Spatially Explicit Cost-Effective Actions for Biodiversity Threat Abatement

Home , Acacia grandifolia, Baloghia marmorata, Collared delma, Cossinia, Eucalyptus beaniana, Eucalyptus conglomerata

Spatially explicit cost-effective actions for biodiversity threat abatement

Nancy Anne Auerbach

BA, MBSc

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2015

School of Biological Sciences

Abstract

Biodiversity decline is indisputable, and rates of future decline depend on whether threats to species persistence are abated. However, current resources for threatened species management are less than required to stop further decline. Management that abates many threats to many species is necessary, yet decisions about how to do this under resource constraints are inherently complex. My thesis incorporates systematic conservation planning and cost-effectiveness analysis in a decision-support framework for prioritising spatially-explicit management actions for many species across a region. By prioritising action where it is expected to provide the greatest benefit to the most species at least cost, my research advances the thinking on decision support, and contributes to the effort to reduce biodiversity decline.

Using information on threats to species that was compiled by the Queensland, Australia government, my research develops a decision-support process for managing threats to threatened species in a bio-diverse regional-scale management area, the Burnett-Mary Natural Resource Management Region. In my thesis, predicted distributions for 65 threatened species are modelled on co-occurring presence-only species locations and ecologically-meaningful environmental data. Three threats are addressed: invasive red fox predation; too frequent and intense fire; and habitat degradation from overgrazing. Indirect threat maps are made by combining predicted distribution models of species vulnerable to specific threats and are used to identify locations where threat- abating actions are most likely to provide benefit to species. Management action costs are estimated for fox control, on the basis of a roadside baiting strategy; proactive fire management, on the basis of vegetation type and proximity to human structures; and stewardship agreements to reduce grazing, on the basis of foregone agriculture profit. Spatially combining the costs and benefits of threat-abating actions leads to the ability to prioritise cost-effective locations for actions. Within the context of prioritising regional threat management actions, my research examines four topics.

Firstly, an understanding of where species are affected by threats determines where to direct management. Indirect threat maps are made by combining threatened species distribution maps, but are sensitive to map scale in guiding spatially-explicit threat management efforts (Chapter 2). Using fine-scale predicted species distribution models to derive indirect threat maps, instead of general range maps, may lead to better regional management decisions.

Funding limitations for threat management require that choices be made when not everything can be done. Transparent and rigorous decision support is provided by prioritising management ranked on the most, to least, cost-effective actions and locations, and is made accessible when calculated with readily-available spreadsheet and mapping software (Chapter 3). Cost-effective management ii

priorities are determined by combining indirect threat maps that locate which actions benefit species with maps quantifying the cost of action. Priorities depend on how the benefits and costs co-vary across a landscape.

The costs and/or benefits of one management action may depend upon those of another action, which will affect the ranking of conservation actions by their cost-effectiveness (Chapter 4). If actions are dependent and species are secured when only one threat is managed, managing for multiple threats results in poor allocation of resources. Alternatively, focusing on individual threats potentially results in inadequate management and failure of conservation outcomes if species only benefit when all the threats are managed. Considering when management action success depends upon which other actions are undertaken will better protect species from threats and guide spending of limited funding.

Lastly, providing an alternative to basic decision support, more complex methods for conservation planning can be used to meet more comprehensive management goals, but spatial priorities are likely to differ (Chapter 5). When the goal includes explicitly managing for all species and threats, priority locations for actions are different than if prioritising management in species-rich areas. Trade-offs between threatened species management plans are inevitable, but examining the implications of different strategies may lead to better decisions.

My research provides four contributions to conservation research by focusing on decision support for threat management that reduces biodiversity decline. 1) Indirect threat maps, created by combining predicted distribution models of threatened species, indicate where action is needed for managing threats at the regional scale. 2) Decisions about where to manage threats are made by prioritising cost-effective actions and can be determined using accessible and commonly-used software. 3) Failing to consider that the success of management actions is likely to depend on what other actions have been undertaken will result in ineffective spending of limited funding or insufficient management of threats. 4) Both simpler and more complex conservation planning methods assist in choosing where to manage, with which actions, at least cost, but spatial priorities will differ. In summary, my research shows that management priorities can be selected by combining threatened species distributions with the costs of abating threats. Transparently chosen management actions that are efficient in using limited resources may better lead to abating threats to biodiversity.

iii

Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my research higher degree candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the General Award Rules of The University of Queensland, immediately made available for research and study in accordance with the Copyright Act 1968.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis

Publications during candidature

Peer reviewed papers

Auerbach, N. A., A. I. T. Tulloch, and H. P. Possingham 2014. Informed actions: Where to cost- effectively manage multiple threats to species to maximize return on investment. Ecological Applications 24:1357-1373.

Tulloch, V. J. D., A. I. T. Tulloch, P. Visconti, B. S. Halpern, J. E. M. Watson, M. C. Evans, N. A. Auerbach, M. Barnes, M. Beger, I. Chadès, S. Giakoumi, E. McDonald-Madden, N. J. Murray, J. Ringma, and H. P. Possingham 2015. Why do we map threats? Linking threat mapping with actions to make better conservation decisions. Frontiers in Ecology and the Environment doi:10.1890/140022.

Conference abstracts

Auerbach, N., M. C. Evans, and H. P. Possingham. 2012a. A comparison of regional spatial patterns of threats to species for conservation management guidance. Paper presentation at World Conference on Natural Resource Modeling. 9-12 July 2012, University of Queensland, Brisbane, Australia.

Auerbach, N., T. Holmes, J. E. M. Watson, L. N. Joseph, J. Szabo, and H. Possingham. 2012b. The state(s) of threatened species prioritisation in Australia. Paper presentation at Society for Conservation Biology Oceania Section Regional Conference. 21-23 September 2012, Charles Darwin University, Darwin, Australia.

Auerbach, N., A. Tulloch, and H. Possingham. 2013a. Cost-effectively managing threats to species. Poster Presentation at Student Conference on Conservation Biology. 21-31 January 2013, University of Queensland, St Lucia, Australia.

Auerbach, N., A. Tulloch, and H. Possingham. 2013b. Where should I do what to cost-effectively manage threats to species? Paper presentation at Connecting Systems, Disciplines, and Stakeholders: 26th International Congress for Conservation Biology. 21-25 July 2013, Baltimore, MD, USA.

Auerbach, N., A. Tulloch, K. A. Wilson, J. R. Rhodes, and H. Possingham. 2014. Evaluating dependencies in threat action prioritization. Paper presentation at Society for Conservation Biology Fiji 2014 Conference. 7-8 July 2014, Suva, Fiji.

Other articles

Mills, M., and N. Auerbach. 2014. Tools for natural resource management and monitoring. Decision Point 80:14.

Publications included in this thesis

– incorporated as Chapter 3.

Contributor Statement of Contribution Conceived the ideas for the paper (45%) Auerbach, N.A. (Candidate) Conducted the analyses (100%) Wrote the paper (80%) Conceived the ideas for the paper (10%) Tulloch, A.I.T. Wrote and edited paper (10%) Conceived the ideas for the paper (45%) Possingham, H.P. Wrote and edited paper (10%)

Contributions by others to the thesis

My PhD thesis is composed of four manuscripts with the following status: Chapter 2 is under revision. Chapter 3 is published. Chapter 4 is in the final preparation for submission. Chapter 5 will be submitted in October 2014. I use the plural first-person pronoun ―’we’ in Chapters 2-5 as per their publication format. In Chapters 1 and 6, I use the singular first-person pronoun – ‘I’ as I am describing my own views of the research as a whole.

Chapter 1

This chapter was written by the Candidate with the editorial input of Hugh Possingham, Kerrie Wilson and Ayesha Tulloch.

Chapter 2

Auerbach, N. A., M. C. Evans, and H. P. Possingham. In prep-b. All threats are not equal: Coarse and fine-scaled indirect threat maps compared.

This chapter is taken from a manuscript by the Candidate, Megan Evans, and Hugh Possingham (Auerbach et al. In prep-b). The ideas for the paper were conceived by all the authors. The Candidate conducted the analyses. The Candidate wrote the paper with the editorial input of Megan Evans and Hugh Possingham.

Chapter 3

This chapter is taken from a manuscript by the Candidate, Ayesha Tulloch, and Hugh Possingham (Auerbach et al. 2014). The ideas for the paper were conceived by the Candidate and Hugh Possingham. The Candidate conducted the analyses. The Candidate wrote the paper with the editorial input of Ayesha Tulloch and Hugh Possingham.

vii

Chapter 4

Auerbach, N. A., K. A. Wilson, A. I. T. Tulloch, J. R. Rhodes, and H. P. Possingham. In review. Accounting for dependencies in threat management can alter conservation priorities. Submitted to Conservation Biology, revise and resubmit expected January 2015.

This chapter is taken from a manuscript by the Candidate, Kerrie Wilson, Ayesha Tulloch, Jonathan Rhodes, and Hugh Possingham (Auerbach et al. In review). The ideas for the paper were conceived by all the authors. The Candidate conducted the analyses. Jeffrey Hanson wrote a key piece of the R coding for the analysis. The Candidate wrote the paper with the editorial input of all the authors.

Chapter 5

Auerbach, N., A. Tulloch, and H. Possingham. In prep-a. Accounting for complementarity and action dependencies in spatial threat-management prioritisation.

This chapter is taken from a manuscript by the Candidate, Ayesha Tulloch, and Hugh Possingham (Auerbach et al. In prep-a). The ideas for the paper were conceived by all the authors; the Candidate conducted the analyses. The Candidate wrote the paper with the editorial input of Ayesha Tulloch and Hugh Possingham.

Chapter 6

This chapter was written by the Candidate with the minor editorial input of Hugh Possingham.

Statement of parts of the thesis submitted to qualify for the award of another degree

None.

viii

Acknowledgements

The Australian Research Council Centre of Excellence for Environmental Decisions, the Australian Government’s National Environmental Research Program, and an Australia Postgraduate Award scholarship provided me with the financial support to conduct my research.

Hugh Possingham has provided me with the unparalleled opportunity of conducting my research in the remarkable Centre of Excellence that he leads with marvellous vision: for this I am extremely grateful. His ability to provide clarity and guidance is extraordinary, and I will always be inspired by his indefatigable application of logic and the scientific method to grave conservation issues. I could not have asked for a better or more appropriate advisor for my PhD research.

I also thank the following people for being supportive of my work: Ayesha Tulloch continues to impress me with her enthusiasm, creativity, intelligence, proficiency, and generosity–she motivated me to do more than I thought possible. Kerrie Wilson provided expert speedy feedback and encouraged and inspired me with all that she has accomplished and all that she will. Greg Baxter, Yvonne Buckley, and Jonathan Rhodes generously gave time for critiquing my research and giving feedback for my mid-candidature and thesis reviews. Ros Taplin started me down the path of a PhD and kept me afloat during a baffling time; her humbleness hides her remarkable talents. Steve Jones gave me a job in environmental science just when I thought it was impossible to work in a field I believe in; his care led me to my topic of research in support of threatened species. Rachel Lyons has the real heavy-weight responsibility of on-the-ground biodiversity conservation; she patiently provided me with the opportunity to try to help.

For efficiently making it possible to navigate the details, I thank Heather Christensen, Jane Campbell, Karen Gillow, Melisa Lewins, Gail Walter, and Waruna Wickramanayake.

I’m most appreciative of the people who generously shared their technical expertise. Steven Phillips, Richard Pearson, and Ramona Maggini provided immensely useful species distribution modelling workshops. Simon Blomberg, Ross Dwyer, Chris Brown, and Claire Runge helped me realise I really could program in R, and Jeffrey Hanson is undoubtedly the most generous and creative R geek in Australia. Rebecca Runting and Matt Watts provided helpful guidance in working with Marxan. Abbey Camaclang, Rob Clemens, Hong Du, Yi Han, and Fei Ng bravely joined our Spatial Ecology Lab Writing Group. Morena Mills, Terry Walshe, David Pannell, Geoff Park, and Matt Watts made our NRM workshop happen. Abbey Camaclang, Fleur Maseyk, and Dirk Hovorka provided helpful comments on my thesis. Grant Dobinson from edit·optia provided proofreading services for my thesis introduction and conclusion.

For taking the time out of their busy schedules to talk with me about fire management, I am grateful to Peter Leeson, Samantha Lloyd, Fiona Gibson, Kerrie Lock, Chris White, and Quinn McCleod.

My gratitude extends to all my friends on the journey who contributed in their own unique ways: Abbey Camaclang, Anna Renwick, Carla Archibald, Claire Runge, Danielle Shanahan, Duan Biggs, Fei Ng, Fleur Maseyk, Hawthorne Beyer, Hong Du, Hsien-Yung Lin, Jane McDonald, Jeffrey Hanson, Jen McGowan, Jeremy Ringma, Joe Bennett, Jude Keyse, Jutta Beher, Kate Helmstedt, Kiran Dhanjal-Adams, Laurel Osborne, Liz Law, Maria Beger, Maria Martinez-Harms, Megan Barnes, Megan Evans, Nathalie Butt, Payal Bal, Ramona Maggini, Rebecca Runting, Rob Clemens, Scott Atkinson, Sean Maxwell, Sugeng Budiharta, Sylvaine Giakoumi, Tal Polak, Tessa Mazor, Tim Holmes, Veronica Gama, and Viv Tulloch.

For keeping me balanced and on the path I am grateful for the wisdom of Leonie Seabrook, Kat Woods, Celie Xie, Adele Bergin, Gynette Kershaw, Mark Togni, Cameron Story, Danny Hill, Kathy Mason, Suzanne Gray, Shaktiprem, and all the dedicated farmers at the Gold Coast Organic Market.

Dirk Hovorka’s support for me was unblinching—I thank him for believing in me, challenging me, sharing dreams with me, and always being there for me. He also made sure we regularly went out to be with the birds and reminded me that what we do does change the World.

Keywords

Biodiversity Prioritisation, Complementarity, Cost-effectiveness Analysis, Decision Support, Nature Conservation, Priority Threat Management, Return on Investment, Systematic Conservation Planning, Threatened Species, Threats

Australian and New Zealand Standard Research Classifications (ANZSRC)

ANZSRC code: 050202, Conservation and Biodiversity, 34%

ANZSRC code: 050209, Natural Resource Management, 33%

ANZSRC code: 149902, Ecological Economics, 33%

Fields of Research (FoR) Classification

FoR code: 0502 Environmental Science and Management, 67%

FoR code: 1499 Other Economics, 33%

Table of Contents

Abstract ...... ii

Declaration by author ...... iv Publications during candidature ...... v Publications included in this thesis ...... vi Contributions by others to the thesis ...... vii Statement of parts of the thesis submitted to qualify for the award of another degree ...... viii Acknowledgements ...... ix Keywords ...... xi Australian and New Zealand Standard Research Classifications (ANZSRC) ...... xi Fields of Research (FoR) Classification ...... xi Table of Contents ...... xiii List of Figures ...... xx List of Tables ...... xxi List of Supplementary Information Figures ...... xxii List of Supplementary Information Tables ...... xxiii List of Abbreviations used in the thesis ...... xxvi 1 Introduction ...... 5

1.1 Threat-abatement action is required for biodiversity conservation ...... 5 1.1.1 Decision-making for biodiversity conservation ...... 5

1.1.2 Economics in conservation ...... 6

1.1.3 Spatial conservation prioritisation ...... 7

1.1.3.1 Prioritisation objectives ...... 8

1.1.4 Addressing research needs ...... 9

1.1.5 Research aims ...... 9

1.2 Advancing understanding of decision support for managing threats to species ...... 10 1.2.1 Primary research questions ...... 11

1.2.2 Theme 1: Manage threats to benefit species ...... 12

1.2.3 Theme 2: Spatially explicit management prioritisation for multiple species and threats under resource constraints ...... 13

1.2.4 Theme 3: Accessibility in decision support ...... 14

xiii

1.2.5 Theme 4: Relationships between actions affect conservation outcomes ...... 14

1.3 Study area ...... 16 1.3.1 Australian natural resource management ...... 16

1.3.2 Burnett-Mary Natural Resource Management Region ...... 17

1.3.3 Species of concern and threats to their persistence and recovery ...... 18

1.4 Chapter summaries ...... 19 1.4.1 Research goal ...... 21

2 All threats are not equal: Coarse- and fine-scaled indirect threat maps compared...... 25

2.1 Abstract ...... 25 2.2 Introduction ...... 25 2.3 Methods ...... 29 2.3.1 Study region ...... 29

2.3.2 Species distributions ...... 30

2.3.3 Species ...... 31

2.3.4 Threats to species ...... 31

2.3.5 Indirect threat map comparison...... 33

2.4 Results ...... 34 2.4.1 Threats to species ...... 34

2.4.2 Spatial extent of threats ...... 34

2.4.3 Spatial location of threats ...... 35

2.4.4 Measures of indirect threat map comparisons...... 35

2.5 Discussion ...... 39 2.5.1 Advantages and disadvantages of SDMs and range maps for indirect threat mapping ...... 40

2.5.2 Recommendations and cautions ...... 41

2.5.3 Conclusion ...... 42

2.6 Acknowledgements ...... 43 2.7 Supplementary Information ...... 43

xiv

2.7.1 Appendix 2A Regional priority threatened species distribution modelling methods and model performance validation and assessment...... 43

2.7.1.1 Species presence records and environmental variables ...... 43

2.7.1.2 Model predictive performance ...... 44

3 Informed actions: Where to cost-effectively manage multiple threats to species to maximise return on investment ...... 53

3.1 Abstract ...... 53 3.2 Introduction ...... 54 3.2.1 Research aim ...... 56

3.3 Methods ...... 57 3.3.1 Case Study ...... 57

3.3.2 Framework for conservation return-on-investment analysis ...... 58

3.3.3 Step 1. Problem definition ...... 58

3.3.4 Step 2. Define realistic estimates of expected benefits ...... 59

3.3.5 Step 3. Integrate realistic cost estimates ...... 61

3.3.6 Step 4. Solve the problem ...... 62

3.3.6.1 Comparing expected benefits in an ROI approach ...... 63

3.3.6.2 Spatial priorities for single or multiple actions using ROI analysis ...... 64

3.4 Results ...... 65 3.5 Discussion ...... 72 3.5.1 Informed investment when budgets are low can yield high expected benefits to species ...... 73

3.5.2 ROI results are sensitive to the way in which expected benefits are defined and how benefits and costs co-vary ...... 73

3.5.3 Multiple-action strategy for multiple species is better than single-action strategy ...... 75

3.5.4 ROI informs where to act ...... 76

3.5.5 Conclusion ...... 77

3.6 Acknowledgments ...... 78 3.7 Supplementary Information ...... 79 xv

3.7.1 Appendix 3A The case study ...... 79

3.7.2 Appendix 3B Model of opportunity costs to cease grazing ...... 80

3.7.3 Appendix 3C Model of management costs to bait foxes ...... 81

3.7.3.1 Net present value of 20-year fox-baiting management program ...... 82

3.7.3.2 Cost of one-year fox-baiting management program ...... 82

3.7.3.3 Labour cost to administer baits ...... 82

3.7.3.4 Cost to operate a vehicle ...... 84

3.7.3.5 Cost to purchase bait ...... 84

3.7.4 Appendix 3D Linear regression statistics for random management ROI curves ...... 86

3.7.5 Appendix 3E What drives early gains in efficiency? ...... 87

3.7.6 Appendix 3F Cost-effectiveness of threat management under different budgets ...... 88

3.7.7 Appendix 3G Variation in management cost sensitivity analysis...... 90

4 Accounting for dependencies in threat management can alter conservation priorities ...... 99

4.1 Abstract ...... 99 4.2 Introduction ...... 99 4.2.1 Research Aim ...... 101

4.3 Methods ...... 102 4.3.1 Case study description ...... 102

4.3.2 Decision problem ...... 103

4.4 Results ...... 104 4.4.1 Benefit dependency analysis ...... 104

4.4.2 Spatially dependent cost analysis ...... 107

4.5 Discussion ...... 108 4.5.1 Conclusion ...... 112

4.6 Acknowledgements ...... 112 4.7 Supplementary Information ...... 113

xvi

4.7.1 Appendix 4A Priority species to be managed and threats to their persistence...... 113

4.7.2 Appendix 4B Detailed methods ...... 115

4.7.2.1 Estimating independent and dependent expected management benefits ...... 115

4.7.2.2 Estimating independent management costs ...... 117

4.7.2.3 Estimating spatially-dependent costs ...... 117

4.7.2.4 Prioritising sites and actions for investment ...... 118

4.7.3 Appendix 4C Model of management costs to maintain fire regimes for biodiversity ...... 119

4.7.3.1 Objective ...... 119

4.7.3.2 Summary ...... 119

4.7.3.3 Justification ...... 119

4.7.3.4 Background ...... 120

4.7.3.5 Method ...... 120

4.7.3.6 Model caveats and acknowledgements ...... 121

4.7.4 Appendix 4D Analyses for assessing similarity between scenarios with and without considering benefit-dependencies in threat management ...... 122

4.7.5 Appendix 4E Analyses to determine difference in spatial configuration of sites selected for fox management under a budget of $10 million, without and accounting for a spatial cost dependency...... 126

4.7.6 Appendix 4F Descriptive statistics for expected benefits (B), costs (C), and cost-effectiveness (CE) of threat management for five scenarios...... 128

4.7.7 Appendix 4G Descriptive statistics for management cost layers ...... 130

4.7.8 Appendix 4H The variation in the expected benefits and costs of threat mitigation at different budgets ...... 130

4.7.9 Appendix 4I Sensitivity analysis with greater number of species with multiple threats ...... 132

5 Accounting for complementarity and action dependencies in spatial threat- management prioritisation ...... 139

5.1 Abstract ...... 139 xvii

5.2 Introduction ...... 139 5.2.1 Research aims ...... 142

5.3 Methods ...... 142 5.4 Results ...... 145 5.4.1 Different sites are prioritised when including dependencies between sites and species managed (complementarity) ...... 145

5.4.2 Similar sites are prioritised when including additional dependencies between management actions ...... 147

5.5 Discussion ...... 149 5.6 Supplementary Information ...... 152 5.6.1 Appendix 5A Supplementary tables and figures ...... 152

6 Conclusions ...... 165

6.1 Potential strategic responses for securing biodiversity persistence ...... 166 6.1.1 Protected areas are necessary but not sufficient...... 166

6.1.2 Ad hoc management may be inadequate ...... 167

6.1.3 Time spent improving knowledge may be counter-productive ...... 168

6.2 Economically based threat-action prioritisation as an alternative ...... 168 6.2.1 Research contributions ...... 168

6.2.1.1 Indirect threat maps derived from the predicted distribution models of threatened species identify threat abatement locations that aids decision support ...... 169

6.2.1.2 Spatially explicit prioritisation for management actions that cost- effectively abate multiple threats to many species ...... 169

6.2.1.3 Incorporating relationships between costs and benefits of management actions clarifies trade-offs ...... 170

6.2.1.4 Accessible decision support ...... 170

6.2.2 Theory and methods supporting conservation decision research...... 171

6.3 Assumptions and opportunities to augment research ...... 172 6.3.1 Data availability ...... 173

6.3.2 Predicted species distribution models ...... 173

xviii

6.3.3 Socio-economic costs ...... 174

6.3.4 Ecological considerations ...... 175

6.3.5 Likelihood of management success ...... 176

6.3.6 Sensitivity analyses ...... 177

6.3.7 Uncertainty ...... 178

6.3.8 Adaptive Management ...... 181

6.4 Concluding remarks ...... 183 7 References ...... 189

xix

List of Figures

Figure 1.1 Structural overview of thesis...... 11 Figure 1.2 Location of Burnett-Mary Natural Resource Management Region in South East Queensland, Australia...... 18 Figure 2.1 The Burnett-Mary Natural Resource Management (NRM) Region...... 30 Figure 2.2 Percentage of study area covered by threat category for indirect threat map datasets...... 36 Figure 2.3 Percentage of 33 nationally threatened species affected by eight threats...... 37 Figure 2.4 Quantity and allocation disagreement summed for total disagreement, proportion correct, and Kappa index of agreement for indirect threat map datasets...... 38 Figure 3.1 Total Return-on-Investment curves show greater expected benefit for the same level of investment when management sites are chosen by ranked cost- effectiveness...... 66 Figure 3.2 Total ROI curves are plotted to show the relationship between investment against the percent of the total benefit expected to be returned from site management with different threat action-abating strategies...... 67 Figure 3.3 The spatial distribution of the expected benefits of management for species who are vulnerable to (a) fox predation or (b) grazing threats are not necessarily the same as cost-effective locations of managing those threats...... 69 Figure 3.4 A greater percentage of species are expected to benefit from simultaneously managing threats with a combined management action strategy, but cost- effectiveness is greatest with initial investment and returns diminish rapidly...... 70 Figure 3.5 (a) Cost-effectiveness guides which threat abatement action to implement with additional investment in management for the greatest expected benefit to species under the constraint of a $10 M budget. (b) Investment in management costs is approximately half-and-half for fox and grazing action...... 71 Figure 3.6 Mapping the most highly-ranked cost-effective sites for acting to abate threats to species shows where to efficiently spend a management budget of $10 million...... 72 Figure 4.1 Return-on-investment curves when optimising for independent threat management compared with optimising for dependencies ...... 105 Figure 4.2 Maps illustrate where to manage, with which strategy, when investing $10 Million in threat abatement ...... 106

Figure 4.3 Return-on-investment curve for cost-independent threat management (invasive predator control) compared with accounting for a spatial cost- dependency...... 108 Figure 5.1Different locations for action are prioritised when we optimistically expect species to benefit proportionally from an action that abates one threat of many...... 146 Figure 5.2 When a dependency between actions is considered in addition to managing for all species through complementarity, similar locations for action are prioritised...... 148 Figure 5.3 Uncertainty maps provide an indication for each planning unit as to whether the same package of management actions was frequently selected for the planning unit ...... 149

List of Tables

Table 2.1 Number of threatened species, categorised by taxa, and data sources for the indirect threat map comparative analysis...... 32 Table 2.2 Detail of data used for comparing the spatial distribution of eight threats to species...... 33 Table 2.3 Percentage and number of species in the Burnett-Mary Natural Resource Management Region affected by eight threats...... 35 Table 4.1 Scenarios evaluated in analyses of threat management benefit- and cost- dependencies...... 104 Table 4.2 Area of management chosen for action varies under an independent versus a dependent management scenario when investing $10 million and different threat mitigation strategies...... 107 Table 5.1 When optimistically expecting species to benefit from management of any threat, a greater area is needed for managing threats to all species at least cost as compared to cost-effectively managing threats in species-rich, rare habitats...... 146 Table 6.1 A taxonomy of uncertainty in ecology and conservation biology...... 179

xxi

List of Supplementary Information Figures

Figure 2A.1 Spatial difference between threats represented by SDMs as compared to range maps...... 49 Figure 3E.1 Results of a sensitivity analysis show return-on-investment (ROI) curves of combined fox and grazing threat management action sites selected by ranked cost-effectiveness...... 87 Figure 3G.1 Total return-on-investment curves for expected benefit over all levels of investment and detail of ROI curves at low levels of investment...... 92 Figure 3G.2 As compared to the original scenario, proportion of investment in fox threat action is higher if grazing management costs are higher or if fox management costs are lower than originally estimated...... 93 Figure 3G.3 Under a management budget of $10M, cost-effective areas for managing the threat of foxes are depicted in red, for managing for grazing in green, and for managing both threats in blue for four management scenarios in which we vary management costs...... 94 Figure 4I.1 Threats versus number of species in original analysis and in sensitivity analysis...... 133 Figure 4I.2 Sensitivity analysis return-on-investment curves when optimising for independent threat management compared with optimising for dependencies, under optimistic and pessimistic management scenarios...... 133 Figure 4I.3 Sensitivity analysis maps illustrate where to manage, with which strategy, when investing $10 million in threat abatement...... 134 Figure 5A.1 Comparisons of most efficient solutions from Marxan with Zone analyses and from the Cost-effectiveness analysis and similarity between the two...... 152 Figure 5A.2 The total cost for implementing the most efficient Marxan with Zones scenario versus target values of 10-50%...... 153 Figure 5A.3 Percentage of area from most efficient MZA-optimistic and CEA- optimistic scenarios versus MZA-optimistic selection frequency...... 154 Figure 5A.4 The percentage of area in the CEA-optimistic scenario that is selected for any management strategy that includes fire, fox, or grazing action is plotted against the selection frequency of management units ...... 155 Figure 5A.5 Frequency (number of species) versus level of representation in MZA- optimistic and CEA-optimistic scenarios...... 156

xxii

List of Supplementary Information Tables

Table 2A.1 One-hundred twenty-nine species were included in the spatial threat analysis...... 45 Table 3A.1 List of priority threatened species in the Burnett-Mary Natural Resource Management region...... 79 Table 3B.1 Notation, value, and description of variables in grazing management cost model...... 80 Table 3C.1 Notation, value, and description of constants in fox management cost model...... 85 Table 3C.2 Labour, vehicle operation, and bait cost details and sources for fox management cost model...... 86 Table 3D.1 Cumulative species benefit can be predicted from cumulative random investment in threat management using linear regression to statistically explain return-on-investment curves...... 86 Table 3F.1 Cost-effectiveness values for discrete budgets and unlimited funding vary by benefit metric and management strategy, and also decrease with increased investment, when threat management sites are selected on the basis of ranked cost-effectiveness...... 88 Table 3F.2 Cost-effectiveness values for discrete budgets and unlimited funding vary by benefit metric and management strategy, and also decrease with increased investment, when threat management sites are randomly selected...... 89 Table 4A.1 Listing of priority species and their vulnerability to the threats of fire frequency and intensit, invasive species predation, and domestic stock- caused habitat degradation ...... 113 Table 4D.1 Fractional weight matrix: Used to count disagreements differentially when calculating weighted kappa statistic...... 122 Table 4D.2 Binary weight matrix: Used to count disagreements differentially when calculating a weighted kappa statistic ...... 123 Table 4D.3 Cohen’s un-weighted and weighted kappa statistic was calculated for comparisons between scenarios...... 123 Table 4D.4 Confusion Matrix: Independent versus Pessimistic Scenario ...... 124 Table 4D.5 Confusion Matrix: Independent versus Optimistic Scenario ...... 124 Table 4D.6 Confusion Matrix: Optimistic versus Pessimistic Scenario ...... 125 Table 4D.7 Percentage of manageable area in agreement of location and action ...... 125

xxiii

Table 4E.1 Compactness measure for sites chosen for cost-effective fox management action under a budget of $10 million...... 126 Table 4E.2 Descriptive statistics for nearest-neighbour distance of sites chosen for cost- effective fox management action under a budget of $10 million...... 127 Table 4F.1 Independent benefit management scenario ...... 128 Table 4F.2 Optimistic benefit management scenario...... 128 Table 4F.3 Pessimistic benefit management scenario ...... 129 Table 4F.4 Cost-independent scenario ...... 129 Table 4F.5 Cost-dependent scenario ...... 130 Table 4G.1 Management costs: reduced fire intensity and frequency, invasive predator control, reduced grazing pressure of domestic stock ...... 130 Table 4I.1 Cohen’s un-weighted and weighted kappa statistic for comparisons between scenarios for the sensitivity analysis...... 134 Table 4I.2 Confusion Matrix: Independent versus Pessimistic Scenario ...... 135 Table 4I.3 Confusion Matrix: Independent versus Optimistic Scenario ...... 135 Table 4I.4 Confusion Matrix: Optimistic versus Pessimistic Scenarios ...... 136 Table 5A.1 Fractional weight matrix used to count disagreements differentially when calculating a weighted kappa statistic ...... 157 Table 5A.2 Binary weight matrix used to count disagreements differentially when calculating a weighted kappa statistic ...... 157 Table 5A.3 Cohen’s un-weighted and weighted kappa statistic compare the most efficient Marxan with Zones analysis scenarios to the Cost-effectiveness analysis scenario ...... 158 Table 5A.4 Cohen’s un-weighted and weighted kappa statistic compare the most efficient Marxan with Zones analysis scenarios to the Cost-effectiveness

analysis scenario...... 158 Table 5A.5 Percentage of target constraints met in Marxan with Zones analyses as compared to Cost-effectiveness analysis ...... 159 Table 5A.6 A comparison of the total cost of the area selected for management in the most efficient Marxan with Zones analysis scenarios ...... 160 Table 5A.7 Cohen’s un-weighted and weighted kappa statistic compare the most efficient optimistic and pessimistic Marxan with Zones analysis over the entire study region and the most frequently selected management units ...... 160

xxiv

Table 5A.8 Zone management area selected in optimistic and pessimistic 30% target management scenarios...... 161 Table 5A.9 Proportion of frequently selected management units selected with high certainty for specified management zones in the optimistic and pessimistic Marxan with Zones scenarios...... 162

xxv

List of Abbreviations used in the thesis

ALA Atlas of Living Australia

CBA Cost-benefit Analysis

CEA Cost-effectiveness Analysis

CUA Cost-utility Analysis

DERM Department of Environment and Resource Management (Queensland)

GIS Geographic Information System

IUCN International Union for Conservation of Nature

MZA Marxan with Zones Analysis

NPV Net Present Value

NRM Natural Resource Management

ROI Return on investment

SDM Species distribution model

SNES Species of National Environmental Significance

xxvi

"The beauty and genius of a work of art may be reconceived, though its first material expression be destroyed; a vanished harmony may yet again inspire the composer; but when the last individual of a race of living beings breathes no more, another heaven and another earth must pass before such a one can be again."

~William Beebe (1877-1962)

Chapter 1 Introduction

1 Introduction

1.1 Threat-abatement action is required for biodiversity conservation

Biodiversity conservation efforts are having a positive effect (Rodrigues 2006, Hoffmann et al. 2010), but species extinction rates are high (Pimm et al. 2014) and an increasing number of species are listed as being at risk (International Union for Conservation of Nature (IUCN) 2013). Despite a high likelihood of a mass extinction event (Barnosky et al. 2011), concerted action to abate threats to biodiversity may reduce the number of species that become extinct in the near future (Wilson 1989, Pimm et al. 2014). However, even though it is within our means to conserve biodiversity (James et al. 2001), not enough funding is allocated for species management (McCarthy et al. 2012). Moreover, while conservation is often focused on single species, such as the most charismatic, or the most endangered species (Leader-Williams and Dublin 2000, Mace et al. 2007), by concentrating efforts on only a few particular species, the needs of other vulnerable species may be neglected especially when there are limited resources for management (Franklin 1993, Simberloff 1998, Possingham et al. 2002). Alternatively, focusing on abating threatening processes that operate across a landscape will better meet the needs of, and reduce risks to, assemblages of species and ecological communities that are affected simultaneously from the threatening processes (Salafsky et al. 2008, Auld and Keith 2009, Keith 2009). In this way, threat management may contribute towards efforts of maximising the persistence of multiple species.

1.1.1 Decision-making for biodiversity conservation

Central to biodiversity conservation is the need to make decisions about which threats to mitigate with what management action(s), and where. Decision theory provides a framework for making effective and efficient conservation choices (Possingham et al. 2001b). Management grounded in decision theory requires, at a minimum, a specific objective and a link between the costs and benefits of actions for achieving that objective (Possingham et al. 2001b, Wilson et al. 2007). The relationship between costs and benefits quantifies where the most benefit can be achieved for the least cost, leading to the ability to prioritise management choices (Bottrill et al. 2008). Conservation planning that has its foundation in decision theory is transparent and repeatable, in contrast to even well-intentioned and considered conservation decisions that are made in a less systematic or structured way (Possingham et al. 2001b, Wilson et al. 2009a, Wilson et al. 2009b). Decision theory supports making choices in a wide range of management dilemmas such as selecting a habitat-protection strategy that minimises risk of population extinction (Haight et al. 2002), determining when to stop managing threatened species (Chadès et al. 2008), choosing which

isolated threatened species populations to manage (McDonald‐Madden et al. 2008), and modifying management depending upon impediments to recovery action success (Ng et al. 2014). Decision makers can make more logical and accountable conservation investment decisions, when actions for addressing biodiversity decline are evaluated within a decision-theory framework.

1.1.2 Economics in conservation

For sustainability of the biosphere, it is increasingly necessary to integrate economics with ecology when making conservation management and policy decisions (Armsworth and Roughgarden 2001). Including the costs of conservation in decision-making enables more efficient allocation of conservation funding (Possingham and Wilson 2005), and is crucial to addressing large-scale and growing environmental challenges affecting ecosystem services (Polasky and Segerson 2009). Ecosystem services are the benefits provided by nature upon which all humans depend (Daily 1997), and the Millennium Ecosystem Assessment (2005) links ecosystems (underpinned by biodiversity) to human well-being. Environmental economics focuses upon allocating scarce resources for the greatest social welfare (Hanley et al. 2007). For example, “’cap and trade” is an economic approach to reducing pollution discharge, with the intent being to provide equitable human welfare by ensuring that the limited resource of clean air is available to all. However, different economic approaches in policy and management emphasise different goals, and the economic goal of maximising social welfare (measured in economic terms) could lead to unintended ecological outcomes (as measured in non-economic terms). For example if the scarcity value of an ecosystem service common good is ignored or exploited, over-depletion results (Garett 1968). Further, typically the value of an ecosystem service is measured monetarily in terms of willingness of the affected individuals to trade an increase or decrease in ecosystem services for a decrease or increase in other goods and services (Just et al. 1982). However, there may be limited understanding by the public of ecosystem services and their contribution to human welfare (National Research Council 2005, United States Environmental Protection Agency (EPA) 2009), resulting in environmental decisions being made without clear understanding of the ramifications. In addition, a narrow focus on economic factors that overrides ecological considerations could inadvertently lead back to the opportunistic conservation choices that systematic conservation planning approaches were developed to avoid (Arponen et al. 2010).

When the objective is to optimise outcomes for biodiversity, socio-economic tools for examining the relationship between the costs and benefits of conservation actions include: Cost-benefit analysis (CBA – the relationship between monetary costs and monetary benefits); cost-effectiveness analysis (CEA – the relationship between monetary costs and non-monetary benefits); cost-utility

analysis (CUA – comparisons between competing alternatives); and portfolio management (prioritising on the basis of risk and uncertainty of likely outcomes) (Hughey et al. 2003). Cost- benefit analysis is useful for targeting high returns on investment in conservation efforts (Duke et al. 2013), but when it is difficult to place a monetary value on the benefits of conservation, cost- effectiveness analysis is a useful approach for maximising conservation outcomes. The use of cost- effectiveness analysis (where costs are defined in monetary units but benefits are defined in terms other than by money) rather than cost-benefit analysis (where both costs and benefits are defined in monetary units) is suitable when estimating peoples’ willingness to pay for species protection is not possible. Cost-effectiveness analysis is intuitive, because it is used whenever people decide how to derive the most value (e.g. in satisfaction) from their money (Murdoch et al. 2007). Cost- effectiveness analysis has also long been used to evaluate health-care options relative to costs to make explicit the beliefs and values underlying allocation decisions that affect human lives (Weinstein and Stason 1977), and is compatible with an evaluation of decisions affecting species loss. By analogy, cost-effectiveness analysis is well placed for weighing the costs and benefits of conservation actions. Optimising for multiple socio-economic objectives in addition to optimising biodiversity outcomes is possible, as is accounting for social, political and economic constraints (Polasky et al. 2008).

1.1.3 Spatial conservation prioritisation

Spatial conservation prioritisation is a decision-theoretic approach for identifying locations where management actions will have conservation outcomes in a specific region (Ferrier and Wintle 2009). Spatial conservation prioritisation frequently focuses on designing efficient networks of protected areas and is known as systematic conservation planning (Margules and Sarkar 2007). Efficient protected areas contain different areas that are chosen to complement each other because they contain different species. Sites chosen depend on the list of previously-chosen sites, and this interaction or dependency between management at different sites is known as complementarity (Kirkpatrick 1983, Margules et al. 1988). Prioritisation requires information on the distribution of biodiversity features, and the costs of conservation action, to find areas that are comprehensive, adequate, representative, and efficient for achieving desired outcomes of either minimising biodiversity loss or maximising biodiversity gain (Margules and Pressey 2000). In addition to designing protected areas, current approaches have extended prioritisation-thinking to encompass conservation planning not only for biodiversity and other natural values, but also for other social and institutional considerations that underlie managing threats to biodiversity. Examples include planning for contributions of diverse land uses in production landscapes to biodiversity conservation (Wilson et al. 2010), aligning development and carbon conservation outcomes (Venter 7

et al. 2012), and designing protected areas that consider both socioeconomic and biodiversity factors (Klein et al. 2009). Fundamental to prioritisation, however, is defining an objective for what conservation outcome the actions are meant to achieve.

1.1.3.1 Prioritisation objectives

Two common classes of conservation prioritisation problems with separate objectives are the maximal-coverage or the minimum-set coverage problems. The objective of solving a maximal- coverage problem is to maximise representation of a conservation objective for a fixed budget (Church et al. 1996). In contrast, the objective of solving a minimum-set coverage problem is to minimise the cost of meeting specified conservation targets, such as a comprehensive representation of a set of conservation features (Pressey 2002). Optimisation algorithms are used to produce spatial prioritisation models, which are primarily tools for decision support (Williams et al. 2004). While there are many choices for optimisation techniques, (Moilanen and Ball 2009), two options that are particularly useful for addressing both types of conservation problems are cost-effectiveness analysis (CEA) and systematic planning with Marxan software. CEA is an iterative greedy algorithm that may not produce an optimal solution, but can yield locally optimal solutions that approximate a global optimal solution, whereas Marxan is widely accepted systematic planning software that makes use of a simulated annealing algorithm that produces multiple near-optimal solutions (Ball et al. 2009). CEA is increasingly recommended for decision support in conservation due to its simplicity and transparency (Naidoo and Adamowicz 2006, Moran et al. 2010, Laycock et al. 2011); a non-monetary measure of the conservation benefit of an action is divided by the cost of that action, and then these ratios are ranked to prioritise where the greatest benefit is expected for an investment in conservation (Hughey et al. 2003, Murdoch et al. 2007). Cost-effectiveness is computed for each candidate action, and for comparison and ranking, all benefits are expressed in the same units. Marxan has been used to address a wide range of requirements for protected areas, including quantifying ecological effectiveness of differential protection zones (Makino et al. 2013) and incorporating dynamic ecological processes site selection of protected areas (Levin et al. 2013), as well as for efficient management, such as reducing illegal activities leading to biodiversity loss (Plumptre et al. 2014). Although Marxan is perceived to be more computationally complex and may require more comprehensive data than CEA, both aim to guide efficient, repeatable, and transparent decision support for biodiversity conservation that is accessible to conservation practitioners facing complex management choices.

1.1.4 Addressing research needs

Despite increasing evidence that multiple actions need to be delivered to maintain or restore populations of species at threat, most systematic conservation planning and prioritisations do not consider threats, but rather only target species. In addition, planning that does account for threats usually focuses only on single threats and actions and, furthermore, rarely are additional interactions between management actions considered outside of complementarity (Moilanen and Ball 2009). Historically, spatial prioritisation has emphasised the identification of protected areas to offset the threat of habitat loss (Margules and Pressey 2000). Although comprehensive species representation in protected areas is desirable and provides the foundation of species conservation (Rodrigues et al. 2004a), protected areas alone are likely insufficient for long-term species persistence or recovery because species are affected by many threats, not only habitat loss. Some approaches that prioritise threat abatement efforts amongst multiple species take into account the costs and benefits of multiple threat actions, as well as the likelihood of success, but are not spatially explicit in locating management actions (Briggs 2009, Joseph et al. 2009, Carwardine et al. 2012, Carwardine et al. 2014). Although decision-support optimisation algorithms can accommodate the complexity of spatially prioritising management plans for multiple species and threats (Moilanen and Ball 2009), it can be difficult for practitioners to translate such models to on- the-ground conservation actions (Knight et al. 2008). Therefore, I pursue addressing these identified shortcomings by conducting research that furthers understanding of decision-support approaches that are accessible and pragmatic for spatially prioritising multiple threat-abating actions for many species.

1.1.5 Research aims

My research aims to inform applied conservation. Within a case-study framework, I examine how a regional-scale biodiversity manager could consider decision-making regarding action(s) for abating threats to multiple species across a landscape. This shift – from the idea that protecting species in reserves, alone, is enough to benefit biodiversity, to the recognition that additional actions are needed for abating multiple threats – diverges from some traditional approaches of identifying potential reserves for biodiversity hotspots. Alone, creating reserves for species protection can be insufficient for threat management that will benefit species; identifying locations for additional threat-abating actions for multiple species at a landscape level may better meet species’ needs. Within this landscape there is a need to prioritise threat abatement actions to achieve the greatest benefits to species within cost constraints. In the effort to minimise overall species loss, efficiently allocating scarce resources to recovery actions across species is required. My research integrates

economics and decision theory to prioritise cost-effective conservation actions for multiple species, but prioritisation is not a simple linear process because threats and actions have complex interactions that affect outcomes. My research also aims to examine the implications of interactions in costs and benefits of conservation actions across the landscape by showing how they affect the configurations of single or multiple threat-abatement actions in a spatial prioritisation of sites for management. In doing so, my research reveals trade-offs in conservation goals.

1.2 Advancing understanding of decision support for managing threats to species

This thesis advances approaches to threat management decision-making in biodiversity conservation using spatial prioritisation. It is structured (Figure 1.1) around four core, cross-cutting themes that are detailed below, and is organised in four research chapters addressing specific aspects of these themes (chapters summarised in Section 1.4). The first core theme is an emphasis on managing threats to benefit species, rather than only managing species. The fundamental objective is providing greater benefit to biodiversity through threat management action. While species are targeted for management more often than threats (Leader-Williams and Dublin 2000, Mace et al. 2007), by managing threats across a landscape it is likely possible to address the needs of many species (Noss 1987, Franklin 1993, Pressey 2004, Salafsky et al. 2008, Auld and Keith 2009, Keith 2009, Carwardine et al. 2012, Carwardine et al. 2014) (Chapters 2–5). Second, decisions need to be made about which management actions are needed to abate threats to multiple species, as well as where they are needed. While the spatial component of planning increases complexity, and so is sometimes avoided (Carwardine et al. 2014, Chadès et al. 2014), the next core theme in my research is about how spatial prioritisation of threat-abatement actions can guide benefits to multiple threatened species (Chapters 3–5). Third, while sophisticated optimisation algorithms may provide more precise answers, decision support that is accessible makes it more likely to implemented for on-ground conservation problems (Knight et al. 2006, Knight et al. 2008). The third theme in my research revolves around decision support that is manageable for being put into practice (Chapters 2–3). The fourth core theme is that the costs and benefits of threat- abating actions are not independent of one another, and to optimise management outcomes, this should be considered when making decisions. Because threats are interactive (Sala et al. 2000, Brook et al. 2008), the management actions that are meant to abate threats are also dependent upon each other. My research explores the implications of relationships between actions to abate threats that affect conservation outcomes, but are seldom considered (Chapters 4–5). While trade-offs are

intrinsic in any decision, the considerations I research in this thesis will lead to more informed management decisions and ultimately to action for biodiversity conservation.

Figure 1.1 Structural overview of thesis.

1.2.1 Primary research questions

Each of my research chapters examines the potential for a systematic and transparent approach to address spatial prioritisation problems in consideration of conservation decision makers who face funding constraints, and who also are constrained by real-world barriers in access to technology, expertise, and data availability. The primary questions my research pursues are as follows: When data are limited, how can we make decisions about spatially managing threats to species (Chapter 2)? Can we provide transparent, accessible, and useful decision support for management on the basis of the relationship between the costs and benefits of conservation actions (Chapter 3)? How do interactions or dependencies between the costs and benefits of conservation actions change investment priorities (Chapter 4)? And how does incorporating complementarity and dependence between management actions change least-cost conservation priorities (Chapter 5)? While recognising that different computational approaches, and more data, may enable more precise or 11

optimal outcomes, I argue that delays in action and ad hoc decision-making that does not consider the needs of many species may be detrimental to species persistence and I aim to provide an actionable alternative.

1.2.2 Theme 1: Manage threats to benefit species

A fundamental shift in thinking about biodiversity conservation has come with the recognition that managing species requires taking action to manage their threats (Wilcove et al. 1998). Priorities first were identified as being regional areas of high endemism (International Union for the Conservation of Nature (IUCN) 1997, Mittermeier and Mittermeier 1997, Olson and Dinerstein 1998), because areas with high numbers of species that cannot be conserved elsewhere are considered valuable to global biodiversity numbers. Subsequent approaches to prioritising global areas of importance were either proactive or reactive to threat susceptibility (Brooks et al. 2006), a measure of vulnerability (Wilson et al. 2005b). A reactive approach identified areas of high-threat vulnerability that are undergoing rapid habitat loss as high priority for protection of species (Myers et al. 2000, Hoekstra et al. 2005, Wilson et al. 2005a). A proactive approach located the remaining less-threatened wilderness areas (Bryant et al. 1997, Sanderson et al. 2002, Mittermeier et al. 2003). However, global species richness hotspots do not correspond with threat, and are therefore not indicative of priority areas in need of management (Orme et al. 2005). In addition, although it is commonly assumed that species-rich areas should be prioritised, the high diversity may be due to the presence of common and widespread species, when it is rare and restricted-area species in most need of conservation action (Lamoreux et al. 2005). Furthermore, priorities based on species richness that do not consider the cost of management for conserving biodiversity limit the ability to allocate restricted resources to conservation actions where it is most needed, most quickly (Possingham and Wilson 2005). Despite this, many prioritisations rely solely on species data without considering which threats are acting in the landscape and which mitigating actions might be needed. Prioritisations based on species richness alone are inadequate because they do not indicate threat, do not identify species in most need of management, and do not locate where to focus funding for action.

Ideally, we could map threatening processes to inform spatial conservation planning (Wilson et al. 2005b, Hof et al. 2011) – but this is difficult to do. Alternatively, overlaying the geographic ranges of species at threat indirectly represents the spatial extent of threatening processes such as overharvesting or invasive species (Orme et al. 2005, Schipper et al. 2008, Evans et al. 2011a). Conceptually these indirect threat maps are similar to species-richness maps, but are more informative because they are specific to threat. Prior research indicates that scale matters when

interpreting species richness maps; when created from overlaying broad-scale range maps, they may introduce spatial bias by overestimating species ranges (Conroy and Noon 1996, Rondinini et al. 2006, Jetz et al. 2008, Pineda and Lobo 2012). Similarly, indirect threat maps created from overlays of fine-grained species distribution maps provide a better indication of threats to species at the landscape scale of threat management than broad-scale maps. It is not commonplace for predicted species distribution models to be used in supporting conservation decision-making for on-ground conservation problems (Guisan et al. 2013), but my research demonstrates that indirect threat maps derived from species distribution models provide the spatial aspect of prioritising threat actions at the fine-scale (Chapters 2–5).

1.2.3 Theme 2: Spatially explicit management prioritisation for multiple species and threats under resource constraints

It is fundamentally complex to manage many threats to species under funding limitations. However, because multiple-action planning is structurally more complicated than single-action planning, it is addressed less often in the literature (Moilanen and Ball 2009) but is vitally needed for on-the- ground conservation. Threats are processes that extend across the landscape both spatially and temporally, and therefore affect multiple species. One implication is that threat mitigation through conservation action may result in the broad benefit of reducing risks to multiple species (Salafsky et al. 2008, Auld and Keith 2009, Keith 2009). Because species are affected by threats in different ways, threat abatement also requires undertaking multiple actions. Management actions are required both within protected areas (Brashares 2010), and in the larger landscape matrix (Franklin 1993) which encompasses developed land surrounding relict habitat and protected areas (Driscoll et al. 2013), and includes both government and private land holdings. Prioritising multiple threat-abating actions across the landscape will lead to greater efficiencies in terms of species managed (Wilson et al. 2006, Pressey et al. 2007, Wilson et al. 2007, Moilanen et al. 2011b).

Conservation actions guided by economic efficiency are also more likely to provide benefit to more species, and the socio-economic limits to threatened species management have emerged as critical dimensions of conservation planning (Orr 1991). This is a central premise of coherent decision making, and deserves prominence. A 'decision', coherently viewed, is defined as an irrevocable allocation of resources. Therefore, by definition, all decisions involve trade-offs in that the selection of one option makes others unavailable. Maximising conservation returns requires prioritisation in the allocation of limited resources where there is the greatest likelihood of successful conservation (Bottrill et al. 2008). Integrating economics into biological decision-making guides more efficient conservation plans (Balmford et al. 2003, Naidoo et al. 2006, Polasky et al. 2008, Moilanen et al.

