Mammalian diversity patterns: effects of bias and scale

Ben Collen

A thesis submitted for the degree of Doctor of Philosophy at Imperial College London, University of London.

September 2005

Abstract

Abstract In recent years, conservation research has dramatically increased appreciation of the underlying mechanisms of changing biodiversity patterns. More specifically, phylogenetic comparative analyses have provided conservation biologists with a more rigorous tool to explore and understand the underlying processes and patterns of contemporary extinction. I use methods that control for phylogenetic non-independence of taxa, to examine several aspects of mammalian diversity patterns.

The first section addresses a number of issues that may bias our current views of patterns of biodiversity. Species description is ongoing, even among well-studied taxa, so how will currently undiscovered species alter our view of biodiversity? Results from chapter two suggest that within groups, we can make some robust generalisations about what undescribed species may be like.

A key tool for conservation is the assessment of extinction risk. As the impact and use of threatened species lists become wider reaching, the imperative to ensure they are robust, objective and reliable becomes greater. In chapter three I find that Red List assessments among multiple users of varying experience are remarkably consistent.

The final category on the IUCN Red List is ‘Extinct’. Though apparently straightforward, applying the category can be problematic, as the disappearance of the final individual of a species is rarely observed. Conservation science must seek more robust and objective means to classify species for when data are sparse. A potential solution to the problem, which utilises a species sighting record, is tested in chapter four.

The second section of this thesis addresses extinction risk and the effect of scale. Global species extinction typically represents the end point in a long sequence of population declines and local extinctions, but there is still little appreciation of how local processes scale up to global patterns. Regional scale analyses presented in chapters five and six, provide some insight and unlike the majority of global analyses, are more able to account for specific threatening processes which, mediated by species biology, are driving extinction. Making progress in understanding the processes

2 Abstract governing species extinction may require robust generalisations about the relationship of extinction between the population and species level. I synthesise these results in chapter seven, with published analyses of extinction risk and show some that correlates are really quite robust to changes in scale.

3 Declaration

Declaration All the work presented in this thesis is my own, with the following acknowledgements:

Chapter 2 This chapter is published in Global Ecology and Biogeography as a senior authored paper with Andy Purvis and John Gittleman. I was responsible for all aspects of the design, analysis and writing of the paper. The other two authors provided data and thoughtful comments on manuscript structure and clarity.

Chapter 3 The idea for this chapter came from Georgina Mace. Many people at the Global Mammal Assessment, African small mammals workshop, and student groups from Imperial College London and Manchester Metropolitan University completed Red List assessments to enable me to write this chapter. All aspects of the design, analysis and writing are my own.

Chapter 4 Part of the data presented in this chapter was collated whilst working at the University of Sydney in Chris Dickman’s lab. Dave Roberts and Andy Solow answered my incessant questions about the technique used. Dave Orme and Marcus Rowcliffe helped me implement it. Analyses and writing are my own.

Chapter 5 This chapter is to be published in Biodiversity and Conservation in a special issue on extinction. Data for this chapter came from the INTAS funded project 99-1483 from E.J. Milner-Gulland and Stephen Ling. Elena Bykova was instrumental in collating and validating information for the database. Andy Purvis, E.J. Milner-Gulland and Stephen Ling provided comments on the structure and clarity of the final write up – they are all coauthors on the final paper. The design, analysis and write up were my own work.

Chapter 6 Data for this chapter was collated during work with Chris Pavey at the Parks and Wildlife Commission of the Northern Territory, Alice Springs, Australia. Marcel Cardillo generated the species lists from Mammal SuperTeam range maps. Data on the environmental variables came from the Pantheria database. All these are fully acknowledged in the text.

Ben Collen

4 Acknowledgements

Acknowledgements My supervisors Georgina Mace and Andy Purvis have offered advice and support throughout – they have been invaluable and have played a large part in constructing my thoughts on conservation biology. Above all, working with them has been great fun. Jon Bielby, Marcel Cardillo and Sarah Adamovicz have all read and improved multiple versions of my frequently grammar free manuscripts, and I thank them for their invaluable input. Jonathan Baillie, Nick Isaac, Sam Turvey, Craig Hilton-Taylor, Rich Grenyer, John Gittleman, Konrad Dolphin, Zoe Cokeliss, Ian Owens and Tim Barraclough have all offered discussion and advice for which I am grateful.

Chris Pavey from the Parks and Wildlife Commission for the Northern Territory and the various members of the Dickman lab at the University of Sydney hosted me during those arduous long hot days of ‘field work’. Chris in particular made me feel very welcome whilst in Alice Springs and for a brief moment I wished we hadn’t beaten you so resoundingly in the Ashes…no, hang on. Andy Solow, Dave Roberts, Andy Beet, Marcus Rowcliffe and Dave Orme answered my questions and offered practical help regarding computing.

I thank E.J. Milner-Gulland, Stephen Ling, Lena Bykova, Mike Hoffmann and the Mammal SuperTeam for data and advice. A large group of people, who I won’t name, took the time and trouble to complete Red List assessments necessary for one of my chapters, for which I am very grateful. NERC has provided the funding. Finally, countless friends and family have almost unfailingly acted interested in what I’m doing – the glazed over eyes gave you away.

5 Table of contents

Table of Contents

Abstract 2 Declaration 4 Acknowledgements 5 Table of Contents 6 List of Tables 8 List of Figures 10

Chapter 1 12

Introduction & chapter rationale 12 Bias in diversity patterns 12 Scaling of extinction 15 Chapter 2 17 Biological correlates of description date in carnivores and primates 17 Introduction 17 Methods 19 Data 19 Phylogenetic analysis 20 Species analysis 22 Results 22 Contrasts analysis 22 Species analysis 23 Discussion 23 Tables and Figures 29 Chapter 3 34

Consistent conservation: examining the reliability of IUCN Red List assessments 34 Introduction 34 Methods 37 Species information 37 Threat assessment protocol & application 37 Analysis 38 Results 39 Application of categories & criteria 39 Error – occurrence, direction, magnitude and type 40 Discussion 41 Consistency of classification 41 In the context of other risk assessment protocols 44 Recommendations 44 Tables and Figures 46 Chapter 4 54

When is a species really extinct? Inferring extinction from a sighting record to inform the IUCN Red List 54 Introduction 54 Methods 56 Data 56

6 Table of contents

Analysis 57 Results 59 Tests of robustness 59 Estimating date of extinction 59 Discussion 60 Concerns over effort 62 Levels of certainty 62 Tables and Figures 64 Chapter 5 71

Regional extinction risk in central Asian vertebrates 71 Introduction 71 Methods 72 Data 72 Analysis 73 Results 75 Discussion 77 Tables and Figures 81 Chapter 6 87 Bioregional patterns of extinction in Australian mammals 87 Introduction 87 Methods 89 Database & data collection 89 Analysis 92 Results 94 Distribution of extinction 94 Predictors of extinction pattern 94 Discussion 95 Patterns of extinction 95 Inferring causes of extinction 97 Methodological issues & implications for risk scaling 99 Tables and Figures 101 Chapter 7 108

Synthesis: does extinction risk scale from a local to global level? 108 Summary 108 Introduction 109 Methods 111 Data collation 111 Analysis 113 Results 114 Discussion 115 Evidence of an effect of scale 115 Methodological concerns 116 Processes & measures 118 Conclusions 119 Tables and Figures 120 Chapter 8 126

Conclusions 126 Bias in diversity patterns 126 The scaling of extinction risk 127 Appendices 129 References 152

7 List of Tables

List of Tables

Table 2.1. Wilcoxon’s signed rank tests of categorical variables predicting description date. 29

Table 2.2. Results of single predictor regressions of contrast data and species data for each order. 30

Table 2.3. Multiple regression model for carnivores predicting description date. 31

Table 2.4. Multiple regression model across carnivores and primates. 31

Table 3.1. IUCN Red List categories and criteria. 46

Table 3.2. Variation in classification amongst the assessors. 47

Table 3.3. Variation within IUCN Red List category. 48

Table 3.4. Analysis of Deviance for the proportion of errors made by assessors in each level of experience. 49

Table 3.5. Proportions of different types of errors made by assessors at each level of experience. 49

Table 4.1. Evaluation of Type I error rate. Date of extinction sequentially estimated for each successive number of sightings starting from the most distant. 64

Table 4.2. Estimated dates of extinction for 10 species listed as Regionally Extinct in NSW. 65

Table 4.3. Estimated extinction dates for 10 Asian bird species considered globally extinct. 66

Table 5.1. Assessment of the contribution of phylogeny to each of the traits. 81

Table 5.2. Single predictor regressions for the phylogenetic analysis of correlates of threat on regional IUCN threat rating. 82

Table 5.3. Single predictor regressions of species data of correlates of threat on regional IUCN threat rating. 83

Table 5.4. Single predictor regressions for the phylogenetic analysis of correlates of threat on regional IUCN threat rating in mammals: game species and non-game species. 84

Table 5.5. Minimum adequate model for mammals for the phylogenetic analysis of correlations of threat on regional IUCN threat rating. 85

8 List of Tables

Table 6.1. Component matrix for PCA of traits relating to (a) speed of life history and (b) environment. 101

Table 6.2. Trait coverage for extant and bioregionally extinct species in each of the bioregional clusters. 102

Table 6.3. Sign tests and Wilcoxon’s signed rank tests on sister clade contrasts for the effect of a number of traits on extinction from a bioregion. 103

Table 6.4. Categorical predictors of order of extinction for each bioregional cluster. 103

Table 6.5. Univariate regressions of predictors against order of extinction for each Bioregion cluster. 104

Table 6.6. Minimum adequate model of order of extinction for the Pit & Finke bioregion cluster. 105

Table 7.1. Survey of studies examining extinction risk using phylogenetic comparative methods. 120

Table 7.2. Analysis of variance for sample size of local and global scale studies. 122

Table 7.3. χ2 values generated for combined probabilities for predictors of threat across studies. 122

Table 7.4. Correlates of extinction risk from local and global level studies. 123

Table 7.5. Logistic regression of sign of association with extinction risk, against sample size, predictor of risk and scale of analysis. 124

Table 7.6. Analysis of Deviance to evaluate the effect of scale and predictor on sign of response. 124

9 List of Figures

List of Figures

Figure 2.1. Frequency distribution for discovery date in (a) primates and (b) carnivores. 32

Figure 2.2. (a) Contrast plot and (b) species data plot, showing the relationship between date of description and geographic range. 33

Figure 3.1. Frequency distribution of Spearman’s rank correlation coefficient for pairwise comparisons of ranks of species classifications between assessors. 50

Figure 3.2. Variation in application of Red List criteria. 50

Figure 3.3. Proportion of errors made by assessors of different levels of experience against (a) SSC listing and (b) Consensus listing. 51

Figure 3.4. Proportion of incorrect listings moving species into higher categories of threat (positive values) and lower categories of threat (negative values) for (a) SSC listing and (b) Consensus listing. 52

Figure 3.5. Magnitude of errors made by assessors of each level of experience for (a) SSC listing and (b) consensus listing. 53

Figure 4.1. Variation in rate of extinction over time, redrawn from the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment 2005). 67

Figure 4.2. Forward run of data sequence for Macrotis lagotis dataset, starting with the oldest four 68 sightings.

Figure 4.3. Distribution of estimated extinction dates from 10,000 repetitions of the Macrotis lagotis 69 dataset, rarefied to 10 data points selected at random.

Figure 4.4. 70 Distribution of most recent sighting date for the 162 mammalian CR species.

Figure 5.1. Contrast plot of relationship between IUCN threat category and dispersal distance in central 86 Asian mammals.

10 List of Figures

Figure 6.1. 106 Interim Biogeographical Regionalisation for Australia (IBRA).

Figure 6.2. Frequency histograms of the proportion of (a) extinct species in each family for Australian mammals, and (b) bioregionally extinct species in each family for species in the bioregional 107 analysis.

Figure 7.1. Distribution of traits predicting extinction tested at a global (black bars) and local (white bars) 125 scale.

11 Chapter 1 Introduction

Chapter 1 Introduction & chapter rationale

In this thesis, I present six research papers evaluating the effects of bias and scale on patterns of mammalian diversity. Each chapter contains an introduction and rationale specific to the topic addressed. The thesis divides naturally into two sections, which I will address briefly. The first examines three issues that may bias our current views of macroecological and biodiversity patterns: the effect of undescribed species, the consequences of inconsistency in conservation assessments, and the results of inaccurate listing of extinct species. The second section tackles the issue of scaling in extinction risk, addressing this by examining correlates of extinction at a regional and a bioregional level. The findings of this section are then synthesised with other published phylogenetic comparative analyses of vulnerability to extinction in a meta-analysis, to address how extinction risk might scale from a population to a global level.

Bias in diversity patterns

Due to the current anthropogenically driven extinction event, biodiversity is being lost at an alarming rate, far in excess of the geologically normal levels (Pimm et al. 1995; Regan et al. 2001). The resources available for conservation are, unfortunately, insufficient to prevent the loss of much of the world’s threatened biodiversity during this crisis. Consequently, conservation science is required to make choices, and must often do so in the context of great ignorance about the species it is looking to conserve, and while human expansion continues to exacerbate the speed and magnitude of biodiversity loss (Groombridge 1992; Mace 1995). In order to better assess how fast biological diversity is disappearing and how it might best be maintained, it is necessary to reliably evaluate key aspects of diversity. Moreover, it is important that the effect that biases may have on biodiversity patterns are accurately diagnosed, so that they can be accounted for. As conservation biologists have become increasingly aware of the magnitude and rate of the current extinction episode, the focus has been shifted to disciplines such as priority setting, conservation planning and monitoring trends, with conservation assessments integral to fulfilling these activities (Purvis et al. 2005b). The effect of this realisation

12 Chapter 1 Introduction has been that conservation science has increasingly had to deal with taxa whose natural history and even species membership remains enigmatic.

Species are a fundamental unit of biodiversity. They are readily recognised, and offer the opportunity to both measure and decipher changes at other levels of complexity (Baillie et al. 2004). Conservation biology depends heavily on species lists, as an accurate measure of biodiversity and a means of keeping score; biodiversity loss is frequently expressed in terms of numbers of extinct species. However, taxonomic knowledge is incomplete, with only 1.7 million of the projected 14 million species on earth described (Hawksworth & Kalin-Arroyo 1995). A reliable knowledge of biodiversity patterns is essential for understanding ecological and evolutionary processes, and the effect that increased anthropogenic pressure is having on them. However, uncertainty in species diversity estimates are apparent across large areas, due to a paucity of systematic surveys (Diniz- Filho et al. 2005) and due to taxonomic uncertainty (Isaac et al. 2004; Mace 2004). Given that even in well studied groups, species description is ongoing (e.g. primates: Jones et al. 2005; Sinha et al. 2005), how might currently undiscovered species alter our view of biodiversity?

By examining traits associated with species description we might be able to infer what undiscovered species are like. If we can do this it enables us to make several advances. Firstly it allows prioritisation of the types of groups and geographic regions that may provide new species, so that systematic surveys can be targeted towards the most likely areas. This means description of new species should be more efficient, both financially and in terms of accessing the key areas where new species description is likely to unfold. Secondly it can help define the uncertainty component surrounding undescribed species, which can then be used in a variety of different ways, such as regional conservation planning (Diniz-Filho et al. 2005) or be incorporated into priority setting selection algorithms. Chapter two examines correlates of description date in carnivores and primates, and briefly explores the ramifications such results may have for future description and for conservation practices such as area selection algorithms.

A further key influence of species on biodiversity patterns in mammals is the distribution of threat. Threatened species classification provides a means to highlight species at high risk of extinction, and so focus attention and elicit conservation measures designed to

13 Chapter 1 Introduction protect species at risk (IUCN 2001). They have become widely used in systematic conservation planning (Pressey et al. 1993), priority setting (Bonn et al. 2002) and monitoring trends (Butchart et al. 2004), providing a benchmark on the severity and extent of anthropogenic impact. Their inception can be traced back to the Red Data Books, conceived by Peter Scott in 1963 (Scott et al. 1987; Collar 1996). For the past four decades, the Red List, through IUCN and the Species Survival Commission, has been at the forefront of species risk assessment, facilitating the identification of the worlds most threatened species. Taxa are evaluated against a coarse scale of extinction risk ranging from Least Concern to Extinct (IUCN 2001). BirdLife International produced the first global threatened species list in 1988 (Collar & Andrew 1988), with the first complete assessment of the group, along with mammals, in 1996 (Baillie & Groombridge 1996). This has been followed more recently with the first complete assessments of amphibians (Stuart et al. 2004). Threatened species lists have become widely used; specifically, to assess potentially adverse impacts on species, to help inform conservation priorities and to form the basis of State of the Environment reports (Possingham et al. 2002a; Butchart et al. 2004; Balmford et al. 2005).

As the uses of threatened species lists become more wide reaching, the imperative to ensure they are robust, objective and consistent becomes greater. The Red List has recently undergone criticism governing the extent of its use (Possingham et al. 2002a - but see Lamoreux et al. 2003), its applicability across multiple taxa (Keith 1998) and the belief that the coarse nature of the categories does not accurately reflect the true continuous nature of extinction risk (Victor & Keith 2004). As individual assessors become increasingly accountable for the species classifications they make (Rohlf 1991), it is of paramount importance that assessments are consistent. If threatened species lists are generated inconsistently, the biodiversity patterns inferred from them, and conservation actions incorporating the information they assimilate, will correspondingly be biased. By controlling for the type and quality of information available, in chapter three, I assess inter-assessor consistency of species classification in a variety of user groups, and evaluate how experience may influence assessment of threat status.

The final category on the IUCN Red List is the Extinct (EX) category, which is applied where there is no reasonable doubt that the last individual of a species has died. At face value, it appears that this is perhaps the simplest category to assign; a species is either

14 Chapter 1 Introduction extinct, or it remains extant. However, in reality, the distinction remains problematic. Extinction is rarely witnessed, and therefore must be inferred (Diamond 1987). If we are unable to do this without bias, then views of biodiversity patterns are distorted. Conservation biologists sensibly adopt a precautionary approach towards the classification of extinction. Given that the repercussions of wrongly classifying a species as extinct could mean the diversion of funds away from a species that urgently requires conservation attention, this is a reasonable approach. However, it inevitably leads to an underestimation of the number of extinct species (Groombridge 1992; Lawton & May 1995; MacPhee & Flemming 1999). In realisation of this problem, a number of techniques have been devised to make more objective, the currently subjective process of assessing species extinction (see Solow 2005). In the final chapter of the section, I address a procedure that allows a more robust examination of the date of extinction of a species. I exemplify the method using sighting records of regionally extinct Australian mammals from New South Wales, and globally extinct Asian birds. I examine whether the results could usefully inform the new Possibly Extinct (PE) category, recently proposed for the IUCN Red List (Butchart et al. in press), for these and other species.

Scaling of extinction

Global species extinction typically represents the end point in a long sequence of population declines and local extinctions. Understanding patterns of extinction, decline and threat is a major goal of conservation biology (Purvis et al. 2000c). Accruing evidence supports the assertion that extinction is non-randomly distributed amongst taxa, with extinction risk characteristically clustered amongst species that share certain traits (Bennett & Owens 1997; Mace & Balmford 2000; Owens & Bennett 2000; Purvis et al. 2000a; Purvis et al. 2000b). These comparative studies have highlighted several traits that universally predispose taxa to elevated risk of extinction at a global level, while population level studies have given us insight into the processes governing population decline and extinction. Nonetheless, there is a growing concern that the majority of attempts at assessing the impact of humans on the planet’s fauna have revolved around rate and occurrence of species extinction (Myers 1979; May et al. 1995; Pimm & Brooks 2000). It appears that in observing large scale patterns, we may be using a blunt instrument to study underlying threatening processes (Purvis & Hector 2000). Population declines and extinction are seen as a more sensitive and responsive indicator of

15 Chapter 1 Introduction biodiversity loss (Ceballos & Ehrlich 2002; Balmford et al. 2003). Making progress in understanding the processes governing species extinction may require making robust generalisations about the relationship of extinction between the population and species level. It would be useful to know if correlates are the same between scales.

While there are several global scale studies of extinction risk, and some at a local scale, there is a lack of intermediate regional scale multi-species comparative analyses. Regional studies have the potential advantage over global studies of being able to identify specific threatening processes, while being made more complex by having to deal with cross border populations and non-breeding phases that are nevertheless dependent on the region for certain resources. More narrowly focussed analyses can be of greater practical use to conservation practitioners (Fisher & Owens 2004), and may allow us to make generalisations about the scaling of extinction risk between population and global levels. In chapters five and six I present two phylogenetic comparative analyses of extinction risk at different scales. In the first, a regional analysis of risk for vertebrates in central Asia, I employ IUCN Regional Red List status as a proxy for extinction risk. In the second, I examine whether a suit of intrinsic biological and extrinsic environmental traits can explain the order of smaller scale, bioregional extinctions in Australia. Finally, in chapter seven, I synthesise these results with other published analyses of extinction risk across a range of taxa and scales, in order to assess whether extinction risk scales from a local to a global level.

16 Chapter 2 Description date correlates

Chapter 2 Biological correlates of description date in carnivores and primates

The manuscript presented in this chapter is published as:

Collen, B., Purvis, A. & Gittleman, J. L. (2004). Biological correlates of description date in carnivores and primates. Global Ecology and Biogeography, 13, 459-467

Introduction It is estimated that we have described just 1.5 - 1.8 million of the approximately 14 million extant species (Wilson 2003), and there is still considerable uncertainty how many species exist (Godfray 2002). If our description of species is inherently non-random, with species in some taxa more likely to be described than those in others, our view of diversity is correspondingly distorted. This matters if, for instance, conservation policies are based on skewed reflections of true diversity patterns. Across higher taxa, studies consistently show that probability of description is not equal for all species within a taxon (Gaston 1991; Allsop 1997; Cabrero-Sanudo & Lobo 2003). Broad scale comparisons among lower taxa have suggested that certain groups may receive a greater degree of taxonomic scrutiny (May 1988), perhaps because they appeal to us more (Purvis et al. 2003), and that some taxa have a higher chance of observation due to larger size (Gaston 1991). Even within taxa, accumulating evidence suggests that some species are more likely to be described than others (Figure 2.1), though explanations are more subtle and vary among groups (Gaston 1991; Gaston & Blackburn 1994; Gaston et al. 1995b; Allsop 1997; Reed & Boback 2002).

If we assume that species remaining to be discovered are more similar to recently described species than to species named in the past, we can gain some insight into which biological attributes undescribed species might possess. The published literature contains several tests of hypotheses relating species attributes to dates of description in a variety of taxa (including butterflies, beetles, reptiles, amphibians and birds). These hypotheses include:

17 Chapter 2 Description date correlates

1. Larger animals are more apparent to taxonomists and collectors, are perhaps easier to collect, and are therefore discovered earlier (Gaston & Blackburn 1994; Gaston et al. 1995a; Reed & Boback 2002; Cabrero-Sanudo & Lobo 2003). 2. Species with large geographical ranges are more likely to be encountered and thus described (Patterson 1994; Blackburn & Gaston 1995; Gaston et al. 1995a; Allsop 1997; Cabrero-Sanudo & Lobo 2003). In the same way, restricted range species are less likely to be discovered, e.g. South American endemic birds were found after those that are not restricted to the continent (Blackburn & Gaston 1995). 3. Animals that attain higher densities are more overt or apparent and thus more likely to be described early (Gaston et al. 1995a). 4. Conspicuous animals (for example those that have large home and day ranges, maintain large group sizes, social animals or species with diurnal activity timings that coincide with human activity) are more likely to be discovered. Many aspects of conspicuousness have been highlighted, dependent on the taxa under study (e.g. plumage in birds: Blackburn & Gaston 1995), but few tested. 5. Geographical location determines encounter rate, with early taxonomists centred in the northern hemisphere, and description patterns reflecting settlement by Europeans (e.g. Australian scarab beetles: Allsop 1997; Palaearctic scarab beetles: Cabrero- Sanudo & Lobo 2003). 6. Species with smaller numbers of related species within the same genus may take a shorter time to distinguish taxonomically (Gaston et al. 1995a).

Even within a relatively intensely studied, charismatic taxon such as the mammals, species description still continues at a high rate. For example, between 1982 and 1993, total world primate species recognised was raised from 179 (Honacki et al. 1982) to 232 (Wilson & Reeder 1993), an increase of 29.6%. This has since been elevated to 358 recognised species (Groves 2001a), a further 34.9% change. Though much of the increase in species discovery is attributable to taxonomic revision and application of new species concepts (Groves 2001b), nevertheless, mammal species new to science continue to be discovered.

Statistical analyses on multispecies datasets can suffer from incomplete information about species’ biology and the potential interrelationship between hypothesised predictor

18 Chapter 2 Description date correlates variables. Additionally, in a comparative study, treating species as statistically independent data points may be invalid as it can lead to pseudoreplication and hence elevated Type I error rates (Harvey & Pagel. 1991). However, continued inquiry of the validity of comparative analyses (e.g. Bjorklund 1997; Harvey & Rambaut 1998) means that it is sensible to present both phylogenetic and non-phylogenetic types of analysis (Freckleton et al. 2002).

In this study I examine the relationships between description date and its possible biological correlates in carnivores and primates using phylogenetic comparative methods. Primates and carnivores make an excellent test case for analysing what determines date of description (i.e. when a species new to science is first formally described) because they are both widely distributed and biologically better understood than many other groups, although some species do remain enigmatic. Like some previous studies, we use single predictor regressions of independent contrasts of biological attributes on date of description to ensure statistical and phylogenetic independence across taxa. However, many biological predictor variables inter-correlate (for example body mass with geographic range size, latitude, density and home range, e.g. Gaston & Blackburn 2000). For this reason I control for interrelated and confounding independent variables, and examine clade differences, using a multiple regression. I find that around one third of the variation in date of description can be explained by my models and discuss whether the results for primates and carnivores are similar to those identified previously for other taxa.

Methods

Data Species lists and description dates for primates (232 species) and carnivores (270 species) were taken from Wilson & Reeder (1993), with dates of description spanning the period from 1758-1991 for primates and 1758-1986 for carnivores. The following continuous predictor variables were tested: adult body mass (kg), current geographical range (km2), home range area (ha), day range length (km), population density (individuals per km2), average population group size and number of congeners. Four binary traits were also tested: sociality (1 = species commonly found in groups larger than one or both parents

19 Chapter 2 Description date correlates plus their litter, 0 = otherwise), activity timing (1 = diurnal species, 0 = other activity schedules), island endemicity (1 = island endemic, 0 = not wholly restricted to islands), and for carnivores whether the species is endemic to the tropics (1 = tropical endemic, 0 = range not wholly restricted to the tropics). Information on these species traits came from (Purvis et al. 2000a). Continuous predictor variables were logarithmically transformed to equalise variances, except for number of congeners, which was ln(n+1) transformed (some species have no congeners). Complete data were not available for all species (see Table 2 for sample sizes).

Phylogenetic analysis Phylogenetically independent contrasts (Felsenstein 1985; Pagel 1992) were generated using the CAIC computer program (Purvis & Rambaut 1995). The phylogenies used (Purvis 1995 – modified as in Bininda-Emonds et al. 1999; Purvis et al. 2000b) were composite estimates of all extant carnivore and primate species recognised by Wilson & Reeder (1993). Neither phylogeny was completely resolved, so contrasts at multiple nodes were computed by calculating a difference between two subnodes after the daughter taxa of each multiple node is split into two monophyletic groups (Pagel 1992; see Purvis & Rambaut 1995 for full explanation). Branch lengths were set to be equal when generating the contrasts because preliminary analysis (not reported), showed that homogeneity of variances was more closely approached when branches were set to the same length, rather than when set to estimates of divergence time. All statistical tests were two tailed (except for sign tests where a priori predictions of the relationship between the independent variable and date of description were used to design the analyses, and to rationalise the use of one-tailed tests) and were performed using R version 1.9.1 (R Development Core Team 2004).

Using the BRUNCH algorithm of CAIC to select non-overlapping and non-nested sets of taxa across the phylogeny, categorical variables were tested in turn against the response variable (date of description). Using this algorithm, all the contrasts in the predictor are set to positive so that under H0, roughly half the contrasts are positive and half negative. Sign tests and Wilcoxon’s signed rank tests were carried out on these contrasts to test for significance, for both orders separately and then for the orders combined.

20 Chapter 2 Description date correlates

Continuous variables were tested against description date using least-squares regression through the origin (Garland et al. 1992). For primates and carnivores separately and then for the two orders combined, each variable was first regressed as a single predictor against date of description, as has been the common approach in previous studies examining description date correlates (Blackburn & Gaston 1995; Reed & Boback 2002). Data points with Studentised t-residuals of greater than ±3 were removed and the regressions repeated to assess if data with large leverage had unduly affected the results (following Jones & Purvis, 1997); however, removing these data never made a difference to the results.

For body mass and geographical range, I tested the linearity of the relationship by testing whether slope of description date was constant along the X-axis. At each node in the phylogeny, the average body mass of the taxon comparison and the relationship at the node between body mass and date of description i.e. the quotient of the body mass contrast and the date contrast – termed the ‘contrast slope’ (Jones & Purvis 1997) - were computed. Spearman’s rank correlation was used to test if these two quantities covaried. Further, runs tests were applied to the signs of the contrast slopes once ranked by average body mass, to test the null hypothesis that the signs of the contrast slopes are distributed randomly with respect to body mass. I then did the same for geographical range

For each order I calculated the residuals of the regression of description date on each predictor variable in turn, and examined the differences between the two orders using t- tests on these residuals. Multiple regression through the origin was then used in order to factor out geographical range, which proved the most important single predictor. Multiple models were constructed using forwards step-wise regression sequentially adding each predictor variable to geographical range, and evaluating if additional variance was explained. Finally, to test for significant differences between the multiple models for the two orders, we fitted a model to the combined data using the predictors implicated in either model and the cross-product order × trait interactions (Garland et al. 1992).

21 Chapter 2 Description date correlates

Species analysis For comparison, non-phylogenetic analyses were performed, treating each species as an independent data point. Variables were first regressed as single predictors. Multiple models were constructed using model simplification by first removing the high order interaction terms (order × trait interaction), and the main effects (Crawley 2002). Linearity was tested by including a quadratic function and examining the model to see if there was any improvement in explanatory power.

Results

Contrasts analysis Wilcoxon’s signed rank tests on categorical variables (Table 2.1) showed activity timing significantly predicts description date in primates, with diurnal species on average described earlier (p=0.008). Tropical endemic carnivores are on average described later (p<0.0001), and there is evidence that non-social (p=0.03) and island endemic (p=0.05) species are on average described later. When both orders were combined, no significant result was apparent for island status, but sociality and activity timing remained significant. Such conservative tests however, contain fewer contrasts than regression analysis.

The regression analysis of single predictors (Table 2.2) shows that, within both the primates and carnivores, species with larger geographical ranges tend to be described earlier (primates: b = -20.74, t158 = -10.12, p<0.001; carnivores: b = -12.63, t181 = -8.91, p<0.001). Slopes in the two orders are significantly different (geographic range: b = -

28.86, t384 = -6.91, p<0.0001; order × geographic range: b338 = 8.11, t = 3.27, p<0.01, see fig. 2a). Within the carnivores, large body mass (b = -11.63, t196 = -3.12, p<0.001) and large home range (b = -5.07, t83 = -2.03, p<0.05) are also significantly correlated with early description date as single predictors.

Within the primate single predictor analysis, in addition to those with large geographical range, species with fewer congeners are described significantly earlier (b = 25.66, t159 =

2.69, p=0.01). A more conservative analysis using the BRUNCH algorithm of CAIC (for a full description of the algorithm see Purvis & Rambaut, 1995) was performed as prior

22 Chapter 2 Description date correlates inspection of the variable revealed homogeneity of variance not closely approached. This proved marginally significant (sign test: n+22, n-10, p=0.05).

Spearman’s rank correlations showed that in both orders, neither geographical range

(primates rs = 0.09, carnivore rs = -0.25) nor body mass (primates rs = -0.03, carnivore rs = -0.04) significantly covaried with the contrast slope. Further, no significant results were observed in the runs tests on the signs of the contrast slopes.

Most of the significant correlates of description date did not explain significant additional variance when geographical range was controlled for in a multiple regression. Body size was significant in carnivores, with species more likely to be described early if they occupy a large geographical range and are large bodied (Table 2.3). No multiple model was apparent for the primates however. Differences between orders therefore concerned body mass (significant only in carnivores). Table 2.4a shows that in the combined model, this difference was itself significant. Order significantly interacted with geographic range, but not with body mass.

Species analysis

When species were treated as independent data points in the single predictor non- phylogenetic analysis (Table 2.2) an additional correlate, day range, was found to be significant in primates (t97 = -2.50, p=0.01). In carnivores the same correlates were significant as in the contrasts analysis. There was no evidence of non-linearity. However, when the orders were combined a multiple regression model including geographical range and body mass was found to significantly predict date of description (Table 2.4b). The model explained around 25% of the variance.

Discussion

My results show that taxonomic description of carnivores and primates has been a biased process with several biological correlates of description date apparently predisposing species to a likelihood of earlier discovery and description. My findings strongly suggest that species with large geographical ranges tend to be described earliest, both in primates and carnivores. In carnivores for example, the median geographical range size for the

23 Chapter 2 Description date correlates first 50% of described species is 5.22 ×106 km2 compared to just 3.37 ×105 km2 for the most recently described 25%. An obvious mechanism is apparent whereby species with larger geographical ranges are more likely to be encountered by collectors and taxonomists, and the result is supported by several previous studies on groups as varied as North American butterflies (Gaston et al. 1995a), South American oscine passerine birds (Blackburn & Gaston 1995) and Australian scarab beetles (Allsop 1997). One possibility confounding the result is that more recently described species have smaller ranges simply because they are more recently described, thus enabling little time for their ranges to be delineated (Blackburn & Gaston 1995). However, this is unlikely to explain the magnitude of the effect found in carnivores and primates.

Our results provide limited support for the other hypotheses listed in the introduction. Large home range size correlates with early description in carnivores, but not independently of geographical range size. In primates, species with fewer congeners were discovered earlier. This regression analysis was repeated using the more conservative BRUNCH algorithm of CAIC and a sign test, and still remained marginally significant. (Gaston et al. 1995a) highlight two potential explanations for a similar result in North American butterflies. Firstly, the taxonomy of smaller genera may take less time to distinguish, being less complex and involved. The result could also arise though as an artefact of taxonomists choosing to work on species-rich groups. Secondly, there could be a relationship between number of congeners and another trait that confounds the result. Number of congeners does correlate with geographical range size, but drops out in the multiple regression model when other predictor variables are controlled for. There is further limited support for primate species with larger day ranges being described earlier. Day range however is not significant in the multiple model, or as a single predictor when phylogeny is controlled for.

I find in the single predictor analysis that large bodied carnivores are described earlier, though not larger bodied primates. In the combined analysis, species with large geographical ranges and large body mass were described earliest, combining to explain around 34% of the variation in description date. When analysed separately by order, large bodied carnivores of a given geographical range were described earlier, however, body mass remains a poor predictor of description date in primates. The median body mass of the first 50% of carnivores described was 7.36 kg (mean value 9.75 kg) compared

24 Chapter 2 Description date correlates to a median value of 4.12 kg (mean value 6.56 kg) for total carnivores described to date. The pattern was less pronounced in primates, with a median body mass of 3.51 kg for the first 50% described, but 3.31 kg (mean values 5.28 kg vs. 5.60 kg, respectively) for the total described by 1993 (not significantly different: t238 = -0.09, p>0.05).

The general lack of consistency between orders, and carnivores’ more size-dependent date of description, can perhaps be explained by habitat. With primates being primarily forest based, size may be of relatively little importance, and other factors (for example behavioural traits) may be more likely to determine human encounter rate and thus description. In the more habitat-diverse carnivores, this may not be the case, though body mass is obviously of only secondary importance to geographical range. Reed & Boback (2002) found similar inconsistencies of body size effect across reptiles and amphibians, and the trait has been found to predict description date in British beetles (Gaston 1991), British birds (described 1966-1990: Gaston & Blackburn 1994), North American butterflies (wingspan: Gaston et al. 1995a) and extinct mammals (Alroy 2003), but not in South American birds (Blackburn & Gaston 1995), Australian Scarab beetles (Allsop 1997), nor Western Palaearctic beetles (Cabrero-Sanudo & Lobo 2003).

A study by Alroy (2003) on fossilised mammal species bears on the carnivore result that larger bodied species were discovered earlier. He explains a similar pattern in fossilised taxa, both in terms of taxonomists wanting to make the big catch, but also that certain mammal taxonomists (e.g. Cope, Matthew, Hibbard) were much more influential than others. In our data set there is no obvious effect of author, though a few authors dominate the description of most species. The lack of consistency of body size results between carnivores and primates is striking, especially given that it seems to pertain in different contexts (e.g. species richness: Gittleman & Purvis 1998; extinction threat: Purvis et al. 2000b). There are no doubt multiple reasons - carnivores have greater intrinsic variability in body mass; with increasing size, carnivores move more quickly and over larger areas; with vast travel distances, habitat diversity is much greater than in primates; and last, in the 19th century there was considerable reward and fame for bagging a large carnivore, and many type specimens were from hunting. All would contribute to large carnivores being discovered first.

25 Chapter 2 Description date correlates

Bergmann’s rule states that, within endothermic vertebrate genera, larger species are found at higher latitudes (Bergmann 1847; Mayr 1956, 1963; Freckleton et al. 2003), because larger bodies are beneficial in colder climates. I find that carnivore species with ranges solely in the tropics were on average described later, reflecting a northern bias in carnivore description with most early taxonomists coming from Europe and North America. So there exists a potential bias, whereby large carnivores were discovered early, but they were looked for in areas that contain larger carnivores. However, when the relationship between body mass and whether the species are tropical endemics, was examined it was not significant (sign test: n+13, n-21, p>0.05) and neither was the relationship between body mass and percentage of total range in tropics (b = -0.08, t158 = - 0.59, p>0.05). To some extent this can therefore be ruled out as a confounding factor.

Of the binary variables tested in primates, only activity timing proved significantly predictive of date of description, with diurnal primates on average being described earlier. The one opposing comparison is between Eulemur mongoz (not diurnal) and E. coronatus (diurnal). It is predicted that diurnal organisms have smaller geographical ranges than those with other activity timings, which could confound the diurnal result. E. coronatus has a geographical range of 6.63 ×103 km2, E. mongoz a range of 2.09 ×104 km2 however, there is no general tendency for range size to covary with activity timing (using the BRUNCH algorithm, Sign test: n+ 5, n- 3, p>0.05). In carnivores, sociality and island status were significantly predictive of description date, with social species described earlier on average, and description of species endemic to islands expected later than those that are less restricted (Allsop 1997). The remaining variable that aimed to capture conspicuousness (i.e. big groups – continuous variable) did not significantly predict description date in either order.

