The Interactive Effects of Fire and Herbivory on Understorey Vegetation and Its Dependent Fauna

The interactive effects of fire and herbivory on understorey vegetation and its dependent fauna

Claire Foster

A thesis submitted for the degree of Doctor of Philosophy of

The Australian National University

August 2015

Declaration

This thesis is my own work, except where otherwise acknowledged

(see Preface and Acknowledgements)

Claire Foster

August 2015

iii

Preface

This thesis is structured as a series of connected papers that have been published in, or were submitted to, peer reviewed journals at the time of thesis submission. These papers are listed at the end of this preface and are referred to in the text by their roman numerals. All papers were intended as stand-alone pieces of work. For this reason, there is some unavoidable repetition between chapters, for example in the background material, and the description of the study area and experimental design. Due to journal copy-editing there may be minor differences between the papers presented here and their published versions.

The formatting and content of this thesis complies with The Australian National

University’s College of Medicine, Biology and Environment guidelines for ‘Thesis by

Compilation’. In line with these guidelines, an Extended Context Statement has been provided at the beginning of the thesis. The Extended Context Statement is not intended to be a complete literature review, but rather a framework for understanding the relationships between all aspects of the research. Paper I provides a detailed review of literature concerning the effects of large herbivores on other animals, and Papers II, III, IV and V each present a review of the literature relevant to the focus of the paper.

I performed the majority of the work for the papers that form this thesis. This included developing research questions and experimental designs, data collection and analysis, and writing manuscripts. My supervisors (Philip Barton, Chloe Sato and David Lindenmayer) and collaborators provided advice on conceptualisation, experimental design, data analysis, and manuscript revisions. The addition of different co-authors to each paper reflects contributions from collaborators, which are detailed below. The author contribution statements below have been agreed to in writing by all authors. Other assistance is acknowledged in the acknowledgments section at the end of each paper.

Paper I: Foster, C.N., Barton, P.S. and D.B. Lindenmayer (2014) The effects of large native

herbivores on other animals. Journal of Applied Ecology. 51, 929-938.

Conceptualisation & design: CNF, DBL, PSB; Data extraction and analysis: CNF; Manuscript

drafting: CNF; Manuscript revision: CNF, PSB, DBL.

Paper II: Foster, C.N., Barton, P.S., Sato, C.F., Wood, J.T., Macgregor, C.I., and D.B.

Lindenmayer (2016) Herbivory and fire interact to affect forest understory habitat, but

not its use by small vertebrates. Animal Conservation. 19, 15-25.

Conceptualisation & design: CNF, DBL, CIM, PSB, JTW; Data collection: CNF, CIM; Data

analysis: CNF, JTW, CFS; Manuscript drafting: CNF; Manuscript revision: CNF, PSB,

DBL, CFS, CIM, and JTW.

Paper III: Foster, C.N, Barton, P.S, Sato, C.F, Macgregor, C.I, and D.B. Lindenmayer (2015)

Synergistic interactions between fire and browsing drive plant diversity in a forest

understory. Journal of Vegetation Science. 26, 1112-1123.

Conceptualisation & design: CNF, DBL, CIM, PSB; Data collection: CNF, CIM; Data analysis:

CNF, CFS; Manuscript drafting: CNF; Manuscript revision: CNF, PSB, DBL, CFS, CIM.

Paper IV: Foster, C.N, Barton, P.S., Wood, J.T. and D.B. Lindenmayer (2015) Interactive

effects of fire and large herbivores on web-building spiders. Oecologia. 179, 237-248.

Conceptualisation & design: CNF, DBL, PSB, JTW; Data collection: CNF, Data analysis: CNF,

PSB, JTW; Manuscript drafting CNF, Manuscript revision: CNF, PSB, DBL, JTW.

Paper V: Foster, C.N, Sato, C.F., Lindenmayer, D.B., and P.S. Barton (Accepted) Integrating

theory into disturbance interaction experiments to better inform ecosystem management.

Global Change Biology, doi:10.1111/gcb.13155.

Conceptualisation & design: CNF, PSB, CFS, DBL; Data extraction: CNF; Manuscript drafting:

CNF, Manuscript revision: CNF, PSB, CFS, DBL.

Acknowledgements

Thank you to my supervisors David Lindenmayer, Chloe Sato, and Philip Barton for their support and guidance during my candidature. David’s confidence in me was pivotal for allowing me to get my experiments up and running within a few months of commencing. The freedom David allowed me in my topics and approach enabled me to complete a PhD that I have found both interesting and thoroughly enjoyable. Chloe’s advice on analyses, discussions on applied management, and general assistance with the whole PhD process has been invaluable. Philip’s advice, constructive feedback and patience with trawling through every first draft have been fantastic. Our regular meetings have helped me stay focussed, and turn my disjointed rambles into something worth writing about. My PhD would have been a very different experience without your advice and encouragement. I would also like to thank Chris

MacGregor, Peter Lane and Jeff Wood for their assistance and advice at various times during my candidature. In particular, Chris MacGregor’s cheerful company, willing assistance and ongoing advice made my fieldwork, and especially the dreaded fence construction, so much more enjoyable. Chris was always happy to keep an eye on my experiments, chainsaw a tree from exclosure fences or help identify a difficult specimen. John Evans also provided invaluable assistance with sorting and identifying invertebrate collections. All of my supervisors and collaborators have been outstanding in providing constructive and timely feedback on my written work, for which I am very grateful.

I would like to thank the Department of Environment and the landowners and co- managers of Booderee National Park, the Wreck Bay Aboriginal Community, for allowing me to conduct this study in Booderee National Park and providing in-kind support. I would particularly like to thank Nick Dexter and Martin Fortescue for their ongoing support, advice and interest in this project, and making the burning of my experimental plots a reality. Thank you also to all the staff of Booderee National Park who were involved in conducting the prescribed burns, especially Stuart James, who gave me training so I could participate myself.

vii

Much of my field work would not have been possible without the assistance of a band of willing volunteers. Thank you to Angela, Stuart, Darren, Dylan, Helen, Emily, Xiying, Laura,

Kevin, Daniela, Anna and Luke for cheerfully putting up with the early starts, leeches and ticks and wrestling with all the gates.

Thank you to the Long Term Ecological Research Network, the Australian Academy of

Science (Margaret Middleton Fund) and the Normal Wettenhall Foundation for granting financial support to this project. Thank you also to the ANU College of Medicine and Biology for providing me with a supplementary scholarship.

My fellow students and the staff at the Fenner School created an environment of encouragement, active discussion and constructive feedback that made my candidature so much more enjoyable. In particular, thank you to Ben Scheele for great discussions, good company, and reading drafts, and Karen Ikin for your friendship, encouragement and the morning coffee run. I would also like to thank Nélida Villaseñor, Darren Le Roux, Will Batson, Laurence

Berry, Nick Engerer, John Evans, Geoff Kay, Miles Kieghley, Megan Evans, Stephanie

Pulsford, Laura Rayner, Chloe Sato, Philip Barton, Ingrid Stirnemann, Martin Westgate, Alessio

Mortelliti, Jenny Pierson and Crid Fraser for numerous instances of assistance and encouragement, as well as many great lunchtime discussions. Thank you also to Claire

Shepherd and Tabitha Boyer for your assistance with navigating the university administration jungle.

Thank you to my family for their love and support throughout this experience. In particular to my partner Stuart, who has provided love, good humour and positivity over the last

3.5 years. Thank you for continuing to visit during long periods of field work, even when I made you get up at five, or count poo for your trouble, and for putting things in perspective when I needed it. Finally, thank you to my Mum, for making the effort to understand what I do, correcting my grammar, and encouraging my curiosity from the very beginning.

viii

Abstract

Interactions between multiple disturbances have been shown to have unexpected, and often undesirable effects on ecosystems and biodiversity. Improving our ability to predict and manage the outcomes of multiple disturbances is therefore an important research priority. In this thesis, I focussed on the interaction between fire and grazing (or browsing) by large herbivores.

Evidence of the individual effects of fire and large herbivores is substantial, but there has been little quantitative synthesis of the effects of native herbivores on biodiversity. Using a systematic review and meta-analysis, I found that high densities of large herbivores usually have negative effects on other animals. However, I found that interactions between large herbivores and episodic disturbances, such as fire, remain poorly understood.

I therefore designed a field experiment to test the interactive effects of prescribed fire and large herbivores on forest flora and fauna. I used full and partial exclosure fences to create a gradient of herbivore pressure across both burnt and unburnt sites, and measured the responses of vegetation, spiders and small vertebrates.

I found that fire and herbivory interacted strongly to affect vegetation, with herbivory limiting the recovery of vegetation from fire. In contrast to the vegetation response, small vertebrates responded to the individual, but not interactive effects of disturbance. I then focussed on the mechanisms driving interactive effects on vegetation, and found that the interaction occurred through both numerically mediated (concentration of herbivores in burnt sites) and functionally moderated (stronger effect of herbivores post-fire) pathways.

The differing responses of plants and animals to fire and herbivory was at odds with existing literature, where the effects of large herbivores on fauna are usually attributed to vegetation changes. I therefore tested to what extent vegetation mediated the effects of fire and herbivory on web-building spiders – a group sensitive to changes in habitat structure.

Vegetation structure partially mediated the negative effect of fire on spider density, while negative effects of large herbivores on spiders were mostly independent of vegetation. Different ix types of web builders differed in their responses, resulting in important changes to the spider community following disturbance.

The results of my experiments highlight the importance of focussing on mechanistic pathways for understanding and managing disturbance interactions. However, in reviewing recently published fire-grazing studies, I found that most reported only net effects of interactions. I demonstrate how by failing to identify mechanistic pathways, or non-linear effects, such studies are limited in their management applications. I describe adjustments to disturbance interaction studies that would improve their ability to inform effective management and advance theory.

Collectively, my research shows that fire and large herbivores can have strong interactive effects on forested ecosystems and their associated biota, and highlights the importance of considering large herbivores in forest fire planning. It also demonstrates the value of a mechanistic understanding of interactions for the management of disturbance regimes. Such considerations are of broad relevance to the management of multiple disturbances, particularly in the context of increasing global change.

Table of Contents

Declaration ...... iii Preface ...... v Acknowledgements ...... vii Abstract ...... ix

Context Statement ...... 1 Introduction ...... 1 Overview of Aims ...... 4 Study Area ...... 7 Summary of Outcomes ...... 9 Synthesis and Management Implications ...... 11 Conclusion ...... 17 References ...... 18

Paper I: Effects of large native herbivores on other animals ...... 25 Summary ...... 26 Keywords ...... 27 Introduction ...... 27 Materials and Methods ...... 29 Results ...... 33 Discussion ...... 39 Acknowledgements ...... 46 References ...... 46 Supporting Information ...... 51

Paper II: Herbivory and fire interact to affect forest understory habitat, but not its use by small vertebrates ...... 81 Abstract ...... 82 Keywords ...... 82 Introduction ...... 82 Materials and methods ...... 85 Results ...... 88 Discussion ...... 95 Acknowledgments ...... 100 References ...... 100 Appendices ...... 105

Paper III: Synergistic interactions between fire and browsing drive plant diversity in a forest understory ...... 115 Abstract ...... 116 Keywords ...... 116 Introduction ...... 117 Methods ...... 119 Results ...... 124 Discussion...... 129 Acknowledgements ...... 134 Literature Cited ...... 134 Supporting Information ...... 138

Paper IV: Interactive effects of fire and large herbivores on web-building spiders ...... 145 Abstract ...... 146 Keywords ...... 146 Introduction ...... 147 Materials and methods ...... 149 Results ...... 156 Discussion...... 161 Acknowledgements ...... 165 References ...... 165 Supplementary Material ...... 171

Paper V: Integrating theory into disturbance interaction experiments to better inform ecosystem management ...... 183 Abstract ...... 184 Keywords ...... 184 Introduction ...... 185 Recent advances in disturbance interaction theory ...... 186 Where are the gaps between theory and experimental studies?...... 190 Bridging the gap to improve management applications ...... 199 Conclusions ...... 203 Acknowledgements ...... 204 Literature Cited ...... 204 Supporting Information ...... 209

xii

Context Statement

Introduction

Ecological disturbance is a key process driving patterns of temporal and spatial heterogeneity in ecosystems worldwide. Disturbance regimes are strongly linked with drivers of global change, and hence many disturbance regimes have been dramatically altered in recent decades, and continue to undergo rapid change (Turner 2010). While there has been increasing interest in interactions between global change drivers (e.g. Tylianakis et al. 2008, Oliver and Morecroft

2014, Mantyka-Pringle et al. 2015), and also global change and disturbance (e.g. Didham et al.

2007, Gibb et al. 2015), the interactions between disturbance processes themselves remain poorly understood (Turner 2010, Buma 2015). This knowledge gap is of critical importance for ecological management as disturbances are becoming increasingly managed in both natural and modified ecosystems, rarely occur in isolation from other disturbance processes, and are predicted to continue to change in frequency and intensity as a result of global change (Turner

2010, Buma 2015). Improving our understanding of both the individual and combined effects of disturbance is therefore imperative for the successful conservation of ecosystems and biodiversity.

Disturbance can be defined as any “relatively discrete event that disrupts the structure of an ecosystem, community or population, and changes resource availability or the physical environment” (White and Pickett 1985). A disturbance regime refers to the spatio-temporal distribution of disturbance, including aspects such as size, intensity, severity and frequency

(Turner et al. 1998). Two of the most widespread and commonly studied disturbance processes in terrestrial ecosystems are fire and herbivory (grazing and browsing by large herbivores).

Fire occurs over a wide range of temperate and tropical landscapes, and fire regimes drive the distribution of vegetation types and dominant species in many ecosystems (Bond and

Keeley 2005). Fire regimes have undergone substantial changes in recent decades, from widespread fire suppression in some ecosystems, to an increase in fire intensity, severity and

1 frequency in others (Bowman et al. 2009, Syphard et al. 2009). Attempts to re-introduce historical fire regimes to contemporary ecosystems can be contentious, particularly in human- dominated landscapes, where wildfire is often perceived as a “devastating” destructive event

(Stephens and Ruth 2005, Stephens et al. 2013). Managing fire for the (often competing) priorities of both life and property protection and biodiversity conservation is therefore a key challenge for managers of many ecosystems (Driscoll et al. 2010, Gibbons et al. 2012).

Browsing and grazing by large mammalian herbivores is also a widespread disturbance process that occurs in most of the world’s ecosystems (Danell et al. 2006). Rangelands, due to their suitability for livestock grazing, have been particularly affected by changes to grazing regimes (Gordon et al. 2004). However, wild herbivore assemblages have also undergone substantial changes in recent decades, with dramatic declines from over-hunting in some areas, and increases to historically high levels in others, due to relaxation of hunting pressure and loss of native predators (Gordon et al. 2004). Changes in herbivore densities can have dramatic effects on ecosystems, from altering dominance and diversity patterns in plant and animal communities (Hester et al. 2006, Suominen and Danell 2006), to changing landscape-scale processes such as nutrient cycling, erosion and hydrological cycles (Asner et al. 2004,

Tanentzap and Coomes 2012). Due to their strong effects on ecosystems, and widespread changes in abundance, large herbivore populations are actively managed for conservation outcomes throughout much of their range (Danell et al. 2006).

As common and widespread disturbances, fire and herbivory co-occur in a wide range of habitats. Understanding how these processes interact in the different ecosystems in which they co-occur is therefore essential if the management of either is to be effective and efficient

(Wisdom et al. 2006, Fuhlendorf et al. 2009). Existing studies of fire-grazing interactions generally focus on one of two interaction types. The first interaction (referred to in this thesis as an interaction chain) involves feedbacks between fire and grazing through space and time: fire leads to a concentration of grazing pressure in burnt areas, which then reduces the chance of future fire in those areas, but increases the risk in unburnt areas where reduced grazing pressure allows greater accumulation of biomass (Fuhlendorf et al. 2009, Kimuyu et al. 2014). Some 2 researchers argue that this coupling of fire and grazing in space and time should be thought of as a single disturbance process, termed “pyric herbivory” (Fuhlendorf et al. 2009). The second type of interaction (referred to in this thesis as an interaction modification) is generally studied at the site-level and occurs when the effect of grazing on plants or animals is modified by fire (or vice versa) (Eby et al. 2014, Koerner and Collins 2014). These interactions often occur via changes to the traits or behaviours of plants or animals. For example, fire can remove the chemical and physical defences of plants, making generally unpalatable plants palatable (Augustine and

McNaughton 1998, Augustine and Derner 2015). After fire, the selectivity of grazers, and therefore their effects on patterns of plant dominance and diversity can be significantly altered

(e.g. Augustine and Derner 2015). While often studied separately, both of these interaction types are likely to be acting when fire and herbivory co-occur in natural systems.

The majority of fire-grazing interaction studies to date have been of domestic herbivores in rangeland ecosystems. This bias is understandable due to the large global extent of livestock grazing in rangelands, as well as the comparative ease of manipulating domestic livestock densities compared with manipulating populations of wild herbivores. However, fire and large herbivores also co-occur, and are actively managed, in forested systems, but here their interactions remain poorly understood (Wisdom et al. 2006, Royo et al. 2010a). Browsing can affect the successional trajectory of forests, resulting in shifts in the abundance and composition of both canopy, and understory species (Olofsson et al. 2005, Royo and Carson 2006, Mathisen et al. 2010, Tanentzap et al. 2012, Hidding et al. 2013, Holm et al. 2013). As fire re-starts succession, and can alter vegetation species composition, plant traits, and herbivore behaviour, there is strong potential for fire to interact with browsing, with possible long-term consequences for the state of forest ecosystems (Royo et al. 2010a, Pekin et al. 2014). Indeed, the few existing studies to date indicate that fire-browsing interactions are prevalent in forests (Royo et al.

2010a, Kerns et al. 2011). The effect of fire-browsing interactions on biodiversity therefore warrants urgent attention from both managers and researchers of forested ecosystems.

In this thesis I investigated how fire and browsing interact to affect understory flora and fauna in a eucalypt forest ecosystem of south-eastern Australia. I focussed on this system as the 3 role of native herbivores in Australian ecosystems remains poorly understood, but existing studies show that the effects of macropod herbivores on both plants and animals can be substantial (Bulinski and McArthur 1999, Meers and Adams 2003, Tuft et al. 2012, Howland et al. 2014). Further, although both fire and macropods are managed in many Australian ecosystems (Clarke 2008, Gowans et al. 2010, Lunney 2010, Wiggins and Bowman 2011,

Enright and Fontaine 2014), they are generally considered as separate issues in contemporary management decisions (Meers and Adams 2003). Fire is usually managed with the aim of reducing the risk of large, high intensity wildfires to prevent life and property loss, although some burning is also done for ecological purposes (Morrison et al. 1996, Driscoll et al. 2010).

Macropod populations are also managed in many Australian ecosystems, either directly through culling (or other control) of overabundant populations (Coulson et al. 2008, Gowans et al. 2010,

Wiggins and Bowman 2011, Howland et al. 2014), or indirectly via ongoing control of both native (dingo, Canis familliaris) and introduced (European red foxes, Vulpes vulpes) predators

(Banks et al. 2000, Coates 2008, Nimmo et al. 2015). The few existing studies of the effects of fire and macropod herbivores on biodiversity indicate that interactive effects are likely to be prevalent (Catling et al. 2001, Meers and Adams 2003, Tuft et al. 2012, Dexter et al. 2013,

Pedersen et al. 2014). The forests of south-eastern Australia therefore provide an opportunity to study general questions about the interactive effects of fire and browsing on patterns of plant and animal diversity, and also to generate insights into how to account for fire-browsing interactions in the management of these forests.

Overview of Aims

This thesis includes three related components (Fig. 1), each addressing different questions, but all of which contribute to the central aim of improving understanding and management of fire- herbivore interactions. These components are: a global review of the effects of large herbivores on other animals (Paper I); a series of field studies to test the interactive effects of fire and herbivory on forest flora and fauna (Papers II, III and IV); and a synthesis (Paper V) which builds on my field studies and the broader literature to critically evaluate the way that empirical studies inform the management of interacting disturbances. 4

Numerous studies have highlighted the important role played by large herbivores in shaping ecosystem structure and function (Danell et al. 2006). However, there has been limited quantitative synthesis of the effect of native herbivores on biodiversity, despite wild herbivore populations being managed for biodiversity conservation in many ecosystems worldwide. In

Paper I, I conducted a quantitative review of the effects of large native herbivores on other animals. This paper aimed to (i) summarise generalities and trends in the effects of large herbivores on different animal groups, (ii) identify gaps and limitations in the literature, and the reasons for these gaps, and (iii) highlight priorities for future research.

Paper I: . Interactions with other processes poorly How do large herbivores affect other animals? understood . Vegetation change cited as a key mechanism, but rarely quantitatively tested . Multi-taxa studies rare, but needed to identify key pathways

Field experiment: Fire-browsing interactions

Paper II: How do fire-herbivore interactions affect forest . Different groups of biota have different vegetation and small vertebrates? responses to disturbance interactions . Fauna responses not predictable from vegetation responses Paper III: By which mechanisms do fire and large . Herbivores, and the fire x herbivore herbivores interact to affect forest plant diversity? interaction can have important effects on fauna that are not vegetation mediated . Mechanistic understanding of interaction types useful for targeted, efficient Paper IV: management, and identifying novel solutions To what extent are effects on animals mediated by vegetation structure?

Paper V: Could a focus on interaction pathways improve the ability of fire-grazing studies to inform ecological management?

Figure 1. Relationships between the key questions explored in this thesis.

Understanding of how the effects of large herbivores plants and animals are modified by management practices and disturbance regimes remains limited, particularly for forested ecosystems (Paper I, Wisdom et al. 2006). However, understanding these interactions is essential for informing effective and efficient management of the effects of large herbivore on biodiversity (Wisdom et al. 2006). I therefore conducted a field experiment to test whether large herbivores and fire interact to affect the forest ecosystem. I used this experiment to: (i) test the interactive effects of fire and large herbivores on forest understory plants and animals (Paper

II), (ii) identify the mechanisms by which fire and large herbivores interacted (Paper III), and

(iii) test the extent to which the effects of fire and large herbivores on other animals were mediated by changes in vegetation (Paper IV). Collectively, these papers (II-IV) aimed to develop an improved, mechanistic understanding of the interaction between fire and large herbivores, and how this affects biodiversity in forested ecosystems.

Several recent studies have highlighted the value of understanding interaction pathways for both predicting and managing the effects of co-occurring stressors or threats (Didham et al.

2007, Boyd and Brown 2015, Doherty et al. 2015). I also found that the multi-taxa, mechanistic approach I took in Papers II, III and IV allowed me to critically evaluate different management options. However, in exploring the fire-grazing literature, I encountered few examples of studies which took a similar mechanistic approach. In Paper V, I examined whether a focus on interaction pathways could improve the ability of fire-grazing studies to inform ecological management. This final paper aimed to: (i) quantify the extent to which fire-grazing studies do/do not take a mechanistic approach, (ii) investigate whether failing to take a mechanistic approach limits the management relevance of studies, and (iii) identify adjustments to disturbance interaction studies that could improve their ability to inform effective ecological management. Overall, Paper V aimed to identify approaches to studying disturbance interactions that provide valuable insights for management, and should therefore form the basis of future studies investigate fire-grazing or fire-browsing interactions.

Study Area

The empirical papers in this thesis (Papers II, III, and IV) are based on a field experiment that I conducted in the Eucalypt forests of Booderee National Park (BNP). A description of the experimental design can be found in Paper II (Paper II, Methods, and Appendix 1), and a map of the study area is included in Paper III (Paper III, Fig. 1). Below, I provide a more detailed description of the study area and its management, as well as some aspects of the methods, which could not be included in individual papers due to journal space limitations.

Booderee National Park is located on a 6 500 ha peninsula of coastal South-Eastern

Australia (35°10′S, 150°40′E). The area has a temperate maritime climate with rainfall (average

1225 mm per annum) occurring throughout the year (Lindenmayer et al. 2008). The vegetation of BNP comprises a spatially heterogeneous mosaic of vegetation types, the most widespread of which are eucalypt forest, heathland, eucalyptus/banksia woodland, and shrubland (Taws 1998,

Lindenmayer et al. 2008). Fire occurs naturally in the ecosystems of south-eastern Australia, and many native species are well adapted to intermittent fire, but not to frequent fire (fire return intervals of < 8-10 years cannot be survived by many species, Morrison et al. 1996, Enright and

Fontaine 2014). Booderee National Park has a complex mosaic of fire histories, which includes a large wildfire in 2003 which burnt over 50% of the park’s area (Lindenmayer et al. 2008).

While large wildfire events occur occasionally in BNP, because of the proximity of urban areas and the presence of important populations of threatened species, small-scale prescribed fires are used to reduce fuel loads and minimise the risk of large wildfires.

Management of BNP also includes an intensive pest control program, targeting the

European red fox, which has been in place since 1999 (Dexter et al. 2007). This program has been successful in increasing the abundances of some native small mammals in the park (Dexter et al. 2007). However, the loss of large predators (dingoes), and the absence of hunting, means that without foxes, predation pressure on larger mammals is low (Lindenmayer et al. 2014).

This has resulted in increases in the abundance of native macropod herbivores, with counts increasing ten-fold between 2003 and 2009 (Dexter et al. 2012, Lindenmayer et al. 2014). The most abundant of these macropods is the swamp wallaby (Wallabia bicolor, a mixed feeder, 7

Davis et al. 2008, Di Stefano and Newell 2008), but eastern grey kangaroos (Macropus giganteus, grazers, Davis et al. 2008), and red-necked wallabies (M. rufogriseus, grazers, Sprent and McArthur 2002) also occur (Dexter et al. 2012, Dexter et al. 2013). Non-native European rabbits (Oryctolagus cuniculus) also occur within BNP, however they are at low abundance and do not appear to be increasing (Dexter et al. 2013, Stutz et al. 2015). The high abundances of macropods currently present in BNP are compromising efforts to restore ex-plantation areas

(Stutz et al. 2015), and also may be limiting the recovery of some plant species following weed control burns (Dexter et al. 2013). Park managers are therefore concerned about what broader impacts these abundant macropods may be having on the flora and fauna of the park, and whether these impacts interact with the prescribed burning program.

In the empirical chapters of this thesis (Papers II, III and IV) I report results from an experiment where I used a factorial combination of prescribed fire (burnt, unburnt) and herbivore exclosure (open, partial exclosure and full exclosure) treatments to test the responses of a range of plant and animal groups to the interactive effects of fire and herbivory (see Paper

II, Methods, and Appendix 1). In this experiment, I monitored the use of sites by macropod herbivores using scat counts, which allowed me to compare quantitatively the activity of herbivores in each treatment, and between sites over time. In these counts, I did not attempt to distinguish between the different macropod species. Most of the scats would have been from swamp wallabies (browsers) as they are the most common macropod in Booderee National

Park. However, some scats would have been from grey kangaroos (grazers), which I occasionally observed at the sites. For this reason, my conclusions and recommendations throughout the thesis relate to macropod herbivores in general and further detail regarding the identity of herbivore species using the sites would need to be obtained before implementing management actions that are targeted toward individual species.

Summary of Outcomes

Paper I: The effects of large native herbivores on other animals

In Paper I, I conducted a systematic review and meta-analysis of the effects of large native herbivores on other animals. I found that, on average, large herbivores had negative effects on the abundance and diversity of other animal groups. However, most studies compared areas of very high herbivore density (overabundance) with areas of very low density (exclusions), which likely biased studies towards detecting negative effects. Most of the effects of large herbivores on other animals were attributed to indirect effects, predominantly through changes in the quantity and/or structure of vegetation, although mediating mechanisms were rarely quantitatively tested.

I identified key knowledge gaps on the effects of large herbivores on other animals including: (i) the prevalence of nonlinear responses to herbivore pressure, (ii) how responses differ between different herbivores, (iii) the spatial and (iv) the temporal variation of responses,

(v) how the effects of herbivores interact with disturbance and land management activities and

(vi) the mechanisms driving cascading effects through ecosystems (including synthesis of multi- taxa responses). This paper was the first quantitative synthesis of the effects of large native herbivores on other animals, and provides important context for the growing literature on the ecological effects of large herbivores.

Paper II: Herbivory and fire interact to affect forest understory habitat, but not its use by small vertebrates.

In Paper II, I tested whether herbivory and fire interacted to affect understory vegetation and small vertebrates. Fire and herbivore exclusion interacted to affect understorey habitat, with vegetation cover, particularly of palatable vegetation, showing poor post-fire recovery on browsed sites. However, despite these interactive effects on vegetation, small vertebrates responded to the individual, not interactive effects of fire and herbivory. Small vertebrate species had contrasting responses to disturbance. For example, the small mammal, brown

9 antechinus (Antechinus stuartii) responded positively to herbivore exclusion, while the delicate skink (Lampropholis delicata) showed the opposite trend.

These results suggest that it may be necessary to manage the macropod herbivore population after fire to prevent the decline of palatable plants, and maintain the dense habitat required by some small mammals. However, as the invasive European rabbit was most active in macropod-free sites after fire, any management must consider both types of herbivores. As different combinations of management benefited different species, and fauna responses were not predictable from vegetation responses, a mix of management practices may be needed to ensure suitable habitat is maintained for all species.

Paper III: Synergistic interactions between fire and browsing drive plant diversity in a forest understory.

In Paper III, I tested the interactive effects of browsing and fire on understorey plant diversity. I found browsing increased plant community dominance, and reduced evenness and diversity in burnt, but not in unburnt sites. This fire–browsing interaction was driven by both numerically mediated and functionally moderated pathways: Fire both increased local browsing intensity, and amplified the per-unit effect of herbivores on the plant community. The altered competitive environment after fire, combined with heavy post-fire browsing created a depauperate understorey, dominated by an unpalatable, fire-resistant fern species (austral bracken, Pteridium esculentum). Due to the ability of bracken to suppress the establishment of other plants, this fern-dominated state may be difficult to reverse. Integrated management of fire and large herbivores may therefore be necessary to prevent the development of a depauperate understorey in areas where herbivores are abundant.

Paper IV: Interactive effects of fire and large herbivores on web-building spiders.

In Paper IV, I tested for the interactive effects of fire and large herbivores on web-building spiders, and the mechanisms by which these effects occurred. I found that fire had a strong negative effect on the density of web-building spiders, which was partly mediated by effects on

10 vegetation structure, while the negative effect of large herbivores on web density was not related to changes in vegetation. Fire amplified the effects of large herbivores on spiders, both by modifying herbivore effects on vegetation, and by increasing herbivore activity. The relative importance of the different pathways driving the effects of disturbance differed between types of web-building spiders, and led to potentially important changes in the composition of this predator guild. This study is one of the first to quantitatively assess the role of vegetation in mediating the effects of fire-browsing interactions on fauna, and highlights the importance of considering the driving mechanisms both when investigating and managing disturbance interactions.

Paper V: Integrating theory into disturbance interaction experiments to better inform ecosystem management.

Theoretical studies have shown that developing efficient, novel solutions to managing multiple stressors relies on a detailed, mechanistic understanding of their interactions. In Paper V, I used a mini-review of the fire-grazing literature to demonstrate that few empirical studies of disturbance interactions were designed to identify the mechanistic pathways driving interactive effects, or to detect non-linear effects (key relationships highlighted in theoretical models as being important for management). I then showed how these widespread limitations among disturbance interaction studies can lead to management recommendations that are ineffective, inefficient, or likely to have negative outcomes. To address this issue, I presented a series of adjustments to the design and interpretation of empirical disturbance interaction studies that would require minimal additional research effort, but would improve the contribution of such investigations to both the theoretical understanding, and applied management of multiple stressors.

Synthesis and Management Implications

My research tested the interactive effects of fire and large herbivores on forest flora and fauna. I found that for vegetation, the effects of fire and browsing were strongly interactive: in the absence of fire, the effects of browsing on vegetation structure and floristics were minimal, but

11 following fire, heavy browsing led to a depauperate understorey dominated by ferns (Papers II,

III). Surprisingly, these interactive effects only partially mediated fauna responses, and both vertebrate and invertebrate fauna were most strongly influenced by the individual, rather than the interactive effects of disturbance (Papers II, IV). These results indicate that at their current densities, the native macropod herbivores may have minimal effects on an undisturbed forest ecosystem, and that management of the herbivore population should be targeted towards post- fire, or early successional vegetation.

While I observed strong interactive effects of fire and herbivory on vegetation, the result that browsing had minimal effects on vegetation in the absence of fire will require further investigation. This is because the experiments in this thesis were of short duration (two years) compared with the lifespans of many understory plants in these forests (time to maturity > 5-8 years for many shrub species, Morrison et al. 1996, Morrison 2002). Herbivory can affect plant communities through mechanisms that can take a number of years to become evident, such as limiting reproductive output and preventing establishment of seedlings (Hester et al. 2006,

Mathisen et al. 2010, Tanentzap et al. 2012), and forest vegetation can be slow to recover after release from chronic herbivory (Royo et al. 2010b, Tanentzap et al. 2011). Therefore, a longer- term study may be needed to detect the full range of herbivory effects. Further, while the plant community exhibited only limited responses to large herbivores in the absence of fire, some fauna (brown antechinus, skinks, spiders) responded to herbivore exclusion regardless of burning treatment. This suggests that some important effects of large herbivores on other fauna were not vegetation-mediated, or, that subtle (undetected) changes in the vegetation in herbivore exclusion sites played an important role in determining fauna occupancy.

Fire and herbivory had strong interactive effects on vegetation. However, it was not possible to conclude whether post-fire herbivory had led the system to an alternate state that would persist long-term, or had simply slowed the recovery of vegetation (Paper III). If post- fire vegetation recovery was simply slowed by browsing, then in the long-term, the individual effects of herbivory may prove to be more important than interactive effects in defining the community structure and floristics. It will therefore be important that monitoring of the 12 experimental plots continues, so that remaining questions regarding the longevity of interactive effects, and herbivore effects on vegetation succession and recruitment can be answered.

My research has shown that small-scale prescribed fires can lead to focal browsing

(localised, high intensity browsing, Fuhlendorf and Engle 2001) by macropod herbivores in burnt forest patches, which simplifies site-level vegetation structure and reduces floristic diversity. Focal grazing (particularly of domestic herbivores) has been found to have net benefits for biodiversity in many grasslands, despite reducing local vegetation structure

(Fuhlendorf et al., 2006; Fuhlendorf et al., 2009; Fuhlendorf et al., 2010; Winter et al., 2012). In grasslands, focal grazing after patch burning can increase the diversity of flora and fauna, because the concentration of herbivores in burnt patches releases vegetation elsewhere from grazing pressure, the contrasting grazing pressures creating a heterogeneous mosaic of vegetation structure, supporting a diversity of species (Fuhlendorf et al. 2010). However, focal browsing following prescribed burning in forests is likely to have markedly different effects on biodiversity.

The effects of focal browsing are likely to differ between grasslands and forests due to general differences between herbivore populations, fire extent and frequency, and plant traits in these ecosystems. Most herbivore populations in forested systems are wild, and free to move across large areas of habitat. When the area of habitat that is burnt is small compared with the total available habitat (which is often the case for prescribed fires in forests), focal browsing in burnt patches is likely to do little to reduce browsing pressure in unburnt areas (Fuhlendorf et al., 2009). Further, even if browsing pressure is reduced temporarily in unburnt areas, the slow- growing nature of many forest plant species and the lack of an evolutionary history of heavy grazing (Milchunas and Lauenroth 1993, Díaz et al. 2007), mean that forest vegetation is likely to be slow to recover following release from browsing pressure (e.g. Royo et al. 2010b,

Tanentzap et al. 2011, Tanentzap et al. 2012). Supporting this reasoning, I found that in

Booderee National Park, the concentration of herbivores in burnt areas lasted only months, but interactive effects on vegetation persisted throughout the study (Papers III, IV). Therefore, small burns which initiate focal browsing of herbivores in forests may lead to the simplification 13 of vegetation of burnt patches, but not allow vegetation complexity or density to increase elsewhere. Thus, focal browsing in forests could potentially lead to the overall decline of vegetation complexity, rather than creating the heterogeneous vegetation structure that promotes biodiversity in grassland systems.

The potential for focal browsing after prescribed fire to lead to simplification of vegetation is of concern for the management of forests in south-eastern Australia for two reasons. The first is that ongoing control of both native (dingoes) and introduced (European red foxes) predators for livestock protection and threatened species conservation mean that some species of macropod herbivores are highly abundant in many forests of south-eastern Australia

(Banks et al. 2000, Coates 2008, Coulson et al. 2008, Gowans et al. 2010, Nimmo et al. 2015).

Where predator control is ongoing, high densities of these herbivores, combined with limited predation risk (predation risk can discourage some herbivores from browsing in open areas,

While and McArthur 2006, Cromsigt et al. 2013, McArthur et al. 2014), means that there is strong potential for fire-browsing interactions such as those observed in this thesis. The second reason for concern is that current fire management policies in south-eastern Australia aim to increase the proportion of fire that occurs through prescribed burning (rather than wildfire)

(Attiwill and Adams 2013, NSW Office of Environment and Heritage 2013). Prescribed fires are usually of lower intensity and severity than wildfire (and so less likely to directly reduce macropod populations), and of smaller spatial extent (and so more likely than wildfire to lead to focal browsing in burnt areas). Further, many plant species in south-eastern Australia are already at risk from increasing fire frequencies, which do not allow sufficient time between return fires for plant recruitment (Morrison et al. 1996). As I found in this thesis that macropod browsing slowed the recovery of vegetation from fire (Paper III), focal browsing after fire may slow plant maturity, and therefore increase the risk of plants being locally extirpated by return fires. This combination of release of macropod herbivores from predation and changing fire regimes is therefore of concern for the conservation of flora and fauna in south-eastern

Australian forests, and warrants further consideration from both researchers and managers.

In this thesis, I have recommended that the management of macropod herbivores be integrated with fire management decisions in forests. Where fire-browsing interactions are considered to be negatively affecting conservation values, forest managers may need to reduce the occurrence or strength of these interactions. Two potential ways managers could reduce the strength of fire-browsing interactions would be to: (i) manage fire to reduce the concentration of herbivores in recently burnt areas (manage the chain interaction), or (ii) target management of the herbivore population to post-fire environments (manage the interaction modification) (Paper

V). The choice between these two strategies will depend on whether the fire-herbivore interaction is being driven by chain or modification effects. In cases such as Booderee National

Park, where fire and herbivory are interacting via both chain effects (fire-driven browsing) and modification effects (herbivore effects being stronger after fire), effective management of the interaction may require a combination of both strategies (see Paper V).

Managing the chain interaction may be an effective way to manage many fire-herbivore interactions, particularly in reserves where high visitor use makes implementing broad-scale macropod control programs difficult. To manage the chain interaction, it may be possible to alter the size, shape or timing of prescribed burns to reduce the concentration of herbivores in burnt patches. However, before this could be implemented effectively, further work is needed to determine how fire influences macropod habitat selection in forests. Previous studies have shown that some small macropods, such as the swamp wallaby, forage less in open areas, even if they contain high quality forage (While and McArthur 2006, Di Stefano et al. 2009, Stutz et al. 2015), and it has also been found that many small savannah herbivores select more strongly for small, rather than large burns (Sensenig et al. 2010). It may therefore be possible to reduce the extent to which small macropods concentrate in burnt areas by increasing the size of individual fires (e.g. burning one larger area, rather than a number of small areas), or reducing the edge:area ratio of burnt areas (so that more of the burnt area is at large distance from the edge).

Another strategy that may be effective in reducing the concentration of herbivores in burnt areas would be to alter the timing of burning, either ensuring that burns are conducted 15 when there is ample fresh pick available elsewhere in the system, or if a number of burns are planned in the same season, ensuring burns are completed close in time, so that herbivores must choose between patches, rather than moving from one freshly burnt patch to another as burns are completed. However, the success of such strategies may be limited in areas such as

Booderee National Park, where terrestrial predators are largely absent, as it is unclear whether swamp wallabies will avoid open areas without the ongoing threat of predation. Changes to prescribed burning must also be considered in the context of broader fire history, as any strategies that result in high fire frequency or short inter-fire intervals could have negative effects on many species (Morrison et al. 1996, Enright and Fontaine 2014). Further study is therefore needed to determine how the size, shape, intensity and/or timing of prescribed burns affect the selection of different macropod species for burnt areas, and whether this varies with the presence or absence of terrestrial predators.

When interaction modification effects are strong, managing for an interaction chain may do little to mitigate interactive effects, and if the strategy includes increasing the size of burns, such management may actually increase interactive effects (see Paper V). Therefore, fire- browsing interactions that are driven by modification effects may be more effectively managed by controlling the herbivore population in recently burnt areas. Targeted population control (e.g. culling) to protect vegetation from browsing in the period immediately following fire may be more efficient than broad scale population control as it may also encourage herbivores to avoid burnt areas (i.e. reduce chain effects) (Cromsigt et al. 2013). Targeted control of macropods in burnt areas may also more closely resemble historic processes in this region, as indigenous

Australians traditionally used fire to attract macropods to areas for easier hunting (Benson et al.

1997, Yibarbuk et al. 2001). Prior to implementing any population control, managers would need to confirm the identity and relative abundance of the macropod species foraging in the burnt habitat, and ensure that control was targeted towards the species contributing to vegetation decline.

In Booderee National Park, management intervention may be required to prevent the simplification of vegetation following prescribed burning. As Booderee National Park attracts 16 high numbers of visitors, managing the fire to reduce focal browsing in burnt areas (managing the interaction chain) may be an effective and achievable alternative to broad scale lethal control of macropods. In the long-term, effective management of the interaction chain requires a better understanding of how the size, shape or intensity of prescribed fires affects site selection by different herbivore species. Until such information is available, managers could consider a number of measures to reduce the likelihood of vegetation simplification after burning:

- Avoid very small sized prescribed burns, particularly in areas of sensitive or high-value

vegetation.

- Consider the timing of burns, and where possible conduct burns when forage is readily

available and widespread within the park.

- Monitor the recovery of vegetation following prescribed burns. If, despite measures to

reduce chain effects, recovery of vegetation is limited (i.e. bracken dominance exceeds

60%, see Paper III), consider short-term measures to deter herbivores from feeding in

burnt areas.

Conclusion

My research has shown that fire and macropod herbivores can interact via both numerically mediated (chain) and functionally moderated (modification) effects to simplify the structure and diversity of forest vegetation, and alter the density and composition of some fauna groups.

Under current management and policy conditions, it is probable that such interactions occur in many of the forests of south-eastern Australia. Large herbivore populations, particularly those that are highly abundant, should therefore be an important consideration for prescribed burn planning in these forests. Future research that determines how long interactive effects persist, and how fire properties and the presence of predators affect macropod site selection should be prioritised to support these management decisions. Overall, this thesis has highlighted the value of a mechanistic understanding of how co-occurring disturbances interact for determining when, where and how disturbance interactions can most effectively and efficiently be managed.

References

Asner, G. P., A. J. Elmore, L. P. Olander, R. E. Martin, and A. T. Harris. 2004. Grazing systems, ecosystem responses, and global change. Annual Review of Environment and Resources 29:261-299. Attiwill, P. M. and M. A. Adams. 2013. Mega-fires, inquiries and politics in the eucalypt forests of Victoria, south-eastern Australia. Forest Ecology and Management 294:45-53. Augustine, D. J. and J. D. Derner. 2015. Patch burn grazing management in a semiarid grassland: Consequences for pronghorn, plains pricklypear, and wind erosion. Rangeland Ecology & Management 68:40-47. Augustine, D. J. and S. J. McNaughton. 1998. Ungulate effects on the functional species composition of plant communities: Herbivore selectivity and plant tolerance. Journal of Wildlife Management 62:1165-1183. Banks, P. B., A. E. Newsome, and C. R. Dickman. 2000. Predation by red foxes limits recruitment in populations of eastern grey kangaroos. Austral Ecology 25:283-291. Benson, J., P. Redpath, and N. Sydney. 1997. The Nature of Pre-European Native Vegetation in South-eastern Australia: A Critique of Ryan, DG, Ryan JR and Starr, BJ (1995), The Australian Landscape: Observations of Explorers and Early Settlers. Royal Botanic Gardens, National Herbarium of New South Wales. Bond, W. J. and J. E. Keeley. 2005. Fire as a global ‘herbivore’: The ecology and evolution of flammable ecosystems. Trends in Ecology & Evolution 20:387-394. Bowman, D. M. J. S., J. K. Balch, P. Artaxo, W. J. Bond, J. M. Carlson, M. A. Cochrane, . . . S. J. Pyne. 2009. Fire in the Earth System. Science 324:481-484. Boyd, P. and C. Brown. 2015. Modes of interactions between environmental drivers and marine biota. Frontiers in Marine Science 2:9. Bulinski, J. and C. McArthur. 1999. An experimental field study of the effects of mammalian herbivore damage on Eucalyptus nitens seedlings. Forest Ecology and Management 113:241-249. Buma, B. 2015. Disturbance interactions: characterization, prediction, and the potential for cascading effects. Ecosphere 6:art70. Catling, P. C., N. Coops, and R. J. Burt. 2001. The distribution and abundance of ground- dwelling mammals in relation to time since wildfire and vegetation structure in south- eastern Australia. Wildlife Research 28:555-565. Clarke, M. F. 2008. Catering for the needs of fauna in fire management: science or just wishful thinking? Wildlife Research 35:385-394. Coates, T. D. 2008. The effect of fox control on mammal populations in an outer urban conservation reserve. Australian Mammalogy 30:51-64.

Coulson, G., C. D. Nave, G. Shaw, and M. B. Renfree. 2008. Long-term efficacy of levonorgestrel implants for fertility control of eastern grey kangaroos (Macropus giganteus). Wildlife Research 35:520-524. Cromsigt, J. P. G. M., D. P. J. Kuijper, M. Adam, R. L. Beschta, M. Churski, A. Eycott, . . . K. West. 2013. Hunting for fear: innovating management of human–wildlife conflicts. Journal of Applied Ecology 50:544-549. Danell, K., R. Bergström, P. Duncan, and J. Pastor. 2006. Large Herbivore Ecology, Ecosystem Dynamics and Conservation. Cambridge University Press, Cambridge. Davis, N. E., G. Coulson, and D. M. Forsyth. 2008. Diets of native and introduced mammalian herbivores in shrub-encroached grassy woodland, south-eastern Australia. Wildlife Research 35:684-694. Dexter, N., M. Hudson, S. James, C. MacGregor, and D. B. Lindenmayer. 2013. Unintended consequences of invasive predator control in an Australian forest: Overabundant wallabies and vegetation change. PloS one 8:e69087. Dexter, N., P. Meek, S. Moore, M. Hudson, and H. Richardson. 2007. Population responses of small and medium sized mammals to fox control at Jervis Bay, southeastern Australia. Pacific Conservation Biology 13:283-292. Dexter, N., D. L. Ramsey, C. MacGregor, and D. Lindenmayer. 2012. Predicting ecosystem wide impacts of wallaby management using a fuzzy cognitive map. Ecosystems 15:1363- 1379. Di Stefano, J. and G. R. Newell. 2008. Diet selection by the swamp wallaby (Wallabia bicolor): Feeding strategies under conditions of changed food availability. Journal of Mammalogy 89:1540-1549. Di Stefano, J., A. York, M. Swan, A. Greenfield, and G. Coulson. 2009. Habitat selection by the swamp wallaby (Wallabia bicolor) in relation to diel period, food and shelter. Austral Ecology 34:143-155. Díaz, S., S. Lavorel, S. U. E. McIntyre, V. Falczuk, F. Casanoves, D. G. Milchunas, . . . B. D. Campbell. 2007. Plant trait responses to grazing – a global synthesis. Global Change Biology 13:313-341. 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. Doherty, T. S., C. R. Dickman, D. G. Nimmo, and E. G. Ritchie. 2015. Multiple threats, or multiplying the threats? Interactions between invasive predators and other ecological disturbances. Biological Conservation 190:60-68. Driscoll, D. A., D. B. Lindenmayer, A. F. Bennett, M. Bode, R. A. Bradstock, G. J. Cary, . . . G. Friend. 2010. Resolving conflicts in fire management using decision theory: asset‐ protection versus biodiversity conservation. Conservation Letters 3:215-223.

