Distribution, Abundance, Social and Genetic Structures of Indo-Pacific Bottlenose Dolphins (Tursiops Aduncus) in Perth Metropolitan Waters, Western Australia

Distribution, abundance, social and genetic structures of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in Perth metropolitan waters, Western Australia

Submitted by Delphine Brigitte Hélène Chabanne MSc (Montpellier, France)

This thesis is presented for the degree of Doctor of Philosophy of Murdoch University School of Veterinary and Life Sciences

2017

Declaration

I declare that this thesis is my own account of my research and contains as its main content work which has not previously been submitted for a degree at any tertiary education institution.

...... Delphine B. H. Chabanne

Abstract

In heterogeneous coastal and estuarine environments, dolphins are exposed to varying levels of human activities. Consequently, it is important to identify and characterise fine-scale population structuring based on ecological, social, spatial and genetic data to develop appropriate conservation and management strategies. This thesis focused on identifying subpopulations of Indo-Pacific bottlenose dolphins (Tursiops aduncus) inhabiting Perth waters, Western Australia (WA). Using spatial and social data collected over four years of boat-based photo-identification surveys, I: i) estimated abundances, survival and movement rates using a Multistate Closed Robust Design approach; and ii) examined the social structure and home range using social association and network analyses. I used microsatellite loci and mtDNA markers to investigate the genetic population structure of dolphins at metropolitan (Perth) and regional (c. 1000 km of coastline) scales. High capture probabilities, high survival and constant abundances described a subpopulation with high fidelity in an estuary. In contrast, low captures, emigration and fluctuating abundances suggested transient use and low fidelity in an open coastline region. Overall, dolphins formed four socially and geographically distinct, mixed-sex subpopulations that varied in association strength, site fidelity and residency patterns. Curiously, home range overlap and genetic relatedness did not affect the association patterns. In Perth metropolitan waters, a source-sink relationship was suggested between a subpopulation inhabiting a semi-enclosed embayment and three other subpopulations, including the estuarine subpopulation. On a broader scale, the Perth metapopulation was genetically distinct from other populations along the WA southwestern coastline, with little to no migration from and into other populations. The subpopulations present in Perth waters should each be regarded as a distinct management unit, with a particular focus on protecting the estuarine subpopulation, which is small, has limited connection with adjacent subpopulations and is more vulnerable because of the intensity and diversity of anthropogenic threats present in the estuary.

iii

Statement on the contribution of others

Supervision Professor Lars Bejder, Doctor Hugh Finn, Professor William Bill Sherwin

Project funding This research was made possible through the financial commitment of the Swan River Trust, with additional financial support from the Fremantle Harbour Ports and Murdoch University.

Stipend Australian Postgraduate Award (APA) Murdoch Strategic Top-up

Contribution to data chapters Chapter 2 – Applying the Multistate Capture-recapture Robust Design to assess metapopulation structure of a marine mammal: Delphine Chabanne collected and analysed the data and wrote the manuscript. Professor Kenneth Pollock advised on the analysis. Professor Kenneth Pollock, Doctor Hugh Finn and Professor Lars Bejder critically reviewed the manuscript.

Chapter 3 – Identifying the relevant local population for environmental impact assessments of mobile marine fauna: Delphine Chabanne collected and analysed the data and wrote the manuscript. Professor Lars Bejder advised on the analysis. Doctor Hugh Finn and Professor Lars Bejder critically reviewed the manuscript.

Chapter 4 – Genetic structure of socially and spatially discrete subpopulations of dolphins (Tursiops aduncus) in Perth, Western Australia: Delphine Chabanne collected and analysed the data and wrote the manuscript. Doctor Simon Allen contributed to the biopsy data collection. Delphine Chabanne, Doctor Celine Frère and Bethan Littleford-Colquhoun conducted the genetic laboratory work. Professor William Sherwin advised on the analysis. Professor William Sherwin, Doctor Hugh Finn and Professor Lars Bejder critically reviewed the manuscript.

Chapter 5 – Population genetic structure and effective population sizes in Indo- Pacific bottlenose dolphin (Tursiops aduncus) in southwestern Australia: Delphine Chabanne analysed the data and wrote the manuscript. Doctor Simon Allen, Anna Sellas, and Dee McElligot contributed to the biopsy data collection. Claire Daniel and Doctor Oliver Manlik conducted the genetic laboratory work. Doctor Oliver Manlik and Professor William Sherwin advised on the analysis. Doctor Oliver Manlik, Professor William Sherwin, Doctor Hugh Finn and Professor Lars Bejder critically reviewed the manuscript.

This thesis is presented as a series of four manuscripts in journal format, in addition to a general introduction and general discussion.

Ethics statement This study was carried out with approval from the Murdoch University Animal Ethics Committee (W2342/10 and R2649/14) and was licensed by the Department of Parks and Wildlife (SF008067, SF008682, SF009286 and SF009874). Biopsy sampling for molecular analyses were carried out as a part of broader study, with data collected in accordance with the Murdoch University Animal Ethics Committee approval (W2076/07; W2307/10; W2342/10 and R2649/14), and collected under research permits (SF005997; SF006538; SF007046; SF007596; SF008480; SF009119: SF009734; SF010223) licensed by Department of Parks and Wildlife.

Acknowledgements

The completion of this PhD marks the end of a long and eventful journey and could not have been possible without the guidance and support from many people, especially during the last few months. I have met so many people throughout this journey: some that were around from the beginning to the end and others that I briefly met along the way. To everyone, thank you so much for your contribution, big or small, it has significantly helped me to get through it.

I would like to thank Lars Bejder and Hugh Finn for not giving up on me when I was about to embark on a different adventure. Your guidance and suggestions during the planning and writing up of the thesis were invaluable and helped me improve my critical thinking and scientific writing (well, I hope so!). I’m particularly grateful to both of you for your understanding and encouragement during my chaotic last few months. To Lars, thank you for trusting me for many years and giving me the opportunity to be part of the team (again). To Hugh, thank you for sharing so many of your anecdotes with me. Our discussions have always helped me to find the right direction and given me more confidence.

Sincere thanks to Bill Sherwin, for accepting to be my genetic mentor and making me part of his team. Bill, I have learnt so much over our Skype conversations and the weekly lab meetings with your students. Also, thanks go to Oliver Manlik for taking the time to discuss genetic data and provide some advice. To Celine Frère, thank you for introducing me to the genetic world. You welcomed me in your lab where I had the privilege to experience the genetic lab work, learn how to process and analyse data and have brainstorming sessions where your expertise was much appreciated.

A special thanks to Kenneth Pollock for teaching me what I know about mark- recapture. It has been a privilege to work with you, Ken. Our discussions on multistate mark-recapture and your enthusiasm for my research have inspired me. I wish you all the best in your retirement.

vii

I am particularly grateful to Holly Raudino for giving me the opportunity to assist her during her PhD in Bunbury (WA). You taught me the essentials I needed to pursue my own project. You helped me to step up in my goals by trusting me and letting me lead the fieldwork when you were writing up your thesis. You took me under your wing and always made sure that I was OK and still do. Your ongoing support means a lot to me.

In that learning process, I’m also very much thankful to Simon Allen for taking the time to teach me how to biopsy a dolphin and for sharing many of your fieldwork experiences. I will forever thank you for caring, listening and giving me advice when unexpected things occurred.

My fieldwork would not have been possible without the dedicated help from many assistants and interns. Thank you all for keeping the motivation up and countless hours collecting and processing data. In alphabetical order, thank you to: Ana Carolina Andrade, Emmeline Audic, Rebecca Bakker, Bronte Bates, Maree Bekkers, Anna Bendz, Martin Binet, Lisa Binks, Alex Brown, Jacklyn Buchanan, Dennis Buffart, Joanna Burgar, Anita Byrone, Rene Byrskov, Elizabeth Chabanne, Sharon Chan, Manon Chautard, Elisa Chillingworth, Lucie Chovrelat, Ana Costa, Alicia day, Aurelien Delume, Brechtje de Schipper, Georgie Dicks, Gwenael Duclos, Vivienne Foroughirad, Julia Friese, Hayley Gamble, Beatriz Gemez, Kate Greenfeld, Olivia Hamilton, Nora Handrek, Daniella Hanf, Emily Hanley, Lisa-Marie Harrison, Lauren Hawkins, Joe Heard, Nikki Hume, Lonneke IJsseldijk, Kim Jaloustre, Camille Jan, Vanessa Jaith, Ben Jenhinson, Maria Carmen Jimenez, Martyn Kelly, Anna Kopps, David Kranitz, Krista Krijger, Alexis Levengood, Wey Yin Lo, Cassandra Magnin, Jim Maureau, Kelsey Mcclellan, Severine Methion, Micol Montana, Lisa Mueller, Jesse Murdoch, Nao Nakamura, Krista Nicholson, Harriet Park, Natasha Prokop, James Raeside, Rita Reis, Nick Riddoch, Nina Schäfer, Ashleigh Shelton, Lisa Shulander, Claire Smart, Megan Smook, Kate Sprogis, Nahiid Stephens, Sam Tacey, Alexandra Thibaudeau, Romain Tscheiller, Julian Tyne, Adam Urwin, Nieki van der Kuijl, Jessica Vermass, Rémi Vignals, Yvette Vignals, Lizette Viljoen, Timothy Walker, Maike Werner, Imogen Webster, Shona, Wharton, Dion Whittle and Renae Williams.

viii

For the genetic data, I would like to thank Celine Frère and Bethan Littleford- Colquhoun for carrying out molecular analyses on the tissue samples I collected in Perth. I would also like to thank people who have been involved in collecting and analysing biopsy samples from other dolphin populations along the southwestern coastline of Western Australia: Simon Allen, Anna Sellas, Dee McElligot, Michael Krützen and Lars Bejder for biopsy sampling dolphins; Claire Daniel and Oliver Manlik for carrying out the molecular analyses.

Massive thanks to my fellow lab mates Simon Allen, Alex Brown, Fredrik Christiansen, Daniella Hanf, Amanda Hodgson, Krista Nicholson, Rob Rankin, Jenny Smith, Josh Smith, Kate Sprogis, Nahiid Stephens, Julian Tyne, for their time, collaboration, support, the many useful, or not, discussions during coffee (hot chocolate for me ) and for sharing their experiences. Special thanks to Julian and Krista, for making the office less empty, checking on me when not in the office and for the trips to the science store for more chocolates; and to Nahiid, for making sure that my belly was full of really yummy food.

I would like to thank the Swan River Trust (now merged with the Department of Parks and Wildlife) for financial and logistic support throughout many years. Among many others, thank you to Marnie Giroud, Rachel Hutton, Jason Menzies and Kerry Trayler for supporting the research work and trusting me. Thank you to Chandra Salgado-Kent and Sarah Marley for checking on me and the good times sharing our respective research to the Dolphin Watch volunteers. Thanks also go to the Fremantle Ports for their annual donations. I would also like to thank the Sailing Club of Fremantle. Access to the boat ramp and the fuel jetty made our long days out on the water much easier.

Big thanks to my family back home and friends from all around the world for their continuous and tremendous support and encouragement as well as their patience and understanding. A special thank you to Nathalie Long for letting me know Holly was looking for assistants. This is how my journey down under started. To my fellow climbers, thank you for adding more adventures in my life and the well-deserved breaks.

This acknowledgement would not be complete if I did not mention Rémi. Thank you for belaying me when climbing this mountain (aka ‘PhD’), for securing me, and for taking the rope when I needed breaks. I had always valued your help and advice, particularly when it came to presentation and design. I also thank you for giving me many precious memories, taking me exploring and climbing real rocks. Although we do not partner anymore, I will always care about you.

Last, but definitely not least, a big thank you to my mum and dad. Your ongoing support and encouragement helped me following my dreams even so it meant going down under. Je vous aime…

Thesis Publications

The following publications are associated with the chapters of this thesis:

Chabanne, D. B. H., Pollock, K. H., Finn, H. and Bejder L. (2017) Applying the Multistate Capture-recapture Robust Design to investigate metapopulation structure. Methods in Ecology and Evolution. DOI: 10.1111/2041-210X.12792. (Chapter 2)

Chabanne, D. B. H., Finn, H. and Bejder L. (2017) Identifying the relevant local population for environmental impact assessments of mobile marine fauna. Frontiers in Marine Science, 4: 148. DOI: 10.3389/fmars.2017.00148. (Chapter 3)

xii

Table of contents

Declaration ...... i Abstract ...... iii Statement on the contribution of others ...... v Acknowledgements ...... vii Thesis Publications ...... xi List of Tables ...... xvii List of Figures ...... xx List of Appendices ...... xxiii Glossary ...... xxv List of abbreviations and symbols ...... xxvi

Chapter 1. General Introduction ...... 1

1.1. Characterising the population structure of small cetaceans in coastal and estuarine environments ...... 1

1.2. The Indo-Pacific bottlenose dolphin ...... 8

1.3. Indo-Pacific bottlenose dolphin population in Perth metropolitan waters 9

1.4. Objectives and structure of the thesis ...... 12

1.5. Ethics statement ...... 13

Chapter 2. Applying the multistate capture-recapture robust design to assess metapopulation structure ...... 15

2.1. Abstract ...... 15

2.2. Introduction ...... 17

2.3. Materials and methods ...... 22 2.3.1. Field methods ...... 22 2.3.2. Statistical methods ...... 26

2.4. Results ...... 27 2.4.1. Effort ...... 27 2.4.2. Model selection ...... 27

2.5. Discussion ...... 33 2.5.1. Spatial heterogeneity...... 33 2.5.2. Apparent survival and abundance estimates ...... 34

xiii

2.5.3. Transitions ...... 35 2.5.4. Limitations ...... 35 2.5.5. Conclusion ...... 37

Appendices ...... 38

Chapter 3. Identifying the relevant local population for environmental impact assessments of mobile marine fauna ...... 57

3.1. Abstract ...... 57

3.2. Introduction ...... 58

3.3. Materials and methods ...... 62 3.3.1. Study area ...... 62 3.3.2. Data collection ...... 65 3.3.3. Association patterns ...... 66 3.3.4. Community structure and dynamic ...... 67 3.3.5. Spatial distribution of communities ...... 68 3.3.6. Residence time ...... 69 3.3.7. Genetic relatedness ...... 69

3.4. Results ...... 70 3.4.1. Effort and group size ...... 70 3.4.2. Community structure and dynamic ...... 71 3.4.3. Association patterns ...... 72 3.4.4. Spatial distribution of communities ...... 79 3.4.5. Residence time ...... 82 3.4.6. Genetic relatedness ...... 82

3.5. Discussion...... 85 3.5.1. Community segregation ...... 87 3.5.2. Using social, ecological, and genetic data to evaluate the impact of developments and activities on the relevant local population: four case studies ... 89 3.5.3. Conclusions ...... 93

Appendices ...... 94

Chapter 4. Genetic structure of socially and spatially discrete subpopulations of dolphins (Tursiops aduncus) in Perth, Western Australia .. 105

4.1. Abstract ...... 105

4.2. Introduction ...... 106

xiv

4.3. Materials and methods ...... 108 4.3.1. Genetic sample collection ...... 108 4.3.2. DNA extraction and sexing...... 110 4.3.3. Genotyping and validation of microsatellites ...... 110 4.3.4. Mitochondrial (mt) DNA sequencing ...... 111 4.3.5. Assessment of genetic differentiation ...... 111 4.3.6. Assessment of genetic population structure...... 111 4.3.7. Analysis of gene flow ...... 112 4.3.8. Assessment of genetic diversity ...... 112

4.4. Results ...... 113 4.4.1. Validation of genotypes and haplotypes ...... 113 4.4.2. Genetic differentiation ...... 115 4.4.3. Genetic population structure ...... 116 4.4.4. Gene flow and dispersal ...... 118 4.4.5. Sex-biased dispersal ...... 118 4.4.6. Genetic diversity ...... 119

4.5. Discussion ...... 122 4.5.1. Genetic structure and gene flow ...... 123 4.5.2. Conservation implications ...... 126 4.5.3. Conclusion ...... 128

Appendices ...... 129

Chapter 5. Population genetic structure and effective population sizes in Indo-Pacific bottlenose dolphin (Tursiops aduncus) along the southwestern coastline of Western Australia ...... 135

5.1. Abstract ...... 135

5.2. Introduction ...... 136 5.2.1. Genetic structure ...... 136 5.2.2. Effective population size and effective/census population size ratio .... 138

5.3. Materials and methods ...... 140 5.3.1. Study site and sample collection ...... 140 5.3.2. DNA extraction and Microsatellite genotyping ...... 142 5.3.3. Genetic diversity ...... 142 5.3.4. Genetic differentiation and population structure ...... 143 5.3.5. Gene flow ...... 144

5.3.6. Effective population (Ne) sizes ...... 144 5.4.1. Genetic diversity ...... 145 5.4.2. Genetic differentiation and population structure ...... 146

5.4.3. Contemporary gene flow ...... 150 5.4.4. Effective population sizes ...... 152

5.5. Discussion...... 154 5.5.1. Genetic diversity ...... 154 5.5.2. Genetic differentiation and structure ...... 155 5.5.3. Contemporary gene flow ...... 156

5.5.4. Contemporary effective population size and Ne/Nc ratio ...... 157 5.5.5. Conclusions and future research ...... 160

Appendices ...... 162

Chapter 6. General Conclusions ...... 175

6.1. Keys research findings for each subpopulation (Figure 6.1) ...... 176 6.1.1. The Swan Canning Riverpark estuarine subpopulation ...... 177 6.1.2. The semi-enclosed embayment subpopulations: Owen Anchorage and Cockburn Sound ...... 178 6.1.3. Bottlenose dolphins occurring along the open Gage Roads coastline ... 179

6.2. Theoretical and practical limitations of the study ...... 182 6.2.1. Fine-scale delineation of subpopulations ...... 182 6.2.2. Ecological characteristics of Gage Roads: are there sufficient data? .... 182 6.2.3. Genetic sampling is not random ...... 183

6.2.4. Contemporary effective population size (Ne) and ratio of effective/census population size (Ne/Nc): unresolved conservation tools ...... 184

6.3. Management recommendations ...... 185

6.4. Concluding remarks ...... 188 References ...... 191

xvi

List of Tables

Table 2.1. Estimates of transition probability ψ (SE) between sites for (a) Scenario 1 (three sites: SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and (b) Scenario 2 (two sites: SCR vs. Coastal). .... 12 Table 3.1. Number of groups and group size (mean (SE), minimum and maximum) by geographic region...... 70 Table 3.2. The mean association indices for the population overall and per community (ComA, ComB, ComC and ComD); the measure of social differentiation (S); and the correlation coefficient of the true and estimated association matrices (r)...... 73 Table 3.3. Association indices within and among sex classes (Mantel test, 0.05 one- side)...... 74 Table 3.4. Average strength, clustering coefficients, and affinity (SE) with comparisons from random calculating using half-weight indices for individuals sighted at least five times...... 75 Table 3.5. Sex class (female, male, unknown sex) permutation tests for preferred (HWI CV) and avoided (proportion of non-zero indices) associations for the population overall and per community...... 76 Table 3.6. Bootstrap mean (and standard error) genetic relatedness (Queller and Goodnight, 1989) for within-community (bold) and with individuals from other communities of bottlenose dolphins (between-community) genotyped with ten microsatellite loci in Perth metropolitan waters, WA...... 84 Table 3.7. Correlation coefficient R and ANOVA test between relatedness coefficient and HWI pairwise within-community...... 84 Table 3.8. Summary comparison of social, temporal, spatial, residency and genetic patterns across the four communities (ComA, ComB, ComC and ComD). Social differentiation is described using measures of strength (i.e., measure of gregariousness); the clustering coefficient (i.e., degree of connection between associates); and affinity (i.e., the strength of the associates)...... 86 Table 4.1. Microsatellite diversity in Indo-Pacific bottlenose dolphins inhabiting Perth metropolitan waters...... 114 Table 4.2. Pairwise fixation indices between four genetic clusters previously defined in STRUCTURE analysis based on ten microsatellite loci and mtDNA control sequences. Microsatellite FST values are above the diagonal; Mitochondrial ΦST values are below. Mitochondrial FST values were similar to ΦST, and thus are not shown here; values that are significant after sequential Bonferroni correction are in bold. Values in italic were significant before correction (P-value < 0.05)...... 115

xvii

S Table 4.3. Pairwise comparisons of Shannon mutual information index ( HUA) among socio-geographic subpopulations based on ten microsatellite loci (allele frequency differences, above diagonal) and mtDNA control region sequences (haplotype frequency differences, below diagonal). Significant P-values after sequential Bonferroni correction and based on statistical testing of 999 random permutations are shown in bold. Values in italic were significant before correction (P-value < 0.05)...... 116 Table 4.4. Mean (standard deviation) of the posterior distribution of the contemporary migration rates (m) in BayesAss (Wilson and Rannala 2003) among four bottlenose dolphin socio-geographic subpopulations in Perth metropolitan waters. The subpopulations of which each dolphin belongs are listed in the rows, while the subpopulations from which they migrated are listed in the columns. Values along the diagonal (in bold) are the proportions of non-immigrants from the origin subpopulation for each generation. Moderate estimated migration rates (m > 0.10) are displayed in italic...... 118

Table 4.5. Sex-specific FST values based on microsatellite loci and mtDNA for all and for each pairwise socio-geographic subpopulations. Significance levels of genetic differentiation between socio-geographic subpopulations are also indicated for FST values (estimated with Arlequin, after Bonferroni correction: * P-value < 0.05)...... 119 Table 4.6. Genetic diversity measures (SE) for bottlenose dolphin socio-geographic subpopulations using microsatellite loci (n = 10) and mtDNA...... 120 Table 4.7. Summary statistics of various tests to detect a recent bottleneck effect based on mtDNA control region (Tajima’s D and Fu’s Fs) and microsatellite loci (SMM: stepwise mutation model. TPM: two-phased model). The Wilcoxon test found no significant heterozygosity excess after Bonferroni correction (Pcrit = 0.012)...... 122 Table 5.1. Genetic diversity for microsatellite loci in bottlenose dolphin sampled in eight localities...... 146

Table 5.2. Genetic differentiation (FST) among eight bottlenose dolphin sampling localities in southern Western Australia...... 147 Table 5.3. Mean (and 95% CI) recent migration rates inferred using BayesAss 3.0 (Wilson and Rannala 2003). The migration rate is the proportion of individuals in a population that immigrated from a source population per generation. Values of CI that do not overlap with zero are in bold. Values along the diagonal (underlining) are the proportions of non-immigrants from the origin subpopulation for each generation...... 151 Table 5.4. Estimates of the contemporary effective population size before correction

(Ne, 95% CI) and after correction (cNe, 95% CI), ratio cNe/Nc (SE) for localities with known census population size (Nc, , 95% CI), and estimated eNc (95% CI) for populations with unknown Nc. The critical value varied according to the sample size xviii

(n > 25, Pcrit = 0.02; n > 50, Pcrit = 0.01). cNe was corrected for overlapping generation by adding 15 or 10% of the estimates based on the Pcrit, respectively. .. 153

xix

List of Figures

Figure 1.1. Practical approaches (with considerations) and scientific information used to identify putative ‘units to conserve’ (subpopulations or local populations) in a population of a small cetacean species...... 7 Figure 1.2. The three areas previously studied for bottlenose dolphins in the metropolitan waters of Perth (Western Australia) and the scale of the current study (red square, 2011-15). Area 1 – Open coastline (more effort in dark purple area than light purple area, Waples 1997); Area 2 – Cockburn Sound (green, Finn 2005); Area 3 – Swan Canning Riverpark (blue, Chabanne et al. 2012). RI = Rottnest Island; SCA = Scarborough...... 11 Figure 2.1. Traditional closed robust design (CRD) vs. multistate closed robust design (MSCRD) approaches to characterise metapopulation structure and dynamics through demographic parameters. Both approaches allow estimation of abundance (N), apparent survival rate (φ) and emigration and immigration [solid arrows] either time varying (t, t+1, t+2, etc.) or constant. In addition, MSCRD models estimate any transition probabilities ψ [dashed arrows] between subpopulations associated with states (i.e., geographic sites)...... 20 Figure 2.2. Map of the metropolitan waters of Perth, Western Australia, showing the systematic survey routes within each site: the estuary SCR = Swan Canning Riverpark and the coastal sites (south to north) CS/OA = Cockburn Sound/Owen Anchorage, and GR = Gage Roads. Within the coastal sites, surveys were conducted by rotating between three pre-defined transect routes (full, long dash and short dash lines) to maximise the coverage...... 24 Figure 2.3. Capture probability (p) yielded by the models for each secondary occasion represented as box plot (min; Quartile 1; median; Quartile 3; max) for each Scenario: 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads); and 2 – two sites (SCR and Coastal)...... 29 Figure 2.4. Seasonal estimated abundances (Ntotal ± 95% confidence intervals) for (a) Scenario 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and (b) Scenario 2 – two sites (SCR and Coastal). Lines between data points have been used for illustrative purposes only; continuity of values is not implied. Sites are as follows: SCR (red), CS/OA (yellow), GR (purple) and Coastal in Scenario 2 (grey)...... 30 Figure 3.1. Maps of the study area showing: (a) the transect routes per geographic regions (GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound) with locations of past, current, and proposed developments (1- pile driving; 2- dredging; 3- desalination; 4- outer harbour); and (b) the bathymetry (in meters) with the locations of the groups sighted during the

systematic surveys conducted from 2011 to 2015, including mixed groups (i.e., mix of individuals from different communities)...... 54 Figure 3.2. Network diagram for 129 bottlenose dolphins using the HWI. The shape of each node indicates its sex (circle: females; square: males; triangle; unsexed), and the colour of each node indicates its unit defined by the modularity of Newman (2006) (purple ComA; red ComB; green ComC; yellow and orange sub-communities D and D’), although three individuals were assigned to two different units depending on the method (Newman vs. Hierarchical linkage). Only links representing affiliations (HWI > 0.16) are shown, and link width is proportional to index weight. Node size is based on the betweenness centrality measure of each individual...... 72 Figure 3.3. Lagged association rates (LAR) for all individuals (black line) and within the communities (ComA purple; ComB red; ComC green; and ComD yellow). The null association rate (dash lines) and jackknife error bars are shown. . 77 Figure 3.4. Lagged association rates (LAR) for males and females (faint colour) bottlenose dolphins of each community ((a)-ComB, (b)-ComC and (c)-ComD). The null association rate (dash lines) and jackknife error bars are shown. Note that the LAR for ComA could not be estimated because of small sample sizes...... 78 Figure 3.5. Study area showing the bathymetry and core areas (a) based on the 50% kernel density and home ranges (b) based on the 95% kernel density estimated for each community using clustered individual sightings. Communities are: ComA- purple; ComB-red; ComC-green; and ComD-yellow...... 71 Figure 4.1. Map of the four geographic regions (in bold) associated with each socio- ecological subpopulation within the Perth metropolitan waters, Western Australia...... 109 Figure 4.2. Bayesian assignment probabilities from STRUCTURE for bottlenose dolphins based on ten microsatellite loci: (a) without prior information, K = 3. (b) With prior information, K = 4. Each vertical line represents one individual, with the strength of that individual to any of the genetic clusters (blue cluster 1; red cluster 2; green cluster 3; purple cluster 4). Individuals are grouped by social subpopulations (GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound) and sorted within subpopulations by the latitude of sampling location from North (left) to South (right) when available...... 117 Figure 4.3. Median-joining network of mtDNA control region haplotypes in Indo- Pacific bottlenose dolphins in Perth metropolitan waters. The size of the circles is proportional to the total number of individuals carrying that haplotype. Different colours denote the four different sampled subpopulations: purple GR = Gage Roads, red SCR = Swan Canning Riverpark, green OA = Owen Anchorage and yellow CS = Cockburn Sound. Number of mutational events between each haplotype is indicated by hash marks...... 121 Figure 5.1. Map of the sampling sites, southwestern Australia, showing the biopsy sample collection sites for bottlenose dolphins (n = 221) from eight localities: blues

xxi

= Perth (PE) including Swan Canning Riverpark (SCR) and Cockburn Sound (CS/OA); black = Mandurah (MH); green = Bunbury (BB); orange = Busselton (BS); purple = Augusta (AU); grey = Albany (AL); and pink = Esperance (ES). .. 141 Figure 5.2. Structure plots showing assignment probability of each dolphin (one individual per column) to the respective populations for different number of clusters, K: (a) with K = 2 and (b) K = 4 for the full dataset (n = 221); and (c) with K = 3 for Augusta, Albany, and Esperance only (n = 53). PE = Perth (including SCR = Swan Canning Riverpark and CS/OA = Cockburn Sound), MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance...... 149 Figure 6.1. Summary overview of the conservation status of bottlenose dolphins in the four geographic regions of Perth metropolitan waters: SCR = Swan Canning Riverpark (red), CS = Cockburn Sound (yellow), OA = Owen Anchorage (yellow), GR = Gage Roads (green)...... 181

xxii

List of Appendices

Appendix A2.1. Survey effort ...... 38 Appendix A2.2. The CloseTest: a method to investigate of the population closure . 39 Appendix A2.3. Data processing for photograph quality and individual distinctiveness ...... 40 Appendix A2.4. Method to estimate proportion of distinctly marked individuals and correction of the marked abundance estimates for consideration of the proportion of unmarked individuals (D3 individuals identified from Q1 and Q2) ...... 41 Appendix A2.5. Test for heterogeneity in capture probabilities by implementing goodness-of-fit tests for multistate models using the program U-CARE ...... 43 Appendix A2.6. Bottlenose dolphin groups ...... 44 Appendix A2.7. Bottlenose dolphin individuals ...... 46 Appendix A2.8. Multistate closed robust design models summary ...... 13 Appendix A2.9. Sighting frequency ...... 48 Appendix A2.10. Coefficient of variations (CV) of capture probability (p) ...... 49 Appendix A2.11. MSCRD analysis using sex as an individual covariate ...... 50 Appendix A3.1. Social dendrogram ...... 94 Appendix A3.2. Tests of differences in distribution of centrality values between communities ...... 95 Appendix A3.3. Network parameters ...... 96 Appendix A3.4. Best-fitted models for the lagged association rate (LAR) ...... 55 Appendix A3.5. Core areas and home ranges ...... 100 Appendix 3.6. Bathymetry ...... 101 Appendix A3.7. Overlap between social communities ...... 102 Appendix A3.8. Best-fitted models for the lagged identification rate (LIR) ...... 55 Appendix A4.1. Linkage disequilibrium ...... 129 Appendix A4.2. Mean of the posterior probabilities (LnP(D)) and ΔK statistic for STRUCTURE ...... 130 Appendix A4.3. Mean assignment indices (AIc) ...... 131 Appendix A4.4. Sex-biased dispersal ...... 132 Appendix A4.5. Distribution of allelic frequencies ...... 133 Appendix A5.1. Genetic diversity for 25 microsatellite loci ...... 162 Appendix A5.2. Bottleneck tests ...... 167

xxiii

Appendix A5.3. Isolation by distance ...... 168 Appendix A5.4. Mean of the posterior probabilities (LnP(D)) and ΔK statistic for STRUCTURE ...... 170 Appendix A5.5. Correlation between sample size n and estimated effective population size (Ne and cNe)...... 171 Appendix A5.6. Abundances estimated for dolphins using the Perth metropolitan waters using single state Closed Robust Design models ...... 172

xxiv

Glossary

The terms defined in the glossary will be highlighted by a dagger (†) when first used in the text:

Term Definition Allele Variant form of a given gene (Taylor et al. 2010).

Allele richness Measure of the number of alleles that takes into account variations in sample size (Greenbaum et al. 2014).

Bottleneck Drastic reduction of the effective population size† of a population (Slatkin 2008).

Effective population Number of breeders in an idealised population (in Hardy- size Weinberg equilibrium†) that would show the same amount of genetic drift† or inbreeding or accumulation of linkage disequilibrium† as the population under consideration (Taylor et al. 2010).

Genetic drift Random variation of allele frequencies in transmission between generations (Taylor et al. 2010; Nielsen and Slatkin 2013).

Haplotype Group of genes in an organism that are inherited together from a single parent (Nei 1987).

Haplotypic diversity Probability that two randomly sampled alleles or haplotypes† are different (Nei 1987).

Hardy-Weinberg Principle stating that allele and genotype frequencies in a equilibrium population remain constant from generation to generation in the absence of other evolutionary influences (Nei 1987).

Linkage Non-random association of alleles at different loci (Slatkin disequilibrium 2008).

Nucleotide diversity Average number of nucleotide differences per site in pairwise comparisons among DNA sequences (Nei and Li 1979).

Null allele Allele (at a microsatellite locus) which is present in a sample but which consistently fails to amplify during polymerase chain reaction (PCR). Amplification of the allele can be inhibited because of a mutation in the primer binding region (Chapuis and Estoup 2007).

Private alleles Alleles that are only found in one population (Szpiech and Rosenberg 2011).

xxv

List of abbreviations and symbols

Abbreviation Meaning Symbol θ Proportion of distinctly marked individuals π Nucleotide diversity φ Survival rate ψ Movement transition

φST Nucleotide differentiation AIc Assignment index correction

AICc Akaike information criterion

AR Allele richness c Probability of recapture CCC Cophenetic coefficient correlation CI Confidence interval CRD Closed robust design CS Cockburn Sound CV Coefficient of variation DMSO Dimethyl sulfoxide EIA Environmental impact assessment

FIS Inbreeding coefficient

FST Fixation index (genetic distance/differentiation) GOF Goodness-of-fit GR Gage Roads h Haplotypic diversity

HO Observed heterozygosity

HE Expected heterozygosity hrs Hours HWE Hardy-Weinberg equilibrium HWI Half-weight index IBD Isolation by distance KDE Kernel density km Kilometre LAR Lagged association index LIR Lagged identification index xxvi

m Migration M Modularity MCMC Markov chain Monte Carlo MR Mark-recapture MS Multistate mark-recapture MSCRD Multistate closed robust design mtDNA mitochondrial DNA n Sample size

N or Nc Abundance -- Number of individuals in the study area (Chapters 2 and 3) or Census population size in genetic context (Chapters 4 and 5)

NA Number of alleles

Ne Effective population size NM Nautical mile OA Owen Anchorage p Probability of capture PA Private alleles QG Queller and Goodnight index R Coefficient of correlation r Frequency of null alleles S Social differentiation SCR Swan Canning Riverpark SE Standard error S HUA Shannon’s mutual information index SMM Stepwise mutation model td Time lag TPM Two-phase model WA Western Australia

xxvii

xxviii

Chapter 1. General Introduction

1.1. Characterising the population structure of small cetaceans in coastal and estuarine environments

Coastal and estuarine ecosystems adjacent to urban centres are challenging environments for small cetaceans (Reeves et al. 2003; Jefferson et al. 2009; Cagnazzi et al. 2013a; Derville et al. 2016). Small cetaceans inhabiting these areas may experience: habitat loss and degradation (Ross 2006; Culloch et al. 2016); incidental mortality from indirect and direct interactions with commercial and recreational fisheries (Dawson and Slooten 1993; Curry and Smith 1997; Slooten et al. 2000; Chilvers et al. 2003; Reeves et al. 2003); disturbance and harassment from vessel interactions and anthropogenic noise (Donaldson et al. 2010; Pirotta et al. 2013; Christiansen et al. 2016; Culloch et al. 2016; Marley et al. 2016); and exposure to environmental contaminants (Todd et al. 2015). These stressors can affect the behaviour, physiology, and health of cetaceans and reduce reproductive success and survival, particularly if stressors exert cumulative or synergistic impacts (Bejder et al. 2006; Van Bressem et al. 2009; Jefferson et al. 2009; Christiansen and Lusseau 2015). Given the complex and heterogeneous distribution of small cetaceans in these habitats, it can be challenging to identify and implement appropriate conservation measures. Therefore, it is vital to improve assessments of the population status, the biological significance of human impacts, and the effectiveness of management approaches (Taylor 2005; Bejder et al. 2009; Berger-Tal et al. 2011). In particular, the methods and theoretical frameworks used to characterise their population structures should be able to identify the appropriate putative ‘units to conserve’ (sensu Taylor 2005).

The heterogeneity of coastal and estuarine environments means that, rather than there being a single continuously distributed population or species, small cetaceans are often distributed as multiple, localised populations or subpopulations (e.g., Brown et al. 2016). Local- or sub-populations are often defined by their association with particular areas and are linked, in varying degrees, to each other, but often have unique characteristics (demographic, ecological or genetic) that make them

1 Chapter 1. General Introduction somewhat distinctive (e.g., Curry and Smith 1997; Sellas et al. 2005; Möller et al. 2007). A subpopulation can be characterised as having: 1) distinctive ecological characteristics (e.g., use of particular prey, foraging tactics, or habitats); 2) strong associations between particular individuals; and 3) natal philopatry (i.e., retention of maternal home range, Rossbach and Herzing 1999; Wells et al. 1999; Connor et al. 2000b). Wells et al. (1999) used the term ‘community’ to define a ‘regional society of animals sharing ranges and associates’. Subpopulation and community definitions overlap to some extent, but a community should still exhibit genetic exchange with other similar units.

While there is much enthusiasm for identifying subdivisions within a species’ range, humans’ perceptions of such subdivisions are often not very biologically meaningful (Taylor, 2002). Nonetheless, it is better to consider subdivisions as humans perceive them, and estimate dispersal between these subdivisions, which is the most useful information for both evolutionary biology and management. This thesis is based on the premise that, in some circumstances, a subpopulation will be an appropriate ‘unit to conserve’. Rather than engaging in an exhaustive discussion of the policy or statutory bases for how management units or ‘units to conserve’ are to be defined, I intend to provide – in the context of a major metropolitan area with multiple anthropogenic stressors present – an empirical framework for the population structure of small cetaceans in an urbanised region. I then discuss potential ‘units to conserve’ in that region, based on the evidence presented and available information on human environmental impacts.

Small cetaceans are long-lived species characterised by slow growth rates, late maturation, and low reproductive rates (Taylor 2002; LeDuc 2009). As a consequence, the persistence of populations depends on their ability to sustain a net positive reproductive output (Caughley 1977). Thus, human activities that either singly or cumulatively affect the health of individuals in a population by repeatedly disturbing ecologically vital processes (e.g., resting, foraging, and reproducing, Lusseau 2003a; Bejder et al. 2006; Tyne et al. 2014) or through direct mortality (e.g., incidental capture in fisheries and boat strikes, Stone and Yoshinaga 2000; Lewison et al. 2004; Baker et al. 2013) may jeopardise the population’s viability.

2 Chapter 1. General Introduction

The level and intensity of human activities vary across regions and habitats, and each subpopulation (which may have differing demographic, ecological and genetic characteristics) has varying levels of exposure to human impacts. Therefore, the identification of ‘units to conserve’ in a population is a fundamental aspect of impact assessment (Taylor 2005), but not currently required under Australian wildlife protection law. In contrast, the United States Marine Mammal Protect Act 1972 specifically requires the identification of marine mammal ‘stocks’ for management. The scientific argument for considering subpopulations as ‘units to conserve’ is based on evidence that subpopulations: (1) are demographically independent from populations in neighbouring areas; (2) maintain unique associations with a particular geographic area or ecosystem; (3) are genetically differentiated; and (4) possess unique cultural traditions (Rendell and Whitehead 2001; Sellas et al. 2005; Taylor 2005; Möller et al. 2007; Fury and Harrison 2008; Urian et al. 2009; Wiszniewski et al. 2009).

Individual-based behavioural and genetic parameters are typically used to characterise the population structure of small cetaceans (Figure 1.1) (e.g., Lusseau and Newman 2004; Tezanos-Pinto et al. 2009; Urian et al. 2009; Wiszniewski et al. 2009; Titcomb et al. 2015; Thompson et al. 2016). An analysis of residency patterns examines how individuals occupy an area over time, allowing for discrimination between animals present only for certain periods and those inhabiting an area continuously (i.e., site fidelity) (Whitehead 2001; Zolman 2002; Chilvers and Corkeron 2003; Chabanne et al. 2012). Mark-recapture models can estimate abundance and apparent survival that define the probability of surviving and staying within a region and can also be interpreted as a measure of residency (Smith et al. 2013; Brown et al. 2016). Analyses of association patterns between individuals can characterise the social structure within an area, including the existence of distinct groupings (Whitehead 1995; Urian et al. 2009; Wiszniewski et al. 2009; Oudejans et al. 2015; Titcomb et al. 2015). Analyses of the kernel density index (KDE) can describe the home range of subpopulations and can be used to infer the degree of spatial distinctiveness (i.e., site fidelity that can be associated with use of particular prey or foraging tactics) (Wiszniewski et al. 2009; Sprogis et al. 2015; Titcomb et al. 2015). Analyses of genetic markers can be used to describe whether or not habitat boundaries and residency in sheltered ecosystems can also promote genetic

3 Chapter 1. General Introduction differentiation, genetic relatedness or lack of dispersal (i.e., genetic structure) between groups of cetaceans ranging over relatively small geographical distances (Möller et al. 2007; Andrews et al. 2010).

Below, I discuss three practical approaches (see Figure 1.1) and considerations that are relevant to the methods applied in this study to characterise the local dolphin population structure in and around Perth metropolitan waters and to identify potential ‘units to conserve’.

Study design for systematic surveys The study of small cetaceans in coastal and estuarine environments requires designs that allow for the systematic collection of individual-specific spatial and social data (Figure 1.1; see Morrison et al. 2008a). Consideration of several practical issues must be made, and should include:  Deciding on the study area and spatial scale(s) of interest (i.e., the overall seascape and particular sub-areas);  Investigating a range of variables, such as the environmental features (e.g., salinity, temperature, and bathymetry) used to delineate sub-areas/habitats in coastal and estuarine ecosystems. While stratification of a study area into blocks for sampling purposes is generally related to logistical considerations, it can have specific research applications (e.g., to generate abundance estimates for different sub-areas/habitats) and may increase the precision of various estimated parameters (Morrison et al. 2008b; Hammond 2010);  Applying a realistic timeframe for sampling (i.e., knowledge of weather conditions in the study area and available daylight hours when designing survey transects); and,  Balancing the study area coverage with field survey time and budget. A survey design should aim for equal coverage probability of the study area (including between blocks or sub-areas). In practice, a design involving a set of zig-zag transect lines has been found to be the most appropriate for boat-based surveys of small cetaceans, because it allows for relatively even coverage and minimises time spent transiting between transects (Hammond 2010).

4 Chapter 1. General Introduction

Photo-identification Photo-identification (Figure 1.1) techniques were first applied to small cetaceans in the early 1970s and are now an integral component of many field studies (Würsig and Jefferson 1990). The objective of this technique is to recognise individuals by using natural marks (or those from anthropogenic sources such as boat strikes and entanglements) recorded in photographs. There are practical challenges to obtaining photographs of good quality and protocols to ensure appropriate sampling of well- marked, as well as un-marked (‘clean fin’) individuals, must be carefully considered (Urian et al. 2014). Nonetheless, photo-identification has proved to be one of the most useful approaches for understanding small cetacean life history (Hammond et al. 1990) and for mark-recapture studies that assess abundance and apparent survival (e.g., Nicholson et al. 2012; Ansmann et al. 2013; Brown et al. 2016; Sprogis et al. 2016). When combined with individual spatial and temporal information, photo- identification can be used to identify site fidelity, movement patterns and home range size, as well as social structure (e.g., Chabanne et al. 2012; Oudejans et al. 2015; Sprogis et al. 2015; Titcomb et al. 2015).

Biopsy sampling Remote biopsy sampling (Figure 1.1), using either a crossbow or modified rifle to propel a small biopsy dart, is now in widespread used for obtaining tissue samples from free-ranging cetaceans (Krützen et al. 2002). Biopsy darting has facilitated genetic analyses to determine the sex of individuals, which is important for monitoring population dynamics (e.g., Frère et al. 2010b; Sprogis et al. 2016). Recognising sex-specific ecological differences (e.g., in ranging patterns) within a population could indicate issues for conservation and management, as one sex may be more susceptible to anthropogenic threats than the other (e.g., Lusseau 2003b; Stensland and Berggren 2007; Sprogis et al. 2015). Secondly, techniques for the use of genetic markers (e.g., microsatellite loci and mitochondrial DNA) are now well- advanced (Frankham 1995a; Wan et al. 2004) and are able to support analyses to investigate the presence of inbreeding depression, population genetic structure, gene flow, effective population size, and evolutionary history, all of which are critical forms of information for understanding the viability of putative ‘units to conserve’ within a seascape.

5 Chapter 1. General Introduction

Inferring demographic or genetic isolation With adequately designed surveys, photo-identification can be used to determine whether populations living in different geographic areas are demographically isolated or not (e.g. Gaspari et al. 2015; Brown et al. 2016), and this can be corroborated with the integration of genetic data (e.g., Andrews et al. 2010).

Chapter 1. General Introduction

Figure 1.1. Practical approaches (with considerations) and scientific information used to identify putative ‘units to conserve’ (subpopulations or

7 local populations) in a population of a small cetacean species.

Chapter 1. General Introduction

1.2. The Indo-Pacific bottlenose dolphin

The Indo-Pacific bottlenose dolphin Tursiops aduncus (Ehrenberg 1832) was recognised as a separate species from the common bottlenose dolphin Tursiops truncatus in the late 1990s, based on differences in genetics, osteology, and external morphology (LeDuc et al. 1999; Wang et al. 1999, 2000a, b). T. aduncus has a wide distribution in the warm temperate to tropical coastal and shallow waters of the Indian Ocean, Indo-Pacific region and western Pacific Oceans, including South Africa in the west, the Red Sea, Persian Gulf and Indo-Malay Archipelago along the rim of the Indian Ocean, to the southern half of Japan, and southward to Australia along both east and west coasts (Rice 1998; Hale et al. 2000; Möller and Beheregaray 2001; Reeves et al. 2003; Wang and Yang 2009; Gibbs et al. 2011; Allen et al. 2016). Along the north-western coastline of Western Australia, both T. truncatus and T. aduncus are documented, with the former found in the offshore, pelagic environment and the latter found in waters < 50 m deep and within approximately 10 km of the coastline (Allen et al. 2016).

T. aduncus is distributed across a wide range of habitats throughout temperate and tropical waters. In Australia, T. aduncus is restricted to inshore areas, such as bays and estuaries, nearshore waters, open coastal environments, and shallow offshore waters including coastal areas around reefs and oceanic islands (e.g., Hale 1997; Bilgmann et al. 2007; Fury and Harrison 2008; Wiszniewski et al. 2010; Chabanne et al. 2012; Sprogis et al. 2015; Allen et al. 2016). In south-eastern Australia, for example, T. aduncus shows a high degree of site fidelity to local areas and appears to reside in relatively small communities or populations (Möller et al. 2002). Despite the potential for long-distance movements within their broad distribution, significant genetic differentiation has been detected both within ocean basins and on a microgeographic scale within localised study sites (e.g., Ansmann et al. 2012; Kopps et al. 2014).

The nearshore distribution of T. aduncus makes them easily accessible, resulting in them being well-known and extensively studied cetacean species (Connor et al. 2000b). The species is not considered threatened in Australian waters, although T. aduncus is still classified as ‘Data deficient’ in the most recent International Union

8 Chapter 1. General Introduction for Conservation and Natural Resources (IUCN) species classification (Hammond et al. 2012). Populations of T. aduncus are exposed to a wide variety of threats that are technically challenging to quantify in terms of their precise impact, but which cumulatively could cause long-term population declines (Manlik et al. 2016). Some of the main threats likely to affect populations of T. aduncus include: habitat degradation and destruction (Ross 2006); pollution and coastal development (Steiner and Bossley 2008); entanglements and boat strikes (e.g., Steiner and Bossley 2008; Donaldson et al. 2010).

1.3. Indo-Pacific bottlenose dolphin population in Perth metropolitan waters

Perth is located in south-west Australia (S31.93, E115.86). The Perth metropolitan waters (as defined for this thesis) encompass approximately 300 km2 of coastal and estuarine ecosystems from southern Cockburn Sound (off Rockingham township) to the reef at Trigg Beach, including the Swan Canning Riverpark (Figure 1.2). The greater Perth metropolitan region continues to the north and south of the study area.

Several independent studies were conducted in the Perth metropolitan region in the last three decades, providing some knowledge of the ecology of T. aduncus associated with different areas and habitats (Waples 1997; Finn 2005; Moiler 2008; Chabanne et al. 2012):

Area 1 (Figure 1.2): Waples (1997) encountered more than 270 free-ranging bottlenose dolphins during boat-based surveys conducted from 1991 to 1993 and across a large area within the Perth metropolitan region (i.e., from the north of Cockburn Sound to the south of Lancelin, including Rottnest Island). Waples (1997) suggested that bottlenose dolphins were present year-round with evidence of different patterns of residency and associations (i.e., some repeated associations and many casual affiliates).

Area 2 (Figure 1.2): In Cockburn Sound, photo-identification studies of bottlenose dolphins were conducted during three periods: 1993-1997 (R. Donaldson, Murdoch University, unpublished data), 2000-2003 (Finn 2005) and a short investigation in 2009 (Ham 2009). Since there are few embayments along the western coastline of

9 Chapter 1. General Introduction

Western Australian, Cockburn Sound is the most intensively utilised coastal area (Environmental Protection Authority 1998). The proximity of the deep and protected waters of the Sound to the capital city of Perth led to intensive developments of the area for industry and maritime operations, including the Outer Harbour for the Port of Fremantle, the Kwinana Industrial Area, and the Australian Navy base at HMAS Stirling on Garden Island. In addition, the Sound supports several commercial and recreational fisheries and aquaculture operations, as well as a range of tourism and recreational activities (Cockburn Sound Management Council 2012). Donaldson (unpublished data) and Finn (2005) described a resident subpopulation of c. 75 bottlenose dolphins and the importance of the Kwinana Shelf as a key foraging habitat. Those studies involved some investigation of conservation issues such as illegal feeding (i.e., unregulated provisioning), entanglements, vessel disturbance, and habitat change related to industrial development and human-induced ecosystem modification affecting the population of bottlenose dolphins in this area (Finn 2005; Finn and Calver 2008; Finn et al. 2008; Donaldson et al. 2010, 2012a, b). The study by Ham (2009) suggested that the abundance of dolphins in the area was stable or at least that no precipitous decline had occurred since the prior investigations by Donaldson (1993-1997) and Finn (2000-2003).

Area 3 (Figure 1.2): In the Swan Canning Riverpark (SCR), a micro-tidal estuary, studies of T. aduncus began in late 2001 and continued with intense surveys up to 2003, followed by more sporadic efforts until the start of my PhD study in 2011. Based on the survey data covering the period from 2000 to 2003, Chabanne et al. (2012) indicated that the community within the SCR was small (c. 17-18 individuals) but characterised by a strong residency pattern based on their year-round presence and their preferential and long-term associations with other individuals within the SCR.

In 2009, six bottlenose dolphins were found dead within a six-month period in the SCR (Holyoake et al. 2010; Stephens et al. 2014). The unusual mortality event raised concerns about the welfare of dolphins inhabiting estuarine environments and resulted in an investigation supported by the Swan River Trust into the cause of the deaths. Post-mortem examination of four dolphins (the two other dolphins were too decomposed) indicated that a suite of factors likely contributed to the mortality of

10 Chapter 1. General Introduction each individual (i.e., skin lesions, active entanglements, secondary infections) (Holyoake et al. 2010). Stephens et al. (2014) also confirmed the presence of cetacean morbillivirus in two of the dead dolphins.

Figure 1.2. The three areas previously studied for bottlenose dolphins in the metropolitan waters of Perth (Western Australia) and the scale of the current study (red square, 2011-15). Area 1 – Open coastline (more effort in dark purple area than light purple area, Waples 1997); Area 2 – Cockburn Sound (green, Finn 2005); Area 3 – Swan Canning Riverpark (blue, Chabanne et al. 2012). RI = Rottnest Island; SCA = Scarborough.

