 2010; Ahumada, Hurtado & Lizcano 2013; Swan et al. 2014; Bowler et al. 2017), behavioural studies (Ridout & Linkie 2009; Rowcliffe et al. 2014) and population density estimates (Karanth & Nichols 1998; Kelly et al. 2008). Camera traps can also assist in rapid assessments of conservation threats (Silveira, Jacomo & Diniz 2003) as species inventories can be made relatively quickly by using multiple cameras over short time periods (Kelly 2008).

There has been a recent interest in evaluating the effectiveness of camera traps under different conditions. Several studies have examined the performance of different models (Swann, Kawanishi & Palmer 2011; Swan, Di Stefano & Christie 2014), and different camera orientations (De Bondi et al. 2010). Guidelines now exist for best-practice camera set-up (Meek & Pittet 2012), optimal software settings and specifications (Rowcliffe et al. 2011), and minimum reporting requirements (Meek et al. 2014). Despite these improvements, questions still arise regarding the best study design to obtain scientifically robust and statistically sufficient data to answer a specific question, while ensuring a study is financially and logistically feasible. Surprisingly few studies have examined how design decisions influence data quality (but see Tobler et al. 2008; Rovero et al. 2013; Si, Kays & Ding 2014; Smith et al. 2017b).

Two important considerations in camera survey design are (a) the number of detection units per site and (b) deployment time. Both factors will likely affect data quality but there is little information about their relative influence. Consequently the number of cameras established and their deployment time vary greatly among studies. We searched the Web of Science for the 100 most recently published camera trapping studies (as of January 16th 2018- see Appendix B-1 for detailed methods). We found the number of cameras deployed had very little variation, with a range of 1-3 cameras established per site (median 1, standard deviation 0.43). ‘Site’ varied between studies but generally refers to a sample location within a study. In contrast, we found substantial variability in how many days cameras were deployed at a site, ranging from 4 - 730 days (median 30, standard deviation 122.8). Twenty-eight percent of these studies did not report the total number of trap nights, and sixteen percent did not report the number of cameras per site.

Based on this review, researchers commonly vary deployment time but not the number of cameras per site, despite the lack of data about the effect of these choices on study objectives. In this study we assessed the trade-off between deploying more detection units and extending the length of the sampling period on two frequently assessed response variables in camera trapping studies – species richness and detection probability.

4.3 METHODS Our study was undertaken in the Central Highlands of Victoria, a topographically variable landscape of tall eucalypt forest in southeast Australia (~37°20’ - 37°55’S and 145°30’ - 146°20’E). The topography is hilly to mountainous (200-1500 m elevation) with moderate to steep slopes common. The climate is temperate with cool wet winters and warm dry summers. Mean annual rainfall is 1110mm and average temperatures range from 7.2 (min) to 18.5oC (max) (Noojee, Station ID: 085153).

STUDY DESIGN

The data used in this study were collected as part of a broader study (See Chapter 3). Ten sites were established in mature forest stands that had not been disturbed by fire or logging since 1939 (Figure 4.1). At each site, 60 m transects were randomly allocated to either north (n=6) or south (n=4) facing slopes, and were established at different distances from gullies to ensure there was no bias associated with the preference of different species for dryer or wetter parts of the landscape. Transects were a minimum of 400 m apart.

Reconyx Hyperfire digital motion camera traps were established at each site to target small, medium and large mammals. We deployed two models of cameras (HC500, HC600) and randomised their allocation at each site. Both models use a passive infrared motion detector and infrared flash to capture images of passing animals. While the flash of the HC600 is invisible, the HC500 flash is visible to animals as a soft red glow that may influence the detectability of some species. Cameras were set to maximum sensitivity and programmed to record images continuously while movement was detected, with five photographs per trigger event. Cameras were deployed in pairs, hereafter referred to as a detection unit, with a total of 6 detection units (12 cameras) per site. The spacing of the cameras was asymmetrical as the study for which the cameras were originally established (Chapter 3) required this particular spacing arrangement. Cameras were fastened to trees 30 cm above the ground and positioned in pairs facing away from each other. A patch of vegetation (1 m × 1 m) was cleared at a distance of 1.5 metres in front of the camera to allow the motion sensor to work effectively and to aid identification of species. Cameras were not baited and were deployed for a minimum of 17 days, between January and June 2016.

DATA ANALYSIS

Data were collected across ten sites, from 120 camera locations. Camera pairs formed a natural detection unit so data from the two cameras in a pair were pooled for analysis. Cameras ran for a minimum of 17 days, so we used this as a cut-off point to standardise across all cameras. Fauna captured in photos were identified to species level, using a reference guide to aid species identification (Menkhorst & Knight 2004). Species identification was completed by two researchers with similar levels of experience. For each photo we recorded the species, date and time of day. For analysis we used species presence/absence per site, per detection unit, per day (24 hour period). We excluded all birds from the analysis as our cameras were set up to detect ground-dwelling mammals. We also excluded any species that were expected to only occur in one of the two surveyed forest type (see, Appendix B-2).

We used a bootstrapping procedure to iteratively sample all possible combinations of detection units (1-6) and camera days (1-17), to investigate the effect of both factors on species richness and detection probability. The procedure randomly resampled the data with replacement (where each value is returned to the set before the next iteration), repeating the process 1000 times for every day by detection unit combination. This approach enabled us to include an extrapolation out to 34 days. Calculations were performed in the R statistical environment (R Core Team

2017) using purpose-written code and the package Vegan (Oksanen et al. 2011). Species richness varied among sites, so we represented our results using proportional species richness (range 0 to 1) to have a consistent measure across all sites. Detectability estimates were used to calculate the number of days required to have 80, 90 or 95 percent confidence of detecting a species if it were present for 1-6 detection units. Probability of detection was calculated for each species based on the full data set using a naïve estimate, i.e. detectability was constant across all surveys. These values were then used to estimate the number of days required to have 80, 90 or 95 percent confidence of detecting a species if it were present for 1-6 detection units.

4.4 RESULTS Thirteen species of mammal were detected from a total of 2,040 camera days, with seven species recorded at sufficient sites for modelling (i.e. >4 sites). These included six native species (agile antechinus, Antechinus agilis, bush rat, Rattus fuscipes, short-beaked echidna, Tachyglossus aculeatus, swamp wallaby, Wallabia bicolor, mountain brushtail possum, Trichosurus cunninghami, common wombat, Vombatus ursinus) and one invasive species (feral cat, Felis catus). All other species detected were excluded from the analysis.

SPECIES RICHNESS

Increasing the number of detection units from 1 to 6 substantially increased the species richness index (Figure 4.2). One camera pair deployed for 34 days detected approximately 40 percent of the species present, whereas six camera pairs deployed for 34 days detected around 80 percent of species. The benefit of longer deployment time decreased the longer cameras were deployed. This was also the case for the addition of detection units. There were considerable benefits associated with initially increasing detection units, however the benefits diminished as additional detection units were added. Increasing the number of detection units from two to three resulted in a similar value of relative species richness (around 0.6) in approximately half the time.

DETECTABILITY

For overall detectability, both the length of deployment and number of detection units was important (Figure 4.3). Irrespective of the number of detection units, more days were required to increase detection probability. For example, to be 80, 90 and 95 percent confident of detecting all seven species, six cameras would need to be deployed for 40, 58 or 72 days respectively (Figure 4.3f). Increasing the number of detection units resulted in more species being detected in a shorter period of time.

For individual species, the best use of both time and detection units varied according to the detectability of species. Increasing detection units or deployment time (or both) resulted in high overall detection probability for the more detectible species (Figure 4.4), but multiple detection units were always needed to achieve high detection probability in a reasonable timeframe (<50 days) for less detectible species (Figure 4.5).

4.5 DISCUSSION Camera traps are a popular tool for detecting animals, enabling researchers to answer a range of questions with relatively less disturbance, effort or cost compared to other trapping techniques (De Bondi et al. 2010; O'Connell, Nichols & Karanth 2010). Despite their popularity, few studies have presented guidelines for improving study designs (but see Meek et al. 2014; Si, Kays & Ding 2014; Galvez et al. 2016; Smith et al. 2017b). The consequence is that new surveys are often

59 designed without guidance or follow untested rules-of-thumb about the sampling effort needed to reach a desired outcome. We compared the relative effect of increasing both detection units and deployment time on species richness and detectability. Our results show that adding extra detection units provided benefits over and above simply leaving cameras out for longer.

Traditionally, methods for increasing species detections in camera trapping studies have involved increasing the length of deployment time, or including the use of lures or bait to attract animals to the camera (Paull, Claridge & Barry 2011; Balme, Hunter & Robinson 2014; Austin et al. 2017). However, there are some potential problems associated with these options. In some studies population closure (i.e. no immigration or emigration) is assumed, but this assumption will be violated if a survey continues for too long. Extended sampling periods may also require more intra-survey maintenance, with batteries and storage cards needing replacement, possibly resulting in more financial output associated with repeated visits. The use of bait or lures may increase the detectability and identification of animals, however baiting may also bias results if animals are attracted over large areas, or their presence at bait stations reduces the likelihood of other species being detected (Rocha, Ramalho & Magnusson 2016). For example, a common goal of camera trapping surveys is to link species presences with site-level covariates (e.g Swan et al. 2015; Smith et al. 2017a), but findings may be faulty if species are attracted to the camera from outside what constitutes a site. Our results show that increasing the number of detection units deployed per site is an effective way of increasing species richness and probability of detection, while reducing the likelihood of unwanted effects associated with longer deployment durations or baiting.

In some situations, increasing the number of detection units and reducing deployment time per site will enable more locations to be sampled within a survey period. Given site is the unit of replication in most camera studies, survey designs that increase the number of sites that can be surveyed without reducing detection probability will be important for increasing statistical power. Assuming cameras are already available and do not need to be purchased, increasing the number of cameras deployed at each site is likely to result in a negligible increase in cost compared to establishing new sites. For example, our results show that a relative species richness value of 0.6 can be achieved by deploying two detection units (four cameras per site) for 34 days, or three detection units (6 cameras per site) for 17 days (Figure 4.2). If we assume access to 100 cameras, the first strategy will mean 25 sites can be surveyed over a 34 day period. In comparison, the second strategy will mean 32 sites can be surveyed over a 34 day period. Increasing the number of detection units per site from 2-3 means an additional 7 sites could be surveyed during the

60 same time period. However, this does not take into account any additional time required to set up additional cameras, but is expected to result in a negligible increase.

SURVEY DESIGN WILL DEPEND ON STUDY OBJECTIVES

The most appropriate survey design will vary depending on the objectives of each study. If the goal is to detect a high proportion of the species present at each site (i.e. biodiversity surveys), our results suggest the best method to achieve this is by using a combination of more cameras and longer deployment times. However, if the goal is to detect a particular species (i.e. targeted surveys) the most effective design will depend on the detectability of that species. If detectability is high, a small number of cameras will suffice. If detectability is low, a greater number of cameras will be required to detect the species with reasonable confidence (e.g. Figure 4.5a).

When considering common species, it is often more effective to conduct intensive surveys at fewer locations, while the opposite is largely true for rare species (e.g. less intensive surveys at more locations) (MacKenzie & Royle 2005; Tobler et al. 2008; Srbek-Araujo & Chiarello 2013). The amount of survey effort required can substantially increase for cryptic or rare species (Galvez et al. 2016). Therefore, the benefits associated with deploying more cameras are more pronounced when targeting elusive or rare animals. Species that are unlikely to be detected at every site or occur in low densities within a study region would benefit from a) more sites being sampled across the region and b) more cameras being deployed per site. Deploying more cameras (both at a site and across a study region) means that more microhabitats are sampled, increasing the chance of detecting species that utilise different habitat components.

Our results are specific to paired camera designs and more research is needed to determine the trade-off between detection units and deployment time when other design features (e.g. camera number and location, camera model, camera orientation, transect length) are varied. Furthermore, other external factors unique to a study region may affect species detections. The set type (i.e. baited, unbaited and type of bait), spatial coverage of cameras and local animal densities will also affect the probability of detection. It will be important for researchers to carefully consider the factors that may influence the detectability of the species in their region from the outset and incorporate these considerations into the planning phase of their study.

4.6 CONCLUSIONS Camera number and deployment time are two critical aspects of study design for every camera trapping survey. In this study we provide new evidence that increasing the number of detection units deployed per site is an effective way to increase the detection of ground-dwelling mammals. In addition, we demonstrate that multiple detection units in combination with longer deployment times were necessary to detect a high proportion of the species present. Increasing the number of detection units or increasing deployment time (or both) resulted in high overall detection probability for more detectible species, but multiple detection units were always needed to achieve high detection probability in a reasonable timeframe (<50 days) for less detectible species. Our findings will help guide future decisions about the number of detection units and length of deployment for camera trapping surveys.

ACKNOWLEDGEMENTS

Thank you to Amy Scott, Matthew Swan, Holly Sitters, Julio Najera for helping with fieldwork; Local land management agencies for facilitating access to sites. Funding was provided by the Victorian Government Department of Environment, Land, Water and Planning, and the Holsworth Wildlife Research Endowment. This research was conducted with ethics approval from the University of Melbourne Animal Ethics Committee (AEC 1413324).

5 THE DEVIL IS IN THE DETAIL: SEX AND FOREST TYPE INFLUENCE POST-FIRE RESOURCE SELECTION IN A SEMI-ARBOREAL MAMMAL

5.1 ABSTRACT Fire is an important natural disturbance process and a major driver of variation in resource availability in many ecosystems. Large, unplanned wildfires frequently burn across broad landscapes, affecting multiple forest types. Understanding how resources differ across a landscape and how this variability drives resource selection will provide insight into how species meet their requirements for survival following major wildfires. Furthermore, differences in the ecological and reproductive roles of males and females can often result in each sex prioritising different resources. We investigated resource selection of a semi-arboreal mammal eight years after a major wildfire. We used GPS telemetry to collect habitat use data across two different forest types that had been severely burnt during the 2009 Black Saturday wildfire in Victoria, Australia. Our objectives were to determine: 1) if the availability of key habitat resources differed between wet and dry forest eight years after a major wildfire, 2) if forest type influenced selection of key resources by mountain brushtail possums, and 3) if male and female mountain brushtail possums differed in their selection of key resources in each forest type. We found that resource availability differed between forest types, predominantly driven by the greater availability of wattles and dead trees in wet forest, and greater availability of live trees in dry forest. Resource selection by mountain brushtail possums 8 years after a large wildfire depended on both sex and forest type, with males (n = 9) and females (n = 7) using resources differently, particularly in terms of food and denning resources. For example, males selected areas with abundant dead trees in dry forest, whereas females selected for this resource in wet forest. Studies that account for the spatial variability of resources and species’ demographic status are likely to provide a better understanding of how major fires influence resource selection, which will be important for the conservation of fauna in fire-prone systems.

5.2 INTRODUCTION Resource selection is a behavioural process that facilitates survival and reproduction, maximises fitness, and can provide fundamental information about how animals meet their resource requirements (Boyce & McDonald 1999). Resource selection is principally driven by resource quality (Manly et al. 2007), but is often mediated by other factors such as competition and predation (Morris 2003), and may differ between individuals depending on their age, sex or reproductive status (Ardia & Bildstein 1997). In addition, selection is a function of both resource use and availability (Manly et al. 2007; Beyer et al. 2010), and so can be strongly influenced by factors affecting the distribution and abundance of important resources, i.e. a ‘functional response’ (Mysterud & Ims 1998).

Fire is an important natural disturbance process (Bowman et al. 2009) and a major driver of variation in resource availability in many forest and woodland ecosystems (Fox 1982; Bond & Keeley 2005; Pastro, Dickman & Letnic 2011). The impact of fire on biodiversity is of growing concern, particularly as climate change is predicted to increase wildfire extent and severity in many terrestrial systems (Cary & Banks 2000; Flannigan et al. 2009). A better understanding of how animals select resources in areas burnt by large wildfires is important for the conservation of biodiversity in flammable systems.

The survival and persistence of animals after fire is largely driven by the abundance and distribution of remaining resources and the regeneration rates of key habitat components. Post- fire regeneration will be influenced by components of the disturbance regime (i.e. fire severity, extent, frequency) and biophysical properties (i.e. topography, substrate, post-fire rainfall). The rate of regeneration often differs between vegetation communities, driven by different responses to disturbance (Noble & Slatyer 1980). For example, the dry and wet sclerophyll forests of south-eastern Australia are characterised by markedly different fire responses (Fairman, Nitschke & Bennett 2016): dry sclerophyll forests are ‘fire-tolerant’ while wet sclerophyll forests are ‘fire- sensitive’. In dry systems the dominant canopy species, Eucalyptus obliqua, is seldom killed by fire, regenerating vegetatively by epicormic shoots. In contrast, the dominant tree species in wet sclerophyll forests (e.g. Eucalyptus regnans, Eucalyptus delegatensis), are frequently killed by fire with the next generation of individuals regenerating from canopy-stored seed. Different fire responses can lead to variation in the availability of resources across a landscape. A better understanding of how this variability drives the selection process for species will provide insight into how species meet their requirements for survival following major wildfires.

Large, unplanned wildfires (>100, 000 ha) frequently affect multiple forest types. For example, in 2009 a high-severity wildfire burnt through more than 250 000 hectares of wet and dry sclerophyll forest north east of Melbourne, southeast Australia. This region is home to numerous arboreal marsupials including the mountain brushtail possum (Trichosurus cunninghami), a medium- sized (2.6-4.2 kg), semi-arboreal, nocturnal marsupial that occurs in a range of forest types along the Great Dividing Range in south-eastern Australia. Mountain brushtail possums survived the wildfire, persisting in many areas burnt at different severities (Banks et al. 2011b; Chia et al. 2015; Berry et al. 2017). However, their use of resources in the post-fire environment and how resource selection might be influenced by the different fire responses of wet and dry forest has not been studied.

We identified four primary resources expected to be important for the mountain brushtail possum: food availability, understorey cover, the presence of suitable den trees, and the presence of live vegetation facilitating arboreal movement. Our objectives were to determine: 1) if the availability of key habitat resources differed between wet and dry forest eight years after a major wildfire, 2) if forest type influenced selection of key resources by mountain brushtail possums, and 3) if male and female mountain brushtail possums differed in their selection of key resources in each forest type. We expected resource availability to differ between wet and dry forest, and predicted that resource selection by the mountain brushtail possum would be influenced by both sex of the animal and forest type.

