The Ecology of the (Setonix brachyurus) (: Marsupialia) in

the Northern Jarrah of .

by

Matt Hayward

A thesis submitted in fulfillment of the requirements for the degree of

Doctor of Philosophy

School of Biological, Earth and Environmental Science, University of N.S.W.

JULY 2002

To Mum

2 “…I saw my first quokka. It came shambling out of the low shrub that surrounded the knoll and stood regarding me insolently. It was a malicious-looking beast, with a rat-like cast of features and a small, mean mouth, but it was small and showed no immediate aggressive tendencies. I was not alarmed. I looked at it and it looked at me and the decencies seemed to require some sore of overture on my part. I broke off a piece of cheese and some apple and tossed them to the quokka. It leaned forward and sniffed the offering, then looked up at me suspiciously with no sign of gratitude. “Go on” I said amiably, “tuck in. Cheese and apple never hurt a quokka.” The quokka leaned forward again and picked up the offering with both its tiny paws. It sniffed again, delicately took the apple in its mouth, chewed for a moment, wrinkled its nose and spat the apple out. “All right,” I said. “So don’t like apples. Try the cheese.” It did. It popped the gorgonzola into its mouth and chewed experimentally. Then it fell on its back in a dead faint. I leaped [sic] to my feet in great agitation. I was then a convinced conservationist and thought creatures such as quokkas should be preserved at all costs… I emptied my haversack, picked up the small, quivering body and gently laid it inside. I carefully left the top flap open so the quokka could breathe, put the haversack on my back and leaped, or rather clambered, onto my bicycle… I went racing down the slope, pedalling hard. Then it became apparent that I was moving so fast that pedalling was beside the point. I stopped and sat in the saddle concentrating on steering the bicycle at what was, for me, a breathtaking speed… Then the quokka began to recover. I became aware of a violent stirring at my back and then obvious movements indicating that it was climbing out of the haversack. I was, for a moment, vaguely concerned that it might fall to the ground at this impossible speed and hurt itself, but all I could do was cling to the handlebars and try to keep the bicycle in the middle of the road. I felt the quokka reach the top of the haversack, then I felt its tiny paws on my neck. Even with the rushing wind of my passage I became aware of a strange smell, a combination of mildewed carpet and gorgonzola… There was something shrieking in my left ear. It was the quokka, panicking. It was climbing around my neck trying, it seemed, to get in front of me… Then the quokka sank its teeth into my left ear. I heard my own yelp of pain over the rushing wind and took my left hand off the handlebar to try and prise the quokka loose… Then it stopped … and just let its dead weight hang on my ear from its clamped jaws.

3 A quokka weighs only a kilogram or two, but you try riding a runaway bicycle down a steep hill with a couple of kilograms of quokka hanging onto your left ear… Tugging away at the quokka’s tail, bellowing in pain and fear, I swung into that bend, bouncing away at a hundred kilometres an hour at least, and I hadn’t the slightest chance in the world of getting around it. The machine bounced once at the beginning of the bend, once more at the edge of the cliff, and then sailed swiftly into the air and down into the sea five metres below… I wasn’t hurt and the quokka seemed to get dislodged in the flight. I poked my head out of the water and saw the wretched thing standing on the water’s edge, glaring at me, and snarling. I bought myself a drink at the pub and the barman asked me what had happened to my ear. I told him I had bumped into a tree. Extract from ’s Revenge (Cook 1987)

4

Abstract

The quokka (Setonix brachyurus Quoy & Gaimard 1830) is a medium-sized, macropodid that is endemic to the mesic, south-western corner of Australia.

While being a tourist icon on Rottnest Island, the is threatened with extinction.

It has been intensively studied on Rottnest Island in the 1960s and 1970s, however very little is known of its ecology on the mainland. Additionally the insular and mainland environments are extremely different suggesting that ecological differences between the two populations are likely. Consequently, this study sought to determine the basic autecology of the quokka and identify what factors have attributed to its threatened conservation status. The northern of was selected as the study region due to it being at the northern limit of extant quokka distribution and because it was thought that the factors threatening the quokka would be exacerbated there.

Fossil deposits suggest that the quokka originally occupied an area of approximately 49,000 km2 in the south-western corner of Australia. Historical literature show that they were widespread and abundant when Europeans colonised the region in

1829 but a noticeable and dramatic decline occurred a century later. The arrival of the red to the region coincided almost exactly with this decline and so it was probably ultimately responsible. Continued predation by both it and the feral are likely to have continued the decline, along with destruction and modification through altered fire regimes. Specific surveys and literature searches show that since the 1950s, the area occupied by the quokka has declined by 45% and since 1990 by 29%. Based on the criteria of the IUCN (Hilton-Taylor 2000), the conservation status of the quokka

i

should remain as vulnerable. An endangered status may be more applicable if the quokkas restriction to patches through its existence as a metapopulation is considered.

Trapping of eight sites supporting quokka populations in the mid-1990s revealed three sites now locally extinct despite the ongoing, six year old, fox control programme.

Another three are at serious risk of extinction. Extant population sizes ranged from one to 36 and population density ranged from 0.07 to 4.3 individuals per hectare. This is considered to be below the carrying capacity of each site.

The overall quokka population size in the northern jarrah forest may be as low as

150 adult individuals, of which half are likely to be female. Even the largest extant populations are highly susceptible to stochastic extinction events. This small size was surprising considering the six year old, introduced predator control programme.

Historically, the restriction to discrete habitat patches, the occasional inter-patch movement, the lack of correlation between the dynamics of each population and reports of frequent localised extinctions and colonisations suggest that the quokka population once existed as part of a classic metapopulation. The massive decline of the quokka in the 1930s pushed the metapopulation structure into a non-equilibrium state such that today, the extant populations are the terminal remnants of the original classic metapopulation.

Wild mainland quokkas breed throughout the year. A significant reduction in the number of births occurs over summer and this coincides with a decline in female body weight. Despite this, the mainland quokka is relatively fecund and is able to wean two offspring per year. The level of recruitment from pouch young to independence was low and this may explain the apparent lack of population increase following the initiation of fox control.

ii

A total of 56 trapped quokkas were fitted with a radio collar. Mean home range size for quokkas was 6.39 ha with a core range of 1.21 ha and this was negatively related to population density. Male home ranges were larger than females but not significantly when the sexual size dimorphism was considered. Nocturnal ranges were larger than diurnal ranges reflecting nocturnal departures from the swamp refugia.

Home range sizes varied seasonally, probably due to changes in the distance required to move to obtain sufficient nutrients and water over the dry summer compared to the wet winter and spring.

Telemetry confirmed trapping results that showed no movement between swamps or populations. Home range centres shifted to the periphery of the swamp following the winter inundation and this may increase the species susceptibility to predation. The lack of dispersal is probably caused by quokka populations existing below carrying capacity and following selection for philopatry under the threat of predation for dispersing individuals. Without dispersal to recolonise or rescue unpopulated patches, the collapse of the original quokka metapopulation appears to have occurred.

On a macrohabitat scale, the quokka in the northern jarrah forest is restricted to

Agonis swamp shrubland that form in the open, upper reaches of creek systems on the western side of the forest. This restriction was probably initially due to the high water requirements of the quokka but is likely to have been exacerbated by increased predation pressure since the arrival of the fox. On a microhabitat scale, the quokka is a habitat specialist, preferring early seral stage swamp habitats, probably for foraging, as part of a mosaic of old age swamp that provides refuge.

Despite the six year old, introduced predator control programme, and are still the major cause of mortality to quokkas. Road kills was the other identifiable

iii

cause. Individuals alive at the start of the study had an 81% chance of staying alive until the end. The likelihood of dying was minimised by grouping together with conspecifics, maximising home range size and maximising the time spent within the swampy refuge. Current rates of adult and juvenile survivorship should allow population recovery and so it seems pouch young mortality, reflected by low recruitment, has inhibited the anticipated population increase following predator control.

The confounding effect of inadequate unbaited controls meant that little statistical evidence was available on the impact of introduced predators on the quokka, however the models provided support for earlier hypotheses of these. The presence of a quokka population at a site was related to the amount of poison baits delivered – reflecting predation pressure, the average age of the swamp and a mosaic of early and late seral stages within the swamp habitat. Recently burnt habitat is thought to provide for quokkas and long unburnt habitat provides refuge from predation.

iv Table of Contents

ABSTRACT...... V. TABLE OF CONTENTS...... IX.

List of Tables...... XVII. List of Figures...... XIX. ACKNOWLEDGEMENTS...... XX. INTRODUCTION...... 1. Aims and scope of the thesis...... 5. Structure of this thesis...... 7.

METHODS AND STUDY SI TES...... 8. Study region...... 8. Location, topography and geomorphology...... 8. Geology and soils...... 9. Drainage...... 10. Vegetation ...... 11.

Anthropogenic disturbances...... 12. Climate...... 13. Why this region?...... 18. Study sites...... 18. Chandler Road...... 19.

Hadfield...... 20. Hoffman...... 21. Holyoake...... 21.

Kesners...... 22.

Rosella Road...... 22. Victor Road...... 23. Wild Pig...... 23. Feral predator control...... 24. DISTRIBUTION AND STATUS OF THE QUOKKA, SETONIX BRACHYURUS (MACROPODIDAE: MARSUPIALIA) AND A DISCUSSION ON THE

DECLINE OF AUSTRALIAN NATIVE ...... 33

Introduction ...... 34 Methods ...... 36 Results ...... 39 Fossil record ...... 39 Distribution and abundance: knowledge prior to 1950...... 53 Distribution and abundance: knowledge between 1950 and 1990...... 53 Distribution and abundance: knowledge since 1990 ...... 54 Distribution and abundance: knowledge from 2000...... 54 Changes in the actual area of distribution...... 55 Discussion ...... 57 Status...... 57 The pattern of decline...... 58 Factors influencing the decline...... 62 A more widespread and long-term decline? ...... 63 Predation...... 65 Predation by the ...... 66 Predation by the ...... 67 Predation by the cat...... 71 Summary of the effect of recently introduced predators...... 73 Human predation...... 77 Summary of the effects of exotic predators ...... 80 Predation by birds...... 82 Predation by reptiles...... 82 Competition with introduced and endemic species...... 83 Climatic influences...... 85 The effects of European colonisation...... 89 Fire...... 91 Disease ...... 94 Chains of extinction...... 95 Conclusion...... 96 LOCAL POPULATION STRUCTURE OF A NATURALLY-OCCURRING METAPOPULATION OF THE QUOKKA SETONIX BRACHYURUS (MACROPODIDAE: MARSUPIALIA) (QUOY & GAIMARD 1830) IN THE NORTHERN JARRAH FOREST OF AUSTRALIA...... 101 Introduction...... 102 Study area...... 104 Vegetation and climate...... 104 Study sites...... 105 Fox baiting...... 108 Methods ...... 109 Trapping...... 109 handling and measurements...... 112 Species identification...... 114 Data analysis...... 114 Results ...... 119 Trapping success and population size...... 119 Population composition...... 121 Morphology and condition...... 122 Reproduction ...... 126 Discussion ...... 128 Metapopulation attributes and its collapse...... 128 Population sizes...... 129 Quokkas and their response to fox control...... 133 Population composition...... 135 Morphology and condition...... 137 Reproduction ...... 142 Conclusion...... 146 HOME RANGE AND MOVEMENTS OF THE QUOKKA, SETONIX BRACHYURUS (MACROPODIDAE: MARSUPIALIA) IN THE NORTHERN JARRAH FOREST OF AUSTRALIA. 149 Introduction...... 150 Methods ...... 152 Collar attachment...... 152 Radio telemetry and triangulation ...... 153 Home range estimation and statistical analyses...... 155 Results ...... 165 Differences between the sexes...... 166 Differences between day and night home range size ...... 166 Differences between the sites ...... 166 Differences between the seasons...... 168 Movements...... 169 Dispersal ...... 170 Discussion ...... 172 Home ranges ...... 172 Factors affecting home range size ...... 176 Movements...... 177 Dispersal ...... 179 Home range estimation techniques ...... 181 Collar success ...... 182 Implications for the metapopulation...... 182 Conclusion...... 184 THE HABITAT IN AND AROUND POTENTIAL QUOKKA (SETONIX BRACHYURUS QUOY & GAIMARD 1830) (MACROPODIDAE: MARSUPIALIA) SWAMPS OF THE SOUTH-WEST OF AUSTRALIA. 186 Introduction...... 187 Methods ...... 189 Data analysis...... 191 Results ...... 192 Differentiation between habitat units...... 193 Description of individual habitat units...... 198 Agonis swamp shrubland habitat units...... 199 Freshly burnt Agonis swamp shrubland...... 199 5 – 9 years since fire...... 200 10 - 14 years since fire ...... 201 15 – 19 years since fire...... 202 20 – 24 years since fire...... 202 Long unburnt swamp (more than 25 years since fire)...... 202 Allocasuarina forest ...... 203 Blackberry thicket ...... 203 Blackbutt open forest...... 203 Bullich – blackbutt open forest...... 204 Bullich swamp forest ...... 204 Cleared area...... 204 Dry heath...... 205 Flooded gum forest...... 205 Introduced pasture ...... 205 Jarrah-Marri open forest ...... 205 Lepidosperma – Hypocalymna swamp ...... 206 Paperbark swamp...... 206 Peppermint forest...... 206 plantation...... 207 Revegetation – dense and sparse ...... 207 River forest ...... 207 Soapbush swamp...... 207 Swishbush swamp...... 208 Tea-tree swamp...... 208 Wandoo ...... 208 Changes with age since fire within the Agonis swamp shrubland...... 208 Discussion ...... 209 The effect of fire on Agonis swamp shrublands ...... 209 HABITAT USE AND PREFERENCES OF THE QUOKKA SETONIX BRACHYURUS (QUOY & GAIMARD 1830) (MACROPODIDAE: MARSUPIALIA) IN THE NORTHERN JARRAH FOREST OF AUSTRALIA. 212 Introduction...... 213 Methods ...... 214 Study sites...... 214 Trapping...... 217 Radio telemetry ...... 217 Habitat use analysis...... 218 Results ...... 219 Available habitat types...... 219 Broad habitat use and preferences...... 223 Specific habitat use and preferences ...... 227 Habitat preference and use at individual sites...... 229 Chandler site...... 229 Hadfield...... 230 Kesners...... 231 Rosella Road...... 232 Victor Road...... 233 Discussion ...... 236 Causes of quokka habitat preferences...... 238 Specific habitat preferences of the quokka...... 244 The impact of habitat on local quokka populations...... 247 Implications for the status of the quokka...... 248 Conclusions ...... 249 MORTALITY AND SURVIVORSHIP OF THE QUOKKA SETONIX BRACHYURUS (MACROPODIDAE: MARSUPIALIA) (QUOY & GAIMARD 1830) IN THE NORTHERN JARRAH FOREST OF AUSTRALIA. 251 Introduction...... 252 Methods ...... 253 General procedures...... 253 Survival estimates...... 255 Modelling the causes of mortality...... 257 Results ...... 260 Overall survivorship...... 260 Comparative period of survivorship (March 1999 – May 2000)...... 264 Generalised models of the features that exacerbate the chance of mortality....265 The impact of mortality on each population ...... 269 Discussion ...... 269 Survivorship ...... 269 Mortality events...... 270 Other factors implicated in quokka mortality...... 275 The future…...... 277 Conclusions ...... 282 MODELLING THE OCCURRENCE OF THE QUOKKA (SETONIX BRACHYURUS QUOY & GAIMARD) (MACROPODIDAE: MARSUPIALIA) IN AUSTRALIA. 283 Introduction...... 284 Methods ...... 285 Determining quokka presence...... 285 Ecological variables investigated ...... 287 Modelling ...... 291 Results ...... 294 Principal component factor analyses...... 294 Comparison between sites with quokka populations present or absent...... 295 Generalised linear and additive modelling...... 296 Discussion ...... 301 Conclusions ...... 305 CONCLUSION 307 Distribution and status...... 307 Population sizes and dynamics...... 308 Reproduction...... 309 Home range and movements...... 309 Habitat use...... 310 Survivorship and mortality...... 311 Modelling quokka presence or absence ...... 311 The prospects for quokkas at each individual site...... 312 Chandler...... 312 Hadfield...... 312 Hoffman...... 313 Holyoake...... 313 Kesners...... 313 Rosella Road ...... 314 Victor Road ...... 315 Wild Pig...... 315 The extinction risk for the quokka ...... 316 Other threatening processes ...... 320 The value of fox control to quokkas...... 321 The future of fox control...... 323 Management recommendations...... 324 Do nothing approach ...... 324 Ex-situ management ...... 324 Individual site management ...... 325 Concentrate valuable resources on sustainable populations...... 326 Summation ...... 328 REFERENCES...... 329. APPENDICES...... 376. APPENDIX A – TRAPPING DATA...... 377 APPENDIX B – HOME RANGE OVERLAP OF QUOKKAS AT EACH SITE. 379 APPENDIX C - HABITAT UNITS AND THEIR DESCRIPTIVE VARIABLES. 383 APPENDIX D - FACTOR SCORES FOR HABITAT DESCRIPTIONS 396.

List of Tables

Table 1. Details of the location, size and baiting history of quokka swamps investigated

during this study. The AGD84 MapDatum was used......

Table 2. Locality records of quokkas in fossil deposits. Table 3. Locality records of quokkas reported in the literature prior to 1950

Table 4. Locality records of quokkas reported in the literature between 1950 and 1990.

Table 5. Locality records of quokkas reported in the literature since 1990

Table 6. Area of quokka distribution during each time period...... 56 Table 7. Percentage change in quokka distribution between periods ...... 56 Table 8. Details of the location, size and baiting history of quokka swamps investigated during this study...... 107 Table 9. Trap nights at each study site. Seasonal primary trapping sessions with number of trap nights in each secondary trapping session are shown. Number of days trapped shown in brackets...... 112 Table 10. Population estimates and capture success of adult quokkas at sites in the northern jarrah forest calculated for the entire study period. The high variability of quokka captures meant presenting seasonal capture data was irrelevant...... 117 Table 11. Sex ratios of all newly captured quokkas recorded at sites in the northern jarrah forest. Only sites with more than one individual KTBA used...... 119 Table 12. Regression analyses between the population indices (captures per 100 trap nights/known to be alive (KTBA)) and estimates (Cormack-Jolly-Seber (C-J-S)) as well as the length of each habitat patch...... 121 Table 13. Mean body mass of quokkas from sites in the northern jarrah forest...... 122 Table 14. Mean overall and core home range sizes (ha) derived by each of the three estimation methods. Individuals with 30 or more fixes (MCP) and 40 or more fixes (harmonic mean and kernel) were used for overall, male, female and site estimates...... 161 Table 15. Statistical test results on home ranges estimated by the kernel method. Male – female comparisons use data from home ranges calculated with more than 40 fixes, while the remainder use data from home ranges with more than 10 fixes. Small sample size precluded doing one large analysis...... 162 Table 16. Results of linear regression of home range size estimated by the kernel method against site population density and body mass...... 163 Table 17. Analysis of variance results for comparisons between the various age classes of the Agonis swamp shrubland habitat types. The individual comparisons using Scheffe’s test are shown in Table 18...... 196 Table 18. Identification of the factors used to differentiate between the Agonis swamp shrubland habitat units. The principal component factor is presented along with the Scheffe’s test probability in brackets. The principal component factors used were structural (SF1, SF2, SF3, SF4), floristic (FF1, FF2) and other habitat variables (OF1, OF2) and the factor scores of each of these is presented in Appendix F. A cross (X) signifies there was no direct differentiation possible between habitat units and so plots of two factor scores were subsequently used (Figure 45)...... 197 Table 19. Brief description of habitat types found at five quokka sites in the northern jarrah forest. Full descriptions are detailed in Chapter 6...... 215 Table 20. Analysis of variance of broad habitat use (swamp or non-swamp), based on area covered by home range, with sex and site nested within...... 224 Table 21. ANOVA table of broad habitat use based on area covered by home range nested with a diurnal / nocturnal time factor...... 224 Table 22. ANOVA table of broad habitat use of quokkas with season nested within. Repeated measures ANOVA was not conducted due to the small sample size of continually monitored ...... 225 Table 23. Analysis of variance results of broad habitat use. Sex and site factors are nested...... 226 Table 24. Variables used in the models of features affecting individual quokka survival. The name, description and range of each is listed. An asterisk identifies variables that showed partial correlations that were subsequently excluded from the analysis...... 257 Table 25. Mortality causes of the 58 collared quokkas with their location site and timing in the northern jarrah forest...... 260 Table 26. Log-rank and normal tests of survivorship. Total number of animals collared in each comparison is shown in brackets. The Bonferroni correction was applied to the comparisons between sites yielding a significant probability of 0.008...... 262 Table 27. Model selection statistics for dead/alive quokkas. The generalised linear models (GLM) are numbered sequentially from the null model with the Akaike Information Criterion (AIC), the delta (D) AIC, the number of variables in the model (K) and the Akaike weights. These statistics are also presented for the generalised additive model (GAM) of the best GLM. Models with a lower AIC are deemed best fitting the data (Burnham and Anderson 2001). Models with a D AIC of less than two (bolded) have substantial support while those with a D AIC greater than seven have no support (Burnham and Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and is essentially the probability that the model is the Kullback-Leibler best model for the data (Burnham and Anderson 2001). Although using a forward and backward stepwise selection procedure, the model did not add any variables once they were removed...... 266 Table 28. Variation explained by the preferred GLM and the GAM of that model. Lower residual deviances indicate a better model. The D2 statistic indicates the amount of variation in the data explained by the model and is analogous to the R2 of regression...... 268 Table 29. Analysis of deviance table of the preferred generalised additive model (GAM 6). A probability of less than 0.05 indicates that the non-parametric smooth spline curve of the GAM fits the data significantly better than a linear fit...... 268 Table 30. A comparison between the proportion of the overall quokka population found at each site and the proportion of the total quokka deaths at each site...... 269 Table 31. The name and description of ecological variables used to model quokka presence or absence as well as the range of values found for each and their unit of measurement. The description of each variable also presents the reasons why this variable was included in the analysis. Those variables that were correlated with others and excluded from analyses are marked with an asterisk...... 288 Table 32. Orthogonal factor scores for all the major habitat types. The variation explained by each variable is shown also. The variables that are important in describing each factor score are bolded. The overall factor analysis was significant (Bartlett’s c2 = 34.462; d.f. = 54; P = 0.982). For later analyses these factors are denoted by NJF1-5...... 295 Table 33. Comparison of the mean (± s.e.) raw values for each variable for sites (66) with quokka populations present or absent. Variables marked with an asterisk were excluded from the models after they were too highly correlated with others. The ‘aspect’ variable was categorised into north-east (1), south-east (2), south-west (3) and north-west (4) for analyses but is presented here as raw data. Only the factor scores for the NJF2 and NJF4 factors are presented in the habitat type variable. The presence and absence of quokka populations was compared using t- tests with the Bonferroni correction factor making the significance probability 0.003 (degrees of freedom is 64 for each)...... 296 Table 34. Model selection statistics for the presence/absence of quokka populations data set. The generalised linear models (GLM) are numbered sequentially from the null model with the Akaike Information Criterion (AIC), the delta (D) AIC, the number of variables in the model (K) and the Akaike weights. These statistics are also presented for the generalised additive model (GAM) of the best GLM. Models with a lower AIC are deemed best fitting the data (Burnham and Anderson 2001) (i.e. Model 11). Models with a D AIC of less than two have substantial support while those with a D AIC greater than seven have no support (Burnham and Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and is essentially the probability that the model is the Kullback-Leibler best model for the data (Burnham and Anderson 2001). Although using a forward and backward stepwise selection procedure, the model added no variables once they were removed...... 298 Table 35. Variation explained by the preferred GLM and the GAM of that model. Lower residual deviances indicate a better model. The D2 statistic indicates the amount of variation in the data explained by the model and is analogous to the R2 of regression...... 300 Table 36. Analysis of deviance table of the generalised additive model (GAM). A probability of less than 0.05 indicates that the non-parametric smooth spline curve of the GAM fits the data significantly better than a linear fit...... 301 Table 37. Attributes thought to correlate with increased extinction risk in mammals and a discussion on whether they apply to the quokka. The predictive attributes come from Purvis et al. (2000) and the references therein...... 317 Table 38. Factor scores for the structural factors and the variance they explain

Table 39. Factor scores for the floristic factors and the variance they explain

Table 40. Factor scores for the other habitat factors and the variance they explain.

List of Figures

Acknowledgements

Most importantly I would like to thank my parents. Their love, encouragement and support provided me with the foundation from which to leap. Without the knowledge that, if life in WA turned pear-shaped (as it almost did), I had a safety blanket waiting for me in , I may not have left a great job and girlfriend to chase a dream. To both my parents I am also honoured to be able to call you my friends.

To Dad, whose pride in my achievements are echoed by my pride is his and my desire to emulate them, I thank you. You are an ideal role model but only the second best kayak paddler in the team.

To Mum, who at times views her life as being one without achievement, you are wrong. Your support for me, your teaching of me, and your instilling in me the merits of positive thinking and the ability to communicate have all acted to shape me into the person I am. I chose well. I may be bestowed with a cap, a gown and a piece of paper but no such triumph would have been possible without you. It is with Dad’s support that

I dedicate this thesis to you.

To Nan, as a third parent throughout my life you are irreplaceable. Thank you for all manner of things that you have done to make me the person I am.

To Mif, your friendship throughout my life is one of its joys. Thank you for that.

To Anna – whose love for me was greater than her need for me and who selflessly sent me to Western Australia with her blessing - I thank you. The pain of our separation has scarred me but you challenged me to be a better person and you may have succeeded.

My desire to chase a life-long dream meant I discarded your love. I am so sorry – I wish it could have been otherwise.

To Beth, whose friendship and companionship made Dwellingup much more fun – I thank you. I’m glad you were there to relate to me my comments while anaesthetised during my elbow relocation.

Paul de Tores was good enough to look outside Western Australia for a student to undertake this project. This project may not have worked without your assistance.

Looking back, we had our moments – most of which were great, and I am extremely glad to have worked with you and learnt from you. Thank you for all this and for remaining my friend throughout the trials and tribulations of thesis writing. I look forward to seeing

Foxglove published and being able to say I was there.

Mick Dillon’s knowledge of the flora and fauna of the northern jarrah forest, and particularly quokkas, constantly amazed me. This document seems to confirm the majority of his theories of the ecology of the quokka. I loved stopping by his office to discuss my plans, my observations, my findings and my theories almost as much as I enjoyed his rough red and ‘ding’ sausages. I could have done without the winter through which he cajoled me into wearing shorts only to reveal at the end of it that his diabetes precluded him from feeling his legs. Try staying out of hospital and enjoy your retirement, you old bastard.

Mike Augee was amazing in his ability not only to remember me from three years earlier but to also track me down to a small country town in south-western New South

Wales and then offer me the project. His departure from the UNSW teaching staff is a

sad loss however the whistled version of God Save the Queen that heralds his return to our corridors is always welcome.

Barry Fox took over my supervision at UNSW after Mike retired and was faultless.

His subsequent retirement and his loss to the UNSW as an ecologist will be tremendous.

Peter Banks was the last to be lumped with me as his student and fortunately wasn’t forced into retirement by my questioning. His knowledge, time availability and easy- going nature made his assistance readily sought and easily received.

I have been educated at, or played sport for, the UNSW for over one-third of my life and I have enormously enjoyed the experience.

The ferals of Operation Foxglove (Beth MacArthur, Jim Cocking, Kathy Himbeck,

Marika Maxwell and Elisabeth White) made my field work in Dwellingup extremely enjoyable. I learnt a lot from all of you. Rob Brazell also assisted with discussions of quokka ecology and logistical support.

To the Fox lab – Jenny Taylor, Karen Ross and Liz Jefferys - thank you so much for your assistance and support during the analysis and write-up stages. Thanks also to Liz for spending a holiday in WA wading through the uncatalogued material in the

Palaeontology section of the Museum of Western Australia. The Friday pub crew also deserves thanks for keeping me sane (to a degree) and entertained.

Lisa Wright worked tirelessly to find all available references pertaining to quokkas and I thank her sincerely for putting up with my almost endless queries.

To the researchers and staff of CALM’s Wildlife Research Centre at Woodvale, thank you for your assistance, friendship and for making me do the cross-cut sawing at the Big Brook relay.

Thanks to the people of Dwellingup for making me feel so welcome during my stay.

In particular I would like to thank the Dillons, Mick and Tracy Tink, Ashy and Sue

Ashcroft, Pat and Lee Steinbacher, the late Taryn Linning, Thommo2, Westy, Rog,

Macca, Hum, Rowdy, Don, Sam, Nads, Johnny Maranta, Bill Beach, Jimmy Finnegan,

Rossco and Jim.

On a more specific note, this project was funded by CALM and when funding was threatened with withdrawal it was guaranteed by the then Executive Director of CALM,

Dr Syd Shea. Funding for the fox control baiting around the known quokka sites was provided by Alcoa World Alumina Australia Ltd. I was funded by an Australian

Postgraduate Award, by work at UNSW and by personal savings. The project was approved by CALMScience Ethics Committee Approval Number CAEC 1/97 and subsequent renewals. Trapping permits were issued by CALM (Licence number

SF002928).

Quokka sites investigated throughout the south-west were identified by Doug Giles,

Bruce Withnell, Ian Wilson, Greg Freebury, Peter Batt and Jim Shugg of CALM district offices.

The Museum of Western Australia provided recent records of quokkas deposited into their collections and allowed Liz Jefferys to search through the uncatalogued material.

Antoinette Tomkinson’s work setting up this study and informing me of background information on the quokka, showing me sites and identifying hardships is acknowledged.

Chantel Ward provided advice on using S-Plus and stepwise GAM and Chris Taggart and Norm MacKenzie provided advice on predictive modelling.

Joan Hayward, Phil Hayward and Eve Sheppard also assisted with data input.

The following people provided valuable criticism and comments on various chapters:

Joan Hayward; Peter Banks, Barry Fox, Paul de Tores, Jenny Taylor, Elizabeth Jefferys,

Mike Augee, Karen Ross, Ben Russell, Anna Beth MacArthur,

The assistance of the following people in the field is also gratefully appreciated.

Most importantly I’d like to thank all the Sydney-siders who travelled to Dwellingup to assist, in particularly, my family – Mum, Dad, Nan and Mif; Dougy Kimber for joining

Dad and I on a great trapping week (a highlight); Wazza Saunders; Shakin’ Stephens;

Nick Valentine and Anna Sherriff.

Name Days Hours Phil Hayward 11 88.5 Doug Kimber 8 70.5 Anna Sherriff 5 67.5 Beth MacArthur 7 42 Bruce Withnell 8 40 Megan Narducci 5 40 Shannon Freeman 5 33.6 Marika Maxwell 5 28 Darren Harvey 5 24 Kathy Himbeck 5 20 Miffy Hayward 3 20 Darren Stephen 3 20 Jim Cocking 4 18.5 Chris Gilbert 3 18 Warren Saunders 2 17 Kate Hassall 2 17 Mary Hayward 2 16 Nick Valentine 3 15 Kate Hassall 2 14 Joan Hayward 2 13 Prue Trollope 1 11 Alex Vella 1 11 Ron Rengger 1 11

Name Days Hours Dave Ferrier 1 8 M. Clapp 1 8 Winston Kay 1 5.5

Finally, I’d like to thank the staff of Lean and Hayward Pty Ltd for putting up with regular questioning from Dad regarding survey techniques.

Chapter 1 Introduction

Introduction

The quokka (Setonix brachyurus) (Quoy and Gaimard 1830) is a small, macropodid marsupial with a head to body length of about half a metre and an average weight of 3.6 kilograms in males and 2.9 kilograms in females (Kitchener 1995) (Figure 1). It is the sole member of its . Despite looking very similar to other , the dentition, skull structure, chromosomes and blood proteins of the quokka exclude it from the genus

Macropus within the Macropodidae (Strahan 1995). These characteristics do not link it with any other species of macropod, although relationships with (Thomas 1888;

Richardson and McDermid 1978; Baverstock et al. 1989) as well as Thylogale (Sharman

1954; Flannery 1989), (Kirsch et al. 1995), Petrogale (Kitchener 1995) and

Lagorchestes (Burk and Springer 2000) have been suggested.

Figure 1. Photograph of a quokka.

1 Chapter 1 Introduction

The quokka was the second species of marsupial to be recorded in Australia (Kitchener

1995). The first written descriptions of the species came from Dutchman Samuel

Volckertzoon in 1658 who described it as “a wild cat resembling a civet-cat but with browner hair” (Kitchener 1995). When his countryman, Willem de Vlamingh landed on an island off on the 30th of December 1696, he described its mammalian inhabitant as “a kind of rat as big as a common cat” and named the island Rottenest (now spelt Rottnest) - meaning ‘rat nest’ (Clarke 1947; Kitchener 1995). The quokka remains a tourist attraction on Rottnest Island today (Figure 2).

Figure 2. A quokka on Rottnest Island.

2 Chapter 1 Introduction

Names recorded for the species by the Nyoongar Aboriginal people of the south-west of Australia include ‘bangcup’, ‘bungeup’ and ‘ban-gup’ (Shortridge 1909; Gould 1973).

However the common name for the species is attributed to the Aboriginal people around

Albany after a medical officer from the King Georges Sound penal settlement in 1829 interpreted the name for the species as ‘quakur’ (Glauert 1950).

The ease in which the quokka can be studied on Rottnest Island has put it at the forefront of marsupial research studies in Australia. Detailed work on the Rottnest quokkas has been conducted at the University of Western Australia under the supervision of

Professors H. Waring and A. R. Main (Sharman 1954; Sharman 1955a, 1955b; Storr 1961;

Holsworth 1964; Storr 1964a, 1964b, 1965b; Holsworth 1967; Kitchener 1972). The knowledge developed from this research has become the foundation of almost all modern work on Australian (Ride 1970). The depth of knowledge on the island population and the lack of knowledge and scarcity of quokka populations on the mainland has often led to the belief that the quokka does not occur on mainland Australia (Barker et al. 1957).

Quokkas were abundant throughout the south-west of Australia for at least 33,000 years (Lundelius 1957). Records of its distribution on the mainland since the European colonisation of Australia are rare although at the time of settlement it was recorded as common (Kitchener 1995). Quokkas were thought to have occurred from the Moore River north of Perth to Esperance on the south coast (Shortridge 1909) where they were recorded them as “Still abundant in suitable swampy localities” on the mainland (Glauert 1932-33).

Quokkas also occurred on Rottnest and Bald Islands.

3 Chapter 1 Introduction

A drastic decline in many critical weight range mammals occurred on the south- western Australian mainland in the late 1930s (Burbidge and McKenzie 1989; Sinclair and

Morris 1996) to such an extent that quokkas were thought to have become extinct (Barker et al. 1957). Anecdotes of multiple dead quokkas in swamps linked this decline with disease (How et al. 1987) although this period of decline corresponded with the establishment of the red fox (Vulpes vulpes) in the south-west (King and Smith 1985).

Quokkas were rediscovered on the mainland in the late 1950s (Barker et al. 1957).

Accounts of their abundance from the 1970s indicated that they were common in forested areas where they occurred in dense vegetation associated with drainage lines (M. Dillon, pers. comm.), as well as similarly dense, coastal heath vegetation. Accounts from the coast indicate that quokkas move outside these areas of dense vegetation (Christensen et al.

1985). Today, the quokka is widespread in the southern forest but only locally common

(Christensen et al. 1985).

Quokkas were thought to have declined in range by at least 50% over the past 50 years

(Maxwell et al. 1996). In 1996, were known from only two of the twelve previously known locations in state forest near Dwellingup while two of the 31 known populations around Manjimup were known to be extinct (Maxwell et al. 1996). The population size at these sites was unknown but anecdotal evidence and results from ad-hoc trapping the

Western Australian Department of Conservation and Land Management (CALM) indicated that the population size may be declining (P. de Tores pers. comm.). Quokkas were still abundant on Rottnest Island.

4 Chapter 1 Introduction

On account of the discussion above, the quokka was listed as vulnerable in the Action

Plan for Australian Marsupials and Monotremes (Maxwell et al. 1996) as well as Western

Australian and Commonwealth legislation. The reasons given for this were that it was thought to have declined in abundance on the mainland and is now restricted to an area of less than 20,000 square kilometres ((Maxwell et al. 1996). It is also listed as vulnerable by the International Union for the Conservation Nature and Natural Resources (Hilton-Taylor

2000). Historical anecdotes suggested this decline was caused by predation by the introduced red fox (Vulpes vulpes), habitat modification and potentially by altered fire regimes (Maxwell et al. 1996). There has been no scientific study to ascertain the accuracy of these anecdotes.

Aims and scope of the thesis

While a great deal of knowledge has been acquired on the Rottnest Island quokka population in the past, there are vast differences between there and the mainland. Main

(1959) stated: ‘Environmentally, the situation on Rottnest Island is completely unlike the densely vegetated swamps now occupied by only a remnant of a formerly widespread population on the mainland’. Rottnest Island is more arid causing quokkas to suffer seasonal mortality, it lacks widespread permanent freshwater and lacks mammalian predators (Shield 1959; Storr et al. 1959; Barker 1961; Holsworth 1964; Storr 1964b;

Packer 1965; Kitchener 1970) and is considered atypical habitat for the quokka (Storr

1963). The extent of quokka occurrence on the mainland is enormous compared to the size of Rottnest and Bald Islands. Additionally, both island populations are susceptible to

5 Chapter 1 Introduction

catastrophic, stochastic events that may threaten their long-term survival. Consequently, knowledge of the general ecology of the quokka on the mainland and, in particular, the reasons for the apparent decline of the species are urgently required to ensure that the quokka does not become another in a long list of Australian fauna either extinct or restricted to offshore islands.

During an assessment of the decline of modern Western Australian mammals, the quokka was placed in the fifth level of priority for conservation initiatives (Burbidge and

McKenzie 1989). This priority recommended monitoring of the species. Such monitoring occurred in the past but was somewhat haphazard occurring on a district by district basis.

This thesis combines the accumulated data from this previous research as well as adding substantial detail to it.

The objectives of this study are to:

· accurately document the historic decline of the quokka and objectively assess its

current conservation status;

· estimate the sizes of known populations of the quokka in the northern jarrah

forest of the Australian mainland;

· determine habitat use and home range of the quokka by investigating in detail

sites in the northern jarrah forest;

· develop a predictive model of the occurrence of quokkas within the northern

jarrah forest;

· document the reproductive biology of wild quokkas on the mainland; and

6 Chapter 1 Introduction

· determine factors influencing the decline of the quokka on the mainland.

Structure of this thesis

This thesis is largely structured as a series of ‘stand-alone’ chapters addressing the aims of the research. Each chapter has its own introduction, methods, results and discussion section and each has been, or will be, submitted to a journal for publication.

7 Chapter 2 Methods and study sites

Methods and study sites

The majority of methods used in this study are described in the chapter that they pertain to. Consequently, this chapter describes the important environmental features of the study region and the study sites.

Study region

Location, topography and geomorphology

The ecology of the quokka was investigated in the northern jarrah forest. This comprises of the northern half of the Jarrah Forest biogeographical zone (Thackway and

Cresswell 1995) and is known as the Darling sub-district (Beard 1980). The jarrah forest covers an area of 46,078 square km (Thackway and Cresswell 1995) and the

Darling sub-district of this covers 18,970 square km. Both areas are dominated by the indigenous jarrah tree ( marginata) (Wallace 1966).

The majority of the jarrah forest occupies the undulating plateau of the Darling

Range (Figure 1) where it extends from just north of Perth to the karri (E. diversicolor) forest over 300 km to the south. The Darling Range varies in width from 30 to almost

60 km and it is bounded to the east by the cleared areas of the agricultural wheatbelt and to the west by the Darling Scarp (Figure 1). The Darling Range is an ancient erosional surface with an average height of around 300 m above sea level (Strelein 1988) but with the highest peaks exceeding 400 m.

8 Chapter 2 Methods and study sites

Figure 1. Map of the northern jarrah forest (shaded grey) showing major towns, the approximate position of the Darling Scarp (grey line running north – south) and the forest block boundaries. The Darling Range largely extends to the eastern edge of the forested region.

Geology and soils

The Darling Scarp is the surface expression of the Darling Fault (Figure 1) that marks the western edge of the Yilgarn Block, a plateau of stable, Archaen crystalline rocks where linear belts of metamorphosed sedimentary and volcanic rocks are invaded by large areas of granite (Biggs, Wilde and Leech 1980). The fault reached its major vertical movement at the - boundary and was associated with the break up of the Gondwana supercontinent (Biggs et al. 1980). Cainozoic laterite covers the majority of the Darling Range and consists of a ferruginous or aluminous layer

9 Chapter 2 Methods and study sites

generally two metres thick overlying a pallid, kaolinitic zone (Biggs et al. 1980). The

Collie Basin consists of and younger sedimentary deposits (Churchward and

McArthur 1980).

The soils of the region are largely of granitic origin with laterites dominating those of the ridges and slopes on the Darling Range (Churchward and McArthur 1980). The soils in the shallow valleys that are typically inhabited by quokkas consist of sandy gravels on the slopes of the minor valleys that fall to orange earths on the swampy floors (Churchward and McArthur 1980).

Drainage

Three major catchments drain the northern jarrah forest. The southern parts of the

Swan River catchment – the Canning, Dale and Avon Rivers – drain the northern section of the jarrah forest (Figure 2). The catchments that flow into the Peel Inlet near

Mandurah drain the central section of the northern jarrah forest (Figure 2). These waterways include the Serpentine, Dandalup and the Murray Rivers and this later river has its headwaters in the Hotham and Williams Rivers in the wheatbelt. The Collie,

Brunswick and Harvey Rivers drain the western side of the Darling Range to the south of the northern jarrah forest while the Arthur River flows along the eastern edge of the region (Figure 2).

10 Chapter 2 Methods and study sites

Figure 2. Map of the northern jarrah forest study region showing major waterways

(black lines), highways (grey-filled lines), important roads (grey lines) and towns (filled circles).

Vegetation

While jarrah lends its name to the region through its general dominance as a canopy species, marri (Corymbia. calophylla) often occurs as a co-dominant species throughout the forest. Western Australian blackbutt (E. patens) and bullich (E. megacarpa) occur

11 Chapter 2 Methods and study sites

on damp soils along smaller watercourses while flooded gum (E. rudis) dominates the vegetation along larger tributaries (Wallace 1966). Dieback induced by the fungus

Phytophthora cinnamoni affects large areas of the jarrah forest (RDAG/DWG 2000), however forest on the moister soils associated with the shallow, upper valleys are generally unaffected.

The Dwellingup and Hester complexes supporting jarrah – marri open, forest occur in the higher rainfall areas before giving way to the Yalanbee complex of wandoo (E. wandoo) to the drier, east (Heddle, Loneragan and Havel 1980). In the shallow, upper valleys typically inhabited by quokkas, these vegetation complexes give way to the

Yarragil, Pindalup and Catterick complexes that are characterised by bullich and blackbutt as well as various shrub species (Heddle et al. 1980) of which the most notable is Agonis linearifolia.

Anthropogenic disturbances

Originally the jarrah forest was distributed over three million hectares (Strelein

1988), however the remaining one and a half million hectares was dedicated as state under the Forests Act 1918-1954 (Wallace 1966). Jarrah is the principal timber tree of Western Australia (Harris 1966) and so has been extensively logged. Logging commenced in the northern jarrah forest in the 1870s and by 1989 approximately half of the jarrah and wandoo forest of the region had been logged twice, with some areas logged up to five times (Heberle 1997). The state forests are managed for many purposes in addition to timber production including water quality protection, conservation, recreation, road construction and mining (McArthur and Mulcahy 1980).

12 Chapter 2 Methods and study sites

Land clearing in the region began in the 19th century and continued until the designation of the majority of the region as state forest and was concentrated on the alluvial soils along the valleys of the Darling Range (McArthur and Mulcahy 1980).

This clearing is generally restricted to areas near townships. The water supply for Perth is drawn from surface catchments along the Darling Range (McArthur and Mulcahy

1980) and the construction of these dams caused the loss of habitat along many larger watercourses of the region. Other disturbances in the northern jarrah forest include mining for bauxite around Waroona, Dwellingup and Jarrahdale, coal around Collie and gold near Boddington. Apiary and firewood collection are also conducted in the forest and infrastructure associated with these, and other anthropogenic disturbances, occur throughout the region.

Climate

The climate of the south-west of Australia is typically Mediterranean with a well- defined pattern of winter rainfall and summer drought (Wallace 1966). The receives less than 1,000 mm of precipitation annually but this increases dramatically on the western side of the forest before declining again to the east (Figure

3). The rapid increase in rainfall at the western edge of the forest (Figure 3) arises through the influence of the Darling Scarp (Figure 4). The Darling Scarp (the western edge of the Darling Range) creates orographic rainfall which results in much higher levels of precipitation at its peak compared to adjacent areas.

Average temperatures in the northern jarrah forest range from about 30oC in the summer down to 10oC in winter (Figure 4). There is also a longitudinal cline with higher temperatures in the north than the south.

13 Chapter 2 Methods and study sites

Monthly average maximum temperatures at Dwellingup (Figure 4) range from

29.7oC in January to 14.9oC in July, while average monthly minimum temperatures peak in February at 14.8oC and are lowest in August with 5.5oC (Bureau of

Meteorology). Mean annual rainfall is 1,200 mm in the north (between Mundaring and

Dwellingup), 1,265 mm at Dwellingup and 1,179 mm near the Victor Road site (Figure

4). The majority of this rain falls in winter and the least falls over summer (e.g. 693 mm in winter and 60 mm in summer at Dwellingup).

The three years of field work associated with this study were conducted during a period of almost average rainfall in the northern jarrah forest. The first year, 1998, experienced one mm less than the 1,265 mm average at Dwellingup, 1999 experienced

58 mm above average and 2000 experienced 41 mm below average (Bureau of

Meteorology).

14 Chapter 2 Methods and study sites

Figure 3. Average annual rainfall isohyets for the study region. Isohyets range from 1,300 mm (unbroken line) in a small region along the western side of the Darling

Scarp near Dwellingup down to 600 mm to the east of the region. Isohyets were supplied by CALM’s Information Services Branch.

15

Figure 4. Map showing climatic information for various towns and sites throughout the northern jarrah forest. Towns are represented with a circle and study sites with a star. The location of the Darling Scarp is shown with the grey line running approximately north – south. Climate data sourced from the Bureau of Meteorology.

17 Chapter 2 Methods and study sites

In contrast to the areas inhabited by quokkas on the mainland, Rottnest Island is exceptionally arid. It too has a Mediterranean climate but with very short, wet winters and extremely dry summers (Figure 6) (Hesp et al. 1983). Mean annual rainfall is 736 mm, of which 93% occurs between April and October, however annual evaporation level is 1,500 mm (Hesp et al. 1983).

Why this region?

Local populations in this area have declined markedly and the quokkas’ distribution has contracted to the south since the 1930s (de Tores et al. In prep.). The general region abuts the two largest cities in Western Australia (Perth and Bunbury) (Figure 3) which suggests a high potential for disturbance and environmental degradation. Known anthropogenic disturbances in the jarrah forest include altered fire regimes, encroachment by , timber harvesting and mining (Thackway and Cresswell

1995). The area was selected for this study as it is at the northern limit of extant quokka distribution.

Study sites

Ten sites within the northern jarrah forest were initially investigated for quokka presence (Figures 4 and 5). All of these lie in the upper reaches of creek systems between 300 and 400 m above sea level and are vegetated by Agonis swamp shrublands.

Nine of these sites supported quokkas in the early 1990s (de Tores et al. In prep.) and there was an unconfirmed report of a quokka sighting at the Hoffman site in the late

18 Chapter 2 Methods and study sites

1990s (A. Danks pers. comm.). The Albany Highway and Gervasse sites were eventually excluded from the project due to feasibility constraints arising from the additional costs incurred for travel and the additional time required to trap these sites that are at the northern and southern extremeties of the study region respectively. The

Gervasse site was trapped regularly as part of a CALM monitoring program and is thought to have a relatively high population density when compared with other northern jarrah forest populations (de Tores et al. In prep.). The abundance of quokka faecal pellets within their characteristic runways at the Albany Highway site compared to other trapped sites indicated a moderate population density. The sites that were trapped as part of this study are discussed below and aerial photographs of each site are presented at the end of this chapter.

Chandler Road

The Chandler Road quokka swamp is situated seven km north of Jarrahdale via

Nettleton Road and Alcoa’s Jarrahdale minesite. Access to Chandler Road occurs to the north of Haul Road No. 1 (Figure 6).

This site forms part of an extensive system of swamp but evidence of quokka presence exists only in the western section, immediately downstream of the Chandler

Dam (M. Dillon & D. Giles pers. comm.). The swamp generally runs from west to east before joining the more major watercourse that flows to the north (Figure 6).

The most significant disturbance to the Chandler site was the construction of a dam to supply water to the bauxite mine in the 1980s. This dam flooded a large portion of the swamp and fragmented the small, upstream section (Figure 6). Following the

19 Chapter 2 Methods and study sites

construction of the dam, mining has occurred to within twenty metres of the swamp in places (Figure 6).

Despite the presence of a nationally threatened species immediately downstream of the dam wall, it was deconstructed in autumn 2000. This resulted in flooding and significant sedimentation of the swamp. As trapping had ceased at this site, there is no indication as to the impact of this event however the absence of footprints in the mud to the west of Chandler Road does not augur well.

Hadfield

The Hadfield site occurs to the south of the northern jarrah forest (Figure 2). It lies approximately 13 km south east of Harvey and 26 km north west of Collie. Access to the site is via the Mornington Road where it crosses the powerline easement (Figure 7).

The watercourse flows to the north-east and eventually reaches the Brunswick River.

While the upper reaches of the Hadfield swamp are relatively undisturbed, the lower section has been fragmented in the past (Figure 7). A powerline easement was created and the associated access track cut an approximately 20 m wide swathe through the swamp. A gas pipeline was added to the easement in 1999 and this widened the disturbed area by approximately 20 m (Figure 7).

A large block of land to the east of the Mornington Road was cleared for pasture

(Figure 7). A house was built on this block in 2000 and a dam was also constructed following swamp vegetation clearance and the subsequent flooding affected a small portion of undisturbed swamp. A fire in 2000 that started on the cleared, private property following the construction of the house escaped and burnt a triangular section

20 Chapter 2 Methods and study sites

of swamp and jarrah forest between the swamp, the powerline easement and the cleared area.

Hoffman

The Hoffman site is situated to the east of Harvey and is accessed by Stromlo Road and Black Form via the Harvey-Quindanning and Tallanalla Roads. This site is in the unbaited control region of Operation Foxglove (Figure 8) and was the site of an unconfirmed quokka observation (A. Danks pers. comm.). This site forms the upper reaches of an extensive swamp system which flows to the east although only minor areas of Agonis linearifolia swamp habitat are dispersed along its length. Forestry, dieback and fire have recently affected the Hoffman site to the east of Stromlo Road.

Holyoake

The Holyoake site had a well-studied quokka population extant until the early 1990s

(M. Dillon pers. comm.) but trapping in the mid-1990s suggested it had since gone extinct (de Tores et al. In prep.). The site is located one kilometre to the east of

Dwellingup and virtually adjacent to the cleared orchards and residential areas of the

Holyoake village. Access to the Holyoake site occurs via Old or New Duncan’s Roads from Dwellingup. The watercourse flows to the south (Figure 9). Being in such proximity to human habitation, this site is not baited to control introduced predators.

The sites’ proximity to human habitation means that it is quite disturbed. The upper reaches of the swamp system have been cleared for agriculture and a dam was constructed in the middle reaches several decades ago (Figure 9). The upper reaches of

21 Chapter 2 Methods and study sites

this site are infested with blackberry (Rubus affinus). Town and cats are also likely to be frequent visitors to the site.

Kesners

The Kesners quokka site is located 10 km north-east of Dwellingup (Figure 2) and is accessed via the Del Park and North Spur Roads. Kesners and Teddy Road either cut or skirt the edge of the swamp (Figure 10). This site is within the Operation Foxglove seasonal baiting treatment (Figure 5). The swamp initially flows west before heading north. A population of quokkas was known to exist at the Kesners site in the 1970s and

1980s but was considered extinct in the early 1990s (M. Dillon pers. comm.). Sign in

1998 suggested a population had recolonised the site.

Kesners is adjacent to Alcoa’s Del Park mine and has disturbances resulting from that. A large conveyor belt runs past the upper-most section of swamp resulting in a large cleared area (Figure 10). This and the noise associated with its operation may inhibit dispersal of quokkas across the ridgetop (Figure 10). Several areas adjacent to the swamp have also been mined. The North Spur Road, a sealed and busy road, cuts a small tributary of the swamp system – fragmenting the habitat (Figure 10).

Rosella Road

The Rosella Road site exists in the former minesite of Alcoa’s Jarrahdale bauxite mine, approximately 12.5 km north of Jarrahdale and halfway between there and

Armadale (Figure 2). Like the Chandler site, the Rosella Road site is situated within the twice-yearly baiting treatment of Operation Foxglove (Figure 5). It is accessed via

22 Chapter 2 Methods and study sites

tracks to the east of Nettleton Road and then along the rehabilitated Haul Road No. 1

(Figure 11). The watercourse flows to the north.

The Rosella Road site has had disturbances associated with mining. The haul road split the swamp in two and actual mining activity has occurred to within twenty metres of the swamp vegetation in several places (Figure 11). Within one kilometre to the south-west lie areas cleared for orchards.

Victor Road

The Victor Road site was the only site with a known quokka population that remained unbaited (Figure 5). It was discovered following a road kill in 1997 (R.

Brazell pers. comm.). The site is situated 16.5 km from Collie and 24 km from Harvey

(Figure 2) and is accessed from Victor Road via the Mornington Road to the north of

Gastaldo Road to the east. The watercourse flows to the west (Figure 12).

Victor Road cuts the swamp near its centre and land has been cleared in the upper reaches of the swamp (Figure 12) and now supports a single-mens accomodation area for one of the Collie mines. A control burn occurred to the west of Victor Road in 1997

(R. Brazell pers. comm.).

Wild Pig

The Wild Pig or Lewis site had a quokka population in the early and mid-1990s however, despite having extensive evidence of quokka activity, trapping in 1998 failed to yield any animals (de Tores et al. In prep.). The site is approximately 15 km north of

Dwellingup and is accessed to the north of Del Park Road via Scarp Road and on to

23 Chapter 2 Methods and study sites

Whittaker’s and North Roads (Figure 5). The Wild Pig site is within the seasonal (4) baiting treatment of Operation Foxglove (Figure 5). The watercourse flows to the west before turning north-west (Figure 13).

The Wild Pig site has been substantially disturbed with from the operations of

Alcoa’s Huntly bauxite mine. Bauxite extraction has occurred to the very edge of the swamp vegetation and the haul road now obliterates the upper reaches of the swamp

(Figure 13). Vegetation clearance in 1998 suggested further mining activities along the length of the swamp were planned.

Feral predator control

Sodium monofluoroacetate (1080) is highly toxic to most , birds and mammals (King 1993). Having evolved in association with the poison-bearing of the genus Gastrolobium, many Western Australian native fauna species, including the quokka, have a relatively high tolerance to the toxin sodium monofluoroacetate (1080), particularly when compared with (King, Oliver and Mead 1981;

Mead et al. 1985; Eason and Frampton 1991; King 1993; Thompson and Fleming 1994;

Short et al. 1997; Saunders, Kay and McLeod 1999; Twigg et al. 2000).

Since 1994, CALM has conducted feral predator control programmes within the south-west including parts of the northern jarrah forest under the auspices of Operation

Foxglove and Western Shield. The northern jarrah forest is subdivided into four baiting regimes or treatments (Figure 5) where broad-scale, aerial fox poisoning occurs at five baits km-2 at frequencies of either two, four or six times per year (de Tores 1994, 1999;

Thomson and Algar 2000). The baits consist of fresh meat injected with 4.5 mg of 1080 which is then dried in a forced-air, drying chamber at 350C for 4 days.

24 Chapter 2 Methods and study sites

Figure 5. Map of the northern jarrah forest showing the Operation Foxglove baiting treatments and the annual frequency at which baits are delivered. Towns are represented by black circles and the location of the study sites are shown in grey

(CALM Information Services Branch).

The local area around the Rosella Road, Chandler Road, Wild Pig Swamp, Kesners,

Hadfield, Albany Highway and Gervasse sites is baited monthly along vehicle tracks.

Baits are delivered from vehicles at a baiting interval of 100 metres. This higher intensity is due to the known occurrence f extant quokka populations in the mid-1990s and the anticipated response of the quokka to a reduction in predation pressure following that of other predation-sensitive species (Kinnear, Onus and Bromilow 1988;

Friend 1990; Friend and Thomas 1994; Kinnear, Onus and Sumner 1998; de Tores

1999; Morris 2000). The remaining three sites were unbaited. Two of these, the

Holyoake and Hoffman sites, were trapped in an attempt to confirm quokka presence

25 Chapter 2 Methods and study sites

and to establish another unbaited control population in addition to Victor Road. The location, length and fox control details of the eight sites trapped are discussed in Table

1.

Table 1. Details of the location, size and baiting history of quokka swamps investigated during this study. The AGD84 MapDatum was used.

Site Latitude Longitude Length Area Monthly fox (S) (E) (m) (ha) control start

Albany Highway 32o14’56” 116o08’57” 750 3.8 March 1998 1

Chandler Road 32o18’24” 116o07’20” 1,340 11.3 March 1998 1

Gervasse 33o22’09” 115o54’44” 750 7.4 October 1997 2

Hadfield 33o11’07” 115o58’25” 1,570 6.8 October 1997 2

Kesners 32o39’00” 116o00’59” 2,510 12.4 January 1999 2

Rosella Road 32o15’34” 116o04’36” 1,690 15.3 March 1998 1

Wild Pig 32o34’07” 116o03’03” 1,200 9.9 March 1999 2

Hoffman 33o02’12” 116o01’34” 2,000 8.0 Unbaited

Holyoake 32o42’14” 116o05’24” 500 2.6 Unbaited

Victor Road 33o16’10” 116o00’49” 930 8.2 Unbaited

1 Fox baiting occurs biannually since 1994 across wider, surrounding area as part of

CALM’s Western Shield/Operation Foxglove programme

2 Fox baiting occurs quarterly since 1994 across wider, surrounding area as part of

CALM’s Western Shield/Operation Foxglove programme

26

Figure 6. Aerial photograph of the Chandler site. The area of the swamp is outlined in white. Figure provided by Department of Land

Administration (DOLA).

27 Chapter 2 Methods and study sites

Figure 7. Aerial photograph of the Hadfield site. The area of the swamp is outlined in white. Figure provided by DOLA.

28

Figure 8. Aerial photograph of the Hoffman site. The area of the swamp is outlined in white. Recently burnt areas appear brownish.

Figure provided by DOLA.

29

Figure 9. Aerial photograph of the Holyoake site. The area of the swamp is outlined in white. Figure provided by DOLA.

30

31

Figure 10. Aerial photograph of the Kesners site. The area of the swamp is outlined in white. Figure provided by DOLA.

32 Chapter 2 Methods and study sites

Figure 11. Aerial photograph of the Rosella Road site. The area of the swamp is outlined in white. Figure provided by DOLA.

33

Figure 12. Aerial photograph of the Victor Road site. The area of the swamp is outlined in white. Figure provided by DOLA.

34

Figure 13. Aerial photograph of the Wild Pig site. The area of the swamp is outlined in white. Figure provided by DOLA.

35 Chapter 2 Methods and study sites

36 Chapter 3 Distribution and status

Distribution and status of the quokka, Setonix brachyurus

(Macropodidae: Marsupialia) and a discussion on the decline

of Australian native mammals.

Abstract. This chapter reports on the distribution and status of the quokka, Setonix brachyurus, in Australia. Literature searches, museum records and field surveys were used to plot the occurrence of the quokka during five periods. Two methods were used to quantify the area of quokka distribution, the minimum convex polygon method

(extent of occurrence) and a derivation of the minimum concave polygon method that excluded uninhabited areas (area of occupancy). Fossil sites indicate that quokkas prehistorically occupied an area of 49,000 km2 in the south-western corner of Australia.

By the time Europeans began their occupation of the continent, the quokka was widespread and abundant and was reported to occupy at least 39,000 km2. An extensive and dramatic population decline occurred in the 1930s. The decline continued and by

1990 had reduced to 28,000 km2 and by the year 2000 to around 20,000 km2. This represents a decline of 45% since 1950 and 29% since 1990. These results suggest that the conservation status of the quokka should remain as vulnerable when it is reassessed.

Predation by the introduced red fox, Vulpes vulpes, was probably the cause of the initial decline, while on-going predation, habitat destruction and modification through altered fire regimes led to its continuation. The decline of other critical weight range

Australian native mammals is discussed in light of the conclusions drawn from the decline of the quokka.

40 Chapter 3 Distribution and status

Introduction

The quokka, Setonix brachyurus (Quoy & Gaimard 1830), is a small macropodid marsupial that is endemic to the south-western corner of Australia. It has become synonymous with Rottnest Island through its appeal to tourists and through the pioneering work on macropods by Main (1959); Shield (1958; 1960; 1961; 1962; 1964;

1968); Sharman (1954; 1955a; 1955b); Holsworth (1964; 1967); Storr (1961; 1964a;

1964b; 1965b); Packer (1963; 1965; 1969; 1977); and Kitchener (1972; 1981) amongst others. Consequently few people are aware that the quokka also occurs on the mainland

(Barker et al. 1957; Kitchener 1995).

When Europeans began exploring Australia, quokkas were reported to extend from the Moore River (31°19’19”S, 115°28’34”E) in the north, south through the jarrah

(Eucalyptus marginata) and karri (E. diversicolor) forests to the south coast and east to

Twin Peaks Island near Esperance (34o07’40”S, 122o15’E) (Shortridge 1909; Glauert

1932-33; Troughton 1967; Kabay and Start 1976) (Figure 1). This distribution encompassed the Swan Coastal Plain, Jarrah Forest, Warren and potentially the

Esperance Plains biogeographical zones (Thackway and Cresswell 1995). As well as

Rottnest Island, quokkas also occur on near Albany (Storr 1965b). Today there is conjecture as to the most easterly occurrence of the quokka when Europeans were exploring the region with Shortridge (1909) stating that they occurred as far east as

Esperance in 1904 while Baynes et al. (1975) suggested that the Hunter River east of

Bremer Bay was their eastern limit.

Despite this previously widespread distribution, today quokka populations are highly scattered and discontinuous. Along the coastal areas of the mainland, the quokka is restricted to moist, dense heath and shrubland (Maxwell et al. 1996). In the jarrah

41 Chapter 3 Distribution and status

forest the quokka inhabits densely vegetated swamp shrublands along the broad and relatively flat, upper reaches of creek systems that are dominated by the shrub Agonis linearifolia (Christensen et al. 1985). Further south in the karri forest they are found in a broader range of habitats from similar densely vegetated, alluvial creeklines to the banks of larger rivers, while also extending up to the ridgetops (Christensen et al. 1985).

The quokka is listed as vulnerable on the Western Australian Wildlife Protection

Act 1950 and subsequent schedules therein. This follows such a listing in the Action

Plan for Marsupials and Monotremes that was based on a suspected reduction in abundance or distribution of at least 20% during the past ten years (Maxwell et al.

1996). This report also estimates the total population at less than 10,000 mature individuals with the population estimated as declining by at least 10% within ten years

(Maxwell et al. 1996).

In 1998 the quokka was thought to be endangered in Western Australia and threatened nationally (Stanger et al. 1998). In 1989 it was thought to have disappeared from more than 50% of its former range and was identified as a species with the highest priority for research (Johnson et al. 1989). The quokka is listed as vulnerable on the

Commonwealth’s newly enacted Environment Protection and Biodiversity

Conservation Act 1999. It is also listed internationally as vulnerable (Hilton-Taylor

2000).

The key recovery objectives for the quokka in the Action Plan for Marsupials and

Monotremes involved downlisting the species from vulnerable to lower risk

(conservation dependent) within five years (i.e. by 2001) (Maxwell et al. 1996).

Consequently, the purpose of this chapter is to identify the past and present distributions of the quokka and assess the status of the species in light of this information. Further,

42 Chapter 3 Distribution and status

this chapter aims to determine whether there are justifiable grounds for downlisting the status of the quokka as intended (Maxwell et al. 1996) and to identify and discuss the potential causes of the decline.

Finally over a decade ago several authors reviewed the modern mass extinction of medium-sized mammals in Australia since European arrival (Burbidge and McKenzie

1989; Flannery 1990a; Morton 1990a; Recher and Lim 1990). This chapter updates those reviews in light of recent studies and of the quokkas’ demise.

Methods

Locality records for quokkas were obtained from three sources: the available literature, records of the Western Australian Museum and field surveys. Records are split into five time periods: fossil records; pre-1950; 1950 – 1990; post-1990 and 2000.

There are various potential biases from this method. Fossilisation is likely to be biased toward the limestone areas along the coast and the dirth of fossil deposits in the granitic

Darling Scarp suggests this region is unfavourable to the process. This bias may have been exacerbated by the increased urbanisation of the coastal fringe of the south-west of

Australia. Survey reports prior to the 1930s have a positive reporting bias, but the decline of the quokka saw negative reporting become increasingly common.

Geographically, there may be some bias toward coastal regions and the wheatbelt to the east of the forest through the increased development of these regions and therefore the increased likelihood of quokka discovery compared to the jarrah forest. Consequently, although these time periods are of different duration, the increasing human population size (as observers and mortality causes of quokkas) and specific field surveys meant that the survey effort during each is considered similar.

43 Chapter 3 Distribution and status

Records from the Western Australian Museum were obtained from a published catalogue of records up until 1981 (Kitchener and Vicker 1981), from a search of their database for subsequent records and from manual searches of the uncatalogued collection of the Palaeontology and Anthropology Sections of the museum (E.Jefferys pers. comm.). Records from the former two sources have not been tabulated but are shown in distribution figures. Field surveys, aimed at identifying new quokka localities and monitoring historical quokka sites, were conducted between September 1998 and

September 2000 with a wide-ranging survey conducted throughout the south-west from

30/7/2000 to 8/8/2000. Potential sites were identified from either the previous occurrences of quokkas or by the presence of the preferred habitat of the species

(Christensen et al. 1985; Maxwell et al. 1996). During field inspections, the presence of quokkas at a site was determined by recent road kills, trapping or by observing scats within the characteristic runways created in the densely vegetated areas that they inhabit

(Chapter 2) (Barker et al. 1957; Storr 1964a; Christensen et al. 1985; Sinclair and

Morris 1996). Differentiating quokka scats from those of other macropods found in the region (Macropus fuliginosus, M. irma, M. eugenii) is extremely imprecise unless they are found within these 15 cm high by 10 cm wide runways (M. Dillon pers. comm.; pers. obs.). The scats of other species that occur sympatrically with the quokka cannot be confused with those of the target species (pers. obs.).

Locations from each of these sources were plotted on geographic information system maps using MapInfo Version 5.5 computer program (1985-1999 MapInfo

Corporation Inc.). The total extent of quokka occurrence (Gaston 1991; Hilton-Taylor

2000) during each of the five time periods was calculated using the minimum convex polygon method (MCP) with the Ranges V computer program (Kenward and Hodder

1992). This method has some inherent problems. Although being objective, it does not

44 Chapter 3 Distribution and status

take into account barriers to a species’ range, such as water bodies or uninhabitable areas. Consequently, the creation of a concave polygon around the known locations and excluding areas known to be outside a species’ distribution may be a more precise measure of a species’ distribution. This area of occupancy (Gaston 1991; Hilton-Taylor

2000) estimate was conducted by excluding the area east of the forest (wheatbelt)

(Gould 1973) and large waterbodies (Figure 1).

The Information Services Branch of the Western Australian Department of

Conservation and Land Management (CALM) supplied annual rainfall isohyets. The

700 mm and 1000 mm annual rainfall isohyets are drawn on all figures.

Results

Fossil record

The prehistoric distribution of the quokka extended from Jurien Bay, south through the Leeuwin-Naturaliste region and east to Albany (Figure 1; Table 2). The distribution has been extended from the most easterly quokka fossil deposit to reflect the probable distribution when Europeans arrived. This is justified as there is no reason to believe that a range expansion came following European arrival because of the quokkas’ restricted habitat preferences (White 1952; Christensen and Kimber 1975; Christensen et al. 1985) and the apparent stability in the number and distribution of those habitats since then. This distribution extends outside the current 700 mm rainfall isohyet although the fossils may have been deposited in wetter periods prehistorically when quokkas may have extended their distribution. The lack of records in the eastern part of the quokkas range, and on the Darling Scarp, probably reflect unfavourable conditions for fossilisation and possibly less palaeontological survey effort.