2011b, Carwardine et al. 2012, McCarthy et al. 2012). Spatial priorities are substantially affected by heterogeneity in management costs that are distributed across a landscape, and so efficiency can be gained in achieving conservation goals by explicitly defining an objective that considers economics (Ando et al. 1998, Polasky et al. 2001, Naidoo et al. 2006, Carwardine et al. 2008). Spatially prioritising multiple actions for many species under funding constraints is complex, but straightforward methods are needed for providing decision support to conservation practitioners. My research extends the thinking of how prioritisation can be done transparently and rigorously (Chapters 2–5), both more simply using indirect threat maps and commonly used technology (Chapters 2–3), and then with the added complexity of considering the relationships between conservation actions (Chapters 4-5).

1.2.4 Theme 3: Accessibility in decision support

If conservation science is to succeed in its mission of conserving biodiversity (Soulé 1985), theory must be put into practice. Limited data access, sparse data, and barriers to implementation deter applications of well-intended conservation practice. Applied conservation science is also hindered by a knowing-doing gap, whereby research findings are not widely accepted, understood, or applied by practitioners (Knight et al. 2008). If a decision-support method for optimising biodiversity outcomes is perceived to be too complex for implementation and thus not used, potentially critically important information that would support decision-making could be disregarded. A challenge to conservation science is thereby identifying practices in decision support that are accessible, transparent and provide practitioners with confidence of their rigour, applicability, and practicality. Decision science provides a protocol for those striving to achieve explicitly stated objectives through understanding the trade-offs between the costs and benefits of conservation actions (Possingham et al. 2001b). Therefore it is a framework well suited for supporting conservation management decisions. In this thesis, benefits are defined as abating threats to species, and indirect threat maps represent spatially explicit abatement opportunities, because threat data are not often available to directly provide this information. My research finds that cost-effectiveness analysis, calculated with common spreadsheet and Geographic Information System software, provides accessible spatial decision- support. Decision-support that is accessible, yet rigorous, may have a higher likelihood of uptake by practitioners than complex approaches (Chapter 2–3).

1.2.5 Theme 4: Relationships between actions affect conservation outcomes

Ecosystems contain coupled components such that a change in one component will result in other changes. In the same manner, conservation planning approaches contain relationships between actions, yet these dependencies are often overlooked. A dependency between management actions 14

occurs when the whole (i.e. a strategy of multiple actions) is not the same as the sum of the parts (single actions). Complementarity is a commonly considered dependency between the benefits of conservation action (Kirkpatrick 1983, Margules et al. 1988), and it is a central principle of systematic conservation planning (Justus and Sarkar 2002). Specifically, if the goal is to conserve as many species as possible in at least one location, protecting two sites with very different species contributes to that goal, because the sites are complementary. The benefits of conserving more of a species’ distribution are reduced once it has been partially represented in a protected area, and the benefit of conserving two sites is not simply the addition of the richness of the species in each site. There is a non-linearity in the benefit function and diminishing returns from conserving more of the same species (Moilanen 2007). To more efficiently represent all species, complementary sites are preferred. The spatial dependency between management sites is also often addressed in systematic planning, because it affects compactness of protected areas (Possingham et al. 2000). An aggregated protected area network provides greater benefits to species than one that is fragmented (Moilanen and Wintle 2006). Other dependencies between costs and benefits are rarely examined though they may have important implications for prioritising threat actions.

Most often, multiple threatening processes interact (Sala et al. 2000, Brook et al. 2008). Dependencies or interactions between threats affect the outcomes of threat management by changing the ways that species respond when multiple threats are present in the landscape (Evans et al. 2011b, Regan et al. 2011). Therefore, the traditional focus on single threats is suboptimal for securing species and ecosystems (Brown et al. 2013b), and efficiency is improved by managing complex threats affecting multiple species (Burgman et al. 2007). For simplification in conservation planning, a common assumption is that threats can be assessed independently and their effects added together to estimate cumulative effects (Halpern et al. 2008, Ban et al. 2010, Vörösmarty et al. 2010). However, because threats interact, the benefits of threat abatement will depend upon which of multiple threats are managed, and benefits may be non-linear. In some cases species will benefit even if one of multiple threats are managed (e.g. appropriate fire regimes can control disease; Regan et al. 2011), while in other cases all threats must be managed for species to benefit (e.g. invasive foxes and rabbits both need control; Evans et al. 2011b). Dependencies between the costs and benefits of threat management are seldom addressed when prioritising conservation actions (Duke et al. 2013). My research expands on the idea of dependencies by specifically examining how dependencies between management actions affect prioritisations (Chapters 4 and 5).

1.3 Study area

This thesis focuses on a natural resource region with species of concern and expert-identified threats – it is a real-world management problem. All analyses were based on existing observational data for the species as found in accessible databases. In general, many threatened species are data-deficient, including little being known of their spatial distribution, which deters focused action. The data sparseness contributes to uncertainty in predicting species distributions, and minimal availability of costing data for management actions further exacerbates difficulties in the analysis. However, management decisions must still be made despite not having full knowledge, and my research realistically reflects the issues faced by managers in this context.

One approach for managers is decision support that transparently considers the trade-offs between the costs and benefits of threat-abating actions. Therefore, my research pursues a better understanding of prioritisation and systematic planning approaches that may inform threat management decisions. The Burnett-Mary Natural Resource Management (NRM) Region case study is the locational foundation of my thesis. A research focus on one natural resource management area with defined species of concern, threats, management actions and management costs enables detailed analysis and comparison of prioritisation and systematic conservation planning in the same context, as opposed to comparisons across circumstances where contextual differences may dominate the analysis. Although research results are particular to the case study, the implications of the research and the methods used are broadly applicable to conservation problems worldwide where there are barriers to overcome in implementing threat management for biodiversity conservation under resource constraints.

1.3.1 Australian natural resource management

Natural resource managers responsible for regional-scale biodiversity conservation are challenged by decisions about where to manage which threats given limitations in funding, technology, expertise, and social/political will. Regional NRM bodies support a community-based board that is responsible for integrated management of regional natural resources (ACIL Tasman 2005). Adaptive management and its participatory approach underpins natural resource management in Australia (Ewing et al. 2000), which has developed through the programs of Landcare, National Action Plan for Salinity and Water Quality, and Natural Heritage Trust, in an attempt to synergise the interests of community, industry and government (Curtis et al. 1999, Paton et al. 2004). A capacity-related characteristic assessment of the Australian NRMs (Robins and Dovers 2007) indicates many NRMs, including the Burnett-Mary NRM, have limited access to scientific research support to strengthen management practice. 16

The boundaries of Australia’s NRM areas are based upon a river catchment and bioregional framework (Ewing et al. 2000, Ewing 2003). Bioregions comprise landscapes that are distinct areas of similar climate, geology, landform, and vegetation and animal communities, and are mapped according to the Interim Biogeographic Regionalisation for Australia (Thackway and Cresswell 1995).

1.3.2 Burnett-Mary Natural Resource Management Region

As a case study of the challenges facing conservation practitioners, my research centres on biodiversity conservation in the Burnett-Mary NRM, one of 14 NRM regions in Queensland and of 56 NRM regions throughout Australia. The Burnett-Mary NRM Region is located in southeast Queensland, Australia (Figure 1.2). Covering 56 000 km2 in extent, the Burnett-Mary NRM Region encompasses watersheds of the Mary, Kolan, Burnett, Auburn, Boyne, Elliot, Gregory, Isis and Burrum Rivers and their tributaries, and is comprised of parts of the South East Queensland and Brigalow Belt bioregions. The Burnett-Mary NRM region contains a diversity of ecosystems, including rainforest, eucalypt woodlands and forest as well as sandy heaths, coastal dune formations, mangroves and salt marsh. The 2010 regional population was estimated to be 290 000, but is projected to reach 350 000 by 2026 (Australian Government 2010). Inhabitants reside primarily near the east coast in the urban centres of Bundaberg, Gympie and Maryborough. Major regional industries include dairying, grazing, forestry, irrigated cropping, fisheries and tourism, and the region has a history of land-use decisions that threaten its biodiversity. For example, the Paradise Parrot (Psephotus pulcherrimus) formerly inhabited the Burnett-Mary NRM, but was last sited in the 1920s and is now presumed extinct due to habitat destruction, among other causes (Olsen 2007a). More current land-use decisions have favoured threatened species: a 2009 Commonwealth decision to reject a proposed dam on the Mary River was based on unacceptable impacts to nationally threatened species including the Australian lungfish (Neoceratodus forsteri), Mary River turtle (Elusor macrurus), and Mary River cod (Maccullochella mariensis) (Australian Broadcasting Corporation (ABC) News 2009). The Burnett-Mary NRM is diverse in species, many of which face varied threats, and so species in the area are need of considered management.

Figure 1.2 Location of Burnett-Mary Natural Resource Management Region in South East Queensland, Australia.

1.3.3 Species of concern and threats to their persistence and recovery

In this case study, the species of concern were identified using a modified version of a species prioritisation method proposed by Marsh et al. (2007) in a recovery strategy coordinated by the State Government of Queensland, Australia in a species prioritisation undertaking (Department of Environment and Resource Management 2010). Species-based conservation efforts may focus on single species approaches such as keystone, umbrella, flagship, or indicator species, or on multi- species approaches such as focal or landscape species (Mace et al. 2007). However, the Back-on- Track species prioritisation framework aims to guide conservation management and species recovery through knowledge transfer by identifying threatened species with high potential for recovery, as well as their common threats and actions that may affect a range of species (Queensland Government 2012). Most of the species are listed under State (Nature Conservation Act 1992) or Commonwealth (Environment Protection and Biodiversity Conservation Act 1999) legislation, or they are species of regional concern whose declining regional status is not reflected in these legislative acts (BMRG, 2008). The framework is intended to encourage multi-species or landscape approaches to conservation. However, even though knowledge transfer is known to be a weak instrument, management within the framework has not been enforced by legislation or enabled by funding incentives and has had limited uptake (Milena Kim, Pers. Comm.).

My research aims to use the compiled information on species, threats, and actions (Department of Environment and Resource Management 2010) to guide spatially explicit, cost-efficient, conservation because although the Back-on-Track framework is descriptive, it does not provide substantive direction for actionable uptake. My research focuses on those species for which I was 18

able to model spatial habitat distribution (models detailed in Chapter 2 and species and threats detailed in Chapter 3 and 4), and three threats affecting the species of concern: too frequent and intense fire, an invasive predator (the red fox, Vulpes vulpes), and habitat degradation caused by domestic stock. These three threats are harmful to native species in Australia and globally (Fleischner 1994, Lowe et al. 2001, Bradstock et al. 2012), and are IUCN-listed threatening processes (International Union for Conservation of Nature (IUCN) 2014). My research estimates the spatially distributed cost of actions to abate these threats, including proactive predator control by lethal baiting, fire management for biodiversity according to vegetation type, and grazing reduction or removal through a stewardship agreement. Management cost models are detailed in Chapter 4 (fox and grazing) and Chapter 5 (fire). My research aims to provide spatially explicit, systematically derived, defensible, and actionable guidance for abating specific threat configurations at multiple sites in the most cost-effective manner.

1.4 Chapter summaries

In brief, the four research chapters of my thesis contribute to the field of spatially explicit systematic planning, through an assessment of choosing management priorities in the Burnett-Mary Natural Resource Management Region of Australia. First, to spatially focus threat management actions, we need to determine where in the landscape species are vulnerable to threats. Indirect threat maps, derived by overlaying habitat maps of species vulnerable to threats, are not yet widely used in systematic conservation planning. However, they can be practical to implement when data access is limited and scale of analysis is carefully considered (Chapter 2). Second, CEA may be a practical prioritisation alternative to more sophisticated systematic conservation planning approaches. I find that CEA grounded in decision science provides practical decision support in threat-management considerations (Chapter 3). Third, the implications of considering dependencies between the costs and benefits of threat management using decision science and CEA are not well understood. I caution that assuming independence in threat management actions may misdirect resource priorities or result in failing to achieve conservation objectives (Chapter 4). Lastly, threat- management prioritisations that consider complementarity will be different from those that do not. Cost-effective management of species-rich rare habitats may lead to not meeting the needs of some species, whereas complementarity explicitly accounts for managing all species but in comparison, requires a larger region be managed (Chapter 5). More detailed descriptions of the four research chapters follow.

Understanding where key threats are likely to affect biodiversity is a crucial step in determining where to implement mitigating actions (Chapter 2). Modelling the distribution of threatening

processes can assist in identifying areas where management action is needed. When threats cannot be directly modelled, the predicted distributions of species affected by a particular threat can be used to indirectly represent the spatial extent and intensity of a threat. Indirect threat maps can then be used in further spatial analysis for decision support. Fine-scale spatial modelling of species habitat requires commitment of scarce time and resources, and it may be more expedient to implement management based on available data, regardless of its scale. Hence, we need to know whether substantial differences exist between indirect threat maps derived from more widely available but coarse-scale geographic range maps as compared to fine-scale predicted species distribution models. I quantify the differences between threats modelled from eight key threats to species and find that indirect threat maps are sensitive to the spatial scale of mapping and the species mapped. I caution that indirect threat maps based on broad-scale geographic range maps may mislead efforts to manage threats at a regional scale.

Conservation practitioners, faced with managing multiple threats to biodiversity and limited funding, must prioritise investment in different management actions, but applied use of sophisticated optimisation approaches is limited. I demonstrate a basic but rigorous cost- effectiveness analysis (CEA) to prioritise where to invest in management to abate threats to species, and examine the returns on investment (Chapter 3). With a method that can be applied using a basic geographic information system and spread sheet calculations, I use CEA to prioritise management actions for two threats to a suite of threatened species. I show how decisions based on cost-effective threat management depend upon how expected benefits to species are defined and how benefits and costs co-vary. A landscape management strategy that implements multiple actions is more efficient than managing only for one threat or more traditional approaches that don’t consider both costs and benefits. My approach provides transparent and logical decision support for prioritising different actions intended to abate threats associated with multiple species; it is of use when managers need a justifiable and repeatable approach to investment.

The costs and benefits of threat management actions are likely not to be independent of one another, but simple cost-effectiveness ranking to prioritise conservation actions assumes that there are no dependencies between those actions. A “dependency” exists when the costs and/or benefits of one management action interact non-additively with those of another action. I assess how dependencies alter priorities for managing threatened species (Chapter 4). I use cost-effectiveness analysis to rank and prioritise three kinds of conservation action, appraising how two types of dependencies alter strategies for managing threatened species: 1) dependencies between the expected benefits to different threatened species of undertaking each action, and 2) a spatial/cost-dependency for a single action. I evaluate the differences in decisions that would be made when investing in priority 20

sites and actions when the benefits of management are assumed to be independent in contrast to when accounting for dependencies. If a species may be secured by managing only one of its threats and we continue to manage as though all threats are equally important and additive, funding could be misdirected to managing multiple threats. If a species may be secured only by managing all threats and we continue to manage as though this is not true, then conservation outcomes may fail. Considering spatial dependencies in management costs may lead to increased investment efficiency, as well as identifying spatially contiguous areas that may increase likelihood of management success. If dependencies in management action costs and benefits are not considered, limited investments may be poorly allocated, or threatened species management may fail. Addressing dependencies in conservation action costs and benefits is more complex than assuming they don’t exist, but it is also more realistic.

Decisions about managing threats to species can be informed by prioritising for cost-effective actions, but incorporating complementarity will change priorities, as will pessimistically assuming species will benefit only when all of the threats to their persistence are managed (Chapter 5). While a simple CEA prioritises cost-effective management in species-rich, rare, habitats (Chapters 3 and 4), it means that management needs for other species may be ignored. Similarly, a cost-minimising approach that does not manage all threats to all species may fail to deliver conservation objectives. Alternatively accounting for managing all species and all threats in a conservative strategy introduces the potential for spending too much on management of threats that may have been mitigated with less effort. Trade-offs between the costs and benefits of threatened species management are inevitable, but examining the implications of different strategies may lead to better decisions.

1.4.1 Research goal

Species persistence and recovery requires active interventions on the part of conservation managers, both in protected areas and in the surrounding habitats. The continuance and expansion of threatening processes creates an urgency to abate threats, if biodiversity is to be supported across the landscape. There is little room for management that produces suboptimal or worse than predicted outcomes. Practitioners are held accountable for management action decisions and resources spent to implement them despite the fact that they themselves are not responsible for the problems they are trying to solve, and therefore are in need of rigorous and transparent decision support for making choices that are defensible to stakeholders. Although indefensible, action may be delayed due to lack of data, lack of knowledge of prioritisation approaches or a sense that computation approaches are overly complex. But lack of action also represents a threat to

biodiversity: actions must be taken given current states of knowledge when the decision problem (objective) so dictates (Martin et al. 2012). Therefore it is the goal of my research to make justifiable, sound and considered decision support accessible to conservation practitioners, such as those managing regional natural resources and who are charged with the critical task of supporting biodiversity. It is my hope that this research can contribute to transparency in managing threats so that fewer species become extinct through inaction.

Chapter 2 All threats are not equal: Coarse- and fine-scaled indirect threat maps compared

2 All threats are not equal: Coarse- and fine-scaled indirect threat maps compared

2.1 Abstract

Understanding where key threats to biodiversity are located is a crucial step in determining where to implement mitigating actions. Mapping the distribution of threatening processes can assist in identifying areas where management action is needed, but producing direct threat maps may be difficult. Alternatively, the modelled distributions of species affected by a particular threat may be used to indirectly represent its spatial extent and intensity when we cannot directly map the threat, itself. Similar to a species richness map, an indirect threat map may then be used to quantitatively inform spatially-explicit management decisions. To derive an indirect threat map, it may be more expedient to implement available species distribution data, regardless of its scale; fine-scale predictive spatial modelling requires commitment of scarce time and resources. Hence, we need to know whether substantial differences exist between indirect threat maps derived from data of dissimilar scales. In this research, we quantify the differences between threats modelled from coarse-scale geographic range maps as compared to fine-scale predicted species distribution models for eight key threats to species in a biodiverse part of Queensland, Australia. We explore the sensitivity of our results in four scenarios in which we vary the threatened species being modelled, the spatial resolution of the data, and/or the number of threat intensity classes. In each case, we discovered substantial differences between the spatial distributions of threats, as measured by the overlap of threats represented by species distribution in the two different data sources. We find that indirect threat maps derived from data on threatened species are sensitive to the assumptions we make about the spatial scale of mapping and the species mapped, especially for more localised threats. We caution that indirect threat maps based on broad-scale geographic range maps may mislead efforts to make decisions about managing threats at a regional scale.

2.2 Introduction

Anthropogenic pressures have fundamentally altered the biosphere, generating threats to biodiversity that might be overcome with concerted investment in effective conservation strategies (Rodrigues 2006, Steffen et al. 2007, Hoffmann et al. 2010). Multiple threats imperil biodiversity, with habitat loss and degradation, overexploitation, introduced species, disease, and fire widely recognised as key threats to species persistence (Wilson 1992, Wilcove et al. 1998, Evans et al. 2011a). Knowing the location and extent of threats can help determine where to mitigate their effects, yet natural resource managers and policy-makers must often make such investment 25

decisions with a poor understanding of the spatial distributions of both the threats and the species those threats imperil.

Threats to species can be spatially modelled and quantified using one of two broad approaches: build a direct model of the threat itself, or indirectly model the threat using the distributions of species known to be affected by a particular threatening process. We refer to spatial models of threats as threat maps. Examples of direct threat maps include cumulative impacts in terrestrial (Walker et al. 1987, Sanderson et al. 2002) and marine (Halpern et al. 2008, Selkoe et al. 2009) ecosystems, vulnerability modelled from past patterns of threatening processes (Neke and Du Plessis 2004, Wilson et al. 2005a, Etter et al. 2006), where disease will spread with climate change (Rödder et al. 2010), and predicted distributions of successful invasive species establishment (Thuiller et al. 2005, Urban et al. 2007). Indirect threat map examples include those where individual species are expected to be affected by the threats of climate change (Thomas et al. 2004) and sea-level rise (Convertino et al. 2011). Indirect threat maps representing threatened species richness (Orme et al. 2005, Schipper et al. 2008) or more specifically those imperilled by specific threats (Schipper et al. 2008, Evans et al. 2011a) have appeal for detecting broad patterns of threats to multiple species, and may inform wide-ranging strategies for conservation effort. Determining where threats overlap with the species they affect leads logically to priorities for action whether using direct threat maps (Vale et al. 2008, Trebilco et al. 2011) or in combining direct and indirect threat mapping (Hof et al. 2011). The direct method of threat mapping is a more straightforward approach, but it can be difficult to implement. Alternatively, the indirect threat mapping approach may be more easily derived, but has not yet been used extensively in quantitative decision-making.

Indirect threat maps have potential for providing base data for systematic conservation planning of threat management actions, contributing to conservation decision-making. Indirect threat maps are surrogates for the spatial occurrence of threats; they are derived by overlaying maps of the spatial ranges of species know to be affected by specific threatening processes, similar to species richness maps (Orme et al. 2005, Schipper et al. 2008, Evans et al. 2011a). Indirect threat maps are not commonly used for conservation analytics, and two debates surround their use. The first debate topic pertains to the use of indirect threat maps instead of direct threat maps. The second debate topic pertains to questioning the reliability of the species distribution data used to represent indirect threats. Indirect threat maps may provide a viable approach to spatially representing the extent of threats to species of concern at the landscape scale, in particular when predicted species distribution models are combined to derive the indirect threat maps, in contrast to deriving them from more widely-available but broad-scale range maps.

Indirectly mapping threats is in contrast to directly mapping them, i.e. mapping where threats are known to occur, or where they are likely to occur. Direct threat maps may be derived from more conventional evidence, such as the area extent of pollutant plumes or the delineation of a proposed development. If the spatial extent of the direct threat does not overlap with the habitat of a species affected by that threat, i.e., they are not co-located, it would be unnecessary to manage the threat out of concern for its effect on that species. Threat management should be planned where both the threat and the species habitat are co-located. Furthermore, if the resolution of direct threat maps is too broad, they are not useful for fine-scale management, because they are inefficient in specifying where management is needed. Broadly mapped threats, alone, do not inform a manager where to manage, but producing high resolution direct threat maps may be difficult to do. In contrast, it is sensible to pay attention to managing threats to species where they live or can live. Waiting to apply management actions while using time and resources to directly map threats could be a costly alternative both in terms of resources utilised and in the potential for greater species loss in the meantime, due to mismanagement. Resources may be better directed to mitigating threats or other management actions.

As an alternative to a direct threat map, an indirect threat map is developed by overlaying distribution maps of the species known to be at risk from a specific threat. The fact that the species are threatened is a signature that tells us about the effect of those threats. If a species is declining due to a threat, the threatening process is interfering with the needs of that species. The signature of a threat in this sense is arguably its most direct measure. The underlying assumption of indirect threat maps is that we should protect species from threats where they are co-located, i.e. where indications are that suitable habitat is present for a threatened species. If a species has been eliminated from portions of its range that have been identified as being suitable habitat, i.e. part of the species distribution map, an indirect threat map could indicate where threat management may yet benefit the species if there is the possibility of assisted or natural recolonisation.

The second debate regarding indirect threat maps centres on the different outcomes resulting from overlaying geographic range maps as compared to overlaying predicted distribution maps, comparable to the different outcomes when deriving species richness maps (Pineda and Lobo 2012). There are known trade-offs to incorporating both of these data types into systematic conservation planning, due in part to the different methods for producing them and in the scale of resolution they are capable of representing (Rondinini et al. 2006, Jetz et al. 2008). Geographic range maps may be more accessible to practitioners, but the relative advantage of using more readily-available range maps is offset by the lack of spatial resolution and subsequent lack of precision (Conroy and Noon 1996, Pineda and Lobo 2012). Indirect threat maps derived from broad-scale range maps may 27

provide general guidance for where to focus attention (Schipper et al. 2008, Evans et al. 2011a), but lack fine-scale decision guidance for addressing specific managerial actions.

If maps are used as the foundation for management decisions, different types of errors in mapping habitat could translate to potentially adverse outcomes. Omission errors are when a species is mapped as being absent when it is present, and commission errors are when a species is mapped as being present when it is absent. Extending these potential errors to indirect threat maps means that omission errors would be when a threat is mapped as being absent when it is present, and commission errors would be when a threat is mapped as being present when it is absent. Both of these errors could mislead managerial decision-making. Not managing a threat where species are actually present may put species at risk (omission errors), and managing a threat where species are not present is a waste of resources that could be better spent on threat management in other vital locations (commission errors). Geographic range maps most often assume homogenous distributions at a relatively coarse scale, and are likely to generate commission errors, leading to recommendations to manage threats in areas where the species of interest do not actually exist. In contrast, predicted distribution maps are inferred from the environmental conditions of known species occurrences, and their commission and omission errors are dependent mostly upon model structure and data resolution (Guisan and Zimmermann 2000, Rondinini et al. 2006).

Decisions about implementing management actions to address threats to biodiversity most typically occur at the local or regional scale. Therefore, indirect threat maps derived from regional-scale data may better contribute to making threat management decisions than when derived from broad-scale data. Although species distribution models are not yet widely reported as being used in on-the- ground conservation decisions (Guisan et al. 2013), they are currently the best choice for predicting spatially explicit maps of habitat suitability (Guisan and Thuiller 2005, Elith and Leathwick 2009, Franklin 2009, Peterson 2011). However, species distribution modelling procedures require time and expertise, both of which may not be available to regional natural resource managers, who often need to make rapid decisions concerning investment in management using the best information currently available. Yet existing species distribution data (such as IUCN or BirdLife International range maps, originally developed for informing decisions at a national or global scale) will not necessarily be suited to regional planning due to potential inaccuracies introduced by down-scaling coarse-scale data (Stoms 1994, Hurlbert and Jetz 2007, Kriticos and Leriche 2010), but see Selkoe et al. (2009).

Concern over the potential inaccuracies introduced by the use of coarse-scale data usually leads to calls for more fine-scale data (Stuart et al. 2010). Although it is often assumed that more and higher

quality information is always better, the urgent need to address threats for biodiversity conservation brings to question how much time and effort should be invested in improving knowledge of species distribution (La Sorte and Hawkins 2007). When accompanied by the major threat of continuing habitat loss, delaying action to collect more biodiversity data for conservation planning can be counter-productive (Grantham et al. 2008). Therefore, it is sensible to explore existing mapped species distribution data options for considered decision-making, prior to investing in time- consuming data compilation for modelling threats.

Using a case study of threats to species in a regional natural resource management area in Australia, our research examines spatial differences in indirect threat maps when using two different data sources—existing as compared to spatially modelled data. Our research addresses a typical scenario for resource managers who need to decide whether to invest resources into modelling for decisions about threat management. We compare indirect threat maps of nationally-listed species based on existing geographic range maps that have been expertly reviewed and are generally available (Department of the Environment 2008a), with indirect threat maps based on predicted species distribution models of regional priority species (Department of Environment and Resource Management 2010) created specifically for our study area using the Maxent species distribution modelling software (Phillips et al. 2006).

The mismatches between these two data sources as to species and scale (a national list and range maps versus a regional list and species distribution models) raises three questions. First, we want to know if the list of threatened species used in modelling threats affects the spatial distribution of those threats; second, whether the spatial scale of the species distribution mapping affects the distribution of threats; and third, whether the mapped intensity of a threat (calculated as the proportion of species affected by a threat) is comparable between data sources. Our research evaluates the sensitivity of indirect threat mapping to the data and assumptions used to make those maps.

2.3 Methods

2.3.1 Study region

Our research centres on the Burnett-Mary Natural Resource Management (NRM) region, located in southeast Queensland, Australia (Figure 2.1) as a representative regional management area. Regional NRM bodies in Australia support a community-based board that is responsible for integrated management of regional natural resources (ACIL Tasman 2005). Natural resource managers are responsible for biodiversity conservation at many levels, but a capacity-related

characteristic assessment of the Australian NRMs (Robins and Dovers 2007) indicates many have limited access to scientific research and technical expertise to support and strengthen management practice. One of 14 Queensland and 56 Australian NRM regions, the Burnett-Mary NRM is approximately 55 000 km2 in extent. This region contains diverse ecosystems and has a history of land use decisions that both directly and indirectly affect the biodiversity of the region. For example, a 2009 Australian Commonwealth decision to reject a proposed dam on the Mary River was based on unacceptable impacts to matters of national environmental significance, namely threatened species including the Australian lungfish (Neoceratodus forsteri), Mary River turtle (Elusor macrurus), and Mary River cod (Maccullochella mariensis) (Australian Broadcasting Corporation (ABC) News 2009).

Figure 2.1 The Burnett-Mary Natural Resource Management (NRM) Region is located on the northeast coast of Australia, within the state of Queensland (QLD). Environmental features affect species distribution and guide natural resource management; catchment (main watershed) boundaries are delineated and shaded topography indicates higher (darker) and lower (lighter) elevations. Map Projection UTM Zone 56.

2.3.2 Species distributions

We used two sources of species distribution maps to create separate indirect threat maps for our comparison: first, we used geographic range maps from the Species of National Significance (SNES) spatial database (DEWHA 2008a) (hereafter referred to as range maps), and second, we used predicted species distribution models (hereafter referred to as SDMs) derived for priority threatened species (DERM 2010) in the Burnett-Mary NRM Region. The SNES range maps have been compiled for nationally threatened species in Australia. The maps were derived from a variety of methods, but generally represent “extent of occurrence” (sensu Gaston and Fuller 2009), with 30

some maps refined by species-specific habitat characteristics to represent “area of occupancy” (sensu Jetz et al. 2008). We modelled SDMs for Burnett-Mary priority threatened species using Maxent presence-only species distribution modelling software (Phillips et al. 2006) and publicly available data (Appendix 2A). We analysed only those areas with remaining or regrowth native vegetation (Queensland Herbarium 2010a, b) that are presumed to provide habitat for species in the region. We used ecologically meaningful variables and rigorous model performance validation and assessment (following Fielding and Bell 1997, Pearson et al. 2007, Kremen et al. 2008). Maxent has been found to be suitable for modelling distribution of species with small sample sizes of presence records (Hernandez et al. 2006, Pearson et al. 2007), as is often found with data documenting threatened species, such as in this study. However, sparse species data has been found to identify primarily strong environmental gradients (Barry and Elith 2006), and conservative use of predictions is recommended in the case of small sample sizes (Wisz et al. 2008, Elith and Leathwick 2009). We defined suitable habitat in the SDMS using a threshold that conservatively minimised both the area predicted and the omission error of the training (species presence) data (Appendix 2A), resulting in identification of areas that are at least as suitable as those sites where a species was recorded as being present (Pearson et al. 2007).

2.3.3 Species

All species in the analysis (Appendix 2A, Table 2A. 1) are threatened at the national, state, and/or regional level, and are a representative subset of different taxa (Table 2.1). We used two species lists for our analysis, and one subset of those lists. The first list comprised regional priority threatened species (Department of Environment and Resource Management 2010) and the second list comprised nationally threatened species (EPBC Act) in the region. A subset of 33 nationally threatened species was common to both lists.

2.3.4 Threats to species

We grouped threats to species into eight categories according to Evans et al. (2011a) after Venter et al. (2006): 1) Disease, 2) Fire and fire suppression, 3) Habitat loss, 4) Introduced species, 5) Native species interactions, 6) Natural causes, 7) Overexploitation, and 8) Pollution. We attributed threats to species from documentation compiled in the Species Profiles and Threats Database (Department of the Environment 2008b), from recovery plans, action plans, species management profiles, or the Burnett-Mary priority species action plan (Department of Environment and Resource Management 2010).

Table 2.1 Number of threatened species, categorised by taxa, and data sources for the indirect threat map comparative analysis.

Regional Nationally In-common priority threatened (SDMs* and Taxa (SDMs*) (Range Maps**) Range Maps**) Amphibians 8 3 2 Birds 6 6 2 Fish 4 4 3 Insects 0 1 0 Mammals 5 2 2 Reptiles 9 5 2 Vascular plants 33 67 22 Total 65 88 33 *SDMs = predicted species distribution models (Appendix 2A) **Range maps = Species of National Environmental Significance range maps (Department of the Environment 2008a)

We modelled spatial distributions of eight threats by overlaying species distribution maps of those species affected by each threat. We calculated models of the proportion of species affected by each threat with respect to the overall threatened species richness within the Burnett-Mary NRM Region using methods adopted from Evans et al. (2011a). We refer to each model of the spatial distribution of a threat, as derived from species affected by that threat, an “indirect threat map.” We analysed data in raster format (500 x 500 m pixel-size) using the Spatial Analyst implemented within the Geographical Information System (GIS) software ArcGIS vs 10. We converted vector (polygon) range maps to raster, matching the data resolution and spatial extent of the Maxent-derived SDMs.

In total, we created eight datasets for our comparative analysis, each comprising eight indirect threat maps, i.e. for each of four scenarios, we created a pair of comparable geographic range map and SDM indirect threat map datasets and each dataset consisted of eight indirect threat maps (described further, below). We up-scaled species’ predicted habitats to subcatchment (n = 1434; x̄ = 38.25 km2) management units (Stein 2007) for the list, up-scaled, and aggregation comparison scenario datasets; if a species was present anywhere within the subcatchment, we assigned the entire subcatchment that species presence. We created proportional indirect threat maps at original and subcatchment data resolutions and then reclassified them into six classes: 1) 0% (of all species threatened), 2) 1-20%, 3) 21-40%, 4) 41-60%, 5) 61-80%, and 6) 81-100%. We masked non- vegetated areas out of the range maps to compare with the SDMS for the direct comparison. We aggregated threat proportions in subcatchment indirect threat maps to two classes: 1) threatened (> 0%) and 2) not threatened (0%) for the aggregation comparison scenario.

2.3.5 Indirect threat map comparison

We investigated whether the spatial distribution of the eight different threats is dependent upon the species distribution data source used to model the threat in four comparison scenarios. In each comparison scenario, we assessed differences between threats modelled from Maxent-derived SDMs as compared to threats modelled from SNES geographic range maps (Table 2.2). We examined the differences in the spatial distribution of eight threats and their intensity (proportion of species affected) in the four comparison scenarios summarised below. First (Table 2.2 list comparison), we compared the indirect threat maps derived from two different priority lists of threatened species at the subcatchment data resolution: a) 65 regional priority threatened species (Department of Environment and Resource Management 2010) (modelled by SDMs) and b) all (88) nationally threatened species (Department of the Environment 2008b) (modelled by range maps). In contrast, for the next three comparisons, we kept the species list identical (33 nationally threatened species) but varied data resolution. Second, we analysed threats modelled with SDMs as compared to range maps with both sources of data scaled to the original (500 m pixel) resolution of the SDMs (Table 2.2 direct comparison). Third, we compared the indirect threat maps derived from the SDMs and range maps at the up-scaled (subcatchment) data resolution (Table 2.2 up-scaled comparison). Fourth, we compared indirect threat maps derived from the SDMs and the range maps with areas of threat intensity aggregated to threatened versus not threatened (Table 2.2 aggregation comparison).

Table 2.2 Detail of data used for comparing the spatial distribution of eight threats to species.

Threat Indirect Comparison Data classes Species threat map Scenario resolution (n) (n spp) dataset Data source Regional priority 1 1 SDMs* threatened (65) List subcatchment 6 All nationally Range Comparison 2 threatened (88) maps** 2 500m pixel In-common 3 SDMs Direct (SDM 6 nationally 4 Range maps Comparison original) threatened (33) 3 In-common 5 SDMs Up-scaled subcatchment 6 nationally 6 Range maps Comparison threatened (33) 4 In-common 7 SDMs Aggregation subcatchment 2 nationally Comparison threatened (33) 8 Range maps *SDMs = predicted species distribution models (Appendix 2A) **Range maps = Species of National Environmental Significance range maps (Department of the Environment 2008a)

We compared indirect threat maps using the Map Comparison Kit (http://www.riks.nl/mck/, Hagen- Zanker et al. 2010) and evaluated them using the Kappa (Cohen 1960) index of agreement, the percentage correct, and the proportion of disagreement represented by quantity and allocation disagreements (Pontius Jr. and Millones 2011), all calculated from cross-tabulation matrices. Quantity disagreement measures the difference between two maps that is due to an imperfect match in the proportions of categories compared, and allocation disagreement measures the difference between two maps that is due to the imperfect match in the spatial allocation of the categories, given the proportions of categories compared in the two maps (Pontius Jr. and Millones 2011). The two components of disagreement measured (quantity and allocation) sum to total disagreement, and the percentage correct equals 100 minus this value.

2.4 Results

2.4.1 Threats to species

First, we assessed the percentage of species in the region that are affected by the eight threats analysed in the study for the three species lists for a non-spatial analysis. Habitat loss is a threat to the greatest percentage of regional priority and nationally threatened species in the study region, and disease threatens the fewest species (Table 2.3). Three threats: introduced species, fire and fire suppression, and overexploitation, affect a high percentage of both regional and nationally threatened species in the region, and affect the second-, third-, and fourth-most species after the threat of habitat loss. Pollution, natural causes, and native species interactions are intermediate-level threats to species in the region.

2.4.2 Spatial extent of threats

The spatial extents of threats, in terms of the proportion of the study area affected, varied amongst the indirect threat maps derived from the two different data sources (Figure 2.2). Habitat loss was modelled as the most severe threat in the region in all four comparative scenarios, both in area and the proportion of species affected. Introduced species, fire and fire suppression, and native species interactions threats are also modelled as affecting almost the entire study area. Only when threats were modelled using the SDM dataset with its original data resolution did no threat cover greater than 70% of the region (Figure 2.2b). Disease was modelled as threatening the smallest area of the study region of the eight threats (Figure 2.2b-d), except when it was predicted using all nationally- threatened species and the geographic range map data source (Figure 2.2a).

Table 2.3 Percentage and number of species in the Burnett-Mary Natural Resource Management Region affected by eight threats, for three sets of species and data sources used in the indirect threat map comparison analysis. Rank (r) indicates order of highest to lowest percentage and number of species affected by threat, by list.

Nationally In-common Regional Priority Threatened (33 species) (65 species) (88 species) SDMs* and SDMs* Range maps** Range maps** Threat (%) (n) r (%) (n) r (%) (n) r Disease 7.7 (5) 8 6.8 (6) 8 6.1 (2) 8 Fire and fire suppression 53.8 (35) 3 45.5 (40) 3 51.5 (17) 3 Habitat loss 86.2 (56) 1 89.8 (79) 1 90.9 (30) 1 Introduced species 64.6 (42) 2 64.8 (57) 2 66.7 (22) 2 Native species interactions 15.4 (10) 5 9.1 (8) 6 24.2 (8) 5 Natural causes 10.8 (7) 7 11.4 (10) 5 18.2 (6) 6 Overexploitation 18.5 (12) 4 22.7 (20) 4 27.3 (9) 4 Pollution 12.3 (8) 6 6.8 (6) 7 12.1 (4) 7 *SDMs = predicted species distribution models (Appendix 2A) **Range maps = Species of National Environmental Significance range maps (Department of the Environment 2008a)

2.4.3 Spatial location of threats

Location and degree of threat vary when modelled by the SDMs as compared to the range maps (Figure 2.3). In general, habitat loss was modelled as threatening a high proportion of species across most of the region and disease was modelled as affecting a low proportion of species only in the east of region. The threats of fire and fire suppression, introduced species, and native species interactions were more concentrated in contiguous areas when modelled by the range maps as compared to the SDMs, where they were more patchily distributed.

2.4.4 Measures of indirect threat map comparisons

The disagreement between SDM- and range map-modelled indirect threat maps was greatest when the list of threatened species is different (Figure 2.4a), and least when threat level was aggregated for the same species (Figure 2.4d and Appendix 2A, Figure 2A. 1. The spatially restricted threat of disease and the broadly distributed threat of habitat loss were most consistently spatially similar in degree and location of threat (Figure 2.4a-d).

Threat % of Level Comparison study Indirect threat maps derived Indirect threat maps derived (% of Scenario area from SDMs* from Range maps** species) 1

List (a) Comparison

Direct (b) Comparison

Up-scaled (c) Comparison

Aggregation (d) Comparison

*SDMs = predicted species distribution models (Appendix 2A) **Range maps = Species of National Environmental Significance range maps (Department of the Environment 2008a)

Figure 2.2 Percentage of study area covered (Y-axis) by threat category (X-axis) for indirect threat map datasets referred to in Table 2.2. Colour of bar section indicates level of threat (intensity) as represented by percentage of total species affected by eight threats: disease, fire (and fire suppression), habloss (habitat loss), introduced (species), native (species interactions), natural (causes), overex (overexploitation), and pollution.

Disease Native species interactions

SDMs* Range maps** SDMs* Range maps** Fire and fire suppression Natural causes

SDMs* Range maps** SDMs* Range maps** Habitat loss Overexploitation

SDMs* Range maps** SDMs* Range maps** Introduced species Pollution

SDMs* Range maps** SDMs* Range maps** *SDMs = predicted species distribution models (Appendix 2A) **Range maps = Species of National Environmental Significance range maps (Department of the Environment 2008a)

Figure 2.3 Percentage of 33 nationally threatened species affected by eight threats, presented only for the up-scaled comparison scenario. Threats maps are represented by predicted SDMs as compared to range maps at the subcatchment scale for six threat-level classes based on proportion of species affected. Map Projection UTM Zone 56.

Comparison Correct Scenario Threat Disagreement (%) (%) Kappa

05 -0.06 1 24 -0.06 List 94 0.08 (a) 14 -0.04 Comparison 25 -0.01 31 -0.05 28 -0.03 58 0.22

92 0.24 38 0.08 2 70 0.06 Direct 27 0.10 (b) 24 0.08 Comparison 55 0.21 47 0.22 82 0.07

91 0.47 31 0.09 3 96 0.29 Up-scaled 45 0.21 (c) 33 0.10 Comparison 39 0.17 33 0.09 59 0.19

94 0.64 96 0.13 4 98 0.02 Aggregation 98 0.13 (d) 97 0.00 Comparison 70 0.40 64 0.24 66 0.28

Figure 2.4 Quantity (percentage of species threatened) and allocation (spatial extent of threat) disagreement summed for total disagreement, proportion correct, and Kappa index of agreement for indirect threat map datasets detailed in Table 2.2. Comparisons are between threats modelled using predicted SDMs as compared to using range maps. Threats analysed are: disease, fire (and fire suppression), habloss (habitat loss), introduced (species), native (species interactions), natural (causes), overex (overexploitation), and pollution.

In the list comparison (Figure 2.4a), only the spatial distribution of habitat loss was similar between threats modelled from the two data sources. The next lowest percentage of disagreement of threats

modelled (42%) was for the threat of pollution. Overall, percentage disagreements between modelled threats in the direct comparison (Figure 2.4b) were less than in the list comparison (Figure 2.4a). For most threats, the greatest proportion of disagreement was due to quantity disagreement: there was an imperfect match in the proportions of threat intensity classes compared (also see Figure 2.2b). Disease indirect threat maps were less dissimilar in the direct comparison (8% disagreement) than in the list comparison (95%). For most threats, disagreement between indirect threat maps in the up-scaled comparison (Figure 2.4c) was proportionally more due to spatial location (allocation disagreement) than differences in threat intensity (quantity disagreement).Overall, however, the agreement/disagreement proportions were similar when the direct comparison (Figure 2.4b) is contrasted with the up-scaled comparison (Figure 2.4c). When threat levels were aggregated to threat/no threat (aggregation comparison; Figure 2.4d), overall, indirect threat maps were more similar when evaluated against the other three comparisons (Figure 2.4a-c); the spatial distribution of five of the indirect threat maps derived from the two data sources were >90% in agreement.

Kappa index values (Figure 2.4) were relatively low for all comparisons between threats modelled by SDMs as compared to range maps, indicating a low level of agreement between the indirect threat maps when accounting for differences in location and degree of threat and corrected for what would be expected by pure chance. Disease was the only threat for which map comparisons exceeded a ‘poor’ agreement threshold of 0.40 (Monserud and Leemans 1992) (Figure 2.4c and Figure 2.4d). Overall, the Kappa index of agreement was higher when the same set of species was used to model the distribution of threats (Figure 2.4b-d). For the majority of comparisons (Figure 2.4), the kappa index values indicate agreement between maps was essentially no better than would be expected by pure chance (K < 0.25).

2.5 Discussion

Natural resource managers are most interested in areas where threats are co-located with species of concern, but threats to species are difficult to map at a fine scale. For example, direct threat maps of grazing, invasive fox, and fire threats for the Burnett-Mary NRM are mapped at a coarse resolution. Land use maps (Department of Environment and Resource Management (DERM) 1999) indicate the majority of the study region is used for production from relatively natural environments (grazing and agriculture), invasive foxes are likely to occur across virtually the entire study region (Department of the Environment 2008a), and fire management approaches are hotly contested across all of Australia (Bradstock et al. 2012). Using threatened species distributions to derive indirect threat maps provides the ability to simultaneously locate threat intensities associated with

species of conservation concern. Furthermore, we believe SDMs have a greater capacity to locate threat intensity in comparison to range maps, because suitable habitat conditions can be rigorously modelled at finer detail. This advantage comes at the cost of data intensive processing and modelling effort. Therefore it is reasonable, as we did in this research, to compare indirect threat map representations using predicted species distribution models and existing range maps, and we did this specifically in terms of listed species and scale.

At the regional scale, our analysis finds that the level and location of threats vary significantly according to the source of our data in the four scenarios in which we compare models of eight major threats to species. Indirect threat maps derived from data on threatened species distributions are sensitive to mapping scale and the list of species used in threat modelling. None of the map comparisons exceeded ‘poor’ Kappa index agreement (Monserud and Leemans 1992), the percentage of agreement between two maps corrected for what would be expected by pure chance (Visser and de Nijs 2006). As well, most comparisons were greater than 40% in disagreement (Pontius Jr. and Millones 2011) as to location of threats and threat intensity.

In comparing SDMs and range maps for representing regional spatial patterns of threats, we conclude that threat maps indirectly derived from SDMs are more useful for regional analysis and management planning. Especially for threats affecting range-restricted species, SDMs can be modelled at a resolution that better captures heterogeneity in distribution (La Sorte and Hawkins 2007). Non-spatial approaches for prioritising threat management (e.g. Carwardine et al. 2012) may be an alternative when resources cannot be dedicated to species distribution modelling to produce indirect threat maps for decision-making. We caution that indirect threat maps based on broad-scale geographic range maps may mislead efforts to prioritise threat management at a regional scale.

2.5.1 Advantages and disadvantages of SDMs and range maps for indirect threat mapping

An advantage of the predicted species distribution modelling process is that the method is reproducible and can be updated with new information. SDMs guide understanding of species distribution when there is imperfect knowledge and where there is lack of data such as systematic grid-based surveys (Hawkins et al. 2008). Finer-resolution models can provide managers a better indication of priority management areas for species of interest (Bombi et al. 2011, Gillingham et al. 2012), and may be modelled at an appropriate spatial scale to match intended conservation applications (Cabeza et al. 2010). It is also increasingly feasible to better address uncertainty in SDM modelling (Beale and Lennon 2012) and guidelines to better mapping are evolving (Benito et al. 2013). On the other hand, expert-derived geographic range maps may have more perceived credibility, and are more generally available than specific project-derived SDMs created from

dedicated resources. However, the often coarse representations of range maps may make them more appropriate for large-scale analyses (Rondinini et al. 2006).

The spatial distribution of species richness has been found to be highly dependent upon measurement resolution (Stoms 1994) and species richness overestimation by range maps decreases at coarser spatial resolutions (Hurlbert and Jetz 2007); indirect threat maps would be expected to follow similar patterns. Conservation decision-making that uses mapped data must carefully consider how uncertainty in data and models can affect conservation outcomes, and may require additional data collection including deliberate sampling and experimentation at multiple scales to assure predictive reliability (Conroy and Noon 1996). This is important because overestimation of threatened species distribution may produce false optimism at best (Wilson et al. 2004, Jetz et al. 2008), and misplaced guidance at worst (Grand et al. 2007), with the risk of leading to a focus on non-representative locations for conservation effort (Rondinini et al. 2006). Broad-scale conservation planning analyses cannot be assumed to predict fine-scale results (Shriner et al. 2006), although analyses based on coarse-scale data may be useful as a rapid assessment method for preliminary identification of important areas for conservation at finer scales (Larsen and Rahbek 2003). Threats modelled from range maps at the national scale do provide a broad picture of the spatial distribution of threats to Australian species without investing resources in finely-scaled SDMs (Evans et al. 2011a).

Threat maps cannot inform us whether areas of higher or lower risk should be prioritised for conservation action (Game et al. 2008). Other modelling techniques map how threats affect species distributions, such as for projections under future climate scenarios (Thomas et al. 2004) or increased fire frequency (Reside et al. 2012). Multi-modelling methods are an improvement on static SDMs (as used in this study) to support conservation biogeography (Franklin 2010). Threat modelling with this method may be sensitive in areas with low species richness (Evans et al. 2011a), and the number of species modelled in this study may be too few to make specific statements about broad threats. Nevertheless, we assume that the spatial extent of summed geographical distributions of species affected by threats represents an indication of threat location.

2.5.2 Recommendations and cautions

Our research does not determine whether it is more appropriate to use SDMs or range maps for threat analysis, the choice will depend on the particular conservation objective at hand. We do find that indirect threat maps differ, depending on species distribution data source used. At the regional level, we conclude that SDMs can be useful for modelling spatial representations of the degree and location of threats to species, although some threats may be better modelled using this technique

than others. It may be most important to invest resources in mapping fine-scale distributions of threats which only affect a few species with restricted habitat, and/or those which are less tightly linked to other types of more easily accessed spatial information. For example, land use is a clear predictor of habitat loss, but the threat of disease may warrant greater investment in better mapping its distribution. We concur with Boitani et al. (2012) in advocating for an increased openness of datasets and the creation of a dynamic database of species distribution maps for threat analysis and setting spatial priorities for action. With expert input, threat maps could inform policy and planning; researchers with modelling expertise could be connected with policymakers, to inform their decision-making (Guisan et al. 2013). On the other hand, due to its ease of use, species distribution modelling is prone to uncritical and sometimes inappropriate application. For example models may suffer from sample selection bias (Yackulic et al. 2012), despite coping strategies that improve predictive accuracy (Phillips and Dudík 2008). However, practitioners may have limited awareness of modelling requirements that are outside of established techniques, even though diagnostic tools for checking model assumptions are improving (Renner and Warton 2012). In addition, understanding of the spatial patterns of species and their threats still may not be enough to decide where to implement mitigating action (Evans et al. 2011b). Furthermore, a careful return-on- investment analysis (Grantham et al. 2008) may reveal limits to the benefits of modelling SDMs or more targeted research could further show where coarse range maps are sufficiently informative for threat mapping. Finally, non-spatial prioritisations based on expert knowledge and feasibility of threat management action can also be a viable alternative to threat mapping (Joseph et al. 2009,

Carwardine et al. 2012, Pannell et al. 2012).

2.5.3 Conclusion

In conclusion, we believe that resource managers would benefit from investing in models of the distribution of priority species for threat analysis at the regional scale, but this decision is based on the availability of resources to do so. Used to model the spatial distribution of location and intensity of threat based upon proportion of species affected by threat, species distribution models provide refined detail, and can enhance what can be represented by geographic range maps, because they are based on habitat suitability (Rondinini et al. 2011). Threat mapping using species distributions and their threats leads logically to decisions about actions to address those threats, but regional decisions on threat management based on coarse range maps may be misleading, especially for threats affecting species with restricted distributions. When directly mapping threats is not easy to do, using indirect threat maps based upon fine-scale SDMs can inform threat management for regionally-threatened species, if used cautiously and with consideration of data source advantages and disadvantages. 42

2.6 Acknowledgements

This research was conducted with the support of funding from the Australian Government’s National Environmental Research Program, the Australian Research Council Centre of Excellence for Environmental Decisions, and an Australia Postgraduate Award scholarship. The Queensland Government provided species presence data (Wildnet (Department of Environment and Resource Management (DERM) 2010) and HERBRECS (Department of Environment and Resource Management (DERM) 2011) databases). We thank Ashley Bunce, Dirk Hovorka, Rachel Lyons, Ramona Maggini, and Roslyn Taplin for discussions leading up to the research, and Richard Pearson and Steven Phillips for their 2011 AMNH SDM workshop (Pearson 2007).

