Bioacoustic Monitoring of Breeding Behaviour in the Endangered

Home , Cacatua, Eucalyptus cladocalyx, Palm cockatoo

Bioacoustic monitoring of breeding behaviour in the endangered Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne

Daniella Teixeira

BMarSt (Hons), MConsBiol

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2020

School of Biological Sciences

Centre for Biodiversity and Conservation Science Abstract

This thesis investigates the use of bioacoustic technology for monitoring breeding behaviours in two endangered subspecies of black-cockatoo, the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne. Over three years, I collected breeding behavioural data, including animal sound data and video footage, at nests of glossy black-cockatoos on Kangaroo Island in South Australia and red-tailed black-cockatoos in the Casterton region of Victoria. This thesis comprises an introduction (Chapter 1) and synthesis and conclusion (Chapter 6) and four research chapters (Chapters 2 to 5).

In Chapter 2, I provide a literature review of the potential use of bioacoustics in monitoring animal behaviour for conservation. I argue that bioacoustic studies should better incorporate knowledge of species’ vocal behaviours, to improve the resolution of context- specific data. Animal behaviour is often relevant to conservation, and bioacoustics could greatly improve our ability to acquire behavioural data. A necessary first step is to understand the vocal behaviours of a species to the extent required for conservation monitoring. To this end, in Chapter 3, I describe the nest-associated vocal behaviours of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Over two breeding seasons, I recorded daily vocal activity at nests using autonomous sound recorders. Combined with behavioural observations, including video footage, I identified vocalisations from six behavioural contexts: birds in flight, while perched, during begging (adult females), during courtship displays (adult males), when entering or sitting near to the nest hollow entrance (adult females), and from nestlings. In total, I describe 14 putative call types for the glossy black-cockatoo and 11 putative call types for the red-tailed black-cockatoo. For both subspecies, the female nest call and nestling calls are the most conspicuous vocal indicators of active nesting.

In Chapter 4, I investigate the utility of bioacoustics for monitoring nest outcomes in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. First, I provide the vocal signal of a successful fledging event for both subspecies. This unique behaviour is clearly indicated in spectrograms and provides a bioacoustic signal with which to confirm nest outcome. I then investigated automated methods to identify nest success or failure using an open-source call recogniser implemented in the monitoR package in R software. Constructed from templates of nestling calls, the recogniser was tested on 3 x 3-

ii hours of sound data from early, mid and late stages of the recording period for each nest monitored (n = 23 for the glossy black-cockatoo; n = 21 for the red-tailed black-cockatoo). Daily nest activity was correctly assigned in 61.9% of survey days analysed for the red-tailed black-cockatoo, and 68.1% of survey days for the glossy black-cockatoo. Importantly, the recogniser successfully detected the fledging event in almost all cases (five out of six fledging events in the glossy black-cockatoo, and two out of three fledging events in the red- tailed black-cockatoo). Precision of individual detections was moderate, with many false positives. Manual verification of outputs was required, making this a semi-automated method.

In Chapter 5, I examine the ontogeny of vocalisations in nestlings of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. To determine whether nests can be aged from sound recordings, I examined changes in acoustic structure and daily call rate during nest development. I determined that nestlings vocalised from about 4 weeks of age, but calls were soft and infrequent until about 6 weeks. Daily call rate increased over time, especially in the final week of nesting. Peak amplitude and low frequency of nestling calls increased significantly with development. Call duration increased significantly for the glossy black-cockatoo, but not for the red-tailed black-cockatoo. Average entropy declined significantly for both subspecies. Aggregate entropy declined significantly for the red-tailed black-cockatoo but not the glossy black-cockatoo. Together, these changes in call rate and acoustic structure provide a useful way to broadly categorise nest age from sound recordings and thereby improve knowledge on nestling survival. This knowledge may be useful in future studies examining the influence of habitat variables, such as food availability, on nest development and nestling survival across landscapes.

This thesis presents novel bioacoustic methods for monitoring breeding in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. I demonstrate that these subspecies exhibit diverse repertoires at nests. I describe the vocal signal of fledging, which is a direct measure of breeding success, and provide an open-source call recogniser to aid sound data processing. Finally, I describe the ontogenetic development of nestling vocalisations. I conclude that bioacoustics has potential to greatly improve nest monitoring of the endangered Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, to the benefit of conservation.

iii

Declaration by author

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

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

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

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co-authors for any jointly authored works included in the thesis.

Publications included in this thesis

Teixeira D., Maron M. & van Rensburg B. J. (2019) Bioacoustic monitoring of animal vocal behavior for conservation. Conservation Science and Practice 1, e72. DOI: 10.1111/csp2.7

Teixeira, D., Hill, R., Barth, M., Maron, M. and van Rensburg, B. J. (2020) Nest-associated vocal behaviours of the south-eastern red-tailed black-cockatoo, Caylptorhynchus banksii graptogyne, and the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus. Austral Ecology DOI: 10.1111/aec.12921

Submitted manuscripts included in this thesis

No other manuscripts have been submitted.

Other publications during candidature

Conference abstracts:

Teixeira, D., Maron, M. and van Rensburg, B. Monitoring breeding of threatened black- cockatoo using bioacoustics. Ecoacoustics for Conservation Workshop, Brisbane, February 2020.

Teixeira, D., Maron, M. and van Rensburg, B. Novel technologies for conservation: bioacoustics and black-cockatoos, Australasian Ornithological Conference, Darwin, July 2019.

Teixeira, D., Maron, M. and van Rensburg, B. Fledge or fail: saving the south-eastern red- tailed black-cockatoo, Victorian Biodiversity Conference, Melbourne, February 2019.

Teixeira, D., Maron, M. and van Rensburg, B. Conservation behavior and bioacoustics, Australasian Society for the Study of Animal Behaviour Conference, Brisbane, July 2018.

Teixeira, D., Roe, P., Maron, M. and van Rensburg, B. Bioacoustics, behaviour and black- cockatoos: a new approach to conservation, International Ecoacoustics Congress, Brisbane, June 2018.

Teixeira, D., Roe, P., Maron, M. and van Rensburg, B. Eavesdropping on black-cockatoo nests: Bioacoustic monitoring of breeding in the endangered south-eastern red-tailed black-

v cockatoo and the Kangaroo Island glossy black-cockatoo, Australasian Ornithological Conference, Geelong, November 2017.

Teixeira, D. Bioacoustics for black-cockatoo conservation, Australasian Chapter for Ecoacoustics Workshop, Brisbane, February 2017.

Contributions by others to the thesis

Berndt van Rensburg and Martine Maron contributed substantially to the research presented in this thesis. Both helped in establishing the project’s aims, design and implementation, in addition to providing feedback on drafts of the thesis chapters and manuscripts. In the field, I received significant help from Richard Hill and Mike Barth. This included locating nesting birds and maintaining field equipment in my absence. Both contributed substantially to discussions about the conservation relevance of my research. Tim Burnard, Evan Roberts, Karleah Berris and Torren Welz also contributed to fieldwork. Paul Roe, Anthony Truskinger, Michael Towsey and Phil Eichinski helped with data storage and contributed to discussions regarding automated processing of sound data. Simon Linke contributed to coding the call recogniser for Chapter 4. Simone Blomberg provided statistical guidance.

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

No works submitted towards another degree have been included in this thesis.

Research Involving Human or Animal Subjects

This work was conducted under animal ethics approval number SBS/076/17/VIC and SBS/DEWNR/219/17 issued by The University of Queensland Animal Ethics Committee.

vii

Acknowledgements

I am deeply grateful for the help I’ve received from so many people during this research. First and foremost, thank you to my academic supervisors, Berndt van Rensburg and Martine Maron. Your support, encouragement, and knowledge has been fundamental to everything I’ve done during this PhD. I am so thankful to have had the opportunity to be mentored by you both. Thank you for inspiring my passion for threatened species conservation.

To Richard Hill and Mike Barth, thank you endlessly. This research would not have been possible without your support in the field. Thank you for ensuring everything ran smoothly when I couldn’t be there. Most importantly, thank you for sharing your knowledge and wisdom on black-cockatoos. Thank you for taking me under your wing, so to speak, and teaching me all you know about the nesting behaviours of the glossy and red-tailed black- cockatoos.

For help in Victoria, I also thank Tim Burnard. From helping me find nests to welcoming me into your home, there is so much to thank you for. Thank you for always encouraging me and my ideas, including the National Black-Cockatoo Forum. I couldn’t have done it without your help. Thank you also to Evan Roberts for helping me find nests. For help on Kangaroo Island, thank you also to Karleah Berris and Torren Welz. Thank you for supporting my research along the way, and giving me so many opportunities to get involved in the glossy recovery efforts. Thank you also to Gay Crowley. I’m so glad our paths crossed and that I’ve been able to glimpse of your incredible knowledge of Kangaroo Island and the glossy black-cockatoos.

Thank you to everyone in the NESP Threatened Species Recovery Hub. I’m so thankful to be part of such a diverse, knowledgeable and passionate conservation community. Australia’s threatened species are in good hands. Thank you to the Hub for financial support during this research. For helping me communicate my research to a wide range of audiences, thank you to Jaana Dielenberg and Nicolas Rakotopare. For help in managing travel and finance, thank you to Heather Christensen.

To my lab mates at UQ, thanks for being so fun and supportive. Thank you to my honours students, Nick Delzoppo, Fiona Hoegh-Guldberg and Ashleigh Gonzalez. I’m so glad to have worked with you on other black-cockatoo and bioacoustics projects. To Debra Stark, Alannah Filer and Lenn Isidore, thanks for making the lab a great place to be.

viii

To everyone at the QUT Ecosounds Lab, thank you. I couldn’t have found a better place to learn about technical side of bioacoustics. Paul Roe and Mike Towsey, thank you for sharing your knowledge on acoustics and recogniser development. For help with storing and managing the huge volumes of sound data I acquired, thank you to Anthony Truskinger and Phil Eichinski. You made the data much more manageable for me.

For help with recogniser development, huge thanks to Simon Linke. I really appreciate your efforts. Thank you to the team at Frontier Labs for help with sound recorders. For statistical help, thank you to Simone Blomberg.

Thank you to everyone in the Glossy Black Conservancy. Thank you for your financial support and for welcoming me into the Conservancy as a member. I’m so glad to be part of a glossy black-cockatoo community that’s close to home.

For accommodation during fieldwork, thank you to Donna Higgins and Tim Burnard, Lynne Nadebaum and Jim Edwards, Robyn Molsher, Gay Crowley and Dave Gillieson. Thank you for welcoming me into your homes and the communities in Casterton and Kangaroo Island.

Thank you to the landowners who allowed me to work on their properties; the Fosters, the Edgars, the Halls, the McKennys, the Zipples, and the Australian Bluegum Plantations.

During this work, I’ve met many people who care deeply about black-cockatoo conservation. To all of you, the whole black-cockatoo community, thank you. Thank you to everyone who has contacted me, connected with me on the Black-Cockatoo Project pages, or contributed in some way to black-cockatoo conservation. For your support during the bushfires on Kangaroo Island, when so much was lost, my deepest thanks. Thank you for your donations, your concern and your words of support. I appreciate it all, and the glossies are better off for it.

To my family, the Teixeiras, the Coopers and beyond, thank you for believing in me and supporting me during the past four years, and well before that. Thank you for always encouraging that spark of curiosity in me.

To Josh, thank you for everything. I wouldn’t be here without you. Thank you for your endless support, encouragement and love. This has been as much your journey as mine, and I couldn’t be happier about that. Thank you for fostering in me a deep love for the world. I hope to experience it all with you.

Financial support

This work was supported by an Australian Postgraduate Award, the National Environmental Science Programme’s Threatened Species Recovery Hub and the Glossy Black Conservancy.

Keywords black-cockatoos, bioacoustics, monitoring, nesting, vocal behaviour, call recogniser, ontogeny, conservation

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 050206 Environmental Monitoring, 70%

ANZSRC code: 050202 Conservation and Biodiversity, 30%

Fields of Research (FoR) Classification

FoR code: 0502, Environmental Science and Management, 100%

Table of Contents

Abstract ...... ii Declaration by author ...... iv Publications included in this thesis ...... v Submitted manuscripts included in this thesis ...... v Other publications during candidature ...... v Contributions by others to the thesis ...... vi Statement of parts of the thesis submitted to qualify for the award of another degree ...... vii Research Involving Human or Animal Subjects ...... vii Acknowledgements ...... viii Financial support ...... x Keywords ...... x Australian and New Zealand Standard Research Classifications (ANZSRC) ...... xi Fields of Research (FoR) Classification ...... xi Table of Contents ...... ii List of Figures ...... v List of Tables ...... viii List of Abbreviations ...... ix Acknowledgement of Traditional Owners ...... x Note regarding Kangaroo Island ...... xi Chapter 1 | Introduction ...... 1 Bioacoustic monitoring for conservation ______1 Study species ______3 Bioacoustic monitoring of vocal behaviour in glossy and red-tailed black-cockatoos _____ 5 Expanding bioacoustics to other subspecies ______7 Thesis aims and structure ______9 Terminology ______14 Chapter 2 | Bioacoustic monitoring of animal vocal behaviour for conservation ...... 18 Abstract ______18 Introduction ______18 Vocal behaviour and conservation ______20 Reproduction and recruitment ______20 Alarm and defence ______24 Sociality and vocal complexity ______26 Challenges and considerations for bioacoustic monitoring programs ______27 Conclusion ______31

Acknowledgements ______32 Chapter 3 | Nest-associated vocal behaviours of the south-eastern red-tailed black- cockatoo, Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black- cockatoo, C. lathami halmaturinus ...... 39 Abstract ______39 Introduction ______39 Methods ______42 Study sites ______42 Acoustic data collection ______43 Behavioural classification ______44 Quantitative structure of vocalisations ______45 Statistical analysis ______46 Results ______47 South-eastern red-tailed black-cockatoo ______47 Kangaroo Island glossy black-cockatoo ______48 Discussion ______49 Conclusion ______55 Acknowledgments ______55 Chapter 4 | Fledge or fail: Nest monitoring of the endangered black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus, using bioacoustics and open-source call recognition ...... 69 Abstract ______69 Introduction ______70 Methods ______73 Sound data collection______73 Fledging vocalisations ______74 Recogniser development ______74 Recogniser performance ______77 Results ______78 Fledging vocalisations ______78 Recogniser performance ______79 Discussion ______80 Acknowledgments ______84 Chapter 5 | Vocal ontogeny of nestling black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus ...... 88 Abstract ______88 Introduction ______88

iii

Methods ______91 Study sites ______91 Sound data collection______91 Spectrographic measurements ______92 Statistical analysis ______93 Results ______93 Vocal ontogeny ______94 Changes in daily call rate ______94 Changes in acoustic structure ______94 Discussion ______95 Acknowledgements ______100 Chapter 6 | Thesis synthesis and conclusion ...... 104 Thesis overview______104 The future of bioacoustic monitoring for conservation ______108 Conservation significance of bioacoustics for nest monitoring of black-cockatoos _____ 110 Recommended methods for bioacoustic nest monitoring ______111 Limitations ______113 Concluding remarks ______115 References ...... 116 Appendices ...... 140 Appendix 1 | Ethics approvals for Victoria and South Australia. ______140 Appendix 2 | Supplementary material for Chapter 2 ______144 Appendix 3 | Supplementary material for Chapter 3 ______146 Appendix 4 | Supplementary material for Chapter 4 ______155

List of Figures

Fig. 1.1: Distribution map of the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne (red) and the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus (green)...... 12

Fig. 1.2: Structure of thesis chapters...... 13

Fig. 1.3: Spectrograms of a perch call of a red-tailed black-cockatoo, Calyptorhynchus banksii (a), and a flight call of a yellow-tailed black-cockatoo, C. funereus (b), demonstrating acoustic features: phrases (P1, P2, P3), pulses (P), harmonics (H), frequency modulation (FM) and chaos (C). X-axis denotes time in seconds. Y-axis denotes frequency in kilohertz (kHz). Spectrograms created in Raven Pro 1.5 (Cornell Lab of Ornithology)...... 16

Fig. 3.1: Vocalisations of the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, at nests. (a) Flight call; (b) Take-off call; (c) Female begging bout, loud, clear harmonics with chaos; (d) Begging bout, soft, high-pitched, showing nonlinearity; (e) Female nest call; (f) Male display call (courtship); (g) Nestling call, subtype 1; (h) Nestling call, subtype 2; (i) Nestling call, subtype 3; (j) Male perch call, subtype 1; (k) Male perch call, subtype 2 (l) Male perch call, subtype 3. Spectrogram parametres: Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap...... 62

Fig. 3.2: Vocalisations of the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, at nests. (a) Flight call; (b) Female begging call, clear harmonics; (c) Begging bout, showing nonlinearity; (d) Male display call (courtship); (e) Female nest call; (f) Nestling call, subtype 1; (g) Nestling call, subtype 2; (h) Nestling call, subtype 3; (i) Male perch call, subtype 1; (j) Male perch call, subtype 2, loud version; (k) Male perch call, subtype 2, soft version; (l) Male perch call, subtype 3; (m) Female perch call, subtype 4; (n) Female perch call subtype 5; (o) Female perch call, subtype 6. Spectrogram parametres: Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap...... 63

Fig. 3.3: Spectrograms of vocalisations and behavioural interactions of adult and nestling south- eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, at nests. Symbols denote sex of adult birds. (a) Nestling call (subtype 1), male perch call (subtypes 1 and 2) and the begging of a beginning bout by the female; (b) Nestling call (subtype 2), male soft perch call (subtype 3) and female begging; (c) Nestling call (subtype 3) showing clear nonlinearity. Female responds with soft nest call and male with soft version of perch subtype 3; (d) Loud female begging bout showing clear subharmonics. Take-off flight by male and female. Other species’

v vocalisations are not indicated. Spectrograms created from video footage using Raven Pro 1.5 (Cornell Lab of Ornithology; Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap). X axis denotes time into the video file...... 64

Fig. 3.4: Example spectrograms of vocalisations and behavioural interactions of adult and nestling Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, at nests. Symbols denote sex of adult birds. (a) Male soft perch call (subtype 2) and female loud perch call (subtype 4); (b) Female loud alarm perch call (subtype 5) and male perch calls (subtypes 1, 2, and 3); (c) Female nest entry call and soft perch call (subtype 6); (d) Nestling call, female flight and nest entry call, and male soft perch (subtype 2); (e) Female begging bout and male response (perch 2). Other species’ vocalisations are not indicated. Spectrograms created from video footage using Raven Pro 1.5 (Cornell Lab of Ornithology; Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap). X axis denotes time into the video file...... 66

Fig. 3.5: Linear discriminant analysis of nest-associated vocalisations of (a) the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus (MANOVA: Wilk’s λ = 0.04, F = 47.362, p < 0.001) and (b) the south-eastern red-tailed black-cockatoo, C. banksii graptogyne (MANOVA: Wilk’s λ = 0.16, F = 22.518, p < 0.001). Dashed lines represent normal confidence intervals. Solid lines represent Euclidean distances...... 68

Fig. 4.1: Spectrograms of a fledging event of (a) the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, and (b) the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus. Loud nestling calls are evident until ~ 30 seconds, followed by the fledging event (~ 32-40 seconds) when the nestling leaves the nest. Nestling and adult calls are evident as the birds call in flight. Calls rapidly attenuate in amplitude with increasing distance of the birds from the nest tree. Spectrograms created using Raven Pro 1.5 (Cornell Lab of Ornithology; Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap)...... 85

Fig 5.1: Box plot of nestling call rate (count of vocalisations per day) by week of development for the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne (RTBC; n = 3 nestlings) and the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus (GBC; n = 7 nestlings). Both subspecies show a marked increase in nestling call rate in the final week of nesting (F = 4.307, p = 0.055 for the south-eastern red-tailed black-cockatoo; F = 8.486, p = 0.006 for the Kangaroo Island glossy black-cockatoo). Solid lines represent weeks when data were only available from a single nest...... 101

Fig 5.2: Spectrograms showing nestling vocal ontogeny of the (a) Kangaroo Island glossy black- cockatoo, Calyptorhynchus lathami halmaturinus, and (b) south-eastern red-tailed black- cockatoo, Calyptorhynchus banksii graptogyne. Each species is represented by a single nest from approximately 5 weeks of age (Day 53 and 54) through to the day of fledging (Day 0). Early vocalisations tend to be soft, but older nestlings commonly give both loud and soft calls. Loud forms are shown here...... 102

Fig. 5.3: Ontogenetic changes in vocalisations of nestling glossy black-cockatoos, Calyptorhynchus lathami halmaturinus, and red-tailed black-cockatoos, Calyptorhynchus banksii graptogyne. (a) Peak amplitude (u), (b) Call duration (seconds), (c) Low frequency (Hz), (d) Aggregate entropy (bits), (e) Average entropy (bits). Asterisks denote significance (* p < 0.05; ** p < 0.01; *** p < 0.001)...... 103

vii

List of Tables

Table 1.1: Threatened species listings of the red-tailed black-cockatoo and the glossy black- cockatoo at the national and relevant state levels (AUS: Environment Protection and Biodiversity Conservation Act 1999; QLD: Nature Conservation Act 1992; NSW: Threatened Species Conservation Act 1995; VIC: Advisory List of Threatened Vertebrate Fauna in Victoria (2013), Flora and Fauna Guarantee Act 1988; SA: National Parks and Wildlife Act 1972; WA: Wildlife Conservation Act 1950; NT: Territory Parks and Wildlife Conservation Act). CR: Critically Endangered, E: Endangered, VU: Vulnerable, NT: Near Threatened...... 8

Table 2.1: Examples from the literature summarising the potential benefits of bioacoustic monitoring programs of animal behaviour over traditional methods. Note that several challenges or benefits are likely to apply to any given monitoring program...... 33

Table 2.2: Summary of vertebrate vocal behaviours that are relevant to conservation...... 36

Table 3.1: Description of nest-associated call types and associated behavioural contexts of the south-eastern red-tailed black-cockatoo (RTBC), Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black-cockatoo (GBC), C. lathami halmaturinus...... 57

Table 4.1: Recogniser performance evaluated at the level of the survey day for the south-eastern red-tailed black-cockatoo (RTBC), Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black-cockatoo (GBC), C. lathami halmaturinus. (a) Nest activity correctly assigned as active (true positive detections returned) or inactive (no detections returned); (b) Nest active but no true positive detections returned; (c) Nest inactive but false positive detections returned...... 86

Table 4.2: Precision of binary point matching call recogniser for detecting nestling calls of the south-eastern red-tailed black-cockatoo (RTBC), Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black-cockatoo (GBC), C. lathami halmaturinus. Precision reported for the nestling calls (% TP nestling) and for nestling and adult calls combined (%TP nestling + adults). Total detections (n total), number of nests for which detections were returned (n nests), and mean number of detections per nest (mean n per nest) are shown. Times 1, 2 and 3 represent early, mid and late stages of the recording periods. Precision = true positives/(true positives + false positives)...... 87

viii

List of Abbreviations

GBC: Glossy black-cockatoo RTBC: Red-tailed black-cockatoo

Acknowledgement of Traditional Owners

I acknowledge the Traditional Owners of the land on which this research took place. I pay my respect to Elders, past, present and emerging.

Note regarding Kangaroo Island

In December 2019 and January 2020, severe wildfires impacted Kangaroo Island, including large habitat areas for the glossy black-cockatoo. Several nesting sites monitored for the research presented in this thesis were impacted. Many nesting hollows were lost. At the time of writing, we do not know the full scale of the impact to the glossy black-cockatoo, but the subspecies’ conservation status is likely to be more critical than when this research took place.

Chapter 1 | Introduction

Bioacoustic monitoring for conservation

Species conservation requires robust data to inform decision-making. Monitoring population trends and responses to environmental change, whether management-based or otherwise, are central to effective conservation. Acquiring data has, historically, been the realm of the field biologist. However, relying on human observers limits the scales at which data can be collected and may, in some cases, preclude comprehensive monitoring (e.g. highly cryptic species). New technologies provide alternative means to efficiently monitor populations and ecosystems across large spatial and temporal scales, greatly expanding our capacities to study living systems. Bioacoustics, the study of animal sounds, is an emerging field of wildlife monitoring, aided by substantial advances in digital technology, power and data storage. While biologists have been recording wildlife vocalisations and soundscapes for decades, today ecologists can record data autonomously for long time periods with minimal human involvement. Methods to analyse these data for a range of conservation-relevant metrics are increasingly being developed (Efford 2011; Efford & Fewster 2013; Stevenson et al. 2015; Buxton et al. 2016). There are now more options to efficiently and repeatedly study animal populations, proving novel insights into species’ responses to altered environments and conservation actions. This is a core principle of adaptive management (Blumstein & Fernández-Juricic 2010) and will undoubtedly see bioacoustics continue to gain popularity for conservation monitoring.

Bioacoustics has huge potential for a wide range of vocal taxa but it has been most widely applied to species that are cryptic, rare or otherwise difficult to observe. Indeed, studies of presence-absence and species richness have increasingly used bioacoustic methods to detect individuals, often outperforming human observers (Venier et al. 2012; Wimmer et al. 2013; Dema et al. 2018). For example, on the island of Hawai’i, bioacoustics recently confirmed the presence of the rarely observed ‘Amikihi honeyeater, Hemignathus virens virens, in a lowland area where it was thought to be precluded by avian malaria (Sebastián- González et al. 2015). Bioacoustics is also helping map the distribution of highly cryptic and threatened birds in remote Australia, including the night parrot, Pezoporus occidentalis, the plains-wanderer, Pedionomus torquatus, and coxen’s fig parrot, Cyclopsitta diophthalma coxeni (N. Leseberg, I. Gynther, pers. com.; DT, pers. obs.). A major benefit of bioacoustics in such cases is the ability to survey for substantially longer time periods than is normally

1 feasible by human observers, thus increasing the detection probability for species that are rare or vocalise infrequently (Zwart et al. 2014; Dema et al. 2018; Znidersic et al. 2020). Nonetheless, bioacoustics is also useful for vocally conspicuous species. For example, bioacoustics has been used to estimate nesting colony size in Forster’s terns, Sterna forsteri (Borker et al. 2014), and density of bell miners, Manorina melanophrys (Lambert & McDonald 2014). In this capacity, bioacoustics may offer many novel options for monitoring highly vocal species for a range of population metrics, including density and occupancy (Dawson & Efford 2009; Marques et al. 2013; Furnas & Callas 2015; Stevenson et al. 2015; Chambert et al. 2018). Further, for species with complex vocal repertoires, bioacoustics may provide options to monitor a range of behavioural contexts.

Bioacoustic monitoring at large scales, while highly informative, generates large volumes of sound files that require processing to extract relevant data. Unfortunately, inefficiencies in this regard remain the field’s major hurdle. Manual processing, where humans listen to sound recordings and view spectrograms, is impractical in most situations. Therefore, algorithms to detect vocalisations from sound data, termed call recognisers, are important for improving the utility of bioacoustics (Sugai et al. 2019). Recognisers are highly diverse, from complex machine learning algorithms to simpler template-matching (image- based) approaches. There are now numerous commercial options (e.g., Kaleidoscope software by Wildlife Acoustics Inc.) and open-source recognisers are also becoming available. However, while recogniser development has greatly increased, their performance and reliability has been mixed (Priyadarshani et al. 2018). Common issues arise from the variability of calls, within species and individuals, and the extraneous sounds that recordings are unavoidably subject to (Brandes 2008; Towsey et al. 2012; Salamon et al. 2016). Moreover, recogniser construction has typically required expert programming skills, which has limited the utility to ecologists responsible for monitoring. If recognisers are used, such as via a commercial interface, their underlying functioning and performance (i.e., reliability of outputs) may not be fully understood. This can lead to spurious data, with potential consequences for management (Lemen et al. 2015; Russo & Voigt 2016). Improving these issues is critical to the success of bioacoustics as a wildlife monitoring tool.

Bioacoustics is typically discussed broadly for its potential to collect data; its ability to record species’ presence in the field and detect individual calls in sound files. However, to be more useful in conservation, the design of bioacoustic programs, both field-based methods

2 and recognisers, must be established in the context of specific research questions and the needs of conservation end-users. Ideally, this would be done through collaborative processes. Studies seeking rare and cryptic species will invariably differ from those measuring conspicuous and abundant species. For instance, in seeking species that vocalise infrequently, recording duration will likely be longer and recognisers more sensitive, since minimising missed detections will be important. In other cases, bioacoustic programs could be made more practical by limiting recording times, the number of recorders deployed and recogniser performance to that necessary for capturing the desired vocal behaviours (e.g., conspicuous frog advertisement calling in early morning). In this capacity, bioacoustics can be targeted to specific population metrics and trigger points required for management; this is necessary for any monitoring program to translate into conservation outcomes (Lindenmayer et al. 2013). Notwithstanding, as technologies continue to advance, it is possible that large scale monitoring networks (e.g., Australia’s new Acoustic Observatory, acousticobservatory.org) and multi-species recognisers will serve a range of research objectives. Nevertheless, only when bioacoustics properly considers the conservation problems at hand and the design needs of practitioners will it truly be an applied conservation tool.

Study species

In this thesis, I focus on bioacoustic monitoring of two black-cockatoo species: the glossy black-cockatoo, Calyptorhynchus lathami, and the red-tailed black-cockatoo, Calyptorhynchus banksii. Specifically, my research concerns two endangered subspecies, the Kangaroo Island glossy black-cockatoo, C. l. halmaturinus, and the south-eastern red-tailed black-cockatoo, C. b. graptogyne. Both subspecies are restricted to a single isolated population, and breeding success is a limiting factor in their recoveries. A high priority for management, therefore, is to understand how breeding success varies by habitat features, especially food availability. The methods with which breeding success is inferred differs between the two subspecies, though data for both are currently limited by resource constraints. This is concerning since both subspecies are highly threatened.

The south-eastern subspecies of red-tailed black-cockatoo comprises approximately 1,400 individuals that occur across 18,000 km2 of south-west Victoria and south-east South Australia (Fig. 1.1). Flock monitoring is conducted annually, and is coordinated by Birdlife Australia and the South-eastern Red-tailed Recovery Team (www.redtail.com.au) (Burnard & Pritchard 2013). Volunteers assist in locating flocks in the landscape. Following this,

3 members of the recovery team revisit flocks to collect basic demographic data. Most importantly, the relative proportion of adult male birds to adult female and juvenile birds (collectively, since they are highly similar in appearance) in each flock is estimated by skilled observers. This provides an index of breeding success; a decrease in the proportion of females and juveniles suggests a decrease in breeding success. A marked decline in females and juveniles since 2015 (mean proportion < 55%) has spurred significant concern from the recovery team (Hill, 2019). Unfortunately, since there has been no routine nest monitoring, we know very little about the habitat features that influence nest site choice and the likelihood of fledging. Management actions to improve breeding are, therefore, limited. The major challenges of directly monitoring nests are the accessibility of nesting areas and the effort required to observe nests through to fledging (or failure). Nests are usually remote, and often on private land.

The Kangaroo Island glossy black-cockatoo exists only on Kangaroo Island in South Australia since disappearing from the mainland by the 1970s (Cleland & Sims 1968; Baird 1986; Higgins 1999) (Fig. 1.1). The subspecies is comparatively well studied and has been intensively managed (Pepper 1997; Garnett et al. 1999; Pepper et al. 2000; Crowley & Garnett 2001; Chapman & Paton 2005; Berris et al. 2018). The population was subject to a large monitoring program from 1995, when the recovery program began, until 2017 when the program’s funding ended. Since the distribution of the birds is well known, annual flock counts (‘censuses’) were conducted each spring by state government officers and volunteers. By 2016, the population comprised at least 373 birds (Berris et al. 2018), a substantial increase from the estimated 115-150 birds in the 1980s (Joseph 1982). This is testament to the success of the intensive on-ground management over many years. In particular, the protection of nests from predatory common brushtail possums, Trichosurus vulpecula, has been critical to the population’s recovery. This is achieved by installing a corrugated iron collar around the trunk of each nest tree, regularly inspecting and repairing collars, and pruning neighbouring trees to prevent canopy access. The installation of artificial nest boxes has also been critical to this success. Additional conservation measures included control of invasive honeybees, which occupy nest hollows, some lethal control of nest competitors, and extensive planting of food trees. While this program has clearly been successful, the resources required were substantial. Since 2017, very little monitoring has occurred, and management has been largely limited to the critical nest maintenance works. The future of the program is currently unclear.

There is currently no consistent method with which to accurately and efficiently monitor large numbers of nests in the south-eastern red-tailed black-cockatoo and the Kangaroo Island glossy black-cockatoo. Further, there is no method with which to directly monitor fledging. Traditional human-observer methods are difficult and resource-intensive, and do not offer long-term options, particularly in the face of uncertain funding. This issue is owing to the challenges of finding nests and in observing their activity through to fledging, primarily because the birds are rare in the landscape, tend to occur remotely and often on private land, and nests are sparsely distributed. For breeding to be feasibly monitored in the future, new methods that are accurate, efficient, and have smaller human resource requirements are needed.

Bioacoustic monitoring of vocal behaviour in glossy and red-tailed black-cockatoos

Bioacoustics may offer new ways to improve monitoring of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Like most cockatoos, the glossy black-cockatoo and the red-tailed black-cockatoo are generally considered noisy birds with harsh calls (Higgins 1999). Their conspicuous, raucous behaviour may make them good candidates for bioacoustic monitoring. The vocal behaviour of adult glossy black-cockatoos is known from a study on the Kangaroo Island subspecies by Pepper (1996), who described nine distinct calls types and their associated behavioural contexts, but suggested the repertoire is probably much larger. There are no detailed studies of the vocalisations of the red-tailed black-cockatoo, and currently only four calls are described in informal accounts (Higgins 1999). Pepper (1996) suggested that, despite their taxonomic closeness, the vocalisations of the red-tailed black-cockatoo and glossy black-cockatoo are markedly different. However, Courtney (1996) noted the juvenile food-begging calls of these species are very similar, and distinct from other black-cockatoo species by being a clear ‘squeak’ rather than a harsh and raspy call. The begging call of the red-tailed black-cockatoo, however, was only briefly observed (Courtney 1996). Moreover, the glossy and red-tailed black-cockatoos are the only black-cockatoos in which food-swallowing vocalisations are absent (Courtney 1996).

Vocal behaviour

The calls described for the red-tailed black-cockatoo are the adult contact call, the adult alarm call, the adult male display call (also known as the advertisement call or the courtship call), and the juvenile begging call. The contact call is typically described as a

5 rolling krurr-rr or krur-rurr-kee that can be heard from afar (Higgins 1999; Menkhorst et al. 2017). The alarm call is usually described as a sharp or abrupt krur-rak (Higgins 1999; Menkhorst et al. 2017). The behavioural contexts of these vocalisations are only broadly described in most accounts (e.g. to maintain contact; when alarmed). The adult male display call is described as beginning with a scratchy krow-row row-row that becomes a loud and rhythmic pulse of kred-kred-kred-kred. This call is said to be given by males during a courtship display (Higgins 1999).

For adults of the Kangaroo Island glossy black-cockatoo, the calls described by Pepper (1996) were the flight call, the alarm call, the feeding call, the allofeed call, the grunt call, the kwee-chuck call (i.e. the adult male display call), the begging call, and the nest call. The flight call was described as loud, flat-pitched, and long in duration. The alarm call was described as being quite variable, which may denote the level of concern. The feeding call is a soft and short call given by paired birds during foraging, and appeared to function in contact maintenance. The allofeed call is similarly soft, but with a rising and falling unit at the end of the call, given by males before allofeeding their mate. The grunt call is a short, atonal call, repeated several times, given by adult males, and appeared to function as a threat display. The kwee-chuck (display) call is given only by males. It comprises a longer tonal unit followed by a short broadband ‘chuck’, repeated many times and was immediately followed by a bow display. In a bow display, the male raises its head briefly before lowering it (‘bowing’) and spreading his tail feathers. Pepper (1996) notes this call was often given by adult males when perched near their mate, and almost entirely during the breeding season. It was usually given after allofeeding and before copulation. It is, therefore, thought to function in pair bond maintenance and courtship. However, unpaired subadult males also gave this call, usually from a prominent perch. Often other males would respond with the same call. More dominant males would sometimes supplant the caller, suggesting the call also functions in establishing dominance. The begging call was described as a repetitive high-pitched ‘squeak’. Begging calls were recorded from both adult females and fledglings, but their structures was not differentiated. Pepper (1996) did note that the begging call was highly variable.