11 Chapter 1. General Introduction

1.4. Objectives and structure of the thesis

In this thesis, I aim to provide the scientific basis for decision-making around ‘units to conserve’ for bottlenose dolphins in the Perth region. There are multiple imperatives for this research, including the concerns raised by the unusual mortality event in 2009 (Holyoake et al. 2010), rapid human population growth in the region (Australian Bureau of Statistics 2015), and proposed coastal developments, for example, the outer harbour development on the Kwinana Shelf (Western Australian Planning Commission 2004) and a desalination plant proposed for the northern metropolitan coast (Mercer 2013). The scientific basis I will provide includes information about population structure (Chapters 2 and 3, Chabanne et al. 2017a, b), encompassing genetic population structure at local- (Chapter 4) and regional-scales (Chapter 5), and information about the size and status of the subpopulations identified. I also illustrate how such information can be applied to assist in decision- making about the appropriate "local population" for the purposes of Environmental Impact Assessment (Chapter 3, Chabanne et al. 2017a).

Specifically, the objectives of my thesis are to: Chapter 1. Develop a systematic and robust photo-identification sampling design using multistate mark-recapture models (i.e., Multistate Closed Robust Design, MSCRD) to estimate the apparent survival, abundance and movement rates of bottlenose dolphins associated with defined geographic regions (Chapter 2, Chabanne et al. 2017b); Chapter 2. Identify local populations of bottlenose dolphins through the integration of social, ecological and genetic data collected during comprehensive and consistent sampling effort. Specifically, I assess the social network, examine the spatial segregation, and evaluate the genetic relatedness of the socio- geographic bottlenose dolphin communities in the study area. I then demonstrate the relevance of the fine-scale population structure for Environmental Impact Assessments (EIA) of coastal developments (Chapter 3, Chabanne et al. 2017a); Chapter 3. Examine whether the social structure of the bottlenose dolphin population is reflected through genetic differentiation (i.e., diversity and structure) using microsatellite loci and mitochondrial DNA markers (Chapter 4);

12 Chapter 1. General Introduction

Chapter 4. Investigate the patterns of gene flow between the population in Perth metropolitan waters and others at a regional-scale (southwestern coastline of Western Australia) using microsatellite loci markers (Chapter 5); and, Chapter 5. Summarise key findings and make recommendations to support the conservation and management of local dolphin populations in Perth metropolitan waters (Chapter 6).

This thesis has been written in the style of a ‘thesis by publication’, following the Murdoch University style guideline for thesis by publication/manuscripts. This chapter (Chapter 1) provides a general introduction and thesis overview. Data chapters (Chapters 2 to 5) have been prepared as stand-alone papers and are presented here with minimal changes from the versions submitted or published, although references are collated across chapters. Finally, in Chapter 6, I synthesise the main findings relevant to stakeholders and government agencies on the current status of bottlenose dolphins in Perth metropolitan waters. I also discuss some of the theoretical and practical limitations of the approaches used in this thesis, and conclude with recommendations for future management strategies.

1.5. Ethics statement

This study was carried out with approval from the Murdoch University Animal Ethics Committee (W2342/10 and R2649/14) and was licensed by the Department of Parks and Wildlife (SF008067, SF008682, SF009286 and SF009874). Biopsy sampling for molecular analyses were carried out as a part of broader study, with data collected in accordance with the Murdoch University Animal Ethics Committee approval (W2076/07; W2307/10; W2342/10 and R2649/14), and collected under research permits (SF005997; SF006538; SF007046; SF007596; SF008480; SF009119: SF009734; SF010223) from the Department of Parks and Wildlife.

Chapter 2. Applying the multistate capture-recapture robust design to assess metapopulation structure

2.1. Abstract

1. Population structure must be considered when developing mark–recapture (MR) study designs as the sampling of individuals from multiple populations (or subpopulations) may increase heterogeneity in individual capture probability. Conversely, the use of an appropriate MR study design which accommodates heterogeneity associated with capture occasion varying covariates due to animals moving between ‘states’ (i.e., geographic sites) can provide insight into how animals are distributed in a particular environment and the status and connectivity of subpopulations. 2. The multistate closed robust design (MSCRD) was chosen to investigate: (i) the demographic parameters of Indo-Pacific bottlenose dolphin (Tursiops aduncus) subpopulations in coastal and estuarine waters of Perth, Western Australia; and (ii) how they are related to each other in a metapopulation. Using four years of year- round photo-identification surveys across three geographic sites, I accounted for heterogeneity of capture probability based on how individuals distributed themselves across geographic sites and characterised the status of subpopulations based on their abundance, survival and interconnection. 3. MSCRD models highlighted high heterogeneity in capture probabilities and demographic parameters between sites. High capture probabilities, high survival and constant abundances described a subpopulation with high fidelity in an estuary. In contrast, low captures, permanent and temporary emigration and fluctuating abundances suggested transient use and low fidelity in an open coastline site. 4. Estimates of transition probabilities also varied between sites, with estuarine dolphins visiting sheltered coastal embayments more regularly than coastal dolphins visited the estuary, highlighting some dynamics within the metapopulation. 5. Synthesis and applications. To date, bottlenose dolphin studies using mark- recapture approach have focussed on investigating single subpopulations. Here, in a heterogeneous coastal-estuarine environment, I demonstrated that spatially structured bottlenose dolphins subpopulations contained distinct suites of individuals and

15 Chapter 2 - Population structure investigated by MSCRD differed in size, demographics and connectivity. Such insights into the dynamics of a metapopulation can assist in local-scale species conservation. The MSCRD approach is applicable to species/populations consisting of recognizable individuals and is particularly useful for characterising wildlife subpopulations that vary in their vulnerability to human activities, climate change or invasive species.

16 Chapter 2 - Population structure investigated by MSCRD

2.2. Introduction

At an individual level, wildlife tends to be neither uniformly nor randomly distributed across land- or sea-scapes but to occur in association with particular environmental features (Legendre and Fortin 1989). At a population-level, species are typically distributed in a series of populations or ‘subpopulations’, as in a metapopulation model (i.e., set of spatially separated populations of the same species which interact at some level, Levins 1969). Emigration and immigration between subpopulations may occur through either permanent additions or subtractions or only the short-term presence or absence of individuals (Brown et al. 2016; Sprogis et al. 2016). Individuals within a population (or subpopulation) may have ranging patterns that overlap or are connected with a particular locality (Sprogis et al. 2015).

Such population structure must be considered when developing mark-recapture (MR) study designs because the sampling of individuals from multiple populations (or subpopulations) may increase heterogeneity in individual capture probability (Brown et al. 2016). Conversely, it is feasible for an appropriate MR study design also to provide insight into how animals are distributed in a particular environment and the status and connectivity of any subpopulations that are present (Brooks and Pollock 2014).

Since its development in the late 1800s (Petersen 1895), the MR approach has been widely used for assessing wildlife abundance, distribution and demographic processes. Here, I attempted to use extensions of a MR study design, the multistate closed robust design (MSCRD), to investigate demographic parameters and connectivity between putative subpopulations that were spatially predefined in a heterogeneous coastal and estuarine environment.

In MR studies, individual-specific encounter (‘capture’) histories may be used to generate capture probabilities, and to estimate apparent survival rates (i.e., the true survival and permanent emigration combined) and abundance (i.e., number of animals in the study area, Lettink and Armstrong 2003). The underlying assumption of homogeneity in individual capture probabilities is often violated because of practical constraints on sampling (see review by Lindberg 2012). Heterogeneity in

17 Chapter 2 - Population structure investigated by MSCRD individual capture probability may be reduced by the inclusion of time-dependent covariates (e.g., year), individual time-constant covariates (e.g., sex), and covariates associated with individual capture occasion (e.g., weight, social affiliations, geographical locations: Pollock et al. 1990).

The closed robust design (CRD) was built using two different temporal scales: (i) two or more open sampling occasions (hereafter ‘primary periods’) in which the time interval between periods is sufficiently long enough to allow for births and immigration, and for losses from deaths and emigration; and, (ii) closed sampling occasions (hereafter ‘secondary occasions’) set within each of the primary periods and where the intervals between occasions are sufficiently short so that no gains and losses are assumed to occur (Pollock 1982). By sampling across multiple temporal scales, CRD models estimate temporary emigration (TE) and immigration between primary periods as well as abundance and apparent survival parameters without having to assume equal probability of capture over the entire study period (Kendall and Pollock 1992; Smith et al. 2013). Thus, biases due to heterogeneity in capture probability are minimised and abundance and apparent survival are estimated from multiple occasions allowing better precision (Kendall 1990). The incorporation of time-constant covariates (e.g., sex) within the CRD models also has advantages in reducing the heterogeneity in capture probability and estimating abundance and apparent survival specific to covariate classes (e.g., males, females).

Another MR study design, the multistate mark-recapture (MS) approach, enables the use of fixed set of categorical ‘states’ that are discrete covariates measured upon capture of the individual, e.g., geographic location, reproductive state (e.g., Hestbeck et al. 1991; Cam et al. 2004). Like time-constant covariates, an advantage of including categorical ‘states’ in MS models is a homogeneity assumption that is state-specific and the ability of the models to provide state-specific estimates for abundance and apparent survival (Lindberg 2012). As well as modelling immigration and emigration to and from an unobservable state (i.e., outside the study area, and thus part of the apparent survival estimates), MS models have a unique feature in which transition among ‘states’ can be estimated (Darroch 1961; Arnason 1972, 1973). The transition between states is the probability that an individual, alive and in the state x, just before t+1, emigrates into the state y. Transition between states may

18 Chapter 2 - Population structure investigated by MSCRD be either temporary or permanent and both contribute to the estimate of transition probability.

Here, I applied the MSCRD approach with ‘states’ referring to geographic sites (see ‘site’ hereafter), which utilizes aspects of MS models and the CRD (Nichols and Coffman 1999) (Figure 2.1) for several reasons. Firstly, the MSCRD allows for greater flexibility in model specifications for individual heterogeneity in capture probability. Critically, heterogeneity can be modelled according to: (i) individual- level characteristics (i.e., a time-constant covariate such as sex), (ii) individual-level responses to capture (i.e., state measured upon capture) or (iii) the relevant temporal scale for captures (primary periods vs. secondary occasions). Secondly, the MSCRD can provide abundance estimates for each ‘state’ within each primary period. Finally, the inclusion of multiple secondary occasions within each primary period increases the capture probability, which improves the precision of the apparent survival estimates and transition probabilities (White et al. 2006; Lindberg 2012).

19 Chapter 2 - Population structure investigated by MSCRD

Figure 2.1. Traditional closed robust design (CRD) vs. multistate closed robust design (MSCRD) approaches to characterise metapopulation structure and dynamics through demographic parameters. Both approaches allow estimation of abundance (N), apparent survival rate (φ) and emigration and immigration [solid arrows] either time varying (t, t+1, t+2, etc.) or constant. In addition, MSCRD models estimate any transition probabilities ψ [dashed arrows] between subpopulations associated with states (i.e., geographic sites).

20 Chapter 2 - Population structure investigated by MSCRD

My aim is to show how the MSCRD, with its innate flexibility in modelling heterogeneity in capture probabilities, can simultaneously provide demographic parameter estimates for multiple putative subpopulations associated with particular sites as well as describe their conservation status and connectivity to other subpopulations. This approach allows for use of: (i) capture probabilities to affirm (or refute) the putative grouping of individuals associated with a particular site as a distinct ‘subpopulation’ (i.e., homogeneity within sites); (ii) estimates of the variation in abundance (i.e., primary period changes in the number of individuals in any geographic site) and apparent survival (i.e., the probability of surviving and staying in any site) to assess the occupancy (or residency) of a group of individuals in that site; and (iii) transition probabilities between sites to describe the interconnectivity of those groupings.

Previous MSCRD studies using site as a ‘state’ have generally had other aims and applications: e.g., detecting changes in transition probabilities before, during and after an environmental perturbation affecting one state (see O'Connell-Goode et al. 2014) or human development activities (see Brooks and Pollock 2014), evaluating individual fitness over time (see Gibson et al. 2014) or quantifying the connectivity (i.e., transition of individuals) between areas exposed to different management regimes (see Lee 2015). Notably, this study aimed to examine the dynamics, status and connectivity of multiple putative subpopulations each associated with a particular site.

To pursue the above aim, I applied the MSCRD approach in a mark-recapture study of Indo-Pacific bottlenose dolphins Tursiops aduncus (‘dolphin’ hereafter) in coastal and estuarine waters near Perth, Western Australia. I defined the geographic sites based on the coastal geography and landforms and the known presence of small resident subpopulations in an estuary (N ≈ 20, Chabanne et al. 2012) and a nearby coastal embayment (N ≈ 75, Finn 2005) (but without knowledge of their connectivity to each other, or to other potential subpopulations in the study area).

I then used individual capture histories obtained from four years of year-round boat- based photo-identification surveys to estimate: (i) capture probabilities per site to evaluate and compare the occupancy pattern of dolphins within sites (i.e., to explore

21 Chapter 2 - Population structure investigated by MSCRD heterogeneity between sites); (ii) apparent survival rates and abundances as to verify putative site-related groupings through the assessment of site fidelity (i.e., close to true survival, stable abundances) and (iii) transition probabilities so as to characterise movement between site-related groupings (i.e., the metapopulation dynamic).

2.3. Materials and methods

2.3.1. Field methods

2.3.1.1. Study area and field sampling design

My study area encompassed an area of 275 km2, extending for 45 km along the coast of Perth and then inland to include the Swan Canning Riverpark (SCR), an estuarine reserve of about 55 km2 (Figure 2.2). Three sites were defined based on coastal geography, principal landforms (estuary, open waters and coastal embayment) and information from previous local studies (Waples 1997; Finn 2005; Chabanne et al. 2012): (i) the estuary (SCR) and two sites in coastal waters, (ii) Gage Roads (GR), a length of open coastline with mostly sandy beaches and small areas of rocky reef and seagrass, and (iii) Cockburn Sound/Owen Anchorage (CS/OA), a semi-enclosed embayment. The northern section of the embayment (OA) is of < 10 m depth, except in a shipping channel (max depth: 14.7 m), with substrates mainly of shell-sand and seagrass. The southern section (CS) has shallow (< 10 m) margins, a deep (c. 20 m) central basin, and seagrass, sand, silt and limestone substrates. In comparison to GR, CS/OA experiences intensive industrial and recreational use, with threats to dolphins including entanglement and illegal feeding (Finn 2005; Donaldson et al. 2010), industrial and harbour development (Finn 2005) and shell-sand dredging (BMT Oceania 2014). For practical reasons (i.e., wind and sea conditions), CS and OA were split and run as two separate sub-sites, although there were jointly sampled in 84% of the secondary occasions (see below).

Between June 2011 and May 2015, I collected year-round mark-recapture data for individual dolphins using boat-based photo-identification surveys following predefined transect routes (Figure 2.2). While the same transect route was conducted in the estuary (due to the confined waters), I rotated between three pre-defined zig-zag

22 Chapter 2 - Population structure investigated by MSCRD transect routes (off-set by 2 km) in the coastal sites to increase sampling coverage (Figure 2.2). Transect routes were designed using Distance 6.0 (Thomas et al. 2009).

In the robust design language, our primary periods corresponded to the four seasons in the Australasian calendar: winter (June to August); spring (September to November); summer (December to February); and autumn (March to May). For this study, I aimed to conduct at least five secondary occasions (i.e., consecutive surveys of the three sites) per primary period (n = 16); however, this was not successful for four primary periods because of weather conditions (Appendix A2.1). If a survey was interrupted because of weather conditions or logistical issues, the survey was cancelled and entirely re-run. Surveys of each site were conducted in random order and at different times of the day.

To limit violation of the closure assumption of a robust design (Pollock 1982; Nichols and Kendall 1995), I aimed to complete a secondary occasion in the shortest possible time (i.e., on consecutive days, mean = 2.60; min = 2; max = 8 days, Appendix A2.1) so as to minimise transitions of the animals (Pollock 1982). When multiple captures occurred for an individual in a secondary occasion, I retained only the first capture for that secondary occasion. I then waited for at least one week (unless weather conditions were excellent and/or I was approaching the end of the season – primary period) before starting another secondary occasion. The break between two secondary occasions was longer than the time needed to successfully complete a secondary occasion (mean = 8.63; min = 0, max = 60 days, Appendix A2.1), thus allowing us to assume independence between secondary occasions. I also left a longer interval between two adjoining primary periods (mean = 47.30; min = 12; max = 80 days, Appendix A2.1) to minimise violation of the assumption between closed and open sampling occasions (Kendall 2004; Brown et al. 2016). The assumption of closure within primary periods was tested with the program CloseTest (Stanley and Burnham 1999, see Appendix A2.2 for explanations).

23 Chapter 2 - Population structure investigated by MSCRD

Figure 2.2. Map of the metropolitan waters of Perth, Western Australia, showing the systematic survey routes within each site: the estuary SCR = Swan Canning Riverpark and the coastal sites (south to north) CS/OA = Cockburn Sound/Owen Anchorage, and GR = Gage Roads. Within the coastal sites, surveys were conducted by rotating between three pre-defined transect routes (full, long dash and short dash lines) to maximise the coverage.

2.3.1.2. Data collection and data processing

To minimise heterogeneity of individual capture probabilities, the vessel was driven at a constant speed (8-12 knots) with at least three observers on-board to maximise the area coverage. However, 3% (10 of 304 surveys) of the surveys were conducted with two observers only. Surveys were conducted in Beaufort sea state ≤ 3. When a dolphin group was encountered along a transect route, I paused the search effort and photographed the dorsal fin of each individual on both sides (if possible) and without

24 Chapter 2 - Population structure investigated by MSCRD regard to the distinctiveness of fins. Photographic effort was conducted by the same person (DBHC) throughout the entire study period. In this study, dolphins were assigned to the same group when seen within approximately 100 m from the boat (Wells et al. 1987; Quintana-Rizzo and Wells 2001) and performing similar activities. Once all dolphins were photographed, the search effort was resumed from where I had departed the transect route. Photographs of each dolphin group were then graded for quality by one to three trained assistants and checked by DBHC for the entire study period. Measures of the quality and individual distinctiveness were done using modified methods developed by Urian et al. (1999, see Appendix A2.3). Each individual was assigned a grade for distinctiveness of their dorsal fins to minimise misidentification and heterogeneity in capture probabilities (Nicholson et al. 2012).

To minimise heterogeneity in captures due to misidentification of non-distinctive fins (D3), only individuals with distinctive fins (D1 and D2, Appendix A2.3) were used in the MSCRD models. Abundance estimates were then adjusted to take into account the proportion of individuals in the population that were unmarked (D3) following the method described in Nicholson et al. (2012) (see Appendix A2.4 for calculation of the proportion of distinctly marked individuals). I also attempted to address individual heterogeneity in capture probabilities by including sex as an individual covariate (i.e., female, male or unknown). However, given that 50% of the individuals were not sexed (n = 169), I acknowledge that estimates obtained through MSCRD models may be overestimated for sexed individuals and underestimated for not sexed individuals (Nichols et al. 2004) and for that reason are not presented here (but see Appendix A2.11 for MSCRD analyses including sex as individual covariate). Calves, typically less than four years of age (Mann and Smuts 1998), were excluded from the analysis because of their dependence on their mothers (i.e., captures must be independent, Pollock et al. 1990). Heterogeneity in capture probabilities was tested by implementing goodness-of-fit tests for multistate models using the program U-CARE (see Appendix A2.5, Pradel et al. 2005; Choquet et al. 2009).

25 Chapter 2 - Population structure investigated by MSCRD

2.3.2. Statistical methods

The multistate closed robust design models were run in MARK (White and Burnham 1999) and estimated four parameters per site: (i) abundance (N), which is the number of individuals present in the study area; (ii) apparent survival rate (φ), which is the probability of surviving and staying in a sample site; (iii) transition probability (ψ), which represents the probability of moving from one site to another; and, (iv) capture probability (p). Although transitions from and to the study area may have occurred (i.e., TE to an unobservable site), models with an unobservable site never reached convergence, and thus are not presented. The modelling approach assumes that no site transitions occurred within a primary period (Arnason 1972, 1973). However, I acknowledge that 2.6% of the captures violated this assumption. Two adjustments were made to minimise this violation. First, if an individual was captured in two different sites within a primary period, I retained captures matching the site of the first capture recorded in that primary period. Results were similar if the last capture was retained instead and therefore are not presented here. Second, I ran the MSCRD models for two different scenarios, including one that involved pooling sites so that transitions between sites were minimised. Scenario 1 represented the three sites as originally described in this study area and Scenario 2 had all of the coastal sites (CS/OA and GR) pooled together into a single Coastal site for comparison with the estuary (SCR).

In MARK, each MSCRD model combination was run with the probability of capture (p) varying by site and/or primary period or constant, and with recapture probability (c) set as equal to first capture probability (p). The abundance (N) was set to vary by site and primary periods [N(site × primary periods)]. Several sub-models for apparent survival (φ) were run (i.e., whether it varied by site and/or primary period or constant). Transition probability between sites (ψ) was also estimated, whether that parameter varied by site and/or primary period or if it did not vary. In MARK, time intervals between primary periods were specified as a fraction of a year (i.e., 0.25) to estimate annual apparent survival and annual transition rates when modelled as time-constant.

26 Chapter 2 - Population structure investigated by MSCRD

Models were ranked using the Akaike information criterion (AICc, Burnham and

Anderson 2002). The model with most support by AICc (highest AICc weight) was selected as the most parsimonious model. Models with ΔAICc < 2 were also considered to have support from the data (Burnham and Anderson 2002).

2.4. Results

2.4.1. Effort

Seventy-six secondary occasions (167 days of boat-based surveys) were completed between June 2011 and May 2015 (see Appendix A2.1). In total, 410 dolphin groups were encountered, ranging in size from one to 32 dolphins (mean = 5.7, SE 0.3; excluding calves, see Appendix A2.6). I individually identified 346 dolphins, of which seven were well marked but were identified from poor-quality photographs, and were therefore excluded from further analyses. Among the 339 individuals, 134 individuals were excluded from the mark-recapture analyses because of insufficiently marked dorsal fins (see Appendix A2.7). The overall proportion of distinctly marked individuals was 0.78 (SE 0.02) and varied from 0.69 (SE 0.06) for individuals captured in GR to 0.80 (SE 0.02) for individuals captured in SCR.

2.4.2. Model selection

Results from the program CloseTest indicated that the population was closed over 13 of the 16 primary periods, indicating that the assumption of population closure was satisfied on > 81% of cases, with no significant gains or losses (Appendix A2.2). Goodness-of fit (GOF) test results, based on multistate and subcomponent tests in U- CARE, suggested an overall heterogeneity in capture probability (χ2 = 216.551, d.f. = 145, P-value < 0.01, see Appendix A2.5 for summary of GOF tests). The estimate of the variation inflation factor ĉ was < 1 (ĉ = 0.75), suggesting no substantial overdispersion, which meant there was no need for Quasi-likelihood (QAICc) adjustments to define the most parsimonious model (Cooch and White 2005). For

Scenario 1 (three sites), the best-fitting model, based on the AICc weight, was that capture probability varied by site and primary period [p(site × primary period)], apparent survival rate varied by primary period but not site [φ(primary period)], and

27 Chapter 2 - Population structure investigated by MSCRD transitions varied between each site [ψ(site)] (see Appendix A2.8). For Scenario 2 (two sites), the best-fitting model was that capture probability varied by site and primary period [p(site × primary period)], apparent survival varied by site [φ(site)] and transitions varied between site [ψ(site)] (see Appendix A2.8). Due to having small numbers of animals, I did not allow for capture probabilities to vary among secondary occasions. Individual heterogeneity in capture probability was therefore not modelled (which can be accommodated in conventional RD analyses, given sufficient data) despite this frequently being found in photo-identification studies of cetacean populations.

2.4.2.1. Capture probabilities

Capture probabilities varied by site and primary period (Figure 2.3). Regardless of the scenario, the SCR had high capture probability (mean, ̂ = 0.30, min = 0.11, max = 0.52, SE 0.03). In contrast, capture probability in GR was low (mean, ̂ = 0.06, min = 0.00, max = 0.12, SE 0.01; Figure 2.3). Sighting frequencies showed that individuals with higher sighting frequency were seen in SCR (>17 sightings), whereas 50% of individuals observed in GR were seen only once (see Appendix A2.9). Probability of captures for CS/OA were moderate (mean, ̂ = 0.15, min = 0.08, max = 0.27, SE 0.01; Figure 2.3). Coefficient of variation (CV) of the estimated capture probability varied by site (Appendix A2.10) with GR having the highest CV (CVmedian = 30%), thus suggesting high heterogeneity in capture probability in comparison to CS/OA for which the CV was lower (CVmedian = 15%).

28 Chapter 2 - Population structure investigated by MSCRD

Figure 2.3. Capture probability (p) yielded by the models for each secondary occasion represented as box plot (min; Quartile 1; median; Quartile 3; max) for each Scenario: 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads); and 2 – two sites (SCR and Coastal).

2.4.2.2. Apparent survival estimates and abundances

Models yielded apparent survival rates ( ) ranging from 0.93 (SE 0.03) to 1 (SE

0.00), although was higher in SCR ( ̂ = 0.98, SE 0.04) than in the pooled

Coastal site ( ̂ = 0.83, SE 0.02) in the Scenario 2.

Total estimated abundances in the SCR were low but stable over the study period

(N̂ = 16, min 10, max 23) (Figure 2.4). Also, individuals were frequently resighted in the SCR (see Appendix A2.9).

No obvious seasonal variation in abundance estimates was detected in the CS/OA site (N̂ = 103, min 71, max 147, Figure 2.4). Abundance estimates in GR varied with the highest in winter 2011 (N̂ = 172, 95% CI 53-561) and autumn

2015 (N̂ = 172, 95% CI 78-381; Figure 2.4). No dolphins were ‘captured’ in GR in summer 2012 and winter 2014.

29 Chapter 2 - Population structure investigated by MSCRD

Figure 2.4. Seasonal estimated abundances (N̂total ± 95% confidence intervals) for (a) Scenario 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and (b) Scenario 2 – two sites (SCR and Coastal). Lines between data points have been used for illustrative purposes only; continuity of values is not implied. Sites are as follows: SCR (red), CS/OA (yellow), GR (purple) and Coastal in Scenario 2 (grey).

30 Chapter 2 - Population structure investigated by MSCRD

2.4.2.3. Transitions

The estimates of the transition probabilities ( ̂) yielded by the model in Scenario 1 suggested that there was very little or no transition between the SCR and GR sites

( ̂ < 0.010) (Table 2.1). The model yielded a higher transition probability from the SCR to CS/OA ( ̂ = 0.151, SE 0.028) than in the opposite direction ( ̂ = 0.028, SE 0.005). Estimates from Scenario 2 also indicated similar transition probabilities with higher transition from the SCR to Coastal sites

̂ = 0.158, SE 0.029) and low rate in the opposite direction ̂ 0.017, SE 0.003).

Chapter 2 - Population structure and dynamic assessed by MSCRD

Transition Into:

From: SCR CS/OA GR Coastal

(a) Scenario 1

SCR 0.840 0.151 (0.028) 0.009 (0.008) - CS/OA 0.028 (0.005) 0.921 0.051 (0.009) - GR 0.000 (0.002)* 0.084 (0.014) 0.916 -

(b) Scenario 2 SCR 0.842 - - 0.158 (0.029) Coastal 0.017 (0.003) - - 0.983

Note: Values in italic represent rates when staying in the same site. * Values estimated were smaller than 0.001.

Chapter 2 - Population structure investigated by MSCRD

2.5. Discussion

Three broad results emerged from the use of a MSCRD with geographic sites as ‘states’ in a complex coastal environment with estuarine, embayment, and open coastline components and with a species known to exhibit fine-scale population structure in such systems. First, the heterogeneity of capture probabilities between sites showed a clear spatial component, consistent with some degree of population structuring. Second, estimates of abundance and apparent survival rate allowed some inference about the status of each site-related grouping (or ‘subpopulation’, in the metapopulation model). Finally, estimates of transition probability between sites indicated some degree of connectivity between those site-related groupings.

2.5.1. Spatial heterogeneity

Differences in capture probability appear to reflect individual variation in the use of (and fidelity to) a site. Heterogeneity in capture probability has also previously been linked to variation in individual or group ranging patterns (Crespin et al. 2008; Urian et al. 2014). Here, the capture probability was high in the estuary (SCR) and low in the open coastline (GR), suggesting that the ranging patterns of dolphins using those sites differ markedly in, e.g., home range size, site fidelity, seasonal or year-round occupancy and habitat use (Sprogis et al. 2015). The capture probabilities for the estuary are consistent with the long-term site fidelity and year-round occupancy reported in Chabanne et al. (2012). In contrast, Waples (1997) suggested that dolphins in the open coastline north of Perth likely range over many kilometres of coastline and are only intermittently present in particular areas, again consistent with the low capture probabilities observed here.

In addition, coefficients of variation for GR were high, suggesting more individual heterogeneity in capture probability due to factors such as large and variable home range sizes or avoidance or attraction responses to boats (Pollock et al. 1990). I acknowledge that the estimates of demographic parameters for GR could be biased with lower estimates of apparent survival leading to underestimated abundances (Pollock et al. 1990; Williams et al. 2002). This outcome indicates the practical difficulties for MSCRD approaches if the ranging patterns (or other characteristics)

33 Chapter 2 - Population structure investigated by MSCRD of the individuals present at a site are such that CVs will be high, even where sampling is relatively intensive and is sustained over multiple years.

In contrast, the low coefficient of variation in the capture probabilities for CS/OA (CV = 15%) suggested that majority of the individuals were equally captured. Furthermore, despite lower capture probabilities than those estimated in SCR, dolphins nonetheless occurred year-round in CS/OA. Differences in capture probabilities between CS/OA and SCR may reflect larger home ranges for individuals in the embayment system (Sprogis et al. 2016). Site configuration may also influence individual detection with a greater likelihood of detecting individuals in narrow areas such as channels or rivers than in wide, unconfined areas such as open water that have with no prominent barriers.

2.5.2. Apparent survival and abundance estimates

Given that the majority of the individuals were not sexed, I acknowledge that estimates of the apparent survival rates obtained through the best-fitting models may be overestimated for sexed individuals and underestimated for unsexed individuals. Also, most of the sexed individuals were those regularly seen during the study period because collection of genetic samples (for which sex determination was one objective) was preferentially undertaken on well-known individuals. While Nichols et al. (2004) demonstrated how to deal with unsexed individuals in capture-recapture analytical approaches, that method could not be applied in this study due to the complexity of the models. Apparent survival of dolphins in SCR was high (0.98), illustrating an almost complete lack of permanent emigration during the study. In addition, consistent abundance estimates across the course of the study (c. 16 dolphins), along with high individual resighting rates, indicated the long-term residency of the SCR subpopulation.

In contrast, an apparent survival estimate of 0.83 in the pooled Coastal site is indicative of permanent emigration of individuals, suggesting both resident and more transient components (Brown et al. 2014; Palmer et al. 2015). Variation in abundance estimates in conjunction with apparent survival rates can assist in making inferences about residency status (Brown et al. 2016). The high variability in

34 Chapter 2 - Population structure investigated by MSCRD abundance estimates in GR coupled with the large number of individuals sighted only once (and not seen in SCR or CS/OA) is consistent with transient occupancy patterns for dolphins at that site. This agrees with other studies indicating that bottlenose dolphins are often more abundant in open coast environments where individuals tend to have larger home range size, which may reflect both food availability and foraging tactics (Sprogis et al. 2015; McCluskey et al. 2016).

2.5.3. Transitions

There was a substantial difference in transition probabilities between the SCR and the coastal sites. The reasons for this are not clear. As the mouth of the SCR estuary is located at the junction between the CS/OA and GR sites, travel distance between sites should not be a factor. Furthermore, transitions from SCR were also limited to CS/OA. One possibility is that the environment of OA (shallow, protected waters with extensive seagrass meadows) may be more suitable habitat for SCR individuals than the GR environment.

Conversely, transitions from CS/OA to SCR were limited, although occasional visitors were documented in the lower reaches of the estuary or further up river, sometimes escorted by SCR males (Connor et al. 1996, 2000a). Those transitions in and out from SCR were consistent with the emigration and reimmigration demographic model reported in Chabanne et al. (2012). The long-term connection between SCR and CS/OA suggests a certain degree of gene flow between the subpopulations.

2.5.4. Limitations

The transitions of animals in and out of the study area (i.e., sometimes referred to as the “edge effect”, Otis et al. 1978) present two significant problems for MSCRD studies. The first is that these transitions increase the heterogeneity of captures in the study area at large (Crespin et al. 2008; Brown et al. 2016). The second is that sites within the study area may differ in the degree to which such edge effects occur. In this study, for example, I found more heterogeneity associated with edge effects in the open coastline (GR) than in the estuary (SCR). When considering predefined

35 Chapter 2 - Population structure investigated by MSCRD sites, it is advisable to consider what proportion of the individuals captured in that site may also be captured in other sites and whether individual heterogeneity in capture probability may differ between sites. This is particularly relevant if sites differ greatly in size or in other features that may limit detectability and the precision of estimation (Burgess et al. 2014; Palmer et al. 2015). Here, I benefitted from existing information on the likely ranging patterns of individuals, but was nonetheless unable to implement a study design that negated the heterogeneity of captures arising from the transitions of individuals into and out of the study area.

The assumption that no transition between sites occurs within a primary period (Arnason 1972, 1973) is difficult to validate, particularly when sites are juxtaposed (i.e., no physical barrier and distance exists). Such violations may result in greater heterogeneity in capture probability between individuals and within sites (i.e., individuals captured within a site do not all have the same survival rate). Here, 2.6% of the captures violated this assumption. The extent to which this assumption can be acceptably violated is unclear, although it has been reported that < 1% of violated occasions would create a small bias (O'Connell-Goode et al. 2014). This issue was dealt with in this study by pooling the coastal sites (Scenario 2) (Schwarz 2002), while also ensuring that all sites were sampled equally (Crespin et al. 2008). However, this procedure may lead to more heterogeneity in capture probabilities and bias survival rates and abundance estimates (Pollock et al. 1990). Here, the survival rate for dolphins in CS/OA was higher than for GR, which was also supported by a consistency in abundance estimates and moderate resighting rates in CS/OA with few individuals being seen only once.

Low capture probabilities make it difficult to obtain reliable estimates of apparent survival rate and abundance (Pollock et al. 1990; Rosenberg et al. 1995). It is advisable that capture probabilities of at least 0.10 per secondary occasions be obtained for reasonable results (Lettink and Armstrong 2003). Although common in studies of wide-ranging and low density species (Harmsen et al. 2010; Palmer et al. 2015), there are few obvious measures for dealing with low capture probabilities other than increasing sampling effort (Pollock et al. 1990; Rosenberg et al. 1995). However, increases in sampling effort involve additional cost (Tyne et al. 2016) and

36 Chapter 2 - Population structure investigated by MSCRD time outlays that must be multiplied by the number of sites (to ensure that each site is surveyed equally).

Finally, I note that the sophistication and utility of MSCRD models continue to evolve, notably in relation to the modelling of TE, which can alleviate some of the large differences in survival estimates (Bailey, Converse & Kendall 2010). I was unable to model TE in this study as the models would not estimate the applicable parameters, due to the small population sizes. Rankin et al. (2016) also discussed issues linked to low capture probabilities and the estimation of TE and suggested use of hierarchical Bayesian models for the Robust design.

2.5.5. Conclusion

This study, which explored the implications of heterogeneity in capture probabilities for MR studies within a MSCRD framework, demonstrated a valuable approach for assessing the dynamic, status and connectivity of multiple subpopulations of a behaviourally plastic species within a heterogeneous environment. I also showed that a MSCRD study design can assess transitions between predefined geographic sites, and thus assist in understanding dynamic processes between subpopulations within a metapopulation. A MSCRD incorporating geographic sites associated with anthropogenic impacts or climate change may be a powerful tool for management and conservation of species that are amenable to a MR study. The short-term transition of individuals between putative subpopulations is particularly relevant for the conservation of highly mobile species (e.g., birds, larger mammals) in environments where anthropogenic pressures vary greatly from one geographic ‘site’ to another.

37 Chapter 2 - Population structure investigated by MSCRD

Appendices

Appendix A2.1. Survey effort

Table A2.1.1. Summary of survey effort and time interval between efforts across primary periods (i.e., seasons) for four years (June 2011-May 2015). A secondary occasion refers to a combination of the four surveys covering the entire study area. “in secondary occasion” refers to the number of days required to complete a secondary occasion, while “out secondary occasion” refers to the number of days between the last survey of the secondary occasion i and the first survey of the secondary occasion i + 1.

Effort Interval time # days in # days out # # days Time Distance secondary secondary Year Seasons secondary between (hrs) occasion occasion (km) occasion seasons (SE) (SE)

Winter 53.1 NA 6 2.17 0.17 11.20 9.01 2011 Spring 42.6 663 5 3.20 1.20 14.50 2.1 12 Summer 43.9 672 5 2.20 0.20 13.25 3.33 27 Autumn 40.8 662 5 2.00 0.00 17.50 14.24 16 2012 Winter 40.5 670 5 2.20 0.20 4.25 2.66 32 Spring 43.2 669 5 3.00 0.45 7.25 1.55 80 Summer 43.9 675 5 3.00 0.55 3.75 1.75 58 Autumn 39.2 663 5 2.00 0.00 4.25 1.31 45 2013 Winter 42.1 678 5 2.00 0.00 9.50 4.77 59 Spring 31.3 534 4 2.00 0.00 7.33 2.19 41 Summer 32.8 531 4 3.75 1.03 7.67 4.26 77 Autumn 36.0 547 4 2.00 0.00 8.67 4.18 38 2014 Winter 45.3 676 5 2.00 0.00 9.00 2.45 63 Spring 26.3 417 3 4.67 0.88 8.50 5.5 57 Summer 44.5 683 5 2.40 0.24 6.75 2.43 52 2015 Autumn 45.3 686 5 3.00 0.77 3.50 0.65 53 TOTAL 650.8 c. 10,000.00 76 2.60 0.20 8.63 1.00 47.33 5.16

38 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.2. The CloseTest: a method to investigate of the population closure

I investigated the assumption of population closure for each primary period separately with all sites converted to a single site using the Stanley and Burnham (1999) and Otis et al. (1978) closure tests, as implemented in the computer program CloseTest (Stanley and Burnham 1999). The two tests were developed on different null hypotheses and, if used in conjunction, allow for better detection and interpretation of closure violations in capture-recapture datasets. The Stanley and Burnham (1999) closure test allows for time variation in capture probabilities in the absence of behavioural responses and heterogeneity. The Otis et al. (1978) closure test allows for investigation of population closure in the presence of heterogeneity in capture probabilities.

Table A2.2.1. Closure tests for each primary period using Stanley and Burnham (1999) and Otis et al. (1978) implemented in the program CloseTest. Null hypothesis (no closure) was rejected when P-value < 0.05 at both tests.

Primary # secondary Otis et al. test Stanley and Burnham

period occasions z-value P-value χ2 test P-value

1 6 1.31 0.9 19.3 0.01

2 5 -3.3 0 10.37 0.11

3 5 insufficient data 4.24 0.12

4 5 -1.46 0.01 15.6 0.01

5 5 -0.7 0.24 13.33 0.02

6 5 1.02 0.85 9.63 0.14

7 5 1.45 0.92 38.59 0.00

8 5 -2.26 0.01 24.83 0.00

9 5 -4.01 0 15.27 0.00

10 4 insufficient data 1.02 0.60

11 4 1.4 0.92 9.68 0.02

12 4 1.01 0.84 6.02 0.20

13 5 -0.71 0.24 10.26 0.07

14 3 3.77 0.99 6.35 0.04

15 5 -1.85 0.03 10.08 0.12

16 5 1.29 0.9 40.12 0

39 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.3. Data processing for photograph quality and individual distinctiveness

Photograph quality was measured using five parameters: clarity-focus, contrast, angle, fin wave effects and the proportion of the frame occupied by the dorsal fin (Urian et al. 1999). Three categories were then defined from the sum of the grades attributed to each parameter for each photo: Q1 – good quality (6-9), Q2 – average quality (10-12) and Q3 – poor quality (> 12). Only Q1 and Q2 photographs were used to identify dolphins and in the analyses.

Individual dolphins were primarily identified based on the nicks and scars on the leading and trailing edges of the dorsal fin (Würsig and Jefferson 1990). Temporary markings such as rake marks and skin lesions only visible for a short-term were also used for individuals regularly sighted. Each dolphin was assigned a grade for distinctiveness of their dorsal fin: D1 - very distinctive fin (i.e., unique features evident even in distant or poor quality photograph); D2 - moderately distinctive fin (e.g., distinctive markings involve subtle nicks); and D3 non-distinctive fin (e.g., fin lacks distinctive markings, a “clean fin”). Mark-recapture analysis was performed using photos of quality Q1 and Q2 for D1 and D2 individuals only.

40 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.4. Method to estimate proportion of distinctly marked individuals and correction of the marked abundance estimates for consideration of the proportion of unmarked individuals (D3 individuals identified from Q1 and Q2)

I calculated mark rates for each site (θ) and adjusted the estimates of population size by following the method described in Nicholson et al. (2012), Tyne et al. (2014), Brown et al. (2016) and Sprogis et al. (2016).

Only sightings where all photographed individuals were identified from a good photo quality (Q1 and Q2) image and without regard to the distinctiveness of their dorsal fins were used to estimate the proportion of marked individuals in the population (θ). Marked rates were calculated for the full study period per site.

Population size estimates were corrected to consider the proportion of unmarked individuals (D3 individuals identified from Q1 and Q2 photos, see Appendix A2.3). The total number of dolphins (per site) with distinctive fins (D1 and D2) was divided by the total number of dolphins (per site) encountered in the selected sightings, as follows: ̂ N̂ N̂ ⁄ , where N̂ is the estimated total population size, N̂ the estimated of the distinctly marked population size, and ̂ the estimate proportion of distinctly marked individuals in the population. All parameters are estimated for each site. The approximate variance for the estimated total population size for overall or per site was derived using the following formula for the standard error of a ratio (Williams et al. 2002):

̂ ̂ ̂ ̂ ( ) (N ) √ N ( ̂ ) , ̂ where n is the number of marked individuals captured within each site.

Log-normal 95% confidence intervals were calculated, with a lower limit of ̂ ̂ ̂ ̂ N N and upper limit of N N , where:

41 Chapter 2 - Population structure investigated by MSCRD

̂ exp ( √ln ( ( ) )) , ̂

(Burnham et al. 1987).

42 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.5. Test for heterogeneity in capture probabilities by implementing goodness-of-fit tests for multistate models using the program U-CARE

Pollock’s closed robust design models do not have a goodness-of-fit (GOF) test to validate the assumptions of equal probabilities of capture and survival between individuals. However, I tested the fit of the data using the program U-CARE (Choquet et al. 2009). Secondary occasions within each primary period were pooled, although the test allows for multiple sites (Pradel et al. 2005).

In U-care, the goodness-of-fit test was divided into three categories: WBWA test for a memory effect; two 3G component tests (3G.Sr and 3G.Sm) for evidence of transience; and two M component tests (ITEC and LTEC) for trap-dependence. If adjustment of the starting model structure was required (based on tests), a variance inflation factor (ĉ) was then estimated by dividing the Pearson statistic of the sum of each test component (χ2) by its degrees of freedom (d.f.), excluding components that were structurally adjusted (Choquet et al. 2009).

Table A2.5.1. Summary of U-CARE test results for bottlenose dolphin study during 2011-2015. Global goodness-of-fit tests were divided into three categories: WBWA test for a memory effect; two test 3G components test for evidence of transience; and 3G.SR for transience; M.ITEC for trap dependence; and. Tests statistics (χ2), corresponding degrees of freedom (d.f.) and P-values are given. See Choquet et al.

(2009) and Pradel et al. (2005) for more details on component tests.

Tests χ2 d.f. P-value WBWA 65.735 26 0.000 3G.Sr 29.324 21 0.106 3G.Sm 34.073 63 0.999 M.ITEC 63.584 21 0.000 M.LTEC 23.835 14 0.048 Note: Significant P-values are in bold.

43 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.6. Bottlenose dolphin groups

Table A2.6.1. Number of bottlenose dolphin groups per site (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and per season of each year (16 seasons, 4 years, aka primary periods).

Season # Group

Year SCR CS/OA GR 2011 Winter 6 13 5 Spring 11 12 5 Summer 5 11 0 2012 Autumn 11 10 5 Winter 13 15 3 Spring 9 18 2 Summer 15 13 1 2013 Autumn 8 17 12 Winter 10 13 6 Spring 3 14 3 Summer 3 16 3 2014 Autumn 7 13 3 Winter 12 15 0 Spring 6 8 2 2015 Summer 9 15 5 Autumn 10 13 1 Sub-Total 138 216 56 TOTAL 410

Chapter 2 - Population structure investigated by MSCRD

Table A2.6.2. Group size (Mean, SE, Min, and Max) per site (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and per season of each year (16 seasons, 4 years, aka primary periods).

Season SCR CS/OA GR Year Mean SE Min Max Mean SE Min Max Mean SE Min Max Winter 3.2 0.6 1.0 8.0 7.5 1.2 1.0 21.0 8.6 2.2 1.0 13.0 2011 Spring 4.2 0.7 1.0 8.0 7.2 1.4 1.0 25.0 6.4 2.6 1.0 14.0 Summer 3.2 0.7 1.0 4.0 5.2 1.4 2.0 11.0 - - - - Autumn 3.7 0.5 1.0 6.0 6.8 1.5 1.0 15.0 6.2 3.0 3.0 18.0 2012 Winter 3.0 0.6 1.0 8.0 3.5 1.0 1.0 20.0 7.0 5.0 2.0 17.0

Spring 3.4 1.7 1.0 9.0 7.1 1.3 1.0 20.0 8.0 6.0 2.0 14.0 Summer 3.3 0.9 1.0 9.0 4.6 1.7 1.0 17.0 17.0 - 17.0 17.0 Autumn 5.8 0.6 3.0 8.0 4.3 0.9 1.0 15.0 6.0 1.1 2.0 13.0 2013 Winter 2.9 0.2 1.0 5.0 6.6 0.4 2.0 14.0 5.0 1.4 1.0 10.0

Spring 4.0 0.9 2.0 7.0 4.6 3.0 1.0 16.0 8.0 3.2 2.0 14.0 Summer 3.7 1.7 2.0 5.0 7.4 0.9 1.0 23.0 5.7 3.2 2.0 12.0 Autumn 3.1 1.1 1.0 8.0 8.4 1.7 2.0 24.0 4.7 2.2 2.0 9.0 2014 Winter 5.5 1.0 1.0 13.0 3.9 0.7 1.0 10.0 - - - -

Spring 4.2 0.7 1.0 8.0 9.9 1.8 3.0 27.0 5.0 4.0 1.0 9.0 Summer 5.0 0.7 1.0 14.0 6.1 2.5 1.0 14.0 13.0 4.2 2.0 22.0 2015 Autumn 4.6 1.0 1.0 11.0 10.4 2.1 1.0 23.0 32.0 - 32.0 32.0 Overall 3.9 0.2 1.0 14.0 6.3 0.4 1.0 27.0 7.6 0.9 1.0 32.0

Chapter 2 - Population structure investigated by MSCRD

Appendix A2.7. Bottlenose dolphin individuals

Table A2.7.1. Summary of the number of identified individuals of bottlenose dolphins (and marked only) by site (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) in each season of each year (16 seasons, 4 years) and from good photo quality only.

Year Season Site SCR CS/OA GR Winter 12 (10) 70 (52) 35 (27) Year 1 Spring 17 (13) 76 (57) 28 (18)

(2011-2) Summer 9 (6) 39 (30) 0 (0)

Autumn 20 (16) 47 (39) 29 (23) Winter 19 (14) 46 (41) 15 (15) Year 2 Spring 16 (10) 81 (61) 14 (14)

(2012-3) Summer 18 (14) 46 (36) 13 (9)

Autumn 16 (10) 59 (39) 49 (33)

Winter 20 (12) 64 (41) 24 (17) Year 3 Spring 8 (5) 52 (36) 21 (20)

(2013-4) Summer 8 (6) 78 (51) 14 (9)

Autumn 16 (10) 66 (43) 14 (7) Winter 22 (15) 42 (21) 0 (0) Year 4 Spring 20 (11) 59 (39) 10 (7)

(2014-5) Summer 24 (14) 63 (42) 54 (32)

Autumn 22 (14) 90 (60) 32 (29)

Overall 37 (25) 218 (139) 170 (107)

Total (all sites confounded) 343 (209)

Chapter 2 - Population structure investigated by MSCRD

Appendix A2.8. Multistate closed robust design models summary

Table A2.8.1. Multistate closed robust design models (in rank order of AICc scores) for each scenario: Scenario 1 with three sites and Scenario 2 with two sites. The table provides an overview of the Akaike Information Criterion corrected for small sample size (AICc), difference in AICc with best-fitting model and AICc weight, the number of parameters used in model fit and the deviance explained.

Models AIC ΔAIC AICc Model Parameters Deviance c c weight likelihood Scenario 1 (three states) φ(period) ψ(site) p(site × period) N(site × period), p=c 3574.0 0.0 0.994 1.0000 114 3326.1 φ(.) ψ(limited-site) p(site × period) N(site × period), p=c 3585.0 11.0 0.004 0.0041 95 3381.3 φ(site) ψ(limited-site) p(site × period) N(site × period), p=c 3587.0 13.0 0.002 0.0015 96 3381.0 φ(.) ψ(.) p(site × period) N(site × period), p=c 3633.0 59.0 0.000 0.0000 95 3429.3 φ(.) ψ(limited-site × period) p(site × period) N(site × period), p=c 3645.8 71.7 0.000 0.0000 138 3340.1 φ (period) ψ(.) p(site × period) N(site × period), p=c 3655.8 81.8 0.000 0.0000 108 3422.0 Scenario 2 (two states) φ(site) ψ(site) p(site × period) N(site × period), p=c 2292.5 0.0 0.923 1.0000 65 2156.3 φ(.) ψ(site) p(site × period) N(site × period), p=c 2297.5 5.0 0.077 0.0834 64 2163.4 φ(site) ψ(site × period) p(site × period) N(site × period), p=c 2309.5 17.0 0.000 0.0002 95 2106.0 φ(period) ψ(site) p(site × period) N(site × period), p=c 2311.4 18.9 0.000 0.0001 79 2144.1 φ(.) ψ(site × period) p(site × period) N(site × period), p=c 2313.1 20.6 0.000 0.0000 93 2114.2 φ(period) ψ(site × period) p(site × period) N(site × period), p=c 2326.5 34.0 0.000 0.0000 107 2095.2 Note: φ apparent survival; ψ transition rate; p probability of capture; p = c probability of capture is equal to recapture; N abundance; (.) constant; (site)

varying by site; (limited-site) varying by site but restricted to symmetric movement such that ψ(1 2) = ψ(2 1); (period) varying by primary period; (site 47 × period) varying by site and primary period; (limited-site × period) varying by site (restricted to symmetric movement) and primary period.

Chapter 2 - Population structure investigated by MSCRD

Appendix A2.9. Sighting frequency

Figure A2.9.1 Sighting frequency of adult/sub-adult dolphins observed in the metropolitan waters of Perth from June 2011 to May 2015: Top – for all; Bottom – per site for individuals seen in more than 10% of the surveys. GR = Gage Roads, SCR = Swan Canning Riverpark, CS/OA =

Cockburn Sound/Owen Anchorage. 48

Chapter 2 - Population structure investigated by MSCRD

Appendix A2.10. Coefficient of variations (CV) of capture probability (p)

Figure A2.10.1. Coefficient of variations (CV) of each capture probability (p) represented as box plot (min; Quartile 1; median; Quartile 3; max) for each Scenario: Scenario 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads); and Scenario 2 – two sites (SCR and Coastal).

