PALAEOECOLOGICAL AND BIOCHRONOLOGICAL STUDIES OF RIVERSLEIGH, WORLD HERITAGE PROPERTY, OLIGO-MIOCENE FOSSIL LOCALITIES, NORTH-WESTERN QUEENSLAND, AUSTRALIA

Kenny James TRAVOUILLON

Thesis submitted for the degree of Doctor of Philosophy at the University of New South Wales, Australia

August, 2008

i N° d‘ordre Année 2008

THESE

présentée

devant l‘UNIVERSITE CLAUDE BERNARD - LYON 1

pour l‘obtention

du DIPLOME DE DOCTORAT

(arrêté du 7 août 2006 et arrêté du 6 janvier 2005)

Soutenance : Novembre 2008

par

M. Kenny James TRAVOUILLON

TITRE :

ETUDES PALEOECOLOGIQUES ET BIOCHRONOLOGIQUES DES GISEMENTS FOSSILIFERES DE L‘OLIGO-MIOCENE DE RIVERSLEIGH, WORLD HERITAGE PROPERTY, NORD-OUEST DU QUEENSLAND, AUSTRALIE

Directeur de thèse : Dr Serge Legendre Prof Michael Archer Dr Suzanne Hand

JURY : Dr. Serge Legendre Directeur de Thèse Pr. Michael Archer Directeur de Thèse Pr. Christophe Lecuyer Examinateur Dr. Gilles Escarguel Examinateur Dr. Sophie Montuire Rapporteur Pr. Philip Gingerich Rapporteur Dr. Manuel Hernandez Fernandez Rapporteur

ii

THE UNIVERSITY OF NEW SOUTH WALES Thesis/Dissertation Sheet

Surname or Family name: Travouillon

First name: Kenny Other name/s: James

Abbreviation for degree as given in the University calendar: PhD

School: BEES Faculty: Science

Title: PALAEOECOLOGICAL AND BIOCHRONOLOGICAL STUDIES OF RIVERSLEIGH, WORLD HERITAGE PROPERTY, OLIGO-MIOCENE FOSSIL LOCALITIES, NORTH-WESTERN QUEENSLAND, AUSTRALIA

Abstract 350 words maximum:

Riversleigh, World Heritage Property, located in North-western Queensland, Australia, contains over 200 fossil bearing localities from the Oligo-Miocene. The study presented here aims at finding new methods to improve the accuracy of palaeoecological and biochronological studies and describe the palaeoenvironmental and chronological settings of the Riversleigh fossil deposits. One of the methods developed in this thesis, Minimum Sample Richness (MSR), determines the minimum number of species that must be present in a fauna to allow meaningful comparisons using multivariate analyses. Using MSR, several Riversleigh localities were selected for a palaeoecological study using the cenogram method to determine the palaeoenvironment during the Oligo-Miocene. Finally, the Numerical ages method was used to refine the relative ages of the Riversleigh localities and a re-diagnosis of the Riversleigh Systems is proposed.

Declaration relating to disposition of project thesis/dissertation

I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future works (such as articles or books) all or part of this thesis or dissertation.

I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is applicable to doctoral theses only).

………………………………………… ………………………………………… ……………………………………… Signature Witness Date

The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use. Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.

FOR OFFICE USE ONLY Date of completion of requirements for Award:

THIS SHEET IS TO BE GLUED TO THE INSIDE FRONT COVER OF THE THESIS

iii ______

RESUME en français

Riversleigh, World Heritage Property, dans le nord-ouest du Queensland en Australie, contient plus de 200 localités fossilifères de l‘Oligo-Miocène. L‘étude présentée ici vise à trouver de nouvelles méthodes pour améliorer l'exactitude des études paléoécologique et biochronologique et décrire les paramètres paléoenvironnementaux et chronologique des localités de Riversleigh. L'une des méthodes développées dans cette thèse, Minimum Sample Richness (MSR), détermine le nombre minimum d'espèces qui doivent être présents dans une faune pour permettre des comparaisons significatives avec les analyses multivariées. En utilisant MSR, plusieurs localités de Riversleigh ont été choisies pour une étude paléoécologique utilisant la méthode des cénogrammes permettant de déterminer les paléoenvironnements pendant l‘Oligo-Miocène. Enfin, la méthode des âges numériques a été utilisée pour raffiner l‘âge relatif des localités de Riversleigh et une réévaluation des Systems de Riversleigh est proposée. ______

NAME AND ADDRESS OF THE AUSTRALIAN LABORATORY:

School of Biological, Earth and Environmental Sciences, University of New South Wales, New South Wales 2052, Australia. ______

INTITULE ET ADRESSE DE L'U.F.R. OU DU LABORATOIRE FRANCAIS:

UMR 5125 PEPS, CNRS, France; Université Lyon 1, Campus de La Doua, Bt. Géode, 69622 Villeurbanne cedex, France ______

KEY WORDS

Palaeoecology, Biochronology, Riversleigh, Australia, Quercy, France, Cenogram, MSR, Numerical Ages

______

MOTS-CLES

Paléoécologie, Biochronologie, Riversleigh, Australie, Quercy, France, Cénogramme, MSR, Age numérique

______

iv Acknowledgements

I would like to thank the many people that provided assistance and support throughout my PhD thesis. First, I would like to thank my supervisors Professor Michael Archer and Dr Serge Legendre, my co-supervisor Dr Sue Hand and our UNSW lab manager Henk Godthelp. I would like to thank Michael Bedward of the New South Wales National Parks and Wildlife Service for writing the R routines essential to chapter 2, for his advice and for reading the manuscript. I also would like to thank Dr Gilles Escarguel for providing his help and advice for chapter 2 and 3. I want to thank the Palaeo lab in Sydney for their help and support, namely Pip Brewer, Dr Mina Bassarova, Dr Karen Roberts, Dr Yamila Gurovich, Jackie Nguyen, Hayley Bates, Elizabeth Price, Dr Kirsten Crosby, Dr Steven Wroe, Dr Vera Weisbecker, Dr Karen Black, Dr Anna Gillespie, and especially Dr Julien Louys, Robin Beck, and Dr Rick Arena whose help has significantly improved my thesis. I want to thank the students at the University of Lyon, namely Jeremy Martin, Dr Benjamin Greselle, Vincent Fernandez, Dr Marie-Anne Heran, Dr Caroline Sassier, Guillaume Suan, and all other students that I met and gave me their support. I thank Prof. Bill Sherwin, Dr Peter Banks and Dr Alistair Poore for their advice. I thank the South Australian Department for Environment and Heritage (particularly Andrew Graham), Queensland National Parks and Services (particularly Noleen Kunst), Victoria‘s Department of Sustainability and Environment for providing the National Park data needed for this study. Thanks to Dr Karen Roberts, Dr Karen Black, Hayley Bates, Dr Anna Gillespie, Dr Anne Musser and Pip Brewer for providing dental measurements of the taxa they are studying. I would like to thank Dr Bernard Cooke, Scott Hocknull and Dr Gilbert Price for their help during my visit to the Queensland Museum. Thanks to Dr Ian Graham and Dr Dioni Cendon for helping find references. Finally, I would like to thank my family for their support and especially my girlfriend Kara Giles, who has supported me throughout this thesis.

v Table of contents

CHAPTER 1: INTRODUCTION 1

1.1 RIVERSLEIGH WORLD HERITAGE AREA: BIOCHRONOLOGY AND PALAEOECOLOGY 2 1.2 AIMS 4 1.3 CHAPTER OUTLINE 5 1.4 REFERENCES 8 CHAPTER 2: FINDING THE MINIMUM SAMPLE RICHNESS (MSR) FOR MULTIVARIATE ANALYSES: IMPLICATIONS FOR PALAEOECOLOGY. 11

2.1 ABSTRACT 12 2.2 INTRODUCTION 12 2.3 METHODS AND MATERIALS 14 2.3.1 Parent Lists 14 2.3.2 Minimum Sample Richness (MSR) 15 2.3.3 Analyses 16 2.4 RESULTS 16 2.5 DISCUSSION 23 2.6 REFERENCES 26 CHAPTER 3: THE USE OF MSR (MINIMUM SAMPLE RICHNESS) FOR SAMPLE ASSEMBLAGE COMPARISONS IN CONJUNCTION WITH A TAXONOMIC DISTINCTNESS MEASURE 29

3.1 ABSTRACT 30 3.2 INTRODUCTION 30 3.3 MATERIAL AND METHODS 34 3.3.1 Data sets 34 3.3.2 Cluster analysis 36 3.3.3 MSR calculation 38 3.3.4 Taxonomic distinctness 39 3.4 RESULTS & DISCUSSION 42 3.4.1 Preliminary cluster analysis 42 3.4.2 Case study 1: Quercy and Limagne Area 43 3.4.3 Case study 2: Riversleigh 58 3.5 CONCLUSIONS 65 3.6 REFERENCES 66 3.7 APPENDIX 70 CHAPTER 4: EXPLAINING THE GAPS IN MAMMALIAN BODY MASS DISTRIBUTIONS (CENOGRAMS) AND THE ECOLOGICAL IMPACT OF INTRODUCED SPECIES IN AUSTRALIA 81

4.1 ABSTRACT 82 4.2 INTRODUCTION 83 4.3 MATERIALS AND METHODS 87 4.3.1 Mammal species database 87 4.3.2 Environmental data 89 4.3.3 Cenograms 92 4.3.4 Arboreal taxa 94 4.3.5 Body mass distribution 94 4.3.6 Analysis 95 4.4 RESULTS 95 4.5 DISCUSSION 114 4.5.1 Limitations 114 4.5.2 Australia‘s shifted gap 118 4.5.3 Explaining the gap 119 4.5.4 Revision of the cenogram method 122

vi 4.5.5 The impact of introduced predators 124 4.5.6 Mid-domain effect 126 4.6 CONCLUSION 126 4.7 REFERENCES 127 CHAPTER 5: PALAEOECOLOGICAL ANALYSES OF RIVERSLEIGH‘S OLIGO-MIOCENE SITES: IMPLICATIONS FOR OLIO-MIOCENE CLIMATE CHANGE IN AUSTRALIA 134

5.1 ABSTRACT 135 5.2 INTRODUCTION 135 5.3 MATERIALS AND METHODS 139 5.3.1 Materials 139 5.3.2 Body mass estimate 141 5.3.3 Cenogram and Body Mass Distribution methods 141 5.3.4 Discriminant Function Analysis (DFA) of Body Mass Distribution data 144 5.4 RESULTS 145 5.4.1 Faunal Zone A Cenograms and Body Mass Distributions 145 5.4.2 Faunal Zone B Cenograms and Body Mass Distributions 146 5.4.3 Faunal Zone C Cenograms and Body Mass Distributions 149 5.4.4 Encore Site Cenogram and Body Mass Distribution 151 5.4.5 Combined Faunal Zone Cenograms and Body Mass Distributions 152 5.4.6 Discriminant function analysis 154 5.5 DISCUSSION 157 5.5.1 Limitations 157 5.5.2 Comments on the use of the revised cenogram method with fossil localities 159 5.5.3 Arboreal species 162 5.5.3 Riversleigh and the climate of northern Australia during the Oligo-Miocene. 163 5.6 CONCLUSION 168 5.7 REFERENCES 169 5.8 APPENDIX 176 CHAPTER 6: ESTIMATION OF RELATIVE AGES OF RIVERSLEIGH LOCAL FAUNAS USING THE NUMERICAL AGES METHOD 182

6.1 ABSTRACT 183 6.2 INTRODUCTION 183 6.3 MATERIALS AND METHODS 189 6.3.1 Materials 189 6.3.2 Numerical Ages method 189 6.4 RESULTS 192 6.5 DISCUSSION 195 6.5.1 Comparison of the estimated relative age sequence with stratigraphy in previous work and implications for biostratigraphic interpretation 195 6.5.2 Comparison with the karst stratigraphic sequence of Arena (2004) 197 6.5.3 Applicability of the Neohelos and Wakaleo lineages for biocorrelation between Riversleigh and other Australian Oligo-Miocene localities. 204 6.6 CONCLUSION 210 6.7 REFERENCES 211 CHAPTER 7: CONCLUSIONS 215

7.1 REFERENCES 220 APPENDIX A 221 SPECIES LIST AND BODY MASS ESTIMATES OF MODERN NATIONAL PARKS AND RESERVES 221 APPENDIX B 242 DENTAL MEASUREMENTS OF UNPUBLISHED RIVERSLEIGH SPECIMEN USED IN THE NUMERICAL AGE METHOD (CHAPTER 6) 242

vii APPENDIX C 243 RAW MSR DATA FOR THE QUERCY AND LIMAGNE AREA AND RIVERSLEIGH WORLD HERITAGE AREA. 243 APPENDIX D 244 RAW NUMERICAL AGES DATA 244 APPENDIX E 245 PUBLICATIONS 245

viii List of figures

Figure 1.1. Map of Australia showing the location of Riversleigh. 3

Figure 2.1. Example of cluster analysis with grouped subsets. 15

Figure 2.2. Plot of minimum sample richness against species richness. 18

Figure 2.3. Plot of minimum sample richness against similarity. 18

Figure 2.4. Plot of minimum sample richness against number of lists. 19

Figure 2.5. 3D plot of minimum sample richness, species richness and similarity 20 between parent lists using Dice's, Jaccard's and Simpson's similarity indices.

Figure 2.6. 3D plot of minimum sample richness, species richness and similarity 20 between parent lists using Raup-Crick's similarity indices.

Figure 2.7. Contour graph of the MSR surface using Dice/Jaccard/Simpson‘s 22 indices.

Figure 2.8. Contour graph of the MSR surface using Raup-Crick‘s index. 22

Figure 3.1. Preliminary neighbour-joining cluster analysis. 45

Figure 3.2. Neighbour-joining clusters for the Quercy sites, using all sites. 47

Figure 3.3. Neighbour-joining clusters for the Quercy sites, using two sites 48 per biochronological unit.

Figure 3.4. Plot of the average taxonomic distinctness and variation 49 in taxonomic distinctness for the Quercy sites from MP16 to MP20.

Figure 3.5. Plot of the average taxonomic distinctness and variation 50 in taxonomic distinctness for the Quercy sites from MP21 to MN02a.

Figure 3.6. Neighbour-joining clusters for the Riversleigh sites. 60

Figure 3.7. Plot of the average taxonomic distinctness and variation 62 in taxonomic distinctness for the Riversleigh sites.

Figure 4.1 Cenogram patterns. 84

Figure 4.2. Map of Australian National Parks and Reserves. 89

Figure 4.3A-F. Cenograms of the mammalian faunas of National Parks. 97-102

Figure 4.3G. Modern and historical cenograms of the mammalian faunas 103 of four National Parks.

ix Figure 4.4. Plot of the largest gap magnitude versus its relative 104 position on the cenogram.

Figure 4.5. Box plot of the number of mammal species in the size gaps. 106

Figure 4.6. Bar graphs of the number of non-arboreal and arboreal 109 mammal species.

Figure 4.7. Bar graphs of the proportion of mammal species in each 110 of the body mass categories of the National Parks.

Figure 4.8. Bar graphs of the proportion of mammal species in each of 110 the body mass categories of Australia and New Guinea.

Figure 4.9. Bar graphs of the proportion of mammal species in each of 112 the body mass categories of four main patterns identified for Rainforest, Temperate forest, riparian woodlands and grasslands and desert.

Figure 4.10. Bar graphs of the proportion of mammal species of 113 historical data versus modern data.

Figure 4.11. New cenogram pattern model. 124

Figure 5.1. New cenogram pattern model. 142

Figure 5.2. Body mass distribution patterns in Australian habitats. 143

Figure 5.3. Cenograms and Body Mass Distribution graphs of Faunal Zone A. 146

Figure 5.4. Cenograms and Body Mass Distribution graphs of Faunal Zone B 147 sites part 1.

Figure 5.5. Cenograms and Body Mass Distribution graphs of Faunal Zone B 148 sites part 2.

Figure 5.6. Cenograms and Body Mass Distribution graphs of Faunal Zone C 150 sites part 1.

Figure 5.7. Cenograms and Body Mass Distribution graphs of Faunal Zone C 151 sites part 2.

Figure 5.8. Cenogram and Body Mass Distribution graph of Encore Site. 152

Figure 5.9. Cenograms and Body Mass Distribution graphs of the combined 153 Faunal Zone A, B and C.

Figure 5.10. Plot of the canonical discriminant function analysis. 156

Figure 6.1. Summary of Riversleigh karst stratigraphy. 186

x Figure 6.2. Seriation of taxa presence within the Riversleigh sites scaled in time 196 using the estimated ages of the sites.

Figure 6.3. Graph of the relative age estimate (in Ma) of the Riversleigh 201 local faunas in Numerical Ages sequence.

xi List of tables

Table 3.1. List of the Quercy and Limagne Area fossil-bearing localities. 35

Table 3.2. List of Riversleigh fossil-bearing localities. 37

Table 3.3. Results of the MSR estimates for the Quercy and Limagne area. 44

Table 3.4. Results of the MSR estimates for Riversleigh. 59

Table 4.1. List of major vegetation groups and habitat types for 91 each National park.

Table 4.2. Kendall‘s R coefficient between cenogram, body mass distribution 114 and arboreal taxa variables and environmental variables.

Table 5.1. Summary of Discriminant Function Analysis results. 155

Table 5.2. Results of the habitat classification of the modern and fossil localities. 156

Table 5.3 Predicted classification of the fossil localities. 157

Table 5.4. Riversleigh species lists, specimen numbers and body mass 176-181 estimates.

Table 6.1. Numerical ages estimated for the Riversleigh localities. 194

Table 6.2. Comparison of the relative age of the Riversleigh sites given by 199 the Numerical Ages method with the Depositional phase model.

Table 6.3. Biocorrelation between Riversleigh faunas and other Australian 207 Oligo-Miocene localities based on shared taxa including members of Wakaleo and Neohelos lineages.

xii

CHAPTER 1: INTRODUCTION

1 1.1 RIVERSLEIGH WORLD HERITAGE AREA: BIOCHRONOLOGY AND

PALAEOECOLOGY

The Riversleigh fossil deposits were inscribed on UNESCO‘s World Heritage

List of geological sites in 1994 (in a serial nomination with Naracoorte Caves, south- eastern South Australia) in recognition of their importance for understanding the evolution of mammals in Australia. Located near Riversleigh Station Homestead

(19°02‘S, 138°45‘E) on the Gregory River (Fig. 1.1), approximately 200 km north-west of Mount Isa, in north-western Queensland, the Riversleigh World Heritage Area contains more than 200 fossil localities, and has been the subject of research efforts for over 40 years (Archer et al., 1989; Archer et al., 1994a; Archer et al., 1997). The geology of the Riversleigh area is complex, with Tertiary freshwater carbonate deposits occurring adjacent to, within and upon Proterozoic siliclastics, Cambrian marine limestone and chert deposits that form the north-eastern edge of the Barkly Tableland

(Archer et al., 1989; Megirian, 1992; Creaser, 1997; Arena, 2004; Archer et al., 2006).

The Riversleigh Tertiary carbonate deposits appear to have been deposited via a complex sequence of fluvial and lacustrine tufa deposits and karst deposits (Archer et al., 1989; Megirian, 1992; Creaser, 1997; Arena, 2004; Archer et al., 2006).

Riversleigh‘s fossil deposits preserve a very diverse fauna, including invertebrates

(arthropods and molluscs), fish, amphibians, reptiles, birds and mammals, as well as some plants in rare localities (Archer et al., 1989; Archer et al., 1997; Arena, 1997;

Archer et al., 2006). Several studies have focused on the palaeoecology of Riversleigh fossil deposits, but interpretation of the palaeohabitats that are represented remains controversial (Archer et al., 1989; Megirian, 1992; Archer et al., 1997; Myers, 2002;

Megirian et al., 2004; Bassarova, 2005).

2

Figure 1.1. Map of Australia showing the location of Riversleigh World Heritage Area.

Most of Riversleigh‘s fossil localities range from late Oligocene to late Miocene in ages, with a few Pliocene, Pleistocene and Holocene localities also known (Archer et al., 1989; Archer et al., 1994a; Archer et al., 1997; Archer et al., 2006). Radiometric dating is underway for some of the fossil localities (Archer et al., 1997; Archer et al.,

2006; Graham et al., 2006), but the principal method of dating for the Riversleigh deposits until now has been relative dating, using stage of evolution interpretations, biocorrelation with other Australian and international fossil localities, and stratigraphic and faunal analyses (Woodburne et al., 1985; Archer et al., 1989; Woodburne et al.,

1993; Archer et al., 1997; Black, 1997; Cooke, 1997; Creaser, 1997; Megirian et al.,

2004; Travouillon et al., 2006). Travouillon et al. (2006) identified one major problem with the current methodologies employed in biochronology and palaeoecology œ namely that palaeontological data are incomplete (Hammer and Harper, 2006) and there are

3 currently few methods to identify whether fossil localities are representative of the original community.

1.2 AIMS

The Riversleigh fossil localities span the broadest temporal range of any fossiliferous formation in the Oligo-Miocene of Australia (Archer et al., 1997; Megirian et al., 2004). This broad temporal range includes a crucial period of Australian climate change between the warm and humid climatic conditions of the early and middle

Miocene to the cool and dry conditions that began in the late Miocene (Frakes et al.,

1987; Archer et al., 1992; Archer et al., 1994; Hill, 1994; McGowran and Li, 1994;

Archer et al., 1995; Archer et al., 1997; Martin, 1998, 2006). For these reasons, understanding of Riversleigh biochronology and palaeoecology is crucial to the study of broader Australian faunal evolution, to establish sound biochronological zonations (e.g. land mammal ages), and to study climatic change through time in Australia.

The first aim of this thesis is to develop a new technique to account for a major problem of fossil data œ incompleteness. This new method, ”Minimum Sample

Richness‘, is developed to help determine whether a fossil locality is representative of the original fauna that lived in the immediate vicinity at the time the fossils were deposited. A representative fossil locality is one that preserves sufficient faunal data to allow accurate and precise palaeoecological and biochronological studies. The new technique is first developed theoretically and then tested on fossil data. Only those

Riversleigh localities identified as representative of the original fauna are used in subsequent palaeoecological and biochronological studies.

4 The second aim of this thesis is to identify habitat types for the representative fossil localities and examine whether these habitat types can be correlated with changes in climatic conditions during the Oligo-Miocene. Before investigating this aim, the method used in this thesis for determining habitat type, the ”Cenogram‘ method

(Legendre, 1986, 1989), is reviewed to determine whether the methodology in its current form is suitable for analysing Australian mammal faunas. Modern mammalian communities from across Australia are used to test this method, previously criticised because it cannot be interpreted within a quantitative statistical framework (Rodríguez,

1999).

The third and final aim of this thesis is to refine the relative age of the

Riversleigh deposits (in the absence of absolute dates), to clarify the biochronological sequence of fossil localities across Australia and to clarify Riversleigh‘s position within this biochronological sequence. The ”Numerical Ages‘ method (Legendre and Bachelet,

1993) is used to estimate the relative ages of Riversleigh‘s fossil deposits, and is compared with previously used methods of relative dating of Australian fossil assemblages (Woodburne et al., 1985; Archer et al., 1989; Woodburne et al., 1993;

Archer et al., 1997; Megirian et al., 2004; Travouillon et al., 2006).

1.3 CHAPTER OUTLINE

Chapter 2: Finding the Minimum Sample Richness (MSR) for multivariate analyses: implications for palaeoecology

Chapter 2 outlines the development and theoretical aspects of the new palaeoecological method Minimum Sample Richness (MSR), developed here to identify

5 fossil assemblages that are representative of the corresponding original faunal communities. A simulation approach was used to investigate the behaviour of commonly used similarity indices (Dice, Jaccard, Simpson and Raup-Crick indices), and the reliability of classifications derived from these indices, when working with incomplete fossil data. Equations are presented that enable the calculation of MSR values for the different similarity indices necessary to achieve 95% classification accuracy from estimates of absolute species richness and beta diversity. The development of this new method is a first step toward more reliable and meaningful biochronological and palaeoecological studies. This chapter has been published in the journal Historical Biology (Travouillon et al., 2007).

Chapter 3: The use of MSR (Minimum Sample Richness) for sample assemblage comparisons in conjunction with a taxonomic distinctness measure

Chapter 3 describes the first application of the MSR methodology to fossil data and identifies one of its many potential usages. This new method is first tested on fossil data from French localities in the Palaeogene and early Neogene of the Quercy and

Limagne area (Massif Central, south-western France), where biochronological relationships are well understood, enabling the strengths and weaknesses of the method to be identified. Another method, ”Taxonomic Distinctness‘ analysis (Warwick and

Clarke, 1995), is used in conjunction with MSR to minimise taxonomic biases. The combination of the two methods is then used to analyse the Riversleigh assemblages to identify those that contain sufficient taxa to represent original communities and that are not taxonomically biased, and hence which can confidently be used in palaeoecological analyses.

6 Chapter 4: Explaining the gaps in mammalian body mass distributions

(cenograms) and the ecological impact of introduced species in Australia

Chapter 4 reviews the uses of the cenogram method for palaeoecological studies and provide the first use of this method for marsupial faunas. Australian national parks and reserves are used to identify patterns in Australian body mass distributions and to revise the cenogram method for Australian faunas. Correlations between environmental variables and cenogram/body mass distribution variables are investigated. Modern ecological issues are also addressed, such as the impact of the ”dingo fence‘, habitat fragmentation and the island effect on the body mass distributions of Australian mammalian faunas.

Chapter 5: Palaeoecological Analyses of Riversleigh‘s Oligo-Miocene Sites:

Implications for Oligo-Miocene climate change in Australia

Chapter 5 presents the results of the application of the revised cenogram/body mass distribution method to Riversleigh fossil faunas. Only fossil assemblages that were identified by the MSR/Taxonomic distinctness analyses as unbiased were included.

Probable habitat types are identified for each of these fossil localities, using cenograms, body mass distribution graphs and Discriminant Function Analysis.

Chapter 6: Estimation of relative ages of Riversleigh local faunas using he

Numerical Ages method

Chapter 6 presents the results of the Numerical Ages method as applied to the

Riversleigh fossil deposits. Age estimates for some Riversleigh deposits are presented and compared to previous geological and biostratigraphical studies Representative index taxa and lineages are identified for the Riversleigh Faunal Zones.

7

Chapter 7: Conclusions

Chapter 7 presents the final conclusions of the research presented in this thesis, as well as implications and suggestions for future research.

1.4 REFERENCES Archer, M., D. A. Arena, M. Bassarova, R. M. D. Beck, K. Black, W. E. Boles, P. Brewer, B. N. Cooke, K. Crosby, A. Gillespie, H. Godthelp, S. J. Hand, B. P. Kear, J. Louys, A. Morrell, J. Muirhead, K. K. Roberts, J. D. Scanlon, K. J. Travouillon and S. Wroe (2006). Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1: 1-17.

Archer, M., H. Godthelp, S. J. Hand and D. Megirian (1989). Fossil mammals of Riversleigh, northwestern Queensland: preliminary overview of biostratigraphy, correlation and environmental change. Australian Zoologist 25: 29-65.

Archer, M., S. J. Hand and H. Godhelp (1992). Back to the future: the contribution of palaeontology to the conservation of Australian forest faunas. Conservation of Australia's forest fauna. D.Lunney. Sydney, Royal Zoological Society for New South Wales: 67-80.

Archer, M., S. J. Hand and H. Godhelp (1994). Patterns in the history of Australia's mammals and inferences about palaeohabitats. . History of the Australian Vegetation. R. Hill. Cambridge, Cambridge University Press: 80-103.

Archer, M., S. J. Hand and H. Godhelp (1995). Tertiary environmental and biotic change in Australia. Paleoclimate and evolution, with emphasis on human origins. G. H. D. E. Vrba, T.C. Partridge, L.H. Burckle. New Haven, Yale University Press: 77-90.

Archer, M., S. J. Hand and H. Godthelp (1994a). Riversleigh, Sydney, 256 pp.

Archer, M., S. J. Hand, H. Godthelp and P. Creaser (1997). Correlation of the Cainozoic sediments of the Riversleigh World Heritage fossil property, Queensland, Australia. . Actes du congrès BiochroM'97, Mémoires et Travaux de l'Ecole Pratique des Hautes Etudes. J.-P. Aguilar, Legendre, S., Michaux, J. Institut de Montpellier. 21: 131-152.

Arena, D. A. (1997). Palaeontology and geology of Dunsinane Site, Riversleigh. Memoirs of the Queensland Museum 41(2): 171-179.

Arena, D. A. (2004). The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle Tertiary. PhD thesis, UNSW, Sydney.

8 Bassarova, M. (2005). Taphonomic and palaeoecological investigations of Riversleigh Oligo-Miocene fossil sites. PhD, UNSW, Sydney.

Black, K. (1997). Diversity and biostratigraphy of the Diprotodontoidea of Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 187-192.

Cooke, B. N. (1997). Biostratigraphic implications of fossil kangaroos at Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 295-302.

Creaser, P. (1997). Oligocene-Miocene Sediments of Riversleigh: The potential significance of topography. Memoirs of the Queensland Museum 41(2): 303-314.

Frakes, L. A., B. MCGowran and J. Bowler (1987). Evolution of Australian environments. Fauna of Australia. G. R. Dyne and D. W. Walton. Canberra, Australian Government Publishing Service. 1A, General articles: 1-16.

Graham, I., E. Price, D. Cendón and J. Woodhead (2006). Understanding Riversleigh's geology: what we know in 2006, and where to next. Riversleigh 2006 Symposium. University of New South Wales, Sydney, Australia, The Riversleigh Society Inc.

Hammer, O. and D. A. T. Harper (2006). Paleontological Data Analysis. Blackwell Publishing, Carlton, 351.

Hill, R. S., Ed. (1994). History of the Australian vegetation: Cretaceous to recent. Cambridge University Press, Cambridge.

Legendre, S. (1986). Analysis of mammalian communities from the late Eocene and Oligocene of Southern France. Palaeovertebrata 16: 191-212.

Legendre, S. (1989). Les communautés de mammifères du Paléogène (Eocène supérieur et Oligocène) d'Europe occidentale: structures, milieux et évolution. Münchner Geowissenschaftliche Abhandlungen (A) 16: 1-110.

Legendre, S. and B. Bachelet (1993). The numerical ages: a new method of datation applied to Paleogene mammalian localities from Southern France. Newsletters on Stratigraphy 29: 137-158.

Martin, H. A. (1998). Tertiary climatic evolution and the development of aridity in Australia. Proceedings of the Linnean Society of New South Wales 119: 1115-1136.

Martin, H. A. (2006). Cenozoic climatic change and the development of arid vegetation in Australia. Journal of Arid Environments 66: 533-563.

McGowran, B. and Q. Li (1994). The Miocene oscillation in southern Australia. Records of the South Australian Museum 27: 197-212.

Megirian, D. (1992). Interpretation of the Carl Creek Limestone, northwestern Queensland. The Beagle: Records of the Northern Territory Museum of Arts and Sciences 9(1): 219-248.

9 Megirian, D., P. Murray, L. Schwartz and C. Von Der Borch (2004). Late Oligocene Kangaroo Well Local Fauna from the Ulta Limestone (new name), and climate of the Miocene oscillation across central Australia. Australian Journal of Earth Sciences 51: 701-741.

Myers, T. J. M. (2002). Palaeoecology of Oligo-Miocene Local Faunas from Riversleigh. University of New South Wales, Sydney.

Rodríguez, J. (1999). Use of cenograms in mammalian palaeocology. A critical review. Lethaia 32: 331-347.

Travouillon, K. J., M. Archer, S. J. Hand and H. Godthelp (2006). Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1: 323-349.

Travouillon, K. J., M. Archer, S. Legendre and S. J. Hand (2007). Finding the Minimum Sample Richness (MSR) for multivariate analyses: implications for palaeoecology. Historical Biology.

Warwick, R. M. and K. R. Clarke (1995). New —biodiversity“ measures reveal a decrease in taxonomic distinctness with increasing stress. Marine Ecology Progress Series 129: 301-305.

Woodburne, M. O., B. J. MacFadden, J. A. Case, M. S. Springer, N. S. Pledge, J. D. Power, J. M. Woodburne and K. B. Springer (1993). Land mammal biostratigraphy and magnetostratigraphy of the Etadunna Formation (Late Oligocene) of South Australia. Journal of Vertebrate Paleontology 13: 483-515.

Woodburne, M. O., R. H. Tedford, M. Archer, W. D. Turnbull, M. D. Plane and E. L. Lundelius (1985). Biochronology of the continental mammal record of Australia and New Guinea. Stratigraphy, Palaeontology, Malacology Papers in Honour of Dr Nell Ludbrook. L. J. M., South Australian Department of Mines and Energy Special Publication. 5: 347œ363.

10

CHAPTER 2: FINDING THE MINIMUM SAMPLE RICHNESS (MSR) FOR MULTIVARIATE ANALYSES: IMPLICATIONS FOR PALAEOECOLOGY.

This chapter has been published:

Travouillon, K. J., Archer, M., Legendre, S. and Hand, S. J. (2007). Finding the Minimum Sample Richness (MSR) for multivariate analyses: implications for palaeoecology. Historical Biology, 19: 315 œ 320.

11 2.1 ABSTRACT

Many techniques have been developed to estimate species richness and beta diversity. Those techniques, dependent on sampling, require abundance or presence/absence data. Palaeontological data is by nature incomplete, and presence/absence data is often the only type of data that can be used to provide an estimate of ancient biodiversity. We used a simulation approach to investigate the behaviour of commonly used similarity indices, and the reliability of classifications derived from these indices, when working with incomplete data. We drew samples, of varying number and richness, from artificial species lists, which represented original life assemblages, and calculated error rates for classifications of the parent lists and samples. Using these results, we estimated the minimum sample richness (MSR) needed to achieve 95% classification accuracy. Results were compared for classifications derived from several commonly used similarity indexes (Dice, Jaccard,

Simpson and Raup-Crick). MSR was similar for the Dice, Jaccard and Simpson indices.

MSR for the Raup-Crick index was often much lower, suggesting that it is preferable for classifying patchy data, however the performance of this index was less stable than the other three in the simulations, which required an even lower MSR. MSR can be found for all presence/absence data from the contour graphs and equations as long as the absolute species richness and the beta diversity can be estimated.

2.2 INTRODUCTION

Dealing with incomplete data is a particular challenge in palaeontological studies. Travouillon et al. (2006) used several multivariate analyses to compare presence/absence data of the Riversleigh Local Faunas (Oligo-Miocene assemblages

12 from north-west Queensland, Australia.), to determine which fossil localities were members of the same assemblage. They noticed that many localities had very low species richness and concluded that they were probably not representative of the original life assemblage. This misrepresentation of the life assemblage resulted in the unexpected grouping of local faunas which are geologically different. Mares and Willig

(1994) investigated this issue of representativeness using recent mammal faunas. They randomly selected species from a known fauna and continued to increase sample size until the correct community was identified. Using this method, they answered two questions: (1) 'How large a sample of species must be drawn from the fauna in order to arrive at a correct decision as to the fauna's community association? '

(2) 'What percentage of the species from a fauna is required in a sample in order to yield a correct determination of the community from which those species were drawn?'

In this study we use a simulation approach to investigate these questions.

Multivariate analyses such as cluster analysis or ordination, which are often used to compare fossil localities, are not statistical tests. They are data reduction methods used to visualise relationships between objects and attributes in complex data. The results found by such analyses may be highly skewed by missing data caused by undersampling or taphonomic processes. It is for this reason that we attempt to identify the relationship between classification accuracy and minimum sample richness. The minimum sample richness (MSR) is defined here as the smallest number of taxa that must be present in a sample to achieve a given level of classification accuracy. We compare four similarity indices to see if different indices generated different MSR values. We used the following indices: Dice (also known as the Sørensen index),

13 Jaccard, Simpson and Raup-Crick (Dice, 1945; Sørensen, 1948; Jaccard, 1912;

Simpson, 1943; Raup and Crick, 1979).

2.3 METHODS AND MATERIALS

2.3.1 Parent Lists

Alternative sets of artificial parent lists of taxa, representing different original life assemblages, were generated for a range of values of richness (A) and similarity between lists or Beta diversity (S). Richness (A) is the number of taxa in a parent list.

The following A values were used; 25, 50, 100, 150, 200, 250, 300 and 350. The number of parent lists used in the analysis (N) varies between analyses. All values of N between 2 and 16 (inclusive) were investigated and the following S values were used; 0,

10, 30, 50, 70, 90, 95 and 100. For example, with an S value of 50 and N value of 3, the first parent list shares 50% of its taxa with the second and the third parent lists. The second parent list shares only S/2 (25%) with the third parent list. For N greater than 3, any subsequent parent lists would continue to be represented by decreasing values of S, i.e. they would share 12%, followed by 0% of the taxa with the first parent list. In each analysis, only one parameter was investigated at a time, so that the two not being investigated were represented by a standard value (A=100, S=50, N=3).

14

Figure 2.1. Example of cluster analysis with 1) correctly grouped subsets; and 2) incorrectly grouped subsets. Numbers 1, 2 and 3 represent the three parent faunas. S1, S2 and S3 are the corresponding subsets to the original faunas.

2.3.2 Minimum Sample Richness (MSR)

The Minimum Sample Richness (MSR) is the smallest number of taxa in a sample needed for that sample to be a reliable indicator of its original life assemblage.

We simulated presence-absence data for sites in a palaeontological field study by drawing replicate subsets from the parent lists. The subsets and parent lists were then compared using cluster analysis to see if they group together correctly (Fig. 2.1). This comparison is replicated 1000 times. The error rate (proportion of replicates in which subsets were grouped with the wrong parent list) was then calculated for a range of subset sizes. If the error rate was greater than 0.05, then the procedure was repeated increasing the size of the subset by one species until an error rate of 0.05 or less was achieved. Data generation and clustering were performed with custom scripts in the R statistics environment (Gentleman and Ihaka, 2005).

15 2.3.3 Analyses

Four analyses were performed in order to investigate how changing the two parameters and the number of parent lists used (A, S and N) affected the MSR. In the first analysis, the value of A was investigated. The MSR for each value of A was recorded using four different similarity indices (Dice's, Jaccard's, Simpson's and Raup-

Crick similarity indices), using the same dataset for each of the indices. The use of the four similarity indices follows Hammer and Harper (2006). We also compared the results using one subset per parent list at a time and using several subsets per parent list at a time (increasing from 6 to 12 subsets with increasing A). In the second and third analyses, the values of S and N were investigated, respectively. For these analyses, we recorded the results using one subset per parent list and using each of the four similarity indices. In the fourth analysis, every possible combination of A and S values used in the first two analyses were investigated and the corresponding MSR values calculated.

2.4 RESULTS

The results of the first analysis are shown in Figure 2.2. For each value of A, there was no difference in MSR using the Dice, Jaccard or Simpson similarity indices.

Using only one subset per parent list, there was a linear relationship between MSR and

A (MSR = 0.34A). Using several subsets drawn from each parent list, the relationship between MSR and A was more complex. For A values between 0 and 75, MSR was higher than for the analyses with a single subset per parent list, but the reverse was true for A values greater than 75. The Raup-Crick index generally achieved much lower

MSW values than the other three indices. For analyses with several subsets drawn from each parent list, the curve resembles that for the other indices but with a substantially lower MSR.

16

In the second analysis, we varied the similarity between parent lists and only one subset was used per list. The results of this analysis are shown in Figure 2.3.

Classifications derived from the Raup-Crick index correctly identified the parent lists with a much smaller MSR than required with the other indices. For, Dice, Jaccard and

Simpson indices, MSR increased steadily with increasing S. When S=50, MSR = 34. In contrast, the MSR using the Raup-Crick index remained quite low when S was less than

50. When S was greater than 50, MSR rapidly increased but remained lower than for the other indices.

The third analysis aimed at testing whether the number of parent lists (N) used in the analysis influenced the MSR. The results are shown in Figure 2.4. Increasing N from 2 to 15 did not affect the results for classifications derived from the Dice, Jaccard and Simpson indices and MSR remained stable at 34 taxa. In contrast, the Raup-Crick measure displayed a sudden jump in MSR at N=11.

17

Figure 2.2. Plot of minimum sample richness (MSR) against absolute species richness (A) where N=3 and S=50, using Dice's, Jaccard's, Simpson's and Raup-Crick‘s similarity indices.

Figure 2.3. Plot of minimum sample richness (MSR) against similarity between parent lists (S) where N=3 and A=100 using Dice‘s, Jaccard‘s, Simpson‘s and Raup-Crick‘s similarity indices.

18

Figure 2.4. Plot of minimum sample richness (MSR) against number of lists (N) where S=50 and A=100 using Dice's, Jaccard's, Simpson's and Raup-Crick‘s similarity indices.

19

Figure 2.5. 3D plot of minimum sample richness (MSR), absolute species richness (A) and similarity between parent lists (S) using Dice's, Jaccard's and Simpson's similarity indices.

Figure 2.6. 3D plot of minimum sample richness (MSR), absolute species richness (A) and similarity between parent lists (S) where N=3 using Raup-Crick's similarity indices.

20 Figure 2.5 shows the joint effects of richness (A) and similarity between parent lists (S) on MSR for the Dice, Jaccard and Simpson indices, while Figure 2.6 shows the corresponding results for the Raup-Crick index. MSR increases with increasing A and S regardless of the similarity index used but Raup-Crick achieved a lower MSR, especially when S was less than 50.

Contour graphs (Fig. 2.7 and 2.8) are obtained by Delaunay triangulation of the

MSR Surface and projection onto the A-S plane. Least-squares 3D-surface fitting is calculated for the Dice/Jaccard/Simpson indices as the combination of an A-MSR linear relation and a S-MSR second order polynomial relation. In order to optimize the correlation between simulated and fitted MSR estimates, the fitted estimates are rounded to the upper integer (ceil-function) when S≤50% and to the lower integer

(floor-function) when S>50%. The resulting prediction functions are:

MSR 107ceil −5 A S 2 +××⋅= 103 −3 if SSA ≤××⋅ %50 {[( ) ] [( ) ]} MSR {[()107floor −5 SA 2 ] [()103 −3 ]} if SSA >××⋅+××⋅= %50

21

Figure 2.7. Contour graph of the MSR surface using Dice/Jaccard/Simpson‘s indices.

Figure 2.8. Contour graph of the MSR surface using Raup-Crick‘s index.

22 The resulting determination coefficient (R2=0.999) indicates that the fitted surface perfectly matches the observed one, thus enabling the use of these two complementary functions to estimate MSR-value for any [S, A]-values. Due to the very nature of the Raup-Crick‘s index (a type-I error rate of a significance test; see below), its MSR-surface was not modelled in the same way; its very high concavity not corresponding to any simple combination of linear, polynomial, exponential or power function. Nevertheless, a rough prediction of a MSR- value for any [S, A]-values can be graphically obtained from the contour graph (Fig. 2.8).

2.5 DISCUSSION

Palaeontological data is by nature incomplete (Hammer and Harper, 2006).

However, an appreciation of how this ”incompleteness‘ can affect analyses and interpretations can be achieved through simulation studies such as that presented here.

Mares and Willig (1994) investigated the minimum sample size which would correctly group a sample to its original faunal list. However, application of their method to other types of data is difficult. We developed a method to provide palaeoecologists with a way to check whether their fossil samples are statistically representative of the original life assemblage.

Very early in the analysis, we came across a challenge: should we be using one subset or several subsets per parent list to determine the MSR? The first analysis we performed aimed at answering this question. We noticed that when we used Dice's,

Jaccard's or Simpson's similarity indices, using one subset required a higher MSR than using several subsets. In addition, we found that using one subset gave a linear relationship whilst several subsets did not. We found that using several subsets gave

23 unstable results (varying with the number of subsets included in the clustering) because the subsets were creating links between each other and the parent lists, lowering the chances of missgrouping the subsets and the parent lists. The values for MSR can be predicted using one subset and the result show that increasing the number of subsets can only increase the chance of a correct grouping. For this reason, we chose to use only one subset per parent list for the remaining analyses.

The results of the analyses showed that the Raup-Crick similarity index behaved very differently to the Dice, Jaccard and Simpson indices in that it performed worse with increasing numbers of subsets or parent lists. The Raup-Crick index is a probabilistic measure which assesses the pattern of species between two samples in terms of a 'random sprinkling of species' hypothesis (Legendre and Legendre, 1998).

The original formulation of the index (Raup and Crick, 1979) used a randomisation procedure to estimate the probability of the observed data, although by recognising that the calculation is equivalent to a 'sampling without replacement' problem an exact value can easily be derived from the hypergeometric distribution (M. Bedward, Pers. Comm.,

2006).The probabilistic nature of the Raup-Crick index is probably the reason why it performed less well than the other indices, as the chance of correctly grouping similar lists decreases with increasing numbers of similar lists. Travouillon et al. (2006) noticed in their cluster analysis that Raup-Crick's index clustered together local faunas that had no species in common. The relationship between the same local faunas was unresolved using the Dice, Jaccard and Simpson indices. Raup-Crick's index can, therefore, be potentially misleading. The index can also be unstable depending on the data and may lead to error.

24 Nonetheless, we recommend use of this index, as it does perform better than the other indices tested. Raup-Crick's index performs much better than the other indices test here when data are sparse. Its probabilistic nature is most likely the reason why it performs better than distance measures in this case. The fewer major groups (i.e. environments or time periods, which are the equivalent to the parent lists in the analysis) and the fewer faunal/floral samples being compared (Equivalent to the subsets in the analysis), the better.

The downside to the method described here to calculate the MSR, is that it requires the absolute species richness and the percentage similarity (Beta diversity) of the faunas being compared to be known. However, there are a number of ways to acquire estimates of absolute species richness. Estimating absolute species richness can be done using accumulation curves or extrapolating from species abundance distribution or from non-parametric estimators (Colwell and Coddington, 1994; Chazdon et al.,

1998; Magurran, 2004). Walther and Martin (2001) reviewed many of these estimators and considered that the two Chao estimators (Chao, 1984; Chao, 1987; Shen et al.,

2003) were the least biased, followed by the two jackknife estimators (Burnham and

Overton, 1979). Chao, Shen and Hwang (2006) applied Laplace‘s boundary-mode approximations (See Erkanli, 1994; Erkanli, 1997) to the Chao estimators to improve the accuracy of their estimations of species richness and allow their use with replicated incidence data (presence/absence data). For palaeontological data, for which abundance data is often unavailable, non-parametric estimators (Chao2 estimator) are the only tools that can be used, as they are the only estimators that are able to use presence/absence data. Colwell (2000) released a computer program called EstimateS which not only provides the tools to estimate the absolute species richness but also provides tools to

25 estimate the similarity between two or more sites (Beta diversity). Again, most of the techniques used to estimate the beta diversity use abundance data (ICE and ACE, Chao et al., 2000; Chao's Abundance-based Jaccard and Sørensen indexes, Chao et al., 2005) but beta diversity can be estimated for presence/absence data using classic diversity indices such as Jaccard, Sørensen (same as Dice) or Bray-Curtis (Magurran, 2004) or using Chao, Shen and Hwang (2006)‘s estimators of beta diversity. Using estimates for absolute species richness and similarity will necessarily only provide an estimate of

MSR. MSR is therefore as accurate as the estimates of absolute species richness and similarity. However, even an estimate of MSR will provide palaeoecologists with a useful tool to check whether their data is representative of the original life assemblage.

Future work will aim at applying the method to palaeontological data and testing the strengths and weaknesses of the method.

2.6 REFERENCES

Burnham, K.P., and Overton, W.S., 1979. Robust estimation of population size when capture probabilities vary among animals. Ecology, 60: 927-936.

Chao, A., 1984. Nonparametric estimation of the number of classes in a population. Scandinavian Journal of Statistical Theory and Applications, 11: 265-270.

Chao, A., 1987. Estimating the population size for capture-recapture data with unequal catchability. Biometrics, 43: 783-791.

Chao, A., Shen, T.-J., and Hwang, W.-H., 1987. Application of Laplace‘s boundary- mode approximations to estimate species and shared species richness. Australian and New Zealand journal of statistics, 48: 117-128.

Chao, A., Hwang, W.-H., Chen, Y.-C., and Kuo, C.-Y., 2000. Estimating the number of shared species in two communities. Statistica Sinica, 10: 227-246.

Chao, A., Chazdon, R.L., Colwell, R.K., and Shen, T.-J, 2005. A new statistical approach for assessing compositional similarity based on incidence and abundance data. Ecology Letters, 8: 148-159.

Chazdon, R.L., Colwell, R.K., Denslow, J.S., and Guariguata, M. R., 1998. In: F. Dallmeier and J.A. Comiskey (Ed.) Forest biodiversity research, monitoring and

26 modelling: conceptual background and old world case studies (Paris: Parthenon Publishing).

Colwell, R.K., and Coddington, J.A., 1994. Estimating terrestrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society of London B, 345: 101- 118.

Colwell, R.K., 2000. EstimateS œ statistical estimation of species richness and shared species from samples Version 7.5. Available online at: http://viceroy.eeb.uconn.edu/EstimateS (accessed 15 July 2006).

Dice, L.R, 1945. Measures of the amount of ecological association between species. Ecology, 26: 297-302.

Erkanli, A., 1994. Laplace approximations for posterior expectations when the mode occurs at the boundary of the parameter space. Journal of the American statistical association, 89: 250-258. Erkanli, A., 1997. Boundary-mode approximations for posterior expectations. Journal of statistical planning and inference, 58: 217-239.

Gentleman, R., and Ihaka, R., 2005. The R Project for statistical computing. Available online at: http://www.r-project.org/ (accessed 10 November 2005).

Hammer, O., and Harper, D., 2006. Paleontological data analysis (UK: Blackwell Publishing).

Jaccard, P., 1912. The distribution of the flora of the alpine zone. New Phytologist, 11: 37-50.

Legendre, P. and L. Legendre. 1998. Numerical ecology (Amsterdam: Elsevier).

Magurran, A. E., 2004. Measuring biological diversity (UK: Blackwell Publishing).

Mares, M.A., and Willig, M.R., 1994. Inferring biome associations of recent mammals from samples of temperate and tropical faunas: Paleoecological considerations. Historical Biology, 8: 31-48.

Raup, D.M., and Crick, R.E., 1979. Measurements of faunal similarities in Paleontology. Journal of Paleontology, 53: 1213-1227.

Shen, T.-J., Chao, A., and Lin, J.-F., 2003. Predicting the number of new species in further taxonomic sampling. Ecology, 84: 798-804.

Simpson, E.H., 1949. Measurement of diversity. Nature, 163: 688.

Sørensen, T., 1948. A method of establishing groups of equal amplitude in plant sociology based on similarity of species content. Det Kongelige Danske Videnskab. Selskab, Biologiske Skrifter, 5: 1-34.

27 Travouillon, K.J., Archer, M., Hand, S. J., and Godthelp, H., 2006. Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special issue 1: 323-249.

Walther, B.A., and Martin, J.-L., 2001. Species richness estimation of bird communities: how to control for sampling effort? Ibis, 143: 413-419.

28

CHAPTER 3: THE USE OF MSR (MINIMUM SAMPLE RICHNESS) FOR SAMPLE ASSEMBLAGE COMPARISONS IN CONJUNCTION WITH A TAXONOMIC DISTINCTNESS MEASURE

29 3.1 ABSTRACT

Minimum Sample Richness (MSR) is defined as the smallest number of taxa that must be recorded in a sample to achieve a given level of inter-assemblage classification accuracy. MSR is calculated from known or estimated richness and taxonomical similarity. In this chapter, MSR is tested for strengths and weaknesses using 167 published mammalian local faunas from the Palaeogene and early Neogene of the Quercy and Limagne area (Massif Central, south-western France), and then applied to 84 fossil faunas from Riversleigh, north-western Queensland in Australia. In many cases, MSR is able to detect the assemblages in the data set that are potentially too incomplete to be used in a comparative, taxonomical similarity-based analysis. In some cases, MSR alone is not able to account for potential biases resulting in the incorrect clustering of some faunas. In those cases, the average taxonomic distinctness and its variation are shown to provide useful complementary information. This analysis reveals that lower than expected average taxonomic distinctness and higher than expected variation in taxonomic distinctness may lead to incorrect clustering of faunas. It is concluded that when used together in the context of comparative analysis between taxonomical assemblages, these methods can screen sample assemblages that are not representative of their underlying original living communities. Ultimately, they can be used to identify which assemblages require further sampling before being included in a comparative analysis.

3.2 INTRODUCTION

Taxonomic comparisons of sample assemblages are generally used to distinguish different communities in time, space and habitat types. In the specific

30 context of palaeobiological studies, the incomplete nature of the fossil record is a major problem in those comparisons, potentially resulting in incorrect interpretation of patterns of similarities. Travouillon et al. (2006) identified such problems in their comparison of the Riversleigh Local Faunas (Oligo-Miocene assemblages from north- west Queensland, Australia.), corresponding to assemblages with mammal species richness varying from 1 to 43 taxa. There are a number of methods to estimate the completeness of a sample assemblage (e.g. species accumulation curves, non-parametric species richness estimators, completeness index…). However, at present there is no methodology available to estimate the minimum number of sampled taxa required to achieve an accurate representation of its original life assemblage in a comparative analytical scheme. Ultimately, there is no need for a complete or near complete data set

(which is often impossible to achieve from palaeontological data) if the available data contains enough information to confidently reflect —true“ taxonomical similarities between assemblages due to environmental and/or chronological identities.

The Minimum Sample Richness (MSR) has been developed for such purpose

(Travouillon et al., 2007). MSR is defined as the average minimum number of taxa that must be identified in a sample assemblage to achieve a given level of classification accuracy using taxonomical similarity-based multivariate analyses. MSR-distributions are known for four classical indices based on presence/absence data: Dice, Jaccard,

Simpson and Raup-Crick. Whilst Travouillon et al. (2007) introduced and discussed the potential of MSR, the aim of this chapter is to further explore the applicability and power of this new tool.

31 MSR is calculated from estimates of absolute species richness and similarity for each compared community (or life assemblage). Once absolute species richness and similarity are estimated for each community and couple of communities, then MSR is calculated and can be used to identify which sample assemblages are species-rich enough to give an accurate representation of their community. This is one strength of the MSR method œ to eliminate from a comparative analysis all sample assemblages possibly disrupting any meaningful similarity pattern.

A key assumption of the MSR method that could potentially introduce error into the results is that each taxon is treated equally. From a sampling point of view, it is well known that taxa are not equal, with some being abundant and others rare. Communities are often differentiated by their rare taxa, rather than their abundant taxa, which are usually found in most communities. Sample assemblages that have more taxa than a calculated MSR, but are taxonomically biased, may therefore give incorrect classification results.

Based on incidence (i.e., presence/absence) data, taxonomical composition biases in sample assemblages can be identified using two complementary diversity measures: the average taxonomic distinctness (AvTD or b+ Warwick and Clarke, 1995;

Clarke and Warwick, 1998; Warwick and Clarke, 1998; Clarke and Warwick, 1999) and the variation in taxonomic distinctness (VarTD or d+ Clarke and Warwick, 2001). On one hand, under controlled sampling settings, these two indices have been shown to be ecologically meaningful: AvTD can detect differences in taxonomic composition caused by trophic diversity changes (e.g., due to habitat pollution; Warwick and Clarke, 1998), whilst VarTD negatively correlates with habitat diversity (Clarke and Warwick, 2001).

32 On the other hand, they both show highly robust statistical sampling properties, including a lack of dependence, in mean value, on sample size, sampling effort and taxonomic identification skills of different workers, which make them particularly attractive in the general context of uncontrolled sampling settings (Clarke and Warwick,

1998, 2001). Thus, because both peculiar local ecological conditions and non-random sample making are susceptible to generating sample assemblages with distorted taxonomical spectra, these two indices are a priori useful to detect taxonomically corrupted sample assemblages possibly disrupting meaningful similarity patterns.

However, neither of these indices has yet been used to detect taxonomical biases in fossil assemblages.

Here, the aim is to explore both the strengths and the weaknesses of the MSR method, used in conjunction with taxonomic distinctness analysis. For this purpose, two distinct mammalian case studies are successively considered focusing on chronological clustering, although this approach could also be used for other purposes (e.g. environmental or biogeographical clustering). First, this two-fold methodology is applied to a data set of 167 sample assemblages from the Palaeogene and early Neogene of the Quercy and Limagne area (Massif Central, south-western France). This data set is well suited to test this methodology because: (i) the biochronological framework underlying this data set is fairly well established (Remy et al., 1987, BiochroM‘97)

(Legendre and Bachelet, 1993; Escarguel et al., 1997), (ii) the mammalian evolutionary history is well resolved at the local and regional levels (Legendre, 1986, 1989;

Escarguel & Legendre, 2006), and (iii) the species richness of the studied assemblages range widely from 1 to 51 taxa. Second, the method is applied to a data set of 84 sample assemblages from the Riversleigh fossil assemblages (north-western Queensland,

33 Australia; Archer et al., 2006; Travouillon et al., 2006). In this biochronologically less constrained and evolutionarily less resolved case, MSR and taxonomic distinctness analyses are used in order to detect and eliminate species-poor and/or taxonomically biased sample assemblages.

3.3 MATERIAL AND METHODS

3.3.1 Data sets

The first case study is based on the data collected for more than 20 years by the third author for Cainozoic mammal communities of the Quercy and Limagne area

(Legendre, 1986, 1987, 1987b, 1987c, 1987d, 1989; Legendre et al., 1991; Legendre and Hartenberger, 1992; Legendre and Bachelet, 1993; Escarguel et al., 1997; Legendre et al., 1997; Legendre and Girard, 1999; Massif Central, south-western France, see

Escarguel and Legendre, 2006; Legendre et al., 2006). The incidence data set is made of the observed occurrence of 350 mammalian (excluding bats) phyletic lineages within

167 local faunas. Each local fauna is assigned to one of the 17 successive biochronological units identified for the studied time interval, including 15 Paleogene reference levels (MP16 to MP30, excluding MP27) and 2 Neogene zones (MN01 and

MN02a, see Legendre and Bachelet, 1993; Aguilar et al., 1997; Escarguel et al., 1997).

For the purpose of this study and in order to be able to test the methodology, it is assumed that each local fauna is correctly assigned to its proper biochronological unit.

Names and biochronological units of each local fauna are listed in Table 3.1.

34 Table 3.1. List of the Quercy and Limagne Area fossil-bearing localities, ordered by biochronological unit (MP16 to MN02a), with Site I.D. used in the analysis and sampled species richness (S.R.).

35 For the second case study, published mammal occurrence lists are used for the

Riversleigh species assemblages (see Archer et al., 2006), following Travouillon et al.

(2006) with a few modifications (see section 3.7 for modifications). Of the 75 sites studied in Travouillon et al. (2006), all are used except the four youngest: Rackham‘s

Roost Site (Pliocene), Terrace Site (Pleistocene), Carrington‘s Cave and Message Stick

Cave (Recent). Thirteen new mammal-bearing sites are then included, namely: Micro

Site, Baker‘s Delight Site, Angela‘s Sinkhole Site, Crusty Meat Pie Site, Group Site,

Golden Steph Site, Panorama Site, Phil‘s Phenomenal Fissure Fill Site, Jeanette‘s

Birthday Site, Black Coffee 2 Site, Neville‘s Pancake Site, Steph‘s Small Reward Site and Rick‘s Sausage Site. Species occurrences in these sites are listed in section 3.7, including corrections, additions and updates to the original lists given by Archer et al.

(2006). Site names are listed in Table 3.2. Biochronological nomenclature follows

Arena (2004) and Travouillon et al. (2006).