5.3 METHODS This study was undertaken in the temperate eucalypt forests of the Victorian Central Highlands, in south-eastern Australia (~-37°75’ latitude; 145°69’ longitude) (Figure 5.1). The region is topographically variable, with elevation ranging from 200 – 1500 m. The climate is temperate with cool wet winters and warm dry summers. Mean annual rainfall is 1110 mm and average annual temperatures range from 7.2 (min) to 18.5oC (max). Large wildfires (>100 000 ha) have periodically occurred throughout this region over the last 100 years, with large, high severity fires occurring in 1939, 1983 and most recently in 2009. During the 2009 fires, a large proportion of the fire-affected area was burnt at high-severity, with patches of low-severity and unburnt forest within the fire perimeter (Figure 5.1). We selected two broad forest types: foothills forest and montane eucalypt forest (hereafter referred to as dry and wet forest, respectively). The dry forest had an overstorey dominated by Messmate Stringybark (E. obliqua) and Broad and Narrow-leaf Peppermint (E. dives and E. radiata), and a diverse open understorey of shrubs, grasses and herbs.

Wet forest was characterised by a Mountain Ash (Eucalyptus regnans) overstorey, with a tall broad- leaved shrubby understorey and a moist, shaded fern-rich ground layer.

SITE SELECTION

We categorised the area burnt by the 2009 fires into three broad fire severity classes, derived from remotely-sensed fire severity maps provided by the local land management agency. These were (1) unburnt areas, (2) areas burnt mostly at low fire severities and (3) areas mostly burnt at high fire severities (Table 5.1). Initially, our objective was to trap possums in each fire severity class to determine if resource selection varied between severity classes. However, after extensive trapping (~2000 trap nights), no animals were trapped in unburnt locations and only one animal was trapped in an area categorised as low severity. Accordingly, the results for this study come from 15 individuals trapped at high severity sites, and 1 individual from low severity. Given the low sample size for the low severity class, we have combined the data across the two severity classes. Possums were successfully trapped in three locations (Figure 5.1).

Table 5.1 Burn severity categories.

Fire severity category Description 1. Unburnt No crown scorch, no understorey burn 2. Low Understorey burnt, light or moderate crown scorch 3. High Complete crown burn

TRAPPING PROTOCOL AND COLLAR DEPLOYMENT

Possums were captured in wire cage traps (30 × 31 × 63 cm), positioned on the ground and baited with apple and peanut butter. Cages were covered with black plastic to provide protection from adverse weather conditions. Traps were opened at dusk and checked at first light. Captured animals were transferred from the cage into a hessian sack and sedated using an intramuscular injection of Zoletil 100 (0.5mg/kg) (Virbac Australia, Sydney). Each individual was weighed, sexed and their tooth-wear estimated following the procedure outlined by Winter (1980). Animals with a tooth-wear rating of six or more were not collared, as this likely represents an animal of increased age, and/or may indicate problems in maintaining body weight and general health due to their advanced tooth wear, making mastication and digestion inefficient.

Figure 5.1 Area burnt by the 2009 wildfire north east of Melbourne, Australia, stratified by fire severity category. Possums were successfully trapped at three locations within the burn perimeter.

Only adults weighing more than 3 kilograms were fitted with collars, and females with dependant pouch young were also excluded. After sedation and collaring animals were returned to their trap, and released at dusk.

Healthy individuals were fitted with a custom-made GPS/VHF collar (Figure 5.2). Collars weighed < 5 percent of body mass in all cases and contained a GPS module (STM32), a 2300mAh lithium-ion battery pack, and a micro-SD card for data storage. We enclosed these components in a 3D printed housing (polyamide - nylon) to protect the device from damage (i.e. from weather or by the animal). This was then attached to a commercial VHF collar (SirTrack, model V5C 163E). Collars were set to record a position every 15 minutes, between 1700 and 0600. After three weeks, collared animals were radio-tracked to den trees using a VHF antenna and receiver. Traps were positioned around the base of the tree to recapture the animal and

68 remove the collar. If collars malfunctioned they were replaced, and on occasion collars were replaced for a second or third sampling period to extend the dataset.

DATA CLEANING AND PRE-PROCESSING

GPS data were screened to remove any fixes recorded while trapping was being conducted, or when the collar battery was severely depleted. Data were also filtered to remove position fixes with low accuracy. Fixes with less than three satellites were discarded, as were fixes with a horizontal dilution of precision (HDOP) value greater than five. Fixes were further screened using the approach of Bjørneraas et al. (2010) to remove erroneous samples based on unrealistic speeds and turning angles. Location data that reflected a turning angle of 170 - 190° and a travelling speed of > 400m per 15 minutes were excluded from the analysis. After data cleaning, we retained 17,103 GPS location data from 16 individuals (~9 males, 7 females).

Figure 5.2 Custom-made GPS circuit board, b) collar with GPS and VHF, c) Mountain brushtail possum with collar.

VEGETATION SURVEYS

We identified four primary resources expected to be important to mountain brushtail possums: food availability, understorey cover, the presence of suitable den trees, and the presence of live vegetation facilitating arboreal movement. These resources were represented by 11 variables described in Table 5.2. We measured all variables in 825 circular plots with a 15-m radius (Figure 5.3) established at the intersections of a 50 × 50 m sampling grid that we superimposed over each of the three study sites using the fishnet tool in ArcMap 10 (ESRI 2011). We used plots with a 15 m radius as this was the distance observers could clearly see given the dense understorey.

We generated an understorey complexity index for each plot in a similar manner to Newsome and Catling (1979), where the percentage cover of litter (<10 cm), coarse woody debris (CWD >10 cm), and vegetation cover < 50 cm and 51 ‒ 200 cm, were estimated. These scores were then summed to give an overall understorey complexity score for each site.

Figure 5.3 Sampling protocol for each of the 825 × 15 m vegetation sampling locations.

Table 5.2 Vegetation survey methods.

Unit of Variable Resource Method measurement

Visual cover assessment using a 5-point rating in 3 × 2 m radius sub-plots (Figure 5.3). Litter cover (<10 cm) Understorey cover Index (0 - 5) 0% = 0; 1-20% = 1; 21-40% = 2; 41-60% = 3; 61-80% = 4; 81-100% = 5

Visual cover assessment using a 5-point rating in 3 × 2 m radius sub-plots 0% = 0; 1- CWD (>10 cm) Understorey cover Index (0 - 5) 20% = 1; 21-40% = 2; 41-60% = 3; 61-80% = 4; 81-100% = 5

Visual cover assessment using a 5-point rating in 3 × 2 m radius sub-plots Vegetation cover <50 cm Understorey cover Index (0 - 5) 0% = 0; 1-20% = 1; 21-40% = 2; 41-60% = 3; 61-80% = 4; 81-100% = 5

Visual cover assessment using a 5-point rating in 3 × 2 m radius sub-plots Vegetation cover 51-200 cm Understorey cover Index (0 - 5) 0% = 0; 1-20% = 1; 21-40% = 2; 41-60% = 3; 61-80% = 4; 81-100% = 5

Visual cover assessment using a 5-point rating in 3 × 2 m radius sub-plots Canopy cover Movement Index (0 - 5) Using a densitometer 0% = 0; 1-20% = 1; 21-40% = 2; 41-60% = 3; 61-80% = 4; 81-100% = 5 Single visual assessment taken from the centre of each plot, using a 3-point rating 0 = no connection, 1 = poorly/sparsely connected foliage or limbs, few connecting Midstorey connectivity Movement Index (0 - 3) bridges; 2 = moderately connected, several lateral limbs of bridges connecting, good foliage cover; 3 = excellent connectivity, high foliage cover and many lateral bridges connecting, good connectivity to surrounding forest.

Basal area live trees Movement m2 / ha Single basal area sweep from the centre of each plot using a dendrometer.

Visual Assessment of the dominant species contributing to midstorey connectivity Main midstorey species Food./movement Species (name) Only recorded when connectivity was >2

Basal area of wattles Food m2 / ha Basal area sweep for all wattles species in 3 x 2 m radius subplots using a dendrometer.

Basal area dead trees Den m2 / ha Single basal area sweep from the centre of each plot using a dendrometer.

Visual assessment and count of large trees (diameter at breast height, DBH >50cm). Categorised into one of five tree forms (see Error! Reference source not found.). Tree form 1 (live) = live healthy trees, no signs of decay; Count, Tree form 2 (live) = live trees with slight decay (typically older trees beginning to Tree forms Den (separately for senesce) each form) Tree form 3 (dead) = dead trees in the early stages of decay Tree form 4 (dead) = dead trees in the mid-stages of decay Tree form 5 (dead) = highly decayed dead trees

RESOURCE SURFACES

We used Pearson’s correlation coefficient to identify highly correlated habitat variables (r ≥ 0.70). Variable pairs with the highest correlation included tree-form two and basal area of dead trees; live tree forms and basal area of live trees; canopy cover and basal area of live trees; and wattle basal area and presence of silver wattles (Appendix C-2). When variables were highly correlated, we chose a single variable from the set that best represented the resource of interest. This resulted in a final set of six explanatory habitat variables with r <0.52: understorey complexity, basal area of wattles, basal area live trees, basal area dead trees, midstorey connectivity, and the composite variable dead tree forms (which included tree forms 3, 4 and 5).

We used Inverse Distance Weighted interpolation in ArcMap 10 (ESRI 2011) to create continuous raster surfaces for each habitat variable. The default settings were used to determine pixel size of output. We estimated cell values by averaging the values of eight actual data points in the neighbourhood of each processing cell, with the assumption that the variable being mapped decreases in influence with increasing distance from its sampled location (Li & Heap 73

2011). The power parameter was set to 2 (the default value). This value controls the significance of surrounding points on the interpolated value. Using this method, we generated a series of spatial layers representing the abundance of each resource across the study area (see Figure 5.5; and Appendix C-4).

DEFINING RESOURCE USE AND AVAILABILITY

We followed a use-verses-availability design (Johnson et al. 2006; Manly et al. 2007) to assess resource selection. The ‘use’ sample comprised the cleaned position fixes of each possum. To generate the ‘available sample’, we generated a 100 percent minimum convex polygon (MCPs) around each individual possum’s full set of position fixes, and generated 1000 points within each MCP. For each used and available location, we extracted the values of each habitat variable from the interpolation maps using the extraction tool in ArcMap. Extracted habitat variables were standardised by dividing by the maximum observed value.

DATA ANALYSIS

We conducted our analysis in two stages. Firstly, we investigated differences in resource availability between wet and dry forest using multivariate techniques. Standardised data from the final set of six habitat variables (understorey complexity, midstorey connectivity, basal area of wattles, basal area of live trees, basal area dead trees and dead tree forms) were used to create a Euclidean resemblance matrix that formed the basis for a non-metric multi-dimensional scaling ordination. We then used ANOSIM to test for differences between forest types, and SIMPER to determine the proportional contributions made by each habitat variable. All analyses were run in PRIMER 6 (Clarke & Gorley 2006).

Secondly, to evaluate differences in mountain brushtail possum selection for each habitat variable between forest types and sexes, we fitted generalised linear mixed models with binomial errors in the R statistical environment (R Core Team 2017), using the package lme4 (Bates, Maechler & Bolker 2011). We created a series of models of the form response ~ forest type × sex × resource, where the response variable was binary (used [1] or available [0]) as per Manly et al. (2007), forest type was a two level predictor variable (dry, wet), sex was a two level predictor variable (male, female) and resource was one of the six habitat variables. A factor identifying each individual possum was included as a random intercept. We built a separate model for each of the six habitat variables and applied the dredge function in the MuMIn package (Barton 2013) to generate all possible model combinations (single, additive or interactive). Competing models for each habitat variable were compared using an information theoretic approach where lower values of Akaike’s information criterion (AICc) are considered the most parsimonious (Burnham & Anderson 2004). Akaike weights were used to indicate the relative likelihood that a model was the best in the set. Conditional and marginal R2 (Nakagawa & Schielzeth 2013) were calculated to assess model fit using the MuMIn package.

5.4 RESULTS An ordination of the data (Figure 5.6) and ANOSIM (global R statistic = 0.198, P< 0.001) showed that resource availability differed between forest types. The SIMPER analysis indicated that dead tree forms (22.5%), basal area of dead trees (21.4%) and basal area of live trees (17%) contributed most to the differences between forest types (Table 5.3). Dead tree forms, basal area of dead trees, basal area of wattles and understory complexity were higher in the wet forest than the dry forest. Basal area of live trees was substantially higher in the dry forest, and midstorey connectivity was similar between forest types.

Figure 5.6 Bi-plot generated from an ordination procedure, demonstrating the difference between wet and dry forest types.

Table 5.3 Differences between habitat variables across the two forest types. Variables are ranked according to those contributing most to the differences. Means and ranges represent values from the raw data, and percentage contributions were calculated using the transformed data. Wet Dry Range Range Percentage Habitat Variable forest forest (min-max) (min-max) Contribution (mean) (mean) Dead tree forms 2.56 0 - 10.6 1.52 0 – 7 22.5 (2-4) Basal area 11 0 - 40 7.61 0 - 26 21.4 dead trees Basal area 8.4 0 - 22 16.4 0 - 52.2 17.0 live trees Basal area 4.32 0 - 20.3 2.17 0 - 19.4 15.8 wattles Midstorey 1.51 0 – 3 1.42 0 - 3 12.0 connectivity Understorey 7.02 0 - 12.5 5.87 0 - 15.9 11.4 complexity

A three-way interaction in habitat selection was strongly supported for three habitat variables (dead tree forms, basal area of live trees and basal area of wattles), indicating that selection of these habitat variables by mountain brushtail possums depended on both forest type and sex (Table 5.4). For understorey complexity, there was some evidence that the three-way interaction was important, but the model without the three-way interactions also had substantial support. Female and male possums both selected for areas with high understorey cover in the wet forest and did not respond to understorey cover in the dry forest (Figure 5.7a and b). Males selected areas with abundant dead tree forms in the dry forest, while females selected for this resource in the wet forest (Figure 5.7e and f). Males and females selected high wattle basal area in the dry forest, but not in the wet forest. The slopes of these relationships differed somewhat between the sexes in the wet forest (Figure 5.7c and d). Midstorey connectivity was best predicted by a model without the three-way interaction: mountain brushtail possums selected for high midstorey cover in the dry forest and low midstorey cover in the wet forest, with little difference between the sexes (Figure 5.8).

Table 5.4 Responses of mountain brushtail possums to forest type, sex and habitat variables (understorey complexity, basal area live trees, basal area dead trees, basal area wattles, dead tree forms, midstorey connectivity) derived from linear mixed models. Levels of forest type (FT) were wet and dry; estimates associated with wet forest represent contrasts with dry, and estimates associated with males represent contrasts with females. Akaike's information criterion (AICc) was used to rank models. Models within two units of the top-ranked model are shown with Akaike weights. Parameter estimates with 95 percent confidence intervals (CI) are displayed. Two measures of fit are included: marginal R2 (R2m) is the variance explained by fixed factors and conditional R2 (R2c) is the variance explained by both fixed and random factors.

[Opposite page]

95 % CI Habitat variable Model AIC Akaike Weight Estimate R2m R2c (lower, upper)

Understorey 0 0.682 0.18 0.49 FT × Sex × UC Complexity FT 1.31 0.97, 1.65

(UC) Sex 1.97 0.48, 3.46

UC 3.84 3.35, 4.33

FT : Sex -2.49 -3.05, -1.93

FT : UC -3.8 -4.33, -3.27

Sex : UC -1.00 -1.66, -0.34

FT : Sex : UC 0.72 -0.01, 1.45

FT × UC + FT × Sex + UC × Sex 1.64 0.3

Dead tree FT × Sex × TF dead 0 1 0.19 0.51

forms FT -2.09 -2.31, -1.87

(TF dead) Sex 0.38 -1.15, 1.91

TF Dead -0.08 -0.53, 0.37

FT : Sex -0.07 -0.52, 0.38

FT : TF Dead 2.68 2.15, 3.21

Sex : TF Dead 2.09 1.42, 2.76

FT : Sex : TF Dead -5.79 -6.67, -4.91

Basal area FT × Sex × Live 0 1 0.27 0.59

live trees FT 0.58 0.34, 0.82

( BA Live) Sex 2.4 0.73, 4.07

BA Live 17.45 15.37, 19.53

FT : Sex -2.62 -3.03, -2.21

FT : BA Live -18.88 -20.98, -16.78

Sex : BA Live -9.09 -11.48, -6.70

FT : Sex : BA Live 6.21 3.74, 8.68

Basal area FT × Sex × Dead 0 1 0.19 0.49

dead trees FT -2.01 -2.21, -1.81

(BA Dead) Sex 0.43 -1.04, 1.90

BA Dead -5.15 -5.68, -4.62

FT : Sex -1.04 -1.45, -0.63

FT : BA Dead 4.53 3.88, 5.18

Sex : BA Dead 4.64 3.93, 5.35

FT : Sex : BA Dead -4.71 -5.65, -3.77

Basal area FT × Sex × Wattles 0 1 0.32 0.66

wattles FT -2.77 -3.06, -2.48

(Wattles) Sex 2.21 0.31, 4.11

BA Wattles -0.76 -1.23, -0.29

FT : Sex -3.37 -4.94, -1.80

FT : BA Wattles 3.97 3.34, 4.60

Sex : BA Wattles -3.8 -4.53, -3.07

FT : Sex : BA Wattles 4.68 3.45, 5.91

Midstorey FT × Connect 0 0.727 0.21 0.52 connectivity FT -4.09 -4.67, -3.51

(Connect) Sex -1.35 -2.83, 0.13

Connect -3.17 -3.61, -2.73

FT : Sex 1.87 1.25, 2.50

FT : Connect 4.11 2.95, 5.27

Sex : Connect 1.03 0.44, 1.63

FT × Sex × Connect 1.98 0.271

Figure 5.8 Model predictions for midstorey connectivity by a) forest type and b) sex. Shaded areas are 95 percent confidence intervals.

5.5 DISCUSSION Understanding how animals select resources after major wildfires aids biodiversity conservation in managed landscapes, and will become increasingly important as wildfires increase in extent and severity under climate change (Bowman et al. 2009; Flannigan et al. 2009; Cary et al. 2012). We found that resource availability for mountain brushtail possums differed substantially between wet and dry forest eight years after a large wildfire in south-eastern Australia, with wet forests characterised by more dead trees and wattles, but fewer live trees than dry forests. Selection of large dead trees and wattles were influenced by both sex and forest type, suggesting that selection is driven by both resource availability and sex-related differences in requirements for denning and food resources. In contrast, selection for dense or connected vegetation only differed between forest type, suggesting that both sexes have similar requirements for vegetation cover. Data on resource availability and demographic class will be necessary for accurate predictions of resource selection following major wildfires.