49 Chapter 2 - Population structure investigated by MSCRD

Appendix A2.11. MSCRD analysis using sex as an individual covariate

Materials and Methods

In MSCRD models, I also attempted to address individual heterogeneity in capture probabilities by including sex as individual covariate (i.e., female, male or unknown). Individuals were sexed in-situ based on the direct observation of the genitals or on the presence of a dependent calf. Additionally, biopsy samples were collected throughout the entire study area between 2007 and 2015 using a remote biopsy system designed for small cetaceans by PAXARMS (Krützen et al. 2002). I then molecularly sexed individuals following the method described in Brown et al. (2014).

In MARK, each MSCRD model combination was run with the probability of capture (p) varying by sex, site and/or primary period or constant, and with recapture probability (c) set as equal to first capture probability (p). The abundance (N) was set to vary by sex, site and primary periods [N(sex × site × primary periods)]. Several sub-models for apparent survival (φ) were run (i.e., whether it varied by sex and/or site or none). Probability of transition between sites (ψ) was also estimated whether it varied by sex and/or site or none. Time intervals between primary periods were specified as a fraction of a year (i.e., seasons). Thus, when apparent survival and transitions probabilities were time-constant estimated, their estimates were annual.

Goodness-of-fit tests were re-run for each sex class: females, males and not sexed individuals following the same protocol as previously (see Appendix A2.5). Models were ranked using the Akaike Information Criterion (AICc, Burnham and Anderson

2002). The model with most support by AICc (highest AICc weight) was selected as the most parsimonious model. Models with ΔAICc < 2 were also considered to have support from the data (Burnham and Anderson 2002).

50 Chapter 2 - Population structure investigated by MSCRD

Results

Among the 339 individuals, 108 females and 61 males were identified although 134 individuals (including 40 females and 11 males) were not included in the mark- recapture analysis because of insufficiently marked dorsal fins.

Model selection and Capture probabilities

Goodness-of fit based on multistate and subcomponent tests in U-CARE indicated 2 no heterogeneity in capture for any of the sex classes (χ females = 75.093, d.f. = 67, P- 2 2 value = 0.233, χ males = 46.780, d.f. = 60, P-value = 0.894, and χ not sexed = 30.719, d.f. = 56, P-value = 0.998, see Table A2.11.1 for summary of GOF tests). I did not adjust the Quasi-likelihood (QAICc) for substantial overdispersion, which meant the most parsimonious model was defined under no adjustment (Cooch and White 2005).

The best fitting model, based on the AICc weight, for Scenario 1 (three sites) was that capture probability varied by site and primary period [p(site × primary period)], apparent survival rate varied by sex [φ(sex)] and transitions between sites varied between each site but not sex [ψ(site)] (see Table A2.11.2). For Scenario 2 (two sites), the best-fitting model was that capture probability varied by site and primary period [p(site × primary period)], apparent survival varied by sex class [φ(sex)] and transitions varied between site and sex [ψ(sex × site)] (see Table A2.11.2). Due to model constraints it was too difficult to apply the sex covariate as variable for capture probabilities. Estimates of capture probabilities and coefficients of variation (CV) of the capture probabilities (Figures not showed) were similar to those obtained in MSCRD models without using sex as individual covariate (i.e., the CV was high for GR and low for CS/OA with CV = 31% and 15%, respectively).

Chapter 2 - Population structure investigated by MSCRD

Table A2.11.1. Summary of U-CARE test results for bottlenose dolphin study during 2011-2015 per sex classes (female, male, not sexed). Global goodness-of-fit (JMV Model) tests were divided into three categories of components: WBWA test for memory; 3G.Sr and 3G.Sm for transience; and M.ITEC and M.LTEC for trap dependence. Tests statistics (χ2), corresponding degrees of freedom (d.f.) and P-values are given. See Choquet et al. (2009) and Pradel et al. (2005) for more details on component tests.

Females Males Not sexed

Tests χ2 d.f. P-value χ2 d.f. P-value χ2 d.f. P-value WBWA 29.798 14 0.008 16.347 14 0.293 6.455 6 0.374

3G.Sr 8.841 8 0.356 5.953 5 0.311 7.849 14 0.897

3G.Sm 18.229 27 0.894 8.575 29 1 9.483 32 1

M.ITEC 14.038 11 0.231 9.721 8 0.285 3.450 2 0.178

M.LTEC 4.118 7 0.766 6.184 4 0.186 3.482 2 0.175

Goodness-of-fit 75.093 67 0.233 46.780 60 0.894 30.719 56 0.998

Note: Significant P-values are in bold

Chapter 2 - Population structure investigated by MSCRD

Table A2.11.2. Multistate closed robust design models (in rank order of AICc scores) for each scenario: Scenario 1 with three sites and Scenario 2 with two sites. The table provides an overview of the Akaike Information Criterion corrected for small sample size (AICc), difference in AICc with best-fitting model and AICc weight, the number of parameters used in model fit and the deviance explained.

AIC Model Models AIC ΔAIC c Parameters Deviance c c weight likelihood Scenario 1 (three states) φ(sex) ψ(site) p(site × period) N(sex × site × period), p=c 5638.0 0.0 0.983 1.0000 164 5267.3 φ(sex × site) ψ(site) p(site × period) N(sex × site × period), p=c 5646.1 8.1 0.017 0.0176 166 5270.3 φ(.) ψ(site) p(site × period) N(sex × site × period), p=c 5656.2 18.2 0.000 0.0001 161 5293.1 φ(site) ψ(site) p(site × period) N(sex × site × period), p=c 5658.3 20.3 0.000 0.0000 164 5287.6 φ(sex) ψ(sex) p(site × period) N(sex × site × period), p=c 5722.4 84.4 0.000 0.0000 159 5364.4 φ(sex × site) ψ(sex) p(site × period) N(sex × site × period), p=c 5732.4 94.4 0.000 0.0000 166 5356.6 Scenario 2 (two states) φ (sex) ψ(sex × site) p(site × period) N(sex × site × period), p=c 4440.4 0.0 0.928 1.0000 105 4213.8 φ (sex × site) ψ(sex × site) p(site × period) N(sex × site × period), p=c 4445.5 5.1 0.072 0.0780 108 4211.9 φ (site) ψ(sex × site) p(site × period) N(sex × site × period), p=c 4459.6 19.2 0.000 0.0001 103 4237.7 φ (sex) ψ(site) p(site × period) N(sex × site × period), p=c 4460.4 20.1 0.000 0.0000 105 4233.9 φ (sex × site) ψ(site) p(site × period) N(sex × site × period), p=c 4465.7 25.3 0.000 0.0000 108 4234.1 φ (.) ψ(sex × site) p(site × period) N(sex × site × period), p=c 4466.8 26.4 0.000 0.0000 103 4244.8 Note: φ apparent survival; ψ transition rate; p probability of capture; p=c probability of capture is equal to recapture; N abundance; (.) constant; (sex) varying by sex; (site) varying by site; (sex × site) varying by sex and site; (sex × site × period) varying by sex, site, and primary period; (site × period) varying by site and primary period.

Chapter 2 - Population structure investigated by MSCRD

Apparent survival estimates and abundances

Given that the majority of the individuals were not sexed, I acknowledge that estimates of the apparent survival rates obtained through the best-fitted models may be overestimated for sexed individuals and underestimated for not sexed individuals. In fact, most of the sexed individuals were those regularly seen during the study period. While Nichols et al. (2004) demonstrated how to deal with not sexed individuals in capture-recapture approaches, this method could not be applied in this study due to the complexity of the models. Regardless of the scenarios, models yielded the same range of apparent survival rates ( ) for females ( ̂ = 0.91-0.94,

SE 0.02) and males ( ̂ = 0.93-0.95, SE 0.02-0.04) and were lower for unknown sex

( ̂ = 0.70-0.74, SE 0.04), although such results were expected as most sexed individuals were resighted multiple times. Biopsy sampling for dolphins is expensive (cost of equipment and time outlays) and, thus, sampling may not be random but target individuals that can be recognized (i.e., that have a distinctive fin) and which are seen multiple times so that individual data (i.e., sex, capture history, grouping) can be used for multiple purposes, including methods that require a certain quantity of data.

No obvious seasonal variation in abundance estimates was detected in the CS/OA site for males and females (N̂ = 44, min 32, max 58; N̂ = 33, min 16, max 48), although estimates for not sexed individuals were more variable

N̂ = 23, min 16, max 56) (Figure A2.11.1). Abundance estimates for females in GR varied with the highest in summer 2013 (N̂ = 91, 95% CI 24- 346, Figure A2.11.1) while there was no record in summer 2012 and winter 2014

(N̂ = 0). Only two not sexed individuals were recorded in the SCR during autumn and winter 2012. In other sites, however, estimated abundances of not sexed individuals varied with GR showing more variable and less precise estimates

(N̂ = 53 individuals, min 0, max 127, SE 0-88) (Figure A2.11.1).

54 Chapter 2 - Population structure investigated by MSCRD

Figure A2.11.1. Seasonal estimated abundances (N̂total ± 95% confidence intervals) by sex (column) and by Scenario (row): 1 – three sites (SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and 2- two sites (SCR and Coastal). Lines between data points have been used for illustrative purposes only; continuity of values is not implied. Sites are: SCR (red), CS/OA (yellow), GR (purple) and Coastal in Scenario 2 (grey). Sexes are: females (circle), males (square) and unknown (triangle).

Movements

The estimates of the transition probabilities ( ̂) yielded by the model in Scenario 1 suggested that there was very little or no movement between the SCR and GR sites -3 ( ̂ < 10 ) (Table A2.11.2). Model yielded higher movement rate from the

SCR to CS/OA ( ̂ = 0.153, SE 0.028) than in the opposite direction

( ̂ = 0.026, SE 0.005). Estimates from Scenario 2 also indicated similar transition rates for females ( ̂ = 0.128, SE 0.040) and males ( ̂ = 0.157, SE 0.005) from the SCR to coastal waters, although rates in the opposite direction were lower ̂ .

55 Chapter 2 - Population structure investigated by MSCRD

Table A2.11.2 Estimates of movement transitions ψ (SE) between sites for (a) Scenario 1 (three sites: SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound/Owen Anchorage, GR = Gage Roads) and (b) Scenario 2 (two sites: SCR vs. Coastal). In comparison to Scenario 1, estimates for Scenario 2 varied by sex. Values in italic represent rates when staying in the same site.

(a)

Movement Into: From: SCR CS/OA GR All

SCR 0.847 0.153 (0.028) 0.000 (0.000)*

CS/OA 0.026 (0.005) 0.917 0.057 (0.010)

GR 0.000 (0.002)* 0.099 (0.015) 0.901 Note: * Values estimated were smaller than 0.001.

(b)

Movement Into: From: SCR Coastal Females

SCR 0.872 0.128 (0.040) Coastal 0.016 (0.005) 0.984 Males

SCR 0.843 0.157 (0.038) Coastal 0.035 (0.005) 0.965 Unknown

SCR 0.351 0.649 (0.277)** Coastal 0.002 (0.002) 0.998

Note: ** ψUnknown moving from SCR to Coastal waters was estimated to 1 when using the last capture approach.

Chapter 3. Identifying the relevant local population for environmental impact assessments of mobile marine fauna

3.1. Abstract

Environmental impact assessments must be addressed at a scale that reflects the biological organisation for the species affected. It can be challenging to identify the relevant local wildlife population for impact assessment for those species that are continuously distributed and highly mobile. Here, I document the existence of local communities of Indo-Pacific bottlenose dolphins (Tursiops aduncus) inhabiting coastal and estuarine waters of Perth, Western Australia, where major coastal developments have been undertaken or are proposed. Using sighting histories from a four-year photo-identification study, I investigated fine-scale, social community structure of dolphins based on measures of social affinity, and network (half-weight index - HWI, preferred dyadic association tests, and lagged association rates - LAR), home ranges, residency patterns (lagged identification rates - LIR), and genetic relatedness. Analyses revealed four socially and spatially distinct, mixed-sex communities. The four communities had distinctive social patterns varying in strength, site fidelity, and residency patterns. Overlap in home ranges and relatedness explained little to none of the association patterns between individuals, suggesting complex local social structures. The study demonstrated that environmental impact assessments for mobile, continuously distributed species must evaluate impacts in light of local population structure, especially where proposed developments may affect core habitats of resident communities or subpopulations. Here, the risk of local extinction is particularly significant for an estuarine community because of its small size, limited connectivity with adjacent communities, and use of areas subject to intensive human use. In the absence of information about fine-scale population structure, impact assessments may fail to consider the appropriate biological context.

57 Chapter 3 - Identifying local population for EIA

3.2. Introduction

Applied wildlife research can improve the scientific basis for environmental impact assessment (EIA) by developing methodologies to evaluate impacts of human activities on wildlife (Morrison et al. 2006; Steidl and Powell 2006; Bejder et al. 2009, 2012; Torres et al. 2016). However, for such evaluations to be effective, they must also be directed at an appropriate scale of biological organisation for the species to be impacted by a proposed development or activity. Here, I describe a methodology to identify the relevant local population for EIA which is suitable for species that are continuously distributed and highly mobile.

While the procedural and formal requirements for EIA are often closely prescribed by statutes, regulations, and associated policy and guidance documents, there are often few specific requirements as to the scientific information necessary for EIA. While many jurisdictions have now developed policies or protocols for biodiversity surveys to allow for the identification of fauna and flora species that may affected by a proposed development or activity, prescriptive guidelines for the conduct of field studies of human impacts on wildlife, as set by the administrative bodies having statutory responsibility for EIA, remain uncommon. As such, it is vital for wildlife researchers to identify best-practice methodologies for field-based impact assessment research, so as to encourage their use in studies undertaken to support EIAs.

Broadly speaking, the aim of EIA is to conduct a detailed assessment of the potential impacts of a proposed development or activity on a particular environment (including the biota occurring there) on which decision-makers can then rely in determining whether the proposed development or activity should be approved and, if so, with what conditions (Glasson et al. 2012). Ideally, the EIA process for a proposed development or activity should consider the range of possible impacts on wildlife in a manner that is species-, site-, and (if applicable) season-specific (Fox et al. 2006). If the assessment methods employed are inappropriate or are inadequately implemented, the EIA outcomes (typically an environmental impact statement or report) may be incomplete and inaccurate. An obvious example is an EIA based on sparse and opportunistic sighting data for a species (Bejder et al. 2012).

58 Chapter 3 - Identifying local population for EIA

To adequately characterise the impact of a proposed development or activity on a particular species, it is necessary to identify the relevant local population which may be impacted (Brittingham et al. 2014; Bastos et al. 2016; Brown et al. 2016). In highly fragmented landscapes, the relevant local population may be straightforward to delineate because of the geographical separation between the area affected by the development and the nearest other sites where the species may be found. For example, the presence of physical barriers (natural or anthropogenic) or long distances between patches may limit dispersal and thus enable isolation of local populations (e.g., natural and anthropogenic barriers for cougars, Sweanor et al. 2000; land-clearing for wombats, Walker et al. 2008; geographical distance for sharks, Sandoval-Castillo and Beheregaray 2015). In contrast, identifying the relevant local population may be more challenging if a species is highly mobile (and therefore able to disperse between even geographically distant sites) or displays migratory behaviour (e.g., between breeding sites and feeding areas), or if it is continuously distributed across the land or seascape (DeYoung 2007).

What constitutes a “local population” is an important though often under-considered aspect of impact assessment research. Some concept of a local population is often implicit in considerations of spatial scale and population structure for EIA. In one sense, the local population may simply comprise the total number of individuals that may be affected by the proposed development or activity. In the simplest scenario, an EIA could proceed on the basis of an estimate of the number of individuals present in the ‘patch’ that the development or activity will affect (Total Individuals Affected). However, it will often be desirable (or necessary) to identify the relevant biological population that may be impacted, in the sense of a group of animals (or ‘subpopulation’) that displays some meaningful degree of genetic, demographic, or spatial discreteness (Population Unit Affected). An effective EIA may therefore require information about population structure, so that decision-makers can evaluate the biological significance of potential impacts - e.g., will the development affect the viability of a distinct population (or population unit) or is the species continuously distributed across the impact area and its surrounds such that little or no population structure is present? A metapopulation framework is often applied to examine interactions between spatially distinct local populations (Levins 1969; Hill et al. 1997; Moilanen and Nieminen 2002).

59 Chapter 3 - Identifying local population for EIA

Weak population structure is often expected for marine wildlife because of the lack of barriers to movement and the broad distributions of many marine species (Waples 1998). Nonetheless, geographic features do exist in the marine environment that may act as natural boundaries and thus contribute to population structuring, such as between estuarine, coastal, and offshore habitats. Bottlenose dolphins (Tursiops spp.), for example, are known to exhibit population (and even species) structure across a gradient from protected inshore environments to deeper, more exposed offshore habitats. For example, estuarine bottlenose dolphins generally exhibit greater site fidelity and year-round residency, have stronger and more enduring associations with conspecifics than do bottlenose dolphins in coastal habitats, and may form distinct “communities” within particular estuaries or embayments (Quintana-Rizzo and Wells 2001). A “community” has been defined as a set of individuals that is behaviourally discrete from neighbouring communities and within which most individuals associate with other members of the community (Wells et al. 1987). I suggest that a dolphin community might constitute a relevant local population for the purposes of EIA, both in terms of comprising the total number of animals that might be affected by a proposed development (Total Individuals Affected) and in terms of representing a population unit of some biological significance (Population Unit Affected). However, the diversity and flexibility of mammalian social behaviour can make it difficult to identify communities for dolphins and other social mammal species (both terrestrial and marine) (Cantor and Whitehead 2013).

As nearshore environments such as embayments and estuaries are a focus point for coastal development, EIAs will often need to consider how proposed developments and activities will affect local dolphin populations. These environments contain shallow and protected habitats that allow bottlenose dolphins to reside year-round (Wells 1986; Brusa et al. 2016). Prey availability in these environments is also more continuous and dependable (Elliott and Whitfield 2011; McCluskey et al. 2016) than in open and coastal regions, where prey is distributed patchily and prey availability is dictated largely by oceanic physical processes (Silva et al. 2008). The key ecological and demographic characteristics of dolphin communities in estuaries and embayments differ from those in coastal areas - e.g., inshore communities tend to be small and to exhibit weak to moderate levels of dispersal and immigration (Titcomb

60 Chapter 3 - Identifying local population for EIA et al. 2015). These characteristics influence their vulnerability or resilience to human impacts (Bejder et al. 2009; Pirotta et al. 2013).

Impact assessment research has been undertaken for a range of developments and activities that may impact on dolphins, including activities such as dredging and pile driving that may exert short-term impacts on dolphin populations (Dungan et al. 2012; Pirotta et al. 2013; Culloch et al. 2016) and those which are more enduring, such as the construction of permanent infrastructure (Jefferson et al. 2009; Cagnazzi et al. 2013b). The coastal and estuarine waters of Perth (Western Australia) have experienced significant development for industrial and other commercial uses. Notably, the Swan Canning Riverpark (SCR) estuary bisects the city, threading through heavily developed residential and agricultural areas (Holyoake et al. 2010) and Cockburn Sound (CS), a sheltered embayment, contains Perth’s main industrial area. Some developments in the region have involved short-term activities (e.g., pile driving activities conducted in the Inner Harbor of the Port of Fremantle, Salgado Kent et al. 2012; Paiva et al. 2015), while others are continuing impacts (e.g., year- round dredging for a shell-sand mining operation, Environmental Protection Authority 2001). In addition, new developments may be undertaken in the near future (e.g., a proposed outer harbor development on Kwinana Shelf, Western Australian Planning Commission 2004; a desalination plant proposed for the northern metropolitan coast, Mercer 2013). In many respects, these developments are exemplars of the types of developments which may impact on dolphins and other wildlife in coastal and estuarine environments.

In this paper, I: (a) describe a methodology to identify local populations of wildlife and then (b) examine its further application in evaluating possible impacts of proposed developments and activities. Firstly, I integrated social, ecological, and genetic data collected during longitudinal field study to identify relevant local populations of Indo-Pacific bottlenose dolphins (T. aduncus) within estuarine and coastal waters of Perth. I employed sampling methodologies that are robust and consistent with best-practice for long-term monitoring, abundance estimation and behavioural study of coastal dolphins (e.g., systematic line-transect survey, photo- identification over a four-year period (2011-2015)) to: (1) assess social networks of bottlenose dolphins through estimates of the half-weight index (HWI) and network

61 Chapter 3 - Identifying local population for EIA analysis; (2) compare the role of sex composition and temporal stability of association in driving social organisation within communities based on lagged association rates (LAR) and preferred associations; (3) examine the spatial segregation of the social communities by assessing home range overlaps in conjunction with bathymetry and habitat differences, to assess the residency patterns; and (4) evaluate the genetic relatedness within and between communities by estimating the relatedness between individuals within and between communities. Secondly, I considered the fine-scale population structure of bottlenose dolphins in the context of past, current, and proposed developments for the region to demonstrate how such information about local populations can also assist in evaluating the possible impacts of proposed developments and activities.

3.3. Materials and methods

3.3.1. Study area

The study area was located in the metropolitan waters of Perth (Western Australia), one of the fastest growing capital cities in Australia (Australian Bureau of Statistics 2015). The study area encompassed 275 km2 and extended from Rockingham to Scarborough along the coast and then inland to include part of the Swan Canning Riverpark (SCR), an estuarine reserve (Figure 3.1). Following a mark-recapture robust design (see Chapter 2, Chabanne et al. 2017b), the study area was subdivided into four geographic regions with three that were defined by the topography and bathymetry of the coastal waters (from South to North, Figure 3.1): (1) Cockburn Sound (CS) – a semi-enclosed embayment with varying depth (< 2 to > 20 m) and with seagrass, sand, silt, or limestone substrates; (2) Owen Anchorage (OA) – an embayment with less than 10 m depth, except in the channel (max depth: 14.7 m), and with a substrate mainly consisting of shell-sand and seagrass; (3) Gage Roads (GR) – an open coastline typified by deep waters (> 10 m), with sandy beaches, rocky reefs, and seagrass patches. The lower section of GR, also deeper (> 20 m), is an anchoring area for ships before entering the Port of Fremantle. The SCR is a micro-tidal estuary which encompasses an area of about 55 km2 and includes two river systems (Swan and Canning rivers) that join near the City of Perth before reaching the Indian Ocean through the Inner Harbour of the Port of Fremantle. While

62 Chapter 3 - Identifying local population for EIA the estuary is mainly shallow (< 10 m), the Inner Harbour section is maintained at 14 m through regular dredging activities (Figure 3.1b). With a Mediterranean climate, the estuary experiences marked temperature and salinity variations through the year, particularly when freshwater flow is weak.

Chapter 3 - Identifying local population for EIA

Figure 3.1. Maps of the study area showing: (a) the transect routes per geographic regions (GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound) with locations of past, current, and proposed developments (1- pile driving; 2- dredging; 3- desalination; 4- outer harbour); and (b) the bathymetry (in meters) with the locations of the groups sighted during the systematic surveys

conducted from 2011 to 2015, including mixed groups (i.e., mix of individuals from different communities). 64

Chapter 3 - Identifying local population for EIA

3.3.2. Data collection

Boat-based surveys were conducted between June 2011 and May 2015 and within each season corresponding to the Australasian calendar (winter: June to August; spring: September to November; summer: December to February; autumn: March to May). Using boat-based photo-identification sampling, I documented individual bottlenose dolphins based on nicks and marks on the dorsal fin (Würsig and Jefferson 1990). Three zig-zag transect routes (offset by 2 km) were designed using Distance 6.0 (Thomas et al. 2009) for each coastal geographic region in order to optimise the coverage (Figure 3.1a). Each route extended to c. 7 km offshore in CS and OA and from 5 to 3 km offshore in GR. In the SCR, I followed the same transect route used during the 2001-03 study (Chabanne et al. 2012) which extended from the mouth of the estuary (Inner Harbour) through the lower reaches and the main basin where the Swan and Canning rivers join. Full details on the robust design sampling structure design and survey methodology (i.e., predefined transect routes, Figure 3.1a) are provided in the Chapter 2 (Chabanne et al. 2017b).

A cycle was defined as a successful completion of a survey within each geographic region of the study area. My goal was to complete a cycle in a minimum period, although a minimum of two days was required because of daylight. During a survey, a sighting was defined as a group of dolphins observed within c. 250 m on either side of the boat along the transect route (Wells et al. 1987; Wells et al. 1999; Quintana- Rizzo and Wells 2001). A group consisted of one to several dolphins. Dolphins were considered to be in a group if they were within 10 m of any individual within the group (a 10 m “chain” rule) and engaged in the same behaviour (Smolker et al. 1992). For each sighting, I recorded the location (southing/easting using a hand-held GPS unit), behaviour, group size, and age-sex composition. Age classes (adult, juvenile, and calf) were based on the body size or the presence of a dependent calf. Individuals were sexed through: (i) molecular analyses of tissue samples (Gilson et al. 1998; Brown et al. 2014) collected via remote biopsy sampling (Krützen et al. 2002); (ii) field observation of the genital regions; or (iii) the presence of a dependent calf (for females). I performed photo-identification using a Nikon D300

with Nikkor lens 70-300 mm or a D7000 with Nikkor lens 80-400 mm. Full details

65 Chapter 3 - Identifying local population for EIA for photo-identification and grading processes are provided in Chapter 2 (Chabanne et al. 2017b).

3.3.3. Association patterns

To analyse patterns of association, I used only high-quality photographic identifications, from groups for which all individuals were identified. I did not consider the distinctiveness of the individuals to avoid small sample sizes. The survey frequency also allowed for use of temporary marks, if required. Individuals present in the same group were assumed to be associated ("gambit of the group", Whitehead 2008a). Calves were excluded from the analysis because they lack identifying marks, are dependent on their mothers and have high natural mortality (Mann et al. 2000). The sampling period was set to one day to minimise sampling time incoherency between successful surveys of the entire study area (i.e., a cycle). Individuals seen multiple times within the same cycle were restricted to the first sighting only. The strength of associations among dyads (i.e., pairs of individuals, n = 8,256) was calculated using the half-weight index (HWI, Cairns and Schwager 1987). Values of HWI range from 0 (never associated) to 1 (always associated). The HWI is frequently used in social structure of cetaceans as it reduces bias due to incomplete identification within encounters (Cairns and Schwager 1987). Only adults and juveniles sighted more than five times over the entire study period were retained for this analysis. The minimum number of sightings was decided by comparing the social differentiation (S, measure of variability of the associations) and Pearson’s correlation coefficient (r, measure of the quality of the representation of the association pattern, Whitehead 2008b) for different sets of data based on a minimum number of sightings per individual, until appropriate values were reached. Specifically, an S < 0.3 indicates that the society is homogeneous, 0.5 < S < 2 indicates that the society shows some strong associations between individuals, and S > 2 indicates that the society generally has weak associations between individuals (Whitehead 2008b). Additionally, an r value near 1 indicates that the representation is excellent, while r ~ 0.8 suggests a good representation and r ~ 0.4 indicates a moderate representation (Whitehead 2008b). All association patterns were analysed using the software SOCPROG 2.6 (Whitehead 2009).

66 Chapter 3 - Identifying local population for EIA

A Monte Carlo permutation test was conducted to examine whether associations within the study area population were different from random (Bejder et al. 1998; Whitehead et al. 2005; Whitehead 2008a). As such, higher coefficients of variation (CV) of real association indices compared to that of randomly permuted data indicated the presence of preferred long-term companions in the studied population (Whitehead 1999). I ran 103 permutations with 103 flips per permutation for the complete dataset and significant variations from random were tested using a two- tailed test (P-value = 0.05). The number of permutations was determined to be sufficient when the P-value stabilised (Bejder et al. 1998). The preferred associations are those for which an association index value is at least twice higher than the mean (Whitehead 2008b). A Mantel test, using 103 permutations, was carried out to examine whether differences in associations occurred between sex classes (two-tailed 0.05 P-value, Schnell et al. 1985).

3.3.4. Community structure and dynamic

To investigate the social structure based on the HWI, I calculated the eigenvector modularity network algorithms to identify cut-off limits to identify possible communities (Newman 2004, 2006). A modularity M > 0.3 indicated that the community division is meaningful (Newman 2004; Whitehead 2009). I used the software NetDraw 2.139 (Borgatti 2002) to visualise the network structure. For comparison, I carried out an average linkage hierarchical cluster analysis that calculated a cophenetic correlation coefficient (CCC). A CCC > 0.8 indicates a good match between the degree of association between individuals and the association matrix (Bridge 1993).

I examined the association levels, some network measures, sex segregation and the temporal stability of associations to highlight potential differences in association patterns between the different communities identified. First, mean and maximum levels of associations were compared. Second, I measured the network strength, clustering coefficient and affinity within communities. The strength is the sum of the association indices of each individual, also defined as a measure of gregariousness (Barrat et al. 2004); the clustering coefficient indicates how well an individual’s

associates are themselves associated; and the affinity is a measure of the strength of

67 Chapter 3 - Identifying local population for EIA an individual’s associates (Whitehead 2016). All the network measures were calculated in SOCPROG 2.6 (Whitehead 2009). Third, a Mantel test (as described above) was then carried out to examine whether differences in associations occur between sex classes within each community. Additionally, tests for preferred or avoided associations were run for each community and per sex classes as described above. And finally, I measured the persistence of associations within each community by calculating the lagged association rates (LAR, Whitehead 1995). The LAR estimates the probability that two individuals sighted together at a given time will still be associated at some time lag later. LARs from each community were compared to the null LAR of the complete dataset (i.e. association value the animals would have if associating randomly, Whitehead 1995). I then tested exponential decay models characterising the patterns of dyadic association over time. The quasi- Akaike information criterion (QAIC) was used for model selection (Whitehead 2007). I used the jackknife method to obtain estimates of precision of the LAR (Efron and Stein 1981). LARs were also estimated and modelled as above for each sex class within each community.

3.3.5. Spatial distribution of communities

I calculated the estimates of kernel density (KDE) in ArcGIS 10.3 and estimated the probability of contours of 50% (i.e., the core of a community) and 95% (i.e., community’s home range defined by the outermost boundaries) by pooling sightings of individuals assigned to the same community. Individuals that were equally observed in two or more geographic regions were excluded from this analysis. I used the kernel interpolation with barriers tool to take into account land barriers to movements (the output grid cell size was set to 200 × 200m and the bandwidth was fixed to 6,000 for each individual, Sprogis et al. 2015). All other steps followed the protocols by MacLeod (2014) and were calculated in the Universal Transverse Mercator (UTM) Zone 50 South projection using the coordinate system World Geodetic System (WGS) 1984 datum. Overlaps home ranges between each community were computed using the Intersect tool in ArcGIS. In order to characterise some factors associated with community structure, I also calculated an asymmetric matrix of pairwise individual home range (95% kernel density) overlaps following the same protocol as above and conducted a Mantel test (103 permutations)

68 Chapter 3 - Identifying local population for EIA to check the correlation between home range overlaps and HWIs for each individual pairs between and within communities.

3.3.6. Residence time

Using the software SOCPROG 2.6 (Whitehead 2009), I assessed the demographic processes within each community by estimating the lagged identification rates (LIR) for each individual within their respective assigned community (Whitehead 2001). This analysis estimated the probability that an individual would be resighted in the study area after a certain time lag (td) in comparison to a randomly chosen individual. I then fitted different models of no movment (i.e., closed populations), emigration and reimmigration, and emigration, reimmigration and mortality to the observed LIR (Whitehead 2001). I used the QAICc to select the most parsimonious model (Burnham and Anderson 2002). The LIR confidence intervals (CI) were obtained using bootstrap replicates (Whitehead 2008b).

3.3.7. Genetic relatedness

Skin samples were collected via remote biopsy sampling (Krützen et al. 2002) over the four-year period (2011-2015) mentioned earlier. However, additional samples collected between 2007 and 2010 were also included for genetic testing. All biopsy samples were stored in DMSO buffer for cryopreservation. Genomic DNA was extracted from all skin samples using the Gentra Puregene Tissue Kit (Qiagen) and following the manufacturer’s protocol. Samples were genotyped at 13 different microsatellite loci: DIrFCB4, DIrFCB5 (Buchanan et al. 1996), LobsDi_7.1, LobsDi_9, LobsDi_19, LobsDi_21, LobsDi_24, LobsDi_39 (Cassens et al. 2005), SCA9, SCA22, SCA27 (Chen and Yang 2008), TexVet5, TexVet7 (Rooney et al. 1999). I followed the PCR conditions as described in Frère et al. (2010a). The single stranded PCR products were run on an ABI 3730 DNA Sequencer (Applied Biosystems). Genotypes were scored using GENEIOUS 9.1.5 (http://www.geneious.com; Kearse et al. 2012) with microsatellite plugin 1.4 (Biomatters Ltd). Each microsatellite locus was checked for null alleles† and scoring errors using the software Micro-Checker 2.2.3 with a confidence level of 95% (Van

Oosterhout et al. 2004). Departures from Hardy-Weinberg equilibrium (HWE) and

69 Chapter 3 - Identifying local population for EIA linkage disequilibrium† were tested using the Markov chain probability test and 104 iterations in Genepop 4.4.3 (Rousset 2008). Significance values for multiple comparisons were adjusted by sequential Bonferroni corrections (Rice 1989). I calculated individual pairwise relatedness within and between social communities using Queller and Goodnight (1989) index (QG) in Coancestry 1.0.1.2 (Wang 2011). Average relatedness coefficients within communities were tested using t-test. I also conducted an ANOVA test to identify the correlation between pairwise relatedness and HWI within each community.

3.4. Results

3.4.1. Effort and group size

A total of 322 group sightings were successfully (i.e., all individuals identified from good quality photos) obtained during the four-year (2011-2015) study period. In total, 315 individual dolphins (excluding calves) were identified.

Average group size was 5 (SE 0.27) individuals (range: 1-31 individuals) (Table 3.1). Although group size was similar across the three coastal geographic regions, it was smaller in SCR, with an average of four individuals and a maximum group size of 14 individuals.

Table 3.1. Number of groups and group size (mean (SE), minimum and maximum) by geographic region.

# Groups Group size

Mean (SE) Min - Max Overall 323 5 (0.27) 1 - 31

GR 44 7 (0.75) 1 - 31 SCR 107 4 (0.48) 1 - 14 OA 77 6 (0.56) 1 - 24 CS 95 7 (0.51) 1 - 27

Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

70 Chapter 3 - Identifying local population for EIA

3.4.2. Community structure and dynamic

After restricting the dataset to those individuals with at least five sightings, 129 individuals were identified (n = 57 females, 44 males, and 28 unsexed). Both community division using the eigenvector method of Newman (2006) and modularity from gregariousness and hierarchical clustering using average linkage methods indicated a meaningful community division with maximum modularity of 0.514 and 0.526 for an HWI of 0.022 and a cophenetic correlation coefficient (CCC) of 0.843, indicating a good match between the degree of association between individuals and the association matrix. Both methods assigned individuals to four communities, although one community was split into two sub-communities (D and D’). From here on, I refer to these communities as ComA, ComB, ComC and ComD. Three individuals (designated as KWL, GIL, MUF, n = 2.3%) were assigned in different communities depending on the method (Figure 3.2 and see Appendix A3.1). Therefore, we used their respective sighting locations to assign them to one community only (KWL in ComB; GIL and MUF both in ComA).

71 Chapter 3 - Identifying local population for EIA

Figure 3.2. Network diagram for 129 bottlenose dolphins using the HWI. The shape of each node indicates its sex (circle: females; square: males; triangle; unsexed), and the colour of each node indicates its unit defined by the modularity of Newman (2006) (purple ComA; red ComB; green ComC; yellow and orange sub-communities D and D’), although three individuals were assigned to two different units depending on the method (Newman vs. Hierarchical linkage). Only links representing affiliations (HWI > 0.16) are shown, and link width is proportional to index weight. Node size is based on the betweenness centrality measure of each individual.

3.4.3. Association patterns

The overall mean HWI and maximum HWI were 0.05 (SE 0.02) and 0.55 (SE 0.19), respectively. The coefficient of correlation (r) between the true and estimated association indices for the entire study population indicated a moderate representation of the data (r = 0.476, SE 0.024) and a well-differentiated society value (S = 1.020, SE 0.033) suggesting that some individuals form strong associations. I ran this analysis within each community (identified by the network

72 Chapter 3 - Identifying local population for EIA and community analysis) and found values of r higher than 0.4 indicating that my analysis is representative of the true patterns (Table 3.2).

Table 3.2. The mean association indices for the population overall and per community (ComA, ComB, ComC and ComD); the measure of social differentiation (S); and the correlation coefficient of the true and estimated association matrices (r).

n Mean Maximum S r restricted* HWI HWI

Overall 129 0.05 (0.02) 0.55 (0.19) 1.020 (0.033) 0.476 (0.024)

ComA 15 0.21 (0.08) 0.59 (0.20) 0.662 (0.090) 0.653 (0.053) ComB 25 0.17 (0.05) 0.61 (0.24) 0.753 (0.062) 0.742 (0.027) ComC 36 0.19 (0.08) 0.56 (0.16) 0.683 (0.069) 0.712 (0.048) ComD 53 0.13 (0.05) 0.51 (0.16) 0.567 (0.073) 0.577 (0.050) Note: Numbers between brackets () are the standard errors of the respective parameters. *Individuals seen at least five times. Some individuals could not be assigned to a community.

Tests of preferred/avoided associations showed a significantly higher CV of observed vs. expected association indices (HWIobserved CV = 2.1405, HWIrandom CV = 1.8797, P-value < 0.001) indicating that long-term preferred companions are present in the overall population. The proportion of non-zero association indices was significantly lower in the observed vs. the expected association indices (observed = 0.2648, random = 0.3069, P-value < 0.001) indicating avoidance between some individuals in the population. More specifically, individuals were more likely to associate with same-sex individuals than among individuals of different sex (Mantel test, HWIwithin = 0.07, SE 0.03; HWIbetween = 0.05, SE 0.02, P-value < 0.001), with preferred associations occurring between females (HWIobserved CV = 1.9989;

HWIrandom CV = 1.7675, P-value < 0.001) and between males (HWIobserved CV =

2.1770; HWIrandom CV = 1.9222, P-value < 0.0001).

The mean HWI was lower in ComD (0.13, SE 0.05) and higher in ComA (0.21, SE 0.08), although stronger dyads were estimated in ComB (maximum HWI = 0.61, SE 0.24) (Table 3.2). Mantel test confirmed that associations were stronger within than

between communities (HWImean, within = 0.16, SE 0.07; HWImean, between = 0.01, SE

73 Chapter 3 - Identifying local population for EIA

0.01, P-value < 0.0001) and individuals were more likely to associate with same-sex individuals than among individuals of different sex in all communities to the exception of ComA (Table 3.3).

Table 3.3. Association indices within and among sex classes (Mantel test, 0.05 one- side).

HWI mean (SE) P-value Within Between Overall 0.07 (0.03) 0.05 (0.02) < 0.001

ComA 0.20 (0.15) 0.20 (0.05) 0.270 ComB 0.21 (0.07) 0.13 (0.05) < 0.001 ComC 0.23 (0.09) 0.16 (0.08) < 0.001 ComD 0.15 (0.06) 0.12 (0.06) 0.008

Significant differences between communities were found in the network measures with ComC and ComD communities having higher strength and affinity (Table 3.4 and see Appendix A3.2 for Mann-Whitney test, P-value < 0.02), although the strength was much higher for ComA community when considering all individuals (i.e., including individuals seen less than five times, see Appendix A3.3). ComA had the highest clustering coefficient; however, this may be biased by the small number of individuals seen more than five times (n = 15), which may result in individuals appearing more connected than they actually are. ComC had the next highest clustering coefficient, indicating a dense network of individuals in that community.

Preferred associations occurred in all communities with all CVs of association indices higher in the observed vs. the random values (Table 3.5). Specifically, preferences occurred between females within ComB, ComC and ComD and between males within ComB and ComD. Although avoidances occurred with the proportions of non-zero indices being lower in the observed vs. the random values, they were not sex-specific in ComC as found in ComD or for males in ComB (Table 3.5).

74 Chapter 3 - Identifying local population for EIA

Table 3.4. Average strength, clustering coefficients, and affinity (SE) with comparisons from random calculating using half-weight indices for individuals sighted at least five times.

Strength Clustering Affinity coefficient

ComA (n=15) Mean 3.73 (1.11)* 0.25 (0.11) 4.88 (0.90) Random 3.51 (1.11) 0.18 (0.04) 5.14 (0.91)

ComB (n=25) Mean 5.01 (1.43)* 0.19 (0.05) 5.83 (0.97) Random 5.08 (1.50) 0.17 (0.04) 6.04 (0.81)

ComC (n=36) Mean 7.91 (2.94) 0.22 (0.05)* 8.49 (0.78) Random 7.90 (2.93) 0.20 (0.06) 8.59 (0.55)

ComD (n=53) Mean 7.31 (2.73)* 0.19 (0.04) 8.01 (0.86) Random 7.26 (2.78) 0.16 (0.03) 8.02 (0.53) Note: n = number of samples; * Significant differences from 103 random networks: P-value < 0.05.

75 Chapter 3 - Identifying local population for EIA

Table 3.5. Sex class (female, male, unknown sex) permutation tests for preferred (HWI CV) and avoided (proportion of non-zero indices) associations for the population overall and per community.

HWI CV Proportion of non-zero indices na Observed Random P-value Observed Random P-value

All 129 2.14054 1.87585 ** 0.26478 0.30717 **

Females 57 1.99886 1.76774 ** 0.27318 0.32263 ** Overall 44 2.17702 1.92031 ** 0.29704 0.33345 ** Males 28 1.88829 1.84766 ** 0.30423 0.31697 ** Unknown

All 15 0.95857 0.90723 * 0.70476 0.72067 NS

5(2) ------Females ComA (2) Males 2 ------8(2) ------Unknown

All 25 0.99969 0.78870 ** 0.83333 0.85316 *

Females 13 0.74870 0.68623 ** 0.91026 0.89579 NS ComB 12 1.18034 0.98124 *** 0.74242 0.81982 *** Males 0 ------Unknown

All 36 0.93709 0.88541 *** 0.73333 0.75045 **

Females 14 0.80479 0.76055 ** 0.7923 0.77871 NS ComC 11 0.72803 0.72823 NS 0.81818 0.81818 NS Males 11 0.83146 0.83155 NS 0.78182 0.77996 NS Unknown

All 53 1.06854 0.99465 ** 0.60958 0.63421 ***

Females 25 0.96490 0.93202 ** 0.64000 0.66447 ** ComD 19 1.20840 1.04314 *** 0.63158 0.68587 *** Males b Unknown 9 ------Note: n = number of individuals; NS = non-significant; *P-value < 0.05; ** P-value < 0.01; *** P-value < 0.001. a Individuals seen at least five times. b Test could not be run because of degenerate matrix.

76 Chapter 3 - Identifying local population for EIA

Lagged association rates (LAR) for each community were higher than the null LARs for the overall population indicating that associations within communities were relatively stable and non-random over the study period (Figure 3.3). The most parsimonious LAR model (based on the QAICc, see Appendix A3.4) showed constant companions for all communities and brief associations described as rapid or casual but lasting for less than a day. However, the casual acquaintances in ComB lasted for only a few days.

Figure 3.3. Lagged association rates (LAR) for all individuals (black line) and within the communities (ComA purple; ComB red; ComC green; and ComD yellow). The null association rate (dash lines) and jackknife error bars are shown.

Female and male LARs (except in ComA) were higher than their respective null LARs, particularly in ComB, indicating that associations between individuals of the same sex were relatively stable over the study period within their respective communities (Figure 3.4). LARs of males were generally higher than the LARs of females indicating that associations between males were stronger than associations between females, although more females were identified in ComC and ComD (sex ratio 0.71:1 and 0.76:1, respectively). Female and male LARs for ComB were higher than LARs of other communities indicating that associations within ComB were

higher than in other communities (this related only to sexed individuals, Figure 3.4).

77 Chapter 3 - Identifying local population for EIA

Figure 3.4. Lagged association rates (LAR) for males and females (faint colour) bottlenose dolphins of each community ((a)-ComB, (b)-ComC and (c)-ComD). The null association rate (dash lines) and jackknife error bars are shown. Note that the LAR for ComA could not be estimated because of small sample sizes.

78 Chapter 3 - Identifying local population for EIA

The most parsimonious LAR models (based on the QAICc, see Appendix A3.4) for each sex class and per community suggested some long lasting associations and others that were of brief duration because of constant companions and rapid dissociations or casual acquaintances lasting less than a day. However, if not constant, female and male associations in ComB still lasted for up to month or few days, respectively, before dissociating (Appendix A3.4).

3.4.4. Spatial distribution of communities

I estimated core areas (50% kernel density) and home ranges (95% kernel density) of each of the four communities using individuals assigned to each respective community. Core area estimates were similar when using individuals seen at least five times and when including all individuals irrespective of their sighting frequency. Similarly, home range estimates were similar when using individuals seen at least five times and when including all individuals irrespective of their sighting frequency (see Appendix A3.5). As such, I have only presented the results using all clustered individuals. The core areas (i.e., 50% kernel density estimated using all individual sightings) of each community were discrete and located in each geographic region (Figure 3.5a) with sizes varying from 6.83 km2 (ComB) to 31.05 km2 (ComC) (Appendix A3.5). The core area of ComB mainly covered shallow waters (83% coverage at < 10 m) while the core area of ComA was mainly in deep water (71% coverage at > 10 m) (see Appendix A3.6). Home ranges were mainly contained within the respective geographic region (Figure 3.5b), with the home range of ComB mainly covering the shallow waters of the SCR (61.4% coverage at < 10 m). Conversely, the home range of ComA covered much of the GR region, which is mainly deeper waters (55.4% coverage > 15 m). I therefore referred each community to a geographic region, namely ComA to GR; ComB to SCR; ComC to OA; and ComD to CS. Seven individuals (seen less than five times) were not assigned to a community because I could not define the geographic region where they were mainly sighted.

Permutation tests indicated significant avoidance between communities in GR, SCR, and CS, but occurrence of some associations with ComC in OA. However, when

tested with all the individuals (including individuals seen less than five times),

79 Chapter 3 - Identifying local population for EIA avoidance tests were all non-significant. There were overlapping home ranges between each of the communities (Figure 3.5b) with 17% (n = 54 groups) of the groups being composed of individuals from different communities. Most of those multi-community groups (78%) were observed in OA and the lower reach section of SCR (Figure 3.1b). The smallest home range overlap occurred between ComA and ComD (2.42 – 7.96 km2) and the largest between ComC and ComD (43.97 – 45.62 km2) (Appendix A3.8). Additionally, 30 and 32% of ComC and ComD home ranges, respectively, overlapped with the core area of ComB, covering the entire Inner Harbour of the Port of Fremantle.

Inspection of the overlap in home ranges for individuals seen at least five times and assigned to a community indicated that there was a clear difference of percentage of overlap from dyads within and between communities, with much higher proportion of dyads sharing > 80% of the home range within communities (18, 62, 47 and 36% for ComA, ComB, ComC and ComD communities, respectively) than between (only 1.1% of the dyads showed that > 80% of the home range was shared, although this was not necessarily the case for both individuals of the dyad).

Home range overlap significantly explained the HWI dyads at the population-level (R = 0.10, P-value < 0.01) and more specifically for dyads allocated to ComA and

ComD communities (R ComA = 0.07, R ComD = 0.07, P-value < 0.01). Measures of HWI dyads for individuals allocated to ComB and ComC were not explained by their home range overlap (P-value > 0.05).

Chapter 3 - Identifying local population for EIA

Figure 3.5. Study area showing the bathymetry and core areas (a) based on the 50% kernel density and home ranges (b) based on the 95% kernel density estimated for each community using clustered individual sightings. Communities are: ComA-purple; ComB-red; ComC-green; and

ComD-yellow.

Chapter 3 - Identifying local population for EIA 3.4.5. Residence time

I examined the residency patterns of all individuals (including individuals seen less than five times) per community. Models consisting of parameters indicating the occurrence of emigration and mortality best fitted the LIR of each of the four communities (based on the QAIC, Appendix A3.10). Parameters showed differences in community sizes and residency. In ComB, 78% of the individuals were described as residing for nearly 18 years (95% CI 10-87 years). Conversely, 58% of the individuals observed in ComA were described as individuals staying for maximum c. five years (95% CI 3-11 years). Another LIR model represented the demography of

ComA best (i.e., emigration, reimmigration and mortality, ΔQAICc ≤ 2) and showed that a minority of individuals that were considered as a community (n = 20%) also spent more time outside the area than in (Appendix A3.10). Most of the individuals in ComC and ComD (n = 63 and 66%, respectively) were described with a long residence, with individuals from ComC staying for 7.5 years (95% CI 4-23 years) and ComD individuals for about 12 years (95% CI 7-47 years).

3.4.6. Genetic relatedness

A total of 107 tissue biopsy samples were collected for genetic analyses. Samples were from individuals identified during the current study and who had been assigned to a community, and were checked for duplicates and removed when allele frequencies were missing for more than four loci. In addition, three loci were removed for further genetic analyses because of scoring errors due to stuttering (SCA27) or because of departure of HWE associated with homozygosity excess (i.e., frequencies of null alleles > 0.05 for TexVet5 and SCA17).

I identified 35 pairs of high relatedness values (QG > 0.5), although I didn’t have prior information on their relatives for most of these. While most pairs (n = 20) were of individuals assigned to the same communities, others were identified from individuals assigned to different communities, although no pairs involved individuals from ComA and ComD. The bootstrap values of within-community pairwise genetic relatedness coefficients averaged 0.012 (SE 0.005, 95% CI 0.007-

82 Chapter 3 - Identifying local population for EIA

0.017) over the four communities, whereas the average pairwise genetic relatedness coefficient of the population was -0.008 (SE 0.002, 95% CI -0.011, -0.006). All communities except ComD had positive mean genetic relatedness coefficient (Table 3.6) and individuals were more related to individuals from the same community than different communities. However, there were no significant differences in relatedness within and between ComA and ComC (t-test, P-value > 0.05, Table 3.6). Significant correlation between pairwise relatedness coefficients and HWI was found in ComB

(ANOVA, P-value < 0.05, Table 3.7), although the correlation was small (R = 0.01).

Chapter 3 - Identifying local population for EIA

Table 3.6. Bootstrap mean (and standard error) genetic relatedness (Queller and Goodnight, 1989) for within-community (bold) and with individuals from other communities of bottlenose dolphins (between-community) genotyped with ten microsatellite loci in Perth metropolitan waters, WA.

ComA ComB ComC ComD ComA 0.048 (0.020) 0.014 (0.098) *** 0.025 (0.010) NS -0.027 (0.007) *

ComB 0.014 (0.098) NS 0.063 (0.012) 0.003 (0.008) * -0.026 (0.005) *

ComC 0.025 (0.010) NS 0.003 (0.008) *** 0.037 (0.036) -0.030 (0.006) *

ComD -0.027 (0.007) *** -0.026 (0.005) *** -0.030 (0.006) *** -0.012 (0.002) Note: t-tests were performed per column to compare within-community mean to between-community means: *P-value < 0.05; **P-value < 0.01; ***P-value < 0.0001); NS = non-significant.

Table 3.7. Correlation coefficient R and ANOVA test between relatedness coefficient and HWI pairwise within-community.

P-value Correlation R (ANOVA test) ComA -0.0050 0.43 ComB 0.0115 0.04* ComC -0.0005 0.35 ComD 0.0009 0.17

Note: * Significant P-value < 0.05.

Chapter 3 - Identifying local population for EIA

3.5. Discussion

It is imperative to characterise the fine-scale population structure of mobile, continuously distributed species so that an EIA is conducted within an appropriate biological context. The first step in that process is to identify the relevant local wildlife populations that will be affected by a proposed development or activity. This study demonstrated that social, spatial, ecological, and genetic information may be used to identify local communities of bottlenose dolphins.

Analyses of association patterns of photo-identified bottlenose dolphins revealed four distinct social mixed-sex communities in which patterns of social organisation, social dynamics, home ranges, and residency differed (see Table 3.8 for a summary). Spatial analyses found these communities occupy discrete core areas associated with different environmental and bathymetric characteristics. As expected, high site fidelity and residency patterns were documented for communities occupying shallow, protected embayment and estuary habitats. Overlap in home ranges (e.g., dyads within ComA-GR and ComD-CS) and genetic relatedness (e.g., dyads within SCR community) explained associations between some dyads. However, overall, such factors explained only little of the associations between individuals, suggesting that other explanatory factors drive community structure at a local scale.