Very little is known about nest-associated vocalisations in either the glossy black- cockatoo or the red-tailed black-cockatoo. The only formal account of a nest vocalisation is the female nest call in the glossy black-cockatoo, described by Pepper (1996). This call is given by females when they are near a nest or entering a nest. Pepper (1996) described this

6 call as a ‘grating’ sound comprising distinct pulses. Nestling vocalisations are not formally described in either species, although Cameron (2009) provides anecdote that glossy black- cockatoo nestlings give a harsh growl from about 50 days of age. Likewise, there are no formal descriptions of the vocalisations that accompany a fledging event for either species. However, anecdotes indicate that fledging is associated with loud vocalising. There is one report from the red-tailed black-cockatoo of parents calling loudly near the nest before fledging (Higgins 1999). The few fledging events that have been observed for the Kangaroo Island glossy black-cockatoo were stimulated by the parents calling loudly and repetitively to their chick (M. Barth, pers. com.). It seems likely, therefore, that there is a vocal signal of fledging in both species. If this is true, bioacoustics may provide the first method with which to directly monitor nest outcome.

Expanding bioacoustics to other subspecies

Bioacoustic methods developed for the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo may be transferrable to other subspecies. This is important because there are several subspecies of glossy and red-tailed black-cockatoo across Australia and many are threatened (Table 1.1). In addition to the south-eastern subspecies, C. b. graptogyne, which is restricted to south-west Victoria and south-east South Australia, red- tailed black-cockatoos comprise four other subspecies across mainland Australia, the taxonomy of which has recently been examined (Ewart et al. 2020). The forest red-tailed black-cockatoo, C. b. naso, exists in south-west Western Australia, while the inland red-tailed black-cockatoo, C. b. samueli, comprises several populations widely distributed across inland Australia. C. b. escondidus is a newly described subspecies, previously considered a western population of C. b. samueli (Ewart et al. 2020). A fifth population, C. b. banksii, occurs across north-eastern Australia, through the Top End, Cape York, eastern Queensland and, very rarely, in New South Wales, although birds to the west of the Carpentarian Barrier were, until recently, considered to be a separate subspecies, C. b. macrorhynchus (Ewart et al. 2020).

There are three recognised subspecies of glossy black-cockatoo (Schodde et al. 1993). The Kangaroo Island subspecies, C. l. halmaturinus, is restricted to Kangaroo Island in South Australia, while two other subspecies occur on the east coast of mainland Australia. The south-eastern glossy black-cockatoo, C. l. lathami, occurs from about Hervey Bay in central Queensland, south to Mallacoota in north-eastern Victoria. The northern glossy black-

7 cockatoo, C. l. erebus, is a Queensland endemic, occurring from about Bundaberg in the south to Paluma, near Townsville, in the north (Schodde et al. 1993; Garnett et al. 2000). This subspecies is very poorly studied.

Table 1.1: Threatened species listings of the red-tailed black-cockatoo and the glossy black-cockatoo at the national and relevant state levels (AUS: Environment Protection and Biodiversity Conservation Act 1999; QLD: Nature Conservation Act 1992; NSW: Threatened Species Conservation Act 1995; VIC: Advisory List of Threatened Vertebrate Fauna in Victoria (2013), Flora and Fauna Guarantee Act 1988; SA: National Parks and Wildlife Act 1972; WA: Wildlife Conservation Act 1950; NT: Territory Parks and Wildlife Conservation Act). CR: Critically Endangered, E: Endangered, VU: Vulnerable, NT: Near Threatened.

Species AUS QLD NSW VIC SA WA NT

Red-tailed black-cockatoo1

C. banksii graptogyne (south-eastern) EN2 EN EN C. b. naso (forest) VU3 VU C. b. banksii (coastal) CR C. b. samueli (inland) VU NT

Glossy black-cockatoo

C. lathami (species level) VU VU C. l. lathami (south-eastern) EN4 VU C. l. halmaturinus (Kangaroo Island) EN EN

Both the glossy and red-tailed black-cockatoo are widely distributed, often occur in remote locations and face ongoing threats, including habitat loss and bushfire. Even subspecies that are thought to be relatively secure may be experiencing population declines; the coastal red-tailed black-cockatoo, for example, appears to be contracting northwards, with very few sightings today in northern New South Wales and southern Queensland.

1 C. b. escondidus is newly described and has not been assessed for its conservation status. 2 Recovery plan in effect under the EPBC Act since 2007. A new recovery plan is currently under review. 3 Recovery plan in effect under the EPBC Act since 2011. 4 Riverina population.

Understanding species’ trajectories may, therefore, require more monitoring in more locations. It is possible that bioacoustics could help facilitate this. If vocal behaviours are similar among subspecies, the methods presented in this thesis may be transferrable to other subspecies, either directly or with some refinement. In this regard, while this thesis focusses on the two most threatened subspecies, as these have the most urgent need for improved monitoring, the applications of this work are not necessarily limited to these subspecies. Rather, this thesis provides fundamental knowledge on the vocalisations of the glossy and red-tailed black-cockatoo, as well as a framework with which to use this in bioacoustic monitoring.

Thesis aims and structure

The overall aim of thesis is to develop bioacoustic methods for monitoring breeding in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Specifically, for each subspecies, I aimed to (a) describe the vocalisations informative for nest monitoring, including those associated with fledging, (b) develop and test call recognisers to detect nest activity and success, and (c) examine variance in nestling vocalisations in relation to age. Importantly, I aimed to make the methods easily accessible to the conservation practitioners responsible for monitoring. This is a guiding tenet throughout the thesis.

This thesis comprises six chapters, namely an introduction chapter, four core chapters, and a thesis synthesis and conclusion chapter. In Chapter 2, I present a literature review of the potential applications of bioacoustics to behavioural monitoring for conservation. The remaining three parts relate to the development of the bioacoustic methods. In Chapter 3, I describe the nest-associated vocalisations, and behavioural contexts, of both subspecies. I discuss the vocal behaviours most suited to a bioacoustic nest monitoring program. In Chapter 4, I develop and test call recognition algorithms to detect nest activity and fledging. In Chapter 5, I provide a preliminary analysis on vocal ontogeny of nestlings, as a tool with which nests can be aged from sound data. In Chapter 6, I conclude that bioacoustics is a useful tool for nest monitoring in these endangered subspecies and discuss limitations.

➢ Chapter 1: Introduction This chapter provides an overview of bioacoustics and the study species. I discuss the current status of the Kangaroo Island glossy black-cockatoo and the south-eastern red- tailed black-cockatoo.

➢ Chapter 2: Bioacoustic monitoring of animal behaviour for conservation In this review, I examine the utility of bioacoustics for monitoring animal behaviour for the purposes of conservation. Bioacoustic studies tend to focus on species’ most conspicuous calls, to examine metrics like presence-absence, but there is great potential for bioacoustics to monitor specific behavioural contexts. I consider vertebrate vocalisations that indicate reproduction and recruitment, alarm and defence, and social behaviour, and demonstrate how this knowledge could be useful for conservation.

This chapter has been published in Conservation Science and Practice.

➢ Chapter 3: Nest-associated vocal behaviours of the south-eastern red-tailed black- cockatoo, Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black- cockatoo, C. lathami halmaturinus In this chapter, I describe the nest-associated vocal behaviours of the south-eastern red- tailed black-cockatoo and the Kangaroo Island glossy black-cockatoo. Using autonomous sound recorders at nests, combined with field observations and video footage, I identified vocalisations characteristic of six behavioural contexts at nests: birds in flight, while perched, during begging (adult females), during courtship displays (adult males), when entering or sitting near to the nest hollow entrance (adult females), and from nestlings. This knowledge is a necessary first step for a comprehensive bioacoustic nest monitoring program.

This chapter has been published in Austral Ecology.

➢ Chapter 4: Fledge or fail: Nest monitoring of the endangered black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus, using bioacoustics and open-source call recognition. In this chapter, I develop and test open-source call recognisers for extracting nesting vocalisations from long-term recordings (breeding season length). Using the information

on vocal behaviour presented in Chapter 3, I built the recognisers to detect loud nestling calls. Following a pilot study, I constructed two template-matching recognisers (one per subspecies) using binary point matching, implemented in the R statistical language. I tested the recognisers in a regime appropriate for the species (early, mid and late stages of the recording period for each nest), such that it could applied in future monitoring programs. I quantified the performance of the recogniser in detecting nest activity (active or inactive) and nest outcome (fledge or fail). In constructing the recognisers, I aimed to create a method that is easily reproducible and straightforward for managers to implement.

This chapter has been prepared for submission to Bioacoustics.

➢ Chapter 5: Vocal ontogeny of nestling black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus In this chapter, I examine the vocal ontogeny of nestlings and discuss how this could be used to extract age-related information from sound recordings at nests. Using daily measurements, I quantified changes in acoustic structure during nest development, from the time when nestlings first vocalised through to fledging. I discuss how vocal ontogeny can provide an additional source of demographic information in bioacoustic studies.

This chapter has been prepared for submission to Austral Ecology.

➢ Chapter 6: Thesis synthesis and conclusion In this chapter, I summarise my thesis findings and provide a framework for bioacoustic nest monitoring of glossy black-cockatoos and red-tailed black-cockatoos. I discuss the potential to improve monitoring in the endangered subspecies examined here, the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, and also how this could translate to other populations of these species. Finally, I discuss future directions for bioacoustic monitoring of these species.

Chapter 1 Background, aims and thesis overview

Chapter 2 Review of bioacoustics for monitoring behaviour for conservation Behavioural background for bioacoustic monitoring programs

Chapter 3 Description of vocal behaviours

Chapter 4 Call recognisers to monitor nest activity and success Applied conservation:

Using bioacoustics to acquire conservation-relevant data from Chapter 5 nests Vocal ontogeny as an additional source of demographic data

Chapter 6 Thesis synthesis and summary and future directions

Fig. 1.2: Structure of thesis chapters.

Terminology

The terminology used in the bioacoustic literature is often inconsistent. For clarity, I generally follow the terminology provided by Bradbury & Vehrencamp (2011) and Köhler et al. (2017) for broad definitions of sound and signal production by animals, and Catchpole & Slater (2008) for describing avian vocalisations:

Vocalisation: Collective term for bird calls and songs. Vocalisations are produced by an animal’s respiratory system. The term also applies to any other vocally active species.

Call: Vocalisations that are usually relatively short and simple in structure. Calls are usually produced by both sexes. Calls appear to relate to specific functions, such as flight or alarm. Cockatoo vocalisations are referred to as calls.

Chaos: Noisy signals, or parts of signals, generated by random processes (non-linear phenomena). Also termed deterministic chaos.

Song: Vocalisations produced by songbirds. Generally, bird songs are longer and more complex than calls. There are many exceptions, however, the many authors have argued that the distinction between a call and a song is arbitrary.

Pulse: Single burst of sound, not further divided into subunits, separate from others by strong amplitude modulation. Pulses can be arranged into pulse groups.

Phrase: Distinct unit of a vocalisation. Vocalisations often comprise several phrases.

Syllable: Distinct unit of a phrase. Syllables are often complex in structure, particularly for songbirds.

Element: Distinct unit of a syllable. An element is usually a single continuous line on a spectrogram.

Frequency: Number of complete oscillation cycles of sound waves per unit time. Measured in Hertz (Hz) or kiloHertz (kHz). Sound audible to humans exists between 20 – 20 000 Hz. Frequency is related to what humans term the ‘pitch’ of a sound. As the repeat rate of sound waves slows down as the vocalisation ends (i.e. the frequency decreases), we hear a lower pitched sound. Variation in frequency during the course of a vocalisation is termed frequency modulation.

Amplitude: Difference between peak pressure (peak of sound wave) and ambient pressure. Proportional to sound intensity. Relative unit of amplitude is the decibel (dB). Often referred to as the sound pressure. A vocalisation that varies in its amplitude is said to have amplitude modulation.

Wavelength: The spatial length of a single cycle of a sound wave in a given medium. The greater the wavelength, the greater the time required to generate the sound. As such, wavelength and frequency are inversely related. Animals that vocalise at low frequencies (e.g. elephants) are therefore able to communicate over much greater distances than those whose vocalisations are high in frequency.

Harmonic: Stacked, separated frequencies in a sound. The lowest frequency is termed the fundamental or reference harmonic. Harmonics are integer multiples of the fundamental harmonic. Collectively, they are referred to as a harmonic series.

Tonal: Sound comprising a single frequency component at any time point.

Power spectrum: Visual representation of sound which shows the amplitude of each frequency component.

Spectrogram: Visual representation of sound which shows frequency and amplitude over time. Also termed a sonogram or audiospectrogram.

a b

P1 P2 P1 P2 P3

P H H FM C

Chapter 2 Author Contributions Chapter 2 is presented as published in Conservation Science and Practice, with some amendments to accommodate suggestions from the examiners of this thesis. Daniella Teixeira conceived the topic of the literature review and wrote the manuscript. Berndt van Rensburg and Martine Maron helped refine the review’s scope and contributed to drafts.

Chapter 2 | Bioacoustic monitoring of animal vocal behaviour for conservation

Abstract

The popularity of bioacoustics for threatened species monitoring has surged. Large volumes of acoustic data can be collected autonomously and remotely with minimal human effort. The approach is commonly used to detect cryptic species and, more recently, to estimate abundance or density. However, the potential for conservation-relevant information to be derived from acoustic signatures associated with particular behaviours is less well- exploited. Animal vocal behaviour can reveal important information about critical life history events. In this paper, we argue that the overlap of the disciplines of bioacoustics, vocal communication and conservation behaviour – thus, ‘acoustic conservation behaviour’ - has much to offer threatened species monitoring. In particular, vocalisations can serve as indicators of behavioural states and contexts that provide insight into populations as it relates to their conservation. We explore the information available from monitoring species’ vocalisations that relate to reproduction and recruitment, alarm and defence, and social behaviour, and how this information could translate into potential conservation benefits. While there are still challenges to processing acoustic data, we conclude that acoustic conservation behaviour may improve threatened species monitoring where vocalisations reveal behaviours that are informative for management and decision-making.

Introduction

Bioacoustic monitoring is a rapidly emerging tool in wildlife conservation, aided by recent advances in technology and analytical approaches (Snaddon et al. 2013). Distinct from the related discipline ecoacoustics, bioacoustics is behaviour-centric and focusses on the acoustic signals of individuals and species, rather than broader ecological processes or environments (Towsey et al. 2014; Sueur & Farina 2015). Potentially suited to any sound- producing species, especially those that are rare, cryptic or otherwise difficult to observe (Zwart et al. 2014; Wrege et al. 2017; Williams et al. 2018), bioacoustic monitoring via autonomous recording units is becoming increasingly popular for measuring metrics such as species presence-absence (Zwart et al. 2014; Sebastián-González et al. 2015), species richness (Celis-Murillo et al. 2012; Wimmer et al. 2013), abundance (Borker et al. 2014; Jaramillo-Legorreta et al. 2017) and density (Efford 2011; Efford & Fewster 2013; Marques et al. 2013; Rogers et al. 2013; Lambert & McDonald 2014; Stevenson et al. 2015).

In contrast, bioacoustics is not commonly used to monitor animal behaviour, despite the extensive literature on animal acoustic communication and vocal behaviour (Bradbury & Vehrencamp 2011). This is unfortunate because vocalisations of anurans, birds and mammals are often well-described for critical behaviours like reproduction and defence (Bradbury & Vehrencamp 2011). Further, new discoveries in animal communication, such as the recent recognition of the ancestral and widespread functions of female birdsong, provide a plethora of opportunities to monitor less well-known behaviours (Odom et al. 2014). Knowledge on vocal behaviours can contribute to conservation outcomes if those behaviours provide insight into a population that exposes a conservation problem or monitors responses to conservation actions. We propose, therefore, that bioacoustic monitoring programs should give greater consideration to animal behaviour (Table 2.1). For some species, bioacoustics may be the only feasible approach with which to acquire behavioural data, such as for cetaceans’ cryptic behaviours (McDonald et al. 2017). Bioacoustics may also be more efficient and cost- effective than traditional methods (Williams et al. 2018).

The discipline of conservation behaviour seeks to apply behavioural knowledge to conservation solutions. Although a relatively young discipline, several books have been published on the topic (Buchholz & Clemmons 1997; Caro 1998; Gosling & Sutherland 2000; Festa-Bianchet & Apollonio 2003; Blumstein & Fernández-Juricic 2010; Berger-Tal & Saltz 2016), and its scope and utility have received considerable attention in recent literature (Caro 1999; Harcourt 1999; Knight 2001; Linklater 2004; Angeloni et al. 2008; Caro & Sherman 2011, 2013; Caro 2016). Early on, conservation behaviour research tended to focus on behaviours that affect population persistence (Anthony & Blumstein 2000), while recent studies have commonly examined behavioural responses to anthropogenic impacts (Slabbekoorn & Ripmeester 2008; Barber et al. 2010; Laiolo 2010; Owens et al. 2012; Shannon et al. 2014; Templeton et al. 2016). To better define the application of animal behaviour to conservation, Berger-Tal et al. (2011) proposed a framework of three interconnected conservation themes to which behaviour is relevant. These themes are defined as: 1. anthropogenic impacts on animal behaviour, which can alter the fitness of individuals; 2. behaviour-based management, in which managers aim to preserve or change a species’ behaviour (e.g. pre-translocation anti-predator training); and 3. behavioural indicators, whereby the behaviours themselves are a source of information about the state of a species or its habitat. Theme 3 pertains to monitoring, which is most relevant to bioacoustic studies of animal behaviour. Specifically, by monitoring vocal behaviours that indicate the state of

19 individuals or groups for metrics that are relevant to conservation (e.g., reproductive status), populations and their trends can be tracked to inform decisions.

Under theme 3, Berger-Tal et al. (2011) provides two pathways that link behavioural domains to conservation outcomes. First, behaviours can reveal threatening processes and act as warning systems for population declines. Vocal behaviours may, for example, signal habitat quality or reproductive success. In the Eurasian eagle owl, Bubo bubo, males with the greatest fecundity and proportion of rats in the diet called significantly earlier during the dusk chorus (Penteriani et al. 2014). Vocal behaviours like these may reveal areas where reproduction is relatively poor, even if many individuals persist there. For species whose behaviours are socially transmitted, the size of groups that a habitat can support may be critically important. Fragmented populations may see the loss of important behaviours (cultural erosion) which can cause declines faster than would otherwise be expected (Laiolo & Jovani 2007; Laiolo & Tella 2007; Hart et al. 2018). In this capacity, habitat ought to support not just a species’ presence, but also the presence of behaviours that facilitate its long-term viability. Bioacoustics may serve to acquire this knowledge. Second, behaviours can provide a means to monitor and evaluate the effectiveness of conservation actions. This process of evaluation is critical for effective conservation planning and adaptive management (Lindenmayer et al. 2013). Monitoring vocal behaviours may facilitate this by examining temporal and spatial trends in behaviours that indicate key life-history stages. An example of this would be examining changes in a population’s breeding success, such as via the presence of infants’ vocalisations, over time following conservation action.

In this review, we examine the potential benefits to conservation from the bioacoustic detection of animal vocal behaviour. We consider vertebrate vocalisations that may facilitate conservation monitoring, namely those associated with reproduction and recruitment, alarm and defence, and social behaviour (Table 2.2). The challenges of processing large acoustic datasets are discussed. We conclude that monitoring vocal behaviour using bioacoustics has much to offer conservation, provided that monitoring programs are explicit in their objectives and the link between them and the vocalisations detected.

Vocal behaviour and conservation

Reproduction and recruitment

Understanding how a species’ reproductive behaviours and breeding success differ among various levels of habitat perturbation can be informative for protected area planning or

20 habitat restoration. If vocalisations reveal that certain environments consistently comprise individuals with poorer reproductive success, then decisions can be made about improving habitat or, alternatively, focussing conservation efforts elsewhere. In male songbirds, mate attraction is a chief function of singing, as is territory defence (Marler & Slabbekoorn 2004; Nowicki & Searcy 2005; Catchpole & Slater 2008). Sexual selection can act strongly on males’ songs, since for many species females choose males according to song characteristics (Andersson 1994). Song complexity, singing rate, singing duration and repertoire size have been shown to influence females’ behaviours (Catchpole & Slater 2008). Moreover, since singing is physically challenging, higher quality males, or those occupying higher quality territories, should sing songs that are closer to a physiological performance limit (Nowicki & Searcy 2005; Liu et al. 2017). This can lead to a reproductive advantage. In the song sparrow, Melospiza melodia, for instance, high-performance songs (a measure of frequency bandwidth relative to trill rate; (Podos et al. 2016) induce more copulation solicitation displays from females (Ballentine et al. 2004). Males with larger vocal repertoires also maintain larger territories and have greater annual and lifetime reproductive success (Hiebert et al. 1989). In such cases, monitoring song traits may help define the environmental factors that support high quality males and reproductive success. Equally, the occurrence of males with poor fitness may be vocally indicated, thus highlighting poorer population outcomes (McGregor et al. 2000).

In mammals, body size is a sexual trait that is commonly signalled in males’ reproductive vocalisations, and many studies highlight females’ preference for larger males; thus, vocalisations that signal body size may indicate fitness. Acoustic allometry shows that larger animals produce lower-pitched calls, since vocal production is highly constrained by the size of the vocal organs and vocal tract (Fletcher 2004; Bowling et al. 2017). Information about body size (among other factors) may be contained in the fundamental (i.e. lowest) frequency of the vibration generated in the larynx (the source) or in the formant frequencies (amplified frequency bands adjacent to the fundamental frequency) generated via resonance filtering of the fundamental frequency through the vocal tract (the filter) (for a review of source-filter theory in mammals see Taylor and Reby (2010). In playback experiments, female koalas, Phascolarctos cinereus, and red deer, Cervus elaphus, prefer male vocalisations that contain formants that indicate a large body size (Charlton et al. 2007; Charlton et al. 2012). In bison, Bison bison, the bellows of larger males contain lower frequency formants and these males secure more copulations (Wyman et al. 2012). Indeed, in

21 many mammals, intense sexual selection for body size has seen the evolution of anatomical novelties that exaggerate body size by lowering formant frequency spacing beyond that predicted allometrically (Charlton & Reby 2016). An example of this is the elongated nose of the male saiga antelope, Saiga tatarica, which effectively extends the vocal tract during roaring, thereby lowering formant frequencies (Frey et al. 2007). The avian syrinx is less anatomically constrained than the multi-purpose (e.g. food processing, respiration) vocal tract of other tetrapods; thus, selection for body size, and exaggeration via tracheal elongation, is likely to manifest in the vocalisations of some birds (Fitch & Hauser 2003). A recent study on banded penguins, Spheniscus sp., found that the ecstatic display song given by males during the breeding season encodes body size in the fundamental frequency (Favaro et al. 2017). Male little blue penguins, Eudyptula minor, produce growl vocalisations that vary significantly in peak frequency with body size (Miyazaki & Waas 2003). Larger males produce eggs and chicks earlier in the season, and their chicks grow faster.

Vocalisations can also provide direct measures of reproductive contexts or events. Many animals vocally announce their readiness to mate, often in conspicuous displays. Like songbirds, male frogs are well known for their advertisement calls during the breeding season, and these can be monitored in relation to local environmental factors (Plenderleith et al. 2017) or latitudinal gradients (Lowe et al. 2016), which may help to predict how a species’ breeding output will be affected by environmental change. Furthermore, vocal activity may indicate reproductive success (Höbel 2017), including how these differ over time or among areas. Conspicuous vocal signals may also accompany copulation events in some species. In African elephants, Loxodonta africana, copulation is signalled by way of the loud and overlapping calls made by many members of the female’s family, for which scientists have termed it ‘the mating pandemonium’ (Payne 2003; Poole 2011). Within a given area, these loud vocalisations could presumably serve to measure copulation events directly. Chimpanzees, Pan troglodytes, most often females, also exhibit copulation vocalisations. These behaviours are given significantly more when a female is with a high- ranking male, and are suppressed when in the presence of high-ranking females (Townsend et al. 2008). However, the number of copulations does not necessarily indicate habitat quality nor recruitment into populations. These assumptions should be tested if copulations are to provide an index of reproductive success or its relationship to habitat factors.

Other vocalisations may provide more direct measures of recruitment. If vocalisations accompany parturition or fledging, success rates could be estimated from monitoring these

22 vocalisations. In killer whales, Orcinus orca, calling behaviour within matriline (mothers and descendants) groups changes markedly after the birth of a calf (Weiß et al. 2006). Additionally, in mammals, infants’ first vocalisations provide a direct measure of birth success. For example, observations of a barbary macaque, Macaca sylvanus, show changes in a female’s moans as labour progresses (increasing in frequency and regularity), and the screaming of the infant following birth (Hammerschmidt & Ansorge 1989). These may act as acoustic signals of recruitment into populations.

In some species, vocalisations may signal the presence of infants or juveniles in groups. Among birds and mammals, it is common for infants and juveniles to learn vocalisations from adults. During learning, the calls of young individuals can be highly variable. For example, in cotton-top tamarins, Saguinus oedipus, the variability in the structure of food-associated vocalisations is higher in infants than in sub-adults and adults (Roush & Snowdon 2001). Infants’ food-associated repertoire decreases over time as they learn which vocalisations are relevant to feeding. Likewise, in leopard seals, Hydrurga leptonyx, juveniles’ vocalisations are less stereotyped than those of adult males, and their repertoires are larger (Rogers 2007). These differences have allowed for the study of age- related differences in spatial distribution, which is otherwise unfeasible with visual surveys (Rogers et al. 2013). The presence of juveniles in a group may also be indicated by changes in the composition or rate of a group’s calls. In the meerkat, Suricata suricatta, the call rate of the adults’ close contact calls decreases when pups are foraging with the group, presumably a response to the loud and continuous begging calls given by the pups (Wyman et al. 2017). Additionally, since frequency is strongly affected by the size of the individual’s vocal anatomy, the presence of young individuals in a group may be detectable by the greater acoustic energy at higher frequencies. This approach predicts the presence of pups in packs of the Iberian wolf, Canis lupus signatus (Palacios et al. 2016). While such a method requires further study in other species, it may prove to be a reliable and simple mechanism by which the presence, and potentially the abundance, of juveniles in a group could be estimated without disturbance from an observer.

Finally, vocalisations can signal not only reproductive success, but also the quality or health of infants or juveniles. Again, these may vary by habitat factors that could be improved through management. The vocalisations of infant birds and mammals, which are often conspicuous and extravagant, may encode signals that vary with individual states, such as hunger. A pertinent example is begging in young birds and mammals, a behaviour that is

23 thought to function largely as signals of an individual’s need or quality (Johnstone & Godfray 2002; Hinde & Godfray 2011). Evidence of the former comes from the many experiments that show hungrier infants to vocalise differently, usually louder or more often (Godfray 1991; Redondo & Castro 1992; Manser et al. 2008; Gladbach et al. 2009; Reers & Jacot 2011; Rector et al. 2014; Klenova 2015). Parents or other adults may respond with greater food provisioning. For example, in cooperatively-breeding meerkats, food-deprived pups beg at greater rates, to which adults respond by increasing food allocation to the hungry pups (Manser et al. 2008). Furthermore, food calls and provisioning vocalisations are known from many birds and mammals (Roush & Snowdon 2001; Evans & Evans 2007; McDonald & Wright 2008), which could provide information on food availability and use (Suzuki & Kutsukake 2017), or even direct measures of feeding events. For example, most cockatoo species (family Cacatuidae) exhibit juvenile food-swallowing vocalisations, given during food transfer (allofeeding) from parent to young (Courtney 1996). In the golden lion tamarin, Leontopithecus rosalia, adults use food-offering calls to encourage juveniles to take prey and later, as juveniles learn foraging skills, to assist with directing juveniles to prey locations (Rapaport 2011). Use of foraging habitat or rate of feeding events may differ among individuals and habitat types (e.g. quality), which may relate to factors in the environment that could be managed.

The honesty of begging signals is hypothesised to be maintained by the associated metabolic and predation costs (Kilner 2001; Rodríguez-Gironés et al. 2001; McDonald et al. 2009; Haff & Magrath 2011; Ibáñez-Álamo et al. 2012; Rector et al. 2014). Indeed, parental alarm calls may suppress infants’ vocalisations, such as in the white-browed scrubwren, Sericornis frontalis (Platzen & Magrath 2004). Alternatively, ‘begging’ may actually be ‘boasting’; that is, high-quality juveniles, who are metabolically more capable, produce conspicuous vocalisations that may facilitate parental provisioning to the strongest young, depending on environmental conditions (Caro et al. 2016; Mock 2016). In either case, differences in the state of juveniles between areas can help mangers decide where actions to improve breeding outcomes are needed, such as by increasing the extent or quality of food resources near breeding sites, or evaluate the success of management interventions.

Alarm and defence

The vocalisations associated with alarm and defence may offer conspicuous vocal signals that can be easily detected in bioacoustic studies. A novel approach that can benefit

24 conservation monitoring is to use prey species’ alarm calls as a proxy of predator abundance. For example, in Australia, sugar gliders, Petaurus breviceps, are a major nest predator of the critically endangered swift parrot, Lathamus discolor (Stojanovic et al. 2014). In a recent study, the occupancy of sugar gliders in the swift parrot’s breeding habitat was estimated by eliciting gliders’ alarm calls from playback of the calls of southern boobook owls, Ninox novaeseelandiae, which depredate the gliders (Allen et al. 2018). Similarly, in South Africa’s Kalahari, the decline of the white-backed vulture, Gyps africanus, over 17 years was estimated from the occurrence of alarm calls that sympatric meerkats give upon sighting a vulture (Thorley & Clutton-Brock 2017).

Indeed, for some species the alarm call can be highly specific to the threat, thereby improving the resolution of threat-specific data, in which case the calls are termed ‘functionally referential’. By definition, functionally referential calls meet two criteria: 1. they refer to, and possibly convey information about, a specific external threat, and 2. they elicit in receivers antipredator behaviours that can be repeated in isolation (Macedonia & Evans 1993; Townsend & Manser 2013; Suzuki 2016; Smith 2017). In early seminal studies, Seyfarth et al. (1980a, 1980b) showed that vervet monkeys, Chlorocebus aethiops, produce unique alarm calls for eagles, leopards and snakes, to which receivers respond with unique antipredator behaviours: monkeys look up or run for cover in response to ‘eagle’ alarm calls, run up trees in response to ‘leopard’ alarm calls, and look down or approach the predator in response to ‘snake’ alarm calls. Urgency of the threat may also be signalled, as in the graded alarm calls of the meerkat (Manser 2001). Functionally referential alarm calls have been reported for several other primate and non-primate mammals (Townsend & Manser 2013), as well as birds from several families (Gill & Bierema 2013; Smith 2017). It is possible, therefore, that for some species alarm calls could provide an index of the costs of specific threats, such as the frequency of fleeing for cover from predators. These costs may be substantial, for instance, if they interfere with important behaviours like foraging. In the blue tit, Cyanistes caeruleus, the presence of model sparrowhawk incited a decrease in foraging and an increase in vocalisation and wing-flicking (Carlson et al. 2017), all of which would incur an energetic cost.

Alarm calling can also indicate aggression, which itself may be a threat to some species. For instance, noisy miners, Manorina melanocephala, produce functionally referential alarm calls for airborne and non-airborne predators, which elicit unique behavioural responses from receivers (Cunningham & Magrath 2017; Farrow et al. 2017).

Mobbing (a deterrent behaviour) is an aggressive and vocally-conspicuous response to non- airborne predators and competitors (Cunningham & Magrath 2017), which is suggested as a reason for the negative relationship between bird species richness and miner abundance (Kutt et al. 2012). By monitoring the alarm calls associated with mobbing, bioacoustics could produce a relatively straightforward way to monitor the occurrence of this behaviour. This would help gauge the severity of this threat among habitats that vary in disturbance level or other factors.

Sociality and vocal complexity

Sociality begets complexity, and animals that live in complex societies are thought to exhibit more complex vocal repertoires because they have more social interactions to navigate. Pro-social behaviours, like cooperation, collaboration and reciprocity (Krams et al. 2012) are posited to be facilitated by complex vocal abilities (Freeberg et al. 2012; Freeberg & Krams 2015). Termed the social complexity hypothesis for vocal communication (Freeberg et al. 2012; Kershenbaum & Blumstein 2017), it is supported both in primates and in birds (McComb & Semple 2005; Leighton 2017). In social species, therefore, vocalisations can indicate a greater range of behavioural and demographic contexts, such as group size or composition by age and gender. Depending on the conservation issue being examined, vocal signals may improve the resolution of datasets and predictive models (e.g. population viability analysis). For example, in sperm whales, Physeter macrocephalus, McDonald et al. (2017) calculated group-level behavioural time-budgets according to the ratio of clicks given in social situations versus those given during foraging. Based on previous descriptions of vocal behaviours (Whitehead & Weilgart 1991), differences in this ratio between sites is suggested to indicate demographic differences, while temporal changes likely reveal differences in food availability. Similarly, in African savanna elephants, the occurrence of three call types (single-voice, clustered and high-frequency) increases significantly with the complexity of group composition, from bull-only groups, to single family groups, to multi- family groups (Payne et al. 2003).

Another important aspect of social behaviour, as it relates to conservation monitoring, is vocal convergence, which is where associated individuals share one or more call types that are distinct from other conspecifics. These ‘affiliative’ group-level vocal signatures are present in many social species (Tyack 2008) and can indicate the segregation of species into distinct social groups or subpopulations. These may reflect variations in a species’ cultural

26 diversity, including reproductive boundaries, which could imply a need for management at sub-specific levels. For instance, the three sympatric but culturally- and acoustically-distinct killer whale, Orcinus orca, groups (ecotypes) that exist in British Colombia are treated separately for conservation purposes (Yurk 2005; CMS 2017). Likewise, acoustic distinctions are genetically supported in subpopulations of Bigg’s (‘transient’ ecotype) killer whale in western Alaska (Sharpe et al. 2017). While vocal convergence may or may not reflect genetic boundaries (Bradbury 2003), regional or geographical dialects can indicate the degree to which groups interact, and therefore inform management. For example, contact calls of the palm cockatoo, Probosciger aterrimus, vary geographically, and the most distinct contact call occurs in a population thought have undergone long-term isolation (Keighley et al. 2017). Thus, the management of this population should be considered separately. For the humpback whales, Megaptera novaeangliae, of Oceania, differences in males’ songs reveal that some populations (vocal clusters) form part of a metapopulation with some, albeit infrequent, interaction, while the east Australian whales are vocal and social outliers (Garland et al. 2015). Thus, vocal convergence may help to explain not only whether groups are socially distinct but also if, and how often, they interact. Moreover, the importance of preserving culture (socially-learned behaviours) within groups of social species, termed ‘culturally significant units’, is increasingly recognised (Whitehead et al. 2004; Ryan 2006; Kühl et al. 2019). Culture can govern the transfer of critical knowledge among individuals, the loss of which may perpetuate extinction risk faster than population size or genetic diversity predicts (Whitehead 2010). For behaviours that are vocally facilitated, bioacoustics may provide an important tool to monitor cultural maintenance or erosion and the relationships to human impacts or other environmental factors (Kühl et al. 2019).

Challenges and considerations for bioacoustic monitoring programs

The diversity of contexts indicated by animal vocalisations increases bioacoustics’ application to conservation monitoring. Much of bioacoustics’ appeal lies in its ability to increase the scale of monitoring and to locate vocal but cryptic species (Browning et al. 2017; Sugai et al. 2019). Behavioural information can also be beneficial, but its potential remains underexplored. Where bioacoustics is used for monitoring, practitioners should consider the relevance of behavioural information available from the sound data being collected. A basic approach may classify calls into broad behavioural (e.g. alarm call) or demographic (e.g. juvenile contact call) categories, while other approaches may apply detailed behavioural knowledge to describe animals’ use of the landscape (e.g. variation among vocally-distinct

27 social groups). Combined with other ecological metrics like soundscape complexity (Borker et al. 2020) or environmental events (e.g. watering; Linke & Deretic 2020), vocal behaviour can provide a rich source of data on how species use their environments and respond to conservation actions. Multitaxa bioacoustic monitoring, though in its infancy, offers even greater potential (Sugai et al. 2019).