3.3.2 Cluster analysis

A preliminary clustering was performed on both case study data sets to determine the quality of the clustering before applying the MSR methodology. The

Neighbor-joining method (Saitou and Nei, 1987) as implemented in the PAST software

(Hammer et al., 2001) was used to cluster assemblages using presence/absence of mammalian taxa. The Neighbor-joining method is more suitable for chronological data because it minimises the total sum of branch lengths (producing a phenogram) rather than producing an ultrametric tree (or dendrogram) as per the more usual hierarchical techniques (e.g. Single and Complete linkage, UPGMA and UPGMC) in which branch lengths are constrained so that the distance from the root to each tip is the same

(Escarguel, 2005).

36 Table 3.2. List of Riversleigh fossil-bearing localities, ordered by biochronological unit (Faunal Zone A, B, C and D), with Site I.D. used in the analysis and sampled species richness (S.R.).

37 Bootstrap values were also calculated for each node of the phenogram, from

1000 replicates.

3.3.3 MSR calculation

For each biochronological unit of each analyzed data set, MSR was calculated using Travouillon et al. (2007)‘s equation:

ceil −5 2 −3 if S≤ 5.0 MSR floor {[( 107 ) SA ] [( 103 )××⋅+××⋅= SA ]}if S > 5.0 , where A is the estimated absolute total species richness and S is the largest similarity value between the sample assemblages recorded in the considered biochronological unit, estimated using the Dice index of similarity (Dice, 1945). A was computed using the Chao-2 non-parametric estimator of total species richness for incidence data (Chao et al., 2006), as available in the SPADE software (Chao and Shen, 2003).

For each biochronological unit, two distinct estimation strategies of A and S were adopted; in both cases the upper and lower bounds of the 95% confidence interval associated to the Chao-2 mean estimate (A) were also calculated. Because estimation of the Chao-2 index requires a minimum of two sample assemblages, a first estimate of A and S was obtained by only considering the two local faunas with the highest sampled species richness. A second estimate was obtained by taking into account all the available local faunas correlated to the biochronological unit. Given that the computation of the Chao-2 estimator is only based on the number of unique and duplicate species (i.e., species found in only one or two sites, respectively), estimates of

A based on the two richest sample assemblages are expected to be lower than that based on all available assemblages. Nevertheless, using all local faunas may not necessarily give a better estimate due to an overestimation of the unique/duplicate ratio, and thus of

38 A as the consequence of the general incompleteness of the fossil record. However, the same problem also applies when considering only two sample assemblages if the species lists have very few species in common as the consequence of —false“ absences.

For each biochronological unit and computation strategy, MSR was then calculated from A and S estimated values. For each estimation strategy, three MSR- estimates were actually calculated: a —mean MSR“, an —upper MSR“ and a —lower

MSR“, using the mean Chao-2 estimate (A) and the upper and lower limits of its associated 95% confidence interval, respectively. Once calculated for each biochronological unit, these MSR values were finally used to construct reduced data sets including only those sample assemblages with species richness greater than or equal to their associated MSR. These reduced data sets were then analyzed using the two-step procedure (similarity matrix computation and additive cluster analysis) described in the previous section. A total of 9 cluster analyses were made: 3 using 2 sites per biochronological unit and 3 using all sites for the Quercy and Limagne data set, and 3 using 2 sites per biochronological unit for the Riversleigh data set.

3.3.4 Taxonomic distinctness

In order to test the sample assemblages for taxonomical biases, the data sets were subjected to a taxonomic distinctness analysis. This method tests the average degree to which species in an assemblage are taxonomically related to each other

(average taxonomic distinctness, b+) and the evenness of the spread of taxa across the taxonomic spectrum (variation in taxonomic distinctness, d+) (Clarke and Warwick,

2001). The computation of these two indices first relies on the construction of a Linnean taxonomical tree as a reasonable proxy of the underlying phylogenetic history (see

39 below for details). In both case studies, a partially resolved taxonomical tree made of 8

Linnean levels was used: species phyletic lineage, genus, family, superfamily, infraorder, suborder, order, and supercohort. The taxonomical path lengths within the tree were defined following a simple linear weighting scheme, with a path length of 1 when contrasting two species lineages from the same genus, 2 for two species lineages from distinct genera, but the same family, and so on up to a path length of 8 when contrasting two species lineages from distinct supercohorts.

For a given sample assemblage made of N distinct taxa, AvTD is the average taxonomical path length measured for each of the N þ (N-1)/2 possible pairs of taxa in the taxonomic tree. Thus, AvTD can be considered as a taxonomic disparity index reflecting the level of phylogenetic heterogeneity of the assemblage: the higher the

AvTD, the more the assemblage is made of several different phylogenetic groups. For instance, a taxonomical assemblage of 10 species from the same order has the same species richness, but a lower taxonomic disparity than an assemblage with 10 species from 10 distinct orders. Thus, unless considering highly specialised, taxonomically impoverished assemblages corresponding to rather rare and atypical environmental conditions (see Warwick & Clarke 2001), sample assemblages characterised by low

AvTD-values are likely to poorly represent their underlying life assemblages.

VarTD is the variance associated with AvTD. A low VarTD-value indicates that the N taxa of the assemblage tend to be —taxonomically equidistant“, whereas a high

VarTD-value illustrates a heterogeneous distribution of the pairwise taxonomical distances. Complementary to AvTD, VarTD can be viewed as a confidence index of the randomness of the sample assemblage when compared to the underlying life

40 assemblage: a low VarTD-value, all the more when associated with a low AvTD-value, is likely to indicate that the studied taxonomic assemblage is not a random sample of its parent life assemblage. Regardless of the origin of the bias, this sample thus carries peculiar taxonomical information and must be used with caution in the similarity analysis.

For both indices, taxonomically biased sample assemblages are readily detected using a standard Monte-Carlo procedure of random resampling of taxa without replacement. This procedure allows the estimation of the f+ and f+ confidence funnels associated with the null hypothesis that a given sample assemblage shows a taxonomical spectrum randomly drawn from the synthetic master list of all taxa identified in the analysed data set (see Clarke and Warwick, 1998, 2001 for methodological details). In this study, the TDA.pro software is used, written in IDL language and referred to in

Escarguel and Legendre (2006), to produce b+ and d+ graphs with their 95% and 99% probability funnels estimated from 1000 random replicates.

For the Quercy and Limagne data set, the taxonomical classification used is based on recent molecular phylogenies that have robustly resolved relationships among extant mammals (e.g. Madsen et al., 2001; Murphy et al., 2001a; Murphy et al., 2001b;

Waddell and Shelley, 2003; Springer et al., 2004; Beck et al., 2006; Nishihara et al.,

2006). However, these molecular studies strongly conflict with morphology-based classifications (e.g. McKenna and Bell, 1997; Kielan-Jaworowska et al., 2004; Rose,

2006), making the affinities of many fossil taxa (for which molecular data are unavailable) uncertain. As a result, Leptictida, Creodonta, Apatotheria and

Nyctitheriidae have been considered as Eutheria incertae sedis œ none of them can be

41 convincingly placed in any of the currently recognised molecular-based superordinal clades. The confidence funnels associated with AvTD and VarTD were estimated for two distinct data subsets: MP16-MP20 and MP21-MN02. Indeed, a major faunal event occurred at the end of the Eocene, the Grande Coupure de Stehlin, changing considerably the taxonomic structure of the Quercy and Limagne mammalian fauna

(Escarguel and Legendre, 2006; Legendre et al., 2006). Separating the data subsets avoids the taxonomic distinctness analysis from finding biases caused mainly by this major change in fauna.

For the Riversleigh data set, the taxonomic classification used follows Archer et al. (2006) for the family, genus and species levels, and Aplin and Archer (1987) for the higher taxonomical levels (supercohort to superfamily).

3.4 RESULTS & DISCUSSION

3.4.1 Preliminary cluster analysis

A preliminary clustering was performed on both case study data sets to determine the quality of the clustering before applying the MSR methodology. The preliminary Neighbor-joining trees are shown in Figure 3.1 (A- Quercy and Limagne and B- Riversleigh). Both preliminary clusters show that the data set is suitable for testing the MSR methodology: in them chronologically similar sites cluster away from each other, and vice-versa, showing that most sites have too few taxa to correctly cluster with their own group.

42 3.4.2 Case study 1: Quercy and Limagne Area

Species richness, similarity and MSR values were calculated and are shown in

Table 3.3. Table 3.3A shows the results calculated by using all sites while Table 3.3B shows the results of using only two sites. Although one might expect the use of all sites to give better estimates than using two sites, the methodology would be circular if all sites were used to calculate an MSR for all sites. Hence this technique focuses on selecting two sites from the same biochronological unit (MP/MN), which are then used to calculate a MSR value, and determining the minimum number of species required to represent the biochronological unit.

Overall, the Chao2 estimates of species richness are much larger using all sites

(Table 3.3A) compared to using two sites (Table 3.3B), except for MP20 and MP30.

Using all sites may also give an overestimate of the species richness due to high number of single occurrences contributed by lesser sampled sites. The opposite can be expected from using two sites, having less single occurrences (more shared occurrences) resulting in an underestimate of species richness. Hence one would expect that the lower Chao2 would be the more accurate estimate for all sites, while the upper Chao2 would be the more accurate estimate for two sites. In terms of similarity, in both cases, all or two sites, the similarity values found using the Dice similarity index gave fairly similar results, ranging from about 30% similarity to 70% similarity. The resulting MSR values calculated from the Chao2 estimates and similarity values are much higher using all sites than two sites. These values were then utilised to select sites from the sites list (see

Table 3.1) and proceed to cluster.

43 Table 3.3. Results of the absolute species richness estimates (using the Chao-2 non- parametric estimator), similarity estimates (using Dice similarity index) and MSR estimates for each biochronological unit identified in the Quercy and Limagne area (see text for computational details).

44

Figure 3.1. Preliminary neighbour-joining cluster analysis for: A) Quercy and Limagne area sites, and B) Riversleigh sites. Non-parametric bootstrap supports are shown at each node. Correspondence between site codes and names is given in Table 3.1 for the Quercy and Limagne data set, and in Table 3.2 for the Riversleigh data set.

45 Sites with species richness above the MSR value were selected. Sites selected using the —all sites“ upper MSR are shown in Figure 3.2A, —all sites“ MSR in Fig. 3.2B and —all sites“ lower MSR in Fig. 3.2C. Only five sites had a species richness higher than the upper MSR (Fig. 3.2A), two from MP16, 1 from MP19 and 2 from MP28. The five sites cluster as expected and the bootstrap values are very robust. In Figure 3.2B, 19 sites had a species richness higher than the MSR value. Again, sites cluster as expected and bootstrap values are high. Using the lower MSR (Fig. 3.2C), 30 sites were selected.

Here, the first few unexpected clusters are identified, such as MP19_11 (Rosières 2) clustering with MP20_4 (Tabarly) instead of MP19_4 (Escamps), MP28_1 (Cournon) clustering with MP30_2 (Coderet C3) instead of MP28_3 (pech Desse) and MP28_4

(Pech du Fraysse), and MN02a_1 (Chavroches) clustering with MN01_6 (Peublanc) instead of MN02a_5 (Montaigu). Note that in each of these cases, these sites are clustering with the only representative site of a biochronological unit (e.g. MP20_4

Tabarly is the only representative of MP20 in this cluster). Possible reasons for such clustering are discussed below.

46

Figure 3.2. Neighbour-joining clusters for the Quercy and Limagne area sites, using all sites to estimate species richness and similarity for the calculation of A) the upper MSR, B) MSR and C) lower MSR. Bootstrap values are shown at each node.

47

Figure 3.3. Neighbour-joining clusters for the Quercy and Limagne area sites, using two sites per biochronological unit to estimate species richness and similarity for the calculation of A) the upper MSR, B) MSR and C) lower MSR. Bootstrap values are shown at each node.

48 Figure 3.3 shows the clusters based on the —2 sites“ upper MSR (Fig. 3.3A),

MSR (Fig. 3.3B) and lower MSR (Fig. 3.3C). The upper MSR, MSR and lower MSR selected 25, 48 and 54 sites respectively. The —2 sites“ upper MSR cluster (Fig. 3.3A) contains only two unexpected clusters, with MP28_1 (Cournon) distant from other

MP28 sites and MN02a_1 (Chavroches) clustering with MN01_6 (Saulcet) instead of

MN02a_5 (Montaigu). The biochronological units have very strong bootstrap values

(above 50% and as high as 100%), showing the robustness of these units. Nodes connecting these units are however much weaker, but the cluster is not expected to correctly order each unit chronologically. Seriation would be the method to use to do so.

The —2 sites“ MSR cluster (Fig. 3.3B) and lower MSR cluster (Fig. 3.3C) are very similar, having the same unexpected clusters for MP18 to MP24 and MP28 to MP30, and having much lower bootstrap values. The Upper MSR would therefore be a better value to use because it minimises the number of incorrect grouping due to low species richness.

Figure 3.4. Plot of the average taxonomic distinctness (left) and variation in taxonomic distinctness (right) with 99% (thin doted lines) and 95% (thick doted lines) probability funnels, for the Quercy and Limagne sites from MP16 to MP20.

49

Figure 3.5. Plot of the average taxonomic distinctness (left) and variation in taxonomic distinctness (right) with 99% (thin doted lines) and 95% (thick doted lines) probability funnels, for the Quercy and Limagne sites from MP21 to MN02a.

Unexpected clusters may not be all caused by an underestimation of the MSR value. The taxonomical distinctness method was applied to this data to identify any potential taxonomical biases that may influence the clustering of the sites. The average taxonomic distinctness, b+ and variation in taxonomic distinctness, d+ are shown in

Figure 3.4 for MP16 to MP20, and Figure 3.5 for MP21 to MN02a. The following sites fell outside the probability funnels in Figures 3.4 and 3.5:

99% probability funnel for average taxonomic distinctness, b+:

MP16_2 (Castrais), MP17a_9 (St-Antonin), MP17b_3 (Pech d'Isabeau), MP17b_4

(Pépénut), MP20_1 (Coyrou 1-2), MP24_3 (Coânac-Château), MP24_5 (Genebrières

1), MP24_11 (Vialenc), MP25_6 (Gari), MP25_8 (La Garrigue), MP25_9 (L'Escoufle),

MP25_20 (Romagnat-MP25), MP25_23 (St-Yvoine), MN02a_3 (Langy), MN02a_6

(Poncenat).

50 95% probability funnel for average taxonomic distinctness, b+:

MP17a_4 (La Cantine 2), MP17b_6 (Rosières 5), MP18_4 (Gousnat), MP19_6 (Le

Puy), MP22_8 (Guirolle), MP23_6 (Mège), MP23_13 (Roqueprune 3), MP24_1 (La

Benissons-Dieu), MP24_6 (Itzac), MP24_7 (Lebratières 14), MP25_3 (Belgarric 1),

MP25_16 (Piatzé), MP25_18 (La Plante 3), MP25_21 (Rousselou), MN01_3 (Gannat sup.), MN02a_2 (Carrière Cluzel).

99% probability funnel for variation in taxonomic distinctness, d+:

MP17a_9 (St-Antonin), MP18_4 (Gosnat), MP19_4 (Escamps), MP19_9 (Palembert),

MP20_1 (Coyrou 1-2), MP22_1 (Baraval), MP22_5 (Coulou), MP22_7 (Gardiol 3),

MP23_6 (Mège), MP24_7 (Lebratières 14), MP24_11 (Vialenc), MP25_3 (Belgarric 1),

MP25_6 (Gari), MP25_9 (L'Escoufle), MP25_11 (Mas de Gaston), MP25_16 (Piatzé),

MP25_17 (Pipet), MP25_24 (La Sauvetat), MP26_2 (Espeyrasse), MP26_4 (Mas de

Pauffié), MP28_6 (Portal).

95% probability funnel for variation in taxonomic distinctness, d+:

MP17a_4 (La Cantine 2), MP17b_3 (Pech d'Isabeau), MP18_5 (Mas de Labat 1),

MP19_14 (Sindou D), MP21_8 (Mas de Labat 2), MP22_10 (La Plante 2), MP23_8

(Pech Crabit 1), MP23_10 (Raynal), MP23_13 (Roqueprune 3), MP23_14 (Roqueprune

4), MP24_3 (Coânac-Château), MP25_8 (La Garrigue), MP25_19 (Rigal Jouet 1),

MP26_1 (La devèze), MP26_3 (Genebrières 2), MP29_9 (Verneuil), MP30_1 (Coderet

C1), MP30_2 (Coderet C3), MP30_3 (Créchy 1-2, MP30_7 Thézels), MN01_6

(Sualcet), MN02a_2 (Carrière Cluzel), MN02a_3 (Langy).

51 Of all sites falling outsides the probability funnels, twenty are of interest, being selected by the MSR method. These are: MP18_4 (Gousnat), MP19_4 (Escamps),

MP19_14 (Sindou D), MP22_1 (Baraval), MP22_10 (La Plante 2), MP23_8 (Pech

Crabit 1), MP24_1 (La Benissons-Dieu), MP24_3 (Coânac-Château), MP24_5

(Genebrières 1), MP24_6 (Itzac), MP24_7 (Lebratières 14), MP24_11 (Vialenc),

MP25_3 (Belgarric), MP25_8 (La Garrigue), MP25_19 (Rigal Jouet 1), MP26_4 (Mas de Pauffié), MP29_9 (Verneuil), MP30_2 (Coderet C3), MN01_6 (Saulcet) and

MN02a_3 (Langy). Instead of removing them from the analysis and assuming that they do influence the clustering, the resulting clusters are examined in detail (Figure 3.3) and the cause of unexpected clusters within the biochronological units visually identified.

In Figure 3.3, MP16, MP17a and MP17b, found in all three clusters, grouped as expected, have high bootstrap values supporting the clusters and none of the sites used is taxonomically biased (average taxonomic distinctness and variation are within expected range). However, MP18, represented by two sites, MP18_4 (Gousnat) and

MP18_9 (Ste-Néboule) in the MSR and lower MSR clusters, does not cluster as expected. MP18_9 (Ste-Néboule) occurs at the base of the MP19 cluster and has a high bootstrap value (92 and 89 in MSR and lower MSR clusters, respectively). In contrast,

MP18_4 (Gousnat) occurs at the base of the MP17a and MP17b clusters, with a very low bootstrap value (33 and 36 in MSR and lower MSR clusters, respectively). MP18_4

(Gousnat) has both a lower than expected average taxonomical distinctness and higher than expected variation in taxonomic distinctness. Here one would therefore conclude that MP18_9 (Ste-Néboule) is correctly clustered close to MP19 while MP18_4

(Gousnat) is incorrectly clustered close to MP17b because it is taxonomically biased.

52 In the case of MP19, two of the four selected sites have higher than expected variation in taxonomical distinctness: MP19_4 (Escamps) and MP19_14 (Sindou D).

This does not seem to affect at all the upper MSR cluster, which has the highest bootstrap value (100) but in the MSR and lower MSR clusters, it clearly affects the cluster. In the MSR cluster, MP19_14 (Sindou D) and MP19_4 (Escamps) are clustered as expected with the other two MP19 sites, but the bootstrap values are relatively low

(23 and 51 respectively). In the lower MSR cluster, the only selected MP20 site

(MP20_4 (Tabarly)) nests within the MP19 cluster, separating MP19_14 (Sindou D) and MP19_4 (Escamps) from the other two, and the bootstrap values remain fairly low.

In this case, one would conclude that the unexpected clusters are caused by the high variation in taxonomical distinctness.

MP21 is represented by three sites, two of which MP21_1 (Aubrelong 1) and

MP21_10 (Ravet) always cluster together and have associated high bootstrap values in all three clusters. MP21_11 (Ronzon), however, always clusters away from other MP21 sites. None of these MP21 sites showed any taxonomic bias, and therefore the only conclusion that can be made is that MP21_11 (Ronzon) has a species richness below the true MSR (assuming that it has been correctly assigned to MP21) that in this analysis was underestimated.

MP22 is also represented by three sites. MP22_9 (Mas de Got) and MP22_10

(La Plante 2) cluster together every time, although MP22_10 (La Plante 2) has a higher than expected variation in taxonomic distinctness, whose impact is demonstrated by the reduction in bootstrap values as more sites are introduced (93 in upper MSR, 76 in MSR and 75 in lower MSR clusters). MP22_1 (Baraval), selected only in the MSR and lower

53 MSR clusters, has also a higher than expected variation in taxonomical distinctness, but it never clusters with other MP22 sites. This particular example may show that a higher than expected variation in taxonomical distinctness has more chance of leading to incorrect clustering when species richness is lower.

This is also apparent in MP23, with MP23_8 (Pech Crabit 1) having a higher than expected variation in taxonomical distinctness but having a high species richness

(second highest species richness in MP23); it always clusters correctly with MP23_2

(Itardies) (highest species richness in MP23) and MP23_7 (Mounayne) (equal third highest species richness in MP23) but has a lower bootstrap value with increasing number of sites. Meanwhile, MP23_12 (Roqueprune 2) does not cluster with any MP23 sites but, because it has the same species richness as MP23_7 (Mounayne) (24 taxa) which does cluster with other MP23 sites, one would have to conclude here that the true

MSR is slightly higher than 24 taxa for MP23.

MP24 had the lowest MSR estimate of the study, having the lowest estimated species richness and the lowest similarity. All selected MP24 sites except one MP24_4

(Digoin) have a lower than expected average taxonomic distinctness, and some also have a higher than expected variation in taxonomic distinctness. Despite the overall low species richness (higher species richness being 10 taxa for MP24_7, Lebratières 14) of the MP24 sites, they generally cluster well together. In the upper MSR cluster, only

MP24_7 (Lebratières 14) is selected and, despite the low bootstrap value of its clustering with MP25 sites, its position is as expected. In the MSR cluster, the MP24 sites all cluster together with relatively low bootstrap values, expect for MP24_1 (La

Benissons-Dieu) which clusters with MP21_11 (Ronzon). The same problematic

54 clustering is found in the lower MSR cluster, but MP24_1 (La Benissons-Dieu) this time clusters with MP24_4 (Digoin) away from other MP24 sites. One would conclude here that the unexpected clusters are caused by underestimated MSR values and that the low bootstrap values are the result of taxonomic distinctness.

In the case of MP25, all four selected sites clustered as expected, despite the fact that two had lower than expected average taxonomic distinctness (MP25_3 (Belgarric 1) and MP25_8 (La Garrigue)) and three of them had high variation in taxonomic distinctness (MP25_3 (Belgarric 1), MP25_8 (La Garrigue) and MP25_19 (Rigal Jouet

1)). Bootstraps values decrease on addition of taxonomically biased sites in the MSR and lower MSR clusters.

MP26‘s two selected sites also cluster as expected but the bootstrap value is relatively low (under 60), probably due to the fact that MP26_4 (Mas de Pauffié) has a higher than expected variation in taxonomic distinctness.

MP28 is represented by 4 selected sites: MP28_1 (Cournon,), MP28_2

(Cournon-les-Soumeyroux), MP28_3 (Pech Desse) and MP28_4 (Pech du Fraysse). All four sites have an average and variation in taxonomic distinctness within expectation and their species richness is fairly high (between 23 and 48 taxa). However, in all three clusters, MP28_3 (Pech Desse) and MP28_4 (Pech du Fraysse) always cluster together, with high bootstrap values, while MP28_1 (Cournon) clusters away from these two sites in the upper MSR cluster, clusters with MP29_2 (Les Chaufours) and MP28_2

(Cournon les Souméroux) in the MSR cluster, and clusters with MP28_2 (Cournon les

Souméroux) in the lower MSR cluster. In this case, the unexpected clustering cannot be

55 explained by MSR or taxonomic distinctness. MP28_1 (Cournon) and MP28_2

(Cournon les Souméroux) are located in the Limagne area, while MP28_3 (Pech Desse) and MP28_4 (Pech du Fraysse) are located in the Quercy area. Although at the start of the case study it was assumed that all sites were correctly assigned to the correct biochronological unit, here it is questioned whether the Limagne MP28 and the Quercy

MP28 do represent the same unit in time, or perhaps they do represent the same unit in time but possibly a different habitat, hence the difference in faunal composition.

MP29 and MP30 are not represented by any sites in the upper MSR cluster. This is quite surprising for MP30 since it contains at least three sites with at least 20 taxa:

MP30_2 (Coderet C3), MP30_3 (Créchy 1-2) and MP30_7 (Thézels). However, only

MP30_2 (Coderet C3) is selected in the MSR and lower MSR cluster, and it was found to have a high variation in taxonomic distinctness (as was one of the MP29 sites,

MP29_9 (Verneuil)). MP29 sites have much lower species richness in general (none over 19 taxa). In the MSR cluster, all three MP29 sites are separated: one clustering with MP28_1 (Cournon) (MP29_2, Les Chaufours), one clustering with MP30_2

(Coderet C3) (MP29_3, Créchy-bas) and MP29_9 (Vernbeuil) on its own. In the lower

MSR cluster, MP29_2 (Les Chaufours), MP29_6 (Mines des Roys) and MP29_9

(Verneuil) cluster together on one hand and MP29_3 (Créchy-bas) clusters with

MP30_2 (Coderet C3) on the other. In both clusters, the groups are unexpected and have very low bootstrap values. In this case, it is difficult to identify which of MSR or taxonomic distinctness might cause the unexpected groups, and is possibly caused by both.

56 MN01 and MN02a are unusual because they only cluster unexpectedly in the upper MSR cluster. This may be due to the fact that MN01_6 (Saulcet) has a higher than expected variation in taxonomical distinctness and is also the only representative of the MN01 sites. This is also supported by the very low bootstrap value (47) between

MN01_6 (Saulcet) and MN02a_1 (Chavroches). Even in the MSR and lower MSR clusters, the bootstrap values are generally low.

In terms of methodology, the MSR values should be calculated using the minimum amounts of sites (two) to avoid circular reasoning. The results show that when two sites are used to calculate the MSR value, species richness may be underestimated, and therefore the MSR value may also be underestimated. For this reason, we recommend using the upper MSR rather than the MSR or lower MSR for clustering.

MSR and lower MSR can still be carefully used, in conjunction with taxonomic distinctness, to identify possible incorrect clustering caused by taxonomic biases.

However, as noted above, it can become very difficult to differentiate incorrect groups caused by taxonomic distinctness and by an underestimation of MSR. Average taxonomic distinctness and variation in taxonomic distinctness seem to have an effect on clustering in reducing the bootstrap values of the clusters and producing incorrect clusters if species richness of the site is relatively low. Single representatives of biochronological units (e.g. MN01_6, Saulcet) also may result in incorrect clustering because they lack a similar site with which to cluster. In the latter case, incorrect clustering is probably caused by the presence of many common species (common to more than one biochronological unit, as opposed to rare or biochronologically unique species). This is the weakness of the MSR and taxonomic distinctness methods, the assumption that all species are treated equally, when they are obviously not. Some

57 species survive longer than others; some species are more common than others. It is probably impossible to develop a statistical tool to account for common/rare taxa. The combined use of MSR and taxonomical distinctness is therefore the best method to date to minimise the chance of incorrectly clustering biochronological units.

3.4.3 Case study 2: Riversleigh

Riversleigh‘s biochronological units, Faunal Zones A, B, C and D, are quite different to Quercy and Limagne Area‘s MP and MN units in terms of time scale.

Riversleigh‘s Faunal Zones are much less precise and, because of the uncertain age of the included sites, therefore span a much longer period of time. To calculate the MSR values of each Faunal Zone, two sites were selected from each Faunal Zone with the highest species richness. These sites are FZA_D and FZA_WH for Faunal Zone A,

FZB_CS and FZB_U for Faunal Zone B and FZC_Gag and FZC_HH for Faunal Zone

C. The results are shown in Table 3.4. Chao2‘s estimates of species richness were particularly high for Faunal Zone A, despite the fact that the two sites used for the estimate have much lower species richness values than the sites used for Faunal Zone B and C. This unrealistic estimate is mainly caused by the large numbers of single occurrences in the two sites, which have quite different faunas perhaps representing two distinct time periods within Faunal Zone A. The estimated species richness of the other two Faunal Zones seems more realistic and is similar to species richness estimates for the Quercy and Limagne Area. In terms of similarity, Riversleigh‘s Faunal Zones are much more dissimilar from each other, with values ranging from 0.27 to 0.4, than the

MP and MN of the Quercy and Limagne Area, where values ranged from 0.3 to 0.7.

Overall, the MSR estimates are lower for Riversleigh‘s Faunal Zones than for the

Quercy and Limagne Area. Riversleigh sites with species richness higher than the

58 lower MSR, MSR and upper MSR values for each Faunal Zones were selected for cluster analysis. FZD_En, being the only Faunal Zone D site, was also selected in all three clusters.

Table 3.4. Results of the absolute species richness estimates (using the Chao-2 non- parametric estimator), similarity estimates (using Dice similarity index) and MSR estimates for each Riversleigh‘s biochronological unit see text for computational details).

The upper MSR, MSR and lower MSR clusters are shown in Figure 3.6A, 3.6B and 3.6C respectively. A total of 10 sites were selected using the upper MSR, 18 sites using MSR and 23 sites using the lower MSR. In contrast with the clusters produced for

Quercy and Limagne Area which were unrooted trees (no particular site was used as an out-group), the clusters presented in Figure 3.6 are rooted, using FZD_En as an out- group, being the only biochronological unit/site for which MSR was not calculated. The methodology developed using the Quercy and Limagne Area data set suggests that the upper MSR cluster will be the most trustworthy and should contain the least number of incorrect clusters, if any.

59

Figure 3.6. Neighbour-joining clusters for the Riversleigh sites, using two sites per biochronological unit to estimate species richness and similarity for the calculation of A) the upper MSR, B) MSR and C) lower MSR. Bootstrap values are shown at each node.

60 Before discussing the clusters in detail, the results of the taxonomic distinctness analysis are examined in order to identify sites that may potentially produce incorrect clusters. The results of the taxonomic distinctness analysis are shown in Figure 3.7. In the average taxonomic distinctness graph, only one site fell outside the 99% probability funnel: FZC_KCB. Three sites fell outside the 95% probability funnel. These sites are

FZA_WH, FZA_G and FZ?_GP. These sites have lower than expected average taxonomic distinctness, meaning that the taxa present in their faunal lists are more closely related to each other than average, hence having a potential bias toward certain taxonomic groups. Only FZC_KCB and FZA_WH have been selected in the MSR clusters, thus potentially affecting clustering.

In the variation in taxonomic distinctness graph, six sites fell outside the 99% probability funnel and two outside the 95% probability funnel. These sites are FZB_JH,

FZB_RSO, FZB_WW, FZC_Gag, FZC_Ring and FZC_Wang for the 99% probability funnel and FZC_CK and FZC_GC for the 95% probability funnel. FZB_RSO,

FZB_WW, FZC_Gag, FZC_Ring, FZC_Wang and FZC_CK are of particular interest because they are selected in the MSR clusters. These sites have higher than expected variation in taxonomic distinctness, meaning that the taxa present in their respective faunal lists are not spread evenly within the overall taxonomic spectrum, hence having potential taxonomic biases. Having identified sites with potential taxonomic biases, the clusters can now be described in detail, Faunal Zone by Faunal Zone, to identify potentially incorrect clusters.

61

Figure 3.7. Plot of the average taxonomic distinctness (left) and variation in taxonomic distinctness (right) with 99% (thin doted lines) and 95% (thick doted lines) probability funnels for the Riversleigh sites.

Faunal Zone A is represented by five sites, FZA_BR, FZA_D, FZA_LSO,

FZA_QL and FZA_WH in the lower MSR cluster, by 2 sites FZA_D and FZA_WH in the MSR cluster and by 1 site FZA_WH in the upper MSR cluster. In the lower MSR cluster (Fig. 3.6C), all Faunal Zone A sites cluster together with bootstrap values varying between 62 and 11. In the MSR cluster (Fig. 3.6B), FZA_D and FZA_WH cluster together with a strong bootstrap value of 70 and cluster with FZC_KCB with a low bootstrap value of 34. In the upper MSR cluster (Fig. 3.6A), FZA_WH clusters as a sister group to all Faunal Zone B sites. FZC_KCB‘s was expected to cluster with Faunal

Zone C sites, but its position varies within the cluster, and bootstrap values do not support strongly its clustering with any Faunal Zone. FZC_KCB and FZA_WH were found to have a lower than expected average taxonomic distinctness and, as seen in the previous case study, this may lower significantly the bootstrap values or result in an incorrect clustering, especially if it is the single representative of a biochronological unit. FZA_WH‘s position in the cluster is more stable and supported by higher bootstrap values than FZC_KCB. Comparing the MSR cluster (Fig. 3.6B) with the lower MSR cluster (Fig. 3.6C), the inclusion of FZA_BR, FZA_LSO and FZA_QL in the cluster resulted in lower bootstrap values and lower confidence in the overall

62 clustering of Faunal Zone A sites. In this case, the lower MSR might be too low, resulting in the inclusion of sites that may not be representative. Overall, only two sites,

FZA_D and FZA_WH, are regarded here to be representative of Faunal Zone A, but caution is needed with FZA_WH. FZC_KCB should be discarded from further studies until its faunal list is no longer biased.

Faunal Zone B sites separate from all other sites in all clusters (Fig. 3.6).

Bootstrap values are higher overall in the upper MSR cluster, followed by the MSR cluster and lower MSR cluster. Faunal Zone B separates from Faunal Zone C sites in all three clusters, but bootstrap values are very low. This is expected because Faunal Zones represent much longer periods of time than the MP and MN of the Quercy and Limagne area. In Figure 3.3, the higher branches of the clusters, connecting larger groups of MP and MN had very low bootstrap values. Clarke and Warwick (1994) showed that cluster analysis was weak at identifying higher levels of relationship and therefore recommend the use of ordination for that purpose. Ordination of the Riversleigh Faunal Zones has shown them to be clearly distinct from each other (Travouillon et al., 2006). In this case study, the main concern is with how much confidence a site can be assigned to a Faunal

Zone using the MSR and taxonomic distinctness methods. FZB_WW and FZB_RSO have a higher than expected variation in taxonomical distinctness, but only FZB_RSO has low bootstrap values. This may be due to its much lower species richness, which would decrease the confidence of its clustering, while FZB_WW with high species richness is less affected. As for Faunal Zone A sites, it is probably wise to avoid using lower MSR values. This means that only six sites might be used confidently as representatives of Faunal Zone B: FZB_CR, FZB_CS, FZB_WW, FZB_U, FZB_NG and FZB_DT.

63

Faunal Zone C has 5 taxonomically biased sites, out of the 9 selected in the lower MSR cluster. As noted above, FZC_KCB has a lower than expected average taxonomic distinctness and is removed from the analysis. FZC_CK, FZC_Gag,

FZC_Ring and FZC_Wang have higher than expected variation in taxonomic distinctness. FZC_Gag‘s position within the three clusters is consistent (Fig. 3.6), always clustering with FZC_LM and FZC_HH. In comparison, the position of

FZC_CK, FZC_Ring and FZC_Wang is not consistent (Fig. 3.6), clustering together with other Faunal Zone C sites in the lower MSR cluster and clustering together with

Faunal Zone B sites in the MSR cluster. This is consistent with the problem seen with

Faunal Zone B sites FZB_WW and FZB_RSO, where FZB_WW was not affected by its taxonomic bias because it has high species richness, while FZB_RSO has a lower species richness. FZC_CK, FZC_Ring and FZC_Wang have low species richness compared to FZC_Gag. From observations made in both case studies, it seems that sites selected by the upper MSR are not affected or at least less affected by taxonomic biases than sites only selected by MSR or lower MSR. From these observations, we recommend including in future studies only sites selected using the upper MSR for maximum confidence, but sites selected by MSR may be used if taxonomically biased sites are removed. Use of the lower MSR is not recommended.

In this study, the MSR of each of the three major Riversleigh biochronological units, Faunal Zones A, B and C, is estimated and any potential taxonomical biases are identified. Of the 84 Riversleigh sites examined in this study, only 7 sites can be confidently used as representative of the Faunal Zones, having sufficient taxa in their faunal list and not showing any taxonomic biases. These sites are: FZB_DT, FZB_NG,

64 FZB_U, FZB_CS, FZC_HH, FZC_COA and FZD_En. A further 7 sites may also be considered as having sufficient taxa (using MSR values) despite some being taxonomically biased. These sites are: FZA_D, FZA_WH, FZB_WW, FZC_AL90,

FZC_Gag and FZC_LM. The inclusion of these 7 sites in future studies should be done with caution because they may lower the confidence of interpretations. Inclusion of other Riversleigh sites in future sites is likely to introduce noise and result in misleading results.

3.5 CONCLUSIONS

In this study, MSR methodology is tested. This technique seeks to eliminate from a given data set all fossil sites that could potentially be too incomplete to use in a cluster analysis. The method is demonstrated to be robust and to provide high confidence to the clustering of fossils sites. This method requires inclusion of at least two sites per biochronological unit, which are used to calculate a species richness estimate for the biochronological unit, as well as an estimate of the similarity between the units compared in the analysis. From the estimates of species richness and similarity, MSR is calculated, giving the minimum number of taxa required for a site to correctly cluster as a representative of the biochronological unit. It is established that the weakness of the MSR method is the assumption that all species are equal. The presence of many common species and few rare species in the faunal list of a site may lead to incorrect clustering. Despite the fact that this weakness cannot be eliminated, it can be reduced significantly if the MSR method is used in conjunction with the taxonomical distinctness analysis. This method is able to identify sites that may be taxonomically biased, having either too many taxa related to each other or unevenly spread within the taxonomical hierarchy. These biases may also be the result of taphonomic processes or

65 emphasis by researchers on specific taxonomic groups at the expense of others. We recommend using sites that are selected with MSR or upper MSR, discarding within such selected sites those that are shown to be taxonomically biased.

Although the MSR method and taxonomic distinctness are tested here on biochronological data, the technique may also be used for other types of data, notably environmental data.

3.6 REFERENCES

Aguilar, J.-P., S. Legendre and J. Michaux, Eds. (1997). Actes du congrès BiochroM‘97, Mémoires et Travaux de l‘Ecole Pratique des Hautes Etudes. Institut de Montpellier.

Aplin, K. P. and M. Archer (1987). Recent Advances in Marsupial systematics with a new syncretic classification. Possums and Opossums, Studies in Evolution. M. Archer. Sydney, Surrey Beatty and Sons.

Archer, M., D. A. Arena, M. Bassarova, R. M. D. Beck, K. Black, W. E. Boles, P. Brewer, B. N. Cooke, K. Crosby, A. Gillespie, H. Godthelp, S. J. Hand, B. P. Kear, J. Louys, A. Morrell, J. Muirhead, K. K. Roberts, J. D. Scanlon, K. J. Travouillon and S. Wroe (2006). Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1: 1-17.

Archer, M., S. J. Hand and H. Godthelp (1994a). Riversleigh, Sydney, 256 pp.

Arena, D. A. (2004). The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle Tertiary. PhD thesis, UNSW, Sydney.

Bates, H. (2007). 'New Miocene Species of Musky Rat Kangaroos (Hypsiprymnodontidae, Macropodoidea) Description, Phylogenetics and a New Hypotheses about Kangaroo Evolution. Honours, University of New South Wales, Sydney.

Beck, R. M. D., O. R. P. Bininda-Emonds, M. Cardillo, F.-G. R. Liu and A. Purvis (2006). A higher-level MRP supertree of placental mammals. BMC Evolutionary Biology 6(93).

Brewer, P., M. Archer, S. Hand and H. Godthelp (2007). A new species of the wombat Warendja from late Miocene deposits at Riversleigh, North-West Queensland, Australia. Palaeontology 50(4): 811-828.

66 Chao, A. and T.-J. Shen (2003). Program SPADE (Species Prediction and Diversity Estimation). Program and User's Guide available at http://chao.stat.nthu.edu.tw.

Chao, A., T.-J. Shen and W.-H. Hwang (2006). Application of Laplace's boundary- mode approximations to estimate species and shared species richness. Australian and New Zealand Journal of Statistics 48: 117-128.

Clarke, K. R. and R. M. Warwick (1994). Change in Marine Communities: An approach to statistical analysis and interpretation. Plymouth Marine Laboratory, Plymouth, UK., 144 pp.

Clarke, K. R. and R. M. Warwick (1998). A taxonomic distinctness index and its statistical properties. Journal of Applied Ecology 35: 523-531.

Clarke, K. R. and R. M. Warwick (1999). The taxonomic distinctness measure of biodiversity: weighting of step lengths between hierarchical levels. Marine Ecology Progress Series 184: 21-29.

Clarke, K. R. and R. M. Warwick (2001). A further biodiversity index applicable to species lists: variation in taxonomic distinctness. Marine Ecology Progress Series 216: 265-278.

Crosby, K. (2002). Studies in the diversity and evolution of Phalangeroid possums (Marsupialia; Phalangerida; Phalangeroidea). PhD, University of New South Wales, Sydney.

Crosby, K. (2007). Rediagnosis of the fossil species assigned to Strigocuscus (Marsupialia, Phalangeridae), with description of a new genus and three new species. Alcheringa 31: 33-58.

Dice, L. R. (1945). Measures of the amount of ecological association between species. Ecology 26: 297-302.

Escarguel, G. (2005). Mathematics and the lifeway of Mesopithecus. International Journal of Primatology 26: 801-823.

Escarguel, G. and S. Legendre (2006). New methods for analysing deep-time meta- community dynamics and their applications to the Paleogene mammals from Quercy and Limagne area (Massif Central, France). Strata, serie 1 13: 245-273.

Escarguel, G., B. Marandat and S. Legendre (1997). Sur l'âge numérique des faunes de mammifères du Paléogène d'Europe occidentale, en particulier celles de l'Eocène inférieur et moyen. Actes du congrès BiochroM'97, Mémoires et Travaux de l'Ecole Pratique des Hautes Etudes. J.-P. Aguilar, Legendre, S., Michaux, J. Institut de Montpellier. 21: 443-460.

Gillespie, A. K. (2007). Diversity and systematics of marsupial lions from the Riversleigh World Heritage Area and the evolution of the thylacoleonidae. Doctor of Philosophy, University of New South Wales, Sydney.

67 Hammer, Ø., D. A. T. Harper and P. D. Ryan (2001). PAST: Palaeontological Statistics software package for education and data analysis. Palaeontologia Electronica 4(1): 9 pp.

Kear, B. P., B. N. Cooke, M. Archer and T. F. Flannery (2007). Implications of a new species of the Oligo-Miocene Kangaroo (Marsupialia: Macropodoidea) Nambaroo, from the Riversleigh World Heritage Area, Queensland, Australia. Journal of Paleontology 81(6): 1147-1167.

Kielan-Jaworowska, Z., R. L. Cifelli and Z. X. Luo (2004). Mammals from the age of dinosaurs: origins, evolution, and structure. Columbia University Press, New York

Legendre, S. (1986). Analysis of mammalian communities from the late Eocene and Oligocene of Southern France. Palaeovertebrata 16: 191-212.

Legendre, S. (1987). Les communautés de mammifères d'Europe occidentale de l'Eocene supérieur et Oligocène: structures et milieux. Münchner Geowissenschaftliche Abhandlungen (A) 10: 301-312.

Legendre, S. (1987b). Les communautés de mammifères et leurs milieux en Europe Occidentale de l'Eocene supérieur et l'Oligocène. Revue Paleobiologie 6: 1983-1988.

Legendre, S. (1987c). Concordance entre paléontologie continentale et les événements paléocéanographiques :exemple des faunes de mammifères du Paléogène du Quercy. Comptes Rendus à l'Académie des Sciences, Paris, (III) 304: 45-50.

Legendre, S. (1987d). Mammalian faunas as paleotemperature indicators: concordance between oceanic and terrestrial paleontological evidence. Evolutionary Theory 8: 77-86.

Legendre, S. (1989). Les communautés de mammifères du Paléogène (Eocène supérieur et Oligocène) d'Europe occidentale: structures, milieux et évolution. Münchner Geowissenschaftliche Abhandlungen (A) 16: 1-110.

Legendre, S. and B. Bachelet (1993). The numerical ages: a new method of datation applied to Paleogene mammalian localities from Southern France. Newsletters on Stratigraphy 29: 137-158.

Legendre, S., J.-Y. Crochet, M. Godinot, J.-L. Hartenberger, B. Marandat, J.-A. Remy, B. Sigé, J. Sudre and M. Vianey-Liaud (1991). Evolution de la diversité des faunes de mammifères d'Europe occidentale au Paléogène (MP 11 à MP 30). Bulletin de la Société géologique de France 162: 867-874.

Legendre, S. and C. Girard (1999). Impact des crises dans les domaines marins et terrestres : évolution de la biodiversité des conodontes (Dévonien, Montagne Noire) et des mammifères (Paléogène, Quercy). Geobios 32: 275-284.

Legendre, S. and J.-L. Hartenberger (1992). Evolution of Mammalian Faunas in Europe during the Eocene and Oligocene. Climatic and biotic evolution. D. R. Prothero and W. A. Berggren, Princeton University Press: 516-528.

68 Legendre, S., C. Mourer-Chauvire, M. Hugueney, E. Maitre, B. Sigé and G. Escarguel (2006). Dynamique de la diversité des mammifères et des oiseaux paléogènes du Massif Central. Strata, serie 1 13: 273-280.

Legendre, S., B. Sigé, J.-G. Astruc, L. d. Bonis, J.-Y. Crochet, C. Denys, M. Godinot, J.-L. Hartenberger, F. Leveque, B. Marandat, C. Mourer-Chauvire, J.-C. Rage, J.-A. Remy, J. Sudre and M. C. Vianey-Liaud (1997). Les phosphorites du Quercy : 30 ans de recherche. Bilan et perspectives. Geobios M.S. 20: 331-345.

Madsen, O., M. Scally, C. J. Douady, D. J. Kao, R. W. DeBry, R. Adkins, H. M. Amrine, M. J. Stanhope, W. W. de Jong and M. S. Springer (2001). Parallel adaptive radiations in two major clades of placental mammals. Nature 409: 610-614.

McKenna, M. C. and S. K. Bell (1997). Classification of Mammals Above the Species Level. . Columbia University Press, New York

Muirhead, J. (1994). Systematics, evolution and palaeobiology of recent and fossil bandicoots (Peramelemorphia, Marsupialia). PhD, University of New South Wales, Sydney.

Murphy, W. J., E. Eizirik, W. E. Johnson, Y. P. Zhang, O. A. Ryder and S. J. O'Brien (2001a). Molecular phylogenetics and the origins of placental mammals. Nature 409(614-618).

Murphy, W. J., E. Eizirik, S. J. O'Brien, O. Madsen, M. Scally, C. J. Douady, E. Teeling, O. A. Ryder, M. J. Stanhope, W. W. de Jong and M. S. Springer (2001b). Resolution of the early placental mammal radiation using Bayesian phylogenetics. Science 294: 2348-2351.

Murray, P. and D. Megirian (2006). Cranial morphology of the Miocene thylacinid Mutpuracinus archibaldi (Thylacinidae, Marsupialia) and relationships within the Dasyuromorphia. Alcheringa Special Issue 1: 229-276.

Nishihara, H., M. Hasegawa and N. Okada (2006). Pegasoferae, an unexpected mammalian clade revealed by tracking ancient retroposon insertions. Proceedings of the National Academy of Sciences of the United States of America 103: 9929-9934.

Pledge, N. S. (2005). The Riversleigh wynyardiids. Memoirs of the Queensland Museum 51(1): 135-169.

Rose, K. D. (2006). The beginning of the age of mammals. The John Hopkins University Press, Baltimore

Saitou, N. and M. Nei (1987). The Neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4: 406-425.

Springer, M. S., M. J. Stanhope, O. Madsen and W. W. de Jong (2004). Molecules consolidate the placental mammal tree. Trends in Ecology and Evolution 19: 430-438.

69 Travouillon, K. J., M. Archer, S. J. Hand and H. Godthelp (2006). Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1: 323-349.

Travouillon, K. J., M. Archer, S. Legendre and S. J. Hand (2007). Finding the Minimum Sample Richness (MSR) for multivariate analyses: implications for palaeoecology. Historical Biology.

Waddell, P. J. and S. Shelley (2003). Evaluating placental inter-ordinal phylogenies with novel sequences including RAG1, gamma-fibrinogen, ND6, and mt-tRNA, plus MCMC-driven nucleotide, amino acid, and codon models. Molecular Phylogenetics and Evolution 28: 197-224.

Warwick, R. M. and K. R. Clarke (1995). New —biodiversity“ measures reveal a decrease in taxonomic distinctness with increasing stress. Marine Ecology Progress Series 129: 301-305.

Warwick, R. M. and K. R. Clarke (1998). Taxonomic distinctness and environmental assessment. Journal of Applied Ecology 35: 532-543.

3.7 APPENDIX

The appendix contains updates and corrections to the Riversleigh mammalian species list published in Archer et al. (2006). Updates and corrections are presented here following the taxonomical listing of Archer et al. (2006) and contains presence additions (new specimen found) or removal (incorrect presence) and change of species status (new species published).

Thylacinidae

Species name: Nimbacinus dicksoni Updates/Corrections: Remove presence from Upper Site Note: No specimen from Upper Site, presence incorrectly included in Archer et al. (1994a).

Species name: Thylacinidae cf. Mutpuracinus archibaldi Updates/Corrections: Change to Mutpuracinus archibaldi Note: Published in Murray and Megirian (2006).

Dasyuridae

Species names: Dasyuridae genus indet. sp. 1, 2, 3 and 4

70 Updates/Corrections: Remove presence from Upper Site Note: Presence noted in Archer et al. (1994a) but no identifiable specimen can be allocated at this stage.

Dasyuromorphia incertae sedis

Species names: Dasyuromorphia new genus new sp. Updates/Corrections: Remove Note: Synonymized with Joculusium muizoni and therefore should be removed.

Notoryctidae

Species names: Notoryctid new genus new sp. Updates/Corrections: Add presence in Camel Sputum Site Note: New identified specimen (R. Beck, pers. comm., 2007).

Yaralidae (Remove Peramelemorphia incertae sedis) (Muirhead, 1994)

Species names: Yaralidae genus 1 sp. 1 Updates/Corrections: Add presence in Boid Site East, Dirk‘s Tower Site, Gotham City Site, Judith‘s Horizontalis Site, Lee Sye‘s Outlook Site, Outa Site, Price Is Right Site, Ringtail Site and Rat Vomit Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 2 sp. 1 Updates/Corrections: Add presence in Cleft of Ages Site, Last Minute Site and Wang Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 2 sp. 2 Updates/Corrections: Add presence in Creaser‘s Rampart Site. Remove presence in Upper Site. Note: New identified specimen. No Upper Site specimen identified (personal observation).

Species names: Yaralidae genus 2 sp. 3 Updates/Corrections: Add presence in Camel Sputum Site, Dirk‘s Tower Site, Micro Site, Mike‘s Menagerie Site, Neville‘s Garden Site, Ross Scott-Orr Site and Wayne‘s Wok Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 2 cf. sp. 3 Updates/Corrections: New species present in Boid Site, Quantum Leap Site and Judith‘s Horizontalis Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 3 sp. 1 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site, Cadbury‘s Kingdom Site, Cleft of Ages Site, Creaser‘s Rampart Site, Dirk‘s Tower Site, Gag Site, Judith‘s Horizontalis Site, Last Minute Site, Outa Site and Price Is Right Site.

71 Note: New identified specimen (personal observation).

Species names: Yaralidae genus 3 cf. sp. 1 Updates/Corrections: New species present in White Hunter Site and VIP Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 4 sp. 1 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site, Creaser‘s Rampart Site, Gag Site, Gotham City Site, Henk‘s Hollow Site, Jim‘s Carousel Site, Main Site and Ross Scott-Orr Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 4 sp. 2 Updates/Corrections: Add presence in Cadbury‘s Kingdom Site, Creaser‘s Rampart Site, Mike‘s Menagerie Site, Price Is Right Site, Ringtail Site, Two Trees Site and Wang Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 4 cf. sp. 2 Updates/Corrections: New species present in Judith‘s Horizontalis and VIP Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 5 sp. 2 Updates/Corrections: Add presence in Gag Site, Henk‘s Hollow Site, Last Minute Site, Neville‘s Garden Site, Ringtail Site and Wayne‘s Wok Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 5 sp. 3 Updates/Corrections: New species present in Lee Sye‘s Outlook Site. Note: New identified specimen (personal observation).

Species names: Yaralidae genus 5 sp. 4 Updates/Corrections: New species present in Gag Site. Note: New identified specimen (personal observation).

Species names: Yarala burchfieldi Updates/Corrections: Add presence in Bitesantenary Site, Cadbury‘s Kingdom Site, Creaser‘s Rampart Site, Dirk‘s Tower Site, Encore Site, Judith‘s Horizontalis Site, Price Is Right Site and Ringtail Site. Remove presence from Keith‘s Chocky Block Site. Note: New identified specimen. No identified specimen from Keith‘s Chocky Block Site (personal observation).

Phascolarctidae

Species names: Litokoala kutjamarpensis Updates/Corrections: Add presence in Camel Sputum Site, Dirk‘s Tower Site and Outa Site. Note: New identified specimen (K. Black, pers. comm., 2007).

Species names Nimiokoala greystanesi

72 Updates/Corrections: Remove presence in Outa Site. Note: Incorrectly identified. Outa Site specimen is Litokoala kutjamarpensis (K. Black, pers. comm., 2007).

Thylacoleonidae

Species names: Wakaleo oldfieldi Updates/Corrections: Add presence in Helicopter Site, Jim‘s Jaw Site, Golden Steph Site and Keith Chocky Block Site. Note: New identified specimen in PhD thesis (Gillespie, 2007).

Species names: Wakaleo vanderleueri Updates/Corrections: Remove presence in Keith Chocky Block Site. Note: Keith Chocky Block Site specimen is Wakaleo oldfieldi (Gillespie, 2007).

Species names: Wakaleo new sp. 1 Updates/Corrections: Add presence in Upper Site, Neville‘s Garden Site, Creaser‘s Rampart Site and Burnt Offering Site. Remove presence in D Site. Note: New identified specimen in PhD thesis, and D Site specimen is Wakaleo cf. new sp. 1 (Gillespie, 2007).

Species names: Wakaleo cf. new sp. 1 Updates/Corrections: Add new species presence in D Site. Note: D Site specimen is cf. new sp. 1 (Gillespie, 2007).

Species names: Priscileo roskellyae Updates/Corrections: Add presence in White Hunter Site and Phenomenal Fissure Fill Site. Note: New identified specimen in PhD thesis (Gillespie, 2007).

Species names: Thylacoleonidae new genus new sp. Updates/Corrections: Change species name to Thylacoleonidae new genus 1 new sp. Note: Addition of new Thylacoleonid species requires change of name of this species (Gillespie, 2007).

Species names: Thylacoleonidae new genus 2 new sp. Updates/Corrections: Add new species presence in Dirk‘s Tower Site. Note: New species identified in PhD thesis (Gillespie, 2007).

Vombatidae

Species name: Warendja sp. 1 Updates/Corrections: Change to Warendja encorensis Note: Published in Brewer et al. (2007).

Wynyardiidae

Species names: Namilamadeta albivenator Updates/Corrections: Remove presence in Lee Sye‘s Outlook Site.

73 Note: Lee Sye‘s Outlook Site specimen is Namilamadeta sp. cf. N. albivenator (Pledge, 2005).

Palorchestidae

Species names: Propalorchestes ponticulus Updates/Corrections: Remove presence from Hiatus Site. Note: Incorrectly identified (K. Black, pers. comm., 2007).

Species names: Propalorchestes novaculacephalus Updates/Corrections: Remove presence from Neville‘s Garden Site, Gag Site and Jim‘s Jaw Site. Note: Incorrectly identified (K. Black, pers. comm., 2007).

Species names: Ngapakaldia bonythoni Updates/Corrections: Add presence in Al Site, Bone Reef Site, Lee Sye‘s Outlook Site, 300m from Bone Reef Site, Upper Burnt Offering Site, Dirk‘s Tower Site, Mike‘s Potato Patch Site and Hiatus South Site. Note: New identified specimen. Al Site Ngapakaldia sp. specimen and 300m from Bone Reef Site Neohelos tirarensis specimen are Ngapakaldia bonythoni (K. Black, pers. comm., 2007).

Species names: Ngapakaldia sp. Updates/Corrections: Remove species presence. Note: Al Site specimen is Ngapakaldia bonythoni (K. Black, pers. comm., 2007).

Diprotodontidae

Species names: Nimbadon lavarackorum Updates/Corrections: Remove presence in Ringtail Site; add presence in Cleft Of Ages Site and Last Minute Site. Note: No specimen from Ringtail Site. New identified specimen in Cleft Of Ages Site and Last Minute Site (K. Black, pers. comm., 2007).

Species names: Silvabestius johnnilandi Updates/Corrections: Add presence in Bone Reef Site and Steph‘s Small Reward Site. Note: New identified specimen (K. Black, pers. comm., 2007).

Species names: Neohelos tirarensis Updates/Corrections: Add presence in Wang Site, Black Coffee 2 Site, Neville‘s Pancake Site, Dirk‘s Tower Site, White Hunter Site and Creaser‘s Rampart Site. Remove presence in Upper Site and 300m from Bone Reef Site. Note: New identified specimen. No specimen from Upper Site. 300m from Bone Reef Site Neohelos tirarensis specimen is Ngapakaldia bonythoni (K. Black, pers. comm., 2007).

Species names: Neohelos sp. A Updates/Corrections: Add presence in Keith‘s Chocky Block Site. Note: New identified specimen (K. Black, pers. comm., 2007).

74 Species names: Neohelos stirtoni Updates/Corrections: Add presence in Golden Steph Site. Note: New identified specimen (K. Black, pers. comm., 2007).

Burramyidae

Species names: Burramys brutyi Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site, Bone Reef Site, Fireside Favourites Site, Gillespie‘s Gully Site, Judith‘s Horizontalis Site, Lee Sye‘s Outlook Site and Micro Site. Note: New identified specimen (personal observation, 2007).

Species names: Cercartetus new sp. Updates/Corrections: Add presence in Last Minute Site and Main Site. Remove presence in Wayne‘s Wok Site. Note: New identified specimen. No specimen from Wayne‘s Wok Site (personal observation, 2007).

Pseudocheiridae

Species names: Paljara nancyhaywardae Updates/Corrections: Add presence in Cadbury‘s Kingdom Site, Cleft Of Ages Site, Dirk‘s Tower Site, Judith‘s Horizontalis Site, Lois & Diedrie 1994 Site, Neville‘s Garden Site and Panorama Site. Note: New identified specimen (K. Roberts, pers. comm., 2007).

Species names: Paljara tirarensae Updates/Corrections: Add presence in Dirk‘s Tower Site, Encore Site, Gag Site, Last Minute Site, Neville‘s Garden Site, Two Trees Site and Upper Site. Note: New identified specimen (K. Roberts, pers. comm., 2007).

Species names: Gawinga aranaea Updates/Corrections: Add presence in Gotham City Site and Cadbury‘s Kingdom Site. Note: New identified specimen (K. Roberts, pers. comm., 2007).

Species names: Pseudocheiridae new genus 1 sp. 1 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site, Cadbury‘s Kingdom Site, Cleft Of Ages Site, Creaser‘s Rampart Site, Camel Sputum Site, Dirk‘s Tower Site, Encore Site, Gag Site, Gotham City Site, Henk‘s Hollow Site, Jeanette‘s Birthday Site, Keith Chocky Block Site, Last Minute Site, Main Site, Neville‘s Garden Site, Price Is Right Site, Ringtail Site, Ross Scott-Orr Site, Upper Site, Wang Site, White Hunter Site and Wayne‘s Wok Site. Note: New species identified (K. Roberts, pers. comm., 2007).

Species names: Pseudochirops sp. 1 Updates/Corrections: Change to Pseudocheiridae new genus 1 sp. 2. Add presence in Alan‘s Ledge 1990 Site, Boid Site East, Crusty Meat Pie Site, Dome Site, Judy‘s Jumping Joint Site, Rick‘s Sausage Site, Rat Vomit Site, Cadbury‘s Kingdom Site, Upper Site, Neville‘s Garden Site and Two Trees Site.

