Fire Effects on Pollinators and Pollination

Fire effects on pollinators and pollination

Julian MacPherson Brown

Submitted in total fulfilment of the requirements of the degree of Doctor of Philosophy

School of Ecosystem and Forest Sciences University of Melbourne June 2016

Abstract

One of the most important knowledge gaps currently inhibiting biodiversity conservation in fire-prone landscapes concerns the interactions between fire and other ecosystem processes. Animal-mediated pollination, whereby animals transport pollen between the male and female parts of flowers, is an important ecosystem processes as it is required for approximately 80-90% of the Earth’s flowering plants to reproduce sexually. The aim of my thesis is to better understand fire effects on pollinators and pollination and their role in fire management. First I synthesise the literature to develop a conceptual model, and then describe my empirical work exploring the ideas underlying this model. The central idea is that fire can influence plant-pollinator interactions through multiple processes operating over different spatial scales, and this was supported by my empirical work (and a study (Ponisio et al. 2016) from North America published as I was finalising my thesis). I collected data from the Mediterranean climate zone of south-eastern Australia, employing a space-for-time substitution design with spatially independent sites along a 75 year post-fire successional gradient. I monitored pollinator visitation to the sexually-deceptive orchid Caladenia tentaculata, capsule set of the food-deceptive orchid Diuris maculata sensu lato, and sampled aerial invertebrate assemblages (focusing on known fly and wasp pollinator groups) across these sites. I found that visitation to C. tentaculata was greatest when the site was recently burnt but the surrounding landscape was long- unburnt. Diuris maculata s.l. capsule set was influenced mostly by rainfall in the growing and flowering season (winter and spring). Flies and wasps exhibited only moderate or weak responses to post-fire age. Interestingly, though, species richness was positively related to fire age diversity within 800 m of sample locations but negatively related to fire age diversity within 200 m. My model could form the basis for simulations of plant and pollinator population dynamics in fire-prone landscapes, parameterised with spatially-explicit empirical data as presented in my thesis.

Declaration

This is to certify that:

i. The thesis comprises only my original work towards the PhD except where indicated in the Preface. ii. Due acknowledgement has been made in the text to all other material used. iii. The thesis is fewer than 100,000 words in length, exclusive of tables, maps, bibliographies and appendices.

Julian Brown

June 2016

Preface

This thesis comprises a general introduction (Chapter 1) to fire ecology and management and the key themes underlying my PhD research, four stand-alone papers presenting my research (Chapters 2-5) that have been or will be submitted for publication, and a general discussion (Chapter 6) of the research chapters with recommendations for future directions. Chapters 2-5 were prepared for publication in collaboration with one or more of my supervisors as co-authors. Therefore there is some overlap in chapter content, particularly with regards to descriptions of the study area, and the pronoun ‘we’ is used instead of ‘I’ in recognition of the co-authors’ contributions. For empirical work I conceived the study designs (with advice from supervisors), conducted the field work (with help from field assistants), processed samples post-field work (with help from lab assistants), and analysed the data (with advice from supervisors and others acknowledged below). I prepared the first drafts of all papers and subsequently responded to comments from supervisors, such that I contributed at least 70% of the work to all chapters.

iii

Acknowledgements

I would like to thank my supervisors Alan York, Fiona Christie, and Michael McCarthy for suggestions and insights about sample design, the practicalities of field work, unfamiliar data analysis techniques, and writing manuscripts for publication. I would also like to thank Julian Di Stefano and Thomas Duff for discussions and advice on data analysis and other aspects of the project, Colin Bower for teaching me baiting techniques, and Dave Pitts, Pauline Rudolph, Ryan Duffy, Mike Duncan, Dave Patience and other DELWP and Parks Vic staff for advice and support in the field.

I would like to thank Mal Brown (Dad), Julio Najera, and John Loschiavo for field assistance, and Julio again and Amanda Ashton for helping me with the long hours of insect sorting.

Finally, I would like to thank my wife, Simmy, for her unwavering support and understanding, for tolerating long periods of absence during field work, and for pretending to be interested in my work when I tried to explain it.

Funding for this project was generously provided by the Department of Environment, Land, Water and Planning (Hawkeye) and the Holsworth Wildlife Research Endowment.

CONTENTS

1. GENERAL INTRODUCTION ...... 1

1.1. The history of fire and life on Earth ...... 1

1.2. The history of humans and fire ...... 3

1.3. Future fire management for biodiversity conservation ...... 4

1.4. Thesis aims and structure ...... 6

2. EFFECTS OF FIRE ON POLLINATORS AND POLLINATION ...... 9

2.1. Abstract ...... 9

2.2. Introduction ...... 10

2.3. Conceptual model...... 11

2.3.1. Pollinator source map ...... 13

2.3.2. Pollination services map ...... 14

2.3.3. Cell networks ...... 14

2.4. Traits ...... 18

2.4.1. Pollinators – cell fire history ...... 18

2.4.2. Pollinators – cell composition and configuration ...... 19

2.4.3. Plants – cell fire history ...... 20

2.5. Management recommendations and future research ...... 22

2.5.1. Immediate actions...... 22

2.5.2. Future research ...... 22

3. FIRE EFFECTS ON POLLINATION IN A SEXUALLY-DECEPTIVE ORCHID ...... 25

3.1. Abstract ...... 25

3.2. Introduction ...... 26

3.3. Methodology ...... 28

3.3.1. Study species...... 28

3.3.2. Study landscape and site selection...... 28

3.3.3. Pollinator hosts (scarab beetles) ...... 30

3.3.4. Pollinator visitation ...... 30

3.3.5. Collection of age class data ...... 31

3.3.6. Data analysis ...... 32

3.4. Results ...... 33

3.5. Discussion ...... 36

4. FIRE, FOOD, AND SEXUAL-DECEPTION IN THE NEIGHBOURHOOD OF VICTORIAN

ORCHIDS ...... 39

4.1. Abstract ...... 39

4.2. Introduction ...... 40

4.3. Methodology ...... 43

4.3.1. Study landscape and site selection...... 43

4.3.2. Data collection ...... 44

4.3.3. Data analysis ...... 46

4.4. Results ...... 50

4.4.1. Flowering and fire history...... 50

4.4.2. Diuris maculata s.l...... 51

4.4.3. Caladenia tentaculata ...... 54

4.5. Discussion ...... 55

4.5.1. Diuris maculata s.l...... 56

4.5.2. Caladenia tentaculata ...... 57

5. SCALE-DEPENDENCE OF FIRE DIVERSITY-SPECIES DIVERSITY RELATIONSHIPS FOR

FLOWER-VISITING FLIES AND WASPS OF SOUTH-EAST AUSTRALIA ...... 59

5.1. Abstract ...... 59

5.2. Introduction ...... 60

5.3. Methodology ...... 62

5.3.1. Study landscape and site selection...... 62

5.3.2. Trapping ...... 64

5.3.3. Collection of habitat data ...... 64

5.3.4. Collection of age class data ...... 65

5.3.5. Data analysis ...... 67

5.4. Results ...... 70

5.4.1. Habitat variables ...... 70

5.4.2. Assemblage ...... 71

5.4.3. Individual taxa ...... 74

5.4.4. Species richness ...... 75

5.5. Discussion ...... 77

6. GENERAL DISCUSSION ...... 82

6.1. Summary and synthesis ...... 82

6.2. Future directions ...... 85

6.2.1. Management of (semi)natural landscapes ...... 85

6.2.2. Management of mixed (semi)natural-agricultural landscapes...... 87

7. REFERENCES ...... 89

vii

viii

List of Tables

Table 3.1: evidence ratios relative to the null (Akaike weight of landscape model/Akaike weight of null model), parameter estimates (with p-value), and R-squared values for the proportion of land covered by AC1-5 for both scarab capture and wasp visitation...... 34

Table 4.1: shows for each floral variable the best model (lowest AICc), evidence ratio relative to the null, R2, and parameter estimates (with p-value) for 1) the difference in the response between AC1 (the reference category) and AC2, AC3, and AC4 (since they were entered as dummy variables), and 2) fire frequency (FF), and 3) the minimum inter-fire interval (min) if they were included in the best model (bold indicates statistically significant effects at α = 0.05)...... 53

Table 4.2: shows for each D. maculata capsule set and C. tentaculata visitation model the parameter estimate and p-value for the floral variable being tested (bold indicates statistically significant effects at α = 0.05), and the change in AICc, Akaike weight, and explained deviance (D2)...... 53

Table 5.1: Responses of species to habitat, local, and landscape variables (with landscape heterogeneity measured at 200, 400, and 800 m). The evidence ratio relative to the null model, and deviance explained (D2 %) is shown for the best model at each level. + indicates positive relationship, – indicates negative relationship...... 75

List of Figures

Figure 2.1: Conceptual model. The pollination services map contains a plant–pollinator interaction network in each cell (only one cell shown) based on the abundance of each pollinator species arriving in the cell and each plant species flowering in the cell. The outcome of each cell’s interaction network (in conjunction with the quality of each interaction partner) for that time-step is: (i) seed production for each resident plant species; and (ii) the floral value for each pollinator species, which is used in conjunction with the nesting value of each cell to calculate offspring production in the pollinator source map. Before the pollination services map is calculated in the next time-step, seeds, offspring, and adults remain dormant, transition to different life stages (e.g. plant 3 commences flowering), senesce, or are killed by fire (direct effects: e.g. plant 1 and pollinator 3), and nest availability changes with fire age...... 17

Figure 3.1: The study landscape. Broken line indicates Victoria-South Australia border. Grey areas represent native vegetation and white areas represent agricultural/residential areas. The insert shows the location of the study landscape (black rectangle) in Victoria (grey area)...... 29

Figure 3.2: mean (with standard error) untransformed A) number of scarabs captured and B) number of wasp visits recorded in each age class (AC). Columns labelled ‘a’ are statistically significantly different (α = 0.05) from columns labelled ‘b’...... 35

Figure 3.3: untransformed wasp visitation as a function of the proportion of land within pollinator flight range covered by AC5 (vegetation burnt greater than 50 years ago) for sites in AC1 (0-3 years post fire) and AC2-5 (4+ years post fire). Regression lines (with confidence bands) produced using R package ‘visreg.’ ...... 35

Figure 4.2: the mean (with standard error) number of flowering individuals of rewarding taxa thought to share pollinators with a) Diuris maculata (to enhance

x clarity only Dillwynia glaberrima, Pulteneae, and Bossiaeeae are shown as they demonstrate the range of between-taxa variation in flowering responses to fire), and b) Caladenia tentaculata (for clarity the combination of all species is not shown as it is qualitatively similar to the Leptospermum myrsinoides response only larger). (AC1 n = 13 sites, AC2 n = 10 sites, AC3 n = 12, AC4 n = 6 sites)...... 51

Figure 4.3: spatial autocorrelation (Moran’s I, with 95% confidence intervals) at a range of distances for the residuals of the a) capsule set model with all (minimally correlated) environmental predictors, b) visitation model with all (minimally correlated) environmental predictors, and c) visitation model with all environmental predictors plus a random effect for site...... 52

Figure 4.4: Partial regression plots showing the log odds of Diuris maculata capsule set as a function of a) the number of flowering Dillwynia glaberrima individuals, and b) winter rainfall...... 54

Figure 4.5: Partial regression plot showing the number of wasp visits to Caladenia tentaculata as a function of the number of Burchardia umbellata flowers...... 55

Figure 5.1: The study landscape. Broken line indicates Victoria-South Australia border. Grey areas represent native vegetation and white areas represent agricultural/residential areas. The insert shows the location of the study landscape (black rectangle) in Victoria (grey area)...... 63

Figure 5.2: illustrates the range of responses exhibited by habitat variables. The mean (with standard error bars) of each habitat variable summed across height categories (for vertically structured variables dead and living vegetative biomass) and intervals for each site is shown for each age class...... 71

Figure 5.3: Bi-plots based on canonical correspondence analysis of the flower-visiting fly and wasp community with respect to A) habitat variables, B) the proportion of land covered by different post-fire age classes within 200 m and C) 800 m. Bm = Bombyliidae, B = Braconidae, I = Ichneumonidae, M = Muscidae, G = Mycomya, S = Syrphidae, Tb = Tabanidae, Tp = Tiphidae, and T = Tachinidae. Location of

xi abbreviation indicates the optima of the unimodal response surface for each morphospecies from the corresponding taxonomic group...... 73

Figure 5.4: mean (with standard error bars) species richness (raw data) in each fire age diversity category (1, 2, or 3 fire age classes) measured within A) 200 m and B) 800 m of sample locations, A is statistically significantly different from B (α = 0.05)...... 76

Figure 5.5: Partial regression plots of species richness as a function of fire age Shannon diversity within A) 200 m (holding Shannon diversity within 800 m constant) and B) 800 m (holding Shannon diversity within 200 m constant)...... 76

Figure 5.6: spatial autocorrelation (Moran’s I, with 95% confidence intervals) at a range of distances for the residuals of the best model for each taxa...... 77

xii

1. GENERAL INTRODUCTION

1.1. The history of fire and life on Earth

Fire on earth is ancient and widespread. As a physical phenomenon, fire requires three elements; heat, oxygen, and fuel. Heat sources such as lightning strikes have probably been present throughout most of Earth’s history, but it was not until early photosynthetic organisms produced an oxygen-rich atmosphere and terrestrial plants (fuel) emerged that fire came into being (Glasspool, Edwards & Axe 2004; Pausas & Keeley 2009). Fire seems to have been continuous since its appearance, though with fluctuating activity in response to atmospheric oxygen concentrations and climatic changes (Scott & Glasspool 2006; Marlon et al. 2008). Fires now occur regularly in most vegetation types (Mouillot & Field 2005), and are expected to become more widespread and prevalent under climate change (Jolly et al. 2015).

A central concept in understanding fire as an ecological process is the fire regime. It has a number of aspects (Gill 1975; Keeley et al. 2011a); the season in which fire occurs, the length of time between successive fires (inter-fire interval), fire intensity, fire size and spatial patterning, and whether it is a peat fire, surface fire, or it reaches the canopy. There are four main factors that influence the fire regime (Keeley et al. 2011a); biomass production (potential fuel), periodic drying (turning potential into available fuel), fuel structure (vertical and horizontal distribution of fuels), and ignition sources (e.g. lightning, arson). Climate is thus an important determinant of fire regimes at large spatial scales, but within a landscape there can be variation in fire regime resulting

1 from, for example, topographic features that inhibit fire spread (Keeley et al. 2011a; McKenzie, Miller & Falk 2011).

While climate and geology strongly influence the origin and distribution of plant species, it is now apparent that across much of the Earth’s surface fire plays an important role (Pausas & Keeley 2009; Keeley et al. 2011a; Pausas & Schwilk 2012). Several traits influence a plant’s ability to survive fire. Many plants possess vegetative structures, such as underground storage organs or aboveground buds insulated by bark, that enable resprouting following tissue damage caused by fire (Lamont, Enright & He 2011; Bowman et al. 2012). Further, some traits allow plants to exploit the flush of resources available in the early post-fire environment. Some plants, both respouters and non-sprouters (i.e. plants lacking fire-survival traits), produce seeds that lie dormant in the soil or in fruiting structures on the parent plant (serotiny) and are stimulated to germinate by fire (Keeley et al. 2011b). Many resprouting plants with limited seed dormancy are stimulated to flower by fire, releasing seeds in the first few years after fire (Lamont & Downes 2011). Whether these traits are adaptations to fire or traits that evolved in response to other selective forces but are adaptive in fire-prone landscapes (i.e. exaptations sensu Gould & Vrba 1982)) has been the subject of much recent debate (Bradshaw et al. 2011; Keeley et al. 2011a), though there is growing consensus that certain forms of resprouting, serotiny, and fire-stimulated flowering are adaptations to fire (Crisp et al. 2011; Midgley & Bond 2011; Bowman et al. 2012; He et al. 2012; Pausas & Schwilk 2012). Regardless of origin, the fitness benefits a given trait confers varies with fire regime (Keeley et al. 2011a); for example, fire-dependent recruitment is only adaptive for non-sprouters when fire intervals are long enough to allow a seed bank to develop but short enough to stimulate germination before seed banks senesce. Moreover, different fire-related traits are beneficial under different fire regimes; for example, resprouters may tolerate shorter inter-fire intervals than non-sprouters (Keith 1996). Thus, there are broad-scale biogeographical patterns of associations between fire regime and vegetation types with suits of fire-related traits (Bond, Woodward & Midgley 2005; Keeley et al. 2011a). Further, topography can drive finer-scale associations within fire-prone landscapes. Examples include patches of fire-sensitive rainforest inhabiting less flammable canyons and scree slopes in a matrix of fire-prone vegetation (Bowman 2000; Wood, Murphy & Bowman 2011), or non-sprouters

2 inhabiting rocky outcrops (which are less frequently burnt) and resprouting species dominating the surrounding forest matrix (Clarke 2002b; Clarke 2002a).

Fire-determined vegetation patterns are not static. Fire may have facilitated the expansion of angiosperms in the Cretaceous (Bond & Scott 2010) and C4 grasses later in the Tertiary (Keeley & Rundel 2005). Both of these expansions might be driven by positive feedbacks between these plant groups and fire, whereby traits that allowed these groups to dominate in open habitats created by fire also enhanced landscape flammability, maintaining open habitats and thus reinforcing the dominance of these plants (Keeley & Rundel 2005; Bond & Scott 2010). Today a similar scenario may be occurring where positive feedbacks between exotic grasses and fire appear to be having negative impacts on more fire-sensitive native plant species (e.g. Bowman et al. 2014). Fire’s role as an environmental filter on ecological communities is also evidenced by recent declines of several plant and animal species following the cessation of Aboriginal fire management practices with European colonisation (Bowman & Panton 1993; Bowman et al. 2012; Bird et al. 2013). Thus, fire appears to have had a long history as an ecological and evolutionary force that continues to the present day.

1.2. The history of humans and fire

Like many species humans evolved in fire-prone landscapes, and there is evidence that fire shaped our evolution and biogeography (Attwell, Kovarovic & Kendal 2015; Parker et al. 2016). However, our relationship with fire has been unique because for us it has become a tool. Pyne (2013) distinguishes two parallel historical narratives of fire use by humans. The first traces the use of fire to cook raw food items into more digestible forms, then earth into more malleable and diverse forms such as iron and glass, and metals into various tools and eventually machines. These machines housed fire and gave us most of the modern comforts of our homes, transport networks, etc., as well as anthropogenic climate change, guns, and other less comfortable things.

The second narrative traces our attempts to direct fire’s behaviour in the landscapes we inhabit, and this is the story framing my thesis. Fire’s ability to influence the

3 movements of organisms (e.g. game fleeing flames during the fire event, and foraging on fresh grass in the early post-fire environment) and make them easier to detect (e.g. clearing vegetation concealing burrows, shallow-rooting tubers, and other food items) was harnessed early by controlling where and when to burn to facilitate hunting and gathering of native species. It was also an effective way of clearing travel routes, communicating, and defending against enemies and unwanted fires. Following domestication of animals and plants, fire was used to promote forage for livestock and to burn agricultural fields, freshly cut from forest or left fallow, as a means of removing (combusting) pests and liberating nutrients. In recent times, agrochemicals have partially replaced fire in agriculture, particularly in the developed world. In areas where forests have production, recreation, and/or spiritual values, or where wildfire possess a threat to human lives and property, fire suppression and prevention through prescribed burning have been used. There is now a growing perception that fire suppression, the cessation of aboriginal burning practices, landscape modification, climatic changes, and species introductions (e.g. flammable grasses) have created fire regimes outside the historical ranges to which biota had adapted and so pose a threat to biodiversity (Driscoll et al. 2010; Bradstock, Williams & Gill 2012a; Spies et al. 2012). Management agencies charged with conserving biodiversity in a number of fire-prone regions now aim to restore historical fire regimes (e.g. North America) or determine and create fire regimes supporting species of management interest (e.g. southern Australia) (Driscoll et al. 2010; Spies et al. 2012).

1.3. Future fire management for biodiversity conservation

The scientific study of fire ecology allows us to understand fire as an ecological and evolutionary force and so predict the ecological consequences of our attempts to direct fire’s behaviour in the landscape. Our future use of fire will depend on what we learn from fire ecology. Driscoll et al. (2010) identified three major knowledge gaps in fire ecology currently inhibiting fire management for biodiversity conservation: 1) a mechanistic understanding of species’ responses to fire regimes; 2) knowledge of how the spatial and temporal arrangement of fires influences the biota; and 3) an

4 understanding of interactions between fire and other ecosystem processes such as predation and pollination. The focus of my thesis is on interactions between fire and animal-mediated pollination because the small number of studies that have investigated fire effects on pollination have all found evidence of such effects (Ne'eman, Dafni & Potts 2000; Potts, Dafni & Ne'eman 2001; Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013; Ponisio et al. 2016).

An understanding of animal-mediated pollination is essential to the management of many plant species. For approximately 80-90% of the Earth’s flowering plants to reproduce sexually they require animals to transport pollen from the male parts of flowers to the female parts of flowers (Ollerton, Winfree & Tarrant 2011). A number of countries are currently experiencing pollinator declines as a result of land use intensification, climate change, alien species, pest, and pathogens (Potts et al. 2010; Vanbergen et al. 2013). When plants lose their pollinators sexual seed production fails, eventually leading to local extinctions unless the plants are capable of self-pollination or vegetative reproduction, but even then they may suffer reduced genetic diversity and inbreeding depression (Wilcock & Neiland 2002; Eckert et al. 2010; Pauw & Bond 2011; Pauw & Hawkins 2011). It is thus important to figure out what we can do to better manage plant-pollinator interactions.

The effects of land use, land management, and disturbance on pollination have been the focus of much recent research in agricultural and urban landscapes. This work has led to the understanding that plant-pollinator interactions in heterogeneous landscapes can be influenced by multiple processes operating over different spatial scales (Kremen et al. 2007; Kennedy et al. 2013; Vanbergen 2014). For instance, pollination services to a target plant can be influenced both the local flower community helping to attract pollinators and the distribution of flowers in the surrounding landscape attracting pollinators away from the target plant (e.g. Holzschuh et al. 2011). Fire creates heterogeneity at multiple spatial scales (McKenzie, Miller & Falk 2011), but at the time I commenced writing my thesis no studies had considered fire effects on pollination through multiple spatial scales. Thus it was unclear whether the multi-scale understanding developed in agricultural and urban landscapes applied to fire-prone landscapes, and this was what I set out to explore. A study published a few months

5 before I completed my thesis found evidence of fire effects on plant-pollinator interactions at local- and landscape-scales (Ponisio et al. 2016), suggesting the multi- scale understanding is applicable in the North American system studied. My research focuses on an Australian system.

1.4. Thesis aims and structure

The aim of my thesis is to better understanding fire effects on pollinators and pollination and their role in fire management for biodiversity conservation. I address two broad questions in different chapters:

1) How might the multi-scale understanding of plant-pollinator interactions in heterogeneous landscapes, based on empirical work from agricultural and urban landscapes, be incorporated into and extend fire management for biodiversity conservation? This is addressed in chapter 2.

In Chapter 2 I develop a conceptual model for understanding fire effects on plant- pollinator interactions. The model is developed by synthesising the scant empirical data, but also spatially explicit models of pollination services (based on the model by Kremen et al. (2007)) and fire-driven population dynamics currently available to managers to facilitate incorporation of my model into ecological management of fire-prone landscapes. I also review phenotypic traits of pollinators and plants that could assist managers in parameterising this model. This chapter has been submitted to and accepted by the Journal of Applied Ecology: Brown, J, York, A, Christie, F, and McCarthy, M (2016) Effects of fire on pollinators and pollination, Journal of Applied Ecology, DOI: 10.1111/1365-2664.12670.

2) Does the multi-scale understanding of plant-pollinator interactions in heterogeneous landscapes apply to plant-pollinator interactions in fire-prone landscapes of the Mediterranean climatic zone of south-eastern Australia? This is addressed in chapters 3, 4, and 5.

Chapter 3 explores the possibility of fire effects on interactions between the Australian terrestrial orchid Caladenia tentaculata and its wasp pollinator at multiple spatial scales. Understanding fire effects on pollination of terrestrial orchids is important as 1) there are numerous threatened species in southern Australia, 2) they often have highly specialised pollination systems making them vulnerable to disruption of pollination, and 3) they often, or are at least commonly thought to, increase flowering in response to fires in appropriate seasons such that regular burning has sometimes been recommended for their management. I test whether relationships between pollinator visitation and post-fire age vary with the scale at which fire age is measured, as evidence of multiple processes operating at different spatial scales, and whether these effects interact. This chapter has been submitted to and accepted by the International Journal of Wildland Fire: Brown, J, York, A, Christie, F (2016) Fire effects on pollination in a sexually- deceptive orchid, International Journal of Wildland Fire, DOI: 10.1071/WF15172

Chapter 4 investigates the effects of fire-driven changes in local floral communities on terrestrial orchids. I initially set out to relate seed production of five orchid species with contrasting pollination systems (some offering food rewards, some mimicking other rewarding species, and some exhibiting sexual-deception) to local- and landscape-scale pyrogenic habitat changes. However, a very dry growing/flowering season with high rates of flower wilting and florivory left only two species with sufficient data for even simple modelling of local-scale effects. Thus the chapter investigates the effects of fire- driven changes in rewarding heterospecifics on two rewardless orchid species (Diuris maculata sensu lato and Caladenia tentaculata) putatively dependent on these co- flowering heterospecifics to attract and support pollinators. This chapter has been submitted to Austral Ecology and is currently in review.