The increasing accessibility of bioacoustic technology risks providing massive volumes of sound data that do not align with a program’s aims. To improve manageability, aims should be made explicit before a monitoring regime is established. Subsequently, the relevant vocal behaviours can be identified. An important practical consideration is the potential impact of ambient noise in the system being examined because processing sound data is usually more difficult where the target vocalisations occur in the same bandwidth as anthropogenic noise or sympatric species’ vocalisations. Where possible, sampling should be designed to avoid acoustic masking. In this capacity, the position and detection range of the sound recorders used is important to consider before deployment. Additionally, the frequency with which the recorders will need to be serviced for power and data storage should be estimated. For example, if a program aims to measure nest survival in a bird species, practitioners should consider whether loud vocalisations, as opposed to soft close-range calls (e.g., between parents and offspring in the nest), provide sufficient data to achieve the program’s aims. If so, sound recorders may be placed at greater distances from the nests, which may improve practicality, and a lower sampling rate, which conserves memory and power, could be used. Efficiency can be further improved if these vocalisations can be temporally stratified or subsampled, either in the recording schedule used (i.e., time recorded per day, if the behaviours are thereby representative) or post-hoc during data processing (Wimmer et al. 2013). Informed by explicit aims, monitoring programs should be designed to record appropriate vocal behaviours while maximising efficiency.

In planning bioacoustic monitoring, practitioners should consider the effort required to collect and process sound data relative to other methods with human observers. Since bioacoustic monitoring typically generates very large volumes of sound data, algorithms to detect vocalisations from sound files, termed call recognisers, are critical to the success of bioacoustics as a wildlife monitoring tool. The reliability of recognisers, however, has been mixed, performing well for some species and poorly for others. An ongoing challenge is defining standardised performance metrics; at present, descriptions of performance are

28 somewhat arbitrary and often subjective. The term accuracy, in particular, is inconsistently used (Knight et al. 2017). To help address this issue, Knight et al. (2017) recommend several metrics to describe recogniser performance, based on rates of true positive detections (detections of the target species), false positive detections (detections of non-target species) and false negative detections (missed detections of the target species). These metrics are precision (true positives / true positives + false positives), recall (true positives / true positives + false negatives), F-score ((β2 + 1) * precision * recall / β2 * precision + recall, where β is user-defined metric that defines the relative priority of recall and precision) and area under the curve (AUC). Adopting these recommendations would allow for recognisers to be more easily compared, although this should be done in context of the recogniser’s objective. For example, recall may not be important for programs in which it is unnecessary to detect all occurrences of a species’ call.

Recogniser performance depends in part on extraneous sources of sound (e.g. other species’ calls) and the overall noisiness of the environment (e.g. anthropogenic noise, wind, rain), as well as the acoustic structure of the vocalisations, which is important in the choice of algorithm (Brandes 2008; Towsey et al. 2012; Cragg et al. 2015; Salamon et al. 2016; Priyadarshani et al. 2018). For instance, Towsey et al. (2012) were able to successfully detect the characteristic ‘whip-crack’ of the eastern whipbird, Psophodes olivaceus (100% recall; 67% precision; 82% accuracy) using syntactic pattern recognition. Accuracy was defined here as the proportion of sound files correctly classified. For the frequency-modulated whistle of the currawong, Strepera graculina, they used hidden Markov models (40% recall; 100% precision; 90% accuracy), while for the pulsatile bellows of male koalas, they used binary template matching (75% recall; 75% precision; 95% accuracy). Recogniser performance is also affected by the distance of the calling individual from the sound recorder, given loss of amplitude as well as attenuation of high frequency sound components (Digby et al. 2013; Heinicke et al. 2015; Sebastián-González et al. 2015). Species that call in the infrasonic (very low frequency) and ultrasonic (very high frequency) ranges should have less acoustic overlap with other species, which may assist recognition. Indeed, Zeppelzauer et al. (2015) reported an accuracy of 88.2% and a false positive rate of 13.7% for detecting the low- frequency rumbles of the African elephant. Their method involved signal enhancement to reduce the masking effects of low frequency noise, particularly wind and rain, which can be problematic for detecting infrasonic vocalisations. Ultrasonic vocalisations are less subject to

29 the masking effects of noise, however reviews of commercially-available bat call detection software suggest poor reliability (Lemen et al. 2015; Russo & Voigt 2016).

Importantly, recogniser development must consider the vocal variability within and between individuals, social groups and populations. If the species being monitored has a large repertoire, researchers or practitioners using bioacoustic methods should determine which vocalisations are of most use considering the program’s aims, and tailor the recogniser towards these. If vocalisations are highly variable, training data must properly represent this variability, such as by using calls from several individuals or groups. This includes the often- variable calls of juveniles (Priyadarshani et al. 2018). Notwithstanding, another challenge may be vocal instability, whereby individuals’ vocalisations vary over time, either through drift or vocal learning. The latter is particularly problematic for species that maintain the ability to learn vocalisations throughout life (open-ended vocal learning). For example, the vocalisations of palm cockatoos, while individually-distinct, are thought to change over time (Zdenek et al. 2018). Indeed, lifelong vocal learning is common in parrots (Bradbury & Balsby 2016), and this should be considered in defining how similar a call must be to the recogniser for it to be detected.

In this regard, recogniser development is inherently about trade-offs; a recogniser that is highly specific to the target species’ calls (high specificity; most detections are true detections of the target species) may return few false positive detections but more false negatives, while a less specific recogniser will return more detections, but many may be false positives (low specificity; detections are more likely to be false positive detections of non- target species). For conservation programs aiming to detect rare or cryptic species or behaviours, specificity should be low. In such cases, manual verification will most likely be necessary, the time commitments for which must be considered (Cragg et al. 2015; Rocha et al. 2015). Conversely, for programs in which detecting every call is not necessary, or where the cessation of a vocal behaviour is of interest (e.g. to indicate the end of breeding), then reducing false positives by increasing specificity will be more important. For this reason, recognisers should ideally be built to align with the project’s aims (Priyadarshani et al. 2018), or should be easily altered as needs be. An example of this is the new template-matching R package monitoR (Katz et al. 2016b; Hafner & Katz 2017) which allows the detection threshold (specificity) to be set by the end-user. Moreover, packages like monitoR, as well as other commercial software, assist enormously in making recogniser development accessible

30 to the non-expert (i.e. people without expert programming skills), which is crucial if bioacoustics is to become widely applied in threatened species monitoring (Sebastián- González et al. 2015; Priyadarshani et al. 2018).

Conclusion

The discipline of conservation behaviour is still young, having only gained momentum over the last two decades. Except for the effects of anthropogenic noise (Shannon et al. 2016), monitoring of animal vocalisations has rarely been considered for its conservation implications, much less applied to solutions. Given the functions and behavioural contexts indicated by many vocalisations, however, we argue that bioacoustics can aid conservation and monitoring by providing an alternate means to capture individual- and group-level data. In particular, if vocal behaviours can inform decision-making as it relates to habitat management, then the potential for conservation achievement is substantial.

The specific acoustic signals monitored will vary according to the species and the problem at hand. For many threatened species, for whom breeding success is often compromised, this requires that vocalisations reveal the environmental factors that influence reproduction. For some, sufficient information may be gathered by monitoring discrete reproductive events, such as copulations or births. For others, more complex programs that monitor populations’ social interactions, distress, movements or culture may be more informative. Before a monitoring program commences, however, it is important that a species’ vocal behaviours are understood for the conservation-relevant data that they can provide. Unfortunately, for some less-studied species, this level of detail does not exist. In these cases, managers must decide whether to invest in acquiring that knowledge, or whether the mere occurrence of certain vocalisations that are already described, or exist in publicly- available repositories (e.g. Macaulay Library, Cornell Lab of Ornithology), can be sufficiently informative (Table A2.1). Nonetheless, even if a species’ repertoire is not fully described, bioacoustic programs can provide such information while monitoring is underway. Programs can later be adjusted, or acoustic data re-analysed, to obtain additional behavioural information, if necessary.

Bioacoustics remains limited by the methods and technologies available to handle and process sound data. There are several dimensions to this problem and addressing them, we believe, will be critical to the widespread uptake and applicability of bioacoustics to conservation monitoring. Firstly, building call-classification recognisers often requires expert

31 programming skills, which may limit the use of bioacoustics by ecologists or managers. To address this, collaboration with computer scientists and programmers should be encouraged. Commercial software that provides a user-friendly platform for non-experts to build recognisers must be used with caution (Russo & Voigt 2016) and its limitations clearly articulated. Secondly, the computing power required to process sound files must be carefully considered. Bioacoustics can obtain excessive volumes of data, the processing of which can limit its efficiency. Lastly, and most importantly, any reporting of bioacoustic monitoring where a recogniser is used should include information on how the recogniser was constructed (i.e. the algorithm used; input data used) and statistics on recogniser performance (Towsey et al. 2012; Knight et al. 2017). This can be done at various levels, such as the individual call, the sound file, or the date of recording, depending on the program’s aims. For example, a false negative detection could be defined as a species’ presence being missed at a site on a given day and not an individual call being missed. To this end, the definitions of performance statistics, which have been used inconsistently in the literature, should also be provided.

In this paper, we suggest that the overlap of three disciplines – vocal behaviour, conservation behaviour, and bioacoustics – can benefit threatened species monitoring. In essence, bioacoustics can help to define and monitor the behaviours that enable a species’ persistence and recovery, and the environments in which these are supported. With a clear conservation direction, monitoring programs should consider the data that species’ vocal behaviours can provide, as well as the relative costs of recording and analysing them. Considering the variety of behaviours that are indicated by vocalisations, and their often pivotal role in fitness, the potential for bioacoustic monitoring of behaviour to support conservation is likely to be substantial.

Acknowledgements

We thank Paul Roe and the members of the Ecosounds Lab at the Queensland University of Technology for their insightful discussions about bioacoustics and animal call classification. We acknowledge the Macaulay Library at the Cornell Lab of Ornithology.

Example application

Potential Challenges of benefit of IUCN traditional Taxa Species Vocal behaviours Data Reference bioacoustic status monitoring monitoring Gladiator frog, Relative importance of Amphibian Hypsiboas LC Male chorusing environmental and social cues to Höbel (2017) Individuals are rosenbergi reproduction. difficult to Improved Sperm whale, Social and foraging time budget observe or Social clicks and McDonald et al. detection of Mammal Physeter VU (ratio of hours per day). Spatial habitat is foraging clicks (2017) individuals macrocephalus and temporal variation. difficult to White-bellied access Variation in breeding activity Bird heron, CR Breeding calls Dema et al. (2018) among nesting sites. Ardea insignis Individually-distinct Variation in aggression towards Behaviour of Improved Olive frog, Babina male advertisement Amphibian LC neighbours and strangers (dear Chuang et al. (2017) interest is ability to adenopleura calls and aggression enemy effect). difficult to monitor cryptic calls observe or or rare African elephant, Translocation-induced stress measure behaviours Mammal Loxodonta VU Rumbles (validated from faecal Viljoen et al. (2015) africana glucocorticoid metabolite levels).

Diel changes in behaviour Forest elephant, Mammal NA Rumbles relative to seismic oil Wrege et al. (2010) Loxodonta cyclotis exploration. Killer whale, Call use within Mammal DD Birth of calves. Weiss et al. (2006) Orcinus orca matrilines Long-distance pant- Chimpanzee, Changes in ranging behaviour Mammal EN hoot calls and non- Kalan et al. (2016) Pan troglodytes (centre of activity). vocal drumming Alarm calls of White-backed meerkats, Suricata Thorley & Clutton- Bird vulture, Gyps CR Population trend through time. suricatta, given when Brock (2017) africanus vultures are sighted Richmond Range Breeding phenology through High resource Improved mountain frog, Male advertisement Amphibian EN time and relationship with abiotic Willacy et al. (2015) requirements efficiency, cost- Philoria calls factors. (e.g. human effectiveness or richmondensis hours, cost) to coverage Australasian Number of calling males. More obtain sufficient (spatial or Bird bittern, Botaurus EN Male boom calls cost-effective than human Williams et al. (2018) data temporal) poiciloptilus observers. Southern yellow- Spatial variation in occupancy. cheeked crested More efficient than human Mammal gibbon, EN Morning calls Vu and Tran (2019) observers in habitats with low Nomascus gibbon density. gabriellae

Abundance estimate. Relative to point-counts, acoustic methods Rock ptarmigan, Individually-distinct Marin-Cudraz et al. Observer bias Reduced bias Bird LC reduced observer bias and Lagopus muta male breeding songs (2019) double-counting of individual birds. Number of males at lek sites, Reduced Western estimated from call rate per site. Disturbance by disturbance to Bird Capercaillie, LC Male lek calls Comparison of bioacoustic and Abrahams (2019) observers sensitive Tetrao urogallus observer data suggests human species disturbance may affect counts.

Table 2.2: Summary of vertebrate vocal behaviours that are relevant to conservation.

Reproduction and recruitment Alarm and defence Sociality and vocal complexity

Behaviours • Vocalisations for mate attraction. • Conspicuous vocalising to warn • Group behaviours and contexts e.g. represented • Complex calls or songs. conspecifics of predators. foraging, fission-fusion, demographic • Reproductive events (e.g. copulation, • Aggression. composition. birth). • Vocal convergence within and between • Infant and juvenile calling. groups. • Infant and juvenile begging. • Cultural transmission of critical behaviours. Relevance to • Define the habitats or other environmental • Monitor predator presence or • Improve resolution of demographic data, conservation factors that support high-quality males and abundance via prey species’ alarm such as age and gender composition of reproductive success. calls. groups. • Provide direct measure of reproductive • Provide an index of the cost of defence • Improve parameterisation of predictive events. e.g. reduction in foraging. models. • Signal the presence of juveniles in groups • Measure the level of threat from • Determine if, and how often, groups by way of the higher-pitched vocalisations aggressive competitor species. interact. Inform whether groups should be of smaller-bodied individuals. • Spatial and temporal variations. treated individually or collectively for • Quality or health of juveniles. management. • Spatial and temporal variations. • Spatial and temporal variations. Key principles • Sexual selection for call complexity: • Functionally referential alarm calls: • Social complexity hypothesis for vocal females prefer complex calls or songs, as calls refer to, and possibly convey communication: animals that live in they indicate high-quality males. information about, a specific external complex societies exhibit complex vocal threat and they elicit in receivers repertoires.

• Acoustic allometry: vocalisations vary by antipredator behaviours that can be • Vocal convergence: associated body size. repeated in isolation. individuals share calls that are distinct • Sexual selection for body size: females from other social groups. May manifest as prefer larger males. regional or geographical dialects. • Infant and juvenile begging: individuals • Culturally significant units. beg more when hungry or if they are of higher metabolic quality. Key references • Catchpole & Slater 2008 • Seyfarth et al. 1980a, b • Freeberg et al. 2012 • Fletcher 2004 • Smith 2017 • Kershenbaum & Blumstein 2017 • Charlton & Reby 2016 • Townsend & Manser 2013 • Tyack 2008 • Godfray 1991 • Macedonia & Evans 1993 • CMS 2017 • Johnstone & Godfray 2002 • Ryan 2006 • Mock 2016 • Whitehead et al. 2004

Chapter 3 Author Contributions

Chapter 3 is presented as published in Austral Ecology, with some amendments to accommodate suggestions from the examiners of this thesis. The order of authors is Daniella Teixeira, Richard Hill, Michael Barth, Martine Maron and Berndt van Rensburg. Daniella Teixeira designed the monitoring program, conducted fieldwork, analysed the data and wrote the manuscript. Richard Hill and Michael Barth helped with fieldwork. Berndt van Rensburg and Martine Maron supervised the project’s design and implementation, and reviewed drafts of the manuscripts.

Chapter 3 | Nest-associated vocal behaviours of the south-eastern red-tailed black- cockatoo, Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black- cockatoo, C. lathami halmaturinus

Abstract

Animal vocalisations can signify diverse behavioural contexts, knowledge of which can be applied in bioacoustic monitoring programs. Australia’s endemic black-cockatoos (Calyptorhynchus sp., family Cacatuidae) are highly vocal species that are threatened in many locations. In this study, we describe the nest-associated vocal behaviours of two endangered subspecies of black-cockatoo, the south-eastern red-tailed black-cockatoo, C. banksii graptogyne, and the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus. Breeding success is limiting their recoveries and nest monitoring is challenging, but vocal recordings might provide valuable long-term information hard to obtain otherwise. We recorded daily vocal activity at wild nests of both cockatoos using autonomous sound recorders. Combined with behavioural observations and video footage, we identified vocalisations characteristic of six behavioural contexts at nests: birds in flight, while perched, during begging (adult females), during courtship displays (adult males), when entering or sitting near to the nest hollow entrance (adult females), and from nestlings. Linear discriminant analysis on 12 acoustic measurements correctly classified 58.4% of calls of the red-tailed black-cockatoo (n = 907 calls from eight nests) and 62.9% of calls of the glossy black-cockatoo (n = 1,632 calls from 11 nests). In both subspecies, the female nest call and nestling calls are the most conspicuous vocal indicators of active nesting, and therefore should be considered for their bioacoustic potential. Other adult vocalisations indicate a range of behavioural contexts that could be informative for monitoring nesting behaviour, and its association to habitat features, in these endangered subspecies.

Introduction

Animal vocalisations can signify diverse behavioural states. For vocal species, sound data collected in bioacoustic studies can therefore indicate particular behavioural contexts, which can benefit conservation if they provide new insights into a population’s state, trajectory or response to management (Chapter 2). For example, critical life history events like mating and recruitment may be detected via context-specific vocalisations (e.g. African elephant copulation; Poole 2011) or changes in group-level vocalisations (e.g. a shift in acoustic energy of Iberian Wolf packs when juveniles are present; Palacios et al. 2016).

Therefore, knowledge on the behavioural contexts associated with particular vocalisations can improve the resolution and application of bioacoustic data beyond population metrics such as species presence-absence, population density or abundance. As technology advances, the efficiency and cost-effectiveness of acquiring data from species’ vocalisations is likely to improve, warranting improved incorporation of vocal behaviour into bioacoustic monitoring programs. To achieve this, the vocal behaviours that could provide conservation-relevant data must be described for the species of interest.

Parrots are among the most social, intelligent and vocally complex avian species (Bradbury & Balsby 2016; Cussen 2017). They are one of few taxa able to learn new vocalisations throughout life (open-ended vocal learning), an ability hypothesised to be an adaptation to their highly social foraging behaviours (Bradbury & Balsby 2016). Australia’s endemic black-cockatoos, comprising five species of genus Calyptorhynchus (family Cacatuidae), are highly social parrots, often found in large, noisy flocks (Higgins 1999). Nesting is semi-communal, with nests often aggregated in the landscape (Johnstone et al. 2013; DT, pers. obs.). Clutches comprise one or two eggs, and overall reproductive output is low. Eggs hatch after about four weeks’ incubation and fledging occurs ten to twelve weeks later (Higgins 1999; Johnstone & Kirkby 2008). Fledgling cockatoos have a long parental dependency period (Higgins 1999), possibly up to 24 months in some populations (forest red- tailed black-cockatoo, C. banksii naso; Johnstone et al. 2013). As in other parrots, young cockatoos likely learn their early vocalisations from their parents, with whom they have contact at the nest, and develop their adult repertoire from social interactions in larger flocks during the fledgling period.

Given their sociality, it is likely that black-cockatoos’ different vocalisations reflect different behavioural contexts. This is shown in other cockatoos, including the closely-related Carnaby’s black-cockatoo, C. latirostris (Saunders 1983), and the palm cockatoo, Probosciger aterrimus (Zdenek et al. 2015). As such, black-cockatoos are good candidates for bioacoustic monitoring of behaviour, with potential benefits for conservation. Indeed, every black-cockatoo species is listed as threatened under state or national legislation in at least part of its range, and bioacoustics could aid monitoring. Currently, monitoring and management vary among species and populations, but often rely on citizen science activities coordinated by non-profit organisations (e.g. Birdlife Australia’s Great Cocky Count). These activities usually involve counting birds in flocks, such as at roosts or drinking sites. Such

40 data can be useful for understanding trends in flock size, demographic structure and occupancy in the landscape. However, methods for collecting data from other contexts, particularly during the breeding season, are limited. If behaviour-specific vocalisations can be reliably identified, then obtaining data from other contexts, such as nesting, should be achievable using bioacoustic methods.

In this study, we investigated the nest-associated vocalisations of two endangered subspecies of black-cockatoo, the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne. We focussed specifically on vocal behaviours associated with nesting because efforts to monitor breeding in these subspecies have been restricted, despite their recoveries being limited by breeding success (Russell et al. 2018; Berris et al. 2018). The Kangaroo Island glossy black- cockatoo’s population size is the smallest of any black-cockatoo (373 individuals counted in the 2016 census; Berris et al. 2018). Until 2017, when the recovery program’s funding was dramatically reduced, up to 50 nests were monitored each breeding season. However, human resource requirements were substantial, and the funding reductions have seen monitoring efforts greatly reduced. For the south-eastern red-tailed black-cockatoo, whose declining population numbers about 1400 individuals, monitoring of breeding has always been limited (Russell et al. 2018). Nests are difficult to monitor because they are remote, rare in the landscape and are often on private land, making it challenging to observe nests through to fledging or failure. Breeding success is inferred from the demographic structure of flocks, and increases in the proportion of male birds over recent years suggests a decrease in breeding output (Russell et al. 2018). For this reason, a high priority for conservation is to develop efficient methods for nest monitoring.

This study aimed to describe the nest-associated vocalisations of the south-eastern red-tailed black-cockatoo and the Kangaroo Island glossy black-cockatoo. For these subspecies, the only adult vocalisation reported from nests is the nest call of the Kangaroo Island glossy black-cockatoo, given by adult females when entering or prospecting a nest hollow (Pepper 1996). Vocalisations of nestlings are not formally described in either red- tailed or glossy black-cockatoos, although Cameron (2009) in studying the eastern subspecies of glossy black-cockatoo, C. lathami lathami, defined late-stage nests as those where nestlings gave “harsh growling” calls upon the return of the nesting female. Here, for both subspecies, we aimed to qualitatively describe the behavioural contexts associated with each

41 call type given at nests and to provide quantitative acoustic measurements for each. We hypothesised that adult and nestling birds give unique vocalisations in various behavioural contexts, and that these are distinct in acoustic structure. This knowledge may reveal important information about critical life history events and, more specifically, the potential for bioacoustics to provide a novel method with which to monitor nesting behaviour.

Methods

Study sites

This study took place from 2016 – 2019. Data for the south-eastern red-tailed black- cockatoo were collected from areas near Casterton and Edenhope in south-west Victoria, Australia. The cockatoos commonly nest in dead and isolated river red gums, Eucalyptus camaldulensis, that occur in livestock paddocks (Fig. A3.1). Given their relative ease of observation and accessibility, paddock trees were the focus of data collection for this study. Nests were located through active searching in spring and summer. Typically, cockatoos in flight were detected by their calls and then followed to identify if they approached a nest hollow. This method allowed active nests to be confirmed without the need for tree-climbing, which is normally unsafe as the trees are dead. A nest was deemed active if a female emerged from a hollow upon a male’s flight calls, which indicates incubation or brooding, or remained in a hollow after sunset (no roosting behaviour observed). Some nests were inspected with a pole-mounted camera at later stages of nesting, at times of day when the female was not present, to confirm their continued activity. Nest height varied, but all were at least 10 metres above ground level. Some nests (>15 metres height) were beyond the reach of the pole- mounted camera.

On Kangaroo Island, data for the glossy black-cockatoo were collected from several nesting areas that are routinely monitored by the state government. To increase the cockatoos’ breeding opportunities, many artificial nest hollows have been installed on the island, and these have been successfully used by the cockatoos for many years (Berris et al. 2018) (Fig. A3.1). Nesting occurs in several habitat types, including conservation estates, roadsides, regenerated woodlands, and residential and agricultural areas (MB, pers. obs.), all of which are represented in this study. No preference was given to either natural or artificial hollows for this study. Nest activity was confirmed via afternoon observations, as in the red- tailed black-cockatoo, via nest inspection with a pole-mounted wireless camera, or via a

42 female’s presence at the nest hollow entrance. Laying is thought to peak in March and April (MB, pers. obs.). For most nests, monitoring began during these months.

Acoustic data collection

For each active nest, an autonomous sound recorder (Frontier Labs Bioacoustic Audio Recorder, https://frontierlabs.com.au/) was installed on the nest tree or a nearby tree. On the nest tree, recorders were installed at approximately 2 metres height. We chose to install recorders at this height, rather than at the hollow, because this is what would be more feasible in a bioacoustic monitoring program. If not on the nest tree, recorders were within 5 metres of the nest tree to ensure that it was closer to the nest of interest than to other nests; therefore, the loudest vocalisations could be confidently assigned to the nest of interest. Recorders were approximately 8 – 30 metres in linear distance from the nest hollow, depending on whether they were fixed to the nest tree or a neighbouring tree, and the height of the hollow in the nest tree. During initial field observations, we determined that cockatoo calls are loud, and clearly audible to human observers at 10 – 20 metres distance, for conspicuous behaviours. Therefore, we considered that the detection space of the sound recorders, though not experimentally tested, was appropriate.

Installed sound recorders were programmed to record for three hours per day, beginning at 2.5 hours before sunset (sunset-based schedule). This time period was chosen because this is when the cockatoos are most active at the nest (RH, MB, DT, pers. obs.). While longer recording periods would capture behaviours outside of this time period, battery limitations currently preclude this as a feasible approach to remote monitoring, although solar-powered options could be used. In addition, once a week, recording began half an hour before sunrise and ended half an hour after sunset (full-day schedule). This was done to capture any unexpected activity at the nest during the day. If a nestling was observed at a nest (i.e. a late-stage nest), we attempted to update schedules to record at the full-day schedule every day, to maximise the likelihood of recording the fledging event. This was not achieved for every nest. Recorders remained in place until after the observed or expected date of fledging, unless nest failure was confirmed sooner. All recordings were made using an omnidirectional microphone, with a fixed gain of 20 dB, at a sample rate of 44.1 kHz. Microphones had an 80 Hz high-pass filter to reduce the effects of low frequency noise (e.g. wind). Recorders were fitted with four rechargeable lithium ion batteries and one 128 GB

SanDisk memory card, which were replenished at approximately 6-week intervals. All recordings were made in uncompressed wave (.wav) format.

Behavioural classification

To describe the behavioural contexts of vocalisations recorded, observations of cockatoos were carried out at and near nest trees. This included nests that were monitored with autonomous sound recorders, as well as other nests that were opportunistically observed. Cockatoos were filmed using a Canon 5D mark III DSLR camera and Canon 100-400 II IS USM telephoto lens with a Rode VideoMic Pro microphone attached. Observation distance was usually a minimum of 10 metres from the cockatoos’ location but varied between sessions. Observations were usually made in the late afternoon and early evening when the cockatoos are most active at the nest. Recording usually commenced when the cockatoos were in plain sight and actively vocalising. Individuals were not marked but could be identified by their association with a nest. Observation length varied and depended primarily on the length of time that the birds were within visual and auditory distance. Observations were usually terminated by the cockatoos flying out of sight or by the female entering the nest hollow. Observations were also terminated in poor weather (rain or high wind). We did not standardise observation time, frequency or recording length among nests, since most observations were opportunistic and dependent on recording conditions and nest status. Observation time varied widely, from several minutes to over one hour. Nests were visited randomly, and for some nests no behavioural observations were made. Where possible, preference was given to nests where a nestling was observable at the hollow entrance. Behavioural observations were conducted in both early and late stages of nest development, although nestlings were only observed at late stages (DT, pers. obs.). Nests were visited more frequently at late stages. Where large nestlings were observed, nests were visited every day (except in poor weather) until fledging. There were no obvious differences in behaviour among nests visited, or even between the two species; behavioural data are, therefore, considered to be sufficiently representative.

Call types and their associated behavioural contexts (Table 3.1) were putatively described from field observations and video footage. Sound files were extracted from videos using Adobe Media Encoder CC 2017. Spectrograms (Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap) and waveforms (oscillograms) were inspected using Raven Pro 1.5 (Cornell Lab of Ornithology, Ithaca, New York) and viewed in

44 tandem with videos to determine the behaviours associated with each call type. Behaviours for both subspecies were classified into broad categories in line with the ethogram provided by Pepper (1996) for the Kangaroo Island glossy black-cockatoo. Vocal behaviours were subsequently classified in recordings made by the autonomous sound recorders by visually and aurally identifying calls of the same structure. Most call types were clearly associated with particular behaviours and were easily identifiable in recordings. However, the various calls associated with perching behaviour were less distinct from each other, for which reason we did not attempt to separate them into unique behavioural categories (instead, they were referred to as perch subtype 1, perch subtype 2, etc.). Since the camera records sound in compressed MPEG-4 AVC/H.264 format, spectrograms created from video data were used only for visual purposes when classifying putative call types and behavioural contexts. Quantitative measurements of each call type (Table A3.1) were only taken from recordings made with the autonomous sound recorders.

Quantitative structure of vocalisations

To efficiently select calls for quantitative analysis, we used recordings made in the final two weeks before fledging (or failure), as this time period represents the complete repertoire of conspicuous vocalisations at nests. This includes nestlings, which can be difficult to detect in in earlier weeks (DT, pers. obs.). Most selections were made from nests where nestlings were recorded. Additional nests were included to increase sample sizes of adult calls, if required (see below). Selections were not made on days or during time periods where recording conditions were poor (e.g. high wind or rain).

For the south-eastern red-tailed black-cockatoo, selections were made from recordings obtained at eight nests. For the Kangaroo Island glossy black-cockatoo, selections were made from 11 nests. Calls were manually selected and annotated from spectrograms (Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap) and waveforms using Raven Pro 1.5. For each selection, upper and lower bounds (i.e. low frequency and high frequency of the selection box) were set via inspection of the spectrogram, while start time and end time were set via inspection of the waveform (Fig. A3.2). For each call type, we aimed to select a minimum of 20 calls per nest, to sufficiently represent within-individual variation in call structure (Fischer et al. 2013). For flight calls, we initially aimed to annotate at least 40 calls per nest, because both male and female adult birds give this call. However,

45 this was difficult to achieve for the glossy black-cockatoo because male and female flight calls are often overlapping and therefore unsuitable for acoustic analysis.

Calls that were selected were chosen ad hoc from those that showed relatively high signal to noise ratio on the spectrogram and were not overlapping with other calls or background noise. Each call selected was categorised by call type (behavioural category), age (adult or nestling) and, except for flight calls, sex (adult birds only). Quantitative measurements recorded for each selected call were: (1) low frequency (Hz), (2) peak frequency (kHz), (3) centre frequency (kHz), (4) aggregate entropy (bits), (5) average entropy (bits), (6) minimum entropy (bits), (7) maximum entropy (bits), (8) delta time (seconds), (9) interquartile range duration (seconds), (10) peak amplitude (U), (11) peak frequency contour average slope (Hz/ms), and (12) peak frequency contour maximum slope (Hz/ms) (Table A3.1). We excluded high frequency measurements (e.g., high frequency, delta frequency, interquartile range bandwidth) because high frequency components were often attenuated by distance. This differed among nests because the distance of the sound recorder to the nest hollow varied, as did the distance between the sound recorder and the vocalising birds when not in the nest hollow (i.e., the birds’ position in the nest tree or nearby trees varied).

Statistical analysis

To examine for differences in putative call types (response variable), we conducted linear discriminant analyses on the acoustic measurements (predictors; Table A3.1) using the MASS package in R (Venables & Ripley 2002; R Core Team 2019). Acoustic measurements were inspected for normality and transformed where necessary. For each subspecies, the data (n = 907 selections of calls for the red-tailed black-cockatoo; n = 1,632 selections of calls for the glossy black-cockatoo) were randomly divided into two separate datasets, one as training data (70% of the original dataset), from which the discriminant models were built, and the other as test data (30% of the original dataset). To account for different units of acoustic predictors, test data were centred and scaled using the caret package (Kuhn 2008). Test data were used to classify call types (Table A3.2). To further confirm model performance, each discriminant model was tested using leave-one-out cross validation on the complete dataset. Results were very similar between the two approaches (Table A3.3). Finally, each discriminant model was tested using a Multivariate Analysis of Variance (MANOVA). Plots were made using ggplot2 in R (Wickham 2016).

Results

For the south-eastern red-tailed black-cockatoo, 23 nests from eight locations were located over two breeding seasons. All nests were located on livestock farms, except for one nest that was in an artificial hollow in a plantation of Australian blue gum, E. globulus. In total, four nests were in artificial nest hollows. The remainder were most often in large, dead river red gums, E. camaldulensis. One nest was found in a live river red gum. Nesting occurred from September through March. Nestlings were recorded at nine nests. For the Kangaroo Island glossy black-cockatoo, data were recorded from 28 nests in eight locations over two breeding seasons. 18 nests were in artificial nest hollows. Natural nest hollows were all in live sugar gums, E. cladocalyx. Nesting occurred from March through November. Nestlings were recorded at 15 nests. Observed behaviours at nests were similar for the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo.

South-eastern red-tailed black-cockatoo

For the red-tailed black-cockatoo, we putatively described 11 call types (including subtypes) from six behavioural categories (Table 3.1 and Fig. 3.1). Flight calls were typically loud and harmonic in structure, given by adult birds (male and female) when flying to and from the nest tree (Fig. 3.1a). We considered the take-off call (adult male and female) to be a subtype of the flight call, differing by having more arched frequency components (downward inflection) (Fig. 3.1b). Begging calls given by adult females were highly variable in structure and amplitude. These calls clearly exhibited nonlinear phenomena, including deterministic chaos, frequency jumps and subharmonics, often within a single bout of begging (Figs. 3.1c and 3.1d). Begging calls typically elicited head-bobbing and allofeeding from the adult male. Early in the nesting period, females could often be heard begging from inside the nest hollow in response to the approaching males’ flight calls. Display calls were sometimes given by males in response to females’ begging. Display calls were highly stereotypical and repetitive (Fig. 3.1f) and involved head-bobbing and fanning of the tail feathers. At two nests, display calls were given soon after alarm calling. Perch calls and nestling calls each comprised three subtypes, differentiated by their apparent loudness and harmonic, chaotic or pulsatile structure. Perch calls (three subtypes) were given by adult males when the adult female was near or inside the nest hollow, presumably to maintain contact with the female (Fig. 3.1j-l). Perch calls were usually the final calls given in the day, except for take-off and flight calls as the male went to roost. Females sometimes called when perched soon after flight, but we

47 considered these calls to be flight calls because they were not obviously different in sound or spectrographic structure. Adult females’ nest call (or nest entry call) was highly pulsatile, sometimes resembling nestlings’ calls (Fig. 3.1e). These calls were not detected every day. The most commonly recorded nestling calls were loud and broadband (Fig. 3.1g). These calls usually began as the parents were flying to the nest. These calls were easily heard up to 30 metres from the nest and were distinct on spectrograms. Two subtypes of nestling calls (nestling subtype 2 and subtype 3; Figs. 3.1h and 3.1j) were quieter and given some time after the parents’ arrival at the nest. These softer varieties were more difficult to detect on spectrograms. Vocal behaviours and interactions between individuals at the nest were evident in spectrograms (Fig. 3.3).

Acoustic measurements were obtained from a total of 939 selections of annotated vocalisations, representing 11 call types (including subtypes) from eight nests, for the south- eastern red-tailed black-cockatoo. Descriptive statistics of each call type’s acoustic measurements are provided in Appendix 3 (Table A3.4). The sample size (i.e. number of annotated vocalisations) of each call type varied between n = 5, for nestling subtype 3, and n = 251 for the flight call. Three call subtypes (perch subtype 3, nestling subtype 3, and take- off) had fewer than 20 annotations and were therefore excluded from subsequent analyses. The final dataset used for analyses, therefore, comprised 907 selections of vocalisations. Linear discriminant analysis correctly classified 58.4% of calls (MANOVA: Wilk’s λ = 0.16, F = 22.518, p < 0.001) (Fig. 3.5). LD1 and LD2 explained 68.0% of the overall variance in the model. Accuracy was highest for the display call (73.7%), followed by the flight call (72.0%), and lowest for nestling subtype 2 (28.6%) (Table A3.2). Nestling subtype 2 was most often misclassified as nestling subtype 1 or begging (Table A3.2). Perch subtype 1, with an accuracy of 34.8%, was commonly misclassified as a flight call (Table A3.2).