Other potentially important factors that were not examined may include other aspects of conspicuousness, both behavioural and morphological. However habitat may confound such traits (Blackburn & Gaston 1995). An obvious candidate predictor for primates and carnivores is pelage colour, with the hypothesis that brighter coats are more perceptible. However, species with a bright pelage may just live in difficult-to-survey habitat – particularly likely in forest dwelling primates where many arboreal species within families have brighter coat colours (e.g. Pongo pygmaeus vs. Gorilla gorilla; Presbytis rubicunda vs. Semnopithecus entellus: Treves 1997).

26 Chapter 2 Description date correlates

There has been debate about whether to control for phylogeny in comparative analyses, with some authors arguing that doing so makes no difference (see Freckleton et al. 2002). In this study phylogeny seems to matter, and phylogenetic methods have been shown to be important in some other taxa, e.g. South American birds (Blackburn & Gaston 1995), North American butterflies (Gaston et al. 1995a), but the methods have not been used in the majority of studies of species description to date.

No discussion of mammalian taxonomy is complete without reference to the phylogenetic species concept and its application. Taxonomic shuffling in the primates in particular has meant an increase over the past 10 years from 232 (Wilson & Reeder 1993) to 358 species (Groves 2001a). Doubtless this increase in species number is primarily due to taxonomic reassessment and application of the phylogenetic, rather than biological species concept (Groves 2001a). Right or wrong, its application may affect analyses like this (see Agapow et al. 2004; Mace 2004). Although species numbers are increasing, many are being raised from synonymy or subspecies (specifically, three quarters of newly recognised species of Neotropical mammals described in the period 1982 – 1994 had already been discovered, and were raised from synonymy: Patterson 1994). So although their ranges will most likely be small, they will not be particularly recently described.

We can look to recent discoveries in primate taxonomy to see if the correlative traits predicted are accurate. Since Wilson & Reeder (1993), night monkeys (Aotus), Galagos (Galagoides udzungwensis and G. rondoensis; (Honess & Bearder 1998)) and mouse lemurs (Microcebus ravelobensis; (Zimmermann et al. 1997)) have spawned additional species descriptions – all are nocturnal clades. As a cautionary note however, diurnal groups such as the Titi monkeys (Callicebus) have undergone extensive revision with the addition of 2 species new to science (van Roosmalen et al. 2002) as well as the elevation of 6 taxa listed by (Groves 2001a) as subspecies to species level - an issue of species concept is therefore apparent.

The findings of this study may be of particular relevance for hotspot selection algorithms, and reflect an avenue for future research. Conservation priority areas are often defined using numbers of endemic species - or at the very least, area selection algorithms are most sensitive to such restricted groups. Small-range specialists or endemics are major

27 Chapter 2 Description date correlates contributors to areas of tropical species diversity (Stevens 1989; Pagel et al. 1991; Hughes et al. 2002). Myers et al. (2000) for example, define hotspots using endemic plant diversity. So the very species to which some priority setting algorithms attach most conservation value are least likely to have been described.

28 Chapter 2 Description date correlates

Tables and Figures Table 2.1. Wilcoxon’s signed rank tests of categorical variables predicting description date (all tests 1 tailed). For island status, 0 = not endemic to islands, 1 = endemic to islands. For activity timing, 0 = not diurnal, 1 = diurnal. For sociality, 0 = not social, 1 = social. For range in tropics, 0 = not tropical endemic, 1 = tropical endemic.

Order Predictor pos/neg contrasts Wilcoxon V Carnivores Sociality 8/22 330* Activity timing 8/17 161.5 Island status 11/16 80* Tropical endemic 35/40 200*** Primates Sociality 6/15 150 Activity timing 1/8 56*** Island status 6/12 72 Combined Sociality 14/37 888** Activity timing 9/25 412.5** Island status 17/28 308 * p<0.05, ** p<0.01, *** p<0.001.

29 Chapter 2 Description date correlates

Table 2.2. Results of single predictor regressions of contrast data and species data for each order. Nc = number of contrasts; Ns = number of species, bcontrast = coefficient of contrast analysis, tcontrast = t-value of contrast analysis, bspecies = coefficient of species analysis, tspecies = t-value of species analysis. All tests are two tailed.

s

e 54*** i c 61** 67** 67 e 58 81 01 p 2. 10. 2. 0. s t 1. - - - - 0. 1.

s

e

i 89 c e 72 96 01 p 85 60 61

s 3. 12. 3. 2. s b 2. - - - - 2. 4. e

vor

i t s

n a

tr 12** 71*** 15* 93 15 ar n 00 71 o 3. 8. 2. 0. 1. C c t 1. - - - - 0. -

t

s

a

63 63 tr n 07 60 65 o 64 15 c 11. 12. 5. 3. 7. b 2. - - - - 4. -

c 3 N 75 197 182 84 4 211 89

s N 88 240 236 99 47 270 109

s

e i c 30 32*** 87 50* 51 e 21 15* p 1. 8. 0. 2. 1. s t 0. - - - - 2. -

s

e

i 21 37 c e 56 89 09 p 76 53 s 3. 14. 1. 16. 5. b 0. - - - - 9. -

s e at

t

s m

a i 00*** r t r 35 43 32 60 n 54 69* P o 1. 10. 0. 1. 0. c t 0. - - - - 2. -

t

s

a

49 74 29 tr 66 n 72 41 001 . o 03 c 11. 20. 1. 16. 4. 0 b 3. - - - - 25. -

p<

c N 86 138 159 110 84 160 136 ***

s 01, . N 96 179 211 131 98 217 170 0

s

y c t i p< i

r nge ns a ne **

r ,

de

c

s e 05 s nge z on onge a i or i a phi 0. c s t t r nge

a

m c a a i e r p< of d

ogr y e oup *

e om a o. r ody r opul

P P B G H D N G

30 Chapter 2 Description date correlates

Table 2.3. Multiple regression model for carnivores predicting description date. Sample size: 206 species, 168 contrasts.

Predictor Coefficient se t Geographical range -11.89 1.55 -7.64*** Body mass -8.69 4.03 -2.16* * p<0.05, ** p<0.01, *** p<0.001

Table 2.4. Multiple regression model across carnivores and primates. (a) Contrast model, sample size: 385 species and 308 contrasts, model explains 34 % of the total variance, (b) species model, d.f = 383. (a) Predictor Coefficient se t Geographical range -10.77 1.59 -6.76*** Body mass -7.67 3.66 -2.10* Order × geographic range -9.15 2.55 -3.59***

(b) Predictor Coefficient se t Geographical range -10.90 1.02 -10.72*** Body mass -3.62 1.50 -2.41* * p<0.05, ** p<0.01, *** p<0.001

31 Chapter 2 Description date correlates

(a)

(b)

Figure 2.1. Frequency distribution for discovery date in (a) primates and (b) carnivores

32 Chapter 2 Description date correlates

(a)

(b)

Figure 2.2. (a) Contrast plot and (b) species data plot, showing the relationship between date of description and geographic range. Solid circles denote carnivores (solid line is regression line); open circles denote primates (dotted line is regression). An ANCOVA test showed a significant effect of order.

33 Chapter 3 Red List assessment consistency

Chapter 3 Consistent conservation: examining the reliability of IUCN Red List assessments

Introduction

The IUCN Red List of Threatened Species provides readily assimilated information to focus attention on the plight of species at risk of extinction (Mace & Lande 1991; Mace 1995; IUCN 2001). It has been described as one attempt to increase the world’s store of knowledge about our biological resources before they are lost (Baillie et al. 2004). In classifying species into categories of extinction risk, the Red List provides information about one of the most informative aspects of the status of biodiversity. It is policy relevant but not policy prescriptive, so informed decisions to target funds towards species, groups or areas of the most urgent need are often made incorporating, though not solely using, its information. For more than four decades, the IUCN Red List through IUCN and the Species Survival Commission (SSC) has been at the forefront of extinction risk assessment, estimating the conservation status of species at a global level, in order to highlight taxa threatened with extinction and therefore promote their conservation. Redeveloped in 1994 as a more quantitative system (IUCN 1994), the IUCN Red List categories and criteria are intended to provide an explicit and objective framework for the classification of the broadest range of species according to their extinction risk. IUCN cite several more specific aims (IUCN 2004): 1. To provide a system that can be applied consistently by different people. 2. To improve objectivity by providing users with clear guidance on how to evaluate different factors which affect the risk of extinction. 3. To provide a system which will facilitate comparisons across widely different taxa. 4. To give people using threatened species lists a better understanding of how individual species were classified.

Whilst its application to non-vertebrates remains incomplete, the Red list has become widely accepted and applied by practitioners at local, regional and global levels, due to its wide applicability, objectivity and simplicity of use, on a broad range of vertebrate taxa

34 Chapter 3 Red List assessment consistency

(Akçakaya et al. 2000). Complete assessments exist for all mammals, birds and amphibians (Baillie et al. 2004; IUCN 2004). Though the Red List has been effective at overcoming the limitations of subjective opinion in extinction risk assessment, and provides a quantitative system that combines different indicators of risk to compare level of threat in a broad range of taxa (see Table 3.1), the aims stated above have rarely been tested. The gathering of a number of Red List user groups at the Global Mammal Assessment, African small mammals workshop (GMA Africa) in December 2003, which formed part of the Global Species Assessment (Baillie et al. 2004), provided an opportunity to examine these aims.

The Red List is the product of a network with over 7000 scientists and conservation practitioners, providing assessments and independent peer review (IUCN 2001; Lamoreux et al. 2003), where transparency, accountability and openness are cornerstones of the process. However, the Red List has been criticised on a number of issues. The limit to the use of threatened species lists in decision-making processes has probably been the most public and most cited criticism (see Possingham et al. 2002b and reply by Lamoreux et al. 2003). Other criticisms have queried the applicability of criteria thresholds across multiple taxa (Keith 1998), particularly for sessile organisms such as vascular plants (Keith 1998), very restricted range species such as Brazilian endemic amphibians (Pimenta et al. 2005) and wide ranging aquatic species such as turtles (Mrosovsky 2003; Pimenta et al. 2005 – but see response by Stuart et al. 2005). Some have argued the categories represent a coarse interval scale not accurately reflecting the continuous distribution of extinction risk and the geographic range criterion is scale dependent (Mrosovsky 2003). There are further concerns that taxa listed as LC can be rare, but not declining (Victor & Keith 2004). The implication is that this precludes a higher listing, though the taxon may still be at threat through the possible effect of stochasticity on small populations. In less well justified evaluations, criticism of the Red Listing process is seen by some as “onerous”, owing to the validation procedures required to include nationally listed species on the global list (Heywood 2003), whilst some are unable to see the value of a “complicated procedure that requires considerable data and expertise”, implying the older system of subjective opinion by experts was sufficient (Burton 2003 – but see reply by Given 2003). The advantage of the modern IUCN system is its transparency, so when discrepancies occur, the causes can be determined

35 Chapter 3 Red List assessment consistency

(Regan et al. 2004). With the previous system based largely on expert subjective judgement, this was not necessarily the case.

Whilst some criticisms are valid, the majority are case or taxon dependent, and lie outside the scope of this chapter. Increasingly though, individual assessors are being held accountable for their assessments and have been required to defend their classifications (Rohlf 1991). This chapter focuses on evaluating the consistency with which assessments are made. The other criticisms form part of a broader discussion of the scope of the Red List. However, if assessors are given the same information and produce vastly divergent classifications, then the system is failing in its aims. Because of the increasingly important role that threat classifications such as the Red List play as part of a toolbox for setting conservation priorities, it is critical that assessments are accurate and repeatable.

Inconsistencies arise from uncertainty, of which there are two types: epistemic uncertainty (Williamson 1994; Regan et al. 2000) and semantic uncertainty (Akçakaya et al. 2000). Epistemic uncertainty arises from incomplete data, extrapolation and the limitations of measurement accuracy, whereas semantic uncertainty arises from miscommunication and vague or inexact definitions, which permit borderline cases. Semantic uncertainty may be tackled by tightening imprecise definitions, or by using a technique that respects the inherent imprecision of the process, such as fussy logic, or fuzzy set theory (Regan et al. 2000). A more fundamental error is that of epistemic uncertainty. Gathering more data, e.g. carrying out more surveys to obtain a more accurate view of population size, can to an extent ameliorate this, but cannot address extrapolation or user error. In order to address how Red List user groups deal with these types of uncertainty, and to determine to what extent multiple assessors produce similar Red List classifications, a comparative assessment was carried out amongst 14 Red List users and 17 student first time user groups. The key to the study was controlling for information provided, in order to evaluate the consistency with which assessors and user groups come to similar conclusions. The Red List protocol (IUCN 2001) was applied by each assessor to a set of 10 marsupial species with varied levels of threat. I compared the congruence of the results with the documented IUCN Red List assessment of the species, and evaluated this against the experience of the assessors. I then examined which criteria assessors applied, and the occurrence, direction, magnitude and type of errors made. Consistent classifications are of great importance to species assessments, as

36 Chapter 3 Red List assessment consistency misclassification and inconsistent listing both compromises the credibility of threatened species lists, which in turn impinges on the conservation actions that are guided by threatened species assessment.

Methods

Species information Information was collated on ten Australian marsupial species. Marsupials were chosen as a test case, as none of the participants had specific knowledge of them. Species were selected at random to represent all the levels of extinction risk in the Red List (see Tables 3.1 and 3.2). Data was placed in a form commonly available to conservation biologists making conservation assessments. Information was presented for each species to varying degrees of detail, on population status and trends, distribution, habitat, species biology, threats and existing protection (see Appendix 9.1). The majority of the data was collated from the IUCN action plan for Australian marsupials and monotremes (Maxwell et al. 1996).

Threat assessment protocol & application

IUCN Red List categories and criteria are decision-based rules based on qualitative and quantitative criteria (IUCN 2001). Possible outcomes are EX, EW, CR, EN, VU, NT, LC and DD (see Table 3.1) and relate to both quantitative changes in, and specific threshold levels of, population size and/or distribution. The Red List categories and criteria version 3.1 (IUCN 2001) were followed and applied to each of the species by each assessor or group.

14 assessors applied the IUCN criteria independently in order to determine the threat status of the 10 Australian marsupials. A further 17 groups of third year undergraduate students from Imperial College London and M.Sc students Manchester Metropolitan University (three - six students per group) also assessed the species during a three hour workshop on the IUCN Red List. Assessors were categorised into groups of successively less experience; those that are responsible for implementing the Red List (reviewers), those that teach others use of the categories and criteria (trainers), those that were being taught to use the categories and criteria at the GMA Africa workshop in order to act as

37 Chapter 3 Red List assessment consistency facilitators (trainees) and first time users of the Red List (students). All were asked to evaluate the information presented to estimate population size, range size, trends and other relevant biological information. From these they made Red Lists assessments, which would later be compared to the documented IUCN status. Both the final category and criteria classification concluded by the assessor were recorded and compared.

The data presented in the action plan for Australian marsupials and monotremes (Maxwell et al. 1996) was not sufficient to come to the same decision as the documented Red List status, in every case (IUCN 2004). Whilst this is an issue in itself that is returned to in the discussion, in order to obtain fair comparisons, a consensus opinion of the true assessment that could be gained from the data available was agreed on by the Red List Programme Officer (C. Hilton-Taylor), Red List author (J. Baillie) and the author of the original categories and criteria (G. Mace). The two listings are referred to herein as the SSC listing and Consensus listing.

Analysis Four aspects of consistency were evaluated; whether the categories and criteria applied correctly, whether the level of experience of the user group affected the outcome, were less experienced assessors likely to make errors of different types and did they make errors of greater magnitude? The results were analysed by proportion of agreement to the true category within species, within Red List category and by assessor. Pairwise comparisons of 14 assessors’ species ranks produced 90 correlations, which were analysed using Spearman’s rank correlation coefficient (student assessments were not included in this section of the analysis because they did not complete all assessments). Data Deficient results were excluded from the analysis, as they do not fit into a meaningful rank system. Agreement within category was assessed by examining the proportion of assessors who classified species correctly within each category. The Jaccard similarity index (Ludwig & Reynolds 1988) was used for comparison. This is calculated as the number of species that have the same risk category between two assessors, divided by the total number of species in that category. This resulted in 90 comparisons of agreement between assessors. The variation in application of the criteria was examined as well as the deviation from the true result.

38 Chapter 3 Red List assessment consistency

Error was measured on four levels; simply whether or not the assessor made a mistake in categorisation, what was the magnitude of that mistake, in what direction was the error made and what type of error was made? For the first, a listing of the species in an incorrect category was scored as a one, a correct listing as a zero. Not all assessors reviewed all of the species, therefore proportional deviation from the true answer was evaluated with an Analysis of Deviance in a GLM with binomial errors, using an F-test to correct for over-dispersion (Crawley 2002). For magnitude of error, it is assumed that categories are a linear scale from LC through to EW (see Purvis et al. 2000b; Purvis et al. 2005a for a discussion). Number of threat categories away from the true answer was recorded. As the DD category does not fit into this linear scale, a species that was incorrectly assessed as DD, or a DD species that was assessed incorrectly as another threat category was given an error value of one. The effect of experience on direction of incorrect listing was evaluated. The incorrect listing of a species was scored as being in a higher or lower threat category than the SSC or Consensus listing. Effect of experience was analysed using a binomial proportions test. Finally, consistency of errors has three components: data interpretation (e.g. conversion of a series of population distributions into a species extent of occurrence), evaluating data correctly against criteria (i.e. rule following) and risk attitude (does experience affect attitude towards risk). The type of error made was recorded and evaluated qualitatively. All data was analysed using R version 1.9 (R Development Core Team 2004).

Results

Application of categories & criteria Assessors assigned the correct category to the species in 71% of cases for both the SSC and the Consensus results. In general there was good agreement between assessors on which risk category should be assigned to each species, both within the SSC listing and the consensus listing (Table 3.2). There were much lower levels of agreement for Petrogale sharmani (SSC: 0.16, Consensus: 0.20) and Bettongia penicillata (SSC: 0.05, Consensus: 0.33). Assessors also consistently failed to classify Bettongia tropica in the correct risk category, according to the consensus listing (proportion of agreement 0.26). When the assessments were evaluated within category, it became clear that some categories had more agreement than others. Table 3.3 shows the variation within

39 Chapter 3 Red List assessment consistency category. The proportion of correct assessments is high for the threatened categories (CR, EN and VU) and for DD, but is lower for NT (SSC only) and LC (Consensus only). Jaccard similarity index proportions gave qualitatively similar results. Rank correlation coefficients for pairwise comparisons between assessors’ ranks were variable (Figure 3.1), ranging from 0.28 to 1.00. However, the comparisons peaked strongly in the 0.8 - 1.0 range. There were no negative correlations.

Closer evaluation of the main criteria used to assign categories (Figure 3.2) reveals that some criteria are more consistently applied in a correct manner than others. Criterion C (relating to small and declining populations) was applied correctly in all of the cases in which it was used. Criterion B (restricted geographic range) was applied accurately in 76% of cases. Criterion A (population reduction) and criterion D (very small or restricted population) were used correctly in only 56% of cases. There were insufficient data to evaluate the application of sub-criteria.

Error – occurrence, direction, magnitude and type Though there was qualitatively greater variation in the errors made by less experienced assessors (Figure 3.3), they were not significantly different from assessors of greater experience (Table 3.4). When the direction of error was considered (Figure 3.4) in comparison to the SSC listing, trainees and students were found to be significantly more likely to elevate species to a higher category of threat, than to move it to a less threatened 2 2 category (binomial proportions test: trainee - χ1 = 22.32, p<0.001; student - χ1 = 20.82, p<0.001). Reviewers were qualitatively more likely to classify a species in a higher category of threat, and trainers were more likely to lower the species to a less threatened category, though these differences were not statistically significant. These patterns remained qualitatively similar, when evaluated against the Consensus listing; students at a 2 reduced level of significance (χ1 = 9.51, p<0.01), trainers were significantly more likely 2 to down-list a species to a lower category of threat than the true level (χ1 = 12.93, p<0.001).

79% of the errors made were single category errors when compared to the SSC listing, and 74% when compared to the Consensus listing. 16% were two category errors (for both SSC and Consensus listings) and 3% and 16% three category errors, for the SSC and

40 Chapter 3 Red List assessment consistency

Consensus respectively. Level of experience had no significant effect on the magnitude of errors made (Figure 3.5 – Analyses of Deviance all p>0.05, not reported). Qualitatively, the less experienced trainees and students made errors of greater magnitude though, as they were the only groups to misclassify by more than two categories in comparison to the SSC listing, and the only groups to misclassify by more than three categories in comparison to the Consensus listing. The majority of student errors involved data interpretation (54% of errors made); trainee errors were mostly concerned with rule following (62%), whereas trainers and reviewers made errors involving their attitude to risk (60% and 58% respectively: see Table 3.5).

Discussion

Consistency of classification The major finding of this study is that although assessors vary in their classification of a species’ extinction risk, variations are not significantly different between user groups of different experience, and tend to be relatively consistent with the documented classification given by the IUCN Red List. It is perhaps surprising though, that even given limited information, assessors classified certain species in every possible risk category. However, this variation is not evenly distributed, with eight from ten species showing relatively high congruence of category listing (>70%). Separate analysis of the Red List categories provides more evidence of where the variation lies. In general, close agreement is shown in the higher threat categories (CR, EN, VU), and within the DD category. Assessors consistently found species classified by IUCN as NT the most difficult to resolve. An explanation may lie in the nature of the criteria. While the threatened categories have strict quantitative standards that must be met in order to classify a species, the NT category is not so strictly bounded. A taxon is considered Near Threatened when ‘it has been evaluated against the criteria but does not qualify for CR, EN or VU now, but is close to qualifying for or is likely to qualify for a threatened category in the near future’ (IUCN 2001). The decision is more subjective. A body of literature shows that subjective judgement is likely to be both biased, and to vary unpredictably between experts (Tversky & Kahneman 1974; Anderson 1998; Keith et al. 2004; Seoane et al. 2005). In practice, species are commonly placed in this category if they come very close, but do not quite meet, one of the VU criteria, e.g. they exhibit a

41 Chapter 3 Red List assessment consistency

25% decline in population size over the last 10 years or 3 generations (J. Baillie, pers. Comm.; see Red List guidelines; IUCN 2001). The advantage of quantitative protocols such as those used in the Red List is that they offer a means of informing expert judgement, which is better than either judgement or a protocol alone. Perhaps in the case of the NT category this is not facilitated, due to the absence of specific quantitative threshold values. Implicitly, it is clearer how to classify a species that meets a criterion, rather than one that fails to meet it.

The other key finding from the study was that the proportion of errors made by an assessor did not seem to be determined by their experience. Based on this evidence, IUCN appears to meet its aim of providing a system that can be consistently applied by different people. Given that many of the students who took part in the study were using the categories and criteria for the first time, this is a good indication of the simplicity and accessibility of the system. However, the simple occurrence of errors neither accounts for the direction in which they go, the magnitude, nor the type. Potentially, conservation practitioners and scientists are more likely to up-list species into higher categories of threat, due to their general concern over the conservation problem. Indeed, this certainly seems to be true for students and possibly for trainees. Reviewers were equally likely to apportion species to both higher and lower categories than the SSC or Consensus listing, while trainers were significantly more likely to down-list a species in comparison to the consensus listing. Regardless, exaggerating the level of threat compromises the credibility of threatened species lists, as is exemplified in the recent debate over the validity of assessments for the GAA (see Pimenta et al. 2005; Stuart et al. 2005). (Akçakaya et al. 2000) relate this phenomenon to a scale of risk tolerance. They argue that a user with a precautionary attitude would accept that a species is safer than endangered only if it is certainly not endangered. Conversely, a user with an evidentiary attitude (the type of attitude adopted by the GAA) would demand substantial evidence of endangerment, before classifying a species in a higher threat category.

There are of course consequences to both attitudes, not least in the inappropriate allocation of conservation resources. In practice, in a group classification, such as one carried out by a Species Specialist Group, a dispute over risk tolerance would be mediated with either an intermediate stance, or a varied response determined on a case- by-case basis. Practitioners may argue that it is more favourable for a species to be

42 Chapter 3 Red List assessment consistency placed in a higher risk category so that something is done about the conservation status of the species, rather than listing a species in a lower category of threat where it is ignored and the species continues to decline unchecked. Some may feel that they are more likely to obtain funding to study a species that is classified in one of the high risk categories, or conversely might want to minimise restrictions on scientific collection by down listing it (Stuart et al. 2005). These are issues that must be faced by IUCN, funding organisations and scientists alike, and support the need for consistency checks.

The magnitude of a mistake compounds the severity of an error. For example, recording a CR species as LC could be more detrimental to the species than recording the same species as EN or VU. The great majority of errors amongst all levels of experience were single category errors. Only inexperienced assessors made errors of ≥3 categories. Errors may occur for a number of reasons but given that I have controlled for the uncertainty caused by assessors having different sources of information, differences are likely to be due to factors such as data interpretation, inadvertent error, dissimilar willingness to estimate parameters and different attitudes towards uncertainty. It is difficult though to differentiate between them without having asked the assessors to give very specific reasoning for their listings. The evaluation of the type of errors made, though qualitative, is somewhat revealing. Experience appears to determine the type of errors made. Less experienced assessors make the majority of errors in data interpretation, i.e. true errors, in comparison to the more experienced assessors whose errors are more likely to be governed by their attitude towards risk. Intermediate level assessors appear to make most errors following the category rules, advocating a requirement for continued training and familiarisation with Red List protocols. The Red List categories and criteria are designed to allow classification even when comparatively sparse information is available (IUCN 2001). While IUCN provides guidelines, the concepts of inference and uncertainty are personal to each assessor. In situations where data is vague or uncertain, these will vary with the level of interpretation each assessor is willing to make. In the process of official Red List assessment, the group variation liable to result from this is resolved by group consensus, for example, amongst the Species Specialist Group. Clearly the assessors who took part in this exercise were unable to group resolve differing opinion, so variation may appear larger than we might hope for.

43 Chapter 3 Red List assessment consistency

In the context of other risk assessment protocols

There have been three other studies that have evaluated the performance of extinction risk assessments protocols. Keith et al. (2004) found that the Red List (IUCN 2001), NatureServe (Master 1991; Master et al. 2000) and the Florida Fish and Game (Millsap et al. 1990) protocols performed similarly in dividing extinct species from extant sister taxa. They found that the IUCN system gave the best separation, although with greatest operator error. O'Grady et al. (2004) compared the ability of the same three protocols to rank 55 mainly vertebrate taxa in order of priority, and of the outcomes to correlate with rankings based on probability of extinction within 100 years (as determined by a PVA). They found all three systems gave positively correlated ranks of the taxa tested. Finally Alvo & Oldham (2000) compared the Red List, TNC Heritage (Master 1991) and

COSEWIC (COSEWIC 2005) systems for 93 Canadian reptiles and amphibians. The major finding showed least concordance was apparent between protocols at low levels of risk. Unlike previous studies, this analysis concentrates on the assessment of within protocol consistency. The three other studies, which assess cross-protocol congruence, all highlight the advantage of familiarity with the risk assessment system in producing consistent classifications. While this is also evident from my findings and doubtless will lead to better assessments, I show that consistent classifications are gained across all levels of experience.

Recommendations

Two courses of action could be recommended for the Red List categories and criteria, in light of the evidence provided by this study. The utility of the NT category is that it enables assessors to flag attention towards a species that does not quite reach the criteria for a threatened category, but which they feel should not be grouped with the most abundant and widespread species of the LC category. The expectation is therefore that in the next round of assessments, the species may move to a higher category of risk, or if conservation intervention is successful, be down-listed to LC. Whilst the temptation may be to make the definition more exact by introducing a somewhat arbitrary numerical threshold (e.g. see Regan et al. 2000), this would result in the loss of generality and the possibility of species experts instigating proactive conservation action by highlighting an at risk species. It is neither possible, nor desirable to remove all aspects of subjective judgement from risk classification systems such as the IUCN Red List. The alternative

44 Chapter 3 Red List assessment consistency but not mutually exclusive option would be for IUCN to create a ‘safe as houses’ category for the most widespread and abundant species. Currently the LC category is particularly broad, ranging from highly invasive pest species with growing populations of many millions of individuals, to species with just over 10,000 mature individuals with a decline rate of 3% per generation (Purvis et al. 2005a). Perhaps in this way, the stigma associated with placing a non-threatened but restricted species in the current LC category would be less. However, the strength of the Red List lies in its simplicity. Whilst the evolution of its categories and criteria in light of scientific debate and new information should be welcomed, extensive reclassification would be detrimental. The bias in assessment towards vertebrate taxa must be addressed, and is a focus of future Red Listing, but for the most part I believe the categories and criteria should be left intact. However, additional guidance and regulation on the use of the NT category may be useful, as different groups could potentially apply it inconsistently.

My main recommendation is one of transparency. If the SSC provides action plans with information on species status, such as the Action plan for monotremes and marsupials (Maxwell et al. 1996), there should be sufficient data contained within the document to come to the same conclusion as the assessors. While decisions are without doubt made with the benefit of expert knowledge, this should be conveyed accurately in the report. IUCN has been similarly upfront with their online database (IUCN 2004), where names of assessors, evaluators and a justification of the species classification are routinely given. Species Specialist Groups must follow suit.

45 Chapter 3 Red List assessment consistency

Tables and Figures

Table 3.1. IUCN Red List categories and criteria for CR, EN, VU. A taxon is considered NT when it does not meet the following criteria, but is close to qualifying, or is likely to qualify for a threatened category in the near future. A taxon is LC if it does not meet any of these categories – widespread and abundant taxa are included in this category. If there is inadequate information to make an assessment, a species is categorised as DD. Taxa that only survive in captivity or as naturalised population(s) well beyond the former range are EW. When there is no reasonable doubt that the last individual has died, a taxon is considered EX (IUCN 2001).

Critically Use any of the criteria A-E Endangered Vulnerable Endangered A. Population reduction Declines measured over the longer of 10 years or 3 generations A1 ≥ 90% ≥ 70% ≥ 50% A2, A3 & A4 ≥ 80% ≥ 50% ≥ 30% Al Observed, estimated, inferred, or suspected in the past based on and specifying any of the following: (a) Direct observation, (b) an index of abundance appropriate to the taxon, (c) a decline in AOO, EOO and/or habitat quality, (d) actual or potential levels of exploitation, (e)

effects of introduced taxa, hybridisation, pathogens, pollutants, competitors or parasites. A2. Observed, estimated, inferred, or suspected in the past where the causes of reduction may not have ceased (a) to (e) under Al A3. Projected or suspected to be met in the future (up to a maximum of 100 years) based on (b) to (e) under Al. A4. Observed, estimated, inferred, projected or suspected (up to a maximum of 100 years) where the time period must include both the past and the future, based on (a) to (e) under Al. B. Geographic range in the form of either B1 (extent or occurrence) OR B2 (area or occupancy) B1. Extent of occurrence < 100 km² < 5,000 km² < 20,000 km² B2. Area of occupancy < 10 km² < 500 km² < 2,000 km²

and 2 of the following 3: (a) Severely fragmented or # locations = 1 ≤ 5 ≤ 10 (b) Continuing decline in any of: (i) extent of occurrence; (ii) area of occupancy; (iii) area, extent and/or quality of habitat; (iv) number of locations or subpopulations; (v) number of mature individuals (c) Extreme fluctuations in any of: (i) extent of occurrence; (ii) area of occupancy; (iii) number of locations or subpopulations; (iv) number of mature individuals C. Small population size and decline Number of mature < 250 < 2,500 < 10,000 individuals and either C1 or C2: C1. An estimated continuing 25% in 3 years or 1 20% in 5 years or 2 10% in 10 years or decline of at least: generation generations 3 generations up to a maximum of 100 years C2. A continuing decline and (a) and/or (b): a (i) # mature individuals in < 50 < 250 < 1,000 all sub-populations: a (ii) or % individuals in one 90% 95% 100% sub-population at least (b) extreme fluctuations in the number of mature individuals D. Very small or restricted population (1) no. of mature individuals < 50 < 250 < 1,000 OR AOO < 20 km² or na na (2) restricted AOO # locations ≤ 5 E. Quantitative Analysis Indicating the probability of ≥ 50% in 10 yrs or 3 ≥ 20% in 20 yrs or 5 ≥ 10% in 100 years extinction in the wild to be: gens. (100 yrs max) gens. (100 yrs max)

46 Chapter 3 Red List assessment consistency

Table 3.2. Variation in classification amongst the assessors. Agreement values are proportions, calculated as the number of assessors listing the species under the same category as the SSC/Consensus, divided by the total.

t s n u e s n m e e s

e 92 88 89 90 00 20 33 26 68 88 gr on A C 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.

t n e m e

e C 92 88 89 90 00 16 05 58 68 88 gr S A S 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.

s or s

s al e s

s ot T A 26 26 27 29 25 25 21 19 19 16

) i i i , i i b( 2a +

)

D i

i ; i v) ) ,

i i i

( s i i )

, i u i d ( b( ( bc s

n 2a 2a 1 1a 2a 1a e s C A C B C B

on D D U D U U R R N C C D D V E C L D C V V

) v

v, ) i i ,

i i b( 1

i v) , i ) i C

i ( D 1a i

( , ; c g ; i B e )

n i 2a i 2a b( ( t s 1d; 1bc i 1+ 1a 1+ 2a l

C A A B B C

C D D U T T U U R N N S S D D V E C N N E V V

a

a

a t ur i s r us at

l i us ni ani l i a at s gal he i t c e u c hi m ur r c ul v c m ni ar e hy be opi ac l a har s r a l s par i pe t m ac i r gi

s e s a a br

y pe

u s i ops x l e hops ous ur i m y am ogal t ongi ongi y c nt r or r t t i oni e hy t t t t as ac c ur ot e e e p S Sm E D B P P B B D Se

47 Chapter 3 Red List assessment consistency

Table 3.3. Variation within IUCN Red List category. Proportion = proportion of assessors who assessed each category correctly, Jaccard = Jaccard similarity index proportions.

Mean agreement SSC Mean agreement Consensus Category Proportion Jaccard Proportion Jaccard CR 1.00 1.00 0.68 0.68 EN 0.77 0.68 0.90 1.00 VU 0.82 0.78 0.82 0.78 NT 0.11 0.29 na na LC na na 0.20 0.29 DD 0.90 0.79 0.90 0.63

48 Chapter 3 Red List assessment consistency

Table 3.4. Analysis of deviance for the proportion of errors made by assessors in each level of experience, shown in Figure 3.3.

SSC Coefficients b se Z (intercept) -1.10 0.37 -3.01** Student 0.32 0.43 0.75 Trainee 0.05 0.49 0.11 Trainer 0.44 0.47 0.92

Consensus Coefficients b se Z (intercept) -1.24 0.38 -3.27** Student 0.61 0.44 1.139 Trainee 0.48 0.49 1.00 Trainer -0.03 0.51 -0.06

Table 3.5. Proportions of different types of errors made by assessors at each level of experience.

Experience Data interpretation Rule following Risk attitude Reviewer 0.17 0.25 0.58 Trainer 0.20 0.20 0.60 Trainee 0.23 0.62 0.15 Student 0.54 0.31 0.15

49 Chapter 3 Red List assessment consistency

Figure 3.1. Frequency distribution of Spearman’s rank correlation coefficient for pairwise comparisons of ranks of species classifications between assessors

A B C D

Figure 3.2. Variation in the application of criteria. Black = % occasions applied correctly, grey = % occasions applied incorrectly. Values are number of times the criterion was applied.

50 Chapter 3 Red List assessment consistency

(a) n = 4 n = 17 n = 5 n = 5

(b) n = 4 n = 17 n = 5 n = 5

Figure 3.3. Proportion of errors made by assessors of different levels of experience against (a) SSC listing and (b) Consensus listing. Horizontal line in box = median, top and bottom of box show the 25 and 75 percentiles respectively, whisker shows 1.5 times the interquartile range. Also see Table 3.4.

51 Chapter 3 Red List assessment consistency

(a)

(b)

Figure 3.4. Proportion of incorrect listings moving species into higher categories of threat (positive values) and lower categories of threat (negative values) for (a) SSC listing and (b) consensus listing. **p<0.01, ***p<0.001 denote the outcome of a binomial proportions test between upward and downward listed species within each level of experience.

52 Chapter 3 Red List assessment consistency

(a)

(b)

Figure 3.5. Magnitude of errors made by assessors of each level of experience for (a) SSC listing and (b) consensus listing. Black bars = single category error, dark grey bars = two category error, white bars = three category error, light grey bars = 4 category error.

53 Chapter 4 Estimating extinction

Chapter 4 When is a species really extinct? Inferring extinction from a sighting record to inform the IUCN Red List

Introduction

The global extinction of a species typically represents the end point in a series of population extinctions, during which, unique evolutionary history is lost at every stage (Pimm et al. 1988; Ceballos & Ehrlich 2002). The loss of the final individual of a species represents the irretrievable loss of a unique form of biodiversity (Diamond 1985; Diamond 1987). Insight into the process of extinction can provide us with the means to identify species at high risk, and highlight groups or regions that may be particularly extinction prone (Pimm et al. 1988; Baillie et al. 2004; Millennium Ecosystem Assessment 2005). Figure 4.1, from the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment 2005) shows how rates of extinction have increased over time, to a current estimate in excess of 1000 times background rate. Rate of extinction and the length of extinct species lists are two markers with which conservation can gauge success (but see Balmford et al. 2003). The severity of the extinction crisis is often expressed in terms of extinct species, with growing lists of extinct species representing continuing loss of biodiversity, whereas the lack of addition to such lists could be seen as indicators of conservation success (Keith & Burgman 2004). Though data and the techniques used to derive current rate estimates have improved, significant uncertainty remains, partly because ambiguity surrounds the extent of extinctions in undescribed taxa, and partly because establishing when the last surviving individual of a species has finally become extinct remains problematic. 784 extinctions have been documented since 1500 AD, but this certainly represents a considerable underestimation of the true extent of extinction during this period (MacPhee & Flemming 1999; Baillie et al. 2004; IUCN 2004).

Extinction is rarely witnessed, and therefore must be inferred (Diamond 1987). If we are unable to do this without bias, then views of biodiversity patterns are distorted. The problem of identifying potentially extinct species is compounded further if a species is cryptic. Scientists are sensibly reluctant to state with certainty if a species is extinct, so as not to facilitate the Romeo effect, giving up on a species too early (Collar 1998) or

54 Chapter 4 Estimating extinction

Lazarus effect, bringing species back from the dead (Keith & Burgman 2004). In essence, scientists must decide how long we should be able to go without seeing a species, before we can safely say it is extinct. The ramifications of getting it wrong can be far reaching. For example, the recent rediscovery of the Ivory-billed woodpecker (Campephilus principalis) in the Big Woods region of Arkansas (Fitzpatrick et al. 2005), led to the immediate allocation of $10 million by US Department of the Interior and Department of Agriculture, for projects to conserve the bird and its habitat (Wilcove 2005). Conservation funds are, understandably, not targeted at species thought to be extinct (Butchart et al. in press), and there are only limited funds to be divided.

