Monitoring the Effects of Climate Change on the Rainforest Birds of Eastern Australia

Monitoring the effects of climate change on the rainforest birds of eastern Australia

Author Leach, Elliot

Published 2017-09

Thesis Type Thesis (PhD Doctorate)

School School of Environment and Sc

DOI https://doi.org/10.25904/1912/1224

Downloaded from http://hdl.handle.net/10072/379569

Griffith Research Online https://research-repository.griffith.edu.au Monitoring the effects of climate change on the rainforest birds of eastern Australia

Elliot Leach BSc (Hons) Environmental Futures Research Institute School of Environment Griffith University

Submitted in fulfilment of the requirements of the degree of Doctor of Philosophy

September 2017

Synopsis

Climate change will significantly affect avian biodiversity on a global scale. Increasing temperatures over the next century will lead to shifts in species distributions, alterations in the timing of breeding and migration, changes in morphology and shifts in genetic frequencies among avian populations. The global hotspots of avian diversity are found in mountainous rainforests, regions which may be difficult to access. Therefore, effective ways of monitoring rainforest bird assemblages are vital, for both ecologists and conservationists.

This thesis addresses the challenge of monitoring the effects of climate change on rainforest bird assemblages. I used two methods, point counts and automated acoustic recording, to sample the rainforest birds occupying three elevational gradients in rainforests on the east coast of Australia. In doing so, I had the following aims: 1) to determine whether biodiversity data from automated acoustic recordings made using automated recording units (ARUs) was comparable to data generated using a traditional method (point counts), 2) to assess the ability of ARUs to monitor cryptic rainforest species for long time periods, 3) to identify birds that could be used as indicator species of elevation for the purpose of long-term climate change monitoring, and 4) to investigate the driving factors of bird species richness and abundance along elevational gradients in Australian rainforests.

Existing studies showed contrasting results when comparing the effectiveness of traditional avian sampling methodologies with ARUs. To address this in an Australian rainforest context, we collected data on the birds of Eungella National Park in central Queensland over two sampling periods. We found that data from point counts and ARUs was broadly similar. On average, point counts detected more species than recordings of the same duration. The respective strengths and weaknesses of point counts and ARUs are complementary, and they should be used simultaneously in future biodiversity surveys.

ARUs can sample remotely, simultaneously, and for long time periods. Using ARUs, we collected a year’s worth of data on two cryptic species inhabiting rainforest in north-eastern New South Wales. Bassian Thrush Zoothera lunulata and Russet-tailed Thrush Z. heinei are secretive inhabitants of wet forests on the eastern coast of Australia. We found that the two species had differential elevational preferences: Bassian Thrush preferred elevations above 900m asl, and Russet-tailed Thrush preferred elevations below 700m asl. Recordings of song indicated that Russet-tailed Thrush bred earlier than Bassian Thrush in 2015. This, along with the elevational preferences of the two species, may be related to temperature. The use of

ARUs enabled us to quantify the elevational preferences and likely breeding times of these cryptic species. Populations of Bassian Thrush in north-eastern New South Wales and southeastern Queensland are likely to decline with increasing temperatures.

Upwards shifts in the elevational ranges of rainforest birds are expected due to increasing global temperatures. Identifying the current elevational distributions of indicator species has been suggested as one way of monitoring such upwards shifts. Previous research in our study region had identified indicator species among various invertebrate and plant taxa, but information on vertebrate indicators was lacking. Using data on the elevational preferences of birds collected over one year, we identified avian indicators of lowland and highland rainforest sites in north-eastern New South Wales. These indicators may be used to detect future shifts in species elevational preferences in the region.

Previous research in tropical rainforests of the Wet Tropics identified temperature as an important driver of bird species’ distributions. There was a comparative lack of information for the subtropical rainforests of north-eastern New South Wales. Our data from elevational gradients in this region indicated that temperature was significantly positively correlated with both avian species richness and abundance. Species richness declined with elevation; there was no consistent elevational pattern in abundance. We found that species’ functional traits mediated their responses to the changes in environmental conditions along the gradient: large-bodied and small-bodied species are likely to be affected in different ways by increasing temperatures.

My research has determined effective ways of monitoring the effects of climate change on rainforest bird assemblages. In doing so, I have also addressed major gaps in the knowledge of two relatively understudied biodiversity hotspots on the eastern coast of Australia. The baseline data presented in this thesis allows future researchers to detect changes in the avian biodiversity of the study regions, and represents a significant contribution to ornithology and climate change research in Australia and internationally.

iii

Statement of originality

This work has not previously been submitted for a degree or diploma in any university. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made in the thesis itself.

______

Elliot Leach

Table of Contents

Synopsis ...... ii

Statement of originality ...... iv

Table of Contents ...... v

List of figures ...... vii

List of tables...... ix

Journal articles arising from this thesis ...... xi

Acknowledgements ...... xii

Chapter 1 ...... 14

1.1 Introduction ...... 14 1.2 Thesis outline ...... 19

Chapter 2 ...... 21

2.1 Abstract...... 22 2.2 Introduction ...... 22 2.3 Methods ...... 24 2.4 Results ...... 26 2.5 Discussion ...... 27 2.6 Acknowledgements ...... 29 2.7 Appendix 2 ...... 30

Chapter 3 ...... 31

3.1 Abstract...... 32 3.2 Introduction ...... 32 3.3 Methods ...... 33 3.4 Results ...... 34 3.5 Discussion ...... 36 3.6 Conclusions ...... 38 3.7 Acknowledgements ...... 38 3.8 Appendix 3 ...... 39

Chapter 4 ...... 40

4.1 Abstract ...... 41 4.2 Introduction ...... 42 4.3 Methods ...... 43 4.4 Results ...... 47 4.5 Discussion ...... 51 4.6 Acknowledgements ...... 55 4.7 Appendices ...... 57

Chapter 5 ...... 72

5.1 Abstract ...... 73 5.2 Introduction ...... 74 5.3 Methods ...... 76 5.4 Results ...... 81 5.5 Discussion ...... 86 5.6 Acknowledgements ...... 90 5.7 Appendices ...... 91

Chapter 6 ...... 118

6.1 Introduction ...... 118 6.2 Monitoring the effects of climate change on rainforest birds ...... 119 6.3 Potential effects of climate change on Australian rainforest birds ...... 122 6.4 The drivers of avian species richness and abundance in Australian rainforests ...... 123 6.5 Conclusions ...... 125

References ...... 126

List of figures

Figure 1.1 Conceptual model of the impacts of climatic change on the birds of Australia’s eastern rainforests. Arrows indicate a direct effect; red question marks indicate gaps in current knowledge...... 15

Figure 2.1 Non-metric multidimensional scaling ordinations (left) and graphs of Sørensen similarity between all possible pairs of sites plotted against the difference in elevation between site pairs for (a) point counts (b) recordings and (c) both methods combined. In all ordinations, inverted triangles 200m sites; open squares 400m sites; closed squares 600m sites; diamonds 800m sites, circles 1000m sites and triangles 1200m sites...... 27

Figure 3.1 Sonograms of recordings made in June 2015–May 2016 from the study area in the Border Ranges and Mebbin National Parks, New South Wales, containing thrush Zoothera spp. calls and songs. Arrows indicate the midpoints of the calls and songs described as follows: (a) two contact calls of a thrush Zoothera sp. from 0 to 1 second and again from c. 5.5 to 6 seconds, in the 6.5–10-kHz frequency band (contact calls of the species discussed here are virtually identical); (b) three examples of Bassian Thrush (BT) song, from 0 to 1 second, from 2.75 to 3.5 seconds and from c. 5.5 to 6.5 seconds (the latter being an example of the pilop-whee-er call) in the 2–3-kHz frequency band; (c) two examples of Russet-tailed Thrush (RT) song, from 1 to 2.25 seconds and from c. 5 to 6 seconds in the 3–4-kHz frequency band...... 35

Figure 3.2 The number of 2-minute recordings per month in which Bassian Thrush (BT) and Russet-tailed Thrush (RT) songs were detected across the entire elevational gradient, 10 June 2015–31 May 2016, in the Border Ranges and Mebbin National Parks, New South Wales... 36

Figure 4.1 Bird species richness in the Gondwanan Rainforests. The total number of species detected along the entire elevational gradient (300m – 1100m asl) in each season by ARUs and point counts is shown in A. The total number of species detected in each elevational band using ARUs and point counts is shown in B...... 48

vii

Figure 4.2 Ordination panes based on non-metric multidimensional scaling (NMDS) displaying turnover in community composition between low-elevation (300-700m asl) and high-elevation (900-1100m asl) sites, based on the ‘total’ ARU (A) and ‘total’ point counts (B) datasets. Low elevation sites are orange and high elevation sites are blue. Site labels are interpreted as follows. The numbers represent the elevation: “3” represents 300m asl. The next two letters identify the gradient: “SC” represents Sheepstation Creek and “BM” represents Bar Mountain. The final letter distinguishes between replicate sites at that elevation and gradient. A map of the study area, including all sites, is available in Leach et al. (2018a)...... 49

Figure 5.1 Map of the study area near the Queensland/New South Wales border on the east coast of Australia. All sites shown lie within Border Ranges National Park (28°230S, 153°30E) with the exception of the 300 m sites on the Bar Mountain gradient, which lie within Mebbin National Park (28°270S, 153°100E). Dark areas of the map indicate low elevations; paler areas indicate high elevations...... 77

Figure 5.2 Relative proportions of each avian feeding guild in each elevational band overall (a), in summer (b) and in winter (c). Summer and winter were defined as December to February and June to August, respectively...... 83

Figure 5.3 Overall bird species richness (a) and median temperature (b) in each elevational band...... 84

Figure 5.4 Predicted abundances for bird species with small body size (a) and average body size (b) with increasing temperature. Temperature was measured during the 3 h following dawn; the temperatures along the x-axis represent the temperature of the early morning. Body sizes are representative: the small bird (Golden Whistler Pachycephala pectoralis) represents the average body size of all species in the dataset with weight <100 g, and the average bird (Noisy Pitta Pitta versicolor) represents the average body size of all species in the dataset. The shaded area represents the 95% confidence interval...... 86

viii

List of tables

Table 2.1 Relative performance of automated acoustic recorders (ARs) versus traditional methods in previous studies ...... 23

Table 3.1 Two-minute sound recordings taken at different elevations in the Border Ranges and Mebbin National Parks, New South Wales: total number of recordings processed and numbers in which Bassian Thrushes (BT) and Russet-tailed Thrushes (RT) were detected ... 36

Table 4.1 Numbers of indicators found in each seasonal dataset using ARUs and point counts. Indicators are individual species or pairs of species...... 50

Table 4.2 Recommended avian indicator sets for the long-term monitoring of lowland and highland sites in the Gondwanan Rainforests ...... 53

Table 5.1 Observed sample coverage (SC), species richness and total abundance at each site ...... 82

Table 5.2 Results of the GLMMs for overall species richness and abundance. The change (Δ) in Akaike’s information criterion (AIC) and the Bayesian information criterion (BIC) between the null model and the final model is also shown ...... 85

Table 5.3 Results of the GLMM testing the importance of environmental variables, trait information and trait-environment interactions on species’ abundances. Canopy refers to canopy cover in %. Significant parameter estimates (p <0.05) are in italics...... 85

Table S5.1b Seasonal changes in the trophic structure of each elevation between summer and winter. At each elevation, the proportion of individuals belonging to each feeding guild was compared between the two seasons using chi-squared tests; significant differences are indicated by italics...... 92

Journal articles arising from this thesis

Chapters 2, 3, 4 and 5 in this thesis are manuscripts that were prepared for scientific publication. Each chapter has now been published in a peer-reviewed journal. The manuscripts for Chapters 2, 4 and 5 were co-authored with other researchers. Appropriate acknowledgements of those who contributed to the research but did not qualify as authors are made in each manuscript or in the Acknowledgements section of this thesis. My contribution to the co-authored manuscripts is outlined at the start of each relevant chapter. The bibliographic details for these manuscripts are:

Chapter 2: Leach, E. C., Burwell, C. J., Ashton, L. A., Jones, D. N. & Kitching, R. L. 2016. Comparison of point counts and automated acoustic monitoring: detecting birds in a rainforest biodiversity survey. Emu, 116, 305-309.

Chapter 3: Leach, E. C. 2016. Notes on the Zoothera thrushes in the Tweed Range of north-eastern New South Wales. Australian Field Ornithology, 33, 240-243.

Chapter 4: Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018. Identifying avian indicators of elevation in the Gondwanan Rainforests of Australia. Pacific Conservation Biology. https://doi.org/10.1071/PC18039

Chapter 5: Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018. Modelling the responses of Australian subtropical rainforest birds to changes in environmental conditions along elevational gradients. Austral Ecology. https://doi.org/10.1111/aec.12586

(Signed) ______(10/7/18) (Countersigned) ______(10/7/18)

Elliot Leach Supervisor: Darryl Jones

(Countersigned) ______(10/7/18) (Countersigned) ______(10/7/18)

Associate Supervisor: Chris Burwell Supervisor: Roger Kitching

Acknowledgements

Firstly, I extend my thanks to my supervisors. Professor Darryl Jones fostered my interest in nature from a very young age, through numerous expeditions close to home and far away. Professor Roger Kitching inspired my interest in ecology as a career when I met him for the first time as part of the IBISCA project in Lamington National Park. Doctor Chris Burwell encouraged what quickly became an obsession with Australian birds on my first trip to Mt. Lewis. Each one of my supervisors has provided me with valuable support, advice and encouragement throughout my PhD. I am truly thankful for their help.

I am also grateful to the Commonwealth Government of Australia for providing an Australian Postgraduate Award (APA) which funded my PhD program. The Griffith School of Environment, the Environmental Futures Research Institute, and the Griffith Graduate Research School all provided additional financial, administrative and academic development support. Birds Queensland provided me with a generous research grant in 2015 to cover the cost of purchasing my BARs. Mark Calder and Michael Maggs from Frontier Labs were extremely generous with their time, expertise and troubleshooting nous: I wish them the best of luck as they continue to develop their Bioacoustic Recorders.

I am grateful to the traditional custodians of the land on which this research was conducted: the Birri Gubba people of Eungella National Park, in Queensland, and the Githabul people of the Border Ranges region in New South Wales. I pay respects to their elders past, present and emerging. I thank the Queensland and New South Wales governments for providing permits to work in Eungella, Border Ranges, and Mebbin National Parks. National Parks staff in both states were very helpful, and should be commended for their work. I also thank Jim O’Brien for rescuing a BAR, showing me around Mebbin, and inviting me over for tea during fieldwork.

I thank the volunteers and scientists involved in the Eungella Biodiversity Project, in particular numerous members of BirdLife Mackay, Gary Hendicott and Jenny Wooding, Kate Brunner, Maya Harrison, Claire Gely, Erica Odell, and Darryl Barnes. I thank Dr Louise Ashton and Dr Sarah Maunsell, as well as their associates, for the establishment of the Border Ranges elevational gradients which I used in my research.

xii

I’ve been lucky enough to work with a number of amazing researchers over the last five years. All have been, and continue to be, motivational figures that have provided me with support, encouragement, and laughter. Thanks go to Dr Akihiro Nakamura, who showed me that I could be a scientist; Dr Bill McDonald, whose knowledge and personality are national treasures; Dr James McBroom, whose statistical knowledge is incredible; and Rob Appleby, who introduced me to scepticism. I would like to extend special thanks to two inspirational women: Dr Louise Ashton and Dr Sarah Maunsell. They taught me that hard work pays off, and I’m lucky to have them as two of my closest friends.

At university, various members of my PhD cohort and lab group have helped me in too many ways to quantify over the years. Thanks to Kyle Barton, Amanda Dawson, Dr Francesca Dem, Dr Casey Hall, Jane Hardwick, Ma Rakeb-ul Islam, Lois Kineen, Erica Odell, Rachakonda Sreekar, Marisa Stone, Tamara Taylor, and Emma Window. Away from university, I have a great group of friends who also deserve recognition: Barkley, Heath Cambie, Matt Cochrane, Gus Daly, Nikolas Haass and Raja Stephenson, Brendan Hogg, Chris Kotmel, Kate Mackie, Michelle and Andrew Mackie (plus Harvey), Mel McGregor, Elie Moubarak, Tom O’Halloran, Stu Pickering, Abigail Pirie, Stuart Richardson, Taya Seidler, Zen Spokes, Fran Thomas, Els Wakefield and, finally, Kit and Diana Walker.

I am forever grateful to my parents, David and Lori Leach, and my sisters, Laetitia and Phoebe, for their love and encouragement. I thank both my extant (James) and extinct (Judith, Edward and Betty) grandparents for being so proud of me. Finally, I thank my partner, Jess Mackie, for her positivity, her love, and for being my best friend.

xiii

Chapter 1

Introduction and Thesis Outline 1.1 Introduction

Ongoing climate change represents arguably the most severe threat to biodiversity on a global scale (Dirzo et al., 2014). Climate change is altering the structure and function of many ecosystems and has been linked to population declines and range shifts in a variety of taxa, including birds (Both et al., 2006; Hurrell and Trenberth, 2010; Chen et al., 2011; Freeman and Class Freeman, 2014). Serious impacts of rising temperature are likely to occur on mountain ranges, where evolutionary history has led to increased sensitivity to temperature among animals and plants (Williams et al., 2003; Şekercioğlu et al., 2008). Unfortunately, due to the same evolutionary history, many global hotspots of avian biodiversity exist in mountainous regions near the equator (Orme et al., 2005; Ruggiero and Hawkins, 2008) and these are seriously threatened by increasing temperatures (Laurance et al., 2011a). In Australia, the forested slopes of the Great Dividing Range support some of the richest avian communities in the country, which are similarly vulnerable to climatic change (Laurance et al., 2011b).

Global climate change will significantly affect the rainforest bird communities of Australia. Exactly what impact(s) the combination of rising temperatures and declining rainfall forecast by the CSIRO and the Bureau of Meteorology (2016) will have on these communities is currently unknown, but some predictions can be made (Fig. 1.1). There is strong evidence that rising temperatures are driving changes in species’ distributions (Root et al., 2003; Parmesan, 2006). Such changes are the result of species attempting to track the ongoing spatial and temporal changes in their preferred microclimatic niches (Hannah et al., 2014). These adaptive movements are typified by species moving upwards in elevation, or towards the poles (Chambers et al., 2005; Parmesan, 2006; Kirchman and van Keuren, 2017). In Australia, there has been researchers have been interested in the impacts of climate change on avifaunal distributions since at least the 1980’s, when Arnold (1988) examined potential changes in faunal distributions in the south-west of Western Australia. Soon after, Bennett et al. (1991) used BIOCLIM to investigate the impacts of climate change on 17 Victorian bird species. Since that time, there have been many Australian studies attempting to forecast shifts in species’ distributions as a result of climate change (partially reviewed in Chambers et al.

(2005)). Of particular interest in the context of this thesis are the studies carried out by Professor Stephen Williams’ research group and others in the Wet Tropics bioregion. These studies have demonstrated that Australian rainforest birds are likely to experience contraction of their distributions as the climate continues to warm (Hilbert et al., 2004; Williams et al., 2003; Shoo et al., 2005; Williams and Middleton, 2008; Williams et al., 2010; Anderson et al., 2013; Scheffers et al., 2017). This will likely be the most obvious and impactful consequence of climatic change for rainforest birds in Australia.

Changes in phenology, in particular the timing and success of breeding events and migrations, are also expected in a warming world. Long-term studies, mainly conducted in Europe, indicate that rising temperatures are associated with advances in the laying dates of many bird species (Dunn and Winkler, 2010). Revealing these trends requires high-quality, long-term datasets such as those of Both et al. (2006) and Charmantier et al. (2008). Datasets such as these are less common in Australia than they are in Europe – there are still large gaps in the scientific knowledge of many Australian bird species (Garnett et al. 2010). Many Australian birds (approximately 40% of land-based residents) undertake seasonal or nomadic migrations (Chan, 2001). As the effects of climate change are not uniform across regions (Garcia et al., 2014), it is possible that species may respond to local migratory cues that have become decoupled from conditions at their destination. Such scenarios can lead to dramatic population declines in some cohorts (Both, 2010).

The changes in distribution and phenology described above are likely to lead to declines in the populations of Australian rainforest birds. Changes in habitat structure or quality, as well as food availability, are the most obvious causative factors. However, the rise of novel communities and inter-specific interactions (Hobbs et al., 2009), as well as failures to adapt physiologically to increasing temperature or reduced water availability are also likely to lead to declines in certain species (Chown et al., 2010). Evolutionary adaptation, which is expected to occur but may be difficult to detect (Hoffman and Sgrò, 2011), may work to counteract these declines, and lead to morphological or physiological adaptation. The utilisation of museum specimens as comparative material may prove valuable when attempts are made to detect changes in genetic frequencies or morphology as a result of climatic change (Gardner et al., 2014; Holmes et al., 2016).

Clearly, there are significant gaps in our knowledge of how Australian rainforest birds are responding to climatic changes. Key to closing these gaps is an improved understanding of

16 the best way in which to monitor avian communities (Krause and Farina, 2016). In addition, understanding how Australian rainforest birds respond to changes in abiotic conditions (such as temperature) is vital for future conservation efforts.

Monitoring avian communities in rainforests

Birds are counted, surveyed and monitored by ornithologists using a bewildering array of methods, for a bewildering number of reasons (Bibby et al., 1992). On land, the most common survey techniques involve an ornithologist searching an area, counting birds from a fixed location, or walking transects. A common extension of the latter methods that allows for density and population estimation is known as Distance sampling (Buckland et al., 2005). Ornithologists also utilise a variety of trapping, banding and tracking techniques to garner information on and monitor avian communities (Bub, 2012). New methods are frequently developed; technological advances have recently led to the use of unmanned aerial vehicles in bird surveys (e.g. Ratcliffe et al., 2015) and satellite imaging has been used to estimate penguin populations in Antarctica (Fretwell et al., 2012).

However, surveying birds in rainforests can be difficult, as visibility is often limited and diversity is high (Anderson, 2011). Many areas of interest are also challenging and / or expensive to access (especially on a regular basis) due to rugged terrain (Thomas and Marques, 2012). Inclement weather, particularly in the tropics, can also impede survey efforts. Traditional methods of surveying birds in rainforests include point counts, line transects and mist netting (Loiselle and Blake, 1991; Haselmayer and Quinn, 2000; Anderson et al., 2013); each of these methods depends on the presence of at least one skilled ornithologist. Using these methods to monitor rainforest avian biodiversity over large spatial distances and over long periods of time is, therefore, often impractical.

Recently there has been an explosion of research interest surrounding the use of automated acoustic monitoring devices for monitoring biodiversity (Depraetere et al., 2012; Aide et al., 2013; Gasc et al., 2013; Gage and Axel, 2014; Sueur et al., 2014; Towsey et al., 2014; Klingbeil and Willig, 2015; Krause and Farina, 2016; Shonfield and Bayne, 2017). There are many advantages to monitoring biodiversity using autonomous recording units (ARUs). ARUs can be programmed to record at specific times of the day, they can remain in the field for extended periods of time (and potentially transmit data to researchers remotely), they are becoming more affordable, and they create a permanent record of the species generating sounds at the time the recording is made (Sedláček et al., 2015; Krause and Farina, 2016).

ARUs, therefore, seem ideal for use by rainforest ecologists who study vocal taxa and are interested in monitoring the impacts of climate change. To date, most research using ARUs has been performed in temperate habitats. Testing the effectiveness of ARUs in tropical and subtropical rainforests is therefore a priority.

Elevational gradients and climate change

Early ecologists (Willdenow, von Humboldt, Darwin and Wallace) recognised in the 1800’s that the diversity of animals and plants changed along environmental gradients, both latitudinal and elevational (Lomolino, 2001). The predictable nature of the changes in environmental conditions along such gradients (i.e. the decline in temperature with increasing latitude or elevation) has led to their widespread use in the fields of biogeography and ecology (Lomolino, 2001; Becker et al., 2007). In particular, the use of elevational gradients in climate change research is now well-established (Malhi et al., 2010; McCain and Colwell, 2011). This is due to the fact that rapid changes in abiotic conditions, caused by increasing elevation, can be found over relatively small distances (Körner, 2007). These changes allow researchers to make ‘space-for-time’ substitutions when investigating how ecological communities respond to changes in climate, or other factors such as wind strength or soil composition (Strong et al., 2011). Continued climatic warming is predicted to result in species shifting their distributions in order to track their preferred climatic conditions (Root et al., 2003; Parmesan, 2006). Along elevational gradients, this will lead to species moving upwards in elevation (Freeman and Class Freeman, 2014). Baseline surveys, to establish the current elevational ranges of species along elevational gradients, are essential to ensure that informed conservation decisions can be made in the future (Williams et al., 2010). For several taxa, including birds, mammals, reptiles and invertebrates, such baseline surveys have already been carried out in the Wet Tropics bioregion (Williams et al., 1996; Yeates and Monteith, 2008). The Gondwana Rainforests of Australia World Heritage Area (hereafter Gondwanan rainforests) protects some 366,131 ha of subtropical rainforest (Kitching et al., 2010b). Baseline surveys of elevational distribution in the Gondwanan rainforests have previously been restricted to invertebrate groups (Kitching et al., 2011) and plants (Laidlaw et al., 2011a). Baseline surveys to determine the elevational preferences of the birds in the Gondwanan rainforests are necessary to inform future research and conservation efforts in the region.

Avian responses to elevation

Changes in elevation can have dramatic effects on the distributions, abundance and species richness of birds (McCain, 2009; Acharya et al., 2011). However, elevation itself rarely has direct effects (Elith and Leathwick, 2009). Rather, it is the changes in more functionally relevant predictors, such as temperature or solar radiation with elevation that affect species (Körner, 2007; Elith and Leathwick, 2009). The decline in temperature with increasing elevation is a universal pattern, and is the most obvious change in abiotic conditions that occurs along elevational gradients (Körner, 2007). It was long assumed that this trend would drive a similar decline in species richness. However, Rahbek (1995) demonstrated through a careful review of the available data that this decline was not necessarily monotonic, and that in fact many elevational gradients had a mid-elevation peak in species richness. A global meta-analysis of bird diversity along elevational gradients inspired by this seminal paper showed four general patterns: (1) decreasing diversity with elevation, (2) low-elevation plateaus in diversity, (3) low-elevation plateaus with mid-peaks, and (4) uni-modal mid- elevational peaks (McCain, 2009). Current climate, particularly the combination of temperature and water availability, was identified as the main driver of bird elevational diversity. Other putative explanations for diversity trends, such as the species-area effect and the mid-domain effect, have little empirical support (McCain, 2009). On humid mountains, diversity patterns (1) and (2) predominate; diversity on dry mountains tends to follow patterns (3) and (4). In Australia, similar research has so far been limited to the Wet Tropics bioregion in northern Queensland (Williams et al., 2010). Studies in other rainforest ‘hotspots’ of bird diversity, such as the Gondwanan rainforests, are necessary to gain a broader understanding of the responses of Australian avifauna to elevation.

1.2 Thesis outline

This thesis addresses the challenge of understanding how best to monitor rainforest bird communities, as well as determining drivers of avian species richness and abundance in subtropical rainforest along the Great Dividing Range. I used a combination of point counts and ARUs to collect data on the avifaunal communities along three elevational gradients in eastern Australian rainforests. In conducting this research, I had the following aims:

1. To generate baseline data on the elevational distributions of birds in Eungella National Park and the northern regions of the Gondwanan rainforests of Australia,

2. To determine the utility of using ARUs in biodiversity surveys and as long-term monitoring tools in Australian rainforests, 3. To identify avian indicators of elevation suitable for use in long term climate change monitoring schemes, and 4. To determine how changes in abiotic conditions along elevational gradients influence the diversity and abundance of Australian rainforest birds, and whether species’ traits play a role in mediating their responses to such changes.

Apart from Chapters 1 and 6, this thesis is organised as a series of published manuscripts. Chapters 2 – 5 are presented in the form of manuscripts that have been prepared for submission to peer-reviewed journals in accordance with Griffith University policy on PhD theses as published and unpublished papers (Griffith University, 2017).

Chapter 2 is a methodological study which compares the effectiveness of a traditional survey technique (point counts) with a relatively new survey method (automated acoustic monitoring) in terms of detecting birds in Eungella National Park, in central Queensland. Fieldwork for Chapters 3 to 5 was conducted in the northern region of the Gondwana Rainforests of Australia World Heritage Area, in the Border Ranges and Mebbin National Parks of north-eastern New South Wales. Chapter 3 is a case study demonstrating the usefulness of ARUs for the detection and monitoring of two cryptic species (thrushes in the genus Zoothera) in subtropical rainforest. Chapter 4 focuses on identifying avian indicators of elevation suitable for long-term monitoring of the effects of climate change on the avifauna of subtropical rainforests in the region. Chapter 5 uses generalised linear mixed models to (1) identify the abiotic factors responsible for determining species richness and abundance of a rainforest bird assemblage, and (2) quantify the trait-environment relationships of this assemblage in subtropical rainforest. Chapter 6 concludes the thesis by providing a general discussion of the findings, as well as suggesting future research directions. Relevant supplementary material is provided following each chapter. To avoid duplication, a combined reference list is provided at the end of the thesis.

Chapter 2

In this chapter, we compare the efficacy of point counts and automated acoustic recording for detecting birds as part of a biodiversity survey in Eungella National Park, Queensland.

Statement of contribution to a co-authored paper

Chapter 2 is a co-authored paper that has been published in the journal Emu. Copyright permission to reproduce this article has been granted by the publisher (Taylor and Francis, online ISSN: 1448-5540). The citation is as follows:

Leach, E. C., Burwell, C. J., Ashton, L. A., Jones, D. N. & Kitching, R. L. 2016. Comparison of point counts and automated acoustic monitoring: detecting birds in a rainforest biodiversity survey. Emu, 116, 305-309. https://doi.org/10.1071/MU15097

My contribution to the paper involved field work, data collection, analysis and preparation of the manuscript. Chris Burwell and Louise Ashton assisted with field work, analysis and editing the manuscript. Darryl Jones assisted with editing the manuscript. Roger Kitching was responsible for the sampling design and assisted with editing the manuscript.

(Signed) ______(21/9/17) (Countersigned) ______(21/9/17)

Elliot Leach Supervisor: Darryl Jones

(Countersigned) ______(21/9/17) (Countersigned) ______(21/9/17)

Associate Supervisor: Chris Burwell Supervisor: Roger Kitching

(Countersigned) ______(21/9/17) Co-author: Louise Ashton

CSIRO PUBLISHING Emu, 2016 , 116, 305 – 309 http://dx.doi.org/10.1071/MU15097

Comparison of point counts and automated acoustic monitoring: detecting birds in a rainforest biodiversity survey

Elliot C. LeachA,D, Chris J. BurwellA,B, Louise A. AshtonC, Darryl N. JonesA and Roger L. KitchingA

AEnvironmental Futures Research Institute, Griffith University, Nathan, Qld 4111, Australia. BBiodiversity Program, Queensland Museum, South Brisbane, Qld 4101, Australia. C Soil Biodiversity Group, Life Sciences Department, Natural History Museum, Cromwell Road, London SW7 5BD, UK. DCorresponding author. Email: [email protected]

2.1 Abstract. To monitor assemblages of animals, ecologists need effective methods for detecting and recording the distributions of species within target areas in restricted periods of time. In this study, we compared the effectiveness of a traditional avian biodiversity assessment technique (point counts) with a relatively new method (automated acoustic recordings) along an elevational gradient in rainforest in central Queensland, Australia. On average, point counts detected more species than acoustic recordings of an equivalent length of time (n = 40, P = <0.001). We suggest these results are driven by the visual detection of additional species during point counts. Despite the fact that point counts detected more species than acoustic recordings, datasets generated by both methods showed similar patterns in the community response to change in elevation. There was significant overlap in the species detected using both methods, but each detected several unique species. Consequently, we recommend the use of both techniques in tandem for future biodiversity assessments, as their respective strengths and weaknesses are complementary.

Additional keywords: Eungella National Park, point counts, rainforest bird diversity.