45

Table 1. Locality records of quokkas in fossil deposits. Date Location Comment Source Prehistoric Mammoth Cave, near Margaret River Quokka fossils ‘exceedingly abundant’ (Glauert 1948) Prehistoric Lake, Museum and Brides Caves, Margaret Quokka fossils found. (Glauert 1948) River Prehistoric Devil’s Lair Cave, near Margaret River Quokka fossils found from 35,000 years before present (ybp) until (Dortch and Merrilees between 1,000 and 6,000 ybp although recent dating techniques have 1971; Baynes et al. 1975; increased this to 48,000 ybp. Balme et al. 1978; Turney et al. 2001) Prehistoric Nannup/Strong’s Cave, near Margaret River Quokka fossils ‘abundant’ (Cook 1960) Prehistoric Turner Brook, near Augusta Quokka fossils found. (Archer and Baynes 1972) Prehistoric Skull Cave, near Augusta Quokka fossils found. (Porter 1979) Prehistoric McIntyre Gully, Gingin Quokka fossils found. (Merrilees 1979) Prehistoric Rainbow Cave, near Margaret River Quokka fossils found. (Lilley 1993) Prehistoric , Vicki Dawn, Orchestra Shell Caves Quokka fossils found. 1 and a quarry near Wanneroo 30 km north of Perth Prehistoric Several sites near Albany Quokka fossils found. 1 Prehistoric Calgardup Cave Quokka fossils found. 1 Prehistoric Caves in the Windy Harbour region including Quokka fossils found. 1 near Fish Creek and in a rock-shelter on Mt. Chudalup south of Northcliffe. Prehistoric Caves near Augusta including Easter, Quokka fossils found. 1 Harley’s, and Deepdene. Prehistoric Cave near northern boundary of Nambung Quokka fossils found making this the most northerly record of quokka 1 National Park near Jurien Bay distribution. Prehistoric In Protomnodon Chamber of Crystal Cave and Quokka fossils found. 1 surrounds Prehistoric Donnelly River mouth, 3 km east in limestone Quokka fossils found. 1 cliff Prehistoric Caves in the Dongara- Moore River region Quokka fossils found. 1 west of Gingin including Echidna and House Caves Prehistoric Grant's Cave, Witchcliffe east of Margaret Quokka fossils found. 1 River Prehistoric Hastings Cave, Jurien Bay Quokka fossils found. 1

46

Date Location Comment Source Prehistoric Jewel and Labyrinth Caves, Augusta Quokka fossils found. 1 Prehistoric Mandalay Beach, south of Northcliffe Quokka fossils found. 1 Prehis toric Margaret River Quokka fossils found. 1 Prehistoric Various caves along McIntyre Gully, Gingin Quokka fossils found. 1 Prehistoric In blown out sand dunes to the south and west Quokka fossils found. 1 of Northcliffe Prehistoric Mill Cave, Ngilgi Cave, between Margaret Quokka fossils found. 1 River and Augusta Prehistoric Moondyne Cave and surrounds, Margaret Quokka fossils found. 1 River Prehistoric On coast south of Scott River, near Walpole Quokka fossils found. 1 Prehistoric Quarry Bay, Augusta Quokka fossils found. 1 Prehistoric Sand dune west of Skippy Rock, Augusta Quokka fossils found. 1 Prehistoric South Lakes Cave, Margaret River Quokka fossils found. 1 Prehistoric Various sites in the Two Peoples Bay Nature Quokka fossils found. 1 Reserve, east of Albany Prehistoric Walpole and surrounds including a cave near Quokka fossils found. 1 the Nornalup Inlet Prehistoric Yeagerup Dunes Quokka fossils found at east and west ends of the dune system. 1 Prehistoric William Bay National Park, Albany Bones found in fossilised soil in exposed petrified forest 1 Prehistoric Windy Harbour Cave Quokka fossils found. 1 Prehistoric Yanchep, Yanderup and Minnie’s Grotto Quokka fossils found. 1 Caves near Yanchep, 30 km north of Perth Prehistoric Caves at Shark Bay, north of Jurien No evidence found. (Glauert 1948) Prehistoric Caves at Esperance No evidence found. (Baynes 1985) Prehistoric Nullarbor caves, east of Esperance No evidence found. (Lundelius 1963; Baynes 1985) 1 E.Jefferys personal communication

47

Figure 1. Estimated original distribution of the quokka. Fossil sites catalogued in the WA Museum are shown as stars and paleontological/archaeological sites (E.Jefferys pers. comm.) shown as filled triangles. Filled area shows the area of occupation while the long dashed line depicts the extent of occurrence. Dotted line is the 700 mm and short dashed line is the 1000 mm annual rainfall isohyet. The Darling Range extends along the western edge of the 1000 mm rainfall isohyet.

48

Table 2. Locality records of quokkas reported in the literature prior to 1950 Date Location Comment Source 1658 Rottnest Island Second Australian native mamma l recorded by Europeans (Holsworth 1964) when Dutchman (in Glauert 1950) Samuel Volckersen described it as ‘a wild cat resembling a civet-cat, but with browner hair’ during the voyage of the Waekend Boey. 1696 Rottnest Island Willem de Vlamingh mistook the quokka for ‘a kind of rat as big as a common cat’. (in Glauert 1950) Post – Yanchep, north of Perth Skeleton with an iron spike penetrating the skull suggesting persistence in the area following (Merrilees 1965) European European arrival arrival Post – Scott’s River, Walpole Abundant in unfossilised surface deposits (Butler 1968) European arrival 1829 King Georges Sound penal Medical officer reports ‘Quakur’ present. (Glauert 1950) settlement, Albany 1830 King Georges Sound, Albany Scientifically described by Frenchmen Quoy and Gaimard during the Astrolabe expedition. (in Glauert 1950) 1837 Swan River settlement (now Kitchener (1978) cautions that this record may have been in the Darling Range to the east of (Glauert 1950) Perth) Perth. 1839-1843 Various ‘Abundant in all the swampy tracts which skirt nearly the whole of Western Australia at a (as reported to John short distance from the sea, and that at Augusta … it inhabits the thickets and is destroyed in Gould by his great numbers … by the natives, who, after firing the bush, place themselves in a clear space collector, John and spear them as they attempt to escape from the fire’. Not found east of the Darling Range. Gilbert in (Gould 1973)) 1866-1869 King Georges Sound, Albany Most common macropodid collected. (Glauert 1950) 1888 King Georges Sound, Perth In the collections of the British Museum of Natural History, London, having been collected by (Thomas 1888) and Augusta John Gilbert for John Gould. 1906 Albany Shortridge’s field notes state “Very plentiful around Albany, but not extending very far (Thomas 1906) inland. It seems to be far more coastal in its range than any of the other wallabies, not appearing to occur anywhere at a great distance from the sea; gregarious. Resembling M. eugenei (sic) in habits.” 1904-1907 Various Specimens collected during the Balston expedition from King River and Big Grove (Albany), (Shortridge 1909) Yallingup (Busselton), Burnside (Margaret River), and Rottnest and Bald Islands. Also recorded on Twin Peak Island off Esperance. Found to be very plentiful ‘among the coastal thickets and swamps of the South-West, not extending inland’. 1902-1910 Margaret River ‘very numerous in the coastal country’ but by 1933 were uncommon. (White 1952) 1919 Blackwood River, east of ‘Numerous and frequently seen’. (Perry 1971) Margaret River

49

Date Location Comment Source 1920 Margaret River While collecting for the US Museum of Natural History, Charles Hoy stated that the quokka, (Short and Calaby “while not rare, is a strictly nocturnal animal and so is seldom seen. It is found mostly in the 2001) swamps.” Hoy also stated the quokka “Frequents the swampy parts of the forest where the undergrowth is the thickest. Makes well defined runways.” 1920 Gingin Brook, Gingin Long standing residents reported quokkas living among the densely vegetated banks and on (Roe 1971) islands in the brook. Until 1926 Logue Brook, near Ya rloop ‘so numerous among the low tangled scrubs … that we were able to stampede them … and (White 1952) watch the Quokkas bouncing across the open tract into scrub on the other side’. 1929 Busselton and along the coast Observed ‘in numbers’. (White 1952) to Augusta 1931 Busselton ‘it was possible to find quokkas in numbers at any time in the low scrub between the coastal (White 1952) dunes and the Vasse estuary’ where they occurred ‘through the low lying scrub country from the coast to the Marri and Jarrah fringes’. Early 1930s Mundaring, in the Darling Quokkas were so numerous that they were classed as ‘vermin’ by the Forests Department (Stewart 1936) Range east of Perth through the damage caused by their feeding on pine seedlings within ‘100 chains’ (two km) of watercourses. 1933 Canal Rocks, north of Recorded as ‘particularly numerous’ but had declined in abundance four years later and had (White 1952) Margaret River disappeared within a few more years. Mid-1930s Bickley, in Darling Range east ‘vanished from its gully haunts in these hills’ where they were particularly plentiful in the (White 1952) of Perth early 1920s, ‘when their narrow and often covered runways went in all directions through the few chains of thick scrub bordering the streams, to which areas they confined themselves’. 1940s South Yunderup, near Present. (Hutchinson 1972) Mandurah 1947 Swan coastal plain around Rare. (Clarke 1947) Perth

50

Figure 2. Distribution of the quokka prior to 1950. Filled area shows the area of occupancy while the long dashed line shows the extent of occurrence including the Esperance record and the solid line shows this excluding the Esperance record. Dotted line is the 700 mm and short dashed line is the 1000 mm annual rainfall isohyet.

51

Table 3. Locality records of quokkas reported in the literature between 1950 and 1990. Date Location Comment Source 1950s Darling Range near Yarloop and Reports of ‘reappearance of the quokka'. (Serventy et al. south of Manjimup 1954) 1950s Darling Range However, records from the Western Australian Museum indicate that ‘the valleys of the Darling (Serventy et al. Ranges, in which the Quokka was once so abundant, have long been deserted by this interesting 1954) species’. 1950s Southern forest, around Found in several localities in the region and believed to occur eastwards to the Franklin River. (Serventy et al. Manjimup 1954) 1954 Toolbrunup, near Bluff Knoll Quokka skull found. (Sharman 1954) 1957 Mainland Newspaper article claims the quokka became extinct on the mainland during the early 1930s. (Hodgkin, E. P. West Australian 17/4/57 in Barker et al. 1957) 1957 Byford, Darling Range south-east Responding to the newspaper report (above) these authors conducted a survey which found (Barker et al. of Perth ‘abundant evidence of Quokka presence’ through ‘tracks, droppings and tunnels in the thick 1957) undergrowth’. Prior to this, the last mainland quokka specimen in the Western Australian Museum was obtained in 1935. 1957 Walpole and near Perth Concluded that surveys of suitable habitat in forested areas would yield more sites. (Barker et al. 1957) 1960 Karridale/Margaret River ‘almost certainly absent’. (Cook 1960) Mid-1960s Byford (south-east of Perth) and Persisted into the 1960s. (Storr 1964a) Bald Island populations 1965 Bald Island and Waychinicup “Quokkas were abundant on Bald Island” where they “occurred everywhere from sea-level to (Storr 1965b) Inlet the peak”. In the Waychinicup Inlet “The dense beds of sedges were intersected by a maze of runways littered with the characteristic faeces of the species”. 1970 Mainland Rare and only known with certainty from a ‘few swampy valleys in the Darling Range close to (Ride 1970) Perth’. 1971-1972 Dwellingup Six quokkas trapped in 12,738 trap-nights. (Schmidt 1973; Schmidt and Mason 1973) 1972 Albany Highway 27 miles from Road kill. (Crabb 1972) Perth 1972-1974 Within 20 km of Dwellingup Quokkas trapped. (Christensen and Kimber 1975) Mid 1970s Various Quokkas collected at Byford, Holyoake (Dwellingup), Muddy Lake (Bunbury), Mowen and (Hart et al. 1986)

52

Date Location Comment Source Stoats Road (Blackwood River, south of Nannup), Bald Island and Waychinicup Inlet (near Albany) during a survey of Salmonella infection in the species. 1977 Stirling Ranges, near Bluff Knoll Quokkas survived ‘in isolated and undisturbed valleys … where water seepage during the dry (Packer 1977) summer maintains fairly luxuriant vegetation’. 1977-1978 Swan coastal plain No evidence of quokkas despite being recorded there in 1933 (Glauert 1932-33). (Kitchener et al. 1978) 1978-1980 Mount Saddleback State Forest No evidence of quokkas. (Nichols and near Boddington Nichols 1984) 1983 Green Range, north of Albany Quokka carcass found. (Kirke 1983) Mid-1980s Dwellingup, Pemberton, Bald and Quokkas collected to determine their tolerance to sodium monofluoroacetate (1080). (Mead et al. Rottnest Islands 1985) Mid-1980s Point D’Entrecasteaux, near Found only here during a survey of fauna throughout the entire coastal area between Busselton (How et al. 1987) Busselton and Albany. Mid-1980s Between Busselton and Albany Survey included forested areas and found quokkas to be widespread, but only locally common. (Christensen et al. 1985) 1986-1988 Torndirrup National Park, Albany No evidence of quokkas. (Smith 1990) 1980s Bold Park, 11km west of Perth No evidence of quokkas although it was suggested they did once occur there. (How and Dell 1990) 1980s Lesmurdie, Darling Range west No evidence of quokkas. (Dell and How of Perth 1988) 1986 Murray-Serpentine River delta, No evidence of quokkas despite being recorded there in the 1940s (Hutchinson 1972). (Brown Cooper et near Mandurah al. 1989)

53

Figure 3. Distribution of the quokka between 1950 and 1990. Stars indicate records of occurrence. Filled area shows the area of occupancy while the solid line shows the extent of extent of quokka occurrence over this period. Dotted line is the 700 mm annual rainfall isohyet and short dashed line is the 1,000 mm annual rainfall isohyet.

54

Table 4. Locality records of quokkas reported in the literature since 1990. Date Location Comment Source 1994 Between Torndirrup National Surveys by the University of Western Australia found no evidence of quokkas. (in Sinclair and Park and Cape Riche, Albany Morris 1996) Mid-1990s Western Australia The ‘last decade has seen some recovery and it is now common in the moister parts of the (Kitchener 1995) extreme south-western part of the state’. Mid-1990s Alcoa mine site, Jarrahdale Trapped by Mick Dillon (pers. comm.) (MD). (Sinclair 1999) Mid-1990s Holyoake, Kesners and Wild Pig Trapped after site identification by MD. (Sinclair 1999) swamps, Dwellingup Mid-1990s Harvey (south of Yarloop) and Trapped after site identification by Rob Brazell (pers. comm.). (Sinclair 1999) Collie Mid-1990s Two Peoples Bay Nature Reserve, Trapped. (Sinclair 1999) Albany Mid-1990s Byford, Mundaring, Manjimup No evidence of quokkas despite previously occurring here. (Sinclair 1999) and Green Range Mid-1990s Stirling Ranges National Park, Road kill. (Sinclair 1999) near Bluff Knoll 1996 Dwellingup Found to be extinct at all but one (Wild Pig Swamp) of 12 previously known sites (MD). (Maxwell et al. 1996) 1996 Manjimup Extinct in at least two of the 31 sites previously identified around Manjimup (MD). (Maxwell et al. 1996) 1996 Mainland Only known to occur at 25 sites on the mainland. (Maxwell et al. 1996)

55

Figure 4. Distribution of the quokka since 1990. Stars indicate records of occurrence. Filled area shows the area of occupancy while the solid line shows extent of occurrence. Dotted line is the 700 mm annual rainfall isohyet and short dashed line is the 1000 mm annual rainfall isohyet.

56

Figure 5. Distribution of the quokka since 2000. Filled stars indicate records of occurrence and open circles show sites surveyed in 2000 without finding quokkas. Filled area shows the area of occupancy while the solid line shows extent of occurrence . The 700 mm (dotted) and 1000 mm (short dashed line) annual rainfall isohyets are shown.

57 Chapter 3 Distribution and status

Distribution and abundance: knowledge prior to 1950

Records of the distribution of the quokka prior to 1950 are listed in Table 2 and shown in Figure 2. This distribution is similar to its prehistoric distribution apart from the large (130 km) contraction to the south. Records extend from the Moore River in the north, south through the jarrah and karri forests, west to the Leeuwin-Naturaliste region between Augusta and Busselton and as far east as Bremer Bay or possibly Twin

Peaks Island, near Esperance (Figure 2). Conjecture exists over this last record as it significantly extends the historical distribution of the quokka compared to the present distribution (Figure 2).

No records are found to the east of the jarrah forest in the area now commonly known as the wheatbelt. This pattern of distribution appears to follow the 700 mm annual rainfall isohyet for coastal areas and the 1000 mm isohyet for the inland, forested areas (Figure 2).

Reports prior to 1950 illustrate that the species was once very abundant (Table 2).

The species declined massively and abruptly in the 1930s (White 1952).

Distribution and abundance: knowledge between 1950 and 1990

Literature records of the occurrence of the quokka between 1950 and 1990 are listed in Table 3. An interesting record during this period is from Breaksea Island near

Albany which extends the number of offshore islands inhabited by quokkas to three

(I.Abbott in Kitchener and Vicker 1981) (Figure 3).

Between 1950 and 1990 the distribution of the quokka continued its decline from the north, and probably also from the east (Figure 3). There are no records north of

Perth, despite having previously occurred there. There are also no records east of

59 Chapter 3 Distribution and status

Bremer Bay. The quokka appears to have been rare and scarce during this time (Table

3).

Distribution and abundance: knowledge since 1990

Records of quokka occurrence since 1990 are recorded in Table 4 and shown in

Figure 4. Since 1990 quokka distribution has continued to decline (Figure 6). There are no records from the coastal plain between Perth and Margaret River. The most northerly records from the Darling Range since 1990 are further south than before. The most easterly record is approximately 50km west of Bremer Bay near the Pallinup

River. Despite no records from the Leeuwin-Naturaliste region, it is likely that quokkas still occurred in the upper reaches of the Blackwood River, north-east of Augusta.

Trapping in 1999 of the one surviving population near Dwellingup found it to be extinct also, however another site containing quokkas was located. The recovery mentioned by

Kitchener (1995) (Table 4) is not apparent from the distribution of locality records

(Figure 4) or by comments on their abundance in the literature.

Distribution and abundance: knowledge from 2000

The field surveys conducted throughout the south-west in 2000 found evidence of quokka presence at 30 of the 97 potential quokka swamps investigated (Figure 5).

Fifty-five of these 97 sites were known to support quokka populations previously and five of these were found to be extinct. The surveys suggest the decline in quokka distribution has continued (Figure 6). The most easterly record is now from Bald Island east of Albany and the Stirling Range population around Bluff Knoll may be isolated.

The most northerly record is in a water catchment area beside the Albany Highway

60 Chapter 3 Distribution and status

approximately 20 kilometres south-east of the Perth metropolitan area. Locality records

in forested areas extend in a narrow band along the Darling Range and are almost

entirely within the 1,000 mm annual rainfall isohyet (Figure 5). Observations of

potential sites suggests that they rarely contain more than 20 individuals (Sinclair 1999).

Changes in the actual area of distribution

Prehistorically quokkas existed across an area of over 116,000 km2 of the south-

western corner of Australia (Table 5). Quokkas may have occurred over almost

150,000 km2 prior to 1950, however when the questionable eastern-most record (Twin

Peaks Island, near Esperance) is excluded this area is reduced to a more likely 93,000

km2 (Table 5). Between 1950 and 1990 this had declined to just under 80,000 km2 and

by end of the 1990s this had further reduced to 60,000 km2. Surveys in 2000 revealed

the decline has continued and the distribution of the quokka now covers less than

50,000 km2.

The more precise estimates of the area that the quokka occupied based on excluding

uninhabited areas revealed a range of almost 50,000 km2 prehistorically and 37,000 km2

before 1950 (39,000 km2 if the questionable Esperance record is included). This area

had declined to 28,000 km2 between 1950 and 1990 and to just over 20,000 km2 during

the 1990s. The decline appears to have continued in the year 2000 (Table 5; Figure 6).

Table 5. Area of quokka distribution during each time period. Distribution period Extent of occurrence (MCP) (km2) Area of occupancy (km2) Prehistoric (excluding Esperance) 116,200 49,030 Pre-1950 (including Esperance) 148,380 38,940 Pre-1950 (excluding Esperance) 92,890 36,620 1950-1990 79,280 28,460 Post 1990 60,930 20,240 2000 49,650 20,220

61 Chapter 3 Distribution and status

The percentage reduction in area of quokka distribution is shown in Table 6. The

distribution of the quokka in 2000 has declined by over 50% from the prehistoric

distribution. Over the past 50 years the extent of occupancy has declined by over 40%.

The decline over the past ten years is almost 40% according to the extent of occurrence

or almost 30% based on the area of occupancy.

Table 6. Percentage change in quokka distribution between periods Distribution period Change (Occurrence) Change (Occupancy) Prehistoric to today - 57.27% - 58.76% 50 year decline (pre-1950 to post-1990) - 58.94% - 48.07% 50 year decline (as above excluding Esperance) - 34.41% - 44.74% 10 year decline (1950-1990 to post 1990) - 23.15% - 28.89% 10 year decline (1950-1990 to 2000) - 37.37% - 28.94%

Figure 6. Map showing the decline of the quokka. The area of quokka occupancy prehistorically is shown as the dotted region, prior to 1950 is hatched, from 1950 until 1990 is light grey and in 2000 is shown as dark grey. 62 Chapter 3 Distribution and status

Discussion

Status

The results of this research validates the vulnerable status attributed to the quokka by Maxwell et al. (1996). A decline in quokka distribution of 29% over the past ten years and 45% over the past 50 years quantifies this. The decline in the distribution of the quokka has not abated and there appears to be no justification for downlisting its status from vulnerable to lower risk (conservation dependent) in 2001 as was aimed for in the recovery objectives (Maxwell et al. 1996).

Conservation areas supporting quokka populations today include the Stirling

Ranges, Walpole-Nornalup, D’Entrecasteaux, Sir James Mitchell and, possibly, the newly established Wellington Weir National Parks; Two Peoples Bay and Mount

Manypeaks Nature Reserves; as well as Rottnest and Bald Islands. The most northerly of these areas is the Wellington Weir National Park, near Collie. The majority of these conservation areas are in coastal habitats. Lane Poole Conservation Reserve, that extends from Dwellingup to Collie, is not thought to contain quokkas. There are no conservation areas supporting quokkas in the northern jarrah forest and only one in the southern jarrah forest. As this area is suffering large reductions in the area of quokka distribution it seems there is a need for the reservation of some existing quokka sites.

The most easterly existing locations of quokka occurrence appear to be adequately conserved in the two nature reserves and the Stirling Ranges National Park.

63 Chapter 3 Distribution and status

The pattern of decline

Quokkas are abundant in the fossil record dating back to the (Vickers-

Rich and Rich 1993) although Flannery (1984) suggests an earlier evolution. Fossils indicate that the quokka has always been restricted to the south-western corner of

Australia although the prehistoric distribution determined in this study (Figure 1) is far greater than that delineated by previous researchers (Shortridge 1909; Glauert 1932-33;

Kitchener 1995) due to the addition of previously unreported fossils. The mainland population was split by rising sea levels prehistorically with the Rottnest quokka population becoming isolated between 6,000 and 8,000 ybp (Churchill 1959) while the

Bald, and possibly Breaksea, Island populations were separated almost 10,000 ybp

(Storr 1965b).

Cave deposits to the east of the area of known quokka occupancy, including the

Esperance and Nullarbor caves, do not contain quokka fossils (Baynes 1985). These may always have been unsuitable for quokkas due to the aridity of the region, even in moister prehistoric times. The fossil records north of Yanchep similarly may have occurred during a range expansion associated with a wetter climate (Balme et al. 1978).

Pleistocene range expansions have been inferred for many species in worldwide (Hewitt

2000) including the quokka (Sinclair 2001).

Quokka fossils in the Devil’s Lair deposit have been dated to 35,000 ybp (Balme et al. 1978), although more recent techniques have increased this to 48,000 ybp (Turney et al. 2001). The fossil record also suggests that quokka abundance varied markedly throughout this time, from a peak 20,000 ybp to a reduction at 12,000 ybp and then a more recent increase (Balme et al. 1978). This change in abundance was attributed to a change in vegetation from shrubs to woodland and finally forest and was thought to

64 Chapter 3 Distribution and status

correspond to a time of increasing temperature and rainfall at the peak followed by a cooler, drier period (Balme et al. 1978). Despite these variations, the quokka was still one of the most abundant species throughout the Devil’s Lair deposit (Balme et al.

1978), however these conclusions may be biased to an extent as the deposit was thought to have been caused by predation by owls and humans (Dortch and Merrilees 1971).

Reading reports of quokka abundance prior to the mid-1930s provides stark contrast to reports thereafter and today (Table 2). Not only was the quokka widespread prior to the 1930s, it was particularly abundant in suitable habitat (Shortridge 1909; White 1952;

Perry 1971; Gould 1973). These comparative reports highlight the quokka’s dramatic decline.

Visual inspection of the distribution over the five time periods highlights the reduction in area occupied by quokkas (Figure 6). The decline has occurred predominately in the north and east, with expanding urban areas along the coast reducing records in the west. The existence of fossil records 130 km north of the northern-most quokka location attributed by early European explorers (Shortridge 1909;

Gould 1973) may be due to their lack of surveys in this region or to a decline following the Pleistocene (Sinclair 2001). The lack of records from around Esperance since 1950 is largely responsible for the approximately 50% decline over that period (Table 6).

The Twin Peaks Island population near Esperance is now considered erroneous as a more recent survey of Twin Peaks Island found tammar wallabies (Macropus eugenii) but not quokkas (Main and Yadav 1971). This is supported by the absence of quokka fossils in deposits from the Esperance Plains coastal region (Baynes 1985).

65 Chapter 3 Distribution and status

Variation in survey effort is unlikely to be responsible for this pattern of decline.

Additional survey effort today will undoubtedly result in the identification of new quokka populations, however the overall pattern is likely to remain.

The decline of the quokka distribution has occurred along the peripheral margins of its range. The decline in the quokka appears likely to continue to occur from the northern periphery at least, as the northern-most population trapped being made up of one lone male (Chapter 4). The next most northerly site has a population of 10 individuals (Chapter 4) and genetic data confirms this population is verging on extinction (Alacs 2001).

Intuitively, this is to be expected considering the poorer quality of preferred habitat at the periphery of a species' range (Gaston 1990) and the fact that populations are more highly fragmented and therefore less likely to be 'rescued' by nearby populations

(Brown and Kodric-Brown 1977). Yet this finding, contrasts starkly with the majority of declining species of which 98% of 245 declining species from around the world maintained remnant populations in some peripheral parts of their range, while 68% of these species were more common in the periphery than the core (Channell and

Lomolino 2000). Only 5 of the 245 species persisted solely in the core of their original distribution (Channell and Lomolino 2000). When this was broken down into continents Africa, with its long history of human evolution, showed no peripheral bias in range contraction compared to the other continents (Channell and Lomolino 2000).

This persistence along the periphery of the range was attributed to the over-riding influence of anthropogenic extinction forces (habitat alteration and introduced species) that render historical distribution patterns irrelevant (Channell and Lomolino 2000).

While it is agreed that populations that survive the longest are those last affected by

66 Chapter 3 Distribution and status

extinction forces (Channell and Lomolino 2000), this is not necessarily because of their geographic location along the edge of their range or presence at high elevations. In

Australia, the majority of declines have occurred in arid areas which are resource- depleted, while resource-rich mesic areas along the coast are thought to have been buffered from such declines (Burbidge and McKenzie 1989). According to the

(Channell and Lomolino 2000) definition, these resource-deficient areas are classed as core areas while the resource-rich mesic areas along the coast are considered as peripheral areas. This is simply not the case as species that have declined in Australia have persisted in areas that provide refuge from extinction forces by means of predator avoidance, nutrient provision or climatic extremes. Furthermore, these mesic areas have been much more severely affected by human occupation than the arid zone as the vast majority of the Australian population live along the coast. Consequently, we would expect the declines to have occurred in this mesic region rather than the arid zone

(Burbidge and McKenzie 1989).

Thus, while the geography of recent extinctions reflects human geography

(Channell and Lomolino 2000), it may as easily reflect the geography of available resources - an abundance of which buffers species from extinction. Perhaps a better way of investigating such phenomena would be to do so at the scale of individual biogeographic zones (Thackway and Cresswell 1995).

Factors influencing the decline

The question that now warrants investigating is why the quokka has declined and continues to do so. There are several factors in the biology of the quokka that apparently increase its risk of extinction. It is restricted to an island continent, which

67 Chapter 3 Distribution and status

affords a higher threat of extinction than species endemic to the other continental land masses (Mace and Balmford 2000). With its small geographic range, the quokka is also more likely to be driven to extinction by spatially-restricted threats (Pimm 1998) which is compounded by evolving in relative isolation from generalist predators (Mace and

Balmford 2000), particularly considering the speed at which predator-naivete is attained

(Berger et al. 2001).

Several studies have specifically assessed the causes of the widespread decline in

Australia’s mammalian fauna since European settlement (Burbidge and McKenzie

1989; Morton 1990a; Recher and Lim 1990; Smith and Quin 1996; Calver and Dell

1998a; Wilson and Friend 1999; Short and Calaby 2001; Short et al. 2002). Earlier studies on the quokka have discussed these factors. White (1952) stated that ‘their virtual disappearance on the mainland (is) ascribed to the spread of foxes, to competition with rabbits and to the destruction of habitat through clearing and bushfires, but many bushmen who knew the animal well consider that these factors were probably supplementary to a rapid decline resulting from disease’. Main (1959) suggested that the spread of urbanisation and agriculture contributed to the decline of the quokka.

These factors are discussed below.

A more widespread and long-term decline?

While the most marked decline of the quokka was recorded during the 1930s

(White 1952), a more widespread decline of many species, particularly in more arid areas, was recorded in the 1880s (Shortridge 1909). Burbidge & McKenzie (1989) attributed the modern decline of Western Australia’s terrestrial, critical weight range (35 g – 5.5 kg), mammalian fauna to environmental changes since European settlement that

68 Chapter 3 Distribution and status

have reduced available productivity by diverting resources to humans and introduced species and reduced vegetative cover through and altered fire regimes. Morton

(1990a) had similar views as to the diversion of resources in the arid zone.

Critical weight range mammals suffered the greatest declines due to their limited mobility and relatively high metabolic requirements, and the arid zone fauna were most at risk with the mesic zone fauna buffered (Burbidge and McKenzie 1989). That study reported a more recent disappearance of several critical weight range mammals from the mesic North Kimberley district and additional long-term research from the Northern

Territory suggests mesic areas are no longer secure (Woinarski et al. 2001). The mesic south-west Australian environment may have buffered the fauna from this initial decline, however it has been suggested that a new phase of decline, an artifact of the first, may now be upon us (Johnson et al. 1989; Recher and Lim 1990). Mesic areas may have more available environmental productivity that allows fauna to cope for longer with the various disturbances that caused the earlier decline in arid areas but today are approaching a similar diversion of environmental resources and a reduction in refugia leading to the a more modern decline. In essence this is the extinction debt

(Tilman et al. 1994). The decline of the quokka concurs with this theory, with their distribution shrinking in the more arid coastal areas to the north and west while the mesic forested areas have survived (Figure 6). Further evidence of this phenomena would arise if the increasingly mesic karri forest supports higher population densities than found in the slightly less mesic northern jarrah forest. The quokka may be further buffered from declines compared to other sympatric critical weight range mammals by its habitat preference of moist, densely vegetated swamps that offer refugia from predation, competition with introduced and fire in a similar manner to rock piles of more arid areas (Burbidge and McKenzie 1989).

69 Chapter 3 Distribution and status

The direct cause of the quokka decline may never be known for certain, however

Hoy in the 1920s (in Short and Calaby 2001) wrote that “the cat has already reached the most remote parts of the Australian continent and as the fox and rabbit are spreading fast, it is only a matter of time before they too overrun the place. When this happens its goodbye to the native fauna for in the desert country, it seems that the native fauna cannot make, even the feeble stand that it can in the more favorable types of country.”

This chapter illustrates the accuracy of this prediction.

Predation

Introduced species present especially serious problems to endemic island fauna, which have evolved for a long time in isolation (Mace and Balmford 2000). The arrival of humans to Australia has seen the introduction of three such predators; the dingo

(Canis lupus dingo), the European red fox (Vulpes vulpes) and the cat (Felis catus).

These eutherian predators are acting on a marsupial prey that has not interacted with medium-sized cursorial predators for the last 20,000 to 30,000 years (Johnson et al.

1989) and even then this indigenous predatory fauna was depauperate (Flannery 1991).

As such native animals, particularly medium to large species which, prior to the arrival of these species were not as threatened by birds of prey, , varanids or the more widespread dasurids, were likely to be predator-naïve (Burbidge and McKenzie 1989).

This is increasingly likely to be the case considering moose (Alces alces) lose their anti-predator response to grey (Canis lupus) within ten generations (Berger et al.

2001). Although such naivete is rapidly lost within one generation in prey species that have evolved with these predators (Berger et al. 2001), it may take substantially longer for effective anti-predatory responses to evolve against completely foreign predators

70 Chapter 3 Distribution and status

such as those now found in Australia. Furthermore, if adults and young are killed rapidly, learning opportunities are diminished and the threat of local extinction is increased (Berger et al. 2001). The most powerful illustration of the impact of naivete to danger is thought to come from the late Quaternary extinctions of more than half of the genera of large land mammals, primarily due to the initial effects brought about by colonising human hunters (Martin and Steadman 1999). Predator naivete is now considered so significant as to be added to Diamond’s (1984) ‘evil quartet’ of extinction causes (Berger et al. 2001).

Predation by the dingo

The dingo (Canis lupus dingo) has been in Australia for about 4,000 years (Corbett

1995a). It is an opportunistic feeder in Western Australia (Whitehouse 1977) with mammals of all sizes that are common to a particular area being preferential prey sources (Corbett 1995b; Vernes et al. 2001). It also switches between prey species in accordance with their availability or abundance (Robertshaw and Harden 1986; Corbett and Newsome 1987; Lundie-Jenkins et al. 1993). The quokka was common prior to the

1930s and so the dingo may have preyed upon it, although the preference of quokkas for densely-vegetated habitat that offered refuge from predation may have minimised this threat. Considering the presence of the dingo in the region prior to the decline of the quokka, there is no reason to believe dingo predation was the cause. Today, dingoes are exceedingly rare in the jarrah forest (pers. obs.; M.Dillon pers. comm.) and are unlikely to be generating any predation pressure on quokkas at all. The dingo was thought to have had little impact on the majority of Australian fauna with the exception of the (Thylacinus cynocephalus) and the ( harrisii)

71 Chapter 3 Distribution and status

(Hope 1984) although (Flannery and Roberts 1999) classify this loss as a second phase of extinction following, and indirectly attributable to, human arrival.

One reason Morton (1990a) discounted introduced predators as the cause of the decline of critical weight range mammals was the apparent lack of substantial changes following the arrival of the dingo. Its impact on Australian native fauna prior to

European arrival may have been minimised through competitive inhibition by a sympatric meso-predator (Soule 1988) - man (Homo sapiens). Similarly, an indirect effect of dingoes may be their own role as a meso-predator (Newsome 2001). Dingoes actually prey on the European red fox (Vulpes vulpes) and the where they occur in sympatry (Marsack and Campbell 1990; Lundie-Jenkins et al. 1993) and are likely to competitively inhibit these species, thereby forcing them to occur at lower densities than would otherwise have occurred. Such behaviour has been mentioned as a theoretical method of suppressing the impact of recently introduced predators (Lundie-Jenkins et al. 1993; Newsome 2001). Another reason for the lack of a substantial impact on native fauna following the arrival of the dingo may be the importance of larger prey species, in terms of biomass eaten (Corbett and Newsome 1987). Alternatively the climate at the time of the arrival of the dingo may have been in a wet period which averted prey suppression (Newsome et al. 1989).

Predation by the red fox

Increasingly, authors are beginning to attribute the widespread demise of

Australia’s terrestrial fauna to the introduction of the red fox (Short et al. 1992; Short and Smith 1994; Short and Turner 1994; Smith and Quin 1996; Short 1998; Short et al.

2002; Kinnear et al. In press). Recent suggestions that surplus killing by exotic

72 Chapter 3 Distribution and status

predators on a naïve and relatively scarce prey with low intrinsic rates of increase may have initiated the precipitous declines explain how this can occur (Short et al. 2002).

The red fox arrived in the south-west of Western Australia in the early 1930s (King and

Smith 1985). This corresponds almost exactly to the dramatic decline of the quokka

(White 1952) as well as numerous other terrestrial Australian native mammals within a critical weight range (Jenkins 1974; Burbidge and McKenzie 1989; Richards and Short

1996). Mr. W.H. Loaring linked the demise of the quokka population at Bickley, east of

Perth, after 1934 to the fox by stating ‘The rabbit began to appear at that time, and the fox followed, and they dwindled away’ (in White 1952).

The dramatic decline of other species in Australia also coincided with the arrival of the fox (Short 1998; Short and Calaby 2001; Short et al. 2002). Other species exhibiting such declines not mentioned in these studies include the bilby ( lagotis) (Jenkins 1974) and the or brush-tailed bettong (Bettongia penicillata)

(Christensen 1980). The woylie declined in the early 1930s to the mid-1940s

(Christensen 1980). This decline was more dramatic in area of distribution than the quokka probably due to the almost cosmopolitan distribution of the woylie across

Australia. Although this decline was apparently over a longer period than the decline of the quokka it does not refute the ‘blitzkrieg’ hypothesis caused by exotic predators

(Mosimann and Martin 1975; Martin and Steadman 1999). This hypothesis of the predatory behaviour of newly invading predators has been questioned following the persistence of threatened native fauna long after the arrival of introduced predators but this seems bound to occur, particularly when they are secondary prey that have access to refugia (Pech et al. 1995). This is something akin to the extinction debt or future ecological cost that arises from habitat destruction (Tilman et al. 1994). Following the initial dramatic decline population densities will be low (Beck 1996) and therefore

73 Chapter 3 Distribution and status

increasingly difficult for predators to capture. Predators may respond to this lower abundance with a functional shift in prey preference (Pech et al. 1995). The threatened species may compound these problems for the predator by being forced into predation refuges such as densely vegetated habitats (e.g. the quokka) or areas supporting poison- bearing plants (e.g. the woylie and at Perup Forest Block and Dryandra Forest

Nature Reserve) (Christensen 1980; Friend 1990). It seems intuitive therefore that a lag time before extinction will be observable and for quokkas this exists from the 1930s until today. This lag-time however, does not infer that the proximal cause of the trend toward extinction was not introduced predators. This may be exacerbated by other disturbance factors that become increasingly important at low population densities as part of the ‘evil quartet’ of extinction forces (Diamond 1984).

The abundance of otherwise threatened species on islands in the absence of introduced predators provides further evidence as to their impact (Short et al. 1992).

Quokkas are super-abundant on Rottnest Island, in the absence of the fox (Shield 1959;

Dunnet 1963; Holsworth 1967), but are threatened with extinction on the mainland.

The long-nosed (Potorous tridactylus) is rare on the mainland but common in

Tasmania (Heinsohn 1968).

The reintroduction of several threatened species to Shark Bay from their island refuge provides more evidence where, in the presence of predator control and exclusion, these threatened species can survive adequately and increase in abundance and range

(Risbey et al. 2000; Short and Turner 2000). Further evidence for the impact of foxes on quokkas comes from the failed reintroduction to the mainland of nearly 700 Rottnest

Island quokkas between 1972 and 1983 near Perth (Short et al. 1992). In the absence of effective predator control only nine quokkas remained in the 254 ha reserve by 1988

74 Chapter 3 Distribution and status

and predation by foxes and feral cats was attributed as the cause (Short et al. 1992).

Other species suffer similarly at the hands of the fox. A population of rufous hare- ( hirsutus) was eradicated by a single fox (Lundie-Jenkins et al.

1993) in a spree of surplus killing (Short et al. 2002). A reintroduction of 85 tammar wallabies (Macropus eugenii) released between 1971 and 1981 failed after they were all killed by introduced predators (Short et al. 2002). The (Macropus parma) is another species thought to have suffered through predation by the fox (Short et al. 1992). In fact a review of reintroductions in Australia reveals only 8% were successful when predators were present compared to an 82% success rate in their absence (Short et al. 1992).

Foxes are still seen as a major threat to quokka populations (Maxwell et al. 1996) yet it is only recently that they were experimentally linked to the suppression of native populations (Kinnear et al. 1988; Kinnear et al. 1998; Kinnear et al. In press).

Additionally, the increase in abundance of numerous other native fauna following fox control highlight the threat (e.g. the removal of the southern brown bandicoot (Isoodon obesulus) from threatened species lists in Western Australia; black-footed rock-wallaby

(Petrogale lateralis) (Kinnear et al. 1988; Kinnear et al. 1998); numbat (Myrmecobius fasciatus) (Friend 1990); the woylie (Christensen 1980; de Tores 1999); the chuditch

(Dasyurus geoffroyii) (Morris 1992); and the common brushtail (Trichosurus vulpecula) and western ringtail ( occidentalis) possums (de Tores 1999). A recent meta-analysis of the effect of the fox on Australian native fauna concluded a detrimental blanket impact (Kinnear et al. In press).

During poisoning of other species, such as rabbits (Oryctolagus cuniculus), foxes are susceptible to secondary poisoning if they scavenge carcasses (Algar and Kinnear

75 Chapter 3 Distribution and status

1996). Between 1952 and 1968 broad-scale poisoning of rabbits occurred in the south- west of Australia and it is likely that fox numbers were reduced throughout this period

(King and Smith 1985). Anecdotes from residents of forested areas of the south-west suggest that quokkas may have increased in abundance in the late 1960s and early 1970s

(M.Dillon pers. comm.).

The restriction of mainland quokkas today to thickly vegetated habitats (Christensen and Kimber 1975) may be due to the refuge from predation offered there (Shield 1968).

The wariness exhibited by mainland quokkas even after years of captivity and compared to their reported fearlessness and curiosity in the 1930s (White 1952) is thought to have been selected for following heavy predation pressure from the fox (Shield 1968).

As shown, the evidence linking the fox to the decline of critical weight range mammals, including the quokka, is damning. The fox arrived in the south-west in the early 1930s (King and Smith 1985) and within five years the quokka had noticeably declined in abundance and distribution. It seems very likely therefore, that foxes were responsible for the initial decline and assisted in continuing their demise.

Predation by the cat

Aborigines from the desert regions believe cats to have either always been present or that they arrived from the west (Burbidge et al. 1988) possibly following early

European shipwrecks (Gaynor 2000). More recent analysis indicates cats arrived with the European colonists and were unable to become independent from humans until native predators around the settlements were eradicated (Abbott 2002) in a form of mesopredator release (Courchamp et al. 1999). Once feral, cats slowly spread outwards reaching the limits of pastoral settlement by 1880 and then rapidly overtaking the rest of

76 Chapter 3 Distribution and status

the continent (Abbott 2002). By 1907, cats were distributed throughout Western

Australia (Shortridge 1936).

The arrival of the feral cat may explain the reported decline of numerous species in the 1880s (Shortridge 1909) either directly through predation or indirectly through the transmission of disease. There are no records in the literature of feral cats killing quokkas on Rottnest Island despite coexisting for decades (Main et al. 1959), however indirect evidence suggests it is possible and even likely. Feral cats have been recorded preying on more than 50 species of Australian native mammals including the young of bridled nailtail wallabies (Onychogalea fraenata) (Horsup and Evans 1993), adult allied rock-wallabies (Spencer 1990) and adult rufous hare-wallabies (Lagorchestes hirsutus)

(Gibson et al. 1994). Adult rock-wallabies are the largest of these and can attain 4.7 kg

(Eldridge and Close 1995). This is approximately the maximum weight of adult male quokkas (Chapter 4) (Shield 1968; Sinclair 1998) and so it seems probable that cats are also able to prey on quokkas.

Interspecific competition of cats by foxes or mesopredator function is thought to be the most likely mechanism limiting cat abundance (Molsher et al. 1999) and in situations where foxes have been controlled cat numbers increase (Risbey and Calver

1998; Risbey et al. 2000). Such a response may be occurring with the large-scale fox eradication programme being conducted in the south-west as part of the Western

Australian Department of Conservation and Land Management’s Western Shield initiative (Thomson and Algar 2000) and compensatory mortality on native fauna by cats may occur when fox density is reduced.

Although being implicated in the extinction of smaller mammals (< 1 kg) in the initial century of European occupation of Australia (Dickman 1993), (Abbott 2002)

77 Chapter 3 Distribution and status

considers these extinctions to be localised and suggests the cat probably caused the decline of some species that would have persisted in the absence of the fox. The lack of extinctions on , in the presence of feral cats but the absence of foxes, also suggests their impact was not the ultimate factor in the demise of most species (Abbott

2002). Furthermore, the lack of cat predation records from Rottnest Island (Main et al.

1959) suggests that cats did not initiate the decline of the quokka. Today the cat may be acting in conjunction with the fox in continuing the predation pressure (Newsome et al.

1989) thereby maintaining the decline of quokka populations. Nonetheless, the cat is not considered to be the driving factor behind the dramatic decline of the quokka considering its likely presence in the area well before the 1930s. The same conclusion has been drawn for other critical weight range mammals (Burbidge and McKenzie

1989).

Summary of the effect of recently introduced predators

The hunting success of these introduced predators seems intuitively to be lower in mesic areas that are more densely vegetated and therein provide more refuge than more open arid regions. This ecological feature may assist in explaining the delay in the decline of native mammals in mesic areas compared to the arid zone. These introduced placental predators may nowadays exist as part of an extinction vortex (Diamond 1989) on a native fauna that is already in low numbers such that several factors are now contributing significantly to drive them towards extinction (Figure 7). More significantly, these declines in mammals from previously buffered regions may be indicative of the collapse of the broader Australian ecosystem.

78 Chapter 3 Distribution and status

The prehistoric dirth of generalist predators (Flannery 1991) occupying the niche that the red fox now fills in Australia may make native fauna more susceptible to extinction as they lack the resilience that is thought to be attained after surviving similar selection pressure in the past which would have filtered out or eradicated predation susceptible species (Balmford 1996). Essentially, this can be conceptualised as a similar process to the development of antibiotic resistance in bacteria. The arrival of humans to Australia could have been an extinction filter but their likely initial preference for the megafauna suggests resilience to such selection pressures was not afforded to smaller species.

79

TIME LINE

Species decline following Extinction vortex of several factors a reduction in productivity (habitat alteration, overexploitation, through the diversion of introduced species, and chains of Arid zone --> resources to humans and extinction) operating in conjunction introduced species, and a to drive a species toward extinction reduction in vegetative – no buffering so the decline is cover by exotic herbivores immediate. and changed fire regimes.

Additional cumulative threats occur on a larger scale than Extinction before as the human population vortex Species remain stable as --> has increased along with their eventually Mesic --> more predictable associated disturbances (habitat initiated zone resources are available to clearance, fragmentation and after the them than in arid areas. altered fire regimes) while under decline was the constant threat of predation delayed and competition from introduced following species. initial Figure 7. Schematic representation of the decline of Australia’s critical weight range mammals (Burbidge and McKenzie 1989)buffering. including the quokka. Initial stability in pre-European times is followed by the arrival of Europeans which results in threats that initiate an extinction ‘vortex’ whereby several factors drive a species into decline (Diamond 1989). For the quokka, these factors are likely to be fox predation, human hunting and habitat alteration arising from changed fire regimes. The individual factor that ultimately forces a species into the precipitous decline of the vortex - where other factors in conjunction propagate the downward spiral – varies for each species but for the quokka is likely to have been the arrival of the European red fox. 80 Chapter 3 Distribution and status

Human predation

Of the longer standing predators in Australia, humans are likely to have been the most significant. There are many references to quokkas being hunted for food or sport up until the 1930s by both Aborigines and Europeans on the mainland (White 1952;

Dortch and Merrilees 1971; Perry 1971; Gould 1973; Baynes et al. 1975) and on

Rottnest Island (Shield 1959; Holsworth 1964). Aborigines used to set fire to the swamps and spear quokkas as they escaped the flames (Gardner 1957; Gould 1973).

Aboriginal hunting of the quokka is likely to have been substantially reduced following the various disease epidemics and conflicts resulting from European colonisation in the

1830s, well before the decline of the quokka.

While the Aboriginal population of Western Australia decreased, the European population increased substantially following colonisation (Figure 8). The population growth of the state over the period of quokka decline almost ceased and the increase in population over the decade preceding and following the decline was not as substantial as it has been since (Figure 8). The decline of the quokka occurred during the Great

Depression and, although evidence of increased hunting of quokkas to supplement the diet of people over this period was not found (Garden 1979), it is possible that human hunting increased.

81 Chapter 3 Distribution and status

1750

1500

1250

1000

750 Population size (000)

500

250

0 1829 1839 1849 1859 1869 1879 1889 1899 1909 1919 1929 1939 1949 1959 1969 1979 1989 1999 Year WA Perth Kalgoorlie/Boulder

Figure 8. Plot of the population of Western Australia (line), Perth (triangles) and the rural regional centre of Kalgoorlie-Boulder (circles) since European settlement (ABS Undated). The Kalgoorlie-Boulder data is presented to illustrate the likely proportion of the population inhabiting rural areas away from Perth. The gray dotted lines show the period over which the quokka noticeably declined and this did not coincide with a period of substantial population increase.

Compounding these threats, the quokka was classed as vermin and controlled by the

Forests Department near Mundaring in the early 1930s due to their impact on pine seedlings (Stewart 1936). Consequently, hunting by the expanding European population may have contributed to the decline of the quokka in the 1930s as part of an extinction vortex (Figure 7) however, considering the declines were ubiquitous throughout the south-west and not focused around towns, it seems unlikely to have been the proximate cause. Reinforcing this conclusion, Burbidge & McKenzie (1989) believed that no Australian terrestrial mammal had been driven to extinction through

82 Chapter 3 Distribution and status

overkill by Europeans and, although humans do still kill quokkas (AAP 1998), this is not likely to be a significant threat to the quokka.

The impact of the more recently arrived placental predators on a predator-naïve fauna can be extended to the arrival of humans (Homo sapiens) to Australia. A great deal of conjecture has arisen from assertions that humans were responsible for the extinction of Australia’s Pleistocene megafauna (Anderson 1990; Bowdler 1990;

Flannery 1990a, 1990b; Grayson 1990; Horton 1990; Martin 1990; Morton 1990b;

O'Connell 1990; Wright 1990; Eldridge 1999) by means of a ‘blitzkrieg’ (Mosimann and Martin 1975; Holdaway and Jacomb 2000), surplus killing (Short et al. 2002) or indirectly (Merrilees 1967; Jones 1969; Flannery 1990a). One of these issues of conjecture is that species of megafauna coexisted with humans for several thousand years (Hope 1984; Dodson et al. 1993). Many late-Quaternary extinctions however, have now been linked to either directly or indirectly to humans (Cassels 1984; Miller et al. 1999; Holdaway and Jacomb 2000).

The modern arrival of the European red fox into Australia can be paralleled with the earlier arrival of humans albeit with less indirect effects and the Recent extinctions may be a continuation or after-shock event resulting from earlier ones which may have been accentuated by the fox (Flannery and Roberts 1999). Again this concept leads back to the extinction debt of habitat loss and degradation (Tilman et al. 1994), but the concept seems increasingly likely to apply to all manner of ecological phenomenon.

The predator-naïve fauna were often devastated in arid areas (Burbidge et al. 1988) but often survived in the mesic areas either through additional resources available there

(Burbidge and McKenzie 1989) or refuge offered from predation (Short et al. 2002).

Native fauna species in mesic areas may coexist with foxes for years before extinction

83 Chapter 3 Distribution and status

occurs, in the absence of human intervention. It seems likely that this time frame will be lower today than occurred in prehistoric times as there are numerous other disturbances acting in unison. These include habitat loss through the expansion of the human population, changed fire regimes, increased densities of introduced predators following the urban sprawl, etc. A surplus killing, ‘blitzkrieg’ is not surprising for invading species that are uncertain as to when or where their next meal will come from and who are prepared to scavenge carcasses at a later date (e.g. foxes and humans). The

Quaternary megafaunal extinctions and those in Australia since Europeans arrived may have been limited to those species that were not fast enough to develop strategies to avoid or cope with the human hunting blitzkreig (Berger et al. 2001). The massive proportion of extinctions in the critical weight range (Burbidge and McKenzie 1989) may be due to their learning capacity which may be low due to their evolution on the nutrient-depauperate Australian continent (Flannery 1994).

Summary of the effects of exotic predators

Morton (1990a) listed several anomalies that he believed discounted introduced predators as the sole or ultimate cause of the extinction of native fauna. The first was that the extinctions only occurred in the arid zone. I would argue that this was due to the paucity of resources, and consequently refuges, available to arid zone fauna, which agrees with the arguments of Burbidge & McKenzie (1989) and Morton (1990a) himself. This may have buffered the fauna of the mesic areas during the initial predation onslaught which arose when predators attained higher densities than the long- term carrying capacity (Short et al. 2002). The fauna of these mesic areas may today be under the same threat (Recher and Lim 1990) as the increased resources there are either

84 Chapter 3 Distribution and status

being used up, destroyed through human disturbance or increasingly utilised by introduced predators.

The second anomaly discounting introduced predators was that small mammals, birds and reptiles escaped the havoc wrought by introduced predators (Morton 1990a).

Small mammals possibly had intrinsic rates of increase that were sufficiently high to cope with this increased depredation. The amount of energy and nutrients provided by small mammals and reptiles may make preferentially preying upon them inefficient although small mammals are preferential prey sources in Europe (Henry 1986). Yet in

Europe medium-sized mammals have evolved strategies to cope with such depredation, which I would argue makes targeting smaller species a more optimal foraging strategy there. Additionally, birds, reptiles and small mammals have the ability to use refuges from medium-sized terrestrial predators that larger prey species often do not (such as flying away or burrowing). Reptiles also have the ability to remain inactive for long periods of time. Finally, these smaller prey species are unlikely to have been as predator-naïve as larger species due to the increased richness and diversity of predators on them (e.g. the majority of birds of prey (Barker and Vestjens 1989/90), varanids, snakes (Cogger 1992), (Dasyurus sp.) (Strahan 1995) depredating small vertebrates compared to only the largest raptors (Barker and Vestjens 1989/90), pythons

(D. Pearson pers. comm.) and tiger (Dasyurus maculatus) (Edgar and Belcher

1995) depredating medium-sized mammals) and hence they evolved efficient anti- predator strategies. Based on the above discussion, I would conclude that introduced predators, especially the red fox, had a massive and detrimental effect on quokkas.

85 Chapter 3 Distribution and status

Predation by birds

Nocturnal birds of prey are reported to have collected the fossils in the Turner

Brook deposit (Archer and Baynes 1972). The owls that occur in the south-west of

Australia are the barking owl, Ninox connivens, and the masked owl, Tyto novaehollandiae. These species are reported to take prey up to the size of rabbits

(Schodde and Tidemann 1997) which weigh up to 1.6 kg (Myers 1995). This is approximately the size of juvenile and small adult female quokkas (Shield and Woolley

1961). The presence of quokka bones below a wedge-tail eagle’s eyrie on Bald Island

(Storr 1965b) indicates an additional diurnal predation threat. The restriction of quokkas to densely vegetated swamps (White 1952; Storr 1964a; Christensen et al.

1985) during the day, however minimises the threat from eagles. Considering this, at least modern, restriction to refugia, predation by birds is not likely to have been the cause of the decline of the quokka in the 1930s as there is unlikely to have been an increase in predation pressure at this time.

Predation by reptiles

On Garden Island, off Perth, a 2.7 kg carpet python () ate a 2.8 kg (D. Pearson pers. comm.). Other medium-sized mammals eaten by carpet pythons include adult , brushtail possums, and bilbies (D.

Pearson pers. comm.). Considering female pythons on Garden Island grow up to six kilograms (D. Pearson pers. comm.), all but the largest quokkas are likely to be at risk.

The species found in the south-west (Varanus rosenbergii, V. gouldii and V. tristis) are not considered large enough to be a threat to medium-sized mammals

(Cogger 1992). As they have all been present in the south west of Australia throughout

86 Chapter 3 Distribution and status

the quokkas existence and there is no reason to believe they initiated the decline in quokka abundance.

Competition with introduced and endemic species

Despite the suggestions of White (1952), the European rabbit (Oryctolagus cuniculus) has not been implicated as a competitor of the quokka by other authors. The ecology of the rabbit is very different to that of the quokka, thereby minimising the likelihood of competition. The rabbit is generally a grazer of green grass and herbage but during drought it may switch to feeding on bark, leaves and roots of shrubs (Myers

1995). The quokka on the other hand is a generalist browser on Rottnest Island and is likely to have a similar diet on the mainland (Storr 1964b). The rabbit seeks refuge in burrows or warrens (Myers 1995) while there are no records of the quokka burrowing or nesting (Kitchener 1995). There is also no previous record of rabbits inhabiting the same dense, moistly-vegetated swamps and coastal heaths that quokkas do (Christensen et al. 1985) and the few rabbits trapped in this study (4 out of 1,014 captures of all species) were only captured at sites close to agricultural or mined areas (Appendix A).

(Morton 1990a) and attributed the ultimate causes of the decline of critical weight range mammals in the arid zone to competition and habitat alterations caused by rabbits and sheep. Conversely, competition with introduced herbivores is not thought to have directly initiated the declines of critical weight range mammals in Western Australia

(Burbidge and McKenzie 1989). Furthermore, the area historically inhabited by quokkas has largely been avoided by exotic herbivores due to their low tolerance to the sodium monofluoroacetate poison in plants of the genus Gastrolobium that is endemic to the south-west (King et al. 1981; Twigg and King 1991; King 1993).

87 Chapter 3 Distribution and status

The widespread decline in desert fauna was attributed to the destruction of areas of refuge by rabbits and other introduced herbivores (Morton 1990a; Lunney 2001).

Introduced herbivores seem unlikely to ever have been abundant in the forested south- west due to the presence of poison plants. On this basis it seems they are unlikely to have significantly directly affected the quokka or other species that have declined in that region (woylie, chuditch, southern brown bandicoot or Gilbert’s potoroo).

Main (1979) presumed Rottnest Island to be inhabited by the tammar and brush (M. irma) wallaby and the (M. fuliginosus), along with quokkas when it was isolated from the mainland. He attributed the persistence of the quokka there to competitive elimination of the slower reproducing species through the quokkas’ ability to withdraw to refuges around the salt lakes in drought and expand their range in favourable conditions (Main 1979) despite requiring a higher quality diet than the tammar (Main and Yadav 1971). Additionally, the earlier reports of quokka abundance

(Table 2) in the presence of these other native macropods suggests they do not out- compete the quokka.

Algar (1986 in Short et al. 1992) attributed the failure of the reintroduction of quokkas near Perth to overgrazing by rabbits and macropods. Another explanation, favoured by Short et al. (1992), was direct predation by introduced predators, although overgrazing may have been a contributory factor through the reduction in the amount of refuge (Morton 1990a).

Climatic influences

The distribution of the quokka since European arrival, largely follows the pattern of rainfall in the south-west of Australia (Figure 1). The coastal quokka populations

88 Chapter 3 Distribution and status

extended to areas that exceeded 700 mm of annual rainfall while inland or forest populations appear to require in excess of 1000 mm of annual precipitation to persist

(Figure 2). The quokka population in the Stirling Ranges survives despite the surrounding area’s aridity because the topography creates supplementary, orographic rainfall. This provides the quokka with moist, densely vegetated areas in the upper reaches of creek systems as habitat (Packer 1977). It seems likely that this seemingly isolated population survived from wetter periods in the Pleistocene before being isolated as the surrounding area dried. The fossil deposits containing quokka bones outside the current 700 mm isohyet may reflect an expanded range during wetter climatic periods.

The relatively high water requirements of the quokka, despite its efficient kidneys, are probably the reason for this link with rainfall (Main and Yadav 1971). The Rottnest

Island population suffers seasonal mortality over summer which has been attributed indirectly to dehydration (Main 1959; Barker 1961; Holsworth 1964; Storr 1964b;

Packer 1968). Nevertheless, quokkas persisted in the fossil record throughout periods of significant climatic variation from warm and wet to glacial aridity while several other forest-dependent mammals (e.g. Gilbert’s potoroo, Potorous gilberti) apparently became locally extinct (Balme et al. 1978). This restriction to increasingly higher rainfall areas may be due to the availability of resources in the Darling Range following recent multiple disturbances (Burbidge and McKenzie 1989) or, in light of the impact introduced predators appear to have had on quokkas, the refuge offered from predation in the denser vegetation there.

The thermoregulatory ability of the quokka is adequate to have met ambient temperature conditions throughout its distribution during the Pleistocene and earlier

(Bartholomew 1954) and so is not a limiting factor on the species’ distribution. While

89 Chapter 3 Distribution and status

there is fossil evidence for a prehistoric range contraction and while Cook (1960) suggested climate change was the cause of the modern decline of the quokka, the climate has not been reported as altering dramatically in the 1930s. Inspection of the long-term rainfall data from sites throughout the south-west shows that the decades preceding and during the 1930s experienced above average rainfall (Figure 9). This trend also occurred to the west of the forest in the wheatbelt (Hobbs 1993). Considering the importance of water to quokkas, it seems unlikely that climate change was the direct cause of the species dramatic decline. Furthermore, the absence of mass extinctions during earlier Pleistocene climate change (Caughley 1987) or on continents with a long history of human occupation (Africa and Europe) (Caughley and Gunn 1996), coupled with the anticipated regionality predicted for modern climate change (Allan et al. 1991;

Mitchell et al. 1994), cast doubt on the impact of climate change on most species.

Another recent theory is that exposure to climatic fluctuations in the past meant that species susceptible to extinction from such threats were filtered out, through their extinction, and those that survived are henceforth relatively resilient to similar threats

(Balmford 1996). Consequently, having survived the climatic fluctuations of the

Pleistocene (Balme et al. 1978), the quokka may be able to cope with future climate change.

Conversely, the above average rainfall during the decline of the quokka may have indirectly contributed. Quokkas move toward the periphery of the swamps as they become inundated with water during the winter rains (Chapter 5). This period of above average rainfall coincided with the arrival of the fox and this may have increased the susceptibility of the quokka to predation (Chapters 5 and 7).

90 Chapter 3 Distribution and status

This may have been exacerbated subsequently by the below average rainfall that has occurred in the south-west of Australia since the early 1970s (Wahlquist 2002) (Figure

9). With climate change predicted to increase the aridity of the south-west, with a tendency to more variabilty (Allan et al. 1991; Mitchell et al. 1994), the water requirements of the quokka (Main and Bakker 1981) may mean extant populations will be threatened by future climate change, even with reduced levels of predation.

91

400

300

200

100

0

-100 Mean precipitation difference (mm) -200

-300 1876 1881 1886 1891 1896 1901 1906 1911 1916 1921 1926 1931 1936 1941 1946 1951 1956 1961 1966 1971 1976 1981 1986 1991 1996 2001 Year

Figure 9. Long-term, mean precipitation difference for the south-west of Australia (Perth, Albany, Margaret River, Dwellingup and Rottnest Island) (grey lines) with a ten year running mean showing the longer term trends in mean rainfall. Precipitation difference was calculated by subtracting the mean rainfall of the five towns that span the south-west from the mean annual rainfall for each year. The grey lines at 1929 and 1935 show the time period in which the quokka is reported to have declined.

92 Chapter 3 Distribution and status

The effects of European colonisation

There is no real evidence to suggest that pre-agricultural societies altered their habitat sufficiently to lead directly to the extinction of any species (Eldridge 1999). After 10,000 years of increasing anthropogenic disturbance, habitat alteration is only now being identified as the primary cause of species loss (Eldridge 1999). The rapid decline of the quokka in the mid-1930s occurred over 100 years after the settlement of the Swan River and King Georges Sound (Albany) by Europeans (Bolton and Hutchinson 1979). The effects of colonisation are unlikely to have directly caused this decline through habitat alteration or loss, considering the decline occurred so long after colonisation (Cook 1963).

The success of quokkas near urbanised areas on Rottnest Island highlights the resilience of the species. The quokkas around the tourist areas of the island appear to have thrived as individuals are frequently observed entering houses to feed (pers. obs.) and with the population that inhabits the rubbish tip found to obtain enough nutrients to allow year- round breeding compared to those animals living away from the settlement (Shield 1964,

1968).

Yet the effects of colonisation are ongoing and cumulative and may become more significant when operating as part of an extinction vortex (Figure 7). While all the early collectors of quokkas reported them as very common up until the 1930s (Shortridge 1909;

White 1952; Gould 1973), it is important to emphasise that they were common but only in quite specific habitats (Storr 1964a, 1965b; Holsworth 1967; Sinclair et al. 1996). Much of the coastal heath and shrubland habitats in the south-west of Australia where quokkas once occurred (Shortridge 1909; White 1952) have been cleared for urban development.

The coastal plain from north of Perth to Busselton has very few pockets of undisturbed vegetation remaining. The remaining fragments are small and highly susceptible to

93 Chapter 3 Distribution and status invasion by introduced predators. Such total clearance of habitat is a far more serious disturbance than degradation or fragmentation alone (Mace and Balmford 2000).

Additionally, the upper reaches of creek systems in the forested areas of the Darling Range have often been cleared for agriculture, dammed to supply water to Perth or split from connecting habitat by roads (pers. obs.) such that the current state of quokka habitat is fragmented.

Mining and logging may have affected the quokka. Calver (1998a) concluded that there is no evidence that quokkas have suffered directly from logging, particularly with the maintenance of streamside reserves that are likely to protect the habitat of the species

(Anon 1999). Logging however, may cause indirect impacts on the quokka through increased predation rates and increasing the area of edge-affected habitat (Calver and Dell

1998b) as well as direct loss or degradation of the habitat.

Mining has directly impacted upon some local quokka populations via the large-scale clearing operations (e.g. associated with the construction of crusher sites for the bauxite mines near Dwellingup) where previously known populations once occurred (M.Dillon pers. comm.). Bauxite mining companies are required to maintain a protected area of approximately 50 metres around waterways in a similar fashion to logging operations

(Anon 1999). This may protect quokkas today as they are restricted to the swamp and nearby areas (Chapters 5 and 6b). Thus, while colonisation may not have directly led to the initial decline of the quokka, it seems likely that it has continued to place significant pressure on the species. Although the quokka is locally extinct at swamps that have been drained, salinised or otherwise altered, it has not become extinct on the mainland overall despite multiple disturbances (Burbidge and McKenzie 1989).

94 Chapter 3 Distribution and status

The impact of anthropogenic disturbance on the mainland since settlement may have more long-term ramifications however. Increasing economic development in a country has been linked to increasing numbers of threatened species (Naidoo and Adamowicz 2001) and, as prosperity of the colony in Western Australia increased, so too did the threat to fauna. Furthermore, population modelling suggests that extinctions are still likely to occur generations after disturbance and as such they represent an extinction debt or future ecological cost of earlier events (Tilman et al. 1994). The decline of the quokka may be the result of such a debt, arising from earlier disturbance or, more concerning, modern disturbances are likely to have future repercussions.

Fire

Aborigines have occupied the south-western corner of Australia for at least the last

30,000 years (Merrilees et al. 1973; Balme et al. 1978) and possibly longer (48,000 years according to Turney et al. 2001). Throughout this time they have significantly managed their environment through the use of fire (Jones 1969) and such modification of the environment is thought to have adversely affected native fauna (Merrilees 1967). Burning releases nutrients from plants to produce a flush of relatively nutritious and digestible growth (Morton 1990a). Aboriginal burning practices may have increased the dependability of certain habitat types, expanded suitable patches and provided corridors for recolonisation of dependable patches (Morton 1990a) and Aborigines used these ecological features as a tool for hunting (Nicholson 1981).

The decline in arid zone mammals has been linked to the alteration of these fire regimes following the departure of Aborigines from their traditional lands into settlements

(Burbidge et al. 1988). This movement is thought to have resulted in an alteration from a

95 Chapter 3 Distribution and status mosaic of small-scale burns to less frequent, but larger scale, wildfires ignited by lightning

(Morton 1990a). These reduced the mosaic of habitats available to animals for refuge and foraging (Bolton and Latz 1978; Johnson et al. 1989). Whether Aborigines moved as a result of the declining abundance of their food sources or their food sources declined following their departure is unknown.

The Nyoongar people of the south-west burnt the jarrah forest at a frequency of every three to four years with low intensity fires, however, moist areas such as swamps would have escaped fire for longer periods (Wallace 1966; Burrows et al. 1995; Ward and

Sneeuwjagt 1999). Quokkas survived with this burning regime until the arrival of

Europeans altered it through fire exclusion policies some 170 years ago (Wilson and

Friend 1999).