RESOURCE AVAILABILITY DIFFERS BETWEEN FOREST TYPES

We used six habitat variables to represent primary resources important for mountain brushtail possums: food availability (basal area of wattles), den trees (basal area of dead trees and dead tree forms), understorey cover (understorey complexity) and upperstorey vegetation to facilitate arboreal movement (midstorey connectivity and basal area of live trees). As expected, the

82 availability of these habitat variables differed between wet and dry forest, reflecting differences in the biophysical properties of the two vegetation communities and their responses to fire.

Large dead trees were more abundant in wet forest. In contrast, live trees were more abundant in dry forest, resulting in a higher canopy cover (Appendix C-3). The wet and dry sclerophyll forests of south-eastern Australia are characterised by markedly different responses to fire. Wet sclerophyll forests are fire-sensitive with fire-killed canopy species taking upwards of 35 years to reach maturity and redevelop canopy cover (Lindenmayer et al. 2000), and burnt forests are often dominated by standing dead trees for many years. In contrast, dry sclerophyll forests are fire- tolerant with canopy species regenerating vigorously from epicormic shoots on the bole and branches (Penman et al. 2017) and re-establishing live foliage cover rapidly after fire.

Mountain brushtail possums rely on large tree hollows for nesting and denning, which commonly occur in senescing or decaying trees (Lindenmayer et al. 1996; Martin 2006). As large dead trees were more abundant in wet forest, denning resources were likely to be more available in this forest type. In contrast, the higher availability of live trees and canopy cover in dry forest would likely facilitate greater arboreal movement in this forest type. Understorey complexity was higher in wet forest and was expected to be particularly important for this species when upperstorey connectivity and vegetation cover was low. Midstorey connectivity was similar between forest types.

Wattles were also more abundant in wet forest. Many Acacia species are disturbance specialists, germinating prolifically after high severity fire, particularly in wet forests dominated by Mountain Ash (Cunningham & Cremer 1965; Adams & Attiwill 1984; Hunt, Unwin & Beadle 1999), partly driven by reduced competition and increased light penetration from the open canopy in this forest type. Silver wattle is the primary food source for the mountain brushtail possum (Seebeck, Warneke & Baxter 1984), and so food availability was likely higher in the wet forest.

RESOURCE SELECTION DEPENDS ON BOTH SEX AND FOREST TYPE

Given that resource availability differed between wet and dry forests and that sex often strongly influences selection (Glutton-Brock & Vincent 1991a; Loe et al. 2006; Main 2008), we predicted that resource selection by mountain brushtail possums would be influenced by both sex and forest type. We found that selection of dead tree forms, basal area of dead trees and wattles were influenced by both sex and forest type. However, selection for understorey complexity, basal area of live trees and midstorey connectivity only differed between wet and dry forest, with similar patterns between males and females.

The variables where selection was influenced by both sex and forest type represented denning and food resources (basal area of dead trees, dead tree forms and basal area of wattles). Differences in the ecological and reproductive roles of males and females lead to sex-based differences in resources selection for many species (Johnson & Bayliss 1981; Glutton-Brock & Vincent 1991b; Ardia & Bildstein 1997; Corti & Shackleton 2002; Bowyer & Kie 2004; Loe et al. 2006; Main 2008). Males often prioritise areas of high forage availability and quality, whereas females often use poorer-quality habitats if doing so increases the likelihood of offspring survival (Main 2008). Environmental and behavioural factors can operate simultaneously to influence sex differences in resource use, and selection is expected to vary among locations where resource availability differs (Orians & Wittenberger 1991; Martin & Martin 2007).

Male mountain brushtail possums selected areas with abundant large dead trees in dry forest (where this habitat variable was less available), with selection remaining constant in the wet forest. In contrast, females selected for abundant dead tree forms in wet forest, but not in dry forest. In this landscape, wet forests that have been burnt by high severity fire support more tree hollows than less severely burnt sites (Berry et al. 2017), providing a key denning resource for mountain brushtail possums and other arboreal mammals (Lindenmayer, Cunningham & Donnelly 1997; Haslem et al. 2012). It is unclear why females responded positively to large dead tree availability in wet forest but not in dry forest where the availability of this key resource is limited ‒ other unmeasured resources that may influence habitat selection (such as predation risk and access to breeding partners) need further investigation.

Both males and females selected areas of high wattle abundance in dry forest where the resource was most limited. Selection of this resource was constant for males in the wet forest, while selection by females decreased with increasing availability. Areas of high wattle abundance in the wet forest may be characterised by increased intra-specific competition, or higher predation risk (e.g. from birds of prey) due to the more open canopy. Females may therefore be prioritising other resources (e.g. shelter) over food availability in the wet forest where wattles are abundant overall.

The variables where selection was primarily influenced by forest type were characteristic of vegetation structure at the ground, midstorey and canopy levels. Possums selected for high understorey complexity and large live trees in the wet forest, but not the dry forest, likely in response to the limited availability of these resources in the wet forest. Following high severity wildfire, wet forests are often dominated by fire-killed mountain ash trees, providing an open canopy with little shelter and upper-storey connectivity for arboreal mammals. For many species,

84 exposed or open areas are characterised by higher perceived predation risk (Coleman & Hill 2014; Mella, Banks & McArthur 2014). Mountain brushtail possums are predominantly arboreal, but come to the ground intermittently to feed, or move to different locations. In landscapes burnt by moderate and high severity fire, mountain brushtail possums prefer intact, live canopy and dense understorey vegetation (Berry et al. 2017), and select for areas of abundant foliage (Lindenmayer, Norton & Tanton 1990). Feeding in trees with high vegetation cover provides possums with a low-quality diet but relative safety. Ground feeding may provide a potentially higher-quality diet (Kerle 1984), but the cost of accessing these resources may be associated with higher risks (Mella, Banks & McArthur 2014).

The availability of midstorey connectivity was similar between forest types, but possums only selected for this resource in dry forest. Eight years after wildfire, canopy cover remained low in the wet forest, and possums selected for areas characterised by large trees, suggesting that the species prioritised the use of intact canopy rather than midstorey connectivity in wet forest.

5.6 CONCLUSIONS Fire is a natural disturbance process, and primarily influences the survival and distribution of animals through its effects on resource availability (Monamy & Fox 2000; Swan et al. 2015). Eight years after a major wildfire in the Victorian Central Highlands, the availability of key habitat resources for mountain brushtail possums differed between forest types, predominantly driven by differences in the availability of wattles, and live and dead trees. Selection of these resources by the mountain brushtail possum was influenced by forest type, with selection also differing between males and females for some resources. Selection of large dead trees and wattles were influenced by both sex and forest type, suggesting that patterns of selection are driven by both resource availability and sex-related differences in requirements for denning and food resources. In contrast, selection for dense vegetation cover at the ground, midstorey and canopy level only differed between forest type, suggesting that both sexes respond similarly to vegetation cover and perceived risk. Studies that account for the spatial variability of resources and species demographic status are likely to provide a better understanding of how major fires influence faunal resource selection, and are key to predicting the effects of disturbance events on biodiversity.

ACKNOWLEDGEMENTS

Thank you to Ben Lehtonen, George Hnanicek, Julio Najera, Kirsten Langmaid, Bess Hartcher, Annalie Dorph and April Glory for helping with fieldwork; Tristam Horn, Blake Allan and Kean Maizel for helping with GPS collar design and development; Bronwyn Hradsky for assistance with GPS data wrangling; and local land management agencies for facilitating access to many sites. Funding was provided by the Victorian Government Department of Environment, Land, Water and Planning, and the Holsworth Wildlife Research Endowment. This research was conducted with ethics approval from the University of Melbourne Animal Ethics Committee (AEC 1513673).

6 SYNTHESIS

Edges are ecologically important environmental features that have been well researched in agricultural and urban landscapes, but remain poorly understood in natural systems where edges result from fire. Fire is an agent of edge creation and a globally important driver of biome distribution and community composition, yet little research has been undertaken to determine the ecological importance of edges in fire-prone landscapes and their implications for fauna. Many studies have examined the ecological importance of burnt and unburnt parts of a landscape, however very few have explicitly considered the interface or transition zone between these patches, or how they might influence the distribution, abundance or movement of species over local or landscape scales. With a currently limited knowledge of how fire edges influence ecosystem processes and an increasing prevalence of fire in the landscape, this research presents new and important information on the effects of fire edges on fauna populations. In this chapter I briefly summarise the key findings from chapters 2-5, discuss implications for management and highlight areas for future research.

EDGES EFFECTS IN FIRE-PRONE LANDSCAPES

The aim of this thesis was to understand the importance of fire edges in influencing ecological patterns and processes in flammable landscapes, and determine the influence of fire edges on fauna. In Chapter 2 I reviewed the literature on fire, fauna, and edge effects to summarise current knowledge of faunal responses to fire edges and identify key knowledge gaps. Fire- generated edge effects were found to differ from other edge types in several ways. The key findings are listed below.

1. Fire edges change over time. They are in a constant state of flux due to post-fire regeneration and the occurrence of new fires, and are expected to be softer and shorter lived than high contrast edges in other contexts (i.e. forest/farmland edges). 2. Fire edges occur at multiple spatial scales, and the strength of edge effects and species responses are expected to vary at different scales. 3. Edges created by fire never occur in isolation from other environmental patterns and processes. Fire-induced edge effects may be exacerbated, diminished or masked by the interaction of other factors. Fire edges are therefore inherently complex due to these interactions with other variables. Species responses to fire edges are expected to be similarly

complex, with negative effects potentially intensified by changed competition levels and predation risk. 4. Existing predictive models are unlikely to perform well in fire-prone systems because they do not consider the agent of edge creation and how that might influence the physical architecture of edges, and they fail to consider the importance of species traits (other than mobility) that may affect a species’ edge response.

In response to these findings I developed a new conceptual model for predicting how fire is likely to shape the physical properties of an edge and influence species’ edge responses. While this model was designed to predict the responses of fauna to fire edges, it is applicable in other disturbance contexts, as components of the disturbance regime can be modified to suit any edge creation process. Inclusion of biophysical properties and aspects of the disturbance regime will increase the power of the model to predict the location and architecture of fire edges. Consideration of species traits that go beyond mobility (e.g. habitat and diet requirements, adaptability to disturbance, susceptibility to other processes such as predation) will also help predict how different species are likely to respond to fire edges.

To direct future research efforts my conceptual model was designed as a Bayesian Network, a statistical framework capable of analysing complex environmental relationships between a range of variables, dealing with data gaps and generating testable hypotheses. Bayesian networks have strong predictive power when empirical datasets are incomplete and where potential drivers are correlated. The capacity of the BayesNet to deal with data gaps and to be used as a hypothesis generation tool will be particularly useful given the current paucity of information about faunal responses to fire edges.

HABITAT USE AT FIRE EDGES OVER TIME

A fire-induced change in structure represents a disruption to the continuity of resources, and is therefore expected to influence the activity patterns of some species, particularly those with a strong reliance on vegetation cover. In Chapter 3, I located fire edges within a series of prescribed burns, and assessed their influence on the activity patterns of ground-dwelling mammals. I used a space-for-time sampling design to examine temporal change at fire edges, and to determine if animal activity patterns were following temporal changes in vegetation structure. To the best of my knowledge, this research is one of the first true fire-edge studies, where fire- edges were not confounded by the location of other modified edges.

In this chapter I demonstrate that fire edges are ecologically important features in flammable landscapes, influencing the activity and selection of burnt edges for many species. The key findings are outlined below.

1. Understorey complexity was reduced on the burnt side of edges for up to two years, and followed a hump-shaped trend over time on the burnt side of the edge. Understorey complexity remained constant on the unburnt side. 2. Despite some activity on the burnt side of edges immediately after fire, small mammals were generally less active on burnt edges for up to 6-7 years after fire. In contrast, larger-bodied animals like mountain brushtail possums and swamp wallabies were more active on the burnt side of edges immediately post-fire, however patterns of high activity declined over time. Wombats had a hump shaped response to time since fire. Fox activity decreased with time since fire and this species was more active on the burnt side of edges than the unburnt, while cats did not respond to time since fire or edge. 3. In general, animal use of the burnt edge over time was not congruent with patterns of vegetation change. Most native species were using the burnt side of edges more than expected based on the availability of structurally complex vegetation in the first two years after fire. However, use of the burnt edge decreased for all native mammals three years post- fire, despite the peak in complex habitat at this time. Furthermore, use of the unburnt side of edges did not remain constant, despite the lack of fluctuation in the habitat structure. These results suggest that species responses to fire edges are not simply related to habitat structure and that other processes are occurring that are driving these patterns. For example, edge effects may extend beyond the burnt edge for some species and influence the use of unburnt habitat for several years. Both foxes and cats were using the burnt side of edges more than expected immediately after fire. Both species are known to exploit recently burnt and open habitats adjacent to dense vegetation, with reduced cover increasing access to structurally complex habitats, potentially providing better hunting opportunities. Increased predation pressure on the burnt side of edges may also be influencing observed patterns in the native species.

The results from this chapter provide several important insights for management of fire-prone landscapes. Increased predation pressure at fire edges could have negative consequences for native fauna if changes in habitat structure results in increasing predation rates. Targeted use of

89 burnt/unburnt edges by invasive predators may also greatly diminish the protective value of small unburned patches of vegetation for native fauna.

We compared resource use (the activity index) to resource availability (understorey complexity index) using a resource selection index (Vanderploeg & Scavia 1979). To my knowledge, this technique has not yet been used in this manner. I found that this was an effective method for comparing the congruence between two different datasets, and may be useful for future studies where researchers wish to compare species response with other explanatory variables.

Habitat structure is often used as a surrogate for animal biodiversity in flammable systems (Haslem et al. 2011). However, our results show that animal use of burnt/unburnt edges did not closely follow patterns of habitat availability, suggesting that a measure of habitat complexity after fire may not be a good predictor of animal occurrence. Future fire management must therefore acknowledge the potential for fire to affect fauna and vegetation differently. Predicting processes that are influenced by edges (such as post-fire recolonisation) will be challenging, but consideration of other factors like predation and competition will improve our understanding of the ecological importance of fire edges and their patterns of temporal change. The results from this study suggest that fire edges are important and unique landscape features, and will require further attention and consideration for future fire management planning.

CAMERA TRAP SURVEY DESIGN

Camera number and deployment time are two critical aspects of study design for every camera trapping survey. The trade-off between these two factors will likely affect data quality, but there is currently little information about their relative influence. In Chapter 4, I assessed the trade-off between deploying more detection units or extending the length of the sampling period on two frequently assessed variables in camera trapping studies – species richness and detection probability. I used a bootstrapping simulation procedure to iteratively sample each possible combination of detection units (1-6 camera pairs) and camera days (1-34), to investigate the effect of both factors on species richness and detection probability. The key findings were:

1. Increasing the number of detection units deployed per site resulted in a substantial increase in the species richness index. 2. For overall detectability, multiple detection units in combination with longer deployment time were necessary to detect a high proportion of the species present. Increasing the number of detection units or increasing deployment time (or both) resulted in high overall detection probability for more detectible species, but multiple detection units

were always needed to achieve high detection probability in a reasonable timeframe (<50 days) for less detectible species.

Camera trapping is a widely used survey technique, and these findings will inform decisions about the number of detection units and deployment time in future camera trapping surveys. Furthermore, the results from this study will also have important implications for the study of fire edges. Studying faunal responses to edges is challenging, particularly when surveying at fine spatial scales. These results show that deploying multiple cameras per site is an effective method for increasing species richness, and this may guide future camera trapping studies at fire edges.

RESOURCE SELECTION AFTER HIGH SEVERITY FIRE

Despite the propensity for large uncontrolled wildfires to affect large areas and burn through multiple vegetation types, little research has been conducted into how post-fire resource selection might vary between different forest types. Determining which resources are important for species to persist in post-fire environments will be important for incorporating the needs of biodiversity into fire management planning. In Chapter 5 I followed a use-versus-availability study design (Manly et al. 2007), where used and available locations were compared to environmental variables to predict resource use by the mountain brushtail possum in two forest types affected by high severity wildfire in 2009. With the help of an electronics engineer, I designed and made customised wildlife trackers for mountain brushtail possums. In this chapter I demonstrate that resource availability differed between wet and dry forest eight years after wildfire and that selection of several resources by the mountain brushtail possum was influenced by both forest type and sex.

The key findings were:

1. Resource availability differed between wet and dry forest eight years after a high severity wildfire, with more dead trees and wattles (food and denning resources) in wet forest, compared to the dry forest, which was characterised by more vegetation cover and live trees (facilitating increased arboreal movement). 2. Selection of large dead trees and wattles were influenced by both forest type and sex, suggesting that patterns of selection are driven by both resource availability and sex-related differences in requirements for denning and food resources.

3. In contrast, selection for dense vegetation cover at the ground, midstorey and canopy level only differed between forest type, suggesting that both sexes respond similarly to vegetation cover and perceived risk.

Differential resource selection is one of the principal processes that enable individuals to coexist. Studies that account for the spatial variability of resources and species demographic status are likely to provide a better understanding of how major fires influence animals resource selection.

KNOWLEDGE GAPS AND FURTHER RESEARCH

This thesis adds to the body of knowledge on the ecological importance of fire edges and their implications for fauna. However, the study of edges in fire-prone systems is still in its infancy. While my conceptual model provides a theoretical understanding of edge creation in flammable landscapes and associated implications for fauna, few data quantifying these processes currently exist. Conversion of the conceptual model into a BayesNet will enable these conceptual advances to be effectively combined with new data as fire-edge research is conducted. The capacity of the model to deal with data gaps and to be used as a hypothesis generation tool will be particularly useful given the current paucity of information about faunal responses to fire edges.

There are, therefore, many avenues for future research, including:

1. Understanding fire as an agent of edge creation The agent of edge creation can strongly influence ecological patterns and processes, but few edge-related studies have considered how the process of creation influences edge effects. Understanding how fires interact with biophysical properties to create edges will be an important component of future fire-edge research.

2. Modelling the spatio-temporal flux of fire edges in flammable landscapes Modelling the effect of fire cycles and plant regeneration rates on the distribution, abundance and architecture of edges will be an important precursor to understanding fire-induced edge effects more broadly, particularly at landscape scales. Understanding how permeability at single edges interact to influence landscape-scale structural and functional connectivity will be important for the conservation of biodiversity in fire-prone systems, particularly when considering landscapes that contain complex and varied fire histories.

92 not yet been studied in flammable ecosystems. Better knowledge of these relationships will aid our understanding of the underlying mechanisms that drive species and community edge responses in fire-prone landscapes.