75 Note: New identified specimen. This species is not in the Pseudochirops genus (K. Roberts, pers. comm., 2007).

Species names: Pseudochirops sp. 2 Updates/Corrections: Remove taxa. Note: Merged with Pseudocheiridae new genus 1 sp. 2. (K. Roberts, pers. comm., 2007).

Species names: Marlu sp. 1 Updates/Corrections: Remove species presence. Note: Lee Sye‘s Outlook Site and Ross Scott-Orr Site specimen are Pseudocheiridae new genus 2 sp. 1 (K. Roberts, pers. comm., 2007).

Species names: Marlu sp. 2 Updates/Corrections: Add presence in Henk‘s Hollow Site. Note: Henk‘s Hollow Site specimen previously Pseudochirops sp. 2 is Marlu sp. 2 (K. Roberts, pers. comm., 2007).

Species names: Marlu sp. 3 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site and Henk‘s Hollow Site. Note: Alan‘s Ledge 1990 Site and Henk‘s Hollow Site specimen previously Marlu cf. sp. 3 are Marlu sp. 3 (K. Roberts, pers. comm., 2007).

Species names: Marlu cf. sp. 3 Updates/Corrections: Remove species presence. Note: Alan‘s Ledge 1990 Site and Henk‘s Hollow Site specimen previously Marlu cf. sp. 3 are Marlu sp. 3 (K. Roberts, pers. comm., 2007).

Species names: Pildra sp. 1 Updates/Corrections: Add presence in Cadbury‘s Kingdom Site, Camel Sputum Site, Gotham City Site, Henk‘s Hollow Site, Last Minute Site, Mike‘s Menagerie Site, Ringtail Site, Rick‘s Sausage Site and White Hunter Site. Note: New identified specimen. Pildra sp. 3 and Pildra sp. 4 specimen are merged with Pildra sp. 1 (K. Roberts, pers. comm., 2007).

Species names: Pildra sp. 2 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site, Cadbury‘s Kingdom Site, Jim‘s Jaw Site, Last Minute Site, Micro Site, Price Is Right Site, Ringtail Site and Wang Site. Remove Encore site Specimen. Note: New identified specimen. Encore Site Specimen is Pseudocheiridae new genus 1 sp. 1 (K. Roberts, pers. comm., 2007).

Species names: Pildra sp. 3 Updates/Corrections: Remove species presence. Note: Merged with Pildra sp. 1 (K. Roberts, pers. comm., 2007).

Species names: Pildra sp. 4 Updates/Corrections: Remove species presence. Note: Merged with Pildra sp. 1 (K. Roberts, pers. comm., 2007).

76 Species names: Pseudocheiridae new genus 2 sp. 1 Updates/Corrections: Add presence in Camel Sputum Site, Dirk‘s Tower Site, Lee Sye‘s Outlook Site and Ross Scott-Orr Site. Remove presence in Gag Site, Jim‘s Jaw Site, Last Minute Site, Neville‘s Garden Site, Ringtail Site and Upper Site. Note: New identified specimen. Specimen from Gag Site, Jim‘s Jaw Site, Last Minute Site, Neville‘s Garden Site, Ringtail Site and Upper Site are Pseudocheiridae new genus 2 sp. 2. Lee Sye‘s Outlook Site and Ross Scott-Orr Site specimen were previously Marlu sp. 1 (K. Roberts, pers. comm., 2007).

Species names: Pseudocheiridae new genus 2 sp. 2 Updates/Corrections: Remove species presence. Note: Merged with Pildra sp. 2 (K. Roberts, pers. comm., 2007).

Phalangeridae

Species names: Trichosurus dicksoni Updates/Corrections: Add presence in Ringtail Site and Angela‘s Sinkhole Site. Note: Omitted presence from Crosby (Crosby, 2002).

Species names: —Strigocuscus“ reidi Updates/Corrections: Change to Onirocuscus reidi. Note: Published in Crosby (2007).

Species names: Phalangeridae new genus 1 sp. 1 Updates/Corrections: Change to Onirocuscus inversus. Remove presence in Creaser‘s Rampart Site. Note: Published in Crosby (2007). Creaser‘s Rampart Site specimen is Onirocuscus sp. cf. O. inversus.

Species names: Onirocuscus sp. cf. O. inversus Updates/Corrections: Add presence in Creaser‘s Rampart Site. Note: Published in Crosby (2007).

Species names: Phalangeridae new genus 1 sp. 2 Updates/Corrections: Change to Onirocuscus rupina. Note: Published in Crosby (2007).

Species names: Phalangeridae new genus 1 sp. 3 Updates/Corrections: Change to Onirocuscus silvicultrix. Remove presence in Dirk‘s Tower Site. Note: Published in Crosby (2007). Dirk‘s Tower Site specimen is Onirocuscus sp. cf. O. silvicultrix.

Species names: Onirocuscus sp. cf. O. silvicultrix Updates/Corrections: Add presence in Dirk‘s Tower Site. Note: Published in Crosby (2007).

Species names: New Phalangeridae Genus 2 sp 1 Updates/Corrections: Add presence in Hiatus South Site and Group Site. Note: Omitted presence from Crosby (Crosby, 2002).

77

Ektopodontidae

Species names: Ektopodon serratus Updates/Corrections: Remove presence. Note: Creaser‘s Rampart Site specimen is Ektopodontidae new genus new sp. (personal observation, 2007).

Species names: Ektopodon sp. cf. E. serratus Updates/Corrections: Remove presence in Neville‘s Garden Site. Note: Neville‘s Garden Site specimen is Ektopodontidae new genus new sp. (personal observation, 2007).

Species names: Chunia sp. Updates/Corrections: Add presence in Neville‘s Garden Site. Note: New identified specimen (personal observation, 2007).

Species names: Ektopodontidae new genus new sp. Updates/Corrections: Add presence in Price Is Right Site. Note: New identified specimen (K. Roberts, pers. comm., 2007). . Balbaridae

Species names: Ganawamaya acris Updates/Corrections: Add presence in Neville‘s Garden Site. Note: New identified specimen (B. Cooke, pers. comm., 2007).

Species names: Balbaroo gregoriensis Updates/Corrections: Change presence in Boid Site with Boid Site East, remove presence in D Site. Note: Presence incorrectly allocated to Boid Site instead of Boid Site East. D Site presence noted in Archer et al. (1994a) but no identifiable specimen can be allocated at this stage.

Species names: Balbaroo fangaroo Updates/Corrections: Remove presence in Keith Chocky Block Site. Note: Keith Chocky Block Site specimen incorrectly assigned to Balbaroo fangaroo (B. Cooke, pers. comm., 2007).

Species names: Balbaroo sp. 4 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site. Note: New identified specimen (B. Cooke, pers. comm., 2007).

Species names: Wururoo sp. 2 Updates/Corrections: Add presence in Alan‘s Ledge 1990 Site. Note: New identified specimen (B. Cooke, pers. comm., 2007).

Species names: Nambaroo sp. 3 Updates/Corrections: Change to Nambaroo gillespieae. Note: Published in Kear et al.(2007).

78

Species names: Nambaroo sp. 5. Updates/Corrections: Change presence in Quantum Leap Site to White Hunter Site. Note: Presence incorrectly allocated to Quantum Leap Site instead of White Hunter Site.

Hypsiprymnodontidae

Species names: Hypsiprymnodon bartholomaii Updates/Corrections: Remove presence from Neville‘s Garden Site. Note: Neville‘s Garden Site specimen is Hypsiprymnodon new sp. 1 (Bates, 2007).

Species names: Hypsiprymnodon new sp. Updates/Corrections: Change species name to Hypsiprymnodon new sp. 1. Add presence in Creaser‘s Rampart Site, Judith‘s Horizontalis Site, Boid Site East, Keith Chocky Block Site, and Neville‘s Garden Site. Remove presence in Cleft Of Ages Site. Note: Addition of new Hypsiprymnodon species requires change of name of this species. New species described and new specimen described in Honours thesis. Cleft Of Ages Site specimen is Hypsiprymnodon new sp. 2 (Bates, 2007).

Species names: Hypsiprymnodon new sp. 2 Updates/Corrections: Add species presence in Cleft of Ages Site and Upper Site. Note: New species described and new specimen described in Honours thesis (Bates, 2007).

Species names: Hypsiprymnodon new sp. 3 Updates/Corrections: Add species presence in Camel Sputum Site. Note: New species described and new specimen described in Honours thesis (Bates, 2007).

Species names: Ekaltadeta ima Updates/Corrections: Add presence in Gone Over Here Site, Jim‘s Carousel Site and Keith Chocky Block Site. Remove presence in Encore Site. Note: New specimen identified. Encore Site specimens are Ekaltadeta jamiemulvaneyi (personal observation, 2007).

Species names: Ekaltadeta jamiemulvaneyi Updates/Corrections: Remove presence in Cleft Of Ages Site. Note: Cleft Of Ages Site specimens are Ekaltadeta ima (personal observation, 2007).

Macropodidae

Species names: Bulungamaya delicata Updates/Corrections: Add presence in Inabeyance Site and Mike‘s Menagerie Site. Note: New identified specimen (B. Cooke, pers. comm., 2007).

Species names: Ganguroo bilamina Updates/Corrections: Add presence in Dome Site. Note: New identified specimen (B. Cooke, pers. comm., 2007).

79 Yalkaparidontidae

Species names: Yalkaparidon coheni Updates/Corrections: Add presence in Ross Scott-Orr Site and Mike‘s Menagerie Site, Note: New identified specimen. (Beck et al, in prep.).

Species names: Yalkaparidon jonesi Updates/Corrections: Remove presence in Last Minute Site. Note: Cannot confirm species presence from an isolated incisor. (Beck et al., in prep.).

80

CHAPTER 4: EXPLAINING THE GAPS IN MAMMALIAN

BODY MASS DISTRIBUTIONS (CENOGRAMS) AND THE

ECOLOGICAL IMPACT OF INTRODUCED SPECIES IN

AUSTRALIA

* This chapter has been submitted to Palaeogeography, Palaeoclimatology,

Palaeoecology:

Travouillon, K. J., and Legendre, S. (in review). Explaining the gaps in mammalian body mass distributions (cenograms) and the ecological impact of introduced species in

Australia. Palaeogeography, Palaeoclimatology, Palaeoecology .

81 4.1 ABSTRACT

Body size distribution and cenogram analyses both use body weight distributions of mammalian species to describe structural patterns within communities.

Using these methods it has been possible to correlate modern mammalian community structure and habitat. In turn these correlations have been used to infer palaeohabitat from analysis of the structure of extinct mammal communities. Continuous body size distributions indicate closed environments, and discontinuous distributions, with a gap between 500 g and 8 kg, indicate open environments. Some quantitative studies have suggested that the gaps present in open environments are not always statistically significant. However, qualitative studies continue to show that the gap in body mass distribution and cenograms does reliably reflect environment type and is more efficient at doing so than use of more traditional methods such as analyses of diet or locomotion.

Very few studies have tried to explain the gaps in open environments or question the results of the quantitative analyses. Considering all other potential explanations for the gap, we suggest that introductions by Europeans of non-indigenous herbivorous and carnivorous mammals into Australian ecosystems have caused extinction of medium- sized native mammals in the body weight range of 35 g to 5500 g.

We used the cenogram method to construct the body size distribution of both contemporary and pre-European invasion lists of mammal taxa from 52 Australian national parks spanning all major environments. All modern Australian open environments showed a gap in body mass distribution.

Historical open environments showed no distinct gap in body mass distribution but had significantly less medium-sized species than closed environments. Large, introduced

Australian mammalian predators have been shown to prefer medium-sized prey over

82 large or small prey and to contribute significantly to the extinction of medium-sized species in open environments.

We suggest that the observed gaps in open environment mammal communities are at least in part the result of non-indigenous species, particularly carnivores, following the European colonization of Australia.

4.2 INTRODUCTION

The cenogram method is one of the most criticised palaeoecological methods in this discipline. The term cenogram was first proposed by Valverde (1964; 1967) for a graph displaying the relationships between the size of predators and the size of their prey species in a mammalian community. His cenograms were constructed by plotting rank ordered taxa versus head-body length. Legendre (1986; 1989) adapted this method for palaeoecological studies. This adaptation excluded bats, as per Valverde‘s method, as well as carnivores. Body sizes were expressed in Log of body mass (g) instead of head- body length. Legendre (1986; 1989) made three visual observations regarding the structure of cenograms and their environments (see Fig. 4.1):

1. Cenograms of open environments have a gap in the medium-sized species (500-

8000 g) whereas closed environments have a continuous distribution.

2. The slope of large species (over 8000 g) is steeper in more arid environments.

3. The slope of small species (under 500 g) is related to minimal temperatures.

Legendre (1986; 1989) recommended the use of these three rules for qualitative comparison between different faunal complexes. In order to formalise the methods

83 outlined by Legendre (1986; 1989), Gingerich (1989) quantified the gaps and the slopes of cenograms.

Figure 4.1 Cenogram patterns identified by Legendre (1986; 1989).

However, a review of this methodology was undertaken by Rodríguez (1999), who showed that the relationships between cenogram patterns and climate are not statistically significant. Rodríguez‘s (1999) analysis does, however, support the fact that the gap in medium-sized mammals relates to vegetation structure in tropical communities. In a further review, Hernández Fernández et al. (2006) indicate that

Rodríguez‘s (1999) concentrated on a quantitative analysis of the each of the cenogram variables (e.g. gap size, slops) rather than a qualitative approach. In addition, Hernández

Fernández et al. (2006) used a qualitative statistical approach in order to infer biomes using cenogram patterns, comparing the efficiency of cenograms in predicting biomes with four other commonly used ecological variables (taxonomic, trophic, locomotion

84 and body size categories). They concluded that body size categories and cenogram variables were the most accurate for identifying biomes.

Gómez Cano et al. (2006) showed that the method remains efficient with a random species loss up to 60-70% in a fossil assemblage.

Body size categories or body size distribution is another commonly used method to describe faunal communities (Holling, 1992). The similarity of these methods with cenograms is obvious, as both rely on the differential distribution of body sizes with respect to biomes. Holling (1992), without knowing the work of Legendre (1986; 1989), found that patterns of gaps and clumps in the body size distribution of mammal and bird communities were correlated with changes in climate and vegetation structure.

Interestingly for this study, Holling‘s (1992) largest gap in distribution data also represented medium-sized animals (sensu Legendre, 1986; 1989). Siemann and Brown

(1999) re-examined Holling‘s (1992) method by testing the gaps in mammalian body size distribution. They compared the magnitudes of gaps in mammalian communities of

North America and Australia to randomly generated models and found that the gaps were similar in structurally dissimilar but adjacent biomes that shared similar species.

They concluded that the structure of body size distributions reflected taxonomic constraints on body size rather than climate/vegetation constraints. Allen et al. ”s

(2006) review on patterns in body mass distributions identified five competing hypotheses (including that of Siemann and Brown, 1999) as follows: Energetic, phylogenetic, biogeographical, textural discontinuity and community interaction hypotheses. This review concluded that each hypothesis only partially explained pattern in body mass distribution and that mechanisms underlying those patterns are more likely to be multicausal and vary with scale, perhaps explaining the different patterns found by

85 Legendre (1986; 1989) and Holling (1992) using one dataset and by Rodríguez (1999) and Siemann and Brown (1999) using another.

However, it could be that the differences in conclusion between these authors could be a reflection that the original datasets used were not actually representative of the area from which they were sourced. Limitations should be addressed regarding species lists compilation and identification of vegetation structure, as well as the understanding of interaction between animals and their environment. Neither Rodríguez

(1999) nor Siemann and Brown (1999) addressed these issues. As Sand-Jensen (2007) points out, the current trend in scientific writing in biology reduces all species to numbers and statistical elements without considering any interesting biological aspects of adaptation, behavior and evolution. The dataset that Siemann and Brown (1999) used was prior to European settlement for both North America and Australia. In their analysis, Australia was found to have no gaps statistically larger than random. However, since European settlement, 117 species are now listed as extinct, threatened, or vulnerable out of 245 (47% of the fauna) in Australia (Department of the Environment and Water ResourcesShort and Smith, 1994; Resources, 1999). Species most affected are the medium-sized terrestrial species in the weight range 35g to 5.5kg (Australia‘s critical weight range) (Short and Smith, 1994).

The aim of this paper is to examine patterns in body size distribution in

Australian datasets through time and space. In particular, it aims to identify factors producing gaps in the body size distributions and to develop new methodologies to improve on the current ones.

86 4.3 MATERIALS AND METHODS

4.3.1 Mammal species database

We compiled a database of recent mammalian species lists from 52 national parks (NP) and reserves across Australia (Fig. 4.2). These parks and reserves were selected to cover all major habitat types in Australia. Mammal species lists were compiled from a number of sources. The Department for Environment and Heritage

(DEH) of South Australia provided data from the Biological Database of SA (BDBSA) for Mount Remarkable, Flinders Ranges, Gawler Ranges, Vulkathunha - Gammon

Ranges and Witjira NPs. The Department of Sustainability and Environment (DSE) of

Victoria provided data for Wyperfeld, Little Desert, Mount Buffalo and Snowy River

NPs. Online databases were used to collect data for Millstream-Chichester, Kalbarri and

Karijini NPs using Western Australia Museum‘s FaunaBase

(http://www.museum.wa.gov.au/faunabase/prod/index.htm), for Mutawintji, Kinchega,

Mungo, Gundabooka, Bundjalung, Yuraygir, Deua, Wadbilliga, South East Forest,

Abercrombie River, Blue Mountains, Ku-ring-gai Chase and Kosciuszko NPs using the

NSW National Parks and Wildlife Services‘ Wildlife atlas database

(http://wildlifeatlas.nationalparks.nsw.gov.au/wildlifeatlas/watlas.jsp), and for Mungkan

Kandju, Iron Range, Mount Barney, Main Range, Boodjamulla, Simpson Desert,

Currawinya, Diamantina and Carnarvon NPs using Queensland Parks and Wildlife

Service‘s Wildlife Online database

(http://www.epa.qld.gov.au/nature_conservation/wildlife/wildlife_online/).

The faunal list of Lamington NP was taken from its website

(http://lamington.nrsm.uq.edu.au). The data from the following parks and reserves is from publications: Fitzgerald River (Chapman, 1995), Prince Regent River Nature

Reserve (Miles and Burbidge, 1975), Purnululu (Woinarski et al., 1992), Stirling Range

87 (Herford et al., 1999), Croajingolong (Anonymous, 1998a), Grampians (McCann,

1985), Yumbarra Conservation Park (Owens et al., 1995), Bookmark Biosphere

Reserve (Anonymous, 1997), Shoalwater and Corio Bays Area Ramsar Site (Schodde et al., 1992; Anonymous, 1999a), Uluru (Burbidge and McKenzie, 1989; Balding,

2004), Gregory (Anonymous, 2001a), Nitmiluk (Anonymous, 2002a), Kakadu

(Anonymous, 1999b), Savage River (Anonymous, 2001b), Ben Lomond (Anonymous,

1998b), Mount Field (Anonymous, 2002b) and Douglas-Apsley NPs (Anonymous,

1993).

Historical data were also collected from the same sources where available (

Stirling Range, Karijini, Mutawintji, Uluru, Gundabooka, Kinchega, Wadbilliga, Ku- ring-gai Chase and Kosciuszko NPs). Interpretations from our historical data is limited because only Uluru NP has a pre-European record (Burbidge and McKenzie, 1989). All other historical data dates back to the opening of the National Park. Average body weight for each mammal species is from Strahan (1995).

88

Figure 4.2. Map of Australian National Parks and Reserves used in this study. Shapes of park are approximate and sizes of parks are not to scale. The position of the dingo fence is also indicated.

4.3.2 Environmental data

Environmental data was collected for each national park from diverse sources.

Annual Rainfall, Mean Annual Maximum (MAMT) and Minimum (MAmT)

Temperature and climate (based on the Koeppen classification) information were taken from the bureau of meteorology website (http://www.bom.gov.au/). When environmental data was not available for a park, data from the closest meteorological station was used. Vegetation data (Major Vegetation Groups in Australia) was taken from the Department of the Environment and Heritage website

(http://www.environment.gov.au/erin/nvis/publications/major-veg-map.html).

Vegetation data are shown in Table 4.1 along with our simplified classifications

(habitat type). Note that major vegetation groups are different to vegetation types:

89 vegetation groups are much broader categories and include several vegetation types. For example, the vegetation group —eucalypt forests“ refers to eucalypt tall open forest, eucalypt open forest and eucalypt low open forest, all of which are open forests. In

Table 4.1, only the name of the broader vegetation groups (such as eucalypt forest) were included (as opposed to listing detailed vegetation types).

In Australia, the vegetation group —closed forests“ specifically refer to rainforests, however, the habitat type —closed“ has also been used in cenogram methods

(Legendre, 1986, 1987) to refer not only to closed forests but also to any type of forests, including open forests. Similarly, in the cenogram method, the habitat type —open“ includes the vegetation groups —shrublands“, —grasslands“ and —deserts“. —Woodlands“, being an intermediate state between closed and open habitats, have been classified as both closed and open. Because these definitions of habitat types can be ambiguous, we have assigned our own classification. Parks containing major vegetation groups referred to as —rainforest“ or —eucalypt forests“ were classified as —closed“ (equivalent to

Legendre‘s closed/humid habitat), parks containing —woodlands“ were classified as

—semi-open“ (equivalent to Legendre‘s open/humid habitat) and parks containing

—shrublands“ and —grasslands“ were classified as —open“ (equivalent to Legendre‘s open/arid habitat).

90 Table 4.1. List of major vegetation groups and habitat types for each National park used in this study.

91 In his study, Rodríguez (1999) also allocated a single habitat type to each

National Park used. For example, he classified the Guadalupe Mountains National Park

(USA) as a desert, a vegetation group which is indeed present at the park. However, this park also contains riparian woodlands and mountaintop forests, which support a number of arboreal species such as squirrels and chipmunks (http://www.nps.gov/gumo).

Therefore, this particular national park exhibits a mix of closed and open habitats. In modern ecology, the dominant vegetation group (in this case desert) is commonly used to describe the overall community of a national park. However, in the cenogram methodology, this way of classifying habitat type may be in fact problematic.

Cenograms use presence/absence data of individual species as opposed to abundance data, (the latter usually used in modern ecology). In the case of Guadalupe Mountains

National Park, abundance data would show a clear majority of desert species, but with presence/absence data of individual species, both forest and desert types would both be represented. The presence of arboreal species in Guadalupe Mountains National Park, as well as species that are only found in forested areas (e.g., black bears, skunks, porcupines) should therefore be acknowledged, particularly when using cenogram methods. Because the gap in open habitats will be obscured by the overlying closed habitat signature, Guadalupe Mountains National Park would be assigned to a —closed“ habitat in a cenogram representation. Therefore, in our study, we classified parks with mixed closed and open vegetation (e.g. Kakadu NP) according to the highest density vegetation group present, e.g. a mix of forest and woodland is classified as —closed“.

4.3.3 Cenograms

We constructed cenograms for each of the parks studied. Legendre‘s (1986;

1989) method has been modified for application to Australia‘s marsupial faunas.

92 Legendre (1986; 1989) did not explicitly state the reason for excluding both bats and carnivores from his cenograms. Myers (2002) used the cenogram method including and excluding carnivorous taxa and showed that the inclusion of carnivorous taxa made no significant difference to results, at least for Australian datasets. In the case of bats, it can be argued that their body size reflects flying constraints, and therefore may have very little ties with body size distributions of terrestrial mammals. For this reason, we excluded only bats from the cenograms. In order to be able to visually discriminate them, we have given the large carnivorous species (dingo, fox, cat, quolls and

Tasmanian devil) a different symbol to illustrate their position in the cenograms. We also represent all large introduced mammals (those larger than the red kangaroo,

Macropus rufus, such as cattle, sheep, camels, deer, horses, goats, pigs, donkeys and buffalo) by a different symbol, as their current distribution, reflects introduction by humans. Finally, because it is easier to visualise, we have used log10 rather than the natural log of average body mass for each species.

Once the cenograms were constructed, we compared the cenogram shapes of each national park or reserve to the patterns found by Legendre (1986; 1989) (Fig. 4.1) as well as to known habitat types (Table 4.1). In order to investigate the relationship between the gap in medium-sized mammals and openness, we plotted the magnitude of the largest gap between two consecutive species within the whole fauna (excluding large introduced mammals, for the reasons mentioned above). These were plotted in log units versus the position of the largest gap within the cenogram, using the mean weight of the two species defining the largest gap in log units. We also compared the total number of mammal species for each habitat type in three different weight ranges:

93 Legendre‘s (1986; 1989) gap range (500g to 8000g); Australia‘s critical weight range

(35g to 5500g); and a new range found by our analyses (100g to 1000g).

4.3.4 Arboreal taxa

We assume here that the presence of arboreal taxa in an area indicates the presence of trees. To our knowledge, no research had attempted to correlate the presence of arboreal taxa with habitat type. Defining arboreal taxa is arguably quite difficult. However, certain taxa require the presence of trees for feeding and predator avoidance. In Australia, for example, possums, koalas, tree-kangaroos, some dasyurids

(quolls, antechinuses and phascogales) and some rodents can be classified as —arboreal“ or —scansorial“. However, dasyurids and rodents are not fully restricted to closed habitats, and are quite capable of surviving in more open habitats (Strahan, 1995). For this reason, we are referring in this study only to possums, koalas and tree-kangaroos as our —arboreal“ taxa, although, no tree-kangaroos are present in the data.

We examined the presence of these arboreal taxa (as opposed to all other taxa; referred to herein as non-arboreal) in different body classes graphically, and compared this with habitat type and cenograms.

4.3.5 Body mass distribution

Cenograms are a type of representation of body mass distribution. However, patterns in cenograms can be difficult to distinguish (e.g. difference in slop between large and small mammals can be hard to see visually). The representation of body mass distributions, following Holling (1992) ‘s method, will allow easier comparisons of body mass patterns when combined with cenograms patterns. Percentages of taxa in 5 body class categories (logarithmic mass in grams of the following ranges: 0-0.99, 1-

1.99, 2-2.99, 3-3.99 and 4-4.99) for the mammal fauna (excluding large introduced

94 herbivores) of each national park were represented as bar graphs. Arboreal species were represented separately on these graphs. We also compared the patterns found for each park with the overall pattern of body mass distribution of all non-volant Australian mammals, as well as between historical and modern faunas. Mammalian body mass distributions of New Guinea‘s closed forest (Flannery, 1995; Bassarova, 2005) was also examined.

4.3.6 Analysis

Rodríguez (1999) used Kendall‘s R coefficient to find the probability of correlation between cenograms and environmental variables. The advantage of using

Kendall‘s R is that it does not assume normal distribution of the data. We used

Rodríguez (1999)‘s method for our data, but modified the variables used. We used climate, major vegetation group (MVG), mean annual maximum temperature (MAMT), mean annual minimum temperature (MAmT), annual rainfall and habitat type (closed, semi-open, open) as our environmental variables. We used the following variables to describe cenogram, body mass distribution and arboreal taxa patterns: magnitude of largest gap; position of largest gap; percentage of taxa in the logarithmic mass ranges of

0-0.99, 1-1.99, 2-2.99, 3-3.99 and 4-4.99; and percentage of arboreal taxa in the logarithmic mass ranges of 0-0.99, 1-1.99, 2-2.99, and 3-3.99.

4.4 RESULTS

Cenograms for 52 Australian national parks and reserves are represented in Fig.

4.3A to 4.3G. In this figure, cenograms have been arranged according to their habitat type in Fig. 4.3, with 4.3A to 4.3D being closed habitats, 4.3D and 4.3E being semi-

95 open and 4.3F being open habitats. Figure 4.3G shows cenograms of the historical data for four national parks (Karijini, Mutawintji, Stirling Range and Uluru National Parks) with their modern equivalents. As expected, most national parks we classified as being closed habitats (Fig. 4.3A to 4.3D) had a pattern similar to that habitat represented in

Fig. 4.1 except for the following exceptions. All the Tasmania national parks (Ben

Lomond, Douglas-Apsley, Mount Field and Savage River NPs) have a gap between log body mass of 2 and 3 (100 to 1000 grams). Several other parks arguably have such a gap (Abercrombie, Kosciuszko, Ku-Ring-Gai Chase, Mount Barney, Mount Buffalo,

Mount Remarkable, Wadbilliga NPs). In the case of Abercrombie and Mount

Remarkable NPs, there are very few mammal species less than 1000 grams (3-4 species).

Semi-open habitats (Fig. 4.3D and 4.3E) were expected to resemble the pattern shown by closed/arid or open/humid cenograms (Fig. 4.1). No closed/arid patterns were found among Australian cenograms. However, the pattern displayed by Australian parks classified as semi-open habitat was very similar to Legendre‘s open/humid pattern (Fig.

4.1). There appears to be little difference between the open/humid and open/arid patterns, but all semi-open habitats display had a distinct gap varying from log body mass of 2 to 3, except for Fitzgerald River NP which showed no distinct gap. Open habitats patterns (Fig. 4.3F) were very similar to Legendre‘s open/arid patterns (Fig.

4.1).

96

Figure 4.3A. Cenograms of the mammalian faunas of eight National Parks classified as closed habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

97

Figure 4.3B. Cenograms of the mammalian faunas of eight National Parks classified as closed habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

98

Figure 4.3C. Cenograms of the mammalian faunas of eight National Parks classified as closed habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

99

Figure 4.3D. Cenograms of the mammalian faunas of eight National Parks classified as closed and semi-open habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

100

Figure 4.3E. Cenograms of the mammalian faunas of eight National Parks classified as semi-open habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

101

Figure 4.3F. Cenograms of the mammalian faunas of eight National Parks classified as open habitats. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

102

Figure 4.3G. Modern and historical cenograms of the mammalian faunas of four National Parks. Light grey circles represent large introduced herbivores; dark grey squares represent large carnivorous species; light grey triangles represent extinct species; black diamonds represent all other mammal species.

103 Cenogram patterns for the modern mammalian faunas of Karijini, Mutawintji,

Stirling Range and Uluru NPs show patterns (Fig. 4.3G) expected for their current habitat type (open, semi-open, closed and open respectively; Fig. 4.1). Historically, however, their patterns were different even though most environmental factors have not changed (i.e. rainfall, climate, major vegetation group and temperatures can be assumed as homologous to conditions present today). Karijini NP‘s historical pattern was closer to Legendre (1989)‘s open/humid pattern while Mutawintji‘s and Uluru‘s historical patterns more closely resembled a closed/humid pattern. The only exception among these is Stirling Range, whose historical pattern, although losing some taxa, still closely resembles the modern pattern.

Figure 4.4. Plot of the largest gap magnitude versus its relative position on the cenogram for each of the National Parks, divided into the three habitat types, closed, semi-open and open.

When a gap occurs in Australian cenograms, it appears to consistently lie between log 2 and 3 (i.e. 100 to 1000 grams) sometimes extending to log 1.5 or 3.5 (i.e.

104 50 to 5000 grams). Legendre‘s (1989) gap was predominantly found between 500 and

8000 grams (log 2.5 to 3.8). For each of the Australian national parks and reserves we plotted the magnitude of the largest gap and its position on the cenogram (Fig. 4.4). This graph clearly demonstrates that the gaps of largest magnitude were almost always features of parks that we classified as open or semi-open habitats, and all were between log 1.5 and 3.5. The majority (all except three) were between log 2 and 3, shifting the gaps in Australian cenograms to lower body masses.

We plotted the total number of species in our three habitat types in three body weight categories (Fig. 4.5): our new observed Australian cenogram gap (100 to 1000 grams); Legendre‘s (1989) gap (500 to 8000 grams); and Australia‘s critical weight range (35 to 5500 grams) identified by Short and Smith (1994). In all three categories, there is significant overlap between semi-open and open habitats, with semi-open habitats having a lower mean number of species than open habitats for the weight ranges 100 to 1000 grams and 35 to 5500 grams. The lower end of the closed habitats also overlaps with the upper end of both semi-open and open habitats for all body weight ranges. However, the body weight range 100 to 1000 grams is the only one that can be described as a true gap, having the lowest mean number of taxa and the closest to zero (with a large number of parks having no taxa in that range).

105

Figure 4.5. Box plot of the number of mammal species in the size gaps 100-1000g, 500- 8000g and 35-5500g, in each of the three habitat types, closed, semi-open and open.

For our investigation of the presence of —arboreal“ taxa in habitat types (Fig.

4.6) we included only two body weight ranges, 100 to 1000 grams and 1000 to 10000 grams, (these ranges were shown to strongly correlate with habitat types in Kendall‘s R

106 results.), which we will refer herein as medium-sized and large respectively. For closed habitats, all parks contained both medium-sized and large arboreal taxa, except for

Stirling Range and Prince Regent River NPs which included only large arboreal taxa.

Most semi-open habitats included large but no medium-sized arboreal taxa. The exceptions were Mutawintji, Mungo and Gawlers Range NPS which contained no arboreal taxa, and Gregory NP which had a medium-sized taxon but no large arboreal taxa. Open habitats generally had no arboreal taxa, with the exception of Kalbarri,

Karijini (modern and historical), Purnululu and historical Uluru NPs. Large non- arboreal taxa were always presents in all habitats, but medium non-arboreal taxa were not always present in semi-open and open habitats. Overall, the patterns in presence of arboreal taxa in Australian habitats were very close to those expected.

In Fig. 4.7, we represented the geographical distribution of the body mass distributions of each national park and reserve. Each bar graph shows the percentage of taxa in each of the body size categories selected (i.e. very small: 0 to 10 grams; small:

10 to 100 grams; medium: 100 to 1000 grams; large: 1000 to 10000 grams; and very large: 10000 to 100000 grams). Figure 4.8 shows the overall body mass distribution of

Australian and New Guinean mammalian faunas. Australia‘s body mass distribution is characterised by having the greatest proportion of its taxa in the small size range, followed by large, medium-sized, very large and very small. The body mass distribution of all mammal species for Australia (except large introduced herbivores) represents a mix of habitats ranging from rainforest to desert. We would expect that a similar distribution pattern found at the scale of a national park would mean mixed habitats.

The body mass distribution pattern for the closed forest of New Guinea is very different from Australia‘s pattern. The majority of taxa are medium-sized, followed by large,

107 then small, with no very large or very small mammals. In Fig. 4.7, only the rainforests of north-eastern Queensland (Iron Range and Mungkan Kandju NPs) and Deua NP have the same peak in medium-sized mammals as the New Guinea distribution. All

Australian parks on the eastern, south-eastern and south-western coasts (all categorised as closed habitats in this study) had their highest distribution peak in large mammals

(i.e. 1 kilogram to 10 kilograms). In contrast, all central and western Australian parks

(categorised as open or semi-open) had their highest distribution peak in small mammals (i.e. 10 grams to 100 grams). Several individual parks had a distribution similar to Australia‘s overall body mass distribution pattern. These parks occur either in transitional areas between forest and grasslands (e.g. Fitzgerald River NP) or include a mix of forest and grassland (i.e. Nitmiluk and Kakadu NPs). Flinders Range, Mungo,

Gundabooka, Kinchega and Mutawintji NPs had a different body mass distribution pattern, with most of their species being large and very large, with very few if any medium, small and very small mammals. Fig. 4.9 summarises the four major patterns in body mass distribution observed in Australia. A peak in medium sized mammals characterises rainforests, a peak in the large mammals characterises open forests, peaks in both small and large (and sometimes also very large) characterise mixed habitats

(such as riparian woodlands) or transitional habitats, and a peak in small mammals characterises grasslands and deserts.

108

Figure 4.6. Bar graphs of the number of mammal species in each of the body mass categories 100-1000g and 1000-10000g for non-arboreal and arboreal species in each of the National Parks studied, and grouped into the three habitat types, closed, semi-open and open.

109

Figure 4.7. Bar graphs of the proportion of mammal species in each of the body mass categories of the National Parks used in this study, with each bar graph representing on the map of Australia the approximate geographical position of each National Park. Body mass categories are in Log body mass, from left to right, 0-0.9, 1-1.99, 2-2.99, 3-3.99, 4-4.99. Bar graphs are coloured to represent non-arboreal taxa in black and arboreal taxa in light grey.

Figure 4.8. Bar graphs of the proportion of mammal species in each of the body mass categories of Australia and New Guinea.

110

Figure 4.9. Bar graphs of the proportion of mammal species in each of the body mass categories of four main patterns identified for Rainforest, Temperate forest, riparian woodlands and grasslands and desert.

Fig. 4.10 plots the historical body mass distribution of Mutawintji and Uluru

NPs versus their current distribution (Stirling Range and Karijini NPs had the same historical distribution as their current distributions). Historical Mutawintji NP had a body mass distribution that resembles the distribution that observed for grasslands and deserts, displaying a peak in small mammals. Most mammals that went extinct at

Mutawintji were small, between 10 and 100 grams. In contrast, Uluru had very different story. In contrast, Uluru‘s historical distribution patterns resemble a mixed habitat distribution with medium-sized and large mammals subsequently becoming extinct.

111

Figure 4.10. Bar graphs of the proportion of mammal species in each of the body mass categories of Mutawintji and Uluru National Parks, comparing historical data versus modern data.

The results of Kendall‘s R to find correlations between our cenogram, body mass distribution, and arboreal taxa variables and environmental variables are shown in Table

4.2. The R coefficient varies between 1 (correlated) and -1 (inversely correlated), with values close to 0 showing no correlation. P-values less than 0.05 indicate that there is less than 5% chance no correlations. Unlike Rodríguez (1999), we found many significant correlations between our cenogram variables and environmental variables.

Magnitude of the largest gap correlated with all environmental variables except maximum (MAMT) and minimum (MAmT) temperatures, with the highest correlation being with annual rainfall. No correlations were found between the average position of the largest gap and any of the environmental variables. This was not unexpected because the average position of the largest gap in closed habitats revealed no particular pattern (see Fig. 4.4). The percentage of species in the very small (log 0-0.99 or 0 to 10 grams) and small (log 1-1.99 or 10 to 100 grams) body mass categories correlated with

112 all environmental variables. Correlations for these two body mass categories range from fairly low correlations (Kendall‘s R values of 0.25) with temperatures, and fairly high correlations (Kendall‘s R values of 0.57) with habitat type. Of all body mass categories, medium sized mammals (log 2-2.99 or 100 to 1000 grams) have the highest Kendall‘s R values (over 0.5) with all the environmental variables, except for temperatures which had low or no correlation. Large mammals (log 3-3.99 or 1000 to 10000 grams) correlated with all variables weakly, with temperatures having the highest Kendall‘s R values (-0.459 and -0.419). Annual rainfall was the only variable correlating (weakly) with very large mammals (log 4-4.99 or 10000 to 100000 grams). Very small arboreal species correlated only with temperatures and only very weakly. Small arboreal species correlated with all environmental variables and Kendall‘s R values were quite high for temperatures, MVG and habitat type. The highest Kendall‘s R values in the analysis were for correlations between medium-sized arboreal species and habitat type (-0.753) and MVG (-0.607). Medium-sized arboreal species also correlated with all other environmental variables. Similar results, but with lower Kendall‘s R values, were found for large arboreal species. Overall, the results of Kendall‘s R analysis showed strong correlations between our cenogram, body mass distribution and arboreal taxa variables versus environmental variables, giving statistical support to the patterns examined visually.

113 Table 4.2. Kendall‘s R coefficient and probability (p-value) of uncorrelated pairs between cenogram, body mass distribution and arboreal taxa variables and environmental variables. MVG, MAMT and MAmT stands for major vegetation groups, mean annual maximum temperature and mean annual minimum temperature, respectively.

4.5 DISCUSSION

4.5.1 Limitations

The first step of any interpretation of statistical results should be the identification of possible errors (Sand-Jensen, 2007). Limitations inherent in the data used in our analyses and their possible effects on our results and their interpretation are as follow. First, the data used in all cenogram and body mass distribution studies, both by ourselves and other authors (e.g. Legendre, 1986; 1989; Rodríguez, 1999) is derived from national parks and reserves. These data are very useful for ecological studies: they are delimited in space, and the fauna and flora are monitored and studied. Rarely are there data of this quality available outside national parks. However, data from national parks are subject to several sources of error, such that they may not be truly representative of patterns between habitats (Louys et al., unpublished manuscript).

Ideally comparisons would be between data from uniform habitats of the exactly the

114 same size, with homogenous vegetation and equal sampling effort. However, in reality each national park is of different size, contains usually more than one habitat, has heterogeneous vegetation and unequal sampling effort. This may in fact not be a problem when patterns found using national parks are compared to fossil sites. Fossil sites are certainly collecting fossils from areas of different sizes, may collect fossils from different habitats (although one may be dominant), and taphonomic biases can be considered as an analogue to modern sampling effort.

In our own analysis, we have identified several sources of error. Firstly, we know that our national park data has not been uniformly sampled. Some parks have had decades of intense sampling effort, while others have had only a few years of sampling effort. Sampling effort information is not readily available, and therefore we can only rely on observation of the raw data to identify possible sampling biases. As yet, there have been very few statistical treatments of this problem (Louys et al., unpublished manuscript). In our study, we suspect that Abercrombie and Mount Remarkable NPs are the most heavily biased in terms of sampling effort because small mammals are poorly represented.

The second potential source of error in our study is the identification of vegetation type/habitat type. We used a single vegetation map to minimise the problem of inconsistencies in vegetation identification, which can frequently occur when these are sampled across several countries (Lawesson, 1994; Louys et al., in prep).

Nevertheless, vegetation maps have been recognized to have several problems, including their robusticity (they do not account for spatial and temporal variability) and generality (simplification of complex interactions between vegetation, climate, fire and

115 grazing) that can potentially be misleading for ecological studies (Bastin and Ludwig,

2006). Further, national parks that contain only one major vegetation group are rare. In our study, Kakadu National Park, at 1,980,400 square km the largest National Park in

Australia, contains at least five different major vegetation types, ranging from rainforest to grasslands. Current statistical analyses require that only one vegetation type describes the data from a single park. This is one possible source of error in Rodríguez‘s (1999) results, as he assigned one vegetation type per park, which can lead to further errors if the vegetation type was not assigned properly. So how does one assign a vegetation type? If, for example, a park contains 10% forest and 90% grassland, that the park would in all likelihood be assigned to grassland, however grasslands may not be the dominant pattern. Using the cenograms patterns, if you were to overlap the pattern of a closed forest with a desert, the resulting pattern would still look like a closed forest, because the medium-sized mammals of the forest would fill in the gap in the desert cenogram. Therefore, in the cenogram method, closed habitats have a dominant pattern over open habitats. In the case of the previous example, the hypothetical park with 10% forest and 90% grassland, using the cenogram method will not assign it to grasslands, as this will not be the pattern observed. As an example, Kakadu National Park displays a closed habitat pattern, yet a large portion of the park is actually grasslands. In our study, we chose to simplify the vegetation types to three habitat types (closed, semi-open and open) because we wanted to minimise error caused by incorrectly assigning a park to the wrong vegetation types. Our results from the Kendall‘s R analysis do show that our assignment of a park to the three habitat types works very well, as they showed the highest correlation with mammalian body mass patterns. However, this does not imply that we assigned correctly all our park to the correct habitat type, and it is possible that we did incorrectly assigned some parks.

116

Assuming that we have correctly assigned most national parks to habitat type, two major points can be made about our cenograms. Firstly, they are not infallible identifiers of habitats in Australian, and secondly, it is often very difficult to identify a pattern visually when the slopes of small or large mammals are made of only a few data points. This method was considered amongst the best methods to infer habitat types along with body mass distributions (Hernández Fernández et al., 2006). However, there are definite inconsistencies in the way cenograms, as defined by Legendre (1986; 1989), have been used in the past by many authors. The weight range of the gap varies depending on the author: for example between 500 grams and 8 kg (Legendre, 1986,

1988, 1989; Legendre and Hartenberger, 1992; Gunnell, 1994; Legendre, 1995; Morgan et al., 1995; Gunnell, 1997; Montuire and Desclaux, 1997; Montuire, 1999; Croft, 2001;

Montuire and Marcolini, 2002; Costeur, 2005; Tsubamoto et al., 2005; Tougard and

Montuire, 2006), 1 and 6 kg (de Bonis et al., 1992), 500 grams and 6 kg (Legendre,

1987b; Ducrocq et al., 1994), 500 grams and 10 kg (Legendre et al., 1997; Wilf et al.,

1998), 500 grams and 25 kg (Gunnell and Bartels, 1994), and at 500 grams (Gingerich,

1989; Maas and Krause, 1994). Similar inconsistencies are also found with the interpretation of the slopes of small and large mammals by the same authors. Despite those inconsistencies in their use, cenograms have consistently been useful palaeoecological reconstructions. Montuire (2000) pointed out that cenograms are only meant to be used to identify whether a fauna is closed or open, and are not accurate predictors of temperatures.

117 4.5.2 Australia‘s shifted gap

In our study, we identified a significant shift in the position of the gap in open habitats between 100 grams to 1000 grams, compared to gap between 500 grams to

8000 grams used by most researchers. This could be partly explained by the inclusion of carnivorous species in our study, which are usually excluded from these types of analyses. Currently, Australia counts only eight large carnivorous species (all of them in the weight range of 500 grams to 8000 grams), the dingo (Canis lupus dingo), the fox

(Vulpes vulpes), the cat (Felis catus), the Tasmanian devil (Sarcophilus harrisii) and the four species of quolls (Dasyurus maculates, Dasyurus viverrinus, Dasyurus geoffroii and Dasyurus hallucatus). However, even if carnivorous species were excluded from cenograms, a maximum of four of those species are found in a single park, and would therefore not account for the fact that the gap would still be shifted down, with the minimum weight being 100 grams instead of 500 grams. Holling‘s

(1992) —Core-Taxa Hypothesis“ (Siemann and Brown, 1999) predicts that the greatest gaps should occur at different body sizes in structurally similar biomes on different continents but at similar sizes in structurally different biomes on the same continent.

Our results do show that Australia does support part of this hypothesis, having its greatest gap at a different position to those of other continents (sensu Siemann and

Brown, 1999), but do not support the other part of the hypothesis, as our closed habitats had their largest gap at different body sizes. In addition, Australia is, as far as we know, the only place in the world which greatest cenogramic gap is different from the rest of the world. Siemann and Brown (1999) accepted Holling‘s (1992) —Core-Taxa

Hypothesis“ only based on comparisons of body mass distributions between Australia and North America and rejected the Textural-Discontinuity Hypothesis which linked patterns of vegetation and mammalian body size. Textural-Discontinuity Hypothesis

118 (Holling, 1992) predicts the exact reverse to the —Core-Taxa Hypothesis“. Siemann and

Brown (1999) concluded that patterns in body mass distributions were the results of geographical ranges and the history of phylogenetic radiations. Our results do not agree with their conclusions as cenogram patterns do reflect the closed/open patterns of the vegetations. Furthermore, the results of the Kendall‘s R correlations showed that body mass distributions correlated with most environmental variables (rainfall, climate… ). It should be expected that Australia would be different from the rest of the world as it remains the only place where the mammalian fauna is composed largely of marsupials, however the patterns found in Australia are extremely similar to the rest of the world, showing only a shift down for the greatest gap of open habitats. Our results therefore support the Textural-Discontinuity Hypothesis and reject the —Core-Taxa Hypothesis“, suggesting that body mass distributions and cenograms reflect patterns in vegetation.

4.5.3 Explaining the gap

Even though most of our closed habitats exhibited a continuous distribution of mammalian body sizes, some showed a small gap in the medium-sized range. Small gaps in body weight distributions are present in all Tasmanian parks (Ben Lomond,

Douglas-Apsley, Mount Field and Savage River NPs). Mammals filling this gap on the mainland are not present in Tasmania and, indeed there are no Tasmanian mammal species (Strahan, 1995) that could fill this gap. Since European arrival over 200 years ago, only the Tasmanian tiger (Thylacinus cynocephalus) is known to have gone extinct in Tasmania (Strahan, 1995). At this point, the cause of the medium-sized mammal gap in Tasmania is unknown. The last land bridge occurred about 12000-13000 years ago

(Alexander, 2005) possibly providing the opportunity for more medium-sized mammals to colonise Tasmania. However, land bridges between Tasmania and the mainland

119 occurred during glacial maxima that perhaps mitigated against successful establishment of Tasmanian population. This hypothesis can only be tested by an examination of the fossil record of Tasmania. Burbidge et al. (1997)‘s study on Australian island species richness demonstrated that the size of the island is highly correlated to its species richness. The gap could therefore be the response to an —island effect“. A similar gap is also found in some mainland national parks (Abercrombie, Kosciuszko, Ku-Ring-Gai

Chase, Mount Barney, Mount Buffalo, Mount Remarkable, Wadbilliga NPs).

Considering that all other parks adjacent to them had a continuous size distribution, two possible explanations can be hypothesised: medium-sized mammals were less well sampled than other mammal sizes, or medium-sized mammals have been well sampled but they are locally going extinct. In the latter case, these absences/extinctions could be the result of habitat fragmentation and edge effects, in turn due to the cleared lands

(urban and agricultural areas) found outside of national park boundaries. Habitat fragmentation and edge effects have been shown to have a serious impact on mammalian communities, resulting in the loss of species or decline in population size due to increasing competition for food and predation (Andrén, 1994; Abensperg-traun et al., 1996; Turner, 1996; Williams and Pearson, 1997; Lidicker, 1999; Debinski and

Holt, 2000; Schneider, 2001). Legendre (1989) did identify an edge effect (calling it an island effect) from the fauna of Transvaal (type 10), a small tropical forest surrounded by open areas. This fauna was identified as being too small to support a forest community, hence showing an open pattern, reflecting the community outside of the forest. The inverse of this pattern was identified in the faunas of Lamto (Ivory Coast) and Caatinga (Brazil) being mixes of closed and open habitats (perhaps similar to that of

Kakadu NP) or open habitats surrounded by closed habitats (not observed in our study).

120 Although Legendre (1989) identified these effects, until now they have not been recognised as independent cenogram patterns themselves.

The cenogram method becomes less credible if both island and edge effects create gaps in the cenograms of closed habitats. In both cases, a closed habitat could be mistakenly identified as an open habitat. In addition, historical data from Uluru and

Mutawintji National Parks show no distinct gaps in either cenogram. In this case, open habitats could be mistakenly identified as closed habitats. Legendre (1986; 1989) only identified aridity and openness as the only environmental —pressures“ affecting cenograms patterns. In our study, we can identify two new environmental —pressures“: pressure caused by island effect, habitat fragmentation and edge effect, and lack of pressure caused by the isolation of Australia for millions of years, resulting in lower predator numbers than the rest of the world (Wroe et al., 2004). When dingoes were introduced by humans to Australia around 3500-4000 years ago (Strahan, 1995), the largest carnivore on the mainland were the Tasmanian tiger (Thylacinus cynocephalus), the Tasmanian devil (Sarcophilus harrisii) and the four quoll species (Dasyurus maculates, Dasyurus viverrinus, Dasyurus geoffroii and Dasyurus hallucatus).At that time, Australia was already lacking carnivorous mammals over 50 kg, which are found on all other continents (Wroe et al., 2004). Following the introduction of dingoes to the mainland, both the Tasmanian tiger and Tasmanian devil went extinct from the mainland, diminishing the overall number of large predators (Johnson and Wroe, 2003).

The introduction of cats and foxes by Europeans increased predation rates, bringing in a new predatory pressure on the Australian mammal fauna.

121 4.5.4 Revision of the cenogram method

Because of these two new pressures (island, edge and fragmentation effects and increased predation), the cenogram method becomes less accurate for infering habitat types, perhaps explaining why Rodríguez (1999)‘s results showed very little correlation between cenogram patterns and environmental variables (using both historical and modern data, and possibly data with island or edge effects) and why Siemann and

Brown (1999) found few or no statistically significant gaps (using mainly historical data). However, used in combination with other methods, an alternative use of the cenogram method may in fact be very informative. In this study, we examined two other ways to infer habitat type: using arboreal species (possums and koalas) and using the percentage of mammals in each of the selected body mass categories. We found that closed habitats contain arboreal species of all sizes (large, medium and small) while semi-open habitats contain only large arboreal species and open habitats contain no arboreal species. With decreasing numbers of trees, there is more competition for food and less shelter from predators, and this helps explain the absence of arboreal species in open areas. The common brushtail possum (Trichosurus vulpecula) is the only arboreal species whose geographical range extends into more open areas, and historically, its geographical range covered most of Australia, even into the arid zone provided trees were present (Strahan, 1995). They are now restricted to forests and woodlands only.

We have also identified patterns in body mass distributions using a bar graph of the percentage of mammals in each of the selected categories (logarithmic mass in grams of the following ranges: 0-0.99, 1-1.99, 2-2.99, 3-3.99 and 4-4.99). The patterns identified, using this method, are able to infer habitat type where cenograms may be misleading. For example, all parks showing a gap in medium-sized mammals due to

122 island or edge effect, showed the same pattern in body mass distribution as any other closed habitat, in having a peak in the large sized mammals. Similarly, the cenogram of historical Mutawintji National Park was a straight line, but the body mass distribution showed a peak in the small sized mammals, characterising open habitats. Overall, the presence of arboreal species presence and the proportion of body mass categories are most likely better methods to infer habitat type, but neither of these methods is able to visualise pressures on mammalian communities in the way that the cenogram method does, by showing the magnitude of the gap in medium-sized mammals. Using all three methods in conjunction should allow accurate identification of habitats as well as visualisation of pressures caused by diverse ecological phenomenon. Fig. 4.11 gives a revised version of the cenogram patterns proposed by Legendre (1986; 1989), including the presence of arboreal taxa. We propose that these patterns should be used, at least for

Australian habitat type identification, in conjunction with the patterns of body mass distribution of Fig. 4.9.

123

Figure 4.11. New cenogram pattern model hypothesised in this study for Australian habitats. Trees represent forests or woodlands, grasses represent shrublands, grasslands and deserts, waves represent islands and possums represent arboreal species.

4.5.5 The impact of introduced predators

Special attention should also be given to some of the patterns found using the body mass distribution method (also seen in cenograms). Mutawintji, Currawinya,

Mungo, Kinchega, Gundabooka, Flinders Range, Bookmark, Gawlers Range, Little

Desert and Yumbarra NPs, all share one pattern in common which separates them from other semi-open/open habitats. They have a relatively low number of small mammal species (or low proportion compared to large mammals) compared to a high number of small mammal species found everywhere else in the arid and semi-arid zone of

Australia. Three possible explanations for this pattern are: sampling for small mammals in those parks was poor; the pattern may be caused by an environment that does not favour small mammals; or it is the result of another unidentified phenomenon.

Historical data from Mutawintji National Park help eliminate two of these explanations.

124 First, Mutawintji NP historically contained many small mammals (12 small and very small species versus only four currently found at Mutawintji NP), eliminating the explanation of poor sampling. Second, if this pattern was caused by an environment unfavourable to small mammals then no changes would be found between historical and modern data. This suggests that the pattern is the result of another phenomenon or

—pressure“ that has resulted in a different pattern in body mass distribution to the other open habitats. Having fewer small species, the cenograms of these ten parks show a larger gap extending into the small mammals, shifting downwards the gap from 1000 grams to 50 grams.

Valverde (1964; 1967) hypothesised that the gap in medium-sized mammals was the result of predator-prey relationships. If this relationship between gap size and gap position is correct, then the position of the gap will correlate with the size of the predators, and the magnitude should reflect the degree of predation. A shift in the gap to smaller sizes should therefore imply that the main predators are also smaller.

Geographically, all the parks showing this pattern are located in the area where dingoes have been excluded by the dingo proof fence (Fig. 4.2). It has been hypothesised that top predators such as the dingo may control smaller predator populations, such as foxes and cats (Risbey et al., 2000; Glen and Dickman, 2005; Johnson et al., 2007). The observed pattern may be the result of the removal of dingoes and increases in the populations of foxes and cats and increasing predation on small and medium-sized mammals. The geographical distributions of sheep and rabbits have also been shown to be correlated with areas of high extinctions, caused by the depletion of food and ground-level shelter by these two species (Burbidge and McKenzie, 1989; Morton,

1990; Short and Smith, 1994; Smith and Quin, 1996; Fisher et al., 2003; Glen and

125 Dickman, 2005; Johnson et al., 2007). The two different patterns observed for semi- open/open habitats in our study may be the result of the combination of presence and absence of dingoes, sheep and rabbits.

4.5.6 Mid-domain effect

The other noteworthy, and cautionary, pattern found in our analyses is that based on historical data from Uluru National Park. Uluru‘s pattern in body mass distribution

(Fig. 4.10) was very similar to that of the overall pattern for the Australian continent today (Fig. 4.8). This pattern might be expected in parks with mixed closed and open habitats, but Uluru‘s habitat is exclusively open. The pattern could be explained by the mid-domain effect (Colwell and Lees, 2000). This model (still being debated Davies et al., 2005; McClain and Etter, 2005) predicts that centrally located areas tend to have higher species richness than other areas because they share common species with all other areas, thus increasing the overall species richness of that area. Historically, many

Australia mammal species‘ distributions extended from west to east, or north to south, meeting in the centre of the continent. Today, most of these very broad distributions have been disrupted and reduced to small isolated areas, with many taxa now locally extinct at Uluru.

4.6 CONCLUSION

This study suggests that the current cenogram methodology cannot reliably infer habitat types from cenogram patterns. However, with new cenogram patterns for modern and historical habitats identified and used in conjunction with two other measures described here (presence of arboreal taxa and body mass distribution),

126 cenogram methodology should prove to be a useful ecological tool. Cenograms are able to visualise pressures causes by several well-recognised ecological phenomena. These pressures can significantly affect the patterns of the cenograms and body mass distributions, and include island effect, habitat fragmentation, edge effect, predator removal (viz. Australian dingo-proof fence) and introduced species (cat, fox, sheep and rabbit). The gap in cenograms is the result of these pressures and its position is correlated with predator size. Australia has smaller predators than the rest of the world and there is a resulting shift downwards in the position of the cenogram gap. Historical data indicates that Australia was under very little environmental pressure before

European arrival, with no distinct gaps in the cenograms of open habitats, evidently reflecting Australia‘s isolation from other continents for millions of years and its lack of large carnivores. Cenogram methodology should prove to be an increasing useful tool both to infer habitat type in Australia for fossil communities and identify pressures caused by diverse pressures on the mammalian community.

4.7 REFERENCES

Abensperg-traun, M., G. T. Smith, G. W. Arnold and D. E. Steven (1996). The effects of habitat fragmentation and livestock-grazing on animal communities in remnants of gimlet Eucalyptus salubris woodland in the Western Australian wheatbelt. I. Arthropods. Journal of Applied Ecology 33: 1281-1301.

Alexander, A., Ed. (2005). The Companion to Tasmanian History. Centre for Tasmanian Historical Studies, University of Tasmania, Hobart.

Allen, C. R., A. S. Garmestani, T. D. Havlicek, P. A. Marquet, G. D. Peterson, C. Restrepo, C. A. Stow and B. E. Weeks (2006). Patterns in body mass distributions: sifting among alternative hypotheses. Ecology Letters 9: 630-645.

Andrén, H. (1994). Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71: 355-366.

Anonymous (1993). Douglas - Apsley National Park Management Plan, Parks and Wildlife Service, Tasmania, Department of Environment and Land

127 Management.http://www.parks.tas.gov.au/publications/tech/douglas_apsley/douglas_ap sley.pdf

Anonymous (1997). Appendix 1: Species for Calperum Station - Bookmark Biosphere Reserve (Australia).

