CONSERVATION GENETICS OF THE

GROWLING GRASS FROG, LITORIA

RANIFORMIS, IN URBANISING LANDSCAPES

CLAIRE CATHERINE KEELY Bachelor of Science (Hons)

Submitted in total fulfillment of the requirements of the degree of

DOCTOR OF PHILOSOPHY

AUGUST 2016

School of BioSciences The University of Melbourne ORCID ID: 0000-0002-7243-4639

Conservation genetics of the Growling Grass Frog, Litoria raniformis, in urbanising landscapes By Claire C. Keely ! 2016 Supervisors: Kirsten Parris, Geoff Heard, Jane Melville and Andrew Hamer Title Page: Litoria raniformis juvenile, photo by C. C. Keely

For Dan and Max

Abstract

The proportion of the world’s human population living in cities and towns (urban areas) grew rapidly over the 20th century. Indeed, the global urban population grew by an order of magnitude during this period, from 220 million to 2.8 billion. By 2030, the global urban population is expected to swell to almost 5 billion. Urbanisation is a key threatening process for amphibians, with the global amphibian assessment listing greater than one-third of the world’s known amphibian species as currently threatened by urbanisation. As is the case for biodiversity more generally, habitat loss and fragmentation represent pervasive impacts of urbanisation for amphibians. The International Union for Conservation of Nature (IUCN) recognises genetic diversity as one of three forms of biodiversity requiring conservation. However, surprisingly few studies have focused on the genetic consequences of urbanisation for amphibians. With the global rate of urbanisation set to steadily increase, and its recognition as a key threatening process to amphibians, the application of genetic techniques will be an important component of conservation planning for these animals.

This thesis investigates the conservation genetics of the Growling Grass Frog, Litoria raniformis, around Melbourne, Australia. This species has declined significantly since the late 1970s, largely due to habitat loss and fragmentation, drought and disease. Remnant populations around Melbourne occur in four main regions, three of which are marked for urban growth, causing further loss, degradation and fragmentation of habitat for L. raniformis.

The aim of this thesis was to assess the genetic structure and diversity of remnant populations of L. raniformis across Melbourne. There were four main objectives:

1. Assess four different genetic sampling techniques for amphibians, using a multi- criteria decision framework and L. raniformis as a case study. 2. Document the genetic structure and diversity of L. raniformis across Melbourne’s urban fringe, using mitochondrial DNA (mtDNA) and microsatellites.

i 3. Investigate the population genetic structure of L. raniformis in the northern region of urbanising Melbourne and develop a model of the landscape determinants of gene flow for the species. 4. Undertake a Bayesian metapopulation viability analysis for L. raniformis, incorporating estimates of gene flow to define population connectivity.

The thesis concludes by outlining directions for further research on the conservation genetics of the Growling Grass Frog and its management.

ii Declaration

This is to certify that:

i. The thesis comprises only my original work towards the PhD except where indicated in the Preface,

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

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

______Claire C. Keely August 2016

iii iv Preface

This thesis was prepared as a collection of publications. Each data chapter (Chapters 2 to 5) constitutes an independent paper. Therefore, there may be some repetition between chapters, however this has been avoided where possible.

These publications have been prepared in co-authorship. However, the contents of these chapters are my own work, except where outlined below. Co-authorship reflects a supervisory role in most cases (K. M. Parris, G. W. Heard, J. Melville, and A. J. Hamer), but the result of collaboration in other instances. J. M. Hale provided genotypes and COI mtDNA sequences for 198 Litoria raniformis individuals (used in Chapters 3 and 4). J. Sumner provided advice on genetics laboratory techniques for Chapters 2 and 3 and A. Moran-Ordonez assisted with GIS analysis (Chapter 4).

The thesis is made up of the following chapters (Chapters 2 to 5) for publication:

Keely, C.C., Heard, G.W., Sumner, J.M., Parris, K.M. and Melville, J. Comparing genetic sampling techniques for amphibians with a multi-criteria decision framework Keely, C.C., Hale, J.M., Heard, G.W., Parris, K.M., Sumner, J.M. Hamer, A.J. and Melville, J. 2015. Genetic structure and diversity of the endangered growling grass frog in a rapidly urbanizing region. Royal Society Open Science 2(8): 1-13. Keely, C.C., Heard, G.W., Hale, J.M., Moran-Ordonez, A., Parris, K.M. and Melville, J. Impacts of urbanisation on the population genetic structure of a threatened amphibian, Litoria raniformis Keely, C.C., Parris, K.M., Melville, J. and Heard, G.W. Integrating genetic connectivity measures with stochastic patch occupancy models for metapopulation management

All research procedures reported in the thesis were approved by The University of Melbourne Animal Ethics Committee (approval number 1011804.1) and conducted under a scientific research permit from the Victorian Department of Environment, Land, Water and Planning (permit number 10005649).

v This study was generously funded by an Australian Research Council Linkage Grant (LP0990161, with the Australian Research Centre for Urban Ecology, Growling Grass Frog Trust Fund, Melbourne Museum, Melbourne Water, Parks Victoria and Victorian Department of Environment, Land, Water and Planning), The National Environmental Research Program, Environmental Decisions Hub, ARC Centre of Excellence for Environmental Decisions, The University of Melbourne (Melbourne Abroad Travelling Scholarship and School of Botany Postgraduate Student Travel Award), Museum Victoria (Museum Victoria 1854 Student Scholarship) and Australian Society of Herpetologists (ASH Student Travel Grant).

vi Acknowledgements

I dedicate this thesis to my incredible husband, Dan, and my beautiful PhD baby, Max.

Dan, I’m sorry I disappeared to do field work for entire summers. Thank you for your endless encouragement and strength, for helping put things in perspective and for always being a welcome distraction from the challenges of a PhD. I will never be able to thank you enough for choosing me.

Max, thank you for being a constant inspiration (and such a good sleeper, so I could address paper revisions when you were only a few weeks old). Your smiles and laughter and watching you grow into the most amazing tiny human have encouraged me to continue working and hopefully make you proud of your mama.

My parents, Cate and Kelvin, thank you for a lifetime of support. From finally giving in and allowing me to have my first pet frog at the age of seven, to entertaining Max Steel for countless days while I completed my PhD.

I wish to thank my supervisors, Kirsten Parris, Geoff Heard, Jane Melville and Andrew Hamer, who have all played important roles during different stages of my PhD. Geoff, I couldn’t have done this without you, thank you for being such an amazing supervisor and for sharing your passion for Growling Grass Frogs with me. We have very different attitudes towards this species and it’s behaviour towards others, however definitely share a love and respect for the little guys. Kirsten and Jane, I’m so happy that I chose two strong, intelligent women to guide me through this process, thank you for being great role models, as well as supervisors. Kirsten, I really admire your scientific integrity and the ongoing commitment and support you provide your students, especially after they’re no longer students. Jane, I really appreciate your ability to turn a piece of writing into a story. Andrew, thank you for some fun field adventures.

Josh Hale, thank you for allowing me to follow on from your PhD work and being so patient as you taught me how to navigate a genetics lab. Your guidance and support, especially at the beginning of my candidature, were so important to me.

vii Thank you to all my friends at Melbourne Museum. Katie Date, my amazing friend and PhD mentor, I cannot imagine these years without you. I’m so lucky to have such an intelligent and caring kindred spirit. Claire McLean (the ‘other me’). Thank you for all your help and encouragement. Thank you also to Jo Sumner for all your guidance and lab support, you always felt like an honarary supervisor and to Maggie, Susi, Sumi, Shandiya, Nat, Zoe, Alice, Bec, Adnan and many more.

I’ve been really lucky to be part of such a friendly and supportive Uni lab. Lab retreats, writing retreats and attending conferences have definitely been highlights and there are far too many people to mention everyone individually. However, I would particularly like to mention Stefano Canessa, my field work companion for the entire first season. Thank you for all your hard work, expert frog-catching skills and always being so incredibly accommodating (your constant reply of ‘as you wish’ would make me smile every time). Reid Tingley, my official mentor, thank you for the interesting stories and always having a smile on your face. Alejandra, thank you for your incredible help with ArcGIS, and Will and John B for your patient computer support. And to the incredible mamas in the lab, especially Emily and Carly, thank you for your advice on balancing PhD and babies and for general inspiration. Thank you also to James, Kim, Chris, Freya, Luke, Laura, Els, Matt, Pia, Casey, Heini, the list goes on...

I had so much fun in the field during my PhD, and would like to thank the amazing people out with me – Stefano, Susi, Nat, Fran, Dani, Geoff and Andrew. Thank you for all the adventures.

During my canditure I undertook part-time work with the Live Exhibits Department of Melbourne Museum. I would especially like to thank my manager, Patrick, for being so accommodating of my limited availability.

My dear friend, Jo Ainley, I would also like to dedicate this thesis to you. Over 10 years we shared countless adventures chasing frogs and I still can’t believe we never will again. I hope that you have finally found peace.

viii Table of Contents

Abstract i Declaration iii Preface v Acknowledgements vii Chapter 1 General introduction 1 Chapter 2 Comparing genetic sampling techniques for amphibians with a multi-criteria decision framework 17 Chapter 3 Genetic structure and diversity of the endangered Growling Grass Frog in a rapidly urbanising region 39 Chapter 4 Impacts of urbanisation on the population genetic structure of a threatened amphibian, Litoria raniformis 77 Chapter 5 Integrating genetic connectivity measures with stochastic patch occupancy models for metapopulation management 101 Chapter 6 General discussion 123 Literature Cited 129 Appendix A Natural history note: Leucism 157

ix x List of Tables

Table 1.1 Studies focusing on the genetic consequences of urbanisation for amphibians. 8 Table 2.1 Considerations for potential future use of each sample type. 29 Table 2.2 Considerations for individual welfare for each sample type. 31 Table 2.3 Scores for each decision criterion when evaluating sampling techniques for the Growling Grass Frog, Litoria raniformis: DNA quantity, genetic accuracy, potential future use of samples, species welfare and individual welfare. (a) Raw scores; (b) Normalized, weighted scores. Using SMART, scores were normalized for each decision criteria on a 0-1 scale, where 1 is the sampling technique considered to be most favorable and 0 the least. A weight was applied to each decision criteria reflecting its relative importance, between 0 and 100, with 100 given to the criteria considered the most important. 32 Table 3.1 Genetic variability of mitochondrial DNA (COI) and microsatellites in populations determined by STRUCTURE. 54 Table 4.1 Summary microsatellite statistics across 11 loci and 9 populations. AR =

mean allelic richness, HO = Observed heterozygosity and HE = Expected heterozygosity. 89 Table 4.2 Estimates of the proportion of individuals in each population that can be assigned as migrants from another population using BayesAss. Immigrant source populations (SOURCE) are listed in the left-hand column, receiving populations (INTO) are listed across the top row. Numbers in parentheses represent the 95% credible intervals. Bold values are the proportion of frogs assigned to the site of capture and are therefore non-migrant individuals. 93 Table 4.3 Variables, deviance information criterion (DIC) results and weight for models of genetic distance. 94 Table 4.4 Coefficient estimates (mean and 95% credible interval) for the three explanatory variables included in the best regression model. 94 Table 5.1 Summary statistics (mean and 95% credible interval) for the posterior distributions of the regression coefficients for the effects of effective

xi wetland area (Aeff), aquatic vegetation cover (V), wetland type (WT) and connectivity (S, log transformed) on the probabilities of persistence and colonisation for Litoria raniformis. The summaries are based on the 5000 samples from the posterior, as used for the simulations of metapopulation viability. 113

xii List of Figures

Figure 1.1 The study species, Litoria raniformis: (a) & (b) adults exhibiting typical colour variations, (c) & (d) L. raniformis habitat north of Melbourne. 13 Figure 1.2 The geographic distribution of Litoria raniformis around Melbourne, Australia. Black circles indicate detection records since 2000. The light blue area is Port Phillip Bay, dark pink indicates the current Melbourne metropolitan area and light pink indicates the future urban growth area. 14 Figure 2.1 Flow chart demonstrating a decision framework for designing a conservation genetics study. 24 Figure 3.1 Geographic distribution and sample sites of Litoria raniformis around Melbourne. Dark grey = current Melbourne metropolitan area, light grey = proposed urban growth area. Closed circles indicate detection records between 2000 and the present. Open circles indicate the location of sampling sites included in this study. The four study regions around Melbourne are indicated and clusters of sites within these regions circled (dashed lines). 44 Figure 3.2 Haplotype network for 495bp of the COI gene. Each pie represents a unique genetic sequence (haplotype) and the area of the pie is proportional to haplotype frequency within the entire data set. Each line represents one mutational step. Small black circles correspond to inferred alleles, missing from the data set. 51 Figure 3.3 Geographic association of COI haplotypes. Each pie represents the haplotypes found in that cluster and the area of the pie is proportional to sample size. Clusters have been assigned letters and haplotypes numbered. See included table for the number of individuals displaying each haplotype. 52 Figure 3.4 Structure bar plot (K=2) with individuals organised by geographic region. Each vertical bar represents a single individual and estimates of Q - the probability an individual belongs to a population of the given colour. 54

xiii Figure 4.1 Populations of Litoria raniformis sampled north of Melbourne. Grey = current Melbourne metropolitan area. (a) All sample sites; (b) Sample sites included in population assignment analyses, grouped into circled clusters. Small, closed circles = sample sites from current study, small open circles = sample sites from previous study. 83 Figure 4.2 Population structure of Litoria raniformis north of Melbourne. Bayesian population assignment analysis using GENELAND. Black dots represent sampling locations from each of the nine clusters. 91 Figure 4.3 Population structure of Litoria raniformis north of Melbourne. Bayesian population assignment analysis using GENELAND. Black dots represent sampling locations. The red and yellow colours represent genetic units, and the contours the posterior probability of belonging to the focal genetic unit. (a) Cluster two; (b) cluster three; (c) cluster six; (d) cluster seven; and (e) cluster nine. (Clusters that did not display genetic sub- division are not presented). 92 Figure 4.4 Relationships between genetic distance and (a) geographic distance, (b) the proportion of urban infrastructure, and (c) stream connectivity. For the continuous variables, the solid lines represent the mean relationship and the broken lines the 95% credible intervals. For the binary variable (stream connectivity), mean estimates are represented by circles and vertical lines represent the 95% credible intervals. 95 Figure 4.5 The estimated multiplicative effect on genetic distance of the three explanatory variables. Mean estimates are represented by circles and vertical lines represent the 95% credible intervals. 96 Figure 5.1 Monitoring sites for Litoria raniformis (closed circles) north of Melbourne, Victoria, Australia, including the Darebin, Merri and Moonee Ponds Creek catchments. Open circles indicate sites sampled during the current study and open squares indicate sites sampled during a previous study. The rectangle indicates the subset of sites used in the simulations of management effectiveness. 106 Figure 5.2 The subset of monitoring sites for Litoria raniformis (black dots) used for the simulations of management effectiveness. Sites likely to be lost under future urbanisation are indicated by dots marked with a cross and potential new sites (to be constructed to offset the impacts of xiv urbanisation) are indicated by white dots. Potential new sites are labelled 1 - 11. Scenario 2 includes new sites 3, 4, 5 and 6. Scenario 3 includes new sites 2, 8, 10 and 11. Scenario 4 includes new sites 1, 2, 3, 4, 5 and 6. Scenario 5 includes new sites 2, 7, 8, 9, 10 and 11. 111 Figure 5.3 Relationships between: (a) the probability of colonisation and genetic connectivity (S, log transformed) for Litoria raniformis, and; (b) the

probability of persistence and effective wetland area (Aeff), wetland type (WT), aquatic vegetation cover (V) and genetic connectivity (S, log transformed). The solid lines represent the mean relationship, and grey shading the 95% credible intervals. Relationships are shown with all other variables held at their mean values. 114 Figure 5.4 The predicted number of wetlands occupied by Litoria raniformis over 30 years for each scenario. The scenarios were: (a) Current conditions are maintained; (b) With sites removed due to future urbanisation (Scenario 1); (c) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a cluster (Scenario 2); (d) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a stepping stone formation (Scenario 3); (e) With sites removed due to future urbanisation and 6 new offset wetlands arranged in a cluster (Scenario 4), and; (f) With sites removed due to future urbanisation, and 6 new wetlands arranged in a stepping stone formation (Scenario 5). The solid line shows the mean estimate and the dashed lines the 95% CI. 115 Figure 5.5 The predicted minimum number of wetlands occupied by Litoria raniformis over 30 years for each scenario. The scenarios were: (a) Current conditions are maintained; (b) With sites removed due to future urbanisation (Scenario 1); (c) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a cluster (Scenario 2); (d) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a stepping stone formation (Scenario 3); (e) With sites removed due to future urbanisation and 6 new offset wetlands arranged in a cluster (Scenario 4), and; (f) With sites removed due to future urbanisation, and 6 new wetlands arranged in a stepping stone formation (Scenario 5). The solid line shows the mean estimate and the dashed lines the 95% CI. Triangles represent mean estimates derived

xv using the updated SPOM with genetic connectivity. Circles represent mean estimates derived from the original model excluding urban barriers. Vertical lines represent the 95% credible intervals. 116

xvi Chapter 1

CHAPTER 1 General introduction

1 Chapter 1

URBANISATION, CONSERVATION GENETICS AND AMPHIBIANS

1. Introduction

The proportion of the world’s human population living in cities and towns (urban areas) grew by an order of magnitude during the 20th century, from 220 million to 2.8 billion. By 2030, the global urban population is expected to swell to 5 billion (United Nations 2010), with 1.2 million km2 of new urban land (Seto et al. 2012). Perhaps unsurprisingly, urbanisation – which can be characterised as the spread of urban infrastructure to support dense human habitation (McDonnell and Pickett 1990, McIntyre and Rango 2009) – represents a burgeoning threat to the conservation of biodiversity. In fact, the majority of species legally recognised as threatened are impacted by human development in various forms (Lynch et al. 2015). Urbanisation is often more lasting than other forms of habitat loss, which may be reversed through restoration (McKinney 2002).

Urbanisation can negatively impact species through a variety of processes, including altering the physical and chemical environment and modifying biotic assemblages. An analysis of >800 avian species found that most lack the appropriate adaptations for exploiting resources and avoiding risks in urban environments !"#$% &'% ($)% *+,-.. A pervasive concern is the loss and fragmentation of habitat. Although most researchers do not distinguish between the processes of habitat loss and fragmentation, the two can have different effects on biodiversity (Fahrig 2003). Empirical studies suggest that habitat loss has large, consistently negative effects on biodiversity (Fahrig 2003). Habitat fragmentation divides large areas of habitat into smaller patches through the removal of original habitat, the reduction of habitat patch size and the increasing isolation of these remnant patches (Andren 1994). These processes can result in the restriction of dispersal through barriers and alterations to the landscape matrix (Lindenmayer and Fischer 2006).

2. The genetic consequences of urbanisation

Landscape fragmentation and habitat loss due to urbanisation can reduce the size and

2 Chapter 1 genetic viability of populations (Irwin and Bockstael 2007, Noel and Lapointe 2010). Small populations in small fragments of habitat are susceptible to the erosion of genetic diversity, including random genetic drift and inbreeding depression (Reed and Hobbs 2004). Numerous studies have demonstrated a direct link between fragmentation in urban environments and lowered genetic diversity. Examples include studies of insects (Vandergast et al. 2007), mammals (Robinson and Marks 2001, McClenaghan and Truesdale 2002, Sato et al. 2014), birds (Sadanandan and Rheindt 2015), reptiles (Gray 1995, Row et al. 2011) and amphibians (Hitchings and Beebee 1998, Munshi-South et al. 2013).

Furthermore, for ground dwelling species, habitat isolation often involves a combination of both the distance between fragments of habitat and the landscape resistance between these patches (Vos and Chardon 1998). Dispersal rates and genetic connectivity between small, remnant populations can be inhibited by the non-habitat matrix surrounding remnant patches. These barriers to dispersal in urban areas can include roads and other infrastructure, such as buildings and fences. Open areas can similarly present barriers to dispersal when they are avoided by migrants (given predation risks or similar), or are physiologically stressful to cross (Hamer and McDonnell 2008). Few studies have examined matrix resistances directly by measuring individual movement across different non-habitat types (Ricketts 2001).

Traditional approaches to the study of fragmented landscapes assumed that the non- habitat matrix surrounding remnant patches was uniform (Ricketts 2001) and incorporated ‘isolation by distance’ models. These models were widely applied and provided powerful explanations of population structure, yet they assumed landscape homogeneity (McRae 2006). One alternative, described by McRae (2006), is the ‘isolation by resistance’ model, which incorporates landscape heterogeneity into isolation by distance analyses by modelling landscape connectivity using circuit theory. This approach objectively identifies important connective elements, by replicating expert opinion (McRae et al. 2008). Models such as this, that incorporate landscape variables, including urban land use, perform significantly better than simple isolation by distance tests in explaining genetic structuring among populations (Storfer et al. 2010).

The negative effects of road and traffic density on genetic connectivity and dispersal, as

3 Chapter 1

well as on mortality levels, have received increasing attention. These effects can contribute to reductions in effective population sizes and therefore reduced genetic diversity and population fitness. Traffic can impose problems for many large and medium sized mammals (Seiler et al. 2004), including badgers (Meles meles) in Britain, for whom road traffic is the largest single cause of recorded death (Clarke et al. 1998). Each year, road traffic kills hundreds of millions of animals (Rytwinski et al. 2016). A study by Lode (2000) on the impact of a motorway in Western France on vertebrate mortality and isolation found animal mortality increased exponentially with traffic volume. Ninety-seven vertebrate species were found dead during a 33 week period, including nine amphibian species and four reptile species. Moreover, these deaths were not restricted to ground-dwelling species, with 56 species of bird and six bat species recorded among the deaths. Shine et al. (2004) found roads modified movement patterns in garter snakes and impaired mate-location abilities. The effects of habitat isolation by roads on population genetic structure have been demonstrated even for highly mobile species (Vandergast et al. 2007). Riley et al. (2006) found a major highway near Los Angeles, USA, presented a formidable barrier to dispersal for coyotes (Canis latrans) and bobcats (Lynx rufus), restricting gene flow even in these wide-ranging species.

Conservation genetics has the potential to be an important tool for the management of threatened species in urban landscapes, improving our ability to make effective recommendations for management (Miller and Hobbs 2002, Noel and Lapointe 2010). Understanding landscape permeability may provide opportunities for reducing isolation, and therefore the extinction risk of fragmented populations (Ricketts 2001). The inclusion of genetic information in management decisions for fragmented urban populations can also facilitate planning for population translocations and augmentation, including the maintenance of corridors and addition of road crossing structures (Bos and Sites 2001, Balkenhol and Waits 2009).

The concept of landscape genetics was formalised by Manel et al. (2003) as a combination of population genetics and landscape ecology, which involves the correlation of genetic discontinuities with landscape and environmental features. It allows us to study the influence of landscape structure on gene flow and spatial genetic variation (Storfer et al. 2007). Landscape genetics is a rapidly evolving field which is playing an increasingly important role in the management and conservation of species

4 Chapter 1 threatened by habitat loss and fragmentation (Holderegger and Wagner 2008, Segelbacher et al. 2010). Despite this, as of 2010, only 7% of landscape genetic studies had been conducted in urban areas (Storfer et al. 2010). Although it can be extremely useful in assessing connectivity in urban species, many of the commonly used analytical approaches assume equilibrium conditions (Manel and Holderegger 2013). Urbanisation produces landscapes comprised of elements that differ strongly in their permeability, characteristics predicted to increase the applicability of landscape genetic analyses (Jaquiery et al. 2011, Munshi-South 2012). Landscape genetics has enormous potential for conservation management, at a time when landscape transformation is creating new challenges for species survival (Sork and Waits 2010). However, it appears that only a few landscape genetic studies to date have been applied to practical conservation management (Keller et al. 2015). One example is the Bighorn Sheep (Ovis canadensis nelsoni) in California, USA. After a study found highways were barriers to gene flow for the species, over- and underpasses were included in the species’ recovery plan (Epps et al. 2005, Keller et al. 2015).

3. Urbanisation as a key threatening process for amphibians

The IUCN Red List of Threatened Species categorises more than one-third of the world’s listed amphibian species as threatened by urbanisation (IUCN 2015). As is the case for biodiversity more generally, habitat loss, fragmentation and isolation represent pervasive impacts of urbanisation for amphibians. The vulnerability of amphibians to habitat fragmentation is multifaceted. The effects of fragmentation appear amplified by relatively low vagilities, narrow habitat tolerances and high mortality rates when moving between habitats (Cushman 2006). Furthermore, amphibians display complex life cycles, with many effectively occupying two niches, developing from an aquatic larva to a terrestrial adult (Hazell et al. 2001). Aquatic breeders, which make up roughly 97% of threatened species (Becker and Loyola 2008), rely on standing water for the development of their eggs and larvae. Amphibian survival in an urban environment therefore relies on the availability of both aquatic and terrestrial habitats. Numerous studies have identified the preservation and connectivity of both aquatic and terrestrial habitats as necessary for the maintenance of local amphibian populations (Semlitsch 1998, Porej et al. 2004, Rothermel 2004, Becker et al. 2007).

5 Chapter 1

Population connectivity in amphibians is predominantly achieved through juvenile dispersal (deMaynadier and Hunter 2000, Rothermel 2004). Many studies have indicated that following metamorphosis, juveniles can disperse over relatively large distances, contributing more to regional persistence than adult dispersal (Sinsch and Seidel 1995, Cushman 2006, Semlitsch 2008). For example, Sinsch (1997) found that almost all Natterjack Toad (Bufo calamita) metamorphs left their natal ponds within a few weeks after emergence and migrated up to 2 km, helping prevent local extinction within a metapopulation network. However, infrastructure and open areas associated with urbanisation can substantially reduce dispersal success and juvenile survival (deMaynadier and Hunter 1999). Juveniles are less experienced than adults with the terrain between natal ponds and terrestrial refuges and may be less able to determine the shortest route (Rothermel 2004). Moreover, their small body size can make them vulnerable to desiccation (Rothermel and Semlitsch 2002). For example, Semlitsch (1981) found high mortality of juvenile Mole Salamanders (Ambystoma talpoideum) during their initial dispersal, with 53% found dead or dying from heat stress or desiccation.

In a fragmented urban landscape species with the ability to disperse both short and long distances appear vulnerable, although in different ways (Cushman 2006). Individuals of species that can disperse over long distances generally will have an increased probability of encountering roads and other anthropogenic barriers. These barriers can disrupt genetic connectivity and dispersal, and lead to increasing mortality levels in amphibian populations. Results from a number of studies support the hypothesis that traffic mortality has a significant negative effect on anuran densities (Fahrig et al. 1995, Cosentino et al. 2014, Marsh et al. 2017). In a study on the effects of road kills on amphibian populations in Denmark (Hels and Buchwald 2001), results indicated that approximately 10% of the adult populations of Spadefoot Toads (Pelobates fuscus), Moor Frogs (Rana arvalis) and Common Frogs (Rana temporaria) were killed by traffic annually at the site investigated. Population projections of Spotted Salamanders (Ambystoma maculatum) estimate that an annual risk of adult road mortality greater than 10% can lead to local population extinctions (Gibbs and Shriver 2005).

Additional effects of road and traffic density may include an increase in pollutants,

6 Chapter 1 vibrations and noise, leading to increased mortality or disrupted amphibian behaviour (Fahrig et al. 1995). In a study by Parris et al. (2009), Southern Brown Tree Frogs (Litoria ewingii) were found to increase the pitch of their advertisement call in response to acoustic interference from traffic noise. This may have interesting evolutionary consequences, as higher frequency calls increase male audibility and may also decrease attractiveness to potential mates (Parris et al. 2009).

4. Conservation genetics of amphibians in urbanising landscapes

Despite the clear threat urbanisation poses to amphibians, few studies have focused on the genetic consequences of urbanisation for these animals (see Table 1.1 for an overview). With the global rate of urbanisation set to steadily increase, and its recognition as a key threatening process to amphibians, the application of genetic techniques will be an important component of conservation planning for these animals. For example, although population size can be estimated by either demographic or genetic methods, demographic estimates are usually both difficult to obtain and imprecise. It is generally accepted that genetic estimates are more accurate and can be more informative for understanding the overall health of a population (Beebee 2005). Genetic measures of the effective population size of amphibians in urban landscapes are particularly useful, as in addition to providing estimates of population viability and extinction risk, they can help to identify whether genetic factors such as inbreeding depression are involved in population declines (Funk et al. 1999). The discrepancy between the actual and effective size of breeding populations in amphibians is generally quite substantial (Schmeller and Merilä 2007). Many amphibian populations have very small effective population sizes, commonly less than 100 (Funk et al. 1999). This suggests that amphibian populations are especially susceptible to loss of genetic diversity through genetic drift and inbreeding, with potentially serious consequences in fragmented urban landscapes (Allentoft and O'Brien 2010). It is, however, important to acknowledge that although many questions are best answered with data on genetic connectivity, there is value in combining these with demographic measures to elucidate the complex role of dispersal in natural populations (Lowe and Allendorf 2010).

7 8 Table 1.1 Studies focusing on the genetic consequences of urbanisation for amphibians

Study Method Results/Conclusions Reference

Genetic substructuring as a result of barriers to Allozyme Genetic diversity & fitness lowest in urban environment, but (Hitchings and Beebee 1997) gene flow in urban common frog (Rana electrophoresis population size did not decrease. temporaria) populations

Effects of fragmentation on genetic diversity & mtDNA Genetic diversity positively related to fragment size, but genetic (Dixo et al. 2009)

structure of Rhinella ornata! (15% of fragmented differentiation was not related to geographic distance and areas was urban) fragment size did not significantly alter patterns of genetic connectivity.

Comparing genetic diversity and fitness of Allozyme analysis Genetic diversity, survival and developmental homeostasis were (Hitchings and Beebee 1998) Chapter 1 common toad (Bufo bufo) populations in urban and and minisatellites significantly lower in small, urban populations than in larger, rural habitats rural populations.

Structure and fragmentation of the growling grass Microsatellites Genetic distance significantly correlated with presence of urban (Hale et al. 2013) frog (Litoria raniformis) in an urbanising barriers between populations. landscape

Genetic population differentiation and connectivity Microsatellites Lower level of gene flow among populations that have (Arens et al. 2007) among fragmented moor frog (Rana arvalis) experienced a higher degree of fragmentation for a longer period. populations Negative effect of roads was found.

Impact of urban fragmentation on the genetic Microsatellites Urban populations were genetically differentiated and allelic (Noel et al. 2007) structure of the eastern red-backed salamander richness and heterozygosity lower. (Plethodon cinereus)

Genetic structure of urban red-backed salamander Microsatellites High genetic differentiation between urban and non-urban (Noel and Lapointe 2010) (Plethodon cinereus) populations in a major city populations. No clear genetic structure detected in urban area, but and nearby islands genetic differentiation was observed at a small spatial scale. No inbreeding for any population, but genetic variation relatively low.

The influence of land use and topographic distance Gel electrophoresis Motorways and railways had significant barrier effect on frog (Reh and Seitz 1990) on the genetic structure of populations of the populations within 3-4km. common frog (Rana temporaria)

Gene flow in wood frog (Lithobates sylvaticus) Microsatellites Despite extensive urbanisation, genetic homogeneity preserved, (Furman et al. 2015) populations in an urban landscape inhabiting both indicating constructed wetlands may preserve gene flow. constructed and natural wetlands Chapter 1

Population differentiation, genetic variation, and Microsatellites All populations genetically differentiated from each other, and the (Munshi-South et al. 2013) bottlenecks among urban populations of the most isolated populations have maintained very little genetic northern dusky salamander (Desmognathus fuscus) variation. A majority of the populations exhibited genetic bottlenecks.

The role of golf courses in maintaining genetic Microsatellites Golf course populations did not differ from natural populations in (Saarikivi et al. 2013) connectivity between common frog (Rana terms of genetic variability or differentiation. Suggests golf temporaria) populations in an urban setting courses prevent isolation and loss of genetic variability within populations.

Assessment of census and effective population size Microsatellites High Ne/N population size ratio obtained in this small and (Alvarez et al. 2015) in urban and isolated populations of fire isolated population suggests the existence of mechanisms of salamander (Salamandra salamandra) genetic compensation.

9

!

10 Genetic structure of marsh frog (Pelophylax Microsatellites Reduced genetic diversity, lower effective population sizes, and (Mikulicek and Pisut 2012) ridibundus) populations in an urban landscape higher genetic differentiation for spatially isolated urban populations. Population structure also shows higher genetic connectivity within water paths than between them, suggesting limited overland dispersal.

The effects of 20 years of highway presence on the Gel electrophoresis Separation by highways has reduced the genetic diversity and (Lesbarreres et al. 2006) genetic structure of agile frog (Rana dalmatina) polymorphism in local populations and resulted in a higher degree populations of population differentiation.

Landscape genetics of the alpine newt (Mesotriton Microsatellites Identifies land-uses that act as dispersal barriers (i.e. urban areas) (Emaresi et al. 2011) alpestris) inferred from a strip-based approach and corridors (i.e. forests). (urban was 1 of 14 land use types investigated) Chapter 1 Effects of roads on patterns of genetic Microsatellites Gene flow across very large roads is rare and bisected red-backed (Marsh et al. 2008) differentiation in red-backed salamanders salamander populations are likely to diverge from one another. (Plethodon cinereus) Smaller roads did not appear to be affecting genetic population structure.

Combining demography and genetic analysis to Microsatellites Demographic simulations and genetic analyses suggested different (Safner et al. 2010) assess population structure of the common frog genetic structuring, likely due to recent landscape fragmentation. (Rana temporaria) in a human-dominated landscape

Population genetics of fire salamanders in a pre- Microsatellites No clear-cut sign of genetic differentiation could be detected. (Straub et al. 2015) Alpine urbanized area (Salzburg, Austria) Habitat alteration effects might take several generations before leading to isolated genetic pools.

!

! Chapter 1

Some amphibian populations are structured as metapopulations (Smith and Green 2005). Most metapopulation studies consider the dynamics of a number of populations linked through dispersal, that may be subject to local extinction and recolonisation events (Marsh and Trenham 2001). This recurrent extinction and recolonisation is expected to reduce levels of genetic diversity, both within populations and across species (Pannell and Charlesworth 2000). However, habitat fragmentation and isolation in urban environments can increase the extinction risk of metapopulations by reducing the genetic input from immigrants and the potential for recolonisation following local extinction (Cushman 2006). Evidence of disrupted metapopulation dynamics of amphibians in urban landscapes comes from several molecular studies. Using assignment tests, a study by Hale (2010) found strong evidence for limited dispersal and gene flow for the endangered Growling Grass Frog (Litoria raniformis) in Melbourne, Australia. A reduction in genetic diversity and consequently fitness has been documented both at a population and species level for amphibian populations in urban environments and significantly higher levels of genetic differentiation between amphibian populations have been found in urban habitats when compared with non- urban habitats (Hitchings and Beebee 1997, Arens et al. 2007, Noel et al. 2007, Noel and Lapointe 2010). For example, Arens et al. (2007) found higher levels of genetic differentiation between populations that had experienced a higher degree of fragmentation due to urbanisation in a study of Moor Frogs (Rana arvalis) in The Netherlands.

Many studies that identify substantial genetic differentiation between amphibian populations due to urban barriers to dispersal similarly report reduced genetic diversity within amphibian populations. Measures of genetic diversity within populations such as allelic richness, haplotype diversity, heterozygosity and polymorphism have been found to be lower in amphibians in urban areas than in non-urban environments (Reh and Seitz 1990, Hitchings and Beebee 1997, Hitchings and Beebee 1998, Arens et al. 2007, Noel et al. 2007, Noel and Lapointe 2010). A study by Reh and Seitz (1990) on the Common Frog (Rana temporaria) in Germany found an influence of urban land use on the genetic structure of populations, with highway separation reducing the average heterozygosity and genetic polymorphism of local populations. Additionally, larval mortality and developmental abnormalities, long considered indicators of loss of genetic diversity and inbreeding, have also been documented in urban amphibian populations

! 11 Chapter 1

(Hitchings and Beebee 1997, Hitchings and Beebee 1998, Carrier and Beebee 2003). A study by Hitchings and Beebee (1997) on urban populations of the Common Frog (Rana temporaria) in Brighton, UK, found elevated levels of physical abnormalities and increased mortalities in developing tadpoles.

