Pseudomys Novaehollandiae)

Testing the decline of the threatened New Holland Mouse (Pseudomys novaehollandiae)

Phoebe Ann Burns

https://orcid.org/0000-0003-1015-3775

Submitted in total fulfilment of the requirements for the Degree of Doctor of Philosophy

School of BioSciences University of Melbourne Parkville, Victoria 3010

January 2019

I conducted the fieldwork component of this research on the traditional lands of the

Gunaikurnai Nation, the Boon wurrung and Wadawurrung people of the Kulin

Nation, and the Gadubanud people. I wrote this thesis on the lands of the Wurundjeri

and Boon wurrung people of the Kulin Nation, the Gunaikurnai Nation, the

Jardwadjali and Djab wurrung people of Gariwerd, and the Tongva and Tlingit

Nations of Turtle Island. I pay my respects to their rich cultural histories and futures,

to their Elders, past, present, and emerging, their relationship with the Land as

custodians, and to the Land itself.

I pay my respects to the Traditional Custodians of these lands who lost their lives at

the hands of invaders in many locations that I visited and the land on which I live. I

acknowledge these atrocities and recognise that the Land and People still bear these

scars.

Abstract

Delineating the distribution of a threatened species is critical for identifying threats and guiding conservation management. The process is challenging, however, especially when a species is rapidly declining, and so changing its distribution. In this context, species distribution modelling (SDM) often lacks the precision needed to inform fine-scale management decisions, but on-ground surveys to test species’ distributions are time and resource intensive. The dilemma can be mitigated to some extent by careful examination of historical data, and optimal monitoring.

The New Holland Mouse (NHM; Pseudomys novaehollandiae) is one of many

Australian rodent species to have undergone drastic distributional declines since

European invasion. Initially recorded in Victoria in 1970, by 2015 NHMs were thought to occur in only 3 of 12 historically occupied regions. I tested this decline with statistical rigour, using extensive Elliott and camera trapping surveys at >500 sites across Victoria. Combining my survey data with 48 years of others’ efforts, I evaluated the utility of standard Elliott trapping surveys and the efficacy of camera trapping for NHMs. I tested whether NHMs were where we would expect based on state-government threatened fauna SDMs, and whether the species’ purported early- successional fire association explained occurrence or abundance.

I confirmed the species’ persistence in 5 of 12 historical regions – including regions where NHMs had not been detected in 5-21 years – and expanded the species’ known distribution in two regions. However, these finds can be attributed to a paucity of

prior survey effort and were partnered with greater declines elsewhere. Elliott trapping surveys were often inadequate to provide statistical confidence in the species’ absence; camera trap surveys provide a viable alternative for distribution assessments. Standard state-government SDMs provided limited guidance as to the true distribution of NHMs and SDMs for declining species should be interpreted with caution. Time-since-fire did not explain the species’ occurrence and poorly explains abundance, though in certain locations inappropriate fire regimes are a threatening process. Predator control, habitat management, and careful reintroductions are key priorities for conservation of NHMs in Victoria.

Declaration

This is to certify that:

i. the thesis comprises only my original work toward the degree of PhD except where

indicated in the preface; ii. due acknowledgement has been made in the text to all other material used; and iii. the thesis is fewer than 100,000 words in length, exclusive of tables, maps,

bibliographies and appendices.

Phoebe A. Burns

January 2019

iii

Preface

This thesis comprises a series of independent publications. Chapters 2-6 comprise co-authored or sole-authored publications that have been or will be submitted for publication in scientific journals. Consequently, there may be some repetition and stylistic differences among chapters. Although chapters 2-5 are co-authored, I performed all fieldwork and analyses (with the exception of developing the species distribution model referenced in chapter 4), wrote all chapters, and incorporated feedback from co-authors where applicable. Specific contributions of each co-author are outlined below.

My supervisors, Dr Ben Phillips (University of Melbourne), Dr Marissa Parrott

(Zoos Victoria), and Dr Kevin Rowe (Museums Victoria) assisted with the development of research questions and provided feedback on all chapters. Dr Phillips also provided guidance and support for statistical analyses in chapters 2, 3, and 5. Dr

Matt White (Arthur Rylah Institute) created species distribution models for chapter 4, and he and Nick Clemann (Arthur Rylah Institute) assisted with development of the research questions, and provided feedback on the chapter. Claire McCall (Wildlife

Unlimited) identified the original hypothesis regarding moon phase in chapter 2 and provided feedback on the chapter.

The thesis includes the following chapters for publication:

Chapter 2 submitted to Diversity and Distributions:

Burns, P. A., McCall, C., Rowe, K. C., Parrott, M. L. and Phillips, B. L. (in review).

Accounting for detectability and abundance in survey design for a declining species.

Chapter 3 published by Australian Mammalogy online on 21 August 2017:

Burns, P. A., Parrott, M. L., Rowe, K. C. and Phillips, B. L. (2018). Identification of threatened Pseudomys species using infrared and white-flash camera surveys.

Australian Mammalogy 40, 188–197 (DOI 10.1071/AM17016).

Chapter 4 submitted to Austral Ecology:

Burns, P. A., Clemann, N., and White, M. (in review). Testing the utility of generic approaches to species distribution modelling for species in decline.

Chapter 5 submitted to the Journal of Mammalogy:

Burns, P. A., and Phillips, B. L. (in review). Time since fire is an over-simplified measure of habitat suitability for the New Holland Mouse (Pseudomys novaehollandiae).

Chapter 6 submitted to Australian Mammalogy:

Burns, P. A. (in review). Testing the decline of the New Holland Mouse (Pseudomys novaehollandiae) in Victoria.

This work was facilitated by generous financial support from an Australian

Government Research Training Program Scholarship, Parks Victoria, Zoos Victoria, the Holsworth Wildlife Research Endowment, the Royal Zoological Society of NSW through a Paddy Pallin Science Grant and the Ethel Mary Read Research Fund, the

Linnaean Society of New South Wales via the Joyce W. Vickery Scientific Research

Fund, and the Field Naturalists Club of Victoria through their Environment Fund.

Parts of this work were also funded by the Department of Environment, Land, Water, and Planning (DELWP) Biodiversity on Ground Actions funding.

I conducted all fieldwork under permit numbers 1000 7493, 1000 7606 and 1000

7834 from the Department of Environment, Land, Water, and Planning, and with the approval of the Museums Victoria Animal Ethics Committee.

Acknowledgments

Thank you to my supervisors Ben Phillips, Marissa Parrott, and Kevin Rowe for your support and guidance over the years. Ben, thank you for convincing me that I can do stats – under your mentorship I’ve grown from having a mild distain for ‘R’, to finding a peculiar enjoyment in spending weeks running models that occasionally converge. Thank you for always reaffirming that my research is valuable and meaningful, and thank you for adding to that value and meaning. You have an incredible knack for instilling confidence in people in the most relaxed, down-to- earth manner and I am so grateful for that. Marissa, thank you for being the ultimate one-woman cheer squad! Your relentless enthusiasm, passion, and optimism is delightful and you’ve brought so much energy and encouragement to this project.

Kevin, thank you for introducing me to the wonderful world of rats, for supporting and encouraging my initial growth as a scientist, fostering my independent nature and encouraging me to challenge myself. Thank you for teaching me more about island biogeography theory than I ever anticipated.

To my weird and wonderful family – thank you for always being there, showing

(feigning?) enthusiasm and interest for my work, and for accepting without question that I spent nine years at uni. Mum, thank you for always balancing me out and for putting up with my erratic behaviour and the mountains of fieldwork gear flooding the house and veranda for years. Thank you also for being my keenest fieldwork vollie ever. Dad, thank you for making up stories about possums and platypuses that inspired a love of nature in me as a kid, and for all the philosophical chats that have

vii

helped shape the way I think. Kate, thank you for your deep-rooted belief in me, endless encouragement, and your willingness to help out in any way possible (Andy, you’re great too). Allan, thank you for reminding me that there is life outside study, keeping me grounded in frivolous joy and feeding me way too many hash browns.

Henry & Willow, thank you for telling me how boring I am when I’m working and for constantly checking when I’ll be finished my ‘assignment’. Nick, thank you for always loving everything I write, for being genuinely excited about my research, and for all the adventuring.

Thank you to my Museums Victoria crew for being there through a crazy wild ride and for somehow infiltrating the rest of my life. Emily, the Rat Gene Queen, thank you for being an ever-willing sounding board for everything and anything. Thank you for all our epic science plans and for the dozens of papers we may never actually write (just because we’ll be too busy writing others). Kiz and Kate, thank you for all the comic relief and PT chats, thank you especially for your amazing lyric-altering abilities. Pete, thank you for being my lab buddy through the early years, thanks for all the laughs, and for the years of oversharing. Heru, terima kasih for your quirkiness, your support, your banana handling, for running trap lines with me in

Bahasa, and for eventually stopping taking photos of me eating. Jon, thank you for bringing a totally different energy to our lab and for being a wonderful conference travel buddy. Molly, thank you for your perfect taste in everything, and for teaching me more than I ever wanted to know about rat guts. Mon, thank you for your level- headedness and making blood parasites interesting. Sakib, thank you for your smoky

viii

enthusiasm. Thank you to the Phillips lab and Room 136 at Melbourne Uni for creating a comfortable, functional space to learn, grow, and relax into science life.

To my many incredible vollies over the years: Thank you for putting up with the rain, hail, leeches in your eyes, leeches in my ears, tea tree thickets, long treks, early mornings, rat poo filled traps, and occasionally, me yelling in my sleep. Thank you to the people who welcomed me into their homes in Gippsland, providing hot showers and hysterical dinner parties when I was covered in ticks and dirt – Lachie & Rach in

Bairnsdale, Ian & Karen in Loch Sport.

Thank you to the many amazing Parkies who facilitated, supported, and disseminated my research and to those who joined me in the field. Thank you to Jim Whelan for sharing your encyclopaedic knowledge of Yanakie Isthmus and for your enthusiasm and encouragement whenever I turned up with random updates. Thank you to Emily

Green, Scott Griggs, and the whole Prom team for welcoming me into the park, letting me do my thing, and being there to help out when needed. Also thanks to the

Parks Victoria office at Tidal River for giving me a room every month for years, and lending voice to my research through public talks and holiday programs. Thank you to Stephen Vorros for taking me to Snake Island. Thank you also to the Halls Gap

Parks Victoria crew – the Smoky Mouse work didn’t manage to squeeze into this thesis, but it’s on its way.

Thank you to all the support provided by DELWP and ARI staff. My interim supervisor, Lucas Bluff, thank you for facilitating a huge chunk of the data collection

presented in this thesis, for your confidence in me, and for your knowledgeable and constructive feedback. Thank you to Matt White for making a dozen New Holland

Mouse SDMs to guide my searches around the state and for taking the time to co- author a chapter with me. Thank you to Finley Roberts and Luke Smith for your enthusiasm for New Holland Mice and for getting out there and continuing this work.

And thank you to Department of Defence staff who facilitated access to DoD properties or acted as chaperones so I didn’t get blown up: Fred Ford, Juan Montoya

Salazar, Stephen Kearns, and Elise Hardiker.

Thank you to everyone who fell in love with New Holland Mice before me, and who built up the vast knowledge base upon which this work is formed. With particular thanks to those who shared their trapping records and NHM musings: Claire McCall,

Jim Reside, Greg Hollis, Bruce Quin, Peter Homan, Peter Menkhorst, Barbara

Wilson, Cath Kemper, Andrew McCutcheon and Jill Poynton of the Mammal Survey

Group of Victoria, and the Field Naturalists Club of Victoria. Special thanks to Claire

McCall and Greg Fyfe who introduced me to my first two New Holland Mice in the same trap.

Chapters 3, 4, 5, and 6 would not have been possible if not for people generously lending me camera traps and bait stations: Wildlife Unlimited, the Arthur Rylah

Institute (Peter Menkhorst & Luke Woodford), Zoos Victoria (Dan Harley & Marissa

Parrott), DELWP Bairnsdale (Finley Roberts & Lucas Bluff), DELWP Orbost (Tony

Mitchell), and Parks Victoria (Scott Griggs & Jim Whelan).

Table of contents

Abstract ...... i Declaration ...... iii Preface ...... iv Acknowledgments ...... vii Table of contents ...... xi List of Tables ...... xiii List of Figures ...... xvi

Chapter 1 Introduction ...... 20 Humans and the Sixth Mass Extinction ...... 21 Conservation efforts thwarted by data deficiencies...... 22 Imperfect detection impedes data interpretation ...... 24 New technologies and methodologies to test and estimate species distributions .. 24 An Australian context ...... 27 The New Holland Mouse ...... 30

Chapter 2 Accounting for detectability and abundance in survey design for a declining species ...... 39 Abstract ...... 40 Introduction ...... 41 Methods ...... 45 Survey data ...... 45 Abundance - detectability modelling ...... 46 Survey design ...... 48 Results ...... 48 Abundance-detectability modelling ...... 48 Survey design ...... 53 Discussion ...... 56 Data Accessibility Statement ...... 60 Appendix 2.1 – Data sources ...... 61

Chapter 3 Identification of threatened rodent species using infrared and white-flash camera traps ...... 62 Abstract ...... 63 Introduction ...... 63 Materials and methods ...... 67 Study sites ...... 67 Camera trap surveys ...... 67 Identifying characteristics ...... 70 Identification trial ...... 73 Results ...... 75 Surveys ...... 75 Identification trial ...... 75 Discussion ...... 80 Supplementary images ...... 66

Chapter 4 Testing the utility of generic approaches to species distribution modelling for species in decline ...... 91 Abstract ...... 92 Introduction ...... 93 Materials and methods ...... 95 Species Distribution Modelling ...... 95 Ground-truthing the model ...... 98 Results ...... 101 Species Distribution Modelling ...... 101 Ground-truthing the model ...... 103 Discussion ...... 108

Chapter 5 Time since fire is an over-simplified measure of habitat suitability for the New Holland Mouse (Pseudomys novaehollandiae)...... 112 Abstract ...... 113 Introduction ...... 113 Methods ...... 116 Data ...... 116 Occupancy-detectability modelling ...... 118 Abundance-detectability modelling ...... 120 Results ...... 122 Occupancy-detectability modelling ...... 122 Abundance-detectability modelling ...... 126 Discussion ...... 134

Chapter 6 Testing the decline of the New Holland Mouse (Pseudomys novaehollandiae) in Victoria ...... 141 Abstract ...... 142 Introduction ...... 142 Methods ...... 146 Targeted surveys ...... 146 Occupancy - detectability analyses...... 147 Area of Occupancy ...... 148 Results ...... 149 Targeted surveys ...... 149 Occupancy - detectability analyses...... 150 Area of Occupancy ...... 151 Discussion ...... 152 Causes of decline ...... 154 Conservation priorities ...... 159

Chapter 7 General Discussion ...... 164 Practical applications ...... 166 Broader implications ...... 167 Future research ...... 169 Conclusion...... 171

References ...... 173

xii

List of Tables

Table 2.1. Parameter estimates from N-mixture model for Pseudomys novaehollandiae, derived from a 48-year live-trapping dataset in Victoria, Australia.

Table 2.2. Minimum survey effort (number of nights), using 30 Elliott traps with no rain, for 95% confidence in species-level detection of Pseudomys novaehollandiae if present (AX <0.05) at minimum abundance (N = 1) and modelled annual peak mean per-site abundance (N = 7).

Table 3.1. Identifying characteristics for three native Australian rodent species (P. novaehollandiae, P. fumeus, R. fuscipes) and two invasive rodent species (R. rattus,

M. musculus) on white-flash and infrared camera trap images

Table 3.2. Sensitivity, specificity, predictive values and error rates for identification of P. fumeus and P. novaehollandiae.

Table 3.3. Parameter estimates for a binomial mixed-effects model of factors affecting identification accuracy for P. fumeus using white-flash and infrared flash.

Table 3.4. Parameter estimates for a binomial mixed-effects model of factors affecting P. novaehollandiae identification accuracy using white-flash and infrared flash.

xiii

Table 4.1. List of explanatory variables used in developing the Species Distribution

Model for Pseudomys novaehollandiae in Victoria, Australia.

Table 4.2. Performance measures for the Pseudomys novaehollandiae Species

Distribution Model across Victoria, Australia.

Table 4.3. Model selection results for unmarked models of Pseudomys novaehollandiae detectability (p) and occupancy (ѱ) across Victoria, Australia.

Table 4.4. Parameter estimates from top ranked occupancy-detectability model for

Pseudomys novaehollandiae across Victoria, Australia.

Table 4.5. Sensitivity and specificity of a Species Distribution Model to the occurrence of Pseudomys novaehollandiae across Victoria, Australia.

Table 5.1. Parameter estimates for occupancy-detectability model for P. novaehollandiae derived from a 48-year Elliott and camera trapping dataset in

Victoria, Australia.

Table 5.2. Parameter estimates from N-mixture model for Pseudomys novaehollandiae, derived from a 48-year live-trapping dataset in Victoria, Australia covering Time Since Fire (TSF) age classes 1-68 years.

xiv

Table 6.1. Model selection results for ‘unmarked’ models of P. novaehollandiae detectability (p) and occupancy (ѱ) across Victoria, Australia.

Table 6.2. Parameter estimates from top ranked occupancy-detectability model for

Pseudomys novaehollandiae in Victoria, Australia.

Table 6.3. Estimated current minimum known and maximum plausible Areas of

Occupancy (AOO) for P. novaehollandiae in distinct management areas across

Victoria.

List of Figures

Figure 1.1 Distribution of Pseudomys novaehollandiae, Australia, based on historical and modern records.

Figure 1.2. Occurrence records of Pseudomys novaehollandiae in Victoria, Australia,

1970 – 2014.

Figure 2.1. Abundance estimates for Pseudomys novaehollandiae derived from N- mixture models a) seasonally; and b) with an increasing number of Elliott traps used.

Figure 2.2. Detectability estimates for Pseudomys novaehollandiae derived from N- mixture models a) at a species-level seasonally; and at an individual-level b) during different moon phases, c) with increasing rainfall during a survey, and d) over the course of a survey.

Fig. 3.1. Camera trap bait station set-up featuring P. novaehollandiae, Providence

Ponds, Victoria. Image captured using a PixController DigitalEye white-flash camera trap.

xvi

Figure. 3.2. Infrared and white-flash camera images of (a, b) P. fumeus, (c, d) P. novaehollandiae, (e, f) M. musculus, (g, h) R. fuscipes, and (i, j) R. rattus.

Figure 4.1. Habitat suitability estimates derived from a correlative Species

Distribution Model for Pseudomys novaehollandiae across Victoria, Australia. Box indicates plausible region of occupancy.

Figure 4.2. Camera trap survey sites across Victoria, Australia. Circles indicate

Pseudomys novaehollandiae detection sites (yellow) and non-detection sites (black).

Figure 4.3. Sum of Sensitivity and Specificity (SSS) of SDM derived habitat suitability estimates for Pseudomys novaehollandiae in Victoria, Australia.

Figure 4.4. Species Distribution Model recognised suitable habitat for Pseudomys novaehollandiae thresholded at a) optimised value based on maximum Sum of

Sensitivity and Specificity (SSS; 56), and b) the minimum non-zero value at which

Pseudomys novaehollandiae were detected (39); and c) estimated area of occupancy derived from minimum convex hulls by region based on detection records 2013-18.

Figure 5.1. Sites included in Pseudomys novaehollandiae Elliott trapping and camera trapping datasets, Victoria, Australia.

xvii

Figure 5.2. Occupancy estimate for Pseudomys novaehollandiae with increasing

Time Since Fire (TSF), derived from an occupancy-detectability model of a 48-year dataset in Victoria, Australia.

Figure 5.3. Detectability estimates for Pseudomys novaehollandiae populations a) seasonally; b) over the course of a survey; and c) with differing amounts of rainfall during a survey.

Figure 5.4. Abundance estimates for Pseudomys novaehollandiae populations a) with Time Since Fire (TSF); b) with rainfall over an 18-month period prior to the survey; and c) with survey effort, derived from N-mixture model of a 48-year live- trapping dataset in Victoria, Australia covering Time Since Fire (TSF) age classes 1-

68 years.

Figure 5.5. Detectability estimates for Pseudomys novaehollandiae populations a) seasonally (individual level); b) with moon brightness; c) over the course of a survey

(per night, not cumulative); and c) with differing amounts of rainfall during a survey, derived from an N-mixture model of a 48-year live-trapping dataset in Victoria,

Australia covering Time Since Fire (TSF) age classes 1-68 years.

Figure 5.6. Abundance estimates for Pseudomys novaehollandiae populations with

Time Since Fire (TSF) over the 1-15 years following a fire event derived from N- mixture model of subset of a 48-year live-trapping dataset in Victoria, Australia.

xviii

Figure 6.1. Occurrence records of Pseudomys novaehollandiae in Victoria, Australia,

1970 – 2014.

Figure 6.2. Camera trap survey sites for Pseudomys novaehollandiae across Victoria,

Australia.

Figure 6.3. Occurrence records of Pseudomys novaehollandiae in Victoria, Australia.

xix

Chapter 1: Introduction

Chapter 1 Introduction

20 Chapter 1: Introduction

Humans and the Sixth Mass Extinction

There is a global biodiversity crisis underway, heralded as Earth’s ‘Sixth Mass

Extinction’ (Barnosky et al. 2011; Ceballos et al. 2015). Since 1900, around 500 vertebrate species have gone extinct (Ceballos et al. 2015), along with an estimated

130,000 species of invertebrate (Régnier et al. 2015). Though we are yet to meet the benchmark of a 75 percent loss in species diversity needed to classify for a mass extinction event, our current extinction rates are estimated somewhere between 100 –

1000 times higher than background rates, and if sustained, may reach the threshold in a shorter period of time than any previous mass extinction (Pimm et al. 2006;

Barnosky et al. 2011; Pimm et al. 2014; Ceballos et al. 2015).

While prior mass extinction events have spanned millions of years and were attributable to slow climatic shifts, volcanic activity, or cataclysmic asteroid impacts, this current crisis is occurring at unprecedented rates and can be primarily attributed to human activities (Barnosky et al. 2011; Pimm et al. 2014; Cafaro 2015). Through a combination of accidental misfortune and wilful negligence, we are wreaking havoc on Earth’s biota; wanton destruction likened to an ‘interspecies genocide’

(Cafaro 2015). Humans have triggered extinctions by spreading disease, providing passage for invasive species, clearing habitat, and by overhunting (Steadman 1989;

Fahrig 1997; Brooks et al. 2002; Skerratt et al. 2007; Bellard et al. 2016).

Increasingly, human-induced climate change is also placing species at risk (Wiens

2016; Waller et al. 2017), as is our pollution of oceans and waterways (Derraik 2002;

Olden et al. 2007).

21 Chapter 1: Introduction

In addition to species-level extinctions, if we also consider species’ distributional declines (i.e. population-level extinctions) our estimates of recent global biodiversity loss are greatly amplified. Globally, more than 26,500 species are listed as threatened

– a substantial underestimate considering that fewer than five percent of species have been assessed (IUCN 2018). Even among those species classified as ‘least concern,’ declines are prevalent (Ceballos et al. 2017). Aside from simply edging individual species closer toward extinction, population-level extinctions are symptomatic of the destabilisation of entire ecosystems and – though not necessarily the most germane way to look at it – the loss of valuable resources and services helpful in sustaining human life and wellbeing (Cafaro 2015; Ceballos et al. 2017). While humans are demonstrating an exceptional aptitude for destroying biodiversity with unprecedented efficiency (Ceballos et al. 2015), we are also capable of halting extinction processes, and in some cases, remedying them (Hoffmann et al. 2010;

Ceballos et al. 2015; Jones et al. 2016).

Conservation efforts thwarted by data deficiencies

Ignorance is a fundamental impediment to biodiversity conservation (Clemann

2015). Aside from poor political and public comprehension, conservation efforts are often obstructed by incomplete or inadequate knowledge of species’ basic taxonomy, ecology, and distribution (Brito 2004; Butchart and Bird 2010; Bland et al. 2015).

Erroneous assessments of a species’ distribution can misinform species’ habitat associations (Guillera-Arroita et al. 2014; Lahoz-Monfort et al. 2014; Guillera-

Arroita 2016) and contribute to local extinctions by preventing implementation of

22 Chapter 1: Introduction legislative protections (Clemann 2015; Nelson et al. 2017) and misdirecting allocation of conservation resources (Kéry 2002; Chadès et al. 2008; Rout et al.

2010; Boakes et al. 2015). Recognising declines in a species’ abundance or distribution is the initial step in most conservation-status evaluations (e.g. IUCN

2012) and subsequently, in allocating conservation efforts to threatened species. To recognise a species’ distributional decline, you need to 1) know it exists, 2) have some idea of its prior distribution, and 3) be monitoring it adequately to notice a decline. All three steps present surprisingly large obstacles.

First, we haven’t yet discovered or described all species; an average of 25 new species of mammal and around 15,000 species of invertebrate are described each year, and there is no expectation those numbers will decrease any time soon (Cardoso et al. 2011; Burgin et al. 2018). Species that lack taxonomic ranking are in the most basic sense unable to be afforded conservation status and protections, and many are expected to go extinct before we are aware of them (Dubois 2003). Some species have only been studied insofar as to describe them, with little knowledge of their ecology, and in some extreme cases the type specimen forms the only known record

(Dubois 2003). For species that have been described and studied to a greater degree, our understanding of their current distributions is typically patchy (Pimm et al.

2014), and our knowledge of their historical distributions often near non-existent, limiting our ability to assess declines (Bilney et al. 2010). Identifying modern-day shifts in species distributions remains challenging because effective long-term monitoring programs capable of registering species declines are exceptionally rare

(Legge et al. 2018).

23 Chapter 1: Introduction

Imperfect detection impedes data interpretation

When survey for a species does occur, the results still may not give a clear indication of the species’ presence or absence from a site: non-detection of a species at a site does not always mean it is not present (MacKenzie et al. 2002; Wintle et al. 2005;

Rout et al. 2010). This is an issue of detectability – the probability of finding a species at a site during a survey, if it is present. Detectability varies widely among and within species; it varies with survey technique, effort and duration (O’Connell et al. 2006; Gibson 2011; Willson et al. 2011; Molyneux et al. 2018), survey timing

(Kéry 2002; Willson et al. 2011; Lillywhite and Brischoux 2012; Gilmore 2016), or weather conditions (Wintle et al. 2005; Otto and Roloff 2011; Smith et al. 2012).

Detectability also varies systematically with abundance (Kéry 2002; McCarthy et al.

2013; Scheele and Gillespie 2018), so we would expect species to become harder to detect as they decline. Determining the minimum survey effort required for confident classification of absence is an important step in adequate survey design, particularly when assessing local extinctions in potential distributional declines (Kéry 2002; Kéry et al. 2009).

New technologies and methodologies to test and estimate species distributions

Traditional survey methods that involve physically capturing animals are important for demographic studies, genetic sampling, specimen collection, understanding disease, and answering other important ecological questions (Sikes and Gannon

2011; Clemann et al. 2014). When a species’ presence or absence is the only factor of interest, however, trapping programs can be cumbersome and resource heavy; in the context of understanding species’ distributions and declines, the financial cost and

24 Chapter 1: Introduction time involved in such surveys can be a serious impediment (Nelson et al. 2009; De

Bondi et al. 2010). New technologies have the potential to improve species’ detectability and reduce the amount of effort required to conduct surveys, while new statistical methods can improve inferences drawn from survey data and provide guidance on species’ probable former distributions.

Emerging field-based technologies such as camera traps, automated acoustic recording devices, drones, and thermal imaging may improve survey efficiency for many species – increasing detection probabilities, lowering survey costs, and allowing greater areas to be surveyed. Surveys using thermal imaging cameras can have drastically better detection probabilities than spotlighting surveys (Focardi et al.

2001); automated recording devices and ‘recogniser’ software cut down on in-field time for avian and amphibian acoustic surveys (Acevedo and Villanueva‐Rivera

2006; Bardeli et al. 2010; Blumstein et al. 2011); unmanned aerial vehicles (i.e. drones) are dramatically reducing costs of aerial surveys (Colefax et al. 2018); and camera traps have been widely adopted as a cost-effective presence/absence survey method for mammals (De Bondi et al. 2010; Roberts 2011; Meek et al. 2015b) and, increasingly, for other taxa (O’Brien and Kinnaird 2008; Ariefiandy et al. 2013;

Welbourne et al. 2015). The appropriateness of each of these tools, however, varies among species and requires assessment prior to broad-scale application – e.g. species’ vocalisations may be poorly distinguished (Waddle et al. 2009), or species may be difficult to identify on camera trap images (Swan et al. 2014a; Meek and

Vernes 2015; Villette et al. 2016).

25 Chapter 1: Introduction

Even with highly efficient techniques, we cannot conceivably survey everywhere – inference must be drawn to some degree, and correlative Species Distribution

Models (SDMs) have the capacity to draw that inference with estimable precision

(Elith et al. 2006). Such SDMs correlate species’ occurrence data with explicit environmental covariates (e.g. climate, topography, vegetation type) to map environmental suitability for species (Smith et al. 2013). This output has many potential applications for threatened species, such as predicting responses to changing climates (Sinclair et al. 2010) or identifying areas of potentially suitable habitat for survey, reintroductions, or translocations (Peterson 2006). Where broad- scale surveys to determine species’ occurrence are impractical, SDMs are increasingly being used to estimate individual species’ distributions or biodiversity hotspots for prioritising conservation funding (Guisan et al. 2013; Liu et al. 2013b;

Farashi et al. 2017), to inform environmental impact assessments (DEPI 2013; OEH

2014), and to select compensatory land for threatened species during offsetting processes (DELWP 2017a).

As species decline, however, the threatening processes causing distributional declines or shifts (e.g. unknown environmental factors, historical extirpations, predation, competition, disease; Jetz et al. 2008; Sinclair et al. 2010) may not be accounted for in standard correlative SDMs. Thus, metrics of habitat suitability drawn from SDMs may provide an increasingly misleading estimate of the species’ plausible distribution. For example, the Cane Toad (Rhinella marina) has triggered catastrophic declines in Northern Quoll (Dasyurus hallucatus) populations across northern Australia (Burnett 1997); a traditional habitat model for the species based on historical records would grossly overestimate the quoll’s current distribution,

26 Chapter 1: Introduction failing to account for the effect of the toad. Neglecting to recognise and address the potential shortcomings of SDMs in describing a species’ current realised niche may lead to under- or over-estimation of the distribution of a threatened species, with attendant adverse conservation outcomes (Loiselle et al. 2003; Clemann 2015).

Assessing the utility of SDM predictions of habitat suitability for estimating distributions of threatened and declining species may help to moderate their application in modelling-reliant management strategies (e.g. DEPI 2013; OEH 2014;

DELWP 2017a) and identify avenues to advance modelling techniques.

An Australian context

In the relatively brief period since the European invasion of Australia, the continent has experienced a catastrophic collapse in its native flora and fauna. In just 230 years, we have witnessed the extinction of at least 95 species, and 1775 are currently

Federally recognised as under threat (Department of Environment and Energy 2018b;

Department of Environment and Energy 2018c). The declines we recognise are likely a fraction of the true declines suffered by many species, because our understanding of species’ current and past distributions is poor (Bilney et al. 2010).

The agents of impact are numerous. Land clearing for grazing, agriculture, housing and roads has drastically reduced many species’ distributions simply by removing habitat (Stevens 2001; Woinarski et al. 2014). For example, less than 0.5 percent of grassy plains ecosystems in south eastern Australia remain intact (Lunt 1991).

Through central and eastern Victoria, logging of native forests threatens the critically endangered Leadbeater’s Possum (Gymnobelideus leadbeateri) and the vulnerable

27 Chapter 1: Introduction

Greater Glider (Petauroides volans; Lindenmayer 1994; Lindenmayer et al. 2016). In

New South Wales and Queensland, lax environmental protection laws permit clearing of critical habitat for populations of the northern-form Koala (Phascolarctos cinereus), as remnant populations fall prey to Domestic Dogs (Canis familiaris) and vehicles (Maron et al. 2015; Shumway et al. 2015; McAlpine et al. 2015). We have introduced novel predators to naïve prey, causing populations of native fauna to crash in dramatic fashion with the arrival of Domestic Cats (Felis catus) and Red Foxes

(Vulpes vulpes; Short et al. 2002; Allen and Leung 2012; Doherty et al. 2016). Other species we brought with us may not eat native fauna, but can change vegetation communities (e.g. overgrazing at Wilsons Promontory; Bennett 1994), destroy habitat (e.g. horses, Equus ferus caballus, in Australian Alps; Porfirio et al. 2017;

Tolsma and Shannon 2018), or establish new regimes to which our native species are unsuited (e.g. cattle, Bos taurus, grazing on Bogong High Plains; Williamson et al.

2014). Unique introduced species such as the cane toad act as both predator and toxic prey, threatening a wide diversity of native fauna (Burnett 1997; Boland 2004;

Woinarski et al. 2010). We spread disease and disease carrying species, triggering rapid declines (e.g. Chytrid fungus; Skerratt et al. 2007; Skerratt et al. 2016) or insidiously slow community change (e.g. Cinnamon fungus; Newell 1998). We ignored, and in many cases have now lost, Indigenous knowledge of fire and land management, instead burning too much or too little, ‘sacrificing’ blocks of unique habitat to meet burning targets, and fundamentally changing the composition and function of vast regions (Bowman 1998; Haslem et al. 2011; Bennett et al. 2014;

Pascoe 2014). We continue to shift the courses of rivers, draining the life out of regions, inappropriately flooding others, contributing to erosion and toxic salinization, and exacerbating the effects of climate change (e.g. Murray Darling

28 Chapter 1: Introduction

Basin; Walker and Thoms 1993; Pittock and Finlayson 2011). And in some instances, we sent species toward extinction by actively hunting them (e.g. the Thylacine,

Thylacinus cynocephalus, and the Toolache Wallaby, Macropus greyi; Short and

Smith 1994; Paddle 2002; Prowse et al. 2013).

Australian mammal species are faring exceptionally poorly. Australia currently holds the record for the most rapid mammal extinction rate of any time or place, ever

(Woinarski et al. 2015). There have been at least 30 extinctions and 109 declines drastic enough to warrant a Federal threatened species listing (Department of

Environment and Energy 2018b). Half of these extinctions and many of these declines are among our rodents: of 64 species recorded in recent history, 14 are now extinct and half the remainder have a threatened species listing (Short and Smith

1994; Morris 2000; Robinson et al. 2000; Seebeck and Menkhorst 2000; Dickman et al. 2000a; Dickman et al. 2000b). Poor historical survey effort, and the relatively recent taxonomic recognition of many native rodent species, limits our understanding of their declines, while haphazard modern survey efforts provide incomplete knowledge of current distributions (Bilney et al. 2010; Burns et al. 2016). Our poor understanding of species’ distributions and abundances and how these have shifted over time impedes our ability to identify and remedy threatening processes (Wayne et al. 2013; Bilney 2014). Effective conservation of our remaining rodent species will benefit from investment in comprehending their declining distributions and the processes responsible for these declines.

29 Chapter 1: Introduction

The New Holland Mouse

The New Holland Mouse (Pseudomys novaehollandiae; Indigenous: Pookila) is a small threatened rodent species native to south eastern Australia, part of the ‘Old

Endemic’ murid radiation that evolved in Australia over the past ~ 4 Mya (Fig. 1.1;

Breed and Ford 2007). The mouse is an exemplar of inadequate historical survey effort and the drastic recent declines experienced by many native species

(Department of Environment and Energy 2018b). Pseudomys novaehollandiae has an eclectic detection history characterised by bursts of detection, and decades of non- detection. While the species may demonstrate boom-bust dynamics (Wilson et al.