Eby, S., D. E. Burkepile, R. W. S. Fynn, C. E. Burns, N. Govender, N. Hagenah, . . . M. D. Smith. 2014. Loss of a large grazer impacts savanna grassland plant communities similarly in North America and South Africa. Oecologia 175:293-303. Enright, N. J. and J. B. Fontaine. 2014. Climate change and the management of fire-prone vegetation in southwest and southeast Australia. Geographical Research 52:34-44. Fuhlendorf, S. D. and D. M. Engle. 2001. Restoring heterogeneity on rangelands: Ecosystem management based on evolutionary grazing patterns. Bioscience 51:625-632. Fuhlendorf, S. D., D. M. Engle, J. A. Y. Kerby, and R. Hamilton. 2009. Pyric herbivory: Rewilding landscapes through the recoupling of fire and grazing. Conservation Biology 23:588-598. Fuhlendorf, S. D., D. E. Townsend, R. D. Elmore, and D. M. Engle. 2010. Pyric-herbivory to promote rangeland heterogeneity: Evidence from small mammal communities. Rangeland Ecology & Management 63:670-678. Gibb, H., N. J. Sanders, R. R. Dunn, S. Watson, M. Photakis, S. Abril, . . . C. L. Parr. 2015. Climate mediates the effects of disturbance on ant assemblage structure. Proceedings of the Royal Society of London B: Biological Sciences 282:20150418. Gibbons, P., L. van Bommel, A. M. Gill, G. J. Cary, D. A. Driscoll, R. A. Bradstock, . . . D. B. Lindenmayer. 2012. Land management practices associated with house loss in wildfires. PloS one 7:e29212. Gordon, I. J., A. J. Hester, and M. Festa-Bianchet. 2004. The management of wild large herbivores to meet economic, conservation and environmental objectives. Journal of Applied Ecology 41:1021-1031. Gowans, S. A., M. S. Gibson, M. E. Westbrooke, and P. Pegler. 2010. Changes in vegetation condition following kangaroo population management in Wyperfeld National Park. Pages 361-370 in G. E. M. Coulson, editor. Macropods: The Biology of Kangaroos, Wallabies and Rat-Kangaroos. CSIRO Publishing, Collingwood, Australia. Hester, A. J., M. Bergman, G. R. Iason, and J. Moen. 2006. Impacts of large herbivores on plant community structure and dynamics. Pages 97-141 in K. Danell, R. Bergström, P. Duncan, and J. Pastor, editors. Large Herbivore Ecology, Ecosystem Dynamics and Conservation. Cambridge University Press, Cambridge. Hidding, B., J.-P. Tremblay, and S. D. Côté. 2013. A large herbivore triggers alternative successional trajectories in the boreal forest. Ecology 94:2852-2860. Holm, J. A., J. R. Thompson, W. J. McShea, and N. A. Bourg. 2013. Interactive effects of chronic deer browsing and canopy gap disturbance on forest successional dynamics. Ecosphere 4:art144. Howland, B., D. Stojanovic, I. J. Gordon, A. D. Manning, D. Fletcher, and D. B. Lindenmayer. 2014. Eaten out of house and home: Impacts of grazing on ground-dwelling reptiles in Australian grasslands and grassy woodlands. PloS one 9:e105966.

Kerns, B. K., M. Buonopane, W. G. Thies, and C. Niwa. 2011. Reintroducing fire into a ponderosa pine forest with and without cattle grazing: understory vegetation response. Ecosphere 2:art59. Kimuyu, D. M., R. L. Sensenig, C. Riginos, K. E. Veblen, and T. P. Young. 2014. Native and domestic browsers and grazers reduce fuels, fire temperatures, and acacia ant mortality in an African savanna. Ecological Applications 24:741-749. Koerner, S. E. and S. L. Collins. 2014. Interactive effects of grazing, drought, and fire on grassland plant communities in North America and South Africa. Ecology 95:98-109. Lindenmayer, D., C. MacGregor, N. Dexter, M. Fortescue, and E. Beaton. 2014. Booderee National Park, The Jewel of Jervis Bay. CSIRO Publishing, Collingwood, Australia. Lindenmayer, D. B., C. MacGregor, A. Welsh, C. Donnelly, M. Crane, D. Michael, . . . A. M. Gill. 2008. Contrasting mammal responses to vegetation type and fire. Wildlife Research 35:395-408. Lunney, D. 2010. A history of the debate (1948-2009) on the commercial harvesting of kangaroos, with particular reference to New South Wales and the role of Gordon Grigg. Australian zoologist 35:383-430. Mantyka-Pringle, C. S., P. Visconti, M. Di Marco, T. G. Martin, C. Rondinini, and J. R. Rhodes. 2015. Climate change modifies risk of global biodiversity loss due to land-cover change. Biological Conservation 187:103-111. Mathisen, K. M., F. Buhtz, K. Danell, R. Bergström, C. Skarpe, O. Suominen, and I. L. Persson. 2010. Moose density and habitat productivity affects reproduction, growth and species composition in field layer vegetation. Journal of Vegetation Science 21:705-716. McArthur, C., P. Banks, R. Boonstra, and J. Forbey. 2014. The dilemma of foraging herbivores: dealing with food and fear. Oecologia 176:677-689. Meers, T. and R. Adams. 2003. The impact of grazing by eastern grey kangaroos (Macropus giganteus) on vegetation recovery after fire at Reef Hills Regional Park, Victoria. Ecological Management & Restoration 4:126-132. Milchunas, D. G. and W. K. Lauenroth. 1993. Quantitative effects of grazing on vegetation and soils over a global range of environments. Ecological Monographs 63:328-366. Morrison, D. A. 2002. Effects of fire intensity on plant species composition of sandstone communities in the Sydney region. Austral Ecology 27:433-441. Morrison, D. A., R. T. Buckney, B. J. Bewick, and G. J. Cary. 1996. Conservation conflicts over burning bush in south-eastern Australia. Biological Conservation 76:167-175. Nimmo, D. G., S. J. Watson, D. M. Forsyth, and C. J. A. Bradshaw. 2015. FORUM: Dingoes can help conserve wildlife and our methods can tell. Journal of Applied Ecology 52:281- 285.

NSW Office of Environment and Heritage. 2013. Living with fire in NSW National Parks: A strategy for managing bushfires in national parks and reserves. Office of Environment and Heritage, Sydney, NSW. Oliver, T. H. and M. D. Morecroft. 2014. Interactions between climate change and land use change on biodiversity: attribution problems, risks, and opportunities. Wiley Interdisciplinary Reviews: Climate Change 5:317-335. Olofsson, J., P. Hulme, L. Oksanen, and O. Suominen. 2005. Effects of mammalian herbivores on revegetation of disturbed areas in the forest-tundra ecotone in northern Fennoscandia. Landscape Ecology 20:351-359. Pedersen, S., H. P. Andreassen, D. A. Keith, C. Skarpe, C. R. Dickman, I. J. Gordon, . . . C. McArthur. 2014. Relationships between native small mammals and native and introduced large herbivores. Austral Ecology 39:236-243. Pekin, B. K., B. A. Endress, M. J. Wisdom, B. J. Naylor, and C. G. Parks. 2014. Impact of ungulate exclusion on understorey succession in relation to forest management in the Intermountain Western United States. Applied Vegetation Science 18:252–260. Royo, A. A. and W. P. Carson. 2006. On the formation of dense understory layers in forests worldwide: consequences and implications for forest dynamics, biodiversity, and succession. Canadian Journal of Forest Research 36:1345-1362. Royo, A. A., R. Collins, M. B. Adams, C. Kirschbaum, and W. P. Carson. 2010a. Pervasive interactions between ungulate browsers and disturbance regimes promote temperate forest herbaceous diversity. Ecology 91:93-105. Royo, A. A., S. L. Stout, D. S. de Calesta, and T. G. Pierson. 2010b. Restoring forest herb communities through landscape-level deer herd reductions: Is recovery limited by legacy effects? Biological Conservation 143:2425-2434. Sensenig, R. L., M. W. Demment, and E. A. Laca. 2010. Allometric scaling predicts preferences for burned patches in a guild of East African grazers. Ecology 91:2898-2907. Sprent, J. A. and C. McArthur. 2002. Diet and diet selection of two species in the macropodid browser-grazer continuum- do they eat what they 'should'? Australian Journal of Zoology 50:183-192. Stephens, S., J. K. Agee, P. Fulé, M. North, W. Romme, T. Swetnam, and M. G. Turner. 2013. Managing forests and fire in changing climates. Science 342:41-42. Stephens, S. L. and L. W. Ruth. 2005. Federal forest-fire policy in the United States. Ecological Applications 15:532-542. Stutz, R. S., P. B. Banks, N. Dexter, and C. McArthur. 2015. Associational refuge in practice: can existing vegetation facilitate woodland restoration? Oikos 124:571-580. Suominen, O. and K. Danell. 2006. Effects of large herbivores on other fauna. Pages 383-412 in K. Danell, R. Bergström, P. Duncan, and J. Pastor, editors. Large Herbivore Ecology, Ecosystem Dynamics and Conservation. Cambridge University Press, Cambridge.

Syphard, A. D., V. C. Radeloff, T. J. Hawbaker, and S. I. Stewart. 2009. Conservation threats due to human-caused increases in fire frequency in Mediterranean-climate ecosystems. Conservation Biology 23:758-769. Tanentzap, A. J., D. R. Bazely, S. Koh, M. Timciska, E. G. Haggith, T. J. Carleton, and D. A. Coomes. 2011. Seeing the forest for the deer: Do reductions in deer-disturbance lead to forest recovery? Biological Conservation 144:376-382. Tanentzap, A. J. and D. A. Coomes. 2012. Carbon storage in terrestrial ecosystems: do browsing and grazing herbivores matter? Biological Reviews 87:72-94. Tanentzap, A. J., K. J. Kirby, and E. Goldberg. 2012. Slow responses of ecosystems to reductions in deer (Cervidae) populations and strategies for achieving recovery. Forest Ecology and Management 264:159-166. Taws, N. 1998. Fire and vegetation management in Jervis Bay Territory, Report. Environment Australia, Canberra. Tuft, K. D., M. S. Crowther, and C. McArthur. 2012. Fire and grazing influence food resources of an endangered rock-wallaby. Wildlife Research 39:436-445. Turner, M. G. 2010. Disturbance and landscape dynamics in a changing world. Ecology 91:2833-2849. Turner, M. G., W. L. Baker, C. J. Peterson, and R. K. Peet. 1998. Factors influencing succession: Lessons from large, infrequent natural disturbances. Ecosystems 1:511-523. Tylianakis, J. M., R. K. Didham, J. Bascompte, and D. A. Wardle. 2008. Global change and species interactions in terrestrial ecosystems. Ecology Letters 11:1351-1363. While, G. M. and C. McArthur. 2006. Distance from cover affects artificial food-patch depletion by macropod herbivores. Wildlife Research 33:565-570. White, P. S. and S. T. A. Pickett. 1985. Natural disturbance and patch dynamics: An introduction. Pages 3-13 in S. T. A. Pickett and P. S. White, editors. The Ecology of Natural Disturbance and Patch Dynamics. Academic Press, New York, USA. Wiggins, N. L. and D. M. J. S. Bowman. 2011. Macropod habitat use and response to management interventions in an agricultural-forest mosaic in north-eastern Tasmania as inferred by scat surveys. Wildlife Research 38:103-113. Wisdom, M. J., M. Vavra, J. M. Boyd, M. A. Hemstrom, A. A. Ager, and B. K. Johnson. 2006. Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildlife Society Bulletin 34:283-292. Yibarbuk, D., P. Whitehead, J. Russell‐Smith, D. Jackson, C. Godjuwa, A. Fisher, . . . D. Bowman. 2001. Fire ecology and Aboriginal land management in central Arnhem Land, northern Australia: a tradition of ecosystem management. Journal of Biogeography 28:325- 343.

Paper I: Effects of large native herbivores on other animals

Numerous studies have highlighted the strong role of large herbivores in shaping ecosystem structure and function. However, there has been little quantitative synthesis of the effect of native herbivores on other animals, despite wild herbivore populations being managed for biodiversity conservation in many ecosystems. In this paper, I consolidated existing knowledge of the effects of large herbivores on other animal groups, analysed the reasons behind current gaps and limitations in the literature, and identified approaches for further research to better inform the management of large herbivores for conservation objectives.

Foster, C.N., Barton, P.S. and D.B. Lindenmayer (2014) The effects of large native herbivores on other animals. Journal of Applied Ecology. 51, 929-938.

Summary

1. Large mammalian herbivores are major drivers of the structure and function of

terrestrial ecosystems worldwide, and changes in their abundance have resulted in many

populations being actively managed. Many empirical studies have identified that

abundant mammalian herbivores can have negative impacts on biodiversity, but there

has been no specific review of the impacts of native mammalian herbivores.

2. We assessed the peer-reviewed literature on the effects of large native herbivores on

other animals. We aimed to quantitatively synthesise current knowledge, identify gaps

and limitations in the literature, and highlight priorities for future research.

3. Most empirical studies of herbivory effects compared only two levels of herbivory

(76%), and meta-analysis showed that very high densities of herbivores, when

compared with very low densities, had mostly negative effects on other animal species.

These negative effects were usually attributed to changes in the quantity and/or

structure of vegetation. Only 24% of papers studied animal responses across a gradient

of herbivore densities; and non-linear responses to herbivory, as well as responses to

low and moderate herbivore densities, remain poorly understood.

4. The literature also was dominated by short-term studies (76% sampled animal responses

for 2 years or less) and there was a high incidence of confounding factors among studies

(38% of studies). In addition, many studies used only coarse metrics to assess effects

(e.g. only 33% of studies assessed species composition) and few included community-

level synthesis (only 31% of studies reported results from more than one animal class).

5. Synthesis and applications. Critical questions remain for both basic ecology and the

management of large native herbivores for biodiversity. Key knowledge gaps include

(1) non-linear responses to herbivore pressure, (2) how responses differ between

different herbivores, (3) the spatial and (4) the temporal variation of responses, (5) how

the effects of herbivores interact with other land management activities, and (6) the

mechanisms driving cascading effects through ecosystems. We identify ways to address

these gaps and emphasise the need for studies which employ contrasts over a gradient of

ecologically relevant herbivore densities and biologically meaningful timeframes.

Keywords

Browsing, biodiversity, ecological cascades, deer, grazing, herbivory, indirect effects, overabundance, species interactions, ungulates

Introduction

Large mammalian herbivores play a central role in the structure and functioning of terrestrial ecosystems (Côté et al. 2004; Danell et al. 2006). Loss of native predators, and changes in land use have led to large herbivores becoming over-abundant in many ecosystems worldwide (Estes et al. 2011; Ripple et al. 2014). In other locations, large herbivore populations are threatened by over-hunting, habitat loss, and habitat fragmentation (Mallon & Zhigang 2009; Festa-Bianchet et al. 2011). Consequently, many species of large herbivores are now actively managed throughout much of their range (Gordon, Hester & Festa-Bianchet 2004). This has resulted in a demand for empirical studies to quantify the effects of large herbivores on ecosystems and biota, and for data to both inform and justify management decisions (e.g. Owen-Smith et al.

2006).

Knowledge of the effects of large herbivores is substantial. Effects include changes to primary production (Frank & McNaughton 1993), nutrient cycling (Molvar, Bowyer &

Ballenberghe 1993), soil properties (Abdel-Magid, Schuman & Hart 1987), fire regimes (Hobbs

1996), plant and animal communities (Goheen et al. 2010), and possibly even climate (Doughty,

Wolf & Field 2010). In this study we focus on the effects of large native herbivores on other animals. The effects of large herbivores on other animals can occur via a range of mechanisms, many of which are indirect. These include consumption of vegetation, modification of the physical environment, provision and re-distribution of resources (e.g. dung, carrion), and direct interference (Schmitz 2008; Marquis 2010). These effects can vary in both temporal and spatial

scale, from short-term impacts on individuals, to landscape-scale shifts in ecosystem state that can be difficult to reverse (Côté et al. 2004; Tanentzap, Kirby & Goldberg 2012). The effects of herbivory on other animals cannot be easily predicted from effects on vegetation as animal responses have been found to differ from plant responses both in shape and magnitude

(Milchunas, Lauenroth & Burke 1998; Wardle et al. 2001; Grandin, Lenoir & Glimskär 2013).

Previous reviews on the effects of large herbivores on other animals have attempted to synthesise these complex effects but have been specific in taxonomic or geographic focus. They include effects on birds (Gill & Fuller 2007), small mammals (Flowerdew & Ellwood 2001), butterflies (Feber et al. 2001) and invertebrates (Stewart 2001), and effects on specific ecosystems such as European heathlands (Rosa García et al. 2013) and the North American arid zone (Jones 2000). Work by Suominen and Danell (2006) took a more holistic approach, providing a qualitative summary of the effects of large herbivores on both invertebrate and vertebrate groups and in a range of ecosystems. However, the volume of literature available on this topic has more than doubled since 2006, which both warrants renewed attention to this topic, and allows the application of quantitative approaches not previously possible.

Furthermore, previous reviews have not distinguished between the effects of native and introduced herbivore species. This is important because the effects of native herbivores may differ from those of introduced species due to their shared evolutionary history with their environment (Milchunas, Sala & Lauenroth 1988; Hobbs 1996). The management goals for native herbivores are also likely to differ, being focussed on establishing an ‘optimum’ herbivore density, rather than on damage limitation or eradication (Côté et al. 2004; Gordon,

Hester & Festa-Bianchet 2004).

We conducted a systematic review and meta-analysis of the published literature on the effects of large native herbivores on other animals. Our review had four principal aims: (1) to synthesise current knowledge on the effects of large native herbivores on other animals, (2) to identify key knowledge gaps or inconsistencies within the literature, (3) to identify the problems which have led to these gaps, and (4) to suggest priorities for future research. While our review focusses on effects on animal assemblages, many of the knowledge gaps and barriers we 28 identify also will be applicable to the study of other effects of large herbivores. While the definition of a ‘large herbivore’ varies between studies, to be consistent with the previous review, we use the definition of Danell et al. (2006), where the term covers all terrestrial mammalian herbivores with an average body mass exceeding 2 kg.

Our quantitative and systematic approach to this review reveals key gaps in our understanding of the effects of large native herbivores, and identifies specific limitations in study design that have led to these gaps. We identify approaches for future research to overcome these limitations and address critical questions for basic ecology and management.

Materials and Methods

We used a systematic approach to obtain a representative sample of the literature on the effects of large herbivores on other animals. Briefly, we used key search terms to identify potentially relevant articles, and selected articles which quantified the effects of large native herbivores on other animals (full details are outlined in Appendix S1 in Supporting Information). This process yielded 96 different studies from 69 different study areas, published between 1983 and 2013.

For details of these papers see Appendix S2 in Supporting Information.

Systematic Review

We used data from these 96 papers to address six specific research questions. Below we briefly outline the methods used to address each of these, for full details see Appendix S1.

1. Which ecosystems and animal response groups are most commonly studied?

To assess geographical and taxonomic differences in study effort, we sorted all papers into categories by two location variables (continent and ecosystem), two focus herbivore variables

(taxonomic group and abundance (as described by authors)), and two animal response group variables (number of groups studied, taxonomic group studied). For each variable, we then tested whether there were differences in the frequency of papers in each category using a Chi- squared analysis. We also sorted papers from the original search to assess whether studies of native herbivores were more or less common than studies of domestic or wild exotic herbivores. 29

2. Are there overall trends in the direction of effects of large herbivores on other animals?

From each experimental study which was not strongly confounded (i.e. quality of evidence score of iii or better as defined by Felton et al. 2010, see Appendix S1), we recorded the abundance, species diversity and species composition responses of animal assemblages to different levels of herbivory. We calculated the mean effect of herbivory for taxonomic groups for which we could extract data from at least five studies. We selected species richness as the measure of species diversity as it was the most commonly used metric. For each taxonomic group in each study, we calculated a log response ratio (highest versus lowest herbivory treatments in study) and its variance (Borenstein et al. 2009, p. 30). We then performed a random effects meta-analysis to calculate mean log response ratios and 95% confidence intervals for each taxonomic group (Borenstein et al. 2009, p. 69). Taxonomic groups were used instead of functional groups as most studies used taxonomic divisions and the infrequent and inconsistent use of functional groups between studies meant sample sizes were small. In addition, where sample sizes were adequate, we found that taxonomic groups were able to provide more detail than functional groups (e.g. Orthoptera and Lepidoptera versus leaf chewing herbivores). We analysed Arthropoda overall, as well as arthropod orders individually, as

Arthropoda were often used as a measure of prey abundance for other taxa.

As species composition is not measurable on a continuous scale, we could not calculate a mean response ratio for species composition. Instead, we assessed whether overall, results for species composition differed from those for species diversity. We examined studies that reported both species diversity (any measure of within-plot species diversity) and species composition responses to herbivory, and compared the numbers of studies reporting a significant effect or no effect for each of these measures. As some studies analysed data from different animal groups separately, we included data from each analysis as a separate data point.

We also assessed whether the effects of large herbivores on other animals could be attributed to any particular mechanism. To do this, for each study, and also for each animal response group in each study, we recorded the mechanisms authors attributed to the observed 30 animal responses. For each response, we also recorded whether the importance of these mechanisms was determined through an explicit test, or through author inference, and the direction of the animal response (positive, negative or ‘other’).

Environmental factors can modify the way organisms respond to herbivory (Wisdom et al. 2006), and the interactions can have stronger effects than the factors alone (Joern & Laws

2013). We therefore recorded the results of all studies investigating the interaction between herbivory and another factor to identify patterns across studies.

3. What is the shape of the animal response to herbivory?

We defined the shape of the herbivory response as the relationship between the animal group response variable (abundance, species diversity etc.) and increasing herbivore density. This could be increasing or decreasing, linear or non-linear etc. We first quantified how many studies were designed to test the shape of the herbivory response. To be counted, a study had to measure the abundance or species diversity of an animal assemblage at a minimum of three different levels of herbivory, and analyse these measurements against a continuous scale of herbivory. Second, we quantified the frequency of non-linear responses to herbivory among the reviewed studies. For each study which measured responses at three or more levels of herbivory, we allocated the response of an animal group to one of the following four categories for both abundance and diversity (any measure of within-plot diversity): (1) linear response, (2) non-linear response, (3) no response/unable to tell from data or (4) no data. For studies which used qualitative categories of herbivore pressure (e.g. low, medium and high) and reported a monotonic response, it was not possible to differentiate between linear and non-linear responses. Therefore, all monotonic responses from categorical studies were assigned to the category ‘unable to tell from data’.

4. Has the design of herbivory studies changed over time?

We assessed the change in experimental approaches over time by examining the design, spatial scale and ‘quality’ of experiments. To assess changes in design, we sorted papers by experimental approach (herbivore exclusion, herbivore density manipulation, simulation of 31

herbivory or natural experiment) and herbivory treatments (binary – two levels of herbivory or gradient – three or more levels of herbivory). We also sorted herbivore exclusion studies according to whether they excluded all large herbivores or selectively excluded particular groups of herbivores. We grouped studies by publication year into five year blocks, and used a

Chi-squared analysis to test whether the frequency of studies across experimental designs differed among time periods.

To assess changes in scale over time, we used the experimental exclosure studies and examined whether, for the same level of replication, the size of exclosures had changed over time. To do this we calculated the mean exclosure size and the number of sampling plots (the number of exclosures per treatment times the number of treatment combinations) in each paper.

We log-transformed both variables to improve linearity. To test for differences over time, we performed a linear regression with groups (with five-year intervals as the groups).

To assess the ‘quality’ of studies over time, we used the quality of evidence categories described by Felton et al. (2010) to classify studies as either not confounded (evidence quality score of i or ii) or confounded (evidence quality score iii or iv). We also grouped studies according to whether they used an experimental or non-experimental approach. We tested whether the frequency of studies with confounding variables had changed over time using a logistic regression in GenStat 15th Edition (VSN International 2012). We allowed interactions between publication year and study approach to allow different slopes to be fitted to experimental and non-experimental studies.

5. Do short-term studies provide a reliable representation of the effect of large herbivores on other fauna?

We assessed whether the duration of sampling of animal responses affected the likelihood of the study detecting a significant effect of herbivory. For each of three variables (abundance, within- plot diversity and composition), we classified the response of each animal group in each study as either ‘significant response to herbivory’ or ‘no response to herbivory’. We conducted a Chi- squared analysis to test whether the frequency of studies detecting a significant result differed

32 between short-term studies (sampling for three years or less, = 1.56, n = 84) and long-term studies (sampling over four or more years, = 6.25, n = 12). We also assessed the prevalence of significant time × herbivory treatment interactions in long-term studies by recording whether or not such an interaction was detected in each study.

6. What do authors recommend as priorities for future research?

One of the principal aims of our review was to identify ways to guide the direction of future research. We therefore recorded research recommendations from each paper.

Results

Most commonly studied ecosystems and response groups

We found significant biases in the geographic location, ecosystem, types of herbivores and animal response groups studied (all P < 0.001, Fig. S1 in Supporting Information). Studies in forested systems of North America and Europe were the most common; with Arthropoda and birds the most frequently studied animal response groups. Our search did not return any studies of the effects of large herbivores on their predators. Only 31% of studies measured the responses of more than one class of animal to herbivory. In 58% of studies, at least one native herbivore was described as over-abundant or at higher than historical levels. In the initial search of the literature, studies on the effects of introduced herbivores (n = 193, domestic = 167, wild =

27) on fauna were twice as common as studies on the effects of native herbivores (n = 99, P <

0.001, Fig. S2 in Supporting Information).

Effects of herbivory on other animals

Meta-analysis revealed significant overall responses to large herbivores for many animal taxa.

The abundance of small mammals, Arthropoda, Araneae, Hemiptera and Lepidoptera were all negatively affected by high levels of herbivory (P < 0.05 for all groups, Fig. 1). The abundance of birds, Coleoptera and Hymenoptera showed no significant overall response to herbivory.

Orthopteran abundance was the only variable showing a positive overall response to herbivory,

however this was not significant (P = 0.21). The effect of herbivory on species richness could be calculated only for birds and Arthropoda and both groups were negatively affected by herbivory (P < 0.001, Fig. 1). Arthropoda richness showed the strongest negative response of all variables, with richness being 66% lower under high than low levels of herbivory.

Figure 1. Abundance (a) and richness (b) responses of animal groups to large native herbivores (ln response ratio and 95% confidence limits). Arthropoda includes only data from studies examining arthropods overall (i.e. not restricted to 2 or 3 focal groups). Data from some studies are included in both Arthropoda and order-level analyses. Response ratio equals abundance (or richness) at high herbivore density / abundance (or richness) at low herbivore density. Values less than zero indicate that the animal group was more abundant (or species rich) at low levels of herbivory. Values with error bars not overlapping zero indicate a significant effect of herbivory (P < 0.05).

The effects of herbivory on the species composition of animal assemblages were assessed in 32 papers (33 %) for 41 animal groups, and effects on both composition and species diversity were analysed for 37 animal groups. Herbivory significantly affected the composition but not the diversity of eight animal groups, while only two animal groups had the reverse result.

Ninety of the 96 reviewed papers discussed mechanisms driving the effects of large herbivores on other animals. Although 84 papers provided conclusions about which mechanisms were likely to have caused the effects they observed, only 48 performed any test of these mechanisms. Of these, 46 confirmed one or more mechanism as important, while only eight reported evidence against a mechanism. Animal responses were most commonly attributed to changes to vegetation biomass and structure (Fig. 2). Most mechanisms were associated with negative effects of large herbivores on other animals, however changes to plant chemistry also commonly associated with positive effects (Fig. 2). Vegetation structure and biomass were the most commonly implicated mechanisms for effects on birds, small mammals, Arthropoda, coleopterans and orthopterans (Table S1 in Supporting Information). Effect on coleopterans were also associated with changes to the physical environment.

Figure 2. The mechanisms that authors considered to be driving the effects of large native herbivores on the (a) abundance and (b) diversity of other animals, and the numbers of animal responses associated with each mechanism. The animal response groups are divided according to whether the mechanism was associated with a negative (black), positive (white) or ‘other’ (grey) response to herbivory.

Nineteen studies (20%) considered how environmental or land management factors interacted with herbivory effects. All three studies that investigated herbivory × productivity interactions reported that herbivory effects were most strongly negative at low productivity

(Suominen et al. 2003; Pringle et al. 2007; Suominen et al. 2008). Recent fire (three papers), addition of logs (two papers), vegetation age (two papers), vegetation density, tree thinning, soil disturbance, nutrient input, rainfall and invasive plants were also found to modify the response to herbivory of at least some animal groups. Studies of fire frequency (two papers) and predator control identified no significant interaction with herbivory.

Shape of the response to herbivory

We found that 23 studies (24%) measured animal response variables at three or more levels of herbivory, but only six of these performed their analyses using herbivore pressure as a quantitative variable. We were able to assess the shape of the response to herbivory for only two of these, and both linear and non-linear responses were reported. Among the 23 studies which measured animal responses to a gradient of herbivore pressures, linear, exponential decay, threshold, intermediate maximum and intermediate minimum response shapes were all reported for at least one animal group. There was no significant difference in the number of animal groups which showed linear and non-linear responses to herbivory for either abundance (P =

0.21) or diversity (P = 0.26).

Experimental approaches

The most common experimental approach was herbivore exclusion, followed by natural experiments (Fig. S3a in Supporting Information). Chi-square comparisons indicated that there was no difference in the distribution of papers across experiment types between time periods (P

= 0.29, Fig. S3b). In contrast, the number of papers using three or more levels of herbivory in their comparisons tended to increase over time; from 10 % of pre-2003 papers to 30% of papers published between 2004 and 2013 (P = 0.03, Fig. S3b). The number of papers including selective exclusion of different herbivores in their design also increased over time (P = 0.017,

Fig. S3c). However, only seven of the 17 papers which had study designs including selective exclusion actually included the selective treatments in their analysis of effects on animals.

The size of exclosures in herbivore exclusion studies varied greatly, from 0.1 m2 cages surrounding individual shrubs to 2100 ha fences encompassing entire watersheds (median =

0.92 ha, IQR = 3.9 ha). There was a significant negative relationship between the log of mean exclosure size and the log of the number of sample plots in a study, with studies employing larger exclosures using lower levels of replication (Fig. 3). This relationship did not differ between time periods, indicating that the scale of exclosure studies has not changed over time.

Figure 3. Relationship between mean exclosure size in a study and the number of study plots (number of replicate exclosures per treatment combination times the number of treatment combinations) in different time periods. The log of the number of sample plots was negatively related to the log of exclosure size (r2 = -0.6, P < 0.001). This relationship did not differ between time periods (P > 0.2).

We identified potential confounding effects in 38% of papers, with 13% of papers confounded to the extent that their data were unusable for meta-analysis (data quality score iv, see Felton et al. 2010). Logistic regression showed that there has been no change in the incidence of confounded studies over time for either experimental or non-experimental studies

(P > 0.05). Experimental studies were far less likely to have problems with confounding factors

(20% of studies) than non-experimental studies (85% of studies, P < 0.001).

Effects of time

Most (76%) papers measured animal responses for two years or less, while only 12% measured animal responses for four years or more. Chi-squared tests showed that there was no difference in the proportion of animal groups showing significant responses to herbivory between short- term (1–3 year) and longer-running studies (P > 0.5 for all variables). However, this analysis had limited power due to studies being heavily skewed across both categories (only 12% ran for more than three years and only 5% found no significant response to herbivory).

Among the 12 studies which measured animal responses for four or more years, five reported a significant year × treatment interaction (Keesing 2000; McShea & Rappole 2000;

Gómez & González-Megías 2007; Mathisen et al. 2012; Parsons, Maron & Martin 2013), and two found significant treatment effects in only some years (Neff, Fettig & VanOverbeke 2007;

Martin & Maron 2012). The remaining five studies did not contain any information on year– treatment interactions.

Priorities for future research

Fifty-one of the 96 papers made explicit recommendations for future research.

Recommendations could be divided into two main categories; those relating to research topic and those relating to study design (Table S2 in Supporting Information). The most frequently suggested topic for future research was the possible cascade of observed herbivory effects to other species or ecosystem processes. Other commonly identified research priorities included studies of the mechanisms of effects, novel habitats or locations, and possible interacting factors. Recommendations regarding the design of future studies were less common but included: studies which measure responses at larger spatial scales and/or examine spatial variation in responses; studies which can detect both direct and indirect relationships; studies which can detect the shape of herbivory responses; and longer-term studies.

Discussion

Our review of the empirical literature has demonstrated that very high herbivore densities usually have negative impacts on other animals. However, there remains only a limited understanding of whether these effects are consistent across taxa (of both herbivores and response animals) or ecosystems, and what factors modify these effects. We also have identified key knowledge gaps around the shape of animal responses to herbivory, the consistency of responses between herbivores, how responses vary with temporal and spatial scale and the mechanisms driving the flow of effects through ecosystems. These gaps are, in part, due to limitations in the design of studies. Below we discuss key findings and research gaps and identify ways to advance future research in this area.

Effects of herbivory

Most animal groups showed a negative overall response to herbivory (Fig. 1). Small mammal abundance and bird species richness were both negatively affected by high levels of herbivory, mostly due to reduced biomass and structural complexity of vegetation and reduction in food resources (Fig. 1, Table S1). This result is consistent with previous reviews (Flowerdew &

Ellwood 2001; Fuller 2001; Suominen & Danell 2006). Overall, bird abundance was not affected by herbivory, which may be due to the small sample size in our analysis. However, the effects of livestock grazing on birds are also mixed (e.g. Milchunas, Lauenroth & Burke 1998) and the difference we found between abundance and diversity effects may reflect a compensatory increase in generalist species under high herbivory (McKinney & Lockwood

1999). Arthropoda, Araneae, Hemiptera and Lepidoptera, also showed negative responses to high herbivory, which was consistent with the summary of Suominen & Danell (2006), and was again attributed to simplified vegetation structure and reduced plant biomass under high herbivory.

Orthopterans (leaf-chewing herbivores), were the only group to show a positive overall response to herbivory, although this effect was not significant. This result differed from the other herbivorous analysed (Hemiptera - sap suckers and Lepidoptera - leaf chewers as larvae),

which responded negatively to the reduced vegetation biomass at high levels of herbivory. The different responses may have occurred because studies of Orthoptera were predominantly from grassland and savanna habitats. In these systems, although grazing reduces grass biomass, it also can increase nutrient turnover and improve the nutritional quality of food available to grasshoppers, while also increasing the heterogeneity of the grass canopy, which assists with predator evasion and thermoregulation (Joern 2004).

We found that Arthropoda richness displayed the strongest negative response to herbivory (Fig. 1), which differs from Suominen and Danell’s (2006) summary that richness is often higher in grazed sites. As only three of the five studies in our analysis corrected richness for abundance this strong negative response may be a slight over-estimate (Gotelli & Colwell

2001). However, it does indicate that effects on richness, at least at this coarse taxonomic level, tended to be negative rather than positive. It is somewhat surprising that such an ecologically diverse group as Arthropoda showed such a strong negative response, but this may partly be explained by the nature of the comparisons in the studies we analysed. Most studies we analysed were exclosure studies, where very high levels of herbivory were compared with very low or no herbivory (Figs. S1, S3). As herbivore density increases, it will almost invariably reach a point where effects on other species become negative, even if effects were positive or negligible at lower densities (Milchunas, Sala & Lauenroth 1988; Suominen & Danell 2006).

By comparing two extreme (and unnatural) herbivore densities, exclosure studies in areas of high herbivore density are therefore pre-positioned to detect negative effects of large herbivores on other animals. At lower levels of herbivory, animal responses may be more varied, both between sites and between animal groups.

Shape of animal responses to herbivory

A wide range of ‘shapes’ of animal responses to herbivory were reported, which reflects the wide range of ecosystems, animal groups and herbivores included in this review. Both grazing theory, and papers we reviewed, indicate that site-specific variables such as productivity

(Pringle et al. 2007; Vesterlund et al. 2012), history of herbivory (Milchunas, Sala & Lauenroth

1988), and the presence of other disturbances (Wisdom et al. 2006), can change the shape, direction and magnitude of animal responses to herbivory. The variation caused by these factors is in addition to the variation in responses between animal response groups and between different herbivores (Pringle et al. 2011).

These variable results highlight the importance of not assuming a linear response to herbivory and caution against extrapolation (or interpolation) from studies which include only binary comparisons. If the relationship between herbivore pressure and an animal assemblage were non-linear, a binary study could detect a positive, negative or no response to herbivory simply depending on which two herbivore pressures were selected for comparison (Rooney &

Waller 2003)(Fig. 4). The use of binary studies to inform the management of large herbivore populations could therefore lead to unexpected and undesirable outcomes. For example, reduction of a highly abundant herbivore population intended to benefit other animals could have no effect, or even a deleterious effect on the target animals, depending on the underlying shape of the herbivory response (see Fig. 4).

Figure 4. The effect of non-linear responses to herbivory on the outcome of experiments including only binary comparisons (74% of reviewed studies). If the ‘true’ response of an animal group to increasing herbivore pressure was unimodal (bold line), a study including only binary comparisons would fail to detect this, and depending on the two herbivore pressures compared, could conclude that an animal was positively (e.g. a vs b), negatively (e.g. b vs c) or not affected (e.g. a vs b) by increasing herbivore pressure. A binary study would also fail to detect a threshold-shaped response (dotted line) and might instead report no effect (e.g. i vs ii) or a negative effect (ii vs iii).

Experimental Approaches

The recent increase in the number of studies using a gradient of herbivore densities indicates that researchers are recognising the limitations of simple binary designs. However, non-binary studies are still in the minority, and animal responses are rarely analysed against a quantitative gradient of herbivory. Studies of functional differences between herbivores are also increasing, but many papers present data only from the two extreme treatments and therefore do not use these studies to their full potential. Further progress in isolating the effects of herbivory from

42 confounding factors is also essential if studies are to detect the ‘true’ response of animals to herbivory.

Due to logistical constraints, experimental studies were usually limited in either spatial scale or replication (Fig. 3). A trade-off is made between gaining detailed information at small spatial scales, or coarser information at large spatial scales. As a result, most studies did not integrate considerations of spatial scale into their design, although a few notable studies found important scale-dependent results (Jay-Robert et al. 2008; Barton et al. 2011). As the effects of herbivory vary with spatial scale (Chaneton & Facelli 1991; Tahmasebi Kohyani et al. 2008), contradictory results can simply be an artefact of the scale of measurements; small-scale responses do not necessary multiply up to similar effects at larger spatial scales (Hester et al.

2006). While understandable, the bias of experimental studies towards low replication at large scales and high replication at small scales means that they are unlikely to detect landscape-scale changes in diversity such as biotic homogenisation (Olden 2006).

The need for long-term studies

Most studies measured animal responses to herbivory over short timeframes of two years or less. However, such studies may give a poor estimate of the true response of an animal assemblage to herbivory, even if herbivory treatments have been in place for a number of years.

Animal populations can fluctuate greatly within short time periods, particularly in short-lived organisms or boom–bust ecosystems (Joern & Pruess 1986; Ernest, Brown & Parmenter 2000;

Ostfeld & Keesing 2000). Short-term studies, especially those with few sampling events, can therefore not be reliably extrapolated to the multi-year times scales of interest to land managers

(Lindenmayer et al. 2012). This is demonstrated by the high incidence of significant time × herbivory interactions in long-running studies, where the effect of herbivory was significant only in some years, and some studies even returned opposite results between years. Although long-term studies are more costly and difficult to conduct, they are necessary if we are to detect and distinguish between short-term populations fluctuations, which may require little

management intervention, and major shifts in ecosystem state which justify significant investment of conservation resources (Andersen et al. 2009; Lindenmayer et al. 2012).

Implications for future work

We have identified key limitations in the design of existing studies that have led to knowledge gaps in the effects of large herbivores on other animals. These gaps include; (1) the shape of animal responses to increasing herbivore pressure, (2) how animal responses differ between different herbivores and the (3) spatial and (4) temporal variation in animal responses (Table 1).

How these effects interact with other land management activities (5), and the mechanisms by which effects cascade to other biota and ecosystem processes (6) also remain key areas for future research. To address these gaps, study designs need to move beyond the traditional “lots or none” herbivore exclusion approach, and employ collaborative and creative study designs

(see Table 1).

Table 1. Knowledge gaps surrounding the effects of large native herbivores on other animals and suggested approaches to address these gaps.

To understand the shape of animal responses to herbivory, studies need to include a gradient of ecologically relevant herbivore pressures and analyse animal responses against herbivory as a continuous variable (Hester et al. 2000). In systems where more than one species of large herbivore is active, studies which measure animal responses to different herbivores are 44 also critical for understanding the functioning of these systems (Wisdom et al. 2006; Pringle et al. 2011). Combining experimental approaches with natural experiments (Hester et al. 2000), and using experimental treatments to their full potential will go a long way to addressing these gaps (Table 1). Considerable potential also lies in collaborations with land managers to measure responses across different areas of active management and establish adaptive management studies (Wisdom et al. 2006; Westgate, Likens & Lindenmayer 2013).

An understanding of the spatial and temporal variation in the effects of large herbivores is essential for the efficient, targeted and timely management of large herbivore populations

(Gordon, Hester & Festa-Bianchet 2004; Wisdom et al. 2006). To detect this variation, comparisons need to extend further than simple between-plot comparisons and employ sampling regimes over biologically meaningful timeframes. Again, combining experimental approaches with natural experiments and replicating experiments over different land management practices or environmental gradients (e.g. Goheen et al. 2013) will be useful approaches (Table 1).

Metrics which examine changes in species composition and species turnover across landscapes can also provide larger-scale insights (e.g. Newman, Mitchell & Kelly 2013), without great increases in sampling effort above what is required for between-plot diversity comparisons. As identified by previous authors (Table S2), multi-taxa studies able to test the mechanisms which drive effects, and how effects cascade through ecosystems, are essential if we are to thoroughly understand the processes underpinning the effects of large herbivores on other animals.

Conclusion

Studies of the effects of large native herbivores on other animals were biased towards short- term, binary comparisons at limited spatial scales. These studies revealed mostly negative effects of high densities of large native herbivores on bird, small mammal and arthropod assemblages. To advance this field, future studies must address key knowledge gaps surrounding non-linear effects, differences in effects between herbivores, spatial and temporal variation, interacting factors and the mechanisms driving effects. These studies must give

careful consideration to study design if they are to give new insights for both basic ecology and management.

Acknowledgements

We thank Peter Lane for statistical advice, Chloe Sato for assistance with figures and the associate editor, Otso Suominen, Benjamin Scheele and an anonymous reviewer for comments on earlier versions of this manuscript.

References

Abdel-Magid, A.H., Schuman, G.E. & Hart, R.H. (1987) Soil bulk density and water infiltration as affected by grazing systems. Journal of Range Management, 40, 307-309. Andersen, T., Carstensen, J., Hernández-García, E. & Duarte, C.M. (2009) Ecological thresholds and regime shifts: Approaches to identification. Trends in Ecology & Evolution, 24, 49-57. Barton, P.S., Manning, A.D., Gibb, H., Wood, J.T., Lindenmayer, D.B. & Cunningham, S.A. (2011) Experimental reduction of native vertebrate grazing and addition of logs benefit beetle diversity at multiple scales. Journal of Applied Ecology, 48, 943-951. Borenstein, M., Hedges, L.V., Higgins, J.P.T. & Rothstein, H.R. (2009) Introduction to Meta- Analysis. John Wiley & Sons Ltd, West Sussex. Chaneton, E.J. & Facelli, J.M. (1991) Disturbance effects on plant community diversity: spatial scales and dominance hierarchies. Vegetation, 93, 143-155. Côté, S.D., Rooney, T.P., Tremblay, J.P., Dussault, C. & Waller, D.M. (2004) Ecological impacts of deer overabundance. Annual Review of Ecology Evolution and Systematics, 35, 113-147. Danell, K., Bergström, R., Duncan, P. & Pastor, J. (2006) Large Herbivore Ecology, Ecosystem Dynamics and Conservation. Cambridge University Press, Cambridge. Doughty, C.E., Wolf, A. & Field, C.B. (2010) Biophysical feedbacks between the Pleistocene megafauna extinction and climate: The first human-induced global warming? Geophysical Research Letters, 37, L15703. Ernest, S.K.M., Brown, J.H. & Parmenter, R.R. (2000) Rodents, plants, and precipitation: spatial and temporal dynamics of consumers and resources. Oikos, 88, 470-482. Estes, J.A., Terborgh, J., Brashares, J.S., Power, M.E., Berger, J., Bond, W.J., Carpenter, S.R., Essington, T.E., Holt, R.D., Jackson, J.B.C., Marquis, R.J., Oksanen, L., Oksanen, T., Paine, R.T., Pikitch, E.K., Ripple, W.J., Sandin, S.A., Scheffer, M., Schoener, T.W.,

Shurin, J.B., Sinclair, A.R.E., Soulé, M.E., Virtanen, R. & Wardle, D.A. (2011) Trophic downgrading of planet Earth. Science, 333, 301-306. Feber, R.E., Brereton, T.M., Warren, M.S. & Oates, M. (2001) The impacts of deer on woodland butterflies: the good, the bad and the complex. Forestry, 74, 271-276. Felton, A., Knight, E., Wood, J., Zammit, C. & Lindenmayer, D. (2010) A meta-analysis of fauna and flora species richness and abundance in plantations and pasture lands. Biological Conservation, 143, 545-554. Festa-Bianchet, M., Ray, J.C., Boutin, S., Côté, S. & Gunn, A. (2011) Conservation of caribou (Rangifer tarandus) in Canada: An uncertain future. Canadian Journal of Zoology-Revue Canadienne De Zoologie, 89, 419-434. Flowerdew, J.R. & Ellwood, S.A. (2001) Impacts of woodland deer on small mammal ecology. Forestry, 74, 277-287. Frank, D. & McNaughton, S. (1993) Evidence for the promotion of aboveground grassland production by native large herbivores in Yellowstone National Park. Oecologia, 96, 157- 161. Fuller, R.J. (2001) Responses of woodland birds to increasing numbers of deer: a review of evidence and mechanisms. Forestry, 74, 289-298. Gill, R.M.A. & Fuller, R.J. (2007) The effects of deer browsing on woodland structure and songbirds in lowland Britain. Ibis, 149, 119-127. Goheen, J.R., Palmer, T.M., Charles, G.K., Helgen, K.M., Kinyua, S.N., Maclean, J.E., Turner, B.L., Young, H.S. & Pringle, R.M. (2013) Piecewise disassembly of a large-herbivore community across a rainfall gradient: The UHURU experiment. PloS one, 8, e55192. Goheen, J.R., Palmer, T.M., Keesing, F., Riginos, C. & Young, T.P. (2010) Large herbivores facilitate savanna tree establishment via diverse and indirect pathways. Journal of Animal Ecology, 79, 372-382. Gómez, J. & González-Megías, A. (2007) Long-term effects of ungulates on phytophagous insects. Ecological Entomology, 32, 229-234. Gordon, I.J., Hester, A.J. & Festa-Bianchet, M. (2004) The management of wild large herbivores to meet economic, conservation and environmental objectives. Journal of Applied Ecology, 41, 1021-1031. Gotelli, N.J. & Colwell, R.K. (2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters, 4, 379-391. Grandin, U., Lenoir, L. & Glimskär, A. (2013) Are restricted species checklists or ant communities useful for assessing plant community composition and biodiversity in grazed pastures? Biodiversity and Conservation, 22, 1415-1434. Hester, A.J., Bergman, M., Iason, G.R. & Moen, J. (2006) Impacts of large herbivores on plant community structure and dynamics. Large Herbivore Ecology, Ecosystem Dynamics and

Conservation (eds K. Danell, R. Bergström, P. Duncan & J. Pastor), pp. 97-141. Cambridge University Press, Cambridge. Hester, A.J., Edenius, L., Buttenschon, R.M. & Kuiters, A.T. (2000) Interactions between forests and herbivores: The role of controlled grazing experiments. Forestry, 73, 381-391. Hobbs, N.T. (1996) Modification of ecosystems by ungulates. The Journal of Wildlife Management, 60, 695-713. Jay-Robert, P., Niogret, J., Errouissi, F., Labarussias, M., Paoletti, E., Luis, M.V. & Lumaret, J.- P. (2008) Relative efficiency of extensive grazing vs. wild ungulates management for dung beetle conservation in a heterogeneous landscape from Southern Europe (Scarabaeinae, Aphodiinae, Geotrupinae). Biological Conservation, 141, 2879-2887. Joern, A. (2004) Variation in grasshopper (acrididae) densities in response to fire frequency and bison grazing in tallgrass prairie. Environmental Entomology, 33, 1617-1625. Joern, A. & Laws, A.N. (2013) Ecological mechanisms underlying arthropod species diversity in grasslands. Annual Review of Entomology, 58, 19-36. Joern, A. & Pruess, K.P. (1986) Temporal constancy in grasshopper assemblies (Orthoptera: Acrididae). Ecological Entomology, 11, 379-385. Jones, A. (2000) Effects of cattle grazing on North American arid ecosystems: A quantitative review. Western North American Naturalist, 60, 155-164. Keesing, F. (2000) Cryptic consumers and the ecology of an African savanna. Bioscience, 50, 205-215. Lindenmayer, D.B., Likens, G.E., Andersen, A., Bowman, D., Bull, C.M., Burns, E., Dickman, C.R., Hoffmann, A.A., Keith, D.A., Liddell, M.J., Lowe, A.J., Metcalfe, D.J., Phinn, S.R., Russell-Smith, J., Thurgate, N. & Wardle, G.M. (2012) Value of long-term ecological studies. Austral Ecology, 37, 745-757. Mallon, D.P. & Zhigang, J. (2009) Grazers on the plains: Challenges and prospects for large herbivores in Central Asia. Journal of Applied Ecology, 46, 516-519. Marquis, R.J. (2010) The role of herbivores in terrestrial trophic cascades. Trophic Cascades: Predators, Prey and the Changing Dynamics of Nature (eds J. Terborgh & J.A. Estes), pp. 109-123. Island Press, Washington, DC. Martin, T.E. & Maron, J.L. (2012) Climate impacts on bird and plant communities from altered animal-plant interactions. Nature Clim. Change, 2, 195-200. Mathisen, K.M., Pedersen, S., Nilsen, E.B. & Skarpe, C. (2012) Contrasting responses of two passerine bird species to moose browsing. European Journal of Wildlife Research, 58, 535- 547. McKinney, M.L. & Lockwood, J.L. (1999) Biotic homogenization: A few winners replacing many losers in the next mass extinction. Trends in Ecology & Evolution, 14, 450-453.