Received 22 September 2015, accepted 26 January 2016, published online 5 April 2016

2.2 Introduction

Traditionally, point counts have been the most commonly used method to rapidly assess the avian biodiversity of an area (Celis-Murillo et al. 2009; Venier et al. 2012). There are, though, a number of problems with point counts: in areas with high species richness, observers may be ‘swamped’ and fail to record some species (Haselmayer and Quinn 2000); differences in skill between observers may manifest as inter-observer effects (Hobson et al. 2002; Celis-Murillo et al. 2009); there may be a lack of trained personnel to conduct surveys (Holmes et al. 2014); birds may respond to the observers presence (Digby et al. 2013) and the cost of extended fieldwork necessary to build long-term datasets is often prohibitive (Thomas and Marques 2012). In recent years, technological advances have led to the development of

22 automated acoustic recorders that may prove an effective, low-cost way of monitoring biodiversity over large spatial and temporal scales (Thomas and Marques 2012; Wimmer et al. 2013).

Table 2.1 Relative performance of automated acoustic recorders (ARs) versus traditional methods in previous studies Traditional methods varied among studies but included call counts (Digby et al. 2013), call playback (McGuire et al. 2011), area searches (Wimmer et al. 2013) and a variety of methods based on point counts (remaining studies). Equivalent effort refers to whether the same amount of time was allocated to ARs and the alternative method during the study.

Authors Equivalent effort? Target Better performing method? Haselmayer and Quinn Yes Community Equal performance (2000) Hobson et al. (2002) Yes Community Equal performance Acevedo et al. (2009) No Community Acoustic recordings McGuire et al. (2011) No One species Equal performance Celis-Murillo et al. (2012) Yes Community Equal performance Venier et al. (2012) Yes Community Traditional method (distance point count) Digby et al. (2013) Yes One species Equal performance Wimmer et al. (2013) No Community Acoustic recordings Holmes et al. (2014) No Three Equal performance species Klinbeil & Willig (2015) Yes Community Traditional methods (point count) Sedláček et. al. (2015) Yes Community Equal performance

In avian biology, acoustic recorders have been primarily used for the detection of rare, poorly-known or elusive species (Goyette et al. 2011; McGuire et al. 2011; Buxton and Jones 2012; Zwart et al. 2014), but also for the identification of individuals within a population (Laiolo et al. 2007; Grava et al. 2008), behavioural studies (Mennill 2011; Osmun and Mennill 2011), detection of migrating birds (Farnsworth and Russell 2007) and community- level diversity studies (Haselmayer and Quinn 2000; Rempel et al. 2005; Wimmer et al. 2013). Some advantages of acoustic recordings are the creation of a permanent record for later analysis (or reanalysis), removal of observer bias, minimisation of disturbance, potential increases in the scale of monitoring schemes, and the ability to sample more easily in difficult

23 conditions (Hobson et al. 2002; Mennill et al. 2012; Rempel et al. 2013; Holmes et al. 2014). So far, the use of acoustic recorders to monitor avian biodiversity in Australia has been limited (but see Cai et al. 2007; Wimmer et al. 2013) and has taken place only in open forest habitats.

Previous comparisons between acoustic recordings and traditional techniques have most commonly found no significant difference between methods with regards to the number of species or individuals detected (Table 2.1). However, the majority of these studies have been conducted in temperate regions. Given the challenges associated with detecting birds in rainforests (Anderson 2011), and the lack of previous studies in Australia, we aimed to determine whether acoustic recorders deployed in subtropical Australian rainforest could be used to generate a dataset comparable to one generated using point counts. A secondary aim was to assess whether this dataset would be suitable for analysis of community-level ecological patterns.

2.3 Methods

Study site

Field work was conducted in two sessions: between the 4th of November and the 1st of December 2013 and between the 19th of March and the 15th of April 2014 in Eungella National Park (21.03°S, 148.64°E) and Pelion State Forest (21.06°S, 148.68°E), approximately 80km west of Mackay on the central Queensland coast. The first sampling session was at the beginning of the wet season and the second at the conclusion. The average annual rainfall in the region is 2199.1mm (Dalrymple Heights weather station, see http://www.bom.gov.au/climate/data/index.shtml), and the average daily maximum temperature between the sampling periods was 31.6°C (temperature was recorded at three sites per elevation using iButtons (DS21 Thermochron® iButton®, Maxim Integrated, CA, USA)). In general, the vegetation of the survey locations can be classified as complex notophyll vine forest (Tracey 1982).

We established a total of 24 sites along a gradient that encompassed the full elevational distribution of rainforest in the study area; four replicate sites in each of six elevational bands of approximately 200, 400, 600, 800, 1000 and 1200m a.s.l. (± 56m), in a design broadly similar to that of Kitching et al. (2011). Replicate sites within each band were separated by at

24 least 400m, with some sites up to 5km distant. At each site a 20 x 20m, permanently marked quadrat was established.

Point Counts

EL conducted two ten-minute point counts at each site, in the morning and afternoon, in each sampling session. Morning counts were undertaken as close as possible to dawn and afternoon counts in the three hours before sunset. The occurrence of all bird species seen or heard at the site during the count period was recorded, but counts of individual birds were not. Morning and afternoon count data were pooled before analysis. Due to inclement weather, counts at the 1200m sites in 2013 were only partially completed.

Automated Audio Recordings

Each of the replicate sites was sampled using a Songmeter SM2+ (unit described in Mennill et al. (2012)) placed in a central position within the 20x20m quadrat in both the 2013 and 2014 sampling sessions. Songmeters were programmed to record for two and a half hours in the morning (beginning 20min before dawn) and for two hours in the afternoon (programmed to finish recording at sunset) at factory default settings.

To compare recordings with point counts, EL listened to five randomly selected two-minute sections from each recording and noted all bird species heard. If ambient environmental noise (e.g. from rain) was too loud to identify the birds that were calling, a different section of the recording was chosen. This generated a dataset that was directly comparable (in terms of elapsed time) to that of the point count data.

Data analysis

Initially we were interested in whether point counts or acoustic recordings were more effective at detecting species. Rather than compare the mean number of species recorded by each method, since species richness was likely to vary between sites and elevations, we compared the mean proportions of the total species detected at sites. For each sampling occasion we pooled the available data from both methods at each site (n=40), calculated the proportions of the total species that were detected by each method, and compared their means using a one-way ANOVA.

We also investigated elevational patterns of community composition shown by each method and the results of both methods combined. To investigate these we first pooled data from both

25 methods and sampling sessions to create a data set based on the presence or absence of species at each site. Using the package Plymouth Routines in Multivariate Ecological Research (PRIMER 6TM) (Clarke and Gorley 2006), non-metric multi-dimensional scaling (NMDS) ordinations were generated to visualise elevational patterns displayed by bird assemblages. Three separate ordinations were generated based on data from either point counts, acoustic recordings or both methods combined. Ordinations were generated from matrices of pairwise comparisons between sites, based on Sørensen’s similarity index values with 999 permutations. We examined the relationship between the compositional turnover in bird assemblages and elevation by plotting the Sørensen similarity values between all pairs of sites against the difference in elevation between the sites. Separate graphs were generated for the three datasets; point counts, acoustic recordings and the combined methods (see Figure 2.1). The relationship between community similarity and elevational distance was assessed using simple linear regression.

2.4 Results

A total of 60 species were detected in our study: 50 species by point counts and 45 by recordings. Of these, 36 species were detected by both methods (Sørensen’s similarity index value of 0.774). Point counts detected 14 unique species (of which six were recorded only once), and recordings detected 8 unique species (of which three were detected once). There were more overall detections in the first sampling occasion (943; November – December 2013) than in the second (735; March 2014). On average, point counts detected a significantly greater proportion of the total species present at a site than acoustic recordings across all sites (n = 40, p = <0.001).

Ordinations based on all three datasets, point counts (Figure 2.1a), acoustic recordings (Figure 2.1b) and both methods combined (Figure 2.1c), showed clear elevational stratification of avian assemblages. In all cases low elevation and highland sites were located at opposite sides of the ordination panes. All three datasets showed that as the elevational distance between a pair of sites increased, community similarity decreased (point counts (R2 = 0.40, p = <0.001); recordings (R2 = 0.44, p = <0.001); combined (R2 = 0.41, p = <0.001)), indicating significant turnover in community composition with changing elevation.

2D Stress: 0.16 90 R² = 0.40 p < 0.001 80 70 60 50 40 30

Sørensen's Similarity Similarity Sørensen's 20 10 0 0 200 400 600 800 1000 Δ elevation (m) (a)

2D Stress: 0.16 90 R² = 0.44 p < 0.001 80 70 60 50 40 30

Sørensen's Similarity Similarity Sørensen's 20 10 0 0 200 400 600 800 1000 Δ elevation (m) (b)

2D Stress: 0.12 100 R² = 0.41 p < 0.001 90 80 70 60 50 40 30

Sørensen's Similarity Similarity Sørensen's 20 10 0 0 200 400 600 800 1000 Δ elevation (m) (c) Figure 2.1 Non-metric multidimensional scaling ordinations (left) and graphs of Sørensen similarity between all possible pairs of sites plotted against the difference in elevation between site pairs for (a) point counts (b) recordings and (c) both methods combined. In all ordinations, inverted triangles 200m sites; open squares 400m sites; closed squares 600m sites; diamonds 800m sites, circles 1000m sites and triangles 1200m sites.

2.5 Discussion

In this study, we found that point counts detected significantly more species than acoustic recordings across an elevational gradient in subtropical rainforest. The detection of ‘quiet’

27 species from visual cues during point counts was at least partially responsible for this result. Obviously, only species that vocalise can be detected by acoustic recorders (Laiolo 2010; Mennill et al. 2012). Due to the short duration of our study we also had to ‘take what we could get’ in terms of the recordings. As an example, increased ambient noise from water flow in a nearby creek at the conclusion of the wet season in 2014 made analysis of the recordings from some 200m sites difficult.

Despite these challenges, we found no differences in the community level patterns (elevational stratification and turnover in species composition with increasing elevational distance between sites) that were generated by data from the two methods (see Figure 2.1). While a direct comparison between methodologies was the focus of this study, we must acknowledge that it is unusual for acoustic recordings to be so short. Indeed, the main advantage of using acoustic recorders is that they can be left in situ for long periods of time (Klingbeil and Willig 2015). Increasing the recording duration will increase the number of species detected but leads to longer processing time when automatic processing is not possible (Wimmer et al. 2013).

In this study, the first of its kind in Australia, we have demonstrated that it is possible to generate a dataset suitable for the analysis of community-level patterns in birds using acoustic recorders. From a management perspective, this is encouraging; advances in technology and manufacturing are leading to the development of significantly cheaper and more efficient bioacoustic recorders and it is now theoretically possible to have a large number of these devices installed at monitoring stations year-round. However, the need to manually process the recordings, the adverse effects that the weather and the ambient acoustic environment can have on recording quality, the cost of equipment and hardware problems are all weaknesses of automated acoustic monitoring which must be considered in future projects. Another potential issue, the inability to accurately generate abundance or density data from acoustic recordings, may be compensated for by using point counts to target certain species or areas of interest. We found that a combination of point counts and acoustic recordings detected more species than either method alone, and that each method contributed several unique species to the pooled dataset.

The permanent record of a community generated by acoustic recording allows for the reanalysis of the data using novel techniques in the future (Sedláček et al. 2015). Each time a recording of a community is made, it has the potential to become part of a larger dataset for

28 future analysis. As climate change and habitat fragmentation continue to threaten avian communities worldwide (Şekercioğlu et al. 2008; Bregman et al. 2014), and in Australia (Anderson et al. 2013; Garnett and Franklin 2014), the baseline data generated by current acoustic recordings may assist in the assessment of future impacts. Acoustic recorders are becoming an important tool for ecological research and monitoring, and we recommend their use alongside traditional methods of biodiversity assessment in the future.

2.6 Acknowledgements

Fieldwork in Eungella was made possible through funding provided by the Mackay Regional Council, and was undertaken with the support of the Griffith University School of Environment. This research was carried out under permit number WITK12193512 issued by the Queensland Department of Environment and Heritage Protection. We thank the many people who volunteered their time as part of the Eungella Biodiversity Survey in 2013 and 2014, in particular numerous members of Birdlife Mackay, Gary Hendicott and Jenny Wooding, Kate Brunner, Maya Harrison, Sarah Maunsell, Claire Gely and Jessica Mackie. We also thank those scientists who made up the research team, particularly Akihiro Nakamura and Bill McDonald.

2.7 Appendix 2

The Songmeter SM2+ units used in this study were set to record using the factory default settings. This meant recording at 16 kHz in stereo, with a bit rate of 16kb and 0dB gain. Files were saved in the lossless .wav format.

Chapter 3

In this chapter, I use automated recording units to remotely monitor the elevational preferences and vocal behaviour of two cryptic thrush species along an elevational gradient in subtropical rainforest in north-eastern New South Wales. The study was conducted in the northern regions of the Gondwana Rainforests of Australia World Heritage Area.

Chapter 3 is a sole authored paper that has been published in the journal Australian Field Ornithology. Copyright permission to reproduce this article has been granted by the publisher (Birdlife Australia). The citation is as follows:

Leach, E. C. 2016. Notes on the Zoothera thrushes in the Tweed Range of north-eastern New South Wales. Australian Field Ornithology, 33, 240-243. https://doi.org/10.20938/afo33240243

My contribution to the paper involved field work, data collection and processing, analysis and preparation of the manuscript. My supervisors edited the manuscript.

(Signed) ______(21/9/17)

Elliot Leach

Australian Field Ornithology 2016, 33, 240–243 http://dx.doi.org/10.20938/afo33240243

Notes on the Zoothera thrushes in the Tweed Range of north-eastern New South Wales

Elliot C. Leach Environmental Futures Research Institute, Griffith University, Nathan QLD 4111, Australia Email: [email protected]

3.1 Abstract. The Zoothera thrush complex is represented on the Australian mainland by the Bassian Thrush Z. lunulata and the Russet-tailed Thrush Z. heinei. These species are sympatric at several locations on the eastern coast. Often, these populations occupy different elevations, with the Bassian Thrush preferring higher elevations, though reasons for this are poorly understood. I present data from automated acoustic recordings made of these species in the Border Ranges and Mebbin National Parks of north-eastern New South Wales between ~300 m and 1100 m above sea-level over a 1-year period from June 2015 to May 2016. Bassian Thrushes were recorded most frequently in October, typically at or above 900 m. Russet-tailed Thrushes were recorded most frequently in August, at or below 700 m. Differences in elevational preference between the species may be driven by several factors including adaptation to cold, avoidance of interspecific competition and avoidance of hybridisation.

3.2 Introduction

The Bassian Thrush Zoothera lunulata and Russet-tailed Thrush Z. heinei are two secretive occupants of forest habitats along the eastern coast of Australia. They are the only native representatives of the worldwide thrush family, Turdidae, that breed on the Australian mainland (Cooper 1959; Higgins et al. 2006). Both species were previously included in the widespread Scaly Thrush Z. dauma complex until Ford (1983) proposed that they were distinct species, based on differences in plumage, song and morphological characters (Higgins et al. 2006). This classification was supported by Christidis & Boles (1994), and has now been accepted by numerous authorities (Christidis & Boles 2008). The two species

32 are sympatric in several regions of Australia (Barrett et al. 2003). In southeastern Queensland and north-eastern New South Wales (NSW), they are sympatric in subtropical rainforest, where the Bassian Thrush (hereafter BT) usually occupies higher elevations than the Russet- tailed Thrush (hereafter RT) (Ford 1983; Gosper & Holmes 2002; Higgins et al. 2006). The ecological processes responsible for the elevational preferences of the species remain poorly understood. In this paper, I present data on both Zoothera species along an elevational gradient in the Border Ranges and Mebbin National Parks of north-eastern NSW generated using automated acoustic recorders.

3.3 Methods

Sites

Data were collected during a 12-month period from 10 June 2015 to 31 May 2016 in two national parks (NP) in north-eastern NSW: Border Ranges NP (28°23′S, 153°3′E) and Mebbin NP (28°27′S, 153°10′E). Sampling occurred along an elevational gradient with a design like that used by Leach et al. (2016). A total of 20 sites was used: four sites in each of five elevational bands, ~300, 500, 700, 900 and 1100 m above sea-level. Sites within each band were separated by at least 400 m, with some sites being up to 16 km distant because of their location on different sides of the Tweed Range.

Recordings

Bioacoustic Audio Recorders (BARs) (see http://www. frontierlabs.com.au) were used to make automated acoustic recordings of the soundscape at these sites as part of an ongoing study in the region. A BAR was deployed at one site per elevational band in each month of sampling. Five BARs were used so that recordings were collected simultaneously at the five different elevations. Each month, the sites were visited to collect data and to move each BAR to a new site within its elevational band. BARs were set to record for five 2-minute periods each day. The first 2-minute recording began 10 minutes before sunrise. Each subsequent recording was then made after a 10-minute delay, such that the last recording each day took place 30 minutes after sunrise.

Analysis

After collecting the data from the BARs, I listened to each 2-minute recording and noted all bird species heard. I then extracted 2-minute recordings in which I detected Zoothera song for further analysis. Recordings that were compromised by high levels of ambient noise (e.g.

33 from heavy rain or strong wind) were not processed; in total, from all elevations, 795 such recordings were excluded. Zoothera contact calls (see Figure 3.1a) were noted but not included in the dataset, as distinguishing the two species based only on contact call is difficult, if not impossible. The recording period (June 2015–May 2016) encompasses the probable breeding and non-breeding seasons of both Zoothera species in the region (BT breeds July–September; RT breeds September–January) (Cooper 1959; Higgins et al. 2006).

3.4 Results

In total, 5825 2-minute recordings were processed. Songs of Zoothera species were detected in 472 2-minute recordings between June 2015 and May 2016. BT songs were detected 239 times and RT songs 248 times. In general, the number of thrush detections increased with elevation, a pattern driven by the increased frequency of BT detections at 900 m and 1100 m sites (see Table 3.1).

Bassian Thrush (BT)

BT song was quite variable, though typically consisted of two or three melodic whistles, which have been previously described as pilop whee-er (Higgins et al. 2006) (see Figure 3.1b). BT songs were detected in all elevational bands except 300 m: four times at 500 m, 12 times at 700 m, 92 times at 900 m and 131 times at 1100 m (Table 3.1). BT songs were detected most frequently from October to the end of December (see Figure 3.2). The low number of songs recorded in November was probably because recordings from the higher elevational bands in that month were often affected by rain, reducing the number of ‘good days’ available for processing; ~40% of records from 900 m and 1100 m sites in November were affected. BT songs were not detected in February or March.

Russet-tailed Thrush (RT)

RT song was typically two whistles: pee-eer hoo, with the first phrase descending in tone and the second remaining steady (see Figure 3.1c). RT songs were detected in all elevational bands: 84 times at 300 m, 75 times at 500 m, 77 times at 700 m, nine times at 900 m and three times at 1100 m (Table 3.1). RT songs were detected most often in August (Figure 3.2). RT song was detected only once in December (at 700 m), and not at all in March.

Time (seconds).

Figure 3.1 Sonograms of recordings made in June 2015–May 2016 from the study area in the Border Ranges and Mebbin National Parks, New South Wales, containing thrush Zoothera spp. calls and songs. Arrows indicate the midpoints of the calls and songs described as follows: (a) two contact calls of a thrush Zoothera sp. from 0 to 1 second and again from c. 5.5 to 6 seconds, in the 6.5–10-kHz frequency band (contact calls of the species discussed here are virtually identical); (b) three examples of Bassian Thrush (BT) song, from 0 to 1 second, from 2.75 to 3.5 seconds and from c. 5.5 to 6.5 seconds (the latter being an example of the pilop-whee-er call) in the 2–3-kHz frequency band; (c) two examples of Russet- tailed Thrush (RT) song, from 1 to 2.25 seconds and from c. 5 to 6 seconds in the 3–4-kHz frequency band.

Elevation (m) Total no. of recordings No. in which BT No. in which RT detected detected

300 1165 0 84

500 1215 4 75

700 1240 12 77

900 1155 92 9

1100 1050 131 3

BT RT

3.5 Discussion

From this study, there is evidence that in north-eastern NSW BT and RT appear to be largely confined to high (900–1100 m) and low–mid (300–700 m) elevations, respectively. Gosper & Holmes (2002) also conducted fieldwork in the region and found BT to be resident at 550–

1100 m and RT to be partially nomadic between 250 and 750 m, though the methodology and locations used differed from those in the present study. Regardless, both studies found that BT prefer high elevations and RT low– mid elevations. This may simply be a consequence of BT being better adapted to cold than RT (which can be seen in the wider distributions of the two species: see Higgins et al. 2006). Reduction of interspecific competition for resources or avoidance of interspecific pair formation are two other plausible explanations for the observed pattern (Hinde 1959; Slabbekoorn & Smith 2002; Irwin et al. 2008).

Interspecific communication (and possible avoidance) between BT and RT is presumably based on vocalisations. I found that BT and RT in north-eastern NSW have very different songs in terms of frequency and structure (though, interestingly, very similar contact calls) (Figure 3.1). Several factors may have contributed to the divergence in signalling between the species. The acoustic adaptation hypothesis proposes that differences in habitat structure may lead to signal divergence (Irwin et al. 2008), as different sonic frequencies can travel more efficiently in different environments (Boncoraglio & Saino 2007). Differences in morphology have implications for the ability to produce certain vocalisations; for example, body mass is negatively correlated with song frequency, and large bills are not well suited to the production of certain sounds (see Seddon 2005). The species recognition hypothesis suggests that difference in song structure (or other traits) is an evolutionary adaptation to reduce hybridisation between closely related species (Sætre et al. 1997; Seddon 2005). In the context of the present study, the differences in habitat structure between the low elevations of the Border Ranges (characterised by palm forest) and the high elevations (characterised by cloud forest dominated by Antarctic Beech Nothofagus moorei) suggest that there may be some support for the acoustic adaptation hypothesis. However, it is necessary to test each hypothesis specifically before conclusions can be drawn.

The two species appear to have different calling ‘seasons’, with RT detected most frequently in August and BT detected most frequently in October. There were no records of either species in March (Figure 3.2). If vocalisation is related to breeding, the delayed peak in the detections of BT may be explained by the fact that conditions at low elevations become suitable for breeding earlier than those at higher elevations, though further work is needed to confirm this.

Generating a dataset on secretive or cryptic species such as that presented in this study can be difficult using traditional fieldwork techniques. Using automated acoustic recorders allows researchers to monitor such populations over long time periods relatively cheaply and easily. There are many other potential applications of such technology; for a broader discussion of the advantages and disadvantages associated with using automated acoustic recorders in ornithological contexts, see Leach et al. (2016). Further investigation of the vocalisations of the Zoothera thrushes in Australia would benefit from the use of automated acoustic recorders.

3.6 Conclusions

This study contributes to the general understanding of the ecology of both BT and RT, in particular by providing empirical evidence for the elevational preferences of the two species in north-eastern NSW. Data such as these are important for conservation purposes: as the climate continues to warm, the populations of many Australian birds (particularly montane species) are predicted to decline (Williams et al. 2003; Chambers et al. 2005; Garnett & Franklin 2014). The use of automated acoustic recorders to monitor cryptic species that are best detected by call (e.g. thrushes Zoothera spp., ground parrots Pezoporus spp.) is recommended in future studies. Fieldwork is continuing at the sites used in this study, and further work will focus on community-level patterns of elevational stratification in the avifauna.

3.7 Acknowledgements

The fieldwork was carried out as part of a PhD project supported by Griffith University, Queensland, under NSW scientific license SL101514. Birds Queensland provided me with a generous research grant in 2015, which allowed me to purchase the BARs used in this study. Gus Daly, Jess Mackie and Chris Burwell read an earlier draft of this paper and their comments, as well as those of Dennis Gosper, Hugh Ford and James Fitzsimons, improved the manuscript. The supervision of Roger Kitching and Darryl Jones is also gratefully acknowledged.

Received 10 May 2016, accepted 12 October 2016, published online 22 December 2016

3.8 Appendix 3

Bioacoustic Recording Units were used to make automated acoustic recordings of the soundscape in this study. These weatherproof units have a single omni-directional microphone, which was set to record at 44.1 kHz with a bit rate of 16kb and a 40dB gain. Files were saved in the lossless .wav format.

Chapter 4

In this chapter we identify avian indicators of elevation suitable for the long-term monitoring of the effects of climate change on subtropical rainforest bird assemblages. The study was conducted in the northern regions of the Gondwana Rainforests of Australia World Heritage Area.

Statement of contribution to a co-authored paper

Chapter 4 is a co-authored paper that has been published by the journal Pacific Conservation Biology. Permission to reproduce this article has been granted by the publisher (CSIRO Publishing, online ISSN: 2204-4604). The citation is as follows:

Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018. Identifying avian indicators of elevation in the Gondwanan Rainforests of Australia. Pacific Conservation Biology. https://doi.org/10.1071/PC18039

My contribution to the paper involved field work, data collection and processing, analysis and preparation of the manuscript. Chris Burwell assisted with field work, data processing and editing the manuscript. Darryl Jones assisted with editing the manuscript. Roger Kitching was involved in the sampling design and assisted with editing the manuscript.

(Signed) ______(10/7/18) (Countersigned) ______(10/7/18)

Elliot Leach Supervisor: Darryl Jones

(Countersigned) ______(10/7/18) (Countersigned) ______(10/7/18)

Associate Supervisor: Chris Burwell Supervisor: Roger Kitching

Identifying avian indicators of elevation in the Gondwanan Rainforests of Australia

Elliot C. Leacha*, Chris J. Burwella,b, Darryl N. Jonesa and Roger L. Kitchinga a Environmental Futures Research Institute, Griffith University, Nathan, Qld 4111, Australia. b Biodiversity Program, Queensland Museum, South Brisbane, Qld 4101, Australia.

* Corresponding author, email address: [email protected], telephone: +61 434 633 218 (E. Leach)

Author ORCIDs: Elliot Leach 0000-0003-2108-4402, Chris Burwell NA, Darryl Jones 0000-0002- 2565-9456, Roger Kitching 0000-0002-6798-6041

Keywords: climate change; birds; elevational gradient; automated acoustic monitoring; indicator species; bioacoustics

4.1 Abstract

Many montane avian communities are likely to be impacted negatively by future climate change. The ability to monitor these ecosystems effectively is therefore a priority. As species are expected to track their preferred climates by moving upwards in elevation, using indicator species of elevations has been suggested as a climate change monitoring strategy and has been explored for a variety of taxa in eastern Australia. Birds have great potential as vertebrate indicators due to their familiarity, detectability, and well-known life histories. We used automated recording units (ARUs) and point counts to sample the avifauna along two elevational gradients in sub-tropical rainforest in north-eastern New South Wales, Australia. We used the indicator value protocol to identify avian indicators of elevation suitable for long-term monitoring. Pairs of species were more reliable than single species as indicators, and searching for indicators of elevational ranges (e.g. 300m – 500m) proved more effective than looking for indicators of single elevations (e.g. 300m). Point counts and ARUs were equally effective at determining indicators of elevations and ARUs performed particularly well in spring. We present avian indicator sets of lowland and highland sites, which provide a baseline for future monitoring of the effects of climate change on the region’s avifauna. The methodology employed here is broadly suitable for similar studies elsewhere. We propose that the use of ARUs to identify indicator species of elevations is an effective strategy for monitoring the effects of climate change on montane avian communities worldwide.

4.2 Introduction

Anthropogenic impacts on biodiversity over the last 500 years are driving a sixth mass extinction event, and have led to the onset of the ‘Anthropocene’ (Crutzen, 2006; Dirzo et al., 2014; Ceballos et al., 2015). Consequently, the ability to monitor ecosystems effectively is an important challenge in conservation biology (Dornelas et al., 2013). Globally, the expected long-term trend is for species to track their climatically-driven niches by moving upwards in elevation or towards the poles (Root et al., 2003; Freeman and Class Freeman, 2014). These shifts in distribution have already been documented for a variety of taxa, including birds, but not all species respond in predictable ways (Auer and King, 2014; Roth et al., 2014; Campos- Cerqueira et al., 2017). Tingley et al. (2012), for example, found that birds in the Sierra Nevada Mountains moved upslope in response to increasing temperature, but downslope in response to increasing precipitation. Detecting such changes can be achieved through large- scale monitoring schemes such as bird atlas projects which provide information on avian populations and distributions over regional or continental scales (Barrett et al., 2003; Balmer et al., 2014). Atlases provide valuable information, but data collection at the large spatio- temporal scales required can be problematic due to difficulties of access and associated costs (Dunn and Weston, 2008). Such approaches often rely on volunteers, potentially introducing bias (Griffioen and Clarke, 2002; Harris et al., 2016). Online databases such as those available from eBird or the Australian Bird and Bat Banding Scheme also provide valuable insights into movements or changes in population structure (Clark, 2017).

At the local scale, various methods for detecting range shifts in avian populations have been proposed and tested, but most require comparisons with historical data, intensive fieldwork, or both (Shoo et al., 2006; Anderson et al., 2013; Freeman and Class Freeman, 2014). A common approach is to sample along environmental gradients. Here, climatic conditions often change in predictable ways, allowing researchers to make ‘space-for-time’ substitutions with their data and forecast potential climate change impacts (Williams et al., 2003; Strong et al., 2011). Elevational gradients are particularly popular, with over 750 results returned on a Web of Science search for the topics “elevational gradient” and “climate” among the biological sciences (EL unpublished data, 2/3/2018).

Including indicator species of elevations in long-term monitoring schemes has been suggested as a way to quantify the impacts of future climatic changes (Nakamura et al., 2016). Indicator species can be used to assess or monitor environmental conditions or as surrogates for a larger

42 community (Dale and Beyeler, 2001; Carignan and Villard, 2002; Butler et al., 2012). Groups or ‘predictor sets’ of species may be more effective as indicators than single species (Butler et al., 2012; Kitching and Ashton, 2013). A good indicator species (or set) is one that has both high specificity (is restricted to the sites of interest) and high fidelity (occurs frequently at these sites) (see Dufrêne and Legendre, 1997). Chambers (2008) determined that birds are especially useful as indicator species due to their ease of detection and identification, well understood taxonomy, high trophic position and the level of public interest in the taxon.

Most birds are readily identified by their vocalisations, and recently, the use of automated recording units (ARUs) in ornithological research has become popular (Shonfield and Bayne, 2017). Using ARUs to identify avian indicators of elevation along mountain ranges, which represent hotspots of avian diversity (Orme et al., 2005), is a cost-effective approach that may be capable of detecting and monitoring long-term responses to climatic change. Such a strategy should be a conservation priority (Krause and Farina, 2016), but has so far been largely unexplored.

The ecosystems of the mountain ranges within the Gondwana Rainforests of Australia World Heritage Area (hereafter, Gondwanan rainforests) are particularly vulnerable to future climate change (Laurance et al., 2011b). We deployed ARUs along two elevational gradients in the Gondwanan rainforests to monitor the avian community remotely over a one year period. To validate the data from the ARUs, we also conducted monthly point counts along each of the gradients. Here, we use these data and the classical Indicator Value Protocol (Dufrêne and Legendre, 1997) to identify avian indicators of elevation in the Gondwanan rainforests that should be suitable for cost-effective, long-term monitoring of the region’s avifauna.

4.3 Methods

Study area

Fieldwork was undertaken in two national parks (NP) in north-eastern New South Wales: Border Ranges NP (28°23′S, 153°3′E) and Mebbin NP (28°27′S, 153°10′E) which together comprise a protected area of subtropical rainforest of 15, 940ha. Both parks are part of the Gondwanan rainforests (Kitching et al., 2010). The region has a moist subtropical climate, with 65-70% of rainfall occurring in the warmer months (October – March) (McDonald, 2010). Below 700-800m above sea level (asl), the forest is generally classified as warm

43 subtropical rainforest, with most sites characterised as either complex notophyll vine forest or wet sclerophyll forest (McDonald, 2010). Above 800m asl there is a gradual transition to cool subtropical rainforest and, above 1000m asl, a further transition to cool temperate rainforest characterised as microphyll mossy forest or thicket. The high elevation forest is dominated by Antarctic Beech (Nothophagus moorei) (McDonald, 2010).

Two elevational gradients (‘Sheepstation Creek’ and ‘Bar Mountain’, hereafter SSC and BM respectively) were established within the study area. Two replicate sites, separated by a minimum of 400m, were located in each of five elevational bands ~300, 500, 700, 900 and 1100m asl along each gradient (see Fig. 5.1). All sites were situated on soils derived from basalt (www.environment.nsw.gov.au/eSpade2WebApp). All fieldwork was undertaken between September 2015 and October 2016 under Scientific Licence number SL101514 issued by the New South Wales National Parks and Wildlife Service.