85 Chapter 3 - Identifying local population for EIA

Table 3.8. Summary comparison of social, temporal, spatial, residency and genetic patterns across the four communities (ComA, ComB, ComC and ComD). Social differentiation is described using measures of strength (i.e., measure of gregariousness); the clustering coefficient (i.e., degree of connection between associates); and affinity (i.e., the strength of the associates).

Parameters ComA ComB ComC ComD Group size Large Small Large Large Population size Large Small Medium Medium

Social network measures

Strength Weak Medium Strong Strong Eigenvector centrality Weak Weak Strong Medium Reach Weak Medium Strong Strong Clustering coefficient Strong Medium Strong Medium Affinity Weak Medium Strong Strong

Temporal associations

High High Medium Medium LAR index Stable Stable Stable Stable Maximum duration of < day days-month < day < day shortest associations

Spatial distribution

Core area - Home range GR SCR OA CS Water depth Deep Shallow Mixed Mixed

Residency

Site fidelity Weak Strong Strong Strong

Very long- Duration Short-term Long-term Long-term term Status Transient Resident Resident Resident

Genetic relatedness a No kin- Kin-selection Weak Weak Weak selection Note: a kin-selection measurement should be used with caution because of the limited number of individuals genetically sampled in ComA, ComC and ComD. GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

86 Chapter 3 - Identifying local population for EIA

3.5.1. Community segregation

Dolphins exhibited a complex structure of associations across the coastal and estuarine seascape of the study area. Network analysis identified four social mixed- sex communities which have few interactions between them. While some constant companionship relationships were identified in all communities, each community is driven by casual acquaintance relationships demonstrating rapid disassociation and frequent re-association (Wells et al. 1987). However, even for constant companions, the shortest associations between individuals within ComB (i.e., SCR community) lasted for a few days to a month between females, suggesting stronger maternal cooperation (Wells et al. 1987; Lusseau et al. 2003) in the estuary. In fact, the entire ComB (females and males) presented a much higher degree of stability (LAR) than occurred in any of the other communities; this may reflect the ecological constraints associated with estuarine systems (Lusseau et al. 2003) and human pressures.

Social segregation of dolphin populations within nearshore and inshore habitats has been documented in Indo-Pacific bottlenose dolphin elsewhere. For example, discrete communities occur in the Port Stephens embayment in south-eastern Australia, with two communities occupying spatially discrete core areas within the embayment (Wiszniewski et al. 2009). Here, despite some home ranges overlapping (95% KDE), each community was associated with a distinct geographic region within the study area, namely: an estuary (SCR – ComB), a semi-enclosed embayment (OA – ComC and CS – ComD), and an open coastline (GR – ComA). The core areas (50% KDE) of the communities did not overlap, suggesting that ecological differences among communities reflect environmental differences in bathymetry, benthic substrate, habitat types, as well as human impacts.

Communities also differed in their sociality. Titcomb et al. (2015) demonstrated that habitat shape could have an effect on the structure and association patterns within communities, by influencing movement patterns and encounters between conspecifics. Here, the densest communities in the network, OA (ComC) and CS (ComD) (i.e., indicated by higher strength and affinity), were found in a semi- enclosed embayment. As the SCR community (ComB) occupies an estuary with narrow channels, encounters between conspecifics may occur on a daily basis.

87 Chapter 3 - Identifying local population for EIA

However, the mean sociality for that community was not as high, which may reflect the limited number of possible associates given its small size (n = 25, excluding two individuals seen less than five times). Conversely, the GR community (ComA) was the least cohesive, with high redundancy in connection expressed by lower strength, which is consistent with individuals occupying larger ranges and, consequently, less frequent encounters with conspecifics.

The LIR analysis identified clear differences in site fidelity and residency pattern between communities. These differences may be related to the difference in habitat structure and prey distribution between open coastline (GR) and more protected embayment and estuary (OA & CS and SCR) habitats. Individuals in open coastlines often have diminished levels of site fidelity and a more extensive home range (Defran and Weller 1999; Oudejans et al. 2015; Sprogis et al. 2015). Here, individuals from GR community (ComA) showed no residency pattern, with most identified individuals seen less than five times and LIR models indicating that individuals spent more time outside the study area. In addition, the large but variable estimates of abundance (see Chapter 2, Chabanne et al. 2017b) suggested that individuals identified in this geographic region (GR) may be members of a larger population located further north. In contrast, residency period was estimated to be more than seven years in OA and CS and 18 years in SCR, which is consistent with other studies indicating that dolphins occupying shallow and protected areas show a high degree of residency and long-term site fidelity (Wells 1986; Sprogis et al. 2015; Brusa et al. 2016) and often belong to relatively small and stable communities (Wells et al. 1987). That latter characteristic is consistent with the small size of the resident communities estimated by the LIR model (SCR ≈ 21 residents; OA ≈ 43; and CS ≈ 64) and the abundances estimated via mark-recapture analyses (SCR ≈ 19 individuals and CS/OA (combined) ≈ 122; see Chapter 2, Chabanne et al. 2017b).

While habitat differences seemed to largely account for association differences, range overlap only weakly predicted the association strength of the two communities at the northern and southern extremes (GR and CS). Likewise, genetic relatedness only explained associations within the SCR community. Kin-based and overlapping associations have been recorded in some bottlenose dolphin populations (Parsons et al. 2003; Frère et al. 2010b), but not in others (Möller 2001). It would therefore

88 Chapter 3 - Identifying local population for EIA appear that other intrinsic or extrinsic factors are likely to drive the social patterns within each community. The KDE method used to calculate the habitat used by each individual is also limited by the sample size (i.e., number of observations per individual). Sprogis et al. (2015), for example, calculated the KDE for individuals seen at least 30 times (Seaman et al. 1999), although other studies indicated that 100- 300 observations per individual were necessary to obtain highly precise representation of space use (Girard et al. 2002).

3.5.2. Using social, ecological, and genetic data to evaluate the impact of developments and activities on the relevant local population: four case studies

Once the relevant local wildlife populations have been identified, social, spatial, ecological, and genetic information about those local populations can then be used to evaluate potential environmental impacts on those populations. The range of impacts that may affect bottlenose dolphins in coastal and estuarine habitats include: habitat degradation, indirect and direct interactions with commercial and recreational fisheries, vessel disturbance, and environmental contaminants (e.g. Dungan et al. 2012; Pirotta et al. 2013; Todd et al. 2015; Culloch et al. 2016).

To demonstrate the utility of such information for evaluating environmental impacts, I consider four case-study examples of developments or activities that may affect dolphins in the context of one of the local communities identified in this study: pile driving (SCR), dredging of seagrass (OA), operation of a desalination plant (GR), and construction of a large harbour (CS).

3.5.2.1. Pile driving in Swan Canning Riverpark (SCR)

Pile driving involves the use of large hammer mounted on a crane to drive piles into the seabed. The process may affect marine mammals by masking underwater sounds, causing behavioural changes (e.g., avoidance), or causing hearing damage or physiological injury (David 2006; Brandt et al. 2011; Erbe 2013).

Paiva et al. (2015) observed a decrease in detection of bottlenose dolphins during pile driving activities in the Inner Harbour at Fremantle (located at the entrance to

89 Chapter 3 - Identifying local population for EIA

SCR), and suggested that dolphins may have been using other areas during periods of pile driving activity. If such avoidance or displacement behaviour occurs, then pile driving may affect dolphins in two relevant ways: (a) a decline or cessation in foraging activity within the harbour, which is a known foraging habitat (Chabanne et al. 2012) and (b) a decline or cessation in the use of the harbour to transit between the estuary and adjacent coastal waters. The latter impact is particularly significant as the harbour is the only access way between the SCR and the adjacent coastal waters. Were movements of dolphins through the harbour to cease or greatly diminish over an extended period of time, then demographic isolation of the SCR community might occur.

Several characteristics of the SCR community make it relatively more vulnerable to long-term demographic isolation and would make local extinction a plausible risk, including: (1) small community size; (2) a recent disease-related mass mortality event (Holyoake et al. 2010); (3) injury and mortality from fishing line entanglement; and (4) exposure to high levels of boat traffic and to occasional harassment.

3.5.2.2. Dredging in Owen Anchorage (OA)

Dredging of marine habitats may occur to create or maintain infrastructure (e.g., shipping channels) or to remove benthic material such as shellsand for commercial purposes. Todd et al. (2015) reviewed the effects of marine dredging activities on marine mammals and concluded that (a) direct impacts (such as vessel collisions and underwater noise emissions) were unlikely because vessel speeds were slow and (b) the low-frequency levels (below 1 kHz) emitted by dredgers should not cause damage to marine mammal auditory systems. However, underwater noise from dredging may mask prey sounds and dolphin vocalisations and lead to displacement (Pirotta et al. 2013), particularly if activities directly impact on marine mammals prey species (Todd et al. 2015). Dolphins may also be attracted to dredging sites if the disturbance facilitates the capture of fish (e.g., Chilvers and Corkeron 2001).

A long-term shellsand dredging operation operates in Owen Anchorage which relies on dredging of suitable substrates (Environmental Protection Authority 2001; BMT

90 Chapter 3 - Identifying local population for EIA

Oceania 2014). The extensive coverage of shallow (< 10 m) sand areas and seagrass meadows (BMT Oceania 2014) sheltered from the oceanic swell in OA means the area is likely to support a broad assemblage of prey species for dolphins (Kendrick et al. 2000; Heithaus and Dill 2002; Hyndes et al. 2003; Finn 2005; Sampey et al. 2011). The current management plan for the dredging operation, developed to meet the requirements of approval conditions imposed in 2002 after an EIA of the operation, focuses on the dredging of areas devoid of seagrass to minimise environmental impacts to benthic habitats and fisheries (BMT Oceania 2014).

The focus on dredging of non-seagrass areas and the overall scale of the dredging operation suggest that impacts on prey availability for dolphins will be localised. Further, impacts from interactions with dredging and transport vessels are unlikely to present a significant risk, as vessel speeds are slow and dolphins do not appear to be attracted to active dredging operations. The OA community identified in this study would be the relevant local population for any EIA of any future proposal to expand the current shellsand dredging operation.

3.5.2.3. Desalination in Gage Roads (GR)

Impact assessments for the operation of desalination plants in southern Australia have reported low risks of impacts for marine mammals (e.g., Wonthagii, Victoria, Minister for Minister for Planning 2009) (e.g., Cape Riche or Binningup, southern Western Australia, Water Water Corporation 2008; Bejder 2011). However, direct and indirect impacts from brine discharges to the benthic environment (and subsequently to local fauna populations) remain unknown in these areas (Bejder 2011). In Binningup, for example, prey availability may have been reduced indirectly from osmoregulation impacts to fish (e.g., cuttlefish, Sepia apama, Dupavillon and Gillanders 2009; Smith and Sprogis 2016) or destruction of fish habitats. Such impacts have been reported elsewhere. In Alicante Bay in the northwestern Mediterranean Sea, for example, a seagrass die-off resulted from physiological stress caused by salinity fluctuations associated with brine discharge from two desalination plants (Garrote-Moreno et al. 2014). Physiologically, dolphins and other marine mammals are highly-evolved osmoregulators, with a kidney structure developed for habitats with a broad salinity range, indicating that higher

91 Chapter 3 - Identifying local population for EIA localised salinities should not cause significant physiological stress if exposure to extreme conditions is not prolonged (Ortiz 2001).

The ecological characteristics of the GR community, principally low site fidelity and more transient behaviour, suggest the operation of a proposed desalination plant in the northern metropolitan waters of Perth (as has been discussed -- see Mercer (2013)) would be unlikely to have as adverse an impact as might occur for a community showing strong site fidelity and near continuous occupancy of the affected area. Nonetheless, environmental change may induce displacement (Dungan et al. 2012) or splitting (Nishita et al. 2015) of the community, which may have adverse ecological impacts because some individuals are essential for maintaining the cohesion of the network (i.e., metapopulation) and controlling the flow of information within it (Lusseau and Newman 2004).

3.5.2.4. Harbour construction in Cockburn Sound (CS)

Two harbour developments have been proposed for the Kwinana Shelf region in CS, a private port and a new Outer Harbour facility for the Port of Fremantle. The construction of harbour facilities presents a range of risks for dolphins, including reduction or displacement of dolphins because of direct and indirect impacts from construction-related activities (e.g., Pirotta et al. 2013; Todd et al. 2015; Culloch et al. 2016). Potential impacts on dolphins include but are not limited to: (1) disturbances or changes in behaviour from construction noise and vibration; (2) displacement due to a change in prey availability resulting from modification or removal of habitat because of dredging or the construction of infrastructure; and (3) health issues arising from changes in water quality (e.g., sedimentation and increased incidence of algal blooms), circulation patterns (e.g., reduced flushing), and increased chemical contaminants (Environmental Protection Authority 1998).

Here, the risks of a harbour development are significant because the core area of the CS community is located in the Kwinana Shelf, an area that has ecological significance for dolphins as foraging habitat and as a nursery area (Finn 2005; Finn and Calver 2008). In particular, the loss of nursing habitat for females and calves may impose substantial fitness costs on individual dolphins through reduced

92 Chapter 3 - Identifying local population for EIA reproductive success. The strong, long-term spatial association of dolphins with the area and the absence of habitat with environmental characteristics similar to the Kwinana Shelf suggest that dolphins would not be able to compensate for the loss of habitat on the Kwinana Shelf by shifting to other, nearby areas. The extent to which dolphins use new harbour facilities for foraging may depend on the harbour designs, the materials used, and the fish assemblages those areas ultimately sustain.

3.5.3. Conclusions

This study emphasises the need for EIAs to focus on the relevant local wildlife populations that will be affected by proposed developments and activities. Here, I applied a methodology of broad utility to identify multiple communities of bottlenose dolphins in a heterogeneous coastal environment and then used information on their social and spatial structures, residency patterns, and abundances used to assess the vulnerability of each community to a particular environmental impact. Such results could also be informative for Marine Spatial Planning and Cumulative Impact Mapping. One local population, the SCR community, appeared to be at some risk of local extinction because of its small size and reliance on an estuary which is only connected to adjacent coastal waters by a heavily-used harbour area.

While mobile marine fauna such as bottlenose dolphins may range over large areas of ocean or coastline, they may also exhibit fine-scale population structure reflecting long-term residency, strong site fidelity, limited ranging patterns and strong, long- term associations with particular conspecifics. Other species may exhibit short-term residency in defined coastal areas, e.g., for breeding or feeding. The proper evaluation of impacts of coastal and estuarine developments therefore requires information about the distribution of species at an individual level (i.e., spatial and temporal scales) and their connection at community level (i.e., metapopulation dynamics).

93 Chapter 3 - Identifying local population for EIA

Appendices

Appendix A3.1. Social dendrogram

Figure A3.1.1. Dendrogram showing average linkage cluster analysis of bottlenose dolphins sighting at least five times in the Perth metropolitan waters (n = 129 individuals, excluding calves). The HWI of 0.022 indicated the best cut-off value for forming four groupings based on the maximum modularity (M = 0.526).

94 Chapter 3 - Identifying local population for EIA

Appendix A3.2. Tests of differences in distribution of centrality values between communities

Table A3.2.1. P-value of the Mann-Whitney tests showing the significant differences in distribution of centrality values (strength, clustering coefficient and affinity) between communities (mean metric values were calculated for individuals sighted at least five times).

ComA ComB ComB ComA ComA ComC vs. vs. vs. vs. vs. vs. ComB ComC ComD ComC ComD ComD

Strength < 0.01 < 0.001 < 0.001 < 0.001 < 0.001 NS Clustering < 0.001 < 0.001 < 0.001 NS NS < 0.001 coefficient Affinity < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 NS

Note: NS = non-significant (P-value > 0.05).

95 Chapter 3 - Identifying local population for EIA

Appendix A3.3. Network parameters

Table A3.3.1. Average strength, clustering coefficients and affinity (SE) with comparisons from random calculating using half-weight indices (including individuals sighted less than five times).

Strength Clustering Affinity coefficient ComA Mean (n = 117) 11.92 (7.96)* 0.46 (0.20)* 13.23 (6.42) Random 11.88 (7.98) 0.44 (0.20) 13.46 (6.20)

ComB Mean (n = 27) 3.72 (1.50)* 0.24 (0.03) 4.30 (0.28) Random 3.70 (1.52) 0.28 (0.04) 4.32 (0.20)

ComC Mean (n = 68) 6.37 (3.85)* 0.27 (0.18) 7.74 (2.18) Random 6.36 (3.89) 0.26 (0.18) 8.10 (1.93)

ComD Mean (n = 96) 6.65 (3.83)* 0.22 (0.10)* 8.13 (2.05) Random 6.62 (3.89) 0.19 (0.10) 8.32 (1.83) Note: n = number of samples; Significant differences from 1000 random networks: P-value < 0.05.

Chapter 3 - Identifying local population for EIA

Appendix A3.4. Best-fitted models for the lagged association rate (LAR)

Table A3.4.1. Best-fitted models for the LAR of each network cluster (ComA, ComB, ComC and ComD), regardless of the sex class. Community Best-fitted models QAIC ΔQAIC Pref. comp % Casual % Time casual

36.41% Rapid dis. + pref. comps 538.7228 0 ComA (SE 0.06) (n = 15) 36.41% 0.07days Pref. comps + casual acqs 540.7228 2 63.59% (SE 0.06) (SE 0.09 days) 36.50% 2.84 days Pref. comps + casual acqs 7385.5975 0 63.50% ComB (SE 0.04) (SE 2.69 days) (n = 25) 36.42% 21.34% 7.07 days Rapid dis. + pref. comps + casual acqs 7386.1142 0.5167 (SE 0.04) (SE 1126278.48) (SE 0.21 days) ComC 26.91% Rapid dis. + Pref. comps 1451.1593 0 (n = 36) (SE 0.05) 19.19% Rapid dis. + pref. comps 2372.1672 0 (SE 0.03) ComD 20.26% 9696.50 days Rapid dis. + casual acqs 2373.3551 1.1879 (n = 53) (SE 0.04) (SE 4,279.91 days) 19.20% 0.03 days Pref. comps + casual acqs 2373.9162 1.7490 80.80% (SE 0.03) (SE 0.04 days)

97 Note: n = number of samples; Rapid dis. = Rapid dissociation; Pref. comps = Preferred companions; Casual acqs = casual acquaintances.

Chapter 3 - Identifying local population for EIA

Table A3.4.2. Best-fitted models for the LAR of females for each network cluster (ComA, ComB, ComC and ComD).

Community Best-fitted models QAIC ΔQAIC Pref. comp % Casual % Time casual

ComB 31.40% 16.49% 33.32 days Rapid dis. + pref. comps + casual acqs 2897.1046 0 (n = 13) (SE 0.05) (SE 91324806) (SE 0.15 days)

24.84 % Rapid dis. + pref. comps 810.6382 0 ComC (SE 0.05) (n = 14) 21.84% 0.04 days Pref. comps + casual acqs 812.6381 1.9999 78.16% (SE 0.05) (SE 0.03 days) 18.53% Rapid dis. + Pref. comps 557.2738 0 ComD (SE 0.04) (n = 25) Pref. comps + casual acqs 559.2738 2 18.52% 0.16 days 81.48 % (SE 0.04) (SE 0.49 days)

Note: n = number of samples; Rapid dis. = Rapid dissociation; Pref. comps = Preferred companions; Casual acqs = casual acquaintances.

100 98

Chapter 3 - Identifying local population for EIA

Table A3.4.3. Best-fitted models for the LAR of males for each network cluster (ComA, ComB, ComC and ComD).

Community Best-fitted models QAIC ΔQAIC Pref. comp % Casual % Time casual

58.34% 3.64 days Pref. comps + casual acqs 4166.3721 0 41.66% (SE 0.04) (SE 0.12 days) ComB 58.58% Rapid dis. + pref. comps 4167.2745 0.9024 (n = 12) (SE 0.04) 58.34% 33.28% 4.19 days Rapid dis. + pref. comps + casual acqs 4168.3165 1.9444 (SE 0.04) (SE 1613350) (SE 0.20 days) 27.83% Rapid dis. + pref. comps 253.1170 0 (SE 0.06) ComC 31.79% 3625.55 days Rapid dis. + casual acqs 254.5568 1.4398 (n =11) (SE 0.15) (SE 1498.51 days) 27.83 % 0.07 days Pref. comps + casual acqs 255.1170 2 72.17% (SE 0.06) (SE 1.05 days)

30.27% Rapid dis. + pref. comps 1124.9691 0 ComD (SE 0.06) (n = 19) 1126.9691 2 30.27% 0.04 days Pref. comps + casual acqs 69.73% (SE 0.06) (SE 0.04 days) Note: n = number of samples; Rapid dis. = Rapid dissociation; Pref. comps = Preferred companions; Casual acqs = casual acquaintances.

Chapter 3 - Identifying local population for EIA

Appendix A3.5. Core areas and home ranges

Table A3.5.1. Core area (50% kernel density, km2) and home range (95% kernel density, km2) of each community using all individuals sighted or only individuals sighted more than five times (restricted).

Core area Home range All Restricted All Restricted

ComA (GR) 19.23 23.67 167.11 172.93

ComB (SCR) 6.83 6.91 66.67 66.65

ComC (OA) 31.05 27.9 137.5 131.04

ComD (CS) 24.35 23.75 125.79 123.48 Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

100 Chapter 3 - Identifying local population for EIA

Appendix 3.6. Bathymetry

Table A3.6.1. Depth coverage (%) of the restricted core areas (50% kernel density) and home ranges (95% kernel density, km2) of each community. Depth classes are: shallower than 5 m; between 5 and 10 m; between 10 and 15 m; and deeper than 15 m.

Core area Home range < 5 5-10 10-15 > 15 < 5 5-10 10-15 > 15 ComA (GR) 40.8 42.1 13.2 3.9 16.4 28.2 31.0 24.4 ComB (SCR) 26.8 2.4 41.5 29.3 26.8 34.6 22.1 16.5 ComC (OA) 19.0 35.0 38.0 8.0 14.4 24.1 24.3 37.2 ComD (CS) 7.4 18.5 33.4 40.7 14.4 20.8 26.8 38.0 Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

101 Chapter 3 - Identifying local population for EIA

Appendix A3.7. Overlap between social communities

Table A3.7.1. Area (km2) of overlap between the four social communities.

From Into GR SCR OA CS ComA (GR) - 55.55 7.96

ComB (SCR) 18.83 - 33.90 14.49 ComC (OA) 33.92 39.93 - 45.62 ComD (CS) 2.42 12.79 43.97 - Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

102

Chapter 3 - Identifying local population for EIA

Appendix A3.8. Best-fitted models for the lagged identification rate (LIR)

Table A3.8.1. Best-fitted models for the LIR for each network cluster (ComA, ComB, ComC and ComD). The 95% confidence intervals are given in italic.

Community size Mean time in Mean time out Mortality Community Best-fitted models QAIC ΔQAIC (N) (days) (days) 68 2061.88 ComA Emigration + mortality 2731.0889 0 60-85 1253.64-3971.52 (n = 117) Emigration + reimmigration 23 9.99 21.17 0.0004 2731.8408 0.6274 + mortality 8-39 4.63-18.18 10.62-41.91 0.0001-0.0007 ComB 21 6773.32 Emigration + mortality 34521.2798 0 (n = 27) 18-24 3685.52-31792.77 ComC 43 2743.37 Emigration + mortality 14181.2691 0 (n = 68) 36-50 1504.16-8333.60 ComD 64 4485.11 Emigration + mortality 21842.4180 0 (n = 96) 56-73 2424.79-17045.43 Note: n = number of individuals.

103

104

Chapter 4. Genetic structure of socially and spatially discrete subpopulations of dolphins (Tursiops aduncus) in Perth, Western Australia

4.1. Abstract

Populations of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in Perth metropolitan waters (Western Australia) are spatially segregated into four discrete social subpopulations, although the level of genetic differentiation within the population is unknown. Contemporary and historical genetic differences among the four socio-geographic subpopulations were assessed using ten microsatellite loci and mitochondrial DNA control region sequence data. Pairwise estimates of genetic differentiation (FST) based on microsatellite alleles revealed some significant differences between the socio-geographic subpopulations, also supported by little or no contemporary gene flow. However, differences were too weak and recent to assign individuals to their original subpopulations using the Bayesian clustering implemented in STRUCTURE. Historically, individuals from the socio-geographic subpopulations originated from two ancestral lineages, although no apparent geographic clustering was detected. The Tajima’s D analysis suggested the occurrence of a historical bottleneck event or selection in at least one socio- geographic subpopulation (Cockburn Sound) associated with a semi-enclosed embayment, although none of the other bottleneck tests agreed. Likewise, high genetic diversity was maintained by moderate asymmetric gene flow (m > 0.10) and could have erased any bottleneck or selection signal. Within this source-sink dynamic, it is important to conserve the source subpopulation as other subpopulations depend on it. The estuarine subpopulation, which acts as a sink, also warrants particular attention for conservation and management because of potential inbreeding and vulnerability to extinction because of its small population size and impacts from multiple anthropogenic threats.

105 Chapter 4 – Genetic structure of socio-geographic subpopulations

4.2. Introduction

Understanding population structure is a challenging yet vital component of wildlife conservation, particularly for continuously distributed marine taxa like cetaceans (Baker et al. 1998; Waples 1998; Thompson et al. 2016). Low levels of population genetic structure are generally expected for cetaceans because of the general lack of geographical barriers to gene flow in marine environments and the high dispersal capacity of cetaceans (Fontaine et al. 2007; Thompson et al. 2016). Nonetheless, clear genetically discrete population units often occur for many cetacean species, even at fine geographical scales (Hoelzel et al. 1998b; Sellas et al. 2005; Andrews et al. 2010; Fernández et al. 2011; Möller et al. 2011; Ansmann et al. 2012; Brown et al. 2014). Genetic differentiation between populations may be a result of isolation by distance or be related to physical features or geographic separation (e.g., Krützen et al. 2004). However, it can also be attributable to complex behavioural factors associated with ecological and environmental processes, such as local resource specialisations, philopatry or social organisation (Gaggiotti et al. 2004; Möller et al. 2007; Andrews et al. 2010; Wiszniewski et al. 2010; Ansmann et al. 2012; Kopps et al. 2014).

For bottlenose dolphins (Tursiops spp.), the geographical scale at which genetic structure appears is highly dependent on the nature of the surrounding environment (Wiszniewski et al. 2010). Little differentiation has been observed (with both nuclear and mitochondrial DNA markers) in large pelagic populations (e.g., Quérouil et al. 2007). By contrast, clear genetic differentiation may occur within populations occupying coastal habitats, despite some reproductive exchange (Sellas et al. 2005; Möller et al. 2007; Rosel et al. 2009; Tezanos-Pinto et al. 2009; Urian et al. 2009; Mirimin et al. 2011). Such differentiation is often related to the coastal and estuarine habitats present, e.g., open coastlines, embayments, lagoons, sounds, tidal marshes and river systems (Sellas et al. 2005; Richards et al. 2013). Where a number of coastal and estuarine habitats are present, and environmental heterogeneity is high, the possibility for fine-scale population subdivision increases (Krützen et al. 2004; Wiszniewski et al. 2010; Ansmann et al. 2012).

106 Chapter 4 – Genetic structure of socio-geographic subpopulations

Strong site fidelity, with resident subpopulations inhabiting estuaries and bays as a consequence of different social systems and behavioural strategies, is a factor that may lead to genetic structure (Hoelzel et al. 1998b; Parsons et al. 2006; Wiszniewski et al. 2009). Social structure can play a critical role in shaping genetic variability within and between populations since it can determine patterns between related and unrelated individuals in space and time (Sugg et al. 1996; Storz 1999). High philopatry and long-term social affiliations between females may result in non- random mating and reduced gene flow among inshore subpopulations, as will a moderate level of male philopatry (Connor et al. 2000a; Krützen et al. 2004; Möller and Beheregaray 2004; Möller et al. 2006).

Dedicated line-transect photo-identification surveys revealed that the population of Indo-Pacific bottlenose dolphins (Tursiops aduncus) inhabiting the coastal and estuarine waters of Perth (Western Australia) consists of small subpopulations that are geographically, socially and demographically differentiated (see Chapters 2 and 3, Chabanne et al. 2017a, b). Three socially stable subpopulations exhibit long-term residency in different habitats, namely an estuary and an embayment. A fourth one, along an open coastline, was described as a transient subpopulation with characteristics suggesting that only a small portion of a larger population located outside the boundaries of the study area has been identified. Despite strong social and spatial segregation between subpopulations, sightings of mixed groups also occurred, suggesting the possibility for interbreeding between subpopulations which would minimise local genetic differentiation. Nonetheless, while an interbreeding scenario may be the most plausible, Sellas et al. (2005), for example, found significant genetic differentiation between Sarasota Bay resident dolphins and nearshore coastal Gulf of Mexico dolphins just outside the Bay with only a small amount of interbreeding, despite sightings of mixed groups.

Understanding whether bottlenose dolphins in the Perth metropolitan area represent several distinct subpopulations that reflect their social and spatial segregation, or a single genetic population, will play a major role in defining management requirements and planning for the conservation of the species in this region. Findings will be extremely valuable since an unusual mortality event occurred in the Swan Canning Riverpark in 2009, with the death of six bottlenose dolphins within a six-

107 Chapter 4 – Genetic structure of socio-geographic subpopulations month period (Holyoake et al. 2010). To this end, I investigated the genetic differentiation and diversity among and within the identified socio-geographic subpopulations (see Chapters 2 and 3, Chabanne et al. 2017a, b) using microsatellite DNA markers and a 416-base pair (bp) fragment of the mitochondrial (mt) DNA control region. Understanding the genetic structure of this population will complement the social and spatial segregation findings and aid in making appropriate decisions for management and enhance conservation of the local subpopulations that are highly impacted by human activities.

4.3. Materials and methods

4.3.1. Genetic sample collection

Systematic boat-based surveys for Indo-Pacific bottlenose dolphins in Perth metropolitan waters, Western Australia, (Figure 4.1) were carried out from June 2011 to May 2015 following a mark-recapture design approach (see Chapter 2, Chabanne et al. 2017b). Analyses of the social structure based on those surveys and photo-identification defined a population subdivided into four socio-ecological subpopulations (see Chapter 3, Chabanne et al. 2017a). Spatial segregation was also described with each subpopulation inhabiting a particular geographic region in the study area: GR – Gage Roads, which represents an open coastline; SCR – Swan Canning Riverpark, which is an enclosed estuary; OA – Owen Anchorage, which is the northern section of a large semi-enclosed embayment that is mainly shallow; and CS – Cockburn Sound, which is the southern section of the large semi-enclosed embayment with deeper waters.

108 Chapter 4 – Genetic structure of socio-geographic subpopulations

Figure 4.1. Map of the four geographic regions (in bold) associated with each socio- ecological subpopulation within the Perth metropolitan waters, Western Australia.

During systematic surveys and other opportunistic surveys, biopsy samples were collected using the PAXARMS remote biopsy system specifically designed for small cetaceans (Krützen et al. 2002). A few individuals observed between 2011 and 2015 were biopsied during opportunistic surveys conducted between 2007 and 2011. Calves assumed to be less than two-year-old (i.e., based on body length and approximate date of birth) were excluded from biopsy sampling. Tissue samples were preserved in saturated NaCl 20% dimethyl sulfoxide until DNA extraction.

109 Chapter 4 – Genetic structure of socio-geographic subpopulations

4.3.2. DNA extraction and sexing

Genomic DNA was extracted from all skin samples using the Gentra Puregene Tissue Kit (Qiagen) and following the manufacturer’s protocol. The sex of sampled individuals was determined using fragments of the ZFX and SRY genes (Gilson et al. 1998) that were amplified using polymerase chain reaction (PCR) with 20-25 ng DNA, 0.15 μl of each primer (ZFX forward and reverse and SRY forward and reverse) and standard PCR reagents. The PCR profile consisted of initial denaturation at 95°C for 4 min, followed by 35 cycles of 94°C for 45 s, 50°C for 45 s and 72°C for 10 min. PCR products were separated on agarose gel to determine sex base on length differences.

4.3.3. Genotyping and validation of microsatellites

Four primers were used to optimize a total of 13 microsatellite loci: DIrFCB4, DIrFCB5 (Buchanan et al. 1996), LobsDi_7.1, LobsDi_9, LobsDi_19, LobsDi_21, LobsDi_24, LobsDi_39 (Cassens et al. 2005), SCA9, SCA17, SCA22, SCA27 (Chen and Yang 2008), TexVet5, TexVet7 (Rooney et al. 1999). I followed the PCR conditions as described in Frère et al. (2010a). Single stranded PCR products were run on an ABI 3730 DNA Sequencer (Applied Biosystems). Using GENEIOUS 9.1 (http://www.geneious.com; Kearse et al. 2012) with the microsatellite plugin 1.4 (Applied Biosystems), bins for each locus were determined, and genotypes scored. Each microsatellite locus was checked for scoring errors using the software Micro- Checker 2.2 with a confidence level of 95% (Van Oosterhout et al. 2004). Samples that matched in sex and microsatellite genotypes were considered duplicates, and only one of each was retained for analyses. Departures from Hardy-Weinberg equilibrium (HWE) and linkage disequilibrium were tested using the Markov chain probability test and 104 iterations in Genepop 4.4 (Rousset 2008). Significance values for multiple comparisons were adjusted by sequential Bonferroni corrections (Rice 1989). I used the software INEST 2.0 (Inbreeding/Null Allele Estimation; Chybicki and Burczyk 2009) to check whether any departure from HWE at a given locus might be explained by the presence of null alleles or by inbreeding and estimated their frequencies using the population inbreeding model. The probability of identity (PI) was calculated using GenAlEx 6.3 (Peakall and Smouse 2006) to

110 Chapter 4 – Genetic structure of socio-geographic subpopulations assess the discriminatory power of the set of microsatellite loci and calculate the probability that two different individuals could share the same genotype.

4.3.4. Mitochondrial (mt) DNA sequencing

Primers dlp1.5 (5′-TCA CCC AAA GCT GRA RTT CTA-3′) and dlp5 (5′-CCA TCG WGA TGT CTT ATT TAA GRG GAA-3′) (Baker et al. 1993) were used to amplify a 412-bp mitochondrial fragment following PCR conditions described in Bacher et al. (2010). Sequences of the mtDNA were manually edited using the program GENEIOUS 9.1.

4.3.5. Assessment of genetic differentiation

Sampled individuals were assigned into their respective socio-geographic subpopulation defined in Chapter 3 (Chabanne et al. 2017a). Genetic differentiation between socio-geographic subpopulations was investigated by calculating pairwise genetic distances FST between microsatellite alleles and between mtDNA haplotypes (Weir and Cockerham 1984). With mtDNA data, I also estimated the nucleotide differentiation φST (Tamara and Nei, 1993) using the program Arlequin 3.5 (Excoffier and Lischer 2010). Significance levels were tested using 104 permutations (AMOVA). Significance values for multiple comparisons were adjusted by sequential Bonferroni corrections (Rice 1989). For both marker sets, I also estimated s pairwise Shannon’s Mutual Information index, HUA using GenAlEx 6.3 (Sherwin et al. 2006) for which tests of significance are performed by random permutations (n = 999 permutations) rather than by the more conservative chi-square test.

4.3.6. Assessment of genetic population structure

To verify the hypothesis that social structure reflects the genetic structure, I used the software STRUCTURE 2.3 (Pritchard et al. 2000) to determine the most likely number of distinct nuclear genetic clusters (K) and compare the genetic cluster assignment with the social assignment of each individual. The software uses a Markov chain Monte Carlo (MCMC) procedure to estimate the mean log probability (LnP(D)) that

111 Chapter 4 – Genetic structure of socio-geographic subpopulations the data fit the hypothesis of K clusters. I ran the analysis using an admixture model (with or without prior information on social subpopulation), and all models were run with correlated allele frequencies (recommended by Falush et al. 2003); a burn-in period set to 105 following by 107 MCMC steps and was independently run ten times for each K cluster. The likely number of clusters (K) was set to values from 1 to 6. As the LnP(D) estimator has been shown to overestimate K, as it frequently plateaus at higher values than biologically meaningful estimates of K, I also calculated the Δk statistic (Evanno et al. 2005).

4.3.7. Analysis of gene flow

Contemporary migration rates among the bottlenose dolphin social subpopulations were estimated using the Bayesian multilocus genotyping approach implemented in the program BayesAss 3.0 (Wilson and Rannala 2003). I reached the acceptance rates of total iterations (i.e., between 20 and 60%) by adjusting the parameters migration rates (m), allele frequencies (a) and inbreeding coefficient (f) to 0.3, 0.5 and 0.6, respectively. Five independent runs were performed using 107 Markov chain Monte Carlo (MCMC) iterations, 106 burn-in and sampled every 103 iterations. Convergence was examined using the software Tracer 1.6 (Rambaut et al. 2013). In addition, sex-biased dispersal was analysed in GenAlEx (Peakall and Smouse 2012), by calculating sex-specific assignment index correction (AIc) and testing difference for statistical significance using the Mann-Whitney U test (Mossman and Waser 1999).

4.3.8. Assessment of genetic diversity

For microsatellites, I assessed the genetic diversity within social subpopulations by † calculating the number of alleles (NA), private alleles (NPA), and observed (HO) and expected (HE) heterozygosities in GenAlEx (Peakall and Smouse 2012). I also estimated the allelic richness* (AR) and inbreeding coefficient (FIS) using FSTAT (Goudet 2001). For mtDNA, identification of the haplotypes, and assessment of haplotypic diversity† (h) and nucleotide diversity† (π) were performed in DnaSP 5.10 (Librado and Rozas 2009). I constructed a haplotype network using median-joining

112 Chapter 4 – Genetic structure of socio-geographic subpopulations implemented in NETWORK 5.0 (Bandelt et al. 1999) to assess genealogical relationship.

Both microsatellites and mtDNA were tested for evidence of recent bottleneck† events. For microsatellites, I used the software BOTTLENECK 1.2 (Cornuet and Luikart 1996) and assessed the significance of Wilcoxon sign rank tests for two models of generation of new alleles: the stepwise mutation model (SMM) and the two-phased model of mutation (TPM; variance = 30, 70% stepwise mutational model, 103 iterations). I also inspected the distribution of allelic frequencies to detect a mode-shift distortion due to the loss of rare alleles (Luikart et al. 1998). For mtDNA, signals of population size reduction (based on the assumption of neutrality) were assessed using Tajima’s D test (Tajima 1989) and Fu’s FS test (Fu 1997) implemented in Arlequin. P-values of both statistics were generated using 103 simulations

4.4. Results

4.4.1. Validation of genotypes and haplotypes

A total of 155 biopsy samples were collected from bottlenose dolphins from 2007 to 2015. Thirteen samples were identified as duplicates through confirmation by both identical microsatellite genotypes and photographic identification. One sample was confirmed to be the same individual via photography but not all loci alleles matched, suggesting an issue in the alignment of the DNA. Another 32 samples were of individuals never seen during the systematic surveys conducted from June 2011 to May 2015 (see Chapters 2 and 3, Chabanne et al. 2017a, b). As I aimed to verify whether or not the social structure previously described among bottlenose dolphins in the study area (see Chapter 3, Chabanne et al. 2017a) was influenced by genetic differentiation, those samples were removed as well as the duplicated ones from the data set, leaving a total of 109 different sampled individuals.

Micro-Checker analysis of microsatellite data did find evidence of scoring errors due to stuttering or large allele dropout at one locus (SCA27). Test for HWE at the population level indicated significant departure for loci TexVet5, and SCA27, both

113 Chapter 4 – Genetic structure of socio-geographic subpopulations due to homozygosity excess that could be explained by the presence of null alleles (Table 4.1). Using INEST program, frequencies of null alleles with consideration of inbreeding for loci TexVet5 and SCA27 were of 0.0978 (SE 0.0010) and 0.1034 (SE 0.0012), respectively. Substantial missing data existed at SCA17 due to allele scoring difficulties. Therefore, I excluded TexVet5, SCA27 and SCA17 from further analyses. After sequential Bonferroni correction, there was no significant linkage disequilibrium between any pair of loci (see Appendix A4.1).

Table 4.1. Microsatellite diversity in Indo-Pacific bottlenose dolphins inhabiting Perth metropolitan waters.

Locus n NA HO HE FIS r DIrFCB4 108 14 0.796 0.833 0.044 0.0196 (0.0005)

LobsDi_19 109 12 0.743 0.746 0.003 0.0123 (0.0005)

LobsDi_21 109 7 0.761 0.729 -0.045 0.0072 (0.0004)

LobsDi_24 109 14 0.798 0.848 0.059 0.0121 (0.0005)

LobsDi_9 105 6 0.657 0.720 0.088 0.0420 (0.0012) TexVet5 107 9 0.589 0.725 0.188 * 0.0978 (0.0010) DIrFCB5 109 8 0.633 0.676 0.064 0.0117 (0.0005)

LobsDi_39 108 11 0.787 0.771 -0.021 0.0060 (0.0003)

SCA22 105 18 0.924 0.895 -0.032 0.0049 (0.0002)

SCA27 108 8 0.241 0.324 0.256 * 0.1034 (0.0012) SCA9 103 17 0.845 0.859 -0.017 0.0063 (0.0003)

TexVet7 108 5 0.472 0.467 -0.012 0.0154 (0.0007)

Note: n = number of screened samples; NA = number of found alleles; HO = observed heterozygosity; HE = expected heterozygosity; FIS = coefficient of inbreeding; r (SE) = frequency of null alleles. * Significant departure from Hardy-Weinberg equilibrium after Bonferroni correction.

All the remaining ten loci were polymorphic with the number of alleles per locus ranging from five to 18. The probability of two unrelated dolphins sharing the same genotype (PI) across all ten loci was very low (max PI = 1.1 × 10-11), indicating a high discriminatory power of the set of microsatellite loci.

Mitochondrial control region sequences of 412-base pairs were aligned for 99 of the 109 samples and defined six unique haplotypes (see section 4.4.6 below).

114 Chapter 4 – Genetic structure of socio-geographic subpopulations

4.4.2. Genetic differentiation

A weak but significant degree of overall genetic differentiation was seen for both marker data sets. The overall average FST for the ten microsatellite loci was 0.020

(P-value < 0.0005). For the mtDNA data set, the overall FST and ΦST values were 0.127 (P-value < 0.006) and 0.128 (P-value < 0.005), respectively.

Pairwise FST comparisons (Table 4.2) based on microsatellite data again revealed significant and low genetic differentiation (FST < 0.03), except between GR and SCR and between GR and OA. MtDNA data, however, returned a significant genetic differentiation only between OA and CS (ΦST = 0.310, P-value < 0.01, Table 4.2).

Table 4.2. Pairwise fixation indices between four genetic clusters previously defined in STRUCTURE analysis based on ten microsatellite loci and mtDNA control sequences. Microsatellite FST values are above the diagonal; Mitochondrial ΦST values are below. Mitochondrial FST values were similar to ΦST, and thus are not shown here; values that are significant after sequential Bonferroni correction are in bold. Values in italic were significant before correction (P-value < 0.05).

Subpopulation GR SCR OA CS GR - 0.016 0.006 0.018 *

-0.063 - 0.023 ** 0.025 *** SCR -0.037 0.017 - 0.020 ** OA 0.186 0.122 0.310 ** - CS Note: * P-value < 0.05, ** P-value < 0.01, *** P-value < 0.001 after sequential Bonferroni correction. GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

115 Chapter 4 – Genetic structure of socio-geographic subpopulations

S Shannon’s mutual information index ( HUA) (Table 4.3) supported some FST and ΦST S results. GR showed no significant differences in allele or mtDNA HUA to any other subpopulation after sequential Bonferroni correction, although significant difference was detected with SCR before correction (P-value < 0.05). All other subpopulations were significantly different to one another.

Subpopulation GR SCR OA CS GR - 0.152 0.140 0.099

0.159 - 0.165 ** 0.152 ** SCR 0.160 0.221 * - 0.119 ** OA CS 0.102 0.204 ** 0.137 ** - Note: * P-value < 0.05; ** P-value <0.01; *** P-value < 0.001 after sequential Bonferroni correction. GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

4.4.3. Genetic population structure

Microsatellite data analysis using STRUCTURE with no prior information showed the highest mean posterior probability (LnP(D)) reached at K = 4 (see Appendix A4.2). However, the ΔK index obtained by the method of Evanno et al. (2005) revealed a modal value of ΔK = 20.2 at K = 3 (see Appendix A4.2), although there was no clear correlation between the genetic clusters and the socio-geographic subpopulations (Figure 4.2a). When prior information on socio-geographic subpopulations was included, the LnP(D) mean was higher at K = 1, although not different at K = 2, 3 or 4, and the modal value of ΔK = 4.05 was found at K = 4 (Appendix A4.2). Individuals assigned to genetic cluster 2 (red) were essentially from the SCR subpopulation, although their mean proportion of membership q was moderate (q = 0.635) (Figure 4.2b). Similarly, all individuals from the GR subpopulation and majority of the individuals from the OA subpopulation (n = 86%) were assigned to genetic cluster 1 (blue) with moderate confidence (q = 0.613 and 0.667,

116 Chapter 4 – Genetic structure of socio-geographic subpopulations respectively), while individuals from the CS subpopulation were mainly of mixed ancestry from genetic clusters 1 (blue), 3 (green) and 4 (purple).

Figure 4.2. Bayesian assignment probabilities from STRUCTURE for bottlenose dolphins based on ten microsatellite loci: (a) without prior information, K = 3. (b) With prior information, K = 4. Each vertical line represents one individual, with the strength of that individual to any of the genetic clusters (blue cluster 1; red cluster 2; green cluster 3; purple cluster 4). Individuals are grouped by social subpopulations (GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound) and sorted within subpopulations by the latitude of sampling location from North (left) to South (right) when available.

As suggested by Pritchard et al. (2000), in order to further assess possible population structure within the three coastal subpopulations (GR, OA, and CS subpopulations), I estimated the number of populations (K) by considering only the individuals from those subpopulations. No population structure was detected (K = 1, results not presented).

117 Chapter 4 – Genetic structure of socio-geographic subpopulations

4.4.4. Gene flow and dispersal

Estimated contemporary migration rates inferred in BayesAss suggested very low gene flow between the majority of the social subpopulations (m = 1-4%, Table 4.4). However, estimated migration rates from OA to other subpopulations were moderate (m = 23% to 28%), suggesting that OA may act as a source with negligible migration in the opposite direction. The proportion of non-immigrants from their respective origin subpopulation was high for OA subpopulation (0.91), while others show lower proportions (from 0.68 to 0.74).

Table 4.4. Mean (standard deviation) of the posterior distribution of the contemporary migration rates (m) in BayesAss (Wilson and Rannala 2003) among four bottlenose dolphin socio-geographic subpopulations in Perth metropolitan waters. The subpopulations of which each dolphin belongs are listed in the rows, while the subpopulations from which they migrated are listed in the columns. Values along the diagonal (in bold) are the proportions of non-immigrants from the origin subpopulation for each generation. Moderate estimated migration rates (m > 0.10) are displayed in italic.

Migration Origin: into: GR SCR OA CS GR 0.69 (0.02) 0.03 (0.02) 0.26 (0.03) 0.02 (0.02) SCR 0.01 (0.01) 0.74 (0.03) 0.23 (0.03) 0.02 (0.01) OA 0.03 (0.02) 0.04 (0.02) 0.91 (0.03) 0.02 (0.02) CS 0.02 (0.01) 0.01 (0.01) 0.28 (0.02) 0.68 (0.01) Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

4.4.5. Sex-biased dispersal

Of the total of 109 dolphins sampled, 56 were females, and 53 were males. Analysis of mtDNA sequences among subpopulations did not reveal significant structuring for females and males (Table 4.5) and the corrected mean assignment indices for females and males did not reveal any sex-biased dispersal (Z = -0.177, P-value =

0.860) (see Appendix A4.3). However, because the differentiation FST is a function of 1/(1+4Nm) (Wright 1931), and mtDNA effective population size Nmt should represent 0.25 of the microsatellite effective population size Ne (Avise et al. 1987), my results suggested restricted female movement with a much lower Nm for mtDNA

118 Chapter 4 – Genetic structure of socio-geographic subpopulations than expected if both males and females have similar dispersal (m) (see Appendix

A4.4 for details). However, values of FST for mtDNA should be used with caution since FSTAT estimated negative values.

Within sex classes, differentiation of nuclear variation, suggested some level of genetic differentiation among males from SCR and CS and among females from SCR, OA and CS subpopulations (Table 4.5).

FST microsatellite FST mtDNA Females Males Female Male All 0.019 0.021 0.142 0.114

GR - SCR 0.003 0.022 0.006 -0.066 GR - OA 0.005 0.007 0.126 0.150 GR - CS 0.017 0.018 0.036 -0.086 SCR - OA 0.027 * 0.034 0.309 0.311 SCR - CS 0.022 * 0.026 * -0.119 0.020 OA - CS 0.022 * 0.023 0.502 -0.061 Note: GR = Gage Roads, SCR = Swan Canning-Riverpark, OA = Owen Anchorage, and CS = Cockburn Sound.

4.4.6. Genetic diversity

Levels of microsatellite diversity were high for all subpopulation samples as measured by both allelic richness (AR) and expected heterozygosity (HE). Allelic richness (AR) ranged from 5.9 to 6.6 and expected heterozygosity (HE) from 69% to

76% (Table 4.6). The average number of alleles (NA) per subpopulation ranged from 6.0 to 9.6 with an overall value of 11.2. Out of a total of 29 private alleles identified, four were found in OA, seven in SCR and 18 in CS. When null alleles were acknowledged (INEST tests), none of the estimated inbreeding (FIS) values were significantly different from zero except for SCR.

119

Chapter 4 – Genetic structure of socio-geographic subpopulations

Table 4.6. Genetic diversity measures (SE) for bottlenose dolphin socio-geographic subpopulations using microsatellite loci (n = 10) and mtDNA.

Microsatellites mtDNA FSTAT n Nf Nm NA NPA AR HE HO FIS n NH h π INEST 11.2 6.5 0.75 0.74 -0.020 NS 0.585 0.014 Overall 109 56 53 - 99 6 (1.5) (0.7) (0.04) (0.04) -0.021 NS (0.044) (0.002)

6.0 5.9 0.69 0.69 -0.004 NS 0.691 0.020 GR 14 7 7 0 11 4 (0.8) (0.8) (0.04) (0.05) -0.025 NS (0.128) (0.006)

7.1 5.9 0.70 0.68 -0.050 NS 0.414 0.017 SCR 24 13 12 7 21 4 (0.9) (0.7) (0.05) (0.05) -0.048 * (0.124) (0.005)

7.4 6.2 0.72 0.74 -0.008 NS 0.667 0.023 OA 22 13 9 4 21 3 (0.8) (0.6) (0.04) (0.05) -0.010 NS (0.050) (0.003)

9.6 6.6 0.76 0.78 -0.014 NS 0.532 0.005 CS 48 23 25 18 46 4 (1.3) (0.7) (0.04) (0.04) -0.013 NS (0.053) (0.002) Note: n = number of samples; Nf = number of females; Nm = number of males; NA = mean number of alleles; NPA = number of private alleles; AR = mean allelic richness; HE = expected heterozygosity; HO = observed heterozygosity; FIS = Inbreeding coefficient calculated both in FSTAT using the original dataset and in INEST using a null-allele corrected dataset; NH = number of haplotypes; h = Haplotypic diversity; π = nucleotide diversity. NS = non-significant; * FIS significantly different to zero P-value < 0.05 two-tailed). GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

120

Chapter 4 – Genetic structure of socio-geographic subpopulations

For the mtDNA, a total of 21 polymorphic sites were found among samples, defining six unique haplotypes. There were two common haplotypes (H1, H2), and a third haplotype (H3) shared by all except SCR. There was one rare haplotype (H6 found in three samples from GR and SCR), as well as two unique haplotypes each found in only one sample (H4 found in CS and H5 found in SCR). As a result, moderate haplotypic (h) but low nucleotide (π) diversity was observed (h: 0.414-0.683; π: 0.005-0.022) (Table 4.6).