In Chapter 5 I shift focus away from pollination to pollinator assemblages, though still continue with the theme of fire effects at multiple spatial scales. Patch mosaic burning – the creation of landscape mosaics of different fire age classes supporting different species and so enhancing faunal species richness – is a common fire management strategy for fauna (Parr & Andersen 2006; Bradstock, Williams & Gill 2012b). It has been argued that the aim of patch mosaic burning is to promote gamma diversity via beta diversity (Farnsworth et al. 2014), but I focus on alpha diversity because it is more

7 directly related to stabilising and enhancing pollination services (Hoehn et al. 2008; Tscharntke et al. 2012). Fire effects on pollinating flies and wasps are poorly understood so I focus on this group (I also captured too few bees for meaningful analysis). Pollinator assemblages are diverse and can exhibit interspecific variation in the scale at which the environment is experienced (e.g. Westphal, Steffan-Dewenter & Tscharntke 2006; Benjamin, Reilly & Winfree 2014). I thus test the hypothesis that interspecific variation in the scale at which fire-driven changes in the environment are experienced results in species diversity responding to pyrogenic heterogeneity at multiple spatial scales. The scale-dependence of species diversity-pyrogenic diversity relationships is poorly understood for fauna generally.

The concluding Chapter 6 synthesises and discusses my research in the context of the overarching and specific questions outlined above, as well as the broader context of anthropogenic fire.

2. EFFECTS OF FIRE ON POLLINATORS AND POLLINATION

2.1. Abstract

Increased incidence of landscape fire and pollinator declines with co-extinctions of dependent plant species are both globally significant. Fire can alter species distributions, but its effects on plant-pollinator interactions are poorly understood so its present and future role in coupled plant-pollinator declines cannot be assessed. We develop a conceptual model of fire effects on plant-pollinator interactions. We review the empirical literature in the context of this model to identify important knowledge gaps regarding the processes underlying these effects and the phenotypic traits of flowering plants and pollinators mediating these effects. Fire generates, and plant–pollinator interactions respond to, heterogeneity at multiple spatial scales. There is evidence of local-scale fire effects on these interactions, but landscape-scale effects are poorly understood. Nest location and floral resource utilization primarily mediate pollinator survival during and after fire. Voltinism and mobility traits are potentially important but poorly studied. Plant traits mediating flowering responses to fire include growth form, phenology, and potentially bud location, seasonal changes in bud exposure, and response to bud damage. We suggest management actions and an agenda for future research to fill knowledge gaps currently inhibiting predictions of fire effects on plant- pollinator interactions. Fire regimes promoting floral diversity at local scales provide a surrogate means of managing pollinators and pollination while empirical research

9 continues. Above-ground nesting, univoltine pollinators may be particularly vulnerable under expected fire regime changes. Improved knowledge of traits mediating exploitation of landscape heterogeneity could be used to enhance persistence of these species. Ultimately, our conceptual framework could be used as a basis for understanding fire effects on aggregate network properties to inform fire management strategies buffering plant–pollinator networks against secondary species extinctions.

2.2. Introduction

Fires occur regularly in most vegetation types (Mouillot & Field 2005) and are expected to become more widespread and prevalent under climate change (Jolly et al. 2015). The history of fire events occurring within a specific area and period of time defines the fire regime, which is characterized by the frequency (inter-fire interval), season, intensity, size, and spatial patterning of fires (Gill 1975; Keeley et al. 2011a). Fire regimes can be manipulated by fire suppression and applying prescribed fire (e.g. Keeley et al. 2011a), though the ability to control wildfire is expected to decrease as fire intensity increases across much of the globe under climate change (Stephens et al. 2012; Flannigan et al. 2013). Fire regimes can act as filters on biotic communities, determining relative abundances and species composition (e.g. Noble & Slatyer 1980; Keeley et al. 2011a). Predicting the ecological consequences of altered prescribed and wildfire regimes requires knowledge of relationships between fire regimes and biota.

Predictions of fire effects on biodiversity are currently limited by substantial knowledge gaps, particularly (i) animal responses to fire, and (ii) interactions between fire and ecosystem processes such as herbivory and pollination that influence plant survival and reproduction (Driscoll et al. 2010). Here we focus on pollinating animals and animal- mediated pollination. Most flowering plants require animals to effect reproduction by transporting pollen to and from mates (Ollerton, Winfree & Tarrant 2011). While some animal-pollinated plants are able to self-pollinate in the absence of pollinator visitation, they may ultimately suffer inbreeding depression and reduced genetic diversity (Wilcock & Neiland 2002; Eckert et al. 2010). Recent observations of pollinator declines in several countries are thus concerning (Potts et al. 2010; Vanbergen et al.

2013). While the role of altered fire regimes in these declines is unclear (Potts et al. 2010), fire management of pollinator habitat might help mitigate pollinator declines in regions where fire is an important ecological process (Willmer 2011). Moreover, a better understanding of fire effects on pollinators and their interactions with plants should improve predictions of plant population responses to fire. For instance, post-fire flowering (reproductive maturity) can precede recovery of pollinators and so pollination by several years (Ne'eman, Dafni & Potts 2000; Geerts, Malherbe & Pauw 2012); however the rate of pollinator recovery can depend on fire characteristics such as size and patchiness (Watson et al. 2012). Thus the risk of subsequent fires occurring before significant reproduction – which can cause local plant extinctions (Keith 1996) – might be mediated by spatial patterns of fire, though this has not been tested. We identify similarly important knowledge gaps and possible management actions by (i) developing a conceptual model of coupled plant-pollinator population dynamic in fire-prone landscapes by synthesizing models of plant-pollinator interactions and fire-driven population dynamics currently available to managers, and (ii) reviewing empirical studies of fire’s effects on plant-pollinator interactions, and the phenotypic traits mediating these effects, in the context of our model. We suggest future research directions for filling important knowledge gaps and developing our model to better inform biodiversity conservation in fire-prone landscapes.

2.3. Conceptual model

Choosing a fire management strategy that is most likely to achieve the general conservation objective of avoiding species extinction is difficult given the long time- scales relevant to population persistence and alternative climate change scenarios. Simulation modelling based on observed responses of plants and animals to fire is advocated and increasingly used to predict the consequences (and effectiveness) of alternative fire management strategies and climatic change on fire regimes and, in turn, biodiversity (e.g. Bradstock et al. 1998; Franklin et al. 2001; Bradstock et al. 2005; Driscoll et al. 2010; Pacifici et al. 2015; Regos et al. 2015). The landscape is typically conceived of as a grid, with each cell having a fire age that either increases or returns to

11 zero (is burnt) in each time-step, and an inter-fire interval across time-steps, depending on fire ignition (planned and unplanned) and spread between cells. The fire history of each cell determines demographic characteristics of plant species such as survival, seed production, and seedling establishment, and habitat suitability for animal species (based on empirically-derived abundance–fire age relationships). As fire history changes across time-steps, cell occupancy changes according to demographic processes and habitat suitability, as well as dispersal from neighbouring occupied cells.

We believe the addition of plant–pollinator interactions could improve predictions of population persistence by improving parameter estimates and model structure. In the existing framework plant reproduction in each cell is typically assumed to occur when the cell reaches a certain fire age (e.g. when flowering commences) (e.g. Bradstock et al. 1998; Franklin et al. 2001). Pollinator visitation and seed set can vary with fire age (Ne'eman, Dafni & Potts 2000; Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013; Bourg, Gill & McShea 2014), suggesting a more accurate conceptualization of plant population dynamics would allow reproduction to vary as a function of pollination services provided to each cell. Pollinator abundance– fire age relationships have been observed (e.g. Potts et al. 2003a; Potts et al. 2003b; Chalmandrier et al. 2013; Rodríguez & Kouki 2015) and so might be used to derive a habitat suitability value for each cell according to its fire age. However, some animals require complementary resources (e.g. nests and forage) from contrasting fire age classes (Woinarski 2005). This has not been investigated in pollinators, but both pollinator nesting substrate and floral resources are known to vary with fire history (De Swardt 1993; Potts et al. 2003a; Potts et al. 2003b; Potts et al. 2005; Moretti et al. 2009; Chalmandrier et al. 2013; Rodríguez & Kouki 2015), and pollinators use other land- cover types in a complementary fashion (e.g. Westrich 1996). Moreover, many pollinators return to nest sites between foraging bouts to provision young, such that floral resources nearer to nests are of greater value. Lonsdorf et al. (2009) developed a grid-based model accounting for these pollinator characteristics and used it to predict pollinator abundance (the “pollinator source map”), and pollination services (the “pollination services map”), across heterogeneous landscapes. We integrate this with the existing conceptualization of population dynamics in fire-prone landscape to develop our conceptual model (Figure 2.1).

2.3.1. Pollinator source map

Lonsdorf et al. (2009) estimated the abundance of each pollinator species in each cell as the cell’s nesting value multiplied by the sum of floral values in surrounding cells. A cell’s floral value was a function of its distance from the nest cell relative to the pollinator’s foraging range and the land-cover types occupying it, with each type assigned a floral resource value. Fire alters the landscape distribution of pollinator floral and nesting resources (e.g. Potts et al. 2003b), but we are aware of only one study relating pollinator abundance to landscape fire age composition; Rubene, Schroeder and Ranius (2015) found bee and wasp species richness increased with the amount of early post-fire (and harvest) age classes in the landscape.

To link pollinators more explicitly to plant fire responses, we suggest an alternative to land-cover type as the basis for estimating floral values. A plant may be absent from a cell due to cumulative effects of the antecedent fire regime (e.g. local extinction following inter-fire intervals shorter than the time required to replenish propagule stores (Keith 1996; Gill & McCarthy 1998)) and failure to recolonize from neighbouring cells (Bradstock et al. 2005). A plant may be present but not flowering due to inappropriate fire history conditions (see ‘Traits’ section). The value of each cell for each pollinator species then depends on the floral resource quality (e.g. nectar sugar content, pollen) of each plant species they interact with in the cell. Determinants of interactions are discussed in the ‘Cell networks’ section below.

Lonsdorf et al.’s model was developed for relatively static landscapes so further additions are necessary for fire-prone landscapes. Adult pollinators active in a given season developed on the resources of the previous season so there should be a time-lag between fire-driven changes in floral resources in one time-step and changes in pollinator abundance in the next (Potts et al. 2003b). In addition to the indirect effects of fire on pollinators through habitat alteration, exposure to heat and smoke can cause direct effects (Whelan et al. 2002; Dafni, Izhaki & Ne'eman 2012). Rates of post-fire recovery depend on within-cell recruitment, or recolonization through dispersal from neighbouring cells (Watson et al. 2012). Thus a cell’s nesting and floral values need to be multiplied by its pollinator abundance in each time-step to allow modelling of the cumulative effects of changing resources, direct effects, and dispersal across time-steps.

These are poorly understood for pollinators in any frequently disturbed landscape, though Le Féon et al. (2013) found bee abundance and species richness were related to the previous five years of crop rotations in the landscape surrounding sites.

2.3.2. Pollination services map

Lonsdorf et al. (2009) used their pollinator source map to derive a “pollination services map” predicting the numbers of each pollinator species arriving in each cell as a function of the abundances of pollinators residing in and distance (relative to pollinator flight ranges) to surrounding cells. The value of each pollinator species to each plant species in the cell then determines overall visitation. Due to the time-lags and direct fire effects described in the previous section, the pollination services map of a particular time-step should be based on the pollinator source map of the previous time-step minus direct effects occurring during the transition to the current time-step. This accounts for effects on pollination resulting from fire-driven changes in pollinator populations, such as observed lags between the post-fire return to flowering of species and recolonization by their pollinators (Ne'eman, Dafni & Potts 2000). However, it does not account for the plant–plant and pollinator–pollinator interactions that can occur between co-occurring species. These are relevant to both pollination and floral resource intake by pollinators and enter our conceptual model as determinants of the plant–pollinator interaction network that forms in each cell in the pollination services map.

2.3.3. Cell networks

In our model the plant species flowering, and the pollinators residing or arriving, in a given cell in a given time-step form an interaction network described by the presence/absence or frequency of interaction between each plant–pollinator species pair (Jordano 1987; Schleuning, Fründ & García 2015). Pair-wise interaction probabilities vary with matching traits (e.g. tongue length vs. nectar tube depth, phenological overlap (Stang et al. 2009; Junker et al. 2013)) and spatio-temporal variation in plant and pollinator abundance (Carstensen et al. 2014; Kaiser-Bunbury et al. 2014; Schleuning, Fründ & García 2015; Trøjelsgaard et al. 2015). Since fire can drive variation in flower and pollinator abundances (e.g. Potts et al. 2003a) we focus on this aspect of network formation.

The probability that a plant–pollinator species pair interacts in a network may increase with the plant’s floral abundance as it becomes more likely the pollinator will encounter the plant, but also because the plant becomes more attractive to and so influences the foraging decisions of the pollinator (Carstensen et al. 2014). The latter process extends to situations where pollinators perceive different plant species as equivalent when evaluating densities of flower patches such that plants with shared pollinators jointly attract visits (Thomson 1981). However, increasing the abundance of one plant species in a network can increase visitation, but also heterospecific pollen transfer (reduced quality), to other plant species in the network (Lopezaraiza–Mikel et al. 2007) unless there is divergence in pollen placement on pollinators (Armbruster, Edwards & Debevec 1994). Moreover, the effects of floral abundance (conspecific or heterospecific) on visitation to a target plant varies with spatial scale; greater floral abundance at local scales often enhances visitation by concentrating pollinators in the target plant’s patch, but at larger scales reduces visitation by diluting pollinators across neighbouring patches (Veddeler, Klein & Tscharntke 2006; Holzschuh et al. 2011; Hegland 2014; Schmid et al. 2015). Empirical studies of the scales at which pollinators respond to floral densities (e.g. Thomson 1981) might be used to define cell sizes in our model such that plant–plant facilitation occurs within cells, but competition occurs between cells. Within-cell competition is also possible when pollinators shared by plant species are able to distinguish between them and preferentially visit the more rewarding/attractive species (Thomson 1981; Rathcke 1988).

The probability that a plant–pollinator species pair interacts can also increase with the pollinator’s abundance as it expands its foraging niche (increases the range of species visited) to reduce intraspecific competition (Fründ et al. 2013; Trøjelsgaard et al. 2015). Similarly, when pollinator species richness increases a greater range of plant species may be visited, though as a result of individual pollinator species shifting and contracting their foraging niches to avoid interspecific competition by reducing niche overlap (Brosi & Briggs 2013; Fründ et al. 2013). Thus, heterospecific pollen transfer, and so seed set, increases in the case of intraspecific, but decreases in the case of interspecific pollinator–pollinator competition (Brosi & Briggs 2013). Less abundant or rewarding plant species or patches (cells) may nonetheless benefit from both forms of pollinator–pollinator competition as pollinators are forced to visit less attractive

15 resources, though this would reduce cell floral values for competitively inferior pollinators that are more likely to be displaced (Bronstein 1995; Tylianakis et al. 2008; Essenberg 2013).

There is some evidence of fire-driven variation in plant–plant and pollinator–pollinator interactions. Four studies reported variation in pollination associated with fire-driven changes in conspecific and/or heterospecific floral abundance (Potts et al. 2006; Pauw 2007; Van Nuland et al. 2013; Bourg, Gill & McShea 2014). However, only Van Nuland et al. (2013) considered the spatial and temporal scales at which these effects occurred by manipulating floral densities at local scales, finding immediate increased visitation with increased local conspecific floral abundance (which was greater in early age classes). Potts, Dafni and Ne'eman (2001) found evidence that a shrub species differed in reward status, flower visitor species composition, but not pollination services between recently burnt and long unburnt sites because competitively inferior pollinator species were displaced to less rewarding individuals.

2.4. Traits

Here we discuss phenotypic traits for use in parameterizing our model. A trait-based approach is justified because there is intraspecific variation in both fire response traits (e.g. Moreira, Tormo & Pausas 2012) and matching traits (e.g. Poisot, Stouffer & Gravel 2015) such that trait rather than species information should be used in applying the grid-based model to a particular management area. Trait-based models have been developed to predict the consequences of environmental change for plant–pollinator interactions, though fire regimes have not been addressed (Lavorel et al. 2013; Schleuning, Fründ & García 2015). We extend these by identifying traits mediating direct responses to fire and indirect responses through mutualistic partners, in the context of our spatially explicit conceptual model. For pollinators we attempt to cover the full range of parameters required to simulate population dynamics, but for plants we limit discussion to traits relevant to pollination since other variables have been addressed elsewhere (e.g. Noble & Slatyer 1980).

2.4.1. Pollinators – cell fire history

Fire history of the cell can be linked directly to pollinator occupancy/abundance through mortality associated with the fire event. While highly mobile adult pollinators may typically evade fire, most species have a relatively immobile immature life stage that may be vulnerable when in exposed locations during fire events. Below-ground nesting pollinators may experience less mortality than above-ground nesting pollinators during fire because of insulation by soil (Williams et al. 2010; Cane & Neff 2011). Fire severity might moderate these effects since high severity fires can kill below-ground bee eggs and pupae (Cane & Neff 2011). Low severity fires can leave patches of vegetation – and their animal occupants – unburnt (Robinson et al. 2013), such that above-ground nesting species may survive such fires. Species with non-overlapping generations (e.g. univoltine species with short-lived adults) are presumably particularly vulnerable to fires that occur in the season during which the population comprises only immobile immatures. Finally, the number of time-steps taken for abundance of species experiencing direct effects to reach pre-fire levels may be less for multivoltine (and to a

18 lesser extent bivoltine) species, since this trait may assist rapid reproduction and population recovery following disturbance (e.g. Draney & Crossley Jr 1999). Moretti et al. (2009) found early fire age classes dominated by multivoltine bees, though early post-fire conditions might simply be more suitable for mutlivoltinism (e.g. longer flowering periods).

Cell fire history can also be linked to nesting and floral suitability. Associations between early sere and below-ground nesting bees are often attributed to greater amounts of exposed soil, and associations between early sere and above-ground nesting bees to the greater amount of dead wood (nesting substrate) when low severity fire damages rather than consumes vegetation (Potts et al. 2005; Moretti et al. 2009; Williams et al. 2010; Rodríguez & Kouki 2015; Rubene, Schroeder & Ranius 2015). However, these associations could equally be explained by direct effects (see previous paragraph). Floral resources would be linked to fire history when plant traits determining their flowering response to fire (see ‘Plant – cell fire history’ section) were also (or were correlated with) matching traits (i.e. the response–effect framework extended to multi-trophic interactions (Lavorel et al. 2013)). There is some evidence of this in the Mediterranean where early age classes are dominated by open access flowers with concentrated nectar and short-tongued pollinators that are more efficient at extracting this resource, whereas later age classes are dominated by long nectar tube flowers with dilute nectar and long-tongued pollinators that are capable of exploiting these flowers (Potts et al. 2003a; Moretti et al. 2009). Finally, a given reward level (e.g. nectar sugar content) may represent a lower floral value for large-bodied pollinators with greater energetic requirements (e.g. Schmid et al. 2015).

2.4.2. Pollinators – cell composition and configuration

An important variable in the grid-based model is foraging range, which increases with pollinator body size (e.g. Greenleaf et al. 2007; Benjamin, Reilly & Winfree 2014). There is evidence this trait mediates responses to landscape composition and configuration. Large pollinators are often less sensitive to habitat patch size (clumped vs. dispersed habitat configuration) presumably because they are capable of utilizing small habitat fragments (isolated cells) (Jauker et al. 2013; Hopfenmueller, Steffan- Dewenter & Holzschuh 2014), though more sensitive to the amount of habitat (number

19 of suitable cells) in the landscape (composition) presumably due to their greater energy requirements (Larsen, Williams & Kremen 2005).

Dispersal is important for recolonization following direct fire effects or a cell’s return to habitat suitability, but is poorly understood for pollinators. A rough distinction might be made between species that return to nesting sites to provision young (many bees and birds) and species that oviposit freely as they move through the landscape (many pollinating flies), the latter being less constrained in their dispersal movements (Jauker et al. 2009; Parsche, Fründ & Tscharntke 2011).

2.4.3. Plants – cell fire history

The time taken to reach reproductive maturity after fire, following either germination or vegetative regrowth (for resprouting species), and subsequently return to a vegetative or dormant (seed or adult) state, determines the relationships between flowering and fire age. Interspecific differences in the onset and cessation of flowering can correspond approximately to growth form. Trees take longer than shrubs, which take longer than herbs, to reach reproductive maturity following fire (Burrows, Wardell-Johnson & Ward 2008), and taller-stature plants can inhibit flowering of lower-stature plants through competition for light and other resources as their canopies recover from fire (Coates & Duncan 2009; Keeley et al. 2011a). For taller-stature plants, flowering capacity may cease with adult senescence, bud bank attrition, or due to a lack of environmental cues associated with fire (Enright et al. 2011; Lamont & Downes 2011).

A fire’s impact on flowering can depend on its intensity in relation to the plant’s requirements for elevated temperatures and responses to tissue damage. Seeds requiring heat shock to germinate can experience greater germination after more intense fires (Knox & Clarke 2006). Similarly, adult fire-stimulated flowering species might increase flowering with fire intensity where temperature is the flowering cue (Lamont & Downes 2011). Lower-stature (competitively inferior) adults may enhance growth and flowering after more severe fires that increase light (through canopy removal) and nutrient availability (e.g. Bowen & Pate 2004; Knox & Clarke 2006; Lamont & Downes 2011). Flowering of evergreen adults may initially increase with fire severity because of greater damage to dominant apical buds (Bowen & Pate 2004), but excessive damage

20 from high severity fires may reduce flowering in species resprouting from poorly protected above-ground or shallow below-ground buds (Clarke et al. 2013).

The effects of a fire on flowering can also depend on the season in which it occurs. Flowering may be reduced in seeds germinating immediately after fire if the time available for growth between fire/germination and the flowering season is short (Hiers, Wyatt & Mitchell 2000). Conversely, seeds germinating in the first appropriate season (e.g. cued by seasonal temperature regimes) after fire may experience reduced survival and growth, and delay flowering for one or more years, as the time between fire and the appropriate season increases (Knox & Clarke 2006; Ooi 2010). Resprouting species that flower before producing foliage can do so within weeks of fire regardless of season (provided resource reserves have been replenished since the previous fire), such that fire season determines flowering season (Le Maitre & Brown 1992). Resprouting adults needing to produce foliage before or with flowers typically bloom during the growing/flowering season after fire and the effects of fire timing may depend on whether plants are evergreen or deciduous and capable of producing multiple stems. For deciduous, single-stemmed plants (e.g. many geophytes) and evergreen resprouters, the flowering response often increases with the length of time between fire and the flowering season (Le Maitre & Brown 1992; Bowen & Pate 2004; Lamont, Wittkuhn & Korczynskyj 2004). This may occur because there is more time for regrowth and bud maturation prior to flowering (Bowen & Pate 2004), and in the case of dormant geophytes (particularly those that store only enough resources for one season’s growth) greater resource allocation to flowering by avoiding fire damage to foliage (Le Maitre & Brown 1992). For deciduous plants capable of producing multiple stems (e.g. rhizomatous herbs), flowering can be enhanced by fires closer to the flowering season, presumably because exposure of apical stems creates opportunities for release of supressed buds (Platt, Evans & Davis 1988; Hiers, Wyatt & Mitchell 2000). Finally, co- occurring species can flower in different months such that phenological period influences the effects of fire season on flowering (Pavlovic, Leicht-Young & Grundel 2011) and therefore the effects of fire in a given season will vary between species with different phenologies.

2.5. Management recommendations and future research

2.5.1. Immediate actions

Since fire effects on plant–pollinator interactions are poorly understood, few management recommendations can be given. Perhaps the most widely applicable is to manage for flower diversity because it has great potential for conservation benefits and at least rudimentary knowledge of flowering responses to fire is widespread. Floral diversity is positively associated with pollinator diversity because a greater range of foraging niche requirements are met and floral resource availability is stabilized across pollinator activity periods (Blüthgen & Klein 2011; Bennett & Gratton 2013; Dorado & Vázquez 2014). In turn, greater pollinator diversity can enhance and stabilize pollination services through pollinator–pollinator interactions and complementarity due to interspecific variation in traits such as flower height preferences (Hoehn et al. 2008; Blüthgen & Klein 2011; Ebeling, Klein & Tscharntke 2011; Brosi & Briggs 2013).

How floral diversity is enhanced may vary between regions depending on knowledge and management approaches. For instance, in North American prairies knowledge of interspecific differences in flowering response to fire season could be used to generate floral diversity through mixed fire season regimes (Hiers, Wyatt & Mitchell 2000). In southern Australia knowledge of the fire ages in which different species flower could be incorporated into existing optimization procedures that determine fire age mosaic compositions maximizing species diversity (Di Stefano et al. 2013; Kelly et al. 2015). In any case, the aim should be to create floral diversity within pollinator foraging range (while being cognisant of pollinator habitat area requirements) which in turn should generate the local pollinator diversity required for enhancement of pollination services (Tscharntke et al. 2012).