2005; Wilson et al. 2017), its sporadic detection history appears more to reflect fluctuations in interest and effort to find the mouse, and also perhaps a widespread failure to distinguish it from the invasive House Mouse (Mus musculus; Mahoney and Marlow 1968). The species’ initial discovery – a female and two young, sheltering under a piece of bark – was a sidenote in John Gould’s foray in the Upper

Hunter, New South Wales in 1840 (the specimens later described by Waterhouse in

1843; Mahoney and Marlow 1968; Mahoney 1974). Nearly half a century passed until the next reliable record – a single specimen from 1886 – then nothing until a chance hand-capture by a park ranger in 1967, by which point the species had been presumed extinct (Mahoney and Marlow 1968; Mahoney 1974). Renewed awareness of the species was followed by a burst of discoveries of new populations in New

South Wales, Victoria, Tasmania, and Flinders Island, eventually including

Queensland in 1996 (Seebeck and Beste 1970; Hocking 1980; Van Dyck and Lawrie

1997).

30 Chapter 1: Introduction

Figure 1.1 Distribution of Pseudomys novaehollandiae, Australia, based on

historical and modern records. Black dots indicate subfossil records outside

modern range, red dot indicates reintroduced population. Adapted from (Burns

et al. in press)

Pseudomys novaehollandiae is a nocturnal rodent that lives in burrow systems, often communally, favouring sandy soils across a variety of habitats (Keith and Calaby

1968; Kemper 1977; Fox and Fox 1986; Pye 1991; Lazenby et al. 2008). Early habitat association studies linked P. novaehollandiae with early post-fire successional vegetation stages in dry sclerophyll forests, coastal heaths and vegetated

31 Chapter 1: Introduction sand dunes (Keith and Calaby 1968; Posamentier and Recher 1974; Cockburn 1980;

Pye 1991). The species has also been recorded in woodlands, shrublands and grasslands, including some mature, undisturbed habitats (Hollis 1996; Wilson and

Laidlaw 2003; Lazenby et al. 2008; McCall et al. 2015). Although omnivorous – eating a variety of seeds, vegetation, fungi, and invertebrates – there is evidence P. novaehollandiae is a preferential granivore when given the choice (Cockburn 1980;

Norton 1987; Wilson and Bradtke 1999). Current management strategies for the species focus primarily on burning regimes to provide early successional habitat

(Parks Victoria 1998; Anon. 2010), although the species’ response to these planned burns and its relationship with time since fire has not been rigorously assessed.

The species differs in size and breeding behaviour along its latitudinal gradient. As with many mammal species, Bergmann’s rule applies: average adult body weight for

P. novaehollandiae more than halves from 28 g in Tasmania to 12 g in Queensland

(Burns et al. in press). The species’ breeding season is longer in the north of its range, and further extended during periods of high food availability (Kemper 1980;

Wilson et al. 2005). Typically, Victorian and Tasmanian populations breed between late Spring and early Autumn with one or two litters (Pye 1991; Wilson et al. 2005), whereas populations in New South Wales breed variably between early Spring and late Autumn, bearing 3-5 litters in a season (Kemper 1980). Gestation lasts 29-32 days and juveniles are weaned after 3-4 weeks (Kemper 1976). Litters vary in size between one and six individuals (Kemper 1976; Kemper 1980). Due to the seasonal nature of breeding, population abundance fluctuates annually (Wilson et al. 2005;

Lock 2005). Behaviour also fluctuates seasonally; the species is poorly trappable over the summer months – male individuals in particular disappearing for months,

32 Chapter 1: Introduction only to reappear in Autumn (Kemper 1980; Fox 1982; Wilson 1991; Wilson et al.

2005).

Since its rediscovery 50 years ago, P. novaehollandiae appears to have suffered numerous local extinctions (Wilson et al. 2017; Lazenby et al. 2018). As we begin to consider the species’ distributional decline, it is important to note that P. novaehollandiae was not recognised throughout much of its range until 130-170 years after the regions had begun being drastically altered by European settlement.

Thus, any decline we can identify between the species’ discovery and present time is likely a fraction of that which occurred prior. Even so, the observed decline is substantial. Populations detected in Victoria seemed to disappear soon after their discovery; by 2015, only three of 12 historically occupied areas were known to still support the species (Fig. 1.2; Quin 1996; Wilson and Laidlaw 2003; Lock 2005;

McCall et al. 2014). The species appears to also have declined appreciably in

Tasmania, where it has not been captured in a decade (Lazenby et al. 2018). Some populations in New South Wales are considered stable, while other historically abundant populations in that State were not detected when recently resurveyed (F.

Ford pers. comm.). It remains to be tested, however, whether historically occupied areas have been resurveyed with enough effort to confirm the species absence: are the apparent declines simply a reduction in abundance, or indicative of true distributional declines? Delineating the species’ current distribution and how that has changed over time will aid us in identifying and ameliorating the processes responsible for its decline. Knowing where the species persists will also aid us in directing management funding and efforts, and in identifying locations for potential reintroductions, if decline is substantial enough to warrant such action.

33 Chapter 1: Introduction

Figure 1.2. Occurrence records of Pseudomys novaehollandiae in Victoria,

Australia, 1970 – 2014. Dates indicate period of recorded occurrence in each

region; light grey shading depicts water bodies. Note: P. novaehollandiae

records at Anglesea 2000-2003 were a product of an unsuccessful

reintroduction attempt and not included here.

Traditionally, as with most small mammals in Australia, P. novaehollandiae are most commonly surveyed using Elliott live traps (e.g. Lock and Wilson 1999; Fox et al.

2003; Wilson et al. 2005), occasionally supplemented with cage or pitfall traps (e.g.

Quin 1996; Fauna Survey Group F.N.C.V., unpub. data) and historically, break-back traps (e.g. Kemper 1980). These methods, while effective at P. novaehollandiae capture, are labour intensive and impose logistical restrictions on potential broad- scale resurvey efforts. Further, no work has been done assessing the species’ detectability with live capture methods and how that may change among regions, with shifts in abundance, under different survey conditions, or with seasonal

34 Chapter 1: Introduction fluctuations in behaviour (i.e. related to resource availability or breeding behaviours).

Camera trapping may provide an efficient alternative to live trapping, facilitating more extensive survey efforts. Our ability to take advantage of camera trap technology is contingent on being able to identify the species in images, which is not a given for P. novaehollandiae (which has been confused with M. musculus in hand;

Quin 1996). For areas where historical survey effort has lacked the power to confirm extinctions, camera traps may provide a valuable tool for resurvey efforts. Regardless of technique, however, to be statistically confident in the species local extinctions, we must determine how much effort is required to reliably detect P. novaehollandiae and how that may change among surveys.

In short, the most pressing questions in P. novaehollandiae conservation and management at present are: where does the species persist? And, if populations have been lost, why? My thesis addresses these questions, and aims to:

1) test the adequacy of historical survey efforts in identifying the species’ absence, and determine minimum requirements for future presence/absence surveys;

2) test the utility of emerging technologies (i.e. camera traps) in detecting P. novaehollandiae;

3) test the utility of species distribution models in estimating and explaining distribution of P. novaehollandiae as a declining species;

4) test the role of a presumed threatening process – inappropriate fire regimes in the form of excessive time-since-fire – in explaining the distribution of P. novaehollandiae; and,

5) test the overall distributional decline of P. novaehollandiae in Victoria and survey

35 Chapter 1: Introduction for the potential occurrence of as yet undetected populations in poorly surveyed areas.

I address these aims in the following chapters:

In Chapter Two, I investigate the adequacy of historical Elliott trapping survey efforts to inform the absence of P. novaehollandiae. I assess the factors that influence the species’ detectability, while accounting for changes in abundance caused by seasonal variation among surveys, differences among sites and regions, and trapping effort. I find that the species’ detectability varies considerably under different survey conditions and with abundance, and that traditional survey efforts were inadequate to provide statistical support for the species’ absence.

In Chapter Three, I test whether camera trapping is a viable technique for P. novaehollandiae occupancy surveys. The challenge being whether P. novaehollandiae can be distinguished from M. musculus in white-flash and infrared camera trap images. I find that P. novaehollandiae is readily identified on both camera trap types, supported by results from an identification trial testing the identification accuracy of 15 ecologists and laypersons with varying levels of relevant expertise. Since completion of this chapter early in my PhD, DELWP

Bairnsdale, Parks Victoria, and several consulting companies have adopted camera trapping as a go-to tool for P. novaehollandiae occupancy surveys, using the guidelines outlined here to aid in identifications.

In Chapter Four, I explore the use of Species Distribution Models (SDMs) to delineate distributions of threatened and declining species. Testing whether standard

36 Chapter 1: Introduction

State government threatened species distribution modelling techniques can provide a useful indication of where species persist, particularly in the context of declining species whose threatening processes are unlikely to be accounted for in the modelling process. I find that standard modelling techniques grossly over-represent the species’ distribution, and when thresholds are tightened, sensitivity to the species’ presence is diminished.

In Chapter Five, I delve in to the potential role of fire as a threatening process in the presumed decline of P. novaehollandiae, testing the hypothesised association between P. novaehollandiae and early successional vegetation. I test whether Time-

Since-Fire (TSF) affects P. novaehollandiae occupancy and abundance, finding little support for the former, and moderate support for the latter. Further, in Chapter Five, I provide guidance for fire management for short- and long-term P. novaehollandiae persistence.

Chapter Six serves both as a stand-alone assessment of the present distribution of P. novaehollandiae in Victoria, and as a synthesis of the work contained within the rest of this thesis. In it, I draw on knowledge gleaned in each other chapter to provide a comprehensive assessment of the status of P. novaehollandiae in Victoria, redefining the species’ known distribution and outlining threatening processes and conservation priorities. Chapter Six is a thorough, statistically rigorous survey of plausible P. novaehollandiae habitat across Victoria, deemed necessary by findings from Chapter

Two, using methodologies from Chapter Three, and habitat selection guided by

Chapter Four. Findings from Chapter Five are incorporated in shaping guidelines for the species’ ongoing management priorities. Foremost, I find that P.

37 Chapter 1: Introduction novaehollandiae has suffered a catastrophic distributional decline within Victoria in the past 40 years.

38 Chapter 2: Survey requirements of a declining species

Chapter 2 Accounting for detectability and abundance in survey design for a declining species

Burns, P. A., McCall, C., Rowe, K. C., Parrott, M. L. and Phillips, B. L. (in review).

Accounting for detectability and abundance in survey design for a declining species.

Diversity and Distributions.

39 Chapter 2: Survey requirements of a declining species

Abstract

Aim

As a species declines in abundance it becomes harder to detect – simply because there are fewer animals to observe. Survey effort that once detected the species may become inadequate and, if this goes unrecognised, a species may prematurely be labelled locally extinct. This can preclude management actions, and in turn contribute to the species’ local extinction. Determining survey requirements for declining populations requires estimation of the survey effort necessary to reliably detect the species, assuming only a single individual is present at a site. As detectability can differ among surveys due to seasonal, behavioural, and environmental variables, these must also be accounted for in detectability analyses and survey planning. To assess distributional declines of a threatened Australian rodent, the New Holland Mouse (Pseudomys novaehollandiae), we determined minimum survey requirements for achieving suitable confidence in the species’ absence.

Location

Victoria, Australia.

Methods

Using a 48-year live trapping data set, we estimated P. novaehollandiae abundance and detectability using N-mixture modelling in a Bayesian framework, testing the effects of seasonal fluctuations and environmental variables to determine ideal conditions for occupancy surveys. We used these findings to assess the adequacy of historical resurvey efforts in confirming local extinctions.

40 Chapter 2: Survey requirements of a declining species

Results

We found that detectability followed seasonal fluctuations distinct from changes in abundance, was strongly reduced by brighter moon phases, and slightly increased over the course of a survey and when raining. When abundance was low, standard historical survey efforts were inadequate to assert the species’ absence with appropriate statistical confidence, indicating that further surveys would be required to test purported local extinctions.

Main conclusions

Our results highlight the potential for considerable intra-species heterogeneity in detectability caused by shifting local abundances and survey conditions, as well as the importance of accounting for detectability when monitoring threatened and declining species.

Introduction

Misinterpretation of ‘absence’ data can have catastrophic consequences for threatened species and their conservation (Kéry 2002; Chadès et al. 2008; Boakes et al. 2015). Whether through insufficient effort, poor survey conditions, or other factors, non-detections may be recorded as absences when a species is in fact present

(MacKenzie et al. 2002; Wintle et al. 2005; Rout et al. 2010). Inaccurate assessments of species’ occurrence and abundance bias species distribution models and misinform habitat associations (Guillera-Arroita et al. 2014; Lahoz-Monfort et al. 2014;

Guillera-Arroita 2016). In situations where threat mitigation and preservation of an occupied area depends on timely proof of a species’ occurrence and abundance (e.g. exclusion zones in logging coupes for critically endangered Leadbeater’s Possum;

Nelson et al. 2017), or where allocation of conservation resources requires

41 Chapter 2: Survey requirements of a declining species confirmation of the species’ presence, false absence data can contribute to local extinctions (Rout et al. 2010; Boakes et al. 2015). To minimise the risk of this occurring, there is an acute need for informed, rationalised survey effort, and critical interpretation of survey data (Reed 1996; Boakes et al. 2016).

The probability of finding a species at a site if it is present, – the species’ detectability – is central to appropriate survey design (Reed 1996; MacKenzie et al.

2002; Guillera-Arroita et al. 2010; Wintle et al. 2012). Detectability varies widely among species, even for closely related taxa (Wintle et al. 2005). Within species, detectability can vary with a multitude of factors, such as seasonal changes in behaviour (Kéry 2002; Willson et al. 2011), habitat quality (Wintle et al. 2005), temperature (Smith et al. 2012), elevation (Kéry 2008), rainfall (Wintle et al. 2005;

Otto and Roloff 2011), wind speed (Wintle et al. 2005), and moon phase (Lillywhite and Brischoux 2012; Gilmore 2016). Survey technique, effort, and duration can also affect detection probabilities (O’Connell et al. 2006; Molyneux et al. 2018). In capture-based surveys, animals may acclimate to the presence of a trap, becoming more trappable or less trappable over the duration of a survey as they overcome trap neophobia, become ‘trap happy’, or develop avoidance behaviour (Tasker and

Dickman 2001; Gibson 2011; Willson et al. 2011). Failure to investigate and understand these factors when planning surveys and interpreting survey results can lead to misinterpretation of species’ abundance and population trends, or even premature assumption of the species’ local extinction (Kéry 2002; Kéry et al. 2009).

Detectability also varies systematically with abundance (Kéry 2002; McCarthy et al.

2013; Scheele and Gillespie 2018). Simply, the more individuals present, the greater

42 Chapter 2: Survey requirements of a declining species the chance of detecting the species; this can be captured in a relationship between individual- and species-level detection probability: . Here PI is � the mean probability of detecting an individual at a site,�� = N1 is− the(1 −number��) of animals present at that site, and PS is the resultant species-level detectability (Royle and

Nichols 2003). Where abundance is expected to be constant over time, it is enough to use survey data to estimate species-level detection. However, where abundance changes over time, species-level detection will vary in concert. The effect of abundance on species-level detection is a particular issue for surveys of declining species. Here, detection declines in concert with the decline of the species: survey effort that was once adequate to detect the species becomes inadequate and may lead to mismanagement (Kéry 2002).

Australia is in the midst of the most rapid mammalian extinction crisis in recorded history (Woinarski et al. 2015), and our understanding of threatened species’ current and past distributions is poor (Bilney et al. 2010). Initial historical detections of

Australian species often happened by chance, or as part of broad-scale surveys with minimal per-site survey effort (e.g. Victorian Fisheries and Wildlife Department surveys of the 1970s; Gilmore 1977; Emison et al. 1978). Haphazard survey effort means that many species can go undetected in an area for decades, not because they are absent necessarily, but because no effort is made to survey for them (Burns et al.

2016; Boakes et al. 2016). When resurvey efforts are later conducted and the species is not detected, we struggle to know when or why the species disappeared, and even whether they are truly extirpated or we have just not tried hard enough to find them

(Reed 1996; Burns et al. 2016; Boakes et al. 2016; Woinarski et al. 2018). This has repeatedly been demonstrated in rediscoveries of ‘Lazarus’ species in Australia –

43 Chapter 2: Survey requirements of a declining species species that have gone undetected for so long that they were falsely presumed extinct, such as the Leadbeater’s Possum (Gymnobelideus leadbeateri), found again in 1961 after 50 years, or the Mountain Pygmy Possum (Burramys parvus), known only from fossil records until 1966 (Short and Smith 1994). While historical knowledge gaps are often irreparable, and we can only hypothesize about prior distributions and the timing and causes of decline for many species, we can still make use of these haphazard detection histories to improve future monitoring efforts, and to identify local extinctions with statistical confidence.

The New Holland Mouse (Pseudomys novaehollandiae), a threatened rodent species native to south east Australia, has a classically haphazard detection history. The small

(12-28 g), nocturnal species was first described in 1843, but went entirely undetected for almost a century between 1886 and 1967 (Mahoney and Marlow 1968). It was first captured in the states of Victoria, Tasmania, and Queensland in 1970, 1976, and

1996 respectively (Seebeck and Beste 1970; Hocking 1980; Van Dyck and Lawrie

1997). Like many native Australian rodents, P. novaehollandiae is readily lured by standard mammal bait (rolled oats, peanut butter and golden syrup) and easily trapped when at high local abundance (e.g. McCall et al. 2015), thus its poor early detection history is likely due to lack of effort, rather than it being an innately cryptic species. Resurvey efforts since these initial discoveries indicate that many populations have declined, potentially to local extinction (Wilson 1996; Wilson et al.

2017; Lazenby et al. 2018). Historically, the standard survey effort for P. novaehollandiae was three consecutive nights of live trapping using 30 Elliott traps, with abundance indices drawn from trap success or capture-mark-recapture data (e.g.

Quin 1996; Atkin and Quin 1999; Lock and Wilson 1999; Wilson et al. 2005). As the

44 Chapter 2: Survey requirements of a declining species species declines in abundance, this effort may no longer be sufficient for detection, and low-abundance populations may be mistakenly considered extinct. Further, abundance estimates may be clouded by seasonal fluctuations in detectability, due both to true shifts abundance and to behavioural changes across the year; dispersal of young occurs in Autumn following the late-spring – early-Autumn breeding season

(Wilson 1991; Wilson et al. 2005; Kemper 1980; Fox 1982). With purported local extinctions at 7 of 12 historical regions of occupancy in Victoria (Quin 1996; Wilson et al. 2017), there is an acute need to delineate the species’ current distribution to guide management efforts and test causes of decline.

Here we assess the effects of seasonal and environmental variables on the abundance and detectability of P. novaehollandiae. We define minimum-abundance survey requirements for reasonable confidence in the species’ absence, and identify key factors to consider in survey planning and the interpretation of historical data.

Methods

Survey data

We collated Victorian P. novaehollandiae Elliott trapping survey data spanning 48 years from the species’ initial detection in the state in 1970 to present day 2018, sourced from historical field notes, recent reports, and our own survey work (see S1 for list of surveys and Appendix 1 for published data sources). Briefly, animals were trapped in Elliott small live mammal traps (Elliott Scientific Co., Upwey, Victoria; size 90 mm × 100 mm× 330 mm) baited with a mixture of peanut butter, golden syrup and rolled oats. Each trap contained non-absorbent nesting material for insulation and was wrapped in a plastic tube for protection during poor weather

45 Chapter 2: Survey requirements of a declining species conditions. We recorded nightly trap results as counts of individuals, nightly trap effort, survey dates, and precise locality. Following quality control measures, we included data from 328 surveys of 130 sites (>200 m apart) in south east Melbourne and Gippsland, collected over 349 unique nights, totalling 35,774 trap nights.

Average survey effort was 31 Elliott traps (range: 8-130 traps) set for 3.5 nights

(range: 1-6 nights).

Abundance - detectability modelling

We used the N-mixture model approach of Royle and Nichols (2003) to jointly estimate site-level abundance (where a site represents the surveyed area and not the local population size) and individual-level detectability. This approach uses repeated count data to estimate individual-level detectability (PI in the above formula) and site-level abundance (N), which together affect the site-level detectability of a species

(PS). The number of individuals captured each night is considered to be a draw from a binomial distribution parameterised with the true number of individuals at the site and individual-level detection probability. The “true” number of individuals at a site is treated as a latent variable, drawn from a Poisson distribution, where the expected value of the Poisson is specified with a linear model (and in this case, a log link).

Individual detectability is also specified as a linear model (and a logit link) in which covariates affect detection.

Our abundance model incorporated an intercept, μN, and two fixed effects: (1) survey effort (E: number of Elliott traps per night, as a proxy for site survey area); and (2) to capture annual cycles in abundance, a first-order Fourier function of day of the year

46 Chapter 2: Survey requirements of a declining species

(D: in radians). We also specified random effects of year of survey (yt: 25 levels), site

(si: 130 levels), and region (rj: 8 levels).

� � 1 2 3 � � � log(� ) = � + � � + � Cos(�) + � Sin(�) + � + � + �

Our individual-detectability model comprised an intercept, P, and four fixed effects

(1) moon brightness (M: 0-1; 0 = new moon, 1 = full moon), (2) night of survey (T:

1–n nights), (3) daily rainfall (Rd: 24-hour rainfall prior to and including each survey night 6am – 6am; Bureau of Meteorology, 2018) as continuous effects, and (4) time of year (D). The effect of time of year was again specified as a first-order Fourier function of day of the year (in radians) to specify annual cycles in detection. To assist interpretation, we graphed seasonal detectability as PS rather than PI, thus accounting for seasonal fluctuations in abundance.

logit(��) = �� + �4� + �5� + �6�� + �7Cos(�) + �8Sin(�)

To aid numeric model fitting, we scaled all continuous variables (except day of the year) to have a mean of zero and unit variance. We used a Bayesian framework and sampled priors using ‘JAGS’ v 4.3.0 (Plummer 2017) via the package ‘rjags’

(Plummer 2016) in the program R v 3.4.3 (R Core Team 2017). We set minimally- informative priors (see Table 2.1), excluded a burn-in of 220,000 iterations, and ran three chains for an additional 80,000 iterations sampling every fifth iteration. We assessed convergence through visual inspection of trace plots and the Gelman-Rubin convergence diagnostic (Gelman and Rubin 1992)

47 Chapter 2: Survey requirements of a declining species

Survey design

To compare the minimum survey effort (nights) required to obtain 95% confidence in

P. novaehollandiae absence under different conditions, we calculated probability of false absence (AX) values and plotted these against night of survey. As PI (and so PS) varies across survey night, T, the probability of a false absence is given by: (1-PS,1)

(1-PS,2)…(1-PS,T), where PS,T is the species-level detection on night, T. To achieve

95% confidence in the species’ absence, AX must fall below 0.05. For all calculations, we set the number of traps to 30 Elliott traps (standard historical survey effort) and no rain during survey (most common state). To examine seasonal variation, we

st st calculated AX values at new moon and full moon for the 1 January (summer), 1

May (autumn), and 1st October (spring), at minimum (N = 1) and modelled annual peak average per-site abundance (N = 7).

Results

Abundance-detectability modelling

Our analysis found seasonal fluctuations in abundance (Fig. 2.1a; Table 2.1), corresponding with the influx of new recruits during the breeding season over summer (Wilson 1991; Wilson et al. 2005; Kemper 1980; Fox 1982). Increasing number of Elliott traps set (reflective of increasing the area surveyed) had minimal effect on observed abundance, with CIs overlapping zero. An increase from eight traps to 100 traps provided an average increase in abundance of two animals per site

(Fig. 2.1b). Much of the variability in abundance was explained by differences among years, sites, and regions (Table 2.1).

48 Chapter 2: Survey requirements of a declining species a)

Figure 2.1. Abundance estimates for Pseudomys novaehollandiae derived from

N-mixture models a) seasonally; and b) with an increasing number of Elliott

traps used. Shading indicates 95% prediction envelopes. Parameters set to

49 Chapter 2: Survey requirements of a declining species

mean values for the calculation of each other parameter. Raw capture data

(captures per night) shown as purple circles (n = 1150).

Table 2.1. Parameter estimates from N-mixture model for Pseudomys

novaehollandiae, derived from a 48-year live-trapping dataset in Victoria,

Australia. Minimally informative priors specified where N(mean, sd) is the

normal distribution; G(alpha, beta) is the Gamma distribution.

Abundance covariates (log scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (N) N (0, 1000) 0.819 0.065 -0.277 1.881

Trap effort (β1) N (0, 1000) 0.049 0.001 -0.066 0.166

Day of year (cos) (β2) N (0, 1000) 0.967 0.018 0.483 1.572

Day of year (sin) (β3) N (0, 1000) 0.682 0.006 0.412 0.978 2 Variance due to Region ( r) G (0.001, 0.001) 1.979 0.037 0.281 6.348 2 Variance due to Site ( s) G (0.001, 0.001) 0.958 0.002 0.645 1.359

2 Variance due to Year ( y) G (0.001, 0.001) 1.808 0.007 0.777 3.411 Detection covariates (logit scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (P) N (0, 1000) -0.256 0.037 -1.324 0.791

Moon (β4) N (0, 1000) -0.434 0.004 -0.627 -0.301

Night of survey (β5) N (0, 1000) 0.17 0.001 0.079 0.277

Nightly rainfall (β6) N (0, 1000) 0.129 0.001 0.052 0.225

Day of year (cos) (β7) N (0, 1000) -2.027 0.008 -2.643 -1.349

Day of year (sin) (β8) N (0, 1000) -1.058 0.007 -1.516 -0.593

Species-level detectability (PS; incorporating seasonal fluctuations in abundance) showed considerable variation within and across seasons, with notably lower detection probability during summer months (Fig. 2.2a; Table 2.1). Brighter moon phases corresponded to lower individual-level detectability (PI; Fig. 2.2b). PI

50 Chapter 2: Survey requirements of a declining species improved over the course of a survey period (Fig. 2.2c), and increased with rainfall during a survey (Fig. 2.2d). a) Captures perCaptures night

Detectability

Day of year

Captures perCaptures night

Detectability

Moon phase

51 Chapter 2: Survey requirements of a declining species c)

Captures per night Captures

Detectability

Rainfall (mm) d)

Captures per night Captures

Detectability

Night of survey Figure 2.2. Detectability estimates for Pseudomys novaehollandiae derived

from N-mixture models a) at a species-level (PS; accounting for seasonal

fluctuations in abundance) seasonally; and at an individual-level (PI) b) during

different moon phases, c) with increasing rainfall during a survey, and d) over

the course of a survey (per night, not cumulative). Shading indicates 95%

52 Chapter 2: Survey requirements of a declining species

prediction envelopes. Parameters set to mean values for the calculation of each

other parameter. Raw capture data (captures per night) shown as purple circles

(n = 1150).

Survey design

The survey effort required to reach 95% confidence in the absence of P. novaehollandiae varied considerably with abundance, season, and moon phase. We derived AX values from the abundance-detectability model for 30 Elliott traps with no rainfall during the survey. With per-site abundance at seven individuals, a single night of survey around a new moon in autumn or spring would have a >95% probability of detecting the species. However, at minimum abundance (one animal per site), the number of consecutive nights survey required for 95% confidence in absence was up to five nights at a new moon in May, and greater in other moon phases (Table 2.2; Fig. 2.3). Summer surveys and surveys around full moons consistently required impractically high numbers of consecutive nights trapping for confidence in the species’ absence (Table 2.2; Fig. 2.3).

53 Chapter 2: Survey requirements of a declining species

Minimum abundance (N = 1) Peak mean abundance (N = 7) new moon full moon new moon full moon 1st January Lower bound 22 >30 6 23 Mean 11 23 3 7 Upper bound 4 8 1 2 1st May Lower bound 5 18 1 5 Mean 3 8 1 2 Upper bound 2 3 1 1 1st October Lower bound 3 7 1 2 Mean 2 4 1 1 Upper bound 2 2 1 1

54 Chapter 2: Survey requirements of a declining species

Minimumabundance Peakmeanabundance y r a u n a J t s 1 y a X M A t s 1 r e b o t c O t s 1

Nightofsurvey

Figure 2.3. Probability of species non-detection (AX) in different seasons and with different moon phases given the presence of Pseudomys novaehollandiae at a site-level abundance of N = 1 (minimum) or N = 7 (annual peak mean abundance). For each month, the predictions for the new moon are shown in grey (darker), and the full moon in blue (lighter), with overlap between the two depicted in blue-grey. Blue and grey lines indicate mean model predictions and shading indicates 95% prediction envelopes. All models assumed 30 Elliott traps with no rain. Red lines indicate AX = 0.05 (horizontal) and standard historical survey effort (vertical; 3 nights).

55 Chapter 2: Survey requirements of a declining species

Discussion

Species’ detectability is intimately related to abundance, and as species decline, previously adequate survey effort can become inadequate (Kéry 2002; Royle and

Nichols 2003; McCarthy et al. 2013). Here we used a 48-year trapping dataset to examine how abundance and detectability varies with site and survey conditions. Our analysis demonstrates considerable variation in the abundance and detectability of the threatened and declining P. novaehollandiae. Our modelling makes inference on individual-level detection, which allows us to calculate survey effort required to detect the species when it is at minimum abundance (one individual at the site) and mean maximum abundance (approximately seven individuals per site). This minimum-abundance survey effort is the appropriate metric if we are to be confident that the species is not present, and is substantially higher than the effort required to detect the species when it is at its mean maximum abundance. Failing to account for how detectability may decrease as populations decline in abundance can massively underestimate the survey effort required to achieve confidence in a species’ absence.

This bias may cause us to falsely affirm the species’ absence at a site and underestimate the species’ occupancy at a regional scale (Thompson 2002; Mazerolle et al. 2007; Cubaynes et al. 2010).

While we can calculate species-level detectability at fixed abundance, abundance itself is a dynamic variable, changing in response to environmental conditions.

Foremost for P. novaehollandiae is the effect of season on abundance. Abundance peaked in late summer (Fig. 2.1a), coinciding with the late stages of the breeding season and recruitment of new individuals into the population, and declined through

56 Chapter 2: Survey requirements of a declining species winter as individuals died (Wilson 1991; Wilson et al. 2005). However, seasonal effects on behaviour complicate the direct link between abundance and detectability; individual-level detectability fell dramatically over summer, so even accounting for the higher abundance, species-level detectability was lowest over summer (Fig.

2.2a). This decline in detectability from November through February is commonly observed in the species, in particular, adult male individuals regularly disappear from the trappable population in late spring, then reappear as second-year males in mid- autumn (Kemper 1980; Fox 1982; Wilson 1991; Wilson et al. 2005). Poor detectability in summer, despite high abundance, leaves mid-autumn to mid-spring as the ideal survey period for P. novaehollandiae.

Independent of abundance and season, individual-level (and so species-level) detection also varied across survey conditions. Detectability varied with moon phase and rainfall, and improved as surveys progressed. After seasonal shifts, moon phase had the strongest effect on the detectability of P. novaehollandiae. In some seasons, surveys conducted during the week of a full moon would require up to five times more survey nights to reach 95% confidence in P. novaehollandiae absence, compared with an otherwise identical survey around a new moon. This is consistent with findings that small mammal activity levels increase on moonless nights and is likely a response to the increased risk of predation on brightly illuminated nights

(Vickery and Bider 1981; Prugh and Golden 2014; Gilmore 2016). Our finding that

P. novaehollandiae detectability improved with increasing rainfall during a survey may have a similar explanation in that cloud cover can reduce illumination. The inter-survey variation in detectability we observed, particularly with relation to moon phase, led to substantial variation in appropriate survey effort; variation that, if

57 Chapter 2: Survey requirements of a declining species ignored, risks considerable wasted survey effort or false absence results. By contrast, understanding patterns of variation in detectability allows us to design efficient monitoring and detection programs.

Given the considerations around naturally-varying abundance and detectability, how should we allocate survey effort for P. novaehollandiae? There are two main survey types to consider: occupancy surveys, where only the species’ presence or absence is of interest; and demographic surveys, where more detailed questions are being asked about abundance, breeding, diet, genetics, etc. Clearly, both survey types should avoid the week around the full moon, because reduced activity here will give low detection rates. For occupancy surveys, trapping is best conducted during the peak species-level detection period in late autumn or mid-spring and should be avoided over the summer months. Where a species’ abundance or detectability is highly variable, as our broad prediction envelopes indicate is the case here, the lowest values should be assumed before non-detection data is interpreted as absence (Reed

1996). Therefore, in defining minimum survey effort, we derive values from the lower end of model predictions. Provided occupancy surveys are conducted at the new moon (irrespective of rainfall), a minimum of five (late-Autumn) or three (mid-

Spring) consecutive nights of survey are necessary to achieve statistical confidence in the species probable absence (a minimum-abundance survey effort). For demographic studies, timing is often determined by the research question; however, surveys should be wisely timed to work with seasonal fluctuations in detectability and abundance. Long-term, multi-year demographic studies require consistency in season(s) of survey to allow comparisons between years; and studies of short-term responses (i.e. to disturbance such as fire) must account for the effect of changing

58 Chapter 2: Survey requirements of a declining species season underneath any perceived response. Furthermore, abundance estimates based on capture numbers should be interpreted in the context of detectability estimates, e.g. via N-mixture models as used here.

Except for times and places of highest detectability, historical resurvey efforts for P. novaehollandiae in Victoria fell well below our newly defined minimum-abundance survey requirements. For instance, broad-scale surveys across historical and hypothesised P. novaehollandiae sites in Victoria in the 1990s typically used 30

Elliott traps per site for three nights in Autumn (Quin 1996; Quin and Williamson

1996; Reside and Hooper 1999). While this survey effort is sufficient if the species is at high abundance, we can now determine that it is an inadequate effort at low abundance. So, in those surveys and almost all others, survey effort has not been adequate to confidently assert absence or local extinction of P. novaehollandiae. This does not mean that the species has not declined or suffered local extinction. Habitat alteration and change in land use make persistence in many locations implausible, and historically-adequate effort that no longer detects a species is an indication of probable decline. Rather, it means that the species may well persist (at low densities) in places where it is currently considered locally extinct. Further surveys are necessary before extinctions can be confirmed in several historical locations.

Failing to account for decreasing detectability as a species declines may contribute to premature assertion of local extinction and non-detection of novel populations, thus undermining conservation efforts. Given Australia’s current mammalian extinction crisis, delays in, or a lack of, threat mitigation and appropriate management of undetected and declining populations could rapidly exacerbate a species’ risk of

59 Chapter 2: Survey requirements of a declining species extinction. This study highlights the enormous intra-specific heterogeneity in detectability, and how incorporating species- and survey-specific detectability estimates into future occupancy and demographic survey design can contribute to survey efficacy.

Data Accessibility Statement

Trapping data available online at: https://figshare.com/s/f324d1b995a45c4a1b2c

60 Chapter 2: Survey requirements of a declining species

Appendix 2.1 – Data sources

Atkin, B. W., & Quin, B. R. (1999). New Holland mouse Pseudomys novaehollandiae (Rodentia: Muridae): Further

findings at Yanakie Isthmus, Wilsons Promontory National Park. The Victorian Naturalist, 116(5), 169–172.

Braithwaite, R. W., & Gullan, P. K. (1978). Habitat selection by small mammals in a Victorian heathland. Austral

Ecology, 3(1), 109–127. https://doi.org/10.1111/j.1442-9993.1978.tb00857.x

Burns, P. A. (2016). Assessing the status of New Holland mouse (Pseudomys novaehollandiae) on Yanakie Isthmus,

Wilsons Promontory National Park (Report to Parks Victoria). Melbourne, Victoria: University of Melbourne.

Gilmore, A. M. (1977). A survey of vertebrate animals in the Stradbroke area of South Gippsland, Victoria. Victorian

Naturalist, 94, 123–128.