The median-joining network showed two main groups of haplotypes with a central core showing a minimum of 17 mutational steps (unsampled or extinct haplotypes) (Figure 4.3). However, there was no clear clustering based on socio-geographic subpopulations.

Figure 4.3. Median-joining network of mtDNA control region haplotypes in Indo- Pacific bottlenose dolphins in Perth metropolitan waters. The size of the circles is proportional to the total number of individuals carrying that haplotype. Different colours denote the four different sampled subpopulations: purple GR = Gage Roads, red SCR = Swan Canning Riverpark, green OA = Owen Anchorage and yellow CS = Cockburn Sound. Number of mutational events between each haplotype is indicated by hash marks.

For mtDNA, Tajima’s and Fu’s tests for selection or population size change did not differ significantly from expected under a neutral model of evolution for the overall and any of the socio-geographic subpopulations, except CS which showed a significantly negative Tajima’s D value (Table 4.7). Similarly, the two tests for a bottleneck using allele frequencies of microsatellite loci (SMM and TPM) did not show significant heterozygosity excess after correcting for multiple comparisons

121 Chapter 4 – Genetic structure of socio-geographic subpopulations

(Table 4.7), suggesting no evidence of a recent decline or colonisation event. Additionally, the distribution of allelic frequencies did not show significant departure from a standard L-shape in the model-shift test, indicating no loss of rare alleles in any subpopulations (see Appendix A4.5).

Table 4.7. Summary statistics of various tests to detect a recent bottleneck effect based on mtDNA control region (Tajima’s D and Fu’s Fs) and microsatellite loci (SMM: stepwise mutation model. TPM: two-phased model). The Wilcoxon test found no significant heterozygosity excess after Bonferroni correction (Pcrit = 0.012).

mtDNA Microsatellites Wilcoxon test (P-value) Tajima's D Fu's Fs SMM TPM Overall 1.120 1.730 0.999 0.652

GR 1.154 5.414 0.188 0.839 SCR 0.863 8.432 0.997 0.313 OA 2.936 13.046 0.999 0.652 CS -1.748* 3.325 0.990 0.186 Note: * significant value, P-value < 0.05; GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

4.5. Discussion

s Despite the significant FST and HUA differentiations found between three out of four socio-geographic subpopulations defined in Chapter 3 (Chabanne et al. 2017a) - SCR, OA and CS - the high gene flow from OA supports several factors indicating the potential for interbreeding (Litz 2007), namely: the high mobility of the species, the small size of the study area (c. 300 km2), the evidence of home range overlapping and the occurrence of mixed groups between subpopulations (see Chapter 3, Chabanne et al. 2017a).

The genetic structure of bottlenose dolphins is typically assessed based on individual spatial information (e.g., Krützen et al. 2004; Natoli et al. 2004; Charlton-Robb et al. 2014; Fruet et al. 2014; Gaspari et al. 2015; Allen et al. 2016). Although spatial segregation was found among bottlenose dolphins in Perth (see Chapter 3, Chabanne et al. 2017a), here I investigated the genetic structure based on the social structure described in the population of bottlenose dolphins in the study area (see Chapter 3,

122 Chapter 4 – Genetic structure of socio-geographic subpopulations

Chabanne et al. 2017a). The social structure was defined from photo-identification data collected between 2011 and 2015 and would, therefore, reflect less than a single generation, defined as the average age of females at the time of the birth of their first female calf, from demographic data (Manlik et al. 2016). Using microsatellite loci and mtDNA markers, I was able to obtain information on distinct timescales. As microsatellite markers are biparentally inherited with relatively fast mutation rates, they allow insights of recent and almost contemporaneous events (i.e., from the last few generations). In contrast, mtDNA markers reflect more historical events due to their relatively slow mutation rates associated with their maternal mode of inheritance (Avise et al. 1987). Comparing microsatellite loci and mtDNA marker variation may also permit the detection of differences in migration rates between sexes, using the assumption that the effective population size Nmt of mtDNA is 0.25 that of microsatellite Ne (Avise et al. 1987).

4.5.1. Genetic structure and gene flow

Pairwise FST comparisons (Table 4.2) of genetic differentiation based on microsatellite loci found significant contemporary differences between SCR, OA, and CS as well as between GR and CS. However, values of FST were weak with a maximum value of 0.025 between SCR and CS. The high variation in microsatellites for all subpopulations (Table 4.6) may provide high statistical power and thus increase the ability to pick up small differences in allele frequencies that are not biologically meaningful (Hedrick 1999; Kalinowski 2002). However, more importantly, it is known that high variation within populations severely depresses

FST between populations (Jost 2008). The microsatellites and mtDNA mutual s information index HUA (Table 4.3) also showed contemporary and historical genetic s differentiation between SCR, OA, and CS that are moderate (0.119 < HUA < 0.165 s and 0.137< HUA < 0.1221 based on microsatellites and mtDNA markers, s respectively). HUA avoids dependence of between-population variation on within- population variation, and relative to FST shows a more robust and predictable response to levels of dispersal, over a wide range of population sizes and dispersal rates (Sherwin et al. 2006). The results suggested weak sex-biased dispersal, at least for females as they may have restricted movement. However, this suggestion was based on non-significant values (Table 4.5). Other studies found no evidence for sex-

123 Chapter 4 – Genetic structure of socio-geographic subpopulations biased dispersal in bottlenose dolphins that have complex social structure and show high site fidelity (Natoli et al. 2004; Sellas et al. 2005; Parsons et al. 2006; Rosel et al. 2009; Ansmann et al. 2012). Here, the four socio-geographic subpopulations showed different social patterns, with stronger and longer-term social patterns associated with high residency patterns in the SCR subpopulation and weaker social patterns with a majority of transient animals in the GR subpopulation (see Chapters 2 and 3, Chabanne et al. 2017a, b).

Contemporary migration rates inferred in BayesAss (Table 4.4) indicated low gene flow with values estimated at < 4% per generation between socio-geographic subpopulations. However, the high asymmetric gene flow (0.23 - 0.28) estimated from OA to other subpopulations suggested that SCR, CS and GR are all genetically dependent on OA, or the other subpopulations all sometimes use OA’s area, which is physically intermediate to the other subpopulations’ areas, although different degrees of demographic independence may occur (see Chapters 2 and 3, Chabanne et al. 2017a, b).

Isolation by distance is not the only factor that can explain genetic differentiation (e.g., Natoli et al. 2005; Sellas et al. 2005; Gaspari et al. 2015; Allen et al. 2016). Variation in oceanographic conditions (e.g., salinity and temperature gradients, habitat type) as well as dissimilarity in prey preference and prey distribution and abundance, may explain genetic differentiation at a fine-scale (Natoli et al. 2005; Gaspari et al. 2015). Similarly, Kopps et al. (2014) provided evidence for fine-scale geographical genetic differentiation driven by socially transmitted behaviour associated with particular habitat. In my study, however, estimates of low gene flow between non-differentiated pairwise subpopulations compared with GR (Tables 4.2 and 4.3) may be an artefact of the small sample size representing GR or indicate that dispersal between populations is at least greater than a few individuals per generation.

With input from prior information (i.e., the geographic region of the socio- geographic subpopulations), the Bayesian technique implemented in STRUCTURE did not reveal the best number of genetic clusters, but varying from one to four (Appendix A4.2). Although four contemporary genetic clusters seemed to be best

124 Chapter 4 – Genetic structure of socio-geographic subpopulations when using the method of Evanno et al. (2005), none of the four socio-geographic subpopulations was strongly assigned to a cluster (proportion of membership q < 0.80, Figure 4.2). The inability of STRUCTURE to correctly assign all individuals to their subpopulation of origin may be due to software limitations in detecting genetic differentiation when FST values are low (FST < 0.025, Table 4.2) (Latch et al. 2006). Relevantly, the same analysis without samples from the SCR locality could not define more than one genetic population. This result is consistent with the low FST (values < 0.02, Table 4.2) (Latch et al. 2006) and that the three localities (OA, GR and CS) in coastal waters define a more continuous environment (i.e., no movement restriction such as the narrow harbour port at the entrance of the SCR estuary). Additionally, I previously documented that movement occurs between the three coastal geographic regions (see Chapter 2, Chabanne et al. 2017b). Louis (2014), for example, found a similar dilemma in regards to defining more than one genetic structure among three social clusters.

Estimates of mtDNA FST or ΦST (Table 4.2) were similar, and both were significantly different to zero only between OA and CS. The dissimilarity in pairwise population differentiation between estimates of microsatellite FST (biparental inheritance) and mtDNA FST (maternal inheritance) may be caused by sex-biased dispersal – however, in this current study, the described weak sex-biased dispersal was not significant. Another possible explanation, as previously mentioned, is that microsatellite markers are highly polymorphic (i.e., each locus act as an independent marker), thus offering higher statistical power than mtDNA markers that are defined by one locus only (Kalinowski 2002). The lack of differentiation in the mtDNA FST was also supported by the median-joining network analysis of mtDNA haplotypes (Figure 4.3) suggesting that the contemporary geographic distributions of individuals associated with distinct socio-geographic subpopulations was not associated with historical processes. The analysis described two lineages with shared haplotypes H2 and H3 (clade defining Tursiops aduncus) differing by at least 15 bases from the other lineage containing another shared haplotype (H1) that was closely related to Tursiops truncatus or even Stenella coeruleoalba (Genbank). In Shark Bay (WA), Krützen et al. (2004) found that the microsatellites of the two main haplotype clades were not significantly different from one another and that interbreeding occurred between both haplotype clades, but suggested some separation of the matrilines

125 Chapter 4 – Genetic structure of socio-geographic subpopulations between the eastern Gulf and other localities. In the current study, the most likely explanation is that the Perth metropolitan waters have been colonised by two distinct mitochondrial DNA lineages (e.g., other Tursiops aduncus populations along the south west coast of Western Australia and offshore Tursiops truncatus populations).

4.5.2. Conservation implications

This study found that the genetic structure of Indo-Pacific bottlenose dolphins in coastal and estuarine waters of Perth seems particularly associated with the current socio-geographic structure but less clearly with historical events. Indeed, there was significant differentiation between some of the four socio-geographic subpopulations for both the FST (Table 4.2) and relatedness index (see Chapter 3, Chabanne et al. 2017a), but the differentiation was weak, thus suggesting a very recent change in genetic structure or, more likely, that there has never been much genetic structure in the last 200 years, and the social structure changes every few years, having little effect on the genetic structure.

Reduction in genetic diversity may occur after a population expansion associated with a stochastic event such as drastic changes in the environment, selective sweeps or founder and bottleneck effects (Hedrick 2011). The mtDNA Tajima’s D analysis (Table 4.7) associated with low nucleotide diversity (π = 0.005, Table 4.6) indicated the occurrence of a historical bottleneck event or selection in CS (Rand 1996), although the failure to detect a similar event in SCR may result from low statistical power of those tests associated with limited sample sizes (Peery et al. 2012). However, it is important to note that the Tajima’s D result does not agree with all other bottleneck tests (i.e., for mtDNA: Fu’s Fs; for microsatellites: the stepwise mutation model (SMM) and the two-phased model of mutation (TPM)). Although there may have been some bottlenecks, the high gene flow from the OA subpopulation to others has probably erased any signal. Historically, the OA and CS embayments as well as the section between Fremantle and Rottnest Island ( < 20 km) were flooded (c. 5,000 years ago) sometime after Gage Roads (c. 7,000 years ago) during the rise in sea levels of the Holocene marine transgression (Brearley 2005). OA may be connected to dolphins at Rottnest Island and from there to dolphins further offshore (i.e., more T. truncatus types). Thus, gene flow may occur not only

126 Chapter 4 – Genetic structure of socio-geographic subpopulations north – south along the coast but also offshore – onshore. The connection between the SCR and coastal areas, however, was more recent with the removal of the rock bar at the river mouth in the late 1800s (Department of Parks and Wildlife 2015), although historical research indicates that dolphins were already present in the SCR (pers. comment S. Graham-Taylor). The SCR is a permanently open estuary with the middle and upper reaches that changes from brackish (conditions still occurring in winter and spring) to marine salinity conditions (Department of Parks and Wildlife 2015). In 2009, the subpopulation of bottlenose dolphins in the SCR experienced an unusual mortality event, with six dolphins found dead (Stephens et al. 2014). In a report associated with this event, seven haplotypes were identified among 13 samples (obtained between 2007-09) of resident individuals from the SCR subpopulation (Holyoake et al. 2010; Chabanne et al. 2012). In my study, however, I defined only four haplotypes (Table 4.6), thus, suggesting a loss of haplotypic diversity potentially associated with the 2009 unusual mortality event (Stephens et al. 2014) or

Microsatellite genetic variation in CS was high in comparison to GR and OA (Table 4.6), but could not be explained by a large population size. Instead, contemporary gene flow from OA (Table 4.4) may have helped to maintain the genetic variation, although influx from other adjacent populations may occur (see, for example, Figure 4.2b showing considerable admixture with a genetic cluster that was not well represented in other socio-geographic subpopulations). In that context, it is notable that some individuals from CS that were reported overlapping with adjacent populations located further south (c. 12 km, pers. comment K. Nicholson), provide an opportunity for interbreeding at a larger scale. Additionally, Manlik et al. (in prep-a) suggested genetic connectivity of bottlenose dolphins from Bunbury (located c.180 km south of Perth) into northern localities, including Perth.

In addition to the possible bottleneck or founder event, the INEST test suggested some inbreeding in SCR (Table 4.6). The high site fidelity, year-round residency pattern, long-term and stable association patterns amongst females and amongst males, low but stable abundance (i.e., number of individuals using the SCR), asymmetric demographic movement (see Chapters 2 and 3, Chabanne et al. 2017a,

127 Chapter 4 – Genetic structure of socio-geographic subpopulations b) and the contemporary gene flow from OA observed for SCR are all characteristics indicating that SCR acts as a sink and OA as a source in a source-sink dynamic.

4.5.3. Conclusion

Despite there being some current social and spatial patterns in bottlenose dolphins inhabiting Perth metropolitan waters (Chapter 3, Chabanne et al. 2017a), there is only weak genetic structure, apparently because of the strong source-sink dynamic between a coastal subpopulation OA (source) and others in the Perth metropolitan region (SCR, CS and GR – sinks). This indicates that OA should be accorded very high conservation status within the Perth metropolitan region.

The SCR is a small subpopulation with fewer than 25 individuals, exhibiting strong ecological affinity to an estuary where significant anthropogenic stressors occur (see Chapters 2 and 3, Chabanne et al. 2017a, b), low genetic variation, and restricted gene dispersal, factors suggesting that inbreeding may occur. The results of this chapter, along with the findings reported in Chabanne et al. (2012) and Chapters 2 and 3 (Chabanne et al. 2017a, b) of this thesis and those related to the 2009 unusual mortality event (Holyoake et al. 2010; Stephens et al. 2014) suggest that the SCR subpopulation is highly vulnerable and that appropriate conservation measures are needed in addition to maintaining its connectivity with the coastal subpopulation OA (source).

It is also possible that other adjacent populations from outside the study area contribute to the genetic material of the bottlenose dolphins in the Perth area. Although Manlik et al. (in prep-a) recently indicated the occurrence of genetic dispersal from Bunbury (c. 180 km south of Perth) to northern populations (including Perth), further studies would still be necessary to investigate the degree of connection between Perth bottlenose dolphins and other populations at regional- scale.

128 Chapter 4 – Genetic structure of socio-geographic subpopulations

Appendices

Appendix A4.1. Linkage disequilibrium

Table A4.1.1. Pairwise loci test for linkage disequilibrium over all four subpopulations. Significant values after Bonferroni correction are in bold (P-value <0.05).

Com_GR Com_SCR Com_OA Com_CS Locus#1 Locus#2 P-value SE P-value SE P-value SE P-value SE DIrFCB4 LobsDi_19 0.0120 0.0005 0.8777 0.001 0.5505 0.005 0.2788 0.0082 DIrFCB4 LobsDi_21 0.8366 0.0022 0.4797 0.0038 0.0025 0.0004 0.5008 0.0104 LobsDi_19 LobsDi_21 0.2357 0.0027 0.0452 0.0018 0.3302 0.0034 0.1281 0.0039 DIrFCB4 LobsDi_24 1.0000 0.0000 0.7299 0.0045 0.7952 0.0069 0.1386 0.0089 LobsDi_19 LobsDi_24 0.0345 0.0013 0.5651 0.0058 0.4794 0.0075 0.5359 0.0081 LobsDi_21 LobsDi_24 0.3489 0.0047 0.4028 0.0067 0.5194 0.0062 0.4309 0.0091 DIrFCB4 LobsDi_9 1.0000 0.0000 0.6494 0.0031 0.9414 0.0010 0.0371 0.0026 LobsDi_19 LobsDi_9 1.0000 0.0000 0.0014 0.0003 0.7765 0.0038 0.0978 0.0028 LobsDi_21 LobsDi_9 0.7722 0.0027 0.1470 0.0033 0.9787 0.0001 0.7034 0.0049 LobsDi_24 LobsDi_9 1.0000 0.0000 0.4475 0.0063 0.1578 0.0047 0.0424 0.0027 DIrFCB4 DIrFCB5 1.0000 0.0000 0.5115 0.0034 0.6933 0.0049 0.8193 0.0065 LobsDi_19 DIrFCB5 0.2996 0.0026 0.3478 0.0036 0.4304 0.0030 0.5041 0.0054 LobsDi_21 DIrFCB5 0.4733 0.0034 0.3615 0.0038 0.2674 0.0026 0.1435 0.0041 LobsDi_24 DIrFCB5 0.2926 0.0043 0.5887 0.0058 0.2768 0.0059 0.2269 0.0068 LobsDi_9 DIrFCB5 1.0000 0.0000 0.9778 0.0006 0.3481 0.0031 0.2314 0.0045 DIrFCB4 LobsDi_39 0.5450 0.0034 0.4976 0.0039 0.6350 0.0052 0.2544 0.0100 LobsDi_19 LobsDi_39 0.5597 0.0027 0.2305 0.0022 0.4340 0.0049 0.3475 0.0067 LobsDi_21 LobsDi_39 1.0000 0.0000 0.1234 0.0028 0.9655 0.0018 0.7517 0.0062 LobsDi_24 LobsDi_39 1.0000 0.0000 0.4556 0.0051 0.1044 0.0052 0.0484 0.0039 LobsDi_9 LobsDi_39 0.1384 0.0026 0.7440 0.0030 0.8900 0.0021 0.0262 0.0016 DIrFCB5 LobsDi_39 0.8210 0.0021 0.2343 0.0023 0.0026 0.0004 0.6036 0.0062 DIrFCB4 SCA22 1.0000 0.0000 0.2592 0.0047 0.5231 0.0083 0.0109 0.0025 LobsDi_19 SCA22 0.4093 0.0040 0.4103 0.0054 0.5318 0.0060 0.0192 0.0023 LobsDi_21 SCA22 1.0000 0.0000 0.8021 0.0032 1.0000 0.0006 0.6081 0.0102 LobsDi_24 SCA22 0.0499 0.0031 1.0000 0.0007 1.0000 0.0000 1.0000 0.0000 LobsDi_9 SCA22 1.0000 0.0000 0.6999 0.0060 0.2665 0.0050 0.2652 0.0079 DIrFCB5 SCA22 0.2445 0.0040 1.0000 0.0004 1.0000 0.0003 0.0655 0.0045 LobsDi_39 SCA22 1.0000 0.0000 0.7091 0.0040 1.0000 0.0000 0.6037 0.0110 DIrFCB4 SCA9 1.0000 0.0000 1.0000 0.0006 0.4509 0.0080 0.5763 0.0132 LobsDi_19 SCA9 1.0000 0.0000 0.1000 0.0040 1.0000 0.0008 0.8635 0.0054 LobsDi_21 SCA9 0.5304 0.0041 0.7322 0.0051 0.1285 0.0040 0.8137 0.0071 LobsDi_24 SCA9 1.0000 0.0000 1.0000 0.0008 1.0000 0.0002 1.0000 0.0000 LobsDi_9 SCA9 1.0000 0.0000 0.0587 0.0030 0.6186 0.0060 0.1308 0.0053 DIrFCB5 SCA9 1.0000 0.0000 0.4561 0.0075 0.5711 0.0067 0.4311 0.0078 LobsDi_39 SCA9 1.0000 0.0000 1.0000 0.0002 1.0000 0.0001 0.8789 0.0061 SCA22 SCA9 1.0000 0.0000 1.0000 0.0000 0.1454 0.0070 0.8350 0.0094 DIrFCB4 TexVet7 0.5372 0.0013 0.0040 0.0000 0.1994 0.0038 0.0825 0.0027 LobsDi_19 TexVet7 0.7540 0.0010 0.8372 0.0012 0.8754 0.0012 0.3985 0.0033 LobsDi_21 TexVet7 0.2634 0.0012 0.2375 0.0014 0.0557 0.0018 0.0420 0.0013 LobsDi_24 TexVet7 0.6379 0.0017 0.5260 0.0028 0.0296 0.0012 0.0757 0.0022 LobsDi_9 TexVet7 0.1314 0.0010 0.3741 0.0029 0.0478 0.0016 0.7639 0.0026 DIrFCB5 TexVet7 1.0000 0.0000 0.0291 0.0005 0.2657 0.0023 0.0075 0.0004 LobsDi_39 TexVet7 0.6971 0.0012 0.1153 0.0017 0.5579 0.0032 0.4428 0.0040 SCA22 TexVet7 1.0000 0.0000 0.0902 0.0011 0.3862 0.0044 0.0174 0.0013 SCA9 TexVet7 0.1363 0.0012 0.5520 0.0036 0.0351 0.0010 0.0452 0.0021 3 6

129 Chapter 4 – Genetic structure of socio-geographic subpopulations

Appendix A4.2. Mean of the posterior probabilities (LnP(D)) and ΔK statistic for STRUCTURE

(a) (b)

-3600 25

-3700 20

-3800 15

P(D) K

-3900 Δ 10

-4000 5

Mean of Ln ofMean -4100 0 0 2 4 6 8 0 2 4 6 8

-3700 5

-3800 4

-3900 3

P(D) Δ -4000 2

-4100 1

Mean of Ln ofMean -4200 0 0 2 4 6 8 0 2 4 6 8 K K Figure A4.2.1. Mean of the estimated posterior probabilities (LnP(D)) and ΔK statistic (Evanno et al. 2005) over ten replicate runs for values of K = 1-10 using the Bayesian method in STRUCTURE ((a) and (b)) without prior location and ((c) and (d)) with prior location.

130 Chapter 4 – Genetic structure of socio-geographic subpopulations

Appendix A4.3. Mean assignment indices (AIc)

Table A4.3.1. Mean assignment indices (AIc) within sexes overall and between pairwise socio-geographic subpopulations. Outcomes of the Mann-Whitney U test are given (Z, P-value).

AIc Test Subpopulation Female Male Z P-value All 0.240 (0.244) -0.244 (0.417) -0.177 0.860

GR 0.307 (0.355) -0.219 (0.302) 1.056 0.291 SCR -0.105 (0.575) 0.105 (0.520) 0.295 0.768 OA -0.045 (0.500) 0.060 (0.600) -0.213 0.831 CS 0.345 (0.295) -0.345 (0.647) -0.077 0.939 Note: GR = Gage Roads, SCR = Swan Canning Riverpark, OA = Owen Anchorage, CS = Cockburn Sound.

131 Chapter 4 – Genetic structure of socio-geographic subpopulations

Appendix A4.4. Sex-biased dispersal

The differentiation FST is a function of 1/(1+4Nm) (Wright 1931), and mtDNA effective population size Nmt should represent 0.25 of the microsatellite effective population size Ne (Avise et al. 1987).

Table A4.4.1. Estimates of the observed and expected 4Nm values for females (overall).

Microsatellite FST mtDNA FST

Estimate FST 0.019 0.142 Estimate of 4Nm (observed) 51.632 6.042 Estimate of 4Nm (expected) (51.632) 12.908

Note: observed: FST = 1/(1+4Nm); expected Nmt = 0.25 Ne (microsatellites).

The difference obtained between the observed and expected values of mtDNA 4Nm suggested restricted female movement with a much lower Nm for mtDNA than expected if both males and females have similar dispersal (m).

I did not calculate the observed and expected for all populations-pairs because some mtDNA FST values were negative, thus showing ambiguity of the positive values (same dataset).

I did not verify the observed and expected values of 4Nm for males because the mtDNA come from the females.

132 Chapter 4 – Genetic structure of socio-geographic subpopulations

Appendix A4.5. Distribution of allelic frequencies

1 Overall GR SCR OA CS

0.8

0.6

0.4

0.2 Proportion of alleles ofProportion

0 0 0.2 0.4 0.6 0.8 1 Allele frequency class

Figure A4.5.1. Distribution of allelic frequencies in the mode shift test for overall and within each socio-geographic community (GR = Gage Roads (purple); SCR = Swan Canning Riverpark (red); OA = Owen Anchorage (green); CS = Cockburn Sound (orange)).

133

134

Chapter 5. Population genetic structure and effective population sizes in Indo-Pacific bottlenose dolphin (Tursiops aduncus) along the southwestern coastline of Western Australia

5.1. Abstract

Information on genetic factors that contribute to population structure, connectivity and effective population size (Ne) is important to inform management strategies for wildlife conservation. In Australia, isolated populations of Indo-Pacific bottlenose dolphin (Tursiops aduncus) have been documented at local and regional-scales. At a local scale, there is only weak genetic structure between socially differentiated subpopulations inhabiting a closed estuary, an embayment and adjacent coastline in Perth, Western Australia. To address regional-scale genetic structure, I evaluated the genetic diversity, population genetic structure and contemporary gene flow of bottlenose dolphins sampled (n = 221) across eight locations spanning c. 1,000 km along the southwestern coastline of Western Australia, using 25 microsatellite loci.

Measures of genetic diversity were similar across localities (HO = 0.532-0.592; HE =

0.513-0.574), but significant genetic differentiation (FST) occurred between all localities, except for two localities in Perth that supported a panmictic population. Contemporary gene flow described two source-sink dynamics with Bunbury acting as source for the nearest northern (Mandurah) and southern (Busselton) populations and Augusta acting as source for Albany and Esperance populations. Gene flow between other localities was negligible (i.e., 95% confidence interval included zero).

The small estimated effective population size of the Perth metapopulation (cNe = 23.0, 95% CI 19.7-27.3) raises conservation concerns, although the estimate may be biased downward by the small sample size and upward by the occurrence of immigration (>10%). Despite the larger sample sizes needed to fully investigate effective population sizes, the current findings have important conservation management implications for bottlenose dolphins inhabiting the Perth estuary and embayment sites as well as those along the adjacent coastline.

135 Chapter 5 – Regional-scale genetic structure

5.2. Introduction

Identifying population structure and genetic connectivity is essential for determining appropriate scales for wildlife conservation and management (e.g., Wiszniewski et al. 2010; Bilgmann et al. 2014; Sandoval-Castillo and Beheregaray 2015; Yannic et al. 2015). In both terrestrial and aquatic environments, gene flow may be limited and populations may become isolated even where geographical or physical barriers are absent and the mobility capacity of a species is high (Irwin 2002; Taylor 2005; Brown et al. 2007). If gene flow is limited, populations are vulnerable to loss of genetic diversity through genetic drift and mutation and to inbreeding depression.

The effective population size (Ne) is an important evolutionary and conservation parameter that can help in the assessment of the genetic vulnerability of a population.

For example, a population with an estimate Ne < 50 would most likely experience inbreeding depression, also be indicative of critical status for the population (Crnokrak and Roff 1999; Palstra and Ruzzante 2008).As the effective population size (Ne) represents the number of breeders in an idealised population, it may only be a portion of the census population size (Nc), which suggests that management strategies should consider both genetic and demographic factors in assessing the status of a population (Frankham 1995b).

5.2.1. Genetic structure

Developing appropriate conservation strategies for wildlife can be challenging if the genetic structure at fine- (or local-) and regional-scales is not known. Studies of the population genetic structure of cetaceans have found a surprising degree of differentiation despite the lack of obvious geographic barriers in the marine environment. While some studies described large panmictic populations (e.g., Kiszka et al. 2012; Moura et al. 2013; Thompson et al. 2016), others populations were subdivided into multiple genetically differentiated subpopulations that were restricted to single areas (e.g., Sellas et al. 2005; Ansmann et al. 2012). At larger scales, such patterns in population genetic structure may be influenced by geographic isolation, local genetic drift or isolation by distance (Hoelzel 1998).

136 Chapter 5 – Regional-scale genetic structure

The genetic population structure of Indo-Pacific bottlenose dolphins (Tursiops aduncus, referred to as ‘bottlenose dolphin’ hereafter) in coastal areas in Australia exhibits a variety of structure patterns and with a range of factors potentially affecting it. At fine or local-scales, genetic structure has been described within large embayments or estuaries along the east coast of Australia as well as in Shark Bay, Western Australia. Factors such as social behaviour, feeding specialisations, or human disturbance explained their structure (e.g., Krützen et al. 2004; Wiszniewski et al. 2010; Ansmann et al. 2012; Kopps et al. 2014). At regional-scales, changes of environmental conditions were associated with differentiation of bottlenose dolphins from the Spencer Gulf (Southern Australia) to coastal populations, the nearest of which was < 100 km from the opening of the Gulf (Bilgmann et al. 2007). Similarly, heterogeneous environments and local adaptation genetically were associated with differentiation of resident populations along the New South Wales (NSW) coast with only 23 km separating two of the populations (Möller et al. 2007; Wiszniewski et al. 2010). Over larger scales (> 100s km of coastline), Allen et al. (2016) documented a gradual differentiation that followed the isolation by distance model among seven populations sampled from Beagle Bay to Coral Bay in northwestern Australia.

The population genetic structure for bottlenose dolphins along the southwestern coastline of Western Australia has not yet been described, other than at local-scale in Perth (see Chapter 4). However, the ecology of bottlenose dolphins has been extensively studied in Bunbury (Smith et al. 2013; Sprogis et al. 2015; Smith et al. 2016; Sprogis et al. 2016) and Perth (Finn 2005; Finn and Calver 2008; Finn et al. 2008; Donaldson et al. 2010; Chabanne et al. 2012; Donaldson et al. 2012a, b). Populations from both localities exhibit year-round and long-term residency patterns and strong social structures, although emigration outside of the respective study areas was suggested (Smith et al. 2013; Sprogis et al. 2015, 2016; see Chapters 2 and 3, Chabanne et al. 2017a, b).

A recent study suggests that, historically, the Bunbury population was a genetic source to adjacent populations located to the north (Perth and Mandurah) and south (Busselton and Augusta) (Manlik et al. in prep-a). At a fine-scale, the Perth population showed weak genetic differentiation between socio-geographic subpopulations in which a source-sink dynamic system was suggested (see Chapters

137 Chapter 5 – Regional-scale genetic structure

3 and 4, Chabanne et al. 2017a). More specifically, the Swan Canning Riverpark (SCR) subpopulation in the estuary acts as a sink while a semi-enclosed embayment subpopulation (OA) acts as a source for SCR and adjacent subpopulations to the north (GR) and south (CS) (see Chapter 4). Presumably because of this strong source-sink dynamic, a Bayesian analysis implemented in STRUCTURE (Pritchard et al. 2000) did not identify a clear genetic structure, with some individuals presenting ancestry from an unknown population. At a regional-scale, the southern coastline of WA offers estuaries, embayments and protected bays for bottlenose dolphins (e.g., the Swan Canning Riverpark and Cockburn Sound in Perth, the Leschenault estuary and Koombana Bay in Bunbury, the Blackwood River estuary in Augusta and many more), suggesting that genetic population structure may also occur throughout this region.

5.2.2. Effective population size and effective/census population size ratio

In addition to the census population size (Nc, i.e., number of living animals), the effective population size (Ne) is an important parameter in understanding the ecology and evolution of natural populations and to inform conservation and management (Sollmann et al. 2013). The effective population size is the reciprocal of the rate of genetic change (inbreeding, heterozygosity, linkage disequilibrium) due to random processes in a finite population. In broad terms, Ne can be thought of as reflecting the mean number of breeding individuals contributing to offspring per generation. Ne can have a profound effect on population genetics as it can indicate whether a population may be at high risk of losing genetic variation (e.g., through genetic drift or inbreeding) which, in turn, may increase the risk of population extinction (Hare et al. 2011). Ne is ideally estimated from sex ratio, variation of lifetime reproductive output, variation of census population size, and other factors.

In an ‘ideal’ population (i.e., equal sex ratio; all animals are equally likely to produce offspring; mating is random; no immigration, emigration, mutation or selection), the ratio effective/census population sizes (Ne/Nc) should be close to one (Frankham

1995b). However, there is no reason why Nc and Ne should be the same. Ne can be larger or smaller than Nc, but Ne estimates are generally lower than Nc (Frankham

1995b; Palstra and Fraser 2012). Comparison of multiple Ne/Nc ratios with variation

138 Chapter 5 – Regional-scale genetic structure of some factors such as sex ratio, may help understanding the ecological factors that drive Ne below Nc, thus enhancing effective conservation and management decision- making (Kalinowski and Waples 2002).

In general, the necessary data to calculate Ne from sex ratio, variation of lifetime reproductive output, or variation of census population size, are not available, so it is necessary to back-calculate Ne from observed effects on genetic patterns (linkage disequilibrium LD, in the current study). Using LD information, researchers can estimate a ‘single-sample’ Ne called LD-Ne, which may also give an indication of Ne for buildup of inbreeding (inbreeding Nb) and for loss of heterozygosity (Waples et al. 2014).

Population Viability Analysis (PVA) can also be used to assess populations at risk and only demographic parameters (e.g., population size, reproduction, survival, Manlik et al. 2016) are required to perform it. However, as demonstrated in Chapter 2 (Chabanne et al. 2017b), estimates of census population size (i.e., abundance) and other parameters (i.e., apparent survival rate) based on mark-recapture analysis require a large amount of data which can be expensive and time-consuming to obtain (Tyne et al. 2016). Further issues may arise because of difficulties in accessing sites (e.g., in remote areas) or when barriers to gene flow for highly mobile animals are not known and hence the appropriate scale for a study area is not known (Gagnaire et al. 2015). For such situations, the ratio Ne/Nc could be used for the purpose of inferring Nc from Ne for inaccessible populations (Frankham 1995b; Luikart et al.

2010) --although first one must calculate the ratio Ne/Nc from other populations of same species that are known or assumed to have similar environmental conditions.

There has been an increasing effort applied to estimating Ne for wildlife populations (e.g., Cronin et al. 2009; Hamner et al. 2012; Brown et al. 2014; Dudgeon and Ovenden 2015; Zachos et al. 2016) However, only a few studies have aimed to evaluate Ne for bottlenose dolphin populations (Ansmann et al. 2013) despite the availability of numerous studies on their population census size (e.g., Chilvers and Corkeron 2003; Fury and Harrison 2008; Mansur et al. 2012; Smith et al. 2013; Webster et al. 2014; Sprogis et al. 2016). With the exception of the Bunbury and Perth dolphin populations (Smith et al. 2013; Sprogis et al. 2016; see Chapter 2,

139 Chapter 5 – Regional-scale genetic structure

Chabanne et al. 2017b), the census population (Nc) size of other populations along the southwestern coastline of WA are still unknown.

In this chapter, I first address the current gap in the knowledge about regional-scale genetic structure for Indo-Pacific bottlenose dolphins along the southwestern coastline of Western Australia (including Perth) by using 25 microsatellite loci to assess the current regional-scale population genetic variation and genetic structure as well as gene flow among eight localities spanning from Perth (WA) to Esperance (c. 1,000 km of coastline). Second, I examined the practicality of the linkage disequilibrium LD method used to estimate the effective population (Ne) size of the genetically distinct populations. I also assessed the ratio between effective population size and census population size (Ne/Nc) to add to the available data for this species in WA (Ansmann et al. 2013), and discussed whether Ne/Nc ratios have sufficient generality to be used in calculating Nc from Ne in other WA bottlenose dolphin populations.

5.3. Materials and methods

5.3.1. Study site and sample collection

Skin biopsy samples from free-living individual dolphins were collected from 2007 to 2011 along the southwestern coastline of Western Australia. Eight localities were sampled between Perth (S31.951; E115.860) and Esperance (S33.862, E121.890), covering about 1,000 km of coastline: the Swan Canning Riverpark (SCR) and Cockburn Sound (CS/OA; includes individuals from Owen Anchorage (OA)); Mandurah (MH); Bunbury (BB); Busselton (BS); Augusta (AU); Albany (AL); and Esperance (ES) (Figure 5.1). During dedicated and opportunistic small boat surveys, samples were taken using a remote biopsy system designed for small cetaceans by PAXARMS (Krützen et al. 2002). Calves assumed to be less than two years old (i.e., based on body length and approximate date of birth) were excluded from biopsy sampling. Tissue samples were preserved in saturated NaCl 20% dimethyl sulfoxide until DNA extraction (Amos and Hoelzel 1991).

140

Chapter 5 – Regional-scale genetic structure

Figure 5.1. Map of the sampling sites, southwestern Australia, showing the biopsy sample collection sites for bottlenose dolphins (n = 221) from

141 eight localities: blues = Perth (PE) including Swan Canning Riverpark (SCR) and Cockburn Sound (CS/OA); black = Mandurah (MH); green =

Bunbury (BB); orange = Busselton (BS); purple = Augusta (AU); grey = Albany (AL); and pink = Esperance (ES).

Chapter 5 – Regional-scale genetic structure

5.3.2. DNA extraction and Microsatellite genotyping

Total genomic DNA was extracted from the samples following standard phenol- chloroform protocol (Davis et al. 1986) or, alternatively, using the Gentra Puregene Tissue Kit (Qiagen). Samples were PCR genotyped at 25 polymorphic microsatellite loci: D22 (Shinohara et al. 1997), KWM12 (Hoelzel et al. 1998a), MK3, MK5, MK6, MK8, MK9 (Krützen et al. 2001), Tur_E12, Tur_F10, Tur4_66, Tur4_80, Tur4_87, Tur4_91, Tur4_98, Tur4_105, Tur4_108, Tur4_111, Tur4_117, Tur4_128, Tur4_132, Tur4_138, Tur4_141, Tur_142, Tur4_153, and Tur4_162, (Nater et al. 2009). Following the method described in Manlik et al. (in prep-b), microsatellite loci were amplified using the Qiagen Multiplex KitTM (Qiagen) in three multiplex polymerase chain reactions. PCR products were run on an ABI 3730 DNA Sequencer (Applied Biosystems) and analysed using GENEIOUS 9.1 (http://www.geneious.com; Kearse et al. 2012). Samples were checked for duplication (i.e., same microsatellite allele sizes, sex, sample locality, photo- identification) and some samples were removed from the dataset, making a total sample of 221.

I used the software Micro-Checker 2.2 (Van Oosterhout et al. 2004) to test for scoring errors due to stuttering and the presence of large allele dropouts across all loci and localities. The software INEST 2.0 (Inbreeding/Null Allele Estimation, Chybicki and Burczyk 2009) was used to estimate the frequency of null alleles at microsatellite loci within each locality and using a population inbreeding model. Manlik et al. (in prep-a) found no significant linkage disequilibrium in any of the locus-pairs. Although Hardy-Weinberg equilibrium (HWE) departures were found for four loci in two different localities, Manlik et al. (in prep-a) indicated that the loci which departed from HWE were different in each of the two localities, and the number of loci that departed was lower than expected, at the 5% significance level. Therefore, all 25 loci were used in the genetic analysis.

5.3.3. Genetic diversity

For each locality, genetic diversity was estimated by calculating the mean number of alleles (NA), observed (HO), expected (HE) and unbiased expected (uHE)

142 Chapter 5 – Regional-scale genetic structure

heterozygosities (Nei 1978) and the number of private allele (NPA) in GenAlEx 6.5

(Peakall and Smouse 2012). Mean allelic richness (AR) and inbreeding coefficient

(FIS) were calculated using FSTAT 2.9 (Goudet 2001). I also estimated FIS using a correction for null alleles in INEST.

Microsatellites were tested for evidence of recent bottleneck events using the software BOTTLENECK 1.2 (Cornuet and Luikart 1996). I assessed the significance of Wilcoxon sign rank tests for two models of generation of new alleles: the stepwise mutation model (SMM) and the two-phased model of mutation (TPM; variance = 30, 70% stepwise mutational model, 103 iterations). I also inspected the distribution of allelic frequencies to detect a mode-shift distortion due to the loss of rare alleles (Luikart et al. 1998).

5.3.4. Genetic differentiation and population structure

Genetic differentiation between pairs of sampling localities was estimated by FST values using Arlequin 3.5 (Excoffier and Lischer 2010). Significance testing was based on 104 permutations in Arlequin. Significance values for multiple comparisons were adjusted by sequential Bonferroni corrections (Rice 1989). I also tested for isolation by distance (IBD) by conducting a Mantel test comparing untransformed pairwise FST with untransformed geographic distances among localities. Geographic distances between sampling localities were measured in the most direct line through the water between the approximate centres of the areas where samples were collected. Significance testing was performed using 104 randomizations in IBDWS 3.2 (Jensen et al. 2005).

I used the Bayesian model-based clustering method STRUCTURE 2.3 (Pritchard et al. 2000) to assess the number of genetic clusters in the dataset (K). Individuals were assigned to a number of clusters within which Hardy-Weinberg equilibrium and linkage equilibrium were achieved. Markov chain Monte Carlo (MCMC) runs were conducted for K values ranging from one to ten, using a burn-in length period of 105 iterations, followed by 106 replicates. Ten independent runs were performed for each K. The analysis was performed with population information (i.e., locality) and using the correlated frequency and admixture models given the close geographical

143 Chapter 5 – Regional-scale genetic structure proximity of some localities. The most likely number of genetically homogeneous clusters (if K ≥ 2) was determined when the mean log probabilities (LnP(D)) among K values reached its maximum value. Because the most likely K was not clearly defined (see results section), I also calculated ΔK, a second-order rate of change of LnP(D), to confirm the most likely K (Evanno et al. 2005).

5.3.5. Gene flow

I used the program BayesAss 3.0 (Wilson and Rannala 2003) to estimate contemporary gene flow (< 5 generations). I conducted five independent runs using 107 iterations, a burn-in length of 106 and a sampling interval of 103 steps. To reach the recommended acceptance rates of total iterations between 20% and 60%, I adjusted the values of continuous parameters such as migration rates (m), allele frequencies (a) and inbreeding coefficient (f) to 0.4, 0.6 and 0.7, respectively. To confirm a convergence, each run was executed with different seed numbers, and trace files were examined for consistent oscillations using the software Tracer 1.6 (Rambaut et al. 2013).

5.3.6. Effective population (Ne) sizes

Using the linkage disequilibrium (LD) method implemented in LDNe 1.3 (Waples and Do 2008), I estimated the contemporary effective population sizes (Ne) of the populations identified by genetic differentiation and population structure. However, populations with less than 25 samples were discarded from the analysis because the

LD method is unreliable with small sample size (Waples and Do 2010). Ne estimates were obtained using the random mating model (Sugg et al. 1996). To avoid bias caused by rare alleles, I specified a criterion Pcrit for which alleles at lower frequencies were excluded. The choice of Pcrit depended on the sample size of each population such that 1/(2n) ≤ Pcrit (Waples and Do 2010). Bias due to overlapping generations (cNe) was also corrected by adjusting the estimates, following the method used in other studies (Ansmann et al. 2013; Louis et al. 2014): when Ne was estimated using Pcrit = 0.02, I adjusted Ne by adding 15% of it; with Pcrit = 0.01, I added 10%. For comparison, I also calculated cNe for Perth using the dataset from Chapter 4 (n = 109 samples). Although the number of samples used in Chapter 4 was

144 Chapter 5 – Regional-scale genetic structure much larger, loci defined in the dataset were different to those used in this study and less numerous (i.e., only ten were defined, see Chapter 4 for details). Genetic laboratory work was performed by two different research groups and at different time (i.e., laboratory work concluded before 2013 for the dataset used in this chapter). Tissue samples collected after 2013 (see Chapter 4) were not included in the database used for the genetic structure of this study.

5.4. Results

5.4.1. Genetic diversity

A total of 221 samples (including 15 from SCR, 24 from CS/OA, 25 from MH, 85 from BB, 19 from BS, 29 from AU, 14 from AL and 10 from ES) were kept after checking for duplicates (Table 5.1). No evidence for scoring errors due to stuttering or large-allele dropout was detected for any of the loci in any of the localities. Evidence for null alleles was found only for locus MK6 in AL, although the frequency of null allele was < 0.05 (r = 0.016) and thus considered negligible (Chapuis and Estoup 2007) (see Appendix A5.1). Manlik et al. (in prep-a) showed no genotypic linkage disequilibrium (LD) in any pairs of loci in any of the localities, indicating that the 25 loci were independently inherited. All 25 microsatellite loci were polymorphic across the entire dataset in each of the localities. Levels of genetic diversity were similar for each locality with average observed heterozygosities ranging from 0.53 in AU to 0.59 in three localities (SCR, BB, BS) (Table 5.1). Allelic richness ranged from 3.28 in AU to 3.70 in CS/OA. I found private alleles in all populations except MH, with frequencies of private alleles varying from 4% in

SCR, CS/OA, and BS (localities with a single private allele) to 16% in BB. FIS values were not significantly different from zero in any locality, although the overall

FIS value corrected for null alleles (INEST analysis) was significantly positive, indicating that there was a deficit of heterozygosity relative to HWE expectations as expected when pooling distant populations (Table 5.1).

145 Chapter 5 – Regional-scale genetic structure

Table 5.1. Genetic diversity for microsatellite loci in bottlenose dolphin sampled in eight localities.

Locality n NA AR NPA HO HE uHE FIS FIS corrected SCR 15 3.920 3.519 1 0.592 0.553 0.571 - 0.035 NS 0.011 NS CS/OA 24 4.400 3.697 1 0.579 0.574 0.587 0.006 NS 0.016 NS MH 25 4.200 3.390 0 0.536 0.547 0.559 0.041NS 0.025 NS BB 85 4.520 3.451 4 0.586 0.565 0.568 - 0.031NS 0.005 NS BS 19 3.880 3.287 1 0.591 0.532 0.546 - 0.062 NS 0.007 NS AU 29 4.160 3.279 3 0.532 0.534 0.543 0.020 NS 0.017 NS AL 14 3.600 3.556 2 0.554 0.513 0.533 - 0.040 NS 0.009 NS ES 10 3.680 3.628 3 0.568 0.525 0.554 - 0.026 NS 0.018 NS Overall 221 4.045 4.235 15 0.567 0.543 0.558 - 0.008 NS 0.016 *-

Note: n = number of samples, NA = mean number of alleles, AR mean allelic richness, NPA = number of private alleles, HO = observed heterozygosity, HE = expected heterozygosity, uHE = unbiased expected heterozygosity, FIS = inbreeding coefficient (all tested using Bonferroni correction: NS = non-significant at P-value > 0.05), FIS corrected = inbreeding coefficient corrected for null alleles (NS = non-significantly different from zero, * = significantly different from zero). SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound, PE = Perth, MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance.

Results from BOTTLENECK indicate an excess of heterozygosity (one-tailed Wilcoxon test for heterozygosity excess relative to numbers of alleles, which are lost faster than heterogeneity in a bottleneck (P-value < 0.05 after Bonferroni correction, Appendix A5.2), which suggests a recent reduction in effective population size of SCR and CS/OA pooled, and BB. Additionally, the graphical representation of the allele frequencies for all 25 polymorphic loci (see Appendix A5.2) shows a deficit of rare alleles (i.e., frequency < 0.1) in AL and ES, causing a mode-shift distortion (i.e., indication of recently bottlenecked populations, Luikart et al. 1998). The representation for all other localities did not show a typical ‘L-shape’ because more alleles were found in intermediate frequency classes than in the low frequency class.

5.4.2. Genetic differentiation and population structure

All FST pairwise comparisons were significant (Table 5.2) with FST values varying

from 0.026 to 0.113, except between SCR and CS/OA where FST was low and non-

significant after sequential Bonferroni correction (FST = 0.008, P-value = 0.0123).

146 Chapter 5 – Regional-scale genetic structure

The Mantel test revealed a positive and significant correlation between FST and geographic distances (R = 0.58, P-value <0.02, Appendix A5.3) among all sample localities. Geographic distance accounted for 34% of the variation in genotypic distance. In detail, I found significant isolation by distance (R = 0.89, P-value < 0.01, Appendix A5.3) between localities bordering the Indian Ocean (SCR, CS/OA, MH, BB and BS) with geographic distance accounting for 80% of the variation in genotypic distance. However, the test was non-significant between localities bordering the Great Australia Bight (AU, AL and ES, R = -0.93, P-value = 0.84, Appendix A5.3).

Table 5.2. Genetic differentiation (FST) among eight bottlenose dolphin sampling localities in southern Western Australia.

SCR CS/OA MH BB BS AU AL

SCR -

CS/OA 0.008*** -

MH 0.039*** 0.030*** -

BB 0.045*** 0.043*** 0.026*** -

BS 0.064*** 0.065*** 0.047*** 0.029*** -

AU 0.028*** 0.033*** 0.061*** 0.060*** 0.060*** -

AL 0.059*** 0.074*** 0.113*** 0.107*** 0.083*** 0.057*** - ES 0.075*** 0.051*** 0.096*** 0.081*** 0.072*** 0.045*** 0.060*** Note: * P-value < 0.05, ** P-value < 0.01, *** P-value < 0.001 after sequential Bonferroni correction. SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound, MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance.

The Bayesian clustering analysis implemented in STRUCTURE showed clustering of bottlenose dolphins from the southwestern coastline of Western Australia. Firstly, the mean posterior probability (LnP(D)) reached a plateau at K = 4 but slightly increased up to K = 7 clusters (Appendix A5.4). The ΔK index obtained by the method of Evanno et al. (2005) indicated two modes at K = 2 and K = 4 (Appendix A5.4). At K = 2 (Figure 5.2a), all individuals sampled in BB were strongly assigned to one genetic cluster (assignment probability q > 0.95) while all other localities were assigned to a second cluster (0.51 < q < 0.97). At larger K value (K = 4, Figure 5.2b), only three genetic clusters seemed relevant (i.e., sporadic representation of the fourth genetic cluster). Individuals sampled in SCR and CS/OA formed one cluster

147 Chapter 5 – Regional-scale genetic structure

(northern cluster with q > 0.70 for 72% of the individuals) while AU, AL and ES formed a second cluster (southern cluster), both clusters being distinctive to the BB genetic cluster. Individuals from MH and BS were of mixed ancestry between BB and the northern genetic cluster or the southern genetic cluster, respectively. Secondly, because of the long distance that separated the southern localities from one another, I ran an independent Bayesian clustering analysis to provide higher statistical power for detecting genetic clusters among AU, AL and ES localities. The LnP(D) and ΔK index was higher at K = 3 (LnP(D) = -2731.28, see Figures A5.4C and D), with majority of the individuals from each locality being assigned to a respective genetic cluster (Figure 5.2c). Altogether, STRUCTURE identified five populations, including SCR and CS/OA pooled into one population (referred to as Perth – PE hereafter). However, because of the admixture found in individuals from

MH and BS and the moderate levels of genetic differentiation with BB (FST ≈ 0.03, Table 5.2), populations MH and BS were considered to be two distinct populations for subsequent analyses.

148 Chapter 5 – Regional-scale genetic structure

Figure 5.2. Structure plots showing assignment probability of each dolphin (one individual per column) to the respective populations for different number of clusters, K: (a) with K = 2 and (b) K = 4 for the full dataset (n = 221); and (c) with K = 3 for Augusta, Albany, and Esperance only (n = 53). PE = Perth (including SCR = Swan Canning Riverpark and CS/OA = Cockburn Sound), MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance.