Kangaroo Island glossy black-cockatoo

We putatively described 14 call types (including subtypes) from six behavioural categories for the Kangaroo Island glossy black-cockatoo (Table 3.1 and Fig. 3.2). Flight calls were typically loud and overlapping, given by the adults when flying to and from the nest (Fig. 3.2a). Begging calls were given by females and usually elicited allofeeding from the male. Begging calls usually contained harmonics (Fig. 3.2b) but were highly variable in structure and showed clear nonlinear phenomena (Fig. 3.2c). Display calls were given by males in the presence of the female. Display calls were highly stereotypical and repetitive

(Fig. 3.2d) and involved head-bobbing and fanning of the tail feathers. Adult females gave nest calls when perched at, or near, the nest hollow entrance, but not on every occasion. Nest calls were highly pulsatile in structure, giving the call a ‘growling’ characteristic (Fig. 3.2e). Perch calls were given by adult males and females when perched on or near the nest tree. They comprised six graded call types (Fig. 3.2i-o). Perch subtypes 2 and 3, given by males, were the most common perch calls observed in the field and in sound recordings (Fig. 3.2j-l). They are usually the final calls recorded on any given day. Female perch calls (subtypes 4, 5 and 6) were less commonly observed (Fig. 3.2m-o). Nestling calls were typically loud, given in response to the parents’ arrival at the nest (subtype 1; Fig. 3.2f). Nestling calls were highly resemblant of the female nest call. Nestling subtype 2 was softer, given when the nestling appeared less stimulated following the arrival of its parents (Fig. 3.2g). Nestling subtype 3 was the softest subtype, given only when the female was close to the nestling (Fig. 3.2h). Vocal behaviours and interactions between individuals at the nest were evident in spectrograms (Fig. 3.4).

Acoustic measurements were obtained from a total of 1,641 selections of annotated vocalisations, representing 14 call types (including subtypes) from 11 nests, for the Kangaroo Island glossy black-cockatoo. Descriptive statistics of each call type’s acoustic measurements are provided in Appendix 3 (Table A3.4). The sample size of call types varied from n = 9, for the take-off call, to n = 303 for nestling subtype 1. Due to low sample size, the take-off call was excluded from further analysis. Therefore, the final dataset used for analyses comprised 1,632 selections. Linear discriminant analysis correctly classified 62.9% of calls (MANOVA: Wilk’s λ = 0.04, F = 47.362, p < 0.001) (Fig. 3.5). LD1 and LD2 explained 81.2% of the overall variance in the model. Nestling subtype 1 had the greatest classification accuracy (87.8%) followed by begging (84.7%) (Table A3.2). Nestling subtype 2 had an accuracy of 73.2% and misclassifications were mostly nestling subtype 1 (Table A3.2). There was misclassification among flight calls and perch calls (Table A3.2). Perch subtypes 1 and 3 had no correct classifications (Table A3.2). Perch subtype 1 was mostly classified as flight and perch subtype 5 was mostly classified as the female nest call (Table A3.2).

Discussion

Bioacoustic sound recordings can provide a rich source of behavioural data for species whose vocal diversity is known (see Chapter 2). Particularly for social species, which usually exhibit more complex repertoires (Freeberg et al. 2012; Leighton 2017), vocalisations

49 can indicate specific behaviours, demographics (e.g. age and sex of the caller) and interactions among individuals. Bioacoustic studies often focus on what vocalisations a species makes; this helps determine presence-absence, a common objective of bioacoustic studies. However, the ability to understand from vocalisations who the signaller is and why they are vocalising (behavioural context) can greatly improve the resolution of data acquired from bioacoustic programs (Chapter 2). Through context-specific vocalisations, bioacoustics could help to monitor species’ behaviours and the relationship to habitat features, and thereby inform conservation decision-making. A necessary first step, then, is to understand a species’ vocal repertoire to the extent required for monitoring or conservation.

In this study, we provide the first descriptions of the diversity of nest-associated vocal behaviours of two endangered subspecies of black-cockatoo, the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne. Through behavioural observations, we found that these subspecies gave distinct vocalisations in each of six behavioural contexts at nests. Specifically, vocalisations were identified from birds in flight, while perched, during begging (females), during courtship displays (males), when entering or sitting near to the nest hollow entrance (females), and from nestlings when in the presence of their parents. This knowledge can be used to develop novel nest monitoring methods using bioacoustic technology. This is important because bioacoustics using remote sound recorders allows for data to be collected at spatial and temporal scales much greater than that feasible by human observers. This offers an advantage for these subspecies as traditional monitoring is limited by human survey effort and available funding. Even where active nests are known, monitoring their subsequent development and outcome (fledging or failure) is difficult or, in many cases, is not achieved. Moreover, bioacoustics could be used to monitor not only known active nests, but also potential nests, such as tree hollows of unknown status, hollows used in previous years, and newly deployed artificial nest hollows.

The female nest call and the nestling calls are the most conspicuous vocal indicators of active nesting in these subspecies. These calls are loud, distinct and are, to the best of our knowledge, the only calls that are unique to active nests (DT, RH, MB, pers. obs.). These calls, therefore, are most relevant to bioacoustic monitoring programs. The female nest call appears to function in close-range communication with the nestling and with the adult male when he is perched on the nest tree. In both subspecies, but especially the Kangaroo Island

50 glossy black-cockatoo, the nest call resembled the nestling call. This may be a product of nestlings learning their calls from the adult female, who is the only parent to enter the nest hollow in these species. Late-stage nestling calls are characteristically loud upon the parents’ arrival to the nest tree (subtype 1) in both subspecies. These loud calls were clear and easily identified in spectrograms of sound recordings. Calls become softer and less-stimulated after the parents’ arrival (subtype 2), but the acoustic structure is otherwise similar. Although discrimination accuracy varied (glossy black-cockatoo: 87.8% and 73.2% for subtypes 1 and 2, respectively; red-tailed black-cockatoo: 54.6% and 28.6% for subtypes 1 and 2, respectively; Table A3.2), nestling calls of both subspecies were distinct to the human ear and unlike other call types, except for some cases of the female’s nest call (DT, pers. obs.). This is of little practical significance, however, since both the nestling call and the female nest call indicate active nesting.

Female begging calls were highly variable within- and between-individuals of both subspecies. Calls showed a range of nonlinear phenomena including deterministic chaos, subharmonics and frequency jumps. The acoustic structure of calls observed on spectrograms often varied substantially within a single begging bout (Figs. 3.1c and 3.2c). Though largely untested in birds, one hypothesis regarding nonlinear sound states that the more variable and random (nonlinear) a call is, the less likely a receiver is able to ignore it (Blumstein & Récapet 2009). That is, nonlinearity in animal communication has possibly evolved to attract and maintain attention to increase fitness. For example, in African elephants, Loxodonta africana, infant roars increase in chaos with the urgency of the situation (Stoeger et al. 2011). In red deer, Cervus elaphus, the harsher the males’ roars, the more attention they receive from potential mates (Reby & Charlton 2012). It is possible, therefore, that the highly nonlinear structure of female begging calls functions to limit habituation by the male and consequently increase his provisioning of the female. Since females are solely responsible for provisioning the nestling, it is plausible that female begging has been subject to strong selection pressures to increase provisioning rates. Likewise, the soft, begging-like nestling call (subtype 3) also appeared to be nonlinear and has possibly evolved to stimulate allofeeding by the female. Moreover, since both subspecies are highly specialised in diet, begging call structure may relate to the availability or quality of food in the habitat near nests and, subsequently, nestling body condition or the likelihood of breeding success. Alternatively, the soft, high-pitched characteristics of the female and nestling begging calls

51 may function to limit detection by nest predators during allofeeding (c.f. loud calls given in other contexts).

The male display call is highly stereotypical and repetitive in both subspecies and involves head-bobbing and tail-fanning. The call contains two elements; the first element is longer in duration and may show harmonics, while the second element is a short, broadband ‘chuck’. Characterised by its rhythm, with each two-element call being repeated over time, the call may function as a signal of male fitness. Recent literature on rhythm in birds, though in its infancy, suggests the possibility of sexual selection for highly rhythmic calls or other sounds. For instance, palm cockatoos, Probosciger aterrimus, are renowned for their drumming behaviour, wherein males use a tool (a stick) to drum on a tree branch or a nest hollow. Drumming is most often directed towards females. Individualised drumming styles, including variations in rhythm, suggest that information about the male may be conveyed to females (Heinsohn et al. 2017). In budgerigars, Melopsittacus undulatus, experiments show that females prefer rhythmic sounds, which may relate to a preference for rhythm (as yet untested) in the head-bobbing sexual display given by males of the species (Hoeschele & Bowling 2016). In the current study, observed display calls were always given by adult males and directed towards their bonded females. Often, females seemingly ignored the display, or lunged to the male, or moved to a different position in the tree (DT, pers. obs.). Copulation sometimes followed the display call and was, in all observations, preceded by it (DT, pers. obs.; copulation not observed in the red-tailed black-cockatoo). Thus, the call appears to be a sexual display by males to elicit copulation, or to reinforce the pair bond, to which females respond variably. Selection for rhythm may be acting on the call, in which case individualised rhythmic features may provide a bioacoustic index of male fitness.

Notwithstanding, the display call appears to have at least two secondary functions. Pepper (1996) noted that, in the Kangaroo Island glossy black-cockatoo, the display call (referred to as the kwee-chuck call) was sometimes given by unpaired juvenile males when perched prominently, suggesting a secondary function in dominance. Pepper (1996) also noted that the call was given after disturbance by human observers. This concords with two opportunistic findings in sound recordings from nests of the south-eastern red-tailed black- cockatoo. In both cases, display calls were given following a period of alarm calling or loud banging sounds in the hollow. Cockatoos were not detected in recordings thereafter, which

52 suggests that display calls may accompany nest failure in some cases (e.g. a predation event). The function of the display call in such a context is not yet clear.

We classified perch calls as any vocalisation that did not resemble another call type and was given by adult birds when perched on the nest tree. These represented a range of graded contact calls, which were sometimes difficult to differentiate on spectrograms, and classification accuracy was mixed (Table A3.2). These results support the hypothesis that animal calls are often graded, variations of each other, rather than distinct categories; thus categorisation is somewhat subjective (Fischer et al. 2016). In parrots, loud contact calls tend to elicit a vocal response from conspecifics and, therefore, are generally thought to function in establishing connections between individuals (Bradbury 2003). This appeared to be the case in this study, where males and females, and sometimes nestlings, would often engage in vocal exchanges while perched on the nest tree (Figs. 3.3 and 3.4). In both subspecies, the males’ soft perch call (referred to as subtype 2) was relatively common and detected most days. This call provided a clear acoustic signal of the birds’ presence at the nest tree. These soft contact calls were given by males after landing on the nest tree, after the female had entered the hollow, as well as in response to female begging. Therefore, its function appeared mostly one-way, directed from male to female, usually without response. Parrots’ soft contact calls often do not elicit responses and are thought to function in coordinating flock movements through vegetation. Indeed, in the glossy black-cockatoo, this soft perch call appears synonymous with the feeding call shown in Pepper (1996) wherein it was noted “mated pairs gave soft, short calls at intervals while foraging”. Like the display call, perch calls appear to function in several behavioural contexts.

Vocal behaviours in this study were described from bioacoustic methods that align with those likely to be feasible in a larger monitoring program for these subspecies. Specifically, sound data were collected from nest trees, usually in the late afternoon as this is when birds are most active at nests (DT, pers. obs.), from approximately 8 – 30 metres distance from the nest hollow. Tree-climbing was avoided as it is unsafe for nests in dead trees, which are important for the south-eastern red-tailed black-cockatoos, and has high human resource costs. Vocalisations were clearer from recorders that were closer to the nest hollows (e.g. where hollows were lower to the ground), although loud calls, including nestling calls, were easily identified in recordings from all distances included in this study. Since loud nestling calls are one of the most useful indicators of active nesting, the approach

53 used here should be appropriate for these subspecies. A limitation is that this requires a sound recorder at every nest tree monitored. However, since the cockatoos often nest in loose aggregations, it is possible that a smaller number of recorders could monitor several nests simultaneously. Nest location could be measured from the time difference in the arrival of calls at each recorder (Stevenson et al. 2015). Designing an appropriate recorder array requires an understanding of the distances at which key vocalisations can be detected by the sound recorders in each habitat type (e.g., forest vs paddock). This was not explicitly examined in this study but warrants attention as it could reduce the number of sound recorders required. Another important consideration is post-processing sound data using automated or semi-automated recognition methods (Blumstein et al. 2011; Priyadarshani et al. 2018; Crump & Houlahan 2017). Although not examined here, nestling calls may be a good candidate for automated detection, since they are loud, distinct and are a good indicator of active nesting.

This study aimed to describe the nest-associated vocal behaviours of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, to provide the knowledge necessary for the development of a bioacoustic nest monitoring program. Nest monitoring is important for understanding how breeding activity varies across the landscape, which can help inform management decisions. For instance, two important conservation actions for these subspecies are managing fire impacts to feeding habitat (especially close to nests) and supplementing natural nest hollows with artificial nest hollows. However, spatial prioritisation of these actions could be better informed by a greater understanding of the habitat features that influence the choice of nest location and the likelihood of fledging success. Acquiring sufficient data to test relevant hypotheses is resource intensive if using traditional human-observer methods. Moreover, using traditional methods, it is not feasible to collect behavioural data such as nest visitation rates by the adult birds or the date of fledging or nest failure. The vocal behaviours described in this study suggest that a wide range of behavioural data could be extracted from sound data. Bioacoustics can, therefore, aid monitoring by reducing human survey effort while also providing a range of behavioural data. With continued advances in recording technology and automated sound processing, it is foreseeable that bioacoustics could provide daily data from potential and active nests, with human field effort limited to the deployment and retrieval of sound recorders.

Conclusion

Both subspecies examined in this study are nationally endangered and breeding success is a limiting factor in their recoveries (Berris et al. 2018; Russell et al. 2018). However, monitoring of breeding is difficult in both subspecies, largely because of accessibility and resource restrictions. This potentially limits conservation decision-making as it pertains to nesting. Bioacoustics may help address this issue. With knowledge on vocal behaviour, bioacoustics can be used to monitor not only nesting activity, but also specific behaviours of the cockatoos at nests. Further, it is possible that calls are individually distinct, as shown in the palm cockatoo (Zdenek et al. 2018) and Carnaby’s black-cockatoo (Saunders 1983), in which case bioacoustics may help monitor nest site fidelity. Indeed, bioacoustics offers a range of monitoring options for black-cockatoos and this study provide a preliminary description of conservation-relevant vocalisations for two highly threatened subspecies.

For bioacoustic nest monitoring of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, it appears that the most useful calls are female nest call and loud nestling calls. These call types are the most conspicuous signs of active nesting, since they are loud, distinct and unique to active nests. They are usually easily identifiable to the human ear and on spectrograms. Additionally, because they are relatively stereotypical and loud, these calls could be the focus of automated or semi-automated detection of nesting activity. As bioacoustic technology and analytical methods continue to advance and become more accessible, large-scale bioacoustic nest monitoring programs could be implemented for the conservation benefit of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo.

Acknowledgments

Many people contributed to this study. In-kind field support was provided by the south- eastern red-tailed recovery team and the Kangaroo Island glossy black-cockatoo recovery program. In particular, we thank Tim Burnard, Evan Roberts, Karleah Berris and Torren Welz. We thank the landowners who permitted us to work on their properties and the citizens who reported their sightings of the black-cockatoos. We thank the members of the Ecosounds Lab at the Queensland University of Technology, especially Prof. Paul Roe, Dr. Anthony Truskinger, Dr. Michael Towsey and Dr. Phil Eichinski, for help with sound data handling and storage. We thank Dr. Simone Blomberg for statistical guidance. This work was

55 supported by an Australian Postgraduate Award, the National Environmental Science Programme’s Threatened Species Recovery Hub and the Glossy Black Conservancy.

Call type Caller Behavioural Description Notes (behaviour) identity context

Flight Adult male Adults give flight Typically loud with harmonics. May contain pulsatile elements. Common at nests. Reliably Did not and female calls when flying recorded every day. differentiate

to and from the Take-off call subtype: Given upon take-off flight. Shorter in duration. May show a downward between the nest tree. inflection. sexes.

RTBC: Clear harmonics, with or without pulsatile elements. Males and female calls are usually not overlapping. Take-off call is usually the final call of the day, given by the male as he leaves the nest tree to roost. Take-off call is common at RTBC nests because the male tends not to roost in the nest tree (isolated paddock tree).

GBC: Often contain pulsatile elements. Male and female flight calls are often overlapping. Take- off call is less common at GBC nests because the male tends to roost in the nest tree or a nearby tree.

Begging Adult Female begs in the Often given when the pair is perched on the nest tree or a nearby tree. During incubation and Highly variable female presence of her brooding, a female may beg from within the nest hollow in response to her mate’s flight call. The in structure. mate to elicit main purpose of begging appears to be to elicit allofeeding from the male, however females also allofeeding. appear to beg to maintain contact with their mate. Calls are given repetitively. Begging bouts can be long in duration. Calls are highly variable in structure within and among individuals. Calls vary from loud to soft. Common at nests.

RTBC: Commonly contain harmonics and deterministic chaos. Some calls appear largely chaotic with little harmonic structure. Subharmonics are sometimes present.

GBC: Usually show harmonics, with or without deterministic chaos and frequency jumps. Subharmonics are sometimes present.

Display Adult male Male displays to Display calls are given by adult males when perched close to their mate. Calls appear to function Termed the (courtship) his mate in in courtship and in maintaining the pair bond. Given prior to copulation. Display involves non- kwee-chuck call courtship, to vocal elements, namely head-bobbing and tail-panning wherein the male’s red tail feathers are for the glossy maintain pair displayed to the female. Call are given repetitively. Display bouts can be long in duration. Call black-cockatoo bonds and to contains two elements. Second element is usually louder. Usually soft. by Pepper

instigate RTBC: First element is harmonic with a dominant frequency band at around 1.8 kHz. Other (1996). copulation. harmonic elements are clear. Harmonic at 3.6 kHz can also be high in energy. First element usually contains some noise. Second element is short and broadband without clear harmonics. Second element is usually louder but elements can be similar in energy.

GBC: First element is soft with most energy around 3.4 kHz. Other harmonic elements are usually low in energy and less clear. Second element is short and broadband with harmonics. Second element typically much louder than the first element.

Nest Adult Given by females Very pulsatile. Varies from loud to soft. Appears to function in communication with her mate and Not given every

(nest entry) female when near the nest the nestling. Female will vocalise once sitting at the nest hollow entrance. Sometimes given when time nest is or entering the perched near the hollow. Can be drawn out and relatively long in duration visited or

nest hollow. RTBC: Guttural, ‘purring’ sound. Pulsatile. entered.

GBC: Guttural, ‘growling’ sound. Can be very similar to nestling call, but pulses are more distinct (less noisy).

Nestling Nestling Given by large Call given in the presence of parents. Calling begins upon parents’ return to the nest. Can Type 3 is soft; nestlings in the resemble females’ nest call. Subtypes 1 and 2 are common once the nestling is close the fledging; can be difficult nest hollow or reliably heard every day. to detect on a when sitting at the recorder.

nest hollow Subtype 1: Very loud, broadband, chaotic. May contain harmonics towards the end of the call. Identified from a entrance. Nestling stimulated upon parents’ first return to the nest. Can be associated with wing-flapping in small sample of older nestlings. nests with video

Subtype 2: Softer call but otherwise similar in structure to subtype 1. Nestling is less stimulated. footage, but expected at every Subtype 3 (begging): Soft, highly variable. Can resemble female begging calls. Appears to nest. function as a close contact call with the adult female. Elicits allofeeding by female. Sounds high- pitched and ‘squeeky’.

RTBC: Very noisy and chaotic. Sounds ‘throaty’ or ‘wheezy’. Not pulsatile. Harmonics sometimes present. GBC: Noisy, chaotic call, often with some harmonic structure. Highly pulsatile. Can be very similar to females’ nest call, but more chaotic and less harmonic.

Perch Adult male Given when Perch calls comprise several, variable contact calls given by adult birds when perched on or near Graded contact (RTBC). perched on or near the nest tree. Appear to maintain contact between male, female and nestling in hollow. In the calls. Soft perch

Adult male the nest tree. GBC, both sexes commonly call when perched. In the RTBC, only male calls are reliably heard in calls are easily and female recordings. Perch calls range from loud to soft. Male perch call (subtype 2) is most common. masked by other (GBC). Recorded often. After female has entered the nest hollow for the night, male will call sounds and intermittently, often for a long duration. Perch calls are often the final vocalisations of the day. therefore appear Subtypes 2 and 3 (soft perch calls) often given in response to female begging. to be uncommon

RTBC: Given by male. Always begins with pulsatile or chaotic elements. Has a ‘crackling’ in sound quality to the sound. recordings.

o Male loud perch call (perch subtype 1): Loud. Sounds ‘trumpet-like’. Clear harmonic structure in second part of call. Often with subharmonics. First part of call is chaotic and pulsatile. First part of call may be lacking, showing only harmonics; typically given soon after landing on the nest tree.

o Male soft perch call (perch subtype 2): Beginning of call is highly pulsatile, with a ‘crackling’ sound quality, ending in harmonics. Can be loud or soft. Typically given when male is perched on the nest tree, including after the female has entered the hollow. o Male soft perch call (perch subtype 3): Soft, guttural. Lacks harmonics but otherwise is similar to subtype 2. Entire call is chaotic or pulsatile.

GBC: Highly variable, graded calls. Subtypes 1, 2 and 3 are given by adult males. Subtypes 4, 5 and 6 are given by adult females. Perch calls by both adult birds are commonly given at nests, often in interactions.

o Male loud perch call (perch subtype 1): Contains harmonics. May contain some pulses and chaos. Less common than soft perch calls. Usually loud but can be relatively soft. May precede take-off. o Male soft perch call (perch subtype 2): Most common adult male soft perch call. Usually very soft, ‘fuzzy’ sound. Often appears as a single, wide frequency band between 2.5-4.5 kHz, ending with a downward inflection. Louder variations have additional high and low frequency bands but maintain the ‘fuzzy’ quality to the sound. Common at nests. Often given in response to female perch calls or begging. Given when the female is in the nest hollow. o Male soft perch call (perch subtype 3): Very similar to subtype 2. Begins as subtype 2 but abruptly changes to end with a louder harmonic element. Some individuals appear to use subtype 3 more often than subtype 2. o Female loud perch call (perch subtype 4): Clear harmonics. Sometimes shows frequency modulation. Often given when the female is perched high on the nest tree or on a nearby tree.

60 o Female loud, alarm perch call (perch subtype 5): Loud, pulsatile call. Can resemble nest entry call. Given when female is alarmed near the nest, usually when defending the nest tree from other birds (e.g. other glossy black-cockatoos or galahs). o Female soft perch call (perch subtype 6): Soft. Resembles male perch call subtype 2 but contains more than one frequency band. Often appears as two dominant frequency bands around 3-3.5 kHz and 6-6.5 kHz. Uncommon and easily masked by other sounds.

a b c

d e

g h i j k l

a b c

d e f g h

i j k l m n o

Nestling Nestling Nestling Nestling Nestling Nestling Nestling

Perch 2 Flight Perch 2 Perch 1 Perch 1 Begging ♂ ♀ Perch 1 ♂ ♂ ♂ Begging ♀ ♂ Perch 2 Perch 2 ♀ Perch 2 ♂ ♂ ♂

Nestling Nestling Nestling Begging Nestling Begging Begging Perch 3 Perch 3 ♀ Perch 3 ♀ ♀ ♂ ♂ ♂

Nestling Nestling Nestling Nest Nestling Nestling Nestling ♀ Perch 3 ♂

d Take-off Begging ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♂

64 nonlinearity. Female responds with soft nest call and male with soft version of perch subtype 3; (d) Loud female begging bout showing clear subharmonics. Take-off flight by male and female. Other species’ vocalisations are not indicated. Spectrograms created from video footage using Raven Pro 1.5 (Cornell Lab of Ornithology; Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap). X axis denotes time into the video file.

Perch 2 Perch 4 ♀ ♂ ♀ ♀ ♂ ♀ ♂ ♂ ♀ ♂

b Perch 5 Perch 5 Perch 5 Perch 5 ♀ ♀ ♀ Overlap ♀ Perch 2 Perch 1 Perch 2 Perch 3 ♂ ♂ ♂ ♂

Nest entry ♀ Perch 6 Nest entry Perch 6 ♀ ♀ ♀

d Flight Nestling Nestling Nestling Nestling Nestling ♀

Nest entry Nest entry ♂ ♂ ♀ ♀

e ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ ♀ Perch 2 Perch 2 Perch 2 Perch 2 ♂ ♂ ♂ ♂

(subtypes 1, 2, and 3); (c) Female nest entry call and soft perch call (subtype 6); (d) Nestling call, female flight and nest entry call, and male soft perch (subtype 2); (e) Female begging bout and male response (perch 2). Other species’ vocalisations are not indicated. Spectrograms created from video footage using Raven Pro 1.5 (Cornell Lab of Ornithology; Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap). X axis denotes time into the video file.

Chapter 4 | Fledge or fail: Nest monitoring of the endangered black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus, using bioacoustics and open-source call recognition

Abstract

Ecologists are increasingly using bioacoustics in wildlife monitoring programs. Remote autonomous sound recorders provide new options to collect data for species and contexts that were previously more difficult. However, post-processing sound files to extract relevant data remains a significant challenge. Detection algorithms, or call recognisers, can aid automation of species detection, but their reliability, in practice, has been mixed. Importantly, complex algorithms often require expert programming skills which reduces their accessibility to ecologists responsible for monitoring. In this study, we investigated the performance of open-source call recognisers provided by the monitoR package in R, a language popular among ecologists. We tested recognisers on sound data collected under natural conditions at nests of two endangered subspecies of black-cockatoo, the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus (n = 23 nests), and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne (n = 21 nests). Specifically, we tested binary point matching recognisers, trained on nestling calls, in confirming daily nest activity (active or inactive) and nest outcome (fledge or fail). Through field observations, we first described the vocal signal associated with fledging. We then tested recognisers on 3 x 3-hour recordings per nest, from early, mid and late stages of the recording period. We consider this an appropriate regime for these subspecies to, in the first instance, confirm nest activity broadly across the nesting season (up to four months in duration). Daily nest activity was correctly assigned in 61.9% of survey days analysed (n = 63 days) for the red-tailed black-cockatoo, and 68.1% of survey days (n = 69 days) for the glossy black-cockatoo. Fledging was successfully detected in five out of six events in the glossy black-cockatoo, and two out of three events in the red-tailed black-cockatoo. Precision of individual detections was moderate, with many false positives. Manual verification of outputs is, therefore, required, however it is not necessary to verify all detections to confirm an active nest (i.e., nest is deemed active when true positives are identified). We conclude that bioacoustics combined with semi-automated post-processing can be an appropriate tool for nest monitoring in these endangered subspecies.

Introduction

Acoustic technologies offer new ways to collect data on wildlife populations and ecosystems (Servick 2014). Data can now be collected over spatial and temporal scales much greater than those historically possible (Towsey et al. 2014; Shonfield & Bayne 2017; Sugai et al. 2019); Australia’s new, permanent continent-wide Acoustic Observatory (https://acousticobservatory.org/) is a case in point. Monitoring with sound recorders is broadly categorised as either ecoacoustics, the study of soundscapes and ecological processes (e.g., by using acoustic indices of soundscape complexity), or bioacoustics, the study of wildlife sounds, usually with a single species focus (Sueur & Farina 2015). Advances in recent decades have seen bioacoustic methods used to locate rare and cryptic species (Frommolt & Tauchert 2014; Sebastián-González et al. 2015; Wrege et al. 2017; Dema et al. 2018; Schroeder & McRae 2020), measure population density and abundance (Dawson & Efford 2009; Marques et al. 2013; Borker et al. 2014; Stevenson et al. 2015; Pérez-Granados et al. 2019), identify individual animals (Ehnes & Foote 2015), localise individuals (Frommolt & Tauchert 2014) and assess species occupancy (Furnas & Callas 2015; Campos- Cerqueira & Aide 2016; Chambert et al. 2018). Indeed, bioacoustics has expanded our capacities to monitor wildlife for a wide range of metrics (Shonfield & Bayne 2017; Sugai et al. 2019). However, bioacoustics faces several big data problems, not the least of which concern processing large volumes of sound files to extract relevant ecological data (Servick 2014). These issues currently limit the utility of bioacoustic methods to ecologists and need to be addressed for the methods to be more widely adopted for wildlife monitoring and conservation.

A key issue in bioacoustics is detecting and classifying species’ calls from sound recordings (Browning et al. 2017; Sugai et al. 2019). Manual processing involves viewing spectrograms to visually and aurally detect calls, however this approach alone is not typically feasible for anything other than short-term recordings. For this reason, ecologists have increasingly used automated or semi-automated detection methods via call recognisers, a general term for various algorithms that interrogate sound recordings to detect calls of interest (Sugai et al. 2019). The choice of algorithm, and subsequent performance, depends on many factors, including the characteristics of the calls to be detected, other species’ calls and environmental noise (Brandes 2008; Towsey et al. 2012; Cragg et al. 2015; Salamon et al. 2016; Crump & Houlahan 2017; Knight et al. 2017; Priyadarshani et al. 2018). Recognisers have been used with varying levels of success. Some have performed relatively

70 well (Kirschel et al. 2009; Frommolt & Tauchert 2014; Zeppelzauer et al. 2015; Große Ruse et al. 2016; Albornoz et al. 2017; Alonso et al. 2017; Ruff et al. 2019) but this is not always the case (Tanttu et al. 2006; Cragg et al. 2015; Heinicke et al. 2015; Schroeder & McRae 2020). Generally, the field of recogniser development, while becoming more established, still faces difficulties; it is not a panacea to bioacoustics’ challenges. Reliable, fully-automated methods are rare and most often semi-automated methods, combining automated recognition and manual verification, are used (Shonfield & Bayne 2017; Sugai et al. 2019).

Although progress in recent years has been substantial, developing a custom call recogniser that performs well requires some expertise in programming and machine learning, which limits accessibility to ecologists and managers responsible for on-ground monitoring (Sebastián-González et al. 2015; Priyadarshani et al. 2018; Sugai et al. 2019). Commercial software partly addresses this issue by providing a more user-friendly interface with which to train algorithms and process sound files. These recognisers can perform well and improve monitoring. For example, Shonfield et al. (2018) found that Hidden Markov model recognisers built using Song Scope software (Wildlife Acoustics Inc., Maynard, MA, USA) allowed for many more owl detections to be acquired. However, performance is not necessarily high, and results can be inconsistent among different software (Duan et al. 2013; Lemen et al. 2015; Rocha et al. 2015; Russo & Voigt 2016; Joshi et al. 2017; Knight et al. 2017; Schroeder & McRae 2020). For commercially-available bat detectors, which are often used by ecological consultants and others involved in monitoring, Russo and Voigt (2016) caution that their use has preceded proper testing and detections should not be accepted without scrutiny. The issue lies partly in that the underlying construction of the algorithms is not easily understood or altered by non-experts. An additional limitation of commercial software is their cost. Acoustic projects can be costly to setup and operate, especially in terms of data analysis and processing (Browning et al. 2017). This may prelude conservation programs, which often operate on small budgets, from investing in commercial software. Together, these issues - namely programming skills, recogniser performance and cost - may see traditional survey methods favoured over bioacoustics, despite the many benefits of acoustic methods for species monitoring.

In this study, we examined the performance of a newly-developed call recogniser package implemented in R software (R Core Team 2019), an open-source statistical language popular among ecologists (Lai et al. 2019). The monitoR package (Katz et al. 2016b; Hafner & Katz 2017) provides two template-matching algorithms: spectrogram cross correlation and

71 binary point matching. Like most commercial options, these recognisers are easily constructed from example training calls (templates) and allow the end-user to manipulate various parameters that alter the recogniser’s performance. This flexibility is important for creating a recogniser in line with a project’s aims (Crump & Houlahan 2017; Chapter 2). One of the most important features for users to define is score threshold, which is the threshold of similarity between templates and raw sound data at which the recogniser returns a detection (termed score cut-off in monitoR) (Knight et al. 2017; Knight & Bayne 2019). Each signal evaluated in raw sound data is assigned a score, which can be taken as the probability that the signal belongs to the species of interest. Though often overlooked and not reported in studies, score threshold can strongly affect recogniser performance, such as trade-offs in the number of true and false positive detections returned, as well as true and false negatives (Brauer et al. 2016; Katz et al. 2016a; Knight & Bayne 2019). Projects seeking to locate cryptic species, for instance, will require a lower detection threshold (i.e., greater sensitivity) to detect rare or faint calls, with the likely trade-off of increased false positive detections (i.e., poorer specificity). Considering a range of score thresholds, recent tests on calls of the common nighthawk, Chordeiles minor, have shown monitoR to perform well relative to other popular recognisers, including those from commercial software (Knight et al. 2017). Although monitoR has only been tested on a small number of species (Katz et al. 2016a; Knight et al. 2017), these results are promising for addressing some difficulties in processing bioacoustic data with reliable, high performing open-source recognisers, implemented in a way familiar to many ecologists.

We tested the performance of recognisers constructed in the monitoR package at detecting daily nest activity (active or inactive) and nest outcome (fledge or fail) in two endangered subspecies of black-cockatoo endemic to south-eastern Australia. The Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne, comprise small and isolated populations whose recoveries are partly constrained by low reproductive output. While the Kangaroo Island glossy black-cockatoo has had some traditional nest monitoring, there has been no routine nest monitoring for the south-eastern red-tailed black-cockatoo. Both populations are remote, and nesting occurs across large spatial areas, especially in the red-tailed black- cockatoo. As such, both would benefit from more efficient nest monitoring methods that reduce the requirement for in-field human observers. Further, there is currently no method with which to directly monitor fledging events and, therefore, estimates of breeding success

72 rely on in-field observations of large nestlings. Here, we apply the nest-associated vocalisations of both subspecies, as described in Chapter 3, to assess the utility of bioacoustics for nest monitoring. Their vocalisations are loud and distinct and are given at predictable times each day. These traits may make them potentially suited to automated or semi-automated methods of call detection. In this study, using sound data collected under natural conditions at wild nests, we aimed to (a) describe the vocalisations associated with a fledging event in both subspecies, (b) develop an open-source call recogniser using the monitoR package in R, and (c) for each nest, test the performance of the recogniser in detecting daily nest activity, over the course of the nesting period, and nest outcome.

Methods

Sound data collection

Sound data were collected over two breeding seasons (2017 and 2018) from 24 nests of the glossy black-cockatoo on Kangaroo Island in South Australia, and 22 nests of the red- tailed black-cockatoo in the Casterton region of Victoria. Each nest tree was fitted with an autonomous sound recorder (Frontier Labs Bioacoustic Audio Recorder, https://frontierlabs.com.au/) for the duration of the nesting period (i.e., until fledging or confirmed failure). If a female cockatoo was incubating an egg or brooding a chick, we installed the recorder on a nearby tree within 10 metres of the nest tree, to minimise disturbance during this sensitive period. Incubation and brooding were confirmed either via nest inspection with a pole-mounted camera or by the presence of a female at a hollow’s entrance following soft tapping on the nest tree. Each sound recorder was programmed to record for three hours per day, concluding at 30 minutes after sunset (sunset-based schedule), as this is when the birds are most active at nests (DT, pers. obs.). Additionally, one day per week, recording commenced at 30 minutes before sunrise and concluded at 30 minutes after sunset (full-day schedule). Technical issues in the first breeding season resulted in some recorders losing their sunset-based schedules, and therefore recorded only during morning schedules. To increase the chances of recording the fledging event, some recorders were updated to record every day at the full-day schedule if a large nestling was observed at the nest hollow entrance during field inspections. All recordings were made using an omnidirectional microphone, with a fixed gain of 20 dB and a sample rate of 44.1 kHz. Microphones had an 80 Hz high-pass filter to reduce the effects of low frequency noise (e.g. wind and traffic). Recorders were fitted with four rechargeable lithium ion batteries and one

128 GB SanDisk memory card, which were replenished at approximately 6-week intervals. All recordings were made in uncompressed wave (.wav) format.

Fledging vocalisations

Vocalisations that accompany fledging were confirmed through behavioural observations and video footage of fledging events in both subspecies (n = 3 for the Kangaroo Island glossy black-cockatoo; n = 1 for the south-eastern red-tailed black-cockatoo). All but one of these nests, which was of the glossy black-cockatoo, had autonomous sound recorders active at the time of fledging. To witness fledging, the lead researcher (DT) followed closely the activity of late stage nests. If a large nestling was seen at a nest, the nest was visited daily, weather permitting, until fledging. Video footage was recorded using a Canon 5D mark III DSLR camera and Canon 100-400 II IS USM telephoto lens with a Rode VideoMic Pro microphone attached. Sound files were extracted from videos using Adobe Media Encoder CC 2017 and viewed in Raven Pro 1.5 to determine the spectrographic patterns of fledging vocalisations. For all other nests where fledging or failure were not observed in the field, we manually viewed spectrograms collected using the autonomous sound recorders to determine the date of fledging or failure. Fledging was confirmed by observing the same spectrographic pattern as that identified from the videos. Failure was confirmed where the presence of the nesting birds’ vocalisations ceased in sound recordings before the time of expected fledging (typically early in the nesting period).