Conservation genetics can provide valuable direction for the management of threatened amphibian species in urbanising landscapes. However, despite its recognised importance, there is scope for the broader application of conservation genetics to fragmented populations which require active management (Jehle 2010). Few studies have attempted to incorporate both demographic and genetic structure to gain strategies for effective conservation management. Population viability analysis is an important tool for predicting the probability of persistence and informing conservation management decisions (Heard et al. 2013). For example, Greenwald (2010a) illustrated how the incorporation of both initial abundance and dispersal estimates derived from genetic data could be used to inform population viability analysis for two species of salamander. Assessing the genetic origins of populations and effectively managing genetic diversity are central to the persistence of these urban populations in both the short- and long-term (Armstrong and Seddon 2008, Tracy et al. 2011).

5. Study species: The Growling Grass Frog

The Growling Grass Frog, Litoria raniformis, is endemic to south-eastern Australia, including Victoria, New South Wales, South Australia and Tasmania (Pyke 2002). It was introduced to New Zealand in 1867, where it is now widely distributed (see Vörös et al. 2008). Litoria raniformis is listed as vulnerable to extinction in Australia under the Environment Protection and Biodiversity Conservation Act 1999. The species was formerly abundant across much of its range (Pyke 2002), however its distribution has declined significantly since the late 1970s (Mahony et al. 2013). Numerous threatening processes have been suggested in the National Recovery Plan for L. raniformis (Clemann and Gillespie 2012). These include loss and degradation of habitat, barriers to movement, disease, predation and pollution.

12 ! Chapter 1

Litoria raniformis is a large, semi-aquatic species that commonly occurs in close proximity to still or slow-flowing creeks and permanent waterbodies, including farm dams, irrigation channels and disused quarries (Clemann and Gillespie 2010; see Fig. 1.1). The species is primarily active during the spring and summer months, when breeding occurs. It appears to prefer water bodies with diverse aquatic vegetation and minimal shading from fringing vegetation (Heard et al. 2008, Heard et al. 2014). Throughout the inland sections of its range, L. raniformis displays ‘patchy population dynamics’ closely linked to flooding. The distribution of individuals within these systems changes over time, depending on the distribution of water (Wassens et al. 2010). Populations from cooler regions exhibit ‘classical metapopulation dynamics’, whereby discrete populations, connected by limited migration, display frequent extinction and recolonisation (Heard et al. 2012a, 2013, 2015).

(a) (b)

(c) (d) Figure 1.1. The study species, Litoria raniformis: (a) & (b) adults exhibiting typical colour variations, (c) & (d) L. raniformis habitat north of Melbourne.

In Victoria, L. raniformis is listed as endangered under the Department of Environment, Land, Water and Planning’s Advisory List of Threatened Vertebrate Fauna in Victoria – 2013 (DSE 2013). In many cases, remaining populations have become geographically

! 13 Chapter 1

isolated from one another and are threatened by habitat loss, degradation and fragmentation. Urban expansion is a key threat in the city of Melbourne, the state capital, where remnant populations occur in the Whittlesea-Hume, Casey-Cardinia, Melton-Caroline Springs and Wyndam regions (Fig. 1.2). Recent expansion of urban growth boundaries will increase the extent of Melbourne by an additional ~40,000 hectares (DPCD 2009). Remnant populations of L. raniformis occur throughout the new urban growth areas (Figure 1.2 and further described in Chapter 3) and the species is known to be sensitive to habitat fragmentation caused by urbanisation (Heard et al. 2010).

Figure 1.2. The geographic distribution of Litoria raniformis around Melbourne, Australia. Black circles indicate detection records since 2000. The light blue area is Port Phillip Bay, dark pink indicates the current Melbourne metropolitan area and light pink indicates the future urban growth area.

THESIS OUTLINE

The primary objective of this thesis is to investigate the role of urbanisation in shaping the genetic structure and diversity of Litoria raniformis in the greater Melbourne region, and to incorporate this information into conservation planning for the species. In

14 ! Chapter 1

Chapter 2, I begin with an important methodological and ethical consideration for this study and others on conservation genetics. I use a multi-criteria decision framework to assess four tissue sampling techniques for amphibians against five decision criteria: DNA quantity, genetic accuracy, potential future use of samples, species conservation and the welfare of individual animals. Structured decision making provides a useful framework for making difficult decisions and is a transparent, repeatable and efficient approach, which increases understanding of the decisions to be made (Moore and Runge 2012). I show that by using a Simple Multi-Attribute Rating Technique, it is possible to quantitatively assess tradeoffs when choosing the optimal genetic sampling technique for a given species. In Chapter 3, I examine broad-scale patterns of the genetic structure of populations of L. raniformis around Melbourne using mitochondrial DNA and microsatellites, and discuss the implications of these results for the conservation of the species in urbanising landscapes across the city.

In Chapter 4, I use microsatellite data and landscape genetic approaches to investigate the determinants of spatial population genetic structure and gene flow for L. raniformis in Melbourne’s northern growth corridor. In this chapter I develop a linear regression model of gene flow (a genetic measure of population connectivity), which can be an important predictive tool for the management of L. raniformis around Melbourne. I used this statistical approach, as it delivers a model of gene flow with direct application to the management of threatened species in urban landscapes. Its power lies in the fact that the model can be used for predictive purposes, allowing the user to investgate a range of management scenarios. In Chapter 5, I incorporate this model of gene flow into an existing stochastic patch occupancy model for L. raniformis, and use this modelling framework to assess both metapopulation viability and potential management options for the species in urbanising landscapes. Metapopulation models are important tools for predicting the probability of species’ persistence in fragmented landscapes. Integrating a model of genetic connectivity with a Bayesian stochastic patch occupancy model (SPOM) is a novel approach, which can improve the accuracy of the SPOM and help provide more realistic estimates of metapopulation viability, as it incorporates barriers to dispersal. Specific case studies involve habitat loss and creation, manipulating both the number of newly created wetlands and the spatial pattern of these wetlands. In Chapter 6, I conclude with an integrated discussion of the findings of each chapter, synthesised into the framework of conservation genetics and its implications for

! 15 Chapter 1

population viability and persistence in altered landscapes. I also discuss future research directions and applications of this work to other study systems.

!

16 ! Chapter 2

CHAPTER 2 Comparing genetic sampling techniques for amphibians with a multi-criteria decision framework

! 17 Chapter 2

ABSTRACT

Conservation genetics provides important information for the management of threatened species. However, some genetic sampling methods may harm animals, leading to a potential conflict between the welfare of the individual study organisms and the welfare of the species as a whole. For example, toe-clipping amphibians for tissue samples remains contentious and many ethics committees no longer allow this technique due to the pain and potential suffering caused to individual animals. Yet, alternatives may also cause pain and suffering and deliver poorer genetic samples. In this case study, I use a multi-criteria decision framework to assess four tissue sampling techniques for amphibians against five decision criteria; DNA quantity, genetic accuracy, potential future use of samples, species conservation and the welfare of individual animals. Under this framework, clipping of toe-webbing was the most viable alternative genetic sampling technique to toe-clipping in anurans, performing well under each decision criterion. While buccal swabbing and skin swabbing performed well on species welfare and conservation criteria, they performed relatively poorly on the sample quality or reuse criteria. The approach demonstrated here is broadly applicable to future conservation genetic studies of amphibians and other animals.

18 ! Chapter 2

INTRODUCTION

Threatened species often have small population sizes and limited distributions; therefore, studies of threatened species usually use non-lethal methods to collect genetic samples (reviewed in Appendix 2.1). However, there is a conflict within many sampling methods between prioritising the welfare of the individual study organisms and the value of the underlying research for the conservation of populations or species (Minteer and Collins 2005). The trade-off between individual welfare and the importance of the proposed research needs careful consideration during the initial planning phase of conservation genetics projects, particularly when determining which sampling technique/s are to be used. However, the process of making these decisions has never been formalised. Thus, a framework that formalises these decision-making processes would be of significant value to researchers, government authorities and ethics committees.

Amphibians are declining rapidly, with one third of species threatened globally (Stuart et al. 2004). Concern for declining frog numbers, combined with significant advances in genetic technology over the last two decades, have resulted in a burgeoning number of conservation genetics studies on these animals. However, frogs present a range of sampling difficulties, which makes a decision framework of particular importance. Since it was first described by Bogert (1947), toe-clipping has been common practice for marking and obtaining tissue samples from amphibians. However, research has found that toe-clipping can reduce the survival of some amphibian species and/or change their behaviour (McCarthy and Parris 2004, Waddle et al. 2008).

Toe-clipping has consequently become a contentious practice amongst researchers (May 2004, Funk et al. 2005, Phillott et al. 2007, Parris and McCarthy 2008, Correa 2013) and many ethics committees now refuse to allow its use for genetic sampling (Grafe et al. 2011). Hence, there is a clear need to find reliable alternative sampling techniques. Less-invasive sampling methods have been trialled for amphibians, including the use of buccal swabs, epidermal swabs, and the collection of moulted epidermal cells and intestinal cells from faeces (see Appendix 2.1). However, these techniques vary in their DNA yield and accuracy, both within and between amphibian species (Taberlet et al. 1999).

! 19 Chapter 2

Decision theory is a framework in which people aim to achieve particular goals, while acknowledging the uncertainty and subjectivity involved in the decision process (Clemen 1996). A set of alternative actions for reaching the objectives is devised, which can be ranked and scored against the measurable decision criteria. Structured decision making provides a useful framework for making difficult decisions and is a transparent, repeatable and efficient approach, which increases understanding of the decisions to be made (Moore and Runge 2012). Multi-criteria decision analysis recognises that objectives, criteria and weights are often subjective, however incorporating weighting and sensitivity analyses can alleviate some of the influence of arbitrary choices (Burgman 2005). Decision theory is receiving increasing attention for its utility in making conservation decisions (Driscoll 2010). However it has been largely overlooked in the field of conservation genetics to date, despite its potential to solve conflicts in genetic sampling designs, such as in the case of amphibians.

I used multi-criteria decision analysis to facilitate selection of tissue sampling techniques for conservation genetic studies of amphibians. Here, I demonstrate the utility of a decision framework for choosing among tissue sampling techniques, using the endangered Growling Grass Frog (Litoria raniformis) as a case study, and including five decision criteria; DNA quantity, genetic accuracy, potential future use of samples, the conservation of the species as a whole and the welfare of individual animals. This framework is broadly applicable across vertebrates, and provides a valuable tool for researchers and ethics committees developing and assessing the design of conservation genetic studies.

METHODS

Study species and tissue sampling

Litoria raniformis is a large, semi-aquatic frog that naturally occurs across south-eastern Australia (Pyke 2002). The species is listed as endangered, having undergone significant range reductions over the past two decades (IUCN 2013). Around Melbourne, remnant populations have become geographically isolated and are threatened by habitat loss, fragmentation and degradation due to urban expansion (Heard et al. 2013).

20 ! Chapter 2

I captured 30 L. raniformis during spotlight surveys conducted in the Austral summer of 2010/2011, at slow-flowing sections of major streams and lentic water-bodies across Melbourne. Samples were collected as part of a wider study on the regional genetic structure of the species (Keely et al. 2015). Four sampling methods were used to collect tissues from each frog: toe-clips, toe-web clips, buccal swabs (following Poschadel and Möller 2004) and epidermal swabs, the majority of which have been frequently used in conservation genetics studies of frogs (Appendix 2.1). A number of these sampling methods are used broadly in vertebrates (Appendix 2.1).

For toe-clips, I removed the toe-pad on the second toe of the left hind limb using surgical scissors. Clipping toe-webbing entailed using surgical scissors to remove a triangular section of toe-webbing (~ 2 mm at the base) between the second and third toes closest to the body on the left hind limb. Sampling toe-webbing has primarily been used in studies of the amphibian chytrid fungus, Batrachochytrium dendrobatidis (Edge et al. 2013). For buccal swabbing, I used a small reptile-sexing probe to gently open the mouth, and then swabbed the buccal cavity using a Tubed Sterile Dryswab tip (Medical Wire and Equipment, UK). Tubed Sterile Dryswab tips were also used to swab the epidermis, spinning the tip 360° while moving it along the skin surface for approximately 30 seconds, including the dorsal and ventral surfaces of the abdomen, thighs, hands and feet.

Toe and web tissue samples were stored in 95% ethanol and kept at -18°C for approximately two weeks, then -80°C for up to 3.5 months. Swab tips were stored dry in sealed cryotubes held at ambient temperature for up to 8 hours, then stored frozen at - 18°C for approximately two weeks, then -80°C for up to 9 months prior to extraction. Latex gloves were worn when taking samples from frogs and replaced between individuals. Tissue sampling equipment was sterilised in 70% ethanol between individuals.

DNA extraction

I extracted Genomic DNA from toe and web tissue samples using a DNeasy Blood & Tissue Kit (QIAGEN, South Korea), following the manufacturer’s animal tissue protocol. I used a QIAamp DNA Investigator Kit (QIAGEN) to extract genomic DNA

! 21 Chapter 2

from buccal and epidermal swabs, following the manufacturer’s protocol for isolation of total DNA from surface and buccal swabs, applying 20 !l Buffer ATE, then 80 !l separately into two microcentrifuge tubes, which were incubated at room temperature for 5 minutes before the final centrifuge. A step was added using a QIAshredder (QIAGEN) to harvest the remaining lysate. The concentration and therefore quantity of double stranded DNA extracted from all samples was measured using a Qubit 2.0 Fluorometer (Invitrogen, USA), using a Qubit dsDNA HS Assay Kit.

DNA amplification, sequencing and genotyping

For mtDNA (mitochondrial) and nuDNA (nuclear) sequencing, I performed PCR amplification in a 20 µl total volume with 2 µl of DNA template, 10 µl GoTaq Hot Start

Polymerase (Promega, USA) (25 mM MgCl2, GoTaq Hot Start Polymerase, 5X Colorless GoTaq Flexi Buffer, 5X Green GoTaq Flexi Buffer), 1 µM forward primer, 1

µM reverse primer and 6 µl dH2O. I amplified a partial sequence of 495 bp of the mitochondrial Cytochrome oxidase I (COI) gene, using the primers Cox (5’- TGATTCTTTGGGCATCCTGAAG-3’) and Coy (5’- GGGGTAGTCAGAATAGCGTCG-3’) (Schneider et al. 1998). COI amplification involved 95°C 2 min, 37 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5

min (adapted from Hale et al. 2011). 1:10 DNA diluted with dH2O was used for all sample types. For the nuclear gene Rhodopsin (Rhod), I amplified a 321 bp sequence, using the primers Rhod 1A (5’- ACCATGAACGGAACAGAAGGYCC-3’) and Rhod 1C (5’- CCAAGGGTAGCGAAGAARCCTTC-3’) (Bossuyt and Milinkovitch 2000). Rhod amplification involved 95°C 2 min, 40 cycles of 95°C 30 s, 60°C 45 s, 72°C 45 s,

followed by 72°C 5 min. 1:10 DNA diluted with dH2O was used for all sample types, except epidermal swab DNA, which was amplified neat. Negative controls were routinely included and checked for contamination. I purified the amplified PCR products using Exo Sap-IT (USB, USA) following manufacturer’s instructions, then sent the purified products to Macrogen (South Korea) for sequencing using a 3730XL DNA sequencer (ABI). I BLAST searched all sequences to confirm identity and aligned sequences using Geneious Pro v. 5.6 (Biomatters).

I used seven polymorphic microsatellite markers for genotyping (Lr1, Lr2, Lr5, Lr6, Lr7, Lr8 and Lr9; Hale et al. 2011), which included three sets of two markers

22 ! Chapter 2 multiplexed during PCR amplification. Four markers incorporated a GTTTCTT ‘pigtail’ added to the 5’ end of the reverse primer to reduce stutter (Lr2, Lr6, Lr8 and Lr9). See Hale et al. (2011) for methods for fluorescently labeling fragments for all loci. I performed PCR amplification in a 10 µl total volume with 1 µl of DNA template, 5 µl

GoTaq Hot Start Polymerase (Promega) (25 mM MgCl2, GoTaq Hot Start Polymerase, 5X Colorless GoTaq Flexi Buffer, 5X Green GoTaq Flexi Buffer), 0.5 µM reverse primer, 0.15 µM forward primer, 0.25 µM fluorescently labeled 454A primer and 3.1 µl dH2O. Amplification involved 95°C 2 min, 42 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5 min (adapted from Hale et al. 2011). Fragment analysis of PCR products were carried out by Macrogen (South Korea) on an Applied Biosystems ABI3730XL DNA analyser using a LIZ-500 size standard. I completed scoring using the Microsatellite Plugin v 1.2 in Geneious Pro v 5.6 (Biomatters), and all samples were screened manually for accuracy.

Decision framework

The decision framework takes the form of a flow chart (Fig. 2.1) that defines the sampling regime that will best achieve the objectives of the researcher, dependent on five criteria (see below). As my case study involved choosing a genetic sampling regime for a threatened species, the objective was to inform conservation management decisions at a species level. I then broke this objective down into separate decision criteria. I considered five decision criteria for assessing the appropriateness of the different genetic sampling techniques for L. raniformis: DNA quantity, genetic accuracy, the potential future use of samples, species welfare and individual welfare. I used DNA extraction and quantification methods that are widely used across many vertebrate taxa, using commercially available kits, ensuring the broad utility of results. I used two separate criteria for assessing welfare at the species and individual levels, following a similar protocol to Parris et al. (2010), detailed below.

! 23 Chapter 2

Figure 2.1. Flow chart demonstrating a decision framework for designing a conservation genetics study.

Criterion 1 DNA quantity

I considered the amount of DNA extracted from each sample type as the DNA quantity. The concentration and therefore quantity of double stranded DNA extracted from all samples was measured using a Qubit 2.0 Fluorometer (Invitrogen).

24 ! Chapter 2

Criterion 2 Genetic accuracy

For genetic accuracy, I considered the amplification success of mitochondrial and nuclear sequences, in addition to the microsatellite genotyping error rate. These are the three techniques most commonly used for conservation genetics studies. Cells contain more mtDNA than nuDNA, therefore mtDNA is generally easier to amplify. There are many errors that can occur during the amplification and microsatellite genotyping of small quantities of DNA, as they are especially sensitive to contamination and errors (Broquet and Petit 2004). Toe-clip samples are considered reliable sources of genetic material, and have been used to obtain DNA from many anuran species, including L. raniformis (Hale et al. 2013). I therefore used DNA extracted from toe-clip samples as my reference.

Criterion 3 Potential future use of samples

I considered the potential for long-term storage of samples for future genetic research. There has been substantial progress in the field of conservation genetics since its foundation in the late 1970s and we can expect incremental advances in all areas of conservation genetics (Frankham 2010), particularly with the increasing use of genomic techniques. Therefore, new genetic techniques may emerge which prompt further analyses of existing samples. I assessed whether multiple DNA extractions were possible from each sample type, and whether the sample could also be used for detecting infection with the amphibian chytrid fungus B. dendrobatidis, based on reports of these applications in the literature.

Criterion 4 Species welfare

Species welfare was defined as the impact of sampling at a population level. The sampling technique with the lowest mortality rate as a result of DNA sampling was considered to have the least impact on the species. There is uncertainty in quantifying mortality rates from genetic sampling techniques, however I calculated values for each technique incorporating the mortality rates allocated by Parris et al. (2010) for toe- clipping and buccal swabbing, then considered the potential mortality rates from toe- web and skin swab sampling relative to these. Using previous research on the

25 Chapter 2

relationship between return rate and the number of toes clipped in mark-recapture studies of anurans (McCarthy and Parris 2004), Parris et al. (2010) concluded that a mortality rate of 2% with one toe clipped was feasible in conservation genetic studies of these animals. Additionally, negative physiological effects have been reported for toe- clipping, including infection, necrosis and reduced weight gain (Golay and Durrer 1994, Davis and Ovaska 2001). Parris et al. (2010) suggested that a mortality rate of 1% of adults was also feasible following buccal swabbing (although the effect could be negligible or larger).

Criterion 5 Individual welfare

Individual welfare was defined as the level of suffering experienced during sample collection, incorporating both pain and stress. There is limited information on the extent of pain caused by genetic sampling procedures, however, frogs probably have rapid and local pain perception (Koeller 2009). Pain responses in amphibians can include vocalisation and rapid startle responses (Machin 1999). Pain was ranked for each technique using two sources of information: previous research on pain responses and my observations of individuals during this study. Urinary corticosterone has been used to objectively measure the degree of stress experienced by amphibians (Narayan et al. 2011). The ranking of individual welfare for each technique was guided by urinary corticosterone concentrations reported for wild cane toads (Rhinella marina) six hours after initial capture for individuals that were only captured and handled versus those captured and toe-clipped (Narayan et al. 2011).

SMART Analysis

To determine the optimal sampling technique, I quantitatively established the tradeoffs between each decision criterion using the Simple Multi-Attribute Rating Technique (SMART; Edwards 1977, Goodwin and Wright 2004). I then applied a weight between 0 and 100 to each decision criterion to reflect its relative importance, with 100 given to the criteria considered the most important. I normalised the scores within each decision criterion on a 0-1 scale, where 1 is the sampling technique considered to be most

favorable and 0 the least. Normalised scores (Nij) were calculated for technique i

26 ! Chapter 2 relative to the decision criterion j using the following formulae adapted from Converse et al. (2013), where Sij is the original score:

Eq. 1

(for criteria where desired original scores were high i.e., DNA amount) or

Eq. 2

(for criteria where desired original scores were low i.e., impact on species welfare)

I then used the following formula to determine the final score for each alternative technique (Fi), as the product of the weight for each decision criteria (Wj) and the normalised score:

Eq. 3 I conducted an analysis to determine the sensitivity of the final decision to uncertainty in the data and rankings by substituting uncertain welfare values for genetic sampling techniques with a range of values, as well as substituting the weights given to each decision criterion, to establish how sensitive the optimal decision was to rankings of each technique.

! 27 Chapter 2

RESULTS

Decision Framework

Criterion 1 DNA quantity

Toe-clipping produced the largest quantity of double stranded DNA (887.13 ± 414.51 ng, mean ± SD), followed by toe-web samples (456.11 ± 259.63 ng, mean ± SD) and buccal swabs (31.41 ± 24.56 ng, mean ± SD). Skin swab samples provided the lowest quantity of DNA (10.67 ± 4.35 ng, mean ± SD). Hence, toe-clipping ranked highest under the DNA quantity criterion.

Criterion 2 Genetic accuracy

Toe-clipping, toe-web clipping and buccal swabbing all ranked equally under this criterion. In each case, 100% of samples accurately amplified a partial sequence of 495 bp of the mitochondrial COI gene and a 321 bp sequence of the nuclear gene Rhod, and accurately genotyped the 7 microsatellite loci for each of the 30 individuals. Skin swab samples accurately amplified a partial sequence of 495 bp of the mitochondrial COI gene for each of the 30 individuals, and accurately amplified a 321 bp sequence of the nuclear gene Rhod for 28 of the 30 individuals. The two unsuccessful skin swab samples failed to amplify after multiple attempts. Further, skin swab samples accurately genotyped the 7 microsatellite loci in only 2 of the 30 individuals (7%). Incorrect readings usually occurred across multiple microsatellite loci. Incorrect genotypes were due to allelic dropouts, PCR failing to amplify and incorrect amplification of some alleles. The accuracy of skin swab samples was given a cumulative score of 67% (the average of 100% for mtDNA sequencing, 93% for nuDNA sequencing and 7% for microsatellite genotyping).

Criterion 3 Potential future use of samples

Each sample type under this criterion was scored as follows (see Table 2.1 for justifications). Both toe-clip and toe-web samples were given a score of 1 for potential

28 ! Chapter 2 future use of samples, as they can be stored long term in museum collections and can be used for detecting B. dendrobatidis infection. Skin swabs cannot be stored long term in museum collections but can be used for detecting B. dendrobatidis infection, so were given a score of 0.5. Buccal swabs received the lowest score of 0, as they cannot be stored long term in museum collections and are not useful for detecting B. dendrobatidis infection, as B. dendrobatidis targets the keratinised layer of the skin and keratinised mouthparts are shed at metamorphosis (Marantelli et al. 2004).

Table 2.1. Considerations for potential future use of each sample type.

Sample type Comparative longevity of samples B. dendrobatidis detection Toe-clip Multiple DNA extractions are possible, depending on the Yes sample size. I was able to collect enough tissue from L. raniformis for at least two DNA extractions per sample. These samples can be stored long term in 95% ethanol at - 80°C and can be added to Museum collections for future use by other researchers. Toe-web Multiple DNA extractions are possible, depending on the Yes sample size. I was able to collect enough tissue from L. raniformis for at least two DNA extractions per sample. These samples can be stored long term in 95% ethanol at - 80°C and can be added to Museum collections for future use by other researchers. Buccal swab Involved the use of the entire swab tip, and therefore multiple No extractions were not possible. Future storage would only apply to the extracted DNA and therefore have a shortened longevity. Skin swab Involved the use of the entire swab tip, and therefore multiple Yes extractions were not possible. Future storage would only apply to the extracted DNA and therefore have a shortened longevity.

Criterion 4 Species welfare

I considered the potential mortality rates from toe-web and skin swab sampling relative to 2% mortality from toe-clipping and 1% mortality from buccal swabbing for Bufo boreas, Rana aurora and Rana temporaria (Parris et al. 2010). I concluded that a mortality rate of 1.5% for toe-webbing was feasible, given that it involved handling and an incision, although not through bone and through less tissue than toe-clipping. I

! 29 Chapter 2

assigned skin swabbing a mortality rate of 0.5% due to handling stress. This sampling method did not involve opening the frog’s mouth, as for buccal swabbing. However, there is anecdotal evidence that handling alone can occasionally cause the death of a frog.

Criterion 5 Individual welfare

Each sample type under this criterion was scored as follows (see Table 2.2 for justifications). I considered toe-clipping to cause the highest level of individual suffering and gave it a ranking of 100%. Toe-webbing was considered the next level of suffering, due to pain and stress associated with cutting through skin. I gave it a ranking of 75%, lower than toe-clipping and higher than skin swabbing. I gave buccal swabbing a ranking of 50%, as I considered it to cause less suffering than both toe-clipping and toe-webbing, and pain levels were assumed to be lower, because no incision was made. I considered skin swabbing the sampling method that would produce the lowest relative level of individual suffering, however, this technique is still known to produce a stress response due to capture and handling in toads (Narayan et al. 2011), and was given a ranking of 25%.

SMART Analysis

Toe-web sampling received the highest total score, once scores had been normalised, both before and after weighting, when considering all five decision criteria (Table 2.3). This was followed by toe-clipping, which received the second highest score, again both before and after weighting. Species welfare was the criterion to which I gave the highest weighting, given its potential to impact the species at a population level. I also gave individual welfare, genetic accuracy and DNA amount high weightings, as these were viewed as important considerations when designing a study. The potential for future use of samples was ranked lower, as it did not directly impact the quality of the research or welfare at either the species or individual levels.

30 Chapter 2

Table 2.2. Considerations for individual welfare for each sample type. Sampling Pain Stress Percentage Technique given Toe-clipping Pain associated with cutting Corticosterone concentration 100 through tissue and bone. of 333 pg/µg measured after Pain responses often capture and toe-clipping observed for L. raniformis (Narayan et al. 2011). during this study. Toe-webbing Pain associated with cutting No known measure from 75 through skin. Pain responses literature search. Assumed sometimes observed for higher than capture and L. raniformis during this handling alone and lower study. than capture and toe- clipping. Buccal swabbing Bleeding has been reported No known measure from 50 from buccal swabbing literature search. Assumed (Pidancier et al. 2003, higher than capture and Poschadel and Möller 2004), handling alone and lower however no bleeding was than capture and toe- observed during the present clipping. study. No incision made. Skin swabbing No incision made. Corticosterone concentration 25 of 146 pg/µg measured after capture and handling (Narayan et al. 2011).

Because individual welfare values for toe-web and buccal swab techniques were uncertain, I additionally performed a sensitivity analysis, using a range of normalised scores from 0 to 1. If toe-web clipping were given a normalised score of 0 for individual welfare (equivalent to toe-clipping), then toe-clipping would become the most favorable genetic sampling technique overall. If buccal swabbing were simultaneously given a normalised score of 1 for individual welfare (equivalent to skin swabbing), then the final rankings would change again and buccal swabbing would become the most favorable genetic sampling technique overall. I also performed a sensitivity analysis on the weights given to each decision criteria. I altered the weight of each criteria independently, trialling weights from 100 to 70 (for potential future use, a minimum weight of 40 was instead trialled). For each altered weight, toe-webbing remained the technique with the highest final ranking; therefore the outcome of the analysis did not change.

! 31 Chapter 2

Table 2.3. Scores for each decision criterion when evaluating sampling techniques for the Growling Grass Frog, Litoria raniformis: DNA quantity, genetic accuracy, potential future use of samples, species welfare and individual welfare. (a) Raw scores; (b) Normalized, weighted scores. Using SMART, scores were normalized for each decision criteria on a 0-1 scale, where 1 is the sampling technique considered to be most favorable and 0 the least. A weight was applied to each decision criteria reflecting its relative importance, between 0 and 100, with 100 given to the criteria considered the most important. (a) Actions Criteria Toe-clip Toe-web Buccal swab Skin swab DNA amount 887 456 31 11 Genetic accuracy 100 100 100 67 Potential future use 1 1 0 0.5 Species welfare 0.02 0.015 0.01 0.005 Individual welfare 1 0.75 0.5 0.25

(b) Actions Criteria Toe-clip Toe-web Buccal swab Skin swab Weights DNA amount 1 0.508 0.023 0 90 Genetic accuracy 1 1 1 0 95 Potential future use 1 1 0 0.5 60 Species welfare 0 0.333 0.667 1 100 Individual welfare 0 0.333 0.667 1 95 Total scores 3 3.174 2.357 2.5 Weighted scores 245 265.655 227.135 225

DISCUSSION

I used a multi-criteria decision framework to assess four tissue sampling techniques for amphibians against five decision criteria, providing a case study of the development and use of multi-criteria decision frameworks in conservation genetics. Under this framework, I found clipping of toe-webbing was the most viable alternative genetic sampling technique to toe-clipping in anurans, performing well under each decision criterion. The sensitivity analysis revealed that the outcomes of the decision analysis

32 Chapter 2 were robust to variations in the weight given to each decision criteria, strengthening this result. Although toe-clipping has been the most widely used technique for tissue sampling amphibians, it is known to influence the behaviour and/or survival of frogs (McCarthy and Parris 2004, Waddle et al. 2008), which has raised welfare concerns. In the decision framework, toe-clipping scored higher for DNA quantity but toe-webbing was more favorable for welfare at the species and individual levels. While altering the scores given to toe-webbing and buccal swabbing for individual welfare (to rank the two techniques equal to toe-clipping and skin swabbing respectively) changed the final outcome of the decision analysis, an argument in favor of equal ranking is difficult to justify. Toe-webbing was not ranked equivalent to toe-clipping for individual welfare because it does not involve cutting through bone. However, it must be acknowledged that a drawback of toe-web samples is that they are limited to species with adequate toe- webbing, which is absent in numerous frogs. Thus, if toe-webbing is not present, toe- clipping or additional alternative methods could be pursued.

Of the other methods I trialled, buccal swabbing has been successfully implemented across all major vertebrate groups, including mammals, birds, fish, reptiles and amphibians (see Appendix 2.1). I found buccal swabs to be a viable genetic sampling technique in anurans, with the technique performing favorably under most welfare and genetic accuracy criteria. However, buccal swabbing places limits on the genetic analyses that can be performed, due to the small quantity of DNA produced from the samples. Moreover, buccal swabbing is limited to species with a mouth cavity large enough to fit a swab tip and may prove detrimental to individuals of small species (Pidancier et al. 2003).

I found the reliability of skin swabbing to vary with the genetic markers used, with microsatellite genotyping having a particularly low accuracy score. A recent study on alpine newts (Ichthyosaura alpestris) and European tree frogs (Hyla arborea) found skin swabbing to be a reliable and efficient genetic sampling technique for use with microsatellites (Prunier et al. 2012). However, the sample size of this study was small, and the results contrast with those of other recent studies. For example, Muller et al. (2013) found skin swab samples contained high levels of foreign DNA contamination. Variation in the skin secretions of anurans may impact DNA quality and PCR inhibitors, such as humic acids, may be present on the skin (Garland et al. 2009).

! 33 Chapter 2

Consequently, skin swabs need specific testing before being incorporated into conservation genetic research on anurans.

An area of ongoing uncertainty, particularly for buccal and skin swabs, is storage conditions and DNA extraction techniques. I stored all swab samples dry at -80°C. Other studies have reported similar swab storage techniques, along with storage in 70% ethanol, lysis buffer and TNE buffer (Goldberg et al. 2003, Poschadel and Möller 2004, Tanaka-Ueno et al. 2006, Prunier et al. 2012) with varying degrees of success. The extraction kit and protocols used here were recommended by the manufacturer specifically for extracting genomic DNA from buccal and epidermal swabs. Prunier et al. (2012) reported the use of Blood and Tissue Kits by the same manufacturer for swab extractions and obtained much higher concentrations of DNA. Future research into different storage and extraction methods is required to establish protocols that maximise DNA yield in general.

Across vertebrate groups, the efficacy of different genetic sampling techniques has been the focus of extensive research in population and conservation genetics over the last two decades, and there is now considerable published information available on this topic for use in decision frameworks such as that presented here (Appendix 2.1). However, this study highlights that further research into stress and pain is vital for informing the design of conservation genetic studies across many vertebrate species. There has been limited investigation into stress levels associated with toe-clipping in herpetofauna. Research on the cane toad (Rhinella marina) found stress levels after toe-clipping were more than twice those after capture and handling alone (Narayan et al. 2011). In contrast, another recent study found no significant differences in corticosterone levels of toads (R. marina) that were handled versus those that were toe-clipped (Fisher et al. 2013). However, the latter study used laboratory toads and measured plasma corticosterone levels, while the former measured urinary corticosterone levels in wild toads. To further refine estimates of individual welfare rankings for decision frameworks such as that presented here, future research should focus on stress levels from toe-clipping (both inter- and intra-specific) compared to stress associated with alternative genetic sampling techniques.

34 ! Chapter 2

The objectives, framework and decision criteria for designing genetic sampling regimes outlined here can be directly applied across many vertebrates and different developmental stages, or may be adapted to match idiosyncrasies of research on particular animals. Additionally, alternative objectives not included in this decision framework could be integrated depending on the conservation objective of the study. These include conservation or detection of an ESU, identifying species or populations at risk due to low genetic diversity, resolving fragmented population structures, understanding species biology, resolving taxonomic uncertainties and the genetic management of captive populations (Frankham et al. 2002). Additional decision criteria for other taxa could include financial costs or sampling effort. For example, genetic sampling techniques for large mammals such as bears can either be noninvasive, including the collection of hair, or invasive, including the collection of tissue samples (Kendall et al. 2009). Noninvasive genetic sampling techniques are likely to require additional laboratory costs, but less sampling effort than blood or tissue collection, as individuals don’t need to be caught, handled, or even observed (Taberlet et al. 1999).

Similarly, the weight given to each decision criterion could be altered to reflect the studies’ objectives. For instance, a study may involve the incidental capture of a non- target threatened species as part of a broad biological survey. In this case, the potential for storage and future use of this sample may be weighted more highly. In Brazil, for example, the scarcity of basic scientific knowledge on taxonomy and systematics of the country’s small mammals is considered a major threat (Costa et al. 2005). Studies on these species may give more weight to decision criteria such as DNA quantity and the potential for storage and future use of samples. Other studies may wish to include consideration of potential long-term storage of samples for future genomic work, including both RNA and DNA preservation. Indeed, the application of genomics in conservation and management is a growing field. While the generation of genomic data, their analyses and interpretation remain challenging, genomic approaches have been shown to provide more statistical power than microsatellites and cost less for genotyping (Shafer et al. 2015, Garner et al. 2016). Additionally, weighting schemes of decision criteria can be modified, based on sensitivity analyses and uncertainty, or non- independence. It is important to note that I have simplified the process for this case study. Decision theory objectives, values and weights are typically determined by stakeholders, experts and decision makers and a consensus is often not easily reached.

! 35 Chapter 2

In this case study I have demonstrated the use of a decision framework to guide the design of conservation genetic studies. This technique provides a formal means of weighing the efficacy of these genetic sampling techniques (using criteria such as DNA quantity, genetic accuracy and its potential future use) against their impacts at both the species and individual level. Simply by considering alternate genetic sampling regimes, objectives and decision criteria, this approach is broadly applicable across many vertebrate groups, and may be used by researchers, government institutions and ethics committees to determine the most appropriate genetic sampling regime for a given study.