Determining whether a species is extinct through extensive survey work may be too costly, and expert opinion may involve considerable subjective bias (van der Ree & McCarthy 2005), so conservation science must strive to find more cost effective and robust alternatives. The technical difficulty in discriminating species extinction lies in answering the question, how long should a species go unseen before we can safely call it extinct (Solow 2005)? Unfortunately the answer is not simple, and relies on the expected rate at which a species would be seen if it were extant. IUCN guidelines state a species should be considered extinct when “there is no reasonable doubt that the last individual has died”, and that this is reliant on “exhaustive surveys in known and/or expected habitat, at appropriate times… throughout its historic range” (IUCN 2001). These surveys must take place over a time scale appropriate to the taxon’s life cycle. Though this is no doubt an improvement on the old cut off which was arbitrarily set at 50 years by CITES (Reed 1996), it remains necessarily ambiguous and open to interpretation.

The problem is analogous to that of estimating time of extinction of a species from its last occurrence in a stratigraphic boundary. It was long assumed that mass extinctions could be read directly from the fossil record. However, the vagaries of the fossil record, differential collection effort and incompleteness of the taxa involved mean it is difficult to reliably infer extinction. Observed stratigraphic ranges are likely to be an underestimate of true range (Benton 1994). A range of techniques have been developed to statistically infer true stratigraphic boundaries (e.g. Strauss & Sadler 1989; Marshall 1994; Solow 2003). In attempting to infer recent extinction conservation biologists are facing similar problems. Methods are required that can reliably conclude whether or not a taxon is extinct from very limited datasets, amongst data that is inconsistent in effort and in taxon

55 Chapter 4 Estimating extinction population size. A number of techniques have been proposed, principally by Solow (see Solow 2005) in order to attempt to make the current qualitative process of classifying species as extinct, more quantitative.

Whilst the current precautionary approach towards the classification of extinction is reasonable given the likely repercussions of getting it wrong, underestimating the number of extinct species will nevertheless bias analyses of extinction. It is therefore appropriate, that techniques which offer a more objective and quantitative assessment of extinction are tested. The principle used in this chapter comes from a technique called optimal linear estimation (Cooke 1980; Roberts & Solow 2003), herein referred to as Cooke’s estimator, a non-parametric method that allows extinction date to be estimated based on the distribution of the k most recent sightings. I test the utility of the technique at two levels; globally and within geographic region. Cooke’s (1980) method was applied to data collated on the sighting records of a number of species of vertebrate taxa, in order to predict whether these species can truly be considered extinct. I address whether such techniques can be used to inform the Extinct and newly proposed Possibly Extinct (Butchart et al. in press) categories of the IUCN Red List, on order to tackle the current underestimation of extinct species.

Methods

Data Sighting records were collated for mammals considered regionally extinct in New South Wales and for Asian birds considered globally extinct. The New South Wales Department of Environment and Conservation Wildlife Atlas of NSW (NSW Department of Environment & Conservation 2005) was the primary source of information for mammals. In this source, data are compiled from a variety of sources including historical reports, DEC staff, survey data and the general public reporting sightings. Whilst these data are extensive, they are certainly patchy, and centred around areas of human habitation, along roads and tracks. However, if a rare species is sighted in the state, it should be reported in this source, and as such is taken as the most reliable available record of known sightings for NSW. For Asian bird species, candidate species information was drawn from (Butchart et al. in press) for Critically Endangered species

56 Chapter 4 Estimating extinction

that were evaluated as Possibly Extinct. Data on sighting records were then compiled from the Asian bird Red Data Book (Collar et al. 2001) and BirdLife’s data zone (www.birdlfe.org) – see Appendix 9.2.

Analysis

The approach used in this chapter to infer time of extinction is based on the Weibull distribution, a two-parameter model. Its origin is in weakest link analysis (Crawley 2002), and thus its use has traditionally been in engineering industrial risk analysis (for example, the failure of a component). The attraction for an analysis aimed at inferring extinction, is due to the result that provided n is large enough (and not necessarily known), it is reasonable to assume that the k most recent sightings has the same Weibull extreme value distribution, regardless of the parent distribution. The following is from

Solow (2005). The optimal linear estimate of time of extinction, TE based on the k most recent sightings has the form:

k ˆ TE = #witn"i+1 (1) i=1

The weight vector w = (w1, w2,…, wk)’ is given by:

! w = (e "# $1e)$1 #$1e (2)

where e is a vector of k 1’s and Λ is the symmetric k-by-k matrix with typical element:

! #(2vˆ + i)#(vˆ + j) " = j $ i (3) ij #(vˆ + i)#( j)

where Γ is the gamma function and:

! k"2 1 t " t vˆ = #log n n"k+1 (4) k "1 i=1 tn " ti+1

is an estimate of the shape parameter of the Weibull extreme value distribution. An

! approximate p-value for testing extinction is given by:

# 1/ vˆ & # T " t & p = exp% "k% n ( ( (5) % T t ( $ $ " n"k+1 ' '

! 57 Chapter 4 Estimating extinction

Finally, under the assumption that the species is extinct, the upper bound of an

approximate 1 - α confidence interval for TE is:

u tn " c(#)tn"k+1 T = (6) E 1" c(#)

where:

! #vˆ $ #log" ' c(") = & ) (7) % k (

Variations in the sighting rate reflect variations in both ‘sightability’, which may be

! connected to abundance, and sighting effort. The main assumption of the technique is that while sighting effort may vary, it never falls to zero, particularly around the time of extinction (Roberts & Solow 2003; Solow 2005). This assumption will be discussed below.

Concerns over the robustness of the method were addressed in three ways. Older sightings are firstly, less likely to be reported and secondly, more likely to be missed during data collection. To test the effect of this, sightings data for the Greater Bilby (Macrotis lagotis: the most data rich species in this study), was assessed by running the data sequence backwards. The three most recent sightings of the 18 available were taken, and the extinction date estimated. Next, the four most recent sightings were taken, and the same process followed, stepping forward through the dataset until all 18 sightings were used. This also allowed evaluation of a second concern, the possibility of elevated Type I errors. The data sequence was run forward with the aim of assessing from the first n observations, whether there was extinction with 95% certainty before the n+1th observation. The third concern was that not all sightings of the species had been collected, either through non-reporting of sightings, or missing the sighting during data collection. Rarefying the full dataset for the Bilby tested its possible effects. 10 dates were selected at random and the date of extinction and confidence limits were estimated. This was repeated 10,000 times and the mean extinction date compared with the estimate from the full dataset.

Extinction dates were estimated for the 10 Asian birds and 10 Australian mammals. Confidence limits surrounding these estimates were calculated. All analyses were

58 Chapter 4 Estimating extinction implemented using R version 1.9.1 (R Development Core Team 2004).

Results

Tests of robustness Running the data sequentially backwards for the Bilby showed that whilst 95% confidence limits are high with few data points, the estimated date of extinction becomes consistent with the use of five data points (Figure 4.2) and remains so up to 18 data points (the full dataset) with 95% CI’s becoming progressively smaller. Assessment of Type I error rate (rejecting the null when it is true) showed the technique predicted extinction of the species before it was next sighted at two points in the dataset (Table 4.1) if the absolute date of extinction is considered. However, 95% confidence intervals were never broken. Rarefication of the Bilby data through randomisation tests also showed congruence between rarefied data and the full set of dates (Figure 4.3). The mean date of extinction estimated from 10,000 repetitions of rarefied sets was the same as the estimated date from the full set (1947). Median estimated date of extinction was higher - 1949.

Estimating date of extinction The most recent sighting of a regionally extinct mammal included in the NSW dataset was 1940. Using Cooke’s estimator, all species except the Bridled nailtail wallaby (Onchoglea fraenata) are predicted to be extinct (Table 4.2), based on estimated date of extinction. The upper 95% confidence interval is high, and only the greater Bilby (Macrotis lagotis) and the Fawn hopping mouse (Notomys cervinus) are predicted to be extinct with 95% certainty. However, if the confidence intervals are relaxed to 80%, five species are predicted to be extinct (M. lagotis, Chaerops ecaudatus, Perameles bougainville, N. cervinus and Pseudomys gouldii), three are not (Bettongia penicillata, O. fraenata and Lagorchestes leporides) and two are close to the present day boundary (B. leseuer and Leporillus conditor).

Of the Asian birds tested, the estimated date of extinction precedes present day in seven species (Table 4.3), four of which are predicted to be extinct with 95% certainty (upper

59 Chapter 4 Estimating extinction

95% CI also precedes present day: Columba argentina, Ophrysia superciliosa, Rhodonessa caryophyllacea and Vanellus macropterus). If confidence limits are relaxed to 80%, Rueck's Blue-flycatcher (Cyornis ruckii) is also predicted to be extinct. The White-eyed River-martin (Eurochelidon sirintarae) and Sulu Bleeding-heart (Gallicolumba menagei) both have upper 80% confidence intervals within the next four decades, so may still be extant. For two species (Otus siaoensis and Ptilinopus arcanus) there was only one record, the type specimen of the species, therefore Cooke’s estimator could not be implemented.

Discussion

Superficially, determining whether a species is extinct might seem a simple task – we either find a species extant, or it is extinct. The reality however, is very different. Not only do taxa considered extinct reappear (e.g. Lowe’s servaline genet, Genetta servalina lowei - Brink et al. 2002; the Omiltemi rabbit, Sylvilagus insonus - Cervantes et al. 2004), but our knowledge of even charismatic, extensively studied taxa such as primates, is still sufficiently poor that new species are still being discovered (Jones et al. 2005; Sinha et al. 2005). Not only does this reveal inadequacies in the way that we have been surveying for species, but also raises the likelihood of an unknown number of species that may have become extinct, before they became known to science. Cooke’s estimator seems to offer a partial solution to help determine extinction in known taxa. For species with sufficient data (>5 sightings), Cooke’s estimator appears to be a useful tool to assess likelihood of extinction, for both regional and global extinctions. For species with low numbers of sightings however, the upper 95% CI’s generally extend so far in the future that they are unhelpful in any meaningful assessment of extinction date. If we are willing to relax these confidence intervals to 80%, they are a little more informative.

The lack of utility of the technique at low values of n does however highlight an important feature of CR species on the IUCN Red List. Almost 13% of mammals listed by IUCN as CR are only known from the holotype. A further 31% are only known from their type locality (see Figure 4.4). For these species there is no substitute for further survey work, and taxonomic re-evaluation. It is impossible to evaluate their likelihood of extinction, other than on a case-by-case basis, assessing the quantity and type of survey work that has been carried out to date, in the region in which they were discovered. This

60 Chapter 4 Estimating extinction lends a particular urgency to the study of their extinction risk. Mammals are relatively well known; for other taxa, estimation of extinction risk may be even harder. There may be a failing of the method though, as well as the data. Cooke’s estimator is designed to consider presence data only, and requires a certain number of data points to work usefully. If its assumptions are met, it appears to work relatively well. But if we consider a hypothetical situation of two point endemics known only from the holotypes – one in the Trans-fly region of Papua New Guinea, the other in Regent’s park, our conclusions about the status of the species ought to be different. However, Cooke’s estimator does not permit consideration of, in effect, sampling effort. The technique would be far more robust if it could consider the history of specimen collection or sightings from the same location, as well as retaining Optimal Linear Estimation’s robustness.

The estimation of Type I error rate is also informative. Type I error appears to be a problem when a series of sightings occur close together, followed by a relatively long gap before the next. In such cases, additional information on sighting effort could be informative. If the intervening period is one of no collection, then the model assumptions are not met. However, such information is often difficult, and in many cases impossible to obtain. A further note of caution is apparent considering recent information on surveys for the pink-headed duck (Rhodonessa caryophyllacea). While the sighting predicts extinction with 95% certainty by 1963, an expedition by BirdLife International in November and December 2004 to Myanmar had possible, though unconfirmed sightings of the species (www.birdlifeindochina.org). Further expeditions to clarify the potential sighting are planned for later this year.

To tackle the problem of potential misidentification of extinct species in the IUCN Red List, Butchart et al. (in press) have proposed a Possibly Extinct (PE) sub category, which could be used to flag CR species that are, on balance of evidence, likely to be extinct, but for which there is a small chance that they may still be extant. In order to make better use of what is sometimes the only available data, it seems that techniques such as Cooke’s estimator may be useful in addition to the framework developed for the PE category, for species that have sufficient sighting information. Some of the results presented in this chapter however, reveal the need for caution.

61 Chapter 4 Estimating extinction

Concerns over effort

One of the main concerns over techniques such as Cooke’s estimator is how they deal with variation in collection or sighting effort. Clearly these are not uniform processes. The advantage of this method is that it explicitly does not assume sighting effort is constant. Variations in sighting rate reflect variation in the ‘sightability’ of the species, which may be connected, amongst other things, to abundance, activity timing, habitat preferences and variations in sighting effort. The basic assumption is that effort does not fall to zero, particularly around the time of extinction (Solow 2005). It is difficult to assess with some early records such as those for NSW mammals, whether this ever is the case. It is certainly unlikely for many species that there was ongoing effort to record them, though retrospective assessment of sightings in some cases is feasible. In certain areas it is possible to trace back to early exploration such as the Horn expedition in 1894, where Baldwin Spencer collected. In other areas it is possible to evaluate sightings back in time by constructing oral histories with local aboriginal communities (Burbidge et al. 1988). These give us a greater degree of accuracy of sighting rates for species which were rare at the time of these early expeditions, in between expeditions and more recent comprehensive survey work. However, the ‘non-zero’ assumption means that Cooke’s estimator cannot be used with most museum archives, where there either is or is not a collection in any given year. Regardless, larger gaps in the sighting record, potentially due to under reporting, will lead to a more conservative estimate of date of extinction. This means that we must be more aware of the desired level of certainty that we are willing to accept, and in cases where it is clear that the assumptions are not met, disregard the method for an alternative. Feasibly, this alternative must be additional survey work, and so evaluation of sightings records could be combined with information on probable distribution, to target co-ordinated and comprehensive future surveys.

Levels of certainty Selecting a level of certainty is paramount to the success of this type of technique. (Reed 1996) highlights the problems of statistical inference in balancing the costs of Type I and Type II error. In this situation, Type I error would mean concluding a species is extinct when in reality it is still extant. The cost to be weighed up is the fact that further action (e.g. converting habitat to an alternative use where it is believed the species is no longer present) could ultimately precipitate the actual extinction of that species. The cost, in this

62 Chapter 4 Estimating extinction case, is determined by the ability of a species to increase from low population levels, something that evidence suggests is likely to be related to reproductive output, population size and density and speed of life history (Forsyth et al. 2004). Species with traits indicative of a slow life history are likely to fare worse (see chapters five, six and seven). The cost of Type II error on the other hand, would be continuing to search for a species that is now extinct; effectively the misallocation of already limited funds. Priority-setting regions, rather than species, therefore focussing searches on areas where there is most uncertainty, could ameliorate this.

Rapid taxonomic change due to species inflation represents a problem for threatened and extinct species lists. With many sub species being elevated to specific level due to a change in species concept rather than new discovery (Hey 2001), conservation biologists should be cautious about their dependency on such lists as accurate biodiversity indicators (Isaac et al. 2004). For potentially extinct species, such changes make tracing back sighting records problematic, as it can become increasingly unclear which species was seen.

There are a number of ways forward. Several different methods of estimating extinction have been proposed (Smith & Weissman 1985; Solow 1993; Burgman et al. 1995; McCarthy 1998; Hall & Wang 1999). The advantage of Cooke’s estimator is that it is better able to place an upper bound on the confidence interval than, for example, likelihood approaches (e.g. Smith & Weissman 1985) or minimum distance approaches (e.g. Hall & Wang 1999), though it could certainly be used in conjunction with other methods that are sensitive to different characteristics of the collection record. Clearly though this method is highly reductive, and should not be used as a sole indicator of extinction, rather it should be used in conjunction with other available information, both quantitative and qualitative. Ultimately, perhaps the IUCN Red List EX category should be altered to reflect a range of certainties (Master 1991; Keith & Burgman 2004) so that uncertainties are more transparent and threatened and extinct species lists are buffered against changes in knowledge or assessment technique. Accounting for bias due to change in knowledge rather than actual change in status, and accurately diagnosing extinction will become increasingly important as we use biodiversity patterns to move towards assessing key objectives, such as the 2010 targets (Butchart et al. 2004; Balmford et al. 2005; Butchart et al. 2005).

63 Chapter 4 Estimating extinction

Tables and Figures Table 4.1. Evaluation of Type I error rate. Date of extinction sequentially estimated for each successive number of sightings starting from the most distant. Discrepancy represents the difference between estimated date of extinction, and the year of next sighting, so negative values suggest Type I error.

# of sightings Estimated date Year of next Discrepancy Upper 95% CI of extinction sighting 4 1975 1879 96 7338 5 1948 1883 65 2789 6 1899 1886 13 2012 7 1895 1893 2 1948 8 1904 1896 8 1957 9 1903 1898 5 1937 10 1902 1898 4 1925 11 1899 1899 0 1909 12 1900 1908 -8 1909 13 1913 1912 1 1934 14 1916 1932 -16 1935 15 1946 1932 14 1989 16 1935 1932 3 1949 17 1933 1932 1 1939 18 1947 1940 7 1970

64 Chapter 4 Estimating extinction

Table 4.2. Estimated dates of extinction for 10 species listed as Regionally Extinct in NSW (NSW Department of Environment & Conservation 2005).

gs n i t gh i

s

# 4 4 18 7 8 4 4 4 4 6

s ’ I 2197 2037 1957 2848 1970 1955 1848 1882 2032 2160 C 80% 1933, 1901, 1940, 2020, 1867, 1860, 1845, 1858, 1906, 1943,

s ’ I 4032 2986 1969 8542 2686 2680 1927 2071 2193 2437 C , 95% 1971, 1896, 1940, 1979, 1863 1857, 1845, 1858, 1897, 1925,

e

at d

on i d t e c n at i

m xt i e t

s E of 1963 1919 1947 2059 1879 1876 1845 1862 1921 1948

t as

l

g n of i

t e at gh i D s 1906 1879 1940 1900 1857 1857 1845 1858 1883 1890

e s l l

i

de a s

i

a i u nv at

l or s r l t di i por nat s i c e nu ue l i

ae e ni audat v s r s ondi goul f r c e bougai e c t e pe l agot e s a l s s c y

e a a e s us

l s i l he s e y l ops c i e ot r i r hogl ongi ongi c am om udom t t e t t r e por ac nc hae ot e e e s p e agor S B B M O C P N P L L

65 Chapter 4 Estimating extinction

Table 4.3. Estimated extinction dates for 10 Asian bird species considered globally extinct (ref). § = two sightings added – see Appendix 9.2.

gs n i t gh i

s

# 9 3 4 3 6 1 1 18 5 7 16

s ’ I 1940 1979 2029 2055 1893 1957 2294 2044 1944 C

80% 1931, 1918, 1988, 1995, 1876, - - 1949, 2003, 1989, 1940,

s

50 ’ I nf 1951 i 2326 26 1923 1963 2869 2116 1949 C

95% 1931, 1918, 1987, 1995, 1876, - - 1949, 1986, 1987, 1940,

of

e

n n at e e d

m m i i d c c e on i t at pe pe c s s

m n i i y y t s xt E e 1934 1923 1995 1997 1880 onl onl 1952 2045 2004 1941

t as

l

g n of i

t e at gh i D s 1931 1918 1986 1995 1876 1866 1953 1949 1964 1987 1940

a e

ac

l i

l ae a us ar

r os

i e nt nage l ophy na i i i e y r a a§ c i r opt

m nt s anus ar i at at r s c i t t i c ge s s k i i ba ac ar upe a ns r r

s don s ar uc m c c i s

r l um a

i aoe s s ba us s i ol he na na l e s y c l i ni m

i nopus c oc l i u l e or us i al phr t ol y ur t hodone ane p ador ador S C C E G O O P R T T V

66 Chapter 4 Estimating extinction

Figure 4.1. Variation in rate of extinction over time, redrawn from the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment 2005). Distant past = average extinction rates estimated from fossil record. Recent past = calculated from known extinctions (lower estimate) or known plus ‘possibly extinct’ species (upper bound). Future = model derived estimates including species-area, rates of shift between threat categories, probability associated with IUCN threat categories, impact of projected habitat loss and correlation of species loss with energy consumption.

67 Chapter 4 Estimating extinction

Figure 4.2. Forward run of data sequence for Macrotis lagotis dataset, starting with the oldest four sightings. Bars are 95% confidence limits.

68 Chapter 4 Estimating extinction

Figure 4.3. Distribution of estimated extinction dates from 10,000 repetitions of the Macrotis lagotis dataset, rarefied to 10 data points selected at random. Dotted line is the estimated extinction date from the full dataset (1947). Mean estimated date of extinction from randomisation test = 1947, median value = 1949.

69 Chapter 4 Estimating extinction

Figure 4.4. Distribution of most recent sighting date for the 162 mammalian CR species (IUCN 2004). Black = species only known from type locality, light grey = species only known from holotype, dark grey = species found more broadly. Data from M. Hoffmann, unpublished – collected as part of the Global Mammal Assessment (Baillie et al. 2004).

70 Chapter 5 Regional extinction patterns

Chapter 5 Regional extinction risk in central Asian vertebrates

The manuscript presented in this chapter is published as:

Collen, B., Bykova, E., Ling, S., Milner-Gulland, E.J. & Purvis, A. (in press). Extinction risk: a comparative analysis of Central Asian vertebrates. Biodiversity and Conservation: Extinction, special Issue

Introduction

Extinction represents the final process in the evolutionary life of a species. Accumulating evidence suggests that present rates of extinction far outstrip those seen in recent history and are comparable to the five mass extinctions observed in the fossil record (May et al. 1995; Pimm et al. 1995; Regan et al. 2001). It is becoming increasingly apparent that the current pattern of threat is often non-random (Raup 1994; Bennett & Owens 1997; McKinney 1997; Purvis et al. 2000a), and that threat characteristically appears clustered amongst groups of similar species or those that possess similar traits, suggesting that there are certain attributes that may predispose species to extinction (Jones et al. 2003). Several studies have tested this hypothesis in a wide range of species at a global level including birds (Bennett & Owens 1997; Owens & Bennett 2000; Jones et al. 2001b), bats (Jones et al. 2001a; Jones et al. 2003), reptiles (Foufopoulos & Ives 1999), carnivores (Woodroffe & Ginsberg 1998; Purvis et al. 2000b; Cardillo et al. 2004b) and primates (Harcourt 1998; Purvis et al. 2000b; Harcourt & Schwartz 2001). All reflect the simple fact that not all species are equally prone to extinction, and that recent extinctions and species declines have been phylogenetically selective. In a world where funding and resources are at best limited, one of the ultimate aims of conservation biology must be to identify such traits which predispose taxa to greater extinction risk, and thus have the opportunity to target restricted funding to those in greatest need, and ultimately act to conserve in a proactive manner.

The IUCN Red List offers a means of classifying species at risk to global extinction, enabling evaluation of relative risk between groups of species, highlighting threatened taxa and promoting their conservation. At a regional level, two options are available to a group wishing to assess species; the first to publish an unaltered subset of the global Red

71 Chapter 5 Regional extinction patterns

List, or second to assess at a local level within a specific region (IUCN 2003). The first option is more straightforward but may not give a true representation of the status of species within a region. The second option presents several problems not experienced with global listing, including how to deal with cross border populations and non-breeding phases that are nevertheless dependent on the region for certain resources.

Recent compilation of a database to assess central Asian vertebrates at a regional level (Milner-Gulland et al. in press) presents an opportunity to test which traits within a group of Asian vertebrates are associated with an elevated risk of regional extinction. The central Asian region includes a variety of high threat (e.g. freshwater-based) and lower- threat (e.g. steppe and montane) ecosystems. It is also an important stopover area for migratory birds, and an area where hunting, and other forms of species utilisation by humans continue. Which species traits are associated with extinction within a region and how do these compare to those at a global level? I examined patterns of threat in a group of central Asian vertebrates using phylogenetic comparative methods. I examine how extinction-promoting traits differ between groups and comment on how results may inform discussion on the effect of scale on studies of threat.

Methods

Data Data were collected for 109 species (40 mammals, 46 birds and 23 reptiles), on the following life history traits: body mass (kg), generation time (years), age at first reproduction (years) and fecundity (average no. of offspring produced by a mature female in a single year). Ecological data was collated on average dispersal distance within the region (km) and regional area of occupancy (km2). All continuous data was logarithmically transformed to equalise variances prior to analysis, except generation time, which was square root transformed so that in the phylogenetic analysis the relationship between contrast and nodal value was minimised (Garland et al. 1992). Data were primarily drawn from the regional database (Milner-Gulland et al. in press) and were selected with the intention of representing a range of taxonomic groups, life histories, endemicity, perceived threat status and knowledge base of the species at the regional scale. Categorical variables were then compiled from the literature on activity

72 Chapter 5 Regional extinction patterns timing (1 = diurnal species, 0 = other activity schedules), sociality (1 = species commonly found in groups larger than one or both parents plus their litter/clutch, 0 = otherwise – for birds, species were scored as social if they were normally indicated as gregarious in Birds of the Western Palaearctic, (Cramp 1980), trophic level (0 = herbivore, 1 = omnivore, 2 = insectivore, 3 = vertebrate eater), habitat (0 = estuarine, 1 = freshwater, 2 = forest, 3 = open, 4 = montane), and whether or not the species was considered a game hunted species (1 = game species, 0 = otherwise; where a game species is defined as a species traditionally utilised for sport hunting, or for meat, fur/skins, horns, fat or traditional medicines). IUCN threat assessments were compiled at a regional level from Milner- Gulland et al. (in press). The IUCN threat categories (v.3.1) were then translated into a variable, treating the levels of threat as a coarse continuous character varying from 0 to 5 (Purvis et al. 2000b) as follows: LC = 0, NT = 1, VU = 2, EN = 3, CR = 4, EX = 5. Data deficient species were excluded from the analysis. A full copy of the dataset is in Appendix 9.3.

Analysis The assumption when performing comparative analyses of multiple species data is that species represent independent data points. If this is not the case then it is necessary to perform some form of phylogenetic correction. I used CONTINUOUS (Pagel 1999; Rambaut & Pagel 2001) to evaluate the strength and statistical significance of phylogenetic signal in our data, by likelihood ratio test. In all cases model A (assuming random walk) was tested. The likelihood was computed in a model where λ (a parameter indicating phylogenetic signal strength) was constrained to zero, and compared to the likelihood of a model where λ was allowed to take its maximum likelihood value. A χ2 test was then carried out to test if λ was significantly greater than zero. If this was the case phylogenetic correction is justified. There was a high proportion of cases where λ was significantly different from zero in our data (Table 1), though as some were not, analyses were carried out using both phylogenetic and non-phylogenetic techniques for comparison.

A composite phylogeny was constructed for each clade. For the birds, this was taken from (Sibley & Ahlquist 1990) and for the reptiles, phylogenetic relatedness was taken from (Estes & Pregill 1988). Interordinal relationships for the mammals were taken from

73 Chapter 5 Regional extinction patterns

(Beck 2003) to establish the relationships between carnivores (Bininda-Emonds et al. 1999), artiodactyls (Price et al. 2005), rodents (Bininda-Emonds et al. in prep) insectivores (Grenyer & Purvis 2003) and bats (Jones et al. 2002).

Using the BRUNCH algorithm of CAIC (Purvis & Rambaut 1995) to select non- overlapping and non-nested sets of taxa across the phylogeny, categorical variables were tested in turn against the response variable (IUCN regional threat status). Using this algorithm, all the contrasts in the predictor are set to positive so that under H0, roughly half the contrasts in the response variable are positive and half negative. Sign tests were carried out on these contrasts to test for significant departures from this expectation, zero value contrasts having first been discarded. For comparison, one-way ANOVA’s were used, for which H0 was that the true mean of response values is identical in all cases of the predictor. All tests were two-tailed and were implemented using R version 1.9.1 (R Development Core Team 2004).

Potential continuous correlates of threat were analysed using single predictor least squares regression through the origin (Garland et al. 1992) on phylogenetically independent contrasts generated using the CRUNCH algorithm of CAIC (Purvis & Rambaut 1995). Branch lengths were set to be equal when generating the contrasts, because preliminary analysis on carnivores (not reported) showed that homogeneity of variances was more closely approached when branches were set to the same length, rather than when set to estimates of divergence time. For other groups divergence estimates were not available. Regression was performed on each predictor in turn. Contrasts having undue influence (i.e. those with studentised t residuals > ±3) were identified and regressions repeated with these data points removed (following Jones & Purvis 1997). Results of this reanalysis never made a qualitative difference so are not reported.

Multiple regression with model simplification was then used to take account of intercorrelated predictor variables, and to generate minimum adequate models (MAMs, Crawley 2002). Significant single predictors were taken as a start group for a multiple regression model. Contrasts for this group were calculated and the maximal model derived. Model simplification was achieved by sequentially dropping the predictor that explained least variance. At each step, contrasts were recalculated in order to prevent missing data increasing the number of possible contrasts. In cases where dropping a

74 Chapter 5 Regional extinction patterns predictor lead to a substantial increase in number of contrasts, previously discarded predictors were introduced and the model retested. Once a MAM was obtained, robustness was assessed by reinserting all discarded predictors back into the model one at a time (including those not in the starting set), to evaluate if additional variance was explained. Non-linearity was then investigated by taking contrasts in x2 and testing if the slope of y on x depends on x, by adding them into the MAM, and simplifying as before. Due to the circularity of listing species under assessment criterion B (based on estimates of geographic range), analyses were repeated with species regionally listed under this criterion removed from the dataset.

Non-phylogenetic analyses were also performed, treating each species as an independent data point, with variables regressed as single predictors. Multiple regression models were constructed using model simplification as above. Additionally, linearity was tested by including a quadratic function and examining the model to see if there was improvement in explanatory power (Crawley 2002).

Results

Table 5.1 shows evidence of a phylogenetic signal in the majority of traits tested; though variation was apparent among groups. λ was significantly greater than zero for all traits except generation time, dispersal distance and fecundity in mammals, for all traits except dispersal distance in birds, and for all traits except threat category in reptiles.

Amongst the 109 threatened species of mammals, birds and reptiles evaluated in single predictor phylogenetic analyses, those with a small area of occupancy were found to be at higher risk of extinction (Table 5.2). Additionally the mammal species with larger dispersal distances were at higher threat (b = 0.39, t26 = 4.29, p<0.001 – also see Figure 5.1). No multiple model was apparent for any of the species groups. Results were slightly altered when species listed under criterion B were removed. For the mammals, area of occupancy became non-significant (b = -0.28, t29 = -1.88, p<0.05) and long dispersal distance remained a significant predictor of elevated extinction risk (b = 0.35, t23 = 3.22, p<0.01. No change in significant predictor variables was observed in the birds, and in reptiles area of occupancy became marginally non-significant (p=0.06).

75 Chapter 5 Regional extinction patterns

For the equivalent analysis using raw species data, which treated all species as independent (Table 5.3), additional single predictors were found to be significant in mammals: body mass (b = 0.18, t40 = 3.20, p<0.01) and age at first reproduction (b =

0.81, t40 = 2.44, p<0.05). Again, results changed slightly when species categorised under IUCN criterion B were removed. Area of occupancy became marginally non-significant

(b = -0.28, t32 = -1.93, p = 0.06). No additional predictors were apparent for birds or reptiles, though area of occupancy became non-significant in birds (b = -0.11, t43 = -1.88, p = 0.07). There was no evidence of non-linearity.

The dispersal distance variable was further investigated in the mammals where species were split into game and non-game species. When only game species were considered, area of occupancy was no longer a significant predictor of threat, and dispersal distance remained significant (Figure 5.1, Table 5.4: b = 0.67, t22 = 4.32, p<0.001). Game species showed slightly different patterns of risk when those assessed under criterion B were removed; in addition to dispersal distance, body mass significantly predicted threat as a single predictor (b = -0.44, t22 = 2.11, p<0.05). In the non-game species, neither dispersal distance nor area of occupancy significantly predicted extinction risk, however body mass did (b = -0.22, t13 =-2.39, p<0.05). No multiple regression models were apparent for either group.

I controlled for dispersal distance using multiple regression (Table 5.5). When all the mammals were considered (Table 5.5a), I found a MAM with dispersal distance squared and area of occupancy as predictors, with species with large dispersal distance and small area of occupancy at higher risk of extinction. However, removal of species assessed under criterion B alters the MAM (Table 5.5b), with large dispersal distance still predicting elevated risk, but independently of the interaction of body mass with generation time, i.e. – species with long dispersal distances and a relatively short generation time for their body size are more threatened.

Sign tests on categorical variables for mammals and birds found no significant predictors of extinction risk, though sample sizes are small. For reptiles, the sample size was too small for tests to have any power. The equivalent non-phylogenetic tests, using one-way

ANOVA, were also all non-significant.

76 Chapter 5 Regional extinction patterns

Discussion

For the 109 species of central Asian vertebrates analysed, I find extinction risk is significantly predicted by dispersal distance in mammals, and by area of occupancy within the region for mammals, birds and reptiles in single predictor analyses. Small area of occupancy may intuitively be associated with increased threat through increasing chances of negative effect by localised stochastic events (Purvis et al. 2000b), and is likely to be associated with other extinction promoting factors such as small population size. Similar traits such as geographic and home range, have been found to predict risk in carnivores (Woodroffe & Ginsberg 1998; Purvis et al. 2000b; Brashares 2003), bats (Jones et al. 2003) and primates (Purvis et al. 2000b). The effect of area of occupancy may not be surprising though, given that some of the criteria used by IUCN to assess threatened status are specifically based on quantitative estimates of area of occupancy or extent of occurrence (criterion B1 or B2). When species assessed under these criteria are removed from the analysis the relationship remains significant in birds, with a similar slope (all bird species: b = -0.16; B1 or B2 species removed: b = -0.17), but not in mammals and reptiles.

If mammals are considered alone, I find a minimum adequate model where species with longer dispersal distances and small areas of occupancy are threatened. This makes intuitive sense and is supported by other studies of mammals (Purvis et al. 2000b; Jones et al. 2003). This relationship perhaps indicates the nature of the species in the analysis, in a predominantly steppe grassland system. The larger dispersal distances of migratory species could result in higher incidence of negative interactions with humans, consequently increased human-wildlife conflict and so greater threat. Longer dispersers may also be more likely to coincide with areas undergoing habitat degradation, though this may be mitigated as dispersal ability means they could disperse their way out of trouble. Further, species with larger dispersal distance may be less likely to have home ranges contained within protected areas and so would be less well buffered against negative human effects, such as habitat destruction, persecution and hunting. However, if species assessed under criterion B1 or B2 (Marmota menzberi, Cervus elaphus and Capra falconeri) are discounted due to the circularity of using area of occupancy as a predictor and a means of assessing risk, I find a minimum adequate model with dispersal distance and the cross product of body mass and generation time. Generation time reflects the

77 Chapter 5 Regional extinction patterns speed of life history of a mammal (Read & Harvey 1989), so the biological interpretation of the model is species with long dispersal distance and relatively low reproductive rate for their body size are more threatened. Similar trends have been seen in a variety of groups with traits indicative of ‘slow’ life history predicting increased risk as species and populations are less able to react to perturbations (Bennett & Owens 1997; Purvis et al. 2000b).

An increased risk of extinction for species with longer dispersal distance is further supported, when the mammals are split into game and non-game species and analysed separately. The positive relationship is highly significant when only game species are considered, but not in non-game species, though an effect of sample size cannot be ruled out. The findings however are contrary to those found in a study of Ghanaian mammals (Brashares 2003). Brashares found a negative correlation between species persistence and the ratio between isolation distance and dispersal distance. He explained this with the fact that isolation limits migration, and therefore species less able to disperse are less susceptible to isolation and thus it follows, more able to persist. Differences between the two studies could result from a scale effect; Brashares’ study examined local populations in protected areas, whereas ours is at a larger, regional level. Further, there are habitat differences between the studies, one based on species from Ghana, in predominantly forest reserves and therefore likely under different threatening processes, and ours on predominantly steppe system.

Body size also predicts status in mammalian game species once B1 or B2 listed species are removed. The relationship is consistent with larger game species being preferentially hunted, and is supported by many studies to date in which large body size predicts an elevated risk of extinction e.g. primates (Harcourt & Schwartz 2001), carnivores (Purvis et al. 2000b; Cardillo et al. 2005) and birds (Owens & Bennett 2000). Owens & Bennett (2000) additionally, show that large body size is only associated with threat in species for which the threatening process is persecution and introduced predators, though not in species for which habitat loss is the major threat. In this study, though large body mass appears associated with game species where persecution is probably the major threat, the direction of the relationship is reversed in non game species, with smaller species at greater risk, though it is not possible to make strong inferences as sample size is small. The relationship is perhaps better explained with the result of the multiple regression

78 Chapter 5 Regional extinction patterns model, controlling for dispersal distance. Rather than body mass or generation time by themselves, it is the interaction of the two traits together that predict elevated risk – so it is perhaps only at larger body mass values that longer generation times lead to elevated risk. The result may also imply size-selective hunting, focusing on bigger species. Certainly it seems to suggest that long generation time is not the only problem of being big. Another, more prosaic, possibility is that generation time does not capture all of the life history effect, and that the non-captured part still correlates with body size. Our findings hint at the fact that it maybe advantageous to try to understand the mechanisms of extinction by process, rather than using overall composite indices that are likely to mask biological traits that may predispose extant species to elevated extinction risk.

Four categorical variables: sociality, activity timing, habitat type and trophic level were tested in the analysis. None were found to significantly predict extinction risk in either birds or mammals. Sample sizes were relatively small, particularly in the phylogenetic analysis where the process of selecting non-nested and non-overlapping sets of taxa reduces the number of contrasts markedly. Such variables have been found to be important in predicting threat in other studies (e.g. Purvis et al. 2000b; Cardillo et al. 2004b).

One potential drawback of this study is that the species analysed are biased by the way they were selected. They do however represent the species for which data are available to make regional IUCN Red list assessments, and thus the species for which relevant data are obtainable. The mammals in the study only represent around 20% of those present in the central Asian region however, with the percentage coverage of reptiles being smaller (22%) and birds smaller still (approximately 7.5%). It would therefore be difficult to make generalisations about how the trends observed relate to the region as a whole. Regional studies like this one may also suffer from analysing species that are on the edge of their range; such species might be at risk regionally but not globally. Indeed, it is possible that some species are represented only by sink populations within the region, dependent upon extra-limital populations for their continued presence.