McShea, W.J. & Rappole, J.H. (2000) Managing the abundance and diversity of breeding bird populations through manipulation of deer populations. Conservation Biology, 14, 1161- 1170. Milchunas, D.G., Lauenroth, W.K. & Burke, I.C. (1998) Livestock grazing: Animal and plant biodiversity of shortgrass steppe and the relationship to ecosystem function. Oikos, 83, 65- 74. Milchunas, D.G., Sala, O.E. & Lauenroth, W.K. (1988) A generalized model of the effects of grazing by large herbivores on grassland community structure. The American Naturalist, 132, 87-106. Molvar, E., Bowyer, R.T. & Ballenberghe, V. (1993) Moose herbivory, browse quality, and nutrient cycling in an Alaskan treeline community. Oecologia, 94, 472-479. Neff, P.K.K., Fettig, S.M. & VanOverbeke, D.R. (2007) Variable response of butterflies and vegetation to elk herbivory: An exclosure experiment in ponderosa pine and aspen-mixed conifer forests. Southwestern Naturalist, 52, 1-14. Newman, M., Mitchell, F.J.G. & Kelly, D.L. (2013) Exclusion of large herbivores: Long-term changes within the plant community. Forest Ecology and Management, 321, 136-144. Olden, J.D. (2006) Biotic homogenization: a new research agenda for conservation biogeography. Journal of Biogeography, 33, 2027-2039. Ostfeld, R.S. & Keesing, F. (2000) Pulsed resources and community dynamics of consumers in terrestrial ecosystems. Trends in Ecology & Evolution, 15, 232-237. Owen-Smith, N., Kerley, G.I.H., Page, B., Slotow, R. & van Aarde, R.J. (2006) A scientific perspective on the management of elephants in the Kruger National Park and elsewhere. South African Journal of Science, 102, 389-394. Parsons, E.W.R., Maron, J.L. & Martin, T.E. (2013) Elk herbivory alters small mammal assemblages in high-elevation drainages. Journal of Animal Ecology, 82, 459-467. Pringle, R.M., Palmer, T.M., Goheen, J.R., McCauley, D.J. & Keesing, F. (2011) Ecological importance of large herbivores in the Ewaso ecosystem. Smithsonian Contributions to Zoology, 632, 44-53. Pringle, R.M., Young, T.P., Rubenstein, D.I. & McCauley, D.J. (2007) Herbivore-initiated interaction cascades and their modulation by productivity in an African savanna. Proceedings of the National Academy of Sciences of the United States of America, 104, 193-197. Ripple, W.J., Estes, J.A., Beschta, R.L., Wilmers, C.C., Ritchie, E.G., Hebblewhite, M., Berger, J., Elmhagen, B., Letnic, M., Nelson, M.P., Schmitz, O.J., Smith, D.W., Wallach, A.D. & Wirsing, A.J. (2014) Status and ecological effects of the world’s largest carnivores. Science, 343, 151. Rooney, T.P. & Waller, D.M. (2003) Direct and indirect effects of white-tailed deer in forest ecosystems. Forest Ecology and Management, 181, 165-176.

Rosa García, R., Fraser, M., Celaya, R., Ferreira, L., García, U. & Osoro, K. (2013) Grazing land management and biodiversity in the Atlantic European heathlands: A review. Agroforestry Systems, 87, 19-43. Schmitz, O.J. (2008) Herbivory from individuals to ecosystems. Annual Review of Ecology Evolution and Systematics, 39, 133-152. Stewart, A.J.A. (2001) The impact of deer on lowland woodland invertebrates: A review of the evidence and priorities for future research. Forestry, 74, 259-270. Suominen, O. & Danell, K. (2006) Effects of large herbivores on other fauna. Large Herbivore Ecology, Ecosystem Dynamics and Conservation (eds K. Danell, R. Bergström, P. Duncan & J. Pastor), pp. 383-412. Cambridge University Press, Cambridge. Suominen, O., Niemela, J., Martikainen, P., Niemela, P. & Kojola, I. (2003) Impact of reindeer grazing on ground-dwelling Carabidae and Curculionidae assemblages in Lapland. Ecography, 26, 503-513. Suominen, O., Persson, I.L., Danell, K., Bergström, R. & Pastor, J. (2008) Impact of simulated moose densities on abundance and richness of vegetation, herbivorous and predatory arthropods along a productivity gradient. Ecography, 31, 636-645. Tahmasebi Kohyani, P., Bossuyt, B., Bonte, D. & Hoffmann, M. (2008) Grazing as a management tool in dune grasslands: Evidence of soil and scale dependence of the effect of large herbivores on plant diversity. Biological Conservation, 141, 1687-1694. Tanentzap, A.J., Kirby, K.J. & Goldberg, E. (2012) Slow responses of ecosystems to reductions in deer (Cervidae) populations and strategies for achieving recovery. Forest Ecology and Management, 264, 159-166. Vesterlund, S.-R., Suominen, O., Bergström, R., Danell, K. & Persson, I.-L. (2012) The impact of simulated moose densities on conifer aphids along a productivity gradient. Ecography, 35, 105-112. VSN International (2012) GenStat for Windows 15th Edition. VSN International, Hemel Hempstead. Wardle, D.A., Barker, G.M., Yeates, G.W., Bonner, K.I. & Ghani, A. (2001) Introduced mammals in New Zealand natural forests: aboveground and belowground consequences. Ecological Monographs, 71, 587-614. Westgate, M.J., Likens, G.E. & Lindenmayer, D.B. (2013) Adaptive management of biological systems: A review. Biological Conservation, 158, 128-139. Wisdom, M.J., Vavra, M., Boyd, J.M., Hemstrom, M.A., Ager, A.A. & Johnson, B.K. (2006) Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildlife Society Bulletin, 34, 283-292.

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Appendix S1. Detailed methods for literature search and meta-analysis.

Appendix S2. Details of the 96 papers included in the systematic review.

Table S1. Mechanisms driving the effects of large herbivores on other animals.

Table S2. Recommendations for future research given by authors.

Figure S1. Distribution of reviewed papers by study system and animal response groups.

Figure S2. Numbers of studies of effects of herbivores on other animals where herbivores are native, wild exotic or domestic livestock.

Figure S3. The distribution of experimental approaches, contrasts and exclusion method used by studies.

Appendix S1. Detailed methods for literature search and meta-analysis.

Literature Search and Paper Selection

We used a four stage systematic approach to obtain a representative sample of the literature on the effects of large herbivores on other animals.

1. We searched the ISI Web of Knowledge Database with the following search string:

Topic=((Herbivor* OR graz* OR brows*) AND (Mammal* OR vertebrate* OR

overabundan* OR density OR deer OR ungulate*) AND (Fauna* OR reptil* OR herp*

OR frog* OR amphibian* OR bird* OR avifauna* OR “small mammal*” OR rodent*

OR invertebrate* OR insect*)). This search returned approximately 9500 results.

2. We reviewed the first 100 of these articles to identify additional key words that were

common to the titles of relevant papers. This enabled us to add an additional criteria

based on the article title: Title=(brows* OR herbivor* OR graz* OR deer OR cascade*

OR trophic). We performed the final search using both of these search strings on the

15th August 2013 and obtained 1453 unique matches.

3. We assessed each article for relevance using according to seven criteria. The criteria

were applied hierarchically, with articles only being assessed if they met all of the

preceding criteria.

4. Criteria:

i. Studied the effect of a native mammalian herbivore on fauna (268 articles)

ii. Empirical study published in English in peer reviewed journal (260 articles)

iii. At least one of the herbivores was native to the study system (91 articles)

iv. At least one of the native herbivores was large (> 2 kg) (85 articles)

v. Study was from a natural (or minimally modified) terrestrial ecosystem (76

articles).

vi. Study measured the impact of herbivory on native terrestrial fauna (70 articles)

vii. Study included contrasts (i.e. compared at least two levels of herbivory of

native herbivores) (67 articles).

5. Finally, we searched the reference list of the 67 articles identified above and located a

further 29 papers which met the seven inclusion criteria.

Classification of Studies

To assess biases in the literature we classified studies according to their study location and herbivores and animal response groups. We classified studies by six variables, each with a number of categories:

i. Continent: North America / South America / Asia / Europe / Africa / Oceania.

ii. Ecosystem: Forest or Woodland / Heath or Shrubland / Savanna or Grassland

iii. Taxonomic group of herbivores: at order level

iv. Abundance of herbivores (as described by author): over-abundant (or at higher than

historical levels) / not over-abundant

v. Number of animal response groups studied: single species / single taxonomic group

(class level) / multiple taxonomic groups

vi. Taxonomic group of animal response groups: class for vertebrates, order for

invertebrates.

We also returned to data collected in stage three of the paper selection process to assess whether studies of the effects of native herbivores were less common than studies of introduced herbivores. We sorted the 268 articles which studied the effects of herbivores on fauna according to whether the herbivores studied were native, wild exotic or livestock (step 3iii). In both of the above processes, if a paper fit into more than one category (e.g. studied both heath and forest habitats) then it was counted in both categories. Multiple papers from the same study site were included separately in this analysis as this was thought to summarise research effort more accurately than if we pooled papers by study site.

Data Extraction and Conversion

For meta-analysis we extracted data from tables, figures and supplementary material on the abundance and species richness of animal groups under different levels of herbivory. Animal groups were taxonomic and grouped by class for vertebrates, order for invertebrates. As many studies did not separate arthropods by order, we also analysed arthropods overall which included any analysis of arthropods where data from three or more orders were combined. As we were interested in community level effects, we excluded data which reported effects on only one or two species (except for the effect of herbivory calculation for small mammals, where studies of one or two species were the most common and so included in the analysis). Data in figures were extracted using the web-based application Web Plot Digitiser V2.5 (Rohatgi 2012).

Papers reported data in a variety of formats and we converted all data to means and standard deviations according to the methods in Table 1. If a study did not include all the information needed to calculate a mean and standard deviation for each level of herbivory then the study could not be included in the analysis. All data conversions were conducted before data were combined. To reduce unexplained variation within the analysis we assessed studies according to the quality of evidence criteria of Felton et al. (2010) and included only experimental studies with a data quality score of iii or better (Table 2).

Studies with > 2 levels of herbivore pressure

In the calculation of mean response ratios we used data from the lowest and highest levels of herbivory compared. This was thought to give the most consistent data across studies as most studies with only two levels of herbivory compared absence (or very low) with very high densities.

Combining Data

A number of studies had split data across years, sites, factorial treatments or taxonomic or functional groups, however for our analysis we needed to generate a single data point per study.

For the purpose of analysis, a study was defined as “a project employing the same treatments at the same geographical location measuring the same animal response group”. By this definition some papers produced multiple data points while in other instances data from multiple papers were combined to give a single data point. We used the following methods to combine data:

Data split by taxonomic or functional group:

Before calculating the response ratio and variation we summed the means to obtain an overall mean. To calculate the summed standard deviation we used the rule: The variance of the sum = the sum of the variances.

Data split by sites:

If the sites were split according to habitat type (e.g. both grassland and forest sites were studied and each habitat was analysed separately), then each habitat type was included as its own data point (as if they were separate studies). If sites were split by criteria other than broad habitat types (e.g. landscape position) we combined data from the different categories using the methods described by Borenstein et al. (2009) for ‘multiple outcomes or time points within a study’, treating the categories as ‘outcomes’. As we did not know the level of correlation between among categories, we took the conservative approach and assumed a correlation coefficient of 1 (which overestimates the variance). As a trial of the same calculations with a correlation coefficient of 0 and 0.5 had little effect on the variation estimates, we believe this assumption had little effect on the final outcome of the analysis.

Data split by a factorial treatment (e.g. in a fire x herbivory factorial design):

We attempted to produce data which was as consistent as possible to studies without factorial treatments. If the factorial treatments were with/without a second factor we used only data from sites without the factor. If this was not possible we then assessed whether the factor affected

herbivore pressure. If it did, we treated the study as if it had > 2 levels of herbivore pressure (as described above). Failing either of these methods we combined data from across factors using the same approach outlines above for data split by sites.

Data split by year or sampling events:

We combined data across time periods using the methods described by Borenstein et al. (2009) for ‘multiple outcomes or time points within a study’. We again assumed a correlation coefficient between time points of 1.

Weighting of Studies within Meta-Analysis

Standard practice when conducting a random-effects meta-analysis is to weight each score by the inverse of its variance, so that more precise scores are given more weight in the analysis

(Borenstein et al. 2009). In addition to this, we also weighted all data points by the square root of the number of years of sampling so that long-running studies were given more weight.

References

Borenstein, M., Hedges, L.V., Higgins, J.P.T. & Rothstein, H.R. (2009) Introduction to Meta- Analysis. John Wiley & Sons Ltd, West Sussex. Felton, A., Knight, E., Wood, J., Zammit, C. & Lindenmayer, D. (2010) A meta-analysis of fauna and flora species richness and abundance in plantations and pasture lands. Biological Conservation, 143, 545-554. Rohatgi, A. (2012) Web Plot Digitiser v2.5. Available from [http://arohatgi.info/WebPlotDigitizer/app/].

Table 1. Methods used to convert different types of data to means and standard errors. Data formats not included below were unable to be converted to means and standard errors so were not included in the analysis. Original Format of Conversion to Mean Original Format of Variation Conversion to SD Abundance or Richness Measure Measure Population estimate Taken as mean Standard Error (SE) SD = SE * sqrt (n) 95% Confidence Interval (CI) SE = CI (1.96*2)

Geometric mean Taken as mean Geometric 95% CI SE = CI (1.96*2)

Median Taken as mean Interquartile Range (IQR) SD = (IQR/1.34)

Median (log scale - ln(m)) Mean = exp((ln(m)+(s2/2)) IQR (ln scale) s(ln scale) = (IQR/1.34) SD = mean*sqrt(exp(s2)-1) P value of t-test SE = (diff in means / t @ (p/2, df)) * (sqrt(n)/sqrt(2))

Raw data Calculate mean Raw data Calculate SD

Table 2. Categories used to assess quality of evidence in reviewed papers. From Felton et al. 2010.

Category Quality of evidence presented i Randomized controlled trial with matched pairs of treatments and controls. Study conducted at an adequate scale for subject taxa ii Controlled trial of adequate scale for study organism. Unpaired treatments and controls iii Unpaired treatments and controls. Scale of study raises potential of confounding effects for the subject taxa considered iv Evidence deemed inadequate due to inherent problems with experimental design

Appendix S2. Details of the 96 papers included in the systematic review.

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Bailey & 2002 North Forest / Cervus Herbivore gradient i Fire 1 Single class Arthropods - Overall Whitham America Woodland canadensis Exclusion, Simulation of Herbivory Bailey & 2003 North Forest / Cervus Herbivore binary i none 2 Multiple Birds Whitham America Woodland canadensis Exclusion, classes Arthropods - Overall, Simulation of Phytophagous Insects Herbivory Bailey & 2006 North Forest / Castor Natural binary iv none 2 Single class Phytophagous Insects Whitham America Woodland canadensis Experiment Baines 1996 Europe Shrubland / Cervus Natural binary ii Predator 1 Multiple Birds Heath elaphus, Experiment Control classes Araneae, Coleoptera, Sheep Diptera, Hymenoptera, Hemiptera, Lepidoptera Baines, 1994 Europe Mixed Cervus Herbivore binary i none 2 Multiple Araneae, Coleoptera, Sage & Habitats elaphus Exclusion classes Diptera, Hemiptera, Baines Hymenoptera, Lepidoptera, Plecoptera Barrett & 2007 North Forest / Odocoileus Natural gradient iii Nutrients 1 Single class Phytophagous Insects Stiling America Woodland virginianus Experiment Barton et 2011 Oceania Grassland / Macropus Herbivore binary i Log 2 Single class Coleoptera al. Savanna giganetus Exclusion Addition Berger et 2001 North Forest / Alces alces Natural binary iii none 1 Single class Birds al. America Woodland Experiment Bergstrom, 2000 Africa Grassland / Mixed Simulation of gradient i none 1 Single class Phytophagous Insects Skarpe & Savanna ungulate Herbivory Danell assemblage Bonal & 2007 Europe Forest / Cervus Herbivore binary iii none 2 single Beetles Munoz Woodland elaphus Exclusion species

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Bressette, 2012 North Forest / Odocoileus Herbivore binary i none 1 Multiple Birds Beck & America Woodland virginianus Exclusion classes Arthropods - Overall, Beauchamp Araneae, Chilopoda, Coleoptera, Diplopoda, Hymenoptera Brooks 1999 North Forest / Odocoileus Manipulation binary iii Disturbanc 3 single Amphibians America Woodland virginianus of herbivore e species density Buesching 2011 Europe Forest / Capreolus Herbivore binary i none 1 Single class Small mammals et al. Woodland capreolus, Exclusion Dama dama, Muntiacus reevesi Bush et al. 2012 Europe Forest / Capreolus Herbivore binary i none 1 Single class Small mammals Woodland capreolus, Exclusion Dama dama, Muntiacus reevesi Casey & 1983 North Forest / Cervus Herbivore binary iii none 1 Single class Birds Hein America Woodland elaphus, Exclusion Odocoileus virginianus, Ovus musimon Christopher 2012 North Forest / Odocoileus Herbivore binary i Weeds 2 Multiple Arthropods - Overall, & Cameron America Woodland virginianus Exclusion classes Acari, Araneae, Coleoptera, Collembola, Hymenoptera Cumming 1997 Africa Forest / Loxodonta Herbivore binary iii none 1 Multiple Bats, Birds et al. Woodland africana Exclusion classes Hymenoptera, Orthoptera Danell & 1985 Europe Forest / Alces alces Simulation of binary i none 1 Multiple Hymenoptera, Huss- Woodland herbivory classes Phytophagous Insects Danell

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Decalesta 1994 North Forest / Odocoileus Manipulation gradient i none 1 Single class Birds America Woodland virginianus of herbivore density Degraaf, 1991 North Forest / Odocoileus Natural binary iii Tree 3 Single class Birds Healy & America Woodland virginianus Experiment Thinning Brooks Den 2004 Europe Forest / Rangifer Herbivore binary i Vegetation 2 Single class Coleoptera, Herder, Woodland tarandus Exclusion Age Phytophagous Insects Virtanen & Roininen Deveny & 2006 North Shrubland / Odocoileus Herbivore binary i none 1 Single class Small mammals Fox America Heath hemionus Exclusion columbianus Feeley & 2006 South Forest / Alouatta Natural gradient iii none 1 Single class Birds Terborgh America Woodland seniculus Experiment Gebeyehu, 2006 Africa Grassland / Mixed native Natural gradient iv none 3 Single class Grasshoppers S. & Savanna ungulate Experiment Samways, assemblage, M. cattle, sheep Gill & Fuller 2007 Europe Forest / Capreolus Herbivore binary i Disturbanc 5 Single class Birds Woodland capreolus, Exclusion e Dama dama, Muntiacus reevesi Goheen et 2004 Africa Grassland / Mixed Herbivore binary i none 1 Multiple Small mammals al. Savanna ungulate Exclusion classes Coleoptera, Hemiptera, assemblage Orthoptera Goheen et 2010 Africa Grassland / Mixed Herbivore gradient i none 3 single Small mammals al. Savanna ungulate Exclusion species assemblage, Cattle Goheen et 2013 Africa Grassland / Mixed native Herbivore binary i Rainfall 3 Multiple Small mammals al. Savanna ungulate Exclusion classes assemblage

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Gomez & 2002 Europe Shrubland / Capra Herbivore binary i none 3 Single class Coleoptera Gonzalez Heath pyrenaica, Exclusion Megias Sheep Gomez & 2007 Europe Shrubland / Capra Herbivore binary i none 6 Single class Phytophagous Insects Gonzalez- Heath pyrenaica Exclusion Megias González- 2004 Europe Shrubland / Capra Herbivore binary iii none 2 Multiple Arthropods - Overall, Megías, Heath pyrenaica, Exclusion classes Coleoptera, Gómez Sheep, Goats Hymenoptera &Sánchez- PiÑero Gough et 2012 North Mixed Rangifer Herbivore binary i Nutrient 1 Multiple Soil Invertebrates al. America Habitats tarandus, Exclusion addition classes mixed rodents Greenwald, 2008 North Mixed Odocoileus Herbivore binary iv none 2 Multiple Amphibians, Reptiles Petit & America Habitats virginianus Exclusion classes Arthropods - Overall, Waite Gastropoda, Isopoda, Myriapoda, Oligochaeta Hagenah, 2009 Africa Grassland / Mixed Herbivore gradient i none 2 Single class Small mammals Prins & Olff Savanna ungulate Exclusion assemblage Herremans 1995 Africa Forest / Loxodonta Natural binary iv none 2 Single class Birds Woodland africana Experiment Hino 2000 Asia Forest / Cervus Natural binary iv none 4 Single class Birds Woodland nippon Experiment Hino 2006 Asia Forest / Cervus Natural gradient iv none 1 Single class Birds Woodland nippon Experiment Holt, Fuller 2010 Europe Forest / Capreolus Herbivore binary i Disturbanc 10 single Birds & Dolman Woodland capreolus, Exclusion e species Dama dama, Muntiacus reevesi

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Holt, Fuller 2011 Europe Forest / Capreolus Herbivore binary i Disturbanc 3 Multiple Birds & Dolman Woodland capreolus, Exclusion e classes Flying Insects, Dama dama, Phytophagous Insects Muntiacus reevesi Holt, Fuller 2013 Europe Forest / Dama dama, Herbivore binary i none 3 single Birds & Dolman Woodland Capreolus Exclusion species capreolus, Muntiacus reevesi Huffman et 2009 North Forest / Cervus Herbivore binary ii none 3 Single class Arthropods - Overall al. America Woodland elaphus, Exclusion Odocoileus hemionus Huntzinger, 2008 North Mixed Lepus Herbivore binary ii none 2 Multiple Gastropoda, Karban & America Habitats californicus, Exclusion classes Lepidoptera, Orthoptera Cushman Odocoileus hemionus Ims et al. 2007 Europe Grassland / Rangifer Manipulation binary i none 1 Multiple Birds, Small Mammals Savanna tarandus of herbivore classes density Jay-Robert 2008 Europe Mixed Cervus Natural binary iv none 2 Single class Coleoptera et al. Habitats elaphus, Experiment sheep Joern 2004 North Grassland / Bos bison Manipulation binary iii Fire 1 Single class Orthoptera America Savanna of herbivore density Joern 2005 North Grassland / Bos bison Manipulation binary iii Fire 1 Single class Orthoptera America Savanna of herbivore density Jonas & 2007 North Grassland / Bos bison Manipulation binary iii Fire 1 Single class Orthoptera Joern America Savanna of herbivore density

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Jonsson et 2010 Africa Grassland / Mixed Herbivore gradient iii none 1 Single class Araneae, Coleoptera, al. Savanna ungulate Exclusion Hymenoptera, assemblage Orthoptera Kanda et al. 2005 Asia Forest / Cervus Natural gradient iv none 1 Single class Coleoptera Woodland nippon Experiment Keesing 1998 Africa Grassland / Mixed Herbivore binary i none 2 Single class Small mammals Savanna ungulate Exclusion assemblage Keesing 2000 Africa Grassland / Mixed Herbivore binary i none 4 Single class Small mammals Savanna ungulate Exclusion assemblage Kuria et al. 2010 Africa Grassland / Mixed Herbivore gradient i none 1 Single class Coleoptera Savanna ungulate Exclusion assemblage, cattle Lessard et 2012 North Forest / Odocoileus Herbivore binary i none 1 Multiple Arthropods - Overall, al. America Woodland virginianus Exclusion classes Acari, Araneae, Coleoptera, Collembola, Hymenoptera Lind et al. 2012 North Forest / Odocoileus Herbivore binary i none 1 Single class Phytophagous Insects America Woodland virginianus Exclusion Little, 2013 Africa Grassland / Mixed native Herbivore binary iii Fire 2 Multiple Birds Hockey & Savanna ungulate Exclusion classes Arthropods – Overall, Jansen assemblage Orthoptera Manning, 2013 Oceania Mixed Macropus Herbivore binary i Log 4 Single class Reptiles Cunningha Habitats giganetus Exclusion Addition, m & Vegetation Lindenmay Density er Martin & 2012 North Forest / Cervus Herbivore binary i none 7 Single class Birds Maron America Woodland canadensis Exclusion Martin, 2011 North Forest / Odocoileus Natural gradient ii none 1 Single class Birds Arcese & America Woodland hemionus Experiment Scheerder

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Martinsen, 1998 North Forest / Castor Natural binary iv none 1 single Coleoptera Driebe & America Woodland canadensis Experiment species Whitham Mathisen & 2011 Europe Forest / Alces alces Manipulation gradient iii Nutrient 1 Single class Birds Skarpe Woodland of herbivore addition density Mathisen et 2012 Europe Forest / Alces alces Manipulation gradient iii none 5 Single class Birds al. Woodland of herbivore density Matlack, 2001 North Grassland / Bos bison, Manipulation binary iii Fire 2 single Small mammals Kaufman & America Savanna Cattle of herbivore species Kaufman density McCauley 2006 Africa Grassland / Mixed Herbivore binary i none 3 Multiple Reptiles, Small et al. Savanna ungulate Exclusion classes Mammals assemblage McCauley 2008 Africa Grassland / Mixed Herbivore binary i none 2 Multiple Small mammals et al. Savanna ungulate Exclusion classes Parasitic Invertebrates assemblage McShea & 2000 North Forest / Odocoileus Herbivore binary i none 9 Single class Birds Rappole America Woodland virginianus Exclusion Melis et al. 2006 Europe Forest / Cervus Natural gradient iv none 1 Single class Coleoptera Woodland elaphus Experiment Melis et al. 2007 Europe Forest / Alces alces Natural gradient iv none 1 Single class Coleoptera Woodland Experiment Mohr, 2005 Europe Forest / Cervus Herbivore binary iii Disturbanc 2 Multiple Arthropods - General, Cohnstaedt Woodland elaphus Exclusion e classes Araneae, Isopoda & Topp Moser & 2000 North Grassland / Cervus Herbivore binary i none 1 Multiple Birds, Small Mammals Witmer America Savanna elaphus, Exclusion classes Cattle Neff, Fettig 2007 North Mixed Cervus Herbivore binary i none 4 Single class Lepidoptera & America Habitats elaphus Exclusion VanOverbe ke

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Newson et 2012 Europe Forest / Capreolus Natural gradient i none 11 Single class Birds al. Woodland capreolus, Experiment Dama dama, Muntiacus reevesi Nkwabi et 2011 Africa Grassland / Connochaete Natural binary iii Fire 1 Multiple Birds al. Savanna s sp. Experiment classes Arthropods - Overall Ogada et 2008 Africa Grassland / Mixed Herbivore gradient i none 2 Multiple Birds al. Savanna ungulate Exclusion classes Arthropods - Overall assemblage Olofsson & 2000 Europe Grassland / Rangifer Herbivore binary ii none 2 Multiple Phytophagous Insects Strengbom Savanna tarandus Exclusion classes Palmer et 2008 Africa Grassland / Mixed Herbivore binary i none 1 Multiple Arthropods Overall, al. Savanna ungulate Exclusion classes Coleoptera, assemblage Hymenoptera Parsons, 2013 North Forest / Odocoileus Herbivore binary i none 6 Single class Small mammals Maron & America Woodland hemionus Exclusion Martin Pedersen, 2007 Europe Forest / Alces alces Manipulation binary ii none 2 Multiple Birds Nilsen & Woodland of herbivore classes Arthropods - Overall Andreassen density Pringle 2008 Africa Grassland / Loxodonta Simulation of binary i none 3 single Reptiles Savanna africana herbivory species Pringle et 2007 Africa Grassland / Mixed Herbivore binary i Nutrients 2 Multiple Reptiles al. Savanna ungulate Exclusion classes Arthropods - Overall, assemblage Coleoptera Rambo & 1999 North Grassland / Cervus Herbivore binary i none 1 Multiple Arthropods - Overall Faeth America Savanna elaphus, Exclusion classes Odocoileus hemionus, Cattle Roininen, 1997 North Mixed Alces alces, Herbivore binary i none 1 Single class Phytophagous Insects Price & America Habitats Lepus Exclusion Bryant americanus

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Saetnan & 2006 Africa Grassland / Mixed Herbivore binary i none 1 Single class Small mammals Skarpe Savanna ungulate Exclusion assemblage Schwenk & 2011 North Forest / Alces alces Natural gradient iii none 2 Single class Lepidoptera Strong America Woodland Experiment Seki and 2013 Asia Forest / Cervus Herbivore binary iv none 1 Multiple Small mammals Koganezaw Woodland nippon Exclusion classes Coleoptera, earthworms a Shibata, 2008 Asia Forest / Cervus Herbivore binary iii none 1 Single class Small mammals Saito & Woodland nippon Exclusion Tanaka Shimazaki 2002 Asia Forest / Cervus Herbivore binary iii none 2 Single class Phytophagous Insects & Miyashita Woodland nippon Exclusion Smit et al. 2001 Europe Mixed Capreolus Herbivore binary i none 1 Single class Small mammals Habitats capreolus, Exclusion Cervus elaphus, Ovis ammon musimon Spalinger et 2012 Europe Grassland / Cervus Natural binary iv none 1 Multiple Orthoptera al. Savanna elaphus, Experiment classes Rupicapra rupicapra Suominen 1999 Europe Forest / Alces alces, Herbivore binary i none 1 Single class Gastropoda Woodland Rangifer Exclusion tarandus Suominen 2003 Europe Forest / Rangifer Herbivore binary i Nutrients 1 Single class Coleoptera et al. Woodland tarandus Exclusion Suominen 2008 Europe Forest / Alces alces Simulation of gradient i Nutrients 1 Multiple Araneae, Hemiptera, et al. Woodland herbivory classes Flying Insects Suominen, 1999 North Mixed Alces alces Herbivore binary i none 1 Multiple Coleoptera, Orthoptera Danell & America Habitats Exclusion classes Bryant

Study Data Study # taxonomic Year Study Herbivore(s) Interacting Authors Location Study Design Contrasts Quality Duration groups Animal Taxa Studied Pub. ecosystem Studied* Factors (continent) Score (years) studied Suominen, 1999 Europe Forest / Alces alces Herbivore binary i none 1 Multiple Arthropods - Overall, Danell & Woodland Exclusion classes Acari, Araneae, Bergström Blattodea, Chilipoda, Coleoptera, Diplopoda, Diptera, Hemiptera, Hymenoptera, Isopoda, Gastropoda, Nemastomatidae, Psocoptera, Phalangiidae Tabuchi, 2010 Asia Forest / Cervus Herbivore binary ii none 1 single Phytophagous Insects Ueda & Woodland nippon Exclusion species Ozaki Takada et 2008 Asia Forest / Cervus Natural gradient iii none 1 Single class Araneae al. Woodland nippon Experiment Warui et al. 2005 Africa Grassland / Mixed Herbivore gradient i none 2 Single class Araneae Savanna ungulate Exclusion assemblage * Wild herbivores (native or exotic) denoted with scientific name. Domestic stock uses common name.

References Bailey, J.K. & Whitham, T.G. (2002) Interactions among fire, aspen, and elk affect insect diversity: reversal of a community response. Ecology, 83, 1701-1712. Bailey, J.K. & Whitham, T.G. (2003) Interactions among elk, aspen, galling sawflies and insectivorous birds. Oikos, 101, 127-134. Bailey, J.K. & Whitham, T.G. (2006) Interactions between cottonwood and beavers positively affect sawfly abundance. Ecological Entomology, 31, 294-297. Baines, D. (1996) The implications of grazing and predator management on the habitats and breeding success of black grouse Tetrao tetrix. Journal of Applied Ecology, 33, 54-62. Baines, D., Sage, R.B. & Baines, M.M. (1994) The implications of red deer grazing to ground vegetation and invertebrate communities of Scottish native pinewoods. Journal of Applied Ecology, 31, 776-783. Barrett, M.A. & Stiling, P. (2007) Relationships among Key deer, insect herbivores, and plant quality. Ecological Research, 22, 268-273. Barton, P.S., Manning, A.D., Gibb, H., Wood, J.T., Lindenmayer, D.B. & Cunningham, S.A. (2011) Experimental reduction of native vertebrate grazing and addition of logs benefit beetle diversity at multiple scales. Journal of Applied Ecology, 48, 943-951. Berger, J., Stacey, P.B., Bellis, L. & Johnson, M.P. (2001) A mammalian predator–prey imbalance: Grizzly bear and wolf extinction affect avian neotropical migrants. Ecological Applications, 11, 947-960. Bergstrom, R., Skarpe, C. & Danell, K. (2000) Plant responses and herbivory following simulated browsing and stem cutting of Combretum apiculatum. Journal of Vegetation Science, 11, 409-414. Bonal, R. & Munoz, A. (2007) Multi-trophic effects of ungulate intraguild predation on acorn weevils. Oecologia, 152, 533-540. Bressette, J.W., Beck, H. & Beauchamp, V.B. (2012) Beyond the browse line: complex cascade effects mediated by white-tailed deer. Oikos, 121, 1749-1760. Brooks, R.T. (1999) Residual effects of thinning and high white-tailed deer densities on northern redback salamanders in southern New England oak forests. Journal of Wildlife Management, 63, 1172-1180. Buesching, C.D., Newman, C., Jones, J.T. & Macdonald, D.W. (2011) Testing the effects of deer grazing on two woodland rodents, bankvoles and woodmice. Basic and Applied Ecology, 12, 207-214. Bush, E.R., Buesching, C.D., Slade, E.M. & Macdonald, D.W. (2012) Woodland recovery after suppression of deer: cascade effects for small mammals, wood mice (Apodemus sylvaticus) and bank voles (Myodes glareolus). PloS one, 7, e31404-e31404. Casey, D. & Hein, D. (1983) Effects of heavy browsing on a bird community in deciduous forest. Journal of Wildlife Management, 47, 829-839.

Christopher, C.C. & Cameron, G.N. (2012) Effects of invasive amur honeysuckle (Lonicera maackii) and white-tailed deer (Odocoileus virginianus) on litter-dwelling arthropod communities. American Midland Naturalist, 167, 256-272. Cumming, D.H.M., Fenton, M.B., Rautenbach, I.L., Taylor, R.D., Cumming, G.S., Cumming, M.S., Dunlop, J.M., Ford, G.S., Hovorka, M.D., Johnston, D.S., Kalcounis, M.C., Mahlanga, Z. & Portfors, C.V. (1997) Elephants, woodlands and biodiversity in southern Africa. South African Journal of Science, 93, 231-236. Danell, K. & Huss-Danell, K. (1985) Feeding by insects and hares on birches earlier affected by moose browsing. Oikos, 44, 75-81. Decalesta, D.S. (1994) Effect of white-tailed deer on songbirds within managed forests in Pennsylvania. Journal of Wildlife Management, 58, 711-718. Degraaf, R.M., Healy, W.M. & Brooks, R.T. (1991) Effects of thinning and deer browsing on breeding birds in New-England oak woodlands. Forest Ecology and Management, 41, 179- 191. Den Herder, M., Virtanen, R. & Roininen, H. (2004) Effects of reindeer browsing on tundra willow and its associated insect herbivores. Journal of Applied Ecology, 41, 870-879. Deveny, A.J. & Fox, L.R. (2006) Indirect interactions between browsers and seed predators affect the seed bank dynamics of a chaparral shrub. Oecologia, 150, 69-77. Feeley, K.J. & Terborgh, J.W. (2006) Habitat fragmentation and effects of herbivore (howler monkey) abundances on bird species richness. Ecology, 87, 144-150. Gebeyehu, S. & Samways, M. (2006) Conservation refugium value of a large mesa for grasshoppers in South Africa. Biodiversity & Conservation, 15, 717-734. Gill, R.M.A. & Fuller, R.J. (2007) The effects of deer browsing on woodland structure and songbirds in lowland Britain. Ibis, 149, 119-127. Goheen, J.R., Keesing, F., Allan, B.F., Ogada, D. & Ostfeld, R.S. (2004) Net effects of large mammals on Acacia seedling survival in an African savanna. Ecology, 85, 1555-1561. Goheen, J.R., Palmer, T.M., Keesing, F., Riginos, C. & Young, T.P. (2010) Large herbivores facilitate savanna tree establishment via diverse and indirect pathways. Journal of Animal Ecology, 79, 372-382. Goheen, J.R., Palmer, T.M., Charles, G.K., Helgen, K.M., Kinyua, S.N., Maclean, J.E., Turner, B.L., Young, H.S. & Pringle, R.M. (2013) Piecewise disassembly of a large-herbivore community across a rainfall gradient: The UHURU experiment. PloS one, 8. Gómez, J. & González-Megías, A. (2002) Asymmetrical interactions between ungulates and phytophagous insects: Being different matters. Ecology, 83, 203-211. Gómez, J. & González-Megías, A. (2007) Long-term effects of ungulates on phytophagous insects. Ecological Entomology, 32, 229-234.

González-Megías, A., Gómez, J. & Sánchez-Piñero, F. (2004) Effects of ungulates on epigeal arthropods in Sierra Nevada National Park (southeast Spain). Biodiversity and Conservation, 13, 733-752. Gough, L., Moore, J.C., Shaver, G.R., Simpson, R.T. & Johnson, D.R. (2012) Above- and belowground responses of arctic tundra ecosystems to altered soil nutrients and mammalian herbivory. Ecology, 93, 1683-1694. Greenwald, K.R., Petit, L.J. & Waite, T.A. (2008) Indirect effects of a keystone herbivore elevate local animal diversity. Journal of Wildlife Management, 72, 1318-1321. Hagenah, N., Prins, H.H.T. & Olff, H. (2009) Effects of large herbivores on murid rodents in a South African savanna. Journal of Tropical Ecology, 25, 483-492. Herremans, M. (1995) Effects of woodland modification by African elephant Loxodonta africana on bird diversity in northern Botswana. Ecography, 18, 440-454. Hino, T. (2000) Bird community and vegetation structure in a forest with a high density of sika deer. Japanese Journal of Ornithology, 48, 197-204. Hino, T. (2006) The impact of herbivory by deer on forest bird communities in Japan. Acta Zoologica Sinica, 52, 684-686. Holt, C.A., Fuller, R.J. & Dolman, P.M. (2010) Experimental evidence that deer browsing reduces habitat suitability for breeding common nightingales Luscinia megarhynchos. Ibis, 152, 335-346. Holt, C.A., Fuller, R.J. & Dolman, P.M. (2011) Breeding and post-breeding responses of woodland birds to modification of habitat structure by deer. Biological Conservation, 144, 2151-2162. Holt, C.A., Fuller, R.J. & Dolman, P.M. (2013) Deer reduce habitat quality for a woodland songbird: Evidence from settlement patterns, demographic parameters, and body condition. Auk, 130, 13-20. Huffman, D.W., Laughlin, D.C., Pearson, K.M. & Pandey, S. (2009) Effects of vertebrate herbivores and shrub characteristics on arthropod assemblages in a northern Arizona forest ecosystem. Forest Ecology and Management, 258, 616-625. Huntzinger, M., Karban, R. & Cushman, H. (2008) Negative effects of vertebrate herbivores on invertebrates in a coastal dune community. Ecology, 89, 1972-1980. Ims, R.A., Yoccoz, N.G., Brathen, K.A., Fauchald, P., Tveraa, T. & Hausner, V. (2007) Can reindeer overabundance cause a trophic cascade? Ecosystems, 10, 607-622. Jay-Robert, P., Niogret, J., Errouissi, F., Labarussias, M., Paoletti, E., Luis, M.V. & Lumaret, J.-P. (2008) Relative efficiency of extensive grazing vs. wild ungulates management for dung beetle conservation in a heterogeneous landscape from Southern Europe (Scarabaeinae, Aphodiinae, Geotrupinae). Biological Conservation, 141, 2879-2887. Joern, A. (2004) Variation in grasshopper (acrididae) densities in response to fire frequency and bison grazing in tallgrass prairie. Environmental Entomology, 33, 1617-1625.

Joern, A. (2005) Disturbance by fire frequency and bison grazing modulate grasshopper assemblages in tallgrass prairie. Ecology, 86, 861-873. Jonas, J.L. & Joern, A. (2007) Grasshopper (Orthoptera : Acrididae) communities respond to fire, bison grazing and weather in North American tallgrass prairie: a long-term study. Oecologia, 153, 699-711. Jonsson, M., Bell, D., Hjalten, J., Rooke, T. & Scogings, P.F. (2010) Do mammalian herbivores influence invertebrate communities via changes in the vegetation? Results from a preliminary survey in Kruger National Park, South Africa. African Journal of Range & Forage Science, 27, 39-44. Kanda, N., Yokota, T., Shibata, E.i. & Sato, H. (2005) Diversity of dung-beetle community in declining Japanese subalpine forest caused by an increasing sika deer population. Ecological Research, 20, 135-141. Keesing, F. (2000) Cryptic consumers and the ecology of an African savanna. Bioscience, 50, 205-215. Keesing, F. (1998) Impacts of ungulates on the demography and diversity of small mammals in central Kenya. Oecologia, 116, 381-389. Kuria, S.K., Villet, M.H., Palmer, T.M. & Stanton, M.L. (2010) A comparison of two sampling methods for surveying mammalian herbivore impacts on beetle communities in the canopy of Acacia drepanolobium in Kenya. African Entomology, 18, 87-98. Lessard, J.-P., Reynolds, W.N., Bunn, W.A., Genung, M.A., Cregger, M.A., Felker-Quinn, E., Barrios-Garcia, M.N., Stevenson, M.L., Lawton, R.M., Brown, C.B., Patrick, M., Rock, J.H., Jenkins, M.A., Bailey, J.K. & Schweitzer, J.A. (2012) Equivalence in the strength of deer herbivory on above and below ground communities. Basic and Applied Ecology, 13, 59-66. Lind, E.M., Myron, E.P., Giaccai, J. & Parker, J.D. (2012) White-tailed deer alter specialist and generalist insect herbivory through plant traits. Environmental Entomology, 41, 1409-1416. Little, I.T., Hockey, P.A.R. & Jansen, R. (2013) A burning issue: Fire overrides grazing as a disturbance driver for South African grassland bird and arthropod assemblage structure and diversity. Biological Conservation, 158, 258-270 Manning, A.D., Cunningham, R.B. & Lindenmayer, D.B. (2013) Bringing forward the benefits of coarse woody debris in ecosystem recovery under different levels of grazing and vegetation density. Biological Conservation, 157, 204-214. Martin, T.G., Arcese, P. & Scheerder, N. (2011) Browsing down our natural heritage: Deer impacts on vegetation structure and songbird populations across an island archipelago. Biological Conservation, 144, 459-469. Martin, T.E. & Maron, J.L. (2012) Climate impacts on bird and plant communities from altered animal-plant interactions. Nature Clim. Change, 2, 195-200.

Martinsen, G.D., Driebe, E.M. & Whitham, T.G. (1998) Indirect interactions mediated by changing plant chemistry: Beaver browsing benefits beetles. Ecology, 79, 192-200. Mathisen, K. & Skarpe, C. (2011) Cascading effects of moose (Alces alces) management on birds. Ecological Research, 26, 563-574. Mathisen, K.M., Pedersen, S., Nilsen, E.B. & Skarpe, C. (2012) Contrasting responses of two passerine bird species to moose browsing. European Journal of Wildlife Research, 58, 535- 547. Matlack, R.S., Kaufman, D.W. & Kaufman, G.A. (2001) Influence of grazing by bison and cattle on deer mice in burned tallgrass prairie. American Midland Naturalist, 146, 361-368. McCauley, D.J., Keesing, F., Young, T.P., Allan, B.F. & Pringle, R.M. (2006) Indirect effects of large herbivores on snakes in an African savanna. Ecology, 87, 2657-2663. McCauley, D.J., Keesing, F., Young, T. & Dittmar, K. (2008) Effects of the removal of large herbivores on fleas of small mammals. Journal of Vector Ecology, 33, 263-268. McShea, W.J. & Rappole, J.H. (2000) Managing the abundance and diversity of breeding bird populations through manipulation of deer populations. Conservation Biology, 14, 1161- 1170. Melis, C., Buset, A., Aarrestad, P.A., Hanssen, O., Meisingset, E.L., Andersen, R., Moksnes, A. & Roskaft, E. (2006) Impact of red deer Cervus elaphus grazing on bilberry Vaccinium myrtillus and composition of ground beetle (Coleoptera, Carabidae) assemblage. Biodiversity and Conservation, 15, 2049-2059. Melis, C., Sundby, M., Andersen, R., Moksnes, A., Pedersen, B. & Røskaft, E. (2007) The role of moose Alces alces; L. in boreal forest – the effect on ground beetles (Coleoptera, Carabidae) abundance and diversity. Biodiversity and Conservation, 16, 1321-1335. Mohr, D., Cohnstaedt, L.W. & Topp, W. (2005) Wild boar and red deer affect soil nutrients and soil biota in steep oak stands of the Eifel. Soil Biology & Biochemistry, 37, 693-700. Moser, B.W. & Witmer, G.W. (2000) The effects of elk and cattle foraging on the vegetation, birds, and small mammals of the Bridge Creek Wildlife Area, Oregon. International Biodeterioration & Biodegradation, 45, 151-157. Neff, P.K.K., Fettig, S.M. & VanOverbeke, D.R. (2007) Variable response of butterflies and vegetation to elk herbivory: An exclosure experiment in ponderosa pine and aspen-mixed conifer forests. Southwestern Naturalist, 52, 1-14. Newson, S.E., Johnston, A., Renwick, A.R., Baillie, S.R. & Fuller, R.J. (2012) Modelling large- scale relationships between changes in woodland deer and bird populations. Journal of Applied Ecology, 49, 278-286. Nkwabi, A.K., Sinclair, A.R.E., Metzger, K.L. & Mduma, S.A.R. (2011) Disturbance, species loss and compensation: Wildfire and grazing effects on the avian community and its food supply in the Serengeti Ecosystem, Tanzania. Austral Ecology, 36, 403-412.