149 Chapter 5 – Regional-scale genetic structure

5.4.3. Contemporary gene flow

Based on the seven populations (i.e., including MH and BS as distinct populations) identified from the FST differentiation and STRUCTURE analyses, the BayesAss analysis of microsatellite genotypes (Table 5.3) indicated high migration from BB into MH (located c. 80 km north from BB) or BS (c. 40 km south from BB) with estimates of 24% and 20% migrants per generation, respectively. Similarly, there was high migration from AU to the southern localities with 23% and 18% of migrants per generation into AL (c. 300 km east from AU) and ES (c. 700 km east from AU), respectively. Other pairwise comparisons indicated lower migration rates ranging from 9% to > 1%, with the majority showing a 95% confidence interval that included zero. Specifically, contemporary migration from and into PE was little or negligible (i.e., 95% confidence interval included zero). The proportion of non- immigrants was the highest in BB (97%) following by PE (86%) and AU (85%).

150

Chapter 5 – Regional-scale genetic structure

Table 5.3. Mean (and 95% CI) recent migration rates inferred using BayesAss 3.0 (Wilson and Rannala 2003). The migration rate is the proportion of individuals in a population that immigrated from a source population per generation. Values of CI that do not overlap with zero are in bold. Values along the diagonal (underlining) are the proportions of non-immigrants from the origin subpopulation for each generation.

From Into PE MH BB BS AU AL ES PE 0.86 0.01 0.04 0.01 0.05 0.01 0.02 (0.79-0.93) (0.00 -0.03) (0.00-0.10) (0.00-0.02) (0.00 -0.11) (0.00-0.02) (0.00-0.05)

MH 0.03 0.68 0.24 0.01 0.02 0.01 0.01 (0.00-0.11) (0.66-0.70) (0.18-0.29) (0.00-0.02) (0.00-0.08) (0.00-0.02) (0.00-0.04)

BB 0.01 0.01 0.97 0.00 0.01 0.00 0.00 (0.00-0.08) (0.00-0.03) (0.90-1.00) (0.00-0.02) (0.00-0.065) (0.00-0.02) (0.00-0.03)

BS 0.01 0.01 0.20 0.68 0.070 0.01 0.01 (0.00-0.09) (0.00-0.03) (0.14-0.25) (0.67-0.69) (0.01-0.13) (0.00-0.03) (0.00-0.04)

AU 0.02 0.01 0.09 0.01 0.85 0.01 0.02 (0.00-0.09) (0.00-0.03) (0.03-0.14) (0.00-0.02) (0.79-0.91) (0-0.02) (0.00-0.05)

AL 0.02 0.02 0.02 0.02 0.23 0.68 0.02 (0.00 -0.09) (0.00-0.04) (0.00-0.07) (0.00-0.03) (0.17-0.29) (0.67-0.70) (0.00-0.05)

ES 0.03 0.02 0.042 0.02 0.18 0.02 0.69 (0.00-0.10) (0.00-0.04) (0.00-0.10) (0.01-0.03) (0.12-0.24) (0.01-0.03) (0.66-0.72)

151 Note: PE = Perth, MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance.

Chapter 5 – Regional-scale genetic structure

5.4.4. Effective population sizes

Using the genetic linkage disequilibrium method and correction for bias introduced by overlapping generations, the corrected effective population sizes (cNe) per locality varied from 23.0 (95% CI 19.7-27.3) in PE1 (i.e., PE locality using data from this chapter, n = 39 samples, 25 loci) to 104.6 (95% CI 85.1-132.3) in BB

(Table 5.4). For comparison, I obtained an estimate cNe of 75.8 (95% CI 64.6-90.2) for PE2, i.e., PE locality when using a larger sample size but fewer loci (n = 109 samples genotyped for 10 loci, see Chapter 4). However, there was a high positive correlation between the sample size n and Ne or cNe (R1 = 0.8830, P-value < 0.05 and

R2 = 0.8723, P-value = 0.05, respectively, see Appendix A5.5). Both slopes were near unity (slope1 = 0.8384 and slope2 = 0.8925), which means perfect prediction.

A seasonal mean census population size (Nc) is known for two populations, being 153.8 (95% CI 114.3-193.3) and 148.2 (95% CI 127.5-168.8) adults and juveniles (i.e., excluding calves) for PE and BB, respectively (i.e., means of the seasonal abundances estimated using the dataset described in Chapter 2, Chabanne et al.

2017b (but see Appendix A5.6), and in Sprogis et al. 2016, respectively). Nc values were not corrected for generation overlap as I did for Ne values.

When using a large sample size (n > 50), the mean ratios cNe/Nc between PE (n = 109, PE2) and BB (n = 85) were in the same range with 0.75 (SE 0.20) and 0.71 (SE 0.10), respectively.

On the assumption (discussed further below) that cNe/Nc ratios can be transferred between populations of this species in WA, I calculated Nc from cNe for two other populations. I used the cNe/Nc ratio from BB because it was based on the same markers as the populations in question. For MH, the estimate of census population size (Nc) based on cNe/Nc could range between 53.7-155.5. For AU, the estimate of census population size (Nc) based on cNe/Nc ratio could range between 52.3-125.8.

152

Chapter 5 – Regional-scale genetic structure

Table 5.4. Estimates of the contemporary effective population size before correction (Ne, 95% CI) and after correction (cNe, 95% CI), ratio cNe/Nc (SE) for localities with known census population size (Nc, , 95% CI), and estimated eNc (95% CI) for populations with unknown Nc. The critical value varied according to the sample size (n > 25, Pcrit = 0.02; n > 50, Pcrit = 0.01). cNe was corrected for overlapping generation by adding 15 or 10% of the estimates based on the Pcrit, respectively.

Estimates of sample size eNc Locality n Pcrit Ne cNe Nc * cNe/Nc estimated from cNe/Nc 20.0 23.0 153.8 0.15 PE1 39 0.02 - (17.1-23.7) (19.7-27.3) (114.3-193.3) (SE 0.04)

105.1 115.6 153.8 0.75 PE2 109 0.01 - (77.9-129.0) (85.7-141.9) (114.3-193.3) (SE 0.01)

50.6 58.2 82.0 MH 25 0.02 unknown - (33.1-96.0) (38.1-110.4 ) (53.7-155.5)

95.1 104.6 148.2 0.71 BB 85 0.01 - (77.4-120.3) (85.1-132.3) (127.5-168.8) (SE 0.10)

46.8 53.8 75.8 AU 29 0.02 unknown - (32.3-77.7) (37.1-89.3) (52.3-125.8) Note: n = sample size; PE1 = Perth using the current study dataset; PE2 = Perth using the dataset described in Chapter 4, MH = Mandurah, BB = Bunbury, AU = Augusta. * mean of the seasonal abundances were estimated using abundances estimated in the Chapter 2 (Chabanne et al. 2017b) and in Sprogis et al. (2016).

153

Chapter 5 – Regional-scale genetic structure

5.5. Discussion

This study provides a key regional-scale contribution to the understanding of the distribution of genetically differentiated populations of Indo-pacific bottlenose dolphins in nearshore areas in Western Australia (Bilgmann et al. 2007; Wiszniewski et al. 2010; Allen et al. 2016). It also investigated whether the LD method can be used in WA for Tursiops species to calculate the ratio between effective Ne

(corrected for generation overlap) and census Nc population size not corrected for generation overlap, a parameter considered as of interest to scientists, managers and governmental agencies.

5.5.1. Genetic diversity

Genetic diversity of bottlenose dolphins along the southwestern coastline of Western Australia was moderate and similar between each locality (Table 5.1). Estimates of the observed (HO) and expected (HE) heterozygosities were found to be in the same range as those obtained for other populations along the north western coastline of

WA (HO 0.538-0.667; HE 0.547-0.613, Allen et al. 2016) as well as along the east coast of South Africa (HE 0.54-0.59, Natoli et al. 2007). Wiszniewski et al. (2010) also estimated HO of 0.54-0.59 for bottlenose dolphins inhabiting the Port Stephens (New South Wales, Australia). However, those values were the lowest in comparison to adjacent populations along the New South Wales coast.

It is generally expected that after a population colonises a new habitat (which may also be described as a founder event, e.g., when an estuary becomes accessible from an adjacent coastline), there will be a loss of microsatellite diversity while the original population maintains its level of genetic diversity (Hoelzel 1998; Hoffman et al. 2009). Here, none of the study populations showed differences in microsatellite genetic diversity, suggesting that a recent founder event has not occurred. However, tests run in BOTTLENECK and the documented allele frequency distribution (Appendix A5.2) indicated that all populations went through a recent reduction of effective population size (Luikart et al. 1998). Although the results may be explained by individuals becoming more genetically closely related to individuals within than between populations (Moura et al. 2013), it is more likely that a continuing

154 Chapter 5 – Regional-scale genetic structure population decline is occurring at regional-scale. If declining, the viability of the populations along the southwestern coastline of WA might be of very high concern, in particular since Manlik et al. (2016) indicated that BB population was projected to decline and at risk of extinction.

5.5.2. Genetic differentiation and structure

Pairwise comparisons based on FST (Table 5.2) showed significant differences in nuclear DNA between all localities except between SCR and CS/OA. The lack of differentiation between SCR and CS/OA is in general agreement with the previous study (see Chapter 4) in which genetic differentiation between the socio-geographic subpopulations in the Perth metropolitan region was weak, although CS/OA samples in this current study included individuals from two of the previously defined socio- geographic subpopulations (OA and CS, Chapter 3, Chabanne et al. 2017a). The identification of individuals from OA (i.e., source population for SCR and CS as defined in the Chapter 4) supported the gene flow from the CS/OA locality defined in the current study to the SCR and may explain the negligible genetic differentiation

(microsatellite FST value of 0.008 in this current study, Table 5.2) in comparison to those found in Chapter 4 (microsatellite FST values of 0.023-0.025, P-value < 0.05, Table 4.4).

As for many mammal species (terrestrial or marine, e.g., Olsen et al. 2014; Yannic et al. 2015; Allen et al. 2016), the genetic population structure of bottlenose dolphins throughout the study area was characterised by a pattern of isolation by distance

(IBD) (Appendix A5.3). The values of microsatellite FST (Table 5.2), as well as mtDNA FST values (Manlik et al. in prep-a), clearly showed an increase of differentiation with an increase of geographic distance from CS/OA to BS, the southernmost of the western coastline populations (Figure 5.1). However, values of

FST between SCR or CS/OA and AU were lower despite a larger geographic distance (> 250 km) suggesting a closer ancestral link between those three localities. Alternatively, the bottlenose dolphins’ high mobility or ranging pattern allows them to travel long distances without interacting with populations that are closer geographically (e.g., if individuals moved along the coast in offshore areas, away from nearshore populations), thus preventing the development of a strong

155 Chapter 5 – Regional-scale genetic structure relationship between gene flow and geography. At a finer-scale, the IBD model (Appendix A5.3) did not explain the differentiation with localities bordering the Great Australian Bight, and all localities, except SCR, showed a larger differentiation from AL than ES (the easternmost locality). It is difficult to explain such differentiation patterns and it may be related to the respective small sample sizes (< 20 samples) affecting the power of the analysis. However, the overall IBD model for all localities along the southwestern coastline of WA was not rejected (Figure A5.2).

STRUCTURE analyses indicated a clear genetic population structure that differentiated the BB population from others (K = 2, Figure 5.2a and Appendix A5.4). However, genetic structure on a finer scale was apparent to the north and the south of the BB population, either by significant genetic differentiation FST between neighbour populations (Table 5.2) or by genetic admixture patterns (Figure 5.2b). This finding provides the basis for the recommendation for distinguishing seven populations. Two further points may be made. First, individuals from SCR and CS/OA were mostly assigned to one genetic cluster, which was in agreement with the weak to negligible FST value (FST < 0.01, P-value > 0.05, Table 5.2). Secondly, individuals from MH and BS were not assigned to one genetic cluster but showed some admixture in which BB represented one ancestral population. Both MH and BS showed a weak but significant differentiation from the central population at BB (FST = 0.026 and 0.029 with P-value < 0.001, respectively, Table 5.2). However, presumably because FST values < 0.03, no distinct genetic cluster was detected in STRUCTURE (Evanno et al. 2005; Latch et al. 2006).

5.5.3. Contemporary gene flow

Analysis of contemporary migration rates using BayesAss (Table 5.3) estimated significant rates of gene flow (i.e., different from zero) from BB to MH and BS as well as from AU to AL and ES. However, migration estimates on the opposite direction were not significant, suggesting asymmetrical migration between these pairs of populations. Migration rates of individual dolphins between all other pairs of populations were negligible. The contemporary migration pattern associated with BB

156 Chapter 5 – Regional-scale genetic structure was similar to that found historically (i.e., across tens to hundreds of generations, Manlik et al. in prep-a).

5.5.4. Contemporary effective population size and Ne/Nc ratio

Contemporary effective population sizes (Ne, Table 5.4) were much larger for the BB population than for others, although this apparent difference is probably due to the relatively low sample size representing the MH and BS populations (Waples and

Do 2010). Indeed, I found that Ne calculating using Waples and Do (2008) LDNe method or the corrected parameter (cNe) were highly correlated with the sample size

(N, see Appendix A5.5). In comparison to Ne estimates for the MH and BS populations, the estimate of Ne for the PE population (SCR and CS/OA combined) was lower than its sample size with a narrow confidence interval. Despite both the PE and BB populations having similar genetic diversities (Table 5.1) and having been through a bottleneck event (i.e., reduction of Ne in both populations, although the method is very dependent on chance, Luikart et al. 1998), the difference of sample size has a clear effect on Ne estimates. When using a larger sample size (n =

109, see Chapter 4), the Ne estimate for the PE population (cNe = 115.6, 95% CI

85.7-141.9) was similar to the Ne estimated for the BB population despite using fewer loci (ten loci, see Chapter 4). Other studies have shown that increasing the numbers of samples could have a greater effect on precision than increasing the number of loci (Dudgeon and Ovenden 2015). However, the comparison may not be adequate because of the use of different markers (i.e., microsatellites but with different loci). Ne estimated for the PE population in this study should be considered as a worst case scenario because a value of Ne < 50 describes populations at high risk of inbreeding depression, and thus extinction (Crnokrak and Roff 1999; Palstra and Ruzzante 2008).

The interpretation of Ne estimated via the linkage disequilibrium method in LDNe (Waples and Do 2008) is strongly influenced by the sample size as well as several biological features, including overlapping generations and gene flow between populations (Hare et al. 2011; Waples and England 2011; Waples et al. 2014). These issues suggest the following points in relation to this study.

157 Chapter 5 – Regional-scale genetic structure

First, I must correct for overlapping generations. Although the data were collected over a period of only five years, there is a high chance that multiple generations were sampled, as bottlenose dolphins are long-lived animals with low fecundity, late maturity and high survivorship to adult stages (Wells and Scott 2009). The Ne estimates obtained are therefore likely to be downwardly biased (Waples and Do 2010; Waples et al. 2014). Following methods described in Waples et al. (2014) and applied in several other studies (e.g., Ansmann et al. 2012; Brown et al. 2014; Louis et al. 2014), I corrected the estimates for a 10-15% downward bias (cNe), depending on the sample size of respective populations. Despite this correction, the estimates did not change much.

Second, I must deal with immigration (i.e., gene flow), which can result in reduced

Ne values because of an apparent increase in the LD signal amongst a mixed sample from the recipient population in which immigrants and residents are genetically differentiated (Waples and England 2011). However, the bias from the “no immigration” assumption can be considered negligible if the immigration rate (i.e., gene flow) is below a 10% threshold (Hastings 1993) and with the LDNe analysis performing adequately up to this threshold (Waples and England 2011). Based on the contemporary gene flow analysis implemented in BayesAss (Table 5.3), the BB population was the only population with an immigration rate falling under the 10% threshold (immigration rate < 3%), suggesting that only the cNe for BB could be described as unbiased. In line with the discussion above, the occurrence of immigration (> 14%) in the PE population would suggest that in addition to the bias associated with the sample size, the cNe estimate is also biased downwards due to immigration (Waples and England 2011).

Third, the estimate of Ne can be severely downwardly biased by sample size. Apparent LD is equally sensitive to random processes in the actual population (which is what one aims to assess), and random processes during sampling (which, of course, one does not want to assess). Therefore unless the sample is much larger than the likely Ne, sample size can have a serious effect on the Ne estimate. In this case, the sample size clearly dominated the estimate, to the extent that the sample size was an almost perfect predictor of the Ne or the Ne corrected for overlap (cNe).

Thus, these Ne estimates are highly unreliable, to say the least.

158 Chapter 5 – Regional-scale genetic structure

Census population (Nc) sizes for the BB and PE populations were obtained for periods covering 2007-2013 in BB (Sprogis et al. 2016) and 2011-2015 in PE (see Chapter 2, Chabanne et al. 2017b), and thus included overlapping generations. My attempt in producing a Ne/Nc ratio, therefore allows correlation from Ne corrected for generation overlap (cNe) and Nc that is not corrected for generation overlap. While defining a constant value for a Ne/Nc ratio may be a particularly valuable tool for wildlife conservation efforts at the population-level, temporal fluctuations occur in effective size and, thus, also in the Ne/Nc ratio (Palstra and Ruzzante 2008). In this study, cNe/Nc ratios were 0.71 for BB and 0.75 for PE populations – however, I acknowledge that: 1) sample size largely influences the estimates of Ne and 2) Ne estimates should apply to Nc from the previous generation if overlapping generations during sampling did not occur (Palstra and Fraser 2012). In addition, the estimate for the PE population was obtained using another dataset (i.e., more skin samples but fewer and different loci).

As in other marine animals (e.g., elasmobranchs, Dudgeon and Ovenden 2015), the ratio Ne/Nc for delphinids is expected to be near to one (Frankham 1995b; Portnoy et al. 2009). Their longevity along with low fecundity associated with long gestation and nursing periods, late maturity and high survivorship parameters may maintain Ne close to Nc (Portnoy et al. 2009). On that basis, the lower ratios found in this study suggest that populations along the southern coastline of WA may have suffered from bottleneck events (i.e., reduction of Ne, Luikart et al. 1998). It is difficult to interpret the Ne/Nc ratio without the availability of other studies for comparison, with the exception of the study by Louis et al. (2014) which reported a Ne/Nc ratio of 5-10% and therefore proposed that the bottlenose dolphin population inhabiting the coastal areas of Atlantic Ocean and the Mediterranean Sea was in a critical state.

Despite those issues, the Ne/Nc ratio has been suggested as a tool to estimate the census population size of other populations (Frankham 1995b; Luikart et al. 2010). Nonetheless, caution is required because of several unresolved assumptions and the different factors affecting populations (e.g., habitat factors or expansion and contraction). In this study, for example, census population size for MH and BS had values ranging from 53.7 to 155.5 and 52.3 to 125.8, respectively. However,

159 Chapter 5 – Regional-scale genetic structure immigration (> 30%, Table 5.3) occurred and the respective sample sizes were small, both downwardly biasing the cNe estimates.

5.5.5. Conclusions and future research

My study suggests that present bottlenose dolphin genetic structure patterns along the southwestern coastline of Western Australia may consist of two metapopulations. In the Indian Ocean, BB acts as a genetic source for the nearest northern (MH) and southern (BS) populations. In the Great Australian Bight, AU acts as a genetic source for AL and ES populations, although limited genetic connectivity was also found between AU and BS (c. 120 km apart). Although results from the bottleneck tests and the distribution of allele frequencies suggest a general decline of bottlenose dolphins in the southwestern region, such declines have only been projected from demographic data for BB population itself (Manlik et al. 2016). Such a trend would have a significant implication for evaluations of the vulnerability of the species at regional-scale and highlights the importance of continued monitoring of these populations, and of obtaining larger genetic sample size for more reliable results.

The population of bottlenose dolphins in Perth was defined by the presence of a panmictic pattern between the SCR and CS/OA. In addition, it is possible that the genetic connectivity between PE and other populations may have changed within the last few generations. While historically the PE population appears to have obtained asymmetric immigration from the BB population (Manlik et al. in prep-a), this study found little to no contemporary connectivity.

The current approach for estimating the effective population size (Ne) and the ratio effective and census population sizes (Ne/Nc) still possesses several critical limitations that must be resolved before the approach can be of broad utility. Thus, despite the high conservation management interest in the development of methods to estimate Ne and Ne/Nc, it appears that the LD method used in this study does not estimate the effective population size but instead is highly correlated with the sample size. Even if the sample size issue is set aside, the assumptions (i.e., overlapping generations and migration) associated with the methods to calculate Ne and Ne/Nc still are largely violated with bottlenose dolphins. In addition, while Waples et al.

160 Chapter 5 – Regional-scale genetic structure

(2014) methods may assist in minimising biases due to generation overlapping, the migration assumption and sample size dependence are still problematic. The difference found between the Ne estimated for PE population using two different datasets also suggested the need to understand better the limits associated with the choice of the markers (e.g., loci).

161 Chapter 5 – Regional-scale genetic structure

Appendices

Appendix A5.1. Genetic diversity for 25 microsatellite loci

Table A5.1.1. Genetic diversity indices for eight sampling localities for all 25 microsatellite loci. n = number of individuals, NA = number of alleles, HO = observed heterozygosity, HE = expected heterozygosity, FIS = inbreeding coefficient, r = frequency of null alleles. SCR = Swan Canning Riverpark, CS/OA = Cockburn Sound, MH = Mandurah, BB = Bunbury, BS = Busselton, AU = Augusta, AL = Albany, ES = Esperance.

Locality Locus n NA HO HE FIS r (95% CI) SCR MK6 15 7 0.733 0.798 0.081 -0.001 0.004 Tur4_117 15 4 0.467 0.490 0.049 -0.002 0.010 Tur4_98 15 2 0.400 0.329 -0.217 -0.001 0.006 Tur4_66 15 3 0.200 0.248 0.192 -0.002 0.010 E12 15 3 0.800 0.640 -0.249 -0.001 0.003 Tur4_108 15 2 0.533 0.457 -0.167 -0.001 0.007 Tur4_128 15 3 0.467 0.481 0.030 -0.001 0.007 Tur4_111 15 3 0.067 0.195 0.659 -0.004 0.035 Tur4_105 14 7 0.857 0.835 -0.026 -0.002 0.017 D22 15 2 0.733 0.500 -0.467 -0.001 0.005 Tur4_87 15 3 0.800 0.655 -0.222 -0.001 0.004 Tur4_91 15 5 0.733 0.757 0.031 -0.001 0.006 Tur4_138 15 4 0.800 0.745 -0.073 -0.001 0.004 Tur4_141 15 7 0.867 0.824 -0.052 -0.001 0.004 F10 15 4 0.733 0.576 -0.273 -0.001 0.003 MK8 15 6 0.600 0.712 0.157 -0.001 0.007 MK3 15 4 0.733 0.605 -0.213 -0.001 0.003 KWM12 15 6 0.733 0.579 -0.267 -0.001 0.004 MK9 15 4 0.533 0.679 0.214 -0.002 0.013 MK5 15 3 0.333 0.543 0.386 -0.004 0.032 Tur4_153 15 2 0.400 0.519 0.229 -0.002 0.012 Tur4_80 15 5 0.533 0.662 0.194 -0.002 0.009 Tur4_132 15 2 0.133 0.129 -0.037 -0.002 0.009 Tur4_142 15 3 0.800 0.640 -0.249 -0.001 0.003 Tur4_162 15 4 0.800 0.693 -0.155 -0.001 0.004 CS/OA MK6 23 8 0.783 0.763 -0.026 -0.001 0.005 Tur4_117 24 4 0.417 0.565 0.263 -0.003 0.023

Tur4_98 24 2 0.375 0.361 -0.040 -0.001 0.008

Tur4_66 24 3 0.292 0.263 -0.110 -0.001 0.007

E12 22 4 0.545 0.696 0.216 -0.005 0.048

Tur4_108 24 2 0.375 0.361 -0.040 -0.001 0.008

162 Chapter 5 – Regional-scale genetic structure

Locality Locus n NA HO HE FIS r (95% CI) CS/OA Tur4_128 21 3 0.286 0.370 0.228 -0.006 0.051 Tur4_111 21 4 0.429 0.361 -0.188 -0.002 0.011

Tur4_105 20 7 0.900 0.818 -0.100 0.000 0.002

D22 24 3 0.750 0.578 -0.298 -0.001 0.003

Tur4_87 24 3 0.625 0.582 -0.075 -0.001 0.005

Tur4_91 24 6 0.708 0.742 0.045 -0.001 0.008

Tur4_138 24 4 0.625 0.645 0.031 -0.001 0.005

Tur4_141 24 7 0.875 0.732 -0.196 0.000 0.002

F10 20 5 0.750 0.789 0.050 -0.002 0.010

MK8 24 6 0.750 0.687 -0.092 -0.001 0.004 MK3 24 5 0.708 0.722 0.019 -0.001 0.005

KWM12 24 6 0.625 0.675 0.074 -0.001 0.006

MK9 24 5 0.500 0.612 0.183 -0.002 0.014

MK5 24 5 0.708 0.701 -0.010 -0.001 0.007

Tur4_153 23 2 0.478 0.512 0.066 -0.003 0.026

Tur4_80 22 5 0.682 0.708 0.037 -0.002 0.008

Tur4_132 20 2 0.250 0.224 -0.118 -0.004 0.024

Tur4_142 22 3 0.591 0.561 -0.054 -0.002 0.012

Tur4_162 20 5 0.600 0.671 0.106 -0.003 0.020

MH MK6 25 8 0.800 0.792 -0.011 -0.001 0.008 Tur4_117 25 4 0.560 0.542 -0.032 -0.001 0.010 Tur4_98 25 2 0.400 0.492 0.186 -0.003 0.023 Tur4_66 25 3 0.120 0.117 -0.029 -0.002 0.015 E12 24 4 0.625 0.650 0.039 -0.003 0.029 Tur4_108 25 2 0.280 0.245 -0.143 -0.002 0.012 Tur4_128 22 4 0.500 0.582 0.141 -0.005 0.044 Tur4_111 22 2 0.182 0.169 -0.077 -0.007 0.069 Tur4_105 20 7 0.650 0.763 0.148 -0.002 0.018 D22 25 3 0.640 0.601 -0.065 -0.001 0.009 Tur4_87 25 3 0.280 0.383 0.270 -0.003 0.023 Tur4_91 25 8 0.800 0.784 -0.020 -0.001 0.006 Tur4_138 25 4 0.840 0.685 -0.226 -0.001 0.004 Tur4_141 25 7 0.720 0.757 0.048 -0.001 0.007 F10 23 5 0.783 0.654 -0.196 -0.001 0.008 MK8 24 4 0.750 0.613 -0.223 -0.002 0.012 MK3 24 4 0.625 0.676 0.075 -0.003 0.023 KWM12 24 6 0.583 0.690 0.155 -0.003 0.036 MK9 24 4 0.458 0.581 0.211 -0.004 0.047 MK5 25 3 0.680 0.666 -0.021 -0.001 0.011 Tur4_153 25 2 0.440 0.482 0.087 -0.002 0.015 Tur4_80 24 5 0.542 0.641 0.155 -0.004 0.038 Tur4_132 24 2 0.167 0.286 0.418 -0.006 0.101 Tur4_142 24 4 0.333 0.392 0.150 -0.004 0.043

163 Chapter 5 – Regional-scale genetic structure

Locality Locus n NA HO HE FIS r (95% CI) MH Tur4_162 23 5 0.652 0.733 0.111 -0.004 0.039 BB MK6 84 8 0.821 0.766 -0.072 0.000 0.001 Tur4_117 84 4 0.560 0.617 0.093 -0.001 0.005 Tur4_98 84 2 0.548 0.503 -0.089 -0.001 0.003 Tur4_66 84 2 0.036 0.035 -0.012 -0.002 0.004 E12 84 4 0.702 0.690 -0.018 -0.001 0.002 Tur4_108 84 2 0.083 0.080 -0.038 -0.003 0.006 Tur4_128 83 4 0.518 0.484 -0.071 -0.003 0.006 Tur4_111 84 3 0.464 0.403 -0.152 -0.002 0.003 Tur4_105 80 7 0.787 0.801 0.017 -0.003 0.006 D22 84 3 0.786 0.631 -0.245 -0.003 0.004 Tur4_87 84 3 0.548 0.506 -0.083 -0.004 0.005 Tur4_91 84 7 0.786 0.781 -0.006 -0.002 0.004 Tur4_138 84 5 0.702 0.746 0.059 -0.003 0.005 Tur4_141 84 8 0.774 0.772 -0.002 -0.002 0.003 F10 81 5 0.728 0.635 -0.147 -0.003 0.004 MK8 84 7 0.488 0.482 -0.013 -0.003 0.004 MK3 84 6 0.857 0.764 -0.121 -0.003 0.003 KWM12 84 7 0.667 0.699 0.046 -0.004 0.005 MK9 84 4 0.607 0.591 -0.028 -0.006 0.009 MK5 84 3 0.464 0.457 -0.015 -0.006 0.007 Tur4_153 84 2 0.464 0.472 0.016 -0.009 0.012 Tur4_80 84 5 0.548 0.583 0.060 -0.007 0.008 Tur4_132 83 2 0.458 0.471 0.028 -0.023 0.033 Tur4_142 83 4 0.554 0.521 -0.063 -0.012 0.016 Tur4_162 82 6 0.695 0.709 0.020 -0.010 0.014 BS MK6 21 6 0.810 0.692 -0.170 -0.001 0.002 Tur4_117 21 4 0.810 0.636 -0.273 -0.001 0.002 Tur4_98 21 2 0.476 0.417 -0.143 -0.001 0.005 Tur4_66 21 3 0.190 0.180 -0.060 -0.001 0.006 E12 20 4 0.750 0.674 -0.113 -0.001 0.009 Tur4_108 20 2 0.200 0.184 -0.086 -0.004 0.033 Tur4_128 20 3 0.600 0.591 -0.016 -0.002 0.016 Tur4_111 20 4 0.400 0.421 0.050 -0.003 0.028 Tur4_105 20 7 0.900 0.822 -0.094 -0.001 0.005 D22 21 3 0.762 0.570 -0.336 -0.001 0.003 Tur4_87 21 3 0.333 0.517 0.355 -0.003 0.018 Tur4_91 21 6 0.857 0.798 -0.075 -0.001 0.003 Tur4_138 21 5 0.762 0.770 0.011 -0.001 0.004 Tur4_141 21 8 0.857 0.844 -0.016 -0.001 0.003 F10 20 5 0.800 0.654 -0.223 -0.001 0.006 MK8 21 3 0.333 0.298 -0.120 -0.001 0.005 MK3 21 6 0.810 0.680 -0.191 0.000 0.002

164 Chapter 5 – Regional-scale genetic structure

Locality Locus n NA HO HE FIS r (95% CI) BS KWM12 21 7 0.762 0.724 -0.053 -0.001 0.003 MK9 21 3 0.429 0.531 0.193 -0.002 0.011 MK5 21 3 0.619 0.562 -0.102 -0.001 0.004 Tur4_153 21 2 0.333 0.498 0.330 -0.002 0.016 Tur4_80 21 2 0.476 0.417 -0.143 -0.001 0.005 Tur4_132 21 2 0.238 0.214 -0.111 -0.001 0.006 Tur4_142 21 4 0.381 0.392 0.027 -0.001 0.008 Tur4_162 21 6 0.810 0.758 -0.068 -0.001 0.003 AU MK6 29 9 0.759 0.784 0.033 -0.001 0.005 Tur4_117 29 3 0.586 0.584 -0.004 -0.001 0.007 Tur4_98 29 2 0.448 0.448 0.000 -0.002 0.010 Tur4_66 29 4 0.138 0.165 0.164 -0.002 0.012 E12 28 5 0.786 0.675 -0.164 -0.001 0.006 Tur4_108 29 2 0.345 0.334 -0.033 -0.001 0.009 Tur4_128 27 3 0.370 0.448 0.173 -0.005 0.058 Tur4_111 27 3 0.185 0.175 -0.057 -0.005 0.040 Tur4_105 25 7 0.720 0.812 0.113 -0.002 0.014 D22 29 3 0.586 0.517 -0.133 -0.001 0.007 Tur4_87 29 2 0.414 0.436 0.051 -0.002 0.011 Tur4_91 29 7 0.690 0.810 0.149 -0.001 0.012 Tur4_138 29 4 0.483 0.553 0.127 -0.002 0.011 Tur4_141 29 7 0.931 0.791 -0.178 0.000 0.002 F10 26 4 0.500 0.645 0.224 -0.006 0.065 MK8 29 6 0.724 0.666 -0.087 -0.001 0.005 MK3 28 6 0.571 0.636 0.102 -0.003 0.028 KWM12 29 5 0.655 0.692 0.053 -0.001 0.010 MK9 29 4 0.517 0.628 0.176 -0.002 0.013 MK5 29 3 0.724 0.602 -0.202 -0.001 0.004 Tur4_153 29 2 0.276 0.374 0.263 -0.002 0.019 Tur4_80 28 3 0.607 0.499 -0.218 -0.002 0.012 Tur4_132 27 2 0.185 0.171 -0.083 -0.005 0.046 Tur4_142 27 4 0.481 0.470 -0.024 -0.003 0.027 Tur4_162 27 4 0.630 0.674 0.066 -0.003 0.024 AL MK6 14 6 0.571 0.824 0.307 0.013 0.018 Tur4_117 14 3 0.714 0.588 -0.215 -0.001 0.007 Tur4_98 14 2 0.143 0.137 -0.040 -0.004 0.021 Tur4_66 14 3 0.500 0.418 -0.197 -0.005 0.014 E12 14 3 0.429 0.371 -0.156 -0.007 0.017 Tur4_108 14 2 0.357 0.302 -0.182 -0.012 0.024 Tur4_128 14 2 0.571 0.516 -0.106 -0.014 0.025 Tur4_111 14 3 0.357 0.319 -0.121 -0.014 0.025 Tur4_105 9 5 0.667 0.667 0.000 -0.034 0.053 D22 14 3 0.500 0.544 0.081 -0.022 0.035

165 Chapter 5 – Regional-scale genetic structure

Locality Locus n NA HO HE FIS r (95% CI) AL Tur4_87 14 3 0.286 0.409 0.302 -0.039 0.060 Tur4_91 14 7 0.857 0.802 -0.068 -0.013 0.018 Tur4_138 14 5 0.643 0.692 0.071 -0.030 0.040 Tur4_141 14 5 0.857 0.780 -0.099 -0.022 0.028 F10 11 4 0.727 0.714 -0.019 -0.077 0.097 MK8 14 4 0.786 0.604 -0.300 -0.030 0.037 MK3 14 3 0.429 0.585 0.268 -0.067 0.085 KWM12 14 4 0.786 0.626 -0.254 -0.038 0.044 MK9 14 4 0.571 0.563 -0.015 -0.041 0.050 MK5 14 4 0.857 0.640 -0.339 -0.034 0.040 Tur4_153 14 2 0.214 0.308 0.304 -0.109 0.135 Tur4_80 14 3 0.500 0.522 0.042 -0.082 0.096 Tur4_132 13 2 0.077 0.077 0.000 -0.282 0.395 Tur4_142 13 3 0.615 0.564 -0.091 -0.119 0.144 Tur4_162 12 5 0.833 0.742 -0.122 -0.075 0.085 ES MK6 9 7 0.667 0.889 0.250 -0.004 0.046 Tur4_117 10 4 0.900 0.744 -0.209 -0.001 0.006 Tur4_98 10 3 0.500 0.467 -0.071 -0.002 0.011 Tur4_66 10 2 0.100 0.100 0.000 -0.003 0.022 E12 9 5 0.556 0.785 0.292 -0.005 0.059 Tur4_108 10 2 0.300 0.267 -0.125 -0.002 0.014 Tur4_128 9 3 0.222 0.215 -0.032 -0.006 0.059 Tur4_111 9 2 0.333 0.292 -0.143 -0.005 0.048 Tur4_105 8 4 0.875 0.759 -0.153 -0.002 0.012 D22 10 3 0.400 0.489 0.182 -0.003 0.024 Tur4_87 10 2 0.600 0.522 -0.149 -0.002 0.012 Tur4_91 10 5 0.800 0.817 0.020 -0.001 0.008 Tur4_138 10 3 0.800 0.528 -0.516 -0.001 0.007 Tur4_141 10 4 0.600 0.594 -0.009 -0.002 0.011 F10 8 4 0.750 0.714 -0.050 -0.003 0.023 MK8 10 5 0.700 0.622 -0.125 -0.001 0.006 MK3 10 5 0.600 0.567 -0.059 -0.001 0.008 KWM12 10 6 0.600 0.772 0.223 -0.002 0.016 MK9 10 3 0.600 0.528 -0.137 -0.002 0.009 MK5 10 3 0.600 0.661 0.092 -0.002 0.012 Tur4_153 10 2 0.300 0.267 -0.125 -0.003 0.017 Tur4_80 10 4 0.700 0.722 0.031 -0.002 0.011 Tur4_132 10 2 0.100 0.100 0.000 -0.003 0.020 Tur4_142 10 3 0.600 0.617 0.027 -0.002 0.012 Tur4_162 8 6 1.000 0.804 -0.244 -0.001 0.007

166 Chapter 5 – Regional-scale genetic structure

Appendix A5.2. Bottleneck tests

Table A5.2.1. Summary statistics of the SMM (stepwise mutation model) and the TPM (2-phased model) tests of bottleneck (Cornuet and Luikart 1996). Wilcoxon test found significant heterozygosity excess (*) relative to numbers of alleles, which are lost faster than heterogeneity in a bottleneck, after Bonferroni correction (Pcrit = 0.007).

Wilcoxon test (P-value) Locality SMM TPM PE 0.560 0.001* MH 0.275 0.019* BB 0.653 0.002* BS 0.491 0.030* AU 0.958 0.020* AL 0.458 0.220* ES 0.974 0.474*

PE MH BB BS AU AL ES

0.4

0.2 Proportion of alleles ofProportion

0 0 0.2 0.4 0.6 0.8 1 Allele frequency class

Figure A5.2.1. Distribution of allele frequencies for each sampling locality of bottlenose dolphins. PE = Perth (blue), MH = Mandurah (black), BB = Bunbury (green), BS = Busselton (orange), AU = Augusta (purple), AL = Albany (grey), ES = Esperance (pink).

167 Chapter 5 – Regional-scale genetic structure

Appendix A5.3. Isolation by distance

Figure A5.3.1 Isolation by distance plot of correlation between untransformed genetic differentiation (FST) vs. untransformed geographic distance (km) among all sampling localities.

Mantel test with all sampling localities indicated a positive and significant correlation between genetic and geographic distances (R = 0.58, P-value < 0.02).

168 Chapter 5 – Regional-scale genetic structure

Appendix A5.3. (ongoing)

Figure A5.3.2. Isolation by distance plot of correlation between untransformed genetic differentiation (FST) and untransformed geographic distance (km) among sampling localities (a) bordering the Indian Ocean (Swan Canning Riverpark (SCR), Cockburn Sound (CS/OA), Mandurah (MH), Bunbury (BB) and Busselton (BS)) and (b) those bordering the Great Australia Bight (Augusta (AU), Albany (AL) and Esperance (ES))

(a) Mantel test between populations bordering the Indian Ocean (Swan Canning Riverpark (SCR), Cockburn Sound (CS/OA), Mandurah (MH), Bunbury (BB) and Busselton (BS)) indicated a positive and significant correlation between genetic and geographic distances (R = 0.89, P-value < 0.01). (b) Mantel test between populations bordering the Great Australia Bight (Augusta (AU), Albany (AL) and Esperance (ES)) indicated a positive and significant correlation between genetic and geographic distances (R = -0.93, P-value = 0.84).

169 Chapter 5 – Regional-scale genetic structure

Appendix A5.4. Mean of the posterior probabilities (LnP(D)) and ΔK statistic for STRUCTURE

(a) (b)

-12000 8

P(D) -12400

K 4 -12800 Δ 2

Mean of Ln ofMean -13200 0 0 2 4 6 8 10 0 2 4 6 8 10

-2600 4

-2700 3

P(D)

-2800 K

2 Δ

-2900 1 Mean of Ln ofMean -3000 0 0 2 4 6 0 2 4 6 K K

Figure A5.4.1. Mean of the estimated posterior probabilities (LnP(D)) and ΔK statistic (Evanno et al. 2005) over ten replicate runs for values of K = 1-10 using the Bayesian method in STRUCTURE and with (a and b) the full dataset (n = 221) or (c and d) excluding samples from Bunbury (BB) locality (n = 136).

170 Chapter 5 – Regional-scale genetic structure

Appendix A5.5. Correlation between sample size n and estimated effective population size (Ne and cNe)

160 140

120

) 100 e

, , cN e

(N 60 40 20

Estimated effective population size size populationeffective Estimated 0 0 20 40 60 80 100 120 Sample size (n)

Figure A5.5.1. Correlation between sample size n and estimated sample size as effective population size (Ne, black) or corrected effective population (cNe, red). Black trendline (R = 0.8384, P-value < 0.05); red trendline (R = 0.8925, P-value = 0.05).

171 Chapter 5 – Regional-scale genetic structure

Appendix A5.6. Abundances estimated for dolphins using the Perth metropolitan waters using single state Closed Robust Design models

Materials and Methods

For the purpose of this chapter, I run single state closed robust design (CRD) models to obtain estimates of the seasonal abundances of dolphins. In this scenario (Scenario 3), the dataset was modified so all individuals appeared to be captured within the same site.

In MARK, each CRD model combination was run with the probability of capture (p) varying by primary period, and with recapture probability (c) set as equal to first capture probability (p). The abundance (N) was set to vary by primary periods [N(primary periods)]. Several sub-models for apparent survival (φ) were run (i.e., whether it varied by primary periods or none). Temporary emigrations and re- immigrations (γ”, γ’) were also estimated whether it varied by primary periods or none, or with other constraints such as: Random model with γ "k = γ'k; a Markovian model where γ"k = γ"k-1 and γ' k = γ'k-1; or a mix of Random and Markovian models; and a No movement model with γ"= γ'= 0.

Models were ranked using the Akaike Information Criterion (AICc, Burnham and

Results

Models selection With no stratification (single site, CRD model), two models best fitted the data and attracted 74% of the AICc weight together. The first model described for constant apparent survival, emigration, and re-immigration while the second one had emigration varying by primary period (Table A5.6.1).

172

Chapter 5 – Regional-scale genetic structure

Table A5.6.1. Single state closed robust design models (in rank order of AICc scores). The table provides an overview of the Akaike Information Criterion corrected for small sample size (AICc), difference in AICc with best-fitting model and AICc weight, the number of parameters used in model fit and the deviance explained.

AIC Model Models AIC ΔAIC c Parameters Deviance c c weight likelihood Scenario 3 (one states) φ(.) γ"(.) γ'(.) p(period) N(period), p=c 1319.5 0.0 0.386 1.0000 35 6087.8 φ(.) γ"(period) γ'(.) p(period) N(period), p=c 1319.6 0.2 0.356 0.9235 49 6058.2 φ(.) γ"( period) γ'(period) p(period) N(period), p=c 1321.7 2.2 0.126 0.3265 61 6034.4 Random - φ(.) γ"(period) γ'(period) p(period) N(period), p=c 1322.2 2.7 0.099 0.2574 48 6062.9 φ(.) γ"(.) γ'(period) p(period) N(period), p=c 1324.5 5.1 0.031 0.0794 48 6065.3 Markovian - φ(period) γ"( period) γ'(.) p(period) N(period), p=c 1330.6 11.2 0.001 0.0038 62 6041.2 Note: φ apparent survival; γ" emigration; γ' re-immigration; p probability of capture; p = c probability of capture is equal to recapture; N

abundance; (.) constant; (period) varying by primary period. Some models have constraints: Random model in which γ "k = γ'k; Markovian model where γ"k = γ"k-1 and γ' k = γ'k-1.

173

Chapter 5 – Regional-scale genetic structure

Abundance estimates

With a capture probability varying from 0.10 to 0.25 (mean = 0.17, SE 0.01), the model yielded a constant apparent survival rate of 0.87 (SE 0.02, 95% CI 0.82-0.91). The model also estimated an emigration rate of 0.09 (SE 0.02, 95% CI 0.05-0.14) and a reimmigration rate of 0.70 (SE 0.09, 95% CI 0.51-0.84). Mean seasonal abundance of dolphins using the study area was 153.8, SE 5.0 (95% CI 114.3-193.3) and ranged from the lowest estimate of 94 (SE 16.8, 95% CI 66-133) in winter 2014 to the highest estimate of 224 (SE 37.9, 95% CI 162-312) in winter 2011, both corrected for unmarked individuals (Figure A5.6.1).

350

300

95% CI) 95%

± 250

200

150

100

50 Estimated abundance (Ntotal (Ntotal abundance Estimated

Winter Winter Winter Winter

Spring Spring Spring Spring

Autumn Autumn Autumn Autumn

Summer Summer Summer Summer 2011 2012 2013 2014 2015

Seasons

Figure A5.6.1. Seasonal estimated abundances (N̂total ± 95% confidence intervals) of dolphins in Perth metropolitan waters (defined as one site).

174

Chapter 6. General Conclusions

I have detailed a set of theoretical and practical methods that were applied in research conducted for this thesis (and for associated projects in the case of the southwestern WA genetics work) to assess the current status of bottlenose dolphins residing in an estuary (the Swan Canning Riverpark - SCR), two embayments (Cockburn Sound and Owen Anchorage – CS and OA, respectively) and an adjacent area of open coastline (Gage Roads - GR) within Perth metropolitan waters.

This study was prompted by an unusual mortality event in the SCR estuary in 2009, and the lack of understanding about the ecological and genetic connectivity between previously defined resident subpopulations of bottlenose dolphins in Perth metropolitan waters (Finn 2005; Chabanne et al. 2012).

Through the four data chapters, I have: 1. Provided estimates of abundance, apparent survival and demographic movement between geographic regions within the Perth metropolitan waters using a multistate modelling approach (Chapter 2, Chabanne et al. 2017b); 2. Combined information on social structure (i.e., social affinity and network), home ranges, residency patterns and genetic relatedness to identify the existence of local populations (i.e., ecological differences, Chapter 3, Chabanne et al. 2017a); 3. Demonstrated that the relationship between socio-geographic structure and genetic structure is not a straightforward relationship but a complex process at temporal and spatial scales (Chapter 4); and 4. Assessed the dispersal along the coastline (i.e., broader regional-scale) that the Perth metropolitan region belongs to and examined the efficacy of the

effective population size (Ne) and the ratio of effective/census population

sizes (Ne/Nc) as conservation tools (Chapter 5).

Taken together, the information in these four data chapters greatly improves the scientific basis for decision-making on the conservation and management of bottlenose dolphins in Perth metropolitan waters. I have indicated in each of the

175 Chapter 6 – General Conclusions chapters, and will address again here, how this information provides valuable and targeted guidance for management authorities and local industries to inform management strategies with the aim of conserving not a single overall population of T. aduncus in Perth metropolitan waters, but multiple distinct subpopulations associated with particular locales. Below is a list of management strategies that will be better informed and improved based on the information from this research:  Environmental Impact Assessment (EIA) of proposed coastal and estuarine developments;  The design and construction-phase management of an "outer harbour" port in CS (Western Australian Planning Commission 2004);  Formal management plans for the SCR and CS/OA (Department of Environment 2003; BMT Oceania 2014);  Management actions for marine reserves (e.g., Shoalwater and Marmion Marine Parks, Department of Conservation and Land Management 1992; Department of Environment and Conservation 2007);  Species-specific management actions (e.g., feeding of dolphins in CS, anti- disturbance or entanglement initiatives, Donaldson et al. 2012a).

In the following sections of this chapter, I summarise the key findings relevant to the four subpopulations described in this thesis. I note that from this point on, however, I will avoid using the term ‘subpopulation’ for the bottlenose dolphins associated with GR (the open coastline area north of Fremantle) based on their ecological characteristics (below). I then discuss some of the limitations I encountered during the study and, specifically, whether they were practical in nature (e.g., relating to the time and logistics of sampling) or related to the principles of the methods used in this research. Finally, I provide specific management recommendations that should be given due consideration by relevant management authorities and local industries.

6.1. Keys research findings for each subpopulation (Figure 6.1)

Considerable differences in population size and ecological parameters (i.e., social affinity, home ranges) were found between the subpopulations inhabiting the metropolitan waters of Perth, particularly between the estuarine subpopulation (SCR) and the dolphins associated with the open coastline north of Fremantle (GR).

176 Chapter 6 – General Conclusions

6.1.1. The Swan Canning Riverpark estuarine subpopulation

Both the MSCRD and social structure analyses (Chapters 2 and 3, Chabanne et al. 2017a, b) showed evidence of a stable, but small subpopulation of bottlenose dolphins in the estuary (< 25 individuals). Their high apparent survival rate (0.98) indicated an almost complete lack of permanent emigration during the study, although the few movements detected in and out of the SCR generally involved individuals from the estuary visiting the semi-enclosed embayments (CS/OA). The inconsistency between those results and the occurrence of genetic dispersal in the opposite direction (from OA to SCR, Chapter 4) refers to different timescales involved: social is on a timescale less than one generation; assignment of individuals to genetic clusters is on a one-generation timescale; and the genetic differentiation s indices (e.g., FST, HUA) are on a multi-generation timescale.

Other characteristics included a high degree of social stability, few interactions with individuals from adjacent subpopulations, individual home ranges almost exclusively encompassed within the estuary, and year-round residency (i.e., occupancy throughout all four seasons) and long-term site fidelity (i.e., several decades, Chabanne et al. 2012). The Fremantle Harbour Port and other areas in the lower reaches of the estuary were defined as hot spots (i.e., core areas representing 50% of the kernel density estimation). All those characteristics were comparable to those reported in a previous study (Chabanne et al. 2012) and suggest that the SCR subpopulation may experience some degree of demographic isolation (Chapters 2 and 3, Chabanne et al. 2017a, b), which would make it vulnerable to long-term population decline, possibly leading to local extinction of bottlenose dolphins in the estuary. The small size of the subpopulation (and particularly the small number of resident females) emphasises the potentially catastrophic effects of extreme stochastic events, such as the unusual mass mortality event in 2009 (Holyoake et al. 2010).

The SCR subpopulation was not found to be genetically isolated, despite the higher potential for inbreeding (Chapter 4) and higher genetic relatedness within that subpopulation than within any of the other adjacent subpopulations (Chapter 3, Chabanne et al. 2017a). Moreover, in the context of all the findings, the SCR

177 Chapter 6 – General Conclusions subpopulation may currently act as a sink within the Perth metropolitan metapopulation and, thus, the SCR subpopulation would not likely survive without the migration of individuals from another source, i.e., the adjacent subpopulations. In the extreme scenario of an effective population size < 50, as was estimated for Perth (Chapter 5), there is a plausible risk of local extinction (Crnokrak and Roff 1999; Palstra and Ruzzante 2008).

6.1.2. The semi-enclosed embayment subpopulations: Owen Anchorage and Cockburn Sound

Both the OA and CS subpopulations presented similar characteristics to each other (Chapters 2 and 3, Chabanne et al. 2017a, b), with stable population sizes (c. 103 individuals combined; 43 and 63 individuals in OA and CS, respectively). Although most individuals from both subpopulations showed a high degree of residency (i.e., year-round, with long-term site fidelity), the apparent survival was lower than 0.9 (Chapter 2, Chabanne et al. 2017b), suggesting that some individuals may move permanently into adjacent waters further south, such as Shoalwater Bay and Warnbro Sound (Green 2011) or perhaps west of Garden Island.

However, the OA and CS subpopulations were socially independent with limited interactions occurring between subpopulations (Chapter 3, Chabanne et al. 2017a). In addition, the home ranges of individuals within each subpopulation were mainly limited to their respective geographic regions and, viewed cumulatively, indicated distinct core areas of ranging that extended east and west of the shipping channel for the OA subpopulation and within the Kwinana Shelf area for the CS subpopulation (see Figure 3.5a, Chapter 3, Chabanne et al. 2017a).