2.5.2. Future research

The landscape ecology approach to empirical research (Wagner & Fortin 2005) is required to understand the effects of fire-generated spatial heterogeneity on pollinator populations and foraging behaviour. Space has so typically been considered only as a substitute for time in studies of fire effects on pollinators and pollination (though see

Ponisio et al. (2016)), despite the fact that pollinators and pollination respond to landscape heterogeneity (Kennedy et al. 2013; Vanbergen 2014) and these responses are an important component of our conceptual model.

Prioritizing species for empirical research might be based on vulnerability under future fire regimes predicted from species traits. Fires are expected to be larger, more severe, more frequent, and occur throughout more of the year across much of the globe under climate change (e.g. Flannigan et al. 2013). This would increase the probability of overlap between severe fire and immobile life stages, reducing abundances particularly of above-ground nesting pollinators with non-overlapping generations, and reduce the amount of time for on-site post-fire population recovery, reducing univoltine pollinators most dramatically.

Consequences of pollination failure for plant persistence can also help prioritize plant species. Bond (1994) suggested these consequences will depend on the plant’s reproductive dependence on pollination (e.g. self-incompatibility) and demographic dependence on seeds (e.g. non-clonality). Carpenter and Recher (1979) suggested a link between reproductive and fire-related traits, whereby non-sprouters are self-compatible in order to ensure high rates of seed set before the next fire. However, some non- sprouters abort all self-sired seeds and instead enhance seed set by successfully attracting pollinators (Lamont, Olesen & Briffa 1998). Resprouters depend less on seed (Bond 1994), but can still be driven to extinction through pollinator decline particularly if non-clonal (Pauw & Bond 2011; Pauw & Hawkins 2011). Simple dichotomies between the consequences of pollination failure for non-sprouters versus resprouters are untenable – reproductive traits need to be considered in addition to fire-response traits.

Ultimately, to guard against extinctions resulting from loss of mutualistic partners (Pauw & Bond 2011), the conceptual framework presented here might be expanded to predict aggregate metrics of plant–pollinator interactions networks such as nestedness (the degree to which specialists interact with generalists) and connectance (number of interactions relative to the number of species). These properties have been linked to the probability of co-extinctions of dependent species (though the nature of the relationships is still being debated (e.g. Vieira & Almeida‐Neto 2015)) and so may be used in addition to or in place of traditional biodiversity measures such as species

23 diversity (Tylianakis et al. 2010; Kaiser-Bunbury & Blüthgen 2015). Ponisio et al. (2016) recently found that plant-pollinator interaction richness increased with the diversity of fire histories within foraging range, though only when the local environment was burnt last by low- or moderate-severity fire. Fire effects on other network properties have not been investigated, though nestedness and connectance have been linked to other forms of landscape diversity (e.g. vegetation types) and local disturbance (e.g. grazing) (Vanbergen et al. 2014; Moreira, Boscolo & Viana 2015). These initial findings suggest that a spatially explicit grid-based model, with each cell’s aggregate network properties determined by its fire history and the fire histories of neighbouring cells, might be developed to predict network properties under alternative fire regimes. Further research aimed at understanding these dynamics could thus enhance the capacity of fire managers to promote ecosystem resilience in the face of diverse environmental changes.

3. FIRE EFFECTS ON POLLINATION IN A SEXUALLY-DECEPTIVE ORCHID

3.1. Abstract

Research into the effectiveness of prescribed fire in managing pollination has only recently begun. The effects of fire on pollination have not been explored in sexually- deceptive systems. Further, the potential for multiple effects operating at different spatial scales have not been explored in any pollination system despite multi-scale effects on pollination observed in agricultural landscapes. We observed the frequency of pollinator visitation to flowers of sexually-deceptive Caladenia tentaculata and related it to the post-fire age class of the vegetation at local- and landscape-scales. We also related the number of the pollinator’s putative larval hosts (scarab beetles) captured at these sites to age class. At the local-scale (i.e. the sample location), visitation was highest in recently burnt sites. At the landscape-scale, positive associations were observed between 1) putative pollinator hosts and vegetation burnt 36-50 years ago, and 2) pollinator visitation and vegetation burnt ≥ 50 years ago. Local- and landscape-scale effects on visitation were synergistic, such that visitation was greatest when fire age was heterogeneous within pollinator foraging range.

3.2. Introduction

Prescribed fire is a widely used tool of conservation and land management (e.g. Bond & Van Wilgen 1996; Penman et al. 2011). The ecological and economic importance of pollination, and recent predictions and observations of pollinator declines (predominantly in Europe and the USA), have motivated researchers to focus attention on the effects of land use and land management practices on pollination (Kremen et al. 2007; Potts et al. 2010). In the context of pollination, fire management has received less attention than other forms of land management, but the small number of studies conducted so far suggest that fire can have a substantial influence on plant-pollinator interactions (Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013; Bourg, Gill & McShea 2014).

Here we explore for the first time relationships between fire and pollinator visitation in a sexually-deceptive pollination system. Sexually-deceptive plants, predominantly terrestrial orchids from Australia and Europe (each with approximately 160 described species), achieve pollination by mimicking the sex pheromone (and potentially other characteristics) of female insects in order to attract male insects (Schiestl et al. 1999; Gaskett 2011). These systems are highly specialised, with the orchid often being pollinated by a single insect species (Peakall et al. 2010; Gaskett 2011). This may contribute to rarity in these orchids by restricting viable populations to areas occupied by the pollinator, necessitating a focus on the ecological requirements of pollinators for the conservation of sexually-deceptive orchids (Brundrett 2007; Swarts & Dixon 2009; Phillips et al. 2014a). Australian sexually-deceptive orchids are pollinated predominantly by parasitoid wasps of the thynnine subfamily (at least 150 orchid species pollinated by 70 thynnine species) (e.g. Stoutamire 1983; Bower 2007; Hopper & Brown 2007; Phillips et al. 2009; Peakall et al. 2010; Gaskett 2011). Theory (Price 1991) and empirical work (Hilszczański et al. 2005; Marino, Landis & Hawkins 2006; Maleque et al. 2010) suggest that many parasitoid wasp groups vary in abundance along post-disturbance successional trajectories in response to changes in host availability. Mateos, Santos and Pujade-Villar (2011) found greater numbers of parasitoid wasp families, along with host individuals, in recently burnt Mediterranean sites. Pollination

26 of sexually-deceptive orchids might thus vary with the availability of successional stages supporting pollinators.

Pollination is influenced by pollinator habitat conditions within flight range of the plant, affecting both pollinator populations and foraging behaviour, but also by more localised factors such as facilitation or competition for pollination with neighbouring plants (e.g. Lammi & Kuitunen 1995; Johnson et al. 2003; Internicola et al. 2006; Internicola et al. 2007; Kremen et al. 2007). For instance, pollination of the Australian sexually- deceptive orchid Caladenia arenicola responded to both landscape-scale (patch size and shape) and local-scale (conspecific abundance and bare ground cover) factors in fragmented landscapes (Newman et al. 2013) (see also Phillips et al. (2014a) for an effect of population size on Drakaea glyptodon pollination). Thus in addition to larger- scale effects on pollinator habitat, local effects of fire on pollination in sexually- deceptive orchids might occur, for instance, through potentially enhanced conspecific flower abundance following appropriately timed fires (Le Maitre & Brown 1992; Jones 2006; Lamont & Downes 2011; Duncan 2012), or by reducing the amount of nearby vegetation structure, enhancing detection by pollinators (Petit & Dickson 2005). The landscape ecology perspective, which includes spatial scale and spatial heterogeneity in analysis (Wagner & Fortin 2005), is required to detect multiple processes operating over different spatial scales. This has rarely been applied in studies of fire effects on pollination to date, most of which (though see Ponisio et al. (2016)) have employed space-for-time substitution designs with fire history considered to be homogeneous at each sample location (i.e. each site has a particular fire age, and the fire age of surrounding vegetation is ignored) (Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013; Bourg, Gill & McShea 2014). If processes operating at different scales exhibit contrasting responses to fire, ignoring fire history heterogeneity could provide misleading conclusions about fire effects on pollination.

Thus we employ a space-for-time substitution design to relate fire age of the sample location to pollinator visitation, but also quantify the proportion of each age class within pollinator flight range of each sample location and relate this to visitation to explore whether relationships differ between spatial scales. We use the same approach to investigate relationships between fire age and the pollinator’s putative larval hosts

27 which are similarly mobile in their adult forms and so potentially responsive to landscape fire age heterogeneity.

3.3. Methodology

3.3.1. Study species

The native Australian orchid Caladenia tentaculata was selected for this study. It is pollinated solely by a parasitoid wasp from the thynnine subfamily (in this case Thynnoides pugionatus) as is typical of several genera of Australian sexually-deceptive orchids (Peakall & Beattie 1996; Phillips et al. 2009; Gaskett 2011). Both C. tentaculata and T. pugionatus are widespread and common in south-east Australia (Jones 2006; Bower 2007), allowing for relatively large sample sizes. The identity and number of larval host species of T. pugionatus have not been recorded, but thynnine wasps typically parasitise the larvae of scarab beetles (family Scarabaeidae) (Ridsdill- Smith 1970; Campbell & Brown 1994; Brown & Phillips 2014).

3.3.2. Study landscape and site selection

The study landscape is an area of approximately 15,000 km2 in south-west Victoria, Australia (Figure 3.1). The area encompasses a number of National Parks and State reserves (including the Grampians National Park) within a predominantly agricultural matrix. The native vegetation is primarily restricted to reserves and consists of sclerophyllous heaths, shrub lands, and woodlands interspersed with open grasslands (Gibbons & Downes 1964; Dodson 2001). The area has a Mediterranean-type climate (hot dry summers and mild wet winters), and has experienced recurring fires throughout the Holocene (review in Dodson 2001). Prescribed fire is applied by land management agencies in an attempt to protect human life and property and achieve ecological objectives (Parks Victoria 2003). The area is suitable for investigation as both the orchid (C. tentaculata) and its pollinator (T. pugionatus) have been detected at a number of localities in this area (Bower 2007), and there is a broad range of fire histories in the region (0-75 years since fire).

Study sites were selected within the ArcGIS environment according to a number of criteria: 1) No non-native vegetation cover within 1 km to minimise the effects of other land cover types, since thynnine wasps have not been observed to travel more than approximately 550m (Menz et al. 2013). 2) Greater than 90% heathy woodland (Ecological Vegetation Class 48 (Cheal 2010)) cover within 1 km. This criteria was used to control for effects of other native vegetation types, with different fire responses and floristics, and because heathy woodland is thought to provide habitat for thynnine wasps (Bower 2007)). 3) No two sites could be closer than 1 km or contain areas of vegetation burnt last by the same fire (to enhance independence). 4) Sites had to be between 100 m and 1,000 m from any road to minimise interference from the public and enhance accessibility. Under these criteria 60 sites were available. Only 52 sites were accessible during 2013 when visitation and scarab data were collected, but all 60 and one additional site were accessible in 2014 when visitation data was collected from an additional nine sites.

3.3.3. Pollinator hosts (scarab beetles)

Potential pollinator hosts (scarab beetles) were sampled to investigate their relationship with post-fire age class. As thynnines parasitise scarab beetle larvae it would be best to sample larvae directly, via soil extraction. However, initial attempts proved inefficient; consequently, adult scarab beetles captured in flight-intercept (malaise) traps were used as an indicator of potential larval abundance. A single flight-intercept trap was placed at each of the 52 sites for three, ten day periods between late September and early November of 2013, corresponding to C. tentaculata’s flowering period in this landscape. Three kinds of traps were used (all using ethylene glycol as the preserving agent) due to limits on trap availability; large and small malaise traps (SLAM traps from Australian Entomological Supplies) and home-made traps similar in dimensions to small malaise traps. Trap types were systematically rotated around sites to reduce effects of trap type on scarab capture. Traps were erected (veins aligned with the cardinal points) at random locations previously chosen in the ArcGIS environment. Traps were only moved from this random point up to 10 m to avoid thickets of tall shrubs and other interferences. Trap contents were later sorted in the laboratory using a binocular microscope and scarab beetles separated and stored in 70% ethanol.

3.3.4. Pollinator visitation

Observations of pollinator visitation to natural populations of C. tentaculata and other sexually-deceptive orchids can be time-consuming (Peakall & Beattie 1996). The preferred approach is to artificially present flowers of the chosen species, because the male wasps deceived into visiting these flowers respond rapidly, peaking within several minutes of presentation (e.g. Stoutamire 1983; Peakall 1990; Peakall & Beattie 1996). Ten C. tentaculata individuals, each with a single flower, were collected from two, large wild populations (these have been lodged at the University of Melbourne Creswick campus). Plants were kept in pots in an attempt to prolong flowering and were used in flower presentations until their flower began to wilt or became damaged.

Flowers were presented at a single location 100 m beyond the trap at each site to reduce interference between traps and flower presentations. They were presented in groups of 4-5 individuals in a consistent spatial pattern of pots clustered tightly together.

Presenting multiple individuals simultaneously was an attempt to minimise changes in attractiveness when flowers were replaced (i.e. through an averaging effect), since there is variation between individual sexually-deceptive orchid flowers in attractiveness to pollinators (Peakall & Beattie 1996). In an attempt to reduce the effects of seasonal variation in wasp activity, presentations were made each time traps were installed and collected (i.e. maximum of 6 between Sept-Nov 2013) at each site, except when raining or below 17°C since thynnine wasps are mostly inactive in these weather conditions (e.g. Bower 2007). Weather permitted one to three presentations per site. Searches within 10 m of each presentation location for flowering conspecifics were made prior to presentation and locations shifted if necessary, because thynnine wasps temporarily learn to avoid the area within 10 m of locations where they have been deceived (Wong, Salzmann & Schiestl 2004). Rain and conspecific flowers prevented any presentations at three sites; consequently data from 49 sites from 2013 and nine sites from 2014 were used in further analysis (there was no significant effect of year on visitation; t = 0.98, p = 0.35). Each presentation lasted 10 minutes and the number of times the flowers were contacted by thynnine wasps within each presentation was recorded. Temperature was also recorded during each presentation using a Kestrel 3000 Pocket Weather Meter and mean values used as a statistical control in analysis because it is known to influence visitation to sexually-deceptive orchids (Bower 2007).

3.3.5. Collection of age class data

Spatial layers in ArcGIS were interrogated to identify the age classes of vegetation at each sample location and the proportion of land covered by each age class within pollinator foraging range of each sample location. Spatial fire history data sourced from the Victorian Department of Environment and Primary Industries was used to determine time since last fire in years for all native vegetation within the study landscape. These data were then categorised into five age classes which corresponded approximately to heathy woodland growth stages as defined by Cheal (2010) based on known vegetation fire responses, though with modifications: the two youngest growth stages were rare so we combined them in a single age class, and the time taken to reach the final growth stage can vary from 45 to 55 years in the study landscape (Cheal 2010) so we used 50+ years for the oldest age class. The age class categories were thus identified in years

31 since fire as: age class 1 (AC1) = 0-3 years (renewal-juvenility), age class 2 (AC2) = 4- 10 years (adolescence), age class 3 (AC3) = 11-35 years (maturity), age class 4 (AC4) = 36-50 years (waning), age class 5 (AC5) = 50+ years (senescence). The proportion of area covered by each age class within 500 m (which approximates the maximum recorded flight range of thynnine wasps (Menz et al. 2013)) of each trap and flower presentation location was determined using the Patch Analyst routine in ArcGIS. These data were then used as predictor variables for exploring relationships between age class and wasp visitation and scarab beetle capture. Ground truthing at the site scale confirmed mapped fire history data was accurate. At larger spatial scales ground truthing was conducted where practical and mapped data was found generally to be accurate though one site mapped as being AC1 contained large patches of unburnt vegetation within 50 m of the trap (see below).

3.3.6. Data analysis

Wasp visitation and scarab beetle capture data were analysed separately. The response variable for wasp visitation was the number of wasp visits recorded during a single flower presentation – the presentation during which the most visits were recorded – at each site (n = 58), because the between-site variation in the number of presentations precluded the use of total visits across all presentations. Wasp visitation data exhibited heteroskedasticity which was overcome with a square root transformation rather than by using Generalised Linear Modelling (GLM) because the data were also over-dispersed and the additional parameter required with adoption of negative binomial models would have detracted from our ability to model more complex responses to fire age. The response variable for scarab capture was the total number of scarab beetles captured at each site and square root transformation was used to overcome heteroskedasticity. The site mapped as AC1 but found to contain unburnt patches was identified as an outlying data point in the scarab capture data set (Cook’s D = 1.7), so was removed for subsequent analyses (leaving n = 51 sites; note that this was also one of the sites not included in visitation analysis due to conspecific presence). All analysis was conducted in the R statistical environment (R Development Core Team 2011).

We first performed ANOVAs with wasp visitation and scarab capture as functions of the age class of sample locations, followed by post-hoc comparisons (with Bonferroni

32 adjustments) if statistically significant (α = 0.05). We then constructed five landscape models for wasp visitation and scarab capture, each with a single predictor corresponding to the proportion of land within 500 m mapped as being in AC1, AC2, AC3, AC4, or AC5. Landscape models were compared in terms of their evidence ratios relative to a null model with an intercept only (i.e. Akaike weight of landscape model/Akaike weight of null model (Burnham & Anderson 2002)), their parameter estimates, and their R-squared values. Finally, the effects on wasp visitation of fire age at the sample location and at the landscape scale differed, so we investigated whether they were additive or interactive by constructing two additional models; both of which had local scale effects represented by a binary predictor that was one when a sample was in the age class with the greatest number of visits and 0 when a sample was in any other age class (a categorical predictor with all five levels would have made interactive models excessively complex for our sample size, and we believe a binary predictor is justified by the ANOVA results), and landscape scale effects represented by the proportion of land covered by the age class with the strongest positive relationship with visitation (see results section), but differed in being an additive vs. interactive combination of these predictors. These models were compared by calculating their evidence ratio (Akaike weight of interaction model/Akaike weight of additive model) and (adjusted) R-squared was calculated for the best model.

We assessed multi-colinearity with tolerance values (Quinn & Keough 2002), which were never greater than 1.6 (indicating no significant multi-colinearity). We also assessed spatial autocorrelation with spline correlograms (using R packed “ncf”) and found no evidence for wasp visitation or scarab capture data (95% confidence intervals for Moran’s I statistic overlapped zero at all scales).

3.4. Results

Scarab capture was weakly related to the post-fire age classes defined in this study. Both the ANOVA relating scarab capture to the fire age of the sample location (Figure 3.2A) and the landscape model comparison (Table 3.1) revealed weak, positive

33 associations with AC4 (though this was not statistically significant at α = 0.05 for the ANOVA; F = 2.01, p = 0.11). Ten morphospecies were identified.

Wasp visitation was related to the age class of the sample location and age class heterogeneity within pollinator flight range, though the age class with greatest effect differed between scales. The ANOVA was statistically significant at α = 0.05 (F = 2.86, p = 0.03). Visitation was highest when the sample location was AC1, though not significantly so relative to AC5 (Figure 3.2B). The landscape model comparison revealed the linear model of wasp visitation as a function of the proportion of land covered by AC5 provided the best explanation (Table 3.1). For the comparison of additive and interactive combinations of local and landscape-scale effects, the binary local-scale predictor was one when a sample was in AC1 and zero when a sample was in any other age class, and the landscape-scale predictor was the proportion of land covered by AC5. The interaction model was 249 times more likely to be the best model than the additive model (Akaike weight of interaction model = 0.996, Akaike weight of additive model = 0.004) and explained 40% of the variation in wasp visitation (adjusted R-squared = 0.40). Figure 3.3 shows wasp visitation increases with the proportion of land in AC5 for sample locations in AC1, but not for locations in AC2-5.

Table 3.1: evidence ratios relative to the null (Akaike weight of landscape model/Akaike weight of null model), parameter estimates (with p-value), and R-squared values for single- predictor models of scarab capture and wasp visitation as functions of the proportion of land covered by AC1-5 within 500 m of sample locations.

Scarabs captured AC1 AC2 AC3 AC4 AC5 Evidence ratio 0.45 0.55 0.75 21.70 0.55 Parameter estimate 0.17 (0.41) -0.16 (0.32) -0.2 (0.20) 0.49 (0.00) -0.19 (0.33) R-squared 0.02 0.02 0.03 0.17 0.02 Wasp visitation AC1 AC2 AC3 AC4 AC5 Evidence ratio 1.74 6.44 1.56 0.33 14.55 Parameter estimate 0.98 (0.07) -1.19 (0.02) -0.77 (0.08) 0.08 (0.88) 1.39 (0.01) R-squared 0.06 0.10 0.05 0.01 0.13

3.5. Discussion

Pollinator visitation to sexually-deceptive C. tentaculata was positively associated with the oldest (AC5) and youngest (AC1) post-fire age classes, though with the oldest age class more so at the scale corresponding to pollinator flight ranges and with the youngest age class more so at smaller spatial scales (i.e. the sample location). Thus there appear to be at least two processes, operating at different spatial scales, through which fire can influence visitation in our study system. Moreover, these processes appear to interact. Between-scale interactive effects on pollinators and pollination have been observed in agricultural landscapes where beneficial farming practices (e.g. organic farming) compensate for low levels of pollinator habitat (semi-natural vegetation) in the surrounding landscape, and conversely landscapes with high levels of pollinator habitat compensate farming practices harmful to pollinators such as pesticide application (Tscharntke et al. 2012; Kennedy et al. 2013; Park et al. 2015). In contrast, our results suggest that local and landscape processes interact not in a compensatory fashion, but synergistically. The benefits of recent fire at the local scale (the mechanism by which this occurs is currently uncertain) are strongest when there is long-unburnt vegetation, which we argue below acts as pollinator habitat, in the surrounding landscape. Since we only recorded visitation, future research should investigate whether the observed synergism emerging from fire age heterogeneity within pollinator flight range translates into enhanced reproductive output and ultimately population persistence.

Further research is needed to determine the local-scale mechanism(s) by which visitation to orchids flowering in the recently burnt environment might be enhanced. One possibility that deserves further attention is that conspecific flower abundance is greater in the recently burnt environment and their collective odour plume attracts male wasps from greater distances (the distance at which an odour plume indicating the presence of a mate is detected by an insect increases with the number of odour sources (Andersson, Löfstedt & Hambäck 2013)). There is mostly anecdotal evidence that C. tentaculata and other sexually-deceptive orchids can exhibit enhanced flowering in the recently burnt environment (and there appeared to be greater numbers of C. tentaculata flowers in recently burnt sites in the present study, though the reduced vegetation

36 structure may simply have made them more visible); however responses are highly variable and fire can be detrimental to flowering (Coates & Duncan 2009; Duncan 2012). Some of this variability may result from seasonal timing of fire relative to annual growth cycles, with beneficial effects being restricted to fires occurring between the replenishment of carbohydrate reserves toward the end of one growing period (around the time of flowering and seed production (Pate & Dixon 1982)) and leaf emergence in the following growing period (Le Maitre & Brown 1992; Jasinge 2014). Temperate Caladenia species generally flower and set seed from spring to summer (September- November for C. tentaculata in our study landscape; personal observation), return to dormancy, and then re-emerge in late autumn-early winter (Dixon & Tremblay 2009). Recent fires (AC1) in our study occurred predominantly in spring (during flowering/seed set), with some in late summer-mid autumn (during dormancy), and so may not have been detrimental to flowering in subsequent years though we did not monitor dormancy-emergence-flowering transition probabilities. Variable flowering responses may also result from variation in the extent to which fire alleviates competition for light, water, or nutrients be removing surrounding vegetative biomass, driven by site-specific conditions such as vegetation community composition (Coates & Duncan 2009) or perhaps fire severity (which is correlated with fire season (Whelan 1995)). Further research is required to understand the effects of fire season on the flowering of terrestrial orchids.

The observed relationship at the larger scale is significant for the conservation of many sexually-deceptive orchids in Australia, which must focus on the ecological requirements of thynnine pollinators (Brundrett 2007; Swarts & Dixon 2009; Phillips et al. 2014a). AC4 (36-50 years since fire) was the only age class to exhibit a positive relationship with the number of scarab beetles (potential thynnine hosts) captured. This age class is characterised by a number of features (Cheal 2010) that could influence scarab abundance such as relatively low soil moisture (which can influence scarab larva mortality (e.g. Davidson, Wiseman & Wolfe 1972; Hassan 1975)) and low levels of bare ground (scarabs are known to avoid ovipositing into bare ground (e.g. Kelsey 1968; Szendrei & Isaacs 2006)). AC5 (50+ years since fire) had the strongest positive effect on wasp visitation. This pattern is consistent with the notion of parasitoid wasps entering succession following establishment of host populations (in AC4 in this case)

(Price 1991). The positive association with AC5 for wasps but not scarabs could be explained by stabilisation of parasitoid-host interactions with host abundance held low at equilibrium (Murdoch, Briggs & Swarbrick 2005) in AC5, causing a reduction in the number of scarabs emerging as adults (and therefore being captured in our traps). These results suggest that long-unburnt vegetation is important for T. pugionatus, potentially as the primary source of its hosts though further work is required (e.g. observing levels of thynnine parasitism of scarab larvae along a post-fire chronosequence).