4. Understanding the implications of multiple interacting edges

In modified landscapes, the magnitude and extent of edge effects increase at locations where multiple edges are present (Fletcher 2005). This is likely to be similar in fire-prone landscapes, where multiple fires result in an intricate network of fire edges, each with a unique trajectory of temporal change. Fire managers often utilise pre-existing fire edges as boundaries for future burns. However, multiple prescribed burns adjacent to each other may result in large parts of a landscape that are influenced by fire-induced edge effects. Therefore, it will be important to understand how multiple fire edges affect fauna, and for fire managers to consider the location of pre-existing fire edges in the landscape when planning future fuel reduction burns.

5. Understanding how edge effects vary in different locations and through time

Fire affects flora and fauna communities differently and it is likely that fire-indicted edge effects will vary in different regions for different taxa. Further research into edges resulting from wildfires, prescribed fires and fuel breaks, for example, will be important for understanding edge effects in flammable landscapes more broadly. The study of fire edges at different spatial scales will also be important.

To conclude, the research presented in this thesis is timely and important as the prevalence of fire in many terrestrial systems is increasing. Mechanistic approaches based on the strength of habitat associations and resource availability may help to clarify the nature and strength of edge effects in fire-prone landscapes and improve predictive models. A better understanding of fire edges in flammable systems and species resource selection in fire-affected landscapes will better enable managers to integrate biodiversity conservation into fire management planning.

7 REFERENCE LIST

Adams, M. & Attiwill, P. (1984) Role of Acacia spp. in nutrient balance and cycling in regenerating Eucalyptus regnans F. Muell. forests. I. Temporal changes in biomass and nutrient content. Australian Journal of Botany, 32, 205-215.

Ahumada, J.A., Hurtado, J. & Lizcano, D. (2013) Monitoring the status and trends of tropical forest terrestrial vertebrate communities from camera trap data: a tool for conservation. PLoS One, 8, e73707.

Alexander, J.D., Seavy, N.E., Ralph, C.J. & Hogoboom, B. (2006) Vegetation and topographical correlates of fire severity from two fires in the Klamath-Siskiyou region of Oregon and California. International Journal of Wildland Fire, 15, 237-245.

Anderson, E.M. & Boutin, S. (2002) Edge effects on survival and behaviour of juvenile red squirrels (Tamiasciurus hudsonicus). Canadian Journal of Zoology, 80, 1038-1046.

Ardia, D.R. & Bildstein, K.L. (1997) Sex-related differences in habitat selection in wintering American kestrels, Falco sparverius. Animal Behaviour, 53, 1305-1311.

Arnold, K.T., Murphy, N.P. & Gibb, H. (2017) Post-fire recovery of litter detritivores is limited by distance from burn edge. Austral Ecology, 42, 94-102.

Arthur, A.D., Catling, P.C. & Reid, A. (2012) Relative influence of habitat structure, species interactions and rainfall on the post-fire population dynamics of ground-dwelling vertebrates. Austral Ecology, 37, 958-970.

Auld, T.D. & O'Connell, M.A. (1991) Predicting patterns of post-fire germination in 35 eastern Australian Fabaceae. Austral Ecology, 16, 53-70.

Austin, C., Tuft, K., Ramp, D., Cremona, T. & Webb, J.K. (2017) Bait preference for remote camera trap studies of the endangered northern quoll (Dasyurus hallucatus). Australian Mammalogy, 39, 72-77.

Baguette, M. & Van Dyck, H. (2007) Landscape connectivity and animal behavior: functional grain as a key determinant for dispersal. Landscape Ecology, 22, 1117-1129.

Balme, G., Hunter, L. & Robinson, H. (2014) Baited camera-trap surveys–Marginally more precise but at what cost? A response to du Preez et al.(2014). Biological Conservation, 179, 144-145.

Banks, S.C., Dujardin, M., McBurney, L., Blair, D., Barker, M. & Lindenmayer, D.B. (2011a) Starting points for small mammal population recovery after wildfire: recolonisation or residual populations? Oikos, 120, 26-37.

Banks, S.C., Knight, E.J., McBurney, L., Blair, D. & Lindenmayer, D.B. (2011b) The effects of wildfire on mortality and resources for an arboreal marsupial: resilience to fire events but susceptibility to fire regime change. PLoS One, 6, e22952.

Barton, K. (2013) MuMIn: multi model inference. R package version 1.9.13.

Bates, D., Mächler, M., Bolker, B. & Walker, S. (2015) Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67.

Bates, D., Maechler, M. & Bolker, B. (2011) lme4: Linear mixed-effeccts models using S4 classes. R package version 0.999375-42.

Bélisle, M. (2005) Measuring landscape connectivity: the challenge of behavioral landscape ecology. Ecology, 86, 1988-1995.

Bell, D.T. (1999) The process of germination in Australian species. Australian Journal of Botany, 47, 475-517.

Berry, L., Lindenmayer, D., Dennis, T., Driscoll, D. & Banks, S. (2017) Fire severity alters spatio–temporal movements and habitat utilisation by an arboreal marsupial, the mountain brushtail possum (Trichosurus cunninghami). International Journal of Wildland Fire, 25, 1291-1302.

Beyer, H.L., Haydon, D.T., Morales, J.M., Frair, J.L., Hebblewhite, M., Mitchell, M. & Matthiopoulos, J. (2010) The interpretation of habitat preference metrics under use–availability designs. Philosophical Transactions of the Royal Society B: Biological Sciences, 365, 2245-2254.

Bjørneraas, K., Van Moorter, B., Rolandsen, C.M. & Herfindal, I. (2010) Screening global positioning system location data for errors using animal movement characteristics. Journal of Wildlife Management, 74, 1361-1366.

Bond, W.J. & Keeley, J.E. (2005) Fire as a global ‘herbivore’: the ecology and evolution of flammable ecosystems. Trends in Ecology & Evolution, 20, 387-394.

Borchert, M. & Borchert, S.M. (2013) Small mammal use of the burn perimeter following a chaparral wildfire in southern California. Bulletin, Southern California Academy of Sciences, 112, 63-73.

Bowler, M.T., Tobler, M.W., Endress, B.A., Gilmore, M.P. & Anderson, M.J. (2017) Estimating mammalian species richness and occupancy in tropical forest canopies with arboreal camera traps. Remote Sensing in Ecology and Conservation, 3, 146-157.

Bowman, D.M., Balch, J., Artaxo, P., Bond, W.J., Cochrane, M.A., D’Antonio, C.M., DeFries, R., Johnston, F.H., Keeley, J.E. & Krawchuk, M.A. (2011) The human dimension of fire regimes on Earth. Journal of biogeography, 38, 2223-2236.

Bowman, D.M., Balch, J.K., Artaxo, P., Bond, W.J., Carlson, J.M., Cochrane, M.A., D’Antonio, C.M., DeFries, R.S., Doyle, J.C. & Harrison, S.P. (2009) Fire in the Earth system. Science, 324, 481-484.

Bowyer, R.T. & Kie, J.G. (2004) Effects of foraging activity on sexual segregation in mule deer. Journal of Mammalogy, 85, 498-504.

Boyce, M.S. & McDonald, L.L. (1999) Relating populations to habitats using resource selection functions. Trends in Ecology & Evolution, 14, 268-272.

Bradstock, R., Bedward, M., Gill, A. & Cohn, J. (2005) Which mosaic? A landscape ecological approach for evaluating interactions between fire regimes, habitat and animals. Wildlife Research, 32, 409-423.

Bradstock, R., Keith, D. & Auld, T. (1995) Fire and conservation: imperatives and constraints on managing for diversity. Conserving biodiversity: threats and solutions, 323, 333.

Bradstock, R.A. (2008) Effects of large fires on biodiversity in south-eastern Australia: disaster or template for diversity? International Journal of Wildland Fire, 17, 809-822.

Bradstock, R.A. (2010) A biogeographic model of fire regimes in Australia: current and future implications. Global Ecology and Biogeography, 19, 145-158.

Bradstock, R.A., Hammill, K., Collins, L. & Price, O. (2010) Effects of weather, fuel and terrain on fire severity in topographically diverse landscapes of south-eastern Australia. Landscape Ecology, 25, 607-619.

Brockett, B., Biggs, H. & Van Wilgen, B. (2001) A patch mosaic burning system for conservation areas in southern African savannas. International Journal of Wildland Fire, 10, 169-183.

Broncano, M.J. & Retana, J. (2004) Topography and forest composition affecting the variability in fire severity and post-fire regeneration occurring after a large fire in the Mediterranean basin. International Journal of Wildland Fire, 13, 209-216.

Burnham, K.P. & Anderson, D.R. (2004) Multimodel inference understanding AIC and BIC in model selection. Sociological Methods and Research, 33, 261-304.

Burrows, N. & Wardell-Johnson, G. (2004) Implementing fire mosaics to promote biodiversity and reduce the severity of wildfires in south-west Australian ecosystems. 11th Annual AFAC Conference and Inaugural Bushfire CRC Conference.

Cadenasso, M.L., Pickett, S.T., Weathers, K.C., Bell, S.S., Benning, T.L., Carreiro, M.M. & Dawson, T.E. (2003a) An interdisciplinary and synthetic approach to ecological boundaries. BioScience, 53, 717-722.

Cadenasso, M.L., Pickett, S.T., Weathers, K.C. & Jones, C.G. (2003b) A framework for a theory of ecological boundaries. BioScience, 53, 750-758.

Cary, G.J. & Banks, J.C. (2000) Fire regime sensitivity to global climate change: an Australian perspective. Biomass Burning and Its Inter-Relationships with the Climate System, pp. 233-246. Springer, Dordrecht.

Cary, G.J., Bradstock, R.A., Gill, A.M. & Williams, R.J. (2012) Global change and fire regimes in Australia. CSIRO Publising, Collingwood, Victoria.

Catchpole, W. (2002) Fire properties and burn patterns in heterogeneous landscapes. Cambridge University Press, Cambridge.

Catling, P., Burt, R. & Forrester, R. (2000) Models of the distribution and abundance of ground- dwelling mammals in the eucalypt forests of north-eastern New South Wales in relation to habitat variables. Wildlife Research, 27, 639-654.

Catling, P., Coops, N. & Burt, R. (2001) The distribution and abundance of ground-dwelling mammals in relation to time since wildfire and vegetation structure in south-eastern Australia. Wildlife Research, 28, 555-565.

Chia, E.K., Bassett, M., Nimmo, D.G., Leonard, S.W., Ritchie, E.G., Clarke, M.F. & Bennett, A.F. (2015) Fire severity and fire-induced landscape heterogeneity affect arboreal mammals in fire-prone forests. Ecosphere, 6, art190.

Claridge, A. & Lindenmayer, D. (1998) Consumption of hypogeous fungi by the mountain brushtail possum (Trichosurus caninus) in eastern Australia. Mycological Research, 102, 269-272.

Claridge, A., Trappe, J. & Claridge, D. (2001) Mycophagy by the swamp wallaby (Wallabia bicolor). Wildlife Research, 28, 643-645.

Claridge, A.W., Tennant, P., Chick, R. & Barry, S.C. (2008) Factors influencing the occurrence of small ground-dwelling mammals in southeastern mainland Australia. Journal of Mammalogy, 89, 916-923.

Claridge, A.W., Trappe, J.M. & Hansen, K. (2009) Do fungi have a role as soil stabilizers and remediators after forest fire? Forest Ecology and Management, 257, 1063-1069.

Clarke, K. & Gorley, R. (2006) PRIMER v6: user manual/tutorial (Plymouth routines in multivariate ecological research). Plymouth: Primer-E Ltd.

Cochrane, M.A. & Laurance, W.F. (2002) Fire as a large-scale edge effect in Amazonian forests. Journal of Tropical Ecology, 18, 311-325.

Coleman, B.T. & Hill, R.A. (2014) Living in a landscape of fear: the impact of predation, resource availability and habitat structure on primate range use. Animal Behaviour, 88, 165-173.

Collins, L., Bradstock, R.A., Tasker, E.M. & Whelan, R.J. (2012) Can gullies preserve complex forest structure in frequently burnt landscapes? Biological Conservation, 153, 177-186.

Comfort, E.J., Clark, D.A., Anthony, R.G., Bailey, J. & Betts, M.G. (2016) Quantifying edges as gradients at multiple scales improves habitat selection models for northern spotted owl. Landscape Ecology, 31, 1227-1240.

Conner, L.M., Castleberry, S.B. & Derrick, A.M. (2011) Effects of mesopredators and prescribed fire on hispid cotton rat survival and cause‐specific mortality. Journal of Wildlife Management, 75, 938-944.

Corti, P. & Shackleton, D.M. (2002) Relationship between predation-risk factors and sexual segregation in Dall's sheep (Ovis dalli dalli). Canadian Journal of Zoology, 80, 2108-2117.

Covert-Bratland, K.A., Block, W.M. & Theimer, T.C. (2006) Hairy woodpecker winter ecology in ponderosa pine forests representing different ages since wildfire. Journal of Wildlife Management, 70, 1379-1392.

Craig, M.D., Stokes, V.L., Hardy, G.E.S. & Hobbs, R.J. (2015) Edge effects across boundaries between natural and restored jarrah (Eucalyptus marginata) forests in south-western Australia. Austral Ecology, 40, 186-197.

Cunningham, T. & Cremer, K. (1965) Control of the understorey in wet eucalypt forests. Australian Forestry, 29, 4-14. da Rosa, C.A., Secco, H., Carvalho, N., Maia, A.C. & Bager, A. (2018) Edge effects on small mammals: Differences between arboreal and ground‐dwelling species living near roads in Brazilian fragmented landscapes. Austral Ecology, 43, 117-126.

De Bondi, N., White, J.G., Stevens, M. & Cooke, R. (2010) A comparison of the effectiveness of camera trapping and live trapping for sampling terrestrial small-mammal communities. Wildlife Research, 37, 456-465.

Di Stefano, J., Anson, J.A., York, A., Greenfield, A., Coulson, G., Berman, A. & Bladen, M. (2007) Interactions between timber harvesting and swamp wallabies (Wallabia bicolor): space use, density and browsing impact. Forest Ecology and Management, 253, 128-137.

Di Stefano, J., McCarthy, M.A., York, A., Duff, T.J., Slingo, J. & Christie, F. (2013) Defining vegetation age class distributions for multispecies conservation in fire-prone landscapes. Biological Conservation, 166, 111-117.

Dickman, C.R. & Steeves, T.E. (2004) Use of habitat by mammals in eastern Australian forests: are common species important in forest management. Conservation of Australia’s forest fauna, 761- 773.

Diffendorfer, J., Fleming, G.M., Tremor, S., Spencer, W. & Beyers, J.L. (2012) The role of fire severity, distance from fire perimeter and vegetation on post-fire recovery of small-mammal communities in chaparral. International Journal of Wildland Fire, 21, 436-448.

Doherty, T.S., Davis, R.A. & van Etten, E.J. (2015) A game of cat-and-mouse: microhabitat influences rodent foraging in recently burnt but not long unburnt shrublands. Journal of Mammalogy, 96, 324-331.

100

Doumas, S.L. & Koprowski, J.L. (2013) Effect of heterogeneity in burn severity on Mexican fox squirrels following the return of fire. International Journal of Wildland Fire, 22, 405.

Dreisig, H. (1984) Control of body temperature in shuttling ectotherms. Journal of Thermal Biology, 9, 229-233.

Driscoll, D.A., Lindenmayer, D.B., Bennett, A.F., Bode, M., Bradstock, R.A., Cary, G.J., Clarke, M.F., Dexter, N., Fensham, R. & Friend, G. (2010) Fire management for biodiversity conservation: key research questions and our capacity to answer them. Biological Conservation, 143, 1928-1939.

Driscoll, M.J.L. & Donovan, T.M. (2004) Landscape context moderates edge effects: Nesting success of wood thrushes in central New York. Conservation Biology, 18, 1330-1338.

Eby, S.L., Anderson, T.M., Mayemba, E.P. & Ritchie, M.E. (2014) The effect of fire on habitat selection of mammalian herbivores: the role of body size and vegetation characteristics. Journal of Animal Ecology, 83, 1196-1205.

Elia, M., Lafortezza, R., Tarasco, E. & Sanesi, G. (2016) Response of beetle communities five years after wildfire in Mediterranean ecosystems. Redia-Giornale Di Zoologia, 99, 107-116.

ESRI, A.D. (2011) Release 10. Documentation Manual. Redlands, CA, Environmental Systems Research Institute.

Eyes, S.A., Roberts, S.L. & Johnson, M.D. (2017) California Spotted Owl (Strix occidentalis occidentalis) habitat use patterns in a burned landscape. Condor, 119, 375-388.

Fagan, W.F., Fortin, M.-J. & Soykan, C. (2003) Integrating edge detection and dynamic modeling in quantitative analyses of ecological boundaries. BioScience, 53, 730-738.

Fahrig, L. (2003) Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evolution, and Systematics, 34, 487-515.

Fahrig, L. (2007) Non‐optimal animal movement in human‐altered landscapes. Functional Ecology, 21, 1003-1015.

Fairman, T.A., Nitschke, C.R. & Bennett, L.T. (2016) Too much, too soon? A review of the effects of increasing wildfire frequency on tree mortality and regeneration in temperate eucalypt forests. International Journal of Wildland Fire, 25, 831-848.

101

Fernandes, P.M., Davies, G.M., Ascoli, D., Fernández, C., Moreira, F., Rigolot, E., Stoof, C.R., Vega, J.A. & Molina, D. (2013) Prescribed burning in southern Europe: developing fire management in a dynamic landscape. Frontiers in Ecology and the Environment, 11, e4-e14.

Figueiredo, M.d.S.L. & dos Santos Fernandez, F.A. (2004) Contrasting effects of fire on populations of two small rodent species in fragments of Atlantic Forest in Brazil. Journal of Tropical Ecology, 20, 225-228.

Fischer, M., Parkins, K., Maizels, K., Sutherland, D.R., Allan, B.M., Coulson, G. & Di Stefano, J. (2018) Biotelemetry marches on: A cost-effective GPS device for monitoring terrestrial wildlife. PLoS One, 13, e0199617.

Fisher, R.J. & Wiebe, K.L. (2006) Nest site attributes and temporal patterns of northern flicker nest loss: effects of predation and competition. Oecologia, 147, 744-753.

Flannigan, M.D., Krawchuk, M.A., de Groot, W.J., Wotton, B.M. & Gowman, L.M. (2009) Implications of changing climate for global wildland fire. International Journal of Wildland Fire, 18, 483-507.

Fletcher, R.J. (2005) Multiple edge effects and their implications in fragmented landscapes. Journal of Animal Ecology, 74, 342-352.

Fordyce, A., Hradsky, B.A., Ritchie, E.G. & Di Stefano, J. (2015) Fire affects microhabitat selection, movement patterns, and body condition of an Australian rodent (Rattus fuscipes). Journal of Mammalogy, 97, 102-111.