In a small-scale study of three sites around Dwellingup, quokkas returned to swamps to forage almost immediately after a burn and became resident after the first year

(Christensen and Kimber 1975). An older (15 years since last burn) site trapped in that study led the authors to conclude that quokkas desert long-unburnt sites (Christensen and

Kimber 1975). A spatial mosaic or patchy burn is preferable to provide areas of refuge and areas of foraging habitat (Christensen and Kimber 1975). Only recently have areas in the jarrah forest been once again managed using frequent, patchy, low-intensity fires (Ward and Sneeuwjagt 1999). On Rottnest Island the increased fire regimes since permanent settlement in 1838 led to vegetation changes from woodland to heath (Pen and Green

1983), particularly in the presence of quokka overgrazing (Storr 1963), yet this has not affected quokka abundance (Johnson et al. 1989).

The small and scattered quokka populations (Sinclair and Morris 1996) are also highly susceptible to catastrophic fires. A wildfire in the Nuyts Wilderness east of Walpole in

96 Chapter 3 Distribution and status

2001 reportedly killed ‘numerous’ quokkas (Anon 2001-2002). A high-intensity wildfire burnt a large portion of the Stirling Range National Park including Bluff Knoll in

November 2000 (pers. obs.) which may have resulted in the demise of that quokka population. A similar fire there in 1991 was thought to have been equally as damaging to quokkas (Sinclair 1999) however they repopulated burnt areas in the proceeding years

(Table 4). Insular populations are similarly threatened.

Indirect impacts of fire also exist. The increased predation rate following fire severely impacts prey (Christensen 1980) and, at low population densities, may threaten localised extinction.

A test of the vegetation mosaic hypothesis on Barrow Island revealed golden bandicoots (Isoodon auratus), common brushtail possums and burrowing

(Bettongia lesueur) were not affected by the presence of a mosaic of seral stages in the absence of predation (Short and Turner 1994). Despite this and considering the findings of

(Christensen and Kimber 1975), it seems that the alteration of fire regimes since Europeans arrived was a factor contributing to the decline of the quokka on the mainland, particularly as part of an extinction vortex. The survival of the quokka in the future may be dependent on human intervention through the management of their habitat with fire.

Disease

Reports from captive quokka colonies show the susceptibility of the species to disease.

Anecdotes suggest that in 1967 a fatal herpes epidemic transmitted by a handler affected a colony in Western Australia (Burnet 1970). Salmonella infections are common in the

Rottnest and Bald Island quokka populations but far less so on the mainland yet these do

97 Chapter 3 Distribution and status not cause mortalities (Hart et al. 1986). Several other parasites and diseases have been isolated from quokkas but none has been reported as causing mass mortality (for a description see Gibb et al. 1966).

The swamps around Northcliffe were reportedly full of quokka carcasses in the 1920s

(G.Gardner pers. comm. to How et al. 1987) which was attributed to disease (How et al.

1987). This date may be questionable in light of other references to quokka carcasses in swamps in the early 1930s (White 1952; Cook 1960) and considering the time since the occurrence. In a recent reanalysis of these mass deaths, surplus killing by foxes was suggested as the cause rather than disease (Short et al. 2002).

The decline in arid zone mammals of the late 1800s mentioned by Shortridge (1909) was also recorded from the Nullarbor Plain (Richards and Short 1996). The protozoan, parasitic disease , passed on from feral cats (Felis catus), may have been the cause (Richards and Short 1996). Toxoplasmosis has been linked to disease in other marsupials including the quokka (Kakulus 1963; Gibb et al. 1966) but (Recher and Lim

1990) considered it only as a contributing factor in their decline. Recent research however, indicates toxoplasmosis may alter the behaviour of its intermediate hosts to increase their susceptibility to predation, thereby ensuring transmission of the parasite to its definitive host – the cat – and ensuring the completion of its life cycle (Berdoy et al. 2000). While the date of arrival of cats on the mainland is unknown (Gaynor 2000), it may have corresponded to the decline of arid zone native mammals in the 1880s (Richards and Short

1996). The decline of quokkas occurred in the 1930s (White 1952), which was many years after cats arrived in the south-west (Gaynor 2000).

Despite the references to quokkas dying from disease in the 1930s (White 1952; Cook

1960), the persistence of quokkas on Rottnest and Bald Islands through this time suggests

98 Chapter 3 Distribution and status that disease was not the major contributor to their decline. No other author since then has linked the decline of the quokka with disease or recorded disease killing wild quokkas and others have not considered disease as a factor associated with the decline of critical weight range macropods altogether (Johnson et al. 1989; Morton 1990a).

Chains of extinction

Diamond’s (1989) fourth category of extinction causes was chains of extinction.

Chains of extinction are secondary extinctions whereby the extinction of one species causes the extinction of another upon which the first depends (Caughley and Gunn 1996).

Although the quokka and Gilbert’s potoroo occurred in sympatry throughout the southern, karri forest region of Australia (see Kabay and Start 1976) and the latter was, until recently, thought to be extinct (Sinclair et al. 1996), there appears no justifiable direct link between the declines of the two. The potoroo still coexists with the quokka in the former’s last stand in the west, Two People’s Bay (Sinclair et al. 1996; Sinclair 1999).

Conclusion

The geographical range of a species is limited at the point where the sum of immigration and birth is at a level below that of emigration and death (Gaston 1990). The quokka is probably no exception with abiotic factors, particularly rainfall, limiting the distribution originally in conjunction with biotic factors such as predation and competition.

Since the 1930s biotic factors have become increasingly important such that predation, in conjunction with changed fire regimes, took over from biotic factors and reduced this distribution.

99 Chapter 3 Distribution and status

Like most ecological problems, extinction is rarely caused by one factor alone

(Gittleman and Gompper 2001). For the quokka, like numerous other native species, the arrival of Europeans to Australia began a slow but precipitous decline in both abundance and distribution (Figure 10). The increasing European population in the south-west of

Australia resulted in anthropogenic disturbances such as vegetation clearance, logging, mining, hunting and fire suppression. Introduced predators have been implicated in 40% of historic extinctions (Caughley and Gunn 1996), as well as those of Australia’s

Pleistocene megafauna (Flannery 1990a; Flannery 1991; Flannery 1994; Flannery and

Roberts 1999), and it seems likely that predation pressure from the introduced red fox on a species already under pressure initiated the negative feedback relationship of an ‘extinction vortex’ (Diamond 1989) (Figure 10). Continued habitat alteration through fire exclusion and colonisation placed more stress on the quokka, while fox and, probably cat, predation continued and led to the current vulnerable status of the species (Figure 7).

Like most Australian native mammals the quokka is probably resilient to individual disturbance factors, but increasingly more susceptible to combined and cumulative factors

(Wilson and Friend 1999). None of the factors discussed above occurred in isolation, rather they occurred more or less in synchrony with the arrival of the fox. It seems likely, therefore, that several factors have contributed to the reduced distribution and vulnerable status of the quokka in a similar pattern to Diamond’s (1989) ‘evil quartet’ of extinction causes (Figure 7) or Morton’s (1990a) flow-charts of extinction forces. This is consistent with the view that species decline is caused by multiple disturbances driven by a resource- based mechanism (Burbidge and McKenzie 1989; Morton 1990a) that essentially lowers the carrying capacity of the land.

100 Chapter 3 Distribution and status

Considering the disagreement as to the ultimate cause of the decline of critical weight range mammals since European arrival (Burbidge and McKenzie 1989; Morton 1990a;

Recher and Lim 1990; Smith and Quin 1996; Wilson and Friend 1999; Lunney et al. 2001;

Short and Calaby 2001; Short et al. 2002) it may be that each individual declining species are impacted to differing degrees by the factors that comprise Diamond’s (1989) ‘evil quartet’. As such, hypotheses claiming to explain mass extinction events of the past and present by highlighting only one cause are likely to be flawed.

Yet for the quokka, there is plenty of evidence linking one factor to the decline.

Islands often act as refuges for threatened species due to the fact that often many of the factors that caused the decline on the mainland are absent (Dickman 1992a). The quokka population on Rottnest Island may therefore provide clues as to the significance of the various factors that led to its decline. Compared to the mainland, the Rottnest population has a higher susceptibility to disease (Salmonella) (Hart et al. 1986); encounters seasonal aridity leading to summer mortality (Waring 1959; Shield 1964) and a period of anoestrus

(Shield 1964). Both areas have populations of feral cats (pers. obs.) and birds of prey

(Storr 1965a) that may prey upon quokkas. Both areas have significant areas of urban development (pers. obs.) and have had contiguous habitat fragmented by roads and fences

(on Rottnest Island) (pers. obs.). The success of the quokka population on Rottnest Island compared to those on the mainland, despite these disturbances, suggests that the species is quite resilient to them. The most significant difference between Rottnest Island and the mainland with regard to quokka abundance appears to be the presence of the red fox on the mainland. Recent interesting advances in our understanding of the impact of the fox on native fauna reinforce this conclusion (Kinnear et al. In press). There is a disparity between the predation efficiency of foxes and the anti-predator defences of a prey fauna

101 Chapter 3 Distribution and status that evolved in relative isolation since the break-up of Gondwana 50-60 mya (Short et al.

2002). The behavioural mechanism causing the dramatic declines of many prey species may have been surplus killing by foxes upon their arrival to a particular area (Short et al.

2002). Considering foxes spread across Australia at up to 140 km per year (Short et al.

2002) and reached a peak in numbers 5-15 years after their arrival in an area (Jarman 1986) it is easy to envisage the rapid carnage they wrought.

It is thought that species with little or no refuge from predators would have suffered most from this predation (Gittleman and Gompper 2001; Short et al. 2002) and the preponderance of critical weight range mammal declines in the arid zone (Morton 1990a) supports this. The quokka today is restricted to areas of refuge (islands and dense vegetation) and, to date, this seems likely to have been the sole factor staving off extinction in the face of the onslaught of fox predation coupled with altered fire regimes and habitat loss. Persistence of threatened species in refugia is not considered a security in the long- term due to the instability associated with small populations (Kinnear et al. In press) and such persistence may simply delay their extinction. The future of threatened prey species may rest with understanding both current extinction threats and historical interactions between them and their predators (Gittleman and Gompper 2001) to determine their ability to cope with variations in predation rates, particularly if prey have high levels of predation in the past (Balmford 1996).

If the decline of native fauna has been buffered in more mesic areas of the mainland compared to more arid areas (Recher and Lim 1990; Woinarski et al. 2001) then it may well be due to the increased availability of resources (nutrients, water, refuge) there. If a new wave of extinction is upon us (Recher and Lim 1990; Woinarski et al. 2001) then, in the 214 years of changes wrought by Europeans, these resources may have become

102 Chapter 3 Distribution and status depleted and therefore no longer provide the buffering that they once did. This may be indicative of an overall ecosystem collapse and the belief that critical weight range mammals in mesic areas are secure (Burbidge and McKenzie 1989) may no longer be true.

103 Chapter 3 Distribution and status

TIME

Prehistorically the quokka was The arrival of Europeans in one of the most abundant 1829 to the south-west and The arrival of the European species in the south-west of their subsequent expansion red fox in the 1930’s seems Australia with populations probably reduced quokka to have been the catalyst to limited only by natural numbers marginally as the push quokka numbers over resources, Aboriginal hunting threat of habitat alteration the edge and into a massive and indigenous predators. through clearing and changed decline. The other factors fire regimes. impacting on quokkas since the arrival of Europeans contribute as an Figure 10. A schematic representation of the decline of the quokka. extinction vortex is created. QUOKKA POPULATION SIZE

The future? With a reduction in predation pressure following fox control and management of the habitat through a return to previous fire regimes it is hoped that the quokka will again become one of the more abundant species in south-western Australia. ?

Chapter 4 Population dynamics

Local population structure of a naturally-occurring

metapopulation of the quokka Setonix brachyurus

(Macropodidae: Marsupialia) (Quoy & Gaimard 1830) in the

northern jarrah forest of Australia.

This chapter has been published in Biological Conservation 110: 343-355

Abstract. Aspects of the population structure of a threatened macropodid marsupial, the quokka (Setonix brachyurus), were investigated using mark-recapture techniques. Seven previously known mainland populations and one unconfirmed site were studied in detail. No evidence of quokka presence was detected at three of these sites where the quokka is now considered to be locally extinct. The five remaining sites were trapped intensively for up to 2½ years. The surviving populations had extremely low numbers, ranging from a single individual at one site to 36 individuals at another.

Population density at these sites ranged from 0.07 to 4.3 individuals per hectare. The majority of these populations also appear to be threatened with localised extinction despite quokkas having a high fecundity. No movement was observed between the widely spaced sites. The remaining quokka populations in the northern jarrah forest may be the terminal remnants of a collapsing metapopulation. The proportions of adults, juveniles and pouch young were similar to those observed on Rottnest Island, however the observed parity in pouched young sex ratios has not been recorded from

Rottnest previously. Quokka condition, based on tail fat storage, varied seasonally.

This study confirmed previous work that found captive mainland quokkas bred throughout the year whereas the two insular populations bred seasonally. A reduction in births on the mainland occurred over summer which coincided with a decline in female

105 Chapter 4 Population dynamics

body condition. The apparent lack of a response by quokkas to the on-going, six-year- old fox control programme occurring in the region may be explained by low recruitment levels and suggests additional management intervention may be required to prevent metapopulation collapse.

Introduction

The quokka, Setonix brachyurus (Quoy & Gaimard 1830), is a small macropodid marsupial that is endemic to the south-western corner of Australia (Chapter 3)

(Kitchener 1995). It is the sole member of the genus whose affinities within the subfamily are extremely uncertain (Baverstock et al. 1989; Kirsch et al.

1995; Burk et al. 1998) but have been reported to lie with the genera Macropus

(Thomas 1888; Richardson and McDermid 1978; Baverstock et al. 1989), Thylogale

(Sharman 1954; Flannery 1989), Dorcopsulus (Kirsch et al. 1995), Petrogale

(Kitchener 1995) or Lagorchestes (Burk and Springer 2000).

The quokka was widespread and abundant when Europeans settled the south-west of Australia in the 1830s (Shortridge 1909; Gould 1973). It was distributed along the coast and high rainfall forested areas of the south-western corner of Western Australia as well as on Rottnest, Bald and Breaksea Islands (Chapter 3) (Shortridge 1909; Gould

1973; Kitchener and Vicker 1981; Kitchener 1995). On the mainland the species suffered a drastic decline in both abundance and distribution during the 1930s (White

1952). This decline has continued (Chapter 3) (de Tores et al. In prep.) and consequently the quokka is listed as a threatened species in accordance with the IUCN criteria for vulnerable (Hilton-Taylor 2000); as threatened fauna (Vulnerable) in accordance with the Commonwealth of Australia’s Environmental Protection and

106 Chapter 4 Population dynamics

Biodiversity Act 1999; and as “fauna which is rare or likely to become extinct” in accordance with the Western Australian Wildlife Conservation Act 1950 and subsequent schedules therein.

The Rottnest Island quokka population is estimated to fluctuate around 5,000 individuals (Waring 1956) with peaks of up to 10,000 individuals (Dickman Undated).

The Bald Island population is estimated to range between 200 and 600 (Main and Yadav

1971). Population estimates from the widely scattered mainland populations are thought to rarely exceed 10 individuals (Sinclair and Morris 1996; de Tores et al. In prep.).

Quokkas have been extensively studied on Rottnest Island, however there is no rational for extrapolating their ecology on the island to the mainland populations. In contrast to the mainland, the Rottnest Island quokka population is subject to seasonal aridity that leads to mortality (Main 1959), an absence of fresh water over summer

(Dunnet 1962; Storr 1964a), a lack of shade (Sadlier 1959 in Storr 1964a; Kitchener

1972) and an absence of the introduced predator – the red fox (Vulpes vulpes) (Dunnet

1962; Sinclair 1999). In addition, mainland and island populations differ in their susceptibility to disease and ability to store fat (Hart et al. 1986), tolerance to the poison sodium monofluoroacetate (Mead et al. 1985), breeding pattern (Shield 1964) and genetics (Sinclair 2001). Despite the long-term knowledge of the decline of the quokka

(Chapter 3) (White 1952) and therefore the apparent need to investigate the causes, research into the species ecology on the mainland has been hampered by the difficulty in locating and capturing animals from the small, widely-scattered and cryptic populations there (Sinclair 1999) and a lack of financial initiative and support.

107 Chapter 4 Population dynamics

The purpose of this study was to derive more reliable estimates of population size at known localities within the northern jarrah forest. In the process, data were collected on the demography of each population, morphological features of captured quokkas and on their reproductive characteristics.

Study area

Vegetation and climate

Jarrah forest is a tall, open forest dominated by Eucalyptus marginata (jarrah) and

Corymbia calophylla (marri). The jarrah forest is a multiple use forest managed, amongst other things, for both forestry and conservation. The jarrah forest is situated on the duricrusted Yilgarn plateau with soils that are dominated by laterite gravels

(Thackway and Cresswell 1995). Agonis swamp shrublands occur on alluvial deposits along the upper reaches of creek systems in the western half of the forest (Thackway and Cresswell 1995) and it is within these densely vegetated swamps that quokkas are restricted (Barker et al. 1957; Storr 1964a; Christensen et al. 1985).

The northern jarrah forest comprises the northern half of the jarrah forest biogeographic region (Thackway and Cresswell 1995) and is known as the Darling sub- district (Beard 1980). It extends from Mundaring in the north to Collie in the south and is bounded to the west by the Darling Range and the east by the cleared agricultural areas of the wheatbelt (Figure 1). It was selected for this study as it forms the northern- most area of extant quokka distribution (Chapter 3) and it was thought that the threatening processes acting on the quokka throughout its range would be most intense and obvious in this region.

108 Chapter 4 Population dynamics

The study area has a Mediterranean climate characterised by hot, dry summers and mild, wet winters (Southern 1979). Monthly average maximum temperatures at

Dwellingup (Figure 1) range from 29.7oC in January to 14.9oC in July, while average monthly minimum temperatures peak in February at 14.8oC and are lowest in August with 5.5oC (Bureau of Meteorology). Mean annual rainfall is 1,200 mm in the north

(between Mundaring and Dwellingup), 1,265 mm at Dwellingup and 1,179 mm near the

Victor Road site (Figure 1). As is characteristic of Mediterranean climates, the majority of this rain falls in winter and the least falls over summer (e.g. 693 mm in winter and 60 mm in summer at Dwellingup). The study was conducted over a period of average rainfall (Bureau of Meteorology).

Study sites

Ten sites were initially investigated for quokka presence (Figure 1). Sites were classed as the swamp, or the upper reaches of creek systems vegetated by Agonis linearifolia, and the surrounding jarrah forest. Nine of these sites supported quokkas in the early 1990s (M.Dillon pers. comm.) (de Tores et al. In prep.) and the Hoffman site had an unconfirmed sighting in the late 1990s (A.Danks pers. comm.). The Albany

Highway and Gervasse sites were eventually excluded from the project due to feasibility constraints. The Gervasse site was already monitored annually (R.Brazell pers. comm.)

(de Tores et al. In prep.) and was thought to have a high population density, while the abundance of quokka faecal pellets at the Albany Highway site compared to other trapped sites indicated a moderate population density (Chapter 2) (pers. obs.).

109

Enlarged area Indian Ocean

(

Figure 1. Location of study sites (stars). Sites underlined were trapped but contained no quokkas, while sites shown in italics were excluded due to feasibility constraints. Towns are shown in bold and their location is depicted by filled circles. The northern jarrah forest is stippled light grey and the Darling Range generally runs along its western boundary. The 1000 mm (dashed line) and 700 mm (dotted line) annual rainfall isohyets are also shown.

110 Chapter 4 Population dynamics

Fox baiting

Having evolved in association with the poison-bearing plants of the genus

Gastrolobium, Western Australian fauna have a relatively high tolerance to the toxin sodium monofluoroacetate (1080), particularly compared to introduced species (King et al.

1981; Eason and Frampton 1991; King 1993; Thompson and Fleming 1994; Short et al.

1997; Saunders et al. 1999; Twigg et al. 2000). The quokka also possesses such a tolerance

(Mead et al. 1985). As such, 1080 is widely used in Western Australia to control introduced pest species as part of the Department of Conservation and Land Management’s

(CALM) large scale fox control programmes (Operation Foxglove and Western Shield). In these programmes vast areas are subject to broadscale, aerial poisoning at five baits per km

(de Tores 1994, 1999; Thomson and Algar 2000).

Additionally, the local area around the Rosella Road, Chandler Road, Wild Pig,

Kesners, Hadfield, Albany Highway and Gervasse sites was baited monthly along vehicle tracks with dried meat baits injected with 4.5 mg of 1080 at intervals of 100 metres. Based on preliminary results of Operation Foxglove this fox control yields substantial decreases in fox density (de Tores 1999). The Holyoake and Hoffman sites were trapped in an attempt to find another unbaited control site in addition to the Victor Road site. The location, length and fox control details of the eight sites trapped are shown in Table 1.

111 Chapter 4 Population dynamics

Table 1. Details of the location, size and baiting history of quokka swamps investigated during this study. Site Latitude Longitude Swamp Swamp Monthly fox control (S) (E) length (m) area (ha) start Chandler Road 32o18’24” 116o07’20” 1,340 11.3 March 1998 1 Hadfield 33o11’07” 115o58’25” 1,570 6.8 October 1997 2 Kesners 32o39’00” 116o00’59” 2,510 12.4 January 1999 2 Rosella Road 32o15’34” 116o04’36” 1,690 15.3 March 1998 1 Wild Pig 32o34’07” 116o03’03” 1,200 9.9 March 1999 2 Hoffman 33o02’12” 116o01’34” 2,000 8.0 Unbaited Holyoake 32o42’14” 116o05’24” 500 2.6 Unbaited Victor Road 33o16’10” 116o00’49” 930 8.2 Unbaited

1 Fox baiting occurred biannually since 1994 across wider, surrounding area as part of CALM’s Western Shield/Operation Foxglove programme 2 Fox baiting occurred quarterly since 1994 across wider, surrounding area as part of CALM’s Western Shield/Operation Foxglove programme

Methods

Trapping

Trapping was conducted at eight sites (Figure 1) wherein the entire Agonis-dominated areas of the creeklines were trapped. Trap stations were situated every 50 to 100 metres along the creekline. At each trap station, one large and one small wire cage trap were placed inside the swamp while a corresponding pair were placed between 30 and 80 metres outside the swamp (Figure 2).

112 Chapter 4 Population dynamics

Figure 2. Schematic representation of trap station layout.

Two sizes of trap were used - primarily to reduce non-target captures in the large traps but also to minimise unequal catchability of individual quokkas. Large wire cage traps measured 0.90 m x 0.45 m x 0.45 m while the small wire cage traps measured 0.59 m x 0.205 m x 0.205 m. Large traps were baited with apples while the small traps were baited with a ‘universal’ mix of peanut butter, rolled oats, honey and pilchards. Large wire cage traps had either a treadle or baited hook mechanism while small traps were all of the treadle variety. The six baited hook mechanism large traps were randomly positioned to avoid bias.

Past studies indicated that quokkas required pre-baiting and a period of familiarisation with traps before they could be successfully captured (M.Dillon pers. comm.) (Dunnet

1963; Kabay and Start 1976; Christensen et al. 1985). This approach was also found to

113 Chapter 4 Population dynamics

minimise unequal catchability of individuals on Rottnest Island (Dunnet 1963).

Consequently, a pre-baiting period of six or seven days was used after which traps were opened and cleared each morning for the following eight days. It is generally accepted that open population estimates require a minimum of three primary trapping periods (e.g. years/seasons/months/weeks) each containing five secondary sampling periods (e.g. days/hours) depending on the target species (Pollock 1982). With a threatened target species like the quokka, which was likely to have inherently low capture probabilities, eight secondary trapping periods (days) were considered necessary in addition to at least seven seasonal primary trapping periods at each site showing evidence of quokkas, to provide a more robust data set. Prior to autumn 1999 trapping was conducted for periods up to 51 consecutive days to obtain individuals for collar attachment and after autumn 2000 to remove these collars (Table 2). Although such extended trapping periods are considered unethical (Monamy and Gott 2001), it was considered more ethical than leaving animals with collars still attached at the end of the study and conducting minimal trapping to get inadequate data.

Table 2. Trap nights at each study site. Seasonal primary trapping sessions with number of trap nights in each secondary trapping session are shown. Number of days trapped shown in brackets.

114 Chapter 4 Population dynamics

Site 1998 1999 2000 Spring Summer Autumn Winter Spring Summer Autumn Winter Spring Chandler 1,636 (30) 880 (16) 540 (9) 480 (8) 480 (8) 480 (8) 480 (8) Hadfield 522 (9) 480 (8) 480 (8) 480 (8) 480 (8) 69 (9) 81 (12) Hoffman 368 (8) Holyoake 54 (6) Kesners 946 (18) 1200 (10) 832 (8) 912 (8) 794 (8) 800 (8) 72 (10) 286 (16) Rosella Road 1,611 (21) 596 (12) 480 (8) 480 (8) 480 (8) 480 (8) 480 (8) Victor Road 540 (9) 480 (8) 480 (8) 480 (8) 478 (8) 90 (10) 168 (24) Wild Pig 194 (9)

Animal handling and measurements

Captured animals were removed from traps and placed in hessian bags. Quokkas were sedated via intramuscular injections with nominal dose rates of 9 mg kg-1 of ketamine and

2.5 mg kg-1 of xylazil which allowed approximately 30 to 60 minutes of sedation. Each animal was weighed and various measurements, including right pes length and tail circumference, were made. The condition of each quokka throughout the study was assessed using the condition index:

3 M / L where M = body mass and L = pes length (Kerle 1998). The tail circumference was also used as a measure of condition. Reproductive status was assessed via testes and pouch inspections. The birth dates of pouch young were calculated by comparing morphological measurements to growth curves of captive quokkas (Shield and Woolley 1961). The same researcher (M.H.) took all measurements to minimise variation.

115 Chapter 4 Population dynamics

Quokkas were placed into three age classes; pouch young (restricted to the pouch) juveniles (i.e. permanently exited the pouch but not yet breeding) and adults. Adult animals were defined as males exceeding 2.5 kg and females exceeding 1.6 kg. This definition was based on observations of the reproductive status of individuals exceeding these weights during this study such as enlarged testes and the presence of pouch young.

This contrasts with previous studies which defined adult females as those exceeding 2.5 kg

(Shield and Woolley 1961; Dunnet 1963; Shield 1968; Sinclair 1998). The birth date of pouch young was estimated by comparing morphological measurements with the growth curves calculated by Shield & Woolley (1961).

New animals were fitted sub-cutaneously between the shoulder blades with a

TROVAN ® transponder microchip (Central Animal Records, Melbourne, Victoria). The unique ten digit alphanumeric code of the transponder was read using a TROVAN® mini- reader. The hole resulting from ear tissue samples, that were taken for genetic analysis, served as a secondary mark to ensure that no marks were missed or lost (an assumption of mark-recapture methods) (White et al. 1982).

Animals were released near the point of capture once they regained consciousness

(usually 1 hour). There was no mortality associated with animal handling or sedation although two adult females died within traps.

Species identification

The following field guides and references were used to identify non-target organisms trapped in this study: mammals (Strahan 1995); reptiles (Storr et al. 1983, 1990, 1999);

116 Chapter 4 Population dynamics

frogs (Cogger 1992); and birds (Slater et al. 1986). These non-target captures are presented in Appendix A.

Data analysis

The recruitment of juveniles into the population was estimated according to the methods of Gilfillan (2001). The number of juveniles expected to enter the population was estimated by counting the number of pouch young per female multiplied by the number of females with pouch young (Gilfillan 2001). This was then compared with the number of juveniles actually captured to get a recruitment percentage (Gilfillan 2001).

Three methods of estimating actual and relative population size were used in order to allow comparison with other studies. Capture success was determined by the number of newly captured individuals divided by the number of available traps (that is, excluding all closed or occupied traps). The minimum number of individuals known to be alive (KTBA) was estimated according to the methods of Krebs (1966), whereby an animal alive at the start and finish of the primary trapping periods was assumed to be alive in between these periods. Due to the low capture rates, the start period was classed as the pooled captures from spring 1998, summer 1998-1999 and autumn 1999 while the finish trapping period was classed as trapping after autumn 2000.

Open population estimates were attained using the ‘deaths but no immigration’ model of the Cormack-Jolly-Seber (C-J-S) method (Seber 1982) with the computer program

JOLLY. The ‘deaths but no immigration’ model is justified considering the philopatry

117 Chapter 4 Population dynamics

observed during radio telemetry (Chapter 5). C-J-S estimates were attempted because of the desire of wildlife managers for an actual population size rather than an index.

The robust-design estimator (Pollock 1982) was also available for population estimates, however results determined using program CAPTURE were not presented after it was found they were similar but had larger errors than those from the Jolly-Seber method (as recommended by Kendall and Pollock 1992). Program MARK (White and Burnham 1999) and its associated goodness-of-fit tests and other methods of model selection (Akaike’s information criterion (Akaike 1973, 1974), etc.) was not used due to the small sample sizes

(McKelvey and Pearson 2001).

Actual population size estimates were valid for this study as each site is thought to support discreet populations within the swamps (i.e. geographically closed) (White et al.

1982). All other assumptions of open population estimation were met (White et al. 1982).

Population density was calculated as the number of individuals estimated by the C-J-S method per hectare of swamp.

The recruitment of juveniles into the population was estimated according to the methods of (Gilfillan 2001). The number of juveniles expected to enter the overall population was estimated by counting the number of pouch young per female multiplied by the number of females with pouch young (Gilfillan 2001) over the entire study period. This was then compared to the number of juveniles actually captured to get a recruitment percentage (Gilfillan 2001).

Mean values are presented along with standard errors. Statistical calculations were analysed using Statview Version 5.0 (SAS Institute Inc. 1998). The log-likelihood test (G-

118 Chapter 4 Population dynamics

statistic) and contingency tables were used to compare age structures and sex ratios. Other tests conducted included t-tests, linear regression, analyses of variance (ANOVA) and covariance (ANCOVA), two-factor nested ANOVA and repeated measures ANOVA. For each of these tests, data were log transformed where necessary. In ANCOVA, where the slopes were not significantly different, interaction terms were deleted and the analyses recalculated to investigate differences in intercepts (Sokal and Rohlf 1969). Significant comparisons were assessed with Scheffe’s post-hoc test at 5% significance level. The chi- square (c2) test was used to investigate differences in female birth rates. The Kolmogorov-

Smirnov test was used to compare the uniformity of birth distributions throughout the year.

The Rosella Road site was excluded from several analyses due to its small sample size.

119

Table 3. Population estimates and capture success of adult quokkas at sites in the northern jarrah forest calculated for the entire study period. The high variability of quokka captures meant presenting seasonal capture data was irrelevant. Site Trap Captures per # individuals a Known to be alive Jolly (C-J-S) Density (# ha-1)b Recaptures nights 100 trap nights Chandler Road 4,976 0.20 8 4 10 0.9 11 Hadfield 2,592 0.99 21 21 29 4.3 69 Hoffman 368 0 0 0 0 0 - Holyoake 54 0 0 0 0 0 - Kesners 5,842 0.40 21 11 36 2.9 22 Rosella Road 4,607 0.03 1 1 C 0.07d 1 Victor Road 2,654 0.51 11 10 9 1.1 78 Wild Pig 194 0 0 0 0 0 - Overall 21,287 0.42 62 47 84 1.1 186 a Adults only. b Based on the C-J-S estimate divided by the area of the swamp (Table 1). c Population size at Rosella Road was too small to calculate using C-J-S method. d Based on one captured individual known to be alive through telemetry.

120

Chandler (8)

Hadfield (35)

Kesners (29)

Rosella (1)

Victor Road (22)

Overall (95)

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Adult Juvenile Pouched young

Figure 3. Age class breakdown (%) of quokkas at first capture for sites in the northern jarrah forest. Only sites with quokkas are shown. Numbers in brackets show the total number of individuals recorded. For short-lived animals with relatively long reproductive lifespans, a pyramid-shaped age breakdown, with more pouch young than juveniles than adults, would indicate an increasing population; an even (33.3% each age category) breakdown would indicate population stability and an inverted pyramid shape would suggest an aging population going into decline.

121 Chapter 4 Population dynamics

Table 4. Sex ratios of all newly captured quokkas recorded at sites in the northern jarrah forest. Only sites with more than one individual KTBA used. Sex ratio (M:F) Site Adults Juveniles Pouch young Overall Chandler Road 3 : 2 1 : 1 0 : 1 1.0 : 1.0 Hadfield 6 : 11 4 : 3 7 : 4 1.0 : 1.1 Kesners 10 : 6 5 : 5 0 : 3 1.1 : 1.0 Victor Road 5 : 4 3 : 1 5 : 4 1.4 : 1.0 Overall 24 : 23 13 : 10 12 : 12 1.1 : 1.0

Results

Trapping success and population size

Seven native mammal species were captured during this study: quokka, southern brown bandicoot (Isoodon obesulus), common (Trichosurus vulpecula), chuditch or western quoll (Dasyurus geoffroii), mardo or yellow-footed (Antechinus flavipes), (Macropus irma) and echidna

(Tachyglossus aculeatus). In addition, four introduced mammal species, the black rat

(Rattus rattus), feral cat (Felis catus), feral pig (Sus scrofa) and European rabbit

(Oryctolagus cuniculus), were caught. Four species of ; Egernia kingii, E. luctuosa, E. napoleonis and Tiliqua (Trachydosaurus) rugosa; three varanids, Varanus gouldii, V. rosenbergi and V. tristis; and the western tiger (Notechis scutatus); were also captured. The number of new individuals and total captures at each site are presented in Appendix A.

More than 21,000 trap nights yielded 281 quokka captures of which 95 were individual animals and 62 of these were mature adults (Table 3). Quokka presence could not be confirmed by trapping or faecal pellet counts at two sites (Holyoake and

122 Chapter 4 Population dynamics

Wild Pig) previously known to support them (M. Dillon pers. comm.) and at another site with suitable habitat and an unconfirmed sighting (Hoffman) (Table 3). This led to the early cessation of trapping at these three sites.

Capture success at sites with quokkas present varied from 0.03 animals per 100 trap nights at Rosella Road to 0.99 at Hadfield (Table 3). The known to be alive estimates varied in a similar fashion to capture success, ranging from one lone male at Rosella

Road to 21 individuals at Hadfield (Table 3). The C-J-S estimates indicate that the

Victor Road site has the smallest calculable population (9) and Kesners the largest (36), with the Rosella Road population being too small to calculate (Table 3). Based on the

C-J-S estimates, Hadfield supported the highest population density (over four adults per hectare), followed by Kesners, then Victor Road, Chandler Road and finally Rosella

Road. No individuals captured at one site were subsequently captured at another.

There was a significant linear relationship between the number of captures per 100 trap nights and the KTBA estimates (Table 5). No other regression of population indice, estimate or swamp characteristic was significant (Table 5).

123 Chapter 4 Population dynamics

Table 5. Regression analyses between the population indices (captures per 100 trap nights/known to be alive (KTBA)) and estimates (Cormack-Jolly-Seber (C-J-S)) as well as the length of each habitat patch. Comparison r2 n Probability Captures per 100 trap nights against KTBA estimates 0.98 6 < 0.0001 Captures per 100 trap nights against C-J-S estimates 0.51 6 0.11 C-J-S estimates against KTBA estimates 0.64 6 0.06 Captures per 100 trap nights against swamp length 0.01 6 0.89 Captures per 100 trap nights against swamp area 0.53 6 0.10 KTBA estimates against swamp length 0.05 6 0.69 KTBA estimates against swamp area 0.43 6 0.16 C-J-S estimates against swamp length 0.44 6 0.23 C-J-S estimates against swamp area 0.13 6 0.55

Population composition

The overall quokka population in the northern jarrah forest was made up of 50.5% adults, 24.2% juvenile individuals and 25.3% pouch young. The proportion of adult, juvenile and pouch young in each population varied between the sites (Figure 3) but not significantly (Contingency À2 = 9.341; d.f. = 20; P = 0.31).

There was no significant variation from parity for the sexes overall (G = 0.35; d.f. = 1; P

> 0.05) or at each age class (Contingency À2 = 0.212; d.f. = 2; P = 0.90). There was also no significant variation from parity of the sexes at each site (Contingency À2 =

17.579; d.f. = 20; P = 0.62) (Table 4).

124 Chapter 4 Population dynamics

Table 6. Mean body mass of quokkas from sites in the northern jarrah forest. Site Male mean (kg) ± Female mean (kg) ± Overall mean (kg) ± S.E. S.E. S.E. Chandler Road 3.54 ± 0.15 2.47 ± 0.23 3.18 ± 0.21 Hadfield 3.81 ± 0.13 2.73 ± 0.07 3.12 ± 0.10 Kesners 3.85 ± 0.13 2.18 ± 0.18 3.37 ± 0.18 Victor Road 3.43 ± 0.15 2.74 ± 0.06 3.02 ± 0.09 Overall 3.68 ± 0.07 2.65 ± 0.05 3.16 ± 0.07

Morphology and condition

Body mass of adult quokkas from the northern jarrah forest averaged 3.16 kg (±

0.07 kg) (Table 6). The species was sexually dimorphic there with males (3.68 kg ±

0.07 kg) significantly larger than females (2.65 kg ± 0.05 kg) (two-factor ANOVA of log10 transformed body mass (site * sex interaction P = 0.1363) F = 45.234, d.f. = 1, P <

0.0001). The heaviest male captured weighed 4.787 kg while the heaviest female weighed 3.379 kg. The same ANOVA revealed there was no significant difference between sites in adult body mass (F = 0.738, d.f. = 3, P = 0.535). There was no significant difference in mass of consecutively captured animals between the seasons

(single factor (sex) repeated measures ANOVA F = 1.28; d.f. = 3; P = 0.331) (Figure 4).

There was no correlation between mean population body mass and latitude of the swamps (r2 = 0.50; n = 5; P = 0.18) although a slight cline was apparent and nor was there a relationship between body mass and population density (r2 < 0.01; n = 56; P =

0.90).

125 Chapter 4 Population dynamics

4200

4000

3800

3600

3400

3200 Mass (g)

3000

2800

2600

2400

2200 Autumn Winter Spring Summer

Figure 4. Seasonal variation in mass (mean ± S.E. in grams) of male (n = 57) (triangles) and female (n = 58) (circles) quokkas from sites in the northern jarrah forest.

There was no difference in condition of quokkas among all the sites (nested

ANOVA of log10 condition with seasonal differences removed (interaction P = 0.24) F

= 1.16, d.f. = 3, P = 0.33). There was no correlation between condition and swamp area

(regression r2 = 0.008; n = 5; P = 0.33) or population density (r2 = 0.006; n = 5; P =

0.39). There was no significant difference in condition indices of males and females (t

= 0.54, d.f. = 123, P = 0.59) (Figure 5). There was no overall difference in body condition seasonally when variation in body mass was removed (interaction P = 0.48)

(ANCOVA F = 1.15, d.f. = 3, P = 0.33) (Figure 5). Repeated measures ANOVA also revealed no significant differences between the sexes seasonally (F = 2.34; d.f. = 1; P =

0.16).

126 Chapter 4 Population dynamics

.146

.144

.142

.140

.138 Condition Index

.136

.134

.132 Autumn Winter Spring Summer

Figure 5. Seasonal variation in condition (mean ± S.E.) of male (open triangles) (n = 57) and female (filled circles) (n = 58) quokkas.

Tail circumference was also used as an indicator of the condition of quokkas. Adult males and females differed significantly in tail circumference (t = -7.412; d.f. = 48; P <

0.0001). Increase in tail circumference compared body weight was twice as rapid in males (regression r2 = 0.78; slope = 0.4; n = 26; P < 0.0001) than females (r2 = 0.26; slope = 0.2, n = 26; P < 0.001). There was a significant difference in tail circumference between sites (ANCOVA with log10 body mass as covariate (interaction P = 0.40) F =

9.24; d.f. = 3; P < 0.0001) with animals at Chandler and Kesners having significantly larger tail circumferences than those at Hadfield and Victor Road (Scheffe’s test P <

0.03 for all). The removal of the log10 body mass covariate interaction (P = 0.199) revealed significant differences between seasons (ANCOVA F = 5.64; d.f. = 3; P =

127 Chapter 4 Population dynamics

0.001) with animals having significantly larger tail circumferences in summer than in all other seasons (Scheffe’s test P < 0.004 for each). A repeated measures ANOVA also showed such a difference for individuals captured in each season (F = 4.091; d.f. = 3; P

= 0.022), however in contrast to the ANCOVA above, animals had significantly larger tail circumferences in autumn than in winter (Scheffe’s test P = 0.03). There was not enough data to compare seasonal variation in tail circumference between the sexes using repeated measures ANOVA. Tail circumference peaked in summer and was lowest in winter for males and spring for females (Figure 6). The rise for females in summer was opposite the variation observed in body mass (Figure 4) and condition index (Figure 5).

Similarly, where both sexes increased body mass in winter (Figure 4), tail circumference declined (Figure 6). There was no correlation between tail circumference and population size (regression r2 = 0.001; n = 95 for all; P = 0.818), density (r2 = 0.014;

P = 0.412), swamp length (r2 = 0.016; P = 0.388) or swamp area (r2 = 0.043; P = 0.147).

85

80

75

70 Tail circumference (mm)

65

Autumn Winter Spring Summer

Figure 6. Seasonal variation (mean ± S.E.) in quokka tail circumference (condition). Males are shown as triangles and females as circles.

128 Chapter 4 Population dynamics

Reproduction

There were 56 captures of 47 adult female quokkas (new or recaptured from a previous trapping session) during this study of which 23 were breeding. Eighty-four percent of these females had pouch young of various ages, ranging from 100% at

Chandler (n = 2) and Kesners (n = 9) to 88% at Hadfield (n = 26) and 68% at Victor

Road (n = 19). These differences were not significant (c2 = 1.596, d.f. = 3, 0.75 < P <

0.50). There was no relationship between the number of females with young and population size (regression r2 = 0.22; n = 47 for both; P = 0.53) or population density (r2

= 0.01; P = 0.80).

The youngest female to give birth was 15 months old and weighed 1.65 kg, while another female of known birth date had a pouch young when trapped at 18 months of age. The average mass of breeding females was 2.77 ± 0.05 kg. It was impossible to estimate the age of males at first breeding, however testes were enlarged and conspicuous at 2.5 kg.

Adult females that were captured each season throughout the study averaged 2.2 young per year (mean = 2.15 at Victor Road ranging to 2.40 at Hadfield). Out of six females captured four times or more over 18 months, four had three consecutive young

(in pouch or at foot) and two had two young. No twins were observed.

The number of juveniles expected to enter the quokka metapopulation was 47. The number of juveniles observed entering the metapopulation was 19 which represented a

40% recruitment rate and was significantly less than expected (c2 = 19.62; d.f. = 3; P <

0.0001). The percentage rate of recruitment differs between the sites with Kesners having the highest rate (89%), then Chandler (50%), Victor Road (38%) and Hadfield

(22%).

129 Chapter 4 Population dynamics

In this study, monthly births were far more evenly distributed throughout the year than for quokkas on Rottnest Island (Shield 1964) (Figure 7), however they were still not uniformly distributed (Kolmogorov-Smirnov test D = 0.19, d.f. = 57, 0.02 < P <

0.05). When these data were pooled into seasons, the variation was even more significant (D = 0.35, d.f. = 57, P < 0.001) with summer being the season of fewest births (Figure 7). The addition of the 31 mainland births recorded by (Shield and

Woolley 1963) made no changes to the observed trends with summer births still occurring at approximately half the rate of the other seasons (summer 12 births, autumn

28, winter 24 and spring 24).

45% 40% 35% 30% 25% 20% 15%

Percentage of births 10% 5% 0% Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 7. Percentage of quokkas born each month in the northern jarrah forest (this study) (filled) (n = 57) and Rottnest Island (unfilled) (n = 299 taken from Shield 1964).

130 Chapter 4 Population dynamics

Discussion

Metapopulation attributes and its collapse

Difficulties in identifying metapopulations of larger mammals stems from the large temporal scales needed to record population turnover through localised extinction and colonisation with patch occupancy data (Elmhagen and Angerbjorn 2001). The restriction to discrete patches and the absence of correlation in life history characteristics between populations (e.g. Fig. 2) (Elmhagen and Angerbjorn 2001), as well as the population turnover south of the jarrah forest (G. Liddelow pers. comm.), suggests the quokka is a species that does exist naturally in metapopulations. This feature may minimise the need for 18 hectares to support a viable population.

Considering there was no movement observed between populations, the degree of isolation of each population in the northern jarrah forest is, however, an issue of concern. Additionally, the densely vegetated, swampy habitat of the quokka, although abundant in the northern jarrah forest, is not contiguous and the eight extant populations are scattered across an area of almost one million hectares (Fig. 1). There is virtually no potential for these highly disjunct populations to ‘rescue’ each other from extinction through repopulation (see Brown and Kodric-Brown 1977) and consequently the populations are unlikely to survive as part of a functioning metapopulation (see Hanski and Gilpin 1991). The arrival of the red fox in the south-west of Australia in the 1930s

(King and Smith 1985) is likely to have selected for philopatry as the predator-naïve dispersing individuals would have been easily killed by foxes when they left their swampy refuge. Historically however, such impediments to movement were not present and quokkas would undoubtedly have moved between patches. This is confirmed by genetic data (Sinclair 2001).

131 Chapter 4 Population dynamics

The genetic differences between quokka populations indicate that they now represent separate management units despite consisting of a single species or evolutionary significant unit (Sinclair 2001). Existing as a metapopulation allays some concern as the minimum viable genetic population size is thought to be lower for metapopulations than for contiguous populations (Gilpin 1991), although this may be underestimated (Hanski et al. 1996). The minimum viable metapopulation is suggested as ten extant populations with more than 20 suitable patches in the network (Hanski et al. 1996). Although there are countless suitable patches to support local quokka populations in the northern jarrah forest there are only eight known populations and these are not connected by colonising and dispersing individuals.

Population sizes

This study has confirmed the anecdotal reports of the low abundance of northern jarrah forest quokka populations. Considering the paucity of recent museum records

(Museum of Western Australia pers. comm.; Chapter 3) (de Tores et al. In prep.) and the reports of population declines elsewhere (Maxwell et al. 1996) this scarcity can be extrapolated to the rest of the mainland. The widely scattered populations rarely exceed

30 individuals and have a maximum density of around four quokkas per hectare of their small and fragmented preferred habitat.

The total adult quokka population at known sites in the northern jarrah forest may be as few as 150 individuals. This includes the estimated 84 individuals as part of this study, as well as estimates for the Gervasse and Albany Highway sites. Gervasse is a high density site with a population estimated at 49 individuals (de Tores et al. In prep.).

The scats and runways present at the Albany Highway site suggest a moderate

132 Chapter 4 Population dynamics

population density (~ 15) when compared to the trapped sites (pers. obs.). Only half of this overall known population of 150 is likely to be breeding females. A detailed survey however, is likely to identify additional quokka populations in the northern jarrah forest but these are unlikely to exceed the individual population sizes estimated in this study.

Quokkas appear to be locally extinct or on the verge of extinction at four of the eight sites trapped in this study (Holyoake, Wild Pig, Hoffman and Rosella Road). The

Rosella Road site has only one male remaining. Trapping in 1997 yielded two (A.

Tomkinson pers. comm.) and in 1995 four other individuals (P. de Tores & R. Buehrig unpubl. data) that were not recaptured during this study. Local extinction here appears imminent.

Quokkas were captured at Holyoake in the early 1990s (M. Dillon pers. comm.) yet by 1995 the population was believed to be extinct (de Tores et al. In prep.). This site is within one kilometre of the Dwellingup residential and agricultural areas and this proximity and associated disturbance may have affected the quokkas there (pers. obs).

A sizeable population of quokkas persisted at Wild Pig Swamp in the early 1990s

(M. Dillon pers. comm.). Evidence of continued presence was detected in 1995 but was unable to be confirmed by trapping (de Tores et al. In prep.). The absence of scats, the deterioration of runways and the lack of captures during this study suggests another localised extinction. Activity associated with bauxite mining has occurred to within 20 metres of the swamp around the upper reaches of this site (pers. obs.) and this disturbance may have contributed to the local decline.

Although possessing areas of Agonis-dominated habitat and having an unconfirmed sighting, the Hoffman site also showed no evidence of quokka presence (scats/runways)

133 Chapter 4 Population dynamics

and none were trapped. If the sighting was of a quokka then this individual appears to have been traveling through and this site is now presumed to be extinct also.

The small population estimates at the Chandler and Victor Road sites also suggest a significant threat of localised extinction. Even the two largest populations estimated in this study (Kesners and Hadfield) are well below the minimum viable population size

(50) often speculated as delineating genetic problems such as inbreeding depression and loss of heterozygosity (Caughley 1994). While loss of genetic diversity probably has occurred in quokkas, molecular data reveals there was a more continuous distribution historically along with a larger population size (Sinclair 2001). The success of the quokka population on Rottnest Island, which has a much lower heterozygosity than mainland populations, suggest there has been no negative impact (Sinclair 2001) attributable to inbreeding depression. This may be explained by the fact that the effective population size of a metapopulation can be substantially smaller than for the same number of individuals in one continuous population (Gilpin 1991).

The low level of captures per 100 trap nights highlights the trap shyness of quokkas.

Although other explanations for low recapture rates such as high mortality rates, dispersal, etc. have been suggested for other species (Krebs et al. 1994), radio telemetry of movements and survival suggests these are unlikely to be operating for quokkas

(Chapters 5 and 7). Trapping around Dwellingup in the early 1970s also yielded a low capture rate - less than 0.05 quokkas per 100 trap nights (calculated from Schmidt 1973;

Schmidt and Mason 1973). This is similar to that observed at the Rosella site in this study. Conversely, a larger-scale study at the same time near Dwellingup attained between 2.2 and 11.5 new captures per 100 trap nights (Christensen and Kimber 1975).

134 Chapter 4 Population dynamics

Potential differences in calculating captures per 100 trap nights mean that results are unlikely to imply a population change.

Both captures per 100 trap nights and KTBA estimates reinforce the scarcity and difficulty in capturing quokkas today and reflect the C-J-S estimates. The absence of a correlation between the C-J-S estimates and the other two methods is probably due to the large variation in recapture rates between each population. This means that the extensive time and effort necessary to derive open population estimates are essential in the future unless some other form of rapid abundance estimate is found.

At Barker’s Swamp, on Rottnest Island, the population density of quokkas between

December 1967 and December 1969 was 16.7 quokkas per hectare (Kitchener 1973).

This was far greater than was recorded for the whole island where density was reported to be between 0.8 and 2.5 per hectare (Main and Yadav 1971). This exceptionally high density around Barker’s Swamp was attributed to recruitment of individuals to the valuable water resource (Kitchener 1973). The Main & Yadav (1971) estimate was similar to that observed for the mainland populations in this study (Table 3).

Kitchener (1973) determined that an area of between 120 and 360 hectares was necessary to support a sustainable quokka population, however in resource-rich areas, such as around water sources, he revised this area to as little as 18 hectares. Although likely to be considered resource-rich areas, none of the swamps studied in the northern jarrah forest covered an area as large as 18 hectares (Table 1). Considering the past reports of high quokka density on the mainland (e.g. White 1952; de Tores et al. In prep.) and the densities reached around resources on Rottnest Island, it seems that the quokka populations investigated in this present study are well below the potential carrying capacity of each site.

135 Chapter 4 Population dynamics

Quokkas and their response to fox control

Perhaps the biggest concern of these small population sizes is that there is no evidence to suggest the populations have shown a response to on-going fox control.

Fox control began in the northern jarrah forest in June 1994 and additional high- intensity baiting commenced at quokkas sites between October 1997 and March 1999

(Table 1). This fox baiting has been shown to decrease fox densities (de Tores 1999).

In other parts of Western Australia, fox control has been shown to result in dramatic increases in native fauna abundance (see references in Chapter 3 and de Tores 1994; de

Tores 1999). For example, fox control was initiated at two reserves in the Western

Australian wheatbelt in 1982 and within two years the low density population of black- footed rock-wallabies (Petrogale lateralis) had almost doubled (Kinnear et al. 1988).

After eight years the populations had increased by up to 400% (Kinnear et al. 1998).

Considering the similarity in fecundity (Sharman and Maynes 1995) and initial population size between quokkas and black-footed rock-wallabies, it could be anticipated that, with the easing of predation pressure quokka numbers would increase in a similar fashion to that observed in the rock-wallabies.

Quokkas may have an even greater propensity to increase than the rock-wallabies considering fox control in the jarrah forest is likely to have led to population increases in other critical weight range mammals (e.g. southern brown bandicoots, chuditch) such that they are no longer considered threatened with extinction. Other fauna, such as the , are also likely to have increased in abundance. With these species increasing in abundance relative to the quokka it is anticipated that the fox

136 Chapter 4 Population dynamics

would functionally shift its prey species preferences to these species which would thereby exert even less predation pressure on quokkas.

Evidence from annual rates of change of quokkas reintroduced to a site south of

Perth (Short et al. 1992) indicates dispensatory predation by foxes and cats which was considered as a density-dependent (Type II) functional response (Sinclair et al. 1998).

With such a response, it was concluded that any reduction in predator density should allow quokka densities to increase, particularly for populations larger than 50 individuals (Sinclair et al. 1998). Such populations are not available in the northern jarrah forest and so a larger predator reduction (possibly to < 25% of current predator densities) may be necessary to allow quokka population increases.

The precautionary principle meant that leaving quokka sites unbaited was not justified considering the status of the species. This meant that only quokka sites identified following the initiation of fox control were left unbaited. As only one of these was found to support quokkas the experimental design was such that direct statistical comparison between treatment and control sites was impossible.

Consequently, while no firm conclusions can be made regarding the quokkas’ need for, and the effectiveness of, fox control, the absence of obvious detectable population increases at the treatment sites suggests several possible alternatives. Firstly it suggests that fox predation is not limiting quokka population growth. Population growth was not tested due to the short term nature of this study relative to the longevity of the quokka.

Secondly it may be that the level of fox control is insufficient to enable quokka populations to respond. Considering the earlier discussion, these two alternatives seem unlikely. Finally there may still be other factors inhibiting quokka population growth

137 Chapter 4 Population dynamics

such as the apparent preference of quokkas for a mosaic of early and late seral stage habitat (Christensen and Kimber 1975).

If the factors limiting population growth and/or contributing to the decline of the quokka are removed, the low population numbers may not necessarily limit population growth (Kinnear et al. 1998). Further, a minimum population size of 50 may not be critical, as the Nangeen Hill rock-wallaby population increased to over 100 individuals from a founder population of 18 in less than eight years from the initiation of fox control (Kinnear et al. 1998). Similarly ten to 20 pairs have been suggested as adequate founder populations for macropodid reintroductions (Johnson et al. 1989), while feral macropodid populations have established from founder populations as small as two

(Maynes 1989). Such populations are still available at some sites in the northern jarrah forest.

Population composition

The age-class structure of quokkas on the mainland (50.5% adult, 24.3% juvenile, and 25.2% pouch young) was very similar to that observed by Dunnet (1963) on

Rottnest Island (49.3% adult, 22.4% juvenile and 28.7% pouch young). This suggests that mortality rates through all age classes are similar between the two regions, however considering the large number of deaths over summer on Rottnest Island (Main et al.

1959) this is of concern for the mainland populations whose mortality sources have previously only been speculated upon (Maxwell et al. 1996)..

The differences among sites in age-class breakdown (Figure 3) are likely to reflect future population growth and therefore the ‘health’ of a population. A preponderance of breeding adults suggests catastrophic events may have removed old and young animals

138 Chapter 4 Population dynamics

from the population while an abundance of sub-adults suggests a high reproductive output and therefore population growth (p 254 in Caughley and Sinclair 1994). With the highest proportion of juveniles and pouch young, Victor Road appears to be the site with the greatest propensity to increase, followed closely by Hadfield. The Kesners site has very few pouched young and may reflect an approach to carrying capacity and equilibrium or a recent catastrophic event. The Chandler site has a much lower proportion of sub-adults and this may indicate it is a population in decline. The lone male that comprises the Rosella Road population appears likely to be the final barrier to localised extinction there. Factors inhibiting the growth of quokka populations appear to be acting in increasing order of severity from Victor Road and Hadfield to Kesners,

Chandler and then Rosella Road. As the Victor Road site is unbaited it reinforces the suggestion that habitat features may be as important as predation in inhibiting quokka population growth.

Yet the overall level of recruitment following weaning is low (40%) and increases from Hadfield (22%) to Victor Road then Chandler and finally Kesners (89%) with the highest recruitment rate. This order is almost opposite that of the overall population composition (Figure 3) and highlights the likely source of the apparent population stability or decline.

The Rottnest Island quokka population has similar numbers of males and females overall (Dunnet 1963) as has been observed on the mainland. While sex ratios throughout each age class appear to be at parity in the northern jarrah forest, they vary on Rottnest Island. One study there found male pouch young to be more common than females at birth, while these ratios were equal once the young left the pouch (Main et al.

1959). Another found females significantly outnumbered males (1.12 : 1.00) at less

139 Chapter 4 Population dynamics

than 100 days old, with this ratio becoming increasingly female-biased over 100 days

(1.56 : 1.00) (Shield 1962). Dead pouch young were biased toward males (0.67 : 1.00)

(Shield 1962). A slightly more recent study of one thousand joeys also found females significantly outnumbered males (1.21 : 1.00), although 204 pouch young less than 40 days old showed slightly more males than females (Shield 1968). On Rottnest Island, therefore, male quokkas outnumber females at birth (primary sex ratio) but a higher male pouch mortality leads to a female bias by the time they leave the pouch (secondary sex ratio). The parity observed on the mainland may mean a greater sample size is required to discern similar trends in pouch young sex ratios or that this study mixed primary and secondary sex ratios by sexing young when they were captured rather than born or weaned. Alternatively differing parental investment strategies (e.g. Trivers and

Willard 1973; Silk 1984; Clutton-Brock and Iason 1986) may explain the variation in sex ratios between the mainland and Rottnest Island.

Morphology and condition

Quokkas have long been known as a sexually dimorphic species (Dunnet 1962;

Sinclair 1998), however the ratio of adult male to adult female body mass found in this study (1.4:1) was larger than observed on Rottnest Island (1.2:1) (Dunnet 1962) and for a smaller sample size from the mainland (1.3:1) (Sinclair 1999). These differences may be an artifact of the increased sample size in this present study compared to earlier studies on the mainland, the seasonal variation quokkas exhibit in body mass and, most importantly, the inclusion of females less than 2.5 kg in mass as adults in this present study.

140 Chapter 4 Population dynamics

Although there was no significant differences in body mass between the sites, it is interesting to note that the unbaited Victor Road site had the lowest body mass overall and for males out of all sites. Increased predation risk causes similar decreases in body weight for field voles (Microtus agrestis) (Carlsen et al. 1999), however the absence of a replicated control in this present study precludes any firm conclusions as to its cause in quokkas.

The seasonal variation in body mass exhibited by quokkas is not unusual in macropodids (Jarman 1989). Similarly the latitudinal cline in body mass observed in this study, while non-significant, was also found across a broader area by Sinclair

(1999). This cline was attributed to an historically continuous distribution (Sinclair

1999) which suggests that all mainland quokka populations originally formed a single metapopulation. Genetic data generally confirm this assertion with molecular variation reflecting recent phylogenetic history consistent with a historically contiguous distribution (Sinclair 1999) although there may have been some historical isolation between the far northern and far southern populations due to their wide separation

(Sinclair 2001).

Despite being widely used (e.g. Kerle 1998; Richards et al. 2001) the validity of condition indices has long been questioned particularly with regard to the use of the cube law of scaling and the repeatability of measurements (see Krebs and Singleton

1993). While the concerns of these authors are valid, the calculation of regression equations of skeletal size against body mass to overcome these concerns for a pooled quokka population of only 62 adults is considered unjustified and likely to result in larger errors. Consequently, the comparison of an observed condition value with a predicted condition value to derive the condition index was not conducted in this study.

141 Chapter 4 Population dynamics

The use of the cube law of scaling has been used previously in measuring the condition of quokkas (Miller and Bradshaw 1979; Bakker and Main 1980), common brushtail possums (Short and Turner 1994; Kerle 1998), golden bandicoots (Isoodon auratus) and burrowing bettongs (Bettongia leseur) (Short and Turner 1994). The absence of significant seasonal variation in condition index may also have been due to the conservative nature of this measure caused by the inappropriate use of the cube root in the calculation (see Krebs and Singleton 1993). This was acknowledged by Bakker

& Main (1980) who determined that a 15% increase in the cubed root of body weight resulted in an increase in condition index by a single unit at the same pes length. The problem of repeatability may be reduced in this study by the larger body size of the quokka compared to the house mice (Mus domesticus) used by Krebs & Singleton

(1993) and by all measurements being taken by the same researcher.

A questioned assumption of condition indices is that higher indices relate to better survivorship, reproductive output, etc. (Krebs and Singleton 1993). Studies have found such a relationship in several species (e.g. Gerhart et al. 1996) including macropodids

(Moss and Croft 1999). Condition indices derived for quokkas on Rottnest Island were at their lowest at the end of a long, dry summer when seasonal mortality was observed

(Miller and Bradshaw 1979; Bakker and Main 1980) and reproductive performance declines with poorer condition (Holsworth 1964; Kitchener 1972). Quokka condition indices also varied with the health of animals as determined by various physiological parameters (Miller and Bradshaw 1979). Our study supports such conclusions as the decline in female condition index is at the same time as reproductive output is reduced.

On Rottnest Island the condition of male quokkas is significantly higher than that of females throughout the year (Miller and Bradshaw 1979) whereas there is no significant

142 Chapter 4 Population dynamics

sexual differences in condition on the mainland. This anomaly may be due to the differences in sexual dimorphism between the island and mainland populations.

Studies to confirm the relationship between the condition index value and tail fat storage were not conducted due to ethical considerations however there is evidence to suggest that quokkas store fat in their tails similar to other macropodids (Moss and

Croft 1999). Bakker & Main (1980) discounted tail length as a condition measure due to the difficulty in locating the “insertion of the tail accurately in fat animals”. This suggests quokkas also possess caudal fat deposits. This was confirmed by Sinclair

(1999) who found captive quokkas on a regular, high nutrient diet had larger tail diameters than wild animals which she attributed to tail fat storage.

While the absence of seasonal fluctuations in condition was similar to that observed in earlier studies on mainland quokkas (Sadlier 1959; in Hart et al. 1986), the seasonal variation in tail circumference observed in this study again suggests that quokkas store fat in their tails (e.g. Bakker and Main 1980; Sinclair 1999). This suggests there is some variation in condition of quokkas on the mainland despite not being significantly shown in the more conservative condition index. Male fat storage peaked in summer at the end of the high rainfall months. Conversely, female tail circumference remained relatively constant throughout the year but with a decline in spring. This decline in fat content following winter may be caused by the costs of maintaining body heat as occurs in mountain brushtail possums (Trichosurus caninus) (Viggers et al. 1998). Not surprisingly fat is also deposited around the tail after body mass is probably regained and is used to sustain the animals over the drier periods from late summer to autumn.

The more rapid growth of tail circumference over this period in males than females

(Figure 6) suggests that the higher costs of reproduction in females may be reducing

143 Chapter 4 Population dynamics

their ability to store fat and that they only do so following the period of reduced breeding over summer (Figure 7). Rottnest Island quokkas possess larger fat stores than mainland animals due to their regular periods of starvation encountered by the insular population (Hart et al. 1986).

Storr (1964a) found that in and around the swamps of the northern jarrah forest there was at least one species of putting on fresh growth at any particular time throughout the year. The lower water content of these plants compared to those on

Rottnest Island meant that protein was able to be more readily absorbed by mainland quokkas which compensated for the relatively low protein content of the plants (Storr

1964a). Consequently, Storr (1964a) attributed the lack of seasonal fluctuation in condition, found by Sadlier (1959 in Storr 1964a) in mainland populations, to the relatively stable availability of nutrients. The condition index used in this present study also revealed no differences in seasonal condition, however the tail circumference – another indicator of condition – did. Considering the continuity of fresh plant growth throughout the year, it seems most likely that different reproductive costs rather than differences in nutrient intake cause the seasonal variation in tail circumference observed between male and female quokkas. This is probably exacerbated by the fact that female quokkas have higher daily water requirements than males (Main and Yadav 1971).

The difference in tail circumference between the sites is interesting with Chandler and Kesners quokkas having larger tail circumferences than quokkas at Victor Road and

Hadfield. The habitat available at each site undoubtedly differs (due to fire history, topography, climate, soils, etc.) and therefore may suit quokkas differently, allowing animals at some sites to store more fat than at others. Interestingly, the Victor Road and

Hadfield sites have been burnt more recently than the other sites (Chapter 6). This

144 Chapter 4 Population dynamics

again lends weight to the suggestion that habitat factors may be as important to quokkas as the easing of predation pressure, particularly since there was no correlation between population density and tail circumference.

Reproduction

The age of first breeding females in this study (15 months) was much older than that observed in captive quokkas (8-9 months) (Shield 1968) but younger than Rottnest

Island (18 – 30 months) (Shield 1964). Captive males start breeding at 12-13 months

(Shield 1968) but on Rottnest Island at over two years (Shield 1964). There was no accurate data from this study on the initiation of breeding in males.

It has been suggested that mammalian females, being the most significant reproductive unit for population persistence, are under the greatest selective pressure to attain early sexual maturity (Main and Bakker 1981). This is likely to be manifested by a smaller size compared to males (Main and Bakker 1981). Since a single male can fertilise numerous females, selection for early maturity in males is likely to be less intense (Main and Bakker 1981). Furthermore the greatest selection for early maturity and the associated increases in sexual dimorphism is thought to occur in species inhabiting unpredictable and highly variable environments (Main and Bakker 1981).

Yet the quokka occurs across a range of environments, from the seasonally arid Rottnest population to the more predictable, mesic mainland populations (Chapter 3). According to the above theory we would expect the quokka to exhibit earlier sexual maturity on

Rottnest Island (arid) compared to the mainland (mesic). Although this theory is true across species boundaries (Main and Bakker 1981) it appears not to hold true for individual species as sexual maturity is reached at older ages on Rottnest Island.

145 Chapter 4 Population dynamics

The data presented here give no indication as to the of quokkas on the mainland, however Rottnest Island quokkas are promiscuous (Dunnet 1962). When two or more males were placed in captivity together one became dominant and eventually killed or mutilated the others (Shield 1968). These factors indicate that the mating system of quokkas may be some form of polygamy. This may be revealed by genetic analysis.

Niven (1971) found a negative relationship between the percentage of females with young and population size during a modeling study using field data from Rottnest

Island. Such a relationship was not found in this present study and this reinforces the earlier suggestion that quokkas are below carrying capacity at each of the known mainland sites.

Although female quokkas on Rottnest Island have occasionally been observed with twins (Shield 1964) this was not found in the northern jarrah forest. With 185 to 200 days from birth to weaning in quokkas (Sharman 1955a, 1955b; Shield 1968), it is biologically possible for a female to produce 1.825 to two young per year. Based on an initiation of breeding at 18 months (pers. obs.) (Shield 1964, 1968) and a longevity of at least 10 years (Holsworth 1964; Shield 1968), and continuous breeding as observed in this study, female quokkas are able produce up to 17 young over a lifetime. This is possible on the mainland but the seasonal anestrous on Rottnest Island means that breeding generally only occurs once a year and begins at two and a half years of age

(Shield 1964). So this number is more like seven or eight over a lifetime for the island population. The higher fecundity on the mainland means there is a high capacity to increase population size. This was not observed in this study despite the relaxation of predation pressure following the initiation of fox control.

146 Chapter 4 Population dynamics

Female quokkas captured seasonally in the northern jarrah forest averaged 2.2 births per year which is more than we have just determined is mathematically possible. This phenomenon arises due to the death of pouch young prior to weaning and their subsequent replacement from a blastocyst. This may be the cause of the absence of population growth following fox control. A high rate of pouch young mortality is also suggested by the percentage of juveniles recruited into the population. Less than one in two pouch young survive to independence which also highlights the importance of pouch young mortality in explaining the continued low population levels.

Quokkas on Rottnest and Bald Islands breed seasonally with the majority of births occurring between February and April (Figure 7) (Shield 1964). This seasonal pattern of breeding was attributed to the seasonal variation in climate causing starvation and inducing anestrous (Shield 1964). This pattern is thought to have evolved to provide the young with high nutrient forage at weaning following the winter rains (Shield and

Woolley 1963). The observation of breeding throughout the year in the northern jarrah forest quokka populations is consistent with the observations of year-round breeding by captive mainland animals (Shield 1964).

The decline in births over summer (Figure 7) is interesting. This corresponds to a period when female body condition is lowest (Figure 5). Female quokkas have much higher daily water requirements than males (9.5% of body weight for females compared to 7.6% for males) (Main and Yadav 1971) and summer is generally a period of very low precipitation (Chapter 2). Additionally, the low protein content of plants around mainland swamps means that water intake must be higher to allow sufficient dry matter to be consumed to maintain condition (Storr 1964a). The reduced water availability at this time is therefore, likely to cause the decline in births. Similar responses in

147 Chapter 4 Population dynamics

reproduction occur in quokkas on Rottnest Island suffering decreased nutrient availability (Holsworth 1964; Kitchener 1972). Furthermore, years of above average rainfall may result in increased births and, conversely, drought years may cause a reduction.