Hollis, G. J. (1996). Survey for the New Holland Mouse (Pseudomys novaehollandiae) in Gippsland Lakes Coastal Park

and Providence Ponds Flora and Fauna Reserve, Gippsland Area, Victoria. 1996. Warragul, Victoria: Department

of Natural Resources and Environment.

Hollis, G. J. (1999). Pre- and post-fire trapping in Gippsland Lakes Coastal Park: an update on monitoring of Pseudomys

novaehollandiae and other small mammals. Warragul, Victoria: Department of Natural Resources and

Environment.

McCall, C., Reside, J., & Collyer, A. (2015). An assessment of the status of New Holland mouse populations in

Gippsland: surveys for the New Holland mouse at Providence Ponds Flora and Fauna Reserve and Gippsland

Lakes Coastal Park. Autumn 2015. Bairnsdale, Victoria: Wildlife Unlimited Pty Ltd.

McCall, C., Reside, J., van Rooyen, A., & Weeks, A. (2014). Surveys for the New Holland Mouse at Dutson Downs:

Genetic Sampling, 2013-2014. Wildlife Unlimited Pty Ltd and cesar Pty Ltd.

Quin, B. R. (1996). New Holland Mouse Pseudomys novaehollandiae (Rodentia: Muridae) in South Gippsland, southern

Victoria part one - distribution and status. The Victorian Naturalist, 113(5), 236–246.

Quin, B. R., & Williamson, R. C. (1996). New Holland Mouse Pseudomys novaehollandiae (Rodentia: Muridae) in

South Gippsland, southern Victoria part two - conservation and management. The Victorian Naturalist, 113(6),

281–288.

Reside, J., & Hooper, D. (1999). Surveys for the New Holland Mouse (Pseudomys novaehollandiae) in Gippsland.

Bairnsdale, Victoria: Wildlife Unlimited Pty Ltd.

Seebeck, J. H., & Beste, H. J. (1970). First Record of the New Holland Mouse (Pseudomys novaehollandiae

(Waterhouse, 1853)) in Victoria. The Victorian Naturalist, 87(10), 280–287.

Wilson, B. A., White, N. M., Hanley, A., & Tidey, D. L. (2005). Population fluctuations of the New Holland mouse

Pseudomys novaehollandiae at Wilson’s Promontory National Park, Victoria. Australian Mammalogy, 27(1), 49–

60. https://doi.org/10.1071/AM05049

61 Chapter 3: Identifying Pseudomys in camera trap images

Chapter 3 Identification of threatened rodent species using infrared and white-flash camera traps

Burns, P. A., Parrott, M. L., Rowe, K. C. and Phillips, B. L. (2018). Identification of threatened Pseudomys species using infrared and white-flash camera surveys.

Australian Mammalogy 40, 188–197 (DOI 10.1071/AM17016).

Manuscript received 18 March 2017, accepted 24 July 2017

62 Chapter 3: Identifying Pseudomys in camera trap images

Abstract

Camera trapping has evolved into an efficient technique for gathering presence/absence data for many species; however, smaller mammals such as rodents are often difficult to identify in images. Identification is inhibited by co-occurrence with similar-sized small mammal species and by camera set-ups that do not provide adequate image quality. Here we describe survey procedures for identification of two small, threatened rodent species – Smoky Mouse (Pseudomys fumeus) and New

Holland Mouse (P. novaehollandiae) – using white-flash and infrared camera traps.

We tested whether observers could accurately identify each species and whether experience with small mammals influenced accuracy. Pseudomys fumeus was ~20 times more likely to be correctly identified on white-flash images than infrared, and observer experience affected accuracy only for infrared images, where it accounted for all observer variance. Misidentifications of P. novaehollandiae were more common across both flash types: false positives (>0.21) were more common than false negatives (<0.09), and experience accounted for only 31% of variance in observer accuracy. For this species, accurate identification appears to be, in part, an innate skill. Nonetheless, using an appropriate setup, camera trapping clearly has potential to provide broad-scale occurrence data for these and other small mammal species.

Introduction

Trapping is a vital tool of small mammal research, monitoring, and management; it provides access to specimens and genetic samples, and data on abundance, demographics, and reproduction (Clemann et al. 2014). However, when species are rare, their distributions are often poorly resolved and their habitat requirements ill-

63 Chapter 3: Identifying Pseudomys in camera trap images defined, so their presence at a site is the primary factor of interest. In these situations, and because trapping surveys across large geographic areas can require a cumbersome amount of time and effort (Nelson et al. 2009; De Bondi et al. 2010), camera trap surveys have emerged as an increasingly cost-effective method for gathering presence/absence data (De Bondi et al. 2010; Roberts 2011; Meek et al.

2015b). Camera traps may also provide substantially higher detection rates for some species (Roberts 2011), and detect a broader range of species (Paull et al. 2012).

Camera traps can be used to identify areas where a species is present, allowing those areas to be targeted for live trapping to obtain more detailed data where required.

However, camera trapping relies on the identification of species from images alone and this can be challenging. Image quality, inappropriate camera trap and bait configuration, and co-occurrence with non-target species that are similar in appearance to target species can lead to misidentification (Meek et al. 2013; Meek et al. 2015a; Meek and Vernes 2015), limit identifications to genus level (Swan et al.

2014b), or force studies to rely on estimates of species identity (Villette et al. 2016).

When a target species co-occurs with other species similar in appearance, the identification characteristics used to distinguish the target need to be sensitive enough to detect when it is present in the images (i.e. prevent false absences), yet specific enough to prevent false presences. Here we are specifically considering whether an observer can detect the species in an image set; whether or not the camera trap detects the species during the survey is a separate matter (e.g. Ballard et al.

2014; McDonald et al. 2015). False presences and false absences are troubling for threatened species management, although the relative costs and risks associated with each type of error differ among species (Field et al. 2004). Mistakenly asserting that

64 Chapter 3: Identifying Pseudomys in camera trap images a species is absent can result in a failure to manage the population or protect the area

(Taylor and Gerrodette 1993), bias species’ distribution models used to inform broader management decisions (Tyre et al. 2003; Gu and Swihart 2004), or cause an underestimation of the species’ stability. The latter may draw resources from other species that are actually at greater risk, while the former issues may contribute to the realisation of the species’ local extinction. False presences can result in an overestimation of the security of a species, or misallocation of limited management resources to areas where the species cannot actually benefit (Chadès et al. 2008). For particularly valuable or threatened species, the costs of false presences are greatly outweighed by the risks of false absences, such that the optimal strategy can be to assume the species is present and manage accordingly (Chadès et al. 2008).

The Smoky Mouse (Pseudomys fumeus) and New Holland Mouse (P. novaehollandiae) are species of national conservation concern (endangered and vulnerable respectively under the Environment Protection and Biodiversity

Conservation Act 1999; Department of Enviroment and Energy 2017) that are endemic to south-eastern Australia. Local extirpations and significant range contractions have been reported in both species in recent decades (Quin 1996; Quin and Williamson 1996; Lock 2005; Burns et al. 2016). Their current distributions remain uncertain, however, as no local extirpations have been verified with statistically robust measures of detectability.

Comprehensive assessment of each species’ current distribution, habitat requirements and threatening factors are necessary to enable targeted conservation management.

Camera trapping may facilitate broad-scale surveys for the species, but only if we

65 Chapter 3: Identifying Pseudomys in camera trap images can actually identify each species from the images. Another Pseudomys species, the

Hastings River Mouse (P. oralis), is not distinguishable on camera trap images from the Bush Rat (Rattus fuscipes) with reasonable confidence due to similarities in size, shape and pelage colouration (Meek and Vernes 2015). Pseudomys fumeus has distinctive colouration that enables accurate identification when using white-flash camera traps (Nelson et al. 2009; Nelson et al. 2010; Holmes 2012). However, P. fumeus is similar in size and shape to R. fuscipes and even juvenile Black Rats (R. rattus), meaning that in infrared imagery, where colouration cannot be used for identification, body shape may not be distinctive enough to allow identification

(Nelson et al. 2009). For P. novaehollandiae, confusion with the invasive house mouse (Mus musculus) and R. fuscipes is a key risk. Mus musculus occurs throughout the range of P. novaehollandiae and is similar in size and appearance, such that in- hand misidentification of M. musculus as P. novaehollandiae was reported in 1992

(Quin 1996).

Here, we aim to develop guidelines for the use of infrared and white-flash camera traps in identifying P. fumeus and P. novaehollandiae, and for distinguishing them from co-occurring rodent species. We test whether observers can identify the species with adequate sensitivity (low false absence rate) and specificity (low false presence rate), and thus whether camera trapping is a suitable tool for broad-scale surveys of

P. fumeus and P. novaehollandiae. We expect that P. fumeus will be more readily identified on white-flash camera trap images than infrared, whereas flash type will have minimal bearing on P. novaehollandiae identification accuracy. Additionally, we test whether observer experience impacts identification accuracy. Overall, we would expect that observers with in-hand experience of either species will have

66 Chapter 3: Identifying Pseudomys in camera trap images greater accuracy in identifying images of the relevant species than those with no experience.

Materials and methods

Study sites

Surveys were conducted at two locations in Victoria, Australia, selected for recent records of P. fumeus or P. novaehollandiae: the Victoria Range, Grampians–Gariwerd

National Park (GGNP; 37°20S, 142°17E), where P. fumeus has been intermittently detected with varied trap success over four decades (Burns et al. 2015); and

Gippsland Lakes Coastal Park (GLCP; 38°6S, 147°29E), where P. novaehollandiae has historically been detected with consistently high trap success (Hollis 1999;

McCall et al. 2015).

Camera trap surveys

In April 2016, we set cameras at four sites in GGNP where we had detected P. fumeus in Elliott traps a month earlier. At each site, we established three bait stations

50–100 m apart (Fig. 3.1). Following Nelson et al. (2009), each bait station consisted of a metal cage (modified cutlery drainer) slotted onto a plastic garden stake with a

100 × 100 mm sheet metal cap (i.e. ant cap, as used in building construction) fixed to the top of the cage for protection from rain (Fig. 3.1). Within each cutlery drainer, we placed six tea strainers filled with peanut butter, oats, golden syrup and vanilla essence as an olfactory lure. This elevated bait station design was chosen to provide profile and ventral images with a clear scale, while showcasing body shape as the animal reaches up or climbs the station. Bait stations were placed so that the bottom of each cutlery drainer was ~20 cm from the ground. We placed two cameras ~20 cm

67 Chapter 3: Identifying Pseudomys in camera trap images above the ground and 1 m from each bait station. The first camera was a

PixController DigitalEye white-flash camera trap (PixController Inc.: Pittsburgh, PA) set to maximum sensitivity, 1-m focus, one image per trigger. The second camera was a Reconyx HC500 Hyperfire infrared camera trap (Reconyx: Holman, WI) set to

‘Trail’ mode settings with ‘High Quality’ night mode, three images per trigger, and high sensitivity. ‘Fast shutter’ night mode may be better suited to small mammal surveys (Meek et al. 2015a). We selected camera models based on what was freely available for loan and would provide adequate image quality. However, previous studies have reported that PixControllers are less reliable with shorter battery lives and higher failure rates than Reconyx (Meek and Vernes 2015), which may result in more site-level false negatives in a field setting. Images produced by PixControllers are equal to or slightly higher-quality than those from Reconyx HC series. Thus, for our study design based on reviewing selected images with known species composition from white-flash and infrared cameras, the issues of reliability of

PixControllers is of limited concern.

In May 2015, we set cameras at 11 sites where P. novaehollandiae had been detected in live traps in the previous month (McCall et al. 2015). At each site, we established two elevated bait stations 50–100 m apart and set one PixController Digital Eye and one Reconyx HC500 Hyperfire at each station. Bait station set-up and camera trap settings were as above.

68 Chapter 3: Identifying Pseudomys in camera trap images

Fig. 3.1. Camera trap bait station set-up featuring P. novaehollandiae,

Providence Ponds, Victoria. Image captured using a PixController DigitalEye

white-flash camera trap.

After 14–17 nights, we retrieved the 18 and 44 camera traps from GGNP and GLCP and downloaded the images. PAB identified the mammal species present in each infrared and white-flash image independently using extensive experience with small mammal identifications. By the completion of this study PAB had identified 240 in- hand captures of P. fumeus and 350 P. novaehollandiae among more than 2000 small mammal captures. Additionally, PAB had identified more than 360 000 images from camera traps specifically targeting P. fumeus or P. novaehollandiae. During the initial stages of identification characteristic development, PAB verified infrared

69 Chapter 3: Identifying Pseudomys in camera trap images identifications against identifications of white-flash images with similar time stamps from the same bait station.

Identifying characteristics

Across most of its range, P. fumeus co-occurs with, and is most likely to be mistaken in camera trap images for, two Rattus species – R. rattus and R. fuscipes. While P. fumeus is smaller than adult Rattus, head and body shape can be similar to that of R. fuscipes and juvenile R. rattus, thus scale is useful only if there are no juvenile Rattus in the study area at the time of survey. We collected P. fumeus images for this trial from an area where R. fuscipes is not present to reduce uncertainty about species identity. The smaller species, P. novaehollandiae, is most likely to be mistaken for M. musculus. Head and body shape, pelage colouration and tail morphology are key characteristics for distinguishing between P. fumeus, P. novaehollandiae and other species (Table 3.1; Fig. 3.2a–j). Non-rodent small mammal species that may co- occur with P. fumeus and P. novaehollandiae (Antechinus sp., Sminthopsis sp.,

Cercartetus sp.) have distinctly pointed snouts that guide discernment from

Pseudomys. Similarly, the swamp rat (R. lutreolus) has a blunt snout and a stocky build, which allow rapid discernment from Pseudomys. An image set to aid identifications can be found in Fig. 3.S1 available as Supplementary material to this paper.

70 Chapter 3: Identifying Pseudomys in camera trap images

Table 3.1. Identifying characteristics for three native Australian rodent species

(P. novaehollandiae, P. fumeus, R. fuscipes) and two invasive rodent species (R.

rattus, M. musculus) on white-flash and infrared camera trap images

P. fumeus P. novaehollandiae M. musculus R. fuscipes R. rattus Nose shape Snubby Snubby Pointed Snubby Pointed Neck Medium Thick Slim Thick Slim Body shape Long, heavy rump Cylindrical, sausage- Tapers towards Cylindrical Heavy rump shaped head Tail Slender, >head– Slender, >head–body Slender, >head– Thick, =head– Thick, >head– body length length body length body length body length Ears Medium Small Small Small, rounded Large

White-flash colouration Ventral White White Cream to brown White–cream Cream–black Dorsal and Soft grey Sandy with dark Most often solid Sandy–brown Cream–black lateral guard hairs most brown notable along dorsal midline Tail White ventral, White ventral, sandy Dark, solid colour Solid brown Solid colour grey dorsal, can dorsal appear pink Feet White fur, pink White fur, pink foot Dark, solid colour Cream–brown Cream–black foot pads pads

71 Chapter 3: Identifying Pseudomys in camera trap images

Figure. 3.2. Infrared and white-flash camera images of (a, b) P. fumeus, (c, d)

P. novaehollandiae, (e, f) M. musculus, (g, h) R. fuscipes, and (i, j) R. rattus.

72 Chapter 3: Identifying Pseudomys in camera trap images

Identification trial

We conducted an identification trial to (1) test whether P. fumeus and P. novaehollandiae are identifiable on camera trap images by people with a range of experience, and (2) to test whether camera trap type affects identification accuracy.

We provided observers with a PDF guide and a series of infrared and white-flash images. Each observer identified and recorded the animals sighted in each image, noted the time taken to identify all images in the series, and recorded their relevant experience with small mammals and camera trapping. Each observer in the P. fumeus trial (n = 12) identified one of three sets containing 63 infrared images (21 triggers) and 50 white-flash images (50 triggers). Observers marked each image as ‘smoky mouse’, ‘other’, ‘unsure’, or ‘empty’. Other species present in the image sets included: R. rattus, R. fuscipes (images from Wilsons Promontory National Park as

R. fuscipes are not present in GGNP), M. musculus, Agile Antechinus (A. agilis) and

Dusky Antechinus (A. mimetes). Each observer in the P. novaehollandiae trial (n =

15) identified one of three sets containing 96 infrared images (32 triggers) and 50 white-flash images (50 triggers). Observers marked each image as ‘New Holland mouse’, ‘house mouse’, ‘other’, ‘unsure’, or ‘empty’. Species present also included the eastern pygmy possum (C. nanus).

We compared each completed ID set with PAB’s identifications (which for these purposes are assumed to be the correct identifications) and recorded discrepancies between observers’ identifications and PAB’s. Using the images for which an ID was made by the observer and by PAB (i.e. not marked by either as ‘unsure’ or ‘empty’), we constructed contingency tables to calculate the sensitivity, specificity, positive

73 Chapter 3: Identifying Pseudomys in camera trap images and negative predictive values, and error rates of the identification characteristics as applied by the observers for each species and each camera trap type. Sensitivity is the proportion of images of the target species that are correctly identified, whereas specificity is the proportion of images of non-target species that are correctly identified as non-target species. The positive predictive value is the proportion of positive identifications that were correct and the negative predictive value is the proportion of negative identifications that were correct.

To test the significance of observer experience and flash type on identification success for each species, we built binomial generalised linear mixed-effects models in the program R 3.3.2 (R Core Team 2016), using the ‘GLMER’ function in the package ‘lme4’. We incorporated flash type, the observer’s level of experience with identifying (1) the target species in-hand, (2) small mammals in-hand, and (3) small mammals on camera as fixed effects, and observer nested within image set as random effects. We excluded a fourth measure of experience – general camera trap image identification – because it was highly correlated with small mammal camera trap image identification. Additionally, for the P. novaehollandiae model, we included whether or not the observer had previously completed a guided P. novaehollandiae identification quiz as a fixed effect. We ran simpler models incorporating only flash type and observer within set to determine the amount of observer variance accounted for by experience. To test the effect of experience on identification success for each flash type and each species, we also ran the above models divided into flash type for each species. To quantify the effects of flash type tested in the models, we calculated the relative risk of misidentifying an image when white-flash, rather than infrared flash, was used for each species.

74 Chapter 3: Identifying Pseudomys in camera trap images

Results

Surveys

At GLCP, we detected eight species of mammal in 166 235 images collected from

420 camera trap-nights across 11 sites. We detected P. novaehollandiae at all 11 sites and M. musculus at 10 sites. Trap-nights varied among cameras because of camera failure (n = 3 PixController) and flat batteries (n = 13 PixController).

At GGNP, we detected seven species of mammal in 21 406 images collected from

368 camera trap-nights across four sites. We detected P. fumeus at three sites, R. rattus at three sites, and M. musculus and A. agilis at all four sites. Trap effort varied among cameras because of camera failure (n = 1 PixController) and flat batteries (n =

1 PixController). Across all sites, most bait station visits resulted in multiple images of each animal, facilitating accurate identifications in varying poses. A small proportion of images were too blurred or obscured for identification.

Identification trial

Pseudomys fumeus

Observers (n = 12) readily identified P. fumeus on white-flash images with near-zero error rates (<0.01; high specificity and sensitivity; Table 3.2). On infrared images, observers reported moderate false negative identifications (0.24; moderate sensitivity) and low false positive identifications (0.08; high specificity; Table 3.2).

White-flash images were identified with significantly greater accuracy than infrared images (z = 6.021, P < 0.001). The relative risk (RR) of misidentification for white- flash images was approximately one-twentieth that for infrared images (RR = 0.053,

95% CI = 0.018–0.157). Increased experience identifying P. fumeus in-hand (z =

75 Chapter 3: Identifying Pseudomys in camera trap images

4.710, P < 0.001) or other small mammal species on camera (z = 5.540, P < 0.001) significantly improved identification accuracy (Table 3.3). Conversely, increased experience with other small mammal species in-hand significantly lowered identification accuracy (z = –4.033, P < 0.001; Table 3.3). Comparison of models with and without experience covariates suggested that experience accounts for 100% of observer variance (Table 3.3). However, when considering only white-flash camera traps we found no significant effect of experience on identification accuracy for P. fumeus, and no measurable difference among observers – as expected, given the remarkably low error rates reported above (Table 3.3). Thus, all observer variance in the combined model is derived from the infrared images, 100% of which is again explained by observer experience (Table 3.3).

Table 3.2. Sensitivity, specificity, predictive values and error rates for

identification of P. fumeus and P. novaehollandiae. For P. fumeus, observers =

12, identifications = 1131. For P. novaehollandiae, observers = 15,

identifications = 1147

P. fumeus P. novaehollandiae

White-flash Infrared White-flash Infrared Sensitivity 1.000 0.765 0.963 0.911 Specificity 0.993 0.918 0.725 0.794 Positive predictive value 0.954 0.682 0.724 0.943 Negative predictive value 1.000 0.944 0.963 0.704 False positive error rate 0.007 0.082 0.275 0.206 False negative error rate 0.000 0.235 0.037 0.089

76 Chapter 3: Identifying Pseudomys in camera trap images

Table 3.3. Parameter estimates for a binomial mixed-effects model of factors affecting identification accuracy for P. fumeus using white-flash and infrared flash. Six models are compared: combined white-flash and infrared, white- flash only, and infrared only, all with and without explanatory variables relating to observer experience. Parameter estimates are reported on the logit scale. *, P < 0.05; **, P < 0.01; ***, P < 0.001

Estimate s.e. z-value P Full model Intercept (observer) 2.761 1.417 1.949 0.051 Flash type (white) 3.671 0.610 6.021 1.73e-09*** P. fumeus in-hand ID experience 1.514 0.321 4.710 2.47e-06*** Small mammal in-hand ID experience –1.396 0.346 –4.033 5.51e-05*** Small mammal camera ID experience 3.066 0.553 5.540 3.03e-08*** Observer var. = 0.000; Set var. = 5.422; deviance = 307.9; d.f. = 1124 Simple model Intercept (observer) 3.790 1.038 3.653 2.60e-04*** Flash type (white) 3.683 0.615 5.988 2.13e-09*** Observer var. = 6.091; set var. = 0.607; deviance = 324.6; d.f. = 1127 Residual deviance = 530.53, residual d.f. = 1130 White-flash full model Intercept (set/observer) 4.962 1.136 4.368 1.25e-05*** P. fumeus in-hand ID experience 0.476 1.014 0.469 0.645 Small mammal in-hand ID experience 0.407 0.884 0.460 0.639 Small mammal camera ID experience –0.582 0.878 –0.663 0.508 Observer var. = 0.000; set var. = 0.603; deviance = 35.7; d.f. = 509 White-flash simple model Intercept (set/observer) 5.140 0.579 8.876 <2e-16*** Observer var. = 0.000; set var. = 0.000; deviance = 36.9; d.f. = 512 Residual deviance = 36.856; residual d.f. = 514 Infrared flash full model Intercept (set/observer) 3.827 2.298 1.666 0.096 P. fumeus in-hand ID experience 2.127 0.737 2.886 0.004** Small mammal in-hand ID experience –2.240 0.976 –2.295 0.022* Small mammal camera ID experience 4.490 1.487 3.019 0.003** Observer var. = 0.000; set var. = 10.990; deviance = 252.7; d.f. = 610 Infrared flash simple model Intercept (set/observer) 5.684 2.893 1.965 0.049* Observer var. = 17.420; set var. = 3.373e-08; deviance = 270.4; d.f. = 613 Residual deviance = 427.91; residual d.f. = 615

77 Chapter 3: Identifying Pseudomys in camera trap images

Pseudomys novaehollandiae

Observers (n = 15) identified P. novaehollandiae with slightly lower false negative rates (higher sensitivity) on white-flash (0.037) than on infrared images (0.089), but with higher false positive rates (lower specificity; white-flash = 0.275, infrared =

0.206; Table 3.2). Overall, infrared images were identified with significantly greater accuracy than white-flash images (z = –2.372, P = 0.018; Table 3.4). The relative risk of misidentification for white-flash images was 1.5 times that for infrared images

(RR = 1.50, 95% CI = 1.13–2.00). Increased experience identifying P. novaehollandiae or other small mammal species on camera or in-hand did not significantly improve identification accuracy. However, comparison of models with and without experience covariates suggests that experience accounts for ~31% of observer variance (Table 3.4). For white-flash images alone, experience accounted for ~58% of variation among observers, and we found a significant positive correlation between experience identifying other small mammal species on camera and identification accuracy for P. novaehollandiae (z = 2.168, P = 0.030; Table 3.4).

For infrared images, experience accounted for 41% of variation among observers although we found no significant correlation between experience and identification accuracy (Table 3.4).

78 Chapter 3: Identifying Pseudomys in camera trap images

Table 3.4. Parameter estimates for a binomial mixed-effects model of factors affecting P. novaehollandiae identification accuracy using white-flash and infrared flash. Six models are compared: combined white-flash and infrared, white-flash only, and infrared only, all with (full model) and without (simple model) explanatory variables relating to observer experience. Parameter estimates are reported on the logit scale. *, P < 0.05; **, P < 0.01; ***, P <

0.001

Estimate s.e. z-value P Full model Intercept (set/observer) 1.680 0.554 3.031 0.002** Flash type (white) –0.435 0.183 –2.372 0.018* P. novaehollandiae in-hand ID experience 0.141 0.517 0.273 0.785 Small mammal in-hand ID experience 0.040 0.292 0.136 0.892 Small mammal camera ID experience 0.406 0.368 1.105 0.269 P. novaehollandiae ID Quiz completion 0.088 0.771 0.114 0.910 Observer var. = 0.865; set var. = 0.280; deviance = 833.2; d.f. = 1139 Simple model Intercept (observer) 2.283 0.461 4.920 8.64e-07*** Flash type (white) –0.432 0.183 –2.362 0.018* Observer var. = 1.260; Set var. = 0.288; deviance = 837.0; d.f. = 1143 Residual deviance = 937.73; residual d.f. = 1146 White-flash full model Intercept (set/observer) 1.460 0.979 1.492 0.136 P. novaehollandiae in-hand ID experience –0.357 0.443 –0.806 0.570 Small mammal in-hand ID experience –0.146 0.257 –0.568 0.420 Small mammal camera ID experience 1.075 0.496 2.168 0.030* P. novaehollandiae ID Quiz completion –0.997 0.959 –1.040 0.298 Observer var. = 0.396; set var. = 2.420; deviance = 369.4; d.f. = 489 White-flash simple model Intercept (set/observer) 1.764 0.748 2.358 0.018* Observer var. = 0.949; set var. = 1.296; deviance = 374.8; d.f. = 493 Residual deviance = 460.64; residual d.f. = 495 Infrared flash full model Intercept (set/observer) 1.497 0.534 2.802 0.005** P. novaehollandiae in-hand ID experience 0.682 0.742 0.919 0.915 Small mammal in-hand ID experience –0.038 0.355 –0.106 0.358 Small mammal camera ID experience 0.276 0.441 0.627 0.531 P. novaehollandiae ID Quiz completion 0.550 0.981 0.560 0.575 Observer var. = 0.924; set var. = 0.174; deviance = 412.7; d.f. = 644 Infrared flash simple model Intercept (set/observer) 2.369 0.505 4.695 2.67e-06*** Observer var. = 1.648; set var. = 0.297; deviance = 418.9; d.f. = 648 Residual deviance = 469.22; residual d.f. = 650

79 Chapter 3: Identifying Pseudomys in camera trap images

Discussion

Our identification trials, across a range of observers with varying small mammal experience, suggest that P. fumeus and P. novaehollandiae are both reasonably identifiable in high-quality infrared and white-flash camera trap images. For both species, pelage colouration and morphology (body, head, neck, and tail) allow discernment from non-target species. As suggested by Nelson et al. (2009) and

Holmes (2012), high-quality white-flash camera trap images are ideal for P. fumeus surveys, with high accuracy even for inexperienced observers. Observers performed exceptionally well when identifying P. fumeus in white-flash images, and performed comparatively poorly when discerning P. fumeus in infrared images, likely due to P. fumeus having distinctive colouration from, but similar body shape to, Rattus spp.

(Nelson et al. 2009). Contrary to expectation, observers in the P. novaehollandiae trial performed modestly better when identifying infrared images than white-flash images, attributable to a higher rate of false positives in white-flash images where M. musculus was identified as P. novaehollandiae. This suggests that for P. novaehollandiae body shape provides a clearer diagnostic than colouration, which may be misleading in the white-flash images (Table 3.1). Towards the north of the species’ range in parts of New South Wales and Queensland, P. novaehollandiae co- occurs with some other rodent species (Pseudomys spp., Melomys spp.) not described here. Further identification trials would be necessary to test whether P. novaehollandiae is readily distinguished from these other species.

While experience had no significant bearing on identification accuracy for P. fumeus in white-flash images, it accounted for all variation among observer identification

80 Chapter 3: Identifying Pseudomys in camera trap images accuracy in infrared image identification. However, white-flash identification for P. fumeus was so straightforward and resulted in so few errors that there was very little variance for which to account. Infrared identification was significantly more challenging, and, as could logically be expected, increasing experience with identifying P. fumeus in-hand and with other small mammal species on camera improved observer skill. However, increased experience with in-hand identifications of other small mammal species correlated significantly with lower observer skill.

This may be due to an over-reliance on typical in-hand identification characteristics by observers with extensive field experience who lacked experience with P. fumeus or camera identification. Because in-hand identification factors such as colouration, ear shape/size, and scent are poorly represented or absent in camera trap images, identification relies on a different suite of characteristics, including body shape and how individuals of a species carry themselves. Descriptively, P. fumeus and R. fuscipes are very similar in infrared images, and were more frequently mistaken by those lacking experience at identifying these less obvious characteristics (i.e. those with less experience of camera image identification). Alternatively, those with more in-hand small mammal experience may have been overconfident in their abilities and risked identifications of more challenging images, which less experienced observers dismissed as ‘unsure’. This effect of overconfidence has been demonstrated in bird surveys, wherein observers with greater experience are more likely to record false- positives of rare species than those with less experience (Farmer et al. 2012). Given the comparative accessibility of white-flash image identification to all observers regardless of experience and confidence, we recommend the use of white-flash camera traps for P. fumeus surveys in preference to infrared cameras.

81 Chapter 3: Identifying Pseudomys in camera trap images

In the P. novaehollandiae trial, observer experience failed to explain most of the variation in identification accuracy among observers. In other systems, training and experience of observers can improve results (anuran call surveys: McClintock et al.

2010) or have no significant effect on observers’ performance (bird call surveys:

Miller et al. 2012). For some tasks, innate skills may limit the efficacy of training.

For example, studies of observer performance in facial recognition have identified that people are inherently good or inherently bad at identifying stimuli (Russell et al.

2009). Observer effects commonly have significant influences on species detection rates (Diefenbach et al. 2003; Lotz and Allen 2007; McClintock et al. 2010), and, in this instance, we showed that variation among observers can remain even after accounting for training and experience. Quantifying observer effects does, however, allow us to choose which observers to retain for maximal effectiveness in ongoing survey work.

Our study supports other research demonstrating the need to conduct specificity and sensitivity trials for small mammal species before dismissing or accepting camera trapping as a suitable survey technique. Other species may not be so readily identified (Meek and Vernes 2015; Villette et al. 2016) or may be better suited to a different camera trap set-up (e.g. orientation: Harley et al. 2014; Taylor et al. 2014).

For example, Meek and Vernes (2015) conducted similar work on P. oralis, a species similar in colouration and body shape to P. novaehollandiae, but closer in size to P. fumeus. They determined that P. oralis was not adequately distinguishable from R. fuscipes on white-flash or infrared images and, as such, camera surveys were not an appropriate surveying technique for the species. While their rejection of infrared cameras is concordant with the findings here for P. fumeus, distinctive pelage

82 Chapter 3: Identifying Pseudomys in camera trap images colouration in P. fumeus allows for use of white-flash cameras although these were deemed unsuitable for P. oralis. White-flash cameras commonly produce more readily identifiable images than infrared cameras when pelage colour is a useful identifying characteristic (Glen et al. 2013; Meek and Vernes 2015). However, when pelage colouration is unreliable or misleading, infrared cameras may perform better, as shown here by the P. novaehollandiae trial. When human observations produce a high level of error, further development of automated recognisers for distinctive pelage textures can improve identification accuracy (Falzon et al. 2014).

While not tested explicitly by this study, our success in identifying P. fumeus and P. novaehollandiae was dependent upon our optimised camera trap settings and placement. Tailoring camera trap type, function, and settings to maximise the number and quality of images produced is critical in surveys of cryptic, rare or elusive small mammals (Meek and Vernes 2015; Fancourt et al. 2017). The identification guide presented in Table 3.1 requires clear, high-resolution images to implement. The

Reconyx and PixController camera models and set-up that we used in this study provided high-quality images suited to small mammal identification. However,

PixController cameras can have poor battery lives and high failure rates, potentially inflating the number of site-level false negatives and making them unsuitable for long-term deployment (Meek and Vernes 2015). Many other camera trap models are designed for large game, and, as such, they can produce images washed out by the flash with resolution too poor to make out the necessary identifying characteristics of smaller species (e.g. Scoutguard: Holmes 2012; Cuddleback – Meek and Vernes

2015). For small mammals, fast trigger and shutter speeds, high-sensitivity triggers, multiple images per trigger, high-resolution images, and a flash muted enough for

83 Chapter 3: Identifying Pseudomys in camera trap images use at short focal distances are preferred camera settings (Rovero et al. 2013; Meek et al. 2015a). When surveying larger mammals, bait stations are commonly set several metres from the camera; this distance produces poor-quality images of small mammal species, reducing confidence in identification. Where the camera trap type allows, a focal distance of 1–1.5 m improves picture quality, with careful positioning to match the detection zone of the camera trap type (Glen et al. 2013). Attaching an external lens (e.g. Monitor Cameras, Meredith, Victoria) to the camera trap can further improve picture clarity by allowing focal length to be reduced below 1 m, as has been demonstrated for P. novaehollandiae (A. McCutcheon, pers. comm.). Use of a trigger delay can reduce the number of images to process and has been used to delineate a ‘hit window’ for abundance estimation (e.g. 5-min delay suggested for pine martens (Martes martes) by Manzo et al. (2012). However, for small mammals a trigger delay reduces the number of identifiable images available, and camera trapping is not a reliable method for estimating abundance of small mammal populations when animals are not individually recognisable (Oliveira-Santos et al.

2010; Foster and Harmsen 2012). Rather, increasing the number of images per trigger may help to attain identifiable images from multiple angles as the animal navigates the bait station (Meek et al. 2015a; Meek and Vernes 2015). Alternatively, setting camera traps to record videos may improve (Villette et al. 2016) or return similar (Glen et al. 2013) identification rates for small mammals. However, video recordings rapidly drain camera batteries and consume storage space on the SD card, limit deployment time, and take longer to screen (Glen et al. 2013; Meek and Vernes

2015).

84 Chapter 3: Identifying Pseudomys in camera trap images

Using the techniques outlined in this study, P. fumeus and P. novaehollandiae are discernable from species co-occurring within their southern ranges by means of camera trap images. To reduce false positives in surveys, white-flash camera traps are recommended for P. fumeus and infrared cameras for P. novaehollandiae. Similar studies are needed to assess the reliability of camera trapping for identification of many small mammal species not yet addressed here or elsewhere (Meek and Vernes

2015), with careful consideration of the most appropriate camera trap and bait station set-up to support identification. Caution must be taken to ensure that cryptic species are accurately identified, and that factors such as imperfect detection and effective sampling area are considered in the design and interpretation of camera trapping surveys (Meek et al. 2015a; Burton et al. 2015). As shown here, the advent of camera trapping as a reliable survey method for P. fumeus, P. novaehollandiae and other threatened small mammal species will facilitate the much-needed resolution of the species’ current distributions, informing management decisions to enhance conservation outcomes. Where questions requiring in-hand captures arise (e.g. abundance, demography, genetics), camera trap surveys can provide preliminary occupancy data to guide traditional trapping surveys.