Ogada, D.L., Gadd, M.E., Ostfeld, R.S., Young, T.P. & Keesing, F. (2008) Impacts of large herbivorous mammals on bird diversity and abundance in an African savanna. Oecologia, 156, 387-397. Olofsson, J. & Strengbom, J. (2000) Response of galling invertebrates on Salix lanata to reindeer herbivory. Oikos, 91, 493-498. Palmer, T.M., Stanton, M.L., Young, T.P., Goheen, J.R., Pringle, R.M. & Karban, R. (2008) Breakdown of an ant-plant mutualism follows the loss of large herbivores from an African Savanna. Science, 319, 192-195. Parsons, E.W.R., Maron, J.L. & Martin, T.E. (2013) Elk herbivory alters small mammal assemblages in high-elevation drainages. Journal of Animal Ecology, 82, 459-467. Pedersen, S., Nilsen, E.B. & Andreassen, H.P. (2007) Moose winter browsing affects the breeding success of great tits. Ecoscience, 14, 499-506. Pringle, R.M. (2008) Elephants as agents of habitat creation for small vertebrates at the patch scale. Ecology, 89, 26-33. Pringle, R.M., Young, T.P., Rubenstein, D.I. & McCauley, D.J. (2007) Herbivore-initiated interaction cascades and their modulation by productivity in an African savanna. Proceedings of the National Academy of Sciences of the United States of America, 104, 193-197. Rambo, J.L. & Faeth, S.H. (1999) Effect of vertebrate grazing on plant and insect community structure. Conservation Biology, 13, 1047-1054. Roininen, H., Price, P.W. & Bryant, J.P. (1997) Response of galling insects to natural browsing by mammals in Alaska. Oikos, 80, 481-486. Saetnan, E.R. & Skarpe, C. (2006) The effect of ungulate grazing on a small mammal community in southeastern Botswana. African Zoology, 41, 9-16. Schwenk, W.S. & Strong, A.M. (2011) Contrasting patterns and combined effects of moose and insect herbivory on striped maple (Acer pensylvanicum). Basic and Applied Ecology, 12, 64-71. Seki, Y. & Koganezawa, M. (2013) Does sika deer overabundance exert cascading effects on the raccoon dog population? Journal of Forest Research, 18, 121-127. Shibata, E.I., Saito, M. & Tanaka, M. (2008) Deer-proof fence prevents regeneration of Picea jezoensis var. hondoensis through seed predation by increased woodmouse populations. Journal of Forest Research, 13, 89-95. Shimazaki, A. & Miyashita, T. (2002) Deer browsing reduces leaf damage by herbivorous insects through an induced response of the host plant. Ecological Research, 17, 527-533. Smit, R., Bokdam, J., den Ouden, J., Olff, H., Schot-Opschoor, H. & Schrijvers, M. (2001) Effects of introduction and exclusion of large herbivores on small rodent communities. Plant Ecology, 155, 119-127.

Spalinger, L.C., Haynes, A.G., Schuetz, M. & Risch, A.C. (2012) Impact of wild ungulate grazing on Orthoptera abundance and diversity in subalpine grasslands. Insect Conservation and Diversity, 5, 444-452. Suominen, O. (1999) Impact of cervid browsing and grazing on the terrestrial gastropod fauna in the boreal forests of Fennoscandia. Ecography, 22, 651-658. Suominen, O., Danell, K. & Bryant, J.P. (1999) Indirect effects of mammalian browsers on vegetation and ground-dwelling insects in an Alaskan floodplain. Ecoscience, 6, 505-510. Suominen, O., Danell, K. & Bergström, R. (1999) Moose, trees, and ground-living invertebrates: indirect interactions in swedish pine forests. Oikos, 84, 215-266. Suominen, O., Niemela, J., Martikainen, P., Niemela, P. & Kojola, I. (2003) Impact of reindeer grazing on ground-dwelling Carabidae and Curculionidae assemblages in Lapland. Ecography, 26, 503-513. Suominen, O., Persson, I.-L., Danell, K., Bergström, R. & Pastor, J. (2008) Impact of simulated moose densities on abundance and richness of vegetation, herbivorous and predatory arthropods along a productivity gradient. Ecography, 31, 636-645. Tabuchi, K., Ueda, A. & Ozaki, K. (2010) Contrasting effects of deer browsing on oviposition preference, neonate survival and potential fecundity of a galling insect. Ecoscience, 17, 379-386. Takada, M., Baba, Y.G., Yanagi, Y., Terada, S. & Miyashita, T. (2008) Contrasting responses of web-building spiders to deer browsing among habitats and feeding guilds. Environmental Entomology, 37, 938-946. Warui, C.M., Villet, M.R., Young, T.P. & Jocque, R. (2005) Influence of grazing by large mammals on the spider community of a Kenyan savanna biome. Journal of Arachnology, 33, 269-279.

Supplementary Tables

Table S1. Mechanisms driving the effects of large herbivores on other animals. Counts indicate the number of papers where authors conclude that mechanism to be important for driving the effect of large herbivores on a particular animal group. The most common mechanism(s) for each taxonomic group appears in bold.

Number of papers concluding mechanism to be important

For Mechanism Small any Birds Arthropoda Araneae Coleoptera Hemiptera Hymenoptera Lepidoptera Orthoptera mammals taxa

Vegetation biomass 48 14 12 6 1 6 1 1 1 4 Vegetation structure 47 21 9 5 3 5 2 4

Via changes in other animals (e.g. prey) 17 7 2 2 2 1

Vegetation chemistry / nutrient quality 13 1 2 1

Vegetation species composition 13 4 1 1 2 1 1 3

Physical environment 10 1 3 2 5 1 2

Direct interference 5 1 2

Nutrient input/cycling (e.g. dung) 2 2

Total number of papers giving any 84 26 14 11 5 17 2 4 1 6 conclusions

Table S2. Recommendations for future research given by authors. We sorted recommendations according to whether they were related to research topic or to the design of future studies. Fifty one of the 96 papers contained specific recommendations for future research.

Recommendation Number of papers Research Topic Cascading effects on other species or ecosystem processes 19 Mechanisms of effects 11 Novel habitats or locations 7 Interacting factors 7 Study Design Larger spatial scales / spatial variation in responses 6 Studies able to detect both direct and indirect relationships 5 Studies able to detect the shape of herbivory responses 3 Long-term studies to distinguish between state shifts and short-term 3 fluctuations

Supplementary Figures

a. 35 Multiple habitats Heath / Shrubland 30 Grassland / Savanna 25 Forest / Woodland

5 Number Number of papers

0 North Europe Africa Asia Oceania South America America Study location b. c. 80 60 Both Not over-abundant 70 Vertebrates Over-abundant 50 Invertebrates 60 50 40 40 30 30 20 20

10 Number Number of papers

0 10 Number Number of papers

0 Single sp. Single class Multiple classes Order of focal herbivores Number of taxa studied

d. e. 25 25

20 20

15 15

10 10

5 5

Number Number of papers Number Number of papers 0 0

Animal response groups: Vertebrate Animal response groups: Invertebrate classes orders Figure S1. Distribution of the 96 reviewed papers by (a) continent and study system, (b) focal herbivores, (c) number of animal response groups studied, (d) animal response groups: class of vertebrates and (e) animal response groups: most common invertebrate orders. Studies which belonged to more than one category were counted in both.

Native

12 12 0

14 1 154

Exotic Livestock

Figure S2. Studies returned by our initial search which studied the effects of terrestrial mammalian herbivores on other animals, sorted by the status of the focal herbivores (native, exotic or livestock). Not all studies included under ‘native’ were included in the final review as a number did not meet all criteria due to factors such as size of herbivores, absence of contrasts etc. (see Appendix S1 for full details).

Figure S3. The distribution of experimental approaches (a.), contrasts (b.) and exclusion method (c.) used by studies published in each five year period. The frequency of studies using each of the experimental approaches did not vary among time periods (P = 0.29). The frequency of studies employing more than two levels of herbivory, and including selective exclusion fences was higher among studies published since 2004 (P = 0.03 and P = 0.017).

Paper II: Herbivory and fire interact to affect forest understory habitat, but not its use by small vertebrates

In Paper I, I found that large herbivores can have substantial effects on the abundance and diversity of other animal groups. However, I also found that few studies have investigated the impacts of Australian native herbivores, and that knowledge of how large herbivores interact with episodic disturbances remains limited. In this paper, I experimentally tested for the interactive effects of prescribed fire and macropod herbivores (kangaroos and wallabies) on forest understorey vegetation and small vertebrate fauna.

Foster, C.N., Barton, P.S., Sato, C.F., Wood, J.T., Macgregor, C.I., and D.B. Lindenmayer (2016) Herbivory and fire interact to affect forest understory habitat, but not its use by small vertebrates. Animal Conservation. 19, 15-25.

Abstract

Herbivory and fire are two disturbances which often co-occur, but studies of their interactive effects are rare outside of grassland ecosystems. We experimentally tested the interactive effects of prescribed fire and macropod herbivory on forest understory vegetation and its vertebrate fauna. Fire and herbivory interacted synergistically to affect forest understory vegetation, with palatable plants showing poor post-fire recovery in un-fenced sites compared with herbivore exclusion sites. Despite this strong interactive effect on vegetation, small vertebrates responded to the individual, and not the interactive effects of disturbance. The native insectivorous mammal Antechinus stuartii was more frequently encountered on large herbivore exclusion sites, as was the introduced European rabbit. In contrast, the small skink Lampropholis delicata was more common on sites with high densities of large herbivores. Skinks, snakes and

European rabbits were also more active on burnt than unburnt sites. Our results suggest that it may be necessary to manage the macropod herbivore population after fire to prevent the decline of palatable plants, and maintain the dense habitat required by some small mammals. However, as the invasive rabbit was most active in macropod-free sites after fire, any management must include control of both types of herbivores. A mix of understory densities may also need to be maintained to ensure the persistence of species preferring more open habitats. Our study demonstrates that interactive effects of disturbance on vegetation communities may not lead to predictable effects on animals, and highlights the importance of considering both multiple stressors, and multiple species, in the management of disturbance regimes.

Keywords

Disturbance interaction, browsing, grazing, indirect effects, kangaroo, synergistic effects

Introduction

Disturbance regimes play a central role in ecosystem dynamics (Willig and Walker, 1999).

However, in many parts of the world, natural disturbance regimes have been disrupted, with unwanted outcomes for biodiversity (Hobbs and Huenneke, 1992, Sinclair and Byrom, 2006).

Thus, many biodiversity conservation programs aim to reinstate natural disturbance regimes

(Fuhlendorf et al., 2010, Noss et al., 2006), but achieving this in systems where multiple disturbances co-occur may be difficult. Co-occurring disturbances can interact to produce distinctly different outcomes from what would be expected based on individual effects (Didham et al., 2007, Tylianakis et al., 2008) and a poor understanding of these interactions can lead to unexpected and undesirable management outcomes (Lindenmayer et al., 2010, Tylianakis et al.,

2008).

The importance of disturbance interactions for structuring grasslands and heathlands has been widely recognised, and re-establishing fire-grazing interactions is identified as a priority for maintaining biodiversity in these habitats (Fuhlendorf et al., 2010, Van Langevelde et al.,

2003). However, understanding of how fire and herbivory interact to affect species in forested habitats remains limited (Foster, Barton and Lindenmayer, 2014, Royo et al., 2010, Wisdom et al., 2006). As the interactive effects of fire and herbivory depend on the scale, intensity and timing of these disturbances, the outcome of interactions can be highly variable (Fuhlendorf et al., 2010, Wisdom et al., 2006). For example, at a local-scale, deer browsing after fire supressed dominant shrub species, increasing herbaceous plant richness in a forest understory (Royo et al.,

2010). Conversely, heavy macropod herbivory following fire limited grass and forb recovery

(Tuft, Crowther and McArthur, 2012). At a larger scale, Bailey and Whitham (2002) found that elk (Cervus canadensis) browsed more heavily in areas of aspen (Populus tremuloides) that burned at high intensity, compared with moderate intensity. This heavy browsing reversed the positive effect of browsing on arthropod richness that occurred after moderate intensity fire.

While such studies indicate that fire-herbivory interactions are likely to be prevalent in forested ecosystems (Royo and Carson, 2006), investigations of animal responses to the combined effects of these disturbances remain rare (Foster et al., 2014, Wisdom et al., 2006).

We combined prescribed fire and large herbivore exclusion treatments to test the interactive effects of fire and herbivory on understory vegetation and small vertebrates in a temperate forest ecosystem. As the management of disturbances is often targeted at plants, with the assumption that this will also cater for the needs of animals (Clarke, 2008), it is important to 83

understand whether such assumptions are valid, and whether fauna respond in a predicable way to disturbances. Our study addressed the following questions: (1) How do fire, herbivory and their interaction affect understory habitat structure at the site level? (2) How do these disturbances affect site occupancy by small vertebrate fauna? We expected that vertebrate species would respond differently to the experimental treatments due to differences in their habitat and dietary preferences, and that these responses would be mediated by changes in vegetation structure. For example, we expected that both fire and herbivory would reduce understory cover, and lead to negative effects on site occupancy by vertebrates preferring dense understory habitats (Table 1). We provide recommendations for biodiversity conservation based on our findings.

Table 1. Predicted effects of fire and large herbivores on habitat and food resources for vertebrates and the corresponding predicted responses of two small mammal and two reptile species, based on their diet and habitat preferences. Habitat preferences are attributes which have been associated with higher abundance in forest habitat for that species.

Predicted response Response group Habitat preference Diet Fire Herbivores

Habitat and food resources Understory cover - - Understory height - - Leaf litter depth - None Fresh plant growth + - Invertebrate prey - -

Vertebrates European rabbit Open understory a Forbs and + - (Oryctolagus cuniculus) grasses b Brown antechinus Dense, complex understory a,c Invertebrates c - - (Antechinus stuartii) Tall understory d Abundant logs d Delicate skink High canopy cover e Invertebrates e - ? (Lampropholis delicata) Deep litter f,g Tall understory g Eastern small-eyed snake Warm diurnal refuge h Skinks h + ? (Cryptophis nigrescens) aCatling and Burt (1995), b Davis, Coulson and Forsyth (2008), c Bennett (1993), d Knight and Fox (2000), e Bragg, Taylor and Fox (2005), f Taylor and Fox (2001), g Howard, Williamson and Mather (2003), h Webb et al. (2004)

Materials and methods

Study site

We conducted our study in Booderee National Park (BNP); a ~ 65 km2 peninsula in south- eastern Australia (35°10′S, 150°40′E). We established sites within Eucalyptus pilularis forest, which is the most widespread vegetation type in BNP (Taws, 1998). An intensive fox (Vulpes vulpes) baiting program has been conducted in BNP since 1999 to protect native species from predation (Dexter et al., 2012). Over the last decade, there has been a tenfold increase in native herbivores in BNP (predominantly swamp wallaby, Wallabia bicolor, and eastern grey kangaroo, Macropus giganteus), which is attributed to reduced predation by foxes

(Lindenmayer et al., 2014). Small-scale exclosure trials indicate that this high abundance of herbivores could be driving changes in vegetation composition (Dexter et al., 2013), and there is concern about flow-on effects for smaller vertebrates, which include a number of threatened species (Dexter et al., 2012). As fire is a naturally occurring disturbance within BNP, occurring both as wildfire and low-intensity prescribed burning (Lindenmayer et al., 2008), it is important to understand how native herbivores interact with fire regimes.

Study design

We quantified the interactive effects of fire and herbivory on vegetation and small vertebrates using a randomised blocked experiment. We combined three levels of large herbivore exclosure and two levels of burning treatment in a factorial design (Appendix 1). We replicated each treatment combination across four blocks to give 24 sites. For the exclosure treatments, we excluded macropod herbivores from 25 × 25 m sites, using 1.1 m tall wire fencing, in June

2012. We created three levels of herbivore exclosure treatment: high activity (open treatment – no fence), intermediate activity (partial treatment – sites were fenced but gates opened and closed at two month intervals to simulate lower herbivore pressure), and no large herbivores

(exclosure treatment). For the burnt treatments, we conducted 50 × 50 m burns in August 2012, with the 25 x 25 m site in the centre of the burnt area. Fire was low-intensity, removing

approximately 95% of understory vegetation and did not reach the canopy (scorch height 1.5 - 4 m) or burn large logs.

Data collection

We surveyed understorey vegetation prior to treatment in April/May 2012 and repeated surveys biannually until May 2014. We sampled four 3 × 3 m quadrats in each site, with each quadrat at least 1.5 m from the fence. We recorded four vegetation variables, representing important attributes of fauna habitat; total understorey projective cover (%), projective cover of bracken

(Pteridium escelentum) (%), understorey height (averaged across 10 locations per quadrat using the stick-and-disc method of Smit et al. (2001); disc 100 mm diameter, 4.7 g weight) and litter depth (averaged across 10 locations per quadrat).

We surveyed macropod herbivore activity by counting scats along two 25 × 2 m transects (100 m-2) per site, summing counts to give one count per site. As macropods defecate more while feeding than resting (Johnson, Jarman and Southwell, 1987), pellet counts can give a comparative measure of macropod feeding pressure between sites (Howland et al., 2014). We surveyed transects every two months from August 2012 to May 2014, removing scats after each survey to avoid re-counting. We also monitored European rabbit (Oryctolagus cuniculus) activity using these transects, counting the number of rabbit diggings every two months from

June 2013 to May 2014.

We surveyed site use by small mammals through live trapping in April/May 2012 and then every six months until May 2014. For each survey, we set eight Elliott traps per site for four consecutive nights. We ear marked animals with a permanent marker to identify recaptures within a survey.

We monitored reptiles using iron sheeting as artificial substrates. We set out four 1 × 1 m sheets per site in July 2012, and checked them on two consecutive mornings approximately every two months from October 2012 to May 2014. To minimise biases due to time of day, we rotated the order of site checking so that each site had one early morning and one late morning check per sampling period. To address non-independence of counts within a survey, we used the 86

maximum value of the two consecutive counts for each species. Weather conditions meant that some surveys returned few individuals. Therefore, for data analysis, we excluded surveys with fewer than three detections for that species.

To measure arthropod prey availability, we sampled ground-dwelling beetles and spiders using pitfall traps, counting the total captures per trap. We deployed four 250ml (100 mm diameter) traps per site (2/3 filled with non-toxic polyethylene glycol solution) for two weeks in November 2012 and 2013.

Data analysis

To assess how fire-herbivory interactions affected habitat structure (question 1), and fauna occupancy (question 2) we tested treatment effects on dependent variables using linear mixed models (LMMs) for vegetation variables and generalised linear mixed models (GLMMs) with

Poisson errors for animal counts. Vegetation variables were understory cover (%), understory cover excluding bracken (%, total understory cover minus cover of bracken), understory height

(m), and leaf litter depth (mm). We analysed understory cover excluding bracken because bracken is a dominant, unpalatable species which could mask responses of other plants. Bracken also provides little of the ground-level structure important for small vertebrates (Bennett, 1993).

Animal count variables were macropod scats, rabbit diggings, antechinus captures, delicate skink (Lampropholis delicata) and eastern small-eyed snake sightings (Cryptophis nigrescens), and spider and beetle captures.

We fitted each dependent variable with the full fixed effects model of herbivores × burning × time, and random effects of block/site/quadrat for vegetation and block/site for animal variables to account for the repeated measures. A first-order auto-regressive covariance structure on the random effects was trialled for the vegetation LMMs but was not used as it did not improve model fit (ΔAIC < 2, Pinheiro and Bates, 2000). We did not define a covariance structure for animal responses as inspection of residuals indicated little evidence of temporal autocorrelation, and methods for fitting such structures with GLMMs are not well developed

(Zuur et al., 2009). We used Akaike information criterion, corrected for small sample size

(AICc), to select the most parsimonious model from all possible subsets of the full model (19 models) (Burnham and Anderson, 2002). If models within two AICc of the top ranked model included predictors not included in the top model, we also discussed these alternate models. We excluded pre-treatment surveys from analyses to avoid spurious time × treatment interactions.

The properties of some animal variables meant that adjustments to the full model were necessary. Specifically, to adjust for over-dispersion of macropod scat data (φ = 6.9), rabbit digging data (φ = 2.7) and arthropod data (φ = 3.1), we added an observation-level random effect to the models for these variables (Harrison, 2014). Further, for the macropod model, we divided the partial herbivore treatment into two categories: partial – open months, and partial – closed months, to better describe this treatment. As there were low numbers of macropod scats in exclosure and partial – closed month treatments, they were excluded from this analysis. For the GLMM of antechinus captures, we ran model selection on all subsets of the full model of herbivores × burning × time, plus an alternate model with ‘season’ substituted for ‘time’ (giving

33 different models for comparison). This accounted for the strong seasonal variation in antechinus abundance (Lazenby-Cohen and Cockburn, 1991). Finally, as low mean counts for the two reptile species meant some models including the burn × time fixed effect did not converge, we excluded five models for the delicate skink and three models for the small-eyed snake from model comparisons. LMMs were performed using the lme function in the package

MASS, GLMMs using the glmer function in the package lme4 and AICc model ranking using the package AICcmodavg, within R version 3.0.1 (R Core Team, 2013).

Results

Vegetation structure

Vegetation structure responded both to the burning and herbivore exclosure treatments, as well as their interaction (Table 2). Understory vegetation cover at unburnt sites declined in open and partial treatments over time, but remained stable in exclosure sites (Fig. 1a). After an initial reduction after fire, a similar decline was observed for burnt, open sites. However, burnt partial

and exclosure sites remained stable. (Fig. 1a, Appendix 2). When bracken was excluded from understory cover, there was a strong exclosure × time interaction, as cover on open and partial exclosure sites declined or remained low, while on full exclosure sites cover increased through time (Fig. 1b, Appendix 2). The burning × time interaction was also important, as non-bracken vegetation increased over time on all burnt sites. Compared with other treatments, burnt, open sites had a high proportion of bracken, with very little non-bracken vegetation present across all time periods (Fig. 1b). Both understory height and leaf litter depth responded to the burning × time interaction, but not to any other interaction terms (Table 2). Understory height was reduced by fire but had recovered after 21 months (Fig. 2a). Litter depth recovered more slowly, remaining lower in burnt than unburnt sites across all time periods (Fig. 2b). The second ranked model for understory height also included an exclosure main effect, where vegetation was slightly taller on exclosure than on open sites (Appendix 2).

Table 2. Model rankings for linear mixed models (LMMs, vegetation) and generalised linear mixed models (GLMMs, animals), testing the fixed effects of burning (B - burnt or unburnt), exclosure (E - open, partial or full exclosure), time (T - sampling event - categorical), and their interactions on vegetation structure and animal occurrence. For brown antechinus, we also ran model ranking on all subsets of a model with season (S - autumn or spring) substituted for time (i.e. B*E*S), to account for the strong seasonal variation in the abundance of this species. K is the number of parameters estimated in the model, Δ AICc is the change in Aikaike’s Information Criterion (corrected for small sample size) from the best-ranked model. AICcWt is the Aikine Weight of the model, LL is the Log-likelihood.

Data Model terms K Δ AICc AICcWt LL LMMs Understory cover B + E + T + B:T + E:T + B:E + B:E:T 28 0 0.97 -1454.7 Cover excluding bracken B + E + T + B:T + E:T 20 0 0.74 -1389.6 Vegetation height B + T + B:T 12 0 0.61 266.5 B + E + T + B:T 14 1.59 0.28 267.8

Litter depth B + T + B:T 12 0 0.82 -429.5 GLMMs Macropod scats B + Ea + T 15 0 0.6 -500.6 Rabbit diggings T 9 0 0.37 -441.0 E + T 11 1.09 0.21 -439.2

B + T 10 1.64 0.16 -440.7

Brown antechinus E + season 6 0 0.36 -172.0 B + E + season 7 1.4 0.18 -171.5

Delicate skink B + E + T 13 1.01 0.46 -162.8 B + T 11 1.35 0.39 -165.3

Eastern small-eyed snake B 4 0 0.67 -60.4 Beetles and spiders T 5 0 0.61 -187.4 B + T 6 1.85 0.24 -187.0 a Compares only partial-open months with open treatment (see methods)

Open Partial Exclosure (a) 60 50 40 30 20 10

Understory cover cover (%)Understory 0 (b) 60 Burnt 50 Unburnt 40 30 20 10 0 Cover excl. (%)brackenexcl. Cover Aut'12 Spr'12 Aut'13 Spr'13 Aut'14 Aut'12 Spr'12 Aut'13 Spr'13 Aut'14 Aut'12 Spr'12 Aut'13 Spr'13 Aut'14

Time period

Figure 1. Vegetation structural responses to exclosure and burning treatments across sampling periods, (a) total understory percent cover, and (b) understory percent cover, excluding bracken (Pteridium esculentum). Values post-treatment are predicted means and estimated SE from the top-ranked models. Pre-treatment data (May 2012) were not included in the LMM, but are presented here (mean and SE) to allow comparison with post-treatment data. Arrows indicate timing of prescribed burning.

(a) 0.6 0.5 0.4 0.3 0.2 Burnt 0.1 Unburnt Understory height (m)heightUnderstory 0 (b) 60 50 40 30 20

Litter depth (mm)depthLitter 10 0 May '12 Nov '12 May '13 Nov '13 May '14

Time period

Figure 2. Response of understory height (a) and leaf litter depth (b) to burning treatment across time periods. Values post-treatment are predicted means and SE from the top-ranked models. Pre-treatment data (May 2012) were not included in the LMM, but are presented here (mean and SE) to allow comparison with post-treatment data. Arrows indicate timing of prescribed burning.

Vertebrate responses

Exclosure treatments successfully excluded macropod herbivores, with very low scat counts in exclosure treatments (푥̅ = 0.71 ± 0.52). When partial treatment gates were open, scat counts were 56% lower in partial than in open treatments (Fig. 3a). Macropods also responded to burning, showing higher activity in burnt than unburnt sites (Table 2, Fig 3). There were three competing models for rabbit diggings, with models including time, time plus exclosure and time plus burning all explaining similar levels of variation in the data (Table 2). Rabbit activity peaked in summer and tended to be higher in sites without macropods (partial and full exclosure treatments) and in burnt sites (Fig. 4, Appendix 2).

Figure 3. Number of macropod scats in burnt and unburnt sites through time in open (a) and partial exclosure sites (b). Note that closed months of the partial exclosures (Nov 2012, Apr 2013 etc.), and full exclosure sites were not included in the analysis as means for this group were too low to allow model fit. Values for open sites and partial-open months are predicted means and estimated SE from the top-ranked model. Arrows indicate timing of prescribed burning (note Aug ’12 counts were after implementation of herbivory treatments but before burning treatment).

Brown antechinus captures were highest in herbivore exclusion sites, and in Autumn

(May) surveys (Table 2, Fig. 5). The second ranked model also included burning as a fixed effect (Table 2), where antechinus captures were slightly lower in burnt than unburnt sites

(Appendix 2). Both the delicate skink and the eastern small-eyed snake were encountered more frequently in burnt sites than unburnt sites (Table 2, Fig. 6, Appendix 2). Delicate skink numbers also tended to be higher in open and partial than in full exclosure sites and were higher in the first survey (three months post-fire), than at any other time (Table 3, Fig, 6). Beetles and spiders captures were 65% higher in the first year (2012, 푥̅ = 31.3 ± 3.1) than the second (2013,

푥̅ = 18.9 ± 1.96), and the second ranked model indicated that captures were also slightly higher in burnt than in unburnt sites (Table 2, Appendix 2).

(a) 25 Open 20 Partial Exclosure 15

0 (b) rabbitNumberdiggings of 25 Burnt 20 Unburnt

Number of rabbitNumberdiggings of Jun '13 Aug '13 Oct '13 Dec '13 Feb '14 Apr '14 Time period

Figure 4. Number of rabbit (Oryctolagus cuniculus) diggings per site in the different exclosure treatments (a) and burning treatments (b) over time. Values are predicted means and estimated SE from the 2nd (a) and 3rd (b) ranked models respectively. The effect of time alone (first-ranked model) can be clearly seen in both plots.

5 Pre-treatment Post-treatment Open Partial 4 Exclosure 3

Number of individualsNumber of 0 May May November Season

Figure 5. Number of individual brown antechinus (Antechinus stuartii) captured per site in different seasons and herbivory treatments. Values post-treatment are predicted means and SE from the top-ranked model based on two years of data (November 2012 to May 2014). Pre-treatment data (May 2012) were not included in the GLMM, but are presented here (mean and SE) to allow comparison with post-treatment data.

Open Partial Exclosure 3.0

Burnt 2.5 Unburnt 2.0

1.5

1.0 Number of skinksNumber of 0.5

0.0 Nov Feb Apr Jun Aug Oct Dec Apr Nov Feb Apr Jun Aug Oct Dec Apr Nov Feb Apr Jun Aug Oct Dec Apr '12 '13 '13 '13 '13 '13 '13 '14 '12 '13 '13 '13 '13 '13 '13 '14 '12 '13 '13 '13 '13 '13 '13 '14

Time period

Figure 6. Number of delicate skinks (Lampropholis delicata) detected under artificial substrates in the different burning and exclosure treatments across time. Values are predicted means and SE from the top-ranked model.

Discussion

Fire and herbivory can interact strongly in space and time to shape the structure of vegetation communities (Koerner and Collins, 2014, Royo and Carson, 2006, Van Langevelde et al.,

2003). However, animal responses to the fire × herbivory interaction are rarely studied (but see

Fuhlendorf et al., 2010, Kimuyu et al., 2014, Kutt and Woinarski, 2007). In our experimental test of the interactive effects of fire and large herbivores, we found that forest understory structure responded to the fire × herbivore exclosure interaction, but vertebrate site occupancy was affected only by the individual effects of disturbance. This suggests that local changes in vegetation structure may not be an adequate predictor of animal responses to disturbance and that animals warrant individual consideration for the management of ecosystems that are subject to both recurring fire and herbivory.

Question 1: Changes to habitat structure

As expected, fire and herbivory both affected habitat structure, and effects differed with disturbance type. The limited recovery of non-bracken vegetation on burnt sites with high densities of large herbivores (Fig. 1), was consistent with previous studies (Meers and Adams,

2003, Tuft et al., 2012). Both of these previous studies attributed the stronger effect of herbivory on burnt sites to greater herbivore pressure, driven by the attraction of herbivores to the fresh plant growth following fire. This is a commonly reported mechanism explaining fire × herbivore interactions (Klop, van Goethem and de Iongh, 2007), for which we also found evidence, as indicated by macropod activity being greatest on recently burnt sites (Fig. 2b).

In contrast to the non-bracken vegetation, bracken recovered well in burnt, open sites, and made up a large proportion of the vegetation in these sites (Fig. 1). Bracken can regenerate rapidly following fire and suppress other plants. However, after an initial post-fire pulse, bracken cover usually declines over time, as other plants become dominant (Tolhurst and

Turvey, 1992). Our results indicate that abundant macropod herbivores may be disrupting this successional process by selectively feeding on more palatable vegetation, maintaining the bracken-dominated understory. As prescribed fire is commonly used in this system to reduce forest fuel loads and promote vegetation heterogeneity and floristic diversity (Morrison et al.,

1996), our results suggest that herbivore management following fire may be important to maintain a heterogeneous forest flora.

Question 2: Habitat use by vertebrates

Despite the strong effect of the fire × herbivory interaction on vegetation, habitat use by vertebrates was affected only by the individual effects of disturbances and not their interaction.

The increased rabbit activity we observed on herbivore exclusion and recently burnt sites was consistent with previous studies of small herbivore responses to large herbivore removal

(Keesing, 1998) and fire (Leigh et al., 1987, Moreno and Villafuerte, 1995). Competitive release of small herbivores following a reduction in large herbivore densities can result in increased herbivory by small herbivores, with subsequent impacts on vegetation communities

(Lagendijk, Page and Slotow, 2012). Our results suggest that control of native macropod herbivores may favour introduced rabbits, particularly after fire when fresh plant growth is abundant. Therefore, management of the native herbivore population should carefully consider

the potential for competitive release of the European rabbit, which is a destructive pest species in Australia (Davey et al., 2006).

Antechinus responded positively to large herbivore exclusion, with more individuals captured in sites without macropod herbivores, regardless of burning treatment. Despite the small size of our exclosures, this response likely indicates a preference for herbivore exclusion areas, as antechinus have small foraging ranges (approximately 0.4 ha for females and 0.9 ha for males, Lazenby-Cohen and Cockburn, 1991). Further, our result is consistent with Pedersen et al. (2014), who found that antechinus captures were negatively correlated with wallaby density.

The preference of antechinus for sites without large herbivores may have been due to the dense understory cover in exclosure sites (Bennett, 1993, Knight and Fox, 2000), providing increased foraging habitat (antechinus are scansorial) and/or greater protection from predators (e.g. Stokes et al., 2004). Other studies have found that some small mammal species prefer habitats with lower densities of large herbivores (Bush et al., 2012, Keesing, 1998, Kutt and Gordon, 2012), but our study is the first to experimentally demonstrate this response to macropod herbivores.

Contrary to other studies from south-eastern Australia (Fox, 1982, Lindenmayer et al.,

2008), we found only weak support for a negative response of antechinus to fire. This was likely due to the low intensity, small-scale fires used in our study. The availability of arthropod prey

(Table 2, Appendix 2), combined with the persistence of logs and the proximity of unburnt vegetation to burnt areas in our study, may have sufficiently maintained habitat quality for antechinus. It is likely that antechinus may have responded more strongly to a larger-scale, or higher intensity burn (Lindenmayer et al., 2008, Penn et al., 2003).

Delicate skinks were more common in open than in herbivore exclusion sites, and both delicate skinks and small-eyed snakes were more common in burnt than unburnt sites. The negative response of delicate skinks to herbivore exclusion was likely due to increased shading from recovering vegetation, which could have reduced the thermal suitability of the environment for this heliothermic species (Howard et al., 2003). The positive response of the delicate skink to burning was contrary to our expectation for this species, which generally

recovers slowly from fire, in response to the slow accumulation of leaf litter (Taylor and Fox,

2001). Our result may therefore reflect a change in micro-habitat use following burning, rather than a change in site occupancy. Burning reduced litter depth, and in the absence of suitable leaf-litter habitat, the delicate skink may have increased its use of the artificial survey substrates for shelter, basking and foraging, as found by Croft, Reid and Hunter (2010).

A lack of alternative refuges also may explain the higher numbers of small-eyed snakes under the artificial substrates in burnt sites. However, there are two other possible explanations for this response: First is that the small-eyed snake was more common after burning as a result of reduced shading of substrates, which provided warmer, more desirable diurnal refuges (Webb et al., 2004). Second, the density of skinks, a key prey item for small-eyed snakes (Shine, 1984), may have attracted snakes to the substrates in burnt sites. Previous studies of small-eyed snakes show that reduced shading after fire can improve the thermal properties of diurnal refuges

(Webb et al., 2005), but large wildfire can lead to population declines, possibly due to increased predation (Webb and Shine, 2008). While the possible biases in our reptile sampling technique mean results should be interpreted with caution, the greater numbers of reptiles in burnt sites and skinks in the open treatments, indicate that maintaining areas of open understory may be important for the persistence of reptiles in these forests (Webb et al., 2005).

Conservation implications

Our results have four key implications relevant to the conservation of ecosystems subject to both recurrent fire and herbivory. First, the dominance of bracken and limited recovery of other vegetation in burnt, open sites indicates that short-term management of abundant macropod herbivores following prescribed fire may be useful for the conservation of structurally and floristically complex vegetation. Second, the preference of antechinus for herbivore exclusion sites suggests that management of native herbivores to promote dense understory habitat is also likely to benefit small mammals dependent on such habitats. Third, the increased rabbit activity we observed on recently burnt sites, and sites without macropods, suggests that any plan to improve the post-fire recovery of vegetation by controlling native herbivores also should

include management of rabbits. Fourth, the positive response of reptiles to the open understory of burnt sites and sites with more macropods differed from the antechinus response, and suggests that a mixed management scenario might be more appropriate.

While our experiments were small-scale, the contrasting responses of different species to herbivory and fire indicate that a mixed management strategy promoting a heterogeneous understory may be important for the persistence of all species in our study. Heterogeneous landscapes have commonly been suggested as a desired goal of land management, as such landscapes are more likely to allow the co-existence of species with different niches, as well as species that require a mix of habitats (Law and Dickman, 1998, Stein, Gerstner and Kreft,

2014). In Australian landscapes, fire patch-mosaics have been recommended to promote fauna diversity, although key questions around the appropriate spatial and temporal scales of such mosaics remain unanswered (Allouche et al., 2012, Clarke, 2008, Driscoll et al., 2010). While not designed to address questions of spatial scale, our study suggests that maintaining a mix of habitat types and conditions may be important for fauna in forested systems.

Although both fire and herbivory are often actively managed in forested systems

(Gordon, Hester and Festa-Bianchet, 2004, Morrison et al., 1996), these processes are usually considered independently (Royo and Carson, 2006, Wisdom et al., 2006). However, the interactive effects of fire and herbivory observed in our study indicate that integrating large herbivore management with fire management practices is likely to be important for achieving vegetation heterogeneity in forests. This could be through the fire-dependent management of herbivores (e.g. controlling large herbivores across only part of a burn or after only some prescribed burns), or through planning fires to consider large herbivore behaviour (e.g. reducing the edge-area ratio of prescribed burns, as macropods can be reluctant to enter open areas and so feed more heavily at the edge (While and McArthur, 2006)). Our study shows the value of experimental studies that quantify disturbance responses both individually and collectively, and highlights the importance of considering both multiple stressors, and multiple species, in the management of disturbance regimes.

Acknowledgments

Staff of BNP conducted prescribed burns. We thank the Department of Environment and the landowners and co-managers of BNP - the Wreck Bay Aboriginal Community. The Margaret

Middleton Fund, The Norman Wettenhall Foundation and the Long Term Ecological Research

Network provided financial support.

References

Allouche, O., Kalyuzhny, M., Moreno-Rueda, G., Pizarro, M. & Kadmon, R. (2012). Area– heterogeneity tradeoff and the diversity of ecological communities. Proceedings of the National Academy of Sciences 109, 17495-17500. Bailey, J. K. & Whitham, T. G. (2002). Interactions among fire, aspen, and elk affect insect diversity: Reversal of a community response. Ecology 83, 1701-1712. Bennett, A. F. (1993). Microhabitat use by the long-nosed potoroo, Potorous tridactylus, and other small mammals in remnant forest vegetation, south-western Victoria. Wildl. Res. 20, 267-285. Bragg, J. G., Taylor, J. E. & Fox, B. J. (2005). Distributions of lizard species across edges delimiting open-forest and sand-mined areas. Austral Ecol. 30, 188-200. Burnham, K. P. & Anderson, D. R. (2002). Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. edn. New York: Springer-Verlag. Bush, E. R., Buesching, C. D., Slade, E. M. & Macdonald, D. W. (2012). Woodland recovery after suppression of deer: cascade effects for small mammals, wood mice (Apodemus sylvaticus) and bank voles (Myodes glareolus). PLoS ONE 7, e31404. Catling, P. & Burt, R. (1995). Studies of the ground-dwelling mammals of eucalypt forests in south-eastern New South Wales: the effect of habitat variables on distribution and abundance. Wildl. Res. 22, 271-288. Clarke, M. F. (2008). Catering for the needs of fauna in fire management: science or just wishful thinking? Wildl. Res. 35, 385-394. Croft, P., Reid, N. & Hunter, J. (2010). Experimental burning changes the quality of fallen timber as habitat for vertebrate and invertebrate fauna: implications for fire management. Wildl. Res. 37, 574-581. Davey, C., Sinclair, A., Pech, R., Arthur, A., Krebs, C., Newsome, A., Hik, D., Molsher, R. & Allcock, K. (2006). Do exotic vertebrates structure the biota of Australia? An experimental test in New South Wales. Ecosystems 9, 992-1008.

100

Davis, N. E., Coulson, G. & Forsyth, D. M. (2008). Diets of native and introduced mammalian herbivores in shrub-encroached grassy woodland, south-eastern Australia. Wildl. Res. 35, 684-694. Dexter, N., Hudson, M., James, S., MacGregor, C. & Lindenmayer, D. B. (2013). Unintended consequences of invasive predator control in an Australian forest: Overabundant wallabies and vegetation change. PloS ONE 8, e69087. Dexter, N., Ramsey, D. L., MacGregor, C. & Lindenmayer, D. (2012). Predicting ecosystem wide impacts of wallaby management using a fuzzy cognitive map. Ecosystems 15, 1363- 1379. Didham, R. K., Tylianakis, J. M., Gemmell, N. J., Rand, T. A. & Ewers, R. M. (2007). Interactive effects of habitat modification and species invasion on native species decline. Trends Ecol. Evol. 22, 489-496. Driscoll, D. A., Lindenmayer, D. B., Bennett, A. F., Bode, M., Bradstock, R. A., Cary, G. J., Clarke, M. F., Dexter, N., Fensham, R., Friend, G., Gill, M., James, S., Kay, G., Keith, D. A., MacGregor, C., Russell-Smith, J., Salt, D., Watson, J. E. M., Williams, R. J. & York, A. (2010). Fire management for biodiversity conservation: Key research questions and our capacity to answer them. Biol. Conserv. 143, 1928-1939. Foster, C. N., Barton, P. S. & Lindenmayer, D. B. (2014). Effects of large native herbivores on other animals. J. Appl. Ecol. 51, 929–938. Fox, B. J. (1982). Fire and mammalian secondary succession in an Australian coastal heath. Ecology 63, 1332-1341. Fuhlendorf, S. D., Townsend, D. E., Elmore, R. D. & Engle, D. M. (2010). Pyric-herbivory to promote rangeland heterogeneity: Evidence from small mammal communities. Rangeland Ecol. Manage. 63, 670-678. Gordon, I. J., Hester, A. J. & Festa-Bianchet, M. (2004). The management of wild large herbivores to meet economic, conservation and environmental objectives. J. Appl. Ecol. 41, 1021-1031. Harrison, X. A. (2014). Using observation-level random effects to model overdispersion in count data in ecology and evolution. PeerJ 2, e616. Hobbs, R. J. & Huenneke, L. F. (1992). Disturbance, diversity, and invasion: Implications for conservation. Conserv. Biol. 6, 324-337. Howard, R., Williamson, I. & Mather, P. (2003). Structural aspects of microhabitat selection by the skink Lampropholis delicata. J. Herpetol. 37, 613-617. Howland, B., Stojanovic, D., Gordon, I. J., Manning, A. D., Fletcher, D. & Lindenmayer, D. B. (2014). Eaten out of house and home: Impacts of grazing on ground-dwelling reptiles in Australian grasslands and grassy woodlands. PLoS ONE 9, e105966.

101

Johnson, C. N., Jarman, P. J. & Southwell, C. J. (1987). Macropod studies at Wallaby Creek. 5. Patterns of defecation by eastern gray kangaroos and red-necked wallabies. Wildl. Res. 14, 133-138. Keesing, F. (1998). Impacts of ungulates on the demography and diversity of small mammals in central Kenya. Oecologia 116, 381-389. Kimuyu, D. M., Sensenig, R. L., Riginos, C., Veblen, K. E. & Young, T. P. (2014). Native and domestic browsers and grazers reduce fuels, fire temperatures, and acacia ant mortality in an African savanna. Ecol. Appl. 24, 741-749. Klop, E., van Goethem, J. & de Iongh, H. H. (2007). Resource selection by grazing herbivores on post-fire regrowth in a West African woodland savanna. Wildl. Res. 34, 77-83. Knight, E. H. & Fox, B. J. (2000). Does habitat structure mediate the effects of forest fragmentation and human-induced disturbance on the abundance of Antechinus stuartii? Aust. J. Zool. 48, 577-595. Koerner, S. E. & Collins, S. L. (2014). Interactive effects of grazing, drought, and fire on grassland plant communities in North America and South Africa. Ecology 95, 98-109. Kutt, A. S. & Gordon, I. J. (2012). Variation in terrestrial mammal abundance on pastoral and conservation land tenures in north-eastern Australian tropical savannas. Anim. Conserv. 15, 416-425. Kutt, A. S. & Woinarski, J. C. Z. (2007). The effects of grazing and fire on vegetation and the vertebrate assemblage in a tropical savanna woodland in north-eastern Australia. J. Trop. Ecol. 23, 95-106. Lagendijk, G., Page, B. R. & Slotow, R. (2012). Short-term effects of single species browsing release by different-sized herbivores on sand forest vegetation community, South Africa. Biotropica 44, 63-72. Law, B. S. & Dickman, C. R. (1998). The use of habitat mosaics by terrestrial vertebrate fauna: implications for conservation and management. Biodiversity & Conservation 7, 323-333. Lazenby-Cohen, K. A. & Cockburn, A. (1991). Social and foraging components of the home range in Antechinus stuartii (Dasyuridae: Marsupialia). Aust. J. Ecol. 16, 301-307. Leigh, J., Wimbush, D., Wood, D., Holgate, M., Slee, A., Stanger, M. & Forrester, R. (1987). Effects of rabbit grazing and fire on a sub-alpine environment .I. Herbaceous and shrubby vegetation. Aust. J. Bot. 35, 433-464. Lindenmayer, D., MacGregor, C., Dexter, N., Fortescue, M. & Beaton, E. (2014). Booderee National Park, The Jewel of Jervis Bay. Collingwood, Australia: CSIRO Publishing. Lindenmayer, D. B., Likens, G. E., Krebs, C. J. & Hobbs, R. J. (2010). Improved probability of detection of ecological “surprises”. Proceedings of the National Academy of Sciences 107, 21957-21962. Lindenmayer, D. B., MacGregor, C., Welsh, A., Donnelly, C., Crane, M., Michael, D., Montague-Drake, R., Cunningham, R., Brown, D., Fortescue, M., Dexter, N., Hudson, M.