Breeding season differences have been observed in other macropods also (Merchant and Calaby 1981; Robertshaw and Harden 1986). Predation pressure by dingoes (Canis lupus dingo) on swamp wallabies (Wallabia bicolor) caused the wallabies to breed throughout the year whereas nearby conspecifics at lower predation risk bred seasonally

(Robertshaw and Harden 1986). While quokkas on the mainland are under much more predation pressure than those on Rottnest, the fact that annual breeding in Rottnest quokkas can be initiated in captivity with ample food (Shield 1964, 1968) suggests nutrition is the cause of these differences rather than predation. This is reinforced by the group of quokkas that inhabit the rubbish tip on Rottnest Island that breed throughout the year due to the year-round abundance of food compared to their seasonally-starved neighbours (Shield 1964, 1968). It is interesting to ponder the likely breeding pattern of local quokka populations at the original limits of the species distribution where rainfall, and subsequently resources, would be limited in a similar fashion as Rottnest Island.

The average mass of breeding females in the northern jarrah forest (mean = 2.77 kg) was similar to that observed on Rottnest Island (mean = 2.75 kg) (Shield 1968). The percentage decline in body mass observed on the mainland, however was far smaller than the 25% decline observed on Rottnest Island from late summer to autumn (Shield

1964). This is driven by the arid conditions on Rottnest over summer and autumn

(Shield 1964). The decline in births during summer in the northern jarrah forest occurs

148 Chapter 4 Population dynamics

at a time when female body mass and condition are at their lowest level but fat storage is highest (Figure 6). By the end of summer the swamps almost entirely dry up leaving few, if any, areas of available water (pers. obs.; M. Dillon pers. comm.). As has been suggested previously, the smaller body mass of females may mean that, in the absence of reproduction, female quokkas can obtain sufficient nutrients from plants with lower water content to deposit fat.

Conclusion

The plight of the quokka on the mainland can be considered from a metapopulation dynamics perspective. There are very few examples in the literature of medium-sized and large mammals occurring naturally as metapopulations (see review by Hastings and

Harrison 1994; Elmhagen and Angerbjorn 2001). Nonetheless, it seems likely that the quokka once naturally existed as a classic metapopulation (Hanski and Gilpin 1991), being restricted to abundant but discrete habitat patches wherein breeding took place and ecological processes, including colonisation and extinction, were uncorrelated while functioning on a local and regional scale (Hanski 1999). This patchiness was not caused by human-induced habitat fragmentation, but rather arose through the occurrence of naturally-occurring islands of the quokkas preferred habitat which increase in their patchiness due to the quokkas preference for a mosaic of seral stages

(Christensen and Kimber 1975). More recently, human-induced fragmentation has further fragmented habitat which has increased the inter-patch distances between populations. Water voles (Arvicola terrestris) in Scotland also exist naturally as classic metapopulations and this structure arises through the similar natural patchiness of their preferred habitat (Aars et al. 2001). Anecdotes from areas south of the northern jarrah

149 Chapter 4 Population dynamics

forest suggest that the turnover of local populations through colonisation and extinction are still occurring there (G. Liddelow pers. comm.).

The northern jarrah forest quokka metapopulation was subsequently threatened through predation and habitat alteration (de Tores et al. In prep.) and turnover of local populations resulted in a negative feedback relationship as the species declined and local populations became extinct. This resulted in a non-equilibrium metapopulation in which local extinction dominated as part of the species overall decline (Harrison 1991).

The collapse of the quokka metapopulation may have been exacerbated if extinction among local populations was correlated (Harrison 1991) or regionally stochastic

(Hanski and Gilpin 1991) as may have occurred with the intense, surplus killing by foxes of a predator-naïve prey immediately after its arrival in a particular area (Short et al. 2002). Predation is one factor that causes localised extinctions to be correlated and this thereby increases the likelihood of metapopulation extinction (Aars et al. 2001).

The northern jarrah forest quokka metapopulation structure today is characterised by excessively large inter-patch distances, such that the metapopulation persistence time

(Hanski and Gilpin 1991) appears very short with few populations remaining and those that remain reduced to very low numbers. Further, the likelihood of localised extinction appears high as there is little potential for a ‘rescue effect’ from immigrants (Brown and

Kodric-Brown 1977) as distances of up to 40 km separate some extant populations.

Such non-equilibrium metapopulations are generally bound for regional extinction

(Harrison 1991) as the overall metapopulation collapses. Stochastic models illustrate that metapopulations consisting of a small number of local populations (as in the northern jarrah forest), that each have with a high risk of extinction (exacerbated in the quokka by the small population sizes), are unlikely to survive for long (Hanski et al.

150 Chapter 4 Population dynamics

1996) as an alternative stable equilibria of localised extinction is attained (Hastings and

Harrison 1994).

Given local extinction has a greater bearing on metapopulation persistence time than colonisation rate (Etienne and Heesterbeek 2001); and the reduced genetic diversity of the Rottnest Island population (Sinclair 2001); the approach we recommend is to manage known populations of quokkas in the northern jarrah forest as separate entities rather than a metapopulation (Hastings and Harrison 1994) until the metapopulation structure is restored. This is also recommended in preference to augmenting these populations with translocated animals from Rottnest Island given the genetic differences between island and mainland populations (Alacs 2001; Sinclair

2001).

Managing populations as separate entities involves identifying all sites containing quokkas, allowing quokka density at each location to increase through natural in situ processes, augmented with feral predator control and, where necessary, also augmented with habitat manipulation to increasing the number of these sites and return their connectivity to a functional state. Clearly for this to occur, all threatening processes must be identified and removed to the extent that each extant population can act as the source for natural colonisation to adjacent habitat patches. Monitoring is also a critical component and such programmes should be designed to ensure changes in population size and structure can be detected.

151 Chapter 5 Home range and movements

Home range and movements of the quokka, Setonix

brachyurus (Macropodidae: Marsupialia) in the northern

jarrah forest of Australia.

Abstract. The home range and movements of the quokka (Setonix brachyurus Quoy

& Gaimard 1830), a vulnerable macropodid marsupial, were investigated using radio telemetry. Three home range estimation methods were used: minimum convex polygons (MCP), harmonic mean and kernel estimators. Fifty-eight quokkas from five sites were radio-collared and monitoring yielded 2,060 fix locations over almost three years. Mean (± S.E.) home range sizes were 6.39 ± 0.77 ha and core ranges averaged

1.21 ± 0.12 ha. Male home ranges averaged 6.92 ± 1.16 ha and females averaged 5.91

± 1.07 ha although these differences were not significant when corrected for body size.

Nocturnal ranges (5.48 ± 0.90 ha) were not significantly larger than diurnal ranges (3.51

± 0.40 ha) despite nocturnal departures from the dense, swampy habitat used by quokkas. There was a negative relationship between home range size and population density in both entire and core home ranges with females exhibiting this relationship far more strongly than males. Home range size peaked in autumn at the end of the dry season and was smallest in spring at the end of the wetter months, probably reflecting variation in the availability of fresh forage. Range centres remained fairly constant up and downstream along the swamp but shifted to the edge of swamps in winter, as the swamps became inundated following rain, and toward the centre in autumn as the swamps dried out. The lack of dispersal found in this study is probably caused by quokka populations existing below carrying capacity. Predation may be suppressing

153 Chapter 5 Home range and movements

population booms which are likely to drive dispersal and is thereby disrupting the metapopulation dynamics. Without dispersal to recolonise or rescue unpopulated patches, this study reinforces the theory that the original quokka metapopulation has collapsed and ameliatory measures are required.

Introduction

For collapsed or collapsing metapopulations, knowledge of the species’ home range size and dispersal distance is important in determining the likelihood of restoring the metapopulation structure and identifying potential management methods. If movements are adequate enough to facilitate population mixing between patches then little management may be required. If movements are too small to allow such mixing then options for restoring connectivity need investigation either through population augmentation or habitat management.

The quokka (Setonix brachyurus Quoy & Gaimard 1830) population in the south- west of Australia is thought to represent the terminal remnants of a once classical metapopulation (as defined by Hanski and Gilpin 1991) that has now collapsed into a non-equilibrium state (Chapter 4) following an extensive decline in the 1930’s (Chapter

3). The quokka is a small, macropodid marsupial that is endemic to this region

(Kitchener 1995). The species is listed as vulnerable on state and federal threatened species legislation (Chapter 3). Today on the Australian mainland local quokka populations are isolated, widely scattered and rarely exceed thirty individuals (Chapter

4). In the jarrah (Eucalyptus marginata) forest of the mainland, quokkas are restricted to the swampy, densely vegetated, upper reaches of creek systems that are dominated by

154 Chapter 5 Home range and movements

the tall shrub Agonis linearifolia (Chapters 3, 4 & 6) (White 1952; Christensen et al.

1985).

There is no published study of the home range size of quokkas on the mainland, however the movements of those on Rottnest Island have been investigated. Home ranges estimated by spool and line using the minimum convex polygon method averaged 0.089 ha in May and 0.036 ha in November (Kitchener 1973). From these estimates, the average annual home range of Rottnest quokkas was calculated as 0.13 ha

(Kitchener 1973). Trapping over ten years on Rottnest Island identified overlapping individual home ranges of six hectares and 50 to 100 individuals inhabiting non- overlapping, group territories of up to 15 hectares (Holsworth 1964). Trapping indicated that very few individuals departed their natal group territory (Holsworth

1964), however radio telemetry discounted this, showing regular movement between groups which was explained by differences in resting and foraging ranges (Nicholls

1971).

Despite a degree of knowledge about the home ranges and movements of the quokka on Rottnest Island, extrapolation to the mainland populations is unfounded due to the vast environmental and ecological differences between the two areas (Chapter 4)

(Main, Shield and Waring 1959; de Tores et al. In prep.). Knowledge of quokka movements is critical in identifying the functional status of the metapopulation following the species decline and ultimately to determine what is required to restore the metapopulation structure. As such, this study aimed to determine movement sizes of quokkas from five local populations on the Australia mainland and concurrently investigate home range size and compare these across space and time for each sex.

155 Chapter 5 Home range and movements

Methods

Collar attachment

Quokkas were trapped at five sites (Chandler, Rosella Road, Kesners, Hadfield and

Victor Road) between spring 1998 and summer 2000/2001 according to the methods described in Chapter 4. Once captured and sedated, the majority of quokkas were fitted with radio collars (Biotrack, Institute of Terrestrial Ecology, Wareham, UK). Three types of collar were used: a brass loop aerial (82%), a whip aerial (8%) and a breakaway variety (10%). The brass loop aerial collars utilised the brass loop of the collar as both the attachment device and the aerial to close the circuit to the transmitter. The whip aerial variety had the brass loop collar but used a whip aerial from the terminal and, for this collar variety, the circuit was not broken if the brass loop broke. The breakaway collars had whip aerials and, as they were designed for juveniles, a perishable rubber collar that was intended to break as the animal grew.

Collars averaged 28 grams when fitted and the maximum collar weight was 37 grams. The maximum proportion of collar weight to body mass was 2.3% and occurred when a 670 gram ‘young at foot’ was fitted with a 15.25 gram breakaway collar. The majority of individuals were fitted with collars that were less than 1% of their body mass which is far less than is generally considered as energetically costly (Berteaux et al. 1996).

Radio telemetry and triangulation

The density of the vegetation inhabited by quokkas meant that directly observing individuals was impossible without entering the swamp which, in turn, was considered likely to alter their behaviour. (After spending the majority of 2½ years in and around

156 Chapter 5 Home range and movements

quokka swamps I only observed three individuals that were not within a trap – I almost ran over two while arriving to check traps at the Kesners site and the other approached to within five metres of me while conducting nocturnal telemetry at Hadfield).

Movements within the swamps were also minimised because of the perceived threat of creating access pathways into the quokka’s refuge for predators, particularly the red fox

(Vulpes vulpes) (as discussed by Barker, Main and Sadlier 1957).

For these reasons triangulation was used to locate radio collared individuals.

Compass bearings were taken from at least three telemetry stations in rapid succession

(generally less than 5 minutes). Telemetry stations were established using a differential global position system (GPS) (Magellan) prior to the disabling of selective availability

(Hulbert and French 2001) that still yielded centimetre or sub-metre accuracy. While this minimised variation, other sources of telemetry variation that are rarely mentioned in the literature were also identified. Compass bearings have an origin at magnetic north, however reference maps are plotted according to grid north. This discrepancy was accounted for by the addition of 6° to all compass bearings after such a correction was calculated by comparing the compass bearing between two GPS surveyed points with the bearing calculated from map coordinates (Australian Map Grid 1984 datum).

Proximity to vehicles was also found to affect compass bearings slightly and so measurements were taken from more than ten metres away from vehicles. Efforts were also made to place telemetry stations away from other sources of electromagnetic variation such as power lines (Parker et al. 1996).

Autocorrelation of locations occurs when the position of an individual is correlated with or dependent upon its previous location and often occurs when fixes are taken over a short period of time (Swihart and Slade 1985). To avoid underestimating home ranges

157 Chapter 5 Home range and movements

caused by the autocorrelation of fixes (Swihart and Slade 1985; Rooney, Wolfe and

Hayden 1998; Otis and White 1999), diurnal and nocturnal locations were taken no more than once per day.

Once bearings to the collared individual were taken, they were entered into the

Locate II computer program (Nams 1990). This program uses a maximum likelihood estimator (MLE) to determine the most-likely location of an animal from three or more compass bearings by weighting bearings nearer to the approximated position of that individual higher than those more distant (Nams 1990). The MLE calculated the deviation of each bearing from the estimated location to get an error angle from which a

95% confidence or error ellipse was determined (Nams 1990). Only fixes with error ellipses of less than one hectare were used in the analyses. This was determined after a preliminary study found such error ellipses were within 15 metres of the true locations when the transmitter position was later surveyed with differential GPS. The locations where quokkas were trapped were also surveyed using differential GPS and were used as nocturnal fixes after incidental observations of traps found no quokkas were captured during the day. This did not lead to any autocorrelated fixes as telemetry was not conducted on trapped animals.

Home range estimation and statistical analyses

Three methods were used to determine home range size using the Ranges V computer program (Kenward and Hodder 1992). All three methods were used after a review of home range studies recommended a combination of techniques (Harris et al.

1990) and to allow comparison with other home range studies that only use one estimation method. The minimum convex polygon method (MCP) involved fitting the

158 Chapter 5 Home range and movements

smallest possible polygon with external angles that exceed 180o to the locational fixes

(Harris et al. 1990). This method has been widely used in the past (e.g. Quin et al.

1992; Goldingay and Kavanagh 1993; Sander, Short and Turner 1997; Newell 1999;

Hanski et al. 2000; Jackson 2000; Mallick, Driessen and Hocking 2000; Meek and

Saunders 2000) and was used here to allow comparison with these and other studies. It is acknowledged that the area and shape of the home range estimated by this method is heavily biased by outlying fixes (Harris et al. 1990).

The harmonic mean method (Dixon and Chapman 1980) estimates the fix density distribution at intersections of an arbitrary grid (Kenward and Hodder 1992). The fix density distribution is equivalent to the probability of encountering a radio-collared individual (Kenward and Hodder 1992). Contours were fitted to actual fixes and fixes were centred between grid intersections. Grid size was left at the Ranges V default setting of 40 x 40. The choice of these three options in harmonic mean analysis may cause variation in home range size estimation (Kenward and Hodder 1992). The

Gaussian estimator used in kernel methods avoids these problems (Kenward and

Hodder 1992) and so was used. Kernel estimation is mathematically robust and produces more consistent results than harmonic mean contouring (Kenward and Hodder

1992). Fix densities at the 40 x 40 grid intersections were derived using a bivariate normal kernel estimator which is less grid-dependent than the harmonic mean function

(Kenward and Hodder 1992). The default smoothing factor of the standard deviation divided by the number of fixes was used (Kenward and Hodder 1992). Core weighting of the kernels was used after the adaptive kernel was found to overly enlarge the home range estimate.

159 Chapter 5 Home range and movements

Utilisation plots (Kenward and Hodder 1992) were used to identify the percentile contour line that conservatively identified the core home range of quokkas. These plot contours around the fixes every five percentiles ranging from the 100 percentile contour down to the 20 percentile contour. When the home range areas of these contours are plotted it often becomes apparent that a plateau in the decline is reached and this determines the core home range of a species (Kenward and Hodder 1992). For each estimation method the 95 percentile contour line was subsequently used to remove outlying locational fixes. This meant that the most distant fix from the range centre was removed after the addition of every 19 fixes. The 95 percentile estimates were classed as an animals’ overall home range in this study and incorporates both diurnal and nocturnal ranges.

Incremental area plots were conducted to determine the number of fixes needed to accurately define a home range for each estimation technique (Kenward and Hodder

1992). Incremental area plots are created by drawing an outline around the first three fixes and then adding successive fixes and drawing the outline again until all fixes in the range have been added (Kenward and Hodder 1992). Incremental area plots are used to determine the number of fixes required to estimate an entire home range. When individual ranges were broken up (such as diurnal/nocturnal or seasonal) then the number of fixes attained were greatly reduced. Consequently, individual home ranges with 10 fixes or more were used to compare diurnal and nocturnal ranges and seasonal ranges (as recommended by Harris et al. 1990). Individual animals that were thought to have dispersed were excluded from analyses.

Home range overlap analysis was not included in this thesis but will be published in the future. Plots of overlapping home ranges are presented in Appendix B.

160 Chapter 5 Home range and movements

The Gaussian kernel estimate of the centre of each home range (Worton 1989) was used to investigate seasonal shifts in range centres in a similar fashion to that used to investigate seasonal bird movements (Griffioen and Clarke 2002). Once the seasonal range centre for each individual was determined the shift along the swamp or across the swamp between seasons was measured using the MapInfo Version 5.5 geographic information system. Movements across and along the swamps were then tested non- parametrically using the sign test (Zar 1996).

Parametric statistical tests used for other comparisons included t-tests, nested

ANOVA, repeated measures ANOVA, ANCOVA corrected for body mass and linear regression using the Statview computer program (SAS Institute 1992-1998). Population densities at each site and body mass for each animal used in linear regressions were taken from Chapter 4. Bonferroni correction was applied to post-hoc comparisons.

Data not conforming to the assumptions of these tests were log transformed. Results are shown as mean ± standard error.

161 Chapter 5 Home range and movements

a. Minimum Convex Polygon (MCP)

b. Harmonic mean

c. Kernel

Figure 1. Incremental area plots of all animals home ranges according to the (a) MCP, (b) harmonic mean and (c) kernel methods. The area (%) of the y-axis refers to the percentage of the total home range made up by a contour fitted using the number of fixes of the x-axis.

162 Chapter 5 Home range and movements

Figure 2. Utilisation plot of all home ranges determined by the kernel estimator. The smaller error bars and the plateau in the decline in home range size at 50% indicates that this is the percentage of location fixes that should be conservatively used to estimate the core range of quokkas.

Figure 3. Example of a utilisation plot on the kernel estimate of the home range of KF2 (adult female). Fixes are shown as an asterisk and the approximate area of the swamp is shaded. Each contour, from the 100 percentile (outer-most contour) to the 20 percentile in increments of 5%, are shown. The 95 percentile contour (overall home range) is bolded and the 50 percentile contour (core home range) is bolded and dashed.

163

Table 1. Mean overall and core home range sizes (ha) derived by each of the three estimation methods. Individuals with 30 or more fixes (MCP) and 40 or more fixes (harmonic mean and kernel) were used for overall, male, female and site estimates.

Minimum Convex Polygons ± S.E. (n) Harmonic mean ± S.E. (n) Kernel ± S.E. (n) Mean Overall Core Overall Core Overall Core

Overall 5.422 ± 0.513 (25) 0.947 ± 0.113 (25) 10.534 ± 1.388 (21) 1.176 ± 0.141 (21) 6.390 ± 0.773 (21) 1.211 ± 0.118 (21) – Male 6.385 ± 0.902 (11) 1.228 ± 0.188 (11) 11.216 ± 2.349 (10) 1.401 ± 0.215 (10) 6.922 ± 1.156 (10) 1.471 ± 0.179 (10) – Female 4.666 ± 0.524 (14) 0.709 ± 0.188 (14) 9.913 ± 1.663 (11) 0.992 ± 0.176 (11) 5.907 ± 1.068 (11) 0.974 ± 0.124 (11) Diurnal 2.841 ± 0.305 (37) 5.014 ± 0.620 (37) 3.506 ± 0.402 (37) – Male 3.039 ± 0.382 (20) 4.892 ± 0.722 (20) 3.601 ± 0.518 (20) – Female 2.709 ± 0.483 (17) 5.174 ± 1.104 (17) 3.374 ± 0.656 (17) Nocturnal 4.558 ± 0.452 (26) 6.857 ± 1.113 (26) 5.483 ± 0.895 (26) – Male 5.610 ± 0.920 (14) 7.684 ± 1.690 (14) 5.679 ± 1.348 (14) – Female 4.083 ± 0.492 (12) 5.892 ± 1.450 (12) 5.253 ± 1.194 (12) Chandler 4.763 ± 0.955 (3) 1.290 ± 0.232 (3) 8.673 ± 3.682 (2) 1.492 ± 0.666 (2) 6.783 ± 2.834 (2) 1.414 ± 0.358 (2) Hadfield 3.277 ± 0.398 (7) 0.90 ± 0.068 (7) 5.680 ± 1.702 (6) 0.598 ± 0.082 (6) 2.358 ± 0.231 (6) 0.658 ± 0.079 (6) Kesners 5.669 ± 0.809 (9) 1.074 ± 0.148 (9) 10.081 ± 1.422 (6) 1.210 ± 0.141 (6) 6.273 ± 0.919 (6) 1.458 ± 0.219 (6) Rosella 10.600 (1) 1.51 (1) 28.194 (1) 2.367 (1) 8.790 (1) 1.882 (1) Victor Road 7.885 ± 0.967 (6) 1.042 ± 0.325 (6) 13.537 ± 1.993 (6) 1.463 ± 0.304 (6) 10.010 ± 0.930 (6) 1.337 ± 0.197 (6) Autumn 3.479 ± 0.449 (32) 7.885 ± 0.967 (32) 7.885 ± 0.967 (32) Winter 2.866 ± 0.421 (18) 3.130 ± 0.385 (18) 3.557 ± 0.613 (18) Spring 1.621 ± 0.293 (17) 2.494 ± 0.915 (17) 2.098 ± 0.532 (17) Summer 1.789 ± 0.382 (19) 3.260 ± 1.472 (19) 2.854 ± 0.724 (19)

164

Table 2. Statistical test results on home ranges estimated by the kernel method. Male – female comparisons use data from home ranges calculated with more than 40 fixes, while the remainder use data from home ranges with more than 10 fixes. Small sample size precluded doing one large analysis. Comparison Test Interaction probability Test statistic Probability Overall home range

Females versus males ANCOVA corrected for population density 0.01 F1, 36 = 3.505 0.0689

Diurnal versus nocturnal Nested ANOVA corrected for sex differences 0.71 F1, 58 = 1.658 0.2029

Repeated measures ANOVA F1, 19 = 4.686 0.0406 a Sites Nested ANOVA corrected for sex differences 0.1761 F3, 58 = 9.418 < 0.0001

Seasons Nested ANOVA corrected for sex differences 0.2768 F3, 78 = 3.798 0.0125

Nested ANOVA corrected for site differences 0.2545 F3, 66 = 6.012 0.2630

Repeated measures ANOVA F3, 19 = 3.938 0.0188 Core home range

Females versus males ANCOVA corrected for population density 0.1798 F1, 19 = 4.351 0.0515

Sites Nested ANOVA corrected for sex differences 0.5243 F3, 58 = 3.731 0.0199 Bonferroni correction meant that probabilities less than 0.007 were classed as significant and are shown in bold. a Rosella Road site excluded due to presence of only one animal (Chapter 4). N/A refers to a significant interaction probability thereby negating further analysis. NB. Nested ANOVA with sex and site interactions and ANCOVA on site differences with population density interaction removed could not be conducted due to inadequate sample size.

165 Chapter 5 Home range and movements

Table 3. Results of linear regression of home range size estimated by the kernel method against site population density and body mass Regression r2 n Probability Overall home ranges against population density 0.660 21 0.0001 *** Overall home ranges – male 0.421 10 0.0424 * Overall home ranges – female 0.877 11 < 0.0001 *** Overall home ranges against body mass 0.011 21 0.6447 Overall home ranges – male 0.008 10 0.8087 Overall home ranges – female 0.006 11 0.8220 Core home ranges against population density 0.378 21 0.0030 ** Core home ranges – male 0.223 10 0.1684 Core home ranges – female 0.668 11 0.0021 ** Core home ranges against body mass 0.122 21 0.1199 Core home ranges – male 0.115 10 0.3384 Core home ranges – female 0.042 11 0.5434

166 Chapter 5 Home range and movements

a

b

Figure 4. Comparison between the three estimation methods for (a) a male quokka (VM1) from the Victor Road site and (b) a female quokka (KF11) from the Kesners site. Individual location fixes are shown with an asterisk. The swamp boundary is shaded. The circle in the top right corner of (b) for all three methods is caused by a discontinuous 95 percentile contour after the two most distant fixes (in the bottom left corner) were excluded. Such discontinuities can indicate multiple range centres or dispersal where larger number of fixes are involved.

167 Chapter 5 Home range and movements

Results

Between November 1998 and November 2000, 2,060 discontinuous fix locations were taken of 58 individuals (33 males and 25 females) at the five sites. These individuals averaged 217 days with a functioning radio transmitter during which an average of 35 fixes were taken (Appendix D). A nested ANOVA found there was no significant difference between the sexes (F1, 49 = 0.817; P = 0.371) or the sites (F 3, 49 = 0.906; P = 0.445) in the number of days individuals were on the air (interaction site * sex F3, 49 = 1.441; P = 0.242).

There was also no significant difference in the number of fixes taken between the sexes

(nested ANOVA F1, 49 = 0.840; P = 0.364) or the sites (F3, 49 = 0.204; P = 0.893) or site * sex (F3, 49 = 2.136; P = 0.107). This means comparisons between sexes, sites and time periods are justified.

Incremental area plots showed approximately 30 fixes were needed to estimate within

10% of the final home range size for the MCP method while 40 fixes were required for the harmonic mean and kernel estimators (Figure 1). Utilisation plots indicated that the 50 percentile home range was a conservative estimate of the core area used by each animal

(Figure 2 and Figure 3).

The home range sizes estimated by the three methods for various periods are shown in

Table 1. Despite being biased by outlying fixes (Harris et al. 1990) the MCP method generally gave the smallest estimate of home range size (Table 1). In the case of quokka home ranges, the biases appeared to affect the shape (particularly length) of the range more than the size of it (Figure 4). The harmonic mean method always gave the largest home range (Table 1 and Figure 4). The kernel estimator with core weighting was between the other two methods and was also generally sensitive enough to avoid areas not utilised (and

168 Chapter 5 Home range and movements by appearance not likely to be utilised) during the animals movements. Based on personal observations of the movements of collared quokkas, the kernel estimator was considered to be the most accurate (Figure 4). It was used to define home range sizes discussed in the text and also in statistical tests. Accordingly, the mean home range size for quokkas in the northern jarrah forest was 6.39 ± 0.77 hectares and the mean core range size was 1.21 ±

0.12 hectares (Table 1).

Differences between the sexes

Male home ranges (6.92 ± 1.16 ha) were not significantly larger than those of females

(5.91 ± 1.07 ha) however core ranges for males (1.47 ± 0.18 ha) were significantly larger than those of female (0.97 ± 0.12 ha) (Table 2). There was no significant difference between the sexes in overall or core home range size when corrected for body mass (Table

2).

Differences between day and night home range size

Diurnal home ranges averaged 3.51 ± 0.40 hectares while nocturnal ranges averaged

5.48 ± 0.90 hectares (Table 14), however these differences were not significant (Table 2).

Differences between the sites

There were highly significant differences in overall home range sizes between the sites

(Table 1 and Table 2) with animals at Hadfield having significantly smaller ranges than all other sites (Scheffe’s test for Chandler P = 0.0288; for Kesners P = 0.0156; and for Victor

169 Chapter 5 Home range and movements

Road P < 0.0001). The core home ranges showed similar results (Table 2) with Hadfield animals having significantly smaller ranges than those at Chandler and Kesners (Scheffe’s test P = 0.0491 and P = 0.0282 respectively).

18

16

14

12

10

8

6

4

2 Overall home range size (ha)

0 Diurnal Nocturnal Figure 5. Box plot of overall (95 percentile) kernel estimate of diurnal and nocturnal home range (females dark grey and males light grey) showing mean, standard error bars and outliers.

Both entire and core home range size were significantly negatively correlated with population density, however this relationship was stronger for females than males (Table 3;

Figure 7). There was no relationship between home range size and body mass (Table 3).

170 Chapter 5 Home range and movements

5

4.5

4

3.5

3

2.5 Seasonal home range size (ha) 2

1.5 Autumn Winter Spring Summer

Figure 6. Seasonal variation in quokka home range size.

Differences between the seasons

Seasonal differences in home range size are shown in Table 1 and Figure 6. There was no significant difference in seasonal home range size although home ranges in autumn were significantly larger than in spring when sexual differences (Bonferroni/Dunn post hoc test with adjusted significance level of 0.0083 P = 0.0040) and site differences

(Bonferroni/Dunn post hoc test P = 0.0012) were examined (Table 2). Repeated measures

ANOVA of seasonal differences (Table 2) showed autumn home ranges significantly larger than those of spring (Bonferroni test P = 0.0033).

171 Chapter 5 Home range and movements

1.2

1.1

1 ) (ha) 10 .9

.8

.7

.6

.5

.4

2 Overall home range size (log Female = 1.2 - 0.2 * Density; r = 0.9 .3 Male = 1.1 - 0.1 * Density; r2 = 0.4 .2 0 .5 1 1.5 2 2.5 3 3.5 4 4.5 Population density ha-1

Figure 7. Regression plot of home range size estimated by the kernel method plotting males (open triangles) and females (closed circles) against population density at each of the five sites (Table 3). Male regression line is dotted and female is dashed.

Movements

There was no significant difference in the direction of movements of individual quokkas upstream or downstream between the seasons. A significant difference was found in movement patterns across the swamp. Between autumn and winter quokkas moved toward the edge of the swamp significantly more than toward the centre (Sign test n = 24; P

172 Chapter 5 Home range and movements

= 0.041). While not significant, quokkas also tended to be more toward the edge of the swamp in spring than they were in autumn (Sign test n = 20; P = 0.096).

Dispersal

Of 48 collared adult quokkas, only one showed any evidence of a shift in home range that may have indicated dispersal. In this case, an adult male (CM2) was initially trapped in the western section of the swamp and was located there for the following two months

(Figure 8). He then relocated approximately one kilometre to the east where he stayed until his collar failed seven months later (Figure 8). This male made three forays back into his initial home range over the first month following the dispersal but never stayed more than a few days. A week after this dispersal was first observed, a new adult male was trapped in the western section where, radio telemetry revealed, he took up residence. The new adult male (3.57 kg) was larger than the displaced male (2.98 kg) suggesting a dominance or agonistic encounter may have initiated the movement. This was the only evidence of adult dispersal found in this study and no animals were observed moving from the swamp system in which they were originally captured to another.

Ten juvenile quokkas were collared at some stage during this study. Two escaped their breakaway collars and were not seen again. Of the remaining seven males and one female that remained on the air or were recaptured, only one male (from Hadfield swamp) had moved more than 100 metres from the natal home range or from where it was first trapped.

This individual was later captured 900 metres downstream as a sub-adult but was not trapped in the following five trapping sessions.

173 Chapter 5 Home range and movements

Figure 8. Overall home range of an adult male from the Chandler swamp (CM2) estimated by the kernel method. The dual contours show a shift in home range that indicates dispersal. Fixes are shown as an asterisk, the approximate area of the swamp is shaded and the dotted area to the left hand side of the figure is a dam. Note how the large distance between the dual range centres made the contours almost circular and because of this dispersal this individual was not included in analyses.

174 Chapter 5 Home range and movements

Figure 9. Kernel estimates of the diurnal and nocturnal home range of an adult female quokka (KF2). Each fix is shown as an asterisk and the approximate area of the swamp is shaded. Note how the overall home range estimate is smaller in length than the nocturnal range due to the exclusion of the most distant outlying fixes.

Discussion

Home ranges

Females on Rottnest Island were assumed to have similar sized home ranges to males

(Kitchener 1973), however in captive situations, quokkas defend territories and males form dominance hierarchies (Packer 1969; Kitchener 1972). In addition, quokkas are sexually size-dimorphic (Chapter 4) and these features suggest that sexual differences in home range size should exist with males having larger ranges than females. There was no significant difference between the sexes in overall home range size however the core home ranges of males were larger than those of females. The absence of sexual differences may be explained by the dominance hierarchy created by male quokkas. The dominant male may

175 Chapter 5 Home range and movements well have a large home range that encompasses the range of more than one female but he may also inhibit subordinate males from having large home ranges. The larger variances in male home range size compared to those of females (Figure 7) may be a result of this.

Future comparisons of the distances moved by male and female quokkas between successive fixes may elucidate this, as males may move from one end of their territory to the other regularly as part of its defence, where females may move in a more autocorrelated manner.

Larger animals tend to have larger home ranges (McNab 1963), as occurs in North

American mammals (Harestad and Bunnell 1979). Although the differences in techniques used to measure home range often make direct comparison impossible, this trend does not necessarily occur in macropodids (Croft 1989), largely due to dietary influences (Norbury,

Sanson and Lee 1989). For example the home range of quokkas is small compared to that of similar-sized macropodids. The long-nosed potoroo (Potorous tridactylus apicalis) is smaller than the quokka but has a home range size at least 40 times larger (Kitchener 1973).

The ability of the quokka to survive in such small areas compared to other small macropodids has been attributed to the likely abundance of food in the swamps of the northern jarrah forest (Norbury et al. 1989). The comparatively low-quality diet of the quokka (clumped browse) (Storr 1964) compared to the potoroo (widely-dispersed fungi)

(Claridge and Cork 1994) however, would be expected to result in a larger home range in order to satisfy the species’ high metabolic requirements (Norbury et al. 1989). This anomaly has been explained by the prediction that quokkas select the most nutritious component of their low-nutrient forage and thereby minimise their home range size

(Norbury et al. 1989), whereas the widely scattered distribution and ephemeral nature of the fungi eaten by the long-nosed potoroo forces it to forage over large areas (Kitchener

176 Chapter 5 Home range and movements

1973). Conversely, the (Wallabia bicolor) of eastern Australia occupies a similar feeding niche to the quokka; being a browser (Hollis, Robertshaw and Harden

1986), essentially solitary with loose feeding aggregations (Jarman and Coulson 1989) and preferring habitats with dense understorey (Troy, Coulson and Middleton 1992).

Consequently the swamp wallaby has a home range size (Troy and Coulson 1993) larger than the quokkas but this is in proportion to their body weight differences. The red-necked (Thylogale thetis) is approximately twice the size of the quokka and, despite preferring areas with dense understorey in rainforest, it has a home range approximately twice as large as the quokka (Johnson 1980).

Quokkas on Rottnest Island moved continuously throughout the night before returning to a regular rest site (Kitchener 1973). Captive quokkas were found to sleep during the day with evening and early morning (generally pre-dawn) peaks in foraging (Packer 1969).

These characteristics suggest that the nocturnal home range of the quokka is likely to be larger than the diurnal range although it hasn’t been quantified in previous studies. The similar sized nocturnal and diurnal home ranges on the mainland, therefore, is surprising.

For mainland quokkas, nocturnal home ranges were generally broader than diurnal ranges but with similar lengths (Figure 7). Such differences were probably caused by the diurnal resting behaviour of quokkas, where individuals appeared to rest within the swamp during the day at a point nearest to where they ceased their nocturnal activities. This concurs with the lack of evidence found during specific searches for ‘nests’ of quokkas on the mainland

(M. Dillon pers. comm.). Such resting behaviour meant that diurnal ranges were as long as nocturnal ranges but only slightly exceeded the boundary of the Agonis-dominated swamp vegetation which minimised the significance of the difference between them.

177 Chapter 5 Home range and movements

Other species of macropodid inhabiting refuges also do not necessarily exhibit larger nocturnal than diurnal home ranges. The red-necked pademelon ranges widely in rainforest during the day but forages in small areas of pasture at night in order to minimise predation risk (Johnson 1980).

Quokkas also appeared to spend the majority of their time when outside the swamp close to the swamp-forest ecotone which further minimises day/night differences in home range size (Table 41). Such preference for edge habitat or ecotonal vegetation may be driven by dietary requirements and/or predation risk reduction. Anecdotes from earlier researchers and forest workers suggest quokkas spent significant amounts of time outside the swamp prior to their dramatic decline in the 1930’s (Stewart 1936; White 1952) and even in more recent times (M. Dillon, A. Start, R. Brazell, B. Withnell and D. Giles pers. comm.). This decline has been linked to predation by introduced species such as the red fox

(Vulpes vulpes) (White 1952; de Tores et al. In prep.) and remnant quokka populations today may have survived through altered habitat use and an increased restriction to the refuge provided in the swamps. Foxes are probably more successful predators in open areas (see discussion in Chapter 7) and numerous prey species released from fox predation pressure increase their time spent in open habitats (Dickman 1992; Banks, Hume and

Crowe 1999; Banks, Newsome and Dickman 2000a). Nonetheless, all historical reports do link quokkas to densely vegetated areas (swamps in the forest and heaths and shrublands along the coast) (White 1952; Gould 1973) and this altered habitat use is not likely to have been over a broad scale (i.e. forest to swamp). These earlier anecdotes may therefore be the result of higher population densities than occur today which simply meant that quokkas were seen more frequently outside the swamps.

178 Chapter 5 Home range and movements

Factors affecting home range size

The negative relationship between home range size and population density is hardly surprising. The observation that females exhibit this relationship more markedly than males in quokkas is interesting however. As sex ratios at maturity are at parity (Chapter 4) this is not an artifact of differing densities of each sex. Males may be more tolerant of other males compared to the tolerance of females to other females; or the dense vegetation may reduce interactions amongst promiscuous males intruding upon adjacent territories while seeking polygynous mating opportunities. The sexual dimorphism exhibited by the quokka on the mainland (Chapter 4) suggests that males have higher energetic costs than females and therefore should have larger home ranges. The dominance hierarchy exhibited by quokkas (Packer 1969; Kitchener 1981) may explain similar home ranges sizes between the sexes and the stronger relationship with population density that females exhibit.

Dominant males may inhibit the movements of subordinate males which may explain the variation in male home range sizes that were observed in this study (Figure 7).

The experimental relaxation of predation pressure has been shown to reduce home range size in South American rodents due to their subsequent preference for resource-rich areas (Lagos et al. 1995). Conversely, voles (Microtus species) move more under reduced predation risk (Norrdahl and Korpimaki 1998; Banks, Norrdahl and Korpimaki 2000b).

Although quokkas at the unbaited Victor Road site have substantially larger home ranges than those at the baited sites, the lack of replicated control sites meant that investigations into the effect of predators on home range size could not be conducted, particularly considering the strong linear effect of population density.

179 Chapter 5 Home range and movements

Home range size appears to follow a harmonic curve - seasonally peaking in autumn and declining to spring before increasing again (Figure 5). In the northern jarrah forest, autumn is the season prior to the onset of the rains (Chapter 4) and so is the peak period of water restriction. Available forage is also likely to be scarce at this time, particularly if quokkas are only selecting the freshest available browse (Norbury et al. 1989). It is hardly surprising then that autumnal home ranges are largest for mainland quokkas considering this is probably the time when their resources are most widely scattered. Conversely, spring is at the end of the highest rainfall period and is a time when water (and probably fresh plant growth) is abundant. Home ranges are also smallest in this season.

This pattern of seasonal variation in home range size on the mainland contrasts with that observed on Rottnest Island where ranges increased in wetter months (Kitchener 1973).

This is probably due to the quokkas’ restriction to water sources over summer on Rottnest

Island where home ranges increase when this restriction is eased as water is more widely available in winter. This may also be an artifact of the small sample size (six) of the

Rottnest study.

Movements

Considering the affect rainfall has on the habitat of quokkas, it is probably out of necessity that they move toward the edge of the swamp between autumn and winter/spring.

The south-west of Western Australia has a Mediterranean climate with hot, dry summers and mild, wet winters (Chapter 2) (Southern 1979). This means that by the beginning of the high rainfall period (winter) each year the swamps are almost completely dry after six months with minimal rain (Chapter 2). Thus, in autumn, quokkas can, and do, move to the

180 Chapter 5 Home range and movements centre of the swamp either searching for water to satisfy their high requirements (Main and

Bakker 1981) or to obtain refuge within the dense vegetation there. The onset of the winter rains greatly reduces the amount of dry land available as the swamp becomes inundated and so quokkas move to the very edge of that vegetation. By doing this quokkas may increase their susceptibility to predation in winter and spring.

Evidence from Rottnest Island suggests that the quokka is “basically nocturnal” with crepuscular, bimodal activity peaks occurring immediately after sunset and before sunrise

(Packer 1965). Such daily movement patterns on Rottnest Island occurred in each season but were less pronounced in the arid summer months when quokkas were recorded travelling to water throughout the entire night (Packer 1965).

Climate plays a significant role in the behaviour of many species through the alteration of foraging patterns, social interactions and activity (Stokes, Slade and Blair 2001). This applies to quokka behaviour on Rottnest Island also. Despite the apparent of quokkas there are variable amounts of activity throughout the daylight hours on Rottnest

Island (Packer 1965). Sunlight inhibited quokka activity and this was most pronounced during ‘wet’ periods (winter) (Packer 1965). Conversely, during the ‘dry’ months activity increased before sunset so as to maximise the available time for feeding and other activities

(Packer 1965). Moonlight was also found to inhibit movement on Rottnest Island (Packer

1965) and probably on the mainland as well. The causes of this are unknown but may relate to the activity of predators in the quokka’s evolutionary past (Chapter 7). Rain was a dominant factor in reducing quokka activity on Rottnest Island (Packer 1965), however barometric pressure was also influential. Falling barometric pressure was associated with high activity during the dry season on Rottnest Island (October to March), while the

181 Chapter 5 Home range and movements clearing conditions associated with rising pressure led to higher activity during the wetter part of the year (Packer 1965). There was not enough data collected from this study to investigate these features of quokka movements on the mainland.

Dispersal

Only one out of 48 radio collared adult quokkas showed any evidence of long distance

(one km) movement or dispersal (Figure 8). Considering the capture of another adult male immediately after this movement it seems possible that this was the usurping of a territory that forced the movement rather than an actual dispersal event. This behaviour may be the catalyst for dispersal. Of the other uncollared adults or adults that were collared for only a short period of time, virtually all were recaptured in the same area as their initial capture suggesting quokkas exhibit very little adult movement.

Several individuals were trapped repeatedly from pouched young through to young-at- foot at which time a radio collar was attached. As discussed below, the breakaway collars proved to be of limited value due to their short-term attachment and the difficulty in recapturing individuals. Nonetheless, one juvenile female and seven juvenile males were collared or recaptured. All of those individuals stayed within the natal swamp and generally within their natal home range. The only individual to disperse from its natal home range was a male that moved approximately one kilometre downstream before losing its collar. This occurred at the Hadfield swamp which had the highest quokka density

(Chapter 4). The fact that this site was considered to be below carrying capacity (Chapter

4) may mean that only small, within-patch, dispersal distances were necessary. The

182 Chapter 5 Home range and movements absence of dispersal at the other mainland sites may also be due to their low population densities.

Dispersal in vole (Microtus agrestis and Clethrionomys glareolus) metapopulations in the Finnish archipelago is driven by overexploitation of resources (Banks et al. In review).

The invasion of the archipelago by American mink (Mustela vison) suppressed population booms and removed a major driving force for dispersal and recolonisation of vacant islands, which thereby disrupted the metapopulation dynamics (Banks et al. In review).

The fundamental interactions between the quokka and the red fox in Australia may similarly explain the absence of dispersal in quokkas and therein reduce the probability of larger source populations rescuing the metapopulation.

Minimal dispersal however, may be a trait of the species, as philopatry was initially thought to be common on Rottnest Island (Dunnet 1962; Holsworth 1964; Kitchener 1970) although, more recently, this has been refuted (Nicholls 1971). In fact, the majority of quokkas on Rottnest Island are assumed to occupy the same home range for their entire lives (Kitchener 1973). If it exists, such philopatry in both sexes is extremely unusual in mammals (Greenwood 1980), however it was explained for quokkas on Rottnest Island by their inhabitance of group territories wherein individual home ranges overlap so broadly as to allow random mating of the 50 – 100 inhabitants (Johnson 1989). On Rottnest Island, the chances of inbreeding were thought to be minimised in such large groups and there may be little selection for regular dispersal in this species (Johnson 1989), although a more likely explanation is a lower dispersal rate than could have been detected during the field studies. It is difficult to imagine this would be an effective strategy on the mainland with such small, extant tiny populations (Chapter 4) and patches possibly not large enough to

183 Chapter 5 Home range and movements support populations of even 50 individuals (Kitchener 1973). Problems associated with inbreeding may already be occurring as is evidenced by the small levels of heterozygocity found in each population (Alacs 2001).

Home range estimation techniques

The home range area determined by the kernel estimator produced mapped home ranges that appear to more accurately depict the area and shape of the creekline known to have been used by the quokkas studied. The locations of quokkas recorded in this study may be biased toward the swamp due to disturbance caused by researcher movements and so the extension of the home range boundary outside the swamp is considered accurate despite the fewer locations there (as shown in Figure 4). The MCP method is biased by outlying fixes (Harris et al. 1990), however for animals inhabiting linear corridors of habitat it appears to be a satisfactory estimation method despite potentially overly elongating home range shapes (Figure 4). The harmonic mean method appeared to place excessive weight on core areas resulting in unduly large circles around them (Figure 4), such that this was considered the poorest performing home range estimator for quokkas.

Collar success

Some problems were experienced with the applicability of the three varieties of radio collar. The breakaway collars proved too weak for even juvenile quokkas and rarely lasted more than one week. The brass loop aerial collars were highly susceptible to damage by the quokkas at the terminal-collar/aerial interface and the vast majority of these failed short of their expected lifespan. This did not significantly effect home range estimates at sites

184 Chapter 5 Home range and movements with high recapture rates (Hadfield and Victor Road) but where the recapture rates were low (Kesners and Chandler Road) (Chapter 4) and the collared individual was never recaptured then often all the initial telemetry data was wasted unless adequate fixes were obtained prior to the loss. Despite this, eventual recapture of most individuals meant that it is highly unlikely that dispersal events were missed. Survival estimates were much more affected by this problem (Chapter 7). This problem was rectified mid-way through the study by the use of whip aerial collars that functioned for much longer than brass loop aerial collars despite suffering similar damage to the collar and transmitter. The whip aerial also meant that collars that animals broke-free from could be relocated and repaired. It is recommended that future studies of macropodids and other species that can inflict significant damage to radio collars use whip aerials, particularly where recapture rates are likely to be low.

Implications for the metapopulation

Considering the linear nature of the quokkas preferred habitat (Chapter 6), the average home range size of just over 6 ha (Table 1) generally equates to approximately 600 metres along the swamp and 100 metres across it (Figure 4). Despite their relatively small home ranges, quokkas appear highly mobile (pers. obs.) and, anecdotally, the longest recorded movement was of an adult male that had been trapped for several years in the one site and was later found dead beside a road approximately 10 km away (R.Brazell pers. comm.).

Movements of 600 metres would allow quokkas to access adjacent swampy habitat either within the same catchment or aross a ridge into an adjacent one and movements of 10 km

185 Chapter 5 Home range and movements obviously would allow access to much more distant patches. That such movements were not observed indicates that the metapopulation structure is no longer functioning.

There are various explanations for the lack of inter-patch movement. The simplest being that inter-patch movement will be restored once individual populations reach the carrying capacity of the habitat (Crone, Doak and Pokki 2001; Banks et al. In review).

Macrohabitat does not appear to be the limiting factor considering the plentiful availability of Agonis linearifolia dominated swamps in the northern jarrah forest. The reported preference of quokkas for early seral stage habitat (less than 15 years since fire)

(Christensen and Kimber 1975) suggests microhabitat factors may be an issue however.

A recent study has found that inter-patch movements in a metapopulation can be inhibited by predation (Banks et al. In review) and this may be exacerbated when long distance movements are required due to the increased risk of encountering a predator

(Banks et al. 2000b). Predation pressure by red foxes (Vulpes vulpes) occurs on the mainland (Chapter 7) and this could explain the observed lack of inter-patch movement by means of limiting population growth and recruitment (Chapter 4).

Conclusion

The small distances moved by quokkas imply a bleak future for the overall metapopulation. Their restriction to the natal swamp and even natal home range means there is very little potential for recolonisation of adjacent patches and the threat of genetic problems associated with inbreeding depression is high (Frankham 1995).

Although the applicability of metapopulation theory to large mammals has been questioned due to the difficulty in observing inter-patch movements, other mechanis ms can

186 Chapter 5 Home range and movements be used to identify whether a species does exist as part of a metapopulation (Elmhagen and

Angerbjorn 2001). The absence of inter-patch movements by the quokka appears unlikely to be due to their size but rather the inhibiting effect of recently arrived predators and possibly the absence of suitable microhabitat patches within dispersal distance. The infrequency of population turnover was probably originally exacerbated by the specific habitat requirements of the quokka (Christensen and Kimber 1975) but instead of colonising new sites the increased mortality risk of dispersing meant that localised extinction occurred. Thus, although no inter-patch movements were recorded, the quokka does exhibit the features (discrete habitat patches supporting breeding populations and ecological processes, such as growth rates, acting on local and regional scales (Chapter 4)) that characterise metapopulations in large mammals (Elmhagen and Angerbjorn 2001).

Therefore, this study adds weight to earlier conclusions that the quokka did originally exist as a classic metapopulation which is now either collapsing or collapsed (Chapter 4).

The causes of the apparent lack of inter-patch movement can only be hypothesised.

The existence of populations well below carrying capacity seems the most likely explanation (Chapter 4). Predation pressure may be exacerbating this problem through the maintenance of low population density and by preying upon dispersing individuals. That predation on dispersing individuals was not observed in this study (Chapter 7) suggests there may have been selection for non-dispersal or philopatry in the quokka since the arrival of the fox into the south-west of Australia in the early 1930’s (King and Smith

1985). The forest surrounding the swamps may now be a population sink such that instead of swamps and forest existing as part of a mosaic of different suitability there is now a small number of small habitat islands occurring within an ocean of uninhabitable forest.

Although the theory has largely been discounted (Nicholls 1971), the final alternative is

187 Chapter 5 Home range and movements that philopatry is a behavioural trait of the species (Holsworth 1964) and when quokkas are in greater abundance there may be adequate mixing within local populations to avoid genetic problems.

From a management perspective, there are several options available. Increases in population density should be worked towards by largely preferring in-situ methods such as the continuation of predator control. Habitat near existing quokka sites must be suitable and free of threats to support the species if, as seems most likely, population density drives dispersal and they subsequently obtain densities that require dispersal. The removal of threats, particularly predation, between habitat patches is also likely to increase dispersal opportunities. Monitoring is also important to determine the maximum distance that quokkas will disperse so that sufficient habitat patches can be managed within this area.

188 Chapter 6 Habitat description

The habitat in and around potential quokka (Setonix

brachyurus Quoy & Gaimard 1830) (Macropodidae:

Marsupialia) swamps of the south-west of Australia.

Abstract. The habitat types in and around 66 potential quokka populations in the northern jarrah forest was described and the effect of fire on Agonis swamp shrublands

– the principal habitat of the quokka - was quantified. Habitat units were mapped from aerial photographs that were imported into a geographic information system. Field surveys were then conducted at each of these sites to ground-truth each of the habitat units. Differentiation between each habitat unit was conducted using TWINSPAN and principal components factor analysis. TWINSPAN differentiated between broad habitat units but did not successfully distinguish between closely related units such as those within the Agonis swamp shrublands that had previously been separated according to the number of years since the last fire. Principal components factor analysis adequately differentiated the majority of swamp habitat units. These formed a continuum structurally and floristically.

The Agonis swamp shrublands appear highly adapted to the pre-European fire regime of Australia with species richness peaking between five and 19 years following a fire. The vegetation within the swamps remains relatively open for the first five years following a fire while being largely dominated by three to four species. Thereafter, vegetation density and species richness increased to a peak in density after 24 years and in species richness after 14. Long unburnt Agonis swamp shrubland habitat returns to intermediate vegetation density levels, although becoming increasingly woody, as a few

189 Chapter 6 Habitat description

species dominate. Such a response to fire reflects adaptations to the frequent, low intensity fire regimes utilised by Aborigines prior to European colonisation.

Introduction

The interaction between an organism and its biotic environment is one of the fundamental tenets of ecology. When that organism is a threatened species, an understanding of this interaction becomes crucial in the management of the species. Yet simply listing a group of habitat units used by an animal is inadequate if the reasons why it is using this habitat are sought. In this case, a thorough scientific investigation into the properties of these habitat units is warranted.

As part of a study into the ecology of the quokka (Setonix brachyurus) (Quoy &

Gaimard 1830), the first aim of this section was to map, differentiate and describe the habitat units within which the species was likely to occur throughout the northern jarrah forest of Australia. The quokka is a medium-sized, macropodid marsupial that is endemic to the south-west of Australia and some offshore islands (Chapter 3). In the jarrah forest of the mainland, the quokka occurs in the densely vegetated, swampy, upper reaches of creek systems dominated by the tall shrub Agonis linearifolia

(Christensen and Kimber 1975). In the karri forest to the south the quokka inhabits both swamps and ridge tops while along the coast and offshore islands it inhabits heaths and shrublands (White 1952; Christensen and Kimber 1975; Christensen et al. 1985;

Kitchener 1995). Variation in quokka abundance within these areas has been linked to changes in the habitat associated with time since fire (Christensen and Kimber 1975) and so the second aim of this study was to examine the effect of fire on the Agonis swamps of the region. Although habitat types within the jarrah forest have previously

190 Chapter 6 Habitat description

been mapped and described (Havel 1975), the classification produced was general and did not examine vegetation age and was therefore not sufficiently detailed to assess quokka habitat.

The effect of fire on jarrah forest communities has also previously been investigated

(Hatch 1959; Wallace 1966; Abbott et al. 1985; Bell 1995; Grant and Loneragan 1999).

Jarrah (Eucalyptus marginata) is recognised as one of the most fire resistant eucalypts

(Wallace 1966). Low intensity fires assist in opening the seed and by removing competition for the seed (Wallace 1966). The seedling develops a lignotuber and then lies dormant for ten to 20 years until an opening in the canopy, possibly caused by fire, enables a single shoot to develop (Wallace 1966). Adult trees develop thick bark up to four centimetres thick (Wallace 1966). Jarrah also exhibits dormant budding, epicornic shooting and a surge in growth rate in the first five years following a fire (Wallace

1966).

While much is known about the effect of fire on the jarrah forest, their focus meant that these previous studies tended to exclude the swamp systems from assessment.

Consequently, this present study aims to describe the potential habitat in and around quokka swamps in the northern jarrah forest and to provide information on the effect of fire on the structure and floristics of these swamps.

Methods

Quokka sites were identified by the known presence of a population, by recent sightings or by road kills. Potential sites were identified by the presence of suitable habitat. Staff from the Western Australian Department of Conservation and Land

Management (CALM) district offices within the quokkas’ distributional range (Chapter

191 Chapter 6 Habitat description

3) were questioned to identify these sites. This allowed a description of all potential habitat units within 200 m of potential swamps that, based on observed movements

(Chapter 5), could be used by quokkas.

Once an area was selected, air photo interpretation was conducted. The air photographs were scanned as high-resolution images by the WA Department of Land

Administration (DOLA) and geo-referenced to known points in the MapInfo

Professional Version 5.5 computer package (MapInfo Corporation Inc. 1985-1999).

Habitat units were then mapped with a polygon.

Once a site was mapped it was ground-truthed to confirm the habitat type of each unit. Ground-truthing initially involved recording structural, floristic and other habitat component data that were later used to differentiate each unit. Quadrats were placed in the centre of each unit to minimise the influence of ecotonal features. After each habitat unit was ‘typed’ it was representatively sampled with between 5 and 40 quadrats depending upon the frequency of each habitat units’ occurrence.

The structure of each unit was determined by estimating the vegetation density at

0.1 m, 0.3 m, 0.5 m, 1.0 m, 1.5 m, 2.0 m, 3.0 m and 5.0 m heights about ground level.

Vegetation density was estimated by the percent cover of stems within a 1 m x 0.2 m quadrat turned horizontal to the ground at each height unit and categorised according to the Braun-Blanquet scale (Braun-Blanquet 1932; Gauch 1986). Floristics at these sites were determined using a 2 m x 1 m quadrat for shrubs, sedges and grasses within a 10 m x 10 m quadrat for trees. The percent cover of each species within and overhanging the quadrats was categorised according to the Braun-Blanquet scale. Species were identified according to Marchant et al. (1987a; 1987b) and nomenclature was based on

CALM’s FloraBase web site (www.calm.wa.gov.au/florabase.html).

192 Chapter 6 Habitat description

Other habitat components measured were:

· leaf litter depth (using a ruler to measure the depth from the top of the litter to the soil surface at three places within the 10 m x 10 m quadrat);

· soil moisture (using a scale of 0 representing dry, 1 representing moisture raised to the surface when trodden on, and 2 for open bodies of water);

· number of stags (dead trees or shrubs);

· percent cover of bare earth;

· percent cover of open water;

· number of cut stumps; and

· the number of logs.

The stags, cut stumps and logs were counted in 10 m x 10 m quadrats and the remainder were recorded in the smaller 2 m x 1 m quadrats. These habitat components are grouped hereafter as ‘other habitat variables’.

The Agonis swamp shrubland habitat units were further subdivided to investigate microhabitat use by the quokka (Figure 1). The number of years since a fire burnt each vegetation unit within the swamp shrubland (that is the age of the habitat unit) was estimated by counting growth rings on the dominant species, Agonis linearifolia. This species dominates swamps in the jarrah forest but is burnt back by fire after which it resprouts. Agonis linearifolia is also thought to have only one growth period per year

(Storr 1964a). A 240 mm straight pruning saw was used to cut the stems of the Agonis.

The pruning saw had a steel blade with hard, chrome-plated finish which was friction free and produced a smooth cut surface. This smooth finish allowed annual growth rings to be counted without further preparation.

193 Chapter 6 Habitat description

40

30

20

10

0 Fresh (< 5) 5-9 10-14 15-19 20-24 > 25 Period (years) # samples

Figure 1. Number of samples (line) and the period of years over which they span

(columns) for each age class within the Agonis swamp shrubland.

Data analysis

Two Way Indicator Species Analysis (TWINSPAN) (Hill 1979) was used a posteriori to see whether the habitat units identified in the field were distinct from one another. This analysis was conducted with a recent version of TWINSPAN that used strict convergence criteria (Oksanen and Minchin 1997). Analysis was performed on percent cover data categorised into the Braun-Blanquet scale. Plant species that occurred at only one site or that occurred at less than 2% cover at less than five sites were excluded from the analysis.

TWINSPAN produces a two-way species-site matrix that shows which species are associated with particular sites (Hill 1979). The matrix is formed by dividing successive reciprocal averaging ordinations into the two most dissimilar groups of sites

194 Chapter 6 Habitat description

and species consecutively until the requested level is attained (Tausch et al. 1995).

Because the degree of differentiation sought differed from broad habitat units outside the swamp, to fine units inside, the dataset was split into two groups: the jarrah forest and Agonis swamp shrublands.

Following this, the separation of the Agonis swamp shrubland habitat units required more detailed investigation. Principal components factor analysis was conducted on structural, floristic and other variables in the swamps using the Statview Version 5.0 computer program (SAS Institute 1992-1998). Variables not conforming to the assumptions of normality were normalised using square-root transformations (small counts) and arcsine transformations (percentages). Analysis of variance was used to determine the significance of differences between the factor scores of each habitat unit.

Results

A total of 92 plant species were identified in and around 66 potential quokka sites located throughout the northern jarrah forest of Australia. Of these, 95.5% were indigenous plant species and 4.5% were exotic. Within Agonis swamp shrublands there were 57 plant species identified including one exotic weed species.

Confirmation that Agonis linearifolia has an annual growth regime occurred when the number of years since a fire was compared with growth rings at sites of known fire history. Growth rings at three sites were used for the comparison: the freshly burnt

Findlay Brook site (one growth ring in six months since a fire); the older Victor Road site (three to four rings in the three and a half years since a fire) and the Kesners site

(ten to 12 rings in the 12 years since a fire). Where growth rings indicated more than one age present within a quadrat, the oldest age was used.

195 Chapter 6 Habitat description

Differentiation between habitat units

TWINSPAN adequately split the broader habitat units within each region. It showed that flooded gum forest, swishbush swamp, low heath, blackberry thicket, ti- tree swamp, jarrah-marri open forest, revegetation, pine plantation, wandoo woodland and introduced pasture were distinct habitat units (Figure 2). Separate from these units were the Agonis swamp shrublands but these only separated into two discernable units – freshly burnt Agonis swamp shrublands and the rest (Figure 2). Another interesting feature here is the lack of differentiation between the bullich, blackbutt and peppermint forests. Appendix D shows the habitat units identified and the variables used to describe them.

196

swamp

Blackberry thicket 1 Flooded gum forest Swishbush1 swamp 1 1 1 1 Low heath2 2 2 1 2 2 2 2 2 1 1 1 1 1 2 2 2 Bullich-Blackbutt open forest / Blackbutt1 open forest1 / Peppermint forest 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 Agonis shrubland 1 1 9 2 2 8 8 6 8 8 3 3 2 2 0 1 1 1 2 2 6 6 7 7 8 8 9 9 9 0 0 7 8 8 1 3 4 6 9 2 2 2 4 5 5 5 5 9 0 0 1 1 1 5 6 7 1 1 1 4 4 5 5 5 1 1 1 1 2 3 3 3 4 4 4 4 5 6 7 7 8 8 8 8 8 9 9 9 9 0 0 1 1 1 1 2 3 3 3 3 3 4 6 6 6 7 8 8 9 9 0 0 0 1 1 3 3 4 4 4 1 2 2 2 3 3 4 5 5 6 6 6 6 7 7 7 8 0 2 3 1 0 5 3 0 2 4 7 8 2 4 7 7 8 9 0 1 4 5 4 5 4 6 0 1 2 5 6 9 7 8 6 2 3 6 5 0 1 5 7 5 6 8 9 8 1 9 0 1 2 9 9 5 2 3 4 1 2 0 1 2 2 3 4 5 1 2 3 4 9 4 5 7 0 1 6 9 4 9 0 2 1 3 5 7 8 4 6 7 9 0 8 3 4 7 8 4 1 3 4 7 8 4 1 2 8 0 6 9 5 7 4 8 9 0 1 1 9 0 5 6 7 8 9 0 3 7 8 0 8 8 0 1 0 1 2 3 1 3 9 2 4 5

Xyris lacera ------3 ------2 2 5 ------5 - - - 2 2 - 2 4 - 2 ------2 - 4 2 5 4 - - - - - 6 5 5 - 4 - - 3 - 6 5 3 6 4 4 - - - - 5 3 - Antheria prolifera ------5 ------3 3 4 4 - 5 - - - - 6 - Juncus microcephalus ------2 ------Oxylobium lineare ------2 ------3 ------2 2 ------2 - - - 3 ------3 ------3 Gahnia decomposita ------3 ------3 ------2 - - 2 - - 3 - - - 2 4 4 ------2 3 2 - - - 3 ------2 ------2 3 3 - 3 - - 5 2 - - - - 4 ------4 - - - - 6 - - - - 3 3 5 2 - - - - 3 Lepidosperma tetraquetrum ------5 5 5 2 6 6 4 4 4 6 6 6 6 6 6 5 5 3 5 6 - - - - - 5 - - 6 6 6 6 6 6 3 - - 2 3 6 - - 6 3 - 3 2 - 6 6 6 6 4 6 6 5 2 6 6 6 3 6 6 5 6 3 6 6 2 3 - - 3 6 6 6 6 2 6 - 6 3 - 6 6 6 6 6 6 6 6 5 3 6 4 6 4 6 5 6 6 6 6 6 3 - 6 5 2 - 3 3 - - - - 3 2 4 3 - - - - 2 2 2 2 - 3 Agonis linearifolia ------4 ------3 5 - 4 4 2 - 2 - 4 2 - 3 3 5 3 2 - - 4 4 - - - 5 6 4 6 6 5 3 2 - 3 - 3 6 6 2 4 4 2 6 6 6 6 4 6 5 6 5 5 5 6 5 5 6 5 6 6 6 6 6 6 5 5 4 6 6 6 5 6 6 6 6 5 6 6 6 6 5 6 6 6 6 6 6 6 6 6 6 6 3 6 6 6 5 5 6 6 6 5 6 6 6 6 6 5 5 6 6 3 6 6 6 6 6 6 6 4 6 5 6 4 Aotus cordifolium ------4 3 3 ------2 ------4 ------2 - - - 3 ------2 4 ------6 - - Comesperma virgatum ------2 ------2 - - 2 ------2 - 2 - - - - 2 - Thomasia pauciflora ------3 5 ------2 - - 3 4 - 2 ------2 - - 2 3 - - - - 2 - - - 5 - 5 4 ------4 - - - - 3 6 5 3 - - - - 2 - 3 - 4 ------3 ------3 ------Dampiera hederacae ------2 - - - 2 ------2 ------2 ------3 ------3 ------2 2 ------2 2 ------Astartea fascicularis ------2 2 - - - 4 - 3 ------4 2 - 4 2 ------2 - - 2 - - 2 - - - 2 ------3 ------3 3 ------2 - - - - 3 - 2 - - - 3 ------2 - 2 - - - 3 ------2 3 2 2 - - 2 3 3 3 - - 3 2 3 2 2 - Boronia molloyiae ------3 - - - - - 2 - - 2 - - 3 - 2 4 2 - - - 4 4 3 - - - 3 ------2 2 ------4 - 4 - - - - 3 ------2 ------3 3 ------Hypocalymna cordifolium ------3 ------6 4 - - - - 6 6 6 6 4 3 - - - - - 6 3 6 6 3 5 - 3 3 6 ------2 - 3 - - 4 ------2 - - - 3 - - - 3 2 ------3 2 ------3 ------2 - Oxylobium lanceolata ------2 3 ------3 6 4 6 ------2 ------Juncus pallidus ------6 - - - 6 ------2 3 - 4 - - - - - 5 ------Leptospermum firmum ------4 ------Rubus affinis ------5 ------3 4 ------Lepidosperma squamatum ------6 ------2 - - - 6 - 2 ------3 ------3 3 4 3 ------3 ------5 Lomandra seracea ------2 ------2 ------4 3 3 ------4 alata ------2 - - 3 ------2 2 2 ------2 ------Acacia divergens ------5 - - - 5 - - - - 3 4 ------2 2 - - - 2 ------4 - 2 ------3 ------3 - - 4 ------2 ------2 - - 2 - - 3 ------2 2 ------3 2 2 Pteridium esculentum ------2 - - - 2 - - - - - 3 - - - 2 - 3 - - - - 3 2 - - 2 2 - 2 - - - - - 3 - 3 2 - - - 2 - - - - - 2 3 - - - 3 - 2 - - - - 5 ------3 3 - 2 - - - - - 2 ------2 3 - - 2 ------3 ------3 ------Thomasia paniculata ------4 3 ------2 3 ------3 ------5 6 ------2 3 ------3 - - 4 ------6 3 3 ------2 - - - - - Grevillea diversifolia ------4 3 3 3 4 ------6 ------Trymalium floribundum ------2 - - 3 - 4 6 - - 4 ------3 6 - - 4 ------3 ------4 ------Banksia littoralis 2 - - 3 ------3 5 5 5 5 - - 2 ------3 ------Lepidosperma tennue ------3 ------3 ------2 - - - - - 3 - - - - 2 ------2 ------6 - 2 ------2 ------ 4 2 ------2 ------Melaleuca rhaphiophylla - - 3 5 3 ------4 - 3 ------3 ------3 ------2 - - - 3 ------4 ------Tetraria capillaris - - - - - 4 ------2 ------3 ------3 ------Leptospermum erubescens - 2 2 2 4 4 5 6 6 6 5 2 ------Eucalyptus patens ------2 - 3 2 - - - 2 - - 6 6 6 ------3 2 3 4 4 2 - - - - 4 2 4 3 3 4 3 4 3 3 - 5 - 5 3 ------4 - - - - 3 ------Eucalyptus megacarpa ------6 5 2 3 4 ------4 ------2 - 2 ------Lepidosperma angustatusm ------4 ------2 ------2 - - - - - 2 - - 3 2 ------2 2 ------3 ------3 6 ------4 ------Hakea trifolia ------2 5 4 ------Drosera sp. - - - - - 2 3 ------Acacia extensa - - 2 - - 2 ------2 ------2 3 ------Viminaria juncea - - - - 2 ------4 ------Callistemon phoenicus ------3 ------3 - - Amphipogon turbinatus ------Drosera sp.2 ------2 2 ------2 Mirbelia dilatata ------3 - - - - 3 - 4 4 - - - 3 2 ------2 ------3 - 2 - - - - 2 - - - 2 ------2 ------3 - - Scaevola calliptera 2 2 ------2 ------Melaleuca incana - - - 2 ------Eucalyptus maculatus ------Pinus sp. ------Eucalyptus wandoo ------Bossiaea eriocarpa ------Acacia pulchella ------2 ------Dryandra nivea ------3 ------Hibbertia montana ------2 ------Hakea undulatum - - - 2 ------Persoonia longifolia - - - 2 ------Sporodora stricta ------3 - - - 2 ------Acacia laterifolia ------3 ------2 ------3 - 2 ------Acacia saligna ------Agonis flexuosa ------5 ------Allocasuarina fraseriana ------Banksia grandis ------2 ------Bossiaea ornata ------Davesia decurrens ------Corymbia calophylla ------2 2 2 - 4 ------3 ------2 ------Eucalyptus marginata ------3 ------4 - 3 ------4 ------2 - - Hakea amplexicaulis ------Hybanthos floribundum ------2 - 2 ------Lasiopetulum floribundum ------2 ------2 ------2 ------2 ------Lepyrodia muiri ------Leucopogon pulchella ------Leucopogon verticulus ------3 ------Macrozamia reidii ------2 ------2 ------2 ------Oxylobium capitatum ------Pimelea suaveolens ------Pimelea sylvestris ------ gracilis ------Conostylis aculeata ------Conostylis setigula ------Allocasuarina fibrosa ------Chorizema ilicifolium ------2 ------2 ------Dryandra sessilis ------Hakea lissocarpa ------Hibbertia acerosa ------Loxocarya cinerea ------Loxocarya flexuosa ------Xanthorrhoea pressei ------2 ------2 ------3 ------Hypocalymna angustifolia ------3 ------2 ------Trymalium ledifolium ------Bossiaea aquifolium ------2 ------Acacia celastrifolia ------Introduced pasture ------Taraxacum officinale ------

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 0 0 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Figure 2. TWINSPAN results for the northern jarrah forest quadrats. Floristic data are grouped into habitat units by the horizontal and vertical lines.