These results may be relevant to other species of thynnine since members of this subfamily typically parasitises scarab larvae (Campbell & Brown 1994; Brown & Phillips 2014). Thynnine species may be vulnerable to increasing fire frequency resulting from fire management practices and climatic changes (Williams et al. 2009; Cary et al. 2012; Stephens et al. 2012; Flannigan et al. 2013) decreasing the availability of long-unburnt vegetation in southern Australian landscapes. Rare sexually-deceptive orchids with distributions limited by the availability of thynnine wasps (Phillips et al. 2014a; Phillips et al. 2015) may thus become rarer in these landscapes.

Our findings add to a growing body of evidence of fire effects on pollination (Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013), though for the first time in a sexually-deceptive pollination system. These findings are significant in the context of declining pollinators and pollination services, since there is often geographic overlap of fire management with these declines (e.g. Pauw & Bond 2011; Burkle, Marlin & Knight 2013). While long term monitoring of pollinator communities and pollination services is rare in Australia (though see Reiter et al. (2016)), there is evidence of anthropogenic declines in pollination and reproductive output of some Australian native plant species (e.g. Cunningham 2000). Should declining pollination services be recognised as a serious conservation issue in Australia, the results of this study suggest that fire-pollination interactions may become an important avenue of research which should consider multiple processes operating at different spatial scales.

4. FIRE, FOOD, AND SEXUAL- DECEPTION IN THE NEIGHBOURHOOD OF VICTORIAN ORCHIDS

4.1. Abstract

The effective use of prescribed fire in biodiversity conservation is currently inhibited by a limited understanding of fire effects on ecosystem processes such as pollination. Orchids inhabiting fire-prone landscapes are likely to be particularly sensitive because they often exhibit highly specialised pollination systems and provide no reward to pollinators, making them dependent on co-flowering heterospecifics to attract and support pollinators. We investigated the hypothesis that fire-driven changes in the local abundance of rewarding heterospecific flowers influence pollination in two rewardless Australian orchid species, Diuris maculata sensu lato and Caladenia tentaculata. Diuris maculata s.l. is thought to achieve pollination by mimicking papilionoid Fabaceae flowers. Caladenia tentaculata attracts male thynnine wasps through sexual deceit, and these wasps forage on the open-access flowers of other taxa. We used a space-for-time substitution design with sites in different stages of post-fire succession where we recorded capsule set in D. maculata s.l., pollinator visitation to Caladenia tentaculata, the floral abundance of rewarding heterospecifics, and abiotic conditions. Many rewarding taxa responded to fire age, but there was only weak evidence that capsule set in D. maculata s.l. was positively related to the local floral abundance of rewarding

39 species. There was evidence of an overriding effect of rainfall on capsule set that may have obscured effects of the floral community. Visitation to C. tentaculata was not positively associated with any rewarding heterospecifics, and was negatively associated with rewarding Burchardia umbellata. Our preliminary findings highlight the need to account for multiple factors when trying to detect fire effects on pollination.

4.2. Introduction

Prescribed fire is a widely used tool in biodiversity conservation and land management (e.g. Bond and Van Wilgen 1996; Penman et al. 2011). Substantial knowledge gaps currently limiting its effectiveness in conservation include interactions with ecosystem processes such as herbivory and pollination that influence plant survival and reproduction (Driscoll et al. 2010). There is a growing body of evidence from fire-prone regions of North America, South Africa, and the Mediterranean Basin suggesting fire can substantially influence pollination (Ne'eman, Dafni & Potts 2000; Potts, Dafni & Ne'eman 2001; Potts et al. 2006; Pauw 2007; Geerts, Malherbe & Pauw 2012; Van Nuland et al. 2013; Bourg, Gill & McShea 2014; Brown et al. 2016). The influence of fire on pollination in other fire-prone regions such as Australia have been poorly studied (although see (Brown, York & Christie 2016)).

Orchids inhabiting the fire-prone landscapes of southern Australia and South Africa might especially benefit from fire management of pollination services. These orchids generally form highly specialised relationships with pollinators that make them vulnerable to loss of pollination services (Brundrett 2007; Pauw 2007; Phillips et al. 2009; Swarts & Dixon 2009; Pauw & Bond 2011; Phillips et al. 2014a; Phillips et al. 2015). Many are geophytes that respond to hot summer fires with enhanced flowering (Lamont & Downes 2011; Duncan 2012), which has led to the notion that burning (in the appropriate season) to promote flowering and attract pollinators benefits orchids (Cropper & Calder 1990; Coates, Lunt & Tremblay 2006; Coates & Duncan 2009). This was recently supported by Pauw (2007) who found that flowering and pollination of the South African orchid Pterygodium catholicum declined as post-fire succession proceeded.

The diversity and complexity of orchid pollination systems may limit the generality of fire management recommendations from studies of individual systems like that of Pauw (2007). Pterygodium catholicum rewards its pollinators with oil, such that when fire enhances P. catholicum flowering it also enhances pollinator resources (Pauw 2007). However, approximately 6,500 orchid species are rewardless (Jersáková, Johnson & Kindlmann 2006), including many fire-stimulated flowering species from the large genera Diuris, Caladenia, Thelymitra, and Disa (Beardsell et al. 1986; Bernhardt & Burns-Balogh 1986; Dafni & Calder 1987; Dafni & Bernhardt 1990; Johnson, Linder & Steiner 1998; Indsto et al. 2006; Phillips et al. 2009). Rewardless species necessarily depend on rewarding species to support pollinator populations, and may also benefit from enhanced local abundance of pollinators foraging on nearby rewarding plants (the magnet species effect (Laverty 1992)). Rewardless species that mimic rewarding (model) species can also experience enhanced pollination in the presence of the model as pollinators learn to associate rewards with floral traits shared by the model and mimic (Renner 2006; Newman, Anderson & Johnson 2012). Fire effects on the pollination of rewardless orchids are poorly understood. There is anecdotal evidence that the rewardless Australian orchid Diuris maculata sensu lato experiences reduced pollination in the early post-fire environment where it flowers more profusely but the rewarding flowers utilised by its pollinators are scarce (Beardsell et al. 1986). There is also evidence that pollinator visitation to the rewardless Australian orchid Caladenia tentaculata is enhanced in recently burnt and long unburnt vegetation relative to mid- succession (Brown, York & Christie 2016), though the mechanism underlying this pattern (e.g. fire-driven changes in the local floral community) is unknown.

Here we present preliminary findings from field observations of two Australian orchid species, from the genera Diuris and Caladenia, as representatives of the two most common strategies used by rewardless orchids to attract pollinators; food-deception and sexual-deception (Jersáková, Johnson & Kindlmann 2006). Food-deception and fire- stimulated flowering are thought to be widespread in the Australian genus Diuris comprising 60-100 species (Dafni & Bernhardt 1990; Indsto et al. 2006; Duncan 2012). Two studies (Beardsell et al. 1986; Indsto et al. 2006) of D. maculata s.l. (including D. maculata sensu stricto from New South Wales and D. pardina from Victoria) found it to 1) be nectarless (although only D. maculata s.s. was tested), 2) share pollinators

(Leioproctus and Trichocolletes bees, and a Gasterupation wasp) with rewarding Daviesia and Pultenea (Fabaceae) species, and 3) share floral morphology (keel flowers) and spectral reflectance (UV nectar guides and similar colour in the bee visual system) with ‘egg and bacon’ peas which are papilionoid Fabaceae species in tribes Mirbelieae (including Daviesia and Pultenea) and Bossiaeeae. Both studies concluded that rewardless D. maculata s.l. (and many other Diuris species) mimic the flowers of egg and bacon peas (or Papilionoideae more broadly). Egg and bacon peas often do not commence flowering until three or more years after fire (e.g. Benson 1985; Knox & Clarke 2004; Burrows, Wardell-Johnson & Ward 2008). It is thus possible that D. maculata s.l. pollination is influenced by the availability of rewarding pea flowers nearby to support pollinator populations, enhance local pollinator abundance, and/or condition pollinators to associate floral signals with rewards. We investigated the hypothesis that fire-driven change in the local availability of flowers of particular egg and bacon pea species, genera, tribe, or the Papilionoideae sub-family will influence D. maculata s.l. pollination.

Sexual-deception and fire-stimulated flowering are common traits among the more than 370 species comprising the Australian genus Caladenia (Phillips et al. 2009; Gaskett 2011; Duncan 2012), including the widespread and common C. tentaculata that is the second focus of this paper. These orchids mimic female wasps of the sub-family Thynnidae to exploit the mate-seeking behaviours of male wasps for pollination (Gaskett 2011). The dependence of these wasps on nectar for completion of the life- cycle suggests the importance of rewarding species to sexually-deceptive orchids for pollinator support (Phillips et al. 2009). While sexually-deceptive orchids are unlikely to benefit from conditioning of pollinators to associate floral traits with food rewards (i.e. since pollinators visit these orchids for sex not food), magnet species effects are possible though have not been explored. Wasps generally are restricted by short mouthparts to collecting easily accessible nectar (Willmer 2011). Accordingly, thynnine wasps have been observed feeding predominantly on open-access flowers of Leptospermum, Eucalyptus, Chamelaucium, Melaleuca (Myrtaceae), Hakea (Proteaceae), and Xanthorrhoea (Xanthorrhoeaceae) (Brown & Phillips 2014). Flowers of these taxa are typically scarce in the early post-fire environment (e.g. Benson 1985; Enright & Goldblum 1999; Burrows, Wardell-Johnson & Ward 2008), though

Xanthorrhoea and other open-access nectariferous species are abundant (e.g. fire- stimulated flowering Burchardia umbellata (Ramsey & Vaughton 2000; Lamont & Downes 2011)) and so may support and attract thynnines while major food sources recover. We investigate the hypothesis that pollinator visitation to C. tentaculata is positively associated with fire-driven changes in the local abundance of open-access nectariferous species.

4.3. Methodology

4.3.1. Study landscape and site selection

The study landscape is an area of approximately 15,000 km2 in south-west Victoria, Australia (Figure 4.1). The area encompasses a number of National Parks and State reserves (including the Grampians National Park) within a predominantly agricultural matrix. The native vegetation is primarily restricted to reserves and consists of sclerophyllous heaths, shrub lands, and woodlands interspersed with open grasslands (Gibbons & Downes 1964; Dodson 2001). The area has a Mediterranean-type climate (hot dry summers and mild wet winters), and has experienced recurring fires throughout the Holocene (reviewed in Dodson 2001). Prescribed fire is applied by land management agencies in an attempt to protect human life and property and achieve ecological objectives (Parks Victoria 2003).

Figure 4.1: The study landscape. Broken line indicates Victoria-South Australia border. Grey areas represent native vegetation and white areas represent agricultural/residential areas. The insert shows the location of the study landscape (black rectangle) in Victoria (grey area). Study sites were selected within the ArcGIS environment according to a number of criteria: 1) mapped as heathy woodland (Ecological Vegetation Class 48), to control for effects of other native vegetation types with different fire responses and floristics. 2) To enhance independence no two sites could contain areas of vegetation burnt last by the same fire. 3) Sites had to be between 50 m and 150 m from any road to reduce edge effects and enhance accessibility. 4) Collectively sites covered a broad range of time- since-fire (spatial fire history data sourced from Victorian Department of Environment and Primary Industries and ground truthing confirmed mapped fire history was accurate) and so represented a chronosequence (1-75 years). 5) Not to be burnt by land management agencies during the current prescribed burning season. 6) At least one of the study species or a species from the same genus had to be detected during initial site visits (see data collection section below) to ensure that the vegetation being sampled was representative of the local floral community experienced by geophytic orchids. Under these criteria (and accessibility issues) 41 sites were used.

4.3.2. Data collection

At each site searches were conducted for C. tentaculata, D. maculata s.l., and related species between 50 m and 150 m from the road/track until: 1) one or more of

44 either species was detected or 2) 10 mins had elapsed (only after one or more study species had been detected at the site). A 20 x 20 m quadrat was then set up centred on the first orchid detected or at the location where time elapsed. This procedure was continued (while ensuring that quadrats were greater than 40 m apart) until three quadrats had been set up within the site. Up to three flowering individuals (depending on availability) of each species were tagged (using white plastic tags inserted into the ground approximately 5 cm from the flower and labelled with the species and A, B, or C) within quadrats. The flower community was surveyed twice in each quadrat, with 10-15 days between repeat visits, from late September until early November of 2014. During each survey the number of flowering individuals of every native species was recorded (no attempt was made to distinguish ramets from genets), and the number of flowers per individual was estimated for each species by counting and averaging the number of flowers on up to three individuals in each quadrat. Tagged orchids were checked for capsule set during flower community surveys and then once more in December (only one plant per quadrat was used for statistical analysis; plant A if it had not been removed through grazing, plant B if A was removed and plant C if both A and B were removed).

Low rainfall can limit capsule set in Diuris species (Indsto et al. 2006) and rainfall varied across the landscape over which capsule set was observed. Interpolated rainfall data were thus obtained from the Queensland Government Department of Science, Information Technology and Innovation’s Scientific Information for Land Owners database (https://www.longpaddock.qld.gov.au/silo/about.html accessed 15/09/2016) to account for rainfall-driven spatial variation in capsule set during statistical analysis. Eight rainfall variables were calculated: total rainfall (mm) in November 2014, October 2014, September 2014, spring 2014, winter 2014, autumn 2014, the six months prior to sampling, and the 12 months prior to sampling.

Insufficient numbers of C. tentaculata individuals were detected during initial site visits so an alternative sampling technique was employed. Observations of visitation to artificially presented sexually-deceptive flowers (baiting) is efficient

45 because the male thynnine wasps deceived into visiting these flowers respond rapidly, peaking within several minutes of presentation (e.g. Peakall 1990; Peakall & Beattie 1996). Ten C. tentaculata individuals, each with a single flower, were collected from two, large wild populations. Plants were kept in pots in an attempt to prolong flowering and were used in flower presentations until their flower began to wilt or became damaged. Flowers were presented at the centre of each of the previously established quadrats (see above) within a sub-set of sites (n = 20) including only those that 1) were surveyed on clear days with temperatures exceeding 17°C since thynnine wasp activity is sensitive to these weather conditions (e.g. Bower 2007), and 2) did not contain naturally occurring C. tentaculata flowers (to avoid intraspecific interactions between bait and natural flowers). Flowers were presented in groups of four individuals in a consistent spatial pattern of pots clustered tightly together. Presenting multiple individuals simultaneously was an attempt to minimise changes in attractiveness when flowers were replaced (i.e. through an averaging effect), since there is variation among individual sexually-deceptive orchid flowers in attractiveness to pollinators (Peakall & Beattie 1996). Searches within 10 m of each presentation location for flowering conspecifics were made prior to presentation and locations shifted if necessary, because thynnine wasps temporarily learn to avoid the area within approximately 10 m of locations where they have been deceived (Wong, Salzmann & Schiestl 2004). Each presentation lasted 10 minutes and the number of times the flowers were contacted by thynnine wasps within each presentation was recorded. Temperature was also recorded during each presentation using a Kestrel 3000 Pocket Weather Meter and mean values used as a statistical control in analysis.

4.3.3. Data analysis

All analysis was conducted in the R statistical environment (R Core Team 2014) using the MuMIn package (Barton 2014) to compare models in an information- theoretic framework (Burnham & Anderson 2002), and followed protocols for assessing spatial autocorrelation and determining the optimal structure for

46 random model components outlined in (Zuur et al. 2009). General linear models were used to model fire effects on floral abundance (since values were non- integers; see below) and square root transformations were required to remove heteroscedasticity and normalise error distributions. Generalised Linear Models (GLM) were used to model capsule set in each quadrat as a binary response (1= yes, 0 = no) for D. maculata s.l. (only quadrats containing tagged D. maculata s.l. plants that flowered and either produced capsules or became desiccated at the end of the monitoring period without producing capsules were used in analysis) and number of visits in each quadrat as the response (negative binomial models were required to account for over-dispersion) for C. tentaculata. All candidate model sets are described in Table S1. Models with multiple predictor variables were found to have low levels of multi-collinearity (variance inflation factors < 2). Partial regression plots for multiple regressions were produced using the visreg package (Breheny & Burchett 2012).

Flowering species entered models individually, and in combination (i.e. as a single predictor) at various taxonomic levels under the assumption that pollinators experience combined species equivalently (in a sensory, cognitive, and behavioural sense (e.g. Thomson 1981)). Thus floral variables relevant to D. maculata s.l. included: 1) Dillwynia glaberrima, 2) Dillwynia sericeae, 3) Dillwynia species combined, 4) Pultenaea species combined, and 5) tribe Mirbelieae (which at the study sites included Aotus, Gompholobium, and Daviesia in addition to Dillwynia and Pultenaea), 6) tribe Bossiaeeae (which at the study sites was only Platylobium obtusangulum), 7) egg and bacon peas (Mirbelieae and Bossiaeeae combined), and 8) sub-family Papilionoideae (egg and bacon peas plus Kennedia prostrata). Floral variables relevant to C. tentaculata (all open-access, nectariferous species present at study sites) included 1) Leptospermum myrsinoides (Myrtaceae), 2) Burchardia umbellata (Colchicaceae), 3) Microseris sp. (Asteraceae), and 4) these species combined with several Hakea (Proteaceae) species and Xanthorrhoea australis (Xanthorrheaceae) which occurred at less than 10 sites and so were not modelled individually. (Note that if there was no literature describing nectar production and

47 floral morphology for species detected in the present study it was assumed they were nectarless and so not included in analysis).

While some herbaceous species produced a similar, relatively small number of flowers per individual regardless of fire age, some shrubby species produced numerous flowers per individual that varied with fire age (i.e. as stem growth occurred along post-fire succession). Pollinators therefore may not have experienced individuals of different taxa and/or in different stages of post-fire recovery as equivalent floral sources. Thus two floral abundance measures were compared for each taxon in each quadrat: 1) the number of flowering individuals, and 2) the number of flowering individuals multiplied by the number of flowers per individual. However, results are presented only for the number of flowering individuals as preliminary analyses demonstrated that results did not vary substantially between floral abundance measures 1 and 2 (although the responses of measure 2 to fire were more exaggerated).

The number of individuals of each floral predictor variable was averaged across quadrats (using the maximum of the two surveys for each quadrat) to obtain a single value for each predictor for each site. Between-site variation was then modelled as a function of fire history. The spatial time-since-fire data (see ‘Study landscape and site selection’ above) were categorised into four age classes which corresponded approximately to heathy woodland growth stages as defined by (Cheal 2010) based on known vegetation fire responses, though with modifications: the two youngest growth stages were rare so we combined them in a single age class, and the oldest age class was rare so it was combined with the second oldest. The age class (AC) categories were thus identified in years since fire as: AC1 (renewal-juvenility) = 0-3 years (n = 13 sites), AC2 (adolescence) = 4-10 years (n = 10 sites), AC3 (maturity) = 11-35 years (n = 12 sites), AC4 (waning- senescence) = 36+ years (n = 6 sites). The same spatial fire history data layer was used to extract for each site the recorded fire frequency (continuous variable with range = 0-5 fires) and minimum inter-fire interval (continuous variable with range = 1-50 years between fire) to be used as predictors since these variables are known to influence population sizes of some of the taxa used in this study (Duff

2010). A separate candidate model set was constructed for each floral variable, consisting of a model containing all single-predictor models, age class and fire frequency, age class and minimum inter-fire interval (the pairwise correlation between fire frequency and minimum inter-fire interval was 0.71 so they were not included in the same model), and a null model with intercept only. These models were compared using AICc (lowest AICc indicates the best model), evidence ratios relative to a null model with intercept only, and R2, and p-values for hypothesis tests were calculated.

The spatially nested sampling design (quadrats within sites) necessitated assessment of spatial autocorrelation. This was performed for each orchid species by inspection of spline correlograms (with 95% pointwise bootstrap confidence intervals) using residuals from the global GLM containing all predictor variables that did not cause multi-collinearity (variance inflation factors <10, see Table S1) with a maximum lag distance of 5,000 m (using the R-package ncf (Bjornstad 2013)). Where spatial autocorrelation was detected, a Generalised Linear Mixed Model (GLMM) with a random effect for site added to the global model was fitted (using R-package glmmADMB for the range of error distributions it allows (Fournier et al. 2012)), assessed for spatial autocorrelation as with the GLM, and then compared to the full GLM (i.e. without the random effect) using Akaike Information Criterion for small sample sizes (AICc) and evidence ratios (Akaike weight of GLMM/Akaike weight of GLM).

Next we assessed the importance of rainfall for D. maculata s.l. capsule set and temperature for C. tentaculata visitation. For capsule set we compared the AICc of eight, single-predictor rainfall models (i.e. one model for each of the eight rainfall variables) to determine the best rainfall predictor, and then calculated the evidence ratio of the best model compared to a null model with intercept only. For visitation we calculated the evidence ratio for a model containing temperature only compared to a null model with a random effect for site (because we detected an effect of site; see ‘Results’ below). If the model containing the environmental predictor was at least twice as likely as the null to be the best model, and the slope parameter was statistically significantly (α = 0.05) different from zero, it was 49 included in all subsequent models as a statistical control and was used as the null model (i.e. model with environmental predictor only) for further comparison.

We then constructed a candidate set of ten D. maculata s.l. capsule set models (corresponding to the eight rewarding flower predictors, conspecific abundance, and a null model) and a candidate set of five C. tentaculata visitation models (corresponding to the four rewarding flower predictors plus a null model). Models were compared within candidate sets using AICc, explained deviance (D2), and evidence ratios.

4.4. Results

4.4.1. Flowering and fire history

For rewarding species thought to share pollinators with D. maculata s.l., fire history models were better than the null for all floral variables except Bossiaeeae and D. glaberrima, and fire history explained a moderate amount of variation (21- 42%) (Table 1). The number of flowering individuals for all taxonomic groupings except Bossiaeeae was higher in AC2 and AC3 compared to AC1 (though the differences were not statistically significant for Bossiaeeae, Pultenea and D. glaberrima), lower in AC4 compared to AC1 (though the difference was not statistically significant for any taxa) (Table 4.1; Figure 4.2a). Flowering individuals increased with minimum inter-fire interval for most taxa, though only statistically significantly so for Papilionoideae and D. sericea. The number of flowering D. maculata s.l. individuals was highest in AC1 and increased with minimum inter-fire interval, though neither effect was statistically significant.

For rewarding species thought to share pollinators with C. tentaculata, fire history models were better than the null for all floral variables except Microseris sp. and explained a moderate amount of variation (31-37%) (Table 1). Floral abundance for L. myrsinoides and all species combined (dominated by L. myrsinoides and Hakeae species) was higher in AC2, AC3, and AC4 compared to AC1 (Figure 1b),

50 and increased with fire frequency and minimum inter-fire interval (Table 1), but for B. umbellata was lower in AC2 and AC3 compared to AC1 (Figure 1b) and increased with fire frequency but not minimum inter-fire interval (Table 1).

Figure 4.2: the mean (with standard error) number of flowering individuals of rewarding taxa thought to share pollinators with a) Diuris maculata (to enhance clarity only Dillwynia glaberrima, Pulteneae, and Bossiaeeae are shown as they demonstrate the range of between-taxa variation in flowering responses to fire), and b) Caladenia tentaculata (for clarity the combination of all species is not shown as it is qualitatively similar to the Leptospermum myrsinoides response only larger). (AC1 n = 13 sites, AC2 n = 10 sites, AC3 n = 12, AC4 n = 6 sites).

4.4.2. Diuris maculata s.l.

Spline correlograms did not indicate statistically significant spatial autocorrelation (95% confidence intervals for Moran’s I statistic overlapped zero at all scales) for D. maculata s.l. capsule set (Figure 4.3a), so no random effect was included. The best rainfall model contained total rainfall during winter 2014, though models containing rainfall during October, autumn, and the previous 6 and 12 months also had substantial support (and pairwise correlations between these predictors ranged from 0.75-0.95). The model containing winter rainfall was 90 times more likely to be the best model (i.e. evidence ratio = 90) compared to the null, the p-value for the significance test that the slope parameter equals zero was 0.006, and deviance explained was 34%. The evidence that winter rainfall influenced capsule set justified its inclusion in all subsequent models and use as the null for subsequent comparisons. The best model – and the only model with substantial support – contained winter rainfall plus the number of flowering D. glaberrima individuals and had an explained deviance of 52% (Table 4.2). The evidence

51 ratio for this model compared with the model containing only winter rainfall, which was the third best after the model containing the number of flowering D. maculata s.l. individuals, was 7.75. However, the effect of D. glaberrima on capsule set was not statistically significantly different from zero (at α = 0.05). Capsule set was positively associated with both D. glaberrima (Figure 4.4a) and winter rainfall (Figure 4.4b).

Figure 4.3: spatial autocorrelation (Moran’s I, with 95% confidence intervals) at a range of distances for the residuals of the a) D. maculata capsule set model with all (minimally correlated) environmental predictors, b) C. tentaculata visitation model with all (minimally correlated) environmental predictors, and c) C. tentaculata visitation model with all environmental predictors plus a random effect for site.