2.7 Supplementary Information

2.7.1 Appendix 2A Regional priority threatened species distribution modelling methods and model performance validation and assessment

2.7.1.1 Species presence records and environmental variables

Species presence records for species distribution modelling were provided by Queensland Wildnet (Department of Environment and Resource Management (DERM) 2010) and the HERBRECS (Department of Environment and Resource Management (DERM) 2011) databases. Twelve environmental variables were used in the final species distribution modelling. An initial variable selection was evaluated in a pair-wise correlation matrix, and to reduce collinearity, those variables correlated with the highest number of other variables were eliminated systematically until no variables were correlated above a Pearson correlation coefficient threshold of 0.75 (after Carvalho et al. 2011a). Climatic variables derived from monthly temperature and rainfall data represent the biologically meaningful parameters (Hijmans et al. 2005) of precipitation seasonality (1) isothermality (2), and the extreme/limiting variables of mean temperature of coldest (3) and wettest (4) quarters, and precipitation of the driest month (5) and wettest quarter (6). Aspect (7) and slope (8) were derived from a Digital Elevation Model (Geoscience Australia 2010a, b); aspect data were reclassified into eight categorical variables (north, northeast, east, etc.) to avoid problems with radially distributed values. Euclidean distance to the nearest water source (9) (Geoscience Australia 2010c) was derived using ESRI ArcInfo version 10 spatial analyst for order 4+ (Strahler 1957) streams. Geological variables used were dominant lithological rock unit (10) (DME 2008), and dominant soil order (11) (Isbell 2002, DERM 2010). The dominant remnant regional ecosystem (12) (Neldner et al. 2005) of pre-clearing vegetation (Queensland Herbarium 2010a) was used to

analyse the species environment in Maxent, then distribution models were “projected” in Maxent to the species environment of current and regrowth (2006) vegetation cover (Queensland Herbarium 2010a, b) to reflect potential habitat in areas mapped as native vegetation. Hinge and categorical features (Phillips and Dudík 2008) were used to model the relationships between species presence records and environmental variables to avoid over-fitting so that the model does not fit the training data so well that it does not generalise (Elith and Leathwick 2009).

2.7.1.2 Model predictive performance

Species presence data were partitioned into training and testing data for cross-validation using either Jackknife validation (Pearson et al. 2007) or K-fold partitioning (Fielding and Bell 1997) for species with n < 20 or n ≥ 20 presence records, respectively. The mean value of the AUC was calculated from the testing run results, and 95% confidence limits were calculated using the t- distribution. Species distribution models were accepted only if the lower 95% confidence limit was greater than 0.6 (after Kremen et al. 2008, supplement). For those species models that passed predictive performance validation (using K-fold or jackknife (p < .05) partitioning), all sample data were used to develop a robust classification rule for the final multivariate models as suggested by (Rencher 1995) in (Fielding and Bell 1997).

The SDMs were individually delineated for suitable and non-suitable habitat based on the Lowest Presence Threshold (LPT) (Pearson et al. 2007) of each model. This threshold was used for the following reasons. The failure to identify rare species habitat could have devastating consequences for at-risk populations (Pearce and Ferrier 2000), and by minimising omission (false negative) error, the possibility of not identifying areas of potentially occupied habitat is reduced. Alternatively, there is danger in identifying suitable habitat areas that are actually unsuitable, i.e. overestimating suitability through commission error, if management is targeted either in areas that are unsuitable for the species of concern or in areas chosen for complementarity that are risky because apparent high species richness occurs because of overlap at range peripheries (Loiselle et al. 2003). A conservative approach is to identify the minimum predicted area possible while also minimising omission error of the training (species presence) data. The LPT is defined by the lowest probability value assigned to the location of presence records used to define the model, and can be ecologically interpreted as the identification of areas that are, at a minimum, as suitable as those sites where a species was recorded as being present (Pearson et al. 2007). A training test omission rate of 0.0 was achieved using a minimum training presence logistic threshold (equivalent to LPT) calculated by Maxent to delineate suitable and unsuitable habitat from logistic probability output maps.

Sixty-five SDMs (of 145 priority species) were accepted for use in the analysis. Habitat for four species (Elseya albagula, Elusor macrurus, Neoceratodus forsteri, and Maccullochella mariensis) was restricted to known inhabited watersheds; otherwise SDM post-processing was not customised for individual species. Data constraints can limit modelling efforts, and reasons why species distributions were unable to be modelled for inclusion in the threat analysis include too few presence records (28 species with zero records, 13 species with one record), models for 29 species with less than seven records were not statistically significant, and nine species with acceptable models did not have documented threats. Whereas it is evident that uncertainty exists in niche-based models (Beale and Lennon 2012), standard assessment of model performance was addressed (Fielding and Bell 1997, Pearson et al. 2007, Kremen et al. 2008) as discussed above, and only models that passed predictive performance validation were used in the analysis. The quality of SDMs is necessarily dependent upon appropriate data, which can be difficult to obtain in many situations (Beale and Lennon 2012), as well as suitably scaled environmental data (Kriticos and Leriche 2010).

Table 2A. 1 One-hundred twenty-nine species were included in the spatial threat analysis. SDMs were derived for 65 regional priority (threatened) species, range maps were available for 88 nationally threatened species, and there were 33 nationally threatened species in common for which both types of distribution maps were available.

Range In- SDMs maps common Taxa Species (65) (88) (33) Amphibians Crinia tinnula x Litoria cooloolensis x Litoria freycineti x Litoria olongburensis x x x Litoria pearsoniana x Litoria pearsoniana (Kroombit Tops) x Mixophyes fleayi x Mixophyes iteratus x x x Taudactylus pleione x Birds Calyptorhynchus lathami x Cyclopsitta diophthalma coxeni x Dasyornis brachypterus x x x Erythrotriorchis radiatus x Geophaps scripta scripta x Grantiella picta x Neochmia ruficauda ruficauda x

Table 2A.1 Continued.

Range In- SDMs maps common Taxa Species (65) (88) (33) Birds cont. Ninox strenua x Pezoporus wallicus wallicus x Turnix melanogaster x x x Fish Maccullochella mariensis x x x Nannoperca oxleyana x x x Neoceratodus forsteri x x x Pseudomugil mellis x Rhadinocentrus ornatus x Insects Phyllodes imperialis (s ssp. - ANIC 3333) x Mammals Kerivoula papuensis x Petaurus australis australis x Pseudomys patrius x Pteropus poliocephalus x x x Xeromys myoides x x x Reptiles Acanthophis antarcticus x Caretta caretta x Chelonia mydas x Denisonia maculata x Egernia rugosa x Elseya albagula x Elusor macrurus x x x Hoplocephalus stephensii x Lampropholis colossus x Nangura spinosa x x x Paradelma orientalis x Strophurus taenicauda x Vascular plants Acacia attenuata x x x Acacia baueri subsp. baueri x Acacia eremophiloides x x x Acacia grandifolia x Acacia porcata x x x Acacia tingoorensis x Acronychia littoralis x Alectryon ramiflorus x x x Archidendron lovelliae x Arthraxon hispidus x Baloghia marmorata x Bertya granitica x

Table 2A.1 Continued.

Range In- SDMs maps common Taxa Species (65) (88) (33) Vascular plants cont. Boronia keysii x Bothriochloa bunyensis x x x Brunoniella spiciflora x Cadellia pentastylis x Callitris baileyi x Calytrix gurulmundensis x Clematis fawcettii x Coopernookia scabridiuscula x Corynocarpus rupestris subsp. arborescens x Cossinia australiana x x x Cryptocarya foetida x Cupaniopsis shirleyana x x x Cycas megacarpa x x x Daviesia discolor x Denhamia parvifolia x x x Dichanthium queenslandicum x Digitaria porrecta x Diuris sheaffiana x Eucalyptus beaniana x Eucalyptus conglomerata x Eucalyptus hallii x x x Eucalyptus virens x Floydia praealta x x x Fontainea rostrata x x x Fontainea venosa x Germainia capitata x Graptophyllum reticulatum x Grevillea venusta x Haloragis exalata subsp. velutina x Homoranthus decumbens x Lasiopetalum sp. (Proston JA Baker 17) x Lepidium hyssopifolium x Lepidium peregrinum x Macadamia integrifolia x Macadamia jansenii x x x Macadamia ternifolia x x x Macadamia tetraphylla x Macrozamia crassifolia x Macrozamia lomandroides x x x Macrozamia parcifolia x x x Macrozamia pauli-guilielmi x x x Medicosma elliptica x x x

Table 2A.1 Concluded.

Range In- SDMs maps common Taxa Species (65) (88) (33) Vascular plants cont. Newcastelia velutina x Pararistolochia praevenosa x Paspalidium grandispiculatum x Phaius australis x Phebalium distans x Plectranthus omissus x Plectranthus torrenticola x Pomaderris clivicola x Pouteria eerwah x Prasophyllum wallum x Pterostylis cobarensis x Quassia bidwillii x x x Romnalda strobilacea x x x Sarcochilus fitzgeraldii x Sarcochilus roseus x Sarcochilus weinthalii x Sophora fraseri x Stemmacantha australis x Syzygium hodgkinsoniae x Thesium australe x Triunia robusta x x x Trymalium minutiflorum x Xanthostemon oppositifolius x Zieria verrucosa x x x

Disease Fire and Habitat loss Introduced fire suppression species

Native species Natural causes Overexploitation Pollution interactions

Figure 2A. 1. Spatial difference between threats represented by SDMs as compared to range maps is shown for aggregation comparison (same 33 species, subcatchment data resolution, and aggregated threat levels). Colours indicate where threat maps are in agreement or in disagreement. Map Projection UTM Zone 56.

Chapter 3 Informed actions: Where to cost-effectively manage multiple threats to species to maximise return on investment

3 Informed actions: Where to cost-effectively manage multiple threats to species to maximise return on investment

3.1 Abstract

Conservation practitioners, faced with managing multiple threats to biodiversity and limited funding, must prioritise investment in different management actions. From an economic perspective, it is routine practice to invest where the highest rate of return is expected. This return- on-investment (ROI) thinking can also benefit species conservation, and researchers are developing sophisticated approaches to support decision-making for cost-effective conservation. However, applied use of these approaches is limited. Managers may be wary of ‘black-box’ algorithms or complex methods that are difficult to explain to funding agencies. As an alternative, we demonstrate the use of a basic ROI analysis for determining where to invest in cost-effective management to address threats to species. This method can be applied using basic geographic information system and spread sheet calculations. We illustrate the approach in a management-action prioritisation for a bio-diverse region of eastern Australia. We use ROI to prioritise management actions for two threats to a suite of threatened species: habitat degradation by cattle grazing and predation by invasive red foxes (Vulpes vulpes). We show how decisions based on cost-effective threat management depend upon how expected benefits to species are defined and how benefits and costs co-vary. By considering a combination of species richness, restricted habitats, species vulnerability, and costs of management actions, small investments can result in greater expected benefit compared with management decisions that consider only species richness. Furthermore, a landscape management strategy that implements multiple actions is more efficient than managing only for one threat or more traditional approaches that don’t consider ROI. Our approach provides transparent and logical decision support for prioritising different actions intended to abate threats associated with multiple species; it is of use when managers need a justifiable and repeatable approach to investment.

Key words: Australia; biodiversity; cost-effectiveness; decision support; natural resource management; multiple species; multiple threats; regional-scale; return on investment; species richness; spatial prioritisation; threatened species.

3.2 Introduction

Species are imperilled by numerous threats (Wilson 1992); abating these threats requires decisive management action. As a conservation practitioner, it is difficult to decide which management action to enact first and where to enact it (Marris 2007), especially given constraints of time (Pimm et al. 1995) and money (James et al. 1999). One choice is to act in the sites that may maximise the number of species protected against impending threats to provide the greatest benefit (Brooks et al. 2006, Evans et al. 2011b). However, this approach ignores the cost of management, and it might lead to managing expensive sites and missing out on cheaper sites that might still benefit many species (Balmford et al. 2000). By incorporating economic costs into conservation decision-making (Naidoo et al. 2006), more species or locations may benefit from management actions (Possingham et al. 2001b, Joseph et al. 2009, Carwardine et al. 2012). Although conservation practitioners are already likely to tacitly consider costs in their decisions, it is difficult to mentally consider the trade- offs in such complex decisions. On the other hand, if a manager has the explicit goal of maximising species benefits from management actions for abating threats at least cost, the dilemma of where to act first can be resolved by placing the problem in a conservation decision theory framework (Possingham 2001a). By using conservation return on investment (ROI) (Murdoch et al. 2007) to reveal which actions are most cost-effective, a manager can make an informed decision to transparently prioritise the necessary actions expected to provide the greatest benefit to species (Bottrill et al. 2008).

The difficulty of deciding how and where to manage multiple threats has led to a number of alternative approaches being developed. Much of the traditional systematic conservation planning literature on threat management addresses only one action at a time for multiple species, such as targeting the threat of habitat loss with reserves (Margules and Pressey 2000). In contrast, economic approaches to spatial prioritisation of conservation (Busch and Cullen 2009, Walsh et al. 2012) generally focus on single-species management with more than one action. Conservation effort allocation needs to be strategic to most efficiently mitigate the multiple threats acting in the landscape. Efficiency in this context is related to the spatial distribution of threats, how difficult the threats are to abate, and how many species might benefit from threat mitigation (Klein et al. 2010, Evans et al. 2011b). Yet doing everything, everywhere, isn’t an option. Simultaneously considering the costs and benefits of multiple threat-management actions leads to an understanding of where to expect the greatest return to species (Wilson et al. 2007, Wilson et al. 2011a). Nevertheless, on-the- ground, fine-scale spatial prioritisation of multiple actions is still not routine (van Teeffelen and Moilanen 2008).

Cost-effectiveness analysis (CEA) is a form of ROI that is commonly used to evaluate human health care options relative to costs (Weinstein and Stason 1977). CEA is also applicable to managers making decisions for investing in single or multiple conservation actions under resource constraints (Hughey et al. 2003, Murdoch et al. 2007). In both cases, benefits are measured in non- monetary units. ROI provides guidance to managers and practitioners by clarifying where the highest rate of conservation return, i.e. the greatest benefit, is expected from an investment in one or more conservation actions. In the ROI framework, managers explicitly specify what outcome they want to achieve (i.e. an objective), define the benefits they expect from conservation action, and estimate the costs it would take to achieve them (Possingham et al. 2001b, Mace et al. 2006).

Research surrounding the use of ROI to explore trade-offs in prioritising conservation efforts is progressing rapidly. Conservation ROI is becoming increasingly complex, by integrating important system considerations. ROI informs efficient land acquisition while achieving species complementarity (Ando et al. 1998, Balmford et al. 2000, Underwood et al. 2008). Additionally incorporating considerations of land conversion and on-going habitat loss threats (Wilson et al. 2006, Withey et al. 2012), off-reserve lands (Polasky et al. 2005) and alternative land uses (Wilson et al. 2010) refines the analysis, as does considering additional actions such as policy incentives (Nelson et al. 2008, Lewis et al. 2011), habitat restoration (Goldstein et al. 2008, Wilson et al. 2011b), and abatement of threats such as invasive species (Wilson et al. 2007). Moreover, multi- objective ROI analyses reveal the contributions of desirable ecosystem services (Daily 1997) such as carbon sequestration and freshwater filtration (Kovacs et al. 2013, Kramer et al. 2013) in addition to the benefits of habitat conservation. These conservation ROI analyses provide informative outcomes in directing management by incorporating increasingly complex information, algorithms, and important dependencies.

However, although the conservation ROI research provides compelling evidence that the approach directs where best to efficiently invest in conservation action, there is less evidence that the ROI approach is being used extensively by conservation practitioners (Murdoch et al. 2007, Boyd et al. 2012). While ROI is also routinely used as a business performance indicator to provide strategic direction (e.g. Venkatraman and Ramanujam 1986), its surprising lack of use to rank and prioritise on-ground conservation action indicates a knowing-doing gap, whereby research findings are not widely accepted and applied by practitioners (Knight et al. 2008). A key limitation to the effectiveness of a ranking system such as ROI is the gap between its availability as a decision- support tool, and its implementation by decision makers (Cullen and White 2013). Developing more sophisticated ROI approaches is undeniably valuable, but the perceived complexity of the analytics could make the solutions less approachable to practitioners. As an alternative, our research seeks to 55

make ROI accessible as a decision-support tool to a manager who makes day-to-day decisions about implementing conservation action at a regional or local scale.

Here we demonstrate an ROI approach as decision support for a conservation practitioner/manager with the conservation objective of recovering threatened species through threat mitigation actions. Time and resource constraints also apply to managers making the effort to gain additional expertise in prioritisation techniques, and those responsible for biodiversity management need quick, transparent, and easy-to-understand decision-support tools that are useful at a fine scale. The calculation logic should be straightforward, so that managers can satisfy stakeholders (donors, funding agencies and the general public) that important decisions are being made in a defensible manner. Using systematic planning software to prioritise spatial conservation actions at least cost or without a budget constraint (e.g. Marxan and Zonation, Ball et al. 2009, Moilanen et al. 2009a) is a widely respected approach. On the other hand, practitioners are still sometimes wary of what are perceived to be complex “black-box” algorithms (e.g. simulated annealing in Marxan), although their use is grounded in quantitative decision theory that provides a solid basis for decision-support applications (Possingham 2001b). Similarly, increasingly complex ROI approaches can be difficult to perform and interpret by practitioners.

Our research provides an alternative both to decision-support solutions that are perceived to be complex, as well as the default position of not using decision support. The approach is a simplified, but rigorous and systematic, way to use ROI to prioritise conservation actions that considers multiple species and multiple threats. We promote a basic ROI based on costs and benefits because it is a straightforward concept to understand, relate to, and communicate to stakeholders and donors. The approach requires only working knowledge of commonly-used technology (a geographic information system and spread sheets) yet provides informative guidance for decision-making by allowing decision makers to explore and easily visualise how uncertainty and variation in costs or benefits might drive different decisions. Due to the ease in which we can visualise what drives investment decisions, our approach makes the strength of ROI accessible to practitioners that are hesitant to implement complex approaches, yet need a transparent way to account for the way they invest limited funding.

3.2.1 Research aim

We describe an ROI decision-making framework to guide natural resource managers in achieving efficient allocation of conservation resources to address multiple threats to multiple threatened species. We also show the difference in decisions made with ROI, and without ROI. We use a case study of a bio-diverse natural resource management region in Australia that is typical of any

administratively-defined location, worldwide, where a manager must make decisions about where to focus action to abate threats to species. Our research informs a conservation practitioner on using a straightforward ROI analysis to evaluate the cost-effectiveness of management action to benefit species at least cost. We model ROI using the relationship between the level of investment in a given management action, and the expected conservation outcome, i.e. threat abatement to species affected. We compare our results to those of alternative approaches to decision-making. Specifically, we explore three questions to guide management action strategies across space:

1) How much better is an ROI approach when compared to more traditional conservation planning approaches of allocating resources or arbitrary action? 2) How do alternative ways of defining the expected benefits of conservation management change the spatial priorities and level of investment required to manage single or multiple threats in an ROI approach? 3) If managers can choose single or multiple management actions across the landscape depending on where actions are most cost-effective how can ROI inform whether one or more action(s) are allocated to an area for a given budget?

We illustrate how to implement our ROI approach using a case study of a regional natural resource management area where multiple species would benefit from actions to abate multiple threats to their persistence. Our goal is to show that an uncomplicated ROI analysis can provide useful decision support for conservation managers.

3.3 Methods

3.3.1 Case Study

To illustrate our framework of using ROI to make investment decisions to abate multiple threats, we use an example from the Burnett-Mary Natural Resource Management (NRM) Region, which is located in southeast Queensland, Australia. Administrative boundaries of Australia’s 56 NRM regions are ecologically-defined bioregions or watershed catchments. Encompassing the catchments of the Burnett and Mary Rivers, the case study region covers 55,000 km2. The region contains diverse ecosystems and has a history of land use decisions that threaten its biodiversity; for example, the now presumed extinct Paradise Parrot (Psephotus pulcherrimus) was last sited in the study area in the 1920s (Olsen 2007b). The regional managers are under pressure to successfully cope with a range of threats to priority threatened species (Department of Environment and Resource Management 2010), yet there are limited financial and technical resources for strengthening management practice (Robins and Dovers 2007) such as in prioritising management 57

action. We divided the region into a grid of 25-ha potential management sites for a fine-grain analysis of species’ habitat and management requirements. We analysed 129,894 sites, those with remaining or regrowth native vegetation (Queensland Herbarium 2010a, b), that are presumed to provide habitat for species in the region.

3.3.2 Framework for conservation return-on-investment analysis

We conducted ROI analysis to measure the increase in conservation outcome per unit cost of management actions taken to reduce threats in the study region. The ROI produces a measure of conservation cost-efficiency (Murdoch et al. 2007). In our case, we define management outcomes, and therefore conservation efficiency, to be abating a given threat to the species that are affected by that threat. Critical steps to applying ROI to a conservation decision framework (Possingham et al. 2001b) are: 1) identifying the problem, 2) defining realistic expected benefits, 3) integrating realistic costs, and 4) combining information on expected benefits and costs to solve the management funding allocation problem.

3.3.3 Step 1. Problem definition

Our problem is determining how and where to cost-effectively mitigate threats to species in a given landscape. We aim to find the most cost-effective management strategy to secure selected species (Appendix 3A) across a typical natural resource management region by addressing two expert- identified threats to those species (Department of Environment and Resource Management 2010): invasive species (fox) predation and habitat degradation (from cattle grazing). The red fox is recognised as one of the world’s worst invasive alien species (Lowe et al. 2001), and there is overwhelming evidence that invasive predators have negative effects on a broad range of native vertebrates in many parts of the world, including Australia (Mack et al. 2000, Burbidge and Manly 2002, Blackburn et al. 2004, Saunders et al. 2010). Red foxes are listed as a Key Threatening Process in Australia by the Environment Protection and Biodiversity Conservation (EPBC) Act 1999, with management actions such as bait poisoning addressed in a national Threat Abatement Plan (Anonymous 1999). Unsustainable grazing is also known to negatively affect the species composition, function, and structure of ecosystems (Fleischner 1994). To mitigate this threat, introduced herbivores are removed or reduced, and can result in increased species richness and abundance of small mammals (Legge et al. 2011), although recovery is likely to be dependent upon time, ecosystem type and extent of habitat alteration (Read and Cunningham 2010).

We set the goal for solving our problem to be: For a given level of investment, maximise the net expected benefit of the actions we take to mitigate threats that imperil a set of target species. As a

frame of reference, we also want to examine the full potential benefit of action if management funding was not limited.

3.3.4 Step 2. Define realistic estimates of expected benefits

Our 20 target species (Appendix 3A; Table 3A. 1) were previously selected by experts for management action on the criteria of probability and consequences of extinction, and potential for affecting successful recovery (Department of Environment and Resource Management 2010). Species include seven vascular plants, one fish, four reptiles, one amphibian, three mammals, and four birds that are threatened at the national, state, and/or regional level. We obtained species’ spatial distribution data across the study area from either of three sources, based on the best data availability for a species (Appendix 3A; Table 3A. 1): 1) species distribution models (SDMs) modelled (Auerbach et al. In prep-b) using Maxent software (Phillips et al. 2006, Dudík et al. 2010); 2) Species of National Environmental Significance (SNES) range maps (Department of the Environment 2008a) or 3) Atlas of Living Australia (ALA) point locations (Atlas of Living Australia (ALA) 2012).

To model the distribution of fox and grazing threats across the landscape, we used indirect threat modelling (see below), i.e. the spatial distribution of a particular threat is represented by the collective distributions of species affected by that threat (Evans et al. 2011a). For the two threats we considered, seven of the selected species are only vulnerable to predation by the red fox, eleven are only vulnerable to habitat degradation caused by cattle grazing, and two species are susceptible to both threats (Department of Environment and Resource Management 2010) (Appendix 3A; Table 3A. 1).

We need to define a measure for the expected benefits to species that captures the management goal (Guikema and Milke 1999). Here, we explore defining biodiversity benefits with the primary goal of threat management. For example, are managers most interested in prioritising threat management in areas with many species (high species richness)? Or do managers also want to target species with rare or vulnerable habitats? Alternative choices will lead to different prioritisations of management actions (Nicholson and Possingham 2006). The traditional approach to systematic conservation planning for reserve selection often uses data on the distribution of species, i.e. species richness, to inform decisions. However, to link these distributions explicitly with actions for threat management, a manager might also want to account for restricted habitats and/or vulnerability to multiple threats, as areas with species more likely to respond to targeted threat management might be considered higher priority for management.

We assume: 59

1) that the region is made up of m sites labelled 푖 = 1, … , 푚 2) that the region contains n threatened species labelled 푗 = 1, … , 푛 and 3) that there are p threats labelled 푘 = 1, … , 푝.

In the first instance let the expected benefit of acting to abate threat k to species j at site i (푏푖푗푘) be a value of one if a species is vulnerable to a threat k, and a value of zero if not. Our indirect threat mapping approach to species’ threat vulnerability (Wilson et al. 2005b) means that if a species is known to be susceptible to a threat, the species is assumed to be vulnerable across all of its mapped habitat. If all values are the same (in this example, a value of one), this vulnerability is portrayed as being spatially homogeneous. However, other values between zero and one could be used to represent spatial heterogeneity of a threat, variability in the vulnerability of a species to a threat, or uncertainty in the benefit of managing for a threat. For simplicity to demonstrate our approach, we make the assumption that a threat to a species is abated if managed (similar to Murdoch et al. (2007)), and the species will receive full benefit. This allows additive calculation of benefits to multiple species. Here, we compare three different ways to define expected benefits that depend on how much detail about the species and their threat is included, using data of increasing levels of complexity. The first defines benefit at a site to be the number of species (species richness) that will be secured by a particular threat abatement action (Metric 1), with the benefit 퐵푖푘of abating threat k at site i for n species formally described as follows:

푛 퐵푖푘 = ∑푗=1 푏푖푗푘 푎푖푗 (3-1) where 푎푖푗 represents the presence of a species j in site i, 푎푖푗 ∈ {0,1} with a value of one indicating species presence; and 푏푖푗푘 ∈ {0,1}, with a value of one indicating that species j is vulnerable to the threat and would benefit from management.

The second metric considers habitat rarity in addition to species richness (Metric 2), with the benefit

퐵′푖푘 of abating threat k at site i for n species calculated as:

푛 푎푖푗 퐵′푖푘 = ∑푗=1 푏푖푗푘 (3-2) 퐴푗 where the total number of occupied sites of each species 푗 is given by:

푚 퐴푗 = ∑푖=1 푎푖푗 (3-3)

This means that the benefit of acting at a site for a species is proportional to the fraction of the range of a species that the site encompasses. If it is the entire range of a species the benefit is large, whereas if it is a small part of the range of a species, the benefit is relatively small.

The third metric adjusts Metric 2 for species richness and habitat rarity by accounting for threat vulnerability (Metric 3), as we may want to manage in less risky areas (Game et al. 2008). Assume that there are multiple threats K at site i and a species receives only partial benefit from a management action that addresses only one threat. Assume that if we abate a threat k at a site, the management action secures the species in that site only proportionally to the number of threats to that species, and the benefit is proportional to the fraction of the range of the species. Here, the benefit 퐵′′푖푘 of abating threat k at site i for n species is:

푛 푏푖푗푘 푎푖푗 퐵′′푖푘 = ∑푗=1 . (3-4) 퐾푗 퐴푗

푏푖푗푘 where Kj is the total number of threats to species j, and weights the benefit by the proportion of 퐾푗 threats acting on the species across its range in the study region. For acting on one threat for a 푏 species vulnerable to only one threat, the value of 푖푗푘 is one, whereas acting on one threat for a 퐾푗 species vulnerable to two threats will result in a value of ½ = 0.5. For our two-action case study,

퐾푗 ∈ {1,2}.

We applied each metric to estimate the cumulative benefit 퐵푖푘 of abating threat k for all species j that occur at site i to examine benefit metrics of increasing complexity. In each metric we assume that each management action is equally effective at abating a specific threat, i.e. we assume that in each case we have perfect control over a given threat, in which case the decision is which basket of controls or threat abatement actions to employ.

3.3.5 Step 3. Integrate realistic cost estimates

Threat abatement comes at a cost, and management and opportunity costs are two specific types of conservation expenditures (Naidoo et al. 2006). Heterogeneity in costs across a landscape has been found to substantially affect results of systematic conservation planning (Ando et al. 1998, Polasky et al. 2001, Naidoo et al. 2006), and there is an increasingly large body of literature that describes how to calculate costs of management over space (e.g. Balmford et al. 2003, Naidoo et al. 2006, Polasky et al. 2008, Moilanen et al. 2011b, Carwardine et al. 2012, McCarthy et al. 2012). We calculated the cost of two possible management actions across the landscape over a 20-year period: removing grazing (an opportunity cost of foregone agricultural profits, e.g., Naidoo and Adamowicz (2006)), and controlling an invasive species (an on-ground management cost, e.g., Wilson et al. (2007)).

We estimated the opportunity cost to cease grazing through a stewardship agreement, based upon foregone agricultural profits (Marinoni et al. 2012) similar to other conservation planning studies 61

(e.g. Naidoo and Iwamura 2007, Carwardine et al. 2008), although landowner bargaining power creates some uncertainty in this cost assumption (Lennox and Armsworth 2013, Lennox et al. 2013). We assumed landowners would accept a financial stewardship payment to cease grazing on land populated with threatened species, to mitigate income volatility (Mouysset et al. 2013); financial security is a known influence on landholders decisions to change existing practices (Pannell et al. 2006). The dominant land use (approximately 65%) in the study region is grazing of natural vegetation (Department of Environment and Resource Management (DERM) 1999). To calculate a stewardship payment, we used a snapshot of agricultural profitability for the year 2005/2006 in Australia (Marinoni et al. 2012), which was based upon costs, revenues, yields and commodity (including livestock production) areas that were derived from land use, detailed census information, water resource use and production costs. Where the overall cost of production outweighed the returns and resulted in negative profits, we set the yearly profit value to a minimum of $10/25 ha to simulate providing landholders with a token payment. The stewardship payment was calculated to be equal to the net present value (NPV) of the opportunity cost of foregone agriculture profits over 20 years at a discount rate of 2% (Appendix 3B).

We estimated the cost of fox control using bait poisoning with sodium fluoroacetate (1080), a substance that naturally occurs in some native Australian vegetation and is used to control non- native predators (Saunders and McLeod 2007). We calculated the cost of reducing fox effects on native fauna with four 14-day baiting campaigns per year modelled as a roadside/grid baiting strategy after Carter et al. (2011). Costs may be higher if longer campaigns are needed for better outcomes. We accounted for labour, transportation, and bait price in the calculation of expenses for this strategy, and calculated the costs to be equal to the NPV of fox-baiting over a period of 20 years at a discount rate of 2% (Appendix 3C).

The cost of abating one threat k at site i is 퐶푖푘. For the action of abating the threat of overgrazing,

퐶푖푘 is the opportunity cost to cease grazing through a stewardship agreement, and for the action of abating the threat of foxes, 퐶푖푘 is the cost of predator control.

3.3.6 Step 4. Solve the problem

Solving the problem of how to maximise the expected benefits of managing target species under the constraint of a given budget is a type of ‘knapsack problem,’ which is a mathematical formulation in combinatorial optimisation that maximises an objective function subject to a single resource constraint (Pisinger and Toth 1998). In other words, here we seek to optimise the allocation of resources to threat management under financial limitations, as in other research addressing the prioritisation of management actions given a restricted conservation budget (e.g. Joseph et al 2009). 62

3.3.6.1 Comparing expected benefits in an ROI approach

To examine how different ways of calculating benefits affect decision-making, we first calculated the cost-effectiveness 푐푒푖푘 of abating threat k, for each site, i, represented by the benefit of acting to mitigate the threat k for all species, divided by the cost of doing this:

퐵푖푘 푐푒푖푘 = ⁄ (3-5) 퐶푖푘

We ranked all sites in order of 푐푒푖푘 for each action and each of the three benefit metrics. ROI analysis requires an evaluation of how much benefit we would expect for a given investment, which 푚 we defined as the cumulative benefit (∑푖=1 퐵푖푘) for each management strategy (and each benefit metric) after sites were ranked by 푐푒푖푘. In addition to managing only for foxes and only for grazing, we included a third management strategy of the combined management actions. We combined the information on each single-threat abatement action (here two actions, grazing and fox control) into a single list that was ranked in order of 푐푒푖푘. If a site had high cost-effectiveness for managing one threat, but lower cost-effectiveness for managing the second threat, it might be selected only for managing the first threat under a small budget, but under a larger budget both threats could be managed.

We then found the benefit of selecting sets of sites for given budgets. Mathematically, the problem formulation for prioritisation using the ‘knapsack’ approach is:

푝 푚

max ∑ ∑ 푥푖푘퐵푖푘 푘=1 푖=1

푝 푚 subject to ∑푘=1 ∑푖=1 푥푖푘퐶푖푘 ≤ 퐵푢푑𝑔푒푡 (3-6) where 푥푖푘 represents the act of selecting and managing site i for threat k and is therefore a value of one if selected or zero if not. We maximised the cumulative benefit of selecting the action by site combinations (out of a total of p × m choices ranked by cost-effectiveness) subject to a budget constraint using a greedy heuristic, finding a near-optimal solution to the knapsack problem in equation 3-6. By removing the budget constraint, this equation will calculate the total ROI (i.e. the cumulative expected benefit) for a given level of investment (calculated as cumulative costs).

To understand the ROI as sites are added to the management site selection based on their original

푐푒푖푘 rank, we need to plot the cumulative expected benefit against the cumulative cost. We did this for each of the three management strategies, and for each of the three benefit metrics. To determine whether ROI is better than an arbitrary approach that does not consider cost-effectiveness, we calculated and plotted cumulative ROI curves for randomly selected management sites. For a more 63

realistic comparison to potential selections by a manager optimising for either benefits or cost, we calculated maximise-benefit and minimise-cost ROI curves (with no other constraints). In these cases, sites were ranked by species richness with disregard of cost for the former, and by cost with disregard of species richness for the latter. We also conducted sensitivity analyses to clarify 1) whether costs or expected benefits drive the differences in cost-effectiveness between sites, by reanalysing the ROI using constant costs or constant benefits and 2) whether results are robust to variation in cost estimates that might be driven by the uncertainties outlined above. To visualise the changes in benefits with cumulative investment, we fitted the ROI curves determined by cost- effective selection of management sites using a cubic smoothing spline in R, and fitted the ROI curves determined by random selection of management sites with linear regression. To compare the distribution of priority sites under different benefit metrics, we mapped the variation in expected benefits and cost-effectiveness for managing each threat across the landscape. We also compared the cost-effectiveness and percentage of total possible expected species benefit with different budgets relative to an unrestricted budget, for each of the three management strategies (foxes, grazing, and combined).

3.3.6.2 Spatial priorities for single or multiple actions using ROI analysis

The steps above demonstrate how to find the return on investment for a given management strategy. Our final aim was to solve the problem of choosing where to implement one or more management actions across the landscape for a given budget, dependent upon the expected benefits of the benefit metric incorporating the most information (Metric 3). We selected the best management strategy, i.e. the one that resulted in the highest and quickest benefits for a given budget based on the results of the ROI approach (described above). We illustrated ranked cost-efficient sites for this ‘best strategy’ under a total budget of $10 million by mapping spatial locations according to whether one or more actions were selected on the basis of cost-effectiveness by the knapsack approach.

For the described analysis (above), we processed the spatial data using a geographic information system (GIS) (ArcGIS v10) and exported the spatially-indexed data (habitat benefit and management cost maps) from the GIS into text files. We then imported the data to spread sheets (Microsoft Excel) for calculating the new ROI attributes. We calculated species’ benefits using the three different metrics, and cost-effectiveness values. Next we sorted data by cost-effectiveness values to determine ROI rankings for each site, and then we cumulatively summed benefits and costs for discrete budgets. In other words, the solution is a greedy-heuristic based on the rank-order of the benefit-cost ratio. We joined the calculated attributes back to the spatial data in the GIS for map display, based upon the unique spatial index number.

3.4 Results

The expected benefit of threat management is high with low initial investment when management sites are selected based on their cost-effectiveness rank. In comparison, benefit is linearly related (linear regression models in all cases R2 > 0.99; p < 0.001; d.f. > 22,517; Appendix 3D) to investment with indiscriminate selection of management sites (Figure 3.1). Most of the expected benefit to threatened species is achieved when less than $10 million is spent in this region over 20 years, if measured using Metric 2 or 3 (~67%; Figure 3.1). This is illustrated by the steepness of the ROI curves when cumulative management cost is less than $10 million, with diminishing returns as more money is invested in managing each threat. When the management goal is focused on species richness only (Metric 1), rather than benefits that include addressing habitat restriction (Metric 2) and vulnerability (Metric 3), the slope of the ROI curve is flatter (Figure 3.1a). Even so, expected benefits to species are more than four times higher if investing $10 million in cost-effective, species-rich sites using a combined threat management strategy (Metric 1; expected benefit = 19.2%), than if the same amount of money is randomly invested (expected benefit = 4.5%; Figure 3.1a). However, by further specifying a management goal that prioritises restricted habitats with less vulnerability in bio-diverse areas (Metric 3), the contrast is even greater: expected species benefit is 16 times higher when $10 million is spent on management sites that are prioritised using cost-effectiveness (Metric 3; expected benefit = 66.4%) as compared to random selection (cumulative benefit = 4.2%; Figure 3.1c).

Returns from investing in more realistic traditional conservation planning scenarios (maximise- benefit or minimise-cost) are expected to be higher than if resources are arbitrarily allocated (as above), but lower than if investing in cost-effective actions prioritised using Metric 3 (Figure 3.2). Particularly in the maximise-benefit case for grazing management action (Figure 3.2b), very expensive sites drive up costs while returning little overall benefit.

When we reanalysed the ROI curves keeping either benefits or costs constant, the comparisons between curves varied depending on the metric used to calculate benefit (Appendix 3E). When benefits are kept constant and costs are spatially variable, cost-effectiveness is necessarily driven by costs. However, we found that with constant benefits, the ROI curve is more similar to the curve that uses the variable costs and expected benefit values for Metric 1, whereas it is more similar to the random allocation of investment curves for Metrics 2 and 3. In contrast, when costs are held constant and cost-effectiveness is driven by benefits, the results are reversed and the ROI curve is more similar to the random allocation of investment curve for Metric 1, whereas the curves for Metrics 2 and 3 are closer to the curves that use the variable costs and expected benefit values. The

maximise-benefit and minimise-cost scenarios appear to closely mirror the constant-benefit approach of each metric. Varied management cost produces similarly shaped ROI curves for implementing cost-effective threat management action (Appendix 3G, Figure 3G. 1).

(a)

(b)

(c)

Figure 3.1 Total Return-on-Investment (ROI) curves show greater expected benefit for the same level of investment (cumulative cost in units of million Australian dollars) when management sites are chosen by ranked cost-effectiveness (solid lines) as compared to arbitrary/random selection (dashed lines) (see Appendix 3D for linear regression statistics). Alternate action strategies are fox- baiting (Fox), grazing control (Grz) or combined fox and grazing control (Cmb). ROI curves also show that choosing management sites based on different management goals also has contrasting results, as seen when prioritising action for areas with (a) high species richness (Metric 1); (b) and restricted habitats (Metric 2); (c) and less threat vulnerability (Metric 3). Of 129,894 (500 m x 500

m) potential management sites, 22,519 of the sites with grazing vulnerability and 86,790 of the sites with fox vulnerability were matched with associated management costs. Out of these sites (109,309 possible action x site combinations), 16,727 of the sites have vulnerability to both threats.

(a)

(b)

(c)

Figure 3.2 Total ROI curves are plotted to show the relationship between investment (cumulative cost: million AUD) against the percent of the total benefit expected to be returned from site management with threat action-abating strategies of (a) fox-baiting; (b) grazing control; (c) combined fox and grazing control. Benefit is calculated on the basis of species richness, habitat range size, and threat vulnerability (Metric 3). As compared to choosing sites based on ranked cost effectiveness, choosing management sites either by cost (ranked low to high with disregard to benefit, i.e. minimise-cost) or benefit (species-richness, ranked high to low with disregard to cost, 67

i.e. maximise-benefit) can result in lesser conservation outcomes for the same level of investment. All are superior to random (arbitrary) action.

The spatial distribution of expected benefits differs from that of cost-effectiveness (푐푒푖푘) values for both fox threat action (Figure 3.3a) and grazing management (Figure 3.3b). The most cost-effective areas generated using benefit metrics that take into account species richness and restricted habitats (Metric 2) plus areas of less threat (Metric 3) vary from those selected using Metric 1 (species richness only), particularly under low budgets (Figure 3.3, cost-effectiveness columns). Differences between the spatial distribution of priority areas chosen using Metric 2 and Metric 3 to determine cost-effectiveness are less prominent, but are still present. Areas with high expected benefit of grazing management (Figure 3.3a) and fox management (Figure 3.3b) also vary markedly across the landscape (Figure 3.3, benefit columns).

Cost-effectiveness of threat management action is highest with initial investments, and decreases as more money is invested, with diminishing returns. Using Metric 3, the most cost-effective strategy with the highest expected species benefit is to simultaneously implement grazing management along with fox-baiting across the landscape (Figure 3.4). For a budget of $10 million, it is almost twice as cost-effective to manage for both threats (∑푐푒푖(푏표푡ℎ 푡ℎ푟푒푎푡푠) = 7.85), than to solely manage for foxes (∑푐푒푖(푓표푥)= 4.52) or grazing (∑푐푒푖(푔푟푎푧푖푛푔)= 4.55; Figure 3.4a). Furthermore, the first $1 million spent on management is the most cost-effective allocation of money to abate threats to species, with rapidly diminishing returns on additional money invested; spending $1 million on a combined management strategy is greater than five times more cost-effective than spending $10 million (∑푐푒푖(푏표푡ℎ 푡ℎ푟푒푎푡푠) = 42.65 vs. 7.85; Figure 3.4a. (Table 3F. 1 and Table 3F. 2 detail ceik values under discrete budgets for each metric in Appendix 3F). Likewise, spending only $1 million delivers more than a third (36%) of the total possible expected benefit to species, i.e. if there was unlimited money to spend on a combined management strategy (Figure 3.4b). In addition, over 66% of the possible expected species benefit is achieved with the first $10 million budgeted for management; an investment of $50 million would increase expected species benefit only by 21% over the initial $10 million investment (Figure 3.4b).

Cost-effectiveness guides which management action to implement with additional investment for the greatest expected benefit to threatened species (Figure 3.5a). For each increment in investment in the best strategy, we plotted what percentage of that budget was allocated to either action. If budgeting $10 Million, approximately half of the investment is in fox management, after initially greater investment in grazing management (Figure 3.5b). Mapping the spatial distribution of cost-

effective sites for management action indicates high priority areas for managing the threats of habitat degradation, invasive species predation, or both (Figure 3.6).

(a) (b)

Figure 3.3 The spatial distribution of the expected benefits of management for species who are vulnerable to (a) fox predation or (b) grazing threats are not necessarily the same as cost-effective 1 (푐푒푖푘) locations of managing those threats . This result indicates that priorities differ depending upon data incorporated into the decision-making process; benefit metrics reflect different emphasis in achieving the management goal of threat abatement. Shading (light to dark) indicates low to high benefit and cost-effectiveness values for fox and grazing management. Expected benefits were estimated using three metrics: Metric 1 addresses species richness, Metric 2 also includes restricted habitats, and Metric 3 also accounts for multi-threat vulnerability.

1Note that each dot in the diagram designates the centre of a 25-ha management unit to aid in visualisation; some dots overlap each other due to map figure size. For display, data were normalised by dividing every benefit or cost-effectiveness value by the maximum possible summed

benefit or cost-effectiveness value, respectively, for each threat and each metric. Maps are shaded using ‘Geometric Intervals’ based on classes delineated by natural data groupings, a balance between highlighting middle and extreme values.

(a)

Foxes 20

effectiveness Grazing -

(benefit/cost) Combined

10 Cost

0 $1M $3M $5M $10M $50M unlimited Management Budget

(b) 100

Foxes 75 Grazing Combined 50

25 Expected benefit (%) benefitExpected

0 $1M $3M $5M $10M $50M unlimited Management Budget

Figure 3.4 A greater percentage of species are expected to benefit from simultaneously managing threats with a combined management action strategy, but cost-effectiveness is greatest with initial investment and returns diminish rapidly. Constrained management budgets for fox, grazing, and combined management ($M = $Million) are plotted versus (a) cost-effectiveness and (b) expected species benefit when management site selection across the region is by ranked cost-effectiveness. Here, the goal is to benefit species by prioritising threat abatement action for areas with high species richness, restricted habitats, and less threat vulnerability (Metric 3).

(a)

(b)

Figure 3.5 (a) Cost-effectiveness guides which threat abatement action to implement with additional investment (cumulative cost: $M = $Million) in management for the greatest expected benefit to species under the constraint of a $10 M budget.* (b) Investment in management costs is approximately half-and-half for fox and grazing action. Areas of high species richness, restricted habitats, and less threat are prioritised (Benefit Metric 3). *Note that only every 50th management site is plotted for easier visualisation.

Figure 3.6 Mapping the most highly-ranked cost-effective sites for acting to abate threats to species shows where to efficiently spend a management budget of $10 million in the Burnett-Mary Natural Resource Management Region of Queensland, Australia. Here, the goal is to benefit species by prioritising threat abatement action for areas with high species richness, restricted habitats, and less threat vulnerability (Metric 3).

3.5 Discussion

The importance of considering conservation cost-effectiveness for species protection through habitat acquisition is well-understood (e.g. Ando et al. 1998). This concept is routinely used in systematic conservation planning (Margules and Pressey 2000) where algorithms are used to identify areas at a fine scale that efficiently meet feature targets at minimal cost (Ball et al. 2009), for producing priority rankings for biodiversity and alternative land uses (Moilanen et al. 2011a), and more recently for multifaceted conservation ROI analyses that consider additional dependencies (e.g. Kovacs et al. 2013, Kramer et al. 2013). However, with added algorithmic sophistication comes additional computational requirements and difficulty in translating results for stakeholders. Simplicity can be achieved through reducing the complexity or amount of data required to inform decision analysis, or through simplifying the decision-making approach itself. In this study we address both, by using a case study to explore the results of using different types of data to inform cost-effective prioritisation of management actions, and secondly by demonstrating a simple cost- effectiveness prioritisation approach that uses a basic ROI as an alternative to more complex analyses. We focus on a fine-scale prioritisation of multiple management actions to benefit multiple threatened species in an administratively-managed natural resource region. Our approach improves on traditional hotspot allocations of funding based only on species and threats that do not account 72

for costs and actions (Brooks et al. 2006) by demonstrating a technique for selecting management actions which optimises ROI. The practical relevance of the technique, in addition to its transparency and ease-of-use encourages uptake and utilisation for management decision-making (Venkatesh and Davis 2000). Our approach allows managers to determine priority locations where one or more actions would be most cost-effective to mitigate threats (Figure 3.6).

3.5.1 Informed investment when budgets are low can yield high expected benefits to species

Small expenditures on management action result in high expected benefits to species when management sites are prioritised by cost-effectiveness (Figure 3.1, Figure 3.2, and Figure 3.4) compared with allocating funding based on species richness alone, or to minimise spending (Figure 3.2). Other research has also found ROI to yield greater conservation outcomes per dollar than either of these approaches (Ferraro 2003, Naidoo et al. 2006, Murdoch et al. 2010). We also compared our ROI results to a baseline of un-informed action, and found that allocating funding in an arbitrary manner (Figure 3.1 and Figure 3.2) yields low returns. It is important to highlight that ROI for threat management is high initially, but returns rapidly diminish. This means that investors receive the highest relative returns at low levels of investment, if money has been spent strategically on the most cost-effective locations and actions first (Figure 3.1, Figure 3.2, and Figure 3.4). This finding supports arguments against delayed action (Grantham et al. 2008, Grantham et al. 2009) and highlights the benefits of acting immediately even when limited funding is available to invest in management.

3.5.2 ROI results are sensitive to the way in which expected benefits are defined and how benefits and costs co-vary

Conservation ROI can be used to focus management decisions, when the management goal is clearly defined by the benefit metric. Our ROI results were sensitive to the metric used to measure benefit to species (Figure 3.1 and Figure 3.3), which indicates that explicitly-stated management goals (and benefits) can effectively direct where the greatest benefit to species is expected to be obtained and at what cost. Differences in the returns on conservation investment and the spatial distribution of priority sites for management actions depended on the level of complexity and type of data used to parameterise benefit metrics (Figure 3.1 and Figure 3.3). Our benefit measures reflected priorities for management action based upon criteria of species richness, restricted habitat area, and number of threats to species (vulnerability). By comparing different benefit metrics we demonstrate that decision-making with multiple criteria relevant to the decision-making context narrows the array of choices about where to manage effectively (Figure 3.3; Metric 1 versus Metric 2 or Metric 3). Managers need to be aware that allocating funding based on species richness alone 73

(a common goal of systematic conservation planning), rather than vulnerability to threats or other relevant ecological information such as rarity of habitat, might prioritise inefficient areas. If appropriate ecological data are available and can be applied as we have demonstrated, an ROI approach can yield enormous gains in conservation efficiency (Murdoch et al. 2010). Most of the priority species in this analysis are not known to be affected by both threats, explaining the small difference in priority locations when the benefit measure accounts for multiple threats in addition species richness and restricted habitat (Figure 3.3; Metric 3 versus Metric 2). Further discrimination would become more apparent if including additional threats (e.g. we did not model the cost of the threat of fire frequency or invasive plants) or species (e.g. we did not include invertebrates in our analysis due to lack of data).

ROI results are sensitive not only to the way in which expected benefits are defined but also to the range in values of management benefits and costs (and how these values co-vary). We demonstrated how to interpret whether the ROI is more sensitive to costs or benefits for a given scenario (Appendix 3E). For our case study, we found that costs drive differences in cost-effectiveness when the range in expected benefits is small but the range in costs is large (e.g. Metric 1, Appendix 3E). ROI is often more dependent on cost differences rather than benefit differences (Ferraro 2003, Murdoch et al. 2007). If the range in expected benefits is larger than the range in costs (in our study Metrics 2 and 3 ranged in expected benefits by four orders of magnitude, whereas costs ranged by three orders of magnitude across the region), expected benefits are more likely to drive the early gains in efficiency (Appendix 3E). Compiling information on both the costs and the expected benefits is therefore crucial to determining where to achieve the greatest conservation outcomes using ROI. However, as a general rule of thumb, calculating cost-effectiveness will be most useful where costs and expected benefits are heterogeneous across the landscape (Ando et al. 1998, Polasky et al. 2001, Naidoo et al. 2006). If cost is spatially homogeneous, it will not change the decision about where to implement a single management action (although it will inform the relative cost-effectiveness of different management action considerations), and expected benefits alone might be used to inform decisions. This is rarely the case, although many previous conservation prioritisations have assumed constant costs or disregarded them completely (e.g. Brooks et al. 2006, Kremen et al. 2008).

We found that ROI results are robust to minor variations in management cost in that the shape of the curve remains similar with a reduction or increase in management costs, but benefits are accrued sooner if costs are lower than estimated or later if costs are higher (Appendix 3G, Figure 3G. 1). However, in considering if management costs might change in the future (Arponen et al. 2010), our method also allows managers to visualise how considerably reducing the cost of an action (e.g. if 74

landholder engagement reduces cost of stewardship agreements, or if innovation reduces cost of fox-baiting) might affect in which action they invest. High uncertainty in costs (i.e. up to 50%) can result in changes to the management actions being prioritised. For instance, we show that if grazing management costs are reduced by 50%, it is much more cost-effective to invest in that action (Appendix 3G, Figure 3G. 2). ROI also identifies high priority sites for initial investment, which are similar despite cost uncertainty (Appendix 3G, Figure 3G. 3). We have demonstrated a framework to account for variation in management costs, and to visualise how uncertainty in costs, or the potential for costs to change could alter (a) the selected actions as (b) where to act.

3.5.3 Multiple-action strategy for multiple species is better than single-action strategy

We found that a strategy that incorporates more than one action is more cost-effective than single- action strategies, and benefits more species for a lower cost (Figure 3.1, Figure 3.2 and Figure 3.4). This is because different actions provide benefits to different species by mitigating threats that act in different ways. This is an important finding given previous studies that have found no difference between overall outcomes of single- and multi-species management (Clark and Harvey 2002, Cullen et al. 2005). Moreover, while the predominant focus of conservation planning is on single- threat mitigation, such as reserve acquisition to stop habitat loss (e.g. Watson et al. 2011), it is clear that multiple management actions are needed to reduce species loss (Butchart et al. 2010). One explanation for previous lack of support for multi-species and multi-action threat management is the tendency to ‘lump’ species together based on their habitat use in the same or similar region (Dobson et al. 1997), rather than grouping them according to vulnerability to specific threats. In this study, we explicitly grouped species based on their common threats as well as their co-location, and in doing so, focus management on eliminating or mitigating those threats as recommended for successful multi-species management (Clark and Harvey 2002).