197

Agonis Swamps 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 5 9 9 9 4 5 5 4 6 6 3 4 7 8 2 8 8 1 1 1 1 1 2 2 2 3 4 4 4 5 5 5 6 6 6 6 7 7 8 9 0 0 0 1 1 2 3 3 3 3 5 1 4 4 6 6 4 4 4 4 7 4 3 4 5 6 6 1 8 5 9 1 1 1 7 0 2 2 2 0 2 7 4 5 6 8 9 5 0 1 9 0 9 1 2 3 7 3 5 1 4 5 0 1 2 9 4 5 8 7 7 8 9 0 5 7 2 5 6 7 3 6 4 9 2 7 8 2 9 1 4 5 7 8 1 3 4 6 4 9 3 4 7 8 8 2 6 2 1 3 8 6 7 2 6 5 7 8 4 1 4 5 6

Oxylobium lanceolata ------2 ------Sporodium stricta ------3 ------Hybanthos floribundum ------2 ------2 ------Lasiopetulum floribumdum ------2 ------Pteridium esculentum - - 3 - 3 ------2 2 3 5 2 2 3 - - - 3 ------3 - - - - - 2 - - 2 - 2 - - - 2 ------3 - - 3 3 3 3 3 - 2 ------Tetraria capitatum ------3 2 ------3 - - 3 ------Bossia aquifolium ------2 ------Thomasia paniculata - - - - - 5 ------2 2 3 3 6 5 3 3 - - - - - 3 4 ------Trymalium floribundum ------3 - 4 ------4 ------3 ------Macrozamia reidii ------2 ------2 ------Agonis flexuosa ------5 ------Chorizema ilicifolium - - - - - 2 2 2 ------Hypocalymna angustifolium - - - - - 3 ------Lepidosperma angustatum - - - - - 3 - 2 2 3 6 3 6 ------3 ------2 ------Viminaria juncea ------4 ------Xanthorrhoea pressei ------3 ------Corymbia calophylla - - - 3 2 3 ------2 ------Conostylis aculeata - - - 2 ------Scaevola calliptera - - 2 ------Thomasia pauciflora - - 2 2 - - 6 5 - 5 3 ------5 - 5 4 - - - - - 4 - - - - - 3 - - - - - 4 ------3 3 - - 2 2 3 3 3 - 3 - - 3 - - - - Banksia littoralis 3 ------3 3 ------3 ------2 ------Lepidosperma tetraquetrum - - 2 - - - - 6 - - 3 - - 6 2 6 6 3 6 6 4 6 6 4 6 6 2 6 6 6 6 6 2 4 4 2 3 6 6 6 6 6 3 6 5 3 6 5 5 4 6 6 6 6 6 3 5 3 2 6 6 - 3 6 5 2 3 3 3 - - 2 5 5 ------Mirbelia dilatatum ------2 2 ------2 - - - - - 2 2 ------3 ------Dampiera hederaceae - - - - 3 ------2 ------2 ------3 - - 3 2 - - 2 2 ------2 ------Lepidosperma tennue - - 3 ------2 - 6 ------2 ------2 - - - 2 ------2 ------5 ------2 ------Acacia lateralis - - - 2 - - - - - 2 - - - 2 ------3 ------2 ------Acacia alata - - 3 ------2 2 2 2 - - - - 2 ------Eucalyptus megacarpa - - - 6 6 ------2 - - - 4 2 ------4 ------6 - 5 - 4 4 - 3 3 ------Eucalyptus patens - - 3 3 ------2 ------4 3 - - - 3 - 3 ------3 - - 3 4 ------Boronia molloyiae - 4 4 ------2 - 4 - - - - 4 - - - 3 3 - - 3 ------2 2 - - - - 3 2 - 3 3 2 - 2 5 5 - - 3 3 - - - - - Hypocalymna cordifolium ------2 3 4 ------2 - 4 - - 3 - - 3 3 - - - - 3 3 - - 3 - - - - 3 - 5 - - - - - 3 2 4 3 ------Melaleuca rhaphiophylla 5 3 4 ------2 ------3 ------3 - - - - 4 ------Juncus microcephalus ------5 ------2 ------6 - - - Agonis linearifolia - - 2 3 6 4 5 4 5 5 4 3 6 6 6 6 6 5 5 5 6 6 6 4 6 5 5 5 6 5 6 6 4 3 5 6 6 6 6 6 5 6 5 6 6 6 3 4 2 6 6 5 6 5 3 4 6 6 6 6 4 6 6 6 5 2 6 6 6 3 2 6 4 - 3 6 5 5 5 6 6 6 6 6 Acacia divergens - - - - - 2 ------3 ------3 ------2 ------4 - - 3 4 - - - - 4 3 2 - 2 5 4 - - - - 4 ------Leptospermum firmum ------4 ------Oxylobium lineare ------2 - - - 2 - 2 2 ------2 3 ------Eucalyptus rudis - 6 ------Eucalyptus marginata - - - - - 3 ------Astartea fascicularis - - 2 3 2 ------2 2 - 3 - - - 2 ------2 ------2 3 ------3 - 3 2 - - 2 - - 2 3 3 4 3 2 - 2 3 3 - 3 3 2 3 3 3 - - Hakea undulatum 2 ------Leptospermum erubescens 2 ------Melaleuca incana 2 ------Persoonia longifolia 2 ------Drosera sp. ------2 ------Gahnia decomposita - 3 - 4 5 ------3 - - - 2 2 3 ------3 ------2 - - 2 2 ------4 - - - - 3 4 - Amphipogon turbinatus ------Callistemon phoenicus ------3 - 3 - - - - - Lomandra seracea ------3 3 2 - 6 5 6 Comespermum virgatum ------Lepidosperma squamatum ------6 - - - - 2 ------3 ------Aotus cordifolium ------2 ------2 ------2 4 ------Acacia extensa ------Antheria prolifera ------5 ------3 ------Xyris lacera ------5 ------2 ------2 4 ------3 ------Juncus pallidus ------6 ------3 ------Rubus affinis ------

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0

Figure 3. TWINSPAN results for the Agonis swamp shrubland quadrats. Separation of flora into distinct habitat units was impossible.

198 Chapter 6 Habitat description

While TWINSPAN was able to differentiate the majority of habitat units, two required further investigation – Agonis swamp shrublands and bullich-blackbutt open forests.

Analysis of variance results on the factor scores are shown in Table 1. The Agonis swamp shrubland habitat units were largely differentiated using individual factors with significant differentiation occurring on 13 out of 15 habitat type comparisons (Table 2). Actual factor scores for each variable are presented in Appendix E.

Table 1. Analysis of variance results for comparisons between the various age classes of the Agonis swamp shrubland habitat types. The individual comparisons using Scheffe’s test are shown in Table 2.

Factor F5, 165 Probability

Structural factor 1 (SF1) 0.914 0.4736

Structural factor 2 (SF2) 6.491 < 0.0001

Structural factor 3 (SF3) 2.148 0.6230

Structural factor 4 (SF4) 21.098 < 0.0001

Floristics factor 1 (FF1) 1.453 0.2080

Floristics factor 2 (FF2) 2.509 0.0321

Floristics factor 3 (FF3) 3.345 0.0066

Floristics factor 4 (FF4) 0.385 0.8589

Floristics factor 5 (FF5) 8.597 < 0.0001

Other habitat factors 1 (OF1) 31.724 < 0.0001

Other habitat factors 2 (OF2) 6.960 < 0.0001

Other habitat factors 3 (OF3) 1.287 0.2718

199 Chapter 6 Habitat description

Table 2. Identification of the factors used to differentiate between the Agonis swamp shrubland habitat units. The principal component factor is presented along with the

Scheffe’s test probability in brackets. The principal component factors used were structural

(SF1, SF2, SF3, SF4), floristic (FF1, FF2) and other habitat variables (OF1, OF2) and the factor scores of each of these is presented in Appendix F. A cross (X) signifies there was no direct differentiation possible between habitat units and so plots of two factor scores were subsequently used (Figure 4).

Habitat unit Fresh 5-9 years old (y.o.) 10-14 y.o. 15-19 y.o. 20-24 y.o. > 25 y.o. Fresh - 5 – 9 year old OF1 % - OF2 % 10 – 14 year old SF2 * X - OF1 * OF 2 * 15 – 19 year old SF2 % SF4 % SF4 % - FF5 % FF5 % FF5 % OF1 % OF2 % 20 – 24 year old SF2 % SF4 * OF1 * FF5 * - OF1 % OF1 % > 25 year old SF4 % SF4 % SF4 % SF4 % X - OF1 % OF1 % OF1 % FF5 % OF1 % * significant at less than 0.05 level.

% significant at less than 0.01 level.

Plotting of two factor scores from most structural and floristic factors allowed the separation of the two Agonis age class comparisons that could not be statistically

200 Chapter 6 Habitat description differentiated using individual factors (Table 2). For example, structural factors two and three and floristics factors one and two differentiated between the 5 – 9 and 10 – 14 and the

20 – 24 and > 25 year old Agonis swamp shrubland habitat units (Figure 4). High scores in structural factor two relate to dense vegetation between two and three metres while high factor scores in structural factor three relate to high density from one to two metres above ground level (Figure 4 and Appendix F). High factor scores for floristics factor one relate to a high species cover of Astartea fascicularis while negative values relate to low cover of

Lepidosperma tetraquetrum. High factor score values for floristics factor two relate to high cover of Agonis linearifolia.

b. a. 1 0.5 Density Agonis 20 - 24 between linearifolia 5 - 9 2 - 3 m. 20 - 24 0.5 15 - 19 0.2

10 -1 4 > 25 15 - 19 0 -0.1 5 - 9 > 25 10 - 14 -0.5 -0.4 Fresh Fresh Floristics factor 2. Structural factor 2 -1 -0.7 -0.5 0 0.5 1 -0.6 -0.3 0 0.3 0.6 Structural factor 3 Density from Lepidosperma Floristics factor 1 Astartea 1 - 2 m. tetraquentrum fascicularis

Figure 4. Factor score plots highlighting the differentiation between the 5 – 9 and the

10 – 14 year old Agonis swamp shrubland habitat units. Plot a. shows structural factor scores 2 and 3. Positive values of factor 2 relate to high vegetation density between two and three metres above ground level while high values of factor 3 relate to high vegetation density between one and two metres. Plot b. shows a plot of floristic factor scores 1 and 2.

Positive values of factor 1 relate to a high cover of Astartea fascicularis and negative 201 Chapter 6 Habitat description values relate to a low cover of Lepidosperma tetraquetrum. High values of factor 2 relate to high cover of Agonis linearifolia.

The bullich-blackbutt forest types also separated using factor analysis with a factor of other habitat variables characterised by a positive relationship towards species richness

(orthogonal solution = 0.8) and a negative relationship to leaf litter depth (-0.6) (ANOVA

F2, 165 = 14.9; P < 0.0001).

Description of individual habitat units

The determination that all the proposed habitat units were distinct from each other is important, despite occurring a posteriori because they can now be used to assess habitat use of fauna occurring in them. Equally important is the observation that the Agonis swamp shrubland habitat units could still be differentiated based on features of the habitat despite generally and obviously existing in a gradual cline from one age class to another. Each unit is identified below and their unique characteristics are briefly described. Actual data summarising each unit are presented in Appendix D.

Agonis swamp shrubland habitat units

The Agonis swamp shrubland habitat units are generally similar in their species composition, all being dominated by Agonis linearifolia with Lepidosperma tetraquetrum amongst the ground layer. Species that also occur in these habitat units include

Hypocalymna cordifolium, Xyris lacera, Thomasia species and the sedge Gahnia

202 Chapter 6 Habitat description decomposita. These units occur on alluvial soils in the broad, gently-sloping, upper reaches of creek systems (Mulcahy 1967; Mulcahy et al. 1972) in the wetter western side of the jarrah forest (DCE 1980).

Freshly burnt Agonis swamp shrubland

The early seral stage of the Agonis swamp is characterised by relatively low density

(Figure 5) and dead stems (stags) of Agonis linearifolia (Figure 6) that exhibit evidence of resprouting at the base of the plant. Very little leaf litter remains following the obviously recent fire and the amount of bare ground is high. Species richness is very low and this is reflected in a relatively low vegetation density. For the first four years following a fire

Lepidosperma tetraquetrum and Astartea fascicularlis are amo ng the few species that coexist with Agonis linearifolia.

5 – 9 years since fire

As the vegetation density at the ground layer increases (Figure 5) so to does species richness and leaf litter depth. The dead stags from the fresh burn have largely gone (Figure

6).

203 Chapter 6 Habitat description

1 Density between 2 -3 m 0.5 20 - 24 15 - 19

10 - 14

0 > 25

5 - 9

-0.5 Structural factor 2 Fresh

-1 -0.5 0 0.5 1 Structural factor 1 Density below 1 m

Figure 5. Plot showing the relationship between structural factors one and two and the differentiation of Agonis swamp shrubland habitat units. The mean point is shown along with standard error bars. Positive values of factor one relate to dense vegetation below one metre and high positive values of factor two relate to high density between two and three metres. The arrows show the increase in vegetation density for both structural factors as time since fire increases until the 20 – 24 year age class, afterwhich the vegetation structure returns to a more intermediate level.

204 Chapter 6 Habitat description

9

8

7

6

5

4

3

2

1

0 Fresh 5-9 10-14 15-19 20-24 >25 Species richness Leaf litter Stags

Figure 6. Line chart showing the relationship between mean (± s.e.) species richness per quadrat, leaf litter depth and the number of dead stems (stags) for each age class of

Agonis swamp shrubland habitat units.

10 - 14 years since fire

Between 10 and 14 years following a fire, species richness per quadrat reaches a peak, leaf litter continues to develop and the amount of bare earth and number of stags decline

(Figure 6). Vegetation density at ground level reaches an intermediate level but vegetation density between two and three metres increases substantially (Figure 5).

15 – 19 years since fire

205 Chapter 6 Habitat description

Species richness declines (Figure 6) while vegetation density at increases further

(Figure 5). Leaf litter depth continues to increase linearly (Figure 6).

20 – 24 years since fire

Species richness continues to fall while litter depth increases (Figure 6). Vegetation at ground level and between two and three metres attains its highest density (Figure 5).

Long unburnt swamp (more than 25 years since fire)

Species richness continues to fall after 25 years following a fire (Figure 6). Leaf litter depth peaks at long unburnt sites. Vegetation density returns to a more intermediate level, akin to that between five and 14 years following a fire (Figure 5). The height of the Agonis swamp shrubland may reach up to 10 to 15 metres.

Other habitat units identified in this study are described below. The forest communities generally occur on laterite deposits on the slopes and ridges of the jarrah forest (Bartle

1987). Again, the data that summarises their unique characteristics is presented in

Appendix E.

Allocasuarina forest

A component of the jarrah-marri open forest this habitat is dominated by Allocasuarina fraseriana while possessing a standard jarrah-marri open forest understorey. Species

206 Chapter 6 Habitat description richness is lower than the surrounding forest and there is very little evidence within this unit of logging (# of cut stumps).

Blackberry thicket

Agonis swamp shrublands downstream of areas cleared for agriculture are often infested with the weed Rubus affinis or blackberry. These areas are very densely vegetated up to a height of 1.5 metres.

Blackbutt open forest

One of the two habitat units surrounding the Agonis swamp shrublands on the ecotone between swamp and forest which proved difficult to differentiate between statistically. It is located on the lower slopes of the jarrah forest and abuts the swamps. This forest type is dominated by blackbutt (Eucalyptus patens) in the absence of E. megacarpa which is a simple way to identify in the field. It has a much greater leaf litter depth (indicating increased time since fire) than the closely allied bullich-blackbutt open forest. It is unlikely that there is some form of succession associated with age occurring between the two considering the growth period required for each tree species but this difference may be an artifact of fire frequency.

Bullich – blackbutt open forest

The presence of bullich (E. megacarpa) in association with blackbutt characterises this habitat unit and it is located in a similar position to the closely allied blackbutt open forest. 207 Chapter 6 Habitat description

As with blackbutt open forest there is little direct evidence of logging (cut stumps) which reflects the restrictions imposed upon logging operations in close proximity to a watercourse (Anon 1999). The relatively high number of logs however indicates that logging occurs nearby.

Bullich swamp forest

Within the Agonis swamp shrublands are habitat units dominated by a bullich overstorey with the shrubs A. linearifolia, Hypocalymna cordifolium, Thomasia species and

Boronia molloyiae interspersed with the sedge Lepidosperma tetraquetrum. This habitat unit has a high species richness and no logging occurs there.

Cleared area

Cleared areas are totally devoid of vegetation (barring some overhanging eucalpts) and include roads, tracks and conveyor belt lines.

Dry heath

A reasonably open heath occurring near Agonis swamp shrublands dominated by

Leptospermum erubescens. This habitat unit predominates in drier areas to the east of the jarrah forest where Agonis swamp shrublands no longer occur. Very little leaf litter is present and large amounts of bare earth is exposed. The ground is often inundated.

208 Chapter 6 Habitat description

Flooded gum forest

Flooded gum forests occur when topography changes from the broad, flatter areas where the Agonis swamps dominate to slightly steeper, narrower, more incised valleys along larger watercourses. A very low vegetation density until above five metres characterises this habitat unit along with the dominance of Eucalyptus rudis.

Introduced pasture

Grassland (often grazed) that has been cleared of its overstorey.

Jarrah-Marri open forest

The dominant vegetation type throughout the study region occurring in the laterite soils of the Darling Range away from watercourses. Jarrah (Eucalyptus marginata) and marri

(Corymbia calophylla) are the dominant canopy species. This habitat unit has a high species richness and a relatively open shrub layer which progressively declines to three metres and then increases above five metres. Numerous logs on the ground and the presence of cut stumps reveal this is the preferred wood for forestry.

Lepidosperma – Hypocalymna swamp

These densely vegetated (below 1.5 metres) areas are a type of Agonis swamp shrubland without A. linearifolia dominating the canopy. Instead L. tetraquetrum and H. cordifolium dominate.

209 Chapter 6 Habitat description

Paperbark swamp

Also within the Agonis swamp are small pockets dominated by the swamp paperbark

(Melaleuca rhaphiophylla). These areas exhibit high floristic richness and the deep leaf litter suggests they may require long periods without fire to develop.

Peppermint forest

A relatively low forest dominated by peppermint (Agonis flexuosa). It occurs along some watercourses in the southern jarrah forest.

Pine plantation

Monoculture of introduced Pinus species in the canopy but with numerous shrubs from the surrounding jarrah forest encroaching. Nonetheless it is still very open until the canopy reaches above five metres.

Revegetation – dense and sparse

Areas showing evidence of soil furrowing to assist revegetation along with exotic plant species. Two types are classified and these are differentiated based on their vegetation density (Appendix D).

River forest

210 Chapter 6 Habitat description

Occurring only along the steep-sided banks of larger, rivers of the jarrah forest, this habitat type is dominated by Grevillea diversifolia and Banksia littoralis.

Soapbush swamp

Dominated by soapbush (Trymalium floribundum) these swampy areas are low in species richness and have a high leaf litter depth.

Swishbush swamp

A very open vegetation type dominated by swishbush (Viminaria juncea). Low litter and high bare earth and moisture further characterise this low-lying habitat unit and suggest regular periods of flooding.

Tea-tree swamp

Agonis swamp shrublands occasionally give way in slightly drier areas to be dominated by Leptospermum firmum. This habitat unit has a similar structure to Agonis swamp shrublands.

Wandoo woodland

In the drier areas to the east of the jarrah forest, wandoo (Eucalyptus wandoo) becomes the dominant canopy species. Being a woodland it is very open structurally and has a low species richness.

211 Chapter 6 Habitat description

Changes with age since fire within the Agonis swamp shrubland

While the Agonis shrubland habitat units tend to form a continuum, there does appear to be a succession occurring. This is particularly noticeable for vegetation structure (Figure

5). Increased time since fire results in increased vegetation density below one metre and between two and three metres until 24 years after which the structure opens out to a more intermediate level (Figure 5). Initially the density increase occurs fastest in the two to three metre heights but after 14 years the density below one metre also accelerates.

There were significant linear relationships between age since fire at three vegetation heights. At 0.1 m above the ground there is a significant but weak positive relationship with increasing vegetation density and time since fire (R2 = 0.05; n = 171; P = 0.0041). At two metres above the ground there was a similarly weak but significant negative relationship (R2 = 0.07; n = 171; P = 0.0008) and a significant positive relationship exists at five metres above the ground (R2 = 0.36; n = 171; P < 0.0001).

Leaf litter depth also increases in a linear fashion with time since fire (R2 = 0.36; n =

171; P < 0.0001) (Figure 6). Species richness peaks when intermediate vegetation density is attained (10 – 14 years after a fire) however it had declined by the time the long unburnt swamp returned to such densities (Figure 6).

Discussion

The effect of fire on Agonis swamp shrublands

212 Chapter 6 Habitat description

The jarrah forest is susceptible to fire for six months of the year in normal conditions but during drought it is considered possible to have wildfire throughout the year (Wallace

1966). Under Aboriginal fire regimes, the jarrah forest is thought to have burnt at low intens ity at a frequency of every three to four years, however it is likely that wet areas such as swamps regularly escaped fire for longer periods (Wallace 1966; Burrows et al. 1995;

Ward and Sneeuwjagt 1999). Jarrah is considered well adapted to fire (Wallace 1966) as is the surrounding flora (Gardner 1957). The Agonis swamps also show characteristics that suggest they have evolved to cope with such regimes.

The immediate impact of a fire on the vegetation within a swamp is devastating, although patches of unburnt habitat generally remain resulting in a mosaic of age classes.

Immediately after a fire the swamps are an open area of blackened, dead stems. Shortly thereafter resprouting of the Agonis linearifolia begins and vegetation density increases.

Plant species richness peaks between five and 19 years after a fire but is highest between ten and 14. A similar pattern of species richness occurs in the understorey of coastal woodland in where richness peaks around five years after a fire (Fox

1988). Floristically, as well as being species rich, the swamps less than 14 years after a fire appear more diverse (although not quantified) in that they have non-dominant plant species exhibiting higher percentage cover than older swamps which are dominated by one or two species.

Species richness in Agonis swamps declines 14 years after a fire and this decline may begin to plane after 25 years. Again this pattern has been observed in other areas, however elsewhere there is an increase in species richness in long unburnt sites (Fox 1988). This may reflect the regularity with which the jarrah forest burnt prior to European settlement.

213 Chapter 6 Habitat description

Although not quantified, the amount of debris and dead plant material may increase from

12 years after a fire and the proportion of thick, woody stems dominates the vegetation density.

Swamps over 25 years of age are characterised by a few, large Agonis plants with a canopy at least five metres high. The vegetation returns to a more intermediate density similar to 5 – 14 years since fire. Species richness is also low with older swamps dominated by A. linearifolia and Lepidosperma tetraquetrum while any other species present generally occur at very low levels.

As discussed above, the trajectory of succession within the swamp reflects that occurring in other ecosystems (Fox 1988). Such trajectories have been applied to small mammal community succession in those same areas and these show similar behaviour where the community exhibits a circuit before returning to similar compositions at intermediary ages (Fox 1990). This form of succession may also be occurring in the Agonis swamps of the jarrah forest.

Leaf litter depth is a good predictor of the number of years since a fire burnt a swamp as, in the swamps studied here, it increased linearly. This contrasts with studies in New

South Wales which found litter accumulated exponentially, almost attaining a plateau after eight years (Fox et al. 1979). Growth rings of Agonis linearifolia are the most precise predictor of the number of years since fire within the swamp.

214 Chapter 7 Habitat use and preferences

Habitat use and preferences of the quokka Setonix brachyurus

(Quoy & Gaimard 1830) (Macropodidae: Marsupialia) in the

northern jarrah forest of Australia.

Abstract. The habitat use and preferences of the quokka (Setonix bracyhurus) were investigated with particular regard to the status of the species and the overall metapopulation. On a macrohabitat level, quokkas are largely restricted to swamp habitat which is thought to be caused by the species’ high water requirements. This requirement for swamp habitat has been exacerbated by increased predation pressure since the 1930s and may have implications for the conservation status of the species.

On a microhabitat scale, quokkas appear to be habitat specialists, preferring early seral stage swamp habitat that has been burnt within the previous ten years, particularly as part of a mosaic of different age classes. This specific preference may derive from a combination of dietary requirements and predation refuge and these preferences may be linked to the burning regimes employed by the Aboriginal people of the south-west of

Australia. The preference for early seral stages suggests a high rate of population turnover within the metapopulation while it still functioned was natural. Considering the conservation status of the quokka and the lack of dispersal from the few remaining sites in the northern jarrah forest however, there appears to be an urgent need for management intervention. Methods to increase the availability of habitat through burning and continued predator control are discussed.

215 Chapter 7 Habitat use and preferences

Introduction

Metapopulations occur when local populations inhabiting discrete patches are interconnected by dispersing individuals (Hanski and Gilpin 1991). The habitat preferred by a metapopulation can occur as a mosaic of high quality areas interspersed with poorer areas which can lead to the restriction of local populations to within these discrete patches. Species that exist naturally as classic metapopulations (Hanski and

Gilpin 1991) are generally habitat specialists (Smith 1974).

Considering that organisms should use habitat that maximises their fitness (Luck

2002), the identification of the habitat preferences of species existing in metapopulations is crucial in determining the quality of these habitat patches, the carrying capacity of a site and to predict the behaviour of the metapopulation when threatened. Such information can also be used by wildlife management agencies in altering and restoring the habitat to alleviate threatening processes. Habitat use is therefore, a critical facet to wildlife management (White and Garrott 1990).

The massive and widespread decline and extinction of medium-sized mammals in

Australia (Lunney and Leary 1988; Burbidge and McKenzie 1989; Morton 1990;

Lunney 2001) has made it imperative to identify and protect the remaining habitat that these species use. The quokka (Setonix brachyurus) (Quoy & Gaimard 1830) probably originally formed a classic metapopulation with local populations restricted to the swamp shrublands that occupy the broad, flat, upper reaches of creek systems while individuals occasionally mixed with adjacent populations (Hayward et al. In review).

This metapopulation structure is now thought to be in a state of collapse (Hayward et al.

In review). Since the initiation of introduced predator control, survival of adults and

216 Chapter 7 Habitat use and preferences

juveniles appears to be sufficient to allow the populations to increase (Chapter 8), however such increases appear not to have occurred (Chapter 4). With predation unlikely to be the sole cause of the continuing low populations (Chapter 8) it is thought that specific habitat requirements of the quokka may be further inhibiting population growth (Hayward et al. In review). A small-scale study in the 1970s on population sizes at three sites with different burning histories indicated that quokka abundance peaked in recently burnt habitat, before declining by 12 years post-fire and deserting the habitat after 15 years (Christensen and Kimber 1975). This chapter aims to determine the habitat preferences of the quokka at mainland sites in the northern jarrah forest and assess the implications of that on the metapopulation dynamics and conservation of the species.

Methods

Study sites

Digitally imaged aerial photographs (supplied by the Western Australian

Department of Land Administration) were imported into a geographic information system (GIS) (MapInfo Professional Version 5.5; MapInfo Corporation Inc. 1985-1999) and geo-referenced to GPS positioned points of sub-metre accuracy. Within the GIS maps of the habitat of five trapped sites (described in Chapter 4) were created and habitat units were described a priori, based on structural, floristic and other habitat characteristics (Chapter 6). Field surveys were then undertaken to ground-truth these habitat units. Broad or macro- habitat units were differentiated based on their floristics

217 Chapter 7 Habitat use and preferences

using a version of the TWINSPAN computer program that incorporated strict convergence parameters (Hill 1979; Oksanen and Minchin 1997). The Agonis swamp shrubland units were separated to a microhabitat level based on the features they acquired with age since fire and these proved more difficult to separate (Chapter 6).

These units were statistically differentiated by ANOVA of the factor scores identified using principal component factor analysis in the Statview computer program (SAS

Institute 1992-1998) (Chapter 6). Brief descriptions of each habitat type found at the five sites supporting quokka populations are shown in Table 1.

Table 1. Brief description of habitat types found at five quokka sites in the northern jarrah forest. Full descriptions are detailed in Chapter 6.

Habitat unit name Description Agonis swamp The Agonis swamp shrubland habitat units are generally similar in their shrubland habitat units species composition, all being dominated by Agonis linearifolia with Lepidosperma tetraquetrum amongst the ground layer. Species that also occur in these habitat units include Hypocalymma cordifolium, Xyris lacera, Thomasia species and the sedge Gahnia decomposita. These units occur on alluvial soils in the broad, gently-sloping, upper reaches of creek systems (Mulcahy 1967; Mulcahy, Churchward and Dimmock 1972) in the wetter, western side of the jarrah forest (DCE 1980). Freshly burnt swamp The early seral stage of the Agonis swamp is characterised by dead stems (stags) of Agonis linearifolia that exhibit evidence of resprouting at the base of the plant. Very little leaf litter remains following the obviously recent fire and the amount of bare ground is high. Species richness is very low and this is reflected in a relatively low vegetation density. For the first four years following a fire Lepidosperma tetraquetrum and Astartea fascicularlis are among the few species that coexist with Agonis linearifolia. 5 - 9 years since fire After 5 to 9 years following a fire, species richness per quadrat reaches a peak, leaf litter continues to develop and the amount of bare earth and number of stags decline. Vegetation density at ground level is still relatively open but very little vegetation occurs over two metres. 10 – 14 years since fire Between 10 and 14 years following a fire, species richness per quadrat reaches a peak, leaf litter continues to develop and the amount of bare earth and number of stags decline. Vegetation density at ground level reaches an intermediate level but vegetation density between two and

218 Chapter 7 Habitat use and preferences

Habitat unit name Description three metres increases substantially. 15 – 19 years since fire Species richness declines while vegetation density at ground level increases further. Leaf litter depth continues to increase linearly. 20 – 24 years since fire Species richness continues to fall while litter depth increases. Vegetation at ground level and between two and three metres attains its highest density. Greater than 25 years Species richness continues to fall after 25 years following a fire. Leaf since fire litter depth peaks at long unburnt sites. Vegetation density returns to a more intermediate level, akin to that between five and 14 years following a fire. The height of the Agonis swamp shrubland may reach up to 10 to 15 metres.

Allocasuarina forest A component of the jarrah-marri open forest this habitat type is dominated by Allocasuarina fraseriana while possessing a standard jarrah-marri open forest understorey. Species richness is lower than the surrounding forest and there is very little evidence within this unit of logging (# of cut stumps). Blackbutt open forest One of the two habitat units surrounding the Agonis swamp shrublands on the ecotone between swamp and forest. This forest type is dominated by blackbutt (Eucalyptus patens) in the absence of E. megacarpa. It has a much greater leaf litter depth (suggesting increased time since fire) than the closely allied bullich-blackbutt open forest. Bullich-Blackbutt open The presence of bullich (E. megacarpa) in association with blackbutt forest characterises this habitat unit. As with blackbutt open forest there is little direct evidence of logging (cut stumps) due to the proximity to a watercourse (Anon. 1999) but the relatively high number of logs reminds that logging occurs nearby. Bullich swamp forest Occasionally within the Agonis swamp shrublands occur habitat units dominated by a bullich overstorey with the shrubs A. linearifolia , Hypocalymna cordifolium, Thomasia species and Boronia molloyiae interspersed with the sedge Lepidosperma tetraquentrum. This habitat unit has a high species richness and no logging occurs there. Cleared area Cleared areas are totally devoid of vegetation (barring some overhanging eucalpts) and include roads, tracks and conveyor belt lines. Dry heath A reasonably open heath occurring near Agonis swamp shrublands dominated by Leptospermum erubescens. Very little leaf litter is present and large amounts of bare earth is exposed. The ground is often inundated. Introduced pasture Exotic grassland (often grazed) that has been cleared of its overstorey. Jarrah-Marri open forest The dominant vegetation type throughout this study occurring in the laterite soils of the Darling Range on the slopes and ridges away from watercourses. Jarrah (Eucalyptus marginata) and marri (Corymbia calophylla) are the dominant canopy species. This habitat type has a

219 Chapter 7 Habitat use and preferences

Habitat unit name Description high species richness despite its relatively open shrub layer. Vegetation density progressively declines from the ground layer to three metres before increasing above five metres. Numerous logs on the ground the cut stumps reveal this is the preferred wood for forestry. Paperbark swamp Also within the Agonis swamp are small pockets dominated by the swamp paperbark (Melaleuca rhaphiophylla ). These areas exhibit high floristic richness and the deep leaf litter suggesting they may require long periods without fire to develop. Pine plantation Monoculture in the canopy of introduced Pinus species but with numerous shrubs from the surrounding jarrah forest encroaching. Nonetheless it is still structurally very open until the canopy is reached above five metres. Revegetation – dense Defined by having areas showing evidence of soil furrowing to assist and sparse revegetation along with the presence of exotic plant species. Two types are classified and these are differentiated based on their vegetation density. Both result from disturbances associated with bauxite mining.

The area covered by each habitat unit at each site was calculated in the GIS package. A broader habitat distinction was also made in which sites were split into swamp (Agonis swamp shrublands, bullich swamp forest and paperbark swamp) and non-swamp (forest) habitat.

Trapping

Previous trapping revealed five sites in the northern jarrah forest supporting small populations of quokkas (Chapter 4). Trapping data were subsequently used to investigate broad habitat use. At each site, 30 traps were placed inside the swamp and

30 were placed between 50 and 100 metres outside (Chapter 4). The number of captures inside the swamp were compared with the number outside at each site using the chi-square test. More specific habitat use was not investigated because of the use of bait in traps that may have attracted animals to those specific habitat types.

220 Chapter 7 Habitat use and preferences

Radio telemetry

Between November 1998 and November 2000, 58 quokkas (33 males and 25 females) were fitted with radio collars and tracked long enough to obtain stable home ranges (Chapter 5). Triangulation using the Locate II computer program (Nams 1990) was used to locate the position of the collared individuals due to the density of the vegetation (Chapter 5). Only location fixes with error ellipses of less than one hectare were used in the analyses after a preliminary study found such ellipses were within 15 metres of the actual location of the transmitter (Chapter 5). Only temporally independent locations were used (Chapter 5).

Habitat use analysis

The habitat available at each site was digitised into the Ranges V computer program

(Kenward and Hodder 1992) for analysis. The availability of each habitat unit was classed as the proportion of the total area encompassed within the 95 percentile kernel estimate of home range (Chapter 5). All habitat within approximately 200 metres either side of the swamp at each site was included.

The preference for each habitat unit was determined with the Jacobs index (Jacobs

1974) using Ranges V (Kenward and Hodder 1992). Jacobs index is a derivation of the electivity index (Ivlev 1961; Krebs 1989) which is independent of the relative abundance of the habitat type (Krebs 1989). It is calculated by the formula:

221 Chapter 7 Habitat use and preferences

r - p D = r p -+ 2rp

where r is the proportion of a habitat type in an animals’ home range and p is the proportion of locational fixes occurring within that habitat type (Jacobs 1974). When solved for D, the Jacobs index ranges from + 1, indicating maximum preference, to – 1, indicating maximum avoidance with values near zero indicating a habitat type that is used by the species in accordance with its availability within the home range (Jacobs

1974). This method has been used previously for another endangered species, the

Iberian lynx (Felis pardinus) (Palomares et al. 2001). Each habitat unit was preferred or avoided if the mean value of the Jacobs index was significantly different from zero according to single group t-tests (Palomares et al. 2001). Habitat units that showed no variation in Jacobs index (i.e. those always – 1 (avoided)) were unable to be tested with t-tests but were considered and shown in figures as being significantly avoided. It is important to stress that habitat units that occupy a large proportion of the home range area but are not recorded as preferred by the Jacobs index should still be considered as important habitat components. For example, a habitat type that occupies 99% of a home range but in which the animal was only recorded in it on 95% of fixes has a Jacobs index value of only 0.67, despite this habitat type clearly being important to the animal.

Comparing overall habitat preferences statistically was difficult due to differences in habitat composition at each site and differences in the number of habitat types available within each individual’s home range. Consequently, habitat types were grouped into young (less than 9 years since fire), intermediate (ten to 19 years since fire) and mature (more than 20 years since a fire). At least two of these groupings

222 Chapter 7 Habitat use and preferences

occurred at each site except the Chandler Road site which was therefore excluded from this analysis. Two sites had all three habitat groupings. Another method I used to get around this difficulty was the ranking of the habitat preferences for each individual which were then averaged to provide an overall indicator of habitat preference within swamp habitats.

Results

Available habitat types

The dominant habitat type at all sites was jarrah-marri open forest (average at all sites = 66%) with smaller areas of sparse revegetation (9%), bullich-blackbutt open forest (5%), introduced pasture (4%)(Figure 1). Agonis swamp shrubland made up 7% of available habitat at all sites (Figure 1). Rosella Road has the largest area of swamp habitat (15 ha), followed by Kesners, Chandler, Victor Road and Hadfield (Figure 2).

Within these smaller areas of swamp however, the Victor Road and Hadfield sites possess a broad mosaic of swamp habitat types (five) as does the Kesners site, while the

Rosella Road site has two swamp age classes and Chandler Road one (Figure 2).

The dominant swamp vegetation obviously depended upon the individual fire history at each site with the youngest habitat age classes occurring at Victor Road (47% of freshly burnt Agonis) and Hadfield (45% 5 - 9 year old) (Figure 2). Agonis swamp shrubland burnt between ten and 14 years ago was the most common (mean of all sites

= 39%) and widespread swamp habitat type occurring at four of the five sites and small pockets of older swamp habitat units are scattered across each site (Figure 2).

223

100%

75%

50% Total habitat

25%

0% Chandler (231ha) Hadfield (209ha) Kesners (100ha) Rosella (206ha) Victor (83ha)

Swamp * Allocas Bl.OF BullSF BBOF Clear Intr.Past JMOF Dry heath Paper Pine Rev.Dense Rev.Sparse

Figure 1. Percentage of all habitat types available at each site. All Agonis swamp habitats are grouped into Swamp. Fuel ages with each Agonis swamp unit are shown in Figure 2. The following abbreviations refer to each habitat type that are defined in the methods section: BullSF – bullich swamp forest; Paper – paperbark swamp; Bl.OF – blackbutt open forest; BBOF – bullich-blackbutt open forest; JMOF – jarrah-marri open forest; Allocas – allocasuarina forest; Pine – pine plantation; Rev.Dense – dense revegetation; Rev.Sparse – sparse revegetation; Intr.Past – introduced pasture; and Clear – cleared areas.

224

100%

75%

50% Swamp habitat

25%

0% Chandler (11ha) Hadfield (7ha) Kesners (13ha) Rosella (15ha) Victor Rd (8ha) Fresh 5 - 9 Agonis 10 - 14 Agonis 15 - 19 Agonis 20 - 24 Agonis > 25 Agonis

Figure 2. Percentage of each Agonis swamp habitat type available at each site. The legend refers to the number of years since a fire (i.e. 5-9Agonis relates to Agonis swamp shrubland that was burnt between five and nine years ago). 225

Broad habitat use and preferences

There were significantly fewer captures of quokkas from all sites combined outside swamps (37) than inside (232) (c2 = 70.68; d.f. = 1; P < 0.0001). There were significantly more males captured outside swamps (26 of the 37 outside in 144 overall captures of

2 males) compared to females (11/37 in 125 female captures) (c = 4.64; d.f. = 1; P = 0.03).

There is a significant difference in the capture locations between the sites, with Victor Road having only six out of 104 captures outside the swamp as the lowest and Chandler Road

2 with eight out of 17 captures outside the highest (c = 29.60; d.f. = 3; P < 0.0001).

Animals at the unbaited Victor Road site had relatively fewer captures outside the swamp

2 (6/104) compared to the baited sites (31/165) (c = 8.70; d.f. = 1; P < 0.0032).

This massive bias toward the swamp is reduced when we look more generally at the habitat covered by home ranges (Figure 50). The average quokka home range is centred upon the Agonis swamp shrubland vegetation with a lesser amount in surrounding non- swamp (forest) vegetation (Figure 4). The ratio of location fixes in the swamp compared to non-swamp habitats differs between each site with quokkas at the Chandler, Rosella and

Victor Road sites using the swamp more than the non-swamp whereas quokkas at the

Hadfield and Kesners sites have a larger percentage of their home range encompassing the surrounding forest (Figure 4). This variation is reflected in the analysis of variance results that show an interaction between the habitat units and sites (Table 2).

226

Table 2. Analysis of variance of broad habitat use (swamp or non-swamp), based on area covered by home range, with sex and site nested within.

ANOVA d.f. Mean square F-value P-value

Habitat group (swamp / non-swamp) 1 0.411 8.379 0.005

Sex 1 0.004 0.076 0.784

Site 3 0.052 1.065 0.367

Habitat group * sex 1 0.001 0.000 0.998

Habitat group * site 3 0.669 13.634 < 0.001

Sex * site 3 0.005 0.07 0.962

Habitat group * sex * site 3 0.082 1.668 0.178

Residual 118 0.049

There was a significant difference between the use of swamp and non-swamp habitats by quokkas diurnally or nocturnally although there was an interaction that indicated this use of habitat differs between day and night (

227

Table 3). More specifically quokkas spent the majority of daylight hours within the swamp compared to outside (Figure 3). This difference ceased at night when a similar percentage of the home range encompasses swamp and non-swamp habitats.

228

Table 3. ANOVA table of broad habitat use based on area covered by home range nested with a diurnal / nocturnal time factor.

ANOVA d.f. Mean square F-value P-value

Habitat units (swamp / non-swamp) 1 0.549 6.851 0.010

Time (diurnal / nocturnal) 1 0.000 0.000 0.974

Habitat unit * time 1 0.906 11.316 0.010

Residual 120 0.080

100

80

60

40 Habitat use (%)

20

0 Total Diurnal Nocturnal Autumn Winter Spring Summer Swamp Non-swamp

Figure 3. Total, diurnal and nocturnal, and seasonal broad habitat use of quokkas based on home range cover calculated from radio telemetry results. Means with one standard error bar are shown. The overlap between swamp and non-swamp arises due to the variation in area covered by home ranges between individuals.

229

Seasonal broad habitat use also shows significant interactions (Table 4). The use of swamp habitat increases throughout the year while the use of non-swamp (forest) habitat declines from autumn to summer (Figure 3).

Table 4. ANOVA table of broad habitat use of quokkas with season nested within.

Repeated measures ANOVA was not conducted due to the small sample size of continually monitored animals.

ANOVA d.f. å of Mean square F-value P-value squares

Habitat units (swamp / non- 1 3.195 3.195 42.379 < 0.001 swamp)

Season (autumn / winter / spring 3 0.030 0.010 0.131 0.942 / summer)

Habitat unit * season 3 0.636 0.213 2.825 0.041

Residual 174 13.119 0.075

Quokkas from all sites use swamp habitat units more than non-swamp irrespective of sex, site or interactions between the two (Table 5).

Table 5. Analysis of variance results of broad habitat use. Sex and site factors are nested.

ANOVA d.f. å of Mean square F-value P-value squares

Habitat units (swamp / non- 1 13.090 13.090 199.61 < 0.001 swamp)

230

Sex 1 0.070 0.070 1.066 0.305

Site 3 0.331 0.110 1.682 0.178

Habitat unit * sex 1 0.030 0.030 0.459 0.501

Habitat unit * site 3 0.185 0.062 0.941 0.426

Sex * site 3 0.050 0.017 0.257 0.856

Habitat unit * sex * site 3 0.041 0.014 0.209 0.890

Residual 72 4.722 0.066

100 90 80 70 60 50 40 30

Broad habitat use (%) 20 10 0 Chandler Hadfield Kesners Rosella Victor Road Swamp Non-swamp

Figure 4. Broad habitat use (swamp and non-swamp) of quokkas at five sites as defined by radio telemetry.

Specific habitat use and preferences

The problem of determining the overall habitat preferences of quokkas within the swamp was solved satisfactorily by grouping swamp age categories into three major seral

231

stages; young (< 10 years since fire), intermediate (10 – 19 years since fire) and mature (>

19 years since fire). Single group t – tests performed on these categories showed young areas of Agonis swamp shrubland to be significantly preferred by quokkas (t = 4.817; d.f. =

19; P = 0.0001) (Figure 5). In contrast, the intermediate and mature categories were used by individual quokkas in accordance with the proportion of each habitat type existing in each home range (intermediate t = 0.469; d.f. = 50; P = 0.6409; mature t = 0.302; d.f. = 46;

P = 0.7641) (Figure 5).

35% 0.4

30% 0.3

0.2 25% 0.1 20% 0 15%

Habitat use -0.1

10% Habitat preference -0.2

5% -0.3

0% -0.4 Young (< 10 years) Intermediate (10 - 19 years) Mature (> 19 years)

Figure 5. Habitat use and preferences for Agonis swamp habitats grouped into three categories of time since fire. Mean and standard error bars are shown. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance

(negative values). The star above the young age category indicates a significant preference according to t – tests on Jacob’s Index.

232

3

2.5

2

1.5 Mean rank 1

0.5

0 Fresh 5 - 9 y.o. 10 - 14 y.o. 15 - 19 y.o. 20 - 24 y.o. > 25 y.o.

Figure 6. Overall mean swamp habitat preference rank (± s.e.) for quokkas at all sites combined. Values of one indicating most preferred swamp habitat. Mean rank was calculated after the preferences for each swamp habitat unit by each individual were ranked from most (1) to least preferred and then the arithmetic mean and standard deviation of these were taken.

The difficulty in determining overall swamp habitat preferences was also overcome to an extent by calculating the mean and standard error of the preference rank of each habitat unit at each site (Figure 6). The most favoured habitats were the early seral stages of the swamps after which there was a decrease in habitat preference within the swamps with time since fire.

Habitat preference and use at individual sites

233

As discussed in the methods section, statistical tests could not be conducted comparing specific habitat use of quokkas. The actual habitat use and preference of quokkas at each site are discussed below.

Chandler site

At the Chandler site the Agonis swamp shrubland burnt ten to 14 years ago encompassed the majority of the average home range (38%) followed by the bullich- blackbutt open forest (Figure 7). The remaining habitat types occupy less than 15% each of the average home range. In contrast to the habitat use, the bullich swamp forest (BSF) is the most preferred habitat type (t = 9.233; d.f. = 5; P = 0.0003) while the jarrah-marri open forest (JMOF) (t = -4.894; d.f. = 2; P = 0.0393), cleared areas and allocasuarina forest

(allocas) were significantly avoided (Figure 7). The large proportion of home range encompassed by the ten to 14 year old Agonis swamp habitat type meant that substantially more than 38% of locations were required within this habitat type for the Jacobs index to register a preference for it. Nonetheless, the fact that there was no significant selection for or avoidance of this habitat type (and others in similar situations) indicates it is important for quokkas.

234

60% 1.5

50% 1

40% 0.5

30% 0

Habitat use (%) 20% -0.5 Habitat preference

10% -1

0% -1.5 Clear JMOF BBOF Bullich Allocas

10-14yoAg Use Preference

Figure 7. Habitat use and preferences (mean and standard deviation) based on Jacobs index of six quokkas at the Chandler site. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance (negative values) and stars indicate significant preference or avoidance of a habitat type according to single group t- tests.

Hadfield

As with the Chandler swamp the high useage of the five to nine year old Agonis swamp habitat type at the Hadfield site and the minor but insignificant preference for it indicates its importance (Figure 8). Conversely, the bullich-blackbutt open forest (BBOF) occupies a large part of quokka home ranges but is still significantly avoided (t = -3.123; d.f. = 10; P =

0.0108). Other habitat units significantly avoided were the cleared areas, introduced 235

pasture and jarrah-marri open forest. While none of the swamp habitat types were significantly selected for, all were used in accordance with their availability within the home range which indicates their importance.

45% 1

40% 0.8 0.6 35% 0.4 30% 0.2 25% 0 20% -0.2 -0.4 15% Habitat use (%)

-0.6 Habitat preference 10% -0.8 5% -1 0% -1.2 BSF Clear JMOF BBOF Pasture 5-9yoAg >25yoAg

15-19yoAg 20-24yoAg Use Preferences

Figure 8. Habitat use and preferences (mean and standard deviation) based on Jacobs index of 14 quokkas at the Hadfield site. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance (negative values) and stars indicate a significant preference or avoidance of a habitat type according to single group t- tests.

Kesners

Like the Hadfield site, all of the swamp habitat types, including bullich swamp forest

(BSF) were used in slightly favourable accordance with their availability within each home 236

range at the Kesners site (Figure 9). The large useage of ten to 14 year old Agonis swamp shrubland indicates its importance to quokkas at this site (Figure 12). The non-swamp habitat types were all avoided with significant t – tests able to be calculated for jarrah-marri open forest (JMOF) (t = -4.192; d.f. = 17; P = 0.0006); bullich-blackbutt open forest (t = -

19.887; d.f. = 17; P < 0.0001); and dry heath (t = -9.014; d.f. = 6; P = 0.0001) (Figure 9).

Areas of dense revegetation following bauxite mining occurred within the home ranges of quokkas at Kesners but the animals avoided this habitat type.

40% 1.5

35% 1 30% 0.5 25% 0 20% -0.5 15%

Habitat use (%) -1 10% Habitat preference

5% -1.5

0% -2 BSF Pine Clear Paper Heath JMOF BBOF >25yoAg 10-14yoAg 20-24yoAg Revege - Dense

Revege - Sparse Use Preferences

Figure 9. Habitat use and preferences (mean and standard deviation) based on Jacobs index of 21 quokkas at the Kesners site. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance (negative values) and stars indicate a significant preference or avoidance of a habitat type according to single group t- tests.

237

Rosella Road

The lone male at the Rosella Road site had a home range that was dominated by eight to 12 year old Agonis swamp shrubland surrounded by jarrah-marri open forest with lesser amounts of older Agonis swamp (Figure 10). No statistics were conducted due to the sample size but it seems this individual preferred the Agonis swamp shrublands and avoided all other habitat types.

45% 0.4

40% 0.2

35% 0 30% -0.2 25% -0.4 20% -0.6

Habitat use (%) 15%

-0.8 Habitat preference 10%

5% -1

0% -1.2 JMOF Allocas Sparse Cleared Revege- 10-14yoAg 20-24 yoAg

BlackbuttOF Use Preference

Figure 10. Habitat use and preferences based on Jacobs index of the one quokka at the

Rosella Road site. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance (negative values). No statistics were conducted due to the sample size.

238

Victor Road

In contrast to the other sites, where generally only significant avoidance was detected, quokkas at the Victor Road site only exhibited statistically significant habitat preferences

(Figure 11). In this case, the youngest (less than five years since a fire) (t - 7.582; d.f. = 8;

P < 0.0001) and oldest (more than 25 years since a fire) (t - 4.478; d.f. = 6; P = 0.0042) of the Agonis swamp shrubland habitat types were preferred (Figure 13). The cleared areas and the jarrah-marri open forest (JMOF) were significantly avoided by quokkas.

70% 1 0.8 60% 0.6 50% 0.4 0.2 40% 0

30% -0.2 -0.4 Habitat use (%) 20%

-0.6 Habitat preference -0.8 10% -1 0% -1.2 BlOF Fresh JMOF Cleared Pasture >25yoAg

10-14yoAg 15-19yoAg 20-24yoAg Use Preferences

Figure 11. Habitat use and preferences based on Jacobs index (mean and standard deviation) of 11 quokkas at the Victor Road site. The horizontal line separates the Jacobs index relating to habitat preference (positive values) or avoidance (negative values) and

239

stars indicate a significant preference or avoidance of a habitat type according to single group t-tests.

Figure 12. Example of the habitat use of a male quokka (KM2). Location fixes are shown as asterisks, the solid line around the fixes shows the 95 percentile kernel home range and the broken line represents the core range.

240

Figure 13. Example of the habitat use by a male quokka (VM1). Location fixes are shown as asterisks, the pink triple dotted line around the fixes shows the 95 percentile kernel home range and the dotted line represents the core range.

Discussion

While many species with formerly continuous distributions are being turned into metapopulations by habitat fragmentation (Hanski and Gilpin 1991), this is not the case for the quokka. All historical reports link the quokka to discrete patches of dense vegetation

(White 1952; Gould 1973) and the results of this study indicates that the quokka is still restricted to these areas.

Today Agonis swamp shrubland habitat types are still common in the relatively undisturbed jarrah forest and, despite some clearing for agriculture and mining (Chapter 3), this habitat type is considered by me to have only declined marginally in availability in the

241

northern jarrah forest since Europeans arrived in the south-west. Furthermore, the observations of quokkas crossing roads (Chapters 5 & 8) and the relatively few major roads cutting quokka habitat suggests that his habitat has not been significantly fragmented. As such, macrohabitat is not limiting the population growth that was anticipated following the initiation of introduced predator control (Chapter 4). These Agonis swamps are not, and were never, contiguous linear stretches of habitat because of the change in vegetation floristics and structure that occurs when topography changes from the broad, shallow upper reaches of creek systems to the steeper-sided valleys of larger waterways (Chapter 6).

While quokkas may have historically traveled through these less favoured habitat types during dispersal, this appears to no longer occur. This may be a function of strong selective pressure for philopatry following the arrival of the red fox (Vulpes vulpes) (Chapter 5).

This predation pressure is likely to have led to dispersal through corridors acting like demographic ‘sinks’ to the population (Soule and Gilpin 1991) until such dispersal effectively ceased (Chapters 4 & 5). Similar findings of reduced pressure to disperse following predation in sub-optimal areas that may be encountered during inter-patch movements within a metapopulation have been found in Finnish voles (Banks et al. In review). The quokkas’ clear preference for discrete patches of habitat support the metapopulation assertion (Chapters 4 and 5). That movement between patches no longer occurs increases the pressure placed on an already threatened metapopulation (Chapters 4 and 5).

The quokka may not necessarily be as strictly restricted to densely vegetated areas on the mainland (Christensen et al. 1985) because these areas are most favourable to it, rather they may be the habitat least ‘favoured’ by the agent of the quokkas decline (i.e. the fox) 242

(Caughley 1994). The term ‘favoured’ here refers not just to habitat preference but also to foraging success and, in this case, the swamps are probably acting as a refuge for quokkas from fox predation.

Causes of quokka habitat preferences

Habitat selection reflects two contraints: obtaining food and avoiding predation (Lima and Dill 1990). Both of these factors are likely to affect quokkas.

While quokkas show a distinct preference for Agonis swamp shrubland habitat types, they also exhibit a more specific preference for early seral stages therein. This explains the earlier observations where quokkas reached peaks of abundance in swamps that have been burnt less than 12 years previously before deserting the habitat after 15 years (Christensen and Kimber 1975). Quokkas were observed foraging within a burnt swamp less than three months after fire (Christensen and Kimber 1975) and were often located diurnally in the recently burnt areas of the Victor Road site within a year of the fire before increasing their nocturnal utilisation to three years (pers. obs.). The capture of three new males at the

Victor Road site late in the trapping programme indicates these may have been recolonising individuals attracted to the freshly burnt swamp. Unfortunately the genetic study (Alacs

2001) did not ascertain the relatedness of individuals at each site, so this could not be confirmed. Numerous other macropodids have also been recorded preferring early seral stages including the rufous hare-wallaby (Lagorchestes hirsutus) (Lundie-Jenkins 1993), tammar wallaby (Macropus eugenii), woylie or brush-tailed bettong (Bettongia penicillata)

(Christensen 1980), western brush wallaby (M. irma), western grey kangaroo (M.

243

fuliginosus) (Underwood and Christensen 1981), (M. giganteus), red-necked wallaby (M. rufogriseus) (Southwell and Jarman 1987) and (M. rufus) (Denny 1985).

The fact that four of the five sites in the northern jarrah forest where quokkas still exist possess a mosaic of burnt areas within the swamp (Figure 2) is also important. It has been suggested that quokkas feed within freshly burnt swamps but do not become resident for at least a year (Christensen and Kimber 1975). Yet considering the small distances quokkas have been recorded moving (Chapter 5) they are likely to require refuge habitat in close proximity to these foraging areas. Thus, the persistence of older, unburnt habitat within freshly burnt swamps is likely to provide this. Such patchiness seems to be common for recent fires observed in the Agonis swamps although historically this was not the case with the swamp either not burning at all or burning entirely (M.Dillon pers. comm.).

Historically, when the swamps did not burn they formed the boundary of the burn and repeated burns progressively encroached on them, reducing their width and therefore the area of the remaining swamp vegetation (P. de Tores pers. comm.). This difference may be due to the historical fire regimes and control burn rotation length compared to the more frequent burning that is preferred in the jarrah forest today (Burrows, Ward and Robinson

1995). It is important to note that the Chandler Road site, which has only one Agonis swamp age class, had a large section of habitat removed when a dam was constructed immediately upstream of the quokka population (Chapter 2) and quokka presence prior to this disturbance may have been due to the presence of a mosaic of habitats that were subsequently lost.

244

The restriction of quokkas to dense, swampy vegetation may be due to the benefit these areas offer as refuge from climatic extremes (Kitchener 1972; Kitchener 1981), as occurs in the spectacled hare wallaby (Lagorchestes conspicillatus) (Burbidge & Main 1971 in Main and Bakker 1981). Quokkas however, have the ability to survive much greater temperatures than those they are likely to face in the wild (Bartholomew 1954). If the preference to swamps was due to thermal refuge alone, then we would expect quokkas to prefer the denser habitat types within the swamp (i.e. those over 15 years since a fire

(Chapter 6)) or at least more so during summer. While quokkas do not prefer denser habitat types within the swamp (Figure 6) they do utilise more swamp habitat in summer than in any other season (Figure 3). Quokkas on Rottnest Island prefer to take refuge during hot days beneath the sedge Gahnia trifida which exhibits much lower radiant heat loads than the majority of other vegetation on the island (Kitchener 1972). Also agonistic encounters and resource defence over cooler, shady areas have been observed in quokkas on Rottnest Island (Kitchener 1972; Kitchener 1981). During this present study, several quokkas extricated themselves from their radio collars and almost all of these were found beneath the canopy of Gahnia decomposita bushes (pers. obs.). This suggests there may be some degree of behavioural reduction of heat stress but this is still considered a comfort mechanism rather than a survival strategy. Gahnia is reasonably common in the Agonis swamp shrublands and so an absence of shelter is not considered to be a limiting factor to mainland quokka populations.

A more likely reason for the habitat preferences of the quokka is its relatively high water requirements (Main and Yadav 1971; Main and Bakker 1981) which necessitates a close proximity to fresh water throughout the year. This high water requirement probably 245

initially ensured quokkas inhabited the upper reaches of creek systems. The arrival of the red fox in the south-west of Australia in the 1930s (King and Smith 1985) is likely to have reduced the habitat breadth of the quokka to show increasingly more distinct preference for the densely vegetated swamps. Although historically known to inhabit swamps (Chapter 3)

(White 1952), prior to the mid -1930s quokkas browsed pine plantations up to two kilometres from swamps (Stewart 1936). Such behaviour has not been reported since then

(de Tores et al. In prep.) which suggests that quokkas have become increasingly restricted to within the swamps while anti-predatory behaviour has resulted in quokkas spending far less time in the surrounding forest.

Yet the increased restriction to smaller areas may be compounding predator-related problems for the quokka. Limited movements by individual microtine voles in the northern hemisphere results in an increased risk of mortality because of the concentration of predator-attracting odours created (Banks, Norrdahl and Korpimaki 2000). Such features occur in quokkas (Chapter 8) and if quokka faeces and urine also act as attractants then the threat of predation would be increased in, or at least around, the swamps, perhaps reducing their value as refugia. This is highly likely considering the importance olfaction plays in the predatory strategy of the red fox (Saunders and Harris 2000).

The restriction to dense vegetation diurnally may be a further predator avoidance strategy. The long-nosed potoroo (Potorous tridactylus apicalis) inhabits dense heathlands which negate the threat of virtually all birds of prey and several larger terrestrial predators

(Heinsohn 1968). Of the several species of diurnal birds of prey recorded during this study only the wedge-tailed eagle (Aquila audax) is considered a potential predator based on known diets (Chapter 8) (Storr 1965; Barker and Vestjens 1989/90). Only the Tasmanian 246

devil (Sarcophilus harrisii), thylacine (Thylacinus cynocephalus) and quolls (Dasyurus species) were considered actual threats to the based on their habitat use (Heinsohn

1968) and all but the latter are likely to have been prehistoric predators on quokkas

(Chapter 8). This predatory niche has today been taken over by the red fox and, possibly, the cat (Felis catus) (Chapter 8).

It has been previously suggested that quokkas spend greater amounts of time at the periphery of the swamp during the wetter months of winter and spring due to the inundation of the swamp (Chapter 5). This is not reflected in habitat use, possibly because some areas of swamp habitat remain dry throughout these times. This confirms the conclusion that home range sizes are largest in autumn because of the increased movements required to satisfy dietary needs at the end of the hot, dry summer-autumn period and therefore necessitate increased foraging outside the swamps (Chapter 5). There also may be an avoidance of these peripheral or ecotonal habitat types (blackbutt and bullich-blackbutt open forests) by quokkas (Figure 7 - Figure 11). This contrasts with other species that use a mosaic of habitats and often prefer ecotones (e.g. the long-nosed potoroo (Bennett 1993), rufous hare-wallaby (Lundie-Jenkins 1993), red-necked wallaby (Macropus rufogriseus)

(Johnson 1987) and the red-necked pademelon (Thylogale thetis) (Johnson 1980)).

This recent restriction of habitat use following the modern, increased threat of predation is important to remember when investigating the habitat use of all Australian fauna, as these may be substantially different from their original, pre-introduced predator habitat. The rufous hare-wallaby (Lagorchestes hirsutus) today is rarely detected more than 120 metres from densely vegetated pungens habitat, however historically

(undefined but referred to as ancient in the study) it inhabited far more open areas (Lundie- 247

Jenkins 1993). This restriction to areas of refuge was attributed as a response to a decline in productivity of the sites (Lundie-Jenkins 1993) although it may also be a predator avoidance strategy. The restriction of the southern (Ningaui yvonneae) to the dense cover afforded by these Triodia hummocks was attributed to the protection they provided against predation (Bos, Carthew and Lorimer 2002) but this restriction may decrease in the absence of introduced predators. Such altered habitat use in the presence of predators has been shown experimentally to occur in rodent species in South America (Lagos et al. 1995) and rabbits (Oryctolagus cuniculus) (Banks, Hume and Crowe 1999) and eastern grey (Macropus giganteus) in Australia (Banks 2001) and occurs conversely when predation pressure is removed (Kinnear, Onus and Bromilow 1988; Kinnear, Onus and

Sumner 1998). This altered habitat use in the presence of predators is known as niche denial, where niches are still available to support viable populations but occupancy is denied through predation (Kinnear, Sumner and Onus In press). Essentially, this mimics an actual decrease in carrying capacity of a site (Kinnear et al. In press).

The male use of non-swamp habitat more that females is unusual. Their size may necessitate increased dietary requirements that can be fulfilled only outside the swamp.

Alternatively, their increased size compared to females (Sinclair 1998; Hayward et al. In review) has previously been suggested as indicating a reduced susceptibility to fox predation (Chapters 3 and 8) by their approaching the upper limit of the critical weight range of species that are indirectly threatened by introduced predators (Burbidge and

McKenzie 1989). Predation may also explain the significantly reduced frequency with which quokkas at the unbaited Victor Road site were trapped outside the swamp compared

248

to those at the baited swamps and this behaviour may explain the high survivorship there

(Chapter 8).

Specific habitat preferences of the quokka

The preference of early seral (< 10 years since fire) habitat by quokkas seems likely to be diet related considering that the increase in structural density with the less preferred, intermediate seral stages (Chapter 6) would confer both increased climatic and predation refuge. This preference may be due to a higher nutrient content of the new growth - as shas been shown to exist for large macropods in the eucalypt forests of south-eastern Australia

(Catling and Burt 1995). Alternatively it may be due to the ease with which quokkas can obtain freshly growing Agonis linearifolia leaves (considered an important food plant)

(unpubl. data.) (Storr 1964) low to the ground and other shrubs.

While there has been no study of the variation in nutrient availability in the swamp plants with age, the seasonal variation in nutrients in plants growing within the swamp is minor with the majority (including A. linearifolia, Astartea fascicularis, Lepidosperma tetraquetrum and Thomasia species) scarcely changing in protein or water content throughout the year (Storr 1964). Surrounding the swamp there is always some species exhibiting higher levels of protein and water throughout the year (Storr 1964). The early seral preference of quokkas may be due to the rapid uptake of nutrients by swamp vegetation following their breakdown during fire. This may be a similar type of relationship to that between the abundance of arboreal, folivorous marsupials and soil

249

fertility in Australia’s south-eastern forests (Braithwaite, Turner and Kelly 1984; Cork and

Catling 1996).

This dietary-derived habitat preference is also probably linked to the persistence of long unburnt sections within the swamps which are likely to provide the refuge that is becoming increasingly apparent as being important to quokka persistence at a site following a fire. The persistence of unburnt patches in swamps is crucial considering the potential increase in predation rates following a fire (Christensen 1980; Calver and Dell 1998).

Another alternative is that quokkas have evolved to cope with the frequent, low- intensity fire regimes implemented by Aborigines (Ward and Sneeuwjagt 1999).

Aborigines have been regarded as doing little to suppress fires (Wallace 1966) but considering the threat to survival from large wildfires and the subsequent vast areas of devastated land it seems imperative that practices to minimise uncontrolled fire were implemented. These dangers were evident during the 1961 Dwellingup fire in Western

Australia which covered over 24 kilometres and almost 150,000 hectares in 15 hours

(Wallace 1966). With such rates of spread and large areas burnt, the threat of perishing during the fire is as great as the threat of starving following it. Early settlers to Western

Australia report the use of fire by Aborigines with John Lort Stokes stating:

“on our way we met a party of natives burning the bush, which they do in sections every year.

The dexterity with which they manage so proverbially a dangerous agent as fire is indeed astonishing” (referred to in Gardner 1957).

Such frequent burning regimes meant that fuel loads around highly populated areas rarely built up to levels that would sustain high intensity fires and as such was a fire control

250

method. Furthermore, Aborigines hunted many species using fire, including the quokka, which they hunted by directly burning the swamp and spearing animals as they fled the flames (Gardner 1957; Gould 1973). However, whether Aborigines burnt the swamps at a frequency that coincided with peaks in quokka abundance or whether quokka abundance peaked in accordance with the Aboriginal hunting regimes is unknown. This direct ignition within the swamp contrasts with modern day low intensity control burns that generally burn the swamp edges but infrequently the entire swamp without leaving a mosaic of age classes

(P. de Tores pers. comm.).