36 ! Chapter 2

APPENDICES

Appendix 2.1. Summary of genetic sampling techniques. Taxon DNA Source Citation Amphibians buccal swabsa, b, c (Goldberg et al. 2003, Pidancier et al. 2003, Poschadel and Möller 2004, Broquet et al. 2007) epidermal swabsa, b (Davis et al. 2002, Prunier et al. 2012) moulted epidermal cellsa (Tanaka-Ueno et al. 2006) intestinal cells from faecesa (Tanaka-Ueno et al. 2006) toe-clipsa, b (Smith et al. 2013) Reptiles scale-clipsa (Sumner et al. 2010) shed skinc (Bricker et al. 1996) tail-tipsa (Stuart-Fox et al. 2001) blooda (Blumenthal et al. 2009) buccal swabsa (Poschadel and Möller 2004) tissue biopsiesa (Formia et al. 2007) carapace scrapingsb (Stephens et al. 2010) cloacal swabsa, b (Miller 2006), toe-clipsb (Noble et al. 2013) faecesa (Jones et al. 2008) egg and foetal tissuea (Jones et al. 2008) carcassesa (Jones et al. 2008) Birds bloodb (Delaney et al. 2010) faecesa, d (Idaghdour et al. 2003) urinee (Nota and Takenaka 1999) feathersb (Segelbacher 2002) egg shellsb (Strausberger and Ashley 2001) egg shell swabsb (Martin-Galvez et al. 2011) embryo samplingc (Lecomte et al. 2006) buccal swabse (Wellbrock et al. 2012) Mammals ear-clipsa, b (Smith and Hughes 2008) hairb, e (Sloane et al. 2000)

! 37 Chapter 2

faecesb, e (Flagstad et al. 2012) buccal swabsa (Tobe and Linacre 2008) liver or kidneya, d (Hafner et al. 2004) urinea, b (Valiere and Taberlet 2000) sloughed skinb (Pierszalowski et al. 2013) regurgitatesa, d (Taberlet and Fumagalli 1996) food wadgesf (Sugiyama et al. 1993) salivab, e, g (Williams et al. 2003) Invertebrates wing-clipsh (Keyghobadi et al. 2009) single leg removalb (Hadrys et al. 2005) partial leg removalb (Holehouse et al. 2003) shed skin and frassa (Feinstein 2004) Fish buccal swabb (Smalley and Campanella 2005) musclea (Hilsdorf et al. 1999) blood (Cummings and Thorgaard 1994) sperm (Cummings and Thorgaard 1994) caudal or anal finsa, i, j (Wasko et al. 2003) scalesb, g (Lucentini et al. 2006) a mtDNA sequencing, b microsatellite genotyping, c DNA extraction, d nuDNA sequencing, e sex determination, f dinucleotide repeat polymorphisms, g RFLPs, h AFLPs, i RAPD, j ribosomal DNA sequencing

38 ! Chapter 3

CHAPTER 3 Genetic structure and diversity of the endangered Growling Grass Frog in a rapidly urbanising region

Manuscript published as: Keely, C.C., Hale, J.M., Heard, G.W., Parris, K.M., Sumner, J.M. Hamer, A.J. and Melville, J. 2015. Genetic structure and diversity of the endangered growling grass frog in a rapidly urbanizing region. Royal Society Open Science 2(8): 1-13.

! 39 Chapter 3

ABSTRACT

Two pervasive and fundamental impacts of urbanisation are the loss and fragmentation of natural habitats. From a genetic perspective, these impacts manifest as reduced genetic diversity and ultimately reduced genetic viability. The Growling Grass Frog (Litoria raniformis) is listed as vulnerable to extinction in Australia, and endangered in the state of Victoria. Remaining populations of this species in and around the city of Melbourne are threatened by habitat loss, degradation and fragmentation due to urban expansion. I used mitochondrial DNA and microsatellites to study the genetic structure and diversity of L. raniformis across Melbourne’s urban fringe, and also screened four nuclear gene regions (POMC, RAG-1, Rhod, CRYBA1). The mitochondrial DNA and nuclear DNA sequences revealed low levels of genetic diversity throughout remnant populations of L. raniformis. However, one of the four regions studied, Cardinia, exhibited relatively high genetic diversity and several unique haplotypes, suggesting this region should be recognised as a separate Management Unit. I discuss the implications of these results for the conservation of L. raniformis in urbanising landscapes, particularly the potential risks and benefits of translocation, which remains a contentious management approach for this species.

40 ! Chapter 3

INTRODUCTION

Urbanisation represents a growing threat to the conservation of biodiversity. The impacts of urbanisation on biodiversity are numerous (McKinney 2002), but a pervasive and fundamental effect is the loss and fragmentation of habitat (Irwin and Bockstael 2007, Noel and Lapointe 2010). From a genetic perspective, these impacts manifest as reduced genetic diversity and ultimately reduced genetic viability, as a result of declines in local population size and changes in the rate of dispersal between habitat patches. Populations in small habitat fragments also suffer from edge-effects and increased exposure to predators (Andren and Angelstam 1988), as well as fine-scale changes in microclimate and other attributes that influence habitat quality (Saunders et al. 1991, Murcia 1995). The resulting small, fragmented populations are susceptible to genetic drift and inbreeding depression (Reed and Hobbs 2004), both of which can lead to the erosion of genetic diversity and genetic viability (Reed and Frankham 2003). These impacts are a pervasive problem in urban areas and as such, conservation genetics has become an important tool for the management of threatened species in urban landscapes.

Assessing the genetic origins of populations and effectively managing genetic diversity are central to their persistence in both the short- and long-term (Armstrong and Seddon 2008, Tracy et al. 2011). Genetic analyses of population structure, dispersal and gene flow can improve our understanding of how species respond to landscape change (Sunnucks and Taylor 2008), and hence provide insights into landscape modifications that may reduce isolation and the extinction risk of fragmented populations (Ricketts 2001). Genetic information is also vital for active manipulation of populations fragmented by urban development, such as population augmentation and translocation (Bos and Sites 2001). In many cases, initiatives such as these seek to restore the fitness of populations exhibiting symptoms of inbreeding depression, by introducing individuals (and hence, genes) from related populations (Hedrick and Kalinowski 2000). However, without prior knowledge of the genetic structure of a species, there is a risk of altering the genetic structure of populations (Moritz 2002), potentially causing problems such as outbreeding depression or hybridisation of divergent evolutionary lineages (Huff et al. 2011). Thus, detailed knowledge of the population genetic structure

! 41 Chapter 3

of a species is essential prior to the initiation of programs aimed at genetic rescue, such as population augmentation and translocation.

The proportion of the world’s human population living in urban areas grew rapidly during the 20th century, from an estimated 220 million to 2.8 billion (UNFPA 2007). Australia is expected to become the world's fastest growing industrialised nation over the next four decades (Population Reference Bureau 2009), and Melbourne – the capital city of the state of Victoria – is the fastest growing major city in Australia (ABS 2013). The Melbourne Metropolitan Area (MMA) currently covers ~7500 square kilometres and has a population of ~4.25 million people (State of Victoria 2013). Recent expansions of the urban growth boundary will increase the MMA by an additional 400 square kilometres over the next 20 years (DPCD 2009). Urban growth is expected to negatively impact the genetic diversity and viability of several threatened species in this region (DPCD 2009).

In this study, I investigated the conservation genetics of the endangered Growling Grass Frog (Litoria raniformis) in the MMA. This species has declined markedly over the last three decades, yet significant remnant populations persist in the low altitude, urban- fringe environments to the south-west (Wyndham), north-west (Melton), north (Hume- Whittlesea) and south-east (Cardinia) of Melbourne (Fig. 3.1, see below). I used mitochondrial DNA (mtDNA: COI & ND4), four nuclear gene regions (POMC, RAG- 1, Rhod, CRYBA1) and microsatellite genotyping to investigate the genetic structure and diversity of L. raniformis across Melbourne. I predict that: (1) genetic diversity of L. raniformis across Melbourne’s urban fringe is low due to past declines and more recent habitat loss and fragmentation from urbanisation; (2) that a genetic bottleneck may have occurred as a result of these processes, and; (3) that genetic structuring is present between remnant populations in Cardinia and the other three regions, due both to the fundamental geographical isolation of this region and the more recent loss of connecting populations (Fig. 3.1).

42 ! Chapter 3

METHODS

Study species

Litoria raniformis is a semi-aquatic species that inhabits still or slow-flowing sections of streams, as well as lentic wetlands (Clemann and Gillespie 2010). This species was abundant across much of south-eastern Australia (Pyke 2002); however, it has declined significantly since the late 1970s and the frog is now listed as endangered (Mahony 1999, IUCN 2013). Although numerous threatening processes drove this decline, the arrival of the chytrid fungus (Batrachochytrium dendrobatidis) in south-eastern Australia and resulting epidemics of chytridiomycosis were almost certainly a major contributor (Mahony et al. 2013). Populations at higher altitude were particularly affected (Hamer et al. 2010), but low altitude populations survived in areas less suitable for chytrid (Murray et al. 2011), allowing functioning metapopulations to be maintained, particularly in regions with a high density and connectivity of wetland habitats; see (Wassens et al. 2008, Heard et al. 2012a, b).

Despite persisting through the declines last century, remnant populations of L. raniformis around Melbourne are now threatened by the rapid pace of urban expansion. Urbanisation has led to wetland loss, degradation and fragmentation in each of the four regions in which significant remnant populations of L. raniformis persist (Wyndham, Melton, Hume-Whittlesea and Cardinia; Fig. 3.1), and this has caused the loss and fragmentation of some populations in recent years (Heard et al. 2010). This process will continue over the next three decades, with each of these regions being designated urban growth areas in which the proposed urban boundaries encompass numerous remaining populations of L. raniformis (Fig. 3.1; (DEPI 2013)).

Study area and field sampling

Tissue samples (N = 377) were collected from remnant populations of L. raniformis in the Wyndham, Melton, Hume-Whittlesea and Cardinia regions (Fig. 3.1). One hundred and seven samples were collected from Hume-Whittlesea as part of a previous study ((Heard et al. 2012a); see below). The remaining samples (N = 270) were collected during the current study between December 2010 and March 2011 across the other three

! 43 Chapter 3

regions. I divided each region into three geographic clusters of sites for the purposes of sampling (twelve clusters altogether, Fig. 3.1). I chose these clusters to maximise the geographic area sampled within each region, and to encompass networks of sites likely to support metapopulations of L. raniformis (Heard et al. 2012a, Heard et al. 2013). Sampling sites (N = 73) included slow-flowing pools along streams, as well as farm dams, swamps and water-treatment ponds. Between 1 and 10 individuals were sampled per site depending on population size and capture success.

Hume-Whittlesea

Melton

Melbourne

Wyndham

Port Phillip Bay Cardinia

km 09182736

Figure 3.1. Geographic distribution and sample sites of Litoria raniformis around Melbourne. Dark grey = current Melbourne metropolitan area, light grey = proposed urban growth area. Closed circles indicate detection records between 2000 and the present. Open circles indicate the location of sampling sites included in this study. The four study regions around Melbourne are indicated and clusters of sites within these regions circled (dashed lines).

Genetic samples were obtained from each frog by clipping a triangular section of toe webbing (~ 2 mm at the base) between the second and third toes closest to the body on the left hind limb. Toe web samples were stored in 95% ethanol and kept at -18°C for approximately two weeks, then -80°C for up to 3.5 months. Further details on these

44 ! Chapter 3 samples may be found in Appendix 3.1. Latex gloves were worn when taking samples and were changed between individuals. Tissue sampling equipment was also sterilised in 70% ethanol between frogs to prevent the spread of pathogens such as the chytrid fungus.

Samples from Hume-Whittlesea were collected between October 2004 and February 2006 as part of a study on the metapopulation dynamics of L. raniformis in this region (Heard et al. 2012a) (Fig. 3.1). In that study, frogs were captured by hand during spotlight surveys and tissue samples obtained by clipping the toe-pad on the left middle digit of the front left limb following standard procedures (see (Hale et al. 2013) for additional details).

DNA extraction and sequencing

I sequenced the mitochondrial gene Cytochrome oxidase I (COI) for all 377 samples and a sub-set of samples (N = 112) were sequenced for NADH dehydrogenase subunit 4 (ND4) encompassing all COI haplotypes and geographic regions. Each of these gene regions have been shown to be variable in previous studies of L. raniformis (Burns and Crayn 2006, Hale 2010). I also sequenced four nuclear gene regions (POMC, RAG-1, Rhod, CRYBA1 for a subset of 19 samples each, encompassing the extent of mtDNA haplotypic diversity) that are known to be variable at an intraspecific level in frogs (Bossuyt and Milinkovitch 2000, Dolman and Phillips 2004, Wiens et al. 2005, Gomez- Mestre et al. 2008).

Samples collected in the Hume-Whittlesea region were extracted and sequenced for COI as part of a previous, as yet unpublished, study (Hale 2010). I sequenced all other gene regions for these samples as part of the current study. Samples collected from all other geographic areas were extracted and sequenced for all gene regions during this study. In all cases, genomic DNA was extracted using a DNeasy Blood & Tissue Kit (QIAGEN) using the manufacturer’s animal tissue protocol.

I performed PCR amplification of the samples from Wyndham, Melton and Cardinia in a 20 µl total volume with 2 µl of DNA template, 10 µl GoTaq Hot Start Polymerase

(Promega) (25 mM MgCl2, GoTaq Hot Start Polymerase, 5X Colourless GoTaq Flexi

! 45 Chapter 3

Buffer, 5X Green GoTaq Flexi Buffer), 1 µM forward primer, 1 µM reverse primer and

6 µl dH2O. I amplified a partial sequence of 495 bp of COI, using the primers Cox (5’- TGATTCTTTGGGCATCCTGAAG-3’) and Coy (5’- GGGGTAGTCAGAATAGCGTCG-3’) (Schneider et al. 1998). Amplification involved 95°C 2 min, 37 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5 min (adapted from Hale et al. 2011). For COI amplification protocol of samples from Hume- Whittlesea, refer to Hale et al. (2010).

I amplified a partial sequence of 665 bp of ND4 for the subset of 112 samples (drawn from samples in all four regions), using the primers ND4-3 (F) (5’- TTAGCAGGAACACTTCTAAAACTAG-3’) and ND4-1 (R) (5’- GAAAGTGTTTAGCTTTCATCTCTAG-3’) (Burns and Crayn 2006). Amplification involved 95°C 2 min, 30 cycles of 95°C 30 s, 50°C 60 s, 72°C 45 s, followed by 72°C 5 min (adapted from Burns and Crayn 2006). Additionally, I screened four nuclear gene regions: proopiomelanocortin A (POMC), recombinase activating gene 1 (RAG-1), Rhodopsin (Rhod), and "-crystallin (CRYBA1), using a subset of 19 samples that each sequenced a different haplotype at the COI gene region (including two samples from the Hume-Whittlesea region). Protocols are provided in Appendix 3.2.

I diluted extracted DNA to a ratio of 1:10 with dH2O in all cases. I routinely included negative controls and checked for contamination. Following successful amplification, I purified PCR products using ExoSAP-IT (USB) following manufacturer’s instructions and sent the purified products to Macrogen (Korea) for sequencing using a 3730XL DNA sequencer (ABI). I BLAST searched all sequences to confirm identity and aligned them using Geneious Pro v. 5.6 (Drummond et al. 2012). I repeat sequenced all novel haplotypes.

Microsatellite genotyping

I genotyped four polymorphic microsatellite markers (Lr2, Lr6, Lr7 and Lr9) as described by Hale et al. (2011), in two sets of two marker multiplex PCR amplifications, for a subset of 117 samples (including 27 samples from the Hume- Whittlesea). The microsatellite loci chosen for this study had previously been developed and successfully genotyped for L. raniformis around Melbourne (Hale et al. 2011).

46 ! Chapter 3

Three markers incorporated a GTTTCTT ‘pigtail’ added to the 5’ end of the reverse primer to reduce variation in stutter (Lr2, Lr6, and Lr9). See Hale et al. (2011) for methods for fluorescently labelling fragments for all loci. The PCR was performed in a 10 µl total volume with 1 µl of DNA template (neat), 5 µl GoTaq Hot Start Polymerase

(Promega) (25 mM MgCl2, GoTaq Hot Start Polymerase, 5X Colourless GoTaq Flexi Buffer, 5X Green GoTaq Flexi Buffer), 0.5 µM reverse primer, 0.15 µM forward primer, 0.25 µM fluorescently labelled 454A primer and 3.1 µl dH2O. Amplification involved 95°C 2 min, 42 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5 min (adapted from Hale et al. (Hale et al. 2011)). Fragment analysis of PCR products were carried out by Macrogen on an Applied Biosystems ABI3730XL DNA analyser using a LIZ-500 size standard. Scoring was completed using Geneious Pro v 5.6, and all samples were screened manually for accuracy.

Data analyses

I estimated completeness of haplotype sampling of populations using the Stirling probability distribution and Bayes’ theorem (Dixon 2006). Median-joining haplotype networks were built using the program Network v. 4.610 (Bandelt et al. 1999) for COI and ND4 sequence data, for individual gene regions and a concatenated dataset. Haplotype networks were used, as they better illustrate intraspecific genetic divergence when the number of mutations between haplotypes is small (Crandall 1994).

I used DnaSP v. 5.10.1 (Librado and Rozas 2009) to detect signatures of a past genetic bottleneck based on mtDNA, estimating: (1) Tajima’s test statistic (D), where a large, positive value of D is consistent with a population that has experienced a recent bottleneck (Tajima 1989); (2) Fu’s test statistic (FS), which assesses the number of rare alleles in the population and departures from conditions of neutrality (Fu 1997), and; (3) the raggedness statistic (r), which quantifies the smoothness of the observed mismatch distribution and indicates whether populations are expanding or contracting. A substantial mismatch is characteristic of a population not at equilibrium (Harpending 1994). Values were determined for both COI separately, and then the COI and ND4 concatenated dataset.

! 47 Chapter 3

For the microsatellite data, I used Micro-Checker (v. 2.2.3) (van Oosterhout et al. 2004) to test for the presence of null alleles, large allele dropout and scoring error due to stuttering. This was calculated across the entire data set prior to analysing the number of genetic populations present and then for each genetic population individually once these

populations had been identified. Expected and observed heterozygosities (HE and HO respectively) were implemented in Genepop v. 4.2 (Rousset 2008) with default settings, to assess for linkage disequilibrium and deviations from Hardy-Weinberg equilibrium (HWE) across the four microsatellite loci for each population. Significance values were altered for multiple simultaneous tests using the false discovery rate (FDR) correction (Benjamini and Hochberg 1995). I calculated average allelic richness for each population, using HP-Rare v. 1.0, which accounts for differences in sample size

(Kalinowski 2005). FST between populations was calculated using GenAlEx v. 6.5 (Peakall and Smouse 2006b).

I estimated the number of distinct genetic populations using microsatellites for a subset of 117 samples in the Bayesian population clustering program Structure v. 2.3.4 (Pritchard et al. 2000). I used a correlated frequency model (to increase the power to detect subtle population structure) with admixture (the recommended starting point). The LOCPRIOR model was used, which can incorporate prior information when the signal of structuring is relatively weak (Hubisz et al. 2009). I used geographic region, (coded as one of the four regions from which samples were collected, see Fig. 3.1), as prior information. Each run had a burn-in of 100,000 Markov Chain Monte Carlo (MCMC) samples, with a further 100,000 samples used to characterise population structure. The number of genetic populations (K) was set to range from 1 to 12 (where 12 is the number of clusters within regions). Simulations were run ten times for each proposed value of K. I am confident that the chains had converged, as several runs at each K with different run lengths gave consistent parameter values. To determine K, I used Structure Harvester v. 0.6.93 (Earl and Vonholdt 2012), which plots the mean

estimated log probability of the data, loge Pr (X|K), as well as the rate of change in the log probability of the data between successive K estimates. The value of K with the

highest rate of change and the largest mean loge Pr (X|K) was selected (Evanno et al. 2005, Pritchard et al. 2010). The alternative inclusion of geographic cluster and COI haplotype as prior information lead to ambiguous results with regard to K and weak structure that did not seem biologically applicable (e.g. testing K = 1 - 12 and a

48 ! Chapter 3

LOCPRIOR of geographic cluster generated K = 3 or 10, and testing K = 1 - 20 and a LOCPRIOR of COI haplotype generated K = 2. Appendix 3.3).

To detect the signature of a past bottleneck from the microsatellite data, I used BOTTLENECK v. 1.2.02 (Cornuet and Luikart 1996, Piry et al. 1999) and M-ratio tests calculated in M_P_VAL and Critical_M (Garza and Williamson 2001) for each of the two populations. In BOTTLENECK I tested for heterozygosity excess, indicative of population size reduction, using a Wilcoxon test applied to a two-phase model (TPM), with 78% single step mutations, as recommended by Peery et al. (Peery et al. 2012) and a variance of 12 (Piry et al. 1999). An FDR adjustment for multiple comparisons was applied. M-ratio tests were calculated in M_P_VAL, using data formatted in FORMATOMATIC v. 0.8.1 (Manoukis 2007). M-ratio tests are based on the ratio of the number of microsatellite alleles to the range in allele size. During a bottleneck, the number of alleles is expected to decline faster than the range in allele size, so the M- ratio is expected to be smaller in bottlenecked populations (Peery et al. 2012).

The M-ratio method is likely to detect older events than tests based on heterozygosity, because more time is required for the M-statistic to reach equilibrium (Parra-Olea et al.

2012). Three parameters are required for M_P_VAL: theta (# = 4Neµ, where Ne = effective population size and µ = per generation mutation rate), the average size of mutations that are not one-step and the proportion of multi-step mutations. I calculated ! for each population identified by the preceding Structure analysis using Migrate v. 3.6.6 (Beerli 2009). I employed a Brownian motion model, using four static heating chains (1, 1.5, 3 and 1,000,000), swapping among chains every 10 steps. For each model, I performed 1,000,000 steps, sampled every 100 steps, with a burn-in length of 100,000 steps. The average size of mutations that are not one-step was set as the default 2.8 (Garza and Williamson 2001) and 0.22 was used as the recommended proportion of multi-step mutations (Peery et al. 2012). Values of M were compared with the critical value (MC), calculated in Critical_M, estimated after 10,000 simulations. In all the analyses using microsatellite data, the low number of markers used could have limited the power of these methods to detect a bottleneck. However, methods based on examining mtDNA haplotypes can be more powerful for detecting bottlenecks than microsatellite methods based on genetic structure (Franks et al. 2011).

49 Chapter 3

RESULTS

Genetic diversity

The posterior probabilities of completeness of haplotype sampling, Pr(n=x),were high for all regions (>0.99), indicating a high likelihood that all haplotypes from each region were sampled. At the finer scale, with regions separated into their component site clusters, there was at least 99% certainty of completeness of sampling for all clusters (Pr(n=x) $ 0.99) except the southern-most cluster from the Cardinia region, at which Pr(n=x) = 0.75. Overall, there was a high probability that all COI haplotypes were sampled.

A haplotype network was constructed from the 377 aligned COI mitochondrial gene regions (Fig. 3.2). In total, there were 22 polymorphic sites, representing 20 distinct haplotypes. Five haplotypes were present in the Wyndham region, seven in Melton, six in Hume-Whittlesea and ten in Cardinia. There was a common ‘central’ haplotype present (haplotype 1), representing 167 of the 377 samples. All observed haplotypes were a maximum of four mutational steps away from this central haplotype, except one sample from the Wyndham region, which was eight mutational steps away (haplotype 20). One haplotype was unique to the Wyndham region (haplotype 20), two Melton haplotypes were unique (haplotypes 13 and 14), two haplotypes were unique to Hume- Whittlesea (haplotypes 8 and 10), and ten haplotypes were unique to Cardinia (all the haplotypes observed for the area: haplotypes 2, 3, 4, 6, 7, 11, 12, 15, 18 and 19). Two additional haplotypes (haplotypes 9 and 17) were shared by individuals from Wyndham, Melton and Hume-Whittlesea. One haplotype (haplotype 16) was shared by individuals in Wyndham and Melton, and one haplotype (haplotype 5) was shared by individuals in Melton and Hume-Whittlesea. I also examined the COI haplotype composition of each cluster of sites (Fig. 3.3). The central haplotype (haplotype 1) was present in each of the three site clusters from Wyndham, Melton and Hume-Whittlesea (clusters A - I). Patterns of haplotype diversity and exchange between site clusters were also similar in Wyndham, Melton and Hume-Whittlesea.

50 ! Chapter 3

17 16 15 18

19 12 13 20

11 14 10 3

1 2 9 8

4

5 6 Cardinia Wyndham Hume-Whittlesea 7 Melton

Figure 3.2. Haplotype network for 495bp of the COI gene. Each pie represents a unique genetic sequence (haplotype) and the area of the pie is proportional to haplotype frequency within the entire data set. Each line represents one mutational step. Small black circles correspond to inferred alleles, missing from the data set.

The haplotype network derived from the 112 aligned ND4 mitochondrial gene regions revealed similar results to those derived from COI (Appendix 3.4). In total there were 26 distinct haplotypes. Seven haplotypes were present in the Wyndham region (three unique to the region), eight in Melton (three unique), seven in Hume-Whittlesea (five unique) and eleven from the Cardinia region (ten unique). The haplotype network derived from the concatenated COI and ND4 gene regions also revealed similar results to those derived from COI alone (Appendix 3.5). In total there were 33 distinct haplotypes. Eight haplotypes were present in the Wyndham region (three unique to the region), nine in Melton (three unique), eleven in Hume-Whittlesea (eight unique) and thirteen from the Cardinia region (all thirteen unique to the region).

1

! 51 Chapter 3

G

1 H I G 10 3 F 9 8 D H E 4 9 1 10 1 4 I 9 1 G 10 3 F B 8 8 C 9 4 7 4 A1 1 3 2 H 4 D 4 9 1 3 1 10 K 6 7 4 I E 9 1 J F 1 L 8 3 7 4 10 4 1 G 1 3 2 9 8 D 4 B 4 5 3 1 km 4 7 2 C 6 010203040 2 E H 1 1 4 A 9 1 2 3 1 10 4 I 1 9 10 3 1 G 18 17 B 4 5 F 4 8 9 8 15 13 K 2 C 2 4 7 4 1 16 11 12 14 1 1 A 1 3 J 2 17 20 2 3 D 4 H 16 1 4 19 1 3 9 1 11 L 6 7 10 14 15 18 17 E 4 I 13 9 1 15 13 K 11 12 14 F J 16 8 17 20 7 4 16 Haplotype A B C4 1 D E F G H I J K L 19 1 24 241 3 22 23 14 15 1 11 26 0 0 0 11 L 14 15 B 4 5 2 13 2 1D 4 4 5 4 0 0 2 0 0 0 0 0 0 2 3 C 12 3 5 7 0 1 1 6 0 0 0 0 0 1 6 E1 4 A0 1 3 4 8 12 0 3 5 0 0 0 5 2 0 3 1 0 0 0 0 0 0 0 0 0 0 6 0 1 0 0 0 1 0 0 0 0 0 0 0 km 18 17 7 0 0 0 2 6 0 0 0 0 0 0 0 010203040 13 K 8 0 0 0 1 0 0 18 0 0 0 0 0 15 B 4 5 16 11 12 14 4 9 20 C0 2 0 0 0 0 14 6 2 0 0 0 J 17 20 1 1 16 10 0 0 0 0 0 0 11 0 4 0 0 0 19km A 11 L 11 2 3 0 0 0 0 0 0 0 0 0 18 0102030400 9 14 13 15 12 1 0 0 0 0 0 0 0 0 0 0 16 0 13 0 0 0 0 0 0 0 0 0 0 5 7 18 17 14 0 0 0 0 0 0 0 0 0 0 8 8 15 13 K 15 0 0 0 0 0 0 0 0 0 7 0 1 J 16 11 12 14 16 0 0 0 0 0 0 0 0 0 4 0 1 17 20 16 17 0 0 0 0 0 0 0 0 0 0 1 2 19 11 L 14 18 0 0 0 0 0 0 0 0 0 1 0 0 13 15 19 0 0 0 0 0 0 0 0 0 0 0 1 20 0 0 0 0 0 0 0 0 0 0 0 1 km 010203040 1 Figure 3.3. Geographic association of COI haplotypes. Each pie represents the haplotypes found in that cluster and the area of the pie is proportional to sample size. Clusters have been assigned letters and haplotypes numbered. See included table for the km 010203040 number of individuals displaying each haplotype.

52 ! Chapter 3

Additionally, three nuclear gene regions were aligned, using a subset of the 19 most variable COI samples. The fourth nuclear gene region, CRYBA1, failed to successfully amplify. Rhod and RAG-1 showed no variation between samples. POMC showed variation for eight individuals, however this variation was not parsimoniously informative, as it occurred at heterozygous sites (refer to Appendix 3.1). Due to the lack of diversity, there was no scope for further population level analyses with the nuclear DNA.

Population structure

The Structure analysis using geographic region as prior information revealed two clear genetic units (!K = 34.740 for K = 2 and !K = 0.024 – 1.145 for K = 3 – 11). Population I contained all samples collected from the Wyndham, Melton and Hume- Whittlesea regions, while Population II contained all samples collected from the Cardinia region in the south-east (Fig. 3.4). Each of these populations contained 18 microsatellite alleles. I rejected the possibility of large allele dropout or scoring errors due to stuttering when testing each locus individually. There was no evidence of linkage disequilibrium (P = 0.025 – 0.045 following FDR correction for multiple comparisons). Overall, each locus generally conformed to HWE across both populations (Table 3.1). Two loci, Lr6 and Lr9, showed significant deviations from HWE, but the pattern was not consistent across populations (Population I Ho = 0.586 for Lr9 and Population II Ho = 0.367 for Lr6). I observed significant homozygote excess, suggesting the presence of null alleles in Population II for Lr6, however I chose to retain all loci for subsequent analyses, given there was no evidence of null alleles in Population I or when analysing the data as one population. Population I had 18 alleles, with an allelic richness of 2.33.

Population II had 18 alleles, and an allelic richness of 2.43. Pairwise FST between the two genetic populations was low (0.004) and not statistically significant (P = 0.176).

! 53 Chapter 3

1

0.8

0.6 Q 0.4

0.2

0 Wyndham Melton Hume-Whittlesea Cardinia

Population I Population II Geographic Region Figure 3.4. Structure bar plot (K=2) with individuals organised by geographic region. Each vertical bar represents a single individual and estimates of Q - the probability an individual belongs to a population of the given colour.

Table 3.1. Genetic variability of mitochondrial DNA (COI) and microsatellites in populations determined by STRUCTURE.

Genetic diversity Population I Population II

mtDNA Number of samples (N) 287 90 Length (bp) 495 495 No. haplotypes 10 10 Haplotype diversity (h) 0.658 (var=0.001, sd=0.028) 0.826 (var=0.000, sd=0.020) Nucleotide diversity (") 0.003 0.006 Average no. nucleotide differences (k) 1.463 (obs var=2.009, CV=0.970) 2.950 (obs var=3.835, CV=0.666) No. polymorphic sites (S) 18 10 Total no. mutations (Eta) 18 11 Tajima’s test statistic (D) -1.251 (not stat sig, P>0.1) 0.954 (not stat sig, P>0.1) Fu’s test statistic (Fs) -0.853 0.659 Raggedness index (r) 0.118 0.105

Microsatellites

Observed heterozygosity (HO) 0.586 0.550 Expected heterozygosity (HE) 0.608 0.636 Allelic richness (AR) 2.330 2.430

54 ! Chapter 3

Genetic bottleneck tests

There was only weak evidence for a genetic bottleneck based on mtDNA data for COI (Table 3.1). The raggedness statistics showed evidence of weak genetic bottlenecks in both populations, however neither population deviated significantly from expectations of neutrality, as revealed by Fu’s Fs statistic and

Tajima’s D statistic. No substantial evidence of bottlenecks was found in either population for concatenated regions of COI and ND4 (Appendix 3.6). A one-tailed Wilcoxon’s test for heterozygosity excess in BOTTLENECK found Population I displayed evidence of a recent genetic bottleneck (P = 0.031), however this did not remain statistically significant following FDR correction for multiple comparisons (P = 0.063). Similarly, the M-ratio tests on microsatellite loci did not detect evidence of a bottleneck in either of the populations. Using ! = 3.1 (Population I) and ! = 0.967 (Population II), the average M-ratio values were 0.817 - 0.888 and were higher than the M-critical values, of 0.695 - 0.697.

DISCUSSION

This study sought to assess the genetic structure and diversity of remnant populations of L. raniformis around Melbourne, Australia’s fastest growing city. I predicted that: (1) genetic diversity of these populations would be low due to past declines and more recent habitat loss and fragmentation; (2) that genetic bottlenecks would be evident, and; (3) that genetic structuring would be present between remnant populations in Cardinia and the other three regions studied.

Patterns of genetic structure and diversity

The mtDNA and nuclear DNA sequences revealed low levels of genetic diversity throughout remnant populations of L. raniformis around Melbourne, supporting my first hypothesis. Additionally, a large proportion of individuals included in this study shared one haplotype (haplotype 1), with few mutations between all the haplotypes present. In comparison, a study that looked at COI sequences across 34 amphibian species, from 6 families and 11 genera, reported an average nucleotide diversity of 0.203 which is

55 Chapter 3

considerably higher than the population-level diversity found in this study (0.003 and 0.006, see Table 3.1) (Smith et al. 2008).

Measures of genetic diversity within populations such as allelic richness, haplotype diversity, heterozygosity values and polymorphism have been found to be lower in amphibians in urban areas than in non-urban environments (Reh and Seitz 1990, Hitchings and Beebee 1997, Hitchings and Beebee 1998, Arens et al. 2007, Noel et al. 2007, Noel and Lapointe 2010). It is difficult to make direct comparisons of genetic diversity between species using microsatellites, which is exacerbated by the low number of microsatellites used for this study. Nevertheless, the measures of heterozygosity and allelic richness are broadly similar to those of other amphibians, including related taxa in Australia (Burns et al. 2004). In a study of the Green and Golden Bell Frog (Litoria aurea), the sister species to L. raniformis, that covered ~1,000 km, expected heterozygosity levels were reported as high (0.43-0.82, mean 0.69) when compared to other amphibian species (Burns et al. 2004). A previous study (Hale 2010) examined the structure and fragmentation of L. raniformis metapopulations in one of the study regions, Hume-Whittlesea. Using data from 11 microsatellite markers, the signature of a recent bottleneck was found at one site. This population was separated from nearby sites by a four-lane highway, which appeared to significantly reduce gene flow. Hale (2010) suggested that urbanisation around Melbourne has the potential to reduce the genetic diversity of L. raniformis due to bottlenecks.

I hypothesised that genetic bottlenecks would be evident in the populations of L. raniformis studied here, both due to past population declines (resulting from chytridiomycosis in particular) and more recent habitat loss and fragmentation due to urbanisation. However, the data did not support this hypothesis, with both mtDNA and nuclear microsatellite loci analyses failing to detect the signature of a bottleneck. For the microsatellite data, I assessed levels of heterozygosity excess, indicative of population size reduction. However, while commonly used, this method is strongly influenced by the more common alleles (Hartl and Clark 2007), which may explain why I did not detect a bottleneck. The ability to detect a prior bottleneck is also influenced by the population size pre-bottleneck, the mutational model and parameters chosen (Williamson-Natesan 2005), statistical power (Cornuet and Luikart 1996, Peery et al. 2012) the magnitude of the bottleneck and how recently it occurred (Williamson-

56 ! Chapter 3

Natesan 2005). Indeed, populations can experience the negative impacts of a bottleneck for many generations before it is detected (Peery et al. 2012).