Fox, B.J. (1982) Fire and mammalian secondary succession in an Australian coastal heath. Ecology, 63, 1332-1341.

Fox, B.J., Taylor, J.E. & Thompson, P.T. (2003) Experimental manipulation of habitat structure: a retrogression of the small mammal succession. Journal of Animal Ecology, 72, 927-940.

Galvez, N., Guillera-Arroita, G., Morgan, B.J.T. & Davies, Z.G. (2016) Cost-efficient effort allocation for camera-trap occupancy surveys of mammals. Biological Conservation, 204, 350-359.

Geary, W.L., Ritchie, E.G., Lawton, J.A., Healey, T.R. & Nimmo, D.G. (2018) Incorporating disturbance into trophic ecology: fire history shapes mesopredator suppression by an apex predator. Journal of Applied Ecology, 1-10.

102

Gelin, M.L., Branch, L.C., Thornton, D.H., Novaro, A.J., Gould, M.J. & Caragiulo, A. (2017) Response of pumas (Puma concolor) to migration of their primary prey in Patagonia. PLoS One, 12, e0188877.

Gill, A.M. (1975) Fire and the Australian flora: a review. Australian Forestry, 38, 4-25.

Glutton-Brock, T. & Vincent, A.C. (1991a) Sexual selection and the potential reproductive rates of males and females. Nature, 351, 58-60.

Glutton-Brock, T. & Vincent, A.C. (1991b) Sexual selection and the potential reproductive rates of males and females. Nature, 351, 58.

Goosem, M. (2001) Effects of tropical rainforest roads on small mammals: inhibition of crossing movements. Wildlife Research, 28, 351-364.

Gorini, L., Linnell, J.D., May, R., Panzacchi, M., Boitani, L., Odden, M. & Nilsen, E. (2012) Habitat heterogeneity and mammalian predator–prey interactions. Mammal review, 42, 55-77.

Greenville, A.C., Wardle, G.M. & Dickman, C.R. (2017) Desert mammal populations are limited by introduced predators rather than future climate change. Royal Society open science, 4, 170384.

Griffiths, A.D. & Brook, B.W. (2014) Effect of fire on small mammals: a systematic review. International Journal of Wildland Fire, 23, 1034-1043.

Griffiths, A.D., Garnett, S.T. & Brook, B.W. (2015) Fire frequency matters more than fire size: Testing the pyrodiversity–biodiversity paradigm for at-risk small mammals in an Australian tropical savanna. Biological Conservation, 186, 337-346.

Harper, K.A., Macdonald, S.E., Burton, P.J., Chen, J., Brosofske, K.D., Saunders, S.C., Euskirchen, E.S., Roberts, D., Jaiteh, M.S. & Essen, P.-A. (2005) Edge influence on forest structure and composition in fragmented landscapes. Conservation Biology, 19, 768-782.

Haslem, A., Avitabile, S.C., Taylor, R.S., Kelly, L.T., Watson, S.J., Nimmo, D.G., Kenny, S.A., Callister, K.E., Spence-Bailey, L.M. & Bennett, A.F. (2012) Time-since-fire and inter-fire interval influence hollow availability for fauna in a fire-prone system. Biological Conservation, 152, 212-221.

Haslem, A., Kelly, L.T., Nimmo, D.G., Watson, S.J., Kenny, S.A., Taylor, R.S., Avitabile, S.C., Callister, K.E., Spence‐Bailey, L.M. & Clarke, M.F. (2011) Habitat or fuel? Implications of long‐

103 term, post‐fire dynamics for the development of key resources for fauna and fire. Journal of Applied Ecology, 48, 247-256.

Hof, A.R., Snellenberg, J. & Bright, P.W. (2012) Food or fear? Predation risk mediates edge refuging in an insectivorous mammal. Animal Behaviour, 83, 1099-1106.

Hovick, T.J., Allred, B.W., Elmore, R.D., Fuhlendorf, S.D., Hamilton, R.G. & Breland, A. (2015) Dynamic disturbance processes create dynamic lek site selection in a prairie grouse. PLoS One, 10, e0137882.

Hradsky, B.A., Mildwaters, C., Ritchie, E.G., Christie, F. & Di Stefano, J. (2017) Responses of invasive predators and native prey to a prescribed forest fire. Journal of Mammalogy, 98, 835-847.

Hummel, S., Meyer, L., Hackländer, K. & Weber, D. (2017) Activity of potential predators of European hare (Lepus europaeus) leverets and ground-nesting birds in wildflower strips. European Journal of Wildlife Research, 63, 102.

Hunt, M.A., Unwin, G.L. & Beadle, C.L. (1999) Effects of naturally regenerated Acacia dealbata on the productivity of a Eucalyptus nitens plantation in Tasmania, Australia. Forest Ecology and Management, 117, 75-85.

Johnson, C. & Bayliss, P. (1981) Habitat selection by sex, age and reproductive class in the red kangaroo, Macropus rufus, in western New South Wales. Wildlife Research, 8, 465-474.

Johnson, C.J., Nielsen, S.E., Merrill, E.H., McDonald, T.L. & Boyce, M.S. (2006) Resource selection functions based on use-availability data: Theoretical motivation and evaluation methods. Journal of Wildlife Management, 70, 347-357.

Jonasson, S. (1983) The point intercept method for non-destructive estimation of biomass. Phytocoenologia, 11, 385-388.

Karanth, K.U. & Nichols, J.D. (1998) Estimation of tiger densities in India using photographic captures and recaptures. Ecology, 79, 2852-2862.

Keeley, J.E., Fotheringham, C. & Baer‐Keeley, M. (2005) Factors affecting plant diversity during post‐fire recovery and succession of mediterranean‐climate shrublands in California, USA. Diversity and Distributions, 11, 525-537.

104

Kelly, L.T., Bennett, A.F., Clarke, M.F. & McCarthy, M.A. (2015) Optimal fire histories for biodiversity conservation. Conservation Biology, 29, 473-481.

Kelly, M.J. (2008) Design, evaluate, refine: camera trap studies for elusive species. Animal Conservation, 11, 182-184.

Kelly, M.J., Noss, A.J., Di Bitetti, M.S., Maffei, L., Arispe, R.L., Paviolo, A., De Angelo, C.D. & Di Blanco, Y.E. (2008) Estimating puma densities from camera trapping across three study sites: Bolivia, Argentina, and Belize. Journal of Mammalogy, 89, 408-418.

Kerle, J. (1984) Variation in the ecology of Trichosurus: its adaptive significance. Possums and gliders, 115-128.

Kingston, S.R. & Morris, D.W. (2000) Voles looking for an edge: habitat selection across forest ecotones. Canadian Journal of Zoology, 78, 2174-2183.

Laurance, W.F., Didham, R.K. & Power, M.E. (2001) Ecological boundaries: a search for synthesis. Trends in Ecology & Evolution, 16, 70-71.

Laurance, W.F., Goosem, M. & Laurance, S.G. (2009) Impacts of roads and linear clearings on tropical forests. Trends in Ecology & Evolution, 24, 659-669.

Leahy, L., Legge, S.M., Tuft, K., McGregor, H.W., Barmuta, L.A., Jones, M.E. & Johnson, C.N. (2016) Amplified predation after fire suppresses rodent populations in Australia’s tropical savannas. Wildlife Research, 42, 705-716.

Lees, A.C. & Peres, C.A. (2009) Gap‐crossing movements predict species occupancy in Amazonian forest fragments. Oikos, 118, 280-290.

Lehman, S.M., Rajaonson, A. & Day, S. (2006) Edge effects and their influence on lemur density and distribution in southeast Madagascar. American Journal of Physical Anthropology, 129, 232-241.

Lehtinen, R.M., Ramanamanjato, J.-B. & Raveloarison, J.G. (2003) Edge effects and extinction proneness in a herpetofauna from Madagascar. Biodiversity and Conservation, 12, 1357-1370.

Leonard, S.W., Bennett, A.F. & Clarke, M.F. (2014) Determinants of the occurrence of unburnt forest patches: Potential biotic refuges within a large, intense wildfire in south-eastern Australia. Forest Ecology and Management, 314, 85-93.

105

Lessa, I.C.M., Ferreguetti, Á.C., Kajin, M., Dickman, C.R. & Bergallo, H.G. (2017) You can’t run but you can hide: the negative influence of human presence on mid-sized mammals on an Atlantic island. Journal of Coastal Conservation, 21, 829-836.

Li, J. & Heap, A.D. (2011) A review of comparative studies of spatial interpolation methods in environmental sciences: performance and impact factors. Ecological Informatics, 6, 228-241.

Lindenmayer, D., Blanchard, W., McBurney, L., Blair, D., Banks, S., Driscoll, D., Smith, A. & Gill, A. (2013) Fire severity and landscape context effects on arboreal marsupials. Biological Conservation, 167, 137-148.

Lindenmayer, D., Cunningham, R., Donnelly, C. & Franklin, J. (2000) Structural features of old- growth Australian montane ash forests. Forest Ecology and Management, 134, 189-204.

Lindenmayer, D., Norton, T. & Tanton, M. (1990) Differences between wildfire and clearfelling on the structure of montane ash forests of Victoria and their implications for fauna dependent on tree hollows. Australian Forestry, 53, 61-68.

Lindenmayer, D., Welsh, A., Donnelly, C. & Cunningham, R. (1996) Use of nest trees by the mountain brushtail possum (Trichosurus caninus)(Phalangeridae: Marsupialia). II. Characteristics of occupied trees. Wildlife Research, 23, 531-545.

Lindenmayer, D.B., Cunningham, R.B. & Donnelly, C.F. (1997) Decay and collapse of trees with hollows in eastern Australian forests: impacts on arboreal marsupials. Ecological applications, 7, 625-641.

Lindenmayer, D.B., Hobbs, R.J., Likens, G.E., Krebs, C.J. & Banks, S.C. (2011) Newly discovered landscape traps produce regime shifts in wet forests. Proceedings of the National Academy of Sciences, 108, 15887-15891.

Linkie, M., Guillera-Arroita, G., Smith, J., Ario, A., Bertagnolio, G., Cheong, F., Clements, G.R., Dinata, Y., Duangchantrasiri, S. & Fredriksson, G. (2013) Cryptic mammals caught on camera: assessing the utility of range wide camera trap data for conserving the endangered Asian tapir. Biological Conservation, 162, 107-115.

Loe, L.E., Irvine, R.J., Bonenfant, C., Stien, A., Langvatn, R., Albon, S.D., Mysterud, A. & Stenseth, N.C. (2006) Testing five hypotheses of sexual segregation in an arctic ungulate. Journal of Animal Ecology, 75, 485-496.

106

MacKenzie, D.I. & Royle, J.A. (2005) Designing occupancy studies: general advice and allocating survey effort. Journal of Applied Ecology, 42, 1105-1114.

Main, M.B. (2008) Reconciling competing ecological explanations for sexual segregation in ungulates. Ecology, 89, 693-704.

Manly, B., McDonald, L., Thomas, D.L., McDonald, T.L. & Erickson, W.P. (2007) Resource selection by animals: statistical design and analysis for field studies. Kluwer Academic Publishers, Dordrect.

Martin, J.K. (2006) Den-use and home-range characteristics of bobucks, Trichosurus cunninghami, resident in a forest patch. Australian Journal of Zoology, 54, 225-234.

Martin, J.K. & Martin, A.A. (2007) Resource distribution influences mating system in the bobuck (Trichosurus cunninghami: Marsupialia). Oecologia, 154, 227-236.

McElhinny, C., Gibbons, P., Brack, C. & Bauhus, J. (2006) Fauna-habitat relationships: a basis for identifying key stand structural attributes in temperate Australian eucalypt forests and woodlands. Pacific Conservation Biology, 12, 89-110.

McGregor, H.W., Legge, S., Jones, M.E. & Johnson, C.N. (2014) Landscape management of fire and grazing regimes alters the fine-scale habitat utilisation by feral cats. PLoS One, 9, e109097.

McGregor, H.W., Legge, S., Jones, M.E. & Johnson, C.N. (2016) Extraterritorial hunting expeditions to intense fire scars by feral cats. Scientific Reports, 6, 22559.

Meek, P., Ballard, G., Claridge, A., Kays, R., Moseby, K., O’brien, T., O’Connell, A., Sanderson, J., Swann, D. & Tobler, M. (2014) Recommended guiding principles for reporting on camera trapping research. Biodiversity and Conservation, 23, 2321-2343.

Meek, P.D. & Pittet, A. (2012) User-based design specifications for the ultimate camera trap for wildlife research. Wildlife Research, 39, 649-660.

Meek, P.D. & Vernes, K. (2016) Can camera trapping be used to accurately survey and monitor the Hastings River mouse (Pseudomys oralis)? Australian Mammalogy, 38, 44-51.

Meers, B.T. & Adams, R. (2003) The impact of grazing by Eastern Grey Kangaroos (Macropus giganteus) on vegetation recovery after fire at Reef Hills Regional Park, Victoria. Ecological Management and Restoration, 4, 126-132.

107

Mella, V., Banks, P. & McArthur, C. (2014) Negotiating multiple cues of predation risk in a landscape of fear: what scares free‐ranging brushtail possums? Journal of Zoology, 294, 22-30.

Mendes-Oliveira, A.C., dos Santos, P.G.P., de Carvalho, O., Montag, L.F.D., de Lima, R.C.S., de Maria, S.L.S. & Rossi, R.V. (2012) Edge effects and the impact of wildfires on populations of small non-volant mammals in the forest-savanna transition zone in Southern Amazonia. Biota Neotropica, 12, 57-63.

Menkhorst, P. & Knight, F. (2004) Field guide to the mammals of Australia. Oxford University Press, Melbourne.

Monamy, V. & Fox, B. (2000) Small mammal succession is determined by vegetation density rather than time elapsed since disturbance. Austral Ecology, 25, 580-587.

Morris, D.W. (2003) Toward an ecological synthesis: a case for habitat selection. Oecologia, 136, 1- 13.

Murcia, C. (1995) Edge effects in fragmented forests: implications for conservation. Trends in Ecology & Evolution, 10, 58-62.

Murray, K.A., Rosauer, D., McCallum, H. & Skerratt, L.F. (2011) Integrating species traits with extrinsic threats: closing the gap between predicting and preventing species declines. Proceedings of the Royal Society of London B: Biological Sciences, 278, 1515-1523.

Mysterud, A. & Ims, R.A. (1998) Functional responses in habitat use: availability influences relative use in trade-off situations. Ecology, 79, 1435-1441.

Nakagawa, S. & Schielzeth, H. (2013) A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution, 4, 133-142.

Nams, V.O. (2011) Emergent properties of patch shapes affect edge permeability to animals. PLoS One, 6, e21886.

Nams, V.O. (2012) Shape of patch edges affects edge permeability for meadow voles. Ecological applications, 22, 1827-1837.

Nappi, A. & Drapeau, P. (2009) Reproductive success of the black-backed woodpecker (Picoides arcticus) in burned boreal forests: are burns source habitats? Biological Conservation, 142, 1381-1391.

108

Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D. & Smouse, P.E. (2008) A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences, 105, 19052-19059.

Newsome, A. & Catling, P. (1979) Habitat preferences of mammals inhabiting heathlands of warm temperate coastal, montane and alpine regions of southeastern Australia. Ecosystems of the World.

Noble, I.R. & Slatyer, R. (1980) The use of vital attributes to predict successional changes in plant communities subject to recurrent disturbances. Vegetation, 43, 5-21.

O'Brien, T., Baillie, J., Krueger, L. & Cuke, M. (2010) The Wildlife Picture Index: monitoring top trophic levels. Animal Conservation, 13, 335-343.

O'Connell, A.F., Nichols, J.D. & Karanth, K.U. (2010) Camera traps in animal ecology: methods and analyses. Springer Science & Business Media, London.

Oksanen, J., Blanchet, F.G., Kindt, R., Legendre, P., O’hara, R., Simpson, G.L., Solymos, P., Stevens, M.H.H. & Wagner, H. (2011) vegan: Community Ecology Package. R package version 1.17-2. R Foundation for Statistical Computing, Vienna. Available: CRAN. R-project. org/package= vegan.(July 2012).

Orians, G.H. & Wittenberger, J.F. (1991) Spatial and temporal scales in habitat selection. The American Naturalist, 137, S29-S49.

Parkins, K., York, A. & Di Stefano, J. (2018) Edge effects in fire‐prone landscapes: Ecological importance and implications for fauna. Ecol Evol, 0, 1-12.

Parr, C.L. & Andersen, A.N. (2006) Patch mosaic burning for biodiversity conservation: a critique of the pyrodiversity paradigm. Conservation Biology, 20, 1610-1619.

Parsons, A.W., Forrester, T., McShea, W.J., Baker-Whatton, M.C., Millspaugh, J.J. & Kays, R. (2017) Do occupancy or detection rates from camera traps reflect deer density? Journal of Mammalogy, 98, 1547-1557.

Pastro, L.A., Dickman, C.R. & Letnic, M. (2011) Burning for biodiversity or burning biodiversity? Prescribed burn vs. wildfire impacts on plants, lizards, and mammals. Ecological applications, 21, 3238-3253.

109

Pastro, L.A., Dickman, C.R. & Letnic, M. (2014) Fire type and hemisphere determine the effects of fire on the alpha and beta diversity of vertebrates: a global meta‐analysis. Global Ecology and Biogeography, 23, 1146-1156.

Paull, D.J., Claridge, A.W. & Barry, S.C. (2011) There's no accounting for taste: bait attractants and infrared digital cameras for detecting small to medium ground-dwelling mammals. Wildlife Research, 38, 188-195.

Pausas, J.G. (1999) Mediterranean vegetation dynamics: modelling problems and functional types. Plant Ecology, 140, 27-39.

Penman, T., Christie, F., Andersen, A., Bradstock, R., Cary, G., Henderson, M., Price, O., Tran, C., Wardle, G., Williams, R. & York, A. (2011) Prescribed burning: how can it work to conserve the things we value? International Journal of Wildland Fire, 20, 721-733.

Penman, T., Kavanagh, R., Binns, D. & Melick, D. (2007) Patchiness of prescribed burns in dry sclerophyll eucalypt forests in south-eastern Australia. Forest Ecology and Management, 252, 24-32.

Penman, T.D., Price, O. & Bradstock, R.A. (2012) Bayes Nets as a method for analysing the influence of management actions in fire planning. International Journal of Wildland Fire, 20, 909- 920.