Recogniser development

We used binary point matching, implemented using the monitoR package in R, as the recognition algorithm for this study (Katz et al. 2016b). We chose to use monitoR because it is open source, easily shared among end-users and, once scripted, does not require expert programming skills. Additionally, the seewave package (Sueur 2019) allows long-duration sound files to be read into R in smaller sizes which greatly reduces the impact of the recogniser on computer memory. Outputs can be inspected and verified in various ways, including in R or by importing detections into other acoustic software. Another approach is to write spectrograms images from R wherever detections are present; this allows for visual verification without the need for any acoustic software. While call playback is not possible with this method, for calls that are visually distinct in structure this may be an efficient approach. For programs that engage citizen scientists to verify detections, having a variety of options is useful. Commercial recognition software is less flexible in this way, and can be

74 costly. For these reasons, we chose to used monitoR exclusively in this study. These are important considerations for making automation accessible to conservation stakeholders.

Binary point matching is a template matching algorithm, wherein templates of the calls to be detected (‘templates’) are compared to spectrograms of sound recordings (‘surveys’). The method of binary point matching used in monitoR is a variation of that described in Towsey et al. (2012) (Katz et al. 2016b). MonitoR also provides spectrogram cross correlation, another template matching algorithm that many previous studies have used. Here, preliminary trials of spectrogram cross correlation returned many more false positive detections than binary point matching, and therefore we did not pursue this option further. Although the two algorithms are similar, they differ in their underlying construction. Spectrogram cross correlation uses all regions of the template (i.e., the entire boxed area around the template call) for matching against the survey spectrograms, and scores each spectrographic frame for Pearson’s correlation coefficient. A score of 0 represents no correlation, while a score of 1 represents a perfect match. Binary point matching, conversely, delimits ‘on’ and ‘off’ regions (call and non-call) of the template and ignores all others. Thus, within a selection of a call (the template) there are fewer regions with which the survey is compared. In binary point matching, each time bin is scored for the difference in mean amplitude between on and off regions. MonitoR allows the user to set score cut-off (or detection threshold), which is the minimum score that will return a positive detection. As such, score cut-off determines the relative proportions of true and false positives, as well as true and false negatives (Katz et al. 2016b).

To determine an appropriate threshold for the nestling calls of the glossy and red-tailed black-cockatoo, and to select appropriate call templates, we performed a pilot study on a small number of surveys. To select sample calls as templates, for each subspecies we randomly selected three nests at which nestling calls were recorded. For each nest, we selected two representative nestling calls that were not masked by other sounds and clearly showed the structure of the call. In creating the binary templates, an amplitude cut-off, above which a pixel is denoted as ‘on’, was set at -30 dB. To select surveys on which to test the recognisers’ performance, for each subspecies we randomly selected two nests at which fledging was recorded. For each nest, we took a 15-minute sound clip that included fledging. Therefore, for each subspecies, the pilot recogniser compared six template nestling calls to 2 x 15-minute sound recordings that included fledging events. The score cut-off was set at 15 for both subspecies. Previous tests of monitoR have examined a range of score cut-offs from

0 through 30 (Katz et al. 2016a). Since nesting black-cockatoos are mostly loud and close to the sound recorder, we did not examine low score cut-offs necessary for detecting faint calls.

Detections were imported into Raven Pro 1.5 for verification (Hann window; window size = 512 samples; hop size = 512 samples; 50% overlap). Detections were categorised as ‘yes’ if they were a nestling (true positive), or ‘no’ in all other cases (false positive). Nestling calls have been previously described by comparing spectrograms to video recordings taken simultaneously (Chapter 3) and this knowledge was used to verify detections in this study. We did not quantify false negative (missed) detections because, in monitoring nest activity in these species, we do not need to detect every individual call; rather, an appropriate false negative would be if no detections were returned for a day where the nest was active, but in the pilot study calls were detected on every day (i.e. no false negatives).

For each subspecies, the pilot recogniser performed relatively well for one nest. At a score cut-off of 18, most detections (i.e., of 18 or greater) at these nests were true positives (Fig. A4.1). For the other nest of each subspecies, more false positive detections were returned at high score cut-offs. For the Kangaroo Island glossy black-cockatoo, one nest returned false positive detections that scored higher than true positives. At a score of 18 or greater, 29 of 35 detections were false positives (c.f. 9 out of 44 detections at the other nest). This is likely a result of the nest hollow being relatively high in the tree which meant that the distance between the nest and the sound recorder was greater than at other nests. Hence, the recorded nestling calls were lower in amplitude than at other nests. Calls from other birds (e.g., corellas, Cacatua spp.) that were closer to the sound recorder were greater in amplitude. For the south-eastern red-tailed black-cockatoo, one nest returned 83 detections of a score of 18 or greater, of which 51 were false positives (c.f. 10 out of 144 detections at the other nest). However, almost all were detections of the adult red-tailed black-cockatoos (not the nestling). Some false detections were of sulphur-crested cockatoos. From these results, we chose a score cut-off of 18 for the final recognisers, as this threshold appeared to represent the best trade-off between true and false negatives. Of the six templates of nestling calls tested in each recogniser, two returned most of the true positive detections (Fig. A4.1), and these templates were retained in the final recognisers. The final recogniser for each subspecies, therefore, included the two best-performing templates and had a detection threshold of 18.

Recogniser performance

A total of 46 nests were monitored in this study (n = 24 for the Kangaroo Island glossy black-cockatoo; n = 22 for the south-eastern red-tailed black-cockatoo). We quantitatively measured the performance of the recogniser on recordings from every nest except those from which the templates were constructed (one nest per subspecies). For most nests, we monitored 3 x 3 hours of sound recordings collected in late afternoons, finishing at 30 mins after sunset. In some cases, technical problems caused recordings to fail, in which case we used recordings collected earlier in the day (e.g. 3 hours commencing 30 minutes before sunrise). Recordings to be tested were chosen randomly from three defined time periods in each nest’s recording schedule: early, mid and late stages, which we categorised as time 1, time 2 and time 3. For nests that fledged, we randomly selected one recording before fledging (time 1) and one recording after fledging (time 3), as well as the recording on the day of fledging (time 2). For nests deemed successful but where the fledging event was not recorded (e.g., batteries lost power before fledging), we randomly chose three recording days from early, mid and late stages. The same was done for unsuccessful nests, noting that the recording period was often shorter for failed nests because sound recorders were removed if failure was confirmed through in-field inspections. The recording period per nest ranged from 6 – 153 days. For nests that were monitored over several months, the test recordings were often separated by weeks or months. For nests monitored for shorter periods, test recordings were temporally closer (sometimes consecutive days). We did not exclude any recording days due to poor conditions (e.g., rain) as this is a true limitation of bioacoustic monitoring of these subspecies. Detections were verified using Raven Pro 1.5 (spectrogram parameters: Hann window; window size = 512 samples; hop size = 512 samples; 50% overlap). Verification was done at two levels: individual detections and recording day. First, detections were categorised as ‘yes’ if they were a nestling call (true positive), and ‘no’ in all other cases (false positive). For false positive detections, we also recorded whether the detection was a different call of the target species (i.e., any adult call of red-tailed or glossy black- cockatoos). We then categorised each recording day (survey day) as: (a) Correctly assigned nest activity (true positive: nest active and true positive detections returned; or true negative: nest inactive and no detections returned), (b) Missed nesting activity (false negative: nest active but no true positive detections returned), or (c) Incorrect detection of nesting activity (nest inactive but false positive detections returned). For survey day analyses, we accepted

77 adult calls as true positive detections. Finally, for each nest where fledging was recorded (n = 6 for the Kangaroo Island glossy black-cockatoo and n = 3 for the south-eastern red-tailed black-cockatoo; excluding nests from which the templates were constructed), we noted whether the recogniser successfully detected the fledging event. Fledging was recorded at seven nests of the glossy black-cockatoo, but one of these were used for template calls and therefore were not tested here. Fledging was recorded at three nests of the red-tailed black- cockatoo and all were tested, as the templates used were from a different nest. Verifications were summarised and plotted using the dplyr (Wickham et al. 2019) and ggplot2 (Wickham 2016) packages in R statistical language (R Core Team 2019).

Results

Fledging vocalisations

For the south-eastern red-tailed black-cockatoo, we identified fledging vocalisations in recordings from three nests, one of which was witnessed in-person (DT, pers. obs.). Fledging was presumed at a further two nests. At one of these, the sound recorder’s batteries expired before fledging, but late-stage nestling calls were present until the final day of recording. At the other nest, we confirmed that the nest hollow was empty, via inspection with a wireless video camera mounted to a telescopic pole, two days after late-stage nestling calls were recorded. It is likely that the nestling fledged at a time of day not captured by the recording schedule used in this study. In total, of the 22 nests monitored for this subspecies, nestlings were recorded at nine nests, fledging was deemed successful for five nests, and at three of those fledging vocalisations were recorded. The outcome of eight nests was uncertain, because recording schedules did not work properly in the first breeding season.

For the Kangaroo Island glossy black-cockatoo, we identified fledging vocalisations in recordings from seven nests, including the two nests where fledging was witnessed. Fledging was presumed at a further six nests. This determination was made by field staff in the recovery program; nests are deemed successful if a late stage nestling is observed and later inspection (with a pole-mounted camera) shows the nest to be empty. In total, of the 24 nests monitored for this subspecies, nestlings were recorded at 15 nests, fledging was deemed successful for 12 nests, and at seven of those fledging vocalisations were recorded. Nest outcome was uncertain in four cases.

For both subspecies, we identified a unique acoustic signal at fledging (Fig. 4.1). At each nest, in the time period immediately before fledgling, the nestling called loudly, to

78 which to adult birds usually responded with perch calls. At fledgling, when the nestling takes flight, all three birds call loudly at the same time, which shows as a sudden cluster of high- amplitude calls. As they fly away from the nest, there is a rapid attenuation in the amplitude of the calls, especially of the high-frequency components. This increase and subsequent decrease in acoustic energy is characteristic of fledging. This rapid modulation of amplitude, as well as presence of the nestling calls, make the signal unique from other behavioural contexts (e.g. regular flight away from the nest). Often, the nestling’s wing beats of its first take-off can also be heard and seen on a spectrogram if viewed at large scale. Additionally, through manual inspection of spectrograms, we found that the cockatoos’ vocalisations were not detected at nests after fledging, which further supports the finding that fledging is vocally indicated. Through in-field observations and video footage, we found that this signal of fledging was present for both the Kangaroo Island glossy black-cockatoo and the south- eastern red-tailed black-cockatoo.

Recogniser performance

At the level of the survey day, nest activity was correctly identified in 61.9% of recording days analysed (nest active and detections returned, or nest inactive and no detections returned) for the red-tailed black-cockatoo (n = 63 survey days) (Table 4.1). Nest activity was missed in 3.2% of recording days (nest active and no detections returned), and in 34.9% of recording days nest activity was incorrectly detected (nest inactive and detections returned). For the glossy black-cockatoo (n = 69 survey days), nest activity was correctly assigned in 68.1% of recording days. Nest activity was missed in 18.8% of days, and incorrectly detected in 13% of days. The recognisers successfully detected the fledging event in five out of six fledging events in the glossy black-cockatoo, and two out of three fledging events in the red-tailed black-cockatoo.

At the level of individual detections, precision was moderate and many false positives were returned. In total, the recogniser returned a total of 2,087 detections for the Kangaroo Island glossy black-cockatoo (from 207 hours of recording from 23 nests) and 3,958 detections for the south-eastern red-tailed black-cockatoo (from 189 hours of recording from 21 nests). Nestling calls were correctly assigned in 7.6% and 18.8% of detections in the red- tailed and glossy black-cockatoo, respectively. When adult calls were considered to be correct detections, recogniser performance improved to 35.3% and 27.9% in the red-tailed and glossy black-cockatoo, respectively. For both subspecies, fledged nests returned the

79 greatest number of correct detections in all time periods, but false positive detections were returned for every nest type (fledged, failed and unsure) and every time period (Table 4.2). The relative proportion of true positive detections was generally highest mid-stage in the recording periods (time 2). This was expected because the recognisers were constructed on late-stage nestling calls and, for successful nests, time period 2 is when these calls were present in the sound recordings. While increasing the detection threshold would reduce the number of false positive detections, the number of true positive detections would also be reduced (Figs. A4.2 and A4.3).

Discussion

In this study, we examined the utility of open-source call recognisers in a bioacoustic nest monitoring program for two endangered subspecies of black-cockatoo, the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Nest monitoring for these subspecies has been limited by the accessibility of nest sites, as both occur in remote regions, and the associated costs of in-field monitoring. Further, directly monitoring fledging has not been viable with human observers, therefore breeding success is estimated from in-field observations at nests (e.g., large nestlings are assumed to fledge), for the glossy black-cockatoo, or post-breeding flocks, for the red-tailed black-cockatoo (Berris et al. 2018; Russell et al. 2018). A useful bioacoustic nest monitoring program is one in which nest status can be determined as active or inactive on any given survey day, and where nest outcome (fledge or fail) can be measured. Nest status can be confirmed by the presence of nest-associated vocalisations; in this study, we chose to focus primarily on nestling calls as they are loud, distinct and not easily confused with other birds (e.g. other nesting cockatoos) (Chapter 3). To provide a direct measure of breeding success, one aim of this study was to determine the vocalisations associated with the fledging event.

Through field observations and reviews of spectrograms of recordings taken at fledging, we determined that fledging is vocally indicated in these subspecies. With this knowledge, we then aimed to test the utility of a bioacoustic nest monitoring program, including post-processing with monitoR, in a regime appropriate for a monitoring program of these subspecies. Specifically, for each nest, we tested the recogniser on three recording days representing early, mid and late stages of the recording period (named time 1, time 2 and time 3). In practice, this approach would help to, in the first instance, confirm nest activity (nesting active or inactive) across the duration of the recording period, which can be up to 4 months if

80 recorders are deployed soon after laying. This information can then inform additional recording days to be interrogated, if any. This is more efficient than using the recogniser on all survey days. With this approach, we were able to determine daily nest activity with a moderate level of success. At the level of the survey day, recogniser performance was similar for both subspecies. Nest activity was correctly assigned in 61.9% of recording days for the red-tailed black-cockatoo and 68.1% of recording days for the glossy black-cockatoo. Most errors were incorrect detections of nest activity, where the nest was inactive but false positive detections were returned. As such, manual verification is required, and recognisers’ outputs should not be accepted without inspection. This concurs with many previous studies that show semi-automated methods to be most reliable (Sugai et al. 2019).

Where the fledging event was recorded (n = 6 nests for the glossy black-cockatoo; n = 3 nests for the red-tailed black-cockatoo), the recogniser successfully detected fledging at five nests of the glossy black-cockatoo and two nests of the red-tailed black-cockatoo. Where fledging was missed, this appears to be a result of a very noisy environment, in the case of the glossy black-cockatoo, or lower amplitude calls from a relatively greater distance between the recorder and the nest, in the case of the red-tailed black-cockatoo. In the latter case, the distance between the birds and the recorder meant that call amplitude as the birds took flight decreased quicker than at other nests. This means that the similarity of the recogniser to the sound file decreased quicker, thus increasing the chance of false negatives. Nonetheless, that fledging was detected at most nests, thereby providing a direct measure of fledging, greatly improves our ability to measure breeding success.

One benefit of bioacoustics over traditional methods is that it allows for more direct measures of some behaviours that are otherwise difficult to record (Chapter 2). For example, vocalisations can indicate copulation in elephants (Payne 2003; Poole 2011), the birth of a calf in killer whales (Weiß et al. 2006), mother-pup reunions in Weddell seals (Collins et al. 2011) and foraging in sperm whales (McDonald et al. 2017). For glossy and red-tailed black- cockatoos, there are at least six behavioural contexts vocally indicated at nests (Chapter 3) and changes through nest development can be observed (Chapter 5). For these species, this vocal complexity offers a range of data that can be collected in bioacoustic monitoring programs. Fledging vocalisations are among the most useful signals for bioacoustic monitoring as it relates to breeding. For conservation, understanding how nest outcome varies by region and resource availability can help prioritise management actions that seek to

81 maximise reproductive success. In this study, the recogniser missed two fledging events, however true positive detections were returned on those survey days. In such cases, when spectrograms are reviewed to verify detections, the fledging event is relatively easy to identify as it is preceded by other calls of the nestling and adults.

Most studies investigating recogniser performance have done so at the level of the individual call or detection, albeit inconsistently (Knight et al. 2017). For comparison, we also quantified performance at this level by manually verifying all detections returned. Relative to the survey day, recogniser performance was poor when considering each detection individually. Of the 2,087 detections for the glossy black-cockatoo, only 18.8% were correctly identified as nestlings. When adult calls were considered true positives, 27.9% were correct. For the red-tailed black-cockatoo, of 3,958 detections 7.6% were correctly identified as nestlings and 35.3% were correct if adult calls were included. False positives (i.e., lower precision) are often generated by calls of sympatric species (Cragg et al. 2015). In this study, false positives were often calls of other cockatoo species, such as the sulphur- crested cockatoo, Cacatua galerita, in Victoria, and the galah, Eolophus roseicapilla, on Kangaroo Island. Poor precision can greatly increase the burden of post-processing if every call requires verification, such as in studies seeking to detect rare or cryptic species (Frommolt & Tauchert 2014; Dema et al. 2018; Schroeder & McRae 2020) or to obtain population metrics from call rate (Borker et al. 2014). While this does not apply to nest monitoring in black-cockatoos, which operates at the level of the survey day, improving precision would reduce the number of days at which nests are incorrectly deemed to be active. This would improve the method overall and options to achieve this should continue to be investigated.

Our findings align with many studies that have tested call recognisers under natural conditions; recognition can be helpful, but it is not yet a perfect solution (Sugai et al. 2019). Difficulties arise from extraneous source of noise and vocalisations of sympatric species, as well as the varying quality of the vocalisations for the species of interest (e.g., with distance from the sound recorder) (Zwart et al. 2014; Cragg et al. 2015; Heinicke et al. 2015; Sebastián-González et al. 2015). More sophisticated methods may improve performance but, even if they exist, they are often not easily available to the people responsible for on-ground monitoring. Currently, options for ecologists are more limited unless they collaborate with computer scientists. Nonetheless, ecologists using call recognisers, whether commercial or open-source, should carefully consider the choice of algorithm and its construction,

82 particularly the training data used. Recogniser construction should ideally be an adaptive process whereby training data are tested and refined to improve performance. In our case, we chose call templates following a pilot study. Our focus was nestling calls, as these are most indicative of active nesting, however after formal testing we now consider that the recognisers could be improved by including some adult calls, since most adult calls detected were from the nesting pair of interest (calls were loud, sustained, and prompted nestling calling). Lastly, in addition to recogniser performance, the design of a bioacoustic program, including its recording schedule, is critically important. While bioacoustics is appealing in its ability to collect more data than other methods, appropriately reducing recording times can lessen the burden of post-processing. Again, the project’s aims should drive survey design.

Like all methods, bioacoustics faces limitations, and it is important to consider these in relation to other monitoring approaches. For the subspecies examined here, nest monitoring with traditional means is limited by poor funding, the relative survey effort required to manually inspect nests, and the remoteness of most nests. Obtaining data from many nests has substantial human resource requirements, and the relative conservation benefits of this approach have not been explored. Further, traditional monitoring does not allow daily data collection, and does not directly measure nest outcome. On Kangaroo Island, motion-sensor cameras have been used successfully at some nests, but this requires tree- climbing and post-processing of images. Tree-climbing can be beneficial for yielding data that are otherwise unattainable (e.g., DNA samples), but in the case of the south-eastern red- tailed black-cockatoo, tree-climbing is unsafe, because most nests are in dead trees, which restricts traditional monitoring to on-ground observers. In this study, we sought to examine the potential utility of bioacoustics to monitor nesting, although we did not undertake a cost- benefit analysis of all possible monitoring approaches. This would be beneficial, particularly for these endangered subspecies for whom conservation action is urgent yet funding is limited. Where bioacoustics is used, the ease and efficiency of processing data with call recognisers relative to manual inspection of spectrograms should be considered (Joshi et al. 2017). Since cockatoo calls are usually very loud, they are easy to visually detect on spectrograms, even when viewed in 15-minute segments. A benefit of manual processing is that expert knowledge on vocal behaviours can more readily describe nest status (e.g. differentiating active nesting from nest prospecting); this is a critical, yet overlooked, aspect of behavioural monitoring with bioacoustics. Nonetheless, high-performance automation could greatly reduce the number of sound files visually inspected by classifying nests as

83 inactive where no cockatoo calls are present. Improvements to recognition methods should, therefore, be explored further.

Black-cockatoo nesting is well suited to bioacoustic methods of monitoring. The birds vocalise loudly, at predictable times each time, and at fairly consistent distances from the sound recorder. In this study, we expanded upon the existing knowledge of the nest- associated vocalisations of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo (Chapters 3 and 5) by identifying that fledging is vocally indicated. Further, using simple open-source call recognisers in combination with manual verification, we were able to detect daily nest activity and nest outcome (fledging or failure) at 44 known nests of both subspecies. For these reasons, we conclude that bioacoustics can be an appropriate tool for nest monitoring in these endangered subspecies. As bioacoustic technology continues to advance and become more affordable, it is foreseeable that large numbers of sound recorders could be deployed at both known and potential nests for these subspecies. The method could also be applied to other less-studied populations of glossy black-cockatoo and red-tailed black-cockatoo. This would allow, for the first time, a comprehensive understanding of nest use across large spatial areas for these species. Lastly, this work demonstrates the utility of a different approach to recogniser testing, whereby the species’ behaviours and research questions directly inform the level at which performance is measured. Applying this to other species and issues will allow for more practical discussions of bioacoustics’ utility to wildlife monitoring.

Acknowledgments

We would like to thank the south-eastern red-tailed recovery team and the Kangaroo Island glossy black-cockatoo recovery program. We especially thank Richard Hill and Mike Barth for help in the field. We also thank the Ecosounds Lab at the Queensland University of Technology, especially Prof. Paul Roe, Dr. Anthony Truskinger, Dr. Michael Towsey and Dr. Phil Eichinski. This work was supported by an Australian Postgraduate Award, the National Environmental Science Programme’s Threatened Species Recovery Hub and the Glossy Black Conservancy.

Day-level verification n days % days

(a) Correctly assigned nest activity (active or inactive) 39 61.9%

(b) Missed active nesting 2 3.2% RTBC (c) Incorrectly detected active nesting 22 34.9%

(a) Correctly assigned nest activity (active or inactive) 47 68.1%

(b) Missed active nesting 13 18.8% GBC

Time 1 Time 2 Time 3 % TP % TP % TP n mean n % TP n mean n % TP mean n % TP n total n nests nestling n total nestling n total n nests nestling nests per nest nestling nests per nest nestling per nest nestling + adults + adults + adults Fledged 5 375 5 75 3.7% 42.4% 812 5 162.4 35.1% 76.6% 181 5 36.2 0% 15.5%

Failed 9 396 9 44 0% 32.1% 439 8 54.9 0% 7.3% 385 7 55 0% 2.9%

RTBC Unsure 7 420 6 70 0% 23.1% 757 7 108 0.1% 38.2% 193 6 32.2 0% 17.1%

Total 21 1191 20 59.6 1% 32.2% 2008 20 100.4 14.2% 47.0% 759 18 42.2 0% 9.5%

Fledged 11 536 7 76.6 25.9% 28.9% 595 11 54.1 33.4% 46.7% 299 6 49.9 13.0% 15.4%

Failed 9 102 6 17 0% 25.5% 33 6 5.5 0% 12.1% 116 6 19.3 0% 3.4%

GBC Unsure 3 154 3 51.3 0% 1.3% 182 3 60.7 8.2% 26.9% 70 2 35 0% 27.1%

Total 23 792 16 49.5 18% 23.1% 810 20 40.5 26.4% 40.9% 485 14 34.6 8% 14.2%

Chapter 5 | Vocal ontogeny of nestling black-cockatoos, Calyptorhynchus banksii graptogyne and C. lathami halmaturinus

Abstract

Bioacoustic technology offers new ways to monitor wild animal populations. If we understand species’ vocalisations, and the factors that contribute to variability in call structure and vocal behaviour, then a wide variety of data are potentially available from sound recordings. Age-related changes can provide demographic data that are often valuable but difficult to collect for threatened species. In this study, we present the vocal ontogeny of nestlings for two endangered black-cockatoos, the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, C. banksii graptogyne. Using data collected remotely at wild nests, we examined changes in nestling vocalisations through to fledging. Nestlings vocalised from about 4 weeks of age, but calls were soft and infrequent until about 6 weeks. Daily call rate increased over time, especially in the final week of nesting. Peak amplitude and low frequency of nestling calls increased significantly with development. Call duration increased significantly for the glossy black-cockatoo, but not for the red-tailed black-cockatoo. Average entropy declined significantly for both subspecies. Aggregate entropy declined significantly for the red-tailed black-cockatoo but not the glossy black-cockatoo. Together, these changes in call rate and acoustic structure provide a useful way to broadly categorise nest age from sound recordings and thereby improve knowledge of nestling survival. This knowledge may be useful in future studies examining the influence of habitat variables, such as food availability, on nest development and nestling survival across landscapes. We conclude that vocal ontogeny can provide an additional source of demographic data in bioacoustic programs, as demonstrated at nests for the glossy black-cockatoo and red-tailed black-cockatoo.

Introduction

Vocal ontogeny, or changes in vocalisations with development, may be a useful indicator of demography in wild animal populations. With the increasing popularity of acoustic technology for wildlife monitoring, interest in animal vocalisations and bioacoustic methods is growing (Köhler et al. 2017; Servick 2014; Snaddon et al. 2013; Sugai et al. 2019). Bioacoustics is the study of wildlife sounds which, in the context of monitoring, typically refers to the use of sound recorders to collect animals’ vocalisations from which population data can be derived. Bioacoustic programs often focus on species’ most common

88 or conspicuous calls (e.g., for confirming species presence-absence), although the diversity of behavioural contexts indicated via vocalisations may also be of interest (Chapter 2). To acquire demographic data, studies can explicitly examine vocalisations from individuals of different age or sex. For example, in the leopard seal, Hydrurga leptonyx, age-related differences in call rate have helped describe differences in distribution among adults, sub- adults and juveniles (Rogers et al. 2013). Further, changes in vocalisations in early life may provide additional demographic detail, such as juvenile development and survival rates, that is otherwise difficult to obtain. To apply this to bioacoustics, we need to understand the changes in vocalisations that species show in this time period.

Since vocal ontogeny is the study of vocal changes through development, it is likely that changes reflect shifts towards more adult-like vocalisations. These changes may occur from a change in call structure (i.e. morphing of infant or juveniles calls into adult calls) or a shift in the call types that comprise the animal’s repertoire (i.e. losing infant or juvenile call types and acquiring adult call types) (Berg et al. 2013; Bond & Diamond 2005). Vocal changes may, in turn, relate to either behavioural or physical factors, such as vocal learning or body size. Anatomically, since vocal production is constrained by the size and shape of an animal’s sound-producing organs, we expect larger-bodied individuals to emit lower frequency vocalisations (Bradbury & Vehrencamp 2011; Fletcher 2004). This allometric relationship is established for many mammals (Bowling et al. 2017; Charlton et al. 2009; Ey et al. 2007; Wyman et al. 2012) and is also reported in some birds (Favaro et al. 2017; Galeotti et al. 1997; Hall et al. 2013; Martin et al. 2011; Potvin 2013; Tubaro & Mahler 1998). Considering ontogeny, it follows that the frequency of juveniles’ calls will decrease with development (e.g., see Stoeger-Horwath et al. 2007). However, for some birds, call frequency appears unrelated to body size (Cardoso et al. 2008; Digby et al. 2013), and may even increase with development in some species (Anderson et al. 2010). This departure from the tight allometric scaling of body size and song frequency is thought to be more likely in species with complex vocal production methods, such as songbirds (Patel et al. 2010). This may also apply to vocally-learning species, including those who maintain the ability to acquire and modify vocalisations for life, such as parrots (Bradbury & Balsby 2016).

Vocal ontogeny may also relate to habitat features, by way of their effect on infant and juvenile development. If growth rates are reduced, ontogenetic changes may also be reduced. In birds and mammals, many studies have shown begging to relate to hunger (Gladbach et al. 2009; Godfray 1991; Manser et al. 2008; Redondo & Castro 1992; Reers &

Jacot 2011). For example, fledglings of the Tangmalm’s owl, Aegolius funereus (Kouba et al. 2015), and the American dipper, Cinclus mexicanus (Middleton et al. 2007), beg at greater intensities in years when food is more limited. For a given species, if body size is vocally indicated then we might expect individuals with greater access to resources to have lower frequency vocalisations. Supporting this hypothesis, Sacchi et al. (2002) experimentally showed that peak frequency was lower for heavier nestlings of the barn swallow, Hirundo rustica. However, in chicks of the thin-billed prion, Pachyptila belcheri, Quillfeldt et al. (2010) found no relationship between frequency measurements and body condition (body mass relative to multi-year mean mass by age). Currently, the relationship between habitat factors, body size and vocalisations is not well understood, yet it warrants attention because it may provide a novel tool with which to monitor species’ responses to habitat variables and management actions.

In this study, we consider two endangered subspecies of black-cockatoo, the south- eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, and the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus. Both subspecies are habitat specialists and, therefore, highly vulnerable to habitat loss and degradation. The south-eastern red-tailed black-cockatoo feeds only on the seeds of stringybark, Eucalyptus aranacea and E. baxteri, its primary food source today, and buloke, Allocasuarina leuhmannii, which has been largely lost from the bird’s distribution (Russell et al. 2018). Fire and drought continue to threaten the remaining stands of stringybark (Koch 2003; Russell et al. 2018). The Kangaroo Island glossy black-cockatoo feeds exclusively on the seeds of the drooping she-oak, Allocasuarina verticillata, with rare exception (Berris et al. 2018). The loss of drooping she- oak on the mainland of South Australia has confined the birds to Kangaroo Island. Remaining habitat is also threatened by fire. Given these subspecies’ specialisation, we expect feeding habitat to have consequences for nesting, which may be vocally indicated. If so, bioacoustic data may indicate not only breeding behaviours and success (see Chapters 3 and 4), but also nestling development, the age of failed nests, and the relationship to food availability. This may help inform conservation planning to improve breeding, such as spatial prioritisation of revegetation works near breeding areas and locations for artificial nest hollow installations.

To successfully use bioacoustics as a tool to monitor habitat impacts on nestling development, a necessary first step is to understand how nestling vocalisations change through time. This study examined vocal ontogeny in nestlings of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Using daily sound recordings

90 collected autonomously from wild nests, we aimed to determine if nestling vocalisations show ontogenetic changes in acoustic structure and call rate over the nesting period. We hypothesised that nestling vocalisations would decrease in frequency during development. Further, we hypothesised that other acoustic measurements would gradually change over time to become more adult-like (see Chapter 3) and that daily call rate would increase over time as nestlings become more vocal. Altogether, using this knowledge we aimed to describe the vocal traits, whether structural or behavioural, that allow nests to be aged from sound data.

Methods

Study sites

Field work was conducted over two breeding seasons in 2017 and 2018. Data were collected from the Casterton region of western Victoria, for the red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, and from Kangaroo Island in South Australia, for the glossy black-cockatoo, C. lathami halmaturinus. For the red-tailed black-cockatoo, searches for active nests were undertaken in spring and summer. Nest searching was targeted towards areas with historical nest records, which were most often livestock farms. For the glossy black-cockatoo, data were collected from nests monitored over autumn and winter as part of the state government’s long-term Glossy Black-Cockatoo Recovery Program. Nests were located on, or near to, livestock farms, roadsides, remnant and non-remnant forests, conservation parks, timber plantations and in residential areas. No preference was given to natural or artificial nest hollows for either subspecies.

Sound data collection

An autonomous sound recorder (Frontier Labs Bioacoustic Audio Recorder, https://frontierlabs.com.au/) was installed at each nest tree. The distance of the sound recorder to the nest hollow varied from approximately 8 – 30 metres. Recorders remained in place until fledging or nest failure. Recorders were scheduled to record for three continuous hours per day, finishing 30 minutes after sunset (sunset-based schedule). Once a week, recording began 30 minutes before sunrise and finished 30 minutes after sunset (full-day schedule). In attempt to record the fledging event, if a large nestling was observed at a nest, some recorders were updated to operate at the full-day schedule every day. All sound recorders used an omnidirectional microphone, with a fixed gain of 20 dB and a sample rate of 44.1 kHz. Microphones had an 80 Hz high-pass filter to reduce the effects of low frequency background noise, such as wind. Recorders were fitted with four rechargeable

91 lithium ion batteries and one 128 GB SanDisk memory card, which were replenished at approximately 6-week intervals. All recordings were made in uncompressed wave (.wav) format.

Spectrographic measurements

Spectrographic measurements were taken of nestling vocalisations from nests at which the date of fledging was known. Selections of vocalisations were made from spectrograms (Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap) using Raven Pro 1.5 software (Cornell Lab of Ornithology, Ithaca, New York). We selected one vocalisation per day for the entire time period that nestling vocalisations were recorded. We chose to sample one vocalisation per day to ensure that each day was evenly represented in the data; from earlier field observations we noted that nestlings appeared to vocalise infrequently until closer to fledging. Since nestlings were observed in the field to vocalise for several weeks, we considered that samples comprising one selection per day would sufficiently capture individual variation (Fischer et al. 2013 found 20 calls per individual was sufficiently representativeness in primates). Where possible, only the loud form of the nestling call (subtype 1; see chapter 3) was selected. We attempted to select only vocalisations that were clear and free of background noise. If this was unachievable for a given day, we selected the clearest vocalisation possible and classified it as ‘distorted’ so that it could be excluded from some analyses (see below). In all cases, we only selected calls for which the low frequency bound was clear. Vocalisations were not selected on days that were deemed too noisy, such as in rain, heavy wind or when heavily masked by other species’ calls.

Since the date of laying or hatching was not known, nests were aged backwards relative to the date of fledging. We classified the day of fledging as “Day 0” and every preceding day relative to that; e.g., the day before fledging was classified as “Day 1”. For each selection, acoustic measurements calculated were low frequency (minimum frequency bound of the selection; units Hz), average entropy (a measure of disorder in the sound, measured as the average for each time slice in the selection; units bits), aggregate entropy (disorder measured across the selection; unit bits), call duration (also termed delta time; units seconds) and peak amplitude (the greatest absolute value in the selection; units u are dimensionless relative to a reference point). Entropy measurements were included because nestling calls were previously found to be relatively high in entropy (Chapter 3), which likely

92 explains their “harsh” sound. Amplitude was included because prior work (Chapter 3) found late-stage nestling calls to be surprisingly loud. Duration was included because late-stage nestling calls are similar in duration to adult flight calls (Chapter 3) but whether this is the case for younger nestling is not known.

Additionally, to examine changes in call rate over time, for each nest we randomly selected one day per week in which to count every nestling vocalisation in the 3-hour sunset- based recording. Calls were counted manually from spectrograms. All forms of nestling call, including soft calls, were counted. Since soft nestling calls can be difficult to distinguish from adult female nest calls (Chapter 3), we did not include any call in which the caller’s identity was uncertain. Counts of nestling calls are therefore likely to be somewhat underestimated.

Statistical analysis

Data were analysed using R statistical language (R Core Team 2019). Acoustic measurements (low frequency, aggregate entropy, average entropy, duration and peak amplitude) were inspected for normality. Peak amplitude was log-transformed. To examine changes in nestling vocalisations over time, we used linear mixed models, implemented with the lme4 package (Bates et al. 2017), with acoustic measurements as the response variables, day as the fixed explanatory factor and individual nest ID as a random factor. To examine the change in call rate over time, we used generalized linear models with call count as the response variable and week of development as the categorical explanatory variable. We then used a post-hoc Tukey test, implemented in the multcomp package (Hothorn et al. 2008), to compare call rate among weeks of development. Plots were made using the ggplot2 package (Wickham 2016).