While the weak genetic differentiation documented between the OA and CS subpopulations (i.e., genetic input from adjacent subpopulation, Chapter 4) may simply reflect the current social structure (Chapter 3, Chabanne et al. 2017a), it is more conservative to treat each subpopulation as a distinct ‘unit to conserve’. A historical genetic event (i.e., bottleneck event) was found for the CS subpopulation, and my results indicated that contemporary gene flow occurs in one direction (from

178 Chapter 6 – General Conclusions

OA to CS), suggesting a more vulnerable subpopulation (CS) and one (OA) that is vital to conserve, for the sake of GR, CS and SCR subpopulations.

6.1.3. Bottlenose dolphins occurring along the open Gage Roads coastline

In contrast to the three other subpopulations, the ecological and genetic characteristics of bottlenose dolphins observed along the open coastline north of Fremantle do not readily support a conclusion that bottlenose dolphins in this region constitute a ‘unit to conserve’. Such structure difference between open coastline and estuarine systems is not unusual and has been reported elsewhere (e.g., California coastline, Defran and Weller 1999; e.g., Texas coastline, Henderson 2004). Nonetheless, the conservation of these dolphins may be critical (as a source population) to the persistence of other subpopulations inhabiting the embayment and estuary systems in Perth metropolitan waters.

Bottlenose dolphins in the GR area along the open coastline north of Fremantle were abundant (the maximum population size estimated was c. 172 individuals), but there was no obvious seasonal pattern occurrence and the majority of the individuals being seen only once within the entire four-year study (Chapter 2, Chabanne et al. 2017b). Although a fission-fusion society seemed to best describe the social structure of dolphins in this area, their ecological characteristics principally indicated weak site fidelity and more transient behaviour that that documented for the three subpopulations to the south (Chapters 2 and 3, Chabanne et al. 2017a, b).

The small sample size of biopsies made it difficult to draw definitive conclusions about the genetic characteristics of the bottlenose dolphins in this area (Chapter 4). While my results suggested a lack of genetic differentiation from the other subpopulations in Perth metropolitan waters, it is most likely that bottlenose dolphins in this area form a larger population that extends further north. This inference is supported by the core area of home ranges being located along the northern edge of the study area (Chapter 3, Chabanne et al. 2017a) and, further, by the findings from previous boat-based photo-identification surveys conducted between 1991 and 1993 (Waples 1997).

179 Chapter 6 – General Conclusions

6.1.4. Perth metropolitan waters dolphin population at a regional-scale

In my thesis, I also examined the genetic structure of bottlenose dolphins along the southwestern coastline of Western Australia using microsatellite markers (Chapter 5) and with a focus on a contemporary temporal scale. It appears that the Perth population has been more isolated in the last few generations, with limited dispersal from adjacent populations to the south. This result suggests a change in the populations’ dynamic, as Manlik et al. (in prep-a) indicated that, historically, the Bunbury population (located c. 180 km south of Perth) was the source population for the southwestern coast, including the Perth metropolitan waters dolphin population.

The weak genetic structure and the source-sink dynamic found at a fine-scale (Chapter 4) suggests that the Perth population might appropriately be described as a metapopulation (cf. Wells and Richmond 1995). Such a conclusion, however, requires further work, including research to: 1) identify the unknown ancestral population found in admixed individuals and 2) assess the genetic connectivity with populations located further north or offshore.

Effective population size estimates for Perth dolphins were not conclusive (Chapter

5). One estimate suggested some risk of local extinction (i.e., Ne < 50) while the second was estimated in the same range as that of the Bunbury dolphin population

(i.e., Ne > 100), although the second estimate was obtained based on a larger sample size and included individuals from the open coastline north of Fremantle (dataset from Chapter 4).

180

Chapter 6 – General Conclusions

Figure 6.1. Summary overview of the conservation status of bottlenose dolphins in the four geographic regions of Perth metropolitan waters: 181

SCR = Swan Canning Riverpark (red), CS = Cockburn Sound (yellow), OA = Owen Anchorage (yellow), GR = Gage Roads (green).

Chapter 6 – General Conclusions

6.2. Theoretical and practical limitations of the study

All the methodological and analytical approaches applied in this thesis have inherent limitations either practicability, in their assumptions, or available sample sizes. However, their evaluation here in the context of this study should be of broader interest for the future study, conservation and management of small cetaceans in coastal and estuarine systems and, in some cases, for terrestrial and marine wildlife generally.

6.2.1. Fine-scale delineation of subpopulations

The photo-identification surveys were designed according to the different habitats (i.e., topography and bathymetry) identified prior to the study (i.e., the four geographic regions). Using the data collected from those surveys, I used the Multistate Closed Robust Design (MSCRD, Chapter 2, Chabanne et al. 2017b) supplemented by social network and home range analyses (Chapter 3, Chabanne et al. 2017a) to identify the presence of subpopulations of Indo-Pacific bottlenose dolphins in Perth metropolitan waters.

While the social and home range analyses clearly suggested four subpopulations, each associated with a different geographic region, the MSCRD approach did not allow inference as to that structure because of a key assumption inherent to the approach – namely, that individuals should not transit between geographic regions during secondary occasions and within a primary period (Arnason 1972, 1973). When considering the four geographic regions within the MSCRD models, 5% of the captures violated this assumption, mostly because of individuals moving between CS and OA. It was then more appropriate (theoretically) to pool data of both CS and OA regions to minimise the violation of the transit assumption and to obtain less bias in the estimates of survival rates and abundances (O'Connell-Goode et al. 2014).

6.2.2. Ecological characteristics of Gage Roads: are there sufficient data?

Mark-recapture, social structure and kernel density estimation (KDE) for home range analyses are generally not suitable for highly mobile populations or populations with

182 Chapter 6 – General Conclusions a large number of transient individuals, mostly because of a lack of individual data (i.e., insufficient sighting history data per individual). In my study, this issue arose for dolphins sighted along the open GR coastline north of Fremantle. The estimates of abundance were variable, with high variance in capture probabilities (Chapter 2, Chabanne et al. 2017b). I was also unable to document their social structure with clarity (Chapter 3, Chabanne et al. 2017a), because few individuals were sighted more than five times (i.e., a threshold used to avoid potential bias from small individual sample size, Jennions and Møller 2003).

In the context of this thesis, however, the results showed a marked difference in the ecology of bottlenose dolphins in that locale compared to those in the other study regions. In future, more capture-recapture data are required to obtain more reliable estimate of abundance and other demographic characteristics for dolphins in the open coast waters.

6.2.3. Genetic sampling is not random

Obtaining tissue samples from free-ranging and highly mobile species that spend much of their time underwater is challenging. In most cases, genetic sampling is conducted during systematic and randomised surveys (e.g., Ansmann et al. 2012). However, the samples themselves are not necessarily randomly collected. For bottlenose dolphins, genetic sampling often targets individuals that can be easily identified (i.e., those with well-marked dorsal fins), as well as those with some photo-identification history (i.e., multiple captures). Time and cost also contribute to the ‘non-random nature of biopsy’ sampling, with multiple samples often taken from the same area or social group, if previous attempts have not disturbed the dolphins. This applies to how genetic sampling was conducted in my study and explains a lack of samples from bottlenose dolphins sighted along the GR coastline north of Fremantle. In some instances, behavioural heterogeneity between age classes and sexes, and perhaps prior exposure to boating activity, may contribute to non-random biopsy sampling (e.g., Quérouil et al. 2010).

The consequences of the non-random nature of the genetic sampling process are numerous and include (but are not limited to): 1) possible inaccurate representation

183 Chapter 6 – General Conclusions of the true sex ratio; 2) limited sex-specific interpretations (e.g., abundances, survival rate); and 3) a lack of power in estimating parameters and defining patterns

(e.g., genetic structure, differentiation FST).

6.2.4. Contemporary effective population size (Ne) and ratio of effective/census population size (Ne/Nc): unresolved conservation tools

Theoretically, the contemporary effective population size (Ne) is an important variable that needs to be estimated to properly evaluate the ecology and evolution of natural populations and, thus, to inform conservation planning and management with smaller Ne values indicating faster loss of genetic variation upon which adaptation depends (Frankham et al. 2010). Similarly, the ratio of effective/census population sizes (Ne/Nc) can be a valuable tool for wildlife conservation efforts at the population-level, especially when data collection may be difficult to undertake

(molecular or mark-recapture). In this research, I aimed to estimate Ne and assess the extinction risk of the overall Perth metropolitan waters dolphin population. I also aimed to estimate the Ne/Nc ratio to investigate the census population size (Nc) of other populations along the southwestern coastline of WA that are not well-studied. Several points can be made around the lack of success in achieving these objectives.

First, the results clearly indicated a high correlation between the sample size (n) and the estimates of Ne for each population, suggesting that Ne was not successfully estimated as the numbers of breeders in a population but, rather, simply another sample size estimate. Therefore, estimates of any parameter associated with Ne (e.g.,

Ne/Nc ratio and estimation of Nc via the ratio) are currently inaccurate because of the sample size correlation.

Second, Ne and the Ne/Nc ratio remain challenging to determine for long-lived and highly mobile animals, despite efforts to correct the estimates for overlapping generations (Ansmann et al. 2013; Brown et al. 2014; Waples et al. 2014) and to define biases due to migration (Waples and England 2011).

184 Chapter 6 – General Conclusions

Third, a large number of molecular samples for any putative population are required in order to obtain reliable estimates and I only had a small number of samples available.

Finally, there are few comparable studies that might assist in evaluating the appropriateness of the parameters used in this study. More work is therefore required in the estimates of Ne and the Ne/Nc ratio to properly interpret and use them as indicators for contemporary conservation planning and management.

6.3. Management recommendations

In my research, I found that all subpopulations around Perth were connected, as in a metapopulation (Wells and Richmond 1995). This is an important dynamic for the long-term persistence of the subpopulations (Wiens 1976), particularly the SCR subpopulation which experienced an unusual mortality event in 2009 (Stephens et al. 2014). It is, therefore, important to minimise anthropogenic activities that would cause habitat fragmentation or otherwise interfere with dolphin movement patterns within the region’s coastal and estuarine matrix.

However, it is also important to monitor each subpopulation as a distinct ‘unit to conserve’, given evidence of variation in ecological and demographic parameters and thus, potentially, differing pressures from local stressors and mortality factors. Additionally, subpopulations represent a functioning ecological component of the environments they are associated with and support economic (e.g., tourism) and cultural values (i.e., dolphins have a high conservation value throughout the Perth metropolitan area because of their iconic status and high public profile). The decline or local extinction of bottlenose dolphins would mean that: i) few (or no) dolphins would be present to maintain the ecological function of dolphins in those ecosystems; and ii) tourism industries and amenity or aesthetic values that depend on the presence of dolphins would be jeopardised.

Therefore, over the long-term, there is a justifiable reasoning that:

185 Chapter 6 – General Conclusions

1) Each subpopulation should be considered as a ‘unit to conserve’. The lack of fine-scale genetic population structure should not lead to the conclusion that no population structure exists. A source-sink dynamic system may explain such results, in which case management strategies should ensure the protection of the source subpopulation. Additionally, any ecological differences between subpopulations must be considered and, in some circumstances, may be as informative as genetic differences (e.g., Taylor 2005).

2) High use areas for dolphins should be considered for the highest level of protection. Core areas have ecological significance for dolphins as foraging habitat, nursery area (Finn 2005; Finn and Calver 2008; Chabanne et al. 2012), breeding (e.g., OA) and passage (i.e., transit) areas (Chapter 3, Chabanne et al. 2017a). In Chapter 3 (Chabanne et al. 2017a), I examined the risks of direct and indirect threats caused by human activities on bottlenose dolphins. If anthropogenic impacts occur in core areas, it may mean that a subpopulation becomes demographically isolated (e.g., SCR subpopulation), or that ecologically vital processes (i.e., foraging, nursing) are repeatedly disturbed, resulting in negative effects on the viability of the subpopulation.

3) The small resident subpopulation in the Swan Canning Riverpark needs additional protection, a recommendation supported by research conducted for this thesis indicating its ‘at risk’ status based on subpopulation characteristics (Figure 6.1) and other recent research on the subpopulation (Marley et al. 2016). Marley et al. (2016) indicated that, despite being frequently exposed to vessel traffic, bottlenose dolphins show varying levels of tolerance to acoustic disturbance associated with particular areas in the Swan Canning Riverpark.

Marine mammals are already protected under Australia’s EPBC Act, but more formal protection may be required. By way of example, the Adelaide Dolphin Sanctuary was established in 2008 in South Australia to conserve bottlenose dolphins facing similar threats as those that occur in the SCR, e.g., entanglements in fishing gear, vessel strikes, and pollution (Department of Environment and Heritage 2008; Steiner and Bossley 2008). The sanctuary, in association with education efforts, has proved to be effective as abundance of dolphins within the

186 Chapter 6 – General Conclusions

sanctuary is increasing (Bossley et al. 2017). Implementation of similar formal protection would help integrate management efforts across stakeholder groups and focus community support on protecting dolphins and their habitat.

4) The semi-enclosed embayment (counting Owen Anchorage and Cockburn Sound as a single system), which includes both the OA and CS dolphin subpopulations, is also important based on sighting frequency, residency and long-term site fidelity and potential impacts on dolphins should be adequately considered in Environmental Impact Assessment for any development proposal. In particular, OA subpopulation acts as a source for all the others subpopulation s. It is therefore critical to protect this subpopulation for the survival of other subpopulations.

Although all of the above can lead to a view of looking at potential disturbances and how changes in those could enhance habitat and potentially increase the population size, consistent on-going monitoring of the subpopulations and their environmental conditions is necessary for their effective conservation (Reeves et al. 2003; Hammond 2010). Moreover, it is important that local management agencies, NGOs, industries and other stakeholders become involved in the development of conservation processes. Educational programs that enhance public awareness about human impacts on dolphins and the marine environment in general are also advisable. Within the last decade, some significant education efforts have been implemented in Perth, notably the Dolphin Watch project (Government of Western Australia 2017). Dolphin Watch is a community-based engagement program that aims to:

 Educate the public about the bottlenose dolphins in the Swan Canning Riverpark. New volunteers are provided with: training that informs how an estuary system functions and how human activities can impact the system; updates on the ecology of the resident subpopulation of dolphins; examples of the impacts of discarded fishing line and vessel noise and collisions on dolphins; and guidance on safe interactions with dolphins.  Engage citizens in the conservation of the resident subpopulation of dolphins in the Swan Canning Riverpark by collecting data on dolphin sightings (i.e., citizen

187 Chapter 6 – General Conclusions

science); recognising individuals observed through ‘Finbook’ (i.e., a catalogue of dorsal fins updated every year: Department of Parks and Wildlife 2016); and contributing sightings/observations and reporting any incidents or abnormal behaviour.

In addition to the recommendations made above, in the course of the chapters I have indicated how coastal and estuarine developments could impact the subpopulations within the metropolitan waters of Perth and the different risk level expected for each subpopulation. Figure 6.1 can assist managers and stakeholders (e.g., Department of Parks and Wildlife, Council of Cockburn Sound, and Fremantle Harbour Ports) in visualizing the status of bottlenose dolphins in the different geographic regions of Perth metropolitan waters in assessing the potential impacts of proposed developments and where management strategies are most needed.

6.4. Concluding remarks

This research improves the scientific basis for environmental decision-making relating to dolphins in the Perth metropolitan area. Critically, the research will inform management decisions for the different subpopulations present, all of which have different ecological and genetic characteristics.

It is therefore essential that an effective management strategy is developed to mitigate the effects of human activities on the bottlenose dolphin subpopulation associated with each geographic region of Perth metropolitan waters. In particular, it is critical to maintain a stable and healthy OA subpopulation as it acts as the source subpopulation to all the other subpopulations in Perth metropolitan waters. Additionally, the resident subpopulation of bottlenose dolphins within the Swan Canning Riverpark presents all the characteristics of being ‘at risk’: small population size, connection with the adjacent subpopulations upon which their persistence depends (i.e., genetic flow), and habitat that is subject to heavy industrial and recreational use for human.

In this thesis, I used diverse methodological approaches, including field data collection and analytical techniques, to study dolphin ecology, behaviour and genetic

188 Chapter 6 – General Conclusions characteristics, which led to the identification of multiple subpopulations in a heterogeneous study area. I recommend that more studies consider such an approach to better inform management and mitigate anthropogenic activities at appropriate spatial scales. Indeed, the multi-strategy approach used in this thesis is broadly applicable to any individually identifiable marine and terrestrial wildlife species.

189

190

References

Allen, S.J., Bryant, K.A., Kraus, R.H., Loneragan, N.R., Kopps, A.M., Brown, A.M., Gerber, L. and Krützen, M. (2016) Genetic isolation between coastal and fishery-impacted, offshore bottlenose dolphin (Tursiops spp.) populations. Molecular Ecology 25: 2735-2753. doi: 10.1111/mec.13622

Amos, W. and Hoelzel, A.R. (1991) Long term preservation of whale skin for DNA analysis. In: A.R. Hoelzel (Ed.) Genetic ecology of whales and dolphins. Report of the International Whaling Commission Special Issue 13. International Whaling Commission, Cambridge, U.K., pp. 99-104.

Andrews, K.R., Karczmarski, L., Au, W.W.L., Rickards, S.H., Vanderlip, C.A., Bowen, B.W., Gordon Grau, E. and Toonen, R.J. (2010) Rolling stones and stable homes: social structure, habitat diversity and population genetics of the Hawaiian spinner dolphin (Stenella longirostris). Molecular Ecology 19: 732- 748. doi: 10.1111/j.1365-294X.2010.04521.x

Ansmann, I.C., Lanyon, J.M., Seddon, J.M. and Parra, G.J. (2013) Monitoring dolphins in an urban marine system: Total and effective population size estimates of Indo-Pacific bottlenose dolphins in Moreton Bay, Australia. PLoS ONE 8: e65239. doi: 10.1371/journal.pone.0065239

Ansmann, I.C., Parra, G.J., Lanyon, J.M. and Seddon, J.M. (2012) Fine-scale genetic population structure in a mobile marine mammal: inshore bottlenose dolphins in Moreton Bay, Australia. Molecular Ecology 21: 4472-4485. doi: 10.1111/j.1365-294X.2012.05722.x

Arnason, A.N. (1972) Parameter estimates from mark-recapture experiments on two populations subject to migration and death. Researches on Population Ecology 13: 33-48. doi: 10.1007/BF02521971

Arnason, A.N. (1973) The estimation of population size, migration rates and survival in a stratified population. Researches on Population Ecology 15: 1-8. doi: 10.1007/BF02510705

Australian Bureau of Statistics (2015) Regional population growth, Australia, 2014- 15. http://www.abs.gov.au/ausstats/[email protected]/mf/3218.0. Accessed 11 April 2016.

Avise, J.C., Arnold, J., Ball, R.M., Bermingham, E., Lamb, T., Niegel, J.E., reeb, C.A. and Saunders, N.C. (1987) Intraspecific phylogeography: the mitochondrial DNA bridge between population genetics and systematics. Annual Review of Ecology and Systematics 18: 489-522. doi: 10.1146/annurev.es.18.110187.002421

Bacher, K., Allen, S., Lindholm, A.K., Bejder, L. and Krützen, M. (2010) Genes or culture: are mitochondrial genes associated with tool use in bottlenose dolphins (Tursiops sp.)? Behaviour Genetics 40: 706-714. doi: 10.1007/s10519-010- 9375-8

191 References

Baker, C.S., Hamner, R.M., Cooke, J., Heimeier, D., Vant, M., Steel, D. and Constantine, R. (2013) Low abundance and probable decline of the critically endangered Maui's dolphin estimated by genotype capture–recapture. Animal Conservation 16: 224-233. doi: 10.1111/j.1469-1795.2012.00590.x

Baker, C.S., Medrano-Gonzalez, L., Calambokidis, J., Perry, A., Pichler, F.B., Rosenbaum, H., Straley, J.M., Urban-Ramirez, J., Yamaguchi, M. and Von Ziegesar, O. (1998) Population structure of nuclear and mitochondrial DNA variation among humpback whales in the North Pacific. Molecular Ecology 7: 695-707. doi: 10.1046/j.1365-294x.1998.00384.x

Baker, C.S., Perry, A., Bannister, J.L., Weinrich, M.T., Abernethy, R.B., Calambokidis, J., Lien, J., Lambertsen, R.H., Urbán Ramírez, J., Vasquez, O., Clapham, P.J., Alling, A., O'Brien, S.J. and Palumbi, S.R. (1993) Abundant mitochondrial DNA variation and world-wide population structure in humpback whales. Proceedings of the National Academy of Sciences 90: 8239-8243.

Bandelt, H.-J., Forster, P. and Rohl, A. (1999) Median-joining networks for inferring intraspecific phylogenies. Molecular Biology and Ecology 16: 37-48. doi: 10.1093/oxfordjournals.molbev.a026036

Barrat, A., Barthelemy, M., Pastor-Satorras, R. and Vespignani, A. (2004) The architecture of complex weighted networks. Proceedings of the National Academy of Sciences 101: 3747-3752. doi: 10.1073/pnas.0400087101

Bastos, R., Pinhanços, A., Santos, M., Fernandes, R.F., Vicente, J.R., Morinha, F., Honrado, J.P., Travassos, P., Barros, P., Cabral, J.A. and Cadotte, M. (2016) Evaluating the regional cumulative impact of wind farms on birds: how can spatially explicit dynamic modelling improve impact assessments and monitoring? Journal of Applied Ecology 53: 1330-1340. doi: 10.1111/1365- 2664.12451

Bejder, L. (2011) Appendix: Cape Riche desalination plant - Marine mammal assessment. GHD, Perth, WA. http://www.grangeresources.com.au/clients/grange/downloads/item147/append ix_e1_marine_mammal_assessment.pdf 13 p.

Bejder, L., Fletcher, D. and Brager, S. (1998) A method for testing association patterns of social animals. Animal Behaviour 56: 719-725. doi: 10.1006/anbe.1998.0802

Bejder, L., Hodgson, A.J., Loneragan, N.R., Allen, S.J. and Cagnazzi, D.D. (2012) Coastal dolphins in north-western Australia: the need for re-evaluation of species listings and short-comings in the Environmental Impact Assessment process. Pacific Conservation Biology 18: 56-63. doi: 10.1071/PC120022

Bejder, L., Samuels, A., Whitehead, H., Finn, H. and Allen, S. (2009) Impact assessment research: use and misuse of habituation, sensitisation and tolerance in describing wildlife responses to anthropogenic stimuli. Marine Ecology Progress Series 395: 177-185. doi: 10.3354/meps07979

192 References

Bejder, L., Samuels, A., Whitehead, H., Gales, N., Mann, J., Connor, R., Heithaus, M., Watson-Capps, J., Flaherty, C. and Krützen, M. (2006) Decline in relative abundance of bottlenose dolphins exposed to long-term disturbance. Conservation Biology 20: 1791-1798. doi: 10.1111/j.1523-1739.2006.00540.x

Berger-Tal, O., Polak, T., Oron, A., Lubin, Y., Kotler, B.P. and Saltz, D. (2011) Integrating animal behavior and conservation biology: a conceptual framework. Behavioral Ecology 22: 236-239. doi: doi:10.1093/beheco/arq224

Bilgmann, K., Moller, L.M., Harcourt, R.G., Gibbs, S.E. and Beheregaray, L.B. (2007) Genetic differentiation in bottlenose dolphins from South Australia: association with local oceanography and coastal geography. Marine Ecology Progress Series 341: 265-276. doi: 10.3354/meps341265

Bilgmann, K., Parra, G.J., Zanardo, N., Beheregaray, L.B. and Möller, L.M. (2014) Multiple management units of short-beaked common dolphins subject to fisheries bycatch off southern and southeastern Australia. Marine Ecology Progress Series 500: 265-279. doi: 10.3354/meps10649

BMT Oceania (2014) Long-term shellsand dredging Owen Anchorage. Dredging and environmental management plan - stage 2 west Success Bank (Report No. 334_575/1_Rev18). Perth, WA 126 p.

Borgatti, S.P. (2002) NetDraw: graph visualization software. Lexington, KY: Harvard Analytic Technologies

Bossley, M.I., Steiner, A., Rankin, R.W. and Bejder, L. (2017) A long-term study of bottlenose dolphins (Tursiops aduncus) in an Australian industrial estuary: increased sightings associated with environmental improvements. Marine Mammal Science 33: 277-290. doi: 10.1111/mms.12368

Brandt, M.J., Diederichs, A., Betke, K. and Nehls, G. (2011) Responses of harbour porpoises to pile driving at the Horns Rev II offshore wind farm in the Danish North Sea. Marine Ecology Progress Series 421: 205-216. doi: 10.3354/meps08888

Brearley, A. (2005) Ernest Hodgkin’s Swanland estuaries and coastal lagoons of south-western Australia. University of Western Australia Press, Crawley, Australia.

Bridge, P.D. (1993) Classification. In: J.C. Fry (Ed.)Biological data analysis. Oxford University Press, Oxford, UK, pp. 219-242.

Brittingham, M.C., Maloney, K.O., Farag, A.M., Harper, D.D. and Bowen, Z.H. (2014) Ecological risks of shale oil and gas development to wildlife, aquatic resources and their habitats. Environmental Science & Technology 48: 11034- 11047. doi: 10.1021/es5020482

Brooks, L. and Pollock, K.H. (2014) Abundance, movements and habitat use of coastal dolphins in the Darwin region: analysis of the first five primary samples (October 2011 to October 2013). http://www.inpex.com.au/media/1859/dolphin-monitoring-report-3.pdf 49 p.

193 References

Brown, A.M., Bejder, L., Pollock, K.H. and Allen, S.J. (2016) Site-specific assessments of the abundance of three inshore dolphin species to inform conservation and management. Frontiers in Marine Science 3: 4. doi: 10.3389/fmars.2016.00004

Brown, A.M., Kopps, A.M., Allen, S.J., Bejder, L., Littleford-Colquhoun, B., Parra, G.J., Cagnazzi, D., Thiele, D., Palmer, C. and Frère, C.H. (2014) Population differentiation and hybridisation of Australian snubfin (Orcaella heinsohni) and Indo-Pacific humpback (Sousa chinensis) dolphins in north-western Australia. PLoS ONE 9: e101427. doi: 10.1371/journal.pone.0101427

Brown, D.M., Brenneman, R.A., Koepfli, K.P., Pollinger, J.P., Mila, B., Georgiadis, N.J., Louis, E.E., Jr., Grether, G.F., Jacobs, D.K. and Wayne, R.K. (2007) Extensive population genetic structure in the giraffe. BMC Biology 5: 57. doi: 10.1186/1741-7007-5-57

Brusa, J.L., Young, R.F. and Swanson, T. (2016) Abundance, ranging patterns, and social behavior of bottlenose dolphins (Tursiops truncatus) in an estuarine terminus. Aquatic Mammals 42: 109-121. doi: 10.1578/AM.42.1.2016.109

Buchanan, F.C., Friesen, M.K., Littlejohn, R.P. and Clayton, J.W. (1996) Microsatellites from the beluga whale Delphinapterus leucas. Molecular Ecology 5: 571-575. doi: 10.1046/j.1365-294X.1996.00109.x

Burgess, G.H., Bruce, B.D., Cailliet, G.M., Goldman, K.J., Grubbs, R.D., Lowe, C.G., MacNeil, M.A., Mollet, H.F., Weng, K.C. and O'Sullivan, J.B. (2014) A re-evaluation of the size of the white shark (Carcharodon carcharias) population off California, USA. PLoS ONE 9: e98078. doi: 10.1371/journal.pone.0098078

Burnham, K.P. and Anderson, D.R. (2002) Model selection and multimodel inference: a practical information-theoretic approach. Springer-Verlag, New York, USA.

Burnham, K.P., Anderson, D.R., White, G.C., Brownie, C. and Pollock, K.H. (1987) Design and analysis methods for fish survival experiments based on release- capture. American Fisheries Society Monograph 5, Bethesda, MD.

Cagnazzi, D., Parra, G.J., Westley, S. and Harrison, P.L. (2013a) At the heart of the industrial boom: Australian snubfin dolphins in the Capricorn Coast, Queensland, need urgent conservation action. PLoS ONE 8: e56729. doi:56710.51371/journal.pone.0056729. doi: 10.1371/journal.pone.0056729

Cagnazzi, D., Parra, G.J., Westley, S. and Harrison, P.L. (2013b) At the heart of the industrial boom: Australian snubfin dolphins in the Capricorn Coast, Queensland, need urgent conservation action. PLoS ONE 8: e56729. doi: 10.1371/journal.pone.0056729

Cairns, S.J. and Schwager, S.J. (1987) A comparison of assocation indices. Animal Behaviour 35: 1454-1469. doi: 10.1016/S0003-3472(87)80018-0

194 References

Cam, E., Oro, D., Pradel, R. and Jimenez, J. (2004) Assessment of hypotheses about dispersal in a long-lived seabird using multistate capture-recapture models. Journal of Animal Ecology 73: 723-736. doi: 10.1111/j.0021- 8790.2004.00848.x

Cantor, M. and Whitehead, H. (2013) The interplay between social networks and culture: theoretically and among whales and dolphins. Philosophical Transactions of the Royal Society B 368: 20120340. doi: 10.1098/rstb.2012.0340

Cassens, I., Van Waerebeek, K., Best, P.B., Tzika, A., Van Helden, A.L., Crespo, E.A. and Milinkovitch, M.C. (2005) Evidence for male dispersal along the coasts but no migration in pelagic waters in dusky dolphins (Lagenorhynchus obscurus). Molecular Ecology 14: 107-121. doi: 10.1111/j.1365- 294X.2004.02407.x

Caughley, G. (1977) Analysis of vertebrate populations. Wiley, Chichester, UK.

Chabanne, D., Finn, H., Salgado-Kent, C. and Bejder, L. (2012) Identification of a resident community of bottlenose dolphins (Tursiops aduncus) in the Swan Canning Riverpark, Western Australia, using behavioural information. Pacific Conservation Biology 18: 247-262. doi: 10.1071/PC120247

Chabanne, D.B.H., Finn, H. and Bejder, L. (2017a) Identifying the relevant local population for environmental impact assessments of mobile marine fauna. Frontiers in Marine Science 4: 148. doi: 10.3389/fmars.2017.00148

Chabanne, D.B.H., Pollock, K.H., Finn, H. and Bejder, L. (2017b) Applying the multistate capture-recapture robust design to assess metapopulation structure of a marine mammal. Methods in Ecology and Evolutiondoi: 10.1111/2041- 210X.12792

Chapuis, M.P. and Estoup, A. (2007) Microsatellite null alleles and estimation of population differentiation. Molecular Biology and Evolution 24: 621-631. doi: 10.1093/molbev/msl191

Charlton-Robb, K., Taylor, A.C. and McKechnie, S.W. (2014) Population genetic structure of the Burrunan dolphin (Tursiops australis) in coastal waters of south-eastern Australia: conservation implications. Conservation Genetics 16: 195-207. doi: 10.1007/s10592-014-0652-6

Chen, L. and Yang, G. (2008) A set of polymorphic dinucleotide and tetranucleotide microsatellite markers for the Indo-Pacific humpback dolphin (Sousa chinensis) and cross-amplification in other cetacean species. Conservation Genetics 10: 697-700. doi: 10.1007/s10592-008-9618-x

Chilvers, B.L. and Corkeron, P.J. (2001) Trawling and bottlenose dolphins' social structure. Proceedings of the Royal Society London B 268: 1901-1905. doi: 10.1098/rspb.2001.1732

195 References

Chilvers, B.L. and Corkeron, P.J. (2003) Abundance of Indo-Pacific bottlenose dolphins, Tursiops aduncus, off Point Lookout, Queensland, Australia. Marine Mammal Science 19: 85-95. doi: 10.1111/j.1748-7692.2003.tb01094.x

Chilvers, B.L., Corkeron, P.J. and Puotinen, M.L. (2003) Influence of trawling on the behaviour and spatial distribution of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in Moreton bay, Australia. Canadian Journal of Zoology 81: 1947-1954. doi: 10.1139/z03-195

Choquet, R., Lebreton, J.-D., Gimenez, O., Reboulet, A.-M. and Pradel, R. (2009) U- CARE: Utilities for performing goodness of fit tests and manipulating CApture-REcapture data. Ecography 32: 1071-1074. doi: 10.1111/j.1600- 0587.2009.05968.x

Christiansen, F. and Lusseau, D. (2015) Linking behavior to vital rates to measure the effects of non-lethal disturbance on wildlife. Conservation Letters 8: 424- 431. doi: 10.1111/conl.12166

Christiansen, F., McHugh, K.A., Bejder, L., Siegal, E.M., Lusseau, D., McCabe, E.B., Lovewell, G. and Wells, R.S. (2016) Food provisioning increases the risk of injury in a long-lived marine top predator. Royal Society Open Science 3: 160560. doi: 10.1098/rsos.160560

Chybicki, I.J. and Burczyk, J. (2009) Simultaneous estimation of null alleles and inbreeding coefficients. Journal of Heredity 100: 106-113. doi: 10.1093/jhered/esn088

Cockburn Sound Management Council (2012) State of Cockburn Sound 2012 Report. Rockingham. https://www.der.wa.gov.au/images/documents/about/committees/CSMC/2012_ State_of_Cockburn_Sound.PDF 40 p.

Connor, R.C., Read, A.J. and Wrangham, R. (2000a) Male reproductive strategies and social bonds. In: J. Mann, R.C. Connor, P.L. Tyack and H. Whitehead (Eds.) The University of Chicago Press, Chicago, IL, pp. 247-269.

Connor, R.C., Richards, A.F., Smolker, R.A. and Mann, J. (1996) Patterns of female attractiveness in Indian ocean bottlenose dolphins. Behaviour 133: 37-69. doi: 10.1163/156853996X00026

Connor, R.C., Wells, R.S., Mann, J. and Read, A.J. (2000b) The bottlenose dolphin, social relationship in a fission-fusion society. In: J. Mann, R.C. Connor, P.L. Tyack and H. Whitehead (Eds.) Cetacean Societies, Field Studies of Dolphins and Whales. The University of Chicago Press, Chicago, IL, pp. 91-126.

Cooch, E.G. and White, G.C. (2005) Goodness of fit testing. In: E.G. Cooch and G.C. White (Eds.) Program MARK - A gentle introduction. pp. 148-187.

Cornuet, J.-M. and Luikart, G. (1996) Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics 144: 2001-2014.

196 References

Crespin, L., Choquet, R., Lima, M., Merritt, J. and Pradel, R. (2008) Is heterogeneity of catchability in capture–recapture studies a mere sampling artifact or a biologically relevant feature of the population? Population Ecology 50: 247- 256. doi: 10.1007/s10144-008-0090-8

Crnokrak, P. and Roff, D.A. (1999) Inbreeding depression in the wild. Heredity 83: 260-270. doi: 10.1038/sj.hdy.6885530

Cronin, M.A., Amstrup, S.C., Talbot, S.L., Sage, G.K. and Amstrup, K.S. (2009) Genetic variation, relatedness, and effective population size of polar bears (Ursus maritimus) in the southern Beaufort Sea, Alaska. Journal of Heredity 100: 681-690. doi: 10.1093/jhered/esp061

Culloch, R.M., Anderwald, P., Brandecker, A., Haberlin, D., McGovern, B., Pinfield, R., Visser, F., Jessopp, M. and Cronin, M. (2016) Effect of construction-related activities and vessel traffic on marine mammals. Marine Ecology Progress Series 549: 231-242. doi: 10.3354/meps11686

Curry, E.B. and Smith, J. (1997) Phylogeographic structure of the bottlenose dolphin (Tursiops truncatus): stock identification and implications for management. In: A.E. Dizon, S.J. Chilvers and W.F. Perrin (Eds.) Molecular genetics of marine mammals. The Society for Marine Mammalogy, Lawrence, KS, pp. 227-247.

Darroch, J.N. (1961) The two-sample capture-recapture census when tagging and sampling are stratified. Biometrika 48: 241-260. doi: 10.2307/2332748

David, J.A. (2006) Likely sensitivity of bottlenose dolphins to pile-driving noise. Water and Environment Journal 20: 48-54. doi: 10.1111/j.1747- 6593.2005.00023.x

Davis, L.G., Dibner, M.D. and Battey, J.F. (1986) Basic methods in molecular biology. Elsevier Science Publishing, New York, NY.

Dawson, S.M. and Slooten, E. (1993) Conservation of Hector’s dolphins: the case and process which led to establishment of the Banks Peninsula Marine Mammal Sanctuary. Aquatic Conservation: Marine and Freshwater Ecosystems 3: 207-221. doi: 10.1002/aqc.3270030305

Defran, R.H. and Weller, D.W. (1999) Occurence, distribution, site fidelity, and school size of bottlenose dolphins (Tursiops truncatus) off San Diego, California. Marine Mammal Science 15: 366-380. doi: 10.1111/j.1748- 7692.1999.tb00807.x

Department of Conservation and Land Management (1992) Marmion Marine Park Management Plan 1992-2002 (Management Plan No 23). Perth, WA 80 p.

Department of Environment (2003) Riverplan: an environmental management framework for the Swan and Canning rivers : comprehensive management plan and implementation strategy for the Environmental Protection (Swan and Canning Rivers) Policy 1998, for public consultation / Government of Western Australia. Western Australia Government, Perth, WA.

197 References

Department of Environment and Conservation (2007) Shoalwater Islands Marine Park Management Plan 2007-2017 (Management Plan Number 58). Perth, WA 104 p.

Department of Environment and Heritage (2008) Adelaide Dolphin Sanctuary Management Plan June 2008. Government of South Australia, Adelaide, SA 15 p.

Department of Parks and Wildlife (2015) Hydrodynamics of the Swan and Canning rivers. https://www.dpaw.wa.gov.au/management/swan-canning-riverpark/171- about-the-river-system/368-hydrodynamics-of-the-swan-and-canning-rivers. Accessed 09 January 2017.

Department of Parks and Wildlife (2016) FinBook. An identification catalogue for dolphins observed in the Swan Canning Riverpark. http://www.riverguardians.com/projects/dolphin-watch/identifying-dolphins. Accessed 15 January 2017.

Derville, S., Constantine, R., Baker, C.S., Oremus, M. and Torres, L.G. (2016) Environmental correlates of nearshore habitat distribution by the critically endangered Māui dolphin. Marine Ecology Progress Series 551: 261-275. doi: 10.3354/meps11736

DeYoung, R.W. (2007) Genetics and applied management: using genetic methods to solve emerging wildlife management problems. In: T.E. Fulbright and D.G. Hewitt (Eds.) Wildlife science - Linking ecological theory and management applications. CRC Press, Broken Sound Parkway, NW, pp. 317-336.

Donaldson, R., Finn, H., Bejder, L., Lusseau, D. and Calver, M. (2012a) Response: social learning of risky behaviour: importance for impact assessments, conservation and management of human–wildlife interactions. Animal Conservation 15: 442-444. doi: 10.1111/j.1469-1795.2012.00601.x

Donaldson, R., Finn, H., Bejder, L., Lusseau, D. and Calver, M. (2012b) The social side of human-wildlife interaction: wildlife can learn harmful behaviours from each other. Animal Conservation 15: 427-435. doi: 10.1111/j.1469- 1795.2012.00548.x

Donaldson, R., Finn, H. and Calver, M. (2010) Illegal feeding increases risk of boat- strike and entanglement in bottlenose dolphins in Perth, Western Australia. Pacific Conservation Biology 16: 157-161. doi: 10.1071/PC100157

Dudgeon, C.L. and Ovenden, J.R. (2015) The relationship between abundance and genetic effective population size in elasmobranchs: an example from the globally threatened zebra shark Stegostoma fasciatum within its protected range. Conservation Genetics 16: 1443. doi:1410.1007/s10592-10015-10752-y

Dungan, S.Z., Hung, S.K., Wang, J.Y. and White, B.N. (2012) Two social communities in the Pearl River Estuary population of Indo-Pacific humpback dolphins (Sousa chinensis). Canadian Journal of Zoology 90: 1031-1043. doi: 10.1139/Z2012-071

198 References

Dupavillon, J.L. and Gillanders, B.M. (2009) Impacts of seawater desalination on the giant Australian cuttlefish Sepia apama in the upper Spencer Gulf, South Australia. Marine Environmental Research 67: 207-218. doi: 10.1016/j.marenvres.2009.02.002

Efron, B. and Stein, C. (1981) The jackknife estimate of variance. The Annals of Statistics 9: 586-596. doi: 10.1214/aos/1176345462

Elliott, M. and Whitfield, A.K. (2011) Challenging paradigms in estuarine ecology and management estuarine. Estuarine Coastal and Shelf Science 94: 306-314. doi: 10.1016/j.ecss.2011.06.016

Environmental Protection Authority (1998) The marine environment of Cockburn Sound - Strategic environmental advice (Bulletin 907). Perth, WA 51 p.

Environmental Protection Authority (2001) Long-term shellsand dredging, Owen Anchorage, Cockburn Cement Ltd (Bulletin 1033). Perth, WA 80 p.

Erbe, C. (2013) International regulation of underwater noise. Acoustics Australia 41: 12-19.

Evanno, G., Regnaut, S. and Goudet, J. (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14: 2611-2620. doi: 10.1111/j.1365-294X.2005.02553.x

Excoffier, L. and Lischer, H.E. (2010) Arlequin suite ver 3.5: a new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10: 564-567. doi: 10.1111/j.1755- 0998.2010.02847.x

Falush, D., Stephens, M. and Pritchard, J.K. (2003) Inference of popualtion structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics 164: 1567-1587. doi: 10.1111/j.1471-8286.2007.01758.x

Fernández, R., Santos, M.B., Pierce, G.J., Llavona, Á., López, A., Silva, M.a., Ferreira, M., Carrillo, M., Cermeño, P., Lens, S. and Piertney, S.B. (2011) Fine-scale genetic structure of bottlenose dolphins, Tursiops truncatus, in Atlantic coastal waters of the Iberian Peninsula. Hydrobiologia 670: 111-125. doi: 10.1007/s10750-011-0669-5

Finn, H. (2005) Conservation biology of bottlenose dolphins (Tursiops sp.) in Perth metropolitan waters. PhD thesis, Murdoch University. 206 p.

Finn, H. and Calver, M.C. (2008) Feeding aggregations of bottlenose dolphins and seabirds in Cockburn Sound, Western Australia. The Western Australian Naturalist 26: 157-172.

Finn, H., Donalson, R. and Calver, M. (2008) Feeding Flipper: a case study of a human-dolphin interaction. Pacific Conservation Biology 14: 215-225. doi: 10.1071/PC080215

199 References

Fontaine, M.C., Baird, S.J., Piry, S., Ray, N., Tolley, K.A., Duke, S., Birkun, A., Jr., Ferreira, M., Jauniaux, T., Llavona, A., Ozturk, B., A, A.O., Ridoux, V., Rogan, E., Sequeira, M., Siebert, U., Vikingsson, G.A., Bouquegneau, J.M. and Michaux, J.R. (2007) Rise of oceanographic barriers in continuous populations of a cetacean: the genetic structure of harbour porpoises in Old World waters. BMC Biology 5: 30. doi: 10.1186/1741-7007-5-30

Fox, A.D., Desholm, M., Kahlert, J., Christensen, T.K. and Krag Petersen, I.B. (2006) Information needs to support environmental impact assessment of the effects of European marine offshore wind farms on birds. Ibis 148: 129-144. doi: 10.1111/j.1474-919X.2006.00510.x

Frankham, R. (1995a) Conservation genetics. Annual Review of Genetics 29: 305- 327. doi: 10.1146/annurev.ge.29.120195.001513

Frankham, R. (1995b) Effective population size/adult population size ratios in wildlife: a review. Genetics Research 66: 95-107. doi: 10.1017/S0016672308009695

Frankham, R., Ballou, J.D. and Briscoe, D.A. (2010) Introduction to Conservation Genetics. Cambridge University Press, Cambridge, UK.

Frère, C.H., Krzyszczyk, E., Patterson, E.M., Hunter, S., Ginsburg, A. and Mann, J. (2010a) Thar she blows! A novel method for DNA collection from cetacean blow. PLoS ONE 5: e12299. doi: 10.1371/journal.pone.0012299

Frère, C.H.H., Krützen, M., Mann, J., Watson-Capps, J.J.J., Tsai, Y.J.J., Patterson, E.M., Connor, R., Bejder, L. and Sherwin, W.B. (2010b) Home range overlap, matrilineal and biparental kinship drive female associations in bottlenose dolphins. Animal Behaviour 80: 481-486. doi: 10.1016/j.anbehav.2010.06.007

Fruet, P.F., Secchi, E.R., Daura-Jorge, F., Vermeulen, E., Flores, P.a.C., Simões- Lopes, P.C., Genoves, R.C., Laporta, P., Tullio, J.C., Freitas, T.R.O., Rosa, L.D., Valiati, V.H., Beheregaray, L.B. and Möller, L.M. (2014) Remarkably low genetic diversity and strong population structure in common bottlenose dolphins (Tursiops truncatus) from coastal waters of the southwestern Atlantic Ocean. Conservation Genetics 15: 879-895. doi: 10.1007/s10592-014-0586-z

Fu, Y.-X. (1997) Statistical tests of neutrality of mutations against population growth, hitchhiking and background selection. Genetics 147: 915-925.

Fury, C.A. and Harrison, P.L. (2008) Abundance, site fidelity and range patterns of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in two Australian subtropical estuaries. Marine and Freshwater Research 59: 1015-2027. doi: 10.1071/MF08109

Gaggiotti, O.E., Brooks, S.P., Amos, W. and Harwood, J. (2004) Combining demographic, environmental and genetic data to test hypotheses about colonization events in metapopulations. Molecular Ecology 13: 811-825. doi: 10.1046/j.1365-294X.2003.02028.x

200 References

Gagnaire, P.A., Broquet, T., Aurelle, D., Viard, F., Souissi, A., Bonhomme, F., Arnaud-Haond, S. and Bierne, N. (2015) Using neutral, selected, and hitchhiker loci to assess connectivity of marine populations in the genomic era. Evolutionary Applications 8: 769-786. doi: 10.1111/eva.12288

Garrote-Moreno, A., Fernandez-Torquemada, Y. and Sanchez-Lizaso, J.L. (2014) Salinity fluctuation of the brine discharge affects growth and survival of the seagrass Cymodocea nodosa. Marine Pollution Bulletin 81: 61-68. doi: 10.1016/j.marpolbul.2014.02.019

Gaspari, S., Scheinin, A., Holcer, D., Fortuna, C., Natali, C., Genov, T., Frantzis, A., Chelazzi, G. and Moura, A.E. (2015) Drivers of population structure of the bottlenose dolphin (Tursiops truncatus) in the eastern Mediterranean Sea. Evolutionary Biology 42: 177-190. doi: 10.1007/s11692-015-9309-8

Gibbs, S.E., Harcourt, R.G. and Kemper, C.M. (2011) Niche differenriarion of bottlenose dolphin species in South Australia revealed by stable isotopes and stomach contents. Wildlife Research 38: 261-270. doi: 10.1071/WR10108

Gibson, D., Blomberg, E.J., Atamian, M. and Sedinger, J.S. (2014) Lek fidelity and movement among leks by male greater sage-grouse Centrocercus urophasianus: a capture-mark-recapture approach. International journal of avian science 156: 729-740. doi: 10.1111/ibi.12192

Gilson, A., Syvanen, M., Levine, K. and Banks, J. (1998) Deer gender determination by polymerase chain reaction: Validation study and application to tissues, bloodstains, and hair forensic samples from California. California Fish and Game 84: 159-169.

Girard, I., Ouellet, J.-P., Courtois, R., Dussault, C. and Breton, L. (2002) Effects of sampling effort based on GPS telemetry on home-range size estimations. The Journal of Wildlife Management 66: 1290-1300. doi: 10.2307/3802962

Glasson, J., Thrivel, R. and Chadwick, A. (2012) Introduction to environmental impact assessement. Routledge, New York, NY.

Goudet, J. (2001) FSTAT, a program to estimate and test gene diversities and fixation indices (version 2.9.3). Available from http://www.unil.ch/izea/softwares/fstat.html.

Government of Western Australia (2017) Dolphin Watch Project. http://www.riverguardians.com/projects/dolphin-watch/. Accessed 15 January 2017.

Green, B.J. (2011) Photo-identification and population characteristics of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in Warnbro Sound and Shoalwater Bay, Western Australia. Honours thesis, Murdoch University. 107 p.

Greenbaum, G., Templeton, A.R., Zarmi, Y. and Bar-David, S. (2014) Allelic richness following population founding events - a stockastic modeling framework incorporating gene flow and genetic drift. PLoS ONE 9: e1115203. doi: 10.1371/journal.pone.0115203

201 References

Hale, P. (1997) Conservation of inshore dolphins in Australia. Asian Marine Biology 14: 83-91.

Hale, P.T., Barreto, A.S. and Ross, G.J.B. (2000) Comparative morphology and distribution of the aduncus and truncatus forms of bottlenose dolphin Tursiops in the Indian and Western Paciﬁc Oceans. Aquatic Mammals 26: 101-110.

Ham, G.S. (2009) Population biology of bottlenose dolphins (Tursiops sp.) in Cockburn Sound, Western Australia. Honours thesis, Murdoch University.

Hammond, P.S. (2010) Estimating the abundance of marine mammals. In: I.L. Boyd, W. Don Bowen and S.J. Iverson (Eds.) Marine Mammal Ecology and Conservation: A Handbook of Techniques. Oxford University Press, New York, pp. 42-67.

Hammond, P.S., Bearzi, G., Bjørge, A., Forney, K.A., Karkzmarski, L., Kasuya, T., Perrin, W.F., Scott, M.D., Wang, J.Y., Wells, R.S. and Wilson, B. (2012) Tursiops aduncus. The IUCN Red List of threatened species 2012: e.T41714A17600466. http://dx.doi.org/10.2305/IUCN.UK.2012.RLTS.T41714A17600466.en. Accessed 11 January 2017.

Hammond, P.S., Mizroch, S.A. and Donovan, G.P. (1990) Individual recognition of cetaceans: use of photo-identification and other techniques to estimate population parameters. International Whaling Commission, Cambridge 440 p.

Hamner, R.M., Oremus, M., Stanley, M., Brown, P., Constantine, R. and Baker, C.S. (2012) Estimating the abundance and effective population size of Maui's dolphins using microsatellite genotypes in 2010-11, with retrospective matching to 2001-07. Department of Conservation, Auckland 44 p.

Hare, M.P., Nunney, L., Schwartz, M.K., Ruzzante, D.E. and Burford, M. (2011) Understanding and estimating effective popualtion size for practical application in marine species management. University of Nebraska, Lincoln 274 p.

Harmsen, B.J., Foster, R.J. and Doncaster, C.P. (2010) Heterogeneous capture rates in low density populations and consequences for capture-recapture analysis of camera-trap data. Population Ecology 53: 253-259. doi: 10.1007/s10144-010- 0211-z

Hastings, A. (1993) Complex interactions between dispersal and dynamics: lessons from coupled logistic. Ecology 74: 1362-1372. doi: 10.2307/1940066

Hedrick, P.W. (1999) Perspective: highly variable loci and their interpretation in evolution and conservation. Evolution 53: 313-318. doi: 10.1111/j.1558- 5646.1999.tb03767.x

Hedrick, P.W. (2011) Genetics of populations. Jones and Bartlett Publishers, Sudbury, Ontario.

202 References

Heithaus, M.R. and Dill, L.M. (2002) Food availability and tiger shark predation risk influence bottlenose dolphin habitat use. Ecology 83: 480-491. doi: 10.2307/2680029

Henderson, E.E. (2004) Behaviour, association patterns and habitat use of a small community of Bottlenose dolphins in San Luis Pass, Texas. Master thesis, Texas A&M University. 99 p.