One or more of these factors may have influenced my ability to detect a population bottleneck in the study species. However, dramatic declines in the sister species L. aurea, similarly found no signature of a bottleneck (Burns et al. 2004). It seems likely, therefore, that the contemporary pattern of genetic diversity of L. raniformis around Melbourne is due to factors other than recent bottlenecks. In the first instance, it may be that remnant populations of this species around Melbourne may not have suffered the drastic reductions in abundance necessary to produce a bottleneck signature (Peery et al. 2012). Secondly, the populations studied here may have retained sufficient local connectivity to obscure or prevent bottlenecks (Keller et al. 2001). Previous research confirms that L. raniformis displays metapopulation dynamics, with occupancy, mark-recapture and genetic data confirming that migration over distances of 1 – 2 km is crucial to the viability of population networks (Heard et al. 2012a, Hale et al. 2013, Heard et al. 2013). Given their persistence, it may be inferred that the populations sampled during this study continue to exchange migrants with nearby populations, and remain part of a functioning metapopulation. This level of connectivity may have been sufficient to obscure or prevent genetic bottlenecks in the populations studied here. I encourage further research on the ecological and evolutionary drivers of the low genetic diversity but apparent absence of bottlenecks among remnant populations of L. raniformis around Melbourne, including more detailed microsatellite analyses.

Genetic structuring was present between remnant populations in Cardinia and the other three regions, due to both the historical and more recent isolation of this region, supporting my third hypothesis. This was demonstrated by both the mtDNA and microsatellite analyses. Of the four main regions in the study, Cardinia exhibited the highest level of genetic diversity and all haplotypes present there were unique to the region. Additionally, the microsatellite analyses revealed Cardinia should be considered a distinct genetic unit from the other regions.

! 57 Chapter 3

Conservation implications

This study indicates populations of L. raniformis from Cardinia in south-eastern Melbourne are genetically distinct from the west and north of Melbourne, due to the presence of unique haplotypes and higher genetic diversity. Under Moritz’s (Moritz 1994) definition, which incorporates reciprocal monophyly, the Cardinia region would not qualify as an Evolutionarily Significant Unit (ESU). However, I suggest that populations from this region should be regarded as a separate Management Unit (MU). Management Units can be used to address current population structure and short-term management issues, and are considered distinct from ESUs (Moritz 1994). Based on my results, I recommend that the maintenance of genetic diversity in the Cardinia region be prioritised and that the region should be considered genetically independent to the remainder of urban Melbourne.

The Victorian Government has classified suitable habitat for L. raniformis within the urban growth areas into two categories: 1. high-quality habitat that will be protected and managed for the species, and; 2. habitat of lower conservation significance that can be destroyed for urban development, but for which compensatory habitat is required (DEPI 2013). Translocation represents a core component of conservation planning for L. raniformis in Melbourne’s urban growth areas (DEPI 2013). Individuals occupying habitat designated for destruction (i.e., that defined as category 2 habitat, as above) will be translocated in cases where there are appropriate locations to receive animals, and where risks, including disease, are considered manageable (DEPI 2013). Under this policy, some of the populations studied here in the Melton, Hume-Whittlesea and Cardinia areas will be translocated during future urban expansion. The appropriateness of amphibian and reptile translocations has been debated widely during the past 20 years and many attempted translocations have been unsuccessful (Germano and Bishop 2009). There is a clear conflict in policies regarding translocations in Victoria. There is currently no evidence that populations of L. raniformis can be successfully relocated (Heard et al. 2010). Indeed, all translocation attempts for this species and the closely related L. aurea have failed or the focal populations are performing poorly (Koehler et al. 2014b), despite L. aurea being the subject of more translocation attempts than any other Australian frog (White and Pyke 2008).

58 ! Chapter 3

If future captive breeding programs are required or translocations are attempted for L. raniformis, it is important that the geographic location of specimens is considered. This will help to maintain the genetic integrity and evolutionary potential of the species around Melbourne, two key objectives of Victoria’s Flora and Fauna Guarantee Act 1988. This appears most important in the south-east Cardinia region, to maintain the levels of genetic diversity and the unique haplotypes found in the region. I caution that I only analysed neutral genetic diversity here, when in fact levels of genetic variability inferred from neutral markers are often poor correlates of quantitative variation for adaptively important traits (Bekessy et al. 2003). Nevertheless, I advise that translocation of individuals between regions should be avoided, and that fine scale genetic studies within regions should be undertaken to determine the long-term viability of translocations within regions. By quantifying genetic structure and diversity of L. raniformis across Melbourne’s urban fringe, using mitochondrial DNA and microsatellite markers, I am able to forewarn managers of the low genetic diversity displayed by remnant populations, identify genetic management units and regional centres of haplotype endemism.

DATA ACCESSIBILITY

DNA sequences generated in this study have GenBank accessions: KP343052- KP343597

! 59

60 APPENDICES ! Appendix 3.1. Museum Victoria registration numbers, localities of specimens from which DNA was extracted and GenBank accession numbers.

Museum ID Location Latitude/longitude COI # ND4 # POMC # RAG-1 # Rhod # NMVZ16504 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343052 KP343429 NMVZ16505 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343053 KP343430 NMVZ16506 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343054 NMVZ16507 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343055 NMVZ16508 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343056 NMVZ16509 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343057 NMVZ16510 Australia:Kings Park, Vic 37.74 S 144.76 E KP343058 KP343431 NMVZ16514 Australia:Kings Park, Vic 37.74 S 144.76 E KP343059 Chapter 3 NMVZ16515 Australia:Burnside Heights, Vic 37.74 S 144.76 E KP343060 NMVZ16516 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343061 KP343432 NMVZ16517 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343062 NMVZ16518 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343063 KP343433 NMVZ16519 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343064 KP343434 NMVZ16520 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343065 NMVZ16521 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343066 KP343435 KP343541 KP343560 KP343579 NMVZ16522 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343067 NMVZ16526 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343068 KP343436 NMVZ16527 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343069 NMVZ16528 Australia:Caroline Springs, Vic 37.74 S 144.7 E KP343070 NMVZ16529 Australia:Caroline Springs, Vic 37.74 S 144.7 E KP343071 NMVZ16530 Australia:Caroline Springs, Vic 37.74 S 144.7 E KP343072 KP343437 NMVZ16534 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343073 KP343438 NMVZ16535 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343074 KP343439 NMVZ16536 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343075 KP343440 NMVZ16537 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343076 NMVZ16538 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343077

!

NMVZ16539 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343078 NMVZ16540 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343079 NMVZ16541 Australia:Caroline Springs, Vic 37.74 S 144.75 E KP343080 NMVZ16542 Australia:Kings Park, Vic 37.74 S 144.76 E KP343081 NMVZ16543 Australia:Burnside Heights, Vic 37.74 S 144.76 E KP343082 NMVZ16544 Australia:Burnside Heights, Vic 37.74 S 144.76 E KP343083 NMVZ16548 Australia:Burnside Heights, Vic 37.74 S 144.76 E KP343084 NMVZ16549 Australia:Burnside Heights, Vic 37.74 S 144.76 E KP343085 NMVZ16550 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343086 NMVZ16551 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343087 NMVZ16552 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343088 KP343441 NMVZ16553 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343089 NMVZ16554 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343090 KP343442 KP343542 KP343561 KP343580 NMVZ16555 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343091 NMVZ16556 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343092

NMVZ16557 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343093 KP343443 KP343543 KP343562 KP343581 Chapter 3 NMVZ16558 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343094 KP343444 NMVZ16562 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343095 KP343445 NMVZ16563 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343096 NMVZ16564 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343097 NMVZ16565 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343098 NMVZ16566 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343099 NMVZ16567 Australia:Officer, Vic 38.07 S 145.42 E KP343100 KP343446 NMVZ16571 Australia:Officer, Vic 38.07 S 145.42 E KP343101 KP343447 NMVZ16572 Australia:Officer, Vic 38.07 S 145.42 E KP343102 KP343448 NMVZ16573 Australia:Officer, Vic 38.07 S 145.42 E KP343103 NMVZ16574 Australia:Officer, Vic 38.07 S 145.42 E KP343104 NMVZ16575 Australia:Officer, Vic 38.07 S 145.42 E KP343105 NMVZ16576 Australia:Officer, Vic 38.07 S 145.42 E KP343106 NMVZ16580 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343107 NMVZ16581 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343108 NMVZ16582 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343109 NMVZ16583 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343110

61 NMVZ16584 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343111

!

62 NMVZ16585 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343112 NMVZ16586 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343113 NMVZ16587 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343114 NMVZ16588 Australia:Caroline Springs, Vic 37.75 S 144.73 E KP343115 NMVZ16589 Australia:Caroline Springs, Vic 37.75 S 144.74 E KP343116 NMVZ16590 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343117 KP343449 NMVZ16591 Australia:Caroline Springs, Vic 37.74 S 144.74 E KP343118 NMVZ16592 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343119 KP343450 NMVZ16596 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343120 NMVZ16597 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343121 KP343451 NMVZ16598 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343122 KP343452 NMVZ16599 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343123 NMVZ16600 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343124 NMVZ16601 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343125 KP343453 NMVZ16602 Australia:Werribee Western Treatment Plant, Vic 38 S 144.64 E KP343126 KP343454

NMVZ16603 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343127 Chapter 3 NMVZ17254 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343128 NMVZ17255 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343129 NMVZ17259 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343130 KP343455 NMVZ17260 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343131 NMVZ17261 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343132 KP343456 NMVZ17262 Australia:Werribee Western Treatment Plant, Vic 38 S 144.63 E KP343133 NMVZ17266 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.65 E KP343134 KP343457 NMVZ17267 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343135 NMVZ17268 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343136 NMVZ17269 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343137 NMVZ17270 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343138 NMVZ17271 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343139 NMVZ17272 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343140 NMVZ17273 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343141 NMVZ17274 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343142 NMVZ17275 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343143 KP343458 NMVZ17276 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343144 NMVZ17277 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343145

!

NMVZ17278 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343146 NMVZ17279 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343147 NMVZ17280 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.67 E KP343148 NMVZ17281 Australia:Pakenham, Vic 38.15 S 145.5 E KP343149 KP343459 NMVZ17285 Australia:Pakenham, Vic 38.15 S 145.5 E KP343150 KP343460 NMVZ17286 Australia:Pakenham, Vic 38.15 S 145.5 E KP343151 NMVZ17287 Australia:Pakenham, Vic 38.15 S 145.5 E KP343152 KP343461 KP343544 KP343563 KP343582 NMVZ17288 Australia:Pakenham, Vic 38.15 S 145.5 E KP343153 KP343462 NMVZ17289 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343154 NMVZ17290 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343155 KP343463 NMVZ17291 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343156 KP343464 NMVZ17292 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343157 KP343465 NMVZ17296 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343158 KP343466 KP343545 KP343564 KP343583 NMVZ17297 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.51 E KP343159 KP343467 NMVZ17298 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343160 KP343468

NMVZ17299 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343161 Chapter 3 NMVZ17303 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343162 KP343469 NMVZ17304 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343163 NMVZ17305 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343164 NMVZ17306 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343165 NMVZ17307 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343166 NMVZ17308 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343167 KP343470 NMVZ17309 Australia:Werribee Western Treatment Plant, Vic 38.05 S 144.51 E KP343168 NMVZ17310 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.52 E KP343169 NMVZ17314 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343170 NMVZ17315 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343171 NMVZ17316 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343172 NMVZ17317 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343173 NMVZ17318 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343174 NMVZ17319 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343175 NMVZ17320 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343176 NMVZ17321 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.54 E KP343177 NMVZ17322 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.54 E KP343178

63 NMVZ17323 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343179

!

64 NMVZ17324 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343180 NMVZ17325 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343181 NMVZ17326 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343182 NMVZ17327 Australia:Werribee Western Treatment Plant, Vic 38.04 S 144.53 E KP343183 KP343471 NMVZ17328 Australia:Pakenham, Vic 38.09 S 145.45 E KP343184 NMVZ17332 Australia:Pakenham, Vic 38.09 S 145.45 E KP343185 NMVZ17333 Australia:Pakenham, Vic 38.09 S 145.45 E KP343186 NMVZ17334 Australia:Pakenham, Vic 38.09 S 145.45 E KP343187 NMVZ17335 Australia:Pakenham, Vic 38.09 S 145.45 E KP343188 NMVZ17336 Australia:Pakenham, Vic 38.09 S 145.45 E KP343189 NMVZ17337 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343190 KP343472 KP343546 KP343565 KP343584 NMVZ17341 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343191 KP343473 NMVZ17342 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343192 NMVZ17343 Australia:Pakenham South, Vic 38.13 S 145.51 E KP343193 KP343474 KP343547 KP343566 KP343585 NMVZ17344 Australia:Pakenham South, Vic 38.13 S 145.51 E KP343194

NMVZ17345 Australia:Pakenham South, Vic 38.13 S 145.51 E KP343195 Chapter 3 NMVZ17346 Australia:Pakenham South, Vic 38.13 S 145.51 E KP343196 NMVZ17347 Australia:Pakenham South, Vic 38.13 S 145.51 E KP343197 KP343475 NMVZ17351 Australia:Nar Nar Goon, Vic 38.08 S 145.55 E KP343198 NMVZ17352 Australia:Nar Nar Goon, Vic 38.08 S 145.55 E KP343199 KP343476 NMVZ17353 Australia:Nar Nar Goon, Vic 38.08 S 145.55 E KP343200 KP343477 NMVZ17354 Australia:Nar Nar Goon, Vic 38.08 S 145.55 E KP343201 NMVZ17355 Australia:Nar Nar Goon, Vic 38.08 S 145.52 E KP343202 NMVZ17356 Australia:Nar Nar Goon, Vic 38.08 S 145.52 E KP343203 NMVZ17357 Australia:Nar Nar Goon, Vic 38.08 S 145.52 E KP343204 NMVZ17358 Australia:Nar Nar Goon, Vic 38.08 S 145.53 E KP343205 NMVZ17362 Australia:Nar Nar Goon, Vic 38.08 S 145.53 E KP343206 KP343478 NMVZ17363 Australia:Nar Nar Goon, Vic 38.08 S 145.54 E KP343207 NMVZ17364 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343208 NMVZ17365 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343209 NMVZ17366 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343210 KP343479 NMVZ17367 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343211 NMVZ17368 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343212 NMVZ17369 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343213

!

NMVZ17370 Australia:Nar Nar Goon, Vic 38.07 S 145.55 E KP343214 NMVZ17371 Australia:Pakenham, Vic 38.07 S 145.51 E KP343215 KP343480 KP343548 KP343567 KP343586 NMVZ17375 Australia:Pakenham, Vic 38.09 S 145.45 E KP343216 NMVZ17376 Australia:Pakenham, Vic 38.09 S 145.45 E KP343217 NMVZ17377 Australia:Pakenham, Vic 38.09 S 145.45 E KP343218 NMVZ17378 Australia:Pakenham, Vic 38.09 S 145.45 E KP343219 KP343481 NMVZ17379 Australia:Pakenham, Vic 38.09 S 145.45 E KP343220 NMVZ17380 Australia:Pakenham, Vic 38.12 S 145.46 E KP343221 KP343482 NMVZ17381 Australia:Pakenham, Vic 38.12 S 145.46 E KP343222 KP343483 NMVZ17382 Australia:Pakenham, Vic 38.12 S 145.46 E KP343223 NMVZ17383 Australia:Pakenham, Vic 38.12 S 145.46 E KP343224 KP343484 KP343549 KP343568 KP343587 NMVZ17384 Australia:Pakenham, Vic 38.12 S 145.46 E KP343225 NMVZ17385 Australia:Rockbank, Vic 37.69 S 144.67 E KP343226 KP343485 KP343550 KP343569 KP343588 NMVZ17389 Australia:Rockbank, Vic 37.69 S 144.67 E KP343227 KP343486 NMVZ17390 Australia:Rockbank, Vic 37.69 S 144.67 E KP343228 KP343487

NMVZ17391 Australia:Rockbank, Vic 37.69 S 144.67 E KP343229 KP343488 Chapter 3 NMVZ17392 Australia:Rockbank, Vic 37.69 S 144.67 E KP343230 KP343489 NMVZ17393 Australia:Rockbank, Vic 37.69 S 144.67 E KP343231 KP343490 NMVZ17394 Australia:Rockbank, Vic 37.69 S 144.67 E KP343232 NMVZ17398 Australia:Rockbank, Vic 37.69 S 144.67 E KP343233 NMVZ17399 Australia:Rockbank, Vic 37.69 S 144.67 E KP343234 KP343491 KP343551 KP343570 KP343589 NMVZ17400 Australia:Rockbank, Vic 37.69 S 144.67 E KP343235 NMVZ17401 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343236 KP343492 NMVZ17402 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343237 NMVZ17403 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343238 NMVZ17407 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343239 NMVZ17411 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343240 NMVZ17412 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343241 NMVZ17413 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343242 KP343493 NMVZ17414 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343243 NMVZ17415 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343244 KP343494 NMVZ17416 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.59 E KP343245 NMVZ17417 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.56 E KP343246

65 NMVZ17418 Australia:Cardinia, Vic 38.09 S 145.43 E KP343247 KP343495

!

66 NMVZ17419 Australia:Cardinia, Vic 38.09 S 145.43 E KP343248 KP343496 NMVZ17420 Australia:Pakenham, Vic 38.09 S 145.46 E KP343249 NMVZ17421 Australia:Pakenham, Vic 38.07 S 145.53 E KP343250 KP343497 NMVZ17422 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343251 KP343498 NMVZ17423 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343252 NMVZ17424 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343253 NMVZ17425 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343254 NMVZ17426 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343255 NMVZ17427 Australia:Nar Nar Goon, Vic 38.07 S 145.53 E KP343256 NMVZ17428 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343257 NMVZ17429 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343258 NMVZ17430 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343259 NMVZ17434 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343260 NMVZ17435 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343261 NMVZ17436 Australia:Werribee Western Treatment Plant, Vic 37.97 S 144.62 E KP343262

NMVZ17437 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.57 E KP343263 KP343499 Chapter 3 NMVZ17441 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.57 E KP343264 KP343500 NMVZ17442 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.57 E KP343265 NMVZ17443 Australia:Werribee Western Treatment Plant, Vic 38 S 144.55 E KP343266 NMVZ17444 Australia:Werribee Western Treatment Plant, Vic 38 S 144.55 E KP343267 NMVZ17445 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.55 E KP343268 NMVZ17446 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.57 E KP343269 NMVZ17447 Australia:Werribee Western Treatment Plant, Vic 37.98 S 144.57 E KP343270 NMVZ17448 Australia:Plumpton, Vic 37.7 S 144.66 E KP343271 NMVZ17449 Australia:Plumpton, Vic 37.7 S 144.66 E KP343272 NMVZ17450 Australia:Plumpton, Vic 37.7 S 144.66 E KP343273 NMVZ17451 Australia:Plumpton, Vic 37.7 S 144.66 E KP343274 NMVZ17452 Australia:Plumpton, Vic 37.7 S 144.66 E KP343275 NMVZ17453 Australia:Plumpton, Vic 37.7 S 144.66 E KP343276 NMVZ17457 Australia:Rockbank, Vic 37.7 S 144.63 E KP343277 KP343501 NMVZ17461 Australia:Rockbank, Vic 37.7 S 144.63 E KP343278 NMVZ17462 Australia:Rockbank, Vic 37.7 S 144.63 E KP343279 NMVZ17463 Australia:Rockbank, Vic 37.7 S 144.63 E KP343280 KP343502 NMVZ17464 Australia:Rockbank, Vic 37.7 S 144.63 E KP343281

!

NMVZ17465 Australia:Pakenham, Vic 38.1 S 145.46 E KP343282 NMVZ17466 Australia:Pakenham, Vic 38.1 S 145.46 E KP343283 NMVZ17467 Australia:Pakenham, Vic 38.1 S 145.46 E KP343284 NMVZ17468 Australia:Pakenham South, Vic 38.14 S 145.53 E KP343285 NMVZ17469 Australia:Pakenham South, Vic 38.14 S 145.53 E KP343286 NMVZ17473 Australia:Nar Nar Goon, Vic 38.12 S 145.54 E KP343287 NMVZ17474 Australia:Nar Nar Goon, Vic 38.12 S 145.54 E KP343288 NMVZ17475 Australia:Bayles, Vic 38.16 S 145.54 E KP343289 NMVZ17476 Australia:Bayles, Vic 38.16 S 145.54 E KP343290 NMVZ17477 Australia:Bayles, Vic 38.16 S 145.54 E KP343291 NMVZ17478 Australia:Bayles, Vic 38.16 S 145.54 E KP343292 NMVZ17479 Australia:Bayles, Vic 38.16 S 145.54 E KP343293 KP343503 KP343552 KP343571 KP343590 NMVZ17480 Australia:Bayles, Vic 38.16 S 145.54 E KP343294 NMVZ17484 Australia:Bayles, Vic 38.16 S 145.54 E KP343295 KP343504 KP343553 KP343572 KP343591 NMVZ17485 Australia:Pakenham, Vic 38.09 S 145.45 E KP343296

NMVZ17486 Australia:Nar Nar Goon, Vic 38.13 S 145.56 E KP343297 KP343505 KP343554 KP343573 KP343592 Chapter 3 NMVZ17487 Australia:Nar Nar Goon, Vic 38.13 S 145.56 E KP343298 KP343506 KP343555 KP343574 KP343593 NMVZ17488 Australia:Nar Nar Goon, Vic 38.13 S 145.56 E KP343299 KP343507 NMVZ17489 Australia:Nar Nar Goon, Vic 38.13 S 145.56 E KP343300 NMVZ17490 Australia:Nar Nar Goon, Vic 38.13 S 145.56 E KP343301 NMVZ17491 Australia:Werribee Western Treatment Plant, Vic 37.99 S 144.57 E KP343302 KP343508 NMVZ17492 Australia:Werribee Western Treatment Plant, Vic 37.99 S 144.57 E KP343303 KP343509 KP343556 KP343575 KP343594 NMVZ17493 Australia:Werribee Western Treatment Plant, Vic 37.99 S 144.57 E KP343304 KP343510 NMVZ17494 Australia:Werribee Western Treatment Plant, Vic 37.99 S 144.57 E KP343305 KP343511 NMVZ17495 Australia:Werribee Western Treatment Plant, Vic 37.99 S 144.57 E KP343306 KP343512 NMVZ17496 Australia:Koo Wee Rup North, Vic 38.15 S 145.55 E KP343307 KP343513 KP343557 KP343576 KP343595 NMVZ17497 Australia:Koo Wee Rup North, Vic 38.15 S 145.55 E KP343308 NMVZ17498 Australia:Koo Wee Rup North, Vic 38.15 S 145.55 E KP343309 NMVZ17499 Australia:Koo Wee Rup North, Vic 38.15 S 145.55 E KP343310 NMVZ17500 Australia:Nar Nar Goon, Vic 38.08 S 145.53 E KP343311 NMVZ17501 Australia:Nar Nar Goon, Vic 38.08 S 145.53 E KP343312 NMVZ17502 Australia:Rockbank, Vic 37.73 S 144.67 E KP343313 NMVZ17506 Australia:Rockbank, Vic 37.73 S 144.67 E KP343314

67 NMVZ17507 Australia:Rockbank, Vic 37.73 S 144.67 E KP343315

!

68 NMVZ17508 Australia:Rockbank, Vic 37.73 S 144.67 E KP343316 NMVZ17509 Australia:Rockbank, Vic 37.73 S 144.67 E KP343317 NMVZ17510 Australia:Rockbank, Vic 37.73 S 144.67 E KP343318 NMVZ17511 Australia:Rockbank, Vic 37.73 S 144.67 E KP343319 NMVZ17512 Australia:Rockbank, Vic 37.73 S 144.67 E KP343320 NMVZ17513 Australia:Rockbank, Vic 37.73 S 144.67 E KP343321 NMVZ31401 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343322 NMVZ31400 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343323 NMVZ31399 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343324 NMVZ31397 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343325 NMVZ31396 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343326 NMVZ31393 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343327 NMVZ30952 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343328 NMVZ31398 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343329 NMVZ31394 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343330 KP343514

NMVZ31391 Australia:Campbellfield, Vic 37.69 S 144.96 E KP343331 KP343515 Chapter 3 NMVZ31384 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343332 NMVZ31386 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343333 KP343516 NMVZ31385 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343334 KP343517 NMVZ31368 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343335 KP343518 NMVZ31371 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343336 NMVZ31381 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343337 NMVZ31363 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343338 KP343519 NMVZ31365 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343339 KP343520 NMVZ31370 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343340 KP343521 NMVZ31358 Australia:Campbellfield, Vic 37.69 S 144.98 E KP343341 KP343522 NMVZ31500 Australia:Somerton, Vic 37.63 S 144.96 E KP343342 NMVZ31516 Australia:Somerton, Vic 37.63 S 144.96 E KP343343 NMVZ31513 Australia:Somerton, Vic 37.63 S 144.96 E KP343344 NMVZ31503 Australia:Somerton, Vic 37.63 S 144.96 E KP343345 NMVZ31498 Australia:Somerton, Vic 37.63 S 144.96 E KP343346 NMVZ31489 Australia:Somerton, Vic 37.63 S 144.96 E KP343347 NMVZ31499 Australia:Somerton, Vic 37.63 S 144.96 E KP343348 NMVZ31515 Australia:Somerton, Vic 37.63 S 144.96 E KP343349

!

NMVZ31509 Australia:Somerton, Vic 37.63 S 144.96 E KP343350 NMVZ31492 Australia:Somerton, Vic 37.63 S 144.96 E KP343351 NMVZ31447 Australia:Somerton, Vic 37.63 S 144.96 E KP343352 NMVZ31474 Australia:Somerton, Vic 37.63 S 144.96 E KP343353 KP343523 NMVZ31473 Australia:Somerton, Vic 37.63 S 144.96 E KP343354 KP343524 NMVZ31469 Australia:Somerton, Vic 37.63 S 144.96 E KP343355 NMVZ31437 Australia:Somerton, Vic 37.63 S 144.97 E KP343356 NMVZ31446 Australia:Somerton, Vic 37.63 S 144.97 E KP343357 NMVZ31463 Australia:Somerton, Vic 37.63 S 144.97 E KP343358 NMVZ31442 Australia:Somerton, Vic 37.64 S 144.97 E KP343359 NMVZ31444 Australia:Somerton, Vic 37.64 S 144.97 E KP343360 NMVZ31452 Australia:Somerton, Vic 37.64 S 144.97 E KP343361 KP343525 NMVZ31445 Australia:Somerton, Vic 37.64 S 144.97 E KP343362 NMVZ31453 Australia:Somerton, Vic 37.64 S 144.97 E KP343363 NMVZ31448 Australia:Somerton, Vic 37.64 S 144.97 E KP343364

NMVZ31434 Australia:Somerton, Vic 37.64 S 144.97 E KP343365 Chapter 3 NMVZ31439 Australia:Somerton, Vic 37.64 S 144.97 E KP343366 NMVZ31467 Australia:Somerton, Vic 37.64 S 144.97 E KP343367 NMVZ31468 Australia:Somerton, Vic 37.64 S 144.97 E KP343368 NMVZ31485 Australia:Somerton, Vic 37.64 S 144.97 E KP343369 NMVZ31490 Australia:Somerton, Vic 37.64 S 144.97 E KP343370 NMVZ31497 Australia:Somerton, Vic 37.64 S 144.97 E KP343371 KP343526 NMVZ31495 Australia:Somerton, Vic 37.64 S 144.97 E KP343372 NMVZ31505 Australia:Somerton, Vic 37.64 S 144.97 E KP343373 KP343527 NMVZ31476 Australia:Somerton, Vic 37.64 S 144.97 E KP343374 NMVZ31502 Australia:Somerton, Vic 37.64 S 144.97 E KP343375 KP343528 NMVZ31496 Australia:Somerton, Vic 37.64 S 144.97 E KP343376 KP343529 NMVZ31487 Australia:Somerton, Vic 37.64 S 144.97 E KP343377 KP343530 NMVZ31477 Australia:Somerton, Vic 37.64 S 144.97 E KP343378 KP343531 NMVZ31432 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343379 NMVZ31416 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343380 NMVZ31403 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343381 NMVZ31424 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343382

69 NMVZ31420 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343383

!

70 NMVZ31428 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343384 NMVZ31430 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343385 NMVZ31405 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343386 NMVZ31433 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343387 NMVZ31419 Australia:Donnybrook, Vic 37.54 S 144.97 E KP343388 NMVZ31418 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343389 NMVZ31422 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343390 NMVZ31408 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343391 NMVZ31426 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343392 NMVZ31409 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343393 NMVZ31412 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343394 NMVZ31415 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343395 NMVZ31414 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343396 NMVZ31421 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343397 NMVZ31410 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343398

NMVZ31546 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343399 Chapter 3 NMVZ31562 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343400 NMVZ31549 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343401 NMVZ31560 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343402 KP343532 NMVZ30870 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343403 NMVZ31574 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343404 NMVZ31533 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343405 NMVZ31564 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343406 NMVZ31561 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343407 NMVZ31556 Australia:Donnybrook, Vic 37.54 S 144.96 E KP343408 NMVZ31548 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343409 NMVZ31539 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343410 NMVZ31571 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343411 NMVZ31534 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343412 NMVZ31545 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343413 NMVZ31537 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343414 NMVZ31554 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343415 KP343533 NMVZ31544 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343416 KP343534 NMVZ31535 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343417 KP343535

!

NMVZ31559 Australia:Donnybrook, Vic 37.54 S 144.95 E KP343418 KP343536 KP343558 KP343577 KP343596 NMVZ31351 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343419 NMVZ31355 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343420 NMVZ31339 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343421 NMVZ31357 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343422 NMVZ31342 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343423 NMVZ31343 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343424 NMVZ31340 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343425 KP343537 NMVZ31356 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343426 KP343538 NMVZ31338 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343427 KP343539 NMVZ31352 Australia:Donnybrook, Vic 37.55 S 144.94 E KP343428 KP343540 KP343559 KP343578 KP343597 Chapter 3 71

! Chapter 3

Appendix 3.2. Nuclear DNA sequencing protocols. I sequenced four nuclear gene regions using a subset of the 19 most variable samples at the COI gene region. For the nuclear gene proopiomelanocortin A (POMC), I amplified a 611 bp sequence, using the primers POMC-1 (5’- GAATGTATYAAAGMMTGCAAGATGGWCCT-3’) and POMC-2 (5’- TAYTGRCCCTTYTTGTGGGCRTT-3’) (Wiens et al. 2005). Amplification involved 95°C 2 min, 40 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5 min.

For the nuclear gene recombinase activating gene 1 (RAG-1), I amplified a 814 bp sequence, using the primers RS1f (F) (5’-TGCAGTCAGTAYCAYAARATGTAC-3’) P. Chippindale (pers com) from (Gomez-Mestre et al. 2008) and R1-GFR (5’- GAAGCGCCTGAACAGTTTATTAC-3’) (Faivovich et al. 2005). Amplification involved 95°C 2 min, 40 cycles of 95°C 30 s, 51.9°C 45 s, 72°C 45 s, followed by 72°C 5 min.

For the nuclear gene Rhodopsin (Rhod), I amplified a 321 bp sequence, using the primers Rhod 1A (5’- ACCATGAACGGAACAGAAGGYCC-3’) and Rhod 1C (5’- CCAAGGGTAGCGAAGAARCCTTC-3’) (Bossuyt and Milinkovitch 2000). Amplification involved 95°C 2 min, 40 cycles of 95°C 30 s, 60°C 45 s, 72°C 45 s, followed by 72°C 5 min.

For the nuclear gene !-crystallin (CRYBA1), I was unsuccessful in amplifying a partial gene region, using the primers CRYB1Ls (5’-CGCCTGATGTCTTTCCGCC-3’) and CRYB2Ls (5’-CCAATGAAGTTCTCTTTCTCAA-3’) (Dolman and Phillips 2004).

Appendix 3.3. Structure results using alternative prior information. I tested geographic cluster and COI haplotype as prior information in my initial Structure analyses. The resulting output was ambiguous with regard to K and indicated weak population structure that did not seem biologically likely.

1. Structure results using 20 COI haplotype groups as prior information. Line graph of Delta K indicates 2 populations most likely (K=2 is the highest value). The 2 populations are represented as green and blue in the bar plot, which has been ordered into the 20 haplotype groups.

72 ! Chapter 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

2. Structure results using 12 geographic clusters as prior information. Line graph of Delta K indicates 3 or 10 populations most likely (K=3 and K=10 are the highest values). These populations are represented as different colours in the bar plots below. These bar plots have been ordered into the 12 geographic clusters, as indicated by labels A-L (refer to Fig. 3.3 for location of clusters A-L around Melbourne).

! 73 Chapter 3

E F D J L K C A B I H G

E F D J L K C A B I H G

Appendix 3.4. Haplotype network for 665bp of the ND4 gene.

Cardinia Wyndham Hume-Whittlesea Melton

74 ! Chapter 3

Appendix 3.5. Haplotype network for 1160bp of COI and ND4 genes concatenated.

Appendix 3.6. Genetic variability of mitochondrial DNA (concatenated regions of COI and ND4).

mtDNA attribute Population I Population II Number of samples (N) 82 30 Length (bp) 1160 1160 No. haplotypes 20 13 Haplotype diversity (h) 0.900 (var=0.00030, sd=0.017) 0.933 (var=0.00037, sd=0.019) Nucleotide diversity (!) 0.00376 0.00631 Average no. nucleotide differences (k) 4.357 (obs var=14.0759, CV=0.8638) 7.317 (obs var=15.4245, CV=0.5412) No. polymorphic sites (S) 42 29 Total no. mutations (Eta) 43 30 Tajima’s test statistic (D) -1.59282 (not stat sig,0.1> P>0.05) -0.12184 (not stat sig, P>0.1) Fu’s test statistic (Fs) -3.62 -0.132 Raggedness index (r) 0.0138 0.0187

!

! 75 Chapter 3

76 ! Chapter 4

CHAPTER 4 Impacts of urbanisation on the population genetic structure of a threatened amphibian, Litoria raniformis

! 77 Chapter 4

ABSTRACT

Habitat loss and fragmentation in urban environments can lower genetic diversity and increase genetic differentiation between populations, leading to inbreeding and the erosion of genetic viability. Landscape genetic models have enormous potential for conservation management in urban landscapes. However, past approaches to fitting these models were generally not appropriate for quantifying the association between genetic and geographic distance, given the lack of independence between pairwise data points. A recent alternative – mixed effect models – account for this non-independence. I applied such a model (the maximum-likelihood population-effects (MLPE) approach) in a comprehensive landscape genetics study of the threatened Growling Grass Frog (Litoria raniformis) in peri-urban Melbourne, Australia. Geographic distance between sites had a strong negative effect on gene flow, as did the extent of urban infrastructure between sites. Stream connectivity had a positive, but small effect on gene flow. This statistical approach delivers a model of gene flow with direct application to the management of a range of threatened species in urban landscapes.

78 ! Chapter 4

INTRODUCTION

Globally, urbanisation is an important cause of landscape fragmentation and habitat loss (Irwin and Bockstael 2007, Noel and Lapointe 2010). Habitat fragmentation is known to have important effects on the genetic structure of populations (Noel et al. 2007), with numerous studies having demonstrated a direct link between fragmentation in urban environments and lowered genetic diversity (Gray 1995, McClenaghan and Truesdale 2002, Vandergast et al. 2007). Furthermore, the genetic consequences of habitat fragmentation can also include lower effective population sizes, potentially leading to inbreeding and genetic drift (Rivera-Ortíz et al. 2014). For ground-dwelling species, habitat isolation often involves a combination of both the distance between habitat fragments and the landscape resistance between these patches (Vos and Chardon 1998). Additionally, barriers to dispersal such as roads and infrastructure (e.g., buildings and fences) can inhibit migration rates and genetic connectivity between small, remnant populations in these urban environments.

Although landscape genetic models have enormous potential for conservation management in urban landscapes (Storfer et al. 2007), there have been important limitations to the methods most commonly used. In the past, many landscape genetic studies relied on Mantel or partial Mantel tests, which test for correlation between two distance matrices, to compare hypotheses of landscape permeability and landscape effects on gene flow (Balkenhol et al. 2009, Manel and Holderegger 2013). Hypotheses are usually based on single or multiple least-cost paths, including those derived from circuit theory or graph theory (McRae and Beier 2007, Garroway et al. 2008). However, these tests estimate correlative strength between genetic distance and explanatory variables, rather than estimating the effect of the variable on genetic distance. Additionally, there is ongoing debate about the statistical validity of Mantel tests, with concerns regarding type I error rates and low power (Guillot and Rousset 2013). Linear regression models have also been used (Holderegger and Wagner 2008, Storfer et al. 2010), but approaches to fitting these models have generally not been appropriate for analyses that test the association between genetic and geographic distance, given the lack of independence between pairwise data points (Storfer et al. 2010).