102

& Gill, A. M. (2008). Contrasting mammal responses to vegetation type and fire. Wildl. Res. 35, 395-408. Meers, T. & Adams, R. (2003). The impact of grazing by eastern grey kangaroos (Macropus giganteus) on vegetation recovery after fire at Reef Hills Regional Park, Victoria. Ecol. Manage. Restor. 4, 126-132. Moreno, S. & Villafuerte, R. (1995). Traditional management of scrubland for the conservation of rabbits Oryctolagus cuniculus and their predators in Doñana National Park, Spain. Biol. Conserv. 73, 81-85. Morrison, D. A., Buckney, R. T., Bewick, B. J. & Cary, G. J. (1996). Conservation conflicts over burning bush in south-eastern Australia. Biol. Conserv. 76, 167-175. Noss, R. F., Franklin, J. F., Baker, W. L., Schoennagel, T. & Moyle, P. B. (2006). Managing fire-prone forests in the western United States. Front. Ecol. Environ. 4, 481-487. Pedersen, S., Andreassen, H. P., Keith, D. A., Skarpe, C., Dickman, C. R., Gordon, I. J., Crowther, M. S. & McArthur, C. (2014). Relationships between native small mammals and native and introduced large herbivores. Austral Ecol. 39, 236-243. Penn, A. M., Sherwin, W. B., Lunney, D. & Banks, P. B. (2003). The effects of a low-intensity fire on small mammals and lizards in a logged, burnt forest. Wildl. Res. 30, 477-486. Pinheiro, J. & Bates, D. (2000). Mixed-Effects Models in S and S-PLUS. New York, NY: Springer-Verlag. R Core Team (2013). A language and environment for statistical computing. Vienna, Austria. URL http://www.R-project.org/: R Foundation for Statistical Computing. Royo, A. A. & Carson, W. P. (2006). On the formation of dense understory layers in forests worldwide: consequences and implications for forest dynamics, biodiversity, and succession. Can. J. For. Res. 36, 1345-1362. Royo, A. A., Collins, R., Adams, M. B., Kirschbaum, C. & Carson, W. P. (2010). Pervasive interactions between ungulate browsers and disturbance regimes promote temperate forest herbaceous diversity. Ecology 91, 93-105. Shine, R. (1984). Reproductive biology and food habits of the Australian elapid snakes of the genus Cryptophis. J. Herpetol. 18, 33-39. Sinclair, A. R. E. & Byrom, A. E. (2006). Understanding ecosystem dynamics for conservation of biota. J. Anim. Ecol. 75, 64-79. Smit, R., Bokdam, J., den Ouden, J., Olff, H., Schot-Opschoor, H. & Schrijvers, M. (2001). Effects of introduction and exclusion of large herbivores on small rodent communities. Plant Ecol. 155, 119-127. Stein, A., Gerstner, K. & Kreft, H. (2014). Environmental heterogeneity as a universal driver of species richness across taxa, biomes and spatial scales. Ecol. Lett. 17, 866-880.

103

Stokes, V. L., Pech, R. P., Banks, P. B. & Arthur, A. D. (2004). Foraging behaviour and habitat use by Antechinus flavipes and Sminthopsis murina (Marsupialia: Dasyuridae) in response to predation risk in eucalypt woodland. Biol. Conserv. 117, 331-342. Taws, N. (1998). Fire and vegetation management in Jervis Bay Territory, Report. Canberra: Environment Australia. Taylor, J. E. & Fox, B. J. (2001). Disturbance effects from fire and mining produce different lizard communities in eastern Australian forests. Austral Ecol. 26, 193-204. Tolhurst, K. G. & Turvey, N. D. (1992). Effects of bracken (Pteridium esculentum (forst. f.) Cockayne) on eucalypt regeneration in west-central Victoria. For. Ecol. Manage. 54, 45- 67. Tuft, K. D., Crowther, M. S. & McArthur, C. (2012). Fire and grazing influence food resources of an endangered rock-wallaby. Wildl. Res. 39, 436-445. Tylianakis, J. M., Didham, R. K., Bascompte, J. & Wardle, D. A. (2008). Global change and species interactions in terrestrial ecosystems. Ecol. Lett. 11, 1351-1363. Van Langevelde, F., Van De Vijver, C. A. D. M., Kumar, L., Van De Koppel, J., De Ridder, N., Van Andel, J., Skidmore, A. K., Hearne, J. W., Stroosnijder, L., Bond, W. J., Prins, H. H. T. & Rietkerk, M. (2003). Effects of fire and herbivory on the stability of savanna ecosystems. Ecology 84, 337-350. Webb, J. K., Pringle, R. M., Shine, R. & Lannoo, M. J. (2004). How do nocturnal snakes select diurnal retreat sites? Copeia 2004, 919-925. Webb, J. K. & Shine, R. (2008). Differential effects of an intense wildfire on survival of sympatric snakes. J. Wildl. Manage. 72, 1394-1398. Webb, J. K., Shine, R., Pringle, R. M. & Lannoo, M. J. (2005). Canopy removal restores habitat quality for an endangered snake in a fire suppressed landscape. Copeia 2005, 894-900. While, G. M. & McArthur, C. (2006). Distance from cover affects artificial food-patch depletion by macropod herbivores. Wildl. Res. 33, 565-570. Willig, M. R. & Walker, L. R. (1999). Disturbance in terrestrial ecosystems: Salient themes, synthesis, and future directions. In Ecosystems of the World 16: Ecosystems of Disturbed Ground: 747-768. Walker, L. R. (Ed.). Amsterdam: Elsevier Science. Wisdom, M. J., Vavra, M., Boyd, J. M., Hemstrom, M. A., Ager, A. A. & Johnson, B. K. (2006). Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildl. Soc. Bull. 34, 283-292. Zuur, A. F., Ieno, E. N., Walker, N. J., Saveliev, A. A. & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R. New York, USA: Springer.

104

Appendices

Appendix 1. Spatial scale and construction of experimental treatments.

Figure 1. Schematic diagram of blocked design showing types and site dimensions of the six treatment combinations (not to scale). Exact spatial arrangement varies between blocks

Figure 2. Example of a recently burnt herbivore exclusion site (3 months post-fire), showing fence design and forest type.

105

Appendix 2. Full model rankings for LMMs and GLMMs, and parameter estimates for fixed-effect terms in top-ranked models.

Table 1. Full model rankings for linear mixed models testing the fixed effects of burning (B - burnt or unburnt), exclosure (E - open, partial or full exclosure), time (T - sampling event - categorical), and their interactions on vegetation structure. K is the number of parameters estimated in the model, Δ AICc is the change in Aikaike’s Information Criterion (corrected for small sample size) from the best-ranked model. AICcWt is the Aikine Weight of the model, LL is the Log- likelihood.

Data Model terms K AICc ΔAICc AICcWt Cum.Wt LL Understory cover Full model (BxExT) 28 2970.0 0 0.97 0.97 -1454.7 B + E + T + B:T + H:T 20 2977.2 7.22 0.03 1 -1467.4 B + E + T + B:E: + B:T + E:T 22 2980.9 10.92 0 1 -1467.0 B + T + B:T 12 2985.8 15.81 0 1 -1480.5 B + E + T + B:T 14 2987.1 17.14 0 1 -1479.0 B + E + T + B:E: + B:T 16 2990.7 20.69 0 1 -1478.6 B + E + T + E:T 17 3088.1 118.14 0 1 -1526.2 B + T 9 3089.4 119.48 0 1 -1535.5 B + E + T 11 3090.7 120.74 0 1 -1534.0 B + E + T + B:E: + E:T 19 3091.7 121.77 0 1 -1525.8 B + E + T + B:E 13 3094.2 124.22 0 1 -1533.6 E + T + E:T 16 3095.8 125.83 0 1 -1531.2 E + T 8 3096.2 126.2 0 1 -1539.9 E + T 10 3098.5 128.5 0 1 -1538.9 B 6 3206.4 236.39 0 1 -1597.1 B + E 8 3207.6 237.58 0 1 -1595.6 B + E + B:E 10 3211.0 241 0 1 -1595.2 intercept 5 3213.1 243.14 0 1 -1601.5 E 7 3215.4 245.38 0 1 -1600.5

Cover excluding bracken B + E + T + B:T + E:T 20 2821.4 0 0.74 0.74 -1389.6 Full model (BxExT) 28 2824.5 3.02 0.16 0.9 -1381.9 B + E + T + B:E: + B:T + E:T 22 2825.4 3.94 0.1 1 -1389.3 B + E + T + B:T 14 2854.7 33.23 0 1 -1412.8 B + T + B:T 12 2855.9 34.47 0 1 -1415.5 B + E + T + B:E + B:T 16 2858.5 37.02 0 1 -1412.5 B + E + T + E:T 17 2902.1 80.7 0 1 -1433.2 B + E + T + B:E: + E:T 19 2906.0 84.56 0 1 -1433.0 E + T + E:T 16 2917.6 96.13 0 1 -1442.0 B + E + T 11 2924.2 102.71 0 1 -1450.7 B + T 9 2925.5 104.02 0 1 -1453.5 B + E + T + B:E 13 2927.9 106.43 0 1 -1450.4 B + E 8 2933.3 111.85 0 1 -1458.5

106

Data Model terms K AICc ΔAICc AICcWt Cum.Wt LL B 6 2934.7 113.23 0 1 -1461.2 B + E + B:E 10 2936.9 115.5 0 1 -1458.2 T 8 2937.9 116.49 0 1 -1460.8 E + T 10 2939.7 118.21 0 1 -1459.5 intercept 5 2947.2 125.73 0 1 -1468.5 E 7 2948.8 127.38 0 1 -1467.3

Vegetation height B + T + B:T 12 -508.1 0 0.61 0.61 266.5 B + E + T + B:T 14 -506.5 1.59 0.28 0.89 267.8 B + E + T + B:E + B:T 16 -504.4 3.71 0.1 0.98 268.9 B + E + T + B:T + H:T 20 -500.0 8.03 0.01 1 271.2 B + E + T + B:E: + B:T + E:T 22 -497.8 10.29 0 1 272.3 Full model (BxExT) 28 -494.3 13.77 0 1 277.4 B + T 9 -448.4 59.67 0 1 233.4 B + E + T 11 -446.9 61.2 0 1 234.8 B + E + T + B:E 13 -444.8 63.24 0 1 235.9 B + E + T + E:T 17 -439.3 68.81 0 1 237.5 T 8 -438.7 69.37 0 1 227.5 B + E + T + B:E: + E:T 19 -437.1 71 0 1 238.6 E + T 10 -435.9 72.12 0 1 228.3 E + T + E:T 16 -428.4 79.66 0 1 230.9 B 6 -331.4 176.68 0 1 171.8 B + E 8 -329.9 178.14 0 1 173.2 B + E + B:E 10 -328.0 180.11 0 1 174.3 intercept 5 -321.7 186.41 0 1 165.9 E 7 -319.0 189.09 0 1 166.6

Litter depth B + T + B:T 12 883.9 0 0.82 0.82 -429.51 B + E + T + B:T 14 887.6 3.7 0.13 0.95 -429.21 B + E + T + B:E + B:T 16 890.4 6.55 0.03 0.99 -428.46 B + E + T + B:T + H:T 20 892.4 8.53 0.01 1 -425.03 B + E + T + B:E: + B:T + E:T 22 895.4 11.52 0 1 -424.29 Full model (BxExT) 28 904.4 20.49 0 1 -421.89 B + T 9 942.8 58.94 0 1 -462.16 B + E + T 11 946.4 62.57 0 1 -461.86 B + E + T + B:E 13 949.2 65.35 0 1 -461.11 B + E + T + E:T 17 952.8 68.9 0 1 -458.54 B + E + T + B:E: + E:T 19 955.7 71.82 0 1 -457.8 T 8 989.9 106.06 0 1 -486.77 E + T 10 994.1 110.21 0 1 -486.74 E + T + E:T 16 1000.3 116.47 0 1 -483.42 B 6 1023.8 139.9 0 1 -505.77 B + E 8 1027.3 143.46 0 1 -505.47 B + E + B:E 10 1030.0 146.18 0 1 -504.72 intercept 5 1070.9 187.05 0 1 -530.38 E 7 1075 191.14 0 1 -530.35 107

Table 2. Parameter estimates (best), standard errors (SE), degrees of freedom (df) and marginal probabilities (P) of fixed-effect terms for linear mixed models within two AICc of the top-ranked model for the vegetation variables.

Model Model fixed effects Factor levels best SE df P a. Understory cover Top ranked model: full model (BxHxT) Burn burnt -31.71 6.78 15 <0.001 Exclosure partial 3.94 6.78 15 0.570 full exclosure -6.25 6.78 15 0.371 Time May'13 2.69 3.14 270 0.392 Nov'13 -10.44 3.14 270 0.001 May'14 -8.75 3.14 270 0.006 Burn x Exclosure burnt.partial -1.66 9.59 15 0.865 burnt.full 4.68 9.59 15 0.632 Burn x Time burnt.May'13 21.65 4.43 270 <0.001 burnt.Nov'13 31.03 4.43 270 <0.001 burnt.May'14 25.03 4.43 270 <0.001 Exclosure x Time partial.May'13 4.63 4.43 270 0.298 full.May'13 19.88 4.43 270 <0.001 partial.Nov'13 7.75 4.43 270 0.082 full.Nov'13 18.44 4.43 270 <0.001 partial.May'14 -3.94 4.43 270 0.375 full.May'14 14.63 4.43 270 0.001 Burn x Exclosure x Time burnt.partial.May'13 7.35 6.27 270 0.242 burnt.full.May'13 -16.68 6.27 270 0.008 burnt.partial.Nov'13 -2.65 6.27 270 0.673 burnt.full.Nov'13 -14.81 6.27 270 0.019 burnt.partial.May'14 14.98 6.27 270 0.018 burnt.full.May'14 -4.62 6.27 270 0.462

b. Understory cover excluding bracken Top ranked model: B + E + T + B:T + E:T Burn burnt -26.60 3.05 17 <0.001 Exclosure partial 3.67 3.73 17 0.340 full exclosure -3.14 3.73 17 0.412 Time May'13 -9.21 2.13 276 <0.001 Nov'13 -13.00 2.13 276 <0.001 May'14 -12.99 2.13 276 <0.001 Burn x Time burnt.May'13 12.83 2.13 276 <0.001 burnt.Nov'13 16.98 2.13 276 <0.001 burnt.May'14 19.24 2.13 276 <0.001 Exclosure x Time partial.May'13 9.93 2.61 276 <0.001 full.May'13 12.61 2.61 276 <0.001 partial.Nov'13 3.71 2.61 276 0.156 full.Nov'13 11.83 2.61 276 <0.001 partial.May'14 3.06 2.61 276 0.241 full.May'14 13.68 2.61 276 <0.001

108

Model Model fixed effects Factor levels best SE df P c. Vegetation height Top ranked model: B + T + B:T Burn burnt -0.301 0.046 19 <0.001 Time May'13 0.047 0.020 282 0.021 Nov'13 0.030 0.020 282 0.132 May'14 0.084 0.020 282 <0.001 Burn x Time burnt.May'13 0.114 0.028 282 <0.001 burnt.Nov'13 0.200 0.028 282 <0.001 burnt.May'14 0.216 0.028 282 <0.001

2nd ranked model: B + E + T + B:T Burn burnt -0.301 0.043 17 <0.001 Exclosure partial 0.036 0.049 17 0.476 full exclosure 0.082 0.049 17 0.112 Time May'13 0.047 0.020 282 0.021 Nov'13 0.030 0.020 282 0.133 May'14 0.084 0.020 282 <0.001 Burn x Time burnt.May'13 0.114 0.029 282 <0.001 burnt.Nov'13 0.200 0.029 282 <0.001 burnt.May'14 0.216 0.029 282 <0.001

d. Litter Depth Top ranked model: B + T + B:T Burn burnt -2.837 0.197 19 <0.001 Time May'13 0.496 0.126 282 <0.001 Nov'13 0.498 0.126 282 <0.001 May'14 0.240 0.126 282 0.059 Burn x Time burnt.May'13 -0.008 0.179 282 0.964 burnt.Nov'13 0.692 0.179 282 <0.001 burnt.May'14 1.275 0.179 282 <0.001

109

Table 3. Full model rankings for generalised linear mixed models testing the fixed effects of burning (B - burnt or unburnt), exclosure (E - open, partial or full exclosure), time (T - sampling event - categorical), and their interactions on animal occurrence. For brown antechinus, we also ran model ranking on all subsets of a model with season (S - autumn or spring) substituted for time (i.e. B*E*S), to account for the strong seasonal variation in the abundance of this species. K is the number of parameters estimated in the model, Δ AICc is the change in Aikaike’s Information Criterion (corrected for small sample size) from the best-ranked model. AICcWt is the Aikine Weight of the model, LL is the Log-likelihood.

Data Model terms K AICc Δ AICc AICcWt Cum.Wt LL Macropod scats B + E + T 15 1035.9 0 0.6 0.6 -500.6 B + E + T 15 1035.9 0 0.6 0.6 -500.6 B + E + T + B:E 16 1038.4 2.52 0.17 0.77 -500.6 E + T 14 1038.7 2.8 0.15 0.92 -503.3 B + E + T + E :T 19 1041.1 5.21 0.04 0.96 -497.7 E + T + E :T 18 1043.8 7.9 0.01 0.98 -500.5 B + E + T + B:E : + E :T 20 1043.8 7.93 0.01 0.99 -497.7 B + T 14 1045.7 9.81 0 0.99 -506.8 T 13 1046.0 10.16 0 1 -508.3 B + E + T + B:T 24 1046.1 10.23 0 1 -492.7 B + E + T + B:E + B:T 25 1049.0 13.17 0 1 -492.6 B + E + T + B:T + H:T 28 1052.6 16.74 0 1 -489.4 B + E + T + B:E : + B:T + E :T 29 1055.8 19.91 0 1 -489.2 B + T + B:T 23 1056.0 20.16 0 1 -499.3 B + E 6 1056.4 20.49 0 1 -521.8 B + E + B:E 7 1058.4 22.57 0 1 -521.7 E 5 1059.3 23.41 0 1 -524.4 Full model (BxE xT) 33 1062.2 26.36 0 1 -485.1 B 5 1066.3 30.43 0 1 -527.9 intercept 4 1066.9 31.02 0 1 -529.3

Rabbit diggings T 9 901.4 0 0.37 0.37 -441.0 E + T 11 902.4 1.09 0.21 0.58 -439.2 B + T 10 903.0 1.64 0.16 0.74 -440.7 B + E + T 12 903.9 2.59 0.1 0.84 -438.8 B + T + B:T 15 904.0 2.69 0.1 0.94 -435.1 B + E + T + B:T 17 905.4 4.03 0.05 0.98 -433.3 B + E + T + B:E 14 908.3 6.96 0.01 1 -438.5 B + E + T + B:E + B:T 19 910.1 8.73 0 1 -433.0 E + T + E:T 21 920.8 19.48 0 1 -435.6 B + E + T + E:T 22 922.7 21.38 0 1 -435.2 B + E + T + B:T + H:T 27 925.0 23.62 0 1 -429.0 B + E + T + B:E: + E:T 24 927.9 26.56 0 1 -434.9 B + E + T + B:E: + B:T + E:T 29 930.5 29.16 0 1 -428.6 intercept 4 939.6 38.29 0 1 -465.7 E 6 940.3 38.98 0 1 -463.9

110

Data Model terms K AICc Δ AICc AICcWt Cum.Wt LL B 5 941.2 39.82 0 1 -465.4 B + E 7 941.7 40.37 0 1 -463.5 B + E + B:E 9 945.7 44.37 0 1 -463.2 Full model (BxExT) 39 958.8 57.4 0 1 -425.4

Brown antechinus E + S 6 356.9 0 0.36 0.36 -172.0 B + E + S 7 358.3 1.4 0.18 0.54 -171.5 S 4 359.8 2.92 0.08 0.62 -175.7 E + T 8 360.4 3.47 0.06 0.68 -171.4 E + S+ E:S 8 360.4 3.49 0.06 0.74 -171.4 B + E + S + B:S 8 360.5 3.65 0.06 0.8 -171.4 B + S 5 361.4 4.52 0.04 0.84 -175.4 B + E + S + B:E 9 361.5 4.6 0.04 0.87 -170.7 B + E + T 9 361.9 4.99 0.03 0.9 -170.9 B + E + S + E:S 9 361.9 5 0.03 0.93 -170.9 T 6 363.1 6.18 0.02 0.95 -175.1 B + S + B:S 6 363.5 6.66 0.01 0.96 -175.3 B + E + S + B:E + B:S 10 363.8 6.96 0.01 0.97 -170.6 B + E + S + B:S + E:S 10 364.2 7.36 0.01 0.98 -170.8 B + T 7 364.8 7.89 0.01 0.99 -174.7 B + E + T + B:E 11 365.3 8.41 0.01 0.99 -170.1 B + E + S + B:E + E:S 11 365.3 8.43 0.01 1 -170.1 B + E + S + B:E + B:S + E:S 12 367.8 10.97 0 1 -170.0 B + E + T + B:T 12 369.3 12.4 0 1 -170.8 Full model (BxExS) 14 371.1 14.21 0 1 -169.0 B + T + B:T 10 371.8 14.95 0 1 -174.6 B + E + T + B:E + B:T 14 373.1 16.2 0 1 -169.9 E + T + E:T 14 373.9 17 0 1 -170.3 B + E + T + E:T 15 375.8 18.89 0 1 -169.9 B + E + T + B:E + E:T 17 380.0 23.11 0 1 -169.1 E 5 382.5 25.58 0 1 -185.9 B + E 6 383.8 26.94 0 1 -185.4 B + E + T + B:T + E:T 18 384.4 27.52 0 1 -169.8 intercept 3 385.5 28.6 0 1 -189.6 B + E + B:E 8 386.9 30.02 0 1 -184.6 B 4 387.0 30.15 0 1 -189.3 B + E + T + B:E + B:T + E:T 20 389.1 32.22 0 1 -168.9 Full model (BxExT) 26 402.4 45.57 0 1 -165.1

Delicate skink B + E + T 13 353.7 0 0.46 0.46 -162.8 B + T 11 354.1 0.34 0.39 0.85 -165.3 B + E + T + B:E 15 356.0 2.24 0.15 1 -161.6 T 10 364.5 10.75 0 1 -171.6 E + T 12 366.7 12.96 0 1 -170.5 B + E 6 376.5 22.82 0 1 -182.0 B 4 377.2 23.5 0 1 -184.5 111

Data Model terms K AICc Δ AICc AICcWt Cum.Wt LL B + E + B:E 8 378.4 24.71 0 1 -180.8 B + E + T + E:T 27 380.7 26.97 0 1 -158.7 B + E + T + B:T + H:T 34 382.7 28.96 0 1 -149.8 B + E + T + B:E: + E:T 29 383.8 30.05 0 1 -157.5 intercept 3 387.8 34.07 0 1 -190.8 E 5 389.7 35.95 0 1 -189.7 E + T + E:T 26 393.2 39.52 0 1 -166.4

Eastern small-eyed snake B 4 129.2 0 0.67 0.67 -60.4 B + E 6 132.8 3.59 0.11 0.78 -59.9 intercept 3 133.5 4.32 0.08 0.86 -63.6 B + T 7 133.7 4.55 0.07 0.93 -59.2 B + T + B:T 10 136.3 7.19 0.02 0.95 -56.9 B + E + B:E 8 136.4 7.28 0.02 0.97 -59.4 E 5 137.3 8.12 0.01 0.98 -63.3 B + E + T 9 137.6 8.46 0.01 0.99 -58.8 T 6 137.9 8.72 0.01 1 -62.5 B + E + T + B:T 12 140.6 11.44 0 1 -56.4 B + E + T + B:E 11 141.7 12.49 0 1 -58.3 E + T 8 142.0 12.83 0 1 -62.2 B + E + T + E:T 15 146.5 17.29 0 1 -55.2 E + T + E:T 14 150.4 21.29 0 1 -58.6 B + E + T + B:T + H:T 18 150.6 21.4 0 1 -52.8 B + E + T + B:E: + E:T 17 151.3 22.12 0 1 -54.7

Beetles and spiders T 5 386.2 0 0.61 0.61 -187.4 B + T 6 388.1 1.85 0.24 0.85 -187.0 B + T + B:T 7 390.8 4.53 0.06 0.92 -187.0 E + T 7 391.5 5.29 0.04 0.96 -187.4 B + E + T 8 393.6 7.4 0.02 0.98 -187.0 intercept 4 395.2 8.97 0.01 0.98 -193.1 E + T + E:T 9 396.0 9.72 0 0.99 -186.6 B + E + T + B:T 9 396.6 10.38 0 0.99 -186.9 B 5 397.0 10.8 0 0.99 -192.8 B + E + T + B:E 10 397.4 11.21 0 1 -185.7 B + E + T + E:T 10 398.4 12.15 0 1 -186.2 E 6 400.2 14.01 0 1 -193.1 B + E + T + B:E + B:T 11 400.8 14.52 0 1 -185.7 B + E + T + B:T + H:T 11 401.7 15.45 0 1 -186.2 B + E 7 402.3 16.08 0 1 -192.8 B + E + T + B:E: + E:T 12 402.9 16.7 0 1 -185.0 B + E + B:E 9 406.1 19.83 0 1 -191.7 B + E + T + B:E: + B:T + E:T 13 406.6 20.39 0 1 -185.0 Full model (BxExT) 15 414.5 28.26 0 1 -184.7

112

Table 4. Parameter estimates (best), standard errors (SE) and marginal probabilities (P) of fixed-effect terms in the generalised linear mixed models within two AICc of the top-ranked model for animal occurrence variables.

Model Model fixed effects Factor levels best SE P a. Macropod scats Top ranked model: B + Ea + T Burn burnt 0.481 0.192 0.012 Exclosure partial (open months) -0.858 0.209 <0.001 Time Nov '12 0.558 0.294 0.058 Feb '13 -0.544 0.246 0.027 Apr '13 -0.016 0.298 0.956 Jun '13 0.443 0.238 0.062 Aug '13 -0.084 0.298 0.777 Oct '13 -0.153 0.242 0.527 Dec '13 -0.819 0.307 0.008 Feb '14 -0.675 0.245 0.006 Apr '14 -0.840 0.308 0.006 b. Rabbit diggings Top ranked model: T Time Aug '13 -0.372 0.195 0.057 Oct '13 -0.340 0.194 0.080 Dec '13 0.316 0.185 0.087 Feb '14 0.717 0.181 <0.001 Apr '14 -0.420 0.196 0.032

2nd ranked model: E + T Exclosure partial 0.5116 0.314 0.104 full exclosure 0.5716 0.315 0.069 Time Aug '13 -0.3712 0.195 0.057 Oct '13 -0.3393 0.194 0.080 Dec '13 0.3155 0.185 0.088 Feb '14 0.7175 0.181 <0.001 Apr '14 -0.4188 0.196 0.033

3rd ranked model: B + T Burn burnt 0.230 0.278 0.408 Time Aug '13 -0.372 0.195 0.056 Oct '13 -0.341 0.194 0.079 Dec '13 0.316 0.185 0.087 Feb '14 0.718 0.181 <0.001 Apr '14 -0.420 0.196 0.032 c. Brown antechinus Top ranked model: E + season Exclosure partial -0.116 0.182 0.524 full exclosure 0.352 0.163 0.031 Season spring -0.750 0.147 <0.001

2nd ranked model: B + E + season Burn burnt -0.132 0.137 0.336 Exclosure partial -0.116 0.182 0.524 full exclosure 0.352 0.163 0.031 Season spring -0.750 0.147 <0.001

113

Model Model fixed effects Factor levels best SE P d. Delicate skink Top ranked model: B + E + T Burn burnt 1.085 0.228 <0.001 Exclosure partial 0.002 0.246 0.995 full exclosure -0.553 0.265 0.037 Time Feb '13 -1.068 0.348 0.002 Apr '13 -2.080 0.528 <0.001 Jun '13 -1.163 0.360 0.001 Aug '13 -1.386 0.393 <0.001 Oct '13 -1.520 0.415 <0.001 Dec '13 -0.470 0.284 0.097 Apr '14 -1.068 0.348 0.002

2nd ranked model: B + T Burn burnt 1.077 0.260 <0.001 Time Feb '13 -1.068 0.345 0.002 Apr '13 -2.080 0.524 <0.001 Jun '13 -1.163 0.358 0.001 Aug '13 -1.386 0.390 <0.001 Oct '13 -1.520 0.412 <0.001 Dec '13 -0.470 0.281 0.095 Apr '14 -1.068 0.345 0.002 e. Eastern Small-eyed snake Top ranked model: B Burn burnt 1.551 0.600 0.010 f. Beetles and Spiders Top ranked model: T Time Nov'13 -0.505 0.131 0.000

2nd ranked model: B + T Burn burnt 0.135 0.152 0.376 Time Nov'13 -0.505 0.131 0.000 a Compares only partial-open months with open treatment (see methods)

114

Paper III: Synergistic interactions between fire and browsing drive plant diversity in a forest understory

In Paper II, I found that fire and herbivory interacted strongly to affect vegetation, with herbivory limiting the recovery of understorey vegetation from fire. In this paper, I focussed more closely on the mechanisms driving the interactive effects of fire and herbivory on plant diversity, and investigated whether these effects occurred via numerically mediated (interaction chain) or functionally moderated (interaction modification) pathways.

Foster, C.N, Barton, P.S, Sato, C.F, Macgregor, C.I, and D.B. Lindenmayer (2015) Synergistic interactions between fire and browsing drive plant diversity in a forest understory. Journal of Vegetation Science. 26, 1112-1123.

115

Abstract

Questions: Does browsing by large herbivores affect forest understory diversity by modifying assemblage dominance? Does fire interact with browsing to affect forest understory diversity?

Does this interaction occur via a numerically-mediated or functionally-moderated pathway?

Location: Booderee National Park, Jervis Bay Territory, south-eastern Australia

Methods: We tested the interactive effects of fire and browsing by native herbivores on understory plant diversity using a randomised blocked experiment in an open eucalypt forest.

We monitored the percentage cover of every vascular plant species in 24 sites over four experimental blocks. We applied a different treatment to each of the six sites in the four blocks.

Treatments were a factorial combination of three levels of herbivory treatment (open, partial exclosure, full exclosure) and two levels of prescribed fire treatment (burnt, unburnt).

Results: Browsing increased plant community dominance and reduced evenness and diversity, but only in burnt sites. Heavy browsing following fire created an understory dominated by an unpalatable, fire resistant fern species (bracken, Pteridium esculentum). This fire-browsing interaction was driven by both numerically-mediated and functionally-moderated pathways:

Fire both increased local browsing intensity, and amplified the per-unit effect of herbivores on the plant community.

Conclusions: The altered competitive environment after fire, combined with heavy post-fire browsing created a depauperate understory, dominated by bracken fern. The ability of bracken to supress the establishment of other plants means that, once established, this fern-dominated understory may be difficult to reverse. Our results demonstrate the key role of fire-browsing interactions in forest vegetation dynamics and highlight the importance of integrating large herbivore management with fire planning in forest ecosystems.

Keywords

Alternate stable states; Browsing; Community structure; Dominance; Disturbance interaction;

Eucalyptus forest; Herbivory; Macropod; Plant diversity; Pteridium esculentum; Understory 116

Nomenclature

Robinson (1991)

Introduction

Herbivory is a core process driving the structure and diversity of plant communities in many ecosystems worldwide (Milchunas et al. 1988; Hester et al. 2006; Borer et al. 2014). In forested systems, selective browsing by large herbivores can have transformative effects, altering canopy tree species dominance and driving cascading effects through the rest of the ecosystem (Royo &

Carson 2006; Holm et al. 2013; Côté et al. 2014). Browsing has also been shown to exert strong effects on forest understory vegetation. However, there is little agreement among studies about whether browsing enhances or reduces the diversity of forest understory plants. Positive effects

(Royo et al. 2010a), negative effects (Rooney & Waller 2003; Jenkins et al. 2014), and no effect

(Kerns et al. 2011) of browsing on understory plant diversity have all been reported, with discrepancies between studies attributed to differences in herbivore densities, ecosystem productivity, disturbance history and other site-dependent variables (Hester et al. 2006; Royo et al. 2010a).

While the net effects of herbivory on plant diversity are variable, it has been suggested that most responses can be explained by effects on plant species dominance (Hester et al. 2006).

In systems where herbivores selectively feed on dominant plant species, they will reduce assemblage dominance, and increase resource availability for competitively inferior plants, thereby increasing diversity (Côté et al. 2004; Hester et al. 2006). However, when the dominant species are unpalatable, herbivores will have the opposite effect. A number of recent studies from grassland systems have supported this hypothesis, finding that the effects of herbivory on ground-level light availability (which is driven by the abundance of dominant species), consistently explained herbivory effects on plant diversity, despite the sites having a wide range of abiotic conditions and disturbance histories (Borer et al. 2014; Eby et al. 2014; Koerner et al.

2014).

117

Understanding the effects of grazing and browsing can be useful for predicting the outcomes of management interventions (Hester et al. 2000). Such predictions, however, are complicated by the fact that herbivory can interact strongly with other episodic disturbance events (e.g. fire, timber harvest, drought; Royo & Carson 2006; Wisdom et al. 2006). These interactions can produce vastly different outcomes than would be predicted from studies of individual effects, and can lead to unintended management outcomes (Tylianakis et al. 2008).

Ecological disturbances can interact via two main pathways: an interaction chain (a numerically-mediated process), and an interaction modification (a functionally-moderated process) (Didham et al. 2007). In the context of disturbance-grazing interactions, a numerically- mediated interaction could occur through disturbance changing the local abundance of herbivores, in turn affecting vegetation. A functionally-moderated interaction would occur if disturbance modified the per unit effect of herbivores on vegetation.

Both these interaction pathways are commonly described in studies of fire-grazing interactions, although they are not often explicitly identified as such. For example, patch burning of grasslands can initiate a numerically-mediated fire-grazing interaction; fire stimulates fresh growth which attracts large herbivores (Allred et al. 2011). The resulting concentration of herbivores increases local grazing intensity, while reducing it elsewhere, which increases vegetation heterogeneity (Fuhlendorf et al. 2006). By modifying the competitive interactions among plant species, or selectively acting on some plant traits, fire can also have a functionally-moderated effect on plant responses to herbivory. For example, in a North

American mesic grassland, annual burning amplified the negative effect of grazing exclusion on plant diversity (Eby et al. 2014). This was due to the dominant grass species in the system being both a post-fire increaser and highly palatable. These two functional traits gave this grass species a competitive advantage following fire and allowed it to dominate the plant community in the absence of large herbivores (Eby et al. 2014).

As disturbances can interact with the life history traits of organisms, as well as with each other, predicting the outcomes of disturbance interactions is problematic, even if plant traits are well understood (Royo & Carson 2006). Therefore, understanding such interactions 118

requires manipulative experiments that are able to test the non-additive effects of these processes (Didham et al. 2007; Tylianakis et al. 2008). Manipulative tests of the interactive effects of herbivory and disturbance have become reasonably common in grassland systems

(e.g. Fuhlendorf et al. 2006; Collins & Calabrese 2012; Koerner & Collins 2014), but remain rare in forested systems (Wisdom et al. 2006; Royo et al. 2010a). The high densities of large herbivores in many forests worldwide (Côté et al. 2004), and the prevalence of active fire management in forest ecosystems (Stephens & Ruth 2005; Boer et al. 2009), mean that improving our understanding of the effects of fire-browsing interactions on vegetation dynamics is of high importance for the management of forest ecosystems (Wisdom et al. 2006).

We studied the effects of fire × browsing interactions on plant diversity in a temperate forest understory, testing the hypothesis that effects on plant diversity would be similar to those described for grassland ecosystems. Our study had three main predictions: (1) browsing by large herbivores would affect plant diversity by modifying assemblage dominance. Specifically, as the dominant understory species in our study system was of low palatability (Pteridium esculentum, Di Stefano & Newell 2008; Fletcher et al. 2011), we predicted that browsing would increase assemblage dominance and reduce diversity. We also predicted that: (2) fire would amplify the effects of browsing on plant diversity, and that (3) this interaction would occur via both numerically-mediated and functionally-moderated pathways. The numerically-mediated effect would be driven by more herbivores browsing in burnt sites, while the functionally- moderated effect would be driven by fire amplifying the effects of herbivory on P. esculentum dominance.

Methods

Study site

We conducted this study in Booderee National Park (BNP); a ~6 500 ha peninsula in south- eastern Australia (35°10′S, 150°40′E, Fig. 1). The most widespread vegetation type in the park is open eucalypt forest (Taws 1998). This forest type is dominated by Eucalyptus pilularis,

119

Corymbia gummifera and E. botryoides in the overstory (> 10 m), Banksia serrata, Acacia longifolia and Monotoca elliptica in the midstory (2 – 10 m) and Lomandra longifolia and P. esculentum in the understory (< 2 m) (Taws 1998; Lindenmayer et al. 2008).

An intensive baiting program targeting the introduced red fox (Vulpes vulpes) has been in place within the park since 1999 to protect native small mammal species from predation

(Dexter et al. 2012). As hunting has been long discontinued and dingoes (Canis lupus dingo) now occur only rarely in the park, the removal of foxes means that predation pressure on herbivores is low (Lindenmayer et al. 2014). Since the removal of foxes, there has been a tenfold increase in the numbers of native macropod herbivores within the park (Dexter et al.

2012; Lindenmayer et al. 2014). The most common of these macropods are the swamp wallaby,

Wallabia bicolor, a generalist browser, and the eastern grey kangaroo, Macropus giganteus, a grazer (Davis et al. 2008; Dexter et al. 2013). Small-scale exclosure trials have indicated that the current high abundance of these herbivores could be driving a decline in some plant species

(Dexter et al. 2013).

Study design

We quantified the interactive effects of fire and large herbivores on understory plant diversity using a randomised, blocked experiment. We examined three levels of herbivory treatment and two levels of burning treatment in a factorial design (Fig. 1). We replicated each of these six treatment combinations across four experimental blocks to give a total of 24 sites. For the herbivory treatments, we manipulated the density of macropods within 25 × 25 m plots using exclosure fences to produce three levels of herbivory: full herbivory (open treatment), intermediate herbivory (partial treatment – plots were fenced but gates opened and closed at two month intervals to simulate lower browsing pressure), and no herbivory (exclosure treatment).

We constructed exclosure fences in June 2012 using 1.1 m tall feral-proof stock fencing, which we found to be effective at excluding macropods (see results). For the burning treatments, we conducted 50 × 50 m, low severity burns across half of the herbivore treatment sites in August

120

2012. Burns removed approximately 95% of understory vegetation from the site and did not reach canopy foliage (tree scorch height 1.5 - 4 m).

Figure 1. Study location and design showing (a) location of Booderee National Park in south- eastern Australia, (b) location of the four experimental blocks (A-D) within Booderee National Park, and (c) a schematic diagram of the factorial design showing the experimental treatments and their arrangement within the four experimental blocks (not to scale, exact spatial arrangement of sites varies between blocks). Dark grey shaded area in (b) shows the extent of the Eucalyptus pilularis forest type within BNP, black outline indicates park boundary, light grey shading depicts ocean.

Data collection

We surveyed understorey vegetation prior to treatment implementation in April/May 2012

(austral autumn) and repeated surveys three months after burning, and then every six months until May 2014. To capture variation within each site, we sampled vegetation using four small quadrats, rather than one large quadrat in each site. We placed four 3 x 3 m quadrats at fixed locations in each site, with one quadrat in each of the four quarters of the site, ensuring each was at least 1.5 m from the fence. For each quadrat, we visually estimated the projective cover of each vascular plant species, using the same observer for all estimates. Prior to any data processing, we averaged the cover values of each plant species across the four quadrats to give

121

one value for the 36 m2 area sampled in each site. We used these site-level data to calculate four standard metrics of community diversity: species richness (species/ 36 m2), species diversity

(Simpson’s reciprocal index - 1/D), evenness (Simpson’s evenness – E1/D), and assemblage dominance (Berger-Parker index – d) (Magurran 2004).

We surveyed site use by macropod herbivores by counting their scats along two 25 × 2 m transects (100 m2) in each site, summing counts to give one overall count per site. We surveyed transects every two months from August 2012 to May 2014. We removed scats from transects after each survey to avoid double counting. For analysis, we summed the two scat counts preceding each vegetation survey to give an estimate of the average level of herbivore activity in a site leading up to each sampling event.

Data analysis

We tested the effects of herbivory, fire and their interaction on plant species richness, diversity, evenness and dominance using linear mixed-models (LMMs) in R (R Core Team 2013). The full model for these analyses included the fixed effects of herbivory × fire × time, plus the pre- treatment value (to account for any pre-existing differences between sites), and random effects of block/site. Instead of the categorical herbivory treatments, we used the scat counts at each site as our measure of herbivory. This allowed us to account for variation in herbivore pressure both between and within levels of the herbivore exclosure treatments, and also to test whether the relationship between herbivore pressure and plant diversity metrics differed between burnt and unburnt sites (i.e. whether there was a functionally moderated interaction). Prior to inclusion in the model as a fixed effect, scat count data were natural log (ln) transformed to reduce skew, and standardised by centring, and dividing by two standard deviations to allow comparison of the main effects of fire and herbivory and aid in interpretation of interaction terms (Gelman & Hill 2007). After checking the fit of the full models by inspecting residual plots (Zuur et al. 2009), we used the “dredge” function in the package “MuMIn” (Barton 2014) to rank all possible subsets of the full model, based on minimising Akaike’s Information

Criterion, corrected for small sample sized (AICc) (Burnham & Anderson 2002).

122

As our study included repeated measures at sites, we tested for temporal autocorrelation between repeated measures at sites using the “acf” function in R (R Core Team 2013).

Autocorrelation for all variables was low, and including a first-order autoregressive correlation structure on the random effects did not improve the fit of the LMMs. Therefore, we did not include correlation structures in our final models (Pinheiro & Bates 2000). We also considered fitting the species richness data with a generalised linear mixed-model (GLMM), with Poisson errors, which is often used for count data. However, this was not used in the final analysis as the characteristics of a Poisson distribution (skewed distribution and heteroscedasticity) were not evident in our data (Zuur et al. 2009), and a trial GLMM showed poor fit at the extremities when compared with the LMM.

To assess whether fire also interacted with herbivory via a numerically-mediated pathway, we tested whether fire increased herbivore activity (as measured from scat counts) within the different exclosure treatments. We used GLMM with a Poisson error distribution and a log-link function for this analysis, with exclosure treatment × fire × time as fixed effects and block/site as random effects. As very low scat counts in exclosure treatments ( = 1.78 ± 0.53,

59% of counts = 0) led to complete separation of factor combinations (Gelman & Hill 2007), the exclosure treatment was excluded from this analysis. Scat count data were found to be over- dispersed, which we accounted for by including an observation-level random effect in our model (Harrison 2014).

We tested for differences in community composition among the six treatment combinations both pre- (May 2012) and post-treatment (May 2014), using nonparametric blocked multi-response permutation procedures (MRBP, McCune & Grace 2002) in PC-ORD

(McCune & Mefford 2006). We excluded rare species (those occurring at 2 or fewer sites) from the site × species matrix and relativized abundances within sites, prior to analysis (McCune &

Grace 2002). If MRBP indicated significant differences in species composition among treatments, we then used indicator species analysis (Dufrene & Legendre 1997) to identify individual species associated with the different treatments. We used the function “multipatt” in the R package “indicspecies” (De Cáceres & Legendre 2009), which allows the identification of 123

species associations with combinations of groups (treatments) rather than just a single group

(De Cáceres et al. 2010). We set the maximum number of groups (treatment combinations) to be combined to three as this allowed us to identify associations with both individual treatment combinations, as well as with complete factors in the factorial design. We accounted for the blocked structure of sites by constraining permutations within, rather than among blocks.

Significance was determined from 9999 permutations of the data.

Results

Effect of browsing and fire on plant diversity metrics

We detected 111 understory plant species over five surveys in the two-year study period

(Appendix S1). In the top-ranked LMMs, fire modified the effect of herbivory on plant community diversity, evenness, and dominance and this interaction was consistent across time

(Table 1, Fig. 2, Appendix S2). Plant species diversity, evenness and dominance were not related to herbivore activity on unburnt sites, but burnt sites had lower community diversity and evenness and higher assemblage dominance as herbivore activity increased (Table 1, Fig. 2).

This indicates a functionally moderated interaction. The difference in community diversity and evenness between burnt and unburnt sites varied through time and was highest 15 months post- fire. In contrast, the highest-ranked model for assemblage dominance indicated that the effect of fire was consistent across time; at high levels of herbivory, the relative abundance of the dominant species was 30 % higher in burnt than unburnt sites (Fig. 2). This increase in assemblage dominance was driven by P. esculentum, which comprised approximately 70 % of the vegetation cover in burnt-open sites, but less than 40 % in burnt and unburnt exclosure sites

(Fig. 3). In contrast to the other diversity metrics, none of the models for species richness with

ΔAICc < 2 included the herbivory × fire interaction (Appendix S2), and the effects of fire were short-term, with species richness of burnt sites returning to the level of unburnt sites within 15 months (Table 1).

124

Table 1. Estimates of coefficients (Est.), their standard errors (SE) and probabilities for the top-ranked linear mixed model testing the effects of fire (burnt or unburnt), large herbivore activity (scats per 100m2, ln transformed), time since fire (3, 9, 15 or 21 months) on plant species richness (species / 36 m2), diversity (Simpson’s reciprocal index, 1/D), evenness (Simpson’s evenness,

E1/D) and dominance (Berger-Parker, d). Pre-treatment values were included as a fixed effect to account for pre-existing differences between sites. Reference states for comparisons were unburnt sites three months post-fire. Random terms for each model were block/site.

Species richness Diversity (1/D) Evenness (E1/D) Dominance (d) Fixed effects Est. SE df P Est. SE df P Est. SE df P Est. SE df P

Intercept 6.15 5.33 66 0.253 2.25 0.72 64 0.003 0.075 0.024 64 0.002 0.266 0.078 67 0.001 Pre-treatment value 0.85 0.19 18 <0.001 0.39 0.20 18 0.060 0.373 0.168 18 0.040 0.296 0.126 18 0.030 Fire -9.40 1.45 18 <0.001 1.04 0.53 18 0.066 0.095 0.020 18 <0.001 0.107 0.031 18 0.003 Herbivores (ln trans.) 0.28 0.48 64 0.558 0.012 0.016 64 0.455 -0.022 0.046 67 0.636 Time since fire 9 mo 0.50 0.78 66 0.523 -0.43 0.48 64 0.378 -0.016 0.020 64 0.410 0.115 0.038 67 0.003 15 mo -0.25 0.78 66 0.749 -0.02 0.48 64 0.973 0.001 0.020 64 0.975 0.063 0.038 67 0.100

21 mo 1.17 0.78 66 0.139 0.13 0.49 64 0.785 0.001 0.020 64 0.973 0.053 0.039 67 0.177

Fire:Herbivores(ln) -1.68 0.64 64 0.011 -0.059 0.021 64 0.006 0.209 0.061 67 0.001 Fire:Time B:9mo 6.67 1.10 66 <0.001 -1.69 0.67 64 0.014 -0.110 0.028 64 <0.001 B:15mo 10.33 1.10 66 <0.001 -1.88 0.67 64 0.007 -0.126 0.028 64 <0.001

B:21mo 10.92 1.10 66 <0.001 -1.67 0.69 64 0.018 -0.122 0.028 64 <0.001

125

Figure 2. Response of plant community diversity (Simpson’s reciprocal index, 1/D, a-d),

evenness (Simpson’s evenness, E1/D, e-h) and dominance (Berger-Parker, d, i-l) to fire (burnt or unburnt) and herbivore activity through time. Values are predicted means and 95 % confidence intervals from the top-ranked model for each metric. Pre-treatment values were fixed at the mean for all predictions (3.15 for diversity, 0.106 for evenness and 0.53 for dominance).

126

Figure 3. Species rank-abundance curves for burnt and unburnt sites under each of the herbivory treatments in the final survey period (May 2014). Species with relative abundances > 0.1 are identified on each plot.