Response Best model ER null R² AC2 AC3 AC4 FF min Flower community for D. maculata Papilionoideae AC + min 2.10 0.25 2.91 (p = 0.01) 2.19 (p = 0.04) -0.75 (p = 0.60) NA 0.05 (p = 0.05) egg and bacon AC + min 2.86 0.26 3.18 (p = 0.01) 2.52 (p = 0.02) -0.16 (p = 0.91) NA 0.05 (p = 0.07) Mirbelieae AC + min 28.81 0.34 3.69 (p = <0.01) 2.41 (p = 0.01) -0.19 (p = 0.88) NA 0.04 (p = 0.08) Bossiaeeae NULL NA NA NA NA NA NA NA Pultenaea FF 1.02 0.06 NA NA NA 0.28 (p = 0.13) NA Dillwynia AC + min 43.00 0.35 3.29 (p = 0.00) 2.00 (p = 0.02) -0.13 (p = 0.91) NA 0.04 (p = 0.06) D. glaberrima NULL NA NA NA NA NA NA NA D. sericea AC + min 24.50 0.33 3.10 (p = <0.01) 1.97 (p = 0.02) -0.04 (p = 0.97) NA 0.04 (p = 0.05) D. maculata min 1.07 0.06 NA NA NA NA 0.05 (p = 0.10) Flower community for C. tentaculata All AC 3.58 0.36 17.13 (p = 0.05) 25.13 (p = 0.01) 18.81 (p = 0.07) 9.00 (p = 0.02) 0.60 (p = 0.05) L. myrsinoides AC 43.60 0.37 23.31 (p = 0.01) 27.89 (p = <0.01) 25.52 (p = 0.01) NA NA B. umbellata AC + FF 19.81 0.31 -1.73 (p = 0.02) -1.38 (p = 0.04) 0.00 (p = 0.99) 0.63 (p = 0.01) NA Microseris sp. NULL NA NA NA NA NA NA NA

Model Estimate p -value Delta AICc Akaike weight D² D. maculata capsule set D. glaberrima 0.35 0.16 0.00 0.62 0.52 D. maculata 0.17 0.28 3.90 0.09 0.42 Rainfall (NULL) 0.14 0.01 4.04 0.08 0.34 Pultenaea 0.08 0.45 5.57 0.04 0.37 Mirbelieae 0.03 0.42 5.70 0.04 0.37 egg and bacon 0.02 0.40 5.79 0.03 0.36 D. sericea 0.03 0.45 6.09 0.03 0.35 Bossiaeeae 0.02 0.55 6.29 0.03 0.35 Dillwynia 0.02 0.62 6.39 0.03 0.34 Papilionoideae 0.01 0.89 6.62 0.02 0.34 C. tentaculata visitation B. umbellata -0.20 0.04 0.00 0.71 0.24 NULL NA NA 3.50 0.12 NA L. myrsinoides -0.01 0.25 4.44 0.08 0.03 Microseris 0.03 0.64 5.59 0.04 0.02 All 0.00 0.66 5.61 0.04 0.04

Figure 4.4: Partial regression plots showing the log odds of Diuris maculata capsule set as a function of a) the number of flowering Dillwynia glaberrima individuals, and b) winter rainfall.

4.4.3. Caladenia tentaculata

Inspection of spline correlograms revealed significant (95% confidence intervals did not overlap zero) positive autocorrelation at distances less than 500 m, and negative autocorrelation at approximately 3000 and 4500 m (Figure 4.3b). The inclusion of a random effect for site in the mixed effect model removed this autocorrelation (95% confidence intervals overlapped zero at all scales; Figure 4.3c), and the full model with this random effect had considerably more support than the full model without this effect (evidence ratio = 65.66). The model containing temperature was less likely to be the best model compared to the null (i.e. intercept plus random effect only) and the p-value for the slope test was 0.60, so temperature was not used in further modelling. There was evidence (evidence ratio = 5.91, p = 0.04) that the model containing B. umbellata was better than the null (D2 for model with B. umbellata as only predictor = 24%), which in turn was better than all other rewarding flower predictor models (Figure 4.5; Table 4.2).

Figure 4.5: Partial regression plot showing the number of wasp visits to Caladenia tentaculata as a function of the number of Burchardia umbellata flowers.

4.5. Discussion

4.5.1. Flowering and fire history

Flowering papilionoid Fabaceae were generally more abundant at 3-10 years post fire (AC2) and to a lesser extent at 11-35 years (AC3), which is consistent with the literature. The number of flowering individuals was lower at sites with short minimum inter-fire intervals for most taxonomic groupings, which may result from population decline following intervals shorter than the time to first significant post-fire flowering and reproduction (e.g. Keith 1996). The best D. maculata s.l. model also contained a positive relationship with minimum inter-fire interval, though this was not significant.

Leptospermum myrsinoides and all open-access flowering species combined exhibited a similar response to Fabaceae taxa but peaked in AC3 as opposed to AC2. The number of flowering Burchardia umbellata individuals was higher at recently burnt sites, which is consistent with the literature (Lamont & Downes 2011), and at frequently burnt sites.

4.5.2. Diuris maculata s.l.

Our hypothesis that fire-driven change in the availability of papilionoid (model) flowers would influence D. maculata s.l. (mimic) pollination received only weak support. The model of D. maculata s.l. capsule set presence/absence as a function of the number of flowering D. glaberrima individuals and winter rainfall was substantially better than any other model (including one with winter rainfall only), but the effect of D. glaberrima was not statistically significant. Moreover, while D. glaberrima displayed the same fire response as most other Fabaceae taxonomic groupings (i.e. peaking in abundance 3-10 years post-fire and declining thereafter), no fire history variable had a statistically significant effect on its floral abundance. Our measure of pollination (presence/absence of capsule set) was relatively coarse and our sample sizes small, such that a more sensitive measure (e.g. proportion of individuals with capsule set) and/or larger sample size may be required to detect effects.

A different sampling design might also improve the chances of detecting effects of fire-driven changes in papilionoid flowers on D. maculata s.l. abundance. The spatial scale (grain size) at which rewarding species abundance is measured can determine whether relationships with visitation to co-flowering rewardless species are detected (Johnson et al. 2003). In fire-prone landscapes fire age is often heterogeneous at scales corresponding to pollinator flight ranges (e.g. Cane & Neff 2011), such that the fire age (and associated resource levels) of the vegetation patch within which a plant is flowering (as quantified in the present study) is not necessarily the only fire age the animals visiting the plant experience. Given that papilionoid flowers are more abundant in some age classes than others, quantification of the relative frequencies of different age classes within foraging range of D. maculata s.l. may more accurately capture the availability of papilionoid flowers to pollinators. Fire age mosaics where pollinators forage on rewarding Papilionoideae in middle age classes and then encounter mostly deceptive orchids in adjacent early age classes is a potential example of the spatial resampling

56 situation that Gigord et al. (2002) argue selects for mimicry. Thus while our results provide only weak support for local-scale (e.g. magnet species) effects of fire on pollination through changes in Papilionoideae abundance, our sampling design may have been inadequate to test for landscape-scale fire effects involving pollinator populations dynamics and/or conditioned foraging preferences.

Temporal limitations of our design might also have played a role. Deceptive orchids generally are pollen-limited (Tremblay et al. 2005), but we detected a moderate effect of rainfall. This could indicate that capsule set was limited more by resources (water) in the year and/or at the spatial scale of our study, or that wetter areas experienced greater pollinator activity. Though we attempted to account for variation in rainfall statistically, the simple measure we used may have been inadequate. Ultimately, experimental manipulation of soil moisture and pollen-supplementation may be required to disentangle pollen- and resource- limitation. The positive effect of rainfall is interesting in its own right, given predicted declines in growing (winter) and flowering season (spring) rainfall in Victoria (CSIRO and Bureau of Meteorology 2015).

4.5.3. Caladenia tentaculata

Our hypothesis that C. tentaculata visitation is positively associated with fire- driven changes in the local abundance of open-access nectariferous species was not supported by the data. We found no effect of the local floral abundance of all nectariferous species combined. We found a negative effect of the local abundance of rewarding B. umbellata flowers, which is interesting because while rewarding heterospecifics have been found to decrease pollination of food-deceptive species (Lammi & Kuitunen 1995) we are not aware of similar effects being described for sexually-deceptive species. Our study was correlative so experimentation is required to confirm interspecific competition for pollination, but our results suggest that visitation in the recently burnt environment is not enhanced by changes in the local floral community. It is possible that thynnine feeding

57 resources were incompletely sampled because thynnine wasps are known to consume the sugary secretions of scale insects (Coccoidea and Diaspididae) and lerps (Psyllidae) (Phillips et al. 2009) which we did not record. Our results must also be interpreted in light of the fact that while wasp activity will vary through time in response to environmental conditions other than temperature, we observed visitation during a single visit to each quadrat. Quadrats closer in space (within sites) were also closer in time such that the random effect of site we detected may have captured some of this variation.

In summary, we did not find strong evidence that in the heathy woodlands of western Victoria fire-driven changes in the local floral community influence capsule set in D. maculata s.l. (at least when moisture is limiting) or pollinator visitation to C. tentaculata. We stress, however, that our findings are preliminary and pertain to local-scale effects of fire under relatively dry conditions. Trends reported in the present paper warrant further investigation, ideally with experimental control of soil moisture and other conditions.

5. SCALE-DEPENDENCE OF FIRE DIVERSITY-SPECIES DIVERSITY RELATIONSHIPS FOR FLOWER- VISITING FLIES AND WASPS OF SOUTH-EAST AUSTRALIA

5.1. Abstract

Fire can alter habitat conditions and in turn faunal assemblages. Patch mosaic burning aims to enhance species diversity by creating landscape mosaics of post-fire age classes supporting different species, though empirical support is mixed. Species diversity- pyrogenic diversity relationships might not be observed when pyrogenic diversity is measured at spatial scales incongruous with the scales at which species experience the environment. Further, since species within assemblages can vary in the scale at which they experience the environment, there may be independent effects of pyrogenic heterogeneity at different scales. The scale-dependence of species diversity-pyrogenic diversity relationships has received scarce attention. We explore the effects of fire- driven habitat change and pyrogenic heterogeneity at multiple spatial scales on aerial invertebrate assemblages in Mediterranean-type vegetation of south-east Australia. We used a space-for-time substitution design of 52 sites in different stages of post-fire succession to relate the abundance and composition of invertebrate species captured in

59 malaise traps to habitat conditions and post-fire age class. We also related species richness to site-scale habitat complexity (associated with fire age) and post-fire age class richness and diversity measured within 200 m and 800 m of traps. The effects of habitat and fire age on species abundance and composition were weak to moderate. Species varied in their association with fire age and the spatial scales at which these associations were strongest. The best models of species richness, with moderate explanatory power, contained pyrogenic heterogeneity at all three spatial scales. Species richness was positively associated with habitat complexity and fire age diversity within 800 m, but negatively associated with fire age diversity within 200 m. We suggest this resulted from species experiencing the environment at 200 m being more specialised on fire age classes and so being more sensitive to habitat area-heterogeneity trade-offs (i.e. for a given area, increasing the diversity of habitat types reduces the area of any single habitat type). This study highlights the importance of considering species diversity- pyrogenic diversity relationships at multiple scales, particularly in speciose assemblages where interspecific variation in spatial scales of experience is likely. For patch mosaic burning to be successful in enhancing species diversity, the scales at which fire age heterogeneity is created must be given careful consideration.

5.2. Introduction

Knowledge of relationships between fauna and fire history are important for biodiversity conservation in fire-prone landscapes. The habitat accommodation model (Fox 1982) posits that post-fire successional changes in habitat drive changes in faunal abundance as species specialise on transient habitat conditions. There is empirical support for this model in some taxa, though habitat- and so abundance-fire age relationships are spatially variable, potentially due to climatic factors (Monamy & Fox 2000; Fox, Taylor & Thompson 2003; Lindenmayer et al. 2008; Monamy & Fox 2010; Di Stefano et al. 2011; Nimmo et al. 2014; Sitters et al. 2014a; Swan et al. 2015). Knowledge of abundance-fire age relationships can be used to inform patch mosaic burning that aims to promote species diversity by creating spatial fire age diversity and so catering for species specialising on different age classes (Parr & Andersen 2006).

Species diversity and fire age diversity are not always positively correlated. This has been attributed to the nature of the study system, e.g. a lack of fire age specialisation or specialisation of all species on the same age classes, but also to the methodological issue of measuring fire age diversity at spatial scales that are not relevant to the study species (e.g. Kelly et al. 2012; Taylor et al. 2012; Farnsworth et al. 2014; Sitters et al. 2014b; Kelly et al. 2015). Fine-grained prescribed fire mosaics tend to enhance vertebrate alpha diversity whereas coarser-grained natural mosaics promote beta diversity (Pastro, Dickman & Letnic 2011; Pastro, Dickman & Letnic 2014). Few studies, however, have attempted to detect the scales at which assemblages respond to fire heterogeneity, focusing instead on the scale at which management currently creates heterogeneity (though see Sitters et al. (2014b)). Without an organism-centred understanding of responses to pyrogenic heterogeneity there is no basis for modifying burning practices to better meet conservation objectives in terms of alpha, beta, and gamma diversity.

Abundance-age class and species diversity-age class diversity relationships are poorly understood for aerial invertebrates in south-east Australia (New et al. 2010). Aerial invertebrates are highly mobile and so provide ecosystem services such as pollination and pest control that have been degraded globally and are in need of improved management (Kremen 2005; Millennium Ecosystem Assessment 2005). Pollination is of particular concern because of recently observed declines in pollinators linked to local extinctions of dependent plant species (e.g. Potts et al. 2010; Pauw & Bond 2011; Pauw & Hawkins 2011; Vanbergen et al. 2013). Bee abundance-fire age relationships have been observed and parallel successional changes in floral and nesting resources (Potts et al. 2003a; Potts et al. 2005; Potts et al. 2006; Campbell, Hanula & Waldrop 2007; Rodríguez & Kouki 2015). A recent study also found bee species richness was positively associated with landscape pyrodiversity within foraging range (Ponisio et al. 2016), providing support for the effectiveness of patch mosaic burning. While bees are generally the most important pollinators, the fire responses of other important pollinators such as flies and wasps (Larson, Kevan & Inouye 2001; Gaskett 2011; Willmer 2011; Phillips et al. 2014a; Phillips et al. 2014b; Orford, Vaughan & Memmott 2015) are poorly understood. While bees typically consume floral resources as both adults and larvae, flies and wasps that consume floral resources as adults often belong to

61 predatory, parasitic, saprotrophic, or other feeding guilds as larvae (Jervis et al. 1993; Woodcock et al. 2014). Thus they not only contribute to other ecosystem services, but also may be more responsive to fire-driven changes in non-floral resources such as insect hosts, leaf litter, or logs (e.g. Mateos, Santos & Pujade-Villar 2011).

We investigate fire effects on fly and wasp abundance and diversity in south-east Australia. We explore abundance-fire age relationships at both site- and landscape- scales (i.e. where fire age is respectively homogeneous and heterogeneous), and abundance-habitat and habitat-fire age relationships to better understand mechanisms through which fire might influence flies and wasps. We also investigate relationships between species diversity and pyrogenic heterogeneity both in terms of fire age diversity (at multiple spatial scales) and fire-driven changes in habitat complexity. We hypothesise that if within assemblages there is interspecific variation in the spatial scale at which the environment is experienced, as has been observed in response to other forms of landscape heterogeneity (e.g. Steffan-Dewenter et al. 2002; Westphal, Steffan- Dewenter & Tscharntke 2006), pyrogenic heterogeneity will influence assemblages at multiple spatial scales. We focus on alpha species diversity because it rather than beta diversity enhances and stabilises ecosystem services such as pollination and pest control (Hoehn et al. 2008; Winfree & Kremen 2009; Blüthgen & Klein 2011; Tscharntke et al. 2012).

5.3. Methodology

5.3.1. Study landscape and site selection

The study landscape is an area of approximately 15,000 km2 in south-west Victoria, Australia (Figure 5.1). The area encompasses a number of National Parks and State reserves (including the Grampians National Park) within a predominantly agricultural matrix. The native vegetation is primarily restricted to reserves and consists of sclerophyllous heaths, shrub lands, and woodlands interspersed with open grasslands (Dodson 2001; Gibbons & Downes 1964). The area has a Mediterranean-type climate (hot dry summers and mild wet winters), and has experienced recurring fires throughout

62 the Holocene (review in Dodson 2001). Prescribed fire is applied by land management agencies in an attempt to protect human life and property and achieve ecological objectives (Parks Victoria 2003).

Study sites were points in the landscape and were selected within the ArcGIS environment according to a number of criteria: 1) No non-native vegetation cover within 1 km to minimise the effects of other land cover types. 2) Greater than 90% heathy woodland (Ecological Vegetation Class 48) cover within 1 km. This criteria was used to control for effects of other native vegetation types, with different fire responses and floristics. 3) To enhance independence no two sites could be closer than 2 km (available information suggests home ranges of most of the study taxa are smaller than this; see ‘Collection of age class data’ below) or contain areas of vegetation burnt last by the same fire. 4) Sites had to be within 1000 m from a road for ease of access, but more than 100 m from the road to reduce edge effects. 5) Not to be burnt by land management agencies during the current prescribed burning season. 6) Collectively sites covered a broad range of time-since-fire (TSF) and so represented a chronosequence (1- 75 years). Under these criteria 60 sites were available but only 52 sites were used due to accessibility difficulties.

5.3.2. Trapping

Flying insects were sampled using flight-intercept (malaise) traps. A single flight- intercept trap was placed at each of the 52 sites for three, ten day periods between late September and early November of 2013. Three kinds of traps were used (all using ethylene glycol as the preserving agent) due to limits on trap availability; large and small malaise (SLAM) traps (from Australian Entomological Supplies) and home-made traps similar in dimensions to small SLAM traps. Trap types were systematically rotated around sites to reduce effects of trap type on insect capture. Traps were erected (veins aligned with the cardinal points) at random locations previously chosen in the ArcGIS environment. Traps were only moved from this random point up to 10 m to avoid thickets of tall shrubs and other interferences.

Trap contents were later sorted into morphospecies within families in the laboratory using a binocular microscope and stored in 70% ethanol. Fly and wasp voucher specimens were verified by taxonomic experts. We analysed a subset of fly and wasp families thought generally to be important pollinators and/or known to be involved in specialised interactions with particular Australian plants. Of the flies, Syrphidae are second only to bees in importance as pollinators, followed by Bombyliidae, Muscidae, and Tachinidae (Larson, Kevan & Inouye 2001; Willmer 2011)). We also chose to analyse Mycetophilidae as they are potentially important pollinators generally (Larson, Kevan & Inouye 2001) and are the sole pollinators of some Australian orchids (Phillips et al. 2014b). Parasitoid wasps generally rely on floral resources in the adult phase, but Tiphidae, Ichneumonidae, and potentially Brachonidae also form highly specialised interactions with rewardless Australian orchids (Adams & Lawson 1993; Gaskett 2011).

5.3.3. Collection of habitat data

We recorded habitat variables at each trapping location. While some Syrphids and Bombyliids collect both nectar and pollen and have long mouthparts allowing them to access rewards in tubular flowers, flies (including Muscidae, Tachinidae, and Mycetophilidae) and parasitic wasps (including Ichneumonidae, Braconidae, and Tiphidae) generally visit flowers for nectar and are restricted by their short mouthparts to open-access flowers (Willmer 2011). Species from the genera Leptospermum were

64 the dominant nectariferous, open-access plants flowering during the sampling period, and species from the fly and wasp families of interest are associated with Leptospermum in Australia (Barraclough 1990; Hingston & McQuillan 2000; Brown & Phillips 2014). The number of flowering individuals was recorded at each site in a 20 x 20 m quadrat centred on the trap three times (each time traps were erected), and then averaged across visits for each site.

Habitat structure was also recorded as flies and parasitoid wasps are highly dependent on non-floral resources, particularly during larval stages. Many of the study taxa are parasitic, but host identities are poorly understood so indirect measures were used. Habitat characteristics were selected on the basis of known associations with invertebrates of southern Australian forests or overseas where information on the focal fly and wasp groups was lacking from Australia. (Yeates, Logan & Lambkin 1999; Schiegg 2000; Hilszczański et al. 2005; McElhinny et al. 2006; Lassau & Hochuli 2007; Boesi, Polidori & Andrietti 2009; Haubruge, Almohamad & Verheggen 2009; Dickinson 2012). These included bare ground, leaf litter, logs, standing dead wood, and live vegetation classified by growth form or height categories. These variables were recorded once at each site along two crossed 40 m transects (forming a + shape) oriented along the N-S and E-W axes. At 1 m intervals along each transect (80 intervals) a structure pole was used to record characteristics at 0, 10, 20, 50, 100 and 200 cm. At 0 cm ground cover was categorised as bare ground, leaf litter, or other. At higher intervals touches on the pole were recorded as dead wood, herbs, shrubs, Xanthorrhoea, or tree. The presence of logs (dead wood > 5 cm diameter) within each quarter of a 1m radius circular plot was also recorded giving a score of 0 to 4 for each interval. Finally, a densiometer was held vertically to record the presence of vegetation above the pole (at the centre of the cross-hair) directly above the interval.

5.3.4. Collection of age class data

Spatial layers in ArcGIS were interrogated to identify the age classes of vegetation surrounding each flight-intercept trap. Spatial fire history data sourced from Victorian Department of Environment, Land, Water and Planning was used to determine time since last fire in years for all native vegetation within the study landscape. These data were then categorised into five age classes which corresponded approximately to heathy

65 woodland growth stages as defined by (Cheal 2010) based on known vegetation fire responses, though with modifications: the two youngest growth stages were rare so we combined them in a single age class, and the time taken to reach the final growth stage can vary from 45 to 55 years in the study landscape (Cheal 2010) so we used 50+ years for the oldest age class. The age class categories were thus identified in years since fire as: age class 1 (AC1) = 0-3 years (renewal-juvenility), age class 2 (AC2) = 4-10 years (adolescence), age class 3 (AC3) = 11-35 years (maturity), age class 4 (AC4) = 36-50 years (waning), age class 5 (AC5) = 50+ years (senescence). We chose Cheal’s classification scheme because it is being assessed as a surrogate for faunal abundance generally in Victoria (MacHunter, Menkhorst & Loyn 2009), but since fauna may perceive different age classes equivalently (e.g. because levels of a particular resource vary little between two or more age classes) we derived an alternative classification scheme of age class variables representing longer periods of post-fire succession though maintaining the boundaries of these categories. They were AC1b = 0-10 years, AC2b = 11-50 years, AC3b = 50+ years. Each site thus had two separate age class categorisations.

To detect relationships between spatial heterogeneity and organism abundance, the scale at which heterogeneity is measured must correspond to the scale at which organisms respond to heterogeneity (Wiens 1989). The spatial scales of response reported in the literature differ between and within flower-visitor taxonomic groups; Tachinids 50-850 m (Roland & Taylor 1997), Mycetophilids and other saproxylic flies 150 m (Schiegg 2000), Syrphids 500-1000 m (Kleijn & Van Langevelde 2006; Meyer, Jauker & Steffan-Dewenter 2009), Ichneumonids and Braconids 600-7000 m (Gibb et al. 2008). We thus measured fire age heterogeneity at multiple scales, by extracting from the fire age class layer the proportion of land covered by each age class within 200, 400, and 800 m of each trap using the Patch Analyst routine in ArcGIS. The lower limit of 200 m was the smallest scale at which mapped age class was sufficiently heterogeneous, i.e. each fire age class had at least five sites with values other than 0 or 100. This was done for both kinds of age class classification, such that each site had a value between 0 and 100 for AC1, AC2, AC3, AC4, AC5, AC1b, AC2b, and AC3b at each spatial scale. Ground truthing was conducted where practical and mapped data was found generally to be accurate. One site mapped as being AC1 contained large patches of unburnt

66 vegetation within 50 m of the trap, though was retained as it was not an outlier for any species (see below).

5.3.5. Data analysis

A single value for each of the nine habitat variables was obtained for each site. The values for standing dead wood, herbs, shrubs, Xanthorrhoea, and trees were all calculated as the number of times each touched the structure pole summed across height categories and intervals. Logs (0-4 at each interval), litter, and bare ground (0-1 at each interval) were summed across intervals. The score for Leptospermum was the maximum number of flowering individuals observed at each site across the three site visits). Scores for each habitat variable were then averaged across sites within age classes. Two indices of habitat complexity were also derived for each site by summing touches across categories and calculating a Shannon Diversity Index for the full set of habitat variables (excluding bare ground because it was highly correlated with leaf litter) (McElhinny et al. 2005). ANOVAs were used to test for an effect of age class (as a categorical variable) on all habitat variables.