! 79 Chapter 4

The variety of different analytical options is one of the most challenging aspects of landscape genetics and there is yet to be a concensus on which methods to use (Balkenhol et al. 2016b). To date, there has been considerable disagreement on the best anaytical methods for evaluating relationships between landscape structure and gene flow and many studies have relied on unvalidated expert opinion in the absence of empirical data on gene flow (Waits et al. 2016). Methods that are being implemented to deal with the non-independence between pairwise data include redundancy analysis (RDA; the multivariate analog of simple linear regression), generalised dissimilarity modelling (GDM; for non-linear responses of genetic data to the landscape), and simple corrections to degrees of freedom in multiple regression methods such as least-cost paths, from the number of connections among demes, to the number of demes (Hall and Beissinger 2014, Balkenhol et al. 2016a). Other approaches that attempt to account for non-independence include the use of a Delaunay triangulation (Goldberg and Waits 2010) and a modified bootstrap procedure (without replacement) in which random sampling is constrained to maintain independence (Worthington Wilmer et al. 2008, Dudaniec et al. 2013). These methods account for errors associated with Mantel and partial Mantel tests’ inability to distinguish between variables that influence genetic structure and other correlated distances (Balkenhol et al. 2009). However, these approaches estimate the uncertainty around the parameters by reducing the size of the data set.

Recent alternatives, which account for non-independence between pairwise data points, include mixed-effect models with an appropriate covariance structure of allele frequencies, such as the maximum-likelihood population-effects model (Manel and Holderegger 2013). Using an empirical genetic dataset of the southern damselfly (Coenagrion mercuriale), Van Strien et al. (2012) used a covariate structure where a proportion of the total variance is the result of the correlation between two pairwise distances involving a common deme. This approach accounts for the non-independence of values in distance matrices, using the full data set, and has found that R2 statistics can be appropriate when selecting the model of best fit (see Edwards et al. 2008, Van Strien et al. 2012). However, such an approach has not yet been applied across a broad range of organisms.

80 ! Chapter 4

Here, I investigated the landscape genetics of the endangered Growling Grass Frog (Litoria raniformis) in an urbanising landscape, and applied the maximum-likelihood population-effects (MLPE) approach to develop predictive models of the landscape drivers of genetic distance for this species. In Melbourne, Australia, remaining populations of L. raniformis have become geographically isolated from one another and are threatened by urban expansion (Heard et al. 2010). Based on previous research (Heard et al. 2012a, b, Heard et al. 2014), I predicted that geographic distance and urban infrastructure would increase genetic distance (Hale et al. 2013). I also hypothesised that stream connectivity between sites would facilitate gene flow for L. raniformis, given its highly aquatic nature and the perceived importance of watercourses as dispersal corridors for the species based on past mark-recapture studies (Hamer and Organ 2008, Heard et al. 2012a).

METHODS

Study area and field sampling

The study species, Litoria raniformis, is a semi-aquatic frog in the family Hylidae. L. raniformis was abundant across much of south-eastern Australia (Pyke 2002), however, its distribution and numbers have declined significantly since the late 1970s (Mahony 1999). In Victoria, L. raniformis is listed as endangered under the State Government’s Advisory List of Threatened Vertebrate Fauna in Victoria – 2007 (DSE 2013).

My study was conducted in urban and semi-urban regions in the northern spread of Melbourne in south-eastern Australia. The Melbourne Metropolitan Area (MMA) currently covers approximately 7500 km2 and has a population of approximately 4.25 million people (State of Victoria 2013). The urban growth boundary will rapidly increase over the next 20 years, to expand the MMA by an additional 400 km2 (DPCD 2009). The study region encompassed a 20 km stretch of the Merri Creek system (including the Merri, Darebin and Moonee Ponds Creeks) ranging from highly urbanised sites surrounded by domestic housing and industry to semi-rural open grazing

! 81 Chapter 4

land (Fig. 4.1). This region displays undulating topography at low elevations, ranging from ~ 65 to 270 m a. s. l.

Samples were collected from sites including slow-flowing pools along creeks, farm dams, flooded quarry pits and swamps. Populations of L. raniformis around Melbourne exhibit ‘classical metapopulation dynamics’, whereby discrete populations are connected by limited migration, and exhibit frequent population extinction and recolonisation (Heard et al. 2012a, 2013, 2015). Three hundred and eighty eight L. raniformis were captured at 26 sites during spotlight surveys conducted between 12 January 2012 and 26 February 2013. These sites grouped together into nine geographic clusters (Fig. 4.1b). I obtained genetic samples from each individual by clipping a triangular section of toe webbing (~ 2 mm at the base) between the second and third toes closest to the body on the left hind limb (see Chapter 2 for justification of this sampling method for L. raniformis). Toe web samples were stored in 95% ethanol and kept at -18°C for approximately two weeks, then -80°C for up to 3.5 months. Latex gloves were worn when taking samples from frogs and replaced between individuals. Tissue sampling equipment was also sterilised in 70% ethanol between individuals.

Additionally, three clusters of populations were sampled across the Merri Creek catchment between October and March of 2004/2005 and 2005/2006 (Hale et al. 2013), as part of a concomitant study on the metapopulation dynamics of L. raniformis in this region (Heard et al. 2012a) (Fig. 4.1b). The 198 additional samples collected as part of this previous study from 12 sites in the study area had been previously genotyped. PCR amplification was performed on a subset of samples from this previous study and all samples from the previous study were re-scored to ensure comparability with sample scores from the current study. Sampling methods for this study are detailed in Hale et al. (2013). Two of the sites from this original study were resampled in the present study due to low sample sizes.

82 ! Chapter 4

¯

! ! !! ! ! !

Merri Creek

!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!! !!!!!!!!!!!!!!!

!!! Moonee Ponds Creek !!!!!! ! !!!!!!!!!!!!!!!!!!!!!!!!!!!!! Darebin Creek !!!!!!! ! !! !!! !!!!!!!!!!!!! ! ! !!!

!!!!!!!!!!!!!! !!!!!!!! ! ! !!!!!!! !!!!! ! !!!!!! ! !!!!!!!!!! !!!!!!!!!!!!!!! ! !!!!!!!!! !!! !!!!!! ! !!!!! !!!!!!! !!!

Port Phillip Bay

km 036912 (a)

km ¯ 02.557.510

! ! ! 9 ! !! !

Darebin Creek

!!! ! 8 !!!!!!!!!!! !!!!!!! !!!!!!!!!!!! Merri Creek

6 !!!! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! ! 7 !!!!! ! !!! !! !!!

! !!!!!!! ! ! 5 !!!

!!!!!!!!! !!!!! ! 2 !!!!!!!! 4 ! !!!!!!! ! ! ! !!!!!! !!!!!!!!!!!!!!! ! 1 !!!!!!!!!!!! !!!!!!!! 3 !!!!! !!! !

Moonee Ponds Creek

(b) Figure 4.1. Populations of Litoria raniformis sampled north of Melbourne. Grey = current Melbourne metropolitan area. (a) All sample sites; (b) Sample sites included in population assignment analyses, grouped into circled clusters. Small, closed circles = sample sites from current study, small open circles = sample sites from previous study.

!

! 83 Chapter 4

Microsatellite analysis

Genomic DNA was extracted from toe and web tissue samples using a DNeasy Blood & Tissue Kit (QIAGEN, South Korea), using the manufacturer’s animal tissue protocol. Eleven polymorphic microsatellite markers were used: nine previously developed for L. raniformis from this study system (Lr1, Lr2, Lr3, Lr4, Lr5, Lr6, Lr7, Lr8 and Lr9; Hale et al. 2011) and two previously developed for the sister species, Litoria aurea (Laurea2A, Laurea5M; Burns and Ferrari 2004). Two additional microsatellite markers previously developed for L. aurea were trialled (Laurea4-49, Laurea4-10; Burns and Ferrari 2004); however, they proved monomorphic for L. raniformis and therefore were not included. Six of these markers were multiplexed for PCR amplification (three sets of two markers: Lr1 with Lr8, Lr2 with Lr7, and Lr6 with Lr9). Five markers incorporated a GTTTCTT ‘pigtail’ added to 5’ end of the reverse primer to reduce variation in stutter (Lr2, Lr3 Lr6, Lr8 and Lr9). See Hale et al. (2011) for methods for fluorescently labeling fragments for all loci.

The PCR was performed in a 10 µl total volume with 1 µl of DNA template, 5 µl GoTaq

Hot Start Polymerase (Promega, USA) (25 mM MgCl2, GoTaq Hot Start Polymerase, 5X Colourless GoTaq Flexi Buffer, 5X Green GoTaq Flexi Buffer), 0.5 µM reverse primer, 0.15 µM forward primer, 0.25 µM fluorescently labeled 454A primer and 3.1 µl

dH2O. For Lr1, Lr2, Lr3, Lr4, Lr5, Lr6, Lr7, Lr8 and Lr9, amplification involved 95°C 2 min, 42 cycles of 95°C 30 s, 50°C 45 s, 72°C 45 s, followed by 72°C 5 min (adapted from Hale et al. 2011). For Laurea2A, the amplification protocol was adapted from Hale (2010), with an initial denaturation of 95°C for 2 min. For Laurea5M, amplification involved 95°C 2 min, 40 cycles of 95°C 20 s, 58°C 60 s, 72°C 45 s, followed by 72°C 5 min (adapted from Burns 2004). I included one negative control per PCR to check for contamination and two positive controls of L. raniformis DNA to standardise genotype readings. Fragment analysis of PCR products were carried out by Macrogen on an Applied Biosystems ABI3730XL DNA analyser using a LIZ-500 size standard. Scoring was completed using the Microsatellite Plugin v 1.2 in Geneious Pro v 5.6 (Biomatters), and all samples were screened manually for accuracy.

84 Chapter 4

Genetic structuring

Micro-Checker v. 2.2.3 (van Oosterhout et al. 2004) was used to test for the presence of null alleles, large allele dropout and scoring error due to stuttering. Expected and observed heterozygosities were estimated in Genepop v. 4.2 (Rousset 2008) with default settings, to assess for linkage disequilibrium and deviations from Hardy-Weinberg equilibrium (HWE) across the eleven microsatellite loci for each of the nine clusters of sites. I corrected significance values using sequential Bonferroni adjustments, and calculated allelic richness of each loci for each cluster, using HP-Rare v. 1.0 (Kalinowski 2005). Sites with small sample sizes (" 4 samples) were excluded from further analyses.

Genetic structure was assessed using the Bayesian clustering program Geneland v. 4.0.4 (Guillot et al. 2005a, Guillot et al. 2005b) through R v. 3.0.3 (R Core Team 2014). Geneland uses MCMC sampling to detect population structure by clustering individuals based on their genotypes and can also incorporate the geographical location of samples. Geneland analyses were completed over the entire data set and for each sampling site cluster separately. Over the entire data set, I performed ten independent runs with 100,000 MCMC iterations (thinning data at every 100th iteration), allowing K to vary from 1 to 20. The maximum number of nuclei was set to the recommended value of roughly 3 # sample size (= 1725) (Guillot et al. 2005a). Uncorrelated allele frequency models were used, as they are less sensitive than correlated allele frequencies to departures from model assumptions, such as isolation by distance (IBD) (Guillot 2012). The error for spatial coordinates was set to 10 m, while other parameters were set to default values. Outputs were post-processed with a burn-in length of 200 iterations. Results (number of genetic units, probability of membership and posterior densities) were checked for consistency across the ten runs, and plots were generated based on the run with the highest posterior probability. I then assessed the genetic structuring within each of the separate site clusters, allowing K to vary from 1 to 10 and the maximum number of nuclei was set to the default value of 300. Other parameters were set as above. The number of inferred populations from the Geneland program is assumed to be more accurate than a Structure analysis, given that Geneland flexibly simulates data displaying IBD patterns (which were clearly evident in this study, see results and Hale et al. (2013)) (Guillot and Santos 2009).

! 85 Chapter 4

To estimate rates of recent migration among populations, I used the Bayesian framework implemented in BayesAss v. 3.03 (Wilson and Rannala 2003). Individuals were assigned to the populations identified by Geneland and the MCMC was run for 5 x 106 iterations, including a burn-in of 1 x 106 iterations, sampled every 2000. Mixing parameters were set to 0.05 for migration rates, so that acceptance rates for proposed changes fell between the optimal range of 20% and 60% (Rannala 2012). I ran the program 5 times using different random seed values, and compared the posterior mean parameters to identify convergence between runs. Estimates of migration rate (m) were taken from the run with the narrowest log-likelihood distribution and populations were considered to be demographically independent when m < 0.1 (Faubet et al. 2007). I calculated 95% credible intervals (CIs) using Tracer v. 1.6 (Rambaut et al. 2003-2014).

Modelling genetic distance

I used three variables in this model as likely determinants of genetic distance between populations: the minimum edge-to-edge Euclidian distance between sites, the proportion of this distance that was urbanised and the connectivity of sites by stream or mapped drainage lines. Additional variables, including elevation, climate and vegetation type, have been important in previous studies of amphibians (Spear et al. 2005, Murphy et al. 2010, Storfer et al. 2010). However these were not included in the analyses, due to the small and relatively flat nature of the study area (~280 sq. km, 65 – 270 m a. s. l.), and minor variations in vegetation, which is primarily grassland, grazed land and open woodland (Bull and Stolfo 2014).

Pairwise FST between populations were estimated using GenAlex v. 6.501 (Peakall and

Smouse 2006). Pairwise genetic distance was calculated as FST / 1 $ FST (Rousset 1997). Pairwise Euclidian distances were calculated using site maps developed from aerial photographs of Melbourne. All urban infrastructure (buildings, roads, and other impervious surfaces) was also mapped in the study area using this aerial imagery, and the proportion of the shortest path between urbanised sites was then calculated. Three binary variables describing stream connectivity were generated for this study using a 2005 hydrology layer available from the Victorian Department of Environment, Land, Water and Planning (DELWP). I first coded sites as connected (1) if they were pools along the same stream, or unconnected (0) otherwise. Secondly, sites were defined as

86 ! Chapter 4 connected if they lay along the same stream or were connected by a mapped drainage line. Thirdly, sites were defined as connected if they met the two criteria above, and/or were within a 200 m riparian corridor along either side of such watercourses. Each landscape variable was derived with the aid of ArcGIS v. 10.2 (ESRI, Redlands, California).

I modelled the relationship between genetic distance (loge transformed) and the above predictors using the MLPE method advocated by Van Strien et al. (2012), using the MCMCglmm package v. 2.19 (Hadfield 2010) for R v. 3.0.3 (R Core Team 2014) to account for non-independence of the pairwise distances. Three global models were fitted to the data; one for each of the three measures of stream connectivity. Pairwise

Euclidian distance (loge transformed) and proportion of urban infrastructure were included in all models a priori, as both have been demonstrated as important for the dispersal of L raniformis (Heard et al. 2012a, Hale et al. 2013). Hence, in each case, genetic distance (GD) between sites i and j was modelled as:

loge(GDij + 0.01) = ! + "1 # loge(Dij) + "2 # Uij + "3 # Sij Eq. 1 where ! is the intercept, "’s are regression coefficients, D is geographic distance between sites, U is the proportion of urban infrastructure along the shortest distance path between sites and S is stream connectivity.

MCMCglmm uses a random effects structure where correlation between all sites is drawn from a common distribution. I generated 100,000 MCMC samples, after discarding the initial 10,000 iterations as a ‘burn in’ and thinned the chain by keeping every 10th iteration. Inferences were therefore drawn from 10,000 samples from the joint posterior distribution of the parameters.

The model was refined and tested post hoc to account for several potential sources of error. Firstly, I removed the most southern site from cluster three (see Fig. 4.1b), as it proved a significant outlier in the relationship between geographic and genetic distance. As this site was a fauna reserve it may have been established via an artificial translocation event rather than natural processes – such translocations have been an

! 87 Chapter 4

ongoing management strategy for this species (Clemann and Gillespie 2012). Secondly, as samples included in the model were collected during two different time periods (either 2004/2006 or 2012/2013), I used the time of collection as an additional variable, to test whether this had an important effect on pairwise genetic distance.

The fit of the models to the data were assessed using the deviance information criterion (DIC; (Spiegelhalter et al. 2002)) and model R2 (Nakagawa and Schielzeth 2013). Both continuous explanatory variables were z-transformed by subtracting the mean and dividing by two standard deviations, allowing the regression coefficients to be directly compared as multiplicative effect sizes (Gelman and Hill 2007). The relative importance of the three explanatory variables was assessed by calculating the multiplicative effect (including 95% credible interval) of each variable of genetic distance, across the range of the variable. The multiplicative effect E of variable i is calculated as:

Ei = exp (!i # rangei) – 0.01 Eq. 2

where ! is the regression coefficient and range is the range of the variable observed during the study. The size of the multiplicative effect is an indication of how much the genetic distance between sites changes with each explanatory variable. For example, an effect size of 0.1 corresponds to a predicted 10-fold decrease in genetic distance as a variable changes from its smallest to highest value. A multiplicative effect of 1 corresponds to no change in genetic distance (i.e. no effect), therefore an explanatory variable with a multiplicative effect considerably different to 1 is likely to have a biologically important effect on the genetic distance between sites (Parris 2006). The 95% credible interval of the effect size includes the range of plausible values for the effect of an explanatory variable. If working within the null hypothesis significance testing framework, a credible interval that does not encompass 1 would indicate that a null hypothesis of no effect could be rejected at alpha = 0.05 (Cumming and Finch 2001).

88 Chapter 4

RESULTS

Microsatellite genotyping

There was no evidence of large allele dropout. The majority of populations and pairs of loci showed no signs of linkage disequilibrium after sequential Bonferroni correction. Linkage disequilibrium was significant for 14 of the possible 495 combinations; however, this was not consistent across populations or loci. One locus for one cluster (Lr4 in cluster eight) showed potential scoring errors due to stuttering. Overall, each locus conformed to HWE across all clusters (Table 4.1); however, cluster eight did show significant heterozygote deficit, indicative of inbreeding. The possible presence of null alleles was observed in cluster one for Lr4, cluster five for Lr6, cluster seven for Laurea2A and cluster eight for Lr3 and Lr4. I chose to retain all loci for subsequent analyses, as given these inconsistent patterns across loci and genetic clusters, it is unlikely that null alleles are an important factor. Estimates of allelic richness ranged from 3.230 (cluster 3) – 4.400 (cluster 9) and had observed heterozygosity ranging from 0.541 (cluster 3) – 0.723 (cluster 9) (Table 4.1). ! Table 4.1. Summary microsatellite statistics across 11 loci and 9 populations. AR = mean allelic richness, HO = Observed heterozygosity and HE = Expected heterozygosity

Population Sample size AR HO HE 1 91 3.630 0.654 0.652 2 52 4.010 0.645 0.644 3 22 3.230 0.541 0.564 4 62 3.830 0.650 0.621 5 32 3.560 0.642 0.649 6 100 4.310 0.700 0.699 7 39 3.410 0.632 0.626 8 89 4.200 0.635 0.677 9 92 4.400 0.723 0.701

! 89 Chapter 4

Genetic structuring

The Geneland analysis revealed nine distinct genetic units when incorporating the entire data set (Fig. 4.2), with each of these units corresponding to the nine geographic sampling clusters. When within-cluster genetic structure was examined, clusters one, four, five and eight did not display internal genetic sub-division. Two genetic units from each of clusters two, three, six and nine were identified (Fig. 4.3a, b, c & e), and three genetic units were identified from cluster seven (Fig. 4.3d). The division of two of the sub-units of cluster seven correlates with a potential barrier to dispersal, as a road separates the site horizontally into two in-stream ‘ponds’ connected by culverts (Koehler et al. 2014a). However, this population was the result of the salvage and translocation of 156 frogs during 2010/11 (Koehler et al. 2014a), therefore any genetic sub-divisions are likely to be an artifact of structuring at the original site, prior to translocation.

Estimation of recent migration rates using BayesAss showed that three of the populations identified by Geneland (corresponding to the site clusters selected a priori) were independent (Table 4.2). They were populations (or clusters) 1, 8 and 9. However, recent migration rates were estimated to be very low for all population pairs, with 95% credible intervals typically having lower bounds near zero. These results are concordant with one of the basic tenets of metapopulation dynamics, whereby discrete populations are connected by limited migration (Heard et al. 2012a), with populations separated by > 1 km sharing few migrants.

Modelling genetic distance

Model selection using deviance information criterion (DIC) values did not identify one model of genetic distance that was clearly superior. Results of all three models were qualitatively similar (Table 4.3), however the simplest, and the model with the lowest DIC (191.468) included stream connectivity as measured by sites connected within a named stream (with no drainage lines or buffer zone). Estimates of the regression coefficients from the top model, as obtained using MCMC sampling, are presented in Table 4.4. The model had an R2 of 0.667 after accounting for pairwise dependencies. The pairwise dependency had a very weak influence on the results (mean correlation =

90 ! Chapter 4

0.021, 2.5% CI = 0.008, 97.5% CI = 0.035). Explanatory variables were only weakly correlated with each other (r = -0.178 - 0.270). The additional model, which included a time variable, had an indistinguishable DIC of difference < 2, (see Burnham and Anderson 2002). I therefore chose not to select this model, given the time difference between sample collection periods was relatively short and I didn’t expect a priori that this would have a significant influence on results.

9

8

7 6

5

2 4 1 3 5825000 5835000 5845000 316000 324000 Figure 4.2. Population structure of Litoria raniformis north of Melbourne. Bayesian population assignment analysis using GENELAND. Black dots represent sampling locations Estimatedfrom each of the nine clusters. cluster membersh

91 Chapter 4

M ap of posterior probabilityCluster Two to belong to cluste y coordinates 5826600 5827200 320000 320500 321000 321500 (a) x coordinates Cluster Three Cluster Six Map of posterior probability to belong to cluste Map of posterior probability to belong to cluste y coordinates y coordinates 5824700 5825000 5832500 5833500 5834500 326800 327000 327200 327400 327600 318500 319500 320500 321500 (b) x coordinates (c) x coordinates Cluster Seven Cluster Nine Map of posterior probability to belong to cluste Map of posterior probability to belong to cluste y coordinates y coordinates

3237205834640 5834670 5834700 323760 323800 318500 319500 320500 5842000 5843000 (d) x coordinates (e) x coordinates Figure 4.3. Population structure of Litoria raniformis north of Melbourne. Bayesian population assignment analysis using GENELAND. Black dots represent sampling locations. The red and yellow colours represent genetic units, and the contours the posterior probability of belonging to the focal genetic unit. (a) Cluster two; (b) cluster three; (c) cluster six; (d) cluster seven; and (e) cluster nine. (Clusters that did not display genetic sub-division are not presented).

92 Table 4.2. Estimates of the proportion of individuals in each population that can be assigned as migrants from another population using !" BayesAss. Immigrant source populations (SOURCE) are listed in the left-hand column, receiving populations (INTO) are listed across the #" top row. Numbers in parentheses represent the 95% credible intervals. Bold values are the proportion of frogs assigned to the site of $" capture and are therefore non-migrant individuals. %"

INTO

SOURCE 1 2 3 4 5 6 7 8 9

1 0.946 (0.920, 0.971) 0.006 (0, 0.019) 0.046 (0.008, 0.087) 0.018 (0.002, 0.038) 0.018 (0.001, 0.043) 0.009 (0, 0.0120) 0.008 (0, 0.025) 0.005 (0, 0.015) 0.004 (0, 0.012)

2 0.005 (0, 0.014) 0.934 (0.893, 0.970) 0.011 (0, 0.032) 0.011 (0, 0.028) 0.017 (0, 0.040) 0.005 (0, 0.013) 0.008 (0, 0.023) 0.006 (0, 0.015) 0.004 (0, 0.013) 3 0.005 (0, 0.015) 0.010 (0, 0.027) 0.848 (0.786, 0.909) 0.007 (0, 0.021) 0.008 (0, 0.025) 0.020 (0.003, 0.040) 0.015 (0, 0.036) 0.004 (0, 0.012) 0.007 (0, 0.018)

4 0.014 (0.002, 0.029) 0.007 (0, 0.021) 0.016 (0, 0.046) 0.923 (0.885, 0.960) 0.017 (0.001, 0.039) 0.007 (0, 0.018) 0.009 (0, 0.027) 0.006 (0, 0.016) 0.005 (0, 0.014) 5 0.004 (0, 0.012) 0.007 (0, 0.020) 0.019 (0, 0.048) 0.012 (0, 0.027) 0.903 (0.855, 0.950) 0.011 (0, 0.025) 0.007 (0, 0.021) 0.012 (0.001, 0.028) 0.006 (0, 0.016) 6 0.013 (0.001, 0.027) 0.017 (0.001, 0.039) 0.024 (0.001, 0.053) 0.007 (0, 0.020) 0.011 (0, 0.032) 0.913 (0.880, 0.946) 0.008 (0, 0.024) 0.006 (0, 0.017) 0.014 (0, 0.032)

7 0.005 (0, 0.014) 0.007 (0, 0.022) 0.014 (0, 0.043) 0.008 (0, 0.025) 0.008 (0, 0.025) 0.009 (0, 0.023) 0.929 (0.887, 0.968) 0.007 (0, 0.020) 0.005 (0, 0.014) Chapter 4

8 0.005 (0, 0.014) 0.007 (0, 0.021) 0.012 (0, 0.034) 0.007 (0, 0.021) 0.009 (80, 0.027) 0.005 (0, 0.014) 0.008 (0, 0.025) 0.949 (0.918, 0.977) 0.007 (0, 0.018)

9 0.004 (0, 0.012) 0.006 (0, 0.017) 0.011 (0, 0.033) 0.007 (0, 0.020) 0.009 (0, 0.027) 0.022 (0.007, 0.041) 0.007 (0, 0.021) 0.006 (0, 0.017) 0.949 (0.918, 0.975) 93 Chapter 4

Table 4.3. Variables, deviance information criterion (DIC) results and weight for models of genetic distance.

Variables DIC Model weight

Geog dist, Urban, Stream 191.468 0.39

Geog dist, Urban, Stream or drain 192.465 0.24

Geog dist, Urban, Stream or drain 200 m 191.625 0.37

Geog dist = geographic distance (loge transformed); Urban = proportion of urban infrastructure; Stream = connectivity of sites by a named stream; Stream or drain = connectivity of sites by a named stream or mapped drainage line; Stream or drain 200 m = connectivity of sites by a named stream or mapped drainage line and/or within a 200 m riparian corridor along either side of such watercourses.

Table 4.4. Coefficient estimates (mean and 95% credible interval) for the three explanatory variables included in the best regression model.

Variable Mean 2.5% 97.5%

Intercept -2.387 -2.496 -2.267

Geog dist 0.693 0.623 0.761

Urban 0.469 0.378 0.562

Stream -0.131 -0.330 0.069

Geog dist = geographic distance (loge transformed); Urban = proportion of urban infrastructure; Stream = connectivity of sites by a named stream.

The regression modelling provided strong support for an important positive effect of geographic distance on genetic distance (Fig. 4.4 & 4.5), which was predicted to be 13.778 times greater between the two furthest sites than between the two closest sites (range = 9 m – 19,612 m), with all other explanatory variables held constant. There was also strong support for an important positive effect of proportion of urban infrastructure on genetic distance (Fig. 4.4 & 4.5). Genetic distance was predicted to increase by 2.110 times across the observed range of this variable (range = 0 – 0.962). Stream connectivity had a negative effect on genetic distance (Fig. 4.4 & 4.5). Genetic distance was predicted to be 1.151 times lower between two sites that were connected than two sites that were not, with all other explanatory variables held constant (as this was a

94 Chapter 4 binary variable, range = 0 – 1). The 95% credible intervals of the effect size did not encompass 1 for geographic distance and proportion of urban infrastructure, however they did encompass 1 for stream connectivity (Fig. 4.5).

(a)

(b)

(c) Figure 4.4. Relationships between genetic distance and (a) geographic distance, (b) the proportion of urban infrastructure, and (c) stream connectivity. For the continuous variables, the solid lines represent the mean relationship and the broken lines the 95% credible intervals. For the binary variable (stream connectivity), mean estimates are represented by circles and vertical lines represent the 95% credible intervals.

95 Chapter 4

'&$

'#"#$

'%$

'!"#$

&$ Effect size Effect

#"#$

%$

!"#$ Geog. Dist. Urban Stream Variable Figure 4.5. The estimated multiplicative effect on genetic distance of the three explanatory variables. Mean estimates are represented by circles and vertical lines represent the 95% credible intervals.

DISCUSSION

Landscape genetic models have enormous potential for conservation management in urban landscapes. Relationships between genetic distance and potential explanatory variables are often tested; however, common approaches only provide estimates of correlative strength, rather than estimating the effect of the explanatory variable on genetic distance. The approach demonstrated in this study addresses the non- independence of variables and avoids problems of statistical validity associated with Mantel and partial Mantel tests (Guillot and Rousset 2013) while providing estimates of the effect of landscape elements on gene flow.

Understanding landscape permeability may provide opportunities for reducing isolation, and therefore the extinction risk of fragmented populations (Ricketts 2001). The inclusion of genetic information in management decisions for fragmented urban populations can also facilitate planning for population translocations and augmentation, including the maintenance of corridors and addition of road crossing structures (Bos and

96 Chapter 4

Sites 2001, Balkenhol and Waits 2009). However, it appears that only a few landscape genetic studies to date have then been applied to practical conservation management (Keller et al. 2015). One example is the Bighorn Sheep (Ovis canadensis nelsoni) in California, USA. After a study found highways were barriers to gene flow for the species, over- and underpasses were included in the species’ recovery plan (Epps et al. 2005, Keller et al. 2015). The lack of practical application to conservation management may partly be due to the increasingly complex statistical analyses often suggested to address some of the challenges associated with existing methods used in landscape genetics (Balkenhol et al. 2016b). Here, I instead built a simple linear model to describe the effects of geographic distance and urban barriers on dispersal rate, which proved to have high predictive power for the study species.

Landscape genetics studies have thus far been primarily descriptive and few have been hypothesis-driven (Waits et al. 2016). This study sought to assess the utility of maximum-likelihood population-effects models in conservation genetic studies in urbanising environments, incorporating three hypotheses: (1) gene flow would decrease with increased geographic distance; (2) urban infrastructure would act as a barrier to gene flow, and; (3) stream connectivity between sites would facilitate gene flow for L. raniformis. The results provided strong support for the first two predictions but, surprisingly, did not identify stream connectivity as an important predictor of genetic distance in this species.

In this study system, edge-to-edge Euclidian distance between wetlands had the largest effect on genetic distance, reflecting a strong positive interaction between genetic distance and geographic distance. This pattern is consistent with previous research on anuran migration, which suggests that adult frogs have limited dispersal (Faubet et al. 2007) and strong site fidelity (Marsh et al. 1999). Around Melbourne, L. raniformis has been found to exhibit infrequent, distance-limited dispersal and high site fidelity during mark recapture studies (Hamer and Organ 2008, Heard et al. 2012a). In fact, the great majority of movements observed by Heard et al. (2012a) during two consecutive breeding seasons occurred over distances less than 200 m. This model strengthens previous knowledge on the species, as it quantifies the relationship between Euclidian distance and genetic distance.

97 Chapter 4

The proportion of urban infrastructure between sites also had an important effect on gene flow for L. raniformis. My prediction that urban infrastructure would act as a barrier to gene flow was supported, as an increase in the proportion of urban infrastructure between sites corresponded to an increase in genetic distance. Importantly, the model also quantifies the effects of urbanisation on the dispersal of the species, rather than estimates based on a priori resistance measures (see Spear et al. 2010). Habitat fragmentation through the construction of infrastructure and the clearing of native vegetation in urban environments can inhibit or discourage amphibian dispersal, reducing the function of population networks (Hamer and McDonnell 2008). Studies have identified substantial genetic differentiation between amphibian populations and lower genetic diversity within populations due to urban barriers to dispersal (Noel and Lapointe 2010, Munshi-South et al. 2013). In addition to the presence of physical barriers in urban areas, amphibian dispersal can be limited by physiological and behavioural constraints, including evaporative water loss and thermoregulation (Blaustein et al. 1994). Furthermore, individuals of L. raniformis possess traits that make them particularly susceptible to urbanisation, including their reliance on water, with > 90% individuals occurring within 5 m of a water body (Heard 2010), and their limited climbing abilities (Pyke 2002).

Although rivers have been identified as barriers to gene flow for some amphibian species, they can also facilitate gene flow in others (Spear et al. 2005). I found that stream connectivity had a small negative effect on genetic distance for L. raniformis, which was contrary to my prediction that stream connectivity between sites would facilitate gene flow. Litoria raniformis is a semi-aquatic species that commonly occurs in close proximity to still or slow-flowing creeks and permanent waterbodies (Clemann and Gillespie 2010) and previous studies suggest the species uses permanent creeks and drainage channels as dispersal corridors (Heard et al. 2012a). Incorporating three different definitions of streams, including drainage lines and riparian buffer zones, into the model generated very similar results. I may have found no important effect of stream connectivity because L. raniformis displays high site fidelity, with limited overland dispersal occurring during particularly wet conditions (Heard and Scroggie 2009), rather than along permanent creeks and drainage channels.

98 ! Chapter 4

While this study was primarily focused on whether urban infrastructure was acting as a barrier to gene flow for the study species, additional landscape variables that either facilitate or hinder gene flow can be incorporated into linear regression models, including a further division of land-cover type. Future studies could assess the relative influence of particular types of urban infrastructure, including roads, housing estates, industrial estates, flat impervious surfaces and open spaces, on amphibian population genetic structure and gene flow. This could include the three urban ecological paradigms of ecology in, of, and for the city, where ecology in focuses on terrestrial and aquatic patches, ecology of incorporates biological, social and built components, and ecology for integrates the two and links these with civic processes (Pickett et al. 2016).

Conclusion

This statistical approach delivers a model of gene flow with direct application to the management of threatened species in urban landscapes. Its power lies in the fact that the model can be used for predictive purposes, including how gene flow is likely to change given future urbanisation scenarios and increases in geographic distance between populations due to habitat loss. In turn, the model can be used to investigate a range of management scenarios to mitigate impacts of urbanisation on population connectivity, including optimal locations for constructed wetlands based on minimising the genetic distance between sites. Landscape genetics studies can provide new insight and play an important role in conservation management (Keller et al. 2015). Predictive tools such as these are a vital resource for managing species afflicted by habitat loss and fragmentation (Hanski 1994).

99 Chapter 4

100 Chapter 5

CHAPTER 5 Integrating genetic connectivity measures with stochastic patch occupancy models for metapopulation management

101 Chapter 5

!"#$%!&$

Habitat loss and fragmentation represent pervasive impacts of urbanisation. Metapopulation models are an important tool for predicting the probability that species affected by habitat loss and fragmentation will persist, and for identifying optimal management regimes. All metapopulation models require some knowledge of the dispersal behaviours of the species in question, for which mark-recapture studies or radio-telemetry have traditionally been used. However, these approaches can be labour- intensive, expensive and are difficult to apply to many taxa. Population genetics provides various ways to measure dispersal and can be used to improve models of metapopulation viability by refining measures of population connectivity. In this novel study, I integrate a model of genetic connectivity and gene flow with a Bayesian stochastic patch occupancy model (SPOM), to assess the metapopulation viability of the endangered Growling Grass Frog (Litoria raniformis) in Melbourne, Australia, a rapidly urbanising landscape. By trialling different management scenarios of wetland loss and creation, I determined that arranging new sites in a ‘cluster’ formation provided a superior management scenario to the alternative ‘stepping stone’ fromation and also had the smallest range of uncertainty, therefore decreasing the estimated risk of local extinction of the species. The inclusion of dispersal barriers in the model, quantified using genetic data, provided higher estimates of the risk of metapopulation extinction. Implementation of this model in such an urban landscape demonstrates how genetic connectivity measures can improve the accuracy of quantitative models used to assess management options for species afflicted by habitat loss and fragmentation.

102 Chapter 5

INTRODUCTION

Habitat loss, fragmentation and isolation represent pervasive impacts of urbanisation for amphibians. Urbanisation is a key threatening process for these animals (Smallbone et al. 2011), with the IUCN Red List of Threatened Species categorising greater than one- third of the world’s listed amphibian species as currently threatened by urbanisation (IUCN 2015). The vulnerability of amphibians to habitat loss and fragmentation is multifaceted. Habitat loss increases local extinction risk and reduces colonisation rates (Cushman 2006). The effects of fragmentation appear amplified by relatively low vagilities, narrow habitat tolerances and low survival rates during dispersal between these habitats (Cushman 2006). Numerous studies have identified the preservation and connectivity of both aquatic and terrestrial habitats as necessary for the maintenance of local amphibian populations (Semlitsch 1998, Porej et al. 2004, Rothermel 2004, Becker et al. 2007).