Effect of fire on herbivore activity

Fire affected herbivore activity levels, with both the first and second ranked models (ΔAICc =

0.46) showing that scat counts were higher on burnt than unburnt sites (Table 2), indicating a numerically-mediated interaction between fire and herbivory. In the first-ranked model, this effect was consistent across time, but in the second-ranked model, this difference was significant only in the period from three to nine months post-fire, where counts in burnt sites were more than double those in unburnt sites (Table 2, Fig. 4).

127

Table 2. Results of generalised linear mixed models testing whether fire affected the level of herbivore activity (scats per 100m2) in the different exclosure treatments (open or partial exclosure) through time (3, 9, 15 or 21 months post-fire). Full exclosure treatments were not included in this analysis due to very low values ( = 1.78 ± 0.53). Reference states for comparisons were unburnt, open sites three months post-fire.

Top ranked model Second ranked model Fixed effects Est. SE Z P Est. SE Z P Intercept 4.26 0.23 18.8 <0.001 4.36 0.25 17.2 <0.001 Fire 0.46 0.20 2.2 0.025 0.27 0.33 0.8 0.408 Exclosure (partial) -1.63 0.20 -8.0 <0.001 -1.63 0.20 -8.1 <0.001 Time since fire 9 months -0.51 0.24 -2.1 0.038 -1.04 0.32 -3.2 0.001 15 months -0.25 0.24 -1.0 0.306 -0.20 0.31 -0.7 0.515 21 months -0.89 0.24 -3.6 <0.001 -0.81 0.31 -2.6 0.010 Fire:Time B:9m 1.01 0.44 2.3 0.022 B:15m -0.09 0.43 -0.2 0.831 B:21m -0.18 0.44 -0.4 0.690

Figure 4. Herbivore activity in each of the experimental treatment combinations over time. Values are the predicted mean scat count and 95 % confidence intervals (a-d) based on the second-ranked generalised linear model. Full exclosure treatments were not included in this analysis due to very low values ( = 1.78 ± 0.53).

Effect of browsing and fire on plant species composition

Species composition did not differ among experimental treatments prior to treatment implementation in May 2012 (MRBP, P = 0.15). However, by May 2014, species composition differed significantly among treatments (MRBP, P = 0.004). Pairwise comparisons revealed that burnt-open sites had significantly different species composition to all other treatment

128

combinations (all P < 0.05). Burnt partial and burnt exclosure sites also differed significantly in composition (P = 0.03).

Indicator species analysis identified six individual species that were associated with the experimental treatments, each of which was positively associated with three of the treatment combinations (all P < 0.05, Appendix S1). The fern P. esculentum was positively associated with both burnt and unburnt open sites, as well as burnt partial sites. Conversely, the grass

Themeda australis and the herb Galium propinquum were associated with both burnt and unburnt exclosure treatments, as well as unburnt partial treatments. A further three species were associated with burnt exclosure treatments but differed in their other associations; the small tree

Synoum glandulosum was also associated with burnt partial and unburnt exclosure sites, the sub-shrub Marsdenia suavolens was also associated with unburnt exclosure sites and burnt open sites, and another small tree Persoonia linearis was associated with all burnt treatments.

Discussion

We tested plant diversity responses to browsing and the fire-browsing interaction using a manipulative experiment in a temperate forest understory. As predicted, where browsing increased plant community dominance, it reduced plant diversity and evenness. Prescribed fire interacted strongly with browsing, with herbivore activity affecting these measures of plant diversity only in burnt sites. This fire-browsing interaction was generated through a combination of numerically-mediated and functionally-moderated pathways; fire both increased herbivore activity levels, and increased the per-unit effect of herbivores on the plant community.

Effect of browsing and fire

Browsing had little effect on the plant community in the absence of fire, but browsing in burnt environments increased community dominance, reduced plant community evenness and diversity, and altered species composition. Specifically, browsing in burnt environments increased the dominance of the fern P. esculentum and reduced the abundance of a number of

129

palatable grasses and herbs. This pattern was consistent with previous studies where plant diversity responses to grazing depended on the effect of grazing on the dominant plant species

(Mathisen et al. 2010; Royo et al. 2010a; Borer et al. 2014). As suggested by Royo et al.

(2010a), the variable effect of browsing on the dominant plant species can potentially explain the wide range of reported effects of large herbivores on forest plant diversity. In systems where large herbivores have been at high abundances for extended periods, understories are largely dominated by a few unpalatable species; this is the case in many forested systems of North-

America (Tremblay et al. 2006; Rooney 2009) and Europe (Kirby 2001) where deer are highly abundant. In such systems, continued browsing maintains the dominance of unpalatable species, while herbivore exclusion can allow palatable species to recover, thereby increasing plant species richness and evenness. In systems where large herbivores are not highly abundant, browsing may help to maintain understory diversity by reducing the dominance of fast-growing palatable species, particularly following disturbance (Royo et al. 2010a).

Despite having strong effects on plant community diversity and evenness after fire, we found that browsing did not affect plant species richness, either with or without burning.

Mathisen et al. (2010) reported a similar result for moose browsing in Swedish pine forest. This may be because much of the vegetation in both the study by Mathisen et al. (2010) and in our study was comprised of shrubs (Appendix S1). Browsing is often not lethal for established individuals of woody plant species (Hester et al. 2006). Thus, in forest systems with woody understories, browsing will affect the relative cover of different species more strongly than the relative number of individuals. In addition, as shrubs are slow-growing and long-lived compared with many herbs and grasses (McFarland 1998), the two-year time frame of our study may have been insufficient for differences in colonisation and extinction rates to become apparent.

Interaction pathways

The stronger effect of herbivory on vegetation after burning occurred both via numerically- mediated and functionally-moderated pathways. Burning both increased the herbivore activity

130

at a site (Fig. 4), and increased the per unit effect of herbivore activity on the vegetation (Fig.

2). The attraction of large herbivores to recently burnt patches has been well documented in grassland ecosystems (Allred et al. 2011), and has also been shown for some forest-dwelling ungulates (Fisher & Wilkinson 2005). However, previous studies of macropod herbivores following fire have found little effect of large wildfires on macropod densities (Garvey et al.

2010; Arthur et al. 2012). These contrasting results indicate that fire-browsing interactions driven by the attraction of herbivores to burnt areas may be more likely to occur following small, patchy fires (as occurs with prescribed burns), than following large-scale wildfires. The effects of prescribed fire on browsing intensity should therefore be considered in fire planning in these systems, especially where herbivores are particularly abundant.

The greater per-unit effect of herbivores on vegetation in burnt than in unburnt sites was likely due to the competitive release of P. esculentum following fire. After fire (and other disturbances), Pteridium species can rapidly regenerate from below-ground rhizomes and dominate recently disturbed environments (Skre et al. 1998; Spencer & Baxter 2006). However, in the absence of further disturbance, the dominance of P. esculentum usually declines with time since fire, as other plants become more established (Spencer & Baxter 2006). In our study, browsing maintained assemblage dominance two years after fire, with P. esculentum comprising almost 70 % of understory vegetation cover in burnt, browsed (open) sites, compared with less than 40 % in burnt, un-browsed sites (Fig. 3). The dominance of P. esculentum could have important consequences for the ability of the vegetation to recover to its pre-fire state, as P. esculentum can form a shade canopy which supresses the establishment of other plants (Tolhurst & Turvey 1992). Therefore, our results suggest that heavy browsing after prescribed fire may lead to an understory dominated by a few fire- and browsing-resistant species, which could be difficult to reverse.

An understory dominated by browse- and fire-resistant plants was suggested by Wisdom et al. (2006) as a likely outcome of low severity forest fires under moderate or high herbivory.

Previous studies of the individual effects of chronic herbivory (Rooney 2001) and reoccurring fire (Spencer & Baxter 2006) have documented transitions towards depauperate understories 131

dominated by ferns (Dennestaedtia and Pteridium respectively). Our study, and an earlier pilot study (Dexter et al. 2013), show that, in concert, fire and herbivory can have synergistic effects, creating a depauperate, fern-dominated understory over a relatively short time period. Our results also show that at low levels of herbivory, differences in vegetation diversity between burnt and unburnt sites are small. Therefore, in forests where predation by foxes and dingoes keeps macropod herbivores at low densities, fire-browsing interactions may have limited effects on vegetation diversity.

Although the timeframe of our study means we cannot predict whether the effects of these fire-browsing interactions will persist in the long-term, alternate stable states driven by chronic herbivory have been documented from a range of forest ecosystems worldwide (Royo

& Carson 2006; Raffaele et al. 2011; Tanentzap et al. 2011; Hidding et al. 2013). These systems are characterised by understories with low floristic diversity which are dominated by a few, unpalatable species (de la Cretaz & Kelty 1999; Horsley et al. 2003). Once such states have developed, legacy effects can prevent the recovery of plant diversity, even if herbivore densities are reduced (Royo et al. 2010g). Preventing the development of such states through the integrated management of disturbances such as herbivory and fire is therefore of key importance for the conservation of diverse forest understory flora (Royo & Carson 2006;

Wisdom et al. 2006).

Can effects on dominance always be used to predict the outcome of fire- browsing interactions?

We were able to predict the outcome of the fire-browsing interaction in our study. However, this was most likely due to the particularities of our study system rather than a predictability of fire-browsing interactions in general. In our study system, the most dominant and widespread understory species is P. esculentum, which is of low palatability (Di Stefano & Newell 2008) and responds positively to fire (Spencer & Baxter 2006). Therefore, both fire and browsing acted to increase the dominance of P. esculentum, and their functionally-moderated effects were synergistic. In addition, fire increased browsing activity, meaning that the numerically-mediated interaction between fire and browsing also had positive synergistic effects. 132

By contrast, in many other systems, the outcomes of fire-browsing interactions will be far less predictable. For example, in a system where fire and browsing act antagonistically (e.g. fire increases dominance, but browsing has the opposite effect, as in Eby et al. 2014), where more than one species is dominant (and fire or browsing responses differ between these species) and/or where functionally-moderated effects are in opposition to numerically-mediated effects, the outcome will depend on the balance of these opposing effects. Further, fire and browsing may interact in more subtle ways, such as modifying the strength of each other’s effects on dominant species. For example, the tissues of plants regenerating after fire often lack the physical and chemical defences of mature plants, meaning generally unpalatable plants can be highly palatable after burning (Augustine & McNaughton 1998). If differences in palatability between species are reduced, herbivores feed less selectively, so the strength of browsing effects on assemblage dominance can be reduced or even eliminated (Augustine & McNaughton

1998). In addition, low severity fires, such as prescribed burns can have different effects on the plant community than high severity fires that often occur as wildfire (Morrison 2002). Different fire severities may therefore have differing effects on dominant species, and so the outcomes of fire-browsing interactions are likely to vary with fire severity. Therefore, predicting the outcome of fire-browsing interactions will only be possible in some ecosystems, and where there is detailed knowledge of the palatability and functional traits of plant species. Outside of these systems, manipulative experiments will continue to be critical to improving understanding of the dynamics of fire-browsing interactions.

Conclusion

In forested systems that are dominated by unpalatable, fire-resistant plant species, prescribed fire and browsing are likely to interact synergistically to increase assemblage dominance and reduce plant community diversity. In such systems, integrated management of fire and large herbivores will be necessary to prevent the development of an understory dominated by a narrow set of fire tolerant, unpalatable species. In forested ecosystems where unpalatable, fire resistant species are not dominant, the outcome of fire-browsing interactions will not be so easy to predict. Therefore manipulative studies which can examine both the individual and 133

interactive effects of these disturbances, and their interaction pathways, will continue to play a central role in developing our understanding of fire-browsing interactions in forest ecosystems.

Acknowledgements

Staff of BNP conducted prescribed burns. We thank the Department of Environment and the landowners and co-managers of BNP - the Wreck Bay Aboriginal Community for supporting this project. The Margaret Middleton Fund, The Norman Wettenhall Foundation and the Long

Term Ecological Research Network provided financial support.

Literature Cited

Allred, B.W., Fuhlendorf, S.D., Engle, D.M. & Elmore, R.D. 2011. Ungulate preference for burned patches reveals strength of fire-grazing interaction. Ecology and Evolution 1: 132- 144. Arthur, A.D., Catling, P. & Reid, A. 2012. Relative influence of habitat structure, species interactions and rainfall on the post-fire population dynamics of ground-dwelling vertebrates. Austral Ecology 37: 958-970. Augustine, D.J. & McNaughton, S.J. 1998. Ungulate effects on the functional species composition of plant communities: Herbivore selectivity and plant tolerance. Journal of Wildlife Management 62: 1165-1183. Barton, K. 2014. MuMIn: Multi-model inference.R package version 1.10.5., http://CRAN.R- project.org/package=MuMIn. Boer, M.M., Sadler, R.J., Wittkuhn, R.S., McCaw, L. & Grierson, P.F. 2009. Long-term impacts of prescribed burning on regional extent and incidence of wildfires—Evidence from 50 years of active fire management in SW Australian forests. Forest Ecology and Management 259: 132-142. Borer, E.T., Seabloom, E.W., Gruner, D.S., Harpole, W.S., Hillebrand, H., Lind, E.M., Adler, P.B., Alberti, J., Anderson, T.M.... 2014. Herbivores and nutrients control grassland plant diversity via light limitation. Nature 508: 517-520. Burnham, K.P. & Anderson, D.R. 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. Springer-Verlag, New York. Collins, S.L. & Calabrese, L.B. 2012. Effects of fire, grazing and topographic variation on vegetation structure in tallgrass prairie. Journal of Vegetation Science 23: 563-575.

134

Côté, S.D., Beguin, J., de Bellefeuille, S., Champagne, E., Thiffault, N. & Tremblay, J.P. 2014. Structuring effects of deer in boreal forest ecosystems. Advances in Ecology 2014: Article ID: 917834. Côté, S.D., Rooney, T.P., Tremblay, J.P., Dussault, C. & Waller, D.M. 2004. Ecological impacts of deer overabundance. Annual Review of Ecology, Evolution and Systematics 35: 113-147. Davis, N.E., Coulson, G. & Forsyth, D.M. 2008. Diets of native and introduced mammalian herbivores in shrub-encroached grassy woodland, south-eastern Australia. Wildlife Research 35: 684-694. De Cáceres, M. & Legendre, P. 2009. Associations between species and groups of sites: indices and statistical inference. Ecology 90: 3566-3574. De Cáceres, M., Legendre, P. & Moretti, M. 2010. Improving indicator species analysis by combining groups of sites. Oikos 119: 1674-1684. de la Cretaz, A. & Kelty, M. 1999. Establishment and control of hay-scented Fern: A native invasive species. Biological Invasions 1: 223-236. Dexter, N., Hudson, M., James, S., MacGregor, C. & Lindenmayer, D.B. 2013. Unintended consequences of invasive predator control in an Australian forest: Overabundant wallabies and vegetation change. PloS ONE 8: e69087. Dexter, N., Ramsey, D.L., MacGregor, C. & Lindenmayer, D. 2012. Predicting ecosystem wide impacts of wallaby management using a fuzzy cognitive map. Ecosystems 15: 1363-1379. Di Stefano, J. & Newell, G.R. 2008. Diet selection by the swamp wallaby (Wallabia bicolor): Feeding strategies under conditions of changed food availability. Journal of Mammalogy 89: 1540-1549. Didham, R.K., Tylianakis, J.M., Gemmell, N.J., Rand, T.A. & Ewers, R.M. 2007. Interactive effects of habitat modification and species invasion on native species decline. Trends in Ecology & Evolution 22: 489-496. Dufrene, M. & Legendre, P. 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecological Monographs 67: 345-366. Eby, S., Burkepile, D.E., Fynn, R.W.S., Burns, C.E., Govender, N., Hagenah, N., Koerner, S.E., Matchett, K.J., Thompson, D.I.... 2014. Loss of a large grazer impacts savanna grassland plant communities similarly in North America and South Africa. Oecologia 175: 293-303. Fisher, J.T. & Wilkinson, L. 2005. The response of mammals to forest fire and timber harvest in the North American boreal forest. Mammal Review 35: 51-81. Fletcher, M.T., Brock, I.J., Reichmann, K.G., McKenzie, R.A. & Blaney, B.J. 2011. Norsesquiterpene glycosides in bracken ferns (Pteridium esculentum and Pteridium aquilinum subsp. wightianum) from eastern Australia: Reassessed poisoning risk to animals. Journal of Agricultural and Food Chemistry 59: 5133-5138.

135

Fuhlendorf, S.D., Harrell, W.C., Engle, D.M., Hamilton, R.G., Davis, C.A. & Leslie Jr, D.M. 2006. Should heterogeneity be the basis for conservation? Grassland bird response to fire and grazing. Ecological Applications 16: 1706-1716. Garvey, N., Ben-Ami, D., Ramp, D. & Croft, D.B. 2010. Survival behaviour of swamp wallabies during prescribed burning and wildfire. Wildlife Research 37: 1-12. Gelman, A. & Hill, J. 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press, Cambridge, UK. Harrison, X.A. 2014. Using observation-level random effects to model overdispersion in count data in ecology and evolution. PeerJ 2: e616. Hester, A.J., Bergman, M., Iason, G.R. & Moen, J. 2006. Impacts of large herbivores on plant community structure and dynamics. In: Danell, K., Bergström, R., Duncan, P. & Pastor, J. (eds.) Large Herbivore Ecology, Ecosystem Dynamics and Conservation, pp. 97-141. Cambridge University Press, Cambridge. Hester, A.J., Edenius, L., Buttenschon, R.M. & Kuiters, A.T. 2000. Interactions between forests and herbivores: The role of controlled grazing experiments. Forestry 73: 381-391. Hidding, B., Tremblay, J.-P. & Côté, S.D. 2013. A large herbivore triggers alternative successional trajectories in the boreal forest. Ecology 94: 2852-2860. Holm, J.A., Thompson, J.R., McShea, W.J. & Bourg, N.A. 2013. Interactive effects of chronic deer browsing and canopy gap disturbance on forest successional dynamics. Ecosphere 4: art144. Horsley, S.B., Stout, S.L. & deCalesta, D.S. 2003. White-tailed deer impact on the vegetation dynamics of a northern hardwood forest. Ecological Applications 13: 98-118. Jenkins, L.H., Jenkins, M.A., Webster, C.R., Zollner, P.A. & Shields, J.M. 2014. Herbaceous layer response to 17 years of controlled deer hunting in forested natural areas. Biological Conservation 175: 119-128. Kerns, B.K., Buonopane, M., Thies, W.G. & Niwa, C. 2011. Reintroducing fire into a ponderosa pine forest with and without cattle grazing: understory vegetation response. Ecosphere 2: art59. Kirby, K.J. 2001. The impact of deer on the ground flora of British broadleaved woodland. Forestry 74: 219-229. Koerner, S.E., Burkepile, D.E., Fynn, R.W.S., Burns, C.E., Eby, S., Govender, N., Hagenah, N., Matchett, K.J., Thompson, D.I.... 2014. Plant community response to loss of large herbivores differs between North American and South African savanna grasslands. Ecology 95: 808-816. Koerner, S.E. & Collins, S.L. 2014. Interactive effects of grazing, drought, and fire on grassland plant communities in North America and South Africa. Ecology 95: 98-109. Lindenmayer, D., MacGregor, C., Dexter, N., Fortescue, M. & Beaton, E. 2014. Booderee National Park, The Jewel of Jervis Bay. CSIRO Publishing, Collingwood, Australia.

136

Lindenmayer, D.B., MacGregor, C., Welsh, A., Donnelly, C., Crane, M., Michael, D., Montague-Drake, R., Cunningham, R., Brown, D.... 2008. Contrasting mammal responses to vegetation type and fire. Wildlife Research 35: 395-408. Magurran, A.E. 2004. Measuring Biological Diversity. Blackwell Publishing, Oxford, UK. Mathisen, K.M., Buhtz, F., Danell, K., Bergström, R., Skarpe, C., Suominen, O. & Persson, I.L. 2010. Moose density and habitat productivity affects reproduction, growth and species composition in field layer vegetation. Journal of Vegetation Science 21: 705-716. McCune, B. & Grace, J.B. 2002. Analysis of Ecological Communities. MjM Software Design, Gleneden Beach, Oregon, USA. McCune, B. & Mefford, B.M.J. 2006. PC-ORD. Multivariate analysis of ecological data.Version 5.10. MjM Software, Gleneden Beach, Oregon, U.S.A. McFarland , D.C. 1998. Fire and the vegetation composition of subtropical heathlands in south- eastern Queensland. Australian Journal of Botany 36: 533-546. Milchunas, D.G., Sala, O.E. & Lauenroth, W.K. 1988. A generalized model of the effects of grazing by large herbivores on grassland community structure. American Naturalist 132: 87-106. Morrison, D.A. 2002. Effects of fire intensity on plant species composition of sandstone communities in the Sydney region. Austral Ecology 27: 433-441. Pinheiro, J. & Bates, D. 2000. Mixed-Effects Models in S and S-PLUS. Springer-Verlag, New York, NY. R Core Team 2013. A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. Raffaele, E., Veblen, T.T., Blackhall, M. & Tercero-Bucardo, N. 2011. Synergistic influences of introduced herbivores and fire on vegetation change in northern Patagonia, Argentina. Journal of Vegetation Science 22: 59-71. Robinson, L. 1991. Field guide to the native plants of Sydney. Kangaroo Press, Kenthurst, NSW, Australia. Rooney, T.P. 2001. Deer impacts on forest ecosystems: a North American perspective. Forestry 74: 201-208. Rooney, T.P. 2009. High white-tailed deer densities benefit graminoids and contribute to biotic homogenization of forest ground-layer vegetation. Plant Ecology 202: 103-111. Rooney, T.P. & Waller, D.M. 2003. Direct and indirect effects of white-tailed deer in forest ecosystems. Forest Ecology and Management 181: 165-176. Royo, A.A. & Carson, W.P. 2006. On the formation of dense understory layers in forests worldwide: consequences and implications for forest dynamics, biodiversity, and succession. Canadian Journal of Forest Research 36: 1345-1362.

137

Royo, A.A., Collins, R., Adams, M.B., Kirschbaum, C. & Carson, W.P. 2010a. Pervasive interactions between ungulate browsers and disturbance regimes promote temperate forest herbaceous diversity. Ecology 91: 93-105. Royo, A.A., Stout, S.L., de Calesta, D.S. & Pierson, T.G. 2010g. Restoring forest herb communities through landscape-level deer herd reductions: Is recovery limited by legacy effects? Biological Conservation 143: 2425-2434. Skre, O., Wielgolaski, F.E. & Moe, B. 1998. Biomass and chemical composition of common forest plants in response to fire in western Norway. Journal of Vegetation Science 9: 501- 510. Spencer, R.-J. & Baxter, G.S. 2006. Effects of fire on the structure and composition of open eucalypt forests. Austral Ecology 31: 638-646. Stephens, S.L. & Ruth, L.W. 2005. Federal forest-fire policy in the United States. Ecological Applications 15: 532-542. Tanentzap, A.J., Bazely, D.R., Koh, S., Timciska, M., Haggith, E.G., Carleton, T.J. & Coomes, D.A. 2011. Seeing the forest for the deer: Do reductions in deer-disturbance lead to forest recovery? Biological Conservation 144: 376-382. Taws, N. 1998. Fire and vegetation management in Jervis Bay Territory, Report. Environment Australia, Canberra. Tolhurst, K.G. & Turvey, N.D. 1992. Effects of bracken (Pteridium esculentum (forst. f.) Cockayne) on eucalypt regeneration in west-central Victoria. Forest Ecology and Management 54: 45-67. Tremblay, J.-P., Huot, J. & Potvin, F. 2006. Divergent nonlinear responses of the boreal forest field layer along an experimental gradient of deer densities. Oecologia 150: 78-88. Tylianakis, J.M., Didham, R.K., Bascompte, J. & Wardle, D.A. 2008. Global change and species interactions in terrestrial ecosystems. Ecology Letters 11: 1351-1363. Wisdom, M.J., Vavra, M., Boyd, J.M., Hemstrom, M.A., Ager, A.A. & Johnson, B.K. 2006. Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildlife Society Bulletin 34: 283-292. Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A. & Smith, G.M. 2009. Mixed effects models and extensions in ecology with R. Springer, New York, USA.

Supporting Information

Additional supporting information may be found in the online version of this article:

Appendix S1. Plant species list

Appendix S2. Tables of model rankings

138

Appendix S1. Plant species list

Appendix S1. List of species sampled in experimental plots and treatment associations of individual species that were identified by indicator species analysis as having a significant association with one or more treatments. Nomenclature follows Robinson (1991). Letter codes for treatment associations are as follows: BO – burnt, open; BP – burnt, partial exclosure; BE – burnt, exclosure; UO – unburnt open; UP – unburnt partial exclosure; UE – unburnt exclosure. Life-form Species Treatment associations P Ferns Calochlaena dubia Pteridium esculentum BO, BP, UO 0.019

Palms Livistona australis

Grasses Austrostipa pubescens Entolasia marginata Entolasia stricta Imperata cylindrica Microlaena stipoides Oplismenus aemulus Oplismenus imbecillis Poa labillardierei Themeda australis BE, UE, UP 0.024

Sedges / Rhizomatous Perennials Ficinia nodosa Hypolaena fastigiata Lepidosperma concavum Lomandra cylindrica Lomandra filiformis Lomandra glauca Lomandra gracilis Lomandra longifolia Lomandra micrantha Lomandra multiflora Restio fastigiatus Xanthorrhoea resinifera

Herbs and Forbs Acianthus sp. Actinotus helianthi Caladenia catenata Commelina cyanea Corybas sp. Dianella caerulea Dichondra repens Galium gaudichaudii Galium propinquum BE, UE, UP 0.042 Gastrodia sesamoides Lagenifera stipitata Patersonia sericea Poranthera microphylla Pratia purpurascens Pterostylis sp. Scaevola ramosissima Schelhammera undulata

139

Life-form Species Treatment associations P Selaginella uliginosa Viola hederacea

Climbing / Scrambling Herbs Cassytha pubescens Comesperma volubile Desmodium brachypodum Desmodium rhytidophyllum Desmodium varians Glycine clandestina Glycine microphylla Glycine tabacina

Sub-shrubs Billardiera scandens Coronidium elatum Dampiera purpurea Dampiera stricta Gonocarpus teucriodes Hardenbergia violacea Opercularia aspera Marsdenia suaveolens BE, BP, UE 0.017

Shrubs Acacia suaveolens Acacia terminalis Acacia ulicifolia Bossiaea ensata Bossiaea heterophylla Breynia oblongifolia Correa reflexa Dillwynia sp. Epacris pulchella Glochidion ferdinandi Hibbertia aspera Hibbertia empetrifolia Hibbertia linearis Hibbertia obtusifolia Hovea linearis Leptomeria acida Leucopogon lanceolatus Melichrus sp. Monotoca elliptica Monotoca scoparia Petrophile pulchella Phyllanthus hirtellus Pimelea linifolia Platylobium formosum Platysace lanceolata Pultenaea paleacea Pultenaea rosmarinifolia Ricinocarpos pinifolius Styphelia triflora Xanthosia pilosa Zieria smithii

Vines Cissus hypoglauca Clematis aristata Eustrephus latifolius Hibbertia dentata Hibbertia scandens Marsdenia rostrata

140

Life-form Species Treatment associations P Parsonsia straminea Sarcopetalum harveyanum Smilax glyciphylla

Trees / Tree Seedlings Acacia longifolia Allocasuarina seedling Banksia serrata Ceratopetalum gummiferum Corymbia gummifera Elaeocarpus reticulatus Eucalyptus seedling Notelaea longifolia Persoonia linearis BE, BO, BP 0.012 Pittosporum undulatum Syncarpia glomulifera Synoum glandulosum BE, BP, UE 0.047

References Robinson, L. 1991. Field guide to the native plants of Sydney. Kangaroo Press, Kenthurst, NSW, Australia.

141

Appendix S2. Model rankings for linear mixed models (LMMs) and generalised linear mixed models (GLMMs).

Table S1. Model rankings for linear mixed models (LMMs) testing the fixed effects of fire (B, burnt or unburnt), herbivory (H, number of scats per 100m2) and time since fire (T, 3, 9, 15 or 21 months) and their interactions, plus the starting condition (Pre), on plant community diversity metrics (species richness, Simpson‘s reciprocal diversity index, Simpson’s evenness, and Berger-Parker dominance). All models included the random effects of block/site. Details are provided for all models with ΔAICc < 4. Data Model Terms K LL AICc ΔAICc AICcWt Species richness Pre + B + T + B:T 12 - 222.09 471.9 0 0.498 Pre + B + H + T + B:T 13 -221.23 472.9 0.96 0.309 Pre + B + H + T + B:H + B:T 14 -221.15 475.5 3.54 0.085

Simpson’s reciprocal diversity (1/D) Pre + B + H + T + B:H + B:T 14 - 150.94 335.1 0 0.158 Pre + T 8 -159.14 335.9 0.87 0.103 H + T 8 -159.19 336 0.96 0.098 Pre + H + T 9 -157.99 336.1 1.01 0.095 B + H + T + B:H + B:T 13 -153.05 336.5 1.47 0.076 T 7 -161.00 337.3 2.21 0.052 Pre + B + H + T + B:H 11 -156.19 337.5 2.46 0.046 B + H + T 9 -158.80 337.7 2.63 0.043 Pre + B + T 9 -158.84 337.8 2.71 0.041 Pre + B + H + T 10 -157.70 338 2.92 0.037 Pre + B + T + B:T 12 -155.14 338 2.97 0.036 B + H + T + B:T 12 -155.32 338.4 3.33 0.03 B + H + T + B:H 10 -157.91 338.4 3.35 0.03 Pre + B + H + T + B:T 13 -154.16 338.8 3.69 0.025 B + T 8 -160.59 338.8 3.78 0.024 B + T + B:T 11 -156.89 338.9 3.85 0.023

Evenness (E 1/D) Pre + B + H + T + B:H + B:T 14 161.53 - 289.9 0 0.534 B + H + T + B:H + B:T 13 158.98 -287.5 2.35 0.165 Pre + B + H + T + B:H + B:T + H: T 17 163.92 -286 3.89 0.076

Dominance (d) Pre + B + H + T + B:H 11 58.75 - 92.4 0 0.302 Pre + B + H + T + B:H + B:T 14 62.52 -91.9 0.51 0.234 Pre + B + H + B:H 8 53.94 -90.2 2.13 0.104 B + H + T + B:H 10 56.09 -89.6 2.77 0.076 B + H + T + B:H + B:T 13 59.58 -88.7 3.65 0.049

142

Table S2. Model rankings for generalised linear mixed models (GLMMs) testing the fixed effects of fire (B, burnt or unburnt), exclosure treatment (E, open or partial), and time since fire (T, 3, 9, 15, or 21 months) on site-level herbivore activity (number of scats per 100m2). All models included the random effects of block/site. Details are provided for all models with ΔAICc < 4. Model terms K LL AICc ΔAICc AICcWt B + Ep + T 9 -281.81 584.9 0 0.330 B + Ep+ T + B:T 12 -277.65 585.4 0.46 0.262 Ep + T 8 -284.02 586.7 1.71 0.141 B + E + T + B:Ep 10 -281.69 587.5 2.58 0.091 B + Ep + T + B:Ep+ B:T 13 -277.50 588.3 3.33 0.062

143

Paper IV: Interactive effects of fire and large herbivores on web-building spiders

Understanding of how the effects of large herbivores cascade through ecosystems is limited (see Paper I), but is important for both detecting and managing impacts. The effects of large herbivores on other animals are often attributed to changes in vegetation (Paper I). However, in Papers II and III, I found that plants and animals differed in their responses to the fire-herbivory interaction. In this paper, I examined the extent to which vegetation mediated the effects of fire and herbivory on web-building spiders – a group that is usually sensitive to changes in habitat structure.

Foster, C.N, Barton, P.S., Wood, J.T. and D.B. Lindenmayer (2015) Interactive effects of fire and large herbivores on web-building spiders. Oecologia. 179, 237-248.

145

Abstract

Altered disturbance regimes are a major driver of biodiversity loss worldwide. Maintaining or recreating natural disturbance regimes is therefore the focus of many conservation programs. A key challenge, however, is to understand how co-occurring disturbances interact to affect biodiversity. We experimentally tested for the interactive effects of prescribed fire and large macropod herbivores on the web-building spider assemblage of a eucalypt forest understorey, and investigated the role of vegetation in mediating these effects using path analysis. Fire had strong negative effects on the density of web-building spiders, which were partly mediated by effects on vegetation structure, while negative effects of large herbivores on web density were not related to changes in vegetation. Fire amplified the effects of large herbivores on spiders, both via vegetation-mediated pathways, and by increasing herbivore activity. The importance of vegetation-mediated pathways and fire-herbivore interactions differed for web density and richness, and also differed between web types. Our results demonstrate that for some groups of web building spiders, the effects of co-occurring disturbance drivers may be mostly additive, whereas for other groups, interactions between drivers can amplify disturbance effects. In our study system, the use of prescribed fire in the presence of high densities of herbivores could lead to reduced densities and altered composition of web-building spiders, with potential cascading effects through the arthropod food web. Our study highlights the importance of considering both the independent and interactive effects of disturbances, as well as the mechanisms driving their effects, in the management of disturbance regimes.

Keywords

Araneae, browsing, disturbance interaction, grazing, synergistic effects

146

Introduction

Disturbance regimes drive the structure and function of ecosystems worldwide, and altered disturbance regimes are an important cause of biodiversity loss (Sinclair and Byrom 2006).

Maintaining or recreating appropriate disturbance regimes is therefore the focus of many conservation and restoration programs (Hobbs and Huenneke 1992; Halme et al. 2013).

Two of the most common and widely studied disturbance drivers in terrestrial systems are herbivory and fire (Bond and Keeley 2005; Danell et al. 2006). Browsing or grazing by large mammalian herbivores has been shown to shape the structure and function of ecosystems, from plant and animal communities, through to nutrient cycles and even climate (Côté et al.

2004; Danell et al. 2006; Foster et al. 2014). Fire is an episodic disturbance, and the frequency, intensity and spatial extent of fires also drives the structure and function of ecosystems

(Thonicke et al. 2001; Bond and Keeley 2005). Both of these disturbances can affect biota either directly (e.g. through direct mortality), or indirectly (e.g. by modifying habitat), or both

(Thonicke et al. 2001; Côté et al. 2004). A key mechanism linking both large herbivores and fire to effects on biological communities is altered vegetation structure and complexity (Bond and Keeley 2005; Foster et al. 2014).

As disturbances rarely occur in isolation, understanding how disturbance drivers interact to affect biota is critical for effective conservation management (Wisdom et al. 2006;

Didham et al. 2007; Mantyka-pringle et al. 2012). Although the ecological effects of disturbance have been widely studied, investigations of the interactive effects of disturbance drivers are much less common (Wisdom et al. 2006; Didham et al. 2007; Foster et al. 2014).

Interactions between fire and herbivory have been reported in a range of ecosystems, and can occur in a number of ways. For example, patterns and intensity of herbivory can affect fuel loads and hence modify the spatial extent or intensity of fire (Wisdom et al. 2006; Kimuyu et al.

2014). Similarly, as many herbivores are attracted to the new growth available in recently burnt areas, fire can affect the spatial distribution and intensity of herbivory (Allred et al. 2011). Fire and herbivory also can interact via what is termed an interaction modification, where fire changes the mode of action or per-unit effect of herbivory on organisms (sensu Didham et al. 147

2007). For example, Royo et al. (2010) found that moderate levels of deer browsing increased understorey plant richness in burnt deciduous forest, but not in unburnt forest. Interactions between disturbance drivers can be synergistic (i.e. effects magnified, e.g. Barton et al. 2011) or antagonistic (i.e. effects diminished or reversed, e.g. Matlack et al. 2001), and by definition, differ from what would be predicted from the additive effects of each driver occurring in isolation (Didham et al. 2007; Crain et al. 2008).

Most studies of fire × large herbivore interactions have investigated effects on vegetation (e.g. Royo et al. 2010; Kerns et al. 2011), and to date only a small number of studies have investigated how these interactions affect animal assemblages. Among these studies, there appears to be a consistency between vegetation and animal responses; most studies which report interactive effects on animals also report interactive effects on vegetation (e.g. Matlack et al.

2001, deer mice; Bailey and Whitham 2002, arthropods), while studies which find no interactive effects on animals also find no interactive effects on vegetation (e.g. Jonas and Joern 2007, grasshoppers; Underwood and Christian 2009, ants).

Web-building spiders are a group of animals that have been found to respond to both fire (Buddle et al. 2000) and large herbivores (Miyashita et al. 2004; Warui et al. 2005), but the responses of these spiders to fire-herbivore interactions have not previously been studied. Web- building spiders respond strongly to changes in vegetation structure (Langellotto and Denno

2004), and are the dominant invertebrate predators in terrestrial food webs (Riechert and

Lockley 1984; Carter and Rypstra 1995). Therefore, disturbance-induced changes in vegetation structure may modify spider densities, which could have important consequences for trophic dynamics (Schmitz 2008). In addition, different types of web-builders may differ in their response to habitat complexity, for example, Halaj et al. (2000) found that sheet weaving spiders, which build complex three-dimensional webs, responded more negatively to habitat simplification than orb-weaving spiders. As different types of web-builders target different prey items (Nyffeler 1999), changes in vegetation structure also may affect food web structure by altering the composition of the predator guild. As disturbance effects can cascade through ecosystems via both trophic and non-trophic pathways (Ohgushi 2005), it is important to 148

understand the extent to which disturbance effects are mediated by changes in vegetation, and whether managing disturbances to maintain vegetation condition will also cater for the requirements of fauna (Clarke 2008).

In this study, we tested for the effects of fire, large macropod herbivores and their interaction on the web-building spider assemblage of a eucalypt forest understory. We used a randomised, blocked experiment, combining prescribed fire and herbivore exclusion treatments, to address three questions: (1) Do large herbivores and fire interact to affect the density or richness of web-building spiders? (2) To what extent are the effects of these disturbances mediated by changes in vegetation? And (3) Do these effects differ between different types of web-builders? We predicted that both fire and large herbivores would reduce web-building spider density and richness by reducing the structural complexity of vegetation, with sheet web- builders responding most strongly to vegetation simplification.

Materials and methods

Study site

We conducted our study in Booderee National Park (BNP); a ~6 500 ha peninsula in south- eastern Australia (35°10′S, 150°40′E, see Online Resource 1). We established sites within the

Eucalyptus pilularis forest of BNP, which is the most widespread vegetation type in the park

(Barton et al. 2014). An intensive baiting program targeting the introduced red fox (Vulpes vulpes) has been in place in BNP since 1999 to protect native small and medium-sized mammal species from predation (Dexter et al. 2012). Loss of native predators and a lack of human hunting mean that without foxes, predation pressure on native macropod herbivores is low

(Lindenmayer et al. 2014). Over the last decade, there has been a tenfold increase in the numbers of these herbivores in BNP (predominantly Wallabia bicolor, a generalist browser and

Macropus giganteus, a grazer, Family Macropodidae) (Dexter et al. 2012; Lindenmayer et al.

2014). A short-term exclosure trial has indicated that this high abundance of native herbivores could be driving a shift in vegetation composition (Dexter et al. 2013). As prescribed fire is

149

commonly used in eucalypt forests to reduce the risk of high intensity wildfire, promote regeneration of senescing vegetation, and/or increase habitat heterogeneity (Williams et al.

1994), it is important to understand how fire interacts with high abundances of herbivores to affect biodiversity.

Study design

We tested the interactive effects of prescribed fire and large herbivores on understorey vegetation and web-building spiders using a randomised blocked experiment. We combined three levels of herbivore treatment (open, partial and exclosure) and two levels of burning treatment (burnt and unburnt) in a factorial design. We replicated each of these six treatment combinations across four experimental blocks to give a total of 24 sites (Online Resource 1).

We created the herbivore treatments by using exclosure fences to reduce the density of macropods within 0.125 ha (25 m × 25 m) plots, to produce three treatments: full herbivory

(open treatment), intermediate herbivory (partial treatment – plots were fenced but gates opened and closed at two month intervals to create a lower browsing pressure) and no herbivory

(exclosure treatment). Exclosure fences were constructed in June 2012 using 1.1 m tall wire fencing, which we found to be effective at excluding macropods (see Results). Smaller animals were observed to move freely through the fence, and other large animals are rare in the park, so effects of exclosure fences were assumed to occur primarily via their effect on macropods. For the burning treatments, small (50 × 50 m), low-intensity burns were conducted across half of the herbivory treatment sites in August 2012.

Data collection

We sampled spider webs and vegetation within four 3 × 3 m plots within each site (one in each of the four quarters of the site). Data were collected three months post-fire (November 2012) and 15 months post-fire (November 2013). Plots were established at least 1.5 m from the edge of the site.

We counted spider webs as a surrogate measure for the web-building spider assemblage. This method was described and tested by Gollan et al. (2010), who found that the 150

diversity of web-types (based on web architecture) was strongly correlated with the diversity of spider genera in a site. As the richness of spider genera can be a viable surrogate for the species richness of spiders (Foord et al. 2013), this method allowed us to assess compositional and diversity responses of spiders, in addition to density responses. We used vaporised water, applied with a pressurised spray mister, to assist in locating webs, and counted and identified all spider webs within each of the four plots in each site. Each web was assigned to one of 32 web types based on their architecture, according to the key of Smith (2008) (see Fig. 1 for examples). From this, we generated measures of web density (number of webs per 3 x 3 m plot), web richness (the number of different web types per plot) and web composition (the assemblage composition of web types) for each plot. To minimise variation due to weather conditions, surveys were delayed for 48 hours following strong wind or rain to allow spiders to rebuild damaged webs.

We recorded the following vegetation variables from each plot; total understorey foliage projective cover (the proportion of ground area covered by foliage held vertically above it, Specht and Morgan 1981), foliage projective cover of vascular plants by life-form, understorey height (measured at 10 evenly spaced locations per plot using the stick and foam disc method of Smit et al. (2001), disc 100 mm diameter, 4.7 g weight), litter depth (measured at 10 evenly spaced locations per plot) and stem density (number of live woody stems at ground level within a 1 × 1 m sub-plot). Life-forms were grouped into the broad structural categories of ferns, grasses (Poaceae), herbs (including forbs and climbers), sedges (which included grass- like perennial herbs) and shrubs (including sub-shrubs and tree seedlings).

We measured the effects of burning and the herbivory treatments on herbivore activity using scat (pellet) counts. Scat counts are a commonly used method to assess the comparative density of macropod herbivores between sites (e.g. Howland et al. 2014; Pedersen et al. 2014).

We counted the number of macropod scats along two 25 × 2 m transects (100 m-2) in each site approximately every two months from August 2012 to December 2013. Counts from the two transects were summed to give one count per site. We removed scats from transect lines on each sampling event to avoid double counting. For each site, we summed the count preceding and 151

that following the spider web counts to give an approximate index of herbivore activity in each site at the time of the web count.

Figure 1. Examples of web types encountered in surveys. One example of each of the four groups of web types is shown: (a) web 9, an orb web, (b) web 28, a sheet web, (c) web 23, a lace web and (d) web 32, a tangle web.

Data analysis

Effects of fire and large herbivores on web density and richness

We used generalised multi-level path models (Shipley 2009) to test the effects of fire and large herbivores on spider web density and richness, and the extent to which these effects were

152

mediated by vegetation changes. Generalised multi-level path analysis uses directional- separation (d-sep) tests to assess the goodness of fit of a hypothesised causal diagram to the patterns of dependence and independence within a dataset (Shipley 2002). This analysis is based on a structural equation modelling (SEM) framework, but has been generalised to accommodate hierarchical designs and non-normal response variables (Shipley 2009).

We hypothesised that the effects of fire, herbivores and their interaction on spiders would be largely vegetation-mediated, but would also occur via other mechanisms that operate independently of vegetation structure (Fig. 2). Our rationale for the construction of this causal model was as follows: Fire can affect spiders through changes in vegetation structure (Brennan et al. 2006), or via other mechanisms that operate independently of vegetation (the direct fire- spider pathway in Fig. 2), such as fire-induced mortality (Bell et al. 2001), or changes in prey availability (York 1999). Similarly, large herbivores have been found to affect spiders by modifying vegetation structure (Miyashita et al. 2004), but could also affect spiders through other mechanisms (the direct herbivore-spider pathway in Fig. 2) such as physical disturbance of webs (Chmiel et al. 2000), or changes to key resources such as arthropod prey (Foster et al.

2014). We also identified three possible pathways by which fire could interact with large herbivores to affect spiders; (1) a chain effect, where fire attracts herbivores to burnt areas, increasing the level of herbivore activity in burnt sites (Allred et al. 2011), (2) an interaction modification, where fire alters plant traits, modifying the effect of herbivores on vegetation

(Augustine and McNaughton 1998), or (3) an interaction modification where fire increases the vulnerability of web spiders to other impacts of large herbivores (e.g. by reducing prey availability, increasing the likelihood of spiders abandoning webs after physical disturbance,

Chmiel et al. 2000).

Confirmatory path analysis does not allow for reciprocal effects among variables

(Shipley 2009), but a number of vegetation variables we measured were likely to be reciprocally related (for example, high grass cover would lead to low average understory height and low stem density in a site). Therefore, we selected three vegetation variables that were not correlated with each other, but made up a large component of the vegetation and were correlated with 153

other vegetation variables, to include in the path analysis (cover of shrubs, sedges and ferns).

While this approach avoided reciprocity between variables, in excluding some vegetation variables, we potentially omitted variation in vegetation structure that could explain spider responses to disturbance. Therefore, we also conducted a sensitivity analysis (sensu Ruffell et al. 2014) to test whether including additional vegetation variables in our path model would increase the extent to which disturbance effects were mediated by vegetation (Online Resource

2). To do this, we used principal components analysis to reduce our nine vegetation variables to seven orthogonal components, and used these as the measures of vegetation structure in our path analysis. We did not use principal components in the main analysis as principal components are less correlated than would be expected by chance, and including them in the path analysis would reduce our chance of rejecting an incorrectly specified path model (Ruffell et al. 2014).

Fire Herbivores

Vegetation-mediated Vegetation effects Structure Other effects Interaction modification Interaction chain Spiders

Figure 2. Hypothesised causal diagram of the effects of fire and large herbivores on web- building spiders. We predicted that the effects of fire and herbivores on spiders would be largely vegetation-mediated, but may also occur via other mechanisms that operate independently of vegetation structure. We also predicted that fire may potentially interact with large herbivores by increasing herbivore activity (an interaction chain), or by modifying the effects of large herbivores (an interaction modification).

For each of web density and web richness, we constructed two causal diagrams, one for each year of the study, with the cover of shrubs, sedges and ferns as measures of vegetation structure (each with their own node). For each of these diagrams, we tested goodness of fit using d-sep tests (Shipley 2009). Once we had tested the goodness of fit of the full path model, 154

we simplified the model to a more parsimonious one using a backward selection approach

(sensu Ruffell et al. 2014). Backward selection based on minimising AIC was used to simplify each sub-model within the full model (Zuur et al. 2009). Each sub-model was a mixed-effects model fitted with maximum likelihood estimation, and included an endogenous variable as the response and its direct causal parents as predictors. Once all sub-models had been simplified, they were grouped back into a single path model and the fit of this model was tested using generalised multilevel path analysis and d-sep tests as described above.

All analyses were completed using R (R Core Team 2013). For GLMMs we used the glmer function in the package lme4, using a Poisson distribution and log-link function (Bates et al. 2014), while for LMMs we used the lme function in the package nlme (Pinheiro et al. 2014).