Acoustic Recordings Automated acoustic recordings were made of the soundscape using five Bioacoustic Recording Units (see www.frontierlabs.com.au). These are weatherproof ARUs with a single omni-directional microphone set to record at 44.1 kHz with a bit rate of 16kb and a 40dB gain. Files were saved in the lossless .wav format. An ARU was deployed in a central location at one site per elevational band in each month of sampling. Five ARUs were used so that recordings were made simultaneously in each elevational band each month. ARUs were set to record for five 2-minute periods each day. The first 2-minute recording began 10 minutes before sunrise. Each subsequent recording was then made after an 8-minute delay (2 minutes ‘on’, 8 minutes ‘off’), so that the last recording each day started 30 minutes after sunrise. In this way, the dawn chorus was effectively captured while minimising recording and processing times. For a discussion on the benefits of making recording ‘snapshots’ such as those described here, see Wimmer et al. (2013). Each month, the sites were visited to collect the recordings and to move each ARU to a new site within its elevational band. All recordings (ca. 222 hours) were listened to by EL, who recorded the presence of all bird species heard. Spectrograms were inspected as necessary to assist with difficult identifications. Recordings in which the sound quality was compromised by strong wind or heavy rain were not processed.

Point counts Each month, EL conducted a ten minute audio-visual survey at a central location in each of the 20 sites. During each survey, the abundance of all species seen or heard within a 100 m radius of a central point was recorded. Counts were undertaken in the 3 hours following dawn, when bird activity was at its peak, and all sites were typically surveyed over a four day period in fine weather conditions. Inclement weather and road closures in the national parks meant that we were unable to sample at the same time each month for the entire year. Consequently, field trips in March and April, as well as in September and October in 2016 were combined. In total, we obtained 11 monthly datasets during the sampling period.

Analysis Acoustic Recordings A daily species list for each site being sampled in each month was generated. Due to poor weather conditions and hardware complications, the number of recordings available for processing from each site sometimes differed. Thus, rather than using direct counts of species present, we standardised the species lists by using detection frequencies for each species. Detection frequencies were defined as the number of days that a species was detected at a site divided by the number of days that suitable recordings were made at that site. Species’ detectability throughout the year was expected to change on a seasonal basis (Gage and Axel, 2014). To identify which season would yield the highest number of significant indicators, we compiled four ‘seasonal’ datasets: spring (September – November 2015), summer (December 2015 – February 2016), autumn (March – May 2016) and winter (June – August 2016). We also created a year-long ‘total’ ARU dataset by calculating the overall detection frequency of each species at each site throughout the study.

Point counts EL conducted a ten minute point count of the avifauna present at each site in each month between September 2015 and October 2016. Due to inclement weather and road closures in the national parks, we were unable to sample in September 2016. Sampling thus resulted in 11 monthly datasets, which were then combined to create ‘seasonal’ summaries for each site like those generated using the ARU data as follows: spring (October and November 2015), summer (December 2015 – February 2016), autumn (April and May 2016) and winter (June – August 2016). Each ‘seasonal’ dataset summed the counts for each species from each

45 contributing month. We also generated a ‘total’ point counts dataset by pooling the four seasonal datasets for each site.

Community composition and sampling sufficiency To examine changes in community composition along the gradient, we used the ‘total’ ARU and ‘total’ point counts datasets to create ordination plots based on non-metric multidimensional scaling (NMDS) using the vegan package ver. 2.4-1 and the ggord package ver. 1.0.0 in R statistical software version 3.3.2 (Oksanen et al., 2016; Beck, 2017; R Core Team, 2017). To determine whether our sampling regime accurately represented the avian community at each elevation, we used data from both methods to create coverage-based rarefaction curves for each elevational band using the iNEXT package ver. 2.0.14 in R (Chao et al., 2014). Rarefaction curves to test the sampling sufficiency of both methods across the entire gradient were also created for each site.

Indicator species protocol The indicator species protocol was developed by Dufrêne and Legendre (1997) and is used to identify indicators of certain habitat types (in this study, elevations). Species are assigned an indicator value score (IndVal), expressed as a percentage, based on their specificity (relative abundance) and fidelity (relative frequency). A maximum IndVal of 100% for a species would be generated if it was present at all sites within an elevational band, and found only within that elevational band. We used the indicspecies package ver. 1.7.1 in R (De Cáceres and Legendre, 2009) to identify indicator species of elevations using the ‘seasonal’ and ‘total’ ARU and point count datasets.

We first looked for individual indicator species of single elevational bands (e.g. 300m, 500m), and then combinations of wider elevational ranges (e.g. 300-500m, 500-700m, 300-700m), excluding the entire elevational range (i.e. 300 – 1100m). Combining elevational bands increases the amount of information available and is a more flexible approach than looking at single elevations (De Cáceres et al., 2010). Next, we examined all possible pairwise species combinations (indicator ‘sets’ sensu Kitching & Ashton (2013)) in each dataset to determine whether species pairs performed better as indicators than individual species, for both single elevational bands and wider elevational ranges, again excluding the entire elevational range. A pair of species, when found together, provides more ecological information than a single species (De Cáceres et al., 2012). We limited the species combinations to pairs, rather than

46 trios or higher-order groups, due to the computation times involved and the problems arising from multiple comparison. We only considered indicator species and pairs of indicator species (hereafter ‘indicators’) with an IndVal above 85% to be reliable (cf. the subjective benchmark (70%) of Van Rensburg et al. (1999)). To determine the statistical significance of these indicators, a random reallocation of sites within site groups was used (Dufrêne and Legendre, 1997); we chose to permute the data 999 times. We summed the number of indicators identified at each single elevation or range of elevations using all datasets to make comparisons between seasons and methods.We then tested whether the indicators generated by each method reliably classified elevational bands in a new dataset. To test the indicators identified using ARU data, we used them to classify elevations in the point count dataset, and vice versa. To perform these analyses, we used the functions indicators and predict.indicators in the indicspecies package in R (De Cáceres and Legendre, 2009).

After the first round of indicators had been identified from the ‘total’ datasets, we examined their IndVals. We then ranked the indicators in descending order of indicator value, and selected a set of five lowland (<700m asl) and five highland (>900m asl) indicators suitable for long-term monitoring, based on their relative IndVals and ecological ‘breadth’ (Butler et al., 2012). To minimise the likelihood of Type 1 errors affecting our indicator sets, we recomputed the significances (9,999 permutations; significance level 0.01) of the selected indicator species in isolation as recommended by Dufrêne and Legendre (1997).

4.4 Results

In total, we detected 65 bird species using the point count methodology during the sampling period; 57 of these were also detected using the automated recording units (ARUs). We found the greatest species richness in spring (Fig. 4.1A). No species detected by ARUs were not also detected using point counts. Coverage-based rarefaction curves showed over 99% sample coverage was achieved by both methods in each of the five elevational bands, indicating that our sampling regime accurately represented the avian community along the gradients. ARU data had higher sample coverage than point count data on a site-by-site basis, which was likely due to the wider ‘reach’ of the recording protocol across the dawn chorus. The lowest sample coverage (point count data at site 300SCA) was 93%. Species richness generally declined with increasing elevation (Fig. 4.1B). NMDS ordinations on both ARU and point

47 count data showed progressive turnover in avian community composition along the gradient from the lowlands to the highlands (Fig. 4.2).

Figure 4.1 Bird species richness in the Gondwanan rainforests. The total number of species detected along the entire elevational gradient (300m – 1100m asl) in each season by ARUs and point counts is shown in A. The total number of species detected in each elevational band using ARUs and point counts is shown in B.

The indicator value protocol identified 31 significant indicators in the seasonal ARU datasets and 109 significant indicators in the seasonal point count datasets (Table 4.1), as well as 262 significant indicators in the ‘total’ ARU dataset and 161 significant indicators in the ‘total’ point counts dataset. Of these, the majority were species pairs rather than single species. In the ‘total’ datasets, the majority of indicators were associated with elevational ranges rather than single elevations. The two methods often found the same indicators for a given elevation or elevational band, but many indicators were unique to each method. For lists of all indicators identified, see Appendices 4.1 – 4.4.

Table 4.1 Numbers of indicators found in each seasonal dataset using ARUs and point counts. Indicators are individual species or pairs of species

ARUs Point Counts

Spr. Sum. Aut. Win. Spr. Sum. Aut. Win.

Elevations

300m 10 9

500m 2 8

700m 8

900m 1

1100m 16

Elevational ranges

300+500m 3 1 1

500+700m 1 4 3 3 5 10

700+900m 8 5

900+1100m 2 1

Lowlands (300-700m) 1 11 1

Mid-elev. (500-900m) 22 1 1

Highlands (700-1100m) 2 1

Totals 22 3 4 3 26 39 11 33

*Spr. = spring, Sum. = summer, Aut. = Autumn, Win. = winter.

To assess how well indicators found using one method predicted elevations in the other dataset, we tested all elevational bands (except for 700m, where no indicators were found using either method). Indicators of 300m generated using ARU data successfully identified all four 300m sites in the point count data. These indicators were also significantly associated with 500m sites, but the relationship was weaker than for 300m sites. Indicators of 1100m

50 sites generated using ARUs successfully identified the target sites in the point count data. Indicators of 500m and 900m generated using ARUs, however, failed to accurately predict the corresponding elevational bands in the point count dataset. Using the point count data, we found indicators of 300m accurately predicted 300m sites in the ARU dataset, although these indicators were also associated with single sites at each of 500m, 700m and 900m. Indicators of all other elevational bands generated using point count data failed to accurately predict the corresponding elevations in the recordings dataset. Indicators generated using ARU data were more effective at classifying elevations in the point count data than vice versa.

In the seasonal ARU datasets, indicators were only associated with elevational ranges. However, in the seasonal point count data, approximately half of all indicators were associated with single elevations; the rest were associated with elevational ranges. Using ARUs, 71% of indicators were found in spring. Point count data had a fairly even spread of indicators among seasons (with the exception of autumn, in which only 11 indicators were found). There were seasonal changes in the indicators associated with elevations and elevational ranges, both in the number of indicators and in the indicator species or indicator pairs themselves (see Appendix 4.4). The Scarlet Myzomela (Myzomela sanguinolenta) and the Spotted Pardalote (Pardalotus punctata) were significantly associated with low elevation sites and, supporting the results of Leach (2016), the Bassian Thrush (Zoothera lunulata) was significantly associated with the highest elevations in the current study. Indicator sets for the long-term monitoring of lowland and highland sites in the Gondwanan rainforests were selected by EL based on relative IndVals, the foraging guild membership of species (Butler et al. 2012), and the likelihood of detectability in the field. These indicator sets are presented in Table 4.2.

4.5 Discussion

We identified avian indicators of elevation using both ARUs and point counts. Both methods identified avian indicators of lowland (300-700m), mid-elevation (500-900m) and highland (700-1100m) sites in the study area. Pairs of species performed better as indicators than single species, most likely due to the wider ‘reach’ of species pairs in terms of ecological variety (De Cáceres et al., 2012; Kitching and Ashton, 2013). Similarly, more indicators of elevational ranges were identified than indicators of single elevations. This may be due to the larger amounts of data, and consequent increase in statistical power available when analysing

51 elevational ranges. The relatively low numbers of indicators associated with single elevations could also be related to the high vagility of the sampled species – the movements of Australian birds in the landscape are often complex, potentially confounding the search for meaningful indicators (Ford, 2013). Most indicators identified were associated with low elevations (300-500m) or high elevations (900-1100m). A potential explanation for this pattern is that at mid-elevations the low-elevation and high-elevation specialists overlap (the mid-domain effect suggested by Colwell & Lees (2000)).

Nakamura et al. (2016) obtained similar results and suggested the relative paucity of indicators at mid-elevations was due to the limited elevational range of the mountains in the study system. From a monitoring perspective, this is not necessarily a problem: future researchers will be able to examine whether the indicators found at low elevations in contemporary studies have moved upslope in response to warming temperatures (Krause and Farina, 2016), and whether the high elevation indicators have declined.

Both ARUs and point counts indicated a steady decline in bird species richness with increasing elevation in the Gondwanan rainforests, a pattern generally associated with avian diversity in humid mountain ranges (McCain, 2009). The relatively low numbers of species present at high elevations are, nonetheless, of great conservation value – these are the species most likely to be adversely affected by climatic warming (Laurance et al., 2011a). Ongoing warming in the study region is predicted to lead to average mean temperatures 0.4 to 1.3 °C above the climate of 1986–2005 by 2030 (CSIRO and Bureau of Meteorology, 2015). Given that small increases in temperature can lead to steep declines in population size amongst upland species (Shoo et al., 2005), which also generally have poor dispersal capabilities (Sodhi, 2002; Laurance et al., 2011a), the long-term persistence of the high-elevation avian community in the study region is unlikely. Given that the majority of remnant habitat in this region is under World Heritage protection (McDonald, 2010), future changes in avian community structure are most likely to be driven by contemporary changes in climate (Lindström et al., 2013).

As such, species that are restricted to particular elevational ranges should play an important role in monitoring or conservation schemes as indicators of climate change (Larsen et al.,

Table 4.2 Recommended avian indicator sets for the long-term monitoring of lowland and highland sites in the Gondwanan rainforests

Lowland sites (300 -700m asl) Highland sites (700 - 1100m asl)

Indicators IndVal p-value Indicators IndVal p-value

Scarlet Myzomela + Spotted Pardalote 1 < 0.001 Bassian Thrush 0.95 0.001

Rose-crowned Fruit-Dove + Wompoo Fruit-Dove 1 < 0.001 Albert’s Lyrebird + Bassian Thrush 0.95 0.001

Mistletoebird + Rose-crowned Fruit-Dove 0.98 < 0.001 Bassian Thrush + Rose Robin 0.94 0.001

Wompoo Fruit-Dove + White-throated Treecreeper 0.97 < 0.001 Bassian Thrush + White-throated Treecreeper 0.94 0.001

Scarlet Myzomela + Yellow-faced Honeyeater 0.97 < 0.01 Bassian Thrush + Wonga Pigeon 0.93 0.001

*Scientific names are as follows. Albert’s Lyrebird (Menura alberti); Bassian Thrush (Zoothera lunulata);Mistletoebird (Dicaeum hirundinaceum); Rose-crowned Fruit-Dove (Ptilinopus regina); Rose Robin (Petroica rosea); Scarlet Myzomela (Myzomela sanguinolenta); Spotted Pardalote (Pardalotus punctatus); White-throated Treecreeper (Cormobates leucophaea); Wonga Pigeon (Leucosarcia melanoleuca); Wompoo Fruit-Dove (Ptilinopus magnificus); Yellow-faced Honeyeater (Caligavis chrysops).

2007; Nakamura et al., 2016). Previous researchers have identified elevational stratification in diverse taxa in our study area including trees (Laidlaw et al., 2011), ants (Burwell and Nakamura, 2011) and moths (Ashton et al., 2016). The utility of these taxa as indicator species of elevation was investigated by Nakamura et al. (2016), who recommended that different taxonomic groups also be investigated to determine the spatial and temporal variation in their distributions.

Indicators identified using ARU data performed better than those derived from point count data at classifying the elevations of target sites in new datasets, successfully identifying sites in low and high elevations. In contrast, the point count indicators only identified low elevation sites. As discussed above, indicators of mid-elevations were not reliably found in either dataset. Reducing the acceptable IndVal (which we set at 85%) may have led to more mid-elevation indicators being identified. However, this represented an undesirable trade-off between increasing the number of indicators and reducing their reliability. We must acknowledge here that the data we collected were potentially spatially correlated – point counts were performed at the same sites where ARUs were set to record. However, the data were otherwise independent, and ARUs were set to begin recording only after the monthly point counts were completed. Additionally, we used a significance level of 0.01 when calculating the significance of our indicator sets to minimise the possibility of Type 1 errors being made.

We used our initial indicator value analysis to guide the selection of our indicator sets. As such, the indicator sets selected for long-term monitoring of lowland and highland sites intentionally represented most feeding guilds present in the sampled community apart from raptors. They included a range of nectarivores, frugivores, insectivores and omnivores, most of which employ contrasting foraging strategies. As such, our chosen indicators generally met the expectations of Butler et al. (2012), who stated that indicators should be representative of the wider community from which they are selected. If such representation is achieved, indicators can ‘condense’ complex information about biodiversity, making change more understandable for policymakers, environmental managers, and the public (Gregory, 2006; Herrando et al., 2016).

The baseline data presented here is a ‘snapshot’ of the avian community of the northern Gondwanan rainforests in 2015-2016 and provides future researchers with a valuable reference sample. Our results indicate that passive acoustic monitoring schemes should

54 prioritise sampling in spring, when most birds are vocalising, and when we detected the most species. In making this recommendation, we acknowledge that our data come from a limited number of sites, and represent a single year of sampling. However, Gage and Axel (2014) found that the soundscape power associated with mid-frequencies (3-7 kHz) (the range in which most birds vocalise) was consistently highest in spring over a four year period.

Extensions of our analyses considering different functional or taxonomic groups may allow for broader application of this methodology (Vandewalle et al., 2010; Trindade-Filho et al., 2012; Nakamura et al., 2016), as will the constant improvement in automated species detection (Ross and Allen, 2014). Designing automatic detection algorithms for every species in this community may prove problematic, due to the complexity of the soundscape and the high levels of false positives associated with this approach (Joshi et al., 2017). Automating data processing, however, would facilitate long-term monitoring, and should be a research priority. Focusing on the creation of recognisers for target species, such as the indicator sets of lowland and highland sites identified in this study, may be a first step towards automation. Recent developments in the design of continental scale acoustic monitoring projects like the Australian Acoustic Observatory (see http://goo.gl/GvsCYm for details) will eventually provide ample data for monitoring purposes, and would benefit from the inclusion of elevational or latitudinal indicators. For now, elevational gradients in the Wet Tropics have been well studied (Williams et al., 2010; Anderson, 2011) and further testing and validation of our methodology could be carried out in that system, or in any montane region with a well understood avifauna. Using ARUs in combination with the indicator value protocol as presented here has the potential to become a cost-effective and powerful way to monitor the effects of climate change on montane avian communities worldwide.

4.6 Acknowledgements

We wish to thank Birds Queensland for a grant supporting the purchase of the Bioacoustic Recording Units used in this study, as well as Frontier Labs for their assistance with the design and maintenance of these devices. EL also thanks Jim O’Brien for rescuing a recorder and returning it safe and sound. Staff of the Kyogle National Parks and Wildlife Office provided EL with assistance during field work, as did numerous volunteers – a special thank

55 you goes to Jess Mackie. The comments of an anonymous reviewer, as well as those of Harry Recher, greatly improved the manuscript.

4.7 Appendices

The following Appendices will be included online as Supplementary Material complementing the published paper and are presented here to aid the readers’ understanding of this chapter.

Appendix 4.1: Species Codes The following are the species codes used in the analysis and in the Supplementary Material in alphabetical order. The taxonomic order and nomenclature follows IOC World Bird Names, version 8.1 (see http://goo.gl/QqvwaU). Appendix 7.5 lists IOC names alongside names from Christidis and Boles (2008).

ABRT Australian Brushturkey (Alectura lathami) ALLY Albert's Lyrebird (Menura alberti) AULO Australian Logrunner (Orthonyx temminckii) BATH Bassian Thrush (Zoothera lunulata) BFCS Black-faced Cuckooshrike (Coracina novaehollandiae) BFMO Black-faced Monarch (Monarcha melanopsis) BRCD Brown Cuckoo-Dove (Macropygia phasianella) BRCU Brush Cuckoo (Cacomantis variolosus) BRGE Brown Gerygone (Gerygone mouki) BRTB Brown Thornbill (Acanthiza pusilla) CROS Crimson Rosella (Platycercus elegans) CSTI Crested Shriketit (Falcunculus frontatus) EMDO Common Emerald Dove (Chalcophaps longirostris) ESPI Eastern Spinebill (Acanthorhynchus tenuirostris) EWHI Eastern Whipbird (Psophodes olivaceus) EYRO Eastern Yellow Robin (Eopsaltria australis) FTCU Fan-Tailed Cuckoo (Cacomantis flabelliformis) GOWH Australian Golden Whistler (Pachycephala pectoralis) GRCA Green Catbird (Ailuroedus crassirostris)

GRFA Grey Fantail (Rhipidura albiscapa) GRST Grey Shrikethrush (Colluricincla harmonica) KIPA Australian King Parrot (Alisterus scapularis) LBSW Large-billed Scrubwren (Sericornis magnirostra) LEHE Lewin's Honeyeater (Meliphaga lewinii) LKOO Laughing Kookaburra (Dacelo novaeguineae) LSTH Little Shrikethrush (Colluricincla megarhyncha) MSTB Mistletoebird (Dicaeum hirundinaceum) NOPI Noisy Pitta (Pitta versicolor) OBOR Olive-backed Oriole (Oriolus sagittatus) PCUR Pied Currawong (Strepera graculina) PRIF Paradise Riflebird (Ptiloris paradiseus) PYRO Pale-yellow Robin (Tregellasia capito) RALO Rainbow Lorikeet (Trichoglossus moluccanus) RCFD Rose-crowned Fruit Dove (Ptilinopus regina) RORO Rose Robin (Petroica rosea) RTTH Russet-tailed Thrush (Zoothera heinei) RUFA Rufous Fantail (Rhipidura rufifrons) SABB Satin Bowerbird (Ptilonorhynchus violaceus) SBCU Shining Bronze Cuckoo (Chrysococcyx lucidus) SCCO Sulphur-crested Cockatoo (Cacatua galerita) SCHE Scarlet Myzomela (Myzomela sanguinolenta) SILV Silvereye (Zosterops lateralis) SPDR Spangled Drongo (Dicrurus bracteatus) SPMO Spectacled Monarch (Symposiachrus trivirgatus) SPPA Spotted Pardalote (Pardalotus punctatus) STPA Striated Pardalote (Pardalotus striatus) STTB Striated Thornbill (Acanthiza lineata) SUFD Superb Fruit Dove (Ptilinopus superbus) THSP Thrush species (sp.) (Zoothera sp.)

TOPI Topknot Pigeon (Lopholaimus antarcticus) VART Varied Triller (Lalage leucomela) WBSW White-browed Scrubwren (Sericornis frontalis) WHPI White-headed Pigeon (Columba leucomela) WOFD Wompoo Fruit Dove (Ptilinopus magnificus) WOPI Wonga Pigeon (Leucosarcia melanoleuca) WTRE White-throated Treecreeper (Cormobates leucophaea) YFHE Yellow-faced Honeyeater (Calivagus chrysops) YTBC Yellow-tailed Black Cockatoo (Calyptorhynchus funereus)

YTSW Yellow-throated Scrubwren (Sericornis citreogularis)

Appendix 4.2: ARUs Here we present the indicators significantly associated with each elevation or elevational range using the ‘total’ ARU dataset. Species codes are available in Appendix 4.1. The total number of species includes all individual species as well as all possible pairwise combinations of species available for analysis. The elevation or elevational range in metres above sea level (a.s.l.) is then shown along with the number of associated indicators (#sps.) that met our criteria – an IndVal of >85%. All analysis was performed using the indicspecies package in R. For more information, such as the code used to run these analyses, please contact [email protected].

Total number of species: 1770 KIPA+RALO 0.889 0.010 ** RALO+SILV 0.889 0.008 ** List of indicators associated with GRFA+SCHE 0.887 0.005 ** each elevation or elevational PCUR+SCHE 0.886 0.005 ** range: BFMO+SCHE 0.886 0.004 ** WBSW+YFHE 0.881 0.034 * 300m a.s.l. #sps. 117 SILV+SPPA 0.879 0.003 ** IndVal p.value BFMO+RALO 0.879 0.009 ** SCHE+SPPA 1.000 0.002 ** RALO+WOFD 0.876 0.009 ** SCHE+YFHE 0.970 0.002 ** RUFA+SCHE 0.874 0.007 ** SPMO+YFHE 0.961 0.002 ** EYRO+YFHE 0.872 0.034 * SCHE+WTRE 0.958 0.003 ** BRCD+YFHE 0.872 0.034 * GOWH+SCHE 0.950 0.003 ** RALO 0.869 0.011 * CROS+SCHE 0.949 0.005 ** AULO+RALO 0.869 0.011 * SPPA+YFHE 0.947 0.003 ** BRCD+RALO 0.869 0.011 * RALO+YFHE 0.943 0.002 ** BRGE+RALO 0.869 0.011 * SCHE+SPMO 0.938 0.003 ** BRTB+RALO 0.869 0.011 * LKOO+SCHE 0.937 0.002 ** EWHI+RALO 0.869 0.011 * RALO+SPPA 0.935 0.007 ** EYRO+RALO 0.869 0.011 * RALO+SCHE 0.934 0.003 ** GOWH+RALO 0.869 0.011 * LBSW+SCHE 0.917 0.005 ** GRFA+RALO 0.869 0.011 * SCHE 0.916 0.005 ** GRST+RALO 0.869 0.011 * EWHI+SCHE 0.916 0.005 ** LBSW+RALO 0.869 0.011 * GRST+SCHE 0.916 0.005 ** LEHE+RALO 0.869 0.011 * LEHE+SCHE 0.916 0.005 ** PCUR+RALO 0.869 0.011 * AULO+SCHE 0.915 0.005 ** RALO+WBSW 0.869 0.011 * SCHE+WBSW 0.910 0.005 ** RALO+WTRE 0.869 0.011 * SBCU+SCHE 0.909 0.006 ** STPA 0.866 0.020 * EYRO+SCHE 0.907 0.005 ** AULO+STPA 0.866 0.020 * WTRE+YFHE 0.907 0.028 * BFMO+STPA 0.866 0.020 * BRCD+SCHE 0.904 0.005 ** BRCD+STPA 0.866 0.020 * SPMO+SPPA 0.900 0.006 ** BRGE+STPA 0.866 0.020 * RTTH+SCHE 0.897 0.017 * BRTB+STPA 0.866 0.020 * YFHE 0.895 0.032 * CROS+STPA 0.866 0.020 * LEHE+YFHE 0.895 0.032 * ESPI+STPA 0.866 0.020 * BRGE+SCHE 0.894 0.005 ** EWHI+STPA 0.866 0.020 * KIPA+SPMO 0.894 0.007 ** EYRO+STPA 0.866 0.020 * SILV+SPMO 0.894 0.004 ** GOWH+STPA 0.866 0.020 * KIPA+SCHE 0.892 0.005 ** GRCA+STPA 0.866 0.020 * LKOO+SPMO 0.891 0.007 ** GRFA+STPA 0.866 0.020 * AULO+YFHE 0.891 0.032 * GRST+STPA 0.866 0.020 * EWHI+YFHE 0.891 0.032 * KIPA+STPA 0.866 0.020 * GRST+YFHE 0.890 0.032 * LBSW+STPA 0.866 0.020 *

LEHE+STPA 0.866 0.020 * BRCD+SPMO 0.866 0.014 * LKOO+RALO 0.866 0.006 ** BRGE+SCCO 0.866 0.016 * LKOO+STPA 0.866 0.020 * BRGE+SPMO 0.866 0.014 * PCUR+STPA 0.866 0.020 * BRTB+SCCO 0.866 0.016 * RALO+STPA 0.866 0.020 * BRTB+SPMO 0.866 0.014 * RCFD+STPA 0.866 0.020 * ESPI+SCCO 0.866 0.014 * RTTH+STPA 0.866 0.020 * ESPI+SPMO 0.866 0.016 * RUFA+STPA 0.866 0.020 * EWHI+SCCO 0.866 0.016 * SBCU+STPA 0.866 0.020 * EWHI+SPMO 0.866 0.014 * SCHE+STPA 0.866 0.020 * EYRO+SCCO 0.866 0.016 * SILV+STPA 0.866 0.020 * EYRO+SPMO 0.866 0.014 * SPMO+STPA 0.866 0.020 * GOWH+SCCO 0.866 0.016 * SPPA+STPA 0.866 0.020 * GOWH+SPMO 0.866 0.014 * STPA+WBSW 0.866 0.020 * GRCA+SCCO 0.866 0.016 * STPA+WOFD 0.866 0.020 * GRCA+SPMO 0.866 0.016 * STPA+WTRE 0.866 0.020 * GRFA+SCCO 0.866 0.016 * STPA+YFHE 0.866 0.020 * GRFA+SPMO 0.866 0.014 * GOWH+YFHE 0.863 0.039 * GRST+SCCO 0.866 0.016 * BFMO+YFHE 0.861 0.024 * GRST+SPMO 0.866 0.014 * ESPI+SCHE 0.861 0.015 * KIPA+SCCO 0.866 0.014 * RALO+SBCU 0.859 0.012 * LBSW+SCCO 0.866 0.016 * BFMO+SPPA 0.857 0.007 ** LBSW+SPMO 0.866 0.014 * GRFA+YFHE 0.854 0.039 * LEHE+SCCO 0.866 0.016 * SPPA 0.853 0.009 ** LEHE+SPMO 0.866 0.014 * AULO+SPPA 0.853 0.009 ** MSTB+SCCO 0.866 0.012 * BRCD+SPPA 0.853 0.009 ** PCUR+SCCO 0.866 0.016 * BRGE+SPPA 0.853 0.009 ** PCUR+SPMO 0.866 0.014 * BRTB+SPPA 0.853 0.009 ** PYRO+SCCO 0.866 0.016 * EWHI+SPPA 0.853 0.009 ** RALO+SPMO 0.866 0.017 * EYRO+SPPA 0.853 0.009 ** RCFD+SCCO 0.866 0.016 * GOWH+SPPA 0.853 0.009 ** RCFD+SPMO 0.866 0.020 * GRFA+SPPA 0.853 0.009 ** RTTH+SCCO 0.866 0.016 * GRST+SPPA 0.853 0.009 ** RTTH+SPMO 0.866 0.017 * LBSW+SPPA 0.853 0.009 ** RUFA+SCCO 0.866 0.016 * LEHE+SPPA 0.853 0.009 ** RUFA+SPMO 0.866 0.013 * PCUR+SPPA 0.853 0.009 ** SBCU+SCCO 0.866 0.014 * SPPA+WBSW 0.853 0.009 ** SBCU+SPMO 0.866 0.016 * SPPA+WTRE 0.853 0.009 ** SCCO+SILV 0.866 0.017 * RALO+RUFA 0.851 0.010 ** SCCO+WBSW 0.866 0.016 * SCCO+WOFD 0.866 0.016 * 300-500m a.s.l. #sps. 75 SCCO+WTRE 0.866 0.016 * IndVal p.value SCCO+YTSW 0.866 0.016 * YTBC 0.867 0.019 * SPMO+WBSW 0.866 0.014 * AULO+YTBC 0.867 0.019 * SPMO+WOFD 0.866 0.014 * BRCD+YTBC 0.867 0.019 * SPMO+WTRE 0.866 0.014 * EWHI+YTBC 0.867 0.019 * KIPA+YTBC 0.864 0.022 * EYRO+YTBC 0.867 0.019 * BRTB+YTBC 0.863 0.019 * GOWH+YTBC 0.867 0.019 * SILV+YTBC 0.863 0.035 * GRFA+YTBC 0.867 0.019 * RCFD+SCHE 0.861 0.012 * GRST+YTBC 0.867 0.019 * BRGE+YTBC 0.860 0.019 * LEHE+YTBC 0.867 0.019 * SCHE+SILV 0.858 0.012 * PCUR+YTBC 0.867 0.019 * BRTB+SCHE 0.858 0.010 ** WBSW+YTBC 0.867 0.019 * SCHE+WOFD 0.857 0.010 ** WHPI+YTBC 0.867 0.019 * GRCA+SCHE 0.855 0.013 * WTRE+YTBC 0.867 0.019 * RALO+RCFD 0.853 0.013 * SCCO 0.866 0.016 * SPMO 0.866 0.014 * AULO+SCCO 0.866 0.016 * AULO+SPMO 0.866 0.014 * BFMO+SCCO 0.866 0.016 * BFMO+SPMO 0.866 0.014 * BRCD+SCCO 0.866 0.016 * 500-700m a.s.l. #sps. 2

IndVal p.value 300-700m a.s.l. #sps. 36 ALLY+WOFD 0.937 0.001 *** IndVal p.value ALLY+RCFD 0.926 0.029 * BFMO+WOFD 0.992 0.002 ** RCFD+WOFD 0.991 0.002 ** RTTH+WOFD 0.982 0.002 ** 900-1100m a.s.l. #sps. 27 WOFD 0.980 0.002 ** IndVal p.value BRGE+WOFD 0.980 0.002 ** BATH 0.953 0.004 ** EWHI+WOFD 0.980 0.002 ** ALLY+BATH 0.953 0.004 ** GRST+WOFD 0.980 0.002 ** AULO+BATH 0.953 0.004 ** LEHE+WOFD 0.980 0.002 ** BATH+BRGE 0.953 0.004 ** AULO+WOFD 0.979 0.002 ** BATH+BRTB 0.953 0.004 ** BRCD+WOFD 0.979 0.002 ** BATH+EWHI 0.953 0.004 ** RUFA+WOFD 0.978 0.003 ** BATH+EYRO 0.953 0.004 ** GOWH+WOFD 0.978 0.003 ** BATH+GOWH 0.953 0.004 ** GRFA+WOFD 0.978 0.002 ** BATH+LEHE 0.953 0.004 ** EYRO+WOFD 0.977 0.002 ** BATH+YTSW 0.953 0.004 ** GRCA+WOFD 0.977 0.002 ** BATH+ESPI 0.952 0.004 ** WOFD+WTRE 0.976 0.003 ** BATH+GRST 0.952 0.004 ** BRTB+WOFD 0.975 0.002 ** BATH+GRFA 0.951 0.004 ** LBSW+WOFD 0.975 0.002 ** BATH+BRCD 0.951 0.004 ** WBSW+WOFD 0.974 0.002 ** BATH+WBSW 0.951 0.004 ** PCUR+WOFD 0.972 0.002 ** BATH+LBSW 0.946 0.008 ** RCFD+RTTH 0.971 0.003 ** BATH+RORO 0.945 0.010 ** KIPA+WOFD 0.969 0.002 ** BATH+WTRE 0.943 0.008 ** BFMO+RTTH 0.967 0.003 ** BATH+KIPA 0.943 0.008 ** PYRO+WOFD 0.949 0.002 ** BATH+PCUR 0.936 0.010 ** NOPI+WOFD 0.939 0.005 ** BATH+GRCA 0.932 0.010 ** WOFD+YTSW 0.936 0.007 ** BATH+SILV 0.931 0.012 * BATH+WHPI 0.921 0.012 * Signif. codes: 0 ‘***’ 0.001 ‘**’ BATH+SABB 0.919 0.013 * 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 BATH+RUFA 0.916 0.014 * BATH+WOPI 0.908 0.010 ** BATH+NOPI 0.901 0.048 *

PYRO+RTTH 0.922 0.027 * MSTB+RTTH 0.915 0.028 * MSTB+RCFD 0.907 0.039 * FTCU+WOFD 0.866 0.025 * RTTH+YTBC 0.866 0.017 * WOFD+YTBC 0.866 0.016 * BFMO+FTCU 0.854 0.027 * FTCU+RCFD 0.853 0.020 * LBSW+YTBC 0.851 0.031 * GRCA+YTBC 0.850 0.031 *

500-900m a.s.l. #sps. 2 IndVal p.value ALLY+RTTH 0.911 0.018 * RORO+WOFD 0.851 0.037 *

700-1100m a.s.l. #sps. 3 IndVal p.value CROS+WOPI 0.921 0.028 * CROS+RORO 0.898 0.034 * BATH+CROS 0.893 0.018 *

Appendix 4.3: Point Counts Here we present the indicators significantly associated with each elevation or elevational range using the ‘total’ point counts dataset. Species codes are available in Appendix 4.1. The total number of species includes all individual species as well as all possible pairwise combinations of species available for analysis. The elevation or elevational range in metres above sea level (a.s.l.) is then shown along with the number of associated indicators (#sps.) that met our criteria – an IndVal of >85%. All analysis was performed using the indicspecies package in R. For more information, such as the code used to run these analyses, please contact [email protected].