My findings can perhaps inform discussion of the effect of scale on extinction risk studies – global scale analyses give the biggest sample sizes but are tempered by the heterogeneity of threatening process. A regional scale is better, but yields smaller sample

79 Chapter 5 Regional extinction patterns size and there is still often difficulty in assigning threatening process to a species. At a local scale, studies examine how certain species fare in a given situation with known threats, but multi-species data with a local response variable (e.g. percentage decline in abundance or range) are rare, and therefore generalisations are difficult. A continuing lack of data means that conservation science is unable to answer questions on how causes and rates of decline and vulnerabilities at a local scale compare with loss of species on a global scale. Correlates of vulnerability are often taxon, threat and region specific (Fisher & Owens 2004), regional studies are therefore of value in bridging the gap between multi-species global studies, and more targeted local studies, particularly in light of the absence of multi-species data at a more local level. However in many cases they have the obvious drawback of still not being able to inspect the way in which species react to specific threatening processes. To be of practical use, future regional models should be more narrowly focussed.

80 Chapter 5 Regional extinction patterns

Tables and Figures

Table 5.1. Assessment of the contribution of phylogeny to each of the traits. Values of λ = 1.0 are consistent with the constant variance model (Brownian motion or random walk) being a correct representation of the data, with more intermediate values suggesting the tree topology over-estimates the covariance amongst species. χ2 tests were used to check if λ is significantly greater than zero.

λ

s 40 00*** 00*** 00*** 00*** 63* e l i 0. 1. 1. 1. 1. - 0. t

s p e e i c R e p s

# 23 21 19 20 22 2 23

λ

s 00*** 00*** 97* 66** 00*** 54 79*** d 1. 1. 0. 0. 1. 0. 0. r

i s B e i c e p s

# 45 44 31 43 43 40 45

λ s al 37* 00*** 15 00** 00 37 79* m 0. 1. 0. 1. 0. 0. 0.

s

e am i c M e 001 . p 0 s

< # p 40 40 40 40 39 27 36

***

on , i t 01

0. y e <

oduc nc p y e nc pr a m t e ** i

s r gor t upa

e t c s di

t s

s 05, a r y l on oc i t a i a f C t 0. s

a m t t r <

of r

a a t p

a e undi pe * ne ai e c s

e ge i r r ody hr e

T T B G A F D A

81 Chapter 5 Regional extinction patterns

Table 5.2. Single predictor regressions for the phylogenetic analysis of correlates of threat on regional IUCN threat rating for (a) Mammals, (b) Birds, and (c) Reptiles. Values in brackets denote results when species Red list assessed under criteria B1 or B2 have been removed from the analysis.

)

)

) 03ns 22**) .

) . 88ns 37*) 2 . . 3 97) 60) 58) 46) 03) 74) 24) 63) 24) 02 60) - ( 1 2 ...... 09 ( - - 29) 0 1 1 1 1 0 0 0 0 0 1 . ( ( - 0. ------0 ( - ( ( ( ( ( ( ( ( ( ( ( ( 72 74 38* 63 53 08 71 26 37* 89 49 07 32 38** 55 72 29*** 0. 1. 2. 1. 1. 1. 0. 0. 2. 0. 0. 0. 1. 3.

t 0. - 0. - 4. ------

) 48 09) 24) 50) 47) 11) 15) 12) 32) 35) 45) 11) 07) 10) 48) 37) 05) ...... 0. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

09 24 40 45 09 11 12 30 33 42 11 07 14 65 57 47 08 e

s 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. - 0.

) 24) 75) 28) 20) 46) 36) 33) 03) 17) 06) 11) 01) 60) 11) ...... 04 03) 35) 0 0 0 0 0 0 0 0 0 0 0 0 0 0 . . - 0. ------0 0 ( - ( ( ( ( ( ( ( ( ( ( ( ( (

( ( ( 18 78 24 19 46 36 30 03 16 12 31 04 62 28 05 32 39

0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. b 0. - 0. - 0. ------

s

e i

c 36) 36) 36) 35) 24) 32) 43) 43) 43) 43) 40) 43) 19) 19) 19) 19) 19) e ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 3) p ( #s 40 40 40 39 27 36 46 46 46 46 42 46 23 23 23 23 3 23

s t

as ) r

t 33) 33) 33) 32) 23) 29) 34) 34) 34) 34 32) 34) 13) 13) 13) 13) 13) ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( on 001

#c 37 37 37 36 26 33 37 37 37 37 34 37 14 14 14 14 - 14 0.

on on on i i i ***p< t t t

01, e e e .

oduc oduc oduc 0 e e e nc nc nc

pr pr pr a a a s m m m t t t e e e i i i

s s s r r r t t t al

s

t t t s di s di s di e **p< s s s

l m

s s s r r r y y y l l l on s on i on i i i t t t a a a i i i t a a a f f f t d t t s s s

p a r a a 05, m m m am t t t r r r i r r e r a a a 0.

B M undi undi undi pe pe R pe

ne ne ne

c s c s c s ) ) e ge i O e ge i O e ge i O ody ody ody e e e a) b c ( B G A F D A ( B G A F D A ( B G A F D A *p<

82 Chapter 5 Regional extinction patterns

Table 5.3. Single predictor regressions of species data of correlates of threat on regional IUCN threat rating for (a) mammals, (b) birds and (c) reptiles. Values in brackets denote results when species Red list assessed under criteria B1 or B2 have been removed from the analysis.

) ) ns ns

00**)

. 06***) 3

. 93 88 - 97**) . . 5 . ( 38) 54) 96) 44) 56) 38) 52) 27) 54*) ( 1 1 ...... 3 . - - 24) 41) 90) ( 1 0 0 0 0 0 0 1 2 . ( . . ( ------( 1 0 0 ( ( ( ( ( ( ( ( ( ( ( 47 39* 50 24 78 35* 66 33 46 68 11*** . 20** 43 44* 59*** 86 02 1. 2. 0. 1. 0 2. 1. 1. 1. 1. 4.

t 3. 1. 2. - 5. - - - - 0. 1. ------

06) 27) 41) 39) 08) 14) 09) 34) 29) 29) 10) 06) 10) 40) 40 32) 06) ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

06 26 33 35 08 11 09 32 28 27 10 06 11 45 44 38 06 e

s 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. - 0.

) 54) 28) 05) 32) 13) 11) 06) 15) 21) 40) 18) ...... 09 18) 34) 05) 43) 12) 0 0 0 0 0 0 0 0 0 0 0 . . . . . - - - - - 0. ------0 0 1 0 0 ( ( ( ( ( - ( ( ( ( ( ( ( ( ( ( ( ( 52 25 04 40 22 13 18 60 65 64 26 18 37 81 43 23 10

0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.

b 0. 0. 0. - 0. - - - - 0. 0. ------001

s

0. e i

c 36) 36) 36) 35) 24) 32) 43) 43) 43) 43) 40) 43) 19) 19) 19) 19) 19) e ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 3) p ( ***p< #s 40 40 40 39 27 36 46 46 46 46 42 46 23 23 23 23 3 23

01, e e e .

0 e e e nc nc nc

a a a s m m m t t t i i i ge ge ge

s s s t t t al i a a a s

s di s di s d e **p<

o o l o m

s s s y y y l l l on s on i on t t t a a a i i i or t a a a t pr d t pr t pr t s s s p a e r a e a e 05, m m m am c r r r i r r r r e r r i

0. t t t d B

M undi undi undi pe pe R pe

s s s ne ne ne

e c s c s c s ) r r r ) e i O e i O e i O ody ody ody r i e i e i e a) b c P ( B G F F D A ( B G F F D A ( B G F F D A *p<

83 Chapter 5 Regional extinction patterns

Table 5.4. Single predictor regressions for the phylogenetic analysis of correlates of threat on regional IUCN threat rating in mammals: game species and non-game species. Values in brackets denote species assessed under IUCN Red list criterion B1 or B2 removed.

89**) ) . 3 95) 32) ( . . 30 11*) 1 1 . - - 0.

2 ( ( (

(

63 45 42 39* 14 68 66 32*** 50 44 1. 1. 1. 2. 0. 1.

t 1. - - 4. - - 0. - 1. -

21) 55) 32) 20) 19) . . . . . 0 0 0 0 0 ( ( ( ( (

20 53 34 15 13 09 72 37 08 22 e s 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.

46) . 08) 21) . . 0 44) 78) 1 0 – . . - - ( 0 0 ( (

( (

86 44 18 22 05 37 33 67 37 18

0. 0. 0. 0. 0. 0. b 0. - - 0. - - 0. - 0. -

s

e i c 21) 21) 21) 21) 21) e ( ( ( ( (

p

#s 24 24 24 24 24 16 15 16 6 13

s t 001

as 0. r t 20) 20) 20) 20) 20) ( ( ( ( (

on

***p< #c 23 23 23 23 23 14 13 14 5 11

s 01, e e e . i 0 e c e nc nc

e a a s m m t t p i i e s s t t s i

c

s di s di

e **p< e

s s y y l l on on t t p a a i i or a a t t s t s s

a a 05, m m c r r e gam r r i - 0. d undi undi pe pe ne ne e c s c s am e i O on e i O ody ody r e e P G B F G D A N B F G D A *p<

84 Chapter 5 Regional extinction patterns

Table 5.5. Minimum adequate model for mammals for the phylogenetic analysis of correlations of threat on regional IUCN threat rating. (a) all mammals and (b) with species classified under IUCN Red List criterion B1 or B2 removed. All mammal 2 2 species. Sample size: 27 spp, 22 contrasts. R = 0.53, R adj = 0.49; B1+B2 species 2 2 removed: Sample size 24spp, 22 contrasts. R = 0.45, R adj = 0.40

(a) Predictor Coefficient se t Dispersal distance2 0.07 0.02 3.70** Area of occupancy -0.28 0.09 -3.01**

(b) Predictor Coefficient se t Dispersal distance 0.47 0.11 4.14*** Body mass*generation time -0.30 0.13 -2.27* *p<0.05, **p<0.01, ***p<0.001

85 Chapter 6 Bioregional extinction patterns

Figure 5.1. Contrast plot of relationship between IUCN threat category and dispersal distance in central Asian mammals. Solid line shows regression line for game species (open circles), dotted line for non-game species (closed circles).

86 Chapter 6 Bioregional extinction patterns

Chapter 6 Bioregional patterns of extinction in Australian mammals

Introduction

Australia has had a remarkable record of mammalian extinction since European settlement in 1788. Current estimates indicate at least 19 of the 308 endemic mammals have disappeared, representing the greatest continental proportion of recent global mammalian extinctions (Short & Smith 1994; Smith & Quin 1996; IUCN 2004). Not only does this lend urgency to the study of extinction risk in Australian fauna, it provides a unique model assemblage for identifying the agents and processes of extinction and decline (Cardillo 2003). However, global extinction does not tell the whole story, rather it represents the culmination of a series of population, local and regional extinctions (Ceballos & Ehrlich 2002; Baillie et al. 2004). In order to understand species level extinction risk, it is important to comprehend more localised extinctions (Daily 1997; Ceballos & Ehrlich 2002; also see chapter 7). With some rare exceptions (e.g. Brashares 2003), data on local population decline is not readily available for multiple species assemblages. Until such a time, alternative avenues for analysing local decline and extinction must be sought. In 2000, the Australian Government department of Environment and Heritage launched a framework for conservation planning and sustainable resource management, called the Interim Biogeographic Regionalisation for Australia (IBRA). The intent was to represent a landscape based approach to classifying Australia’s land surface, reflecting a unifying set of environmental influences, which shape the distribution of organisms and their interaction with the physical environment (Australian Government Department of the Environment and Heritage 2005). This recent move to partition Australia into bioregions has provided both an opportunity to evaluate extinction on a more localised scale and the prospect of examining the largest sample of recently extinct continental mammal fauna.

Extinction is a selective process, therefore searching for the target of selectivity is important to designing effective ways to counteract current anthropogenically elevated rates (Purvis et al. 2000c). With few exceptions (Cardillo & Bromham 2001; Johnson et al. 2002; Cardillo 2003; Fisher et al. 2003), studies of extinction risk in Australian fauna

87 Chapter 6 Bioregional extinction patterns have, on the whole, failed to control for phylogenetic relatedness. Therefore, results could be affected by pseudoreplication and non-independence, caused by taxa inheriting characteristics from common ancestors (Harvey & Pagel. 1991). Nevertheless, a number of factors have been highlighted that may be associated with increased levels of risk. These can be logically split into 3 categories: ecological traits, life history traits and environmental factors.

Among the ecological predictors of heightened global risk that have been examined, habitat specialisation (Fisher et al. 2003), dietary specialisation (Lunney et al. 1997; Dickman et al. 2000) and large home ranges (Cardillo 2003) have all been implicated in increasing species susceptibility. Among the life history traits associated with global extinction in Australian fauna, species with larger body mass are hypothesised to be at greater risk of extinction than smaller species (Short & Smith 1994; Cardillo 2003). Body mass correlates with a number of other extinction promoting traits, such as slow life history, low population density and larger home range size (McKinney 1997), and may make taxa more prone to persecution by hunters (e.g. Wilcox 1980; Weaver et al. 1996). The idea of a Critical Weight Range also gained popularity in the Australian conservation literature, where species between 35 and 5500g were believed to have elevated risk of extinction (Burbidge & McKenzie 1989). Cardillo & Bromham (2001) were able to show that this pattern is partly artefactual, due to the distribution of body size in Australian mammals, with extinct and endangered species randomly distributed with respect to body size, except in small species, which are more likely to persist. Moreover, litter size has proved particularly important (Smith & Quin 1996; Cardillo 2003) with the premise that species with smaller litters are less able to respond to population declines via increased reproductive output. Extrinsic environmental factors, such as presence of introduced predators and domesticated grazers, have been identified by several studies as a primary drivers of exacerbated extinction rates (Burbidge & McKenzie 1989; Dickman et al. 1993; Smith & Quin 1996; Lunney 2001; Fisher et al. 2003). Further, the fluctuation of ground water levels have been shown to correlate with the last records of some now locally extinct species, suggesting that rock pile dwelling species may be able to take advantage of rocky areas as important refuges during periods of low ground water (Burbidge & McKenzie 1989; Braithwaite & Muller 1997 – but see Lunney et al. 1997).

88 Chapter 6 Bioregional extinction patterns

Australia has experienced a wave of extinction and decline, ostensibly associated with European settlement, and the consequent spread of introduced predators, grazers and altered land use. Mammal decline has been most pronounced in arid central Australia, with similar losses among states dominated by large areas of arid land (Dickman et al. 2000). Australia’s arid zone, the location of this study, includes the Great Sandy, Little Sandy, Gibson, Tanami Simpson, Sturt’s Stony, Tirari and Great Victoria Deserts, and comprises just over 70% of the Australian mainland. The history of serious biological collection, and thus our knowledge of early mammal assemblages in the region, began in 1894 with the Horn expedition (Baynes & Johnson 1996), although many earlier expeditions took naturalists with them. Central Australia was visited between 1931 and 1935 by Finlayson, and again in 1955 (Finlayson 1961), where he became alarmed at the apparent decline of many of the once common species. Systematic surveys (Burbidge & Fuller 1979; McKenzie & Youngson 1983; Gibson 1986; Gibson & Cole 1988, 1993) and more recently a comprehensive analysis of historical Aboriginal knowledge of the region’s fauna, collected through oral history (Burbidge et al. 1988), have greatly added to knowledge of the region, and document large numbers of faunal declines and extinctions.

In this study I test whether the documented pattern of bioregional extinction can be explained as a random process. To do so I identify patterns of extinction across eight Australian bioregional clusters, predominantly in the low-productivity arid zone of the Northern Territory and Western and South Australia. These present a unique opportunity to test if extinction correlates at a regional level compare to those found in Australia as a whole. I test key ecological, life history and environmental traits to predict order of extinction across bioregional clusters, using methods that control for phylogenetic non- independence of taxa.

Methods

Database & data collection

Bioregional boundaries came from the Australian Natural Resource Atlas Interim Biogeographic Regionalisation for Australia (IBRA) (National Land & Water Audit 2005). Two changes were made to the IBRA bioregional boundaries for this analysis, as

89 Chapter 6 Bioregional extinction patterns it is difficult to assign extinction dates to species in these regions separately. The Pitlands region was created from the amalgamation of IBRA’s Central ranges and Great Victorian Desert, and the Western Deserts region created from the Great Sandy Desert and Gibson Desert bioregions (see Figure 6.1: in the context of this chapter, I shall refer to these new unions as bioregions, though they are no longer bioregions in strict definition). These areas share a high proportion of fauna and habitat. Dates of extirpation were estimated for bioregionally extinct species in these eight bioregions, collated from a range of sources including last sightings, oral history (from Burbidge et al. 1988) and museum collection archives. While date of last sighting is not necessarily an accurate reflection of extinction date (Solow 1993; see Chapter 4), there were inadequate records to implement a method such as optimal linear estimation (Cooke 1980; Roberts & Solow 2003), without drastically reducing the size of the dataset and introducing a further aspect of uncertainty. Therefore order of extinction, reflected by the date of last sighting, was used as the response variable because it is more robust to uncertainties surrounding the actual estimated dates of last sighting (see analysis).

Though numbers of both extant and extinct mammals in each bioregion are freely available (National Land & Water Audit 2005), as yet the identities of the extant species are not. In order to generate extant species inventories for each bioregion, I therefore overlaid bioregional boundaries (available from Australian Government Department of the Environment and Heritage 2005) onto species distribution maps for the world’s mammals (Mammal SuperTeam, 2005; Sechrest 2003; Cardillo et al. 2005) using ArcGIS and the ESRI ‘multi_clip’ python script (http://arcscripts.esri.com/), to generate a list of species found in each bioregion. To guard against inaccurate listing of species in a bioregion due to the more coarse resolution of the global distribution maps, bioregional species inventories were manually compared to the estimated past distributions given in (Strahan 1998) and corrected for obvious anomalies. Full species lists for each bioregion are given in Appendix 9.4.

Where possible, trait data for all species in the analysis were collated from primary literature, augmented with information from secondary literature and the Patheria database (see Appendix 9.4; Mammal SuperTeam 2005, see Cardillo et al. 2005). Key traits were selected a priori to represent three predictor types (life history, ecology, environment). Data were collated on the continuous life history traits: body mass (kg),

90 Chapter 6 Bioregional extinction patterns gestation length (days), age at sexual maturity (years), inter birth interval (days) and litter size; and the continuous ecological variables home range (km2) and current global geographic range (km2). As an indicator of the southern-most distribution of a species range, the most southerly latitude of range was recorded from the global species range distribution (minimum latitude). Globally extinct species (eight species in this analysis) were excluded from this section of analysis because their ranges may have extended further south, which could artifactually affect the association with extinction risk. Categorical variables were generated to represent breadth of habitat and diet. Habitat was categorised into the following: mangrove/shoreline, riparian, monsoon forest, savannah, low open woodland, shrubland, Chenopod shrubland, tussock grassland, hummock grassland, rocky areas, wetlands/floodplains, subterranean and sand dunes. A value of one was assigned for the presence of any of these habitat types in a species range and the total for each species summed, giving a habitat breadth score of 1-13. Diet was categorised into vertebrates, invertebrates, fruit, flower products (flower/nectar/pollen), plant products (leaves/branch/bark/grass), seeds and roots/tubers. Again a value of one was given for the presence of any of these diet components and the total for each species summed, giving a diet breadth score of 1-7. Trophic level was scored (1 = herbivore, 2 = omnivore: vertebrate and/or invertebrate plus any other categories and 3 = carnivore: vertebrate and/or invertebrate only) and whether of not the species utilises rock was scored (1 = rock user, 0 = not). Environmental variables were collated on mean precipitation (mm), mean temperature (°C), mean potential evapotranspiration rate (PET/mm) and actual evapotranspiration rate (AET/mm) across the species geographic range, from the Pantheria database (Mammal SuperTeam 2005).

Inter-correlation of predictor variables can influence results in regression analyses (Crawley 2002), and the limited number of species in an analysis such as this can be problematic as there are potentially a large number of predictors. Analyses on mammals suggest that life-history traits combine into two orthogonal axes representing investment in reproductive speed and investment in offspring biomass (Cardillo et al. 2005; Bielby et al. in prep). Therefore, for further analysis, Principal Components Analysis (PCA) was used to reduce the predictor set to a manageable number of traits within trait types. I used litter size to represent life history output. PCA showed that one of either inter birth interval or age at sexual maturity (which are closely correlated), combined with gestation length represent speed of life history (Table 6.1a). A greater quantity of data was

91 Chapter 6 Bioregional extinction patterns available for age at sexual maturity so this trait was used. PCA of range environment data (Table 6.1b) showed less clear cut results, but given the existing hypothesis of the effect of ground water on Australian mammal extinction (Braithwaite & Muller 1997), precipitation and AET were selected as the most relevant for this test. Data were not available on each trait for all species. Table 6.2 shows the data coverage by bioregion. All continuous trait values were ln transformed prior to analysis to equalise variances, except litter size, which was ln+1 transformed to prevent zero values occurring (many marsupials have a litter size of one). Minimum latitude was untransformed.

Analysis

The pattern of extinction of Australian mammals was first investigated to establish whether it was randomly distributed among families. A random distribution of the 19 extinct Australian species (IUCN 2004) amongst the 308 native species was simulated. Families in which the 19 randomly assigned extinct species were distributed were recorded, and the simulation repeated 1000 times. The null expectation that the observed distribution was not significantly different from random was evaluated using a χ2 test. The same process was repeated to establish whether the 38 (of 112) bioregionally extinct species were randomly distributed amongst families.

Order of extinction within bioregions was assessed using a hierarchical cluster analysis on estimated date of extinction for shared species. Results indicated that four groups, one containing the Pitlands and Finke bioregions, a second with Burt plains, MacDonnell ranges, Tanami desert and Simpson Strzelecki dunefields (herein referred to as BMTS), and a further two with the Western Deserts and Flinders ranges considered separately. This result was used as justification to average extinction dates for shared species within bioregional clusters. Subsequent analyses were performed on these four bioregional clusters.

Phylogenetic relationships of the marsupials were drawn from (Cardillo et al. 2004a) and other mammal orders (bats and murids) from Bininda-Emonds et al. (in prep) with interordinal relationships from (Beck 2003). No branch length data were available; therefore all contrasts were calculated assuming equal branch lengths. The first set of analyses aimed to evaluate key differences between bioregionally extinct species and

92 Chapter 6 Bioregional extinction patterns their extant sister species. The null expectation is that extinct species may have, for example, a larger body mass than their extant sister species, in about half the comparisons. A sign test was used to assess the significance of departures from this null hypothesis. The magnitude of the difference of continuous variables between sister clades is also informative. Wilcoxon’s signed ranks test was used to see if contrasts calculated using algorithms from (Pagel 1992), implemented using CAIC (Purvis & Rambaut 1995), differed significantly from zero.

Categorical variables were analysed against the response variable (order of extinction).

Contrasts for each sister species were calculated using the BRUNCH algorithm implemented by CAIC (Purvis & Rambaut 1995), to select non-overlapping and non- nested sets of taxa across the phylogeny. Using this algorithm, all the contrasts in the predictor are set to positive so that under H0, roughly half the contrasts in the response variable are positive and half negative. A sign test was used to assess the significance of departures from this null expectation for each of the predicted categorical extinction promoting traits. Zero value contrasts were discarded. Again, to assess the magnitude of the difference between sister clades, Wilcoxon’s signed ranks test was used to see if contrasts differed significantly from zero. As my measure of habitat breadth is meaningful on a continuous scale (larger values imply a broader habitat breadth), it was also treated as a continuous variable, though using the more conservative BRUNCH algorithm of CAIC. Wilcoxon’s signed ranks tests were used to evaluate if breadth of habitat significantly predicted order of extinction.

Linear least-squares univariate and multivariate regressions through the origin (Garland et al. 1992) were fitted on independent contrasts calculated using the CRUNCH algorithm implemented by CAIC for species in each of the four bioregional clusters, in the same manner as the heuristic search procedure used in Chapter five and other extinction risk analyses (Purvis et al. 2000b; Jones et al. 2003; Cardillo et al. 2004b). Regression was performed on each predictor in turn against the response variable, order of extinction. Contrasts with undue influence (i.e. those with studentised t residuals > ±3) were identified and regressions repeated with these data points removed (following Jones & Purvis 1997). When removing a contrast made a qualitative difference, the results are reported.

93 Chapter 6 Bioregional extinction patterns

Multiple regressions with model simplification were then used to generate minimum adequate models (MAMs, Crawley 2002). Significant single predictors were taken as a starting set for a multiple regression model. Contrasts for this group were calculated and the maximal model derived. Model simplification was achieved by sequentially dropping the predictor that explained least variance. At each step, contrasts were recalculated and the assumptions of regression (linearity and homogeneity of errors) visually inspected. In cases where dropping a predictor lead to a substantial increase in number of contrasts, previously discarded predictors were introduced and the model re-tested. MAM robustness was assessed by reinserting all discarded predictors back into the model one at a time (including those not in the starting set), to evaluate if additional variance was explained. Non-linearity of predictors was then investigated by adding a quadratic term into the MAM, and simplifying as before. For the life history traits, following Owens & Bennett (2000) treatment of gestation length, the residuals of the relationship between body mass and gestation length, was used as a predictor, due to its covariance with body. All tests were implemented using R version 1.9.1 (R Development Core Team 2004).

Results

Distribution of extinction Variation in extinction risk between Australian mammal families was significantly different from random (Figure 6.2a: χ2 = 37.78, df = 3, p<0.001). For species present in the eight bioregions, the bioregional extinction between families was also significantly different from random (Figure 6.2b: χ2 = 29.01, df = 3, p<0.001). Results from the sister species analysis revealed a significant relationship, independent of phylogenetic relatedness, between bioregional extinctions in species with a larger home range (Table 6.3: Wilcoxon’s signed rank test p<0.05), and with a more southerly distribution (Table 6.3: Wilcoxon’s signed rank test p<0.001), than their extant sister species.

Predictors of extinction pattern

There were no significant categorical predictors of extinction order (habitat breadth, diet breadth, trophic level and use of rock) in any of the four bioregional clusters, though sample size was small (Table 6.4). When habitat breadth was treated as a continuous

94 Chapter 6 Bioregional extinction patterns variable, its effect remained non-significant (Wilcoxon’s signed ranks tests; results not reported). In the southern most bioregion, the Flinders ranges, species are more likely to become bioregionally extinct early, if their continental distribution extends further south

(Table 6.5: b = -0.41, t42 = -5.02, p<0.001). Moving north-westwards through the bioregional clusters, in the Pitlands & Finke cluster, early extinction is associated with small global geographic range (b = 0.44, t38 = 3.32, p<0.01) and low mean AET within a species range (b = 3.61, t36 = 3.55, p<0.01). This relationship remains significant when global geographic range is controlled for in a multiple regression model (Table 6.6). Within the BMTS cluster of bioregions, species with long gestation lengths for their body mass (b = -0.54, t36 = -2.17, p<0.05) and with a more southerly latitudinal distribution (b

= -0.13, t47 = -2.16, p<0.05) appear to suffer earlier bioregional extinction. For species in the Western Deserts bioregional cluster early extinction is associated with small global geographic range (b = 0.26, t46 = 2.52, p<0.05).

Discussion

Patterns of extinction Evidence presented in this analysis suggests that variation in the extinction of Australian mammals is not simply the consequence of random processes. Rather, specific influences of ecological, life history and environmental factors have promoted their demise at a bioregional scale. There are some unifying features across the species in all bioregional clusters, primarily reflecting environmental factors. In agreement with Fisher et al. (2003) this suggests that it is geography, rather than intrinsic biological traits, that principally predicts extinction in Australian mammals. Broadly speaking, and particularly in the south eastern most bioregions in this analysis, it suggests that the impact of threats has been so profound, that species have been impacted regardless of their biology.

Evaluating each bioregion in turn further reveals the pattern of extinction. Following the bioregions north-westwards from the most south easterly (Flinders ranges – see Figure 6.1) in the direction of the advancing front of historical anthropogenic impacts (Flannery 1989; Dickman 1993; Short 1998), aspects of species biology and ecology become increasingly important in more north westerly bioregions, whereas to the southeast,

95 Chapter 6 Bioregional extinction patterns environmental aspects best predict extinction. In the Flinders ranges, where the wave of extinction hit earliest, and subsequently the greatest proportion of extinctions have occurred (60 % of the original native mammal fauna), the southerly limit of distribution of extant species range is the only significant predictor of order of extinction. Smith & Quin (1996) report a significant relationship between body mass and date of extinction in this bioregion. However, when phylogenetic relatedness is controlled for, the relationship is not significant (b = -0.28, t = -0.08, p = 0.93). Moving northwest, in the Pit & Finke bioregional cluster, species with low mean AET in their range are more likely to become extinct earlier. AET reflects the joint availability of energy and water, and so correlates with the energy available to plants, and therefore is regarded as a good index of productivity (Townsend et al. 2000). One plausible explanation is species that became extinct early in the Pitlands and Finke bioregions were those in areas of low productivity, which may have initially impacted herbivores, and had knock on effects to higher trophic levels.

Additionally, species with a small global geographic range are likely to become extinct first within this bioregional cluster. This reveals two things; firstly, in these two bioregional clusters where the major threat processes are grazing pressure and habitat degradation/fragmentation (National Land & Water Audit 2005), environmental variables best predict the order of extinction. A possible explanation is that threatening processes have been so indiscriminate and hard-hitting that extinction has occurred across species, regardless of ecology and life history. This could be the case if ground water levels are a major driver of local extinctions, as suggested by Burbidge & McKenzie (1989). This may also suggest that the breakdown of aboriginal land management practices is important (Burbidge et al. 1988; Burbidge & McKenzie 1989). Secondly, species that have a globally restricted geographic range, which has been found to predict global extinction in a wide variety of organisms (e.g. birds - Jones et al. 2001b; primates and carnivores - Purvis et al. 2000b; hoverflies - Sullivan et al. 2000; see chapter seven) are also likely to go extinct at a bioregional level, possibly because there is often a positive association between global range size and local population density (Blackburn & Gaston 2002). This has implications for the scaling of extinction risk, which I will return to in chapter seven.

96 Chapter 6 Bioregional extinction patterns

In the BMTS bioregional cluster the most southerly distributed species are the most likely to have become extinct early. Additionally, species with long gestation length for their body mass are more susceptible to decline and extinction. In the most north-westerly bioregional cluster, the Western deserts, global geographic range is the only significant predictor of order of extinction. Small geographic range size is hypothesised to increase risk of extinction due to its relationship with small population size (Gaston 1994) and therefore the increased likelihood of a detrimental effect of stochastic fluctuation in environment and demography. Both Fisher et al. (2003) and Johnson et al. (2002) found restricted range Australian marsupials were in fact, less extinction prone, offering the explanation that many are geographically restricted forest species in relatively intact forests, in comparison to the more extensive and more degraded inland arid zones. The vast majority of species in the current analysis are arid zone species, which perhaps explains the significant association with geographic range.

Inferring causes of extinction

The two more northerly bioregions are predominantly threatened by invasive predatory species, while the more southerly clusters are predominated by grazing pressure and habitat fragmentation (National Land & Water Audit 2005). Globally, habitat fragmentation is one of the major threatening processes affecting mammals; combined these threats account for 59.3 % of all recorded mammal threats (Mace & Balmford 2000). Each threat process exerts fundamentally dissimilar selective pressures. Invasive predatory species act by increasing mortality among native fauna, whereas habitat fragmentation acts by reducing environmental carrying capacity. The findings from the BMTS cluster supports the assertion that reproductive rate is an important determinant of risk in the face of invasive species predation (Owens & Bennett 2000), as species that can reproduce faster are more able to sustain elevated predator induced mortality. Small global geographic range predicts bioregional extinction in the two bioregional clusters where habitat fragmentation is the primary threat. This concurs with recent findings in ungulates (Price & Gittleman in review), where geographic range only correlates with extinction risk in species threatened with habitat fragmentation, and not with those threatened by hunting.

97 Chapter 6 Bioregional extinction patterns

Determining causality in the face of multiple threats is complex. Only in the Finke bioregion, does the apparent first extinction preceded cat invasion, thus ruling out cats as the cause for at least some bioregional extinctions. However, some indication is given by the results of the ecological and life history traits. If cats were causally associated with bioregional extinction, it is likely that early extinctions would be those with small body masses within the prey range of cats (Smith & Quin 1996), and it is therefore likely that body mass would be significantly associated with order of extinction. Some authors have argued that cats invaded prior to European settlement (Macknight 1976; Burbidge et al. 1988; Newsome 1995; McKay 1996), however, doubt has been cast on this hypothesis by a comprehensive analysis of the origin and spread of the cat by Abbott (2002). Other authors have queried the role of foxes in Australian regional extinction, stating that fox invasion is unable to explain the significant number of declines that occurred prior to its introduction (Burbidge & McKenzie 1989; Tunbridge 1991).

Low mean AET, perhaps as a surrogate for low productivity (Townsend et al. 2000), is associated with elevated risk of extinction in the Pitlands and Finke bioregional cluster, bioregions that are primarily threatened by habitat degradation and grazing pressure from introduced rabbits and domestic livestock (National Land & Water Audit 2005). Habitat degradation is likely to impact habitat specialists more heavily. Whilst my measure of habitat breadth did not significantly predict order of extinction, the power of the analysis was necessarily low due to the small value of n associated with picking non-nested subsets across a phylogeny in categorical trait analyses. Although multiple drivers of extinction no doubt impact the bioregion, these examples highlight how under different major threatening processes, extinction promoting traits may vary.

Prevalence of bioregional or localised extinctions are likely to be higher in heavily impacted locations, where species may have been exposed to elevated threat levels (Kerr & Currie 1995; McKinney 2001, but see Balmford 1996). Owing to the origin of most threats at points of European settlement in Australia, largely on the coast in the southeast, the bioregions used in this analysis represent a cline in increasing intensity of threat and timing of extinction episodes. This is reflected in the proportion of species extinct in each of the clusters (moving north-westwards, Flinders = 60.9%, Pit & Finke = 50.0 %, BMTS = 41.7%, Western deserts = 30.1%), though to make robust conclusions, area should also be controlled for. Qualitatively, the impact of species biology on extinction

98 Chapter 6 Bioregional extinction patterns patterns appears more important in regions of lower threat pressure. In regions of greater pressure, biological correlates of susceptibility become swamped, such that the only apparent predictors of risk are environmental.

Unfortunately, I was unable to account for other threatening processes that are hypothesised to have adversely affected Australia’s native fauna, such as change in fire regimes (Gill 1989; Vernes & Haydon 2001) and specific information on invasive species (e.g. effect surplus killing, Short et al. 2002; domestic stock distribution, Lunney 2001; impact of foxes, Kerle et al. 1992), principally due to the lack of specific information at a bioregional level. Such threats are undoubtedly important explanatory factors, particularly in the arid zone, in understanding Australia’s local extinction patterns. They should be the targets of future analyses.

Methodological issues & implications for risk scaling

There are several traits that have been shown to correlate with extinction at a global level that do not correlate at a bioregional level (e.g. home range and litter size, Cardillo 2003). This is plausibly an artefact of variation. Any trait relating to order of extinction will have greater force when lineages vary markedly in that trait (Gittleman & Purvis 1998). At a bioregional scale with fewer species, variation will likely be lower, so there is less power than at continental or global levels to find significant correlation. Further, strength of topology varies systematically. Marsupial relationships were generated from 158 source trees (Cardillo et al. 2004a), whereas rodent relationships, for example, are generated from very few, and primarily comprise polytomies, therefore adding little to the analysis. To generalise, even within the marsupials, more is known about better studied groups, and less about rarer taxa and recent divergences. Branch length may also affect the outcomes of phylogenetic comparative analyses (Gittleman & Purvis 1998; Purvis et al. 2000b; Cardillo et al. 2004b), but were unavailable for this analysis.

Certain traits predict extinction at both bioregional and global levels for Australian fauna, and in common with global level studies, this chapter shows extrinsic environmental factors primarily determine interspecific variation in bioregional extinction risk. Therefore, in synthesising my results with these previous studies, the identification of threat processes at a species level is paramount to the further analysis of extinction

99 Chapter 6 Bioregional extinction patterns promoting traits in Australian fauna. The only option available in this analysis was to characterise primary threatening processes by bioregion. While this may hint at potential patterns, it is unlikely to yield clear cut results, as multiple drivers of threat impact such comparatively large areas. However, it does serve to exemplify the heterogeneous distribution of threat across different bioregographic areas, and therefore highlights the need to characterise major threatening processes by species, as well as by region. If a modelling approach such as this is to be of value to conservation, which typically acts at the population level, it is critical that we can establish the relationship between smaller scale management units, such as bioregions, and broader scale global species extinction. In order to rapidly gauge the loss of biodiversity, we must have responsive indicators. If global extinction represents the least responsive indicator of biodiversity loss, then regional extinction is certainly an improvement, but more responsive still, are population declines. Progress in conservation may therefore benefit from an ability to make robust generalisations about the relationship between extinction and decline at these scales.

100 Chapter 6 Bioregional extinction patterns

Tables and Figures

Table 6.1. Component matrix for PCA of traits relating to (a) speed of life history and (b) environment.

(a) Component 1 2 Gestation length -0.051 0.997 Inter birth interval 0.956 -0.082 Age at sexual maturity 0.950 0.136 Variance explained by component 1: 60.61%, variance explained by component 2: 33.96%

(b) Component 1 2 Precipitation 0.627 0.763 Temperature 0.712 -0.670 AET 0.814 0.563 PET 0.792 -0.581 Variance explained by component 1: 54.75%, variance explained by component 2: 42.13%

101 Chapter 6 Bioregional extinction patterns

Table 6.2. Trait coverage for extant and bioregionally extinct species in each of the bioregional clusters. Values represent the number of species for which information on each trait was found, values in brackets are the number of species represented as a proportion of the total in that bioregional cluster. BMTS = Burt Plain, MacDonnell Ranges, Tanami and Simpson Strzelecki Dunefields.