By using a simple ROI approach, we were able to untangle the reasons behind one or more actions being selected at a given site (due to its cost-effectiveness) to find a single solution (e.g.Figure 3.6) to a decision problem where there are multiple threats and associated actions. In our case study, both reducing grazing pressure and controlling for foxes across the landscape was warranted, although both management actions are not always required or cost-effective at the same location (Figure 3.5 and Figure 3.6). This makes our approach much easier to follow and to translate the results of planning for multiple actions compared with more complex systematic conservation planning tools or ROI algorithms. The other approaches can produce multiple optimal solutions in which the trade-offs between the benefits and costs of doing different actions are obscured or at least potentially difficult to interpret by some.

Our approach assumes that there is no interaction between the threats being evaluated, and that their effects are additive. We locate cost-effective sites for managing both threats, although realistically, the effects of threats may be synergistic (Brook et al 2008) or antagonistic (Brown et al. 2013b), and more threats exist in the study area than the two we examined. Dependencies or interactions can be added to ROI, but for simplicity of demonstrating our approach we did not do this. We have shown that managing more than one threat is more effective than managing threats independently. Accounting for dependencies would increase the efficiency, due to the possibility of cost-sharing across similarly-managed species’ projects (Briggs 2009, Joseph et al. 2009).

3.5.4 ROI informs where to act

Conservation ROI provides the means to visualise where particular management actions are most cost-effective to implement given a budget (Figure 3.6), aiding in stakeholder engagement. Alternative management actions affect species differently, and it is valuable to identify where which management action can maximise conservation benefits (van Teeffelen and Moilanen 2008). We incorporate ‘indirect’ threat maps in our analysis, assuming that different levels of species richness (derived from the summed habitat models of species that are known to be susceptible to a particular threat) can be used to represent areas of greater and lesser threat vulnerability (Wilson et al. 2005b). We do this because natural resource managers are most concerned with acting on threats where species of interest occur or are likely to occur. An extensive literature discusses decisions about appropriate species for management focus, e.g. from focal (Lambeck 1997) to indicator (Tulloch et al. 2011) species; we chose previously identified priority threatened species (Department of Environment and Resource Management 2010). The choice of species for which to prioritise threat management is especially important when developing indirect threat maps for decision-making and conservation ROI, because the distribution of threats is dependent on the species and their associated vulnerabilities, as well the scale and level of detail in the predicted distributions (Auerbach et al. In prep-b).

Our ROI approach is particularly useful for ubiquitous threats – those that occur across most of the area of interest. It might be less useful for heterogeneous, localised, or patchy threats (e.g. a pocket of weed invasion, or disease), where we might need to know the precise location of the threat and not just the species vulnerable to it. However, certainty in threat data is difficult to come by, and our ROI approach provides first-pass guidance for decision-making in an uncertain world as to where management action could be focused.

We show that the sites that have the highest expected benefit of conducting threat management may not be the most cost-effective (Figure 3.3). Prioritisations that ignore costs can therefore lead to

scarce resources being used inefficiently (Ando et al. 1998, Balmford et al. 2000) and opportunities to achieve conservation goals may be lost (Naidoo et al. 2006). Ultimately, however, assessing cost- effectiveness may be just one part of the conservation decision-making process. Cost-efficiency in conservation does not necessarily translate to effectiveness in management, i.e. work in practice (Arponen et al. 2010). This is because the least expensive solution may not be the best for long- term species persistence (Cabeza and Moilanen 2001, Haight and Travis 2008), may be in marginal or infeasible locations (Gaston et al. 2001), or may depend on other socioeconomic concerns that were not considered (McBride et al. 2007, Knight et al. 2010). These additional considerations might enhance benefits (Di Minin et al. 2013), exacerbate threats (Faith and Walker 1996), or increase costs (Armsworth et al. 2006). Incorporating additional considerations such as these is possible in our ROI approach. For instance, feasibility or likelihood of successful management (or failure; a consideration that will decrease benefits in a heterogeneous way across the landscape) can be included in analyses by incorporating a probability of success into the benefit metric (Joseph et al. 2009, Tulloch et al. 2011, Tulloch et al. 2013a). Alternatively, species importance for conservation might be considered by including a weighting factor in the benefit metric based on phylogenetic distinctiveness (e.g. Isaac et al. 2007, Joseph et al. 2009). For simplicity in presentation, we minimised consideration of these additional factors.

3.5.5 Conclusion

Our research demonstrates an ROI conservation decision-support approach that can be implemented to guide investment in cost-effective management action for abating threats to species. A cost- effectiveness analysis is straightforward: the calculation is based on the expected benefits to species for abating threats relative to the costs of action. Expected benefits are calculated using the spatial distribution of the species of interest, knowledge about which species are affected by a given threat, and additional management concerns, such as restricted habitats and vulnerability. Our approach integrates economics with conservation decision science to inform threat management for stakeholders interested in biodiversity conservation. Resource managers with limited capacity can apply the approach to any set of species and management actions to address abating threats to those species, employing commonly-used GIS and spread sheet software. Decisions about where to efficiently act are informed by cost-effectively investing limited resources. Spatial conservation ROI analysis transparently and accountably locates the least expensive areas to manage specific threats for the greatest expected benefit to species. The ROI approach presented here provides decision support despite having only estimates of species benefits from, and management costs for, abating threats. Efficient use of limited resources to act on abating multiple threats to multiple species, informed by ROI, can be expected to deliver greater conservation outcomes to threatened 77

species at less cost. A complex algorithm may not always prove a better solution to a conservation problem (Possingham et al. 2000), and a simple approach has more potential for uptake by conservation practitioners due to its ease-of-use and transparency.

3.6 Acknowledgments

This research was conducted with the support from the Australian Government’s National Environmental Research Program, the Australian Research Council Centre of Excellence for Environmental Decisions, and an Australia Postgraduate Award scholarship. The Queensland Government provided species presence data (WILDNET (Department of Environment and Resource Management (DERM) 2010) and HERBRECS (Department of Environment and Resource Management (DERM) 2011) databases) used in species distribution modelling for this research. We appreciate suggestions by anonymous reviewers and the editor that improved the research. We thank Rachel Lyons for her interest in, and discussions about, the case study research.

3.7 Supplementary Information

These appendices provide details of our case study region, methods to derive cost models, more detailed results, and sensitivity analyses.

3.7.1 Appendix 3A The case study

This appendix provides details of the species, taxa, threats, and types of species distribution maps used in the analysis.

Table 3A. 1 List of priority threatened species in the Burnett-Mary Natural Resource Management region (Department of Environment and Resource Management 2010) analysed for cost-effective threat management.

Species Taxa Threat1 Map Type2 Fox Grz Ttl Acacia eremophiloides vascular plant 0 1 1 SDM Acacia porcata vascular plant 0 1 1 SDM Acanthophis antarcticus reptile 0 1 1 SDM Alectryon ramiflorus vascular plant 0 1 1 SDM Caretta caretta reptile 1 0 1 SDM Corynocarpus rupestris ssp arborescens vascular plant 0 1 1 SDM Dasyornis brachypterus bird 1 1 2 SDM Dasyurus hallucatus mammal 1 0 1 SNES Dasyurus maculatus mammal 1 0 1 SNES Elseya albagula reptile 1 0 1 SDM Elusor macrurus reptile 1 0 1 SDM Esacus magnirostris bird 1 0 1 ALA Maccullochella mariensis fish 0 1 1 SDM Mixophyes iteratus frog 0 1 1 SDM Pararistolochia praevenosa vascular plant 0 1 1 SDM Phaius australis vascular plant 0 1 1 SNES Romnalda strobilacea vascular plant 0 1 1 SDM Sterna albifrons bird 1 0 1 SNES Turnix melanogaster bird 0 1 1 SDM Xeromys myoides mammal 1 1 2 SDM

1Threat (0: not a threat; 1: threat); Fox (Foxes); Grz (Grazing); Ttl (Total number of threats in study); 2Species distribution map types: SDM (Species distribution models modelled using Maxent (Phillips et al. 2006, Dudík et al. 2010) software (Auerbach et al. In prep-b)); SNES (Species of National Environmental Significance range maps (Department of the Environment 2008a)); ALA (Atlas of Living Australia point locations (Atlas of Living Australia (ALA) 2012)).

3.7.2 Appendix 3B Model of opportunity costs to cease grazing

This appendix provides details on our modification of the agriculture profit model by Marinoni et al. (2012). We modelled the net present value of opportunity costs for a 20-year stewardship program. Our objective was to develop a spatial estimation of the opportunity costs that are needed for a program to abate the threat of grazing to priority species in a regional location. Here, we implement a program whereby landowners are compensated for foregone agricultural profits.

We assume:

1) that the region is made up of a matrix of sites M with dimensions x (columns) and y (rows)

2) that each site Mxy is 500 x 500 meters in size 3) the constants presented in Table 3B. 1

4) and that the cost estimated, below, is for one management site Mxy.

Let the net present value of the cost to administer a grazing-abatement stewardship program (G) over time period (t) for site Mxy be

푁 퐺 푁푃푉 (푑, 푁 ) = ∑ 푝 푡 (3B-1) 푝 푡=1 (1+푑)푡

where the opportunity cost with time of cash flow (t) over the total number of periods (Np) is discounted using rate (d).

Table 3B. 1 Notation, value, and description of variables in grazing management cost model.

Notation Value Description t 1 time of cash flow in years

Np 20 total number of periods for stewardship program d 0.02 discount rate G net cash flow agricultural profit (Marinoni et al. 2012)

3.7.3 Appendix 3C Model of management costs to bait foxes

This appendix provides details for the fox management cost model. We modelled the net present value of management costs for a 20-year fox-baiting program. Management action for threats is not always feasible throughout a species’ mapped distribution. For example, because fox-baiting is a risk to domestic pets, it is risky to undertake near habitation, and in some states of Australia, illegal to do so. We limited our analysis to fox-baiting management greater than 2 km from urban areas, even though foxes are managed at the urban edge in other areas (Mason and Olsen 2000, Olsen et al. 2005). We accounted for a variable cost of fox baiting due to effort required for ground-baiting in accessible locations near roads, and less accessible locations away from roads and major urban centres. Recommendations for frequency of baiting campaigns vary, e.g. from at least two times a year (Saunders and McLeod 2007), to monthly intervals of several months (Carter et al. 2011), to six times a year (de Tores et al. 1998). For wildlife recovery in Western Australia, baits are laid by aerial or ground operations at least four times a year at an intensity of five baits per km2 (Saunders and McLeod 2007). Whereas aerial baiting is allowed in Queensland, our study area is more highly-populated and fragmented than a comparable fox-control area in Western Australia. We determined a ground-baiting approach, with greater control to replace ‘taken’ baits and remove aged, sub-lethal, baits (to prevent poison resistance in foxes) based on research by Gentle (2005) was labour-intensive, but necessary. We calculated costs assuming a bait station density of one station per .25 km2 (500 m apart) to simplify calculations, address legislative guidelines for Queensland (Saunders and McLeod 2007; Table 11.4) that baits be placed close enough together for uptake yet not too close that foxes cache unconsumed baits, and take into consideration that increasing the bait density from 5 to 10 baits per square kilometre did not result in increased uptake (Thomson and Algar 2000). We based our preferred baiting regime for optimal native species recovery management to be four campaigns of 14 days in length per year, reasoning that such an intensive effort is necessary to control foxes for biodiversity conservation. We also tested reducing the duration of the baiting campaigns to 7 days and increasing them to 28 days but do not use these models in the present analysis. We used ArcGIS 10 Spatial Analyst and raster data with a 500 x 500-m resolution.

Our objective was to develop a spatial estimation of the management costs that are needed for a program to abate the threat of foxes to priority species in a regional location. Here, we implement a design strategy whereby poison-bait stations are administered in locations equally spaced in a grid pattern radiating from roadsides. We estimate the net present value (NPV) of this program strategy over a 20-year time period, and our strategy is to implement a program of four 14-day baiting campaigns per year. We incorporate the management costs of labour to administer bait stations, 81

vehicle operation, and bait costs. Therefore, our site estimates for fox-management cost in our return-on-investment analysis is the NPV of a comprehensive fox-baiting management program.

We assume:

1) that the region is made up of a matrix of sites M with dimensions x (columns) and y (rows)

2) that each site Mxy is 500 x 500 meters in size 3) the constants presented in Table 3C. 1 4) the costs presented in Table 3C. 2

5) and that each cost estimated, below, is for one management site Mxy.

3.7.3.1 Net present value of 20-year fox-baiting management program

Let the net present value of the cost to administer a one-year fox baiting program (Fb) over time period (t) for site Mxy be

푁 퐹 푁푃푉 (푑, 푁 ) = ∑ 푝 푏푡 (3C-1) 푝 푡=1 (1+푑)푡

where the management cost with time of cash flow (t) over the total number of periods (Np) is discounted using rate (d).

3.7.3.2 Cost of one-year fox-baiting management program

Let the total management cost for administering a one-year fox-baiting program (Fb) outside an urban buffer zone area (Ub) be

Fb = ((Lb + Ob + Bb) * Ub) * Ny (3C-2)

where costs include labour to administer baits (Lb), operate a vehicle (Ob), and purchase bait (Bb) for a set number of campaigns per year (Ny).

3.7.3.3 Labour cost to administer baits

Let the labour cost to administer one bait station per site (Lb) in one 14-day baiting campaign located in a region (A) be

Lb = (Lr + La + Ld) * A (3C-3)

where labour costs include administration of a bait stations either at a site located immediately next to a road (Lr), or away from a road (La), and also includes the pro-rated cost of driving to a group of bait stations (Ld). For each individual site, either (Lr = 0) or (La = 0), depending upon its spatial position xy.

Let the labour cost to administer one bait station in a site located immediately next to a road (Lr) be

Lr = Ls where Dr == 0 (3C-4)

where labour cost (Ls) is a function of the site’s position relative to the road (DR).

Let the labour cost to administer a bait station in a site (Ls) be

Ls =((Tl +( Nr * Tr) + (Nc * Tc) + Tv)/60) * Cl (3C-5)

where labour cost is a function of time (in minutes) needed to lay (Tl), check (Tc), replace (Tr) and remove (Tv) baits, the number of site visits needed to check (Nc) and replace (Nr) baits, and the cost of standard wages paid per hour (Cl).

Let the labour cost to administer a bait station in a site located some distance away from a road (La) be

La = Df * Cl * Ta * Nt (3C-6)

where the labour cost is a function of a road distance function (Df) described below, wages per hour

(Cl), estimated time to administer the bait (Ta) including driving from the previously visited station

500 m in distance, and the number of trips to a bait station needed for a 14-day baiting campaign

(Nt).

Let road distance function (Df ) be

∞ 1 퐷푓 = ∑퐷 (3C-7) 푟 퐷푟

where distance from a road (Dr) is used to modify labour costs for increasing effort needed to administer bait stations.

Let the pro-rated labour cost to drive to a group of bait stations (Ld) be

Ld = (((DU * Nk * Cl)/Ns) * Nt * Nu)/Nd (3C-8)

where the labour cost is a function of the distance from the nearest urban centre (DU), the distance to drive each site (Nk), wages per hour (Cl), the driving speed (Ns), the number of trips to a bait station needed for a 14-day baiting campaign (Nt), the number of trips in a day (Nu), and the number of bait stations that can be administered in one day (Nd).

3.7.3.4 Cost to operate a vehicle

Let the pro-rated vehicle operating cost (Ob) to drive to a group of bait stations located in a region (A) be

Ob = ((((DU * Nk *Cv) + (Nk * Cv)) * Nt * Nu)/Nd) * A (3C-9)

where the vehicle operating cost is a function of the distance from the nearest urban centre (DU), the distance to drive each site (Nk), the vehicle operating costs per kilometre (Cv), the number of trips to a bait station needed for a 14-day baiting campaign (Nt), the number of trips from and to the nearest urban centre in a day (Nu), and the number of bait stations that can be administered in one day (Nd).

3.7.3.5 Cost to purchase bait

Let the total cost of purchasing bait for one bait station (Bb) located in a region (A) be

Bb = (Cb * Nb) * A (3C-10)

where the bait cost is a function of the individual bait cost (Cb) and the number of baits estimated to be needed per station (Nb) to account for a 25% uptake by foxes.

Table 3C. 1 Notation, value, and description of constants in fox management cost model.

Notation Value Description Number constants

Nt 5 number of trips driving to bait stations needed for a 14-day campaign (1 to lay baits, 0 to replace, 3 to check, 1 to remove) Nr 0 number of times replacing bait in a 14-day campaign (0)

Nc 3 number of times needed to check bait in a 14-day campaign (every 3-4 days)

Nb 1.75 number of baits needed per bait station for a 14-day campaign (Gentle 2007, table 7.3, 14-d, 25% uptake, 75.3/43 = 1.75) Ny 4 number of baiting campaigns in one year

Nd 40 number of bait stations estimated to be administered in a day

Nu 2 number of trips to/from urban centre to bait station on road (i.e. a return trip)

Nk 0.5 length of site in kilometres

Ns 60 average speed in km/hour to drive vehicle from urban centre

Np 20 total number of periods for fox-baiting program d 0.02 discount rate Cost constants

Cl 23 cost in dollars per hour for labour

Cb 1.45 cost in dollars per bait at $29/20 baits

Cv 0.63 cost in dollars per km to operate vehicle Time constants

Ta 0.5 time in hours allowed to access bait stations away from road (500-m increments)

Tl 8.17 time in minutes to lay bait

Tr 6.08 time in minutes to replace bait

Tc 5.38 time in minutes to check bait

Tv 6.08 time in minutes to remove bait t 1 time of cash flow in years Raster constants

A 0,1 The value of site Mxy representing its presence as being either inside (A=1) or outside (A=0) of a region. Ub 0,1 The value of site Mxy representing its presence as being outside (Ub=1) a 2 km buffer around urban areas (Ub = NA inside urban buffer). Du 0,∞ The value of site Mxy representing its distance in number of sites from the nearest urban centre U (calculated as Euclidean distance in meters and reclassified by number of sites). Dr 0,∞ The value of site Mxy representing its distance in number of sites from the nearest road R (calculated as Euclidean distance in meters and reclassified by number of sites).

Table 3C. 2 Labour, vehicle operation, and bait cost details and sources for fox management cost model.

Detail Cost Source

Labour costs adjusted to 2012 salary based on a $23.00 per hour (Reserve Bank of 3.4% inflation rate Australia (RBA) 2012)

Labour costs based on 2010 hourly rate for farm, $21.50 per hour (Australian Bureau of forestry, and garden workers Statistics (ABS) 2010)

Vehicle operating costs based upon business travel $0.63 per (Australian Taxation reimbursement for a fuel-efficient 1.6-litre engine kilometre Office (ATO) 2012) automobile (e.g. Hyundai Accent, Kia Soul)

Foxoff ® manufactured meat baits (Animal Control $29.00 includes (Livestock Health and Technologies, Melbourne, VIC, Australia) GST per 20 baits Pest Authority 2012)

3.7.4 Appendix 3D Linear regression statistics for random management ROI curves

This appendix provides details of our analysis comparing the return on investment for choosing threat management sites based on cost-effectiveness to randomly choosing management sites.

Table 3D. 1 Cumulative species benefit can be predicted from cumulative random investment in threat management using linear regression to statistically explain return-on-investment curves.

Management Benefit strategy metric intercept slope DF R-squared p-value 1 1.37E+02 4.53E-04 22517 0.9988 <2.2e-16 Grazing 2 7.70E-02 8.58E-08 22517 0.9962 <2.2e-16 3 9.55E-03 8.36E-08 22517 0.9992 <2.2e-16 1 2.13E+01 6.55E-04 86788 1.0000 <2.2e-16 Foxes 2 -1.21E-01 5.16E-08 86788 0.9992 <2.2e-16 3 -8.67E-02 4.51E-08 86788 0.9992 <2.2e-16 Combined 1 2.19E+01 5.98E-04 109307 0.9999 <2.2e-16 Foxes and 2 4.93E-02 6.02E-08 109307 0.9997 <2.2e-16 Grazing 3 -1.29E-01 5.57E-08 109307 0.9994 <2.2e-16 86

3.7.5 Appendix 3E What drives early gains in efficiency?

Here we present the results of a sensitivity analysis to clarify whether costs or benefits drive the differences in cost-effectiveness between sites. We reanalyse the ROI using constant costs or constant benefits.

(a)

(b)

(c)

Figure 3E. 1 Results of a sensitivity analysis show return-on-investment (ROI) curves of combined fox and grazing threat management action sites selected by ranked cost-effectiveness under 1) constant costs and variable benefits (dot-dash lines) versus 2) variable costs and constant benefits (dotted lines), as compared to ROI curves under variable costs and variable benefits for management action sites selected by 3) ranked cost-effectiveness (solid lines) versus 4) randomly selected management action sites (dashed lines). In our case study, early gains in efficiency are driven by variable costs under benefit Metric 1, but are driven by variable benefits under Metrics 2

and 3. Metric 1 (a) prioritises action for areas with high species richness, Metric 2 (b) includes restricted habitats, and Metric 3 (c) also accounts for less threat vulnerability.

3.7.6 Appendix 3F Cost-effectiveness of threat management under different budgets

Cost-effectiveness of threat management is calculated as the cumulative value of the expected benefit divided by the cumulative management costs of abating threats under different management practices. Management sites were ranked by cost-effectiveness values (Table 3F. 1) as compared to arbitrary selection of management sites (Table 3F. 2) for limited budgets and unlimited spending. Cost-effectiveness ratios are scaled for relative comparison of values between three different benefit metrics used to measure expected benefit of managing for species richness (Metric 1), and restricted habitats (Metric 2), and less threat vulnerability (Metric 3).

Table 3F. 1 Cost-effectiveness values for discrete budgets and unlimited funding vary by benefit metric and management strategy1, and also decrease with increased investment, when threat management sites are selected on the basis of ranked cost-effectiveness.

1 grz = ∑ 푐푒푖(푔푟푎푧푖푛푔); fox = ∑ 푐푒푖(푓표푥); both = ∑ 푐푒푖(푐표푚푏푖푛푒푑 푡ℎ푟푒푎푡푠)

Budget Metric 1 Metric 2 Metric 3 ($M) grz fox both grz fox both grz fox both 1 4.61 2.82 4.62 29.37 26.49 48.10 29.01 21.13 42.65 3 3.13 2.55 3.38 12.90 11.53 20.57 12.62 9.71 18.55 5 2.60 2.29 2.99 8.59 8.27 13.96 8.4 7.14 12.71 10 1.77 2.01 2.46 4.65 5.28 8.58 4.55 4.52 7.85 50 0.52 1.12 1.32 0.99 1.38 2.28 0.96 1.20 2.07 unlimited 0.52 0.66 0.60 0.99 0.51 0.60 0.97 0.45 0.55

Table 3F. 2 Cost-effectiveness values for discrete budgets and unlimited funding vary by benefit metric and management strategy1, and also decrease with increased investment, when threat management sites are randomly selected.

1 grz = ∑ 푐푒푖(푔푟푎푧푖푛푔); fox = ∑ 푐푒푖(푓표푥); both = ∑ 푐푒푖(푐표푚푏푖푛푒푑 푡ℎ푟푒푎푡푠)

Budget Metric 1 Metric 2 Metric 3 ($M) grz fox both grz fox both grz fox both 1 0.17 0.67 0.67 1.07 0.39 0.51 1.29 0.46 0.52 3 0.36 0.66 0.62 1.01 0.42 0.54 0.93 0.40 0.53 5 0.41 0.66 0.58 0.94 0.44 0.64 0.92 0.41 0.47 10 0.47 0.66 0.60 0.83 0.43 0.64 0.82 0.44 0.44 50 0.45 0.66 0.61 0.86 0.49 0.62 0.84 0.41 0.53 unlimited 0.52 0.66 0.60 0.99 0.51 0.60 0.97 0.45 0.55

3.7.7 Appendix 3G Variation in management cost sensitivity analysis

Here we present results from a sensitivity analysis of prioritising action by cost- effectiveness when the modeled cost of fox and grazing management is varied by 50%. We illustrate the analysis using the benefit metric assessed in the study that accounts for species richness, habitat rarity, and vulnerability to threats. We set up four different scenarios with variation in costs and explore how they might affect the final decision as compared to the original modeled cost scenarios:

Scenario 1: Landholders are more difficult to bargain with than we assume (Lennox and Armsworth 2013, Lennox et al. 2013) and costs of stewardship agreements to reduce grazing threats are higher (50% greater investment required). Fox management costs are equal to original model estimates. Scenario 2: Landholders are easier to bargain with than we originally assumed (e.g. due to poor weather-induced conditions for cattle grazing) and costs of stewardship agreements to reduce grazing threats are lower (50% less investment required). Fox management costs are equal to original model estimates. Scenario 3: Our estimate of fox management using four 14-day baiting campaigns each year does not provide optimal outcomes and more baiting might be required (Carter et al. 2011), increasing management costs (50% greater investment required). Grazing management costs are equal to original model estimates. Scenario 4: We relax our estimate of the baiting intensity needed for fox management based on alternative recommendations (Saunders and McLeod 2007) and costs are reduced (50% lower investment required). Grazing management costs are equal to original model estimates.

We investigated:

(a) Does the shape of the ROI curve change with cost variation? (b) Does the level of investment in different management actions change with cost variation? (c) Does the spatial distribution of investment change with cost variation?

We found that results are robust to moderate levels of uncertainty in cost in that the shapes of the ROI curves remain similar, and there are diminishing returns with increasing investment (Figure 3G. 1a). Furthermore, expected benefits can be seen to accrue sooner if management costs are reduced (Scenarios 2 and 4), or later if costs are increased (Scenarios 1 and 3) at low levels of investment (Figure 3G. 1b). Proportion of investment in management action reflects the differences in costs as compared to the original model (Figure 3G. 2). Spatial selection of sites based on cost-

effective threat management is similar even if costs are uncertain (Figure 3G. 3), particularly for initial investments.

In each figure, four management scenarios in which management costs are varied are compared to the originally modelled scenario. Scenario 1 depicts priorities for management action based on cost- effectiveness if grazing management costs are 50% higher than originally estimated, Scenario 2 if grazing management costs are 50% lower than estimated, Scenario 3 if fox management costs are 50% higher than estimated, and Scenario 4 if fox management costs are 50% lower than estimated. The expected benefit is based on Metric 3 which prioritises for high species richness, restricted habitats and less threat vulnerability.

Figure 3G. 1 Total return-on-investment (ROI) curves (a) for expected benefit over all levels of investment and (b) detail of ROI curves at low levels of investment ($1.5Million) (b). Benefits are accrued sooner if management costs are lower than originally estimated and later if management costs are higher.

Figure 3G. 2 As compared to the original scenario, proportion of investment in fox threat action is higher if grazing management costs are higher (Scenario 1) or if fox management costs are lower (Scenario 4) than originally estimated. Alternatively, proportion of investment in grazing threat action is higher if grazing management costs are lower (Scenario 2) or fox management costs are higher (Scenario 3) than originally estimated.

Figure 3G. 3 Under a management budget of $10M, cost-effective areas for managing the threat of foxes are depicted in red, for managing for grazing in green, and for managing both threats in blue for four management scenarios in which we vary management costs. The greatest difference when comparing management sites selected for action to the original cost scenario model is if management costs are lower than estimated for either grazing (Scenario 2) or foxes (Scenario 4). However, initial sites selected for investment (at $0.5M) are similar across all scenarios, meaning that the same high priority sites are selected for action even if management costs are higher or lower across the landscape than originally estimated.

Chapter 4 Accounting for dependencies in threat management can alter conservation priorities

4 Accounting for dependencies in threat management can alter conservation priorities

4.1 Abstract

Conservation prioritisation often assumes that the benefits and costs of actions for abating threats to species are independent so their cost-effectiveness can be assessed separately. In reality, there are many dependencies between and within the benefits and costs of actions for mitigating threatening processes. Our aim is to assess how accounting for dependencies in either the benefits or costs of conservation actions alters investment strategies for managing threatened species in a bio-diverse natural resource management region in Southeast Queensland, Australia. We use cost-effectiveness analysis to rank and prioritise three threat abatement actions across a landscape: altered fire management, invasive predator control, and reduced grazing pressure. We first assume each species benefits proportionately from the number of their threats that are managed (the traditional additive independent-benefits approach), and contrast this with scenarios where species benefit either when any one of their threats is managed, or only when all of their threats are managed (contrasting non-additive, dependent-benefit approaches). We also evaluate differences between managing locations in isolation compared with situations where there are local economies of scale from taking actions close to each other. We discover that accounting for dependencies affects threat management investment strategies, as well as alters decisions about where to manage with which actions. Our results clarify the trade-offs between the benefits and costs of choosing to manage one threat or all, as compared to independent threat management. If a species may be secured by managing only one of its threats and we continue to manage as though all threats are equally important and additive, funding could be squandered on managing multiple threats. If a species may be secured only by managing all threats and we continue to manage as though this is not true, then conservation outcomes may fail. Considering spatial dependencies in management costs may lead to increased investment efficiency. Traditional assumptions of management action costs and benefits could result in misplaced investments or idealistic expectations of threatened species management.

4.2 Introduction

Investment in conservation actions is imperative for reducing the rate of species extinction (Rodrigues 2006, Hoffmann et al. 2010). Understanding how and where to reduce known threats, and thereby promote species persistence, is consequently a major focus of conservation science (Salafsky et al. 2008, Auld and Keith 2009). Threats, and the costs and benefits of managing them, 99

have traditionally been assessed independently and added together to assess effects, for example in cumulative threat mapping (Halpern et al. 2008, Ban et al. 2010, Vörösmarty et al. 2010). Although multiple threatening processes usually interact (Sala et al. 2000, Brook et al. 2008), interrelationships are rarely addressed when prioritising multiple conservation actions (Duke et al. 2013). This is despite evidence that dependencies between threats affect the outcomes of threat management (Evans et al. 2011b, Regan et al. 2011, Brown et al. 2013b, Brown et al. 2013a), that focusing on single threats is suboptimal for securing species and ecosystems (Brown et al. 2013b), and that recognising threat syndromes to characterise complex threats affecting multiple species may improve management efficiency (Burgman et al. 2007).

Limited funds for conservation necessitates choosing the kind and location of actions to abate threats (Bottrill et al. 2009, Carwardine et al. 2012). Cost-effectiveness analysis (CEA) is one approach that can be used to prioritise actions on a project-by-project basis (Murdoch et al. 2007). The approach of integrating economics with biological decision-making guides more efficient distribution of scarce resources, and CEA is increasingly recommended for decision support in conservation (Hughey et al. 2003, Naidoo and Adamowicz 2006, Moran et al. 2010, Laycock et al. 2011, Auerbach et al. 2014). At a minimum, CEA in conservation requires identifying the actions needed to abate threats, the expected benefit to species from those actions, and a cost estimate for the actions. Cost-effectiveness is calculated as the ratio between the expected benefit of a conservation action (expressed in non-monetary terms) and the cost of that action (Metrick and Weitzman 1998, Weitzman 1998). Assuming independent costs and benefits simplifies a CEA (e.g. Auerbach et al. 2014, Ng et al. 2014). However, overlooking the dependencies between and within costs and benefits of actions neglects the biological and economic complexity of many conservation problems (Polasky et al. 2001).

A dependency between actions occurs when the whole (i.e. a strategy of multiple actions) is not the same as the sum of the parts (single actions). The principle of species complementarity (Kirkpatrick 1983, Margules et al. 1988) is the most commonly explored dependency in conservation planning problems. With complementarity, the benefit of conserving two sites is not simply the addition of the richness of the species in each site, but the benefits of conserving more of the distribution of a species are reduced once that species has been partially represented. In effect, there are diminishing returns, i.e. a non-linearity in the benefit function, from conserving more of the same species (Moilanen 2007). To address this dependency, complementary sites – sites that complement each other because they contain very different species – are preferred. In a conservation planning context, complementarity is a dependency that is often considered (Justus and Sarkar 2002), as is

100

spatial dependency in the costs and benefits of reserve system configurations (Possingham et al. 2000). The implications of other dependencies are rarely examined.

There are potentially important consequences of dependencies in threat management, although they have not been systematically explored. Some research on threat management examines the benefits of managing single threats, such as the influence of grazing on biodiversity (McIntyre et al. 2003), or the benefits and costs of managing single threats to multiple species, such as habitat loss (Underwood et al. 2008, Withey et al. 2012). Other research examines the benefits and costs of managing multiple independent threats to multiple species (Wilson et al. 2007). However, not accounting for dependencies when they do exist might lead to a false belief in the ability of a single management action to effectively recover populations, when in fact species might require more than one threatening process to be mitigated before they respond in a positive way (Evans et al. 2011b). On the other hand, more examples of recognised dependencies between the benefits of conservation actions exist and include when the recovery of species vulnerable to invasive predators requires both the removal of predators and the provision of adequate habitat complexity including shelter and refuge (e.g. through removal of invasive herbivores, fire control, and/or site restoration ) (Hobbs 2001, Zavaleta et al. 2001, Russell et al. 2003, Robley et al. 2013), when vegetation recovery requires both grazing and fire control (Burgman et al. 2007, Hobbs and Cramer 2008, Keith 2011) or when effectiveness of invasive species control is dependent upon coordination and management of adjacent sites (Epanchin-Niell et al. 2009, McLeod et al. 2010, Carter et al. 2011).

Our research is unique in its consideration of the importance in accounting for dependencies in both the costs and benefits of multiple conservation actions at a fine spatial scale. We consider dependencies in the benefits of managing three threats that are all known to detrimentally affect native species: too frequent and intense fire (Bradstock et al. 2012), an invasive predator (red fox, Vulpes vulpes) (Lowe et al. 2001), and habitat degradation caused by domestic stock (Fleischner 1994). We examine how considering the dependencies between the benefits and costs of actions to abate these threats affects strategies for investing in conservation action.

4.2.1 Research Aim

The aim of this research is to evaluate how, and how much, the additional complexity of considering relevant dependencies in the benefits and costs of management actions changes the decisions that would be made in a resource allocation problem as compared to circumstances where the dependencies are not included. We use a case study in which a conservation practitioner must make decisions about investing in the most cost-effective actions for managing multiple threats to species across a landscape. We first consider three scenarios to evaluate benefit dependencies,

101

where the benefit of a threat-mitigating action to a species is defined by which of its threats are managed. We assess how prioritisation decisions for investment in management actions that account for dependencies differ from those using traditional assumptions of independent costs and benefits of threat management. We compare one scenario where the expected benefits of threat management are independent and additive, with two other scenarios where the expected benefits are dependent and non-additive. Second, we choose one management action to enact and compare two scenarios for evaluating a cost dependency. We explore whether coordinated decisions in which the costs of management are dependent upon if the action has been carried out in nearby locations are more efficient than a traditional approach that ignores local economies of scale in threat management. We evaluate the benefits of accounting for dependencies in terms of differences in total expenditure, overall expected benefits, and recommended locations for undertaking actions.

4.3 Methods

4.3.1 Case study description

We evaluated dependencies in threat management actions in a case study of a Natural Resource Management (NRM) region in southeast Queensland, Australia. We divided the region (approximately 55,000 km2) into a grid of 25-ha management sites for a fine-grain analysis. We analyse only those sites (n = 129,894) with remaining or regrowth native vegetation habitat (Queensland Herbarium 2010a, b). Seventy-two species prioritised for NRM and state government threat management (Department of Environment and Resource Management 2010) are listed under State (Nature Conservation Act 1992) or Commonwealth (Environment Protection and Biodiversity Conservation Act 1999) legislation, or are species of regional concern. Potential species distributions were derived from species distribution models (Maxent; Phillips et al. 2006), range maps (Department of the Environment 2008a) or point locations (Atlas of Living Australia (ALA) 2012) (Auerbach et al. 2014). We addressed the management of three IUCN-listed threats (International Union for Conservation of Nature (IUCN) 2014) identified by experts as affecting the priority species (Department of Environment and Resource Management 2010): too frequent and intense fire, an invasive predator (red fox, Vulpes vulpes), and habitat degradation caused by domestic stock (Appendix 4A, Table 4A. 1). We assessed dependencies between the expected benefits of actions to abate threats, and between the costs of one action management because in reality there are complex interactions between these threats and the outcomes of managing for them.

102

4.3.2 Decision problem

We evaluated the effects of considering dependencies in management costs and benefits by solving the problem within a decision-theoretic framework (Possingham et al. 2001b). Our objective was to maximise benefits for a set of species, constrained by a budget, and given different options about which threat-mitigating action(s) were carried out at a particular site. Targeted actions in our case study were proactive fire management for biodiversity according to vegetation type, predator control by lethal baiting, and grazing reduction or removal through a stewardship agreement.

To assess the effects of benefit and cost dependencies within our decision-theoretic framework, we set up five likely prioritisation scenarios (Table 4.1). In each case, one baseline scenario that represents the case whereby the expected benefits and costs of actions are independent and additive, was contrasted with alternative scenarios in which the expected benefits or costs of actions are dependent (non-additive). In valuing the overall benefit of managing multiple threats to multiple species, the calculation of benefits was non-additive in two cases: if a species is secured in a site when only one threat of many is managed (optimistic), or if all threats must be managed (pessimistic). In valuing the overall costs of managing threats to species, the calculation of costs was non-additive if the cost of managing one site is dependent upon whether other sites are managed. We solved these integer linear programming problems with a greedy-heuristic on the basis of CEA rank-order of the ratio between the expected benefits and costs of threat management (after Murdoch et al. 2007). Methods for estimating independent and dependent threat management benefits and costs and prioritising sites and actions are detailed in Appendix 4B.

We optimised for cost-effective investments in independent versus dependent threat management and plotted return-on-investment (ROI) curves as investment (cumulative summed cost) against expected benefit, fitted with a cubic smoothing spline in R. (Expected benefits are expressed as a percentage of the benefit expected with a $10 million budget when managing threats in the independent scenario.) We then mapped priority sites and actions and determined the area selected for each optimised management strategy for a limited budget of $10 million. Dissimilarities in outcomes between scenarios are expressed either as differences in expected benefits for the same investment, or as differences in investment for the same expected benefit. (All costs are expressed in Australian dollars.) We mapped similarity between prioritisation outcomes and evaluated solution similarity taking chance agreement into account by calculating Cohen’s kappa (Cohen 1960) and weighted kappa measures (Cohen 1968) using fractional and binary weight matrices to count disagreements differentially (Appendix4D; Table 4D. 1 and Table 4D. 2).

103

Table 4.1 Scenarios evaluated in analyses of threat management benefit- and cost-dependencies.

Scenario Name Dependency Action benefit to Property Threat(s) species Management Cost 1 Independent Benefit Proportional to Additive for actions Fire, Fox, benefit number of threats in chosen strategy Grazing management to a species that scenario are managed 2 Optimistic Benefit Full benefit if Additive for actions Fire, Fox, management one or more in chosen strategy Grazing scenario threats to a species are managed 3 Pessimistic Benefit Full benefit only Additive for actions Fire, Fox, management if all threats to in chosen strategy Grazing scenario species are managed 4 Cost- Spatial-cost Full benefit if fox Not dependent on Fox independent threat to species whether adjacent is managed properties have been managed 5 Cost- Spatial-cost Full benefit if fox Reduced if an Fox dependent threat to species adjacent property is is managed managed

To assess ROI for investing in fox management both with, and without the cost-dependency, we set a budget of $10 million to prioritise sites and plotted ROI curves. To test whether the cost- dependent sites were grouped closely together, we estimated a measure of their ‘compactness’ (Possingham et al. 2000; p. 299) by calculating the ratio of boundary length to area of sites prioritised for management under the $10 million budget. We also measured nearest-neighbour distances between the centre-points of management sites with the Geospatial Modelling Environment (http://www.spatialecology.com) ‘pointdistances’ function.

4.4 Results

4.4.1 Benefit dependency analysis

Managing only one of all potential threats to a species (our optimistic scenario) resulted in a higher expected return on investment when compared to managing threats independently (independent scenario) or managing all threats to achieve species recovery (pessimistic scenario) (Figure 4.1). When a $10 Million investment in this optimistic scenario was optimised, the expected benefit to species was 13% greater than if managing threats independently (Figure 4.1 ‘a’; optimistic versus independent scenario curves). Alternatively expressed, a $5.7 Million investment in managing

104

threats optimistically would return the same level of expected benefit as a $10 Million investment in managing threats independently (Figure 4.1 ‘b’).

Figure 4.1 Return-on-investment curves when optimising for independent threat management compared with optimising for dependencies, under optimistic and pessimistic management scenarios. The Y-axis is expressed as a proportion of the benefit expected with a $10 million budget when managing threats in the independent scenario. Letters label differences in expected benefit and investment when compared with assuming threats are independent and additive (a: 13%; b: - $4.3 Million; c: -5.6%; d: $2.6 Million).

If we need to manage all threats to achieve secure species (our pessimistic scenario) we would have a lower expected return on investment compared with managing threats independently (Figure 4.1). For a $10 Million investment, the expected benefit to species was nearly 5.6% lower in the pessimistic scenario than if we assumed independent benefits of threat management (Figure 4.1 ‘c’; pessimistic versus independent scenario curves). In other words, a $12.6 Million investment in managing all threats to species would return the same level of expected benefits as a $10 Million investment in independently managing threats (Figure 4.1 ‘d’).

Optimal strategies for threat mitigation differed between the scenarios both in spatial location and in the suite of actions undertaken at each site (Figure 4.2). Compared with the independent threats scenario, when investing $10 million a greater area was identified for single actions under the dependent, optimistic scenario (5 757 versus 3 866 km2), and a greater area was identified for

105

multiple actions under the dependent, pessimistic scenario (806 versus 772 km2) (Table 4.2). Strategies including fire management were prioritised as being the most cost-effective for the majority of sites chosen for action in all scenarios (Figure 4.2; Table 4.2). Using similarity maps we discovered that the sites and actions selected in the pessimistic scenario had a high level of agreement with those selected in the independent scenario (kunweighted = 0.91; Appendix 4D, Table 4D. 3). The sites and actions selected in the optimistic scenario had a lower level of agreement with those selected in the independent scenario (kunweighted = 0.73; Appendix 4D, Table 4D. 3). A value of 1 for kappa indicates perfect agreement, and a value of 0 indicates agreement is no better than expected by chance. Out of the independent, optimistic and pessimistic scenarios, the most similar pair were the pessimistic and independent scenarios (Figure 4.2f), and the most dissimilar pair were the pessimistic and optimistic scenarios (Figure 4.2d).

Figure 4.2 Maps illustrate where to manage, with which strategy*, when investing $10 Million in threat abatement. Sites across the landscape are prioritised where it is most cost-effective to act for the greatest expected benefit to species, either when assuming (a) that management benefit dependencies don’t exist (Independent scenario) or when accounting for dependencies in the expected benefit of threat management action in (b) Optimistic or (c) Pessimistic scenarios.

*Threat-abatement strategies include reduced fire intensity and frequency (Fire), invasive predator control (Fox), reduced grazing pressure of domestic stock (Grz), and combinations thereof. 106

Table 4.2 Area of management chosen for action varies under an independent versus a dependent (optimistic or pessimistic) management scenario when investing $10 million and different threat mitigation strategiesa. Also included are sums of: total counts, strategies include that include particular actions, and strategies of single and multiple actions. Total potential management area is 32,471 km2.

Number of Threat species Action area selected (km2) mitigation vulnerable to Threat management scenario strategy threat(s)b Independent Optimistic Pessimistic Fire 40 3,833 5713 3,665 Foxes 6 14 13 13 Grazing 14 20 31 21 Fire and foxes 2 333 103 315 Foxes and grazing 1 6 7 6 Grazing and fire 8 211 621 250 Fire, foxes and grazing 1 223 36 236

All action 72 4,639 6,523 4,504 No action 0 27,833 25,949 27,967

Fire+ 51 4,599 6,472 4,465 Foxes+ 10 575 158 569 Grazing+ 24 460 695 511

One action NA 3,866 5,757 3,699 Two actions NA 550 730 570 Multiple actions NA 772 766 806 aThreat-mitigation strategies include reduced fire intensity and frequency (Fire), invasive predator control (Foxes), reduced grazing pressure of domestic stock (Grazing), and combinations thereof bThreats to species include too frequent and intense fire, fox predation, and habitat degradation from grazing (Department of Environment and Resource Management 2010).

4.4.2 Spatially dependent cost analysis

Prioritising predator control investment in adjacent sites was marginally more cost-effective than ad-hoc management (Figure 4.3). In addition, the cost-dependent scenario was more spatially compact than the cost-independent scenario and the boundary length-to-area ratio was smaller (Appendix 4E, Table 4E. 1). Furthermore, when accounting for the spatial cost-dependency, for a 107

$10 million investment more sites were managed (1,994.25 km2 versus 1,806 km2) and the average distance (mean ± s.e.) to the nearest neighbouring site was smaller (582.52 ± 3.97 m versus 590.92 ± 4.58 m), although not statistically different (two-group t-test, p=0.1659; descriptive statistics Appendix 4E, Table 4E. 2).

Figure 4.3 Return-on-investment curve for cost-independent threat management (invasive predator control) compared with accounting for a spatial cost-dependency. Adjacent sites are less expensive to manage: As compared to investing $5 million in a cost-independent scenario, $350,000 less investment is needed (labelled ‘X’, above) for the same overall level of expected benefit.

4.5 Discussion

Our analysis focuses on how investment outcomes differ when optimising for the dependencies inherent in the benefits and costs of abating threats to species. Interrelationships are rarely addressed in the selection of cost-effective conservation actions (Duke et al. 2013). We defined a dependency as existing when the costs and/or benefits of one action interact non-additively with the costs and benefits of another action. Another way of saying this is that there is a dependency between two actions when the cost and/or benefit of one action changes when the other action is taken. This is almost always the case, but because accounting for dependencies can complicate an analysis and interpretation, and data on them are often lacking, such dependencies are often ignored. We develop and apply an approach for accounting for two different types of dependencies

108

in a typical spatial conservation prioritisation problem, to clarify the trade-offs that need be considered.

Our research shows that expected payoffs for investment decisions vary depending on whether or not dependencies are explicitly considered in the expected benefits of threat management. As may be expected, investment costs for threatened species management are lower if we optimistically expect to secure species by abating any one their threats (Figure 4.1). This might be a dangerous approach if in fact dependencies are present such that species can only be secured by managing all threats. In contrast, if we pessimistically expect that abating all threats is necessary to benefit threatened species the required investment to achieve conservation targets is greater, because expected benefits of managing one threat are unrealistic (Figure 4.1). Our results concur with previous findings that expected benefits may be overestimated when investing in abating only one threat to species if dependencies exist (Evans et al. 2011b). At the other extreme, spending an entire budget on managing multiple threats, when managing one threat or managing threats independently would be sufficient, runs the risk of managing some sites too intensively, and not leaving enough money for managing threats in other sites (Figure 4.1 and Figure 4.2; Table 4.2). Importantly, accounting for dependencies affects threat management investment strategies, as well as alters decisions about where to manage with which actions (Figure 4.2).

In a landscape in which 83% of the species are affected by a single threat (that might be different between species; Table 4.2), we found that accounting for dependencies optimistically changed the sites to be managed more than pessimistically assuming all threats must be managed to secure species (Figure 4.2). This is because optimistic management aims only to manage one of all the threats that might affect a species, so using cost-effectiveness analysis just chooses the least expensive sites and actions (Table 4.2 shows that more area has been selected for management under the optimistic scenario, but these sites are inexpensive fire management sites). In comparison, managing threats independently or pessimistically is more conservative. By assuming more threats must be managed to secure species, these scenarios place more value on strategies with multiple actions that require higher investment and are more restrictive in where in the landscape this can be done (Figure 4.2).

Investment decisions may also vary depending on whether spatial dependencies have been incorporated in the costs of threat management. We demonstrated a way of accounting for spatial- cost dependencies when investigating where to allocate funding across a landscape. However, we found that in our study area prioritising management in nearby sites was only marginally more cost- effective than not considering the spatial relationship of management costs. The modest differences

109

between the cost-independent and cost-dependent scenarios may be partly attributed to the degree of habitat fragmentation in our study region. If priority areas are small patch sizes and distances between potential management sites are greater than 5km, predator control effectiveness may be limited (McLeod et al. 2010). Our study shows that in a large number of cases managers can most likely achieve almost the same outcomes by ignoring spatial dependencies as including them, especially when budgets are low (in this study below $1 Million, Figure 4.3). Spatial cost- dependencies are likely to be more important in landscapes that are more connected. If managers are seeking spatial cost-dependency savings in highly-fragmented landscapes, we suggest that alternative actions not considered in this study, such as habitat restoration to restore connectivity may be also be important to consider as a threat-abatement measure in combination with predator control (Zavaleta et al. 2001).

Our analysis is among the first to account for dependencies in the context of investing in managing multiple species and threats, and the benefits and costs of management actions to abate those threats across a landscape at a fine scale. Although threats and their management can be additive in effects, they are more likely to be synergistic or antagonistic (Brook et al. 2008, Crain et al. 2008, Darling and Côté 2008), and successful conservation actions must address these interactions accordingly. Threats are synergistic when their combined effect is greater than would be expected from their additive effects and antagonistic when it is less. Assuming no interactions between the benefits of managing for two threats could lead to an incorrect identification of sites requiring management (Brown et al. 2013a). However, the possibility of either synergistic or antagonistic interactions between multiple threats (Paine et al. 1998, Folt et al. 1999) makes prioritising their management complicated (Sala et al. 2000). The consequences of ignoring synergisms, such as between climate change and habitat loss (Mantyka-Pringle et al. 2012) could be missed opportunities in areas where action could abate not only one threat, but also its interaction with another threat. The consequences of ignoring antagonisms could be negative outcomes if reducing one threat worsens another threat (Banks et al. 1998, Didham et al. 2007, Evans et al. 2011b, Brown et al. 2013b).

There are other positive and negative aspects of dependencies in conservation benefits and costs that we do not consider here. First, the expected benefit of managing threats to species that are mutualistic (Connor 1995) is greater than additive. For example, in our case study, reducing fire and grazing pressure benefits both the vine (Pararistolochia praevenosa) and the butterfly (Ornithoptera richmondia) that depends on it. Benefits to species are greater than expected when considering action complementarity, because more species are protected more efficiently (Underwood et al. 2008). Furthermore, the efficiency of the solution can be enhanced through achieving the greatest expected benefit to the most species for the least cost (Briggs 2009, Joseph et 110

al. 2009, Carwardine et al. 2012). Second, the effects of dependencies can be to the detriment of well-intended conservation actions. Threats can be displaced and the benefits of conservation action can be less than might be expected when land-supply and land-use restrictions increase the probability of conversion or effect on any remaining developable and unprotected areas - displacement, resulting in a net loss of environmental benefits (Wu 2000, Newburn et al. 2005, Armsworth et al. 2006, Oliveira et al. 2007, Ewers and Rodrigues 2008, Tóth et al. 2011).

Other dependencies affecting cost effectiveness not considered include when investment in one action could either increase or decrease the likelihood of success of another action (e.g. Nelson et al. 2009), and when habitat quality differentially contributes to population viability (Haight et al. 2002). These factors can be introduced as parameters into our equations, such as by weighting the benefit of an action for each species by the likelihood of the action delivering the desired outcome (Joseph et al. 2009). Furthermore, ideally all threats are considered when planning a conservation strategy to maximise the probability of success (Allan et al. 2013), and potentially multiple actions may be appropriate for a given threat (Carwardine et al. 2012, Carwardine et al. 2014). In our study we considered three major threats to species, and selected a single best action to address each threat. Other alternative actions are available, for example habitat restoration to design shelter for species from invasive predators. Although we did not consider these alternatives, our approach can be translated to consider multiple actions to address each threat by simply expanding the number of potential strategies and weighting these by the likelihood of each strategy to deliver the desired outcome (in this study, species security from threat). However, accounting for more actions requires more prior information on the likely benefits of each action for recovering each species, either from experts or empirical data (e.g. Carwardine (2014)). These data might not always be easy to collect. We explored two extreme cases of dependencies between management actions and assumed binary outcomes, but these outcomes are most likely a continuum. Decision makers can assess whether it is worth collecting more information on different potential actions to address each threat, or on the level of dependencies between the threats, using value-of-information analysis (e.g. Raiffa 1993, Polasky and Solow 2001). For future research, an alternative approach would be to construct a model of stochastic dependency of consequences (threat abatement) under assumptions of independence and non-independence using Bayesian Belief Networks. This analysis would provide the attribute of explicitly and transparently describing, via alternative models, how the management actions co-vary. Further this would yield testable (under adaptive management) predictions.