We investigated relationships between the flower-visiting fly and wasp assemblage, and habitat and age class predictors, while accounting for spatial variation using partial constrained ordination. Abundance data were log transformed and only morphospecies detected at more than one site were included in analysis to reduce the effect of rare species. Detrended Correspondence Analysis revealed long gradients (3.8 SD units) so we assumed a unimodal distribution and used Canonical Correspondence Analysis (CCA) (ter Braak, Prentice & Caswell 1988; Šmilauer & Lepš 2014). Forward selection was used to determine the variation in assemblage composition explained by trap spatial coordinates and their extensions with the following power terms to form trend-surface polynomials (Legendre & Legendre 2012); easting (E), northing (N), E2, N2, E x N, E2 x N, and N2 x E. We used Monte-Carlo permutation tests (999 permutations) of the statistical significance of the relationship between the species matrix and each spatial predictor variable. Significant variables (p<0.05) were then used as covariates in three partial CCAs relating assemblage composition to 1) the least inter-correlated set of habitat variables (Leptospermum, Litter, and Logs), 2) AC1, AC2, AC3, AC4, and AC5 measured at the 200 m scale and 3) the 800 m scale. Separate ordinations were used for

67 each scale to assess whether different subsets of the assemblage responded at each scale. Forward selection was again used to determine the marginal and conditional effect of each habitat and fire variable on assemblage composition, with significant variables (p<0.05) added to the model. Ordination results are presented as bi-plots with axes corresponding to the directions of the greatest assemblage variability explained by habitat and fire variables, accounting for broad-scale spatial effects. Analyses were performed using CANOCO for Windows Version 5.

We investigated relationships between individual taxa and habitat, local-scale, and landscape-scale age class using an information-theoretic approach (Burnham & Anderson 2002). Morphospecies detected at ten or more sites, or morphospecies pooled within families where no morphospecies met this criterion but there were less than five individuals, were included in analysis of individual taxa. The total number of individuals captured at each site across all trapping periods was used as the response variable in each case. Data were analysed using Generalised Linear Models (GLM). All data sets were checked for outliers using scatterplots and Cook’s D for GLM. Where over-dispersion was present negative binomial models were used, otherwise Poisson models were used. All analysis was conducted in the R statistical environment (R Development Core Team 2011).

Single-predictor variable models were used for all taxa because most were detected at less than 20 sites. The nine habitat variables were used for each taxa. Local fire age models consisted of a single, binary predictor for each age class (e.g. 1 = sample point is mapped as AC2, 0 = sample point is not mapped as AC2) from both classification schemes (n = 7 models). This was preferred to a model consisting of age class as a single continuous predictor because abundance-age class relationships are often nonlinear (e.g. Sitters et al. 2014a), or age class as a categorical variable because model complexity would have been too high (four dummy variables) considering most taxa were detected at less than 40 sites. Moreover, it facilitated comparisons with corresponding landscape fire age models by making the number of parameters (K) equal. There were 21 landscape fire age models for the individual taxa (seven age class categories × three spatial scales), each consisting of a single, continuous predictor model for each age class (e.g. proportion of AC2 within 200 m of the sample point).

Thus there were 38 models (including a null model with intercept only) for each taxa. Akaike Information Criterion for small sample sizes (AICc) was used to select the best model for each response. Evidence ratios were then calculated for the best model from each set relative to the null model (e.g. Akaike weight best habitat model/Akaike weight null model). Explained deviance was also calculated for the best model from each set.

We also investigated relationships between species richness, habitat complexity, and landscape fire age heterogeneity using an information theoretic approach. The total number of morphospecies captured at each site across all trapping periods was used as the response variable. We calculated two fire heterogeneity measures, the number of fire age classes (using only the first categorisation; i.e. AC1-5) present within 200 m and 800 m of sample location (400 m was not included as it was highly correlated with measures at the 200 and 800 m scales). These were categorical predictors because species diversity-environmental diversity relationships can be non-linear (Fahrig et al. 2011; Laanisto et al. 2013). Since different aerial invertebrate taxa may perceive the landscape at different spatial scales (e.g. Steffan-Dewenter et al. 2002) we included models containing additive combinations of fire age richness and Shannon diversity measured at each scale (i.e. to test the effect of fire age richness/diversity at one scale on one species subset while accounting for the effect of fire age diversity at another scale on a different species subset), alone and with habitat complexity, giving 21 models (including a null model with intercept only) in the candidate set: additive habitat complexity index (AHC), Shannon diversity habitat complexity index (SHC), Richness 200, Richness 800, Richness 200 + Richness 800, Shannon 200, Shannon 800, Shannon 200 + Shannon 800, AHC + Richness 200, AHC + Richness 800, AHC + Shannon 200, AHC + Shannon 800, AHC + Richness 200 + Richness 800, AHC + Shannon 200 + Shannon 800, SHC + Richness 200, SHC + Richness 800, SHC + Shannon 200, SHC + Shannon 800, SHC + Richness 200 + Richness 800, SHC + Shannon 200 + Shannon 800. Variance inflation factors were less than 1.5 for all models, indicating multi-colinearity was not a problem (Quinn & Keough 2002). Akaike Information Criterion for small sample sizes was used for model comparison, and evidence ratio relative to the null model and explained deviance was calculated for models with substantial support (ΔAICc < 2).

Finally, it was possible that home ranges or dispersal distances for some groups were larger than the minimum 2 km buffer between sites but less than the extent of the study landscape (e.g. dispersal distance of 7 km for an Ichneumonidae (Gibb et al. 2008)). Spatial autocorrelation was assessed using spline correlograms (with 95% pointwise bootstrap confidence intervals) of the residuals of the best model for each taxa.

5.4. Results

The flight-intercept traps detected 637 individual Syrphidae from five morphospecies, 24 Bombyliidae from five morphospecies, 177 Tachinidae from 24 morphospecies, 108 Muscidae from 15 morphospecies, 45 Ichneumonidae from 17 morphospecies, 9 Tiphidae from 9 morphospecies, 75 Brachonidae from 33 morphospecies, and 4634 Mycetophilidae from 53 morphospecies. Mycomya is the only Mycetophilidae genus known to form highly specialised interactions with Pterostylis orchids (Phillips et al. 2014b), so to obtain a clearer picture of how species in this genus respond to fire, and to prevent the relatively high abundance of Mycetophilidae dominating ordination patterns, only the 38 individuals from 5 morphospecies of the genus were used in analysis (except in calculating species richness).

5.4.1. Habitat variables

Habitat variables exhibited a range of relationships with fire age (Figure 5.2). Habitat variables exhibiting statistically significant (α = 0.05) effects of age class were: Leptospermum (F = 10.51, p = <0.001), litter (F = 16.45, p = 0.00), standing dead wood (F = 5.06, p = <0.001), logs (F = 4.08, p = 0.01), shrubs (F = 11.33, p = <0.001), bare ground (F = 6.71, p = <0.001), AHC (F = 8.11, p = <0.001), and SHC (F = 3.76, p = 0.01). Statistically non-significant effects were observed for: herbs (F = 1.07, p = 0.38), trees (F = 1.76, p = 0.15), and xanthorrhoea (F = 0.40, p = 0.81). Bare ground decreased whereas litter cover increased with fire age (only litter is shown because these variables were highly correlated; Pearson correlation coefficient = -0.81). Shrubs and standing dead wood both showed a hump-backed relationship with fire (i.e. was highest in middle age classes). Leptospermum showed a somewhat similar pattern, though the

70 reduction in the first age class was much more pronounced and there was some evidence of a middle-age class dip. Log cover was highest in the oldest age class and there was some evidence of a decline in the middle-age class. Complexity indices were lowest in AC1.

5.4.2. Assemblage

None of the three ordinations contained variables collectively explaining more than 10% of the variation in assemblage composition; however there were predictor variables

71 with statistically significant effects within each ordination. Two habitat predictors were significant; litter explained 3.7% of variation (pseudo-F = 1.9, p = 0.001), logs explained 3% (pseudo-F = 1.5, p = 0.016). Three age class predictors were significant at 200 m; AC1 explained 3.5% (pseudo-F = 1.8, p = 0.006), AC4 explained 3% (pseudo-F = 1.5, p = 0.034), and AC5 explained 2.8% (pseudo-F = 1.5, p = 0.044). Two age class predictors were significant at 800 m; AC1 explained 3.4% (pseud-F = 1.7, p = 0.004) and AC4 explained 3% (pseudo-F = 1.5, p = 0.034).

In the habitat ordination (Figure 5.3A), log cover separated Mycomya sp. 3, and to a lesser extent some Muscidae, Tachinidae, and Ichneumonidae species (all with positive associations with log cover), from Ichneumonidae sp. 4, Braconidae sp. 3 and some Muscidae, Tachinidae, and Tiphidae species. Litter separated Braconidae sp. 3, the Tiphidae species, and to a lesser extent some Muscidae and Tachinidae species (negative relationships), from some Ichneumonidae sp. 2 and 5, some Tachinidae, and to a lesser extent Mycomya and Muscidae species.

To an extent, the patterns observed in the fire age at 200 m ordination (Figure 5.3B) reflected those observed in the habitat ordination. Bracondiae sp. 3 and the Tiphidae species were positively associated with AC1, which contains low levels of litter and logs with which these species were negatively associated in the habitat ordination. Mycomya sp. 3 and 5, which were positively associated with logs and litter respectively, were positively associated with AC5. Similarly, Ichneumondiae sp. 2 and 5 which were positively associated with litter were also positively associated with AC4. However, age class associations did not reflect habitat associations in all cases. For instance, Muscidae sp. 5 had the strongest positive association with AC5 but was not strongly positively associated with litter or logs.

The relative positions of species along gradients changed between the 200 m and 800 m ordinations (Figures 5.3B and 5.3C; AC5 is shown in 5.3C, even though it was not statistically significant, for comparisons with 5.3B). For instances, Ichneumonidae 5 and 2 diverged along the AC4 gradient between 200 and 800 m. There were also instances of species reversing their relative positions along gradients between the 200 and 800 m ordinations, such as Bombyliidae 2 exhibiting a weaker and then stronger association with AC4, compared to Tachinidae 4.

5.4.3. Individual taxa

A range of responses was observed at the species level (Table 5.1). Habitat predictors had greater support (larger evidence ratios) than fire age predictors for Mycomya which had a negative relationship with bare ground cover, Tachinidae sp. 2 which increased with log cover, and Braconidae sp. 16 which increased with Xanthorrhoea. Syrphidae sp. 1, Tachinidae sp. 3, Muscidae sp. 7, and Bombyliidae all had habitat models with substantially more support than the null (two AICc units greater), though less support than fire age predictor models. No habitat model had substantially more support than the null for Tachinidae sp. 23.

Responses to fire age varied between taxa. Braconidae sp. 16 was the only taxa for which no fire age model had substantially more support than the null model. The strongest effect of fire age for Syrphidae sp. 1 and Tachinidae sp. 3 was a negative response to AC1. Tachinidae sp. 2 and sp. 23 on the other hand responded negatively to middle age classes (AC2b). The strongest response of Muscidae sp. 7 was negative to AC4, whereas for Bombyliidae it was positive to AC4. Mycomya was the only group for which there was evidence of specialisation on AC5.

Most taxa (Syrphidae sp. 1, Mycomya, Tachinidae sp. 2, Tachinidae sp. 3) responded most strongly to fire age within approximately 40 m, providing no evidence that mapped fire age heterogeneity was relevant to these taxa. Tachinidae sp. 23, Muscidae sp. 7, and Bombyliidae, on the other hand, responded most strongly at landscape-scales. Thus fire age heterogeneity was relevant to these taxa.

Syrphidae sp. 1 Tachinidae sp. 2 Tachinidae sp. 3 Tachinidae sp. 23 Scale Model ER D2 % Model ER D2 % Model ER D2 % Model ER D2 % Habitat +Litter 3.91 8 +Logs 11.1 14 +Litter 4.12 15 NULL NULL NULL Local -AC1 32.5 15 -AC2b 2.22 8 -AC1 5.73 16 NULL NULL NULL 200 -AC1 25.6 14 -AC2b 1.66 7 -AC1 1.34 8 NULL NULL NULL 400 -AC1 12.46 12 -AC2b 1.75 7 NULL NULL NULL -AC2b 2.41 12 800 -AC1 5.95 9 NULL NULL NULL NULL NULL NULL NULL NULL NULL Muscidae sp. 7 Mycomya Bombyliidae Braconidae sp. 16 Scale Model ER D2 % Model ER D2 % Model ER D2 % Model ER D2 % Habitat -Shrubs 44.34 11 -Bare 102.23 24 +Bare 3.86 17 +Xanth 3.29 12 Local -AC4 72.2 4 +AC5 8.88 18 -AC4 1.19 8 NULL NULL NULL 200 -AC4 38.2 3 +AC5 6.56 13 +AC4 4.19 18 NULL NULL NULL 400 -AC4 13 5 +AC5 2.14 10 +AC4 5.17 19 NULL NULL NULL 800 -AC4 34 10 -AC1b 2.41 9 +AC4 7 20 NULL NULL NULL

5.4.4. Species richness

Only two models had substantial support: species richness as a function of 1) the additive habitat complexity measure, fire age diversity within 200 m, and fire age diversity within 800 m (ER = 71.4, D2 = 26%); and 2) the additive habitat complexity measure, fire age Shannon diversity within 200 m, and fire age Shannon diversity within 800 m (ER = 22.8, D2 = 17%). Species richness increased with additive habitat complexity (parameter estimate = 0.01, p = 0.02), was lower when there was fire age diversity within 200 m (parameter estimate for Shannon 200 from best model = -0.36, p = 0.02), and higher when there was fire age diversity within 800 m (parameter estimate for Shannon 800 from best model = 0.28, p = 0.03) (Figures 5.4 and 5.5).

Figure 5.6: spatial autocorrelation (Moran’s I, with 95% confidence intervals) at a range of distances for the residuals of the best model for each taxa.

5.5. Discussion

We investigated relationships between fly and wasp abundance and species richness, habitat conditions, and fire age heterogeneity at multiple spatial scales. Most habitat features that we assessed varied with fire age, but regression models indicated that these habitat features explained only small to moderate amounts of variation in fly and wasp abundance. Mycomya species abundances were moderately associated with litter and/or

77 log cover (or negatively associated with bare ground which itself was negatively associated with litter cover), or older age classes where these habitat conditions were most abundant. Schiegg (2000) found positive associations between logs and Mycetophilidae and other saproxylic flies at a similar spatial scale (150 m). These habitat variables were probably indirect measures as Mycetophilidae typically feed on the fruiting bodies of saprotrophic fungi (Colless 1970; Dickinson 2012). Bombyliids were moderately to weakly positively associated with bare ground and AC4 at larger spatial scales. Australian Bombyliidae often parasitise the larvae of ground-nesting Hymenoptera and Diptera (Irwin & Yeates 1995; Yeates, Logan & Lambkin 1999) such that easy access to soil may facilitate larval development. Bare ground was not highest in AC4 so other Bombyliid or host resources may have been most abundant in this age class. The weak responses exhibited by most other taxa suggest there were unmeasured habitat conditions more directly related to organism abundance that exhibit weak or contrasting responses to our fire age categories, or that variation is not driven by habitat over this range of sites. Future research should aim to determine more direct habitat variables such as insect host abundance and, if these are related to time-since-fire, derive a more invertebrate-centred (rather than vegetation-centred (Cheal 2010)) fire age classification scheme.

There was evidence of interspecific variation in habitat and age class associations from both ordination and regression analyses. Habitat and fire age heterogeneity explained small, but statistically significant amounts of variation in assemblage composition, and the best habitat and fire age predictors in regression models varied between species. Interspecific variation is not surprising because even within families larval requirements vary. The variability within Muscidae could result from the diversity of larval feeding strategies (e.g. coprophagy, necrophagy, parasitism (Skidmore 1985)). The variability within Ichneumonidae, Braconidae, and Tachinidae, which are all parasitic (Gauld 1984; Stireman III, O'Hara & Wood 2006), could result from variation in insect host resource requirements and fire-driven changes in these requirements, as observed for Braconids during post-harvest succession (Maleque et al. 2010).

Ordinations and regressions also indicated variation in the scale at which pyrogenic heterogeneity was experienced. Species associated with a given age class often changed

78 their positions relative to each other along the gradient when it was measured at different spatial scales. This suggests interspecific variation in the scale of response because the optima of species responding at the same scale would move along that gradient in unison as the scale of measurement changed. Similarly the scale at which abundance-age class relationships were strongest in regression analyses varied between species. Again, this is not surprising because interspecific variation in the scale of response is known within Tachinidae (Roland & Taylor 1997), Braconidae (Gibb et al. 2008), and other aerial invertebrate groups (Steffan-Dewenter et al. 2002; Westphal, Steffan-Dewenter & Tscharntke 2006).

From the observed inter-specific variation in 1) habitat and fire age associations and 2) the spatial scale at which these associations were strongest, we would expect species diversity-pyrogenic heterogeneity relationships at multiple scales. Consistent with our hypothesis, the best model of species richness, which had moderate explanatory power, included habitat diversity at three scales; habitat complexity (which varied with site fire age) within 40 m, fire age diversity within 200 m, and fire age diversity within 800 m. Species richness had a positive relationship with our additive habitat complexity index, which would be expected because niche space increases with habitat heterogeneity and so accommodates a greater diversity of species (MacArthur & MacArthur 1961; Rosenzweig 1995). Species richness had a positive relationship with fire age diversity at 800 m, but a negative relationship at 200 m. Fire age diversity could have a positive effect at 800 m because species responding at this scale differ in their age class preferences, and a negative effect on species responding at 200 m because they all specialise on the same age class such that fire age diversity equates to habitat loss (Kelly et al. 2012; Farnsworth et al. 2014). Alternatively, since for a given area the amount of habitat available for each species decreases as the diversity of habitats for different species increases, hump-shaped species diversity-environmental diversity relationships are expected (Allouche et al. 2012). These curves are expected to reach their inflexion points at lower levels of environmental diversity for more specialised species as they are more susceptible to stochastic extinctions (Allouche et al. 2012). Thus, species responding at 200 m may have been more specialised, without necessarily depending on the same fire age, and so experienced negative fire age diversity effects at lower levels of fire age diversity. The greater amount of variation in assemblage

79 composition explained by age class composition at 200 m compared to 800 m lends some support to the latter explanation since it indicates species responding at 200 m were more divergent and/or specialised in their age class associations. Equally, though, it could indicate our age class classification scheme was less appropriate for species responding at 800 m. Further research is warranted because these alternatives have different management implications where the aim is to maximise alpha diversity to enhance and stabilise ecosystem services. If species responding at 200 m specialise on a particular age class, alpha diversity of these species would be greater in areas with large amounts of this age class. If species responding at 200 m are instead more specialised, promoting any one age class would not be expected to promote alpha diversity, though gamma diversity might still be enhanced via beta diversity through the creation of larger scale (e.g. 800 m) fire age diversity.

It is possible the opposing relationships at 200 m and 800 m were in part caused by individual species responding at both scales but through different mechanisms. Fire age diversity at 800 m could have represented a larger number of habitat types, but fire age diversity at 200 m fragmentation of each habitat type (micro-fragmentation sensu Tamme et al. (2010) and Laanisto et al. (2013)). Disentangling the relative influences of different species responding at different scales (as discussed in the previous paragraph) vs. the same species responding at both scales also has implications for management. If there is both interspecific variation in the scale of experience and micro-fragmentation effects, the creation of heterogeneity at a given scale could benefit one group of species while being detrimental to another.

The relationships between species diversity and any single measure of pyrogenic heterogeneity were weak. Our habitat complexity index may have been a weak predictor for several reasons. It was an additive combination of touches across habitat variables which assumes variables are interchangeable, whereas the observed between-taxa differences in variable associations suggest some of these variables are non- substitutable (Burgman et al. 2001; McCarthy et al. 2004). It is also an estimate of the amount of vertical structure which may present a flight barrier and so be avoided by aerial invertebrates, at least at the scale of movement decisions (Lassau & Hochuli 2005). The habitat variables we used were not strong predictors of species abundance so

80 their combined effects would be weak. The strength of relationships with fire age diversity at 200 and 800 m might have been weak because our fire age categorisation was not appropriate for the study species. Further, equal weighting was given to each age class whereas there may have been variation between age classes in the number of species they supported such that unequal weighting of age classes (i.e. relative to the number of species each supported) would have been more appropriate (Kelly et al. 2015). Finally, each measure of pyrogenic heterogeneity, corresponding to a particular scale, may have been weakly related to species richness because it explained only the variation in the subset of the species experiencing the environment at that scale. Studies of speciose assemblages, where the potential for interspecific variation in spatial scales of experience is great, should test for independent effects of pyrogenic heterogeneity measured at different scales as each may explain a unique component of variation in species diversity.

6. GENERAL DISCUSSION

The aim of my thesis was to better understand fire effects on pollinators and pollination and their role in fire management for biodiversity conservation. I addressed two broad questions: 1) How might the multi-scale understanding of plant-pollinator interactions in heterogeneous landscapes, based on empirical work from agricultural and urban landscapes, be incorporated into and extend fire management for biodiversity conservation (Chapter 2)? 2) Does the multi-scale understanding of plant-pollinator interactions in heterogeneous landscapes apply to plant-pollinator interactions in fire- prone landscapes of the Mediterranean climatic zone of south-eastern Australia (Chapters 3, 4, and 5)? In this chapter I briefly summarise how each previous chapter addresses these questions and provides additional insights into fire effects on pollinators and pollination, then discuss future directions.

6.1. Summary and synthesis

In Chapter 2 I developed a conceptual model for understanding fire effects on plant- pollinator interactions in a way that facilitates incorporation of empirical work into simulation modelling to assist land management decisions. This approach is currently taken to predict population dynamics in fire-prone landscapes, but I took the extra step of incorporating the processes governing plant-pollinator interactions to enhance the realism of simulations. I also reviewed phenotypic traits of pollinators and plants that

82 could assist managers in parameterising this model. The literature suggests that above- ground nesting, univoltine pollinators may be particularly vulnerable under expected fire regime changes, and that improved knowledge of traits mediating the exploitation of landscape heterogeneity could be used to enhance the persistence of these species.

Chapter 3 explored the possibility of fire effects through multiple processes, operating over different spatial scales, on the interaction between the Australian terrestrial orchid Caladenia tentaculata and its wasp pollinator. I found that visitation to the orchid was positively associated with recently burnt vegetation at the local-scale (the sample location) and long-unburnt vegetation at the landscape scale (the circular area within 500 m of the orchid, approximating the maximum known flight range of thynnine wasp pollinators). Further, these effects were interactive such that the positive effect of long- unburnt vegetation at the landscape-scale was greatest when the local environment was recently burnt, and vice versa. Thus this chapter provided evidence that the multi-scale understanding of plant-pollinator interactions developed in agricultural and urban landscapes applies also in fire-prone landscapes of south-east Australia. Future research should investigate whether these effects translate into recruitment and so influence population dynamics.

In Chapter 4 I initially set out to investigate multi-scale effects on seed production of five orchid species, but flower wilting and predation resulting from dry weather allowed only simple modelling of two species. I investigated the effects of fire-driven changes in the local abundance of rewarding heterospecific flowers on pollination of two rewardless orchid species, Diuris maculata s.l. and Caladenia tentaculata, putatively dependent on these heterospecifics to attract and support pollinators. Rainfall had an overwhelming effect on capsule set in D. maculata s.l., which is interesting because it suggests that under the drying climate predicted for Victoria, seed production of this orchid may be limited more by moisture than pollination (in opposition to the general trend toward pollen-limitation in deceptive orchids (Tremblay et al. 2005)). There was some evidence of competition for pollinator visitation between C. tentaculata and a co- flowering heterospecific (B. umbellata) that exhibited greatest floral abundance in the recently burnt environment, though this did not appear to translate into lower visitation in recently burnt sites (Chapter 3). While no strong effect of fire age on visitation was

83 detected, the pattern was similar (highest in the youngest and oldest age classes) to that found in Chapter 3 and was probably not as strong due to the smaller sample size.

Chapters 3 and 4 considered together suggest that fire may influence C. tentaculata visitation through multiple processes at local scales. Negative effects of competition for pollination between C. tentaculata and B. umbelatta (which need to be tested with manipulative experiments before they are confirmed) were overwhelmed by positive effects of some as yet unconfirmed local-scale mechanism which is strongest in a landscape context of high proportions of long-unburnt vegetation. Future research could test whether there are local-scale conditions (e.g. fire severity, inter-fire interval) under which visitation is reduced in the recently burnt environment.

In Chapter 5 I focused on the flower-visiting fly and wasp assemblage, investigating relationships between species abundance, habitat features, and fire age, and between species richness, habitat complexity, and fire age diversity. I investigated the hypothesis that interspecific variation in the scale at which fire-driven changes in the environment are experienced results in species diversity responding to pyrogenic heterogeneity at multiple spatial scales. All abundance-fire age relationships were weak or moderate, suggesting my age class classification – based on the system currently used for fire management in Victoria (Cheal 2010) – was not the best that could be derived for these taxa (a potential focus for future research), or that these taxa do not exhibit strong relationships with fire age. There was some evidence of between-species variation in habitat and age class preferences, as well as the scales at which the environment was experienced. Consistent with the stated hypothesis, relationships between fire heterogeneity and species richness occurred at multiple spatial scales, though relationships were moderate. Species richness was positively related to local-scale (sample location) habitat complexity (which was greatest in older age classes), negatively related to fire age diversity within 200 m, and positively related to fire age diversity within 800 m. I suggested the contrasting responses at 200 and 800 m might result from subsets of the assemblages varying in the spatial scale at which they experience the environment also varying in the degree of specialisation on fire age classes, though this remains to be tested. Alternatively, individual species experienced the environment at multiple spatial scales.