In the absence of the red fox, the quokkas on Rottnest Island exhibit a resilience and a capacity to rapidly respond to habitat changes arising following fire. Following a major fire on the island in 1956 the vegetation composition and structure altered (Pen and Green

1983). This was exacerbated by overgrazing by the quokkas (Storr 1963). Despite this, the quokka population has increased and is now dependant on the winter growth of plants

(Wake 1980). Clearly, under circumstances without external population limiting forces, the quokka is well able to cope with habitat change.

The preference for a mosaic of habitats may also be due to these Aboriginal burning practices. The higher frequency of burning conducted by Aborigines in the jarrah forest would have reduced the amount of accumulated leaf litter and debris, thereby starving the fire and slowing its rate of spread. This would have provid ed a greater propensity for areas to remain unburnt as smaller fire breaks and lower water content of plants would be needed to divert such fires. The swamp, particularly areas close to the centre, would therefore be far less susceptible to entirely burning, if fire did indeed penetrate it, and a mosaic of ages would be created providing the quokka with refuge and foraging areas. 251

A test of the importance to fauna of a vegetation mosaic was conducted on Barrow

Island where golden bandicoots (Isoodon auratus), common brushtail possums

(Trichosurus vulpecula) and burrowing bettongs (Bettongia leseur) were not affected by the presence of a mosaic of seral stages in the absence of predation (Short and Turner 1994).

In the complete absence of predators, a mosaic of seral stages may also proved to be irrelevant for quokkas.

The impact of habitat on local quokka populations

Having determined that more recently burnt habitat types (< 10 years) are most preferred (Figure 6) we would anticipate that the sites with the largest area of these (Victor

Road and Hadfield) would support the greatest population density. Trapping revealed

Hadfield does support the highest population density (4.3 individuals per hectare) (Chapter

4) as expected, however the Victor Road site has a much lower population density (1.1)

(Chapter 4). We would also anticipate that sites with the greatest mosaic of swamp habitats would possess higher population densities, assuming the spatial scale of the mosaics is similar. The Hadfield, Kesners and Victor Road sites each have five swamp habitat types.

The former two also possess the highest density but again the Victor Road site is anomalous (Chapter 4). This discrepancy may be due to the presence of introduced predator control through the use of poison fox baits at the Hadfield site whereas the Victor

Road remains unbaited (Hayward et al. In review). The Victor Road site also has the largest area of long unburnt swamp (Figure 2) and this may have allowed this population to

252

persist through the refuge offered there following fire while the regenerating swamp nearby provided high nutrient forage.

Quokkas at Hadfield and Kesners spend more time outside the swamp than inside relative to other sites (Figure 4). This also may reflect the reduced predation pressure acting upon each population as the broader area surrounding these two sites is baited for introduced predators at a frequency of at least double of the Chandler, Rosella or Victor

Road sites (Chapters 2 & 4). There is no other obvious distinguishing feature of the available habitat at each site to provide alternative explanations (Figure 1).

Implications for the status of the quokka

This increased restriction to the swamps creates significant implications for the status of the quokka. The area of occupancy is used to determine the status of threatened species

(Hilton-Taylor 2000) including the quokka (de Tores et al. In prep.) and is defined as the area occupied by a taxon that excludes unsuitable areas (Hilton-Taylor 2000). If the forest surrounding Agonis swamp shrublands is reclassified as unsuitable following the arrival of the fox then the quokka may now be classed as under a greater threat of extinction than previously thought. For example, the quokka was only known to occur at 25 sites on the mainland in 1996 (Maxwell, Burbidge and Morris 1996) and in 2000 at least two of these had gone extinct and two others had extremely small populations while only one new population was located (Chapter 4). If each habitat that supports a quokka population is considered as the same size (as there is no link between swamp length and population size

(Chapter 4)), then this represents a decline of 4% in four years and possibly up to 12%.

253

Other more detailed surveys of quokka populations suggest the decline is even more disturbing. Between 1990 and 1999, 79 local quokka populations were identified, however by 2000 only 35 of these were considered to still support quokkas (Chapter 3). This represents a decline of 56% in under ten years which would classify the quokka as endangered according to IUCN guidelines (Hilton-Taylor 2000) rather than vulnerable, as it is currently listed (de Tores et al. In prep.). The same considerations are probably warranted for other threatened species occurring as metapopulations or at least restricted to discrete patches.

Conclusions

The quokkas preference for early seral stage habitat and the suggestion that the habitat breadth of the quokka may have declined following an increase in predation pressure from introduced predators add to the concerns about the persistence of the overall metapopulation. The quokkas’ preference for early seral stage habitat (< 10 years since fire) means that the turnover rate of local populations is likely to always have been high, with animals emigrating from, or not surviving in, older swamps. Having said that, it is unlikely that the Agonis swamp habitat types could be considered as a sink habitat (Pulliam

1988) but rather as a mobile scale of habitat which, at young ages provides for a source population but with aging this tends toward a sink. With greater restriction to within discrete habitat patches and less movement between patches, it is likely that the collapse of the metapopulation, suggested previously through small population sizes (Chapter 4) and philopatry (Chapter 5), is imminent or has already occurred.

254

On a positive note, this research provides wildlife management agencies with specific options for managing habitat for the quokka. Fire has become an increasingly important tool for wildlife managers throughout Australia (Lundie-Jenkins 1993). The use of fire to create a mosaic of burnt areas seems likely to improve the quality of the habitat

(Christensen and Kimber 1975). While increasing the quality of the habitat matrix generally buffers against global extinction, modeling suggests that the threat of creating a metapopulation within an existing source population actually increases the probability of the simultaneous extinction of all sub-populations (Vandermeer and Carvajal 2001) due to spatial correlation (Hanski 1999). This is unlikely to occur in the small habitat patches occupied by the quokka on the mainland but needs to be carefully considered prior to implementing habitat management strategies that aim to increase dispersal.

Controlled or hazard reduction burning regimes targeting the quokka should be on maximum ten year rotation and probably more frequent. Most importantly however, these fires should be patchy to ensure areas of refuge and forage are available to surviving quokkas. It may also be beneficial to use very small, specific burns at extant quokka sites to minimise direct mortality. Nearby swamps could be burnt on a larger scale but at the same recommended frequency to increase their attractiveness if dispersal occurs once extant sites attain high densities.

The absence of dispersal is a further issue of concern (Chapter 5). Creating corridors of specific habitat type will not alter the likelihood of dispersal in quokkas, however corridors of reduced predation pressure may.

255

It is important to remember that these recommendations are specifically for the benefit of the quokka. There are several other species occurring in sympatry with the quokka that possess different habitat requirements to those outlined here (Russell and Rowley 1998).

This study provides no indication as to the optimal annual timing of burning for quokkas although cooler spring burns may further slow the rate of spread and thereby the number of deaths caused by fire. Radio telemetry investigating the survivorship and movements of quokkas during and after a fire would provide a great deal of useful information.

256 Chapter 8 Mortality and survivorship

Mortality and survivorship of the quokka Setonix brachyurus

(Macropodidae: Marsupialia) (Quoy & Gaimard 1830) in the

northern jarrah forest of Australia.

Abstract. The potential for a threatened species to increase in abundance following the initiation of predator control was investigated by assessing the causes of eight known deaths of 58 radio collared quokkas (Setonix brachyurus) using characteristic features on carcasses. Despite the predator control, predation was still the main cause of death followed by road kills. The non-parametric Kaplan-Meier method modified for staggered entry of individuals was used to estimate survivorship. Males and females were affected differently from each mortality cause, however their overall survivorship did not differ significantly. Individuals alive at the beginning of the 25-month study had a 61% chance of surviving to the end. This represented an average annual survivorship of 0.81. Quokka deaths appear to be occurring in accordance with the size of each population. There was no significant difference in survivorship between adults and juveniles or for each year. A generalised model highlighted grouping with conspecifics, increased home range size and maximising time spent within refuge-providing vegetation as features that minimise the likelihood of quokkas dying. A simple life-history model suggests that current rates of adult and juvenile survivorship can result in allow population recovery. Pouch young mortality has apparently inhibited the anticipated quokka population increase since the initiation of predator control and the expulsion of pouch young by females when threatened with predation may be a primary predator avoidance strategy of the species.

256 Chapter 8 Mortality and survivorship

Introduction

Survival is a critical life-history process (Krebs 1989) particularly for threatened species. The interaction between births, migration and survival directs the growth of a population. For species existing as metapopulations, the extinction of local populations is a natural process and, assuming colonisation and extinction rates are in equilibrium, of little concern (Harrison 1991). When a metapopulation is in a state of collapse, the extinction of these local populations is of much greater concern, particularly when very few extant populations are known.

The quokka, Setonix brachyurus (Quoy & Gaimard 1830), is a threatened macropodid marsupial that probably once existed naturally as a classic metapopulation (Hayward et al.

In review). It is a small wallaby that is endemic to the south-western corner of Australia

(Chapter 3). The species suffered a dramatic decline in the 1930’s, largely due to predation from the introduced red fox (Vulpes vulpes), although habitat alteration and disease have also been implicated (Chapter 3) (White 1952; Cook 1960). This led to the collapse of the metapopulation structure into a non-equilibrium state (as described by Harrison 1991) such that the conservation status of the quokka is now listed as vulnerable (Hilton-Taylor 2000).

On Rottnest Island the species suffers seasonal mortality over summer (Main 1959) that is attributed to protein starvation that arises after feeding on nutrient-deficient succulents in attempts to derive sufficient moisture to survive through the dry season (Storr 1964b). This chapter reports on both the proximate and distal causes of mortality and estimates survivorship for quokkas from sites within the northern jarrah forest metapopulation on the

Australian mainland.

257 Chapter 8 Mortality and survivorship

Methods

General procedures

Quokkas were captured at five sites (Chandler, Rosella Road, Kesners, Hadfield and

Victor Road) as outlined in Chapter 4. Captured animals were sedated (Hayward et al. In review) and a radio transmitter was fitted (Chapter 5). Radio collars (Biotrack, Institute of

Terrestrial Ecology, Wareham, UK) were configured to maximise signal strength and battery life while minimising weight. The collars incorporated movement sensitive circuitry with a 2½ hour immobile period required to trigger the mortality pulse frequency of approximately 110 pulses per minute compared to 55 pulses per minute in ‘live’ mode.

Radio telemetry was conducted in accordance with the methods outlined in Chapter 5.

Between November 1998 and November 2000 animals were regularly monitored with an intensive period between March 1999 and May 2000 that was designed to allow direct comparison between the survivorship of animals at each site. This comparative monitoring period occurred when all sites had collared animals and ceased when collar removal began.

The location of each collared individual was determined at least monthly and generally more frequently (up to 4 times per week).

For animals whose collar or remains were recovered, the cause of death was determined by characteristic markings on the carcass (Augee et al. 1996; de Tores 1999).

Predation was attributed to the red fox if there was minimal marking on the collar or if marked showing evidence of large canid teeth; if the gut had been eaten or cached in an excavated hole; and if the fur and uneaten remains were cached (Augee et al. 1996; de

258 Chapter 8 Mortality and survivorship

Tores 1999). Predation was attributed to feral cats (Felis catus) if the carcass was mangled; if there was evidence of bite marks to the back of the skull which often included the removal of the brain; if the gut remained near the carcass; and if there was no evidence of caching or simple covering of the carcass with debris (Augee et al. 1996; de Tores 1999).

The western quoll or chuditch (Dasyurus geoffroii), although present on each site, was not considered a potential predator of quokkas considering its largely insectivorous diet and it only being known to prey on mammals significantly smaller in size than quokkas

(Soderquist and Serena 1994). Dingoes (Canis lupus dingo) were not considered as potential predators of the quokka because of their scarcity in the northern jarrah forest

(pers. obs.; M. Dillon pers. comm.). Nonetheless, carcasses were carefully examined for evidence that could be attributable to chuditch or dingoes.

Raptor predation was characterised by carcasses located below trees and intact skulls with skin peeled back and soft tissue removed (Augee et al. 1996). The regular monitoring meant that predation by carpet python (Morelia spilota) would have been identified by the presence of the collared animal still within the snake (Augee et al. 1996). Other evidence of python predation would have been the collar found within a regurgitated pellet containing crushed or fractured bones. Road killed individuals were located on road verges, had obvious massive trauma and showed no evidence of predation. Carcasses not able to be satisfactorily placed in one of these categories were classed as unknown mortality events.

Carcasses with minimal decomposition or trauma were lodged at the Museum of Western

Australia, Perth.

Survival estimates

259 Chapter 8 Mortality and survivorship

Survivorship is the probability of an animal in a population staying alive for a finite period from the beginning of a study (Pollock et al. 1989a). Several different methods that estimate survivorship were initially investigated. Those that determined between period or seasonal survivorship (e.g. those in Program Mark (White and Burnham 1999)) were dismissed due to the high seasonal variability in capture/recapture rates and low numbers of mortality events. The Kaplain-Meier method (product limit estimator) was consequently used as the assumptions of the alternative parametric estimators were not always met in this study. The Kaplain-Meier method is non-parametric and does not assume normality, equal variances or the exact date of death to be known (Pollock et al. 1989a; Kendall and Pollock

1992). As trapping was assumed to be unbiased and almost all captured individuals were collared, these individuals were considered a random samplfe of the population. Other assumptions satisfied were that survival times for each individual were independent and that censoring was random (Pollock et al. 1989a; Pollock et al. 1989b). The concept of

Kaplain-Meier estimates have been extended to allow staggered entry of animals whereby individuals can enter the study at different times while the time variable is measured from the collaring date of the first group of animals (Pollock et al. 1989a). Any newly collared animals are assumed to have the same survival rate as those previously collared (Pollock et al. 1989a).

Fifty-eight animals from the five populations were used in the survival estimates. An additional eight individuals were excluded from the analyses as their collars ceased transmitting within the recommended seven day conditioning period (Pollock et al. 1989a).

Two deaths associated with trapping (Chapter 4) were excluded (right-censored) from the analyses as these would negatively bias the survival estimates (Pollock et al. 1989a).

260 Chapter 8 Mortality and survivorship

Animals whose collar ceased transmitting but whose fate was unknown were classified as censored. Considerable effort was made to locate censored individuals in order to reduce the confidence intervals associated with the survivorship estimates. Although the majority of censored animals were later recaptured without their collars they were still classified as censored in the analyses to avoid bias (Pollock et al. 1989a; Pollock et al. 1989b).

Survival estimates from the Kaplain-Meier method were compared using the log-rank

(Chi-square) test (Pollock et al. 1989a). This test takes into consideration the assumption that censoring is random but may be violated by a predator killing an animal and simultaneously destroying the transmitter (Pollock et al. 1989a). Consequently, the most conservative modification of the log-rank test (with a modified variance of the number of deaths) (Pollock et al. 1989a) was used in this study. Comparisons between survival estimates at the time of the final sample were conducted using the normal test statistic equation (Pollock et al. 1989a). The log-rank test compared survival functions in their entirety while the approximation to the normal (Z) test compared the survival curves at the end of the 15 month comparative period. The probability for accepting or rejecting null hypotheses were obtained from tables in (Zar 1996). Individual comparisons were not conducted on the Rosella Road site as the only collared animal was censored after eight months.

Modelling the causes of mortality

Generalised models (Nelder and Wedderburn 1972; Hastie and Tibshirani 1990) were used to identify features of the ecology of quokkas that increased their chances of dying.

Using alive/dead as the binomial response variable with a logit link function (Yee and

261 Chapter 8 Mortality and survivorship

Mitchell 1991) , a generalised linear model (GLM) was run on the variables listed in Table

1.

Table 1. Variables used in the models of features affecting individual quokka survival.

The name, description and range of each is listed. An asterisk identifies variables that showed partial correlations that were subsequently excluded from the analysis.

Variable name Description Range of values Sex A categorical variable separating males and females. 1(X) – 2(C) Body mass The mass of quokkas in grams was used to determine 800 – 4,787 if the dominance status of each individual somehow affected the chances of dying. A dominance hierarchy occurs in Rottnest Island quokkas and is based on body mass (Packer 1969). Tail circumference* The circumference of the base of the tail in 47 – 101 millimetres. This measure is considered to be an indicator of body condition (Bakker and Main 1980), but was highly correlated with body mass. Body condition Calculated by dividing the cube root of body mass by 0.11 – 0.16 the pes length, this measure is also considered to indicate the condition of an animal (see discussion in Chapter 4). Distance to nearest The distance in metres from the range centre of the 10 – 600 male study individual to the nearest male quokka. This was calculated in the Ranges V computer program (Kenward and Hodder 1992) and then measured in the MapInfo Professional Version 5.5 computer package (MapInfo Corporation Inc. 1985-1999). Both this and the distance to the nearest female may indicate some sort of dominance hierarchy where less dominant individuals or isolated individuals may inhabit more predation-prone habitats. Distance to nearest As above but to the nearest female. 10 – 600 female Interfix distance The average distance in metres between successive 50 – 300 location fixes was used to investigate whether animals that move a lot are more susceptible to mortality events compared to those that are more sedentary. This was calculated in the Ranges V computer program.

262 Chapter 8 Mortality and survivorship

Variable name Description Range of values Home range The area in hectares that is covered by the 95 0.3 – 18.0 percentile kernel home range contour calculated in the Ranges V computer program (Chapter 5). Home range size is generally negatively related to population density (Chapter 5), however the inclusion of home ranges calculated on only ten fixes in this study allowed all dead individuals to be included yet meant this relationship no longer existed. The home range variable was used to investigate the impact of an individuals’ movements on mortality susceptibility. Percentage of fixes The percentage of location fixes for each individual 11 – 100 within the swamp that occur within the swamp habitat types (Chapter 6b). The denser vegetation here may convey some protection from predation due to the refuge it offers (Chapters 3, 6b and 8). Population density The number of quokkas per hectare of swamp 0.07 – 4.30 (calculated in Chapter 4) was used to investigate whether mortality was density-dependent. Baits * The total number of baits laid in the vicinity of each 0 – 37 quokka swamp was intended to be used as an indicator of the susceptibility of quokkas to predation, however the high correlation between it and population density excluded it. Roads The number of kilometres of roads and tracks crossing 1.6 – 4.0 or adjacent to the quokka swamps was intended to be used as an indicator of the susceptibility to road kill, however the high correlation between it and population density precluded this.

Stepwise addition and deletion of variables was employed to determine which variables explained the majority of the deviance in the chances of quokkas dying. Variables were removed if their Cp statistic was lower that that of the current model until the final GLM had lower Cp statistics than the remaining variables (Hastie and Tibshirani 1990). Akaike’s information criterion (AIC) (Akaike 1973, 1974) was used to measure the goodness of fit of each of the models (Gill 2000) with models having the lowest AIC fitting the data best

(Burnham and Anderson 2001). Models with a delta (D) AIC, calculated by subtracting the

263 Chapter 8 Mortality and survivorship

AIC of the preferred model from each other model, of less than two have substantial support while those with a D AIC greater than seven have no support (Burnham and

Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and are essentially the probability that the model is the Kullback-Leibler best model for the data

(Burnham and Anderson 2001).

The amount of deviance explained by the overall model is described by D2 (Hastie and

Tibshirani 1990) which is calculated by the formula:

2 (Null deviance - deviance) x 100 D = . Null deviance

The percentage deviance is analogous to R2 in regressions (Yee and Mitchell 1991).

The GLMs that were highly supported were then compared to a generalised additive model (GAM) to determine whether cubic spline curves, with local behaviour similar to kernel smoothers, better fitted the data. Analysis of deviance was used to determine if the spline curve fitted the data better than parametric terms (Hastie and Tibshirani 1990) which are preferable over non-parametric curves to maximise parsimony (Yee and Mitchell 1991).

Table 2. Mortality causes of the 58 collared quokkas with their location site and timing in the northern jarrah forest.

Fate Males Females Sites and approximate date of death Predator Fox 0 2 Hadfield (20/5/99); Kesners (26/8/99) Cat 0 1 Kesners (16/12/99) Road kills 2 0 Kesners (29/8/99 and 15/3/00) Unknown 2 1 Hadfield (2/4/99); Kesners (1/10/00); Chandler (27/1/99)

264 Chapter 8 Mortality and survivorship

Results

Overall survivorship

The 58 collared individuals were monitored over the 25 month study period with eight known mortality events occurring (Table 2). When the number of individuals at risk each month were added for the 25 month study we get a total of 450 risks of mortality (termed n). Of the known causes of death, predation (3) was the major cause closely followed by road kills (2) (Table 2). In the absence of other evidence, the three unknown mortality causes may also be attributable to predators. The results from the small sample indicate that these two causes of death affect the sexes differently, with predation only acting upon females while only males at the Kesners site were killed in road accidents. Despite the existing predator control programme (de Tores 1994, 1999), foxes and cats still prey on quokkas. Deaths were recorded throughout the year (Table 2).

The Kaplain-Meier survivorship estimate for all quokkas for the entire 25 month study was 0.6065 (95% Confidence Intervals (CI) = 0.3429 – 0.8702) (Figure 1). This equates to an annual survivorship of 0.8111. This means that an individual alive at the beginning of the study had a 61% chance of surviving for 25 months or an 81% chance of surviving for a year. There was no significant difference between the survival functions between years for all sites combined (November 1998 until October 1999 and November 1999 and October

2000) (

265 Chapter 8 Mortality and survivorship

Table 3).

1.0

0.8

0.6

0.4

0.2 Kaplain-Meier Survivorship (S[t]) 0.0 Jul-99 Jul-00 Apr-99 Oct-99 Apr-00 Oct-00 Jan-99 Jun-99 Jan-00 Jun-00 Feb-99 Mar-99 Feb-00 Mar-00 Nov-98 Dec-98 Aug-99 Sep-99 Nov-99 Dec-99 Aug-00 Sep-00 Nov-00 May-99 May-00

Figure 1. Kaplain-Meier survivorship plot of quokkas from all sites combined over 25 months beginning 8th November 1998. Confidence limits (95 percentile) are shown as dashed lines. The comparative period began in March 1999 and extended to May 2000.

266 Chapter 8 Mortality and survivorship

Table 3. Log-rank and normal tests of survivorship. Total number of animals collared in each comparison is shown in brackets. The Bonferroni correction was applied to the comparisons between sites yielding a significant probability of 0.008.

Log -rank test Z statistic Comparison À2 d.f. Probability Z d.f. Probability All sites 7.2551 4 0.25>P>0.10 2.46 4 0.75>P>0.50 Hadfield (17) v Kesners (26) 2.4916 1 0.25>P>0.10 3.07 1 0.10>P>0.05 Hadfield (17) v Victor Road (10) 0.5556 1 0.50>P>0.25 1.06 1 0.50>P>0.25 Hadfield (17) v Chandler (4) 0.4444 1 0.75>P>0.50 1.06 1 0.50>P>0.25 Kesners (26) v Victor Road (10) 4.6486 1 0.05>P>0.02* 4.92 1 0.05>P>0.03 Kesners (26) v Chandler (4) 1.2437 1 0.50>P>0.25 4.92 1 0.05>P>0.03 Victor Road (10) v Chandler (4) a a Males (33) v Females (25) 0.1781 1 0.75>P>0.50 1.14 1 0.50>P>0.25 Adults (48) v Immatures (10) 0.9699 1 0.50>P>0.25 2.61 1 0.25>P>0.10 11/1998-10/1999 v 11/1999-10/2000 0.2849 1 0.75>P>0.50 a no mortality events were recorded at either site during the comparative period and so no statistical comparison was possible.

267 Chapter 8 Mortality and survivorship

1.0

0.8

0.6

0.4

0.2 Kaplain-Meier Survivorship

0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Months Hadfield Victor Road Kesners Chandler Rosella

Figure 2. Survival functions of quokkas at the five sites in the northern jarrah forest between March 1999 and May 2000. The Rosella and Chandler sites were conservatively assumed to have a survival function of 0 and a variance of 1 after all collared animals were censored.

1.0

0.8

0.6

0.4

0.2

Kaplain-Meier Survivorship Males Females 0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Month

Figure 3. Survival functions of male and female quokkas in the northern jarrah forest between March 1999 and May 2000.

268 Chapter 8 Mortality and survivorship

1.0

0.8

0.6

0.4

0.2 Adults Sub-adults Kaplain-Meier Survivorship

0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Months

Figure 4. Survival functions of adult and immature quokkas in the northern jarrah forest between March 1999 and May 2000.

Comparative period of survivorship (March 1999 – May 2000)

Quokkas at each site were intensively monitored over a 15 month comparative period from March 1999 to May 2000 to allow comparison between sites. There was no difference between the survival curves of all sites combined (Figure 2) (log-rank test) or for the final survival estimate (Z statistic) (

269 Chapter 8 Mortality and survivorship

Table 3). Testing between the Chandler and Victor Road sites was not conducted because no mortality events occurred at either during the comparative period.

There was no significant difference between the survival curves or estimates of males

(sum of monthly individuals at risk n = 155) and females (n = 163) (Figure 3) (

270 Chapter 8 Mortality and survivorship

Table 3). Males had a final survival estimate of 0.8167 (95% CI = 0.6100 – 1.0200) while that for females was 0.7545 (95% CI = 0.5335 – 0.9754). There was also no significant difference between the survival curves and estimates of adults (n = 283) and immatures (n

= 58) (Figure 4) (

271 Chapter 8 Mortality and survivorship

Table 3). Adult survivorship was 0.7653 (95% CI = 0.5891 - 0.9416) during the comparative 15 months and during that period no collared juveniles died.

Generalised models of the features that exacerbate the chance of mortality

The preferred GLM (GLM 7) had a D AIC nine units less than the null model after the stepwise addition and removal of variables (

272 Chapter 8 Mortality and survivorship

Table 4). The equation of the preferred GLM was:

Dead / Alive ~ -0.004792611 Distance to nearest female + 0.2422600 Home range size +

0.03632949 Proportion of fixes within the swamp – 0.3212067.

This means that the likelihood of mortality is:

· decreased with proximity to the nearest female conspecific and increases for isolated individuals;

· decreased with increasing home range size and so was increased for sedentary individuals; and

· decreased with the proportion of time spent within the refuge of the dense swamp vegetation (Figure 5).

GLM 6 also had strong support with a D AIC of 0.9384 (

273 Chapter 8 Mortality and survivorship

Table 4). This model included body condition whereby individuals in poorer condition, surprisingly, had less chance of dying (Figure 5).

GLM 7 has an Akaike weight of 0.446, which means there is a 44.6% probability that this model is the Kullback-Leibler best model for the data (

274 Chapter 8 Mortality and survivorship

Table 4) (Burnham and Anderson 2001). Together with the other model with substantial support (GLM 6), these two models have a 72.5% probability of best fitting the data.

275 Chapter 8 Mortality and survivorship

Table 4. Model selection statistics for dead/alive quokkas. The generalised linear models (GLM) are numbered sequentially from the null model with the Akaike Information Criterion (AIC), the delta (D) AIC, the number of variables in the model (K) and the Akaike weights. These statistics are also presented for the generalised additive model (GAM) of the best GLM. Models with a lower AIC are deemed best fitting the data (Burnham and Anderson 2001). Models with a D AIC of less than two (bolded) have substantial support while those with a D AIC greater than seven have no support (Burnham and Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and is essentially the probability that the model is the Kullback-Leibler best model for the data (Burnham and Anderson 2001). Although using a forward and backward stepwise selection procedure, the model did not add any variables once they were removed.

Model AIC DAIC K Akaike weights 1. Null model Dead/Alive ~ Sex + Body mass + Body 56.6113 9.0799 9 0.0048 condition + Distance to nearest male + Distance to nearest female + Distance between successive location fixes + Home range size + Proportion of fixes inside the swamp + Population density 2. Null model with Body mass deleted 54.6115 7.0801 8 0.0129 3. Model 2 with Distance between fixes deleted 52.6342 5.1028 7 0.0348 4. Model 3 with Distance to nearest male deleted 51.0845 3.5531 6 0.0755 5. Model 4 with Population density deleted 49.7505 2.2191 5 0.1471 6. Model 5 with Sex deleted 48.4698 0.9384 4 0.2790 7. Model 6 with Body condition deleted 47.5314 0 3 0.4460 GAM on GLM 7 72.8930 25.3616 3 0.0001

276 Chapter 8 Mortality and survivorship

a. b.

c. d.

Figure 5. Plots of the partial linear fitted curves of the variables in the final generalised linear model (a – Distance to nearest female, b – Home range and c – Proportion of fixes inside swamp) as well as the variables included in the other model that showed substantial support (d – Body condition). In each case, the variable is plotted while all other variables are held constant. Only the proportion of fixes inside the swamp and home range size variables exhibited a positive relationship.

277 Chapter 8 Mortality and survivorship

The preferred model (GLM 7) explained only 15.1% of the deviance of the data and the addition of the body condition variable only explained a total of 19.2% (Table 5).

Table 5. Variation explained by the preferred GLM and the GAM of that model.

Lower residual deviances indicate a better model. The D2 statistic indicates the amount of variation in the data explained by the model and is analogous to the R2 of regression.

Model Null Residual d.f. N D2 deviance deviance GLM 6 – Dead / Alive ~ -0.004792611 45.3042 38.4698 50 54 15.1 Distance to nearest female + 0.2422600 Home range size + 0.03632949 Proportion of fixes within the swamp – 0.3212067 GLM 7 – GLM 6 + Body condition 45.3042 36.6113 49 54 19.2 GAM 6 – A GAM of GLM 6 45.3042 20.1940 41 54 55.4 GAM 7 – GAM 6 + Body condition 45.3042 12.2192 37 54 73.0

The GAM on the preferred model explained far more deviance (55.4%) and when the body condition variable was included 73.0% of the deviance was explained (Table 5).

Despite the increase in explanatory power, none of the variables were better fitted by a spline curve than the linear fit of the GLM (

278 Chapter 8 Mortality and survivorship

Table 6).

279 Chapter 8 Mortality and survivorship

Table 6. Analysis of deviance table of the preferred generalised additive model (GAM

6). A probability of less than 0.05 indicates that the non-parametric smooth spline curve of the GAM fits the data significantly better than a linear fit.

Variable d.f. Non-parametric Non-parametric Prob. d.f. c2 Intercept 1 Body condition 1 2.8 2.3539 0.473 Distance to nearest female 1 3.0 6.1614 0.101 Home range 1 2.9 6.3128 0.091 Proportion of fixes inside swamp 1 2.8 5.3179 0.133

The impact of mortality on each population

Sites with larger populations have proportionally more deaths than those with smaller populations (

280 Chapter 8 Mortality and survivorship

Table 7).

281 Chapter 8 Mortality and survivorship

Table 7. A comparison between the proportion of the overall quokka population found at each site and the proportion of the total quokka deaths at each site.

Site Percentage (%) of total Percentage (%) of Ratio of deaths estimated quokka total deaths (8) to population population (84) 1 size Chandler 11.9 12.5 0.10 Hadfield 34.5 25.0 0.07 Kesners 42.9 62.5 0.14 Rosella Road 0.01 0 0 Victor Road 10.7 0 0 1 population based on Cormack-Jolly-Seber estimate (Chapter 4).

Discussion

Survivorship

The similar survivorship values for each sex is unusual for a sexually size-dimorphic species like the quokka (Chapter 4). Such species generally exhibit lower male survivorship due to the costs associated with competition for mates, dispersal and increased food demands (Owen-Smith 1993). The absence of dispersal in quokkas and similar home range sizes between the sexes (Chapter 5) may explain the similar survival of the sexes.

The similarity in the survival curves of all sites combined was due to the high confidence intervals caused by censored animals. When pairs of sites were compared only the Kesners site differs significantly from the Victor Road site which had no mortality events. Very little can be read into this due to the large confidence intervals and small sample sizes at many of the sites. Comparisons between adults and juveniles suffered

282 Chapter 8 Mortality and survivorship similar problems but these were compounded by the inefficiency of the breakaway collars applied to juveniles (see Chapter 5 for discussion).

Mortality events

The inclusion of the two trap deaths to the overall survivorship estimate reduces it from

0.61 to 0.55. These were both females and they may be more susceptible to trap mortality, via broken necks, than males due to the more massive musculature around the chest and neck of males (M.H. unpubl. data) which may limit neck injuries caused by crashing into the ends of the trap while attempting to escape. The two trap related mortality events, both in the large cage traps used, represented less than 1% of the 281 quokka captures. Both of these animals had rigor mortis when the traps were cleared in the mornings and the author is yet to find a satisfactory way of reducing such trap deaths. Conversely, the capture myopathy found in quokkas on Rottnest Island (Kakulus 1961) and in earlier unpublished studies on the mainland (A. Tomkinson pers. comm.) was not observed during this study.

Natural mortality events for the mainland quokka populations in the past derived from predation, disease, starvation and the direct and indirect effects of fire (Chapter 3 and 4).

While mortality events associated with starvation and fire were not recorded in this study various forms of predation, and possibly disease, were.

The three quokkas dying of unknown causes may have succumbed to disease, although predation may be more likely. Wild quokkas have been found to be susceptible to infection by toxoplasmosis parasites (Gibb et al. 1966) and Salmonella bacteria (Iveson and Hart

1983; Hart et al. 1986). Captive quokkas are known to be parasitised by the genera

283 Chapter 8 Mortality and survivorship

Progamotaenia (Cestoda tapeworms); Austrostrongylus, Breinlia, Cloacina, Dipetalonema,

Filaria, Microfilaria (Nematode flatworms); and Entamoeba and Toxoplasma protozoans

(Collins 1973). A Herpes virus was transferred from an infected handler to a captive quokka colony (Burnet 1970). Post mortem examination of the least decomposed of these three - a female that showed no evidence of predation - revealed fat reserves, a full stomach and bladder, and an absence of obvious internal parasitic cysts suggesting disease was not the cause. While this death occurred in summer the quokka may have succumbed to dehydration considering the species high water requirements (Main and Yadav 1971), however as water was plentiful at the Chandler site throughout the year and as the animal had a full bladder this seems unlikely. While the other two quokkas whose cause of death could not be categorised may have succumbed to disease, this seems unlikely considering the threat of predation and the rapid decomposition of their carcasses.

Nocturnal birds of prey, such as the masked owl (Tyto novaeholliandiae) or barking owl (Ninox connivens), are reported to have caused cave deposits containing fossilised quokka bones (Archer and Baynes 1972). While both species are capable of capturing prey the size of small quokkas (Archer and Baynes 1972), only the masked owl regularly roosts in caves (Schodde and Tidemann 1997) and so it is more likely to have been the cause of these deposits. While both species are present in the study area (Schodde and Tidemann

1997) , neither were observed during this study. The change in abundance of nocturnal birds of prey since the introduction of cats and foxes is unknown but it may be that their levels are such that they no longer pose a significant threat to quokkas.

Quokka bones found below a wedge-tail eagle (Aquila audax) eyrie on Bald Island

(Storr 1965b) suggest diurnal birds of prey are also natural threats on the mainland.

284 Chapter 8 Mortality and survivorship

Wedge-tail eagles, little eagles (Hieraaetus morphnoides), square-tailed kites (Lophoictinia isura), brown goshawks (Accipiter fasciatus), collared sparrowhawks (A. cirrhocephalus), peregrine falcons (Falco peregrinus) and brown falcons (F. berigora) were observed on or near the study sites but only the wedge-tail is considered a potential predator based on known diets (Barker and Vestjens 1989/90). Ospreys (Pandion haliaetus) have been recorded preying on Rottnest Island quokkas (Storr 1965a) but were not observed near the study sites. Despite these previous records there was no evidence of raptor predation on quokkas in this study.

Carpet pythons (Morelia spilota) are predators of many native species including southern brown bandicoots (Isoodon obesulus), common brushtail possums (Trichosurus vulpecula), bilbies (Macrotis lagotis); woylies (Bettongia penicillata) and tammar wallabies (Macropus eugenii) (D. Pearson pers. comm.). This prey size includes all but the largest male quokkas. Carpet pythons occur in the south-west and, consequently, may prey upon quokkas. This was not observed in this present study and, with the potential that the decline in python abundance in the wheatbelt and coastal areas of the south-west over the past 40 years (Pearson 1993) has extended to the jarrah forest in between, they are not considered a major predator of the quokka.

The Aboriginal, and later European, inhabitants of the south-west were recorded as hunting quokkas regularly (Stewart 1936; Whittell 1954; Gardner 1957; Gould 1973). The reduction in density and the removal from a traditional lifestyle of the Nyoongar people of south-western Australia has meant that such predation events are no longer a threat.

Despite recent records of humans killing quokkas (AAP 1998) no predation event on quokkas attributable to humans was recorded in this study.

285 Chapter 8 Mortality and survivorship

Today, these indigenous predators have largely been replaced by introduced placental predators - the dingo, European red fox and the feral cat. The dingo occurs at such low numbers in the jarrah forest that it is not considered a significant threat to quokkas. The fox and cat are considered significant threats to the survival of quokkas (Chapter 3) and the only predation events recorded in this study were attributed to these species.

Although the sexes were not significantly different in overall survivorship, predator derived mortality may not be acting evenly on the two. Over the 25 month period of this study, male quokkas were not observed to be preyed upon (although this cannot be ruled out entirely as the sample size was small and the two highly decomposed male carcasses may have succumbed to predation). Conversely three of the four female deaths were attributed to predation by introduced predators.

Male quokkas (mean body mass = 3.68 kg Chapter 4) are closer than females (mean body mass – 2.65 kg Chapter 4) to the upper limit of the 0.035 – 5.5 kg critical weight range of mammals susceptible to a range of extinction factors that were exacerbated by predation (Burbidge and McKenzie 1989). This larger size may enable males to reduce their risk of predation through increased strength and speed in avoiding predators.

Alternatively, there may be behavioural differences between the sexes driving this, such as males dominating areas of refuge from predation pressure. This is all assuming this difference is not an artifact of small sample sizes. Burbidge & McKenzie (1989) considered the larger sex of other sexually dimorphic species at the lower limit of the critical weight range to be at risk and similar phenomena may be occurring at the other end of the scale.

286 Chapter 8 Mortality and survivorship

Quokkas may be most susceptible to predation in the wetter months when swamps become inundated and home range centres shift toward their periphery (Chapters 3 and 5).

Two predation events occurred in this period and both were attributed to foxes (Table 2).

Obviously such small sample sizes are inconclusive, however it may be that the size and hunting method of the fox – a taller, slightly more coarsing predator than the cat, despite being considered as a felid-like predator (Henry 1986) - is negated by the dense vegetation of mainland swamps compared to the cat – a shorter, stockier, ambush predator – which may be equally as successful inside the swamps as outside. Yet the abundance of quokkas in more open habitats on Rottnest Island, in the presence of feral cats, suggests this is unlikely the sole cause and the body mass of the quokka may minimise the threat from cats.

In addition to recently arrived predators, other causes of mortality have appeared since

European arrival. Today, road kills are one of the most common ways of identifying new quokka sites throughout the mainland. From our study it appears that males may be most susceptible to vehicle caused mortality. In this study the two road kills occurred on the same stretch of sealed road which cut off a small section of the Kesners swamp where a female was known to occur. Both males were captured on either side of this road prior to their deaths. A third road-killed quokka that was found at a new site during this study was also a male.

Male-biased road kills are common in numerous species of macropodid (Coulson

1997). Again, with the small sample size of this study, conclusions must be regarded tentatively but with sex ratios at parity (Chapter 4) one would expect equal numbers of males and females to die from roa d kills (Coulson 1997). The male biased road kills in quokkas may be explained by males having larger home ranges (Chapter 5) or by females

287 Chapter 8 Mortality and survivorship positioning their home ranges according to physical boundaries (including roads) and male home ranges overlapping these.

Other factors implicated in quokka mortality

While predators, particularly the red fox, may inflict the killing blow to quokkas, there are various other ecological features of dead individuals that made them vulnerable to predation. The modelling conducted as part of this study highlighted these. Of the eight mortality events recorded during this study, all but two may have been a result of predation

(Table 2). Although the sample size was small, close proximity of an individual to the nearest conspecific female conveyed a decreased threat of mortality compared to those individuals that were separate from other females (Figure 5). Considering males form dominance heirarchies (Packer 1969) this is likely to mean that being distant from other females implies a solitary nature. Similarly, solitary behaviour in eastern grey kangaroos increased predation risk (Banks 2001) and the benefits of grouping or flocking are thought to include increased vigilance of predators (Pulliam 1973; Lima 1987; Dehn 1990).

Although predators are attracted to the odour of their prey (Dickman 1992b; Lagos et al.

1995; Banks et al. 2000b), there is no evidence to suggest that the grouping behaviour of quokkas increases the risk of predation.

Increasing home range size conveys a decreased risk of mortality (Figure 5). This may be due to larger animals being more dominant (Packer 1969) and thus range more widely while securing preferred foraging areas and, for males, obtaining more mating opportunities. Although this increased range could increase the risk of encountering a predator (Lagos et al. 1995; Norrdahl and Korpimaki 1998; Banks et al. 2000b), this

288 Chapter 8 Mortality and survivorship appears to be offset by appropriate habitat use. Such habitat use involves spending a high proportion of time within the refuge afforded by the dense, swampy vegetation which results in a decreased risk of mortality (Figure 5). If the majority of time is spent outside the swamp then the risk of mortality is increased. Considering the majority of the food plants of the quokka exist within the swamp margins (Storr 1964a), the benefit of remaining with the swamp may also be diet related.

While the variables used in the preferred model are easily explained, the increased mortality likelihood with increasing body condition is confusing (Figure 5). Despite the concerns as to whether such condition indices actually relate any ecological benefit to an individual (Krebs and Singleton 1993), it seems counter-intuitive that individuals in poorer condition should have a lower likelihood of dying that those in better condition. Whether this response is related to the low dominance rank (Packer 1969) associated with poorer conditioned animals which are forced to range widely to avoid agonistic encounters with conspecifics or due to the restriction of these animals to within the swamp is unknown.

The future…

A simple life-history model using the 0.8111 survivorship calculated here shows the possible changes in population size from the initiation of fox baiting in 1994 to the estimated population on 150 in 2001 (Chapter 4) and into the future (Figure 6). Female quokkas are able to produce 1.8 to two young per year (Sharman 1955a, 1955b; Shield

1964). This means each parent can theoretically be replaced annually, although considering the high rate of pouched young loss on the mainland, a more realistic birth rate is 1.4

(Chapter 4). Without births supplementing the population, quokka numbers are expected to

289 Chapter 8 Mortality and survivorship halve every three years without other contributing factors (Figure 6). With the theoretically possible birth rate of two young per year the population can treble in three years, however with the more realistic birth rate a more moderate increase is expected (Figure 6). A birth rate of 1.23 is required to maintain stable population levels in the absence of emigration and immigration. No movement between sites was observed during this study so the assumption of no movements between sites is justified. At these rates of population increase, the quokka population size when fox control began may have been as low as seven individuals although, based on the more conservative birth rate, is more likely to have been about 60 individuals.

1000

800

600

400 Predator control begins Estimated population 200

0 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 Years from 2001 Birth rate of 0 Birth rate of 1.4 Birth rate of 2 Birth rate of 1.23

Figure 6. Life-history model of the estimated population change from the initiation of predator control to 2001 (Year 0) and into the future. This simple model is based on

290 Chapter 8 Mortality and survivorship multiplying the estimated population of 150 (estimated in Chapter 4) in 2001 by the overall annual survivorship value and then by various birth rates. It assumes no movements between populations, which has been shown (Chapter 5), and constant survivorship between years, sexes and age classes.

That such dramatic increases in population size are possible (Figure 6) but have not been observed (Chapter 4) despite apparently adequate adult and juvenile survival indicates mortality may be occurring during the pouch life. Evidence from the high number of young per female per year and low juvenile recruitment show that mainland quokkas do indeed exhibit a high pouch young mortality rate (Chapter 4).

The behaviour of female quokkas within traps may elucidate the reasons for this.

When the trapped animals became aware of the researcher approaching they generally tried to escape during which larger pouch young were often released from the pouch. The pouch young would flounder on the ground and make a loud hissing noise. Whether the young were actually physically expelled by the mother or became dislodged during these violent, evasive movements is unknown but considering the muscular control female quokkas have over the pouch opening (highlighted when the muscle becomes relaxed during sedation) it seems likely that this is a behavioural response rather than accidental.

If this release is found to be a physical action by the mother then it can be viewed as a useful predator-avoidance strategy. When a female quokka becomes aware of a life threatening interaction with a predator she may expel her offspring whose subsequent noise attracts the attention of the approaching predator. When faced with the choice of a certain

291 Chapter 8 Mortality and survivorship meal without any more effort or a possible meal from the fleeing adult, it seems likely that the predator would chose the easy meal.

In fact, such behaviour has been observed in other marsupials. Swamp wallabies

(Wallabia bicolor) have been reported expelling their pouch young when chased by dingoes

(Robertshaw and Harden 1986). These authors found a high proportion of young swamp wallabies in the diet of the dingo which they concluded were pouch young consumed after being ejected by the mother (Robertshaw and Harden 1985; Robertshaw and Harden 1986).

Grey kangaroos also expel their pouch young when chased by red foxes (P. Banks pers. comm.) (Banks 1997). Expulsion of pouch young when in a trap and approached by a researcher has also been observed in the woylie or brush-tailed bettong (Bettongia penicillata ) and in southern brown bandicoots.

The suggestion has been made that this behaviour arose because these predators had worked out how to ‘farm’ their prey by chasing the females and causing them to expel the young (Newsome 2001). Yet there seems little selective pressure on the part of the predator for this behaviour to evolve compared to the prey. It would seem more advantageous for the predator to evolve strategies to increase hunting success on adult prey because they provide a larger meal rather than for predators to actively conserve their food production facilities (breeding females). From the prey’s perspective, the selection pressure to drive this behaviour is large as it allows the adult female to survive another day, breed again and thereby pass on her genetic information. Furthermore, jettisoning the pouch young seems likely to allow increased speed and agility thereby maximising survival chances. Jettisoning the young is also less costly for marsupials because of their ability to replace the lost young with a delayed blastocyst (Sharman 1955b).

292 Chapter 8 Mortality and survivorship

This behaviour may be a strategy that evolved to avoid predation by carnivores that utilise coursing hunting strategies such as the thylacine (Thylacinus cynocephalus)

(Rounsevell and Mooney 1995) and the more recently arrived dingo (Marsack and

Campbell 1990). Having coped with such predation strategies in the past the quokka may be resilient to similar strategies today however the red fox is considered a more felid-like canine (Henry 1986). Consequently, this predator-avoidance strategy may be less effective against them, particularly considering the foxes preferred prey size of medium-sized mammals (Catling 1988; Brunner et al. 1991; Risbey et al. 1999; Read and Bowen 2001).

The success of this anti-predator behaviour in the face of surplus killing by a newly arrived, exotic predator (Short et al. 2002) is likely to have been minimal considering the speed at which the predator could kill the young before re-focusing attention on the nearby and predator naïve mother. This may be exacerbated by the behaviour of the mother if she remains near the young as has been observed in eastern grey kangaroos (Macropus giganteus) (P.Banks pers. comm.). The quokka has been described as “remarkably fearless” (White 1952) and such predator-naivete may have intensified the predation pressure they were under. This bias in favour of the predator is likely to have resulted in the rapid demise of numerous Australian native animals including the quokka upon the arrival of the fox (White 1952; de Tores et al. In prep.).

This predator avoidance strategy and the associated loss of pouch young may explain the absence of an obvious population increase eight years after the initiation of fox control

(Chapter 4). It is acknowledged that there are other pouch young mortality factors such as inadequate nutrition during drought (which was not a factor during this study) and separation from the mother following short excursions outside the pouch (Russell 1974).

293 Chapter 8 Mortality and survivorship

This latter factor cannot be discounted but is difficult to envisage it as being of such significance to have masked the anticipated population increase (Chapter 4).

Conclusions

Based on simple birth versus survivorship comparisons (Figure 6) the endangerment of the quokka appears to have been alleviated. Adult survivorship following fox control is at levels that will allow population increases assuming pouch young are successfully weaned.

That this noticeable population increase has not been apparent to date highlights the likelihood that pouch young mortality is limiting this increase (Chapter 4). As deaths appear to be density dependent, conservation efforts should continue to focus on all of the extant populations in the northern jarrah forest, except possibly the Rosella Road site. On- going predator control appears crucial as, even with today’s reduced predation intensity, introduced carnivores are still a major cause of quokka mortality. Additionally the causes and level of juvenile and pouch young mortality is unknown but are likely to explain why the expected population increase has not been observed (Chapter 4).

From a conservation perspective, the recommendations formed around the finding that local extinction probability has a greater influence on metapopulation persistence time than the rate of colonisation is important (Etienne and Heesterbeek 2001). This is fortunate due to the minimal likelihood of increasing the colonisation rate between local quokka populations because of the large distances separating them (Chapters 4 & 5). In attempting to reinstate metapopulation function, extinction probability at known sites should continue to be minimised and colonisation should be directed from these to adjacent, threat-free patches in accordance with the recommendations of (Etienne and Heesterbeek 2001).

294 Chapter 9 Modelling presence/absence

Modelling the occurrence of the quokka (Setonix brachyurus

Quoy & Gaimard) (Macropodidae: Marsupialia) in Australia.

Abstract. The presence or absence of quokka populations at 66 sites in the northern jarrah forest was investigated using generalised linear and additive models.

Stepwise addition and removal of variables in the linear model saw three of the 13 uncorrelated variables used explain 34% of the deviance in the data. These variables were the number of poison meat baits delivered per hectare, the average age of the swamp and a habitat factor score pertaining to large areas of recently (5 – 9 years ago) burnt Agonis swamp habitat type. A generalised additive model was run on this final list of variables and this explained 52% of the deviance as two of the smoothed curves exhibited a significantly better fit to the data than the linear fit. Two other variables also showed substantial support; distance to disturbance and a habitat factor score that related to a site possessing swamp habitat that was burnt 15 to 19 years ago. The addition of these variables explained 38% of the deviance in the linear model and 74% in the additive model. It is concluded that if more were known about the metapopulation structure of the species then more of the deviation in the data may have been explained. The model provides indirect evidence for the causes of the 1930s decline in quokka numbers and for the species continued vulnerable status. This is largely due to the identification by the model of the importance of fox predation and a mosaic of early seral stage habitat (< 10 years since fire), probably for foraging, and late seral stage habitat (> 19 years since fire), for refuge.

289 Chapter 9 Modelling presence/absence

Introduction

Mathematical models to describe or predict the presence or absence of flora and fauna are becoming important tools for conservation managers (Buckland and Elston

1993; Ginsberg et al. 1995; Catling and Coops 1999; Carrete et al. 2000; Catling et al.

2000; Milsom et al. 2000; Kanowski et al. 2001). Their importance is intensified when the species being modeled is threatened with extinction (Short 1982). Such models provide information on the potential distribution of species using climatic data (Sutherst and Maywald 1985; Nix 1986) or various habitat features (Short 1982; Catling and

Coops 1999; Catling et al. 2001). Habitat suitability models have also become an important tool (Nelson and Buech 1996; Reading et al. 1996; Jorgensen et al. 1998;

Dorgeloh 2001; Loukmas and Halbrook 2001).

The quokka (Setonix brachyurus) (Quoy & Gaimard 1830) is a medium-sized, macropodid marsupial that is endemic to the south-western corner of Australia

(Kitchener 1995; de Tores et al. In prep.; Hayward et al. In review). This species suffered an extensive decline in distribution and abundance in the early 1930s and is today listed as vulnerable (de Tores et al. In prep.). Increasing amounts of evidence have been gleaned from autecological studies as to the causes of the precipitous decline of the quokka and its existing status (Chapters 4, 5, 7 and 8). Yet although these studies have intimated the causes, no definitive answer has become evident. This study provides an opportunity to support or reject the previous hypotheses as to the threatening processes affecting the quokka; namely that fox (Vulpes vulpes) predation and habitat requirements are crucial factors affecting the presence and persistence of quokka populations (de Tores et al. In prep.; Hayward et al. In review). Consequently,

290 Chapter 9 Modelling presence/absence

this study aimed to determine what ecological features explained the presence or absence of quokka populations in the northern jarrah forest.

Methods

Determining quokka presence

Quokka presence at a site was determined by investigating the presence of the species’ faecal pellets within the characteristic runways they create (Liddelow Undated) through their preferred, densely vegetated, Agonis swamp habitat (Chapter 7). Fresh scats located along the 10 cm by 15 cm quokka runways indicated quokka presence.

Although the accuracy of this method has been questioned (Alacs et al. 2002), I still consider it the most valid, efficient and accurate method of determining quokka presence or absence in broad-scale surveys.

Surveys were conducted at 66 sites possessing Agonis swamp shrublands within the northern jarrah forest (Figure 1). At each site, ten transects were walked from the edge of the swamp vegetation into the watercourse at the centre of the swamp over a one kilometre section. The presence of fresh faecal pellets within the characteristic runways were used as signs of presence. Due to the relative uniformity of the swamps, transects were generally of a similar length (~ 30 - 50 m).

291 Chapter 9 Modelling presence/absence

Figure 1. Location of sites investigated for quokka presence in this study. Quokka populations were found at sites denoted with filled stars while sites depicted with open circles showed no evidence of quokka presence. The shaded area represents the distribution of the quokka in the northern jarrah forest in 2000 (Chapter 3). The dashed line shows the 1,000 mm annual rainfall isohyet and the dotted line shows the 700 mm isohyet.

Sites investigated were identified before field surveys through either knowledge of an extant quokka population, previously possessing a quokka population, having

292 Chapter 9 Modelling presence/absence

quokka road-kills nearby or by occurring in the upper reaches of creek systems.

Throughout the range of extant quokka distribution in the northern jarrah forest (Figure

1) therefore, a full range of each variable was sampled, including the specific habitat variables. While there is no estimate of the number of sites falsely recorded as not having quokkas present, the intensity of the survey within the one kilometre stretch of swamp investigated suggests that the likelihood of false absences are low, particularly when pellet counts at sites with known quokka populations are considered.

Ecological variables investigated

At each site various ecological variables were measured (Table 1). Not all of these variables were included in the final analyses due to high partial correlations between them. The slope variable was correlated with aspect (r = -0.318; P = 0.027) and ecotone

(r = -0.289; P = 0.046). The mosaic variable was correlated with the overall number of habitat types (r = 0.744; P < 0.001) and disturbed habitat types (r = 0.347; P = 0.015).

Data were either log- or arcsine transformed to improve normality and heterogeneity of variances. The variables themselves are shown in Table 1. These explanatory variables were compared between sites with quokkas present and absent using t-tests with Bonferroni corrections. They were subsequently used to investigate the response of the presence or absence of quokka populations at sites in the northern jarrah forest of Western Australia.

293

Table 1. The name and description of ecological variables used to model quokka presence or absence as well as the range of values found for each and their unit of measurement. The description of each variable also presents the reasons why this variable was included in the analysis. Those variables that were correlated with others and excluded from analyses are marked with an asterisk.

Variable Description Range and units Aspect * A categorical variable determined by the map bearing in degrees along the down slope direction of the 1 (north-east), watercourse. This variable was included due to the potential for southerly sites to have different 2 (south-east), ecological attributes than northerly facing ones. 3 (south-west), 4 (north-west) Baiting The total number of poisoned fox baits laid per hectare surrounding the swamp since feral animal control 0 – 37 ha-1 began. Higher levels of baiting in the northern jarrah forest have been shown to result in significant increases in the density of predation-sensitive native fauna and generally reduced densities of foxes (de Tores 1999). The range of baiting levels arose due to variations in the annual frequency of bait delivery within the Western Shield and Operation Foxglove programmes which range from an unbaited control region through biannual, quarterly and two monthly up to specific sites (e.g. around quokka populations) that are baited monthly (de Tores 1999; Thomson and Algar 2000; Hayward et al. In review). Distance from The distance from the interface between the state forest and cleared, residential or agricultural lands and 0 – 13 km forest edge the centre of each particular swamp. This variable was used to investigate the finding that foxes reinvaded areas near the forest boundary rapidly after baits were laid (de Tores 1999). Distance to nearest The distance from the centre of the swamp to the nearest source of disturbance (mining, agriculture, urban 0 – 13 km disturbance areas, etc.). This variable was used to investigate the effect of disturbance on quokkas. Disturbed habitats The number of habitat units at each site that are classified as disturbed (cleared or revegetation; Chapter 0 – 4 * 6a). This variable was also intended to investigate the effect of disturbance on quokkas. Ecotone * The percentage of the site that is made of up ecotonal habitat types (bullich swamp forest, bullich- 0 – 27% blackbutt open forest, blackbutt open forest) (Chapter 6a). This variable was included as the preference of quokkas for a mosaic of habitat types (Christensen and Kimber 1975) suggested the species may be an ecotonal specialist. Habitat number * The number of different habitat types found at a site. Again this variable was intended to investigate the 1 – 14

294

Variable Description Range and units mosaic preference of quokkas (Christensen and Kimber 1975). Habitat type The percentage of each habitat type (according to the methods in Chapter 6a) that made up the one km -1 - 1 section of swamp and an area 100 metres wide either side of the swamp. Air photo interpretation was used to delineate habitat unit boundaries which were subsequently ground-truthed during the quokka surveys. Habitat units were described according to floristic, structural and other biological features (Chapter 6a). The percentage that each habitat unit made up the entire study site was subsequently combined using principal components factor analysis. Swamp habitat factors were denoted as NJFSwF1-3, while forest habitat factors were denoted as NJF1-3. This variable was used following previous findings that quokkas prefer a mosaic of freshly burnt and long unburnt areas of swamp (Chapter 7) (Christensen and Kimber 1975). Rainfall Mean annual rainfall was derived from distances from rainfall isohyets (Chapter 2). Rainfall was thought 700 – 1,400 to be an important descriptive variable after it was found that all extant quokka populations are found in mm areas receiving more than 1,000 mm of precipitation per year (Chapter 3) (Figure 1). Slope The angle from horizontal to the ground that the one km length of swamp fell by, measured from 0 – 8o topographical maps. Although all Agonis swamps exist in the relatively flat, broad upper reaches of creek systems along the eastern edge of the northern jarrah forest (Chapter 3), they appear to change into different habitat types where the terrain becomes too steep or too flat (pers. obs.). Consequently, the slope of each site was used as a descriptive variable. Swamp age The average number of years since fire burnt the swamp (Chapter 7). The area of each swamp habitat unit 0 (just burnt) – at a site were multiplied by the average age of that unit and then divided by the total area of swamp to 33 years obtain an average swamp age. This meant that sites with a mosaic of burn ages could be compared with sites with one age class. With age since fire previously identified as a factor affecting quokka presence (Chapter 7) (Christensen and Kimber 1975), the average swamp age was used to determine whether there was a preference for any particular age class (Chapter 7). Swamp mosaic The number of swamp habitat units found at each site was considered as a variable in conjunction with 1 – 5 value swamp age following the habitat preferences of quokkas. Swamp length The number of kilometres that the creek line retained the characteristic Agonis vegetation. This was 1 – 5 km determined from raster images of aerial photographs that were imported into the MapInfo computer

295

Variable Description Range and units program and geo-referenced. Tributaries were added to the total swamp length.

296 Chapter 9 Modelling presence/absence

Modelling

Three methods were used to model quokka presence or absence at particular sites.

Forward and backward stepwise selection of variables in a generalised linear model (GLM)

(Nelder and Wedderburn 1972) was initially used to identify the ecological variables explaining the most deviance in the presence/absence data. The principle behind the GLM is that the systematic component of the linear model can be transformed to create an analytical framework resembling the standard linear model but which accommodates a wide variety of non-normal and non-interval measured outcome variables (Gill 2000). The assumptions underlying linear model theory require that the error component be distributed independently with a mean of zero and a constant variance (Eisenhart 1947; Zar 1996).

Although quite robust, these assumptions often cannot be met if the response variable comes from a non-normal distribution (Cochran 1947; Eisenhart 1947). GLMs generalize linear models by assuming a transformation of the expected (or average) response is a linear function of the predictors, and the variance of the response is a function of the mean response. GLMs use a ‘link’ function to define the relationship between the systematic component of the data and the response variable such that asymptotic normality and constant variance are no longer required (Gill 2000). This allows models to be loosened from the restrictions of standard linear theory (Gill 2000). The generalised approach unifies the diverse probabalistic forms into a consolidated exponential form which facilitates a more rigorous and thorough theoretical treatment of the principles underlying linear models (Gill 2000). In the case of this research a binomial distribution

(presence/absence) has been used with a logit link function.

297 Chapter 9 Modelling presence/absence

There are several methods of assessing how well a particular GLM fits the data, however no single measure provides the ‘right answer’ and whenever possible more than one approach is recommended (Gill 2000). Akaike’s information criterion (AIC) (Akaike

1973, 1974) is a commonly used measure to select between candidate models by minimising the negative likelihood of including a variable while taking into account the number of parameters (Gill 2000). Models with a lower AIC are deemed best fitting the data (Burnham and Anderson 2001). Models with a delta (D) AIC, calculated by subtracting the AIC of the preferred model from each other model, of less than two have substantial support while those with a DAIC greater than seven have no support (Burnham and Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and are essentially the probability that the model is the Kullback-Leibler best model for the data (Burnham and Anderson 2001).

Deviance residuals are also used to measure goodness of fit of a model (Yee and

Mitchell 1991). The residual deviance serves the same role as the residual sum of squares in the linear model. To test whether a model with fewer variables is applicable given a larger model, the increase in deviance can be examined and compared to a chi-square distribution with degrees of freedom equal to the difference in the number of variables in the two models (Yee and Mitchell 1991). Smaller deviances indicate better fit (Yee and

Mitchell 1991).

The Cp statistic (the likelihood version of the AIC statistic) provides a criterion for determining whether a GLM is improved by dropping or adding terms (Hastie and

Tibshirani 1990). If any term has a Cp statistic lower than that of the current model the

298 Chapter 9 Modelling presence/absence

term with the lowest Cp statistic is dropped. If the current model has the lowest Cp statistic, then the model is not improved by dropping any term.

The amount of deviance explained by the overall model is described by D2 (Hastie and

Tibshirani 1990) which is calculated by the formula:

2 (Null deviance - deviance) x 100 D = . Null deviance

The percentage deviance is analogous to R2 in regressions (Yee and Mitchell 1991).

Additive models extend the notion of a linear model by allowing some or all linear functions of the predictor variables to be replaced by arbitrary smooth functions of the predictors. Consequently, the final equation identified by GLM was then compared to a generalised additive model (GAM) (Hastie and Tibshirani 1990). GAMs are a non- parametric extension of GLMs but with more flexibility which allow a wider range of response curves to be modeled (Yee and Mitchell 1991). The ‘smooth spline’, a cubic smoothing spline with local behavior similar to that of kernel-type smoothers, was used in this analysis. GAMs allow the available data to determine the shape of the response curve rather than being limited to the available parametric shapes by using scatterplot smoothers

(Yee and Mitchell 1991).

GAMs have three major advantages over other forms of curve fitting. Firstly, it is unnecessary to group continuous variables (Yee and Mitchell 1991). Secondly, they allow the effect of several explanatory variables on the response variable to be examined simultaneously, and finally they allow hypothesis testing upon the variables (Yee and

Mitchell 1991). A problem with GAMs is that they cannot provide equations with which to predict future presence or absence whereas GLMs can (Yee and Mitchell 1991).

299 Chapter 9 Modelling presence/absence

Analysis of deviance tests whether the smooth functions used in GAMs are significantly better than parametric terms (Hastie and Tibshirani 1990). If a parametric curve is ‘statistically allowable’ then this is preferable to a non-parametric curve for reasons of parsimony (Yee and Mitchell 1991).

Results

Principal component factor analyses

When all of the dominant habitats in and around the potential quokka swamps are included in a principal components factor analysis, factor one (NJF1) is characterised by bullich-blackbutt open forest and bullich swamp forest and explains 15.4% of the variation

(

300 Chapter 9 Modelling presence/absence

Table 2). Factor two (NJF2) is characterised by having a high proportion of jarrah - marri open forest and 15 - 19 year old Agonis swamp shrubland habitat types (

301 Chapter 9 Modelling presence/absence

Table 2). This factor explained 13.7% of the variation. Factor three (NJF3) explained

12.2% of the variation and was characterised by a high proportion of Agonis swamp shrubland burnt between ten and 14 years ago and between 20 and 24 years ago. Factor four (NJF4) explained 11.4% of the variation and was characterised by areas of swamp burnt between five and nine years ago (

302 Chapter 9 Modelling presence/absence

Table 2). Factor five (NJF5) was characterised by large areas of long unburnt swamp and blackbutt open forest and explained 11.1% of the variation.

303 Chapter 9 Modelling presence/absence

Table 2. Orthogonal factor scores for all the major habitat types. The variation explained by each variable is shown also. The variables that are important in describing each factor score are bolded. For later analyses these factors are denoted by NJF1-5.

Variable Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Freshly burnt swamp -0.256 0.272 -0.157 -0.336 -0.484 Five to nine year old swamp -0.122 -0.010 -0.029 0.849 -0.067 Ten to 14 year old swamp 0.028 0.008 0.831 -0.292 -0.097 15 to 19 year old swamp -0.044 0.715 -0.218 -0.188 -0.092 20 to 24 year old swamp -0.155 -0.025 0.662 0.357 0.015 Long unburnt swamp 0.127 0.137 -0.144 0.016 0.656 Blackbutt open forest -0.331 -0.046 0.051 -0.332 0.646 Bullich – blackbutt open forest 0.793 -0.078 -0.175 -0.013 0.054 Bullich swamp forest 0.784 -0.037 0.096 -0.098 0.009 Jarrah – marri open forest 0.112 0.758 0.266 0.195 0.176

Variation explained 15.4% 13.7% 12.2% 11.4% 11.1%

Comparison between sites with quokka populations present or absent

There was no significant differences between sites with quokkas present or absent for each of the variables used in the modelling when the Bonferroni correction was applied (

304 Chapter 9 Modelling presence/absence

Table 3). Only the number of fox baits per hectare (P = 0.005) variable approaches the corrected significance level of 0.003.

305 Chapter 9 Modelling presence/absence

Table 3. Comparison of the mean (± s.e.) raw values for each variable for sites (66) with quokka populations present or absent. Variables marked with an asterisk were excluded from the models after they were too highly correlated with others. Only the factor scores for the NJF2 and NJF4 factors are presented in the habitat type variable. The presence and absence of quokka populations was compared using t-tests with the Bonferroni correction factor making the significance probability 0.003 (degrees of freedom is 64 for each).

Variable Quokkas Quokkas t value Prob. present absent Baiting (number of baits ha –1) 13 ± 3 5 ± 1 -2.939 0.005 Distance from forest edge (km) 2 ± 1 1 ± 1 1.676 0.099 Distance to nearest disturbance (km) 2.4 ± 0.4 2.6 ± 0.4 1.320 0.191 Disturbed habitats * 1.6 ± 0.2 1.4 ± 0.2 -1.044 0.300 Ecotone (%) * 7 ± 2 5 ± 1 -1.268 0.210 Habitat number * 4.5 ± 0.6 4.1 ± 0.3 -0.575 0.567 Habitat type – (NJF2) 0.026 ± 0.011 ± -0.036 0.892 0.166 0.162 - (NJF4) -0.247 ± 0.100 ± 1.283 0.204 0.058 0.170 Rainfall (mm) 1,240 + 1 1,220 ± 1 -0.658 0.513 Slope (angle from horizontal) 4.1 ± 0.3 3.2 ± 0.3 -1.909 0.061 Swamp age (years since fire) 12.7 ± 1.6 7.6 ± 1.2 -0.145 0.886 Swamp mosaic (# swamp habitat types) 1.74 ± 0.24 1.70 ± 0.18 -0.109 0.913

Generalised linear and additive modelling

The stepwise addition and removal of variables in the GLM resulted in a preferred model (Model 11) with three variables, ten less than the null model possessed (Table 4). The null model had an Akaike information criterion (AIC) value of over 70 and this represented a DAIC almost 13 units higher than that of the preferred model. The preferred model

(Model 11) had the equation:

306 Chapter 9 Modelling presence/absence

Quokka presence / absence = 0.1300651 Baits ha-1 + 0.09683564 Swamp age + 1.140151 NJF4 –

2.87852.

As this equation indicates, the variables that are important in explaining quokka presence or absence are the number of baits per hectare, the average age of the swamp and the NJF4 habitat unit factor score. The NJF4 factor score is characterised by possessing large areas of swamp burnt between five and nine years ago.

Models 9 and 10 also possess DAIC values of less than two (Table 4) which signifies they too have substantial support (Burnham and Anderson 2001). Together, these two models exhibit a 66.7% probability of being the best model according to the Akaike weights (Table 4). The two variables that are included in these models are the distance from the swamp to a disturbance factor and NJF2. The NJF2 factor score is characterised by having areas that have been burnt between 15 and 19 years ago (

307 Chapter 9 Modelling presence/absence

Table 2). Plots of this factor indicate there is a negative relationship between quokka presence and the NJF2 variable which signifies an avoidance of this habitat type (Figure 2).

All other variables exhibit a positive relationship ( Figure 2).