) ) ) ) ) ) ) ) ) ) ) ) ) ) )

s ) ) t t 00 81 63 63 75 00 00 00 00 00 75 63 63 63 63 c r 56 44 n e 1. 0. 0. 0. 0. 1. 1. 1. 1. 1. 0. 0. 0. 0. 0. i s ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 0. 0.

e ( ( xt D E 16 9 7 13 10 10 12 16 16 16 16 16 12 10 10 10 10 16

n

r

) ) ) ) ) ) ) ) ) ) ) ) ) ) ) e

t t s 00 45 28 83 72 26 87 92 89 92 92 00 87 85 85 85 85 . 1. 0. 0. 0. 0 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. an ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

We xt E 53 24 15 44 38 14 46 49 47 49 49 53 46 45 45 45 45 53

) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) )

t 00 72 52 88 72 44 80 88 80 00 92 96 80 72 72 72 72 c n 1. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. i s ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

r xt e E 25 18 13 22 18 11 20 22 20 25 23 24 20 18 18 18 18 25 d

n i ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) l

t F 00 46 37 90 63 34 93 88 51 71 78 98 93 90 90 90 90 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. an ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

xt E 41 19 15 37 26 14 38 36 21 29 32 40 38 37 37 37 37 41

) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) )

t 00 65 59 82 71 65 82 00 00 00 00 00 82 71 71 71 71 c n

1. 0. 0. 0. 0. 0. 0. 1. 1. 1. 1. 1. 0. 0. 0. 0. 0. i e ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

k xt n E 17 11 10 14 12 11 14 17 17 17 17 17 14 12 12 12 12 17 i F

&

t ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) )

i

7 ) t P 00 53 29 88 76 9 94 88 88 91 00 97 94 94 94 94 26 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. an ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 0.

( xt E 34 18 10 30 26 9 33 32 30 30 31 34 33 32 32 32 32 34

) ) ) ) ) ) ) ) ) ) ) ) ) ) ) )

3 ) t 00 55 50 91 64 82 91 95 00 95 95 82 7 73 73 73 c 41 n 1. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. i

( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 0.

S ( xt T E 22 12 11 20 14 9 18 20 21 22 21 21 19 17 17 17 17 22

M ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) )

B t 00 50 31 85 69 26 91 93 78 85 89 00 91 89 89 89 89 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. an ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (

xt E 54 27 17 46 37 14 49 50 42 46 48 54 49 48 48 48 48 54

y nge t i a r

ur l c t

a

va h ude r m phi h

t

e a i l h

l t dt ngt e nt a a h

e dt i l

ve on l

e s

ogr a i ur xua e dt p s e t t h nge e l e t a a br

z or a a p a ge

r s i t on e t r c um r t br i S s l m

t c

t

a bi

i pi e br

t m a r pe a

i

k r t d al e he t c T t T s bi ni e e m oba c e ophi t e i e ge om l a i i E ody oc ot r i r e r E nt P B G I L A H G T H N D R M P T A P T

102 Chapter 6 Bioregional extinction patterns

Table 6.3. Sign tests and Wilcoxon’s signed rank tests on sister clade contrasts for the effect of a number of traits on extinction from a bioregion. # pos = number of contrasts where sister species with a larger value in the predictor are extinct in a bioregion, # neg = number of contrasts where sister species with a smaller value in the predictor are extinct in a bioregion.

Predictor # pos # neg Sign test p Wilcoxon V Body mass 8 8 1.00 73 Gestation length 3 7 0.34 20 Litter size 7 3 0.34 34 Age at sexual maturity 7 5 0.77 46 Home range 10 1 0.01 59* Global geographic range 4 9 0.27 38 Minimum latitude 13 1 0.002 102*** Mean ppt 6 7 1.00 42 Mean AET 7 6 1.00 43 *p<0.05, **p<0.01, ***p<0.001

Table 6.4. Categorical predictors of order of extinction for each bioregional cluster. # pos = number of contrasts where the clade with a larger value of the predictor are extinct earlier in a bioregion, # neg = number of contrasts where the clade with smaller value in the predictor are extinct earlier in a bioregion.

Bioregional Predictor # pos # neg Sign test p Wilcoxon V cluster

habitat breadth 7 4 0.55 39 s

r s diet breadth 6 5 1.00 43 nde i nge trophic level 1 6 0.13 6 l a F r rock user ------only 5 informative nodes------

habitat breadth 8 3 0.23 50 nke

nds diet breadth 4 4 1.00 19 i a l F t trophic level 4 3 1.00 17 i

P & rock user 3 3 1.00 13 habitat breadth 8 4 0.39 58

S diet breadth 5 6 1.00 38.5 T trophic level 5 4 1.00 26 M

B rock user 4 2 0.69 13

habitat breadth 7 3 0.34 39.5 n s r t e r diet breadth 5 3 0.73 21 t e s s e

e trophic level ------only 5 informative nodes------

W D rock user 3 3 1.00 11

103 Chapter 6 Bioregional extinction patterns

Table 6.5. Univariate regressions of predictors against order of extinction for each Bioregion cluster. Tests for gestation length, litter size and age at sexual maturity are all the residuals of the trait on body mass.

Bioregion Trait Nspecies Ncontrasts b se t Cluster Body mass 66 47 0.03 0.06 0.48 Gestation length 37 31 1.11 0.76 1.46 Litter size 59 44 0.11 0.41 0.27

s Age at sexual maturity 44 35 -0.78 0.86 -0.91

nge Home range 25 21 -0.37 0.79 -0.47 a r Global geographic range 58 42 0.49 0.32 1.53 s r Minimum latitude 58 42 -0.41 0.08 -5.02*** nde

i Mean ppt 55 40 1.53 1.61 0.95 l

F Mean AET 55 40 -0.58 2.03 -0.28 Body mass 51 38 -0.19 0.28 -0.67 Gestation length 29 27 -0.60 0.47 -1.26

Litter size 44 34 0.28 0.29 -0.96

nke Age at sexual maturity 38 31 -0.30 0.28 -1.07 i F

Home range 20 17 -0.09 0.65 -0.14 &

Global geographic range 47 38 0.44 0.13 3.32**

nds Minimum latitude 47 38 0.10 0.07 1.37 a l

t Mean ppt 44 35 1.03 0.84 1.23 i

P Mean AET 44 36 3.61 1.02 3.55** Body mass 76 52 -0.15 0.24 -0.62 Gestation length 39 36 -0.54 0.25 -2.17* Age at sexual maturity 51 37 -0.49 0.28 -1.76 Litter size 66 47 -0.12 0.22 -0.54 Home range 23 19 0.14 0.46 0.30 Global geographic range 67 48 0.11 0.16 0.70

S Minimum latitude 67 47 -0.13 0.06 -2.16* T Mean ppt 64 46 -0.08 0.81 -0.10 M

B Mean AET 64 46 0.85 1.15 0.74 Body mass 53 69 -0.26 0.23 -1.12 Gestation length 33 29 -0.29 0.29 -1.01

Litter size 57 43 -0.07 0.22 -0.32 s t

r Age at sexual maturity 48 41 -0.19 0.25 -0.73 e s

e Home range 24 20 -0.55 0.59 -0.93 D Global geographic range 58 46 0.26 0.10 2.52* n r

e Minimum latitude 58 47 0.02 0.06 0.37 t s

e Mean ppt 55 45 0.81 0.77 1.05

W Mean AET 55 45 1.97 1.01 1.96 *p<0.05, **p<0.01, ***p<0.001

104 Chapter 6 Bioregional extinction patterns

Table 6.6. Minimum adequate model of order of extinction for the Pit & Finke bioregion cluster, r2 = 0.33.

Predictor Coefficient se t Global Geographic range 0.60 0.19 3.15** Mean AET 2.46 1.20 2.05* *p<0.05, **p<0.01, ***p<0.001

105 Chapter 6 Bioregional extinction patterns

Figure 6.1. Interim Biogeographical Regionalisation for Australia (IBRA). Bioregions represent a landscape-based approach to classifying the land surface from a range of continental data on environmental attributes. 85 bioregions have been delineated, each reflecting a unifying set of major environmental influences that shape the occurrence of flora and fauna and their interaction with the physical environment. Numbered bioregions used in this analysis are 1 = Tanami, 2 = Burt Plain, 3 = MacDonnell ranges, 4 = Simpson Strzelecki Dunefields, 5 = Pitlands region, 6 = Finke, 7 = Western deserts, 8 = Flinders ranges.

106 Chapter 6 Bioregional extinction patterns

(a)

(b)

Figure 6.2. Frequency histograms of the proportion of (a) extinct species in each family for Australian mammals, and (b) bioregionally extinct species in each family for species in the bioregional analysis. Predicted values are based on 1000 simulations. Black bars are the predicted values, white bars are the observed.

107 Chapter 7 Scaling of extinction risk

Chapter 7 Synthesis: does extinction risk scale from a local to global level?

Summary

Global species extinction typically represents the end point in a long sequence of population declines and local extinctions. While comparative studies of extinction risk in contemporary species have found universal traits that may predispose taxa to an elevated risk of extinction, and local population level studies have given us insight into the process of population decline and extinction, there is still little appreciation of how local processes scale up to global patterns. How frequently do local scale studies, in which the response variable relates to population or metapopulation dynamics, and global studies, in which it relates to species dynamics, come to differing conclusions and why? In an effort to evaluate current understanding, I survey comparative studies of extinction risk. Despite possible differences in the effect of distinct threatening processes at different scales, I find the main predictors of risk scale between local and global levels. Large body size, low population density, small geographic range and a generalist mode of living are consistently correlated with elevated risk of extinction across taxa. I find no evidence that these or other traits tested vary systematically in their association with risk between scales. Some support of a change in significant predictive traits between scales is apparent; home range consistently predicts local, though not global extinction, and slow life history predicts global, but not local extinction. While the ultimate threats to species are anthropogenic, knowledge of intrinsic biological traits can provide a general view of the level of risk that both populations and species may face. However, the interaction of intrinsic species biology with anthropogenic threat remains unclear. Whilst we may speculate on possible impacts, we are largely unable to generalise at a broad scale. Conservation biology must not only describe which species are at risk, and why, but also prescribe ways to counteract this. In order to gauge current depletion, we must analyse decline and extinction of both populations and species. Comparative studies provide a medium to test vulnerability in a wide range of taxa, from a population through to a global level.

108 Chapter 7 Scaling of extinction risk

Introduction

In recent years, conservation research has dramatically increased appreciation of the underlying mechanisms of changing biodiversity patterns. More specifically, phylogenetic comparative analyses have provided conservation biologists with a more rigorous tool with which to explore and understand the underlying processes and patterns of contemporary extinction. However, more than a decade on from May’s (May 1994) discussion of the problem of scale in ecology, a knowledge gap still exists in how population processes scale up to the species level. The majority of attempts at assessing the impact of humans on our planet’s fauna have revolved around the rate and occurrence of species extinction (Myers 1979; May et al. 1995; Pimm & Brooks 2000). However, concerns have been raised at this focus on species, due to the difficulty of quantifying extinction (see chapter four) and the relative insensitivity of extinction rates to short-term changes of human impacts (Balmford et al. 2003; also see Waller & Rooney 2003 and Balmford et al. response). Practical conservation typically acts at the population or metapopulation level, yet studies of extinction are mostly concerned with species decline. Population declines are more sensitive than population extinctions and in turn, population extinctions are more responsive indicators of biodiversity loss than species extinctions (Ceballos & Ehrlich 2002), affording conservation practitioners greater opportunity to introduce pre-emptive, rather than remedial, actions. Therefore, making progress in conservation may require us make robust generalisations about the relationship of extinction between these scales. It would be useful to know if correlates are, in fact, the same between scales.

Throughout studies on taxa as disparate as vascular plants (Schwartz & Simberloff 2001) and the major groups of vertebrates (birds - Bennett & Owens 1997; mammals - Purvis et al. 2000b; reptiles - Lockwood et al. 2002; amphibians – Baillie et al. 2004) one key finding holds true; threat susceptibility is not evenly distributed amongst taxa. Part of this is attributable to the possession of key intrinsic biological traits that predispose species to an elevated risk of extinction; threat is characteristically clustered amongst similar species or those that possess similar traits (Gaston & Blackburn 1995; Russell et al. 1998; Purvis et al. 2000b). Another part is due to the extrinsic nature of threat. If a species is present at a site of high human pressure, it is likely to be exposed to a greater level of threat (Raup 1991; Duncan & Lockwood 2001; McKinney 2001 – but see Balmford 1996).

109 Chapter 7 Scaling of extinction risk

Further, some groups may be restricted to highly disturbed locations, and so may experience a higher than average threat intensity, regardless of their biology. Growing evidence suggests that the interaction between threat (from external pressure) and susceptibility better explains extinction risk, and that distinct traits are associated with vulnerability to each threatening process (Owens & Bennett 2000; Fisher et al. 2003; Cardillo et al. 2004b; Isaac & Cowlishaw 2004).

In spite of such findings, there is little understanding of how local processes scale up to global patterns. Are traits that predispose a species to extinction (the final occurrence in the evolutionary history of a species), the same as those that predispose a population to decline to zero (cumulatively, the cause of a species extinction)? There is some evidence, that with particular traits and threatening processes, alternative outcomes might be expected. For example, the ability to disperse widely may, at a global level, allow a species to disperse its way out of trouble – an association with reduced risk. At a more local level however, the same trait might decrease population persistence – an association with elevated risk. Authors undertaking studies into extinction risk at all scales usually have the same basic conceptual model in mind (pressure × trait = extinction risk; see Purvis et al. 2005a). In practice, the two ends of the scale may not be so comparable as it appears.

It is plausible that two different processes govern species and population extinction, the former mediated by species dynamics, and the latter by meta-population dynamics. Fisher and Owens’ (2004) review of the comparative method in conservation biology details a number of key studies, and highlight some of the important taxon-specific variables that have been tested, across scales. I build on their approach, incorporating additional literature, the findings presented in this thesis, and information on the sign of response which has, to date, not been tested, to ascertain whether consistent patterns of risk-predisposing traits are produced by different studies across a range of spatial scales. I address three questions: do traits examined depend on scale of analysis, do significant predictors of extinction risk depend on scale, and do the signs of those predictors depend on scale?

110 Chapter 7 Scaling of extinction risk

Methods

Data collation A meta-analysis was conducted on all independent variables from 33 comparative studies of extinction risk, covering a wide taxonomic spread (mammals, birds, reptiles, fish and invertebrates). To be included in this survey, studies from primary scientific literature had to use a response variable explicitly intended to capture extinction risk, and use a technique to account for phylogenetic relatedness amongst species. Studies were first divided into local or global level analyses (see Table 1). Global studies were defined as those in which extinction would mean the total loss of a species. Local studies were defined as those in which if the majority of the taxon under study were lost, other individuals of the species would still exist elsewhere in the world, i.e. not species extinction. Local studies more generally evaluated declining vs. non-declining species, or used an index of change in abundance. While this is also true of some global studies, the assumption in these was ultimately that species extinction was the outcome, whereas in local studies, it was not (see Appendix 9.5 for a summary of decisions made).

Due to the wide variety of predictive traits that are tested in comparative analyses of extinction risk, predictor variables were classified into categories, each representing a trait that might predispose a species to elevated risk. These were:

1. Speed of life history – analyses on mammals suggest that life-history traits combine on two orthogonal axis representing investment in reproductive speed, and investment in offspring biomass (Cardillo et al. 2005; Bielby et al. in prep), therefore I consider them separately. A slow life history is predicted to correlate with elevated risk of extinction, as slow reproducing species take longer to compensate for increased mortality with increased fecundity (MacArthur & Wilson 1967). Examples of traits placed in this category include gestation length (e.g. Owens & Bennett 2000), age at sexual maturity (e.g. Jones et al. 2003) and inter birth interval (e.g. Harcourt & Schwartz 2001). 2. Reproductive output – a low output is expected to correlate with high risk, as species with lower offspring numbers are less able to quickly recover from lowered population abundance, for example in response to elevated mortality due to introduced predators, or hunting, and are therefore especially vulnerable to higher

111 Chapter 7 Scaling of extinction risk

mortality. Examples include litter size (e.g. Cardillo 2003) and number of litters per year (e.g. Grenyer & Purvis in press). 3. Diet breadth - a broad diet should imply a generalist strategy and therefore a lower extinction risk. Species with broader diet options are buffered against loss of preferred food items. This may be measured, for example, by number of items in diet (e.g. Harcourt et al. 2002) or the Shannon-Weiner diversity index based on diet consumption (Reed & Shine 2002). 4. Morphological and behavioural specialisation – it is hypothesised that specialist species are less able to adapt to environmental change than generalist species. Morphological specialisations (e.g. wing aspect ratio: Jones et al. 2001a; Jones et al. 2003) or behavioural specialisation (e.g. terrestriality: Isaac & Cowlishaw 2004, forest dependent adult form: Koh et al. 2004a) were the most common reported. These were pooled on the basis that if a species is specialised, however that may be defined, it is more at risk. 5. Habitat specificity - if species are confined to a specific habitat type and are ecologically inflexible, they are at greater risk as they cannot move to alternative areas should their habitat be degraded or destroyed (Norris & Harper 2004). Studies typically measure presence in a particular habitat type as a binary variable (e.g. Duncan & Lockwood 2001). 6. Mating system – certain mating systems may render a species at greater risk to hunting as they make them more obvious to humans (Greene et al. 1998), for example a harem mating system (Brashares 2003). Certain mating behaviours have been found to be significant contributors to population declines and stochasticity (Beissinger 1997), such as defending large territories containing multiple females, particularly in small populations (Calsbeek et al. 2002). Therefore, following Fisher & Owens (2004) my classification includes measures of the extent of sexual selection (e.g. testis size in birds, but not dichromatism, following Morrow & Pitcher 2003, and male combat behaviour in reptiles Reed & Shine 2002).

Other predictors, such as body mass, geographic range, home range and population density were commonly measured and therefore considered independently. For each trait in a study, the direction of association of the trait and measure of extinction risk were recorded, along with type of statistical test (scored as univariate or multivariate) with associated p-values, and sample size. In analyses where multiple taxa had been

112 Chapter 7 Scaling of extinction risk evaluated, separate results were used whenever possible (e.g. primates and carnivores were considered separately in Purvis et al. 2000b). Under one null hypothesis p-values are uniformly distributed between 0 and 1, and if a p value is <0.05 then it is given in a study. If no p-value is presented then it must lie between 0.05 and 1. Its expectation is therefore 0.525. Consequently, where the result of a test was given as ns, or where no p- value was given but the trait was obviously tested, a p-value of 0.525 was recorded as a best guess. Where no direction of association between predictor, and measure of risk was recorded, the result was discarded.

Analysis

The first set of analyses examined the basic differences between local and global studies. I initially tested if traits examined depend on scale, and whether scale had an effect on whether multivariate methods were used, using Fisher’s exact tests. The effect of scale was evaluated against the number of predictors tested, using Wilcoxon’s signed ranks test. Two-way ANOVA was used to test whether sample size varied between scale and predictor variable. Sample size was ln transformed to equalise variances. Then, to test if there were consistent predictors of extinction risk across studies I used Fisher’s technique for combining probabilities from independent tests of significance (Sokal & Rolf 1995).

The next set of analyses were designed to determine whether scale had an effect on the significance of the relationship between the predictive trait and extinction risk. Logistic regression was performed, with a starting set of the variables: scale, sample size and predictor. A maximal model was fitted and simplified, removing non-significant terms to obtain a minimum adequate model. To evaluate change in predictive power during model simplification, the change in significance was evaluated between successively simplified models using a Chi square test in an analysis of deviance table (Crawley 2002). I then modelled the effect of scale on direction of response (i.e. a positive or negative association with extinction risk). Because different predictor variables are expected to have different associations with extinction risk, sign of response was recoded as a binary variable, reflecting whether the sign was in the theoretically predicted direction (1) or not (0). Following Crawley (2002), because unique values of one or more explanatory variables for each individual case were not available, the data were reduced to proportions, and analysed using an Analysis of Deviance. A maximal model was fitted as

113 Chapter 7 Scaling of extinction risk before, and simplified to produce a minimum adequate model. Overdispersion was evaluated, and where appropriate, accounted for using an empirical scale parameter. An F-test was used to compare between the maximal and simpler models. All tests were implemented using R version 1.9.1 (R Development Core Team 2004).

Results

Table 7.1 shows the number and dispersion of studies by taxon and region. Most major vertebrate taxa are covered, along with specific examples of invertebrates (Carabids - Kotze & O'Hara 2003; butterflies - Koh et al. 2004b; hoverflies - Sullivan et al. 2000) and one plant taxon (Vamosi & Vamosi 2005). The first and simplest set of questions concerned basic differences between studies at a global and local level. Figure 7.1 shows the distribution of traits tested globally and locally. Binomial proportions tests between scales for each predictor variable, showed that global scale studies assessed the effect of 2 2 body size (χ1 = 6.32, p<0.01) and traits describing reproductive output (χ1 = 4.36, p<0.05) more than those at a local scale. The remainder of the predictor variables were not significantly affected by scale of study, though among the ecological variables (see Figure 7.1 and Table 7.4), diet breadth was only marginally non-significant (binomial proportions test: p=0.07). Sample size of study varies significantly with scale of analysis and with predictor variable tested; local scale studies have on average, smaller sample sizes (two-way ANOVA, Table 7.2). Number of predictors examined did not vary significantly between scales (Wilcoxon’s signed ranks test: W = 130, p>0.05). Across all variables tested, global studies did not use significantly more multivariate statistical methods, with 7/13 local level studies employing multivariate statistics, compared with 11/20 global studies (Fisher’s exact test: p=0.5).

Fisher’s test for combining probabilities of independent tests indicates there are consistent predictors of threat across studies and within scales (Table 7.3a and 7.3b). In global studies, body mass, geographic range, reproductive output, speed of life history, population density, specialisation, mating system and trophic level all significantly predict extinction risk. At a local level, reproductive output and speed of life history become non-significant, and home range and habitat specificity become significant. There is also some evidence of the importance of multivariate statistical methods, with

114 Chapter 7 Scaling of extinction risk the overall significance of traits often attributable to the more statistically rigorous multivariate analyses (Table 7.3a and 7.3b).

Table 7.4 shows the distribution of number of tests for each predictor of extinction and the direction of association with extinction risk, for local and global studies. Using logistic regression, I find no evidence of an effect of scale on the significance of studies (Table 7.5). There was however, an effect of sample size with significant outcomes tending to result from larger sample sizes, and evidence of consistent predictors of risk, across scales for speed of life history and body size. The final question addressed whether there was any evidence that sign of association of predictor variable with extinction risk, depended on scale. There was no support for an effect of scale (Table 7.6), and there are no indications that the expected sign of the predictors varies between traits. Fisher’s exact tests on differences in the proportion of results giving the expected sign for each predictor showed no difference between scales (not reported – raw proportions given in Appendix 9.5).

Discussion

Evidence of an effect of scale The analyses show very little support for a systematic effect of scale of study on either the significance of the relationship of biological correlates and extinction risk, or the direction of association with extinction risk. The majority of predictors tested have the same relationship with risk, regardless of the scale of study. My results indicate that species have a higher risk of extinction, from local to global scales, if they have a small geographic range, large body mass, low population density or are in some way specialised. At both global and local levels, these traits reflect a lack of resilience to anthropogenic effects through population size (large geographic range and high abundance) or through the absence of more generalist habits offering alternatives when options are limited.

However, there is some evidence of inconsistent patterns between scales, in the types of biological, life history and ecological variables that predict elevated risk. Large home range correlates closely with increased susceptibility in studies of local decline and

115 Chapter 7 Scaling of extinction risk extinction, but the association is not found at the global scale. As an indicator of spatial requirements, this may be because geographic range is a much better proxy for risk at a larger scale, or that statistical analyses are swamped by its greater effect. This is unlikely though as they do not correlate well. A more plausible explanation is that home range at a larger scale is actually a rather crude measure of habitat resource use (Carbone et al. 2005), whereas at a smaller scale it is more indicative of spatial requirements.

The reverse pattern is apparent in studies that measure speed of life history and reproductive output, which consistently predict elevated risk at a global level but never at a local level (though at a local level, n is small). Traits indicative of fast life histories (e.g. short gestation period, young age at first reproduction, short inter birth interval) allow a species to compensate for increased mortality and therefore reduce their risk of extinction (MacArthur & Wilson 1967). In studies of local decline and extinction, it may be that aspects of specific threats, such as human hunting preference, are more important drivers of extinction, which subsequently become masked at a global level, when the ability to reproduce quickly becomes more apparent. Further, choice of site and species for local studies is perhaps more likely to identify ecological than life history traits, as these are more easily measured.

There are also some discrepancies between the types of predictor that are routinely analysed at different scales. Key traits such as body size, geographic and home range, speed of life history, reproductive output and specialisation are tested in the majority of studies across global and local analyses. Some traits, however, are much more prevalent at certain scales. Measures of dispersal and of conspicuousness for example, are tested only at local level. Conversely, the effect of anthropogenic indicators such as human population density and invasive species effects, such as interaction with introduced predators and grazers, are tested in global, but not local scale studies of decline and extinction. In general, local scale analyses show less congruence in the nature of the predictors tested, almost certainly because of the site based nature of these studies.

Methodological concerns While global and local scale studies appear largely comparable, there are some aspects, which reflect a need for caution when interpreting results. Local scale studies have

116 Chapter 7 Scaling of extinction risk smaller sample sizes from which to draw conclusions, inevitably leading to a reduction in statistical power. Further, because smaller scale studies necessarily examine the effect of extinction risk in fewer species, they may suffer from lower variance than analyses conducted at a global level across greater numbers of species in a given taxon. Subsequent loss of statistical power suggests it is more difficult to obtain significant results at this smaller spatial scale, particularly in light of the fact that, in phylogenetic analyses, any trait relating to extinction risk will have greater force when lineages vary markedly in that trait (Gittleman & Purvis 1998).

This survey of studies of extinction risk shows an important difference between local and global scales concerning the identification of the threatening processes affecting the study taxon. Whereas at local scale it is relatively common to determine which threatening processes are acting on the species under study, at a global scale it is rare to do so, highlighting a potential bias between scales. Hence, small scale studies are more likely to focus on problem areas, whereas global studies are in effect, liable to be an average of threatened and non-threatened areas. This has relevance when combining measures of extinction risk across scales. For example, Isaac & Cowlishaw (2004) demonstrate how vulnerability of primates to local extinction is highly variable, depending on the anthropogenic process that threatens it. It is therefore important that we establish exactly which threatening processes are affecting a species, so we can be sure we are comparing like with like. However this is complicated by the fact that most species face multiple pressures, which may interact in complex ways.

As with many areas of biology, ecology and evolution, conservation biology faces a ‘file drawer’ problem (Rosenthal 1979). It is difficult to evaluate the frequency with which non-significant results go unreported. Even when they are, p-values seldom given. This is of particular relevance to meta-analysis because non-significant results are as, if not more informative, than significant predictors. This problem could be partially overcome if sufficient data existed to complete a more formal meta-analysis, which would allow the calculation of how many studies of zero effect size would be needed to make the overall result non-significant (Rosenthal 1991).

Vertebrate taxa are still the focus of the majority of extinction risk analyses (85% of the studies reviewed in this synthesis). While conservation science possesses plentiful

117 Chapter 7 Scaling of extinction risk species lists built in the Linnaean hierarchy, these are no surrogate for phylogenetic information (Mace et al. 2003). It is very difficult to infer how these results may compare to those of insects, plants, crustaceans and fungi, which constitute the majority of species.

Processes & measures

One of the most striking patterns, alluded to above, is that only small scale studies are able to specifically report threatening process invoked. Global level analyses can rarely separate out individual threatening processes (Owens & Bennett 2000 is a notable exception), so results are the product of averaging threatening processes across the study taxa. Conversely, local scale studies appear to target explicit areas where specific threatening processes are apparent. Where they specify, global studies are generally concerned with the comparatively irreversible effect of habitat loss or degradation (Mace & Balmford 2000; Stuart et al. 2004). In specific cases in local level analyses, more reversible threats such as persecution and human conflict (e.g. Woodroffe & Ginsberg 1998) or hunting (e.g. Brashares 2003) are invoked.

Metrics evaluating relative risk of extinction vary with both scale and complexity of study. At their simplest, studies may choose to compare trait values between threatened and non-threatened taxa (e.g. Koh et al. 2004a; Vamosi & Vamosi 2005). Using a potentially more powerful approach, studies may use an ordinal scale of perceived threat e.g. IUCN Red List status (Master 1991; IUCN 2004). At a smaller scale, studies may use a measured species response e.g. population decline in abundance (Brashares 2003), presence/absence (Foufopoulos & Ives 1999), or even a composite measure of risk such as hurricane sensitivity (Jones et al. 2001a). This difference represents the major stumbling block in assessing the scaling of extinction risk. Different response variables are used at different scales. Large scale studies typically employ a proxy or perceived measure of risk, whereas smaller scale studies focus on population decline measures. Using a proxy at a global scale is advantageous as it affords greater statistical power due to increased sample size, and the ability to form more general conclusions. The trade off, however, is that population level declines are generally easier to quantify and more precise than those at a global level (Balmford et al. 2003). Different scale studies may therefore be examining fundamentally different processes - the difficulty lies in assessing

118 Chapter 7 Scaling of extinction risk whether or not we should expect the same outcomes. However, the results of this analysis imply such differences may not matter.

A further complication is that the processes driving declines (e.g. habitat loss or fragmentation) may be different to those driving the final steps to extinction (e.g. over hunting). Since correlates of vulnerability to habitat loss will differ from correlates of ‘small populations’ (Caughley 1994) studies ought to distinguish the stage being evaluated. In practice, at both global and local scales, we are primarily studying declines. In order that local population extinction scales to species extinction, local declines must scale up to species declines. However, metapopulation theory suggests this is not always the case, and high rates of local extinction might not necessarily be linked with a high probability of species extinction (Nee 1994). So different scale studies might be examining different processes, and in doing so are arguably using incomparable metrics.

Conclusions

At the two ends of the spectrum, it is feasible conservation biologists are examining extinction risk in completely different ways, such that we should not expect the outcomes to be the same or even similar. However, in considering the studies available, I find little evidence to suggest that there is a difference in general outcomes globally and locally. Given this, if the results indicated prove correct then large species with restricted geographic range and low population density are universally at greater risk of extinction risk, regardless of the scale at which we focus. This implies a very robust result for at least some extinction correlates. Though ultimate causes of threats to species are anthropogenic, knowledge of intrinsic ecology and life history traits can provide a general view of the level of risk that populations and species may face. What is unclear from many of these studies is the interaction of intrinsic species biology with anthropogenic threat. Whilst we can speculate that overharvesting takes its heaviest toll on slow reproducers, whereas habitat loss is of most detriment to habitat specialists, we cannot make broad scale conclusions. Conservation biology must not only describe which species are at risk, and why, but also prescribe ways to counteract this. In order to gauge current depletion, we must analyse decline and extinction of both populations and species. Comparative studies provide a medium to test vulnerability in a wide range of taxa, from a population through to a global level.

119 Chapter 7 Scaling of extinction risk

Tables and Figures Table 7.1. Survey of studies examining extinction risk using phylogenetic comparative methods. Sample size is the mean averaged over all tests accounting for phylogeny within the study; GR = geographic range; IUCN = ordinal scale of risk based on IUCN threat rating; RDB = Red Data Book; § = more traits available but only one taken.

)

) 1 )

) 5 s

0 ) ) s ) 4 9 ) ) 0 2 5 e 3 0 9 7 0 2 r 0 0 0 ( 0 1 9 0

p

) 0 0 ( 0 2 )

9 0 d b 2 )

2 ( 2 n ) 2

n ( ) 1 2 ( i

o ( 4 2

r ( ( b 0

( ) 4 i o s 0 0 w ) r u t 0 0 s 3 s 0 s a d e t 0 0 3 b w i 0 l 2 o 0 n 0 e h h 2 2 0 ( k k v 0 o e 0

( s ( 2 c n r m 0 .

c c i ( t 2 n 2 . l . l

n i w a u 2 a ( l l o ( y

l a

( e h P . P

w . a a e O V L t

l t

. l

B B r o t t l

e e R a a e e

& e &

a c & & & C

t t t

& & r r t K t i n e n o & e

n t w l e e s u e s n & o l s a r e o y y i r o o i s n & s o c r e c n t e n c v e d e v n r i n l h f s m r a n r r e f n h o s e a w a u r a a u e a u a i o o s R J B G M O C P V D G H I D S F J

s e l b

a i d r e a t

s v 0 4 2 2 0 0 e # t 1 2 1 5 3 1 1 3 7 1 1 8 3 7 1 5

e z i

s e

g e l a p r

e 5 5 6 3 2 2 m v 8 4 6 4 5 7 2 1 4 9 5 4 4

a

A s 1 4 3 6 9 6 2 2 - 5 - 4 5 1 1 2

) s s

e e p )

i i

l l p p ) a a i i s r r p p ( e e s m m p

n n a a a s

4 f f e e x 2 9 0 9 3 g g a 5 3 2 3 0 2 5 1 1 0 5 6 3 t 2 3 0 0 5 7 0 5 1 8 2 6 3 5 4 1 6 2 # 9 2 1 1 9 2 5 4 1 ( 1 6 ( 6 2 3 1 4

n

o ) i s g t e a n i ( d t s

e s n r r e u o p

c / , h

s

o e n s g g g g g g g g g g g g g g r r o & o n n n n n n n n n n n n n n

i u

l p i i i i i i i i i i i i i i t t

d g l t h h h h h h h h h h h h h h t u e t t t t t t t t t t t t t t n u a a c i t y y y y y y y y y y y y y y k c e i e i r r r r r r r r r r r r r r g o s r r b e e e e e e e e e e e e e e g r v a g h v v v v v v v v v v v v v v e o n l T i E E E E H p E E E E E E A E E E E

. k s s

v

i

e e e r e

d

c n n c f e i i R n l l n o n a

a c c G e d c / t e e d i n a d d n r N N N N N N N N N N N

u e t

u r C C C C C C C C C p C p C t b e R b h o a o o U U U U U U U U U U U M I I T n I I I I I I I G A P + I P I

a a & i i

l l

e l l l l l l l l l l l l l a a e a a a a a a a a a a a a a p r r t t p b b b b b b b b b b b b b o A s s r o o o o o o o o o o o o o o l l l l l l l l l l l l l S u u u c S G G G G G G G G G G G G G E U A A

s

s m e r t e a p

m s i h o r i s

i P g s

F n y

& r a s

A e s s s s e R t l l s r

e e

a a a r r u o s s i i & o o e e o v

p p W i i t t v v s

t u u i i a a c e s s s s h a c s s e t n n s s s r r d d d d e x m m t t r r a o r r r r e i i a a s i a a i i i i a a a r r r k n T B B B B B C C D F I P P S B M M

120 Chapter 7 Scaling of extinction risk

Table 7.1. cont/…

) ) )

1 4 1 ) 0 0 0 9 0 0 0 9

) 2 2 2 ) 9

g ( ( ( 4 3 ) 1

r ( s ) 0 ) 0 z n e

t s m r 2 0 0 0 s ) b ) r

e a u 0 2 0 e s 2 a ) b a r

( h ( b 0 0 v n

1 1 b ) p

i I w

r

k 2 2 )

0 0 m a 4 3 ) ( ( h e n c

r G 8

0 0 0 o 0 i ) c 3 . a & p a r e ( 9

2 2 l l 0 0 r 1 0 S ( ( s . n & 9 B H a 2 a 2

B 0 l ' i 0

. ( . ( o

1 t e l l l 0 a h 2 e H & ( . O

f e s & ( l u a a c 2 t & S f t t

e

( a r r e t t o n n o & o o r

& n

l l r e e u u t p e a ) a y & l l a s n

e i i d r e o o o i v e 8 s s h

c

e i f r d e z c s c

l o d d s e e 9 l t n h r

f l r r e r r s l u a

o n n 9 e o a u o a o

a a o e r a u o

o o 1

R C C C K S R H B F D C W ( J K J H N

s e l b

a i d r e a t

s v 7 0 § 0 0 e # t 2 3 6 5 3 1 1 8 4 7 1 2 8 1 1 9 2

e z i

s e

g e l a p r

e 5

m v 2 6 3 0 0 7 5 3 a

A s - 1 2 - 9 1 8 2 1 - - 9 3 - 2 8 2

) )

p

a ) ) p b s n u a i ( r m r

u ) e u a m

( p n x S B 7 s ( ( e a 6 9 4 4 1 7 6 8 9 t g 2 6 0 7 4 5 7 1 5 3 0 0 1 1 2 9 1 # 2 6 1 3 2 7 2 4 3 1 1 1 5 4 5 4 9 ( 1

t

a n t

i a s b

s ) m a o s u l h

e h t (

t

y a s t t + n b s c i

e i e l p b n v

c f s a

o s e s s

o i n

h s s s g g g g g g g g n t d

r c o o o o a n n n n n n n n o

e i

y l c l l p i i i i i i i i i t s l c

g t d i s i t t t h h h h h h h h t u n r c e t t t t t t n t t r l a a a a i a a a t d t t a u y y y y y y y a y k h l e i i i g r r r r r r r r r o v o c t m u r b b b m e e e e e e g e r i e s i t s v o r a u a a h v v v v v v o v v t r e r n n n T i E E E E E E L E H I a E H S H H I d P

y

t y i s s t l i e e i i i v b r r e e i ) a t

o o c c

l

i

e b t g g e n n a s

k c g o e e c e e n n y s t t r n s s n r n i e o i a a p i a e e b b t a s r s o

c c

a a h e g d n n x u g n f / / e i i t s s c z e e e n l l o o e e i

’ ’ n r a t i c c u t ( c c s y t a n n a y c t e e s l b i c c n n a a c i l d i

d d a r N N e e N s n h h r a n B t i s k s r a a n t c C p C p p C a e s e e r r e u l D i x o t t o o r r o U U s U M S S I P R I D P P E R P H L P I I

s s

t d t

o s n s e p a d r l s t s n o i

f a o

s l

u n a s d H r i

a d d e

e

u e r y a n n f g o t i l

a n a a i B c

g s

l l e e

i a s r d a a a o a a r r h +

A l i i i

R r e o e e s e l l l s t

e e l l e a p a a a v e a Z Z t c o a a a p p i r r r i t a b e n e r h t t t d l b r o o d t t d g s s s a c w w r r e m n a o t o e o n n u u u l h e e i u u c r u i u u a e S A A c E E A G p G M N N P P S S S B

s s s l l l a a a

, m m m s m m m d

r a a a s i e b m m m

k

, s s l l l s s a s e

a a a e d e l

n r i i s i i s i s r

a s l s l r r r i o e e d

t t s t f f e i t t b v

r l s s s r m e d

a i a b i l i

s s s e e e e e a t i d a t n r r r s m p t d d d v x m t m t r p r r r r n r r r a a i i p o a a e i a a u i i e e l e a r r e T T T M r C H E P T R B L C B B B P B

121 Chapter 7 Scaling of extinction risk

Table 7.2. Analysis of variance for sample size of local and global scale studies.

Mean Source df. SS square F -ratio p Predictor 8 25.92 3.24 2.81 <0.01 Scale 1 64.64 64.64 55.97 <0.0001 Error 108 124.74 1.16

Table 7.3. χ2 values generated for combined probabilities for predictors of threat across studies, for (a) Global scale studies, and (b) Local scale studies.