85 Chapter 3: Identifying Pseudomys in camera trap images

Supplementary images

86 Chapter 3: Identifying Pseudomys in camera trap images

87 Chapter 3: Identifying Pseudomys in camera trap images

88 Chapter 3: Identifying Pseudomys in camera trap images

89 Chapter 3: Identifying Pseudomys in camera trap images

90 Chapter 4: Utility of SDMs for declining species

Chapter 4 Testing the utility of generic approaches to species distribution modelling for a species in decline

Burns, P. A., Clemann, N., and White, M. (in review). Testing the utility of generic approaches to species distribution modelling for a species in decline.

91 Chapter 4: Utility of SDMs for declining species

Abstract

Habitat suitability estimates derived from Species Distribution Models (SDMs) are increasingly used to guide management of threatened species. Poorly estimating species’ ranges can lead to underestimation of threatened status, undervaluing of remaining habitat, and misdirection of conservation funding. We aimed to evaluate the utility of a SDM, similar to the models used to inform government regulation of habitat in our study region, in estimating the contemporary distribution of a threatened and declining species. We developed a presence-only SDM for the endangered New Holland Mouse (Pseudomys novaehollandiae) across Victoria,

Australia. We conducted extensive camera trap surveys across model-predicted and expert-selected areas to generate an independent dataset for use in evaluating the model, determining confidence in absence data from non-detection sites with occupancy and detectability modelling. We assessed the predictive capacity of the model at thresholds based on 1) Sum of Sensitivity and Specificity (SSS), and 2) the

Lowest Presence Threshold (LPT; i.e. the lowest non-zero model-predicted habitat suitability value at which we detected the species). We detected P. novaehollandiae at 40 of 472 surveyed sites, with strong support for the species’ probable absence from non-detection sites. Based on our post-hoc optimised SSS threshold of the

SDM, 25% of our detection sites were falsely predicted as non-suitable habitat and

75% of sites predicted as suitable habitat did not contain the species at the time of our survey. One occupied site had a model predicted suitability value of zero, and at the LPT, 88% of sites predicted as suitable habitat did not contain the species at the time of our survey. Our findings demonstrate that application of generic SDMs in both regulatory and investment contexts should be tempered by considering their

92 Chapter 4: Utility of SDMs for declining species limitations and currency. Further, we recommend engaging species-experts in the extrapolation and application of SDM outputs.

Introduction

Species Distribution Models (SDMs) are increasingly used in threatened species conservation and management planning to prioritise the allocation of conservation investments (Guisan et al. 2013; Liu et al. 2013b), identify biodiversity ‘hot spots’

(Farashi et al. 2017), inform environmental impact assessments and environmental protection (DEPI 2013; OEH 2014), and select compensatory land when species’ habitat is compromised or destroyed (DELWP 2017a). SDMs take species occurrence data and correlate these with spatially explicit environmental covariates, such as climate and vegetation, to map environmental suitability for a given species (Smith et al. 2013). Although this plausibly suitable habitat may or may not be occupied, SDM measures of habitat suitability are often used to estimate a species’ current or future distribution (McPherson and Jetz 2007; Elith and Leathwick 2009; Sinclair et al.

2010).

Species may be precluded from ostensibly suitable habitat by numerous unidentified factors, including: geographic barriers, historical extirpations, issues such as predation, competition, disease, or unknown environmental factors (Sinclair et al.

2010; Jarnevich et al. 2015). This means SDMs that do not recognise these factors are likely to over-estimate the species’ distribution. These issues are compounded for threatened or rare species: threatened species are more likely to be affected by unidentified or unaccounted for threatening processes that are poorly rendered by the typical suite of environmental predictor data such as climate averages (Jetz et al.

93 Chapter 4: Utility of SDMs for declining species

2008); species with few occurrence records or small range sizes are particularly susceptible to overestimation of suitable habitat areas in standard presence-only methods because of the effects of background absence allocation (Somodi et al.

2017); and many threatened species’ ranges have declined before reliable distribution records were established (Bilney et al. 2010). Similarly, a species’ observed distribution may not reflect its full potential ecological space – correlative models based on observed records may underestimate a species’ potential range (e.g. as with invasive species; Phillips et al. 2008), or areas of occupancy may not have been surveyed (Jarnevich et al. 2015).

Over- and under-estimation of a threatened species’ distribution can lead to its further endangerment. When distributions are exaggerated, a species’ level of threat may be underrated and protections not implemented (Jetz et al. 2008; IUCN 2012; Ramesh et al. 2017), or funding and regulatory effort may be inefficiently deployed to places that do not support the species (Loiselle et al. 2003). Small patches of suitable habitat are generally of high conservation value to rare or threatened species (Wintle et al. 2018), however, overestimating a species’ distribution diminishes the perceived value of individual remnant habitat patches. Threatened species’ habitat destruction is often justified by the idea that the area to be destroyed represents a small portion of the species’ total area of occupancy (e.g. Looby and Gilmore 2016), and over- generous model interpretations may facilitate these erroneous assumptions. Similarly, when occupied habitat is misidentified as unsuitable, agencies regulating land-use and development may sanction the loss of important remnant habitat (Clemann 2015;

Nelson et al. 2017), or fail to implement conservation management (Kéry 2002; Rout et al. 2010).

94 Chapter 4: Utility of SDMs for declining species

Standard model testing methods rely on sub-setting the data used to make the model

(Bahn and McGill 2013). As such, tests of model accuracy are prone to error in the same way the models themselves are. While these measures of model accuracy are usually appropriate in a statistical framework, they do not necessarily provide an adequate test for the validity of using model estimates of habitat suitability to estimate species’ current distributions. Here, we instead use ground-truthing to test the capacity of SDM predictions derived from generic (or typical) methods to estimate contemporary occupancy of a threatened and declining rodent species, the

New Holland Mouse (Pseudomys novaehollandiae). We conduct broad-scale on- ground surveys across plausible habitat and contrast measures of model sensitivity and specificity derived from ground-truthing the model with standard model-based measures of accuracy. In the context of our findings, we provide advice on improved approaches.

Methods

Species Distribution Modelling

A SDM for P. novaehollandiae was created for Victoria, Australia (see Fig. 4.1) in

2014. This model was developed using a generic approach which was used for all terrestrial vertebrate fauna indigenous to the State of Victoria. The following sections briefly describe the data and methods used.

95 Chapter 4: Utility of SDMs for declining species

Environmental data

In all, 20 environmental predictors were used as explanatory variables (see Table 4.1) and these were mapped at 75 m x 75 m resolution. Resampling of data that resolved below 75 m was undertaken bilinearly.

Table 4.1. List of explanatory variables used in developing the Species

Distribution Model for Pseudomys novaehollandiae in Victoria, Australia. The

last six variables were derived from Landsat data.

Environmental Variable AnisHeatRug (Diurnal Anisotrophic Heating × Ruggedness) EvaporationJan (Evaporation in January) EvaporationJul (Evaporation in July) InsolationTotal (Total insolation) MaxTempJan (Maximum temperature in January) MinTempJul (Minimum temperature in July) RainfallJan (Rainfall in January) RainfallJul (Rainfall in July) VisibleSky (Visible sky Index) Wetness (Terrain Wetness Index) log_saline (logarithm of vertical height above saline water) log_major (logarithm of vertical height above major river) log_minor (logarithm of vertical height above any watercourses) ND_KTh (Radiometics, normalized difference between standardized potassium and thorium) Ratio_KinvTh (Radiometics, potassium/thorium ratio with inverse operation) Ratio_KTh (Radiometics, potasium/thorium ratio) Red (Median Landsat band: 1989 2000) NIR (Median Landsat near infrared band: 1989 2000) EVI (Median Landsat enhanced vegetation index: 1989 2000) NDVI (Median Landsat Normalised Difference Vegetation Index: 1989 2000)

96 Chapter 4: Utility of SDMs for declining species

Species data

Observation records of P. novaehollandiae made between 1974 and July 2014 from across the study area were collated from State and National databases (Atlas of

Living Australia 2014; DELWP 2014). We removed all records deemed unsuitable for modelling. These data included:

• All records obtained through the identification of hairs found in predator scats.

• Records that were flagged with a potential spatial error exceeding 200 m.

• Records located in modified habitats considered grossly unsuitable for the

species such as urban areas and crops.

• Records that failed a semi-automated expert vetting process, whereby outliers

were identified (see Liu et al. 2018) and sent to a taxon expert for verification.

As of 2014 a total of 264 presence records passed these tests and were incorporated into the model.

‘Background’ absence or pseudo-absence data were selected from a pool of 1 million points randomised within land-cover delimited strata, but covering our entire spatial domain: equal numbers of random background absences were apportioned to 10 land-cover types to minimise bias promoted by non-random patterns in the geographic space.

Modelling

The Random Forest (RF) modelling methods employed were similar to those documented by (Liu et al. 2013b), differing primarily in the updated dataset used.

Random forest modelling uses an ensemble of decision trees based on different subsets of the data to essentially ‘vote’ on the best model (Breiman 2001). We

97 Chapter 4: Utility of SDMs for declining species implemented a routine that supplied 50 bootstrap sub-samples to the RF where each model had tree depths of 25 leaves (i.e. decisions). Each bootstrap sub-sample comprised 16 presence records and 32 background absence records or twice the number of presences (see Barbet-Massin et al. 2012; Liu et al. 2016). Each of the presence records were randomly selected from 16 data ‘bins’ determined by clustering the presence records in the geographic space using the ‘kmeans’ function in R (R Core Team 2017). A hold-out or test dataset, comprising 30% of the presence and absence data was initially used to test model performance. Following testing, all of the presence and absence data were then used to select boot-strap samples to make the final spatially-explicit ensemble model.

Ground-truthing the model

Targeted surveys

Between August 2015 and April 2018, we conducted camera trapping surveys in 472 sites across Gippsland and Anglesea, Victoria. Sites were commonly > 500 m apart, with a minimum distance of 100 m apart for areas where habitat structure or composition changed rapidly. We selected sites in areas with 1) historical P. novaehollandiae detections, 2) a high probability of occupancy as predicted by the

SDM, or 3) where on-ground inspection suggested likely occurrence due to habitat suitability despite lower SDM predictions. We included a mixture of locations where the species was known previously to occur, and locations where the species had never previously been recorded but were environmentally and geographically logical based on prior understanding of the species (focal region highlighted in Fig. 1).

98 Chapter 4: Utility of SDMs for declining species

At each site, we attached one Reconyx Hyperfire HC550 white-flash, HC500 infrared, or PC850 white-flash camera to a tree at 0.10 – 0.30 m from the ground.

Each camera was pointed at the base of an elevated bait station set approximately

0.90 m from the camera, with the bait station at a height of approximately 0.10 m.

Elevated bait stations consisted of a recycled plastic garden stake and alternately one of either a cutlery drainer containing four tea-strainers, or a vented PVC tube. Each station was baited with a mixture of oats, peanut butter, golden syrup, and vanilla extract. Cameras were set to capture 3 images per trigger with 1 second delay between images, no delay between triggers, and ‘fast shutter’ night mode. Cameras were left in position for a minimum of 14 nights. Resultant images were identified as per guidelines in Chapter 3 - Burns et al. (2018).

Occupancy - detectability analyses

To infer the absence of P. novaehollandiae from a site with any degree of confidence, we must assess the species’ detectability – the likelihood of recording P. novaehollandiae if it is present at a site – during our surveys. We estimated site-level occupancy (ѱ) and detectability (p) simultaneously, using repeat count data (i.e. nightly detections or non-detections; (MacKenzie et al. 2002). Detectability of P. novaehollandiae varies seasonally, over the course of a survey, with moon phase, and with rainfall during a survey (Chapter 2 - Burns et al. in review); we were unable to account for seasonal shifts in detectability due to our sampling pattern, but incorporated the other three factors in our detectability analysis. Our most complex detectability model comprised an intercept, P, a binary fixed effect of moon brightness (M: 1 = full moon, 0 = other), and continuous effects of night of survey

99 Chapter 4: Utility of SDMs for declining species

(T: 1-n) and nightly rainfall (R: 24-hour rainfall prior to and including each survey night 6am – 6am; Bureau of Meteorology, 2018).

� 1 2 3 Our occupancy model logit incorporated(�) = � + an � intercept,� + � � +O,� and� the effect of model predicted habitat suitability (S; SDM-derived value between 0 and 1) as used in site selection.

To aid numeric model fitting, welogit scaled(��) all= �continuous� + �4� variables to have a mean of zero and unit variance. We used the ‘occu’ function in the package ‘unmarked’ (Fiske et al. 2015) in the program ‘R’ v 3.4.3 (R Core Team 2017). We ranked models based on AIC values, eliminated those with ΔAIC > 4 (Burnham and Anderson 2002) then eliminated models with uninformative additional complexity (Arnold 2010). This left only one candidate model remaining.

To determine whether the absence data reported here is statistically supported, we calculated probability of false absence (pfa) estimates for each site, based on survey- specific nightly detection probabilities. Because detectability (p) varies across survey night, T, pfa is given by: (1-p,1) (1-p,2)…(1-p,T). Where p,T is the species-level detection on night T. To achieve 95% confidence in the species’ absence, pfa must fall below 0.05.

Ground-truthed model evaluation

We assessed the capacity of the of the model (based on false positives and false negatives) to predict habitat occupancy for P. novaehollandiae. To elicit a binary

‘opinion’ from the SDM we converted the continuous habitat suitability estimates to a

100 Chapter 4: Utility of SDMs for declining species binary, suitable vs unsuitable output at two suitability estimate values. The first used the Sum of Sensitivity and Specificity (SSS) method, optimising the threshold for maximum SSS and minimum difference between Sensitivity and Specificity (Liu et al. 2005; Liu et al. 2013a). The second threshold was set at the minimum non-zero

SDM value at which the species was detected – i.e. maximum sensitivity without setting specificity at zero.

Results

The SDM and ‘within’ SDM testing

Modelled suitability estimates ranged from 0 to 86 (Fig. 4.1). Employing a hold-out data set we evaluated the P. novaehollandiae SDM across the entire south-eastern

Australian study area. Performance measures are given in Table 4.2. The performance measures are uniformly high; however, this is typical of species that have very low prevalence across a given study area (McPherson et al. 2004).

101 Chapter 4: Utility of SDMs for declining species

Figure 4.1. Habitat suitability estimates derived from a correlative Species

Distribution Model for Pseudomys novaehollandiae across Victoria, Australia.

Box indicates plausible region of occupancy.

Table 4.2. Performance measures for the Pseudomys novaehollandiae Species

Distribution Model across Victoria, Australia.

Area Under the True Skill Overall Sensitivity Specificity Kappa RO Curve Statistic accuracy 0.977 0.982 0.997 0.959 0.832 0.982

102 Chapter 4: Utility of SDMs for declining species

Ground-truthing the model

Targeted surveys

We recorded 315,792 images over 8575 camera trap nights, and detected P. novaehollandiae at 40 of 472 sites across Victoria (Fig. 4.2).

Figure 4.2. Camera trap survey sites across Victoria, Australia. Circles indicate

Pseudomys novaehollandiae detection sites (yellow) and non-detection sites

(black).

Occupancy - detectability analyses

The top ranked model incorporated detection covariates of moon brightness and night of survey, and the occupancy covariate of SDM-derived habitat suitability value (Table 4.3; Table 4.4). Detectability was lower in the five days around a full moon and improved over the course of a survey (Table 4.4). Occupancy improved with model-predicted habitat suitability (Table 4.4). Probability of false absence (pfa) at each site fell below the a priori threshold of 0.05 after a maximum of eight nights, sites conferring very high support for the species’ absence from sites at which it was not detected. Hereafter non-detections are treated as absences.

103 Chapter 4: Utility of SDMs for declining species

Table 4.3. Model selection results for unmarked models of Pseudomys

novaehollandiae detectability (p) and occupancy (ѱ) across Victoria, Australia.

Parameters: intercept (.), moon brightness (moon), night of survey (night),

rainfall during survey (rain), SDM predicted habitat suitability (SDM).

Model logLikelihood AICc ΔAICc p(moon+night) ѱ(SDM) -584.93 1179.98 0 p(moon+night+rain) ѱ(SDM) -584.58 1181.35 1.36 p(moon) ѱ(SDM) -589.21 1186.50 6.51 p(.) ѱ(SDM) -598.33 1202.70 22.72 p(moon+night) ѱ(.) -615.14 1238.37 58.39 p(moon+night+rain) ѱ(.) -614.80 1239.73 59.75 p(moon) ѱ(.) -619.43 1244.90 64.92 p(moon+rain) ѱ(.) -619.01 1246.11 66.13 p(.) ѱ(.) -628.55 1261.13 81.15

Table 4.4. Parameter estimates from top ranked occupancy-detectability model

for Pseudomys novaehollandiae across Victoria, Australia. Parameter estimates

are reported on the logit scale.

95% C. I. Detection covariate Estimate SE Lower Upper -0.543 0.146 -0.829 -0.257 Moon brightness -0.944 0.269 -1.471 -0.417 �� Night of survey 0.032 0.011 0.010 0.054 Occupancy covariate Estimate SE Lower Upper -3.170 0.281 -3.721 -2.619 SDM predicted habitat suitability 1.550 0.235 1.089 2.011 ��

Ground-truthed model evaluation

The Sum of Sensitivity and Specificity (SSS) of the SDM to the occurrence of P. novaehollandiae was maximised at a habitat suitability estimate of 56, with a sensitivity of 0.75 and specificity of 0.79 (Table 4.5; Fig. 4.3; Fig. 4.4a). At this

104 Chapter 4: Utility of SDMs for declining species value, a quarter of occupied sites were falsely predicted as unoccupied/unsuitable habitat, and three quarters of predicted habitable sites were unoccupied (Table 4.5).

The SSS thresholded model failed to recognise the suitable habitat of an entire isolated area of occupancy (Providence Ponds Flora and Fauna Reserve; see (Chapter

6 - Burns in review), as well as several other locations closer to recognised habitat.

The minimum non-zero SDM habitat suitability estimate at which the species was detected was 39 (Fig. 4.4b); at this value, sensitivity was predictably high (0.98; n=1 site at model prediction of 0) and specificity exceptionally low (0.33; Table 4.5). This threshold incorporated large areas of blatantly inappropriate habitat, as identified through expert ground-truthing. Both inferences of the species’ distribution differed from expert estimates drawn from recent detection records and survey results (see

Chapter 6 - Burns, in prep). ; Fig. 4.4c).

Table 4.5. Sensitivity and specificity of a Species Distribution Model to the

occurrence of Pseudomys novaehollandiae across Victoria, Australia, at two

thresholds: Sum of Sensitivity and Specificity (SSS; 56), and minimum non-

zero habitat suitability estimate at which the species was detected (39).

Threshold Sensitivity Specificity False positives (%) False negatives (%) SSS 0.75 0.79 75.4 25.0 Minimum non-zero 0.98 0.33 88.3 2.5 detection

105 Chapter 4: Utility of SDMs for declining species

1.58

1.48

1.38

1.28 SSS

1.18

1.08

0.98 0 20 40 60 80 SDM habitat suitability value

Figure 4.3. Sum of Sensitivity and Specificity (SSS) of Species Distribution

Model (SDM) derived habitat suitability estimates for Pseudomys

novaehollandiae in Victoria, Australia.

106 Chapter 4: Utility of SDMs for declining species b)

Figure 4.4. Species Distribution Model recognised suitable habitat for

Pseudomys novaehollandiae thresholded at a) optimised value based on

maximum Sum of Sensitivity and Specificity (SSS; 56), and b) the minimum

non-zero value at which Pseudomys novaehollandiae were detected (39); and

c) estimated area of occupancy derived from minimum convex hulls by region

based on detection records 2013-18.

107 Chapter 4: Utility of SDMs for declining species

Discussion

SDM accuracy is typically tested using data novel to the model, derived either through sub-setting the dataset (Bahn and McGill 2013) or by using completely different datasets assembled for the same species (Randin et al. 2006). Here, we used a variant of the entirely new dataset: we collected a comprehensive contemporary survey dataset to ground-truth a model derived from a biased, dated, and largely incidental dataset (i.e. incomplete and spatially biased sampling conducted between

1970-2014, including records of extinct populations). We recorded numerous new detection sites for P. novaehollandiae that had either not been surveyed before, or not previously surveyed with sufficient effort and/or rigour to detect the species (see

Chapter 6 - Burns in review; Chapter 2 - Burns et al. in review). The vast majority of our survey sites yielded no P. novaehollandiae detections. Our approach demonstrates that estimating species’ occurrence from SDMs has the capacity to exaggerate species’ current distributions, particularly for species in decline, while simultaneously neglecting to recognise some occupied locations. Although our within-model measures of model fit indicated strong support for model accuracy, on- ground, our SSS thresholded SDM failed to predict the suitability of habitat for P. novaehollandiae at 25 percent of detection sites – including the entirety of one of the species’ five remaining management areas for the state (Chapter 6 - Burns in review)

– while 75 percent of sites with modelled suitable habitat did not contain the species.

Some of these false-positives occurred in locations that the species was known to occupy historically or may have previously been occupied, but that no longer support the species. These extirpations are likely due to threatening processes such as invasive predators, habitat fragmentation, and the reduction in habitat suitability over

108 Chapter 4: Utility of SDMs for declining species time (Chapter 6 - Burns in review). Using standard correlative SDMs to estimate species’ realised distributions can be increasingly misleading as species decline in abundance and distribution through time.

In selecting a model threshold that would include most of the areas occupied by our study species across the landscape (i.e. sensitivity close to 1), the SDM delivered a gross exaggeration of P. novaehollandiae occurrence. First and foremost, this suggests we need an improved modelling technique and to be more discerning with our selection of explanatory variables. Techniques such as dynamic occupancy modelling (Santika et al. 2014), accounting for imperfect detection (Lahoz-Monfort et al. 2014; Guillera-Arroita 2016), or mechanistic approaches (Mathewson et al.

2017) may provide better estimates of this and other threatened species’ distributions.

However, the technique and variables we used were selected to reflect the current methodology uniformly applied in threatened fauna habitat management in the study area (intended to influence management planning and investment to 2037; DELWP

2017b), and standard correlative methods commonly applied elsewhere (Jarnevich et al. 2015). Therefore, we discuss the conservation implications of model interpretation and how to improve outcomes with the existing modelling method.

As both over- and under-estimation of a threatened species’ distribution can have poor conservation outcomes, depending on the situation (Kéry 2002; Loiselle et al.

2003; Rout et al. 2010; Clemann 2015), our results encourage the use of application- specific thresholding approaches that engage the Precautionary Principle (Cooney and Dickson 2012). Where a species’ plausible habitat is at risk, it is safer to apply a low threshold and assume presence, unless there is adequate evidence to assert

109 Chapter 4: Utility of SDMs for declining species absence (i.e. MacKenzie et al. 2002; Guillera-Arroita et al. 2014). Even so, a species’ current absence from a site does not always indicate future absence – areas of model- predicted suitable habitat may facilitate reintroductions (Peterson 2006) or natural migrations following mitigation of threatening processes (Scheele et al. 2014).

Potential future occupancy is however, not a reasonable trade-off for current occupancy: unaided recolonization is rare, and translocations often fail (Viggers et al.

1993; Fischer and Lindenmayer 2000; Seigel and Dodd 2002). As such, when selecting areas to offset or mitigate the destruction of threatened species’ habitat, only areas in which the species’ presence is confirmed and at remediable risk should be considered in line with the concept of ‘no net loss’ (Curran et al. 2014; Spash

2015). If the question is one of prioritising areas for the distribution of limited management efforts, assuming the species presence may be cost-ineffective and lead to misapplication of resources better delivered elsewhere (Loiselle et al. 2003;

Chadès et al. 2008). In such cases, areas that are confirmed to contain the target species should be prioritised for investment. When assessing a species’ status, areas should not be assumed occupied based on model predictions alone (Jetz et al. 2008;

Ramesh et al. 2017). Presented with a broad estimate of the species’ distribution

(Fig. 4.1), we found that expert judgement based on historical detections, ecological understanding, and on-ground habitat assessment was critical in narrowing down potential sites and areas for survey, and also in identifying potentially occupied sites in areas with low model-estimated habitat suitability.

SDMs can provide useful frameworks to guide site selection for survey efforts, inform prioritisation of areas for protection (Wilson et al. 2011) or future occupation

(Peterson 2006; Scheele et al. 2014), and improve understanding of species’ habitat

110 Chapter 4: Utility of SDMs for declining species associations (Wilson et al. 2011). By necessity, however, model predictions represent a gross oversimplification of the factors governing a species’ occurrence across the landscape (Jetz et al. 2008; Sinclair et al. 2010). As such, the use of SDMs to estimate species’ present-time distributions is complex and inherently prone to error, as demonstrated here, with potentially dire consequences for threatened species

(Kéry 2002; Loiselle et al. 2003; Rout et al. 2010; Clemann 2015). Some of this error can be mitigated by engaging the Precautionary Principle to adjust interpretation of model outputs to reflect the threatened species’ best interests in a given scenario. We recommend use of generous thresholds when considering areas to protect or survey, and more stringent thresholds when contemplating a species’ status. In all instances, engaging a species-expert to inform the development and interpretation of SDMs will absorb some of the implicit uncertainty, and facilitate retention or elaboration of uncertainty where it remains necessary.

111 Chapter 5: TSF and NHMs

Chapter 5 Time since fire is an over-simplified measure of habitat suitability for the New Holland Mouse (Pseudomys novaehollandiae)

Burns, P. A. and Phillips, B. L. (in review). Time since fire is an over-simplified measure of habitat suitability for the New Holland Mouse (Pseudomys novaehollandiae).

112 Chapter 5: TSF and NHMs

Abstract

Fire has shaped much of the Australian landscape, and alterations to natural or historical fire regimes are implicated in the decline of many native mammal species.

Time Since Fire (TSF) is a common metric used to understand vegetation and faunal responses to fire, but may not capture the complexity of successional changes following fire. The New Holland Mouse (Pseudomys novaehollandiae), a threatened and declining rodent species native to south eastern Australia, is traditionally considered an early post-fire successional species. Here we use a 48-year dataset to test whether this posited association with early TSF is upheld, and whether the species occurrence and abundance are governed by TSF. We find no support for TSF influencing the species’ occurrence, and that abundance of P. novaehollandiae is also poorly explained by TSF. We suggest that it is not helpful to consider the species as early successional and that fire planning for P. novaehollandiae conservation is best considered at a local scale. Additionally, we provide guidelines for maximizing individual survival and persistence during and after planned burns.

Introduction

Fire is a common form of disturbance in the Australian landscape and is often invoked as a major driver of community composition (Fox and McKay 1981; Fox

1982; Fox 1990; Catling et al. 2002; Kelly and Brotons 2017). Altered fire regimes are considered one of several key factors contributing to catastrophic declines and extinctions in Australian mammal species since European invasion (Lindenmayer

2015; Woinarski et al. 2015). Understanding how species and ecological communities respond to fire is integral to developing appropriate conservation

113 Chapter 5: TSF and NHMs strategies for fire and land management agency responses during and after bushfires, as well as intelligent design of planned burns (Driscoll et al. 2010). Mammal species’ responses to fire are broadly underpinned by changes in vegetation structure; itself having a complex relationship with fire. Therefore assessing a species’ response to fire simply as a function of Time Since Fire (TSF) may prove less informative than more direct measures of habitat suitability (Monamy and Fox 2000; Di Stefano et al.

2011; Nimmo et al. 2014; Swan et al. 2015).

The New Holland Mouse (Pseudomys novaehollandiae; Rodentia: Muridae) is a threatened and rapidly declining native Australian rodent species (Chapter 6 - Burns in review; Lazenby et al. 2018) and traditionally considered to be associated with early successional vegetation classes (Keith and Calaby 1968; Posamentier and

Recher 1974; Cockburn 1980). Likely due to their use of underground burrow systems for shelter, P. novaehollandiae can persist in situ following fire, with no or low immediate mortality (Kemper 1990; Lazenby et al. 2008). A reduction in resource availability (i.e. food, shelter) and increased predation risk in the months following a fire may, however, cause delayed fire-related mortality (e.g. 26% mortality in the two months following a low intensity 1975 fire; Kemper 1990) and individuals may move out of their pre-fire home ranges to occupy unburned habitat

(Kemper 1977). In extreme cases, isolated populations may be extirpated by severe fire events (e.g. Ash Wednesday fires at Anglesea, south-eastern Australia; Lock and

Wilson 1999). Abundance is reported to peak anywhere between 0 – 8 years post-fire

(Fox and McKay 1981; Fox 1982; Fox 1990; Lock and Wilson 1999; Wilson et al.

2017), and literature considers much older fire classes to be inappropriate habitat or anomalous records (Fox and McKay 1981; Wilson et al. 2017). As a result, TSF is a

114 Chapter 5: TSF and NHMs major consideration in conservation and management planning for the species (Parks

Victoria 1998; Seebeck et al. 2003; Anon. 2010; Smith and Bluff 2018).

Breaking with this ‘early-successional’ depiction, however, P. novaehollandiae is also regularly found in mature undisturbed habitats, sometimes at great abundance

(Vance 2001; Wilson and Laidlaw 2003; Lazenby et al. 2008; McCall et al. 2015).

The diversity of post-fire responses reported for P. novaehollandiae hint at a complex relationship with fire that may be location-specific and governed more by factors such as fire intensity and coverage, local vegetation types, habitat structure, and rainfall patterns rather than by TSF itself – a complexity that is now well recognized in other species (Di Stefano et al. 2011; Nimmo et al. 2014; Swan et al. 2015;

McCall et al. 2015; Chia et al. 2016). Prior studies defining TSF associations in P. novaehollandiae populations have covered limited post-fire age classes, been geographically localized, or focused on a single fire event (e.g. Fox 1982; Fox 1990;

Hollis 1999; Tasker and Dickman 2004; Lock and Wilson 2017), limiting application of findings to each specific time and place. In order to understand recent declines and manage P. novaehollandiae in the long-term, it is important to identify the potential role of fire in mediating the species’ disjunct distribution and erratic population fluctuations. As such, there is a clear need for a larger-scale assessment of the relationship between P. novaehollandiae occupancy, abundance, and TSF.

Here, we use a 48-year dataset to test whether P. novaehollandiae occurrence and abundance are governed by TSF in eastern Victoria, Australia. Detectability – the probability of recording a species at a site during a survey if it is present – must be taken into account to reliably test covariates of occupancy or abundance (Kéry et al.

115 Chapter 5: TSF and NHMs

2009). Detectability of P. novaehollandiae varies with seasonal behaviours, moon brightness, rainfall, and over the duration of a survey, while abundance also varies seasonally (Chapter 2 - Burns et al. in review). Differences in survey conditions across the dataset could confound our ability to estimate the species abundance or even occurrence and thus our ability to define relationships with other covariates (i.e.

TSF), therefore we account for the species’ nightly detection probabilities. Further, we use our findings and the broader literature to provide interim guidelines on planned burning precautions for P. novaehollandiae.

Methods

Data

We used the live trapping dataset detailed in Chapter 2 - Burns et al. (in review), spanning 48 years from the species’ initial detection in Victoria in 1970 to April

2018, sourced from historical field notes, recent reports, and our own survey work.

Elliott small mammal traps (Elliott Scientific Co., Upwey, Victoria; size 90 mm ×

100 mm× 330 mm) were used across all surveys, baited with a mixture of peanut butter, golden syrup or honey, and rolled oats, with non-absorbent nesting material for insulation. For each survey, we recorded nightly trap results as counts of individuals, nightly trap effort, survey dates, and precise locality. Average survey effort was 31 (range: 8-100) Elliott traps set for 3.5 (range: 1-6) nights. We included data from 328 surveys of 130 sites (>200 m apart) in south east Victoria (Fig. 5.1), collected over 349 unique nights, totalling 35,774 trap nights, spanning 26 years.

For our camera trapping dataset, we used a subset of the camera survey work conducted in Chapter 6 - Burns (in review), in which 472 sites in plausible P.

116 Chapter 5: TSF and NHMs novaehollandiae habitat across Victoria were surveyed between August 2015 and

April 2018. Cameras were baited with peanut butter, golden syrup, oats, and vanilla essence following guidelines in Chapter 3 - Burns et al. (2018). As the camera trapping dataset explored large regions with no historic P. novaehollandiae records, we excluded sites in regions where the species had never been detected or was no longer extant to reduce the impact of extraneous factors (e.g. geographic barriers, habitat loss and fragmentation) confounding the species’ occurrence, leaving 135 sites for inclusion (Fig. 5.1).

Figure 5.1. Sites included in Pseudomys novaehollandiae Elliott trapping and

camera trapping datasets, Victoria, Australia. Note: this is not representative of

the species’ entire trapping history in the state; survey events were only used if

nightly trap effort and results were available and the species was extant in the

region at time of survey.

117 Chapter 5: TSF and NHMs

We extracted time since the last fire for each site from the Victorian state government fire history GIS layer (DELWP 2018); where a site had no fire recorded since records began (~1950), we assigned a TSF value as if it had last burned in 1950. Fire history data are imperfect, with a temporal bias toward false negatives in older time periods, and false positives in recent times – due respectively to poor historical reporting of fires and recent overrepresentation of burn coverage within fire scars (Bluff 2014).

This means some of our older TSF age class data could be more recently burned than is recorded, and that some of our more recent data for younger age classes may not have actually burned during the last recorded fire for a site. We reduced the likelihood of these errors confounding our results by using additional sources of fire history data to correct fire histories where available, drawing on contemporary survey reports (Hollis 1996; McCall et al. 2015), land management agency knowledge of local fire history (Jim Whelan, Parks Victoria, Wilsons Promontory

National Park, pers. comms.), or our own ground-truthing using methods from (Bluff

2014) where applicable.

Occupancy-detectability modelling

We used nightly detection records (repeated count data) from the combined Elliott trapping and camera trapping datasets to jointly estimate site occupancy and species- level detectability (MacKenzie et al. 2002). Eight sites were sampled in both the

Elliott trapping and camera trapping datasets (in separate years), so 257 sites representing 130 Elliott trap sites and 135 camera trap sites were included in the occupancy analysis. True occupancy was specified as a latent variable, drawn from a

Bernoulli distribution. The expected value of the Bernoulli was specified with a linear model (and a logit link). The occupancy model incorporated an intercept, μO,

118 Chapter 5: TSF and NHMs

TSF (F: in years), and random effects of site ( ; 257 levels) and region ( ; 10

� � levels). Starting with a quadratic (the minimum model� required to estimate a peak� in occupancy), we ran models with increasing complexity of TSF effects until the highest order coefficient for TSF was no longer significant (i.e. the 95% credible intervals overlapped zero: a model with a third-order polynomial for TSF had a non- significant cubic term, thus we reverted to a second order polynomial to explain

TSF).

2 � � 1 2 � � logit(� ) = � + � � + � � + � + �

Detectability was also specified as a linear model (and a logit link). The linear model incorporated an intercept, μP, a first-order Fourier function of day of the year (D: in radians) to specify annual cycles in detection; night of survey (T: 1-n nights); and daily rainfall (Rd: 24-hour rainfall prior to and including each survey night 6am –

6am; Bureau of Meteorology, 2018) as continuous effects, and fixed effects of moon phase (Mk: full moon or other; full moon defined as the 5-day period around a full moon) and method (Cl: Elliott trap or camera trap). Moon phase was incorporated as a fixed effect because the model would not converge with full moon as a continuous linear response.

logit(�) = �� + �3Cos(�) + �4Sin(�) + �5� + �6�� + �� + ��

To aid numeric model fitting, we scaled all continuous variables (except day of the year) to have a mean of zero and unit variance. We fit the models in the Bayesian

119 Chapter 5: TSF and NHMs inference program ‘JAGS’ v 4.3.0. (Plummer 2017) via the package ‘rjags’ (Plummer

2016) in the program ‘R’ v 3.4.3 (R Core Team 2017). The detection or non-detection of the species each night was drawn from a Bernoulli distribution parameterized with the species-level detection probability, and true occupancy at the site. We burned in three chains of the models with 120,000 iterations and sampled every fifth iteration of a further 100,000 runs. Convergence was assessed using visual assessment of trace plots and the Gelman-Rubin convergence diagnostic (Gelman and Rubin 1992).