The number of baits per hectare shows a linear relationship with increased likelihood of quokka populations found at increased levels of baiting (Figure 2). The NJF4 variable also exhibits a linear relationship with increased amounts of NJF4, or five to nine year old

Agonis swamp shrubland, resulting in an increased probability of quokka presence (Figure

2). The average swamp age variable is best fitted with a spline curve (

308 Chapter 9 Modelling presence/absence

Table 6) and shows sites with medium-term average ages have the greatest likelihood of quokka presence (Figure 2). There is also an increase at longer unburnt sites, but this relationship is driven by one site.

Table 4. Model selection statistics for the presence/absence of quokka populations data set. The generalised linear models (GLM) are numbered sequentially from the null model with the Akaike Information Criterion (AIC), the delta (D) AIC, the number of variables in the model (K) and the Akaike weights. These statistics are also presented for the generalised additive model (GAM) of the best GLM. Models with a lower AIC are deemed best fitting the data (Burnham and Anderson 2001) (i.e. Model 11). Models with a D AIC of less than two have substantial support while those with a D AIC greater than seven have no support (Burnham and Anderson 2001). Akaike weights provide a weight of evidence in favour of a model and is essentially the probability that the model is the Kullback- Leibler best model for the data (Burnham and Anderson 2001). Although using a forward and backward stepwise selection procedure, the model added no variables once they were removed.

Model AIC DAIC Akaike weights 1. Null model Pres/Abs ~ Slope + Distance to 72.6195 12.8377 0.0001 forest edge + Length of Swamp + Baits + Age + Mosaic + Distance to disturbance + Rainfall + NJF1 + NJF2 + NJF3 + NJF4 + NJF5 2. Null model with NJF1 deleted 70.9039 11.1221 0.0009 3. Model 2 with Mosaic deleted 69.4129 9.6311 0.0022 4. Model 3 with Slope deleted 67.9952 8.2134 0.0047 5. Model 4 with Rainfall deleted 66.7131 6.9313 0.0095 6. Model 5 with Distance to forest edge deleted 65.3131 5.5313 0.0181 7. Model 6 with NJF3 deleted 63.9917 4.2099 0.0364 8. Model 7 with Swamp length deleted 62.0052 2.2234 0.0705 9. Model 8 with NJF5 deleted 61.5839 1.8021 0.1903 10. Model 9 with Distance to disturbance 60.3629 0.5811 0.2349 deleted

309 Chapter 9 Modelling presence/absence

Model AIC DAIC Akaike weights 11. Model 10 with NJF2 deleted 59.7818 0 0.4325 GAM 108.1921 48.4103 0

The distance from a site to the nearest disturbance exhibits a linear response with increased likelihood of quokka presence at less disturbed sites (Figure 2). The NJF2 variable was best fitted with a spline curve and shows a decreased likelihood of quokka occurrence with higher levels of NJF2, which essentially means higher levels of Agonis swamp shrubland burnt between 15 and 19 years ago and jarrah – marri open forest (Figure 2).

b

a

c d

310 Chapter 9 Modelling presence/absence

e

Figure 2. Plots of the spline curve fitted to each of the variables in the final generalised linear model (a – baits ha –1, b – swamp age and c – NJF4) as well as the variables included in the other three models that showed substantial support (d – distance to disturbance; and e – NJF2). The spline curves generated in the GAM are used because of the difficulty in comparing the results of GLM and GAM and because linear fits can still be determined visually with the GAM plots. In each case, the variable is plotted while all other variables are held constant. All but NJF2 (e) exhibit a positive relationship.

311 Chapter 9 Modelling presence/absence

The GAM of the preferred model has a much greater AIC and therefore DAIC than the preferred GLM (Model 11) largely due to the differences in calculating the degrees of freedom and hence residual deviance of each (Table 5). The preferred GLM in explains 34% of the deviance exhibited in the data while the addition of the next three most important variables only show minor increases in the D2 value (Table 5).

Table 5. Variation explained by the preferred GLM and the GAM of that model. Lower residual deviances indicate a better model. The D2 statistic indicates the amount of variation in the data explained by the model and is analogous to the R2 of regression.

Model Null Residual d.f. N D2 deviance deviance GLM 11 – Quokka presence / absence ~ 78.54666 51.78178 61 66 34.07 0.1300651 Baits ha-1 + 0.09683564 Swamp age + 1.140151 NJF4 – 2.87852 GLM 10 – GLM 11 + NJF2 78.54666 50.73874 59 66 35.40 GLM 9 – GLM 10 + distance to 78.54666 48.70155 58 66 38.00 disturbance GAM of GLM 11 78.54666 37.96865 47.7 66 51.66 GAM of GLM 10 78.54666 20.76929 48.4 66 73.56 GAM of GLM 9 78.54666 18.44360 44.5 66 76.51

Two of the smoothing spline curves fitted to the variables differ significantly from a linear curve (swamp age and NJF2) while the remaining three variables are best represented by the linear fit (

312 Chapter 9 Modelling presence/absence

Table 6). Cumulatively, the mix of linear and spline curves results in a better fit to the data such that the GAM of GLM 9 explains 77% of the deviance while that of the preferred

GLM 11 explains 52% of the deviance ( Table 5).

313 Chapter 9 Modelling presence/absence

Table 6. Analysis of deviance table of the generalised additive model (GAM). A probability of less than 0.05 indicates that the non-parametric smooth spline curve of the GAM fits the data significantly better than a linear fit.

Variable d.f. Non-parametric d.f. Non-parametric c2 Probability Intercept 1 Baits 1 2.9 1.46148 0.676 Swamp age 1 2.9 11.12808 0.010 NJF4 1 2.9 3.58138 0.297 Distance to disturbance 1 2.8 2.57415 0.429 NJF2 1 2.9 9.15231 0.025

Discussion

The models derived here support previous work or hypotheses regarding the ecology of the quokka in the northern jarrah forest. The number of introduced predator control baits per hectare is considered an indirect method of determining the predation pressure on the quokka following findings that the abundance and survivorship of predation-sensitive species respond significantly to increased levels of baiting and that fox densities are generally lower at areas with higher levels of baiting (de Tores 1999). Consequently, the model’s determination that predator control is an important attribute in explaining the presence of quokka populations supports previous but unsubstantiated supposition on the effect of the red fox (Vulpes vulpes) on the quokka (Hayward et al. In review). With only one unbaited control site (Chapter 2) the aim of this overall study, to determine the impact of the fox on quokkas (Chapter 1), has been confounded due to the lack of replication. This modelling exercise provides further support to indicate that the fox was a major cause of the initial 1930s decline of the quokka (Chapter 3) and its continued low abundance

314 Chapter 9 Modelling presence/absence

(Hayward et al. In review) by preying upon animals directly at a swamp (Chapter 8) or during inter-patch movements (Chapter 5).

Quokkas are more likely to exist at swamps with intermediate ages since fire and sites with long average ages since fire. There are difficulties in interpreting this on its own because of the calculation of the average swamp age takes into account the proportion of the swamp covered by habitat units of different ages. Very few sites possessed only one age class and so the average age value generally reflects a mosaic of habitat units. When we look at the actual habitat factor scores that were also important in explaining quokka presence and absence, this interpretation becomes easier.

Habitat features have previously been identified as important in sustaining a quokka population with a preference for a mosaic of freshly burnt and unburnt areas (Christensen and Kimber 1975; Hayward et al. In review). The NJF4 variable relates to the presence of

Agonis swamp shrubland that has been burnt between five and nine years ago. When this variable is viewed with the other important factor score, NJF2, we see further weight added to the earlier hypotheses. When model 8 is considered, with a D AIC of only 2.2 (Table 4), we see that long unburnt swamp habitat (NJF5) also has a reasonable amount of support as an explanatory variable. All these features highlight the importance of a mosaic of age classes within the swamp with early (< 10 years since fire) and late (> 19 years) seral stages important for quokka presence and intermediate seral stages avoided.

The reasons for this preference have been suggested as a mix of dietary requirements provided by the fresh regrowth that occurs in recently burnt swamps as long as areas within those swamps remain unburnt to provide shelter and refuge from predation (Chapter 7).

The evolution of this attribute may have arisen to cope with the frequent, patchy, low-

315 Chapter 9 Modelling presence/absence intensity fire regime employed by the local Aboriginal people of the south-west (Wallace

1966; Burrows et al. 1995; Ward and Sneeuwjagt 1999). The alteration of this burning regime through the fire exclusion policy of the European colonists (Wilson and Friend

1999) undoubtedly placed significant pressure on the quokka metapopulation (Chapter 3).

The quokka population plummeted when other factors, particularly the arrival of the fox, combined with altered fire regimes in the 1930s (Chapter 3).

It is also interesting to examine the variables that were not considered to be important for the presence of quokka populations. The poor explanatory power of the swamp length variable may be due to the historically ephemeral nature of a quokka population’s persistence at a site before moving off to colonise a new patch when its habitat requirements were no longer met. Small areas of swamp could possess the heterogeneous mosaic of burnt swamp habitats required by the quokka as easily as large and, because of this habitat specificity, it is likely that quokka populations were, and probably still are, focused in small parts of swamp systems. Despite this, conservation biology and metapopulation theory suggest that larger, mainland populations would be buffered from stochastic causes of extinction more than smaller, island populations (MacArthur and

Wilson 1963, 1967; Diamond 1975; Caughley and Gunn 1996). Yet island biogeographic theory (MacArthur and Wilson 1967) seems unlikely to apply to quokkas (Hayward et al.

In review) whereas source – sink metapopulation dynamics is not governed by the size of the habitat patch but by its quality (Harrison 1991; Caughley and Gunn 1996).

Consequently, small areas of the swamp may be able to provide patches of adequate quality to sustain populations as easily as larger areas.

316 Chapter 9 Modelling presence/absence

Considering the relatively large areas and numbers of anthropogenic disturbances to be found in the northern jarrah forest, it is concerning to note that close proximity to disturbance resulted in quokkas being absent from a site. Quokkas are frequently observed around the settlement areas of Rottnest Island (pers. obs.) and the impact of disturbance on the mainland reiterates the belief that the two populations are extremely different (Hayward et al. In review). Nonetheless, mainland populations have survived in close proximity to towns until recently (Chapters 3 and 4). While quokkas can coexist with humans, direct habitat loss through urbanisation or mining does cause localised (e.g. Wild Pig Swamp discussed in Chapter 4), and possibly regional (e.g. between Perth and Bunbury), extinction

(Figure 1). Furthermore, the cumulative effects of disturbances and other factors can herald an extinction vortex (Diamond 1989).

The link between quokkas in the northern jarrah forest and high rainfall has been previously identified (Chapters 3, 4 and 5) however this was not found to be important in explaining the presence of or absence of quokka populations. This may be because the preferred habitat of the quokka in the northern jarrah forest only exists within the high rainfall belt and, within that area, other factors become more important.

The models may have had more explanatory power had more been known of the metapopulation dynamics of the quokka. Susan Harrison studied a butterfly metapopulation and the pattern of patch occupancy she found could not be explained by indicators of habitat quality but rather by the proximity to the source population (reviewed in Harrison 1991). Proximity to the nearest source population was not used in this analysis because of the virtual impossibility of knowing which were source populations, if indeed

317 Chapter 9 Modelling presence/absence the source populations had been located or if the quokka exists as a source-sink type of metapopulation (Hayward et al. In review).

Conclusions

The significance of the relationship between fox control and quokka presence/absence should not be understated. Ever since Kinnear and colleagues (Kinnear et al. 1988) reported such a link between foxes and the decline of Australian native fauna, the validity of such conclusions has been questioned, largely through the inadequacy of experimental controls (Hone 1999). Similar problems were encountered when trying to investigate the impact of predation pressure on the ecology of the quokka (Chapters 4, 5, 7 and 8). Yet this modelling exercise has shown that the long-held supposition that fox predation affects quokka populations (Hayward et al. In review) is justified, in conjunction with a mosaic of seral stages within the Agonis swamp. The reasons for the continued low abundance of quokkas in the northern jarrah forest (Hayward et al. In review) appears likely to be a combination of continuing predation pressure even with existing fox control measures, the habitat quality of swamps based on a mosaic of seral stages, and probably also inadequate time to exhibit noticeable population increases.

These findings may not be limited to a simple presence/absence response. Using population estimates from faecal pellet counts (unpubl. data), actual quokka abundance may be modelled to investigate the variables that affect this response variable. Such modelling exercises may be used in conjunction with bioclimatic models (Busby 1991) to develop a deeper insight into the factors affecting quokka presence. These findings may also be extended to attempt remote sensing of habitats suitable for quokkas using either

318 Chapter 9 Modelling presence/absence videography (Coops and Catling 1997a, 1997b; Coops et al. 1998; Catling and Coops 1999;

Coops and Catling 2000) or spectral analysis (Behn et al. 2001). Such methods may allow far more sites supporting quokka populations to be discovered.

319 Chapter 10 Conclusion

Conclusion

Key findings from each chapter of the thesis are discussed below.

Distribution and status

The quokka originally occupied an area of approximately 49,000 km2 in the south- western corner of Australia. Quokkas were widespread and abundant when Europeans colonised the region in 1829 but a noticeable and dramatic decline occurred a century later.

The arrival of the red fox to the region coincided almost exactly with this decline and so it was probably ultimately responsible. Continued predation by both the fox and the feral cat continued the decline, along with habitat destruction and modification through altered fire regimes. Since the 1950s, the area occupied by the quokka has declined by 45% and since

1990 by 29%. Based on the criteria of the IUCN (Hilton-Taylor 2000), the conservation status of the quokka should remain as vulnerable. Yet if each Agonis swamp which the quokka is increasingly restricted to is considered alone as the area of occupation, and if each one is assumed to be the same size, then the 56% extinction of local quokka populations at 79 known sites since 1990 suggests an endangered status may be more applicable.

Population sizes and dynamics

Trapping of eight sites supporting quokka populations in the mid-1990s revealed three sites have become locally extinct despite the ongoing, six year old, fox control programme.

Another three are at serious risk of extinction. Extant population sizes ranged from one Chapter 10 Conclusion

(i.e. effectively extinct) to 36 and population density ranged from 0.07 to 4.3 individuals per hectare. This is considered to be below the carrying capacity of each site.

The overall quokka population size in the northern jarrah forest may be as low as 150 adult individuals, of which half are likely to be female. Even the largest extant populations are highly susceptible to stochastic extinction events. This small size was surprising considering the six year old, introduced predator control programme.

Historically, the restriction to discrete habitat patches, the occasional inter-patch movement, the lack of correlation between the dynamics of each population and reports of frequent localised extinctions and colonisations suggest that the quokka population once existed as part of a classic metapopulation (Hanski and Gilpin 1991). Very few studies have described the metapopulation structure of medium-sized or large mammals (Hastings and Harrison 1994; Elmhagen and Angerbjorn 2001). Due to their space requirements

(McNab 1963) larger organisms seem less likely to exist in a metapopulation. The increased movements required to satisfy their metabolic needs mean that they are less likely to be restricted to discrete habitat patches and, if they are, then they are capable of readily moving between patches. On this basis it is hardly surprising that, to date, the quokka is one of the larger animal species occurring naturally (that is not through human- derived fragmentation) as a metapopulation. The massive decline of the quokka in the

1930s pushed the metapopulation structure into a non-equilibrium state such that today, the extant populations are the remnants of the original classic metapopulation.

The age class composition of the overall population is very similar to that on Rottnest

Island which suggests that mortality rates for all ages are similar between the two regions. Chapter 10 Conclusion

This may not be surprising considering the seasonal starvation on Rottnest Island (Main

1959) compared to the threat of predation on the mainland.

Reproduction

Wild mainland quokkas breed throughout the year, as occurs in captive mainland quokkas, but this contrasts with the insular population (Shield 1964). A significant reduction in the number of births occurs over summer and this coincides with a decline in female body condition. Despite this, the mainland quokka is relatively fecund and is able to wean two offspring per year. The level of recruitment from pouch young to independence was low and this may explain the apparent lack of population increase following the initiation of fox control.

Home range and movements

Mean home range size for quokkas was 6.39 ha with a core range of 1.21 ha and this was negatively related to population density. Male home ranges were larger than females but not significantly when the sexual size dimorphism was considered. Nocturnal ranges were larger than diurnal ranges reflecting nocturnal departures from the swamp refugia.

Home range sizes varied seasonally, probably due to changes in the distance required to move to obtain sufficient nutrients and water over the dry summer compared to the wet winter and spring.

Telemetry confirmed trapping results that showed no movement between swamps or populations. Home range centres shifted to the edge of the swamp following the winter Chapter 10 Conclusion inundation and this may increase the species susceptibility to predation at this time. The lack of dispersal is probably caused by quokka populations existing below carrying capacity. Without dispersal to recolonise or rescue unpopulated patches, the collapse of the original quokka metapopulation appears to have occurred.

Habitat use

On a macrohabitat scale, the quokka in the northern jarrah forest is restricted to Agonis swamp shrubland habitats that form in the open, upper reaches of creek systems on the western side of the forest. This restriction was probably initially due to the high water requirements of the quokka (Main and Bakker 1981) and this restricted habitat use and the associated movements between the habitat, meant quokkas existed in metapopulations.

This restriction is likely to have been exacerbated by increased predation pressure since the arrival of the fox. On a microhabitat scale, the quokka is a habitat specialist preferring early seral stage swamp habitats for foraging as part of a mosaic of older age swamp that provides refuge. Considering the small distances quokkas were found to move (Chapter 5) they are likely to require refuge habitat in close proximity to foraging areas.

Survivorship and mortality

Despite the six year old introduced predator control programme, foxes and cats are still the major non-human derived cause of mortality to quokkas. Road kills was the other major cause. Individuals alive at the start of the study had an 81% chance of staying alive for a year. The likelihood of dying was minimised by grouping together with conspecifics, Chapter 10 Conclusion maximising home range size and maximising the time spent within the swampy refuge.

Current rates of adult and juvenile survivorship should allow population recovery and so it seems pouch young mortality, reflected by low recruitment, has inhibited the anticipated population increase following predator control.

Modelling quokka presence or absence

The confounding effect of inadequate unbaited controls meant that little statistical evidence was available on the impact of introduced predators on the quokka, however the models provided support for earlier hypotheses of these. The presence of a quokka population at a site was related to the amount of poison baits delivered, the average age of the swamp and a mosaic of early and late seral stages within the swamp habitat. Poison baits are thought to relate to an decrease in predation pressure while recently burnt habitat provides food and long unburnt habitat provides refuge.

The prospects for quokkas at each individual site

Chandler

The quokka population at Chandler Road may already have gone extinct or emigrated.

Following the cessation of trapping in winter 1999 the Chandler Dam was drained and demolished at the request of the W.A. Water Authority. This led to the inundation of the downstream section of the swamp and silt was deposited to a depth of up to 40 cm for up to 100 metres downstream. No quokka footprints were observed in this area where they Chapter 10 Conclusion had once been relatively common. Furthermore, the maturing age structure of the swamp habitat suggests that the lifespan of this population may be reaching its limit.

Hadfield

The Hadfield site supports the highest known population density in the northern jarrah forest and, as such, appears most likely to survive. The destruction of swamp habitat in the downstream reaches may restrict migration and immigration and this may affect the longevity of this local population. It is surprising that the destruction of critical habitat for a threatened species can still occur, even on private property. The wildfire that began from this property highlights the ease at which stochastic events may arise and may eliminate this, and all, quokka populations. This is the ideal site to manage using in-situ conservation measures and these are discussed below.

Hoffman

The observation of a quokka at this site (A.Danks pers. comm.) was not confirmed and there is only relatively small amounts of suitable habitat present at the site. On this basis it is thought that there is no quokka population extant at the Hoffman site. Nonetheless, the swamp system at Hoffman is very large and quokkas may exist in other, unsurveyed, parts despite not being evidenced during such surveys.

Holyoake Chapter 10 Conclusion

The absence of fox control and the proximity to the residential and agricultural areas of

Dwellingup and the disturbances associated with this probably led to the localised extinction of the quokka here. Alternatively, this population had been known to be extant for a long period of time (at least 20 years; M. Dillon pers. comm.) and its extinction may simply be a natural extinction event as part of a metapopulation. The fact that very few new populations have been discovered means such events are of great concern.

Kesners

The Kesners site (Figure 1) has also supported a quokka population in the past however there was no evidence of quokkas during surveys in the early 1990s (M.Dillon pers. comm.). The quokka may have recolonised this site following the fire in the late

1980s and the population composition dominated by adults suggests that this population has surpassed its peak and is now heading toward localised extinction too. The disturbances associated with mining may have exacerbated the most recent extinction event here, however there seems to be no evidence of an extinction debt arising from this.

The quokka appears quite resilient to surviving near rehabilitated areas following disturbance. With the Hadfield site, Kesners also offers the best opportunity for in-situ conservation strategies. Chapter 10 Conclusion

Figure 1. Overview of the Kesners quokka swamp.

Rosella Road

The lone male at Rosella Road indicates localised extinction has effectively already occurred there. The absence of early seral stage habitat suggests this may be a natural event although the disturbance associated with mining may have expedited this process.

This would be a good site to monitor recolonisation by the quokka once a patchy burn has run through it. Alternatively, following such a burn this may be an appropriate site for translocation, although its disturbed nature and proximity to humans may impinge upon the success.

Chapter 10 Conclusion

Victor Road

Being the only extant population in the study that is unbaited, the Victor Road site highlights the importance of habitat to the quokka as a source of food and as refuge from predation. Over two and a half years, no quokka was recorded as being predated or dying there and this is probably due to behavioural strategies used by individuals to minimise predation risk but which have manifested at the population level. The ability of this population to maximise the time spent within the swamp, a factor shown to minimise mortality, while obtaining sufficient nutrients from the recently burnt sections have been successful. Despite this success, the precautionary approach should be to instigate introduced predator control at this site too.

Wild Pig

The population of quokkas previously known to occur at the Wild Pig site has apparently now gone locally extinct. The feral predator control programme may have been instigated too late or, more likely, the disturbance associated with the bauxite mine (Figure

2) may have been too great. There appears to be little likelihood of quokkas recolonising this site until mining has ceased and the mined areas have been rehabilitated. The recolonisation of the Kesners site following mine rehabilitation indicates this is possible. Chapter 10 Conclusion

Figure 2. Photograph showing the proximity of the mining disturbance to the Wild Pig quokka swamp.

The extinction risk for the quokka

Although there is general agreement that Diamond’s (1984) ‘evil quartet’ (habitat loss/alteration, over-exploitation, introduced species and chains of extinction) are the main processes driving extinction, the attributes of species that are prone to extinction are less well known (Purvis et al. 2000). Chapter 10 Conclusion

Table 1 below lists such attributes that correlate with vulnerability to extinction (Purvis et al. 2000) and compares the ecology of the quokka to them.

Table 1. Attributes thought to correlate with increased extinction risk in mammals and a discussion on whether they apply to the quokka. The predictive attributes come from Purvis et al. (2000) and the references therein.

Attribute correlating with extinction risk Part of the ecology of the quokka? i) Small population size as demographic stochasticity, Yes: today quokka population sizes are very small (Chapter 4). Even local catastrophe, slow adaptation rate, ‘mutational prior to the decline in the 1930s each local quokka population making up meltdown’ and inbreeding are all more serious for the metapopulation is likely to be relatively small (maybe 50 populations with few individuals. individuals).

ia) Small geographic range, which is often related to Yes: although prior to the population crash in the 1930s the quokka had small population size. a much larger range than today, it still only comprised the mesic south- western corner of Australia (Chapter 3). Today, with the small, isolated populations, this range is a fraction of its original size. ii) Island endemics which are likely to have small No: the insular populations are in fact the largest in population size and geographic ranges and subsequently small populations. possibly also individual population range (Chapter 4). The introduction They may also have evolved in isolation from predators of foxes to the islands however, is likely to be catastrophic. If Australia and competitors increasing their vulnerability to the is considered as an island rather than as a continental land mass (as I effects of introduced species and over-exploitation. have done) then this attribute is met. iii) Species at higher trophic levels are more vulnerable No: the quokka is not a higher trophic level species. to chains of extinction or the cumulative effects of disturbance than those at lower trophic levels. iv) Species with ‘slow’ life histories; such as small litter No: although the quokka has small litter sizes (1), it has early sexual sizes, slow growth rates, late sexual maturity, long maturity, short gestation (as do all marsupials) and continuous breeding gestation and long intervals between successive births; such that each female is capable of producing 17 progeny over a ten- are less able to cope with factors causing increased year lifetime (Chapter 4) mortality. v) Species with complex social structures for mating, No: there is no evidence of complex social structures on the mainland foraging or group defence because their persistence and the group territorial defence thought to exist on Rottnest Island

Attribute correlating with extinction risk Part of the ecology of the quokka? depends on a unit larger than the individual. (Holsworth 1964, 1967) has largely been discredited (Nicholls 1971; Kitchener 1972). vi) Species with large home ranges are susceptible to No: the quokka has small home ranges that reduce in size with habitat loss, alteration and edge effects. increasing population density (Chapter 5) and so home ranges are likely to have been smaller than recorded in this study prior to the 1930s population decline. vii) Diurnal species as they tend to be larger, social and No: the quokka is considered nocturnal or crepuscular (Packer 1965), also easier to hunt. although early records do suggest it was easy to hunt (Gould 1973; Kabay and Start 1976). viii) Large body sized species which tend to have low No: the quokka is a medium-sized mammal (Chapter 4), although population densities, slow life histories, larger home despite this it did become an important food of Aborigines (Chapter 3). ranges and a propensity to be persecuted by humans. Chapter 10 Conclusion

The quokka has only two of the attributes thought to convey an increased risk of extinction (small population size and small geographic range) (Table 1), although they are two of the four statistically significant attributes (Purvis et al. 2000). Purvis et al.’s model explains 48% of the variation and so it may be there are other attributes that may increase the ability to predict extinction risk.

Australian fauna, and marsupials in particular, may prove to be an anomaly to this type of modelling analysis as the majority of the larger bodied species went extinct in the

Pleistocene (Flannery 1990) and so Aborigines were forced to hunt smaller prey species.

This may be because Australia can be considered as a large island rather than as a continental land mass. Many parts of Australia are also largely deficient in nutrients and suffer aridity and unpredictability of rainfall. The marsupial fauna is generally deficient in species at higher trophic levels (Flannery 1991), has relatively fast and simple life histories compared to placental mammals and are infrequently strictly diurnal which would apparently preclude them from the risk of extinction under the Purvis et al. (2000) model. Yet Australia has suffered the highest rate of extinction for any country on the planet (Caughley and Gunn 1996) and medium-sized mammals have suffered the most

(Burbidge and McKenzie 1989). Clearly there are ecological attributes unique to marsupials or Australia that convey an increased extinction risk compared to placentals.

Old genera of Australian marsupials have a greater risk of extinction than more recently evolved species irrespective of body size or habitat use although intermediate size and non-forested habitat use are also factors (Johnson, Delean and Balmford 2002).

The evolutionary age of the quokka is poorly known (Chapter 3), however if it is found to be an old lineage then it may have a disproportionally high extinction risk particularly Chapter 10 Conclusion when coupled with its intermediate body size. Being the sole member of its genus, the quokka is phylogenetically distinct and its loss will result in a much greater depletion of biological diversity and evolutionary history than for a member of a more speciose genus

(Johnson et al. 2002).

With two, reasonably large, island populations and reported, but not validated, security in the southern parts of its range (G. Liddelow pers. comm.) the quokka appears secure for the near future. Why the populations in the south are not exhibiting population declines is unknown, but it may be that these more mesic areas have been buffered to an extent from earlier declines and they may only now be facing them. Ascertaining the status of populations in the south is important for any conclusions to be drawn on the species long-term future.

Other threatening processes

As well as introduced predators and altered fire regimes, there are other threatening processes that may indirectly affect quokkas. Feral pigs (Sus scrofa) were only captured once (at Chandler Road ~ Appendix A) and were only observed once each at Kesners and

Victor Road, but their diggings within the swamps were common throughout the majority of study sites and they are widely dispersed and common throughout the jarrah forest (P. de Tores pers. comm.). These animals have the potential to substantially alter the structure of the swamps by opening up the lower strata and such activities may have contributed to the localised extinction of the Holyoake swamp in the early 1990s as it aged (M. Dillon, pers. comm.). Efforts to control or eradicate pigs from the northern jarrah forest should be continued. Chapter 10 Conclusion

Climatic conditions may also indirectly affect quokkas. Currently the south-west of

Australia is in a 20 – 30 year period of below average rainfall (Chapter 3) and a return to above average rainfall may indirectly affect quokkas. Increased rainfall over winter caused quokkas to move to the edge of their swampy refugia (Chapter 5) which, in turn, may increase their susceptibility to predation (Chapter 7). Extended periods of above average rainfall may have indirectly assisted the fox in decimating quokka populations in the 1930s by reducing their time spent within the refuge (Chapter 3). Yet this increased mortality rate may be compensated for, to an extent, by the higher birth rates associated with above average rainfall (Holsworth 1964; Kitchener 1972).

The effect of global warming on the quokka may be similar. Furthermore, the quokkas’ restriction to specific habitats, including its affiliation with Agonis linearifolia swamp shrublands, means that its response to global warming may hinge upon the response of these environments.

The value of fox control to quokkas

While largely circumstantial, the evidence of the threat posed to native fauna by the introduced red fox and for the value of fox control is substantial. Significantly reducing fox numbers, or even removing foxes altogether, and thereby reducing predation has resulted in large increases in native fauna populations in Western Australia. Species to have benefited from fox baiting include the black-footed rock-wallaby (Petrogale lateralis) (Kinnear, Onus and Bromilow 1988; Kinnear, Onus and Sumner 1998),

Rothschild’s rock-wallaby (P. rothschildi) (de Tores 1994, 1999), the numbat

(Myrmecobius fasciatus) (Friend 1990), the tammar wallaby (Macropus eugenii), the Chapter 10 Conclusion western brush wallaby (M. irma), the brush-tailed bettong or woylie (Bettongia penicillata), the common brushtail possum (Trichosurus vulpecula), the western ringtail possum (Pseudocheirus occidentalis) (de Tores 1994, 1999) and the chuditch (Dasyurus geoffroii) (Morris 1992). Additionally translocations and re-introductions of native fauna have generally failed to establish or barely persist unless effective fox control is incorporated (Short et al. 1992). Despite this overwhelming evidence it has not been possible to irrefutably link the fox to the decline of native fauna (de Tores 1994), generally and like this study, because of the inadequacies of experimental controls used

(e.g. Hone 1999).

Similarly the value of fox control to the quokka is circumstantial. The inability to locate quokka populations, other than that at Victor Road, that were not already being baited meant that experimental controls for this study were inadequate. Nonetheless the quokka population at the Victor Road site had numerous anomalous features of its ecology that could be explained by the increased predation pressure. The fact that the quokka was not driven to extinction following the arrival of the fox, like so many other species, indicates it does have mechanisms to survive. The primary mechanism was probably a preference for habitat that offered refuge from predation, however the continuation, albeit slower, of the species’ decline suggests other factors, such as early seral stages for foraging, have become increasingly important. Without fox control, it seems likely that the quokka will eventually be driven to extinction.

Perhaps the future of Tasmania’s native mammals will provide the ultimate answer to the impact of the fox on Australia’s fauna. Anecdotes suggest a breeding fox population has become established there (Hodge 2002) and, unless they are exterminated, this Chapter 10 Conclusion founder population may well result in a similar decimation of critical weight range mammals to that which occurred on the mainland following European colonisation.

The future of fox control

The eradication of the dingo in the south-west of Australia by pastoralists may have inadvertently exacerbated the impact of the fox on native fauna there. Fox numbers are much lower in the presence of the dingo on the South Australian side of the Fence than the New South Wales side and so the dingo may be acting as a mesopredator

(Newsome 2001). Fox control led to increases in feral cat numbers in Shark Bay (Risbey et al. 2000) and as such the fox may be similarly acting as a mesopredator. The return of the dingo to areas identified as requiring feral predator control has been suggested as a solution to the costs of reducing fox numbers while it is acknowledged that this will lead to its own set of problems (Newsome 2001). It seems unlikely that poisoning will occur forever in Australia and, unless fertility control of feral predators is attained, only those native fauna that can cope with the increased predation pressure may survive.

Management recommendations

There are four major alternatives for the management of the quokka and these are considered below.

Do nothing approach Chapter 10 Conclusion

The first option is to do nothing specific to save the quokka. This would be based upon the assumption that the quokka will increase in abundance through the reduced predation pressure arising from the broad-scale, predator control programme that is being conducted as well as the standard control burning practices and regimes of forest managers. The broad-scale baiting programme is not specifically aimed at the quokka and so is considered a ‘do nothing’ option. While population increases may occur, the response of the quokka is likely to be much slo wer leaving the species susceptible to stochastic forces for much longer than necessary. Current burning regimes appear also to be inadequate to sufficiently restore habitat.

Ex-situ management

Quokkas are already held in captive colonies in Western Australia including the Perth

Zoo, Karakamia Sanctuary and the University of Western Australia’s Shenton Park facility. In addition, the Rottnest and Bald Island populations can be managed on an ex- situ basis. These populations can be used if in-situ methods fail. The genetic differences between island and mainland populations (Sinclair 1999; Alacs 2001; Sinclair 2001) are likely to be a major problem and impediment to effective, long-term ex-situ conservation unless adequate breeding practices are maintained. If relocations are required to augment the mainland stock, then predator recognition trials to teach predator-naïve individuals to be predator-aware are recommended (McLean, Lundie-Jenkins and Jarman 1996;

McLean, Holzer and Studholme 1999; Blumstein et al. 2000; McLean et al. 2000).

Chapter 10 Conclusion

Individual site management

Each individual population can be managed with the aim of building numbers into a source population so that, when the carrying capacity of each site is reached, individuals can recolonise adjacent habitat patches. For this to be achieved, the monthly predator control around each population should be maintained and possibly increased in area to include adjacent swamps. It would be hoped that once carrying capacity at each site is attained then dispersal and recolonisation would occur without management intervention.

Habitat management would also be required which would involve small scale burns being conducted at each site to create a mosaic of freshly burnt and long unburnt swamp.

Following such burns, the addition of food (kangaroo pellets) to sites may maximise the rate of increase of the population at low density (Pech, Sinclair and Newsome 1995).

Additional mosaic burning of adjacent habitat patches would increase the chances of recolonisation being successful. To ensure populations are resilient to fox predation if introduced predator control ceases, individual population sizes of at least 50 may be necessary (Sinclair et al. 1998). As such figures are conceptual and as dispersal is likely to be driven by high population densities, the estimates of Kitchener (1972) may be used as indications of the densities to be sought on the mainland before anticipating dispersal or intervening (~ 18 individuals ha-1). Further study on the carrying capacity of quokka sites is also important.

The threat to quokka populations from road kills may be reduced through the use of wildlife underpasses that also serve as cross-drainage structures. Appropriately designed underpasses seem likely to be readily used by the quokka because of its preference for densely vegetated areas. If such structures are used they should be monitored to ensure Chapter 10 Conclusion they are used and don’t become predator attractants (Foster and Humphrey 1995;

Rodriguez, Crema and Delibes 1996; Clevenger and Waltho 2000).

Concentrate valuable resources on sustainable populations

This option would involve the relocation of quokkas from the smaller populations to the larger ones. These larger populations would then be managed individually (see above) until local metapopulatio n structure is restored.

Perhaps the most significant conclusion of this study is the need for wildlife researchers to monitor threatened species following management actions. Simply assuming that the Western Shield fox control programme was successful for quokkas may have led to the species’ habitat requirements being overlooked as a source of their continued scarcity. As such, for each of these options, long-term monitoring of the response of the quokka is essential.

These management options may be considered from the view of Caughley’s (1994) dicotomy of the declining- or small- population paradigms. I prefer the declining- population paradigm for the existing state of quokka abundance because of the perception that conservation can still occur for the species in situ. In particular, I would recommend focussing conservation resources at the Hadfield, Kesners, Victor Road and, possibly,

Chandler sites and rebuilding the metapopulation from these sources.

Quokka populations also appear inadequately preserved in conservation estate, particularly in the northern third of the species distribution (Chapter 3). Reserve design focusing on the quokka in this area should focus on existing local populations that have Chapter 10 Conclusion sufficient numbers of swamps surrounding them to allow a small metapopulation to re- form. While a vertebrate-biased, species-centred approach to biodiversity conservation is considered inadequate and obsolete (Bright and Morris 2000) compared to the value of landscape and ecosystem conservation, the two are not necessarily mutually-exclusive.

By promoting the reservation of areas for the conservation of the quokka in the northern jarrah forest, broader landscapes and ecosystems in and around riparian systems are conserved. Furthermore, these conservation areas will promote connectivity to nearby conservation estate. By creating a conservation reserve in an area of high quokka abundance, Rodriguez’s (2002) concerns of a low probability of long-term persistence that occurs through the reservation of land containing small, remnant populations at a species’ periphery may be minimised.

Summation

The decline of the quokka in the 1930s was devastating in size and scale. This decline was probably caused by the red fox in conjunction with altered fire regimes and habitat loss. Quokka populations continued to dwindle as part of a non-equilibrium metapopulation such that they are now considered as vulnerable to extinction.

The existing introduced predator control programme may be enough for the species to, once again, become abundant, however appropriate habitat management with fire is also likely to be imperative. While attempts are made to resurrect the metapopulation structure, continued monitoring is crucial to ensure that the quokka remains an icon of

Western Australian fauna. References

References

AAP (1998). Youth charged for 'quokka soccer'. In The Newcastle Herald, pp 3. Tuesday, December 8, 1998

Aars, J., Lambin, X., Denny, R. and Griffin, A.C. (2001). Water vole in the Scottish uplands: distribution patterns of disturbed and pristine populations ahead and behind the American mink invasion front. Animal Conservation 4, 187-194.

Abbott, I. (2002). Origin and spread of the cat, Felis catus, on mainland Australia, with a discussion of the magnitude of its early impact on native fauna. Wildlife Research 29, 51-74.

Abbott, I., van Heurck, P. and Wong, L. (1985). Responses to long-term fire exclusion: physical, chemical and faunal features of litter and soil in a Western Australian forest. Australian Forestry 47, 237-242.

ABS (Undated). Australian Historical Population Statistics I. Population size and growth., Rep. No. 3105.0.65.001 www.abs.gov.au/austats/abs%40nsf/c08c69053a26f3e2ca256865007b861a. (Australian Bureau of Statistics: Canberra).

Akaike, H. (1973). Information theory and an extension of the maximum likelihood principle. In 'Proceedings of the Second International Symposium on Information Theory' (Eds. N. Petrov and F. Csadki) pp. 267-281. (Akademiai Kiado: Budapest).

Akaike, H. (1974). A new look at the statistical model identification. IEEE Transactions on Automatic Control AC 19, 716-723.

Alacs, E. (2001). Conservation Genetics of the Quokka, Setonix brachyurus. Honours thesis, Murdoch University, Perth, Western Australia. References

Alacs, E., Alpers, D., de Tores, P., Dillon, M. and Spencer, P.B.S. (2002). Identifying the presence of quokkas (Setonix brachyurus) using cytochrome b analyses from faeces collected in the field. Wildlife Research In Review.

Algar, D. and Kinnear, J.E. (1996). Secondary poisoning of foxes following a routine 1080 rabbit-baiting campaign in the Western Australian wheatbelt. CALMScience 2, 149-152.

Allan, R.J., Mitchell, C.D. and Pittock, A.B., Eds. (1991). The Greenhouse Effect - Regional Implications for Western Australia. (CSIRO Division of Atmospheric Research: Perth).

Anderson, A. (1990). Reply to Flannery. Archaeology in Oceania 25, 63-64.

Anon (2001-2002). Bush Telegraph - rare birds survive Nuyts fire. Landscope 17, 7.

Anon (1999). Forest Management Plan Ministerial Conditions, Western Australian Department of Conservation and Land Management.

Anon. (1999). Forest Management Plan Ministerial Conditions, Western Australian Department of Conservation and Land Management.

Archer, M. and Baynes, A. (1972). Prehistoric mammal faunas from two small caves in the extreme south-west of Western Australia. Journal of the Royal Society of Western Australia 55, 80-90.

Augee, M.L., Smith, B. and Rose, S. (1996). Survival of wild and hand-reared ringtail possums (Pseudocheirus peregrinus) in bushland near Sydney. Wildlife Research 23, 99-108.

Bakker, H.R. and Main, A.R. (1980). Condition, body composition and total body water estimation in the quokka, Setonix brachyurus (Macropodidae). Australian Journal of Zoology 28, 395-406. References

Balme, J., Merrilees, D. and Porter, J.K. (1978). Late Quaternary mammal remains, spanning 30,000 years, from excavations in Devil's Lair, Western Australia. Journal of the Royal Society of Western Australia. 61, 33-65.

Balmford, A. (1996). Extinction filters and current resilience: the significance of past selection pressures for conservation biology. Trends in Research of Ecology and Evolution 11, 193-196.

Banks, P.B. (1997). Predator-prey interactions between foxes, rabbits and native mammals of the Australian Alps. Ph.D. thesis, University of Sydney, Sydney.

Banks, P.B. (2001). Predation-sensitive grouping and habitat use by eastern grey kangaroos: a field experiment. Animal Behaviour 61, 1013-1021.

Banks, P.B., Hume, I. and Crowe, O. (1999). Behavioural, morphological and dietary response of rabbits to predation risk from foxes. Oikos 85, 247-256.

Banks, P.B., Newsome, A.E. and Dickman, C.R. (2000a). Predation by red foxes limits recruitment in populations of eastern grey kangaroo. Austral Ecology 25, 283- 291.

Banks, P.B., Norrdahl, K. and Korpimaki, E. (2000b). Nonlinearity in the predation risk of prey mobility. Proceedings of the Royal Society of London (Series B) 267, 1621-1625.

Banks, P.B., Norrdahl, K., Nordstrom, M. and Korpimaki, E. (In review). Impacts of feral mink predation and weather on a vole metapopulation in the outer archipelago of the Baltic Sea.

Barker, R.D. and Vestjens, W.J.M. (1989/90). The Food of Australian Birds. CSIRO Publishing, Canberra.

Barker, S. (1961). Studies on marsupial nutrition. III. The copper-molybdenum- inorganic sulphate interaction in the Rottnest quokka, Setonix brachyurus (Quoy & Gaimard). Australian Journal of Biological Science 14, 646-658. References

Barker, S., Main, A.R. and Sadlier, R.M. (1957). Recent capture of the quokka (Setonix brachyurus) on the mainland. The Western Australian Naturalist 6, 53-55.

Bartholomew, G.A. (1954). Temperature regulation in the macropod marsupial, Setonix brachyurus. Australian Journal of Science, 26-40.

Bartle, J. (1987). The geology of the jarrah forest. Resource Notes, 1-4.

Baverstock, P.R., Richardson, B.J., Birrell, J. and Krieg, M. (1989). Albumin immunologic relationships of the Macropodidae (Marsupialia). Systematic Zoology 38, 38-50.

Baynes, A. (1985). The original mammal fauna of the Nullarbor and southern peripheral regions: evidence from skeletal remains in superficial cave deposits. In 'A Biological Survey of the Nullarbor Region South and Western Australia in 1984' (Eds. N.L. McKenzie and A.C. Robinson) pp. 139-151. (South Australian Department of Environment and Planning: Adelaide).

Baynes, A., Merrilees, D. and Porter, J.K. (1975). Mammal remains from the upper levels of a late Pleistocene deposit in Devil's Lair, Western Australia. Journal of the Royal Society of Western Australia 58, 97-126.

Beard, J.S. (1980). A new phytogeographic map of Western Australia. Western Australian Herbarium Research Notes 3, 37-58.

Beck, M.W. (1996). On discerning the cause of late Pleistocene megafaunal extinctions. Palaeobiology 22, 91-103.

Behn, G., McKinnell, F.H., Caccetta, P. and Vernes, T. (2001). Mapping forest cover, Kimberley region of Western Australia. Australian Forestry 64, 80-87.

Bell, D.T. (1995). Influence of fire on the seed germination ecology of species of the jarrah forest. CALMScience 4, 212. References

Bennett, A.F. (1993). Microhabitat use by the long-nosed potoroo, Potorous tridactylus, and other small mammals in remnant forest vegetation of south-western Victoria. Wildlife Research 20, 267-285.

Berdoy, M., Webster, J.P. and Macdonald, D.W. (2000). Fatal attraction in rats infected with Toxoplasma gondii. Proceedings of the Royal Society of London (Series B) 267, 1591-1594.

Berger, J., Swenson, J.E. and Persson, I.-L. (2001). Recolonizing carnivores and naive prey: conservation lessons from Pleistocene extinctions. Science 291, 1036-1039.

Berteaux, D., Masseboeuf, F., Bonzom, J.-M., Bergeron, J.-M., Thomas, D. and Lapierre, H. (1996). Effect of carrying a radiocollar on expenditure of energy by meadow voles. Journal of Mammalogy 77, 359-363.

Biggs, E.R., Wilde, S.A. and Leech, R.E.J. (1980). Geology, Mineral Resources and Hydrogeology of the Darling System, Western Australia. In 'Atlas of Natural Resources, Darling System, Western Australia.' pp. 3-20. (Department of Conservation and Environment: Perth, Western Australia).

Blumstein, D.T., Daniel, J.C., Griffin, A.S. and Evans, C.S. (2000). Insular tammar wallabies (Macropus eugenii) respond to visual but not acoustic cues from predators. Behavioural Ecology 11, 528-535.

Bolton, B.L. and Latz (1978). The western hare-wallaby, Lagorchestes hirsutus (Gould) (Macropodidae), in the Tanami Desert. Australian Wildlife Research 5, 285-293.

Bolton, G.C. and Hutchinson, D.E. (1979). The beginning. In 'Environment and Science' (Ed. B.J. O'Brien) pp. 1-21. (University of Western Australia: Perth).

Bos, D.G., Carthew, S.M. and Lorimer, M.F. (2002). Habitat selection by the small dasyurid Ningaui yvonneae (Marsupialia: ) in . Austral Ecology 27, 103-109.

Bowdler, S. (1990). Reply to Flannery. Archaeology in Oceania 25, 61-63. References

Braithwaite, L.W., Turner, J. and Kelly, J. (1984). Studies on the arboreal marsupial fauna of eucalypt forests being harvested for woodpulp at Eden, N.S.W. III. Relationships between faunal densities, eucalypt occurrence and foliage nutrients, and soil parent materials. Australian Wildlife Research 11, 41-48.

Braun-Blanquet, J. (1932). Plant Sociology: The Study of Plant Communities. Hafner, London.

Bright, P.W. and Morris, P.A. (2000). Rare mammals, research and realpolitik: priorities for biodiversity and ecology? In 'Priorities for the Conservation of Mammalian Diversity: Has the Panda had its Day?' (Eds. A. Entwistle and N. Dunstone) pp. 141-158. (Cambridge University Press: Cambridge, UK).

Brown Cooper, R., Robinson, D. and Maryan, B. (1989). The amphibia, reptile and mammal fauna of the Murray-Serpentine River delta, south west, Western Australia. The Western Australian Naturalist 18, 40-51.

Brown, J.H. and Kodric-Brown, A. (1977). Turnover rates in insular biogeography: effect of immigration on extinction. Ecology 58, 445-449.

Brunner, H., Moro, D., Wallis, R. and Andrasek, A. (1991). Comparison of the diets of foxes, dogs and cats in an urban park. Victorian Naturalist 108, 34-37.

Buckland, S.T. and Elston, D.A. (1993). Empirical models for the spatial distribution of wildlife. Journal of Applied Ecology 30, 478-495.

Burbidge, A.A., Johnson, K.A., Fuller, P.J. and Southgate, R.I. (1988). Aboriginal knowledge of the mammals of the central deserts of Australia. Australian Wildlife Research 15, 9-39.

Burbidge, A.A. and McKenzie, N.L. (1989). Patterns in the modern decline of Western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation 50, 143-198. References

Burk, A. and Springer, M.S. (2000). Intergenetic relationships among Macropodoidea (Metatheria: Diprodontia) and the chronicle of kangaroo evolution. Journal of Mammalian Evolution 7, 213-237.

Burk, A., Westerman, M. and Springer, M.S. (1998). The phylogenetic position of the musky rat-kangaroo and the evolution of bipedal hopping in kangaroos (Macropodidae: Diprodontia). Systematic Biology 47, 457-474.

Burnet, S.M. (1970). Changing Patterns: an atypical autobiography. Sun Books, Melbourne.

Burnham, K.P. and Anderson, D.J. (2001). Kullback-Leibler information as a basis for strong inference in ecological studies. Wildlife Research 28, 111-119.

Burrows, N.D., Ward, B. and Robinson, A.D. (1995). Jarrah forest fire history from stem analysis and anthropological evidence. Australian Forestry 58, 7-16.

Busby, J.R. (1991). BIOCLIM - a bioclimatic analysis and prediction system. In 'Nature Conservation: Cost effective biological surveys and data analysis.' (Eds. C.R. Margules and M.P. Austin) pp. 64-68. (CSIRO: Melbourne).

Butler, W.H. (1968). Remains of Sarcophilus the "Tasmanian" Devil (Marsupialia, Dasyuridae) from coastal dunes south of Scott River, Western Australia. The Western Australian Naturalist 11, 87-89.

Calver, M.C. and Dell, J. (1998a). Conservation status of mammals and birds in southwestern Australian forests. I. Is there evidence of direct links between forestry practices and species decline and extinction? Pacific Conservation Biology 4, 296-314.

Calver, M.C. and Dell, J. (1998b). Conservation status of mammals and birds in southwestern Australian forests. II. Are there unstudied, indirect or long-term links between forestry practices and species decline and extinction. Pacific Conservation Biology 4, 315-325. References

Carlsen, M., Lodal, J., Leirs, H. and Jensen, T.S. (1999). The effect of predation risk on body weight in the field vole, Microtus agrestis. Oikos 87, 277-285.

Carrete, M., Sanchez-Zapata, J.A. and Calvo, J.F. (2000). Breeding densities and habitat attributes of golden eagles in southeastern Spain. Journal of Raptor Research 34, 48-52.

Cassels, R. (1984). The role of prehistoric man in the faunal extinctions of New Zealand and other Pacific Islands. In 'Quaternary Extinctions - a prehistoric revolution.' (Eds. P.S. Martin and R.G. Klein) pp. 741-767. (The University of Arizona Press: Tucson, USA).

Catling, P.C. (1988). Similarities and contrasts in the diets of foxes, Vulpes vulpes, and cats, Felis catus, relative to fluctuating prey populations and drought. Australian Wildlife Research 15, 307-317.

Catling, P.C. and Burt, R.J. (1995). Studies of the ground-dwelling mammals of eucalypt forests in south-eastern New South Wales: the effect of habitat variables on distribution and abundance. Wildlife Research 22, 271-288.

Catling, P.C., Burt, R.J. and Forrester, R.I. (2000). Models of the distribution and abundance of ground-dwelling mammals in the eucalypt forests of north-eastern New South Wales in relation to habitat variables. Wildlife Research 27, 639-654.

Catling, P.C. and Coops, N.C. (1999). Prediction of the distribution and abundance of small mammals in the eucalypt forests of south-eastern Australia from airborne videography. Wildlife Research 26, 641-650.

Catling, P.C., Coops, N.C. and Burt, R.J. (2001). The distribution and abundance of ground-dwelling mammals in relation to time since wildfire and vegetation structure in south-eastern Australia. Wildlife Research 28, 555-564. References

Caughley, G. (1987). Introduction to the sheep rangelands. In 'Kangaroos: their ecology and management in the sheep rangelands of Australia.' (Eds. G. Caughley, N. Shepherd and J. Short) pp. 1-13. (Cambridge University Press: Sydney).

Caughley, G. (1994). Directions in conservation biology. Journal of Animal Ecology 63, 215-244.

Caughley, G. and Gunn, A. (1996). Conservation biology in theory and practice. Blackwell Science, Carlton, Victoria.

Caughley, G. and Sinclair, A.R.E. (1994). Wildlife Ecology and Management. Blackwell Science, Massachusetts, USA.

Channell, R. and Lomolino, M.V. (2000). Dynamic biogeography and conservation of endangered species. Nature 403, 84-86.

Christensen, P.E.S. (1980). The biology of Bettongia penicillata Gray, 1837, and Macropus eugenii (Desmarest, 1817) in relation to fire., Rep. No. 91. (Forests Department of Western Australia: Perth, Western Australia).

Christensen, P.E.S., Annels, A., Liddelow, G. and Skinner, P. (1985). Vertebrate fauna in the southern forests of Western Australia., Rep. No. 94. (Forests Department of Western Australia: Perth).

Christensen, P.E.S. and Kimber, P.C. (1975). Effect of prescribed burning on the flora and fauna of south-western Australian forests. Proceedings of the Ecological Society of Australia. 9, 85-106.

Churchill, D.M. (1959). Late Quaternary eustatic changes in the Swan River District. Journal of the Royal Society of Western Australia. 42, 53-55.

Churchward, H.M. and McArthur, W.M. (1980). Landforms and Soils of the Darling System, Western Australia. In 'Atlas of Natural Resources, Darling System, Western Australia.' pp. 25-33. (Department of Conservation and Environment: Perth, Western Australia). References

Claridge, A.W. and Cork, S.J. (1994). Nutritional value of hypogeal fungal sporocarps for the long-nosed potoroo (Potorous tridactylus), a forest-dwelling mycophagous marsupial. Australian Journal of Zoology 42, 701-710.

Clarke, J.R. (1947). Anatomy of the quokka (Setonix brachyurus [Quoy & Gaimard]). I. External morphology and large intestine. Journal of the Royal Society of Western Australia 33, 59-148.

Clevenger, A.P. and Waltho, N. (2000). Factors influencing the effectiveness of wildlife underpasses in Banff National Park, Alberta, Canada. Conservation Biology 14, 47-56.

Clutton-Brock, T.H. and Iason, G.R. (1986). Sex ratio variation in mammals. The Quarterly Review of Biology 61, 339-375.

Cochran, W.G. (1947). Some consequences when the assumptions for the analysis of variance are not satisfied. Biometrics 3, 22-38.

Cogger, H.G. (1992). Reptiles and Amphibians of Australia. Reed Books, Sydney.

Collins, L.R. (1973). Monotremes and Marsupials - a reference for zoological institutions. Smithsonian Institution Press, Washington.

Cook, D.L. (1960). Some mammal remains found in caves near Margaret River. The Western Australian Naturalist 7, 107-108.

Cook, D.L. (1963). The fossil vertebrate fauna of Strongs' Cave, Boranup, Western Australia. The Western Australian Naturalist 8, 153-162.

Coops, N., Culvenor, D., Preston, R. and Catling, P. (1998). Procedures for predicting habitat and structural attributes in eucalypt forests using high spatial resolution remotely sensed imagery. Australian Forestry 61, 244-252. References

Coops, N.C. and Catling, P.C. (1997a). Predicting the complexity of habitat in forests from airborne videography for wildlife management. International Journal of Remote Sensing 18, 2677-2682.

Coops, N.C. and Catling, P.C. (1997b). Utilising airborne multispectral videography to predict habitat complexity in eucalypt forests for wildlife management. Wildlife Research 24, 691-703.

Coops, N.C. and Catling, P.C. (2000). Estimating forest habitat complexity in relation to time since fire. Austral Ecology 25, 344-351.

Corbett, L.K. (1995a). Dingo. In '' (Ed. R. Strahan) pp. 696-698. (Reed Books: Sydney, Australia).

Corbett, L.K. (1995b). The Dingo in Australia and Asia. UNSW Press, Sydney.

Corbett, L.K. and Newsome, A.E. (1987). The feeding ecology of the dingo. III. Dietary relationships with widely fluctuating prey populations in arid Australia: an hypothesis of alteration of predation. Oecologia 74, 215-227.

Cork, S.J. and Catling, P.C. (1996). Modelling distributions of arboreal and ground- dwelling mammals in relation to climate, nutrients, plant chemical defences and vegetation structure in the eucalypt forests of southeastern Australia. Forest Ecology and Management 85, 163-175.

Coulson, G. (1997). Male bias in road-kills of macropods. Wildlife Research 24, 21-25.

Courchamp, F., Langlais, M. and Sugihara, G. (1999). Cats protecting birds: modelling the mesopredator release effect. Journal of Animal Ecology 68, 282-292.

Crabb, D. (1972). Quokka on the Albany Highway. The Western Australian Naturalist 12, 141. References

Croft, D.B. (1989). Social organisation of the Macropodoidea. In 'Kangaroos, Wallabies and Rat-kangaroos' (Eds. G. Grigg, P.J. Jarman and I. Hume), Vol. 2 pp. 505-525. (Surrey Beatty & Sons Pty Ltd: Sydney, Australia).

Crone, E.E., Doak, D.F. and Pokki, J. (2001). Ecological influences on the dynamics of a field vole metapopulation. Ecology 82, 831-843.

DCE (1980). Atlas of Natural Resources - Darling System, Western Australia. Western Australian Department of Conservation and Environment (DCE), Perth. de Tores, P. (1994). Fox Control Manual - operational guidelines for control of the red fox, Vulpes vulpes, through the use of Sodium monofluoroacetate or '1080' on CALM managed estate and in other CALM programs. (Western Australian Department of Conservation and Land Management: Perth). de Tores, P. (1999). Control and Ecology of the Red Fox in Western Australia - prey response to 1080 baiting over large areas. (CALMScience, Wildlife Research Centre, Department of Conservation and Land Management, Western Australia: Wanneroo). de Tores, P.J., Hayward, M.W., Dillon, M.J., Liddelow, G. and Brazell, R. (In prep.). Review of the distribution and conservation status of the quokka, Setonix brachyurus, (Macropodidae: Marsupialia). Journal of Biogeography.

Dehn, M.M. (1990). Vigilance for predators: detection and dilution effects. Behavioural Ecology and Sociobiology 26, 337-342.

Dell, J. and How, R.A. (1988). Mammals of the Darling Scarp, near Perth. The Western Australian Naturalist 17, 86-93.

Denny, M. (1985). The red kangaroo and the arid environment. In 'The Kangaroo Keepers' pp. 55-57. (University of Press: Brisbane).

Diamond, J.M. (1975). The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7, 129-146. References

Diamond, J.M. (1984). Historic extinctions: a rosetta stone for understanding prehistoric extinctions. In 'Quaternary Extinctions - a prehistoric revolution.' (Eds. P.S. Martin and R.G. Klein) pp. 824-862. (The University of Arizona Press: Tucson, USA).

Diamond, J.M. (1989). Overview of recent extinctions. In 'Conservation for the Twenty- first Century.' (Eds. D. Western and M.C. Pearl) pp. 37-54. (Wildlife Conservation International, New York Zoological Society: New York, USA).

Dickman, C.R. (1992a). Conservation of mammals in the Australasian region: the importance of islands. In 'Australia and the Global Environmental Crisis' (Eds. J.N. Coles and J.M. Drew) pp. 175-214. (Academic Press: Canberra).

Dickman, C.R. (1992b). Predation and habitat shift in the house mouse, Mus domesticus. Ecology 73, 313-322.

Dickman, C.R. (1993). Raiders of the last ark: cats in island Australia. Australian Natural History 24, 44-53.

Dickman, C.R. (Undated). Rottnest Island - the quokka. (Information Pamphlet produced for the Rottnest Island Authority: Thomson Bay, Rottnest Island).

Dixon, K.R. and Chapman, J.A. (1980). Harmonic mean measure of animal activity areas. Ecology 61, 1040-1044.

Dodson, J., Fullagar, R., Furby, J., Jones, R. and Prosser, I. (1993). Humans and megafauna in a late Pleistocene environment from Cuddie Springs, north western New South Wales. Archaeology in Oceania 28, 94-99.

Dor geloh, W.G. (2001). A draft habitat suitability model for roan antelope in the Nylsvley Nature Reserve, South Africa. African Journal of Ecology 39, 313-316.

Dortch, C.E. and Merrilees, D. (1971). A salvage excavation in Devil's Lair, Western Australia. Journal of the Royal Society of Western Australia. 54, 103-113. References

Dunnet, G.M. (1962). A population study of the quokka, Setonix brachyurus (Quoy & Gaimard) (Marsupialia). II. Habitat, movements, breeding and growth. CSIRO Wildlife Research 7, 13-32.

Dunnet, G.M. (1963). A population study of the quokka, Setonix brachyurus Quoy & Gaimard (Marsupialia). III. The estimation of population parameters by means of the recapture technique. CSIRO Wildlife Research 8, 78-117.

Eason, C.T. and Frampton, C.M. (1991). Acute toxicity of sodium monofluoroacetate (1080) baits to feral cats. Wildlife Research 18, 445-449.

Edgar, R. and Belcher, C. (1995). Spotted-tailed quoll (Dasyurus maculatus). In 'Mammals of Australia' (Ed. R. Strahan) pp. 67-68. (Reed Books: Sydney, Australia).

Eisenhart, C. (1947). The assumptions underlying the analysis of variance. Biometrics 3, 1-21.

Eldridge, M.D.B. and Close, R.L. (1995). Allied rock-wallaby, Petrogale assimilis. In 'The Mammals of Australia' (Ed. R. Strahan) pp. 365-366. (The Australian Museum/Reed Books: Sydney).

Eldridge, N. (1999). Cretaceous meteor showers, the human ecological 'niche', and the sixth extinction. In 'Extinctions in Near Time' (Ed. R.D.E. MacPhee) pp. 1-16. (Kluwer Academic/Plenum Publishers: New York).

Elmhagen, B. and Angerbjorn, A. (2001). The applicability of metapopulation theory to large mammals. Oikos 94, 89-100.

Etienne, R.S. and Heesterbeek, J.A.P. (2001). Rules of thumb for conservation of metapopulations based on a stochastic winking-patch model. The American Naturalist 158, 389-407.

Flannery, T.F. (1984). Kangaroos: 15 million years of Australian bounders. In 'Vertebrate Zoogeography and Evolution in Australasia: Animals in Space and Time.' (Eds. References

M. Archer and G. Clayton) pp. 817-835. (Hesperian Press: Carlisle, Western Australia.).

Flannery, T.F. (1989). Phylogeny of the Macropodoidea; a study in convergence. In 'Kangaroos, Wallabies and Rat-kangaroos.' (Eds. G.C. Grigg, P.J. Jarman and I.D. Hume), Vol. 1 pp. 1-46. (Surrey Beatty & Sons Pty Ltd: Sydney, New South Wales, Australia.).

Flannery, T.F. (1990a). Pleistocene faunal loss: implications of the aftershock for Australia's past and future. Archaeology in Oceania 25, 45-67.

Flannery, T.F. (1990b). Reply. Archaeology in Oceania 25, 64-65.

Flannery, T.F. (1991). The mystery of the Meganesian meat-eaters. Australian Natural History 23, 722-729.

Flannery, T.F. (1994). The Future Eaters: An Ecological History of the Australasian Lands and People. Reed Books, Sydney.

Flannery, T.F. and Roberts, R.G. (1999). Late Quaternary extinctions in Australasia: an overview. In 'Extinctions in Near Time.' (Ed. R.D.E. MacPhee) pp. 239-255. (Kluwer Academic/Plenum Publishers: New York).

Foster, M.L. and Humphrey, S.R. (1995). Use of highway underpasses by Florida panthers and other wildlife. Wildlife Society Bulletin 23, 95-100.

Fox, B.J. (1990). Changes in the structure of mammal communities over successional time scales. Oikos 59, 321-329.

Fox, B.J., Fox, M.D. and McKay, G.M. (1979). Litter accumulation after fire in a eucalypt forest. Australian Journal of Botany 27, 157-165.

Fox, M.D. (1988). Understorey changes following fire at Myall Lakes, New South Wales. Cunninghamia 2, 85-95. References

Frankham, R. (1995). Inbreeding and extinction: a threshold effect. Conservation Biology 9, 792-799.

Friend, J.A. (1990). The numbat Myrmecobius fasciatus (Myrmecobiidae): history of decline and potential for recovery. Proceedings of the Ecological Society of Australia. 16, 369-377.

Friend, J.A. and Thomas, N.D. (1994). Reintroduction and the numbat recovery programme. In 'Reintroduction Biology of Australian and New Zealand Fauna' (Ed. M. Serena) pp. 189-198. (Surrey beatty & Sons Pty Ltd: Chipping Norton, Sydney, Australia).

Garden, D.S. (1979). Northam: An Avon Valley History. Oxford University Press, Melbourne.

Gardner, C.A. (1957). The fire factor in relation to the vegetation of Western Australia. The Western Australian Naturalist 5, 166-173.

Gaston, K.J. (1990). Patterns in the geographical ranges of species. Biological Review 65, 105-129.

Gaston, K.J. (1991). How large is a species' geographic range? Oikos 61, 434-437.

Gauch, H.G.J. (1986). Multivariate Analysis in Community Ecology Cambridge University Press, Cambridge, UK.

Gaynor, A. (2000). Report on the history of the arrival of the feral cat population in Western Australia. CALMScience 3, 149-179.

Gerhart, K.L., White, R.G., Cameron, R.D. and Russell, D.E. (1996). Estimating fat content of caribou from body condition scores. Journal of Wildlife Management 60, 713-718. References

Gibb, D.G.A., Kakulas, B.A., Perrett, D.H. and Jenkyn, D.J. (1966). Toxoplasmosis in the Rottnest quokka (Setonix brachyurus). Australian Journal of Experimental Biology and Medical Science 44, 665-672.

Gibson, D.F., Lundie-Jenkins, G., Langford, D.G., Cole, J.R., Clarke, D.E. and Johnson, K.A. (1994). Predation by feral cats, Felis catus, on the rufous hare-wallaby, Lagorchestes hirsutus, in the Tanami Desert. Australian Mammalogy 17, 103- 107.

Gilfillan, S. (2001). An ecological study of a population of Pseudantechinus macdonnellensis (Marsupialia: Dasyuridae) in central Australia. II. Population dynamics and movements. Wildlife Research 28, 481-492.

Gill, J. (2000). Generalized Linear Models: A Unified Approach. Sage University, Thousand Oaks, CA.

Gilpin, M. (1991). The genetic effective size of a metapopulation. Biological Journal of the Linnean Society 42, 165-175.

Ginsberg, J.R., Alexander, K.A., Creel, S., Kat, P.W., McNutt, J.W. and Mills, M.G.L. (1995). Handling and survivorship of African wild dog (Lycaon pictus) in five ecosystems. Conservation Biology 9, 665-674.

Gittleman, J.L. and Gompper, M.E. (2001). The risk of extinction - what you don't know will hurt you. Science 291, 997-999.

Glauert, L. (1932-33). The distribution of the marsupials in Western Australia. Journal of the Royal Society of Western Australia 19, 17-32.

Glauert, L. (1948). The cave fossils of the south-west. The West Australian Naturalist 1, 100-104.

Glauert, L. (1950). The development of our knowledge of the marsupials of Western Australia. Journal of the Royal Society of Western Australia. 34, 115-134. References

Goldingay, R.L. and Kavanagh, R.P. (1993). Home-range estimates and habitat of the yellow-bellied glider ( australis) at Waratah Creek, New South Wales. Wildlife Research 20, 387-404.

Gould, J. (1973). Kangaroos. The Macmillan Company of Australia Pty Ltd, Melbourne.

Grant, C.D. and Loneragan, W.A. (1999). The effects of burning on the understorey composition of 11-13 year-old rehabilitated bauxite mines in Western Australia. Plant Ecology 145, 291-305.

Grayson, D.K. (1990). Reply to Flannery. Archaeology in Oceania 25, 60-61.

Greenwood, P.J. (1980). Mating systems, philopatry and dispersal in birds and mammals. Animal Behaviour 28, 1140-1162.

Griffioen, P.A. and Clarke, M.F. (2002). Large-scale bird-movement patterns evident in eastern Australian atlas data. Emu 102, 99-125.

Hanski, I. (1999). Metapopulation Ecology. Oxford University Press, UK.

Hanski, I. and Gilpin, M. (1991). Metapopulation dynamics: brief history and conceptual domain. Biological Journal of the Linnean Society 42, 3-16.

Hanski, I., Moilanen, A. and Gyllenberg, M. (1996). Minimum viable metapopulation size. The American Naturalist 147, 527-541.

Hanski, I.K., Stevens, P.C., Ihalempia, P. and Selonen, V. (2000). Home-range size, movements, and nest-site use in the Siberian flying squirrel, Pteromys volans. Journal of Mammalogy 81, 798-809.

Harestad, A.S. and Bunnell, F.L. (1979). Home range and body weight - a reevaluation. Ecology 60, 389-402.

Harris, A.C. (1966). Forestry in Western Australia., Rep. No. Bulletin 63. (Forests Department of Western Australia: Perth). References

Harris, S., Cresswell, W.J., Forde, P.G., Trewhella, W.J., Woollard, T. and Wray, S. (1990). Home-range analysis using radio-tracking data - a review of problems and techniques particularly as applied to the study of mammals. Mammal Review 20, 97-123.

Harrison, S. (1991). Local extinction in a metapopulation context: an empirical evaluation. Biological Journal of the Linnean Society 42, 73-88.