Anonymous (1998a). Mammals of Croajingolong National Park. , Printed by Parks Victoria, Australia

Anonymous (1998b). Ben Lomond National Park Management Plan, Parks and Wildlife Service, Tasmania, Department of Primary Industries, Water and Environment.http://www.parks.tas.gov.au/publications/tech/ben_lomond/ben_lomond.p df

Anonymous (1999a). Shoalwater and Corio Bays Area Information Sheet on Ramsar Wetlands presented in application for Ramsar designation, Ramsar Wetlands

Anonymous (1999b). Kakadu National Park Plan of Management, Kakadu Board of Management Parks Australia.http://www.environment.gov.au/parks/publications/pubs/kakplan4.pdf

Anonymous (2001a). Gregory National Park Draft Plan of Management, Parks and Wildlife Commission of the Northern Territory.http://www.nt.gov.au/nreta/parks/management/pdf/gregorypom.pdf

Anonymous (2001b). Savage River National Park and Savage River Regional Reserve Draft Management Plan, Parks and Wildlife Service, Tasmania, Department of Primary Industries, Water and Environment.http://www.parks.tas.gov.au/publications/tech/savage/savage.pdf

Anonymous (2002a). Nitmiluk (Katherine Gorge) National Park Plan of Management, Parks and Wildlife Commission of the Northern Territory.http://www.nt.gov.au/nreta/parks/management/pdf/nitmiluk_cover_foreword_ contents.pdf

Anonymous (2002b). Mount Field National Park, Marriotts Falls State Reserve & Junee Cave State Reserve Management Plan Parks and Wildlife Service, Tasmania, Department of Tourism, Parks, Heritage and the Arts.http://www.parks.tas.gov.au/publications/tech/mtfield_management_plan/mtfield.p df

Balding, G. (2004). Animals: a checklist of the vertebrate fauna of Uluru-Kata Tjuta National Park, Department of the Environment and Heritage

Bassarova, M. (2005). Taphonomic and palaeoecological investigations of Riversleigh Oligo-Miocene fossil sites. PhD, UNSW, Sydney.

Bastin, G. N. and J. A. Ludwig (2006). Problems and prospects for mapping vegetation condition in Australia's arid rangelands. Ecological Management & Restoration 7(S1): S71-74.

128

Burbidge, A. A. and N. L. McKenzie (1989). Patterns in the modern decline of western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation 50: 143-198.

Burbidge, A. A., M. R. Williams and I. Abbott (1997). Mammals of Australian Islands: Factors Influencing Species Richness. Journal of Biogeography 24(6): 703-715.

Chapman, A. (1995). A biological survey of the Fitzgerald area, Western Australia. Part 6: Terrestrial mammals. CALMScience Supplement 3: 83-94.

Colwell, R. K. and D. C. Lees (2000). The mid-domain effect: geometric constraints on the geography of species richness. Trends in Ecology and Evolution 15: 70-76.

Costeur, L. (2005). Cenogram analysis of the Rudabánya mammalian community: palaeoenvironmental interpretations. Palaeontographia Italica 90: 303-307.

Croft, D. A. (2001). Cenozoic environmental change in South America as indicated by mammalian body size distributions (cenograms). Diversity and Distributions 7: 271- 287.

Davies, T. J., R. Grenyer and J. L. Gittleman (2005). Phylogeny can make the mid- domain effect an inappropriate null model. Biological letters 1: 143-146. de Bonis, L., G. Bouvrain, D. Geraads and G. Koufos (1992). Diversity and paleoecology of Greek late Miocene mammalian faunas. Palaeogeography, Palaeoclimatology, Palaeoecology 91: 99-121.

Debinski, D. M. and R. D. Holt (2000). A survey and overview of habitat fragmentation experiments. Conservation Biology 14(2): 342-355.

Ducrocq, S., Y. Chaimanee, V. Duteethorn and J.-J. Jaeger (1994). Ages and Paleoenvironment of Miocene mammalian faunas from Thailand. Palaeogeography, Palaeoclimatology, Palaeoecology 108: 149-163.

Fisher, D. O., S. P. Blomberg and I. P. F. Owens (2003). Extrinsic versus intrinsic factors in the decline and extinction of Australian marsupials. Proceedings of the Royal Society of London B 270: 1801-1808.

Flannery, T. (1995). Mammals of New Guinea. Reed Books, Australia

Gingerich, P. D. (1989). New earliest Wasatchian mammalian Fauna from the Eocene of Northwestern Wyoming: composition and diversity in a rarely sampled High- Floodplain assemblage. Papers in Palaeontology. University of Michigan 28: 1-97.

Glen, A. S. and C. R. Dickman (2005). Complex interactions among mammalian carnivores in Australia, and their implications for wildlife management. Biological Reviews 80: 387-401.

129 Gómez Cano, A. R., B. A. García Yelo and M. Hernández Fernández (2006). Cenogramas, análisis bioclimático y muestreo en faunas de mamíferos: implicaciones para la aplicación de métodos de análisis paleoecológico. . Estudios Geológicos 62: 135-144.

Gunnell, G. F. (1994). Paleocene mammals and faunal analysis of the Chappo type locality (Tiffanian), Green River Basin, Wyoming. Journal of Vertebrate Palaeontology 14: 81-104.

Gunnell, G. F. (1997). Wasatchian-Bridgerian (Eocene) paleoecology of the western interior of North America: changing paleoenvironments and taxonomic composition of omomyid (Tarsiiformes) primates. Journal of Human Evolution 32: 105-132.

Gunnell, G. F. and W. S. Bartels (1994). Early Bridgerian (middle Eocene) vertebrate paleontology and paleoecology of the southern Green River basin, Wyoming. Contributions to Geology, University of Wyoming 30: 57-70.

Herford, I., K. Gillen, M. Lloyd, C. Hine, L. McCaw, G. Keighery and J. Allan (1999). Stirling Range and Porongurup National Parks Management Plan, Department of Conservation and Land Management.http://www.naturebase.net/pdf/nature/management/stirling- porongorup_nps.pdf

Hernández Fernández, M., M. T. Alberdi, B. Azanza, P. Montoya, J. Morales, M. Nieto and P. Peláez-Campomanes (2006). Identification problems of arid environments in the Neogene-Quaternary mammal record of Spain. Journal of Arid Environments 66: 585- 608.

Holling, C. S. (1992). Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62(4): 447-502.

Johnson, C. N., J. L. Isaac and D. O. Fisher (2007). Rarity of a top predator triggers continent-wide collapse of mammal prey: dingoes and marsupials in Australia. Proceedings of the Royal Society of London B 274: 341-346.

Johnson, C. N. and S. Wroe (2003). Causes of extinction of vertebrates during the Holocene of mainland Australia: arrival of the dingo, or human impact? The Holocene 13(6): 1009-1016.

Lawesson, J. E. (1994). Some comments on the Classification of African Vegetation. Journal of Vegetation Science 5(3): 441-444.

Legendre, S. (1986). Analysis of mammalian communities from the late Eocene and Oligocene of Southern France. Palaeovertebrata 16: 191-212.

Legendre, S. (1987). Les communautés de mammifères d'Europe occidentale de l'Eocene supérieur et Oligocène: structures et milieux. Münchner Geowissenschaftliche Abhandlungen (A) 10: 301-312.

130 Legendre, S. (1987b). Les communautés de mammifères et leurs milieux en Europe Occidentale de l'Eocene supérieur et l'Oligocène. Revue Paleobiologie 6: 1983-1988.

Legendre, S. (1988). Contributions à l'étude du gisement Miocene supérieur de Montredon (Hérault) les grands Mammifères. 8 - Analyse paléoécologique de la Faune Mammalienne. . Palaeovertebrata Memoire Extraordinaire: 177-186.

Legendre, S. (1989). Les communautés de mammifères du Paléogène (Eocène supérieur et Oligocène) d'Europe occidentale: structures, milieux et évolution. Münchner Geowissenschaftliche Abhandlungen (A) 16: 1-110.

Legendre, S. (1995). Le garouillas et les sites contemporaines (Oligocène, MP 25) des Phosphorites du Quercy (Lot, Tarn-et-Garonne, France) et leurs Faunes de vertébrés. 14. Paleoecology. Paleontographica Abt. (A) 236: 327-343.

Legendre, S., C. Badgley and P. D. Gingerich (1997). Mammifères, taille et climat: les faunes "locales" de mammifères sont-elles au plan de la taille une collection d'espèces prises au hasard dans l'intervalle souris-éléphant ou sont-elles des communautiés structurées par le milieu ? Geobios MS 21: 41.

Legendre, S. and J.-L. Hartenberger (1992). Evolution of Mammalian Faunas in Europe during the Eocene and Oligocene. Climatic and biotic evolution. D. R. Prothero and W. A. Berggren, Princeton University Press: 516-528.

Lidicker, W. Z. (1999). Responses of mammals to habitat edges: an overview. Landscape Ecology 14: 333-343.

Louys, J., K. J. Travouillon, M. Bassarova and H. Tong (in prep). The Use of Game and Nature Reserves, National Parks in Palaeoecological Analyses: Assumptions, Limitations and Application.

Maas, M. C. and D. W. Krause (1994). Mammalian turnover and community structure in the Paleocene of North America. Historical Biology 8: 91-128.

McCann, P. (1985). Mammals, Reptiles, Amphibia and Fishes of Grampians National Park., Deptartment of Conservation, Forests, and Lands, National Parks Service, Victoria

McClain, C. R. and R. J. Etter (2005). Mid-domain models as predictors of species diversity patterns: bathymetric diversity gradients in the deep sea. Oikos 109: 555-566.

Miles, J. M. and A. A. Burbidge (1975). A biological survey of the Prince Regent Reserve, North-West Kimberley, Western Australia. Wildlife Research Bulletin Western Australia 3: 1-116.

Montuire, S. (1999). Mammalian faunas as indicators of environmental and climatic changes in Spain during the Pliocene-Quaternary Transition. Quaternary Research 52: 129-137.

131 Montuire, S. (2000). Letter to the editor: Reply to Letter to the editor by J. Rodríguez. Quaternary Research 54: 436.

Montuire, S. and E. Desclaux (1997). Palaeoecological analysis of mammalian faunas and environmental evolution in South of France during the Pleistocene. Boreas 26: 355- 365.

Montuire, S. and F. Marcolini (2002). Palaeoenvironmental significance of the mammalian faunas of Italy since the Pliocene Journal of Quaternary Science 17(1): 87- 96.

Morgan, M. E., C. Badgley, G. F. Gunnell, P. D. Gingerich, J. W. Kappelman and M. C. Maas (1995). Comparative palaeoecology of Paleogene and Neogene mammalian faunas: body-size structure. Palaeogeography, Palaeoclimatology, Palaeoecology 115: 287-317.

Morton, S. R. (1990). The impact of European settlement on the vertebrate animals of arid Australia: a conceptual model. Proceedings of the Ecological Society of Australia 16: 201-213.

Myers, T. J. M. (2002). Palaeoecology of Oligo-Miocene Local Faunas from Riversleigh. University of New South Wales, Sydney.

Owens, H. M., T. J. Hudspith, A. C. Robinson, I. Dobrzinski, D. M. Armstrong, L. P. Pedler and P. J. Lang (1995). A biological survey of Yumbarra Conservation Park South Australia in March 1995, Department of Environment and Natural Resources, South Australia.http://deh.sa.gov.au/biodiversity/pdfs/biosurvey/yumbarra/yumbarra.pdf

Resources, D. o. t. E. a. W. (1999). EPBC Act List of Threatened Fauna.http://www.environment.gov.au/cgi- bin/sprat/public/publicthreatenedlist.pl?wanted=fauna

Risbey, D. A., M. C. Calver, J. Short, J. S. Bradley and I. W. Wright (2000). The impact of cats and foxes on the small vertebrate fauna of Heirisson Prong, Western Australia. II. A field experiment. Wildlife Research 27: 223-235.

Rodríguez, J. (1999). Use of cenograms in mammalian palaeocology. A critical review. Lethaia 32: 331-347.

Sand-Jensen, K. (2007). How to write consistently boring scientific literature. Oikos 116: 723-727.

Schneider, M. F. (2001). Habitat loss, fragmentation and predator impact: spatial implications for prey conservation. Journal of Applied Ecology 38: 720-735.

Schodde, R., P. C. Catling, M. I. J., G. C. Richards and J. C. Wombey (1992). The Land Vertebrate Fauna of the Shoalwater Bay Training Area, Queensland, CSIRO Division of Wildlife Ecology. Canberra.

132 Short, J. and A. Smith (1994). Mammal decline and recovery in Australia. Journal of Mammalogy 75(2): 288-297.

Siemann, E. and J. H. Brown (1999). Gaps in mammalian body size distributions reexamined. Ecology 80(8): 2788-2792.

Smith, A. P. and D. G. Quin (1996). Patterns and causes of extinction and decline in Australian conilurine rodents. Biological Conservation 77: 243-267.

Strahan, R. (1995). The mammals of Australia. Reed New Holand, Sydney

Tougard, C. and S. Montuire (2006). Pleistocene paleoenvironmental reconstructions and mammalian evolution in South-East Asia: focus on fossil faunas from Thailand. Quaternary Science Reviews 25: 126-141.

Tsubamoto, T., N. Egi, M. Takai, C. Sein and M. Maung (2005). Middle Eocene ungulate mammals from Myanmar: A review with description of new specimens. Acta Palaeontologica Polonica 50(1): 117-138.

Turner, I. M. (1996). Species loss in fragments of tropical rain forest: a review of the evidence. Journal of Applied Ecology 33(2): 200-209.

Valverde, J. A. (1964). Remarques sur la structure et l‘évolution des communautés de vertébrés terrestres. 1. Structure d‘une communauté 2. Rapport entre prédateurs et proies. La Terre et la Vie 111: 121œ154.

Valverde, J. A. (1967). Estructura de una communidad de vertebrados terrestres. Monografías de la Estación Biológica de Doñana 1: 1œ129.

Wilf, P., K. C. Beard, K. S. Davies-Vollum and J. W. Norekjo (1998). Portrait of a Late Paleocene (Early Clarkforkian) Terrestrial Ecosystem: Big Multi Quarry and Associated Strata, Washakie Basin, Southwestern Wyoming. Palaios 13: 514-532.

Williams, S. E. and R. G. Pearson (1997). Historical rainforest contractions, localized extinctions and patterns of vertebrate endemism in the rainforests of Australia's wet tropics. Proceedings of the Royal Society of London B 264: 709-716.

Woinarski, J., K. Menkhorst, N. Gambold and R. Braithwaite (1992). Mammals of the Bungle Bungle area. A Survey of the Wildlife and Vegetation of Purnululu (Bungle Bungle) National Park and Adjacent Area. J. C. Z. Woinarski. Perth, Department of Conservation and Land Management. Research Bulletin No. 6: 53-67.

Wroe, S., C. Argot and C. Dickman (2004). On the rarity of big fierce carnivores and primacy of isolation and area: tracking large mammalian carnivore diversity on two isolated continents. Proceedings of the Royal Society of London B 271: 1203-1211.

133

CHAPTER 5: PALAEOECOLOGICAL ANALYSES OF

RIVERSLEIGH‘S OLIGO-MIOCENE SITES:

IMPLICATIONS FOR OLIO-MIOCENE CLIMATE CHANGE

IN AUSTRALIA

134 5.1 ABSTRACT

During the Cenozoic, Australian environments changed from being dominated by warm and humid rainforests to mainly arid/semiarid habitats comparable to those found today. Northern Australia‘s Oligo-Miocene palaeoenvironment had been analysed using mammalian assemblages from Riversleigh, World Heritage Area, northwestern

Queensland. Limitations of fossil assemblages were minimised using MSR and taxonomic distinctness analyses because these identify under-sampled, unrepresentative and taxonomically-biased fossil assemblages which can then be removed from the overall analysis. Cenogram and body mass distribution methodologies are used to help determine the palaeohabitats of seventeen sites from Riversleigh which span the late

Oligocene to early Miocene, middle Miocene and early late Miocene. The results suggest a change of environment through time, with the late Oligocene of Riversleigh is represented by an open forest, the early Miocene is represented by rainforest, middle

Miocene is represented by rainforest and the early late Miocene is represented by open forest.

5.2 INTRODUCTION

Since the Cretaceous, Australia has undergone major climatic changes (Hill,

1994; Martin, 2006). At the beginning of the Cenozoic, Australia was covered with rainforests (Hill, 1994; Martin, 2006). The history of the Australian vegetation is derived from studies in palaeobotany and palynology (Martin, 1990, 1994, 1998, 2006).

However, the majority of sites yielding fossil plant and pollen remains are located in central or southern Australia (Martin, 2006). Our understanding of northern Australia‘s vegetation history is in part based on ”reflective‘ interpretation of the palaeovegetation

135 structure based on study of the structure of teeth and the habitats of modern relatives of the vertebrates found in the fossil deposits of northern Australia (Archer et al., 1994;

Hill, 1994).

The early Cenozoic mammalian record from Australia is not complete, with no mammal-bearing sites known for the Palaeocene, late Eocene or early Oligocene, and only a single site known for the early Eocene (Murgon). However, from the late

Oligocene onwards the mammalian record is relatively continuous (Woodburne et al.,

1985; Archer et al., 1989; Rich, 1991; Tedford et al., 1992; Woodburne et al., 1993;

Hill, 1994; Archer et al., 1997; Archer et al., 1999; Long et al., 2002; Megirian et al.,

2004). Interpretations of Oligo-Miocene climates in Australia are controversial (Archer et al., 1989; Megirian, 1992; McGowran and Li, 1994; Archer et al., 1997; Creaser,

1997; Megirian et al., 2004; Martin, 2006). Studies (McGowran, 1986; Frakes et al.,

1987; McGowran and Li, 1994) of marine foraminifera, sea level changes and rainfall have been used to infer a sequence of changing conditions from the late Oligocene

(relatively cool and dry icehouse conditions), to the early and middle Miocene

(relatively warm and wet greenhouse conditions) and the late Miocene (relatively cool and dry icehouse conditions). However, palaeobotanical and palynological studies

(Martin, 1990; Hill, 1994; Martin, 1994, 2006), while agreeing in large part with the conclusions just noted have also led to the different conclusion that the Oligocene was relatively wet. There is no evidence of grasslands until the Pliocene in Australia

(Martin, 1994; Archer et al., 1994b).

The interpretation of the climate of the late Oligocene using faunas from the

Northern Territory (i.e. Kangaroo Well Local Fauna), South Australia (Etadunna

136 Formation and Namba Formation), and northwestern Queensland (Riversleigh Faunal

Zone A) is controversial. Megirian (1992) interpreted Riversleigh‘s Carl Creek

Limestone as a calcilclastic alluvial floodplain, which could only have accumulated under relatively dry or semi-arid conditions. However, Creaser (1997) argued that calclithites are forming today in the wet rainforest environment on the Huon Terraces in

Papua New Guinea. Megirian et al. (2004) further argues that extant crocodilian and terrestrial gastropod analogs for taxa present in the Ulta Limestone (Kangaroo Well

Local Fauna), Etadunna Formation, Camfield beds (Bullock Creek Local Fauna) and

Riversleigh‘s Carl Creek Limestone indicate temperatures ranging from 14°C to 20°C, and median average rainfall under 600mm.. Megirian et al.‘s (2004) argument has yet to be challenged. The late Oligocene faunas from central as well as northern Australia contain few arboreal taxa relative to those of early to mid Miocene age from the same areas (Woodburne et al., 1985; Woodburne et al., 1993; Benbow et al., 1995; Archer et al., 1997; Megirian et al., 2004; Archer et al., 2006; Travouillon et al., 2006).

Associated plant macrofossils from Dunsinane Site (late Oligocene) at Riversleigh have been interpreted as either temperate open forest or woodland (Arena, 1997). However, further plant macrofossils have been recovered from Dunsinane Site recently (H.

Godthelp, pers. comm. 2008) such as Nothofagus which are normally rainforest trees and Burdekin Plums which are only known today from rainforest environments.

Recently, Arena (2008) provided evidence indicating the possibility that the plant and vertebrate fossils of Dunsinane Site are not the same age.

Early to Middle Miocene sites are represented by Riversleigh‘s Faunal Zone B

(Early Miocene) and Faunal Zone C (Middle Miocene), Bullock Creek Local Fauna

(Middle Miocene) and Kutjamarpu Local Fauna (See Woodburne et al., 1993; Archer et

137 al., 1997; Megirian et al., 2004; Travouillon et al., 2006). The palaeoenvironments of

Riversleigh‘s Faunal Zones B and C have been argued to be relatively warmer, wet, closed forest environments based on several lines of evidence: relatively high faunal diversity (Archer et al., 1989; Archer et al., 1997); high numbers of sympatric arboreal taxa (Archer et al., 1989; Archer et al., 1997; Myers et al., 2001; Roberts et al., 2007); and presence of rainforest-indicative groups such as Hypsiprymnodon species (Flannery and Archer, 1987; Bates, 2007), Dactylopsila possums (Archer et al., 1999), pseudocheirid possums (Archer et al., 2006; Roberts et al., 2007; Travouillon et al., in prep.), lyre-birds (Menura species) and log runners (Boles, 1993, 1995, 1997). Megirian et al. (2004) dismissed all of these reasons for interpreting these palaeoenvironments to have been wet, closed forests on the basis of the presence of a single species of snail in the D Site fauna (Faunal Zone A) which he suggests indicates a relatively dry environment. However, this gastropod is not found anywhere else at Riversleigh hence whatever the merits of this argument may be, they do not appear to apply to any sites other than D Site. Because D Site is a member of Faunal Zone A (Archer et al., 1997;

Travouillon et al., 2006) which is interpreted to be late Oligocene in age, its environment may be different to those of the early to middle Miocene assemblages from

Riversleigh which are interpreted to represent relatively warmer, wet closed forest environments. Currently, the palaeoenvironment of the middle Miocene Bullock Creek

Local Fauna has been subject to relatively less palaeoenvironmental analysis. However, at least the larger taxa in this fauna have been reasonably well-studied and more rigorous palaeoecological interpretations should be possible (Murray and Megirian,

1992, 2000; Murray et al., 2000; Schwartz and Megirian, 2004).

138 The late Miocene of northern Australia is represented by the Encore Local Fauna from Riversleigh and the Alcoota and Ongeva Local Faunas from the Northern

Territory. These faunas include the first mammals that appear to display adaptations

(e.g. taller crowns and thicker enamel in the teeth of herbivores and even hypselodonty in vombatids) to processing the more abrasive vegetation that would characterise drier habitats (Woodburne, 1967; Murray and Megirian, 1992; Megirian et al., 1996; Archer et al., 1997; Myers et al., 2001; Travouillon et al., 2006; Brewer et al., 2007).

In summary, the climate of Australia during the Oligo-Miocene remains a controversial topic. Palaeoecological studies of mammalian assemblages can provide robust evidence for palaeoclimatic conditions, and hence are potentially highly significant to this debate. The palaeoecological study, presented here uses data from

Riversleigh‘s Oligo-Miocene sites, because these sites cover a long time period and preserve a taxonomically highly diverse mammalian fauna. Furthermore, many of these sites have already been the subject of studies focusing on other aspects of palaeoecology

(Myers, 2002; Bassarova, 2005); the results of these studies can be compared with those presented here.

5.3 MATERIALS AND METHODS

5.3.1 Materials

Over 200 fossil-bearing localities have been found in the Riversleigh World

Heritage Area, 88 (including 84 Oligo-Miocene, 1 Pliocene, 1 Pleistocene and 2

Holocene fossil-bearing localities) of which have specimens identified to species

(Archer et al., 2006; Travouillon et al., 2006; Travouillon et al., in prep.). Specifically, this study concentrates on 14 localities (see Chapter 3) that have been shown to be of

139 palaeoecological significance through MSR (Minimum Sampling Richness) and

Taxonomic distinctness methods (Travouillon et al., 2007; Travouillon et al., in prep.).

Using these methods taxonomical biases in the species lists were identified (probably caused by taphonomic, collection or publication biases), as well as identifying the minimum number of taxa required in a locality to confidently differentiate it from another locality from a different time or environment (Travouillon et al., 2007;

Travouillon et al., in prep.).

The selected sites include: D Site (10 species); White Hunter Site (23 species);

Creaser‘s Rampart Site (19 species); Camel Sputum Site (43 species); Dirk‘s Tower Site

(31 species); Neville‘s Garden Site (35 species); Upper Site (41 species); Wayne‘s Wok

Site (39 species); Alan‘s Ledge 1990 Site (14 species); Cleft Of Ages Site (17 species);

Gag Site (35 species); Henk‘s Hollow Site (26 species); Last Minute Site (17 species);

Encore Site (20 species).

The species lists of these localities are taken from Archer et al. (2006), updated by Travouillon et al. (in prep.). Although the presence of a dasyurid, a possum

(petauroid?) and some peramelemorphians have been previously recorded in D Site

(Rich, 1991; Archer et al., 1994a), they have not been included in this study because there are no further taxonomic information and we have not been able to locate the specimens. The relative age of these localities is based on biocorrelation (Archer et al.,

1989; Archer et al., 1994a; Archer et al., 1997) and on multivariate analyses

(Travouillon et al., 2006; Travouillon et al., in prep.), separating the Oligo-Miocene localities into 4 separate Faunal Zones: A, B, C and D (Arena, 2004; Travouillon et al.,

2006). The selected localities have been assigned to the following Faunal Zones: D Site and White Hunter Site as Faunal Zone A, Creaser‘s Rampart, Camel Sputum, Dirk‘s

140 Tower, Neville‘s Garden, Upper and Wayne‘s Wok Sites as Faunal Zone B, Alan‘s

Ledge 1990, Cleft Of Ages, Gag, Henk‘s Hollow and Last Minute Sites as Faunal Zone

C, and Encore Site as Faunal Zone D.

5.3.2 Body mass estimate

The body weight of the fossil species is estimated using Myers‘ (2001) allometric relationships between cranial/dental measurements and body mass for predicting marsupial body mass. The highest ranked equation was selected to estimate body mass from one of four datasets (—Diprotodontians only“, —Dasyuromorphian species“, —all species excluding dasyuromorphians“ and —All-species“, sensu Myers,

2001) and depending on available measurements. For example, Barinya wangala‘s body size (dasyurid), known from a skull and several lower jaws, was estimated from the highest ranked equation in the dasyuromorphian species dataset, using measurements from the upper molar row length (UMRL). When more than one specimen is available, averages for the same measurement were used. In the case of a fossil locality missing the required information to estimate the body mass of its species (e.g. species represented by immeasurable teeth) weights were estimated from other contemporaneous fossil localities. These estimated body masses are listed for each species of each site in the appendix.

5.3.3 Cenogram and Body Mass Distribution methods

We follow Travouillon and Legendre‘s (in review) revised methodology and nomenclature (Chapter 4) of the cenogram method (Legendre, 1986, 1989). Cenograms were built for each locality by putting the logarithm (base 10) of the mean body mass of each species in each site on the Y-axis, and rank ordered from largest to smallest on the

X-axis. Bats were excluded as per the original cenogram methodology (Legendre, 1986,

141 1989), however carnivores were retained because mammalian carnivore size has been shown to correlate with the position of the gap in cenograms of open environments

(Myers, 2002; Travouillon and Legendre, in review). Following Travouillon and

Legendre (in review), arboreal species (koalas and possums) were specifically indicated, because the presence of arboreal taxa in different body size categories has been shown to correlate with habitat type. Legendre (1986; 1989) originally described 4 different cenogram patterns, differentiating closed humid, closed arid, open humid and open arid habitats. Travouillon and Legendre (in review) identified new cenogram patterns and concluded that the gap in cenogram patterns reflected pressure on the environment (e.g. habitat fragmentation, island effect), not necessarily caused by the opening of the vegetation. The revision of the cenogram patterns are shown in Figure

5.1.

Figure 5.1. New cenogram pattern model modified from Travouillon and Legendre (in review) for Australian habitats, showing patterns in —closed“ and —open“ environments, following Legendre (1989). Trees represent forests or woodlands, grasses represent shrublands, grasslands and deserts, waves represent islands and possums represent arboreal species. The dotted line represents mammalian species ordered in order of mass. The dashed lines separate large, medium and small species.

142

We also constructed for each locality a body mass distribution graph, following

Travouillon and Legendre (in review). Body mass distribution graphs are built by representing the proportion of species in each of the following body mass categories: log10 0-0.99 (0 to 9.99 grams; very small), log 1-1.99 (10 to 99.99 grams; small), log 2-

2.99 (100 to 999.99 grams; medium-sized), log 3-3.99 (1000 to 9999.99 grams; large), log 4-4.99 (10000 to 99999,99 grams; very large), log 5-5.99 (100000 to 999999.99 grams; mega fauna).

Figure 5.2. Body mass distribution patterns in Australian habitats, modified from Travouillon and Legendre (in review). Bar graphs represent the proportion of mammal species in each of the body mass categories of four main patterns identified for rainforest, open forest, woodland and grassland and desert. Arboreal species are represented in light grey. The trend lines represent the overall shape of the body mass distribution.

Travouillon and Legendre (in review) showed that this type of graphical representation was more accurate at identifying habitat type than cenograms (Chapter

143 4). However, body mass distribution graphs are unable to identify environmental pressures, responsible for unusual patterns, which cenograms are able to identify. The combination of the use of cenograms and body mass distribution graphs are therefore recommended. Body mass distribution patterns for Australian habitats are shown in

Figure 5.2.

5.3.4 Discriminant Function Analysis (DFA) of Body Mass Distribution data

We selected 16 modern mammal lists (excluding bats) from Travouillon and

Legendre (in review), representing the four habitat types of Figure 5.2. We avoided selecting any list containing any evidence of BMD pattern distortion, caused by the dingo fence, habitat fragmentation, island effect and mixed habitats (Travouillon and

Legendre, in review). Grassland and desert habitats are represented by Millstream-

Chichester, Uluru, Mutawintji (historical), Witjira and Simpson Desert national parks; woodland habitats are represented by Boodjamulla, Currawinya, Gregory and

Gundabooka national parks; open forest habitats are represented by South East Forest,

Blue Mountains, Kosciuszko, Snowy River and Mount Barney national parks; and rainforest habitats are represented by Iron Range National Park and the rainforest mammal list of New Guinea (Travouillon and Legendre, in review). A Discriminant

Function Analysis (DFA) was run by defining functions from the four habitats described above using a data matrix of proportions of arboreal and non-arboreal species in each of the body mass categories (see section 5.3.2). The resulting functions, calculated using the Mahalonobis distance in the SPSS software, were then used to classify fossil sites as one of the four habitat types. Faunas collected from fossil sites are usually time-averaged, and so tend to have more species in certain body mass categories than observed in modern communities (Behrensmeyer et al., 2000). As a result, we used

144 the proportion of the total number of species from the fauna that fall within a particular body mass category, rather than simply the number of species in that category. We also standardised the data by square rooting each proportion. In modern communities, woodlands, grasslands and deserts are characterised by an absence of arboreal species in some or all body mass categories. However, fossil sites may have similar absences in the same body mass categories, but these absences are not necessarily reflecting the habitat type, but rather missing data or taphonomic biases. To avoid such misclassifications for the fossil sites, all absences were replaced by question marks. The classification of fossil sites is therefore less accurate with increasing missing values.

5.4 RESULTS

5.4.1 Faunal Zone A Cenograms and Body Mass Distributions

Cenograms and Body Mass Distribution (BMD) graphs of Faunal Zone A sites, represented by D-Site and White Hunter Site, are shown in Fig. 5.3. D-Site contains no identified species under 1kg (log10 3) and no identified arboreal species (Fig. 5.3A and

5.3C). This missing data prevents the identification of D-Site‘s habitat type using the cenogram and BMD methods. White Hunter Site has no missing data, containing both species under 1kg and arboreal species (Fig. 5.3B and 5.3D); although relatively few are represented (5 species) compared to mammals over 1kg (17 species). White Hunter

Site‘s cenogram (Fig. 5.3B) shows no gap in the medium-sized mammals (log10 2-2.99) and arboreal species are present in the small (log10 1-1.99) and medium-sized (log10 2-

2.99) mammal categories but not in the large mammal (log10 3-3.99), suggesting a

—closed/forested“ environment. In addition, the largest proportion of mammals in White

Hunter Site (Fig. 5.3D) is in the large mammal category (log10 3-3.99), suggesting an open forest habitat.

145 Although there is definite evidence for missing data for Faunal Zone A sites, cenogram and BDM of White Hunter Site suggest at least an open forest habitat, while habitat type cannot be identified for D-Site.

Figure 5.3. Cenograms (A and B) and Body Mass Distribution graphs (C and D) of Faunal Zone A sites (D Site: A and C; White Hunter Site: B and D). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

5.4.2 Faunal Zone B Cenograms and Body Mass Distributions

Faunal Zone B (Figs. 5.4 and 5.5), represented by six sites, Creaser‘s Rampart

(Figs. 5.4A and 5.4D),Camel Sputum (Figs. 5.4B and 5.4E), Dirk‘s Tower (Figs. 5.4C and 5.4F), Neville‘s Garden (Figs. 5.5A and 5.5D), Upper (Figs. 5.5B and 5.5E) and

Wayne‘s Wok (Figs. 5.5C and 5.5F) Sites. There are no distinct gaps in medium-sized mammals (log10 2-2.99) for any of the six cenograms. Arboreal species are present in the small (log10 1-1.99), medium-sized (log10 2-2.99) and large (log10 3-3.99)

146 mammal categories. The absence of gaps and the presence of arboreal taxa suggest either rainforest or open forest. The BMD graphs of Camel Sputum Site (Fig. 5.4E) has equal proportions of large (log10 3-3.99) and medium-sized (log10 2-2.99) mammals.

These two BMD categories are the only characters separating rainforest and open forest with rainforest having the largest proportion of species in the medium-sized mammal category in contrast to open forest which has the largest proportion in the large mammal category.

Figure 5.4. Cenograms (A, B and C) and Body Mass Distribution graphs (D, E and F) of Faunal Zone B sites (Creaser‘s Rampart Site: A and D; Camel Sputum Site: B and E; Dirk‘s Tower Site: C and F). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

147

Figure 5.5. Cenograms (A, B and C) and Body Mass Distribution graphs (D, E and F) of Faunal Zone B sites (Neville‘s Garden Site: A and D; Upper Site: B and E; Wayne‘s Wok Site: C and F). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

The proportion of arboreal species in most Faunal Zone B sites is much larger than observed in modern BMD graphs (Fig. 5.2), especially in the medium-sized category. Creaser‘s Rampart (Fig. 5.4D), Dirk‘s Tower (Fig. 5.4F), Neville‘s Garden

(Fig. 5.5D), Upper (Fig. 5.5E), and Wayne‘s Wok (Fig. 5.5F) Sites have larger proportions of species in the medium-sized mammals, and the overall distribution resembles the pattern for modern tropical rainforest.

148 Overall, cenograms and BMD graphs of Faunal Zone B Sites suggest rainforest habitat.

5.4.3 Faunal Zone C Cenograms and Body Mass Distributions

Faunal Zone C Site cenograms and BMD graphs are shown in Figures 5.6 and

5.7. All five Faunal Zone C Site cenograms (Figs 5.6A, 5.6B, 5.6C, 5.7A, 5.7B) have no distinct gap and have arboreal in the small, medium-sized and large mammal categories, suggesting a closed environment, either rainforest or open forest habitat. Each of the

Faunal Zone C Site BMD graph (Figs 5.6D, 5.6E, 5.6F, 5.7C, 5.7D) is different and suggests different habitats. Cleft Of Ages (Fig. 5.6E), Gag Site (Fig. 5.6F), Henk‘s

Hollow (Fig. 5.7C) and Last Minute (Fig. 5.7D) sites BMD graphs, having the largest proportion of mammals in their medium-sized category, suggests rainforest habitat.

Alan‘s Ledge 1990 Site has no species over 100 kg (Fig. 5.6A). It has relatively low species richness, suggesting missing data. Alan‘s Ledge 1990 Site has equal proportion of mammals in both the medium-sized and large mammal categories suggesting either of rainforest or open forest. In all Faunal Zone C Site BMD graphs, proportions of arboreal species are unusually high suggesting missing data in non-arboreal species.

Overall, cenograms and BMD graphs of Faunal Zone C does also suggest rainforest habitat as seen in Faunal Zone B.

149

Figure 5.6. Cenograms (A, B and C) and Body Mass Distribution graphs (D, E and F) of Faunal Zone C sites (Alan‘s Ledge 1990 Site: A and D; Cleft Of Ages Site: B and E; Gag Site: C and F). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

150

Figure 5.7. Cenograms (A and B) and Body Mass Distribution graphs (C and D) of Faunal Zone C sites (Henk‘s Hollow Site: A and C; Last Minute Site: B and D). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

5.4.4 Encore Site Cenogram and Body Mass Distribution

The shape of Encore Site‘s (Fig. 5.8A) cenogram is very similar to White

Hunter‘s cenogram shape (Fig. 5.3B). The cenogram has no distinct gap and arboreal species are present in small, medium-sized and large mammal categories, suggesting open forest or rainforest. The largest proportion of mammals in the BMD graph (Fig.

5.8B) is in the large mammal category, suggesting open forest habitat. The absence of non-arboreal species in the medium-sized mammal category, and the relatively poor representation of mammal in the small category suggest that the data is incomplete. The high representation of very large (log10 4-4.99) species also suggests open forest habitat.

151 Pending the availability of further data, open forest is the most likely habitat type for Encore Site.

Figure 5.8. Cenogram (A) and Body Mass Distribution graph (B) of Encore Site. Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

5.4.5 Combined Faunal Zone Cenograms and Body Mass Distributions

The results of the combination of the cenograms and BMD graphs of each

Faunal Zone are shown in Fig. 5.9. The combination of the three Faunal Zone A Sites

(Fig. 5.9A and 5.9D) brings no new information. It confirms, however, the suggestion of open forest (no gap and presence of arboreal species) and shows the under- representation of small and medium-sized mammals.

The combined Faunal Zone B Site cenogram (Fig. 5.9B) shows an over representation of species of similar size compared to all modern cenograms. This may suggest that sites assigned to Faunal Zone B may sample different faunas from different times as a result of time averaging or alternatively they may sample a much larger area containing more than one habitat type. Faunal Zone B site BMD graphs suggested rainforest habitat, while the combined Faunal Zone B site BMD graph (Fig. 5.9E) suggests open forest, having its largest proportion of mammals in the large category.

152 The proportion of arboreal species in the combined Faunal Zone B site BMD graph is very high, suggesting missing data for non-arboreal species.

Figure 5.9. Cenograms (A, B and C) and Body Mass Distribution graphs (D, E and F) of the combined Faunal Zone A, B and C sites (Faunal Zone A: A and D; Faunal Zone B: B and E; Faunal Zone C: C and F). Arboreal species are represented by light triangles and carnivores by dark squares in cenograms. Bar graphs represent the proportion of mammal species in each of the body mass categories. Arboreal species are represented in light grey.

The cenogram of the combined Faunal Zone C sites (Fig. 5.9C) also shows an over representation of species of similar size. The combined Faunal Zone C site BMD graph (Fig. 5.9F) suggests rainforest habitat due to the high proportion of medium-sized

153 mammals, but high proportion of arboreal species also suggest missing data for non- arboreal species.

5.4.6 Discriminant function analysis

The DFA yielded three functions from the four habitat types using 5 informative variables, selected by the analysis. A summary of DFA statistics is shown in Table 1 for all three functions. Function 1 accounted for 99.93% of the variance, while function 2 and 3 accounted for only 0.056% and 0.017% of the variance respectively (Table 5.1).

Only 5 of the 12 variables were found to be informative. These are the proportion of small, medium and large arboreal species and the proportion of small and medium non- arboreal species (Table 5.1). The classification analysis (Table 5.2) resulted in 100% of modern mammal lists being correctly classified to each habitat type on both original counts and cross-validated counts, confirming the validity of the classification of habitat types based on the functions used to define them. 41% of fossil localities were classified as rainforest while 59% were classified as open forest (Table 5.2). Function 1 in the canonical discriminant function analysis plot (Figure 5.10) separates rainforest and open forest from grassland/desert and woodland habitats, which the latter are only separated by function 2. Fossil localities are separating along function 1 and 2, but some localities show very large values on both axes, probably caused by missing data (e.g. Cleft of

Ages Site). Fossil localities and their classifications are shown in Table 5.3, along with missing values, which decrease the confidence of the classification (more missing values = less confidence). D Site, combined Faunal Zone A sites, Camel Sputum Site,

Dirk‘s Tower Site, Neville‘s Garden Site, Wayne‘s Wok Site and combined Faunal

Zone B sites are all classified as rainforest habitats, while all other sites are classified as open forest. Most of those habitat classifications are contrary to the visual interpretation

154 of the Body Mass distributions, switching from rainforest habitat (visual interpretation) to open forest habitat (DFA) and vise versa. In both cases, visual interpretation and

DFA, however, Riversleigh sites are classified as either rainforest or open forest, rejecting the possibility of any of the sites being a woodland, grassland or desert habitat.

Small and medium arboreal species are the two main variables separating rainforest from open forest habitats. However, we have identified that the proportion of arboreal taxa is much higher than in modern communities in some fossil localities, possibly due to an under representation of non-arboreal taxa. These fossil localities with higher proportion of arboreal taxa do corresponding to sites with contrary classification. For this reason, the results of the DFA should be treated with caution, and visual interpretations are probably more reliable.

Table 5.1. Summary of Discriminant Function Analysis results.

155 Table 5.2. Results of the habitat classification of the modern and fossil localities, showing both the original count and the cross-validated count.

Figure 5.10. Plot of the canonical discriminant function analysis, using functions 1 and 2 of the DFA. WH is White Hunter Site, CS is Camel Sputum Site, CR is Creaser‘s Rampart Site, DT is Dirk‘s Tower Site, NG is Neville‘s Garden Site, U is Upper Site, WW is Wayne‘s Wok Site, AL90 is Alan‘s Ledge 1990 Site, COA is Cleft Of Ages Site, Gag is Gag Site, HH is Henk‘s Hollow Site, LM is Last Minute Site, En is Encore Site and FZA, FZB and FZC are Faunal Zone A, B, and C, respectively.

156 Table 5.3 Predicted classification of the fossil localities.

5.5 DISCUSSION

5.5.1 Limitations

Palaeoecological studies are subject to a number of limitations. They are based on fossil deposits that may under-sample the communities they represent, sample more than one community, or have had too much of their originally more complete record destroyed by erosion or missed during collection. In a recent application of cenogram methodology, Tougard and Montuire (2006) attempted to analyse the palaeoenvironment of seven Quaternary mammal faunas; however, only one fauna had sufficient specimens to infer habitat type. The other six faunas were heavily biased toward large mammals, missing medium-sized and small species (Tougard and

Montuire, 2006). Our results showed a similar bias in D Site, missing species under 1 kg. In most cases, this large mammal bias could be the result of sorting bias (large fossils are picked first because they are more easily found) which results in a publishing bias of large mammals. Animal body mass distributions have been shown to change as more species are described (Blackburn and Gaston, 1994). This is generally not the case for the Riversleigh faunas because the whole of the limestone blocks collected are acid-

157 processed; hence no taxon in the site, large and small, can be missed as a result of

”collection /sorting biases. However, biases of other kinds may be involved. D Site, for example, as most Faunal Zone A sites, yields very few fossils of small or medium-sized taxa, most likely a result of taphonomic biases or because the sites themselves were not as effectively trapping/capturing small mammals. Alternatively, the sites are not preserving diagnostic features of small or medium-sized taxa for any species level identification.

Taphonomic studies have been undertaken for Camel Sputum, Mike‘s

Menagerie, Ringtail, Encore, Quantum Leap, Bitesantennary (Bassarova, 2005), Keith‘s

Chocky Block (Morrell, 2002), White Hunter (Myers, 2002) and Price Is Right

(Roberts, 2004) Sites. These were all found to be autochthonous assemblages and time averaging was minimal. Although body size biases were not identified in these studies

(Morrell, 2002; Myers, 2002; Roberts, 2004; Bassarova, 2005), neither were these examined in any detail. Apart from the taphonomic studies of White Hunter and

Quantum Leap Sites (Bassarova, 2005; Myers, 2002), which showed no body mass biases, no other Faunal Zone A sites have been the subject of previous taphonomic studies.

One advantage of the cenogram method is that large body mass biases can be visually identified with ease, thereby avoiding incorrect interpretations. Missing data

(up to 40%) has been shown to have no impact on cenogram patterns (Gómez Cano et al., 2006). However, this impact was calculated from randomly removed data.

Taphonomic or sampling biases are rarely random events, and they will most likely cause a bias towards certain body mass categories, changing the overall shape of the

158 cenogram (Blackburn and Gaston, 1994). For example, Bitesantennary Site, being a cave deposit, contains mostly bats (Bassarova, 2005), which are not used in the cenogram method. All other mammals present in Bitesantennary Site are small mammals. In such a case, use of cenogram methodology would produce obviously flawed results. It is imperative to know the limitation of the cenogram methodology before interpreting a dataset because differences may well reflect factors other than differences in the composition of the faunas sampled. We applied the MSR method and taxonomic distinctness analysis previously to minimise these biases and insure that the dataset can be used for palaeoecological studies (Clarke and Warwick, 1994; Warwick and Clarke, 1995; Clarke and Warwick, 1998; Warwick and Clarke, 1998; Clarke and

Warwick, 1999, 2001; Travouillon et al., 2007; Travouillon et al., in prep.).

5.5.2 Comments on the use of the revised cenogram method with fossil localities

Although fossil localities have more limitations than modern communities, the revised cenogram methodology, combined with body mass distribution graphs, is able to determine at least one or two most probable habitat types for each fossil locality if sufficient data is available and if strong body mass biases are absent. Some inconsistencies were found between the results of the DFA and the visual identification of habitat types of the fossil localities, especially for Faunal Zone C sites. Myers (2002) performed a similar DFA analysis and identified problems with the use of DFA on fossil data. Myers‘ (2002) DFA resulted in inconsistent habitat type classification for his fossil localities. Myers (2002) suggested that recent and fossils faunas are not directly comparable because fossil faunas are the results of time-averaging. This is in congruence with the results given by the combined Faunal Zone cenograms (Fig. 5.9), showing over representations of mammalian species unlike modern communities.

159 Taphonomical biases are almost certainly another reason for inconsistent habitat classification. As mentioned earlier, the DFA separated rainforest from open forest based on two variables: proportion of small and medium arboreal taxa. Several fossil localities have higher proportion of arboreal taxa than modern communities, probably caused by missing data in non-arboreal taxa. For this reason, several fossil localities were classified as open forest while their body mass distribution suggests a rainforest habitat. The classification of the DFA is therefore misleading because it is heavily influenced by missing data.

Previously, the cenogram method was only able to identify whether a fossil locality represented a closed or open habitat and whether it was a humid or arid environment (Legendre, 1986, 1989), and it relied essentially on visual interpretations of cenograms. Cenograms were for a long time criticised for not being statistically rigorous (Rodríguez, 1999). Nevertheless, their use in palaeoecological studies has continued, and a new methodology using quantitative data from cenograms was developed to support their use (Gómez Cano et al., 2006; Hernández Fernández et al.,

2006). Hernández Fernández et al. (2006)‘s method relies heavily on the quantification of the magnitude of cenogram gaps. This method was earlier criticized by Siemann and

Brown (1999), using historical data from arid Australia, which was shown to contain no distinct gaps in its cenogram. Travouillon and Legendre (in review) demonstrated that gaps were only present in cenograms if —environmental pressures“ were present and sufficiently detrimental. In the case of Australia, where carnivorous mammals are today relatively less common than they are in communities in the rest of the world (Wroe et al., 2004), aridity, as an —environmental pressure“, was not sufficient to create a gap.

Travouillon and Legendre (in review) proposed a new use for cenograms, as a visual

160 tool to measure environmental pressure rather than as a tool for interpreting palaeohabitats. Used in conjunction with body mass distribution graphs and identification of arboreal species, which provide a more reliable means of identifying habitat types, cenograms become a consolidating tool used to check body mass biases, rather than an identification tool. A palaeoecological study of the La Venta palaeoassemblage (Columbia, South America) found significant correlation between rainfall and several faunal attributes (Kay and Madden, 1997). Among those attributes, the number of species between 100 g and 10kg and the number of arboreal species were significantly correlated to rainfall, supporting our use of both body mass and arboreality.

Gaps in cenograms are the result of combined environmental pressures and may be caused by aridity, habitat fragmentation and high predation on medium-sized species

(Travouillon and Legendre, in review). As demonstrated by Travouillon and Legendre

(in review), one of these pressures alone is not enough to create a gap. Hence, the presence of a gap in medium-sized mammals can be confidently attributed to open environment, but the presence of no gap does not necessarily mean that the environment is closed.

In the case of Riversleigh, none of the cenograms show a gap. This means that the environmental pressure was low, either because it was a closed environment, such as a rainforest or an open forest, or it was more arid (woodland or shrubland/grassland) but there was low predation on medium-sized mammals. To determine which of these two possible scenarios is more likely to be the correct one, carnivorous species were represented on cenograms. The recorded presence of carnivorous species ranges from 0

161 to 4 species within Riversleigh Sites, with up to 2 sympatric thylacinid species (e.g.

Camel Sputum Site) and up to 3 sympatric thylacoleonid species found within a site

(e.g. Dirk‘s Tower). The combined datasets shows up to eight sympatric carnivorous species (Faunal Zone B), including thylacinids, thylacoleonids and dasyurids, which does not support a scenario of low predation pressures.

The size of the predators is also quite important for interpretations of cenograms.

Valverde (1964; 1967) suggested that the position of the gap in cenogram patterns reflects the size of the predators. This was supported by Australian data (Travouillon and Legendre, in review) which showed a shift in the position of the gap toward smaller mammals due to the lack of very large predators (e.g. lion sized) in modern Australia.

For Riversleigh, the largest predators are thylacoleonids of the genus Wakaleo, ranging in size between 15 to 35 kg, only slightly larger than a modern dingo. The gap in cenogram patterns would be expected in the same body mass range as it is today, between 100 grams and 1 kg.

5.5.3 Arboreal species

The new cenogram methodology relies heavily on the presence of arboreal species for habitat identification. Only koalas and possums were considered as arboreal species in the original method. Other arboreal species such as tree kangaroos are restricted to small areas of rainforest in northern Queensland and were not present in the national parks studied. While some modern dasyurids and rodents spend time in trees, they are better classified as scansorial, spending a large proportion of time on the ground. In this study, it is assumed that Oligo-Miocene koalas and possums were also arboreal. To our knowledge, no published post-cranial remains have been confidently

162 assigned to either of Oligo-Miocene koalas or possums. However, arboreality is almost certainly plesiomorphic for Diprotodontia, with macropodoids and vombatiforms secondarily terrestrial; although arboreal, koalas are usually believed to have evolved from a terrestrial vombatiform ancestor (Aplin and Archer, 1987; Szalay, 1994; but see

Weisbecker and Archer, in press). The assumption of arboreality for Oligo-Miocene koalas and possums is therefore quite reasonable until evidence shows otherwise.

5.5.3 Riversleigh and the climate of northern Australia during the Oligo-Miocene.

The palaeoenvironment of all Riversleigh Oligo-Miocene sites was originally interpreted to be rainforest mainly because of the high mammal diversity and the presence of groups of animals that today live only in rainforest (Archer et al., 1989).

Mammal diversity, which may be high in fossil assemblages because of time averaging

(Behrensmeyer et al., 2000), may not be the most accurate indicators of habitat. In the case of Riversleigh, time averaging has been shown to have a minimal impact

(Bassarova, 2005). The presence of habitat specific animals is, however, a much better argument to support the hypothesis that Riversleigh had rainforest habitats. Megirian et al. (2004) challenged the hypothesis of Riversleigh rainforest palaeohabitats on the basis of the presence of a gastropod from D Site interpreted by them to tolerate no more than 600mm of rainfall a year, which would be too little to support rainforest. They also pointed out that fossil plants recovered at Riversleigh (Arena, 1997) do not support the rainforest hypothesis of Archer et al. (1989). More recent collections of plants from

Dunsinane have, however, revealed undoubted rainforest taxa such as Burdekin Plums

(Pleiogynium; H. Godthelp, pers. comm. 2007)) There are also illogical assertions in the arguments of Megirian et al. (2004)‘s arguments. Firstly, the gastropod used as a palaeoclimatic indicator is only found at D Site. Similarly, the fossil plants were only

163 recovered from Dunsinane Site (Arena, 1997), and D Site as well as Dunsinane Site have been interpreted on the basis of their mammal species and multivariate analysis to be Faunal Zone A sites (Arena, 1997; Travouillon et al., 2006). These analyses showed a much closer relationship between these Sites and White Hunter Site, identified as late

Oligocene on the basis of biocorrelation such as the ilariid, Kuterintja ngama (Archer et al., 1997; Arena, 1997; Myers and Archer, 1997; Travouillon et al., 2006), with South

Australia‘s late Oligocene Ngama Local Fauna (Etadunna Formation). While Megirian et al. (2004) acknowledge the fact that Riversleigh sites are of different ages (Murray et al., 2000), they nevertheless illogically assert on the basis of the gastropod from D Site and plants from Dunsinane, that none of the Oligo-Miocene sites of Riversleigh could represent rainforest. Megirian et al. (2004) attempt to back up their conclusion by arguing that Riversleigh sites contain what they claim to be anomalous associations of non-contemporaneous taxa including Litokoala kanunkaensis and species of the pseudocheirid genera Pildra and Marlu, concluding that Riversleigh sites are mixed faunas. They do not demonstrate any geological evidence for this conclusion. Certainly, nothing in our experience working every year at Riversleigh for 30 years would support this assertion. Second, subsequent to publication of Megirian et al.‘s (2004) study, the species Litokoala kanunkaensis was sunk within Litokoala kutjamarpensis on the basis of newly-recovered dental material from Riversleigh (Louys et al., 2007). Hence

Litokoala kutjamarpensis spans the late Oligocene Ngama and early Miocene

Kutjamarpu Local Faunas of South Australia as well as Riversleigh Faunal Zones B and

C (early and middle Miocene respectively), making it a relatively long-lived species. No koalas have yet been recovered in Faunal Zone A sites to date.

164 In the case of Pildra and Marlu, Riversleigh species are different from species found in the late Oligocene of South Australia, and represent later stages of evolution

(K. Roberts, pers. comm.), which does not support Megirian et al. (2004)‘s argument.

One would assume that if Riversleigh sites were in fact mixed faunas, species at different stage of evolution would be found in the same site. However, there is no record of any such occurrences. Well-studied lineages such as Neohelos (Murray et al.,

2000) and Wakaleo (Gillespie, 2007) do not have any recorded overlapping species at different stages of evolution, making the possibility of faunal mixing less likely. In fact,

Megirian et al. (2004)‘s argument about faunal mixing is based on the fact that they consider it unlikely that the temporal range of species regarded as index taxa is greater than previously thought. Although mean species lifespans for mammals appear to be about 2.5 million years (Alroy, 2000; Vrba and DeGusta, 2004; van Dam et al., 2006), there are numerous examples of species with much longer lifespans suggesting stasis

(Prothero and Heaton, 1996). Evolutionary stasis is a controversial topic. Two broad hypotheses for species lifespans have been proposed: gradualism and punctuated equilibrium (Gould and Eldredge, 1993; Alroy, 1996, 2000; Gould, 2002; Cressman and

Garay, 2006). Assuming faunal mixing did not occur at Riversleigh, examples supporting both hypotheses, gradualism (e.g. the genus Neohelos) and punctuated equilibria (e.g. the stasis of Litokoala kutjamarpensis) can be found at Riversleigh.

However, another hypothesis could explain the —unusual“ lifespan of some of

Riversleigh‘s species: species validity (Alroy, 2002). Alroy (2002) has pointed out that with time (and with more data), many species names are eventually found to be synonyms or invalid. Fossil mammal species are usually erected using dental characters only. In many cases - for example Oligo-Miocene, Pliocene and recent species of the marsupial genus Burramys (Brammall and Archer, 1997), closely related species are

165 distinguished by relatively very minor dental differences. Burramys brutyi from the

Oligo-Miocene Riversleigh deposits appears to show morphological stasis (at least in terms of tooth size) through time. A metric analysis of B. brutyi specimens from

Riversleigh Faunal Zones A, B and C revealed no significant difference in size variation between sites (Brammall and Archer, 1997). Cranial and postcranial material of B. brutyi has not been recognised from any of the Riversleigh sites, and so it is unclear if this apparent morphological stasis was restricted to the dentition or whether it extended to other anatomical regions. Nevertheless, it is clear that species longevity could reflect morphological stasis rather than faunal mixing.

Megirian et al. (2004) do not address the presence of a number of rainforest- restricted animals in Riversleigh sites, which are furthermore not found anywhere else in Australia during the Oligo-Miocene. Rainforest-restricted hyspiprymnodontids are found only in Faunal Zone B and C sites of Riversleigh (Flannery and Archer, 1987;

Bates, 2007). Similarly, lyrebirds (Menura) and logrunners (Orthonyx), birds today that only occur in rainforest (Boles, 1997), bubble-nesting frogs (e.g. Lechriodus), frogs today that require perpetual humidity (Tyler et al., 1990), are also only found in

Riversleigh‘s Faunal Zone B and C. Although pseudocheirids are not currently restricted to rainforest habitats, rainforest habitats are differentiated from all other habitats by having higher abundance of sympatric pseudocheirid species, with up to five sympatric species versus zero to two in all other habitats (Archer et al., 1989).

Pseudocheirids are represented in almost all Australian Oligo-Miocene sites. However, they don‘t occur in the same abundance everywhere. In the late Oligocene, the Ngama

Local Fauna (South Australia: Marlu sp. cf. kutjamarpensis and Pildra magnus) and the

Kangaroo Well Local Fauna (Central Australia: Marlu sp. cf. kutjamarpensis and Pildra

166 sp. cf. magnus) both have just two sympatric pseudocheirid species. In contrast,

Riversleigh‘s White Hunter Site has three (Megirian et al., 2004; Travouillon et al.,

2006; Travouillon et al., in prep.). In the early Miocene, the Kutjamarpu Local Fauna

(Leaf Locality) has up to five (Paljara tirarensae, Pildra tertius, Marlu kutjamarpensis,

Marlu sp. 3 and 4) and Riversleigh‘s Faunal Zone B (e.g. Camel Sputum Site) has up to seven sympatric pseudocheirids (Travouillon et al., 2006; Roberts et al., 2007;

Travouillon et al., in prep.). In the middle Miocene, Riversleigh‘s Faunal Zone C (e.g.

Gag Site) has up to 9 species while Bullock Creek Local Fauna has zero (Travouillon et al., 2006; Roberts et al., 2007; Travouillon et al., in prep.). Finally, in the early late

Miocene from Riversleigh (Encore Site), there are 3 sympatric pseudocheirids while

Alcoota has only one (Travouillon et al., 2006; Roberts et al., 2007; Travouillon et al., in prep.). Based on the evidence of pseudocheirids, the climate of northern Australia during the late Oligocene and late Miocene can be inferred to have been relatively drier than during the middle Miocene, and at all times between the late Oligocene and late

Miocene apparently relatively drier than Riversleigh.. Our cenogram and body mass distribution analyses support this general pattern of climate change within the

Riversleigh region, with Riversleigh‘s Faunal Zone A (late Oligocene) and D (late

Miocene) interpreted to represent woodland or open forest and Faunal Zone B (early

Miocene) and C (middle Miocene) interpreted to represent open forest or rainforest. Our results are also in agreement with McGowran and Li‘s (1994) hypothesis, based on stable oxygen isotope ratio of foraminifera, that the late Oligocene and late Miocene of

Australia were characterised by ”Icehouse‘ conditions, whereas the early and middle

Miocene were characterised by ”Greenhouse‘ conditions. In fact, McGowran and Li

(1994) pointed out that the different interpretations of Riversleigh‘s climate (e.g.