Total number of species: 2556 1100m a.s.l #sps. 3 IndVal p.value List of indicators associated with BATH+WOPI 0.935 0.003 ** each elevation or elevational BATH+NOPI 0.913 0.004 ** range: BATH+RORO 0.882 0.004 **

300m a.s.l #sps. 4 IndVal p.value SCHE+SPPA 1.000 0.002 ** 300-500m a.s.l #sps. 54 EWHI+SCHE 0.933 0.002 ** IndVal p.value PCUR+SCHE 0.882 0.003 ** SCHE 0.935 0.002 ** PRIF+SCHE 0.877 0.005 ** SPMO 0.935 0.001 *** AULO+SCHE 0.935 0.002 ** AULO+SPMO 0.935 0.001 *** 900m a.s.l #sps. 25 BFMO+SCHE 0.935 0.002 ** IndVal p.value BFMO+SPMO 0.935 0.001 *** THSP 0.866 0.022 * BRCD+SCHE 0.935 0.002 ** ALLY+THSP 0.866 0.022 * BRCD+SPMO 0.935 0.001 *** AULO+THSP 0.866 0.022 * BRGE+SCHE 0.935 0.002 ** BRCD+THSP 0.866 0.022 * BRGE+SPMO 0.935 0.001 *** BRGE+THSP 0.866 0.022 * BRTB+SCHE 0.935 0.002 ** BRTB+THSP 0.866 0.022 * BRTB+SPMO 0.935 0.001 *** ESPI+THSP 0.866 0.022 * EYRO+SCHE 0.935 0.002 ** EWHI+THSP 0.866 0.022 * EYRO+SPMO 0.935 0.001 *** EYRO+THSP 0.866 0.022 * GOWH+SCHE 0.935 0.002 ** GOWH+THSP 0.866 0.022 * GOWH+SPMO 0.935 0.001 *** GRCA+THSP 0.866 0.022 * GRFA+SCHE 0.935 0.002 ** GRFA+THSP 0.866 0.022 * GRFA+SPMO 0.935 0.001 *** GRST+THSP 0.866 0.022 * GRST+SCHE 0.935 0.002 ** KIPA+THSP 0.866 0.022 * GRST+SPMO 0.935 0.001 *** LEHE+THSP 0.866 0.022 * LEHE+SCHE 0.935 0.002 ** PCUR+THSP 0.866 0.022 * LEHE+SPMO 0.935 0.001 *** RORO+THSP 0.866 0.022 * MSTB+SCHE 0.935 0.002 ** RUFA+THSP 0.866 0.022 * MSTB+SPMO 0.935 0.001 *** SABB+THSP 0.866 0.022 * NOPI+SPMO 0.935 0.001 *** SBCU+THSP 0.866 0.022 * PYRO+SPMO 0.935 0.001 *** SILV+THSP 0.866 0.022 * RCFD+SCHE 0.935 0.002 ** THSP+WBSW 0.866 0.022 * RCFD+SPMO 0.935 0.001 *** THSP+WHPI 0.866 0.022 * RUFA+SCHE 0.935 0.002 ** THSP+WTRE 0.866 0.022 * RUFA+SPMO 0.935 0.001 *** THSP+YTSW 0.866 0.022 * SCHE+SILV 0.935 0.002 **

SCHE+WBSW 0.935 0.002 ** MSTB+RCFD 0.984 0.002 ** SCHE+WOFD 0.935 0.002 ** WOFD+WTRE 0.969 0.001 *** SCHE+WTRE 0.935 0.002 ** WOFD 0.968 0.001 *** SILV+SPMO 0.935 0.001 *** BRGE+WOFD 0.968 0.001 *** SPMO+WBSW 0.935 0.001 *** LEHE+WOFD 0.968 0.001 *** SPMO+WOFD 0.935 0.001 *** GRFA+WOFD 0.966 0.001 *** SPMO+WTRE 0.935 0.001 *** GOWH+WOFD 0.965 0.001 *** SPMO+YTSW 0.935 0.001 *** AULO+WOFD 0.962 0.001 *** RCFD+RUFA 0.906 0.012 * SILV+WOFD 0.962 0.001 *** RUFA+WOFD 0.884 0.025 * BRCD+WOFD 0.960 0.001 *** EWHI+SPMO 0.866 0.014 * GRFA+RCFD 0.959 0.002 ** GRCA+SPMO 0.866 0.013 * BRTB+WOFD 0.958 0.001 *** KIPA+SCHE 0.866 0.006 ** RCFD 0.956 0.004 ** KIPA+SPMO 0.866 0.008 ** LEHE+RCFD 0.956 0.004 ** LBSW+SCHE 0.866 0.007 ** BRGE+RCFD 0.955 0.004 ** LBSW+SPMO 0.866 0.015 * GOWH+RCFD 0.955 0.004 ** NOPI+SCHE 0.866 0.009 ** AULO+RCFD 0.953 0.004 ** PCUR+SPMO 0.866 0.013 * BRCD+RCFD 0.953 0.004 ** PRIF+SPMO 0.866 0.010 ** EYRO+WOFD 0.953 0.002 ** PYRO+SCHE 0.866 0.011 * RCFD+WTRE 0.953 0.004 ** RTTH+SPMO 0.866 0.005 ** BRTB+RCFD 0.952 0.005 ** SCHE+SPMO 0.866 0.011 * GRST+WOFD 0.952 0.001 *** SCHE+YTSW 0.866 0.010 ** RCFD+SILV 0.951 0.005 ** MSTB+WOFD 0.947 0.002 ** EYRO+RCFD 0.946 0.006 ** 500-700m a.s.l #sps. 2 GRST+RCFD 0.946 0.004 ** IndVal p.value PCUR+WOFD 0.943 0.001 *** GRCA+RCFD 0.935 0.003 ** PRIF+RCFD 0.941 0.003 ** GRCA+WOFD 0.927 0.002 ** PRIF+WOFD 0.930 0.005 ** PCUR+PRIF 0.926 0.007 ** LBSW+WOFD 0.925 0.007 ** EWHI+WOFD 0.924 0.004 ** 900-1100m a.s.l #sps. 22 BFMO+WOFD 0.921 0.003 ** IndVal p.value WOFD+YTSW 0.921 0.005 ** ALLY+RORO 0.907 0.004 ** KIPA+RCFD 0.913 0.009 ** RORO+SBCU 0.907 0.006 ** EWHI+RCFD 0.912 0.022 * BATH 0.866 0.010 ** MSTB+PRIF 0.908 0.022 * ALLY+BATH 0.866 0.010 ** RCFD+YTSW 0.908 0.017 * AULO+BATH 0.866 0.010 ** WBSW+WOFD 0.906 0.018 * BATH+BRCD 0.866 0.010 ** KIPA+WOFD 0.904 0.007 ** BATH+BRGE 0.866 0.010 ** PYRO+WOFD 0.894 0.013 * BATH+BRTB 0.866 0.010 ** PYRO+RCFD 0.892 0.018 * BATH+ESPI 0.866 0.010 ** BATH+EWHI 0.866 0.010 ** BATH+EYRO 0.866 0.010 ** 500-900m a.s.l #sps. 4 BATH+GOWH 0.866 0.010 ** IndVal p.value BATH+GRFA 0.866 0.010 ** GRCA 0.953 0.023 * BATH+GRST 0.866 0.010 ** BRGE+GRCA 0.953 0.023 * BATH+KIPA 0.866 0.010 ** GRCA+LEHE 0.953 0.023 * BATH+LEHE 0.866 0.010 ** GRCA+LBSW 0.910 0.025 * BATH+RUFA 0.866 0.010 ** BATH+SBCU 0.866 0.010 ** BATH+SILV 0.866 0.010 ** 700-1100m a.s.l #sps. 4 BATH+WBSW 0.866 0.010 ** IndVal p.value BATH+WTRE 0.866 0.011 * ALLY+ESPI 0.898 0.036 * BATH+YTSW 0.866 0.010 ** EWHI+RORO 0.890 0.016 * ESPI+RORO 0.866 0.030 * RORO+SABB 0.866 0.018 * 300-700m a.s.l #sps. 45 IndVal p.value Signif. codes: 0 ‘***’ 0.001 ‘**’ RCFD+WOFD 1.000 0.001 *** 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Appendix 4.4: Seasonal Changes Here we present the indicators significantly associated with each elevation or elevational range using the ‘seasonal’ ARU and point counts datasets. Species codes are available in Appendix 4.1. The total number of species includes all individual species as well as all possible pairwise combinations of species available for analysis in that season. The elevation or elevational range in metres above sea level (a.s.l.) is then shown along with the number of associated indicators (#sps.) that met our criteria – an IndVal of >85%. All analysis was performed using the indicspecies package in R. For more information, such as the code used to run these analyses, please contact [email protected].

ARUs Spring Summer

Total number of species: 1770 Total number of species: 1770

List of indicators associated with List of indicators associated with each elevation or elevational each elevation or elevational range: range:

500-900m a.s.l. #sps. 22 500-700m a.s.l. #sps. 1 IndVal p.value IndVal p.value BFMO+RCFD 1.000 0.046 * WHPI+WOFD 0.98 0.033 * BFMO+WOFD 0.993 0.046 * RCFD 0.982 0.023 * 300-700m a.s.l. #sps. 3 AULO+RCFD 0.982 0.023 * IndVal p.value BRCD+RCFD 0.982 0.023 * RCFD+YTSW 0.993 0.044 * BRGE+RCFD 0.982 0.023 * PRIF+SILV 0.983 0.043 * EWHI+RCFD 0.982 0.023 * PYRO+RUFA 0.978 0.045 * EYRO+RCFD 0.982 0.023 * GOWH+RCFD 0.982 0.023 * 500-900m a.s.l. #sps. 1 GRFA+RCFD 0.982 0.023 * IndVal p.value GRST+RCFD 0.982 0.023 * WBSW+WOFD 0.978 0.027 * LEHE+RCFD 0.982 0.023 * RCFD+WOFD 0.982 0.023 * RCFD+WHPI 0.981 0.018 * LBSW+RCFD 0.981 0.046 * Autumn RCFD+RUFA 0.981 0.046 * RCFD+WTRE 0.979 0.018 * Total number of species: 1770 PCUR+RCFD 0.976 0.046 * BRTB+RCFD 0.976 0.046 * List of indicators associated with each elevation or elevational KIPA+RCFD 0.972 0.046 * range: PRIF+RCFD 0.972 0.046 * RCFD+WBSW 0.971 0.046 * 500-700m a.s.l. #sps. 4 Indval p.value ALLY+WOFD 1 0.046 * PRIF+WOFD 1 0.046 * PYRO+WOFD 1 0.046 * WOFD+YTSW 1 0.046 *

Winter EYRO+RORO 1.000 0.003 ** GRST+RORO 1.000 0.003 ** Total number of species: 1770 RORO 0.894 0.011 * BRGE+RORO 0.894 0.011 * List of indicators associated with EWHI+RORO 0.894 0.011 * each elevation or elevational GOWH+RORO 0.894 0.011 * range: LEHE+RORO 0.894 0.011 * RORO+YTSW 0.894 0.011 * 500-700m a.s.l. #sps. 30 AULO+RORO 0.866 0.022 * IndVal p.value CROS+RORO 0.866 0.021 * THSP 1.000 0.044 * GRST+WOPI 0.866 0.017 * ALLY+THSP 1.000 0.044 * NOPI+RORO 0.866 0.022 * AULO+THSP 1.000 0.044 * RORO+SBCU 0.866 0.021 * BRCD+THSP 1.000 0.044 * RORO+SILV 0.866 0.016 * BRGE+THSP 1.000 0.044 * RORO+WOPI 0.866 0.017 * BRTB+THSP 1.000 0.044 * CSTI+PRIF 1.000 0.044 * 300-500m a.s.l #sps. 3 CSTI+THSP 1.000 0.044 * IndVal p.value ESPI+THSP 1.000 0.044 * SPMO 0.866 0.019 * EWHI+THSP 1.000 0.044 * GOWH+SPMO 0.866 0.013 * EYRO+THSP 1.000 0.044 * LEHE+SPMO 0.866 0.019 * GOWH+THSP 1.000 0.044 * GRCA+THSP 1.000 0.044 * 500-700m a.s.l #sps. 3 GRFA+THSP 1.000 0.044 * IndVal p.value GRST+THSP 1.000 0.044 * GRCA+PRIF 0.892 0.007 ** LBSW+THSP 1.000 0.044 * GRCA 0.891 0.009 ** LEHE+THSP 1.000 0.044 * GRCA+LBSW 0.866 0.019 * LKOO+THSP 1.000 0.044 * MSTB+THSP 1.000 0.044 * 700-1100m a.s.l #sps. 2 PCUR+THSP 1.000 0.044 * IndVal p.value PRIF+THSP 1.000 0.044 * BRTB+YTSW 0.877 0.027 * PYRO+THSP 1.000 0.044 * EWHI+YTSW 0.868 0.042 * RTTH+THSP 1.000 0.044 * SABB+THSP 1.000 0.044 * THSP+WBSW 1.000 0.044 * THSP+WTRE 1.000 0.044 * Summer THSP+YTSW 1.000 0.044 * Total number of species: 1378 GRCA+LKOO 0.965 0.044 * LKOO+SABB 0.965 0.044 * List of indicators associated with CSTI+LKOO 0.943 0.044 * each elevation or elevational range: Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 300m a.s.l #sps. 10 IndVal p.value SCHE 0.935 0.001 *** BRCD+SCHE 0.935 0.001 *** Point Counts BRGE+SCHE 0.935 0.001 *** LEHE+SCHE 0.935 0.001 *** Spring RCFD+SCHE 0.935 0.001 *** AULO+SCHE 0.926 0.002 ** Total number of species: 1596 BFMO+SCHE 0.926 0.002 ** GOWH+SCHE 0.913 0.003 ** List of indicators associated with each elevation or elevational GRFA+SCHE 0.913 0.005 ** range: EWHI+SCHE 0.866 0.017 *

500m a.s.l #sps. 1 IndVal p.value 300-500m a.s.l #sps. 1 GRCA+SPMO 0.866 0.015 * IndVal p.value BFMO+RCFD 0.874 0.010 ** 1100m a.s.l #sps. 16 IndVal p.value 500-700m a.s.l #sps. 5 BRTB+RORO 1.000 0.003 **

IndVal p.value IndVal p.value GRCA+RCFD 0.946 0.001 *** KIPA+RORO 0.894 0.008 ** RCFD+YTSW 0.929 0.003 ** WOFD+YTSW 0.904 0.004 ** 900-1100m a.s.l #sps. 1 GRCA+KIPA 0.901 0.008 ** IndVal p.value GRCA+WOFD 0.899 0.004 ** BRTB+RORO 0.866 0.015 *

700-900m a.s.l #sps. 8 IndVal p.value Winter EYRO+WHPI 0.935 0.004 ** WHPI+YTSW 0.926 0.003 ** Total number of species: 1128 WHPI 0.903 0.012 * WHPI+WTRE 0.886 0.008 ** List of indicators associated with BRGE+WHPI 0.885 0.015 * each elevation or elevational BRCD+WHPI 0.879 0.016 * range: LEHE+WHPI 0.879 0.015 * AULO+WHPI 0.862 0.018 * 500m a.s.l #sps. 8 IndVal p.value 300-700m a.s.l #sps. 11 GOWH+GRCA 1.000 0.002 ** IndVal p.value GOWH+WBSW 0.913 0.004 ** BRCD+RCFD 0.949 0.005 ** GOWH+PRIF 0.866 0.015 * RCFD 0.946 0.005 ** GOWH+WOFD 0.866 0.015 * LEHE+RCFD 0.944 0.005 ** GRCA+GRFA 0.866 0.021 * BRGE+RCFD 0.939 0.005 ** GRCA+KIPA 0.866 0.015 * RCFD+WOFD 0.913 0.005 ** GRCA+PYRO 0.866 0.023 * WOFD 0.891 0.020 * KIPA+WOFD 0.866 0.015 * LEHE+WOFD 0.891 0.020 * BRCD+WOFD 0.890 0.021 * 700m a.s.l #sps. 16 BRGE+WOFD 0.890 0.023 * IndVal p.value GOWH+WOFD 0.880 0.023 * EYRO+PRIF 1.000 0.002 ** GRFA+RCFD 0.880 0.026 * ESPI+PRIF 0.894 0.004 ** PCUR+PRIF 0.894 0.004 ** 500-900m a.s.l #sps. 1 CROS+ESPI 0.866 0.010 ** IndVal p.value CROS+EYRO 0.866 0.010 ** GRCA+YTSW 0.868 0.05 * CROS+PCUR 0.866 0.010 ** ESPI+WHPI 0.866 0.016 * 700-1100m a.s.l #sps. 1 PRIF+WHPI 0.866 0.016 * Indval p.value EYRO+YTSW 0.953 0.002 ** 300-500m a.s.l #sps. 1 Indval p.value AULO+GOWH 0.866 0.007 **

Autumn 500-700m a.s.l #sps. 10 Total number of species: 1081 Indval p.value PRIF 0.935 0.001 *** List of indicators associated with AULO+PRIF 0.935 0.001 *** each elevation or elevational BRGE+PRIF 0.935 0.001 *** range: BRTB+PRIF 0.935 0.001 *** LEHE+PRIF 0.935 0.001 *** 300m a.s.l #sps. 9 PRIF+WTRE 0.935 0.001 *** IndVal p.value PRIF+YTSW 0.935 0.001 *** BRCD+GRFA 0.866 0.026 * EWHI+PRIF 0.866 0.005 ** BRCD+PCUR 0.866 0.021 * MSTB+PRIF 0.866 0.006 ** BRCD+SCHE 0.866 0.025 * PRIF+WBSW 0.866 0.008 ** BRCD+SPPA 0.866 0.022 * BRCD+WTRE 0.866 0.021 * GRFA+PCUR 0.866 0.026 * 700-900m a.s.l #sps. 5 PCUR+SCHE 0.866 0.025 * IndVal p.value PCUR+SPPA 0.866 0.022 * BRTB+ESPI 0.935 0.001 *** PCUR+WHPI 0.866 0.026 * ESPI+MSTB 0.935 0.001 *** ESPI+EYRO 0.920 0.001 *** 900m a.s.l #sps. 1 ESPI+LEHE 0.889 0.009 **

AULO+ESPI 0.882 0.004 **

300-700m a.s.l #sps. 1 IndVal p.value AULO+WTRE 0.913 0.021 *

700-1100m a.s.l #sps. 1 IndVal p.value ESPI+YTSW 0.889 0.017 *

Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Appendix 4.5: Species Names The taxonomic order and nomenclature in this paper follows IOC World Bird Names, version 8.1. Some readers may be more familiar with the taxonomy of Australian birds proposed by Christidis and Boles (2008). Since that time there have been a number of taxonomic revisions published in the literature. Please see http://goo.gl/QqvwaU for a list of all bird species found in Australia based on the 2018 IOC checklist. The following table provides a comparison between the taxonomy and nomenclature used in Christidis and Boles (2008), and the taxonomy and nomenclature used in version 8.1 of the IOC World Bird Names.

The common names as given in Christidis and Boles (2008) are listed in alphabetical order.

Christidis and Boles Taxonomy of Australian Birds (2008) IOC World Bird Names version 8.1 (2018) Common Name Scientific Name Common Name Scientific Name Albert's Lyrebird Menura alberti Albert's Lyrebird Menura alberti Australian Brush-turkey Alectura lathami Australian Brushturkey Alectura lathami Australian King-Parrot Alisterus scapularis Australian King Parrot Alisterus scapularis Australian Logrunner Orthonyx temminckii Australian Logrunner Orthonyx temminckii Bassian Thrush Zoothera lunulata Bassian Thrush Zoothera lunulata Black-faced Cuckoo-shrike Coracina novaehollandiae Black-faced Cuckooshrike Coracina novaehollandiae Black-faced Monarch Monarcha melanopsis Black-faced Monarch Monarcha melanopsis Brown Cuckoo-Dove Macropygia amboinensis Brown Cuckoo-Dove Macropygia phasianella Brown Gerygone Gerygone mouki Brown Gerygone Gerygone mouki Brown Thornbill Acanthiza pusilla Brown Thornbill Acanthiza pusilla Brush Cuckoo Cacomantis variolosus Brush Cuckoo Cacomantis variolosus Crested Shrike-tit Falcunculus frontatus Crested Shriketit Falcunculus frontatus Crimson Rosella Platycercus elegans Crimson Rosella Platycercus elegans

Acanthorhynchus Eastern Spinebill Acanthorhynchus tenuirostris Eastern Spinebill tenuirostris Eastern Whipbird Psophodes olivaceous Eastern Whipbird Psophodes olivaceous Eastern Yellow Robin Eopsaltria australis Eastern Yellow Robin Eopsaltria australis Emerald Dove Chalcophaps indica Common Emerald Dove Chalcophaps indica Fan-tailed Cuckoo Cacomantis flabelliformis Fan-tailed Cuckoo Cacomantis flabelliformis Golden Whistler Pachycephala pectoralis Australian Golden Whistler Pachycephala pectoralis Green Catbird Ailuroedus crassirostris Green Catbird Ailuroedus crassirostris Grey Fantail Rhipidura albiscapa Grey Fantail Rhipidura albiscapa Grey Shrike-thrush Colluricincla harmonica Grey Shrikethrush Colluricincla harmonica Large-billed Scrubwren Sericornis magnirostra Large-billed Scrubwren Sericornis magnirostra Laughing Kookaburra Dacelo novaeguineae Laughing Kookaburra Dacelo novaeguineae Lewin's Honeyeater Meliphaga lewinii Lewin's Honeyeater Meliphaga lewinii Little Shrike-thrush Colluricincla megarhyncha Little Shrikethrush Colluricincla megarhyncha Mistletoebird Dicaeum hirundinaceum Mistletoebird Dicaeum hirundinaceum Noisy Pitta Pitta versicolor Noisy Pitta Pitta versicolor Olive-backed Oriole Oriolus sagittatus Olive-backed Oriole Oriolus sagittatus Pale-yellow Robin Tregellasia capito Pale-yellow Robin Tregellasia capito Paradise Riflebird Ptiloris paradiseus Paradise Riflebird Ptiloris paradiseus Pied Currawong Strepera graculina Pied Currawong Strepera graculina Rainbow Lorikeet Trichoglossus haematodus Rainbow Lorikeet Trichoglossus moluccanus Regent Bowerbird Sericulus chrysocephalus Regent Bowerbird Sericulus chrysocephalus Rose Robin Petroica rosea Rose Robin Petroica rosea Rose-crowned Fruit-Dove Ptilinopus regina Rose-crowned Fruit Dove Ptilinopus regina Rufous Fantail Rhipidura rufifrons Rufous Fantail Rhipidura rufifrons Russet-tailed Thrush Zoothera heinei Russet-tailed Thrush Zoothera heinei Satin Bowerbird Ptilonorhynchus violaceus Satin Bowerbird Ptilonorhynchus violaceus Scarlet Honeyeater Myzomela sanguinolenta Scarlet Myzomela Myzomela sanguinolenta Shining Bronze-Cuckoo Chrysococcyx lucidus Shining Bronze Cuckoo Chrysococcyx lucidus

Silvereye Zosterops lateralis Silvereye Zosterops lateralis Spangled Drongo Dicrurus bracteatus Spangled Drongo Dicrurus bracteatus Spectacled Monarch Symposiachrus trivirgatus Spectacled Monarch Symposiachrus trivirgatus Spotted Pardalote Pardalotus punctatus Spotted Pardalote Pardalotus punctatus Striated Pardalote Pardalotus striatus Striated Pardalote Pardalotus striatus Striated Thornbill Acanthiza lineata Striated Thornbill Acanthiza lineata Sulphur-crested Cockatoo Cacatua galerita Sulphur-crested Cockatoo Cacatua galerita Superb Fruit-Dove Ptilinopus superbus Superb Fruit Dove Ptilinopus superbus Topknot Pigeon Lopholamus antarcticus Topknot Pigeon Lopholamus antarcticus Varied Triller Lalage leucomela Varied Triller Lalage leucomela White-browed Scrubwren Sericornis frontalis White-browed Scrubwren Sericornis frontalis White-headed Pigeon Columba leucomela White-headed Pigeon Columba leucomela White-throated Treecreeper Cormobates leucophaea White-throated Treecreeper Cormobates leucophaea Wompoo Fruit-Dove Ptilinopus magnificus Wompoo Fruit Dove Ptilinopus magnificus Wonga Pigeon Leucosarcia picata Wonga Pigeon Leucosarcia melanoleuca Yellow-faced Honeyeater Lichenostomus chrysops Yellow-faced Honeyeater Caligavis chrysops Yellow-tailed Black-Cockatoo Calyptorhynchus funereus Yellow-tailed Black Cockatoo Calyptorhynchus funereus Yellow-throated Scrubwren Sericornis citreogularis Yellow-throated Scrubwren Sericornis citreogularis

Chapter 5

In this chapter we examine the drivers of avian species richness and abundance along elevational gradients in subtropical rainforest. We also explore the influence of species’ functional traits on their responses to changing environmental conditions. The study was conducted in the northern regions of the Gondwana Rainforests of Australia World Heritage Area.

Statement of contribution to a co-authored paper

Chapter 5 is a co-authored paper that has been published by the journal Austral Ecology. Permission to reproduce this article has been granted by the publisher (Wiley-Blackwell, online ISSN: 1442-9933). The citation is as follows:

Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018. Modelling the responses of Australian subtropical rainforest birds to changes in environmental conditions along elevational gradients. Austral Ecology. https://doi.org/10.1111/aec.12586

(Signed) ______(9/7/18) (Countersigned) ______(9/7/18)

Elliot Leach Supervisor: Darryl Jones

(Countersigned) ______(9/7/18) (Countersigned) ______(9/7/18)

Associate Supervisor: Chris Burwell Supervisor: Roger Kitching

Modelling the responses of Australian subtropical rainforest birds to changes in environmental conditions along elevational gradients

* Corresponding author, email address: [email protected], telephone: +61 434 633 218 (E. Leach)

Author ORCIDs: Elliot Leach 0000-0003-2108-4402, Chris Burwell NA, Darryl Jones 0000-0002- 2565-9456, Roger Kitching 0000-0002-6798-6041

Keywords: climate change; altitude; fourth-corner problem; traits; bird assemblage; generalised linear mixed model; GLMM

5.1 Abstract

Montane birds face significant threats from a warming climate, so determining the environmental factors that most strongly influence the composition of such assemblages is of critical conservation importance. Changes in temperature and other environmental conditions along elevational gradients are known to strongly influence the species richness and abundance of bird assemblages occupying mountains. However, the role of species-specific traits in mediating the responses of bird species to changing conditions remains poorly understood. We aimed to determine whether different bird species responded differently to changing environmental conditions in a relatively understudied biodiversity hotspot in subtropical rainforest on the east coast of Australia. We examined patterns in avian species richness and abundance along two rainforest elevational gradients using monthly point counts between September 2015 and October 2016. Environmental data on temperature, wetness, canopy cover and canopy height was collected simultaneously, and trait information on body size and feeding guild membership for each bird species was obtained from the Handbook of Australian, New Zealand and Antarctic Birds. We used a generalised linear mixed modelling (GLMM) framework to determine the drivers of species richness and abundance and to quantify species’ trait-environment interactions. GLMMs indicated that temperature alone was significantly positively correlated with species richness and abundance. Species richness declined with increasing elevation. When modelling abundance, we found that feeding guild membership did not significantly affect species’ responses to environmental conditions. In

73 contrast, the predicted abundance of a species was found to depend on its body size, due to significant positive interactions between this trait, temperature and canopy cover. Our findings indicate that large-bodied birds are likely to increase in abundance more rapidly than small-bodied birds with continued climatic warming. These results underline the importance of temperature as a driving factor of avian community assembly along environmental gradients.

5.2 Introduction

The subtropical rainforests of the east coast of Australia support some of the country’s most biodiverse avian communities (Metcalfe, 2010). The general restriction of these communities to mountainous regions (primarily along the Great Dividing Range) means they are seriously threatened by increasing global temperatures (Laurance et al., 2011), despite their formal and informal protection (Kitching et al., 2010, Birdlife Australia, 2017). Increasing temperatures are likely to cause upwards shifts in the climatic optima of species within mountain ecosystems (Foster, 2001). This will lead to the alteration of ecological communities as new species invade previously unsuitable elevations (Pauchard et al., 2016) and ultimately to the likely extirpation of mountaintop specialists (e.g. Hilbert et al., 2004). Globally, avian diversity is concentrated in mountainous regions, particularly near the equator, and faces similar threats (Orme et al., 2005, Urban, 2015). Climatic conditions, particularly the combination of temperature and water availability, strongly influence species diversity along elevational gradients (Fjeldså et al., 2012, Dossa et al., 2013). Climate influences birds both directly (through physiological constraints) and indirectly (by determining vegetative heterogeneity and food availability) (Dossa et al., 2013, Sundqvist et al., 2013, Ferger et al., 2014). Along humid mountain ranges there is a general decline in avian species richness with elevation (McCain, 2009), although the mechanisms responsible for this pattern are still under discussion (Herzog et al., 2005, Acharya et al., 2011). ‘Hump-shaped’ patterns with peaks in diversity at mid-elevations are also found, particularly along more extensive gradients (Rahbek, 1995, Williams et al. 2010, Ferger et al., 2014, Colwell et al., 2016). As continuing climatic changes are predicted to lead to shifts in species’ distributions, understanding the mechanisms responsible for the determination of diversity and community structure has become a central challenge in community ecology, and is of critical conservation importance (Acharya et al., 2011, Freeman and Class Freeman, 2014, Howard et al., 2015).