Random effects for all GLMMs and LMMs were sites within blocks (block/site). When used as predictors, count variables were ln-transformed and continuous variables were centred on their means. Response variables were checked for over-dispersion and model residuals were inspected to verify that the data met model assumptions (Zuur et al. 2009). After deciding on the final model for both web abundance and richness, we calculated path coefficients as the estimated slopes of the variables within each of the sub-models (Shipley 2009). Unstandardized path coefficients can be interpreted as the change in the response variable for a one unit change in the predictor variable (Aiken and West 1991). However, in the presence of an interaction, these path coefficients are conditional effects. Because all predictors were centred, the conditional effects can be interpreted as the effect of the predictor on the response variable when the interacting predictor is held at its mean value, or in the case of categorical predictors

(fire and herbivore exclosure treatments) at the control value (unburnt and open treatments respectively). The interaction term indicates the amount of change in the slope of the regression of the response on the predictor with a one-unit change in the value of the interacting predictor.

Effects of fire and large herbivores on different web types

To test whether different types of webs responded differently to large herbivores and fire, we estimated unstandardized regression coefficients when the density of that web type was

155

substituted for overall web density in the full path diagram. Each of the individual web types was assigned to one of four groups based on similarity of web characteristics - orb webs, sheet webs, lace webs and tangle webs (Fig. 1, Online Resource 3). Such groupings are commonly used in studies of web-building spiders, and while there is some overlap, these categories broadly sort spiders into groups of families (e.g. Halaj et al. 2000). Orb weavers are generally in the families Araneidae, Tetragnathidae, and Uloboridae, sheet weavers in the families

Linyphiidae, Theridiidae and Stiphidiidae, lace weavers belong to Desidae, and tangle web spinners are generally in the family Theridiidae (Online Resource 3). We used a separate

GLMM for each web type grouping in each year with a Poisson distribution and log-link function and random effects of block/site. We used backward selection as described above to simplify the model for each web type in each year.

We also analysed the effect of our treatments on the composition of individual web types using partial (or conditioned) canonical correspondence analysis in R (R Core Team 2013), using the “cca” function in the package “vegan” (Oksanen et al. 2013), and using the Bray-

Curtis dissimilarity measure. This analysis allowed us to partial out the spatial variation associated with the experimental blocks, before analysing the variation in web type composition that was associated with our experimental treatments (herbivory × fire × year) (Borcard et al.

1992). We then ran permutation tests using the function “anova.cca” to test the significance of our constraints (treatment combinations) using 10 000 permutations of the data. To reduce the incidence of zero values in the data caused by rare web types, we pooled web counts to the site level for this analysis and excluded web types occurring at two or fewer sites.

Results

Density and richness of spider webs

We counted 3687 spider webs, representing 28 different web types, across the two sampling periods (Online Resource 3). Generalised multilevel path analysis indicated that herbivores and

156

fire affected spider web density via both vegetation-mediated and other pathways, but effects on web type richness were of short duration and not mediated by changes in vegetation.

For web density, the simplified path models (2012: χ2 = 34.7, df = 30, P = 0.25, 2013: χ2

= 29.0, df = 28, P = 0.41) gave as good fit as the full models (2012: χ2 = 25.9, df = 22, P = 0.26,

2013: (χ2 = 28.5, df = 22, P = 0.16) in both years. The simplified model for web density in 2012 indicated that fire had a strong negative effect on web density, which was partly mediated by negative effects of fire on vegetation cover (Fig. 3a). In contrast, the small, but significant, negative effect of large herbivores on web density was not mediated by changes in vegetation structure. By 2013, the effects of fire on vegetation variables were weaker, which was associated with a recovery in vegetation (Fig. 3b). Both the vegetation-mediated effects, and the other effects, of fire on web-density were reduced in 2013 compared with 2012. While the effect of fire on shrub cover was reduced in 2013, shrub cover remained a strong determinant of web density in 2013. In 2013 there also was a marginally significant interaction between fire and herbivores, where herbivore activity was positively associated with fern cover in burnt, but not in unburnt sites (Fig. 3b, Online Resource 2 - Figs. OR3a,b). High fern cover was associated with higher web density in the path model (Fig. 3a). As in 2012, in 2013 there was a small negative effect of herbivores on web density which was not explained by vegetation structure.

Sensitivity analysis indicated that including all vegetation variables in the path model did not increase the amount of variation in web density that was explained by vegetation-mediated pathways (Online Resource 2).

Model fits for web type richness were similar to those for web density, where our simplified model (2012: χ2 = 35.6, df = 34, P = 0.39, 2013: χ2 = 30.2, df = 34, P = 0.65) gave as good a fit as the full model (2012: χ2 = 23.8, df = 22, P = 0.35, 2013: χ2 = 25.0, df = 22, P =

0.30) in both years of the study. The simplified path models indicated that the effects of disturbance on web-type richness were not mediated by changes in vegetation structure (Fig.

3c,d). Fire had a short-term negative effect on web-type richness which was no longer evident in 2013 (15 months after fire). Large herbivores had a small negative effect on web-type richness in both years, and this effect was strongest in 2012, when herbivore activity was 157

elevated in burnt sites (Fig. 3c,d). As with web density, including all vegetation variables in the path analysis did not increase the strength of vegetation-mediated effects on web type richness

(Online Resource 2).

(a) Web density Partial Exclosure (b) Web density Partial Exclosure 2012 2013 -1.7 *** -3.4*** -1.0 *** -4.0***

0.62*** -0.14^ Fire Herbivores Fire Herbivores

0.03^

-0.08*** -0.15*** -0.04* -0.03^ -0.06* 0.03 -0.004

Shrub Sedge Fern Shrub Sedge Fern

1.13*** 0.96** 1.8*** 0.43^

Web density Web density -0.61*** -0.13* -0.46*** -0.09*

(c) Web type richness Partial Exclosure (d) Web type richness Partial Exclosure 2012 2013 -1.7 *** -3.4*** -1.0 *** -4.0***

0.62*** -0.14^ Fire Herbivores Fire Herbivores

0.03^

-0.08*** -0.15*** -0.04* -0.03^ -0.06* 0.03 -0.004

Shrub Sedge Fern Shrub Sedge Fern

Web type richness Web type richness -0.46*** -0.06* -0.03

Driver makes effect Driver makes effect Positive effect Negative effect more negative more positive

Figure 3. The effects of fire, herbivores and their interaction on web-building spider density (a,b) and richness (c,d), and the extent to which these effects were mediated by vegetation. Arrows represent causal paths between the experimental treatments (square boxes) and plant and animal responses (rounded boxes). Path coefficients are unstandardized partial regression coefficients. Dashed arrows indicate non-significant relationships, significance levels of path coefficients: ^ P ≤ 0.1, * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001. Herbivore activity was ln transformed when used as a predictor.

158

Responses of different web types

Substituting different web type groupings into the full model in place of web density revealed key differences in the factors affecting different types of webs. Orb web density was higher on sites with higher shrub cover, a relationship which mediated the negative effects of fire on orb webs in both years (Table 1). In 2013, orb webs also were positively associated with higher fern cover. In contrast, while some of the effects of fire on sheet webs were vegetation-mediated, there was a strong negative effect of fire on sheet webs in both 2012 and 2013, which was not vegetation mediated (Table 1). Sheet webs also responded negatively to herbivore activity on both years, a response which was also not explained by changes in vegetation. Tangle webs and lace webs showed only weak or inconsistent responses to the experimental treatments, with tangle webs showing a shrub-mediated negative response to fire and a negative response to herbivore activity in 2012, but no responses to disturbance in 2013 (Table 1).

Table 1. Estimated coefficients (and SE) of parameters for the four different types of webs, when substituted for overall web density in the path diagrams for 2012 and 2013. All models included the blocking structure of block/site as a random effect. Significance levels: ^ P ≤ 0.1, * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001

Model term Estimate (se)

Orb webs Sheet webs Tangle webs Lace webs

2012 2013 2012 2013 2012 2013 2012 2013 Shrub 1.23 2.87 - - 2.18 - - -11.1 cover (0.64)^ (0.63)*** (0.77)** (4.0)** Sedge - - 2.11 -1.34 - - - -2.9 cover (0.37)*** (0.49)** (1.9) Fern Cover -1.42 1.55 - -1.25 - - - - (0.90) (0.41)*** (0.45)** Fire -0.76 - -1.6 -1.84 - -0.07 - - (0.24)** (0.23)*** (0.29)*** (0.12) Herbivores -0.13 - -0.19 -0.19 -0.17 0.02 - - (0.08)^ (0.07)** (0.08)* (0.05)*** (0.05) Fire * ------0.12 - - herbivores (0.07)

159

Compositional differences in web-types reflected the strongest effects in the analysis of web types. The first two axes of the partial CCA analysis accounted for 14.8 % and 12.4 % of the variation in web type composition, respectively. There was a significant interaction between year and burning treatment (P = 0.03, Fig. 4). There were also significant differences in web type composition between burning treatments (P < 0.01), and years (P < 0. 01), but not between herbivory treatments or any of its interaction terms (all P > 0.05). The composition of burnt sites appeared to be a sub-set of unburnt sites, with the two most common sheet webs (web 25 and 27, belonging to the families Theridiidae and Linyphiidae respectively), strongly associated with unburnt sites (Fig. 4).

Figure 4. Site scores (linear combinations of variable scores) for axis 1 and 2 of the partial canonical correspondence analysis on the distribution of individual web types with respect to the experimental treatments (burning × herbivory × year), after blocking effects had been partialed out. Ellipses indicate one standard deviation from the centroid of each burning × year treatment combination. Numbers identify individual web types: 1-20 – orb webs, 21-24 lace webs, 25-29 sheet webs, 30-32 tangle webs (see Online Resource 3 for individual descriptions). Overlapping web numbers are replaced with points (+).

160

Discussion

We used a combination of burning and herbivore exclusion treatments to test the combined effects of fire and large herbivores on web-building spiders in a forest understory. Fire and large herbivores interacted via a chain effect, where fire increased herbivore activity, which in turn affected the web-building spider assemblage. Fire and large herbivores also interacted via vegetation-mediated pathways to affect spiders, although this interaction was only marginally significant. The extent to which vegetation structure mediated disturbance effects and the importance of interactive effects differed for web density and richness, and also between web types. Strong effects of disturbance on web-building spiders that were not mediated by vegetation indicate that managing disturbances to maintain vegetation structure is unlikely to adequately address the needs of fauna in this system.

Fire, but not herbivore effects on web density were vegetation-mediated

Web density was most strongly affected by fire, and much of this effect was mediated by changes in vegetation structure. The loss and subsequent recovery of vegetation after fire (Fig.

3) was mirrored by a loss and partial recovery of spider web density, with spiders likely responding to changes in web-site availability and litter accumulation that occurred with changes in vegetation cover (Brennan et al. 2006; Podgaiski et al. 2013). However, fire also had strong effects on web density that were not explained by vegetation responses, and these effects persisted in the second year of the study (Fig. 3a,b). As our experimental burns were small (50 x

50 m), and many spiders are able to rapidly disperse via aerial ballooning (Bell et al. 2005;

Langlands et al. 2011), this strong residual effect of fire is unlikely to be limited by recolonization ability. We suggest that other changes that can occur after fire, but were not measured this study, such as reduced soil/litter moisture, or reduced abundances of arthropod prey, may have limited the re-establishment of high densities of spiders following fire

(Neumann and Tolhurst 1991; York 1999).

Fire and large herbivores interacted to increase the cover of ferns, which had a weak positive effect on web density. This interaction likely occurred due to selective browsing by

161

herbivores allowing the less palatable bracken to dominate in burnt, browsed environments

(Augustine and McNaughton 1998). This increased cover of ferns was positively associated with web density, with fern cover likely providing suitable web sites in an environment where cover of other lifeforms remained low (Online Resource 4). Overall, very little of the effect of large herbivores on spiders was mediated by vegetation, which contrasts with a number of previous studies, which have attributed negative effects of large herbivores on spiders to changes in vegetation structure (e.g. Miyashita et al. 2004; Warui et al. 2005).

Path analysis revealed that large herbivores negatively affected web density (and also web type richness) via non-vegetation-mediated pathways, likely through the physical disturbance of webs. Web damage is one of the main triggers for a spider to abandon a web site

(Chmiel et al. 2000). Repeated web disturbance may therefore have reduced web density by causing spiders to move out of sites with high herbivore activity, or by reducing spider fitness, as spiders which move web sites must expend considerable energy in web reconstruction

(Rypstra 1983; Chmiel et al. 2000), and are more vulnerable to predation when moving between sites (Lubin et al. 1993). The effect of herbivores on web density was greatest in 2012 (Fig. 3a) and this was likely due to the chain interaction between fire and herbivores, where herbivore activity was higher in burnt sites. This type of interaction has been reported by many previous studies, where burning focusses herbivore activity in burnt patches (Klop et al. 2007; Allred et al. 2011). This greater level of herbivore activity in recently burnt sites would have led to greater rates of web disturbance, and hence the stronger negative effects on spiders.

Disturbance effects on web type richness were not vegetation-mediated

In contrast to web density, the effects of disturbance on web type richness were of short duration and were not mediated by effects on vegetation structure (Fig. 3c,d). The short-lived effect of fire suggests that spiders were able to rapidly re-colonise sites as the vegetation recovered from fire, a result that is not surprising given the small scale of the burns in our study

(50 x 50 m) and the high capacity of spiders to disperse via both ground movement and aerial ballooning (Bell et al. 2005; Langlands et al. 2011). Such short-term effects of fire on spider

162

richness are consistent with previous studies from fire-prone environments, where post-fire recovery of spiders can be rapid (e.g. Brennan et al. 2006; Podgaiski et al. 2013). Spider richness also was directly affected by herbivore activity, where sites with higher herbivore activity had slightly (but significantly) lower web richness (Fig. 3a). This is likely due to a few web types being particularly vulnerable to physical disturbance, causing them to occur only rarely on sites with high herbivore activity.

Responses differed between web types

Different web types showed clear differences in their response to fire, large herbivores, and their interaction. Orb web weavers showed a strong post-fire recovery which was largely mediated by vegetation, while sheet web weavers showed a strong negative response to fire, with little recovery after 15 months. These differences are attributable to key differences in the ecology of these different types of web spinners. Firstly, orb weavers tend to build their webs in higher strata of the vegetation than sheet weavers (Janetos 1982), and so are more likely to be able to escape being killed by a low intensity fire. Second, orb weavers tend to have high dispersal capabilities, allowing them to re-colonise rapidly following disturbance (Bell et al.

2005). Third, sheet weavers often have a high proportion of litter arthropods in their diets, compared with orb weavers which target aerial prey (Harwood et al. 2003). Therefore, the dry litter conditions which usually occur after fire may have supported low abundances of the decomposers which are key prey items for sheet weavers (Neumann and Tolhurst 1991; York

1999). Fourth, orb webs are more efficient at prey capture than sheet webs (Zschokke et al.

2006), and so orb-web weavers may be able to persist with low post-fire prey densities than sheet web builders. Finally, orb web builders may be better able to use the post-fire vegetation than sheet web builders. Orb weavers responded positively to the recovery of fern cover in burnt sites in 2013, which likely allowed their rapid recovery after fire. In contrast sheet web density was negatively related to the cover of ferns. An abundance of ferns may provide suitable structure for orb weavers to build webs spanning open spaces (Rypstra 1983; Halaj et al. 2000), but may not provide sufficient ground-level structure for sheet webs (Janetos 1982, pers. obs.).

163

While orb weavers, sheet weavers and tangle weavers all responded negatively to the high level of herbivore activity in burnt sites in 2012, sheet weavers were the only group to show a consistent negative response in 2013, when herbivore activity was lower. Many sheet webs are larger than tangle and lace webs, are constructed between, rather than within, plants, and are built close to the ground (e.g. web 28, Fig. 1b), all of which would make them vulnerable to trampling and physical disturbance by herbivores. Further, sheet webs have lower rates of prey capture than orb webs, as well as greater costs of initial web construction

(Zschokke et al. 2006), which may cause sheet weavers to be more likely to abandon a site after web disturbance (Chmiel et al. 2000). As sheet web building spiders were more strongly affected by disturbance and were slower to recover than other types of web builders, disturbance caused a shift in the composition of this important predator guild. As different types of webs target different types of arthropod prey (Nyffeler 1999; Harwood et al. 2003), these compositional changes to the spider assemblage could have important cascading effects through the arthropod food web. Understanding how the effects we observed for spiders affect the rest of the arthropod community therefore remains a key area for future research.

Conclusions

Our study has shown that fire and large herbivores can interact both via vegetation-mediated pathways, and via chain effects, to affect web-building spiders, an important component of the forest understory fauna. In our study system, the use of prescribed fire in areas with high densities of native herbivores could lead to reduced densities and altered composition of web- building spiders, with potential cascading effects through arthropod food webs (Riechert and

Lockley 1984; Carter and Rypstra 1995). Managing the ecosystem to reduce the effects of herbivores on vegetation post-fire (i.e. managing the interaction modification) may be ineffective in mitigating the combined effects of disturbance, as the chain effect of fire on herbivore activity also had important short-term effects on spiders. Our results reinforce the importance of considering both independent and interactive effects, as well as the different types of interactions, when managing disturbance regimes (Wisdom et al. 2006; Didham et al.

2007; Crain et al. 2008). 164

The differing strength of vegetation-mediated pathways between web types observed in our study revealed the importance of understanding the mechanisms driving effects for both predicting interactions and managing disturbance effects (Didham et al. 2007; Crain et al.

2008). In this system, monitoring and managing the effects of disturbance on vegetation (a common approach to management of disturbances such as fire (Clarke 2008)), is unlikely to detect or prevent important changes in the spider assemblage. Understanding of mechanistic pathways is also essential if studies are to be used to inform management in un-studied locations (Ruffell et al. 2014). To identify the mechanisms driving disturbance interactions, long-term, multi-taxon studies, which simultaneously address multiple stressors, will be needed

(Tylianakis et al. 2008; Foster et al. 2014). When applied to such studies, analytical approaches such as path analysis will give valuable insights into the importance of different interaction pathways. This mechanistic understanding will be useful, not only in predicting the outcomes of interacting disturbance drivers, but also in identifying appropriate actions to manage their effects on biodiversity (Didham et al. 2007; Crain et al. 2008).

Acknowledgements

Staff at Booderee National Park conducted prescribed burns. Chris Macgregor assisted with fence construction and plant identification. Many volunteers assisted with data collection. Chloe

Sato provided valuable feedback on the manuscript. The Margaret Middleton Fund, The

Norman Wettenhall Foundation and the Long Term Ecological Research Network provided financial support. The experiments comply with the current laws of Australia where the research was performed (ANU Animal Ethics protocol A2012/24 and Booderee National Park Research

Permit BDR12/00005).

References

Aiken LS, West SG (1991) Multiple regression: Testing and interpreting interactions. Sage Publictions, Newbury Park, USA Allred BW, Fuhlendorf SD, Engle DM, Elmore RD (2011) Ungulate preference for burned patches reveals strength of fire-grazing interaction. Ecol. Evol. 1:132-144

165

Augustine DJ, McNaughton SJ (1998) Ungulate effects on the functional species composition of plant communities: Herbivore selectivity and plant tolerance. J. Wildl. Manage. 62:1165-1183 Bailey JK, Whitham TG (2002) Interactions among fire, aspen, and elk affect insect diversity: Reversal of a community response. Ecology 83:1701-1712 Barton P, Ikin K, Smith A, MacGregor C, Lindenmayer D (2014) Vegetation structure moderates the effect of fire on bird assemblages in a heterogeneous landscape. Landscape Ecol. 29:703-714 Barton PS, Manning AD, Gibb H, Wood JT, Lindenmayer DB, Cunningham SA (2011) Experimental reduction of native vertebrate grazing and addition of logs benefit beetle diversity at multiple scales. J. Appl. Ecol. 48:943-951 Bates D, Maechler M, Bolker B, Walker S (2014) lme4: Linear mixed-effects models using Eigen and S4, R package version 1.1-7, http://CRAN.R-project.org/package=lme4 Bell JR, Bohan DA, Shaw EM, Weyman GS (2005) Ballooning dispersal using silk: world fauna, phylogenies, genetics and models. Bull. Entomol. Res. 95:69-114 Bell JR, Wheater CP, Cullen WR (2001) The implications of grassland and heathland management for the conservation of spider communities: a review. J. Zool. 255:377-387 Bond WJ, Keeley JE (2005) Fire as a global ‘herbivore’: The ecology and evolution of flammable ecosystems. Trends Ecol. Evol. 20:387-394 Borcard D, Legendre P, Drapeau P (1992) Partialling out the spatial component of ecological variation. Ecology 73:1045-1055 Brennan KEC, Ashby L, Majer JD, Moir ML, Koch JM (2006) Simplifying assessment of forest management practices for invertebrates: How effective are higher taxon and habitat surrogates for spiders following prescribed burning? For. Ecol. Manage. 231:138-154 Buddle CM, Spence JR, Langor DW (2000) Succession of boreal forest spider assemblages following wildfire and harvesting. Ecography 23:424-436 Carter PE, Rypstra AL (1995) Top-down effects in soybean agroecosystems: Spider density affects herbivore damage. Oikos 72:433-439 Chmiel K, Herberstein ME, Elgar MA (2000) Web damage and feeding experience influence web site tenacity in the orb-web spider Argiope keyserlingi Karsch. Anim. Behav. 60:821-826 Clarke MF (2008) Catering for the needs of fauna in fire management: science or just wishful thinking? Wildl. Res. 35:385-394 Côté SD, Rooney TP, Tremblay JP, Dussault C, Waller DM (2004) Ecological impacts of deer overabundance. Annu. Rev. Ecol., Evol. Syst. 35:113-147 Crain CM, Kroeker K, Halpern BS (2008) Interactive and cumulative effects of multiple human stressors in marine systems. Ecol. Lett. 11:1304-1315

166

Danell K, Bergström R, Duncan P, Pastor J (2006) Large Herbivore Ecology, Ecosystem Dynamics and Conservation. Cambridge University Press, Cambridge Dexter N, Hudson M, James S, MacGregor C, Lindenmayer DB (2013) Unintended consequences of invasive predator control in an Australian forest: Overabundant wallabies and vegetation change. PloS ONE 8:e69087 Dexter N, Ramsey DL, MacGregor C, Lindenmayer D (2012) Predicting ecosystem wide impacts of wallaby management using a fuzzy cognitive map. Ecosystems 15:1363-1379 Didham RK, Tylianakis JM, Gemmell NJ, Rand TA, Ewers RM (2007) Interactive effects of habitat modification and species invasion on native species decline. Trends Ecol. Evol. 22:489-496 Foord SH, Dippenaar-Schoeman AS, Stam EM (2013) Surrogates of spider diversity, leveraging the conservation of a poorly known group in the Savanna Biome of South Africa. Biol. Conserv. 161:203-212 Foster CN, Barton PS, Lindenmayer DB (2014) Effects of large native herbivores on other animals. J. Appl. Ecol. 51:929–938 Gollan JR, Smith HM, Bulbert M, Donnelly AP, Wilkie L (2010) Using spider web types as a substitute for assessing web-building spider biodiversity and the success of habitat restoration. Biodivers. Conserv. 19:3141-3155 Halaj J, Ross DW, Moldenke AR (2000) Importance of habitat structure to the arthropod food- web in Douglas-fir canopies. Oikos 90:139-152 Halme P, Allen KA, Auniņš A et al. (2013) Challenges of ecological restoration: Lessons from forests in northern Europe. Biol. Conserv. 167:248-256 Harwood JD, Sunderland KD, Symondson WOC (2003) Web-location by linyphiid spiders: prey-specific aggregation and foraging strategies. J. Anim. Ecol. 72:745-756 Hobbs RJ, Huenneke LF (1992) Disturbance, diversity, and invasion: Implications for conservation. Conserv. Biol. 6:324-337 Howland B, Stojanovic D, Gordon IJ, Manning AD, Fletcher D, Lindenmayer DB (2014) Eaten out of house and home: Impacts of grazing on ground-dwelling reptiles in Australian grasslands and grassy woodlands. PLoS ONE 9:e105966 Janetos AC (1982) Foraging tactics of two guilds of web-spinning spiders. Behav. Ecol. Sociobiol. 10:19-27 Jonas JL, Joern A (2007) Grasshopper (Orthoptera : Acrididae) communities respond to fire, bison grazing and weather in North American tallgrass prairie: A long-term study. Oecologia 153:699-711 Kerns BK, Buonopane M, Thies WG, Niwa C (2011) Reintroducing fire into a ponderosa pine forest with and without cattle grazing: understory vegetation response. Ecosphere 2:art59

167

Kimuyu DM, Sensenig RL, Riginos C, Veblen KE, Young TP (2014) Native and domestic browsers and grazers reduce fuels, fire temperatures, and acacia ant mortality in an African savanna. Ecol. Appl. 24:741-749 Klop E, van Goethem J, de Iongh HH (2007) Resource selection by grazing herbivores on post- fire regrowth in a West African woodland savanna. Wildl. Res. 34:77-83 Langellotto G, Denno R (2004) Responses of invertebrate natural enemies to complex- structured habitats: a meta-analytical synthesis. Oecologia 139:1-10 Langlands PR, Brennan KEC, Framenau VW, Main BY (2011) Predicting the post-fire responses of animal assemblages: Testing a trait-based approach using spiders. J. Anim. Ecol. 80:558-568 Lindenmayer D, MacGregor C, Dexter N, Fortescue M, Beaton E (2014) Booderee National Park, The Jewel of Jervis Bay. CSIRO Publishing, Collingwood, Australia Lubin Y, Ellner S, Kotzman M (1993) Web relocation and habitat selection in desert widow spider. Ecology 74:1915-1928 Mantyka-pringle CS, Martin TG, Rhodes JR (2012) Interactions between climate and habitat loss effects on biodiversity: a systematic review and meta-analysis. Global Change Biol. 18:1239-1252 Matlack RS, Kaufman DW, Kaufman GA (2001) Influence of grazing by bison and cattle on deer mice in burned tallgrass prairie. Am. Midl. Nat. 146:361-368 Miyashita T, Takada M, Shimazaki A (2004) Indirect effects of herbivory by deer reduce abundance and species richness of web spiders. Ecoscience 11:74-79 Neumann FG, Tolhurst K (1991) Effects of fuel reduction burning on epigeal arthropods and earthworms in dry sclerophyll eucalypt forest of west-central Victoria. Aust. J. Ecol. 16:315-330 Nyffeler M (1999) Prey selection of spiders in the field. J. Arachnol. 27:317-324 Ohgushi T (2005) Indirect interaction webs: Herbivore-induced effects through trait change in plants. Annu. Rev. Ecol., Evol. Syst. 36:81-105 Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O'Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H (2013) vegan: Community Ecology Package, R package version 2.0-10. URL http://CRAN.R-project.org/package=vegan Pedersen S, Andreassen HP, Keith DA, Skarpe C, Dickman CR, Gordon IJ, Crowther MS, McArthur C (2014) Relationships between native small mammals and native and introduced large herbivores. Austral Ecol. 39:236-243 Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team (2014) nlme: Linear and nonlinear mixed effects models, R package version 3.1-118, http://CRAN.R- project.org/package=nlme

168

Podgaiski LR, Joner F, Lavorel S, Moretti M, Ibanez S, Mendonça MdS, Jr., Pillar VD (2013) Spider trait assembly patterns and resilience under fire-induced vegetation change in South Brazilian grasslands. PLoS ONE 8:e60207 R Core Team (2013) A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/ Riechert SE, Lockley T (1984) Spiders as biological control agents. Annu. Rev. Entomol. 29:299-320 Royo AA, Collins R, Adams MB, Kirschbaum C, Carson WP (2010) Pervasive interactions between ungulate browsers and disturbance regimes promote temperate forest herbaceous diversity. Ecology 91:93-105 Ruffell J, Didham RK, Barrett P, Gorman N, Pike R, Hickey-Elliott A, Sievwright K, Armstrong DP (2014) Discriminating the drivers of edge effects on nest predation: Forest edges reduce capture rates of ship rats Rattus rattus, a globally invasive nest predator, by altering vegetation structure. PLoS ONE 9:e113098 Rypstra A (1983) The importance of food and space in limiting web-spider densities; a test using field enclosures. Oecologia 59:312-316 Schmitz OJ (2008) Effects of predator hunting mode on grassland ecosystem function. Science 319:952-954 Shipley B (2002) Cause and correlation in biology: a user's guide to path analysis, structural equations and causal inference, Second edition. Cambridge University Press, Cambridge, U.K. Shipley B (2009) Confirmatory path analysis in a generalized multilevel context. Ecology 90:363-368 Sinclair ARE, Byrom AE (2006) Understanding ecosystem dynamics for conservation of biota. J. Anim. Ecol. 75:64-79 Smit R, Bokdam J, den Ouden J, Olff H, Schot-Opschoor H, Schrijvers M (2001) Effects of introduction and exclusion of large herbivores on small rodent communities. Plant Ecol. 155:119-127 Smith H (2008) BugWise Web2Spider supplement, Available at http://publications.australianmuseum.net.au/pdf/1561_complete.pdf. Accessed 15 October 2012 Specht RL, Morgan DG (1981) The balance between the foliage projective covers of overstorey and understorey strata in Australian vegetation. Aust. J. Ecol. 6:193-202 Thonicke K, Venevsky S, Sitch S, Cramer W (2001) The role of fire disturbance for global vegetation dynamics: Coupling fire into a Dynamic Global Vegetation Model. Global Ecol. Biogeogr. 10:661-677 Tylianakis JM, Didham RK, Bascompte J, Wardle DA (2008) Global change and species interactions in terrestrial ecosystems. Ecol. Lett. 11:1351-1363 169

Underwood EC, Christian CE (2009) Consequences of prescribed fire and grazing on grassland ant communities. Environ. Entomol. 38:325-332 Warui CM, Villet MR, Young TP, Jocque R (2005) Influence of grazing by large mammals on the spider community of a Kenyan savanna biome. J. Arachnol. 33:269-279 Williams J, Whelan R, Gill A (1994) Fire and environmental heterogeneity in southern temperate forest ecosystems: Implications for management. Aust. J. Bot. 42:125-137 Wisdom MJ, Vavra M, Boyd JM, Hemstrom MA, Ager AA, Johnson BK (2006) Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildl. Soc. Bull. 34:283-292 York A (1999) Long-term effects of frequent low-intensity burning on the abundance of litter- dwelling invertebrates in coastal blackbutt forests of southeastern Australia. J. Insect Conserv. 3:191-199 Zschokke S, Hénaut Y, Benjamin SP, García-Ballinas JA (2006) Prey-capture strategies in sympatric web-building spiders. Canadian Journal of Zoology 84:964-973 Zuur AF, Ieno EN, Walker NJ, Saveliev AA, Smith GM (2009) Mixed effects models and extensions in ecology with R. Springer, New York, USA

170

Supplementary Material

Online Resource 1 – Study location and experimental design

Site Description

This study was conducted within the Eucalyptus pilularis forest of BNP (Fig. OR1). We selected this vegetation type as it is the most widespread vegetation type in BNP and it occurs widely through coastal areas of Eastern Australia. In BNP this vegetation is dominated by an overstorey of Eucalyptus pilularis, Corymbia gummifera and E. botryoides (Taws 1998;

Lindenmayer et al. 2008). The midstorey is comprised predominantly of Banksia serrata,

Acacia longifolia and Monotoca elliptica, with Synoum glandulosum and Ziera arborescens becoming more prevalent in moist areas. The understorey is dominated by Lomandra longifolia and Pterideum esculentum.

Figure OR1. Location of study site within the Eucalyptus pilularis forest type (grey) within Booderee National Park (bold line). Map shows the arrangement of the 24 study sites into four blocks with one replicate of each of the six treatment combinations in each block. Abbreviations for levels of herbivory and burning treatments: E = exclosure, P = partial, O = open, B = burnt, U = unburnt.

171

Site selection

Sites were selected using a three step process. (1) We selected the four experimental blocks by using vegetation maps of BNP to identify areas of continuous Eucalyptus pilularis forest which had not been recently burnt (>8 years since fire) and which were large enough to contain six experimental plots (approx. 12ha), and then maximising the distance between selected blocks

(minimum of 2 km between blocks, see Fig OR1). (2) We selected six sites within each block which both minimised the total distance between plots (to reduce variation between plots) and maintained a minimum distance of 150 m between plots (to reduce the chances of interference between treatments, See Fig. OR2). (3) We then randomly allocated the six treatments to the six sites within the following logistical constraints: burnt sites could not be > 50 m from an access trail (due to operational requirements) and fenced sites could not be adjacent to an existing long-term monitoring sites (to avoid interfering with other studies). Pre-treatment vegetation surveys indicated there were no significant differences in vegetation structure between treatments prior to commencement of the study.

Figure OR2. Schematic diagram of blocked design showing types and site dimensions of the six treatment combinations (not to scale). Exact spatial arrangement varies between blocks (see

Fig. OR1).

172

References

Lindenmayer DB et al. (2008) Contrasting mammal responses to vegetation type and fire. Wildl. Res. 35:395-408 Taws N (1998) Fire and vegetation management in Jervis Bay Territory, Report. Environment Australia, Canberra

173

Online Resource 2 – Reanalysis of path model including all vegetation variables

In our main path analysis, we included only three of our nine vegetation variables in the hypothesised causal diagram (cover of shrubs, sedges and ferns). This was because many of the nine vegetation variables were highly correlated and potentially reciprocally related. However,

Generalised Multilevel Path Analysis is unable to handle reciprocal relationships between variables (Shipley 2009). By including only three vegetation variables in the path analysis, we were omitting a large amount of the measured variation in vegetation structure. Therefore, any significant effects of vegetation structure on spiders that were not captured by the three selected variables, would be attributed to “direct” rather than vegetation-mediated effects of disturbance on spiders. To test whether the significant main effects of disturbance that were not explained by vegetation in our main analysis were in fact due to additional variation in vegetation structure, we conducted a ‘sensitivity analysis’ including all of the nine vegetation variables in the analysis.

This sensitivity analysis involved re-running the path analysis to include all of the vegetation variables we measured. However, to avoid the issue of reciprocity between variables, we first had to generate a set of uncorrelated vegetation variables (which we did using principal components analysis), and then use these as the vegetation structural variables in the path analysis. We did not use this approach in the main analysis, as principal components are less correlated than would be expected by chance. Including principal components in the path analysis, could therefore compensate for unacceptably high levels of correlation among other independence claims, and reduce our chances of rejecting an incorrectly specified causal model.

As traditional PCA does not allow for hierarchically structured data, we used the

“phyl.pca” function in the “phytools” package in R to conduct a hierarchically structured phylogenetic PCA of the vegetation data, where plots were nested within sites, within blocks

(see Ruffell et al. 2014 for a more detailed description of this method). We conducted a separate

PCA for each year of data, and based the PCA on the correlation (rather than the covariance) matrix as vegetation variables were measured on different scales. We used the first seven of the 174

principal components in the path analysis, which accounted for 96% of the variation in vegetation variables in each year (Table OR1).

Table OR1. Details of the seven principal components of the vegetation variables used in the path analysis.

PC1 PC2 PC3 PC4 PC5 PC6 PC7 2012 Variation explained (cum. %) 33.9% 49.5% 62.3% 74.0% 83.3% 91.2% 96.0% Variable Loadings Total understory cover -0.95 -0.06 -0.05 0.12 -0.06 -0.03 -0.18 Stem density -0.56 0.15 0.14 -0.35 -0.03 0.70 0.14 Understory height -0.77 -0.31 0.28 0.08 0.13 -0.16 -0.05 Litter depth -0.59 0.50 -0.19 0.09 0.31 -0.24 0.45 Fern cover -0.51 -0.58 0.44 -0.29 0.01 -0.19 0.11 Grass cover 0.10 -0.53 -0.18 0.55 0.53 0.30 0.03 Sedge cover -0.54 0.41 0.20 0.61 -0.29 0.07 -0.14 Shrub cover -0.55 0.06 -0.65 -0.35 0.22 -0.06 -0.29 Herb cover -0.21 -0.50 -0.57 0.13 -0.55 0.03 0.24

2013 Variation explained (cum. %) 28.2% 44.8% 58.0% 70.3% 80.7% 89.3% 96.0% Variable Loadings Total understory cover 0.91 -0.05 0.21 -0.09 -0.14 -0.22 -0.03 Stem density 0.56 -0.07 0.05 0.15 -0.09 0.80 0.11 Understory height 0.83 0.10 -0.19 -0.11 0.00 -0.08 -0.12 Litter depth 0.24 0.68 -0.16 0.30 0.23 -0.13 0.54 Fern cover 0.58 -0.58 0.42 -0.11 0.21 -0.18 0.24 Grass cover 0.01 -0.35 -0.23 0.51 -0.71 -0.17 0.16 Sedge cover 0.25 0.73 0.45 0.06 -0.33 -0.04 -0.27 Shrub cover 0.42 0.06 -0.79 -0.36 -0.05 -0.03 -0.09 Herb cover 0.27 -0.12 -0.15 0.76 0.43 -0.07 -0.35

The full path model including all seven principal components had a high level of support for web density in 2012 (χ2 = 54.9, df = 74, P = 0.95), and in 2013 (χ2 = 50.88, df = 74, P = 0.98), and also had a high level of support when claims among principal components were excluded

(2012: χ2 = 29.6, df = 32, P = 0.59; 2013: χ2 = 30.8, df = 32, P = 0.53). Similarly, models for web-type richness were well supported in both years (2012: χ2 = 54.9, df = 74, P = 0.95; 2013:

χ2 = 46.5, df = 74, P = 0.99), and when independence claims among principal components were excluded (2012: χ2 = 29.6, df = 32, P = 0.58; 2013: χ2 = 26.4, df = 32, P = 0.75).

175

The models including the seven principal components gave qualitatively similar results to our main path analysis. The negative effects of fire on web density were largely mediated by vegetation, but also occurred via the “direct” pathway, while negative effects of herbivores on web density occurred mostly via the “direct” pathway (Fig. OR3a,b). The negative effects of fire and herbivores on web type richness were not mediated by changes in vegetation structure

(Fig. OR3c,d). The strength of “direct” pathways in the analysis including all vegetation variables was not reduced compared with our main analysis. This indicates that our main analysis captured sufficient variation in vegetation structure to provide a reliable estimate of the strength of vegetation-mediated versus other effects of disturbance.

References

Ruffell J et al. (2014) Discriminating the drivers of edge effects on nest predation: Forest edges reduce capture rates of ship rats Rattus rattus, a globally invasive nest predator, by altering vegetation structure. PLoS ONE 9:e113098 Shipley B (2009) Confirmatory path analysis in a generalized multilevel context. Ecology 90:363-368

176

Figure OR3. Effects of fire, herbivores and their interaction on web-building spider density (a,b) and richness (c,d), and the extent that effects were mediated by vegetation. Arrows represent causal paths between the experimental treatments (square boxes) and plant/animal responses (rounded boxes). Path coefficients are unstandardized partial regression coefficients. Grey arrows are negative effects, black arrows positive effects, and blue connectors are interactions. Dashed arrows indicate non-significant relationships, significance levels of path coefficients: ^ P ≤ 0.1, * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001. Herbivore activity was ln transformed when used as a predictor.

177

Figure OR3 cont.

178

Online Resource 3 – Functional group and associated taxa of the recorded web types

Table OR2. Functional group of web types recorded in surveys, and the taxa constructing each type in the region of the study (adapted from Smith 2008).

Sites Individuals Functional Web present recorded Brief description Taxa commonly constructing web group number 2012 2013 2012 2013

Orb webs

Araneidae: Argiope keyserlingi & some W1§ Decorated - silk X 0 2 0 2 other Argiope spp. Decorated - silk vertical Araneidae: Araneus eburnus and Araneus W2 4 1 4 1 line bradleyi

W3§ Decorated, sloping Uloboridae: Zosis & other genera 0 1 0 2

Decorated - line of prey W4§ Araneidae: Cyclosa 0 1 0 1 and eggs Araneidae: Argiope keyserlingi (juv.) & W5 Decorated with doily 3 0 5 0 some other Argiope spp. Araneidae: Phonognatha graeffei & other W9 Leaf at web centre 15 23 43 314 Phonognatha Leaf away from web W10 Araneidae: Araneus dimidiatus 8 8 15 11 centre

W12§ Sloping, missing sector Araneidae: Arachnura higginsi 1 2 1 2

Pie slice, vertical, retreat Araneidae: Phonognatha graeffei (juv.) & W13 22 24 180 303 at hub other Phonognatha Pie slice, signal line to W14 Araneidae: Metepeira 22 17 221 155 retreat

W15 Offset hub, horizontal Uloboridae: Philoponella & other genera 5 1 9 1

Nephilidae: Nephila plumipes and Nephila W16§ Plain, many radial lines 1 0 1 0 edulis (juveniles)

W17 Plain, web hub filled Araneidae: juveniles of many genera 10 18 25 47

Araneidae: Eriophora & other genera W18 Plain, hole at hub 4 10 6 13 Tetragnathidae: Tetragnatha & Leucauge

W19 Horizontal, hole at hub Tetragnathidae: Tetragnatha & Leucauge 10 7 23 12

Horizontal, v. small, W20§ Theridiosomatidae 1 0 1 0 cone

Lace webs

On bark, funnel-like Desidae: Badumna longinqua, Badumna W21 1 3 1 3 entrance insignis Radiates from hole in W22 Desidae: Paramatachia 11 17 19 51 twig Fans out from silk W23 Desidae: Badumna 23 4 169 6 retreat

W24 Web around nest Desidae: Phryganoporus candidus 7 7 10 8

179

Sites Individuals Functional Web present recorded Brief description Taxa commonly constructing web group number 2012 2013 2012 2013

Sheet webs

"Witch's hat", irregular Theridiidae: Achaearanea mundula & W25 20 22 492 475 weave related spp. "Witch's hat", regular W26 Araneidae: Cyrtophora 3 0 7 0 weave Small sheet in low W27 Linyphiidae 20 21 127 119 vegetation Platform, burrow top of W28 Stiphidiidae: Corasoides australis 14 7 41 27 sheet Stiphidiidae: Therlinya, Pillara, Couranga, Sloping sheet, retreat W29 Stiphidion & Jamberoo 7 13 13 17 below sheet Theridiidae: Steatoda

Tangle webs

Silk retreat, web Theridiidae: Achaearanea, Theridion & W30 20 24 83 169 between leaves others Silk retreat, lines to firm Theridiidae: Latrodectus, Achaearanea & W31§ 1 0 1 surface Steatoda Theridiidae: Achaearanea, Theridion & W32 No silk retreat 24 24 220 231 other genera § These rare web types were excluded from the CCA analysis.

References

Smith H (2008) BugWise Web2Spider supplement, Available at http://publications.australianmuseum.net.au/pdf/1561_complete.pdf. Accessed 15 October 2012

180

Online Resource 4 – Response of vegetation to disturbance treatments

Figure OR4. Vegetation response to burning and herbivory treatments (selection of ‘typical’ sites).

181

Paper V: Integrating theory into disturbance interaction experiments to better inform ecosystem management

Theoretical studies of ecological disturbance have highlighted the value of understanding interaction pathways for both predicting and managing the effects of co-occurring disturbances. In this paper, I synthesised the findings from my experiments (Papers II, III and IV), and critically evaluated related literature on fire-grazing interactions, to assess the ability of studies to test the key relationships and mechanisms that are important for ecological management. Based on this evaluation, I identified several ways that future studies can be improved to better inform the management of disturbance interactions.

Foster, C.N, Sato, C.F., Lindenmayer, D.B., and P.S. Barton (Accepted) Integrating theory into disturbance interaction experiments to better inform ecosystem management. Global Change Biology, doi:10.1111/gcb.13155.

183

Abstract

Managing multiple, interacting disturbances is a key challenge to biodiversity conservation, and one that will only increase as global change drivers continue to alter disturbance regimes.

Theoretical studies have highlighted the importance of a mechanistic understanding of stressor interactions for improving the prediction and management of interactive effects. However, many conservation studies are not designed or interpreted in the context of theory, and instead focus on case-specific management questions. This is a problem as it means that few studies test the relationships highlighted in theoretical models as being important for ecological management. We explore the extent of this problem among studies of interacting disturbances by reviewing recent experimental studies of the interaction between fire and grazing in terrestrial ecosystems. Interactions between fire and grazing can occur via a number of pathways; one disturbance can modify the other’s likelihood, intensity or spatial distribution, or one disturbance can alter the other’s impacts on individual organisms. The strength of such interactions will vary depending on disturbance attributes (e.g. size or intensity), and this variation is likely to be non-linear. We show that few experiments testing fire-grazing interactions are able to identify the mechanistic pathway driving an observed interaction, and most are unable to detect non-linear effects. We demonstrate how these limitations compromise the ability of experimental studies to effectively inform ecological management. We propose a series of adjustments to the design of disturbance interaction experiments that would enable tests of key theoretical pathways, and provide the deeper ecological understanding necessary for effective management. Such considerations are relevant to studies of a broad range of ecological interactions, and are critical to informing the management of disturbance regimes in the context of accelerating global change.

Keywords

Antagonisms, compound effects, disturbance ecology, fire, grazing, higher order effects, multiple stressors, terrestrial ecosystems, synergisms

184

Introduction

Disturbance is a major driver of change in ecosystems, and is central to both fundamental and applied ecology. As climate change and anthropogenic pressures are driving substantial changes to disturbance regimes such as fire, severe weather and biological invasions, understanding the factors affecting disturbances and their ecological impacts will be essential to the effective future management of terrestrial ecosystems (Dale et al., 2001; Turner, 2010). Recent advances in disturbance theory include improved understanding of the spatial variability of disturbance effects (Turner, 2010), and the potential for cross-scale effects (Peters et al., 2007), and legacy effects (Essl et al., 2015) to produce unanticipated changes in ecosystems. Another key advancement, and one with strong management applications, is the acknowledgement of the prevalence of disturbance interactions (Wisdom et al., 2006; Turner, 2010), their ability to produce ecological surprises (Lindenmayer et al., 2010; Doherty et al., 2015), and the range of mechanistic pathways by which these interactions can occur (Didham et al., 2007; Buma,

2015). An important question, however, is whether empirical studies have kept pace with theoretical advances, and the implications this has for informing effective ecological management.

As theoretical models are widely used in conservation decision making, advances in disturbance theory have the potential to guide substantial improvements in the management of disturbance regimes (Driscoll & Lindenmayer, 2012). For example, several recent papers have demonstrated how a disaggregated, mechanistic understanding of the effects of disturbance and land use change on ecosystems could lead to more effective, and potentially novel, solutions for ecological management (Didham et al., 2007; Fuhlendorf et al., 2009; Peters et al., 2011). For theory to effectively inform management, however, it needs to be paired with empirical studies that test its applicability to the local context (Driscoll & Lindenmayer, 2012). Yet, while there are many excellent examples of such studies (e.g. Mandle & Ticktin, 2012; Kimuyu et al.,

2014), the majority of empirical conservation studies are not well integrated with theory, and this potentially limits both theory development and management applications (Belovsky et al.,

2004; Fazey et al., 2005; Barot et al., 2015).

185

Here, we examine how well the experimental literature has integrated recent theoretical developments on disturbance interactions, and how this affects management implications. We focus on examples from studies of fire-grazing interactions as they are among the most commonly studied disturbance interactions in terrestrial ecosystems, and as fire and grazing are both commonly managed in terrestrial ecosystems, much research has focussed on how prescribed fire (e.g. burn frequency) and grazing regimes (e.g. stocking density) can best be managed for biodiversity conservation. We use a mini-review of recent fire-grazing interaction experiments to highlight the gaps between theoretical and experimental studies, and the effect this has on the ability of experimental studies to inform management decisions. We then discuss how relatively simple changes to the way interaction experiments are designed and interpreted will strengthen their links with the theoretical literature, and in doing so, will improve the ability of experimental studies to provide useful information for the management of multiple disturbances.