Megirian, 1992; Archer et al., 1989, 1997; Creaser, 1997) are both supported by their

167 foraminiferal evidence, but reflect different time periods: Riversleigh was colder and drier during Faunal Zones A and D and warmer and wetter during Faunal Zones B and

C. Recent geological investigation of the Riversleigh fossil deposits by Arena (2004) suggests a complex sequence of fluvial and karst processes resulting in four depositional phases of tufa and cave deposits, the first (late Oligocene) and last (early late Miocene) phases being relatively drier that the middle two phases. This is in agreement with

McGowran and Li (1994) and the results of our cenogram and body mass distribution analyses. Dating of the Riversleigh sites is currently underway and should test the current hypothetical ages of the Faunal Zones established primarily on the basis of biocorrelation (I. Graham, 2008, pers. comm.).

5.6 CONCLUSION

The study of the palaeoecology of the Riversleigh Faunal Zones of the Oligo-

Miocene using the new cenogram methodology was able to identify with a degree of confidence the habitat type of each Faunal Zone. The new methodology was found to work well despite limitations of the fossil record, which were minimized using MSR and taxonomic distinctness analyses. The late Oligocene Faunal Zone A is identified as an open forest habitat; early Miocene Faunal Zone B as a rainforest habitat, and middle

Miocene Faunal Zone C as a rainforest habitat; and finally early late Miocene Encore

Site as an open forest habitat. These results confirm and support earlier studies and help resolve some contradictory arguments made about the climate of Australia‘s Oligo-

Miocene. However, further dating of Australian Oligo-Miocene sites is necessary, along with further palaeoecological investigations of Australia‘s Oligo-Miocene vertebrate assemblages, before a more substantial understanding about environmental change can emerge.

168 5.7 REFERENCES

Alroy, J. (1996). Constant extinction, constrained diversification, and uncoordinated stasis in North American mammals. Palaeogeography, Palaeoclimatology, Palaeoecology 127: 285-311.

Alroy, J. (2000). New methods for quantifying macroevolutinary patterns and processes. Paleobiology 26: 707-733.

Alroy, J. (2002). How many named species are valid? Proceedings of the National Academy of Sciences 99: 3706-3711.

Aplin, K. P. and M. Archer (1987). Recent advances in marsupial systematics with a new syncretic classification. Possums and Opossums, Studies in Evolution. M. Archer. Sydney, Surrey Beatty and Sons.

Archer, M., D. A. Arena, M. Bassarova, R. M. D. Beck, K. Black, W. E. Boles, P. Brewer, B. N. Cooke, K. Crosby, A. Gillespie, H. Godthelp, S. J. Hand, B. P. Kear, J. Louys, A. Morrell, J. Muirhead, K. K. Roberts, J. D. Scanlon, K. J. Travouillon and S. Wroe (2006). Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1: 1-17.

Archer, M., R. Arena, M. Bassarova, K. Black, J. Brammall, B. Cooke, P. Creaser, K. Crosby, A. Gillespie, H. Godthelp, M. Gott, S. J. Hand, B. Kear, A. Krikmann, B. Mackness, J. Muirhead, A. Musser, T. Myers, N. Pledge, Y. Wang and S. Wroe (1999). The evolutionary history and diversity of Australian mammals. Australian Mammalogy 21: 1-45.

Archer, M., H. Godthelp, S. J. Hand and D. Megirian (1989). Fossil mammals of Riversleigh, northwestern Queensland: preliminary overview of biostratigraphy, correlation and environmental change. Australian Zoologist 25: 29-65.

Archer, M., S. J. Hand and H. Godhelp (1994). Patterns in the history of Australia's mammals and inferences about palaeohabitats. . History of the Australian Vegetation. R. Hill. Cambridge, Cambridge University Press: 80-103.

Archer, M., S. J. Hand and H. Godhelp (1994b). Patterns in the history or Australia‘s mammals and inferences about palaeohabitats. History of the Australian Vegetation: Cretaceous to Recent. R. S. Hill. Cambridge, Cambridge University Press: 80-196.

Archer, M., S. J. Hand and H. Godthelp (1994a). Riversleigh, Sydney, 256 pp.

Archer, M., S. J. Hand, H. Godthelp and P. Creaser (1997). Correlation of the Cainozoic sediments of the Riversleigh World Heritage fossil property, Queensland, Australia. . Actes du congrès BiochroM'97, Mémoires et Travaux de l'Ecole Pratique des Hautes Etudes. J.-P. Aguilar, Legendre, S., Michaux, J. Institut de Montpellier. 21: 131-152.

Arena, D. A. (1997). Palaeontology and geology of Dunsinane Site, Riversleigh. Memoirs of the Queensland Museum 41(2): 171-179.

169

Arena, D. A. (2004). The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle Tertiary. PhD thesis, UNSW, Sydney.

Arena, D. A. (2008). Exceptional preservation of plants and invertebrates by phosphatization, Riversleigh, Australia. Palaios 23: 495-502.

Bassarova, M. (2005). Taphonomic and palaeoecological investigations of Riversleigh Oligo-Miocene fossil sites. PhD, UNSW, Sydney.

Bates, H. (2007). 'New Miocene species of musky rat kangaroos (Hypsiprymnodontidae, Macropodoidea) description, phylogenetics and a new hypotheses about kangaroo evolution. Honours, University of New South Wales, Sydney.

Behrensmeyer, A. K., S. M. Kidwell and R. A. Gastaldo (2000). Taphonomy and paleobiology. Paleobiology 26(4 supplement): 103-147.

Benbow, M. C., N. F. Alley, R. A. Callan and D. R. Greenwood (1995). Geological history and palaeoclimate. The Geology of South Australia, the Phanerozoic,. J. F. Dexel and W. V. Preiss. Adelaide, Geological Survey of South Australia Bulletin. 2: 208-217.

Blackburn, T. M. and K. J. Gaston (1994). Animal body size distribution change as more species are described. Proceedings of the Royal Society of London B 257(1350): 293-297.

Boles, W. E. (1993). A logrunner, Orthonyx (Passeriformes: Orthonychidae) from the Miocene of Riversleigh, north-western Queensland. . Emu 93: 44-49.

Boles, W. E. (1995). A preliminary analysis of the Passeriformes from Riversleigh, northwestern Queensland, Australia, with the description of a new species of lyrebird. . Courier Forchungsinstitut Senckenberg 181: 163-170.

Boles, W. E. (1997). Riversleigh birds as palaeoenvironmental indicators. Memoirs of the Queensland Museum 41(2): 241-246.

Brammall, J. and M. Archer (1997). A new Oligocene-Miocene species of Burramys (Marsupialia. Burramyidae) from Riversleigh. northwestern Queensland. Memoirs of the Queensland Museum 41(2): 247-268.

Brewer, P., M. Archer, S. Hand and H. Godthelp (2007). A new species of the wombat Warendja from late Miocene deposits at Riversleigh, North-West Queensland, Australia. Palaeontology 50(4): 811-828.

Clarke, K. R. and R. M. Warwick (1994). Change in Marine Communities: An approach to statistical analysis and interpretation. Plymouth Marine Laboratory, Plymouth, UK., 144 pp.

170 Clarke, K. R. and R. M. Warwick (1998). A taxonomic distinctness index and its statistical properties. Journal of Applied Ecology 35: 523-531.

Clarke, K. R. and R. M. Warwick (1999). The taxonomic distinctness measure of biodiversity: weighting of step lengths between hierarchical levels. Marine Ecology Progress Series 184: 21-29.

Clarke, K. R. and R. M. Warwick (2001). A further biodiversity index applicable to species lists: variation in taxonomic distinctness. Marine Ecology Progress Series 216: 265-278.

Creaser, P. (1997). Oligocene-Miocene Sediments of Riversleigh: The potential significance of topography. Memoirs of the Queensland Museum 41(2): 303-314.

Cressman, R. and J. Garay (2006). A game-theoretic model for punctuated equilibrium: Species invasion and stasis through coevolution. BioSystems 84: 1-14.

Flannery, T. F. and M. Archer (1987). Hypsiprymnodon bartholomaii (Potoroidae: Marsupilaia), a new species from the Miocene Dwornamor Local Fauna and a reassessment of the phylogenetic position of H. moschatus. . Possums and Opossums: Studies in Evolution, . M. Archer. Sydney, Surrey Beatty & Sons and the Royal Zoological Society of New South Wales: 749-758.

Frakes, L. A., B. MCGowran and J. Bowler (1987). Evolution of Australian environments. Fauna of Australia. G. R. Dyne and D. W. Walton. Canberra, Australian Government Publishing Service. 1A, General articles: 1-16.

Gillespie, A. K. (2007). Diversity and systematics of marsupial lions from the Riversleigh World Heritage Area and the evolution of the thylacoleonidae. Doctor of Philosophy, University of New South Wales, Sydney.

Gómez Cano, A. R., B. A. García Yelo and M. Hernández Fernández (2006). Cenogramas, análisis bioclimático y muestreo en faunas de mamíferos: implicaciones para la aplicación de métodos de análisis paleoecológico. . Estudios Geológicos 62: 135-144.

Gould, S. J. (2002). The Structure of Evolutionary Theory. Mass.: Harvard University Press, Cambridge

Gould, S. J. and N. Eldredge (1993). Punctuated equilibrium comes of age. Nature 366: 223-227.

Hernández Fernández, M., M. T. Alberdi, B. Azanza, P. Montoya, J. Morales, M. Nieto and P. Peláez-Campomanes (2006). Identification problems of arid environments in the Neogene-Quaternary mammal record of Spain. Journal of Arid Environments 66: 585- 608.

Hill, R. S., Ed. (1994). History of the Australian vegetation: Cretaceous to recent. Cambridge University Press, Cambridge.

171 Kay, R. F. and R. H. Madden (1997). Mammals and rainfall: paleoecology of the middle Miocene at La Venta (Colombia, South America). Journal of Human Evolution 32: 161- 199.

Legendre, S. (1986). Analysis of mammalian communities from the late Eocene and Oligocene of Southern France. Palaeovertebrata 16: 191-212.

Legendre, S. (1989). Les communautés de mammifères du Paléogène (Eocène supérieur et Oligocène) d'Europe occidentale: structures, milieux et évolution. Münchner Geowissenschaftliche Abhandlungen (A) 16: 1-110.

Long, J., M. Archer, T. Flannery and S. Hand (2002). Prehistoric mammals of Australia and New Guinea. One hundred million years of evolution. University of New South Wales Press, Sydney, 244.

Louys, J., K. Black, M. Archer, S. J. Hand and H. Godhelp (2007). Descriptions of koala fossils from the Miocene of Riversleigh, northwestern Queensland and implications for Litokoala (Marsupialia,Phascolarctidae). Alcheringa 31: 99-110.

Martin, H. A. (1990). The palynology of the Namba Formation in the Wooltana-1 bore, Callabonna Basin (Lake Frome), South Australia, and its relevance to Miocene grasslands in central Australia. Alcheringa 14: 247-255.

Martin, H. A. (1994). Australian Tertiary phytogeography: evidence from pollen. History of the Australian Vegetation: Cretaceous to Recent. R. S. Hill. Cambridge, Cambridge University Press: 104-142.

Martin, H. A. (1998). Tertiary climatic evolution and the development of aridity in Australia. Proceedings of the Linnean Society of New South Wales 119: 1115-1136.

Martin, H. A. (2006). Cenozoic climatic change and the development of arid vegetation in Australia. Journal of Arid Environments 66: 533-563.

McGowran, B. (1986). Cainozoic oceanic events: the Indo-Pacific biostratigraphic record. Palaeogeography, Palaeoclimatology, Palaeoecology 55: 247-265.

McGowran, B. and Q. Li (1994). The Miocene oscillation in southern Australia. Records of the South Australian Museum 27: 197-212.

Megirian, D. (1992). Interpretation of the Carl Creek Limestone, northwestern Queensland. The Beagle: Records of the Northern Territory Museum of Arts and Sciences 9(1): 219-248.

Megirian, D., P. Murray, L. Schwartz and C. Von Der Borch (2004). Late Oligocene Kangaroo Well Local Fauna from the Ulta Limestone (new name), and climate of the Miocene oscillation across central Australia. Australian Journal of Earth Sciences 51: 701-741.

172 Megirian, D., P. Murray and R. T. Wells (1996). The late Miocene Ongeva Local Fauna of central Australia. The Beagle. Records of the Museums and Art Galleries of the Northern Territory 13: 9-37.

Morrell, A. (2002). The Geology and Palaeontology of Keith's Chocky Block Site, Riversleigh, World Heritage Property, northwest Queensland. Honours, University of New South Wales, Sydney.

Murray, P. and D. Megirian (1992). Continuity and contrast in middle and late Miocene vertebrate communities from Northern Territory. Proceedings of the 1991 Conference on Australasian Vertebrate Evolution, Palaeontology and Systematics, Alice Springs, The Beagle. Records of the Northern Territory Museum of Arts and Sciences.

Murray, P. and D. Megirian (2000). Two new genera and three new species of Thylacinidae (Marsupialia) from the Miocene of the Northern Territory, Australia. The Beagle: Records of the Northern Territory Museum of Arts and Sciences 16: 145-162.

Murray, P., D. Megirian, T. H. Rich, M. Plane, K. Black, M. Archer, S. Hand and P. Vickers-Rich (2000). Morphology, systematics and evolution of the marsupial genus Neohelos Stirton (Diprotodontidae, zygomaturinae). MAGNT Research Report 6: 141 pp.

Murray, P., D. Megirian, T. H. Rich, M. Plane and P. Vickers-Rich (2000). Neohelos stirtoni. A new species of Zygomaturine (Diprotodonta: Marsupialia) from the mid- Tertiary of the Northern Territory, Australia. Records of the Queen Victoria Museum 105: 1-47.

Myers, T. J. (2001). Prediction of marsupial body mass. Australian Journal of Zoology 49: 99-118.

Myers, T. J. and M. Archer (1997). Kutjerintja ngama (Marsupialia, Ilariidae): a revised systematic analysis based on material from the late Oligocene of Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 379-392.

Myers, T. J., K. Crosby, M. Archer and M. Tyler (2001). The Encore local Fauna, a late Miocene assemblage from Riversleigh, northwestern Queensland. Memoirs of the Association of Australasian Palaeontologists 25: 147-154.

Myers, T. J. M. (2002). Palaeoecology of Oligo-Miocene Local Faunas from Riversleigh. University of New South Wales, Sydney.

Prothero, D. R. and T. H. Heaton (1996). Faunal stability during the Early Oligocene climatic crash. Palaeogeography, Palaeoclimatology, Palaeoecology 127: 257-283.

Rich, T. H. (1991). Monotremes, placentals, and marsupials: their record in Australia and its biases. Vertebrate Palaeontology of Australasia. M. J. M. VIckers-Rich P., Baird R. F. & Rich T. H. Melbourne, Pioneer Design Studio and Monash University Publications Committee: 893œ1070.

173 Roberts, K. K. (2004). The Taphonomy and Geological Context of Price Is Right Site, Riversleigh, World Heritage Area, northwestern Queensland. Honours, University of New South Wales, Sydney.

Roberts, K. K., M. Archer, S. J. Hand and H. Godhelp (2007). New Genus and Species of extinct Miocene Ringtail Possums (Marsupialia: Pseudocheiridae). American Museum Novitates 350: 15pp.

Rodríguez, J. (1999). Use of cenograms in mammalian palaeocology. A critical review. Lethaia 32: 331-347.

Schwartz, L. and D. Megirian (2004). A new species of Nambaroo (Marsupialia; Macropodoidea) from the Miocene Camfield Beds of Northern Australia with observation on the phylogeny of the Balbarinae. Journal of Vertebrate Palaeontology 24(3): 668-675.

Siemann, E. and J. H. Brown (1999). Gaps in mammalian body size distributions reexamined. Ecology 80(8): 2788-2792.

Szalay, F. S. (1994). Evolutionary history of the marsupials and an analysis of osteological characters. Cambridge University Press, Cambridge

Tedford, R. H., R. T. Wells and S. F. Barghoorn (1992). Tirari Formation and contained faunas, Pliocene of the Lake Eyre Basin, South Australia. Proceedings of the 1991 Conference on Australasian Vertebrate Evolution, Palaeontology and Systematics, Alice Springs, The Beagle. Records of the Northern Territory Museum of Arts and Sciences.

Tougard, C. and S. Montuire (2006). Pleistocene paleoenvironmental reconstructions and mammalian evolution in South-East Asia: focus on fossil faunas from Thailand. Quaternary Science Reviews 25: 126-141.

Travouillon, K. J., M. Archer, S. J. Hand and H. Godthelp (2006). Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1: 323-349.

Travouillon, K. J., M. Archer, S. Legendre and S. J. Hand (2007). Finding the Minimum Sample Richness (MSR) for multivariate analyses: implications for palaeoecology. Historical Biology.

Travouillon, K. J., G. Escarguel, S. Legendre, M. Archer and S. J. Hand (in prep.). The use of MSR (Minimum Sample Richness) for fossil fauna comparisons in conjunction with a taxonomic distinctness measure.

Travouillon, K. J. and S. Legendre (in review). Explaining the gaps in body mass distributions (cenograms): statistical visualisation of the ecological impact of introduced species. Palaeogeography, Palaeoclimatology, Palaeoecology.

Tyler, M. J., S. J. Hand and V. J. Ward (1990). Analysis of the frequency of Lechriodus intergerivus Tyler (Anura:Leptodactylidae) in Oligo-Miocene local faunas of

174 Riversleigh Station, Queensland. Proceedings of the Linnean Society of New South Wales 112: 105-167.

Valverde, J. A. (1964). Remarques sur la structure et l‘évolution des communautés de vertébrés terrestres. 1. Structure d‘une communauté 2. Rapport entre prédateurs et proies. La Terre et la Vie 111: 121œ154.

Valverde, J. A. (1967). Estructura de una communidad de vertebrados terrestres. Monografías de la Estación Biológica de Doñana 1: 1œ129. van Dam, J. A., H. A. Aziz, M. A. A. Sierra, F. J. Hilgen, L. W. van den Hoek Ostende, L. J. Lourens, P. Mein, A. J. van der Meulen and P. Pelaez-Campomanes (2006). Long- period astronomical forcing of mammal turnover. Nature 443: 687-691.

Vrba, E. S. and D. DeGusta (2004). Do species populations really start small? New perspectives from the Late Neogene fossil record of African mammals. Philosophical Transactions of the Royal Society B 359: 285-293.

Warwick, R. M. and K. R. Clarke (1995). New —biodiversity“ measures reveal a decrease in taxonomic distinctness with increasing stress. Marine Ecology Progress Series 129: 301-305.

Warwick, R. M. and K. R. Clarke (1998). Taxonomic distinctness and environmental assessment. Journal of Applied Ecology 35: 532-543.

Weisbecker, V. and M. Archer (in press). Parallel evolution of hand anatomy in kangaroos and vombatiform marsupials: functional and evolutionary implications. Palaeontology.

Woodburne, M. O. (1967). The Alcoota Fauna, Central Australia. Bureau of Mineral Resources, Geology and Geophysics, Australia Bulletin 87.

Woodburne, M. O., B. J. MacFadden, J. A. Case, M. S. Springer, N. S. Pledge, J. D. Power, J. M. Woodburne and K. B. Springer (1993). Land mammal biostratigraphy and magnetostratigraphy of the Etadunna Formation (Late Oligocene) of South Australia. Journal of Vertebrate Paleontology 13: 483-515.

Woodburne, M. O., R. H. Tedford, M. Archer, W. D. Turnbull, M. D. Plane and E. L. Lundelius (1985). Biochronology of the continental mammal record of Australia and New Guinea. Stratigraphy, Palaeontology, Malacology Papers in Honour of Dr Nell Ludbrook. L. J. M., South Australian Department of Mines and Energy Special Publication. 5: 347œ363.

Wroe, S., C. Argot and C. Dickman (2004). On the rarity of big fierce carnivores and primacy of isolation and area: tracking large mammalian carnivore diversity on two isolated continents. Proceedings of the Royal Society of London B 271: 1203-1211.

175

5.8 APPENDIX

Table 5.4. Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

176 Table 5.4 (continued…) Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

177 Table 5.4 (continued…) Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

178 Table 5.4 (continued…) Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

179 Table 5.4 (continued…) Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

180 Table 5.4 (continued…) Riversleigh sites with species list, specimen numbers and body mass estimates using equations from Myers (2001).

181

CHAPTER 6: ESTIMATION OF RELATIVE AGES OF

RIVERSLEIGH LOCAL FAUNAS USING THE NUMERICAL

AGES METHOD

182 6.1 ABSTRACT

The Numerical Ages method is used in this study to resolve a sequence of relative ages of faunas collected from fossil localities at Riversleigh, northwestern

Queensland, Australia. While age relationships of Riversleigh faunas have previously been determined using biocorrelation, they have yet to be accurately dated with absolute methods, and thus the relative ages of faunas of similar age are uncertain. This method uses regressions of dental measurements from mammalian taxa to estimate the relative ages of the faunas. While this method has been shown to work well as an age estimation tool for faunas from European localities, age estimation requires calibration of the resulting faunal sequence using absolute dates. Our results provide relative age relationships for a sequence of 40 fossil faunas calibrated with the best available dates.

The relative age relationships are compared with stratigraphy in previous work and implications for biostratigraphic interpretation of the Riversleigh Local Faunas and other Australian Oligo-Miocene faunas are discussed.

6.2 INTRODUCTION

Fossil vertebrate faunas have been collected from more than 200 localities in

Cenozoic carbonate deposits at the Riversleigh World Heritage Property in north- western Queensland, Australia. Age estimation of these deposits has been achieved primarily by biocorrelation of the fossil vertebrate faunas they contain with other faunas from Riversleigh, from elsewhere in Australia and from France (e.g. presence/absence and stage-of-evolution of taxa) (Archer et al., 1989; Archer et al., 1997). The

Riversleigh faunas are thus considered to range in age from late Oligocene to early-late

Miocene, with a Pliocene fauna known from a cave deposit in the Cambrian limestone

183 terrain and a Pleistocene fauna recognised from alluvial terraces along the Gregory

River (Archer et al., 1989; Archer et al., 1991; Archer et al., 1995; Archer et al., 1997).

The Riversleigh Tertiary carbonates contain fluvial and cave deposits with a complex depositional history (Archer et al., 1989; Megirian, 1992; Archer et al., 1994;

Archer et al., 1997; Creaser, 1997; Arena, 2004). After detailed lithological study of the

Riversleigh carbonates, Megirian (1992) concluded they were primarily fluvioclastic deposits, with minor tufa and cave deposits, formed in an arid alluvial fan system during the Miocene.

However, because fossil vertebrate faunas from these Riversleigh deposits appear to represent a longer period spanning the late Oligocene to the early-late

Miocene, this suggested that deposition of fluvioclastic and cave deposits in such an alluvial fan setting might have occurred throughout this time in more than one cyclic event (Archer et al., 1989; Archer et al., 1997; Creaser, 1997).

Arena (2004) revised previous diagnoses of Riversleigh lithotypes and reconciled biostratigraphic and geological interpretations in these earlier works with recent advances in karst stratigraphic concepts and karst processes. The cyclic alluvial fan model was rejected and the Riversleigh Tertiary carbonates were interpreted as a karst terrain with a history of development spanning the late Oligocene to the present that could be divided into 4 karst phases sensu Bosák et al. (1989). Each phase is characterised by particular dominant depositional processes beginning with a primary episode of widespread cyclic fluvio-lacustrine tufa deposition, followed by karst terrain development and vadose cave infill with an intervening episode of fluvio-lacustrine tufa

184 deposition during the middle Miocene. These karst depositional phases were related to the record of tectonic and climatic events during this period.

Where karst terrains form in terrestrial freshwater carbonates, such as at

Riversleigh, stratigraphic information may be poorly preserved and standard stratigraphic concepts such as superposition do not necessarily apply. Therefore, it can be difficult to distinguish between sediments of different age and to determine the boundaries of in situ deposits containing local faunas (Archer et al., 1989; Arena, 2004).

Because of these challenges facing stratigraphic interpretation of the Riversleigh deposits, Archer et al. (1989) introduced the 'Systems' stratigraphic concept that combined geological, geographic and faunal information to derive a stratigraphic sequence of fossil faunas. This sequence was divided into time periods called 'System'

A, B and C. The validity of the 'System' concept was disputed by Megirian (1994) because it combined geological and faunal characteristics.

However, Arena (2004) found that the characteristics used to erect the

Riversleigh 'Systems' allowed them to be redefined as stratigraphically valid Faunal

Zones A-C (corresponding to the same time periods represented by 'Systems' A-C), and stratigraphically valid karst Depositional Phases 1-4 characterised by geological and taphonomic characteristics of the deposits containing those faunas (see Fig. 6.1).

Travouillon et al. (2006) recognised a distinct faunal type previously included in Zone

C as representative of a more recent time period that was assigned to Riversleigh Faunal

Zone D.

185

Figure 6.1. Summary of Riversleigh karst stratigraphy from Arena (2004).

Distinction of faunal zones from depositional phases is important because a depositional phase may contain more than one faunal zone or a faunal zone could transcend more than one depositional phase (Arena, 2004). Separation of criteria defining faunal zones from geological criteria avoids estimations of age relationships being made based on circular arguments.

While Riversleigh faunas have been assigned to the broad time periods encompassed by Zones A- D on the basis of their faunal composition, resolution of possible temporal relationships of faunas within those time periods has not yet been achieved. At this stage, no absolute dates for Riversleigh fossil faunas have yet been published. However, a radiometric dating technique currently being applied to

186 Riversleigh deposits has produced a tentative date for Neville‘s Garden Site of 17.89 +/-

0.41 Ma (Graham et al., 2006).

The Numerical Ages method, developed by Legendre and Bachelet (1993), can be used to estimate relative ages of mammalian faunas based on patterns of evolutionary change exhibited by representatives of lineages within those faunas. By using regressions of measurements of mammalian teeth, this method can thus be used to calculate relative ages with an associated statistical error for each fossil fauna.

Numerical ages can be estimated from the relative ages using assigned ages, called bioages, obtained from independent dating methods. As such, this method can be used to assign tentative age relationships to fossil faunas until they can be calibrated using absolute dates.

The method of Legendre and Bachelet (1993), originally tested on French

Palaeogene mammalian faunas, was subsequently extended to most European

Palaeogene faunas by Escarguel et al. (1997) and then calibrated using magnetostratigraphic dates by Legendre and Lévêque (1997). While the numerical ages are relatively well accepted for European faunas, this method has never been applied outside Europe.

Because the Numerical Ages method uses trends in the change in size of homologous teeth over time, it can be used to resolve a sequence of relative ages of specimens according to how their corresponding measurements fit the trend. While the relative sequence produced can therefore be justified provided initial assumptions hold

187 true, in the absence of sufficient absolute dates for specimens in the sequence, age estimation is not possible.

The assumptions of the Numerical Ages method are as follows:

1. Validity of any trend in tooth size over time for a lineage is dependent on the

relative ages assigned to the specimens being correct. While the exact ages of

the time periods they span are yet to be confirmed by absolute dating, the broad

age ranges and sequential relative age relationships of Faunal Zones A-D sensu

(Archer et al., 1989; Archer et al., 1997; Arena, 2004; Travouillon et al., 2006)

are supported by a number of independent analyses of Riversleigh local faunas

and mammalian evolutionary lineages (Black, 1997; Cooke, 1997; Travouillon

et al., 2006; Gillespie, 2007; Black, 2008).

2. The rate of change in size exhibited by each taxon is assumed to be relatively

constant over time because specimens in the analysis are considered to be

representative samples from communities that on average exhibit evolutionary

size-changes at relatively constant rate over time at the community level.

The aims of this study are to:

1. use the Numerical Ages method to resolve a possible sequence of relative

ages for a sample of Riversleigh faunas and thereby improve understanding

of their temporal relationships;

2. use this sequence to test the karst stratigraphic sequence of Arena (2004);

and

3. to demonstrate the applicability of a technique used to establish Land

Mammal Ages on other continents to the Australian fossil record.

188

6.3 MATERIALS AND METHODS

6.3.1 Materials

The Numerical Ages method requires mammalian tooth area data for each identified specimen of each species from each fauna. We compiled a database of dental measurements for non-flying mammalian species (147 species) from 84 Riversleigh faunas (see chapter 3, appendices B, D and E for a complete list of species, sites, specimen numbers and raw measurements). Dental measurements, maximum length and width of each deciduous premolar, premolar and molar, were obtained by direct measurement and from the literature (see chapter 3 and appendix E for a complete list of publication). Dental measurements were taken using a Wild Heerbrugg MMS 235 measuring device attached to a Leica Wild M3B microscope.

6.3.2 Numerical Ages method

We followed the methodology developed by Legendre and Bachelet (1993), and revised by Escarguel et al. (1997). The method uses regressions of teeth area versus a theoretical age, called bioage. The bioage is strictly deducted from a minimal hypothesis of a ”reasonable‘ age for a biostratigraphic zone by taking the middle value of the time interval it is considered to represent. In the case of Riversleigh, previous studies have used biocorrelation to divide sites into 4 faunal zones: Faunal Zone A is late Oligocene,

Faunal Zone B is Early Miocene, Faunal Zone C is Middle Miocene, and Faunal Zone

D is late Miocene (Archer et al., 1989; Archer et al., 1997; Travouillon et al., 2006).

189 Bioages were assigned to each Faunal Zone as follows:

-The bioage of Faunal Zone A was set at 25.72 Ma, the middle of the late

Oligocene (Gradstein et al., 2004).

-The bioage of Faunal Zone B was set at 19.5 Ma, the middle of the early

Miocene (Gradstein et al., 2004).

-The bioage of Faunal Zone C was set at 13.78 Ma, the middle of the middle

Miocene (Gradstein et al., 2004).

-The bioage of Faunal Zone D was set at 8.47 Ma, the middle of the late

Miocene (Gradstein et al., 2004).

Analysis was conducted using the computer program Age_Num.bas, which was custom-designed for this method by Legendre and Bachelet (1993). Numerical ages are calculated using the inverse least squares linear regression from each tooth area of each species versus the bioage and the age of a locality is the mean of all the calculated numerical ages. The probability level of the regression where r is different from 0 was set at p <0.25. This new age calculated from the first regression is then used as an age value in a second step. This step is necessary to reduce temporal biases from the originally assigned values. The final numerical age for the fauna is then estimated from the mean of all calculated ages for each tooth category of each species.

190 Following Escarguel et al. (1997), we assigned a minimum of 3 species and 5 distinct tooth area for the calculation of the average numerical age. Legendre and

Lévêque (1997) pointed out that numerical ages can be calibrated and validated using absolute dates for some localities to find more accurate relative dates for all other localities. Here, we will calibrate and validate Riversleigh numerical ages using two absolute dates from two localities: Neville‘s Garden Site, 17.89 +/- 0.41 Ma (Graham et al., 2006) and, Ngama Local Fauna, 24.8 Ma, biocorrelated with White Hunter Site

(Woodburne et al., 1993; Archer et al., 1997; Myers and Archer, 1997).

The Numerical Ages method uses patterns of change in size occurring within the same mammalian species (Legendre and Bachelet, 1993; Escarguel et al., 1997;

Legendre and Lévêque, 1997). Lineages within species from Riversleigh have not yet been identified in the same way they have been in the European faunas. However, a number of possible evolutionary trends have been identified among successive species within individual marsupial genera. In this study, we use four of these documented intrageneric lineages in place of intraspecific lineages because they represent a succession of individual species through time that are not known to occur in contemporaneous assemblages:

1. Ekaltadeta: Ekaltadeta ima œ Ekaltadeta jamiemulvaneyi (Wroe, 1996)

2. Neohelos: Neohelos tirarensis œ Neohelos stirtoni œ Neohelos sp. C (Murray et al.,

2000; Black, 2008)

191 3. Wakaleo: Wakaleo sp. 1 œ Wakaleo oldfieldi œ Wakaleo vanderleueri (Gillespie,

2007)

4. Yalkaparidon: Yalkaparidon coheni œ Yalkaparidon jonesi (Beck et al., in prep.)

Other taxa included in this analysis were not assigned to lineages due to the uncertainty of their ancestor-descendant relationships. These individual species can still contribute to the Numerical Ages analysis, because some (e.g. Litokoala kutjamarpensis) have been shown to have long fossil records (see chapter 5), and may have changed in size through time. Hence, all Riversleigh taxa have been included in this study, regardless of whether they can be grouped into putative lineages or not.

6.4 RESULTS

The relative age estimates for the Riversleigh faunas are given in Tables 6.1. Of the 84 faunas included in this study, only 40 provided sufficient data (from at least 3 species and 5 tooth categories) to allow estimation of relative age relationships. The majority of these faunas contained numbers of species and/or tooth categories that were barely above the minimum requirements to calculate a numerical age were used in the calculation, decreasing the overall confidence in those relative age estimates. This is reflected in the standard error, where the faunas with the most data, Encore, Neville‘s

Garden, Wayne‘s Wok, Mike‘s Menagerie, Camel Sputum and Upper Site Local Faunas

(LF) have the lowest standard error, while most faunas have standard errors ranging from 0.18 to 3.16, with Jim‘s Jaw Site LF having the highest standard error. Only 16 out of the 27 families included in this study were useful for calculation, with Yaralidae

192 (bandicoots) being the most useful followed by Pseudocheiridae (ringtail possums),

Burramyidae, Phalangeridae and Diprotodontidae. The resulting relative age relationship estimates of the faunas generally agree with the chronological order of the

Riversleigh Faunal Zones A-D and refine the age of the localities

Fireside‘s Favourites LF initially assigned to Zone C is placed younger than

Encore LF (Zone D) by the analysis. While the taxonomic composition of this fauna seems to indicate affinity with Zone C faunas (Travouillon et al., 2006), tooth measurements analysed here suggests affinity with Zone D faunas. However, its placement may be an artefact of the poor sample size (only 3 species present).

Further calibration of the relative ages may not be required as the relative age of

White Hunter Site 24.81 +/- 0.60 Ma matches the age of Ngama Local Fauna, 24.8 Ma and the relative age of Neville‘s Garden Site, 18.95 +/- 0.35 Ma, is almost within the range of its absolute date, 17.89 +/- 0.41 Ma (Graham et al., 2006). While these two dates provide some validation of the age estimates provided by the Numerical Ages method, additional absolute dates are essential for adequate validation of these results.

193

Table 6.1. Numerical ages estimated for the different Riversleigh localities. Faunal Zones, bioages, standard errors (SE), number of tooth categories (N teeth) and the number of species (N sp.) used in the calculation are also provided. Families which contributed to the calculation are also indicated in the last column (Bal: Balbaridae; Bur: Burramyidae; Das: Dasyuridae; Dip: Diprotodontidae; Hyp: Hypsiprymnodontidae; Mac: Macropodidae; Orn: Ornithorhynchidae; Pal: Palorchestidae; Pet: Petauridae; Pha: Phalangeridae; Phc: Phascolarctidae; Pse: Pseudocheiridae; Thy: Thylacoleonidae; Wyn: Wynyardiidae; Yal: Yalkaparidontidae; Yar: Yaralidae).

194 6.5 DISCUSSION

6.5.1 Comparison of the estimated relative age sequence with stratigraphy in previous work and implications for biostratigraphic interpretation

The order of the faunal sequences produced here generally agrees with expectations based on previous biocorrelation and faunal studies (Archer et al., 1989;

Archer et al., 1997; Myers et al., 2001; Myers, 2002), and multivariate analyses by

Travouillon et al. (2006; in prep.). The sequence produced by application of the seriation technique to Riversleigh faunas (Travouillon et al., 2006) is best for direct comparison to the results of this analysis, because the seriation technique provides a relative position in time to all fossil localities based on their faunal composition.

Significantly, the relative positions of the Riversleigh localities in the seriation and in the numerical ages calculated here are quite similar despite the use of very different methodology.

One of the disadvantages of the seriation method used by Travouillon et al.

(2006) is that it does not quantify the magnitude of spacing between the subsequent faunas in the sequence. This means that there is no way to know whether two faunas adjacent in the sequence could be of similar age or if they might be separated by a relatively long period of time. Multivariate analyses do the exact opposite, giving an indication of distance of similarity, however the order is not 2 dimensional and cannot reflect time (Travouillon et al., 2006).

195

Figure 6.2. Seriation of taxa presence within the Riversleigh sites of Table 6.1, scaled in time using the estimated ages of the sites.

196 The Numerical Ages method is useful because it provides both a relative age sequence and scale of the spacing. Providing that the assumptions behind the Numerical

Ages method are true, it can provide useful information that other relative dating method previously used on the Riversleigh localities have been unable to provide.

Travouillon et al.‘s (2006) seriation is modified in Fig. 6.2, to include only sites that have estimated ages (from Table 6.1), and the seriation is scaled in time using the same estimated dates. This scaled seriation provides a visual tool that illustrates differences in species representation through time at Riversleigh; gaps in the sequence indicate time periods from which no faunas for that time period are currently known/dated.

6.5.2 Comparison with the karst stratigraphic sequence of Arena (2004)

The karst stratigraphy of Arena (2004), which is yet to be published, is based on the assumption that time periods represented by Faunal Zones (=“Systems“) A, B and C are correct. The age of each deposit is inferred from the age of the vertebrate fauna it contains. Radiometric dating would be ideal to test the temporal range of each

Depositional Phase but radiometric dating of the Riversleigh sites is still in progress

(Graham et al., 2006).

Because they are dated using Faunal Zones and karst deposits are subject to inherent host/infill relationships where a given deposit must post-date the host deposit, in which it occurs, the Numerical Ages method can be used to test the validity of the

Depositional Phases.

It is therefore possible to test this model using the logical relationships inherent in the stratigraphic sequence. Here, we identify a —Host Phase“ and —Deposit Phase“ for

197 each deposit that has yielded a fauna used in the Numerical Ages analysis. The relative age sequence produced by the Numerical Ages analysis can then be examined to see if the stratigraphic relationships of Arena (2004) are upheld.

For example, a cave deposit assigned to Phase 2 must always be younger than the host sediment it was deposited within. This means that if the host deposit belongs to

Phase 1, then the incised cave must belong to Phase 2, 3 or 4; or if the host sediment belongs to Phase 3, the incised cave must belong to either Phase 3 or 4 (see Fig. 6.1 for a summary of the Depositional Phase model).

The temporal ordering of the Riversleigh faunas, along with the estimated ages using the Numerical Ages method, are represented in Table 6.2, with corresponding lithostratigraphic information for each source locality.

While the Numerical Ages estimates do not completely agree with the

Depositional Phase model, the level of incongruence between these two temporal scenarios is minimal. There is no reason to dismiss either the Numerical Ages estimates or the Depositional Phase model with current knowledge. The results of this analysis provide useful new information about the timing of the depositional phases recognised by Arena (2004).

198 Table 6.2. Comparison of the relative age of the Riversleigh sites given by the Numerical Ages method with the Depositional phase model. Host phase and deposit type (cave or tufa) information were collected from field notes (D. A. Arena Pers. Comm. 2008), while Depositional Phase information gathered from Arena (2004).

199

These results are significant because they suggest vadose-dominated

Depositional Phase 2 conditions might have occurred during Faunal Zone A time in the late Oligocene, at least as early as the age of the Quantum Leap LF, and extended through Faunal Zone B (early Miocene) and into Faunal Zone C (middle Miocene). Due to the uncertainty of the accuracy of its relative age, and lack of lithostratigraphic data, the results for Fireside Favourites Site are not evaluated here.

While the order of the Riversleigh faunas produced by the Numerical Ages method (Table 6.2) is generally congruent with the Depositional Phase model for the

Faunal Zones, a small number of faunas/deposits do not. Quantum Leap Site has been interpreted as a vadose cave deposit assigned to Depositional Phase 2, because it has been deposited within the fluvio-lacustrine tufa assigned to Depositional Phase 1.

Therefore it is expected to be younger than White Hunter Site, which is considered part of the primary fluvio-lacustrine tufa of Depositional Phase 1; however the results of this analysis indicate reverse age relationships (Fig. 6.3).

Similarly, two faunas assigned to Faunal Zone C - Gotham City LF and Last

Minute LF - previously interpreted as Depositional Phase 4 vadose cave deposits formed within fluvio-lacustrine tufa deposits of Depositional Phase 3 are indicated to pre-date, rather than post-date faunas from Gag and Ringtail sites, which were interpreted as having been formed during Depositional Phase 3 (Fig. 6.3).

200

Figure 6.3. Graph of the relative age estimate (in Ma) of the Riversleigh local faunas in Numerical Ages sequence. Red lines show regions of overlap between the relative age of Quantum Leap and White Hunter LFs, and Gotham City and Last Minute LFs Abbreviations: G= G-Site, HS= Hiatus, D= D-Site, QL= Quantum Leap, SB= Sticky Beak, WH= White Hunter, BR= Bone Reef, LSO= Lee Sye‘s Outlook, CR= Creaser‘s Rampart, MM= Mike‘s Menagerie, JH= Judith‘s Horizontalis, RV= Rat Vomit, U= Upper Site, PIR= Price Is Right, Bite= Bitesantennary, RSO= Ross Scott-Orr, WW= Wayne‘s Wok , NG= Neville‘s Garden, MPP= Mike‘s Potato Patch, DT= Dirk‘s Towers, BSE= Boid Site East, CS= Camel Sputum, Ina= Inabeyance, Out= Outasite, KCB= Keith‘s Chocky Block, CK= Cadbury‘s Kingdom, Main= Main Site, GC= Gotham City, AL90= Alan‘s Ledge 1990, Wang= Wang Site, JC= Jim‘s Carousel, Gag= Gag Site, LM= Last Minute, Ring= Ringtail Site, COA= Cleft Of Ages Site, HH= Henk‘s Hollow Site, Dome= Dome Site, JJS= Jim‘s Jaw Site, En= Encore Site and FF= Fireside Favourites.

201 These differences between the karst stratigraphic sequence of Arena (2004) and the relative age estimates produced here by the Numerical Ages method might be explained by the following reasons:

1. Accuracy of the Numerical Ages method - standard error and age-range overlaps

The relative age relationships provided by this analysis might not reflect actual age relationships (or at least those inherent in previous stratigraphic interpretation) due to limitations of the accuracy of method. In a number of cases, including the faunas listed above, the possible age range (including the standard error) of faunas adjacent in the sequence overlaps. As a result, these faunas can be validly reordered within the bounds of each age range estimate into a sequence congruent with the stratigraphy of

Arena (2004).

2. Assumption of the Numerical Ages method regarding constant rate of evolutionary change

The evolutionary history of change in tooth-size in taxa used in this analysis might in reality involve non-constant rates of change over time and/or reversals of size- change trends over time at the population level, which is contrary to one of the starting assumptions of the Numerical Ages method. If this is the case, then the analysis could have produced a sequence matching faunas to a hypothetical regression that does not agree with the physical relationships between deposits used to determine the lithostratigraphic sequence.

202 3. Geological interpretation of source localities

If the results of the Numerical Ages analysis are accurate then this could mean the interpretation of some individual deposits and deposit/host relationships and their resulting assignations to depositional phases might be incorrect. For example, White

Hunter Site could in reality be a Phase 2 vadose cave deposit rather than a Phase 1 fluvio-lacustrine tufa deposit, as has previously been interpreted. If this is the case then the relative age relationships of White Hunter LF and Quantum Leap LF provided by this analysis would be valid. Furthermore - and significantly - this would then provide new information about the timing of the onset of palaeoenvironmental conditions favouring vadose cave processes (e.g. lowering of water table and cessation of widespread fluvio-lacustrine tufa deposition) that distinguish Phase 1 deposits from

Phase 2 deposits. Such a result would indicate that Phase 2 conditions might then have commenced prior to the age of the Quantum Leap LF. Similarly, discrepancies in the resultant ordering of Gotham City, Last Minute, Ringtail and Gag LFs, could be due to incorrect geological interpretation of deposits and/or host/deposit relationships in some cases.

4. Complexity of karst processes and karst stratigraphy

If geological interpretations of source localities of faunas are correct, and the results of the Numerical Ages analysis are accurate, this could indicate that transitions between depositional phases, and conditions occurring during depositional phases, might have been complex rather than being simply dominated by either phreatic or vadose processes. While the karst depositional phases recognised by Arena (2004) are

203 characterised by a dominance of either vadose or phreatic depositional processes sensu

Bosák et al. (1989), this does not exclude other processes from these events. Phreatic and vadose processes can occur concurrently in karst. For example, tufa barrage systems and perched water tables can lead to the formation of localised subaqueous tufa deposits higher in the profile than the regional water table would normally allow, while vadose cave processes can occur simultaneously elsewhere in the karst. Additionally, penecontemporaneous formation of vadose and phreatic deposits can occur if conditions

(e.g. water table level) fluctuate rapidly, possibly due to climatic changes. Arena (2004) noted that the complexity of the deposits at Gag Plateau makes stratigraphic interpretation difficult and that deposits containing middle Miocene faunas assigned to

Phase 3 could indicate either a single or multiple high-water table events resulting in a complicated series of cave and non-cave deposits that requires further detailed investigation.

The results presented here can consequently be further tested and revised as the knowledge of Riversleigh‘s geology and fauna improves.

6.5.3 Applicability of the Neohelos and Wakaleo lineages for biocorrelation between

Riversleigh and other Australian Oligo-Miocene localities.

Index fossils are fossil taxa that identify and date strata or successions of strata in which they are found. Typically they combine morphologic distinctiveness with relatively common occurrence, broad geographic distribution and a restricted stratigraphic range (Salvador, 1994). While the most valuable index fossils are usually planktonic organisms, mammals may also be used as stratigraphic indicators. For example, the European Palaeogene biochronological scale is based on rodents and

204 paleotherid perissodactyls (Legendre and Bachelet, 1993). At this stage, the fossil records of the majority of Australian mammalian lineages are too incomplete for them to be used to erect such a biochronological scale.

However, the two recently documented lineages of Wakaleo (marsupial lion) and Neohelos (diprotodontid) (Murray et al., 2000; Gillespie, 2007; Black, 2008) used in this work, can provide a stratigraphic biocorrelation between Riversleigh and other

Australian Oligo-Miocene localities. Because the successions of species in the Wakaleo and Neohelos lineages occur at different times to each other, this staggering can potentially provide a very useful biocorrelative tool. The co-occurrence of the Wakaleo and Neohelos lineages within some of Riversleigh faunas and their resultant suggested biocorrelation with other Australian Oligo-Miocene localities are shown in Table 6.3.

There are at least five Wakaleo species that represent a single stage-of-evolution lineage: Wakaleo pitikantensis (previously Priscileo pitikantensis), Wakaleo sp. 1,

Wakaleo oldfieldi, Wakaleo vanderleueri and Wakaleo alcootaensis (Clemens and

Plane, 1974; Archer and Rich, 1982; Megirian, 1986; Rauscher, 1987; Murray and

Megirian, 1990, 1992; Gillespie, 2007). Wakaleo pitikantensis is the oldest and most pleisiomorphic of the Wakaleo species, occurring only in the Ngapakaldi Local Fauna of the Etadunna Formation (South Australia), providing no correlation with other Oligo-

Miocene localities (Rauscher, 1987; Gillespie, 2007). Wakaleo sp. 1, Wakaleo oldfieldi and Wakaleo vanderleueri all occur at Riversleigh (see Table 6.3). Wakaleo oldfieldi also occurs in the Kutjamarpu Local Fauna of the Wipajiri Formation in South Australia and Wakaleo vanderleueri also occurs in the Bullock Creek Local Fauna of the

Camfield Beds in the Northern Territory (Clemens and Plane, 1974; Megirian, 1986;

205 Murray and Megirian, 1990, 1992; Gillespie, 2007). The most derived and youngest species of the Wakaleo lineage, Wakaleo alcootaensis occurs in the Alcoota Local

Fauna of the Waite Formation in the Northern Territory (Archer and Rich, 1982;

Gillespie, 2007).

There are at least three species of Neohelos at Riversleigh that represent a single stage-of-evolution lineage (Table 6.3): Neohelos tirarensis, Neohelos stirtoni and

Neohelos sp. C (Stirton, 1967; Murray et al., 2000; Murray et al., 2000; Black, 2008).

Neohelos tirarensis also occurs in the Kutjamarpu Local Fauna and Neohelos stirtoni in the Bullock Creek Local Fauna (Stirton, 1967; Murray et al., 2000; Murray et al., 2000;

Black, 2008). A smaller and pleisiomorphic form of Neohelos tirarensis is found at

Riversleigh in Faunal Zone A, but it was not found to be distinct enough from the later forms to be formally erected as a different species (Murray et al., 2000; Black, 2008). A

Neohelos/Kolopsis species has also been identified from one tooth from Encore Local

Fauna (Faunal Zone D) and may or may not be a part of the lineage (Myers et al., 2001;

Black, 2008).

206 Table 6.3. Biocorrelation between Riversleigh faunas and other Australian Oligo-Miocene localities based on shared taxa including members of Wakaleo and Neohelos lineages. Presence of Wakaleo and Neohelos lineages were taken from Gillespie (2007) and Black (2008). Jaw Junction LF is also included because of the presence of Neohelos sp. C. Presence for all other taxa were taken from Travouillon et al. (2006). LF = Local Fauna.

207 Riversleigh‘s Faunal Zone A contains Wakaleo sp. 1 and the plesiomorphic form of Neohelos tirarensis, Faunal Zone B also contains Wakaleo sp. 1 but has the more derived form of Neohelos tirarensis, which differentiates Faunal Zone B from A. White

Hunter LF (Riversleigh Faunal Zone A) has been biocorrelated with Ngama Local

Fauna from the Etadunna Formation through the presence of the Ilariid Kutjerintja ngama (Table 6.3) (Archer et al., 1997; Myers and Archer, 1997). Neither Wakaleo nor

Neohelos species have been identified from the Ngama Local Fauna, and therefore contribute no further biocorrelation between this local fauna and the Riversleigh localities. Megirian et al. (2004) identified Neohelos tirarensis in the Kangaroo Well

Local Fauna (Table 6.3) from the Ulta Limestone in the Northern Territory. However, this specimen is too fragmentary to assist with biocorrelation and all other taxa present in Kangaroo Well LF are either too fragmented (cf.) or represent new taxa not found in any other localities (Table 6.3).

The occurrence of Wakaleo and Neohelos species lineages in Faunal Zone C of

Riversleigh is more complex. Faunal Zone C is characterised by the presence of

Wakaleo oldfieldi in the Wakaleo lineage and the subsequently presence of Neohelos tirarensis, followed by Neohelos stirtoni and continued with Neohelos sp. C in the

Neohelos lineage.

Wakaleo oldfieldi and Neohelos tirarensis, co-occur in the Keith‘s Chocky

Block LF (a combination previously unknown at Riversleigh), which therefore correlates it with the Kutjamarpu Local Fauna (Gillespie, 2007; Black, 2008). This then suggests the Kutjamarpu Local Fauna is more strongly correlated with Riversleigh

208 Faunal Zone C, which has been interpreted as middle Miocene in age (Archer et al.,

1989; Archer et al., 1997).

Most Faunal Zone C faunas with relative age estimates in this analysis that are younger than Keith‘s Chocky Block LF contain either Wakaleo oldfieldi or Neohelos stirtoni. While the two species have not yet been found together in these sites, this suggests their co-occurrence can be expected. All other taxa correlating Kutjamarpu

Local Fauna and Riversleigh LFs also support a middle Miocene age for Kutjamarpu

LF.

The combination of Wakaleo vanderleueri and Neohelos stirtoni is unique to

Bullock Creek Local Fauna and suggest a younger age than most of Riversleigh‘s

Faunal Zone C faunas (Gillespie, 2007; Black, 2008). Faunal Zone C Jaw Junction LF is unique due to the presence of Neohelos sp. C, which is more derived than Neohelos stirtoni (Murray et al., 2000; Black, 2008). This suggests that Jaw Junction LF is younger than other Riversleigh Zone C faunas, and that there is relatively close temporal relationship between this fauna and Bullock Creek LF because they both share the thylacinid Mutpuracinus archibaldi. The Neohelos sp. present at Encore Site may or may not be part of the Neohelos lineage recognised here, and therefore does not assist with biocorrelation of this site (Myers et al., 2001; Black, 2008). Nevertheless, several mammal species identified at Encore Site, suggest is aged intermediate between the

Bullock Creek Local Fauna/Riversleigh‘s Faunal Zone C and Alcoota Local Fauna

(Myers et al., 2001).

Travouillon et al.‘s (2006) list of taxa that define each of Riversleigh‘s Faunal

Zones can be used to biocorrelated new Riversleigh sites with few identified taxa to a

209 Faunal Zone. As specimens are found, this list will require updating in order to maintain current knowledge. For example, the phascolarctid Litokoala kutjamarpensis (Table 6.3) was previously only known from Faunal Zone C (Archer et al., 2006; Travouillon et al.,

2006) but has now been found in some Faunal Zone B sites (Louys et al., 2007;

Travouillon et al., in prep.). While L. kutjamarpensis should no longer be considered an index taxa for Faunal Zone C, some species do remain excellent index taxa, such as the diprotodontid Nimbadon lavarackorum (Black, 2008) and the ringtail possum Marlu kutjamarpensis (Roberts, 2008) for Faunal Zone C.

6.6 CONCLUSION

The Numerical Ages method was used in this study to provide relative age estimates for some Riversleigh faunas. Trends in tooth size change identified here produce a sequence that generally agrees with that in previously established Faunal

Zones and provides a tentative sequence of relative ages for faunas within those zones.

This sequence generally agrees with the karst stratigraphy of Arena (2004) and helps resolve the timing of events within that sequence. In particular, this analysis suggests that vadose-dominated Depositional Phase 2 conditions might have occurred during

Faunal Zone A time in the late Oligocene, at least as early as the age of the Quantum

Leap LF, and extended through Faunal Zone B (early Miocene) and into Faunal Zone C

(middle Miocene).

This method can be used to help clarify age relationships of the Riversleigh faunas once absolute dates have been established for an adequate number of faunas.

210 Species of Wakaleo and Neohelos have been identified as index taxa that can refine the age of Riversleigh faunas, as their temporal relationships with other

Australian Oligo-Miocene fossil faunas. Recognition of additional mammalian evolutionary lineages may be able to further refine biocorrelation between Oligo-

Miocene fossil sites in Australia.

6.7 REFERENCES

Archer, M., D. A. Arena, M. Bassarova, R. M. D. Beck, K. Black, W. E. Boles, P. Brewer, B. N. Cooke, K. Crosby, A. Gillespie, H. Godthelp, S. J. Hand, B. P. Kear, J. Louys, A. Morrell, J. Muirhead, K. K. Roberts, J. D. Scanlon, K. J. Travouillon and S. Wroe (2006). Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1: 1-17.

Archer, M., H. Godthelp, S. J. Hand and D. Megirian (1989). Fossil mammals of Riversleigh, northwestern Queensland: preliminary overview of biostratigraphy, correlation and environmental change. Australian Zoologist 25: 29-65.

Archer, M., S. Hand and H. Godhelp (1991). Riversleigh - The story of animals in the ancient rainforest of inland Australia. Reed Books, Balgowlah

Archer, M., S. J. Hand and H. Godhelp (1994). Patterns in the history of Australia's mammals and inferences about palaeohabitats. . History of the Australian Vegetation. R. Hill. Cambridge, Cambridge University Press: 80-103.

Archer, M., S. J. Hand and H. Godhelp (1995). Tertiary environmental and biotic change in Australia. Paleoclimate and evolution, with emphasis on human origins. G. H. D. E. Vrba, T.C. Partridge, L.H. Burckle. New Haven, Yale University Press: 77-90.

Archer, M., S. J. Hand, H. Godthelp and P. Creaser (1997). Correlation of the Cainozoic sediments of the Riversleigh World Heritage fossil property, Queensland, Australia. . Actes du congrès BiochroM'97, Mémoires et Travaux de l'Ecole Pratique des Hautes Etudes. J.-P. Aguilar, Legendre, S., Michaux, J. Institut de Montpellier. 21: 131-152.

Archer, M. and T. H. Rich (1982). Results of the Ray E. Lemley Expeditions. Wakaleo alcootaensis n. sp. (Thylacoleonidae: Marsupialia), a new marsupial lion from the Miocene of the Northern Territory, with a consideration of the early radiation of the family. Carnivorous Marsupials. M. Archer. Sydney, Royal Zoological Society of New South Wales: 495-502.

211 Arena, D. A. (2004). The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle Tertiary. PhD thesis, UNSW, Sydney.

Beck, R. M. D., K. J. Travouillon and M. Archer (in prep.). The osteology and systematics of the enigmatic Australian Oligo-Miocene metatherian Yalkaparidon (Yalkaparidontidae; Yalkaparidontia; Australidelphia; Marsupialia).

Black, K. (1997). Diversity and biostratigraphy of the Diprotodontoidea of Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 187-192.

Black, K. (2008). Diversity, phylogeny and biostratigraphy of Diprotodontoids (Marsupialia: Diprotodontidae, Palorchestidae) from the Riversleigh World Heritage Area. PhD, University of New South Wales, Sydney.

Bosák, P., D. C. Ford and J. Glazek (1989). Terminology. Paleokarst. A systematic and regional review. P. Bosák, D. C. Ford, J. Glazek and I. Horálek. Amsterdam-Praha, Elsevier-Academia: 25-32.

Clemens, W. A. and M. Plane (1974). Mid-Tertiary Thylacoleonidae (Marsupialia, Mammalia). Journal of Paleontology 48: 652-660.

Cooke, B. N. (1997). Biostratigraphic implications of fossil kangaroos at Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 295-302.

Creaser, P. (1997). Oligocene-Miocene Sediments of Riversleigh: The potential significance of topography. Memoirs of the Queensland Museum 41(2): 303-314.

Escarguel, G., B. Marandat and S. Legendre (1997). Sur l'âge numérique des faunes de mammifères du Paléogène d'Europe occidentale, en particulier celles de l'Eocène inférieur et moyen. Actes du congrès BiochroM'97, Mémoires et Travaux de l'Ecole Pratique des Hautes Etudes. J.-P. Aguilar, Legendre, S., Michaux, J. Institut de Montpellier. 21: 443-460.

Gillespie, A. K. (2007). Diversity and systematics of marsupial lions from the Riversleigh World Heritage Area and the evolution of the thylacoleonidae. Doctor of Philosophy, University of New South Wales, Sydney.

Gradstein, F., J. Ogg and A. Smith, Eds. (2004). A Geologic time scale 2004. Cambridge University Press.

Graham, I., E. Price, D. Cendón and J. Woodhead (2006). Understanding Riversleigh's geology: what we know in 2006, and where to next. Riversleigh 2006 Symposium. University of New South Wales, Sydney, Australia, The Riversleigh Society Inc.

Legendre, S. and B. Bachelet (1993). The numerical ages: a new method of datation applied to Paleogene mammalian localities from Southern France. Newsletters on Stratigraphy 29: 137-158.

212 Legendre, S. and F. Lévêque (1997). Etalonnage de l'échelle biochronologique mammalienne du Paléogène d'Europe occidentale : vers une intégration à l'échelle globale. Actes du congrès BiochroM‘97, Mémoires et Travaux de l‘Ecole Pratique des Hautes Etudes. L. S. M. J. Aguilar J.-P. Institut de Montpellier. 21: 461-473.

Louys, J., K. Black, M. Archer, S. J. Hand and H. Godhelp (2007). Descriptions of koala fossils from the Miocene of Riversleigh, northwestern Queensland and implications for Litokoala (Marsupialia,Phascolarctidae). Alcheringa 31: 99-110.

Megirian, D. (1986). The dentary of Wakaleo vanderleueri (Thylacoleonidae: Marsupialia). The Beagle. Records of the Museums and Art Galleries of the Northern Territory 3: 71-79.

Megirian, D. (1992). Interpretation of the Carl Creek Limestone, northwestern Queensland. The Beagle: Records of the Northern Territory Museum of Arts and Sciences 9(1): 219-248.

Megirian, D. (1994). Approaches to marsupial biochronology in Australia and New Guinea. Alcheringa 18: 259-274.