In a fragmented urban landscape, species with the ability to disperse both short and long distances appear vulnerable, although in different ways (Cushman 2006). Species that can disperse over long distances generally will have an increased probability of encountering roads and other anthropogenic barriers. These barriers can disrupt genetic connectivity and dispersal and increase mortality rates in amphibian populations. For example, Carr and Fahrig (2001) found that populations of the vagile leopard frog (Rana pipiens) were more vulnerable to traffic volume and the density of roads in the landscape than those of the less vagile green frog (Rana clamitans), because the former encountered roads more frequently and therefore had higher rates of traffic mortality. In contrast, species with low vagilities may also be imperilled by habitat fragmentation over time, as once their populations become isolated by fragmentation, they may ultimately face extinction (Cushman 2006).

Metapopulation models are important tools for predicting the probability of species persistence in fragmented landscapes, and for informing conservation management decisions (Greenwald 2010a). Stochastic patch occupancy models (SPOMs) are useful for characterising metapopulations that persist in a stochastic equilibrium between local extinction and colonisation when all local populations have some risk of extinction through time (Moilanen and Hanski 1998). SPOMs model changes in the occupancy of

103 Chapter 5

habitat patches arising from this recurrent local population extinction and colonisation (Sjögren-Gulve and Hanski 2000, Hanski 2002). By implicitly modelling the relationship between metapopulation viability and patch distribution and quality, SPOMs provide a means of assessing different habitat management options for threatened species, including habitat protection, enhancement and creation (Dreschler et al. 2003, Westphal et al. 2003).

All metapopulation models require some knowledge of the dispersal behaviours of the species in question; specifically, the rate at which dispersal declines with distance and the effect of different landscape types or structures on dispersal rate (Hanski and Gilpin 1997). Traditionally, mark-recapture or radio-telemetry studies have been used to estimate these rates; however, such studies can be labour-intensive, expensive and difficult to apply to many species (e.g., due to small body size, low return rates or a short seasonal activity period) (Marsh and Trenham 2001, Fujiwara and Caswell 2002, Petit and Valiere 2006). Additionally, long-distance dispersal records are difficult to obtain and are often underestimated in mark-recapture or telemetry studies (Porter and Dooley 1993, Marsh and Trenham 2001). Population genetics provides a number of ways to measure dispersal; both directly, through the assignment of individuals to parents or populations of origin, and indirectly, through genotype frequency assessments (Broquet and Petit 2009). Genetic information can also provide a relatively quick estimate of population connectivity, when time does not allow for a long-term mark-recapture study (Greenwald 2010b); it also estimates effective dispersal (dispersal that leads to gene flow) rather than just the rate of migration (Broquet and Petit 2009).

Nevertheless, few landscape genetic studies have been directly integrated into models of metapopulation dynamics. Here, I integrate a model of genetic connectivity and gene flow with a Bayesian SPOM to assess the metapopulation viability of an endangered amphibian in a rapidly urbanising landscape. I then use this model to simulate the occupancy dynamics of L. raniformis within the northern region of Melbourne, for a period of 30 years assuming current conditions. Additionally, I test five scenarios of wetland loss and replacement, manipulating both the number of newly created wetlands to be added and the spatial pattern of these wetlands. These scenarios are based on realistic management strategies that may be implemented in this species in the future, providing an important contribution both to the ongoing management of L. raniformis

104 Chapter 5 and a poweful approach to improving the accuracy of SPOMs by integrating genetic information.

METHODS

Study species, study area and threatening processes

Litoria raniformis is a large, semi-aquatic frog distributed across south-eastern Australia (Pyke 2002). The species occupies slow-flowing pools along streams, as well as farm dams, flooded quarry pits, swamps and billabongs. Populations from temperate regions exhibit ‘classical metapopulation dynamics’, in which discrete populations connected by limited migration display frequent extinction and recolonisation (Heard et al. 2012a, 2013, 2015). Litoria raniformis is listed as endangered, having undergone significant range reductions over the past two decades (IUCN 2015). Around Melbourne, the rapid decline of L. raniformis may have resulted from metapopulation collapse in some regions, precipitated by the species’ susceptibility to chytridiomycosis (Heard et al. 2014, 2015) and drought (Heard et al. 2012a, b), but ultimately driven by habitat loss and fragmentation due to agricultural and urban expansion (Heard et al. 2013).

I conducted the study in the north of Melbourne, including the Darebin, Merri and Moonee Ponds Creek catchments (Fig. 5.1). This region displays undulating topography at low elevations, ranging from approximately 73 to 327 m a. s. l. This is one of the four main regions around Melbourne where remnant populations of L. raniformis can be found. Urbanisation has led to wetland loss, degradation and fragmentation in each of the regions where significant remnant populations of L. raniformis persist (Heard et al. 2010). This process will continue over the next three decades, with each of these regions designated as urban growth areas (DEPI 2013).

105 Chapter 5

km !! 01234 ! ¯ ! Legend ! Sites sampled this study Sites sampled previous study ! Sites ! Streams ! Current urban area Future urban growth area ! ! ! ! !

! !! !!

! !! !! !!!! ! !!!!! ! ! !! ! ! ! ! Merri Creek ! !! ! !! ! ! ! ! ! ! !! ! ! !! ! ! ! ! !! ! ! ! ! ! ! !! ! ! !! ! ! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! !!! ! ! !! ! !! !! ! !!! ! ! !! ! !! ! ! ! ! ! ! ! ! !!! ! ! ! ! !! Darebin Creek !! !! ! ! ! ! !! !! ! !! ! ! ! ! !! ! ! ! ! ! ! !! ! !! ! ! Moonee Ponds Creek !! !! !!

Figure 5.1. Monitoring sites for Litoria raniformis (closed circles) north of Melbourne, Victoria, Australia, in the Darebin, Merri and Moonee Ponds Creek catchments. Open circles indicate sites sampled during the current study and open squares indicate sites sampled during a previous study. The rectangle indicates the subset of sites used in the simulations of management effectiveness.

Existing Stochastic Patch Occupancy Model

Management options to offset the impacts of urbanisation on L. raniformis, including the creation of new wetlands, need to be compared and critiqued. Heard et al. (2013) developed a Bayesian regression-based SPOM for L. raniformis for this purpose, parameterised using occupancy data collected at 167 wetlands monitored across

106 Chapter 5 northern Melbourne between the 2001/2002 and 2006/2007 breeding seasons. In this model, the logit of the annual probability of population persistence is a linear, additive function of effective wetland area, aquatic vegetation cover and connectivity, whereas the logit of the annual probability of colonisation of vacant wetlands is a linear function of connectivity alone.

Effective area of each wetland (Aeffi) is defined as:

Aeffi = log(Ai) ! Hi ⁄ Hmax Eq. 1

where Ai is the surface area of each wetland i, Hi is wetland hydroperiod (on a scale between 1 and 4, where 1 refers to intermittent wetlands that fill during high-rainfall periods and 4 refers to permanent wetlands that never dry out), and Hmax is the maximum hydroperiod score (4). Aquatic vegetation cover for each wetland is defined as the mean cover of each of the main three strata of vegetation: emergent, submergent and floating forms. Taking the mean cover of these vegetation types provides an index of vegetation diversity as well as cover, because sites with a monoculture of one stratum are down-weighted.

Connectivity is defined as the distance-weighted number of surrounding populations within a 1000 m radius that could contribute migrants (Heard et al. 2013). Specifically, the connectivity of each wetland in each year (Si,t) is defined as:

Si,t = "wi,j ! oj,t-1 Eq. 2

where wi,j is a weighting function, and oj,t-1 is the occupancy status of each neighbour j in the preceding year (1 if the wetland is occupied, 0 if not). The weighting function, wi,j, defines a negative power relationship between the probability of dispersal from wetland j to i and their centre-to-centre distance (di,j, measured at 10 m increments):

-0.720 wi,j = (0.100!di,j) Eq. 3!

This model was updated by Rose et al. (2016) using a further 5 years of occupancy

107 Chapter 5

monitoring data (Heard et al. 2015). An additional effect of site type (pool along a stream or off-stream [lentic] wetland) was added to account for apparently higher rates of population extinction along streams (due primarily to flooding and resulting effects on survival rates (Heard et al. 2014)). Two further refinements to this model were pursued during the present study to assess the impacts of urbanisation on metapopulation viability; an effect of urbanisation of wetland buffers on the probability of persistence, and an effect of urban barriers on dispersal rate and site connectivity.

A model of genetic connectivity

In response to limitations of existing measures of connectivity for L. raniformis in the study area, I developed a model of genetic distance based on pairwise geographic distance and extent of urban barriers (detailed in Chapter 4). I derived the model from genetic sampling of 388 individuals of L. raniformis captured at 19 sites in the study area between the 2011/2012 and 2012/2013 breeding seasons (Fig. 5.1 and detailed in Chapter 4), plus a further 198 samples collected previously from 12 sites in the study area (Heard et al. 2012a, Hale et al. 2013; see Fig. 5.1).

I analysed eleven polymorphic microsatellite markers. Nine that were previously developed for L. raniformis (Lr1, Lr2, Lr3, Lr4, Lr5, Lr6, Lr7, Lr8 and Lr9; Hale et al. 2011) and two that were previously developed for the sister species, Litoria aurea (Laurea2A, Laurea5M; Burns and Ferrari 2004) (see Chapter 4 for details).

I estimated pairwise FST for each site using GenAlex v. 6.501 (Peakall and Smouse 2006a), and developed a linear model of pairwise genetic distance (GD = Fst / 1 - Fst) based on the fundamental constraints on dispersal in this system: pairwise geographic distance and urban barriers (Chapter 4). Pairwise geographic distance was defined as the minimum edge-to-edge Euclidian distance between sites (D) calculated (in meters) using aerial imagery and ArcGIS v. 10.2 (ESRI). Urban barriers (U) were represented by the proportion of this path that was covered by urban infrastructure (buildings, roads, and other impervious surfaces), as estimated from aerial imagery. The fitted model of genetic distance (Chapter 4) took the form:

108 Chapter 5

log (GD + 0.01) = -5.686 + 0.342 ! (log [D ]) + 0.803 ! U Eq. 4 e ij e ij ij

Model extensions and incorporation of genetic connectivity

I updated the existing SPOM for L. raniformis in the study area to incorporate genetic estimates of connectivity (genetic distance) and fitted it to the extended occupancy dataset described above, incorporating 2,011 surveys of 190 sites across northern Melbourne (Fig. 5.1) conducted over 11 seasons (2001/2002 - 2011/2012).

Genetic estimates of connectivity were incorporated into the model by replacing the weighting function (w) from Eq. 2 with one based on the above model of genetic distance (GD). Specifically, for each site pair i and j, genetic distance was calculated as:

GDij =exp(-5.686 + 0.342(loge[Dij]) + 0.803 !Uij) - 0.01 Eq. 5

Genetic distance was then linearly transformed to range from 1-0, matching the weighting function used by Heard et al. (2013). Hence, the weighting for each site pair i and j became:

wi,j = (0.129 – GDij)/0.129 Eq. 6 where 0.129 was the maximum estimated genetic distance between any pair of sites in the study area.

I fitted the model to the occupancy monitoring dataset using the Bayesian state-space approach of Royle and Kéry (2007), implemented with MCMC sampling in OpenBUGS v.3.2.3 (Thomas et al. 2006). This approach accounts for imperfect detection during surveys and missing surveys at some sites in some years, allowing propagation of the resulting uncertainty in site occupancy status through to the connectivity measure (see Yackulic et al. 2012, Heard et al. 2013, Eaton et al. 2014). As in Rose et al. (2016), I modelled the annual probability of population persistence as a function of wetland effective area, aquatic vegetation cover, site type and connectivity, while the annual probability of colonisation was modelled as a function of connectivity alone. However,

109 Chapter 5

to incorporate an effect of urbanisation of wetland buffers on population persistence, I down-weighted site area in a linear fashion by multiplying it by the proportion of the terrestrial buffer (defined as a 100 m radius around each site) that remained free of urban infrastructure. I used vague, normally distributed priors (N[0, 0.01]) for all model parameters. Convergence was assessed by visually inspecting the two Markov chains with over-dispersed initial values, and achieved within 30,000 MCMC samples. I extracted the next 5000 estimates of the parameters of the persistence and colonisation functions.

Simulating metapopulation trajectories under alternate management regimes

I used the updated SPOM to assess metapopulation viability for L. raniformis in the central region of the study area. The region incorporates 45 sites, including pools along

the Merri, Central and Edgars Creeks (Fig. 5.2). I began by simulating the occupancy dynamics within this region for a period of 30 years assuming current conditions (wetland number, attributes and locations remain unchanged). I then tested five additional scenarios of wetland loss and replacement (Fig. 5.2). For all scenarios, I removed 11 sites earmarked for destruction under existing urban growth plans for the study area (DEPI 2013). In the first scenario, no replacement wetlands were constructed. In the remainder, either four or six new wetlands were included to mitigate the impacts of wetland loss arising from urbanisation. I chose either a ‘clustered’ or ‘stepping-stone’ approach to wetland locations (Fig. 5.2 and Appendix 5.1), where clustering was intended to increase the robustness of particular wetland networks, and the ‘stepping-stone’ arrangement sought to improve connectivity across the focal region. New wetland locations were selected from maps of hydrologically and logistically feasible sites developed by the Victorian Department of Environment, Land, Water and Planning, for the purpose of investigating scenarios for L. raniformis wetland creation (DELWP, unpublished data). Each new wetland was allocated the surface area mapped by DELWP, and set to a hydroperiod of 3 (semi-permanent). I set aquatic vegetation cover to 50%, following draft guidelines for wetland construction for L. raniformis in Melbourne’s urbanising regions (DELWP unpublished). Sites were set to be unoccupied at the beginning of the simulation. For each scenario, I calculated the proportion of urban infrastructure surrounding and between sites (for down weighting effective wetland area and for genetic connectivity) using aerial imagery and mapping

110 Chapter 5

of the proposed urban growth in the study area (DELWP 2015). ! ! !

!

! ! ! ! ¯ ! 4 ! !( 1 !( !( 5 ! ! 2 !( !( 3 !( 6 ! ! ! > ! ! !( 7

! ! ! ! !( 8 ! Merri Creek !( 9 ! ! > > > >> ! > > ! !

!( 10 ! Legend !( New sites > > Removed sites !! ! ! Sites ! Streams Edgars Creek Future urban growth area Current urban area

! Central Creek !( 11

!! ! ! > > ! ! km 00.61.2

! ! Figure 5.2. The subset of monitoring! sites! for Litoria raniformis (black dots) used for the simulations of management effectiveness. Sites likely to be lost under future urbanisation are indicated by dots marked with a cross and potential new sites (to be constructed to offset the impacts of urbanisation) are indicated by white dots. Potential new sites are labelled 1 - 11. Scenario 2 includes new sites 3, 4, 5 and 6. Scenario 3 includes new sites 2, 8, 10 and 11. Scenario 4 includes new sites 1, 2, 3, 4, 5 and 6. Scenario 5 includes new sites 2, 7, 8, 9, 10 and 11.

111 Chapter 5

I ran 5000 simulations of the occupancy dynamics over 30 years (beginning from the last observed occupancy pattern), corresponding to one simulation for each of the 5000 sets of parameter estimates derived during model fitting. For each simulation, I recorded the minimum number of occupied wetlands across the 30 year time period and the number of occupied wetlands for each of these years. For comparison with the updated model, I also ran corresponding simulations using the previous model structure (but fitted to the same occupancy dataset), which excluded effects of urbanisation on wetland effective area and genetic connectivity. Simulations were run in R v. 3.0.3 (R Core Team 2014).

RESULTS

SPOM extensions

There was a positive relationship between the connectivity estimates derived from my model of genetic distance and the annual probabilities of population persistence and colonisation for L. raniformis (Table 5.1, Fig. 5.3). Colonisation had a strong positive relationship with connectivity, increasing by 85% (95% credible interval: 66 - 92%) over the range of this variable, while persistence had a weaker relationship, increasing by 23% (95% CI: 13 - 35%). The probability of persistence displayed strong positive relationships with wetland effective area and aquatic vegetation cover (increasing by 85% (95% CI: 67 - 97%) and 50% (95% CI: 29 - 69%) over the range of these variables, respectively), and populations in lentic wetlands had 16% (95% CI: 1 - 16%) higher probability of persistence each year than those in pools along streams (Fig. 5.3b).

Metapopulation viability

Simulations in which current conditions were maintained for the next 30 years revealed that the focal metapopulations of L. raniformis are currently close to their equilibrium rate of occupancy. Although there was considerable uncertainty around the predictions, on average the occupancy rate increased only slightly over this period (from 22 to 24 occupied sites; Fig. 5.4a). The lower 95% credible interval for the minimum number of occupied wetlands across the simulation period was 10 for the focal set of sites, indicating that these metapopulations are fairly robust under current conditions and have

112 Chapter 5 a very small probability of extinction (Fig. 5.5). Wetland destruction associated with planned urban development (entailing the removal of 5 dams and 6 quarries), reduced the mean predicted number of occupied sites to 15 over the next 30 years (Fig. 5.4b). However, the risk of decline was substantially higher, with the lower bound of the 95% CI of the minimum number of occupied wetlands over 30 years dropping to just 4 (Fig. 5.5).

Table 5.1. Summary statistics (mean and 95% credible interval) for the posterior distributions of the regression coefficients for the effects of effective wetland area (Aeff), aquatic vegetation cover (V), wetland type (WT) and connectivity (S, log transformed) on the probabilities of population persistence and colonisation for Litoria raniformis. The summaries are based on the 5000 samples from the posterior distribution, as used for the simulations of metapopulation viability.

Probability of Parameter Effect of Code name Mean 2.5% 97.5%

Colonisation "# - alpha.gamma -1.703 -2.174 -1.103

Colonisation $#1 S beta.gamma1 3.014 1.873 5.097

Persistence "% - alpha.phi 1.434 0.810 2.048

Persistence $%1 Aeff beta.phi1 3.371 2.238 4.685

Persistence $%2 V beta.phi2 2.372 1.305 3.668

Persistence $%3 WT beta.phi3 -1.094 -2.376 -0.054

Persistence $%4 S beta.phi4 1.141 0.187 2.279

113 Chapter 5

1.0

0.8

0.6

0.4

0.2

Probability of colonisation 0.0

0 1 2 3 4 5 6 7 8 9 10 Connectivity (a)

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0

1 2 3 4 5 6 7 8 9 10 11 12 Lentic Lotic

Effective area (loge square metres) Wetland type

1.0 1.0

0.8 0.8 Probability of persistence

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0

0 10 20 30 40 50 60 0 1 2 3 4 5 6 7 8 9 10 Aquatic vegetation cover (%) Connectivity (b) Figure 5.3. Relationships between: (a) the probability of colonisation and genetic connectivity (S, log transformed) for Litoria raniformis, and; (b) the probability of

persistence and effective wetland area (Aeff), wetland type (WT), aquatic vegetation cover (V) and genetic connectivity (S, log transformed). The solid lines represent the mean relationship, and grey shading the 95% credible intervals. Relationships are shown with all other variables held at their mean values.

114 Chapter 5

# occupied wetlands occupied # Number of occupied wetlands Number of occupied wetlands # occupied wetlands occupied # 0 10 20 30 40 0 10 20 30 40

0 5 10 15 20 25 30 0 5 10 15 20 25 30 (a) "#$%Year! (b) "#$%Year!

Number of occupied wetlands # occupied wetlands wetlands occupied # 0 10 20 30 40 Number of occupied wetlands # occupied wetlands occupied # 0 10 20 30 40

0 5 10 15 20 25 30 0 5 10 15 20 25 30 "#$%Year! (c) "#$%Year ! (d)

Number of occupied wetlands Number of occupied wetlands # occupied wetlands occupied # # occupied wetlands occupied # 0 10 20 30 40 0 10 20 30 40

0 5 10 15 20 25 30 0 5 10 15 20 25 30 (e) "#$%Year ! (f) "#$%Year! Figure 5.4. The predicted number of wetlands occupied by Litoria raniformis over 30 years for each scenario. The scenarios were: (a) Current conditions are maintained; (b) With sites removed due to future urbanisation (Scenario 1); (c) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a cluster (Scenario 2); (d) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a stepping stone formation (Scenario 3); (e) With sites removed due to future urbanisation and 6 new offset wetlands arranged in a cluster (Scenario 4), and; (f) With sites removed due to future urbanisation, and 6 new wetlands arranged in a stepping stone formation (Scenario 5). The solid line shows the mean estimate and the dashed lines the 95% CI.

! 115 Chapter 5

Wetland creation (up to 6 new wetlands) to offset the removal of these 11 wetlands increased the mean number of occupied wetlands across the simulation period, but there was considerable uncertainty for all wetland creation scenarios (see Fig. 5.4c-f). Scenario 4 (6 new wetlands arranged in a cluster) showed the largest increase in the mean number of occupied wetlands, rising from 15 in the first year to 21 (95% CI = 15 – 30) in the thirtieth year, compared with 24 (95% CI = 14 – 33) in the thirtieth year, under current conditions. The minimum number of occupied wetlands throughout the 30 year time period was similar for all habitat offsetting schemes, ranging from a mean of 9 (95% CI = 3 – 16) for scenario 3 (4 new wetlands arranged in a stepping stone formation), to 12 (95% CI = 5 – 18) for scenario 4 (6 new wetlands arranged in a cluster) (Fig. 5.5). Importantly however, the lower bound of the minimum number of occupied wetlands was lower in the best performing offsetting scenario than under current conditions, decreasing from 10 to 5 (Fig. 5.5).

&!"&!"

%#"%#"

%!"%!"

$#"$#"

$!"$!"

!"#"$%$&'((%)"*+&,"-*.& #"#"

!"!" ''"()*+,-"''" (.,/0123"$"(.,/0123"$"(.,/0123"%"(.,/0123"%"(.,/0123"&"(.,/0123"&"(.,/0123"4"(.,/0123"4"(.,/0123"#" ,(*#/0"1&

Figure 5.5. The predicted minimum number of wetlands occupied by Litoria raniformis over 30 years for each scenario. The scenarios were: (a) Current conditions are maintained; (b) With sites removed due to future urbanisation (Scenario 1); (c) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a cluster (Scenario 2); (d) With sites removed due to future urbanisation and 4 new offset wetlands arranged in a stepping stone formation (Scenario 3); (e) With sites removed due to future urbanisation and 6 new offset wetlands arranged in a cluster (Scenario 4), and; (f) With sites removed due to future urbanisation, and 6 new wetlands arranged in a stepping stone formation (Scenario 5). The solid line shows the mean estimate and the dashed lines the 95% CI. Triangles represent mean estimates derived using the updated

116 Chapter 5

SPOM with genetic connectivity. Circles represent mean estimates derived from the original model excluding urban barriers. Vertical lines represent the 95% credible intervals.

Model comparison

Comparing the above simulations with corresponding simulations using the previous SPOM for L. raniformis suggests that the updated model is pessimistic, both with regard to mean estimates of occupancy trajectories and uncertainty around those estimates. The mean minimum number of occupied wetlands under current conditions decreased by 5 with the updated model (from 23 to 18), and the width of the 95% CI of this metric increased from 10 to 15 (see Fig. 5.5). This trend was similar when comparing results for the alternate wetland creation schemes; the mean minimum number of occupied wetlands ranged between 14 – 16 using the previous model and 9 – 12 using the model developed here, and 95% CIs for this metric between 2 and 7 wetlands higher for the previous model (Fig. 5.5).

DISCUSSION

Incorporating genetic connectivity estimates into metapopulation models

Here, I have demonstrated an effective method of incorporating genetic estimates of dispersal into a stochastic patch occupancy model (SPOM). This can help provide more realistic estimates of metapopulation viability, as it directly incorporates barriers to dispersal that limit recolonisation rates and hamper migration-based ‘rescue effects’ for declining populations. This is an important advance, because the dispersal rates of animals are notoriously difficult to characterise and incorporate into metapopulation models (Hanski 1999). Given the difficulties associated with the more traditional methods of calculating population connectivity estimates, such as mark-recapture and radiotelemetry, it is not surprising that few studies have attempted to directly incorporate these estimates into occupancy models. Instead, it appears the standard approach is to downgrade connectivity measures by an arbitrary amount. This is often achieved by re-scaling connectivity to account for variations in landscape permeability and developing cost-distance metrics (Moilanen and Hanski 1998, Vos et al. 2000,

117 Chapter 5

Duggan et al. 2011). Landscape genetics provides several solutions for calculating dispersal rates and thereby directly incorporating the effects of landscape permeability into connectivity metrics.

Firstly, genetic estimates of dispersal rate can be significantly less expensive and less time consuming than traditional techniques such as mark-recapture and radio-telemetry. The focal species of this study - the endangered frog Litoria raniformis - provides a useful example. Past estimates of the decay in dispersal rate with distance for this species were derived from 131 recaptures of 807 frogs marked across 31 sites during 314 surveys. By comparison, the dispersal model used here was derived using genetic sampling of 575 frogs, captured during 119 surveys of 29 sites; a 62% reduction in survey effort.

Secondly, statistical advances in landscape genetics provide various tools in which to estimate the effects of landscape structure on dispersal rate. Varying landscape permeability is a key consideration for the dynamics of species in fragmented landscapes (Ricketts 2001), but traditional approaches have required extensive information on movement rates to incorporate landscape permeability into metapopulation models (e.g. Ovaskainen 2004). Here, I built a simple linear model to describe the effects of geographic distance and urban barriers on dispersal rate, which proved to have high predictive power for my study species (Chapter 4). However, genetic information potentially allows quite complex models of dispersal rate to be constructed, incorporating multiple landscape elements (Cushman et al. 2006) and spatial scales (Dudaniec et al. 2013).

Thirdly, landscape genetic data provide estimates of effective dispersal and gene flow rather than solely an estimate of the number of migrants a particular site may receive in a given time period. In doing so, genetic information provides a more direct estimate of the interaction of neighbouring populations and the contribution of dispersal to metapopulation viability (Gaggiotti et al. 2004). This may be particularly important for species that have low survival rates, such as many amphibians. In this instance, it is possible that relatively high migration rates over short distances may entail little inter- population mating and interaction (due to high mortality rates post dispersal), masking the benefits of landscape connectivity for population and metapopulation persistence.

118 Chapter 5

Estimates of dispersal based on inter-population genetic distance – as used here – overcome this issue.

Comparing my updated SPOM to that previously developed for this species (Heard et al. 2013, Rose et al. 2016) revealed the benefits of my more nuanced model of dispersal. While the rank order of each management scenario remained the same in terms of metapopulation viability, the minimum number of occupied wetlands was consistently lower for simulations constructed using the model that incorporated genetic connectivity. The original SPOM predicted the mean minimum number of occupied wetlands to be 23% higher given current conditions and 28% – 54% higher for each future scenario. This result emphasises the importance of including barriers to dispersal, such as urban infrastructure, when assessing the future viability of metapopulations. In particular, my simulations revealed that the inclusion of dispersal barriers, quantified using genetic data, provided higher estimates of the risk of metapopulation extinction, as revealed by the lower bound of the 95% credible interval for the minimum number of occupied wetlands. Estimates of the lower bound of the 95% CI decreased 41% – 67% between the original SPOM and my updated model. However, the updated model did have a wider CI than the original SPOM, which could suggest that it is either more difficult to estimate parameters for the updated model, that it is providing more realistically uncertain predictions, or a combination of these.

Assessment of management options for Litoria raniformis

In this study, I manipulated both the number of newly created wetlands to be added and the spatial pattern of these wetlands. Compared with a ‘do nothing’ scenario following proposed urban development, adding six new wetlands increased occupancy after 30 years by 41% and 20% (for ‘cluster’ and ‘stepping stone’ formations respectively), while adding four new wetlands increased occupancy by 23% and decreased occupancy by 1% (for ‘cluster’ and ‘stepping stone’ formations respectively). Hence, arranging sites in a ‘cluster’ formation provided a superior management scenario than the alternative ‘stepping stone’. Although the model is based on genetic connectivity, and it might therefore be hypothesised that the ‘stepping stone’ formation would perform better, my finding supports the observation that migration primarily occurs over very small spatial scales in this species. The scenario in which six newly created wetlands

119 Chapter 5

were arranged in a ‘cluster’ also had the smallest range of uncertainty, therefore decreasing the estimated risk of local extinction of the species.

While the addition of wetlands following urbanisation offset the losses of existing populations to some extent, even the best performing scenario performed worse than current conditions. Projections of the mean number of occupied wetlands under current conditions and for the best performing scenario were similar; however, the lower bound of the mean minimum number of occupied wetlands for the best performing scenario decreased to six. Therefore, for the scenarios assessed here (which are based upon mapping of potential wetlands by the relevant government authority and are realistic in terms of proposed investment levels), wetland creation will not completely offset the increased risk of metapopulation extinction. This is concerning in itself, but compounded by the fact that I did not attempt to model changes in the quality of existing habitat as urbanisation progresses. Such changes are difficult to predict, but may include altered hydrology, decreased water quality and increased pollution, predation, competition, disease transmission and physical disturbance by humans (Hamer and McDonnell 2008). All these changes are known to have detrimental effects on amphibian populations. As such, my simulations of metapopulation viability for L.!raniformis following urban expansion may be optimistic, raising further concerns about the scale of wetland creation and management required to offset the impacts of urbanisation on this species.

The model of population genetic distance I developed (Chapter 4) provides clear evidence that urban infrastructure presents important barriers to dispersal for L. raniformis around Melbourne. However, reductions in metapopulation viability revealed by the updated SPOM may be smaller in my chosen study landscape than in other areas, given the future urbanisation in this area maintains reserves and undeveloped land and does not include the addition of many new roads. Other areas marked for future urbanisation will include additional barriers and further fragmentation of dispersal routes by roads and other infrastructure. Hence, the inclusion of barriers in the updated metapopulation model will have more of an impact in predicting the viability of the species in these areas.

120 ! Chapter 5

Conclusion

My approach demonstrates a powerful means of integrating genetic measures of connectivity with metapopulation modelling. While the case study reveals the utility of such integrated models for assessing management options for an endangered frog, I have used standard techniques in both the genetic component of this study (pairwise

FST) and the occupancy modelling (MacKenzie et al. 2006, Royle and Kéry 2007) that are widely applicable to other systems and taxa. As such, I expect my approach will prove broadly applicable to the development and application of future metapopulation models, and could facilitate conservation decision making for many other species in fragmented landscapes.

121 Chapter 5

APPENDICES

Appendix 5.1. The subset of monitoring sites for Litoria raniformis (black dots) used for the simulations of management effectiveness. Sites likely to be lost under future urbanisation are indicated by dots marked with a cross and potential new sites (to be constructed to offset the impacts of urbanisation) are indicated by white dots. Potential new sites are labelled 1 - 11. a. Scenario 2 includes new sites 3, 4, 5 and 6., b. Scenario 3 includes new sites 2, 8, 10 and 11, c. Scenario 4 includes new sites 1, 2, 3, 4, 5 and 6, and d. Scenario 5 includes new sites 2, 7, 8, 9, 10 and 11. ! ! ! ! ! !

! !

! ! ! ! ! ! ¯ ! ! ¯ ! ! (4 ! ! ( 5 ! ! ! ! 2 ( ( 3 ( 6 ! ! ! > ! ! ! > ! ! ! !

! ! ! ! ! ! ! ! ( 8 ! ! Merri Creek ! Merri Creek ! ! > ! > > > > > > > >> ! ! > > > > ! ! ! !

! ( 10 ! Legend Legend ( New sites > ( New sites > > Removed sites > !! Removed sites !! ! Sites ! ! ! ! Edgars Creek Sites ! Streams Streams Edgars Creek Future urban growth area Future urban growth area Current urban area Current urban area

! ! Central Creek Central Creek ( 11

!! ! !! ! ! ! > > > ! > ! ! km ! km 00.61.2 00.61.2

(a) ! (b) ! ! ! ! ! ! ! ! ! ! ! ! !

! !

! ! ! ! ! ! ¯ ! ! ¯ ! ! 4 ! ! ( 1 ( ( 5 ! ! ! ! 2 ( 2 ( ( 3 ( 6 ! ! ! > ! ! ! > ! ! ! ! ( 7

! ! ! ! ! ! ! ! ( 8 ! ! Merri Creek ! Merri Creek ( 9 ! ! > ! > > > > > >> > > ! ! > > > > ! ! ! !

! ( 10 ! Legend Legend ( ( New sites > New sites > > > Removed sites Removed sites !! !! ! ! ! Sites ! Sites ! ! Edgars Creek Streams Edgars Creek Streams Future urban growth area Future urban growth area Current urban area Current urban area

! ! Central Creek Central Creek ( 11

! !! ! ! ! ! ! > > > ! > ! ! ! km km 00.61.2 00.61.2

! (c) ! (d) ! ! ! ! ! !

122 Chapter 6

CHAPTER 6 General discussion

123 Chapter 6

Understanding the genetic structure and diversity of an endangered species is fundamental to understanding its conservation requirements. The research presented in this thesis makes an important contribution to our understanding of the ways in which amphibians in general – and Litoria raniformis in particular – are impacted by urbanisation. It also demonstrates the value of genetic information for quantitative models of metapopulations that can be used to assess management decisions for threatened species in urbanising landscapes. My contribution is summarised and discussed below.

SUMMARY OF FINDINGS

Chapter 2: Comparing genetic sampling techniques for amphibians with a multi- criteria decision framework

In Chapter 2, I used a multi-criteria decision framework to assess four tissue sampling techniques for amphibians against five decision criteria; DNA quantity, genetic accuracy, potential future use of samples, species conservation and the welfare of individual animals. Under this framework, I identified clipping of toe-webbing as the most viable alternative genetic sampling technique to toe-clipping for the focal species, as it performed well under each decision criterion. Buccal swabbing and skin swabbing performed well under species welfare and conservation criteria, but performed relatively poorly under sample quality and reuse criteria. This study informed the genetic sampling technique I chose for collecting samples for Chapters 3 and 4, and the approach is broadly applicable to future conservation genetic studies of amphibians and other animals.

Chapter 3: Genetic structure and diversity of the endangered Growling Grass Frog in a rapidly urbanising region

I used mitochondrial DNA (mtDNA) and microsatellites to investigate the genetic structure and diversity of L. raniformis across Melbourne’s urban fringe for Chapter 3. I found low levels of genetic diversity throughout remnant populations of the species. However, one of the four regions studied, Cardinia, exhibited high genetic diversity

124 Chapter 6 relative to the other three regions sampled and exhibited several unique haplotypes, suggesting this region should be recognised as a separate Management Unit (MU). This study forewarns managers of the low genetic diversity displayed by remnant populations, and identifies genetic MUs and regional centres of haplotype endemism for L. raniformis around Melbourne.

Chapter 4: Impacts of urbanisation on the population genetic structure of a threatened amphibian, Litoria raniformis ! In Chapter 4, I conducted a comprehensive landscape genetic study of L. raniformis in the northern region of Melbourne and developed a predictive model of the landscape drivers of genetic distance for the species. I also used microsatellites to investigate the fine-scale genetic structuring of L. raniformis in the region. Geographic distance between sites had a strong negative effect on gene flow, as did the extent of urban infrastructure. My statistical approach delivered a model of gene flow that will be an important predictive tool for the management of L. raniformis around Melbourne and demonstrated a statistical approach that may be applied to understand the drivers of gene flow in threatened amphibians generally.

Chapter 5: Integrating genetic connectivity measures with stochastic patch occupancy models for metapopulation management ! Stochastic patch occupancy models (SPOMs) can be used to predict the persistence of metapopulations in fragmented landscapes. In Chapter 5, I integrated my model of genetic connectivity and gene flow (developed in Chapter 4), with a Bayesian SPOM to assess the metapopulation viability of L. raniformis around Melbourne. For the focal metapopulation, I found significant investment in new wetlands may be required to offset the impacts of urbanisation in the region. This study shows how genetic connectivity measures can improve quantitative models used to assess management options for species afflicted by habitat loss and fragmentation. The approach represents a useful alternative to characterising dispersal dynamics in these models when traditional means, such as mark-recapture studies or radio-telemetry, are impractical.