Results

For the south-eastern red-tailed black-cockatoo, the date of fledging was known and recorded for three nests. Successful fledging was suspected for a further two nests, but the fledging event was not recorded and, therefore, the date of fledging was unable to be determined. Nine nests failed, and nest outcome was undetermined for a further nine nests. Nest outcomes were undetermined either because the sound recorder’s batteries expired sooner than predicted, or because the recording schedule malfunctioned. Of the 23 nests located, all were on livestock farms except for one that was in an artificial nest hollow on a timber plantation. In total, four nests were in artificial nest hollows. Natural hollows were all in river red gums, Eucalyptus camaldulensis. For the Kangaroo Island glossy black-cockatoo,

93 the date of fledging was known for eight nests. At one of these, the fledging event was not recorded but a field inspection on the day after the last nestling vocalisations were recorded revealed that the nest hollow was empty. This suggests that fledging occurred outside of the recording period, such as later in the evening or early the following morning. Fledging was suspected for a further four nests. Nine nests failed and the outcome was undetermined for four nests. All the 28 nests monitored were in sugar gums, E. cladocalyx, including artificial nest hollows.

Vocal ontogeny

For both subspecies, the maximum length of time that nestling vocalisations were detected in recordings was 8 weeks, or from approximately 4 weeks of age, assuming a 12- week nesting period (Higgins 1999). However, at many nests, vocalisations were relatively rare earlier than about 6 weeks before fledging, with many days absent of nestling vocalisations. The shortest time period that nestlings were detected was 17 days, in both subspecies. Nestlings were reliably detected most days from about 4 – 5 weeks before fledging, although calls were somewhat muffled until the nestling began sitting at the nest hollow entrance.

Changes in daily call rate

Examining changes in call rate over the nesting period, we found a marked increase in daily call rate in the final week of nesting (Fig. 5.1). For the Kangaroo Island glossy black- cockatoo, call rate was significantly higher in the final week of nesting than in weeks 1 through to 6 (weeks before fledging) (week 1, z-value = 3.8, p < 0.01; week 2, z-value = 4.9, p < 0.001; week 3, z-value = 3.5, p < 0.05; week 4, z-value = 3.9, p < 0.01; week 5, z-value = 3.7, p < 0.01; week 6, z-value = 3.7, p < 0.001). No significant differences were found for the south-eastern red-tailed black-cockatoo.

Changes in acoustic structure

Nestling vocalisations changed in acoustic structure with development (Fig. 5.2). Peak amplitude increased significantly over time (t-value = -7.878, p < 0.001 for the glossy black-cockatoo; t-value = -5.877, p < 0.001 for the red-tailed black-cockatoo) (Fig. 5.3a). There was a significant increase in call duration for the glossy black-cockatoo over time (t- value = -8.909; p < 0.001), but not for the red-tailed black-cockatoo (t-value = 1.34, p > 0.05) (Fig. 5.3b). For both species, there was a significant increase in low frequency over time (Fig.

5.3c) (t-value = -7.63, p < 0.001 for the glossy black-cockatoo; t-value = -2.078, p < 0.05 for the red-tailed black-cockatoo). Average entropy and aggregate entropy declined significantly over time for the red-tailed black-cockatoo (t-value = 3.067, p < 0.001 for average entropy; t- value = 2.112, p < 0.05 for aggregate entropy). Average entropy declined significantly for the glossy black-cockatoo, but there was no change in aggregate entropy (t-value = 2.717, p < 0.01 for average entropy; t-value = -0.505, p > 0.05 for aggregate entropy) (Fig. 5.3d and 5.3e).

Discussion

This study aimed to describe vocal ontogeny in nestling black-cockatoos, to enable the collection of more diverse data from bioacoustic programs. For the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, we found that nestling vocalisations undergo ontogenetic changes during development, which may be used to estimate nest age from sound data. Nestling vocalisations are first detected in sound recordings from about 4 weeks of age (i.e., 8 weeks prior to fledging). However, early vocalisations are soft and can be difficult to detect. From about 6 weeks, louder vocalisations can be detected and, from 7 – 8 weeks, are present most days. The presence of loud vocalisations is shown by an increase peak amplitude with development in both subspecies. In addition to anatomical growth, amplitude increases are expected from changes in nestling behaviour, from remaining inside the hollow when young, to regularly sitting at the nest hollow entrance when older. This would substantially increase the sound amplitude received by a recorder, and this is clearly noticeable in spectrograms and recordings. This increase, therefore, is a good indicator of a large nestling.

Nestlings become more vocal with age, particularly in the week of fledging. Daily call rate increased markedly in the final week of development for both subspecies, and this was significant for the glossy black-cockatoo. This suggests that call rate could indicate nest survival into the final week of development. Although fledging vocalisations can provide a direct measure of a fledging event (Chapter 4), and therefore nest success, if this specific event is not recorded, high call rates may serve as a suitable alternative. More importantly, if nestling vocalisations cease in recordings and an increase in call rate is not observed, then nest failure should not be dismissed as a possible nest outcome, even if a large nestling was recorded. This presents an advantage over traditional methods used for the Kangaroo Island glossy black-cockatoo, which tend to overestimate nest success because fledging is assumed

95 if a large nestling is observed in the field (M. Barth, pers. com.). Although survival of large nestlings is expected to be high, there is evidence that even late nests can fail (J. Pepper, M. Barth, pers. com.). Moreover, this reduces the need for multiple follow-up surveys by field staff, and also presents the first viable method for aging nests of the south-eastern red-tailed black-cockatoo, for which intensive field monitoring has not been possible.

Together, the observed changes in amplitude and call rate provide a useful method with which to broadly categorise nest age, specifically (a) egg or young nestling, up to approximately four weeks (no nestling calls); (b) young nestling up to approximately six weeks (soft calls, infrequent, from inside the hollow); (c) middle-aged nestling up to about 10 or 11 weeks (mostly soft calls, some infrequent loud calls, often distorted from being inside the hollow, becoming more frequent); (d) late-stage nestling from 11 weeks through to fledging (many clear loud calls, some soft calls, high call rate especially noticeable in final week). Given the variation observed, it does not appear that nests could be confidently aged to the day, or even the week (with the exception of the final week of nesting when call rate increases markedly), using the measurements examined here. However, for the purposes of conservation monitoring, broad categorisation is likely to be sufficient; for example, to distinguish unviable eggs or unfit hatchlings (early stage failure, which may indicate inbreeding depression) from other potential causes of failure (e.g., predation in late stages). In this sense, it is the combination of quantitative (e.g. call rate) and qualitative (e.g. level of distortion from being inside the nest hollow) observations that would allow for nests to be broadly, but confidently, aged.

For the glossy black-cockatoo, call duration may also be useful for categorising nest age. Duration increased significantly with nest development for the glossy black-cockatoo, while for the red-tailed black-cockatoo there was no change. These results support our hypothesis that vocalisations become more adult-like over time, since adult calls are relatively long in the glossy black-cockatoo but not the red-tailed black-cockatoo (Chapter 3). This also supports findings from other parrot studies, such as the green-rumped parrotlet, Forpus passerinus, in which nestling vocalisations increase in duration with development but then decrease to become adult-like by the time of fledging (Berg et al. 2013). In our study, the changes may represent a shift in repertoire, rather than acoustic structure per se. Nestlings of both subspecies have a loud form (subtype 1) and a soft form (subtype 2) of vocalisation (Chapter 3). In the glossy black-cockatoo, the soft form is substantially shorter than the loud form, which is similar in duration to adult flight calls. For consistency we attempted to

96 measure only loud calls but calls from younger nestlings were characteristically soft. As such, the increase in call duration, as well as peak amplitude, likely represents a shift from a repertoire comprising only the soft form of call to one that includes both soft and loud forms. In the red-tailed black-cockatoo, the soft form is only marginally shorter in duration than the loud call, therefore any change in duration was expected to be minimal. This is similar to the kea, Nestor notabilis, an alpine parrot endemic to New Zealand, whose juvenile (not nestling) calls do not appear to be less-developed morphs of adult calls, but rather distinct, socially- learned calls that are shorter in duration and more variable in structure (Bond & Diamond 2005).

For both subspecies, low frequency changed over time but, unexpectedly, there was a positive correlation between low frequency and day of development. Assuming an increase in nestling body size with time, this contrasts many studies that show larger-bodied animals to have lower frequency vocalisations (Bradbury & Vehrencamp 2011; Ey et al. 2007; Favaro et al. 2017; Liu et al. 2017; Martin et al. 2011; Miyazaki & Waas 2003; Potvin 2013; Tubaro & Mahler 1998). However, despite being widely referenced in the literature, this relationship has, in recent reviews, been suggested to be more complex than is often acknowledged and, in adult animals, can be subject to strong sexual selection (Charlton & Reby 2016; Garcia et al. 2017; Patel et al. 2010; Rodríguez et al. 2014). In birds, it appears that species with more complex vocalisations are less-tightly bound to allometric scaling (Patel et al. 2010). For example, the sophisticated vocal production methods of songbirds are thought to contribute to the weak or non-existent relationship of body size and frequency observed in some species (Cardoso et al. 2008). Black-cockatoos are parrots, social birds with complex repertoires and lifelong abilities to acquire and modify vocalisations (Bradbury 2003; Bradbury & Balsby 2016). This plasticity probably contributes to more variable relationships between anatomy and acoustic structure of vocalisations. Indeed, our results are similar to findings for the kea, where call frequencies of the kee-ah call are higher in adult calls than juvenile calls (Bond & Diamond 2005). Since parrots learn calls from their conspecifics, we expect that nestlings and juveniles acquire their vocalisations by copying individuals with whom they make contact; in the kea, this pattern manifests in geographically-distinct juvenile vocalisations (Bond & Diamond 2005). In both black-cockatoo species, only the adult female enters the nest hollow, from whom nestlings probably learn their early vocalisations. This supported by the evidence that the female nest call is similarly low in frequency, as well as in apparent sound to the human ear, especially in the glossy black-cockatoo (Chapter 3). Nestling call

97 frequency may increase over time as older nestlings respond more to the higher-pitched adult flight calls.

We also examined nestling calls for changes in entropy through time. Entropy is a measure of nonlinearity or disorder in a sound; pure tones have low entropy while noisy, broadband calls have high entropy (Allen et al. 2018; Blumstein & Chi 2012; Fischer et al. 2016; Reers & Jacot 2011). Cockatoo calls are characteristically noisy or chaotic (Fletcher 2000) and those of nestling black-cocaktoos are especially so (Chapter 3). We expected entropy to decline with development, as adult calls contain less entropy than nestling calls (Chapter 3). Indeed, a significant decline in average and aggregate entropy was shown for the red-tailed black-cockatoo. For the glossy black-cockatoo, average entropy also declined significantly, but the change was less pronounced that than for the red-tailed black-cockatoo. There was no change in aggregate entropy. In interpreting these results, we need to consider that entropy may relate to two different, and possibly opposing, factors: age (as a function of anatomy) and call predictability (as a function of selection). For some species and call types, nonlinearities decrease with age, which likely relates to the poorer control of vocal production in younger animals (Berg et al. 2013; Fitch et al. 2002; Wilden et al. 1998). Alternatively, entropy may be adaptive by making calls more difficult to locate, as suggested for the high entropy, low amplitude nest calls of adult female kea (Wein et al. 2019). Entropy has also been associated with hunger (Boucaud et al. 2016; Klenova 2015; Reers & Jacot 2011).

In this capacity, the unpredictability hypothesis suggests that calls with more random structures (i.e., greater nonlinearity in acoustic structure; Wilden et al. 1998) prevent habituation and are, therefore, more likely to stimulate a response from conspecifics (Blumstein & Récapet 2009). To this end, we may see a decline in entropy with growth, as observed most clearly for the red-tailed black-cockatoo, while at the same time a level of selection to maintain entropy as a mechanism of unpredictability. Maintaining a level of entropy may reflect the need for young cockatoos to minimise habituation by parents, since young black-cockatoos are dependent on their parents for food for many months post- fledging. This may be the case for the glossy black-cockatoo, for which changes in entropy were less significant than the red-tailed black-cockatoo. Properly testing these hypotheses requires a careful experimental design that considers entropy in relation to food provisioning rates and body morphometrics.

Both subspecies are highly specialised in diet (Berris et al. 2018; Russell et al. 2018), for which reason we expect food to limit nesting in some capacity. Food availability may, therefore, be vocally indicated, as shown in adult zebra finches, Taeniopygia guttata (Ritschard & Brumm 2012; Zann & Cash 2008). In this study, we were unable to examine the influence of food availability on vocal ontogeny, as sample sizes were small and nests were spatially aggregated. Addressing this issue requires a larger sample size of nests across a greater range of nesting areas, ideally in a balanced experimental design. This would be interesting for future bioacoustic studies to investigate because if food availability is vocally indicated, this would provide a tool with which to examine nestling development in relation to habitat features. Notwithstanding, if habitat variables do not influence nestling vocal ontogeny, this should not be taken to mean that habitat does not influence nesting. Habitat effects may manifest in the populations in other ways. For instance, poorer food availability may reduce the rate of fledging success or limit the number of nests supported by an area. Nests that are established may not vocally reflect poor food availability; for instance, older birds with greater knowledge of the landscape may be better at maintaining nests even if food is limited. This is partly supported by evidence from Carnaby’s black-cockatoo, Calyptorhynchus latirostris, where older females have better nest survival rates (Saunders et al. 2016), although the influence on vocalisations is not known.

The results of this study rest on the assumption that the nests monitored represent the broader population. First, data used in this study came from successful nests, because to examine vocal ontogeny we needed to confidently age nests. However, it is possible that unsuccessful nests differed in their vocal development. Testing differences in ontogeny between successful and unsuccessful nests in wild populations is difficult because it requires that the date of laying or hatching is known. Future studies could potentially use in-nest cameras to confirm these dates. Until then, categorising nest age in bioacoustic programs for these subspecies assumes that changes in vocal structure and development are similar between successful and unsuccessful nests. Second, the vocal ontogeny observed in this study represents changes as recorded in a bioacoustic monitoring regime; specifically, sound recorders placed within 8 – 30 metres of nest hollows, active for three hours per day. This may differ from changes in vocalisations if soft, close-range calls were recorded. For instance, it is possible that nestlings are vocal earlier than the age at which they were detected in this study. We did not attempt to describe such changes, since the aim of this study was to determine ontogenetic changes that could be used to inform nest age in a bioacoustic

99 monitoring program. The results should be interpreted in this context. Since black-cockatoos often vocalise loudly, bioacoustic programs should not need to use in-nest recorders to collect data relevant to conservation, such as nest survival or age of failure. Bioacoustics is a relatively non-invasive tool for monitoring, and in this study we demonstrate that demographic data can be collected from nests with minimal disturbance.

Advances in bioacoustic technology are paving new ways for conservation monitoring, allowing for more data to be collected with less human involvement (Browning et al. 2017; Darras et al. 2019). Bioacoustics is especially useful for rare or cryptic species, or those that are otherwise constrained by resources or logistics. Sound recordings can provide a wide variety of data, such as species presence-absence, density and behaviour, in addition to environmental metrics like the Acoustic Complexity Index (a measure of a soundscape’s complexity) and rainfall (Darras et al. 2016; Sugai et al. 2019). In this study, we show that bioacoustics can provide demographic data for nestlings of two endangered bird subspecies, the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. There are various constraints on monitoring and management for these birds, and bioacoustics offers a new way to address some of these. By providing demographic data from nests, in addition to other important metrics like nest survival (Chapter 3), bioacoustics can help inform management decisions, such as the location for artificial nest hollows and surrounding food habitat restoration. Bioacoustics has much to offer conservation monitoring and, in this study, we provide an additional source of knowledge about nest-associated vocalisations in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo, from which larger-scale monitoring programs could benefit.

Acknowledgements

For help in the field, we thank Richard Hill from the south-eastern red-tailed recovery team and Mike Barth from the Kangaroo Island glossy black-cockatoo recovery program. We thank the landowners who allowed this research on their properties. We thank the members of the Ecosounds Lab at the Queensland University of Technology for help with sound data handling and storage. We thank Dr. Simone Blomberg from The University of Queensland for statistical help. This project was supported by an Australian Postgraduate Award, the National Environmental Science Programme’s Threatened Species Recovery Hub and the Glossy Black Conservancy.

100

Fig 5.1: Box plot of nestling call rate (count of vocalisations per day) by week of development for the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne (RTBC; n = 3 nestlings) and the Kangaroo Island glossy black-cockatoo, C. lathami halmaturinus (GBC; n = 7 nestlings). Both subspecies show a marked increase in nestling call rate in the final week of nesting (F = 4.307, p = 0.055 for the south-eastern red- tailed black-cockatoo; F = 8.486, p = 0.006 for the Kangaroo Island glossy black-cockatoo). Solid lines represent weeks when data were only available from a single nest.

101

(a) Kangaroo Island glossy black-cockatoo

44 37 22 16 9 0 54 29 (b) South-eastern red-tailed black-cockatoo

53 44 37 29 22 16 9 0 Days before fledging Fig 5.2: Spectrograms showing nestling vocal ontogeny of the (a) Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and (b) south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne. Each species is represented by a single nest from approximately 5 weeks of age (Day 53 and 54) through to the day of fledging (Day 0). Early vocalisations tend to be soft, but older nestlings commonly give both loud and soft calls. Loud forms are shown here.

102

Glossy black-cockatoo Red-tailed black-cockatoo

a *** ***

*** b

* c ***

* d

*** ** e

103

Chapter 6 | Thesis synthesis and conclusion

Thesis overview

This thesis presents a bioacoustic method for nest monitoring of the endangered Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne. The methods presented have been developed and tested under natural conditions over two breeding seasons. For both subspecies, fieldwork was conducted in areas that have historically supported nesting. These methods are, therefore, considered to be broadly representative of the birds’ breeding behaviours, vocalisations and acoustic environment.

My overall aim in this thesis was to develop and validate bioacoustic methods to confirm daily nest activity and nest outcome. Critically, this work was undertaken to maximise the utility of bioacoustics to ecologists responsible for monitoring these subspecies of black- cockatoo. Through collaboration with relevant end-users, I achieved this by completing a detailed study of the subspecies’ vocal behaviours at nests, by collecting sound data in a regime appropriate for the birds’ behaviours, and by developing and testing open-source call recognisers that directly address current research questions. The major outcomes of this research are summarised as follows:

• A critical review of bioacoustics’ potential to improve monitoring of animal behaviour for conservation. I discuss how vocalisations can indicate not only a species’ presence, but also the behavioural contexts of individuals. Such nuances offer a greater range conservation-relevant data to be acquired from species’ vocalisations. • Description of the nest-associated vocal behaviours of the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. Vocalisations described are those that can be easily monitored in a bioacoustic nest monitoring program, negating the need for any in-nest data collection. I provide the quantitative acoustic structure of all vocalisations, as well as detailed descriptions of each vocalisation’s behavioural context. • Description of the vocalisations associated with fledging in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. The pattern of vocal behaviour at fledging is unique and provides the first direct measure of fledging in these subspecies.

104

• Description of ontogenetic changes in nestling vocalisations through to fledging. This provides a novel way to age nests from sound data. • A verified call recogniser to detect nest activity and fledging in the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. The recogniser is open-source, fully scripted and annotated, and easily implemented in R statistical language.

In Chapter 2, I reviewed the bioacoustic and animal behaviour literature to ascertain the status of the discipline I termed acoustic conservation behaviour. This combines conservation behaviour, defined as the application of behavioural knowledge to conservation solutions, and bioacoustics, the study of animal sound. Since bioacoustics is still gaining traction as a wildlife monitoring tool, much of its focus in the literature has been on species detection, typically via their most common or conspicuous vocalisations. Few studies have directly applied bioacoustics to behavioural monitoring, especially for conservation purposes, despite a long interest in animal vocalisations such as birdsong (Bradbury 2003; Catchpole & Slater 2008). In the review, I examined how knowledge of vocal behaviours, both existing and potential, could be used to improve conservation monitoring. Specifically, I discussed vocalisations associated with reproduction and recruitment, alarm and defence, and social behaviour. I concluded that animal behaviour has much to offer conservation, and bioacoustics provides a novel tool with which to collect useful data. This requires that species’ vocalisations are described for the behaviours of interest, which I addressed for glossy and red-tailed black-cockatoos in Chapter 3.

In Chapter 3, I provided descriptions of the nest-associated vocalisations of the Kangaroo Island glossy black-cockatoo and the red-tailed black-cockatoo. Using sound recordings collected remotely at nests, and verified by field observations and video footage, I identified vocalisations for six behavioural categories: in flight, while perched, during begging (adult females), during courtship displays (adult males), when entering or sitting near to the nest hollow entrance (adult females), and from nestlings. The female nest call and nestling calls are the most conspicuous indicators of nesting and are, to the best of my knowledge, unique to active nests. These vocalisations, therefore, are those most useful for bioacoustic nest monitoring. Nestling vocalisations are especially useful because they are often loud and reliably heard most days. Nonetheless, other adult vocalisations can also provide a range of data on nest activity and behaviour. For example, future programs may deploy sound recorders at potential nests to examine nest usage or uptake of artificial hollows. In such

105 cases, adult vocalisations could indicate nest prospecting behaviour, in addition to any subsequent nesting. This is also important to gauge early nest activity when nestlings are not yet heard in sound recordings, which I discussed in Chapter 5.

Vocalisations were described from sound recordings collected near active nests, most often from recorders attached to the nest tree at 1.5 – 2.5 metres above ground. The vocalisations described are, therefore, those available in a bioacoustic program that avoids the need for tree-climbing and in-nest monitoring. This approach was taken to determine if appropriate data could be collected with limited human involvement. This was critical for the south-eastern red-tailed black-cockatoo, since most nests occur in dead trees that are unsafe to climb (R. Hill, pers. com.). While tree-climbing has been routine in the nest monitoring of the Kangaroo Island glossy black-cockatoo, this has required substantial resource commitments (human, time, and financial). Indeed, since 2017, funding reductions have seen most nest monitoring precluded in favour of more critical nest maintenance works (e.g., repairing predator-exclusion tree collars) (K. Berris, pers. com.). With the methods presented in this thesis, bioacoustics could now largely replace human-based nest monitoring, thus providing more funds for other management actions.

In Chapter 4, I examined the practicalities and performance of a bioacoustic nest monitoring program, conducted under natural conditions, for the Kangaroo Island glossy black-cockatoo and the south-eastern red-tailed black-cockatoo. In this chapter, I described the vocal signature of fledging, which provides the first efficient method for directly monitoring nest outcome in these subspecies. I then developed and tested call recognisers for their ability to detect nest activity throughout the nesting period, over several breeding seasons. I also tested their ability to detect a successful fledging event. I used binary point matching, a template-matching recogniser, implemented in the monitoR package in R statistical language (Katz et al. 2016; Hafner & Katz 2017; R Core Team 2019). The recogniser for each subspecies was constructed on templates of nestling calls. To construct the recognisers, I first undertook a pilot study to select suitable high-performing templates of nestling calls (i.e., templates that returned most true positive detections), as well as a score threshold that efficiently balances true and false positives. I then tested the recognisers on sound recordings collected from successful and unsuccessful nests of both subspecies (n = 23 nests for the Kangaroo Island glossy black-cockatoo; n = 21 nests for the south-eastern red- tailed black-cockatoo). For each nest, I tested the recogniser on 3 x 3-hour sound recordings collected in early, mid and late stages of the recording period. I chose this approach because

106 it determines a nest’s status (active or inactive) broadly across the nesting period, which can last four months, following which additional survey days can be examined to determine date of fledging or failure. This is more efficient, in the first instance, than examining every recording day (e.g., if nests failed early).

In the literature, recogniser performance is usually reported at the level of the individual call. Common performance metrics are the proportion of detections that are true positives (precision) and the proportion of available calls that are detected (recall) (Knight et al. 2017; Priyadarshani et al. 2018). However, for black-cockatoo nest monitoring, performance at the level of the individual call is less important than the survey day. Specifically, it is important that if a nest is active, at least one true positive detection is returned for the survey day, and if a nest is not active, that no detections are returned. Detecting all calls is not important. For this reason, I used the survey day as the primary performance unit. With this approach, daily nest activity was correctly assigned as active or inactive in 61.9% of survey days analysed (n = 63 days) for the red-tailed black-cockatoo, and 68.1% of survey days (n = 69 days) for the glossy black-cockatoo. Errors mostly occurred from false positive detections. Indeed, the recognisers returned 2,087 detections returned for the Kangaroo Island glossy black-cockatoo and 3,958 detection for the south-eastern red-tailed black-cockatoo, of which only 18.8% and 7.6% were nestling calls, respectively. When adult calls were considered to be correct detections, recogniser performance improved to 35.3% and 27.9% in the red-tailed and glossy black-cockatoo, respectively. In this regard, improving recogniser precision would reduce the number of survey days in which nests are incorrectly deemed active. The recognisers successfully detected fledging at five out of six events recorded for the glossy black- cockatoo, and two out of three events for the red-tailed black-cockatoo. I concluded that, while the recognisers are useful in confirming nest activity, some manual verification is still required; thus, this can be considered a semi-automated method.

In Chapter 5, I examined vocal ontogeny of nestlings, as a method with which to age nests from sound data. I took daily measurements of nestling calls for every nest at which the date of fledging was known. I determined that nestling vocalisations are recorded from about 4 weeks of age but were common from about 6 weeks. Daily call rate increased with development, particularly in the week of fledging. Peak amplitude also increased over time. Altogether, these changes provide a method with which to broadly categorise nest age, specifically (a) egg or young nestling, up to approximately 4 weeks (no nestling calls); (b) young nestling up to approximately 6 weeks (soft calls, infrequent, from inside the hollow);

107

(c) middle-aged nestling up to about 10 or 11 weeks (mostly soft calls, some infrequent loud calls, often distorted from being inside the hollow, becoming more frequent); (d) late-stage nestling from 11 weeks through to fledging (many clear loud calls, some soft calls, high call rate especially noticeable in final week). Overall, I concluded that nests can be broadly aged from sound data, and thus provide an additional source of demographic information in bioacoustic monitoring programs. Future studies could benefit from this knowledge by examining how nest development relates to important habitat features, such as food availability.

The future of bioacoustic monitoring for conservation

Bioacoustics offers huge potential for a wide range of data to be collected in conservation monitoring programs (Browning et al. 2017; Shonfield & Bayne 2017; Sugai et al. 2019). For animals with diverse vocal repertoires, bioacoustics can provide detailed, context-specific data that can aid conservation. However, this approach is currently underutilised, with most applications focussing on species detection (presence-absence) (Frommolt & Tauchert 2014; Sebastián-González et al. 2015; Wrege et al. 2017; Dema et al. 2018; Schroeder & McRae 2020) and, more recently, metrics like abundance, density and occupancy (Dawson & Efford 2009; Marques et al. 2013; Borker et al. 2014; Furnas & Callas 2015; Stevenson et al. 2015; Campos-Cerqueira & Aide 2016; Chambert et al. 2018; Pérez-Granados et al. 2019). There are still many limitations to large-scale bioacoustic programs; for behavioural studies, a lack of knowledge on the contexts of a species’ vocalisations and methods with which to distinguish these in sound recordings remain a challenge. In this thesis, I examined the potential for bioacoustics to provide data on breeding behaviours of two threatened subspecies of black-cockatoo. To this end, I described the nest- associated vocalisations of both subspecies examined, including the vocal ontogeny of nestlings, and developed call recognisers to detect active nesting and fledging events.

Automated species detection is a priority for improving bioacoustics’ utility as a wildlife monitoring tool (Sugai et al. 2019). Currently, even where species’ vocalisations are well understood, detecting them in sound recordings can be extremely difficult. Recognisers are highly diverse, both in their underlying construction and their implementation. To be useful in conservation, recognisers need to be high-performing and easily implemented. I examined these issues for the glossy black-cockatoo and the red-tailed black-cockatoo, by building and testing open-source recognisers in the R statistical language, a program popular

108 among ecologists. This was done in consultation with conservation end-users, and the final recognisers are relatively straightforward to implement for those familiar with basic programming in R. However, their performance is moderate, with many false positive detections. This is a common issue for recognisers as natural soundscapes can be highly complex. The influences of wind, rain, other species’ calls, as well as the amplitude of the target species’ calls, can all contribute to false detections (Brandes 2008; Towsey et al. 2012; Cragg et al. 2015; Salamon et al. 2016; Crump & Houlahan 2017; Knight et al. 2017; Priyadarshani et al. 2018). Complex algorithms may improve performance, but these are largely the domain of expert computer programmers and are, therefore, often beyond the scope of conservation programs. Currently, in choosing a recogniser, conservation practitioners must trade-off performance, ease of implementation and cost (Priyadarshani et al. 2018). Commercial software provides the most user-friendly options, but these have financial implications. Open-source recognisers are freely available, but require a level of programming experience. Additionally, both commercial and open-source options are limited in the number of algorithms available and, therefore, the options to manipulate performance. These are significant problems and solutions are necessary for bioacoustics to progress as a conservation discipline.

Nonetheless, as acoustic technologies and post-processing methods continue to advance, the scope of research questions able to be examined will increase. The scale at which acoustic data can now be collected is unprecedented (Towsey et al. 2014; Shonfield & Bayne 2017; Sugai et al. 2019). Continued advancements in technology will undoubtedly see this continue, placing bioacoustics firmly in the realm of big data. It is foreseeable that large- scale networks of acoustic sensors will serve multiple research purposes and may, in some cases, negate case-by-case monitoring programs. More than ever, acoustic programs will be interdisciplinary. To ensure conservation objectives are met, bioacoustic practitioners should work collaboratively with conservation end-users when designing monitoring programs. Conservation end-users should, in turn, work with animal behaviourists to establish libraries of species’ vocalisations that help answer conservation questions. These libraries should not only capture the diversity of behavioural contexts, but also demographic variations. Moreover, the acoustic signatures of threats warrant greater attention. This includes non- biological sounds (e.g. gunshots; Wrege et al. 2017), as well as vocalisations of invasive species (e.g., cane toads, Rhinella marina; Brodie et al. 2020) and referential alarm calls of prey species (i.e., calls that are unique to a predator type; Thorley & Clutton-Brock 2017).

109

Sophisticated recognition methods capable of distinguishing many species, their behavioural and demographic contexts, as well as threats, could see acoustic methods revolutionise wildlife monitoring.

Conservation significance of bioacoustics for nest monitoring of black-cockatoos

The methods developed in this thesis can improve nest monitoring of glossy black- cockatoos and red-tailed black-cockatoos, including subspecies not explicitly examined here (see Chapter 1 for a list of all subspecies of glossy and red-tailed black-cockatoo). Assuming that other subspecies’ behaviours and vocalisations are similar, the methods should be directly transferrable to other populations. This is important as both species are threatened in other locations, and routine nest monitoring is lacking. For populations that are monitored, most data are collected from flocks (typically roost sites), as these are most easily observed. While the demographic structure of flocks can indicate breeding output, it alone cannot identify the habitat features that drive breeding; thus, management decisions to improve breeding are limited.

By directly monitoring nests, bioacoustics offers a new approach to understand breeding success across landscapes. Such data are unavailable from flock data alone. Critically, bioacoustics can help determine not only where nesting occurs in the landscape, but also how success varies between locations. If implemented on a large scale, bioacoustic nest monitoring could help address an important conservation question: how does food availability influence breeding success? Other landscape features may also be important. This is important in spatially prioritising management actions, such as prescribed burning in feeding habitat and artificial nest hollow installations. Since glossy and red-tailed black- cockatoos are dietary specialists, variations in food availability are thought to impact breeding. Fire is a major concern for both species as it can dramatically reduce the extent and quality of food available. For stringybark, Eucalyptus arenacea and E. baxteri, the main food of the south-eastern red-tailed black-cockatoo, fire impacts can persist for a decade, during which time the trees are unsuitable for feeding (Koch 2003). On Kangaroo Island, historical wildfires resulted in high mortality of drooping she-oak, Allocasuarina verticillata, the glossy black-cockatoo’s only food source (Pepper et al. 1993), and the cockatoos have never returned to some burnt areas (K. Berris, pers. com.). With climatic threats increasing, supporting breeding through strategic artificial nest hollow installations is likely to become more important. Understanding how food influences breeding is critical for choosing

110 locations that maximise breeding output. In this regard, bioacoustics could contribute important data for decision-making.

Recommended methods for bioacoustic nest monitoring

Based on the research presented in this thesis, as well as practical experience in the field, I recommend the following methods for bioacoustic nesting monitoring for the glossy black- cockatoo and red-tailed black-cockatoo:

Field methods

• Sound recorders: GPS-enabled recorders are desirable because they allow for sunset-based recordings. If GPS is not available, daily recording durations may need to be extended to cover the times of day when the birds are most active at nests, across the duration of the nesting period. Alternatively, sound recorders could be serviced more regularly to update the recording schedule. • Memory cards: I recommend using memory cards whose capacity aligns with the recorder’s battery life. In this study, I used 128 GB cards, but in all cases batteries expired before the memory cards were at capacity. Fast write speeds are necessary to ensure high quality sound recordings. • Recorder location: The clearest recordings were obtained from sound recorders that were closest to nests. As such, I recommend recorders be positioned on the nest tree, as close to the nest as possible. However, all relevant behaviours were sufficiently captured when positioned 1.5 – 2.5 metres above the ground and, therefore, there is no requirement to place recorders high in the nest tree. • Recording schedule: 3 continuous hours per day, concluding 30 minutes after sunset, sufficiently captured most breeding behaviours of interest. However, fledging was missed at several nests and may have occurred later in the evening, after the recording period ended. Therefore, I recommend recording for up to 1 hour after sunset. Although I collected full day recordings once a week, vocalisations were rare at all times except around sunset. Vocalisations could be heard early in the morning by these were often highly obscured by other species’ vocalisations. For the purposes of nest monitoring, sunset-based recordings appear to be sufficient.

111

I recorded at a sampling rate of 44.1 kHz, because I took structural measurements of calls. Lower sample rates would be sufficient to detect nesting activity and would extend the time period between servicing. If sound recorders allow for alternate day recordings, memory and battery life could be further extended by recording every other day early in the nesting period. Daily recording could commence later in the nesting period to capture fledging.

• Servicing: The frequency at which sound recorders need to be serviced depends on many factors including battery life, memory capacity, and the recording schedule used. In this study, recorders failed after approximately 7 weeks; thus, servicing was undertaken at approximately 6-week intervals to minimise the likelihood of missed recording days. As technology continues to advance, the time between servicing will inevitably increase and it is therefore difficult to generalise. Practitioners should consult manufacturers’ guidelines.

Post-processing

• Recogniser: In the first instance, the recogniser should be used on a subset of survey days that covers most of the recording period. In this study, I used the recogniser on 3 x 3-hour recordings from early, mid and late stages of the recording period. The recogniser’s outputs can then be used to inform subsequent survey days to be examined. For instance, if no true positive detections are returned from mid and late stages of the recording period, the nest is likely to have failed. If the date of failure is of interest, additional recordings between the early and mid-stage recordings could be examined; otherwise, no additional days need to be examined, thus reducing post-processing time. • Fledging: Fledging can be identified by inspecting spectrograms for the vocal pattern associated with the event (Chapter 3). Using the recogniser’s outputs, late- stage survey days can be progressively examined for fledging. Fledging can be further verified by inspecting proceeding survey days for nestling calls; an absence of nestling calls thereafter provides further verification that fledging occurred. • Verification: Recogniser outputs need to be manually verified. There are numerous options to achieve this. In this study, I imported the detections as a text file into Raven Pro 1.5 software for verification. This allows for detections to be

112

easily viewed on spectrograms and categorised for detection type (true or false positive), as well as any other relevant information (e.g. the type of false positive detection).

Limitations

This project faced several limitations that impacted some analyses presented in this thesis. The most significant of these is the small sample sizes of nests. In the first breeding season for the south-eastern red-tailed black-cockatoo, data were lost from several nests where sunset-based recordings failed. This malfunction meant that no fledging events were captured for this season, which reduced the number of nests examined for ontogeny (Chapter 5). Unfortunately, I was not aware of this problem until after the first breeding season ended, when sound data were collected and examined. Subsequent firmware upgrades rectified this issue during the second breeding season. Nonetheless, the data presented in this thesis are considered to be suitable because this research focussed largely on vocal patterns within nests. Specifically, I focussed on the diversity of vocalisations present at nests, the changes in nestling calls through development, and the performance of recognisers at detecting activity at nests. I did not test hypotheses that would have been severely impacted by small sample sizes or spatially-aggregated nests. As discussed in the relevant chapters, testing conservation-relevant hypotheses (e.g., the influence of food availability on vocalisations) requires a more careful experimental design.