Megirian, D., P. Murray, L. Schwartz and C. Von Der Borch (2004). Late Oligocene Kangaroo Well Local Fauna from the Ulta Limestone (new name), and climate of the Miocene oscillation across central Australia. Australian Journal of Earth Sciences 51: 701-741.

Murray, P. and D. Megirian (1990). Further observations on the morphology of Wakaleo vanderleueri (Marsupialia: Thylacoleonidae) from the mid-Miocene Camfield Beds, Northern Territory. The Beagle. Records of the Museums and Art Galleries of the Northern Territory 7: 91-102.

Murray, P. and D. Megirian (1992). Continuity and contrast in middle and late Miocene vertebrate communities from Northern Territory. Proceedings of the 1991 Conference on Australasian Vertebrate Evolution, Palaeontology and Systematics, Alice Springs, The Beagle. Records of the Northern Territory Museum of Arts and Sciences.

Murray, P. and D. Megirian (1992). Continuity and contrast in Middle and Late Miocene vertebrate communities from the Northern Territory. The Beagle. Records of the Museums and Art Galleries of the Northern Territory 9(1): 195-218.

Murray, P., D. Megirian, T. H. Rich, M. Plane, K. Black, M. Archer, S. Hand and P. Vickers-Rich (2000). Morphology, systematics and evolution of the marsupial genus Neohelos Stirton (Diprotodontidae, zygomaturinae). MAGNT Research Report 6: 141 pp.

Murray, P., D. Megirian, T. H. Rich, M. Plane and P. Vickers-Rich (2000). Neohelos stirtoni. A new species of Zygomaturine (Diprotodonta: Marsupialia) from the mid- Tertiary of the Northern Territory, Australia. Records of the Queen Victoria Museum 105: 1-47.

213 Myers, T. J. and M. Archer (1997). Kutjerintja ngama (Marsupialia, Ilariidae): a revised systematic analysis based on material from the late Oligocene of Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum 41(2): 379-392.

Myers, T. J., K. Crosby, M. Archer and M. Tyler (2001). The Encore local Fauna, a late Miocene assemblage from Riversleigh, northwestern Queensland. Memoirs of the Association of Australasian Palaeontologists 25: 147-154.

Myers, T. J. M. (2002). Palaeoecology of Oligo-Miocene Local Faunas from Riversleigh. University of New South Wales, Sydney.

Rauscher, B. (1987). Priscileo pitikantensis, a new genus and species of thylacoleonid marsupial (Marsupialia: Thylacoleonidae) from the Miocene Etadunna Formation, South Australia. Possums and Opossums: Studies in Evolution. M. Archer. Sydney, Surrey Beatty and Sons and the Royal Zoological Society of New South Wales: 423- 432.

Roberts, K. K. (2008). Oligo-Miocene pseudocheirid diversity and the early evolution of ringtail possums (Marsupialia). PhD, University of New South Wales, Sydney.

Salvador, A., Ed. (1994). International Stratigraphic Guide. Second Edition. The International Union of Geological Sciences and The Geological Society of America, Inc., Colorado.

Stirton, R. A. (1967). A diprotodontid from the Miocene Kutjamarpu Fauna, South Australia. Tertiary Diprotodontidae from Australia and New Guinea. R. Stirton, M. Woodburne and M. Plane, Bureau of the Mineral Resources, Geology and Geophysics, Australia, Bulletin 85: 45-51.

Travouillon, K. J., M. Archer, S. J. Hand and H. Godthelp (2006). Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1: 323-349.

Travouillon, K. J., G. Escarguel, S. Legendre, M. Archer and S. J. Hand (in prep.). The use of MSR (Minimum Sample Richness) for fossil fauna comparisons in conjunction with a taxonomic distinctness measure.

Woodburne, M. O., B. J. MacFadden, J. A. Case, M. S. Springer, N. S. Pledge, J. D. Power, J. M. Woodburne and K. B. Springer (1993). Land mammal biostratigraphy and magnetostratigraphy of the Etadunna Formation (Late Oligocene) of South Australia. Journal of Vertebrate Paleontology 13: 483-515.

Wroe, S. (1996). An investigation of phylogeny in the giant extinct rat kangaroo Ekaltadeta (Propleopinae, Potoroidae, Marsupialia). Journal of Paleontology 70(4): 681-690.

214

CHAPTER 7: CONCLUSIONS

215 Rich fossil vertebrate faunas from the Riversleigh World Heritage Area in northwestern Queensland, and particularly its mammalian assemblages, have provided important data for studies of Australian Cenozoic biochronology, palaeoecology and faunal evolution. Nevertheless, as Archer et al. (1995) have pointed out, Australia‘s

Cenozoic fossil record remains patchy and incomplete. Further, in Australia and elsewhere, most palaeoecological and biochronological studies acknowledge the fact that fossil assemblages are typically represented by too few specimens to provide statistically significant interpretations of those palaeocommunities (Hammer and

Harper, 2006), and Riversleigh is no exception. Until now, there has been no methodology available to identify the minimum number of species required in a fossil assemblage to significantly interpret its palaeoecology and biochronology (Travouillon et al. (2006).

The Minimum Sample Richness (MSR) method (Chapters 2 and 3) was developed herein to identify the required minimum number of species in a fossil assemblage for statistically significant interpretations of palaeocommunities. MSR is calculated from estimation of species richness and beta diversity in the palaeocommunities being investigated. The MSR method was tested using data from the

Quercy and Limagne areas of France to identify its strengths and weaknesses. While

MSR removes most of the biases related to incompleteness of the data, taxonomically biased fossil assemblages remained problematic. For this reason, the average taxonomic distinctness and its variation in the fossil assemblages was measured to identify and remove taxonomically biased assemblages. Used together, MSR and taxonomic distinctness methods can screen for fossil assemblages that are not representative of a community (i.e. are biased). These two methods are applied to Riversleigh‘s fossil

216 vertebrate faunas to identify which Riversleigh assemblages are unbiased and therefore suitable for meaningful palaeoecological and biochronological studies. Seventeen

Riversleigh assemblages were identified as being taxonomically unbiased and containing enough taxa to be representative of that palaeocommunity. These Riversleigh faunas were then used in a subsequent palaeoecological study described in Chapter 5.

An alternative use of the MSR and taxonomic distinctness methods is proposed by

Louys et al. (in prep) as a way to test whether the faunal lists of national parks and reserves around the world truly represent those local communities.

In Chapter 4 the Cenogram Method (Legendre, 1986, 1989) was revised for

Australian communities and revealed a number of differences between placental and marsupial modern communities, including the position of the gap in the cenogram pattern for arid communities. This study indicates that the cenogram method alone cannot reliably infer habitat type from visual interpretations. However, when combined with the presence of arboreal taxa and body mass distribution graphs, it may identify habitat type accurately. For modern communities in Australia‘s open habitats, the gap in body mass distributions/cenograms is shifted downwards to lower body weights, from the 8 kg to 500 g range to the 1 kg to 100 g range. These gaps were identified as being the result of environmental pressures including aridity, island effect, habitat fragmentation, edge effect, predator removal, and introduced species. For Australia, patterns in body mass distributions were found to be disrupted by the presence of the

5,320 km long dingo fence. On the western side of the fence, dingoes are present and sheep are absent and here the body mass distribution pattern is similar to patterns observed worldwide for open habitats. However, on the eastern side of the fence, where dingoes are absent and sheep are present, patterns are disrupted by high levels of

217 extinctions in medium-sized and small native species. This study has important implications for modern Australian ecology and conservation. For palaeocommunities, aridity is most likely to be the only source of environmental pressure, because all other identified pressures appear to be the results of human impact.

While the cenogram method has been applied previously to some Riversleigh faunas (Myers, 2002), habitat types could not be successfully identified for most assemblages. The revised cenogram method on selected Oligo-Miocene Riversleigh faunas (see Chapter 3) was, however, able to identify habitat types for those assemblages (Chapter 5). Riversleigh‘s late Oligocene Faunal Zone A was identified as representing an open forest habitat; the early Miocene Faunal Zone B and middle

Miocene Faunal Zone C identified as rainforest habitats; and Riversleigh‘s only early late Miocene Encore assemblage as representative of an open forest habitat.

A technique for estimating relative age relationships of Riversleigh faunas is demonstrated in Chapter 6 as a complementary procedure to the radiometric dating of

Riversleigh fossil deposits that is currently underway (Graham et al., 2006). The

Numerical Ages method uses regressions of dental measurements from fossil specimens as a measure of time. This analysis has provided a sequence of relative ages for 40

Riversleigh fossil faunas, providing the most detailed estimation of biochronology of

Riversleigh faunas thus far. This sequence generally agrees with the karst stratigraphic

(Depositional Phase) and biostratigraphic (Faunal Zone) models of Arena (2004), which provides a robust hypothetical framework for the Cenozoic geological history of

Riversleigh. Furthermore, this sequence can help refine the timing of geological events with the karst stratigraphic sequence, with implications for the interpretation of

218 palaeoenvironmental conditions for some faunas. Once sufficient absolute dates have been obtained, this method can be used to generate a 'calibrated' sequence of faunas from both dated and undated localities.

Species of lineages from genus Wakaleo and Neohelos have been identified and demonstrated to be useful index taxa. They provide biocorrelative links between several

Oligo-Miocene faunas from localities both at Riversleigh and elsewhere in Australia.

The outcomes of this work have the potential to advance the findings of previous

Riversleigh research by helping to clarify the biochronological relationships and palaeoecology of Riversleigh faunas, and to thus improve understanding of Australia‘s complex Oligo-Miocene vertebrate fossil record and the record of climate change during the Cenozoic.

Future research at Riversleigh will focus on the now-possible radiometric dating of many key Riversleigh fossil deposits, and their faunal assemblages. These radiometric dates, combined with detailed biochronological data, may provide sufficient information to establish preliminary Land Mammal Ages (see Gradstein et al., 2004) for

Australia, for at least part of the Cenozoic. Key in this endeavour will be further discovery of Riversleigh and other fossil assemblages to fill the narrowing gaps in the

Australian fossil record and constantly improving biochronologies. The methods developed and revised in this thesis will be particularly important not only in this future

Australian research, but will also have worldwide applicability in palaeoecological as well as modern ecological studies.

219 7.1 REFERENCES

Archer, M., S. J. Hand and H. Godhelp (1995). Tertiary environmental and biotic change in Australia. Paleoclimate and evolution, with emphasis on human origins. G. H. D. E. Vrba, T.C. Partridge, L.H. Burckle. New Haven, Yale University Press, New Haven: 77-90.

Arena, D. A. (2004). The geological history and development of the terrain at the Riversleigh World Heritage Area during the middle Tertiary. PhD thesis, University of New South Wales, Sydney.

Gradstein, F., J. Ogg and A. Smith, Eds. (2004). A Geologic Time Scale 2004. Cambridge University Press, Cambridge.

Graham, I., E. Price, D. Cendón and J. Woodhead (2006). Understanding Riversleigh's geology: what we know in 2006, and where to next. Riversleigh 2006 Symposium. University of New South Wales, Sydney, Australia, The Riversleigh Society Inc., Sydney, Australia.

Hammer, O. and D. A. T. Harper (2006). Paleontological Data Analysis. Blackwell Publishing, Carlton, 351 pp.

Legendre, S. (1986). Analysis of mammalian communities from the late Eocene and Oligocene of Southern France. Palaeovertebrata 16: 191-212.

Legendre, S. (1989). Les communautés de mammifères du Paléogène (Eocène supérieur et Oligocène) d'Europe occidentale: structures, milieux et évolution. Münchner Geowissenschaftliche Abhandlungen (A) 16: 1-110.

Louys, J., K. J. Travouillon, M. Bassarova and H. Tong (in prep). The use of game and nature reserves, national parks in palaeoecological analyses: assumptions, limitations and application.

Myers, T. J. M. (2002). Palaeoecology of Oligo-Miocene Local Faunas from Riversleigh. PhD thesis, University of New South Wales, Sydney.

Travouillon, K. J., M. Archer, S. J. Hand and H. Godthelp (2006). Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1: 323-349.

220

APPENDIX A

SPECIES LIST AND BODY MASS ESTIMATES OF

MODERN NATIONAL PARKS AND RESERVES

221 Abercrombie National Park Thylogale billardierii 5225 Trichosurus vulpecula 2598 Species name Weight (g) Vombatus ursinus 26000 Acrobates pygmaeus 12 Antechinus stuartii 26 Blue Mountains National Park Bos taurus 447617 Capra hircus 50517 Weight Macropus giganteus 15199 Species name (g) Macropus rufogriseus 16607 Acrobates pygmaeus 12 Ornithorhynchus Antechinus flavipes 44 anatinus 1249 Antechinus stuartii 26 Oryctolagus cuniculus 1580 Antechinus swainsonii 52 Petauroides volans 1237 Bos taurus 447617 Petaurus breviceps 127 Canis lupus dingo 13647 Pseudocheirus Capra hircus 50517 peregrinus 878 Cercartetus nanus 24 Rattus rattus 280 Dasyurus maculatus 4583 Sus scrofa 75829 Equus Caballus 511000 Tachyglossus aculeatus 3742 Felis catus 3937 Trichosurus vulpecula 2598 Hydromys chrysogaster 676 Vombatus ursinus 26000 Isoodon obesulus Vulpes vulpes 6111 obesulus 771 Wallabia bicolor 14866 Lepus capensis 4000 Macropus giganteus 15199 Ben Lomond National Park Macropus robustus 17048 Macropus rufogriseus 16607 Weight Mus musculus 16 Species name (g) Ornithorhynchus Antechinus swainsonii 52 anatinus 1249 Bettongia gaimardi 1660 Oryctolagus cuniculus 1580 Cercartetus lepidus 7 Perameles nasuta 967 Cercartetus nanus 24 Petauroides volans 1237 Dasyurus maculatus 4583 Petaurus australis 561 Dasyurus viverrinus 1070 Petaurus breviceps 127 Felis catus 3937 Petaurus norfolcensis 230 Isoodon obesulus 771 Petrogale penicillata 4922 Macropus rufogriseus 16607 Phascolarctos cinereus 5758 Ornithorhynchus Pseudocheirus anatinus 1249 peregrinus 878 Oryctolagus cuniculus 1580 Pseudomys Perameles gunnii 650 novaehollandiae 17 Petaurus breviceps 127 Rattus fuscipes 125 Potorous tridactylus 1097 Rattus lutreolus 120 Pseudocheirus Rattus rattus 280 peregrinus 877 Sminthopsis murina 21 Pseudomys higginsi 67 Sus scrofa 75829 Sarcophilus harrisii 6928 Tachyglossus aculeatus 3742 Sminthopsis leucopus 23 Thylogale thetis 5158

222 Trichosurus vulpecula 2598 Pseudomys Vombatus ursinus 26000 hermannsburgensis 12 Vulpes vulpes 6111 Rattus rattus 280 Wallabia bicolor 14866 Sminthopsis crassicaudata 15 Boodjamulla National Park Sminthopsis murina 21 Tachyglossus aculeatus 3742 Species name Weight (g) Trichosurus vulpecula 2598 Bos taurus 447617 Vulpes vulpes 6111 Canis lupus dingo 13647 Equus Caballus 511000 Bundjalung National Park Felis catus 3937 Hydromys Weight chrysogaster 676 Species name (g) Macropus agilis 14457 Acrobates pygmaeus 12 Macropus antilopinus 25446 Antechinus flavipes 44 Macropus robustus 17048 Antechinus stuartii 26 Macropus rufus 41821 Antechinus swainsonii 52 Mus musculus 16 Bos taurus 447617 Onychogalea Canis lupus 13647 unguifera 6595 Capra hircus 50517 Petrogale Dasyurus maculatus 4583 purpureicollis 5700 Felis catus 3937 Petropseudes dahli 1600 Hydromys chrysogaster 676 Planigale ingrami 4 Isoodon macrourus 1520 Pseudomys delicatulus 9 Lepus capensis 4000 Pseudomys desertor 25 Macropus giganteus 15199 Pseudomys nanus 34 Macropus parryi 13267 Rattus villosissimus 132 Macropus robustus 17048 Sminthopsis macroura 20 Macropus rufogriseus 16607 Sus scrofa 75829 Melomys burtoni 54 Tachyglossus Melomys cervinipes 70 aculeatus 3742 Mus musculus 16 Zyzomys argurus 41 Oryctolagus cuniculus 1580 Perameles nasuta 967 Bookmark Biosphere Reserve Petauroides volans 1237 Petaurus australis 561 Weight Petaurus breviceps 127 Species name (g) Petaurus norfolcensis 230 Capra hircus 50517 Phascogale tapoatafa 170 Hydromys chrysogaster 676 Phascolarctos cinereus 5758 Lepus capensis 4000 Planigale maculata 11 Macropus fuliginosus 12669 Pseudocheirus Macropus robustus 17048 peregrinus 878 Macropus rufus 41821 Pseudomys Oryctolagus cuniculus 1580 novaehollandiae 17 Ovis aries 63246 Rattus fuscipes 125 Pseudomys bolami 14 Rattus lutreolus 120

223 Rattus rattus 280 Pseudomys Rattus tunneyi 76 gracilicaudatus 73 Sminthopsis murina 21 Pseudomys patrius 14 Sus scrofa 75829 Rattus fuscipes 125 Tachyglossus aculeatus 3742 Rattus tunneyi 76 Trichosurus caninus 3354 Sminthopsis murina 21 Trichosurus vulpecula 2598 Sus scrofa 75829 Vulpes vulpes 6111 Tachyglossus aculeatus 3742 Wallabia bicolor 14866 Trichosurus vulpecula 2598 Vulpes vulpes 6111 Carnarvon National Park Wallabia bicolor 14866 Zyzomys argurus 41 Weight Species name (g) Croajingolong National Park Acrobates pygmaeus 12 Aepyprymnus rufescens 3240 Weight Antechinus flavipes 44 Species name (g) Bos taurus 447617 Acrobates pygmaeus 12 Canis lupus dingo 13647 Antechinus stuartii 26 Dasyurus hallucatus 520 Antechinus swainsonii 52 Dasyurus maculatus 4583 Canis lupus 13647 Equus Caballus 511000 Cercartetus nanus 24 Felis catus 3937 Cervus unicolor 167428 Hydromys chrysogaster 676 Dasyurus maculatus 4583 Isoodon macrourus 1520 Felis silvestris 3937 Macropus dorsalis 10198 Hydromys chrysogaster 676 Macropus giganteus 15199 Isoodon obesulus 771 Macropus parryi 13267 Macropus giganteus 15199 Macropus robustus 17048 Macropus rufogriseus 16607 Macropus rufogriseus 16607 Mus musculus 16 Macropus rufus 41821 Ornithorhynchus Melomys cervinipes 70 anatinus 1249 Mus musculus 16 Oryctolagus cuniculus 1580 Ornithorhynchus Perameles nasuta 967 anatinus 1249 Petauroides volans 1237 Oryctolagus cuniculus 1580 Petaurus australis 561 Perameles nasuta 967 Petaurus breviceps 127 Petauroides volans 1237 Phascogale tapoatafa 170 Petaurus australis 561 Phascolarctos cinereus 5758 Petaurus breviceps 127 Potorous longipes 1889 Petaurus norfolcensis 230 Potorous tridactylus 1097 Petrogale herberti 4979 Pseudocheirus Phascolarctos cinereus 5758 peregrinus 878 Planigale maculata 11 Pseudomys fumeus 70 Planigale tenuirostris 6 Rattus fuscipes 125 Pseudocheirus Rattus lutreolus 120 peregrinus 878 Rattus rattus 280 Pseudomys delicatulus 9 Sminthopsis leucopus 23

224 Sus scrofa 75829 Felis catus 3937 Tachyglossus aculeatus 3742 Hydromys chrysogaster 676 Trichosurus caninus 3354 Macropus fuliginosus 12669 Trichosurus vulpecula 2598 Macropus giganteus 15199 Vombatus ursinus 26000 Macropus robustus 17048 Vulpes vulpes 6111 Macropus rufus 41821 Wallabia bicolor 14866 Mus musculus 16 Oryctolagus cuniculus 1580 Deua National Park Phascolarctos cinereus 5758 Planigale tenuirostris 6 Weight Sminthopsis Species name (g) crassicaudata 15 Antechinus stuartii 26 Sminthopsis macroura 20 Dasyurus maculatus 4583 Sus scrofa 75829 Macropus giganteus 15199 Tachyglossus aculeatus 3742 Macropus rufogriseus 16607 Trichosurus vulpecula 2598 Mastacomys fuscus 122 Vulpes vulpes 6111 Ornithorhynchus anatinus 1249 Diamantina National Park Oryctolagus cuniculus 1580 Perameles nasuta 967 Weight Petauroides volans 1237 Species name (g) Petaurus australis 561 Antechinomys laniger 24 Petaurus breviceps 127 Bos taurus 447617 Petaurus norfolcensis 230 Camelus dromedarius 529150 Phascogale tapoatafa 170 Canis lupus dingo 13647 Phascolarctos cinereus 5758 Capra hircus 50517 Potorous tridactylus 1097 Dasyuroides byrnei 110 Pseudocheirus Equus Caballus 511000 peregrinus 878 Felis catus 3937 Pseudomys fumeus 70 Hydromys chrysogaster 676 Pseudomys oralis 95 Leggadina forresti 20 Rattus fuscipes 125 Macropus giganteus 15199 Rattus rattus 280 Macropus robustus 17048 Sminthopsis leucopus 23 Macropus rufus 41821 Sus scrofa 75829 Macrotis lagotis 1414 Tachyglossus aculeatus 3742 Mus musculus 16 Vombatus ursinus 26000 Notomys cervinus 35 Wallabia bicolor 14866 Oryctolagus cuniculus 1580 Planigale gilesi 9 Currawinya National Park Planigale tenuirostris 6 Pseudomys Weight hermannsburgensis 12 Species name (g) Rattus villosissimus 132 Bos taurus 447617 Sminthopsis Canis lupus dingo 13647 crassicaudata 15 Capra hircus 50517 Sminthopsis macroura 20 Equus Caballus 511000 Sus scrofa 75829 Tachyglossus aculeatus 3742

225 Douglas-Apsley National Park Pseudomys albocinereus 30 Weight Pseudomys occidentalis 34 Species name (g) Pseudomys shortridgei 74 Antechinus swainsonii 52 Rattus fuscipes 125 Bettongia gaimardi 1660 Rattus rattus 280 Cercartetus lepidus 7 Sminthopsis granulipes 25 Cercartetus nanus 24 Sminthopsis Dasyurus maculatus 4583 griseoventer 22 Dasyurus viverrinus 1070 Tachyglossus aculeatus 3742 Felis catus 3937 Tarsipes rostratus 10 Isoodon obesulus 771 Trichosurus vulpecula 2598 Macropus rufogriseus 16607 Vulpes vulpes 6111 Ornithorhynchus anatinus 1249 Flinders Ranges National Park Oryctolagus cuniculus 1580 Perameles gunnii 650 Weight Petaurus breviceps 127 Species name (g) Potorous tridactylus 1097 Bos taurus 447617 Pseudocheirus Canis lupus dingo 13647 peregrinus 877 Capra hircus 50517 Pseudomys higginsi 67 Felis catus 3937 Sarcophilus harrisii 6928 Macropus fuliginosus 12669 Sminthopsis leucopus 23 Macropus robustus 17048 Thylogale billardierii 5225 Macropus rufus 41821 Trichosurus vulpecula 2598 Mus musculus 16 Vombatus ursinus 26000 Oryctolagus cuniculus 1580 Ovis aries 63246 Fitzgerald River National Park Petrogale xanthopus 7000 Planigale tenuirostris 6 Weight Sminthopsis Species name (g) crassicaudata 15 Bettongia penicillata 1300 Sminthopsis macroura 20 Bos Taurus 447617 Tachyglossus aculeatus 3742 Cercartetus concinnus 13 Trichosurus vulpecula 2598 Equus caballus 511000 Vulpes vulpes 6111 Felis catus 3937 Hydromys chrysogaster 676 Gawler Ranges National Park Isoodon obesulus 771 Macropus eugenii 6423 Weight Macropus fuliginosus 12669 Species name (g) Macropus irma 8000 Capra hircus 50517 Macrotis lagotis 1414 Cercartetus concinnus 13 Mus musculus 16 Felis catus 3937 Notomys mitchellii 52 Macropus fuliginosus 12669 Oryctolagus cuniculus 1580 Macropus robustus 17048 Parantechinus apicalis 63 Macropus rufus 41821 Phascogale calura 51 Mus musculus 16

226 Notomys mitchellii 52 Vulpes vulpes 6111 Oryctolagus cuniculus 1580 Wallabia bicolor 14866 Ovis aries 63246 Petrogale xanthopus 7000 Gregory National Park Pseudomys hermannsburgensis 12 Weight Sminthopsis Species name (g) crassicaudata 15 Bos taurus 447617 Sminthopsis dolichura 14 Bubalus bubalis 728697 Vulpes vulpes 6111 Canis lupus 13647 Equus asinus 324037 Grampians National Park Equus Caballus 511000 Felis catus 3937 Weight Hydromys chrysogaster 676 Species name (g) Leggadina forresti 20 Acrobates pygmaeus 12 Macropus agilis 14457 Antechinus flavipes 44 Macropus antilopinus 25446 Antechinus stuartii 26 Macropus robustus 17048 Antechinus swainsonii 52 Onychogalea unguifera 6595 Capra hircus 50517 Petaurus breviceps 127 Cercartetus nanus 24 Petrogale brachyotis 4500 Cervus elaphus 120565 Planigale maculata 11 Felis catus 3937 Pseudomys delicatulus 9 Hydromys chrysogaster 676 Pseudomys nanus 34 Isoodon obesulus 771 Rattus villosissimus 132 Macropus fuliginosus 12669 Sminthopsis macroura 20 Macropus giganteus 15199 Tachyglossus aculeatus 3742 Macropus rufogriseus 16607 Zyzomys argurus 41 Mus musculus 16 Ornithorhynchus Gundabooka National Park anatinus 1249 Oryctolagus cuniculus 1580 Weight Petaurus breviceps 127 Species name (g) Petaurus norfolcensis 230 Antechinomys laniger 24 Petrogale penicillata 4922 Canis lupus 13647 Phascolarctos cinereus 5758 Capra hircus 50517 Potorous tridactylus 1097 Felis catus 3937 Pseudocheirus Leporillus apicalis 150 peregrinus 878 Leporillus conditor 350 Pseudomys fumeus 70 Macropus fuliginosus 12669 Pseudomys shortridgei 74 Macropus giganteus 15199 Rattus lutreolus 120 Macropus robustus 17048 Rattus rattus 280 Macropus rufus 41821 Sminthopsis Mus musculus 16 crassicaudata 15 Oryctolagus cuniculus 1580 Sus scrofa 75829 Petrogale penicillata 4922 Tachyglossus aculeatus 3742 Sminthopsis murina 21 Trichosurus vulpecula 2598 Sus scrofa 75829

227 Tachyglossus aculeatus 3742 Kakadu National Park Trichosurus vulpecula 2598 Vulpes vulpes 6111 Weight Species name (g) Iron Range National Park Antechinus bellus 43 Bos taurus 447617 Weight Bubalus bubalis 728697 Species name (g) Canis lupus 13647 Acrobates pygmaeus 12 Conilurus penicillatus 150 Antechinus leo 50 Dasyurus hallucatus 520 Bos taurus 447617 Equus asinus 324037 Canis lupus dingo 13647 Equus Caballus 511000 Dactylopsila trivirgata 315 Felis catus 3937 Dasyurus hallucatus 520 Hydromys chrysogaster 676 Echymipera rufescens 1000 Isoodon auratus 423 Equus Caballus 511000 Isoodon macrourus 1520 Felis catus 3937 Lagorchestes Hydromys chrysogaster 676 conspicillatus 2683 Isoodon macrourus 1520 Leggadina Isoodon obesulus lakedownensis 17 peninsulae 771 Macropus agilis 14457 Macropus agilis 14457 Macropus antilopinus 25446 Macropus antilopinus 25446 Macropus bernardus 16912 Melomys burtoni 54 Macropus robustus 17048 Melomys capensis 110 Melomys burtoni 54 Mesembriomys gouldii 626 Mesembriomys gouldii 626 Perameles nasuta 967 Mesembriomys Petaurus breviceps 127 macrurus 267 Petrogale coenensis 4472 Notomys alexis 35 Phalanger Onychogalea unguifera 6595 intercastellanus 1817 Parantechinus bilarni 23 Phascogale tapoatafa 170 Petaurus breviceps 127 Pogonomys Petrogale (Perodorcas) mollipilosus 62 concinna 1349 Pseudomys delicatulus 9 Petrogale brachyotis 4500 Rattus fuscipes 125 Petropseudes dahli 1600 Rattus leucopus 120 Phascogale tapoatofa Rattus sordidus 154 pirata 170 Rattus tunneyi 76 Planigale ingrami 4 Sminthopsis archeri 16 Planigale maculata 11 Sminthopsis virginiae 40 Pseudomys calabyi 19 Spilocuscus maculatus 2324 Pseudomys delicatulus 9 Sus scrofa 75829 Pseudomys nanus 34 Tachyglossus aculeatus 3742 Rattus colletti 61 Thylogale stigmatica 4628 Rattus rattus 280 Uromys Rattus tunneyi 76 caudimaculatus 541 Sminthopsis bindi 11 Wallabia bicolor 14866 Sminthopsis virginiae 40

228 Sus scrofa 75829 conspicillatus Tachyglossus aculeatus 3742 Leggadina Trichosurus vulpecula 2598 lakedownensis 17 Xeromys myoides 56 Macropus robustus 17048 Zyzomys argurus 41 Macropus rufus 41821 Zyzomys maini 94 Macrotis lagotis 1414 Mus musculus 16 Kalbarri National Park Ningaui ridei 9 Ningaui timealeyi 4 Weight Notomys alexis 35 Species name (g) Notomys longicaudatus 100 Canis lupus dingo 13647 Petrogale rothschildi 5250 Capra hircus 50517 Pseudantechinus Felis catus 3937 woolleyae 19 Macropus eugenii 6423 Pseudomys chapmani 10 Macropus fuliginosus 12669 Pseudomys desertor 25 Macropus robustus 17048 Pseudomys Macropus rufus 41821 hermannsburgensis 12 Mus musculus 16 Pseudomys nanus 34 Notomys alexis 35 Rattus tunneyi 76 Pseudomys Sminthopsis macroura 20 albocinereus 30 Sminthopsis ooldea 11 Rattus rattus 280 Sminthopsis youngsoni 10 Sminthopsis Trichosurus crassicaudata 15 arnhemensis 1442 Sminthopsis dolichura 14 Trichosurus vulpecula 2598 Sminthopsis granulipes 25 Vulpes vulpes 6111 Sminthopsis hirtipes 16 Zyzomys argurus 41 Sus scrofa 75829 Tachyglossus aculeatus 3742 Kinchega National Park Tarsipes rostratus 10 Trichosurus vulpecula 2598 Weight Vulpes vulpes 6111 Species name (g) Antechinomys laniger 24 Karijini National Park Capra hircus 50517 Equus asinus 324037 Weight Felis catus 3937 Species name (g) Hydromys chrysogaster 676 Bos taurus 447617 Macropus fuliginosus 12669 Camelus dromedarius 529150 Macropus giganteus 15199 Canis lupus dingo 13647 Macropus robustus 17048 Dasykaluta Macropus rufus 41821 rosamondae 28 Macrotis lagotis 1414 Dasyurus hallucatus 520 Mus musculus 16 Equus asinus 324037 Oryctolagus cuniculus 1580 Equus Caballus 511000 Planigale gilesi 9 Felis catus 3937 Planigale tenuirostris 6 Lagorchestes 2683 Sminthopsis 15

229 crassicaudata Tachyglossus aculeatus 3742 Sus scrofa 75828 Trichosurus caninus 3354 Tachyglossus aculeatus 3742 Trichosurus vulpecula 2598 Trichosurus vulpecula 2598 Vombatus ursinus 26000 Vulpes vulpes 6111 Vulpes vulpes 6111 Wallabia bicolor 14866 Kosciuszko National Park Ku-ring-gai Chase National Park Weight Species name (g) Weight Antechinus agilis 26 Species name (g) Antechinus flavipes 44 Acrobates pygmaeus 12 Antechinus stuartii 26 Antechinus stuartii 26 Antechinus swainsonii 52 Canis lupus Dingo 13647 Bos Taurus 447617 Capra hircus 50517 Burramys parvus 42 Cercartetus nanus 24 Canis lupus 13647 Cervus elaphus 120565 Capra hircus 50517 Dasyurus maculatus 4583 Cercartetus nanus 24 Equus Caballus 511000 Cervus unicolor 167428 Felis catus 3937 Dasyurus maculatus 4583 Hydromys chrysogaster 676 Dasyurus viverrinus 1070 Isoodon macrourus 1520 Equus caballus 511000 Isoodon obesulus Felis catus 3937 obesulus 771 Hydromys chrysogaster 676 Macropus robustus 17048 Isoodon obesulus 771 Macropus rufogriseus 16607 Lepus capensis 4000 Mus musculus 16 Macropus giganteus 15199 Ornithorhynchus Macropus robustus 17048 anatinus 1249 Macropus rufogriseus 16607 Oryctolagus cuniculus 1580 Mastacomys fuscus 122 Perameles nasuta 967 Mus musculus 16 Petaurus breviceps 127 Ornithorhynchus Phascolarctos cinereus 5758 anatinus 1249 Pseudocheirus Oryctolagus cuniculus 1580 peregrinus 878 ovis aries 63246 Pseudomys Perameles nasuta 967 novaehollandiae 17 Petauroides volans 1237 Rattus fuscipes 125 Petaurus Australis 561 Rattus lutreolus 120 Petaurus breviceps 127 Rattus norvegicus 320 Phascogale tapoatafa 170 Rattus rattus 280 Phascolarctos cinereus 5758 Tachyglossus aculeatus 3742 Pseudocheirus Trichosurus vulpecula 2598 peregrinus 878 Vombatus ursinus 26000 Pseudomys fumeus 70 Vulpes vulpes 6111 Rattus fuscipes 125 Wallabia bicolor 14866 Rattus rattus 280 Sus scrofa 75829

230 Lamington National Park Wallabia bicolor 14866

Weight Little Desert National Park Species name (g) Acrobates pygmaeus 12 Weight Aepyprymnus rufescens 3240 Species name (g) Antechinus flavipes 44 Capra hircus 50517 Antechinus subtropicus 26 Cercartetus concinnus 13 Antechinus swainsonii 52 Cercartetus lepidus 7 Canis lupus dingo 13647 Dama dama 47350 Cercartetus nanus 24 Hydromys chrysogaster 676 Dasyurus maculatus 4583 Lepus capensis 4000 Felis catus 3937 Macropus fuliginosus 12669 Hydromys chrysogaster 676 Macropus rufogriseus 16607 Isoodon macrourus 1520 Mus musculus 16 Lepus capensis 4000 Ornithorhynchus Macropus dorsalis 10198 anatinus 1249 Macropus parryi 13267 Oryctolagus cuniculus 1580 Macropus rufogriseus 16607 Pseudomys Melomys cervinipes 70 apodemoides 20 Mus musculus 16 Sminthopsis Ornithorhynchus crassicaudata 15 anatinus 1249 Tachyglossus aculeatus 3742 Perameles nasuta 967 Trichosurus vulpecula 2598 Petauroides volans 1237 Vulpes vulpes 6111 Petaurus australis Wallabia bicolor 14866 australis 561 Petaurus breviceps 127 Main Range National Park Petaurus norfolcensis 230 Petrogale penicillata 4922 Weight Phascogale tapoatafa 170 Species name (g) Phascolarctos cinereus 5758 Acrobates pygmaeus 12 Planigale maculata 11 Aepyprymnus rufescens 3240 Potorous tridactylus 1097 Antechinus flavipes 44 Pseudocheirus Antechinus subtropicus 26 peregrinus 878 Canis lupus dingo 13647 Pseudomys Dasyurus maculatus 4583 gracilicaudatus 73 Felis catus 3937 Pseudomys oralis 95 Hydromys chrysogaster 676 Rattus fuscipes 125 Isoodon macrourus 1520 Rattus lutreolus 120 Macropus dorsalis 10198 Sminthopsis murina 21 Macropus giganteus 15199 Tachyglossus aculeatus 3742 Macropus parryi 13267 Thylogale stigmatica 4628 Macropus robustus 17048 Thylogale thetis 5158 Macropus rufogriseus 16607 Trichosurus caninus 3354 Melomys cervinipes 70 Trichosurus vulpecula 2598 Mus musculus 16 Vulpes vulpes 6111 Ornithorhynchus 1249

231 anatinus Pseudomys chapmani 10 Perameles nasuta 967 Pseudomys delicatulus 9 Petauroides volans 1237 Pseudomys Petaurus australis 561 hermannsburgensis 12 Petaurus breviceps 127 Rattus rattus 280 Petaurus norfolcensis 230 Sminthopsis macroura 20 Petrogale penicillata 4922 Vulpes vulpes 6111 Phascogale tapoatafa 170 Zyzomys argurus 41 Phascolarctos cinereus 5758 Potorous tridactylus 1097 Mount Barney National Park Pseudocheirus peregrinus 878 Weight Pseudomys Species name (g) gracilicaudatus 73 Acrobates pygmaeus 12 Pseudomys Aepyprymnus rufescens 3240 novaehollandiae 17 Antechinus flavipes 44 Pseudomys oralis 95 Antechinus subtropicus 26 Rattus fuscipes 125 Bos taurus 447617 Rattus lutreolus 120 Canis lupus 13647 Rattus rattus 280 Cercartetus nanus 24 Rattus tunneyi 76 Dasyurus maculatus 4583 Sminthopsis murina 21 Felis catus 3937 Tachyglossus aculeatus 3742 Isoodon macrourus 1520 Thylogale stigmatica 4628 Lepus capensis 4000 Thylogale thetis 5158 Macropus dorsalis 10198 Trichosurus caninus 3354 Macropus giganteus 15199 Trichosurus vulpecula 2598 Macropus parryi 13267 Wallabia bicolor 14866 Macropus robustus 17048 Macropus rufogriseus 16607 Millstream-Chichester National Park Melomys cervinipes 70 Mus musculus 16 Weight Ornithorhynchus Species name (g) anatinus 1249 Canis lupus dingo 13647 Perameles nasuta 967 Dasykaluta Petauroides volans 1237 rosamondae 28 Petaurus breviceps 127 Dasyurus hallucatus 520 Petaurus norfolcensis 230 Felis catus 3937 Petrogale penicillata 4922 Leggadina Phascogale tapoatafa 170 lakedownensis 17 Phascolarctos cinereus 5758 Macropus robustus 17048 Pseudocheirus Macropus rufus 41821 peregrinus 878 Mus musculus 16 Rattus fuscipes 125 Ningaui timealeyi 4 Rattus lutreolus 120 Petrogale rothschildi 5250 Rattus rattus 280 Pseudantechinus roryi 30 Sminthopsis murina 21 Pseudantechinus Tachyglossus aculeatus 3742 woolleyae 19 Thylogale thetis 5158

232 Trichosurus caninus 3354 Potorous tridactylus 1097 Trichosurus vulpecula 2598 Pseudocheirus Wallabia bicolor 14866 peregrinus 877 Pseudomys higginsi 67 Mount Buffalo National Park Rattus lutreolus 120 Sarcophilus harrisii 6928 Weight Tachyglossus aculeatus 3742 Species name (g) Thylogale billardierii 5225 Acrobates pygmaeus 12 Trichosurus vulpecula 2598 Antechinus agilis 26 Vombatus ursinus 26000 Antechinus swainsonii 52 Canis familiaris 13647 Mount Remarkable National Park Cercartetus nanus 24 Felis catus 3937 Weight Hydromys chrysogaster 676 Species name (g) Oryctolagus cuniculus 1580 Canis familiaris 13647 Petauroides volans 1237 Capra hircus 50517 Petaurus australis 561 Dasyurus geoffroii 1051 Petaurus breviceps 127 Felis catus 3937 Pseudocheirus Macropus euginii 6423 peregrinus 878 Macropus fuliginosus 12669 Rattus fuscipes 125 Macropus robustus 17048 Tachyglossus aculeatus 3742 Macropus rufus 41821 Trichosurus caninus 3354 Mus musculus 16 Trichosurus vulpecula 2598 Oryctolagus cuniculus 1580 Vombatus ursinus 26000 Petrogale xanthopus. 7000 Vulpes vulpes 6111 Planigale tenuirostris 6 Wallabia bicolor 14866 Pseudocheirus peregrinus 877 Mount Field National Park Rattus rattus 280 Tachyglossus aculeatus 3742 Weight Trichosurus vulpecula 2598 Species name (g) Vulpes vulpes 6111 Bettongia gaimardi 1660 Cercartetus lepidus 7 Mungkan Kandju National Park Cercartetus nanus 24 Dasyurus maculatus 4583 Weight Dasyurus viverrinus 1070 Species name (g) Felis catus 3937 Bos taurus 447617 Hydromys chrysogaster 676 Canis lupus dingo 13647 Isoodon obesulus 771 Dactylopsila trivirgata 315 Macropus rufogriseus 16607 Dasyurus hallucatus 520 Mus musculus 16 Equus Caballus 511000 Ornithorhynchus Felis catus 3937 anatinus 1249 Hydromys chrysogaster 676 Orystolagus cuniculus 1580 Isoodon macrourus 1520 Perameles gunnii 650 Lagorchestes Petaurus breviceps 127 conspicillatus 2683

233 Macropus agilis 14457 Macropus fuliginosus 12669 Macropus antilopinus 25446 Macropus giganteus 15199 Macropus giganteus 15199 Macropus robustus 17048 Melomys burtoni 54 Macropus rufus 41821 Melomys capensis 110 Mus musculus 16 Mesembriomys gouldii 626 Notomys cervinus 35 Petaurus breviceps 127 Notomys fuscus 35 Planigale maculata 11 Notomys longicaudatus 100 Pseudocheirus Oryctolagus cuniculus 1580 peregrinus 878 Perameles bougainville 220 Rattus sordidus 154 Petrogale xanthopus 7000 Spilocuscus maculatus 2324 Planigale gilesi 9 Sus scrofa 75829 Planigale tenuirostris 6 Tachyglossus aculeatus 3742 Pseudomys australis 65 Trichosurus vulpecula 2598 Pseudomys bolami 14 Uromys Pseudomys gouldii 50 caudimaculatus 541 Pseudomys hermannsburgensis 12 Mungo National Park Rattus villosissimus 132 Sminthopsis Weight crassicaudata 15 Species name (g) Sminthopsis macroura 20 Capra hircus 50517 Tachyglossus aculeatus 3742 Felis catus 3937 Vulpes vulpes 6111 Macropus fuliginosus 12669 Macropus giganteus 15199 Nitmiluk National Park Macropus rufus 41821 Mus musculus 16 Weight Ningaui yvonneae 6 Species name (g) Oryctolagus cuniculus 1580 Bubalus bubalis 728697 Rattus villosissimus 132 Bos taurus 447617 Sminthopsis Canis lupus 13647 crassicaudata 15 Dasyurus hallucatus 520 Sminthopsis murina 21 Equus asinus 324037 Tachyglossus aculeatus 3742 Equus Caballus 511000 Vulpes vulpes 6111 Felis catus 3937 Hydromys chrysogaster 676 Mutawintji National Park Isoodon macrourus 1520 Lagorchestes Weight conspicillatus 2683 Species name (g) Leggadina forresti 20 Capra hircus 50517 Macropus agilis 14457 Canis lupus 13647 Macropus antilopinus 25446 Dasycercus cristicauda 101 Macropus bernardus 16912 Felis catus 3937 Macropus robustus 17048 Isoodon auratus 423 Melomys burtoni 54 Leggadina forresti 20 Mus musculus 16 Leporillus conditor 350 Onychogalea unguifera 6595

234 Parantechinus bilarni 23 Purnululu National Park Petaurus breviceps 127 Petrogale brachyotis 4500 Weight Plangale maculata 11 Species name (g) Pseudocheirus dahli 1600 Bubalus bubalis 728697 Pseudomys delicatulus 9 Bos taurus 447617 Pseudomys nanus 34 Camelus dromedarius 529150 Rattus rattus 280 Canis lupus 13647 Rattus tunneyi 76 Dasyurus hallucatus 520 Sminthopsis bindi 11 Equus asinus 324037 Sminthopsis virginiae 40 Equus Caballus 511000 Sus scrofa 75829 Felis catus 3937 Tachyglossus aculeatus 3742 Hydromys Trichosurus vulpecula 2598 chrysogaster 676 Zyzomys argurus 41 Lagorchestes conspicillatus 2683 Prince Regent River Nature Reserve Leggadina forresti 20 Macropus agilis 14457 Weight Macropus robustus 17048 Species name (g) Macrotis lagotis 1414 Bos Taurus 447617 Onychogalea Canis lupus 13647 unguifera 6595 Conilurus penicillatus 150 Petrogale brachyotis 4500 Dasyurus hallucatus 520 Petropseudes dahli 1600 Felis catus 3937 Planigale ingrami 4 Hydromys chrysogaster 676 Planigale maculata 11 Isoodon auratus 423 Pseudantechinus Isoodon macrourus 1520 ningbing 19 Macropus agilis 14457 Pseudomys Macropus antilopinus 25446 delicatulus 9 Macropus robustus 17048 Pseudomys desertor 25 Mesembriomys Pseudomys laborifex 16 macrurus 267 Pseudomys nanus 34 Petrogale brachyotis 4500 Rattus tunneyi 76 Petropseudes dahli 1600 Sminthopsis Pseudomys delicatulus 9 macroura 20 Pseudomys nanus 34 Sus scrofa 75829 Rattus tunneyi 76 Tachyglossus aculeatus 3742 Tachyglossus aculeatus 3742 Zyzomys argurus 41 Wyulda squamicaudata 1643

Zyzomys argurus 41 Savage River National Park Zyzomys woodwardi 114

Weight

Species name (g)

Antechinus minimus 52

Antechinus swainsonii 52

235 Cercartetus lepidus 7 Macropus parryi 13267 Cercartetus nanus 24 Macropus robustus 17048 Dasyurus maculatus 4583 Macropus rufogriseus 16607 Dasyurus viverrinu 1070 Melomys burtoni 54 Hydromys Melomys cervinipes 70 chrysogaster 676 Mus musculus 16 Isoodon obesulus 771 Ornithorhynchus Macropus rufogriseus 16607 anatinus 1249 Mastacomys fuscus 122 Oryctolagus Ornithorhynchus cuniculus 1580 anatinus 1249 Petauroides volans 1237 Perameles gunnii 650 Petaurus australis 561 Petaurus breviceps 127 Petaurus breviceps 127 Potorous tridactylus 1097 Petaurus norfolcensis 230 Pseudocheirus Petrogale inornata 4170 peregrinus 877 Phascogale tapoatafa 170 Pseudomys higginsi 67 Phascolarctos Rattus lutreolus 120 cinereus 5758 Sarcophilus harrisii 6928 Planigale maculata 11 Sminthopsis leucopus 23 Pseudocheirus Tachyglossus peregrinus 878 aculeatus 3742 Pseudomys Thylogale billardierii 5225 delicatulus 9 Trichosurus Pseudomys vulpecula 2598 gracilicaudatus 73 Vombatus ursinus 26000 Rattus tunneyi 76 Sus scrofa 75829 Shoalwater and Corio Bays Area Tachyglossus Ramsar Site aculeatus 3742 Trichosurus Weight vulpecula 2598 Species name (g) Vulpes vulpes 6111 Aepyprymnus Wallabia bicolor 14866 rufescens 3240 Bos taurus 447617 Simpson Desert National Park Canis lupus 13647 Capra hircus 50517 Weight Dasyurus hallucatus 520 Species name (g) Equus Caballus 511000 Bos taurus 447617 Felis catus 3937 Camelus dromedarius 529150 Hydromys Canis lupus dingo 13647 chrysogaster 676 Dasyuroides byrnei 110 Isoodon macrourus 1520 Felis catus 3937 Lagorchestes Macropus rufus 41821 conspicillatus 2683 Mus musculus 16 Macropus agilis 14457 Ningaui ridei 9 Macropus dorsalis 10198 Notomys alexis 35 Macropus giganteus 15199 Notomys fuscus 35

236 Oryctolagus Sminthopsis leucopus 23 cuniculus 1580 Tachyglossus aculeatus 3742 Pseudomys Trichosurus caninus 3354 hermannsburgensis 12 Trichosurus vulpecula 2598 Sminthopsis Vombatus ursinus 26000 crassicaudata 15 Vulpes vulpes 6111 Sminthopsis Wallabia bicolor 14866 youngsoni 10 South East Forest National Park Snowy River National Park Species name Weight (g) Weight Acrobates pygmaeus 12 Species name (g) Antechinus stuartii 26 Acrobates pygmaeus 12 Antechinus Antechinus agilis 26 swainsonii 52 Antechinus swainsonii 52 Bos taurus 447617 Bos taurus 447617 Canis lupus 13647 Canis familiaris 13647 Capra hircus 50517 Capra hircus 50517 Cercartetus nanus 24 Cercartetus nanus 24 Cervus unicolor 167428 Cervus unicolor 167428 Dasyurus maculatus 4583 Dasyurus maculatus 4583 Felis catus 3937 Dasyurus viverrinus 1070 Hydromys Equus Caballus 511000 chrysogaster 676 Felis catus 3937 Isoodon obesulus 771 Hydromys chrysogaster 676 Macropus giganteus 15199 Isoodon obesulus 771 Macropus robustus 17048 Macropus giganteus 15199 Macropus rufogriseus 16607 Macropus robustus 17048 Ornithorhynchus Macropus rufogriseus 16607 anatinus 1249 Mastacomys fuscus 122 Oryctolagus Mus musculus 16 cuniculus 1580 Ornithorhynchus Ovis aries 63246 anatinus 1249 Perameles nasuta 967 Oryctolagus cuniculus 1580 Petauroides volans 1237 Ovis aries 63246 Petaurus australis 561 Perameles nasuta 967 Petaurus breviceps 127 Petauroides volans 1237 Petrogale penicillata 4922 Petaurus australis 561 Phascogale tapoatafa 170 Petaurus breviceps 127 Phascolarctos Petrogale penicillata 4922 cinereus 5758 Phascolarctos cinereus 5758 Potorous longipes 1889 Potorous longipes 1889 Potorous tridactylus 1097 Pseudocheirus Pseudocheirus peregrinus 878 peregrinus 878 Pseudomys fumeus 70 Pseudomys fumeus 70 Rattus fuscipes 125 Rattus fuscipes 125 120 Rattus lutreolus Rattus lutreolus 120 Rattus rattus 280

237 Rattus rattus 280 Uluru National Park (historical) Sminthopsis leucopus 23 Sus scrofa 75829 Weight Tachyglossus Species name (g) aculeatus 3742 Antechinomys laniger 24 Trichosurus caninus 3354 Bettongia lesueur 1500 Trichosurus Bettongia penicillata 1300 vulpecula 2598 Canis lupus 13647 Vombatus ursinus 26000 Chaeropus ecaudatus 419 Vulpes vulpes 6111 Dasycercus cristicauda 101 Wallabia bicolor 14866 Dasyurus geoffroii 1051 Isoodon auratus 423 Stirling Range National Park Lagorchestes asomatus 1500 Lagorchestes Species name Weight (g) conspicillatus 2683 Antechinus flavipes 44 Lagorchestes hirsutus 1658 Bettongia lesueur 1500 Leggadina forresti 20 Canis lupus 13647 Leporillus apicalis 150 Capra hircus 50517 Macropus robustus 17048 Dasyurus geoffroii 1051 Macropus rufus 41821 Felis catus 3937 Macrotis lagotis 1414 Isoodon obesulus 771 Macrotis leucura 368 Macropus eugenii 6423 Myrmecobius fasciatus 454 Macropus fuliginosus 12669 Ningaui ridei 9 Macropus irma 8000 Notomys alexis 35 Macrotis lagotis 1414 Notomys amplus 100 Mus musculus 16 Notomys longicaudatus 100 Myrmecobius fasciatus 454 Notoryctes typhlops 53 Onychogalea lunata 3500 Onychogalea lunata 3500 Oryctolagus cuniculus 1580 Perameles eremiana 760 Parantechinus apicalis 63 Petrogale lateralis 4656 Perameles bougainville 220 Phascogale calura 51 Phascogale calura 51 Pseudantechinus Phascogale tapoatafa 170 macdonnellensis 30 Pseudocheirus Pseudomys desertor 25 occidentalis 1000 Pseudomys fieldi 39 Rattus fuscipes 125 Pseudomys Rattus rattus 280 hermannsburgensis 12 Setonix brachyurus 3231 Rattus villosissimus 132 Sus scrofa 75829 Sminthopsis hirtipes 16 Tarsipes rostratus 10 Sminthopsis macroura 20 Trichosurus vulpecula 2598 Sminthopsis ooldea 11 Vulpes vulpes 6111 Sminthopsis psammophila 36 Sminthopsis youngsoni 10 Tachyglossus aculeatus 3742 Trichosurus vulpecula 2598 Zyzomys pedunculatus 92

238 Uluru National Park (modern) Pseudomys hermannsburgensis 12 Weight Sminthopsis Species name (g) crassicaudata 15 Camelus dromedarius 529150 Sminthopsis macroura 20 Canis lupus 13647 Sminthopsis murina 21 Dasycercus cristicauda 101 Tachyglossus aculeatus 3742 Felis catus 3937 Vulpes vulpes 6111 Macropus robustus 17048 Macropus rufus 41821 Wadbilliga National Park Mus musculus 16 Ningaui ridei 9 Weight Notomys alexis 35 Species name (g) Notoryctes typhlops 53 Acrobates pygmaeus 12 Oryctolagus cuniculus 1580 Antechinus stuartii 26 Pseudantechinus Antechinus swainsonii 52 macdonnellensis 30 Canis lupus 13647 Pseudomys desertor 25 Dasyurus maculatus 4583 Pseudomys Felis catus 3937 hermannsburgensis 12 Macropus giganteus 15199 Rattus villosissimus 132 Macropus rufogriseus 16607 Sminthopsis hirtipes 16 Ornithorhynchus Sminthopsis macroura 20 anatinus 1249 Sminthopsis ooldea 11 Oryctolagus cuniculus 1580 Sminthopsis Perameles nasuta 967 psammophila 36 Petauroides volans 1237 Sminthopsis youngsoni 10 Petaurus australis 561 Tachyglossus aculeatus 3742 Petaurus breviceps 127 Vulpes vulpes 6111 Phascogale tapoatafa 170 Phascolarctos cinereus 5758 Vulkathunha - Gammon Ranges Pseudocheirus National Park peregrinus 878 Rattus fuscipes 125 Weight Rattus lutreolus 120 Species name (g) Tachyglossus aculeatus 3742 Capra hircus 50517 Trichosurus vulpecula 2598 Equus asinus 324037 Vombatus ursinus 26000 Equus Caballus 511000 Vulpes vulpes 6111 Felis catus 3937 Wallabia bicolor 14866 Leggadina forresti 20 Macropus fuliginosus 12669 Witjira National Park Macropus robustus 17048 Macropus rufus 41821 Weight Mus musculus 16 Species name (g) Oryctolagus cuniculus 1580 Bos taurus 447617 Ovis aries 63246 Camelus dromedarius 529150 Petrogale xanthopus 7000 Canis lupus dingo 13647 Planigale tenuirostris 6 Dasycercus cristicauda 101 Pseudomys bolami 14

239 Equus asinus 324037 Vulpes vulpes 6111 Equus Caballus 511000 Wallabia bicolor 14866 Felis catus 3937 Leggadina forresti 20 Yumbarra Conservation Park Macropus robustus 17048 Macropus rufus 41821 Weight Mus musculus 16 Species name (g) Ningaui ridei 9 Canis familiaris dingo 13647 Notomys alexis 35 Cercartetus concinnus 13 Oryctolagus cuniculus 1580 Felis catus 3937 Planigale gilesi 9 Lasiorhinus latifrons 24658 Planigale tenuirostris 6 Macropus fuliginosus 12669 Pseudantechinus Mus musculus 16 macdonnellensis 30 Ningaui yvonneae 6 Pseudomys australis 65 Notomys mitchellii 52 Pseudomys desertor 25 Oryctolagus cuniculus 1580 Pseudomys Sminthopsis hermannsburgensis 12 crassicaudata 15 Rattus villosissimus 132 Sminthopsis dolichura 14 Sminthopsis Tachyglossus aculeatus 3742 crassicaudata 15 Vulpes vulpes 6111 Sminthopsis macroura 20 Tachyglossus aculeatus 3742 Yuraygir National Park Vulpes vulpes 6111 Weight Wyperfeld National Park Species name (g) Acrobates pygmaeus 12 Weight Aepyprymnus rufescens 3240 Species name (g) Antechinus flavipes 44 Bos taurus 447617 Antechinus stuartii 26 Camelus dromedarius 529150 Bos taurus 447617 Canis familiaris 13647 Canis lupus 13647 Capra hircus 50517 Dasyurus maculatus 4583 Cercartetus concinnus 13 Equus Caballus 511000 Cercartetus lepidus 7 Felis catus 3937 Equus Caballus 511000 Hydromys chrysogaster 676 Felis catus 3937 Isoodon macrourus 1520 Macropus fuliginosus 12669 Lepus capensis 4000 Macropus giganteus 15199 Macropus giganteus 15199 Macropus rufus 41821 Macropus parryi 13267 Mus musculus 16 Macropus rufogriseus 16607 Ningaui yvonneae 6 Melomys burtoni 54 Notomys mitchelli 52 Melomys cervinipes 70 Oryctolagus cuniculus 1580 Mus musculus 16 Ovis aries 63246 Oryctolagus cuniculus 1580 Pseudomys apodemoides 20 Perameles nasuta 967 Sminthopsis murina 21 Petauroides volans 1237 Tachyglossus aculeatus 3742 Petaurus australis 561 Trichosurus vulpecula 2598

240 Petaurus breviceps 127 Phalanger gymnotis 2705 Petaurus norfolcensis 230 Phalanger orientalis 2488 Phascogale tapoatafa 170 Phalanger sericeus 2003 Pseudocheirus Phalanger vestitus 1850 peregrinus 878 Pogonomys loriae 95 Pseudomys Pogonomys macrourus 45 gracilicaudatus 73 Pseudocheirops corinnae 1123 Pseudomys Pseudocheirulus novaehollandiae 17 canescens 300 Rattus fuscipes 125 Pseudocheirulus forbesi 649 Rattus lutreolus 120 Rattus leucopus 171.5 Rattus rattus 280 Rattus steini 155 Rattus tunneyi 76 Spilocuscus maculatus 3200 Sminthopsis murina 21 Stenomys niobe 46 Sus scrofa 75829 Stenomys verecundus 108 Tachyglossus aculeatus 3742 Tachyglossus aculeatus 4500 Trichosurus vulpecula 2598 Uromys caudimaculatus 646 Vulpes vulpes 6111 Xenuromys barbatus 1000 Wallabia bicolor 14866 Zaglossus bruijn 9633

Rainforest of New Guinea

Species name Weight (g) Anisomys imitator 510 Antechinus melanurus 48 Chiruromys vates 45 Crossomys moncktoni 165 Dactylopsila trivirgata 404 Dasyurus albopunctatus 627 Dendrolagus dorianus 9581 Dendrolagus spadix 9100 Distoechurus pennatus 44 Echymipera kalubu 959 Leptomys elegans 85 Lorentzimys nouhuysi 14 Mallomys aroaensis 1664 Melomys levipes 100 Melomys lutillus 100 Melomys rubex 60 Microhydromys richardsoni 10 Microperoryctes longicaudata 538 Murexia longicaudata 140 Myoictis melas 200 Parahydromys asper 540 Peroryctes raffrayana 839 Petaurus breviceps 128 Phalanger carmelitae 2048

241

APPENDIX B

DENTAL MEASUREMENTS OF UNPUBLISHED

RIVERSLEIGH SPECIMEN USED IN THE NUMERICAL

AGE METHOD (CHAPTER 6)

ATTACHED CD ONLY See file —Appendix B- Riversleigh Dental Measurements“.