The ways in which bird species respond to changes in environmental conditions may be mediated by functional traits, such as feeding guild membership or body size (Ferger et al., 2014, Howard et al., 2015). For example, Gray et al. (2007) reported that avian feeding guilds responded differentially to environmental disturbance, and Howard et al. (2015) found that there was a negative correlation between body mass and sensitivity to climate in European birds. Accounting for trait-based variation in species responses to climatic conditions is therefore an important part of forecasting community-level changes in biodiversity (McGill et al., 2006, Katuwal et al., 2016). Associating species traits and environmental variables using species abundance data is a challenge in multivariate ecological analysis known as the fourth corner problem. Briefly, the fourth-corner problem relates to the study of trait-environment associations using three matrices: (1) abundance data across species; (2) environmental data across sites; and (3) trait data across species (Brown et al., 2014). The matrix created by the trait-environment interaction coefficients is the missing ‘fourth corner’. For more detailed summaries of the fourth-corner problem, including useful figures, we refer the reader to Legendre et al. (1997) and Brown et al. (2014).

Traditional methods of solving the fourth-corner problem, such as exploratory ordination or algorithmic approaches, may be difficult to interpret and do not explicitly account for the statistical properties of the data (Brown et al., 2014, Warton et al., 2015). Recent developments in statistical methodology and computing power have led to improvements in the analysis of multivariate abundance data and the development of model-based approaches (Warton et al., 2015, Hui, 2016). A model-based approach has several advantages over traditional methods: it accounts for the uncertainty inherent in ecological data such as counts of species, the models have clear assumptions that can be checked, model selection procedures are available and models allow for formal inferences to be made (Jamil et al., 2013, Warton et al., 2015). Recently, a number of researchers independently proposed model- based approaches to solving the fourth-corner problem based on generalised linear mixed modelling (GLMM) (Pollock et al., 2012, Jamil et al., 2013) and generalised linear modelling (GLM) (Brown et al., 2014) frameworks. Predicting species abundance as a function of environmental variables, species traits, and their associated interactions, as suggested by these researchers, links multivariate ecological analysis with species distribution modelling (Elith and Leathwick, 2009, Pollock et al., 2012). This approach allows for important ecological relationships to be quantified directly, and for predictions to be made under current and future conditions (Warton et al., 2015).

Forecasting the potential impacts of global climatic change on rainforest birds requires an understanding of how their community structure is influenced by environmental conditions, and how species traits mediate their responses to these conditions (Elith and Leathwick, 2009, Jankowski et al., 2013). Elevational gradients provide researchers interested in such questions with ideal study systems due to the predictable changes in environmental conditions across small spatial scales that occur along these gradients (Able and Noon, 1976). Research on changes in avian diversity, abundance and community composition with increasing elevation has a rich history beginning with the seminal works of Diamond (1973), Able and Noon (1976) and Terborgh (1977). Few researchers, however, have examined trait- environment relationships and how they relate to avian abundance. To address these questions, we established two elevational gradients in subtropical rainforest in north-eastern New South Wales, Australia and conducted point counts of the birds along these gradients on a monthly basis for one year. Here, we use a GLMM-based approach to determine (1) which environmental variables are associated with spatial changes in avian species richness and abundance; (2) whether species-specific traits mediate responses to environmental conditions; and (3) which members of the avian community are most sensitive to changes in temperature.

5.3 Methods

Study area and sites

All fieldwork was undertaken in two national parks (Border Ranges NP (28°23′S, 153°3′E) and Mebbin NP (28°27′S, 153°10′E)) in north-eastern New South Wales, Australia. At low elevations, the average annual rainfall in the study region is 1555mm, the average maximum summer temperature is 29°C, and the average minimum winter temperature is 6°C (Birdlife International, 2017). The study area and sites are illustrated in Fig. 5.1; see also Appendix 5.1 Table S5.1a. We established two elevational gradients in subtropical rainforest encompassing the full forested elevational range within the region. Each gradient consisted of two replicate sites, separated by a minimum of 400m, in each of five elevational bands ~300, 500, 700, 900 and 1100m above sea level (asl).

Bird surveys

In each month between September 2015 and October 2016, EL conducted a 10 minute audio- visual survey of the avifauna at each site. The abundance of all species seen or heard within a 100m radius of the central point was recorded, with each individual recorded only once during the survey period. Counts were undertaken in the three hours following dawn, and all sites were typically surveyed over a four-day period in fine weather conditions. Due to inclement weather and road closures in the national parks, we were unable to sample in September 2016. Sampling proceeded without interruption in the other months, and we obtained 11 monthly datasets suitable for analysis. Sampling effort was equal at each site.

Species recorded less than five times (a total of 21 species) were excluded from analyses. The majority of these species were typically inhabitants of wet eucalypt forests, as opposed to subtropical rainforests, and their presence at our study sites was considered accidental. Birds recorded flying over the sites, but not landing, were also excluded. Following these exclusions, the abundances of 42 bird species typically found in Australian subtropical rainforests (Appendix 5.2) were used for analysis.

Environmental data

Before starting each point count, EL recorded data on the following variables at each site on each sampling occasion: temperature in °C (measured at ground level using a digital thermometer); wetness (in four classes: 0 – leaf litter dry, 1 – leaf litter damp, 2 – leaf litter wet, 3 – leaf litter and undergrowth dripping wet); canopy cover (estimated in % from a new position within 20m of the central point on each sampling occasion); and canopy height (measured using a laser range finder from a new position within 20m of the central point on each sampling occasion). These are ecologically meaningful predictors sensu Elith and Leathwick (2009). The average canopy height and average canopy cover were calculated for each site and these values were used for analysis.

Trait data

As species traits may mediate species’ responses to changes in environmental conditions (McGill et al., 2006; Katuwal et al., 2016), we obtained information from the Handbook of Australian, New Zealand and Antarctic Birds (HANZAB) (Marchant and Higgins, 1990; Marchant and Higgins, 1993; Higgins and Davies, 1996; Higgins, 1999; Higgins et al., 2001; Higgins and Peter, 2002; Higgins et al., 2006) on feeding guild and body size for each bird species. We designated species as carnivorous, frugivorous, granivorous, insectivorous, nectarivorous or omnivorous based on their primary food source. Body size was based on the total weight in grams given for each species in HANZAB, and was log-transformed before analysis to improve normality and homogeneity. For a list of all species used in the analysis, and their associated traits (feeding guild and body size), see Appendix S5.2.

Data analysis

To evaluate sampling sufficiency, we calculated sample-based species accumulation and sample completeness curves using the package iNEXT version 2.0. 12 (Hsieh et al., 2016) in R software version 3.4.1 (R Core Team, 2017). Within each elevational band we graphed the

78 proportions of individuals within each feeding guild, both across the whole sampling period and in summer and winter, defined as December to February and June to August respectively. Chi-squared tests were used to quantify changes in trophic community structure between seasons. Graphs of species richness and overall abundance in each elevational band were also generated using the packages ggplot2 version 2.2.1 (Wickham, 2016) and ggpubr version 0.1.4 (Kassambara, 2017).

Modelling species richness and abundance

First, the relationships between the environmental variables, species richness and overall abundance were visually inspected following graphing to examine the community- environment relationships (Murtaugh, 2007, Warton, 2008). We also graphed pairs of environmental variables to check for any collinearity among predictors (Zuur et al., 2010). As we analysed count data obtained through repeated measures at sites, the data were not normally distributed. Rather than attempting to transform the data to normality, we chose to use a statistical approach that matched our data (Bolker et al., 2009). To assess the influence of the environmental variables on counts of species richness and abundance, we fitted GLMMs with a Poisson error distribution and a logarithmic link function using function glmer from package lme4 version 1.1-13 (Bates et al., 2015). Possible autocorrelation between sites was controlled for by allowing ‘Site’ to vary randomly through the inclusion of a random intercept effect in the model. In this way, each site was allowed a different intercept; we assumed that most correlation occurred at the within-site scale as opposed to the between-site scale (Parsons et al., 2016). We also allowed temperature to vary randomly by site by including a random slope effect. To select a minimal model for each response variable (species richness and overall abundance), we used backward selection, starting with a saturated model that included all predictor variables, a random intercept effect for sites and a random slope effect for temperature at sites. We then removed the random slope effect and compared the two models using likelihood ratio testing. For the species richness model, the inclusion of the random slope effect did not significantly improve the fit and we continued the model selection procedure by dropping non-significant predictor variables one at a time. Each new model was compared with the previous one by examining the AIC and BIC of the models and by likelihood ratio testing. For the species abundance model, the inclusion of a random slope effect significantly improved the fit of the model. The model selection procedure then followed the same steps applied to the species richness model. Once the final models were determined, we examined diagnostic plots to confirm that the assumptions of

79 homogeneity, normality and independence were not violated (Steel et al., 2013). The data and R code for these analyses are given in Appendix 5.3.

Abundance and trait-environment relationships

We followed the GLMM approach of Jamil et al. (2013) by fitting models to the data that predicted the abundance of species as a function of the environmental variables and trait data described above, as well as their interactions. To account for repeated measures at sites, as well as possible autocorrelation, we included a random intercept effect for Site in the model. As we were interested in how traits mediated the response of individual species to environmental conditions, we also included a random intercept term for ‘Species’. In this way, different species were allowed to have different total abundances via species-specific intercept terms (Brown et al., 2014). To account for apparent overdispersion, we included an observation level random effect (Harrison, 2014). Before fitting the GLMMs, we log- transformed ‘Body Size’, reclassified ‘Wetness’ into a binary variable (wet (classes 1-3) and dry (class 0)), and scaled the quantitative predictor variables (‘Temperature’, ‘Canopy’ and ‘Canopy Height’) to improve numerical stability and assist with the interpretation of model coefficients. We utilised the tiered forward selection approach of Jamil et al. (2012) for model selection, using code that we modified from that provided by Jamil et al. (2013).

Briefly, the tiered model selection procedure was as follows. In the first tier, the environmental variables were added to a null model with only random effects. The terms were added in the order which most decreased the information from the Species random effect term, as determined by the SigAIC. SigAIC is defined as minus two times log- likelihood plus 3.84 times degrees freedom and only reduces when a single parameter is significant at the 0.05 significance level (as determined by a likelihood ratio test) (Jamil et al., 2012). The process of adding environmental terms continued until the SigAIC no longer decreased. In the second tier, the main effects of the environmental variables selected in the first tier were added to the model. The included environmental variables were thus components of both fixed and random effects. Next, the trait-environment terms that most decreased the SigAIC were added to the model in a stepwise fashion. If the main effect for trait was not yet included, it was added jointly and then the SigAIC was evaluated. This process continued until the SigAIC no longer decreased. In the final tier, the model was evaluated and any non-significant trait-environment terms were removed. To determine the influence of traits on species responses to environmental variables, we compared the variance

80 estimates of the random slopes with respect to the environmental variables in models with and without trait-environment interactions as suggested by Jamil et al. (2013). We then used QQ plots to check the normality of the random effects, and used the simulateResiduals and plotSimulatedResiduals functions from the DHARMa package version 0.1.5. (Hartig, 2017) to check that the final model met the assumption of normality and that overdispersion had been accounted for (Steel et al., 2013). The data and R code for these analyses is given in Appendix 5.4.

5.4 Results

In total, we recorded 3,116 individuals from 63 species representing 29 families and analysed data from 2,963 individuals of 42 species that represented 21 families (Appendix 5.2). Observed sample completeness curves showed that over 87% sample coverage was achieved at all sites, indicating that our samples accurately represented the avian community (Table 5.1; Appendix 5.5 Figs. S5.5a – S5.5e). The most abundant and frequently detected species was Lewin’s Honeyeater Meliphaga lewinii (355 birds; 348 detections). Proportionally, insectivores were the dominant feeding guild; approximately 60% of species sampled were insectivorous. Omnivores represented 25% of species and frugivores (8%), granivores (3%), carnivores (3%) and nectarivores (1%) were represented by relatively few species. The trophic structure of the community changed with elevation. Insectivore abundance increased proportionally with elevation, as did that of granivores and carnivores, whereas frugivore abundance decreased with elevation and nectarivores were only present at 300m and 500m sites (Fig. 5.2a). Community composition (in terms of feeding guild representation) changed significantly on a seasonal basis (Fig. 5.2b & c; Appendix 5.1 Table S5.1b). In winter, insectivores significantly increased proportionally at 300m, 500m and 1100m sites (2 tests: p = 0.048, <0.001, 0.01 respectively), whereas omnivores declined at 1100m (p = <0.001), as did frugivores at 500m and 900m (p = <0.01). Nectarivores, although present in low numbers at 500m in summer, were completely absent from that elevation in winter. The abundance of all guilds declined at high elevations in winter. Overall species richness (Fig. 5.3a) and median temperature (Fig. 5.3b) generally declined with elevation. There was no consistent trend in overall abundance along the gradient (Table 5.1).

Table 5.1 Observed sample coverage (SC), species richness and total abundance at each site

Site Observed SC Richness Abundance 300BMA 0.89 32 155 300BMB 0.94 26 142 300SCA 0.94 32 151 300SCB 0.93 27 116 500BMA 0.92 31 187 500BMB 0.96 29 169 500SCA 0.93 34 167 500SCB 0.93 29 132 700BMA 0.92 31 129 700BMB 0.94 30 163 700SCA 0.93 30 155 700SCB 0.96 31 166 900BMA 0.98 28 166 900BMB 0.93 29 146 900SCA 0.93 30 145 900SCB 0.93 30 144 1100BMA 0.94 28 142 1100BMB 0.95 25 144 1100SCA 0.87 26 113 1100SCB 0.94 27 131

Species richness and abundance

Backwards selection using an information-theoretic framework revealed Temperature as the only environmental variable that significantly influenced species richness and abundance along the elevational gradient. The optimal model for species richness included a random effect for Site and a fixed effect of Temperature (Table 5.2). The optimal model for abundance included a random slope effect that allowed Temperature to vary within Site (with correlated intercepts) and a fixed effect of Temperature (Table 5.2). Species richness and abundance increased with increasing temperature. Species richness increased by 1.03 with a one degree Celsius increase in temperature, and abundance increased by 1.027 individuals with a one degree Celsius increase in temperature. Diagnostic plots for both models validated the assumptions of homogeneity, normality and independence (Appendix 5.3).

Abundance and trait-environment relationships

We used the tiered forward selection procedure to fit a parsimonious regression model that included all significant environmental and trait variables, as well as significant trait- environment interaction terms. In the first tier of the forward-selection process, the environmental variables Canopy, Temperature and Canopy Height were selected as species- dependent random effects. Random effects for Site and Observation were also included. Wetness did not influence abundance, and did not significantly reduce SigAIC. In the second tier, the fixed effects of Canopy, Temperature and Canopy Height were added to the model.

Figure 5.3 Overall bird species richness (a) and median temperature (b) in each elevational band.

The next phase of the second tier then evaluated which trait-environment relationships significantly reduced SigAIC. The relationship between Body Size and Temperature and the relationship between Body Size and Canopy both reduced the SigAIC. These interactions, as well as a fixed effect for Body Size, were added to the model. Feeding Guild did not significantly interact with any environmental variables, and no other significant interactions between Body Size and environmental variables were found, so no interactions were deleted in the third tier. In comparison with the null model (which included only random terms), the final model reduced SigAIC by 220.68. The explained deviance of the final model (calculated by removing random effects; see Parsons et al. (2016)) was 82.4%, indicating a relatively good fit.

Response Intercept Temperature Δ AIC Δ BIC

Estimate p value Estimate p value Richness 1.695 < 0.001 0.034 < 0.001 6.9 23.8 Abundance 2.23 < 0.001 0.027 < 0.001 301.6 325.4

Body Size was positively related to Temperature (p < 0.001) and Canopy (p = 0.011) (Table 5.3), indicating that large-bodied species are more likely to increase in abundance with higher temperatures and increased canopy cover than small-bodied species. Examples of the interaction between Temperature and Body Size with respect to abundance are given in Fig. 5.4. Body Size explained 44.9% and 16.6% of the variance in species responses to Temperature and Canopy respectively.

Parameter Estimate z value p value (>|z|)

Intercept -0.811 -1.872 0.061 Canopy -0.406 -2.430 0.015 Temperature -0.135 -1.363 0.172 Canopy Height 0.061 1.254 0.209 Body Size -0.25 -2.287 0.022 Temperature : Body Size 0.096 3.619 < 0.001 Canopy : Body Size 0.106 2.525 0.011

5.5 Discussion

Species richness and abundance

We aimed to determine which biotic and abiotic factors most strongly influenced the community structure of rainforest birds along an elevational gradient. The results of our GLMMs indicated that temperature was the most important abiotic factor measured, and was positively correlated with both species richness and abundance. Species richness declined with elevation, supporting the results of McCain (2009) for humid mountains, but contrasting with the results of Williams et al. (2010), who described a hump-shaped pattern in species richness in rainforests of the Wet Tropics bioregion in northern Queensland. The trophic

86 structure of the community at different elevations changed on a seasonal basis, with species in most feeding guilds appearing to engage in elevational migration. We suggest that species in this study may not have responded directly to temperature changes from summer to winter, but rather to numerous unmeasured correlates of temperature, such as primary productivity, food availability or habitat geometry (Loiselle and Blake, 1991, Williams et al. 2010, Ferger et al., 2014, Martin and Proulx, 2016). This conclusion is supported by the results of Londoño et al. (2017), who found that the elevational distributions of adult birds in the Peruvian Andes were not constrained by the costs of thermoregulation. Dingle (2004) suggested that elevational migrations were largely related to birds tracking shifts in feeding resources (Joshi et al., 2012). Information on food resources available to birds in the region is scarce, but Lambkin et al. (2011) found that the abundance of higher Diptera (flies) at mid and high elevations in nearby Lamington National Park (LNP) declined in winter. In our study, the abundance of insectivorous birds at mid to high elevations also declined in winter, suggesting a possible link between food availability in winter and insectivore abundance. Laidlaw et al. (2011) described a general decline in floristic diversity with elevation in LNP, but the phenology of plants in the region (i.e. when and why they flower and produce fruit) is still poorly understood (Williams and Adam, 2010). Waterhouse et al. (2002) also indicated that changes in the interactions between temperature, forest stand structure and forage productivity with elevation affect the distributions of bird species. Describing the influence of the floral community on trends in frugivore, granivore and nectarivore richness and abundance requires further study.

As it is impossible to measure every factor that could influence a community, or single species, (Brehm et al., 2003), we chose ecologically meaningful predictors to include in our models (Elith and Leathwick, 2009). We used a measure of ‘wetness’ as a proxy for water availability but found no significant relationship between this variable and species richness or abundance, despite the results of a meta-analysis (McCain, 2009) suggesting that a combination of temperature and water availability was the main driver of avian diversity along elevational gradients. This result has two possible explanations: either water availability was not a limiting factor for birds in the study area, or our measure of ‘wetness’ did not sufficiently approximate the true water availability. Ives and Helmus (2011) advocated the inclusion of phylogenetic information in analysis of community structure, but Li and Ives (2017) recently demonstrated that such an approach may lead to inflated Type I error rates with small (<60) numbers of species. Additional environmental predictors that

87 may be considered in future studies include simultaneous measures of food (e.g. fruits, flowers, invertebrates) richness and abundance, the inclusion of phylogenetic information in the model (Li and Ives, 2017), and information on habitat geometry (Martin and Proulx, 2016). However, the inclusion of too many predictors can easily lead to overfitting problems (Babyak, 2004).

The results of our GLMMs suggest that subtropical rainforest birds in Australia may increase in abundance with a warming climate, and that the species richness in the study area may also increase. However, this statement does not consider the effects of increasing temperature on the ecosystem as a whole. As just one example, the species restricted to higher elevations will likely face population declines due to increasing unsuitability of their habitat and increased competition from lowland species (Hilbert et al., 2004, Jankowski et al., 2010). Additionally, species with different functional traits may respond in different ways to changes in environmental conditions.

Abundance and trait-environment relationships

To determine whether traits played a role in mediating species’ responses to environmental conditions along the elevational gradient, we attempted to quantify species’ trait-environment relationships. Surprisingly, we found no relationship between feeding guild membership and environmental variables, despite detecting seasonal shifts in the trophic community that were presumably related to temperature. Previous researchers have linked fruit availability to frugivore diversity and abundance (Loiselle and Blake, 1991, Ferger et al., 2014), and habitat complexity and season to insectivore diversity (Shiu and Lee, 2003, Jankowski et al., 2013, Ferger et al., 2014). Our results may have been affected by the low numbers of species assigned to guilds other than insectivores and omnivores, as well as the patchy elevational distribution of some guilds. Compared with the insectivores, frugivores in our data were quite large in size (being mainly columbids). Therefore, it is also possible that the effect of feeding guild on abundance was ‘captured’ within the effect of body size in our model.

Body size had a significant positive relationship with temperature and canopy cover, and explained a large amount of variation in species responses to temperature. The abundance of large-bodied species increased more rapidly with temperature and increasing canopy cover than that of small-bodied species. Large-bodied species are able to thermoregulate more efficiently due to their smaller surface-area to volume ratio and higher fat reserves (Ding et al., 2005). As a consequence, they have more ‘spare’ energy to invest in foraging or

88 reproduction than do small-bodied species, and may be better placed to take advantage of greater resource availability at warmer sites. Janes (1994) found that the body size of insectivorous birds in North America declined with elevation, and demonstrated that this relationship was correlated with a similar decline in arthropod body size, but similar studies in Australia are yet to be undertaken. Howard et al. (2015) found that abundances of small- bodied birds in Europe were more strongly affected by climatic conditions than large-bodied birds. Increases in temperature may, therefore, have disproportionately adverse effects on the populations of small-bodied species across varied habitats, although Anderson et al. (2013) showed that temperature sensitivity in Australian rainforest birds was not limited to small- bodied species. Sitters et al. (2016) found that frugivores and omnivores in southern Australia responded positively to increasing canopy cover and suggested that this interaction was due to food availability. A similar explanation may be drawn from our results, in which large- bodied species (mainly frugivores and omnivores) responded more strongly to increases in canopy cover than did small-bodied species (mainly insectivores). If this trend is driven by increasing food supply, then interspecific competition may influence species’ responses to changes in environmental conditions if large birds competitively exclude small birds. However, assessing rates of interspecific competition goes beyond the scope of our study.

Including the trait information in the GLMM significantly improved the fit of the final model and allowed us to quantify ecologically meaningful interactions. We used a limited number of traits and environmental variables to avoid overfitting. With a larger dataset, ideally one collected over several years, the inclusion of additional trait parameters (e.g. foraging strategies), or environmental variables (e.g. food availability) as predictors may reveal more subtle patterns.

Conclusions

Our results underline the importance of temperature in structuring the avian community of Australia’s subtropical rainforests. Substantial increases in annual temperature are expected on the east coast of Australia, with increases in annually averaged temperature projected to be within the range of 2.8 to 5°C by 2090 under a high-emissions scenario (CSIRO and Bureau of Meteorology, 2016). Strong et al. (2011) measured the adiabatic lapse rate in the study region and found that temperature declined by roughly 1.5°C per 200m increase in elevation. Therefore, the projected temperature increase in the study area by 2090 corresponds to a roughly 400m to 650m upwards shift in current climatic conditions along the elevational

89 gradient. Species such as the Bassian Thrush (Zoothera lunulata), which are currently restricted to the mountaintops in the study area, are likely to become locally extinct in the face of such dramatic change (Leach, 2016). Adaptive southerly movements by other species will be constrained by the fact that the forested mountains of the Great Dividing Range are ‘islands’ of suitable habitat in an ‘ocean’ of unsuitable lowland habitat, most of which was historically cleared for agriculture (Becker et al., 2007, Kitching et al., 2010). The prognosis for small-bodied rainforest specialists on the east coast of Australia in a warming climate is grim. Our results demonstrate the value of including trait information in models seeking to explain species responses to environmental conditions. Further validation of these relationships requires detailed field observations of many species over multiple years (Sitters et al., 2016), and for similar studies to be conducted along other mountain ranges. Nevertheless, our analyses add to a growing body of literature stressing the importance of temperature in determining the diversity, abundance and distribution of birds along elevational gradients.

5.6 Acknowledgements

We thank two anonymous reviewers, whose suggestions improved an earlier version of this manuscript. EL thanks James McBroom, Sarah Maunsell, Rachakonda Sreekar and Rob Appleby for helpful discussions regarding analysis. Staff of the Kyogle National Parks and Wildlife Office provided EL with assistance during field work, as did numerous volunteers – special thanks go to Jess Mackie. Adam Higgins and John Gunning kindly provided the photographs of the Golden Whistler and Noisy Pitta in Fig. 5.4. All fieldwork was undertaken under Scientific Licence number SL101514 issued by the New South Wales National Parks and Wildlife Service. EL was supported by an Australian Postgraduate Award during this research.

5.7 Appendices

The following Appendices are available online as Supporting Information for this paper. They are presented here to aid the reader’s understanding of this chapter.

Appendix 5.1

This Appendix contains two tables, Table S5.1a and Table S5.1b.

Table S5.1a The GPS coordinates of the sites used in this study. These sites form two elevational gradients in subtropical rainforest in northern New South Wales, Australia: the Bar Mountain (BM) gradient, and the Sheepstation Creek (SSC) gradient. Each gradient consisted of two replicate sites, separated by a minimum of 400m, in each of five elevational bands: ~300, 500, 700, 900 and 1100m asl. All sites were located within a protected area of approximately 15, 940ha, made up by two National Parks (Border Ranges NP and Mebbin NP). For a view of the NP and the surrounding landscape, see https://goo.gl/RpCn95.

Site Latitude Longitude 300BMA -28.4838 153.1721 300BMB -28.4853 153.1671 500BMA -28.4937 153.132 500BMB -28.4925 153.1359 700BMA -28.4879 153.1321 700BMB -28.4902 153.129 900BMA -28.4816 153.1433 900BMB -28.4758 153.1464 1100BMA -28.4598 153.1302 1100BMB -28.4608 153.1245 300SCA -28.4078 153.0219 300SCB -28.404 153.022 500SCA -28.4105 153.0318 500SCB -28.4068 153.0307 700SCA -28.3991 153.0507 700SCB -28.3982 153.0553 900SCA -28.3868 153.0681 900SCB -28.3868 153.0787 1100SCA -28.3759 153.1033 1100SCB -28.3739 153.0972

Guild 300 500 700 900 1100 Carnivore 0.993 0.816 0.988 0.492 1.00 Frugivore 0.056 0.002 0.383 <0.001 0.333 Granivore 0.993 0.912 0.889 1 0.611 Insectivore 0.048 <0.001 0.144 0.069 0.010 Nectarivore 0.758 1 NA NA NA Omnivore 0.665 0.147 0.052 0.818 <0.001

Appendix 5.2: Species used in analysis and their traits

Count per elevational band Species Body ID Common Name Family Scientific Name 300m 500m 700m 900m 1100m Guild Code Mass (g) 1 ALLY Albert's Lyrebird Menuridae Menura alberti 1 2 2 7 2 Insectivore 930 2 AULO Australian Logrunner Orthonychidae Orthonyx temminckii 14 18 20 32 23 Insectivore 59 3 BATH Bassian Thrush Turdidae Zoothera lunulata 0 0 0 2 7 Insectivore 100 4 BFMO Black-faced Monarch Monarchidae Monarcha melanopsis 11 11 4 6 6 Insectivore 23 5 BRCD Brown Cuckoo-Dove Columbidae Macropygia phasianella 22 17 13 15 10 Frugivore 240 6 BRGE Brown Gerygone Acanthizidae Gerygone mouki 36 97 85 74 59 Insectivore 6 7 BRTB Brown Thornbill Acanthizidae Acanthiza pusilla 11 26 25 27 41 Insectivore 7 8 CROS Crimson Rosella Psittacidae Platycercus elegans 1 2 7 4 7 Granivore 135 9 ESPI Eastern Spinebill Meliphagidae Acanthorhynchus tenuirostris 6 4 9 23 9 Omnivore 11 10 EWHI Eastern Whipbird Psophodidae Psophodes olivaceus 19 36 24 24 24 Insectivore 61 11 EYRO Eastern Yellow Robin Petroicidae Eopsaltria australis 22 14 24 33 29 Insectivore 20 12 FTCU Fan-tailed Cuckoo Cuculidae Cacomantis flabelliformis 1 2 2 0 0 Insectivore 50 13 GOWH Australian Golden Whistler Pachycephalidae Pachycephala pectoralis 21 33 27 25 36 Insectivore 25 14 GRCA Green Catbird Ptilonorhynchidae Ailuroedus crassirostris 2 25 32 11 4 Omnivore 210 15 GRFA Grey Fantail Rhipiduridae Rhipidura albiscapa 25 30 17 11 19 Insectivore 8 16 GRST Grey Shrikethrush Pachycephalidae Colluricincla harmonica 11 16 12 14 19 Carnivore 65 17 KIPA Australian King Parrot Psittacidae Alisterus scapularis 7 13 25 17 18 Granivore 243 18 LBSW Large-billed Scrubwren Acanthizidae Sericornis magnirostra 5 10 16 5 10 Insectivore 10 19 LEHE Lewin's Honeyeater Meliphagidae Meliphaga lewinii 88 86 70 65 46 Omnivore 34 20 MSTB Mistletoebird Dicaeidae Dicaeum hirundinaceum 12 12 16 7 6 Frugivore 9 21 NOPI Noisy Pitta Pittidae Pitta versicolor 3 4 2 6 3 Carnivore 95 22 PCUR Pied Currawong Artamidae Strepera graculina 4 2 3 9 7 Omnivore 300 23 PRIF Paradise Riflebird Paradisaeidae Ptiloris paradiseus 6 7 9 3 0 Omnivore 120 24 PYRO Pale-yellow Robin Petroicidae Tregellasia capito 10 11 4 3 2 Insectivore 14

25 RCFD Rose-crowned Fruit-Dove Columbidae Ptilinopus regina 6 14 3 3 0 Frugivore 105 26 RORO Rose Robin Petroicidae Petroica rosea 0 1 4 3 5 Insectivore 8 27 RTTH Russet-tailed Thrush Turdidae Zoothera heinei 2 4 3 2 0 Insectivore 85 28 RUFA Rufous Fantail Rhipiduridae Rhipidura rufifrons 7 18 3 12 8 Insectivore 10 29 SABB Satin Bowerbird Ptilonorhynchidae Ptilonorhynchus violaceus 6 2 14 7 5 Omnivore 220 30 SBCU Shining Bronze-Cuckoo Cuculidae Chrysococcyx lucidus 1 5 0 9 7 Insectivore 25 31 SCHE Scarlet Myzomela Meliphagidae Myzomela sanguinolenta 30 9 0 0 0 Nectarivore 8 32 SILV Silvereye Zosteropidae Zosterops lateralis 29 28 27 32 34 Omnivore 11 33 SPMO Spectacled Monarch Monarchidae Symposiachrus trivirgatus 13 5 0 0 0 Insectivore 12 34 SPPA Spotted Pardalote Pardalotidae Pardalotus punctatus 6 3 5 0 0 Insectivore 9 35 STTB Striated Thornbill Acanthizidae Acanthiza lineata 25 0 5 0 0 Insectivore 7 36 TOPI Topknot Pigeon Columbidae Lopholaimus antarcticus 1 0 1 2 0 Frugivore 525 37 WBSW White-browed Scrubwren Acanthizidae Sericornis frontalis 15 14 18 18 22 Insectivore 14 38 WHPI White-headed Pigeon Columbidae Columba leucomela 0 5 13 15 1 Frugivore 420 39 WOFD Wompoo Fruit-dove Columbidae Ptilinopus magnificus 5 24 7 4 0 Frugivore 450 40 WTRE White-throated Treecreeper Climacteridae Cormobates leucophaea 33 12 18 9 11 Insectivore 22 41 YFHE Yellow-faced Honeyeater Meliphagidae Calivagus chrysops 24 1 1 0 0 Omnivore 17 42 YTSW Yellow-throated Scrubwren Acanthizidae Sericornis citreogularis 11 42 43 62 50 Insectivore 17 Total abundance 552 665 613 601 530 Total diversity 39 39 38 36 31

Appendix 5.3: R code for fitting GLMMs to determine the effects of environmental factors on avian abundance and species richness

Fitting a GLMM to determine the effects of environmental factors on avian abundance library(readr) library(ggplot2) library(lme4) ## Loading required package: Matrix

Get data in, treat Wetness as a factor, remove NA’s…

TA <- read_csv("~/PhD/Analysis/traits paper/GLMMs/traitsabund.csv", col_types = cols(Date = col_date(format = "%d/%m/%Y"))) TA$Wetness <- as.factor(TA$Wetness) TA <- na.omit(TA) head(TA) ## # A tibble: 6 x 7 ## Site Date Temperature Wetness Canopy CanHeight Abundance ## ## 1 300SCA 2015-09-29 17.0 2 95 25 14 ## 2 300SCB 2015-09-29 17.0 2 83 22 17 ## 3 300BMA 2015-09-30 18.5 0 75 25 11 ## 4 1100BMB 2015-09-30 14.1 1 90 20 12 ## 5 300BMB 2015-09-30 14.9 1 81 24 12 ## 6 1100BMA 2015-09-30 15.1 1 90 18 13

Now let’s look at the relationships between the response and predictors using ggplot aplot1 <- ggplot(TA, aes(Temperature, Abundance)) #This fits a regression line to each site's relationship with temperature # aplot1 + geom_point(aes(colour = factor(Site)))+ geom_smooth(aes(group = Site, colour = factor(Site)), method = "lm", se = F)

It looks like the most appropriate model will include random intercept and slope! Temperature varies in different ways at different sites.