111

4.5.1 Conclusion

Our research explores a method of incorporating dependencies to enhance the reality of problem formulation for dealing with multiple threats that has been traditionally approached in an overly- simplistic manner. Threats determine the suite of actions needed for species management; effective conservation actions address the relative vulnerability of species and areas to threats (Wilson et al. 2005a, Pressey et al. 2007). The traditional assumption is that threats, and the effects of managing them, are additive (e.g. Halpern et al. 2008, Ban et al. 2010, Vörösmarty et al. 2010). However, particularly when it is known that two or more threats are affecting a system, it is likely that the effects of threats to most species are interactive (Sala et al. 2000, Darling and Côté 2008). When acting under the constraint of a limited budget, considering dependencies in the costs and expected benefits of managing interactive threats can make a difference to investment decisions. We recommend that taking into account the inherent dependencies of actions can allow the needs of threatened species to be more efficiently addressed.

4.6 Acknowledgements

This research was conducted with the support from the Australian Government’s National Environmental Research Program, the Australian Research Council Centre of Excellence for Environmental Decisions, and an Australia Postgraduate Award scholarship. KAW and HPP were supported by an Australian Research Council Future Fellowship and Laureate Fellowship respectively. The Queensland Government provided species presence data (WILDNET (Department of Environment and Resource Management (DERM) 2010) and HERBRECS (Department of Environment and Resource Management (DERM) 2011) databases). We thank J. Hanson, R. Dwyer, and C. Brown for key R coding insights, P. Leeson, S. Lloyd, F. Gibson, K. Lock, C. White, and Q. McCleod for guidance in costing fire-management, I. Chadès for math clarifications, and C. Brown for discussion on an early draft.

112

4.7 Supplementary Information

4.7.1 Appendix 4A Priority species to be managed and threats to their persistence.

Table 4A. 1 Listing of priority species and their vulnerability (1=Yes; 0=No) to the threats of fire frequency and intensity (Fire), invasive species predation (Fox), and domestic stock-caused habitat degradation (Grz) (Department of Environment and Resource Management 2010).

Priority species Common Name Taxa Kingdom Fire Fox Grz Taudactylus pleione Kroombit Tinkerfrog amphibian animal 1 0 0 Mixophyes iteratus Giant Barred Frog amphibian animal 0 0 1 Calyptorhynchus lathami Glossy Black-Cockatoo bird animal 1 0 0 lathami (E) Cyclopsitta diophthalma Coxen's Fig-Parrot bird animal 1 0 0 coxeni Dasyornis brachypterus Eastern Bristlebird bird animal 1 1 1 Esacus magnirostris Beach Stone-Curlew bird animal 0 1 0 Grantiella picta Painted Honeyeater bird animal 0 0 1 Ninox strenua Powerful Owl bird animal 1 0 0 Pezoporus wallicus Ground Parrot bird animal 1 0 0 wallicus Poephila cincta subsp Black-Throated Finch bird animal 0 0 1 cincta (White-Rumped ssp) Rostratula australis Australian Painted Snipe bird animal 0 0 1 Sterna albifrons Little Tern bird animal 0 1 0 Turnix melanogaster Black-Breasted Button- bird animal 1 0 1 Quail Ornithoptera richmondia Richmond Birdwing butterfly animal 1 0 0 Phyllodes imperialis Pink Underwing Moth butterfly animal 1 0 0 smithersi Maccullochella mariensis Mary River Cod fish animal 0 0 1 Dasyurus hallucatus Northern Quoll mammal animal 1 1 0 Dasyurus maculatus Spotted-Tailed Quoll mammal animal 1 1 0 (Southern ssp) Kerivoula papuensis Golden-Tipped Bat mammal animal 1 0 0 Petaurus australis Yellow-Bellied Glider mammal animal 1 0 0 australis (Southern ssp) Xeromys myoides Water Mouse mammal animal 0 1 1 Acanthophis antarcticus Common Death Adder reptile animal 1 0 1 Caretta caretta Loggerhead Turtle reptile animal 0 1 0 Delma torquata Collared Delma reptile animal 1 0 1 Elseya albagula White-Throated Snapping reptile animal 0 1 0 Turtle Elusor macrurus Mary River Turtle reptile animal 0 1 0 Lampropholis colossus Bunya Mountains reptile animal 1 0 0 Sunskink

113

Table 4A.1 Continued.

Priority species Common Name Taxa Kingdom Fire Fox Grz Natator depressus Flatback Turtle reptile animal 0 1 0 Strophurus taenicauda Golden-Tailed Gecko reptile animal 1 0 0 Pararistolochia Richmond Birdwing Vine climber plant 0 0 1 praevenosa Cycas megacarpa cycad plant 1 0 0 Macrozamia crassifolia cycad plant 1 0 0 Macrozamia lomandroides cycad plant 1 0 0 Macrozamia longispina cycad plant 1 0 0 Macrozamia parcifolia cycad plant 1 0 0 Macrozamia pauli- cycad plant 1 0 0 guilielmi Blandfordia grandiflora Christmas Bells herb plant 1 0 0 Macarthuria complanata herb plant 1 0 0 Picris conyzoides herb plant 0 0 1 Pratia podenzanae herb plant 0 0 1 Rhaponticum australe Native Thistle herb plant 1 0 1 Romnalda strobilacea n/a herb plant 0 0 1 Xerothamnella herbacea herb plant 0 0 1 Diuris parvipetala orchid plant 1 0 1 Phaius australis Lesser Swamp Orchid orchid plant 0 0 1 Sarcochilus weinthalii Blotched Sarcochilus orchid plant 1 0 0 Cyperus clarus sedge plant 0 0 1 Eleocharis blakeana sedge plant 0 0 1 Acacia attenuata Whipstick Wattle shrub plant 1 0 0 Acacia baueri subsp. Tiny Wattle shrub plant 1 0 0 Baueri Acacia eremophiloides n/a shrub plant 1 0 1 Acacia porcata n/a shrub plant 1 0 1 Acacia tingoorensis shrub plant 1 0 0 Apatophyllum olsenii shrub plant 1 0 0 Boronia keysii Key's Boronia shrub plant 1 0 0 Denhamia parvifolia Small-leaved Denhamia shrub plant 1 0 0 Homoranthus decumbens shrub plant 1 0 0 Lasiopetalum sp. (Proston shrub plant 1 0 0 JA Baker 17) Melaleuca groveana shrub plant 1 0 0 Phebalium distans shrub plant 1 0 0 Pomaderris clivicola shrub plant 1 0 0 Pomaderris shrub plant 1 0 0 coomingalensis Swainsona fraseri shrub plant 1 0 0 Zieria verrucosa shrub plant 1 0 0 Alectryon ramiflorus Isis Tamarind tree plant 1 0 1 Cadellia pentastylis Ooline tree plant 1 0 0 Callitris baileyi Bailey's Cypress tree plant 1 0 0 114

Table 4A.1 Concluded.

Priority species Common Name Taxa Kingdom Fire Fox Grz Corynocarpus rupestris tree plant 0 0 1 ssp arborescens Cossinia australiana Cossinia tree plant 1 0 0 Eucalyptus broviniensis tree plant 1 0 0 Eucalyptus pachycalyx ssp Shiny-Barked Gum tree plant 1 0 0 Waajensis Quassia bidwillii Quassia tree plant 1 0 0

4.7.2 Appendix 4B Detailed methods

4.7.2.1 Estimating independent and dependent expected management benefits

Our first aim was to assess dependencies between the expected benefits of management actions. We assume that all species vulnerable to the same threat would benefit from targeted action to abate that threat (Briggs 2009), at the site level (Boyd et al. 2008). The region comprises M sites indexed 푖 = 1 … 푀 and contains habitat for N threatened species indexed 푗 = 1 … 푁. There are P threats indexed 푘 = 1 … 푃 and the total number of threats to a species j is Kj. The expected benefit of abating one threat k to a species j at a site i is 푏푖푗푘 where 푏푖푗푘 ∈ {0,1}, with a value of 1 indicating that species j is vulnerable to the threat and would benefit from management.

We assume there are P possible actions at each site, e.g. P = 3 for grazing, fox, and fire management, with a one-to-one relationship between threats and actions. If P = 3, there are eight (= 2P) possible management strategies that can be carried out at a site, indexed q = 0 … 7 in this case. A management strategy q is a combination of actions at a site, i.e. a “project.” Assume 푞 =

{푌1, 푌2, 푌3} where ykq = 1 if we abate threat k in project q and Ykq = 0 if we don’t abate threat k in project q; for example project 푞 = 0 {0,0,0} is do nothing, 푞 = 7 is do everything {1,1,1}. Let Xiq be a control variable that tells us if we do project q at site i. When Xiq = 1 we carry out project q at site i, otherwise we do not.

In the independent benefit management scenario, the value vijq of each management strategy q for species j at site i, is the sum of benefits for management strategy q divided by the number of threats K to that species j.

푃 ∑푘=1 푏푖푗푘푌푘푞 푣푖푗푞 = 푋푖푞 (4-1) 퐾푗

115

The expected benefit is therefore weighted by the proportion of threats to a species, e.g. if a species is vulnerable to three threats, 퐾푗 = 3 and if the strategy is to manage all three threats at a site

3 푞 = 7 {1,1,1}, then 푣 = = 1. 푖푗푞 3

In some cases managing one threat will be good enough to deliver a major benefit to a species. The optimistic management scenario assumes that a species benefits significantly if any of its threats are managed in strategy q:

푣푖푗푞 = 1 when 푦푘푞 = 1 for any 푏푖푗푘 = 1 for all 푘 = 1 … 푃. (4-2)

Alternatively, the pessimistic management scenario assumes that a species benefits only if all of its threats are managed in strategy q:

푣푖푗푞 = 1 when 푦푘푞 = 1 whenever 푏푖푗푘 = 1 for all 푘 = 1 … 푃. (4-3)

Assume that if rij = 1, then site i is suitable for species j and if rij = 0 the site i is not suitable for species j. We calculate the total number of sites within the study region for every species 푗 as:

푀 푅푗 = ∑푖=1 푟푖푗 (4-4)

푟푖푗 We can use to weight the value vijq of doing each management strategy q at site i for a species j 푅푗 by the proportion of the potential distribution of the given species that the site encompasses. The benefit of an action to abate a threat becomes relatively small if one site is a small part of the mapped species distribution, but if one site is the entire mapped distribution of a species and the species is vulnerable to only one threat, the benefit of the action is large. The total benefit of enacting strategy 푞 at site i for n species is then defined as the sum of the weighted values vijq of that management strategy for each species j that occurs at that site:

푁 푟푖푗 퐵푖푞 = ∑푗=1 (푣푖푗푞 . ) . (4-5) 푅푗

The total benefit of a project at a site is thus derived from the management benefit to all species vulnerable to a particular threat or suite of threats at that site.

116

4.7.2.2 Estimating independent management costs

We estimated the cost of managing fire for biodiversity by accounting for variation in vegetation communities (Queensland Parks and Wildlife Service Fire Management System (QPWS) 2012), proximity to human habitation (Australian Bureau of Statistics (ABS) 2006), and approximations of prescribed burn and wildfire suppression costs in rural, semi-rural, and peri-urban settings (Appendix S3). We estimated the cost for invasive red fox control using a roadside-baiting strategy (Carter et al. 2011) with four 14-day baiting campaigns per year (Auerbach et al. 2014, Appendix C). We estimated the opportunity cost of entering into a stewardship agreement to cease grazing to be foregone profit from agriculture (Marinoni et al. 2012, Auerbach et al. 2014, Appendix B). We then calculated costs for 20 years of management using a low discount rate of 2% encouraging investment in management projects offering returns at distant dates (Johnson and Hope 2012), i.e. the benefits of intergenerational biodiversity persistence.

Assuming independence in costs, and that managing a site abates the threat for all species at the site, the cost of managing threat(s) k for all species j at site i using action strategy q is the sum of the costs of all actions implemented by that strategy:

푃 퐶푖푞 = 푋푖푞 ∑푘=1 퐶푖푘 푌푘푞 (4-6)

4.7.2.3 Estimating spatially-dependent costs

Our second aim was to explore the effect of a spatial dependency in the cost of a single- management action to abate one threat. We chose invasive predator management because until recently, most invasive fox control was carried out in an ad hoc manner, with little cooperation or coordination between managers (Gentle et al. 2007). More recently, both the frequency and spatial coverage of fox control has been found to be positively correlated with lamb (McLeod et al. 2010) and native animal (Western Shield, unpub. data) survival, indicating that spatial coordination of fox control would improve its effectiveness and likely reduce costs (Epanchin-Niell et al. 2009, Carter et al. 2011).

In the cost-dependent scenario, we assumed that if we chose to manage sites that are close together costs will be reduced. Here, we separated the fixed and variable costs of fox management and calculated the total cost 푍푖푞 of managing fox threat k with Action P using strategy q in site i as:

25−푧 푍 = 퐹 + 푉 ∙ ( ) (4-7) 푖푞 푖푞 푖푞 25

117

where 퐹푖푞 represents the fixed costs of bait purchase and labour for administering the bait. The variable costs 푉푖푞 include labour for driving between bait stations and vehicle operation. The variable 푧 is the number of neighbouring sites (in a 5 site x 5 site window surrounding the selected site) selected for fox control. This distance was chosen to reflect the 0.2 km minimum distance between properties found to facilitate effective fox-baiting among landholders (McLeod et al. 2010).

4.7.2.4 Prioritising sites and actions for investment

To assess the outcome of optimising for benefit-dependencies, we prioritised which management strategy would be most cost-effective to implement at each site under three benefit scenarios (independent, dependent/optimistic, or dependent/pessimistic) and using independent costs. We calculated the cost-effectiveness 푐푒푖푞 of abating threat(s) k using management action strategy q for each site i:

퐵푖푞 푐푒푖푞 = ⁄ . (4-8) 퐶푖푞

We then ranked all strategy-by-site combinations from highest to lowest 푐푒푖푞 value, i.e. the highest priority sites and management strategies are those that are most cost-effective under overall budget,

Maximise the net benefit across all species

푀 푁 7 ∑푖=1 ∑푗=1 ∑푞=0 푥푖푞 퐵푖푗푞 (푋), (4-9)

Subject to meeting our budget constraint

푀 7 ∑푖=1 ∑푞=0 푥푖푞퐶푖푞 (푋) ≤ 퐷, (4-10)

And the fact exactly one management strategy occurs at every place

7 ∑푞=0 푥푖푞 = 1 ∀ 푖. (4-11)

118

To assess the outcome of accounting for a spatial cost-dependency, we prioritised where fox management would be most cost-effective to implement under two cost scenarios (cost-independent and cost-dependent), using the independent benefit estimate. The prioritisation for the cost- independent scenario was also calculated using equation 9, but with only one choice (fox management) for action A at each site. For the cost-dependent scenario, we re-calculated the cost- effectiveness 푐푒푖푞 of abating fox threat k using management strategy q at each site i by dividing the benefit, 퐵푖푞 (Equation 5), by the spatially-dependent cost 푍푖푞 :

퐵푖푞 푐푒푖푞 = ⁄ (4-12) 푍푖푞 and selected the most cost-effective site as being the highest priority. Next we re-calculated the cost

푍푖푞 of managing surrounding sites (Equation 7) by reducing the cost of managing sites near the selected site. We then re-ranked the remaining sites on the basis of their cost-effectiveness, before selecting the next-highest ranked site and iteratively repeated this procedure until all sites were allocated a cost-effectiveness ranking.

4.7.3 Appendix 4C Model of management costs to maintain fire regimes for biodiversity

4.7.3.1 Objective

The objective was to estimate the cost of fire management for biodiversity in the Burnett-Mary NRM over a period of 20 years.

4.7.3.2 Summary

Method takes into account recommended fire frequency (based on vegetation fire regime), number of structures surrounding each 25-ha management unit, and prescribed fire cost estimates in rural, semi-rural, and peri-urban areas. Net present value (NPV) of fire management for the 20-year period is based on per annum cost (modelled cost divided by 20) and a discount rate of 2%.

4.7.3.3 Justification

Individual species of flora and fauna can be threatened by fire regimes (interval time and intensity) inappropriate for their life cycles: too much fire can be damaging, as can too little fire. Fire affects both individuals of a species and their habitats. Too much fire can directly kill individuals or indirectly affect them by destroying habitat--cover and food sources. Too little fire can result in reduction of fire-dependent habitat plants and too much growth of invasive plants that crowd out desirable food and cover plants or that can result in overly intense fires of accumulated biomass.

119

Individual flora species can be fire-intolerant, or fire-dependent: too much fire can destroy individuals; too little fire may stunt the life cycle of individuals if they need fire to reproduce.

4.7.3.4 Background

Interest in managing Australian landscapes for biodiversity conservation using fire is growing. Historical evidence suggests Aboriginal fire management was complex and intended to encourage growth of sustenance-providing species. More recently, arson has been a source of damaging fires, climate change is expected to escalate fire frequency, invasive plant growth has amplified fire intensity, and the potential for human and structural collateral fire damage is of increasing concern (Bradstock et al. 2002, Bradstock et al. 2012).

4.7.3.5 Method

We estimate the cost of managing fire for biodiversity by spatially accounting for vegetation type (Queensland Parks and Wildlife Service Fire Management System (QPWS) 2012), proximity to human habitation (Australian Bureau of Statistics (ABS) 2006), and approximations of prescribed burn and wildfire suppression costs in rural, semi-rural, and peri-urban settings.

To account for vegetation, we first determined the number of prescribed fires needed in a 20 year period based on the vegetation fire regime group recommendation. For vegetation in fire regime groups that should not be burned, we assumed wildfire suppression would be needed once in 20 years.

To account for variation in fire management costs depending on proximity to human habitation, we calculated the number of dwellings (structures) per ha from ABS Mesh Block data, then summed the number of structures per 25 ha management unit by aggregating the data. We sourced the cost of prescribed burns in three representative regions, and therefore needed to estimate the number of structures characteristic of rural, semi-rural, and peri-urban settings. To do this, we created point locations for the three sites for which we had prescribed fire cost data, and determined the number of structures within a 25-km radius buffer of each of the three sites. We then fitted a curve between the estimated cost of prescribed fire per ha for the three locations and the number of structures in their respective buffer areas. Next, we used focal statistics (ArcGIS spatial analyst) to determine the number of structures in a 25-km buffer similar surrounding each management unit. (Note: We separately analysed Fraser Island and the mainland, because their physical separation would prevent fire spread. We combined the analyses in the end). We then calculated the management cost for one fire in each management unit, using the equation determined from the cost/structure curve that

120

represents a per-hectare cost of a prescribed fire (based on the number of structures in a 25-km radius buffer). We then multiplied this value by 25 for each 25-ha management unit.

We estimated that wildfire suppression was 9% greater in cost than prescribed burns, based upon estimated rural wildfire suppression costs relative to estimated rural prescribed burn costs.

We calculated the cost of fire management for biodiversity over a 20-year period in each management unit by multiplying the one-fire cost by the recommended 20-year burn frequency. We multiplied by 1.09 the management units that were predicted to need wildfire suppression to represent the greater cost estimated for wildfire suppression relative to prescribed fire.

Finally, we determined the NPV cost of administering fire for biodiversity over a 20 year period with a discount rate of 2% on the basis of the cost of fire management per annum (we divided the 20-year fire cost by 20).

4.7.3.6 Model caveats and acknowledgements

Fire management cost estimates were modelled with best knowledge at the time of writing, but the model should not be used for prescriptive purposes. We acknowledge Peter Leeson (Queensland Parks and Wildlife Service), Samantha Lloyd (Southeast Queensland Fire and Biodiversity Consortium), Fiona Gibson (University of Western Australia), Kerrie Lock (South Queensland Department of Defence), Chris White (Anchor Point Group LLC), and Quinn McCleod (Rocky Mountain Fire) for generously offering suggestions on approaches for fire management costing.

121

4.7.4 Appendix 4D Analyses for assessing similarity between scenarios with and without considering benefit-dependencies in threat management

Table 4D. 1 Fractional weight matrix: Used to count disagreements differentially when calculating weighted kappa statistic (Cohen 1968). Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and are set to zeroes. Off-diagonal cells contain weights indicating differing degrees of disagreement, where a value of one indicates total disagreement and values between zero and one indicate partial agreement.

Mgmnt Fire & Fox & Grz & Fire, Fox Strategy* Fire Fox Grz Fox Grz Fire & Grz None Idx 1 2 3 4 5 6 7 8 Fire 1 0 1.00 1.00 0.50 1.00 0.5 0.67 1.00 Fox 2 1.00 0 1.00 0.50 0.50 1.00 0.67 1.00 Grz 3 1.00 1.00 0 1.00 0.50 0.50 0.67 1.00 Fire & Fox 4 0.50 0.50 1.00 0 0.05 0.50 0.33 1.00 Fox & Grz 5 1.00 0.50 0.50 0.50 0 0.05 0.33 1.00 Grz & Fire 6 0.50 1.00 0.50 0.50 0.5 0 0.33 1.00 Fire, Fox,Grz 7 0.67 0.67 0.67 0.33 0.33 0.33 0 1.00 None 8 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0

*Threat-management strategies include reduced fire intensity and frequency (Fire), invasive predator control (Fox), reduced grazing pressure of domestic stock (Grz), and combinations thereof

122

Table 4D. 2 Binary weight matrix: Used to count disagreements differentially when calculating a weighted kappa statistic (Cohen 1968). Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and are set to zeroes. Off-diagonal cells contain weights indicating differing degrees of disagreement, where a value of one indicates total disagreement. For a relaxed assessment of partial agreement, values of even partial agreement in actions were also set to zero.

Management Fire & Fox & Grz & Fire, Fox strategy* Fire Fox Grz Fox Grz Fire & Grz None idx 1 2 3 4 5 6 7 8 Fire 1 0 1 1 0 1 0 0 1 Fox 2 1 0 1 0 0 1 0 1 Grz 3 1 1 0 1 0 0 0 1 Fire & Fox 4 0 0 1 0 0 0 0 1 Fox & Grz 5 1 0 0 0 0 0 0 1 Grz & Fire 6 0 1 0 0 0 0 0 1 Fire, Fox & Grz 7 0 0 0 0 0 0 0 1 None 8 1 1 1 1 1 1 1 0

*Threat-management strategies include reduced fire intensity and frequency (Fire), invasive predator control (Fox), reduced grazing pressure of domestic stock (Grz), and combinations thereof

Table 4D. 3 To evaluate solution similarity while taking chance agreement into account, Cohen’s un-weighted and weighted kappa statistic (Cohen 1960, 1968) was calculated for comparisons between scenarios.

Kappa Kappa Kappa Scenario comparison Na (unweighted) (weighted)b (weighted)c Optimistic versus Independent 129884 .73 .77 .80 Pessimistic versus Independent 129884 .91 .92 .92 Optimistic versus Pessimistic 129884 .72 .76 .78 a number of management units. bfractional weight matrix (Table 4D.1) c binary weight matrix (Table 4D.2)

123

Table 4D. 4 Confusion Matrix: Independent versus Pessimistic Scenario

Independent actions (columns); Pessimistic actions (rows) 1 2 3 4 5 6 7 8 sum 1 13626 0 0 56 0 195 0 1453 15330 2 0 47 0 7 0 0 0 0 54 3 0 0 73 0 0 3 0 5 81 4 76 4 0 1196 0 1 6 48 1331 5 0 0 0 0 22 0 1 0 23 6 0 0 0 0 0 799 45 0 844 7 0 0 0 0 0 0 891 0 891 8 959 0 9 0 0 0 0 110362 111330 sum 14661 51 82 1259 22 998 943 111868 129884 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

Table 4D. 5 Confusion Matrix: Independent versus Optimistic Scenario

Independent actions (columns); Optimistic actions (rows) 1 2 3 4 5 6 7 8 sum

1 14510 0 0 51 0 769 0 0 15330 2 0 29 0 24 0 0 1 0 54 3 0 0 58 0 6 17 0 0 81 4 920 0 0 330 0 73 8 0 1331 5 0 0 0 0 14 0 9 0 23 6 0 0 0 0 3 825 16 0 844 7 0 0 0 0 6 776 109 0 891 8 7422 21 66 5 0 22 0 103794 111330 sum 22852 50 124 410 29 2482 143 103794 129884 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

124

Table 4D. 6 Confusion Matrix: Optimistic versus Pessimistic Scenario

Optimistic actions (columns); Pessimistic actions (rows) 1 2 3 4 5 6 7 8 sum

1 14069 0 0 847 0 0 0 7936 22852 2 0 29 0 0 0 0 0 21 50 3 0 0 60 0 0 0 0 64 124 4 37 22 0 318 0 0 0 33 410 5 0 0 1 0 14 3 6 5 29 6 555 0 21 85 0 982 824 15 2482 7 0 0 0 9 8 13 113 0 143 8 0 0 0 0 0 0 0 103794 103794 sum 14661 51 82 1259 22 998 943 111868 129884 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

Table 4D. 7 Percentage of manageable area in agreement of location and action

Disagree Agree in part Agree Agree no action Independent versus Optimistic 5.80 2.06 12.22 79.91 Independent versus Pessimistic 1.90 0.30 12.82 84.97 Optimistic versus Pessimistic 6.22 1.87 12.00 79.91

125

4.7.5 Appendix 4E Analyses to determine difference in spatial configuration of sites selected for fox management under a budget of $10 million, without and accounting for a spatial cost dependency.

Table 4E. 1 Compactness measure for sites chosen for cost-effective fox management action under a budget of $10 million calculated with a compactness ratio (Possingham et al. 2000; p. 299). For this ratio, the boundary length of the selected management areas is compared to the circumference of a circle, the most compact shape possible, with the same area as the selected management areas.

푏표푢푛푑푎푟푦 푙푒푛𝑔푡ℎ 푐표푚푝푎푐푡푛푒푠푠 푟푎푡푖표 = 2√휋 푥 푎푟푒푎

boundary length (km),

i.e. summed boundary length boundary length compactness 2 solution perimeters area (km ) /area 2√π x area ratio

Without spatial 8135.00/1433.40 8135.00

dependency 8135.00 1433.40 = 5.67 2√π x 1433.40 60.61

With spatial 6231.31/1529.63 6231.31

dependency 6231.31 1529.63 = 4.07 2√π x 1529.63 44.94

R1 = 21.36 c = 134.141 134.141/1433.40 = .094 8135/134.141= 60.6

R2 = 22.06 c = 138.537 138.537/1529.63 = .091 6231.31/138.537 = 44.9

126

Table 4E. 2 Descriptive statistics for nearest-neighbour distance of sites chosen for cost-effective fox management action under a budget of $10 million.

a) Without spatial cost-dependency

Distance to nearest neighbour (m) n sites 7224 min 500 max 12020.82 range 11520.82 sum 4268807.84 median 500 mean 590.92 SE.mean 4.58 CI.mean.0.95 8.98 var 151526.38 std.dev 389.26 coef.var 0.66

b) With spatial cost-dependency

Distance to nearest neighbour (m) n sites 7977 min 500 max 12020.82 range 11520.82 sum 4646778.30 median 500 mean 582.52 SE.mean 3.97 CI.mean.0.95 7.78 var 125739.62 std.dev 354.60 coef.var 0.61

127

4.7.6 Appendix 4F Descriptive statistics for expected benefits (B), costs (C), and cost- effectiveness (CE) of threat management for five scenarios.

Table 4F. 1 Independent benefit management scenario

B C CE nbr.val 316779 316779 316779 min 0.000004 82.4 1.01E-10 max 0.125023 461803.9 0.000399 range 0.125019 461721.5 0.000399 sum 60.40262 4.03E+08 0.231587 median 0.000033 794.7 1.47E-07 mean 0.000191 1270.6 7.31E-07 SE.mean 2.2E-06 8.9 8.4E-09 CI.mean.0.95 4.3E-06 17.5 1.65E-08 var 1.5E-06 25352662 2.2E-11 std.dev 0.001226 5035.1 4.73E-06 coef.var 6.430301 4 6.469252

Table 4F. 2 Optimistic benefit management scenario

B C CE nbr.val 210254 210254 210254 min -0.0002 -1.1E-11 1.02E-10 max 0.125038 461803.9 0.000399 range 0.125236 461803.9 0.000399 sum 62.90318 2.04E+08 0.235583 median 8.31E-05 380.9317 2.73E-07 mean 0.000299 972.0369 1.12E-06 SE.mean 3.4E-06 13.33013 1.34E-08 CI.mean.0.95 6.8E-06 26.12672 2.62E-08 var 2.5E-06 37360514 3.8E-11 std.dev 0.001581 6112.325 6.14E-06 coef.var 5.284399 6.288161 5.476068

128

Table 4F. 3 Pessimistic benefit management scenario

B C CE nbr.val 303486 303486 303486 min 7.7E-06 82.6 1.01E-10 max 0.125008 461803.9 0.000399 range 0.125 461721.3 0.000399 sum 57.57424 3.98E+08 0.205997 median 3.71E-05 934.3 1.36E-07 mean 0.00019 1311.2 6.79E-07 SE.mean 2.3E-06 9.3 8.52E-09 CI.mean.0.95 4.5E-06 18.3 1.67E-08 var 1.6E-06 26410506 2.2E-11 std.dev 0.001254 5139.1 4.7E-06 coef.var 6.610343 3.9 6.918262

Table 4F. 4 Cost-independent scenario

B C CE

nbr.val 118538 118538 118538 min 4.14E-06 974.5 4.82E-10 max 0.017583 11772.2 1.74E-05 range 0.017578 10797.7 1.74E-05 sum 6.737381 2.20E+08 0.004381 median 4.14E-06 1405.3 3.46E-09 mean 5.68E-05 1857.3 3.70E-08 SE.mean 1.36E-06 3.2 8.80E-10 CI.mean.0.95 2.68E-06 6.4 1.73E-09 var 2.20E-07 1251314 9.20E-14 std.dev 0.00047 1118.6 3.03E-07 coef.var 8.267874 0.6 8.200924

129

Table 4F. 5 Cost-dependent scenario

B C CE nbr.val 118538 118538 118538 min 4.14E-06 927.91 5.41E-10 max 0.017583 11690.92 1.74E-05 range 0.017578 10763.01 1.74E-05 sum 6.737381 1.71E+08 0.004858 median 4.14E-06 950.09 4.46E-09 mean 5.68E-05 1445.26 4.10E-08 SE.mean 1.36E-06 3.02 9.19E-10 CI.mean.0.95 2.68E-06 5.92 1.80E-09 var 2.20E-07 1080344 1.00E-13 std.dev 0.00047 1039.4 3.16E-07 coef.var 8.267874 0.72 7.716864

4.7.7 Appendix 4G Descriptive statistics for management cost layers

Table 4G. 1 Management costs: reduced fire intensity and frequency (Fire), invasive predator control (Fox), reduced grazing pressure of domestic stock (Grz)

Fire Fox Grz nbr.val 220612 210023 144279 nbr.na 230188 240777 306521 min 82.4 966.2 163.5 max 14310.7 11772.19 461803.9 range 14228.2 10805.99 461640.4 sum 89863534.5 373193138.3 308556315 median 204.7 1372.15 551 mean 407.3 1776.92 2138.6 SE.mean 1.3 2.37 31.8 CI.mean.0.95 2.6 4.64 62.2 var 379985.4 1175803.93 145506247.3 std.dev 616.4 1084.34 12062.6 coef.var 1.5 0.61 5.6

4.7.8 Appendix 4H The variation in the expected benefits and costs of threat mitigation at different budgets

Variation in spatial locations chosen for action accounts for most of the differences in benefits between independent versus dependent threat management, especially under a low investment budget in optimistic threat management (14% of expected benefit, $2.5 million budget; Fig. 4H.1a,

130

different sites and/or actions) and to a lesser degree for pessimistic threat management (Fig. 4H.1b). Although expected benefits are different if managing for threats independently versus accounting for dependencies (as above), the cost of threat action at a particular site is the same amount in all scenarios. However, in the optimistic scenario where species benefit from management of one threat, 47% ($23.5 million) of a $50 budget would be spent unnecessarily (Fig. 4H.1c) for the same expected benefit from investing in independent threat management. In the pessimistic management scenario where all threats must be managed, 56% ($27.8 million) additional investment to the $50 million budgeted for independent threat management would be required for the same expected benefit of threat mitigation.

Figure 4H. 1 The variation in the expected benefits and costs of threat mitigation at different budgets is graphed when comparing the outcomes of dependent vs independent threat management. Benefits vary both in spatial location and also within the same action strategy undertaken at each site. The Y-axes of (a) and (b) are expressed as a proportion of the benefit expected when managing threats in the independent scenario under different budgets. ‘Missed opportunities’ for management (a) are revealed when evaluating outcomes of the optimistic vs the independent scenario and ‘Inflated expectations’ of management (b) when assessing outcomes of the pessimistic vs the independent scenario. Management costs are different when additional sites or actions are prioritised in the dependency scenarios (c). As compared to independent threat management, ‘Overinvestment’ is a risk if species may benefit from management of one threat, and ‘Underinvestment’ is a risk if actually all threats must be managed for a positive outcome.

131

4.7.9 Appendix 4I Sensitivity analysis with greater number of species with multiple threats

To examine the implications of dependencies between the benefits of management actions when more species are vulnerable to multiple threats, we conducted an additional sensitivity analysis scenario. In the sensitivity analysis, all species distributions are identical to the original analysis, but in contrast, the vulnerability of all the plants was changed to be to both fire and grazing threats (Figure 4I. 1). Results show that there is a greater difference in expected return on investment when management actions are assumed to be independent as compared to if there are actually dependencies between management benefits (Figure 4I. 2). Managing only one of all potential threats to a species (our optimistic scenario) resulted in a higher expected return on investment when compared to managing threats independently (independent scenario) or managing all threats to achieve species recovery (our pessimistic scenario) (Figure 4I. 2). When an investment of $10 Million in this optimistic scenario was optimised, the expected benefit to species was 50% greater than if managing threats independently (Figure 4I. 2‘a’; optimistic versus independent scenario curves). A $2.2 Million investment in managing threats optimistically would return the same level of expected benefit as a $10 Million investment in managing threats independently (Figure 4I. 2‘b’). If we need to manage all threats to achieve secure species (our pessimistic scenario) we would have a lower expected return on investment compared with managing threats independently (Figure 4I. 2). For a $10 Million investment, the expected benefit to species was 31% lower in the pessimistic scenario than if we assumed independent benefits of threat management (Figure 4I. 2‘c’; pessimistic versus independent scenario curves). We would require a $43.2 Million investment in managing all threats to species to return the same level of expected benefits as a $10 Million investment in independently managing threats (Figure 4I. 2‘d’). Optimal strategies for threat mitigation differed between the scenarios both in spatial location and in the suite of actions undertaken at each site (Figure 4I. 3). We found that the sites and actions selected in the pessimistic scenario had a lower level of agreement with those selected in the independent scenario in the sensitivity analysis, as compared to the original analysis (original analysis kunweighted = 0.91; sensitivity analysis kunweighted = 0.51; Table 4I. 1). The similarity in sites and actions selected in the optimistic scenario as compared to those selected in the independent scenario was equivalent in the sensitivity and original analyses

(kunweighted = 0.73; Table 4I. 1). In contrast to the original analysis, in the sensitivity analysis, the optimistic and independent scenarios were the most similar pair (Figure 4I. 3e), and the pessimistic and optimistic scenarios were the most dissimilar pair (Figure 4I. 3d).

132

Figure 4I. 1 Threats versus number of species in original analysis and in sensitivity analysis.

Figure 4I. 2 Sensitivity analysis return-on-investment curves when optimising for independent threat management compared with optimising for dependencies, under optimistic and pessimistic management scenarios. The Y-axis is expressed as a proportion of the benefit expected with a $10 million budget when managing threats in the independent scenario. Letters label differences in expected benefit and investment when compared with assuming threats are independent and additive (a: 50%; b: -$7.8 Million; c: -31%; d: $33.2 Million).

133

Figure 4I. 3 Sensitivity analysis maps illustrate where to manage, with which strategy*, when investing $10 million in threat abatement. Sites across the landscape are prioritised where it is most cost effective to act for the greatest expected benefit to species, either when assuming (a) that management benefit dependencies don’t exist (Independent scenario) or when accounting for dependencies in the expected benefit of threat management action in (b) Optimistic or (c) Pessimistic scenarios.

*Threat-abatement strategies include reduced fire intensity and frequency (Fire), invasive predator control (Fox), reduced grazing pressure of domestic stock (Grz), and combinations thereof.

Table 4I. 1 To evaluate solution similarity while taking chance agreement into account, Cohen’s un- weighted and weighted kappa statistic (Cohen 1960, 1968) was calculated for comparisons between scenarios for the sensitivity analysis.

Kappa Kappa Kappa Scenario comparison na (unweighted) (weighted)b (weighted)c Optimistic versus Independent 129886 0.73 0.76 0.78 Pessimistic versus Independent 129886 0.51 0.54 0.57 Optimistic versus Pessimistic 129886 0.40 0.44 0.47 a number of management units. bfractional weight matrix (Table 4D. 1) c binary weight matrix (Table 4D. 2)

134

Table 4I. 2 Confusion Matrix: Independent versus Pessimistic Scenario

Columns: Independent actions; Rows: Pessimistic actions

1 2 3 4 5 6 7 8 sum 1 4451 0 0 354 0 754 9 7852 13420 2 0 25 0 21 0 0 0 0 46 3 0 0 5 0 4 62 3 42 116 4 151 5 0 304 0 39 64 604 1167 5 0 0 0 0 9 0 10 0 19 6 0 0 0 0 0 913 252 0 1165 7 0 0 0 0 0 0 975 0 975 8 1897 24 17 38 3 160 8 110831 112978 sum 6499 54 22 717 16 1928 1321 119329 129886 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

Table 4I. 3 Confusion Matrix: Independent versus Optimistic Scenario

Columns: Independent actions; Rows: Optimistic actions 1 2 3 4 5 6 7 8 sum 1 13270 0 0 69 0 81 0 0 13420 2 0 41 0 5 0 0 0 0 46 3 0 0 102 0 5 9 0 0 116 4 797 0 0 360 0 9 1 0 1167 5 0 0 0 0 19 0 0 0 19 6 0 0 0 1 5 1159 0 0 1165 7 0 0 0 0 9 840 126 0 975 8 7239 7 639 21 1 0 0 105071 112978 sum 21306 48 741 456 39 2098 127 105071 129886 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

135

Table 4I. 4 Confusion Matrix: Optimistic versus Pessimistic Scenarios

Columns: Optimistic actions; Rows: Pessimistic actions 1 2 3 4 5 6 7 8 sum 1 5312 0 0 423 0 795 46 14730 21306 2 0 32 0 16 0 0 0 0 48 3 0 0 6 0 3 102 2 628 741 4 52 5 0 247 0 3 17 132 456 5 0 0 0 0 10 2 27 0 39 6 0 0 0 0 0 1002 1096 0 2098 7 0 0 0 0 0 0 127 0 127 8 1135 17 16 31 3 24 6 103839 105071 sum 6499 54 22 717 16 1928 1321 119329 129886 1 (Fire); 2 (Fox); 3 (Grazing); 4 (Fire & Fox); 5 (Fox & Grazing); 6 (Grazing & Fire); 7 (Fire, Fox, & Grazing)

136

Chapter 5 Accounting for complementarity and action dependencies in spatial threat-management prioritisation

5 Accounting for complementarity and action dependencies in spatial threat-management prioritisation

5.1 Abstract

Resources for managing threats to species are limited. Basic cost-effectiveness analysis provides decision support by prioritising where to act for abating threats within a budget. However, management prioritised by cost effectiveness may be inefficient in meeting the needs of all species because sites are chosen simply by rank order, regardless of which species have been managed in higher ranked sites ignoring the conservation planning principle of complementarity. A more complex approach will minimise cost subject to the constraint that all species benefit because management sites are chosen depending on whether species are being managed in previously selected areas. The two approaches should result in different priorities. This research assesses how priorities for threat management vary between a simple cost-effectiveness and a more complex complementarity approach. We evaluate differences in spatial prioritisations of threat-abating actions for a bio-diverse natural resource management region of Australia by comparing scenarios derived using a basic cost-effectiveness ranking approach as compared to a more complex cost- minimising optimisation approach. We also compare two scenarios where we consider that the benefits of threat management are likely to be non-additive: an optimistic situation where species benefit from managing any threat to that species in isolation, as compared to a pessimistic situation where all threats must be abated before species benefit. Management priorities determined from basic cost-effectiveness analysis may under-protect some species of concern, because the focus is on intensive threat management in cost-effective, species-rich, rare habitats. In contrast, a cost- minimising approach to prioritising threat management in sites that depend on all species benefitting requires a more distributed land area be managed, thus different priority sites. In addition, prioritising comprehensive threat management for all species was twice the price of optimistically accepting that sometimes desired conservation outcomes will fail due to insufficient management action. Trade-offs between the costs and benefits of threatened species management are inevitable. Our research will interest those with limited budgets who must prioritise investments in threat management.

5.2 Introduction

Financial needs for biodiversity conservation are unmet (Bakker et al. 2010, McCarthy et al. 2012, Waldron et al. 2013), yet the need for abating threats to an increasing number of species with conservation significance is pressing (Vié et al. 2009). Because resources are limited, finding the 139

most cost-effective and efficient sites and actions for threat abatement establishes where to prioritise management (Wilson et al. 2009b). Decision makers faced with choosing between actions, species, or sites for threat management also have to choose which tool to use to maximise conservation benefits for biodiversity – these tools differ in their complexity and their ability to adequately protect biodiversity assets.

Basic cost-effectiveness analysis (CEA) prioritises conservation actions by ranking the ratio between the expected benefit of a conservation action (expressed in non-monetary terms) and the financial cost of that action (Metrick and Weitzman 1998, Weitzman 1998). By including conservation costs in priority-setting, CEA is transparent in evaluating how restricted budgets may be invested (Wilson et al. 2006, Murdoch et al. 2007). CEA is evolving as a systematic conservation prioritisation approach because it is efficient, repeatable and transparent (Hughey et al. 2003, Naidoo and Adamowicz 2006, Moran et al. 2010, Laycock et al. 2011). Managers may be more likely to have confidence their investments are well-placed when prioritisations also include additional management values such as likelihood of conservation success (Briggs 2009, Joseph et al. 2009, Carwardine et al. 2012, Carwardine et al. 2014), making certain all species are managed by choosing sites that efficiently represent all species (species complementarity) (Underwood et al. 2008), or managing threats efficiently (Chapter 4).

Although prioritisation approaches such as CEA can successfully produce a ranked list of options for decision makers, they do not fully accommodate the non-additive costs and benefits of actions within and between sites. The principle of complementarity is an important management value that commonly underpins spatial prioritisation efforts (Kirkpatrick 1983, Margules and Pressey 2000) because overall, it is a way to efficiently represent each conservation feature while avoiding redundancies (Vane-Wright et al. 1991, Tulloch et al. 2013a, Bennett et al. 2014). Species complementarity is a core feature of systematic conservation planning software that identifies conservation priorities at least cost (Moilanen 2007, Ball et al. 2009) in environments from marine to terrestrial (e.g. Fernandes et al. 2005, Kremen et al. 2008). Complementarity has been integrated into more comprehensive CEA for identification of protected areas (Underwood et al. 2008, Wilson et al. 2010, Withey et al. 2012); but is not commonly considered in basic CEA (Murdoch et al. 2007). A basic CEA is simpler to compute than more complex cost-minimising approaches that incorporate complementarity, but is still flexible in its capacity to fulfil specific objectives such as maximising benefit across all species while localising actions in species-rich, rare habitats (Auerbach et al. 2014 (Chapter 3)).

140

Landscape-scale prioritisations of conservation actions are likely to diverge if they do, or do not include complementarity (Chadès et al. 2014). A more complex approach to prioritising action that incorporates complementarity in least-cost management solutions will address the needs of more species than a basic CEA, but is arguably less transparent. Evaluating the differences between priorities detected in a basic and a more complex prioritisation method would clarify the trade-offs of two alternative management approaches that could be taken when acting under a limited budget. One of the approaches prioritises actions against threats for all species, while the other does not explicitly account for managing all species.

While complementarity is the most common dependency that is addressed in systematic conservation planning (Justus and Sarkar 2002), additional dependencies in conservation actions that may affect management priorities are often overlooked. For instance, while it is commonly assumed that threats are independent (Halpern et al. 2008, Ban et al. 2010, Vörösmarty et al. 2010), this approach disregards evidence that dependencies between threats may affect the outcomes of management (Evans et al. 2011b, Regan et al. 2011, Brown et al. 2013b, Brown et al. 2013a). Mitigating one threat at a site where species are affected by multiple threats might have no effect at all. For example, both invasive foxes and rabbits must be managed where they occur together, or threatened species may still decline (Evans et al. 2011b). In other cases, mitigating one of many threats can have a positive effect on species that are threatened. For example, appropriate fire management combats disease in grass trees (Xanthorrhoea resinosa Pers.) (Regan et al. 2011) and reducing fertilised crop runoff benefits sea-grasses that are also at risk from climate change (Brown et al. 2013a, Brown et al. 2013b). Dependencies in management actions are non-additive, because species may not benefit equally from each threat that is managed.

Considering some of the many dependencies between the costs and benefits of conservation actions will probably change priorities for threat management. A prioritisation approach that chooses sites complementing those that are already selected ensures that all species are managed addresses the principle of complementarity, which is a dependency in the benefits of actions within and between sites. Similarly, other dependencies in the cost and benefits of threat management will affect priorities for conservation action, but are rarely considered in systematic planning. An approach assuming that managing threats independently will secure species could be overly-optimistic, and much greater investment may be needed to abate all the threats necessary to secure a species. Alternatively assuming that managing all threats to species is necessary because management success is dependent on what other actions are also taken could be overly-pessimistic, because less investment may be needed. Making explicit how alternate approaches to prioritising management

141

actions differ when dependencies in the benefits of those actions are taken into account will clarify trade-offs in the costs and benefits of threat management.

5.2.1 Research aims

Our research in this chapter aims to answer two questions about prioritising threat management while considering the costs of management actions and dependencies in the benefits of threat actions: First, how do two approaches to prioritising threat management differ when one approach accommodates a dependency (complementarity) between the benefits of managing different sites, so that all species are managed? Second, how does including an additional dependency between the benefits of threat management actions (all threats must be managed for a species to benefit) change priorities? We use a regional-area case study that prioritises three threat-abating actions for species with conservation significance at a fine-scale. Understanding how dependencies alter management choices will lead to clearer decisions.

5.3 Methods

We evaluated differences between threat-action prioritisation strategies derived from cost- effectiveness analysis (Metrick and Weitzman 1998, Weitzman 1998) that did not account for dependencies in benefits (Chapter 4), and Marxan with Zones (Watts et al. 2009) that does account for dependencies in benefits (hereafter referred to as CEA and MZA respectively). We used a case study of managing threats to species in the Burnett-Mary Natural Resource Management (NRM) region in southeast Queensland, Australia. We divided the approximately 55,000 km2 region into management planning units only in areas with remaining or regrowth native vegetation habitat (Queensland Herbarium 2010a, b). Management units were 25-ha in size for the CEA (n = 129,894), but were aggregated into 100-ha or smaller management units for the MZA (n = 44,373), enabling the Marxan analyses to be carried out in a reasonable time (each time we add a single planning unit the solution space doubles). For the Marxan analyses, we first calculated the proportion of habitat for each species in each planning unit, and summed the un-aggregated management costs in aggregated planning units.

Species in the region were identified for NRM and state government management. Most of the species are listed under State (Nature Conservation Act 1992) or Commonwealth (Environment Protection and Biodiversity Conservation Act 1999) legislation, or are species of regional concern. Maps of potential habitat for 72 priority species were derived from predictive distributions (Chapter 2 Appendix 2A) modelled with Maxent software (Phillips et al. 2006, Dudík et al. 2010), Species of

142

National Environmental Significance range maps (Department of the Environment 2008a) or Atlas of Living Australia (Atlas of Living Australia (ALA) 2012) buffered point locations.

Listed threats (International Union for Conservation of Nature (IUCN) 2014) to each species were identified by experts (Department of Environment and Resource Management 2010). We addressed the management of three key threats to the species: frequent and intense fire, an invasive predator (red fox, Vulpes vulpes), and habitat degradation caused by domestic stock (Chapter 4 Table 4.1 and Table 4A. 1). We considered three threat abatement actions; proactive fire management, predator control, and grazing reduction or removal. We estimated fire management costs (Chapter 4 Appendix 4C) on the basis of variation in vegetation communities (Queensland Parks and Wildlife Service Fire Management System (QPWS) 2012), proximity to human habitation (Australian Bureau of Statistics (ABS) 2006), and prescribed burn and wildfire suppression costs in rural, semi- rural, and peri-urban settings. We estimated invasive red fox control costs (Chapter 3 Appendix 3C) on the basis of a roadside-baiting strategy (Carter et al. 2011) and four 14-day baiting campaigns per year. We estimated the cost of a stewardship agreement to reduce or remove grazing (Chapter 3 Appendix 3B) on the basis of the opportunity cost of foregone agricultural profit (Marinoni et al. 2012). We discounted the three threat management costs across the study region over a 20-year period at the rate of 2% to favour early actions (Johnson and Hope 2012).

Our first prioritisation scenario does not incorporate any dependencies in the benefits of management actions and is referred to as the CEA-optimistic scenario. We prioritised management unit/action combinations in species-rich, rare habitats using a greedy-heuristic based on rank-order of the ratio between the expected benefits and costs of threat management under budget constraints (methods are detailed in Chapter 4 in section 4.7.2 Appendix 4B, following methods for independent benefit management scenario, termed ‘CEA-optimistic scenario’ in present Chapter 5). Cost-effective management strategies (a “project”) allocated to planning units could be a single threat-abatement action (fire, fox, or grazing management), combinations of actions, or no action. We maximised the benefit B given what proportion of each species’ habitat was being managed at a particular planning unit i and which threat-mitigating action(s) q were carried out at cost C, so the conservation problem is:

푀 푁 7 max ∑푖=1 ∑푗=1 ∑푞=0 푥푖푞 퐵푖푗푞 (푋), (5-1) where j is the index of species, subject to meeting our budget constraint D

푀 7 ∑푖=1 ∑푞=0 푥푖푞퐶푖푞 (푋) ≤ 퐷, (5-2)

143

with one management project q per site

7 ∑푞=0 푥푖푞 = 1 ∀ 푖. (5-3)

where xiq is a control variable equal to 1 if management project q is undertaken at site i, and 0 if not; X is the matrix of management projects that can be undertaken at site i.

In our second and third prioritisation scenarios (referred to as MZA-optimistic and MZA- pessimistic), we found near-optimal allocations of resources for managing threats to species to meet conservation targets using a simulated annealing algorithm (Moilanen and Ball 2009). We used Marxan with Zones (Watts et al. 2009), an extension of the original Marxan software (Ball et al. 2009), which has enabled users to take different packages of actions in different sites, and account for their costs, in a systematic planning framework. In the second prioritisation scenario (MZA- optimistic), we optimistically assumed that any independent management action would provide a benefit that was proportional to the number of threats to which a species is vulnerable. In the third scenario (MZA-pessimistic), we pessimistically assumed that species benefitted fully only when all their threats were managed in a combination of actions, i.e. the success of management actions are dependent upon each other. In both scenarios our goal was minimising the cost of comprehensively managing for all species through complementarity, a dependency between the benefits of efficiently managing different sites for all species. To accommodate complementarity, we set equivalent management target constraints to manage the habitat of all species equally, i.e. as a percentage of each species’ mapped habitat. To simplify the analyses, we did not choose the option to cluster management units although this is commonly done in spatial planning; this would have introduced a third dependency--a spatial dependency between management sites. Because the Marxan algorithm finds near-optimal prioritisation scenarios (Moilanen and Ball 2009), it is common to produce multiple solutions. We refer to the solution with the lowest objective function value as being the most efficient scenario. Management sites frequently selected across these multiple good solutions have a higher likelihood of being necessary sites to achieve complementarity; we refer to them as high selection frequency sites.