Abundance-detectability modelling

To assess the relationship between P. novaehollandiae abundance and TSF, we analysed 1) the complete Elliott trapping dataset with TSF 1-68 years, and 2) a subset of the data with TSF 1-15 years. We included the latter as most of the prior emphasis on the relationship between P. novaehollandiae abundance and TSF has focused on this age bracket, and the broader analysis may be unable to pick up the nuance in this time period. We used the N-mixture model approach of Royle and Nichols (2003) to model abundance and individual detectability simultaneously. True abundance was specified as a latent variable, drawn from a Poisson distribution. The expected value of the Poisson was specified with a linear model (and a log link). Detectability was specified as a linear model (and a logit link).

For the full TSF 1-68 years dataset, the linear model of abundance incorporated an intercept, μN, TSF (F: in years); long-term rainfall (Ra: cumulative for 18 months prior to survey; Bureau of Meteorology 2018) and trap effort (E: number of Elliott traps) as continuous effects; a first-order Fourier function of day of the year (D: in radians) to specify annual cycles in detection; and random effects of year ( ; 25

�� 120 Chapter 5: TSF and NHMs levels), site ( ; 130 levels) and region ( ; 8 levels). Models alternately incorporating

� � cumulative long� -term rainfall for the 6� and 12 months prior to a survey were also tested, returning weakly negative associations with P. novaehollandiae abundance, thus we retained the 18-month interval supported for other small mammal species elsewhere (Hale et al. 2016). Again, we ran models with increasing complexity of

TSF until the complexity was no longer significant (i.e. a model with a fourth-order polynomial for TSF had a non-significant quartic term, thus we reverted to a third order polynomial to explain TSF).

2 3 log(��) = �� + �7Cos(�) + �8Sin(�) + �9� + �10� + �11� + �12�� + �13�

� � � + � + � + �

The linear model of detectability incorporated an intercept, μP; a first-order Fourier function of day of the year (D: in radians) to specify annual cycles in detection; and continuous effects of moon brightness (0-1, where 0 = new moon, 1 = full moon), night of survey (T: 1-n nights), and daily rainfall (Rd: 24-hour rainfall prior to and including each survey night 6am – 6am; Bureau of Meteorology 2018)).

logit(��) = �� + �14Cos(�) + �15Sin(�) + �16� + �17� + �18��

The TSF 1-15 years data subset lacked the resolution to clearly elucidate nightly detection effects (moon brightness, rainfall, and night of survey), seasonal effects on abundance and detectability, and the effect of trap effort and long-term rainfall on abundance estimates. As the TSF 1-15 years data is a subset of the full TSF 1-68

121 Chapter 5: TSF and NHMs years dataset, we used output from the full dataset model in place of estimating these parameters again. The abundance model was otherwise as above, including newly estimated intercepts, random effects of year ( ; 20 levels), site ( ; 77 levels) and region ( ; 3 levels), but in this case with TSF (�F�) as only a second-order�� polynomial.

� �

2 log(��) = �� + �7 Cos(�) + �8Sin(�) + �19� + �20� + �12�� + �13� + �� + ��

� + �

To aid numeric model fitting, we scaled all continuous variables (except day of the year) to have a mean of zero and unit variance. We ran the models in the Bayesian inference program ‘JAGS’ as above. The number of individuals captured each night was drawn from a binomial distribution parameterized with the individual-level detection probability, and true number of individuals at the site. For the complete 1-

68 dataset, we burned in three chains of the models with 500,000 iterations and sampled every fifth iteration of a further 100,000 runs. For the 1-15 data subset, we burned in three chains of the models with 700,000 iterations and sampled every fifth iteration of a further 100,000 runs. Convergence was determined as above.

Results

Occupancy-detectability modelling

Occupancy showed only a very weak (possibly absent) relationship with TSF (Fig.

5.2). Detectability showed a strong seasonal fluctuation (Fig 5.3a), and a negligible improvement over the course of a survey and with rainfall during a survey (Fig. 5.3b, c). Moon phase explained more variance than any other detection covariate, with detectability substantially reduced around full moons (Table 5.2). The dataset for the

122 Chapter 5: TSF and NHMs occupancy-detectability modelling incorporated both Elliott trapping data (as in the abundance-detectability modelling), and camera trapping data. Overall, Elliott trap surveys had higher nightly detection probabilities than camera trap surveys, reflective of greater survey area (mean = 31 Elliott traps : 1 camera trap; Table 5.2).

Figure 5.2. Occupancy estimate for Pseudomys novaehollandiae with

increasing Time Since Fire (TSF), derived from an occupancy-detectability

model of a 48-year dataset in Victoria, Australia (n = 258 sites, 3265 survey

nights). Shading indicates 95% prediction envelope.

123 Chapter 5: TSF and NHMs a

124 Chapter 5: TSF and NHMs c

Figure 5.3. Detectability estimates for Pseudomys novaehollandiae

populations a) seasonally; b) over the course of a survey (per night, not

cumulative); and c) with differing amounts of rainfall during a survey.

Estimates derived from occupancy-detectability model of a 48-year Elliott and

camera trapping dataset in Victoria, Australia (n = 258 sites, 3265 survey

nights). Shading indicates 95% prediction envelopes. Other parameters set to

mean values for the calculation of each parameter.

125 Chapter 5: TSF and NHMs

Table 5.1. Parameter estimates for occupancy-detectability model for

Pseudomys novaehollandiae derived from a 48-year Elliott and camera

trapping dataset in Victoria, Australia (n = 258 sites, 3265 survey events).

Occupancy covariates (logit scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μO) N (0, 1000) 0.985 1.592 -1.728 4.893

TSF (β1) N (0, 1000) -0.332 0.281 -0.874 0.240 2 TSF (β2) N (0, 1000) -1.104 0.489 -2.177 -0.228

2 Variance due to Site ( s) G (0.001, 0.001) 0.056 0.016 0.029 0.091

2 Variance due to Region ( r) G (0.001, 0.001) 0.167 0.169 0.017 0.600 Detection covariates (logit scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μP) N (0, 1000) 1.046 0.281 0.566 1.665

Day of year (cos) (β4) N (0, 1000) -0.629 0.280 -1.209 -0.110

Day of year (sin) (β5) N (0, 1000) 0.473 0.216 0.076 0.929

Night of survey (β6) N (0, 1000) 0.092 0.196 -0.258 0.514

Nightly rainfall (β7) N (0, 1000) 0.189 0.152 -0.039 0.533 Moon (full) (M) N (0, 1000) -2.040 0.317 -2.715 -1.466 Method (Elliott trap) (C) N (0, 1000) 1.146 0.394 0.405 1.961

Abundance-detectability modelling

The abundance-detectability modelling incorporated only Elliott trapping data.

Abundance of P. novaehollandiae showed a peak around 10-25 years post-fire for the

0-68 year analyses, however this finding may be an artefact of the model and should be treated with extreme caution given a lack of surveys (and lack of representation in the landscape) of the 30-40 years age bracket (Fig. 5.4a). The species showed strong seasonal fluctuations in abundance (Fig. 5.4b), a slight increase in abundance with increasing rainfall in the 18 months prior to a survey (Fig. 5.4c), and a slight increase

126 Chapter 5: TSF and NHMs in abundance with increasing survey effort – reflective of greater survey area (Fig.

5.4d) (Table 5.2). Detectability showed seasonal fluctuations (Fig. 5.5a), a strong reduction in detectability with increasing moon brightness (Fig. 5.5b), and negligible improvement over the course of a survey (Fig. 5.5c) and with rainfall during a survey

(Fig. 5.5d). Regional, site-specific, and inter-annual differences explained significant amount of variance in P. novaehollandiae abundance (Table 5.2). a

127 Chapter 5: TSF and NHMs b

128 Chapter 5: TSF and NHMs d

Figure 5.4. Abundance estimates for Pseudomys novaehollandiae populations

a) with Time Since Fire (TSF); b) with rainfall over an 18-month period prior

to the survey; and c) with survey effort, derived from N-mixture model of a 48-

year live-trapping dataset in Victoria, Australia, covering Time Since Fire

(TSF) age classes 1-68 years (n = 130 sites, 1159 survey nights). Shading

indicates 95% prediction envelopes. Other parameters set to mean values for

the calculation of each parameter; abundance estimates are proportional rather

than absolute.

129 Chapter 5: TSF and NHMs a

130 Chapter 5: TSF and NHMs c

Figure 5.5. Detectability estimates for Pseudomys novaehollandiae

populations a) seasonally (individual-level – differs from species-level

detectability as presented in Fig. 5.3a); b) with moon brightness; c) over the

131 Chapter 5: TSF and NHMs course of a survey (per night, not cumulative); and d) with differing amounts of rainfall during a survey, derived from an N-mixture model of a 48-year live- trapping dataset in Victoria, Australia, covering Time Since Fire (TSF) age classes 1-68 years. Shading indicates 95% prediction envelopes. Other parameters set to mean values for the calculation of each parameter.

132 Chapter 5: TSF and NHMs

Table 5.2. Parameter estimates from N-mixture model for Pseudomys novaehollandiae, derived from a 48-year live-trapping dataset in Victoria,

Australia, covering Time Since Fire (TSF) age classes 1-68 years (n = 130 sites, 1159 survey nights). Minimally informative priors specified where

N(mean, sd) is the normal distribution; G(alpha, beta) is the Gamma distribution.

0-68 years Abundance covariates (log scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μN) N (0, 1000) 0.547 0.506 -0.462 1.465 Day of year (cos) (β1) N (0, 1000) 0.782 0.252 0.308 1.297 Day of year (sin) (β2) N (0, 1000) 0.827 0.166 0.533 1.178 TSF (β3) N (0, 1000) -1.471 0.350 -2.162 -0.818 2 TSF (β4) N (0, 1000) 0.123 0.239 -0.357 0.569 3 TSF (β5) N (0, 1000) 0.936 0.309 0.342 1.543 18-month rainfall (β6) N (0, 1000) 0.149 0.068 0.016 0.282 Trap effort (β7) N (0, 1000) 0.125 0.060 0.010 0.243 2 Variance due to Region ( r) G (0.001, 0.001) 1.581 1.250 0.276 4.789 2 Variance due to Site ( s) G (0.001, 0.001) 0.854 0.165 0.572 1.216 2 Variance due to Year ( y) G (0.001, 0.001) 2.524 0.941 1.113 4.750 0-68 years Detection covariates (logit scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μP) N (0, 1000) -0.146 0.528 -1.167 0.873 Day of year (cos) (β8) N (0, 1000) -1.861 0.324 -2.449 -1.181 Day of year (sin) (β9) N (0, 1000) -1.292 0.232 -1.722 -0.808 Moon (β10) N (0, 1000) -0.442 0.088 -0.630 -0.308 Night of survey (β11) N (0, 1000) 0.190 0.051 0.100 0.300 Nightly rainfall (β12) N (0, 1000) 0.126 0.043 0.050 0.220 0-15 years Abundance covariates (log scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μN) N (0, 1000) 1.427 0.845 -0.150 3.056 TSF (β1) N (0, 1000) 0.747 0.208 0.362 1.183 2 TSF (β2) N (0, 1000) -0.873 0.186 -1.255 -0.522 2 Variance due to Region ( r) N (0, 1000) 1.258 1.793 0.052 5.533 2 Variance due to Site ( s) G (0.001, 0.001) 1.000 0.285 0.552 1.658 2 Variance due to Year ( y) G (0.001, 0.001) 1.978 1.146 0.582 4.889 0-15 years Detection covariates (logit scale) 95% C.I. Variable Prior Mean S.E. Lower Upper

Intercept (μP) N (0, 1000) -0.948 0.605 -2.217 0.191

133 Chapter 5: TSF and NHMs

The 1-15 year data subset showed a potential peak in abundance 6-11 years post-fire, though with substantial uncertainty such that a much more muted response remains very possible (Fig. 5.6). Again, regional, site-specific, and inter-annual differences explained significant amount of variance in P. novaehollandiae abundance (Table

5.2).

Figure 5.6. Abundance estimates for Pseudomys novaehollandiae populations

with Time Since Fire (TSF) over the 1-15 years following a fire event derived

from N-mixture model of subset of a 48-year live-trapping dataset in Victoria,

Australia (n = 94 sites, 519 survey nights). Shading indicates 95% prediction

envelopes. Other parameters set to mean values for the calculation of each

parameter; abundance estimates are proportional rather than absolute.

Discussion

We collated a large dataset of P. novaehollandiae survey events, spanning multiple decades and sites, across a broad spatial scale. Applying contemporary statistical

134 Chapter 5: TSF and NHMs methods, we conducted the most rigorous tests to date of a hypothesized relationship between TSF and P. novaehollandiae abundance and occupancy. We found that the occurrence of P. novaehollandiae in Victoria was only weakly governed by TSF, if at all. Abundance of the species is also poorly explained by TSF, with the model clearly indicating a peak in abundance occurring in the 0-15 year dataset, but with substantial uncertainty as to the size of that effect (Fig. 5.6). More data will be required to determine the size of this abundance peak, but whether it represents the favoured niche for the species, or simply a post-disturbance demographic response remains to be seen. Certainly, seasonal and inter-annual fluctuations, and differences among sites and regions appear to account for considerable variance in P. novaehollandiae abundance. The wide prediction envelopes in our work, the difference in responses observed when analysing the 1-68 or 1-15 TSF datasets, and the diversity of post-fire responses reported elsewhere for P. novaehollandiae (e.g.

Fox and McKay 1981; Kemper 1990; Wilson and Laidlaw 2003), indicate that the species’ response to fire events does not simplify neatly to a common time-series.

Though P. novaehollandiae may exhibit preferences for certain vegetation structures

(Fox and Fox 1978; Wilson and Laidlaw 2003), the relationship between TSF and these structures is not generalizable across regions or fire events. Individual fire events are accompanied by unique vegetation assemblages, rainfall patterns pre- and post-burn, and burn severities, patterns, and patch sizes, all of which combine to inform how the landscape and thus the local fauna respond in the days and decades following a fire (Swan et al. 2015; Lawes et al. 2015; Chia et al. 2016); P. novaehollandiae is no exception. Considering the occurrence or abundance of P. novaehollandiae in basic terms of TSF is an over-simplification that may mislead management efforts.

135 Chapter 5: TSF and NHMs

Despite the lack of a clear relationship with P. novaehollandiae occurrence across the state, fire remains a useful tool and an important consideration in habitat management for P. novaehollandiae in Victoria. P. novaehollandiae occurs in a broad range of vegetation types (e.g. open-forest, woodland, heathland, shrubland, grassland) with differing species compositions and fire tolerances (Wilson and

Laidlaw 2003), and so fire management strategies need to be individually tailored to each site and adapted to local rainfall patterns and population dynamics. In some locations, fire and strategic fire suppression may be used to control weedy species

(e.g. Coastal Tea Tree Leptospermum laevigatum at Wilsons Promontory), promote suitable vegetation structures, and cultivate desirable food plants. Because the omnivorous diet of P. novaehollandiae is highly variable among seasons and locations (Watts and Braithwaite 1978; Cockburn 1980; Norton 1987; Wilson and

Bradtke 1999), further research is necessary to understand how the species’ diet in extant populations is reflected in local vegetation requirements and how this changes with fire. While there is no uniform maximum fire interval necessary for P. novaehollandiae persistence, maximum fire intervals warrant consideration in areas where desirable native plant species require fire for germination or to maintain species diversity. Similarly, minimum fire intervals should reflect the requirements of key plant species – which in turn should be identified locally.

As there is no global ‘ideal’ fire interval for P. novaehollandiae, the species can potentially persist through a variety of fire management strategies, whether regimes target retention of longer-unburnt vegetation or incorporate fuel reduction burns.

Given the species’ threatened status and clear trend of local population extinctions

136 Chapter 5: TSF and NHMs

(Chapter 6 - Burns in review), however, land managers need to be confident that their proposed management regimes are locally suitable for P. novaehollandiae, taking a broader consideration of the habitat, diet and predation drivers of the species.

Where prescribed burns are necessary or desirable, and expected vegetation outcomes (in terms of both composition and structure) are conducive to long-term P. novaehollandiae persistence, fires should be designed to maximize short-term P. novaehollandiae survival and in situ persistence. While reports of the species’ immediate response during and after fires vary (Kemper 1990; Lock and Wilson

1999; Lazenby et al. 2008), here we draw inference from the limited literature, the species’ biology, and findings from similar species to develop guidelines for burning in P. novaehollandiae occupied areas. We encourage consideration of the following key points, noting that our recommendations are compatible with broader landscape- scale management aims and that recommendations 2) and 3) reflect the requirements of many native species (Lindenmayer et al. 2009; Kelly et al. 2015; Doherty et al.

2015; Doherty et al. 2017; Hradsky et al. 2017):

1) seasonal timing; P. novaehollandiae are seasonal breeders, sometimes producing multiple litters per season (Kemper 1980); in Victoria breeding generally occurs between mid-Spring to mid-Autumn (Wilson 1991; Wilson et al. 2005).

Given their short lifespan of 1-3 years (Lock and Wilson 2017), the loss of a single breeding season may cause a population to crash. We suggest caution in the timing of planned burns to avoid disruption and stress during pregnancy, lactation, and initial dispersal of young; in Victoria, this translates to planned burn timing between mid-

April – mid-September. Where exceptions are necessary – e.g. large, intense

November burns at Wilsons Promontory designed to control Coastal Tea Tree – we

137 Chapter 5: TSF and NHMs suggest additional precautions: extensive pre-burn surveying is recommended, and exclusion zones (see below) should be preserved around P. novaehollandiae records to minimize disruption of breeding individuals. Further research into how the species responds to fire in the short-term at a local level is needed.

2) fire perimeter size and patchiness; Although P. novaehollandiae have been recorded persisting in situ post-fire, individuals have also been observed travelling to and from unburned areas to forage, while maintaining nests within a fire scar

(Lazenby et al. 2008), or moving from burned to unburned habitat permanently

(Kemper 1977). This suggests that unburned refuge habitat may be important for P. novaehollandiae in the initial post-fire environment either as a food source, or a sheltered area to forage with lower predation risk. Large, severe fires with high coverage have the capacity to extirpate the species from a region (Lock and Wilson

1999). We recommend small, patchy burns in P. novaehollandiae habitat to maximize individual survival and facilitate rapid recolonization from nearby unburned habitat, if required. Individual P. novaehollandiae home range estimates vary considerably among locations (0.07 – 1.69 ha) and can increase following fire

(Kemper 1977; Lock and Wilson 1999). Where large burns are necessary, we suggest protecting pockets of unburned habitat > 3 ha in size within the fire perimeter as potential refuges. The maximum distance an individual P. novaehollandiae has been recorded travelling is 1 km (Lazenby et al. 2008), and there are several other incidental records of dispersals 100-500 m (Lock and Wilson 1999; P. Burns, unpub. data). Until targeted research better illuminates the species’ dispersal abilities, we recommend the distance between occupied burned and non-burned habitat should be less than 500 m.

138 Chapter 5: TSF and NHMs

3) predator control; With reduced vegetative cover comes increased predation risk for small ground-dwelling mammals, and fires can lure predators in from the surrounding landscape (McGregor et al. 2016; McGregor et al. 2017). P. novaehollandiae are preyed upon by several native and invasive birds, reptiles, and mesopredators (Slip and Shine 1988; Bilney et al. 2010; Anon. 2010; Davis et al.

2015), with cats and foxes the primary concern throughout their Victorian range

(though P. novaehollandiae has not yet been recorded in their scat; Chapter 6 - Burns in review). We recommend increased cat and fox control in the weeks preceding a planned burn, and that renewed control efforts commence immediately following a burn. Control efforts should remain elevated until ground cover recovers in the fire scar.

Considerable gaps remain in our understanding of how to best manage fire to support

P. novaehollandiae. Many of these gaps can be addressed by incorporating monitoring elements into management programs. Pre- and post-burn P. novaehollandiae scat sampling to assess shifts in diet following fire events could allow us to evaluate whether seasonal timing of burns may be better guided by dietary needs rather than breeding behaviours. Additionally, the live trapping required for P. novaehollandiae scat sampling can be teamed with individual mark- recapture analyses to assess survivorship, site fidelity, and abundance pre- and post- fire, over time compiling a dataset for use in assessing appropriate fire perimeter sizes and burn severities. Predator density estimates prior to and at intervals throughout control programs, teamed with stomach-contents analyses of cat and fox carcasses, could provide a measure of program efficacy in reducing predator numbers

139 Chapter 5: TSF and NHMs following fire, while also potentially providing valuable evidence of predation on P. novaehollandiae.

Although P. novaehollandiae have demonstrated early successional characteristics in some scenarios, considering P. novaehollandiae habitat preferences in terms of TSF dangerously oversimplifies the species’ management requirements. Long-term fire management in P. novaehollandiae occupied areas should reflect local vegetation requirements rather than a global best estimate of ‘ideal’ TSF. Individual burns should be designed to maximize P. novaehollandiae persistence, particularly focusing on seasonal timing, small patchy burn areas, and intensive control of invasive predators, while also incorporating monitoring programs to evaluate the efficacy of these measures and guide adaptation where necessary, as described above.

Despite there being no simple overarching governance of P. novaehollandiae occupancy and abundance by fire, cautious, locally-adapted fire management is necessary to ensure persistence of the species through fire events and in the landscape long-term.

140 Chapter 6: The decline of New Holland Mice

Chapter 6 Testing the decline of the New Holland Mouse (Pseudomys novaehollandiae) in Victoria

Burns, P. A. (in review). Testing the decline of the New Holland Mouse (Pseudomys novaehollandiae) across Victoria.

141 Chapter 6: The decline of New Holland Mice

Abstract

Many Australian rodent species have become extinct or undergone substantial range contractions since European invasion. Limited and haphazard surveying efforts across much of Australia mean we are unsure of many species’ current and former ranges, hampering our ability to identify and remedy causes of decline. The New

Holland Mouse (Pseudomys novaehollandiae) is an endangered rodent species native to south east Australia that is suspected of undergoing rapid and dramatic range contractions and local extinctions in recent decades. Here, I reassess the species’ distribution across Victoria using extensive survey efforts and, subsequently, provide an assessment of key threatening processes and management priorities. In only 40 years, P. novaehollandiae has been lost from seven of the 12 isolated areas where the species was once recorded in Victoria. Habitat loss and fragmentation, invasive predators, and potentially disease and genetic inbreeding have contributed to the species’ rapid and continuing decline. Conservation priorities include cat and fox control, exclusion of rabbit poison-baiting, targeted fire and habitat management, and reintroduction to historically occupied regions where threatening processes have been mitigated.

Introduction

Australia is in the midst of the most rapid mammalian extinction crisis ever recorded in any time or place (Lindenmayer 2015; Woinarski et al. 2015). Since European invasion, at least 30 mammal species have gone extinct and another 109 have declined such that they have a Federal threatened species listing (Woinarski et al.

2015; Department of Environment and Energy 2018b). The worst affected mammalian taxon – rodents – have suffered around 15 extinctions and more than half

142 Chapter 6: The decline of New Holland Mice of the remaining species have declined drastically in distribution (Short and Smith

1994; Morris 2000; Robinson et al. 2000; Seebeck and Menkhorst 2000; Dickman et al. 2000a; Dickman et al. 2000b). Our knowledge of many species’ current distributions is inadequate, and our knowledge of historical distributions is weaker still, leading us to underestimate the severity of recent declines (Bilney et al. 2010).

Meanwhile, a lack of systematic monitoring effort (both historical and contemporary), hampers our ability to identify and manage threatening processes

(Chapter 2 - Burns et al. in review; Bilney et al. 2010; Burns et al. 2016).

The Australian native New Holland Mouse (Pseudomys novaehollandiae; Indigenous name: Pookila) is an exemplar of inadequate historical survey effort and drastic recent distributional declines. First recorded in 1840 in New South Wales, the species went undetected between 1886 and 1967, and was presumed extinct for many decades until a chance hand-capture heralded the species ‘rediscovery’ (Mahoney and Marlow 1968; Mahoney 1974). This 81 year gap in the species’ recorded history was primarily due to a lack of surveying effort – the species went undetected because no-one sought it, and if found inadvertently, it was likely misidentified as Mus musculus (Mahoney and Marlow 1968). Despite the species’ probable occurrence in some of the places earliest occupied by Europeans in Victoria (around the

Mornington Peninsula, Gippsland, and Geelong), approximately 135 years of sheep grazing, land clearing, and settlement occurred before P. novaehollandiae was first observed in the state in 1970 (Seebeck and Beste 1970; Boyce 2011). In that time, large areas of habitat were lost; for example, small, isolated patches occupied by the species in Tyabb, Cranbourne and Langwarrin (Fig. 6.1) were likely remnants of a widespread distribution of P. novaehollandiae through the Western Port Bay and

143 Chapter 6: The decline of New Holland Mice

Mornington Peninsula areas. The lack of early survey effort prevents adequate estimation of the species’ true decline since European invasion in Victoria. Following the 1970 detection, P. novaehollandiae was quickly discovered across several other

Victorian locations in south east Melbourne, Gippsland, and Anglesea (Seebeck and

Beste 1970; Gilmore 1977; Braithwaite and Gullan 1978; Cockburn 1980; Kentish

1982). However, many of these populations disappeared within a few years of their initial discovery (Fig. 6.1), and surveying efforts waned. When targeted efforts to find the species were again made at many historical sites in the 1990s, P. novaehollandiae appeared to have become locally extinct (Wilson 1996; Quin 1996;

Reside and Hooper 1999; Seebeck et al. 2003).

Figure 6.1. Occurrence records of Pseudomys novaehollandiae in Victoria,

Australia, 1970 – 2014. Dates indicate period of recorded occurrence in each

region; light grey shading depicts water bodies. Note: P. novaehollandiae

records at Anglesea 2000-2003 were a product of an unsuccessful

reintroduction attempt and not included here.

144 Chapter 6: The decline of New Holland Mice

Even when surveys for P. novaehollandiae occurred, however, they were not guaranteed to record the species if they were there. As species decline in abundance, the probability of detecting them if present – their detectability – also declines (Kéry

2002; McCarthy et al. 2013; Scheele and Gillespie 2018). This means that defining and implementing adequate survey effort can be particularly challenging for rare or declining species. Historical survey efforts may no longer be sufficient to detect the species, and when this is not identified and addressed in survey planning, the species may be prematurely labelled locally extinct, potentially leading to the removal, or a halt in commencement, of much needed conservation actions (Kéry 2002; Rout et al.

2010; Boakes et al. 2015).

For P. novaehollandiae, detectability varies with seasonal fluctuations in abundance and behaviour, and with environmental factors such as moon phase (Chapter 2 -

Burns et al. in review). If surveys are inopportunely timed, or the species is at low local abundance (as must be assumed in occupancy surveys), standard historical survey efforts using Elliott live small mammal traps do not provide statistical support for asserting the species’ absence when undetected (Chapter 2 - Burns et al. in review). Evaluating the historical survey efforts exerted across Victoria shows that the distributional decline of P. novaehollandiae, while probable, is as yet statistically unsupported. The advent of camera trapping, and its efficacy for P. novaehollandiae

(Chapter 3 - Burns et al. 2018), provides an opportunity to more rigorously test this distributional decline – camera traps can be safely deployed for longer periods than

Elliott traps, improving cumulative detection probabilities and providing greater

145 Chapter 6: The decline of New Holland Mice confidence in species’ absence from a site when not detected (Nelson et al. 2009; De

Bondi et al. 2010).

To test the purported distributional decline of P. novaehollandiae across Victoria, I conducted extensive camera trap surveys across historical and potential regions of occupancy. In light of my findings, I assessed the current status of P. novaehollandiae in Victoria, evaluated potential causes of decline, and outlined ongoing conservation concerns, and outline management priorities for the species.

Methods

Targeted surveys

Between August 2015 and April 2018, I conducted camera trapping surveys at 472 sites across Gippsland and Anglesea, Victoria. Sites were commonly > 500 m apart, with a minimum distance of 100 m apart. I selected sites in areas with 1) historical P. novaehollandiae detections, 2) a high probability of occupancy as predicted by generic SDMs (Chapter 4 - Burns et al. in review), or 3) where on-ground inspection suggested likely occurrence due to habitat suitability. I included a mixture of regions where the species was known previously to occur, and regions where the species had never previously been recorded but that were environmentally and geographically logical. Overall, detection effort was directed to places that were plausibly occupied.

Areas within Gippsland Water property at Dutson Downs were excluded; recent survey data is available elsewhere (P. Homan unpub. data; McCall et al. 2014). Areas around Langwarrin, Tyabb, and Cranbourne were also excluded based on low probability of persistence due to urbanisation of the region.

146 Chapter 6: The decline of New Holland Mice

At each site, I attached one Reconyx Hyperfire HC550 white-flash, HC500 infrared, or PC850 white-flash camera to a tree at 0.10 – 0.30 m from the ground. Each camera was pointed at the base of an elevated bait station set approximately 0.90 m from the camera, with the bait station at a height of approximately 0.10 m. Elevated bait stations consisted of a recycled plastic garden stake and alternately one of either a cutlery drainer containing four tea-strainers, or a vented PVC tube. Each station was baited with a mixture of oats, peanut butter, golden syrup, and vanilla extract.

Cameras were set to capture 3 images per trigger with 1 second delay between images, no delay between triggers, and ‘fast shutter’ night mode. Cameras were left in position for a minimum of 14 nights. Resultant images were identified as per guidelines in (Chapter 3 - Burns et al. 2018).

Occupancy - detectability analyses

To infer the absence of P. novaehollandiae from a site with any degree of confidence, we must assess the species’ detectability – the probability of detecting P. novaehollandiae in a survey if it is present at a site. I estimated occupancy (ѱ) and site-level detectability (p) simultaneously, using repeat observations of nightly detections / non-detections (MacKenzie et al. 2002). My occupancy model estimated a global probability of occupancy, ѱ. Detectability of P. novaehollandiae, however, was allowed to vary with survey-level variables. The most complex detectability model comprised an intercept, P, a fixed effect of moon brightness (M: 1 = full moon, 0 = other), and continuous effects of night of survey (T: 1 – n) and nightly rainfall (R: 24-hour rainfall prior to and including each survey night 6 am – 6 am;

Bureau of Meteorology, 2018).

147 Chapter 6: The decline of New Holland Mice

logit(�) = �� + �1� + �2� + �3�

To aid numeric model fitting, I scaled all continuous variables to have a mean of zero and unit variance. I used the ‘occu’ function in the package ‘unmarked’ (Fiske et al.

2015) in the program ‘R’ v 3.4.3 (R Core Team 2017). I ranked models based on AIC values, eliminated those with ΔAIC > 4 (Burnham and Anderson 2002), then eliminated models with uninformative additional complexity (Arnold 2010). This left only one candidate model remaining.

To determine whether the absences for each site and year are statistically supported, I calculated probability of false absence (pfa) values for each site, based on survey- specific nightly detection probabilities. Because detectability (p) varies across survey night, T, pfa is given by: (1-p,1) (1-p,2)…(1-p,T). Where p,T is the species-level detection on night, T. To achieve high confidence in the species’ absence, pfa should be low, and here I use the criterion that it must be less than 0.05.

Area of Occupancy

Two Area of Occupancy (AOO) estimates were calculated for P. novaehollandiae in

Victoria: a minimum known AOO, and a maximum plausible AOO. To calculate the minimum known AOO, I established 100 m buffers around occurrence records from

2013-2018 (both from data collected here and sources listed in Chapter 2 - Burns et al. in review) then created minimum convex polygons around these buffered records, clustering by geographic isolation and by distinct management areas. Geographic isolation was defined by a maximum recorded P. novaehollandiae dispersal distance of 1 km (Lazenby et al. 2008) and evidence that gene flow is impeded between P.

148 Chapter 6: The decline of New Holland Mice novaehollandiae populations separated by pasture land and non-native vegetation

(e.g. plantations; McCall et al. 2014). 'Maximum plausible AOO’ expanded the above minimum convex polygons to contain nearby modelled suitable habitat, accounting for areas where P. novaehollandiae occurrence was feasible but untested due to inaccessibility (Chapter 4 - Burns et al. in review). Given the low proportion of modelled suitable habitat actually occupied by P. novaehollandiae across the landscape, this likely presents a substantial over-estimation of the species’ true occurrence (Chapter 4 - Burns et al. in review). All mapping and estimations were conducted in the program QGIS (QGIS Development Team 2014). Records pre-2013 were excluded from this analysis as the majority of sites have since been resurveyed and P. novaehollandiae not detected. Knowledge of P. novaehollandiae occurrence in several historically occupied locations (e.g. Langwarrin, Tyabb, Cranbourne, Reeves

Beach / Mcloughlins Beach, Hummock Island, Mullungdung) stems from only one or two sites. This lack of dimensionality prevents meaningful estimation of the species’ recent historical AOO in Victoria, therefore I do not explore it here.

Results

Targeted surveys

The cameras recorded 315,792 images over 8575 trap nights, and detected P. novaehollandiae at 40 of 472 sites across Wilsons Promontory National Park,

Providence Ponds Flora and Fauna Reserve, Gippsland Lakes Coastal Park, and The

Lakes National Park (Sperm Whale Head; Fig. 6.2). These included the first record of P. novaehollandiae in Wilsons Promontory National Park in five years, and the first record in The Lakes National Park in 21 years. See Appendix 1 for a full list of management areas surveyed and distribution of sites among areas.

149 Chapter 6: The decline of New Holland Mice

Figure 6.2. Camera trap survey sites for Pseudomys novaehollandiae across

Victoria, Australia. Circles indicate detection sites (black) and non-detection

sites (dark grey); light grey shading depicts water bodies.

Occupancy - detectability analyses

The top ranked model incorporated detection covariates of moon brightness and night of survey. (Table 6.1; Table 6.2). Overall, occupancy across the study area was low (0.087 ± 0.013 s.e.). Detectability was significantly lower in the five days around a full moon and improved over the course of a survey (Table 6.2). A maximum of 8 nights survey were required for the probability of false absence (pfa) to fall below the a priori threshold of 0.05, conferring very high support for the species’ absence from sites at which P. novaehollandiae was not detected. Hereafter non-detections are treated as absences.

150 Chapter 6: The decline of New Holland Mice

Table 6.1. Model selection results for P. novaehollandiae detectability (p) and

occupancy (ѱ) across Victoria, Australia. Parameters: intercept (.), moon

brightness (moon), night of survey (night), rainfall during survey (rain).

Model logLikelihood AICc ΔAICc p(moon+night) ѱ(.) -615.14 1238.37 0 p(moon+night+rain) ѱ(.) -614.80 1239.73 1.36 p(moon) ѱ(.) -619.43 1244.90 6.53 p(moon+rain) ѱ(.) -619.01 1246.11 7.74 p(.) ѱ(.) -628.55 1261.13 22.76

Table 6.2. Parameter estimates from top-ranked occupancy-detectability model

for Pseudomys novaehollandiae across Victoria, Australia. Parameter estimates

are reported on the logit scale.