Chapters 3 and 5 are different variations on the patch mosaic burning theme, and results of both provide insight into possible plant-pollinator dynamics in the context of my spatially explicit grid-based model. Chapter 3 suggests fire mosaics of recently-burnt and long-unburnt vegetation at the scale of pollinator foraging ranges may be important for pollination of C. tentaculata. In the context of my grid-based model, pollination (assuming visitation translates into pollination) in a cell would be maximised when the cell is recently burnt (0-3 years post-fire) and neighbouring cells within pollinator flight range (approximately 500 m) are long unburnt (50+ years post-fire). Chapter 5 suggests that in order to enhance fly and wasp diversity in each cell (i.e. alpha diversity), the effects of fire mosaics should be considered at multiple scales. In the context of my model, the number of fly and wasp species arriving in a cell should be greatest when neighbouring cells within 200 m are in the same age class but there is age class diversity between cells within 800 m (though the effect would be moderate unless age class classifications could be improved).

My findings, in conjunction with a recent study from North America which found interactive effects of local-scale fire severity and landscape-scale pyrodiversity on plant- pollinator interactions (Ponisio et al. 2016), suggest fire effects on pollination can be understood within the framework of multiple processes operating over different spatial scales developed in agricultural and urban landscapes (Kremen et al. 2007; Kennedy et al. 2013). Predictions of the plant and pollinator population dynamics resulting from these processes under alternative fire management strategies and climate change scenarios could be made using the simulation model outlined in Chapter 2, provided spatially-explicit empirical data are available. These simulations could be used to identify species expected to be vulnerable under particular fire regimes.

6.2. Future directions

6.2.1. Management of (semi)natural landscapes

The management practice of patch mosaic burning aims to create landscapes with a diversity of fire histories to cater for the varying needs of a diversity of species (Parr &

Andersen 2006). Fire managers in south-eastern Australia aim to enhance species diversity because there is a growing body of empirical work describing positive relationships between species diversity and ecosystem stability (McCarthy 2012). Fire management of ecosystem stability in south-eastern Australia thus currently focuses on determining the proportion of the landscape that should be covered by each age class in order to maximise species diversity (McCarthy 2012; Di Stefano et al. 2013; Kelly et al. 2015). However, enhancing the stability of ecosystems in a landscape may require managers to do more than enhance species diversity in the landscape. The diversity- stability link is generally thought to occur when the species contributing to a given ecosystem function, such as pollination, differ in their responses to spatial and temporal environmental variation. For instance, overall pollinator abundance and pollination services have been shown to remain more stable in more species rich pollinator communities (though species identity is also important) in the face of changing temperature, wind speed, and landscape composition (Winfree & Kremen 2009; Bartomeus et al. 2013; Brittain, Kremen & Klein 2013; Cariveau et al. 2013; Rogers, Tarpy & Burrack 2014). Immobile plants flowering at particular locations in the landscape are most likely to experience the stabilising effects of pollinator diversity when that diversity exists at each location, because decline in one pollinator species cannot be compensated by other pollinator species if they are in different parts of the landscape. Thus fire age diversity is most likely to stabilise pollination if it creates local pollinator diversity, though this hypothesis remains to be tested. Chapter 5 and Ponisio et al. (2016) suggest that local pollinator diversity can be enhanced when fire history is heterogeneous within pollinator movement ranges. Chapter 5 also suggests there is a lower limit to the scale at which fire heterogeneity promotes species diversity. Thus future research aimed at providing fire managers with the knowledge required to manage ecosystem stability should determine the spatial scales at which fire heterogeneity needs to be created to promote the species interactions that often underlie the stabilising effects of biodiversity.

In addition to maintaining a diversity of post-fire age classes in the landscape, in order to promote interactions between specialised plants and their pollinators it may be necessary to maintain particular age classes and particular spatial arrangements of age classes. Vegetation burnt greater than 50 years ago appeared to be important for C.

86 tentaculata’s wasp pollinator, and having this age class adjacent to recently burnt vegetation appeared to maximise this interaction (Chapter 3). Interestingly, the Mycomya species that also appeared to depend on vegetation burnt greater than 50 years ago have recently been shown to be the sole pollinators, through sexual-deception, of the southern Australian orchid Pterostylis sanguinea (Phillips et al. 2014b). Pollination by sexual-deception of Mycomya is thought to be widespread in the Pterostylis genus (Gaskett 2011). While the flowering of many Pterostylis species is inhibited or unaffected by fire, it is enhanced or dependent on fire in other Pterostylis species (Lamont & Downes 2011; Duncan 2012). It is worth investigating whether mosaics of recently burnt vegetation adjacent to very long-unburnt vegetation enhance pollination of some Pterostylis species, and other highly specialised sexually-deceptive species, as they appear to for C. tentaculata.

6.2.2. Management of mixed (semi)natural-agricultural landscapes

Fire became less important in agricultural landscapes with the introduction of agrochemicals that serve similar functions of pest control and fertilisation (Pyne 2013); but a better understanding of fire effects on pollination and other ecosystem services might renew its importance. Remnant vegetation in agricultural landscapes enhances the magnitude and stability of pollination services to crops (Garibaldi et al. 2011; Kennedy et al. 2013; Park et al. 2015). Remnant vegetation in historically fire-prone landscapes that now include agricultural land could be better managed with fire to promote pollination (and pest control). For instance, Lonsdorf et al. (2009)’s landscape model of pollination services could be modified to include the fire histories of remnant vegetation in determining pollinator habitat suitability (based on empirical pollinator-fire history relationships).

Further benefits might result from integrating the fire-driven dynamics of remnant vegetation with the agricultural-driven dynamics of fields and pastures (e.g. crop rotations, harvesting). For instance, mass-flowering crops can dilute pollinators across the landscape and reduce visitation to simultaneously flowering plants in adjacent non- crop habitats, but when crop and non-crop flowering occurs at different times of the year (but within pollinator activity periods) or during different years they collectively provide a more stable pollinator resource (Holzschuh et al. 2011; Kovács-Hostyánszki

87 et al. 2013; Russo et al. 2013). Burning of remnant vegetation to stimulate germination and flowering, and crop rotations including mass-flowering crops in adjacent fields, could be timed so that crop and post-fire flowering occur in alternating years such that crops support large numbers of pollinators emerging during post-fire flowering and vice versa (competition would also avoided).

Fire- and agriculture-driven dynamics are analogous in that multiple processes operate over and interact between different spatial scales. For instance, pollinator recovery following local disturbance by fire or pesticide application can be enhanced by long- unburnt (Watson et al. 2012) or remnant vegetation (Park et al. 2015), respectively, in the surrounding landscape. Moreover, ecologists working in fire-prone and agricultural landscapes have both called for managing not just the visible mosaic of fire ages or current crops, but also what fire ecologist call the “invisible” mosaic – the history of recurrent fires creating a landscape with mean and variance of the proportion of land covered by each age class, and patches with different inter-fire intervals (e.g. Bradstock et al. 2005) – and agricultural ecologist the “hidden” mosaic – the history of crop management and rotation of each field in the landscape, similarly with mean and variance of land covered by each crop type and frequency of disturbance in each field (Vasseur et al. 2013). The spatially-explicit grid-based model outlined in Chapter 2 incorporates these long-term landscape dynamics (see Bradstock et al. 2005) and so could be used to simulate population dynamics and crop yield under combined fire and agricultural disturbance regimes (though the fire behaviour component of such models would also need to be updated with knowledge of fire spread between agricultural land and remnant vegetation) to determine ideal systems of integrated cropping and burning.

7. REFERENCES

Adams, P. & Lawson, S. (1993) Pollination in Australian Orchids: a Critical- Assessment of the Literature 1882-1992. Australian Journal of Botany, 41, 553-575.

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.

Andersson, P., Löfstedt, C. & Hambäck, P.A. (2013) How insects sense olfactory patches–the spatial scaling of olfactory information. Oikos, 122, 1009-1016.

Armbruster, W.S., Edwards, M.E. & Debevec, E.M. (1994) Floral character displacement generates assemblage structure of Western Australian triggerplants (Stylidium). Ecology, 75, 315-329.

Attwell, L., Kovarovic, K. & Kendal, J.R. (2015) Fire in the Plio-Pleistocene: the functions of hominin fire use, and the mechanistic, developmental and evolutionary consequences. Journal of Anthropological Sciences, 93, 1-20.

Barraclough, D. (1990) Field observations of Senostoma spp.(Diptera: Tachinidae) at Katoomba, NSW. Australian Journal of Entomology, 29, 247-252.

Bartomeus, I., Park, M.G., Gibbs, J., Danforth, B.N., Lakso, A.N. & Winfree, R. (2013) Biodiversity ensures plant–pollinator phenological synchrony against climate change. Ecology Letters, 16, 1331-1338.

Barton, K. (2014) MuMIn: Multi-model inference. R package version 1.10.5, http://CRAN.R-project.org/package=MuMIn.

Beardsell, D., Clements, M., Hutchinson, J. & Williams, E. (1986) Pollination of Diuris maculata R Br (Orchidaceae) by floral mimicry of the native legumes

Daviesia spp and Pultenaea scabra R Br. Australian Journal of Botany, 34, 165-173.

Benjamin, F.E., Reilly, J.R. & Winfree, R. (2014) Pollinator body size mediates the scale at which land use drives crop pollination services. Journal of Applied Ecology, 51, 440-449.

Bennett, A.B. & Gratton, C. (2013) Floral diversity increases beneficial arthropod richness and decreases variability in arthropod community composition. Ecological Applications, 23, 86-95.

Benson, D. (1985) Maturation periods for fire-sensitive shrub species in Hawkesbury sandstone vegetation. Cunninghamia, 1, 49.

Bernhardt, P. & Burns-Balogh, P. (1986) Floral mimesis inThelymitra nuda (Orchidaceae). Plant Systematics and Evolution, 151, 187-202.

Bird, R.B., Tayor, N., Codding, B.F. & Bird, D.W. (2013) Niche construction and Dreaming logic: aboriginal patch mosaic burning and varanid lizards (Varanus gouldii) in Australia. Proceedings of the Royal Society of London B: Biological Sciences, 280, 20132297.

Bjornstad, O.N. (2013) ncf: spatial nonparametric covariance functions. R package version 1.1-5, http://CRAN.R-project.org/package=ncf.

Blüthgen, N. & Klein, A.-M. (2011) Functional complementarity and specialisation: the role of biodiversity in plant–pollinator interactions. Basic and Applied Ecology, 12, 282-291.

Boesi, R., Polidori, C. & Andrietti, F. (2009) Searching for the right target: oviposition and feeding behavior in Bombylius bee flies (Diptera: Bombyliidae). Zoological Studies, 48, 141-150.

Bond, W. (1994) Do mutualisms matter? Assessing the impact of pollinator and disperser disruption on plant extinction. Philosophical Transactions of the Royal Society B: Biological Sciences, 344, 83-90.

Bond, W. & Van Wilgen, B. (1996) Fire and Plants, 1 edn. Springer, Netherlands.

Bond, W.J. & Scott, A.C. (2010) Fire and the spread of flowering plants in the Cretaceous. New Phytologist, 188, 1137-1150.

Bond, W.J., Woodward, F.I. & Midgley, G.F. (2005) The global distribution of ecosystems in a world without fire. New Phytologist, 165, 525-538.

Bourg, N.A., Gill, D.E. & McShea, W.J. (2014) Fire and canopy removal effects on demography and reproduction in turkeybeard (Xerophyllum asphodeloides), a rare temperate forest herb. Journal of Sustainable Forestry, 34, 71–104.

Bowen, B.J. & Pate, J.S. (2004) Effect of season of burn on shoot recovery and post‐ fire flowering performance in the resprouter Stirlingia latifolia R. Br.(Proteaceae). Austral ecology, 29, 145-155.

Bower, C. (2007) Pollinators of Threatened sexually deceptive spider orchids (Caladenia subgenus Calonema) in South West Victoria. Australian Orchid Foundation.

Bowman, D., Murphy, B.P., Burrows, G.E., Crisp, M.D., Bradstock, R., Gill, A. & Williams, R. (2012) Fire regimes and the evolution of the Australian biota. Flammable Australia: fire regimes, biodiversity and ecosystems in a changing world (ed. R. Williams, Gill, AM, Bradstock, RA), pp. 27-47. CSIRO Publishing

Bowman, D. & Panton, W. (1993) Decline of Callitris intratropica RT Baker & HG Smith in the Northern Territory: implications for pre-and post-European colonization fire regimes. Journal of biogeography, 20, 373-381.

Bowman, D.M. (2000) Australian rainforests: islands of green in a land of fire. Cambridge University Press.

Bowman, D.M., MacDermott, H.J., Nichols, S.C. & Murphy, B.P. (2014) A grass–fire cycle eliminates an obligate‐seeding tree in a tropical savanna. Ecology and evolution, 4, 4185-4194.

Bradshaw, S.D., Dixon, K.W., Hopper, S.D., Lambers, H. & Turner, S.R. (2011) Little evidence for fire-adapted plant traits in Mediterranean climate regions. Trends in plant science, 16, 69-76.

Bradstock, R., Bedward, M., Gill, A. & Cohn, J. (2005) Which mosaic? A landscape ecological approach for evaluating interactions between fire regimes, habitat and animals. Wildlife Research, 32, 409-423.

Bradstock, R., Bedward, M., Kenny, B. & Scott, J. (1998) Spatially-explicit simulation of the effect of prescribed burning on fire regimes and plant extinctions in shrublands typical of south-eastern Australia. Biological conservation, 86, 83-95.

Bradstock, R., Williams, R. & Gill, A. (2012a) Flammable Australia: fire regimes, biodiversity and ecosystems in a changing world. CSIRO Publishing.

Bradstock, R.A., Williams, R.J. & Gill, A.M. (2012b) Future fire regimes of Australian ecosystems: new perspectives on enduring questions of management. Flammable Australia: fire regimes, biodiversity and ecosystems in a changing world (ed. R. Williams, Gill, AM, Bradstock, RA), pp. 307-324. CSIRO Publishing.

Breheny, P. & Burchett, W. (2012) visreg: Visualization of regression models. R package version, 2.0-0.

Brittain, C., Kremen, C. & Klein, A.M. (2013) Biodiversity buffers pollination from changes in environmental conditions. Global Change Biology, 19, 540-547.

Bronstein, J.L. (1995) The plant—pollinator landscape. Mosaic landscapes and ecological processes, pp. 256-288. Springer.

Brosi, B.J. & Briggs, H.M. (2013) Single pollinator species losses reduce floral fidelity and plant reproductive function. Proceedings of the National Academy of Sciences, 110, 13044-13048.

Brown, G.R. & Phillips, R.D. (2014) A review of the diet of flower wasps (Hymenoptera: Thynnidae: Thynninae). Northern Territory Naturalist, 25, 50-63.

Brown, J., York, A. & Christie, F. (2016) Fire effects on pollination in a sexually deceptive orchid. International Journal of Wildland Fire, DOI: 10.1071/WF15172.

Brown, J., York, A., Christie, F. & McCarthy, M. (2016) Effects of fire on pollinators and pollination. Journal of Applied Ecology, DOI: 10.1111/1365- 2664.12670.

Brundrett, M.C. (2007) Scientific approaches to Australian temperate terrestrial orchid conservation. Australian Journal of Botany, 55, 293-307.

Burkle, L.A., Marlin, J.C. & Knight, T.M. (2013) Plant-pollinator interactions over 120 years: loss of species, co-occurrence, and function. Science, 339, 1611- 1615.

Burnham, K.P. & Anderson, D.R. (2002) Model selection and multimodel inference: a practical information-theoretic approach. Springer Science & Business Media.

Burrows, N., Wardell-Johnson, G. & Ward, B. (2008) Post-fire juvenile period of plants in south-west Australia forests and implications for fire management. Journal of the Royal Society of Western Australia, 91, 163-174.

Campbell, A. & Brown, G. (1994) Distribution of parasitoids of scarab larvae in relation to remnant vegetation: a preliminary analysis. Memoirs of the Queensland Museum. Brisbane, 36, 27-32.

Campbell, J., Hanula, J. & Waldrop, T. (2007) Effects of prescribed fire and fire surrogates on floral visiting insects of the blue ridge province in North Carolina. Biological conservation, 134, 393-404.

Cane, J.H. & Neff, J.L. (2011) Predicted fates of ground-nesting bees in soil heated by wildfire: Thermal tolerances of life stages and a survey of nesting depths. Biological conservation, 144, 2631-2636.

Cariveau, D.P., Williams, N.M., Benjamin, F.E. & Winfree, R. (2013) Response diversity to land use occurs but does not consistently stabilise ecosystem services provided by native pollinators. Ecology Letters, 16, 903-911.

Carpenter, F.L. & Recher, H.F. (1979) Pollination, reproduction, and fire. American Naturalist, 113, 871-879.

Carstensen, D.W., Sabatino, M., Trøjelsgaard, K. & Morellato, L.P.C. (2014) Beta diversity of plant-pollinator networks and the spatial turnover of pairwise interactions. PloS one, 9, e112903.

Cary, G., Bradstock, R.A., Gill, A.M. & Williams, R.J. (2012) Global change and fire regimes in Australia. Flammable Australia: fire regimes, biodiversity and ecosystems in a changing world (ed. R. Williams, Gill, AM, Bradstock, RA), pp. 149-169. CSIRO Publishing.

Chalmandrier, L., Midgley, G.F., Barnard, P. & Sirami, C. (2013) Effects of time since fire on birds in a plant diversity hotspot. Acta Oecologica, 49, 99-106.

Cheal, D.C. (2010) Growth stages and tolerable fire intervals for Victoria's native vegetation data sets. Victorian Government Department of Sustainability and Environment.

Clarke, P.J. (2002a) Habitat insularity and fire response traits: evidence from a sclerophyll archipelago. Oecologia, 132, 582-591.

Clarke, P.J. (2002b) Habitat islands in fire‐prone vegetation: do landscape features influence community composition? Journal of biogeography, 29, 677-684.

Clarke, P.J., Lawes, M., Midgley, J., Lamont, B., Ojeda, F., Burrows, G., Enright, N. & Knox, K. (2013) Resprouting as a key functional trait: how buds, protection and resources drive persistence after fire. New Phytologist, 197, 19-35.

Coates, F. & Duncan, M. (2009) Demographic variation between populations of Caladenia orientalis–a fire-managed threatened orchid. Australian Journal of Botany, 57, 326-339.

Coates, F., Lunt, I.D. & Tremblay, R.L. (2006) Effects of disturbance on population dynamics of the threatened orchid Prasophyllum correctum DL Jones and implications for grassland management in south-eastern Australia. Biological conservation, 129, 59-69.

Crisp, M.D., Burrows, G.E., Cook, L.G., Thornhill, A.H. & Bowman, D.M. (2011) Flammable biomes dominated by eucalypts originated at the Cretaceous- Palaeogene boundary. Nature communications, 2, 193.

Cropper, S. & Calder, D. (1990) The floral biology ofThelymitra epipactoides (Orchidaceae), and the implications of pollination by deceit on the survival of this rare orchid. Plant Systematics and Evolution, 170, 11-27.

Cunningham, S.A. (2000) Depressed pollination in habitat fragments causes low fruit set. Proceedings of the Royal Society of London. Series B: Biological Sciences, 267, 1149-1152.

Dafni, A. & Bernhardt, P. (1990) Pollination of terrestrial orchids of southern Australia and the Mediterranean region. Systematic, ecological, and evolutionary implications. Evol. Biol, 24, 193-252.

Dafni, A. & Calder, D. (1987) Pollination by deceit and floral mimesis inThelymitra antennifera (Orchidaceae). Plant Systematics and Evolution, 158, 11-22.

Dafni, A., Izhaki, I. & Ne'eman, G. (2012) The effect of fire on biotic interactions in Mediterranean basin ecosystems: pollination and seed dispersal. Israel Journal of Ecology & Evolution, 58, 235-250.

Davidson, R., Wiseman, J. & Wolfe, V. (1972) Environmental stress in the pasture scarab Sericesthis nigrolineata Boisd. II. Effects of soil moisture and temperature on survival of first-instar larvae. Journal of Applied Ecology, 9, 799-806.

De Swardt, D. (1993) Factors affecting the densities of nectarivors in Protea roupeuae woodland. Ostrich, 64, 172-177.

Di Stefano, J., McCarthy, M.A., York, A., Duff, T.J., Slingo, J. & Christie, F. (2013) Defining vegetation age class distributions for multispecies conservation in fire-prone landscapes. Biological conservation, 166, 111-117.

Di Stefano, J., Owen, L., Morris, R., Duff, T. & York, A. (2011) Fire, landscape change and models of small mammal habitat suitability at multiple spatial scales. Austral ecology, 36, 638-649.

Dickinson, C.H. (2012) Biology of plant litter decomposition. Elsevier.

Dixon, K. & Tremblay, R.L. (2009) Biology and natural history of Caladenia. Australian Journal of Botany, 57, 247-258.

Dodson, J. (2001) Holocene vegetation change in the mediterranean-type climate regions of Australia. The Holocene, 11, 673-680.

Dorado, J. & Vázquez, D.P. (2014) The diversity–stability relationship in floral production. Oikos, 123, 1137-1143.

Draney, M.L. & Crossley Jr, D. (1999) Relationship of habitat age to phenology among ground-dwelling Linyphiidae (Araneae) in the Southeastern United States. Journal of Arachnology, 27, 211-216.

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. (2010) Fire management

for biodiversity conservation: key research questions and our capacity to answer them. Biological conservation, 143, 1928-1939.

Duff, T.J. (2010) Modelling vegetation responses to environmental influences in a natural forest system. PhD, University of Melbourne.

Duncan, M. (2012) Response of Orchides to Bushfire: Black Saturday Victoria 2009: Natural Values Fire Recovery Program. Department of Sustainability and Environment Heidelberg, VIC, Australia.

Ebeling, A., Klein, A.-M. & Tscharntke, T. (2011) Plant–flower visitor interaction webs: temporal stability and pollinator specialization increases along an experimental plant diversity gradient. Basic and Applied Ecology, 12, 300- 309.

Eckert, C.G., Kalisz, S., Geber, M.A., Sargent, R., Elle, E., Cheptou, P.-O., Goodwillie, C., Johnston, M.O., Kelly, J.K. & Moeller, D.A. (2010) Plant mating systems in a changing world. Trends in ecology & evolution, 25, 35-43.

Enright, N., Fontaine, J., Westcott, V., Lade, J. & Miller, B. (2011) Fire interval effects on persistence of resprouter species in Mediterranean-type shrublands. Plant Ecol, 212, 2071-2083.

Enright, N. & Goldblum, D. (1999) Demography of a non-sprouting and resprouting Hakea species (Proteaceae) in fire-prone Eucalyptus woodlands of southeastern Australia in relation to stand age, drought and disease. Plant Ecology, 144, 71-82.

Essenberg, C.J. (2013) Scale-dependent shifts in the species composition of flower visitors with changing floral density. Oecologia, 171, 187-196.

Fahrig, L., Baudry, J., Brotons, L., Burel, F.G., Crist, T.O., Fuller, R.J., Sirami, C., Siriwardena, G.M. & Martin, J.L. (2011) Functional landscape heterogeneity and animal biodiversity in agricultural landscapes. Ecology Letters, 14, 101- 112.

Farnsworth, L.M., Nimmo, D.G., Kelly, L.T., Bennett, A.F. & Clarke, M.F. (2014) Does pyrodiversity beget alpha, beta or gamma diversity? A case study using reptiles from semi‐arid Australia. Diversity and distributions, 20, 663-673.

Flannigan, M., Cantin, A.S., de Groot, W.J., Wotton, M., Newbery, A. & Gowman, L.M. (2013) Global wildland fire season severity in the 21st century. Forest Ecology and Management, 294, 54-61.

Fournier, D.A., Skaug, H.J., Ancheta, J., Ianelli, J., Magnusson, A., Maunder, M.N., Nielsen, A. & Sibert, J. (2012) AD Model Builder: using automatic differentiation for statistical inference of highly parameterized complex nonlinear models. Optimization Methods and Software, 27, 233-249.

Fox, B.J., Taylor, J.E. & Thompson, P.T. (2003) Experimental manipulation of habitat structure: a retrogression of the small mammal succession. Journal of animal ecology, 72, 927-940.

Franklin, J., Syphard, A.D., Mladenoff, D.J., He, H.S., Simons, D.K., Martin, R.P., Deutschman, D. & O'Leary, J.F. (2001) Simulating the effects of different fire regimes on plant functional groups in Southern California. Ecological Modelling, 142, 261-283.

Fründ, J., Dormann, C.F., Holzschuh, A. & Tscharntke, T. (2013) Bee diversity effects on pollination depend on functional complementarity and niche shifts. Ecology, 94, 2042-2054.

Garibaldi, L.A., Steffan‐Dewenter, I., Kremen, C., Morales, J.M., Bommarco, R., Cunningham, S.A., Carvalheiro, L.G., Chacoff, N.P., Dudenhoeffer, J.H. & Greenleaf, S.S. (2011) Stability of pollination services decreases with isolation from natural areas despite honey bee visits. Ecology Letters, 14, 1062-1072.