! 125 Chapter 6

FUTURE DIRECTIONS AND RECOMMENDATIONS

This research has enhanced our undertsanding of the consequences of urbanisation for L. raniformis around Melbourne, and for other taxa in urban environments. However, further research into the effects of urbanisation on metapopulations of L. raniformis is required. Around Melbourne, additional sampling and fine-scale population genetic analyses, specifically in the Cardinia region, would be useful. These populations exhibited relatively high genetic diversity and several unique haplotypes, are more geographically isolated than other populations around Melbourne, and are the remnants of a much larger group of historical populations throughout the west and central Gippsland area to the east of Melbourne.

While this study was primarily focused on quantifying the effect of urban infrastructure on gene flow of L. raniformis around Melbourne, additional landscape variables that either facilitate or hinder gene flow for the species could be incorporated into future modelling. In particular, a refined understanding of the relative influence of particular urban features, such as roads, housing estates, industrial estates, flat impervious surfaces and open spaces, on the genetic structure and gene flow of L. raniformis around Melbourne would provide insights into the value of mitigating particular barrier types during future urban planning. Additionally, this could include the three urban ecological paradigms of ecology in, of, and for the city, where ecology in focuses on terrestrial and aquatic patches, ecology of incorporates biological, social and built components, and ecology for integrates the two and links these with civic processes (Pickett et al. 2016).

Incorporating genetic information into a SPOM delivers a powerful model for assessing metapopulation viability and can provide valuable direction for the conservation management of L. raniformis. Although urbanisation remains a major threat to the species, this is one of the few conservation genetic studies that integrates genetic information into a specific model of metapopulation dynamics for direct use in management. Wetland creation programs are set to dominate conservation planning for this species across Melbourne over the next 30 years (DEPI 2013). My model will

126 Chapter 6 provide significant guidance for optimising the location of these wetlands, and ultimately maximise the likelihood of preserving remnant metapopulations of L. raniformis across the city given the resources and management options available.

A better understanding of wetland creation scenarios could include the manipulation of numerous additional site attributes, including financial cost, wetland area, hydroperiod and the proportion of aquatic vegetation cover, and provide further insights for conservation planning and management. Additionally, alternative management actions such as habitat enhancement of existing wetlands could be considered. Changes in habitat quality due to urbanisation are difficult to predict, however, habitat degradation is likely and has detrimental effects on L. raniformis populations. Therefore management scenarios incorporating habitat degradation (and mitigation through direct management) will provide more realistic predictions of metapopulation viability over time.

Lastly, translocation is another key component of conservation planning for L. raniformis (DEPI 2013). However, there is currently no evidence that populations of this species can be successfully relocated. Indeed, all translocation attempts to date have failed or the focal populations are performing poorly (Koehler et al. 2014a). If future captive breeding programmes are required or translocations are attempted for L. raniformis, it is important that the geographical location of source populations is considered, to maintain the genetic integrity and evolutionary potential of the species around Melbourne. This appears most important in the Cardinia region, but also applies to less diverse regions. Nevertheless, I advise that translocation of individuals between regions should be avoided and that fine-scale genetic studies within regions should be undertaken to determine the long-term viability of translocations within regions.

127 Chapter 6

128 Literature Cited

ABS. 2013. Regional Population Growth, Australia, 2012, cat. no. 3218.0. Australian Bureau of Statistics, Canberra. Allentoft, M. E. and J. O'Brien. 2010. Global amphibian declines, loss of genetic diversity and fitness: a review. Diversity 2:47-71. Alvarez, D., A. Loureno, D. Oro, and G. Velo-Anton. 2015. Assessment of census (N) and effective population size (Ne) reveals consistency of Ne single-sample estimators and a high Ne/N ratio in an urban and isolated population of fire salamanders. Conservation Genetics Resources 7:705-712. Andren, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71:355-366. Andren, H. and P. Angelstam. 1988. Elevated predation rates as an edge effect in habitat islands: experimental evidence. Ecology 69:544-547. Arens, P., T. van der Sluis, W. P. C. van't Westende, B. Vosman, C. C. Vos, and M. J. M. Smulders. 2007. Genetic population differentiation and connectivity among fragmented Moor frog (Rana arvalis) populations in The Netherlands. Landscape Ecology 22:1489-1500. Armstrong, D. P. and P. J. Seddon. 2008. Directions in reintroduction biology. Trends in Ecology & Evolution 23:20-25. Balkenhol, N., S. Cushman, A. Storfer, and L. Waits. 2016a. Landscape genetics: concepts, methods, applications. John Wiley & Sons. Balkenhol, N., S. A. Cushman, L. P. Waits, and A. Storfer. 2016b. Current status, future opportunities, and remaining challenges in landscape genetics. Landscape Genetics: Concepts, Methods, Applications:247-256. Balkenhol, N. and L. P. Waits. 2009. Molecular road ecology: exploring the potential of genetics for investigating transportation impacts on wildlife. Molecular ecology 18:4151-4164. Balkenhol, N., L. P. Waits, and R. J. Dezzani. 2009. Statistical approaches in landscape genetics: an evaluation of methods for linking landscape and genetic data. Ecography 32:818-830. Bandelt, H. J., P. Forster, and A. Rohl. 1999. Median-joining networks for inferring intraspecific phylogenies. Molecular Biology and Evolution 16:37-48.

129 Becker, C. G., C. R. Fonseca, C. F. B. Haddad, R. F. Batista, and P. I. Prado. 2007. Habitat split and the global decline of amphibians. Science 318:1775. Becker, C. G. and R. D. Loyola. 2008. Extinction risk assessments at the population and species level: implications for amphibian conservation. Biodiversity and Conservation 17:2297-2304. Beebee, T. J. C. 2005. Conservation genetics of amphibians. Heredity 95:423-427. Beerli, P. 2009. How to use migrate or why are markov chain monte carlo programs difficult to use? Pages 42-79 in G. Bertorelle, M. W. Bruford, H. C. Hauffe, A. Rizzoli, and C. Vernesi, editors. Population genetics for animal conservation. Cambridge University Press, Cambridge, UK. Bekessy, S. A., R. A. Ennos, M. A. Burgman, A. C. Newton, and P. K. Ades. 2003. Neutral DNA markers fail to detect genetic divergence in an ecologically important trait. Biological Conservation 110:267-275. Benjamini, Y. and Y. Hochberg. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological) 57:289-300. Blaustein, A. R., D. B. Wake, and W. P. Sousa. 1994. Amphibian declines: Judging stability, persistence and susceptibilityof populations to local and global extinctions. Conservation Biology 8:60-71. Blumenthal, J., F. Abreu-Grobois, T. Austin, A. Broderick, M. Bruford, M. Coyne, G. Ebanks-Petrie, A. Formia, P. Meylan, and A. Meylan. 2009. Turtle groups or turtle soup: dispersal patterns of hawksbill turtles in the Caribbean. Molecular ecology 18:4841-4853. Bogert, C. M. 1947. A field study of homing in the Carolina toad. American Museum Novitates 1355:1-24. Bos, D. H. and J. W. Sites, Jr. 2001. Phylogeography and conservation genetics of the Columbia spotted frog (Rana luteiventris; Amphibia, Ranidae). Molecular ecology 10:1499-1513. Bossuyt, F. and M. C. Milinkovitch. 2000. Convergent adaptive radiations in Madagascan and Asian ranid frogs reveal covariation between larval and adult traits. Proceedings of the National Academy of Sciences 97:6585-6590. Bricker, J., L. Bushar, H. Reinert, and L. Gelbert. 1996. Purification of high quality DNA from shed skins. Herpetological Review 27:133-133.

130 Broquet, T., L. Berset-Braendli, G. Emaresi, and L. Fumagalli. 2007. Buccal swabs allow efficient and reliable microsatellite genotyping in amphibians. Conservation Genetics 8:509-511. Broquet, T. and E. Petit. 2004. Quantifying genotyping errors in noninvasive population genetics. Molecular ecology 13:3601-3608. Broquet, T. and E. J. Petit. 2009. Molecular estimation of dispersal for ecology and population genetics. Annual Review of Ecology, Evolution, and Systematics 40:193-216. Bull, M. and G. Stolfo. 2014. Flora of Melbourne. Hyland House Publishing, Carlton, Victoria. Burgman, M. A. 2005. Risks and Decisions for Conservation and Environmental Management. Cambridge University Press, Cambridge. Burnham, K. P. and D. R. Anderson. 2002. Model Selection and Multi-model Inference: A Practical Information-theoretic Approach. 2nd edition. Springer, New York. Burns, E. and G. Ferrari. 2004. Microsatellite loci for the green and golden bell frog (Litoria aurea). Conservation Genetics 5:421-423. Burns, E. L. 2004. Phylogeography, Population History and Conservation Genetics of the Endangered Green and Golden Bell Frog (Litoria aurea). PhD thesis. The University of New South Wales, New South Wales. Burns, E. L. and D. M. Crayn. 2006. Phylogenetics and evolution of bell frogs (Litoria aurea species-group, Anura: Hylidae) based on mitochondrial ND4 sequences. Molecular Phylogenetics and Evolution 39:573-579. Burns, E. L., M. E. Eldridge, and B. A. Houlden. 2004. Microsatellite variation and population structure in a declining Australian hylid Litoria aurea. Molecular ecology 13:1745-1757. Carr, L. W. and L. Fahrig. 2001. Effect of road traffic on two amphibian species of differing vagility. Conseravtion Biology 15:1071-1078. Carrier, J. A. and T. J. C. Beebee. 2003. Recent, substantial, and unexplained declines of the common toad Bufo bufo in lowland England. Biological Conservation 111:395-399. Clarke, G. P., P. C. L. White, and S. Harris. 1998. Effects of roads on badger Meles meles populations in south-west England. Biological Conservation 86:117-124.

131 Clemann, N. and G. Gillespie. 2010. Draft National Recovery Plan for the Growling Grass Frog Litoria raniformis. Department of Sustainability and Environment, Melbourne. Clemann, N. and G. Gillespie. 2012. National Recovery Plan for the Southern Bell Frog Litoria raniformis. Department of Sustainability and Environment, Melbourne. Clemen, R. T. 1996. Making hard decisions: an introduction to decision analysis. Duxbury Press, Pacific Grove. Converse, S. J., C. T. Moore, M. J. Folk, and M. C. Runge. 2013. A matter of tradeoffs: Reintroduction as a multiple objective decision. The Journal of Wildlife Management 77:1145-1156. Cornuet, J. and G. Luikart. 1996. Description and power analysis of two tests for detecting recent population bottlenecks from allele frequency data. Genetics 144:2001-2014. Correa, D. T. 2013. Population declines: toe-clipping vital to amphibian research. Nature 493:305-305. Cosentino, B. J., D. M. Marsh, K. S. Jones, J. J. Apodaca, C. Bates, J. Beach, K. H. Beard, K. Becklin, J. M. Bell, C. Crockett, G. Fawson, J. Fjelsted, E. A. Forys, K. S. Genet, M. Grover, J. Holmes, K. Indeck, N. E. Karraker, E. S. Kilpatrick, T. A. Langen, S. G. Mugel, A. Molina, J. R. Vonesh, R. J. Weaver, and A. Willey. 2014. Citizen science reveals widespread negative effects of roads on amphibian distributions. Biological Conservation 180:31-38. Costa, L. P., Y. L. R. Leite, S. L. Mendes, and A. D. Ditchfield. 2005. Mammal conservation in Brazil. Conservation Biology 19:672-679. Crandall, K. A. 1994. Intraspecific cladogram estimation: accuracy at higher levels of divergence. Systematic Biology 43:222-235. Cumming, G. and S. Finch. 2001. A primer on the understanding, use, and calculation of confidence intervals that are based on central and noncentral distributions. Educational and Psychological Measurement 61:532-574. Cummings, S. and G. Thorgaard. 1994. Extraction of DNA from fish blood and sperm. Biotechniques 17:426-430. Cushman, S. A. 2006. Effects of habitat loss and fragmentation on amphibians: a review and prospectus. Biological Conservation 128:231-240. Cushman, S. A., K. S. McKelvey, J. Hayden, and M. K. Schwartz. 2006. Gene flow in complex landscapes: Testing multiple hypotheses with causal modeling.

132 Davis, A., L. B. Slusher, and J. T. Beneski. 2002. A novel method for the extraction and amplification of DNA from amphibians. Herpetological Review 33:104-105. Davis, T. M. and K. Ovaska. 2001. Individual recognition of amphibians: effects of toe clipping and fluorescent tagging on the salamander Plethodon vehiculum. Journal of Herpetology 35:217-225. Delaney, K. S., S. P. Riley, and R. N. Fisher. 2010. A rapid, strong, and convergent genetic response to urban habitat fragmentation in four divergent and widespread vertebrates. PloS one 5:e12767. DELWP. 2015. Planning scheme Urban Growth Boundary.in D. o. E. The State of Victoria, Land, Water & Planning, editor. deMaynadier, P. G. and M. L. Hunter. 1999. Forest canopy closure and juvenile emigration by pool-breeding amphibians in Maine. Journal of Wildlife Management 63:441-450. deMaynadier, P. G. and M. L. Hunter. 2000. Road effects on amphibian movements in a forested landscape. Natural Areas Journal 20:56-65. DEPI. 2013. Sub-Regional Species Srategy for the Growling Grass Frog. Department of Environment and Primary Industries, East Melbourne, Victoria. Dixo, M., J. P. Metzger, J. S. Morgante, and K. R. Zamudio. 2009. Habitat fragmentation reduces genetic diversity and connectivity among toad populations in the Brazilian Atlantic Coastal Forest. Biological Conservation 142:1560-1569. Dixon, C. J. 2006. A means of estimating the completeness of haplotype sampling using the Stirling probability distribution. Molecular ecology notes 6:650-652. Dolman, G. and B. Phillips. 2004. Single copy nuclear DNA markers characterized for comparative phylogeography in Australian wet tropics rainforest skinks. Molecular ecology notes 4:185-187. DPCD. 2009. Delivering Melbourne's Newest Sustainable Communities. Strategic Impact Assessment Report. Department of Planning and Community Development, Melbourne. Dreschler, M., K. Frank, I. Hanski, R. B. O'Hara, and C. Wissel. 2003. Ranking metapopulation extinction risk: from patterns in data to conservation management decisions. Ecological Applications 13:990-998.

133 Driscoll, D. A., et al. 2010. Resolving conflicts in fire management using decision theory: asset-protection versus biodiversity conservation. Conservation Letters 3:215-223. Drummond, A., B. Ashton, S. Buxton, M. Cheung, A. Cooper, J. Heled, M. Kearse, R. Moir, S. Stones-Havas, and S. Sturrock. 2012. Geneious v5. 6. Biomatters Ltd., Auckland, New Zealand. DSE. 2013. Advisory List of Threatened Vertebrate Fauna in Victoria - 2013. Department of Sustainability and Environment, East Melbourne, Victoria. Dudaniec, R. Y., J. R. Rhodes, J. Worthington Wilmer, M. Lyons, K. E. Lee, C. A. McAlpine, and F. N. Carrick. 2013. Using multilevel models to identify drivers of landscape-genetic structure among management areas. Molecular ecology 22:3752-3765. Duggan, J. M., R. L. Schooley, and E. J. Heske. 2011. Modeling occupancy dynamics of a rare species, Franklin's ground squirrel, with limited data: are simple connectivity metrics adequate? Landscape Ecology 26:1477-1490. Earl, D. A. and B. M. Vonholdt. 2012. STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources 4:359-361. Eaton, M. J., P. T. Hughes, J. E. Hines, and J. D. Nichols. 2014. Testing metapopulation concepts: effects of patch characteristics and neighborhood occupancy on the dynamics of an endangered lagomorph. Oikos 123:662-676. Edge, C. B., M. K. Gahl, D. G. Thompson, and J. E. Houlahan. 2013. Laboratory and field exposure of two species of juvenile amphibians to a glyphosate-based herbicide and Batrachochytrium dendrobatidis. Science of The Total Environment 444:145-152. Edwards, L. J., K. E. Muller, R. D. Wolfinger, B. F. Qaqish, and O. Schabenberger. 2008. An R2 statistic for fixed effects in the linear mixed model. Statistics in medicine 27:6137-6157. Edwards, W. 1977. How to use multiattribute utility measurement for social decisionmaking. Systems, Man and Cybernetics, IEEE Transactions on 7:326- 340. Emaresi, G., J. Pellet, S. Dubey, A. Hirzel, and L. Fumagalli. 2011. Landscape genetics of the Alpine newt (Mesotriton alpestris) inferred from a strip-based approach. Conservation Genetics 12:41-50.

134 Epps, C. W., P. J. PalsbRll, J. D. Wehausen, G. K. Roderick, R. R. Ramey, and D. R. McCullough. 2005. Highways block gene flow and cause a rapid decline in genetic diversity of desert bighorn sheep. Ecology Letters 8:1029-1038. Evanno, G., S. Regnaut, and J. Goudet. 2005. Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular ecology 14:2611-2620. Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, Evoluation and Systematics 34:487-515. Fahrig, L., J. H. Pedlar, S. E. Pope, P. D. Taylor, and J. F. Wegner. 1995. Effect of road traffic on amphibian density. Biological Conservation 73:177-182. Faivovich, J., C. F. B. Haddad, P. C. A. Garcia, D. R. Frost, J. A. Campbell, and W. C. Wheeler. 2005. Systematic review of the frog family Hylidae, with special reference to Hylinae: phylogenetic analysis and taxonomic revision. Bulletin of the American Museum of Natural History 294:1-240. Faubet, P., R. S. Waples, and O. E. Gaggiotti. 2007. Evaluating the performance of a multilocus Bayesian method for the estimation of migration rates. Molecular ecology 16:1149-1166. Feinstein, J. 2004. DNA sequence from butterfly frass and exuviae. Conservation Genetics 5:103-104. Fisher, K. J., K. J. Guilfoyle, and K. A. Hatch. 2013. Stress Induced by Toe-Clipping in Cane Toads (Rhinella marina). Copeia 2013:539-542. Flagstad, y., N. Pradhan, L. G. Kvernstuen, and P. Wegge. 2012. Conserving small and fragmented populations of large mammals: Non-invasive genetic sampling in an isolated population of Asian elephants in Nepal. Journal for Nature Conservation 20:181-190. Formia, A., A. C. Broderick, F. Glen, B. J. Godley, G. C. Hays, and M. W. Bruford. 2007. Genetic composition of the Ascension Island green turtle rookery based on mitochondrial DNA: implications for sampling and diversity. Endangered Species Research 3:145-158. Frankham, R. 2010. Where are we in conservation genetics and where do we need to go? Conservation Genetics 11:661-663. Frankham, R., J. D. Ballou, and D. A. Briscoe. 2002. Introduction to Conservation Genetics. Cambridge University Press, Cambridge, UK.

135 Franks, S. J., P. D. Pratt, and N. D. Tsutsui. 2011. The genetic consequences of a demographic bottleneck in an introduced biological control insect. Conservation Genetics 12:201-211. Fu, Y.-X. 1997. Statistical tests of neutrality of mutations against population growth, hitchhiking and background selection. Genetics 147:915-925. Fujiwara, M. and H. Caswell. 2002. Estimating population projection matrices from multi-stage mark-recapture data. Ecology 83:3257-3265. Funk, W. C., M. A. Donnelly, and K. R. Lips. 2005. Alternative views of amphibian toe clipping. Nature 433:193. Funk, W. C., D. A. Tallmon, and F. W. Allendorf. 1999. Small effective population size in the long toed salamander. Molecular ecology 8:1633-1640. Furman, B. L., B. R. Scheffers, M. Taylor, C. Davis, and C. A. Paszkowski. 2015. Limited genetic structure in a wood frog (Lithobates sylvaticus) population in an urban landscape inhabiting natural and constructed wetlands. Conservation Genetics 17:1-12. Gaggiotti, O., S. Brooks, W. Amos, and J. Harwood. 2004. Combining demographic, environmental and genetic data to test hypotheses about colonization events in metapopulations. Molecular ecology 13:811-825. Garner, B. A., B. K. Hand, S. J. Amish, L. Bernatchez, J. T. Foster, K. M. Miller, P. A. Morin, S. R. Narum, S. J. O’Brien, and G. Roffler. 2016. Genomics in conservation: case studies and bridging the gap between data and application. Trends Ecol Evol 31:81-83. Garroway, C. J., J. Bowman, D. Carr, and P. J. Wilson. 2008. Applications of graph theory to landscape genetics. Evolutionary Applications 1:620-630. Garza, J. and E. Williamson. 2001. Detection of reduction in population size using data from microsatellite loci. Molecular ecology 10:305-318. Gelman, A. and J. Hill. 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press. Germano, J. M. and P. J. Bishop. 2009. Suitability of Amphibians and Reptiles for Translocation. Conservation Biology 23:7-15. Gibbs, J. P. and W. G. Shriver. 2005. Can road mortality limit populations of pool- breeding amphibians? Wetlands Ecology and Management 13:281-289. Golay, N. and H. Durrer. 1994. Inflammation due to toe-clipping in natterjack toads (Bufo calamita). Amphibia-Reptilia 15:81-83.

136 Goldberg, C. S., M. E. Kaplan, and C. R. Schwalbe. 2003. From the Frog's Mouth: Buccal Swabs for Collection of DNA from Amphibians. Herpetological Review 34:220-221. Goldberg, C. S. and L. P. Waits. 2010. Comparative landscape genetics of two pond- breeding amphibian species in a highly modified agricultural landscape. Molecular ecology 19:3650-3663. Gomez-Mestre, I., J. J. Wiens, and K. M. Warkentin. 2008. Evolution of adaptive plasticity: risk-sensitive hatching in neotropical leaf-breeding treefrogs. Ecological Monographs 78:205-224. Goodwin, P. and G. Wright. 2004. Decision analysis for management judgment. 3rd edition. Wiley, West Sussex, England. Grafe, T. U., M. M. Stewart, K. P. Lampert, and M.-O. Rodel. 2011. Putting Toe Clipping into Perspective: A Viable Method for Marking Anurans. Journal of Herpetology 45:28-35. Gray, E. M. 1995. DNA fingerprinting reveals a lack of genetic variation in northern populations of the western pond turtle (Clemmys marmorata). Conservation Biology 9:1244-1255. Greenwald, K. R. 2010a. Genetic data in population viability analysis: case studies with ambystomatid salamanders. Animal Conservation 13:115-122. Greenwald, K. R. 2010b. Genetic data in population viability analysis: guidelines for future research. Animal Conservation 13:129-130. Guillot, G. 2012. Population genetic and morphometric data analysis using R and the Geneland program. Guillot, G., A. Estoup, F. Mortier, and J. F. Cosson. 2005a. A spatial statistical model for landscape genetics. Genetics 170:1261-1280. Guillot, G., F. Mortier, and A. Estoup. 2005b. GENELAND: a computer package for landscape genetics. Molecular ecology notes 5:712-715. Guillot, G. and F. Rousset. 2013. Dismantling the Mantel tests. Methods in Ecology and Evolution 4:336-344. Guillot, G. and F. Santos. 2009. A computer program to simulate multilocus genotype data with spatially autocorrelated allele frequencies. Molecular Ecology Resources 9:1112-1120. Hadfield, J. D. 2010. MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R package. Journal of Statistical Software 33:1-22.

137 Hadrys, H., W. Schroth, B. Schierwater, B. Streit, and O. M. Fincke. 2005. Tree hole odonates as environmental monitors: Non-invasive isolation of polymorphic microsatellites from the neotropical damselfly Megaloprepus caerulatus. Conservation Genetics 6:481-483. Hafner, M. S., T. A. Spradling, J. E. Light, D. J. Hafner, and J. R. Demboski. 2004. Systematic revision of pocket gophers of the Cratogeomys gymnurus species group. Journal of Mammalogy 85:1170-1183. Hale, J., G. Heard, K. Smith, M. Blacket, J. Austin, and J. Melville. 2011. Microsatellite loci for the endangered growling grass frog (Litoria raniformis), with cross amplification in other Australian frog species. Conservation Genetics Resources 3:593-595. Hale, J., G. Heard, K. Smith, K. Parris, J. Austin, M. Kearney, and J. Melville. 2013. Structure and fragmentation of growling grass frog metapopulations. Conservation Genetics 14:313-322. Hale, J. M. 2010. Human-induced changes in the population genetic structure of frogs in south-eastern Australia. PhD thesis, The University of Melbourne, Victoria. Hall, L. A. and S. R. Beissinger. 2014. A practical toolbox for design and analysis of landscape genetics studies. Landscape Ecology 29:1487-1504. Hamer, A. J., S. J. Lane, and M. J. Mahony. 2010. Using probabilistic models to investigate the disappearance of a widespread frog-species complex in high- altitude regions of south-eastern Australia. Animal Conservation 13:275-285. Hamer, A. J. and M. J. McDonnell. 2008. Amphibian ecology and conservation in the urbanising world: a review. Biological Conservation 141:2432-2449. Hamer, A. J. and A. Organ. 2008. Aspects of the ecology and conservation of the growling grass frog Litoria raniformis in an urban-fringe environment, southern Victoria. Australian Zoologist 34:393-407. Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal Ecology 63:151-162. Hanski, I. 1999. Metapopulation Ecology. Oxford University Press, Oxford. Hanski, I. 2002. Metapopulations of animals in highly fragmented landscapes and population viability analysis. Pages 86-108 in S. R. Beissinger and D. R. McCullough, editors. Population Viability Analysis. The University of Chicago Press, Chicago.

138 Hanski, I. and M. E. Gilpin. 1997. Metapopulation biology: ecology, genetics, and evolution. Academic press San Diego. Harpending, H. 1994. Signature of ancient population growth in a low-resolution mitochondrial DNA mismatch distribution. Human Biology 66:591-600. Hartl, D. L. and A. G. Clark. 2007. Principles of population genetics. 4th edition. Sinauer associates Sunderland, MA. Hazell, D., R. Cunnningham, D. Lindenmayer, B. Mackey, and W. Osborne. 2001. Use of farm dams as frog habitat in an Australian agricultural landscape: factors affecting species richness and distribution. Biological Conservation 102:155- 169. Heard, G. W. 2010. Pattern, process and the conservation of threatened amphibian metapopulations. PhD thesis, La Trobe University, Bundoora, Victoria. Heard, G. W., M. A. McCarthy, M. P. Scroggie, J. B. Baumgartner, and K. M. Parris. 2013. A Bayesian model of metapopulation viability, with application to an endangered amphibian. Diversity and Distributions 19:555-566. Heard, G. W., P. Robertson, and M. P. Scroggie. 2008. Microhabitat preferences of the endangered Growling Grass Frog (Litoria raniformis) in southern Victoria, Australia. Australian Zoologist 34:414-425. Heard, G. W. and M. P. Scroggie. 2009. Assessing the impacts of urbanisation on Growling Grass Frog metapopulations. Report to the Victorian Department of Sustainability and Environment Arthur Rylah Institute for Environmental Research, Heidelberg, Victoria. Heard, G. W., M. P. Scroggie, and N. Clemann. 2010. Guidelines for managing the endangered Growling Grass Frog in urbanising landscapes. Department of Sustainability and Environment, Heidelberg, Victoria. Heard, G. W., M. P. Scroggie, N. Clemann, and D. S. Ramsey. 2014. Wetland characteristics influence disease risk for a threatened amphibian. Ecological Applications 24:650-662. Heard, G. W., M. P. Scroggie, and B. S. Malone. 2012a. Classical metapopulation theory as a useful paradigm for the conservation of an endangered amphibian. Biological Conservation 148:156-166. Heard, G. W., M. P. Scroggie, and B. S. Malone. 2012b. The life history and decline of the threatened Australian frog, Litoria raniformis. Austral Ecology 37:276-284.

139 Heard, G. W., C. D. Thomas, J. A. Hodgson, M. P. Scroggie, D. S. L. Ramsey, and N. Clemann. 2015. Refugia and connectivity sustain amphibian metapopulations afflicted by disease. Ecology Letters 18:853-863. Hedrick, P. W. and S. T. Kalinowski. 2000. Inbreeding Depression in Conservation Biology. Annual Review of Ecology & Systematics 31:139-162. Hels, T. and E. Buchwald. 2001. The effects of road kills on amphibian populations. Biological Conservation 99:331-340. Hilsdorf, A., D. Caneppele, and J. E. Krieger. 1999. Muscle biopsy technique for electrophoresis analysis of fish from the genus Brycon. Genetics and Molecular Biology 22:547-550. Hitchings, S. P. and T. J. C. Beebee. 1997. Genetic substructuring as a result of barriers to gene flow in urban Rana temporaria (common frog) populations: implications for biodiversity conservation. Heredity 79:117-127. Hitchings, S. P. and T. J. C. Beebee. 1998. Loss of genetic diversity and fitness in Common Toad (Bufo bufo) populations isolated by inimical habitat. Journal of evolutionary biology 11:269-283. Holderegger, R. and H. H. Wagner. 2008. Landscape genetics. BioScience 58:199-207. Holehouse, K. A., R. L. Hammond, and A. F. G. Bourke. 2003. Non-lethal sampling of DNA from bumble bees for conservation genetics. Insectes Sociaux 50:277-285. Hubisz, M. J., D. Falush, M. Stephens, and J. K. Pritchard. 2009. Inferring weak population structure with the assistance of sample group information. Molecular Ecology Resources 9:1322-1332. Huff, D. D., L. M. Miller, C. J. Chizinski, and B. Vondracek. 2011. Mixed-source reintroductions lead to outbreeding depression in second-generation descendents of a native North American fish. Molecular ecology 20:4246-4258. Idaghdour, Y., D. Broderick, and A. Korrida. 2003. Faeces as a source of DNA for molecular studies in a threatened population of great bustards. Conservation Genetics 4:789-792. Irwin, E. G. and N. E. Bockstael. 2007. The evolution of urban sprawl: evidence of spatial heterogeneity and increasing land fragmentation. Proceedings of the National Academy of Sciences 104:20672-20677. IUCN. 2013. IUCN Red List of Threatened Species. Version 2013.1. IUCN. 2015. The IUCN Red List of Threatened Species. Version 2015-4.

140 Jaquiery, J., T. Broquet, A. Hirzel, J. Yearsley, and N. Perrin. 2011. Inferring landscape effects on dispersal from genetic distances: how far can we go? Molecular ecology. Jehle, R. 2010. Herpetology and conservation genetics. Animal Conservation 13:72-73. Jones, R., J. Cable, and M. W. Bruford. 2008. An evaluation of non-invasive sampling for genetic analysis in northern European reptiles. The Herpetological Journal 18:32-39. Kalinowski, S. T. 2005. HP-RARE 1.0: a computer program for performing rarefaction on measures of allelic richness. Molecular ecology notes 5:187-189. Keely, C. C., J. M. Hale, G. W. Heard, K. M. Parris, J. Sumner, A. J. Hamer, and J. Melville. 2015. Genetic structure and diversity of the endangered growling grass frog in a rapidly urbanizing region. 2:1-13. Keller, D., R. Holderegger, M. van Strien, and J. Bolliger. 2015. How to make landscape genetics beneficial for conservation management? Conservation Genetics 16:503-512. Keller, L. F., K. J. Jeffery, P. Arcese, M. A. Beaumont, W. M. Hochachka, J. N. Smith, and M. W. Bruford. 2001. Immigration and the ephemerality of a natural population bottleneck: evidence from molecular markers. Proceedings of the Royal Society of London. Series B: Biological Sciences 268:1387-1394. Kendall, K. C., J. B. Stetz, J. Boulanger, A. C. Macleod, D. Paetkau, and G. C. White. 2009. Demography and genetic structure of a recovering grizzly bear population. The Journal of Wildlife Management 73:3-16. Keyghobadi, N., L. A. Crawford, and S. A. Maxwell. 2009. Successful analysis of AFLPs from non-lethally sampled wing tissues in butterflies. Conservation Genetics 10:2021-2024. Koehler, S., D. Gilmore, and D. Newell. 2014a. Translocation of the threatened Growling Grass Frog Litoria raniformis: a case study. Australian Zoologist 37:321-336. Koehler, S. L., D. C. Gilmore, and D. A. Newell. 2014b. Translocation of the threatened Growling Grass Frog Litoria raniformis: a case study. Australian Zoologist:1- 16. Koeller, C. A. 2009. Comparison of buprenorphine and butorphanol analgesia in the eastern red-spotted newt (Notophthalmus viridescens). Journal of the American Association for Laboratory Animal Science 48:171.

141 Lecomte, N., G. Gauthier, L. Bernatchez, and J.-F. Giroux. 2006. A nondamaging blood sampling technique for waterfowl embryos. Journal of Field Ornithology 77:67- 70. Lesbarreres, D., C. R. Primmer, T. Lode, and J. Merila. 2006. The effects of 20 years of highway presence on the genetic structure of Rana dalmatina populations. Ecoscience 13:531-538. Librado, P. and J. Rozas. 2009. DnaSP v5: a software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25:1451-1452. Lindenmayer, D. B. and J. Fischer. 2006. Habitat fragmentation and landscape change: an ecological and conservation synthesis. Island Press, Washington. Lode, T. 2000. Effect of a motorway on mortality and isolation of wildlife populations. AMBIO: A Journal of the Human Environment 29:163-166. Lowe, W. H. and F. W. Allendorf. 2010. What can genetics tell us about population connectivity? Molecular ecology 19:3038-3051. Lucentini, L., S. Caporali, A. Palomba, H. Lancioni, and F. Panara. 2006. A comparison of conservative DNA extraction methods from fins and scales of freshwater fish: A useful tool for conservation genetics. Conservation Genetics 7:1009-1012. Lynch, M. J., M. A. Long, and P. B. Stretesky. 2015. Anthropogenic development drives species to be endangered: Capitalism and the decline of species. Pages 117-146 Green Harms and Crimes. Springer. Machin, K. L. 1999. Amphibian Pain and Analgesia. Journal of Zoo and Wildlife Medicine 30:2-10. MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines. 2006. Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence. Elsevier, San Diego. Mahony, M. 1999. Review of the declines and disappearances within the bell frog species group (Litoria aurea species group) in Australia. Pages 81-93 in A. Campbell, editor. Declines and Disappearances of Australian Frogs. Environment Australia, Canberra. Mahony, M. J., A. J. Hamer, E. J. Pickett, D. J. McKenzie, M. P. Stockwell, J. I. Garnham, C. C. Keely, M. L. Deboo, J. O'Meara, C. J. Pollard, S. Clulow, F. L. Lemckert, D. S. Bower, and J. Clulow. 2013. Identifying conservation and research priorities in the face of uncertainty: A review of the threatened bell frog

142 complex in eastern Australia. HERPETOLOGICAL CONSERVATION AND BIOLOGY 8:519-538. Manel, S. and R. Holderegger. 2013. Ten years of landscape genetics. Trends in Ecology & Evolution 28:614-621. Manel, S., M. K. Schwartz, G. Luikart, and P. Taberlet. 2003. Landscape genetics: combining landscape ecology and population genetics. Trends in Ecology and Evolution 18:189-197. Manoukis, N. C. 2007. FORMATOMATIC: a program for converting diploid allelic data between common formats for population genetic analysis. Molecular ecology notes 7:592-593. Marantelli, G., L. Berger, R. Speare, and L. Keegan. 2004. Distribution of the amphibian chytrid Batrachochytrium dendrobatidis and keratin during tadpole development. Pacific Conservation Biology 10:173. Marsh, D. M., B. J. Cosentino, K. S. Jones, J. J. Apodaca, K. H. Beard, J. M. Bell, C. Bozarth, D. Carper, J. F. Charbonnier, and A. Dantas. 2017. Effects of roads and land use on frog distributions across spatial scales and regions in the Eastern and Central United States. Diversity and Distributions. Marsh, D. M., E. H. Fegraus, and S. Harrison. 1999. Effects of breeding pond isolation on the spatial and temporal dynamics of pond use by the tungara frog, Physalaemus pustulosus. Journal of Animal Ecology 68:804-814. Marsh, D. M., R. B. Page, T. J. Hanlon, R. Corritone, E. C. Little, D. E. Seifert, and P. R. Cabe. 2008. Effects of roads on patterns of genetic differentiation in red- backed salamanders, Plethodon cinereus. Conservation Genetics 9:603-613. Marsh, D. M. and P. C. Trenham. 2001. Metapopulation dynamics and amphibian conservation. Conservation Biology 15:40-49. Martin-Galvez, D., J. M. Peralta-Sanchez, D. A. Dawson, A. M. Martin-Platero, M. Martinez-Bueno, T. Burke, and J. J. Soler. 2011. DNA sampling from eggshell swabbing is widely applicable in wild bird populations as demonstrated in 23 species. Molecular Ecology Resources 11:481-493. May, R. M. 2004. Ethics and amphibians. Nature 431:403. McCarthy, M. A. and K. M. Parris. 2004. Clarifying the effect of toe clipping on frogs with Bayesian statistics. Journal of Applied Ecology 41:780-786.