Penman, T.E., Cawson, J.G., Murphy, S. & Duff, T.J. (2017) Messmate stringybark: bark ignitability and burning sustainability in relation to fragment dimensions, hazard score and time since fire. International Journal of Wildland Fire, 26, 866-876.

Peters, D.P., Gosz, J.R., Pockman, W.T., Small, E.E., Parmenter, R.R., Collins, S.L. & Muldavin, E. (2006) Integrating patch and boundary dynamics to understand and predict biotic transitions at multiple scales. Landscape Ecology, 21, 19-33.

Pimm, S.L., Jones, H.L. & Diamond, J. (1988) On the risk of extinction. American Naturalist, 132, 757-785.

Pires, A.S., Fernandez, F.A.S., de Freitas, D. & Feliciano, B.R. (2005) Influence of edge and fire- induced changes on spatial distribution of small mammals in Brazilian Atlantic forest fragments. Studies on Neotropical Fauna and Environment, 40, 7-14.

110

Plavsic, M. (2014) Proximate and ultimate drivers of small‐mammal recolonization after fire: microhabitat conditions, rainfall and species traits. Animal Conservation, 17, 573-582.

R Core Team (2017) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Ridout, M.S. & Linkie, M. (2009) Estimating overlap of daily activity patterns from camera trap data. Journal of Agricultural, Biological, and Environmental Statistics, 14, 322-337.

Ries, L., Fletcher, R.J., Battin, J. & Sisk, T.D. (2004) Ecological responses to habitat edges: mechanisms, models, and variability explained. Annual Review of Ecology, Evolution, and Systematics, 35, 491-522.

Ries, L. & Sisk, T.D. (2004) A predictive model of edge effects. Ecology, 85, 2917-2926.

Ritchie, E.G. & Johnson, C.N. (2009) Predator interactions, mesopredator release and biodiversity conservation. Ecology letters, 12, 982-998.

Robinson, N.M., Leonard, S.W., Bennett, A.F. & Clarke, M.F. (2014) Refuges for birds in fire- prone landscapes: The influence of fire severity and fire history on the distribution of forest birds. Forest Ecology and Management, 318, 110-121.

Robinson, N.M., Leonard, S.W., Ritchie, E.G., Bassett, M., Chia, E.K., Buckingham, S., Gibb, H., Bennett, A.F. & Clarke, M.F. (2013) Refuges for fauna in fire‐prone landscapes: their ecological function and importance. Journal of Applied Ecology, 50, 1321-1329.

Rocha, D.G., Ramalho, E.E. & Magnusson, W.E. (2016) Baiting for carnivores might negatively affect capture rates of prey species in camera-trap studies. Journal of Zoology, 300, 205-212.

Rovero, F., Zimmermann, F., Berzi, D. & Meek, P. (2013) " Which camera trap type and how many do I need?" A review of camera features and study designs for a range of wildlife research applications. Hystrix, the Italian Journal of Mammalogy, 24, 148-156.

Rowcliffe, J.M. & Carbone, C. (2008) Surveys using camera traps: are we looking to a brighter future? Animal Conservation, 11, 185-186.

Rowcliffe, J.M., Carbone, C., Jansen, P.A., Kays, R. & Kranstauber, B. (2011) Quantifying the sensitivity of camera traps: an adapted distance sampling approach. Methods in Ecology and Evolution, 2, 464-476.

111

Rowcliffe, J.M., Kays, R., Kranstauber, B., Carbone, C. & Jansen, P.A. (2014) Quantifying levels of animal activity using camera trap data. Methods in Ecology and Evolution, 5, 1170-1179.

Šálek, M., Kreisinger, J., Sedláček, F. & Albrecht, T. (2010) Do prey densities determine preferences of mammalian predators for habitat edges in an agricultural landscape? Landscape and Urban Planning, 98, 86-91.

Santos, X., Bros, V. & Mino, A. (2009) Recolonization of a burned Mediterranean area by terrestrial gastropods. Biodiversity and Conservation, 18, 3153-3165.

Savadogo, P., Sawadogo, L. & Tiveau, D. (2007) Effects of grazing intensity and prescribed fire on soil physical and hydrological properties and pasture yield in the savanna woodlands of Burkina Faso. Agriculture, Ecosystems & Environment, 118, 80-92.

Schultz, C.B., Franco, A. & Crone, E.E. (2012) Response of butterflies to structural and resource boundaries. Journal of Animal Ecology, 81, 724-734.

Seebeck, J., Warneke, R. & Baxter, B. (1984) Diet of the bobuck, Trichosurus caninus (Ogilby)(Marsupialia: Phalangeridae) in a mountain forest in Victoria. Possums and gliders, 145-154.

Si, X., Kays, R. & Ding, P. (2014) How long is enough to detect terrestrial animals? Estimating the minimum trapping effort on camera traps. PeerJ, 2, e374.

Silveira, L., Jacomo, A.T.A. & Diniz, J.A.F. (2003) Camera trap, line transect census and track surveys: a comparative evaluation. Biological Conservation, 114, 351-355.

Sitters, H., Christie, F.J., Di Stefano, J., Swan, M., Penman, T., Collins, P.C. & York, A. (2014) Avian responses to the diversity and configuration of fire age classes and vegetation types across a rainfall gradient. Forest Ecology and Management, 318, 13-20.

Smith, D.A.E., Smith, Y.C.E., Ramesh, T. & Downs, C.T. (2017a) Camera-trap data elucidate habitat requirements and conservation threats to an endangered forest specialist, the Spotted Ground Thrush (Zoothera guttata). Forest Ecology and Management, 400, 523-530.

Smith, J., Legge, S., James, A. & Tuft, K. (2017b) Optimising camera trap deployment design across multiple sites for species inventory surveys. Pacific Conservation Biology, 23, 43-51.

Smucker, K.M., Hutto, R.L. & Steele, B.M. (2005) Changes in bird abundance after wildfire: importance of fire severity and time since fire. Ecological applications, 15, 1535-1549.

112

Spencer, R.J., Cavanough, V., Baxter, G. & Kennedy, M. (2005) Adult free zones in small mammal populations: response of Australian native rodents to reduced cover. Austral Ecology, 30, 868-876.

Srbek-Araujo, A.C. & Chiarello, A.G. (2013) Influence of camera-trap sampling design on mammal species capture rates and community structures in southeastern Brazil. Biota Neotropica, 13, 51-62.

Stephens, S.L., McIver, J.D., Boerner, R.E., Fettig, C.J., Fontaine, J.B., Hartsough, B.R., Kennedy, P.L. & Schwilk, D.W. (2012) The effects of forest fuel-reduction treatments in the United States. BioScience, 62, 549-560.

Sutherland, E.F. & Dickman, C.R. (1999) Mechanisms of recovery after fire by rodents in the Australian environment: a review. Wildlife Research, 26, 405-419.

Swan, M., Christie, F., Sitters, H., York, A. & Di Stefano, J. (2015) Predicting faunal fire responses in heterogeneous landscapes: the role of habitat structure. Ecological applications, 25, 2293-2305.

Swan, M., Di Stefano, J. & Christie, F. (2014) Comparing the effectiveness of two types of camera trap for surveying ground-dwelling mammals. Camera Trapping: Wildlife Management and Research, 123.

Swan, M., Di Stefano, J., Christie, F., Steel, E. & York, A. (2014) Detecting mammals in heterogeneous landscapes: implications for biodiversity monitoring and management. Biodiversity and Conservation, 23, 343-355.

Swan, M., Galindez-Silva, C., Christie, F., York, A. & Di Stefano, J. (2016) Contrasting responses of small mammals to fire and topographic refugia. Austral Ecology, 41, 437-445.

Swann, D.E., Kawanishi, K. & Palmer, J. (2011) Evaluating types and features of camera traps in ecological studies: a guide for researchers. Camera traps in animal ecology, pp. 27-43. Springer, London.

Taylor, R.S., Watson, S.J., Nimmo, D.G., Kelly, L.T., Bennett, A.F. & Clarke, M.F. (2012) Landscape‐scale effects of fire on bird assemblages: does pyrodiversity beget biodiversity? Diversity and Distributions, 18, 519-529.

113

Thompson, M., Medlin, G., Hutchinson, R. & West, N. (1989) Short-term effects of fuel reduction burning on populations of small terrestrial mammals. Wildlife Research, 16, 117-129.

Tobler, M., Carrillo‐Percastegui, S., Leite Pitman, R., Mares, R. & Powell, G. (2008) An evaluation of camera traps for inventorying large‐and medium‐sized terrestrial rainforest mammals. Animal Conservation, 11, 169-178.

Torre, I. & Díaz, M. (2004) Small mammal abundance in Mediterranean post-fire habitats: a role for predators? Acta Oecologica, 25, 137-142.

Torres, J., Marques, J., Alves, P., Costa, H. & Honrado, J. (2017) Local lithological drivers of post-fire vegetation recovery and implications for fire-prone regions. Ecological Research, 32, 37-49.

Tuff, K., Tuff, T. & Davies, K. (2016) A framework for integrating thermal biology into fragmentation research. Ecology letters, 19, 361-374.

Van de Vijver, C., Poot, P. & Prins, H. (1999) Causes of increased nutrient concentrations in post-fire regrowth in an East African savanna. Plant and Soil, 214, 173-185.

Van der Ree, R. & Loyn, R.H. (2002) The influence of time since fire and distance from fire boundary on the distribution and abundance of arboreal marsupials in Eucalyptus regnans- dominated forest in the Central Highlands of Victoria. Wildlife Research, 29, 151-158.

Van Wagner, C. (1968) The line intersect method in forest fuel sampling. Forest Science, 14, 20-26.

Van Wilgen, B., Higgins, K. & Bellstedt, D. (1990) The role of vegetation structure and fuel chemistry in excluding fire from forest patches in the fire-prone fynbos shrublands of South Africa. The Journal of Ecology, 78, 210-222.

Vanbianchi, C.M., Murphy, M.A. & Hodges, K.E. (2017) Canada lynx use of burned areas: Conservation implications of changing fire regimes. Ecol Evol, 7, 2382-2394.

Vanderploeg, H. & Scavia, D. (1979) Two electivity indices for feeding with special reference to zooplankton grazing. Journal of the Fisheries Board of Canada, 36, 362-365.

Vetter, D., Rücker, G. & Storch, I. (2013) A meta-analysis of tropical forest edge effects on bird nest predation risk: Edge effects in avian nest predation. Biological Conservation, 159, 382-395.

114

Vierling, K.T., Lentile, L.B. & Nielsen-Pincus, N. (2008) Preburn characteristics and woodpecker use of burned coniferous forests. Journal of Wildlife Management, 72, 422-427.

Villasenor, N.R., Blanchard, W., Driscoll, D.A., Gibbons, P. & Lindenmayer, D.B. (2015) Strong influence of local habitat structure on mammals reveals mismatch with edge effects models. Landscape Ecology, 30, 229-245.

Villaseñor, N.R., Driscoll, D.A., Escobar, M.A., Gibbons, P. & Lindenmayer, D.B. (2014) Urbanization impacts on mammals across urban-forest edges and a predictive model of edge effects. PLoS One, 9, e97036.

Watson, S., Taylor, R., Nimmo, D., Kelly, L., Clarke, M. & Bennett, A. (2012) The influence of unburnt patches and distance from refuges on post‐fire bird communities. Animal Conservation, 15, 499-507.

Webb, C.T., Hoeting, J.A., Ames, G.M., Pyne, M.I. & LeRoy Poff, N. (2010) A structured and dynamic framework to advance traits‐based theory and prediction in ecology. Ecology letters, 13, 267-283.

Whittington, J., Hebblewhite, M. & Chandler, R.B. (2017) Generalized spatial mark‐resight models with an application to grizzly bears. Journal of Applied Ecology.

Whittington, J., Hebblewhite, M. & Chandler, R.B. (2018) Generalized spatial mark–resight models with an application to grizzly bears. Journal of Applied Ecology, 55, 157-168.

Wiens, J.A. (1976) Population responses to patchy environments. Annual review of ecology and systematics, 7, 81-120.

Williamson, K., Doherty, H. & Di Stefano, J. (2012) Changes in the Relative Density of Swamp Wallabies (Wallabia bicolor) and Eastern Grey Kangaroos (Macropus giganteus) in Response to Timber Harvesting and Wildfire. New Advances and Contributions to Forestry Research. InTech.

Wilsey, B.J. (1996) Variation in use of green flushes following burns among African ungulate species: the importance of body size. African Journal of Ecology, 34, 32-38.

Wilson, J.W., Stirnemann, R.L., Shaikh, Z.S. & Scantlebury, M. (2010) The response of small mammals to natural and human-altered edges associated with Afromontane forests of South Africa. Forest Ecology and Management, 259, 926-931.

115

Winter, J. (1980) Tooth wear as an age index in a population of the brush-tailed possum, Trichosurus vulpecula (Kerr). Wildlife Research, 7, 359-363.

Wood, S. & Scheipl, F. (2014) gamm4: Generalized additive mixed models using mgcv and lme4. R package version 0.2-3.

Yarrow, M.M. & Salthe, S.N. (2008) Ecological boundaries in the context of hierarchy theory. BioSystems, 92, 233-244.

Youngentob, K.N., Yoon, H.-J., Coggan, N. & Lindenmayer, D.B. (2012) Edge effects influence competition dynamics: A case study of four sympatric arboreal marsupials. Biological Conservation, 155, 68-76.

Zuur, A., Ieno, E., Walker, N., Saveliev, A. & Smith, G. (2009) Mixed effects models and extensions in ecology with R. Statistics in Society, 173, 938-939.

116

8 APPENDICES

117

A. Supplementary material for Chapter 3

A-1 Native species habitat preferences and vegetation associations

[Opposite page]

118

119

A-2 Model estimates for habitat variables. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size) between the model and the best model, Akaike weight is the likelihood of the model being the best in the set; R2 is the proportion of variance explained by the model. Models within 2 AICc units of the best model and the null model are presented for each variable, with estimates and 95 percent confidence intervals. Reference levels for estimates are TSF 76 and unburnt. That is, the estimates associated with time since fire represent contrasts with long unburnt (76 years TSF), and estimates associated with edge represent contrasts with unburnt. Response Akaike Marginal Conditional Model AIC Estimate 95% CI Variable C weight R2 R2 Dead material 0- TSF × EDGE 0.00 0.995 0.58 0.72 100cm

B v UB: TSF 0 v 76 -0.193 (-0.235, -0.148)

B v UB: TSF 1_2 v 76 -0.266 (-0.300, -0.133)

B v UB: TSF 3 v 76 -0.198 (-0.210, -0.094)

B v UB: TSF 6_7 v 76 0.164 (0.150, 0.283)

NULL 35.03 0.000 0.00 0.43

Litter TSF × EDGE 0.00 0.962 0.65 0.78

B v UB: TSF 0 v 76 -0.363 (-0.528, -0.198)

B v UB: TSF 1_2 v 76 -0.200 (-0.343, -0.057)

B v UB: TSF 3 v 76 -0.068 -0.233, -0.057)

B v UB: TSF 6_7 v 76 -0.100 (-0.290, 0.090)

NULL 50.40 0.000 0.00 0.40

Shrub 0-50cm NULL 0.00 0.465 0.00 0.60

120

TSF 1.93 0.177 0.15 0.60

Wire Grass TSF × EDGE 0.00 1.000 0.33 0.78 0-100cm

B v UB: TSF 0 v 76 -0.444 (-0.619, -0.270)

B v UB: TSF 1_2 v 76 -0.256 (-0.407, -0.106)

B v UB: TSF 3 v 76 0.078 (-0.096, 0.252)

B v UB: TSF 6_7 v 76 0.236 (0.036, 0.436)

NULL 20.79 0.000 0.00 0.42

Bracken fern 0- TSF × EDGE 0.00 0.462 0.25 0.55 100cm

B v UB: TSF 0 v 76 -0.240 (-0.508, 0.029)

B v UB: TSF 1_2 v 76 -0.043 (-0.276, 0.189)

B v UB: TSF 3 v 76 0.394 (0.126, 0.663)

B v UB: TSF 6_7 v 76 -0.140 (-0.448, 0.167)

NULL 0.90 0.295 0.00 0.29

Coarse woody NULL 0.00 0.359 0.00 0.05 debris

TSF 0.18 0.328 0.12 0.12

121

A-3 Model estimates for species. ΔAICc is the difference in Akaike’s Information Criterion (adjusted for small sample size) between the model and the best model, Akaike weight is the likelihood of the model being the best in the set; R2 is the proportion of variance explained by the model. Models within 2 AICc units of the best model and the null model are presented for each variable, with estimates and 95 percent confidence intervals. Reference levels for estimates are TSF 76 and unburnt. That is, the estimates associated with time since fire represent contrasts with long unburnt (76 years TSF), and estimates associated with edge represent contrasts with unburnt. Akaike Conditional Species Model AICc Marginal R2 Estimate 95% CI weight R2

Swamp wallaby TSF × EDGE 0.00 0.875 0.04 0.13

B v UB: TSF 0 v 76 0.744 (0.210, 1.287)

B v UB: TSF 1_2 v 76 0.814 (0.364, 1.270)

B v UB: TSF 3 v 76 0.105 (-0.421, 0.631)

B v UB: TSF 6_7 v 76 0.546 (-0.261, 1.383)

NULL 20.79 0.000 0.00 0.09

Mountain Brushtail TSF ×EDGE 0.00 0.607 0.10 0.35 Possum

B v UB: TSF 0 v 76 1.637 (0.722, 2.614)

B v UB: TSF 1_2 v 76 1.053 (0.074, 2.085)

B v UB: TSF 3 v 76 1.139 (0.049, 2.284)

B v UB: TSF 6_7 v 76 1.028 (-0.676, 2.834)

NULL 9.23 0.006 0.00 0.31

Wombat TSF 0.00 0.338 0.04 0.13

122

TSF 0 v 76 -0.269 (-6.77, 0.123)

TSF 1_2 v 76 0.683 (0.395, 0.980)

TSF 3 v 76 0.498 (0.158, 0.838)

TSF 6_7 v 76 -0.048 (-0.521, 0.394)

NULL 0.43 0.237 0.00 0.13

TSF + EDGE 1.05 0.200 0.04 0.13

EDGE 1.20 0.186 0.00 0.13

Agile Antechinus TSF × EDGE 0.00 0.824 0.33 -

B v UB: TSF 0 v 76 1.857 (-0.115, 4.882)

B v UB: TSF 1_2 v 76 0.929 (-1.244, 4.019)

B v UB: TSF 3 v 76 -2.281 (-7.257, 0.338)

B v UB: TSF 6_7 v 76 -0.922 (-2.135, 0.237)

NULL 5.84 0.045 0.11 0.21

Bush Rat EDGE 0.00 0.757 0.05 0.15

B v UB 0.906 (0.382, 1.469)

NULL 9.84 0.006 0.00 0.04

123

A-4 Number of species detections per survey method, total detections and if the species was included in the formal analysis.