95% C.I. Covariate Estimate SE Lower Upper -2.350 0.163 -2.669 -2.031 -0.543 0.146 -0.829 -0.257 �� Moon brightness -0.944 0.269 -1.472 -0.417 �� Night of survey 0.032 0.011 0.011 0.054 (�1)

(�2)

Area of Occupancy

The total minimum known AOO for P. novaehollandiae was 4354 ha spread across

16 discrete (i.e. reproductively isolated) patches within five distinct management areas; AOO estimates differed widely among areas, with Gippsland Lakes Coastal

Park accounting for more than half the species’ Victorian AOO (Table 6.3). Given the patchy distribution of P. novaehollandiae even within occupied areas, the figures presented here are likely an overestimate. The maximum plausible AOO was 6310 ha spread across 12 discrete patches in the five areas. The vast majority (~97%) of P.

151 Chapter 6: The decline of New Holland Mice novaehollandiae occupied habitat in Victoria occurs within 30 km of Lake

Wellington in the Gippsland Lakes district.

Table 6.3. Estimated current minimum known and maximum plausible Areas

of Occupancy (AOO) for Pseudomys novaehollandiae in distinct management

areas across Victoria, Australia.

Minimum known AOO Maximum plausible Region estimate (ha) AOO estimate (ha) Wilsons Promontory National Park 127 480 Gippsland Lakes Coastal Park 2732 3160 Providence Ponds Flora and Fauna Reserve 286 310 Dutson Downs (vested in Gippsland Water) 488 1630 The Lakes National Park (Sperm Whale Head) 720 730 Total 4353 6310

Discussion

These broad-scale surveys across Victoria depict the probable local extinction of P. novaehollandiae from seven of 12 historically occupied areas over a 40-year period

(Fig. 6.3), adding statistically rigorous support to the ongoing narrative of the species’ decline (Wilson 1996; Wilson et al. 2017; Lazenby et al. 2018). Additional survey effort, directed at model-predicted areas of suitable habitat, marginally expanded the known distribution of P. novaehollandiae within some existing regions

(The Lakes National Park / Sperm Whale Head and Gippsland Lakes Coastal Park).

However, these discoveries can be attributed to a paucity of prior survey effort and were partnered with greater declines elsewhere (i.e. at Wilsons Promontory National

Park and Providence Ponds Flora and Fauna Reserve). Further survey effort at

Dutson Downs will likely uncover additional populations, however, this does not

152 Chapter 6: The decline of New Holland Mice lessen the severity of the species’ decline. Overall, P. novaehollandiae occupy an alarmingly small portion of their former known range, the vast majority of which is now centred around Gippsland Lakes Coastal Park. Based on IUCN listing criteria

(IUCN 2012), P. novaehollandiae in Victoria classify as ‘Endangered’ based on a low

AOO (<100 km2), severe fragmentation of remaining populations, and extreme fluctuations in the number of mature individuals (Chapter 2 - Burns et al. in review).

Figure 6.3. Occurrence records of Pseudomys novaehollandiae in Victoria,

Australia. Dates indicate period of recorded occurrence in each area; grey

(extinct) and black (extant) circles indicate detection sites; light grey shading

depicts water bodies. Note: P. novaehollandiae records at Anglesea 2000-2003

were a product of an unsuccessful reintroduction attempt and not included here.

2015-18 data also includes findings from McCall et al. (2015) and P. Homan,

unpub. data.

153 Chapter 6: The decline of New Holland Mice

Causes of decline

Habitat loss and fragmentation

Habitat loss is the most obvious cause of P. novaehollandiae decline in Victoria

(Wilson 1996; Lazenby et al. 2018). Since European invasion, large tracts of suitable habitat in south east Melbourne and Gippsland have been lost to farming or housing, while some previously occupied sites near Anglesea were cleared for mining operations (Wilson et al. 2017). Aside from the direct loss of occupiable habitat, breaking regions into small patches has left populations vulnerable to stochastic events (e.g. predation, disease), and the isolation of these patches means P. novaehollandiae are unable to recolonise areas after such events. Extirpation of populations at Tyabb, Langwarrin, and Cranbourne was likely a product of the small size of remnant habitat patches in those locations and a high density of invasive predators in the fragmented semi-urban, semi-rural south east Melbourne landscape.

As these small populations became geographically isolated by agricultural and urban development, they would also have become genetically isolated (as shown at Dutson

Downs; McCall et al. 2014), potentially losing diversity and suffering from inbreeding. Low genetic diversity in the Anglesea population of P. novaehollandiae may have reduced the species’ fitness and contributed to its local extirpation from the region (Myroniuk 1997; Wilson et al. 2017); genetic inbreeding is plausibly a contributing factor in other local extinctions in Victoria, but this is yet to be tested.

There are also subtler processes of P. novaehollandiae habitat loss and fragmentation than direct habitat clearing. Encroachment of Coastal Tea Tree (Leptospermum laevigatum) is primarily responsible for the range contraction of P. novaehollandiae

154 Chapter 6: The decline of New Holland Mice on the Yanakie Isthmus at Wilsons Promontory, caused by natural but undesirable vegetation succession facilitated by changing land management practices (fire regimes, grazing pressures from livestock, feral herbivores, and overabundant native herbivores; Bennett 1994). Vegetation succession and sand dune erosion likely also played a role in the species’ extirpation from Reeves Beach / Mcloughlins Beach and

Hummock Island (Quin and Williamson 1996), though the small, isolated nature of these populations already placed them at high risk of stochastic events and genetic inbreeding.

The concept of habitat fragmentation as a critical threat to P. novaehollandiae persistence also helps in part to explain the relative stability of remnant populations across Victoria. Gippsland Lakes Coastal Park contains the largest and seemingly most stable P. novaehollandiae populations, likely because it represents the largest contiguous blocks of suitable habitat. Whereas populations at Providence Ponds and

Wilsons Promontory are clustered in much smaller remnant patches of suitable habitat, and have much more erratic detection histories over the past decade and generally lower capture rates (see data sources listed within Chapter 2 - Burns et al. in review). Sperm Whale Head / The Lakes National Park has not been adequately live trapped to understand population dynamics on the peninsula, however the camera trapping results reported here and high habitat suitability (Chapter 4 - Burns et al. in review) hint at a similar dynamic to Gippsland Lakes Coastal Park. Only a small portion of suitable habitat at Dutson Downs has been repeatedly surveyed, with highly varied results among surveys (McCall et al. 2014; P. Homan, unpub. data).

Nevertheless, each remaining population and habitat patch is of extremely high conservation value, regardless of small size (Wintle et al. 2018), and efforts should

155 Chapter 6: The decline of New Holland Mice be made both to prevent any further decline and to expand the species’ current range through assisted recolonization of historically occupied areas where threats have been removed.

Fire

Populations of P. novaehollandiae have demonstrated a broad array of post-fire responses; persisting in situ (Kemper 1990; Lazenby et al. 2008) or being completely extirpated (Lock and Wilson 1999), and booming in early-, mid- or late-successional vegetation (Fox and McKay 1981; Wilson and Laidlaw 2003; Wilson et al. 2017).

While abundance can vary in response to changes in vegetation structure and composition following fire, responses are specific to individual locations and fire events (Chapter 5 - Burns and Phillips in review). Similarly, though occupancy is not limited to any specific post-fire age class, inappropriately large or severe burns have the capacity to extirpate populations, and the complete exclusion of fire can trigger habitat loss through sub-optimal successional processes (Chapter 5 - Burns and

Phillips in review). Though no local extinctions are exclusively attributed to inappropriate fire regimes, planned burning and tactical fire suppression have the capacity to improve or maintain habitat suitability for P. novaehollandiae in remnant habitat.

Drought

Disentangling the potential role of drought from other causes of P. novaehollandiae decline is challenging as historical data is largely clustered in time and clouded by contemporaneous local threatening processes. When accounting for factors such as seasonal fluctuations in abundance and variance among sites and regions, P. novaehollandiae abundance can decline slightly with decreasing rainfall in the 18

156 Chapter 6: The decline of New Holland Mice months prior to a survey (Chapter 5 - Burns and Phillips in review). However, this slight effect is reversed when testing the relationship between rainfall in the 6 and 12 months prior to a survey (Chapter 5 - Burns and Phillips in review), muddying interpretation of a relationship. Prolonged periods of below-average rainfall are unlikely to cause extirpations in isolation, however, abundance can decline during drought (Lock and Wilson 2017), making populations more vulnerable to other threatening processes. With a drier climate predicted for south east Australia over the coming decades (Hennessy et al. 2005), drought may place increasing pressure on remnant P. novaehollandiae populations already afflicted by other threatening processes.

Invasive predators

Predation by feral Cats (Felis catus), Red Foxes (Vulpes vulpes), and Dogs (Canis familiaris) has likely contributed to P. novaehollandiae regional extinctions and declines (Seebeck et al. 2003). Low reproductive rates and the local and global scarcity of P. novaehollandiae mean that even occasional predation events could be catastrophic for a population. Predation pressures may be exacerbated following fire events when burrow entrances are exposed and P. novaehollandiae individuals need to travel further to forage and lack the protection of vegetative cover (Lazenby et al.

2008).

The precise impact of predation on P. novaehollandiae, though potentially significant, is hard to quantify. Despite fox and cat scat surveying in P. novaehollandiae habitat across years and regions, P. novaehollandiae have only been recorded in dog scat (Hoskins 1997; Davis et al. 2015; Lock and Wilson 2017). This

157 Chapter 6: The decline of New Holland Mice may be explained by P. novaehollandiae making up only a small component of any predator’s diet, due to rarity and small body size, and does not necessarily reflect threat level (Allen and Leung 2012). Predation events are most likely rare but extreme – i.e. multiple fatalities at a burrow entrance. The species has also been recorded in the diet of native reptiles and nocturnal birds of prey (Slip and Shine

1988; Bilney et al. 2010).

Invasive rodents and disease

House Mice (Mus musculus) and Black Rats (Rattus rattus) coexist with P. novaehollandiae throughout much of their Victorian range, sometimes in great abundance. Competition with M. musculus does not appear to be an issue (Cockburn

1980; Fox and Pople 1984; Fox and Gullick 1989; Haering and Fox 1997), and nothing has been done to investigate potential impacts of predation by R. rattus. The largest threat both species likely pose is as vectors of novel disease (Abbott 2006;

Morand et al. 2015), however, so little screening has been conducted for such pathogens, that we cannot estimate the prevalence or diversity of parasites with which native rodents have been infected, nor do we have any knowledge of subsequent symptomology (Thompson et al. 2010). Pathogens carried by other species may pose a threat too, for example, Toxoplasma gondii – a virulent parasite spread by cats – is likely present throughout much of the mouse’s former range

(Dickson 2018), and other Pseudomys species have shown high mortality rates when infected with the parasite (N. Hughes, pers. comm.). Rapid declines observed throughout the species’ range could be explained by disease, but targeted research is necessary to test for both the presence and impact of invasive pathogens.

158 Chapter 6: The decline of New Holland Mice

Rabbit poison-baiting

Many native species are susceptible to Pindone and 1080 oat and carrot poison- baiting, and native by-kill is generally considered unavoidable in European Rabbit

(Oryctolagus cuniculus) poison-baiting programs (Sharp and Saunders 2004). Based on LD50 testing of other Pseudomys species, P. novaehollandiae are presumed vulnerable to Pindone and 1080 poisoning (McIlroy 1982; Eisler 1995; Fisher 1999), and readily consume both oats (as bait during Elliott trapping) and carrots (in captivity). Pseudomys novaehollandiae feeding behaviour means they are highly likely to ingest lethal doses of poison, regardless of protective measures such as baiting in the middle of tracks, or coating rather than impregnating oat baits. Baiting programs are, necessarily, designed to eradicate or significantly reduce the abundance of a target species (i.e. rabbits); thus, it is unsurprising that susceptible non-target species are also at risk, particularly if they occur in lower abundance or in rare, isolated populations (Morris 2002). Because P. novaehollandiae commonly persist in small numbers and disjunct habitat patches; poison-baiting has the capacity to extirpate the species from areas which they cannot recolonise unaided. Wide- spread Pindone oat and 1080 carrot baiting on the Yanakie Isthmus in the 1990s and early 2000s may have contributed to the decline of P. novaehollandiae at Wilsons

Promontory, exacerbated by the isolation of populations by Coastal Tea Tree encroachment.

Conservation priorities

Further loss of any P. novaehollandiae habitat, no matter how small a patch, would be a catastrophic misstep given the swathes of habitat already lost and the inherent risks in further fragmentation. Moving forward, the focus should be on sustaining

159 Chapter 6: The decline of New Holland Mice and improving existing P. novaehollandiae habitat and remedying local issues in previously occupied areas to enable assisted recolonizations. Key conservation priorities include:

Fire management

Planned burns should be designed with careful consideration of the local habitat requirements and post-fire behaviour of P. novaehollandiae (Chapter 5 - Burns and

Phillips in review). The species occurs in a wide variety of habitat types, with differing species compositions and vegetation structures (Wilson and Laidlaw 2003;

Seebeck et al. 2003), thus vegetation requirements must be assessed at a local level.

Planned burns are best conducted between mid-April – mid-September to avoid disturbance during breeding, lactation, and initial dispersal of young (Chapter 5 -

Burns and Phillips in review; Kemper 1980; Pye 1991; Wilson et al. 2005). Fires should be small and patchy, with less than 500 m distance between burned and non- burned occupied habitat (Chapter 5 - Burns and Phillips in review). Where larger fires are unavoidable, unburned habitat patches > 3 ha in size should be maintained within the fire perimeter to meet the above guideline (Chapter 5 - Burns and Phillips in review).

Predator control

Long-term investment in comprehensive cat and fox control in known and plausible areas of occupancy should help to reduce predation pressures and, potentially, lessen the spread of Toxoplasmosis by cats. Cat and fox control should be heightened prior to and immediately following planned burns, and maintained at a high level until ground cover has returned (Chapter 5 - Burns and Phillips in review).

160 Chapter 6: The decline of New Holland Mice

Exclusion of Pindone oat or 1080 carrot baiting

Where rabbit control measures are necessary in areas of known or plausible P. novaehollandiae occupancy, poison-baiting must not be used. Target-specific methods such as release of Rabbit Haemorrhagic Disease or shooting (where practical) are recommended instead.

Reintroductions

There is an urgent need to geographically spread the risk faced by P. novaehollandiae. Reintroductions are only suitable in areas where the threatening processes that led to the species’ initial decline can be identified and remedied. The

Royal Botanic Gardens Annex at Cranbourne presents an ideal initial reintroduction site for the species in Victoria as the area is now fully fenced and free from cats and foxes. Genetic issues that may have contributed to the species’ extirpation from the site can be controlled for through careful selection of founder individuals from across the species’ range, following genetic assessment. The success of a recent P. novaehollandiae introduction at Mulligans Flat in the ACT bodes well for careful reintroduction of the species in other parts of their prior range (Abicair et al. in review).

In recent decades, P. novaehollandiae has been extirpated from seven of 12 historically occupied areas in Victoria, representing a fraction of the catastrophic distributional decline and fragmentation likely faced by the species since European invasion. This is in additional to declines in Tasmania (Lazenby et al. 2018), and probable declines in New South Wales (F. Ford, pers. comm.), making P. novaehollandiae conservation a national issue. Long-term conservation of the

161 Chapter 6: The decline of New Holland Mice species requires halting the decline in remnant populations and remedying range contractions where possible. This involves dedicated investment in ongoing habitat management and predator control, carefully considered reintroductions, and continued research into other probable causes of decline including novel disease and genetic inbreeding. Though the species’ decline to date has been considerable, with long-term management and assistance, P. novaehollandiae can persist in Victoria.

162 Chapter 6: The decline of New Holland Mice

Appendix 6.1. Summary of sites surveyed for Pseudomys novaehollandiae (NHMs) by management area across Victoria, Australia.

Region Management Area No. of No. of sites (West-East) (Parks and Reserves) sites with NHMs Anglesea Heath 11 Anglesea Great Otway National Park 26 Mornington HMAS Cerberus 12 Cape Liptrap Coastal Park 4 Drumdlemara H8 Bushland Reserve 2 Mcloughlins Beach – Seaspray Coastal Reserve 2 South Nooramunga Marine and Coastal Park (Mcloughlins Beach) 4 Gippsland Nooramunga Marine and Coastal Park (mainland) 12 Nooramunga Marine and Coastal Park (Snake Island) 10 Wilsons Promontory National Park 91 9 Gelliondale State Forest 13 Giffard (Rifle Range) Fauna Reserve 18 Central Holey Plains State Park 46 Gippsland Mullungdung Flora and Fauna Reserve & State Forest 18 Won Wron Fauna Reserve & State Forest 16 Woodside H28 Bushland Reserve 9 Blond Bay Wildlife Reserve 19 Dutson Bombing Range 5 Gippsland Lakes Coastal Park 57 21 Gippsland Lakes Coastal Park (Boole Poole Peninsula) 6 Gippsland Gippsland Lakes Coastal Park (Glomar Beach) 6 Lakes Moormung Flora and Fauna Reserve 10 Providence Ponds Flora and Fauna Reserve 37 4 Raymond Island Gippsland Lakes Reserve 2 The Lakes National Park (Sperm Whale Head) 7 6 Wattle Point Gippsland Lakes Reserve 3 Cape Conran Coastal Park 9 East Ewing Morass Wildlife Reserve 8 Gippsland Lake Tyers State Park & Lakes Entrance – Lake Tyers Coastal Reserve 6 Marlo Coastal Reserve 3 Total 472 40

163 Chapter 7: General Discussion

Chapter 7 General Discussion

164 Chapter 7: General Discussion

In this thesis I demonstrated, with statistical rigour, the substantial distributional decline of the New Holland Mouse (Pseudomys novaehollandiae) in Victoria over the past 40 years. I surveyed 518 sites across Victoria with 8,552 Elliott trap nights and 10,742 camera trap nights, recording 538 P. novaehollandiae captures and

482,017 images. Combining my survey data with the efforts of ecologists from the past five decades, I used a dataset of 35,255 Elliott trap nights to assess the trapping effort required to achieve statistical confidence in the absence of P. novaehollandiae from a site. I identified that the necessary level of live-trapping effort had rarely been implemented and was largely impractical, except in a few specific months during the new moon (Chapter 2). To overcome the logistical impediment posed by Elliott trapping, I turned to camera trapping for broad-scale surveys. First though, I verified the appropriateness of camera trapping as a survey technique for P. novaehollandiae: running trials to determine whether P. novaehollandiae (and the Smoky Mouse, P. fumeus) were identifiable in camera trap images and whether white-flash or infrared camera traps were better suited to the species (Chapter 3). After finding that P. novaehollandiae was readily identifiable in both white-flash and infrared camera trap images, I proceeded to survey their current distribution across the state of Victoria, while also collecting data to answer further questions about the utility of Species

Distributions Models (SDMs) in recognising the species’ current distribution, and the species’ relationship with Time Since Fire (TSF). I identified that potential shifts in species’ distributions through time need to be accounted for in the interpretation of

SDMs when using these to estimate the distribution of P. novaehollandiae and other declining species (Chapter 4). Combining the Elliott trapping dataset from Chapter 2 and my camera trapping results, I determined that TSF – one of the major variables considered in P. novaehollandiae management – is a gross oversimplification of the

165 Chapter 7: General Discussion species’ habitat requirements and provides a strikingly poor explanation of its abundance and occurrence. Finally, I recognised the local extirpation of P. novaehollandiae from 7 of 12 historically occupied management areas, explored the causes of this decline, and defined management priorities required to conserve the species (Chapter 6).

Practical applications

Many of the findings presented within this thesis have already been embraced by conservation and land management agencies. My identification guides (Chapter 3) are used by staff from DELWP, Parks Victoria, and private consulting companies, and I have trained Department of Environment, Land, Water and Planning (DELWP;

Victoria) staff in camera trapping protocols for P. novaehollandiae. My recommendations regarding P. novaehollandiae and fire events (Chapter 5) have aided in the planning of controlled burns at Wilsons Promontory National Park, and are informing broader fire management goals for DELWP in Gippsland. Parks

Victoria at Wilsons Promontory have committed to excluding Pindone oat rabbit baiting from areas of known or potential P. novaehollandiae occurrence (Chapter 6).

My findings regarding the species’ status and threatening processes (Chapter 6) have informed priorities and conservation actions for P. novaehollandiae in Zoos

Victoria’s 2019-2024 Wildlife Conservation Masterplan. I have also applied my findings from Chapter 6 in DELWP’s Conservation Status Assessment Project, as an independent reviewer of the status of P. novaehollandiae on the state’s Advisory List of Rare or Threatened Vertebrate Fauna in accordance with the Commonwealth approved ‘Common Assessment Method’ (Department of Environment and Energy

2018a). In the near future, my research will directly inform conservation actions for

166 Chapter 7: General Discussion

P. novaehollandiae, including the proposed reintroduction of the species to the Royal

Botanic Gardens Cranbourne Annex in conjunction with Zoos Victoria.

Over the course of my PhD, my research has garnered significant public interest, contributing to improving public recognition and awareness of P. novaehollandiae and other Australian rodent species. My research has been published in print and online through The Age, Sydney Morning Herald, and local newspapers, as well as online through ABC News. I have presented my findings on ABC Gippsland radio, and my work was the subject of an episode of the Radio National program Off Track.

I have presented my research to community groups (i.e. Field Naturalists Clubs and

‘Friends of’ National Park groups), in public seminars hosted by Parks Victoria, to children in wildlife-focused holiday programs, and to a Department of Defence environmental working group. I have also presented my findings at several local and international scientific conferences and to a P. novaehollandiae conservation forum.

Broader implications

This thesis focuses on the decline of one species among the potentially millions to be affected in the Sixth Mass Extinction (Barnosky et al. 2011; Ceballos et al. 2015).

While it may seem a drop in the ocean, species-specific studies such as this are important; obviously for the species itself, but also for similar or interlinked species to which the wider concepts may apply. Additionally, autecological studies such as mine inform the broader narrative; providing the raw material for meta-analyses, and qualitative syntheses.

167 Chapter 7: General Discussion

My findings regarding the strong influence of moon phase on P. novaehollandiae detectability (Chapter 2) are relatively novel in Australia, and similar processes likely occur among many other small mammal species. During brighter moon phases, small mammals are often at higher risk of predation, and in response to this they can display less exploratory behaviour and higher vigilance (Kotler et al. 2010), behaviours which may lessen detectability. Similarly, the strong seasonal shift I identified in P. novaehollandiae detectability (Chapter 2) could be expected for other seasonal breeders, but has rarely been considered in surveys for small mammals. My work on this topic adds to the growing body of literature encouraging the adoption of detectability analyses in ecological surveys (e.g. Wintle et al. 2005; Guillera-Arroita et al. 2010; Guillera-Arroita et al. 2014). Chapter 4 speaks to the space between government policy and scientific understanding, seeking to help bridge the gap.

While to modellers it is obvious that SDMs do not necessarily reflect a species’ current distribution, threatened species’ policy increasingly relies on the erroneous assumption that they do (e.g. DEPI 2013; OEH 2014; DELWP 2017a). As such,

Chapter 4 provides a valuable, explicit demonstration of the limitations of current

SDM methods for estimating distributions of threatened and declining species.

As with all basic research, unanticipated outcomes have also emerged. In the course of this study, a new species was discovered – the Burns’ Mouse Mite (Laelaps burnsae; M. Kwak, pers. comm.) – and we learned of the role of P. novaehollandiae as a host of the Rove Beetle (Myotyphlus newtoni; Jenkins Shaw et al. 2017).

Laelaps burnsae appears to be host-specific to P. novaehollandiae, so its conservation status is intimately linked to that of P. novaehollandiae, and efforts to

168 Chapter 7: General Discussion conserve the mouse in the wild will in turn conserve the mite (M. Kwak, pers. comm.).

Future research

Pseudomys novaehollandiae is comparatively well studied for an Australian rodent species, however, much remains to be learned that will aid in conservation of it and similar species. Priorities for future P. novaehollandiae research include:

Population genetics: Small, geographically isolated remnant P. novaehollandiae populations in Victoria may be at risk from inbreeding. At the phylogeographic scale, mitochondrial analyses suggest that the Victorian and New

South Wales populations have been separated relatively recently (Myroniuk 1997;

Czarny 2005). At the population-level, analyses indicate that contemporary habitat fragmentation now presents impenetrable geographic barriers to reproduction

(McCall et al. 2014). Given that the species’ distribution is now highly fragmented, even within local management areas, we need to assess the degree of inbreeding within these isolated remnant populations. This will in turn inform possible need for genetic rescue via translocations between populations (Tallmon et al. 2004;

Frankham 2015; Whiteley et al. 2015).

Dietary analyses: P. novaehollandiae diet varies seasonally and among locations (Cockburn 1980; Wilson and Bradtke 1999), however, dietary studies have been limited in sample size or seasonal timing and there remains much to ascertain.

Understanding variability within the species’ diet may help us to understand its habitat requirements in more detail, providing clearer and more suitable targets for land managers. Further, if we learn how P. novaehollandiae diet changes following fire, and how this differs among fire events and locations, we will be better equipped

169 Chapter 7: General Discussion to predict the species’ response to individual fire events and to implement supportive measures (e.g. supplemental feeding).

Disease: We have no knowledge of the potential role of disease in local extinctions and decline of P. novaehollandiae across its range. This is true for

Australian mammals in general, and it remains very possible that disease has, and continues to play, an important role in the decline of Australian mammal species.

Future research should consider whether transmission of disease from invasive mammals (rodents and other species) to P. novaehollandiae has occurred, the potential impacts, and whether these may be remedied. In particular, the susceptibility and response of P. novaehollandiae and other Pseudomys species to

Toxoplasmosis is worthy of investigation as most Australian mammals appear to be particularly sensitive to this pathogen, and it is known to be at high prevalence across the entire range of P. novaehollandiae (Dickson 2018).

Regular monitoring: Long-term monitoring is of immense value in adaptive conservation management (Lindenmayer et al. 2012b). My thesis, in many ways, is a testament to the value of long-term monitoring data. The historical data I worked with was collected by numerous people over five decades, and these data in aggregate have provided important information. Using this information, we can now design effective systematic monitoring regimes for known populations. In the years to come, systematic monitoring will provide the data needed to enhance our understanding of how and why populations fluctuate in both local abundance and broader prevalence in the landscape, equipping us to evaluate and adapt management actions, and to identify and combat further causes of decline.

170 Chapter 7: General Discussion

Conclusion

The research presented within this thesis provides a thorough assessment of the current status and recent decline of P. novaehollandiae in Victoria and the threatening processes that have led to this decline (Chapter 6), as well as guidance on monitoring methodologies (Chapters 2 & 3) and conservation and management priorities (Chapters 5 & 6). My research demonstrates the value of modern survey technologies for large resurvey efforts (Chapters 2 & 3), and cautions against misapplication of SDM outputs as standard SDM techniques can fail to recognise distributional shifts in declining species (Chapter 4). More broadly, this work serves as a demonstration of how our knowledge of many species’ current distributions is limited. Compared with numerous other declining small mammal species, P. novaehollandiae has been the subject of substantial research effort over the past five decades. Even so, I uncovered populations that had not previously been recorded, or not seen in two decades. There is much we do not yet know about the species and its requirements, and we are yet to abate its decline.

Numerous processes contributing to the Sixth Mass Extinction have been clearly identified, and combatting many of these major threats requires large-scale political action (e.g. climate change, deforestation, pollution; Derraik 2002; Fearnside 2016;

Wiens 2016). Guiding individual species through these global threatening processes and nuanced local threats, however, often requires incisive efforts based on species- specific information (Hulme 2005). Targeted research and adaptive management programs are necessary to ensure that conservation efforts for threatened species are effective and efficient (Lindenmayer et al. 2012a; Legge et al. 2018). Few species have been assessed for a conservation status, and fewer still are understood well

171 Chapter 7: General Discussion enough to facilitate design of adequate conservation programs (IUCN 2018).

Research that provides insight into species’ conservation status, monitoring requirements and management priorities, such as that contained in this thesis, is vital to global biodiversity conservation.

172 References

References Abbott, I. (2006). Mammalian faunal collapse in Western Australia, 1875-1925: the hypothesised role of epizootic disease and a conceptual model of its origin, introduction, transmission, and spread. Australian Zoologist 33, 530–561. doi:10.7882/AZ.2006.024

Abicair, K., Manning, A. D., Ford, F., Newport, J., and Banks, S. C. (in review). Habitat selection and genetic diversity of a reintroduced ‘refugee species’.

Acevedo, M. A., and Villanueva‐Rivera, L. J. (2006). From the field: using automated digital recording systems as effective tools for the monitoring of birds and amphibians. Wildlife Society Bulletin 34, 211–214. doi:10.2193/0091-7648(2006)34[211:UADRSA]2.0.CO;2

Allen, B. L., and Leung, L. K.-P. (2012). Assessing predation risk to threatened fauna from their prevalence in predator scats: Dingoes and rodents in arid australia. PLOS ONE 7, e36426. doi:10.1371/journal.pone.0036426

Anon. (2010). Approved conservation advice for Pseudomys novaehollandiae (New Holland Mouse). s266B of the Environment Protection and Biodiversity Conservation Act 1999. Australian Government, Canberra.

Ariefiandy, A., Purwandana, D., Seno, A., Ciofi, C., and Jessop, T. S. (2013). Can camera traps monitor komodo dragons a large ectothermic predator? PLOS ONE 8, e58800. doi:10.1371/journal.pone.0058800

Arnold, T. W. (2010). Uninformative Parameters and Model Selection Using Akaike’s Information Criterion. Journal of Wildlife Management 74, 1175– 1178. doi:10.2193/2009-367

Atkin, B. W., and Quin, B. R. (1999). New Holland mouse Pseudomys novaehollandiae (Rodentia: Muridae): Further findings at Yanakie Isthmus, Wilsons Promontory National Park. The Victorian Naturalist 116, 169–172.

Atlas of Living Australia (2014). Atlas of Living Australia. Available at: https://www.ala.org.au/ [accessed 1 July 2014]

Bahn, V., and McGill, B. J. (2013). Testing the predictive performance of distribution models. Oikos 122, 321–331. doi:10.1111/j.1600- 0706.2012.00299.x

Ballard, G., Meek, P. D., Doak, S., Fleming, P. J. S., and Sparkes, J. (2014). Camera traps, sand plots and known events: what do camera traps miss? In ‘Camera Trapping in Wildlife Research and Management’. pp. 189–202. (CSIRO Publishing: Melbourne.)

Barbet-Massin, M., Jiguet, F., Albert, C. H., and Thuiller, W. (2012). Selecting pseudo-absences for species distribution models: how, where and how many?

173 References

Methods in Ecology and Evolution 3, 327–338. doi:10.1111/j.2041- 210X.2011.00172.x

Bardeli, R., Wolff, D., Kurth, F., Koch, M., Tauchert, K.-H., and Frommolt, K.-H. (2010). Detecting bird sounds in a complex acoustic environment and application to bioacoustic monitoring. Pattern Recognition Letters 31, 1524– 1534. doi:10.1016/j.patrec.2009.09.014

Barnosky, A. D., Matzke, N., Tomiya, S., Wogan, G. O. U., Swartz, B., Quental, T. B., Marshall, C., McGuire, J. L., Lindsey, E. L., Maguire, K. C., Mersey, B., and Ferrer, E. A. (2011). Has the Earth’s sixth mass extinction already arrived? Nature 471, 51–57. doi:10.1038/nature09678

Bartoń, K. (2018). ‘MuMIn: Multi-Model Inference.’ Available at: https://CRAN.R- project.org/package=MuMIn

Bellard, C., Cassey, P., and Blackburn, T. M. (2016). Alien species as a driver of recent extinctions. Biology Letters 12, 20150623. doi:10.1098/rsbl.2015.0623

Bennett, A., Nimmo, D., and Clarke, M. (2014). Burnoff policies could be damaging habitats for 100 years. The Conversation. Available at: http://theconversation.com/burnoff-policies-could-be-damaging-habitats-for- 100-years-30240 [accessed 18 December 2018]

Bennett, L. T. (1994). The expansion of Leptospermum laevigatum on the Yanakie Isthmus, Wilson’s Promontory, under changes in the burning and grazing regimes. Australian Journal of Botany 42, 555. doi:10.1071/BT9940555

Bilney, R. J. (2014). Poor historical data drive conservation complacency: The case of mammal decline in south‐eastern Australian forests. Austral Ecology 39, 875–886. doi:10.1111/aec.12145

Bilney, R. J., Cooke, R., and White, J. G. (2010). Underestimated and severe: Small mammal decline from the forests of south-eastern Australia since European settlement, as revealed by a top-order predator. Biological Conservation 143, 52–59. doi:10.1016/j.biocon.2009.09.002

Bland, L. M., Collen, B., Orme, C. D. L., and Bielby, J. (2015). Predicting the conservation status of data-deficient species. Conservation Biology 29, 250– 259. doi:10.1111/cobi.12372

Bluff, L. (2014). Verification of time-since-fire in Gippsland from charring retained on stringybark trees. Fire and adaptive management report No. 92. Department of Environment and Primary Industries, Melbourne.

Blumstein, D. T., Mennill, D. J., Clemins, P., Girod, L., Yao, K., Patricelli, G., Deppe, J. L., Krakauer, A. H., Clark, C., Cortopassi, K. A., Hanser, S. F., McCowan, B., Ali, A. M., and Kirschel, A. N. G. (2011). Acoustic monitoring in terrestrial environments using microphone arrays: applications,

174 References

technological considerations and prospectus. Journal of Applied Ecology 48, 758–767. doi:10.1111/j.1365-2664.2011.01993.x

Boakes, E. H., Fuller, R. A., McGowan, P. J. K., and Mace, G. M. (2016). Uncertainty in identifying local extinctions: the distribution of missing data and its effects on biodiversity measures. Biology Letters 12. doi:10.1098/rsbl.2015.0824

Boakes, E. H., Rout, T. M., and Collen, B. (2015). Inferring species extinction: the use of sighting records Ed D. Hodgson. Methods in Ecology and Evolution 6, 678–687. doi:10.1111/2041-210X.12365

Boland, C. R. J. (2004). Introduced cane toads Bufo marinus are active nest predators and competitors of rainbow bee-eaters Merops ornatus: observational and experimental evidence. Biological Conservation 120, 53–62. doi:10.1016/j.biocon.2004.01.025

Bowman, D. M. J. S. (1998). Tansley Review No. 101 The impact of Aboriginal landscape burning on the Australian biota. New Phytologist 140, 385–410.

Boyce, J. (2011). ‘1835: The Founding of Melbourne and the Conquest of Australia’. (Black Inc.: Melbourne, Victoria.)

Braithwaite, R. W., and Gullan, P. K. (1978). Habitat selection by small mammals in a Victorian heathland. Austral Ecology 3, 109–127. doi:10.1111/j.1442- 9993.1978.tb00857.x

Breed, B., and Ford, F. (2007). ‘Native Mice and Rats’. (CSIRO Publishing: Collingwood, Victoria.)

Breiman, L. (2001). Random Forests. Machine Learning 45, 5–32. doi:10.1023/A:1010933404324

Brito, D. (2004). Lack of adequate taxonomic knowledge may hinder endemic mammal conservation in the Brazilian Atlantic Forest. Biodiversity and Conservation 13, 2135–2144. doi:10.1023/B:BIOC.0000040005.89375.c0

Brooks, T. M., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B., Rylands, A. B., Konstant, W. R., Flick, P., Pilgrim, J., Oldfield, S., Magin, G., and Hilton-Taylor, C. (2002). Habitat loss and extinction in the hotspots of biodiversity. Conservation Biology 16, 909–923. doi:10.1046/j.1523- 1739.2002.00530.x

Bureau of Meteorology (2018). Climate Data Online. Available at: http://www.bom.gov.au/climate/data/index.shtml?bookmark=136 [accessed 24 April 2018]

Burgin, C. J., Colella, J. P., Kahn, P. L., and Upham, N. S. (2018). How many species of mammals are there? Journal of Mammalogy 99, 1–14. doi:10.1093/jmammal/gyx147

175 References

Burnett, S. (1997). Colonizing Cane Toads cause population declines in native predators: reliable anecdotal information and management implications. Pacific Conservation Biology 3, 65–72. doi:10.1071/pc970065

Burnham, K. P., and Anderson, D. R. (2002). ‘Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach’. (Springer Science & Business Media: Sydney.)