143 McClenaghan, L. R. and H. D. Truesdale. 2002. Genetic structure of endangered Stephens' kangaroo rat populations in southern California. Southwestern Naturalist 47:539-549. McDonnell, M. J. and S. T. A. Pickett. 1990. Ecosystem structure and function along urban-rural gradients: an unexploited opportunity for ecology. Ecology 71:1232- 1237. McIntyre, N. E. and J. J. Rango. 2009. Arthropods in urban ecosystems: community patterns as functions of anthropogenic land use. Pages 233-242 in M. J. McDonnell, A. K. Hahs, and J. H. Breuste, editors. Ecology of Cities and Towns: A Comparative Approach. Cambridge University Press, Cambridge, UK. McKinney, M. L. 2002. Urbanization, biodiversity, and conservation. BioScience 52:883-890. McRae, B. H. 2006. Isolation by resistance. Evolution 60:1551-1561. McRae, B. H. and P. Beier. 2007. Circuit theory predicts gene flow in plant and animal populations. Proceedings of the National Academy of Sciences 104:19885. McRae, B. H., B. G. Dickson, T. H. Keitt, and V. B. Shah. 2008. Using circuit theory to model connectivity in ecology, evolution, and conservation. Ecology 89:2712- 2724. Mikulicek, P. and P. Pisut. 2012. Genetic structure of the marsh frog (Pelophylax ridibundus) populations in urban landscape. European journal of wildlife research 58:833-845. Miller, H. 2006. Cloacal and buccal swabs are a reliable source of DNA for microsatellite genotyping of reptiles. Conservation Genetics 7:1001-1003. Miller, J. R. and R. J. Hobbs. 2002. Conservation where people live and work. Conservation Biology 16:330-337. Minteer, B. A. and J. P. Collins. 2005. Why we need an "ecological ethics". Frontiers in Ecology and the Environment 3:332-337. Moilanen, A. and I. Hanski. 1998. Metapopulation dynamics: effects of habitat quality and landscape structure. Ecology 79:2503-2515. Moore, J. L. and M. C. Runge. 2012. Combining Structured Decision Making and Value-of-Information Analyses to Identify Robust Management Strategies. Conservation Biology 26:810-820.

144 Moritz, C. 1994. Defining 'evolutionarily significant units' for conservation. Trends in Ecology and Evolution 9:373-375. Moritz, C. 2002. Strategies to protect biological diversity and the evolutionary processes that sustain it. Systematic Biology 51:238-254. Muller, A. S., P. P. Lenhardt, and K. Theissinger. 2013. Pros and cons of external swabbing of amphibians for genetic analyses. European journal of wildlife research 59:609-612. Munshi-South, J. 2012. Urban landscape genetics: canopy cover predicts gene flow between white-footed mouse (Peromyscus leucopus) populations in New York City. Molecular ecology 21:1360-1378. Munshi-South, J., Y. Zak, and E. Pehek. 2013. Conservation genetics of extremely isolated urban populations of the northern dusky salamander (Desmognathus fuscus) in New York City. PeerJ 1:e64. Murcia, C. 1995. Edge effects in fragmented forests: implications for conservation. Trends in Ecology & Evolution 10:58-62. Murphy, M. A., R. Dezzani, D. S. Pilliod, and A. Storfer. 2010. Landscape genetics of high mountain frog metapopulations. Molecular ecology 19:3634-3649. Murray, K. A., R. W. Retallick, R. Puschendorf, L. F. Skerratt, D. Rosauer, H. I. McCallum, L. Berger, R. Speare, and J. VanDerWal. 2011. Assessing spatial patterns of disease risk to biodiversity: implications for the management of the amphibian pathogen, Batrachochytrium dendrobatidis. Journal of Applied Ecology 48:163-173. Nakagawa, S. and H. Schielzeth. 2013. A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution 4:133-142. Narayan, E. J., F. C. Molinia, C. Kindermann, J. F. Cockrem, and J. M. Hero. 2011. Urinary corticosterone responses to capture and toe-clipping in the cane toad (Rhinella marina) indicate that toe-clipping is a stressor for amphibians. General and Comparative Endocrinology 174:238-245. Noble, D. W., J. S. Keogh, and M. J. Whiting. 2013. Multiple mating in a lizard increases fecundity but provides no evidence for genetic benefits. Behavioral Ecology 24:1128-1137. Noel, S. and F.-J. Lapointe. 2010. Urban conservation genetics: Study of a terrestrial salamander in the city. Biological Conservation 143:2823-2831.

145 Noel, S., M. Ouellet, P. Galois, and F.-J. Lapointe. 2007. Impact of urban fragmentation on the genetic structure of the eastern red-backed salamander. Conservation Genetics 8:599-606. Nota, Y. and O. Takenaka. 1999. DNA extraction from urine and sex identification of birds. Molecular ecology 8:1237-1238. Ovaskainen, O. 2004. Habitat-specific movement parameters estimated using mark- recapture data and a diffusion model. Ecology 85:242-257. Pannell, J. R. and B. Charlesworth. 2000. Effects of Metapopulation Processes on Measures of Genetic Diversity. Philosophical Transactions: Biological Sciences 355:1851-1864. Parra-Olea, G., K. R. Zamudio, E. Recuero, X. Aguilar-Miguel, D. Huacuz, and L. Zambrano. 2012. Conservation genetics of threatened Mexican axolotls (Ambystoma). Animal Conservation 15:61-72. Parris, K. M. 2006. Urban amphibian assemblages as metacommunities. Journal of Animal Ecology 75:757-764. Parris, K. M., S. C. McCall, M. A. McCarthy, B. A. Minteer, K. Steele, S. Bekessy, and F. Medvecky. 2010. Assessing ethical trade-offs in ecological field studies. Journal of Applied Ecology 47:227-234. Parris, K. M. and M. A. McCarthy. 2008. Toe clipping of anurans for mark-recapture studies: acceptable if justified Herpetological Review 39:148-149. Parris, K. M., M. Velik-Lord, and J. M. A. North. 2009. Frogs Call at a Higher Pitch in Traffic Noise. Ecology and Society 14:Article No.: 25. Peakall, R. and P. E. Smouse. 2006a. GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular ecology notes 6:288-295. Peakall, R. and P. E. Smouse. 2006b. GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular ecology notes 6:288-295. Peery, M. Z., R. Kirby, B. N. Reid, R. Stoelting, E. Doucet-Beer, S. Robinson, C. Vasquez-Carrillo, J. N. Pauli, and P. J. Palsboll. 2012. Reliability of genetic bottleneck tests for detecting recent population declines. Molecular ecology 21:3403-3418. Petit, E. and N. Valiere. 2006. Estimating population size with noninvasive capture- mark-recapture data. Conservation Biology 20:1062-1073.

146 Phillott, A. D., L. F. Skerrat, K. R. McDonald, F. L. Lemckert, H. B. Hines, J. M. Clarke, R. A. Alford, and R. Speare. 2007. Toe-clipping as an acceptable method for identifying individual anurans in mark recapture studies. Herpetological Review 38:305-308. Pickett, S. T., M. L. Cadenasso, D. L. Childers, M. J. McDonnell, and W. Zhou. 2016. Evolution and future of urban ecological science: ecology in, of, and for the city. Ecosystem Health and Sustainability 2. Pidancier, N., C. Miquel, and C. Miaud. 2003. Buccal swabs as a non-destructive tissue sampling method for DNA analysis in amphibians. Herpetological Journal 13:175-178. Pierszalowski, S. P., M. Ferrari, D. Glockner-Ferrari, S. Mizroch, P. J. Clapham, and B. R. Dickerson. 2013. Investigating the feasibility of using DNA from sloughed skin for individual identification and kinship analysis in humpback whales (Megaptera novaeangliae). Marine mammal science 29:533-541. Piry, S., G. Luikart, and J. Cornuet. 1999. Bottleneck: a computer program for detecting recent reductions in the effective population size using allele frequency data. Journal of Heredity 90:502-503. Population Reference Bureau. 2009. World Population data Sheet. Porej, D., M. Micacchion, and T. E. Hetherington. 2004. Core terrestrial habitat for conservation of local populations of salamanders and wood frogs in agricultural landscapes. Biological Conservation 120:399-409. Porter, J. H. and J. L. Dooley, Jr. 1993. Animal Dispersal Patterns: A Reassessment of Simple Mathematical Models. Ecology 74:2436-2443. Poschadel, J. R. and D. Möller. 2004. A versatile field method for tissue sampling on small reptiles and amphibians, applied to pond turtles, newts, frogs and toads. Conservation Genetics 5:865-867. Pritchard, J., X. Wen, and D. Falush. 2010. Documentation for structure software: Version 2.3. Universtiy of Chicago, Chicago:1-38. Pritchard, J. K., M. Stephens, and P. Donnelly. 2000. Inference of population structure using multilocus genotype data. Genetics 155:945-959. Prunier, J., B. Kaufmann, O. Grolet, D. Picard, F. Pompanon, and P. Joly. 2012. Skin swabbing as a new efficient DNA sampling technique in amphibians, and 14 new microsatellite markers in the alpine newt (Ichthyosaura alpestris). Molecular Ecology Resources 12:524-531.

147 Pyke, G. H. 2002. A review of the biology of the southern bell frog Litoria raniformis (Anura: Hylidae). Australian Zoologist 32:32-48. R Core Team. 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Rambaut, A., M. A. Suchard, W. Xie, and A. J. Drummond. 2003-2014. Tracer v 1. 6: An MCMC Trace Analysis Tool. Institute of Evolutionary Biology, University of Edinburgh http://tree.bio.ed.ac.uk/software/tracer/. Rannala, B. 2012. BayesAss edition 3.0 user's manual. University of California Davis. Reed, D. H. and R. Frankham. 2003. Correlation between fitness and genetic diversity. Conservation Biology 17:230-237. Reed, D. H. and G. R. Hobbs. 2004. The relationship between population size and temporal variability in population size. Animal Conservation 7:1-8. Reh, W. and A. Seitz. 1990. The influence of land use on the genetic structure of populations of the common frog Rana temporaria. Biological Conservation 54:239-249. Ricketts, T. H. 2001. The matrix matters: effective isolation in fragmented landscapes. The American Naturalist 158:87-99. Riley, S. P. D., J. P. Pollinger, R. M. Sauvajot, E. C. York, C. Bromley, T. K. Fuller, and R. K. Wayne. 2006. A southern California freeway is a physical and social barrier to gene flow in carnivores. Molecular ecology 15:1733-1742. Rivera-Ortíz, F. A., R. Aguilar, M. D. C. Arizmendi, M. Quesada, and K. Oyama. 2014. Habitat fragmentation and genetic variability of tetrapod populations. Animal Conservation. Robinson, N. A. and C. A. Marks. 2001. Genetic structure and dispersal of red foxes (Vulpes vulpes) in urban Melbourne. Australian Journal of Zoology 49:589-601. Rose, L. E., G. W. Heard, Y. E. Chee, and B. A. Wintle. 2016. Cost-effective conservation of an endangered frog under uncertainty. Conservation Biology 30:350-361. Rothermel, B. B. 2004. Migratory success of juveniles: A potential constraint on connectivity for pond-breeding amphibians. Ecological Applications 14:1535- 1546. Rothermel, B. B. and R. D. Semlitsch. 2002. An experimental investigation of landscape resistance of forest versus old-field habitats to emigrating juvenile amphibians. Conservation Biology 16:1324-1332.

148 Rousset, F. 1997. Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance. Genetics 145:1219-1228. Rousset, F. 2008. Genepop'007: a complete re-implementation of the genepop software for Windows and Linux. Molecular Ecology Resources 8:103-106. Row, J. R., R. J. Brooks, C. A. MacKinnon, A. Lawson, B. I. Crother, M. White, and S. C. Lougheed. 2011. Approximate Bayesian computation reveals the factors that influence genetic diversity and population structure of foxsnakes. Journal of evolutionary biology 24:2364-2377. Royle, J. A. and M. Kéry. 2007. A Bayesian state-space formulation of dynamic occupancy models. Ecology 88:1813-1823. Rytwinski, T., K. Soanes, J. A. Jaeger, L. Fahrig, C. S. Findlay, J. Houlahan, R. Van Der Ree, and E. A. van der Grift. 2016. How effective is road mitigation at reducing road-kill? A meta-analysis. PloS one 11. Saarikivi, J., T. Knopp, A. Granroth, and J. MerilD. 2013. The role of golf courses in maintaining genetic connectivity between common frog (Rana temporaria) populations in an urban setting. Conservation Genetics 14:1057-1064. Sadanandan, K. R. and F. E. Rheindt. 2015. Genetic diversity of a tropical rainforest understory bird in an urban fragmented landscape. The Condor 117:447-459. Safner, T., C. Miaud, O. Gaggiotti, S. Decout, D. Rioux, S. Zundel, and S. Manel. 2010. Combining demography and genetic analysis to assess the population structure of an amphibian in a human-dominated landscape. Conservation Genetics:1-13. Sato, J. J., T. Kawakami, Y. Tasaka, M. Tamenishi, and Y. Yamaguchi. 2014. A few decades of habitat fragmentation has reduced population genetic diversity: A case study of landscape genetics of the large Japanese field mouse, Apodemus speciosus. Mammal Study 1:1-10. Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological consequences of ecosystem fragmentation: a review. Conservation Biology 5:18-32. Schmeller, D. S. and J. Merilä. 2007. Demographic and genetic estimates of effective population and breeding size in the amphibian Rana temporaria. Conservation Biology 21:142-151. Schneider, C., M. Cunningham, and C. Moritz. 1998. Comparative phylogeography and the history of endemic vertebrates in the Wet Tropics rainforests of Australia. Molecular ecology 7:487-498.

149 Segelbacher, G. 2002. Noninvasive genetic analysis in birds: testing reliability of feather samples. Molecular ecology notes 2:367-369. Segelbacher, G., S. A. Cushman, B. K. Epperson, M.-J. Fortin, O. Francois, O. J. Hardy, R. Holderegger, P. Taberlet, L. P. Waits, and S. Manel. 2010. Applications of landscape genetics in conservation biology: concepts and challenges. Conservation Genetics 11:375-385. Seiler, A., J. Helldin, and C. Seiler. 2004. Road mortality in Swedish mammals: results of a drivers' questionnaire. Wildlife Biology 10:225-233. Semlitsch, R., D. 1998. Biological delineation of terrestrial buffer zones for pond- breeding salamanders. Conservation Biology 12:1113-1119. Semlitsch, R. D. 1981. Terrestrial activity and summer home range of the mole salamander Ambystoma talpoideum. Canadian Journal of Zoology 59:315-322. Semlitsch, R. D. 2008. Differentiating migration and dispersal processes for pond- breeding amphibians. The Journal of Wildlife Management 72:260-267. Seto, K. C., B. Guneralp, and L. R. Hutyra. 2012. Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proceedings of the National Academy of Sciences 109:16083-16088. Shafer, A. B., J. B. Wolf, P. C. Alves, L. Bergstrom, M. W. Bruford, I. Brannstrom, G. Colling, L. Dalen, L. De Meester, and R. Ekblom. 2015. Genomics and the challenging translation into conservation practice. Trends in Ecology & Evolution 30:78-87. Shine, R., M. Lemaster, M. Wall, T. Langkilde, and R. Mason. 2004. Why Did the Snake Cross the Road? Effects of Roads on Movement and Location of Mates by Garter Snakes (Thamnophis sirtalis parietalis). Ecology and Society 9:9. Sinsch, U. 1997. Postmetamorphic dispersal and recruitment of first breeders in a Bufo calamita metapopulation. Oecologia 112:42-47. Sinsch, U. and D. Seidel. 1995. Dynamics of local and temporal breeding assemblages in a Bufo calamita metapopulation. Australian Journal of Ecology 20:351-361. Sjögren-Gulve, P. and I. Hanski. 2000. Metapopulation viability analysis using occupancy models. Ecological Bulletins 48:53-71. Sloane, M., P. Sunnucks, D. Alpers, L. Beheregaray, and A. Taylor. 2000. Highly reliable genetic identification of individual northern hairy-nosed wombats from single remotely collected hairs: a feasible censusing method. Molecular ecology 9:1233-1240.

150 Smallbone, L. T., G. W. Luck, and S. Wassens. 2011. Anuran species in urban landscapes: Relationships with biophysical, built environment and socio- economic factors. Landscape and Urban Planning. Smalley, J. V. and J. J. Campanella. 2005. Buccal swabbing and extraction of high quality sunfish (Lepomis) DNA for use in PCR analysis. Biotechniques 38:188- 192. Smith, K. L., J. M. Hale, M. R. Kearney, J. J. Austin, and J. Melville. 2013. Molecular patterns of introgression in a classic hybrid zone between the Australian tree frogs, Litoria ewingii and L. paraewingi: evidence of a tension zone. Molecular ecology. Smith, M. A. and D. M. Green. 2005. Dispersal and the metapopulation paradigm in amphibian ecology and conservation: are all amphibians metapopulations? . Ecography 28:110-128. Smith, M. A., N. A. Poyarkov, Jr., and P. D. N. Hebert. 2008. CO1 DNA barcoding amphibians: take the chance, meet the challenge. Molecular Ecology Resources 8:235-246. Smith, S. and J. Hughes. 2008. Microsatellite and mitochondrial DNA variation defines island genetic reservoirs for reintroductions of an endangered Australian marsupial, Perameles bougainville. Conservation Genetics 9:547-557. Sol, D., C. Gonzalez-Lagos, D. Moreira, J. Maspons, and O. Lapiedra. 2014. Urbanisation tolerance and the loss of avian diversity. Ecology Letters 17:942- 950. Sork, V. L. and L. Waits. 2010. Contributions of landscape genetics – approaches, insights, and future potential. Molecular ecology 19:3489-3495. Spear, S. F., N. Balkenhol, M. J. Fortin, B. H. McRae, and K. Scribner. 2010. Use of resistance surfaces for landscape genetic studies: considerations for parameterization and analysis. Molecular ecology 19:3576-3591. Spear, S. F., C. R. Peterson, M. D. Matocq, and A. Storfer. 2005. Landscape genetics of the blotched tiger salamander (Ambystoma tigrinum melanostictum). Molecular ecology 14:2553-2564. Spiegelhalter, D. J., N. G. Best, B. P. Carlin, and A. Van Der Linde. 2002. Bayesian measures of model complexity and fit. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 64:583-639.

151 State of Victoria. 2013. Plan Melbourne, Metropolitan Planning Strategy. Victorian State Government, Melbourne, Victoria. Stephens, S. H., V. C. White, and J. R. A. Bremer. 2010. Assessment of Genetic Tissue Sampling Methods for the Critically Endangered Kemp's Ridley Sea Turtle (Lepidochelys kempii). Marine Turtle Newsletter 128:19-23. Storfer, A., M. A. Murphy, J. S. Evans, C. S. Goldberg, S. Robinson, S. F. Spear, R. Dezzani, E. Delmelle, L. Vierling, and L. P. Waits. 2007. Putting the 'landscape' in landscape genetics. Heredity 98:128-142. Storfer, A., M. A. Murphy, S. F. Spear, R. Holderegger, and L. P. Waits. 2010. Landscape genetics: where are we now? Molecular ecology 19:3496-3514. Straub, C., F. Pichlmuller, and V. Helfer. 2015. Population genetics of fire salamanders in a pre-Alpine urbanized area (Salzburg, Austria). SALAMANDRA 51:245- 251. Strausberger, B. and M. Ashley. 2001. Eggs yield nuclear DNA from egg-laying female cowbirds, their embryos and offspring. Conservation Genetics 2:385-390. Stuart, S. N., J. S. Chanson, N. A. Cox, B. E. Young, A. S. L. Rodrigues, D. L. Fischman, and R. W. Waller. 2004. Status and trends of amphibian declines and extinctions worldwide. Science 306:1783-1786. Stuart-Fox, D. M., C. J. Schneider, C. Moritz, and P. J. Couper. 2001. Comparative phylogeography of three rainforest-restricted lizards from mid-east Queensland. Australian Journal of Zoology 49:119-127. Sugiyama, Y., S. Kawamoto, O. Takenaka, K. Kumazaki, and N. Miwa. 1993. Paternity discrimination and inter-group relationships of chimpanzees at Bossou. Primates 34:545-552. Sumner, J., J. K. Webb, R. Shine, and J. S. Keogh. 2010. Molecular and morphological assessment of Australia's most endangered snake, Hoplocephalus bungaroides, reveals two evolutionarily significant units for conservation. Conservation Genetics 11:747-758. Sunnucks, P. and A. C. Taylor. 2008. The application of genetic markers to landscape management. Pages 211-233 in W. C. Christopher Pettit, Ian Bishop, Kim Lowell, David Pullar and David Duncan, editor. Landscape Analysis and Visualisation. Springer Verlag, Berlin. Taberlet, P. and L. Fumagalli. 1996. Owl pellets as a source of DNA for genetic studies of small mammals. Molecular ecology 5:301-305.

152 Taberlet, P., L. P. Waits, and G. Luikart. 1999. Noninvasive genetic sampling: Look before you leap. Trends in Ecology and Evolution 14:323-327. Tajima, F. 1989. The effect of change in population size on DNA polymorphism. Genetics 123:597-601. Tanaka-Ueno, T., M. Matsui, Y. Ueno, and O. Takenaka. 2006. Noninvasive sampling methods for genotyping amphibians: new application for conservation genetics. Herpetological Review 37:312-316. Thomas, A., B. O'Hara, U. Ligges, and S. Sturtz. 2006. Making BUGS Open. R News 6:12-17. Tobe, S. S. and A. M. Linacre. 2008. A multiplex assay to identify 18 European mammal species from mixtures using the mitochondrial cytochrome b gene. Electrophoresis 29:340-347. Tracy, L. N., G. P. Wallis, M. G. Efford, and I. G. Jamieson. 2011. Preserving genetic diversity in threatened species reintroductions: how many individuals should be released? Animal Conservation 14:439-446. UNFPA. 2007. State of world population 2007: unleashing the potential of urban growth. United Nations Population Fund, New York, USA. United Nations. 2010. World Urbanization Prospects: The 2009 revision. United Nations, New York. Valiere, N. and P. Taberlet. 2000. Urine collected in the field as a source of DNA for species and individual identification. Molecular ecology 9:2150-2152. van Oosterhout, C., W. F. Hutchinson, D. P. Wills, and P. Shipley. 2004. Micro- Checker: software for identifying and correcting genotyping errors in microsatellite data. Molecular ecology notes 4:535-538. Van Strien, M. J., D. Keller, and R. Holderegger. 2012. A new analytical approach to landscape genetic modelling: least-cost transect analysis and linear mixed models. Molecular ecology 21:4010-4023. Vandergast, A. G., A. J. Bohonak, D. B. Weissman, and R. N. Fisher. 2007. Understanding the genetic effects of recent habitat fragmentation in the context of evolutionary history: phylogeography and landscape genetics of a southern California endemic Jerusalem cricket (Orthoptera: Stenopelmatidae: Stenopelmatus). Molecular ecology 16:977-992. Vörös, J., A. Mitchell, B. Waldman, S. Goldstien, and N. J. Gemmell. 2008. Crossing the Tasman Sea: inferring the introduction history of Litoria aurea and Litoria

153 raniformis (Anura: Hylidae) from Australia in to New Zealand. Austral Ecology 33:623-629. Vos, C. C. and J. P. Chardon. 1998. Effects of habitat fragmentation and road density on the distribution pattern of the moor frog Rana arvalis. Journal of Applied Ecology 35:44-56. Vos, C. C., C. J. F. ter Braak, and W. Nieuwenhuizen. 2000. Incidence function modelling and conservation of the tree frog (Hyla arborea) in the Netherlands. Ecological Bulletin 48:165-180. Waddle, J. H., K. G. Rice, F. J. Mazzotti, and H. F. Percival. 2008. Modeling the effect of toe clipping on treefrog survival: beyond the return rate. Journal of Herpetology 42:467-473. Waits, L. P., S. A. Cushman, and S. F. Spear. 2016. Applications of Landscape Genetics to Connectivity Research in Terrestrial Animals. Landscape genetics: concepts, methods, applications. Wiley:199-219. Wasko, A. P., C. Martins, C. Oliveira, and F. Foresti. 2003. Non-destructive genetic sampling in fish. An improved method for DNA extraction from fish fins and scales. Hereditas 138:161-165. Wassens, S., A. Hall, W. Osborne, and R. J. Watts. 2010. Habitat characteristics predict occupancy patterns of the endangered amphibian Litoria raniformis in flow- regulated flood plain wetlands. Austral Ecology 35:944-955. Wassens, S., R. J. Watts, A. Jansen, and D. Roshier. 2008. Movement patterns of southern bell frogs (Litoria raniformis) in response to flooding. Wildlife Research 35:50-58. Wellbrock, A. H., C. Bauch, J. Rozman, and K. Witte. 2012. Buccal swabs as a reliable source of DNA for sexing young and adult Common Swifts (Apus apus). Journal of Ornithology 153:991-994. Westphal, M. I., M. Pickett, W. M. Getz, and H. P. Possingham. 2003. The use of stochastic dynamic programming in optimal landscape reconstruction for metapopulations. Ecological Applications 13:543-555. White, A. W. and G. H. Pyke. 2008. Frogs on the hop: translocations of Green and Golden Bell Frogs Litoria aurea in Greater Sydney. Australian Zoologist 34:249-260.

154 Wiens, J. J., J. W. Fetzner Jr, C. L. Parkinson, and T. W. Reeder. 2005. Hylid frog phylogeny and sampling strategies for speciose clades. Systematic Biology 54:778-807. Williams, C. L., K. Blejwas, J. J. Johnston, and M. M. Jaeger. 2003. A Coyote in Sheep's Clothing: Predator Identification from Saliva. Wildlife Society Bulletin 31:926-932. Williamson-Natesan, E. G. 2005. Comparison of methods for detecting bottlenecks from microsatellite loci. Conservation Genetics 6:551-562. Wilson, G. A. and B. Rannala. 2003. Bayesian inference of recent migration rates using multilocus genotypes. Genetics 163:1177. Worthington Wilmer, J., C. Elkin, C. Wilcox, L. Murray, D. Niejalke, and H. Possingham. 2008. The influence of multiple dispersal mechanisms and landscape structure on population clustering and connectivity in fragmented artesian spring snail populations. Molecular ecology 17:3733-3751. Yackulic, C. B., J. Reid, R. Davis, J. E. Hines, J. D. Nichols, and E. Forsman. 2012. Neighborhood and habitat effects on vital rates: expansion of the Barred Owl in the Oregon Coast Ranges. Ecology 93:1953-1966.

155 156 APPENDIX A Natural history note: Leucism

NATURAL HISTORY NOTES 297

!"#$%"&'%&(")$%*"+'!"#$%&'()*"#+,,*-#$)./!"#$%&'()%!*)19 6*(#6(4*016.*$#1-%,%;,>*=$%-=+1E,J:,51B%?>4*%56,5%4BB:>-%7,5:>,;;D% ,B>4--% -4:56E=,-5=>7% F:-5>,;1,$% 17B;:817*% A,>5-% 4?% [=<% !4:56% e,;=-$%"1B54>1,$%C,-+,71,$%,78%!4:56%F:-5>,;1,%/]DG=%&''&(%F:-5>(% a44;(%\&T\&WVb.(%H75>48:B=8%A4A:;,5147-%,;-4%4BB:>%17%[=<%a=,E ;,78% /"4>4-% =5% ,;(% &''b% F:-5>,;(% YB4;(% \\TQ&\WQ&P.(% !*)16*(# 6(4*9 016.*$%1-%;1-5=8%,-%=78,7*=>=8%@D%56=%HfZ[(%N7%&V%`,7:,>D%&'2&$% <=% B,:*65% ,% ;=:B1-51B% !"# 6(4*016.*$# 17% ,% <=5;,78% 17% 3:7844>,$% O=;@4:>7=$%"1B54>1,$%F:-5>,;1,%/\U(Ug!$%2V)('\\gY.(%H5%<,-%?4:78% 17%*448%A6D-1B,;%B47815147$%?;4,517*%,+47*%=+=>*=75%9=*=5,5147% 7=,>%56=%<,5=>^-%=8*=(%C<4%,8:;5%?=+,;=%!"#6(4*016.*$$%4?%5DA1B,;% *>==7%,78%@>4<7%B4;4>,5147%<=>=%,;-4%B,:*65%17%56=%-,+=%<,E 5=>% @48D(% C6=% ;=:B1-51B% 1781918:,;% <,-% ,7% ,8:;5% +,;=% /QV% ++% !"#.%17%@>==817*%B47815147%/01*(%2.(%H5%<,-%=751>=;D%D=;;4<%<156%,% !"#$%2(%e448%0>4*%/!*):12();$#$%&'()*+,$.%17%,+A;=I:-%<156%,%*>,918% A17G%517*=$%=IB=A5%?4>%8,>G;D%A1*+=75=8%=D=-$%,78%>,1-=8%@>4<7% ?=+,;=%!A455=8%!,;,+,78=>%/-.2%$)1.(#.(+,&(),..$%3,5%#,G=$%F;E 7:A51,;% A,8-% 47% 15-% 56:+@-(% C6=% 84>-,;% -G17% <,-% 5>,7-;:B=75(% *47J:17%]>4917B1,;%],>G$%N75,>14$%Z,7,8,(% e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e1;8;(% c=-(% \\T))UW)QV.(%C4% <=;;%B6,>,B5=>1K=8%L,91,7%@=6,914>-%<61B6%17B;:8=%+1-81>=B5=8% 4:>%G74<;=8*=$%561-%1-%56=%?1>-5%84B:+=75=8%B,-=%4?%;=:B1-+%17%!"# +,517*%<156%B47-A=B1?1B-$%7=B>4*,+D$%4>%,+A;=I:-%<156%17,71E 6(4*016.*$"#F%-=B5147%4?%54=E<=@%<,-%8=A4-15=8%17%56=%51--:=%B4;E +,5=% 4@M=B5-(% N@-=>9,5147-% 4?% 175=>-A=B1?1B% ,+A;=I:-% @=5<==7% ;=B5147%4?%O:-=:+%"1B54>1,$%F:-5>,;1,%/[O"a%2UQ'&.( ,7:>,7-%,>=%>=B4>8=8%<156%-4+=%>=*:;,>15D%/O=-6,G,%2PPQ(%0;4>1E e=%56,7G%d=4??>=D%e(%_=,>8%?4>%B4++=75-%47%561-%745=(%N:>% 8,%!B1=75(%)PR&STUVWU)X%Y,547%=5%,;(%2PPP(%Z,7(%01=;8%[,5(%22\T)2&W <4>G% <,-% ?:78=8% @D% ,7% F:-5>,;1,7% c=-=,>B6% Z4:7B1;% #17G,*=% )2\X%]=,>;%=5%,;(%&'')(%F+(%O18(%[,5(%2)VT2&QW2\VX%L^F+4>=%=5%,;(% d>,75%/#]'PP'2Q2.%<156%56=%F:-5>,;1,7%c=-=,>B6%Z=75>=%?4>%f>E &''P(%_=>A=54;(%Z47-(%314;(%VT\&)W\\'.(%3D%B475>,-5$%>=A4>5-%4?%17E @,7%YB4;4*D$%d>4<;17*%d>,--%0>4*%C>:-5%0:78$%O=;@4:>7=%O:-=E 5=>-A=B1?1B%A,1>17*-%4?%-,;,+,78=>-%,>=%>,>=%,78%>=-5>1B5=8%54%-A=E :+$%O=;@4:>7=%e,5=>$%],>G-%"1B54>1,$%,78%"1B54>1,7%L=A,>5+=75% B1?1B%-A=B1=-%4>%<=;;EB6,>,B5=>1K=8%6D@>181K,5147%=9=75-%/"=>>=;;% 4?%!:-5,17,@1;15D%,78%Y791>47+=75( 2PP'(%`(%a44;(%#478(%&&2TVV2WV)2X%"=>>=;;%2PPV(%Y94;:5147%VbTP&2W P&).(%F7%=I6,:-519=%-=,>B6%4?%56=%?4>+,;%;15=>,5:>=%>=9=,;=8%?=<% 4@-=>9,5147-%4?%,+A;=I:-%@=5<==7%,7%,7:>,7%,78%B,:8,5=$%17E B;:817*%/,01#2,01%,78%3(&(.(456(#$(&(.(456(%/O,>B4%,78%#1E K,7,%&''&(%Y564;(%YB4;(%Y94;(%2VT2Wb.$%,;564:*6%561-%A6=74+=747% 1-%;1G=;D%:78=>>=A4>5=8( H5%6,-%@==7%A>4A4-=8%56,5%L,91,7%@=6,914>-%+,D%>=-:;5%?>4+% 56=%61*6;D%B4+A=51519=%7,5:>=%4?%,7:>,7%>=A>48:B5147%B6,>,B5=>E 1K=8%@D%-=,-47,;%/51+=%-=7-1519=.%,78%=IA;4-19=%>=A>48:B519=%,*E *>=*,5147-%/O=-6,G,%2PPQ$%17"#+*)"X%c=,817*%2PbV(%`(%a44;(%#478(% &'\TP)W2'2.(% d19=7% 56=% -6,>=8% @>==817*% 6,@15,5% ,78% -D7B6>4E 71K=8%@>==817*%A=>148-%4?%!"#$%&'()*+,$%,78%-"#.(+,&(),.8%+,5=% +1-18=751?1B,5147%-==+-%,%;1G=;D%B,:-=(%e61;=%B;4-=%B47?17=+=75% +,D%6,9=%?,B1;15,5=8%56=%L,91,7%@=6,914>%>=A4>5=8%6=>=$%A>=91E 4:-%4@-=>9,5147-%4?%175=>-A=B1?1B%,+A;=I:-%17%456=>%-A=B1=-%,?E ?1>+%56,5%561-%@=6,914>%1-%7,5:>,;;D%4BB:>>17*(%C6=%?,B5%56,5%56=% ?>4*%=I61@15=8%,%A=B54>,;%*1>8;=%B;,-A%?:>56=>%-:**=-5-%56,5%15%<,-%

,55=+A517*%54%=7*,*=%17%74>+,;%,+A;=I:-%>,56=>%56,7%,7%,BB1E !"#$%2(%F8:;5%+,;=%;=:B1-51B%!*)16*(#6(4*016.*$#?>4+%O=;@4:>7=$%"1BE 8=75,;%4>%5=+A4>,>D%=??4>5( 54>1,$%F:-5>,;1,( 017,7B1,;%-:AA4>5%A>4918=8%@D%56=%0>1=78-%4?%F;*47J:17%],>G(% O:B6%56,7G-%?4>%17EG178%-:AA4>5%4<17*%54%56=%F;*47J:17%e1;8;1?=% c=-=,>B6%!5,5147$%c4>D%YBG=7-<1;;=>$%,78%3>,8%!5=17@=>*( CLAIRE C. KEELY, School of Botany, University of Melbourne, Victoria PATRICK D MOLDOWAN (e-mail: [email protected]), Univ. 3010, Australia (e-mail: [email protected]); SUSANA P. MAL! Guelph, Guelph, Ontario, Canada N1G 2W1, DAVID L LEGROS (e-mail: DONADO, Department of Sciences, Museum Victoria, Melbourne, Victoria [email protected]), Laurentian University, Sudbury, Ontario, Canada, 3001, Australia (e-mail: [email protected]). P3E 2C6, and GLENN J TATTERSALL (e-mail: [email protected]), St. Catharines, Ontario, Canada, L2S 3A1.

Herpetological Review 44(2), 2013

157

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Keely, Claire Catherine

Title: Conservation genetics of the Growling Grass Frog, Litoria raniformis, in urbanising landscapes

Date: 2016

Persistent Link: http://hdl.handle.net/11343/127252

File Description: Conservation genetics of the Growling Grass Frog, Litoria raniformis, in urbanising landscapes

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