INCLUDED SURVEY EDGE TSF TSF TSF TSF TSF Total IN SPECIES METHOD POSITION 0 1_2 3 6_7 76 detections ANALYSIS (Y/N) Burnt 7 1 1 2 23 34 Elliot Agile Y trapping Antechinus Unburnt 11 10 4 0 15 40

Burnt 3 1 4 5 6 19 Bush rat Y Unburnt 11 7 10 7 11 46

Burnt 0 0 0 0 6 6 Dusky N antechinus Unburnt 5 0 0 1 3 9

Burnt 67 137 52 18 68 342 Camera Swamp wallaby Y trapping Unburnt 36 70 51 12 74 243

Burnt 20 79 35 13 34 181 Wombat Y Unburnt 18 59 33 13 37 160

Mountain Burnt 59 27 17 4 9 116 Brushtail Y possum Unburnt 24 19 11 3 18 75

Burnt 0 1 2 13 17 33 Long nosed N bandicoot Unburnt 0 3 11 6 14 34

Burnt 0 0 0 0 0 0 Wild dog N Unburnt 3 1 0 0 2 6

Burnt 2 8 9 0 6 25 Sambar deer N Unburnt 0 7 10 0 4 21

Burnt 1 0 0 0 0 1 Echidna N Unburnt 0 2 2 3 8 15

Burnt 6 2 4 1 7 20 Cat Y Unburnt 7 3 9 4 4 27

Burnt 17 14 3 2 2 38 Red fox Y Unburnt 7 10 2 0 0 19

124

B. Supplementary material for Chapter 4

B-1 Search terms for the literature review and table of results.

We searched the Web of Science for the 100 most recently published camera trapping studies. We used the following search criteria: TOPIC (camera trap* OR remote camera*) AND (wildlife OR mammal OR animal), with our search conducted on January 16th 2018. We sorted by date published (most recent) and excluded studies that did not collect terrestrial camera trapping data to make inferences on animal occurrence, abundance or behaviour. We also excluded studies containing only video surveillance, those unavailable in English, or those that presented only review or opinion data. Our final set of the 100 most recently published camera trapping papers is a snapshot of camera trapping paper published across several reputable, peer-reviewed journals with the purpose of providing an example of how many researchers vary the length of camera deployment, as opposed to the number of cameras deployed per site, which is less frequently varied.

List of the 100 most recent camera trapping papers used in the literature review.

Camera Number per Paired days per Total trap Baited # Reference sample (Y/N) site nights (Y/N) point (average) 1 Colyn, R., Radloff, F. & O’Riain, M. (2018) Camera trapping mammals in the scrubland’s of the Cape Floristic Kingdom—the importance of 1 N 69 3450 N effort, spacing and trap placement. Biodiversity and Conservation, 27, 503-520.

2 Whittington, J., Hebblewhite, M. & Chandler, R.B. (2017) Generalized spatial mark‐resight models with an application to grizzly bears. 1 N 365 56552 N Journal of Applied Ecology.

3 Gelin, M.L., Branch, L.C., Thornton, D.H., Novaro, A.J., Gould, M.J. & Caragiulo, A. (2017) Response of pumas (Puma concolor) to 2 Y 0 11442 N migration of their primary prey in Patagonia. PLoS One, 12, e0188877

4 Parsons, A.W., Forrester, T., McShea, W.J., Baker-Whatton, M.C., Millspaugh, J.J. & Kays, R. (2017) Do occupancy or detection rates from 1 N 21 26635 Y camera traps reflect deer density? Journal of Mammalogy, 98, 1547-1557

5 Findlay, M.A., Briers, R.A., Diamond, N. & White, P.J. (2017) Developing an empirical approach to optimal camera-trap deployment at 1 N 0 1720 Y mammal resting sites: evidence from a longitudinal study of an otter Lutra lutra holt. European Journal of Wildlife Research, 63, 96

125

6 Hummel, S., Meyer, L., Hackländer, K. & Weber, D. (2017) Activity of potential predators of European hare (Lepus europaeus) leverets and 2 Y 50.4 6454 N ground-nesting birds in wildflower strips. European Journal of Wildlife Research, 63, 102

7 Giordano, A.J., Tumenta, P.N. & Iongh, H.H. (2017) Camera‐trapping confirms unheralded disappearance of the leopard (Panthera pardus) 1 N 15.5 798 Y from Waza National Park, Cameroon. African Journal of Ecology, 55, 722-726

8 Lessa, I.C.M., Ferreguetti, Á.C., Kajin, M., Dickman, C.R. & Bergallo, H.G. (2017) You can’t run but you can hide: the negative influence 1 N 25 1029 Y of human presence on mid-sized mammals on an Atlantic island. Journal of Coastal Conservation, 21, 829-836

9 Lingaraja, S., Chowdhary, S., Bhat, R. & Gubbi, S. (2017) Evaluating a survey landscape for tiger 1 N 66 4655 N abundance in the confluence of the Western and Eastern Ghats. Current Science (00113891), 113

10 Mendes-Oliveira, A.C., Peres, C.A., Maués, P.C.R.d.A., Oliveira, G.L., Mineiro, I.G., de Maria, S.L.S. & Lima, R.C. (2017) Oil palm 2 Y 60 6720 N monoculture induces drastic erosion of an Amazonian forest mammal fauna. PLoS One, 12, e0187650

11 Oberosler, V., Groff, C., Iemma, A., Pedrini, P. & Rovero, F. (2017) The influence of human disturbance on occupancy and activity patterns 1 N 0 1978 N of mammals in the Italian Alps from systematic camera trapping. Mammalian Biology-Zeitschrift für Säugetierkunde, 87, 50-61

12 Greenville, A.C., Wardle, G.M. & Dickman, C.R. (2017) Desert mammal populations are limited 1 N 730 18250 N by introduced predators rather than future climate change. Royal Society open science, 4, 170384

13 Rodgers, T.W., Xu, C.C., Giacalone, J., Kapheim, K.M., Saltonstall, K., Vargas, M., Yu, D.W., Somervuo, P., McMillan, W.O. & Jansen, P.A. (2017) Carrion fly‐derived DNA metabarcoding 1 N 0 39151 N is an effective tool for mammal surveys: evidence from a known tropical mammal community. Molecular ecology resources

14 Heim, N., Fisher, J.T., Clevenger, A., Paczkowski, J. & Volpe, J. (2017) Cumulative effects of climate and landscape change drive 1 N 0 0 Y spatial distribution of Rocky Mountain wolverine (Gulo gulo L.). Ecol Evol, 7, 8903-8914

15 Kolowski, J.M. & Forrester, T.D. (2017) Camera trap placement and the potential for bias due to 2 Y 21 1371 N trails and other features. PLoS One, 12, e0186679

16 Rodrigues, T.F., Kays, R., Parsons, A., Versiani, N.F., Paolino, R.M., Pasqualotto, N., Krepschi, V.G. & Chiarello, A.G. (2017) Managed forest as 1 N 30 6240 N habitat for gray brocket deer (Mazama gouazoubira) in agricultural landscapes of southeastern Brazil. Journal of Mammalogy, 98,

126

1301-1309

17 Mugerwa, B., du Preez, B., Tallents, L.A., Loveridge, A.J. & Macdonald, D.W. (2017) Increased foraging success or competitor 1 N 50 1250 N avoidance? Diel activity of sympatric large carnivores. Journal of Mammalogy, 98, 1443-1452

18 Suzuki, K.K. & Ando, M. (2017) Seasonal changes in activity patterns of Japanese flying 1 N 32.4 7317 N squirrel Pteromys momonga. Behavioural processes, 143, 13-16

19 Janzen, M., Visser, K., Visscher, D., MacLeod, I., Vujnovic, D. & Vujnovic, K. (2017) Semi- automated camera trap image processing for the 1 N 0 0 N detection of ungulate fence crossing events. Environmental Monitoring and Assessment, 189, 527

20 Lombardi, J.V., Mengak, M.T., Castleberry, S.B. & Terrell, V.K. (2017) Mammal Occurrence in Rock Outcrops in Shenandoah National Park: 1 N 7 1985 N Ecological and Anthropogenic Factors Influencing Trap Success and Co-Occurrence. Natural Areas Journal, 37, 507-514

21 Buzzard, P.J., Li, X. & Bleisch, W.V. (2017) The status of snow leopards Panthera uncia, and high 1 N 0 6300 N altitude use by common leopards P. pardus, in north-west Yunnan, China. Oryx, 51, 587-589

22 Chen, P., Gao, Y., Wang, J., Pu, Q., Lhaba, C., Hu, H., Xu, J. & Shi, K. (2017) Status and conservation of the Endangered snow leopard 1 N 0 38 N Panthera uncia in Qomolangma National Nature Reserve, Tibet. Oryx, 51, 590-593

23 Buzzard, P.J., MaMing, R., Turghan, M., Xiong, J. & Zhang, T. (2017) Presence of the snow leopard Panthera uncia confirmed at four sites in 1 N 0 3216 N the Chinese Tianshan Mountains. Oryx, 51, 594- 596

24 Kachel, S.M., McCarthy, K.P., McCarthy, T.M. & Oshurmamadov, N. (2017) Investigating the potential impact of trophy hunting of wild 1 N 0 4869 N ungulates on snow leopard Panthera uncia conservation in Tajikistan. Oryx, 51, 597-604

25 Rahman, D.A., Gonzalez, G. & Aulagnier, S. (2017) Population size, distribution and status of 1 N 0 5500 N the remote and Critically Endangered Bawean deer Axis kuhlii. Oryx, 51, 665-672

26 Garrote, G., Bueno, J.F., Ruiz, M., de Lillo, S., Martín, J.M., López, G. & Simón, M.Á. (2017) First breeding record of a one-year-old female 1 N 0 50 N Iberian lynx. European Journal of Wildlife Research, 63, 79

27 Zahner, V., Bauer, R. & Kaphegyi, T.A. (2017) Are Black Woodpecker (Dryocopus martius) tree cavities in temperate Beech (Fagus sylvatica) 1 N 0 0 N forests an answer to depredation risk? Journal of Ornithology, 158, 1073-1079

127

28 Bender, L.C., Rosas-Rosas, O.C. & Weisenberger, M.E. (2017) Seasonal occupancy of sympatric larger carnivores in the southern 1 N 0 0 N San Andres Mountains, south-central New Mexico, USA. Mammal Research, 62, 323-329

29 Smith, D.A.E., Smith, Y.C.E., Ramesh, T. & Downs, C.T. (2017) Camera-trap data elucidate habitat requirements and conservation threats to 1 N 0 5796 N an endangered forest specialist, the Spotted Ground Thrush (Zoothera guttata). Forest Ecology and Management, 400, 523-530

30 Dorning, J. & Harris, S. (2017) Dominance, gender, and season influence food patch use in a 1 N 40 0 N group-living, solitary foraging canid. Behavioral Ecology, 28, 1302-1313

31 Gomez, A., Salazar, A. & Vargas, F. (2016) Towards automatic wild animal monitoring: identification of animal species in camera-trap 1 N 0 0 N images using very deep convolutional neural networks. arXiv preprint arXiv:1603.06169

32 Pudyatmoko, S. (2017) Free-ranging livestock influence species richness, occupancy, and daily behaviour of wild mammalian species in Baluran 1 N 78 1562 N National Park, Indonesia. Mammalian Biology- Zeitschrift für Säugetierkunde, 86, 33-41

33 Mathai, J., Sollmann, R., Meredith, M.E., Belant, J.L., Niedballa, J., Buckingham, L., Wong, S.T., Asad, S. & Wilting, A. (2017) Fine-scale distributions of carnivores in a logging 1 N 45 14814 N concession in Sarawak, Malaysian Borneo. Mammalian Biology-Zeitschrift für Säugetierkunde, 86, 56-65.

34 Jumeau, J., Petrod, L. & Handrich, Y. (2017) A comparison of camera trap and permanent 1 N 64 384 N recording video camera efficiency in wildlife underpasses. Ecol Evol, 7, 7399-7407

35 Collett, R.A. & Fisher, D.O. (2017) Time‐lapse camera trapping as an alternative to pitfall 1 N 14 0 N trapping for estimating activity of leaf litter arthropods. Ecol Evol, 7, 7527-7533

36 Bowler, M.T., Tobler, M.W., Endress, B.A., Gilmore, M.P. & Anderson, M.J. (2017) Estimating mammalian species richness and 1 N 0 3147 N occupancy in tropical forest canopies with arboreal camera traps. Remote Sensing in Ecology and Conservation, 3, 146-157

37 Fidino, M. & Magle, S.B. (2017) Using Fourier series to estimate periodic patterns in dynamic 1 N 28 20025 N occupancy models. Ecosphere, 8

38 Robinson, L., Cushman, S.A. & Lucid, M.K. (2017) Winter bait stations as a multispecies 1 N 39 0 Y survey tool. Ecol Evol, 7, 6826-6838

39 Sukumal, N., Dowell, S.D. & Savini, T. (2017) Micro-habitat selection and population recovery 1 N 20 6976 N of the Endangered Green Peafowl Pavo muticus in western Thailand: implications for conservation guidance. Bird Conservation

128

International, 27, 414-430

40 Smith, Y.C.E., Smith, D.A.E., Ramesh, T. & Downs, C.T. (2017) The importance of microhabitat structure in maintaining forest 1 N 21 5250 N mammal diversity in a mixed land-use mosaic. Biodiversity and Conservation, 26, 2361-2382

41 Gray, E.L., Dennis, T.E. & Baker, A.M. (2017) Can remote infrared cameras be used to differentiate small, sympatric mammal species? A case study of the black-tailed dusky antechinus, 1 N 16 725 Y Antechinus arktos and co-occurring small mammals in southeast Queensland, Australia. PLoS One, 12, e0181592

42 Thorne, E.D., Waggy, C., Jachowski, D.S., Kelly, M.J. & Ford, W.M. (2017) Winter habitat associations of eastern spotted skunks in 1 N 8 0 Y Virginia. The Journal of Wildlife Management, 81, 1042-1050

43 Rahman, D.A., Gonzalez, G., Haryono, M., Muhtarom, A., Firdaus, A.Y. & Aulagnier, S. (2017) Factors affecting seasonal habitat use, and 1 N 0 0 N predicted range of two tropical deer in Indonesian rainforest. Acta Oecologica, 82, 41-51

44 Després‐Einspenner, M.L., Howe, E.J., Drapeau, P. & Kühl, H.S. (2017) An empirical evaluation of camera trapping and spatially explicit capture‐ 1 N 0 0 N recapture models for estimating chimpanzee density. American journal of primatology, 79

45 Luja, V.H., Navarro, C.J., Torres Covarrubias, L.A., Cortés Hernández, M. & Vallarta Chan, I.L. (2017) Small Protected Areas as Stepping-Stones 1 N 0 4240 N for Jaguars in Western Mexico. Tropical Conservation Science, 10, 1940082917717051

46 Beca, G., Vancine, M.H., Carvalho, C.S., Pedrosa, F., Alves, R.S.C., Buscariol, D., Peres, C.A., Ribeiro, M.C. & Galetti, M. (2017) High 1 N 30 0 N mammal species turnover in forest patches immersed in biofuel plantations. Biological Conservation, 210, 352-359

47 Wang, T., Feng, L., Yang, H., Han, B., Zhao, Y., Juan, L., Lü, X., Zou, L., Li, T. & Xiao, W. (2017) A science-based approach to guide Amur 2 Y 0 85454 Y leopard recovery in China. Biological Conservation, 210, 47-55

48 Pratas‐Santiago, L.P., Gonçalves, A.L., Nogueira, A.J. & Spironello, W.R. (2017) Dodging the moon: The moon effect on activity allocation of 1 N 120 14728 N prey in the presence of predators. Ethology, 123, 467-474

49 Silva, A.P., Curveira‐Santos, G., Kilshaw, K., Newman, C., Macdonald, D.W., Simões, L.G. & Rosalino, L.M. (2017) Climate and 2 Y 80 8610 Y anthropogenic factors determine site occupancy in Scotland's Northern‐range badger population: implications of context‐dependent responses under environmental change. Diversity and

129

Distributions, 23, 627-639

50 Gecchele, L.V., Bremner-Harrison, S., Gilbert, F., Soultan, A., Davison, A. & Durrant, K.L. (2017) A pilot study to survey the carnivore 1 N 0 0 N community in the hyper-arid environment of South Sinai mountains. Journal of Arid Environments, 141, 16-24

51 Aya-Cuero, C., Rodríguez-Bolaños, A. & Superina, M. (2017) Population density, activity patterns, and ecological importance of giant 1 N 30 1335 N armadillos (Priodontes maximus) in Colombia. Journal of Mammalogy, 98, 770-778

52 Diete, R.L., Meek, P.D., Dickman, C.R., Lisle, A. & Leung, L.K.-P. (2017) Diel activity patterns of northern Australian small mammals: variation, 1 N 0 2246 N fixity, and plasticity. Journal of Mammalogy, 98, 848-857

53 Zhang, J., Hull, V., Ouyang, Z., Li, R., Connor, T., Yang, H., Zhang, Z., Silet, B., Zhang, H. & Liu, J. (2017) Divergent responses of sympatric 1 N 0 0 N species to livestock encroachment at fine spatiotemporal scales. Biological Conservation, 209, 119-129

54 Rathbone, D.A. & Barrett, S. (2017) Vertebrate browsing impacts in a threatened montane plant 1 N 0 304 N community and implications for management. Ecological Management & Restoration, 18, 164-171

55 Spehar, S.N. & Rayadin, Y. (2017) Habitat use of Bornean Orangutans (Pongo pygmaeus morio) in an Industrial Forestry Plantation in East 1 N 30 2065 Y Kalimantan, Indonesia. International Journal of Primatology, 38, 358-384

56 Thapa, K. & Kelly, M.J. (2017) Density and carrying capacity in the forgotten tigerland: 2 Y 0 2097 N Tigers in the understudied Nepalese Churia. Integrative zoology, 12, 211-227

57 Farhadinia, M.S., Hunter, L.T., Jourabchian, A., Hosseini-Zavarei, F., Akbari, H., Ziaie, H., Schaller, G.B. & Jowkar, H. (2017) The critically endangered Asiatic cheetah Acinonyx jubatus 1 N 120 49827 N venaticus in Iran: a review of recent distribution, and conservation status. Biodiversity and Conservation, 26, 1027-1046