Hestbeck, J.B., Nichols, J.D. and Malecki, R.A. (1991) Estimates of movement and site fidelity using mark-resight data of wintering Canada geese. Ecology 72: 523-533. doi: 10.2307/2937193

Hill, D., Hockin, D., Price, D., Tucker, G., Morris, R. and Treweek, J. (1997) Bird disturbance: improving the quality and utility of disturbance research. Journal of Applied Ecology 34: 275-288. doi: 10.2307/2404876

Hoelzel, A.R. (1998) Genetic structure of cetacean populations in sympatry,parapatry and mixed assemblages: implications for conservation policy. Journal of Heredity 89: 451-458. doi: 10.1093/jhered/89.5.451

Hoelzel, A.R., Dahlheim, M.E. and Stern, S.J. (1998a) Low genetic variation among killer whales (Orcinus orca) in the Eastern North Pacific and genetic differentiation between foraging specialists. Journal of Heredity 89: 121-128. doi: 10.1093/jhered/89.2.121

Hoelzel, A.R., Potter, C.W. and Best, P.B. (1998b) Genetic differentiation between parapatric 'nearshore' and 'offshore' populations of the bottlenose dolphin. Proceedings of the Royal Society London B 265: 1177-1183. doi: 10.1098/rspb.1998.0416

Hoffman, J.I., Dasmahapatra, K.K., Amos, W., Phillips, C.D., Gelatt, T.S. and Bickham, J.W. (2009) Contrasting patterns of genetic diversity at three different gene markers in a marine mammal metapopulation. Molecular Ecology 18: 2961-2978. doi: 10.1111/j.1365-294X.2009.04246.x

Holyoake, C., Finn, H., Stephens, N., Duignan, P., Salgado, C., Smith, H., Bejder, L., Linke, T., Daniel, C., Lo, H.N., Ham, G.S., Moiler, K., Allen, S., Bryant, K. and McElligott, D. (2010) Technical report on the bottlenose dolphin (Tursiops aduncus) unusual mortality event within the Swan Canning Riverpark, June-October 2009. Murdoch University, Perth, WA, Perth, WA 234 p.

Hyndes, G.A., Kendrick, A.J., MacArthur, L.D. and Stewart, E. (2003) Differences in the species- and size-composition of fish assemblages in three distinct seagrass habitats with differing plant and meadow structure. Marine Biology 142: 1195-1206. doi: 10.1007/s00227-003-1010-2

Irwin, D.E. (2002) Phylogeographic breaks without geographic barriers to gene flow. Evolution 56: 2383-2394. doi: 10.1111/j.0014-3820.2002.tb00164.x

203 References

Jefferson, T.A., Hung, S.K. and Würsig, B. (2009) Protecting small cetaceans from coastal development: Impact assessment and mitigation experience in Hong Kong. Marine Policy 33: 305-311. doi: 10.1016/j.marpol.2008.07.011

Jennions, M.D. and Møller, A.P. (2003) A survey of the statistical power of research in behavioral ecology and animal behavior. Behavioral Ecology 14: 438-445. doi: 10.1093/beheco/14.3.438

Jensen, J.L., Bohonak, A.J. and Kelley, S.T. (2005) Isolation by distance, web service. http://ibdws.sdsu.edu/

Jost, L.O.U. (2008) GST and its relatives do not measure differentiation. Molecular Ecology 17: 4015-4026. doi: 10.1111/j.1365-294X.2008.03887.x

Kalinowski, S.T. (2002) Evolutionary and statistical properties of three genetic distances. Molecular Ecology 11: 1263-1123. doi: 10.1046/j.1365- 294X.2002.01520.x

Kalinowski, S.T. and Waples, R.S. (2002) Relationship of effective to census size in fluctuating population. Conservation Biology 16: 129-136. doi: 10.1046/j.1523- 1739.2002.00134.x

Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, S., Buxton, S., Cooper, A., Markowitz, S., Duran, C., Thierer, T., Ashton, B., Mentjies, P. and Drummond, A. (2012) Geneious basic: an integrated and extendable desktop software platform for the organization and analysis of sequence data. Bioinformatics 28: 1647-1649. doi: 10.1093/bioinformatics/bts199

Kendall, W.L. (1990) The robust design for capture-recapture studies: analysis using program Mark. International Wildlife Management Congress 357-360.

Kendall, W.L. (2004) Coping with unobservable and mis-classified states in capture- recapture studies. Animal Biodiversity and Conservation 27: 97-107.

Kendall, W.L. and Pollock, K.H. (1992) The Robust Design in capture-recapture studies: a review and evaluation by Monte Carlo simulation. In: D.R. McCullough and R.H. Barrett (Eds.) Wildlife 2001: Populations. Springer Netherlands, Dordrecht, pp. 31-43.

Kendrick, G.A., Hegge, B.J., Wyllie, A., Davidson, A. and Lord, D.A. (2000) Changes in seagrass cover on Success and Parmelia Banks, Western Australia between 1965 and 1995. Estuarine, Coastal and Shelf Science 50: 341-353. doi: 10.1006/ecss.1999.0569

Kiszka, J., Simon-Bouhet, B., Gastebois, C., Pusineri, C., Ridoux, V. and Simon- Bouchet, B. (2012) Habitat partitioning and fine scale population structure among insular bottlenose dolphins (Tursiops aduncus) in a tropical lagoon. Journal of Experimental Marine Biology and Ecology 416-417: 176-184. doi: 10.1016/j.jembe.2012.03.001

204 References

Kopps, A.M., Ackermann, C.Y., Sherwin, W.B., Allen, S.J., Bejder, L. and Krützen, M. (2014) Cultural transmission of tool use combined with habitat specializations leads to fine-scale genetic structure in bottlenose dolphins. Proceedings of The Royal Society B 281: 20133245. doi: 10.1098/rspb.2013.3245

Krützen, M., Barre, L.M., Moller, L.M., Heithaus, M.R., Simms, C. and Sherwin, W.B. (2002) A biopsy system for small cetaceans: darting success and wound healing in Tursiops spp. Marine Mammal Science 18: 863-878. doi: 10.1111/j.1748-7692.2002.tb01078.x

Krützen, M., Sherwin, W.B., Berggrem, P. and Gales, N. (2004) Population structure in an inshore cetacean revealed by microsatellite and mtDNA analysis: bottlenose dolphins (Tursiops sp.) in Shark Bay, Western Australia. Marine Mammal Science 20: 28-47. doi: 10.1111/j.1748-7692.2004.tb01139.x

Krützen, M., Valsecchi, E., Connor, R.C. and Sherwin, B. (2001) Characterization of microsatellite loci in Tursiops aduncus. Molecular Ecology Notes 1: 170-172. doi: 10.1046/j.1471-8278.2001.00065.x

Latch, E.K., Dharmarajan, G., Glaubitz, J.C. and Rhodes, O.E. (2006) Relative performance of Bayesian clustering software for inferring population substructure and individual assignment at low levels of population differentiation. Conservation Genetics 7: 295-302. doi: 10.1007/s10592-005- 9098-1

LeDuc, R.G. (2009) Delphinids, Overview. In: W.F. Perrin, B. Würsig and J.G.M. Thewissen (Eds.) Encyclopedia of Marine Mammals. Academic Press, San Diego, CA, pp. 298-302.

LeDuc, R.G., Perrin, W.F. and Dizon, A.E. (1999) Phylogenetic relationships among the Delphinid cetaceans based on full cytochrome B sequences. Marine Mammal Science 15: 619-648. doi: 10.1111/j.1748-7692.1999.tb00833.x

Lee, D.E. (2015) Demography of giraffe in the fragmented Tarangire ecosystem. PhD thesis, Dartmouth College. 140 p.

Legendre, P. and Fortin, M.J. (1989) Spatial pattern and ecological analysis. Plant Ecology 80: 107-138. doi: 10.1007/BF00048036

Lettink, M. and Armstrong, D.P. (2003) An introduction to using mark-recapture analysis for monitoring threatened species. Department of Conservation Technical Series 28A: 5-32.

Levins, R. (1969) Some demographic and genetic consequences of environmental heterogeneity for biological control. Bulletin of the Entomological Society of America 15: 237-240. doi: 10.1093/besa/15.3.237

Lewison, R., Crowder, L., Read, A. and Freeman, S. (2004) Understanding impacts of fisheries bycatch on marine megafauna. Trends in ecology & evolution 19: 598-604. doi: 10.1016/j.tree.2004.09.004

205 References

Librado, P. and Rozas, J. (2009) DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25: 1451-1452. doi: 10.1093/bioinformatics/btp187

Lindberg, M.S. (2012) A review of designs for capture-mark-recapture studies in discrete time. Journal of Ornithology 152: 355-370. doi: 10.1007/s10336-010- 0533-9

Litz, J.A. (2007) Social structure, genetic structure, and persistent organohalogen pollutants in bottlenose dolphins (Tursiops truncatus) in Biscayne Bay,Florida. PhD thesis, University of Miami. 156 p.

Louis, M. (2014) Social, ecological and genetic structures of bottlenose dolphins, Tursiops truncatus, in the Normano-Breton gulf and in the North-East Atlantic. PhD thesis, University de La Rochelle. 298 p.

Louis, M., Viricel, A., Lucas, T., Peltier, H., Alfonsi, E., Berrow, S., Brownlow, A., Covelo, P., Dabin, W., Deaville, R., de Stephanis, R., Gally, F., Gauffier, P., Penrose, R., Silva, M.a., Guinet, C. and Simon-Bouhet, B. (2014) Habitat- driven population structure of bottlenose dolphins, Tursiops truncatus, in the North-East Atlantic. Molecular Ecology 23: 857-874. doi: 10.1111/mec.12653

Luikart, G., Allendorf, F.W., Cornuet, J.-M. and Sherwin, W.B. (1998) Distortion of allele frequency distributions provides a test for recent population bottlenecks. The American Genetic Association 89: 238-247. doi: 10.1093/jhered/89.3.238

Luikart, G., Ryman, N., Tallmon, D.A., Schwartz, M.K. and Allendorf, F.W. (2010) Estimation of census and effective population sizes: the increasing usefulness of DNA-based approaches. Conservation Genetics 11: 355-373. doi: 10.1007/s10592-010-0050-7

Lusseau, D. (2003a) Effects of tour boats on the behaviour of bottlenose dolphins: using Markov chains to model anthropogenic impacts. Conservation Biology 17: 1785-1793. doi: 10.1111/j.1523-1739.2003.00054.x

Lusseau, D. (2003b) Male and female bottlenose dolphins Tursiops spp . have different strategies to avoid interactions with tour boats in Doubtful Sound , New Zealand. Marine Ecology Progress Series 257: 267-274. doi: 10.3354/meps257267

Lusseau, D. and Newman, M.E.J. (2004) Identifying the role that animals play in their social networks. Proceedings of the Royal Society London B 271: S477- S481. doi: 10.1098/rsbl.2004.0225

Lusseau, D., Schneider, K., Boisseau, O.J., Haase, P., Solooten, E., Dawson, S.M. and Slooten, E. (2003) The bottlenose dolphin community of Doubtful Sound features a large proportion of long-lasting associations. Behavioral Ecology and Sociobiology 54: 396-405. doi: 10.1007/s00265-003-0651-y

MacLeod, C.D. (2014) An introduction to using GIS in marine biology. Supplementary workbook four. Investigating home ranges of individual animals. Pictish Beast Publications, Glasgow, U.K.

206 References

Manlik, O., Daniel, C., Bejder, L., Allen, S. and Sherwin, W.B. (in prep-a) Demography and genetics suggest reversal of dolphin source-sink dynamics, with implications for conservation.

Manlik, O., Krützen, M., Kopps, A.M., Mann, J., Bejder, L., Allen, S.J., Frère, C., Connor, R.C. and Sherwin, W.B. (in prep-b) Population viability is better reflected by MHC diversity than by neutral genetic diversity: a case study of two bottlenose dolphin populations.

Manlik, O., McDonald, J.A., Mann, J., Raudino, H.C., Bejder, L., Krützen, M., Connor, R.C., Heithaus, M.R., Lacy, R.C. and Sherwin, W.B. (2016) The relative importance of reproduction and survival for the conservation of two dolphin populations. Ecology and Evolution 6: 3496-3512. doi: 10.1002/ece3.2130

Mann, J., Connor, R.C., Barre, L.M. and Heithaus, M.R. (2000) Female reproductive success in bottlenose dolphins (Tursiops sp.): life history, habitat, provisioning, and group-size effects. Behavioral Ecology 11: 210-219. doi: 10.1093/beheco/11.2.210

Mann, J. and Smuts, B.B. (1998) Natal attraction: allomaternal care and mother- infant separations in wild bottlenose dolphins. Animal Behaviour 55: 1097- 1113. doi: 10.1006/anbe.1997.0637

Mansur, R.M., Strindberg, S. and Smith, B.D. (2012) Mark-resight abundance and survival estimation of Indo-Pacific bottlenose dolphins, Tursiops aduncus, in the Swatch-of-No-Ground, Bangladesh. Marine Mammal Science 28: 561-578. doi: 10.1111/j.1748-7692.2011.00520.x

Marley, S.A., Salgado Kent, C.P. and Erbe, C. (2016) Occupancy of bottlenose dolphins (Tursiops aduncus) in relation to vessel traffic, dredging, and environmental variables within a highly urbanised estuary. Hydrobiologia 792: 243-263. doi: 10.1007/s10750-016-3061-7

McCluskey, S.M., Bejder, L. and Loneragan, N.R. (2016) Dolphin prey availability and calorific value in an estuarine and coastal environment. Frontiers in Marine Science 3: 30. doi: 10.3389/fmars.2016.00030

Mercer, D. (2013) Dry climate limits water options. The West Australian. The West Australian, Perth.

Minister for Planning (2009) Victoria desalination project assessment under Environment Effects Act 1978. http://www.dtpli.vic.gov.au/__data/assets/pdf_file/0010/231040/Final_Assess ment_-_Victorian_Desalination_Project.pdf 104 p.

Mirimin, L., Miller, R., Dillane, E., Berrow, S.D., Ingram, S., Cross, T.F. and Rogan, E. (2011) Fine-scale population genetic structuring of bottlenose dolphins in Irish coastal waters. Animal Conservation 14: 342-353. doi: 10.1111/j.1469- 1795.2010.00432.x

207 References

Moilanen, A. and Nieminen, M. (2002) Simple connectivity measures in spatial ecology. Ecology 84: 1131-1145. doi: 10.2307/3071919

Moiler, K. (2008) Bottlenose dolphins (Tursiops sp.) - a study of patterns in spatial and temporal use of the Swan River, Western Australia. Honours thesis, Curtin University, WA. 128 p.

Möller, L., Valdez, F.P., Allen, S., Bilgmann, K., Corrigan, S. and Beheregaray, L.B. (2011) Fine-scale genetic structure in short-beaked common dolphins (Delphinus delphis) along the East Australian Current. Marine Biology 158: 113-126. doi: 10.1007/s00227-010-1546-x

Möller, L.M. (2001) Social organisation and genetic relationships of coastal botlenose dolphins in southeastern Australia. PhD thesis, Macquarie University, Australia. 392 p.

Möller, L.M., Allen, S.J. and Harcourt, R.G. (2002) Group characteristics, site fidelity and seasonal abundance of bottlenose dolphins Tursiops aduncus in Jervis Bay and Port Stephens, south-eastern Australia. Australian Mammalogy 24: 11-21. doi: 10.1071/AM02011

Möller, L.M. and Beheregaray, L.B. (2001) Coastal bottlenose dolphins from South Eastern Australia are Tursiops aduncus according to sequences of the mitochondrial DNA control region. Marine Mammal Science 17: 249-263. doi: 10.1111/j.1748-7692.2001.tb01269.x

Möller, L.M. and Beheregaray, L.B. (2004) Genetic evidence for sex-biased dispersal in resident bottlenose dolphins (Tursiops aduncus). Molecular Ecology 13: 1607-1612. doi: 10.1111/j.1365-294X.2004.02137.x

Möller, L.M., Beheregaray, L.B., Allen, S.J. and Harcourt, R.G. (2006) Association patterns and kinship in female Indo-Pacific bottlenose dolphins (Tursiops aduncus) of southeastern Australia. Behavioral Ecology and Sociobiology 61: 109-117. doi: 10.1007/s00265-006-0241-x

Möller, L.M., Wiszniewski, J., Allen, S.J. and Beheregaray, L.B. (2007) Habitat type promotes rapid and extremely localised genetic differentiation in dolphins. Marine and Freshwater Research 58: 640-648. doi: 10.1071/MF06218

Morrison, M.L., Block, W.M., Strickland, M.D., Collier, B.A. and Peterson, M.J. (2008a) Experimental designs. In: B.N. Anderson, R.W. Howarth and L.R. Walker (Eds.) Wildlife study design. Springer, New York, USA, pp. 77-136.

Morrison, M.L., Block, W.M., Strickland, M.D., Collier, B.A. and Peterson, M.J. (2008b) Sampling survey strategies. In: B.N. Anderson, R.W. Howarth and L.R. Walker (Eds.) Wildlife study design. Springer, New York, USA, pp. 137- 197.

Morrison, M.L., Marcot, B.G. and Mannan, R.W. (2006) Wildlife-habitat relationship: concepts and applications. Isanld Press, Washington, DC.

208 References

Mossman, C.A. and Waser, P.M. (1999) Genetic detection of sex-biased dispersal. Molecular Ecology 8: 1063-1067. doi: 10.1046/j.1365-294x.1999.00652.x

Moura, A.E., Natoli, A., Rogan, E. and Hoelzel, A.R. (2013) Atypical panmixia in a European dolphin species (Delphinus delphis): implications for the evolution of diversity across oceanic boundaries. Journal of evolutionary biology 26: 63- 75. doi: 10.1111/jeb.12032

Nater, A., Kopps, A.M. and Krützen, M. (2009) New polymorphic tetranucleotide microsatellites improve scoring accuracy in the bottlenose dolphin Tursiops aduncus. Molecular Ecology Resources 9: 531-534. doi: 10.1111/j.1755- 0998.2008.02246.x

Natoli, A., Birkun, A., Aguilar, A., Lopez, A. and Hoelzel, A.R. (2005) Habitat structure and the dispersal of male and female bottlenose dolphins (Tursiops truncatus). Proceedings of The Royal Society B 272: 1217-1226. doi: 10.1098/rspb.2005.3076

Natoli, A., Peddemors, V.M. and Hoelzel, A.R. (2004) Population structure and speciation in the genus Tursiops based on microsatellite and mitochondrial DNA analyses. Journal of evolutionary biology 17: 363-375. doi: 10.1046/j.1420-9101.2003.00672.x

Natoli, A., Peddemors, V.M. and Hoelzel, A.R. (2007) Population structure of bottlenose dolphins (Tursiops aduncus) impacted by bycatch along the east coast of South Africa. Conservation Genetics 9: 627-636. doi: 10.1007/s10592- 007-9379-y

Nei, M. (1978) Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89: 583-590.

Nei, M. (1987) Molecular Evolutionary Genetics. Columbia University Press, New York, US.

Nei, M. and Li, W.-H. (1979) Mathematical model for studying genetic variation in terms of restriction endonucleases. Proceedings of the National Academy of Sciences USA 76: 5269-5273. doi: 10.1073/pnas.76.10.5269

Newman, M.E.J. (2004) Analysis of weighted networks. Physical Review 70: 056131. doi: 10.1103/PhysRevE.70.056131

Newman, M.E.J. (2006) Modularity and community structure in networks. Proceedings of the National Academy of Sciences 103: 8577-8582. doi: 10.1073/pnas.0601602103

Nichols, J.D. and Coffman, C.J. (1999) Demographic parameter estimation for experimental landscape studies of small mammal populations. In: G.W. Barrett and J.D. Peles (Eds.) Landscape Ecology of Small Mammals. Springer-Verlag, New York, USA, pp. 287-309.

209 References

Nichols, J.D. and Kendall, W.L. (1995) The use of multi-state capture-recapture models to address questions in evolutionary ecology. Journal of Applied Statistics 22: 835-846. doi: 10.1080/02664769524658

Nichols, J.D., Kendall, W.L., Hines, J.E. and Spendelow, J.A. (2004) Estimation of sex-specific survival from capture-recapture data when sex is not always known. Ecology 85: 3192-3201. doi: 10.1890/03-0578

Nicholson, K., Bejder, L., Allen, S.J., Krützen, M. and Pollock, K.H. (2012) Abundance, survival and temporary emigration of bottlenose dolphins (Tursiops sp.) off Useless Loop in the western gulf of Shark Bay, Western Australia. Marine and Freshwater Research 63: 1059-1068. doi: 10.1071/MF12210

Nielsen, O.R. and Slatkin, M. (2013) Genetic drift and mutation. In: An Introduction to Population Genetics: Theory and Applications. Sinauer Associates, Sunderland, U.S., pp. 21-34.

Nishita, M., Shirakihara, M. and Amano, M. (2015) A community split among dolphins: the effect of social relationships on the membership of new communities. Scientific Reports 5: 17266. doi: 10.1038/srep17266

O'Connell-Goode, K.C., Lowe, C.L. and Clark, J.D. (2014) Effects of a flooding event on a threatened black bear population in Louisiana. Animal Conservation 17: 476-485. doi: 10.1111/acv.12114

Olsen, M.T., Andersen, L.W., Dietz, R., Teilmann, J., Härkönen, T. and Siegismund, H.R. (2014) Integrating genetic data and population viability analyses for the identification of harbour seal (Phoca vitulina) populations and management units. Molecular Ecology 23: 815-831. doi: 10.1111/mec.12644

Ortiz, R.M. (2001) Review - Osmoregulation in marine mammals. The Journal of experimental biology 204: 1831-1844.

Otis, D.L., Burnham, K.P., White, G.C. and Anderson, D.R. (1978) Statistical inference from capture data on closed animal populations. Wildlife Monographs 62: 3-135.

Oudejans, M.G., Visser, F., Englund, A., Rogan, E. and Ingram, S.N. (2015) Evidence for distinct coastal and offshore communities of bottlenose dolphins in the North East Atlantic. PLoS ONE 10: e0122668. doi: 10.1371/journal.pone.0122668

Paiva, E.G., Salgado Kent, C.P., Gagnon, M.M., McCauley, R. and Finn, H. (2015) Reduced detection of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in an Inner Harbour channel during pile driving activities. Aquatic Mammals 41: 455-468. doi: 10.1578/AM.41.4.2015.455

Palmer, C., Brooks, L., Parra, G.J., Rogers, T., Glasgow, D. and Woinarski, J.C.Z. (2015) Estimates of abundance and apparent survival of coastal dolphins in Port Essington harbour, Northern Territory, Australia. Wildlife Research 41: 35-45. doi: 10.1071/WR14031_CO

210 References

Palstra, F.P. and Fraser, D.J. (2012) Effective/census population size ratio estimation: a compendium and appraisal. Ecology and Evolution 2: 2357-2365. doi: 10.1002/ece3.329

Palstra, F.P. and Ruzzante, D.E. (2008) Genetic estimates of contemporary effective population size: what can they tell us about the importance of genetic stochasticity for wild population persistence? Molecular Ecology 17: 3428- 3447. doi: 10.1111/j.1365-294X.2008.03842.x

Parsons, K.M., Durban, J.W., Claridge, D.E., Balcomb, K.C., Noble, L.R. and Thompson, P.M. (2003) Kinship as a basis for alliance formation between male bottlenose dolphins, Tursiops truncatus, in the Bahamas. Animal Behaviour 66: 185-194. doi: 10.1006/anbe.2003.2186

Parsons, K.M., Durban, J.W., Claridge, D.E., Herzing, D.L., Balcomb, K.C. and Noble, L.R. (2006) Population genetic structure of coastal bottlenose dolphins (Tursiops truncatus) in the northern Bahamas. Marine Mammal Science 22: 276-298. doi: 10.1111/j.1748-7692.2006.00019.x

Peakall, R. and Smouse, P.E. (2006) GenAlEx 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6: 288-295. doi: 10.1111/j.1471-8286.2005.01155.x

Peakall, R. and Smouse, P.E. (2012) GenAlEx 6.5: genetic analysis in excel. Population genetic software for teaching and research. Bioinformatics 28: 2537-2539. doi: 10.1093/bioinformatics/bts460

Peery, M.Z., Kirby, R., Reid, B.N., Stoelting, R., Doucet-Beer, E., Robinson, S., Vasquez-Carrillo, C., Pauli, J.N. and Palsboll, P.J. (2012) Reliability of genetic bottleneck tests for detecting recent population declines. Molecular Ecology 21: 3403-3418. doi: 10.1111/j.1365-294X.2012.05635.x

Petersen, C.G.J. (1895) The yearly immigration of young plaice into Limfjord form the German Sea. Report of the Danish Biological Station 6: 5-84. doi:

Pirotta, E., Laesser, B.E., Hardaker, A., Riddoch, N., Marcoux, M. and Lusseau, D. (2013) Dredging displaces bottlenose dolphins from an urbanised foraging patch. Marine Pollution Bulletin 74: 396-402. doi: 10.1016/j.marpolbul.2013.06.020

Pollock, K.H. (1982) A capture-recapture design robust to unequal probability of capture. The Journal of Wildlife Management 46: 752-757. doi: 10.2307/3808568

Pollock, K.H., Nichols, J.D., Brownie, C. and Hines, J.E. (1990) Statistical inference for capture-recapture experiments. Wildlife Monographs 107: 3-97. doi: 10.2307/3830560

Portnoy, D.S., McDowell, J.R., McCandless, C.T., Musick, J.A. and Graves, J.E. (2009) Effective size closely approximate the census size in the heavily exploited western Atlantic population of the sandbar shark, Carcharhinus

211 References

plumbeus. Conservation Genetics 10: 1697-1705. doi: 10.1007/s10592-008- 9771-2)

Pradel, R., Gimenez, O. and Lebreton, J.D. (2005) Principles and interest of GOF tests for multistate capture-recapture models. Animal Biodiversity and Conservation 28: 189-204.

Pritchard, J.K., Stephens, M. and Donnelly, P. (2000) Inference of population structure using multilocus genotype data. Genetics 155: 945-959.

Queller, D.C. and Goodnight, K.F. (1989) Estimating relatedness using genetic markers. Evolution 43: 258-275. doi: 10.2307/2409206

Quérouil, S., Freitas, L., Dinis, A., Alves, F., Cascão, I., Prieto, R., Silva, M.A., Magalhães, S., Matos, J.A. and Santos, R.S. (2010) Sex bias in biopsy samples collected from free-ranging dolphins. European Journal of Wildlife Research 56: 151-158. doi: 10.1007/s10344-009-0299-7

Quérouil, S., Silva, M.a., Freitas, L., Prieto, R., Magalhães, S., Dinis, A., Alves, F., Matos, J.a., Mendonça, D., Hammond, P.S. and Santos, R.S. (2007) High gene flow in oceanic bottlenose dolphins (Tursiops truncatus) of the North Atlantic. Conservation Genetics 8: 1405-1419. doi: 10.1007/s10592-007-9291-5

Quintana-Rizzo, E. and Wells, R.S. (2001) Resighting and association patterns of bottlenose dolphins (Tursiops truncatus) in the Cedar Keys, Florida: insights into social organization. Canadian Journal of Zoology 79: 447-456. doi: 10.1139/z00-223

Rambaut, A., Suchard, M. and Drummond, A. (2013) Tracer. MCMC trace analyis tool. http://beast.bio.ed.ac.uk/

Rand, D.M. (1996) Neutrality tests of molecular markers and the connection between DNA polymorphism, demography, and conservation biology. Conservation Biology 10: 665-671. doi: 10.1046/j.1523-1739.1996.10020665.x

Rankin, R.W., Nicholson, K.E., Allen, S.J., Krützen, M., Bejder, L. and Pollock, K.H. (2016) A Full-capture hierarchical bayesian model of Pollock's closed robust design and application to dolphins. Frontiers in Marine Science 3: 25. doi:10.3389/fmars.2016.00025.

Reeves, R.R., Smith, B.D., Crespo, E.A. and Di Sciara, N. (2003) Dolphins, Whales and Porpoises: 2002-2010 Conservation Action Plan for the World's Cetaceans. IUCN/SSC Cetacean Specialist Group, Gland, Switzerland and Cambridge, UK.

Rendell, L. and Whitehead, H. (2001) Culture in whales and dolphins. Behavioral and Brain Sciences 24: 309-324. doi: 10.1017/S0140525X0100396X

Rice, D.W. (1998) Marine mammals of the world: systematics and distribution. Allen Press, Inc.

212 References

Rice, W.R. (1989) Analyzing tables of statistical tests. Evolution 43: 223-225. doi: 10.2307/2409177

Richards, V.P., Greig, T.W., Fair, P.a., McCulloch, S.D., Politz, C., Natoli, A., Driscoll, C.a., Hoelzel, a.R., David, V., Bossart, G.D. and Lopez, J.V. (2013) Patterns of population structure for inshore bottlenose dolphins along the Eastern United States. Journal of Heredity 104: 765-778. doi: 10.1093/jhered/est070

Rooney, A.P., Merritt, D.B. and Derr, J.N. (1999) Microsatellite diversity in captive bottlenose dolphins (Tursiops truncatus). Journal of Heredity 90: 228-253. doi: 10.1093/jhered/90.1.228

Rosel, P.E., Hansen, L. and Hohn, A.A. (2009) Restricted dispersal in a continuously distributed marine species: common bottlenose dolphins Tursiops truncatus in coastal waters of the western North Atlantic. Molecular Ecology 18: 5030- 5045. doi: 10.1111/j.1365-294X.2009.04413.x

Rosenberg, D.K., Overton, W.S. and Anthony, R.G. (1995) Estimation of animal abundance when capture probabilities are low and hetergeneous. Journal of Wildlife Management 59: 252-261. doi: 10.2307/3808938

Ross, G.J.B. (2006) Review of the conservation status of Australia’s smaller whales and dolphins. Department of the Environment and Water Resources, Canberra, ACT. 124 p.

Rossbach, K.A. and Herzing, D.L. (1999) Inshore and offshore bottlenose dolphin (Tursiops truncatus) communities distinguished by association patterns near Grand Bahama Island, Bahamas. Canadian Journal of Zoology 77: 581-592. doi: 10.1139/z99-018

Rousset, F. (2008) GENEPOP'007: a complete re-implementation of the genepop software for Windows and Linux. Molecular Ecology Resources 8: 103-106. doi: 10.1111/j.1471-8286.2007.01931.x

Salgado Kent, C., McCauley, R.D., Parnum, I.M., Gavrilov, A.N., Kent, C.P.S. and Gavrilov, N. (2012) Underwater noise sources in Fremantle inner harbour: dolphins, pile driving and traffic. Australian Acoustical Society 1-7.

Sampey, A., Fromont, J. and Johnston, D.J. (2011) Demersal and epibenthic fauna in a temperate marine embayment, Cockburn Sound, Western Australia: determination of key indicator species. Journal of Royal Society of Western Australia 94: 1-18.

Sandoval-Castillo, J. and Beheregaray, L.B. (2015) Metapopulation structure informs conservation management in a heavily exploited coastal shark (Mustelus henlei). Marine Ecology Progress Series 533: 191-203. doi: 10.3354/meps11395

Schnell, G.D., Watt, D.J. and Douglas, M.E. (1985) Statistical comparison of proximity matrices: applications in animal behaviour. Animal Behaviour 33: 239-253. doi: 10.1016/S0003-3472(85)80138-X

213 References

Schwarz, C. (2002) Capture-recapture sampling designs. Encyclopedia of Environmetrics. doi: 10.1002/9780470057339.vac003.pub2

Seaman, D.E., Millspaugh, J.J., Kernohan, B.J., Brundige, G.C., Raedeke, K.J. and Gitzen, R.A. (1999) Effects of sample size on kernel home range estimates. The Journal of Wildlife Management 63: 739-747. doi: 10.2307/3802664

Sellas, A.B., Wells, R.S. and Rosel, P.E. (2005) Mitochondrial and nuclear DNA analyses reveal fine scale geographic structure in bottlenose dolphins (Tursiops truncatus) in the Gulf of Mexico. Conservation Genetics 6: 715-728. doi: 10.1007/s10592-005-9031-7

Sherwin, W.B., Jabot, F., Rush, R. and Rossetto, M. (2006) Measurement of biological information with applications from genes to landscapes. Molecular Ecology 15: 2857-2869. doi: 10.1111/j.1365-294X.2006.02992.x

Shinohara, M., Domingo-Roura, X. and Takenaka, O. (1997) Microsatellites in the bottlenose dolphin Tursiops truncatus. Molecular Ecology 6: 695-696. doi: 10.1046/j.1365-294X.1997.00231.x

Silva, M.a., Prieto, R., Magalhães, S., Seabra, M.I., Santos, R.S. and Hammond, P.S. (2008) Ranging patterns of bottlenose dolphins living in oceanic waters: Implications for population structure. Marine Biology 156: 179-192. doi: 10.1007/s00227-008-1075-z

Slatkin, M. (2008) Linkage disequilibrium - understanding the evolutionary past and mapping the medical future. Nature Reviews Genetics 9: 477-485. doi: 10.1038/nrg2361

Slooten, E., Fletcher, D. and Taylor, B.L. (2000) Accounting for uncertainty in risk assessment: case study of Hector's dolphin mortality due to gillnet entanglement. Conservation Biology 14: 1264-1270. doi: 10.1046/j.1523- 1739.2000.00099-411.x

Smith, H., Frère, C., Kobryn, H. and Bejder, L. (2016) Dolphin sociality, distribution and calving as important behavioural patterns informing management. Animal Conservation 19: 462-471. doi: 10.1111/acv.12263

Smith, H.C., Pollock, K., Waples, K., Bradley, S. and Bejder, L. (2013) Use of the robust design to estimate seasonal abundance and demographic parameters of a coastal bottlenose dolphin (Tursiops aduncus) population. PLoS ONE 8: e76574. doi: 10.1371/journal.pone.0076574

Smith, H.C. and Sprogis, K.R. (2016) Seasonal feeding on giant cuttlefish (Sepia apama) by Indo-Pacific bottlenose dolphins (Tursiops aduncus) in southwestern Australia. Australian Journal of Zoology 64: 8-13. doi: 10.1071/ZO15075

Smolker, R.A., Richards, A.F., Connor, R.C. and Pepper, J.W. (1992) Sex differences in patterns of association among Indian Ocean bottlenose dolphins. Behaviour 123: 38-69. doi: 10.1163/156853992X00101

214 References

Sollmann, R., Gardner, B., Parsons, A.W., Stocking, J.J., McClintock, B.T., Simons, T.R., Pollock, K.H. and O'Connell, A. (2013) A spatial mark–resight model augmented with telemetry data. Ecology 94: 553-559. doi: 10.1890/12-1256.1

Sprogis, K.R.-A., Pollock, K.H., Raudino, H.C., Allen, S.J., Kopps, A.M., Manlik, O., Tyne, J.A. and Bejder, L. (2016) Sex-specific patterns in abundance, temporary emigration and survival of Indo-Pacific bottlenose dolphins (Tursiops aduncus) in coastal and estuarine waters. Frontiers in Marine Science 3: 12. doi:10.3389/fmars.2016.00012

Sprogis, K.R., Raudino, H.C., Rankin, R., MacLeod, C.D. and Bejder, L. (2015) Home range size of adult Indo-Pacific bottlenose dolphins (Tursiops aduncus) in a coastal and estuarine system is habitat and sex-specific. Marine Mammal Science 32: 287-308. doi: 10.1111/mms.12260

Stanley, T.R. and Burnham, P.K. (1999) A closure test for time-specific capture- recapture data. Environmental and Ecological Statistics 6: 197-209. doi: 10.1023/A:1009674322348

Steidl, R.J. and Powell, B.F. (2006) Asessing the effects of human activities on wildlife. The George Wright Forum 23: 50-58.

Steiner, A. and Bossley, M. (2008) Some reproductive parameters of an estuarine population of Indo-Pacific bottlenose dolphins (Tursiops aduncus). Aquatic Mammals 34: 84-92. doi: 10.1578/AM.34.1.2008.84

Stensland, E. and Berggren, P. (2007) Behavioural changes in female Indo-Pacific bottlenose dolphins in response to boat-based tourism. Marine Ecology Progress Series 332: 225-234. doi: 10.3354/meps332225

Stephens, N., Duignan, P.J., Wang, J., Bingham, J., Finn, H., Bejder, L., Patterson, A.P. and Holyoake, C. (2014) Cetacean morbillivirus in coastal Indo-Pacific bottlenose dolphins, Western Australia. Emerging Infectious Disease Journal 20: 666-670. doi: 10.3201/eid2004.131714

Stone, G.S. and Yoshinaga, A. (2000) Hector's dolphin Cephalorhynchus hectori calf mortalities may indicate new risks from boat traffic and habituation. Pacific Conservation Biology 6: 162-170. doi: 10.1071/PC000162

Storz, J.F. (1999) Genetic consequences of mammalian social structure. Journal of Mammalogy 80: 553-569. doi: 10.2307/1383301

Sugg, D.W., Chesser, R.K., Dobson, F.S. and Hoogland, J.L. (1996) Population genetics meets behavioral ecology. Trends in ecology & evolution 11: 338- 342. doi: 10.1016/0169-5347(96)20050-3

Sweanor, L.L., Logan, K.A. and Hornocker, M.G. (2000) Cougar disperal patterns, metapopultion dynamics, and conservation. Conservation Biology 14: 798- 808. doi: 10.1046/j.1523-1739.2000.99079.x

215 References

Szpiech, Z.A. and Rosenberg, N.A. (2011) On the size distribution of private microsatellite alleles. Theoretical Population Biology 80: 100-113. doi: 10.1016/j.tpb.2011.03.006

Tajima, F. (1989) Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123: 585-595.

Taylor, B.L. (2002) Conservation biology. In: W.F. Perrin, B. Würsig and H. Thewissen (Eds.) Encyclopedia of marine mammals. Academic Press, San Diego, CA, pp. 273-276.

Taylor, B.L. (2005) Identifying units to conserve. In: J.E. Reynolds III, W.F. Perrin, R.R. Reeves, S. Montgomery and T.J. Ragen (Eds.) Marine Mammal Research: Conservation beyond Crisis. The Johns Hopkins University Press, Baltimore, MD, pp. 149-164.

Taylor, B.L., Martien, K. and Morin, P. (2010) Identifying units to conserve using genetic data. In: I.L. Boyd, W. Don Bowen and S.J. Iverson (Eds.) Marine mammal ecology and conservation: a handbook of techniques. Oxford University Press, New York, NY, pp. 306-324.

Tezanos-Pinto, G., Baker, C.S., Russell, K., Martien, K., Baird, R.W., Hutt, A., Stone, G., Mignucci-Giannoni, A.a., Caballero, S., Endo, T., Lavery, S., Oremus, M., Olavarría, C. and Garrigue, C. (2009) A worldwide perspective on the population structure and genetic diversity of bottlenose dolphins (Tursiops truncatus) in New Zealand. Journal of Heredity 100: 11-24. doi: 10.1093/jhered/esn039

Thomas, L., Laake, J.L., Rexstad, E., Strindberg, S., Marques, F.F.C., Buckland, S.T., Borchers, D.L., Anderson, D.R., Burnham, K.P., Burt, M.L., Hedley, S.L., Pollard, J.H., Bishop, J.R.B. and Marques, T.A. (2009) Distance 6.0. release 1. http://www.ruwpa.st-and.ac.uk/distance/

Thompson, K.F., Patel, S., Baker, C.S., Constantine, R. and Millar, C.D. (2016) Bucking the trend: genetic analysis reveals high diversity, large population size and low differentiation in a deep ocean cetacean. Heredity 116: 277-285. doi: 10.1038/hdy.2015.99

Titcomb, E.M., O'Corry-Crowe, G., Hartel, E.F. and Mazzoil, M.S. (2015) Social communities and spatiotemporal dynamics of association patterns in estuarine bottlenose dolphins. Marine Mammal Science 31: 1314-1337. doi: 10.1111/mms.12222

Todd, V.L.G., Todd, I.B., Gardiner, J.C., Morrin, E.C.N., MacPherson, N.A., DiMarzio, N.A. and Thomsen, F. (2015) A review of impacts of marine dredging activities on marine mammals. ICES Journal of Marine Science 72: 328-340. doi: 10.1093/icesjms/fsu187

Torres, A., Jaeger, J.A.G. and Alonso, J.C. (2016) Assessing large-scale wildlife responses to human infrastucture development. Proceedings of the National Academy of Sciences 113: 8472-8477. doi: 10.1073/pnas.1522488113

216 References

Tyne, J.A., Loneragan, N.R., Johnston, D.W., Pollock, K.H., Williams, R. and Bejder, L. (2016) Evaluating monitoring methods for cetaceans. Biological Conservation 201: 252-260. doi: 10.1016/j.biocon.2016.07.024

Tyne, J.A., Pollock, K.H., Johnston, D.W. and Bejder, L. (2014) Abundance and survival rates of the Hawai'i Island associated spinner dolphin (Stenella longirostris) stock. PLoS ONE 9: e86132. doi:86110.81371/journal.pone.0086132

Urian, K., Gorgone, A., Read, A., Balmer, B., Wells, R.S., Berggren, P., Durban, J., Eguchi, T., Rayment, W. and Hammond, P.S. (2014) Recommendations for photo-identification methods used in capture-recapture models with cetaceans. Marine Mammal Science 31: 298–321. doi: 10.1111/mms.12141

Urian, K.W., Hofmann, S., Wells, R.S. and Read, A.J. (2009) Fine-scale population structure of bottlenose dolphins (Tursiops truncatus) in Tampa Bay, Florida. Marine Mammal Science 25: 619-638. doi: 10.1111/j.1748-7692.2009.00284.x

Urian, K.W., Hohn, A.A. and Hansen, L.J. (1999) Status of the photo-identification catalog of coastal bottlenose dolphins of the western north Atlantic: report of a workshop of catalog contributors. NOAA Administrative Report NMFS_SEFSC.https://www.sefsc.noaa.gov//P_QryLDS/download/TM588_T M-425.pdf?id=LDS 22 p.

Van Bressem, M.-F.M.-F., Van Waerebeek, K., Aznar, F.J., Raga, J.A., Jepson, P.D., Duignan, P.P., Deaville, R., Flach, L., Viddi, F., Baker, J.R., Di Beneditto, A.P., Echegaray, M.M., Genovo, T., Reyes, J., Felix, F., Gaspar, R., Ramos, R., Peddemors, V.M., Sanino, G.P., Siebert, U. and Genov, T. (2009) Epidemiological pattern of tattoo skin disease: a potential general health indicator for cetaceans. Diseases of Aquatic Organisms 85: 225-237. doi: 10.3354/dao02080

Van Oosterhout, C., Hutchinson, W.F., Wills, D.P.M. and Shipley, P. (2004) Micro- Checker: software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes 4: 535-538. doi: 10.1111/j.1471- 8286.2004.00684.x

Walker, F.M., Sunnucks, P. and Taylor, A.C. (2008) Evidence for habitat fragmentation altering within-population processes in wombats. Molecular Ecology 17: 1674-1684. doi: 10.1111/j.1365-294X.2008.03701.x

Wan, Q.H., Wu, H., Fujihara, T. and Fang, S.G. (2004) Which genetic marker for which conservation genetics issue? Electrophoresis 25: 2165-2176. doi: 10.1002/elps.200305922

Wang, J. (2011) COANCESTRY: a program for simulating, estimating and analysing relatedness and inbreeding coefficients. Molecular Ecology Resources 11: 141-145. doi: 10.1111/j.1755-0998.2010.02885.x

Wang, J.Y., Chou, L.-S. and White, B.N. (1999) Mitochondrial DNA analysis of symaptric morphotypes of bottlenose dolphins (genus: Tursiops) in Chinese

217 References

waters. Molecular Ecology 8: 1603-1612. doi: 10.1046/j.1365- 294x.1999.00741.x

Wang, J.Y., Chou, L.-S. and White, B.N. (2000a) Differences in the external morphology of two sympatric species of bottlenose dolphins (Genus Tursiops) in the waters of China. Journal of Mammalogy 81: 1157-1165. doi: 10.1644/1545-1542(2000)081<1157:DITEMO>2.0.CO;2

Wang, J.Y., Chou, L.-S. and White, B.N. (2000b) Osteological differences between two sympatric forms of bottlenose dolphins (genus Tursiops) in Chinese waters. Journal of Zoology London 252: 147-162. doi: 10.1111/j.1469- 7998.2000.tb00611.x

Wang, J.Y. and Yang, S.-C. (2009) Indo-Pacific bottlenose dolphin (Tursiops aduncus). In: W.F. Perrin, B. Würsig and J.G.M. Thewissen (Eds.) Encyclopedia of Marine Mammals. Acadenic Press, San Diego, California, pp. 602-607.

Waples, K.A. (1997) The rehabilitation and release of bottlenose dolphins from Atlantis Marine Park, Western Australia. PhD thesis, Texas A&M University. 237 p.

Waples, R.S. (1998) Separating the wheat from the chaff: patterns of genetic differentiation in high gene flow species. Journal of Heredity 89: 438-450. doi: 10.1093/jhered/89.5.438

Waples, R.S., Antao, T. and Luikart, G. (2014) Effects of overlapping generations on linkage disequilibrium estimates of effective population size. Genetics 197: 769-780. doi: 10.1534/genetics.114.164822

Waples, R.S. and Do, C. (2010) Linkage disequilibrium estimates of contemporary Ne using highly variable genetic markers: a largely untapped resource for applied conservation and evolution. Evolutionary Applicationsl 3: 244-262. doi: 10.1111/j.1752-4571.2009.00104.x

Waples, R.S. and Do, C.H.I. (2008) ldne: a program for estimating effective population size from data on linkage disequilibrium. Molecular Ecology Resources 8: 753-756. doi: 10.1111/j.1755-0998.2007.02061.x

Waples, R.S. and England, P.R. (2011) Estimating contemporary effective population size on the basis of linkage disequilibrium in the face of migration. Genetics 189: 633-644. doi: 10.1534/genetics.111.132233

Water Corporation (2008) Southern seawater desalination project. Report and recommendations of the Environmental Protection Authority. Environmental Protection Authority, Perth, Western Australia 70 p.

Webster, I., Cockcroft, V.G. and Cadinouche, A. (2014) Abundance of the Indo- Pacific bottlenose dolphin Tursiops aduncus off south-west Mauritius. African Journal of Marine Science 36: 293-301. doi: 10.2989/1814232X.2014.946448

218 References

Weir, B.S. and Cockerham, C.C. (1984) Estimating F-statistics for the analysis of population structure. Evolution 38: 1358-1370. doi: 10.1111/j.1558- 5646.1984.tb05657.x

Wells, J.V. and Richmond, M.E. (1995) Populations, metapopulations, and species populations: what are they and who should care? Wildlife Society Bulletin 23: 458-462. doi: 10.2307/3782955

Wells, R.S. (1986) Population structure of bottlenose dolphins: Behavioral studies of bottlenose dolphins along the central west coast of Florida. Contract Report to National Marine Fisheries Service, Southeast Fisheries Center. Miami, FL 58 p.

Wells, R.S., Boness, D.J. and Rathbun, G.B. (1999) Behavior. In: J.E. Reynolds and S.A. Rommel (Eds.) Biology of Marine Mammals. Smithsonian Institution Press, Washington and London, pp. 324-422.

Wells, R.S. and Scott, M.D. (2009) Bottlenose dolphins: Tursiops aduncus and Tursiops truncatus. In: W.F. Perrin, B. Wursig and J. Thewissen (Eds.) Academic Press, San Diego, pp. 122-128.

Wells, R.S., Scott, M.D., Irvine, A.B. and Genoways, H. (1987) The social structure of free-ranging bottlenose dolphins. In: H.H. Genoways (Ed.) Current Mammalogy. Plenum, New York, pp. 247-305.

Western Australian Planning Commission (2004) Fremantle Ports outer harbour project - information brochure. Perth, Western Australia 11 p.

White, G., Kendall, W. and Barker, R. (2006) Multistate survival models and their extensions in program MARK. Journal of Wildlife Management 70: 1521- 1529. doi: 10.2193/0022-541X(2006)70[1521:MSMATE]2.0.CO;2

White, G.C. and Burnham, P.K. (1999) Program MARK: Survival estimation from populations of marked animals. Bird Study 46 Supplement: 120-138. doi: 10.1080/00063659909477239

Whitehead, H. (1995) Investigating structure and temporal scale in social organizations using identified individuals. Behavioural Ecology 6: 199-208. doi: 10.1093/beheco/6.2.199

Whitehead, H. (1999) Testing association patterns of social animals. Animal Behaviour 57: F26-F29. doi: 10.1006/anbe.1999.1099

Whitehead, H. (2001) Analysis of animal movement using opportunistic individual- identifications: application to sperm whales. Ecology 82: 1417-1432. doi: 10.2307/2679999

Whitehead, H. (2007) Selection of models of lagged identification rates and lagged association rates using AIC and QAIC. Communications in Statistics- Simulation and Computation 36: 1233-1246. doi: 10.1080/03610910701569531

219 References

Whitehead, H. (2008a) Analyzing animal societies: quantitative methods for vertebrate social analysis. Chicago University Press, Chicago, IL.

Whitehead, H. (2008b) Precision and power in the analysis of social structure using associations. Animal Behaviour 75: 1093-1099. doi: 10.1016/j.anbehav.2007.08.022

Whitehead, H. (2009) SOCPROG programs: analysing animal social structures. Behavioral Ecology and Sociobiology 63: 765-778. doi: 10.1007/s00265-008- 0697-y

Whitehead, H. (2016) SOCPROG: programs for analyzing social structure. http://whitelab.biology.dal.ca/SOCPROG/Manual.pdf

Whitehead, H., Bejder, L. and Ottensmeyer, A. (2005) Testing association patterns: issues arising and extensions. Animal Behaviour 69: e1-e6. doi: 10.1016/j.anbehav.2004.11.004

Wiens, J.A. (1976) Population responses to patchy environments. Annual Review of Ecology and Systematics 7: 81-120. doi: 10.2307/2096862

Williams, K.B., Nichols, D.J. and Conroy, J.M. (2002) Analysis and management of animal populations. Academic Press, San Diego, California, USA.

Wilson, G.A. and Rannala, B. (2003) Bayesian inference of recent migration rates using multilocus genotypes. Genetics 163: 1177-1191.

Wiszniewski, J., Allen, S.J. and Moller, L.M. (2009) Social cohesion in a hierarchically structured embayment population of Indo-Pacific bottlenose dolphins. Animal Behaviour 77: 1449-1457. doi: 10.1016/j.anbehav.2009.02.025

Wiszniewski, J., Beheregaray, L.B., allen, S.J., Moller, L.M. and Möller, L.M. (2010) Environmental and social influences on the genetic structure of bottlenose dolphins (Tursiops aduncus) in Southeastern Australia. Conservation Genetics 11: 1405-1419. doi: 10.1007/s10592-009-9968-z

Wright, S. (1931) Evolution in mendelian populations. Genetics 16: 97-159.

Würsig, B. and Jefferson, T.A. (1990) Methods of photo-identification for small cetaceans. Report of the International Whaling Commission Special Issue 12: 43-52.

Yannic, G., St-Laurent, M.-H., Ortego, J., Taillon, J., Beauchemin, A., Bernatchez, L., Dussault, C. and Côté, S.D. (2015) Integrating ecological and genetic structure to define management units for caribou in Eastern Canada. Conservation Genetics 17: 437-453. doi: 10.1007/s10592-015-0795-0

Zachos, F.E., Frantz, A.C., Kuehn, R., Bertouille, S., Colyn, M., Niedzialkowska, M., Perez-Gonzalez, J., Skog, A., Sprem, N. and Flamand, M.C. (2016) Genetic structure and effective population sizes in European red deer (Cervus

220 References

elaphus) at a continental scale: insights from microsatellite DNA. Journal of Heredity 107: 318-326. doi: 10.1093/jhered/esw011

Zolman, E.S. (2002) Residence patterns of bottlenose dolphins (Tursiops truncatus) in the Stono River estuary, Charleston County, South Carolina, U.S.A. Marine Mammal Science 18: 879-892. doi: 10.1111/j.1748-7692.2002.tb01079.x

221