Burns, P. A. (in review). Testing the decline of the New Holland Mouse (Pseudomys novaehollandiae) across Victoria. Australian Mammalogy.

Burns, P. A., and Phillips, B. L. (in review). Time since fire is an over-simplified measure of habitat suitability for the New Holland Mouse (Pseudomys novaehollandiae). Journal of Mammalogy.

Burns, P. A., Clemann, N., and White, M. (in review). Testing the utility of generic approaches to species distribution modelling for species in decline. Austral Ecology.

Burns, P. A., Kemper, C. M., and Wilson, B. A. (in press). New Holland Mouse (Pseudomys novaehollandiae). In ‘The Mammals of Australia’ 4th edition. (Eds S. Van Dyck and R. Strahan.)

Burns, P. A., McCall, C., Rowe, K. C., Parrott, M. L., and Phillips, B. L. (in review). Accounting for detectability and abundance in survey design for a declining species. Diversity and Distributions.

Burns, P. A., Parrott, M. L., Rowe, K. C., and Phillips, B. L. (2018). Identification of threatened Pseudomys species using infrared and white-flash camera surveys. Australian Mammalogy 40, 188–197. doi:10.1071/AM17016

Burns, P. A., Rowe, K. M. C., Holmes, B. P., and Rowe, K. C. (2016). Historical resurveys reveal persistence of smoky mouse (Pseudomys fumeus) populations over the long-term and through the short-term impacts of fire. Wildlife Research 42, 668–677. doi:10.1071/WR15096

Burton, A. C., Neilson, E., Moreira, D., Ladle, A., Steenweg, R., Fisher, J. T., Bayne, E., and Boutin, S. (2015). Wildlife camera trapping: a review and recommendations for linking surveys to ecological processes Ed P. Stephens. Journal of Applied Ecology 52, 675–685. doi:10.1111/1365-2664.12432

Butchart, S. H. M., and Bird, J. P. (2010). Data Deficient birds on the IUCN Red List: What don’t we know and why does it matter? Biological Conservation 143, 239–247. doi:10.1016/j.biocon.2009.10.008

Cafaro, P. (2015). Three ways to think about the sixth mass extinction. Biological Conservation 192, 387–393. doi:10.1016/j.biocon.2015.10.017

176 References

Cardoso, P., Erwin, T. L., Borges, P. A. V., and New, T. R. (2011). The seven impediments in invertebrate conservation and how to overcome them. Biological Conservation 144, 2647–2655. doi:10.1016/j.biocon.2011.07.024

Catling, P. C., Coops, N., and Burt, R. J. (2002). 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.

Ceballos, G., Ehrlich, P. R., and Dirzo, R. (2017). Biological annihilation via the ongoing sixth mass extinction signalled by vertebrate population losses and declines. Proceedings of the National Academy of Sciences 114, E6089– E6096. doi:10.1073/pnas.1704949114

Ceballos, G., Ehrlich, P. R., Barnosky, A. D., García, A., Pringle, R. M., and Palmer, T. M. (2015). Accelerated modern human–induced species losses: Entering the sixth mass extinction. Science Advances 1, e1400253. doi:10.1126/sciadv.1400253

Chadès, I., McDonald-Madden, E., McCarthy, M. A., Wintle, B., Linkie, M., and Possingham, H. P. (2008). When to stop managing or surveying cryptic threatened species. Proceedings of the National Academy of Sciences 105, 13936–13940. doi:10.1073/pnas.0805265105

Chia, E. K., Bassett, M., Leonard, S. W. J., Holland, G. J., Ritchie, E. G., Clarke, M. F., and Bennett, A. F. (2016). Effects of the fire regime on mammal occurrence after wildfire: Site effects vs landscape context in fire-prone forests. Forest Ecology and Management 363, 130–139. doi:10.1016/j.foreco.2015.12.008

Clemann, N. (2015). Cold-blooded indifference: a case study of the worsening status of threatened reptiles from Victoria, Australia. Pacific Conservation Biology 21, 15. doi:10.1071/PC14901

Clemann, N., Rowe, K. M. C., Rowe, K. C., Raadik, T. A., Gomon, M., Menkhorst, P. W., Sumner, J., Bray, D., Norman, M., and Melville, J. (2014). Value and impacts of collecting vertebrate voucher specimens, with guidelines for ethical collection. Memoirs of Museum Victoria 72, 141–153.

Cockburn, A. (1980). The diet of the New Holland mouse (Pseudomys novaehollandiae) and the house mouse (Mus musculus) in a Victorian coastal heathland. Australian Mammalogy 3, 31–34.

Colefax, A. P., Butcher, P. A., Kelaher, B. P., and Handling editor: Howard Browman (2018). The potential for unmanned aerial vehicles (UAVs) to conduct marine fauna surveys in place of manned aircraft. ICES Journal of Marine Science 75, 1–8. doi:10.1093/icesjms/fsx100

Cooney, R., and Dickson, B. (2012). ‘Biodiversity and the Precautionary Principle: Risk, Uncertainty and Practice in Conservation and Sustainable Use’. (Earthscan.)

177 References

Cubaynes, S., Pradel, R., Choquet, R., Duchamp, C., Gaillard, J.-M., Lebreton, J.-D., Marboutin, E., Miquel, C., Reboulet, A.-M., Poillot, C., Taberlet, P., and Gimenez, O. (2010). Importance of accounting for detection heterogeneity when estimating abundance: the case of French wolves. Conservation Biology 24, 621–626. doi:10.1111/j.1523-1739.2009.01431.x

Curran, M., Hellweg, S., and Beck, J. (2014). Is there any empirical support for biodiversity offset policy? Ecological Applications 24, 617–632.

Czarny, T. (2005). Genetic study of the New Holland mouse - final report. Animal Gene Resource and Storage Centre of Australia Physiology Department, Monash University, Clayton, Victoria.

Davis, N. E., Forsyth, D. M., Triggs, B., Pascoe, C., Benshemesh, J., Robley, A., Lawrence, J., Ritchie, E. G., Nimmo, D. G., and Lumsden, L. F. (2015). Interspecific and geographic variation in the diets of sympatric carnivores: Dingoes/Wild Dogs and Red Foxes in South-Eastern Australia. PLOS ONE 10, e0120975. doi:10.1371/journal.pone.0120975

De Bondi, N., White, J. G., Stevens, M., and 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. doi:10.1071/WR10046

DELWP - Department of Environment, Land, Water and Planning (2018). ‘Fire history of records of fires primarily on public land’. (State Government of Victoria.) Available at: https://www.data.vic.gov.au/data/dataset/fire-history- records-of-fires-primarily-on-public-land [accessed 1 July 2018]

DELWP (2017a). Protecting Victoria’s environment - biodiversity 2037. The State of Victoria Department of Environment, Land, Water and Planning, Victoria. Available at: https://nla.gov.au/nla.obj-465260268 [accessed 11 September 2018]

DELWP (2017b). Protecting Victoria’s environment: biodiversity 2037 : supporting technical supplement. The State of Victoria Department of Environment, Land, Water and Planning, Victoria. Available at: http://nla.gov.au/nla.obj- 613181469 [accessed 11 September 2018]

DELWP (2014). Victorian Biodiversity Atlas. Available at: https://vba.dse.vic.gov.au/ [accessed 1 July 2014]

Department of Enviroment and Energy (2017). Environment Protection and Biodiversity Conservation Act 1999. Australian Government, ACT. Available at: http://www.environment.gov.au/cgi-bin/sprat/public/sprat.pl [accessed 3 July 2017]

Department of Environment and Energy (2018a). Common Assessment Method. Department of the Environment and Energy. Available at:

178 References

http://www.environment.gov.au/biodiversity/threatened/cam [accessed 7 January 2019]

Department of Environment and Energy (2018b). EPBC Act List of Threatened Fauna. Available at: http://www.environment.gov.au/cgi- bin/sprat/public/publicthreatenedlist.pl [accessed 1 October 2018]

Department of Environment and Energy (2018c). EPBC Act List of Threatened Flora. Available at: http://www.environment.gov.au/cgi- bin/sprat/public/publicthreatenedlist.pl?wanted=flora [accessed 1 October 2018]

DEPI - Department of Environment and Primary Industries (2013). Biodiversity information tools used in Victoria’s native vegetation permitted clearing regulations. The State of Victoria Department of Environment and Primary Industries.

Derraik, J. G. . (2002). The pollution of the marine environment by plastic debris: a review. Marine Pollution Bulletin 44, 842–852. doi:10.1016/S0025- 326X(02)00220-5

Di Stefano, J., Owen, L., Morris, R., Duff, T., and York, A. (2011). Fire, landscape change and models of small mammal habitat suitability at multiple spatial scales. Austral Ecology 36, 638–649. doi:10.1111/j.1442-9993.2010.02199.x

Dickman, C. R., Leung, L. K. P., and Van Dyck, S. M. (2000a). Status, ecological attributes and conservation of native rodents in Queensland. Wildlife Research 27, 333–346. doi:10.1071/WR97130

Dickman, C. R., Lunney, D., and Matthews, A. (2000b). Ecological attributes and conservation of native rodents in New South Wales. Wildlife Research 27, 347–355. doi:10.1071/WR97133

Dickson, J. A. (2018). The distribution of Toxoplasma gondii in Australia. Honours Thesis, University of Melbourne Parkville, Australia.

Diefenbach, D. R., Brauning, D. W., and Mattice, J. A. (2003). Variability in grassland bird counts related to observer differences and species detection rates. The Auk 120, 1168. doi:10.1642/0004- 8038(2003)120[1168:VIGBCR]2.0.CO;2

Doherty, T. S., Dickman, C. R., Nimmo, D. G., and Ritchie, E. G. (2015). Multiple threats, or multiplying the threats? Interactions between invasive predators and other ecological disturbances. Biological Conservation 190, 60–68. doi:10.1016/j.biocon.2015.05.013

Doherty, T. S., van Etten, E. J. B., Davis, R. A., Knuckey, C., Radford, J. Q., and Dalgleish, S. A. (2017). Ecosystem responses to fire: identifying cross-taxa contrasts and complementarities to inform management strategies. Ecosystems 20, 872–884. doi:10.1007/s10021-016-0082-z

179 References

Doherty, T. S., Glen, A. S., Nimmo, D. G., Ritchie, E. G., and Dickman, C. R. (2016). Invasive predators and global biodiversity loss. Proceedings of the National Academy of Sciences 113, 11261–11265. doi:10.1073/pnas.1602480113

Driscoll, D. A., Lindenmayer, D. B., Bennett, A. F., Bode, M., Bradstock, R. A., Cary, G. J., Clarke, M. F., Dexter, N., Fensham, R., Friend, G., Gill, M., James, S., Kay, G., Keith, D. A., MacGregor, C., Russell-Smith, J., Salt, D., Watson, J. E. M., Williams, R. J., and York, A. (2010). Fire management for biodiversity conservation: Key research questions and our capacity to answer them. Biological Conservation 143, 1928–1939. doi:10.1016/j.biocon.2010.05.026

Dubois, A. (2003). The relationships between taxonomy and conservation biology in the century of extinctions. Comptes Rendus Biologies 326, 9–21. doi:10.1016/S1631-0691(03)00022-2

Eisler, R. (1995). Sodium monofluoroacetate (1080) hazards to fish, wildlife, and invertebrates: a synoptic review. No. 30. US Department of the Interior, National Biological Service.

Elith, J., H. Graham, C., P. Anderson, R., Dudík, M., Ferrier, S., Guisan, A., J. Hijmans, R., Huettmann, F., R. Leathwick, J., Lehmann, A., Li, J., G. Lohmann, L., A. Loiselle, B., Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., McC. M. Overton, J., Townsend Peterson, A., J. Phillips, S., Richardson, K., Scachetti-Pereira, R., E. Schapire, R., Soberón, J., Williams, S., S. Wisz, M., and E. Zimmermann, N. (2006). Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29, 129– 151. doi:10.1111/j.2006.0906-7590.04596.x

Elith, J., and Leathwick, J. R. (2009). Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics 40, 677–697. doi:10.1146/annurev.ecolsys.110308.120159

Emison, W. B., Porter, J. W., Norris, K. C., and Apps, G. J. (1978). Survey of the vertebrate fauna of the Grampians–Edenhope area of southwestern Victoria. Memoirs of the National Museum of Victoria 39, 281–363.

Fahrig, L. (1997). Relative effects of habitat loss and fragmentation on population extinction. The Journal of Wildlife Management 61, 603. doi:10.2307/3802168

Falzon, G., Meek, P. D., and Vernes, K. (2014). Computer-assisted identification of Australian rodents in camera trap imagery. In ‘Camera Trapping in Wildlife Research and Management’. pp. 299–306. (CSIRO Publishing: Melbourne.)

Fancourt, B. A., Sweaney, M., and Fletcher, D. B. (2017). More haste, less speed: pilot study suggests camera trap detection zone could be more important than

180 References

trigger speed to maximise species detections. Australian Mammalogy. doi:10.1071/AM17004

Farashi, A., Shariati, M., and Hosseini, M. (2017). Identifying biodiversity hotspots for threatened mammal species in Iran. Mammalian Biology 87, 71–88. doi:10.1016/j.mambio.2017.06.002

Farmer, R. G., Leonard, M. L., and Horn, A. G. (2012). Observer effects and avian- call-count survey quality: Rare-species biases and overconfidence. The Auk 129, 76–86. doi:10.1525/auk.2012.11129

Fearnside, P. M. (2016). Brazilian politics threaten environmental policies. Science 353, 746–748. doi:10.1126/science.aag0254

Field, S. A., Tyre, A. J., Jonzen, N., Rhodes, J. R., and Possingham, H. P. (2004). Minimizing the cost of environmental management decisions by optimizing statistical thresholds. Ecology Letters 7, 669–675. doi:10.1111/j.1461- 0248.2004.00625.x

Fischer, J., and Lindenmayer, D. B. (2000). An assessment of the published results of animal relocations. Biological Conservation 96, 1–11. doi:10.1016/S0006- 3207(00)00048-3

Fisher, P. (1999). Factors affecting the risk of poisoning to New Holland Mouse (Pseudomys novaehollandiae) from rabbit oat baits. Victorian Institute of Animal Science, Victoria.

Fiske, I., Chandler, R., Miller, D., Royle, A., Kery, M., Hostetler, J., and Hutchinson, R. (2015). ‘Package “unmarked”’.

Focardi, S., De Marinis, A. M., Rizzotto, M., and Pucci, A. (2001). Comparative evaluation of thermal infrared imaging and spotlighting to survey wildlife. Wildlife Society Bulletin (1973-2006) 29, 133–139.

Fox, B. (1982). Fire and mammalian secondary succession in an Australian coastal heath. Ecology 63, 1332–1341. doi:10.2307/1938861

Fox, B. J. (1990). Changes in the structure of mammal communities over successional time scales. Oikos 59, 321–329.

Fox, B. J., and Fox, M. D. (1978). Recolonization of coastal heath by Pseudomys novaehollandiae (Muridae) following sand mining. Austral Ecology 3, 447– 465. doi:10.1111/j.1442-9993.1978.tb01191.x

Fox, B. J., and Gullick, G. (1989). Interspecific competition between mice: A reciprocal field manipulation experiment. Austral Ecology 14, 357–366. doi:10.1111/j.1442-9993.1989.tb01444.x

Fox, B. J., and McKay, G. M. (1981). Small mammal responses to pyric successional changes in eucalypt forest. Australian Journal of Ecology 6, 29–41.

181 References

Fox, B. J., and Pople, A. R. (1984). Experimental confirmation of interspecific competition between native and introduced mice. Austral Ecology 9, 323– 334. doi:10.1111/j.1442-9993.1984.tb01370.x

Fox, B. J., Taylor, J. E., and Thompson, P. T. (2003). Experimental manipulation of habitat structure: a retrogression of the small mammal succession. Journal of Animal Ecology 72, 927–940. doi:10.1046/j.1365-2656.2003.00765.x

Fox, M. D., and Fox, B. J. (1986). The effect of fire frequency on the structure and floristic composition of a woodland understorey. Austral Ecology 11, 77–85. doi:10.1111/j.1442-9993.1986.tb00919.x

Frankham, R. (2015). Genetic rescue of small inbred populations: meta-analysis reveals large and consistent benefits of gene flow. Molecular Ecology 24, 2610–2618. doi:10.1111/mec.13139

Gelman, A., and Rubin, D. B. (1992). Inference from iterative simulation using multiple sequences. Statistical Science 7, 457–472. doi:10.1214/ss/1177011136

Gibson, L. A. (2011). The importance of incorporating imperfect detection in biodiversity assessments: a case study of small mammals in an Australian region: Imperfect detection in biodiversity assessments. Diversity and Distributions 17, 613–623. doi:10.1111/j.1472-4642.2011.00766.x

Gilmore, A. M. (1977). A survey of vertebrate animals in the Stradbroke area of South Gippsland, Victoria. The Victorian Naturalist 94, 123–128.

Gilmore, S. E. R. (2016). The influence of illumination and moon phase on activity levels of nocturnal mammalian pests in New Zealand. Masters thesis, Lincoln University. Available at: https://researcharchive.lincoln.ac.nz/handle/10182/7811 [accessed 7 March 2018]

Glen, A. S., Cockburn, S., Nichols, M., Ekanayake, J., and Warburton, B. (2013). Optimising camera traps for monitoring small mammals Ed T. Deschner. PLoS ONE 8, e67940. doi:10.1371/journal.pone.0067940

Gu, W., and Swihart, R. K. (2004). Absent or undetected? Effects of non-detection of species occurrence on wildlife–habitat models. Biological Conservation 116, 195–203. doi:10.1016/S0006-3207(03)00190-3

Guillera-Arroita, G. (2016). Modelling of species distributions, range dynamics and communities under imperfect detection: advances, challenges and opportunities. Ecography 39, 001–015. doi:10.1111/ecog.02445

Guillera-Arroita, G., Lahoz-Monfort, J. J., MacKenzie, D. I., Wintle, B. A., and McCarthy, M. A. (2014). Ignoring imperfect detection in biological surveys is dangerous: a response to ‘Fitting and interpreting occupancy models’’’ Ed E. P. White. PLoS ONE 9, e99571. doi:10.1371/journal.pone.0099571

182 References

Guillera-Arroita, G., Ridout, M. S., and Morgan, B. J. T. (2010). Design of occupancy studies with imperfect detection: Design of occupancy studies. Methods in Ecology and Evolution 1, 131–139. doi:10.1111/j.2041- 210X.2010.00017.x

Guisan, A., Tingley, R., Baumgartner, J. B., Naujokaitis-Lewis, I., Sutcliffe, P. R., Tulloch, A. I. T., Regan, T. J., Brotons, L., McDonald-Madden, E., Mantyka- Pringle, C., Martin, T. G., Rhodes, J. R., Maggini, R., Setterfield, S. A., Elith, J., Schwartz, M. W., Wintle, B. A., Broennimann, O., Austin, M., Ferrier, S., Kearney, M. R., Possingham, H. P., and Buckley, Y. M. (2013). Predicting species distributions for conservation decisions Ed H. Arita. Ecology Letters 16, 1424–1435. doi:10.1111/ele.12189

Haering, R., and Fox, B. J. (1997). Habitat use by sympatric populations of Pseudomys novaehollandiae and Mus domesticus in coastal heathland. Austral Ecology 22, 69–80. doi:10.1111/j.1442-9993.1997.tb00642.x

Hale, S., Nimmo, D. G., Cooke, R., Holland, G., James, S., Stevens, M., De Bondi, N., Woods, R., Castle, M., Campbell, K., Senior, K., Cassidy, S., Duffy, R., Holmes, B., and White, J. G. (2016). Fire and climatic extremes shape mammal distributions in a fire-prone landscape Ed A. Syphard. Diversity and Distributions. doi:10.1111/ddi.12471

Harley, D. K. P., Holland, G. J., Hradsky, B. A., and Antrobus, J. S. (2014). The use of camera traps to detect arboreal mammals: lessons from targeted surveys for the cryptic Leadbeater’s possum (Gymnobelideus leadbeateri). In ‘Camera Trapping in Wildlife Research and Management’. pp. 233–243. (CSIRO Publishing: Melbourne.)

Haslem, A., Kelly, L. T., Nimmo, D. G., Watson, S. J., Kenny, S. A., Taylor, R. S., Avitabile, S. C., Callister, K. E., Spence-Bailey, L. M., Clarke, M. F., and Bennett, A. F. (2011). Habitat or fuel? Implications of long-term, post-fire dynamics for the development of key resources for fauna and fire: Habitat or fuel? Post-fire resource dynamics. Journal of Applied Ecology 48, 247–256. doi:10.1111/j.1365-2664.2010.01906.x

Hennessy, K. J., Lucas, C., Nicholls, N., Bathols, J., Suppiah, R., and Ricketts, J. (2005). ‘Climate change impacts on fire-weather in south-east Australia’. (CSIRO Marine and Atmospheric Research: Aspendale, Victoria.)

Hocking, G. J. (1980). The occurrence of the New Holland Mouse, Pseudomys novaehollandiae (Waterhouse), in Tasmania. Wildlife Research 7, 71–77. doi:10.1071/WR9800071

Hoffmann, M., Hilton-Taylor, C., Angulo, A., Böhm, M., Brooks, T. M., Butchart, S. H. M., Carpenter, K. E., Chanson, J., Collen, B., Cox, N. A., Darwall, W. R. T., Dulvy, N. K., Harrison, L. R., Katariya, V., Pollock, C. M., Quader, S., Richman, N. I., Rodrigues, A. S. L., Tognelli, M. F., Vié, J.-C., Aguiar, J. M., Allen, D. J., Allen, G. R., Amori, G., Ananjeva, N. B., Andreone, F., Andrew, P., Ortiz, A. L. A., Baillie, J. E. M., Baldi, R., Bell, B. D., Biju, S. D., Bird, J.

183 References

P., Black-Decima, P., Blanc, J. J., Bolaños, F., Bolivar-G, W., Burfield, I. J., Burton, J. A., Capper, D. R., Castro, F., Catullo, G., Cavanagh, R. D., Channing, A., Chao, N. L., Chenery, A. M., Chiozza, F., Clausnitzer, V., Collar, N. J., Collett, L. C., Collette, B. B., Fernandez, C. F. C., Craig, M. T., Crosby, M. J., Cumberlidge, N., Cuttelod, A., Derocher, A. E., Diesmos, A. C., Donaldson, J. S., Duckworth, J. W., Dutson, G., Dutta, S. K., Emslie, R. H., Farjon, A., Fowler, S., Freyhof, J., Garshelis, D. L., Gerlach, J., Gower, D. J., Grant, T. D., Hammerson, G. A., Harris, R. B., Heaney, L. R., Hedges, S. B., Hero, J.-M., Hughes, B., Hussain, S. A., M, J. I., Inger, R. F., Ishii, N., Iskandar, D. T., Jenkins, R. K. B., Kaneko, Y., Kottelat, M., Kovacs, K. M., Kuzmin, S. L., Marca, E. L., Lamoreux, J. F., Lau, M. W. N., Lavilla, E. O., Leus, K., Lewison, R. L., Lichtenstein, G., Livingstone, S. R., Lukoschek, V., Mallon, D. P., McGowan, P. J. K., McIvor, A., Moehlman, P. D., Molur, S., Alonso, A. M., Musick, J. A., Nowell, K., Nussbaum, R. A., Olech, W., Orlov, N. L., Papenfuss, T. J., Parra-Olea, G., Perrin, W. F., Polidoro, B. A., Pourkazemi, M., Racey, P. A., Ragle, J. S., Ram, M., Rathbun, G., Reynolds, R. P., Rhodin, A. G. J., Richards, S. J., Rodríguez, L. O., Ron, S. R., Rondinini, C., Rylands, A. B., Mitcheson, Y. S. de, Sanciangco, J. C., Sanders, K. L., Santos-Barrera, G., Schipper, J., Self-Sullivan, C., Shi, Y., Shoemaker, A., Short, F. T., Sillero-Zubiri, C., Silvano, D. L., Smith, K. G., Smith, A. T., Snoeks, J., Stattersfield, A. J., Symes, A. J., Taber, A. B., Talukdar, B. K., Temple, H. J., Timmins, R., Tobias, J. A., Tsytsulina, K., Tweddle, D., Ubeda, C., Valenti, S. V., Dijk, P. P. van, Veiga, L. M., Veloso, A., Wege, D. C., Wilkinson, M., Williamson, E. A., Xie, F., Young, B. E., Akçakaya, H. R., Bennun, L., Blackburn, T. M., Boitani, L., Dublin, H. T., Fonseca, G. A. B. da, Gascon, C., Lacher, T. E., Mace, G. M., Mainka, S. A., McNeely, J. A., Mittermeier, R. A., Reid, G. M., Rodriguez, J. P., Rosenberg, A. A., Samways, M. J., Smart, J., Stein, B. A., and Stuart, S. N. (2010). The impact of conservation on the status of the world’s vertebrates. Science 330, 1503–1509. doi:10.1126/science.1194442

Hollis, G. J. (1999). Pre- and post-fire trapping in Gippsland Lakes Coastal Park: an update on monitoring of Pseudomys novaehollandiae and other small mammals. Department of Natural Resources and Environment, Warragul, Victoria.

Hollis, G. J. (1996). Survey for the New Holland Mouse (Pseudomys novaehollandiae) in Gippsland Lakes Coastal Park and Providence Ponds Flora and Fauna Reserve, Gippsland Area, Victoria. 1996. Department of Natural Resources and Environment, Warragul, Victoria.

Holmes, B. (2012). Detection and distribution of smoky mice (Pseudomys fumeus) using remote digital cameras in the Grampians National Park 2012. Unpublished Internal Report. Parks Victoria, Halls Gap, Victoria.

Hoskins, M. (1997). Fox baiting for the protection of the New Holland Mouse. Report to the New Holland Mouse Recovery Team. Parks Victoria, Foster, Victoria.

184 References

Hradsky, B. A., Mildwaters, C., Ritchie, E. G., Christie, F., and Di Stefano, J. (2017). Responses of invasive predators and native prey to a prescribed forest fire. Journal of Mammalogy 98, 835–847. doi:10.1093/jmammal/gyx010

Hulme, P. E. (2005). Adapting to climate change: is there scope for ecological management in the face of a global threat? Journal of Applied Ecology 42, 784–794. doi:10.1111/j.1365-2664.2005.01082.x

IUCN (2012). ‘IUCN Red List Categories and Criteria: Version 3.1’ Second. (IUCN: Gland, Switzerland and Cambridge, UK.) Available at: http://www.iucnredlist.org/technical-documents/categories-and-criteria/2001- categories-criteria#critical [accessed 5 October 2015]

IUCN (2018). The IUCN Red List of Threatened Species. Version 2018-2. Available at: http://www.iucnredlist.org [accessed 21 December 2018]

Jarnevich, C. S., Stohlgren, T. J., Kumar, S., Morisette, J. T., and Holcombe, T. R. (2015). Caveats for correlative species distribution modelling. Ecological Informatics 29, 615. doi:10.1016/j.ecoinf.2015.06.007

Jenkins Shaw, J., Kwak, M. L., Burns, P. A., and Solodovnikov, A. (2017). New host records of the rove beetle ‘Myotyphlus newtoni’ Solodovnikov and Jenkins Shaw, 2016 (Coleoptera: Staphylinidae: Staphylininae), with notes on its association with mammals. Australian Entomologist, The 44, 209.

Jetz, W., Sekercioglu, C. H., and Watson, J. E. M. (2008). Ecological correlates and conservation implications of overestimating species geographic ranges. Conservation Biology 22, 110–119. doi:10.1111/j.1523-1739.2007.00847.x

Jones, H. P., Holmes, N. D., Butchart, S. H. M., Tershy, B. R., Kappes, P. J., Corkery, I., Aguirre-Muñoz, A., Armstrong, D. P., Bonnaud, E., Burbidge, A. A., Campbell, K., Courchamp, F., Cowan, P. E., Cuthbert, R. J., Ebbert, S., Genovesi, P., Howald, G. R., Keitt, B. S., Kress, S. W., Miskelly, C. M., Oppel, S., Poncet, S., Rauzon, M. J., Rocamora, G., Russell, J. C., Samaniego-Herrera, A., Seddon, P. J., Spatz, D. R., Towns, D. R., and Croll, D. A. (2016). Invasive mammal eradication on islands results in substantial conservation gains. Proceedings of the National Academy of Sciences 113, 4033–4038. doi:10.1073/pnas.1521179113

Keith, K., and Calaby, J. H. (1968). The New Holland mouse, Pseudomys novaehollandiae (Waterhouse), in the Port Stephens district, New South Wales. Wildlife Research 13, 45–48. doi:10.1071/CWR9680045

Kelly, L. T., Bennett, A. F., Clarke, M. F., and McCarthy, M. A. (2015). Optimal fire histories for biodiversity conservation: fire and biodiversity conservation. Conservation Biology 29, 473–481. doi:10.1111/cobi.12384

Kelly, L. T., and Brotons, L. (2017). Using fire to promote biodiversity. Science 355, 1264–1265. doi:10.1126/science.aam7672

185 References

Kemper, C. M. (1977). Biology of the New Holland Mouse. PhD thesis, Macquarie University New South Wales.

Kemper, C. M. (1976). Reproduction of Pseudomys novaehollandiae (Muridae) in the laboratory. Australian Journal of Zoology 24, 159–167. doi:10.1071/ZO9760159

Kemper, C. M. (1980). Reproduction of Pseudomys novaehollandiae (Muridae) in the wild. Wildlife Research 7, 385–402. doi:10.1071/WR9800385

Kemper, C. M. (1990). Small mammals and habitat disturbance in open forest of coastal New-South-Wales . I. Population parameters. Wildlife Research 17, 195–206. doi:10.1071/WR9900195

Kentish, K. M. (1982). A new record of the New Holland Mouse (Pseudomys novaehollandiae) from Anglesea, Victoria. Victorian Naturalist 99, 128–129.

Kéry, M. (2008). Estimating abundance from bird counts: binomial mixture models uncover complex covariate relationships. The Auk 125, 336–345. doi:10.1525/auk.2008.06185

Kéry, M. (2002). Inferring the absence of a species: a case study of snakes. The Journal of Wildlife Management 66, 330–338. doi:10.2307/3803165

Kéry, M., Dorazio, R. M., Soldaat, L., Strien, A. V., Zuiderwijk, A., and Royle, J. A. (2009). Trend estimation in populations with imperfect detection. Journal of Applied Ecology 46, 1163–1172. doi:10.1111/j.1365-2664.2009.01724.x

Kotler, B. P., Brown, J., Mukherjee, S., Berger-Tal, O., and Bouskila, A. (2010). Moonlight avoidance in gerbils reveals a sophisticated interplay among time allocation, vigilance and state-dependent foraging. Proceedings of the Royal Society B: Biological Sciences 277, 1469–1474. doi:10.1098/rspb.2009.2036

Lahoz-Monfort, J. J., Guillera-Arroita, G., and Wintle, B. A. (2014). Imperfect detection impacts the performance of species distribution models: Imperfect detection impacts species distribution models. Global Ecology and Biogeography 23, 504–515. doi:10.1111/geb.12138

Lawes, M. J., Murphy, B. P., Fisher, A., Woinarski, J. C. Z., Edwards, A. C., and Russell-Smith, J. (2015). Small mammals decline with increasing fire extent in northern Australia: evidence from long-term monitoring in Kakadu National Park. International Journal of Wildland Fire 24, 712. doi:10.1071/WF14163

Lazenby, B. T., Bell, P., Driessen, M. M., Pemberton, D., and Dickman, C. R. (2018). Evidence for a recent decline in the distribution and abundance of the New Holland mouse (Pseudomys novaehollandiae) in Tasmania, Australia. Australian Mammalogy. doi:10.1071/AM18003

186 References

Lazenby, B. T., Pye, T., Richardson, A., and Bryant, S. A. (2008). Towards a habitat model for the New Holland Mouse Pseudomys novaehollandiae in Tasmania - population vegetation associations and an investigation into individual habitat use. Australian Mammalogy 29, 137–148.

Legge, S., Scheele, B. C., Woinarski, J. C. Z., Garnett, S. T., Keith, D. A., Lintermans, M., Robinson, N. M., and Lindenmayer, D. B. (2018). Summary: monitoring extent and adequacy for threatened biodiversity. In ‘Monitoring Threatened Species and Ecological Communities’. pp. 127–133. (CSIRO Publishing: Australia.)

Lillywhite, H. B., and Brischoux, F. (2012). Is it better in the moonlight? Nocturnal activity of insular cottonmouth snakes increases with lunar light levels. Journal of Zoology 286, 194–199. doi:10.1111/j.1469-7998.2011.00866.x

Lindenmayer, D. B. (2015). Continental-level biodiversity collapse. Proceedings of the National Academy of Sciences 112, 4514–4515. doi:10.1073/pnas.1502766112

Lindenmayer, D. B. (1994). Timber Harvesting Impacts on Wildlife: Implications for Ecologically Sustainable Forest Use. Australian Journal of Environmental Management 1, 56–68. doi:10.1080/13221698.1994.11978480

Lindenmayer, D. B., Blair, D., McBurney, L., and Banks, S. C. (2016). Ignoring the science in failing to conserve a faunal icon – major political, policy and management problems in preventing the extinction of Leadbeater’s possum. Pacific Conservation Biology 21, 257–265. doi:10.1071/PC15022

Lindenmayer, D. B., Gibbons, P., Bourke, M., Burgman, M., Dickman, C. R., Ferrier, S., Fitzsimons, J., Freudenberger, D., Garnett, S. T., Groves, C., Hobbs, R. J., Kingsford, R. T., Krebs, C., Legge, S., Lowe, A. J., Mclean, R., Montambault, J., Possingham, H., Radford, J., Robinson, D., Smallbone, L., Thomas, D., Varcoe, T., Vardon, M., Wardle, G., Woinarski, J., and Zerger, A. (2012a). Improving biodiversity monitoring. Austral Ecology 37, 285–294. doi:10.1111/j.1442-9993.2011.02314.x

Lindenmayer, D. B., Likens, G. E., Andersen, A., Bowman, D., Bull, C. M., Burns, E., Dickman, C. R., Hoffmann, A. A., Keith, D. A., Liddell, M. J., Lowe, A. J., Metcalfe, D. J., Phinn, S. R., Russell‐Smith, J., Thurgate, N., and Wardle, G. M. (2012b). Value of long-term ecological studies. Austral Ecology 37, 745–757. doi:10.1111/j.1442-9993.2011.02351.x

Lindenmayer, D. B., MacGregor, C., Wood, J. T., Cunningham, R. B., Crane, M., Michael, D., Montague-Drake, R., Brown, D., Fortescue, M., Dexter, N., Hudson, M., and Gill, A. M. (2009). What factors influence rapid post-fire site re-occupancy? A case study of the endangered Eastern Bristlebird in eastern Australia. International Journal of Wildland Fire 18, 84. doi:10.1071/WF07048

187 References

Liu, C., Berry, P. M., Dawson, T. P., and Pearson, R. G. (2005). Selecting thresholds of occurrence in the prediction of species distributions. Ecography 28, 385– 393.