To contrast priority sites selected when complementarity is not included with those selected when it is, we first compared the CEA-optimistic prioritisation scenario to the MZA-optimistic scenario. We determined the implementation cost of the most efficient MZA-optimistic scenario and set this cost to be the budget constraint for the CEA-optimistic scenario. To contrast priority sites selected when a second dependency between the benefits of management actions was included in the analysis (in addition to complementarity), we next compared the MZA-optimistic scenario to the

144

MZA-pessimistic scenario. In both comparisons of the effects of dependencies on prioritisation of management actions, we evaluated how similar the priority sites and actions were between the two scenarios when taking chance agreement into account (Cohen's kappa measure; Cohen 1960). We also tested for partial agreement in actions with a weighted kappa measure (Cohen 1968) using a fractional weight matrix, where partial agreement in actions were weighted proportionally, as well as a binary weight matrix, where partial agreement in actions were equivalent in weight to total agreement (Appendix 5A; Table 5A.1 and Table 5A.2). We evaluated similarity in the locations of management actions across the entire study region and in a subset of only the management region with high (≥ 90%) selection frequency in 100 scenarios.

To evaluate the differences in management actions chosen when complementarity is and is not included, we also assessed area of habitat managed for each species in the CEA-optimistic and MZA-optimistic scenarios. We classified species on the basis of the proportion of the study region that was mapped as potential habitat for the species. We measured certainty in prioritised management actions (see Levin et al. 2013) for both the MZA-optimistic and MZA-pessimistic scenarios with an approach commonly used in remote sensing (Eastman 2009).

5.4 Results

5.4.1 Different sites are prioritised when including dependencies between sites and species managed (complementarity)

Different site and action combinations were prioritised in the scenario derived using cost- effectiveness analysis as compared to the least-cost scenario where all species are managed because sites were selected using complementarity (Figure 5.1). In both scenarios, we optimistically assumed that species would benefit from management of any one threat. The Kappa statistic indicates how similar scenarios are to each other, and would be a value of one if the scenarios were perfectly similar, and would be a value of zero when similarity is the same as would be expected by chance (Monserud and Leemans 1992). Regardless of species management target value, the spatial distribution of management action strategies and locations were dissimilar (Kunweighted and Kweighted <= 0.012 in all comparisons; (Appendix 5A; Figure 5A. 1 and Table 5A.3). With a species management target of 30% for every species, the scenarios differed in priority actions for the majority of management units (57%), with agreement in taking no action in 31% of management units and sharing at least one of the three management actions in common in the remaining 12% of locations (total agreement for just over 1%) (Figure 5.1c). A greater extent of management area was necessary for meeting targets when complementarity is accommodated and all species are managed, as compared to when sites are prioritised only by cost-effectiveness (Table 5.1). 145

a) Marxan with Zones b) Cost-effectiveness c) Similarity map scenario analysis scenario

Figure 5.1 Different locations for action are prioritised when we optimistically expect species to benefit proportionally from an action that abates one threat of many as well as a) manage for all species through complementarity (MZA-optimistic); or b) manage for cost-effective, species-rich, rare habitats (CEA-optimistic); as shown in c) the similarity of management actions prioritised in the two scenarios. Management area is targeted to be 30% of each species’ habitat.

Table 5.1 When optimistically expecting species to benefit from management of any threat, a greater area is needed for managing threats to all species at least cost (MZA-optimistic) as compared to cost-effectively managing threats in species-rich, rare habitats (CEA-optimistic).

MZA- optimistic cost and MZA- CEA- MZA- CEA- optimistic optimistic optimistic MZA- optimistic CEA- target budget scenario optimistic scenario optimistic constraint constraint area area cost area area cost (%) ($ million) (km2) ($/km2) (km2) ($/km2) 10 2.20 9,430 233 1,113 1,977 20 7.28 13,318 547 3,711 1,962 30 15.07 19,662 766 7,103 2,122 40 26.25 25,757 1,019 11,237 2,336 50 48.81 29,746 1,641 17,812 2,740 146

When species habitat targets were ≤30%, less than 25% of the CEA scenario priority management areas overlapped areas of high selection frequency (≥ 90%) in the MZA-optimistic scenario (Figure 5A. 3). Moreover, agreement of management actions was uncommon in high selection frequency areas (Figure 5A. 4), and the spatial distribution of optimal management actions and locations were dissimilar in the high selection-frequency (≥ 90%) areas (Figure 5A. 4; Kappa and weighted Kappa ≤ 0.013 in all comparisons). Species’ targets were met more efficiently (the area prioritised for managing a species was nearly equivalent to the target set, rather than prioritising a greater area for management than the target amount) in the MZA-optimistic scenario (when complementarity was accommodated) than in the CEA scenario (Figure 5A. 5; Table 5A.5a). However, in the CEA scenario (that did not include complementarity) comparable targets were met for 95% of species with rare habitats (habitat ≤ 1% of study area) at a 30% target constraint (Table 5A.5b).

5.4.2 Similar sites are prioritised when including additional dependencies between management actions

The pessimistic scenario (where a species needs all its threats abated to receive any benefit) was twice the price ($30 versus $15 Million) of the optimistic scenario (where species benefitted proportionally from managing any threat) although a similar-sized area of management is necessary (Table 5A.6). The spatial distribution of actions in the pessimistic and optimistic scenarios were more similar than dissimilar, with agreement in 55% of the planning units (action in 24%, non- action in 31%), and partial action agreement in 20% of the planning units (Figure 5.2). The spatial distribution of planning units with high selection frequency was similar between the two scenarios, especially when weighted for partial agreement in actions (Kweighted = 0.97; Table 5A.7). However, multiple-action zones were selected for 26% more of the region in the pessimistic scenario, where all threats must be abated for species to benefit, as compared to the optimistic scenario (Table 5A.8b; 58.82% versus 32.33%). Certainty in zone classification was similar for the two scenarios (Figure 5.3). Approximately one-third of the high-selection frequency management units were selected with high certainty for the grazing management zone in the optimistic scenario, and for the grazing and fire management zone in the pessimistic scenario, and roughly half were selected for the ‘no management’ zone in both scenarios (Table 5A.9).

147

(a) Marxan with Zones (b) Marxan with Zones (c) Similarity map optimistic scenario pessimistic scenario

Figure 5.2 When a dependency between actions is considered in addition to managing for all species through complementarity, similar locations for action are prioritised in a) a scenario where we optimistically expect species to benefit proportionally from an action that abates one threat of many (MZA-optimistic); and b) a scenario where we pessimistically expect that species benefit only if all their threats are managed (MZA-pessimistic); as shown in c) the similarity of management actions prioritised in the two scenarios. Management area is targeted to be 30% of each species’ habitat.

148

(a) Uncertainty map for ( b) Uncertainty map for Marxan with Zones Marxan with Zones optimistic scenario pessimistic scenario

Figure 5.3 Uncertainty maps provide an indication for each planning unit as to whether the same package of management actions was frequently selected for the planning unit (Levin et al. 2013) whether a) optimistically expecting species to benefit proportionally from any action that abates one threat of many (MZA-optimistic); or b) pessimistically expecting that species benefit only if all their threats are managed (MZA-pessimistic). Areas selected most frequently for no management action are not shaded. Management area is targeted to be 30% of each species’ habitat.

5.5 Discussion

This research contrasts and compares two approaches to decision support for spatially prioritising conservation actions for multiple species and threats. We found large differences between the locations and actions prioritised for managing threats when chosen by cost-effectiveness, as compared to when chosen by accommodating complementarity, which is a dependency in the benefits of managing species in different sites (Figure 5.1). Prioritising management by cost- effectiveness could lead to inefficient spending of conservation funds if the goal is to maximise the number of species managed, largely because some species for which actions are cheap are being “over-conserved”. When complementarity is not explicitly considered, management may not be prioritised for all species (Table 5A.5). Alternatively requiring that all species be managed with complementarity could lead to inefficient spending if the goal is to maximise the management of rare habitats. When sites and actions were prioritised by cost-effectiveness, 30% targets were not met for nearly a third (28%) of the species, but 95% of species with rare habitats (<=1% of the region) were protected (Table 5A.5) in a little over one-third (36%) of the management area (Table 5.1). 149

We also found large differences in the cost of managing species when comparing a scenario where we optimistically expect species to benefit from any threat management as compared to a scenario where we pessimistically expect that the success of threat management actions are dependent on each other and that species only benefit when all their threats are managed (Table 5A.6). Requiring that all threats to all species be managed was twice the price of assuming species will benefit from individual threat management, although threats were managed in similar locations (Figure 5.2) with high certainty (Figure 5.3).

The dissimilarity between two scenarios that do and do not accommodate complementarity (Figure 5.1 and Table 5A.1; Table 5A.3 and Table 5A.4) shows how important it is to be specific in choosing a management goal that describes desired outcomes (Possingham et al. 2001b). Both the simpler and more complex methods provide viable spatial decision support to cost-effectively implement management action, but result in different action prioritisations because of differences in goals that target either species richness and habitat rarity (CEA) or species diversity (MZA that accommodates complementarity). For the same overall cost, the approach that incorporates complementarity prioritises least-cost management sites meeting targets for abating threats to all species. In contrast, simple CEA (with no complementarity) prioritises multiple actions in fewer higher-cost sites that are offset by threatened species richness and habitat rarity. Thus, including complementarity results in needing a larger management landscape to meet all species’ needs, compared with managing fewer species needs more intensively in a smaller management area (Table 5.1 and Table 5A.5). Our results concur with other research that shows that prioritisation approaches that incorporate complementarity have greater conservation efficiency in terms of species protected per dollar invested, compared with ignoring complementarity (Underwood et al. 2008, Withey et al. 2012). An approach that includes complementarity minimises the risk of under- representing species in prioritisation planning by necessitating a larger management area with fewer management actions. However, this approach assumes that full species benefits will accrue from a limited selection of threat management actions.

In contrast, the scenarios for optimistically managing threats independently and pessimistically managing all threats (by incorporating a second dependency to complementarity) are similar in management area needed and locations managed, but actions prioritised at the similar sites are different (Figure 5.2). The spatial similarity between the scenarios indicates that whether we are optimistic or pessimistic about the ability of a single action to assist species affected by multiple threats, management location selections are comparable. However, a pessimistic strategy addresses more of the threats in similar locations. Because we also found high certainty in specific actions to undertake in locations selected with high frequency, it is then a question of choosing to manage 150

more or less intensively (Figure 5.3 and Table 5A.9). Expected benefits may be overestimated when investing in abating only one threat to species if dependencies in management actions exist (Evans et al. 2011b; Chapter 4). Alternatively, when managing one threat or managing threats independently benefits species, spending an entire budget on managing multiple threats may mean that some sites are managed too intensively and threats may be more effectively managed in alternate sites (Chapter 4).

Accounting for dependencies comes at a price. If complementarity is considered in prioritising management, the cost comes in terms of the area of land needed to be managed. With a goal of abating threats in 30% of all species’ habitats, roughly two-thirds more of the region would need to be managed when choosing sites based on complementarity, in contrast to identifying species-rich, rare habitats with basic CEA (19,662 versus 7,103 km2; Table 5.1). Accounting for spatial heterogeneity in cost increases efficiency (Ando et al. 1998), and the cost per km2 managed was one-third the cost when species complementarity is included as compared to the basic CEA where it is not included ($766 versus $2,122/km2 for 30% targets; Table 5.1). Comprehensive threat management comes at a high price; it was approximately twice the price to manage all threats as compared to optimistically expecting species to benefit from management of any threat ($30.02 versus $15.07 Million; Table 5A.6).

CEA is a transparent, repeatable, and easy-to-use method for spatially prioritising efficient threat- abatement management (Wilson et al. 2007, Evans et al. 2011b, Auerbach et al. 2014), but incorporating complementarity is necessary to explicitly choose management areas for each species. As multiple threatening processes usually interact (Sala et al. 2000, Brook et al. 2008) disregarding dependencies in threat management actions will likely result in less than optimal management. The optimistic assumption that single management actions alone will accrue full species benefits may fail to produce expected conservation outcomes (Chapter 4). In a best case scenario, greater investment to effectively reduce all threats to all species may be optimal. However, our research informs practitioners about the relative trade-offs in costs and benefits of different strategies. Efficiency can be expressed in terms of species, area, or threats managed per dollar spent. By accounting for complementarity and a dependency in threat management actions, the trade-offs in efficiency become clearer because the costs of threatened species management are explicitly and transparently considered.

151

5.6 Supplementary Information

5.6.1 Appendix 5A Supplementary tables and figures

(a) Marxan with (b) Cost- (c) Similarity maps Zones solutions effectiveness Analysis solution

Figure 5A. 1 Comparisons of (a) most efficient solutions from Marxan with Zone analyses (MZA) constrained by 10-50% targets for each species, and (b) from the Cost-effectiveness analysis (CEA) constrained by budgets equivalent to the cost of implementing each MZA solution, and (c) similarity between the two. 152

Figure 5A. 2 The total cost for implementing the most efficient Marxan with Zones scenario versus target values of 10-50%. Targets are specified for percentage of species habitat managed across the natural resource management region.

153

Figure 5A. 3 Percentage of area from most efficient MZA-optimistic and CEA-optimistic (by target/budget constraint) scenarios versus MZA-optimistic selection frequency. High selection frequency is an indication of management units that have a higher likelihood of being necessary areas for a good scenario (aka irreplaceability). Y-axes for 10-30% targets are standardised to 50%; Y-axes for 40-50% targets are standardised to 100%.

154

10%

20%

30%

40%

50%

Figure 5A. 4 The percentage of area in the CEA-optimistic scenario that is selected for any management strategy that includes fire, fox, or grazing action is plotted against the selection frequency of management units in the MZA-optimistic scenario for any zone that includes the same action. Y-axes are not standardised.

155

Figure 5A. 5 Frequency (number of species) versus level of representation in MZA-optimistic and CEA-optimistic scenarios.

156

Table 5A. 1 Fractional weight matrix used to count disagreements differentially when calculating a weighted kappa statistic (Cohen 1968). Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and are set to zeroes. Off-diagonal cells contain weights indicating differing degrees of disagreement, where a value of one indicates total disagreement and values between zero and one indicate partial agreement.

Fire, Management Fire & Fox & Grz & Fox & Strategy Fire Fox Grz Fox Grz Fire Grz None Index 1 2 3 4 5 6 7 8 Fire 1 0 1.00 1.00 0.50 1.00 0.5 0.67 1.00 Fox 2 1.00 0 1.00 0.50 0.50 1.00 0.67 1.00 Grz 3 1.00 1.00 0 1.00 0.50 0.50 0.67 1.00 Fire & Fox 4 0.50 0.50 1.00 0 0.05 0.50 0.34 1.00 Fox & Grz 5 1.00 0.50 0.50 0.50 0 0.05 0.34 1.00 Grz & Fire 6 0.50 1.00 0.50 0.50 0.5 0 0.34 1.00 Fire, Fox & Grz 7 0.67 0.67 0.67 0.34 0.34 0.34 0 1.00 None 8 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0

Table 5A. 2 Binary weight matrix used to count disagreements differentially when calculating a weighted kappa statistic (Cohen 1968). Weight matrix cells located on the diagonal (upper-left to bottom-right) represent agreement and are set to zeroes. Off-diagonal cells contain weights indicating differing degrees of disagreement, where a value of one indicates total disagreement. For a relaxed assessment of partial agreement, values of even partial agreement in actions were also set to zero.

Fire, Management Fire & Fox & Grz & Fox & strategy Fire Fox Grz Fox Grz Fire Grz None index 1 2 3 4 5 6 7 8 Fire 1 0 1 1 0 1 0 0 1 Fox 2 1 0 1 0 0 1 0 1 Grz 3 1 1 0 1 0 0 0 1 Fire & Fox 4 0 0 1 0 0 0 0 1 Fox & Grz 5 1 0 0 0 0 0 0 1 Grz & Fire 6 0 1 0 0 0 0 0 1 Fire, Fox & Grz 7 0 0 0 0 0 0 0 1 None 8 1 1 1 1 1 1 1 0

157

Table 5A. 3 To evaluate scenario similarity while taking chance agreement into account, Cohen’s un-weighted and weighted kappa statistic (Cohen 1960, 1968) compare the most efficient Marxan with Zones analysis (MZA-optimistic) scenarios to the Cost-effectiveness analysis (CEA- optimistic) scenario over the entire study region for five target constraints (n=number of management units).

Kappa Kappa Kappa Target (%) n (unweighted) (weighted)1 (weighted)2 10 129894 0.005 0.008 0.011 20 129894 0.004 0.007 0.011 30 129894 0.005 0.008 0.012 40 129894 0.003 0.005 0.008 50 129894 0.005 0.005 0.006

1 fractional weight matrix (Table 5A.1) 2 binary weight matrix (Table 5A.2)

Table 5A. 4 To evaluate scenario similarity while taking chance agreement into account, Cohen’s un-weighted and weighted kappa statistic compare the most efficient Marxan with Zones analysis (MZA-optimistic) scenarios to the Cost-effectiveness analysis (CEA-optimistic) scenario over the most frequently selected management units (selection frequency ≥ 90%) for five target constraints

(n=number of management units).

Kappa Kappa Kappa Target (%) n (unweighted) (weighted)1 (weighted)2 10 68374 -0.001 0.001 0.003 20 53404 0.000 0.003 0.006 30 54025 0.003 0.008 0.013 40 79846 0.002 0.005 0.009 50 77624 0.005 0.007 0.009

1 fractional weight matrix (Table 5A.1) 2 binary weight matrix (Table 5A.2)

158

Table 5A. 5 Percentage of target constraints met in a) Marxan with Zones analyses (MZA- optimistic) as compared to b) Cost-effectiveness analysis (CEA-optimistic) for all species, and for species separated into habitat area-classes representing the percentage species’ habitat area relative to the entire study area.

(a) Targets met (%) Constraint Habitat class2 Target (%) Overall1 0 - 0.1 0.1 - 1 1 - 10 10 - 100 10 99 94 100 100 100 20 97 94 100 95 100 30 99 94 100 100 100 40 96 88 100 95 100 50 92 88 95 90 93

No. of Species 72 16 21 20 15

1unmet MZA targets were due to habitat restrictions in potential management areas for all cases but one species in the 50% target analysis 2Mapped habitat area relative to study area (%)

(b) Targets met (%) Constraint Habitat class1 Budget ($Million) Overall 0 - 0.1 0.1 - 1 1 - 10 10 - 100 2.20 50 88 76 30 0 7.28 62 94 90 50 7 15.07 72 94 95 75 13 26.25 79 94 100 80 33 48.81 93 100 100 90 80

No. of Species 72 16 21 20 15

1Mapped habitat area relative to study area (%)

159

Table 5A. 6 A comparison of the total cost of the area selected for management in the most efficient Marxan with Zones analysis (MZA) scenarios under optimistic vs pessimistic threat management scenarios with a species target constraint of 30%. Attempting to meet the 30% target for one species raised the most efficient scenario cost more than $6 Million; in pessimistic scenario ‘a’, 24% of the species’ habitat was selected for management, in ‘b’, 28%.

MZA Target Management Constraint Scenario Targets met MZA scenario cost MZA scenario area (%) (#/72)1 ($ Million) (km2) 30 Optimistic 71 15.07 19,662 30 Pessimistic a 68 30.20 19,478 30 Pessimistic b 68 36.47 19,416

1unmet MZA targets were due to habitat restrictions in potential management areas with the exception of one species in the pessimistic scenarios.

Table 5A. 7 To evaluate scenario similarity while taking chance agreement into account, Cohen’s un-weighted and weighted kappa statistic compare the most efficient optimistic and pessimistic Marxan with Zones analysis (MZA) over the entire study region and the most frequently selected management units (selection frequency ≥ 90%) (n=number of management units).

Kappa Kappa Kappa Comparison n (unweighted) (weighted)1 (weighted)2 Total area of most efficient scenario 44373 0.385 0.457 0.554 High selection frequency areas (>= 90%) 14461 0.574 0.749 0.969

1 fractional weight matrix (Table 5A.1) 2 binary weight matrix (Table 5A.2)

160

Table 5A. 8 Zone management was selected in optimistic and pessimistic 30% target management scenarios. Management area selected for each zone is represented in a) km2 and b) as a percentage of potential management area. Potential management area was determined by the combination of species distribution, known threats to species, and areas costed for threat management. Only areas with remnant vegetation where habitat may be managed for the three threats (fire, foxes and grazing) were considered in the analysis, amounting to ~60% (~32,473 km2) of the entire region (~55,156 km2).

(a)

Zone Zone name Optimistic Pessimistic 1 FireMgmnt 53.75 4.25 2 FoxMgmnt 83.50 45.00 3 GrzMgmnt 9,025.50 328.75 4 FireFoxMgmnt 9,785.00 9,689.50 5 FoxGrzMgmnt 45.25 141.25 6 GrzFireMgmnt 668.75 8,828.50 7 FireFoxGrzMgmnt - 440.75 AnyMgmnt 19,661.75 19,478.00

8 NoMgmnt 12,811.75 12,995.50

PossibleMgmnt 32,473.50 32,473.50 RegionTtl 55,156.25 55,156.25

(b) zone zonename Optimistic Pessimistic 1 FireMgmnt 0.17 0.01 2 FoxMgmnt 0.26 0.14 3 GrzMgmnt 27.79 1.01 4 FireFoxMgmnt 30.13 29.84 5 FoxGrzMgmnt 0.14 0.43 6 GrzFireMgmnt 2.06 27.19 7 FireFoxGrzMgmnt 0.00 1.36 AnyMgmnt 60.55 59.98

8 NoMgmnt 39.45 40.02

PossibleMgmnt 58.88 58.88

161

Table 5A. 9 Proportion of frequently selected management units1 selected with high certainty2 for specified management zones in the optimistic and pessimistic Marxan with Zones scenarios. Species targets in both analyses were 30%.

Management Zone Optimistic scenario Pessimistic scenario Fire 0.00 0.00 Foxes 0.00 0.00 Grazing 0.30 0.00 Fire & Foxes 0.08 0.06 Foxes & Grazing 0.00 0.00 Grazing & fire 0.05 0.33 Fire, Fox, and Grazing 0.00 0.02 None 0.50 0.51 Uncertain 0.07 0.08 Total 1.00 1.00

1management units with >= 90% selection frequency 2zones selected with highest certainty (Levin et al. 2013)

162

Chapter 6

Conclusions

6 Conclusions

Evidence of biodiversity decline in Australia and the wider world is unequivocal, with future rates of decline being dependent upon how and where threats expand, and whether those threats are countered (Pimm et al. 2014). A variety of approaches to conservation will be needed if there is to be success in addressing threats to species and maximising biodiversity persistence. However, at the heart of the matter is the need to not only prioritise species for management, identify threats, and create protected areas to offset habitat loss, but also to take direct threat mitigation action (Pressey et al. 2004, Wilson et al. 2006, Pressey et al. 2007, Wilson et al. 2007). Fast and efficient action is needed to abate the threats that endanger many species.

Conservation decision-making is complex, requiring that practitioners make management decisions about securing threatened species with limited information, but with many criteria to consider. Numerous species are imperilled by concurrent threat processes occurring across the landscape, while funding is inadequate for managing all species’ needs. While complex models may be more revealing of underlying patterns and processes than simple models, natural resource management bodies are constrained by time, funding, expertise and limited technical capacity, and often to do not have the luxury to invest in more complex modelling. Little is known about many threatened species’ distribution and their ecological needs, but delaying management while unravelling the complexity may mean that biodiversity continues to decline, so resources may be better spent on more immediate management. The heterogeneous distribution of species’ habitat and the costs of threat management across the landscape suggest that there are more-efficient and less-efficient locations to manage. In addition, there is incomplete understanding of the trade-offs involved with investing in threat management: the implications of the dependencies between the costs and benefits of management actions may not be apparent unless explicitly considered. Given these circumstances, securing biodiversity requires that threats be managed under conditions of uncertainty – with the possibility that management may be misplaced or ineffective, and with constraints limiting what can be done.

In this concluding chapter I begin by outlining some of the many difficulties encountered in choosing a strategic approach for managing the biodiversity crisis at a landscape level. Included in the next sections are descriptions and implications of some of the options for responding to these difficulties, including: prioritising and acquiring protected areas; implementing ad hoc management; or seeking additional data to improve knowledge. Following this discussion, I describe how economically based threat-action prioritisation provides an alternative, and how my research contributes to the conservation literature. In sum, as an alternative, my research 165

recommends that management decisions informed by spatial prioritisation of cost-effective actions to abate threats to multiple species may ameliorate some of the problematic implications of the other approaches. My research advances thinking in several areas of the conservation research literature: 1) by engaging predicted species distribution models in conservation decision-making to provide specificity in locations where actions are needed to abate specific threats; 2) providing an approach for multiple-species and multiple-threat action prioritisation based on rigorous and transparent methods; 3) focusing on accessible decision support; and 4) exposing the implications of disregarding dependencies between the costs and benefits of management actions. I then take a step back and discuss the major assumptions and limitations of my research, as well as future opportunities to augment the research to broaden its scope. I conclude the chapter with my personal perspective on major challenges in the field of conservation research.

6.1 Potential strategic responses for securing biodiversity persistence

Responses to the complex issues surrounding biodiversity conservation include: creating protected areas, managing opportunistically, and delaying action to improve knowledge. Each of these approaches has the potential to improve conservation outcomes, but some implications of these responses are described below.

6.1.1 Protected areas are necessary but not sufficient

Identifying and prioritising potential protected areas (Rodrigues et al. 2004b, Watson et al. 2011 and many others), and then acquiring them, is one response to reducing the rate of biodiversity declines. Protected areas are the cornerstone of biodiversity conservation efforts and address the threat of habitat loss, but an underlying, and often incorrect, assumption is that protected areas are sufficient for biodiversity protection. Protected areas are isolated patches of habitat within a larger landscape matrix that may require management for movement and dispersal between patches, resource availability among patches, and abiotic environment effects on patches from outside (Driscoll et al. 2013). Integrating ecological and biophysical processes such as animal movement and hydrological flows between terrestrial, freshwater, or marine areas requires that corridors, buffer zones and connectivity be incorporated in protected area designs (Beger et al. 2010). However, socio-economic limitations of an approach focused on protected areas include the inability to secure sufficient resources for land purchase; unavailability of biodiverse habitat (i.e. the land is not for sale or available land has limited biodiversity potential); or restrictions on the location of a desired protected area due to political or social considerations (Carpenter et al. 2006, Carpenter et al. 2009). Protected areas are also not immune from threats such as invasive species either acting within their borders or invading from external and unmanageable sources (Brandon et 166

al. 1998, Terborgh et al. 2002, Ervin 2003), and climate change is a pervasive threat that cannot be managed within protected area boundaries (Peters and Darling 1985). Protected areas do little to mitigate threats that occur outside their boundaries, such as on private or other government-held lands in which species would benefit from appropriate threat mitigation. For example, reconciling the need for human food production alongside biodiversity conservation involves land management outside of protected areas to mitigate habitat loss. However, while biodiversity may be able to co- exist with wildlife-friendly, low-yield farming (land sharing), accommodating biodiversity on high- yield agriculture properties may still require that particular areas be set aside for no production (land sparing) (Green et al. 2005, Phalan et al. 2011). Protected areas are necessary, but alone they are insufficient to abate all threats to biodiversity.

6.1.2 Ad hoc management may be inadequate

Ad hoc management involves performing management actions based on managerial or personal knowledge, historical managerial involvement or past precedent, or may be impromptu or opportunistic management that depends on circumstances that arise. For example, opportunistic management may consist of using easily obtained broad-scale maps to represent species distribution or assuming the benefits of management actions are independent from each other. Ad hoc management may be effective in part, but may not be efficient in achieving the maximum biodiversity benefit from investment. Past acquisition of protected areas has often been ad hoc (Pressey et al. 1993, Pressey 1994), and the global network of protected areas does not effectively or adequately protect species diversity (Rodrigues et al. 2004a, Rodrigues et al. 2004b, Watson et al. 2011). Another tendency of an ad hoc approach is to concentrate management on well-studied, highly charismatic species (Sitas et al. 2009). Whereas focusing attention on flagship species may be a strategic approach to raising public awareness and conservation funding (Leader-Williams and Dublin 2000), it is telling that regardless of this attention, even these species are experiencing population declines, diminishing ranges, and local extinctions due to unmanaged threats (Bennett 2011). Over-spending limited funds on the most-charismatic or most-threatened species where recovery efforts are large and chances of success are small has the detrimental side-effect of relegating less well-known species to continue to decline (Possingham et al. 2002). Management that is not systematically evaluated can lack transparency and rigour, making it difficult to defend to stakeholders that require accountability. Neither ad hoc management, nor managing single species alone, will be adequate for managing the needs of the number of species identified as at risk and in need of threat mitigation.

167

6.1.3 Time spent improving knowledge may be counter-productive

Often decision makers can be hesitant to act in the absence of full understanding, and improving knowledge via more data collection to guide management is a technique commonly employed. However, there are difficulties in acquiring existing data on endangered species: observations of rare species are difficult to obtain, the data are sensitive, and in some cases researchers may be unwilling to share data (as discussed below). More extensive data would enable more informed decisions to conserve species, but would require more time and hence delay actions. Help in choosing between alternative management actions prior to making a decision could be aided by value-of-information analysis, which evaluates the amount a decision maker would be willing to pay for information to reduce uncertainty (e.g. Raiffa 1993, Polasky and Solow 2001). However, time spent improving knowledge would delay management action for the sake of increased certainty, which may be counterproductive due to the opportunity costs involved (Grantham et al. 2009). There is a need for rigorous approaches that provide effective and timely decision support despite data limitations. My research considers the implications stemming from taking the common approaches, and contributes an accessible decision-support alternative that can guide conservation choices.

6.2 Economically based threat-action prioritisation as an alternative

An alternative to the three potential responses previously described (procuring protected areas, managing opportunistically, or delaying to improve knowledge) is the thrust of this thesis: conservation action is informed by placing the problem in a decision-science framework (Possingham et al. 2001b) and specifying an objective of maximising the benefits to multiple species across the landscape through cost-effective threat management action. Threats are ubiquitous processes, requiring management both inside and outside of protected areas. Management actions addressing threatening processes serve the needs of many species, rather than focusing on single species. Extending systematic conservation planning to prioritise cost-effective threat action addresses multiple threats to multiple species in a rigorous, repeatable, transparent and defensible way. My research contributes an alternative approach for securing biodiversity persistence by providing a framework that prioritises multiple actions for multiple species, and responds specifically to four needed areas of research in the conservation literature.

6.2.1 Research contributions

My research makes economically based prioritisation thinking more accessible for managerial decision-making that will lead to threat mitigation and ultimately to reduction in biodiversity loss.

168

Threats to biodiversity occur at a global scale, but delivery of conservation management actions to maximise persistence of biological diversity must happen at the landscape level to be effective in conserving local populations. Specifically, this thesis focuses on four aspects of conservation planning where research is needed to further understanding of prioritising management actions.

6.2.1.1 Indirect threat maps derived from the predicted distribution models of threatened species identify threat abatement locations that aids decision support

First, action is required to manage threats to benefit species. Threats are processes that are spatially distributed across the landscape, so threat mitigation has the potential to benefit multiple species. It is insufficient to identify threatened species and potential protected areas. Threats must be abated for vulnerable species to recover and persist in the long term. While predicted species distribution models are proposed to support conservation decision making, there is scarce evidence in the literature of their use in supporting solutions for on-the-ground conservation problems (Guisan et al. 2013). When direct threat maps are unavailable, indirect threat maps derived from the overlay of landscape-scale predicted distribution models of species affected by specific threats locates where those threats can be managed. With careful consideration of uncertainty in data and models, mapped data can contribute meaningfully to conservation decision-making (Guisan et al. 2013). The research in Chapters 2–5 is based on the assumption that threat abatement will benefit species, and that locations for threat management are identified using indirect threat maps, which in turn provides a basis for management decision support.

6.2.1.2 Spatially explicit prioritisation for management actions that cost-effectively abate multiple threats to many species

Second, my research illustrates spatially locating cost-effective actions to benefit species of concern, based upon the spatial co-occurrence of species distributions, their threats, and the costs of actions to abate those threats. The most common conservation prioritisation problems addressed in the literature pertain to the spatial allocation of a single conservation action, even though generally such real-life problems are rare (Moilanen et al. 2009c). Optimal land-use planning addresses socio- economic aspects of biodiversity conservation outside of protected areas (Polasky et al. 2005, Holzkämper and Seppelt 2007, Polasky et al. 2008, Wilson et al. 2010), but does not explicitly address threat management. There is a lack of specificity in directing conservation actions, and a need for greater integration of socio-economic and ecological disciplines and conservation beyond protected areas and into human-occupied landscapes (Heller and Zavaleta 2009). Conservation managers need decision support for identifying what conservation actions to do where, and which is the most cost-effective combination, although this is a complex spatial problem when there are 169

multiple species and multiple threats (Joseph 2007, Carwardine 2009, Moilanen et al. 2009b). Despite the complexity, in Chapters 3–5, my research demonstrates how to prioritise management actions that cost-effectively abate multiple threats to many species.

6.2.1.3 Incorporating relationships between costs and benefits of management actions clarifies trade-offs

There is a need to recognise the implications of the interactive aspects of the costs and benefits of threat management, to optimise outcomes to the benefit of multiple species. This requires that we examine the implications of dependencies between actions, an examination which is rarely done. Although interactions and negative and positive feedbacks between threat-abating action are challenging to model (Pressey et al. 2007), they can make a difference in whether management investments can be expected to result in positive conservation outcomes, and they are rarely addressed in conservation cost effectiveness analyses (Duke et al. 2013). In Chapters 4–5, my research demonstrates that, given the limited resources dedicated to biodiversity persistence and recovery, decisions about where to take action can be guided by an exploration of the trade-offs in the costs and benefits of threatened species management.

6.2.1.4 Accessible decision support

Barriers to implementing prioritisation models and computationally intensive techniques may prevent regional managers from implementing research findings and from taking timely action. There is a research-implementation gap between conservation planning and its real-world application (Knight et al. 2008). Prioritisation and systematic planning approaches that are accessible, that can be based on existing (but frequently sparse) data, and that may inform threat management decisions and immediate actions may reduce barriers to implementation. Because technical and computational expertise may be limited, decision support that is intelligible, intuitive, and reasonable is needed for transparent, repeatable, and defensible decisions. There is a need for models that can be implemented quickly despite sparse data, and to enable decision support for practitioners working with restricted capacity. In Chapter 2, I discuss deriving indirect threat maps, and in Chapter 3, I demonstrate basic cost-effectiveness analysis – both of which are accessible methods for practitioners to use for decision support.

My research provides a rational approach to prioritising threat management with commonly-used technology and publicly available data at a landscape level, where conservation action is needed to support local populations of species. The approach is transparent and defensible to stakeholders, and encourages thinking about the implications of often overlooked dependencies in threat

170

management action. The difficulties that arise with making management decisions under constraints need urgent attention, so that actionable choices can be made to support biodiversity conservation.

6.2.2 Theory and methods supporting conservation decision research

My research has utilised a variety of tools and paradigms: decision science, fine-scale species distribution models, cost-effectiveness analysis, and systematic conservation planning. A decision- science framework is a structured protocol for problem solving (Shea et al. 1998), guiding conservation practitioners attempting to achieve explicitly stated management objectives (Possingham et al. 2001b). Species distribution predictions are central to many conservation applications (Ferrier 2002). Providing the opportunity for conservation practitioners to map species distributions, Maxent (Phillips et al. 2006) is publicly available software for species distribution modelling that performs well using existing data from museums and herbaria (Elith et al. 2006). Cost-effectiveness analysis, also referred to as Return-on-Investment thinking (Murdoch et al. 2007), is a common sense approach for deciding which conservation measures to take with a limited budget, because it is decision-making approach that people use every day. Whenever anyone decides which product to buy with their limited money, or in which activity to participate with their limited time, they are engaging in a cost-effectiveness analysis by combining benefits and costs. Publicly available systematic conservation planning software programs designed for decision support, such as Marxan (Ball et al. 2009), Zonation (Moilanen 2007) and other similar programs reviewed in Moilanen et al. (2009d), are becoming more commonly used for spatially oriented decision support in managing biodiversity. Although applications of these software programs to prioritise multiple management actions at the fine-scale are rare (e.g. Moilanen et al. 2011b), my research shows how fine-scale action prioritisation can be achieved in a relatively transparent fashion. An advantage to using systematic conservation planning software is that complementarity is implicitly incorporated. Other software programs, such as INFFER (Pannell et al. 2010) and RobOff (Pouzols and Moilanen 2013), are emerging non-spatial decision-support tools also focusing on what actions to take, although outputs are not spatially localised; the RobOff software has (to-date) omitted an explicit spatial structure to reduce data demands (Pouzols et al. 2012). These systematic conservation planning methods are becoming more accessible to conservation researchers and practitioners with guidance as to their use provided in workshops and tutorials. The research presented in this thesis reveals how threat-management prioritisation is optimised by combining decision science, fine-scale species distribution models, basic cost-effectiveness analysis, and systematic conservation planning, although acknowledging research assumptions helps identify its limitations.

171

6.3 Assumptions and opportunities to augment research

The research in this thesis reveals additional areas where systematic conservation planning approaches can be improved and also made more accessible. Research in economically grounded, fine-scale, spatial conservation planning is still data-intensive, with requirements for detailed, spatially referenced, species distribution, threat, and management cost data. A major assumption of my research is that the species distribution, threat, and management cost models that I derived are sufficiently accurate to produce reliable analyses.

Because spatial conservation planning is underpinned by predictive models, certain assumptions are required: that in minimising omission and commission errors, species distribution models are sufficiently predictive of threatened species occupancy (Loiselle et al. 2003, Wilson et al. 2005c); that threats can be spatially targeted using indirect threat maps; and that the costs of managing the threats to those species can be estimated. However modelling always contains elements of uncertainty and error so that: “Essentially all models are wrong, but some are useful” (Box and Draper, p. 424). While this research acknowledges limitations, the alternative of doing nothing due to poor information will not solve pressing conservation problems. Detailed knowledge about species distributions and ecology is poor, leading to calls for gathering more data (Stuart et al. 2010). However, although more information would likely be useful, delays in taking action can be counter-productive when threats are ongoing (Grantham et al. 2008). Nevertheless, ground-truthing of management sites would be essential before taking action, and the results of this thesis should be evaluated in this context and should not be taken as prescriptive. The primary purpose is to explore and contrast approaches.

In the following section, I will further discuss limitations of my data, models, and analyses, major assumptions, and opportunities to augment my research in the future. First, I will discuss data matters, as data are the foundation of any analysis. Location data for threatened species can be rare and sparse. While I compiled data for this thesis from a variety of heterogeneous sources, I primarily used publicly available data, similar to a natural resource manager with limited resources. I next discuss the predicted species distribution models derived from the threatened species data. Issues can be compounded in spatial analyses because they build on base data. Modelling with sparse data is challenging, but necessary. Next I discuss some of the future opportunities for further research, including more comprehensively accounting for the socio-economic costs of conservation; integrating more ecological considerations into the analyses; including likelihood of management success into analyses; and conducting further sensitivity analyses to test the robustness of the conclusions. I also include discussions on uncertainty and adaptive management. These research

172

avenues would introduce more complexity to the analyses, as well as broaden the scope of this research.

6.3.1 Data availability

Locating the data upon which to practically base models is difficult, and in some cases the data do not exist. I describe some of the data access issues I encountered in assembling data for this thesis. For many threatened species, accurately recorded presence/absence locations are sparse, and existing records are sensitive, out of genuine concern for species under threat. Ecological data on species of interest can be lacking. Open-access data, such as historical records from museum collections, often have poor spatial accuracy or the locations may be generalised to protect species. Both data collected for specific purposes or opportunistically acquired may be spatially biased. Data repositories are distributed and use different formats, and coordination between repositories is rare. As well, data governance can be restrictive, and access to proprietary data or data considered sensitive is strictly controlled. As an example, user-specific licensing is required for government databases such as the Species of National Environmental Significance range maps (Department of the Environment 2008a) and sensitive WILDNET species location data (Department of Environment and Resource Management (DERM) 2010), and so online data sources such as the Atlas for Living Australia (Atlas of Living Australia (ALA) 2012) are likely to be highly valuable resources for biodiversity conservation research. Additional data issues arise for spatially modelling the costs for abating threats, and I further discuss incorporating socio-economic costs and ecological considerations to strengthen models below. Management costs are often not documented and there may be sensitivity in revealing financial information. Most of the data access issues encountered in compiling data for this thesis would also be problematic for conservation professionals. Therefore there is a great need for more examples of creative modelling for deriving spatially meaningful information from sparse data.

6.3.2 Predicted species distribution models

Predicted species distribution models are inferred from the environmental conditions of known species occurrences (Guisan and Zimmermann 2000). However, species presence records for endangered species are sparse by nature, and records can be difficult to access, as previously described. The use of sparse datasets to model predicted species distributions is based on certain assumptions. Species presence data records are assumed to be accurately recorded, habitat can be described by modelled environmental data, and that unsurveyed habitat will have the same occupancy patterns as surveyed habitat. An alternative option would have been to base the research on a smaller suite of more data-rich species, such as keystone, umbrella, flagship, focal or landscape 173

species (Mace et al. 2007). However, my goal was to research systematic conservation planning for threatened species at a scale pertinent to natural resource managers. Therefore, I chose to base the analysis on the actual threatened species in the Burnett-Mary NRM, despite data limitations.

Predicted species distribution modelling is becoming more accessible to researchers and practitioners, and when used in a decision-making context, species distribution models can effectively guide conservation actions (Guisan et al. 2013). In particular, Maxent species distribution modelling software (Phillips et al. 2006) has rigorously derived default settings (Phillips and Dudík 2008) and is available as freely distributed software with an online-tutorial (http://www.cs.princeton.edu/~schapire/maxent/) and discussion group (http://groups.google.com/group/Maxent). Maxent has been found to be suitable for modelling species distributions from few presence records (Hernandez et al. 2006, Pearson et al. 2007), although sparse species data may identify only the strongest environmental gradients (Barry and Elith 2006), and highly conservative use of predictions and restriction of use to exploratory modelling is recommended in the case of small sample sizes (Wisz et al. 2008, Elith and Leathwick 2009). My goal was to perform the analyses with the data and software environment likely to be most accessible to a conservation practitioner. To reflect what practitioners would encounter, I used environmental and species presence datasets that are publically available. As described above, species presence data is sparse in general, and threatened species data is sparse in particular as well as being sensitive (and thus restricted) data. Although I used only species records with a high accuracy, incorporating bias adjustments into the models using recommended methods (Phillips et al. 2009) might have increased the representativeness of the models. However, over-fitting models when data are sparse is also a risk. I assessed model predictive performance (supplementary information of Chapter 2) and scrutinised which models to implement.

Because of deficiencies in species location data, I was unable to model all of the priority species in the region (Department of Environment and Resource Management 2010). Given sufficient time and resources, additional surveys to obtain more species presence data points would improve model accuracy and allow more species distributions to be modelled. My goal was to explore decision- support approaches using existing data so that actions can be undertaken now, rather than in some unspecified future. Using predicted species distribution modelling for decision support ties theory and practice, and may contribute to positive conservation outcomes (Guisan et al. 2013).

6.3.3 Socio-economic costs

It is not common practice to explicitly incorporate economic costs of management into conservation planning, but my research assumes that relative management and opportunity costs of threat

174

mitigation can be estimated for the purpose of prioritising cost-effective actions. An increasingly large body of literature describes estimation of financial costs of management over space (e.g. Balmford et al. 2003, Naidoo et al. 2006, Polasky et al. 2008, Moilanen et al. 2011b, Carwardine et al. 2012, McCarthy et al. 2012). Conservation costs include acquisition, management, transaction, damage, and opportunity costs, and the different types of costs don’t necessarily align with each other, i.e. a high acquisition cost does not imply a high management cost (Naidoo et al. 2006). In addition, costs of conservation are offset by returns in ecosystem services (Daily 1997) such as carbon sequestration and freshwater filtration (Kovacs et al. 2013, Kramer et al. 2013) plus the existence value of biodiversity (Mace et al. 2012). I have provided the methods I used to model management costs in the supplementary information of Chapter 3 (fox management) and Chapter 4 (fire management); I refer to the opportunity cost of grazing management as foregone agriculture profit (Marinoni et al. 2012). To cover more possibilities of the true costs involved in abating threats, future research could be informed by accounting for additional costs, and offsets to costs, in addition to the management and opportunity cost models I used in my analyses.

6.3.4 Ecological considerations

Including a larger selection of ecological considerations into my analyses would create more complete scenarios. Quantitative targets, i.e. minimal areas intended for protection, often inform conservation planning because they are transparent, simply conveyed, and measureable (Carwardine et al. 2009). In the absence of species-specific population viability analyses (Possingham et al. 2001a) to estimate the area needed for sustaining populations, alternative ways to quantify target areas have been proposed. For example, determining species-specific conservation targets would deliver more equitability amongst species as compared to employing uniform targets in systematic conservation planning, or no targets at all in cost-effectiveness analysis (CEA) such as in this thesis. Depending on information availability, area needed for threat management could be derived on the basis of mammal life history characteristics (Burns et al. 2013), modified by home range sizes (Wilson et al. 2010), dispersal distances (Faleiro et al. 2013), population viability (Haight et al. 2002), or extinction risk (Burgman et al. 2001). More research would be needed on methods to estimate target area needs for other taxa including birds, amphibians, reptiles, fish and insects. Experts may contribute to evaluating the size of management areas needed for individual species to persist and recover, and ideally, protected and managed areas should be designed to maximise species persistence (Haight and Travis 2008). However, incomplete knowledge regarding ecological requirements for species would hamper this additional area of research, and may require substitution of known information from similar species to those in the analysis. Alternatively, differential target sizes could be calculated depending on species vulnerability (Lombard et al. 175

2003, Pressey et al. 2003), although in some cases even managing 100% of the remaining habitat of a species may be insufficient to offset the extinction debt (Tilman et al. 1994, Brooks et al. 1999).

The research could be extended to include complementarity into the CEA (Underwood et al. 2008, Withey et al. 2012). To more efficiently distribute management actions amongst species in a CEA, the area of habitat managed for each species could be calculated sequentially as cost-efficient management sites were selected in rank order. Upon reaching a specified area of management for a species, benefit for further managing that species would be minimised. This analysis could be specifically incorporated in the CEA analysis of Chapter 5. Other ecological considerations that could be integrated include protected area design guidelines such as maximising connectivity and minimising the edge-to-area ratio of the threat management areas (Diamond 1975, Margules et al. 1982). While more computationally complex (Calabrese and Fagan 2004) to incorporate in the basic CEA conducted in this research, these guidelines are straightforward to incorporate into Marxan systematic conservation planning analyses (Ball et al. 2009, Watts et al. 2009) and can be formulated as spatial objectives in integer programming (IP) models (Haight and Snyder 2009). In addition, Zonation systematic conservation planning software could be implemented for a comparative analysis and alternative management planning scenarios (Moilanen 2007). Another unexplored area in this research is to optimise solutions for likelihood of persistence (Moilanen and Cabeza 2002, Haight and Travis 2008). While increasing the complexity of the models, incorporating more ecological considerations may be critical to increase the likelihood that management will have positive conservation outcomes.

6.3.5 Likelihood of management success

Including an estimate of the likelihood of management success enhances prioritisations for threat abatement action (Briggs 2009, Joseph et al. 2009, Wilson et al. 2011a, Carwardine et al. 2012, Carwardine et al. 2014). There is a clear trade-off between funding more cost-efficient management that is likely to succeed, as compared to less cost-efficient management for high-value species that has a lower likelihood of success (Joseph et al. 2009). Probability of functional species persistence also affects optimal management strategies (Chadès et al. 2014). Furthermore, although incorporating likelihood of success may make little difference to which of multiple actions is best for abating the single threat of habitat loss (protected areas versus restoration), it may affect the spatial location of the investments (Wilson et al. 2011a). It follows that integrating likelihood of management success could change which action or combination of actions is the best investment for managing threatened species at a particular location. Regional or species experts could contribute estimates of likelihood of management success (Burgman et al. 2011). Alternatively, success could

176

be evaluated by actual performance metrics -- data gathered through monitoring programs that relate to the achievement (or lack of achievement) of management objectives. I based this thesis research on information in the “Burnett-Mary Back-on-Track Recovery Plan,” and threats to species documented within that plan were informed by expert opinion (Department of Environment and Resource Management 2010). Thus, the threats-to-species data I used is indeed reflective of regional experts. Including likelihood of management success and probability for species persistence in this work would improve its robustness. However, the challenge would be estimating uncertainty in management effectiveness at a fine spatial scale across landscapes. Other work that incorporates uncertainty in the ability of threat management to benefit species assumes homogenous threats (Carwardine et al. 2012, Carwardine et al. 2014, Chadès et al. 2014)and a single feasibility value for each action, but in reality it is much more complex.

6.3.6 Sensitivity analyses

Sensitivity analysis is used to investigate how changing the parameters and assumptions of any model affects the conclusions drawn from the model (Pannell 1997, Burgman et al. 2005). A case study of one real-world scenario doesn’t necessarily allow research to explore the potential range or limits of how economically based management prioritisation operates. The benefit of basing research on one case study is that data requirements are minimised, the case reflects the situation faced by real-world managers, and time can be spent on an in-depth exploration of the sole case study. However, additional analyses could be conducted in parallel landscapes, with different species, and different threats and actions to abate threats, and conclusions drawn from the analyses could be compared with one another. Examining results for differences would further understanding about patterns depending on context. Alternatively, sensitivity analysis can be used to evaluate the drivers of any patterns observed in the analysis. The exploration of how dependencies between the benefits and costs of threat-abating actions affect where and how we would manage (Chapter 4) would benefit from a more thorough sensitivity analysis. I found that although differences are apparent, they were less extreme than we expected. Closer examination of the data reveals that while the research question is important (do dependencies make a difference in where and how we would manage threats?), my analysis does not provide a comprehensive answer. In part, my research was limited by the availability of management cost data, and I examined interactions between the costs and benefits of three management actions. In reality, the species of concern in the case study are affected by many more threats, which we could not represent in the analysis. Because of this, the analysed data represents fewer species and fewer threats than reality, so the real complexity is not adequately represented. While a simpler data set is beneficial in reducing the complexity of the situation, my analysis is reflective of a scenario derived from limited data, and it 177

did not allow me to fully explore the implications of interactions between multiple species, threats, actions, and the costs and benefits of actions. Therefore this research would be strengthened by sensitivity analyses that either compares my results with those from another system, or I could further explore the implications from ‘what-if’ scenarios, where I modified the data of the real-life system to represent other potential realities as in Appendix 4I. In this sensitivity analysis, I examined what the management implications of dependences were if all the plants in the case study are all affected by both fire and grazing, which is not an unlikely scenario. Interpreting the results of sensitivity analyses will build on the research of Chapter 4, and will strengthen understanding of answers to the research questions.

6.3.7 Uncertainty

Uncertainty is a pervasive feature of natural resource management, but an understanding of its nature may lead to accommodating, and/or reducing, its effects on decision making. One taxonomy (Regan et al. 2002) classifies the main types of uncertainty in ecology and conservation biology into two main categories: epistemic and linguistic uncertainty (uncertainty in determinate facts, as compared to in language). The two categories of uncertainty are further subdivided into 11 main types (Table 6.1). In natural resource management, the main sources of uncertainty are: environmental variation, limited system observability, partial controllability of management actions, and structural uncertainty (Conroy et al. 2011). In Regan et al.’s (2002) taxonomy (Table XX), these epistemic uncertainties may be categorised as types of natural variation, measurement error, subjective judgement, and model uncertainty. Natural variation is often regarded as a source of epistemic uncertainty because environmental systems change with respect to time, space, and other independent variables. When represented in a model, there is difficulty in estimating an environmental parameter’s value (Regan et al. 2002). Uncertainty in measurement error arises because a natural system can be only partially observed, and a quantified description must be estimated. Subjective judgement may be necessary due to partial controllability in applying management actions directly and with great precision, requiring an estimate of its efficacy over time. Model uncertainty pertains to the uncertainty in predicting system responses to specific management actions (Conroy et al. 2011). Models are necessarily a simplification that preserves the essential features of the problem (Levins 1966), and represent only the variables and processes that are chosen as relevant for the purpose of understanding a problem--constructs are simplifications of complex processes (Regan et al. 2002).