Since data were collected repeatedly from nests, there is the potential that statistical assumptions of independence were violated. In Chapter 3, I used a linear discriminant analysis to classify call types. Standard methods for linear discrimination do not offer alternatives for violations of assumptions of independence. In behavioural research, it is common practice to repeatedly sample individual animals for the behaviours of interest (here, vocal behaviours), usually because of the substantial challenges of acquiring large independent sample sizes. This is particularly problematic for studies of wild animals, as exemplified in this study. Locating nests of these endangered cockatoos was difficult. Most nests did not survive to the age where the full range of vocalisations was present (e.g. nestling calls only occur in the final weeks of nesting). While this does not justify statistical violations, it highlights the limitations associated with real-world field work of rare species. I could have excluded linear discriminant analysis, making this study entirely qualitative, as is often the case in studies of vocal behaviour. However, I chose to include a quantitative

113 analysis in attempt to improve the robustness of this work and reduce subjectivity as much as I could. To capture as much variation as possible, I manually annotated over 2,500 calls as precisely as I could. I believe, therefore, that this is the best approach I could have taken with the data I had. Further, previous work examining the impact of violating assumptions of independence in discriminant analyses shows that its greatest risk is overstating discriminability; that is, you are more likely to make a type 1 error where behaviours are incorrectly deemed to be different (Mundry & Sommer 2007). In this study, vocal behaviours were often not well discriminated, especially for the red-tailed black-cockatoo; this was true even for vocalisations that sounded distinct. As such, I rely heavily on qualitative descriptions, since the overarching outcome of this work is to describe behavioural states (as a way to monitor nesting) and not to discriminate vocalisations, per se.

Bioacoustic studies typically require significant time investments in post-processing, before any analyses can take place. This was the case in this project. The time required to describe vocal behaviours (Chapter 3) from field observations, videos and spectrograms was substantial. This limited the number of annotations for each call type. It also meant that sample sizes were unbalanced for call types and nests. Vocalisations that are rare or soft were difficult to detect. As such, some vocalisations were not included in the statistical analyses presented. Sophisticated recognition methods may, in future, help address these issues.

Moreover, the analyses presented in Chapters 3 and 5 were performed on structural measurements of calls made by hand while viewing spectrograms. Accordingly, there will be a degree of human error associated with these measurements. This issue was mitigated in two ways. First, I personally made all measurements; thus, error should be consistent among the data. Second, spectrogram parametres were consistent for all measurements. Window size is most critical, since this trades off resolution between the time and frequency domains. I selected a window size of 1024 samples, which was applied to all spectrograms examined.

Finally, conducting research on wild animals under natural conditions presented many challenges. Most nests monitored did not survive to fledging, which limited analyses in Chapters 3, 4, and 5. In Chapter 3, I focussed exclusively on late-stage nests to describe vocal behaviours, as this time period included loud nestling calls that are largely absent from other time periods. Thus, the only nests able to be included were those that successfully fledged or failed late. In Chapter 4, although every nest was included in testing the recognisers’ performance at detecting daily nest activity, only those where fledging was recorded could be

114 tested for fledging detection. Likewise, in Chapter 5, vocal ontogeny could only be described from nests of known age, which could only be determined for those where the date of fledging was known. Lastly, throughout this project, severe weather impacted many survey days by obscuring or entirely masking vocalisations. In these cases, no acoustics measurements could be made. Nonetheless, these are true limitations of bioacoustic nest monitoring programs for these subspecies and are, therefore, important to consider.

Concluding remarks

Bioacoustic technology is changing the landscape of wildlife monitoring. Data can now be acquired from vocal taxa at scales formerly unthinkable. Though challenges remain, bioacoustics has certainly expanded our capacities to monitor species in ways that help address key conservation questions. By increasing survey effort with less human involvement, bioacoustics offers great potential for monitoring species, whether raucous and common, or rare and inconspicuous. As post-processing methods advance, the ability to monitor many species and their threats simultaneously, as well as responses to management, will be hugely advantageous to conservation agendas. In this thesis, I provided a framework for nest monitoring of glossy black-cockatoos and red-tailed black-cockatoos using autonomous sound recorders and semi-automated recognisers of breeding vocal behaviours. I focussed on the two most threatened subspecies, the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, for whom data on breeding has been difficult and costly to attain. Notwithstanding, the methods are likely transferrable to other subspecies of glossy and red-tailed black-cockatoo, some of which are also threatened. Applied on a large scale, bioacoustics could help address key knowledge gaps on the drivers of breeding success in these species, and thereby inform conservation decision-making. Black-cockatoo vocalisations provide a wealth of behavioural information and now, for the first time, can be utilised to achieve conservation outcomes.

115

References

Albornoz E.M., Vignolo L.D., Sarquis J.A. & Leon E. (2017) Automatic classification of Furnariidae species from the Paranaense Littoral region using speech-related features and machine learning. Ecological Informatics 38, 39-49.

Allen M., Webb M.H., Alves F., Heinsohn R. & Stojanovic D. (2018) Occupancy patterns of the introduced, predatory sugar glider in Tasmanian forests. Austral Ecology 43, 470- 5.

Alonso J.B., Cabrera J., Shyamnani R., Travieso C.M., Bolaños F., García A., Villegas A. & Wainwright M. (2017) Automatic anuran identification using noise removal and audio activity detection. Expert Systems with Applications 72, 83-92.

Anderson M. G., Brunton D. H. & Hauber M. E. (2010) Reliable information content and ontogenetic shift in begging calls of grey warbler nestlings. Ethology 116, 357-65.

Andersson M. (1994) Sexual Selection. Princeton University Press, New Jersey.

Angeloni L., Schlaepfer M.A., Lawler J.J. & Crooks K.R. (2008) A reassessment of the interface between conservation and behaviour. Animal Behaviour 75, 731-7.

Anthony L.L. & Blumstein D.T. (2000) Integrating behaviour into wildlife conservation: the multiple ways that behaviour can reduce Ne. Biological Conservation 95, 303-15.

Baird R.F. (1986) Historical records of the glossy black cockatoo Calyptorhynchus lathami and red-tailed black cockatoo C. magnificus in south-eastern Australia. South Australian Ornithologist 30, 38-45.

Ballentine B., Hyman J. & Nowicki S. (2004) Vocal performance influences female response to male bird song: an experimental test. Behavioral Ecology 15, 163-8.

Barber J.R., Crooks K.R. & Fristrup K.M. (2010) The costs of chronic noise exposure for terrestrial organisms. Trends in Ecology & Evolution 25, 180-9.

Bates D., Maechler M., Bolker B. & Walker S. (2017) Package 'lme4': Linear Mixed-Effects Models using 'Eigen' and S4.

Berg K. S., Beissinger S. R. & Bradbury J. W. (2013) Factors shaping the ontogeny of vocal signals in a wild parrot. Journal of Experimental Biology 216, 338-45.

116

Berger-Tal O., Polak T., Oron A., Lubin Y., Kotler B.P. & Saltz D. (2011) Integrating animal behavior and conservation biology: A conceptual framework. Behavioral Ecology 22, 236-9.

Berger-Tal O. & Saltz D. (2016) Conservation behavior: applying behavioral ecology to wildlife conservation and management. Cambridge : Cambridge University Press.

Berris K., Barth M., Mooney T., Paton D.C., Kinloch M., Copley P., Maguire A., Crowley G. & Garnett S. (2018) From the brink of extinction: successful recovery of the glossy black-cockatoo on Kangaroo Island. In: Recovering Australian Threatened Species: A Book of Hope (eds. Garnett S, Latch P, Lindenmayer D. B. & Woinarski J. C. Z.). CSIRO Publishing, Victoria, Australia.

Blumstein D. T. & Chi Y. Y. (2012) Scared and less noisy: glucocorticoids are associated with alarm call entropy. Biology Letters 8, 189-92.

Blumstein D.T. & Fernández-Juricic E. (2010) A primer of conservation behavior. Sinauer Associates Inc., MA, United States of America.

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. & Kirschel A. N. G. (2011) Acoustic monitoring in terrestrial environments using microphone arrays: Applications, technological considerations and prospectus. Journal of Applied. Ecology 48, 758-67.

Blumstein D. T. & Récapet C. (2009) The sound of arousal: The addition of novel nonlinearities increases responsiveness in marmot alarm calls. Ethology 115, 1074- 81.

Bond A. B. & Diamond J. (2005) Geographic and ontogenetic variation in the contact calls of the kea (Nestor notabilis). Behaviour 142, 1-20.

Borker A.L., Buxton R.T., Jones I.L., Major H.L., Williams J.C., Tershy B.R. & Croll D.A. (2020) Do soundscape indices predict landscape-scale restoration outcomes? A comparative study of restored seabird island soundscapes. Restoration Ecology 28, 252-60.

117

Borker A.L., McKown M.W., Ackerman J.T., Eagles-Smith C.A., Tershy B.R. & Croll D.A. (2014) Vocal Activity as a Low Cost and Scalable Index of Seabird Colony Size. Conservation Biology 28, 1100-8.

Boucaud I. C. A., Aguirre Smith M. L. N., Valère P. A. & Vignal C. (2016) Incubating females signal their needs during intrapair vocal communication at the nest: a feeding experiment in great tits. Animal Behaviour 122, 77-86.

Bowling D.L., Garcia M., Dunn J.C., Ruprecht R., Stewart A., Frommolt K.H. & Fitch W.T. (2017) Body size and vocalization in primates and carnivores. Scientific Reports 7.

Bradbury J. W. (2003) Vocal Communication in Wild Parrots. In: Animal Social Complexity (eds de Waal F. B. M. and Tyack P. L.) pp. 293 - 316. Harvard University Press, United States of America.

Bradbury J. W. & Balsby T. J. S. (2016) The functions of vocal learning in parrots. Behavioral Ecology and Sociobiology 70, 293-312.

Bradbury J.W. & Vehrencamp S.L. (2011) Principles of Animal Communication: Second Edition. Sinauer Associates Inc., Sunderland, MA U.S.A.

Brandes T.S. (2008) Automated sound recording and analysis techniques for bird surveys and conservation. Bird Conservation International 18, S163-S73.

Brauer C., L. , Donovan T., M. , Mickey R., M. , Katz J. & Mitchell B., R. (2016) A Comparison of Acoustic Monitoring Methods for Common Anurans of the Northeastern United States. Wildlife Society Bulletin (2011-) 40, 140-9.

Brodie S., Yasumiba K., Towsey M., Roe P. & Schwarzkopf L. (2020) Acoustic monitoring reveals year-round calling by invasive toads in tropical Australia. Bioacoustics, 1-17.

Browning E., Gibb R., Glover-Kapfer P. & Jones K.E. (2017) Passive acoustic monitoring in ecology and conservation. In: WWF Conservation Technology Series 1(2). WWF-UK, Woking, United Kingdom.

Buchholz R. & Clemmons J.R. (1997) Behavioral approaches to conservation in the wild. Cambridge University Press, Cambridge; New York.

118

Burnard T. & Pritchard R. (2013) National Recovery Plan for the South-eastern Red-tailed Black-cockatoo, Calyptorhynchus banksii graptogyne: Final draft August 2013. South-eastern Red-tailed Recovery Team.

Buxton R.T., Brown E., Sharman L., Gabriele C.M. & McKenna M.F. (2016) Using bioacoustics to examine shifts in songbird phenology. Ecology and Evolution. 6: 4697– 4710.

Cameron M. (2009) The influence of climate on Glossy Black-cockatoo reproduction. Pac. Conservation Biology 15, 65-71.

Campos-Cerqueira M. & Aide T.M. (2016) Improving distribution data of threatened species by combining acoustic monitoring and occupancy modelling. Methods in Ecology and Evolution 7, 1340-8.

Cardoso G. C., Mamede A. T., Atwell J. W., Mota P. G., Ketterson E. D. & Price T. D. (2008) Song Frequency Does Not Reflect Differences in Body Size among Males in Two Oscine Species. Ethology 114, 1084-93.

Carlson N.V., Pargeter H.M. & Templeton C.N. (2017) Sparrowhawk movement, calling, and presence of dead conspecifics differentially impact blue tit (Cyanistes caeruleus) vocal and behavioral mobbing responses. Behavioral Ecology and Sociobiology 71, 133.

Caro S.M., Griffin A.S., Hinde C.A. & West S.A. (2016) Unpredictable environments lead to the evolution of parental neglect in birds. Nature Communications 7, 10985.

Caro T. (1998) Behavioral Ecology and Conservation Biology. Oxford University Press, New York.

Caro T. (1999) The behaviour–conservation interface. Trends in Ecology & Evolution 14, 366-9.

Caro T. (2016) Behavior and conservation, conservation and behavior. Current Opinion in Behavioral Sciences 12, 97-102.

Caro T. & Sherman P.W. (2011) Endangered species and a threatened discipline: behavioural ecology. Trends in Ecology & Evolution 26, 111-8.

119

Caro T. & Sherman P.W. (2013) Eighteen reasons animal behaviourists avoid involvement in conservation. Animal Behaviour 85, 305-12.

Catchpole C.K. & Slater P.J.B. (2008) Bird Song: Biological themes and variations 2nd ed. Cambridge University Press, New York.

Celis-Murillo A., Deppe J.L. & Ward M.P. (2012) Effectiveness and utility of acoustic recordings for surveying tropical birds. Journal of Field Ornithology 83, 166-79.

Chambert T., Waddle J.H., Miller D.A.W., Walls S.C. & Nichols J.D. (2018) A new framework for analysing automated acoustic species detection data: Occupancy estimation and optimization of recordings post-processing. Methods in Ecology and Evolution 9, 560-70.

Chapman T.F. & Paton D.C. (2005) The glossy black-cockatoo (Calyptorhynchus lathami halmaturinus) spends little time and energy foraging on Kangaroo Island, South Australia. Australian Journal of Zoology 53, 177-83.

Charlton B.D., Ellis W.A.H., Brumm J., Nilsson K. & Fitch W.T. (2012) Female koalas prefer bellows in which lower formants indicate larger males. Animal Behaviour 84, 1565-71.

Charlton B.D. & Reby D. (2016) The evolution of acoustic size exaggeration in terrestrial mammals. Nature Communications 7.

Charlton B.D., Reby D. & McComb K. (2007) Female red deer prefer the roars of larger males. Biology Letters 3, 382-5.

Charlton B. D., Zhihe Z. & Snyder R. J. (2009) The information content of giant panda, Ailuropoda melanoleuca, bleats: acoustic cues to sex, age and size. Animal Behaviour 78, 893-8.

Cleland J. & Sims E. (1968) Food of the glossy black-cockatoo. South Australian Ornithologist 25, 47-52.

CMS (2017) Conservation Implications of Animal Culture and Social Complexity. 12th Meeting of the Conference of the Parties. UN Environment, Convention on the Conservation of Migratory Species of Wild Animals, Manila, Philippines.

120

Collins K.T., McGreevy P.D., Wheatley K.E. & Harcourt R.G. (2011) The influence of behavioural context on Weddell seal (Leptonychotes weddellii) airborne mother-pup vocalisation. Behavioural Processes 87, 286-90.

Cornell Lab of Ornithology (2019), Macaulay Library. Available at: https://www.macaulaylibrary.org/. Last accessed 25/04/2019.

Courtney J. (1996) The juvenile food-begging calls, food-swallowing vocalisations and begging postures in Australian cockatoos. Australian Bird Watcher 16, 236-49.

Cragg J.L., Burger A.E. & Piatt J.F. (2015) Testing the effectiveness of automated acoustic sensors for monitoring vocal activity of marbled murrelets Brachyramphus marmoratus. Marine Ornithology 43, 151-60.

Crowley G.M. & Garnett S.T. (2001) Food value and tree selection by glossy black- cockatoos Calyptorhynchus lathami. Austral Ecology 26, 116-26.

Crump P.S. & Houlahan J. (2017) Designing better frog call recognition models. Ecology and Evolution.

Cunningham S. & Magrath R.D. (2017) Functionally referential alarm calls in noisy miners communicate about predator behaviour. Animal Behaviour 129, 171-9.

Cussen V. A. (2017) Psittacine cognition: Individual differences and sources of variation. Behavioural Processes 134, 103-9.

Darras K., Batáry P., Furnas B. J., Grass I., Mulyani Y. A. & Tscharntke T. (2019) Autonomous sound recording outperforms human observation for sampling birds: a systematic map and user guide. Ecological Applications 29, e01954.

Darras K., Pütz P., Fahrurrozi, Rembold K. & Tscharntke T. (2016) Measuring sound detection spaces for acoustic animal sampling and monitoring. Biological Conservation 201, 29-37.

Dawson D.K. & Efford M.G. (2009) Bird population density estimated from acoustic signals. Journal of Applied Ecology 46, 1201-9.

Dema T., Towsey M., Sherub S., Sonam J., Kinley K., Truskinger A., Brereton M. & Roe P. (2018) Acoustic detection and acoustic habitat characterisation of the critically

121

endangered white-bellied heron (Ardea insignis) in Bhutan. Freshwater Biology, 1 - 12.

Digby A., Towsey M., Bell B.D. & Teal P.D. (2013) A practical comparison of manual and autonomous methods for acoustic monitoring. Methods in Ecology and Evolution 4, 675-83.

Duan S., Zhang J., Roe P., Wimmer J., Dong X., Truskinger A. & Towsey M. (2013) Time probabilistic automaton: a bridge between Raven and Song Scope for automatic species recognition. In: Twenty-Fifth Innovation Applications of Artificial Intelligence (eds. Muñoz-Avila H & Stracuzzi DJ), pp. 1519-24, Bellevue, Washington, USA.

Efford M.G. (2011) Estimation of population density by spatially explicit capture-recapture analysis of data from area searches. Ecology 92, 2202-7.

Efford M.G. & Fewster R.M. (2013) Estimating population size by spatially explicit capture- recapture. Oikos 122, 918-28.

Ehnes M. & Foote J.R. (2015) Comparison of autonomous and manual recording methods for discrimination of individually distinctive Ovenbird songs. Bioacoustics 24, 111-21.

Evans C.S. & Evans L. (2007) Representational signalling in birds. Biology Letters 3, 8-11.

Ewart K.M., Lo N., Ogden R., Joseph L., Ho S.Y.W., Frankham G.J., Eldridge M.D.B., Schodde R. & Johnson R.N. (2020) Phylogeography of the iconic Australian red- tailed black-cockatoo (Calyptorhynchus banksii) and implications for its conservation. Heredity. DOI: https://doi.org/10.1038/s41437-020-0315-y

Ey E., Pfefferle D. & Fischer J. (2007) Do age- and sex-related variations reliably reflect body size in non-human primate vocalizations? A review. Primates 48, 253-67.

Farrow L.F., Doohan S.J. & McDonald P.G. (2017) Alarm calls of a cooperative bird are referential and elicit context-specific antipredator behavior. Behavioral Ecology 28, 724-31.

Favaro L., Gamba M., Gili C. & Pessani D. (2017) Acoustic correlates of body size and individual identity in banded penguins. PLoS ONE 12. DOI: 10.1371/journal.pone.0170001

122

Festa-Bianchet M. & Apollonio M. (2003) Animal behavior and wildlife conservation. Island Press, Washington, DC.

Fischer J., Noser R. & Hammerschmidt K. (2013) Bioacoustic field research: a primer to acoustic analyses and playback experiments with primates. American Journal of Primatology. 75, 643-63.

Fischer J., Wadewitz P. & Hammerschmidt K. (2016) Structural variability and communicative complexity in acoustic communication. Animal Behaviour 134, 229- 37.

Fitch W.T. & Hauser M. (2003) Unpacking “Honesty”: Vertebrate Vocal Production and the Evolution of Acoustic Signals. In: Acoustic communication (eds. Simmons A.M., Popper A.N. & Fay R.R.), pp. 65 - 137. Springer, United States of America.

Fitch W. T., Neubauer J. & Herzel H. (2002) Calls out of chaos: The adaptive significance of nonlinear phenomena in mammalian vocal production. Animal Behaviour 63, 407-18.

Fletcher N.H. (2004) A simple frequency-scaling rule for animal communication. Journal of the Acoustical Society of America 115, 2334-8.

Freeberg T. M., Dunbar R. I. M. & Ord T. J. (2012) Social complexity as a proximate and ultimate factor in communicative complexity. Philosophical Transactions of the Royal Society B: Biological Sciences 367, 1785-801.

Freeberg T.M. & Krams I. (2015) Does social complexity link vocal complexity and cooperation? Journal of Ornithology 156, 125-32.

Frey R., Volodin I. & Volodina E. (2007) A nose that roars: anatomical specializations and behavioural features of rutting male saiga. Journal of Anatomy 211, 717-36.

Frommolt K.H. & Tauchert K.H. (2014) Applying bioacoustic methods for long-term monitoring of a nocturnal wetland bird. Ecological Informatics 21, 4-12.

Furnas B.J. & Callas R.L. (2015) Using Automated Recorders and Occupancy Models to Monitor Common Forest Birds Across a Large Geographic Region. The Journal of Wildlife Management 79, 325-37.

123

Galeotti P., Saino N., Sacchi R. & MØller A. P. (1997) Song correlates with social context, testosterone and body condition in male barn swallows. Animal Behaviour 53, 687- 700.

Garcia M., Herbst C. T., Bowling D. L., Dunn J. C. & Fitch W. T. (2017) Acoustic allometry revisited: morphological determinants of fundamental frequency in primate vocal production. Scientific Reports 7.

Garland E.C., Goldizen A.W., Lilley M.S., Rekdahl M.L., Garrigue C., Constantine R., Hauser N.D., Poole M.M., Robbins J. & Noad M.J. (2015) Population structure of humpback whales in the western and central South Pacific Ocean as determined by vocal exchange among populations. Conservation Biology 29, 1198-207.

Garnett S.T., Britton P. & Crowley G.M. (2000) A northward extension of range of the glossy black-cockatoo Calyptorhynchus lathami. Sunbird 30.

Garnett S.T., Pedler L.P. & Crowley G.M. (1999) The breeding biology of the Glossy Black- Cockatoo Calyptorhynchus lathami on Kangaroo island, South Australia. Emu 99, 262-79.

Gill S.A. & Bierema A.M.-K. (2013) On the Meaning of Alarm Calls: A Review of Functional Reference in Avian Alarm Calling. Ethology 119, 449-61.

Gladbach A., Büßer C., Mundry R. & Quillfeldt P. (2009) Acoustic parameters of begging calls indicate chick body condition in Wilson's storm-petrels Oceanites oceanicus. Journal of Ethology 27, 267-74.

Godfray H.C.J. (1991) Signalling of need by offspring to their parents. Nature 352, 328-30.

Gosling M. & Sutherland W.J. (2000) Behaviour and conservation Cambridge University Press, New York.

Große Ruse M., Hasselquist D., Hansson B., Tarka M. & Sandsten M. (2016) Automated analysis of song structure in complex birdsongs. Animal Behaviour 112, 39-51.

Haff T.M. & Magrath R.D. (2011) Calling at a cost: Elevated nestling calling attracts predators to active nests. Biology Letters 7, 493-5.

Hafner S.D. & Katz J. (2017) Package 'monitoR': Acoustic Template Detection in R.

124

Hall M. L., Kingma S. A. & Peters A. (2013) Male Songbird Indicates Body Size with Low- Pitched Advertising Songs. PLOS ONE 8, e56717.

Hammerschmidt K. & Ansorge V. (1989) Birth of a Barbary Macaque (Macaca sylvanus): Acoustic and Behavioural Features. Folia Primatologica 52, 78-87.

Harcourt A.H. (1999) The behavior–conservation interface. Trends in Ecology & Evolution 14, 490.

Hart P.J., Sebastián-González E., Tanimoto A., Thompson A., Speetjens T., Hopkins M. & Atencio-Picado M. (2018) Birdsong characteristics are related to fragment size in a neotropical forest. Animal Behaviour 137, 45-52.

Heinicke S., Kalan A.K., Wagner O.J.J., Mundry R., Lukashevich H. & Kühl H.S. (2015) Assessing the performance of a semi-automated acoustic monitoring system for primates. Methods in Ecology and Evolution 6, 753-63.

Heinsohn R., Zdenek C. N., Cunningham R. B., Endler J. A. & Langmore N. E. (2017) Tool- assisted rhythmic drumming in palm cockatoos shares key elements of human instrumental music. Science Advances 3, e1602399.

Hiebert S.M., Stoddard P.K. & Arcese P. (1989) Repertoire size, territory acquisition and reproductive success in the song sparrow. Animal Behaviour 37, 266-73.

Higgins P. J. (1999) Parrots to Dollarbird. Handbook of Australian, New Zealand and Antarctic Birds, Oxford University Press, Melbourne.

Hill R. (2019) Population monitoring of the South-eastern Red-tailed Black-cockatoo: Flock count report to Birdlife Australia. Department of Environment Land, Water and Planning, Victoria.

Hinde C.A. & Godfray H.C.J. (2011) Quality, need, or hunger; Begging the question. Behavioral Ecology 22, 1147-8.

Höbel G. (2017) Social facilitation is a better predictor of frog reproductive activity than environmental factors. Biotropica 49, 372-81.

Hoeschele M. & Bowling D. L. (2016) Sex Differences in Rhythmic Preferences in the Budgerigar (Melopsittacus undulatus): A Comparative Study with Humans. Frontiers in Psychology 7.

125

Hothorn T., Bretz F. & Westfall P. (2008) Simultaneous Inference in General Parametric Models. Biometrical Journal 50, 346-63.

Ibáñez-Álamo J.D., Arco L. & Soler M. (2012) Experimental evidence for a predation cost of begging using active nests and real chicks. Journal of Ornithology 153, 801-7.

Jaramillo-Legorreta A., Cardenas-Hinojosa G., Nieto-Garcia E., Rojas-Bracho L., Ver Hoef J., Moore J., Tregenza N., Barlow J., Gerrodette T., Thomas L. & Taylor B. (2017) Passive acoustic monitoring of the decline of Mexico's critically endangered vaquita. Conservation Biology 31, 183-91.

Johnstone R. & Kirkby T. (2008) Distribution, status, social organisation, movements and conservation of Baudin's Cockatoo (Calyptorhynchus baudinii) in South-west Western Australia. Records of the Western Australian Museum 25, 107 - 18.

Johnstone R. E., Kirkby T. & Sarti K. (2013) The breeding biology of the forest red-tailed black cockatoo Calyptorhynchus banksii naso Gould in South-Western Australia. I. Characteristics of nest trees and nest hollows. Pacific Conservation Biology 19, 121- 42.

Johnstone R.A. & Godfray H.C.J. (2002) Models of begging as a signal of need. In: The evolution of begging: competition, cooperation and communication (eds. Wright, J. & Leonard, M.L.), pp. 1-20. Kluwer Academic Publishers, The Netherlands.

Joseph L. (1982) The glossy black-cockatoo on Kangaroo Island. Emu 82, 46-9.

Joshi K.A., Mulder R.A. & Rowe K.M.C. (2017) Comparing manual and automated species recognition in the detection of four common south-east Australian forest birds from digital field recordings. Emu 117, 233-46.

Katz J., Hafner S.D. & Donovan T. (2016a) Assessment of Error Rates in Acoustic Monitoring with the R package monitoR. Bioacoustics 25, 177-96.

Katz J., Hafner S.D. & Donovan T. (2016b) Tools for automated acoustic monitoring within the R package monitoR. Bioacoustics 25, 197-210.

Keighley M.V., Langmore N.E., Zdenek C.N. & Heinsohn R. (2017) Geographic variation in the vocalizations of Australian palm cockatoos (Probosciger aterrimus). Bioacoustics 26, 91-108.

126

Kershenbaum A. & Blumstein D.T. (2017) Introduction to the special column: Communication, cooperation, and cognition in predators. Current Zoology 63, 295-9.

Kilner R.M. (2001) A growth cost of begging in captive canary chicks. Proceedings of the National Academy of Sciences of the United States of America 98, 11394-8.

Kirschel A.N.G., Earl D.A., Yao Y., Escobar I.A., Vilches E., Vallejo E.E. & Taylor C.E. (2009) Using songs to identify individual mexican antthrush Formicarius moniliger: Comparison of four classification methods. Bioacoustics 19, 1-20.

Klenova A.V. (2015) Chick begging calls reflect degree of hunger in three auk species (Charadriiformes: Alcidae). PLoS ONE 10.

Knight E.C. & Bayne E.M. (2019) Classification threshold and training data affect the quality and utility of focal species data processed with automated audio-recognition software. Bioacoustics 28, 539-54.

Knight E.C., Hannah K.C., Foley G.J., Scott C.D., Brigham R.M. & Bayne E. (2017) Recommendations for acoustic recognizer performance assessment with application to five common automated signal recognition programs. Avian Conservation and Ecology 12.

Knight J. (2001) If they could talk to the animals. Nature 414, 246.

Koch P. (2003) Factors influencing food availability for the endangered south-eastern Red- tailed Black Cockatoo Calyptorhynchus banksii graptogyne in remnant stringybark woodland, and implications for management. In: Department of Environmental Biology. University of Adelaide.

Köhler J., Jansen M., Rodríguez A., Kok P.J.R., Toledo L.F., Emmrich M., Glaw F., Haddad C.F.B., Rödel M.O. & Vences M. (2017) The use of bioacoustics in anuran taxonomy: Theory, terminology, methods and recommendations for best practice. Zootaxa 4251, 1-124.

Kouba M., Bartoš L., Černý M. & Šťastný K. (2015) The reliability of using counts of vocal begging young to estimate the number of surviving juvenile Tengmalm's Owls (Aegolius funereus) at the end of the post-fledging period. Ecological Informatics 27, 39-43.

127

Krams I., Krama T., Freeberg T.M., Kullberg C. & Lucas J.R. (2012) Linking social complexity and vocal complexity: a parid perspective. Philosophical Transactions: Biological Sciences 367, 1879-91.

Kühl H.S., Boesch C., Kulik L., Haas F., Arandjelovic M., Dieguez P., Bocksberger G., McElreath M.B., Agbor A., Angedakin S., Ayimisin E.A., Bailey E., Barubiyo D., Bessone M., Brazzola G., Chancellor R., Cohen H., Coupland C., Danquah E., Deschner T., Diotoh O., Dowd D., Dunn A., Egbe V.E., Eshuis H., Fernandez R., Ginath Y., Goedmakers A., Granjon A.-C., Head J., Hedwig D., Hermans V., Imong I., Jeffery K.J., Jones S., Junker J., Kadam P., Kambere M., Kambi M., Kienast I., Kujirakwinja D., Langergraber K., Lapuente J., Larson B., Lee K., Leinert V., Llana M., Maretti G., Marrocoli S., Mbi T.J., Meier A.C., Morgan B., Morgan D., Mulindahabi F., Murai M., Neil E., Niyigaba P., Ormsby L.J., Pacheco L., Piel A., Preece J., Regnaut S., Rundus A., Sanz C., van Schijndel J., Sommer V., Stewart F., Tagg N., Vendras E., Vergnes V., Welsh A., Wessling E.G., Willie J., Wittig R.M., Yurkiw K., Zuberbuehler K. & Kalan A.K. (2019) Human impact erodes chimpanzee behavioral diversity. Science, eaau4532.

Kuhn M. (2008) Building predictive models in R using the caret package. Journal of Statistical Software 28, 1-26.

Kutt A.S., Vanderduys E.P., Perry J.J. & Perkins G.C. (2012) Do miners (Manorina spp.) affect bird assemblages in continuous savanna woodlands in north-eastern Australia? Austral Ecology 37, 779-88.

Lai J., Lortie C.J., Muenchen R.A., Yang J. & Ma K. (2019) Evaluating the popularity of R in ecology. Ecosphere 10, e02567.

Laiolo P. (2010) The emerging significance of bioacoustics in animal species conservation. Biological Conservation 143, 1635-45.

Laiolo P. & Jovani R. (2007) The emergence of animal culture conservation. Trends in Ecology and Evolution 22, 5.

Laiolo P. & Tella J.L. (2007) Erosion of animal cultures in fragmented landscapes. Frontiers in Ecology and the Environment 5, 68-72.

128

Lambert K.T.A. & McDonald P.G. (2014) A low-cost, yet simple and highly repeatable system for acoustically surveying cryptic species. Austral Ecology 39, 779-85.

Leighton G.M. (2017) Cooperative breeding influences the number and type of vocalizations in avian lineages. Proceedings of the Royal Society B: Biological Sciences 284.

Lemen C., Freeman P.W., White J.A. & Andersen B.R. (2015) The Problem of Low Agreement among Automated Identification Programs for Acoustical Surveys of Bats. Western North American Naturalist 75, 218-25.

Lindenmayer D.B., Piggott M.P. & Wintle B.A. (2013) Counting the books while the library burns: why conservation monitoring programs need a plan for action. Frontiers in Ecology and the Environment 11, 549-55.

Linke S. & Deretic J. A. (2020) Ecoacoustics can detect ecosystem responses to environmental water allocations. Freshwater Biology 65, 133-41.

Linklater W.L. (2004) Wanted for Conservation Research: Behavioral Ecologists with a Broader Perspective. BioScience 54, 352-60.

Liu J.P., Ma L.K., Zhang Z.Q., Gu D.H., Wang J.J., Li J.J., Gao L.J. & Hou J.H. (2017) Maximum frequency of songs reflects body size among male dusky warblers Phylloscopus fuscatus (Passeriformes: Phylloscopidae). European Zoological Journal 84, 186-92.

Lowe K., Castley J.G. & Hero J.-M. (2016) Calling phenology and detectability of a threatened amphibian (Litoria olongburensis) in ephemeral wetlands varies along a latitudinal cline: Implications for management. Austral Ecology 41, 938-51.

Macedonia J.M. & Evans C.S. (1993) Essay on Contemporary Issues in Ethology: Variation among Mammalian Alarm Call Systems and the Problem of Meaning in Animal Signals. Ethology 93, 177-97.

Manser M.B. (2001) The Acoustic Structure of Suricates' Alarm Calls Varies with Predator Type and the Level of Response Urgency. Proceedings: Biological Sciences 268, 2315-24.

Manser M.B., Madden J.R., Kunc H.P., English S. & Clutton-Brock T. (2008) Signals of need in a cooperatively breeding mammal with mobile offspring. Animal Behaviour 76, 1805-13.

129

Marler P. & Slabbekoorn H. (2004) Nature's Music: The Science of Birdsong. Elsevier Academic Press, USA.

Marques T.A., Thomas L., Martin S.W., Mellinger D.K., Ward J.A., Moretti D.J., Harris D. & Tyack P.L. (2013) Estimating animal population density using passive acoustics. Biological Reviews 88, 287-309.

Martin J. P., Doucet S. M., Knox R. C. & Mennill D. J. (2011) Body size correlates negatively with the frequency of distress calls and songs of Neotropical birds. Journal of Field Ornithology 82, 259-68.

McComb K. & Semple S. (2005) Coevolution of vocal communication and sociality in primates. Biology Letters 1, 381-5.

McDonald E.M., Morano J.L., DeAngelis A.I. & Rice A.N. (2017) Building time-budgets from bioacoustic signals to measure population-level changes in behavior: a case study with sperm whales in the Gulf of Mexico. Ecological Indicators 72, 360-4.

McDonald P.G., Wilson D.R. & Evans C.S. (2009) Nestling begging increases predation risk, regardless of spectral characteristics or avian mobbing. Behavioral Ecology 20, 821-9.

McDonald P.G. & Wright J. (2008) Provisioning vocalizations in cooperative bell miners (Manorina melanophrys): more than a simple stimulus for nestling begging? The Auk 125, 670-8.

McGregor P.K., Peake T.M. & Gilbert G. (2000) Communication behaviour and conservation. In: Behaviour and Conservation (eds. Gosling L.M. & Sutherland W.J.). Cambridge University Press, United Kingdom.

Menkhorst P., Rogers D., Clarke R., Davies J., Marsack P. & Franklin K. (2017) The Australian Bird Guide. CSIRO Publishing, Victoria, Australia.

Middleton H. A., Green D. J. & Krebs E. A. (2007) Fledgling Begging and Parental Responsiveness in American Dippers (Cinclus mexicanus). Behaviour 144, 485-501.

Miyazaki M. & Waas J.R. (2003) Correlations between body size, defensive behaviour and reproductive success in male Little Blue Penguins Eudyptula minor: implications for female choice. Ibis 145, 98-105.

Mock D.W. (2016) Some begging is actually bragging. Nature 532, 180-1.

130

Mundry R. & Sommer C. (2007) Discriminant function analysis with nonindependent data: consequences and an alternative. Animal Behaviour 74, 965-76.

Nowicki S. & Searcy W.A. (2005) Song and Mate Choice in Birds: How the Development of Behavior Helps Us Understand Function. The Auk 122, 1-14.

Odom K.J., Hall M.L., Riebel K., Omland K.E. & Langmore N.E. (2014) Female song is widespread and ancestral in songbirds. Nature Communications 5, 3379.