242

APPENDIX C

RAW MSR DATA FOR THE QUERCY AND LIMAGNE

AREA AND RIVERSLEIGH WORLD HERITAGE AREA.

ATTACHED CD ONLY See file —Appendix C- Quercy MSR Data“ and —Appendix C œ Riversleigh MSR Data“.

243

APPENDIX D

RAW NUMERICAL AGES DATA

ATTACHED CD ONLY See file —Appendix D - Numerical Age data“.

244

APPENDIX E

PUBLICATIONS

ARCHER, M., ARENA, D.A., BASSAROVA, M., BECK, R.M.D., BLACK, K., BOLES, W.E., BREWER, P., COOKE, B.N., CROSBY, K., GILLESPIE, A., GODTHELP, H., HAND, S.J., KEAR, B.P., LOUYS, J, MORRELL, A., MUIRHEAD, J., ROBERTS, K.K., SCANLON, J.D., TRAVOUILLON, K.J. & WROE, S., 2006. Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1, 1-17.

TRAVOUILLON K.J., ARCHER, M., HAND, S.J. & GODTHELP, H., 2006. Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1, 323-349.

245 Current status of species-level representation in faunas. from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland

MICHAEL ARCHER1, DERRICK A. ARENA*1, MINA BASSAROVA1, ROBIN M.D. BECK1, KAREN BLACK1, WALTER E. BOLES2, PHILLIPA BREWER1, BERNARD N. COOKE3, KIRSTEN CROSBY1, ANNA GILLESPIE1, HENK GODTHELP1, SUZANNE J. HAND1, BENJAMIN P. KEAR4, JULIEN LOUYS1, ADAM MORRELL5, JEANETTE MUIRHEAD1, KAREN K. ROBERTS1, JOHN D. SCANLON6, KENNY J. TRAVOUILLON*1 and STEPHEN WROE1

ARCHER, M., ARENA, D.A., BASSAROVA, M., BECK, R.M.D., BLACK, K., BOLES, W.E., BREWER, P., COOKE, B.N., CROSBY, K., GILLESPIE, A., GODTHELP, H., HAND, S.J., KEAR, B.P., LOUYS, J, MORRELL, A., MUIRHEAD, J., ROBERTS, K.K., SCANLON, J.D., TRAVOUILLON, K.J. & WROE, S., 2006. Current status of species-level representation in faunas from selected fossil localities in the Riversleigh World Heritage Area, northwestern Queensland. Alcheringa Special Issue 1, 1-17. ISBN 0 9757894 5 7.

Current lists of species-level representation in faunas from 80 Cenozoic fossil localities at the Riversleigh World Heritage Area have been compiled by review of recorded occurrences of taxa obtained from both published and unpublished sources. More than 290 species-level taxa are represented, comprising mammals, amphibians, reptiles, birds, fishes, molluscs and crustaceans. The data are presented for the purpose of ongoing palaeoecological and biochronological studies.

1School of Biological, Earth and Environmental Sciences, University of New South Wales, N.S.W. 2052; 2Australian Museum, 6 College Street, Sydney, N.S.W. 2010; 3School of Natural Resource Sciences, Queensland University of Technology, GPO Box 2434, Brisbane, QLD 4001; 4School of Earth and Environmental Sciences, University of Adelaide, Adelaide, S.A. 5005 and Earth Sciences Section, South Australian Museum, North Terrace, Adelaide, S.A. 5000; 5Richmond Marine Fossil Museum, Richmond, QLD 482; 6Riversleigh Fossil Centre, Outback at Isa, P.O. Box 1094, Mount Isa, QLD 4825; submitted 19.8.2005, accepted 21.08.2006. *corresponding authors.

Key words: Species list, Riversleigh, Local Fauna.

FOSSIL faunas have been collected from more chert deposits that form the northeastern edge than 200 localities at Riversleigh and have been of the Barkly Tableland in northwestern the subject of research efforts during the last 40 Queensland. The depositional history of these years (Archer et al. 1989, 1994, 1997). Ongoing deposits appears to involve a complex sequence fieldwork and exploration continues to identify of fluvial and karst processes (Megirian 1992, an increasing number of additional fossil Archer et al. 1989, 1994, 1997; Arena 2005). occurrences that have not yet been sampled, Faunas from these deposits are considered to and the vertical and lateral extents of potentially range in age from the late Oligocene to early late fossiliferous deposits in the area are not yet fully Miocene (Archer et al. 1989, 1994, 1997). Fossil understood. deposits at Riversleigh are also known to occur The Riversleigh Tertiary carbonate deposits as infills within Cambrian limestone deposits occur adjacent to, within and upon Proterozoic (Rackham’s Roost Site, Pliocene), Pleistocene siliclastics and Cambrian marine limestone and alluvial terraces (Terrace Site) along the Gregory ISBN 0 9757894 5 7/2006/17 $3.00 ©AAP River, and in active caves in the area (Carrington’s 2 ARCHER ET AL. ALCHERINGA

Cave and Message Stick Cave, both Holocene). (Ornithorhynchidae) and a single species Generic and species-level systematics Obdurodon dicksoni. Altogether, 159 species of follows that used by authors contributing to marsupials representing five orders are present, Archer et al. (1999) or that used by the author of comprising two families of Dasyuromorphia the species name. Fossil material used for the (Thylacinidae, 9; Dasyuridae, 8 taxa; 3 taxa not compilation of these faunal lists is on loan from allocated to a family); one family of Noto- the Queensland Museum. ryctemorphia (Notoryctidae, 1 taxon); two families of Peramelemorphia (Peramelidae, 2 taxa; Methods Yaralidae, 1; 9 taxa not allocated to a family); 18 The relationships between fossil taxa and source families of Diprotodontia (Phascolarctidae, 6 taxa; localities at Riversleigh were obtained from all Thylacoleonidae, 5; Vombatidae, 4; Ilariidae, 1; available literature. Additional unpublished data Wynyardiidae, 4; Palorchestidae, 6; were obtained for further taxa that are currently Diprotodontidae, 10; Maradidae, 1; Burramyidae, under study and have not yet been formally 2; Pseudocheiridae, 20; Pilkipildridae, 2; described. Phalangeridae, 10; Ektopodontidae, 4; Whereas a great number of specimens Miralinidae, 2; Balbaridae, 20; Hypsiprymno- collected from Riversleigh has been identified to dontidae, 5; Potoroidae, 2; Macropodidae, 19; the level of family and genus, only taxa identified one family of Yalkaparidontia (Yalkaparidontidae, to species-level have been included in the faunal 2 taxa); 1 taxon not allocated to a family); and two lists presented here. Taxa identified as being new genera of marsupials not allocated to order, distinct at species-level but not named as species suborder or family. Placentals comprise 46 species are referred to as ‘sp.’. Taxa that are currently in six families of microchiropteran bats under study and have not yet been formally (Megadermatidae, 4 taxa; Hipposideridae, 15; described in published literature are referred to Emballonuridae, 2; Molossidae, 2; Vesper- as ‘sp 1’, ‘sp 2’, etc. References to unpublished tilionidae, 5; Mystacinidae, 2) and one family of taxa in these faunal lists are not intended to rodents (Muridae, 16 taxa). The relationship constitute or pre-empt formal descriptions of between the family Yingabalanaridae (1 taxon) these taxa. and other mammal groups is not clear. Amphibia Each fossil-bearing locality at Riversleigh are represented by 19 species in two families of is identified by a unique name and referred to frogs (Hylidae, 9 taxa; Leptodactylidae, 10). There as a ‘Site’. Each site yields a distinct local are 34 species of Reptilia comprising two families fauna. The fauna from Microsite has also been of turtles (Chelidae, 5 taxa; Meiolaniidae, 3), five referred to as the Nooraleeba Local Fauna families of lizards (Pygopodidae, 1 taxon; (Archer & Hand 1984), the fauna from Gag Site Agamidae, 3; Scincidae, 4; Varanidae, 1; as the Dwornamor Local Fauna (Archer & Typhlopidae, 1), three families of snakes (Boidae, Hand 1984), and the fauna collected from Sites 2 taxa; Elapidae, 1; Madtsoiidae, 4) and two sub- A, B (= BMR Locality M103 and Sample Q11 families of crocodilians (Mekosuchinae, 8 taxa; locality of McMichael [1968]), C, D, E, G, H Crocodylinae, 1). Birds are represented by 24 and locality Q12 of McMichael (1968) has species from fourteen families comprising been referred to as the Riversleigh Local Fauna Casuariidae (1 taxon), Dromornithidae (2 taxa), (Tedford 1967, Stirton et al. 1968, Flannery et Ciconiidae (1 taxon), Accipitridae (1 taxon), al. 1982, Archer et al. 1994). Locality names Rallidae (1 taxon), Cacatuidae (1 taxon), and their abbreviations used in this work are Psittacidae (1 taxon), Apodidae (1 taxon), listed in Table 1. Halcyonidae (1 taxon), an undetermined family of Passeriformes (4 taxa), Menuridae (1 taxon), Results Oriolidae (1 taxon), Orthonychidae (1 taxon) and Monotremes are represented by only one family Meliphagidae (7 taxa). A single family of lung ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 3

fish (Neoceratodontidae) is represented by 2 taxa. The species-occurrence lists for the Invertebrates include two families of Gastropoda Riversleigh localities are shown in Table 2. Each (Planorbidae, 1 taxon; Camaenidae, 1) and two occurrence is represented in the table as a black families of Ostracoda (Limnocytheridae, 1 taxon; box. Absence of a species is represented by a Cyprididae, 1). blank box.

Table 1. List of source sites for Riversleigh faunas and their abbreviations used in this work.

4 ARCHER ET AL. ALCHERINGA

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (95, 107) (97, 103) (8, 9) (99) (149) (92) (92) (97, 103) (8, 9, 61) (101, 151) (142) (10, 109) (148) (60) (104) incertae sedis (143) (99) incertae sedis (100, 102) T. macknessi (8, 9, 61) (147) sp. (111) Mutpuracinus archibaldi new sp. (8, 9) sp. 1 (8, 9) sp. 2 (8, 9) cf. sp. cf. sp. 1 (40, 111) new sp. (8, 9) new sp. 1 (16) new sp. 1 (61)    TAXON Dasyuromorphia MAMMALIA - MONOTREMATA Ornithorhynchidae Obdurodon dicksoni Dasyuridae Barinya wangala Dasyuridae genus indet. sp. 1 (2,Dasyuridae 8) genus indet. sp. 2 (2,Dasyuridae 8) genus indet. sp. 3 (2,Dasyuridae 8) genus indet. sp. 4 (2,Ganbulanyi 8) djadjinguli (145) Planigale Sminthopsis Joculusium muizoni Mayigriphus orbus Dasyuromorphia new genus new sp. (146) Notoryctidae Notoryctid new genus new sp. (65) Peramelidae ?Perameles Peramelemorphia new genus 1 sp. 1Peramelemorphia (98) new genus 2 sp. 1Peramelemorphia (98) new genus 2 sp. 2Peramelemorphia (98) new genus 2 sp. 3Peramelemorphia (98) new genus 3 sp. 1Peramelemorphia (98) new genus 4 sp. 1Peramelemorphia (98) new genus 4 sp. 2Peramelemorphia (98) new genus 5 sp. 1Peramelemorphia (98) new genus 5 sp. 2Yaralidae (98) Yarala burchfieldi Phascolarctidae Litokoala kutjamarpensis Litokoala Litokoala garyjohnstoni Priscileo roskellyae Thylacoleonidae new genus new sp. (61) Vombatidae Vombatidae genus 1 sp. 1 (40) Vombatidae genus 2 sp. 1 (40) Rhizophascolonus crowcrofti Warendja Kuterintja ngama (110) MAMMALIA - MARSUPIALIA Thylacinidae Wabulacinus ridei Maximucinus muirheadae Badjicinus turnbulli Thylacinus macknessi Thylacinus ?Perameles Peramelemorphia Nimiokoala greystanesi (21) Phascolarctidae new genus new sp. (16) Phascolarctos Thylacoleonidae Wakaleo oldfieldi Wakaleo vanderleueri Wakaleo Ilariidae Ngamalacinus timmulvaneyi Nimbacinus dicksoni Muribacinus gadiyuli Thylacinidae Archer et al., Current status of species-level representation... Table 2 - Page 1 of 6

ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 5

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (113) (18, 20, 105) (8, 114, 113) (18, 106) (113) (15) (79) (22) (22) (113) (17, 18, 125) (11) (48, 55) (117) N. albivenator (17) (18) (38) (15) (9, 15) (12, 79,126) (48, 55) (18, 52) (51) (116) (15) (19) (108) (39) sp. 1 (117) sp. 2 (117) sp. cf. sp. (18) incertae sedis sp. (22, 127) sp. (6) sp. (48) new sp. (8, 9) sp. A (108) sp. C (108) sp. (111)    sp. 1 (14) sp. 1 (117) sp. 2 (117) sp. 3 (117) sp. 4 (117) sp. 1 (117) sp. 2 (117) sp. 3 (117) cf. sp. 3 (117) sp. 4 (117) TAXON Pilkipildridae Djilgaringa gillespieae ?Djilgaringa Phalangeridae Wyulda asherjoeli Namilamadeta superior Djaludjangi yadjana “Strigocuscus” reidi Namilamadeta crassirostrum Palorchestidae Propalorchestes ponticulus Paljara tirarensae Paljara Marlu kutjamarpensis Marlu Pildra Pildra Pildra Pseudocheiridae new genus 2 sp. 1Pseudocheiridae (117) new genus 2 sp. 2Petauroidea (117) “Trichosurus” dicksoni Trichosurus Burramyidae Burramys brutyi Cercartetus Pseudocheiridae Paljara nancyhaywardae Paljara maxbourkei Gawinga aranaea Pseudochirops Pseudochirops Petropseudes dahli Marlu Marlu Marlu Marlu Pildra Phalangeridae new genus 1 sp. 1Phalangeridae (48) new genus 1 sp. 2Phalangeridae (48) new genus 1 sp. 3Phalangeridae (48) new genus 2 sp. 1Phalangeridae (48) new genus 2 sp. 2 (48) Wynyardiidae Namilamadeta albivenator Namilamadeta Ngapakaldia bonythoni Neohelos Phalangeridae new genus 3 sp. 1 (48) Ngapakaldia Palorchestes anulus Neohelos tirarensis Neohelos stirtoni Neohelos Neohelos Diprotodon optatum Maradidae Marada arcanum Palorchestes azael Diprotodontidae Nimbadon lavarackorum Silvabestius michaelbirti Silvabestius johnnilandi Silvabestius Propalorchestes novaculacephalus Archer et al., Current status of species-level representation... Table 2 - Page 2 of 6

6 ARCHER ET AL. ALCHERINGA

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (56) (8) (42, 88) (54) (8, 114) (141) (86) snewini (41) (56) (56) (1) (50) (41) (49) (7, 47) B. delicata (63) (56) (7) (56) (8, 126) (41) (42, 88) (43) (45) (43) (53) (44) E. serratus new sp. (150) sp. (111) (42) G. Bilamina (8) (5, 141, 144) sp. 4 (41) sp. cf. sp. cf. P. incertae sedis sp. cf. sp. 2 (41) sp. 3 (41) sp. 4 (41) sp. 5 (41) sp. 6 (41) sp. 7 (41) sp. 8 (41) sp. 2 (41, 87) sp. (8) sp. 2 (41) sp. cf. sp. 2 (41) sp. 1 (8) sp. 2 (8) sp. (8) sp. 3 (41) sp. 4 (41) sp. 2 (41) sp. 3 (41)    sp. (64) TAXON Macropus Macropus agilis Rhizosthenurus flanneryi Sthenurus Yalkaparidontidae Yalkaparidon coheni Yalkaparidon jonesi Marsupialia New genus sp. 1 (13) Protemnodon Macropus Wanburoo Wakiewakie lawsoni Ganguroo Ganguroo new sp. (111) Wanburoo hilarus Gumardee pascuali Chunia Ektopodontidae new genus new sp. (64) Miralinidae Durudawiri inusitatus Ganguroo bilamina Ektopodontidae Ektopodon serratus Ektopodon Durudawiri anfractus Ganguroo New genus sp. 2 (13) Balbaridae Ganawamaya ornata Ganawamaya acris Ganawamaya aediculis Ganawamaya Balbaroo gregoriensis Nowidgee Bulungamaya delicata Bulungamaya Balbaroo fangaroo Balbaroo Balbaroo Galanarla tessellata Wururoo dayamayi Wururoo Nambaroo Potoroidae Bettongia moyesii Bettongia Macropodidae Wabularoo naughtoni Nowidgee matrix Wururoo Nambaroo couperi Nambaroo Nambaroo Nambaroo Nambaroo Nambaroo Nambaroo Hypsiprymnodontidae Hypsiprymnodon bartholomaii Hypsiprymnodon Hypsiprymnodon Ekaltadeta ima Ekaltadeta jamiemulvaneyi Archer et al., Current status of species-level representation... Table 2 - Page 3 of 6

ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 7

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (76) (71) (73) (68) (80) (77) (4) (71) (72) (76) (8) (8) B. nooraleebus (70) (66) (72) (74) (81) (69) H. bernardsigei (67) (70) new sp. (76) sp. cf. (75) (78, 82) sp. (8) sp. (8) (82) new sp. (76) sp. (8) sp. cf. new sp. 1 (8, 9) new sp. 2 (8, 9) new sp. 3 (8, 9) new sp. 4 (8, 9) new sp. 5 (8, 9) new sp. 6 (8, 9) new sp. 7 (8, 9) incertae sedis sp. 1 (3, 69, 80) sp. 2 (3, 69, 80) new sp. 1 (9) new sp. 2 (9) sp. (96)    Potwarmus Chalinolobus Scotorepens TAXON Archerops annectens Brachipposideros watsoni Brachipposideros nooraleebus Riversleigha williamsi Xenorhinos halli Brevipalatus mcculloughi Miophyllorhina riversleighensis Rhinonycteris tedfordi Taphozous Molossidae Hydromops riversleighensis MAMMALIA - PLACENTALIA Megadermatidae Macroderma gigas Macroderma godthelpi Megaderma richardsi Hipposideridae Hipposideros winsburyorum Hipposideros bernardsigei Hipposideros Brachipposideros Brachipposideros ?Rhinonycteris aurantius Rhinonycteris Emballonuridae Taphozous Petramops creaseri Yingabalanara richardsoni Macroderma malugara Mystacinidae Icarops paradox Icarops aenae Leuconoe Vespertilionidae genus indet. sp. 1 (8) Vespertilionidae genus indet. sp. 2 (8) Muridae cf Vespertilionidae cf cf Leggadina Leggadina Pseudomys Pseudomys Pseudomys Pseudomys Pseudomys Pseudomys Pseudomys Zyzomys rackhami (62) Zyzomys argurus (64) Hydromys chrysogaster Muridae genus indet. sp. 1 (8) Muridae genus indet. sp. 2 (8) Muridae new genus new sp. (8) MAMMALIA Yingabalanaridae Archer et al., Current status of species-level representation... Table 2 - Page 4 of 6

8 ARCHER ET AL. ALCHERINGA

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (130, 133) (6, 9) (57) (130) (84) (120) (59) (120, 124) (84) (123) (129, 133) (130) lesueurii (46) (122) (128, 135) (119) (85) (119) sp. (6) platyceps L. tasmaniensis (121) sp. (9) striolata frerei (121) sp. (58) sp. (58) E. M. spilota (123) (131) sp. cf. P. sp. (46) (132) T. pusilla . sp. cf. sp. (59) sp. (134) sp. 1 (133) sp. 2 (133) sp. cf. M. sp. (58)    sp. (8) sp. cf sp. cf. E. sp. cf. sp. cf. sp. 1 (8, 133) sp. 2 (8, 133) sp. 3 (8, 133) sp. 4 (8, 133) sp. 5 (8, 133) sp. 6 (8, 133) sp. 7 (8, 133) sp. 8 (8, 133) sp. (133) TAXON Nanowana schrenki Wonambi barriei Elapidae Incongruelaps iteratus Madtsoiidae Nanowana godthelpi Egernia Physignathus Scincidae Egernia Tiliqua pusilla Tiliqua Physignathus LEPIDOSAUROMORPHA - SQUAMATA Pygopodidae Pygopus hortulanus Agamidae Sulcatidens quadratus Varanidae Varanus Typhlopidae ?Rhamphotyphlops Boidae Morelia riversleighensis Morelia Meiolaniidae Warkalania carinaminor ?Warkalania Meiolania Pseudemydura ?Emydura ?Pseudemydura Emydura Kyarranus Kyarranus Lechriodus intergerivus Limnodynastes antecessor Crinia TEMNOSPONDYLI - ANURA Hylidae Litoria magna Litoria Litoria Litoria Litoria Litoria Litoria Litoria Litoria Leptodactylidae Crinia presignifera Crinia remota Limnodynastes ornatus Australobatrachus TESTUDOMORPHA - TESTUDINES Chelidae Elseya lavarackorum Madtsoiidae new genus sp. 1 (118) Limnodynastes Archer et al., Current status of species-level representation... Table 2 - Page 5 of 6

ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 9

WH

WW

Wang

VIP

VD

UBO

U

TT

Ter

TB

SD

SB

RSO

Ring

RV

RRR

RR

Roo

QQ

QL

PIR

Pha

Out

NP

NG

Mim

MPP

MM

Micro

MSC

Mesa

Mel

Main

LSO

LL

LD94

LM

KCB

KJ

JH

JJS

JC

JA

JJJ

JJ

Ina

HSS

HS

HH

Hel

GC

GOH

GG

GLG

Gag

G

FF

FT

En

Dun

Dome

DT

D

CR

COA

CC

CS

CK

BO

BR

Boles

BSE

Boid

Bob

Bite

AR

AA

AL

AL90

300BR

SITE     (93) (126) (83, 137) (32, 37) (31) (89, 90, 91) (136) (30) (24) (115) (34) (140) (26) (139) (27) (138) N. gregoryi (8, 94) (8, 94) (137) (33) (35) (23, 28, 112) sp. cf. (83, 137) sp. (93) (137)    sp. (25) sp. (138) TAXON ARCHOSAUROMORPHA - CROCODYLIA Mekosuchinae Baru wickeni Baru huberi Baru Mekosuchus whitehunterensis Cypridopsis Trilophosuchus rackhami Mekosuchus sanderi Quinkana meboldi Crocodylinae Crocodylus johnstoni Pallimnarchus gracilis ARCHOSAUROMORPHA - AVES Casuariidae Emuarius gidju Dromornithidae Barawertornis tedfordi Dromornithidae new genus A sp.1 (32) Ciconiidae Ciconia louisebolesae Accipitridae Pengana robertbolesi Rallidae Gallinula disneyi Cacatuidae Cacatua MOLLUSCA - GASTROPODA Planorbidae Physastra rodingae Psittacidae Melopsittacus undulatus Apodidae Collocalia buday Meliphagidae Meliphagidae genus indet. sp.1 (36) Meliphagidae genus indet. sp.2 (36) Meliphagidae genus indet. sp.3 (36) Meliphagidae genus indet. sp.4 (36) Meliphagidae genus indet. sp.5 (36) Meliphagidae genus indet. sp.6 (36) Meliphagidae genus indet. sp.7 (36) DIPNOMORPHA - DIPNOI Neoceratodontidae Neoceratodus Mioceratodus anemosyrus Camaenidae Meracomelon lloydi CRUSTACEA - OSTRACODA Limnocytheridae Progomphocythere mawsontalenti Cyprididae Halcyonidae Halcyonidae genus indet. sp. (29) PASSERIFORMES family indet Passeriformes genus indet. sp. 1 (8,Passeriformes 27) genus indet. sp. 2 (8,Passeriformes 27) genus indet. sp. 3 (8,Corvitalusoides 27) grandiculus Menuridae Menura tyawanoides Orthonychidae Orthonyx kaldowinyeri Oriolidae Longmornis robustirostrata Archer et al., Current status of species-level representation... Table 2 - Page 6 of 6 10 ARCHER ET AL. ALCHERINGA

Discussion Australia Pty Ltd; the Australian Geographic The data presented here are a product of the Society; the Queensland Museum; the contributions of many workers over several Australian Museum; the Royal Zoological decades to the ongoing cumulative research Society of New South Wales; the Linnean Society effort. Because these species lists are based of New South Wales; Century Zinc Pty Ltd; the only on the occurrence of taxa identified at Riversleigh Society Inc.; and private supporters species-level and do not include all identified including Elaine Clark, Margaret Beavis, Martin specimens, caution should exercised in the Dickson, Sue & Jim Lavarack and Sue and Don interpretation of the data in terms of the Scott-Orr. Many thanks to Tim Holt, Steve presence or absence of taxa. Taxa that have Salisbury and Dirk Megirian for confirming the been assigned to higher-level groups may not presence of several taxa. Vital assistance in the be listed here because specimens have not yet field has come from many hundreds of volunteers been, or are not able to be, identified beyond as well as staff and postgraduate students of the level of genus. the University of New South Wales. Some differences in relative diversity References between faunas may be due to differences in ARCHER, M., 1979. Wabularoo naughtoni gen. et sp. nov., sampling from localities, which can vary in an enigmatic kangaroo (Marsupialia) from the middle terms of the relative quantities of material that Tertiary Carl Creek Limestone of northwestern have been collected and/or processed and Queensland. Memoirs of the Queensland Museum 19, 299-307. identified so far. There is additional material ARCHER, M., 1982. Review of the dasyurid (Marsupialia) from other localities yet to be processed. As fossil record, integration of data bearing on such the species lists presented here do not phylogenetic interpretation and suprageneric constitute complete faunal lists, but provide classification. In Carnivorous marsupials, M. ARCHER ed., Royal Zoological Society of New South Wales, current samples of known species-level Sydney, 2, 397-443. representation within faunas from selected ARCHER, M., ARENA, R., BASSAROVA, M., BLACK, K., BRAMMALL, localities. Meanwhile, this type of data can be J., COOKE, B., CREASER, P., CROSBY, K., GILLESPIE, A., used for palaeoecological studies. These data GODTHELP, G., GOTT, M., HAND, S.J., KEAR, B., KRIKMANN, A., MACKNESS, B., MUIRHEAD, J., MUSSER, A., MYERS, T., are used by Travouillon et al. (in press) to PLEDGE, N., WANG, Y. & WROE, S., 1999. The investigate the temporal relationship of each evolutionary history and diversity of Australian of Riversleigh’s Local Faunas and test the mammals. Australian Mammalogy 21, 1-45. Riversleigh “System” biostratigraphic concept ARCHER, M., EVERY, R., GODTHELP. H., HAND, S.J. & SCALLY, K., 1990. Yingabalanaridae, a new family of introduced by Archer et al. (1989). enigmatic mammals from Tertiary deposits of Riversleigh, northwestern Queensland. Memoirs of Acknowledgments the Queensland Museum 28, 193-202. ARCHER, M. & FLANNERY, T., 1985. Revision of the extinct Vital support for research at Riversleigh has come gigantic rat kangaroos (Potoroidae: Marsupialia), from the Australian Research Grant Scheme with description of a new Miocene genus and species (grants to M. Archer); the National Estate Grants and a new Pleistocene species of Propleopus. Journal of Paleontology 59, 1331-1349. Scheme (Queensland) (grants to M. Archer and ARCHER, M., GODTHELP, H., HAND, S.J. & MEGIRIAN, D., 1989. A. Bartholomai); the University of New South Fossil mammals of Riversleigh, northwestern Wales; the Commonwealth Department of Queensland: preliminary overview of biostratigraphy, Environment, Sports and Territories; the correlation and environmental change. Australian Zoologist 25, 29-65. Queensland National Parks and Wildlife Service; ARCHER, M. & HAND, S.J., 1984. Background to the search for the Commonwealth World Heritage Unit; ICI Australia’s oldest mammals. In Vertebrate zoogeography

Table 2. Species lists for the faunas from 80 Riversleigh sites. Black boxes represent presence of species at the corresponding locality; blank boxes represent absence of species. The numbers attached to the taxon names correspond to references listed in the appendix. [see preceeding six pages, 4-9]. ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 11

& evolution in Australasia, M. Archer and G. Clayton Queensland. Memoirs of the Queensland Museum eds., Hesperian Press, Perth, 517-565. 41, 187-192. ARCHER, M., HAND, S.J. & GODTHELP, H., 1988. A new order BLACK, K., 2006. Description of new material for of Tertiary zalambdodont marsupials. Science 239, Propalorchestes novaculacephalus (Marsupialia: 1528-1531. Palorchestidae) from the mid Miocene of Riversleigh, ARCHER, M., HAND, S.J. & GODTHELP, H., 1994. Riversleigh. northwestern Queensland. Alcheringa 30, 351-361. Reed Books, Sydney. 256 pp. BLACK, K., in press. Maradidae: a new family of ARCHER, M., HAND, S.J., GODTHELP, H. & CREASER, P., 1997. vombatomorphian marsupial from the late Oligocene Correlation of the Cainozoic sediments of the of Riversleigh, northwestern Queensland. Alcheringa. Riversleigh World Heritage fossil property, BLACK, K. & ARCHER, M., 1997a. Nimiokoala gen. nov. Queensland, Australia. Mémoires et Travaux de (Marsupialia, Phascolarctidae) from Riversleigh, l’Institut de Montpellier de l’Ecole Pratique des northwestern Queensland. Memoirs of the Queensland Hautes Etudes 21, 131-152. Museum 41, 209-228. ARCHER, M., JENKINS, F., HAND, S.J., MURRAY, P. & GODTHELP, BLACK, K. & ARCHER, M., 1997b. Silvabestius gen. nov., a H., 1992. Description of the skull and non-vestigial primitive zygomaturine (Marsupialia, dentition of a Miocene platypus (Obdurodon Diprotodontidae) from Riversleigh, northwestern dicksoni n. sp.) from Riversleigh, Australia, and the Queensland. Memoirs of the Queensland Museum problem of monotreme origins. In Platypus and 41, 193-208. echidnas, M. AUGEE ed., The Royal Zoological BOLES, W.E., 1992. Revision of Dromaius gidju Patterson Society of NSW, Sydney, 15-27. and Rich 1987 from Riversleigh, northwestern ARCHER, M., TEDFORD, R. & RICH, T., 1987. The Queensland, Australia, with a reassessment of its Pilkipildridae, a new family and four new species of generic position. Natural History Museum of L. A. ?petauroid possums (Marsupialia: Phalangerida) from County, Science Series No.36 195-208. the Australian Miocene. In Possums and opossums: BOLES, W.E., 1993a. A logrunner, Orthonyx studies in evolution, M. ARCHER ed., Surrey Beatty & (Passeriformes: Orthonychidae) from the Miocene Sons and The Royal Zoological Society of New South of Riversleigh, north-western Queensland. Emu 93, Wales, Sydney, 2, 607-627. 44-49. ARENA, D. A., 1997. The palaeontology and geology of BOLES, W.E., 1993b. A new cockatoo (Psittaciformes: Dunsinane Site, Riversleigh. Memoirs of the Cacatuidae) from the Tertiary of Riversleigh, Queensland Museum 41, 171-179. northwestern Queensland, and an evaluation of rostral ARENA, D.A., 2005. The geological history and characters in the systematics of parrots. Ibis 135, 8- development of the Riversleigh terrain during the 18. middle Tertiary. Conference on Australasian BOLES, W.E., 1993c. Pengana robertbolesi, a peculiar Vertebrate Evolution, Palaeontology and Systematics bird of prey from the Tertiary of Riversleigh, (CAVEPS) Abstracts, 11. northwestern Queensland, Australia. Alcheringa 17, BASSAROVA, M., 1999. Description and phylogenetic 19-25. analysis of new species of ringtail possums BOLES, W.E., 1995. A preliminary analysis of the (Pseudocheiridae, Marsupialia) of the genus Paljara Passeriformes from Riversleigh, northwestern from Oligo-Miocene deposits of the Riversleigh World Queensland, Australia, with the description of a new Heritage Property, northwestern Queensland. species of lyrebird. Courier Forchungsinstitut Unpublished Honours thesis, University of New South Senckenberg 181, 163-170. Wales, Sydney. BOLES, W.E., 1997a. Hindlimb proportions and BASSAROVA, M., ARCHER, M. & HAND, S.J., 2001. New Oligo- locomotion of Emuarius gidju (Patterson & Rich, Miocene pseudocheirids (Marsupialia) of the genus 1987) (Aves, Casuariidae). Memoirs of the Paljara from deposits of the Riversleigh World Queensland Museum 41, 235-240. Heritage property, northwestern Queensland. BOLES, W.E., 1997b. A kingfisher (Halcyonidae) from Memoirs of the Association of Australasian the Miocene of Riversleigh, northwestern Palaeontologists 25, 61-75. Queensland, with comments on the evolution of BLACK, K., 1992. Systematics and palaeobiology of fossil kingfishers in Australo-Papua. Memoirs of the phascolarctids from Riversleigh. Unpublished Queensland Museum 41, 229-234. Honours yhesis, University of New South Wales, BOLES, W.E., 1998. A Budgerigar, Melopsittacus undulatus Sydney. from the Pliocene of Riversleigh, north-western BLACK, K., 1997a. A new species of Palorchestidae Queensland. Emu 98, 32-35. (Marsupialia) from the late middle to early late BOLES, W.E., 1999. A new songbird (Aves: Passeriformes: Miocene Encore Local Fauna, Riversleigh, Oriolidae) from the Miocene of Riversleigh, northwestern Queensland. Memoirs of the Queensland northwestern Queensland, Australia. Alcheringa 23, Museum 41, 181-185. 51-56. BLACK, K., 1997b. Diversity and biostratigraphy of the BOLES, W.E., 2000. Investigations on Australian Tertiary Diprotodontoidea of Riversleigh, northwestern avifauna, with an emphasis on the fossil birds of 12 ARCHER ET AL. ALCHERINGA

Riversleigh, northwestern Queensland. Unpublished CROSBY, K., 2002a. Studies in the diversity and evolution PhD thesis, University of New South Wales, Sydney. of phalangeroid possums (Marsupialia; BOLES, W.E., 2001. A swiftlet (Apodidae: Collocaliini) Phalangerida; Phalangeroidea). Unpublished PhD from the Oligo-Miocene of Riversleigh, northwestern Thesis, University of New South Wales, Sydney. Queensland. Memoirs of the Association of CROSBY, K., 2002b. A second species of the possum Australasian Palaeontologists 25, 45-52. Durudawiri (Marsupialia: Miralinidae) from the early BOLES, W.E., 2005a. A review of the Australian fossil Miocene of Riversleigh, northwestern Queensland. storks of the genus Ciconia (Aves: Ciconiidae), with Alcheringa 26, 333-340. the description of a new species. Records of the CROSBY, K. & ARCHER, M., 2000. Durudawirines, a new Australian Museum 57, 165-178. group of phalangeroid marsupials from the Miocene BOLES, W.E., 2005b. A new flightless gallinule (Aves: of Riversleigh, northwestern Queensland. Journal Rallidae: Gallinula) from the Oligo-Miocene of of Paleontology 74, 327-335. Riversleigh, northwestern Queensland, Australia. CROSBY, K., NAGY, M. & ARCHER M., 2001. Wyulda Records of the Australian Museum 57, 179-190. asherjoeli, a new phalangerid (Diproto- BOLES, W.E., 2005c. Fossil honeyeaters (Meliphagidae) dontia:Marsupialia) from the early Miocene of from the late Tertiary of Riversleigh, north-western Riversleigh, northwestern Queensland. Memoirs of Queensland. Emu 105, 21-26. the Association of Australasian Palaeontologists 25, BOLES, W.E., 2005d. A new songbird (Aves: Passeriformes) 77-82. from the mid-Tertiary of Riversleigh, northwestern DAVIS, A.C. & ARCHER, M., 1997. Palorchestes azael Queensland. Conference on Australasian Vertebrate (Mammalia, Palorchestidae) from the late Evolution, Palaeontology and Systematics (CAVEPS) Pleistocene Terrace Site Local Fauna, Riversleigh, Abstracts, 19. northwestern Queensland. Memoirs of the Queensland BRAMMALL, J.R., 1999. A new petauroid possum from the Museum 41, 315-320. Oligo-Miocene of Riversleigh, northwestern FLANNERY, T.F. & ARCHER, M, 1987a. Bettongia moyesi, a Queensland. Alcheringa 23, 31-50. new and plesiomorphic kangaroo (Marsupialia: BRAMMALL, J. & ARCHER, M., 1997. A new Oligocene- Potoroidae) from Miocene sediments of Miocene species of Burramys (Marsupialia, northwestern Queensland. In Possums and Burramyidae) from Riversleigh northwestern opossums: studies in evolution, M. ARCHER ed., Surrey Queensland. Memoirs of the Queensland Museum Beatty & Sons and The Royal Zoological Society of 41, 247-268. New South Wales, Sydney, 2, 759-67. COOKE, B.N., 1997a. Biostratigraphic implications of fossil FLANNERY, T.F. & ARCHER, M., 1987b. Hypsiprymnodon kangaroos at Riversleigh, northwestern Queensland. bartholomaii (Potoroidae: Marsupialia), a new Memoirs of the Queensland Museum 41, 295-302. species from the Miocene Dwornamor Local Fauna COOKE, B.N., 1997b. New Miocene bulungamayine and a reassessment of the phylogenetic position of kangaroos (Marsupialia: Potoroidae) from H. moschatus. In Possums and opossums: studies in Riversleigh, northwestern Queensland. Memoirs of evolution, M. ARCHER ed., Surrey Beatty & Sons and the Queensland Museum 41, 281-294. The Royal Zoological Society of New South Wales, COOKE, B.N., 1997c. Two new balbarine kangaroos and Sydney, 2, 749-58. lower molar evolution within the subfamily. Memoirs FLANNERY, T.F. & ARCHER, M., 1987c. Strigocuscus reidi of the Queensland Museum 41, 269-280. and Trichosurus dicksoni, two new fossil phalangerids COOKE, B.N., 1999. Wanburoo hilarus gen. et sp. nov., a (Marsupialia: Phalangeridae) from the Miocene of lophodont bulungamayine kangaroo (Marsuplialia: northwestern Queensland. In Possums and Macropodoidea: Bulungamayinae) from the Miocene opossums: studies in evolution, M. ARCHER ed., Surrey deposits of Riversleigh, northwestern Queensland. Beatty & Sons and The Royal Zoological Society of Records of the Western Australian Museum New South Wales, Sydney, 2, 527-36. Supplement No.57, 239-253. FLANNERY, T.F., ARCHER, M. & PLANE, M., 1983. Middle COOKE, B.N., 2000. Cranial remains of a new species of Miocene kangaroos (Macropodoidea: Marsupialia) balbarine kangaroo (Marsupialia: Macropodoidea) from three localities in northern Australia, with a from the Oligo-Miocene freshwater limestone deposits description of two new subfamilies. BMR Journal of of Riversleigh World Heritage Area, northern Australia. Australian Geology & Geophysics 7, 287-302. Journal of Paleontology 74, 317-326. GAFFNEY, E.S., 1996. The postcranial morphology of COVACEVICH, J. COUPER, P., MOLNAR, R., WITTEN, G. & YOUNG, Meiolania platyceps and a review of the W., 1990. Miocene dragons from Riversleigh: new Meiolaniidae. Bulletin of the American Museum of data on the history of the family Agamidae (Reptilia: Natural History No. 229, 1-166. Squamata) in Australia. Memoirs of the Queensland GAFFNEY, E.S., ARCHER, M. & WHITE A., 1989. Chelid turtles Museum 29, 339-360. from the Miocene freshwater limestones of Riversleigh CREASER, P., 1997. Oligocene-Miocene sediments of Station, northwestern Queensland, Australia. American Riversleigh: the potential significance of topography. Museum Novitates No. 2959, 1-10. Memoirs of the Queensland Museum 41, 303-314. GAFFNEY, E.S., ARCHER, M. & WHITE, A., 1992. Warkalania, ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 13

a new meiolaniid turtle from the Tertiary HAND, S.J., 1998b. Xenorhinos, a new genus of Old World Riversleigh deposits of Queensland, Australia. The leaf-nosed bats (Microchiroptera: Hipposideridae) Beagle 9, 35-48. from the Australian Miocene. Journal of Vertebrate GILLESPIE, A., 1997. Priscileo roskellyae sp. nov. Paleontology 18, 430-439. (Thylacoleonidae, Marsupialia) from the Oligocene- HAND, S.J. & ARCHER, M., 2005. A new hipposiderid genus Miocene of Riversleigh, northwestern Queensland. (Microchiroptera) from an early Miocene bat Memoirs of the Queensland Museum 41, 321-328. community in Australia. Paleontology 48, 371-383. GODTHELP, H., 1997. Zyzomys rackhami sp. nov. (Rodentia, HAND, S.J., ARCHER, M. & GODTHELP, H., 1997. First record Muridae) a rockrat from Pliocene Rackham’s Roost of Hydromops (Microchiroptera: Molossidae) from Site, Riversleigh, northwestern Queensland. Memoirs Australia: its biocorrelative significance. Mémoires of the Queensland Museum 41, 329-333. et Travaux de l’Institut de Montpellier de l’Ecole GODTHELP, H., ARCHER, M., HAND, S.J., PLANE, M.D., 1989. Pratique des Hautes Etudes 21, 153-162. New potoroine from Tertiary Kangaroo Well Local HAND, S., ARCHER, M. & GODTHELP, H., 2001. New Miocene Fauna, N.T. and description of upper dentition of Icarops material (Microchiroptera, Mystacinidae) potoroine Wakiewakie lawsoni from Upper Site Local from Australia, with a revised diagnosis of the genus. Fauna, Riversleigh. Conference on Australasian Memoirs of the Association of Australasian Vertebrate Evolution, Palaeontology and Systematics Palaeontologists 25, 139-146. (CAVEPS) Abstracts, 6. HAND, S.J., ARCHER, M., GODTHELP, H., RICH, T. & PLEDGE, GOTT, M., 1988. A Tertiary marsupial mole (Marsupialia: N., 1993. Nimbadon, a new genus and three new Notoryctidae) from Riversleigh, northwestern species of Tertiary zygomaturines (Marsupialia: Queensland and its bearing on notoryctemorphian Diprotodontidae) from northern Australia, with a phylogenetic systematics. Unpublished Masters reassessment of Neohelos. Memoirs of the Thesis, University of New South Wales, Sydney. Queensland Museum 33, 193-210. HAND, S.J., 1985. New Miocene megadermatids HAND, S.J. & GODTHELP, H., 1999. First Australian Pliocene (Chiroptera: Megadermatidae) from Australia with species of Hipposideros (Microchiroptera: comments on megadermatid phylogenetics. Hipposideridae). Records of the Western Australian Australian Mammalogy 8, 5-43. Museum Supplement No. 57, 299-306. HAND, S.J., 1990. First Tertiary molossid (Micro- HAND, S.J. & KIRSCH, J. A.W., 2003. Archerops, a new chiroptera: Molossidae) from Australia: its annectent hipposiderid genus (Mammalia: phylogenetic and biogeographic implications. Microchiroptera) from the Australian Miocene. Memoirs of the Queensland Museum 28, 175-192. Journal of Paleontology 77, 1139-1151. HAND, S.J., 1993. First skull of a species of Hipposideros HAND, S.J., MURRAY, P., MEGIRIAN, D., ARCHER, M. & GODTHELP, (Brachipposideros) (Microchiroptera: Hippo- H., 1998. Mystacinid bats (Microchiroptera) from sideridae), from Australian Miocene sediments. the Australian Tertiary. Journal of Paleontology 72, Memoirs of the Queensland Museum 33, 179-192. 538-545. HAND, S.J., 1995. First record of the genus Megaderma HOLT, T.R. & SALISBURY, S.W., 2005. New crocodilian Geoffroy (Microchiroptera: Megadermatidae) from remains from Hiatus A site (Early Miocene), Australia. Palaeovertebrata 24(1-2), 47-66. Riversleigh, north-western Queensland. Conference HAND, S.J., 1996. New Miocene and Pliocene on Australasian Vertebrate Evolution, Palaeontology megadermatids (Mammalia, Microchiroptera) from and Systematics (CAVEPS) Abstracts, p 42. Australia, with comments on broader aspects of HUTCHINSON, M.N., 1992. Origins of the Australian scincid megadermatid evolution. Geobios 29, 365-37. lizards: a preliminary report on the skinks of HAND, S.J., 1997a. Hipposideros bernardsigei, a new Riversleigh. The Beagle 9, 61-69. hipposiderid (Microchiroptera) from the Miocene HUTCHINSON, M.N., 1997. The first fossil pygopod and a reconsideration of the monophyly of related (Squamata, Gekkota), and a review of mandibular species groups. Münchner Geowissenschaftliche variation in living species. Memoirs of the Abhandlungen A 34, 73-92. Queensland Museum 41, 355-366. HAND, S.J., 1997b. New Miocene leaf-nosed bats KEAR, B.P., 2002. Phylogenetic implications of (Microchiroptera: Hipposideridae) from Riversleigh, macropodid (Marsupialia: Macropodoidea) northwestern Queensland. Memoirs of the Queensland postcranial remains from Miocene deposits of Museum 41, 335-349. Riversleigh, northwestern Queensland. Alcheringa HAND, S.J., 1997c. Miophyllorhina riversleighensis gen. 26, 299-318. et sp. nov., a Miocene leaf-nosed bat KEAR, B.P., ARCHER, M. & FLANNERY, T.F., 2001a. (Microchiroptera: Hipposideridae) from Riversleigh, Bulungamayine (Marsupialia: Macropodidae) Queensland. Memoirs of the Queensland Museum postcranial elements from the late Miocene of 41, 351-354. Riversleigh, northwestern Queensland. Memoirs of HAND, S.J., 1998a. Riversleigha williamsi gen. et sp. nov., the Association of Australasian Palaeontologists 25, a large Miocene hipposiderid (Microchiroptera) from 103-122. Riversleigh, Queensland. Alcheringa 22, 259-276. KEAR, B.P., ARCHER, M. & FLANNERY, T.F., 2001b. 14 ARCHER ET AL. ALCHERINGA

Postcranial morphology of Ganguroo bilamina MUIRHEAD, J. & ARCHER, M., 1990. Nimbacinus dicksoni, Cooke, 1997 (Marsupialia: Macropodidae) from the a plesiomorphic thylacine (Marsupialia: middle Miocene of Riversleigh, northwestern Thylacinidae) from Tertiary deposits of Queensland Queensland. Memoirs of the Association of and the Northern Territory. Memoirs of the Australasian Palaeontologists 25, 123-138. Queensland Museum 28, 203-221. KEMP, A., 1991. Australian Mesozoic and Cainozoic MUIRHEAD, J. & FILAN, S., 1995. Yarala burchfieldi, a lungfish. In Vertebrate Palaeontology of Australasia, plesiomorphic bandicoot (Marsupialia, P. VICKERS-RICH, J.M. MONAGHAN, R.F. BAIRD, & T.H. Peramelemorphia) from Oligo-Miocene deposits of RICH eds, Pioneer Design Studio, Lilydale, 465-496. Riversleigh, northwestern Queensland. Journal of KEMP, A., 1992. New cranial remains of neoceratodonts Paleontology 69, 127-134. (Osteichthyes: Dipnoi) from the late Oligocene to MUIRHEAD, J. & GILLESPIE, A.K., 1995. Additional parts of middle Miocene of northern Australia, with comments the type specimen of Thylacinus macknessi on generic characters for Cenozoic dipnoans. Journal (Marsupialia: Thylacinidae) from Miocene deposits of Vertebrate Paleontology 12, 284-293. of Riversleigh, northwestern Queensland. Australian KEMP, A., 1997. A revision of Australian Mesozoic and Mammalogy 18, 55-60. Cenozoic lungfish of the family Neoceratodontidae MUIRHEAD, J. & WROE, S., 1998. A new genus and species, (Osteichthyes: Dipnoi), with a description of four Badjcinus turnbulli (Thylacinidae: Marsupialia), new species. Journal of Paleontology 71, 713-733. from the late Oligocene of Riversleigh, northern LOUYS, J., BLACK, K., ARCHER, M., HAND, S.J. & GODTHELP, Australia, and an investigation of thylacinid H., in press. Descriptions of koala material from the phylogeny. Journal of Vertebrate Paleontology 18, middle Miocene of Riversleigh, northwestern 612-626. Queensland and its implications for the genus MURRAY, P.F., 1986. Propalorchestes novaculacephalus Litokoala (Marsupialia, Phascolarctidae). gen. et sp. nov., a new palorchestid (Diproto- Alcheringa. dontoidea: Marsupialia) from the middle Miocene MCKENZIE, K.G., ENGELBRETSEN, M., ARCHER, M. & PRICE, Camfield Beds, Northern Territory, Australia. The E., 2004. Ostracoda from the Miocene, Riversleigh Beagle 3(1), 195-211. World Heritage deposits, Queensland, including MURRAY, P., 1990. Primitive marsupial tapirs Progomphocythere, n. gen., with a discussion of (Propalorchestes novaculacephalus Murray and P. palaeoenvironments and age. Bollettino della Societa ponticulus sp. nov.) from the mid-Miocene of north Palaeontologica Italiana 43(1-2), 321-330. Australia (Marsupialia: Palorchestidae). The Beagle MC MICHAEL, D.F., 1968. Non-marine Mollusca from 7(2), 39-51. Tertiary rocks in northern MURRAY, P. & MEGIRIAN, D., 2000. Two new genera and Australia. Bulletin of the Bureau of Mineral three new species of Thylacinidae (Marsupialia) Resources, Geology & Geophysics (Australia) from the Miocene of the Northen Territory, 80, 133-159. Australia. The Beagle 16, 145-162. MEGIRIAN, D., 1992. Interpretation of the Miocene Carl MURRAY, P.F., MEGIRIAN, D., RICH, T.H., PLANE, M., BLACK, Creek Limestone, northwestern Queensland. The K., ARCHER, M., HAND, S.J. & VICKERS-RICH, P., 2000. Beagle, Records of the Northern Territory Museum Morphology, systematics and evolution of the of Arts and Sciences 9, 219-248. marsupial genus Neohelos Stirton (Diprotodontidae, MENU, H., HAND, S.J. & SIGÉ, B., 2002. Oldest Australian Zygomaturinae). Museums and Art Galleries of the vespertilionid (Microchiroptera) from the early Northern Territory Research Report 6, 1-141. Miocene of Riversleigh, Queensland. Alcheringa 26, MUSSER, A. & ARCHER, M., 1998. New information about 319-331. the skull and dentary of the Miocene platypus MUIRHEAD, J., 1992. A specialised thylacinid, Thylacinus Obdurodon dicksoni, and a discussion of macknessi, (Marsupialia: Thylacinidae) from ornithorhynchid relationships. Philosophical Miocene deposits of Riversleigh, northwestern Transactions of the Royal Society of London B 353, Queensland. Australian Mammalogy 15, 67-75. 1063-1079. MUIRHEAD, J., 1994. Systematics, evolution and MYERS, T.J. & ARCHER, M., 1997. Kuterintja ngama palaeobiology of recent and fossil bandicoots (Marsupialia, Ilariidae): a revised systematic analysis (Peramelemorphia, Marsupialia). Unpublished PhD based on material from the late Oligocene of thesis, University of New South Wales, Sydney. Riversleigh, northwestern Queensland. Memoirs of MUIRHEAD, J., 1997. Two new early Miocene thylacines the Queensland Museum 41, 379-392. from Riversleigh, northwestern Queensland. Memoirs MYERS, T.J., CROSBY, K., ARCHER, M. & TYLER, M., 2001. of the Queensland Museum 41, 367-377. The Encore Local Fauna, a late Miocene assemblage MUIRHEAD, J., 2000. Yaraloidea (Marsupialia, from Riversleigh, northwestern Queensland. Peramelemorphia), a new superfamily of marsupial Memoirs of the Association of Australasian and a description and analysis of the cranium of the Palaeontologists 25, 147-154. Miocene Yarala burchfieldi. Journal of Paleontology PATTERSON, C. & RICH, P.V., 1987. The fossil history of 74, 512-523. the emus, Dromaius (Aves: Dromaiinae). Records ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 15

of the South Australian Museum 21, 85-117. TRAVOUILLON, K.J., ARCHER, M., HAND, S.J. & GODTHELP, H., PLEDGE, N.S., 2005. The Riversleigh wynyardiids. Memoirs in press. Multivariate analyses of the Riversleigh of the Queensland Museum 51, 135-169. local faunas, north-western Queensland. Alcheringa. PLEDGE, N.S., ARCHER, M., HAND, S.J. & GODTHELP, H., 1999. TYLER, M.J., 1989. A new species of Lechriodus (Anura: Additions to knowledge about ektopodontids Leptodactylidae) from the Tertiary of Queensland, (Marsupialia: Ektopodontidae): including a new with a redefinition of the ilial characteristics of the species Ektopodon litolophus. Records of the Western genus. Transactions of the Royal Society of South Australian Museum Supplement 57, 255-264. Australia 113, 15-21. RICH, P.V., 1979. The Dromornithidae, an extinct family TYLER, M.J., 1990. Limnodynastes Fitzinger (Anura: of large ground birds endemic to Australia. Bulletin Leptodactylidae) from the Cainozoic of Queensland. of the Bureau of Mineral Resources, Geology and Memoirs of the Queensland Museum 28, 779-784. Geophysics (Australia) 184, 1-196. TYLER, M.J., 1991a. A large new species of Litoria (Anura: ROBERTS, K.K., ARCHER, M., HAND, S.J. & GODTHELP, H., in Hylidae) from the Tertiary of Queensland. press. A new genus and species of extinct Miocene Transactions of the Royal Society of South Australia ringtail possums (Marsupialia: Pseudocheiridae). 115, 103-105. American Museum Novitates. TYLER, M.J., 1991b. Crinia Tschudi (Anura: SCANLON, J.D., 1996. Studies in the palaeontology and Leptodactylidae) from the Cainozoic of Queensland, systematics of Australian snakes. Unpublished PhD with the description of a new species. Transactions thesis, University of New South Wales, Sydney. of the Royal Society of South Australia 115, 99-101. SCANLON, J.D., 1997. Nanowana gen. nov., small madtsoiid TYLER, M.J., HAND, S.J., & WARD, V.J., 1990. Analysis of snakes from the Miocene of Riversleigh: sympatric the frequency of Lechriodus intergerivus Tyler species with divergently specialised dentition. (Anura: Leptodactylidae) in Oligo-Miocene Local Memoirs of the Queensland Museum 41, 393-412. Faunas of Riversleigh Station, Queensland. SCANLON, J.D., 2001. Montypythonoides: the Miocene Proceedings of the Linnean Society of New South snake Morelia riversleighensis (Smith and Plane, Wales 112, 105-109. 1985) and the geographical origin of pythons. WHITE, A.W., 1997. Cainozoic turtles from Riversleigh, Memoirs of the Association of Australasian northwestern Queensland. Memoirs of the Queensland Palaeontologists 25, 1-35. Museum 41, 413-422. SCANLON, J.D. & LEE, M.S.Y., 2000. The Pleistocene WHITE, A.W. & ARCHER, M., 1994. Emydura serpent Wonambi and the early evolution of snakes. lavarackorum, a new Pleistocene turtle (Pleurodira: Nature 403, 416-420. Chelidae) from fluviatile deposits at Riversleigh, SCANLON, J.D., LEE, M.S.Y. & ARCHER, M., 2003. Mid- northwestern Queensland. Records of the South Tertiary elapid snakes (Squamata, Colubroidea) from Australian Museum 27, 159-167. Riversleigh, northern Australia: early steps in a WILLIS, P.M.A., 1993. Trilophosuchus rackhami gen. et continent-wide adaptive radiation. Geobios 36, 573- sp. nov., a new crocodilian from the early Miocene 601. limestones of Riversleigh, northwestern Queensland. SHEA, G.M. & HUTCHINSON, M.N., 1992. A new species of Journal of Vertebrate Paleontology 13, 90-98. lizard (Tiliqua) from the Miocene of Riversleigh, WILLIS, P.M.A., 1997. New crocodilians from the late Queensland. Memoirs of the Queensland Museum Oligocene White Hunter Site, Riversleigh, 32, 303-310. northwestern Queensland. Memoirs of the Queensland SMITH, M. J. & PLANE, M.D., 1985. Pythonine snakes Museum 41, 423-438. (Boidae) from the Miocene of Australia. BMR Journal WILLIS, P.M.A., 2001. New crocodilian material from of Australian Geology & Geophysics 9, 191-95. the Miocene of Riversleigh (northwestern STIRTON, R.A., 1967. The Diprotodontidae from the Queensland, Australia). In Crocodilian biology and Ngapakaldi Fauna, South Australia. Bulletin of the evolution, G.C. GRIGG, F. SEEBACHER, & C.E. FRANKLIN Bureau of Mineral Resources, Geology & Geophysics eds, Surrey Beatty & Sons, Chipping Norton, Sydney, (Australia) 85, 1-44. 64-74. STIRTON, R.A., TEDFORD, R.H. AND WOODBURNE, M.O., 1968. WILLIS, P.M.A. & ARCHER, M., 1990. A Pleistocene Australian Tertiary deposits containing terrestrial longirostrine crocodilian from Riversleigh: first fossil mammals. University of California Publications in occurrence of Crocodylus johnstoni Krefft. Memoirs Geological Sciences 77, 1-30. of the Queensland Museum 28, 159-163. TEDFORD, R.H., 1967. Fossil mammal remains from the WILLIS, P.M.A. & MOLNAR, R.E., 1997. A review of the Tertiary Carl Creek Limestone, north-western Plio-Pleistocene crocodilian genus Pallimnarchus. Queensland. Bulletin of the Bureau of Mineral Proceedings of the Linnean Society of New South Resources, Geology and Geophysics (Australia) 92, Wales 117, 223-242. 217-237. WROE, S., 1996a. An investigation of phylogeny in the THOMSON, S., WHITE, A. & GEORGES, A., 1997. Re-evaluation of giant extinct rat kangaroo Ekaltadeta (Propleopinae, Emydura lavarackorum: identification of a living fossil. Potoroidae, Marsupialia). Journal of Paleontology Memoirs of the Queensland Museum 42, 327-335. 70, 681-690. 16 ARCHER ET AL. ALCHERINGA

WROE, S., 1996b. Muribacinus gadiyuli, (Thylacinidae: (11) Archer et al. 1987. Marsupialia), a very plesiomorphic thylacinid from (12) Arena 1997. the Miocene of Riversleigh, northwestern Queensland, and the problem of paraphyly for the (13) Arena, D.A., unpublished observations 2005. Dasyuridae (Marsupialia). Journal of Paleontology (14) Bassarova 1999. 70, 1032-1044. (15) Bassarova et al. 2001. WROE, S., 1997a. Mayigriphus orbus gen. et sp. nov., a (16) Black 1992. Miocene dasyuromorphian from Riversleigh, northwestern Queensland. Memoirs of the Queensland (17) Black 1997a. Museum 41, 439-448. (18) Black 1997b. WROE, S., 1997b. Stratigraphy and phylogeny of the (19) Black, in press. giant extinct rat kangaroos (Propleopinae, (20) Black, 2005. Hypsiprymnodontidae, Marsupialia). Memoirs of the Queensland Museum 41, 449-456. (21) Black & Archer 1997a. WROE, S., 1998a. A new ‘bone–cracking’ dasyurid (22) Black & Archer 1997b. (Marsupialia), from the Miocene of Riversleigh, (23) Boles 1992. northwestern Queensland. Alcheringa 22, 277-284. (24) Boles 1993a. WROE, S., 1998b. New marsupicarnivoran fossil material from Tertiary deposits in Queensland and its (25) Boles 1993b. significance in the interpretation of Australian (26) Boles 1993c. marsupicarnivore evolution. Unpublished PhD thesis, (27) Boles 1995. University of New South Wales, Sydney. (28) Boles 1997a. WROE, S., 1999. The geologically oldest dasyurid, from the Miocene of Riversleigh, north-west Queensland. (29) Boles 1997b. Palaeontology 42, 501-527. (30) Boles 1998. WROE, S., 2001a. A new genus and species of (31) Boles 1999. dasyuromorphian from the Miocene of Riversleigh, (32) Boles 2000. northern Australia. Memoirs of the Association of Australasian Palaeontologists 25, 53-59. (33) Boles 2001. WROE, S., 2001b. Maximucinus muirheadae, gen. et. sp. (34) Boles 2005a. nov. (Thylacinidae : Marsupialia), from the Miocene (35) Boles 2005b. of Riversleigh, north-western Queensland, with (36) Boles 2005c. estimates of body weights for fossil thylacinids. Australian Journal of Zoology 49, 603-614. (37) Boles 2005d. WROE, S. & ARCHER, M., 1995. Extraordinary (38) Brammall 1999. diphyodonty-related change in dental function for a (39) Brammall & Archer 1997. tooth of the extinct marsupial Ekaltadeta ima (40) Brewer P., unpublished observations 2005. (Propleopinae, Hypsiprymnodontidae). Archives of Oral Biology 40, 597-603. (41) Cooke 1997a. WROE, S. & MUSSER, A., 2001. The skull of Nimbacinus (42) Cooke 1997b. dicksoni (Thylacinidae : Marsupialia). Australian (43) Cooke 1997c. Journal of Zoology 49, 487-514. (44) Cooke 1999. (45) Cooke 2000. APPENDIX (46) Covacevich et al. 1990. List of references used for Table 2: (47) Creaser 1997. (1) Archer 1979. (48) Crosby 2002a. (2) Archer 1982. (49) Crosby 2002b. (3) Archer et al. 1999. (50) Crosby & Archer 2000. (4) Archer et al. 1990. (51) Crosby et al. 2001. (5) Archer & Flannery 1985. (52) Davis & Archer 1997. (6) Archer et al. 1989. (53) Flannery & Archer 1987a. (7) Archer et al. 1988. (54) Flannery & Archer 1987b. (8) Archer et al. 1994. (55) Flannery & Archer 1987c. (9) Archer et al. 1997. (56) Flannery et al. 1983. (10) Archer et al. 1992. (57) Gaffney 1996. ALCHERINGA SPECIES LISTS FROM RIVERSLEIGH 17

(58) Gaffney et al. 1989. (105) Murray 1986. (59) Gaffney et al. 1992. (106) Murray 1990. (60) Gillespie 1997. (107) Murray & Megirian 2000. (61) Gillespie A., unpublished observations 2005. (108) Murray et al. 2000. (62) Godthelp 1997. (109) Musser & Archer 1998. (63) Godthelp et al. 1989. (110) Myers & Archer 1997. (64) Godthelp H., unpublished observations 2005. (111) Myers et al. 2001. (65) Gott 1988. (112) Patterson & Rich 1987. (66) Hand 1985. (113) Pledge 2005. (67) Hand 1990. (114) Pledge et al. 1999. (68) Hand 1993. (115) Rich 1979. (69) Hand 1995. (116) Roberts et al., in press. (70) Hand 1996. (117) Roberts K.K., unpublished observations 2005. (71) Hand 1997a. (118) Scanlon 1996. (72) Hand 1997b. (119) Scanlon 1997. (73) Hand 1997c. (120) Scanlon 2001. (74) Hand 1998a. (121) Scanlon & Lee 2000. (75) Hand 1998b. (122) Scanlon et al. 2003. (76) Hand & Archer 2005. (123) Shea & Hutchinson 1992. (77) Hand et al. 1997. (124) Smith & Plane 1985. (78) Hand et al. 2001. (125) Stirton 1967. (79) Hand et al. 1993. (126) Stirton et al. 1968. (80) Hand & Godthelp 1999. (127) Tedford 1967. (81) Hand & Kirsch 2003. (128) Thomson et al. 1997. (82) Hand et al. 1998. (129) Tyler 1989. (83) Holt & Salisbury 2005. (130) Tyler 1990. (84) Hutchinson 1992. (131) Tyler 1991a. (85) Hutchinson 1997. (132) Tyler 1991b. (86) Kear 2002. (133) Tyler et al. 1990. (87) Kear et al. 2001a. (134) White 1997. (88) Kear et al. 2001b. (135) White & Archer 1994. (89) Kemp 1991. (136) Willis 1993. (90) Kemp 1992. (137) Willis 1997. (91) Kemp 1997. (138) Willis 2001. (92) Louys et al. 2006. (139) Willis & Archer 1990. (93) McKenzie et al. 2004. (140) Willis & Molnar 1997. (94) McMichael 1968. (141) Wroe 1996a. (95) Megirian D., pers.comm. 2005. (142) Wroe 1996b. (96) Menu et al. 2002. (143) Wroe 1997a. (97) Muirhead 1992. (144) Wroe 1997b. (98) Muirhead 1994. (145) Wroe 1998a. (99) Muirhead 1997. (146) Wroe 1998b. (100) Muirhead 2000. (147) Wroe 1999. (101) Muirhead & Archer 1990. (148) Wroe 2001a. (102) Muirhead & Filan 1995. (149) Wroe 2001b. (103) Muirhead & Gillespie 1995. (150) Wroe & Archer 1995. (104) Muirhead & Wroe 1998. (151) Wroe & Musser 2001. Multivariate analyses of Cenozoic mammalian faunas from. Riversleigh, northwestern Queensland

KENNY J. TRAVOUILLON, MICHAEL ARCHER, SUZANNE J. HAND and HENK GODTHELP

TRAVOUILLON K.J., ARCHER, M., HAND, S.J. & GODTHELP, H., 2006. Multivariate analyses of Cenozoic mammalian faunas from Riversleigh, north-western Queensland. Alcheringa Special Issue 1, 323-349. ISBN 0 9757 5 7.