7.3(a) All Global predictor records n Univariate n Multivariate n Body mass 76.57** 22 44.21** 12 49.06*** 10 Geographic range 79.28*** 8 39.97*** 4 43.91*** 4 Reproductive output 118.57*** 17 53.91*** 7 64.67*** 10 Speed of life history 73.32** 22 34.54 12 38.78** 10 Home range 5.15 4 3.87 3 one study 1 Population density 30.09** 6 8.61 3 21.48** 3 Morpho+behav. Specialisation 109.57*** 16 49.26*** 9 60.31*** 7 Mating system 24.80** 4 9.03 2 15.77** 2 Diet 17.85 5 7.46 2 10.39 3 Invasive sp interaction 12.34 3 - - 12.34 3 Trophic level 28.67** 5 11.93* 2 21.35** 3

7.3(b) All Local predictor records n Univariate n Multivariate n Body mass 109.18*** 20 49.25*** 9 59.93*** 11 Geographic range 110.44*** 14 54.34*** 10 56.10*** 4 Reproductive output 0.94 2 0.94 2 - - Speed of life history 8.30 3 7.01 2 one study 1 Home range 45.05*** 6 18.06** 2 26.98*** 4 Population density 29.89*** 5 9.21 2 20.67** 3 Morpho+behav. Specialisation 95.39*** 21 61.05*** 11 34.34* 10 Habitat type 20.81** 3 20.81** 3 - - Dispersal 9.67 3 9.67 3 - - Taxon distinctiveness 13.26* 2 13.26* 2 - - *p<0.05, **p<0.01, ***p<0.001.

122 Chapter 7 Scaling of extinction risk

Table 7.4. Correlates of extinction risk from local and global level studies. + and - indicate the direction of association between the trait and extinction risk. Values presented are the number of tests for each predictor of extinction risk, ---- = trait not tested at given scale.

t

gn - - - - -

i - - - - - an s 2 1 1 1 1 1 1 2 1 1 - - - - - c - - - - - i o f i n

gn - - - - -

i

------3 5 1 1 2 1 1 1 s - - - - - 12 ------

al on - - - - -

- - - - - 7 2 1 1 3 1 1 1 1 + - - - - - N ob - - - - - l

G

- - - - -

t

------1 5 4 8 2 2 3 1 ------an c i f i

- - - - -

- - - - - gn 8 1 1 1 1 3 9 1 + - - - - - i - - - - - S

gn t - - -

i - - - s 2 1 2 1 6 1 2 ------an o c i n f i - - -

- - -

- 1 1 1 1 4 1 2 4 1 - - - gn i - - - s -

- - -

al on - - - 4 3 1 1 1 1 1 1 6 + - - - N - - - oc L

t - - -

- - - - 4 2 2 2 1 - - - an - - - c i f i

- - -

- - -

gn 8 1 1 4 6 + i ------S

on i

t s a s s i l ne a i

ve c

i y t p e

y

l s pe t y put

y

or i t s t t nc s i

n i i

s s ab t e n on ns c i nge v. out i i s i ns hi

l t a f ne m l

i de de r a a di ha e e

de

c h ve t

f var c c i ve c ol i

e t i l of be

s e dt pe

e ys l nge z i l on

a uous i s pop

s or e z i a a phi phi e s c i t t c of r s t a a n s c a r ve a nc i pi e a br d oduc t r ude ng pho+ e t i d e e t pe i s t ogr ogr bi pr s e t oup xonom s e ophi e a or i e e om i r a um ody e ons e r a r a pe opul r P B S R M D G G H L P D G H T M C T P H R

e

y s e

s

u

s or u t e r s c e ogy i

al r h i h ol t

c e at ou O f s E p i e S L R

123 Chapter 7 Scaling of extinction risk

Table 7.5. Logistic regression of sign of association with extinction risk, against sample size, predictor of risk and scale of analysis. Coefficient b se z value Sample size 0.004 0.002 2.32* Home range -0.24 0.72 -0.34 Reproductive output -0.55 0.54 -1.03 Speed of life history -1.48 0.46 -3.21** Morpho+behav. specialisation -0.54 0.44 -1.22 Body size -0.85 0.43 -1.97* Geographic range 0.11 0.51 -0.21 Population density 0.59 0.70 -0.85 *p<0.05, **p<0.01, ***p<0.001.

Table 7.6. Analysis of Deviance to evaluate the effect of scale and predictor on sign of response. Coefficient b se z Scale 0.43 0.57 0.75 Morpho+behav. specialisation 1.17 0.54 2.15 Body size 0.89 0.46 1.92 Geographic range 1.49 0.67 2.21 Population density 2.43 0.57 0.75 *p<0.05, **p<0.01, ***p<0.001.

124 Chapter 7 Scaling of extinction risk

s r e h t O

y g o l o c E

e s u

e c r u o s e R

e s u

l a i t a p S

y r o t s i H

e f i L

Figure 7.1. Distribution of traits predicting extinction tested at a global (black bars) and local (white bars) scale. Bars represent proportion of studies assessing the trait. 2-sample test for equality of proportions with continuity correction were carried out between local and global values for each predictor. *p<0.05, **p<0.01.

125 Chapter 8 Conclusions

Chapter 8 Conclusions

Bias in diversity patterns

The research presented in chapters two, three and four all highlight the potential for bias to shape our view of biodiversity. Chapter two consistently reveals how taxonomic description of carnivores and primates has been a biased process with large geographic range size predisposing a species to an increased likelihood of earlier discovery and description. Larger bodied carnivores are also more likely to be discovered early. There are though, some complicating factors; the prevalence of taxonomic inflation (Patterson 1994; Alroy 2003; Isaac et al. 2004; Mace 2004), particularly among the primates (Groves 2001b, 2001a), reveals that the majority of new species are being raised from synonymy or subspecies, so although their ranges will likely be relatively small, they will not be particularly recently described. Taxonomic inflation is biased towards certain groups, so the pattern observed may yet be changed. If it were not biased, taxonomy could be treated as a random error in studies of macroecology and conservation biology. More pressingly, these findings are of particular relevance for hotspot selection algorithms. Area selection algorithms are most sensitive to restricted range species; the very species that this analysis shows are least likely to have been described. This suggests the areas selected by such algorithms may depend critically on survey intensity, rather than actual richness. The implication is that metrics for hotspot selection should be more robust to differences in sampling effort, or should at the very least, consider them explicitly.

Chapter three goes someway to allaying recent criticism levelled at threatened species lists. Although assessors vary in their classification of a species’ extinction risk, variation is not significantly affected by experience, and tends to be relatively consistent with documented Red List classification. Differences of opinion are expected in extinction risk assessment, owing to varying approaches to inference, projection and how precautionary the assessor is. In reality, differences of opinion are resolved through group consensus. The study highlights two issues; the first is the nature of the NT category, and the second is the issue of transparency. Implicitly, it is more straightforward to assess when a species meets a criterion, rather than when it fails to, as

126 Chapter 8 Conclusions is the case for the NT category. Careful consideration is required when applying the category to a taxon, so that assessments are consistent both within and between taxa. Transparency is one of the cornerstones of the IUCN Red Listing process, but this transparency must extend to the reports and action plans produced by Species Specialist Groups in the SSC network. IUCN have been similarly upfront with the online database (www.redlist.org), where names of assessors and evaluators, along with a justification of the species classification are routinely given.

In chapter four I have tested a method designed to make the currently subjective approach to classifying species as EX, a more objective and empirical process. For species with sufficient data, the technique appears to usefully predict extinction date, though for data poor taxa, it performs less beneficially. The technique may augment the type of approach advocated by Butchart et al. (in press), informing the newly proposed Possibly Extinct category, thereby combining an objective empirical method with the subjective process required for such data poor assessments. However, the non-zero assumption of the method remains an issue that needs to be resolved. The technique would benefit from incorporating information on sample effort and sightings from the same location, whilst maintaining the robustness of optimal linear estimation. However, of the CR mammals, 13% are only known from the holotype and 31.5% are only known from their type locality. With such an absence of information, there is no substitute for survey work. Priority setting regions, rather than species, and therefore focussing searches on areas where there is most uncertainty, could be of benefit.

The scaling of extinction risk

Chapters five and six investigate correlates of extinction, firstly at the regional level in central Asia, and secondly at the bioregional level in Australia. Like previous studies, I identify robust predictors of extinction, which supports the assertion that extinction risk is non-randomly distributed among taxa, and is characteristically clustered amongst species that share particular traits. However, in doing so, these chapters highlight the variation that can exist between taxa and regions. Whilst in the group of central Asian vertebrates tested, it appears that aspects of intrinsic biological traits reflecting dispersal ability, range and speed of life history are important in predicting regional extinction, the analysis of Australian mammals reveals that on the whole, extrinsic environmental factors best

127 Chapter 8 Conclusions predict bioregional extinction. Both studies expose the importance of having detailed knowledge of the threatening processes operating, in sub-global analyses of risk.

Phylogenetic comparative analyses have provided conservation biologists with a rigorous tool with which to explore contemporary species extinction. However, the majority of attempts to understand the impact of the current extinction spasm on our planets fauna revolve around rate and occurrence of species extinction (Myers 1979; Lawton & May 1995; May et al. 1995; Pimm & Brooks 2000). These could be considered blunt instruments with which to assess the response of biodiversity to increasing anthropogenic pressure. If evaluating population decline and extinction provides a more responsive indication of biodiversity loss (Daily 1997; Ceballos & Ehrlich 2002), then it would be pertinent to resolve whether local decline and extinction scales to a global level. In chapter seven I synthesise the results of these studies with other phylogenetic comparative analyses of extinction to address this issue. It appears that correlates of local level decline are quite similar to those at a global level. While the ultimate threats to species persistence are anthropogenic, knowledge of intrinsic biological traits can provide a general view of the level of risk that both populations and species may face. However, these analyses highlight the lack of knowledge and study of the interaction between species biology and threat. The ultimate aim of conservation biology must be not only to describe which species are at risk, but also to prescribe ways to thwart their decline. To gauge current depletion and estimate possible future scenarios, we must continue to measure the decline and extinction of both populations and species – comparative analyses provide a robust tool and a rigorous statistical framework with which to do this.

128 Appendices

Appendix 9.1 Data presented for Red List analysis in chapter 3

Species information presented for 10 marsupial species, compiled from Maxwell et al. (1996). Data was given to the assessors along with details of Red List categories and criteria (IUCN 2001).

Case study 1 Species: Sminthopsis archeri Common name: Chestnut Dunnart Family: Dasyuridae

Distribution: Known only from seven specimens collected from the lower Archer River, the Iron Range and near Mapoon on Cape York peninsula (fig. 9.1.1). Also occurs in New Guinea, where little is known of range or abundance. Population: Information on population trends and numbers are unknown. Habitat: Tall stringyback woodlands on red earth soils of the laterite-bauxite plateau, Fig 9.1.1. Current known distribution where canopy species include Erythrophleum of Sminthopsis archeri chlorostachys and Eucalyptus nesophylla , with an understorey of Parinari nonda, Planchonia careyi, Grevillea parallela , and Acacia rothia . Threats: Current threats are unknown

Case study 2 Species: Echymipera echinista Common name: Fly river bandicoot Family: Peroryctidae

Distribution: Restricted to Papua New Guinea. Population: No current population data is available. Habitat: Lowland rain and freshwater swamp forest. Biology: One of the New Guinean spiny Bandicoots (from an Indian word meaning literally “pig-rat”), small to medium-sized ground-dwelling marsupials with a long pointed heads and compact body. Long hind legs and short front legs give them a resemblance to kangaroos or rabbits. Locomotion by hopping and gestation is brief, lasting as little as 12. 5 days. Threats: Intrinsic factors, restricted range. Protection: None at present.

129 Appendices

Case study 3 Species: Dasyurus maculatus Common name: Spotted-tailed Quoll Family: Dasyuridae

Distribution: Found in southeast Queensland, this species is rare having undergone a large contraction of its former range. Extant from Eastern NSW to as far west as Warrumbungles NP. Most abundant populations believed to be in north-eastern NSW.

Population: Total population estimated in 1996 to be approx 9000 individuals. Reasonably abundant in a number of localised areas, but severely fragmented. Decline rate estimated at 10-15% between 1977 and 1987. Largest subpopulation estimated at 750 individuals. Habitat: Forests, woodlands, wet forest alliance, rainforest, coastal heaths and coastal wet scrub, estuarine areas and rocky Fig 9.1.2. Current known distribution of headlands. D. maculatus. Biology: Up to 4kg in weight, solitary and primarily nocturnal. Threats: Historic threat during the 1900s include decline in habitat quality (fragmentation), human induced habitat loss (habitat loss), intentional poisoning, invasive species (predators), accidental mortality (Strychnine baiting for dingoes). Recently non- target mortality from trapping and poisoning, but also direct persecution is depleting populations. Estimated forest loss as a result of clearing within its former range in south- east Queensland is over 70%, with the majority of loss occurring over the last 20 years.

130 Appendices

Case study 4 Species: Burramys parvus Common name: Mountain pygmy possum Family: Burramyidae

Distribution: Unlike other arboreal or scansorial members of the family, B. parvus is terrestrial, and confined in its distribution to areas above the winter snowline (approx. 1370m asl). Due to the receding snowline since the last glacial, it is now restricted to 3 genetically isolated populations, separated by deep river valleys in south eastern NSW and Victoria (fig 9).

Population: NSW population distributed in patches confined to an area 30km by 8km. Kosciuszko NP supports a second population in around 8 km2 and a third population in Victoria is supported by around 2 km2 of intact habitat. Habitat: Rocky areas with associated shrubby heathland characterised by mountain plum-pine Podocarpus lawrencei. Figure 9.1.3. Current known distribution of

Burramys parvus. Biology: Average adult weight 40g. B. parvus is nocturnal and has a diet composed of seeds, fruits, worms and other invertebrates. This species may not be strictly arboreal but is an adept climber that is terrestrial. Threats: This species' habitat is highly restricted and has been fragmented or destroyed by road construction, dam/aqueduct construction and development of the infrastructure for the downhill skiing industry. The habitat is subject to weed invasion and fox and cat predation may pose a potential that is as yet to be substantiated. Protection: A government-endorsed Management Plan exists in Victoria. Monitoring of populations within ski resorts and control areas has been initiated.

131 Appendices

Case study 5 Species: Potorous gilbertii Common name: Gilbert’s Potoroo Family: Potoroidae

Distribution: Currently known to occur in small, restricted pockets in Two Peoples Bay Nature Reserve, Western Australia. Population: The single population known in Two peoples Bay Nature Reserve is estimated to contain around 40 individuals all contained in one sub- population. Habitat: Occurs in dense thickets and rank vegetation bordering swamps and Figure 9.1.4. Current known distribution of running streams. P. gilberti. Biology: Small nocturnal rat- kangaroo, bearing some resemblance to a bandicoot. The body, but not the tail, is densely furred. Weight 0.9 – 1.2 kg. Threats: Reasons for the decline remain unknown. Predation by foxes and cats, and changed fire regimes resulting in loss of dense vegetation, may have led to a decline in Gilbert’s potoroo. Dieback caused by Phytopthora cinnamomi may also be a serious threat to the survival of the species by altering the vegetation structure or eliminating plants, which are food sources themselves or hosts to the mycorrhizal fungi on which the potoroo feeds.

Case study 6 Species: Petrogale sharmani Common name: Mt Claro Rock-wallaby Family: Macropodidae

Distribution: This species is restricted to the Seaview and Coane Ranges, west of Ingham in north-eastern Queensland, Australia (an area of about 2,000 km2). Common at some localities. Population: No population data available Habitat: Rocky outcrops, boulder piles, gorges, cliff lines and rocky slopes Threats: No decline observed, but extremely restricted distribution makes this species vulnerable to habitat loss and degradation and/or chance events. Threats include habitat loss for development, and competition from domestic and wild introduced herbivores. Vulnerable to possible effects of any climate change which favours more populous species of rock-wallabies with adjacent ranges.

132 Appendices

Case study 7 Species: Bettongia pencillata Common name: Brush-tailed Bettong Family: Potoroidae

Distribution: Population numbers and range have decreased dramatically during this century, it being recorded as extant only in the "western Centre" and south-west WA by Finlayson (1958). It had disappeared from central Australia by 1960 (Burbidge et al. 1988). The St Francis Island population (of B. p. penicillata ) became extinct some time Figure 9.1.5. Current distribution of B. between the 1920s and 1971. pencillata Despite unconfirmed sightings on the Eyre Peninsula in the 1980s (of B. p. penicillata), it is believed the species survives naturally only as populations of B. p. ogilbyi in south- west WA. Natural populations remain at Dryandra Woodland, Perup NR (and adjacent areas as far as Kingston Forest and near Lake Muir), and at Tutanning NR. Sightings in Fitzgerald River NP have not been confirmed, despite extensive trapping.

Population: Translocated populations occur at Batalling Forest, Boyagin NR and Julimar CP (WA) and Venus Bay Island A, Wedge Island, St Peter Island, mainland Venus Bay CP and Yookamurra Sanctuary (SA). Twenty translocations into areas within the northern Jarrah forest of WA took place in 1995 as part of research into the effectiveness of different rates of aerial fox baiting (Operation Foxglove), and another translocation took place to Karakamia Sanctuary. These translocations are too recent for their success to be measured. Habitat: Forests and open woodlands Threats: Fox and cat predation and habitat destruction and alteration, including changes to fire regimes, competition with domestic and feral introduced herbivores.

133 Appendices

Case study 8 Species: Bettongia tropica Common name: Northern Bettong Family: Potoridae

Distribution: Known from three localities: (1) The western side of the Lamb Range between Tinaroo and Mareeba, in an area about 15 km x 2-3 km in extent. Recent records exist for (2) Mt Windsor and (3) Carbine Tableland, both are about 60 km north of the Lamb Range population – both the same size as Lamb range. The species appears rare at the latter two sites, and a Figure 9.1.6. Current distribution of B. recent resurvey of the Mt Windsor site failed tropica to detect any animals. Population: Detail unclear, but large fluctuations known to have taken place since the 1950s - the Dawson Valley Queensland population is almost certainly extinct. Habitat: Predominantly open woodland with a grassy understorey, on granite-derived soils, some of which are of low fertility, and close to the western edge of wet tropical rainforest. Threats: Unknown, but probably intensification of land use, and possible changed fire regimes.

Case study 9 Species: Dactylopsila megalura Common name: Great-tailed Triok Family: Petauridae

Distribution: Central and western New Guinea. Population: Population decline of 40% observed over the past 10 years. Biology: This species inhabits rainforests and is nocturnal and arboreal. During the day Dactylopsila nests in dry leaf nests in hollows. The diet is largely composed of insects, fruits and leaves. Threats: Population decline caused by exploitation – direct action in the form of management intervention appears to have halted this decline.

134 Appendices

Case study 10 Species: Setonix brachyurus Common name: Quokka Family: Macropodidae

Distribution: Area of occupancy on the mainland has probably been reduced by at least 50% over the last 50-60 years and would not exceed 22,000 km2.

Population: Population size estimated at 9000 mature individuals, with a predicted continuing decline rate of 15% over the next 10 years. Figure 9.1.7. Current distribution of Habitat: restricted to swamps with dense Setonix brachyurus vegetation. Threats: Much of the decline coincided with the arrival of the fox in the south-west of Western Australia in the late 1920s and they probably now occupy a narrower habitat niche than previously with predation confining remaining mainland populations to dense vegetation. Clearing and burning of remnant swamp habitat has probably also contributed to their decline through increased exposure to fox predation.

135 Appendices

Appendix 9.2 Details of sightings record presented in chapter 4

Data compiled on sightings record for regionally extinct NSW mammals (NSW Department of Environment & Conservation 2005) and globally extinct Asian birds (Collar et al. 2001).

Species Sighting record Bettongia penicillata 1857, 1863, 1882, 1906 Bettongia leseuer 1879, 1857, 1890, 1941 Macrotis lagotis 1844, 1844, 1845, 1857, 1879, 1883, 1886, 1893, 1896, 1898, 1898, 1899, 1908, 1912, 1932, 1932, 1932, 1940 Onchoglea fraenata 1840, 1841, 1857, 1900 Chaerops ecaudatus 1838, 1844, 1845, 1857 Perameles bougainville 1800, 1840, 1841, 1857 Notomys cervinus 1800, 1845, 1845, 1846 Pseudomys gouldii 1840, 1850, 1856, 1858 Leporillus conditor 1844, 1844, 1857, 1883 Lagorchestes leporides 1857, 1890, 1863, 1863, 1863, 1857 Columba argentina 1899, 1901, 1902, 1908, 1913, 1925, 1928, 1930, 1931 Cyornis ruckii 1880, 1917, 1918 Eurochelidon sirintarae 1960, 1978, 1980, 1986 Gallicolumba menagei 1891, 1945, 1995 Ophrysia superciliosa 1836, 1865, 1868, 1869, 1870, 1876 Otus siaoensis 1866 Ptilinopus arcanus 1953 Rhodonessa caryophyllacea 1903, 1905, 1908, 1910, 1916, 1917, 1921, 1922, 1923, 1924, 1925, 1930, 1932, 1935, 1940, 1947, 1948, 1949 Tadorna cristata 1877, 1882, 1914, 1916, 1964, 1983§, 1987§ Vanellus macropterus 1872, 1907, 1909, 1914, 1916, 1917, 1919, 1920, 1921, 1922, 1923, 1925, 1927, 1938, 1939, 1940

§ - The reports of sightings in China are unclear. They state: “Heilongjiang unnamed forest river in the Da Hinggan Ling (Greater Xingan mountains), near the border of Heilongjiang and Inner Mongolia, where two were seen by an old forester and hunter (with 20 years experience), May 1987 (Zhao Zhengije 1992, 1995); Da Hinggan Ling (Greater Xingan mountains), female seen by taxidermist, April 1983 (Zhao Zhengije 1992).” Several of the reports listed in Zhao Zhengjie (1992) were not included in Zhao Zhengjie (1995). It is unclear whether these omissions were deliberate, and based upon a re-evaluated of the data, or accidental (Collar et al. 2001). The analysis was therefore performed both with, and without the sightings.

136 Appendices

Appendix 9.3 Species data presented in chapter 5 Data presented in regional analysis of extinction risk in Asian vertebrates (chapter five). Regional threat coding follows (Purvis et al. 2000b). Continuous variables are ln transformed, except for generation time, which is square root transformed. See text for trait definitions.

Reptile? 0 1 0 0 1 1 0 0 0 0 0 0 1 0 0 0

Bird? 0 0 1 0 0 0 1 1 1 1 0 1 0 0 0 0

Mammal? 1 0 0 1 0 0 0 0 0 0 1 0 0 1 1 1

Endemic? 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0

Diurnal? 1 1 1 0 0 0 1 1 1 1 0 1 0 1 1 0

Social? 0 0 0 0 0 0 1 1 0 0 1 1 0 1 1 0

Trophic level 3 1 3 0 2 2 0 0 3 3 2 0 2 0 0 3

Habitat 4 4 4 4 4 4 2 2 4 4 4 1 4 4 4 4

1 1 8 9 0 5 3 8 8 9 2 0 5 0 1 9

Area of 5 1 7 7 4 1 5 6 8 6 6 9 . . . . . 6 . . . . . 9 2 . 9 . . . . . 1 4 2 2 0 3 2 3 3 0 1 2 occupancy 7 9 6 6 1 1 1 1 1 1 1 1 1 1 1 1

9 0 1 6 0 0 1 6 9 0

Dispersal 6

7 0 3 3 6 9 1 0 0 ...... 0 5 8 5 5 7 3 8 4 3

distance -

2 9 0 5 0 9 9 9 9 0 9 8 9 0

9 6 0 2 1 6 7 7 6 0 3 1 6 1

Fecundity ...... 0 1 0 1 1 0 1 1 0 0 1 0 0 1

9 0 0 9 0 0 9 0 9 9 9 0

Age at first 6

1 0 6 1 0 3 1 6 6 6 1 ...... 0 1 0 0 1 0 1 1 0 0 0 1

reproduction -

Generation 4 5 0 1 0 4 4 3 0 5 4

2 9 0 4 0 2 2 8 0 6 2 ......

time 2 0 1 1 1 2 2 2 2 2 2

9 1 0 5 6 0 9 3 9 0 0 0 5 0 2 6 2 2

7 4 3 6 0 5 1 8 0 3 6 1 . . . Body mass ...... 1 7 7 3 0 2 0 1 1 1 2 0 3 3 2 - - -

Regional threat 5 0 1 0 2 3 2 3 1 2 1 2 3 3 3 3

s i s

i i a s l s

d i u l e n t

v e e

s i m a i

o

f s l u

o k s z s s s

u u l t i s i h c s i t u r u j i c

r a t v r c l u u e o p

d r e l c a e e a r a e s a e h o v a c l o a a n

b g n t s

Binomial r u e l b c r

a s o y u a o i a s c i h u x j y a f l

t i r l c x t g

l l c a m y e u a d h a e x l m e s l a r r l c h t g y n s a f e i y

u e i

s a a n u n h h a a a p t i a a c a t r r l l o o p p c r r i i o p a i e e b n n o o a p p r r y s s u u r n i a l s s a a a c g g l l l n n q q a r u A A A A A A A A A A B B B C C C

137 Appendices

Appendix 9.3 cont/...

Reptile? 0 0 0 0 0 0 1 0 1 0 0 1 0 0 0 1

Bird? 0 1 1 1 1 1 0 0 0 1 0 0 1 0 0 0

Mammal? 1 0 0 0 0 0 0 1 0 0 1 0 0 1 1 0

Endemic? 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0

Diurnal? 1 1 1 1 1 1 1 0 1 0 0 1 0 1 1

Social? 1 0 1 1 0 0 0 0 0 0 0 0 1 1 0

Trophic level 0 1 1 3 3 3 3 2 2 0 1 3 1 2 0 1

Habitat 3 2 4 2 2 4 4 4 4 2 2 4 2 4 4 4

9 0 3 5 3 9 3 2 5 1 9 4 5 9 0

Area of 8 7 6 8 7 5 5 8 5 6 8

6 0 ...... 0 . . . 4 . . . . . 3 0 3 3 3 3 3 2 0 2 3 occupancy 6 9 7 9 1 1 1 1 1 1 1 1 1 1 1

Dispersal 0 1 1 1 6 7 0 0 0 1

4 0 0 3 1 4 6 3 3 9 ......

distance 3 5 8 7 8 7 7 2 2 3

0 9 0 5 9 0 1 5 9 5 0 0 0 0 1

0 3 1 2 2 0 0 5 7 2 2 1 0 0 6

Fecundity ...... 0 1 1 1 1 0 2 1 1 1 2 1 0 0 1

9 9 0 9 0 0 0 5 9 9 0 0 9 0 1

Age at first 2

6 0 6 1 1 1 2 6 3 1 0 6 1 4 ...... 0 0 0 0 1 1 1 1 0 1 1 0 0 1 0

reproduction -

Generation 0 1 4 3 0 0 1 3 5 3 6 1

0 4 2 7 0 0 4 7 4 7 1 4 ......

time 2 1 2 1 2 2 1 1 2 1 3 1

5 8 0 9 4 1 5 4 5 5 2 0 4 0 1 3 9 5 1 2

3 2 2 6 3 4 ...... Body mass ...... 0 3 5 0 2 4 4 5 1 1 0 2 5 ------

Regional threat 3 1 3 1 2 2 0 2 0 0 2 0 1 1 4 0

s u l

a i h n p n

e

a s c

a

u o t m i

t a

i s c a a e a

l r k a t

u s s

Binomial i s e c u i m a e i u r i

l v r n s d a h

c e i a n s e g i l u m n c s i i u r

l r i u l s u n l u n a q b n m e

e h n o r s o o a o o o r s r a r i p p y c i g

c g b r o m t t a o i

i s m a

l l m a

g l r k s o s c n a r a e o s u r e u d u s r b u n h t u i a a a s c m s e a i i o r e d a i i s u i u m m b s n n s s u a i u n s m i c v a u m o o c e c a l l s i t u r g o o h e c c r e e h h i i i o r r y c n p q r C C C C C C C C C C D E E E E E

138 Appendices

Appendix 9.3 cont/...

Reptile? 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0

Bird? 0 0 0 0 1 1 0 0 1 0 1 1 1 1 1 0

Mammal? 0 0 0 0 0 0 1 1 0 1 0 0 0 0 0 1

Endemic? 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Diurnal? 1 1 1 0 1 1 0 0 1 1 1 1 1 1 1 0

Social? 0 0 0 0 0 0 0 0 1 1 0 0 0 0

Trophic level 1 1 3 2 3 3 3 3 1 0 1 3 3 3 3 2

Habitat 4 4 4 4 4 4 4 4 4 4 2 4 4 4 2 4

6 1 4 7 3 6 9 6 1 4 2 3 4 9 9 1

Area of 7 3 4 3 8 9 5 1 7 5 9 2 2 2 3 ...... 3 ...... 3 0 4 2 0 2 2 3 2 1 1 3 2 occupancy 8 7 7 1 1 1 1 1 1 1 1 1 1 1 1 1

9 9 8 2 1 0 0 2 1

Dispersal 0

2 4 3 0 3 3 8 0 ...... 4 8 6 4 8 5 5 7 8

distance -

9 0 0 9 1 0 9 0 8 9 0 0 0 9 9

7 1 0 7 6 1 3 5 0 6 0 0 0 6 3

Fecundity ...... 1 1 3 0 1 1 1 1 2 0 0 0 0 0 1

Age at first 0 0 0 1 0 9 0 9 9 0 0 0 9 9

0 0 1 4 0 6 0 6 7 1 1 1 3 6 ......

reproduction 0 0 1 0 0 0 0 0 1 1 1 1 1 0

Generation 1 5 1 1 0 6 0 0 4 4

4 4 4 4 4 1 0 0 2 2 ......

time 1 2 1 1 2 3 2 2 2 2

6 9 7 9 6 6 4 1 4 9 2 8 9 8 1 9 7 3 6 9 9 3 7 6 3 6 9 1 7 0 6 1 ...... Body mass ...... 4 6 0 2 1 0 0 4 0 1 0 3 1 2 1 1 ------

Regional threat 3 3 0 1 1 2 1 2 3 1 4 2 2 2 3 0

s

s s u s a u c u l

i n e a h i

a n l t s

s p

e a m s o a e o s s l y l o i l r c

Binomial u m r d u l i s i t p

i c u n n o e c t n y r o a i a n a c a c t a

t t e h u n b r r b n s u o i u

t i f p y l r e

r r e s i a u r g a a l a s s r g g c a u l g b i s i i m b u c s s o g r n e a

u r u u n s i l n r c u u s s a i a a t t

a e m h a u s s l t u h i e e h a t c n p e a a o a l p h e e c m l i i e e t l c e i o o a a a s l e s s s n i i x m m c c i i z l l p p m b l l u l l e e y a a r y y a a e r r r u a a e e r E E E E F F F F F G G G G H H H

139 Appendices

Appendix 9.3 cont/...

Reptile? 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0

Bird? 1 0 0 1 1 1 1 0 0 1 0 0 0 0 0 1

Mammal? 0 1 1 0 0 0 0 1 1 0 1 1 1 1 0 0

Endemic? 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0

Diurnal? 1 0 0 1 1 1 1 0 0 1 1 0 0 0 1 1

Social? 0 0 1 0 1 0 1 0 0 1 0 1

Trophic level 3 3 0 2 3 3 3 3 3 1 1 3 3 2 3 1

Habitat 3 4 4 2 2 1 2 2 3 2 4 4 4 4 4 4

2 5 2 3 5 6 0 9 8 3 3 2 5 9 8 2

Area of 7 0 7 6 9 1 8 0 6 . 5 . 2 1 . 2 9 . 9 7 ...... 2 3 3 0 2 3 0 3 2 occupancy 9 9 6 1 8 9 7 1 1 1 1 1 1 1 1 1

3 0 0 0 2 7 1 3 1 1 9 1 1

Dispersal 5

6 1 3 9 4 1 9 3 0 8 6 3 ...... 2 7 1 2 8 7 3 2 7 5 2 1 7

distance -

0 9 2 0 0 2 2 0 0 9 5 1 3 0 4 9 0 6 9 1 1 9 9 1 1 7 2 4 6 0 4 6

Fecundity ...... 0 0 0 1 1 0 0 1 1 1 1 0 2 0 2 0

2 0 0 0 0 0 0 9 0 9 0 0 9 9 9

Age at first 2

1 1 1 0 0 1 6 1 5 0 1 6 3 3 ...... 0 1 1 1 0 0 1 0 1 0 0 1 0 1 1

reproduction -

Generation 0 0 3 1 5 3 4 0 3 4 4

0 0 7 4 4 7 2 0 7 2 2 ......

time 2 2 1 1 2 1 2 2 1 2 2

6 5 7 2 5 8 7 9 8 8 0 5 2 5 3 9 3 0 7 9 1 6 4 2 5 8 2 7 0 1 1 7 ...... Body mass ...... 0 1 1 0 0 0 4 0 3 2 0 1 3 1 2 1 ------

Regional threat 1 4 1 2 2 1 2 3 3 3 3 3 1 1 1 1

s i r t s

i o i r s

i t r s

s e

s u i l i

u s h u n r s t t a

g u i

Binomial e n t s u a h n s i r a a

u n p a n t b t

n a e

n s z e m

e i

a n c e s s a n p t a g i

a e t r c p u e a r y i n

t a

h e e a c a h s d n c m c v a n y

n t i a u r n u e n l x a m a h o h t t

i a i

r d r y e e r t n c e u

a x r a b i r o x h l y a o l

a s i l o a r n v i e s s s

t r i a t a m m t e o t l x u u u r a r r r s l o s a s j d r r r t e n i i a a e u y i y y t a a a a u y b H H H I L L L L L M M M M M N O

140 Appendices

Appendix 9.3 cont/...

Reptile? 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 0

Bird? 0 1 0 1 1 1 1 1 0 0 0 0 0 0 0 1

Mammal? 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0

Endemic? 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0

Diurnal? 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1

Social? 1 0 0 1 1 1 1 0 0 0 0 0 0 0 0

Trophic level 0 1 3 1 3 3 3 1 2 2 2 2 2 2 2 2

Habitat 4 2 3 4 2 2 2 2 4 4 4 4 4 4 4 3

6 2 2 2 3 2 5 5 8 2 1 5 1 0 5 0

Area of 8 5 1 1 2 2 3 0 . 3 . . . . . 6 7 . 5 2 3 6 7 ...... 1 1 1 1 1 1 3 occupancy 9 9 7 2 9 6 8 7 5 1 1 1 1 1 1 1

Dispersal 6 0 1 1 1 1 1 1 1

5 6 0 6 9 3 9 3 4 ......

distance 3 7 5 4 6 7 6 7 0

6 1 9 1 6 9 1 9 9 9 9 0 0 9 9 9 2 6 6 6 6 6 6 6 3 6 3 1 1 6 3 3

Fecundity ...... 0 1 0 1 0 0 1 0 1 0 1 1 1 0 1 1

2 6 1 0 0 0 0 0 9 0 0 1 9 0 0 0

Age at first 2 1 8 0 1 0 1 1 6 1 0 4 6 0 0 0 ...... 0 0 0 0 1 0 1 1 0 1 0 0 0 0 0 0

reproduction - -

Generation 8 4 2 1 0 8 8 0 0 2 1

5 2 9 4 0 5 5 0 0 2 4 ......

time 1 2 2 1 1 1 1 1 1 1 1

3 6 2 0 2 7 9 8 4 2 8 8 0 7 2 3 9 2 3 1 4 0 3 7 8

4 5 3 0 1 ...... Body mass . . . . . 0 3 0 5 5 5 6 6 5 1 3 3 2 2 1 ------

Regional

threat 3 3 4 1 2 2 1 1 3 4 0 3 3 3 5

s s s s s i

u u

u u u s t i t w p e n u a h a o c l o e l a c s s k c i a

u a u i t u u u s c d a s l c i s i l e a m l s g t y a a s r a e t g e t o o o s h

Binomial y h m m r r s s o r

s u s p r

p s s s s s s s

e c s u p u u u u u u u u x t x s c o u d l l l l l l l i e a a r r o l n a a a a a a a r r e

c a p c o m h h h h h h h t i o

u p a p p p p p p p e p c s s m e e e e e e e e i u n l o a u u o

s c c c c c c c q r r c g

n n i i a s o o o o o o o e c r

a a r v n n n n n n n n h e a s c c t l e u s y y y y y y y s e e u i y n s a o r r r r r r r l l c v x a a e e h h h h h h h h h i O O P P P P P P P P P P P P P P

141 Appendices

Appendix 9.3 cont/...

Reptile? 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1

Bird? 1 1 1 1 1 0 0 0 0 0 0 0 0 1 0 0

Mammal? 0 0 0 0 0 1 1 1 1 1 1 1 1 0 1 0

Endemic? 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1

Diurnal? 1 1 1 1 1 0 1 0 0 0 0 1 1 0 0

Social? 1 1 1 1 1 1 0 0 0 0 0 0 1 1 0

Trophic level 3 1 1 0 0 2 0 1 1 1 1 0 1 0 2 2

Habitat 2 2 4 2 2 4 4 4 4 4 3 4 4 4 4 4

0 3 9 2 7 6 3 1 0 2 0 5 5 4

Area of 1 1 5 1 1 0 1 7 1 2

. . 7 ...... 5 8 0 . . . . 3 0 3 4 0 3 0 0 3 3 occupancy 8 9 8 9 1 1 1 1 1 1 1 1 1 1

0 0 1 7 7 5 5 0

Dispersal 4

6 0 4 4 4 5 7 ...... 0 7 8 7 7 3 6 4

distance -

9 9 9 0 0 0 8 2 2 9 0 5 0 0 8

3 3 3 1 1 0 5 9 8 7 1 2 1 0 4

Fecundity ...... 1 1 1 1 1 0 0 0 1 1 1 1 1 0 1

9 0 9 1 0 0 9 0 0 0 0 0 0 9 9 6

Age at first 2 0 6 4 0 0 6 0 0 0 0 0 0 6 6 9 ...... 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

reproduction -

Generation 2 3 3 0 0 4 0

2 7 7 0 0 1 0 ......

time 1 1 1 2 1 1 2

1 5 4 1 4 1 4 2 4 9 3 3 2 0 0 5 1 3 9 8 6 8 1 9 3 8 5 9

0 3 ...... Body mass . . 0 0 1 0 4 4 3 5 0 0 0 3 3 0 3 ------

Regional threat 3 3 3 0 0 2 4 2 2 3 2 1 0 3 2 2

s u l y

t s c

o s a r i d s

e o s n

d

t i i

u e s

s r p s l s c a s u u s i i e e o a u i n l l

n d

Binomial l

d n i

p d s i a s l i s t a a s t l i c o t k r t p n e i l u t p s r i a

e n a e e s e n e a o i o h i h t s e i a d p b p

h p r i c t i c c c n u n s n o t l l r i a e u s s l c e a u a r i a a o t h s e b u u

f n g h l p a t t h c e i i

s s t t a p a s o o p c u s g e e i a i a p d l l o s e g g o t b i e l n d x

c c a l o c i r n n o t a x m a i i a o o h v a o a l r e t n a g r r r g p p e d i d i l l r a e r l r e e e a l l o t t h a e a a a e o p p y P P P P P R S S S S S S S S T T

142 Appendices

Appendix 9.3 cont/...