178

Categories of Description Main types of uncertainty in uncertainty category

Epistemic Uncertainty associated with knowledge of the measurement error, (uncertainty in state of a system and including uncertainty due systematic error, natural determinate to limitations of measurement devices, variation, facts) insufficient data, extrapolations and inherent randomness, model interpolations, and variability over time or space uncertainty, and

subjective judgment

Linguistic Uncertainty that arises because natural language vagueness, context (uncertainty in and scientific vocabulary is underspecified, dependence, ambiguity, language) ambiguous, vague, context dependent, and/or indeterminacy of theoretical exhibits theoretical indeterminacies terms, and underspecificity

Table 6.1 A taxonomy of uncertainty in ecology and conservation biology (from Regan et al. 2002).

Uncertainty in determining the truth about the effects of management on a system will affect decision-making. Therefore, reducing uncertainty in a model of the system of interest, or at least understanding how uncertainty affects the system model, is important. Whereas model uncertainty is known to be difficult to quantify, let alone eliminate, general approaches for accommodating model uncertainty include representing the natural variability in parameters by probability distributions and intervals, and validating and revising the model by observation (Regan et al. 2002). Deriving alternative models of system dynamics, each with associated measures of relative credibility, is a common approach to understanding the effects of model uncertainty (Conroy et al. 2011). Validation and revision by observation will be later discussed in the context of adaptive management, but knowledge gained by monitoring the results of management actions help to reduce measurement error and subjective judgement uncertainties, and will reduce model uncertainty.

A major assumption in this thesis research is that species distribution models based on available data adequately represent spatial variability in both species and their threats. If the species distribution models are to be used prescriptively, a more thorough uncertainty analysis is warranted, but was beyond the scope of the current research. Using the uncertainty taxonomy framework of Regan et al. (2002), Elith et al. (2002) describe types of uncertainty associated with species distribution models, and suggest methods for communicating these uncertainties to stakeholders.

179

Other research has focused on improving robustness of conservation planning to uncertainties in species distribution models (Moilanen et al. 2006a, Moilanen and Wintle 2006, Moilanen et al. 2006b, Carvalho et al. 2011b, Bagchi et al. 2013, Tulloch et al. 2013b).

Despite its importance, in most cases uncertainty in species distribution modelling is ignored because it is hard to deal with (Elith and Leathwick 2009). However, species distribution modelling is a rapidly evolving field. For example, a recent tool for simulating uncertainty in species distribution models, specifically uncertainty in data quality, derives probabilistic spatial predictions of predicted presence using a Monte Carlo process (Gould et al. 2014). The tool compares a species distribution derived when ignoring uncertainty to simulations of the effects of uncertainty from different sources, and provides probabilistic, spatially explicit illustrations of how uncertainty affects model projections, enabling explicit communication of the impacts of uncertainty on spatial prediction. Gould et al. (2014) found that uncertainty may blur the margins of a predicted distribution, as well as skew the results. For species distribution model performance validation and assessment in this thesis (section 2.7.1, Appendix 2A), my approach for reducing measurement error uncertainty was to restrict data input to only species presences recorded after 1950 with locational accuracy <500 m (a distance equivalent to the raster cell resolution used in the analysis), and to restrict models to those passing rigorous statistical assessment (Fielding and Bell 1997, Pearson 2007, Kremen et al. 2008). My approach could be further strengthened by correcting for sampling bias (systematic error uncertainty) (Phillips and Dudík 2008). Further research is required for addressing other sources of uncertainty in predicted species distributions specific to cost- effectiveness analysis for managing threats.

It is evident that uncertainty is an important consideration in predicting responses of natural resource management (Conroy et al. 2011). In this thesis, it was not possible to entirely eliminate these sources of uncertainty. For the purpose of understanding the thesis problem, I used a simple construct for representing a more complex process. I chose the most relevant variables for understanding the problem of prioritising threat management by cost-effectiveness as being the spatial distribution of multiple species and their threats, and the cost of managing those threats. To address the issue of partial observability of species and threat occurrence (measurement error uncertainty) would likely require time and resource-intensive efforts for gathering observations of rarely observed species. Furthermore in simplifying the model, I make the assumption that management is fully effective in reducing threats (subjective judgement uncertainty), although admittedly only partial controllability of management actions is likely. However, I discuss the utility of including management feasibility (likelihood of success) in section 6.3.5. Including a spatial estimate of management feasibility would increase the realism of the problem, but also the 180

complexity. Examples of uncertainty treatment in other research examining threat management prioritisation include altering cost-effectiveness values by interval percentages to determine robustness to changes in estimates of costs, benefits, and feasibility (Carwardine et al. 2012, Carwardine et al. 2014) and examining performance of a cost-effectiveness ranking approach relative to solutions with complementary for different species persistence thresholds (Chadès et al. 2014).

Formal development of how uncertainty is to be considered in decision-science is an important topic of research that is of great interest to many in the field of conservation. However, the focus of research in this thesis was on decision-support for prioritising threat management at a fine scale, based on cost-effectiveness of management actions. This decision-support is intended to provide managers with a mechanism for choosing actions within constraints of a small budget, limited data, and limited opportunities to implement complex decision-support approaches that may be perceived as being inaccessible. Nevertheless, the thesis does include discussion about model/structural/parameter uncertainty: The utility of sensitivity analyses is discussed in section 6.3.6, a sensitivity analysis clarifying the drivers of cost effectiveness is in section 3.7.5, Appendix 3E, a sensitivity analysis about variation in management cost is in section 3.7.7 Appendix 3G, and section 3.6 describes the results of testing three different benefit metrics and their effect on return on investment in management action. In addition, subsequent to the submission of the thesis, a sensitivity analysis was conducted for the dependency analysis in Chapter 4, as proposed in section 6.3.6; a manuscript including the sensitivity analysis is under peer-review. The differences in return on investment between scenarios of benefit dependence were greater when species are affected by more threats, but followed similar patterns. Although the treatment of uncertainty in this thesis was not exhaustive, future research could examine further effects of uncertainty.

6.3.8 Adaptive Management

To reduce uncertainty, adaptive management (Holling 1978, Walters 1986) is proposed as a structured, iterative process for decision making in which management practice is improved by integrating new information about the system that has been gained by learning which actions are successful. A key aspect of adaptive management is that it addresses uncertainty by incrementally improving imperfect knowledge about system response to ongoing management (Keith et al. 2011). To implement adaptive management, a practitioner would explicitly define their goals, develop and implement more than one alternative strategy for achieving those goals, monitor and evaluate the results, and modify their management based on the evidence of a defined improvement or decline of the system as influenced by the management action(s) (McCarthy and Possingham 2007). Key to

181

the practice of adaptive management is grounding the alternative strategies for achieving management goals in a comparative experimental and observational framework. The purpose of managing with two or more alternative strategies is that it serves to spread risk of failure while increasing understanding of the system’s response to the different actions (Lindenmayer and Burgman 2005).

While the underlying intent of adaptive management has been widely recognised in the scientific literature and management agencies as being a logical approach to embracing and reducing uncertainty, it has not often been successfully implemented (Walters 1997). Furthermore, even with the goal of practicing adaptive management in mind, many practitioners still engage in what may be characterised as trial and error management, with ad hoc approaches to continuous management improvement (Duncan and Wintle 2008). The problem with trial and error management, which usually involves using only one management strategy until it fails, is the lack of formal models underpinning the system being managed, lack of recognition of uncertainty, and lack of a plan for learning through replication or comparative analysis of results (Duncan and Wintle 2008). On the other hand, when management actions are common-sense, the value of additional reduction in uncertainty may be negligible: in certain cases, adaptive management may not be the best approach, depending upon criteria of temporal and spatial scale, dimensions of uncertainty, costs, benefits and risks; and institutional support (Gregory et al. 2006). However, if resources to attempt adaptive management are available, it is likely the best approach for reducing uncertainty in management actions (Shea et al. 1998).

Structural or model uncertainty in a management problem may be represented by alternative models of system dynamics, and used as a guide for predicting management outcomes in an adaptive management framework. Once management actions are chosen for implementation, monitoring the management outcomes would increase system knowledge, and help further reduce model uncertainty. Future research that combines cost effectiveness analysis with adaptive management planning could involve initially examining predicted outcomes from alternative management actions for each individual threat, rather than the return-on-investment outcomes from any one chosen management approach to a single threat as examined in this thesis. Interval estimates of likelihood of management success and feasibility, as well as estimates of the probability of species persistence could be included. In addition, the cost of monitoring management outcomes could be included in the estimate of management costs. Wise selection of indicator species for monitoring management success could make monitoring costs more affordable (Tulloch et al. 2011). The information determined by monitoring the success or failure of management actions over time could then be used to improve the system model and reduce uncertainty. If examined at a fine spatial 182

scale, such dimensionality would increase computational overhead, would be informative for strategic management, and is an interesting problem for future research.

6.4 Concluding remarks

I conclude this thesis with my closing remarks on the state of biodiversity and what I philosophically consider to be the largest challenges in the field of conservation. I believe there is a need not only to innovatively draw upon science to make biodiversity conservation more effective, but to expose, and change the modus operandi of environmental degradation (Noss and Cooperrider 1994) that is contributing to the dire straits faced by threatened species in particular and biodiversity in general. Biodiversity is undervalued both for its own sake, as well as for its contributions to human welfare (Rands et al. 2010, Noss et al. 2012), and is continuing to decline.

Referring to observed and ongoing changes in the environment, indigenous elders offer the insight that “the Earth is faster now” (Krupnik and Jolly 2002). Current species extinction rates have been estimated to be 100–1000 times their pre-human levels (Pimm et al. 2014), and the rate of biodiversity loss may signal the crossing of a critical planetary threshold with unknown, but potentially profound, consequences to humanity (Rockström et al. 2009). The current geological current epoch has been referred to as the ‘Anthropocene’ (Crutzen 2002) because human beings are significantly altering the Earth’s natural processes (Steffen et al. 2007). By fundamentally and irreversibly affecting biodiversity and ecosystem processes (Ellis and Ramankutty 2008, Ellis et al. 2010), new and potentially less-desirable ecosystem states are emerging. Planetary stewardship is both a necessity, and a responsibility that people have the power to fulfil (Steffen et al. 2011), and preventing further irreversible losses in biodiversity is a priority (Chapin III et al. 2000). Conservation of biodiversity is imperative not only because biodiversity drives many ecosystem services (Mace et al. 2012, Tilman 2012) and can be appreciated as a public good (Rands et al. 2010), but also because of its fundamental, intrinsic, and sacred value (Soulé 1985, Noss and Cooperrider 1994).

Sacred values are defined as those “that a moral community treats as possessing transcendental significance that precludes…trade-offs” (Tetlock 2003, p 320). However, in a world of limited resources, trade-offs are unavoidable – both taboo trade-offs that pit sacred against secular values, as well as tragic trade-offs that pit sacred values against each other (Tetlock 2003). Biodiversity conservation is considered to be a sacred value, yet finite resources are allocated to its cause; one estimate is that half of one percent of American charitable giving is directed to protection of vulnerable species (Soulé 2014). Although effective biodiversity conservation lies within human means (James et al. 2001), funding falls at least an order of magnitude short (McCarthy et al. 2012). 183

This disconnect in valuing biodiversity forces both taboo trade-offs, i.e. there’s not enough money to save everything, as well as tragic trade-offs, i.e. which species do we attempt to manage and which do we observe become extinct through inaction (Marris 2007).

Whereas this research wrestles with the taboo trade-off known as triage or “prioritising the allocation of limited resources to maximise conservation returns, relative to the conservation goals, under a constrained budget” (Bottrill et al. 2008 p 649), it is underpinned with the additional goal of exposing and calling attention to the unacknowledged tragic trade-offs that occur when this taboo trade-off is not made explicit. Doing nothing is a decision not to act, and lack of resources to act means that the needs of many species are not being addressed. With the transparency of conscious decisions, we cannot profess ignorance of our unacknowledged actions. This research aims to bring about mindful decisions for biodiversity conservation.

The problems of conserving biodiversity are complex and meet the characteristics of ‘wicked’ problems (Rittel and Webber 1973), requiring urgent innovation in conservation practices (Game et al. 2014). The central motivation in the field of conservation science is determining the most- effective ways to undertake biodiversity conservation action based on rigorous science (Soulé 1985, Marris 2014). The field has long addressed a moving target as threats multiply and spread, species of concern increase in number, and as approaches to conservation and management actions evolve. However, I believe our further challenge is overcoming the status quo of a societal focus on materialism that leads to environmental degradation (Kilbourne and Pickett 2008) with disregard for, or unawareness of, its effects on biodiversity. My research does not provide a response to addressing the disregard for biodiversity brought about by the tragedy of the commons and pursuit of personal gain over the welfare of the environment (Garett 1968). However, my research does address increasing awareness of how limited funding that could be directed to any number of ends could be spent to benefit more species, while being transparent and accountable. It is my conviction that the trade-offs exposed by implementing systematic and transparent decision support and prioritisation will increase awareness of how limited money is spent on biodiversity management (Myers et al. 2000, Bottrill et al. 2009). Research in cost-effective conservation can provide direction for benefitting biodiversity in terms that can be communicated to economically rational policy makers and managers who have an influence on conservation outcomes. It is my sincere hope is that my research can make a difference that leads to positive conservation outcomes.

184

185

References

7 References

ACIL Tasman. 2005. Institutional Arrangements in the Australian Water Sector. National Water Commission, Canberra. Allan, J. D., P. B. McIntyre, S. D. Smith, B. S. Halpern, G. L. Boyer, A. Buchsbaum, G. Burton, L. M. Campbell, W. L. Chadderton, and J. J. Ciborowski. 2013. Joint analysis of stressors and ecosystem services to enhance restoration effectiveness. Proceedings of the National Academy of Sciences 110:372-377. Ando, A., J. Camm, S. Polasky, and A. Solow. 1998. Species distributions, land values, and efficient conservation. Science 279:2126-2128. Anonymous. 1999. Threat abatement plan for predation by the Red Fox. Environment Australia, Canberra. Armsworth, P. R., G. C. Daily, P. Kareiva, and J. N. Sanchirico. 2006. Land market feedbacks can undermine biodiversity conservation. Proceedings of the National Academy of Sciences of the United States of America 103:5403-5408. Armsworth, P. R., and J. E. Roughgarden. 2001. An invitation to ecological economics. Trends in Ecology & Evolution 16:229-234. Arponen, A., M. Cabeza, J. Eklund, H. Kujala, and J. Lehtomäki. 2010. Costs of integrating economics and conservation planning. Conservation Biology 24:1198-1204. Atlas of Living Australia (ALA) 2012. Atlas of Living Australia: Sharing biodiversity knowledge. – Global Biodiversity Information Facility ALA. http://www.ala.org.au Accessed 13 August 2012. Auerbach, N., A. Tulloch, and H. Possingham. In prep-a. Accounting for complementarity and action dependencies in spatial threat-management prioritisation. Auerbach, N. A., M. C. Evans, and H. P. Possingham. In prep-b. All threats are not equal: Coarse and fine-scaled indirect threat models compared. Auerbach, N. A., A. I. T. Tulloch, and H. P. Possingham. 2014. Informed actions: Where to cost- effectively manage multiple threats to species to maximise return on investment. Ecological Applications 24:1357-1373. Auerbach, N. A., K. A. Wilson, A. I. T. Tulloch, J. R. Rhodes, and H. P. Possingham. In review. Accounting for dependencies in threat management can alter conservation priorities. Conservation Biology. Auld, T. D., and D. A. Keith. 2009. Dealing with threats: Integrating science and management. Ecological Management & Restoration 10:S79-S87.

189

Australian Broadcasting Corporation (ABC) News 2009. Garrett rejects Traveston Dam. http://www.abc.net.au/news/2009-11-11/garrett-rejects-traveston-dam/1138124 Accessed 12 November 2011. Australian Bureau of Statistics (ABS) 2006. 1209.0.55.002 - Mesh Blocks Digital Boundaries, Australia, 2006. – ABS. http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/1209.0.55.0022006?OpenDocum ent Accessed 04 June 2013. Australian Bureau of Statistics (ABS) 2010. 6306.0 - Employee Earnings and Hours, Australia; 1d. Average weekly cash earnings and hours paid for, full-time non-managerial adult employees, Queensland—Detailed occupation (ANZSCO). – ABS. http://abs.gov.au/AUSSTATS/[email protected]/DetailsPage/6306.0May%202010?OpenDocument Accessed 10 September 2012. Australian Government 2010. Burnett Mary - Natural Resource Management region. – Caring for Our Country, http://www.bmrg.org.au/about/our-region/ Accessed 13 December 2010. Australian Taxation Office (ATO) 2012. D1 - Work-related car expenses 2012. –. http://www.ato.gov.au/individuals/content.aspx?doc=/content/00313836.htm Accessed 10 September 2012. Bagchi, R., M. Crosby, B. Huntley, D. G. Hole, S. H. Butchart, Y. Collingham, M. Kalra, J. Rajkumar, A. Rahmani, and M. Pandey. 2013. Evaluating the effectiveness of conservation site networks under climate change: accounting for uncertainty. Global Change Biology 19:1236-1248. Bakker, V. J., J. K. Baum, J. F. Brodie, A. K. Salomon, B. G. Dickson, H. K. Gibbs, O. P. Jensen, and P. B. McIntyre. 2010. The changing landscape of conservation science funding in the United States. Conservation Letters 3:435-444. Ball, I. R., H. P. Possingham, and M. E. Watts. 2009. Marxan and relatives: Software for spatial conservation prioritization. Pages 185-195 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Priorities Action: Quantitative Methods and Computational Tools. Oxford University Press, Oxford. Balmford, A., K. J. Gaston, S. Blyth, A. James, and V. Kapos. 2003. Global variation in terrestrial conservation costs, conservation benefits, and unmet conservation needs. Proceedings of the National Academy of Sciences of the United States of America 100:1046-1050. Balmford, A., K. J. Gaston, A. S. L. Rodrigues, and A. James. 2000. Integrating costs of conservation into international priority setting. Conservation Biology 14:597-605.

190

Ban, N. C., H. M. Alidina, and J. A. Ardron. 2010. Cumulative impact mapping: Advances, relevance and limitations to marine management and conservation, using Canada's Pacific waters as a case study. Marine Policy 34:876-886. Banks, P. B., C. R. Dickman, and A. E. Newsome. 1998. Ecological costs of feral predator control: foxes and rabbits. The Journal of wildlife management:766-772. Barnosky, A. D., N. Matzke, S. Tomiya, G. O. Wogan, B. Swartz, T. B. Quental, C. Marshall, J. L. McGuire, E. L. Lindsey, and K. C. Maguire. 2011. Has the Earth's sixth mass extinction already arrived? Nature 471:51-57. Barry, S., and J. Elith. 2006. Error and uncertainty in habitat models. Journal of Applied Ecology 43:413-423. Beale, C. M., and J. J. Lennon. 2012. Incorporating uncertainty in predictive species distribution modelling. Philosophical Transactions of the Royal Society of London B Biological Sciences 367:247-258. Beger, M., H. S. Grantham, R. L. Pressey, K. A. Wilson, E. L. Peterson, D. Dorfman, P. J. Mumby, R. Lourival, D. R. Brumbaugh, and H. P. Possingham. 2010. Conservation planning for connectivity across marine, freshwater, and terrestrial realms. Biological Conservation 143:565-575. Benito, B. M., L. Cayuela, and F. S. Albuquerque. 2013. The impact of modelling choices in the predictive performance of richness maps derived from species‐distribution models: guidelines to build better diversity models. Methods in Ecology and Evolution 4:327-335. Bennett, E. L. 2011. Another inconvenient truth: the failure of enforcement systems to save charismatic species. Oryx 45:476-479. Bennett, J. R., G. Elliott, B. Mellish, L. N. Joseph, A. I. T. Tulloch, W. J. M. Probert, M. M. I. Di Fonzo, J. M. Monks, H. P. Possingham, and R. Maloney. 2014. Balancing phylogenetic diversity and species numbers in conservation prioritization, using a case study of threatened species in New Zealand. Biological Conservation Restoration and Sustainability 174:47-54. Blackburn, T. M., P. Cassey, R. P. Duncan, K. L. Evans, and K. J. Gaston. 2004. Avian Extinction and Mammalian Introductions on Oceanic Islands. Science 305:1955-1958. Boitani, L., L. Maiorano, D. Baisero, A. Falcucci, P. Visconti, and C. Rondinini. 2012. What spatial data do we need to develop global mammal conservation strategies. Philosophical Transactions of the Royal Society of London B Biological Sciences 366:2623-2632. Bombi, P., L. Luiselli, and M. D’Amen. 2011. When the method for mapping species matters: defining priority areas for conservation of African freshwater turtles. Diversity and Distributions 17:581-592.

191

Bottrill, M. C., L. N. Joseph, J. Carwardine, M. Bode, C. Cook, E. T. Game, H. Grantham, S. Kark, S. Linke, and E. McDonald-Madden. 2009. Finite conservation funds mean triage is unavoidable. Trends in Ecology & Evolution 24:183-184. Bottrill, M. C., L. N. Joseph, J. Carwardine, M. Bode, C. Cook, E. T. Game, H. Grantham, S. Kark, S. Linke, E. McDonald-Madden, R. L. Pressey, S. Walker, K. A. Wilson, and H. P. Possingham. 2008. Is conservation triage just smart decision making? Trends in Ecology & Evolution 23:649-654. Box, G. E. P., and N. R. Draper. 1987. Empirical Model Building and Response Surfaces, . John Wiley & Sons, New York. Boyd, C., T. M. Brooks, S. H. Butchart, G. J. Edgar, G. A. Da Fonseca, F. Hawkins, M. Hoffmann, W. Sechrest, S. N. Stuart, and P. P. Van Dijk. 2008. Spatial scale and the conservation of threatened species. Conservation Letters 1:37-43. Boyd, J., R. Epanchin-Niell, and J. Siikamäki. 2012. Conservation return on investment analysis, a review of results, methods, and new directions. Discussion Paper, Resources for the Future, Washington, D.C. Bradstock, R. A., A. M. Gill, and R. J. Williams, editors. 2012. Flammable Australia: Fire Regimes, Biodiversity and Ecosystems in a Changing World. CSIRO Publishing, Canberra. Bradstock, R. A., J. E. Williams, and A. M. Gill, editors. 2002. Flammable Australia: The Fire Regimes and Biodiversity of a Continent. Cambridge University Press, Cambridge. Brandon, K., K. H. Redford, and S. E. Sanderson, editors. 1998. Parks in Peril. Island Press, Washington DC. Brashares, J. S. 2010. Filtering Wildlife. Science 329:402-403. Briggs, S. V. 2009. Priorities and paradigms: directions in threatened species recovery. Conservation Letters 2:101-108. Brook, B. W., N. S. Sodhi, and C. J. A. Bradshaw. 2008. Synergies among extinction drivers under global change. Trends in Ecology & Evolution 23:453-460. Brooks, T. M., R. A. Mittermeier, G. A. B. da Fonseca, J. Gerlach, M. Hoffmann, J. F. Lamoreux, C. G. Mittermeier, J. D. Pilgrim, and A. S. L. Rodrigues. 2006. Global biodiversity conservation priorities. Science 58:58-61. Brooks, T. M., S. L. Pimm, and J. O. Oyugi. 1999. Time lag between deforestation and bird extinction in tropical forest fragments. Conservation Biology 13:1140-1150. Brown, C. J., M. I. Saunders, H. P. Possingham, and A. J. Richardson. 2013a. Interactions between global and local stressors of ecosystems determine management effectiveness in cumulative impact mapping. Diversity and Distributions:1-9.

192

Brown, C. J., M. I. Saunders, H. P. Possingham, and A. J. Richardson. 2013b. Managing for interactions between local and global stressors of ecosystems. PLoS ONE 8:e65765. doi:65710.61371/journal.pone.0065765. Bryant, D., D. Nielsen, and L. Tangley. 1997. Last frontier forests: ecosystems and economies on the edge. World Resources Institute, Washington, DC. Burbidge, A. A., and B. F. J. Manly. 2002. Mammal extinctions on Australian islands: causes and conservation implications. Journal of Biogeography 29:465-473. Burgman, M., A. Carr, L. Godden, R. Gregory, M. McBride, L. Flander, and L. Maguire. 2011. Redefining expertise and improving ecological judgment. Conservation Letters 4:81-87. Burgman, M., D. Lindenmayer, and J. Elith. 2005. Managing landscapes for conservation under uncertainty. Ecology 86:2007-2017. Burgman, M. A., D. A. Keith, S. D. Hopper, D. Widyatmoko, and C. Drill. 2007. Threat syndromes and conservation of the Australian flora. Biological Conservation 134:73-82. Burgman, M. A., H. P. Possingham, A. J. J. Lynch, D. A. Keith, M. A. McCarthy, S. D. Hopper, W. L. Drury, J. A. Passioura, and R. J. Devries. 2001. A method for setting the size of plant conservation target areas. Conservation Biology 15:603-616. Burnett Mary Regional Group (BMRG). 2008. Burnett Mary Regional Group Multi Species Recovery Strategy, Unpublished document. Burns, E. S., S. F. Tóth, and R. G. Haight. 2013. A modeling framework for life history-based conservation. Biological Conservation 158:14-25. Busch, J., and R. Cullen. 2009. Effectiveness and cost-effectiveness of yellow-eyed penguin recovery. Ecological Economics 68:762-776. Butchart, S. H. M., M. Walpole, B. Collen, A. Van Strien, J. P. W. Scharlemann, R. E. A. Almond, J. E. M. Baillie, B. Bomhard, C. Brown, J. Bruno, K. E. Carpenter, G. M. Carr, J. Chanson, A. M. Chenery, J. Csirke, N. C. Davidson, F. Dentener, M. Foster, A. Galli, J. N. Galloway, P. Genovesi, R. D. Gregory, M. Hockings, V. Kapos, J.-F. Lamarque, F. Leverington, J. Loh, M. A. McGeoch, L. McRae, A. Minasyan, M. H. Morcillo, T. E. E. Oldfield, D. Pauly, S. Quader, C. Revenga, J. r. Sauer, B. Skolnik, D. spear, D. Stanwell-Smith, S. N. Stuart, A. Symes, M. Tierney, T. D. Tyrrell, J.-C. Vié, and R. Watson. 2010. Global biodiversity: Indicators of recent declines. Science 328:1164-1168. Cabeza, M., A. Arponen, L. Jäättelä, H. Kujala, A. van Teeffelen, and I. Hanski. 2010. Conservation planning with insects at three different spatial scales. Ecography 33:54-63. Cabeza, M., and A. Moilanen. 2001. Design of reserve networks and the persistence of biodiversity. Trends in Ecology & Evolution 16:242-248.

193

Calabrese, J. M., and W. F. Fagan. 2004. A comparison-shopper's guide to connectivity metrics. Frontiers in Ecology and the Environment 2:529-536. Carpenter, S. R., R. De Fries, T. Dietz, H. A. Mooney, S. Polasky, W. V. Reid, and R. J. Scholes. 2006. Millennium ecosystem assessment: research needs. Carpenter, S. R., H. A. Mooney, J. Agard, D. Capistrano, R. S. DeFries, S. Díaz, T. Dietz, A. K. Duraiappah, A. Oteng-Yeboah, and H. M. Pereira. 2009. Science for managing ecosystem services: Beyond the Millennium Ecosystem Assessment. Proceedings of the National Academy of Sciences 106:1305-1312. Carter, A., G. W. Luck, and S. P. McDonald. 2011. Fox-baiting in agricultural landscapes in south- eastern Australia: a case-study appraisal and suggestions for improvement. Ecological Management & Restoration 12:214-223. Carvalho, S., J. C. Brito, E. J. Crespo, and H. P. Possingham. 2011a. Incorporating evolutionary processes into conservation planning using species distribution data: a case study with the western Mediterranean herpetofauna. Diversity and Distributions 17:408-421. Carvalho, S. B., J. C. Brito, E. G. Crespo, M. E. Watts, and H. P. Possingham. 2011b. Conservation planning under climate change: toward accounting for uncertainty in predicted species distributions to increase confidence in conservation investments in space and time. Biological Conservation 144:2020-2030. Carwardine, J. 2009. Cost-effective conservation planning. PhD Thesis. University of Queensland, Brisbane. Carwardine, J., C. J. Klein, K. A. Wilson, R. L. Pressey, and H. P. Possingham. 2009. Hitting the target and missing the point: target-based conservation planning in context. Conservation Letters 2:3-10. Carwardine, J., S. Nicol, S. van Leeuwen, B. Walters, J. Firn, A. Reeson, T. G. Martin, and I. Chadès. 2014. Priority threat management for Pilbara species of conservation significance. CSIRO Ecosystem Sciences, Brisbane. Carwardine, J., T. O’Connor, S. Legge, B. Mackey, H. P. Possingham, and T. G. Martin. 2012. Prioritizing threat management for biodiversity conservation. Conservation Letters 5:196- 204. Carwardine, J., K. A. Wilson, G. Ceballos, P. R. Ehrlich, R. Naidoo, T. Iwamura, S. A. Hajkowicz, and H. P. Possingham. 2008. Cost-effective priorities for global mammal conservation. Proceedings of the National Academy of Sciences of the United States of America 105:11446-11450.

194

Chadès, I., E. McDonald-Madden, M. A. McCarthy, B. Wintle, M. Linkie, and H. P. Possingham. 2008. When to stop managing or surveying cryptic threatened species. Proceedings of the National Academy of Sciences 105:13936-13940. Chadès, I., S. Nicol, S. van Leeuwen, B. Walters, J. Firn, A. Reeson, T. Martin, and J. Carwardine. 2014. Complementarity threat management priorities save more species. Conservation Biology. Chapin III, F. S., E. S. Zavaleta, V. T. Eviner, R. L. Naylor, P. M. Vitousek, H. L. Reynolds, D. U. Hooper, S. Lavorel, O. E. Sala, S. E. Hobbie, M. C. Mack, and S. Diaz. 2000. Consequences of Changing Biodiversity. Nature 405:234-242. Church, R. L., D. M. Stoms, and F. W. Davis. 1996. Reserve selection as a maximal covering location problem. Biological Conservation 76:105-112. Clark, J. A., and E. Harvey. 2002. Assessing multi-species recovery plans under the Endangered Species Act. Ecological Applications 12:655-662. Cohen, J. 1960. A coefficient of agreement for nominal scales. Educational and psychological measurement 20:37-46. Cohen, J. 1968. Weighted kappa: Nominal scale agreement provision for scaled disagreement or partial credit. Psychological Bulletin 70:213-220. Connor, R. C. 1995. The benefits of mutualism: a conceptual framework. Biological Reviews 70:427-457. Conroy, M. J., and B. R. Noon. 1996. Mapping of species richness for conservation of biological diversity: conceptual and methodological issues. Ecological Applications:763-773. Conroy, M. J., M. C. Runge, J. D. Nichols, K. W. Stodola, and R. J. Cooper. 2011. Conservation in the face of climate change: the roles of alternative models, monitoring, and adaptation in confronting and reducing uncertainty. Biological Conservation 144:1204-1213. Convertino, M., P. Welle, R. Muñoz-Carpena, M. L. Chu-Agor, R. A. Fischer, and I. Linkov. 2011. Epistemic uncertainty in predicting shorebird biogeography affected by sea-level rise. Ecological Modelling 240:1-15. Crain, C. M., K. Kroeker, and B. S. Halpern. 2008. Interactive and cumulative effects of multiple human stressors in marine systems. Ecology Letters 11:1304-1315. Crutzen, P. 2002. Geology of mankind. Nature 415:23. Cullen, R., E. Moran, and K. F. D. Hughey. 2005. Measuring the success and cost effectiveness of New Zealand multiple-species projects to the conservation of threatened species. Ecological Economics 53:311-323.

195

Cullen, R., and P. C. White. 2013. Prioritising and evaluating biodiversity projects. Wildlife Research 40:91-93. Curtis, A., A. Britton, and J. Sobels. 1999. Landcare networks in Australia: State-sponsored participation through local organizations. Journal of Environmental Planning and Management 42:5-21. Daily, G., editor. 1997. Nature's Services: Societal Dependence on Natural Ecosystems. Island Press, Washington D.C. Darling, E. S., and I. M. Côté. 2008. Quantifying the evidence for ecological synergies. Ecology Letters 11:1278-1286. de Tores, P., K. Himbeck, M. Dillon, J. Cocking, B. MacArthur, and S. Rosier. 1998. Large scale fox control in the northern Jarrah forest of southwest Western Australia. Pages 275-281 in 11th Australian Vertebrate Pest Conference. Agriculture Western Australia, Bunbury, WA. Department of Environment and Resource Management 2010. Burnett Mary Natural Resource Management Region - Back on Track Actions for Biodiversity. – Queensland Government, https://www.ehp.qld.gov.au/wildlife/prioritisation-framework/bot-biodiversity- documents.html Accessed 09 August 2012. Department of Environment and Resource Management (DERM) 1999. Land Use Mapping 1999 Queensland (ISO 19139). – Queensland Government Information Service (QGIS). http://dds.information.qld.gov.au/dds/ Accessed 29 June 2011. Department of Environment and Resource Management (DERM) 2010. WildNet (Database). – Queensland Government. Accessed 28 October 2010. Department of Environment and Resource Management (DERM) 2011. HERBRECS (Database). – Queensland Government, Queensland Herbarium. Accessed 14 June 2011. Department of Mines and Energy (DME) 2008. Queensland Geological Mapping (Polygonised vector) Data - Regional –Queensland Government Department of Employment, Economic Development and Innovation (DEEDI). Department of Natural Resources and Mines (DNRM) 2010. Dominant Soils of Queensland (ISO 19139). – Queensland Government Information Service (QGIS). http://dds.information.qld.gov.au/dds/. Department of the Environment, Water, Heritage, and the Arts (DEWHA) 2008a. Species of National Environmental Significance (Database). – Australian Government. Department of the Environment, Water, Heritage, and the Arts (DEWHA) 2008b. Species Profile and Threats (Database). – Australian Government. http://www.environment.gov.au/cgi- bin/sprat/public/sprat.pl Accessed 4 January 2011.

196

Di Minin, E., D. C. MacMillan, P. S. Goodman, B. Escott, R. Slotow, and A. Moilanen. 2013. Conservation businesses and conservation planning in a biological diversity hotspot. Conservation Biology 27:808-820. Diamond, J. M. 1975. The island dilemma: Lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7:129-146. Didham, R. K., J. M. Tylianakis, N. J. Gemmell, T. A. Rand, and R. M. Ewers. 2007. Interactive effects of habitat modification and species invasion on native species decline. Trends in Ecology & Evolution 22:489-496. Dobson, A. P., J. P. Rodriguez, W. M. Roberts, and D. S. Wilcove. 1997. Geographic distribution of endangered species in the United States. Science 275:550-553. Driscoll, D. A., S. C. Banks, P. S. Barton, D. B. Lindenmayer, and A. L. Smith. 2013. Conceptual domain of the matrix in fragmented landscapes. Trends in Ecology & Evolution 28:605-613. Dudík, M., S. J. Phillips, and R. E. Schapire. 2010. Maximum Entropy Modeling of Species Geographic Distributions. Department of Computer Science, Princeton University. Duke, J. M., S. J. Dundas, and K. D. Messer. 2013. Cost-effective conservation planning: Lessons from economics. Journal of Environmental Management 125:126-133. Duncan, D. H., and B. A. Wintle. 2008. Towards adaptive management of native vegetation in regional landscapes. Pages 159-182 Landscape analysis and visualisation. Springer. Eastman, J. R. 2009. IDRISI Taiga Guide to GIS and Image Processing. Clark Labs, Clark University, USA. Elith, J., C. H. Graham, R. P. Anderson, M. Dudík, S. Ferrier, A. Guisan, R. J. Hijmans, F. Huettmann, J. R. Leathwick, A. Lehmann, J. Li, L. G. Lohmann, B. A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, J. M. M. Overton, A. T. Peterson, S. J. Phillips, K. Richardson, R. Scachetti-Pereira, R. E. Schapire, J. Soberón, S. Williams, M. S. Wisz, and N. E. Zimmermann. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29:129-151. Elith, J., and J. R. Leathwick. 2009. The contribution of species distribution modelling. Pages 70-93 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial conservation prioritization: quantitative methods and computational tools. Oxford University Press, New York. Ellis, E. C., K. K. Goldewijk, S. Siebert, D. Lightman, and N. Ramankutty. 2010. Anthropogenic transformation of the biomes, 1700-2000. Global Ecology and Biogeography 19:589-606. Ellis, E. C., and N. Ramankutty. 2008. Putting people in the map: anthropogenic biomes of the world. Frontiers in Ecology and the Environment 6:439-447.

197

Epanchin-Niell, R. S., M. B. Hufford, C. E. Aslan, J. P. Sexton, J. D. Port, and T. M. Waring. 2009. Controlling invasive species in complex social landscapes. Frontiers in Ecology and the Environment 8:210-216. Ervin, J. 2003. Rapid assessment of protected area management effectiveness in four countries. Bioscience 53:833-841. Etter, A., C. McAlpine, K. Wilson, S. Phinn, and H. P. Possingham. 2006. Regional patterns of agricultural land use and deforestation in Colombia. Agriculture Ecosystems & Environment 114:369-386. Evans, M. C., R. A. Fuller, J. E. M. Watson, O. Venter, S. C. Bennett, P. R. Marsack, and H. R. Possingham. 2011a. The spatial distribution of threats to species in Australia. Bioscience 61:281-289. Evans, M. C., H. P. Possingham, and K. A. Wilson. 2011b. What to do in the face of multiple threats? Incorporating dependencies within a return on investment framework for conservation. Diversity and Distributions 17:437-450. Ewers, R. M., and A. S. L. Rodrigues. 2008. Estimates of reserve effectiveness are confounded by leakage. Trends in Ecology & Evolution 23:113-116. Ewing, S. A. 2003. Catchment management arrangements. Pages 393-412 in S. Dovers and S. Wild River, editors. Managing Australia's Environment. Federation Press, Leichhardt. Ewing, S. A., R. B. Grayson, and R. M. Argent. 2000. Science, citizens, and catchments: decision support for catchment planning in Australia. Society & Natural Resources 13:443-459. Faith, D. P., and P. Walker. 1996. Integrating conservation and development: effective trade-offs between biodiversity and cost in the selection of protected areas. Biodiversity and Conservation 5:431-446. Faleiro, F. V., R. B. Machado, and R. D. Loyola. 2013. Defining spatial conservation priorities in the face of land-use and climate change. Biological Conservation 158:248-257. Fernandes, L., J. Day, A. Lewis, S. Slegers, B. Kerrigan, D. Breen, D. Cameron, B. Jago, J. Hall, D. Lowe, J. Innes, J. Tanzer, V. Chadwick, L. Thompson, K. Gorman, M. Simmons, B. Barnett, K. Sampson, G. De'Atho, B. Mapstone, H. Marsh, H. Possingham, I. Ball, T. Ward, K. Dobbs, J. Aumend, D. Slater, and K. Stapleton. 2005. Establishing representative no-take areas in the Great Barrier Reef: Large-scale implementation of theory on marine protected areas. Conservation Biology 19:1733-1744. Ferraro, P. J. 2003. Assigning priority to environmental policy interventions in a heterogeneous world. Journal of Policy Analysis and Management 22:27-43.

198

Ferrier, S. 2002. Mapping spatial pattern in biodiversity for regional conservation planning: where to from here? Systematic Biology 51:331-363. Ferrier, S., and B. A. Wintle. 2009. Quantitative approaches to spatial conservation prioritization: Matching the solution to the need. Pages 1-15 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Prioritization: Quantitative Methods & computational Tools. Oxford University Press, New York. Fielding, A. H., and J. F. Bell. 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation 24:38-49. Fleischner, T. L. 1994. Ecological costs of livestock grazing in western North America. Conservation Biology 8:620-644. Folt, C., C. Chen, M. Moore, and J. Burnaford. 1999. Synergism and antagonism among multiple stressors. Limnology and Oceanography 44:864-877. Franklin, J. 2009. Mapping species distributions: spatial inference and prediction. Cambridge University Press, Cambridge. Franklin, J. 2010. Moving beyond static species distribution models in support of conservation biogeography. Diversity and Distributions 16:321-330. Franklin, J. F. 1993. Preserving biodiversity: species, ecosystems, or landscapes? Ecological Applications 3:202-205. Game, E. T., E. McDonald-Madden, M. L. Puotinen, and H. P. Possingham. 2008. Should we protect the strong or the weak? Risk, resilience, and the selection of marine protected areas. Conservation Biology 22:1619-1629. Game, E. T., E. Meijaard, D. Sheil, and E. McDonald‐Madden. 2014. Conservation in a wicked complex world; challenges and solutions. Conservation Letters 7:271-277. Garett, H. 1968. Tragedy of the commons. Science 162:1243-1248. Gaston, K. J., and R. A. Fuller. 2009. The sizes of species' geographic ranges. Journal of Applied Ecology 46:1-9. Gaston, K. J., A. S. L. Rodrigues, B. J. van Rensburg, P. Koleff, and S. L. Chown. 2001. Complementary representation and zones of ecological transition. Ecology Letters 4:4-9. Gentle, M. N. 2005. Factors affecting the efficiency of fox (Vulpes vulpes) baiting practices on the central tablelands of New South Wales. University of Sydney, Sydney, AU. Gentle, M. N., G. R. Saunders, and C. R. Dickman. 2007. Poisoning for production: how effective is fox baiting in south-eastern Australia? Mammal Review 37:177-190.

199

Geoscience Australia 2010a. Digital Elevation Model (25m) Burnett Catchment (Data Package) (ISO 19139). – Department of Environment and Resource Management Queensland Government Information Service. http://dds.information.qld.gov.au/dds/. Geoscience Australia 2010b. Digital Elevation Model (25m) South East Queensland (Data Package) (ISO 19139). – Department of Environment and Resource Management Queensland Government Information Service. http://dds.information.qld.gov.au/dds/. Geoscience Australia 2010c. Ordered Drainage 100K - Queensland (ISO 19139). – Department of Environment and Resource Management Queensland Government Information Service. http://dds.information.qld.gov.au/dds/. Gillingham, P. K., B. Huntley, W. E. Kunin, and C. D. Thomas. 2012. The effect of spatial resolution on projected responses to climate warming. Diversity and Distributions 18:990- 1000. Goldstein, J. H., L. Pejchar, and G. C. Daily. 2008. Using return-on-investment to guide restoration: a case study from Hawaii. Conservation Letters 1:236-243. Gould, S. F., N. J. Beeton, R. Harris, M. F. Hutchinson, A. M. Lechner, L. L. Porfirio, and B. G. Mackey. 2014. A tool for simulating and communicating uncertainty when modelling species distributions under future climates. Ecology and Evolution. Grand, J., M. P. Cummings, T. G. Rebelo, T. H. Ricketts, and M. C. Neel. 2007. Biased data reduce efficiency and effectiveness of conservation reserve networks. Ecology Letters 10:364-374. Grantham, H. S., A. Moilanen, K. A. Wilson, R. L. Pressey, T. G. Rebelo, and H. P. Possingham. 2008. Diminishing return on investment for biodiversity data in conservation planning. Conservation Letters 1:190-198. Grantham, H. S., K. A. Wilson, A. Moilanen, T. Rebelo, and H. P. Possingham. 2009. Delaying conservation actions for improved knowledge: how long should we wait? Ecology Letters 4:293-301. Green, R. E., S. J. Cornell, J. P. Scharlemann, and A. Balmford. 2005. Farming and the fate of wild nature. Science 307:550-555. Gregory, R., D. Ohlson, and J. Arvai. 2006. Deconstructing adaptive management: criteria for applications to environmental management. Ecological Applications 16:2411-2425. Guikema, S., and M. Milke. 1999. Quantitative decision tools for conservation programme planning: practice, theory and potential. Environmental Conservation 26:179-189. Guisan, A., and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993-1009.

200

Guisan, A., R. Tingley, J. B. Baumgartner, I. Naujokaitis‐Lewis, P. R. Sutcliffe, A. I. Tulloch, T. J. Regan, L. Brotons, E. McDonald‐Madden, C. Mantyka‐Pringle, T. G. Martin, J. R. Rhodes, R. Maggini, S. A. Setterfield, Jane Elith, M. W. Schwartz, B. A. Wintle, O. Broennimann, M. Austin, S. B. Ferrier, M. R. Kearney, H. P. Possingham, and Y. M. Buckley. 2013. Predicting species distributions for conservation decisions. Ecology Letters 16:1424-1435. Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147-186. Hagen-Zanker, A., J. Hurkens, and R. Vanhout 2010. Map Comparison Kit 3 User Manual. – Research Institute for Knowledge Systems, http://www.riks.nl/mck/index.php?page=downloads. Haight, R. G., B. Cypher, P. A. Kelly, S. Phillips, H. P. Possingham, K. Ralls, A. M. Starfield, P. J. White, and D. Williams. 2002. Optimizing habitat protection using demographic models of population viability. Conservation Biology 16:1386-1397. Haight, R. G., and S. A. Snyder. 2009. Integer programming methods for reserve selection and design.in A. Moilanen, K. A. Wilson, and H. Possingham, editors. Spatial conservation prioritization: quantitative methods and computational tools. Oxford University Press, New York. Haight, R. G., and L. E. Travis. 2008. Reserve design to maximize species persistence. Environmental modeling & assessment 13:243-253. Halpern, B., S. Walbridge, K. A. Selkoe, C. V. Kappel, M. Fiorenza, C. D'Agrosa, J. F. Bruno, K. S. Casey, C. Ebert, H. E. Fox, R. Fujita, D. Heinemann, H. S. Lenihan, E. M. P. Madin, M. T. Perry, E. R. Selig, M. Spalding, R. Seneck, and R. Watson. 2008. A global map of human impact on marine ecosystems. Science 319:948-952. Hanley, N., J. Shogren, and B. White. 2007. Environmental economics in theory and practice. Palgrave, London. Hawkins, B. A., M. Rueda, and M. Á. Rodriguez. 2008. What do range maps and surveys tell us about diversity patterns? Folia Geobotanica 43:355. Heller, N. E., and E. S. Zavaleta. 2009. Biodiversity management in the face of climate change: A review of 22 years of recommendations. Biological Conservation 142:14-32. Hernandez, P. A., C. H. Graham, L. L. Master, and D. L. Albert. 2006. The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29:773-785.

201

Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965-1978. Hobbs, R. J. 2001. Synergisms among habitat fragmentation, livestock grazing, and biotic invasions in southwestern Australia. Conservation Biology 15:1522-1528. Hobbs, R. J., and V. A. Cramer. 2008. Restoration ecology: interventionist approaches for restoring and maintaining ecosystem function in the face of rapid environmental change. Annual Review of Environment and Resources 33:39-61. Hoekstra, J. M., T. M. Boucher, T. H. Ricketts, and C. Roberts. 2005. Confronting a biome crisis: global disparities of habitat loss and protection. Ecology Letters 8:23-29. Hof, C., M. B. Araújo, W. Jetz, and C. Rahbek. 2011. Additive threats from pathogens, climate and land-use change for global amphibian diversity. Nature 480:516-521. Hoffmann, M., C. Hilton-Taylor, A. Angulo, M. Böhm, T. M. Brooks, S. H. M. Butchart, K. E. Carpenter, J. Chanson, B. Collen, N. A. Cox, W. R. T. Darwall, N. K. Dulvy, L. R. Harrison, V. Katariya, C. M. Pollock, S. Quader, N. I. Richman, A. S. L. Rodrigues, M. F. Tognelli, J.-C. Vié, J. M. Aguiar, D. J. Allen, G. R. Allen, G. Amori, N. B. Ananjeva, F. Andreone, P. Andrew, A. L. A. Ortiz, J. E. M. Baillie, R. Baldi, B. D. Bell, S. D. Biju, J. P. Bird, P. Black-Decima, J. J. Blanc, F. Bolaños, W. Bolivar-G., I. J. Burfield, J. A. Burton, D. R. Capper, F. Castro, G. Catullo, R. D. Cavanagh, A. Channing, N. L. Chao, A. M. Chenery, F. Chiozza, V. Clausnitzer, N. J. Collar, L. C. Collett, B. B. Collette, C. F. C. Fernandez, M. T. Craig, M. J. Crosby, N. Cumberlidge, A. Cuttelod, A. E. Derocher, A. C. Diesmos, J. S. Donaldson, J. W. Duckworth, G. Dutson, S. K. Dutta, R. H. Emslie, A. Farjon, S. Fowler, J. Freyhof, D. L. Garshelis, J. Gerlach, D. J. Gower, T. D. Grant, G. A. Hammerson, R. B. Harris, L. R. Heaney, S. B. Hedges, J.-M. Hero, B. Hughes, S. A. Hussain, J. I. M., R. F. Inger, N. Ishii, D. T. Iskandar, R. K. B. Jenkins, Y. Kaneko, M. Kottelat, K. M. Kovacs, S. L. Kuzmin, E. L. Marca, J. F. Lamoreux, M. W. N. Lau, E. O. Lavilla, K. Leus, R. L. Lewison, G. Lichtenstein, S. R. Livingstone, V. Lukoschek, D. P. Mallon, P. J. K. McGowan, A. McIvor, P. D. Moehlman, S. Molur, A. M. Alonso, J. A. Musick, K. Nowell, R. A. Nussbaum, W. Olech, N. L. Orlov, T. J. Papenfuss, G. Parra-Olea, W. F. Perrin, B. A. Polidoro, M. Pourkazemi, P. A. Racey, J. S. Ragle, M. Ram, G. Rathbun, R. P. Reynolds, A. G. J. Rhodin, S. J. Richards, L. O. Rodríguez, S. R. Ron, C. Rondinini, A. B. Rylands, Y. S. d. Mitcheson, J. C. Sanciangco, K. L. Sanders, G. Santos-Barrera, J. Schipper, C. Self-Sullivan, Y. Shi, A. Shoemaker, F. T. Short, C. Sillero-Zubiri, D. L. Silvano, K. G. Smith, A. T. Smith, J. Snoeks, A. J. Stattersfield, A. J. Symes, A. B. Taber,

202

B. K. Talukdar, H. J. Temple, R. Timmins, J. A. Tobias, K. Tsytsulina, D. Tweddle, C. Ubeda, S. V. Valenti, P. P. v. Dijk, L. M. Veiga, A. Veloso, D. C. Wege, M. Wilkinson, E. A. Williamson, F. Xie, B. E. Young, H. R. Akçakaya, L. Bennun, T. M. Blackburn, L. Boitani, H. T. Dublin, G. A. B. d. Fonseca, C. Gascon, T. E. L. Jr., G. M. Mace, S. A. Mainka, J. A. McNeely, R. A. Mittermeier, G. M. Reid, J. P. Rodriguez, A. A. Rosenberg, M. J. Samways, J. Smart, B. A. Stein, and S. N. Stuart. 2010. The impact of conservation on the status of the world's vertebrates. Science 330:1503-1509. Holling, C. S. 1978. Adaptive environmental assessment and management. John Wiley, New York, New York. Holzkämper, A., and R. Seppelt. 2007. Evaluating cost-effectiveness of conservation management actions in an agricultural landscape on a regional scale. Biological Conservation 136:117- 127. Hughey, K. F. D., R. Cullen, and E. Moran. 2003. Integrating economics into priority setting and evaluation in conservation management. Conservation Biology 17:93-103. Hurlbert, A. H., and W. Jetz. 2007. Species richness, hotspots, and the scale dependence of range maps in ecology and conservation. Proceedings of the National Academy of Sciences of the United States of America 104:13384-13389. International Union for Conservation of Nature (IUCN). 2013. The IUCN Red List of Threatened Species. Version 2013.2. International Union for Conservation of Nature (IUCN) 2014. Threats Classification Scheme (Version 3.2). – IUCN, Gland Switzerland, http://www.iucnredlist.org/technical- documents/classification-schemes/threats-classification-scheme Accessed 12 June 2014. International Union for Conservation of Nature (IUCN), World Wildlife Fund (WWF). 1997. Centres of plant diversity, a guide and strategy for their conservation. The Americas. IUCN Publication Unit Cambridge, Cambridge. Isaac, N. J. B., S. T. Turvey, B. Collen, C. Waterman, and J. E. M. Baillie. 2007. Mammals on the EDGE: conservation priorities based on threat and phylogeny. PLoS ONE 2:e296. Isbell, R. 2002. The Australian Soil Classification. CSIRO, Canberra. James, A., K. J. Gaston, and A. Balmford. 2001. Can we afford to conserve biodiversity. Bioscience 51:43-52. James, A. N., K. J. Gaston, and A. Balmford. 1999. Balancing the Earth's accounts. Nature 401:323-324.