58 Rovero, F., Owen, N., Jones, T., Canteri, E., Iemma, A. & Tattoni, C. (2017) Camera trapping surveys of forest mammal communities in the 1 N 48.6 11462 N Eastern Arc Mountains reveal generalized habitat and human disturbance responses. Biodiversity and Conservation, 26, 1103-1119

59 Chauvenet, A.L., Gill, R.M., Smith, G.C., Ward, A.I. & Massei, G. (2017) Quantifying the bias in density estimated from distance sampling and 1 N 0 0 N camera trapping of unmarked individuals. Ecological Modelling, 350, 79-86

130

60 Carricondo-Sanchez, D., Odden, M., Linnell, J.D. & Odden, J. (2017) The range of the mange: Spatiotemporal patterns of sarcoptic mange in 2 Y 0 14660 Y red foxes (Vulpes vulpes) as revealed by camera trapping. PLoS One, 12, e0176200

61 O’Connor, K.M., Nathan, L.R., Liberati, M.R., Tingley, M.W., Vokoun, J.C. & Rittenhouse, T.A. (2017) Camera trap arrays improve detection 1 N 0 0 N probability of wildlife: Investigating study design considerations using an empirical dataset. PLoS One, 12, e0175684

62 Trageser, S.J., Ghose, A., Faisal, M., Mro, P., Mro, P. & Rahman, S.C. (2017) Pangolin 1 N 0 0 N distribution and conservation status in Bangladesh. PLoS One, 12

63 Berry, B., Schooley, R.L. & Ward, M.P. (2017) Landscape context affects use of restored grasslands by mammals in a dynamic 1 N 0 0 Y agroecosystem. The American Midland Naturalist, 177, 165-182

64 Brewster, K., Henke, S.E., Hilton, C. & Ortega-S Jr, A. (2017) Use of Remote Cameras to Monitor the Potential Prevalence of Sarcoptic Mange in 1 N 0 0 N Southern Texas, USA. Journal of wildlife diseases, 53, 377-381

65 Caruso, N., Vidal, E.L., Guerisoli, M. & Lucherini, M. (2017) Carnivore occurrence: do 1 N 35 7054 N interview-based surveys produce unreliable results ? Oryx, 51, 240-245

66 Guarda, N., Gálvez, N., Leichtle, J., Osorio, C. & Bonacic, C. (2017) Puma Puma concolor density 2 Y 40 680 Y estimation in the Mediterranean Andes of Chile. Oryx, 51, 263-267

67 Greene, D.U. & McCleery, R.A. (2017) Reevaluating fox squirrel (Sciurus niger) 1 N 6 1280 N population declines in the southeastern United States. Journal of Mammalogy, 98, 502-512

68 Sollmann, R., Mohamed, A., Niedballa, J., Bender, J., Ambu, L., Lagan, P., Mannan, S., Ong, R.C., Langner, A., Gardner, B. & Wilting, 2 Y 0 0 N A. (2017) Quantifying mammal biodiversity co- benefits in certified tropical forests. Diversity and Distributions, 23, 317-328

69 Hausser, Y., Tagand, R., Vimercati, E., Mermod, S. & Fischer, C. (2017) Comparing survey methods to assess the conservation value of a 1 N 0 726 N community-managed protected area in western Tanzania. African Journal of Ecology, 55, 1-11

70 Martin, E.H., Ndibalema, V.G. & Rovero, F. (2017) Does variation between dry and wet seasons affect tropical forest mammals' 1 N 30 915 N occupancy and detectability by camera traps? Case study from the Udzungwa Mountains, Tanzania. African Journal of Ecology, 55, 37-46

71 McCollum, K.R., Conway, A.L., Lee, M.B. & Carroll, J.P. (2017) Occupancy and demographics 1 N 21 897 N of red river hog Potamochoerus porcus on Tiwai Island, Sierra Leone. African Journal of Ecology, 55, 131

47-55

72 Aguiar-Silva, F.H., Jaudoin, O., Sanaiotti, T.M., Seixas, G.H., Duleba, S. & Martins, F.D. (2017) Camera Trapping at Harpy Eagle Nests: 1 N 0 2660 N Interspecies Interactions Under Predation Risk. Journal of Raptor Research, 51, 72-78

73 Farris, Z.J., Kelly, M.J., Karpanty, S., Murphy, A., Ratelolahy, F., Andrianjakarivelo, V. & Holmes, C. (2017) The times they are a changin': Multi- 2 Y 0 2243 N year surveys reveal exotics replace native carnivores at a Madagascar rainforest site. Biological Conservation, 206, 320-328

74 Ramesh, T., Kalle, R. & Downs, C.T. (2017) Staying safe from top predators: patterns of co- 1 N 24-46 4920 N occurrence and inter-predator interactions. Behavioral Ecology and Sociobiology, 71, 41

75 Kays, R., Parsons, A.W., Baker, M.C., Kalies, E.L., Forrester, T., Costello, R., Rota, C.T., Millspaugh, J.J. & McShea, W.J. (2017) Does 1 N 0 42872 N hunting or hiking affect wildlife communities in protected areas? Journal of Applied Ecology, 54, 242-252

76 Curveira-Santos, G., Marques, T.A., Björklund, M. & Santos-Reis, M. (2017) Mediterranean mesocarnivores in spatially structured managed 1 N 0 6496 N landscapes: community organisation in time and space. Agriculture, Ecosystems & Environment, 237, 280-289

77 Nagy-Reis, M.B., Nichols, J.D., Chiarello, A.G., Ribeiro, M.C. & Setz, E.Z. (2017) Landscape use 1 N 0 5198 N and co-occurrence patterns of neotropical spotted cats. PLoS One, 12, e0168441

78 Molyneux, J., Pavey, C., James, A. & Carthew, S. (2018) The efficacy of monitoring techniques for 1 N 0 1764 N detecting small mammals and reptiles in arid environments. Wildlife Research, 44, 534-545

79 Colyn, R.B., Campbell, A.M. & Smit-Robinson, H.A. (2017) The application of camera trapping to assess Rallidae species richness within 1 N 0 566 N palustrine wetland habitat in South Africa. Ostrich, 88, 235-245

80 Gubbi, S., Harish, N., Kolekar, A., Poornesha, H., Reddy, V., Mumtaz, J. & Madhusudan, M. (2017) From intent to action: a case study for the 1 N 0 11268 N expansion of tiger conservation from southern India. Global Ecology and Conservation, 9, 11-20

81 Stephen, V., Evans, M., Soffian, A.B.M. & Goossens, B. (2017) The first recorded activity pattern for the Sunda stink-badger Mydaus 2 Y 0 24506 N javanensis (Mammalia: Carnivora: Mephitidae) using camera traps. Raffles Bulletin of Zoology

82 McLean, W.R., Goldingay, R.L. & Westcott, D.A. (2017) Visual lures increase camera-trap 1 N 21 0 Y detection of the southern cassowary (Casuarius casuarius johnsonii). Wildlife Research, 44, 230-237

132

83 Gómez-Ramírez, M.Á., Gutiérrez-González, C.E. & López-González, C.A. (2017) Ocelots 2 Y 0 88508 N thrive in a non-typical habitat of northwestern Mexico. Endangered Species Research, 32, 471-478

84 Soofi, M., Ghoddousi, A., Hamidi, A.K., Ghasemi, B., Egli, L., Voinopol-Sassu, A.-J., Kiabi, B.H., Balkenhol, N., Khorozyan, I. & Waltert, M. (2017) Precision and reliability of 1 N 42.8 1585 N indirect population assessments for the Caspian red deer Cervus elaphus maral. Wildlife Biology, wlb. 00230

85 White, J.D., Bronner, G.N. & Midgley, J.J. (2017) Camera-trapping and seed-labelling reveals widespread granivory and scatter-hoarding of 1 N 27 0 Y nuts by rodents in the Fynbos Biome. African Zoology, 52, 31-41

86 Austin, C., Tuft, K., Ramp, D., Cremona, T. & Webb, J.K. (2017) Bait preference for remote camera trap studies of the endangered northern 1 N 5 0 Y quoll (Dasyurus hallucatus). Australian Mammalogy, 39, 72-77

87 Kittle, A.M., Watson, A.C. & Fernando, T.S.P. (2017) The ecology and behavior of a protected 2 Y 0 40 N area Sri Lankan leopard (Panthera pardus kotiya) population. Trop Ecol, 57, 71-86

88 Nichols, M., Glen, A.S., Garvey, P. & Ross, J. (2017) A comparison of horizontal versus vertical camera placement to detect feral cats and 2 Y 0 0 Y mustelids. New Zealand Journal of Ecology, 41, 145- 150

89 Ferreras, P., Díaz‐Ruiz, F., Alves, P. & Monterroso, P. (2017) Optimizing camera‐ trapping protocols for characterizing 1 N 30 2275 Y mesocarnivore communities in south‐western Europe. Journal of Zoology, 301, 23-31

90 Jędrzejewski, W., Puerto, M.F., Goldberg, J.F., Hebblewhite, M., Abarca, M., Gamarra, G., Calderón, L.E., Romero, J.F., Viloria, Á.L. & Carreño, R. (2017) Density and population 1 N 376 12302 Y structure of the jaguar (Panthera onca) in a protected area of Los Llanos, Venezuela, from 1 year of camera trap monitoring. Mammal Research, 62, 9-19

91 Wang, Y., Guan, L., Piao, Z., Wang, Z. & Kong, Y. (2017) Monitoring wildlife crossing structures along highways in Changbai Mountain, China. 1 N 0 2330 N Transportation research part D: transport and environment, 50, 119-128

92 Haverland, M.B. & Veech, J.A. (2017) Examining the occurrence of mammal species in natural areas within a rapidly urbanizing region 1 N 7 963 Y of Texas, USA. Landscape and Urban Planning, 157, 221-230

93 Rovero, F. & Ahumada, J. (2017) The Tropical Ecology, Assessment and Monitoring (TEAM) 1 N 30 0 N Network: An early warning system for tropical rain forests. Science of the Total Environment, 574,

133

914-923

94 Law, S.J., De Kort, S.R. & van Weerd, M. (2016) Morphology, Activity Area, and Movement Patterns of the frugivorous monitor lizard 2 Y 0 58 N Varanus bitatawa. Herpetological Conservation and Biology, 11, 467-475

95 Rayan, D.M. & Linkie, M. (2016) Managing conservation flagship species in competition: 2 Y 0 33727 N tiger, leopard and dhole in Malaysia. Biological Conservation, 204, 360-366

96 Combrink, X., Warner, J.K. & Downs, C.T. (2016) Nest predation and maternal care in the Nile crocodile (Crocodylus niloticus) at Lake St 3 Y 0 * N Lucia, South Africa. Behavioural processes, 133, 31- 36

97 Hernandez-Santin, L., Goldizen, A.W. & Fisher, D.O. (2016) Introduced predators and habitat structure influence range contraction of an 1 N 0 2761 Y endangered native predator, the northern quoll. Biological Conservation, 203, 160-167

98 Paschoal, A.M.O., Massara, R.L., Bailey, L.L., Kendall, W.L., Doherty, P.F., Hirsch, A., Chiarello, A.G. & Paglia, A.P. (2016) Use of 1 N 80 * N Atlantic Forest protected areas by free‐ranging dogs: estimating abundance and persistence of use. Ecosphere, 7

99 Ahmad, S., Hameed, S., Ali, H., Khan, T.U., Mehmood, T. & Nawaz, M.A. (2016) Carnivores’ diversity and conflicts with humans in Musk 1 N 0 1025 Y Deer National Park, Azad Jammu and Kashmir, Pakistan. European Journal of Wildlife Research, 62, 565-576

100 Noble, T.J., Lortie, C.J., Westphal, M. & Butterfield, H.S. (2016) A picture is worth a thousand data points: an imagery dataset of 2 Y 4 0 N paired shrub-open microsites within the Carrizo Plain National Monument. Gigascience, 5

134

B-2 Species detected during the study, and the number of detections per site. Animals that were detected at fewer than four sites were excluded from the analysis.

Number of sites INCLUDED IN SPECIES detected at Total detections ANALYSIS (Y/N) (n=10) Swamp wallaby 10 142 Y Wombat 9 71 Y Bush rat 8 72 Y Mountain brushtail 7 27 Y possum Agile antechinus 5 6 Y Cat 5 11 Y Echidna 4 8 Y Sambar deer 3 10 N Long nosed bandicoot 3 31 N Red fox 2 2 N Rabbit 1 1 N Wild dog 1 2 N

135

C. Supplementary material for Chapter 5

C-1 COLLAR DEVELOPMENT DETAILS

The microTracker is a wildlife activity tracking device incorporating the following features:

- STM32L151CC low power ARM microcontroller - microSD card slot for data storage - microSD card (8GB) - ADXL345B 3-axis accelerometer for activity measurement - MPL3115A2 high resolution barometer for altitude calculation - Magnetic switch sensor & status indicator LEDs - GPS and battery connectors - USB, battery charging, and programming “pogo pin” connections - Expansion connectors (I2C & UART) for optional add-ons, such as an OLED display, or long-range wireless communications - Lithium Ion battery pack (3.7V, e.g. NCA103450 approx. 2300mAh) - GPS receiver module (56 channel u-blox, GP-735T, 38x9x7mm) and 150mm cable

The PCB (excluding battery and GPS receiver) is 38 x 23 x 4.5 mm (Figure C-1a).

Figure C-1a Circuit board developed to record three-dimensional movements by arboreal species (descriptions of each component are listed in Table C-1a)

136

Table C-1a Descriptions of the components attached to the circuit board (Fig. C1)

Code Description Barom Barometric pressure sensor Accel Accelerometer. The x-axis is left to right (positive to the left), y-axis is up and down (positive downwards), and the z-axis from above to below the PCB (positive above). LEDs LED lights to indicate operation status Mag magnetic switch sensor used to turn the device on, off or into sleep mode and put in and out of sleep mode USB/PROG/ Connectors for upgrading firmware OLED/DEBUG BAT Battery connection GPS Connection for the external GPS antenna

ENCLOSURE

We enclosed the PCB and battery inside a small, waterproof casing. The casing was printed in polyamide nylon using a three-dimensional printer. This casing was then coated in an epoxy resin to increase durability. Each GPS housing had two removable ends, one of which had 7 small holes (Figure C-2) to allow air flow to access the altimeter. These holes were covered internally by a small piece of Tyvek (waterproof paper) to ensure the device remained waterproof. Each housing was attached to a commercial VHF wildfire collar for large possums (SirTrack model V5C 163E).

137

Figure C-1b Commercial VHF collar and 3D printed enclosures for the PCB and battery, and final completed collars.

ALTITUDE CALCULATIONS

Altitude calculation requires a known reference pressure at a known reference altitude recorded at a similar time to the pressure recorded by the tracker (i.e. recorded by another tracker at a fixed reference altitude).

138

Equation 1 is used for calculating atmospheric pressure (P) in pascals at a different altitude (h) in metre (for more detail see the MPL3115A2 datasheet & application notes http://www.nxp.com/files/sensors/doc/app_note/AN4528.pdf)

By inserting the values above into the equation we can used equation 2 to calculate altitude based on known atmospherc pressure.

The GPS tracker records in Pascals, which are 1/100th of a mbar (i.e. hPa). Thus a reading of 100014.50 Pa is 1000.1450 mbar, and with standard baseline of 1013.25 mbar at mean sea level, the altitude would be:

A = 44330.77 * (1 – pow(1000.1450/1013.25, 0.190263)) = 44330.77 * (1 – pow(0.98706637, 0.190263)) = 44330.77 * (1 - 0.9975262) = 44330.77 * 0.002473778 = 109.66m

139

Fgure C-1c An example of two dimensional movement data (A) collected from a collar deployed on a mountain brushtail possum in the Central Highlands of Victoria. Addition of the altitude caluclation results in a three-dimensional movement path (B).

140

C-2 Correlation matrix showing Pearson’s correlation coefficients for the explanatory variables. The composite variables (13-16) have the individual components that contribute to the new variable listed in brackets.

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Understorey complexity 1 1 score 2 Midstorey connectivity 0.24 1

3 Silver wattles -0.05 0.37 1

4 Basal area- Wattles 0.15 0.29 0.60 1

5 Canopy cover -0.21 -0.33 -0.16 -0.32 1

6 Basal area- Live trees -0.21 -0.29 -0.11 -0.30 0.73 1

7 Basal area- Dead trees 0.17 0.19 0.18 0.44 -0.18 -0.17 1

8 Tree form- Live -0.09 -0.16 -0.03 -0.25 0.54 0.69 -0.16 1

9 Tree form- 1 0.27 0.02 0.03 0.04 0.02 -0.06 0.19 -0.04 1

10 Tree form- 2 0.24 0.25 0.24 0.49 -0.31 -0.28 0.66 -0.11 0.16 1

11 Tree form- 3 0.20 0.20 0.08 0.21 -0.32 -0.34 0.16 -0.22 -0.05 0.25 1

12 Tree form- 4 0.13 0.10 0.04 0.17 -0.09 -0.07 0.29 0.03 0.09 0.20 0.02 1

13 Overstorey (5, 8, 9) -0.16 -0.26 -0.07 -0.16 0.87 0.84 0.21 0.59 0.06 -0.06 -0.29 0.03 1

14 Midstorey (3, 4) 0.03 0.45 0.97 0.78 -0.24 -0.19 0.29 -0.11 0.04 0.35 0.14 0.09 -0.12 1

15 Live Trees (8, 9) 0.08 -0.13 -0.01 -0.18 0.46 0.55 -0.02 0.81 0.55 0.01 -0.22 0.07 0.52 -0.07 1

16 Dead Trees (10, 11, 12) 0.29 0.27 0.17 0.43 -0.37 -0.35 0.53 -0.16 0.09 0.69 0.68 0.61 -0.17 0.28 -0.08 1 Correlated variables >0.6 are in bold

141

C-3 Box plots of the raw data collected during the vegetation surveys in wet and dry forest

142

143

C-4a Interpolation map of understory complexity at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

144

C-4b Interpolation map of midstorey connectivity at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

145

C-4c Interpolation map of wattle basal area at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

146

C-4d Interpolation map of basal area of live trees at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

147

C-4e Interpolation map of basal area of dead trees at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

148

C-4f Interpolation map of dead tree forms at a site with both wet and dry forest. The blue polygon represents the wet forest, external to the blue section is dry forest. Values have been standardised to range between 0-10, representing low-high availability of each resource.

149

150

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Parkins, Kate Anna

Title: Edge effects in fire prone landscapes: ecological importance and implications for fauna

Date: 2018

Persistent Link: http://hdl.handle.net/11343/217142

File Description: Edge effects in fire prone landscapes: ecological importance and implications for fauna

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.