Hart, R.P., Bradshaw, S.D. and Iveson, J.B. (1986). Salmonella infections and animal conditions in the mainland and Bald Island populations of the quokka (Setonix brachyurus: Marsupialia). Journal of the Royal Society of Western Australia. 69, 7-11.

Hastie, T.J. and Tibshirani, R.J. (1990). Generalized Additive Models. Chapman & Hall, London.

Hastings, A. and Harrison, S. (1994). Metapopulation dynamics and genetics. Annual Review of Ecology and Systematics 25, 167-188.

Hatch, A.B. (1959). The effect of frequent burning on the jarrah (Eucalyptus marginata) forest soils of Western Australia. Journal of the Royal Society of Western Australia 42, 97-100.

Havel, J.J. (1975). Site-vegetation mapping in the northern jarrah forest (Darling Range). 1. Definition of site-vegetation types. Forests Department Bulletin 86.

Hayward, M.W., de Tores, P.J., Dillon, M.J. and Fox, B.J. (In review). Local population structure of a naturally-occurring metapopulation of the quokka (Setonix brachyurus Macropodidae: Marsupialia). Biological Conservation.

Heberle, G. (1997). Timber harvesting of Crown land in the south-west of Western Australia: an historical review with maps. CALMScience 2, 203-224.

Heddle, E.M., Loneragan, O.W. and Havel, J.J. (1980). Vegetation Complexes of the Darling System, Western Australia. In 'Atlas of Natural Resources, Darling References

System, Western Australia.' pp. 37-72. (Department of Conservation and Environment: Perth, Western Australia).

Heinsohn, G.E. (1968). Habitat requirements and reproductive potential of the macropod marsupial Potorous tridactylus in Tasmania. Mammalia 32, 30-43.

Henry, J.D. (1986). Red Fox: the Catlike Canine Smithsonian Institution Press, Washington, D.C.

Hesp, P.A., Wells, M.R., Ward, B. and Riches, J.R.H. (1983). Land Resource Survey of Rottnest Island: an aid to land use planning., Rep. No. 4086. (Department of Agriculture: Perth, Western Australia).

Hewitt, G. (2000). The genetic legacy of the Quaternary ice ages. Nature 405, 907-913.

Hill, M.O. (1979). TWINSPAN: A FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Cornell University, New York.

Hilton-Taylor, C., (Compiler) (2000). 2000 IUCN Red List of Threatened Species IUCN (International Union for the Conservation of Nature and Natural Resources), Gland, Switzerland.

Hobbs, R.J. (1993). Effects of landscape fragmentation on ecosystem processes in the Western Australian wheatbelt. Biological Conservation 64, 193-201.

Hodge, A. (2002). Introduced foxes threaten wildlife. In The Sun Herald, pp 21. 17/2/2002

Holdaway, R.N. and Jacomb, C. (2000). Rapid extinction of the moas (Aves: Dinornithiformes): model, test, and implications. Science 287, 2250-2254.

Hollis, C.J., Robertshaw, J.D. and Harden, R.H. (1986). Ecology of the swamp wallaby (Wallabia bicolor) in north-eastern New South Wales. I. Diet. Australian Wildlife Research 13, 355-366. References

Holsworth, W.N. (1964). Marsupial behaviour with special reference to population homeostasis in the quokkas on the west end of Rottnest Island. PhD Thesis, University of Western Australia, Nedlands, Western Australia.

Holsworth, W.N. (1967). Population dynamics of the quokka, Setonix brachyurus, on the west end of Rottnest Island, Western Australia. I. Habitat and distribution of the quokka. Australian Journal of Zoology 15, 29-46.

Hone, J. (1999). Fox control and rock-wallaby population dynamics - assumptions and hypotheses. Wildlife Research 26, 671-673.

Hope, J. (1984). The Australian Quaternary. In 'Vertebrate Zoogeography and Evolution in Australasia' (Eds. M. Archer and G. Clayton). (Hesperian Press: Carlisle, Western Australia).

Horsup, A. and Evans, M. (1993). Predation by feral cats, Felis catus, on an endangered marsupial, the bridled nailtail wallaby, Onychogalea fraenata. Australian Mammalogy 16, 83-84.

Horton, D. (1990). Reply to Flannery. Archaeology in Oceania 25, 59-60.

How, R.A. and Dell, J. (1990). Vertebrate fauna of Bold Park. The Western Australian Naturalist 18, 122-130.

How, R.A., Dell, J. and Humphreys, W.F. (1987). The ground vertebrate fauna of coastal areas between Busselton and Albany, Western Australia. Records of the Western Australian Museum. 13, 553-574.

Hulbert, I.A.R. and French, J. (2001). The accuracy of GPS for wildlife telemetry and habitat mapping. Journal of Animal Ecology 38, 869-878.

Hutchinson, B. (1972). Notes on the fauna of Yunderup in earlier days. The Western Australian Naturalist 12, 59-61. References

Iveson, J.B. and Hart, R.P. (1983). Salmonella on Rottnest Island: implications for public health and wildlife management. Journal of the Royal Society of Western Australia 66, 15-20.

Ivlev, V.S. (1961). Experimental Ecology of the Feeding of Fishes. Yale University Press, New Haven, Conneticutt.

Jackson, S.M. (2000). Home-range and den use of the , Petaurus gracilis. Wildlife Research 27, 49-60.

Jacobs, J. (1974). Quantitative measurement of food selection - a modification of the forage ratio and Ivlev's electivity index. Oecologia 14, 413-417.

Jarman, P.J. (1986). Exotic mammals, indigenous mammals and land-use. Proceedings of the Ecological Society of Australia 10, 45-61.

Jarman, P.J. (1989). Sexual dimorphism in Macropodoidae. In 'Kangaroos, Wallabies and Rat-kangaroos.' (Eds. G. Grigg, P. Jarman and I. Hume) pp. 433-447. (Surrey Beatty & Sons Pty Ltd: Sydney, Australia).

Jarman, P.J. and Coulson, G. (1989). Dynamics and adaptiveness of grouping in macropods. In 'Kangaroos, Wallabies and Rat-kangaroos.' (Eds. G.C. Grigg, P.J. Jarman and I.D. Hume) pp. 527-547. (Surrey Beatty & Sons: Chipping Norton, New South Wales).

Jenkins, C.F.H. (1974). The decline of the dalgite (Macrotis lagotis) and other wild life in the Avon Valley. The Western Australian Naturalist 12, 169-172.

Johnson, C.N. (1987). Macropod studies at Wallaby Creek. IV. Home range and movements of the red-necked wallaby. Wildlife Research 14, 125-132.

Johnson, C.N. (1989). Dispersal and philopatry in the Macropodoids. In 'Kangaroos, Wallabies and Rat-kangaroos' (Eds. G.C. Grigg, P.J. Jarman and I.D. Hume), Vol. 2 pp. 593-601. (Surrey Beatty & Sons Pty Ltd: Sydney, Australia). References

Johnson, C.N., Delean, J.S.C. and Balmford, A. (2002). Phylogeny and the selectivity of extinction in Australian marsupials. Animal Conservation 5, 135-142.

Johnson, K.A. (1980). Spatial and temporal use of habitat by the red-necked pademelon, Thylogale thetis (Marsupialia: Macropodidae). Australian Wildlife Research 7, 157-166.

Johnson, K.A., Burbidge, A.A. and McKenzie, N.L. (1989). Australian Macropodoidea: status, causes of decline and future research and management. In 'Kangaroos, Wallabies and Rat-kangaroos.' (Eds. G.C. Grigg, P.J. Jarman and I.D. Hume), Vol. 2 pp. 641-657. (Surrey Beatty & Sons Pty Ltd: Sydney, Australia).

Jones, R. (1969). Fire-stick farming. Australian Natural History 3, 224-228.

Jorgensen, E. E., Demarais, S., Sell, S.M. and Lerich, S.P. (1998). Modeling habitat suitability for small mammals in Chihuahuan desert foothills of New Mexico. Journal of Wildlife Management 62, 989-996.

Kabay, E.D. and Start, A.N. (1976). Results of the search for the potoroo in south west and south coast of Western Australia. (Department of Fisheries and Wildlife: Perth).

Kakulus, B.A. (1961). Myopathy affecting the Rottnest quokka (Setonix brachyurus) reversed by alpha-Tocopherol. Nature 191, 402-403.

Kakulus, B.A. (1963). Influence of the size of enclosure on the development of myopathy in the captive Rottnest quokka. Nature 198, 673-674.

Kanowski, J., Hopkins, M.S., Marsh, H. and Winter, J.W. (2001). Ecological correlates of folivore abundance in north Queensland rainforests. Wildlife Research 28, 1-8.

Kendall, W.L. and Pollock, K.H. (1992). The robust design in capture-recapture studies: a review and evaluation by monte carlo simulation. In 'Wildlife 2001: Populations.' (Eds. D.R. McCullough and R.H. Barrett) pp. 1163-1170. (Elsevier Science Publishers Ltd: London, England). References

Kenward, R.E. and Hodder, K.H. (1992). Ranges V - an analysis system for biological location data. Institute of Terrestrial Ecology, Furzebrook Research Station., Wareham, Dorset, UK.

Kerle, J.A. (1998). The population dynamics of a tropical possum, Trichosurus vulpecula arnhemensis Collett. Wildlife Research 25, 171-181.

King, D.R. (1993). 1080 and Australian fauna., Rep. No. 8. (Western Australian Agricultural Protection Board: Perth).

King, D.R., Oliver, A.J. and Mead, R.J. (1981). Bettongia and fluoroacetate: a role for 1080 in fauna management. Australian Wildlife Research 8, 529-536.

King, D.R. and Smith, L.A. (1985). The distribution of the European red fox (Vulpes vulpes) in Western Australia. Records of the Western Australian Museum. 12, 197-205.

Kinnear, J.E., Onus, M.L. and Bromilow, R.N. (1988). Fox control and rock-wallaby population dynamics. Australian Wildlife Research 15, 435-450.

Kinnear, J.E., Onus, M.L. and Sumner, N.R. (1998). Fox control and rock-wallaby population dynamics - II. An update. Wildlife Research 25, 81-88.

Kinnear, J.E., Sumner, N.R. and Onus, M.L. (In press). The red fox in Australia - an exotic predator turned biocontrol agent. Biological Conservation, 25.

Kirke, A. (1983). Quokka, Setonix brachyurus, at Green Range. The Western Australian Naturalist 15, 146.

Kirsch, J.A.W., Lapointe, F.-J. and Foeste, A. (1995). Resolution of portions of the kangaroo phylogeny (Marsupialia: Macropodidae) using DNA hybridization. Biological Journal of the Linnean Society 55, 309-328. References

Kitchener, D.J. (1970). Aspects of the response of the quokka to environmental stress. Bachelor of Science Honours Thesis, University of Western Australia., Nedlands, Western Australia.

Kitchener, D.J. (1972). The importance of shelter to the quokka, Setonix brachyurus (Marsupialia), on Rottnest Island. Australian Journal of Zoology 20, 281-299.

Kitchener, D.J. (1973). Notes on home range and movement in two small macropods, the potoroo (Potorous apicalis) and the quokka (Setonix brachyurus). Mammalia 37, 231-240.

Kitchener, D.J. (1981). Factors influencing selection of shelter by individual quokkas, Setonix brachyurus (Marsupialia), during hot summer days on Rottnest Island. Australian Journal of Zoology 29, 875-884.

Kitchener, D.J. (1995). Quokka (Setonix brachyurus). In 'Mammals of Australia' (Ed. R. Strahan) pp. 401-403. (Reed Books: Sydney, Australia).

Kitchener, D.J., Chapman, A. and Barron, G. (1978). Mammals of the northern Swan Coastal Plain. In 'Faunal Studies of the Northern Swan Coastal Plain.' pp. 54-82. (The Western Australian Museum for the Department of Conservation and Environment: Perth).

Kitchener, D.J. and Vicker, E. (1981). Catalogue of modern mammals in the Western Australian Museum 1895 to 1981. The Western Australian Museum, Perth.

Krebs, C.J. (1966). Demographic changes in fluctuating populations of Microtus californicus. Ecological Monographs 36, 239-273.

Krebs, C.J. (1989). Ecological Methodology. Harper Collins Inc., New York.

Krebs, C.J. and Singleton, G.R. (1993). Indices of condition for small mammals. Australian Journal of Zoology 41, 317-323. References

Krebs, C.J., Singleton, G.R. and Kenney, A.J. (1994). Six reasons why feral house mouse populations might have low recapture rates. Wildlife Research 21, 559-567.

Lagos, V.O., Contreras, L.C., Meserve, P.L., Gutierrez, J.R. and Jaksic, F.M. (1995). Effects of predation risk on space use by small mammals: a field experiment with a neotropical rodent. Oikos 74, 259-264.

Liddelow, G. (Undated). How to spot your quokka. (Forests Department of Western Australia: Como, Perth, Western Australia).

Lilley, I. (1993). Recent research in southwestern Western Australia: a summary of initial findings. Australian Archaeology 36, 34-41.

Lima, S.L. (1987). Vigilance while feeding and its relation to the risk of predation. Journal of Theoretical Biology 124, 303-316.

Lima, S.L. and Dill, L.M. (1990). Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68, 619-640.

Loukmas, J.J. and Halbrook, R.S. (2001). A test of the mink habitat suitability index model for riverine systems. Wildlife Society Bulletin 29, 821-826.

Luck, G.W. (2002). Determining habitat quality for the cooperatively breeding rufous treecreeper, Climacteris rufa. Austral Ecology 27, 229-237.

Lundelius, E.J. (1957). Additions to knowledge of the ranges of Western Australian mammals. The Western Australian Naturalist 5, 173-182.

Lundelius, E.L.J. (1963). Vertebrate remains from the Nullarbor Caves, Western Australia. Journal of the Royal Society of Western Australia. 46, 75-80.

Lundie-Jenkins, G. (1993). Ecology of the rufous hare-wallaby, Lagorchestes hirsutus Gould (Marsupialia: Macropodidae), in the Tanami Desert, Northern Territory. I. Patterns of habitat use. Wildlife Research 20, 457-476. References

Lundie-Jenkins, G., Corbett, L.K. and Phillips, C.M. (1993). Ecology of the rufous hare- wallaby, Lagorchestes hirsutus Gould (Marsupialia: Macropodidae), in the Tanami Desert, Northern Territory. III. Interactions with introduced mammal species. Wildlife Research 20, 495-511.

Lunney, D. (2001). Causes of the extinction of native mammals of the western division of New South Wales: an ecological interpretation of the nineteenth century historical record. Rangeland Journal 23, 44-70.

Lunney, D. and Leary, T. (1988). Effect of European man on fauna in the south-east of New South Wales. Australian Journal of Ecology 25, 100-116.

Lunney, D., Matthews, A. and Grigg, J. (2001). The diet of Antechinus agilis and A. swainsonii in unlogged and regenerating sites in Mumbulla State Forest, south- eastern New South Wales. Wildlife Research 28, 459-464.

MacArthur, R.H. and Wilson, E.O. (1963). An equilibrium theory of insular zoogeography. Evolution 17, 373-387.

MacArthur, R.H. and Wilson, E.O. (1967). The theory of island biogeography. Princeton University Press, Princeton, New Jersey.

Mace, G.M. and Balmford, A. (2000). Patterns and processes in contemporary mammalian extinction. In 'Priorities for the Conservation of Mammalian Diversity: Has the Panda had its Day?' (Eds. A. Entwistle and N. Dunstone) pp. 27-52. (Cambridge University Press: Cambridge, UK).

Main, A.R. (1959). Rottnest Island: the Rottnest Biological Station and recent scientific research. Rottnest Island as a location for biological studies. Journal of the Royal Society of Western Australia 42, 66-67.

Main, A.R. (1979). The fauna. In 'Environment and Science' (Ed. B.J. O'Brien) pp. 77-93. (University of Western Australia: Perth). References

Main, A.R. and Bakker, H.R. (1981). Adaptation of macropod marsupials to aridity. In 'Ecological Biogeography of Australia.' (Ed. A. Keast) pp. 1491-1520. (Dr W. Junk: The Hague -Boston-London).

Main, A.R., Shield, J.W. and Waring, H. (1959). Recent studies on marsupial ecology. In 'Biogeography and Ecology in Australia' (Eds. A. Keast, R.L. Crocker and C.S. Christian) pp. 315-331. (Dr. W. Junk: The Hague, Netherlands).

Main, A.R. and Yadav, M. (1971). Conservation of macropods in reserves in Western Australia. Biological Conservation 3, 123-133.

Mallick, S.A., Driessen, M.M. and Hocking, G.J. (2000). Demography and home range of the eastern barred bandicoot ( gunnii) in south-eastern Tasmania. Wildlife Research 27, 103-115.

Marchant, N.G., Wheeler, J.R., Rye, B.L., Bennett, E.M., Lander, N.S. and Macfarlane, T.D. (1987a). Flora of the Perth Region. Part One. Western Australian Herbarium / Department of Agriculture, Perth.

Marchant, N.G., Wheeler, J.R., Rye, B.L., Bennett, E.M., Lander, N.S. and Macfarlane, T.D. (1987b). Flora of the Perth Region. Part Two. Western Australian Herbarium / Department of Agriculture, Perth.

Marsack, P. and Campbell, G. (1990). Feeding behaviour and diet of dingoes in the Nullarbor region, Western Australia. Australian Wildlife Research 17, 349-357.

Martin, P.S. (1990). Reply to Flannery. Archaeology in Oceania 25, 58-59.

Martin, P.S. and Steadman, D.W. (1999). Prehistoric extinctions on islands and continents. In 'Extinctions in Near Time' (Ed. R.D.E. MacPhee) pp. 17-55. (Kluwer Academic/Plenum Publishers: New York).

Maxwell, S., Burbidge, A.A. and Morris, K.D. (1996). The Action Plan for Australian Monotremes and Marsupials. Wildlife Australia, Canberra. References

Maynes, G.M. (1989). Zoogeography of the Macropodoidea. In 'Kangaroos, Wallabies and Rat-kangaroos.' (Eds. G.C. Grigg, P.J. Jarman and I.D. Hume), Vol. 1 pp. 47- 66. (Surrey Beatty & Sons Pty Ltd: Sydney, New South Wales, Australia.).

McArthur, W.M. and Mulcahy, M.J. (1980). Land use in the Darling System, Western Australia. In 'Atlas of Natural Resources, Darling System, Western Australia.' pp. 77-85. (Department of Conservation and Environment: Perth, Western Australia).

McKelvey, K.S. and Pearson, D.E. (2001). Population estimation with sparse data: the role of estimators versus indices revisited. Canadian Journal of Zoology 79, 1754-1765.

McLean, I.G., Holzer, C. and Studholme, B.J.S. (1999). Teaching predator-recognition to a naive bird: implications for management. Biological Conservation 87, 123-130.

McLean, I.G., Lundie-Jenkins, G. and Jarman, P.J. (1996). Teaching an endangered mammal to recognise predators. Biological Conservation 75, 51-62.

McLean, I.G., Schmitt, N.T., Jarman, P.J., Duncan, C. and Wynne, C.D.L. (2000). Learning for life: training marsupials to recognise introduced predators. Behaviour 137, 1361-1376.

McNab, B.K. (1963). Bioenergetics and the determination of home range size. The American Naturalist 97, 133-140.

Mead, R.J., Twigg, L.E., King, D.R. and Oliver, A.J. (1985). The tolerance to fluoroacetate of geographically separated populations of the quokka (Setonix brachyurus). Australian Zoologist 21, 503-511.

Meek, P.D. and Saunders, G. (2000). Home range and movement of foxes (Vulpes vulpes) in coastal New South Wales, Australia. Wildlife Research 27, 663-668.

Merchant, J.C. and Calaby, J.H. (1981). Reproductive biology of the red-necked wallaby (Macropus rufogriseus banksianus) and Bennett's wallaby (M. r. rufogriseus) in captivity. Journal of Zoology (London) 194, 203-217. References

Merrilees, D. (1965). Quokka at Yanchep in historic time. The Western Australian Naturalist 10, 18.

Merrilees, D. (1967). Man the destroyer: late Quaternary changes in the Australian marsupial fauna. Journal of the Royal Society of Western Australia. 51, 1-24.

Merrilees, D. (1979). The prehistoric environment in Western Australia. Journal of the Royal Society of Western Australia. 62, 109-128.

Merrilees, D., Dix, W.C., Hallam, S.J., Douglas, W.H. and Berndt, R.M. (1973). Aboriginal man in southwestern Australia. Journal of the Royal Society of Western Australia. 56, 44-55.

Miller, G.H., Magee, J.W., Johnson, B.J., Fogel, M.L., Spooner, N.A., McCulloch, M.T. and Ayliffe, L.K. (1999). Pleistocene extinction of Genyornis newtoni: human impact on Australian megafauna. Science 283, 205-208.

Miller, T. and Bradshaw, S.D. (1979). Adrenocortical function in a field population of a macropodid marsupial (Setonix brachyurus, Quoy and Gaimard). Journal of Endocrinology 82, 159-170.

Milsom, T.P., Langton, S.D., Parkin, W.K., Peel, S., Bishop, J.D., Hart, J.D. and Moore, N.P. (2000). Habitat models of birds species' distribution: an aid to the management of coastal grazing marshes. Journal of Applied Ecology 37, 706-727.

Mitchell, C.D., Hennessy, K.J. and Pittock, A.B., Eds. (1994). The Greenhouse Effect - Regional Implications for Western Australia. (CSIRO Division of Atmospheric Research: Perth, Western Australia).

Molsher, R.L., Newsome, A. and Dickman, C.R. (1999). Feeding ecology and population dynamics of the feral cat (Felis catus) in relation to the availability of prey in central-eastern New South Wales. Wildlife Research 26, 593-607.

Monamy, V. and Gott, M. (2001). Practical and ethical considerations for students conducting ecological research involving wildlife. Austral Ecology 26, 293-300. References

Morris, K.D. (1992). Return of the chuditch. Landscope 8, 10-15.

Morris, K.D. (2000). The status and conservation of native rodents in Western Australia. Wildlife Research 27, 405-419.

Morton, S.R. (1990a). 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.

Morton, S.R. (1990b). Reply to Flannery. Archaeology in Oceania 25, 57-58.

Mosimann, J.E. and Martin, P.S. (1975). Simulating overkill by Paleoindians. American Science 63, 304-313.

Moss, G.L. and Croft, D.B. (1999). Body condition of the red kangaroo (Macropus rufus) in arid Australia: the effect of environmental condition, sex and reproduction. Australian Journal of Ecology 24, 97-109.

Mulcahy, M.J. (1967). Landscapes, laterites and soils in southwestern Australia. In 'Landform Studies from Australia and .' pp. 211-230.

Mulcahy, M.J., Churchward, H.M. and Dimmock, G.M. (1972). Landforms and soils on an uplifted peneplain in the Darling Range, Western Australia. Australian Journal of Soil Research 10, 1-14.

Myers, K. (1995). Rabbit, Oryctolagus cuniculus. In 'Mammals of Australia' (Ed. R. Strahan) pp. 703-705. (Reed Books: Sydney, Australia).

Naidoo, R. and Adamowicz, W.L. (2001). Effects of economic prosperity on numbers of threatened species. Conservation Biology 15, 1021-1029.

Nams, V.O. (1990). Locate II User's Guide. Pacer Computer Software, Truro, Nova Scotia.

Nelder, J.A. and Wedderburn, R.W.M. (1972). Generalized linear models. Journal of the Royal Statistical Society, Series A 135, 370-385. References

Nelson, M.D. and Buech, R.R. (1996). A test of 3 models of Kirtland's warbler habitat suitability. Wildlife Society Bulletin 24, 89-97.

Newell, G.R. (1999). Home range and habitat use by Lumholtz's tree-kangaroo (Dendrolagus lumholtzi) within a rainforest fragment in north Queensland. Wildlife Research 26, 129-145.

Newsome, A.E. (2001). The biology and ecology of the dingo. In 'A Symposium on the Dingo' (Eds. C.R. Dickman and D. Lunney) pp. 20-33. (Royal Zoological Society of New South Wales: Mosman, New South Wales).

Newsome, A.E., Parer, I. and Catling, P.C. (1989). Prolonged prey suppression by carnivores - predator-removal experiments. Oecologica 78, 458-467.

Nicholls, D.G. (1971). Daily and seasonal movements of the quokka, Setonix brachyurus (Marsupialia), on Rottnest Island. Australian Journal of Zoology 19, 215-226.

Nichols, O.G. and Nichols, F.M. (1984). The reptilian, avian and mammalian fauna of the Mount Saddleback State Forest, Western Australia. The Western Australian Naturalist 15, 179-189.

Nicholson, P.H. (1981). Fire and the Australian Aborigine - an enigma. In 'Fire and the Australian Biota' (Eds. A.M. Gill, R.H. Groves and I.R. Noble) pp. 55-76. (Australian Academy of Science: Canberra).

Niven, B.S. (1971). Mathematics of populations of the quokka, Setonix brachyurus (Marsupialia). II. A stochastic model for quokka populations. Australian Journal of Zoology 19, 393-399.

Nix, H.A. (1986). A biogeographic analysis of Australian elapid snakes. In 'Atlas of Elapid Snakes of Australia.' (Ed. R. Longmore) pp. 4-15. (Australian Government Publishing Service: Canberra).

Norbury, G.L., Sanson, G.D. and Lee, A.K. (1989). Feeding ecology of the Macropodoidea. In 'Kangaroos, Wallabies and Rat-kangaroos' (Eds. G. Grigg, P.J. References

Jarman and I. Hume) pp. 169-178. (Surrey Beatty & Sons Pty Ltd: Sydney, Australia).

Norrdahl, K. and Korpimaki, E. (1998). Does mobility or sex of voles affect risk of predation by mammalian predators? Ecology 79, 226-232.

O'Connell, J.F. (1990). Reply to Flannery. Archaeology in Oceania 25, 56-57.

Oksanen, J. and Minchin, P.R. (1997). Instability of ordination results under changes in input data order: explanations and remedies. Journal of Vegetation Science 8, 447-454.

Otis, D.L. and White, G.C. (1999). Autocorrelation of location estimates and the analysis of radiotracking data. Journal of Wildlife Management 63, 1039-1044.

Owen-Smith, N. (1993). Comparative mortality rates of male and female kudus: the costs of sexual size dimorphism. Journal of Animal Ecology 62, 428-440.

Packer, W.C. (1963). Homing behaviour in the quokka, Setonix brachyurus (Quoy and Gaimard) (Marsupialia). Journal of the Royal Society of Western Australia 46, 28- 37.

Packer, W.C. (1965). Environmental influences on daily and seasonal activity in Setonix brachyurus (Quoy and Gaimard) (Marsupialia). Animal Behaviour 13, 270-283.

Packer, W.C. (1968). Eosinophils and populations density in the marsupial Setonix. Journal of Mammalogy 49, 124-126.

Packer, W.C. (1969). Observations on the behaviour of the marsupial Setonix brachyurus (Quoy and Gaimard) in an enclosure. Journal of Mammalogy 50, 8-20.

Packer, W.C. (1977). A metabolic study of the quokka, Setonix brachyurus, in varying regimes of temperature and humidity. In 'The Biology of Marsupials.' (Eds. B. Stonehouse and D. Gilmore) pp. 77-99. (The Macmillan Press: London). References

Palomares, F., Delibes, M., Revilla, E., Calzada, J. and Fedriani, J.M. (2001). Spatial ecology of Iberian lynx and abundance of European rabbits in southwestern Spain. Wildlife Monographs 148.

Parker, N., Pascoe, A., Moller, H. and Maloney, R. (1996). Inaccuracy of a radio-tracking system for small mammals: the effect of electromagnetic interference. Journal of Zoology (London) 239, 401-406.

Pearson, D.J. (1993). Distribution, status and conservation of pythons in Western Australia. In 'Herpetology in Australia - a diverse discipline.' (Eds. D. Lunney and D. Ayres) pp. 383-395. (Royal Zoological Society of New South Wales.: Mosman, New South Wales, Australia).

Pech, R.P., Sinclair, A.R.E. and Newsome, A.E. (1995). Predation models for primary and secondary prey species. Wildlife Research 22, 55-64.

Pen, L.J. and Green, J.W. (1983). Botanical exploration and vegetational changes on Rottnest Island. Journal of the Royal Society of Western Australia. 66, 20-24.

Perry, D.H. (1971). Observations of a young naturalist in the lower Blackwood Valley in the year 1919. The Western Australian Naturalist 12, 46-47.

Pimm, S.L. (1998). Extinctions. In 'Conservation Science and Action.' (Ed. W.J. Sutherland) pp. 20-38. (Blackwell Science: Oxford).

Pollock, K.H. (1982). A capture-recapture design robust to unequal probability of capture. Journal of Wildlife Management 46, 752-757.

Pollock, K.H., Winterstein, S.R., Bunck, C.M. and Curtis, P.D. (1989a). Survival analysis in telemetry studies: the staggered entry design. Journal of Wildlife Management 53, 7-15.

Pollock, K.H., Winterstein, S.R. and Conroy, M.J. (1989b). Estimation and analysis of survival distributions for radio-tagged animals. Biometrics 45, 99-109. References

Porter, J.K. (1979). Vertebrate remains from a stratified Holocene deposit in Skull Cave, Western Australia, and a review of their significance. Journal of the Royal Society of Western Australia 61, 109-117.

Pulliam, H.R. (1973). On the advantages of flocking. Journal of Theoretical Biology 38, 419-422.

Pulliam, H.R. (1988). Sources, sinks, and population regulation. The American Naturalist 132, 652-661.

Purvis, A., Gittleman, J.L., Cowlishaw, G. and Mace, G.M. (2000). Predicting extinction risk in declining species. Proceedings of the Royal Society of London (Series B) 267, 1947-1952.

Quin, D.G., Smith, A.P., Green, S.W. and Hines, H.B. (1992). Estimating the home ranges of sugar gliders (Petaurus breviceps) (Marsupialia: ), from grid- trapping and radiotelemetry. Wildlife Research 19, 471-487.

RDAG/DWG (2000). Phytophthora dieback - you can help. (Roleystone Dieback Action Group and the Dieback Working Group: Perth).

Read, J. and Bowen, Z. (2001). Population dynamics, diet and aspects of the biology of feral cats and foxes in arid South Australia. Wildlife Research 28, 195-203.

Reading, R.P., Clark, T.W., Seebeck, J.H. and Pearce, J. (1996). Habitat suitability index model for the eastern barred bandicoot, Perameles gunni. Wildlife Research 23, 221-235.

Recher, H.F. and Lim, L. (1990). A review of current ideas of the extinction, conservation and management of Australia's terrestrial vertebrate fauna. Proceedings of the Ecological Society of Australia. 16, 287-301.

Richards, J.D. and Short, J. (1996). History of the disappearance of the native fauna from the Nullarbor Plain through the eyes of long time resident Amy Crocker. The Western Australian Naturalist 21, 89-95. References

Richards, J.D., Short, J., Prince, R.I.T., Friend, J.A. and Courtenay, J.M. (2001). The biology of banded (Lagostrophus fasciatus) and rufous (Latorchestes hirsutus) hare-wallabies (: Macropodidae) on Dorre and Bernier Islands, Western Australia. Wildlife Research 28, 311-322.

Richardson, B.J. and McDermid, E.M. (1978). A comparison of genetic relationships within the Macropodidae as determined from allozyme, cytological and immunological data. Australian Mammalogy 2, 43-51.

Ride, W.D.L. (1970). A Guide to the Native Mammals of Australia. Oxford University Press, Melbourne.

Risbey, D.A. and Calver, M.C. (1998). Can current control strategies against introduced predators endanger native mammals? In 11th Australian Vertebrate Pest Conference. Promaco Conventions Pty Ltd, Lord Forest Hotel, Bunbury, Western Australia.

Risbey, D.A., Calver, M.C. and Short, J. (1999). The impact of cats and foxes on the small vertebrate fauna of Heirisson Prong, Western Australia. I. Exploring potential impact using diet analysis. Wildlife Research 26, 621-630.

Risbey, D.A., Calver, M.C., Short, J., Bradley, J.S. and Wright, I.W. (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.

Robertshaw, J.D. and Harden, R.H. (1985). The ecology of the dingo in north-eastern New South Wales. II. Diet. Australian Wildlife Research 12, 39-50.

Robertshaw, J.D. and Harden, R.H. (1986). The ecology of the dingo in north-eastern New South Wales. IV. Prey selection by dingoes, and its effect on the major prey species, the swamp wallaby, Wallabia bicolor (Desmarest). Australian Wildlife Research 13, 141-163. References

Rodriguez, A., Crema, G. and Delibes, M. (1996). Use of non-wildlife passages across a high speed railway by terrestrial vertebrates. Journal of Applied Ecology 33, 1527-1540.

Rodriguez, J.P. (2002). Range contraction in declining North American bird populations. Ecological Applications 12, 238-248.

Roe, R. (1971). Trial excavation in a small cave, Gingin. The Western Australian Naturalist 11, 183-184.

Rooney, S.M., Wolfe, A. and Hayden, T.J. (1998). Autocorrelated data in telemetry studies: time to independence and the problem of behavioural effects. Mammal Review 28, 89-98.

Rounsevell, D.E. and Mooney, N. (1995). Thylacine (Thylacinus cynocephalus). In 'Mammals of Australia' (Ed. R. Strahan) pp. 164-165. (Reed Books: Sydney, Australia).

Russell, E.M. (1974). Recent ecological studies on Australian marsupials. Australian Mammalogy 1, 189-211.

Russell, E.M. and Rowley, I. (1998). The effects of fire on a population of red-winged fairy-wrens Malurus elegans in karri forest in southwestern Australia. Pacific Conservation Biology 4, 197-208.

Sadlier, R.M.F.S. (1959). Comparative aspects of the ecology and physiology of Rottnest and Byford populations of the quokka (Setonix brachyurus Quoy and Gaimard). Unpublished B.Sc. (Hons) thesis, University of Western Australia, Nedlands, Western Australia.

Sander, U., Short, J. and Turner, B. (1997). Social organisation and warren use of the burrowing bettong, Bettongia lesueur (Macropodoidea: ). Wildlife Research 24, 143-157. References

Saunders, G. and Harris, S. (2000). Evaluation of attractants and bait preferences of captive red foxes (Vulpes vulpes). Wildlife Research 27, 237-243.

Saunders, G., Kay, B. and McLeod, L. (1999). Caching of baits by foxes (Vulpes vulpes) on agricultural lands. Wildlife Research 26, 335-340.

Schmidt, W. (1973). Fire and fauna in the northern jarrah forest of Western Australia. The Western Australian Naturalist 12, 162-164.

Schmidt, W. and Mason, M. (1973). The effect of prescribed burning on the fauna of the jarrah forest., Rep. No. 11. (Forests Department of Western Australia.: Perth, Western Australia.).

Schodde, R. and Tidemann, S.C., Eds. (1997). The Reader's Digest Complete Book of Australian Birds, 2nd edn.

Seber, G.A.F. (1982). The Estimation of Animal Abundance and Related Parameters., 2nd edn. Charles Griffin & Co. Ltd, London, UK.

Serventy, D.L., Glauert, L., Carnaby, I.C., Loaring, W.H., Aitken, R. and Jones, A.D. (1954). The recent increase of the rarer native mammals. The Western Australian Naturalist 4, 128-141.

Sharman, G.B. (1954). The relationships of the quokka (Setonix brachyurus). The Western Australian Naturalist 4, 159-168.

Sharman, G.B. (1955a). Studies on marsupial reproduction. II. The oestrous cycle of Setonix brachyurus. Australian Journal of Zoology. 3, 44-55.

Sharman, G.B. (1955b). Studies on marsupial reproduction. III. Normal and delayed pregnancy in Setonix brachyurus. Australian Journal of Zoology 3, 56-70.

Sharman, G.B. and Maynes, G.M. (1995). Rock-wallabies. In 'Mammals of Australia' (Ed. R. Strahan) pp. 363-364. (Reed Books: Sydney, Australia). References

Shield, J.W. (1958). Reproduction of the quokka, Setonix brachyurus. Unpublished PhD Thesis, University of Western Australia, Perth.

Shield, J.W. (1959). Rottnest Island: the Rottnest Biological Station and recent scientific research. 15 - Rottnest field studies concerned with the quokka. Journal of the Royal Society of Western Australia 42, 76-78.

Shield, J.W. (1960). Gestation time for delayed birth in the quokka. Nature 185, 163-164.

Shield, J.W. (1961). The development of certain external characters in the young of the macropod marsupial Setonix brachyurus. Anatomical Record 140, 289-293.

Shield, J.W. (1962). The sex-ratio of pouch young, yearlings and adults of the macropod marsupial, Setonix brachyurus. Australian and New Zealand Journal of Obstetrics and Gynaecology 4, 161-164.

Shield, J.W. (1964). A breeding season difference in two populations of the Australian macropod marsupial (Setonix brachyurus). Journal of Mammalogy 45, 616-625.

Shield, J.W. (1968). Reproduction of the quokka, Setonix brachyurus, in captivity. Journal of Zoology, London 155, 427-444.

Shield, J.W. and Woolley, P. (1961). Age estimation by measurement of pouch young of the quokka (Setonix brachyurus). Australian Journal of Zoology 9, 14-23.

Shield, J.W. and Woolley, P. (1963). Population aspects of delayed birth in the quokka (Setonix brachyurus). Proceedings of the Zoological Society of London 141, 783- 789.

Short, J. (1982). Habitat requirements of the brush-tailed rock-wallaby, Petrogale penicillata, in New South Wales. Australian Wildlife Research 9, 239-246.

Short, J. (1998). The extinction of rat-kangaroos (Marsupialia: Potoroidae) in New South Wales, Australia. Biological Conservation 86, 365-377. References

Short, J., Bradshaw, S.D., Giles, J., Prince, R.I.T. and Wilson, G.R. (1992). Reintroduction of macropods (Marsupialia: Macropodoidea) in Australia - a review. Biological Conservation 62, 189-204.

Short, J. and Calaby, J.H. (2001). The status of Australian mammals in 1922 - collections and field notes of museum collector Charles Hoy. Australian Zoologist 31, 533- 562.

Short, J., Kinnear, J.E. and Robley, A. (2002). Surplus killing by introduced predators in Australia - evidence for ineffective anti-predator adaptations in native prey species? Biological Conservation 103, 283-301.

Short, J. and Smith, A.P. (1994). Mammal decline and recovery in Australia. Journal of Mammalogy 75, 288-297.

Short, J. and Turner, B. (1994). A test of the vegetation mosaic hypothesis - a hypothesis to explain the extinction of Australian mammal species. Conservation Biology 8, 439-449.

Short, J. and Turner, B. (2000). Reintroduction of the burrowing bettong Bettongia lesueur (Marsupialia: Potoroidae) to mainland Australia. Biological Conservation 96, 185-196.

Short, J., Turner, B., Risbey, D.A. and Carnamah, R. (1997). Control of feral cats for nature conservation. II. Population reduction by poisoning. Wildlife Research 24, 703-714.

Shortridge, G.C. (1909). An account of the geographical distribution of the marsupials and monetremes of south-west Australia, having special reference to the specimens collected during the Balston Expedition of 1904-1907. Proceedings of the Zoological Society of London. 1909, 803-848.

Shortridge, G.C. (1936). Field notes (hitherto unpublished) on Western Australian mammals south of the Tropic of Capricorn (exclusive of Marsupialia and References

Monotremata), and records of specimens collected during the Balston Expeditions (November 1904 to June 1907). Proceedings of the Zoological Society of London 1936, 743-749.

Silk, J.B. (1984). Local resource competition and the evolution of male-biased sex ratios. Journal of Theoretical Biology 108, 203-213.

Sinclair, A.R.E., Pech, R.P., Dickman, C.R., Hik, D., Mahon, P. and Newsome, A.E. (1998). Predicting effects of predation on conservation of endangered species. Conservation Biology 12, 564-575.

Sinclair, E.A. (1998). Morphological variation among populations of the quokka, Setonix brachyurus (Macropodidae: Marsupialia), in Western Australia. Australian Journal of Zoology 46, 439-449.

Sinclair, E.A. (1999). Genetic variation in two endangered marsupials, the quokka, Setonix brachyurus, and Gilbert's potoroo, Potorous gilbertii. PhD Thesis, University of Western Australia, Perth.

Sinclair, E.A. (2001). Phylogeographic variation in the quokka, Setonix brachyurus (Marsupialia: Macropodidae): implications for conservation. Animal Conservation 4, 325-333.

Sinclair, E.A., Danks, A. and Wayne, A.F. (1996). Rediscovery of Gilbert's potoroo, Potorous tridactylus, in Western Australia. Australian Mammalogy 19, 69-72.

Sinclair, E.A. and Morris, K.D. (1996). Where have all the quokkas gone? Landscope 11, 49-53.

Slater, P., Slater, P. and Slater, R. (1986). The Slater Field Guide to Australian Birds. Rigby Publishers, Sydney.

Smith, A.P. and Quin, D.G. (1996). Patterns and causes of extinction and decline in Australian conilurine rodents. Biological Conservation 77, 243-267. References

Smith, A.T. (1974). The distribution and dispersal of pikas: consequences of insular population structure. Ecology 55, 1112-1119.

Smith, V.M. (1990). The terrestrial vertebrate fauna of the Torndirrup National Park. The Western Australian Naturalist 18, 82-91.

Soderquist, T.R. and Serena, M. (1994). Dietary niche of the western quoll, Dasyurus geoffroii , in the jarrah forest of Western Australia. Australian Mammalogy 17, 133-136.

Sokal, R.R. and Rohlf, F.J. (1969). Biometry - the principles and practice of statistics in biological research. W.H.Freeman and Co., San Francisco.

Soule, M.E. (1988). Reconstructed dynamics of rapid extinctions of chaparral-requiring birds in urban habitat islands. Conservation Biology 2, 75-92.

Soule, M.E. and Gilpin, M.E. (1991). The theory of wildlife corridor capability. In 'Nature Conservation 2: The Role of Corridors' (Eds. D.A. Saunders and R.J. Hobbs) pp. 3-8. (Surrey Beatty & Sons: Sydney, Australia).

Southern, R.L. (1979). The atmosphere. In 'Environment and Science' (Ed. B.J. O'Brien) pp. 183-225. (University of Western Australia: Perth).

Southwell, C.J. and Jarman, P.J. (1987). Macropod studies at Wallaby Creek III. The effect of fire on pasture utilisation by macropodids and . Australian Wildlife Research 14, 117-124.

Spencer, P.B.S. (1990). Evidence of predation by a feral cat, Felis catus (Carnivora: Felidae) on an isolated rock-wallaby colony in tropical Queensland. Australian Mammalogy 14, 143-144.

Stanger, M., Clayton, M., Schodde, R., Wombey, J. and Mason, I. (1998). CSIRO List of Australian vertebrates: a reference with conservation status. CSIRO Publishing, Collingwood, Victoria. References

Stewart, D.W.R. (1936). Notes on marsupial damage in pine plantations. Australian Journal of Forestry 1, 41-44.

Stokes, M.K., Slade, N.A. and Blair, S.M. (2001). Influences of weather and moonlight on activity patterns of small mammals: a biogeographical perspective. Canadian Journal of Zoology 79, 966-972.

Storr, G.M. (1961). Some field aspects of nutrition in the quokka (Setonix brachyurus). Unpublished PhD Thesis, University of Western Australia, Nedlands, Western Australia.

Storr, G.M. (1963). Some factors inducing change in the vegetation of Rottnest Island. The Western Australian Naturalist 9, 15-22.

Storr, G.M. (1964a). The environment of the quokka (Setonix brachyurus) in the Darling Range, Western Australia. Journal of the Royal Society of Western Australia 47, 1-2.

Storr, G.M. (1964b). Studies on marsupial nutrition - IV Diet of the quokka, Setonix brachyurus (Quoy & Gaimard), on Rottnest Island, Western Australia. Australian Journal of Biological Science 17, 469-481.

Storr, G.M. (1965a). The avifauna of Rottnest Island, Western Australia. I. Marine birds. Emu 64, 48-60.

Storr, G.M. (1965b). Notes of Bald Island and the adjacent mainland. The Western Australian Naturalist 9, 187-196.

Storr, G.M., Green, J.W. and Churchill, D.M. (1959). Rottnest Island: the Rottnest Biological Station and recent scientific research. 13 - The vegetation of Rottnest Island. Journal of the Royal Society of Western Australia 42, 70-71.

Storr, G.M., Smith, L.A. and Johnstone, R.E. (1983). Lizards of Western Australia. II. Dragons and monitors Western Australian Museum, Perth. References

Storr, G.M., Smith, L.A. and Johnstone, R.E. (1990). Lizards of Western Australia. III. Geckos and pygopods. Western Australian Museum, Perth.

Storr, G.M., Smith, L.A. and Johnstone, R.E. (1999). Lizards of Western Australia. I. ., Revised edition edn. Western Australian Museum, Perth.

Strahan, R., Ed. (1995). Mammals of Australia. (Reed Books: Sydney, Australia).

Strelein, G.J. (1988). Site Classification in the Southern Jarrah Forest of Western Australia Department of Conservation and Land Management, Perth, Western Australia.

Sutherst, R.W. and Maywald, G.F. (1985). A computerised system for matching climates in ecology. Agricultural Ecosystems and Environment 13, 281-299.

Swihart, R.K. and Slade, N.A. (1985). Testing for independence of observations in animal movements. Ecology 66, 1176-1184.

Tausch, R.J., Charlet, D.A., Weixelman, D.A. and Zamudio, D.C. (1995). Patterns of ordination and classification instability resulting from changes in input data order. Journal of Vegetation Science 6, 897-902.

Thackway, R.M. and Cresswell, I.D. (1995). An Interim Biogeographic Regionalisation for Australia: a framework for setting priorities in the national reserves system cooperative program. Australian Nature Conservation Agency, Canberra.

Thomas, O. (1888). Catalogue of the Marsupialia and Monotremata in the collection of the British Museum (Natural History). British Museum, London.

Thomas, O. (1906). On mammals collected in south-west Australia for Mr. W. E. Balston., 468-478.

Thompson, J.A. and Fleming, P.J.S. (1994). Evaluation of the efficacy of 1080 poisoning of red foxes using visitation to non-toxic baits as an index of fox abundance. Wildlife Research 21, 27-39. References

Thomson, P.C. and Algar, D. (2000). The uptake of dried meat baits by foxes and investigations of baiting rates in Western Australia. Wildlife Research 27, 451- 456.

Tilman, D., May, R.M., Lehman, C.L. and Nowak, M.A. (1994). Habitat destruction and the extinction debt. Nature 371, 65-66.

Trivers, R.L. and Willard, D.E. (1973). Natural selection of parental ability to vary the sex ratio of offspring. Science 179, 90-92.

Troughton, E. (1967). Furred Animals of Australia. Angus and Robertson, Sydney.

Troy, S. and Coulson, G. (1993). Home range of the swamp wallaby, Wallabia bicolor. Wildlife Research 20, 571-577.

Troy, S.K., Coulson, G. and Middleton, D. (1992). A comparison of radiotracking and line transect techniques to determine habitat preferences in the swamp wallaby, Wallabia bicolor, in south-eastern Australia. In 'Wildlife Telemetry - Remote Monitoring and Tracking of Animals.' (Eds. I.G. Priede and S.M. Swift) pp. 651- 660. (Ellis Horwood: Chichester, England).

Turney, C.S.M., Bird, M.I., Fifield, L.K., Roberts, R.G., Smith, M., Dortch, C.E., Grun, R., Lawson, E., Ayliffe, L.K., Miller, G.H., Dortch, J. and Cresswell, R.G. (2001). Early human occupation at Devil's Lair, southwestern Australia 50,000 years ago. Quaternary Research 55, 3-13.

Twigg, L.E., Eldridge, S.R., Edwards, G.P., Shakeshaft, B.J., dePreu, N.D. and Adams, N. (2000). The longevity and efficacy of 1080 meat baits used for dingo control in central Australia. Wildlife Research 27, 473-481.

Twigg, L.E. and King, D.R. (1991). The impact of fluoroacetate-bearing vegetation on native Australian fauna: a review. Oikos 61, 412-430.

Underwood, R.J. and Christensen, P.E.S. (1981). Forest fire management in Western Australia., Rep. No. 1. (Western Australian Forests Department: Perth). References

Vandermeer, J. and Carvajal, R. (2001). Metapopulation dynamics and the quality of the matrix. The American Naturalist 158, 211-220.

Vernes, K., Dennis, A. and Winter, J. (2001). Mammalian diet and broad hunting strategy of the dingo (Canis familiaris dingo) in the wet tropical rain forests of northeastern Australia. Biotropica 33, 339-345.

Vickers-Rich, P. and Rich, T.H. (1993). Wildlife of Gondwana., Melbourne.

Viggers, K.L., Lindenmayer, D.B., Cunningham, R.B. and Donnelly, C.F. (1998). Estimating body condition in the mountain brushtail possum, Trichosurus caninus. Wildlife Research 25, 499-509.

Wahlquist, A. (2002). El Nino's 27-year curse worsens Perth's big dry. In The Weekend Australia, pp 20. 18/19 May 2002

Wallace, W.R. (1966). Fire in the jarrah forest environment. Journal of the Royal Society of Western Australia. 49, 33-44.

Ward, D.J. and Sneeuwjagt, R. (1999). Aboriginal fire: its relevance to present day management of the jarrah forest of south-western Australia. In Proceedings of 'Bushfire 99' Conference, pp. 54-55, Albury, New South Wales, Australia.

Waring, H. (1956). Marsupial studies in Western Australia. Australian Journal of Science 18, 66-73.

Waring, H. (1959). Rottnest Island: the Rottnest Biological Station and recent scientific research. Introduction. Journal of the Royal Society of Western Australia 42, 65- 66.

White, G.C., Anderson, D.R., Burnham, K.P. and Otis, D.L. (1982). Capture-recapture and removal methods for sampling closed populations. Los Alamos National Laboratory, Los Alamos, New Mexico. References

White, G.C. and Burnham, K.P. (1999). Program MARK: survival estimation from populations of marked animals. Bird Study 46 (supplement), S120-139.

White, G.C. and Garrott, R.A. (1990). Analysis of Wildlife Radio-tracking Data. Academic Press, San Diego.

White, S.R. (1952). The occurrence of the quokka in the south-west. The Western Australian Naturalist 3, 101-103.

Whitehouse, S.J.O. (1977). The diet of the dingo in Western Australia. Australian Wildlife Research 4, 145-150.

Whittell, H.M.M. (1954). John Gilbert's notebook on marsupials. The Western Australian Naturalist 4, 105-114.

Wilson, B.A. and Friend, G.R. (1999). Responses of Australian mammals to disturbance: a review. Australian Mammalogy 21, 87-105.

Woinarski, J.C.Z., Milne, D.J. and Wanganeen, G. (2001). Changes in mammal populations in relatively intact landscapes of Kakadu National Park, Northern Territory, Australia. Austral Ecology 26, 360-370.

Worton, B.J. (1989). Kernel methods for estimating the utilization distribution in home- range studies. Ecology 70, 164-168.

Wright, R. (1990). Reply to Flannery. Archaeology in Oceania 25, 56.

Yee, T.W. and Mitchell, N.D. (1991). Generalized additive models in plant ecology. Journal of Vegetation Science 2, 587-602.

Zar, J.H. (1996). Biostatistical Analysis, 3rd edn. Prentice-Hall Inc., New Jersey, USA.

Appendices

Appendices

Appendix A – Trapping data

Trapping data for all species captured at each site trapped during this study. New refers only to new animals while all refers to new captures plus recaptures. Trap nights are the same as shown in Chapter 4.

Common name Scientific name Chandler Hadfield Hoffman Holyoake Kesners Rosella Wild Pig Victor Road New All New All New All New All New All New All New All New All

Mammals Mardo Antechinus flavipes – females 12 21 3 3 7 9 6 8 2 2 - males 4 7 9 14 22 28 3 4 6 10 Chuditch / Dasyurus geoffroii western quoll – females 3 5 - males 6 15 1 1 7 12 Cat b Felis catus – females 1 2 1 1 1 1 1 1 - males 1 2 1 1 1 1 1 1 Southern brown Isoodon obesulus bandicoot – females 24 199 8 44 1 1 2 18 30 223 1 2 8 29 - males 58 516 34 151 2 3 17 64 47 328 1 1 8 29 European rabbit Oryctolagus b cuniculus – females 2 2 1 1 1 Black rat b Rattus rattus – females 15 33 11 34 59 131 18 39 14 30 - males 11 29 13 35 72 146 28 63 23 70 Quokka a Setonix brachyurus – females 4 6 17 61 15 20 8 47 - males 4 14 15 40 15 34 1 2 11 56 Pig b Sus scrofa 1 Echidna Tachyglossus 2 2 2

Common name Scientific name Chandler Hadfield Hoffman Holyoake Kesners Rosella Wild Pig Victor Road New All New All New All New All New All New All New All New All aculeata Common Trichosurus 12 102 brushtail vulpecula 4 12 4 10 3 15 5 7 possum – females - males 10 53 13 59 1 1 9 47 13 71 20 157

Lizards Egernia luctuosa 1 1 2 2 1 1 3 4 King’s skink Egernia kingii 3 11 38 64 3 3 20 39 4 9 Bobtail skink Tiliqua rugosa 37 87 7 8 2 2 1 1 33 50 69 127 9 10 Gould’s goanna Varanus gouldii 2 2 4 4 1 1 1 1 2 3 Racecourse V. rosenbergii 1 1 goanna V. tristis * 2 2

Amphibians Crinia georgiana 1 1

Snakes Western tiger Notechis ater 2 snake

Birds Pacific black Anas superciliosa 1 1 duck Australian Corvus coronoides 4 5 2 27 2 2 raven Australian Gymnorhina tibicen 2 magpie Red-winged Malurus elegans 2 fairy-wren Grey currawong Strepera versicolor 2 1 Silvereye Zosterops lateralis 5 * potential southerly range extension according to Storr et al. (1999) and Cogger (1992). a includes pouch young and juveniles Appendix B – Home range overlap of quokkas at each site.

Figure 1. Home range overlap of quokkas at the Chandler site. Male home ranges are shown in blue and females in pink. Different colours represent different habitat types.

Figure 2. Home range overlap of quokkas at the Hadfield site. Male home ranges are shown in blue and females in pink. Different colours represent different habitat types.

Figure 3. Home range overlap of quokkas at the Kesners site. Male home ranges are shown in brown and females in purple. For ease of interpretation habitat types are not shown in this figure but the swamp generally runs through the middle of each home range.

Figure 4. Seasonal home range overlap of the male quokka at the Rosella Road site. Different colours represent different habitat types.

Appendix C - Habitat units and their descriptive variables.

The habitat units are in an order of time since fire for Agonis swamp shrublands and then alphabetical order for the remainder. The

absence of indicator species (or those species which characterise a habitat unit) within a habitat unit is shown (-). The density of vegetation at

eight heights is shown. For other descriptive variables, relevant means are shown with standard error in brackets.

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps Freshly burnt Agonis linearifolia 5.0 m 3.3 ± 0.3 20.3 ± 6.3 ± 0.6 ± 1.2 ± 0.1 2.0 ± 1.8 ± 0.0 ± 0.0 Agonis swamp Lepidosperma 2.4 0.7 0.4 0.6 0.5 3.0 m shrubland tetraquetrum 2.0 m Astartea fascicularis 1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

5 – 9 year old A. linearifolia 5.0 m 5.3 ± 0.3 31.4 ± 0.4 ± 0.8 ± 0.9 ± 0.1 0.6 ± 1.5 ± 0.0 ± 0.0 Agonis swamp L. tetraquetrum 3.0 m 3.2 0.2 0.2 0.3 0.4 shrubland G. decomposita 2.0 m T. paniculata 1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

10 – 14 year old A. linearifolia 5.0 m 5.7 ± 0.4 33.8 ± 0.4 ± 0.5 ± 1.0 ± 0.2 0.3 ± 1.1 ± 0.0 ± 0.0

Agonis swamp L. tetraquetrum 3.0 m 4.1 0.2 0.1 0.2 0.4 shrubland Hypocalymma 2.0 m

cordifolium 1.5 m

Xyris lacera 1.0 m

G. decomposita 0.5 m T. pauciflora 0.3 m 0.1 m

0 1 2 3 4 5 6

15 – 19 year old A. linearifolia 5.0 m 5.1 ± 0.3 40.1 ± 0.5 ± 0.9 ± 0.8 ± 0.2 0.3 ± 1.6 ± 0.0 ± 0.0

Agonis swamp L. tetraquetrum 3.0 m 2.8 0.2 0.2 0.2 0.6 shrubland H. cordifolium 2.0 m X. lacera 1.5 m G. decomposita 1.0 m T. pauciflora 0.5 m 0.3 m

0.1 m

0 1 2 3 4 5 6

20 – 24 year old A. linearifolia 5.0 m 4.4 ± 0.3 64.2 ± 0.5 ± 1.0 ± 0.8 ± 0.1 0.2 ± 1.0 ± 0.0 ± 0.0

Agonis swamp L. tetraquetrum 3.0 m 6.3 0.2 0.2 0.2 0.5 shrubland H. cordifolium 2.0 m

X. lacera 1.5 m

G. decomposita 1.0 m

T. pauciflora 0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

> 25 year old A. linearifolia 5.0 m 4.1 ± 0.4 81.0 ± 0.1 ± 1.0 ± 0.7 ± 0.2 0.1 ± 0.4 ± 0.0 ± 0.0 Agonis swamp L. tetraquetrum 3.0 m 8.0 0.1 0.4 0.1 0.2 shrubland A. fascicularis 2.0 m H. cordifolium 1.5 m G. decomposita 1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Allocasuarina Allocasuarina fraseriana 5.0 m 6.3 ± 1.1 29.8 ± 0.8 ± 1.5 ± 0.0 ± 0.0 2.5 ± 0.0 ± 0.8 ± 0.3 forest 3.0 m 3.0 0.8 1.2 2.5 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Blackberry thicket Rubus affinus 5.0 m 6.3 ± 2.3 10.2 ± 0.0 ± 0.0 ± 1.0 ± 0.6 0.0 ± 0.0 ± 0.0 ± 0.0

3.0 m 1.9 0.0 0.0 0.0 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Blackbutt open Eucalyptus patens 5.0 m 5.8 ± 0.4 52.0 ± 0.6 ± 2.0 ± 0.5 ± 0.2 0.4 ± 1.5 ± 0.1 ± 0.1 forest E. megacarpa (-) 3.0 m 6.2 0.3 0.4 0.3 0.6

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Bullich-Blackbutt E. megacarpa 5.0 m 7.9 ± 0.3 22.5 ± 1.2 ± 2.3 ± 0.0 ± 0.0 1.2 ± 0.0 ± 0.2 ± 0.1 open forest E. patens 3.0 m 2.2 0.4 0.9 0.6 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Bullich swamp E. megacarpa 5.0 m 8.0 ± 0.6 34.4 ± 0.3 ± 1.4 ± 0.8 ±0.3 0.0 ± 1.8 ± 0.0 ± 0.0 forest A. linearifolia 3.0 m 3.7 0.2 0.6 0.0 1.3 H. cordifolium 2.0 m

Boronia molloyiae 1.5 m

L. tetraquetrum 1.0 m T. pauciflora 0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Cleared area Occasional overhanging 5.0 m 5.0 ± 1.2 0.1 ± 0.0 ± 0.0 ± 0.0 ± 0.0 6.0 ± 0.0 ± 0.0 ± 0.0 eucalypts 3.0 m 0.1 0.0 0.0 0.0 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Coastal heath Spinifex hirsutus 5.4 ± 0.6 19.5 ± 0.0 ± 0.1 ± 0.2 ± 0.1 0.3 ± 0.0 ± 0.0 ± 0.0 5.0 m Loxocarya flexuosa 3.0 m 2.8 0.0 0.1 0.3 0.0 Acacia littoralis 2.0 m Agonis flexuosa 1.5 m Lomandra seracea 1.0 m Melaleuca subacerosa 0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Coastal low open E. marginata 5.0 m 7.8 ± 1.1 21.4 ± 0.0 ± 0.0 ± 0.2 ± 0.2 0.0 ± 0.0 ± 0.0 ± 0.0 forest Hakea hypericoides 3.0 m 3.6 0.0 0.0 0.0 0.0

Hypocalymna 2.0 m

angustatum 1.5 m

Acacia pentadenia 1.0 m Agonis flexuosa 0.5 m Banksia grandis 0.3 m 0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps Coastal shrubland Acacia littoralis 5.0 m 5.6 ± 0.6 54.7 ± 0.0 ± 0.2 ± 0.9 ± 0.4 0.0 ± 0.9 ± 0.0 ± 0.0 Agonis flexuosa 3.0 m 9.8 0.0 0.2 0.0 0.7

Adenanthos obovata 2.0 m L. angustatum 1.5 m

Boronia alata 1.0 m 0.5 m

0.3 m 0.1 m

0 1 2 3 4 5 6

Dry heath Leptospermum 5.0 m 4.1 ± 0.6 9.9 ± 0.0 ± 0.4 ± 1.2 ± 0.3 2.8 ± 3.2 ± 0.0 ± 0.0 erubescens 1.6 0.0 0.3 1.0 0.9 3.0 m

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Flooded gum E. rudis 3.7 ± 0.6 9.8 ± 0.0 ± 0.1 ± 1.1 ± 0.3 4.4 ± 0.9 ± 0.0 ± 0.0 5.0 m forest 2.6 0.0 0.1 1.5 0.6 3.0 m

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Introduced Introduced grasses 5.0 m 5.3 ± 2.6 2.9 ± 0.0 ± 0.0 ± 0.3 ± 0.3 0.0 ± 0.0 ± 0.0 ± 0.0 pasture 3.0 m 1.4 0.0 0.0 0.0 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Jarrah-Marri open E. marginata 5.0 m 8.8 ± 0.4 23.6 ± 0.4 ± 2.8 ± 0.0 ± 0.0 1.4 ± 0.0 ± 1.2 ± 0.3 forest Corymbia calophylla 3.0 m 3.2 0.1 0.5 0.4 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Karri forest E. diversicolor 5.0 m 6.7 ± 0.4 72.3 ± 0.2 ± 2.3 ± 0.0 ± 0.0 0.5 ± 0.0 ± 0.4 ± 0.2 L. gladiatum 3.0 m 6.8 0.1 0.5 0.3 0.0 Trymalium floribundum 2.0 m Acacia pentadenia 1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Lepidosperma – L. tetraquentrum 5.0 m 6.3 ± 0.7 45.2 ± 0.7 ± 1.0 ± 1.0 ± 0.6 0.0 ± 0.0 ± 0.0 ± 0.0

Hypocalymna H. cordifolium 3.0 m 1.2 0.7 0.0 0.0 0.0 swamp Agonis linearifolia (-) 2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Paperbark swamp Melaleuca rhaphiophylla 5.0 m 6.1 ± 0.5 51.2 ± 0.3 ± 0.4 ± 0.9 ± 0.2 0.7 ± 1.2 ± 0.0 ± 0.0 A. linearifolia 3.0 m 8.2 0.2 0.2 0.4 0.6 T. paniculata 2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Peppermint forest A. flexuosa 5.0 m 5.3 ± 0.3 39.5 ± 0.1 ± 0.8 ± 0.3 ± 0.2 0.0 ± 0.8 ± 0.0 ± 0.0 L. tetraquentrum 5.8 0.1 0.3 0.0 0.5 3.0 m

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Pine plantation Pinus species 5.0 m 6.3 ± 2.2 35.2 ± 0.0 ± 0.0 ± 0.0 ± 0.0 3.8 ± 0.0 ± 2.5 ± 1.2

3.0 m 9.9 0.0 0.0 1.2 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Revegetation – E. maculata 5.0 m 6.3 ± 0.7 33.3 ± 0.3 ± 0.7 ± 0.0 ± 0.0 0.0 ± 0.0 ± 0.0 ± 0.0 dense Melaleuca incana 3.0 m 4.6 0.3 0.3 0.0 0.0 Acacia celastrifolia 2.0 m Bossiaea aquifolium 1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Revegetation – E. maculata 5.0 m 5.7 ± 0.8 21.4 ± 0.2 ± 0.8 ± 0.0 ± 0.0 3.3 ± 0.0 ± 1.2 ± 0.8 sparse M. incana 3.0 m 5.4 0.2 0.7 0.4 0.0 A. celastrifolia 2.0 m B. aquifolium 1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

River forest Grevillia diversifolia 5.0 m 5.7 – 1.0 46.4 ± 0.6 ± 1.4 ± 1.7 ± 0.2 0.0 ± 3.0 ± 0.0 ± 0.0 Banksia littoralis 3.0 m 5.2 0.2 0.4 0.0 0.7

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Soapbush swamp Trymalium floribundum 5.0 m 4.5 ± 0.6 74.5 ± 0.3 ± 0.8 ± 0.8 ± 0.3 0.0 ± 0.0 ± 0.0 ± 0.0

3.0 m 13.6 0.3 0.3 0.0 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Southern Trymalium floribundum 5.0 m 6.0 ± 0.4 70.8 ± 0.0 ± 0.9 ± 0.3 ± 0.2 0.3 ± 0.3 ± 0.1 ± 0.1 soapbush swamp Acacia penaadenia 3.0 m 9.1 0.0 0.4 0.3 0.2 Chorilaena quercifolia 2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps Swishbush swamp Viminaria juncea 5.0 ± 2.0 9.2 ± 0.5 ± 0.0 ± 2.0 ± 0.0 3.0 ± 5.0 ± 0.0 ± 0.0 5.0 m 1.5 0.5 0.0 0.0 1.0 3.0 m

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Ti-tree swamp Leptospermum firmum 5.8 ± 1.0 44.0 ± 2.0 ± 0.8 ± 1.0 ± 0.4 0.0 ± 1.3 ± 0.0 ± 0.0 5.0 m 6.3 1.3 0.3 0.0 0.8 3.0 m

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Wandoo E. wandoo 5.0 m 3.0 ± 0.0 14.0 ± 0.0 ± 0.8 ± 0.0 ± 0.0 0.0 ± 0.0 ± 0.0 ± 0.0 woodland Bossiaea eriocarpa 3.0 m 1.2 0.0 0.3 0.0 0.0

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Habitat unit Dominant species and Vegetation density Species Litter # of # of Ground Bare Open # of cut indicators depth stags logs moisture earth water stumps

Wattie swamp Agonis juniperiana 5.0 m 2.0 ± 0.0 52.3 ± 0.0 ± 0.0 ± 2.0 ± 0.0 4.0 ± 5.5 ± 0.0 ± 0.0

3.0 m 6.3 0.0 0.0 0.0 0.5

2.0 m

1.5 m

1.0 m

0.5 m

0.3 m

0.1 m

0 1 2 3 4 5 6

Appendices

Appendix D - Factor scores for habitat descriptions

Table 1. Factor scores for the structural factors and the variance they explain.

Variable Factor 1 Factor 2 Factor 3 Factor 4 Density at 10 cm 0.870 0.080 -0.023 0.097 Density at 30 cm 0.934 0.025 0.175 0.043 Density at 50 cm 0.772 -0.035 0.475 -0.010 Density at 1 m 0.333 0.016 0.856 0.061 Density at 1.5 m 0.046 -0.095 0.877 -0.116 Density at 2 m -0.024 0.445 0.427 -0.557 Density at 3 m 0.076 0.938 -0.110 0.017 Density at 5 m 0.090 0.068 0.037 0.943 Variance explained 37.5% 20.2% 14.9% 10.3%

Table 2. Factor scores for the floristic factors and the variance they explain.

Variable Factor Factor Factor Factor Factor 1 2 3 4 5 Agonis linearifolia 0.010 0.800 0.068 0.010 -0.046 Aotus cordifolia -0.053 -0.075 -0.095 0.653 -0.187 Astartea fascicularis 0.877 0.037 -0.009 -0.047 -0.011 Boronia molloyiae 0.209 -0.419 0.031 -0.102 0.559 Dampiera hederacea 0.307 0.290 0.629 -0.203 0.041 Hypocalymna cordifolium -0.160 0.056 -0.133 -0.051 0.720 Lepidosperma tetraquentrum -0.493 0.392 0.114 -0.247 0.360 Thomasia paniculata -0.237 -0.063 0.719 -0.081 -0.243 Thomasia pauciflora 0.021 0.117 -0.496 -0.562 -0.313 Xyris lacera 0.135 0.425 -0.245 0.639 -0.020 Variance explained 15.0% 13.7% 12.6% 10.7% 10.0% Appendices

Table 3. Factor scores for the other habitat factors and the variance they explain.

Variable Factor 1 Factor 2 Factor 3 Species richness 0.066 -0.858 -0.016 Leaf litter -0.719 0.060 0.301 Ground moisture 0.343 0.194 0.341 Stags 0.651 0.364 -0.034 Logs -0.075 -0.083 0.900 Open water 0.160 0.612 -0.010 Bare earth 0.678 -0.022 0.242 Variance explained 24.9% 15.8% 15.3%