Mammalian faunal lists of Riversleigh fossil sites were compared using several techniques to assess the age relationships between 75 Riversleigh sites and the central Australian Ngama, Kutjamarpu and the Northern Territory Bullock Creek Local Faunas. Presence/ absence data of the sites’ mammalian faunas were compared in terms of faunal similarity using cluster analysis, ordination, seriation and cladistics (the latter method has not previously been used to compare such data). The analysis was repeated using only sites with eight mammal taxa or more. The analyses also tested and supported the Riversleigh System concept introduced by Archer et al. (1989). The analyses confirmed the placement of several sites previously assigned (Creaser 1997, Arena 2004) to a System based on their geology and topography. However, the analyses could not confirm the assignment of sites with less than eight taxa. Sites with eight mammal taxa or more were used as diagnostic sites for a preliminary description of the “Systems” or “Faunal Zones” (sensu Arena 2004).

Kenny J. Travouillon, Michael Archer, Suzanne J. Hand & Henk Godthelp, School of Biological, Earth and Environmental Sciences, University of New South Wales, New South Wales 2052; received 19.8.2005, accepted 20.9.2006.

Key words: Biochronology, multivariate analysis, Riversleigh, Australia, Oligo-Miocene.

VERTEBRATE FOSSIL DEPOSITS in the stratigraphic unit of worldwide significance’ or Riversleigh World Heritage Area, located in Lawn ‘the rocks formed during a period of geological Hill National Park, northwestern Queensland, time’: Bates & Jackson 1984). Megirian (1994) have been excavated since the late 1970s. Over challenged the Systems of Archer et al. (1989) 300 sites have been recorded, spanning the late by pointing out that it involves a combination of Oligocene to the Holocene. The faunal geological and faunal concepts rather than just composition from each site has been collected, one or the other. Moreover, Archer et al. (1997) processed and interpreted to represent a unique used both geological and faunal data in their local fauna (LF), assemblage or local biocorrelation of the Riversleigh assemblages. palaeocommunity (Archer et al. 1997). Archer et Radiometric dating of Riversleigh sites, al. (1989) also introduced the term “System” to using U/Pb isotopes recovered from calcite define three sequential combinations of rock/ samples, is in progress (Elizabeth Price, pers. faunal assemblages interpreted to span the late comm. 2006). In the interim and to date, Oligocene to late Miocene sediments of biocorrelation (comparison of fossil taxa in Riversleigh. System A was interpreted as late undated deposits with others in securely-dated Oligocene, System B as early Miocene and deposits) and stage of evolution have been the System C as middle to early late Miocene. This principle tools used to estimate the relative ages informal use of the term ‘System’ differs from the of Riversleigh fossil faunas (Archer et al. 1995, more traditional geological term ‘System’ which Archer et al. 1997, Murray & Megirian 2000, is a rock-specific term (i.e. ‘a major chrono- Myers & Archer 1997, Woodburne et al. 1994). ISBN 0 9757 5 7/2006/27 $3.00 ©AAP The Kutjamarpu LF (South Australia), the Ngama 324 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

LF (South Australia) and the Bullock Creek LF characterisable as Systems A, B and C; (Northern Territory) are the principle sources of 3. Systems A, B and C are sequential in time the correlation. The Riversleigh System A LFs with A being the oldest and C being the have been demonstrated to correlate with the youngest; and Ngama LF (late Oligocene, magneto- 4. Specific LFs, groups of LFs or Systems at stratigraphically dated at about 24-26 Ma), Riversleigh accumulated at periods of time System B LFs with the Kutjamarpu LF (early that correlate with the Bullock Creek, Miocene) and System C with the Bullock Creek Kutjamarpu and Ngama LFs. LF (middle Miocene) (Archer et al. 1995, Archer In addition, research carried out in this study et al. 1997, Archer et al. 1999, Murray & Megirian enables tests of current hypotheses about the 2000, Myers & Archer 1997, Woodburne et al. relationships of individual assemblages (e.g. 1994). Topographic and stratigraphic data have whether Keith’s Chocky Block Site is referable to been very useful in assessing the relative ages System B or C [Morrell 2002]). We also compare of the sites based primarily on demonstrated or the results given by different analyses and by inferred superposition (Creaser 1997, Arena different similarity indices and assess the 2004). approach most suitable for data of this kind. Statistical techniques such as cluster analysis and ordination have been used to interpret similarities of individual local faunas (LFs) at Materials and Methods specific taxonomic levels (de Bonis et al. 1992, Data compilation Shi 1993, Bennington & Bambach 1996, Bonuso The data examined consist of the lists of land et al. 2002, Elewa, 2004, Fenerci-Masse et al., mammals identified to species level only from 75 2004, Myers 2002, Palombo et al. 2002, Peláez- Riversleigh sites and from three non-Riversleigh Campomanes et al. 2003). Although the statistical Australian sites, Kutjamarpu, Ngama and Bullock method used by different authors may be slightly Creek, compiled from published and unpublished different worldwide (i.e. using different similarity sources. Bats were excluded from the analysis indices or different types of ordination), because there is evidence that they may skew fundamentally it is a well established method. the results due to potential taphonomic biases Rich et al. (1991), for example, used cluster (Hernández Fernández & Peláez-Campomanes analysis (Simpson’s Coefficient) at the generic 2003). The Riversleigh sites used are listed in level to assess the taxic similarity of Australian Table 1. The raw data were extracted from the Cainozoic fossil vertebrate sites including a updated species list compiled by Archer et al. number from Riversleigh. In the appendix of (2006) for the Riversleigh faunas. The species Murray et al. (2000), Rich also performed a cluster lists used for Kutjamarpu, Ngama and Bullock analysis on Neohelos populations to extract Creek are listed in the Appendix. The compiled taxonomic and biochronological meaning from data consists of all mammalian taxa identified the data. Similarly, Megirian et al. (2004) used a before August 2005. A total of 215 mammal cluster analysis to compare the faunal similarity species (280 species from Riversleigh and 37 from of Riversleigh sites. Their results are somewhat the non-Riversleigh sites) were included as different from ours and are compared to our presence/absence data (0 indicates absence; 1 results and thoroughly reviewed in the indicates presence). Although abundance data Discussion. are usually found to be more descriptive of the In this study we test four main hypotheses: fauna in palaeontological studies than presence/ 1. Riversleigh sites accumulated fossils at absence data (Johnson & McCormick 1999), it is different periods of time; 2. Riversleigh sites accumulated fossils Table 1. List of sites with abbreviations ordered by during three main rock/taxa intervals Systems (as found in the literature). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 325 326 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

very difficult to collect abundance data from the taxonomic similarity (de Bonis et al. 1992, Shi Riversleigh material. Aside from being time- 1993, Bennington & Bambach 1996, Bonuso et consuming (with over 30,000 Riversleigh al. 2002, Elewa 2004, Palombo et al. 2002, Peláez- specimens now registered), this process Campomanes et al. 2003). Cluster analysis was involves assignment of all specimens to performed at the species level only, using the taxonomic groups. The majority of postcranial unweighted pair group method in the software fossils from Riversleigh have not been assigned PAST. In the unweighted pair group method, a to a taxon because most of the specimens are number of similarity indices can be used. disassociated and isolated postcranials are According to Hammer (2002) and Hammer & generally not diagnostic at the species level. Harper (2006), Dice’s, Jaccard’s, Simpson’s and Raup-Crick’s similarity indices are the most Data analysis suitable indices to use for presence/absence The data were first entered in an Excel spread data. Each of these indices is used for the sheet and then transferred to the appropriate clusters. program for the analysis. There are several Ordination methods to measure sample size (e.g. NIS, number Ordination is also a very widely used technique of identified specimens) and determine whether for comparing assemblages using taxonomic fossil sites are a representative sample of the similarity (de Bonis et al. 1992, Shi 1993, Bonuso original community. However, as outlined above et al. 2002, Myers 2002, Elewa 2004, Fenerci- in the data compilation section, determining NIS Masse et al. 2004). Hammer (2002) and Hammer values for each site is problematic. For this & Harper (2006) showed that ordination is most reason, the size of sample (total number of effective when it is compared with cluster species) for each site was compared using a bar analysis. There are a number of ordination graph. This can be used to gauge how much methods that can be used. For presence/absence confidence may be placed in the results (small data, Hammer (2002) and Hammer & Harper (2006) size of sample = low confidence). Four types of recommend Principal Coordinate analysis (PCO) analyses were conducted: cluster analysis, as one of the most suitable. Although it is usually ordination, seriation and cladistics. The four used for abundance data, Principal Component analyses approached the data in different ways. Analysis (PCA) is also recommended by Ordination, cluster and cladistic analyses Brenchley & Harper (1998) because it works just examined the taxonomic similarity of individual as well for presence/absence data. PCA and PCO sites and grouped them according to their were both performed on the dataset. similarity. Seriation ordered each site in a sequence, showing a direction in time. Each Seriation analysis was performed twice with two different Seriation is an ordination often used on sets of data. The first set of data contained all stratigraphic data which rearranges data in the the sites shown in Table 1 (78 sites in total). The form of a range chart with species in columns second set of data contained only sites that had and samples (or sites) in rows. This is achieved 8 mammal taxa or more (24 sites in total). We by minimizing the range zones of taxa, which arbitrarily selected 8 taxa as the minimum for the places similar samples in adjacent rows (Brower second set of data after examining the results of & Burroughs 1982). This is done by performing the first set. The results of the two sets of a series of iterations that includes the following analyses were compared to explore the effect of steps: calculation of mean position of presences low sample size. in rows, followed by ordering rows according to Cluster analysis these means; then calculation of mean position Cluster analysis is one of the most widely used of presences in columns, followed by ordering techniques for comparing assemblages using columns according to these means. There are ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 327

two types of seriations: constrained and within the shortest trees already obtained, but unconstrained. In the constrained seriation, without increasing the number of trees saved. samples have a known order (i.e. stratigraphic position) and therefore only taxa are rearranged. The unconstrained version of seriation is used in this analysis because it rearranges both taxa and samples. The seriation was performed using the software PAST. To accommodate the software, sites were moved to columns and taxa to rows.

Cladistics Although primarily used to analyse the relationships of taxa, cladistics has been used for other purposes. For example, O’Brien & Lyman (2003) explored the application of cladistics to archaeology by considering artefacts as human phenotypic characters. It has not been used previously to cluster fossil sites, which is why its potential value to do this was explored in the present study. Cladistics is typically used to group taxa on the basis of their shared derived characteristics. Usually, features unique to one taxon (autapomorphies) and characters common to all taxa (symplesio- morphies) are excluded from the phylogenetic analyses because they are uninformative. Only derived features shared by one or more taxa (synapomorphies) are used in the analysis. In this study, the taxon presence/absence replaces the derived/primitive character state and site names replace taxon names. Using the cladistic concept of autapomorphy, all taxa unique to a single assemblage were removed from the data set. “Symplesiomorphic” taxa do not occur in this data set because no taxa are present in all sites. The analysis was performed using PAUP, version 4.0b10, for Windows, and the perl script PerlRat.pl (the perl script is available from Olaf Bininda-Emonds’ website, http://www2.uni- jena.de/%7Eb6biol2/). The parsimony ratchet (Nixon 1999) was used to analyse the data. The ratchet settings used were 50 batches of 200 replicates, with 25% of the characters randomly Fig. 1. Number of mammalian taxa identified in each of upweighted by a factor of two in each replicate. the Riversleigh site assemblages, and the Bullock Creek, This was then followed by a heuristic search Kutjamarpu and Ngama Local Faunas. 328 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Fig. 2. Cluster Analysis on presence/absence data at the species level, using Dice’s similarity index on all sites (corresponding Systems are shown above sites).

Fig. 3. Cluster Analysis on presence/absence data at the species level, using Jaccard’s similarity index on all sites (corresponding Systems are shown above sites). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 329

Fig. 4. Cluster Analysis on presence/absence data at the species level, using Simpson’s similarity index on all sites (corresponding Systems are shown above sites).

Fig. 5. Cluster Analysis on presence/absence data at the species level, using Raup-Crick’s similarity index on all sites (corresponding Systems are shown above sites). 330 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

The results are summarised in the 50% majority- suited to this kind of data than the first two rule consensus tree. The heuristic method was indices because, unlike Jaccard and Dice indices, used on the basis that it is more suitable for a it is totally insensitive to the size of the larger large data set because it does not conduct an sample (Hammer & Harper 2006). The results exhaustive search (Forey et al. 1992). As for all found using Raup-Crick’s index (Fig. 5) are binary characters, characters (or taxa) were somewhat different. Raup-Crick uses a unordered in the analysis because there are no randomisation method to cluster similar sites evolutionary sequences in presence/absence (Hammer & Harper 2006) and therefore all sites data. Similarly, the trees are unrooted to prevent are placed in a cluster regardless of similarity. It any bias in finding the “ancestral” site. To measure explains why sites that the other indices left the fitness of the data, the consistency index (CI) unresolved (e.g. Ter, RRR, JJ…) due to lack of and retention index (RI) were calculated. Bootstrap similarity are clustered together in the Raup-Crick and jack-knife analyses were used to estimate cluster. However, the three main clusters confidence intervals for the trees. representing the three Systems are more clearly resolved with Raup-Crick’s index, although some Results and Discussion sites do not cluster with their putative System. The total number of mammal taxa present in Results of the cluster analyses performed each site is shown in Fig. 1. Most sites have a on sites with eight mammal taxa or more using very small sample size, with more than 50% of Dice’s, Jaccard’s, Simpson’s and Raup-Crick’s sites having less than 10 taxa. It is difficult to indices are shown in Figs 6, 7, 8 and 9. The results identify the minimum number of species required found by each index is very similar, finding the to be confident in the results, but the smaller same main clusters. WH and D, the only two the sample the higher chance of error. Results System A sites always cluster together. JC, HH, are discussed in relation to the current Gag, KCB, COA and En, putative System C sites, understanding of Systems nomenclature and also cluster together in each analysis. However, included sites (Table 1). CS, WW, NG, U, RSO, Ina, DT, MM, and CR (putative System B sites) always cluster together Cluster analysis with Wang and Ring. Non-Riversleigh sites (BC, Cluster analyses performed on all sites at the Kut and Nga) cluster outside of the Systems, species level using Dice’s, Jaccard’s, Simpson’s and RR (Pliocene) remains unresolved outside and Raup-Crick’s similarity indices are shown of the clusters. in Figs 2, 3, 4 and 5, respectively. The clusters produced by Dice’s (Fig. 2) and Jaccard’s (Fig. Ordination 3) indices are almost identical, despite a few The Principal Components Analysis (PCA) inversions of branches and that the cluster performed on all sites (Fig. 10) shows some branches are slightly shorter using Dice. This separation between the majority of sites of was expected because Dice’s index is less Systems A, B and C. However, sites with a smaller sensitive than Jaccard’s index to differences in sample size tend to clutter near the (0, 0) sample size (Hammer & Harper 2006). Most sites coordinates. Sites with larger sample sizes (U, of the same System cluster together (e.g. Gag, CS, WW, NG, Gag, HH) fall further to the right of HH, LM, JC, KCB, COA, En). However, some (0; 0) coordinates. This means that the first sites do not cluster with sites of the same eigenvector, which accounts for 19.5% of the System (e.g. Wang is interpreted as a System C variation, contains most of the species present site, but clusters with System B sites in the in those sites. The second eigenvector only analysis). Cluster analysis performed using accounts for 8.3% of the variation but it is the Simpson’s index (Fig. 4) gives similar but slightly most useful to distinguish the Systems. more resolved clusters. It is probably better ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 331

Fig. 6. Cluster Analysis on presence/absence data at the species level, using Dice’s similarity index on sites with 8 mammal taxa or more (corresponding Systems are shown above sites).

Fig. 7. Cluster Analysis on presence/absence data at the species level, using Jaccard’s similarity index on sites with 8 mammal taxa or more (corresponding Systems are shown above sites). 332 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Fig. 8. Cluster Analysis on presence/absence data at the species level, using Simpson’s similarity index on sites with 8 mammal taxa or more (corresponding Systems are shown above sites).

Fig. 9. Cluster Analysis on presence/absence data at the species level, using Raup-Crick’s similarity index on sites with 8 mammal taxa or more (corresponding Systems are shown above sites). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 333

Fig. 10. Principal Components Analysis on presence/absence data at the species level on all sites (with convex hulls).

Fig. 11. Principal Coordinates Analysis on presence/absence data at the species level, using Dice’s similarity index on all sites (with convex hulls). 334 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Fig. 12. Principal Coordinates Analysis on presence/absence data at the species level, using Jaccard’s similarity index on all sites (with convex hulls).

Fig. 13. Principal Coordinates Analysis on presence/absence data at the species level, using Simpson’s similarity index on all sites (with convex hulls). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 335

Fig. 14. Principal Coordinates Analysis on presence/absence data at the species level, using Raup-Crick’s similarity index on all sites (with convex hulls).

Fig. 15. Principal Components Analysis on presence/absence data at the species level on sites with 8 mammal taxa or more (with convex hulls). 336 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Fig. 16. Principal Coordinates Analysis on presence/absence data at the species level, using Dice’s similarity index on sites with 8 mammal taxa or more (with convex hulls).

Fig. 17. Principal Coordinates Analysis on presence/absence data at the species level, using Jaccard’s similarity index on sites with 8 mammal taxa or more (with convex hulls). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 337

Fig. 18. Principal Coordinates Analysis on presence/absence data at the species level, using Simpson’s similarity index on sites with 8 mammal taxa or more (with convex hulls).

Fig. 19. Principal Coordinates Analysis on presence/absence data at the species level, using Raup-Crick’s similarity index on sites with 8 mammal taxa or more (with convex hulls). 338 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

In contrast, the Principle Coordinates System B; 3, System C; 4, Pliocene; 5, Pleistocene; analyses (PCO) using Dice’s (Fig. 11) and 6, Recent; 7, Unknown. Jaccard’s (Fig. 12) indices shows no clear All sites (seriation criterion = 0.28): separation between Systems. The convex hulls VIP1,AL1,HSS1, Nga, LL1, JA1, Boid2, LSO1, Dun1,

(a set of points is the smallest convex set that SB1, HS1, BO1, Roo7, JJJ1, AR7, Mesa7, Boles1, includes the points) of each System are covering GOH3, D1, WH1, G1, QL1, UBO1, 300BR7, FT7, BR1, each other, showing no difference between the DT2, JH2, MM2, WW2, BSE2, U2, CS2, RSO2, Bite2,

Systems. The first two eigenvalues account only Ina2, Hel2, RV2, GG1, MPP2, PIR2, NG2, CK3, CR2, for 6.9% and 5.8% of the variation, respectively Wang3, VD2, Out2, QQ3, Bob3, JJS3, COA3, TB2, for Dice, and 5.4% and 4.8% of the variation, KCB3, MIM3, Kut, Ring3, SD3, Gag3, AL903, KJ3, respectively for Jaccard. Main3, HH3, GC3, LD943, LM3, JC3, BC, FF3, En3,

In the case of Simpson’s (Fig. 13) and Raup- Dome3, TT3, Pha3, JJ3, RRR7, RR4, Ter5, CC6, MSC6 Crick’s (Fig. 14) indices, the Systems separate Sites with eight mammal taxa or more more clearly, with still some overlap. Coordinates (seriation criterion = 0.55):

1 (8.3% of the variation for Simpson and 9.3% Nga, Kut, WH1, D1, CR2, DT2, WW2, MM2, CS2, for Raup-Crick) and 2 (7.6% of the variation for Ina2, RSO2, NG2, U2, Wang3, KCB3, COA3, Ring3,

Simpson and 8.9 for Raup-Crick) are equally Gag3, HH3, LM3, JC3, En3, BC, RR4. useful to determine the relationship between the The seriation shows the chronological Systems, and unlike PCA. As in cluster analysis, sequence in which System A sites are followed the Simpson and Raup-Crick indices are not as by System B sites, which are then followed by heavily influenced by sample size as Dice and System C sites. In the seriation with all sites, only Jaccard, explaining why there is less overlap. a few sites are grouped with a different System. The PCA (Fig. 15), and the PCO using Dice’s Boid (B) and GOH (C) are grouped with System A (Fig. 16), Jaccard’s (Fig. 17), Simpson’s (Fig. 18) sites, GG (A) is grouped with System B and TB and Raup-Crick (Fig. 19) indices performed on (B) with System C sites. The sample size is quite sites with eight taxa or more all showed very low for these sites (Boid has 3 taxa, GOH, GG and similar results. Unlike cluster analysis, all the TB have 1 taxon each) which might explain the sites grouped with their own System without error. CK (C) and Wang (C) group with System B overlapping the convex hulls. The first and sites, or CR (B), VD (B) and Out (B) group with second eigenvalue accounts for 19.3 % and 8.4% System C sites. Nga, Kut and BS are placed among of variation in the PCA respectively, 11.9% and System A, B and C sites respectively. RR 8.4% in the PCO with Dice, 9.2% and 6.7% in the (Pliocene), Ter (Pleistocene) and recent sites (CC, PCO with Jaccard, 15.8% and 10.3% in the PCO MSC) follow at the end of the seriation. with Simpson, and 21.1% and 13.5% in the PCO In the seriation using sites with eight with Raup-Crick. BC, Kut, Nga and RR grouped mammal taxa or more, sites followed the System away from the Systems. A, B and C chronological order. In contrast to the seriation with all sites, Kut grouped with Nga Seriation before System A sites. The criterion in this Seriation based on presence/absence data at the seriation (0.55) is also much higher than in the species level is given in Fig. 20 for all sites and seriation with all sites (0.28). This means that the Fig. 21 for sites with eight mammal taxa or more. seriation using sites with eight mammal taxa or Due to the large size of the seriation, Figs 20-21 more is much better resolved and more reliable. have been reduced to fit the page. Taxon and site names are therefore unreadable. Site names Cladistics are recapitulated below and the order of the sites The parsimony Ratchet analysis resulted in 10000 is as shown in the seriation. Subscripts refer to shortest trees using all sites and 156 shortest putative System or age as follows:1, System A; 2, trees using sites with eight mammal taxa or more. ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 339

The trees are summarised below in the 50% or more, 83 out of 195 characters were informative majority-rule consensus trees (Figs 22-23). For (i.e. 112 uninformative). Sites containing all sites, 94 out of 215 characters were informative uninformative characters (300BR to VIP in fig. (i.e. 121 uninformative); for sites with eight taxa 22) are left unresolved in the tree. Indices for the

Fig. 20. Seriation of all sites based on presence/absence data at the species level. 340 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

majority rule consenses were as follows: All sites, CI=0.5012, RI=0.4121; sites with eight taxa or more, CI==0.5821, RI=0.4400. Both values are quite low, indicating that the tree describes the data set poorly. The two trees (all sites and sites with eight mammal taxa or more) resulting from the parsimony analysis are quite similar to results from the cluster analysis. However, the parsimony analysis left more sites unresolved than cluster analysis, due to the removal of uninformative characters. Bootstrap and jack- knife analyses were also performed to test the confidence of the branching. From the bootstrap and jack-knife analyses, there are four well- supported branches suggesting relationships between: - CS, U, WW and NG (all sites and sites with eight mammal taxa or more); - GAG and HH (all sites and sites with eight mammal taxa or more); - Bob and QQ (all sites only); and - Kut and Ng (sites with more than eight mammal taxa only)

Comparison of analyses and size of samples The fauna of each site is a sample of a local community (or local fauna) that has then been compared with other sites. Sites with a small number of species have typically been less well sampled than sites with a large number of species. If the number of species in a site is low due to undersampling, there is a greater chance of error in the similarity comparison. With increasing numbers of species, there is increasing confidence in the results of the comparison. We performed the similarity analyses on all sites and on sites with eight mammal taxa or more to investigate that particular sampling problem. Interpreting the results and making conclusions based on the analyses using all sites could lead to severe error. We used four similarity indices to compare the results given by each of them. In the all site analysis, the results given by Dice’s and Jaccard’s indices were almost identical, although

Fig. 21. Seriation of sites with 8 mammal taxa or more based on presence/absence data at the species level. ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 341

Dice was slightly less sensitive to sample size. the appropriate palaeoenvironment. Simpson’s and Raup-Crick’s indices identified The cladistics method was not particularly the Systems much better than the other two useful in identifying Systems. In the analysis indices. This is in agreement with Hammer's using sites with eight mammal taxa or more, it (2002) and Hammer & Harper's (2006) recommend- clustered small groups such as En and COA, or ations: Simpson’s and Raup-Crick’s indices are CR and MM, outside the System branches. The more suitable when sampling is considered removal of unique species (autapomorphies) from incomplete. In the analysis based on sites with the data set may not be the best method for eight mammal taxa or more, the results given by faunal comparison. In fact, it is recommended the four indices were almost identical, indicating (Etter 1999) to include unique (rare) species that sampling error was reduced or perhaps even because they may be highly important and even removed. characteristic of some of the samples. For Ordination was one of the most useful example, in the analyses En groups with System methods for comparing faunal lists. The Systems C. Myers et al. (2001) demonstrated that En were distinguished in both PCA and PCO, but (Encore Local Fauna) was younger than other they were more clearly separated by PCO. As System C sites. It contains taxa that are more Clarke and Warwick (1994) noted, PCO is more derived than System C taxa and these taxa are flexible in defining dissimilarity than PCA, but unique to En (11 unique taxa out of 19 in total). both suffer from poor distance preserving (the In cladistics analysis, these taxa are treated as proximity between data points does not uninformative, resulting in grouping En with COA accurately reflect their similarities). This based on the eight remaining taxa in common phenomenon is shown in our analysis by the with System C sites. In ordination, cluster grouping of RR with Kut, Nga and BC, although analysis and seriation, En is placed between RR does not share any species with them. COA and BC, or JC and BC. Cluster analysis was not as successful as ordination in defining the Systems, with many Bullock Creek, Kutjamarpu and Ngama Local sites remaining unassigned in the all sites Faunas analyses. Clarke and Warwick (1994) pointed out The Ngama Local Fauna (Ng) has been that cluster analysis was weak at working out magnetostratigraphically dated at about 24-26 relationships at higher levels and it is always mya, in the late Oligocene (Woodburne et al. recommended to use it in conjunction with 1994). The biocorrelation of a taxon shared ordination. This means that the higher branches between Ngama and Riversleigh’s System A of the cluster, which supposedly represent the White Hunter Site (Kuterintja ngama) indicated Systems, are less reliable than the larger that they were of a similar age (Myers & Archer groupings in ordination. Ultimately, ordination 1997, Archer et al. 1997). Murray & Megirian should be used to identify groups (in our case (1992) and Murray et al. (2000) demonstrated Systems) and cluster analysis to identify that Bullock Creek (BC) might be middle Miocene similarities within those groups. in age. BC shares a number of taxa with Seriation was very useful at identifying the Riversleigh’s System C sites (Neohelos stirtoni chronological order of the Systems by placing and Propalorchestes novaculacephalus) and sites in a sequence according to their similarities. the Encore Local Fauna (Wakaleo vanderleueri) System A becomes the oldest followed by suggesting that they are of a similar age. System B and then by System C. Assuming that Woodburne et al. (1994) suggested that the seriation placed the site in the correct Kutjamarpu, like Ngama, was late Oligocene in chronological order, the results of seriation could age, but Archer et al. (1997) argued that it shares be used to make predictions about what species more taxa with Riversleigh’s System B sites than might be expected to be found in a site assuming System A, and hence is probably early Miocene. 342 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 343

In the analyses, Ngama and Kutjamarpu analysis (fig. 18 and table 8), the authors used a always group together either with System A sites presence/absence matrix in a similarity cluster or distantly from all Riversleigh sites. Ngama has analysis to compare mid-Cenozoic formations. more taxa in common with Kutjamarpu (two: The variables used in the matrix are diverse in Bulungamaya sp. A and Peramelemorphian sp. nature: presence of a sediment type, an aquatic B) than Riversleigh (one: Kuterintja ngama) plant, an aquatic vertebrate and the presence at hence its position in the analyses. Kutjamarpu different taxonomic levels (order, family, genus shares six of its taxa with Riversleigh sites: and species) of aquatic invertebrates, amphi- Wakaleo oldfieldi is known from KCB and COA bious vertebrates, terrestrial invertebrates and (System C), Wakiewakie lawsoni from U (System terrestrial vertebrates. One of the assumptions B), Litokoala kutjamarpensis from Gag, GC, HH of multivariate analyses is that each variable is and JC (System C), Marlu kutjamarpensis from given the same weight. This means that all Gag, Ring and Wang (System C), Rizo- variables should be equal and is why multivariate phascolonus crowcrofti from BR, DT and COA analyses are more conventionally used to (System A, B and C) and Neohelos tirarensis compare morphometric measurements or to from 300BR, BR, BO, CS, D, Dun, FT, Ina, JJS, compare assemblages using taxa at the same KCB, MM, NG, PIR, SB, U, UBO and WW taxonomical level (Etter 1999, Hammer 2002, (Systems A, B and C). These Kutjamarpu species Hammer & Harper 2006). The variables used in occur in all three Riversleigh Systems, making the analysis of Megirian et al. are unlikely to be biocorrelation problematic. of equal value and the results given are unsupported. Comparison with the literature In the second analysis (Megirian et al. 2004, Rich et al. (1991) compared the taxonomic fig. 20), data are used from Murray & Megirian composition of Australian Cenozoic terrestrial (2000) to cluster eight Riversleigh sites (D, WH, mammalian sites based on the number of genera MM, NG, CS, Ina, Gag and HH). The sample size shared in common between the sites using of the sites ranged from eight taxa (CS) to two Simpson’s coefficient. They found that taxa (MM). The size of sample is low because all Kutjamarpu and Ngama clustered together with unique species were removed from the data set. all other South Australian Oligo-Miocene sites. Etter (1999) recommends leaving rare species in All three Riversleigh Systems clustered together the data set because these species may be and were closely related to Bullock Creek. characteristic of the sample. In Megirian et al. Performing a similarity analysis at the generic (2004), all the sites clustered according to (or familial) level over a long time span, such as Systems except MM (System B) which clustered carried out by Rich et al. (1991), provides a good with D and WH (System A). The authors understanding of the major chronological concluded that the Riversleigh site assemblages groups. However, genera (and families) exist might be diachronous and dismissed the utility significantly longer than species, and have of the Systems nomenclature. However, those turnover times that can be too long to be of authors did not consider the likelihood that their biocorrelative value for the Oligo-Miocene. results could be undermined by low sample sizes. Similarity analyses performed on shorter time In our analysis, MM’s 12 mammal taxa were spans require the use of species level taxa to included, six times more than in the Megirian et distinguish smaller changes in time. al. (2004) analysis, and MM always clusters with Megirian et al. (2004) performed two System B sites (in the eight mammal taxa or more analyses using Riversleigh sites. In their first analysis). Moreover, recent excavations at Riversleigh have demonstrated that that CS and Fig. 22. 50% majority-rule consensus of 10000 trees of the unordered analysis of all sites (corresponding Systems MM sites are confluent and are almost certainly are next to site names). part of the same deposit. 344 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Revision of the “Systems” from sites with eight mammal taxa or more) show at least three distinct groups (sometimes four Archer et al. (1989) defined the “Systems” as with Encore separate) and therefore supports regionally clustered sites that appear to be the hypotheses and concept of Faunal Zones. superpositionally-related and/or space-related. The diagnostic characters of each faunal zone A “system” is also a stratigraphic analogue of have yet to be defined. As a preliminary the chronological term “period”. To avoid description, sites with eight mammal taxa or more confusion, Arena (2004) proposed using two can be used as diagnostic sites for each Faunal terms to describe Riversleigh sites rather than Zone. D and WH are representatives of Faunal “System”: the faunal concept of System A, B Zone A, CR, CS, DT, Ina, MM, NG, RSO, U and and C would be replaced by “Faunal Zone” A, B WW are representatives of Faunal Zone B, COA, and C and the geological concept by Gag, HH, JC, KCB, LM, Ring and Wang are “Depositional Phase” A, B and C. Arena (2004, representatives of Faunal Zone C and En is 2005) also found that Riversleigh deposits could representative of Faunal Zone D. Taxa unique be interpreted as having been formed and to a Faunal Zone, and those found in two or modified during successive stages of karst more Faunal Zones are listed below. For single development divided into four phases. Encore occurrences of taxa, sites names are given after Site is referred to by Arena (2004, 2005) as Faunal the name of the taxon in brackets. Taxa that are Zone D and Depositional Phase D. not found in the diagnostic sites but are found The results of our analyses (using the data in other sites are in parentheses.

Fig. 23. 50% majority-rule consensus of 156 trees of the unordered analysis of sites with 8 mammal taxa or more (corresponding Systems are shown next to sites). ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 345

Taxa found in Faunal Zone A only: C), Pseudochirops sp. 1 (and En), Pseudochirops Badjicinus turnbulli (WH), Kuterintja ngama sp. 2 (HH), Paljara sp. 1 (Gag), Marlu (WH), (Namilamadeta sp. cf. N. albivenator), kutjamarpensis, Marlu sp. 2 (JC), Marlu cf. sp. (Ngapakaldia sp.), (Silvabestius michaelbirti), 3, Marlu sp. 4, Pildra sp. 1, Pildra sp. 3 (Gag), (Silvabestius johnnilandi), Silvabestius sp. (D), Pildra sp. 4 (LM), Pseudocheiridae new genus 2 (Marada arcanum), Ganawamaya aediculis sp. 2 (Gag), “Strigocuscus” reidi, “Trichosurus” (WH), Balbaroo sp. 3 (WH), Galanarla dicksoni, Phalangeridae new genus 1 sp. 2 tessellata, Wururoo dayamayi (WH), Nambaroo (KCB), (Phalangeridae new genus 3 sp. 1), couperi (WH), Nambaroo sp. 8, Nowidgee sp. 2 Balbaroo sp. 4, Ekaltadeta jamiemulvaneyi (and (WH), Gumardee pascuali (D) En), Bettongia moyesii, Ganguroo sp. 2, Wanburoo hilarus, Wanburoo sp. 2, Taxa found in Faunal Zone B only: Yalkaparidon jonesi Wabulacinus ridei (CS), Ngamalacinus timmulvaneyi, Dasyuridae genus indet. sp. 1 (U), Taxa found in Faunal Zone D (En) only: Dasyuridae genus indet. sp. 2 (U), Dasyuridae Thylacinus sp. cf. T. macknessi, Ganbulanyi genus indet. sp. 3 (U), Dasyuridae genus indet. djadjinguli, Mayigriphus orbus, Phascolarctos sp. 4 (U), Notoryctidae new genus new sp., sp., Wakaleo vanderleueri, Warendja sp. 1, Peramelemorphia new genus 2 sp. 2, Palorchestes annulus, Neohelos sp., Peramelemorphia new genus 2 sp. 3, Trichosurus sp., Ganguroo new sp., Peramelemorphia new genus 5 sp. 2 (U), Rhizosthenurus flanneryi, Marsupialia new (Litokoala garyjohnstoni), Phascolarctidae new genus sp. 2 genus new sp., Thylacoleonidae new genus new sp. (NG), Vombatidae genus 1 sp. 1, Taxa found in Faunal Zones A and B: Namilamadeta sp. (U), Paljara maxbourkei (CS), Wakaleo new sp. 1, Namilamadeta albivenator, Paljara tirarensae (WW), Gawinga aranaea, Namilamadeta crassirostrum, Ngapakaldia ?Djilgaringa sp. (U), Wyulda asherjoeli, bonythoni, Marlu sp. 1, Balbaroo gregoriensis, Phalangeridae new genus 1 sp. 3, Ektopodon Nambaroo sp. 3, Nambaroo sp. 5, Wabularoo sp. cf. E. serratus, Chunia sp., Ektopodontidae naughtoni, Yalkaparidon coheni new genus new sp., Durudawiri inusitatus, Durudawiri anfractus, Ganawamaya ornate Taxa found in Faunal Zones B and C: (WW), Ganawamaya acris, Ganawamaya sp. 4 Obdurodon dicksoni, Thylacinus macknessi, (CR), (Wururoo sp. 3), Nambaroo sp. 2, Barinya wangala, Peramelemorphia new genus Nambaroo sp. 4 (WW), Nambaroo sp. 6, 1 sp. 1, Peramelemorphia new genus 2 sp. 1, Nambaroo sp. 7 (U), (Bulungamaya sp. cf. B. Peramelemorphia new genus 3 sp. 1, delicata), (Ganguroo sp. cf. G. bilamina), Peramelemorphia new genus 4 sp. 1, Wakiewakie lawsoni (U), Yingabalanara Peramelemorphia new genus 4 sp. 2, Yarala richardsoni burchfieldi, Priscileo roskellyae, Vombatidae genus 2 sp. 1, Propalorchestes novacula- Taxa found in Faunal Zone C only: cephalus, Cercartetus new sp., Paljara Maximucinus muirheadae (Ring), Muribacinus nancyhaywardae, Marlu sp. 3, Pildra sp. 2, gadiyuli, Thylacinidae cf. Mutpuracinus Djaludjangi yadjana, Djilgaringa gillespieae, archibaldi (JJ), Joculusium muizoni (Gag), Phalangeridae new genus 1 sp. 1, Phalangeridae Dasyuromorphia new genus new sp. (Gag), new genus 2 sp. 2, Wururoo sp. 2, Hypsi- Peramelemorphia new genus 5 sp. 1, Litokoala prymnodon bartholomaii, Hypsiprymnodon kutjamarpensis, (Litokoala new sp. 1), Wakaleo new sp., Ganguroo bilamina oldfieldi, Nimbadon lavarackorum, Neohelos sp. A (COA), Neohelos stirtoni, (Neohelos sp. Taxa found in Faunal Zones A, B and C: 346 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

Nimbacinus dicksoni, Nimiokoala greystanesi, private supporters including Elaine Clark, Rhizophascolonus crowcrofti, Neohelos Margaret Beavis, Martin Dickson, Sue & Jim tirarensis, Pseudocheiridae new genus 2 sp. 1, Lavarack and Sue & Don Scott-Orr. Vital Phalangeridae new genus 2 sp. 1, Balbaroo assistance in the field has come from many fangaroo, Nowidgee matrix, Bulungamaya hundreds of volunteers as well as staff and delicata postgraduate students of the University of New South Wales. Special thanks to Robin Beck, for Taxa found in Faunal Zones A, B, C and D: providing his help and knowledge for the Burramys brutyi, Ekaltadeta ima cladistic analysis. Many thanks to Rick Arena, Karen Roberts, Mina Bassarova, Pip Brewer, Zac Conclusions Kirkham, Dr Kirsten Crosby, Dr Bernie Cooke, The four hypotheses tested were supported in Karen Black, Anna Gillespie, Dr John Scanlon this study, indicating that Riversleigh sites and Neville Pledge for helping in the gathering accumulated fossils at different periods of time of the data. We thank two referees whose and during four main faunal intervals comments helped refine and strengthen the characterisable as Faunal Zones (sensu Arena arguments presented here. 2004, 2005) A (oldest), B, C and D (youngest), which are sequential in time. Faunal Zone A References ARCHER, M., ARENA, D.A., BASSAROVA, M., BECK, R., BLACK, correlates with Ngama LF and Faunal Zone C K., BOLES, W.E., BREWER, P., COOKE, B.N., CROSBY, K., correlates with Bullock Creek LF. The GILLESPIE, A., GODTHELP, H., HAND, S.J., KEAR, B., LOUYS, biostratigraphic position of Kutjamarpu LF J, MORRELL, A., MUIRHEAD, J., ROBERTS, K.K., SCANLON, remains ambiguous based on current data. J.D., TRAVOUILLON, K.T. & WROE, S. (2006). Current status of species-level representation in faunas from The small sample size for most sites was a selected fossil localities in the Riversleigh World limitation for all techniques and a possible source Heritage Area, northwestern Queensland. Alcheringa. of error in assessing site similarity. Using sites ARCHER, M., ARENA, R., BASSAROVA, M., BLACK, K., BRAMMALL, with eight mammal taxa or more in the analysis J., COOKE, B., CREASER, P., CROSBY, K., GILLESPIE, A., GODTHELP, H., GOTT, M., HAND, S.J., KEAR, B., KRIKMANN, reduced or possibly even removed this limitation A., MACKNESS, B., MUIRHEAD, J., MUSSER, A., MYERS, T., and error. The latter sites were used to produce a PLEDGE, N., WANG, Y. & WROE, S., 1999. The preliminary description of each Faunal Zone. evolutionary history and diversity of Australian mammals. Australian Mammalogy 21, 1-45. ARCHER, M., GODTHELP, H., HAND, S.J. & MEGIRIAN, D., 1989. Acknowledgments Fossil mammals of Riversleigh, northwestern Vital support for research at Riversleigh has come Queensland: preliminary overview of biostratigraphy, from the Australian Research Grant Scheme correlation and environmental change. Australian Zoologist 25, 29-65. (grants to M. Archer and S.J. Hand); the National ARCHER, M., HAND, S.J. & GODTHELP, H., 1991. Australia’s Estate Grants Scheme (Queensland) (grants to lost world. New Holland Publishers Pty Ltd, Sydney. M. Archer and A. Bartholomai); the University 264 pp. of New South Wales; the Commonwealth ARCHER, M., HAND, S.J. & GODTHELP, H., 1995. Tertiary Environmental and Biotic change in Australia. In Department of Environment, Sports and Paleoclimate and evolution, with emphasis on Territories; the Queensland National Parks and human origins, E.S. VRBA, G.H. DENTON, T.C. Wildlife Service; the Commonwealth World PARTRIDGE & L.H. BURCKLE eds, Yale University Press, Heritage Unit; ICI Australia Pty Ltd; the New Haven, 77-90. ARCHER, M., HAND, S.J., GODTHELP, H. & CREASER, P., 1997. Australian Geographic Society; the Queensland Correlation of the Cainozoic sediments of Riversleigh Museum; the Australian Museum; the Royal World Heritage Fossil Property, Queensland, Zoological Society of New South Wales; the Australia. In Actes du Congrès BiocroM’97, J.-P. Linnean Society of New South Wales; Century AGUILAR, S. LEGENDRE & J. MICHAUX eds, Ecole Pratique des Hautes Etudes Institut de Montpellier, Zinc Pty Ltd; Mount Isa City Council, Montpellier, France, 131-152. Outback@Isa, the Riversleigh Society Inc.; and ARENA, R., 2004. The geological history and development ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 347

of the Riversleigh terrain during the middle Tertiary. Oxford. xi + 191 pp. Unpublished PhD Thesis, University of New South HAMMER, O., 2002. Palaeontological community and Wales, Sydney. diversity analysis. From a course in Zürich. http:// ARENA, D.A., 2005. The geological history and folk.uio.no/ohammer/past/community.pdf. development of the Riversleigh terrain during the HAMMER, O. & HARPER, D., 2006. Paleontological Data middle Tertiary. Conference on Australasian Analysis. Blackwell Publishing, UK. 351 pp. Vertebrate Evolution, Palaeontology and Systematics HAND, S., ARCHER, M., RICH, T. & PLEDGE, N., 1993. (CAVEPS) Abstracts, 11. Nimbadon, a new genus and three species of Tertiary BATES, R.L. & JACKSON, J.A. eds, 1984. Dictionary of zygomaturines (Marsupialia, Diprotodontidtidae) geological terms: prepared by the American from northern Australia, with a reassessment of Geological Institute, Anchor Press/Doubleday, New Neohelos. Memoirs of the Queensland Museum 33, York. 193-210. BENNINGTON, J.B. & BAMBACH, R.K., 1996. Statistical testing HERNÁNDEZ FERNÁNDEZ, M. & PELÁEZ-CAMPOMANES, P., 2003. for paleocommunity recurrence: are similar The bioclimatic model: a method of palaeoclimatic assemblages ever the same? Palaeogeography, qualitative inferred based on mammal associations. Palaeoclimatology, Palaeoecology 127, 107-133. Global Ecology and Biogeography 12, 507-517. BLACK, K., 1992. Systematics and palaeobiology of fossil JOHNSON, K.G. & MCCORMICK, T., 1999. The Quantitative phascolarctids from Riversleigh. Unpublished Description of Biotic Change Using Palaeontological BSc(Hons) thesis, University of New South Wales, Databases. In Numerical Palaeobiology, D.A.T. Sydney. HARPER ed, J. WILLEY & Sons, Chichester, 227-248. BONUSO, N., NEWTON, C.R., BROWER, J.C. & IVANY, L.C., LONG, J., ARCHER, M., FLANNERY, T. & HAND, S., 2002. 2002. Statistical testing of community patterns: Prehistoric mammals of Australia and New Guinea. uppermost Hamilton Group, Middle Devonian (New UNSW Press, Sydney. 244 pp. York State: USA). Palaeogeography, MEGIRIAN, D., 1994. Approaches to marsupial Palaeoclimatology, Palaeoecology 187, 1-24. biochronology in Australia and New Guinea. BRENCHLEY, P.J. & HARPER, D.A.T., 1998. Palaeoecology: Alcheringa 18, 259-274. ecosystems, environments and evolution. Chapman MEGIRIAN, D., MURRAY, P., SCHWARTZ, L. & VON DER BORCH, & Hall, London. 402 pp. C., 2004. Late Oligocene Kangaroo Well Local BROWER, J.C. & BURROUGHS, W.A., 1982. A simple Method Fauna from the Ulta Limestone (new name), and for Quantitative Biostratigraphy. In Quantitative climate of the Miocene oscillation across central Stratigraphic Correlation, J.M. CUBITT & R.A. REYMENT Australia. Australian Journal of Earth Sciences 51, eds, 61-83. 701-741. CLARKE, K.R. & WARWICK, R.M., 1994. Change in Marine MORRELL, A., 2002. The Geology and Palaeontology of Communities: An approach to statistical analysis Keith’s Chocky Block Site, Riversleigh, World and interpretation. Plymouth Marine Laboratory, Heritage Property, northwest Queensland. Plymouth, UK. 144 pp. Unpublished BSc (Hons) thesis, University of New CREASER, P., 1997. Oligocene-Miocene sediments of South Wales, Sydney. Riversleigh: the potential significance of topography. MURRAY, P. & MEGIRIAN, D., 1992. Continuity and contrast Memoirs of the Queensland Museum 41, 303-314. in Middle and Late Miocene vertebrate communities DE BONIS, L., BOUVRAIN, G., GERAADS, D. & KOUFOS, G., from the Northern Territory. The Beagle, Records 1992. Diversity and paleoecology of Greek late of the Northern Territory Museum of Arts and Science Miocene mammalian faunas. Palaeogeography, 9, 195-218 Palaeoclimatology, Palaeoecology 91, 99-121. MURRAY, P. & MEGIRIAN, D., 2000. Two new genera and ELEWA, A.M.T., 2004. Quantitative analysis and three new species of Thylacinidae (Marsupialia) from palaeoecology of Eocene Ostracoda and benthic the Miocene of the Northen Territory, Australia. foraminifera from Gebel Mokattam, Cairo, Egypt. The Beagle, Records of the Northern Territory Palaeogeography, Palaeoclimatology, Palaeo- Museum of Arts and Science 16, 145-162. ecology 211, 309-323. MURRAY, P., MEGIRIAN, D., RICH, T., PLANE, M., BLACK, K., ETTER, W., 1999. Community Analysis. In Numerical ARCHER, M., HAND, S. & VICKERS-RICH, P., 2000. Palaeobiology, D.A.T. HARPER ed, J. Wiley & Sons, Morphology, systematics and evolution of the Chichester, 285-355. marsupial genus Neohelos Stirton (Diprotodontidae, FENERCI-MASSE, M., MASSE, J-P. & CHAZOTTES, V., 2004. Zygomaturinae). Northern Territory Museum of Arts Quantitative analysis of rudist assemblages : a key for and Science Research Report 6, 1-141. palaeocommunity reconstructions. The late MYERS, T.J.M., 2002. Classification and Ordination Barremian record from SE France. Palaeogeography, analysis of selected Riversleigh Local Faunas. In Palaeoclimatology, Palaeoecology 206, 133-147. Palaeoecology of Oligo-Miocene Local Faunas FOREY, P.L., HUMPHRIES, C.J., KITCHING, I.L., SCOTLAND, R.W., from Riversleigh, Unpublished PhD thesis, University SIEBERT, D.J. & WILLIAMS, D.M., 1992. Cladistics, A of New South Wales, Sydney, 85-106. practical Course in Systematics. Clarendon Press, MYERS, T. & ARCHER, M., 1997. Kutjerintja ngama 348 TRAVOUILLON, ARCHER, HAND and GODTHELP ALCHERINGA

(Marsupialia, Ilariidae): a revised systematic analysis APPENDIX based on material from the late Oligocene of Riversleigh, northwestern Queensland. Memoirs of Bullock Creek Local Fauna: the Queensland Museum 41, 379-392. MYERS, T., CROSBY, K., ARCHER, M., Tyler, M., 2001. The 5Nimbacinus richi

Encore Local Fauna, a late Miocene assemblage from 5Mutpuracinus archibaldi Riversleigh, northwestern Queensland. Memoir of 7Nimbadon whitelawi the Association of Australasian Palaeontologists 25, Neohelos stirtoni 147-154. 8 NIXON, K. C., 1999. The Parsimony Ratchet, a new method 1Propalorchestes novaculacephalus

for rapid Parsimony analysis. Cladistics 15, 407- 1Wakaleo vanderleueri 414. 1Balbaroo campfieldensis O’BRIEN, M.J. & LYMAN, R.L., 2003. Cladistics and Nambaroo bullockensis Archaeology. The University of Utah Press, Salt 4 Lake City. 288 pp. PALOMBO, M. R., AZANZA, B. & ALBERDI, M. T., 2002. Italian Kutjamarpu Local Fauna: mammal biochronology from the latest Miocene to 1Ankotarinja sp. A the middle Pleistocene: a multivariate approach. Ankotarinja sp. B Geologica Romana 36, 335-368. 1 PELÁEZ-CAMPOMANES, P., MORALES, J., ÁLVAREZ SIERRA, M.A., 1Keeuna sp. A ZANZA RAILE ARCÍA AREDES ERNÁNDEZ A , B., F , S., G P , I., H 3Wakamatha tasselli FERNÁNDEZ, M., HERRÁEZ, E., NIETO, M., PÉREZ, B., 3Dasyuridae genus A sp. B QUIRALTE, V., SALESA, M. J., SÁNCHEZ, I. M. & SORIA, D., Peramelemorphian sp. A 2003. Updated biochronology of the Miocene 3 mammal faunas from the Madrid basin (Spain). In 3Peramelemorphian sp. B

Distribution and migration of Tertiary mammals in 3Peramelemorphian sp. C Eurasia. A volume in honour of Hans De Bruijn, 3Peramelemorphian sp. D J.W.F. REUMER, & W. WESSELS eds, Deinsea 10, 431- Peramelemorphian sp. E 441. 3 RICH, T.H, ARCHER, M., HAND, S.J., GODTHELP, H., MUIRHEAD, 1Litokoala kutjamarpensis LEDGE LANNERY OODBURNE J., P , N.S., F , T.F., W , M.O., 1Neohelos tirarensis CASE, J.A., TEFORD, R.H., TURNBULL, W.D., LUNDELIUS, 1Wakaleo oldfieldi E.L. Jr., RICH, L.S.V., WHITELAW, M.J., KEMP, A. & Rhizophascolonus crowcrofti RICH, P.V., 1991. Australian Mesozoic and Tertiary 1 Terrestrial Mammal Localities. In Vertebrate 1Burramyid sp. A ICKERS ICH Palaeontology of Australasia, P. V -R , J.M. 1Burramyid sp. B MONAGHAN, R.F. BAIRD & T.H. RICH eds, 1005-1058. 1Paljara tirarensae ROBERTS, K., 2004. The Taphonomy and Geological Marlu kutjamarpensis Context of Price Is Right Site, Riversleigh, World 1 Heritage Area, northwestern Queensland. 1Pildra tertius

Unpublished BSc(Hons) thesis, University of New 1Ektopodon serratus South Wales, Sydney. 6Ektopodon litolophus SCHWARTZ, L.R.S. & MEGIRIAN, D., 2004. A new species of Balbaroo sp. A Nambaroo (Marsupialia; Macropodoidea) from the 1 Miocene Camfield Beds of Northern Australia with 1Balbaroo sp. B

observation on the phylogeny of the Balbarinae. 3Nambaroo sp. D Journal of Vertebrate Paleontology 24, 668-675. 1Bulungamaya sp. A SHI, G.R., 1993. Multivariate data analysis in Bulungamaya sp. B palaeoecology and paleobiogeography - review. 1 Palaeogeography, Palaeoclimatology, 1Wakiewakie lawsoni

Palaeoecology 105, 199-234. 1Pinaroo sp. C WOODBURNE, M.O., MACFADDEN, B.J., CASE, J.A., SPRINGER, 1Macropodine genus W sp. A M.S., PLEDGE, N.S., POWER, J.D., WOODBURNE, J.M. & SPRINGER, K.B., 1994. Land mammal biostratigraphy Ngama Local Fauna: and magnetostratigraphy of the Etadunna Formation (Late Oligocene) of South Australia. Journal of 2Obdurodon sp. cf. O. insignis

Vertebrate Paleontology 13, 483-515. 2Dasylurinja kokuminola ALCHERINGA MULTIVARIATE ANALYSIS OF RIVERSLEIGH FAUNAS 349

2Peramelemorphian sp. B

1Litokoala sp. cf. L. kutjamarpensis (previously named kanunkaensis)

1Kuterintja ngama

1Burramys wakefieldi 1 Rich et al. 1991

1Marlu sp. cf. M. kutjamarpensis 2 Pledge pers. comm. 2005

1Pildra magnus 3 Woodburne et al. 1994

1Ektopodon stirtoni 4 Schwartz & Megirian 2004

1Nambaroo sp. B 5 Murray & Megirian 2000

2Bulungamaya sp. A 6 Long et al. 2002

1Pinaroo sp. B 7 Hand et al. 1993

1Purtia sp. A 8 Murray et al. 2000