What about the effect of Wetness? aplot2 <- ggplot(TA, aes(Wetness , Abundance)) + geom_point(aes(colour = factor(Site))) aplot2

It looks like abundance generally declines with Wetness. aplot3 <- ggplot(TA, aes(Canopy , Abundance)) + geom_point(aes(colour = factor(Site))) aplot3

No clear pattern here… aplot4 <- ggplot(TA, aes(CanHeight , Abundance)) + geom_point(aes(colour = factor(Site))) aplot4

Or here.

Time to build some models! We first try a maximal model, with a random intercept and slope and including all variables.

97 model1 <- glmer(Abundance ~ Temperature + Wetness + Canopy + CanHeight + (Temperature|Site), family = poisson, data=TA, nAGQ = 0) summary(model1) ## Generalized linear mixed model fit by maximum likelihood (Adaptive ## Gauss-Hermite Quadrature, nAGQ = 0) [glmerMod] ## Family: poisson ( log ) ## Formula: ## Abundance ~ Temperature + Wetness + Canopy + CanHeight + (Temperature | ## Site) ## Data: TA ## ## AIC BIC logLik deviance df.resid ## 1350.9 1384.9 -665.5 1330.9 210 ## ## Scaled residuals: ## Min 1Q Median 3Q Max ## -2.7491 -0.8883 -0.1181 0.6938 5.1989 ## ## Random effects: ## Groups Name Variance Std.Dev. Corr ## Site (Intercept) 0.1512705 0.38894 ## Temperature 0.0006139 0.02478 -0.99 ## Number of obs: 220, groups: Site, 20 ## ## Fixed effects: ## Estimate Std. Error z value Pr(>|z|) ## (Intercept) 1.948297 0.274756 7.091 1.33e-12 *** ## Temperature 0.023806 0.007551 3.153 0.00162 ** ## Wetness1 -0.080065 0.041771 -1.917 0.05527 . ## Wetness2 -0.104989 0.064769 -1.621 0.10502 ## Wetness3 -0.176549 0.083815 -2.106 0.03517 * ## Canopy 0.002689 0.002908 0.925 0.35512 ## CanHeight 0.005522 0.004589 1.203 0.22891 ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Correlation of Fixed Effects: ## (Intr) Tmprtr Wtnss1 Wtnss2 Wtnss3 Canopy ## Temperature -0.388 ## Wetness1 -0.007 -0.004 ## Wetness2 -0.061 0.003 0.243 ## Wetness3 -0.149 0.225 0.200 0.134 ## Canopy -0.811 -0.033 -0.044 0.010 -0.006 ## CanHeight -0.114 -0.017 -0.024 0.027 0.072 -0.314

So it looks like Temperature and the wettest measures of Wetness significantly affect abundance…

Let’s begin the model selection procedure by dropping the random slope and comparing fits using a likelihood ratio test. model2 <- glmer(Abundance ~ Temperature + Wetness + Canopy + CanHeight + (1|Site), family = poisson, data=TA, nAGQ = 0) #summary(model2) #as model2 is 'nested' within model1 (it's just missing the random slope) we can compare using likelihood ratios# anova(model1, model2) ## Data: TA ## Models: ## model2: Abundance ~ Temperature + Wetness + Canopy + CanHeight + (1 | ## model2: Site)

## model1: Abundance ~ Temperature + Wetness + Canopy + CanHeight + (Temperature | ## model1: Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## model2 8 1357.5 1384.6 -670.74 1341.5 ## model1 10 1350.9 1384.9 -665.47 1330.9 10.549 2 0.00512 ** ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Looks like model1 is a significantly better fitting model, according to the anova results and the AIC & BIC’s - leave random slope in. Using backwards selection, we remove the non-significant terms from model1 and compare models further.

First, we drop CanHeight (it is the least significant variable). model3 <- glmer(Abundance ~ Temperature + Wetness + Canopy + (Temperature|Site), family = poisson, data=TA, nAGQ = 0) #summary(model3) anova(model1, model3) ## Data: TA ## Models: ## model3: Abundance ~ Temperature + Wetness + Canopy + (Temperature | Site) ## model1: Abundance ~ Temperature + Wetness + Canopy + CanHeight + (Temperature | ## model1: Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## model3 9 1349.9 1380.4 -665.93 1331.9 ## model1 10 1350.9 1384.9 -665.47 1330.9 0.9257 1 0.336

So having CanHeight in doesn’t significantly improve the fit - we can drop it. Now let’s see what happens if we drop Canopy, which was not significant in model3. model4 <- glmer(Abundance ~ Temperature + Wetness + (Temperature|Site), family = poisson, data=TA, nAGQ = 0) #summary(model4) anova(model3, model4) ## Data: TA ## Models: ## model4: Abundance ~ Temperature + Wetness + (Temperature | Site) ## model3: Abundance ~ Temperature + Wetness + Canopy + (Temperature | Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## model4 8 1349.6 1376.8 -666.82 1333.6 ## model3 9 1349.9 1380.4 -665.93 1331.9 1.781 1 0.182

So having Canopy in doesn’t significantly improve the fit - we can drop it. Note that the AIC only goes down by 0.3 though!

Finally, let’s see what happens if we remove Wetness. model5 <- glmer(Abundance ~ Temperature + (Temperature|Site), family = poisson, data=TA, nAGQ = 0) #summary(model5) anova(model4, model5) ## Data: TA ## Models: ## model5: Abundance ~ Temperature + (Temperature | Site)

## model4: Abundance ~ Temperature + Wetness + (Temperature | Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## model5 5 1350.9 1367.8 -670.43 1340.9 ## model4 8 1349.6 1376.8 -666.82 1333.6 7.2197 3 0.06522 . ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Wetness also doesn’t improve the fit - we can drop it as well. model5 becomes our final model. The AIC is slightly higher for model5 cf. model4, but the BIC is substantially lower.

Now, let’s check that we haven’t violated the assumptions of our final model. par(mfrow=c(2,2)) plot(resid(model5)~ fitted(model5)) abline(h=0) hist(resid(model5)) qqnorm(resid(model5)) qqline(resid(model5))

Residual plot looks ok, the histogram of the residuals also looks good and the QQ plot is acceptable as well.

In conclusion (from model 5): For every degree that temperature increases, abundance will increase by 1.027 individuals.

100

Fitting a GLMM to determine the effects of environmental factors on avian species richness

Get data in, treat Wetness as a factor, remove NA’s traitsrich <- read_csv("~/PhD/Analysis/traits paper/GLMMs/traitsrich.csv", col_types = cols(Date = col_date(format = "%d/%m/%Y"))) Rich <- traitsrich Rich$Wetness <- as.factor(Rich$Wetness) Rich <- na.omit(Rich) head(Rich) ## # A tibble: 6 x 8 ## Site Sample Date Temperature Wetness Canopy CanHeight Richness ## ## 1 300SCA 1 2015-09-29 17.0 2 95 25 11 ## 2 300SCB 1 2015-09-29 17.0 2 83 22 11 ## 3 1100BMA 1 2015-09-30 15.1 1 90 18 9 ## 4 1100BMB 1 2015-09-30 14.1 1 90 20 8 ## 5 300BMA 1 2015-09-30 18.5 0 75 25 8 ## 6 300BMB 1 2015-09-30 14.9 1 81 24 10

Let’s have a look at the response / predictor interactions like we did before. rplot1 <- ggplot(Rich, aes(Temperature, Richness)) rplot1 + geom_point(aes(colour = factor(Site)))+ geom_smooth(aes(group = Site, colour = factor(Site)), method = "lm", se = F)

Similarly to the abundance GLMM it looks like the most appropriate model will be a random intercept and slope!

101 rplot2 <- ggplot(Rich, aes(Wetness , Richness)) + geom_point(aes(colour = factor(Site))) rplot2

Richness seems to decline with increasing levels of Wetness…

rplot3 <- ggplot(Rich, aes(Canopy , Richness)) + geom_point(aes(colour = factor(Site))) rplot3 rplot4 <- ggplot(Rich, aes(CanHeight , Richness)) + geom_point(aes(colour = factor(Site))) rplot4

After plotting the relationships between canopy cover and species richness and canopy height and species richness, we again find no clear patterns in the graphs so omit them here.

Time to build some models! Again, we start with a model that includes a random slope and intercept, + all predictors… rmodel1 <- glmer(Richness ~ Temperature + Wetness + Canopy + CanHeight + (Temperature|Site), family = poisson, data=Rich, nAGQ = 0) summary(rmodel1) ## Generalized linear mixed model fit by maximum likelihood (Adaptive ## Gauss-Hermite Quadrature, nAGQ = 0) [glmerMod] ## Family: poisson ( log ) ## Formula: ## Richness ~ Temperature + Wetness + Canopy + CanHeight + (Temperature | ## Site) ## Data: Rich ## ## AIC BIC logLik deviance df.resid ## 1060.2 1094.1 -520.1 1040.2 210 ## ## Scaled residuals:

102

## Min 1Q Median 3Q Max ## -2.03418 -0.52661 0.01066 0.52603 2.60070 ## ## Random effects: ## Groups Name Variance Std.Dev. Corr ## Site (Intercept) 0.000e+00 0.00e+00 ## Temperature 1.392e-18 1.18e-09 NaN ## Number of obs: 220, groups: Site, 20 ## ## Fixed effects: ## Estimate Std. Error z value Pr(>|z|) ## (Intercept) 1.496348 0.273241 5.476 4.34e-08 *** ## Temperature 0.031702 0.006238 5.082 3.74e-07 *** ## Wetness1 0.009823 0.049624 0.198 0.843 ## Wetness2 -0.027554 0.076629 -0.360 0.719 ## Wetness3 -0.077371 0.100558 -0.769 0.442 ## Canopy 0.001208 0.003095 0.390 0.696 ## CanHeight 0.005523 0.004909 1.125 0.261 ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Correlation of Fixed Effects: ## (Intr) Tmprtr Wtnss1 Wtnss2 Wtnss3 Canopy ## Temperature -0.287 ## Wetness1 -0.017 0.031 ## Wetness2 -0.072 0.010 0.270 ## Wetness3 -0.170 0.343 0.214 0.141 ## Canopy -0.842 -0.035 -0.063 0.003 -0.016 ## CanHeight -0.058 -0.078 -0.014 0.035 0.057 -0.361

Begin the model selection procedure, drop the random slope and compare fits rmodel2 <- glmer(Richness ~ Temperature + Wetness + Canopy + CanHeight + (1|Site), family = poisson, data=Rich, nAGQ = 0) #summary(rmodel2) #as model2 is 'nested' within model1 (it's just missing the random slope) we can compare using likelihood ratios# anova(rmodel1, rmodel2) ## Data: Rich ## Models: ## rmodel2: Richness ~ Temperature + Wetness + Canopy + CanHeight + (1 | ## rmodel2: Site) ## rmodel1: Richness ~ Temperature + Wetness + Canopy + CanHeight + (Temperature | ## rmodel1: Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## rmodel2 8 1056.2 1083.3 -520.08 1040.2 ## rmodel1 10 1060.2 1094.1 -520.08 1040.2 0 2 1

Looks like the random slope is not important to leave in (it has no significant effect on fit), plus removing shrinks AIC and BIC. We can now remove the non-significant terms from rmodel2 and compare models further.

First, we drop Canopy (it is the least significant variable) rmodel3 <- glmer(Richness ~ Temperature + Wetness + CanHeight + (1|Site), family = poisson, data=Rich, nAGQ = 0) #summary(rmodel3) anova(rmodel2, rmodel3) ## Data: Rich

103

## Models: ## rmodel3: Richness ~ Temperature + Wetness + CanHeight + (1 | Site) ## rmodel2: Richness ~ Temperature + Wetness + Canopy + CanHeight + (1 | ## rmodel2: Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## rmodel3 7 1054.3 1078.1 -520.16 1040.3 ## rmodel2 8 1056.2 1083.3 -520.08 1040.2 0.1524 1 0.6962

So having Canopy in doesn’t significantly improve the fit - we can drop it. Wetness is the next least significant, drop that next. rmodel4 <- glmer(Richness ~ Temperature + CanHeight + (1|Site), family = poisson, data=Rich, nAGQ = 0) #summary(rmodel4) anova(rmodel3, rmodel4) ## Data: Rich ## Models: ## rmodel4: Richness ~ Temperature + CanHeight + (1 | Site) ## rmodel3: Richness ~ Temperature + Wetness + CanHeight + (1 | Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## rmodel4 4 1049.2 1062.8 -520.60 1041.2 ## rmodel3 7 1054.3 1078.1 -520.16 1040.3 0.884 3 0.8293

So having Wetness in doesn’t significantly improve the fit - we can drop it. What about CanHeight? rmodel5 <- glmer(Richness ~ Temperature + (1|Site), family = poisson, data=Rich, nAGQ = 0) #summary(rmodel5) anova(rmodel4, rmodel5) ## Data: Rich ## Models: ## rmodel5: Richness ~ Temperature + (1 | Site) ## rmodel4: Richness ~ Temperature + CanHeight + (1 | Site) ## Df AIC BIC logLik deviance Chisq Chi Df Pr(>Chisq) ## rmodel5 3 1049.3 1059.5 -521.64 1043.3 ## rmodel4 4 1049.2 1062.8 -520.60 1041.2 2.0911 1 0.1482

AIC and BIC do not decline much between rmod4 and rmod5, but CanHeight doesn’t improve the fit - we can drop it as well. rmodel5 becomes our final model. Now let’s check the assumptions again… par(mfrow=c(2,2)) plot(resid(rmodel5)~ fitted(rmodel5)) abline(h=0) hist(resid(rmodel5)) qqnorm(resid(rmodel5)) qqline(resid(rmodel5))

104

It all looks pretty good!

In conclusion (from rmodel5): For every degree that temperature increases, Richness will increase by 1.035 species.

105

Appendix 5.3: R code for fitting a GLMM that predicts the abundance of a species as a function of environmental variables, the species’ traits, and the trait-environment interactions.

Here we are performing Jamil’s forward selection with overdispersion accounted for.

#load required libraries# library(readr) library(lme4) ## Loading required package: Matrix library(ggplot2) library(lattice) library(DHARMa)

First we load the data, and log transform body size. tsumt <- read_csv("~/PhD/Analysis/traits paper/glmm jamil/practise/abundglmmpractise/tsumt.csv") ## Parsed with column specification: ## cols( ## Site = col_character(), ## Date = col_character(), ## Temperature = col_double(), ## Wetness = col_integer(), ## Canopy = col_integer(), ## CanHeight = col_integer(), ## Species = col_character(), ## Guild = col_character(), ## BodySize = col_integer(), ## Count = col_integer() ## ) tsumt$Wetness <- as.factor(tsumt$Wetness) tsumt$Guild <- as.factor(tsumt$Guild) tsumt$logbody <- log(tsumt$BodySize)

Now name the quantitative variables, the environmental variables, and the trait variables.

(quant.vars = colnames(tsumt)[c(3,5,6)]) ## [1] "Temperature" "Canopy" "CanHeight" (EnvBlock = colnames(tsumt)[c(3:6)]) ## [1] "Temperature" "Wetness" "Canopy" "CanHeight" (TraitBlock = colnames(tsumt)[c(8,11)]) ## [1] "Guild" "logbody"

Scale the quantitative variables to zero mean, unit variance for stability and easier interpretation for (var in quant.vars) { tsumt[,var]=scale(tsumt[,var]) }

We add an observation-level random effect to account for overdispersion tsumt$rows = 1:nrow(tsumt)

The data is now ready to be analysed, let’s have a look. head(tsumt)

106

## # A tibble: 6 x 12 ## Site Date Temperature Wetness Canopy CanHeight Species ## ## 1 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 BRTB ## 2 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 EWHI ## 3 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 GOWH ## 4 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 BFMO ## 5 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 GRFA ## 6 300BMA 30/09/2015 0.9683286 0 -1.912511 0.01011785 LBSW ## # ... with 5 more variables: Guild , BodySize , Count , ## # logbody , rows

Now we load the model selection functions provided by Jamil et. al. (2013).

# function to calculate SigAIC value of the model SigAIC<-function(mod,Penalty=qchisq(0.95,1)){ LL <- logLik(mod) ll <- as.numeric(LL) df <- attr(LL, "df") AIC= as.numeric(-2*ll+ Penalty*df) AIC }

# model selection tier 1, function to select the Random effects (environmental variables) Model_select_tier1R<-function(start.model, terms, block){

Sigaic<-NA M<-list() for(j in 1:length(block)) { M[[j]]<-update(start.model,as.formula(paste(". ~ . + (",terms,"+", block[j],"|Species)-(",terms,"|Species)"))) } LL1<-unlist(lapply(M,SigAIC)) k<- order(LL1,decreasing=FALSE) [1] terms <-paste(terms, block[k], sep = "+") list( B=block[-k],term=block[k], terms = terms , new.model=M[[k]],SigAic = LL1[k]) }

# model selection tier 2, function to select the trait + trait:environment interaction terms Model_select_tier2F<-function(start.model, block1,block2){

M = list() for(j in 1:length(block1)){ M[[j]] <- update(start.model, as.formula(paste(". ~ . + ",block2[j],"+", block1[j],":",block2[j]))) } LL1<-unlist(lapply(M,SigAIC)) k<- order(LL1,decreasing=FALSE) [1] list(B1=block1[-k], B2=block2[-k] , new.model= M[[k]], term1 = block1[k],term2= block2[k]) }

Ok; the data is ready now and we have our functions. We can now begin the model selection procedure.

107

Tier 1: select random effect terms to include in the model

Our null model includes a random effect for species, site and observation. The observation-level random effect is a way of accounting for overdispersion.

(start.model <- glmer(Count ~ (1|Species) + (1|Site) + (1|rows), family=poisson, data=tsumt, nAGQ = 0)) ## Generalized linear mixed model fit by maximum likelihood (Adaptive ## Gauss-Hermite Quadrature, nAGQ = 0) [glmerMod] ## Family: poisson ( log ) ## Formula: Count ~ (1 | Species) + (1 | Site) + (1 | rows) ## Data: tsumt ## AIC BIC logLik deviance df.resid ## 11675.594 11704.119 -5833.797 11667.594 9235 ## Random effects: ## Groups Name Std.Dev. ## rows (Intercept) 0.59900 ## Species (Intercept) 0.99253 ## Site (Intercept) 0.09106 ## Number of obs: 9239, groups: rows, 9239; Species, 42; Site, 20 ## Fixed Effects: ## (Intercept) ## -1.622 SigAIC(start.model) ## [1] 11682.96 terms = "1" i = 0 modellist = list() envvar = NULL Sig.AIC = NULL SigAIC1 = SigAIC(start.model ) SigAIC0 = SigAIC1 + 1 block = as.vector(EnvBlock) random.terms = NULL Sig.AICR = SigAIC(start.model ) while (SigAIC1 < SigAIC0){ i = i+1 SigAIC0 = SigAIC1 Model.sel<- Model_select_tier1R(start.model, terms, block) start.model<- Model.sel$new.model SigAIC1= SigAIC(start.model) block<-Model.sel$B terms<-Model.sel$terms # save model and terms if (SigAIC1 < SigAIC0){ final.modelR = start.model random.termsF = c(random.terms, Model.sel$term) } modellist[[i]] =start.model random.terms = c(random.terms, Model.sel$term) Sig.AICR = c(Sig.AICR, SigAIC1) print(c(Model.sel$term, SigAIC1)) } ## [1] "Canopy" "11566.4032663548" ## [1] "Temperature" "11500.1323935881" ## [1] "CanHeight" "11482.4069396203" ## [1] "Wetness" "11483.139811515"

We can see that by including the variables ‘Canopy’, ‘Temperature’ and ‘CanHeight’ in the model, we reduce the SigAIC substantially. However, including the environmental variable ‘Wetness’ does not significantly improve the fit of the model.

108 end.tier1.formula = formula(final.modelR) print(end.tier1.formula) ## Count ~ (1 | Site) + (1 | rows) + (1 + Canopy + Temperature + ## CanHeight | Species)

This model includes only those random terms that significantly improve the fit of the model.

Tier 2: add Fixed Effects for all selected random terms

We now add the fixed effects for all random terms selected in Tier 1. formcc= as.character(end.tier1.formula) formcc = paste(formcc[2],formcc[1], formcc[3]) for (term in random.termsF){ formcc = paste(formcc,"+",term) } start.tier2.formula = as.formula(formcc)

(start.model <- glmer(eval(start.tier2.formula), family=poisson, data=tsumt, nAGQ = 0)) ## Generalized linear mixed model fit by maximum likelihood (Adaptive ## Gauss-Hermite Quadrature, nAGQ = 0) [glmerMod] ## Family: poisson ( log ) ## Formula: Count ~ (1 | Site) + (1 | rows) + (1 + Canopy + Temperature + ## CanHeight | Species) + Canopy + Temperature + CanHeight ## Data: tsumt ## AIC BIC logLik deviance df.resid ## 11442.334 11556.433 -5705.167 11410.334 9223 ## Random effects: ## Groups Name Std.Dev. Corr ## rows (Intercept) 0.50097 ## Species (Intercept) 1.07005 ## Canopy 0.44908 0.23 ## Temperature 0.22584 -0.51 0.03 ## CanHeight 0.24461 -0.19 -0.66 0.17 ## Site (Intercept) 0.06124 ## Number of obs: 9239, groups: rows, 9239; Species, 42; Site, 20 ## Fixed Effects: ## (Intercept) Canopy Temperature CanHeight ## -1.73614 -0.01761 0.22034 0.05983 print(start.tier2.formula) ## Count ~ (1 | Site) + (1 | rows) + (1 + Canopy + Temperature + ## CanHeight | Species) + Canopy + Temperature + CanHeight

The model now includes random effects and the fixed effects of the environmental variables. We can begin the next phase of the model selection, adding the fixed effects of traits and the trait * environmental variable interaction terms.

Tier 2: search for best Trait + Trait X Environment interactions block1 = rep(random.termsF, each= length(TraitBlock)) block2 = rep(TraitBlock, times= length(random.termsF)) i = 0 modellist = list() envvar = NULL trait = NULL Sig.AICF = NULL SigAIC1 = SigAIC(start.model )

109

SigAIC0 = SigAIC1 + 1 while (SigAIC1 < SigAIC0){ i = i+1 SigAIC0 = SigAIC1 Model.sel<- Model_select_tier2F(start.model, block1,block2) start.model<- Model.sel$new.model SigAIC1= SigAIC(start.model) block1<-Model.sel$B1 block2<-Model.sel$B2 # save model and terms modellist[[i]] = Model.sel$new.model envvar = c(envvar, Model.sel$term1) trait = c(trait, Model.sel$term2) Sig.AICF = c(Sig.AICF, SigAIC1) if (SigAIC1 < SigAIC0){ final.model = start.model } print(c(Model.sel$term1,Model.sel$term2,SigAIC1)) } ## [1] "Temperature" "logbody" "11464.725512057" ## [1] "Canopy" "logbody" "11462.2753761516" ## [1] "CanHeight" "logbody" "11465.8266173903" M0 = final.model

We can see that the interaction between canopy height and body size does not significantly improve the fit of the model.

Tier 3: examine final model, delete any non-significant trait X environment terms and refit summary(M0) ## Generalized linear mixed model fit by maximum likelihood (Adaptive ## Gauss-Hermite Quadrature, nAGQ = 0) [glmerMod] ## Family: poisson ( log ) ## Formula: Count ~ (1 | Site) + (1 | rows) + (1 + Canopy + Temperature + ## CanHeight | Species) + Canopy + Temperature + CanHeight + ## logbody + Temperature:logbody + Canopy:logbody ## Data: tsumt ## ## AIC BIC logLik deviance df.resid ## 11427.3 11562.8 -5694.6 11389.3 9220 ## ## Scaled residuals: ## Min 1Q Median 3Q Max ## -1.1595 -0.4799 -0.3014 -0.1632 10.4755 ## ## Random effects: ## Groups Name Variance Std.Dev. Corr ## rows (Intercept) 0.251209 0.50121 ## Species (Intercept) 1.020568 1.01023 ## Canopy 0.170668 0.41312 0.40 ## Temperature 0.027789 0.16670 -0.45 -0.33 ## CanHeight 0.058376 0.24161 -0.24 -0.66 0.30 ## Site (Intercept) 0.003761 0.06133 ## Number of obs: 9239, groups: rows, 9239; Species, 42; Site, 20 ## ## Fixed effects: ## Estimate Std. Error z value Pr(>|z|) ## (Intercept) -0.81126 0.43347 -1.872 0.061266 . ## Canopy -0.40628 0.16716 -2.430 0.015078 * ## Temperature -0.13500 0.09905 -1.363 0.172878

110

## CanHeight 0.06139 0.04897 1.254 0.209940 ## logbody -0.25063 0.10958 -2.287 0.022190 * ## Temperature:logbody 0.09600 0.02653 3.619 0.000296 *** ## Canopy:logbody 0.10603 0.04200 2.525 0.011585 * ## --- ## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 ## ## Correlation of Fixed Effects: ## (Intr) Canopy Tmprtr CnHght logbdy Tmprt: ## Canopy 0.314 ## Temperature -0.327 -0.186 ## CanHeight -0.083 -0.273 0.057 ## logbody -0.930 -0.271 0.310 0.015 ## Tmprtr:lgbd 0.292 0.155 -0.931 -0.006 -0.322 ## Canpy:lgbdy -0.272 -0.902 0.165 0.035 0.284 -0.169 SigAIC(M0) ## [1] 11462.28

As there are no non-significant interaction terms, we conclude that M0 is the final model! We can now check assumptions using QQplots (need packages ggplot2 and lattice for this step).

To check the normality of the random effects… reStack<-stack(ranef(M0)$Species) print(qqmath(~values|ind,reStack,scales=list(relation="free"), prepanel = prepanel.qqmathline, panel = function(x, ...) { panel.qqmathline(x, ...) panel.qqmath(x, ...) }))

There’s a little bit of weirdness associated with canopy % and canopy height - maybe due to few low levels of each in the data? Temperature and the intercept look ok though. Now let’s test for test for normality and overdispersion using the package DHARMa.

111 simulationOutput <- simulateResiduals(M0, n=250) plotSimulatedResiduals(simulationOutput)

The QQ plot looks good! There is also no evidence of zero-inflation or overdispersion. So the row effect was effective in controlling for these problems.

References

Jamil, T., Opdekamp, W., Van Diggelen, R. & Ter Braak, C. J. 2012. Trait-environment relationships and tiered forward model selection in linear mixed models. International Journal of Ecology, 2012.

Jamil, T., Ozinga, W. A., Kleyer, M. & Ter Braak, C. J. F. 2013. Selecting traits that explain species- environment relationships: a generalized linear mixed model approach. Journal of Vegetation Science, 24, 988-1000.

112

Appendix 5.5

This Appendix contains five figures.

Figure S5.5a Species accumulation and sample completeness curves for 300m sites.

113

Figure S5.5b Species accumulation and sample completeness curves for 500m sites.

114

Figure S5.5c Species accumulation and sample completeness curves for 700m sites.

115

Figure S5.5d Species accumulation and sample completeness curves for 900m sites.

116

Figure S5.5e Species accumulation and sample completeness curves for 1100m sites.

117

Chapter 6

General discussion and conclusions

6.1 Introduction

Mountainous regions worldwide are hotspots of avian diversity and endemism (Ruggiero and Hawkins, 2008; Jenkins et al., 2013). Climate change, however, seriously threatens the stability and persistence of montane avian communities (Williams et al., 2003; Laurance et al., 2011a; Harris et al., 2014). The aims of this thesis were to determine effective ways of monitoring the impacts of climate change on rainforest bird communities of eastern Australia, and to gain a better understanding of the driving factors behind the assemblage of these communities along elevational gradients. The baseline data presented in this thesis complements that collected on arthropods and plants during both the Eungella Biodiversity Survey (Odell et al., 2015; Ashton et al., 2016; Burwell and Nakamura, 2016) and the IBISCA Queensland project (Kitching et al., 2011; Laidlaw et al., 2011; Maunsell et al., 2013), and will become an invaluable resource for detecting future regional changes in biodiversity (Shoo et al., 2014). In Fig. 1.1 (Chapter 1), I identified gaps in our current knowledge of the impacts of climate change on the rainforest birds of eastern Australia. The phenology of these communities was poorly known, and the elevational changes in species’ distributions that may lead to novel communities and eventual population decline were previously largely unquantified.

In this thesis, I have shown that:

 Using point counts and automated recording units (ARUs) in tandem is an effective and practical way of rapidly surveying Australian rainforest birds (Chapter 2);  ARUs can be used to monitor cryptic rainforest species such as Zoothera thrushes remotely for long time periods, providing insights into their behaviour and phenology, including the annual timing of breeding-specific vocalisations (Chapter 3);  Bird species can be used as indicators for long-term monitoring of the effects of climate change in Australian subtropical rainforests (Chapter 4);  Bird species richness and abundance along elevational gradients in subtropical Australian rainforest is strongly influenced by changes in temperature (Chapter 5);  Large-bodied and small-bodied birds in subtropical Australian rainforests respond differently to changes in temperature (Chapter 5).

118

The data underlying these findings has gone some way to addressing the gaps in knowledge identified in Fig. 1.1, but there is still much work to be done. This chapter concludes the thesis by providing a general discussion of the main themes, outlining possible limitations of this research, and making suggestions regarding future research directions.

6.2 Monitoring the effects of climate change on rainforest birds

Detecting and monitoring birds in rainforests is often a difficult task, due to inclement weather, access problems, poor visibility and high diversity (Anderson, 2011). However, the future conservation of these ecosystems, which face many and varied threats, relies on the establishment of accurate, long-term monitoring programs (Magurran et al., 2010). The contributions of citizen scientists to large-scale, volunteer-based data repositories such as eBird or the Birdlife Australia Atlas project are becoming an increasingly valuable resource for conservation (Gibbs et al., 2011; Callaghan and Gawlik, 2015). Nevertheless, there are limitations to these sources of data, and other approaches are necessary. Increasingly, researchers are making use of autonomous recording units (ARUs) for avian ecological research, with more than 60 articles involving ARUs published over the last 10 years (Shonfield and Bayne, 2017). ARUs have many advantages over traditional survey methods, but the most compelling in the current context is the potential for ARUs to monitor biodiversity in remote regions autonomously for long periods of time (perhaps, eventually, indefinitely) (Shonfield and Bayne, 2017). I found that data generated by ARUs deployed in Australian rainforests for two months (Chapter 2) and one year (Chapter 4) accurately represented the composition of the avian community in the study areas over those time periods. As such, I suggest that using ARUs to monitor Australian rainforest bird communities is a viable long-term monitoring strategy, capable of detecting changes in species’ phenology or distributions, and warrants further exploration.