Liu, C., Newell, G., and White, M. (2016). On the selection of thresholds for predicting species occurrence with presence-only data. Ecology and Evolution 6, 337–348. doi:10.1002/ece3.1878

Liu, C., White, M., and Newell, G. (2018). Detecting outliers in species distribution data. Journal of Biogeography 45, 164–176. doi:10.1111/jbi.13122

Liu, C., White, M., and Newell, G. (2013a). Selecting thresholds for the prediction of species occurrence with presence-only data. Journal of Biogeography 40, 778–789. doi:10.1111/jbi.12058

Liu, C., White, M., Newell, G., and Griffioen, P. (2013b). Species distribution modelling for conservation planning in Victoria, Australia. Ecological Modelling 249, 68–74. doi:10.1016/j.ecolmodel.2012.07.003

Lock, M. L. (2005). Population ecology and captive breeding of Pseudomys novaehollandiae at Anglesea. PhD thesis, Deakin University Geelong, Victoria.

Lock, M. L., and Wilson, B. A. (1999). The distribution of the New Holland mouse (Pseudomys novaehollandiae) with respect to vegetation near Anglesea, Victoria. Wildlife Research 26, 565–577. doi:10.1071/WR97050

Lock, M., and Wilson, B. A. (2017). Influence of rainfall on population dynamics and survival of a threatened rodent (Pseudomys novaehollandiae) under a drying climate in coastal woodlands of south-eastern Australia. Australian Journal of Zoology 65, 60–70. doi:10.1071/ZO16084

Loiselle, B. A., Howell, C. A., Graham, C. H., Goerck, J. M., Brooks, T., Smith, K. G., and Williams, P. H. (2003). Avoiding Pitfalls of Using Species Distribution Models in Conservation Planning. Conservation Biology 17, 1591–1600. doi:10.1111/j.1523-1739.2003.00233.x

Looby, M., and Gilmore, D. (2016). Loch Car Park upgrades and works: Flora and fauna assessment. Report for Mount Hotham Alpine Resort Management Board. Biosis Pty Ltd, Wangaratta.

Lotz, A., and Allen, C. R. (2007). Observer bias in anuran call surveys. Journal of Wildlife Management 71, 675–679. doi:10.2193/2005-759

Lunt, I. D. (1991). Management of remnant lowland grasslands and grassy woodlands for nature conservation: a review. The Victorian Naturalist 108, 56–66.

MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Andrew Royle, J., and Langtimm, C. A. (2002). Estimating site occupancy rates when detection

188 References

probabilities are less than one. Ecology 83, 2248–2255. doi:10.1890/0012- 9658(2002)083[2248:ESORWD]2.0.CO;2

Mahoney, J. A. (1974). The Australian rodent specimens (Muridae) of J. E. Gray’s: list of the specimens of Mammalia in the collection of the British Museum (1843). Australian Mammalogy 1, 213–242.

Mahoney, J. A., and Marlow, B. J. (1968). The rediscovery of the New Holland Mouse. Australian Journal of Science 31, 221–223.

Manzo, E., Bartolommei, P., Rowcliffe, J. M., and Cozzolino, R. (2012). Estimation of population density of European pine marten in central Italy using camera trapping. Acta Theriologica 57, 165–172. doi:10.1007/s13364-011-0055-8

Maron, M., Laurance, B., Pressey, B., Catterall, C. P., Watson, J., and Rhodes, J. (2015). Land clearing in Queensland triples after policy ping pong. The Conversation. Available at: https://theconversation.com/land-clearing-in- queensland-triples-after-policy-ping-pong-38279 [accessed 23 December 2018]

Mathewson, P. D., Moyer‐Horner, L., Beever, E. A., Briscoe, N. J., Kearney, M., Yahn, J. M., and Porter, W. P. (2017). Mechanistic variables can enhance predictive models of endotherm distributions: the American pika under current, past, and future climates. Global Change Biology 23, 1048–1064. doi:10.1111/gcb.13454

Mazerolle, M. J., Bailey, L. L., Kendall, W. L., Andrew Royle, J., Converse, S. J., and Nichols, J. D. (2007). Making great leaps forward: Accounting for detectability in herpetological field studies. Journal of Herpetology 41, 672– 689. doi:10.1670/07-061.1

McAlpine, C., Lunney, D., Melzer, A., Menkhorst, P., Phillips, S., Phalen, D., Ellis, W., Foley, W., Baxter, G., de Villiers, D., Kavanagh, R., Adams-Hosking, C., Todd, C., Whisson, D., Molsher, R., Walter, M., Lawler, I., and Close, R. (2015). Conserving koalas: A review of the contrasting regional trends, outlooks and policy challenges. Biological Conservation 192, 226–236. doi:10.1016/j.biocon.2015.09.020

McCall, C., Reside, J., and Collyer, A. (2015). An assessment of the status of New Holland mouse populations in Gippsland: surveys for the New Holland mouse at Providence Ponds Flora and Fauna Reserve and Gippsland Lakes Coastal Park. Autumn 2015. Wildlife Unlimited Pty Ltd, Bairnsdale, Victoria.

McCall, C., Reside, J., van Rooyen, A., and Weeks, A. (2014). Surveys for the New Holland Mouse at Dutson Downs: Genetic Sampling, 2013-2014. Wildlife Unlimited Pty Ltd and cesar Pty Ltd.

McCarthy, M. A., Moore, J. L., Morris, W. K., Parris, K. M., Garrard, G. E., Vesk, P. A., Rumpff, L., Giljohann, K. M., Camac, J. S., Bau, S. S., Friend, T.,

189 References

Harrison, B., and Yue, B. (2013). The influence of abundance on detectability. Oikos 122, 717–726. doi:10.1111/j.1600-0706.2012.20781.x

McClintock, B. T., Bailey, L. L., Pollock, K. H., and Simons, T. R. (2010). Experimental investigation of observation error in anuran call surveys. Journal of Wildlife Management 74, 1882–1893. doi:10.2193/2009-321

McDonald, P. J., Griffiths, A. D., Nano, C. E. M., Dickman, C. R., Ward, S. J., and Luck, G. W. (2015). Landscape-scale factors determine occupancy of the critically endangered central rock-rat in arid Australia: The utility of camera trapping. Biological Conservation 191, 93–100. doi:10.1016/j.biocon.2015.06.027

McGregor, H. W., Cliff, H. B., and Kanowski, J. (2017). Habitat preference for fire scars by feral cats in Cape York Peninsula, Australia. Wildlife Research 43, 623–633. doi:10.1071/WR16058

McGregor, H. W., Legge, S., Jones, M. E., and Johnson, C. N. (2016). Extraterritorial hunting expeditions to intense fire scars by feral cats. Scientific Reports 6, 22559. doi:10.1038/srep22559

McIlroy, J. C. (1982). The sensitivity of Australian animals to 1080 poison IV. Native and introduced rodents. Wildlife Research 9, 505–517. doi:10.1071/wr9820505

McPherson, J. M., and Jetz, W. (2007). Effects of species’ ecology on the accuracy of distribution models. Ecography 30, 135–151. doi:10.1111/j.2006.0906- 7590.04823.x

McPherson, J. M., Jetz, W., and Rogers, D. J. (2004). The effects of species’ range sizes on the accuracy of distribution models: ecological phenomenon or statistical artefact? Journal of Applied Ecology 41, 811–823. doi:10.1111/j.0021-8901.2004.00943.x

Meek, P. D., Ballard, G.-A., and Fleming, P. J. S. (2015a). The pitfalls of wildlife camera trapping as a survey tool in Australia. Australian Mammalogy 37, 13– 22. doi:10.1071/AM14023

Meek, P. D., Ballard, G.-A., Vernes, K., and Fleming, P. J. S. (2015b). The history of wildlife camera trapping as a survey tool in Australia. Australian Mammalogy 37, 1–12. doi:10.1071/AM14021

Meek, P. D., and Vernes, K. (2015). Can camera trapping be used to accurately survey and monitor the Hastings River mouse (Pseudomys oralis)? Australian Mammalogy 38, 44–51. doi:10.1071/AM15016

Meek, P. D., Vernes, K., and Falzon, G. (2013). On the reliability of expert identification of small-medium sized mammals from camera trap photos. Wildlife Biology in Practice 9, 1–19. doi:10.2461/wbp.2013.9.4

190 References

Miller, D. A. W., Weir, L. A., McClintock, B. T., Grant, E. H. C., Bailey, L. L., and Simons, T. R. (2012). Experimental investigation of false positive errors in auditory species occurrence surveys. Ecological Applications 22, 1665–1674. doi:10.1890/11-2129.1

Molyneux, J., Pavey, C. R., James, A. I., and Carthew, S. M. (2018). The efficacy of monitoring techniques for detecting small mammals and reptiles in arid environments. Wildlife Research 44, 534–545. doi:10.1071/WR17017

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

Morand, S., Bordes, F., Chen, H.-W., Claude, J., Cosson, J.-F., Galan, M., Czirják, G. Á., Greenwood, A. D., Latinne, A., Michaux, J., and Ribas, A. (2015). Global parasite and Rattus rodent invasions: The consequences for rodent- borne diseases. Integrative Zoology 10, 409–423. doi:10.1111/1749- 4877.12143

Morris, K. D. (2002). The eradication of the black rat (Rattus rattus) on Barrow and adjacent islands off the north-west coast of Western Australia. In ‘Turning the Tide: The Eradication of Invasive Species : Proceedings of the International Conference on Eradication of Island Invasives’. (IUCN.)

Morris, K. D. (2000). The status and conservation of native rodents in Western Australia. Wildlife Research 27, 405–419.

Myroniuk, P. (1997). Conservation genetics of the New Holland mouse Pseudomys novaehollandiae, Waterhouse, 1843 (Rodentia: Muridae) : an analysis of mitochondrial DNA control region variation. Deakin University Geelong, Victoria. Available at: http://dro.deakin.edu.au/view/DU:30027747 [accessed 9 March 2015]

Nelson, J. L., Durkin, L. K., Cripps, J. K., Scroggie, M. P., Bryant, D. B., Macak, P. V., and Lumsden, L. F. (2017). Targeted surveys to improve Leadbeater’s Possum conservation. Arthur Rylah Institute for Environmental Research Technical Report Series No. 278. Department of Environment, Land, Water and Planning, Heidelberg, Victoria.

Nelson, J. L., Main, M. L., Chick, R. R., and Scroggie, M. P. (2010). The status of smoky mouse populations at historic sites in Victoria, and an assessment of two non-invasive survey techniques. Unpublished report to the Department of Environment, Water, Heritage and the Arts, and the Goulburn Broken Catchment Management Authority. Arthur Rylah Institute for Environmental Research, Heidelberg, Victoria.

Nelson, J., Menkhorst, P. W., Howard, K., Chick, R. R., and Lumsden, L. (2009). The status of smoky mouse populations at some historic sites in Victoria, and survey methods for their detection. Unpublished report number 2009/17. Arthur Rylah Institute for Environmental Research, Heidelberg, Victoria.

191 References

Newell, G. R. (1998). Characterization of vegetation in an Australian open forest community affected by cinnamon fungus (shape Phytophthora cinnamomi): implications for faunal habitat quality. Plant Ecology 137, 55–70. doi:10.1023/A:1008056432001

Nimmo, D. G., Kelly, L. T., Farnsworth, L. M., Watson, S. J., and Bennett, A. F. (2014). Why do some species have geographically varying responses to fire history? Ecography 37, 805–813. doi:10.1111/ecog.00684

Norton, T. W. (1987). The Ecology of Small Mammals in Northeastern Tasmania . II. Pseudomys novaehollandiae and the introduced Mus musculus. Wildlife Research 14, 435–441. doi:10.1071/WR9870435

O’Brien, T. G., and Kinnaird, M. F. (2008). A picture is worth a thousand words: the application of camera trapping to the study of birds. Bird Conservation International 18, S144–S162. doi:10.1017/S0959270908000348

O’Connell, A. F., Talancy, N. W., Bailey, L. L., Sauer, J. R., Cook, R., and Gilbert, A. T. (2006). Estimating site occupancy and detection probability parameters for meso- and large mammals in a coastal ecosystem. Journal of Wildlife Management 70, 1625–1633. doi:10.2193/0022- 541X(2006)70[1625:ESOADP]2.0.CO;2

OEH - Office of Environment and Heritage (2014). BioBanking Assessment Methodology 2014. Office of Environment and Heritage for the NSW Government, Sydney, NSW.

Olden, J. D., Hogan, Z. S., and Zanden, M. J. V. (2007). Small fish, big fish, red fish, blue fish: size-biased extinction risk of the world’s freshwater and marine fishes. Global Ecology and Biogeography 16, 694–701. doi:10.1111/j.1466- 8238.2007.00337.x

Otto, C. R. V., and Roloff, G. J. (2011). Using multiple methods to assess detection probabilities of forest-floor wildlife. The Journal of Wildlife Management 75, 423–431. doi:10.1002/jwmg.63

Paddle, R. (2002). ‘The Last Tasmanian Tiger: The History and Extinction of the Thylacine’. (Cambridge University Press.)

Parks Victoria (1998). ‘The Lakes National Park and the Gippsland Lakes Coastal Park management plan’. (Parks Victoria: East Melbourne.)

Pascoe, B. (2014). ‘Dark Emu: Aboriginal Australia and the Birth of Agriculture’. (Magabala Books: Broome, Western Australia.)

Paull, D. J., Claridge, A. W., and Cunningham, R. B. (2012). Effective detection methods for medium-sized ground-dwelling mammals: a comparison between infrared digital cameras and hair tunnels. Wildlife Research 39, 546–553. doi:10.1071/WR12034

192 References

Peterson, A. T. (2006). Uses and requirements of ecological niche models and related distributional models. doi:10.17161/bi.v3i0.29

Phillips, B. L., Chipperfield, J. D., and Kearney, M. R. (2008). The toad ahead: challenges of modelling the range and spread of an invasive species. Wildlife Research 35, 222. doi:10.1071/WR07101

Pimm, S. L., Jenkins, C. N., Abell, R., Brooks, T. M., Gittleman, J. L., Joppa, L. N., Raven, P. H., Roberts, C. M., and Sexton, J. O. (2014). The biodiversity of species and their rates of extinction, distribution, and protection. Science 344, 1246752–1246752. doi:10.1126/science.1246752

Pimm, S., Raven, P., Peterson, A., Şekercioğlu, Ç. H., and Ehrlich, P. R. (2006). Human impacts on the rates of recent, present, and future bird extinctions. Proceedings of the National Academy of Sciences 103, 10941–10946. doi:10.1073/pnas.0604181103

Pittock, J., and Finlayson, C. M. (2011). Australia’s Murray–Darling Basin: freshwater ecosystem conservation options in an era of climate change. Marine and Freshwater Research 62, 232–243. doi:10.1071/MF09319

Plummer, M. (2017). ‘JAGS’.

Plummer, M. (2016). ‘rjags’. (CRAN.)

Porfirio, L. L., Lefroy, T., Hugh, S., and Mackey, B. (2017). Monitoring the impacts of feral Horses on vegetation condition using remotely sensed fPAR: a case study in Australia’s alpine parks. Parks 23, 9–20.

Posamentier, H. G., and Recher, H. F. (1974). The status of Pseudomys novaehollandiae (The New Holland mouse). The Australian Zoologist 18, 66–71.

Prowse, T. A. A., Johnson, C. N., Lacy, R. C., Bradshaw, C. J. A., Pollak, J. P., Watts, M. J., and Brook, B. W. (2013). No need for disease: testing extinction hypotheses for the thylacine using multi-species metamodels. Journal of Animal Ecology 82, 355–364. doi:10.1111/1365-2656.12029

Prugh, L. R., and Golden, C. D. (2014). Does moonlight increase predation risk? Meta-analysis reveals divergent responses of nocturnal mammals to lunar cycles Ed S. Boutin. Journal of Animal Ecology 83, 504–514. doi:10.1111/1365-2656.12148

Pye, T. (1991). The New Holland mouse (Pseudomys novaehollandiae) (Rodentia : Muridae) in Tasmania: a field study. Wildlife Research 18, 521–531. doi:10.1071/WR9910521

QGIS Development Team (2014). ‘QGIS Geographic Information System’. (Open Source Geospatial Foundation Project.) Available at: http://qgis.osgeo.org

193 References

Quin, B. R. (1996). New Holland Mouse Pseudomys novaehollandiae (Rodentia: Muridae) in South Gippsland, southern Victoria part one - distribution and status. The Victorian Naturalist 113, 236–246.

Quin, B. R., and Williamson, R. C. (1996). New Holland Mouse Pseudomys novaehollandiae (Rodentia: Muridae) in South Gippsland, southern Victoria part two - conservation and management. The Victorian Naturalist 113, 281– 288.

R Core Team (2017). ‘R: A language and environment for statistical computing’. (R Foundation for Statistical Computing: Vienna, Austria.) Available at: http://www.R-project.org/

Ramesh, V., Gopalakrishna, T., Barve, S., and Melnick, D. J. (2017). IUCN greatly underestimates threat levels of endemic birds in the Western Ghats. Biological Conservation 210, 205–221. doi:10.1016/j.biocon.2017.03.019

Randin, C. F., Dirnböck, T., Dullinger, S., Zimmermann, N. E., Zappa, M., and Guisan, A. (2006). Are niche-based species distribution models transferable in space? Journal of Biogeography 33, 1689–1703. doi:10.1111/j.1365- 2699.2006.01466.x

Reed, J. M. (1996). Using statistical probability to increase confidence of inferring species extinction. Conservation Biology 10, 1283–1285.

Régnier, C., Achaz, G., Lambert, A., Cowie, R. H., Bouchet, P., and Fontaine, B. (2015). Mass extinction in poorly known taxa. Proceedings of the National Academy of Sciences 112, 7761–7766. doi:10.1073/pnas.1502350112

Reside, J., and Hooper, D. (1999). Surveys for the New Holland Mouse (Pseudomys novaehollandiae) in Gippsland. Wildlife Unlimited Pty Ltd, Bairnsdale, Victoria.

Roberts, N. J. (2011). Investigation into survey techniques of large mammals: surveyor competence and camera-trapping vs. transect-sampling. Bioscience Horizons 4, 40–49. doi:10.1093/biohorizons/hzr006

Robinson, A. C., Kemper, C. M., Medlin, G. C., and Watts, C. H. S. (2000). The rodents of South Australia. Wildlife Research 27, 379–404. doi:10.1071/wr97044

Rout, T. M., Heinze, D., and McCarthy, M. A. (2010). Optimal allocation of conservation resources to species that may be extinct. Conservation Biology 24, 1111–1118. doi:10.1111/j.1523-1739.2010.01461.x

Rovero, F., Zimmermann, F., Berzi, D., and 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. doi:10.4404/hystrix-24.2-8789

194 References

Royle, J. A., and Nichols, J. D. (2003). Estimating abundance from repeated presence-absence data or point counts. Ecology 84, 777–790. doi:10.1890/0012-9658(2003)084[0777:EAFRPA]2.0.CO;2

Russell, R., Duchaine, B., and Nakayama, K. (2009). Super-recognizers: People with extraordinary face recognition ability. Psychonomic Bulletin & Review 16, 252–257. doi:10.3758/PBR.16.2.252

Santika, T., McAlpine, C. A., Lunney, D., Wilson, K. A., and Rhodes, J. R. (2014). Modelling species distributional shifts across broad spatial extents by linking dynamic occupancy models with public-based surveys. Diversity and Distributions 20, 786–796. doi:10.1111/ddi.12189

Scheele, B. C., and Gillespie, G. R. (2018). The extent and adequacy of monitoring for Australian threatened frog species. In ‘Monitoring Threatened Species and Ecological Communities’. (Eds S. Legge, N. Robinson, D. B. Lindenmayer, B. C. Scheele, D. M. Southwell, and B. A. Wintle.) pp. 57–68. (CSIRO Publishing: Clayton South, Victoria.)

Scheele, B. C., Guarino, F., Osborne, W., Hunter, D. A., Skerratt, L. F., and Driscoll, D. A. (2014). Decline and re-expansion of an amphibian with high prevalence of chytrid fungus. Biological Conservation 170, 86–91. doi:10.1016/j.biocon.2013.12.034

Seebeck, J. H., and Beste, H. J. (1970). First record of the New Holland Mouse (Pseudomys novaehollandiae (Waterhouse, 1853)) in Victoria. The Victorian Naturalist 87, 280–287.

Seebeck, J. H., Menkhorst, P. W., Wilson, B. A., and Lowe, K. W. (2003). New Holland Mouse (Pseudomys novaehollandiae) - Flora and Fauna Guarantee Action Statement no. 74. No. 74. Department of Sustainability and Environment, East Melbourne.

Seebeck, J., and Menkhorst, P. (2000). Status and conservation of the rodents of Victoria. Wildlife Research 27, 357–369.

Seigel, R. A., and Dodd, C. K. (2002). Translocations of amphibians: proven management method or experimental technique? Conservation Biology 16, 552–554. doi:10.1046/j.1523-1739.2002.01275.x

Sharp, T., and Saunders, G. (2004). Ground baiting of rabbits with Pindone. No. RAB004. NSW Department of Primary Industries.

Short, J., Kinnear, J. E., and Robley, A. (2002). Surplus killing by introduced predators in Australia—evidence for ineffective anti-predator adaptations in native prey species? Biological Conservation 103, 283–301. doi:10.1016/S0006-3207(01)00139-2

Short, J., and Smith, A. (1994). Mammal decline and recovery in Australia. Journal of Mammalogy 75, 288–297. doi:10.2307/1382547

195 References

Shumway, N., Lunney, D., Seabrook, L., and McAlpine, C. (2015). Saving our national icon: An ecological analysis of the 2011 Australian Senate inquiry into status of the koala. Environmental Science & Policy 54, 297–303. doi:10.1016/j.envsci.2015.07.024

Sikes, R. S., and Gannon, W. L. (2011). Guidelines of the American Society of Mammalogists for the use of wild mammals in research. Journal of Mammalogy 92, 235–253. doi:10.1644/10-MAMM-F-355.1

Sinclair, S. J., White, M. D., and Newell, G. R. (2010). How useful are species distribution models for managing biodiversity under future climates? Ecology and Society 15. Available at: https://www.jstor.org/stable/26268111 [accessed 10 September 2018]

Skerratt, L. F., Berger, L., Clemann, N., Hunter, D. A., Marantelli, G., Newell, D. A., Philips, A., McFadden, M., Hines, H. B., Scheele, B. C., Brannelly, L. A., Speare, R., Versteegen, S., Cashins, S. D., and West, M. (2016). Priorities for management of chytridiomycosis in Australia: saving frogs from extinction. Wildlife Research 43, 105–120. doi:10.1071/WR15071

Skerratt, L. F., Berger, L., Speare, R., Cashins, S., McDonald, K. R., Phillott, A. D., Hines, H. B., and Kenyon, N. (2007). Spread of Chytridiomycosis has caused the rapid global decline and extinction of frogs. EcoHealth 4, 125–134. doi:10.1007/s10393-007-0093-5

Slip, D. J., and Shine, R. (1988). Feeding habits of the Diamond Python, Morelia s. spilota: ambush predation by a Boid snake. Journal of Herpetology 22, 323. doi:10.2307/1564156

Smith, A. B., Santos, M. J., Koo, M. S., Rowe, K. M. C., Rowe, K. C., Patton, J. L., Perrine, J. D., Beissinger, S. R., and Moritz, C. (2013). Evaluation of species distribution models by resampling of sites surveyed a century ago by Joseph Grinnell. Ecography 36, 1017–1031. doi:10.1111/j.1600-0587.2013.00107.x

Smith, A. L., Bull, C. M., and Driscoll, D. A. (2012). Post-fire succession affects abundance and survival but not detectability in a knob-tailed gecko. Biological Conservation 145, 139–147. doi:10.1016/j.biocon.2011.10.023

Smith, L., and Bluff, L. A. (2018). New Holland Mouse pre and post burn monitoring program. DELWP Internal report, Gippsland Risk and Evaluation Team. Department of Environment, Land, Water and Planning, Bairnsdale, Victoria.

Somodi, I., Lepesi, N., and Botta‐Dukát, Z. (2017). Prevalence dependence in model goodness measures with special emphasis on true skill statistics. Ecology and Evolution 7, 863–872. doi:10.1002/ece3.2654

Spash, C. L. (2015). Bulldozing biodiversity: The economics of offsets and trading- in Nature. Biological Conservation 192, 541–551. doi:10.1016/j.biocon.2015.07.037

196 References

Steadman, D. W. (1989). Extinction of birds in Eastern polynesia: A review of the record, and comparisons with other Pacific Island groups. Journal of Archaeological Science 16, 177–205. doi:10.1016/0305-4403(89)90065-4

Stevens, W. (2001). Declining biodiversity and unsustainable agricultural production-common cause, common solution? text. Department of the Parliamentary Library, Canberra. Available at: https://www.aph.gov.au/About_Parliament/Parliamentary_Departments/Parli amentary_Library/pubs/rp/rp0102/02RP02 [accessed 23 December 2018]

Swan, M., Christie, F., Sitters, H., York, A., and Di Stefano, J. (2015). Predicting faunal fire responses in heterogeneous landscapes: the role of habitat structure. Ecological Applications 25, 2293–2305. doi:10.1890/14-1533.1

Swan, M., Di Stefano, J., and Christie, F. (2014a). Comparing the effectiveness of two types of camera trap for surveying ground-dwelling mammals. In ‘Camera Trapping: Wildlife Management and Research’. pp. 123–130. (Csiro Publishing.)

Swan, M., Di Stefano, J., Christie, F., Steel, E., and York, A. (2014b). Detecting mammals in heterogeneous landscapes: implications for biodiversity monitoring and management. Biodiversity and Conservation 23, 343–355. doi:10.1007/s10531-013-0604-3

Tallmon, D. A., Luikart, G., and Waples, R. S. (2004). The alluring simplicity and complex reality of genetic rescue. Trends in Ecology & Evolution 19, 489– 496. doi:10.1016/j.tree.2004.07.003

Tasker, E., and Dickman, C. (2001). A review of Elliott trapping methods for small mammals in Australia. Australian Mammalogy 23, 77–87.

Tasker, E. M., and Dickman, C. R. (2004). Small mammal community composition in relation to cattle grazing and associated burning in eucalypt forests of the Northern Tablelands of New South Wales. In ‘Conservation of Australia’s Forest Fauna’. (Ed D. Lunney.) pp. 721–740. (Royal Zoological Society of New South Wales: Mosman, NSW.)

Taylor, B. D., Goldingay, R. L., and Lindsay, J. M. (2014). Horizontal or vertical? Camera trap orientations and recording modes for detecting potoroos, bandicoots and pademelons. Australian Mammalogy 36, 60. doi:10.1071/AM13012

Taylor, B. L., and Gerrodette, T. (1993). The uses of statistical power in conservation biology: the Vaquita and Northern Spotted Owl. Conservation Biology 7, 489–500. doi:10.1046/j.1523-1739.1993.07030489.x

Thompson, R. C. A., Lymbery, A. J., and Smith, A. (2010). Parasites, emerging disease and wildlife conservation. International Journal for Parasitology 40, 1163–1170. doi:10.1016/j.ijpara.2010.04.009

197 References

Thompson, W. L. (2002). Towards reliable bird surveys: accounting for individuals present but not detected. The Auk 119, 18. doi:10.1642/0004- 8038(2002)119[0018:TRBSAF]2.0.CO;2

Tolsma, A., and Shannon, J. (2018). Assessing the Impacts of Feral Horses on the Bogong High Plains, Victoria. Unpublished client report for Parks Victoria. Arthur Rylah Institute for Environmental Research, Department of Environment, Land, Water and Planning, Heidelburg, Victoria.

Tyre, A. J., Tenhumberg, B., Field, S. A., and Possingham, H. P. (2003). Improving precision and reducing bias in biological surveys: estimating false-negative error rates. Ecological Applications 13, 1790–1801.

Van Dyck, S. M., and Lawrie, B. (1997). The New Holland Mouse Pseudomys novaehollandiae (Rodentia: Muridae), an addition to the mammal fauna of Queensland. Memoirs of the Queensland Museum 42, 367–376.

Vance, A. (2001). Population dynamics and habitat selection of the New Holland mouse, Pseudomys novaehollandiae, in Gippsland Lakes Coastal Park, Loch Sport. Honours Thesis, University of Melbourne Melbourne, Victoria.

Vickery, W. L., and Bider, J. R. (1981). The influence of weather on rodent activity. Journal of Mammalogy 62, 140–145. doi:10.2307/1380484

Viggers, K. L., Lindenmayer, D. B., and Spratt, D. M. (1993). The Importance of Disease in Reintroduction Programmes. Wildlife Research 20, 687–698. doi:10.1071/wr9930687

Villette, P., Krebs, C. J., Jung, T. S., and Boonstra, R. (2016). Can camera trapping provide accurate estimates of small mammal (Myodes rutilus and Peromyscus maniculatus) density in the boreal forest? Journal of Mammalogy 97, 32–40. doi:10.1093/jmammal/gyv150

Waddle, J. H., Thigpen, T. F., and Glorioso, B. M. (2009). Efficacy of automatic vocalisation recognition software for anuran monitoring. Herpetological Conservation and Biology 4, 384–388.

Walker, K. F., and Thoms, M. C. (1993). Environmental effects of flow regulation on the lower river Murray, Australia. Regulated Rivers: Research & Management 8, 103–119. doi:10.1002/rrr.3450080114

Waller, N. L., Gynther, I. C., Freeman, A. B., Lavery, T. H., and Leung, L. K.-P. (2017). The Bramble Cay melomys Melomys rubicola (Rodentia : Muridae): a first mammalian extinction caused by human-induced climate change? Wildlife Research 44, 9. doi:10.1071/WR16157

Watts, C. H. S., and Braithwaite, R. W. (1978). The diet of Rattus Lutreolus and five other rodents in Southern Victoria. Wildlife Research 5, 47. doi:10.1071/WR9780047

198 References

Wayne, A. F., Maxwell, M. A., Ward, C. G., Vellios, C. V., Ward, B. G., Liddelow, G. L., Wilson, I., Wayne, J. C., and Williams, M. R. (2013). Importance of getting the numbers right: quantifying the rapid and substantial decline of an abundant marsupial, Bettongia penicillata. Wildlife Research 40, 169. doi:10.1071/WR12115

Welbourne, D. J., MacGregor, C., Paull, D., and Lindenmayer, D. B. (2015). The effectiveness and cost of camera traps for surveying small reptiles and critical weight range mammals: a comparison with labour-intensive complementary methods. Wildlife Research 42, 414–425. doi:10.1071/WR15054

Whiteley, A. R., Fitzpatrick, S. W., Funk, W. C., and Tallmon, D. A. (2015). Genetic rescue to the rescue. Trends in Ecology & Evolution 30, 42–49. doi:10.1016/j.tree.2014.10.009

Wiens, J. J. (2016). Climate-related local extinctions are already widespread among plant and animal species Ed A. Barnosky. PLOS Biology 14, e2001104. doi:10.1371/journal.pbio.2001104

Williamson, G. J., Murphy, B. P., and Bowman, D. M. J. S. (2014). Cattle grazing does not reduce fire severity in eucalypt forests and woodlands of the Australian Alps. Austral Ecology 39, 462–468. doi:10.1111/aec.12104

Willson, J. D., Winne, C. T., and Todd, B. D. (2011). Ecological and methodological factors affecting detectability and population estimation in elusive species. The Journal of Wildlife Management 75, 36–45. doi:10.1002/jwmg.15

Wilson, B. A. (1996). The distribution and status of the New Holland Mouse, Pseudomys novaehollandiae (Waterhouse 1843) in Victoria. Australian Mammalogy 19, 31–46.

Wilson, B. A. (1991). The ecology of Pseudomys novaehollandiae (Waterhouse, 1843) in the Eastern Otway Ranges, Victoria. Wildlife Research 18, 233–247. doi:10.1071/WR9910233

Wilson, B. A., and Bradtke, E. (1999). The diet of the New Holland mouse, Pseudomys novaehollandiae (Waterhouse) in Victoria. Wildlife Research 26, 439–451. doi:10.1071/WR97062

Wilson, B. A., and Laidlaw, W. S. (2003). Habitat characteristics for New Holland mouse Pseudomys novaehollandiae in Victoria. Australian Mammalogy 25, 1–11. doi:10.1071/AM03001

Wilson, B. A., Lock, M., and Garkaklis, M. J. (2017). Long-term fluctuations in distribution and populations of a threatened rodent (Pseudomys novaehollandiae) in coastal woodlands of the Otway Ranges, Victoria: a regional decline or extinction? Australian Mammalogy 40, 281–293. doi:10.1071/AM17036

199 References

Wilson, B. A., White, N. M., Hanley, A., and Tidey, D. L. (2005). Population fluctuations of the New Holland mouse Pseudomys novaehollandiae at Wilson’s Promontory National Park, Victoria. Australian Mammalogy 27, 49–60. doi:10.1071/AM05049

Wilson, C. D., Roberts, D., and Reid, N. (2011). Applying species distribution modelling to identify areas of high conservation value for endangered species: A case study using Margaritifera margaritifera (L.). Biological Conservation 144, 821–829. doi:10.1016/j.biocon.2010.11.014

Wintle, B. A., Kavanagh, R. P., McCarthy, M. A., and Burgman, M. A. (2005). Estimating and dealing with detectability in occupancy surveys for forest owls and arboreal marsupials. Journal of Wildlife Management 69, 905–917. doi:10.2193/0022-541X(2005)069[0905:EADWDI]2.0.CO;2

Wintle, B. A., Kujala, H., Whitehead, A., Cameron, A., Veloz, S., Kukkala, A., Moilanen, A., Gordon, A., Lentini, P. E., Cadenhead, N. C. R., and Bekessy, S. A. (2018). Global synthesis of conservation studies reveals the importance of small habitat patches for biodiversity. Proceedings of the National Academy of Sciences, 201813051. doi:10.1073/pnas.1813051115

Wintle, B. A., Walshe, T. V., Parris, K. M., and McCarthy, M. A. (2012). Designing occupancy surveys and interpreting non-detection when observations are imperfect: Interpreting non-detections in occupancy surveys. Diversity and Distributions 18, 417–424. doi:10.1111/j.1472-4642.2011.00874.x

Woinarski, J., Burbidge, A., and Harrison, P. (2014). ‘Action Plan for Australian Mammals 2012’. (CSIRO Publishing: Collingwood, Victoria.) Available at: http://www.publish.csiro.au.ezp.lib.unimelb.edu.au/pid/7010.htm [accessed 21 July 2015]

Woinarski, J. C. Z., Armstrong, M., Brennan, K., Fisher, A., Griffiths, A. D., Hill, B., Milne, D. J., Palmer, C., Ward, S., Watson, M., Winderlich, S., and Young, S. (2010). Monitoring indicates rapid and severe decline of native small mammals in Kakadu National Park, northern Australia. Wildlife Research 37, 116–126. doi:10.1071/WR09125

Woinarski, J. C. Z., Burbidge, A. A., and Harrison, P. L. (2015). Ongoing unraveling of a continental fauna: Decline and extinction of Australian mammals since European settlement. Proceedings of the National Academy of Sciences 112, 4531–4540. doi:10.1073/pnas.1417301112

Woinarski, J. C. Z., Burbidge, A. A., and Harrison, P. L. (2018). The extent and adequacy of monitoring for Australian threatened mammal species. In ‘Monitoring Threatened Species and Ecological Communities’. (Eds S. Legge, N. Robinson, D. B. Lindenmayer, B. C. Scheele, D. M. Southwell, and B. A. Wintle.) pp. 21–42. (CSIRO Publishing: Clayton South, Victoria.)

200

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Burns, Phoebe Ann

Title: Testing the decline of the threatened New Holland Mouse (Pseudomys novaehollandiae)

Date: 2019

Persistent Link: http://hdl.handle.net/11343/222131

File Description: Testing the decline of the threatened New Holland Mouse (Pseudomys novaehollandiae)

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.