Owens J.L., Stec C.L. & O’Hatnick A. (2012) The effects of extended exposure to traffic noise on parid social and risk-taking behavior. Behavioural Processes 91, 61-9.

Palacios V., Lopez-Bao J.V., Llaneza L., Fernandez C. & Font E. (2016) Decoding Group Vocalizations: The Acoustic Energy Distribution of Chorus Howls Is Useful to Determine Wolf Reproduction. PLoS ONE 11, 12.

Patel R., Mulder R. A. & Cardoso G. C. (2010) What Makes Vocalisation Frequency an Unreliable Signal of Body Size in Birds? A Study on Black Swans. Ethology 116, 554-63.

Payne K.B. (2003) Sources of social complexity in the three elephant species. In: Animal Social Complexity (eds. de Waal F.B.M. & Tyack P.L.), pp. 57-85. Harvard University Press, United States of America.

Payne K.B., Thompson M. & Kramer L. (2003) Elephant calling patterns as indicators of group size and composition: The basis for an acoustic monitoring system. African Journal of Ecology 41, 99-107.

Penteriani V., Delgado M.D.M., Stigliano R., Campioni L. & Sánchez M. (2014) Owl dusk chorus is related to the quality of individuals and nest-sites. Ibis 156, 892-5.

Pepper J.W. (1997) A survey of the South Australian glossy black-cockatoo (Calyptorhynchus lathami halmaturinus) and its habitat. Wildlife Research 24, 209- 23.

Pepper J. W. (1996) The behavioural ecology of the glossy black-cockatoo Calyptorhynchus lathami halmaturinus. PhD Thesis, The University of Michigan.

131

Pepper J.W., Male T. & Roberts G.E. (1993) Report on a follow-up study of the impact of the November 1991 fire on the glossy black cockatoo in Flinders Chase National Park. National Parks and Wildlife Service of South Australia.

Pérez-Granados C., Bota G., Giralt D., Barrero A., Gómez-Catasús J., Bustillo-De La Rosa D. & Traba J. (2019) Vocal activity rate index: a useful method to infer terrestrial bird abundance with acoustic monitoring. Ibis 161, 901-7.

Platzen D. & Magrath R.D. (2004) Parental alarm calls suppress nestling vocalization. Proceedings of the Royal Society B: Biological Sciences 271, 1271-6.

Plenderleith T.L., Stratford D., Lollback G.W., Chapple D.G., Reina R.D. & Hero J.-M. (2017) Calling phenology of a diverse amphibian assemblage in response to meteorological conditions. International Journal of Biometeorology.

Podos J., Moseley D.L., Goodwin S.E., McClure J., Taft B.N., Strauss A.V.H., Rega-Brodsky C. & Lahti D.C. (2016) A fine-scale, broadly applicable index of vocal performance: Frequency excursion. Animal Behaviour 116, 203-12.

Poole J. (2011) Behavioral Contexts of Elephant Acoustic Communication. In: The Amboseli Elephants: A long-term perspective on a long-lived mammal (eds. Moss C., Croze H. & Lee P.C.), pp. 125 - 61. Univeristy of Chicago Press, Chicago and London.

Potvin D. A. (2013) Larger body size on islands affects silvereye Zosterops lateralis song and call frequency. Journal of Avian Biology 44, 221-5.

Priyadarshani N., Marsland S. & Castro I. (2018) Automated birdsong recognition in complex acoustic environments: a review. Journal of Avian Biology 49, jav-01447.

Quillfeldt P., Poisbleau M., Mundry R. & Masello J. F. (2010) Are acoustical parameters of begging call elements of thin-billed prions related to chick condition? Acta Ethologica 13, 1-9.

R Core Team. (2019) R: A language and environment for statistical computing. R Foundation for Statistical Computing., Vienna, Austria.

Rapaport L.G. (2011) Progressive parenting behavior in wild golden lion tamarins. Behavioral Ecology 22, 745-54.

132

Reby D. & Charlton B. D. (2012) Attention grabbing in red deer sexual calls. Animal Cognition 15, 265-70.

Rector M.E., Walsh C.J., Kouwenberg A.L., Fitzsimmons M.G. & Storey A.E. (2014) Signals of need and quality: Atlantic puffin chicks can beg and boast. Behavioral Ecology 25, 496-503.

Redondo T. & Castro F. (1992) Signalling of Nutritional Need by Magpie Nestlings. Ethology 92, 193-204.

Reers H. & Jacot A. (2011) The effect of hunger on the acoustic individuality in begging calls of a colonially breeding weaver bird. BMC Ecology 11.

Ritschard M. & Brumm H. (2012) Zebra finch song reflects current food availability. Evolutionary Ecology 26, 801-12.

Rocha L.H.S., Ferreira L.S., Paula B.C., Rodrigues F.H.G. & Sousa-Lima R.S. (2015) An evaluation of manual and automated methods for detecting sounds of maned wolves (Chrysocyon brachyurus Illiger 1815). Bioacoustics 24, 185-98.

Rodríguez R. L., Araya-Salas M., Gray D. A., Reichert M. S., Symes L. B., Wilkins M. R., Safran R. J. & Höbel G. (2014) How acoustic signals scale with individual body size: common trends across diverse taxa. Behavioral Ecology 26, 168-77.

Rodríguez-Gironés M.A., Zúñiga J.M. & Redondo T. (2001) Effects of begging on growth rates of nestling chicks. Behavioral Ecology 12, 269-74.

Rogers T.L. (2007) Age-related differences in the acoustic characteristics of male leopard seals, Hydrurga leptonyx. The Journal of the Acoustical Society of America 122, 596- 605.

Rogers T. L., Ciaglia M. B., Klinck H. & Southwell C. (2013) Density Can Be Misleading for Low-Density Species: Benefits of Passive Acoustic Monitoring. PLoS ONE 8, e52542.

Roush R.S. & Snowdon C.T. (2001) Food transfer and development of feeding behavior and food-associated vocalizations in cotton-top tamarins. Ethology 107, 415-29.

133

Ruff Z.J., Lesmeister D.B., Duchac L.S., Padmaraju B.K. & Sullivan C.M. (2019) Automated identification of avian vocalizations with deep convolutional neural networks. Remote Sensing in Ecology and Conservation n/a.

Russell V.-J., Hill R., Burnard T., Perryman B., Copley P., Gilkes K., Maron M., Baker-Gabb D., Pritchard R. & Koch M. (2018) Science, community and commitment underpin the road to recovery for the red-tailed black-cockatoo. In: Recovering Australian Threatened Species: A Book of Hope (eds. Garnett S., Latch P., Lindenmayer D. B. & Woinarski J. C. Z.). CSIRO Publishing, Victoria, Australia.

Russo D. & Voigt C.C. (2016) The use of automated identification of bat echolocation calls in acoustic monitoring: A cautionary note for a sound analysis. Ecological Indicators 66, 598-602.

Ryan S.J. (2006) The Role of Culture in Conservation Planning for Small or Endangered Populations. Conservation Biology 20, 1321-4.

Sacchi R., Saino N. & Galeotti P. (2002) Features of begging calls reveal general condition and need of food of barn swallow (Hirundo rustica) nestlings. Behavioral Ecology. 13, 268-73.

Salamon J., Bello J.P., Farnsworth A., Robbins M., Keen S., Klinck H. & Kelling S. (2016) Towards the automatic classification of avian flight calls for bioacoustic monitoring. PLoS ONE 11.

Saunders D. A. (1983) Vocal repertoire and individual vocal recognition in the short- billed white-tailed black cockatoo, Calyptorhynchus funereus latirostris Carnaby. Australian Wildlife Research 10, 527-36.

Saunders D. A., Dawson R. & Nicholls A. O. (2016) Breeding failure and nestling body mass as a function of age of breeding females in the endangered Carnaby's Cockatoo, Calyptorhynchus latirostris. Australian Zoologist 38, 171-82.

Schodde R., Mason I.J. & Wood J.T. (1993) Geographical differentiation in the glossy black- cockatoo Calyptorhynchus lathami (Temminck) and its history. Emu 93, 156-66.

Schroeder K.M. & McRae S.B. (2020) Automated auditory detection of a rare, secretive marsh bird with infrequent and acoustically indistinct vocalizations. Ibis n/a.

134

Sebastián-González E., Pang-Ching J., Barbosa J.M. & Hart P. (2015) Bioacoustics for species management: Two case studies with a Hawaiian forest bird. Ecology and Evolution.

Servick K. (2014) Eavesdropping on ecosystems. Science 343, 834-7.

Seyfarth R.M., Cheney D.L. & Marler P. (1980a) Monkey Responses to Three Different Alarm Calls: Evidence of Predator Classification and Semantic Communication. Science 210, 801-3.

Seyfarth R.M., Cheney D.L. & Marler P. (1980b) Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Animal Behaviour 28, 1070-94.

Shannon G., Angeloni L.M., Wittemyer G., Fristrup K.M. & Crooks K.R. (2014) Road traffic noise modifies behaviour of a keystone species. Animal Behaviour 94, 135-41.

Shannon G., McKenna M.F., Angeloni L.M., Crooks K.R., Fristrup K.M., Brown E., Warner K.A., Nelson M.D., White C., Briggs J., McFarland S. & Wittemyer G. (2016) A synthesis of two decades of research documenting the effects of noise on wildlife. Biological Reviews 91, 982-1005.

Sharpe D.L., Castellote M., Wade P.R. & Cornick L.A. (2017) Call types of Bigg’s killer whales (Orcinus orca) in western Alaska: using vocal dialects to assess population structure. Bioacoustics, 1-26.

Shonfield J. & Bayne E.M. (2017) Autonomous recording units in avian ecological research: current use and future applications. Avian Conservation and Ecology 12.

Shonfield J., Heemskerk S. & Bayne E.M. (2018) Utility of Automated Species Recognition For Acoustic Monitoring of Owls. Journal of Raptor Research 52, 42-55, 14.

Slabbekoorn H. & Ripmeester E.A.P. (2008) Birdsong and anthropogenic noise: implications and applications for conservation. Molecular Ecology 17, 72-83.

Smith C.L. (2017) Referential signalling in birds: the past, present and future. Animal Behaviour 124, 315-23.

Snaddon J., Petrokofsky G., Jepson P. & Willis K.J. (2013) Biodiversity technologies: tools as change agents. Biology Letters 9.

135

Stevenson B.C., Borchers D.L., Altwegg R., Swift R.J., Gillespie D.M. & Measey G.J. (2015) A general framework for animal density estimation from acoustic detections across a fixed microphone array. Methods in Ecology and Evolution 6, 38-48.

Stoeger A. S., Charlton B. D., Kratochvil H. & Fitch W. T. (2011) Vocal cues indicate level of arousal in infant African elephant roars. Journal of the Acoustical Society of America 130, 1700-10.

Stoeger-Horwath A. S., Stoeger S., Schwammer H. M. & Kratochvil H. (2007) Call repertoire of infant African elephants: First insights into the early vocal ontogeny. Journal of the Acoustical Society of America 121, 3922-31.

Stojanovic D., Webb M.H., Alderman R., Porfirio L.L. & Heinsohn R. (2014) Discovery of a novel predator reveals extreme but highly variable mortality for an endangered migratory bird. Diversity and Distributions 20, 1200-7.

Sueur J. & Farina A. (2015) Ecoacoustics: the Ecological Investigation and Interpretation of Environmental Sound. Biosemiotics 8, 493-502.

Sueur J. (2019) Package 'seewave': Sound Analysis and Synthesis.

Sugai L.S.M., Silva T.S.F., Ribeiro J.W., Jr & Llusia D. (2019) Terrestrial Passive Acoustic Monitoring: Review and Perspectives. BioScience 69, 15-25.

Suzuki T.N. (2016) Semantic communication in birds: evidence from field research over the past two decades. Ecological Research 31, 307-19.

Suzuki T., N. & Kutsukake N. (2017) Foraging intention affects whether willow tits call to attract members of mixed-species flocks. Royal Society Open Science 4, 170222.

Tanttu J.T., Turunen J., Selin A. & Ojanen M. (2006) Automatic feature extraction and classification of crossbill (Loxia spp.) flight calls. Bioacoustics 15, 251-69.

Taylor A.M. & Reby D. (2010) The contribution of source-filter theory to mammal vocal communication research. Journal of Zoology 280, 221-36.

Templeton C.N., Zollinger S.A. & Brumm H. (2016) Traffic noise drowns out great tit alarm calls. Current Biology 26, R1173-R4.

Thorley J.B. & Clutton-Brock T. (2017) Kalahari vulture declines, through the eyes of meerkats. Ostrich 88, 177-81.

136

Townsend S.W., Deschner T. & Zuberbühler K. (2008) Female Chimpanzees Use Copulation Calls Flexibly to Prevent Social Competition. PLoS ONE 3, e2431.

Townsend S.W. & Manser M.B. (2013) Functionally Referential Communication in Mammals: The Past, Present and the Future. Ethology 119, 1-11.

Towsey M., Parsons S. & Sueur J. (2014) Ecology and acoustics at a large scale. Ecological Informatics 21, 1-3.

Towsey M., Planitz B., Nantes A., Wimmer J. & Roe P. (2012) A toolbox for animal call recognition. Bioacoustics 21, 107-25.

Tubaro P. L. & Mahler B. (1998) Acoustic Frequencies and Body Mass in New World Doves. The Condor 100, 54-61.

Tyack P.L. (2008) Convergence of calls as animals form social bonds, active compensation for noisy communication channels, and the evolution of vocal learning in mammals. Journal of Comparative Psychology 122, 319-31.

Venables W. N. & Ripley B. D. (2002) Modern Applied Statistics with S. Fourth Edition. Springer, New York.

Venier L.A., Holmes S.B., Holborn G.W., McIlwrick K.A. & Brown G. (2012) Evaluation of an automated recording device for monitoring forest birds. Wildlife Society Bulletin 36, 30-9.

Wein A., Schwing R. & Huber L. (2019) Kea parrot mothers Nestor notabilis produce nest- specific calls with low amplitude and high entropy. Ibis DOI: 10.1111/ibi.12753

Weiß B.M., Ladich F., Spong P. & Symonds H. (2006) Vocal behavior of resident killer whale matrilines with newborn calves: The role of family signatures. Journal of the Acoustical Society of America 119, 627-35.

Whitehead H. (2010) Conserving and managing animals that learn socially and share cultures. Learning & Behavior 38, 329-36.

Whitehead H., Rendell L., Osborne R.W. & Würsig B. (2004) Culture and conservation of non-humans with reference to whales and dolphins: review and new directions. Biological Conservation 120, 427-37.

137

Whitehead H. & Weilgart L. (1991) Patterns of Visually Observable Behaviour and Vocalizations in Groups of Female Sperm Whales. Behaviour 118, 275-96.

Wickham H. (2016) ggplot2: Elegant Graphics for Data Analysis., Springer-Verlag, New York.

Wickham H., François R., Henry L. & Müller K. (2019) dplyr: A Grammar of Data Manipulation.

Wilden I., Herzel H., Peters G. & Tembrock G. (1998) Subharmonics, biphonation, and deterministic chaos in mammal vocalization. Bioacoustics 9, 171-96.

Williams E.M., O'Donnell C.F.J. & Armstrong D.P. (2018) Cost-benefit analysis of acoustic recorders as a solution to sampling challenges experienced monitoring cryptic species. Ecology and Evolution 8, 6839-48.

Wimmer J., Towsey M., Roe P. & Williamson I. (2013) Sampling environmental acoustic recordings to determine bird species richness. Ecological Applications 23, 1419-28.

Wrege P.H., Rowland E.D., Keen S. & Shiu Y. (2017) Acoustic monitoring for conservation in tropical forests: examples from forest elephants. Methods in Ecology and Evolution 8, 1292-301.

Wyman M.T., Mooring M.S., McCowan B., Penedo M.C.T., Reby D. & Hart L.A. (2012) Acoustic cues to size and quality in the vocalizations of male North American bison, Bison bison. Animal Behaviour 84, 1381-91.

Wyman M.T., Rivers P.R., Muller C., Toni P. & Manser M.B. (2017) Adult meerkats modify close call rate in the presence of pups. Current Zoology 63, 349-55.

Yurk H. (2005) Vocal culture and social stability in resident killer whales (Orcinus orca). University of British Colombia.

Zann R. & Cash E. (2008) Developmental stress impairs song complexity but not learning accuracy in non-domesticated zebra finches (Taeniopygia guttata). Behavioral Ecology and Sociobiology 62, 391-400.

Zdenek C. N., Heinsohn R. & Langmore N. E. (2015) Vocal complexity in the palm cockatoo (Probosciger aterrimus). Bioacoustics 24, 253-67.

138

Zdenek C.N., Heinsohn R. & Langmore N.E. (2018) Vocal individuality, but not stability, in wild palm cockatoos (Probosciger aterrimus). Bioacoustics 27, 27-42.

Zeppelzauer M., Hensman S. & Stoeger A.S. (2015) Towards an automated acoustic detection system for free-ranging elephants. Bioacoustics 24, 13-29.

Znidersic E., Towsey M., Roy W.K., Darling S.E., Truskinger A., Roe P. & Watson D.M. (2020) Using visualization and machine learning methods to monitor low detectability species—The least bittern as a case study. Ecological Informatics 55, 101014.

Zwart M.C., Baker A., McGowan P.J.K. & Whittingham M.J. (2014) The use of automated bioacoustic recorders to replace human wildlife surveys: An example using nightjars. PLoS ONE 9.

139

Appendices

Appendix 1 | Ethics approvals for Victoria and South Australia.

140

141

142

143

Appendix 2 | Supplementary material for Chapter 2

Table A2.1: Taxon summary of vocal behaviours represented in the audio collection of the Macaulay Library (www.macaulaylibrary.org; accessed 03/04/2019). Only species for whom vocalisations were categorised for behaviour (courtship, foraging, flying) are included.

Number Vocal behaviour Taxa Family of species Courtship Amphibian Ranidae 1 (display or copulation) Bird Cotingidae 3 Leiotrichidae 1 Pipridae 5 Scolopacidae 1 Trochilidae 6 Tyrannidae 1 Foraging Bird Cacatuidae 1 (including feeding young) Cardinalidae 1 Cinclidae 1 Corvidae 1 Furnariidae 3 Hirundinidae 1 Icteridae 2 Laniidae 1 Muscicapidae 1 Parulidae 5 Passerellidae 3 Passeridae 1 Phalacrocoracidae 1 Picidae 3 Podicipedidae 1 Polioptilidae 1 Prunellidae 1 Psittacidae 2 Regulidae 1 Strigidae 1

144

Timaliidae 1 Tityridae 2 Troglodytidae 2 Turdidae 2 Tyrannidae 4 Vangidae 1 Vireonidae 1 Flying Bird Accipitridae 3 Alaudidae 3 Alcedinidae 1 Apodidae 2 Ardeidae 1 Burhinidae 1 Cacatuidae 1 Corvidae 2 Falconidae 1 Fringillidae 4 Furnariidae 1 Gruidae 1 Hirundinidae 2 Laridae 2 Meropidae 1 Mimidae 1 Monarchidae 1 Motacillidae 2 Psittacidae 13 Recurvirostridae 1 Scolopacidae 4 Thraupidae 2 Trochilidae 1 Troglodytidae 1 Turdidae 3

145

Appendix 3 | Supplementary material for Chapter 3

a b

Fig. A3.1: Pictures of typical nesting habitat for the south-eastern red-tailed black-cockatoo, an isolated dead river red gum, Eucalyptus camaldulensis, in a livestock paddock (a), and the Kangaroo Island glossy black-cockatoo, an isolated sugar gum, Eucalyptus cladocalyx, in a modified woodland (b). In image (b), the white artificial nest hollow and the possum- exclusion iron collar can be seen.

146

Table A3.1: Quantitative acoustic measurements calculated for each call selected in recordings made with the autonomous sound recorders. Definitions adapted from the user manual for Raven Pro software.

Acoustic measurement Definition

Low frequency (Hz) The lowest frequency bound of the selection. Peak frequency (Hz) The frequency at which greatest energy occurs within the selection. Centre frequency (Hz) The frequency at which the selection is divided into two parts of equal energy. Aggregate entropy (bits) A measure of overall disorder (chaos) in a sound. A higher value corresponds to greater disorder. A pure tone has zero entropy. Aggregate entropy measures the energy distribution across a selection. Average entropy (bits) The average entropy of each time slice within a selection. Describes the typical disorder within a spectrum of the selection. Minimum entropy (bits) The minimum entropy for a spectrogram slice that occurs within the selection. Maximum entropy (bits) The maximum entropy for a spectrogram slice that occurs within the selection. Delta time (s) Duration of the selection. The difference between start time and end time of the selection. Interquartile range duration (s) The difference between the first and third quartile times. The first quartile divides the selection into two time intervals containing 25% and 75% of the energy in the selection. The third quartile divides the selection into two time intervals containing 75% and 25% of the energy in the selection. Peak amplitude (U) The greatest absolute amplitude value in the selection (i.e., greatest of maximum amplitude and minimum amplitude). Raven Pro amplitude measurements are dimensionless (U). They are calculated relative to an arbitrary reference point, to enable comparisons between selections made with the same recording equipment and settings. Peak frequency contour (PFC) Tracks the 'pitch' of the sound in the selection by measuring its slope (i.e., the change in frequency over time). This measurement average slope (Hz/ms) provides the average slope of the peak frequency across spectrogram slices in the selection. Peak frequency contour (PFC) Tracks the 'pitch' of the sound in the selection by measuring its slope (i.e., the change in frequency over time). This measurement maximum slope (Hz/ms) provides the maximum slope of the peak frequency in the selection.

147

b DT

Fig. A3.2: Example waveform (a) and spectrogram (b) of adult and nestling vocalisations of the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, with selection boxes used for acoustic measurements. Arrows demonstrate low frequency (LF), peak frequency (PF) and delta time (DT) on a nestling call. Other acoustic measurements were generated automatically for the area within each selection box in Raven Pro 1.5 software (Cornell Lab of Ornithology). Numbers represent the order in which selection boxes were drawn (other boxes in this sound recording are not shown here). Spectrogram parametres: Hann window; window size = 1024 samples; hop size = 512 samples; 50% overlap.

148

Table A3.2: Linear discriminant analysis confusion matrix and classification accuracy for each call type of (a) the south-eastern red- tailed black-cockatoo (RTBC; n = 907 selections) and (b) the Kangaroo Island glossy black-cockatoo (GBC; n = 1,632 selections). Model trained and tested on 70% and 30% of the dataset, respectively. Take-off subtype of the flight call, nestling subtype 3 and perch subtype 3 (RTBC only) were not included in the analysis due to low sample size.

Begging Display Flight Nest entry Nestling1 Nestling2 Perch1 Perch2 Perch3 Begging 19 3 8 0 1 4 2 2 0 Display 1 14 2 0 0 0 0 1 0 Flight 8 1 54 1 6 3 11 7 0 FlightT 0 0 0 0 0 0 0 0 0 Nest entry 2 0 0 11 4 3 0 3 0

Nestling1 7 0 4 1 18 4 0 3 0

Nestling2 1 1 0 2 4 6 0 2 0 RTBC Nestling3 0 0 0 0 0 0 0 0 0 Perch1 0 0 6 0 0 1 8 1 0 Perch2 2 0 1 5 0 0 2 19 0 Perch3 0 0 0 0 0 0 0 0 0 Accuracy 47.5% 73.7% 72.0% 55.0% 54.6% 28.6% 34.8% 50.0% 0 Begging Display Flight Nest entry Nestling1 Nestling2 Perch1 Perch2 Perch3 Perch4 Perch5 Perch6 Begging 61 10 2 1 0 0 1 3 2 1 2 6

Display 3 9 1 0 0 0 0 2 0 0 0 0

Flight 0 0 51 1 5 1 9 17 11 2 3 1 GBC Nest entry 1 0 2 15 0 1 1 0 1 0 4 0 Nestling1 0 0 6 5 79 9 0 1 0 0 0 0 Nestling2 0 0 3 4 4 30 0 0 1 2 1 0

149

Perch1 0 0 0 0 1 0 0 0 1 0 0 0 Perch2 1 1 0 0 0 0 0 31 0 1 0 4 Perch3 0 0 3 0 1 0 5 2 17 3 0 3 Perch4 0 0 4 0 0 0 0 2 0 3 0 0 Perch5 0 0 1 0 0 0 0 1 0 0 0 0 Perch6 6 0 1 0 0 0 0 7 0 1 0 8 Accuracy 84.7% 45.0% 68.9% 57.7% 87.8% 73.2% 0.00% 47.0% 51.5% 23.1% 0.00% 36.4%

150

Table A3.3: Linear discriminant analysis confusion matrix and classification accuracy for each call type of (a) the south-eastern red- tailed black-cockatoo and (b) the Kangaroo Island glossy black-cockatoo. Model tested using leave-one-out cross validation using the complete dataset. Take-off subtype of the flight call and nestling subtype 3 were not included in the analysis due to low sample size.

Begging Display Flight Nest entry Nestling1 Nestling2 Perch1 Perch2 Perch3 Begging 66 10 19 0 12 12 4 8 0 Display 11 38 4 1 2 1 0 2 0 Flight 27 4 189 2 19 11 35 14 0 FlightT 0 0 0 0 0 0 0 0 0 Nest entry 5 1 0 44 12 8 0 4 0

Nestling1 11 1 8 4 59 16 3 4 0

Nestling2 6 11 2 4 4 20 0 5 0 RTBC Nestling3 0 0 0 0 0 0 0 0 0 Perch1 0 0 16 2 0 0 24 9 0 Perch2 9 0 13 12 2 5 11 81 0 Perch3 0 0 0 0 0 0 0 0 0 Accuracy 48.9% 58.5% 75.4% 63.8% 53.6% 27.4% 31.2% 63.8% Begging Display Flight Nest entry Nestling1 Nestling2 Perch1 Perch2 Perch3 Perch4 Perch5 Perch6 Begging 206 34 3 7 1 4 2 8 7 2 2 17 Display 19 31 2 0 0 1 0 2 0 0 0 1

Flight 0 0 184 3 9 2 38 47 45 9 10 1

GBC Nest entry 2 0 7 54 4 8 1 2 4 1 11 1 Nestling1 0 0 15 14 264 34 2 4 0 1 8 0 Nestling2 1 0 5 8 22 86 0 2 2 2 3 3 Perch1 0 0 4 0 2 1 2 2 3 1 0 0

151

Perch2 1 2 0 0 0 1 0 120 2 3 0 20 Perch3 1 2 20 0 1 1 11 17 44 10 1 13 Perch4 0 0 6 0 0 0 0 4 3 10 0 0 Perch5 0 0 0 2 0 0 0 0 0 0 0 0 Perch6 13 0 2 1 0 0 0 13 2 5 0 18 Accuracy 84.8% 44.9% 74.2% 60.7% 87.1% 62.3% 3.6% 54.3% 39.3% 22.7% 0.00% 24.3%

152

Table A3.4: Descriptive statistics (mean ± SE) of acoustic parameters of call types of the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus, and the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne, at nests. Nestling subtype 3 not shown due to low sample size. PFC PFC Centre Aggregate Average Min Max Delta IQR Peak Low Freq Peak Freq Average Max Call Type Freq Entropy Entropy Entropy Entropy Time Duration Amplitude (Hz) (Hz) Slope Slope (Hz) (bits) (bits) (bits) (bits) (s) (s) (U) (Hz/ms) (Hz/ms) Begging 875.74 ± 3049.56 ± 3233.53 ± 4.37 ± 2.88 ± 5.94 ± 1877.14 ± 1.92 ± 167.58 ± 4.69 ± 0.06 0.32 ± 0 0.1 ± 0 n = 243 23.33 49.09 36.89 0.05 0.06 0.04 77.03 0.28 7.11 Display 876.64 ± 3445.94 ± 3345.45 ± 4.54 ± 2.88 ± 6.05 ± 0.32 ± 0.12 ± 1944.8 ± 1.83 ± 146.23 ± 5.19 ± 0.05 n = 69 70.74 67.22 45.52 0.07 0.08 0.05 0.01 0.01 205.26 0.44 10.57 Flight 577.58 ± 3900.64 ± 3843.85 ± 4.68 ± 3.27 ± 5.92 ± 0.75 ± 3489.48 ± 0.89 ± 222.3 ± 5.22 ± 0.04 0.24 ± 0 n = 248 4.42 32.84 23.94 0.04 0.04 0.03 0.01 167.79 0.13 9.24

Take-off 567.46 ± 3919.04 ± 3761.13 ± 4.05 ± 2.74 ± 5.71 ± 0.61 ± 4839.67 ± 0.28 ± 205.26 ± 4.72 ± 0.14 0.2 ± 0.03 n = 9 16.09 97.1 76.64 0.12 0.17 0.16 0.05 756.4 0.24 65.93 Nest entry 484.41 ± 3270.14 ± 3244.98 ± 5.36 ± 4.51 ± 6.17 ± 0.59 ± 4311.6 ± 0.93 ± 257.58 ±

5.76 ± 0.06 0.2 ± 0.01 cockatoo

- n = 89 8.69 70.83 33.71 0.05 0.07 0.05 0.02 447.71 0.39 12.1 Nestling 1 507.27 ± 3359.46 ± 3499.18 ± 5.54 ± 4.31 ± 6.37 ± 11903.99 ± 1.37 ± 231.56 ± 6.17 ± 0.03 0.8 ± 0.01 0.21 ± 0 n = 303 4.39 48.77 29.59 0.02 0.04 0.02 457.22 0.11 7.86 Nestling 2 534.63 ± 2825.53 ± 3002.79 ± 5.44 ± 4.24 ± 6.38 ± 0.59 ± 4610.28 ± 0.89 ± 192.78 ± 5.95 ± 0.04 0.17 ± 0

Glossyblack n = 138 17.85 65.94 32.3 0.04 0.06 0.04 0.01 287.57 0.25 9.99 Perch 1 610.76 ± 3819.84 ± 3889.82 ± 4.57 ± 3.32 ± 5.87 ± 0.67 ± 4140.2 ± 0.48 ± 201.5 ± 5.09 ± 0.07 0.2 ± 0.01 n = 56 9.44 126.74 65.96 0.06 0.08 0.06 0.02 351.27 0.36 13.55 Perch 2 1632.37 ± 3425.83 ± 3438.69 ± 4.2 ± 3.04 ± 5.55 ± 0.62 ± 1984.57 ± 0.04 ± 110.39 ± 4.59 ± 0.04 0.19 ± 0 n = 221 67.22 31.86 21.03 0.04 0.04 0.05 0.01 146.05 0.12 8.14 Perch 3 614.14 ± 3581.43 ± 3601.43 ± 4.53 ± 3.23 ± 5.91 ± 0.64 ± 0.19 ± 2619.02 ± 0.42 ± 221.51 ± 5.19 ± 0.05 n = 112 8.22 74.52 37.62 0.05 0.07 0.05 0.01 0.01 129.13 0.25 11.35

153

Perch 4 678.79 ± 3472.72 ± 3394.42 ± 4.08 ± 2.72 ± 0.64 ± 2870.75 ± 1.07 ± 218.77 ± 4.53 ± 0.12 5.83 ± 0.1 0.2 ± 0.01 n = 44 48.39 95.78 67.16 0.1 0.12 0.02 275.21 0.36 25.19 Perch 5 553.4 ± 3590.51 ± 3580.66 ± 5.12 ± 3.99 ± 5.99 ± 0.68 ± 0.21 ± 6164.86 ± 224.47 ± 5.58 ± 0.11 1 ± 0.28 n = 35 14.28 124.76 72.99 0.1 0.14 0.06 0.04 0.01 848.93 19.57 Perch 6 1155.03 ± 3236.38 ± 3218.34 ± 4.11 ± 2.75 ± 5.83 ± 0.45 ± 1395.18 ± 0.49 ± 155.5 ± 4.3 ± 0.11 0.13 ± 0 n = 74 90.43 43.96 28.6 0.08 0.08 0.07 0.01 107.85 0.31 16.29 Begging 754.78 ± 2678.41 ± 2844.93 ± 4.7 ± 3.11 ± 6.15 ± 0.39 ± 4661.61 ± 0.37 ± 151.29 ± 5.41 ± 0.06 0.13 ± 0 n = 135 37.05 99.55 79.5 0.05 0.06 0.05 0.01 500.44 0.32 6.52 Display 673.53 ± 1999.61 ± 2234.82 ± 4.38 ± 3.14 ± 0.28 ± 3825.09 ± -1.16 ± 138.68 ± 4.95 ± 0.1 5.8 ± 0.05 0.07 ± 0 n = 65 40.67 107.08 81.69 0.08 0.08 0.01 388.85 0.45 8.98 Flight 593.12 ± 3041.41 ± 3125.32 ± 4.48 ± 5.93 ± 10280.29 ± 212.19 ± 5.16 ± 0.04 3.3 ± 0.03 0.4 ± 0.01 0.13 ± 0 0.79 ± 0.2

n = 251 9.96 54.88 40.81 0.03 0.04 451.15 8.94

Take-off 572.85 ± 2806.49 ± 3075.66 ± 4.45 ± 3.36 ± 5.99 ± 0.38 ± 12468.08 ± 1.09 ± 269.55 ± 5.17 ± 0.21 0.1 ± 0.01 n = 12 28.31 272.93 253.7 0.17 0.13 0.14 0.02 2176 0.85 51.54

cockatoo Nest entry 423.84 ± 2240.7 ± 2308.73 ± 5.58 ± 4.56 ± 6.33 ± 0.13 ± 4580.93 ± 1.09 ± 180.2 ± - 5.98 ± 0.06 0.4 ± 0.02 n = 69 22.49 78.38 63.82 0.06 0.09 0.05 0.01 461.81 0.36 11.03 Nestling 1 457.65 ± 2333.02 ± 2491.2 ± 4.94 ± 3.73 ± 5.91 ± 0.41 ± 8481.12 ± -0.36 ± 125.78 ± 5.65 ± 0.05 0.14 ± 0 n = 110 17.51 51.33 37.16 0.05 0.07 0.05 0.01 739.31 0.25 8.4

Nestling 2 546.63 ± 2493.13 ± 2499.62 ± 5.02 ± 3.85 ± 5.93 ± 0.36 ± 6422.25 ± 0.18 ± 147.51 ± tailed black tailed

- 5.59 ± 0.05 0.12 ± 0 n = 73 34.27 61.16 49.35 0.05 0.07 0.04 0.01 710.04 0.33 14.42

Red Perch 1 570.8 ± 3177.41 ± 3151.12 ± 4.62 ± 3.29 ± 6.18 ± 0.47 ± 8021.9 ± 1.85 ± 194.91 ± 5 ± 0.08 0.14 ± 0 n = 77 17.51 127.06 94.08 0.07 0.06 0.06 0.01 567.92 0.32 11.13 Perch 2 600.73 ± 2328.98 ± 2386.96 ± 3.36 ± 6.32 ± 0.51 ± 0.13 ± 5485.15 ± 234.42 ± 5.25 ± 0.07 5 ± 0.06 1.3 ± 0.26 n = 127 10.87 87.89 70.37 0.06 0.05 0.01 0.01 413.91 13.59 Perch 3 590.43 ± 3399.37 ± 3109.39 ± 5.66 ± 4.68 ± 6.49 ± 0.46 ± 0.14 ± 6957.73 ± -0.01 ± 278.21 ± 6.06 ± 0.19 n = 15 38.89 320.81 162.3 0.15 0.19 0.13 0.02 0.01 677.41 1.09 26.12

154

Appendix 4 | Supplementary material for Chapter 4

South-eastern red-tailed black-cockatoo

Kangaroo Island glossy black-cockatoo

Fig. A4.1: Results of pilot recogniser testing. Each row represents one nest (n = 2 nests per species). For each nest, counts of true positive (TP) and false positive (FP) detections are shown by score cut-off and training templates. Score cut-off is the threshold of similarity between templates and raw sound data at which the recogniser returns a detection. For each subspecies, six templates were tested but only those that returned detections are shown.

155

Fig. A4.2: Frequency distribution of true positive (TP) and false positive (FP) detections by score cut-off (detection threshold) for (a) nestling calls only and (b) nestling and adult calls, for the south-eastern red-tailed black-cockatoo, Calyptorhynchus banksii graptogyne.

156

Fig. A4.3: Frequency distribution of true positive (TP) and false positive (FP) detections by score cut-off (detection threshold) for (a) nestling calls only and (b) nestling and adult calls, for the Kangaroo Island glossy black-cockatoo, Calyptorhynchus lathami halmaturinus.

157