In other countries and in other habitats, ARUs have been deployed for time periods of up to five years (Frommolt and Tauchert, 2014; Gage and Axel, 2014). To be efficient, automated approaches to analysis must deal with the massive amounts of data generated over these time frames (Venier et al., 2017). Automated approaches, so far, have typically involved the use of one or more ‘acoustic indices’, which attempt to quantify the diversity of biotic sounds present in the recordings (Pieretti et al., 2011; Towsey et al., 2014). There have been more than 20 acoustic indices proposed in recent years (Sueur et al., 2014), but how well these

119 actually represent avian diversity or community structure is still a matter of debate; for example, one recent paper found only tenuous links between several indices and species richness (Mammides et al., 2017). More research into the use of acoustic indices for monitoring biodiversity, especially in Australian contexts, is necessary in order to determine whether they can be incorporated into future climate change monitoring schemes. An alternate approach to the ‘big-data’ problem presented by ARUs is the development of algorithms that can automatically detect species of interest in long-term recordings (Bardeli et al., 2010; Digby et al., 2013). So far, these methods have had mixed success (Joshi et al., 2017; Venier et al., 2017). In some cases, when bird sounds are loud, distinctive and made in relatively quiet environments (e.g. nightjar calls), automated recognisers can perform very well (Zwart et al., 2014). Unfortunately, these ideal scenarios are rare, and most studies attempting to use automatic recognisers report high error rates (Joshi et al., 2017).

In Australian rainforests, where acoustic complexity can be very high, the use of automated recognisers for species identification and monitoring has not yet been tested. The challenges inherent in such an approach must be overcome, however, if the effectiveness of ARUs as devices for the long-term monitoring of avian biodiversity is to be maximised. The use of indicators, as proposed in Chapter 4, may simplify the design of automatic recognisers at the community scale. Thankfully, the use of ARUs as a part of biodiversity surveys like those described in this thesis leads to permanent recordings being made of the vocal avian community – another major benefit associated with their use (Shonfield and Bayne, 2017). Advances in computing power, software development and analytical techniques may, in the future, allow for retroactive investigations of the baseline acoustic datasets collected today (Sedláček et al., 2015).

Traditional methods of monitoring rainforest birds, such as the point count methodology adopted in this thesis, still contribute valuable information to monitoring programs. In particular, estimates of species abundance generated through distance-based point counts can be reliably used to predict changes in population size (Shoo et al., 2005). Abundance estimates are currently difficult, but not impossible, to generate using data from acoustic recordings. So far, studies doing so have been limited to one (Lambert and McDonald, 2014), or four (Darras et al., 2018) species; community-level abundance estimation from acoustic data is a formidable challenge. Abundance estimates along elevational gradients, however they are generated, provide a logical direction for future research into the effects of climate change on the rainforest birds of eastern Australia. Along extensive elevational gradients,

120 such as those found in Papua New Guinea, the Himalaya or the Andes, species are often restricted to specific elevational ranges (Acharya et al., 2011; Freeman and Class Freeman, 2014; Londoño et al., 2017). However, along less extensive elevational gradients, such as those surveyed in this thesis, species can occupy the entire elevational range. Obviously, there are exceptions: the two species of Zoothera thrushes found on the east coast of Australia, for example, have marked preferences in their elevational distributions (explored in Chapter 3). For species found at all elevations however, such as Lewin’s Honeyeater Meliphaga lewinii, detecting climate change-induced upwards shifts in distribution without abundance data is problematic. Estimating the elevational peak in Lewin’s Honeyeater abundance, however, may allow for the detection of future shifts in the distribution of that species’ population (Shoo et al., 2006) if comparisons with the baseline data presented in this thesis are made.

Similarly, examining seasonal changes in the abundance of species along elevational gradients may be an effective way of detecting seasonal elevational migration in Australian rainforest birds. Many Australian bird species are known to engage in this behaviour (for e.g. see Dingle, 2004). Quantifying the extent of seasonal movements is difficult, but important: those species which are elevational migrants may be better placed to undertake adaptive behavioural movements in response to warming temperatures. Indeed, a recent article showed that species with the ability to move into the forest canopy (a comparatively minor ‘elevational migration’) were better able to exploit changes in local climate than non-arboreal species (Scheffers et al., 2017). Species which do not engage in seasonal elevational migration are of greater conservation concern, and identification of these species should be a research priority. A better understanding of seasonal elevational migration in the Australian avifauna will provide conservation actors with vital information that is currently lacking.

In this thesis, I used a combination of traditional observational methods and automated acoustic monitoring to generate baseline data on the community structure and elevational preferences of rainforest birds in eastern Australia. Each method was found to have advantages and disadvantages (discussed in detail in Chapter 2). In Chapter 4, data from both methods were used to describe a suite of avian indicators suitable for the long-term monitoring of climate change impacts on the birds of the Gondwanan rainforests, demonstrating that both methods, used in tandem, can be complementary. When estimating abundance or population size is a research priority, I recommend that point counts or other distance-based methodologies be used. For daily monitoring, monitoring over long time

121 periods and spatial scales, or simultaneous monitoring at different elevations along an elevational gradient, I recommend the use of ARUs.

6.3 Potential effects of climate change on Australian rainforest birds

Three influential reviews have described four likely effects of climate change on species worldwide (Parmesan and Yohe, 2003; Root et al., 2003; Parmesan, 2006). These effects can be summarised as follows: alteration in species density and spatial distribution; shifts in phenology; changes in morphology; and changes in genetic frequencies within populations (evolutionary responses). In Australia, climate change research concerning birds is still in its fledgling phase, despite some notable contributions (e.g. Shoo et al., 2005; Anderson, 2011). It is likely that the first indication of climate change effects on rainforest birds will come from the detection of shifts in species’ distributions and densities; it may be argued that these changes are already taking place, and are simply undocumented. It is possible that some species may be able to adapt to the expected changes in climatic conditions, but others are certainly at risk of extinction (Parmesan, 2006). In particular, high-elevation specialists such as the Golden Bowerbird Prionodura newtonia are under extreme pressure due to the dramatic declines in suitable breeding habitat forecast under various warming scenarios (Hilbert et al., 2004). In Chapter 3 I proposed that localised extinction of the Bassian Thrush Zoothera lunulata in south-east Queensland and north-east New South Wales is a likely consequence of continued climatic warming in the region. There is some hope for such species: research in Peru has indicated that endothermy allows for flexibility in the responses of bird species to temperature change along elevational gradients (Forero-Medina et al., 2011). Nevertheless, conservation priority should be assigned to high-elevation specialists in future climate change mitigation programs.

Changes in the timing of migration and breeding of Australian bird species have received relatively little attention in comparison with the birds of Europe or North America, largely due to the paucity of long-term datasets available (Beaumont et al., 2006). However, there is some evidence that the timing of migratory behaviour in several Australian species, including some that occupy rainforests, has advanced significantly over the last few decades in response to warming temperatures (Beaumont et al., 2006). Additionally, Green and Pickering (2002) reported earlier arrivals of elevational migrants in the alpine zone in the 1980’s and 1990’s, and Chambers (2005) reported advances in arrival time for two species

122 commonly recorded at the Eyre Bird Observatory in Western Australia. Repetition of these studies may identify further phenological changes. The continuation of Birdlife Australia’s Atlas Project (Barrett et al., 2003), the rise in the use of eBird in Australia, and baseline data, such as that generated in this thesis, may also assist researchers concerned with changes in the phenology of Australian birds (Shoo et al., 2014).

Morphological changes in Australian birds as a consequence of climate change are difficult to quantify. To my knowledge, only Gardner et al. (2009) have attempted to do so. They found evidence of a decline in body size of eight Australian passerines of various body sizes over the last 100 years by measuring museum specimens. The decline in body size may be an example of micro-evolutionary change in the context of Bergmann’s Rule, which predicts an increase in body size with a decline in temperature (Gardner et al., 2009). Overseas, related research has identified similar trends in body size (Yom-Tov et al., 2006; Teplitsky et al., 2008). In Chapter 5, I found that bird species with different body sizes had differential responses to temperature changes. As a potential consequence of these varying relationships to temperature over time, evolutionary responses favouring smaller body sizes may develop in avian populations (Sheridan and Bickford, 2011), representing local adaptation to changing conditions (Mackey et al., 2008).

Evolutionary adaptation to climate change in Australia’s avifauna may be the only way for threatened species to persist if they are unable to make adaptive movements, or tolerate increasing temperatures (Forero-Medina et al., 2011; Hoffmann and Sgrò, 2011). However, if environmental conditions change too quickly, extinction rather than evolution is a likely outcome (Hoffmann and Sgrò, 2011). More research on the evolutionary potential of Australia’s birds, and in particular the effects of genetic isolation, is necessary for the creation of effective conservation strategies.

6.4 The drivers of avian species richness and abundance in Australian rainforests

Understanding the mechanisms responsible for the assemblage of avian communities is a central challenge in community-level ornithology. Local climate is often regarded as the primary driver of avian community assemblage (McCain, 2009; Howard et al., 2015). However, disentangling the effects of multiple potential drivers can be difficult (Kraft et al., 2011). To answer these questions, researchers commonly relate data on avian populations or communities to changes in environmental conditions along environmental gradients, both

123 longitudinal and elevational (Körner, 2007; McCain, 2009). In Australia, Anderson et al. (2013) demonstrated that temperature was an important driver of rainforest bird distributions along elevational gradients in the Wet Tropics bioregion. In Chapter 5, I found that temperature was significantly related to species richness and abundance patterns within the rainforest bird assemblage of the northern Gondwanan rainforests. These results from Australian montane ecosystems are similar to those described in a global meta-analysis of bird elevational diversity, which incorporated data from 78 elevational gradients (McCain, 2009). In that study, temperature and water availability were determined to be the primary drivers of avian species richness along elevational gradients – and temperature had a stronger effect on ‘wet’ mountains.

On the east coast of Australia, annually averaged temperature is expected to increase by up to 5°C by 2090 under a high-emissions scenario; an intermediate-emissions scenario yields an increase of up to 2.6°C (CSIRO and Bureau of Meteorology, 2016). These temperature increases will lead to an upwards shift in current climatic conditions of approximately 400 and 650m asl respectively, effectively sealing the fate of high-elevation specialists (such as the Bassian Thrush) in the region. The effects of climatic change on rainfall patterns are less certain, but the forecast change in temperature alone will substantially disrupt the distributions of species along elevational gradients and lead to the alteration of entire ecosystems (Laurance et al., 2011a; CSIRO and Bureau of Meteorology, 2016). Determining the likely ‘winners’ and ‘losers’ amongst the Australian rainforest avifauna as the climate continues to warm will require detailed, long-term field observations (Sitters et al., 2016). Variation in species-specific functional traits may influence response to temperature, as reported in Chapter 5, and the inclusion of relevant trait information in predictive models is important (McGill et al., 2006; Jamil et al., 2013). The inclusion of phylogenetic information (Li and Ives, 2017), behavioural examinations of inter-specific interactions (Jankowski et al., 2010) and a more accurate measure of water availability in models examining species’ responses to changes in environmental conditions are important future directions for research in this field, but overfitting is a problem that must be acknowledged (Babyak, 2004). Ecological research into the use of latent variable models, which can account for missing covariates and interspecific interactions, has only just begun (see for e.g. Warton et al., 2015; Niku et al., 2017). Eventually, however, this class of model may provide the flexibility necessary to address the challenges associated with modelling species responses to changing environmental conditions.

124

6.5 Conclusions

The rainforest ecosystems of the eastern coast of Australia are likely to face significant disruptions as a result of increasing global temperatures (Laurance et al., 2011b). While many of these forests are protected as part of the National Reserve System (NRS), significant tracts of forest without formal protection remain. The east coast of Australia has a particularly high proportion of climate change refugia in comparison to other areas of the continent (Reside et al., 2013). Australia’s NRS requires significant expansion in order to function as part of an effective, continent-wide conservation plan (Mackey et al., 2008). If possible, entire areas of remnant habitat along the east coast (and elsewhere), particularly along elevational gradients, should be protected and incorporated into the NRS to give species their ‘best chance’ of persisting in the face of multiple threatening processes (Blake and Loiselle, 2000; Hall et al., 2009). Adding to the NRS and concurrently restoring degraded landscapes will facilitate improvements in landscape-scale habitat connectivity, potentially allowing for adaptive movements among the avian assemblage (Soulé et al., 2004; Mackey et al., 2008). There are hopeful signs: the total area of land protected in the NRS almost doubled between 1997 and 2012, to cover a total area of 1.19 million km2 (15.5% of the continent) (Legge, 2015), and the Great Eastern Ranges Initiative is actively working to improve habitat connectivity across the entire latitudinal distribution of the Great Dividing Range (Pulsford et al., 2013). However, monitoring the outcomes of conservation activities remains a challenge (Legge, 2015), and recent research has indicated that one-third of protected land globally is under intense human pressure (Jones et al., 2018). I sincerely hope that the information and data presented in this thesis will contribute to the future conservation of Australia’s rainforest avifauna.

125

References

Able, K. P. & Noon, B. R. 1976. Avian community structure along elevational gradients in the northeastern United States. Oecologia, 26, 275-294. Acharya, B. K., Sanders, N. J., Vijayan, L. & Chettri, B. 2011. Elevational gradients in bird diversity in the eastern Himalaya: an evaluation of distribution patterns and their underlying mechanisms. PLoS ONE, 6, e29097. Aide, T. M., Corrada-Bravo, C., Campos-Cerqueira, M., Milan, C., Vega, G. & Alvarez, R. 2013. Real-time bioacoustics monitoring and automated species identification. PeerJ, 1, e103. Anderson, A. S. 2011. Influences of the past, present and future climate on the structure and diversity of rainforest bird assemblages in north-eastern Australia. Ph.D., James Cook University. Anderson, A. S., Storlie, C. J., Shoo, L. P., Pearson, R. G. & Williams, S. E. 2013. Current analogues of future climate indicate the likely response of a sensitive montane tropical avifauna to a warming world. PLoS ONE, 8, e69393. Arnold, G. W. 1988. Possible effects of climate change on wildlife in Western Australia. In: Pearman, G. I. (ed.) Greenhouse: Planning for Climate Change. pp. 375–386. Melbourne: CSIRO Australia. Ashton, L. A., Odell, E. H., Burwell, C. J., Maunsell, S. C., Nakamura, A., Mcdonald, W. J. F. & Kitching, R. L. 2016. Altitudinal patterns of moth diversity in tropical and subtropical Australian rainforests. Austral Ecology, 41, 197-208. Auer, S. K. & King, D. I. 2014. Ecological and life‐history traits explain recent boundary shifts in elevation and latitude of western North American songbirds. Global Ecology and Biogeography, 23, 867-875. Babyak, M. A. 2004. What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosomatic Medicine, 66, 411-421. Balmer, D., Gillings, S., Caffrey, B., Swann, B., Downie, I. & Fuller, R. 2014. Bird Atlas 2007-11: The Breeding and Wintering Birds of Britain and Ireland, Thetford, British Trust for Ornithology.

126

Bardeli, R., Wolff, D., Kurth, F., Koch, M., Tauchert, K. H. & Frommolt, K. H. 2010. Detecting bird sounds in a complex acoustic environment and application to bioacoustic monitoring. Pattern Recognition Letters, 31, 1524-1534. Barrett, G., Silcocks, A., Barry, S., Cunningham, R. & Poulter, R. 2003. The New Atlas of Australian Birds, Melbourne, Royal Australasian Ornithologists Union. Bates, D., Mächler, M., Bolker, B. & Walker, S. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67, 48. Beaumont, L. J., Mcallan, I. a. W. & Hughes, L. 2006. A matter of timing: changes in the first date of arrival and last date of departure of Australian migratory birds. Global Change Biology, 12, 1339-1354. Beck, M. W. 2017. ggord: Ordination Plots with ggplot2. https://zenodo.org/badge/latestdoi/ 35334615 Becker, A., Körner, C., Brun, J.-J., Guisan, A. & Tappeiner, U. 2007. Ecological and land use studies along elevational gradients. Mountain Research and Development, 27, 58-65. Bibby, C. J., Burgess, N. D. & Hill, D. A. 1992. Bird Census Techniques, Sydney: Academic Press. Birdlife Australia. 2017. Key Biodiversity Areas (KBAs) - Nature’s Hotspots [Online]. Birdlife Australia. Available: http://www.birdlife.org.au/projects/KBA [Accessed July 2017]. Birdlife International. 2017. Important Bird Areas factsheet: Scenic Rim [Online]. Birdlife International. Available: http://datazone.birdlife.org/site/factsheet/24332 [Accessed July 2017]. Blake, J. G. & Loiselle, B. A. 2000. Diversity of birds along an elevational gradient in the Cordillera Central, Costa Rica. The Auk, 117, 663-686. Bolker, B. M., Brooks, M. E., Clark, C. J., Geange, S. W., Poulsen, J. R., Stevens, M. H. H. & White, J.-S. S. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology & Evolution, 24, 127-135. Both, C., Bouwhuis, S., Lessells, C. M. & Visser, M. E. 2006. Climate change and population declines in a long-distance migratory bird. Nature, 441, 81-3. Both, C. 2010. Food availability, mistiming, and climatic change. In: Møller, A. P., Fiedler, W. & Berthold, P. (eds.) Effects of Climate Change on Birds. pp. 129-147. Oxford: Oxford University Press.

127

Bregman, T. P., Şekercioğlu, Ç. H. & Tobias, J. A. 2014. Global patterns and predictors of bird species responses to forest fragmentation: Implications for ecosystem function and conservation. Biological Conservation, 169, 372-383. Brehm, G., Homeier, J. & Fiedler, K. 2003. Beta diversity of geometrid moths (Lepidoptera: Geometridae) in an Andean montane rainforest. Diversity and Distributions, 9, 351- 366. Brown, A. M., Warton, D. I., Andrew, N. R., Binns, M., Cassis, G. & Gibb, H. 2014. The fourth-corner solution – using predictive models to understand how species traits interact with the environment. Methods in Ecology and Evolution, 5, 344-352. Bub, H. 2012. Bird trapping and bird banding: a handbook for trapping methods all over the world, Ithaca, NY: Cornell University Press. Buckland, S. T., Anderson, D. R., Burnham, K. P. and Laake, J. L. 2005. Distance Sampling. In: Armitrage, P. and Colton, T. (eds.) Encyclopedia of Biostatistics. Volume 2. DOI: 10.1002/0470011815.b2a16019 Burwell, C. J. & Nakamura, A. 2011. Distribution of ant species along an altitudinal transect in continuous rainforest in subtropical Queensland, Australia. Memoirs of the Queensland Museum, 55, 391-412. Burwell, C. J. & Nakamura, A. 2016. Can changes in ant diversity along elevational gradients in tropical and subtropical Australian rainforests be used to detect a signal of past lowland biotic attrition? Austral Ecology, 41, 209-218. Butler, S. J., Freckleton, R. P., Renwick, A. R. & Norris, K. 2012. An objective, niche-based approach to indicator species selection. Methods in Ecology and Evolution, 3, 317- 326. Buxton, R. T. & Jones, I. L. 2012. Measuring nocturnal seabird activity and status using acoustic recording devices: applications for island restoration. Journal of Field Ornithology, 83, 47-60. Cai, J., Ee, D., Pham, B., Roe, P. & Zhang, J. 2007. Sensor network for the monitoring of ecosystem: bird species recognition. 3rd International Conference on Intelligent Sensors, Sensor Networks and Information, 2007 Melbourne, Australia. IEEE, 293- 298. Campos-Cerqueira, M., Arendt, W. J., Wunderle, J. M. & Aide, T. M. 2017. Have bird distributions shifted along an elevational gradient on a tropical mountain? Ecology and Evolution, 10.1002/ece3.3520.

128

Callaghan, C. T. & Gawlik, D. E. 2015. Efficacy of eBird data as an aid in conservation planning and monitoring. Journal of Field Ornithology, 86, 298-304. Carignan, V. & Villard, M.-A. 2002. Selecting Indicator Species to Monitor Ecological Integrity: A Review. Environmental Monitoring and Assessment, 78, 45-61. Ceballos, G., Ehrlich, P. R., Barnosky, A. D., García, A., Pringle, R. M. & Palmer, T. M. 2015. Accelerated modern human–induced species losses: Entering the sixth mass extinction. Science Advances, 1. Celis-Murillo, A., Deppe, J. L. & Allen, M. F. 2009. Using soundscape recordings to estimate bird species abundance, richness, and composition. Journal of Field Ornithology, 80, 64-78. Chambers, L. E. 2005. Migration dates at Eyre Bird Observatory: links with climate change? Climate Research, 29, 157-165. Chambers, L. E., Hughes, L. & Weston, M. A. 2005. Climate change and its impact on Australia’s avifauna. Emu, 105, 1-20. Chambers, S. A. 2008. Birds as environmental indicators: review of literature, Melbourne, Parks Victoria. Chan, K. 2001. Partial migration in Australian landbirds: a review. Emu, 101, 281-292. Chao, A., Gotelli, N. J., Hsieh, T. C., Sander, E. L., Ma, K. H., Colwell, R. K. & Ellison, A. M. 2014. Rarefaction and extrapolation with Hill numbers: a framework for sampling and estimation in species diversity studies. Ecological Monographs, 84, 45-67. Charmantier, A., Mccleery, R. H., Cole, L. R., Perrins, C., Kruuk, L. E. B. & Sheldon, B. C. 2008. Adaptive phenotypic plasticity in response to climate change in a wild bird population. Science, 320, 800-803. Chen, I.-C., Hill, J. K., Ohlemüller, R., Roy, D. B. & Thomas, C. D. 2011. Rapid range shifts of species associated with high levels of climate warming. Science, 333, 1024-1026. Chown, S. L., Hoffmann, A. A., Kristensen, T. N., Angilletta, M. J., Stenseth, N. C. & Pertoldi, C. 2010. Adapting to climate change: a perspective from evolutionary physiology. Climate Research, 43, 3-15. Christidis, L. & Boles, W. 2008. Systematics and Taxonomy of Australian Birds. Melbourne, CSIRO Publishing. Clark, C. J. 2017. eBird records show substantial growth of the Allen's Hummingbird (Selasphorus sasin sedentarius) population in urban Southern California. The Condor, 119, 122-130.

129

Clarke, K. R. & Gorley, R. N. 2006. PRIMER v6: User Manual/Tutorial, Plymouth, Primer-E Ltd. Colwell, R. K., Gotelli, N. J., Ashton, L. A., Beck, J., Brehm, G., Fayle, T. M., Fiedler, K., Forister, M. L., Kessler, M., Kitching, R. L., Klimes, P., Kluge, J., Longino, J. T., Maunsell, S. C., Mccain, C. M., Moses, J., Noben, S., Sam, K., Sam, L., Shapiro, A. M., Wang, X. & Novotny, V. 2016. Midpoint attractors and species richness: Modelling the interaction between environmental drivers and geometric constraints. Ecology Letters, 19, 1009-1022. Colwell, R. K. & Lees, D. C. 2000. The mid-domain effect: geometric constraints on the geography of species richness. Trends in Ecology and Evolution, 15, 70-76. Crutzen, P. J. 2006. The “Anthropocene”. In: Ehlers, E. & Krafft, T. (eds.) Earth System Science in the Anthropocene: Emerging Issues and Problems. pp. 13-18 Berlin: Springer. CSIRO and Bureau of Meteorology. 2016. Climate change in Australia [Online]. Available: www.climatechangeinaustralia.gov.au [Accessed August 2017]. Dale, V. H. & Beyeler, S. C. 2001. Challenges in the development and use of ecological indicators. Ecological Indicators, 1, 3-10. Darras, K., Furnas, B., Fitriawan, I., Mulyani, Y. & Tscharntke, T. 2018. Estimating bird detection distances in sound recordings for standardising detection ranges and distance sampling. Methods in Ecology and Evolution, doi:10.1111/2041- 210X.13031. De Cáceres, M. & Legendre, P. 2009. Associations between species and groups of sites: indices and statistical inference. Ecology, 90, 3566-3574. De Cáceres, M., Legendre, P. & Moretti, M. 2010. Improving indicator species analysis by combining groups of sites. Oikos, 119, 1674-1684. De Cáceres, M., Legendre, P., Wiser, S. K. & Brotons, L. 2012. Using species combinations in indicator value analyses. Methods in Ecology and Evolution, 3, 973-982. Depraetere, M., Pavoine, S., Jiguet, F., Gasc, A., Duvail, S. & Sueur, J. 2012. Monitoring animal diversity using acoustic indices: Implementation in a temperate woodland. Ecological Indicators, 13, 46-54. Diamond, J. M. 1973. Distributional ecology of New Guinea birds. Science, 179, 759-769. Digby, A., Towsey, M., Bell, B. D. & Teal, P. D. 2013. A practical comparison of manual and autonomous methods for acoustic monitoring. Methods in Ecology and Evolution, 4, 675-683.

130

Ding, T.-S., Yuan, H.-W., Geng, S., Lin, Y.-S. & Lee, P.-F. 2005. Energy flux, body size and density in relation to bird species richness along an elevational gradient in Taiwan. Global Ecology and Biogeography, 14, 299-306. Dingle, H. 2004. The Australo-Papuan bird migration system: another consequence of Wallace's Line. Emu, 104, 95-108. Dirzo, R., Young, H. S., Galetti, M., Ceballos, G., Isaac, N. J. B. & Collen, B. 2014. Defaunation in the Anthropocene. Science, 345, 401-406. Dornelas, M., Magurran, A. E., Buckland, S. T., Chao, A., Chazdon, R. L., Colwell, R. K., Curtis, T., Gaston, K. J., Gotelli, N. J., Kosnik, M. A., Mcgill, B., Mccune, J. L., Morlon, H., Mumby, P. J., Øvreås, L., Studeny, A. & Vellend, M. 2013. Quantifying temporal change in biodiversity: challenges and opportunities. Proceedings of the Royal Society B: Biological Sciences, 280. Dossa, G. G. O., Paudel, E., Fujinuma, J., Yu, H., Chutipong, W., Zhang, Y., Paz, S. & Harrison, R. D. 2013. Factors determining forest diversity and biomass on a tropical volcano, Mt. Rinjani, Lombok, Indonesia. PLoS ONE, 8, e67720. Dufrêne, M. & Legendre, P. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs, 67, 345-366. Dunn, A. M. & Weston, M. A. 2008. A review of terrestrial bird atlases of the world and their application. Emu, 108, 42-67. Dunn, P. O. & Winkler, D. W. 2010. Effects of climate change on timing of breeding and reproductive success in birds. In: Møller, A. P., Fiedler, W. & Berthold, P. (eds.) Effects of Climate Change on Birds. pp. 113-128. Oxford: Oxford University Press. Elith, J. & Leathwick, J. R. 2009. Species distribution models: ecological explanation and prediction across space and time. Annual Review of Ecology, Evolution, and Systematics, 40, 677-697. Farnsworth, A. & Russell, R. W. 2007. Monitoring flight calls of migrating birds from an oil platform in the northern Gulf of Mexico. Journal of Field Ornithology, 78, 279-289. Ferger, S. W., Schleuning, M., Hemp, A., Howell, K. M. & Böhning-Gaese, K. 2014. Food resources and vegetation structure mediate climatic effects on species richness of birds. Global Ecology and Biogeography, 23, 541-549. Fjeldså, J., Bowie, R. C. K. & Rahbek, C. 2012. The role of mountain ranges in the diversification of birds. Annual Review of Ecology, Evolution, and Systematics, 43, 249-265.

131

Ford, H. A. 2013. Are we underestimating the threat to Australia's migratory land birds? Pacific Conservation Biology, 19, 303-311. Forero-Medina, G., Terborgh, J., Socolar, S. J. & Pimm, S. L. 2011. Elevational ranges of birds on a tropical montane gradient lag behind warming temperatures. PLoS ONE, 6, e28535. Foster, P. 2001. The potential negative impacts of global climate change on tropical montane cloud forests. Earth-Science Reviews, 55, 73-106. Freeman, B. G. & Class Freeman, A. M. 2014. Rapid upslope shifts in New Guinean birds illustrate strong distributional responses of tropical montane species to global warming. Proceedings of the National Academy of Sciences, 111, 4490-4494. Fretwell, P. T., Larue, M. A., Morin, P., Kooyman, G. L., Wienecke, B., Ratcliffe, N., Fox, A. J., Fleming, A. H., Porter, C. & Trathan, P. N. 2012. An Emperor Penguin population estimate: the first global, synoptic survey of a species from space. PLoS ONE, 7, e33751. Frommolt, K.-H. & Tauchert, K.-H. 2014. Applying bioacoustic methods for long-term monitoring of a nocturnal wetland bird. Ecological Informatics, 21, 4-12. Gage, S. H. & Axel, A. C. 2014. Visualization of temporal change in soundscape power of a Michigan lake habitat over a 4-year period. Ecological Informatics, 21, 100-109. Garcia, R. A., Cabeza, M., Rahbek, C. & Araújo, M. B. 2014. Multiple dimensions of climate change and their implications for biodiversity. Science, 344. Gardner, J. L., Heinsohn, R. & Joseph, L. 2009. Shifting latitudinal clines in avian body size correlate with global warming in Australian passerines. Proceedings of the Royal Society B: Biological Sciences. Garnett, S., Szabo, J. & Dutson, G. 2011. The action plan for Australian birds 2010, Melbourne, Victoria: CSIRO Publishing. Garnett, S. & Franklin, D. (eds.) 2014. Climate Change Adaptation Plan for Australian Birds, Collingwood, Victoria: CSIRO Publishing. Gasc, A., Sueur, J., Jiguet, F., Devictor, V., Grandcolas, P., Burrow, C., Depraetere, M. & Pavoine, S. 2013. Assessing biodiversity with sound: Do acoustic diversity indices reflect phylogenetic and functional diversities of bird communities? Ecological Indicators, 25, 279-287. Gibbs, H. M., Chambers, L. E. & Bennett, A. F. 2011. Temporal and spatial variability of breeding in Australian birds and the potential implications of climate change. Emu, 111, 283-291.

132

Goyette, J. L., Howe, R. W., Wolf, A. T. & Robinson, W. D. 2011. Detecting tropical nocturnal birds using automated audio recordings. Journal of Field Ornithology, 82, 279-287. Grava, T., Mathevon, N., Place, E. & Balluet, P. 2008. Individual acoustic monitoring of the European Eagle Owl Bubo bubo. Ibis, 150, 279-287. Gray, M. A., Baldauf, S. L., Mayhew, P. J. & Hill, J. K. 2007. The response of avian feeding guilds to tropical forest disturbance. Conservation Biology, 21, 133-141. Green, K. & Pickering, C. M. 2002. A potential scenario for mammal and bird diversity in the Snowy Mountains of Australia in relation to climate change. In: Körner, C. & Spehn, E. M. (eds.) Mountain Biodiversity: A Global Assessment. pp. 239-247. London: Parthenon Publishing. Gregory, R. 2006. Birds as biodiversity indicators for Europe. Significance, 3, 106-110. Griffioen, P. A. & Clarke, M. F. 2002. Large-scale bird-movement patterns evident in eastern Australian atlas data. Emu, 102, 99-125. Griffith University. 2017. Inclusion of papers within the thesis. [Online]. Griffith University. Available: https://www.griffith.edu.au/higher-degrees-research/current-research- students/thesis/preparation/inclusion-of-papers-within-the-thesis [Accessed September 2017]. Hall, J., Burgess, N. D., Lovett, J., Mbilinyi, B. & Gereau, R. E. 2009. Conservation implications of deforestation across an elevational gradient in the Eastern Arc Mountains, Tanzania. Biological Conservation, 142, 2510-2521. Hannah, L., Flint, L., Syphard, A. D., Moritz, M. A., Buckley, L. B. & Mccullough, I. M. 2014. Fine-grain modeling of species’ response to climate change: holdouts, stepping-stones, and microrefugia. Trends in Ecology & Evolution, 29, 390-397.

Harris, J. B. C., Dwi Putra, D., Gregory, S. D., Brook, B. W., Prawiradilaga, D. M., Sodhi, N. S., Wei, D. & Fordham, D. A. 2014. Rapid deforestation threatens mid-elevational endemic birds but climate change is most important at higher elevations. Diversity and Distributions, 20, 773-785. Harris, S. A., Shears, N. T. & Radford, C. A. 2016. Ecoacoustic indices as proxies for biodiversity on temperate reefs. Methods in Ecology and Evolution, 7, 713-724. Harrison, X. A. 2014. Using observation-level random effects to model overdispersion in count data in ecology and evolution. PeerJ, 2, e616.