Reptile? 0 0 0 0 0 0 1 1 0

Bird? 1 1 1 1 0 0 0 0 0

Mammal? 0 0 0 0 1 1 0 0 1

Endemic? 0 0 0 0 0 0 0 0 0

Diurnal? 1 1 1 1 0 0 1 1 0

Social? 1 1 0 0 0 0 0

Trophic level 2 0 0 1 3 1 3 2 3

Habitat 3 4 4 4 4 3 4 4 4

7 4 2 0 6 0 6 1 0

Area of 1 7 4 4 5 6 9 . 2 7 ...... 1 3 1 1 3 4 3 occupancy 6 8 1 1 1 1 1 1 1

Dispersal 0 0 1 0 1 8 1

6 6 9 4 8 0 4 ......

distance 7 7 6 3 2 2 0

9 9 9 9 9 7 4 9 3 3 7 7 7 6 4 0 7 5

Fecundity ...... 1 1 1 1 0 0 3 1 1

Age at first 0 0 0 0 9 0 0

0 0 0 1 3 1 0 ......

reproduction 0 0 0 1 1 1 0

Generation 1 1 4 0 1 5

4 4 2 0 6 4 ......

time 1 1 2 2 3 2

6 9 9 6 5 2 8 5 0 8 3 2 8 9 6 5 5 6 . . . . Body mass . . . . . 3 0 3 0 0 0 3 4 0 - - - -

Regional threat 1 3 3 2 3 1 0 1 0

s

i u s s

i n u a i d a t n p a

s e s r Binomial s u b a a i u

g c t i p e

i e

s x s s s r e i n a i a o u u e i r n t l l r s c p l l g t o c r

n a a e r h s u a t

u l g g a

p u

i a e x o o s n a s r i a a a u a m p e r r r c s r r r t t t p n r a i o e e e e T T T T U U V V V

143 Appendices

Appendix 9.4 Species data presented in chapter 6 Data presented in analysis of bioregional extinction in Australian mammals (chapter 6). See text for trait definitions. Continuous traits are ln transformed, except minimum latitude. NPB = not present in bioregion.

Rock user? 0 0 0 0 1 0 1 0 1 1 1 1 0 0 1 0

Diet breadth 3 3 5 2 3 1 2 1 1 1 2 2 2 2

Habitat breadth 3 1 3 4 7 4 7 2 5 4 4

Trophic Level 2 1 2 2 2 3 2 3 3 3 3 3 3 3

7 1 0 2 2 2 7 2 7 0 0 0 9

0 0 1 8 2 1 9 2 1 3 3 2 8

Mean PET ...... 7 7 7 6 7 7 6 7 7 7 7 7 6

2 2 4 2 8 0 2 0 1 9 9 4 3

4 4 4 7 2 0 0 5 8 4 4 6 9

Mean AET ...... 6 6 6 5 6 6 6 6 5 5 5 5 5

Mean 5 4 9 5 9 8 6 6 4 3 3 9 7

1 0 1 9 3 2 0 3 3 4 4 3 9 ...... temperature 5 5 5 4 5 5 5 5 5 5 5 5 4

4 4 0 1 9 3 2 8 8 2 2 2 8

0 0 2 4 0 5 6 1 2 0 0 0 7

Mean ppt ...... 4 4 4 3 4 3 3 4 3 3 3 3 3

2 3 7 4 3 0 4 4 6 6 6 8 1 7 1 5 9 1 8 1 6 6 2 9 3 0 2 1

Minimum ......

9 8 2 5 3 7 3 3 9 4 9 6 5 3 latitude 3 3 2 2 3 2 4 4 2 3 2 2 2 3 ------

Global 4 7 1 2 6 1 3 8 5 0 0 3 4 0 3 7 7 3 8 8 6 1 8 6 6 8 9

...... 8 geographic . 4 3 0 3 4 5 4 4 3 1 1 2 1 5 range 1 1 1 1 1 1 1 1 1 1 1 1 1

2 4 1 6 2 1 0 6 5

0 . . . . Home range . 5 0 1 2 1 - - - -

9 6 5 4 9 5 5 7 2

Age at sexual 3 2 5 2 0 3 3 2 0

...... 0 0 0 0 0 0 0 0 0

maturity ------

4 6 9 9 9 3 9 0 0 7 0 0 9 9 9 4 3 2 6 6 6 7 6 1 1 8 1 1 7 7 9 9

Litter size ...... 1 2 0 0 0 1 0 1 1 0 1 1 1 1 1 1

Inter birth 1 0 8 4 0 6 6 0 0

2 9 7 6 8 1 1 9 9 ...... interval 5 5 4 4 4 5 5 5 5

Gestation 3 6 7 1 6 3 8 8 1 7

3 0 9 8 9 5 4 4 9 8 ...... length 3 3 2 3 3 4 3 3 3 2

8 1 7 6 5 1 7 1 1 9 6 2 1 6 9 7 2 1 0 2 6 6 2 7 7 1 1 4 4 2 7 0 ...... Body mass . . . . 4 3 0 4 3 1 4 4 4 2 2 3 0 0 1 0 ------

t t t t t t 3 3 8 8 n n n n n n Western B B B B B B 6 6 a 5 a a a a a 5 P P P P P P t t t t t t 9 9 9 9 x x x x x x N N N N N N deserts 1 1 1 1 e e e e e e

t t 5 5 5 . . . 1 9 n n B B B B B B B B B 7 8 7 0 6 a a P P P P P P P P P t t 5 5 3

Pit & Finke 9 9 x x 9 9 9 N N N N N N N N N 1 1 e e 1 1 1

t t t t t 6 0 5 5 8 0 0 n n n n n B B B B 4 9 7 9 1 6 8 a a a a a P P P P t t t t t

Flinders 9 8 8 8 9 9 8 x x x x x N N N N 1 1 1 1 1 1 1 e e e e e

3 5 5

t t t t t t 3 2 2 5 n n n n n n . . . B B B B B B 3 a a a a a a 3 8 9 P P P P P P t t t t t t

BMTS 9 6 2 2 x x x x x x N N N N N N 1 9 9 9 e e e e e e 1 1 1

s s s s s

s

e u u u u u n

s s s u s a a

n s b b b b a s o

u t a u u e s i s s s t u d t u o o o o d h s

a a s c c m r u e l l l l u u i

u i i e i n t l n a p u t p r r u s t i

y t

i s l o o o o e r

Binomial n g g s i r i r

e e e o o s a a a l r r a o i h e i s n n n n a i e g n n n r r u c c c k u r u i i i i b p a i d o c e d i m p c n t f p e e l l l o l e i y y y y o o l i c t m i c f o o e u t t n e r a r s s s s a l a a a a a a t u r s t t r v u g n m n r o s a i b g c o a a a a s a e h h h h h h c n e e a y e o i i y e a o c r o l e o A p A f B l B p C c C c C j C e C g C m C n C p D b D c D r D g

144 Appendices

Appendix 9.4 cont/…

Rock user? 1 0 0 1 0 0 1 0 0 0 0 0 0 0

Diet breadth 4 4 1 1 1 1 1 3 2 2 3 1 3 2

Habitat

breadth 3 5 2 2 1 2 3

Trophic

Level 2 2 3 3 3 3 2 2 2 2 1 2 1

0 0 2 1 9 3 4 8 0 0 0 3 2

2 1 1 2 7 7 9 0 1 2 6 3 8

Mean PET ...... 7 6 7 7 6 6 6 7 7 7 6 7 6

5 6 4 9 9 4 8 6 0 9 8 2 1

4 8 5 0 9 2 1 6 4 3 9 3 5

Mean AET ...... 6 5 5 6 5 6 6 6 6 6 5 6 5

Mean 1 6 9 6 8 6 5 3 3 9 2 5 3

3 2 2 3 8 7 0 3 2 3 7 4 0 ......

temperature 5 4 5 5 4 4 5 5 5 5 4 5 5

5 6 6 3 3 2 9 2 5 3 3 1 9

1 8 9 6 7 0 7 8 2 1 9 8 8

Mean ppt ...... 4 3 2 3 3 4 3 4 4 4 3 3 2

2 1 0 5 4 4 4 6 4 8 4 5 1 7 3 6 1 5 6 6 6 1 6 5 6 8 0 2

Minimum ......

5 1 5 6 3 3 3 5 3 7 3 4 5 2

latitude 2 4 3 3 4 4 4 2 4 1 4 2 2 3 ------

Global 7 0 4 0 9 1 9 6 7 3 8 7 6 0 2 1 4 4 8 1 1 5 1 0 7 4 3

...... 5

geographic . 3 1 4 5 3 3 4 5 5 1 2 4 1 4 range 1 1 1 1 1 1 1 1 1 1 1 1 1

1 8 2 1 5 6 3 9 0 1 6 4

3 5 . . . . . Home range . . 3 1 2 4 0 1 0 - - - - -

4 4 6 0 0 8 4 1

Age at sexual 1 1 1 1 9 0 1

4 ...... 0 0 0 1 0 0 0 0

maturity ------

0 3 2 9 9 9 1 8 0 9 9 9 9 9

0 9 9 6 6 6 5 2 2 6 6 6 6 6

Litter size ...... 2 1 0 0 0 0 1 1 1 0 0 0 0 0

Inter birth 0 0 1 4 3 3 0

9 9 1 7 0 8 9 ......

interval 5 5 4 4 5 4 5

Gestation 2 3 7 0 8 1 0 3

0 5 2 6 4 7 4 0 ......

length 3 4 5 3 2 2 3 3

4 9 1 9 1 0 0 9 2 8 9 1 0 1 0 7 6 2 1 5 7 8 8 1 1 4 6 4 1 5 1 2 ...... Body mass ...... 0 5 5 5 4 0 0 0 0 1 1 0 1 0 1 3 ------

t t t

3 3 8 3 n n n Western B B B B B B B B B a a a 7 6 6 6 P P P P P P P P P t t t 9 9 9 9 x x x N N N N N N N N N deserts 1 1 1 1 e e e

t t t 5 0 0 n n n B B B B B B B B B B a a a 5 7 5 P P P P P P P P P P t t t

Pit & Finke 9 9 9 x x x N N N N N N N N N N 1 1 1 e e e

t t t t t t t 9 0 0 0 n n n n n n n B B B B B 0 0 8 0 a a a a a a a P P P P P t t t t t t t

Flinders 9 9 8 9 x x x x x x x N N N N N 1 1 1 1 e e e e e e e

3

t t t t t 3 9 5 7 5 n n n n n . B B B B B B 4 4 9 2 a a a a a 7 P P P P P P t t t t t

BMTS 9 9 9 9 6 x x x x x N N N N N N 1 1 1 1 9 e e e e e 1

i s

s k u s s t

c n s u u u t a

o o

u s r l t n s s r s a

i n e f s u r u u s c t r l r u i i t r l l t i t r u u a u h e i u t l

t u a e s l v m l l s g s s s s s g a v

r u a a u i u r e e e e e e o a t s e t t t t

v v r h b p r s t

Binomial u b a u s s s s

s e s l a o s s s s s s s e s e e e n l s y

d a i i u u u u u u u i h u h h h e n n g t c h r r m c c c c c s c c c c d i o o i i i i i r o a i r r r r u u o o s s s s s p s d d r o r y y o o o o p s i m e e e e e y s s o o o d t t t t t g g g g s p r n o a a i y o o p p p p p p a a a a a s h o s s e D D E E E E E H H c I I L a L c L L l L

145 Appendices

Appendix 9.4 cont/…

Rock user? 1 0 1 1 0 0 0 0 1 0 0 0 1 0 0 0

Diet breadth 3 1 1 2 1 2 1 1 5 2 1 1 1

Habitat

breadth 5 3 3 4 2 1 1 2 3 6 3 2

Trophic

Level 2 1 1 3 1 1 1 1 2 2 3 3 2 3

4 2 0 1 9 2 9 0 8 6 1 8 1 6

2 3 2 9 7 9 0 4 1 1 2 2 1 0

Mean PET ...... 7 7 6 6 6 6 7 7 7 7 7 7 7 7

0 8 1 0 1 3 1 0 5 4 4 3 3 3

7 3 5 7 8 7 3 7 0 8 8 5 4 9

Mean AET ...... 5 6 4 6 5 5 6 6 6 5 5 6 6 5

Mean 8 6 4 8 2 0 9 2 3 4 5 3 2 1

3 4 8 1 9 1 1 5 3 3 4 4 2 2 ......

temperature 5 5 4 5 4 5 5 5 5 5 5 5 5 5

2 1 3 3 1 9 8 2 5 5 8 6 4 5

1 0 8 9 6 2 8 5 5 1 2 2 5 2

Mean ppt ...... 3 4 2 4 3 3 3 4 3 3 3 4 4 3

9 3 3 1 5 3 8 6 8 8 6 8 3 5 8 8 5 2 2 4 2 6 7 5 1 0 1 8

Minimum ......

0 3 0 1 2 8 2 4 6 6 2 7 9 8

latitude 3 2 3 2 3 3 4 1 3 2 2 1 3 3 ------

Global 1 6 3 5 9 1 1 2 8 0 7 2 7 6 4 4 9 4 3 6 5 7 4 1 5 7 1

...... 7

geographic . 4 3 4 1 4 4 0 5 5 4 1 7 5 3 range 1 1 1 1 1 1 1 1 1 1 1 1 1

6 7 0 7 7 2 3 0 0 1 8 4 5 7 8 9

9 6 9 4 ...... Home range . . . . 3 2 0 1 0 0 0 0 1 1 ------

Age at 2 2 4 5 7 0 8 9 1 6 0 8 9 6 0 5 5 5 0 8 1

0 4 5 5 5 4 2 4 1 ......

sexual ...... 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 maturity ------

9 7 9 9 9 9 9 9 9 0 0 9 9

3 3 6 6 6 6 6 6 6 1 1 6 6

Litter size ...... 1 1 0 0 0 0 0 0 0 1 1 0 0

Inter birth 0 2 9 6 7 6

9 9 8 5 4 1 ......

interval 5 5 5 5 5 5

Gestation 0 7 3 7 3 0 4 0 8 1 7

5 7 8 3 4 6 5 5 6 2 6 ......

length 3 3 4 3 3 3 3 3 2 5 4

1 2 8 0 4 0 7 9 1 6 4 1 6 1 1 7 9 0 0 9 0 2 2 7 7 6 2 5 1 5 4 1 ...... Body mass ...... 3 4 2 1 1 4 4 0 2 1 3 3 3 3 0 2 ------

t t t t t t t t 8 1 8 3 n n n n n n n n Western B B B B a a 4 6 a a a a 6 a a 6 P P P P t t t t t t t t 9 9 9 9 x x x x x x x x N N N N deserts 1 1 1 1 e e e e e e e e

t t t t t 5 . 9 0 0 n n n n n B B B B B B B 9 2 7 6 a a a a a P P P P P P P t t t t t 0

Pit & Finke 9 9 9 x x x x x 9 N N N N N N N 1 1 1 e e e e e 1

t t t t t 0 0 0 0 0 5 n n n n n B B B B B a 9 0 2 a a a 2 1 a 8 P P P P P t t t t t

Flinders 8 9 9 9 9 8 x x x x x N N N N N 1 1 1 1 1 1 e e e e e

5

t t t t t t t t 7 5 5 7 n n n n n n n n . B B B B 2 4 6 a a a a a a a a 9 P P P P t t t t t t t t

BMTS 9 9 9 8 x x x x x x x x N N N N 1 1 1 9 e e e e e e e e 1

i

i s

r u s

a

s s u

s i s c i t o i e u a i c l t t

s n a

g s e n s s i e n a r i s i i

e u r b c g t l a u g u s i i i r g

b f i o g c l s s s

p i o g s u a o u g u u u u u f s

a e f r r a g i n i r r e a

Binomial m u l l e a a s b e e s s s s s s s r r

t t s r n n n o u s s e u u u u u u e s i i l i i e p p p t c w l t t d p p p p p p u d d e b o o i p e t i o o o o o o o o o o c a a r o a i r r r r r r r r r e m m d m i i g g o c c c c c c c c c r r r n r e c n g g p a a a a a a a a a i h o o y k a s e e e l c a a L L l L M M M M M M M M M M s M M p M f

146 Appendices

Appendix 9.4 cont/…

Rock user? 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0

Diet

breadth 1 3 0 3 3 2 1 1 1 1 1 1 1 1 1

Habitat

breadth 4 4 3 2 2 3 3 3 4 6 3 5 5 1

Trophic

Level 3 2 2 2 1 1 3 3 3 3 3 3 3 3 1

3 9 5 2 9 9 5 0 3 3 4 6

9 1 1 9 1 1 8 2 1 1 9 2

Mean PET ...... 6 7 7 6 7 7 6 7 7 7 6 7

2 8 4 8 0 9 3 3 9 0 2 5

1 6 6 2 2 6 0 7 9 0 5 1

Mean AET ...... 6 5 5 6 5 5 5 6 6 5 6 6

Mean 7 7 5 1 0 5 0 0 8 9 0 0

9 3 3 2 3 3 0 5 3 2 1 4 ......

temperature 4 5 5 5 5 5 5 5 5 5 5 5

7 6 2 7 6 4 7 4 4 1 0 0

0 0 0 2 7 7 2 4 8 5 6 2

Mean ppt ...... 4 3 3 4 2 2 3 3 4 3 3 4

4 9 5 9 0 1 4 2 6 4 3 2 6 2 5 6 2 1 7 1 9 6 0 7

Minimum ......

4 1 5 2 1 8 6 3 0 3 3 4

latitude 3 3 2 1 3 2 3 2 2 4 4 1 ------

Global 0 1 3 8 4 3 2 1 0 5 4 3 7 9 9 7 2 3 2 2 8 5 8

...... 9

geographic . 4 4 0 2 2 3 1 5 5 4 2 8 range 1 1 1 1 1 1 1 1 1 1 1

3 8 3 4 9 7

Home range . . . 1 4 2 - - -

Age at 9 7 9 9 3 4 7 1 2 6 6 8 6 6

sexual ...... 0 0 1 0 0 1 1 maturity ------

1 4 1 9 5 0 3 9 9 2 0

6 9 6 3 2 4 5 6 6 9 1

Litter size ...... 1 1 1 1 1 1 1 0 0 0 1

Inter birth 0 0 9 3 0

9 9 4 7 6 . . . . .

interval 5 5 3 3 3

Gestation 4 0 8 3 6 9

6 6 4 7 5 5 ......

length 2 2 3 3 3 3

1 4 0 5 2 0 5 0 0 1 3 2 6 6 1 5 6 4 3 3 1 3 1 9 9 7 8 1 9 9 5 2 ...... Body mass . 4 3 2 3 2 2 3 2 2 4 4 5 4 4 4 1 ------

t t t t t t t t t

0 5 n n n n n n n n n Western B B B B B a a a a a a a a a 5 4 P P P P P t t t t t t t t t 9 9 x x x x x x x x x N N N N N deserts 1 1 e e e e e e e e e

t t t t t t 5 5 n n n n n n B B B B B B B B 3 4 a a a a a a P P P P P P P P t t t t t t

Pit & Finke 9 9 x x x x x x N N N N N N N N 1 1 e e e e e e

t t t t t t 5 5 5 0 n n n n n n B B B B B B 7 7 7 9 a a a a a a P P P P P P t t t t t t

Flinders 8 8 8 8 x x x x x x N N N N N N 1 1 1 1 e e e e e e

t t t t t t t t t 0 9 0 5 n n n n n n n n n B B B 2 3 3 5 a a a a a a a a a P P P t t t t t t t t t

BMTS 9 9 9 9 x x x x x x x x x N N N 1 1 1 1 e e e e e e e e e

i

y s i a s

o i t i u n

p r i n a r f s l n

o i f l i o n l o s u i t

r e o s u h b u y n s s

l e u l i h i l u

p e n s i

v c a g x p c a r y a a u t s e r c t e s t i

b g s s i s s e

l m e

d l s u Binomial a s s s u u u i s s s i c f a a m l l l a d e e n r s i i i u u u t t s s s s s s g e u i i i i n h h h c c y y y y y y o e e e a u e y y m p p p i c c c h c a m m m m m m r r e i i i o o o i r c t t t t t t g o o o o o o o o h g o y t t t t t t t t c c c c c c n n n n i o o o o o o o o y y y y y y m r o i N N N N N N l N N N N N N N a N N t O

147 Appendices

Appendix 9.4 cont/…

Rock user? 0 0 0 0 1 1 0 0 0 0 0 1 0 1 1 0

Diet breadth 2 3 1 3 1 3 2 2 2 2 1 1 2 2

Habitat

breadth 3 2 2 3 1 3 2 4 3 1

Trophic

Level 1 2 3 2 1 1 3 1 3 3 3 3 3 1

2 1 2 6 6 8 1 1 2 0 7

3 2 0 3 0 0 1 3 2 0 1

Mean PET ...... 7 7 7 7 7 7 7 7 7 7 7

8 8 6 2 4 3 1 3 3 7 2

4 4 5 5 4 4 6 4 5 6 4

Mean AET ...... 6 6 5 6 6 6 5 6 6 5 6

Mean 7 7 0 0 5 2 9 2 2 0 6

4 3 2 5 1 1 2 4 3 4 0 ......

temperature 5 5 5 5 5 5 5 5 5 5 5

9 9 3 6 6 8 0 7 7 7 0

0 2 0 1 1 9 1 9 1 0 1

Mean ppt ...... 4 4 3 4 4 3 3 3 4 3 4

1 5 0 2 7 9 3 4 4 1 3 4 0 3 3 6 7 3 1 6 0 1 9 6

Minimum ......

5 5 7 3 5 8 9 4 6 3 9 3

latitude 2 1 1 3 1 3 3 3 2 3 2 4 ------

Global 2 2 7 7 8 3 9 9 5 9 8 0 9 0 0 6 8 8 7 8 0 7 1

...... 5

geographic . 3 2 2 2 3 3 3 3 4 2 4 4 range 1 1 1 1 1 1 1 1 1 1 1

6 6 1 8 6 6 2 6 8 6

Home range . . . . . 2 3 4 4 4 - - - - -

5 3 6 9 3 4 2 5 0 7 0

Age at sexual 1 1 4 1 3 2 1

1 3 6 0 ...... 1 0 0 0 0 0 0 0 0 0 0

maturity ------

9 0 9 0 9 9 8 7 9 8 6 4 0 4 8

6 1 6 1 6 6 0 0 6 0 9 1 2 9 0

Litter size ...... 0 1 0 1 0 0 2 2 0 2 1 2 2 1 1

Inter birth 8 8 0 2 3 1 3

5 2 9 9 8 5 5 ......

interval 5 5 5 5 5 4 5

Gestation 8 5 8 4 3 6 8

4 4 4 3 5 7 9 ......

length 2 3 3 3 3 2 2

7 3 4 4 3 1 2 2 2 9 3 1 1 1 4 8 4 4 0 9 6 6 1 4 1 6 1 1 6 3 1 8 ...... Body mass . . . . 1 1 2 2 1 4 5 4 5 3 3 0 1 1 2 1 ------

t t t t t t

3 n n n n n n Western B B B B B B B B B a a a a a a 7 P P P P P P P P P t t t t t t 9 x x x x x x N N N N N N N N N deserts 1 e e e e e e

t t t 0 n n n B B B B B B B B B B B B 4 a a a P P P P P P P P P P P P t t t

Pit & Finke 9 x x x N N N N N N N N N N N N 1 e e e

t t t t t t t 0 0 n n n n n n n B B B B B B B 0 a a 0 a a a a a P P P P P P P t t t t t t t

Flinders 9 9 x x x x x x x N N N N N N N 1 1 e e e e e e e

5

t t t t t t 7 9 n n n n n n . B B B B B B B B 2 a a a a a a 2 P P P P P P P P t t t t t t

BMTS 9 6 x x x x x x N N N N N N N N 1 9 e e e e e e 1

s

a i f s

a r

a t u a s

n t s t a i s p i p a a

r l a o i o s s e

m l s o u r a i i c h l i u u a u m

r p i t s s s r c n n a u s e e v e a n i n i t u r c t l g a n a o e e i a h h r e t e a n e e l r i l e e l x c c i t l c g m l

Binomial l l l s e b s s i e e r e a e e e a a e e e e t t u e e a h v s l l a n l l l l g l l a r s g g l n n c n n u a a e e n y i a a a a o e i u o o o a a o r f r g g e o g g g g h e i a c c c l m m d d d i i i i u g o o c r d l s s s u g a a u u u n n n n a r r e y c e o a a a t t t r r g u e e e r a a a a n a n o e e e e e h h h l l l l s s s n o e i O u P b P P P P P P P c P P P P P m P w P p

148 Appendices

Appendix 9.4 cont/…

Rock user? 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0

Diet breadth 3 4 2 2 4 0 2 1 2 4 4 3 1 1 2

Habitat

breadth 2 2 4 1 2 2 3 2 3 4 3 5 8 3

Trophic

Level 2 1 2 1 2 2 1 1 1 2 2 2 3 3 3

9 7 7 6 7 8 4 2 7 2 3 3 0 8

2 9 2 1 1 3 3 3 1 7 2 3 3 5

Mean PET ...... 7 6 7 7 7 7 7 7 7 6 7 7 7 6

4 9 7 9 4 4 6 1 1 5 7 1 6 2

4 4 4 6 6 0 4 4 4 2 5 1 3 8

Mean AET ...... 6 5 6 5 5 6 6 6 6 6 6 6 6 5

Mean 4 1 9 7 6 1 6 6 1 3 2 6 5 3

4 2 3 3 3 5 4 4 3 7 3 4 4 2 ......

temperature 5 5 5 5 5 5 5 5 5 4 5 5 5 5

1 3 9 9 3 5 3 7 2 9 1 5 6 5

0 8 0 9 0 4 9 9 1 0 2 5 9 1

Mean ppt ...... 4 2 4 2 3 3 3 3 4 4 4 3 3 4

4 8 7 9 8 3 5 6 8 9 4 9 6 7 7 6 0 7 1 9 0 9 9 2 1 6 3 6 7 8

Minimum ......

7 4 7 4 4 1 0 8 1 1 3 8 8 3 4

latitude 1 3 1 2 2 3 2 1 3 3 4 2 2 2 1 ------

Global 3 7 5 9 5 5 0 2 3 4 3 0 3 6 8 0 3 3 0 0 8 8 0 7 4 4 7

...... 8 5 6

geographic . . . 2 3 4 4 5 1 3 4 2 3 4 3 3 9 9 range 1 1 1 1 1 1 1 1 1 1 1 1

0 1 6 9

Home range . . 5 3 - -

Age at 0 9 6 6 9 2 9 7 2 5 0 4 1 9 7 2 6 6 6 3 5 6 9 4 9 0 5 1

2 0 ......

sexual . . 0 0 1 1 1 1 0 1 1 1 1 0 1 0 0 maturity ------

1 9 3 8 9 0 9 8 9 3 5 9 1

6 3 4 3 3 1 3 3 6 6 8 6 6

Litter size ...... 1 1 1 1 1 1 1 1 0 1 1 0 1

Inter birth 0 9

9 3 . .

interval 5 3

Gestation 4 0 7 5 8 1 1 0

8 4 3 4 1 2 1 0 ......

length 3 3 3 3 3 3 3 5

3 7 2 9 0 2 2 1 5 1 2 4 1 1 2 0 7 1 4 6 1 4 4 5 3 5 0 2 0 7 2 2 Body mass ...... 2 4 4 3 3 4 4 4 3 0 1 2 2 4 3 4 ------

t t t t t t t t t t n n n n n n n n n n Western B B B B B B a a a a a a a a a a P P P P P P t t t t t t t t t t x x x x x x x x x x

deserts N N N N N N e e e e e e e e e e

t t t t t t n n n n n n B B B B B B B B B B a a a a a a P P P P P P P P P P Pit & Finke t t t t t t x x x x x x N N N N N N N N N N e e e e e e

t t t t t t 5 5 n n n n n n B B B B B B B B 7 a 7 a a a a a P P P P P P P P t t t t t t

Flinders 8 8 x x x x x x N N N N N N N N 1 1 e e e e e e

t t t t t t t t t t t 5 n n n n n n n n n n n B B B B 0 a a a a a a a a a a a P P P P t t t t t t t t t t t

BMTS 9 x x x x x x x x x x x N N N N 1 e e e e e e e e e e e

i x s

s r i

s e s n l i u o f i

u t t i o a

s s m r i s r a r

m n u l t e i a d n o s e o s l l n u s s

t h b u

g e u o e a s p l a c

i o a i s r t a b d f j l n i a o s e s

u s

l r a c e u i s s s s s s s s s s o

Binomial s b e a s r i d l s y y y y y y y y y t t s l u m r u p i l s s c i t n u m m m m m m m m m a i a l v o u u u i a n n o o o t o o o o o o t n l c h p p e s s i a t d d d a d d d d d d n o o v o o s u u c i n c u u u u u u u u u a t t n m r r s i i t t i v r c e e e e e e e e e l r e e a s s s s s s s s s t t a a h a e e u a m r l P P P d P P P h P P P P P R R R a S f S c

149 Appendices

Appendix 9.4 cont/…

Rock user? 0 0 1 0 0 0 1 0 1 1 1 1

Diet breadth 2 1 1 1 1 2 1 1 1 3 2

Habitat 0 3 3 4 4 3 2 6 4 1 3 breadth 1

Trophic Level 3 3 3 3 3 3 3 2 3 3 1 1

9 6 0 6 1 4 1 8 3 6 2 5 0 1 1 1 2 1 2 0 1 2 0 2

Mean PET ...... 7 7 7 7 7 7 7 7 7 7 7 7

2 3 6 4 2 3 7 7 9 2 8 8 8 6 4 5 9 6 7 9 0 3 2 6

Mean AET ...... 5 5 5 5 5 6 5 5 6 6 6 5

Mean 4 3 9 4 8 2 3 4 8 9 1 7 2 3 2 3 3 3 4 2 2 3 1 3 ......

temperature 5 5 5 5 5 5 5 5 5 5 5 5

1 5 1 6 3 2 5 8 4 9 2 7 3 0 8 8 3 5 1 6 7 8 9 0

Mean ppt ...... 3 3 2 2 3 4 3 3 3 3 3 3

5 9 2 7 0 5 2 3 4 9 6 1 8 0 4 8 4 4 9 1 6 6 9 8 ...... Minimum . 8 7 1 9 3 0 5 9 3 5 7 2

latitude 3 2 3 2 3 2 2 3 4 2 2 2 ------

Global 9 4 4 3 7 5 8 7 7 6 5 6 3 7 0 9 3 8 0 5 8 4 6 3 geographic ...... 5 4 4 3 5 2 4 5 5 4 4 2 range 1 1 1 1 1 1 1 1 1 1 1 1

0 2 7 7 8 1

Home range . . . 0 0 1 - - -

7 7 1 8 0 7 1 0 2 5 1

Age at sexual 0 0 1 0 5 3

6 0 4 0 4 ...... 1 0 1 1 0 0 1 0 0 0 0

maturity ------

0 4 6 1 3 7 9 9 9 1 9

1 9 6 0 6 8 6 6 6 7 3

Litter size ...... 2 1 1 2 1 1 0 0 0 0 1

Inter birth 7 1 1 1 0 5 1

5 2 2 5 9 1 6 ......

interval 4 5 5 4 5 7 5

Gestation 7 6 5 6 9 8 7 6

7 4 6 4 7 8 6 8 ......

length 2 2 2 2 2 2 4 2

4 9 2 1 1 1 9 2 1 9 8 0 1 6 0 9 5 6 6 8 8 3 5 1 ...... Body mass . . . 4 3 4 3 4 4 3 3 2 1 3 1 ------

t t t t t t t t t t t 5 Western n n n n n n n n n n n a a a a a a a a a a 3 a t t t t t t t t t t t 9 x x x x x x x x x x x

deserts 1 e e e e e e e e e e e

t t t t t t t t t 0 5 n n n n n n n n n B 7 5 a a a a a a a a a P t t t t t t t t t

Pit & Finke 9 9 x x x x x x x x x N 1 1 e e e e e e e e e

t t t t t n n n n n B B B B B B B a a a a a P P P P P P P Flinders t t t t t x x x x x N N N N N N N e e e e e

5

t t t t t t t t t 2 5 n n n n n n n n n . B 8 a a a a a a a a a 1 P t t t t t t t t t

BMTS 9 8 x x x x x x x x x N 1 9 e e e e e e e e e 1

s r

e a e s e p i g i l i d t

l a i r n l r i o

t l a

i h l o s s s

a h t u u s s s s s s s u

i i i i i i t

s s a

Binomial a s s

u

i

s s s s s s

s a u r d u u a s l a n p p p p p p a o n l r u s u o o l u u o o o o o o o d t s a u y z z u a s i c g i h h h h h h c a o r o o c o t t t t t t y g n m g i e e r h a h h n n n n n n h n l r u o g i i i i c i i c p c d p p z u u o i l d n a a a a a r y m m m m m m c e e o u o S S S l S m S S y T a T T g T T v Z p

150 Appendices

Appendix 9.5 Data presented in chapter seven

Scale of studies in analysis presented in chapter seven.

Global Local Bennett & Owens (1997) Brashares (2003) Cardillo & Bromham (2001) Duncan & Blackburn (2004) Cardillo (2003) Foufopoulos & Ives (1999) Cardillo et al. (2004b) Harcourt & Schwartz (2001) Cassey (2001) Harcourt (1998) Collen et al. (in press) Isaac & Cowlishaw (2004) Dulvy & Reynolds (2002) Jones et al. (2001b) Duncan & Lockwood (2001) Jones et al. (2001a) Fisher et al. (2003) Koh et al. (2004b) Gaston & Blackburn (1995) Kotze & O'Hara (2003) Grenyer & Purvis (in press) Sullivan et al. (2000) Harcourt et al. (2002) Woodroffe & Ginsberg (1998) Johnson et al. (2002) Jones et al. (2003) Morrow & Pitcher (2003) Norris & Harper (2004) Owens & Bennett (2000) Purvis et al. (2000b) Reed & Shine (2002) Safi & Kerth (2004)

Raw numbers of studies testing a hypothesis that a predictor determines extinction risk for local and global studies. Values are numbers of studies showing the expected sign of response.

Global Local Predictor Expected sign Total Expected sign Total Body mass 11 18 12 13 Geographic range 9 10 7 9 Home range 4 4 4 4 Population density 4 5 5 5 Specialisation 12 14 7 10 Speed of life history 25 26 0 3 Reproductive output 16 18 0 1

151 References

References

Abbott I. (2002) Origin and spread of the cat, Felis catus, on mainland Australia, with a discussion of the magnitude of its early impact on native fauna. Wildlife Research, 29, 51-74

Agapow P.-M., Bininda-Emonds O.R.P., Crandall K.A., Gittleman J.L., Mace G.M., Marshall J.C. & Purvis A. (2004) The impact of species concept on biodiversity studies. Quarterly Review of Biology, 79, 161-179

Akçakaya H.R., Ferson S., Burgman M.A., Keith D.A., Mace G.M. & Todd C.R. (2000) Making consistent IUCN classifications under uncertainty. Conservation Biology, 14, 1001-1013

Allsop P.G. (1997) Probability of describing an Australian scarab beetle: influence of body size and distribution. Journal of Biogeography, 24, 717-724

Alroy J. (2003) Taxonomic inflation and body mass distributions in north American fossil mammals. Journal of Mammalogy, 84, 431-443

Alvo R. & Oldham M.J. (2000) A review of the status of Canada's amphibian and reptile species: a comparison of three ranking systems. The Canadian Field Naturalist, 114, 520-540

Anderson J.L. (1998) Embracing uncertainty: the interface of Baysian statistics and cognitive psychology. Conservation Ecology (online), 2, http://www.consecol.org/vol2/iss1/art2/

Australian Government Department of the Environment and Heritage (2005) The Interim Biogeographic Regionalisation for Australia (IBRA). URL: http://www.deh.gov.au/parks/nrs/ibra/.

Baillie J.E.M. & Groombridge D. (1996) 1996 IUCN Red List of threatened animals. IUCN, Gland, Switzerland.

Baillie J.E.M., Hilton-Taylor C. & Stuart S.N. (2004) 2004 IUCN Red List of threatened species: a global species assessment. IUCN, Gland, Switzerland and Cambridge, UK.

Balmford A. (1996) Extinction filters and current resilience: The significance of past selection pressures for conservation biology. Trends in Ecology & Evolution, 11, 193-196

Balmford A., Bennun L.A., ten Brink B., Cooper D., Côté I.M., Crane P., Dobson A., Dudley N., Dutton I., Green R.E., Gregory R.D., Harrison J., Kennedy E.T., Kremen C., Leader-Williams N., Lovejoy T.E., Mace G.M., May R., Mayaux P., Morling P., Phillips J., Redford K., Ricketts T.H., Rodrígues J.P., Sanjayan M., Schei P.J., van Jaarsveld A.S. & Walther B.A. (2005) The Convention on Biological Diversity's 2010 target. Science, 307, 212-213

152 References

Balmford A., Green R.E. & Jenkins M. (2003) Measuring the changing state of nature. Trends in Ecology & Evolution, 18, 326-330

Baynes A. & Johnson K.A. (1996) The contributions of the Horn Expedition and cave deposits to knowledge of the original mammal fauna of central Australia. In: Exploring central Australia: Society, the environment and the 1894 Horn expedition (eds. Morton SR & Mulvaney DJ), pp. 168-186. Surrey, Beatty & Sons, Chipping Norton, NSW

Beck R. (2003) Interordinal relationships in mammals. M.Sc Thesis, Biological Sciences, Imperial College London, London, UK.