203

Jetz, W., C. Sekercioglu, and J. E. M. Watson. 2008. Ecological correlates and conservation implications of overestimating species' geographic ranges. Conservation Biology 22:110- 119. Johnson, L. T., and C. Hope. 2012. The social cost of carbon in US regulatory impact analyses: an introduction and critique. Journal of Environmental Studies and Sciences 2:205-221. Joseph, L. N. 2007. Theory to guide the management of species for conservation. PhD Thesis. University of Queensland, Brisbane. Joseph, L. N., R. F. Maloney, and H. P. Possingham. 2009. Optimal allocation of resources among threatened species: a project prioritization protocol. Conservation Biology 23:328-338. Just, R., D. Hueth, and A. Schmitz. 1982. Applied welfare economics and public policy. Prentice Hall, NJ. Justus, J., and S. Sarkar. 2002. The principle of complementarity in the design of reserve networks to conserve biodiversity, a preliminary history. Journal of Biosciences 27. Keith, D. A. 2009. The interpretation, assessment and conservation of ecological communities. Ecological Management & Restoration 10:S3-S15. Keith, D. A. 2011. Functional traits: their roles in understanding and predicting biotic responses to fire regimes from individuals to landscapes. Pages 97-125 in R. A. Bradstock, A. M. Gill, and R. J. Williams, editors. Flammable Australia: Fire regimes, biodiversity and ecosystems in a changing world. CSIRO, Melbourne. Keith, D. A., T. G. Martin, E. McDonald-Madden, and C. Walters. 2011. Uncertainty and adaptive management for biodiversity conservation. Biological Conservation 144:1175-1178. Kilbourne, W., and G. Pickett. 2008. How materialism affects environmental beliefs, concern, and environmentally responsible behavior. Journal of Business Research 61:885-893. Kirkpatrick, J. B. 1983. An iterative method for establishing priorities for the selection of nature reserves: an example from Tasmania. Biological Conservation 25:127-134. Klein, C. J., N. C. Ban, B. S. Halpern, M. Beger, E. T. Game, H. S. Grantham, A. Green, T. J. Klein, S. Kininmonth, E. Treml, K. Wilson, and H. P. Possingham 2010. Prioritizing land and sea conservation investments to protect coral reefs. PLoS ONE 5(8):e12431. doi:12410.11371/journal.pone.0012431. e12431. doi:10.1371/journal.pone.0012431. Klein, C. J., C. Steinback, M. Watts, A. J. Scholz, and H. P. Possingham. 2009. Spatial marine zoning for fisheries and conservation. Frontiers in Ecology and the Environment 8:349-353. Knight, A. T., R. M. Cowling, M. Difford, and B. M. Campbell. 2010. Mapping human and social dimensions of conservation opportunity for the scheduling of conservation action on private land. Conservation Biology Series (Cambridge) 24:1348-1358.

204

Knight, A. T., R. M. Cowling, M. Rouget, A. Balmford, A. T. Lombard, and B. M. Campbell. 2008. Knowing but not doing: Selecting priority conservation areas and the research- implementation gap. Conservation Biology 22:610-617. Knight, A. T., A. Driver, R. M. Cowling, K. Maze, P. G. Desmet, A. T. Lombard, M. Rouget, M. A. Botha, A. F. Boshoff, and J. Castley. 2006. Designing systematic conservation assessments that promote effective implementation: best practice from South Africa. Conservation Biology 20:739-750. Kovacs, K., S. Polasky, E. Nelson, B. L. Keeler, D. Pennington, A. J. Plantinga, and S. J. Taff. 2013. Evaluating the Return in Ecosystem Services from Investment in Public Land Acquisitions. PLoS ONE 8:e62202. Kramer, D. B., T. Zhang, K. S. Cheruvelil, A. Ligmann-Zielinska, and P. A. Soranno. 2013. A multi-objective, return on investment analysis for freshwater conservation planning. Ecosystems:1-15. Kremen, C., A. Cameron, A. Moilanen, S. J. Phillips, C. D. Thomas, H. Beentje, J. Dransfield, B. L. Fisher, F. Glaw, T. C. Good, G. J. Harper, R. J. Hijmans, D. C. Lees, E. Louis Jr., R. A. Nussbaum, C. J. Raxworthy, A. Razafimpahanana, G. E. Schatz, M. Vences, D. R. Vieites, P. C. Wright, and M. L. Zjhra. 2008. Aligning conservation priorities across taxa in Madagascar with high-resolution planning tools. Science 320:222-225. Kriticos, D. J., and A. Leriche. 2010. The effects of climate data precision on fitting and projecting species niche models. Ecography 33:115-127. Krupnik, I., and D. Jolly. 2002. The Earth Is Faster Now: Indigenous Observations of Arctic Environmental Change. Frontiers in Polar Social Science. Arctic Research Consortium of the United States, 3535 College Road, Suite 101, Fairbanks, AK 99709. La Sorte, F. A., and B. A. Hawkins. 2007. Range maps and species richness patterns: errors of commission and estimates of uncertainty. Ecography 30:649-662. Lambeck, R. J. 1997. Focal species: a multi-species umbrella for nature conservation. Conservation Biology 11:849-856. Lamoreux, J. F., J. C. Morrison, T. H. Ricketts, D. M. Olson, E. Dinerstein, M. W. McKnight, and H. H. Shugart. 2005. Global tests of biodiversity concordance and the importance of endemism. Nature 440:212-214. Larsen, F. W., and C. Rahbek. 2003. Influence of scale on conservation priority setting - a test on African mammals. Biodiversity and Conservation 12:599-614.

205

Laycock, H., D. Moran, J. C. R. Smart, D. G. Raffaelli, and P. C. L. White. 2011. Evaluating the effectiveness and efficiency of biodiversity conservation spending. Ecological Economics 70:1789-1796. Leader-Williams, N., and H. T. Dublin. 2000. Charismatic megafauna as 'flagship species'. Pages 53-81 in A. Entwistle and N. Dunstone, editors. Priorities for the Conservation of Mammalian diversity: has the Panda had its Day? Cambridge University Press, Cambridge. Legge, S., M. S. Kennedy, R. Lloyd, S. A. Murphy, and A. Fisher. 2011. Rapid recovery of mammal fauna in the central Kimberley, northern Australia, following the removal of introduced herbivores. Austral Ecology 36:791-799. Lennox, G. D., and P. R. Armsworth. 2013. The ability of landowners and their cooperatives to leverage payments greater than opportunity costs from conservation contracts. Conservation Biology 27:625-634. Lennox, G. D., K. J. Gaston, S. Acs, M. Dallimer, N. Hanley, and P. R. Armsworth. 2013. Conservation when landowners have bargaining power: Continuous conservation investments and cost uncertainty. Ecological Economics 93:69-78. Levin, N., J. E. Watson, L. N. Joseph, H. S. Grantham, L. Hadar, N. Apel, A. Perevolotsky, N. DeMalach, H. P. Possingham, and S. Kark. 2013. A framework for systematic conservation planning and management of Mediterranean landscapes. Biological Conservation 158:371- 383. Levins, R. 1966. The strategy of model building in population biology. American Scientist 54:421- 431. Lewis, D. J., A. J. Plantinga, E. Nelson, and S. Polasky. 2011. The efficiency of voluntary incentive policies for preventing biodiversity loss. Resource and Energy Economics 33:192-211. Lindenmayer, D., and M. Burgman. 2005. Practical conservation biology. CSIRO publishing. Livestock Health and Pest Authority. 2012. Fox baiting program winter 2012. NSW, AU. Loiselle, B. A., C. A. Howell, C. H. Graham, J. M. Goerck, T. M. Brooks, K. G. Smith, and P. H. Williams. 2003. Avoiding pitfalls of using species distribution models in conservation planning. Conservation Biology 17:1591-1600. Lombard, A. T., R. M. Cowling, R. L. Pressey, and A. G. Rebelo. 2003. Effectiveness of land classes as surrogates for species in conservation planning for the Cape Floristic Region. Biological Conservation 112:45-62. Lowe, S., M. Browne, S. Boudjelas, and M. De Poorter. 2001. 100 of the world's worst invasive alien species: a selection from the global invasive species database. Species Survival Commission, World Conservation Union, Auckland, NZ.

206

Mace, G. M., K. Norris, and A. H. Fitter. 2012. Biodiversity and ecosystem services: a multilayered relationship. Trends in Ecology & Evolution 27:19-26. Mace, G. M., H. P. Possingham, and N. Leader-Williams. 2006. Prioritizing choices in conservation. Pages 17-34 in D. MacDonald, editor. Key Topics in Conservation Biology. Oxford University Press, UK. Mace, G. M., H. P. Possingham, and N. Leader-Williams. 2007. Prioritizing choices in conservation. Pages 17-34 in D. W. Macdonald and K. Service, editors. Key Topics in Conservation Biology. Blackwell Publishing, Oxford. Mack, R. N., D. Simberloff, W. M. Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic Invasions: Causes, Epidemiology, Global Consequences, and Control. Ecological Applications 10:689-710. Makino, A., C. J. Klein, M. Beger, S. D. Jupiter, and H. P. Possingham. 2013. Incorporating Conservation Zone Effectiveness for Protecting Biodiversity in Marine Planning. PLoS ONE 8:e78986. Mantyka-Pringle, C. S., T. G. Martin, and J. R. Rhodes. 2012. Interactions between climate and habitat loss effects on biodiversity: a systematic review and meta‐analysis. Global Change Biology 18:1239-1252. Margules, C., and S. Sarkar. 2007. Systematic conservation planning. Cambridge University Press. Margules, C. R., A. J. Higgs, and R. W. Rafe. 1982. Modern biogeographic theory: are there any lessons for nature reserve design? Biological Conservation 24:115-128. Margules, C. R., A. O. Nicholls, and R. L. Pressey. 1988. Selecting networks of reserves to maximise biological diversity. Biological Conservation 43:63-76. Margules, C. R., and R. L. Pressey. 2000. Systematic conservation planning. Nature 405:243-253. Marinoni, O., J. Navarro-Garcia, S. Marvanek, D. Prestwidge, D. Clifford, and L. A. Laredo. 2012. Development of a system to produce maps of agricultural profit on a continental scale: An example for Australia. Agricultural Systems 105:33-45. Marris, E. 2007. Conservation priorities: what to let go. Nature 450:152-155. Marris, E. 2014. ‘New conservation’is an expansion of approaches, not an ethical orientation. Animal Conservation. Marsh, H., A. Dennis, H. Hines, A. Kutt, K. McDonald, E. Weber, S. Williams, and J. Winter. 2007. Optimizing allocation of management resources for wildlife. Conservation Biology 21:387-399.

207

Martin, T. G., S. Nally, A. A. Burbidge, S. Arnall, S. T. Garnett, M. W. Hayward, L. F. Lumsden, P. Menkhorst, E. McDonald‐Madden, and H. P. Possingham. 2012. Acting fast helps avoid extinction. Conservation Letters 5:274-280. Mason, N., and S. Olsen. 2000. Sydney-north regional fox control programme: Conserving Sydney’s green web. Pages 70–75 in New South Wales Pest Animal Control Conference: Practical Solutions to Pest Management Problems. New South Wales Agriculture, Orange, NSW. McBride, M. F., K. A. Wilson, M. Bode, and H. P. Possingham. 2007. Incorporating the effects of socioeconomic uncertainty into priority setting for conservation investment. Conservation Biology 21:1463-1474. McCarthy, D. P., P. F. Donald, J. P. W. Scharlemann, G. M. Buchanan, A. Balmford, J. M. H. Green, L. A. Bennun, N. D. Burgess, L. D. C. Fishpool, S. T. Garnett, D. L. Leonard, R. F. Maloney, P. Morling, H. M. Schaefer, A. Symes, D. A. Wiedenfeld, and S. H. M. Butchart. 2012. Financial costs of meeting global biodiversity conservation targets: Current spending and unmet needs. Science 338:946-949. McCarthy, M. A., and H. P. Possingham. 2007. Active adaptive management for conservation. Conservation Biology 21:956-963. McDonald‐Madden, E., P. W. Baxter, and H. P. Possingham. 2008. Subpopulation triage: how to allocate conservation effort among populations. Conservation Biology 22:656-665. McIntyre, S., K. Heard, and T. G. Martin. 2003. The relative importance of cattle grazing in subtropical grasslands: does it reduce or enhance plant biodiversity? Journal of Applied Ecology 40:445-457. McLeod, L. J., G. R. Saunders, S. R. McLeod, M. Dawson, and R. van de Ven. 2010. The potential for participatory landscape management to reduce the impact of the red fox (Vulpes vulpes) on lamb production. Wildlife Research 37:715-721. Metrick, A., and M. L. Weitzman. 1998. Conflicts and choices in biodiversity preservation. The Journal of Economic Perspectives 12:21-33. Millennium Ecosystem Assessment. 2005. Ecosystems and human well-being: Biodiversity synthesis. World Resources Institute, Washington, DC. Mittermeier, R., C. G. Mittermeier, T. M. Brooks, J. D. Pilgrim, W. R. Konstant, G. Da Fonseca, and C. Kormos. 2003. Wilderness and biodiversity conservation. Proceedings of the National Academy of Sciences 100:10309-10313.

208

Mittermeier, R. A., and C. G. Mittermeier. 1997. Megadiversity: earth's biologically wealthiest nations. Megadiversidad: los países biológicamente más ricos del mundo. CEMEX Conservation International and Agrupacion Sierra Madre, Monterrey, México. Moilanen, A. 2007. Landscape zonation, benefit functions and target-based planning: Unifying reserve selection strategies. Biological Conservation 134:571-579. Moilanen, A., B. J. Anderson, F. Eigenbrod, A. Heinemeyer, D. B. Roy, S. Gillings, P. R. Armsworth, K. J. Gaston, and C. D. Thomas. 2011a. Balancing alternative land uses in conservation prioritization. Ecological Applications 21:1419-1426. Moilanen, A., and I. R. Ball. 2009. Heuristic and approximate optimization methods for spatial conservation prioritization. Pages 58-69 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Prioritization. Oxford University Press, Oxford. Moilanen, A., and M. Cabeza. 2002. Single-species dynamic site selection. Ecological Applications 12:913-926. Moilanen, A., H. Kujala, and J. R. Leathwick. 2009a. The zonation framework and software for conservation prioritization. Pages 196-210 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Prioritization: Quantitative Methods & Computational Tools. Oxford University Press, Oxford. Moilanen, A., J. R. Leathwick, and J. M. Quinn. 2011b. Spatial prioritization of conservation management. Conservation Letters 4:383-393. Moilanen, A., H. Possingham, and K. A. Wilson. 2009b. Spatial conservation prioritization: Past, present, and future. Pages 260-267 in A. Moilanen, K. A. Wilson, and H. Possingham, editors. Spatial Conservation Prioritization. Oxford University Press, New York. Moilanen, A., H. P. Possingham, and S. Polasky. 2009c. A mathematical classification of conservation prioritization problems. Pages 16-27 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Prioritization: Quantitative Methods & Computational Tools. Oxford University Press, New York. Moilanen, A., M. C. Runge, J. Elith, A. Tyre, and Y. Carmel. 2006a. Planning for robust reserve networks using uncertainty analysis. Ecological Modelling 199:115-124. Moilanen, A., K. A. Wilson, and H. P. Possingham, editors. 2009d. Spatial Conservation Prioritization: Quantitative Methods & Computational Tools. Oxford University Press, New York. Moilanen, A., and B. A. Wintle. 2006. Uncertainty analysis favours selection of spatially aggregated reserve networks. Biological Conservation 129:427-434.

209

Moilanen, A., B. A. Wintle, J. Elith, and M. A. Burgman. 2006b. Uncertainty analysis for regional- scale reserve selection. Conservation Biology 20:1688-1697. Monserud, R. A., and R. Leemans. 1992. Comparing global vegetation maps with the Kappa statistic. Ecological Modelling 62:275-293. Moran, D., H. Laycock, and P. C. L. White. 2010. The role of cost-effectiveness analysis in conservation decision-making. Biological Conservation 143:826-827. Mouysset, L., L. Doyen, and J. Jiguet. 2013. How does economic risk aversion affect biodiversity. Ecological Applications 23:96-109. Murdoch, W. W., S. Polasky, K. A. Wilson, H. P. Possingham, P. Kareiva, and R. Shaw. 2007. Maximizing return on investment in conservation. Biological Conservation 139:375-388. Murdoch, W. W., J. Ranganathan, S. Polasky, and J. Regetz. 2010. Using return on investment to maximize conservation effectiveness in Argentine grasslands. Proceedings of the National Academy of Sciences 107:20855-20862. Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, and J. Kent. 2000. Biodiversity hotspots for conservation priorities. Nature 403:853-858. Naidoo, R., and W. L. Adamowicz. 2006. Modeling opportunity costs of conservation in transitional landscapes. Conservation Biology 20:490-500. Naidoo, R., A. Balmford, P. J. Ferraro, S. Polasky, T. H. Ricketts, and M. Rouget. 2006. Integrating economic costs into conservation planning. Trends in Ecology & Evolution 21:681-687. Naidoo, R., and T. Iwamura. 2007. Global-scale mapping of economic benefits from agricultural lands: Implications for conservation priorities. Biological Conservation 140:40-49. National Research Council. 2005. Valuing ecosystem services: Toward better environmental decision-making. National Academies Press, Washington, DC. Neke, K., and M. Du Plessis. 2004. The threat of transformation: Quantifying the vulnerability of grasslands in South Africa Conservation Biology 18:466-477. Neldner, V. J., B. A. Wilson, E. J. Thompson, and H. A. Dillewaard. 2005. Methodology for Survey and Mapping of Regional Ecosystems and Vegetation Communities in Queensland. Version 3.1. Queensland Herbarium, Environmental Protection Agency, Brisbane. Nelson, E., G. Mendoza, J. Regetz, S. Polasky, H. Tallis, D. Cameron, K. M. A. Chan, G. C. Daily, J. Goldstein, P. M. Kareiva, E. Lonsdorf, R. Naidoo, T. H. Ricketts, and M. Shaw. 2009. Modeling multiple ecosystem services, biodiversity conservation, commodity production, and tradeoffs at landscape scales. Frontiers in Ecology and the Environment 7:4-11. Nelson, E., S. Polasky, D. J. Lewis, A. J. Plantinga, E. Lonsdorf, D. White, D. Bael, and J. J. Lawler. 2008. Efficiency of incentives to jointly increase carbon sequestration and species

210

conservation on a landscape. Proceedings of the National Academy of Sciences 105:9471- 9476. Newburn, D., S. E. Reed, P. Berck, and A. M. Merenlender. 2005. Economics and land-use change in prioritizing private land conservation. Conservation Biology 19:1411-1420. Ng, C. F., H. Possingham, C. McAlpine, D. de Villiers, H. J. Preece, and J. R. Rhodes. 2014. Impediments to the success of management actions for species' recovery. PLoS ONE 9:e92430. Nicholson, E., and H. Possingham. 2006. Objectives for multiple-species conservation planning. Conservation Biology 20:871-881. Noss, R. F. 1987. From plant communities to landscapes in conservation inventories: A look at The Nature Conservancy (USA). Biological Conservation 41:11-37. Noss, R. F., and A. R. Cooperrider. 1994. Saving Nature's Legacy: protecting and restoring biodiversity. Island Press. Noss, R. F., A. P. Dobson, R. Baldwin, P. Beier, C. R. Davis, D. A. Dellasala, J. Francis, H. Locke, K. Nowak, R. Lopez, C. Reining, S. C. Trombulak, and G. Tabor. 2012. Bolder thinking for conservation. Conservation Biology 26:1-4. Oliveira, P. J., G. P. Asner, D. E. Knapp, A. Almeyda, R. Galván-Gildemeister, S. Keene, R. F. Raybin, and R. C. Smith. 2007. Land-use allocation protects the Peruvian Amazon. Science 317:1233-1236. Olsen, P. 2007a. Glimpses of Paradise The Quest for the Beautiful Parrakeet. National Library of Australia, Canberra. Olsen, P. 2007b. Glimpses of paradise: the quest for the beautiful parrakeet. National Library of Australia, Canberra. Olsen, S., N. Mason, and R. Ali. 2005. The Sydney-north regional fox control programme conserving Sydney’s green web. Pages 93–95 in 3rd NSW Pest Animal Control Conference. NSW Department of Primary Industries, Coffs Harbour, NSW. Olson, D. M., and E. Dinerstein. 1998. The Global 200: a representation approach to conserving the Earth’s most biologically valuable ecoregions. Conservation Biology 12:502-515. Orme, C. D. L., R. G. Davies, M. Burgess, F. Eigenbrod, N. Pickup, V. A. Olson, A. J. Webster, T.- S. Ding, P. C. Rasmussen, R. S. Ridgely, A. J. Stattersﬁeld, P. M. Bennett, T. M. Blackburn, K. J. Gaston, and I. P. F. Owens. 2005. Global hotspots of species richness are not congruent with endemism or threat. Nature 436:1016-1019. Orr, D. W. 1991. The Economics of Conservation. Conservation Biology 5:439-441.

211

Paine, R. T., M. J. Tegner, and E. A. Johnson. 1998. Compounded perturbations yield ecological surprises. Ecosystems 1:535-545. Pannell, D. J. 1997. Sensitivity analysis of normative economic models: Theoretical framework and practical strageties. Agricultural Economics 16:139-153. Pannell, D. J., G. R. Marshall, N. Barr, A. Curtis, F. Vanclay, and R. Wilkinson. 2006. Understanding and promoting adoption of conservation practices by rural landholders. Australian Journal of Experimental Agriculture and Animal Husbandry 46:1407-1424. Pannell, D. J., A. M. Roberts, G. Park, J. Alexander, A. Curatolo, and S. Marsh. 2012. Integrated assessment of public investment in land-use change to protect environmental assets in Australia. Land Use Policy 29:377-387. Pannell, D. J., A. M. Roberts, G. Park, A. Curatolo, and S. P. Marsh 2010. INFFER (Investment Framework for Environmental Resources): Practical and Theoretical Underpinnings http://cyllene.uwa.edu.au/~dpannell/dp1001.htm Accessed 7 November 2010. Paton, S., A. Curtis, G. McDonald, and M. Woods. 2004. Regional natural resource management: is it sustainable. Australasian Journal of Environmental Management 11:259-267. Pearce, J., and S. Ferrier. 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecological Modelling 133:225-245. Pearson, R. G. 2007. Species distribution modeling for conservation educators and practitioners. Synthesis. – American Museum of National History, http://ncep.amnh.org Accessed 1 October 2011. Pearson, R. G., C. J. Raxworthy, M. Nakamura, and A. T. Peterson. 2007. Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. Journal of Biogeography 34:102-117. Peters, R. L., and J. D. Darling. 1985. The greenhouse effect and nature reserves. Bioscience:707- 717. Peterson, A. T. 2011. Ecological niches and geographic distributions (MPB-49). Princeton University Press, Princeton. Phalan, B., M. Onial, A. Balmford, and R. E. Green. 2011. Reconciling food production and biodiversity conservation: land sharing and land sparing compared. Science 333:1289-1291. Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231-259. Phillips, S. J., and M. Dudík. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161-175.

212

Phillips, S. J., M. Dudík, C. H. Graham, A. Lehmann, J. R. Leathwick, and S. Ferrier. 2009. Sample selection bias and presence-only distribution models: implications for background and pseudo-absence data. Ecological Applications 19:181-197. Pimm, S. L., C. N. Jenkins, R. Abell, T. M. Brooks, J. L. Gittleman, L. N. Joppa, P. H. Raven, C. M. Roberts, and J. O. Sexton. 2014. The biodiversity of species and their rates of extinction, distribution, and protection. Science 344. Pimm, S. L., G. J. Russell, J. L. Gittleman, and T. M. Brooks. 1995. The future of biodiversity. Science 269:347-350. Pineda, E., and J. M. Lobo. 2012. The performance of range maps and species distribution models representing the geographic variation of species richness at different resolutions. Global Ecology and Biogeography 21:935-944. Pisinger, D., and P. Toth. 1998. Knapsack problems. Pages 299-428 in D.-Z. Du and P. M. Pardalos, editors. Handbook of combinatorial optimization. Kluwer Academic Publishers, Dordrecht, The Netherlands. Plumptre, A. J., R. A. Fuller, A. Rwetsiba, F. Wanyama, D. Kujirakwinja, M. Driciru, G. Nangendo, J. E. Watson, and H. P. Possingham. 2014. Efficiently targeting resources to deter illegal activities in protected areas. Journal of Applied Ecology 51:714-725. Polasky, S., J. D. Camm, and B. Garber-Yonts. 2001. Selecting biological reserves cost-effectively: An application to terrestrial vertebrate conservation in Oregon. Land Economics 77:68-78. Polasky, S., E. Nelson, J. Camm, B. Csuti, P. Fackler, E. Lonsdorf, C. A. Montgomery, D. White, J. L. Arthur, B. Garber-Yonts, R. G. Haight, J. Kagan, A. Starfield, and C. Tobalske. 2008. Where to put things? Spatial land management to sustain biodiversity and economic returns. Biological Conservation 141:1505-1524. Polasky, S., E. Nelson, E. Lonsdorf, P. Fackler, and A. Starfield. 2005. Conserving species in a working landscape: Land use with biological and economic objectives. Ecological Applications 15:13887-11401. Polasky, S., and K. Segerson. 2009. Integrating ecology and economics in the study of ecosystem services: some lessons learned. Annual Review of Resource Economics 1:409-434. Polasky, S., and A. R. Solow. 2001. The value of information in reserve site selection. Biodiversity and Conservation 10:1051-1058. Pontius Jr., R. G., and M. Millones. 2011. Death to Kappa: birth of quantity disagreement and allocation disagreement for accuracy assessment. International Journal of Remote Sensing 32:4407-4429.

213

Possingham, H., D. B. Lindenmayer, and M. A. McCarthy. 2001a. Population viability analysis. Pages 831-844 in S. Levin, editor. The Encyclopedia of Biodiversity. Academic Press, San Diego, CA. Possingham, H. P. 2001a. The business of biodiversity: Applying decision theory principles to nature conservation. TELA Series No. 9. The Australian Conservation Foundation, Melbourne, AU. Possingham, H. P. 2001b. Models, problems and algorithms: perceptions about their application to conservation biology? Pages 1-6 in MODSIM2001: International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand Inc, The Australian National University, Canberra, ACT. Possingham, H. P., S. J. Andelman, M. A. Burgman, R. A. Medellin, L. L. Master, and D. A. Keith. 2002. Limits to the use of threatened species lists. Trends in Ecology & Evolution 17:503- 507. Possingham, H. P., S. J. Andelman, B. R. Noon, S. Trombulak, and H. R. Pulliam. 2001b. Making smart conservation decisions. Pages 225-244 in M. E. Soulé and G. H. Orians, editors. Conservation Biology: Research Priorities for the Next Decade. Island Press, Washington, D.C. Possingham, H. P., I. R. Ball, and S. Andelman. 2000. Mathematical methods for identifying representative reserve networks. Pages 291-305 in S. Ferson and M. A. Burgman, editors. Quantitative methods for conservation biology. Spring-Verlag, New York. Possingham, H. P., and K. A. Wilson. 2005. Biodiversity--turning up the heat on hotspots. Nature 436:919-920. Pouzols, F. M., M. A. Burgman, and A. Moilanen. 2012. Methods for allocation of habitat management, maintenance, restoration and offsetting, when conservation actions have uncertain consequences. Biological Conservation 153:41-50. Pouzols, F. M., and A. Moilanen. 2013. RobOff: software for analysis of alternative land‐use options and conservation actions. Methods in Ecology and Evolution 4:426-432. Pressey, R. L. 1994. Ad hoc reservations: Forward or backward steps in developing representative reserve systems? Conservation Biology 8:662-668. Pressey, R. L. 2002. The first reserve selection algorithm--a retrospective on Jamie Kirkpatrick's 1983 paper. Progress in Physical Geography 26. Pressey, R. L. 2004. Conservation planning and biodiversity: assembling the best data for the jo. Conservation Biology 13.

214

Pressey, R. L., M. Cabeza, M. E. Watts, R. M. Cowling, and K. A. Wilson. 2007. Conservation planning in a changing world. Trends in Ecology & Evolution 22:583-592. Pressey, R. L., R. M. Cowling, and M. Rouget. 2003. Formulating conservation targets for biodiversity pattern and process in the Cape Floristic Region, South Africa. Biological Conservation 112:99-127. Pressey, R. L., C. J. Humphries, C. R. Margules, R. I. Vane-Wright, and P. H. Williams. 1993. Beyond opportunism: Key principles for systematic reserve selection. Trends in Ecology & Evolution 8:124-128. Pressey, R. L., M. E. Watts, and T. W. Barrett. 2004. Is maximizing protection the same as minimizing loss? Efficiency and retention as alternative measures of the effectiveness of proposed reserves. Ecology Letters 7:1035-1046. Queensland Government. 2012. Back on Track species prioritisation framework. Brisbane. Queensland Herbarium 2010a. Biodiversity status of pre-clearing and 2006b regional ecosystems SE Qld (ISO 19139). – Department of Environment and Resource Management QGIS (Queensland Government Information Service). http://dds.information.qld.gov.au/dds/ Accessed 28 March 2012. Queensland Herbarium 2010b. Vegetation Management Act High Value Regrowth Vegetation Version 2.0 (ISO 19139). – Department of Environment and Resource Management QGIS (Queensland Government Information Service). http://dds.information.qld.gov.au/dds/ Accessed 28 March 2012. Queensland Parks and Wildlife Service Fire Management System (QPWS) 2012. Fire regime groups. – Department of National Parks, Recreation, Sport and Racing QGIS (Queensland Government Information Service). http://dds.information.qld.gov.au/dds/ Accessed 14 May 2013. Raiffa, H. 1993. Decision analysis: introductory lectures on choices under uncertainty. 1968. MD computing: computers in medical practice 10:312. Rands, M. R. W., W. M. Adams, L. Bennun, S. H. M. Butchart, A. Clements, D. Coomes, A. Entwistle, I. Hodge, V. Kapos, J. P. W. Scharlemann, W. J. Sutherland, and B. Vira. 2010. Biodiversity conservation: Challenges beyond 2010. Science 329:1298-1303. Read, J. L., and R. Cunningham. 2010. Relative impacts of cattle grazing and feral animals on an Australian arid zone reptile and small mammal assemblage. Austral Ecology 35:314-324. Regan, H. M., M. Colyvan, and M. A. Burgman. 2002. A taxonomy and treatment of uncertainty for ecology and conservation biology. Ecological Applications 12:618-628.

215

Regan, H. M., D. A. Keith, T. J. Regan, M. G. Tozer, and N. Tootell. 2011. Fire management to combat disease: turning interactions between threats into conservation management. Oecologia 167:873-882. Rencher, A. C. 1995. Methods of Multivariate Analysis. Wiley, New York. Renner, I. W., and D. I. Warton. 2012. Equivalence of MAXENT and Poisson Point Process Models for species distribution modeling in ecology. Biometrics DOI: 10.1111/j.1541- 0420.2012.01824.x. Reserve Bank of Australia (RBA) 2012. Inflation Calculator. – RBA. http://www.rba.gov.au/calculator/annualDecimal.html Accessed 10 September 2012. Reside, A. E., J. VanDerWal, A. Kutt, I. Watson, and S. Williams. 2012. Fire regime shifts affect bird species distributions. Diversity and Distributions 18:213-225. Rittel, H. W., and M. M. Webber. 1973. Dilemmas in a general theory of planning. Policy sciences 4:155-169. Robins, L., and S. Dovers. 2007. NRM regions in Australia: the 'Haves' and the 'Have Nots'. Geographical Research 45:273-290. Robley, A., L. Woodford, M. Lindeman, and R. Sorre. 2013. The influence of fire, fox control and habitat structure on the presence of native mammals in East Gippsland Arthur Rylah Institute for Environmental Research, Victoria, Australia. Rockström, J., W. Steffen, K. Noone, Å. Persson, F. S. Chapin III, E. Lambin, T. M. Lenton, M. Scheffer, C. Folke, H. Schellnhuber, B. Nykvist, C. A. De Wit, T. Hughes, S. van der Leeuw, H. Rodhe, S. Sörlin, P. K. Snyder, R. Costanza, U. Svedin, M. Falkenmark, L. Karlberg, R. W. Corell, V. J. Fabry, J. Hansen, B. Walker, D. Liverman, K. Richardson, P. Crutzen, and J. Foley. 2009. Planetary boundaries: Exploring the safe operating space for humanity. Ecology and Society 14:32. Rödder, D., J. Kielgast, and S. Lötters. 2010. Future potential distribution of the emerging amphibian chytrid fungus under anthropogenic climate change. Diseases of Aquatic Organisms 92:201-207. Rodrigues, A. S., S. J. Andelman, M. I. Bakarr, L. Boitani, T. M. Brooks, R. M. Cowling, L. D. Fishpool, G. A. da Fonseca, K. J. Gaston, and M. Hoffmann. 2004a. Effectiveness of the global protected area network in representing species diversity. Nature 428:640-643. Rodrigues, A. S. L. 2006. Are global conservation efforts successful? Science 313:1051-1052. Rodrigues, A. S. L., H. R. Akçakaya, S. J. Andelman, M. I. Bakarr, L. boitani, T. M. Brooks, J. S. Chanson, L. D. C. Fishpool, G. A. B. D. Fonseca, K. J. Gaston, M. Hoffmann, P. A. Marquet, J. D. Pilgrim, R. L. Pressey, J. Schipper, W. Sechrest, S. N. Stuart, L. G.

216

Underhill, R. W. Waller, M. E. J. Watts, and X. Yan. 2004b. Global gap analysis: Priority regions for expanding the global protected-area network. Bioscience 54:1092-1100. Rondinini, C., M. D. Marco, F. Chiozza, G. Santulli, D. Baisero, P. Visconti, M. Hoffmann, J. Schipper, S. N. Stuart, M. F. Tognelli, G. Amori, A. Falcucci, L. Maiorano, and L. Boitani. 2011. Global habitat suitability models of terrestrial mammals. Philosophical Transactions of the Royal Society of London B Biological Sciences 36:2633-2641. Rondinini, C., K. A. Wilson, L. Boitani, H. Grantham, and H. P. Possingham. 2006. Tradeoffs of different types of species occurrence data for use in systematic conservation planning. Ecology Letters 9:1136-1145. Russell, B. G., B. Smith, and M. L. Augee. 2003. Changes to a population of common ringtail possums after bushfire. Wildlife Research 30:389-396. Sala, O. E., F. S. C. III, J. J. Armesto, E. Berlow, J. Bloomfield, R. Dirzo, E. Huber-Sanwald, L. F. Huenneke, R. B. Jackson, A. KinZig, R. Leemans, D. M. Lodge, H. A. Mooney, M. Oesterheld, N. L. Poff, M. T. Sykes, B. H. Walker, M. Walker, and D. H. Wall. 2000. Global Biodiversity Scenarios for the Year 2100. Science 287:1770-1774. Salafsky, N., D. Salzer, A. J. Stattersfield, C. Hilton-Taylor, R. Neugarten, S. H. M. Butchart, B. Collen, N. Cox, L. L. Master, and S. O'Connor. 2008. A standard lexicon for biodiversity conservation: unified classifications of threats and actions. Conservation Biology 22:897- 911. Sanderson, E. W., M. Jaiteh, M. A. Levy, K. H. Redford, A. V. Wannebo, and G. Woolmer. 2002. The human footprint and the last of the wild. Bioscience 52:891-904. Saunders, G., and L. McLeod. 2007. Improving fox management strategies in Australia. Bureau of Rural Sciences, Canberra. Saunders, G. R., M. N. Gentle, and C. R. Dickman. 2010. The impacts and management of foxes Vulpes vulpes in Australia. Mammal Review 40:181-211. Schipper, J., J. S. Chanson, F. Chiozza, N. A. Cox, M. Hoffmann, V. Katariya, J. Lamoreux, A. S. L. Rodrigues, S. N. Stuart, H. J. Temple, J. Baillie, L. Boitani, T. E. L. Jr., R. A. Mittermeier, A. T. Smith, D. Absolon, J. M. Aguiar, G. Amori, N. Bakkour, R. Baldi, R. J. Berridge, J. Bielby, P. A. Black, J. J. Blanc, T. M. Brooks, J. A. Burton, T. M. Butynski, G. Catullo, R. Chapman, Z. Cokeliss, B. Collen, J. Conroy, J. G. Cooke, G. A. B. d. Fonseca, A. E. Derocher, H. T. Dublin, J. W. Duckworth, L. Emmons, R. H. Emslie, M. Festa- Bianchet, M. Foster, S. Foster, D. L. Garshelis, C. Gates, M. Gimenez-Dixon, S. Gonzalez, J. F. Gonzalez-Maya, T. C. Good, G. Hammerson, P. S. Hammond, D. Happold, M. Happold, J. Hare, R. B. Harris, C. E. Hawkins, M. Haywood, L. R. Heaney, S. Hedges, K.

217

M. Helgen, C. Hilton-Taylor, S. A. Hussain, N. Ishii, T. A. Jefferson, R. K. B. Jenkins, C. H. Johnston, M. Keith, J. Kingdon, D. H. Knox, K. M. Kovacs, P. Langhammer, K. Leus, R. Lewison, G. Lichtenstein, L. F. Lowry, Z. Macavoy, G. M. Mace, D. P. Mallon, M. Masi, M. W. McKnight, R. A. Medellín, P. Medici, G. Mills, P. D. Moehlman, S. Molur, A. Mora, K. Nowell, J. F. Oates, W. Olech, W. R. L. Oliver, M. Oprea, B. D. Patterson, W. F. Perrin, B. A. Polidoro, C. Pollock, A. Powel, Y. Protas, P. Racey, J. Ragle, P. Ramani, G. Rathbun, R. R. Reeves, S. B. Reilly, J. E. R. III, C. Rondinini, R. G. Rosell-Ambal, M. Rulli, A. B. Rylands, S. Savini, C. J. Schank, W. Sechrest, C. Self-Sullivan, A. Shoemaker, C. Sillero- Zubiri, N. D. Silva, D. E. Smith, C. Srinivasulu, P. J. Stephenso, N. v. Strien, B. K. Talukdar, B. L. Taylor, R. Timmins, D. G. Tirira, M. F. Tognelli, K. Tsytsulina, L. M. Veiga, J.-C. Vié, E. A. Williamson, S. A. Wyatt, Y. Xie, and B. E. Young. 2008. The status of the world's land and marine mammals: diversity, threat, and knowledge. Science 322:225- 230. Selkoe, K. A., B. S. Halpern, C. M. Ebert, E. C. Franklin, E. R. Selig, K. S. Casey, J. Bruno, and R. J. Toonen. 2009. A map of human impacts to a ‘‘pristine’’ coral reef ecosystem, the Papahānaumokuākea Marine National Monument. Coral Reefs 28:635-650. Shea, K., P. Amarasekare, P. Kareiva, M. Mangel, J. Moore, W. W. Murdoch, E. Noonburg, A. M. Parma, M. A. Pascual, H. P. Possingham, C. Wilcox, and D. Yu. 1998. Management of populations in conservation, harvesting and control. Trends in Ecology & Evolution 13:371- 375. Shriner, S. A., K. R. Wilson, and C. H. Flather. 2006. Reserve networks based on richness hotspots and representation vary with scale. Ecological Applications 16:1660-1673. Simberloff, D. 1998. Flagships, umbrellas, and keystones: Is single-species management passé in the landscape era? Biological Conservation 83:247-257. Sitas, N., J. Baillie, and N. Isaac. 2009. What are we saving? Developing a standardized approach for conservation action. Animal Conservation 12:231-237. Soulé, M. 2014. The "New Conservation".in G. Weuerthner, E. Crist, and T. Butler, editors. Keeping the Wild: Against the Domestication of the Earth. Island Press, San Francisco. Soulé, M. E. 1985. What is conservation biology? Bioscience 35:727-734. Steffen, W., P. J. Crutzen, and J. R. McNeill. 2007. The anthropocene: are humans now overwhelming the great forces of nature? Ambio 36:614-621. Steffen, W., Å. Persson, L. Deutsch, J. Zalasiewicz, M. Williams, K. Richardson, C. Crumley, P. Crutzen, C. Folke, L. Gordon, M. Molina, V. Ramanathan, J. Rockström, M. Scheffer, H. J.

218

Schellnhuber, and U. Svedin. 2011. The anthropocene: From global change to planetary stewardship. Ambio 40:739-761. Stein, J. L. 2007. A continental landscape framework for systematic conservation planning for Australian rivers and streams. PhD dissertation. Australian National University, Canberra, Australia. Stoms, D. M. 1994. Scale dependence of species richness maps. Professional Geographer 45:346- 358. Strahler, A. N. 1957. Quantitative analysis of watershed geomorphology. Transactions, American Geophysical Union 38:913-920. Stuart, S. N., E. O. Wilson, J. A. McNeely, R. A. Mittermeier, and J. P. Rodriguez. 2010. The barometer of life. Science 328:177. Terborgh, J., C. van Schaik, L. Davenport, and M. Rao, editors. 2002. Making Parks Work: Strategies for Preserving Tropical nature. Island Press, Washington DC. Tetlock, P. E. 2003. Thinking the unthinkable: Sacred values and taboo cognitions. Trends in cognitive sciences 7:320-324. Thackway, R., and I. Cresswell. 1995. An interim beogeographic regionalisation for Australia: a framework for establishing the national system of reserves, Version 4.0. Australian Nature Conservation, Canberra. Thomas, C. D., A. Cameron, R. E. Green, M. Bakkenes, B. L. J., Y. C. Colllingham, B. F. N. Erasmus, M. F. d. Siqueira, A. Grainger, L. Hannah, L. Hughes, B. Huntley, A. S. van Jaarsveld, G. F. Midgley, L. Miles, M. A. Ortega-Huerta, A. T. Peterson, O. L. Phillips, and S. E. Williams. 2004. Extinction risk from climate change. Nature 427:145-148. Thomson, P. C., and D. Algar. 2000. The uptake of dried meat baits by foxes and investigations of baiting rates in Western Australia. Wildlife Research 27:451-456. Thuiller, W., D. M. Richardson, P. Pyšek, G. F. Midgley, G. O. Hughes, and M. Rouget. 2005. Niche-based modelling as a tool for predicting the risk of alien plant invasions at a global scale. Global Change Biology 11:2234–2250. Tilman, D. 2012. Biodiversity & environmental sustainability amid human domination of global ecosystems. Daedalus 141:108-120. Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 370:66. Tóth, S. F., R. G. Haight, and L. W. Rogers. 2011. Dynamic reserve selection: Optimal land retention with land-price feedbacks. Operations Research 59:1059-1078.

219

Trebilco, R., B. Halpern, J. M. Flemming, C. Field, W. Blanchard, and B. Worm. 2011. Mapping species richness and human impact drivers to inform global pelagic conservation prioritisation. Biological Conservation 144:1758-1766. Tulloch, A., H. Possingham, and K. Wilson. 2011. Wise selection of an indicator for monitoring the success of management actions. Biological Conservation 144:141-154. Tulloch, A. I. T., I. Chadès, and H. P. Possingham. 2013a. Accounting for complementarity to maximize monitoring power for species management. Conservation Biology 27:988-999. Tulloch, V. J., H. P. Possingham, S. D. Jupiter, C. Roelfsema, A. I. Tulloch, and C. J. Klein. 2013b. Incorporating uncertainty associated with habitat data in marine reserve design. Biological Conservation 162:41-51. Underwood, E. C., M. R. Shaw, K. A. Wilson, P. Kareiva, K. R. Klausmeyer, M. F. McBride, M. Bode, S. A. Morrison, J. M. Hoekstra, and H. P. Possingham. 2008. Protecting biodiversity when money matters: Maximizing return on investment. PLoS ONE 3:e1515. United States Environmental Protection Agency (EPA). 2009. Valuing the protection of ecological systems and services (EPA-SAB-09-012). EPA Science Advisory Board, Washington, DC. Urban, M. C., B. L. Phillips, D. K. Skelly, and R. Shine. 2007. The cane toad's (Chaunus [Bufo] marinus) increasing ability to invade Australia is revealed by a dynamically updated range model. Proceedings of the Royal Society of London B Biological Sciences 274:1413-1419. Vale, M. M., M. Cohn-Haft, S. Bergen, and S. L. Pimm. 2008. Effects of future infrastructure development on threat status and occurrence of Amazonian birds. Conservation Biology 22:1006-1015. van Teeffelen, A. J. A., and A. Moilanen. 2008. Where and how to manage: optimal selection of conservation actions for multiple species. Biodiversity Informatics 5:1-13. Vane-Wright, R. I., C. J. Humphries, and P. H. Williams. 1991. What to protect? --Systematics and the agony of choice. Biological Conservation 55:235-254. Venkatesh, V., and F. D. Davis. 2000. A theoretical extension of the technology acceptance model: four longitudinal field studies. Management science 46:186-204. Venkatraman, N., and V. Ramanujam. 1986. Measurement of business performance in strategy research: a comparison of approaches. Academy of management review 11:801-814. Venter, O., N. N. Brodeur, L. Nemiroff, B. Belland, I. J. Dolinsek, and J. W. A. Grant. 2006. Threats to endangered species in Canada. Bioscience 56:903-910. Venter, O., H. P. Possingham, L. Hovani, S. Dewi, B. Griscom, G. Paoli, P. Wells, and K. A. Wilson. 2012. Using systematic conservation planning to minimize REDD+ conflict with agriculture and logging in the tropics. Conservation Letters.

220

Vié, J.-C., C. Hilton-Taylor, and S. N. Stuart, editors. 2009. Wildlife in a Changing World - An Analysis of the 2008 IUCN Red List of Threatened Species. IUCN, Gland, Switzerland. Visser, H., and T. de Nijs. 2006. The map comparison kit. Environmental Modelling & Software 21:346-358. Vörösmarty, C. J., P. McIntyre, M. O. Gessner, D. Dudgeon, A. Prusevich, P. Green, S. Glidden, S. E. Bunn, C. A. Sullivan, and C. R. Liermann. 2010. Global threats to human water security and river biodiversity. Nature 467:555-561. Waldron, A., A. O. Mooers, D. C. Miller, N. Nibbelink, D. Redding, T. S. Kuhn, J. T. Roberts, and J. L. Gittleman. 2013. Targeting global conservation funding to limit immediate biodiversity declines. Proceedings of the National Academy of Sciences 110:12144-12148. Walker, D. A., P. J. Webber, E. F. Binnian, K. R. Everett, N. D. Lederer, E. A. Nordstrand, and M. D. Walker. 1987. Cumulative impacts of oil fields on northern Alaskan landscapes. Science 238:757-761. Walsh, J. C., K. A. Wilson, J. Benshemesh, and H. P. Possingham. 2012. Unexpected outcomes of invasive predator control: the importance of evaluating conservation management actions. Animal Conservation 15:319-328. Walters, C. 1997. Challenges in adaptive management of riparian and coastal ecosystems. Conservation Ecology 1(2):1. http://www.consecol.org/vol1/iss2/art1 Accessed 9 Jan 2015. Walters, C. J. 1986. Adaptive management of renewable resources. McGraw Hill, New York. Watson, J. E. M., M. C. Evans, J. Carwardine, R. A. Fuller, L. N. Joseph, D. B. Segan, M. F. J. Taylor, R. J. Fensham, and H. P. Possingham. 2011. The capacity of Australia's protected- area system to represent threatened species. Conservation Biology 25:324–332. Watts, M. E., I. R. Ball, R. S. Steward, C. J. Klein, K. Wilson, C. Steinback, L. Reinaldo, L. Kircher, and H. P. Possingham. 2009. Marxan with zones: Software for optimal conservation based land- and sea-use zoning. Environmental Modelling & Software 24:1513-1521. Weinstein, M. C., and W. B. Stason. 1977. Foundations of cost-effectiveness analysis for health and medical practices. New England Journal of Medicine 296:716-721. Weitzman, M. L. 1998. The Noah's Ark problem. Econometrica 66:1279. Wilcove, D. S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 1998. Quantifying threats to imperiled species in the United States. Bioscience 48:607-615. Williams, J. C., C. S. ReVelle, and S. A. Levin. 2004. Using mathematical optimization models to design nature reserves. Frontiers in Ecology and the Environment 2:98-105. Wilson, E. O. 1989. Threats to biodiversity. Scientific American 261:108-116.

221

Wilson, E. O. 1992. The Diversity of Life. Harvard University Press, Cambridge. Wilson, K. A., M. Cabeza, and C. J. Klein. 2009a. Fundamental concepts of spatial conservation prioritization. Pages 16-27 in A. Moilanen, K. A. Wilson, and H. P. Possingham, editors. Spatial Conservation Prioritization. Oxford University Press, Oxford. Wilson, K. A., J. Carwardine, and H. P. Possingham. 2009b. Setting conservation priorities. Annals of the New York Academy of Sciences 1162:237-264. Wilson, K. A., M. C. Evans, M. Di Marco, D. C. Green, L. Boitani, H. P. Possingham, F. Chiozza, and C. Rondinini. 2011a. Prioritizing conservation investments for mammal species globally. Philosophical Transactions of the Royal Society of London B Biological Sciences 366:2670-2680. Wilson, K. A., M. Lulow, J. Burger, Y.-C. Fang, C. Andersen, D. Olson, M. O’Connell, and M. F. McBride. 2011b. Optimal restoration: accounting for space, time and uncertainty. Journal of Applied Ecology 48:715-725. Wilson, K. A., M. F. McBride, M. Bode, and H. P. Possingham. 2006. Prioritising global conservation efforts. Nature 440:337-340. Wilson, K. A., E. Meijaard, S. Drummond, H. S. Grantham, L. Boitani, G. Catullo, L. Christie, R. Dennis, I. Dutton, A. Falcucci, L. Maiorano, H. P. Possingham, C. Rondinini, W. R. Turner, O. Venter, and M. Watts. 2010. Conserving biodiversity in production landscapes. Ecological Applications 20:1721-1732. Wilson, K. A., A. Newton, C. Echeverria, C. Weston, and M. Burgman. 2005a. A vulnerability analysis of the temperate forests of south central Chile. Biological Conservation 122:9-21. Wilson, K. A., R. L. Pressey, A. Newton, M. Burgman, H. P. Possingham, and C. Weston. 2005b. Measuring and incorporating vulnerability into conservation planning. Environmental Management 35:527-543. Wilson, K. A., E. C. Underwood, S. A. Morrison, K. R. Klausmeyer, W. W. Murdoch, B. Reyers, G. Wardell-Johnson, P. A. Marquet, P. W. Rundel, M. F. McBride, R. L. Pressey, M. Bode, J. M. Hoekstra, S. Andelman, M. Looker, C. Rondinini, P. Kareiva, M. R. Shaw, and H. P. Possingham. 2007. Conserving biodiversity efficiently: What to do, where, and when. PLoS Biology 5:1850-1861. Wilson, K. A., M. I. Westphal, H. P. Possingham, and J. Elith. 2005c. Sensitivity of conservation planning to different approaches to using predicted species distribution data. Biological Conservation 22:99-112. Wilson, R. J., C. D. Thomas, R. Fox, D. B. Roy, and W. E. Kunin. 2004. Spatial patterns in species distributions reveal biodiversity change. Nature 432:393-396.

222

Wisz, M. S., R. J. Hijmans, J. Li, A. T. Peterson, C. H. Graham, and A. Guisan. 2008. Effects of sample size on the performance of species distribution models. Diversity and Distributions 14:763-773. Withey, J. C., J. J. Lawler, S. Polasky, A. J. Plantinga, E. J. Nelson, P. Kareiva, C. B. Wilsey, C. A. Schloss, T. M. Nogeire, A. Ruesch, J. Ramos Jr, and W. Reid. 2012. Maximising return on conservation investment in the conterminous USA. Ecology Letters 15:1249-1256. Wu, J. 2000. Slippage effects of the conservation reserve program. American Journal of Agricultural Economics 82:979-992. Yackulic, C. B., R. Chandler, E. F. Zipkin, J. A. Royle, J. D. Nichols, E. H. Campbell Grant, and S. Veran. 2012. Presence-only modelling using MAXENT: when can we trust the inferences? Methods in Ecology and Evolution doi: 10.1111/2041-210x.12004. Zavaleta, E. S., R. J. Hobbs, and H. A. Mooney. 2001. Viewing invasive species removal in a whole-ecosystem context. Trends in Ecology & Evolution 16:454-459.

223