Recent advances in disturbance interaction theory

The study of disturbance interactions has evolved largely from work on multiple stressor effects in marine systems, which is based on the ‘additive’ model of interactions (Folt et al., 1999;

Crain et al., 2008). Under this model, interactive effects occur when the combined effects of two environmental stressors is greater than (synergism) or less than (antagonism) the sum of their individual effects (Folt et al., 1999). Although the additive model has formed the basis of countless experimental studies of the effects of multiple stressors (reviewed by Brook et al.,

2008; Crain et al., 2008; Darling & Côté, 2008; Jackson et al., 2015; Przeslawski et al., 2015), current theory recognises disturbance interactions as complex, spatially and temporally variable processes (Peters et al., 2004; Wisdom et al., 2006; Turner, 2010; Buma, 2015). This theory includes two recent advances which have the potential to dramatically improve our ability to predict and manage the outcomes of disturbance interactions; (1) characterisation of the different mechanistic pathways driving interactions, and (2) increasing acknowledgement of the likelihood of non-linear interactive effects.

186

Several recent syntheses have focussed on the mechanistic pathways by which multiple disturbances (or stressors) interact, and the value of identifying interaction pathways for designing effective and novel management interventions (Didham et al., 2007; Tylianakis et al.,

2008; Oliver & Morecroft, 2014; Boyd & Brown, 2015; Buma, 2015; Doherty et al., 2015). A variety of terms have been used to describe these different interaction pathways, but despite differences in the stressors discussed, most studies separate interactions into two broadly similar

“types” (Table 1). We adopt the terminology of Didham et al. (2007), who separate interactions into (1) interaction chain effects, where one disturbance affects the occurrence or magnitude of a second disturbance, and both disturbances have a direct effect on the response variable, and

(2) interaction modification effects, where the per-unit effect of one disturbance on the response variable depends on the environmental context of a second disturbance (Fig. 1). We use the definitions of Didham et al. (2007), as they are among the first presented in the literature, and while originally developed in the context of habitat modification and invasive species interactions, have been successfully adapted to characterise interactions between other stressors, both biotic and abiotic (e.g. Mandle & Ticktin, 2012; Oliver & Morecroft, 2014). Regardless of the terminology and definitions used, a focus on identifying the mechanistic pathways driving disturbance interactions has the potential to provide valuable insights for ecological management, and potentially lead to novel solutions for processes that may be difficult to manage directly (Didham et al., 2007; Peters et al., 2011; Boyd & Brown, 2015; Doherty et al.,

2015). For example, feral predators have substantial impacts on many native species, but broad- scale control is often unfeasible. However, as predation success is modified by habitat complexity, it may be possible to conserve native species by improving fire management to maintain complex habitats (i.e. managing the interaction modification) (Doherty et al., 2015).

The theoretical literature also acknowledges the likelihood that many interactions will have non-linear effects on ecosystems, potentially leading to alternate states (Peters et al., 2004;

Wisdom et al., 2006; Oliver & Morecroft, 2014; Kayler et al., 2015). These non-linear effects can occur both in chain (Didham et al., 2007) and modification interactions (Dunne, 2010).

Further, when chain and modification interactions both occur, their net effects may become

187

non-linear through cross-scale interactions (Peters et al., 2007). When the effects of disturbances and their interactions are non-linear, knowledge of key points, such as when a small change in disturbance intensity has a large effect on biodiversity (or vice versa), are potentially invaluable for informing targeted and effective management (Peters et al., 2004;

Huggett, 2005; Groffman et al., 2006; Didham et al., 2007). For example, the effect of fire on the susceptibility of forests to pine-beetle attacks is non-linear: low severity fire can increase the susceptibility of forests to pine beetle outbreaks (Kulakowski & Jarvis, 2013), while high- severity, stand-replacing fires can reduce forest susceptibility to the same pests (Kulakowski et al., 2003). Knowledge of these non-linear effects, and estimates of the threshold at which effects switch from negative to positive would allow managers to adjust fire management practices, and/or conduct target management of local beetle outbreaks, to minimise the risk of widespread pine-beetle infestations.

188

Table 1. Summary of the terms and definitions used to distinguish interaction “types” in recent reviews of ecological interactions. Types are sorted according to how they align with the definitions of interaction chains and interaction modifications provided by Didham et al. (2007).

Paper Context Interaction "types" and their definitions

Disturbances/ Ecosystem Interaction chain Interaction modification stressors

Didham et Broad range Invasive Interaction chain Interaction modification al. (2007) of both species and (numerically (functionally moderated): the per terrestrial and habitat mediated): Total impact capita impact of one driver is marine modification changes because of a dependent on the level of effect of a ecosystems numerically scaled second driver response (of the first driver) to a second driver variable

Oliver and Terrestrial Climate Interaction chain: one Interaction modification: the per Morecroft ecosystems change and driver capita effect of one driver changes (2014) land use changes the magnitude depending on the level of another change of another driver and driver both drivers have a direct effect on the response variable.

Ban et al. Coral reefs Multiple Stressor interactions: Synergistic effects: where two (2014) stressors, both where one stressor stressors interact to produce an effect biotic and directly influences on a response variable that is greater abiotic another than purely additive

Doherty et Terrestrial Invasive Numerical impact: Functional Compounding al. (2015) ecosystems predators and where a disturbance impact: where a impacts: where disturbance increases the abundance disturbance the effect of of the invasive species changes the per- disturbance on a capita impact of population an invasive changes its ability species to cope with invasion

Boyd and Marine Environmental Physiochemical Physiological Food web Brown ecosystems drivers, interactions: interactions: interactions: (2015) focussing on interactions that occur where an where interactive warming and directly between drivers, organism’s effects on ocean outside of organisms physiological organisms occur acidification response to one indirectly, via driver changes changes in food with the level of webs or another driver ecosystem functioning

Buma Broad range Wide range of Linked interactions: Compound disturbance: Recovery (2015) of both disturbances one disturbance from a second disturbance is more or terrestrial and increases or decreases less likely, or at altered speed, due to marine the likelihood, spatial the initial disturbance ecosystems extent, intensity or severity of a second disturbance

Jackson Marine, Multiple Interactions among Combined impacts: where the (2015) terrestrial and invasive invaders: where the presence of one invader influences freshwater animals presence of one invader another's impact on the ecosystem ecosystems influences the performance of a second invader

189

Figure 1. The two major pathways by which fire and grazing can interact to affect biota (adapted from Didham et al. 2007). (a) An interaction chain effect, where fire affects the spatial distribution or intensity of grazing, which in turn has a direct effect on the response variable (e.g. plant or animal abundance), but the per-unit effect of grazing on the response variable remains unchanged. A chain interaction also can occur in the opposite direction, with grazing affecting the spatial location, spread, and intensity of fire. (b) An interaction modification, where the per-unit effect of grazing on the response variable is modified by the occurrence of fire. This interaction can also occur where grazing modifies the effect of fire on the response variable. The response curves shown are only one example of many possible relationships. The slope and direction of responses, and the direction and strength of interactions will vary depending on the ecosystem, disturbances and organisms (response variables) studied.

Where are the gaps between theory and experimental studies?

Many experimental studies of disturbance interactions aim to inform management decisions, citing conservation challenges as motivation for the study, or providing management recommendations based on their findings. However, few studies focussing on conservation problems are well integrated with theory (Fazey et al., 2005; Barot et al., 2015), despite the acknowledged value of using a strong conceptual framework to guide empirical research

190

(Driscoll & Lindenmayer, 2012; Barot et al., 2015), and the popularity of conceptual models for guiding management decisions (Williams, 2011). To assess the extent of this problem among studies of disturbance interactions, we focussed on the two recent conceptual advancements highlighted previously: (1) the mechanisms driving interactions and (2) non-linear interactive effects on ecosystems or biota, and quantified the extent to which these concepts have been addressed in recent studies of fire-grazing interactions. To do this, we reviewed 50 recent papers that focussed on fire-grazing interactions and were published between 2011 and 2015

(Supporting Information, Appendix S1). This mini-review was conducted to obtain a representative sample of recent studies of fire-grazing interactions (most of which were experimental) which we could use to answer specific questions about the current literature, and is not intended as a comprehensive literature review. The 2011 cut-off was selected as a number of key conceptual studies were published in the years 2007-2008 (e.g. Didham et al., 2007;

Peters et al., 2007; Brook et al., 2008; Crain et al., 2008; Darling & Côté, 2008), and the 2011 cut-off allowed some lag time for these ideas to become integrated into experimental studies.

Mechanisms driving interactions

Do experimental studies identify interaction pathways?

Fire and grazing can interact via a number of different pathways. For example, fire can affect the intensity and spatial distribution of grazing as many herbivores preferentially graze in burnt areas (a chain interaction often referred to as fire-driven grazing or pyric herbivory) (e.g. Allred et al., 2011). Similarly, grazing can affect the location and intensity of fires by modifying the distribution of fine fuels (also a chain interaction) (e.g. Kimuyu et al., 2014). At a more local scale, fire and grazing, by altering the composition, traits or condition of biological communities or individual organisms, can alter each other’s effects on those organisms (an interaction modification) (e.g. Eby et al., 2014). However, few of the studies we reviewed identified the mechanistic pathways driving the fire-grazing interaction, despite 80% of these papers citing management applications as a motivation for the study. Of the 50 studies reviewed, only one quantified the relative contributions of both chain and modification effects

191

in the interaction (Mandle & Ticktin, 2012), and 30 did not quantify any pathway, reporting only net effects of the interaction (Fig. 2). Of these 30 studies, five included description of more than one potential pathway driving the observed effects, 15 described only one potential pathway, and 10 did not provide any ecological definition of how the “interaction” might be occurring (Fig. 2).

Overall, 30% of the fire-grazing studies we reviewed discussed only one interaction type, but did not provide evidence that this interaction type was driving the observed effects.

For example, 13 of the 22 studies that defined only chain interactions (mostly fire-driven grazing) actually tested for net interactive effects, and did not quantify the extent to which fire affected herbivory (Fig. 2). As it has been demonstrated that burning does not always affect herbivore site selection (McGranahan et al., 2012), the interaction detected by such studies may be caused by a chain, or a modification, or both. By defining only one interaction type, but testing for net effects, these studies make the implicit assumption that the defined interaction is driving the fire-grazing interaction, which may or may not be true.

192

Figure 2. A comparison of the interaction pathways described or defined by studies of fire- grazing interactions, versus the pathways actually tested in the study (chain = interaction chain, mod. = interaction modification, net = tests for interactions that do not isolate interaction pathways, and hence may result from chain, or modification effects or a combination of the two). Numbers are based on a review of 50 fire-grazing interaction studies published between 2011 and 2015. * Indicates combinations where the definition used was appropriate for the interactions tested (n = 13).

How does not identifying the interaction pathway affect management applications?

Studies which make untested assumptions about interaction pathways could potentially be harmful, rather than useful for land management. We illustrate this with an example, based on a recent study by Foster et al. (2015), who found that fire and herbivory by native macropods interacted via both an interaction chain, and an interaction modification, to reduce the diversity of forest understory vegetation. If the authors had measured the net effect of the interaction, and based their management recommendations on an assumption about the interaction pathway, then these recommendations would be likely to lead to little improvement, or even deterioration of biodiversity (Fig. 3). For example, if the authors had assumed that the interaction occurred through an interaction chain, they may have recommended changing fire management to prevent herbivores concentrating in burnt areas (e.g. by increasing the area burnt). However, due to the presence of an interaction modification, reducing grazing pressure in burnt sites by 193

increasing burn size is unlikely to mitigate all interactive effects, and may actually increase negative effects on plants by increasing the proportion of the population that is exposed to the interactive effects (Fig. 3).

Similarly, the converse assumption (that the interaction occurred solely through an interaction modification) also could lead to ineffective management recommendations. If an interaction modification was assumed, the authors may have recommended control measures to reduce the abundance of macropod herbivores. However, due to the presence of an interaction chain, where herbivores were attracted to burnt areas, general population control may do little to reduce grazing pressure in burnt areas, and therefore not mitigate negative interactive effects

(Fig. 3). As population control of native herbivores is both resource-intensive and socially unpopular (Nugent et al., 2011), a control program that does not achieve the desired benefits for biodiversity is not only a waste of resources, but could compromise the ability of managers to implement similar programs in the future. This example demonstrates how failure to integrate theoretical concepts on interaction pathways into the design and interpretation of experimental studies could limit their ability to effectively inform the management of multiple disturbances.

194

Figure 3. Schematic diagram illustrating how incorrect assumptions about interaction pathways can lead to ineffective, or even damaging, management interventions. In the presence of an interaction chain herbivores congregate in burnt areas (grey shading), while in the presence of an interaction modification the per-herbivore impact is greater in burnt than unburnt areas (larger size of herbivore indicates larger per-herbivore impact). Managing for one interaction (e.g. an interaction chain), when the observed effects are predominantly driven by a different interaction pathway (e.g. an interaction modification), can result in unexpected outcomes. Note that the ‘outcomes’ in this example are based on the study of Foster et al. (2015), where fire and grazing interacted to negatively affect plant diversity, and so the aim of management interventions would be to maintain plant diversity by reducing interactive effects.

Non-linear interactive effects on ecosystems and organisms

How well do experimental studies test for nonlinear effects?

Among the 50 papers we reviewed, there were two common study designs used to investigate fire-grazing interactions; patch burning experiments (15 papers) and factorial experiments (27 papers). Few studies (12%) of either type were able to detect nonlinear effects. Most patch- burning studies compared “homogenous” grazed areas with “heterogeneous” patch burnt and grazed areas, under constant grazing intensity and burn size. Only two of the 15 studies using a

195

patch-burning design varied any characteristics of the burns (e.g. size of the burnt area), and none tested the how interaction strength was affected by grazer density or species. While not testing for non-linear interactions, a further three studies did test the consistency of interactive effects across sites with different management histories.

All of the studies that identified an interaction modification (nine studies), and many studies which tested net effects (18 studies), used a factorial study design, combining different levels of fire (most often burnt/unburnt) and grazing (most often grazed/un-grazed) at the plot level. To detect non-linear interactive effects, factorial studies must use more than two treatment levels for at least one disturbance (i.e. include treatment levels in addition to grazed/un-grazed and burnt/un-burnt, or measure natural variation within these levels), and ideally include that factor as a continuous variable in analyses. However, of the 27 factorial studies we reviewed, only nine used more than two levels for either fire or grazing, and only three of these used either factor as a continuous variable in analyses. Thus, only three of 27 factorial studies and two of 15 patch-burning studies were of a design that could detect non- linear interactive effects.

How does not identifying non-linear effects affect management applications?

The management implications of failing to detect non-linear effects differ slightly between chain and modification interactions, but most relate to the ability to extrapolate (or interpolate) results to untested situations or locations. Studies using patch-burn designs are testing for the ecological effects of a chain interaction (“pyric herbivory”), where herbivores are attracted to the fresh growth after fire, resulting in high grazing intensity in burnt patches (Fuhlendorf et al.,

2009). However, although it is acknowledged that the extent to which fire drives grazing patterns may vary with the intensity, duration and timing of grazing, as well as with the size, shape, intensity and timing of fire (Fuhlendorf et al., 2009; Sensenig et al., 2010; Allred et al.,

2011) (Fig. 4), most patch-burning studies did not test whether any of these factors affected the strength of the fire-grazing interaction (but see McGranahan et al., 2013; Hovick et al., 2014), and the consistency of such relationships remains poorly understood. Most patch-burning

196

studies are therefore able to detect the presence of an interaction at the levels of burning and grazing selected for that experiment, but cannot reliably be used for inference beyond this specific combination of conditions. A manager who is faced with different conditions from those tested in a study is therefore provided with little useful information about whether changing stocking rates, harvest quotas or prescribed burns will improve biodiversity outcomes.

By comparison, studies that tested effects over a gradient of fire sizes, or a range of stocking densities, would be able to give insights into the probable outcomes of changing management practices, as well as improve our fundamental understanding of the fire-grazing interaction.

Figure 4. Generalised relationships between burn size, and the extent to which herbivores select for the burnt area, demonstrating how chain interactions may vary with disturbance properties such as spatial scale, and between ecosystems (Fuhlendorf et al., 2009). Selection for burnt areas also varies between herbivore species (Sensenig et al., 2010), and under certain conditions, herbivores may select against burnt areas (e.g. if dense post-fire regrowth hinders herbivore movements, Wan et al., 2014).

Studies of interaction modifications are also often limited in their ability to inform changes to management practices. This is because most studies are based on factorial experiments using binary treatments (i.e. 2 × 2 factorial designs). When the effects of disturbances, and their interactions, are non-linear, the key points of interest from a management perspective are the points at which changes in one disturbance will have very large

(e.g. points of inflection), or very small (e.g. beyond thresholds), effects on the variable of interest (Groffman et al., 2006)(Fig. 5a). Knowledge of the existence of these points, and their

197

approximate values, would allow managers to identify: when management is most likely, or least likely to be successful; what level of management is required to achieve an outcome; or how management should differ in the presence of other disturbances (Groffman et al., 2006).

However, factorial studies employing binary contrasts cannot detect these points (Fig. 5), and cannot be generalised beyond the levels chosen for study (Inouye, 2001), and hence are of limited use for predicting the outcome of changes to management.

Factorial experiments employing binary comparisons are also problematic because whether they detect synergisms or antagonisms can be entirely dependent on the levels chosen for study (Fig. 5)(Dunne, 2010), but the management implications of synergisms and antagonisms are very different. Synergistic interactions can accelerate biodiversity decline, but synergisms may be more likely to respond to management, as managing one disturbance can mitigate the effects of both (Brown et al., 2013; Piggott et al., 2015). By contrast, mitigating the effects of antagonistic interactions usually requires both disturbances to be managed concurrently (Didham et al., 2007; Brown et al., 2013). Therefore, if the results of factorial studies are used to inform management, but the underlying relationships are nonlinear, unintended outcomes are likely to result.

As an example, Figure 5a presents a hypothetical non-linear interaction between fire and herbivory. In this example, a factorial study would detect; (1) a synergism if the herbivore densities compared were low and moderate; (2) no interaction if herbivore densities were low and high; and (3) an antagonism if herbivore densities were moderate and high (Fig. 5). If a factorial study compared low and high herbivore densities, the authors would likely conclude that fire and grazing were not interacting, and that reducing grazing intensity is likely to be the most effective action to increase the response variable. As can be seen in Fig. 5a, even if herbivore density were halved in this situation, there would be little change in the response variable in burnt environments, resulting in a waste of limited conservation resources. As field experiments tend to employ extreme treatments, with large effect sizes (Foster et al., 2014), and synergisms are more likely to be detected when individual effect sizes are small (Folt et al.,

1999; Piggott et al., 2015), factorial experiments may also under-estimate the occurrence of 198

synergisms in natural ecosystems. Clearly, single-level factorial studies can miss a large (and potentially important) part of the picture if used to study non-linear interactions. Therefore, researchers who aim to inform more efficient and effective management of terrestrial ecosystems through disturbance interaction studies should consider whether non-linear effects are likely, and if so, explore alternatives to factorial designs which include a gradient of biologically and management-relevant treatment levels.

Figure 5. When interactive effects are non-linear, differences in interpretation can occur when examining disturbances as continuous or factorial treatments. (a) An hypothetical, non- linear, interaction modification between fire and herbivory, showing the key points of management interest (turning points, shifts in marginal rates, and limits; open circles) and how these vary across different levels of fire and herbivore density (low density = L, moderate density = M, high density = H). (b) The likely results of factorial studies in the same study system, showing how the experimental results will vary depending on the treatment levels chosen for comparison (L v M => synergism, L v H => additive effects, M v H => antagonism).

Bridging the gap to improve management applications

There is a lack of integration of theory into the design and interpretation of disturbance interaction experiments. We have demonstrated this by reviewing recent studies of fire-grazing interactions. We have also shown that this lack of integration means that interaction experiments, despite often being motivated by management problems, fall short of their potential in effectively informing the management of multiple disturbances. Land managers, faced with complex systems, competing demands, and a mandate for transparent decision

199

making, often rely on system and management models to inform management practices

(Underwood, 1995; Williams, 2011). A core role of experimental studies should therefore be to improve our understanding of the relationships on which such models are based (Underwood,

1995). As we have argued in this paper, to do this requires experimental studies of disturbance interactions to more effectively, and broadly, engage with the theoretical literature. Closing the gap between the theoretical and experimental literature on disturbance interactions is not straightforward – interactions are complex, multiscale processes which can be difficult to manipulate experimentally. However, there are some adjustments that could be made to the design and interpretation of disturbance interaction studies which would both improve links with theory, and generate more useful information for the management of multiple disturbances.

While we have focussed on studies of terrestrial disturbance interactions in this paper, similar experimental designs are used to study many other types of interactions, in a range of ecosystems. The problems and solutions presented below will therefore be relevant to the study of many kinds of ecological interactions.

Study design

1. Focus on testing the key relationships and interaction pathways identified in theoretical

models as important for management. Most studies report the net effects of an

experimentally manipulated interaction (Fig. 2), but it is difficult to generalise the

results of such studies, and they often do not test the key relationships that make up

management models. It is not always feasible, or desirable, to test all interaction

pathways in a single study. However, future studies of disturbance interactions should

explicitly identify the interaction pathway(s) being tested, and ensure that the study is

conducted at an appropriate spatial and temporal scale. Such experimental tests of

theoretically important relationships and pathways would allow theoretical models to be

verified, which could then be used to inform decisions at management-relevant scales

(Inouye, 2001; Belovsky et al., 2004; Cottingham et al., 2005; Denny & Benedetti-

Cecchi, 2012).

200

2. Quantify variation in disturbances/stressors, as well as their ecological effects. Few

experimental studies measure or analyse disturbances as continuous variables.

However, environmental variation, and/or chain interactions mean that even if

disturbance treatments are experimentally imposed, disturbance intensity/extent is

likely to vary within treatments (i.e. between replicates). If measured, this variation can

be used to a researchers’ advantage, as by pairing this quantitative measure of

disturbance directly with responses, researchers will be able to separately quantify chain

and modification effects, and test for non-linear effects (Oliver & Morecroft, 2014).

3. Spread replicates across gradients, rather than pooling within treatments. Many studies

are based on the additive model of interactions, and employ factorial designs, where

single levels are replicated multiple times. However, as demonstrated in Fig. 5, factorial

designs are inappropriate if interactions are non-linear, as the results may be subjective

(dependent on the treatment levels selected by researchers), and therefore misleading.

Alternative regression-based study designs, where replicates are instead spread across a

gradient of levels of a factor (e.g. stocking densities / fire sizes), would be better able to

detect and describe nonlinear effects (Cottingham et al., 2005). It would not be feasible,

or even desirable, to test all combinations of two disturbance gradients in a single study.

However, a gradient could be achieved for a single disturbance without a substantial

increase in research effort compared with a factorial design. For example, a study with

24 sites in a factorial design would have 6 replicates for each of four treatment

combinations (i.e. a 2 × 2 design). The same number of sites, if employed a regression-

based design, could be spread across 12 different levels of one disturbance, and the two

levels of the second disturbance (i.e. a 12 × 2 design).

4. Extend gradients outside their current “natural” range. Most of the reviewed studies

used binary comparisons of treatments in their experiments (e.g. burnt/unburnt or patch

burnt/uniformly burnt), which are usually at “moderate” or average levels of

disturbance size or intensity. However, chain interactions with other disturbances, or

with external global change drivers such as climate, have the potential to generate

conditions, and levels of variation, that are outside the current natural range (Thornton 201

et al., 2014; Kayler et al., 2015). Studies which test responses to conditions both within,

and beyond current natural limits are therefore necessary to highlight potential

thresholds or state-changes that are currently unlikely, but that may occur with these

novel combinations of conditions (Belovsky et al., 2004; Kayler et al., 2015). For such

studies, simulation and modelling tools will be useful for informing how far outside

current ranges treatment gradients should extend (Kayler et al., 2015).

Interpretation

1. Clearly define the interaction being tested. Many studies do not clearly define which

aspect of an interaction they are testing (see Fig. 2). This makes it difficult for readers

to interpret results. Clearly defining which pathway/aspect of an interaction is being

tested, and ensuring that the design and analyses used actually test this component of

the interaction, will allow the study to be interpreted within a broader context.

2. Frame and interpret studies in a broad theoretical context. Many studies of disturbance

interactions focus their discussion and interpretation solely on the specific interaction

being studied (e.g. most of the reviewed papers focussed exclusively on the fire-grazing

literature in framing their studies). However, theoretical studies of interactions are often

more generalised, or are focussed on other interactions or ecosystems, and so go

unacknowledged in studies that are very specific in approach. Exploring and testing

ideas developed from studies of different types of interactions and different systems is

key to bridging the gap between theory and experimental studies (Peters et al., 2011;

Buma, 2015).

3. Discuss results in the context of existing theoretical models. System models and

management models are frequently used to inform management decisions. Yet many

relationships in such models have not been experimentally verified (Knapp et al.,

2011). Interpreting results in the context of existing theoretical models, and discussing

whether results support theoretical pathways is key to improving these models, and

hence management decisions.

202

Conclusions

Managing ecological interactions is a key challenge to biodiversity conservation, and one that is becoming increasingly important as global change drives rapid shifts in threats, abiotic conditions, and disturbance regimes (Brook et al., 2008; Turner, 2010). This challenge also presents an opportunity, through which a detailed, mechanistic understanding of ecological interactions can be used to develop novel solutions for biodiversity conservation. For example, failure to account for interactions between habitat modification and invasive species can cause invasive control programs to be ineffective, or even harmful for native species (e.g. Norbury et al., 2013). However, a mechanistic understanding of habitat-predator interactions is now being used to develop management interventions that may be both more efficient and effective than lethal control for protecting wildlife from introduced predators (Didham et al., 2007; Doherty et al., 2015).

Carefully executed empirical studies, which are well integrated with theory, are essential for developing the detailed understanding that these novel solutions require. For example, novel strategies to conserve frogs that are threatened by the disease Chytridiomycosis have been identified, not from solution-focussed studies, but from well-designed empirical studies, grounded in theory, that investigated the multiple interacting processes affecting disease prevalence and impacts (Scheele et al., 2014; Heard et al., 2015). The suggestions we have presented for the design and interpretation of disturbance interaction studies aim to guide research that will provide a similar understanding for managing multiple stressors. By gaining a deeper, mechanistic understanding of interactions, such studies will be able to provide context- specific information to guide management while also testing and refining theoretical models, which can then be translated to other processes and ecosystems.

Many studies of disturbance interactions aim to improve the management of ecosystems. However, studies which focus on specific management problems and fail to integrate relevant theory are often limited in their broader management applications. There are many studies of ecological interactions that are well integrated with the theoretical literature

(e.g. Mandle & Ticktin, 2012; Kimuyu et al., 2014), but such studies remain the exception, 203

rather than the rule. Until this trend is reversed, disturbance interaction experiments will continue to fall short of their potential in informing effective ecological management.

Acknowledgements

Ben Scheele provided valuable feedback on the manuscript. This research was financially supported by The Margaret Middleton Fund, The Norman Wettenhall Foundation and the Long

Term Ecological Research Network. P.S.B. was supported by an ARC DECRA Fellowship,

D.B.L. was supported by an ARC Laureate Fellowship.

Literature Cited

Allred BW, Fuhlendorf SD, Engle DM, Elmore RD (2011) Ungulate preference for burned patches reveals strength of fire-grazing interaction. Ecology and Evolution, 1, 132-144.

Ban SS, Graham NAJ, Connolly SR (2014) Evidence for multiple stressor interactions and effects on coral reefs. Global Change Biology, 20, 681-697.

Barot S, Abbadie L, Couvet D, Hobbs RJ, Lavorel S, Mace GM, Le Roux X (2015) Evolving away from the linear model of research: a response to Courchamp et al. Trends in Ecology & Evolution, 30, 368-370.

Belovsky GE, Botkin DB, Crowl TA et al. (2004) Ten suggestions to strengthen the science of ecology. Bioscience, 54, 345-351.

Boyd P, Brown C (2015) Modes of interactions between environmental drivers and marine biota. Frontiers in Marine Science, 2, 9.

Brook BW, Sodhi NS, Bradshaw CJA (2008) Synergies among extinction drivers under global change. Trends in Ecology & Evolution, 23, 453-460.

Brown CJ, Saunders MI, Possingham HP, Richardson AJ (2013) Managing for interactions between local and global stressors of ecosystems. PloS one, 8, e65765.

Buma B (2015) Disturbance interactions: characterization, prediction, and the potential for cascading effects. Ecosphere, 6, art70.

Cottingham KL, Lennon JT, Brown BL (2005) Knowing when to draw the line: designing more informative ecological experiments. Frontiers in Ecology and the Environment, 3, 145- 152.

204

Crain CM, Kroeker K, Halpern BS (2008) Interactive and cumulative effects of multiple human stressors in marine systems. Ecology Letters, 11, 1304-1315.

Dale VH, Joyce LA, McNulty S et al. (2001) Climate Change and Forest Disturbances. Bioscience, 51, 723-734.

Darling ES, Côté IM (2008) Quantifying the evidence for ecological synergies. Ecology Letters, 11, 1278-1286.

Denny M, Benedetti-Cecchi L (2012) Scaling up in ecology: mechanistic approaches. Annual Review of Ecology, Evolution, and Systematics, 43, 1-22.

Didham RK, Tylianakis JM, Gemmell NJ, Rand TA, Ewers RM (2007) Interactive effects of habitat modification and species invasion on native species decline. Trends in Ecology & Evolution, 22, 489-496.

Doherty TS, Dickman CR, Nimmo DG, Ritchie EG (2015) Multiple threats, or multiplying the threats? Interactions between invasive predators and other ecological disturbances. Biological Conservation, 190, 60-68.

Driscoll DA, Lindenmayer DB (2012) Framework to improve the application of theory in ecology and conservation. Ecological Monographs, 82, 129-147.

Dunne RP (2010) Synergy or antagonism—interactions between stressors on coral reefs. Coral Reefs, 29, 145-152.

Eby S, Burkepile DE, Fynn RWS et al. (2014) Loss of a large grazer impacts savanna grassland plant communities similarly in North America and South Africa. Oecologia, 175, 293- 303.

Essl F, Dullinger S, Rabitsch W, Hulme PE, Pyšek P, Wilson JRU, Richardson DM (2015) Historical legacies accumulate to shape future biodiversity in an era of rapid global change. Diversity and Distributions, 21, 534-547.

Fazey I, Fischer J, Lindenmayer DB (2005) What do conservation biologists publish? Biological Conservation, 124, 63-73.

Folt C, Chen C, Moore M, Burnaford J (1999) Synergism and antagonism among multiple stressors. Limnology and Oceanography, 44, 864-877.

Foster CN, Barton PS, Lindenmayer DB (2014) Effects of large native herbivores on other animals. Journal of Applied Ecology, 51, 929–938.

205

Foster CN, Barton PS, Sato CF, MacGregor CI, Lindenmayer DB (2015) Synergistic interactions between fire and browsing drive plant diversity in a forest understorey. Journal of Vegetation Science, 26, 1112-1123.

Fuhlendorf SD, Engle DM, Kerby JAY, Hamilton R (2009) Pyric herbivory: Rewilding landscapes through the recoupling of fire and grazing. Conservation Biology, 23, 588- 598.

Groffman P, Baron J, Blett T et al. (2006) Ecological thresholds: the key to successful environmental management or an important concept with no practical application? Ecosystems, 9, 1-13.

Heard GW, Thomas CD, Hodgson JA, Scroggie MP, Ramsey DSL, Clemann N (2015) Refugia and connectivity sustain amphibian metapopulations afflicted by disease. Ecology Letters, 18, 853-863.

Hovick TJ, Elmore RD, Fuhlendorf SD, Engle DM, Hamilton RG (2014) Spatial heterogeneity increases diversity and stability in grassland bird communities. Ecological Applications, 25, 662-672.

Huggett AJ (2005) The concept and utility of ‘ecological thresholds’ in biodiversity conservation. Biological Conservation, 124, 301-310.

Inouye BD (2001) Response surface experimental designs for investigating interspecific competition. Ecology, 82, 2696-2706.

Jackson MC (2015) Interactions among multiple invasive animals. Ecology, 96, 2035–2041.

Jackson MC, Loewen CJG, Vinebrooke RD, Chimimba CT (2015) Net effects of multiple stressors in freshwater ecosystems: a meta-analysis. Global Change Biology, Early view, doi: 10.1111/gcb.13028.

Kayler ZE, De Boeck HJ, Fatichi S et al. (2015) Experiments to confront the environmental extremes of climate change. Frontiers in Ecology and the Environment, 13, 219-225.

Knapp CN, Fernandez-Gimenez ME, Briske DD, Bestelmeyer BT, Wu XB (2011) An assessment of state-and-transition models: Perceptions following two decades of development and implementation. Rangeland Ecology and Management, 64, 598-606.

206

Kulakowski D, Jarvis D (2013) Low-severity fires increase susceptibility of lodgepole pine to mountain pine beetle outbreaks in Colorado. Forest Ecology and Management, 289, 544-550.

Kulakowski D, Veblen TT, Bebi P (2003) Effects of fire and spruce beetle outbreak legacies on the disturbance regime of a subalpine forest in Colorado. Journal of Biogeography, 30, 1445-1456.

Lindenmayer DB, Likens GE, Krebs CJ, Hobbs RJ (2010) Improved probability of detection of ecological “surprises”. Proceedings of the National Academy of Sciences, 107, 21957- 21962.

Mandle L, Ticktin T (2012) Interactions among fire, grazing, harvest and abiotic conditions shape palm demographic responses to disturbance. Journal of Ecology, 100, 997-1008.

McGranahan DA, Engle DM, Fuhlendorf SD, Winter SJ, Miller JR, Debinski DM (2012) Spatial heterogeneity across five rangelands managed with pyric-herbivory. Journal of Applied Ecology, 49, 903-910.

McGranahan DA, Engle DM, Fuhlendorf SD, Winter SL, Miller JR, Debinski DM (2013) Inconsistent outcomes of heterogeneity-based management underscore importance of matching evaluation to conservation objectives. Environmental Science & Policy, 31, 53-60.

Norbury G, Byrom A, Pech R, Smith J, Clarke D, Anderson D, Forrester G (2013) Invasive mammals and habitat modification interact to generate unforeseen outcomes for indigenous fauna. Ecological Applications, 23, 1707-1721.

Nugent G, McShea WJ, Parkes J et al. (2011) Policies and management of overabundant deer (native or exotic) in protected areas. Animal Production Science, 51, 384-389.

Oliver TH, Morecroft MD (2014) Interactions between climate change and land use change on biodiversity: attribution problems, risks, and opportunities. Wiley Interdisciplinary Reviews: Climate Change, 5, 317-335.

Peters DPC, Bestelmeyer B, Turner MG (2007) Cross–scale interactions and changing pattern– process relationships: Consequences for system dynamics. Ecosystems, 10, 790-796.

Peters DPC, Lugo AE, Chapin FS et al. (2011) Cross-system comparisons elucidate disturbance complexities and generalities. Ecosphere, 2, art81.

Peters DPC, Pielke RA, Bestelmeyer BT, Allen CD, Munson-McGee S, Havstad KM (2004) Cross-scale interactions, nonlinearities, and forecasting catastrophic events.

207

Proceedings of the National Academy of Sciences of the United States of America, 101, 15130-15135.

Piggott JJ, Townsend CR, Matthaei CD (2015) Reconceptualizing synergism and antagonism among multiple stressors. Ecology and Evolution, 5, 1538-1547.

Przeslawski R, Byrne M, Mellin C (2015) A review and meta-analysis of the effects of multiple abiotic stressors on marine embryos and larvae. Global Change Biology, 21, 2122- 2140.

Scheele BC, Hunter DA, Grogan LF et al. (2014) Interventions for reducing extinction risk in Chytridiomycosis‐threatened amphibians. Conservation Biology, 28, 1195-1205.

Sensenig RL, Demment MW, Laca EA (2010) Allometric scaling predicts preferences for burned patches in a guild of East African grazers. Ecology, 91, 2898-2907.

Thornton PK, Ericksen PJ, Herrero M, Challinor AJ (2014) Climate variability and vulnerability to climate change: a review. Global Change Biology, 20, 3313-3328.

Turner MG (2010) Disturbance and landscape dynamics in a changing world. Ecology, 91, 2833-2849.

Tylianakis JM, Didham RK, Bascompte J, Wardle DA (2008) Global change and species interactions in terrestrial ecosystems. Ecology Letters, 11, 1351-1363.

Underwood AJ (1995) Ecological Research and (and Research into) Environmental Management. Ecological Applications, 5, 232-247.

Wan HY, Rhodes AC, St. Clair SB (2014) Fire severity alters plant regeneration patterns and defense against herbivores in mixed aspen forests. Oikos, 123, 1479-1488.

Williams BK (2011) Adaptive management of natural resources—framework and issues. Journal of Environmental Management, 92, 1346-1353.

Wisdom MJ, Vavra M, Boyd JM, Hemstrom MA, Ager AA, Johnson BK (2006) Understanding ungulate herbivory-episodic disturbance effects on vegetation dynamics: Knowledge gaps and management needs. Wildlife Society Bulletin, 34, 283-292.

208

Supporting Information

Appendix S1. List of the 50 fire-grazing studies reviewed.

Studies were identified by searching the ISI Web of Knowledge Database for articles including the terms fire AND (grazing or browsing), and which were published between 2011 and the date of the search. This search was performed on the 17th April 2015 and returned 268 results, 50 of which were empirical studies that focussed at least in part on the interaction between fire and large herbivores.

Allred BW, Fuhlendorf SD, Engle DM, Elmore RD (2011) Ungulate preference for burned patches reveals strength of fire-grazing interaction. Ecology and Evolution, 1, 132-144. Allred BW, Scasta JD, Hovick TJ, Fuhlendorf SD, Hamilton RG (2014) Spatial heterogeneity stabilizes livestock productivity in a changing climate. Agriculture Ecosystems & Environment, 193, 37-41. Augustine DJ, Derner JD (2012) Disturbance regimes and mountain plover habitat in shortgrass steppe: Large herbivore grazing does not substitute for prairie dog grazing or fire. Journal of Wildlife Management, 76, 721-728. Augustine DJ, Derner JD (2014) Controls over the strength and timing of fire-grazer interactions in a semi-arid rangeland. Journal of Applied Ecology, 51, 242-250. Blackhall M, Veblen TT, Raffaele E (2015) Recent fire and cattle herbivory enhance plant-level fuel flammability in shrublands. Journal of Vegetation Science, 26, 123-133. Bock CE, Jones ZF, Kennedy LJ, Bock JH (2011) Response of rodents to wildfire and livestock grazing in an Arizona desert grassland. American Midland Naturalist, 166, 126-138. Dalldorf KN, Swanson SR, Kozlowski DE, Schmidt KM, Shane RS, Fernandez G (2013) Influence of livestock grazing strategies on riparian response to wildfire in northern Nevada. Rangeland Ecology & Management, 66, 34-42. Doxon ED, Davis CA, Fuhlendorf SD, Winter SL (2011) Aboveground macroinvertebrate diversity and abundance in sand sagebrush prairie managed with the use of pyric herbivory. Rangeland Ecology & Management, 64, 394-403. Eby S, Burkepile DE, Fynn RWS et al. (2014a) Loss of a large grazer impacts savanna grassland plant communities similarly in North America and South Africa. Oecologia, 175, 293-303. Eby SL, Anderson TM, Mayemba EP, Ritchie ME (2014b) The effect of fire on habitat selection of mammalian herbivores: the role of body size and vegetation characteristics. Journal of Animal Ecology, 83, 1196-1205.

209

Endress BA, Wisdom MJ, Vavra M, Parks CG, Dick BL, Naylor BJ, Boyd JM (2012) Effects of ungulate herbivory on aspen, cottonwood, and willow development under forest fuels treatment regimes. Forest Ecology and Management, 276, 33-40. Field JP, Breshears DD, Whicker JJ, Zou CB (2011) Interactive effects of grazing and burning on wind- and water-driven sediment fluxes: rangeland management implications. Ecological Applications, 21, 22-32. Hierro JL, Clark KL, Branch LC, Villarreal D (2011) Native herbivore exerts contrasting effects on fire regime and vegetation structure. Oecologia, 166, 1121-1129. Hovick TJ, Elmore RD, Fuhlendorf SD (2014) Structural heterogeneity increases diversity of non-breeding grassland birds. Ecosphere, 5, art62. Hovick TJ, Elmore RD, Fuhlendorf SD, Engle DM, Hamilton RG (2015) Spatial heterogeneity increases diversity and stability in grassland bird communities. Ecological Applications, 25, 662-672. Hovick TJ, Elmore RD, Wallred B, Fuhlendorf SD, Dahlgren DK (2014) Landscapes as a moderator of thermal extremes: a case study from an imperiled grouse. Ecosphere, 5, art35. Johansson MU, Granstrom A (2014) Fuel, fire and cattle in African highlands: traditional management maintains a mosaic heathland landscape. Journal of Applied Ecology, 51, 1396-1405. Kerns BK, Buonopane M, Thies WG, Niwa C (2011) Reintroducing fire into a ponderosa pine forest with and without cattle grazing: understory vegetation response. Ecosphere, 2, art59. Kimuyu DM, Sensenig RL, Riginos C, Veblen KE, Young TP (2014) Native and domestic browsers and grazers reduce fuels, fire temperatures, and acacia ant mortality in an African savanna. Ecological Applications, 24, 741-749. Koerner SE, Collins SL (2013) Small-scale patch structure in North American and South African grasslands responds differently to fire and grazing. Landscape Ecology, 28, 1293- 1306. Koerner SE, Collins SL (2014) Interactive effects of grazing, drought, and fire on grassland plant communities in North America and South Africa. Ecology, 95, 98-109. Limb RF, Fuhlendorf SD, Engle DM, Kerby JD (2011) Growing-season disturbance in tallgrass prairie: Evaluating fire and grazing on Schizachyrium scoparium. Rangeland Ecology & Management, 64, 28-36. Lunt ID, Zimmer HC, Cheal DC (2011) The tortoise and the hare? Post-fire regeneration in mixed Eucalyptus-Callitris forest. Australian Journal of Botany, 59, 575-581. Mandle L, Ticktin T (2012) Interactions among fire, grazing, harvest and abiotic conditions shape palm demographic responses to disturbance. Journal of Ecology, 100, 997-1008. Masunga GS, Moe SR, Pelekekae B (2013) Fire and grazing change herbaceous species composition and reduce beta diversity in the Kalahari sand system. Ecosystems, 16, 252- 268. 210

McGranahan DA, Engle DM, Fuhlendorf SD, Winter SJ, Miller JR, Debinski DM (2012) Spatial heterogeneity across five rangelands managed with pyric-herbivory. Journal of Applied Ecology, 49, 903-910. McGranahan DA, Engle DM, Fuhlendorf SD, Winter SL, Miller JR, Debinski DM (2013) Inconsistent outcomes of heterogeneity-based management underscore importance of matching evaluation to conservation objectives. Environmental Science & Policy, 31, 53- 60. McGregor HW, Legge S, Jones ME, Johnson CN (2014) Landscape management of fire and grazing regimes alters the fine-scale habitat utilisation by feral cats. PloS one, 9, e109097. McNew LB, Prebyl TJ, Sandercock BK (2012) Effects of rangeland management on the site occupancy dynamics of prairie-chickens in a protected prairie preserve. Journal of Wildlife Management, 76, 38-47. Monroe AP, O'Connell TJ (2014) Winter bird habitat use in a heterogeneous tallgrass prairie. American Midland Naturalist, 171, 97-115. Moranz RA, Fuhlendorf SD, Engle DM (2014) Making sense of a prairie butterfly paradox: The effects of grazing, time since fire, and sampling period on regal fritillary abundance. Biological Conservation, 173, 32-41. Nayak RR, Vaidyanathan S, Krishnaswamy J (2014) Fire and grazing modify grass community response to environmental determinants in savannas: Implications for sustainable use. Agriculture Ecosystems & Environment, 185, 197-207. Nield AP, Monaco S, Birnbaum C, Enright NJ (2015) Regeneration failure threatens persistence of Persoonia elliptica (Proteaceae) in Western Australian jarrah forests. Plant Ecology, 216, 189-198. Osem Y, Lavi A, Rosenfeld A (2011) Colonization of Pinus halepensis in Mediterranean habitats: consequences of afforestation, grazing and fire. Biological Invasions, 13, 485-498. Pekin BK, Wisdom MJ, Endress BA, Naylor BJ, Parks CG (2014) Ungulate browsing maintains shrub diversity in the absence of episodic disturbance in seasonally-arid conifer forest. PloS one, 9, e86288. Pillsbury FC, Miller JR, Debinski DM, Engle DM (2011) Another tool in the toolbox? Using fire and grazing to promote bird diversity in highly fragmented landscapes. Ecosphere, 2. Prober SM, Thiele KR, Speijers J (2013) Management legacies shape decadal-scale responses of plant diversity to experimental disturbance regimes in fragmented grassy woodlands. Journal of Applied Ecology, 50, 376-386. Radloff FGT, Mucina L, Snyman D (2014) The impact of native large herbivores and fire on the vegetation dynamics in the Cape renosterveld shrublands of South Africa: insights from a six-yr field experiment. Applied Vegetation Science, 17, 456-469.

211

Ramirez AR, Pratt RB, Jacobsen AL, Davis SD (2012) Exotic deer diminish post-fire resilience of native shrub communities on Santa Catalina Island, southern California. Plant Ecology, 213, 1037-1047. Richardson AN, Koper N, White KA (2014) Interactions between ecological disturbances: burning and grazing and their effects on songbird communities in northern mixed-grass prairies. Avian Conservation and Ecology, 9, 5. Scasta JD, Engle DM, Talley JL, Weir JR, Stansberry JC, Fuhlendorf SD, Harr RN (2012) Pyric-herbivory to manage horn flies (Diptera: Muscidae) on cattle. Southwestern Entomologist, 37, 325-334. Schutz AEN, Bond WJ, Cramer MD (2011) Defoliation depletes the carbohydrate reserves of resprouting Acacia saplings in an African savanna. Plant Ecology, 212, 2047-2055. te Beest M, Mpandza NJ, Olff H (2015) Fire and simulated herbivory have antagonistic effects on resistance of savanna grasslands to alien shrub invasion. Journal of Vegetation Science, 26, 114-122. Thomas-Van Gundy M, Rentch J, Adams MB, Carson W (2014) Reversing legacy effects in the understory of an oak-dominated forest. Canadian Journal of Forest Research, 44, 350-364. Traore S, Tigabu M, Jouquet P, Ouedraogo SJ, Guinko S, Lepage M (2015) Long-term effects of Macroternies termites, herbivores and annual early fire on woody undergrowth community in Sudanian woodland, Burkina Faso. Flora, 211, 40-50. Tuft KD, Crowther MS, McArthur C (2012) Fire and grazing influence food resources of an endangered rock-wallaby. Wildlife Research, 39, 436-445. Wan HY, Rhodes AC, St Clair SB (2014) Fire severity alters plant regeneration patterns and defense against herbivores in mixed aspen forests. Oikos, 123, 1479-1488. Winter SL, Fuhlendorf SD, Goad CL, Davis CA, Hickman KR (2011) Topoedaphic variability and patch burning in sand sagebrush shrubland. Rangeland Ecology & Management, 64, 633-640. Winter SL, Fuhlendorf SD, Goad CL, Davis CA, Hickman KR, Leslie DM, Jr. (2012) Restoration of the fire-grazing interaction in Artemisia filifolia shrubland. Journal of Applied Ecology, 49, 242-250. Winter SL, Hickman KR, Goad CL, Fuhlendorf SD, Gregory MS (2013) Seasonal fires, bison grazing, and the tallgrass prairie forb Arnoglossum plantagineum raf. Natural Areas Journal, 33, 327-338.

212

213