Beissinger S.R. (1997) Integrating behaviour into conservation biology: potentials and limitations. In: Behavioural approaches to conservation in the wild (eds. Clemmons JR & Buchholz R), pp. 3-32. Cambridge University Press, Cambridge, UK

Bennett P.M. & Owens I.P.F. (1997) Variation in extinction risk among birds: chance or evolutionary predisposition. Proceedings of the Royal Society of London B, 264, 401-408

Benton M.J. (1994) Palaeontological data and identifying mass extinctions. Trends in Ecology & Evolution, 9, 181-185

Bergmann C. (1847) Uber die verlhaltnisse der warmeokonomie der thiere zu ihrer grosse. Gottinger Studien, 3, 595-708

Bininda-Emonds O.R.P., Gittleman J.L. & Purvis A. (1999) Building large trees by combining phylogenetic information: a complete phylogeny of the extant Carnivora (Mammalia). Biological Review, 74, 143-175

Bjorklund M. (1997) Are "comparative methods" always necessary? Oikos, 80, 607-612

Blackburn T.M. & Gaston K.J. (1995) What determines the probability of discovering a species? A study of South American oscine passerine birds. Journal of Biogeography, 22, 7-14

Blackburn T.M. & Gaston K.J. (2002) Linking patterns in macroecology. Journal of Animal Ecology, 70, 338-352

Bonn A., Rodriguez A.S.L. & Gaston K.J. (2002) Threatened and endemic species: are they good indicators of patterns of biodiversity on a national scale? Ecology Letters, 5, 733-741

Braithwaite R.W. & Muller W.J. (1997) Rainfall, groundwater and refuges: predicting extinctions of Australian tropical mammals. Australian Journal of Ecology, 22, 57-67

153 References

Brashares J.S. (2003) Ecological, behavioural and life-history correlates of mammal extinctions in West Africa. Conservation Biology, 17, 733-743

Brink H., Topp-Jørgensen J.E., Marshall A.R. & Fanning E. (2002) First record in sixty- eight years of Lowe's servaline genet (Genetta servalina lowei). Oryx, 36, 324

Burbidge A.A. & Fuller P.J. (1979) Mammals of the Warburton region, Western Australia. Records of the Western Australian Museum, 8, 57-73

Burbidge A.A., Johnson K.A., Fuller P.J. & Southgate R.I. (1988) Aboriginal knowledge of the mammals of the central deserts of Australia. Australian Wildlife Research, 15, 9-39

Burbidge A.A. & McKenzie N.L. (1989) Patterns in the modern decline of Western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation, 50, 143-198

Burgman M.A., Grimson R.C. & Ferson S. (1995) Inferring threat from scientific collections. Conservation Biology, 9, 923-928

Burton J. (2003) On Red Lists and IUCN. Plant Talk, 32, 4-5

Butchart S.H.M., Stattersfield A.J., Baillie J.E.M., Bennun L.A., Stuart S.N., Akçakaya H.R., Hilton-Taylor C. & Mace G.M. (2005) Using Red List Indices to measure progress towards the 2010 target and beyond. Philosophical Transactions of the Royal Society of London B, 360, 255-268

Butchart S.H.M., Stattersfield A.J., Bennun L.A., Shutes S.M., Akçakaya H.R., Baillie J.E.M., Stuart S.N., Hilton-Taylor C. & Mace G.M. (2004) Measuring global trends in the status of biodiversity: Red List Indices for birds. PLoS Biology, 2, 2294-2304

Butchart S.H.M., Stattersfield A.J. & Brooks T.M. (in press) Going or gone: defining 'Possibly Extinct' species to give a truer picture of recent extinctions. Bulletin of the British Ornithologists' Club

Cabrero-Sanudo F.J. & Lobo J.M. (2003) Estimating the number of species not yet described and their characteristics: the case of the Western Palaearctic dung beetle species (Coleoptera, Scarabaeoidea). Biodiversity and Conservation, 12, 147-166

Calsbeek R., Alonzo S.H., Zamudio K. & Sinervo B. (2002) Sexual selection and alternative mating behaviours generate demographic stochasticity in small populations. Proceedings of the Royal Society of London B, 269, 157-164

Carbone C., Cowlishaw G., Isaac N.J.B. & Rowcliffe J.M. (2005) How far do animals go? Scaling of day range in mammals. American Naturalist, 165

Cardillo M. (2003) Biological determinants of extinction risk: why are smaller species less vulnerable? Animal Conservation, 6, 1-7

154 References

Cardillo M., Bininda-Emonds O.R.P., Boakes E. & Purvis A. (2004a) A species-level phylogenetic supertree of marsupials. Journal of Zoology, 264, 11-31

Cardillo M. & Bromham L. (2001) Body size and risk of extinction in Australian mammals. Conservation Biology, 15, 1435-1440

Cardillo M., Mace G.M., Jones K.E., Bielby J., Bininda-Emonds O.R.P., Sechrest W., Orme C.D.L. & Purvis A. (2005) Multiple causes of high extinction risk in large mammal species. Science, 309, 1239-1241

Cardillo M., Purvis A., Sechrest W., Gittleman J.L., Bielby J. & Mace G.M. (2004b) Human population density and extinction risk in the world's carnivores. PLoS Biology, 2, 909-914

Cassey P. (2001) Determining variation in the success of New Zealand land birds. Global Ecology & Biogeography, 10, 161-172

Caughley G. (1994) Directions in Conservation Biology. Journal of Animal Ecology, 63, 215-244

Ceballos G. & Ehrlich P.R. (2002) Mammal population losses and the extinction crisis. Science, 296, 904-907

Cervantes F.A., Lorenzo C. & Gonzáles-Cózatl F.X. (2004) The Omiltemi rabbit (Sylvilagus insonus) is not extinct. Mammalian Biology, 69, 61-64

Collar N.J. (1996) The reasons for Red Data Books. Oryx, 30, 121-130

Collar N.J. (1998) Extinction by assumption; or, the Romeo error on Cebu. Oryx, 32, 239-244

Collar N.J., Andreev A.V., Chan S., Crosby M.J., Subramanya S. & Tobias J.A. (2001) Threatened Birds of Asia: the BirdLife International Red Data Book. BirdLife International, Cambridge, UK.

Collar N.J. & Andrew P. (1988) Birds to watch: The ICBP world check-list of threatened birds. International Council for Bird Preservation, Cambridge, UK.

Collen B., Bykova E., Ling S., Milner-Gulland E.J. & Purvis A. (in press) Extinction risk: a comparative analysis of central Asian vertebrates. Biodiversity and Conservation, special issue

Cooke P. (1980) Optimal linear estimation of bounds of random variables. Biometrika, 67, 257-258

COSEWIC (2005) Committee on the status of endangered wildlife in Canada. URL: http://www.cosewic.gc.ca.

Cramp S. (1980) Handbook of the birds of Europe, the Middle East and north Africa - The birds of the Western Palearctic. Oxford University Press, Oxford, UK.

155 References

Crawley M.J. (2002) Statistical Computing: An introduction to data analysis using S- Plus. Wiley, Chichester, UK.

Daily G.C. (1997) Nature's services: societal dependence on natural ecosystems. Island Press, Washington DC, USA.

Diamond J.M. (1985) How many unknown species are yet to be discovered? Nature, 315, 538-539

Diamond J.M. (1987) Extant unless proven extinct? Or, extinct unless proven extant? Conservation Biology, 1, 77-79

Dickman C.R. (1993) The Biology and management of native rodents of the arid zone in NSW. NSW National Parks & Wildlife service, Hurstville, Australia.

Dickman C.R., Lunney D. & Matthews A. (2000) Ecological attributes and conservation of native rodents in New South Wales. Wildlife Research, 27, 347-355

Dickman C.R., Pressey R.L., Lim L. & Parnaby H.E. (1993) Mammals of particular conservation concern in the western division of New South Wales. Biological Conservation, 65, 219-248

Diniz-Filho J.A.F., Bastos R.P., Rangel T.F.L.V.B., Bini L.M., Carvalho P. & Silva R.J. (2005) Macroecological correlates and spatial patterns of anuran description dates in Brazilian cerrdo. Global Ecology & Biogeography, 14, 469-477

Dulvy N. & Reynolds J.D. (2002) Predicting extinction vulnerability in skates. Conservation Biology, 16, 440-450

Duncan J.R. & Lockwood J.L. (2001) Extinction in a field of bullets: a search for causes in the decline of the world's freshwater fishes. Biological Conservation, 102, 97- 105

Duncan R.P. & Blackburn T.M. (2004) Extinction and endemism in the New Zealand avifauna. Global Ecology & Biogeography, 13, 509-517

Estes R. & Pregill G. (1988) Phylogenetic relationships of the lizard families: essays commemorating Charles L. Camp. Stanford University Press, Stanford (Ca), USA.

Felsenstein J. (1985) Phylogenies and the comparative method. American Naturalist, 125, 1-15

Finlayson H.H. (1961) On Central Australian mammals. Part IV - the distribution and status of Central Australian species. Records of the South Australian Museum, 11, 141-191

156 References

Fisher D.O., Blomberg S.P. & Owens I.P.F. (2003) Extrinsic versus intrinsic factors in the decline and extinction of Australian marsupials. Proceedings of the Royal Society of London B, 270, 1801-1808

Fisher D.O. & Owens I.P.F. (2004) The comparative method in conservation biology. Trends in Ecology and Evolution, 19, 391-398

Fitzpatrick J.W., Lammertink M., Luneau M.D., Gallagher T.W., Harrison B.R., Sparling G.M., Rosenberg K.V., Rohrbaugh R.W., Swarthout E.C.H., Wrege P.H., Swarthout S.B., Dantzker M.S., Charif R.A., Barksdale T.R., Remsen J.V., Simon S.D. & Zollner D. (2005) Ivory-billed Woodpecker (Campephilus principalis) persists in continental North America. Science, 308, 1460-1462

Flannery T.F. (1989) Australia's vanishing mammals. RD Press, Sydney, Australia.

Forsyth D.M., Duncan R.P., Bomford M. & Moore G. (2004) Climatic suitability, life- history traits, introduction effort and the establishment and spread of introduced mammals in Australia. Conservation Biology, 18, 557-569

Foufopoulos J. & Ives A.R. (1999) Reptile extinctions on land-bridge islands: life-history attributes and vulnerability to extinction. The American Naturalist, 153, 1-25

Freckleton R.P., Harvey P.H. & Pagel M. (2002) Phylogenetic analysis and comparative data: a test and review of evidence. The American Naturalist, 160, 712-726

Freckleton R.P., Harvey P.H. & Pagel M. (2003) Bergmann's rule and body size in mammals. The American Naturalist, 161, 821-825

Garland T., Harvey P.H. & Ives A.R. (1992) Procedures for analysis of comparative data using phylogenetically independent contrasts. Systematic Biology, 41, 18-32

Gaston K.J. (1991) Body size and probability of description: the beetle fauna of Britain. Ecological Entomology, 16, 505-508

Gaston K.J. (1994) Rarity. Chapman & Hall, London, UK.

Gaston K.J. & Blackburn T.M. (1994) Are newly described bird species small bodied? Biodiversity Letters, 2, 16-20

Gaston K.J. & Blackburn T.M. (1995) Birds, body size and the threat of extinction. Philosophical Transactions of the Royal Society of London B, 347, 205-212

Gaston K.J. & Blackburn T.M. (2000) Pattern & Process in Macroecology. Blackwell Science, Oxford, UK.

Gaston K.J., Blackburn T.M. & Loder N. (1995a) Which species are described first? The case of North American butterflies. Biodiversity and Conservation, 4, 119-127

157 References

Gaston K.J., Scoble M.J. & Crook A. (1995b) Patterns in species description: a case study using the Geometridae (Lepidoptera). Biological Journal of the Linnaean Society. Biological Journal of the Lineman Society, 55, 225-237

Gibson D.F. (1986) A Biological Survey of the Tanami desert in the Northern Territory. Conservation Commission of the Northern Territory, Alice Springs, NT, Australia.

Gibson D.F. & Cole J.R. (1988) A Biological Survey of the Northern Simpson Desert. A Consultancy report to the conservation commission of the Northern Territory, Alice Springs, NT, Australia.

Gibson D.F. & Cole J.R. (1993) Vertebrate fauna of the West MacDonnell Ranges, Northern Territory. Conservation Commission of the Northern Territory, Alice Springs, NT, Australia.

Gill A.M. (1989) Priorities for fire research in Australia. Bulletin of the Ecological Society of Australia, 19, 74-76

Gittleman J.L. & Purvis A. (1998) Body size and species-richness in carnivores and primates. Proceedings of the Royal Society of London B, 265, 113-119

Given D. (2003) On Red Lists and IUCN: a response. Plant Talk, 34, 6

Godfray H.C.J. (2002) Challenges for taxonomy. Nature, 417, 17-19

Greene C., Umbanhowar J., Mangel M. & Caro T. (1998) Animal breeding systems, hunter selectivity and consumptive use in wildlife conservation. In: Behavioural ecology and conservation biology (ed. Caro T), pp. 271-305. Oxford University Press, Oxford, UK

Grenyer R. & Purvis A. (2003) A composite species-level phylogeny of the 'Insectivora' (Mammalia: Order Lipotyphla Haeckel, 1866). Journal of Zoology, 260, 245-257

Grenyer R. & Purvis A. (in press) Extinction of the "Insectivora": biological and macroecological correlates of extinction risk. Animal Conservation

Groombridge B. (1992) Global Biodiversity: status of the World's living resources. Chapman & Hall, London, UK.

Groves C.P. (2001a) Primate taxonomy. Smithsonian Institution Press., Washington DC, USA.

Groves C.P. (2001b) Why taxonomic stability is a bad idea, or why are there so few species of primates (or are there?). Evolutionary Anthropology, 10, 192-198

Hall P. & Wang J.Z. (1999) Estimating the end point of a probability distribution using minimum-distance methods. Bernoulli, 5, 177-189

158 References

Harcourt A.H., Coppeto S.A. & Parks S.A. (2002) Rarity, specialization and extinction in primates. Journal of Biogeography, 29, 445-456

Harcourt A.H. & Schwartz M.W. (2001) Primate Evolution: A biology of Holocene extinction and survival on the southeast Asian Sunda shelf islands. American Journal of Physical Anthropology, 114, 4-17

Harcourt S.A. (1998) Ecological indicators of risk for primates, as judged by species' susceptibility to logging. In: Behavioural Ecology and Conservation Ecology (ed. Caro T), pp. 56-79. Oxford University Press, New York, USA

Harvey P.H. & Pagel. M.D. (1991) The comparative method in evolutionary biology. Oxford University Press, Oxford, UK.

Harvey P.H. & Rambaut A. (1998) Phylogenetic extinction rates and comparative methodology. Proceedings of the Royal Society of London Series B-Biological Sciences, 265, 1691-1696

Hawksworth D.L. & Kalin-Arroyo M.T. (1995) Magnitude and distribution of biodiversity. In: Global Biodiversity Assessment (eds. Heywood VH & Watson RT). United Nations Environment Programme & Cambridge University Press, Cambridge, UK

Hey J. (2001) Genes, categories and species: The evolutionary and cognitive cause of the species problem. Oxford University Press, New York, USA.

Heywood V. (2003) Red Listing: too clever by half? Plant Talk, 31, URL: http://www.plant-talk.org/stories/31vheywd.html

Honacki J.H., Kinnman K.E. & Koeppl J.W. (1982) Mammal species of the world: a taxonomic and geographic reference. Allen Press, Inc. and The Association of Systematics Collections, Kansas, USA.

Honess P.E. & Bearder S.K. (1998) Two new primate species from Tanzania: Galagoides udzungwensis and Galagoides rondoensis. African Primates, 2, 11-15

Hughes T.P., Bellwood D.R. & Connolly S.R. (2002) Biodiversity hotspots, centres of endemicity, and the conservation of coral reefs. Ecology Letters, 5, 775-784

Isaac N.J.B. & Cowlishaw G. (2004) How species respond to multiple extinction threats: evidence from primates. Proceedings of the Royal Society of London B, 271, 1135-1141

Isaac N.J.B., Mallet J. & Mace G.M. (2004) Taxonomic inflation: its influence on macroecology and conservation. Trends in Ecology and Evolution, 19, 464-469

IUCN (1994) IUCN Red List categories. IUCN, Gland, Switzerland.

IUCN (2001) IUCN Red List Categories and Criteria version 3.1. IUCN, Gland, Switzerland.

159 References

IUCN (2003) Guidelines for application of IUCN Red list criteria at regional levels version 3.0. IUCN, Gland, Switzerland.

IUCN (2004) 2004 IUCN Red List of threatened species. IUCN, Gland, Switzerland.

Johnson C.N., Delean S. & Balmford A. (2002) Phylogeny and the selectivity of extinction in Australian marsupials. Animal Conservation, 5, 135-142

Jones K.E., Barlow K.E., Vaughan N., Rodrigues-Duran A. & Gannon M.R. (2001a) Short-term impacts of extreme environmental disturbance on the bats of Puerto Rico. Animal Conservation, 4, 59-66

Jones K.E. & Purvis A. (1997) An optimum body size for mammals? Comparative evidence from bats. Functional Ecology, 11, 751-756

Jones K.E., Purvis A. & Gittleman J.L. (2003) Biological correlates of extinction risk in bats. The American Naturalist, 161, 601-614

Jones K.E., Purvis A., MacLarnon A., Bininda-Emonds O.R.P. & Simmons N.B. (2002) A phylogenetic supertree of the bats (Mammalia: Chiroptera). Biological Review, 77, 223-259

Jones M.J., Sullivan M.S., Marsden S.J. & Lindsay M.D. (2001b) Correlates of extinction risk of birds from two Indonesian islands. Biological Journal of the Linnaean Society, 73, 65-79

Jones T., Ehardt C.L., Butynski T.M., Davenport T.R.B., Mpunga N.E., Machaga S.J. & De Luca D.W. (2005) The highland mangabey Lophocebus kipunji: a new species of African monkey. Science, 308, 1161-1164

Keith D.A. (1998) An evaluation and modification of World Conservation Union Red List criteria for classification of extinction risk in vascular plants. Conservation Biology, 12, 1076-1090

Keith D.A. & Burgman M.A. (2004) The Lazarus effect: can the dynamics of extinct species lists tell us anything about the status of biodiversity? Biological Conservation, 117, 41-48

Keith D.A., McCarthy M.A., Regan H., Regan T., Bowles C., Drill C., Craig C., Pellow B., Burgman M.A., Master L.L., Ruckelshaus M., Mackenzie B., Andelman S.J. & Wade P.R. (2004) Protocols for listing threatened species can forecast extinction. Ecology Letters, 7, 1101-1108

Kerle J.A., Foulkes J.N., Kimber R.G. & Papenfus D. (1992) The decline of the Brushtail possum, Trichosurus vulpecula (Kerr 1798), in arid Australia. Rangeland Journal, 14, 107-127

Kerr J.T. & Currie D.J. (1995) Effects of human activity on global extinction risk. Conservation Biology, 9, 1528-1538

160 References

Koh L.P., Dunn R.R., Sodhi N.S., Colwell R.K., Proctor H.C. & Smith V.S. (2004a) Species coextinctions and the biodiversity crisis. Science, 305, 1632-1634

Koh L.P., Sodhi N.S. & Brook B.W. (2004b) Ecological correlates of extinction proneness in tropical butterflies. Conservation Biology, 18, 1571-1578

Kotze D.J. & O'Hara R.B. (2003) Species decline - but why? Explanations of carabid beetle (Coleoptera, Carabidae) declines in Europe. Oecologia, 135, 138-148

Lamoreux J., Akcakaya H.R., Bennun L., Collar N.J., Boitani L., Brackett D., Brautigam A., Brooks T.M., de Fonseca G.A.B., Mittermeier R.A., Rylands A.B., Gardenfors U., Hilton-Taylor C., Mace G., Stein B.A. & Stuart S. (2003) Value of the IUCN Red List. Trends in Ecology & Evolution, 18, 214-215

Lawton J.H. & May R.M. (1995) Extinction rates. Oxford University Press, Oxford, UK.

Lockwood J.L., Russell G.J., Gittleman J.L., Daehler C.C., McKinney M.L. & Purvis A. (2002) A metric for analyzing taxonomic patterns of extinction risk. Conservation Biology, 16, 1137-1142

Ludwig J.A. & Reynolds J.F. (1988) Statistical Ecology: A primer on methods and computing. John Wiley and Sons, New York, USA.

Lunney D. (2001) Causes of the extinction of native mammals of the western division of New South Wales: an ecological interpretation of the nineteenth century historical records. Rangeland Journal, 23, 44-70

Lunney D., Curtin A.L., Fisher D., Ayres D. & Dickman C.R. (1997) Ecological attributes of the threatened fauna of New South Wales. Pacific Conservation Biology, 3, 13-26

MacArthur R.H. & Wilson E.O. (1967) The Theory of Island Biogeography. Princeton, New Jersey.

Mace G.M. (1995) Classification of threatened species and its role in conservation planning. In: Extinction rates (eds. Lawton JH & May RM), pp. 197-213. Oxford University Press, Oxford, UK

Mace G.M. (2004) The role of taxonomy in species conservation. Philosophical Transactions of the Royal Society of London Series B-Biological Sciences, 359, 711-719

Mace G.M. & Balmford A. (2000) Patterns and processes in contemporary mammalian extinction. In: Priorities for conservation of mammalian diversity: has the panda had its day? (eds. Entwhistle A & Dunstone N). Cambridge University Press, Cambridge, UK

Mace, G. M. & Lande, R. (1991) Assessing extinction threats - toward a reevaluation of IUCN threatened species categories. Conservation Biology, 5, 148-157

161 References

Mace G.M., Gittleman J.L. & Purvis A. (2003) Preserving the Tree of Life. Science, 300, 1707-1709

Macknight C.C. (1976) The voyage to Marege: Macassan Trepangers in Northern Australia. Melbourne University Press, Melbourne, Australia.

MacPhee R.D.E. & Flemming C. (1999) Requiem Aeternam: the last five hundred years of mammalian species extinctions. In: Extinctions in near time: causes, contexts and consequences (ed. MacPhee RDE), pp. 333-371. Kluwer Academic & Plenum Publishers, New York, USA

Mammal SuperTeam (2005) Pantheria database. URL: http://www.columbia.edu/~kj2107/superteam/.

Marshall C.R. (1994) Confidence intervals on stratigraphic ranges: partial relaxation of the assumption of randomly distributed fossil horizons. Palaeobiology, 20, 459- 469

Master L.L. (1991) Assessing threats and setting priorities for conservation. Conservation Biology, 5, 559-563

Master L.L., Stein B.A., Kutner L.S. & Hammerson G.A. (2000) Vanishing assets: conservation status of U.S. species. In: Precious heritage: the status of biodiversity in the United States (eds. Stein BA, Kutner LS & Adams JS), pp. 93- 118. Oxford University Press, New York, USA

Maxwell S., Burbidge A.A. & Morris K. (1996) Action plan for Australian marsupials and monotremes: Australasian marsupial and monotreme specialist group, IUCN Species Survival Commission. IUCN, Gland, Switzerland.

May R.M. (1988) How many species are there on earth? Science, 241, 1441 - 1449

May R.M. (1994) The effects of spatial scale on ecological questions and answers. In: Large-scale ecology and conservation biology (eds. Edwards PJ, May RM & Webb NR), pp. 1-18. Blackwell Scientific, Oxford, UK

May R.M., Lawton J.H. & Stork N.E. (1995) Assessing extinction rates. In: Extinction rates (eds. Lawton JH & May RM), pp. 1-24. Oxford University Press, Oxford, UK

Mayr E. (1956) Geographical character gradients and climatic adaptation. Evolution, 10, 105-108

Mayr E. (1963) Animal species and evolution. Harvard University Press, Cambridge, Massatusettes, USA.

McCarthy M.A. (1998) Identifying declining and threatened species with museum data. Biological Conservation, 83, 9-17

162 References

McKay G.M. (1996) Feral cats in Australia: origins and impacts. In: Unwanted Aliens (ed. Diekman B), pp. 9-16. Nature Conservation Council of NSW, Sydney, Australia

McKenzie N.L. & Youngson W.K. (1983) Mammals. In: Wildlife of the Great Sandy Desert, Western Australia (eds. Burbidge AA & McKenzie NL), pp. 62-93. Department of Fisheries and Wildlife, Perth, Australia

McKinney M.L. (1997) Extinction vulnerability and selectivity: combining ecological and paleontological views. Annual Review of Ecology and Systematics, 28, 495- 516

McKinney M.L. (2001) Role of human population size in raising bird and mammal threat among nations. Animal Conservation, 4, 45-57

Millennium Ecosystem Assessment (2005) Ecosystems and human well-being: synthesis. Island Press, Washington DC, USA.

Millsap B.A., Gore J.A., Runde D.E. & Cerulean S.I. (1990) Setting priorities for the conservation of fish and wildlife species in Florida. Wildlife Monographs, 111, 1- 57

Milner-Gulland E.J., Kreuzberg-Mukhina E., Grebot B., Ling S., Bykova E., Abdusalamov I., Bekenov A., Gardenfors U., Hilton-Taylor C., Salnikov V. & Stogova L. (in press) Application of IUCN red listing criteria at the regional and national levels: A case study from Central Asia. Biodiversity and Conservation

Morrow E.H. & Pitcher T.E. (2003) Sexual selection and the risk of extinction in birds. Proceedings of the Royal Society of London B, 270, 1793-1799

Mrosovsky N. (2003) Predicting extinctions: fundamental flaws in IUCN's Red List system, exemplified by the case of sea turtles. Department of Zoology, University of Toronto, Toronto, Canada.

Myers N. (1979) The sinking ark: a new look at the problem of disappearing species. Pergamon Press, New York, USA.

Myers N., Mittermeier R.A., Mittermeier C.G., de Fonseca G.A.B. & Kent J. (2000) Biodiversity hotspots for conservation priorities. Nature, 403, 853-858

National Land & Water Audit (2005) Australian Natural Resource Atlas. URL: http://audit.ea.gov.au/anra/.

Newsome A. (1995) Cat. In: The mammals of Australia (ed. Strahan R), pp. 700-702. Reed Books, Sydney, Australia

Norris K. & Harper N. (2004) Extinction processes in hot spots of avian biodiversity and the targeting of pre-emptive conservation action. Proceedings of the Royal Society of London B, 271, 123-130

163 References

NSW Department of Environment & Conservation (2005) Atlas of NSW Wildlife. URL: http://www.nationalparks.nsw.gov.au/.

O'Grady J.J., Burgman M.A., Keith D.A., Master L.L., Andelman S.J., Brook B.W., Hammerson G.A., Regan T. & Frankham R. (2004) Correlations among extinction risks assessed by different systems of threatened species categorization. Conservation Biology, 18, 1624-1635

Owens I.P.F. & Bennett P.M. (2000) Ecological basis of extinction risk in birds: habitat loss versus human persecution and introduced predators. Proceedings of the National Academy of Sciences USA, 97, 12144-12148

Pagel M. (1999) Inferring the historical patterns of biological evolution. Nature, 401, 877-884

Pagel M.D. (1992) A method for the analysis of comparative data. Journal of Theoretical Biology, 156, 431-442

Pagel M.D., May R.M. & Collie A.R. (1991) Ecological aspects of the geographic distribution and diversity of mammal species. American Naturalist, 137, 791-815

Patterson B.D. (1994) Accumulating knowledge on the dimensions of biodiversity: systematic perspectives on Neotropical mammals. Biodiversity Letters, 2, 79-86

Pimenta B.V.S., Haddad C.F.B., Nascimento L.B., Cruz C.A.G. & Pombal (Jnr) J.P. (2005) Comment on "Status and trends of amphibian declines and extinctions worldwide". Science, 309, 1999

Pimm S.L. & Brooks T.M. (2000) The sixth extinction: how large, where and when? In: Nature and human society. The quest for a sustainable world (ed. Raven PH), pp. 46-62. National Academy Press, Washington DC, USA

Pimm S.L., Jones H.L. & Diamond J. (1988) On the risk of extinction. The American Naturalist, 132, 757-785

Pimm S.L., Russell G.J., Gittleman J.L. & Brooks T.M. (1995) The future of biodiversity. Science, 269, 347-350

Possingham H.P., Andelman S.J., Burgman M.A., Medellin R.A., Master L.L. & Keith D.A. (2002a) Limits to the use of threatened species lists. Trends in Ecology and Evolution, 17, 503-506

Possingham H.P., Andelman S.J., Burgman M.A., Medellín R.A., Master L.L. & Keith D.A. (2002b) Limits to the use of threatened species lists. Trends in Ecology & Evolution, 17, 503-507

Pressey R.L., Humphries C.J., Margules C.R., Vane-Wright R.I. & Williams P. (1993) Beyond opportunism: key principles for systematic reserve selection. Trends in Ecology & Evolution, 8, 124-128

164 References

Price S.A., Bininda-Edmonds O.R.P. & Gittleman J.L. (2005) A complete phylogeny of the whales, dolphins and even-toed hoofed mammals (certartiodactyla). Biological Review, 80, 445-473

Price S.A. & Gittleman J.L. (in review) Bushmeat hunting, habitat loss and global extinction in the Artiodactyla. Proceedings of the Royal Society of London B

Purvis A. (1995) A composite estimate of Primate phylogeny. Philosophical Transactions of the Royal Society of London B, 348, 405-421

Purvis A., Agapow P.-M., Gittleman J.L. & Mace G.M. (2000a) Nonrandom extinction and the loss of evolutionary history. Science, 288, 328-330

Purvis A., Cardillo M., Grenyer R. & Collen B. (2005a) Correlates of extinction risk: phylogeny, biology, threat and scale. In: Phylogeny and Conservation (eds. Purvis A, Brooks TM & Gittleman JL), pp. 295-316. Cambridge University Press, Cambridge, UK

Purvis A., Gittleman J.L. & Brooks T.M. (2005b) Phylogeny and Conservation. Cambridge University Press, Cambridge, UK.

Purvis A., Gittleman J.L., Cowlishaw G. & Mace G.M. (2000b) Predicting extinction risk in declining species. Proceedings of the Royal Society of London B, 267, 1947- 1952

Purvis A. & Hector A. (2000) Getting the measure of biodiversity. Nature, 405, 212-219

Purvis A., Jones K.E. & Mace G.M. (2000c) Extinction. BioEssays, 22, 1123-1133

Purvis A., Orme C.D.L. & Dolphin K. (2003) Why are most species small bodied? A phylogenetic view. In: Macroecology: concepts and consequences (eds. Blackburn TM & Gaston KJ), pp. 155-173. Blackwell Scientific, Oxford, UK

Purvis A. & Rambaut A. (1995) Comparative analysis by independent contrasts (CAIC): an Apple Macintosh application for analysing comparative data. Computational Applications in Bioscience, 11, 247-251

R Development Core Team (2004) R: A language and environment for statistical computing. In. R Foundation for Statistical Computing, Vienna, Austria

Rambaut A. & Pagel M. (2001) CONTINUOUS. University of Oxford, Oxford, UK.

Raup D.M. (1991) Extinction: Bad genes or bad luck? Norton, New York.

Raup D.M. (1994) The role of extinction in evolution. Proceedings of the National Academy of Sciences USA, 91, 6758-6763

Read A.F. & Harvey P.H. (1989) Life history differences among eutherian radiations. Journal of Zoology, 219, 329-353

165 References

Reed M.J. (1996) Using statistical probability to increase confidence of inferring species extinction. Conservation Biology, 10, 1283-1285

Reed R.N. & Boback S.M. (2002) Does body size predict dates of species description among North American and Australian reptiles and amphibians? Global Ecology & Biogeography, 11, 41-47

Reed R.N. & Shine R. (2002) Lying in wait for extinction: ecological correlates of conservation status among Australian elapid snakes. Conservation Biology, 16, 451-461

Regan H., Colyvan M. & Burgman M.A. (2000) A proposal for fuzzy Internatinal Union for the Conservation of Nature (IUCN) categories and criteria. Biological Conservation, 92, 101-108

Regan H.M., Lupia R., Drinnan A.N. & Burgman M.A. (2001) The currency and tempo of extinction. American Naturalist, 157, 1-10

Regan T.J., Master L.L. & Hammerson G.A. (2004) Capturing expert knowledge for threatened species assessments: a case study using NatureServe conservation status ranks. Acta Oecologia, 26, 95-107

Roberts D.L. & Solow A.R. (2003) When did the Dodo become extinct? Nature, 426, 245

Rohlf D.J. (1991) Six biological reasons why the endangered species act doesn't work - and what to do about it. Conservation Biology, 5, 273-282

Rosenthal R. (1979) The "file drawer problem" and tolerance for null results. Psychological Bulletin, 86, 638-641

Rosenthal R. (1991) Meta-Analytic procedures for social research. Sage Publications, Newbury Park, California, USA.

Russell G.J., Brooks T.M., McKinney M.M. & Anderson C.G. (1998) Present and future taxonomic selectivity in bird and mammal extinctions. Conservation Biology, 12, 1365-1376

Safi K. & Kerth G. (2004) A compartive analysis of specialization and extinction risk in temperate-zone bats. Conservation Biology, 18, 1293-1303

Schwartz M.W. & Simberloff D. (2001) Taxon size predicts rates of rarity in vascular plants. Ecology Letters, 4, 464-469

Scott P., Burton J.A. & Fitter R. (1987) Red Data Books: the historical background. In: The road to extinction: problems of categorizing the status of taxa threatened with extinction (eds. Fitter R & Fitter M), pp. 1-5. IUCN, Gland, Switzerland & Cambridge, UK

Sechrest W. (2003) Global diversity, endemism and conservation of mammals. Ph.D. Thesis, Department of Biology, University of Virginia, Charlottesville.

166 References

Seoane J., Bustamante J. & Diaz-Delgado R. (2005) Effect of expert opinion on the predictive ability of environmental models of bird distribution. Conservation Biology, 19, 512-522

Short J. (1998) The extinction of rat-kangaroos (Marsupialia: Potoroidae) in New South Wales, Australia. Biological Conservation, 86, 365-377

Short J., Kinnear J.E. & Robley A. (2002) Surplus killing by introduced predators in Australia - evidence for ineffective anti-predator adaptations in native prey species? Biological Conservation, 103, 283-301

Short J. & Smith A. (1994) Mammal decline and recovery in Australia. Journal of Mammalogy, 75, 288-297

Sibley C.G. & Ahlquist J.E. (1990) Phylogeny and classification of birds. Tale University Press, New Haven, Connecticut, USA.

Sinha A., Datta A., Madhusudan M.D. & Mishra C. (2005) The Arunachal macaque Macaca munzala: a new species from western Arunachal Pradesh, northeastern India. International Journal of Primatology, 26, 977-989

Smith A.P. & Quin D.G. (1996) Patterns and causes of extinction and decline in Australian conilurine rodents. Biological Consevation, 77, 243-267

Smith R.L. & Weissman I. (1985) Maximum likelihood estimation of the lower tail of a probability distribution. Journal of the Royal Statistical Society Series B, 47, 285

Sokal R.R. & Rolf F.J. (1995) Biometry: the principles and practice of statistics in biological research. W. H. Freeman, New York, USA.

Solow A.R. (1993) Inferring extinction from sighting data. Ecology, 74, 962-964

Solow A.R. (2003) Estimation of stratigraphic ranges when fossil finds are not randomly distributed. Palaeobiology, 29, 181-185

Solow A.R. (2005) Inferring extinction from a sighting record. Mathematical Biosciences, 195, 47-55

Stevens G.C. (1989) The latitudinal gradient in geographic range: how so many species coexist in the tropics. American Naturalist, 133, 240-256

Strahan R. (1998) The Mammals of Australia. New Holland Publishers Ltd, Sydney, Australia.

Strauss D. & Sadler P.M. (1989) Classical confidence intervals and Bayesian probability estimates ends of local taxon ranges. Mathematical Geology, 21, 411-427

167 References

Stuart S.N., Chanson J.S., Cox N.A. & Young B.E. (2005) Response to comment on "Status and trends of amphibian declines and extinctions worldwide". Science, 309, 1999

Stuart S.N., Chanson J.S., Cox N.A., Young B.E., Rodrigues A.S.L., Fischman D.L. & Waller R.W. (2004) Status and trends of amphibian declines and extinctions worldwide. Science, 306, 1783-1786

Sullivan M.S., Gilbert F., Rotheray G., Croasdale S. & Jones M. (2000) Comparative analysis of correlates of Red data book status: a case study using European hoverflies (Diptera: Syrphidae). Animal Conservation, 3, 91-95

Townsend C.R., Harper J.L. & Begon M. (2000) Essentials of Ecology. Blackwell Science, Maiden, Massachusetts, USA.

Treves A. (1997) Primate Natal Coats: A preliminary analysis of distribution and function. American Journal of Physical Anthropology, 104, 47-70

Tunbridge D. (1991) The Story of the Flinders ranges mammals. Kangaroo Press, Kenthurst, NSW, Australia.

Tversky A. & Kahneman D. (1974) Judgement under uncertainty: heuristics and biases. Science, 185, 1124-1131

Vamosi J.C. & Vamosi S.M. (2005) Present day risk of extinction may exacerbate the lower species richness of dioecious clades. Diversity and Distributions, 11, 25-32 van der Ree R. & McCarthy M.A. (2005) Inferring persistence of indigenous mammals in response to urbanisation. Animal Conservation, 8, 309-319 van Roosmalen M.G.M., van Roosmalen T. & Mittermeier R.A. (2002) A taxonomic review of the Titi monkeys, genus Callicebus Thomas, 1903, with the description of two new species, Callicebus bernhardi and Callicebus stephennashi, from Brazilian Amazonia. Neotropical Primates, 10, (supplement) 1-52

Vernes K. & Haydon D.T. (2001) Effect of fire on northern bettong (Bettongia tropica) foraging behaviour. Austral Ecology, 26, 649-659

Victor J.E. & Keith M. (2004) The Orange List: a safety net for biodiversity in South Africa. South African Journal of Science, 100, 139-141

Waller D.M. & Rooney T.P. (2003) Nature is changing in more ways than one. Trends in Ecology & Evolution, 19, 6-7

Weaver J.L., Paquet P.C. & Ruggiero L.F. (1996) Resilience and conservation of large carnivores in the Rocky Mountains. Conservation Biology, 10, 964-976

Wilcove D.S. (2005) Rediscovery of the Ivory-billed Woodpecker. Science, 308, 1422- 1423

168 References

Wilcox B.A. (1980) Insular ecology and conservation. In: Conservation biology: an evolutionary-ecological perspective (eds. Soule ME & Wilcox BA). Sinauer Associates, Sunderland, Massachusetts, USA

Williamson T. (1994) Vagueness. Routledge, London, UK.

Wilson D.E. & Reeder D.M. (1993) Mammal species of the World: A taxonomic and geographic reference. second edition edn. Smithsonian Institution Press, Washington DC, USA.

Wilson E.O. (2003) The encyclopedia of life. Trends in Ecology and Evolution, 18, 77-80

Woodroffe R. & Ginsberg J.R. (1998) Edge effects and the extinction of populations inside protected areas. Science, 280, 2126-2128

Zhao Zhengije (1992) The history and controversy on the study of the Crested Shelduck. China Nature, 4, 17-19

Zhao Zhengije (1995) A handbook for the birds of China 1. Non-passerines. Jilin Science and Technology Press, Changchun, China.

Zimmermann E., Ehresmann P., Zietemann V. & Radespiel U. (1997) A new primate species in north-western Madagascar: the golden-brown mouse lemur (Microcebus ravelobensis). Primate eye, 63, 26

169