133

Hartig, F. 2017. DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. R package version 0.1.5. https://CRAN.R- project.org/package=DHARMa. Haselmayer, J. & Quinn, J. S. 2000. A comparison of point counts and sound recording as bird survey methods in Amazonian southeast Peru. The Condor, 102, 887-893. Herrando, S., Brotons, L., Anton, M., Páramo, F., Villero, D., Titeux, N., Quesada, J. & Stefanescu, C. 2016. Assessing impacts of land abandonment on Mediterranean biodiversity using indicators based on bird and butterfly monitoring data. Environmental Conservation, 43, 69-78. Herzog, S. K., Kessler, M. & Bach, K. 2005. The elevational gradient in Andean bird species richness at the local scale: a foothill peak and a high-elevation plateau. Ecography, 28, 209-222. Higgins, P. J. (ed.) 1999. Handbook of Australian, New Zealand, and Antarctic Birds. Volume 4: Parrots to Dollarbird, Melbourne, Australia: Oxford University Press. Higgins, P. J. & Davies, S. J. J. F. (eds.) 1996. Handbook of Australian, New Zealand and Antarctic Birds. Volume 3: Snipe to Pigeons, Melbourne: Oxford University Press. Higgins, P. J. & Peter, J. M. (eds.) 2002. Handbook of Australian, New Zealand and Antarctic Birds. Volume 6: Pardalotes to Shrike-thrushes, Melbourne: Oxford University Press. Higgins, P. J., Peter, J. M. & Cowling, S. J. (eds.) 2006. Handbook of Australian, New Zealand and Antarctic Birds. Volume 7: Boatbill to Starlings, Melbourne: Oxford University Press. Higgins, P. J., Peter, J. M. & Steele, W. K. (eds.) 2001. Handbook of Australian, New Zealand and Antarctic Birds. Volume 5: Tyrant-flycatchers to Chats, Melbourne: Oxford University Press. Hilbert, D. W., Bradford, M., Parker, T. & Westcott, D. A. 2004. Golden bowerbird (Prionodura newtonia) habitat in past, present and future climates: predicted extinction of a vertebrate in tropical highlands due to global warming. Biological Conservation, 116, 367-377. Hobbs, R. J., Higgs, E. & Harris, J. A. 2009. Novel ecosystems: implications for conservation and restoration. Trends in Ecology & Evolution, 24, 599-605. Hobson, K. A., Rempel, R. S., Hamilton, G., Turnbull, B. & Van Wilgenburg, S. L. 2002. Acoustic surveys of birds using electronic recordings: new potential from an omnidirectional microphone system. Wildlife Society Bulletin, 30, 709-720.

134

Hoffmann, A. A. & Sgrò, C. M. 2011. Climate change and evolutionary adaptation. Nature, 470, 479. Holmes, M. W., Hammond, T. T., Wogan, G. O. U., Walsh, R. E., Labarbera, K., Wommack, E. A., Martins, F. M., Crawford, J. C., Mack, K. L., Bloch, L. M. & Nachman, M. W. 2016. Natural history collections as windows on evolutionary processes. Molecular Ecology, 25, 864-881. Holmes, S. B., Mcilwrick, K. A. & Venier, L. A. 2014. Using automated sound recording and analysis to detect bird species-at-risk in southwestern Ontario woodlands. Wildlife Society Bulletin, 38, 591-598. Howard, C., Stephens, P. A., Pearce-Higgins, J. W., Gregory, R. D. & Willis, S. G. 2015. The drivers of avian abundance: patterns in the relative importance of climate and land use. Global Ecology and Biogeography, 24, 1249-1260. Hsieh, T. C., Ma, K. H. & Chao, A. 2016. iNEXT: an R package for rarefaction and extrapolation of species diversity (Hill numbers). Methods in Ecology and Evolution, 7, 1451-1456. Hui, F. K. C. 2016. boral – Bayesian ordination and regression analysis of multivariate abundance data in r. Methods in Ecology and Evolution, 7, 744-750. Hurrell, J. W. & Trenberth, K. E. 2010. Climate Change. In: Møller, A. P., Fiedler, W. & Berthold, P. (eds.) Effects of Climate Change on Birds. pp. 9-29. Oxford: Oxford University Press. Ives, A. R. & Helmus, M. R. 2011. Generalized linear mixed models for phylogenetic analyses of community structure. Ecological Monographs, 81, 511-525. Janes, S. W. 1994. Variation in the species composition and mean body size of an avian foliage-gleaning guild along an elevational gradient: correlation with arthropod body size. Oecologia, 98, 369-378. Jamil, T., Opdekamp, W., Van Diggelen, R. & Ter Braak, C. J. 2012. Trait-environment relationships and tiered forward model selection in linear mixed models. International Journal of Ecology, 2012. Jamil, T., Ozinga, W. A., Kleyer, M. & Ter Braak, C. J. F. 2013. Selecting traits that explain species–environment relationships: a generalized linear mixed model approach. Journal of Vegetation Science, 24, 988-1000. Jankowski, J. E., Merkord, C. L., Rios, W. F., Cabrera, K. G., Revilla, N. S. & Silman, M. R. 2013. The relationship of tropical bird communities to tree species composition and

135

vegetation structure along an Andean elevational gradient. Journal of Biogeography, 40, 950-962. Jankowski, J. E., Robinson, S. K. & Levey, D. J. 2010. Squeezed at the top: Interspecific aggression may constrain elevational ranges in tropical birds. Ecology, 91, 1877-1884. Jenkins, C. N., Pimm, S. L. & Joppa, L. N. 2013. Global patterns of terrestrial vertebrate diversity and conservation. Proceedings of the National Academy of Sciences, 110, E2602-E2610. Jones, K. R., Venter, O., Fuller, R. A., Allan, J. R., Maxwell, S. L., Negret, P. J. & Watson, J. E. M. 2018. One-third of global protected land is under intense human pressure. Science, 360, 788-791. Joshi, K. A., Mulder, R. A. & Rowe, K. M. C. 2017. Comparing manual and automated species recognition in the detection of four common south-east Australian forest birds from digital field recordings. Emu, 117, 233-246. Joshi, K. K., Bhatt, D. & Thapliyal, A. 2012. Avian diversity and its association with vegetation structure in different elevational zones of Nainital district (Western Himalayan) of Uttarakhand. International Journal of Biodiversity and Conservation, 4, 364-370. Kassambara, A. 2017. ggpubr: 'ggplot2' Based Publication Ready Plots. R package version 0.1.4. https://CRAN.R-project.org/package=ggpubr. Katuwal, H. B., Basnet, K., Khanal, B., Devkota, S., Rai, S. K., Gajurel, J. P., Scheidegger, C. & Nobis, M. P. 2016. Seasonal changes in bird species and feeding guilds along elevational gradients of the central Himalayas, Nepal. PLoS ONE, 11, e0158362. Kirchman, J. J. & van Keuren, A. E. 2017. Altitudinal range shifts of birds at the southern periphery of the boreal forest: 40 years of change in the Adirondack mountains. The Wilson Journal of Ornithology, 129, 742-753. Kitching, R. L. & Ashton, L. A. 2013. Predictor sets and biodiversity assessment: the evolution and application of an idea. Pacific Conservation Biology, 19, 418-426. Kitching, R. L., Braithwaite, R. & Cavanaugh, J. 2010a. The Gondwana Rainforests of Australia – Introduction. In: Kitching, R. L., Braithwaite, R. & Cavanaugh, J. (eds.) Remnants of Gondwana: A Natural and Social History of the Gondwana Rainforests of Australia. pp. 1-6. Baulkham Hills, New South Wales: Surrey Beatty & Sons. Kitching, R. L., Braithwaite, R. & Cavanaugh, J. (eds.) 2010b. Remnants of Gondwana: A Natural and Social History of the Gondwana Rainforests of Australia, Baulkham Hills, New South Wales: Surrey Beatty & Sons.

136

Kitching, R. L., Putland, D., Ashton, L. A., Laidlaw, M. J., Boulter, S., Christensen, H. S. & Lambkin, C. L. 2011. Detecting biodiversity changes along climatic gradients: the IBISCA-Queensland Project. Memoirs of the Queensland Museum: Nature, 55, 235- 250. Klingbeil, B. T. & Willig, M. R. 2015. Bird biodiversity assessments in temperate forest: the value of point count versus acoustic monitoring protocols. PeerJ, 3, e973. Körner, C. 2007. The use of ‘altitude’ in ecological research. Trends in Ecology & Evolution, 22, 569-574. Kraft, N. J. B., Comita, L. S., Chase, J. M., Sanders, N. J., Swenson, N. G., Crist, T. O., Stegen, J. C., Vellend, M., Boyle, B., Anderson, M. J., Cornell, H. V., Davies, K. F., Freestone, A. L., Inouye, B. D., Harrison, S. P. & Myers, J. A. 2011. Disentangling the drivers of β diversity along latitudinal and elevational gradients. Science, 333, 1755-1758. Krause, B. & Farina, A. 2016. Using ecoacoustic methods to survey the impacts of climate change on biodiversity. Biological Conservation, 195, 245-254. Laidlaw, M. J., Mcdonald, W. J., Hunter, R. J. & Kitching, R. L. 2011a. Subtropical rainforest turnover along an altitudinal gradient. Memoirs of the Queensland Museum, 55, 271-290. Laidlaw, M. J., Mcdonald, W. J. F., Hunter, R. J., Putland, D. A. & Kitching, R. L. 2011b. The potential impacts of climate change on Australian subtropical rainforest. Australian Journal of Botany, 59, 440-449. Laiolo, P. 2010. The emerging significance of bioacoustics in animal species conservation. Biological Conservation, 143, 1635-1645. Laiolo, P., Vögeli, M., Serrano, D. & Tella, J. L. 2007. Testing acoustic versus physical marking: two complementary methods for individual-based monitoring of elusive species. Journal of Avian Biology, 38, 672-681. Lambert, K. T. & Mcdonald, P. G. 2014. A low-cost, yet simple and highly repeatable system for acoustically surveying cryptic species. Austral Ecology, 39, 779 - 785. Lambkin, C. L., Boulter, S. L., Starick, N. T., Cantrell, B. K., Bickel, D. J., Wright, S. G., Power, N., Schutze, M. K., Turco, F. & Nakamura, A. 2011. Altitudinal and seasonal variation in the family-level assemblages of flies (Diptera) in an Australian subtropical rainforest: one hundred thousand and counting! Memoirs of the Queensland Museum, 55, 315-331.

137

Larsen, F. W., Bladt, J. & Rahbek, C. 2007. Improving the Performance of Indicator Groups for the Identification of Important Areas for Species Conservation. Conservation Biology, 21, 731-740. Laurance, W. F., Carolina Useche, D., Shoo, L. P., Herzog, S. K., Kessler, M., Escobar, F., Brehm, G., Axmacher, J. C., Chen, I. C., Gámez, L. A., Hietz, P., Fiedler, K., Pyrcz, T., Wolf, J., Merkord, C. L., Cardelus, C., Marshall, A. R., Ah-Peng, C., Aplet, G. H., Del Coro Arizmendi, M., Baker, W. J., Barone, J., Brühl, C. A., Bussmann, R. W., Cicuzza, D., Eilu, G., Favila, M. E., Hemp, A., Hemp, C., Homeier, J., Hurtado, J., Jankowski, J., Kattán, G., Kluge, J., Krömer, T., Lees, D. C., Lehnert, M., Longino, J. T., Lovett, J., Martin, P. H., Patterson, B. D., Pearson, R. G., Peh, K. S. H., Richardson, B., Richardson, M., Samways, M. J., Senbeta, F., Smith, T. B., Utteridge, T. M. A., Watkins, J. E., Wilson, R., Williams, S. E. & Thomas, C. D. 2011a. Global warming, elevational ranges and the vulnerability of tropical biota. Biological Conservation, 144, 548-557. Laurance, W. F., Dell, B., Turton, S. M., Lawes, M. J., Hutley, L. B., Mccallum, H., Dale, P., Bird, M., Hardy, G., Prideaux, G., Gawne, B., Mcmahon, C. R., Yu, R., Hero, J.-M., Schwarzkopf, L., Krockenberger, A., Douglas, M., Silvester, E., Mahony, M., Vella, K., Saikia, U., Wahren, C.-H., Xu, Z., Smith, B. & Cocklin, C. 2011b. The 10 Australian ecosystems most vulnerable to tipping points. Biological Conservation, 144, 1472-1480. Leach, E. C. 2016. Notes on the Zoothera thrushes in the Tweed Range of north-eastern New South Wales. Australian Field Ornithology, 33. Leach, E. C., Burwell, C. J., Ashton, L. A., Jones, D. N. & Kitching, R. L. 2016. Comparison of point counts and automated acoustic monitoring: detecting birds in a rainforest biodiversity survey. Emu, 116, 305-309. Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018a. Modelling the responses of Australian subtropical rainforest birds to changes in environmental conditions along elevational gradients. Austral Ecology, https://doi.org/10.1111/aec.12586 Leach, E. C., Burwell, C. J., Jones, D. N. & Kitching, R. L. 2018b. Identifying avian indicators of elevation in the Gondwanan Rainforests of Australia. Pacific Conservation Biology, https://doi.org/10.1071/PC18039 Legendre, P., Galzin, R. & Harmelin-Vivien, M. L. 1997. Relating behaviour to habitat: solutions to the fourth-corner problem. Ecology, 78, 547-562.

138

Legge, S. 2015. A plea for inserting evidence-based management into conservation practice. Animal Conservation, 18, 113-116. Li, D. & Ives, A. R. 2017. The statistical need to include phylogeny in trait-based analyses of community composition. Methods in Ecology and Evolution, 10.1111/2041- 210X.12767. Lindström, Å., Green, M., Paulson, G., Smith, H. G. & Devictor, V. 2013. Rapid changes in bird community composition at multiple temporal and spatial scales in response to recent climate change. Ecography, 36, 313-322. Loiselle, B. A. & Blake, J. G. 1991. Temporal variation in birds and fruits along an elevational gradient in Costa Rica. Ecology, 72, 180-193. Lomolino, M. V. 2001. Elevation gradients of species-density: historical and prospective views. Global Ecology and Biogeography, 10, 3-13. Londoño, G. A., Chappell, M. A., Jankowski, J. E. & Robinson, S. K. 2017. Do thermoregulatory costs limit altitude distributions of Andean forest birds? Functional Ecology, 31, 204-215. Mackey, B. G., Watson, J. E. M., Hope, G. & Gilmore, S. 2008. Climate change, biodiversity conservation, and the role of protected areas: An Australian perspective. Biodiversity, 9, 11-18. Magurran, A. E., Baillie, S. R., Buckland, S. T., Dick, J. M., Elston, D. A., Scott, E. M., Smith, R. I., Somerfield, P. J. & Watt, A. D. 2010. Long-term datasets in biodiversity research and monitoring: assessing change in ecological communities through time. Trends in Ecology & Evolution, 25, 574-582. Malhi, Y., Silman, M., Salinas, N., Bush, M., Meir, P. & Saatchi, S. 2010. Introduction: Elevation gradients in the tropics: laboratories for ecosystem ecology and global change research. Global Change Biology, 16, 3171-3175. Mammides, C., Goodale, E., Dayananda, S. K., Kang, L. & Chen, J. 2017. Do acoustic indices correlate with bird diversity? Insights from two biodiverse regions in Yunnan Province, south China. Ecological Indicators, 82, 470-477. Marchant, S. & Higgins, P. J. (eds.) 1990. Handbook of Australian, New Zealand and Antarctic Birds. Volume 1: Ratites to Ducks, Melbourne: Oxford University Press. Marchant, S. & Higgins, P. J. (eds.) 1993. Handbook of Australian, New Zealand and Antarctic Birds. Volume 3: Raptors to Lapwings, Melbourne: Oxford University Press.

139

Martin, C. A. & Proulx, R. 2016. Habitat geometry, a step toward general bird community assembly rules in mature forests. Forest Ecology and Management, 361, 163-169. Maunsell, S. C., Kitching, R. L., Greenslade, P., Nakamura, A. & Burwell, C. J. 2013. Springtail (Collembola) assemblages along an elevational gradient in Australian subtropical rainforest. Australian Journal of Entomology, 52, 114-124. McCain, C. M. 2009. Global analysis of bird elevational diversity. Global Ecology and Biogeography, 18, 346-360. McCain, C. M. & Colwell, R. K. 2011. Assessing the threat to montane biodiversity from discordant shifts in temperature and precipitation in a changing climate. Ecology Letters, 14, 1236-1245. McDonald, B. 2010. The Tweed Caldera Group. In: Kitching, R. L., Braithwaite, R. & Cavanaugh, J. (eds.) Remnants of Gondwana: A Natural and Social History of the Gondwana Rainforests of Australia. pp. 39-50. Baulkham Hills, New South Wales, NSW: Surrey Beatty & Sons. McGill, B. J., Enquist, B. J., Weiher, E. & Westoby, M. 2006. Rebuilding community ecology from functional traits. Trends in Ecology & Evolution, 21, 178-185. McGuire, A., Johnston, G., Robertson, J. & Kleindorfer, S. 2011. Comparison of survey methods for detection of the elusive Western Whipbird Psophodes nigrogularis with notes on its distribution. South Australian Ornithologist, 37, 49-59. Mennill, D. J. 2011. Individual distinctiveness in avian vocalizations and the spatial monitoring of behaviour. Ibis, 153, 235-238. Mennill, D. J., Battiston, M., Wilson, D. R., Foote, J. R. & Doucet, S. M. 2012. Field test of an affordable, portable, wireless microphone array for spatial monitoring of animal ecology and behaviour. Methods in Ecology and Evolution, 3, 704-712. Metcalfe, P. 2010. The Birds. In: Kitching, R. L., Braithwaite, R. & Cavanaugh, J. (eds.) Remnants of Gondwana: A Natural and Social History of the Gondwana Rainforests of Australia. pp. 163-172. Baulkham Hills, New South Wales: Surrey Beatty & Sons. Murtaugh, P. A. 2007. Simplicity and complexity in ecological data analysis. Ecology, 88, 56-62. Nakamura, A., Burwell, C. J., Ashton, L. A., Laidlaw, M. J., Katabuchi, M. & Kitching, R. L. 2016. Identifying indicator species of elevation: Comparing the utility of woody plants, ants and moths for long-term monitoring. Austral Ecology, 41, 179-188.

140

Niku, J., Warton, D. I., Hui, F. K. C. & Taskinen, S. 2017. Generalized linear latent variable models for multivariate count and biomass data in ecology. Journal of Agricultural, Biological and Environmental Statistics, 22, 498-522. Odell, E. H., Ashton, L. A. & Kitching, R. L. 2015. Elevation and moths in a central eastern Queensland rainforest. Austral Ecology, 41, 133-144. Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., Mcglinn, D., Minchin, P. R., O'hara, R. B., Simpson, G. L., Solymos, P., Henry, M., Stevens, H., Szoecs, E. & Wagner, H. H. 2016. vegan: Community Ecology Package. https://CRAN.R- project.org/package=vegan Orme, C. D. L., Davies, R. G., Burgess, M., Eigenbrod, F., Pickup, N., Olson, V. A., Webster, A. J., Ding, T.-S., Rasmussen, P. C., Ridgely, R. S., Stattersfield, A. J., Bennett, P. M., Blackburn, T. M., Gaston, K. J. & Owens, I. P. F. 2005. Global hotspots of species richness are not congruent with endemism or threat. Nature, 436, 1016+. Osmun, A. E. & Mennill, D. J. 2011. Acoustic monitoring reveals congruent patterns of territorial singing behaviour in male and female tropical wrens. Ethology, 117, 385- 394. Parmesan, C. 2006. Ecological and evolutionary responses to recent climate change. Annual Review of Ecology, Evolution, and Systematics, 37, 637-669. Parmesan, C. & Yohe, G. 2003. A globally coherent fingerprint of climate change impacts across natural systems. Nature, 421, 37-42. Parsons, D. F., Suthers, I. M., Cruz, D. O. & Smith, J. A. 2016. Effects of habitat on fish abundance and species composition on temperate rocky reefs. Marine Ecology Progress Series, 561, 155-171. Pauchard, A., Milbau, A., Albihn, A., Alexander, J., Burgess, T., Daehler, C., Englund, G., Essl, F., Evengård, B., Greenwood, G. B., Haider, S., Lenoir, J., Mcdougall, K., Muths, E., Nuñez, M. A., Olofsson, J., Pellissier, L., Rabitsch, W., Rew, L. J., Robertson, M., Sanders, N. & Kueffer, C. 2016. Non-native and native organisms moving into high elevation and high latitude ecosystems in an era of climate change: new challenges for ecology and conservation. Biological Invasions, 18, 345-353. Pieretti, N., Farina, A. & Morri, D. 2011. A new methodology to infer the singing activity of an avian community: the Acoustic Complexity Index (ACI). Ecological Indicators, 11, 868-873.

141

Pollock, L. J., Morris, W. K. & Vesk, P. A. 2012. The role of functional traits in species distributions revealed through a hierarchical model. Ecography, 35, 716-725. Pulsford, I., Howling, G., Dunn, R. & Crane, R. 2013. The Great Eastern Ranges Initiative: a continental scale lifeline connecting people and nature. In: Fitzsimons, J., Pulsford, I. & Wescott, G. (eds.) Linking Australia’s Landscapes: Lessons and opportunities from large-scale conservation networks. pp. 123-134. Collingwood, Victoria: CSIRO Publishing. R Core Team. 2017. R: A language and environment for statistical computing. http://www.R- project.org/. Rahbek, C. 1995. The elevational gradient of species richness: a uniform pattern? Ecography, 18, 200-205. Ratcliffe, N., Guihen, D., Robst, J., Crofts, S., Stanworth, A. & Enderlein, P. 2015. A protocol for the aerial survey of penguin colonies using UAVs. Journal of Unmanned Vehicle Systems, 3, 95-101. Rempel, R. S., Francis, C. M., Robinson, J. N. & Campbell, M. 2013. Comparison of audio recording system performance for detecting and monitoring songbirds. Journal of Field Ornithology, 84, 86-97. Rempel, R. S., Hobson, K. A., Holborn, G., Van Wilgenburg, S. L. & Elliott, J. 2005. Bioacoustic monitoring of forest songbirds: interpreter variability and effects of configuration and digital processing methods in the laboratory. Journal of Field Ornithology, 76, 1-11. Reside, A. E., Vanderwal, J., Phillips, B. L., Shoo, L. P., Rosauer, D. F., Anderson, B. J., Welbergen, J. A., Moritz, C., Ferrier, S., Harwood, T. D., Williams, K. J., Mackey, B. G., Hugh, S. & Williams, S. E. 2013. Climate change refugia for terrestrial biodiversity: Defining areas that promote species persistence and ecosystem resilience in the face of global climate change Gold Coast, National Climate Change Adaptation Research Facility. Root, T. L., Price, J. T., Hall, K. R., Schneider, S. H., Rosenzweig, C. & Pounds, J. A. 2003. Fingerprints of global warming on wild animals and plants. Nature, 421, 57-60. Ross, J. C. & Allen, P. E. 2014. Random Forest for improved analysis efficiency in passive acoustic monitoring. Ecological Informatics, 21, 34-39. Roth, T., Plattner, M. & Amrhein, V. 2014. Plants, Birds and Butterflies: Short-Term Responses of Species Communities to Climate Warming Vary by Taxon and with Altitude. PLoS ONE, 9, e82490.

142

Ruggiero, A. & Hawkins, B. A. 2008. Why do mountains support so many species of birds? Ecography, 31, 306-315. Scheffers, B. R., Shoo, L., Phillips, B., Macdonald, S. L., Anderson, A., Vanderwal, J., Storlie, C., Gourret, A. & Williams, S. E. 2017. Vertical (arboreality) and horizontal (dispersal) movement increase the resilience of vertebrates to climatic instability. Global Ecology and Biogeography, 26, 787-798. Sedláček, O., Vokurková, J., Ferenc, M., Djomo, E. N., Albrecht, T. & Hořák, D. 2015. A comparison of point counts with a new acoustic sampling method: a case study of a bird community from the montane forests of Mount Cameroon. Ostrich, 1-8. Şekercioğlu, Ç. H., Schneider, S. H., Fay, J. P. & Loarie, S. R. 2008. Climate change, elevational range shifts, and bird extinctions. Conservation Biology, 22, 140-150. Sheridan, J. A. & Bickford, D. 2011. Shrinking body size as an ecological response to climate change. Nature climate change, 1, 401. Shiu, H.-J. & Lee, P.-F. 2003. Seasonal variation in bird species richness along elevational gradients in Taiwan. Acta Zoologica Taiwanica, 14, 1-21. Shonfield, J. & Bayne, E. M. 2017. Autonomous recording units in avian ecological research: current use and future applications. Avian Conservation and Ecology, 12(1), 14. Shoo, L. P., O’mara, J., Perhans, K., Rhodes, J. R., Runting, R. K., Schmidt, S., Traill, L. W., Weber, L. C., Wilson, K. A. & Lovelock, C. E. 2014. Moving beyond the conceptual: specificity in regional climate change adaptation actions for biodiversity in South East Queensland, Australia. Regional Environmental Change, 14, 435-447. Shoo, L. P., Williams, S. E. & Hero, J.-M. 2005. Climate warming and the rainforest birds of the Australian Wet Tropics: using abundance data as a sensitive predictor of change in total population size. Biological Conservation, 125, 335-343. Shoo, L. P., Williams, S. E. & Hero, J.-M. 2006. Detecting climate change induced range shifts: where and how should we be looking? Austral Ecology, 31, 22-29. Sitters, H., York, A., Swan, M., Christie, F. & Di Stefano, J. 2016. Opposing responses of bird functional diversity to vegetation structural diversity in wet and dry forest. PLoS ONE, 11, e0164917. Sodhi, N. S. 2002. A comparison of bird communities of two fragmented and two continuous southeast Asian rainforests. Biodiversity & Conservation, 11, 1105-1119. Soulé, M. E., Mackey, B. G., Recher, H. F., Williams, J. E., Woinarski, J. C. Z., Driscoll, D., Dennison, W. C. & Jones, M. E. 2004. The role of connectivity in Australian conservation. Pacific Conservation Biology, 10, 266-279.

143

Steel, E. A., Kennedy, M. C., Cunningham, P. G. & Stanovick, J. S. 2013. Applied statistics in ecology: common pitfalls and simple solutions. Ecosphere, 4, 1-13. Strong, C. L., Boulter, S. L., Laidlaw, M. J., Maunsell, S. C., Putland, D. & Kitching, R. L. 2011. The physical environment of an altitudinal gradient in the rainforest of Lamington National Park, southeast Queensland. Memoirs of the Queensland Museum: Nature, 55, 251-270. Sueur, J., Farina, A., Gasc, A., Pieretti, N. & Pavoine, S. 2014. Acoustic indices for biodiversity assessment and landscape investigation. Acta Acustica united with Acustica, 100, 772-781. Sundqvist, M. K., Sanders, N. J. & Wardle, D. A. 2013. Community and ecosystem responses to elevational gradients: processes, mechanisms, and insights for global change. Annual Review of Ecology, Evolution, and Systematics, 44, 261-280. Teplitsky, C., Mills, J. A., Alho, J. S., Yarrall, J. W. & Merilä, J. 2008. Bergmann's rule and climate change revisited: Disentangling environmental and genetic responses in a wild bird population. Proceedings of the National Academy of Sciences, 105, 13492-13496. Terborgh, J. 1977. Bird species diversity on an Andean elevational gradient. Ecology, 58, 1007-1019. Thomas, L. & Marques, T. A. 2012. Passive acoustic monitoring for estimating animal density. Acoustics Today, 8, 35-44. Tingley, M. W., Koo, M. S., Moritz, C., Rush, A. C. & Beissinger, S. R. 2012. The push and pull of climate change causes heterogeneous shifts in avian elevational ranges. Global Change Biology, 18, 3279-3290. Towsey, M., Wimmer, J., Williamson, I. & Roe, P. 2014. The use of acoustic indices to determine avian species richness in audio-recordings of the environment. Ecological Informatics, 21, 110-119. Tracey, J. G. 1982. The vegetation of the humid tropical region of north Queensland, Melbourne, CSIRO. Trindade-Filho, J., Sobral, F. L., Cianciaruso, M. V. & Loyola, R. D. 2012. Using indicator groups to represent bird phylogenetic and functional diversity. Biological Conservation, 146, 155-162. Urban, M. C. 2015. Accelerating extinction risk from climate change. Science, 348, 571-573. Van Rensburg, B. J., Mcgeoch, M. A., Chown, S. L. & Van Jaarsveld, A. S. 1999. Conservation of heterogeneity among dung beetles in the Maputaland Centre of Endemism, South Africa. Biological Conservation, 88, 145-153.

144

Vandewalle, M., De Bello, F., Berg, M. P., Bolger, T., Dolédec, S., Dubs, F., Feld, C. K., Harrington, R., Harrison, P. A. & Lavorel, S. 2010. Functional traits as indicators of biodiversity response to land use changes across ecosystems and organisms. Biodiversity and Conservation, 19, 2921-2947. Venier, L. A., Holmes, S. B., Holborn, G. W., Mcilwrick, K. A. & Brown, G. 2012. Evaluation of an automated recording device for monitoring forest birds. Wildlife Society Bulletin, 36, 30-39. Venier, L. A., Mazerolle, M. J., Rodgers, A., Mcilwrick, K. A., Holmes, S. & Thompson, D. 2017. Comparison of semiautomated bird song recognition with manual detection of recorded bird song samples. Avian Conservation and Ecology, 12. Warton, D. I. 2008. Raw data graphing: an informative but under-utilized tool for the analysis of multivariate abundances. Austral Ecology, 33, 290-300. Warton, D. I., Blanchet, F. G., O’hara, R. B., Ovaskainen, O., Taskinen, S., Walker, S. C. & Hui, F. K. C. 2015a. So many variables: joint modeling in community ecology. Trends in Ecology & Evolution, 30, 766-779. Warton, D. I., Foster, S. D., De’ath, G., Stoklosa, J. & Dunstan, P. K. 2015b. Model-based thinking for community ecology. Plant Ecology, 216, 669-682. Waterhouse, F. L., Mather, M. H., & Seip, D. 2002. Distribution and abundance of birds relative to elevation and biogeoclimatic zones in coastal old-growth forests in southern British Columbia. – British Columbia Journal of Ecosystems and Management, 2, 1-13. Wickham, H. 2016. ggplot2: Elegant Graphics for Data Analysis, New York, Springer International Publishing. Williams, S., Pearson, R. & Walsh, P. 1996. Distributions and biodiversity of the terrestrial vertebrates of Australia's Wet Tropics: a review of current knowledge. Pacific Conservation Biology, 2, 327-362. Williams, G. & Adam, P. 2010. The flowering of Australia's rainforests: a plant and pollination miscellany, Collingwood, Victoria, CSIRO Publishing. Williams, S. E., Bolitho, E. E. & Fox, S. 2003. Climate change in Australian tropical rainforests: an impending environmental catastrophe. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270, 1887-1892. Williams, S. E. & Middleton, J. 2008. Climatic seasonality, resource bottlenecks, and abundance of rainforest birds: implications for global climate change. Diversity and Distributions, 14, 69-77.

145

Williams, S. E., Shoo, L. P., Henriod, R. & Pearson, R. G. 2010. Elevational gradients in species abundance, assemblage structure and energy use of rainforest birds in the Australian Wet Tropics bioregion. Austral Ecology, 35, 650-664. Wimmer, J., Towsey, M., Roe, P. & Williamson, I. 2013. Sampling environmental acoustic recordings to determine bird species richness. Ecological Applications, 23, 1419- 1428. Yeates, D. K. & Monteith, G. B. 2008. The invertebrate fauna of the Wet Tropics: diversity, endemism and relationships. In: Stork, N. & Turton, S. M. (eds.) Living in a Dynamic Tropical Forest Landscape. pp. 178-191. Carlton, Victoria: Blackwell Publishing. Yom-Tov, Y., Yom-Tov, S., Wright, J., J. R. Thorne, C. & Du Feu, R. 2006. Recent changes in body weight and wing length among some British passerine birds. Oikos, 112, 91- 101. Zuur, A. F., Ieno, E. N. & Elphick, C. S. 2010. A protocol for data exploration to avoid common statistical problems. Methods in Ecology and Evolution, 1, 3-14. Zwart, M. C., Baker, A., Mcgowan, P. J. K. & Whittingham, M. J. 2014. The use of automated bioacoustic recorders to replace human wildlife surveys: an example using nightjars. PLoS ONE, 9, e102770.

146