Gaskett, A. (2011) Orchid pollination by sexual deception: pollinator perspectives. Biological Reviews, 86, 33-75.

Geerts, S., Malherbe, S.D. & Pauw, A. (2012) Reduced flower visitation by nectar- feeding birds in response to fire in Cape fynbos vegetation, South Africa. Journal of Ornithology, 153, 297-301.

(2008) Responses of parasitoids to saproxylic hosts and habitat: a multi- Gibb, H.,scale Hilszczański, study using J., experimental Hjältén, J., Danell, logs. Oecologia, K., Ball, J.,155, Pettersson, 63-74. R. & Alinvi, O.

Gibbons, F.R. & Downes, R.G. (1964) A study of the land in south-western Victoria. Soil Conservation Authority Melbourne.

Gigord, L.D., Macnair, M., Stritesky, M. & Smithson, A. (2002) The potential for floral mimicry in rewardless orchids: an experimental study. Proceedings of the Royal Society of London B: Biological Sciences, 269, 1389-1395.

Gill, A.M. (1975) Fire and the Australian flora: a review. Australian forestry, 38, 4- 25.

Gill, A.M. & McCarthy, M.A. (1998) Intervals between prescribed fires in Australia: what intrinsic variation should apply? Biological conservation, 85, 161-169.

Glasspool, I.J., Edwards, D. & Axe, L. (2004) Charcoal in the Silurian as evidence for the earliest wildfire. Geology, 32, 381-383.

Gould, S.J. & Vrba, E.S. (1982) Exaptation—a missing term in the science of form. Paleobiology, 8, 4-15.

Greenleaf, S.S., Williams, N.M., Winfree, R. & Kremen, C. (2007) Bee foraging ranges and their relationship to body size. Oecologia, 153, 589-596.

Hassan, S. (1975) Effects of high temperature and soil moisture on survival of first- instar larvae of the scarab Anoplognathus porosus (Dalman)(Coleoptera). Journal of Applied Ecology, 12, 749-754.

Haubruge, É., Almohamad, R. & Verheggen, F.J. (2009) Searching and oviposition behavior of aphidophagous hoverflies (Diptera: Syrphidae): a review. Base, 13.

He, T., Pausas, J.G., Belcher, C.M., Schwilk, D.W. & Lamont, B.B. (2012) Fire‐adapted traits of Pinus arose in the fiery Cretaceous. New Phytologist, 194, 751-759.

Hegland, S.J. (2014) Floral neighbourhood effects on pollination success in red clover are scale‐dependent. Functional Ecology, 28, 561-568.

Hiers, J.K., Wyatt, R. & Mitchell, R.J. (2000) The effects of fire regime on legume reproduction in longleaf pine savannas: is a season selective? Oecologia, 125, 521-530.

J.P. & Danell, K. (2005) Parasitoids (Hymenoptera, Ichneumonoidea) of Hilszczański,saproxylic J., Gibb, beetles H., Hjältén, are affected J., Atlegrim, by forest O., Johansson,successional T., stage Pettersson, and dead R.B., wood Ball, characteristics in boreal spruce forest. Biological conservation, 126, 456- 464.

Hingston, A.B. & McQuillan, P.B. (2000) Are pollination syndromes useful predictors of floral visitors in Tasmania? Austral ecology, 25, 600-609.

Hoehn, P., Tscharntke, T., Tylianakis, J.M. & Steffan-Dewenter, I. (2008) Functional group diversity of bee pollinators increases crop yield. Proceedings of the Royal Society of London B: Biological Sciences, 275, 2283-2291.

Holzschuh, A., Dormann, C.F., Tscharntke, T. & Steffan-Dewenter, I. (2011) Expansion of mass-flowering crops leads to transient pollinator dilution and reduced wild plant pollination. Proceedings of the Royal Society of London B: Biological Sciences, rspb20110268.

Hopfenmueller, S., Steffan-Dewenter, I. & Holzschuh, A. (2014) Trait-Specific Responses of Wild Bee Communities to Landscape Composition, Configuration and Local Factors. PloS one, 9, e104439.

Hopper, S.D. & Brown, A.P. (2007) A revision of Australia's hammer orchids (Drakaea: Orchidaceae), with some field data on species-specific sexually deceived wasp pollinators. Australian Systematic Botany, 20, 252-285.

Indsto, J.O., Weston, P.H., Clements, M.A., Dyer, A.G., Batley, M. & Whelan, R.J. (2006) Pollination of Diuris maculata (Orchidaceae) by male Trichocolletes venustus bees. Australian Journal of Botany, 54, 669-679.

Internicola, A.I., Juillet, N., Smithson, A. & Gigord, L.D. (2006) Experimental investigation of the effect of spatial aggregation on reproductive success in a rewardless orchid. Oecologia, 150, 435-441.

Internicola, A.I., Page, P.A., Bernasconi, G. & Gigord, L.D. (2007) Competition for pollinator visitation between deceptive and rewarding artificial inflorescences: an experimental test of the effects of floral colour similarity and spatial mingling. Functional Ecology, 21, 864-872.

Jasinge, N. (2014) The effect of seasonal burning on three Australian native orchids. MSc Thesis, RMIT University.

Jauker, B., Krauss, J., Jauker, F. & Steffan-Dewenter, I. (2013) Linking life history traits to pollinator loss in fragmented calcareous grasslands. Landscape Ecology, 28, 107-120.

Jauker, F., Diekoetter, T., Schwarzbach, F. & Wolters, V. (2009) Pollinator dispersal in an agricultural matrix: opposing responses of wild bees and hoverflies to landscape structure and distance from main habitat. Landscape Ecology, 24, 547-555.

Jersáková, J., Johnson, S.D. & Kindlmann, P. (2006) Mechanisms and evolution of deceptive pollination in orchids. Biological Reviews, 81, 219-235.

Jervis, M., Kidd, N., Fitton, M., Huddleston, T. & Dawah, H. (1993) Flower-visiting by hymenopteran parasitoids. Journal of natural history, 27, 67-105.

Johnson, S., Linder, H. & Steiner, K. (1998) Phylogeny and radiation of pollination systems in Disa (Orchidaceae). American Journal of Botany, 85, 402-402.

Johnson, S.D., Peter, C.I., Nilsson, L.A. & Ågren, J. (2003) Pollination success in a deceptive orchid is enhanced by co-occurring rewarding magnet plants. Ecology, 84, 2919-2927.

Jolly, W.M., Cochrane, M.A., Freeborn, P.H., Holden, Z.A., Brown, T.J., Williamson, G.J. & Bowman, D.M. (2015) Climate-induced variations in global wildfire danger from 1979 to 2013. Nature communications, 6, 7537.

Jones, D.L. (2006) Native orchids of Australia. Reed Natural History Australia.

Jordano, P. (1987) Patterns of mutualistic interactions in pollination and seed dispersal: connectance, dependence asymmetries, and coevolution. American Naturalist, 129, 657-677.

Junker, R.R., Blüthgen, N., Brehm, T., Binkenstein, J., Paulus, J., Schaefer, H.M. & Stang, M. (2013) Specialization on traits as basis for the niche-breadth of flower visitors and as structuring mechanism of ecological networks. Functional Ecology, 27, 329-341.

Kaiser-Bunbury, C.N. & Blüthgen, N. (2015) Integrating network ecology with applied conservation: a synthesis and guide to implementation. AoB Plants, 7, plv076.

Kaiser-Bunbury, C.N., Vázquez, D.P., Stang, M. & Ghazoul, J. (2014) Determinants of the microstructure of plant–pollinator networks. Ecology, 95, 3314-3324.

Keeley, J.E., Bond, W.J., Bradstock, R.A., Pausas, J.G. & Rundel, P.W. (2011a) Fire in Mediterranean ecosystems: ecology, evolution and management. Cambridge University Press.

Keeley, J.E., Pausas, J.G., Rundel, P.W., Bond, W.J. & Bradstock, R.A. (2011b) Fire as an evolutionary pressure shaping plant traits. Trends in plant science, 16, 406-411.

Keeley, J.E. & Rundel, P.W. (2005) Fire and the Miocene expansion of C4 grasslands. Ecology Letters, 8, 683-690.

Keith, D. (1996) Fire-driven extinction of plant populations: a synthesis of theory and review of evidence from Australian vegetation. Proceedings-Linnean Society of New South Wales, pp. 37-78. LINNEAN SOCIETY OF NEW SOUTH WALES.

Kelly, L., Nimmo, D., Spence-Bailey, L., Taylor, R., Watson, S., Clarke, M. & Bennett, A. (2012) Relative influence of the properties of fire mosaics on small mammals: a landscape perspective. Journal of Applied Ecology, 49, 412-421.

Kelly, L.T., Bennett, A.F., Clarke, M.F. & McCarthy, M.A. (2015) Optimal fire histories for biodiversity conservation. Conservation Biology, 29, 473-481.

Kelsey, J. (1968) Oviposition preference by Costelytra zealandica (White) II. New Zealand journal of agricultural research, 11, 206-210.

Kennedy, C.M., Lonsdorf, E., Neel, M.C., Williams, N.M., Ricketts, T.H., Winfree, R., Bommarco, R., Brittain, C., Burley, A.L. & Cariveau, D. (2013) A global quantitative synthesis of local and landscape effects on wild bee pollinators in agroecosystems. Ecology Letters, 16, 584-599.

Kleijn, D. & Van Langevelde, F. (2006) Interacting effects of landscape context and habitat quality on flower visiting insects in agricultural landscapes. Basic and Applied Ecology, 7, 201-214.

Knox, K. & Clarke, P. (2006) Fire season and intensity affect shrub recruitment in temperate sclerophyllous woodlands. Oecologia, 149, 730-739.

Knox, K.J. & Clarke, P.J. (2004) Fire response syndromes of shrubs in grassy woodlands in the New England Tableland Bioregion. Cunninghamia, 8, 348- 353.

Kovács-Hostyánszki, A., Haenke, S., Batáry, P., Jauker, B., Báldi, A., Tscharntke, T. & Holzschuh, A. (2013) Contrasting effects of mass-flowering crops on bee pollination of hedge plants at different spatial and temporal scales. Ecological Applications, 23, 1938-1946.

Kremen, C. (2005) Managing ecosystem services: what do we need to know about their ecology? Ecology Letters, 8, 468-479.

Kremen, C., Williams, N.M., Aizen, M.A., Gemmill‐Herren, B., LeBuhn, G., Minckley, R., Packer, L., Potts, S.G., Steffan‐Dewenter, I. & Vazquez, D.P. (2007) Pollination and other ecosystem services produced by mobile organisms: a conceptual framework for the effects of land‐use change. Ecology Letters, 10, 299-314.

Laanisto, L., Tamme, R., Hiiesalu, I., Szava-Kovats, R., Gazol, A. & Pärtel, M. (2013) Microfragmentation concept explains non-positive environmental heterogeneity–diversity relationships. Oecologia, 171, 217-226.

Lammi, A. & Kuitunen, M. (1995) Deceptive pollination of Dactylorhiza incarnata: an experimental test of the magnet species hypothesis. Oecologia, 101, 500- 503.

Lamont, B.B. & Downes, K.S. (2011) Fire-stimulated flowering among resprouters and geophytes in Australia and South Africa. Plant Ecology, 212, 2111-2125.

Lamont, B.B., Enright, N.J. & He, T. (2011) Fitness and evolution of resprouters in relation to fire. Plant Ecology, 212, 1945-1957.

Lamont, B.B., Olesen, J.M. & Briffa, P.J. (1998) Seed production, pollinator attractants and breeding system in relation to fire response—are there reproductive syndromes among co-occurring Proteaceous shrubs? Australian Journal of Botany, 46, 377-385.

Lamont, B.B., Wittkuhn, R. & Korczynskyj, D. (2004) TURNER REVIEW No. 8. Ecology and ecophysiology of grasstrees. Australian Journal of Botany, 52, 561-582.

Larsen, T.H., Williams, N.M. & Kremen, C. (2005) Extinction order and altered community structure rapidly disrupt ecosystem functioning. Ecology Letters, 8, 538-547.

Larson, B., Kevan, P. & Inouye, D. (2001) Flies and flowers: taxonomic diversity of anthophiles and pollinators. The Canadian Entomologist, 133, 439-465.

100

Lassau, S.A. & Hochuli, D.F. (2007) Associations between wasp communities and forest structure: Do strong local patterns hold across landscapes? Austral ecology, 32, 656-662.

Laverty, T.M. (1992) Plant interactions for pollinator visits: a test of the magnet species effect. Oecologia, 89, 502-508.

Lavorel, S., Storkey, J., Bardgett, R.D., Bello, F., Berg, M.P., Roux, X., Moretti, M., Mulder, C., Pakeman, R.J. & Díaz, S. (2013) A novel framework for linking functional diversity of plants with other trophic levels for the quantification of ecosystem services. Journal of Vegetation Science, 24, 942-948.

(2013) Solitary bee abundance and species richness in dynamic agricultural Le Féon,landscapes. V., Burel, F.,Agriculture, Chifflet, R., Ecosystems Henry, M., & Ricroch,Environment, A., Vaissière, 166, 94 -B.E.101. & Baudry, J.

Le Maitre, D. & Brown, P. (1992) Life cycles and fire-stimulated flowering in geophytes. Fire in South African mountain fynbos, pp. 145-160. Springer.

Legendre, P. & Legendre, L.F. (2012) Numerical ecology. Elsevier.

Lindenmayer, D.B., Wood, J.T., MacGregor, C., Michael, D.R., Cunningham, R.B., Crane, M., Montague‐Drake, R., Brown, D., Muntz, R. & Driscoll, D.A. (2008) How predictable are reptile responses to wildfire? Oikos, 117, 1086-1097.

Lonsdorf, E., Kremen, C., Ricketts, T., Winfree, R., Williams, N. & Greenleaf, S. (2009) Modelling pollination services across agricultural landscapes. Annals of botany, 103, 1589–1600.

Lopezaraiza–Mikel, M.E., Hayes, R.B., Whalley, M.R. & Memmott, J. (2007) The impact of an alien plant on a native plant–pollinator network: an experimental approach. Ecology Letters, 10, 539-550.

MacArthur, R.H. & MacArthur, J.W. (1961) On bird species diversity. Ecology, 42, 594-598.

MacHunter, J., Menkhorst, P. & Loyn, R. (2009) Towards a process for integrating vertebrate fauna into fire management planning. Arthur Rylah Institute for Environmental Research, Department of Sustainability and Environment.

Maleque, M.A., Maeto, K., MAKINO, S.I., Goto, H., Tanaka, H., Hasegawa, M. & Miyamoto, A. (2010) A chronosequence of understorey parasitic wasp assemblages in secondary broad‐leaved forests in a Japanese ‘satoyama’landscape. Insect Conservation and Diversity, 3, 143-151.

Marino, P.C., Landis, D.A. & Hawkins, B.A. (2006) Conserving parasitoid assemblages of North American pest Lepidoptera: Does biological control by native parasitoids depend on landscape complexity? Biological Control, 37, 173-185.

101

Marlon, J.R., Bartlein, P.J., Carcaillet, C., Gavin, D.G., Harrison, S.P., Higuera, P.E., Joos, F., Power, M. & Prentice, I. (2008) Climate and human influences on global biomass burning over the past two millennia. Nature Geoscience, 1, 697- 702.

Mateos, E., Santos, X. & Pujade-Villar, J. (2011) Taxonomic and functional responses to fire and post-fire management of a Mediterranean Hymenoptera community. Environmental management, 48, 1000-1012.

McCarthy, M. (2012) Review of resilience concepts and their measurement for fire management. Fire and adaptive management, report no. 90 Department of Sustainability and Environment, Victorian State Government, Melbourne, Australia.

McElhinny, C., Gibbons, P., Brack, C. & Bauhus, J. (2005) Forest and woodland stand structural complexity: its definition and measurement. Forest Ecology and Management, 218, 1-24.

McElhinny, C., Gibbons, P., Brack, C. & Bauhus, J. (2006) Fauna-habitat relationships: a basis for identifying key stand structural attributes in temperate Australian eucalypt forests and woodlands. Pacific Conservation Biology, 12, 89-110.

McKenzie, D., Miller, C. & Falk, D.A. (2011) The landscape ecology of fire. Springer Science & Business Media.

Menz, M.H., Phillips, R.D., Dixon, K.W., Peakall, R. & Didham, R.K. (2013) Mate- searching behaviour of common and rare wasps and the implications for pollen movement of the sexually deceptive orchids they pollinate. PloS one, 8, e59111.

Meyer, B., Jauker, F. & Steffan-Dewenter, I. (2009) Contrasting resource-dependent responses of hoverfly richness and density to landscape structure. Basic and Applied Ecology, 10, 178-186.

Midgley, J. & Bond, W. (2011) Pushing back in time: the role of fire in plant evolution. New Phytologist, 191, 5-7.

Millennium Ecosystem Assessment (2005) Millennium ecosystem assessment: ecosystems and human well-being (biodiversity synthesis). 2005.

Monamy, V. & Fox, B. (2000) Small mammal succession is determined by vegetation density rather than time elapsed since disturbance. Austral ecology, 25, 580-587.

Monamy, V. & Fox, B.J. (2010) Responses of two species of heathland rodents to habitat manipulation: vegetation density thresholds and the habitat accommodation model. Austral ecology, 35, 334-347.

102

Moreira, B., Tormo, J. & Pausas, J.G. (2012) To resprout or not to resprout: factors driving intraspecific variability in resprouting. Oikos, 121, 1577-1584.

Moreira, E., Boscolo, D. & Viana, B. (2015) Spatial Heterogeneity Regulates Plant- Pollinator Networks across Multiple Landscape Scales. PloS one, 10, e0123628.

Moretti, M., De Bello, F., Roberts, S.P. & Potts, S.G. (2009) Taxonomical vs. functional responses of bee communities to fire in two contrasting climatic regions. Journal of animal ecology, 78, 98-108.

Mouillot, F. & Field, C.B. (2005) Fire history and the global carbon budget: a 1× 1 fire history reconstruction for the 20th century. Global Change Biology, 11, 398-420.

Murdoch, W., Briggs, C.J. & Swarbrick, S. (2005) Host suppression and stability in a parasitoid-host system: experimental demonstration. Science, 309, 610- 613.

Ne'eman, G., Dafni, A. & Potts, S.G. (2000) The effect of fire on flower visitation rate and fruit set in four core-species in east Mediterranean scrubland. Plant Ecology, 146, 97-104.

New, T., Yen, A., Sands, D., Greenslade, P., Neville, P., York, A. & Collett, N. (2010) Planned fires and invertebrate conservation in south east Australia. Journal of insect conservation, 14, 567-574.

Newman, B.J., Ladd, P., Brundrett, M. & Dixon, K.W. (2013) Effects of habitat fragmentation on plant reproductive success and population viability at the landscape and habitat scale. Biological conservation, 159, 16-23.

Newman, E., Anderson, B. & Johnson, S.D. (2012) Flower colour adaptation in a mimetic orchid. Proceedings of the Royal Society of London B: Biological Sciences, rspb20112375.

Nimmo, D., Kelly, L., Farnsworth, L., Watson, S. & Bennett, A. (2014) Why do some species have geographically varying responses to fire history? Ecography, 37, 805-813.

Noble, I.R. & Slatyer, R. (1980) The use of vital attributes to predict successional changes in plant communities subject to recurrent disturbances. Succession, pp. 5-21. Springer.

Ollerton, J., Winfree, R. & Tarrant, S. (2011) How many flowering plants are pollinated by animals? Oikos, 120, 321-326.

Ooi, M.K. (2010) Delayed emergence and post-fire recruitment success: effects of seasonal germination, fire season and dormancy type. Australian Journal of Botany, 58, 248-256.

103

Orford, K.A., Vaughan, I.P. & Memmott, J. (2015) The forgotten flies: the importance of non-syrphid Diptera as pollinators. Proceedings of the Royal Society of London B: Biological Sciences, 282, 20142934.

Pacifici, M., Visconti, P., Scepi, E., Hausmann, A., Attorre, F., Grant, R. & Rondinini, C. (2015) Fire policy optimization to maximize suitable habitat for locally rare species under different climatic conditions: A case study of antelopes in the Kruger National Park. Biological conservation, 191, 313-321.

Park, M.G., Blitzer, E., Gibbs, J., Losey, J.E. & Danforth, B.N. (2015) Negative effects of pesticides on wild bee communities can be buffered by landscape context. Proceedings of the Royal Society B: Biological Sciences, 282, 20150299.

Parker, C.H., Keefe, E.R., Herzog, N.M., O'connell, J.F. & Hawkes, K. (2016) The pyrophilic primate hypothesis. Evolutionary Anthropology: Issues, News, and Reviews, 25, 54-63.

Parks Victoria (2003) Grampians National Park Management Plan. Parks Victoria.

Parr, C.L. & Andersen, A.N. (2006) Patch mosaic burning for biodiversity conservation: a critique of the pyrodiversity paradigm. Conservation Biology, 20, 1610-1619.

Parsche, S., Fründ, J. & Tscharntke, T. (2011) Experimental environmental change and mutualistic vs. antagonistic plant flower–visitor interactions. Perspectives in Plant Ecology, Evolution and Systematics, 13, 27-35.

Pate, J.S. & Dixon, K.W. (1982) Tuberous, cormous and bulbous plants: Biology of an adaptive strategy in Western Australia. University of Western Australia Press.

Pausas, J.G. & Keeley, J.E. (2009) A burning story: the role of fire in the history of life. BioScience, 59, 593-601.

Pausas, J.G. & Schwilk, D. (2012) Fire and plant evolution. New Phytologist, 193, 301-303.

Pauw, A. (2007) Collapse of a pollination web in small conservation areas. Ecology, 88, 1759-1769.

Pauw, A. & Bond, W.J. (2011) Mutualisms matter: pollination rate limits the distribution of oil‐secreting orchids. Oikos, 120, 1531-1538.

Pauw, A. & Hawkins, J.A. (2011) Reconstruction of historical pollination rates reveals linked declines of pollinators and plants. Oikos, 120, 344-349.

104

Pavlovic, N.B., Leicht-Young, S.A. & Grundel, R. (2011) Short-term effects of burn season on flowering phenology of savanna plants. Plant Ecology, 212, 611- 625.

Peakall, R. (1990) Responses of male Zaspilothynnus trilobatus Turner wasps to females and the sexually deceptive orchid it pollinates. Functional Ecology, 4, 159-167.

Peakall, R. & Beattie, A.J. (1996) Ecological and genetic consequences of pollination by sexual deception in the orchid Caladenia tentactulata. Evolution, 50, 2207-2220.

Peakall, R., Ebert, D., Poldy, J., Barrow, R.A., Francke, W., Bower, C.C. & Schiestl, F.P. (2010) Pollinator specificity, floral odour chemistry and the phylogeny of Australian sexually deceptive Chiloglottis orchids: implications for pollinator‐driven speciation. New Phytologist, 188, 437-450.

Penman, T., Christie, F., Andersen, A., Bradstock, R., Cary, G., Henderson, M., Price, O., Tran, C., Wardle, G. & Williams, R. (2011) Prescribed burning: how can it work to conserve the things we value? International Journal of Wildland Fire, 20, 721-733.

Petit, S. & Dickson, C.R. (2005) Grass-tree (Xanthorrhoea semiplana, Liliaceae) facilitation of the endangered pink-lipped spider orchid (Caladenia syn. Arachnorchis behrii, Orchidaceae) varies in South Australia. Australian Journal of Botany, 53, 455-464.

Phillips, R.D., Faast, R., Bower, C.C., Brown, G.R. & Peakall, R. (2009) Implications of pollination by food and sexual deception for pollinator specificity, fruit set, population genetics and conservation of Caladenia (Orchidaceae). Australian Journal of Botany, 57, 287-306.

Phillips, R.D., Peakall, R., Hutchinson, M.F., Linde, C.C., Xu, T., Dixon, K.W. & Hopper, S.D. (2014a) Specialized ecological interactions and plant species rarity: The role of pollinators and mycorrhizal fungi across multiple spatial scales. Biological conservation, 169, 285-295.

Phillips, R.D., Peakall, R., Retter, B.A., Montgomery, K., Menz, M.H., Davis, B.J., Hayes, C., Brown, G.R., Swarts, N.D. & Dixon, K.W. (2015) Pollinator rarity as a threat to a plant with a specialized pollination system. Botanical Journal of the Linnean Society, 179, 511-525.

Phillips, R.D., Scaccabarozzi, D., Retter, B.A., Hayes, C., Brown, G.R., Dixon, K.W. & Peakall, R. (2014b) Caught in the act: pollination of sexually deceptive trap- flowers by fungus gnats in Pterostylis (Orchidaceae). Annals of botany, 113, 629-641.

Platt, W.J., Evans, G.W. & Davis, M.M. (1988) Effects of fire season on flowering of forbs and shrubs in longleaf pine forests. Oecologia, 76, 353-363.

105

Poisot, T., Stouffer, D.B. & Gravel, D. (2015) Beyond species: why ecological interaction networks vary through space and time. Oikos, 124, 243-251.

Ponisio, L.C., Wilkin, K., M'Gonigle, L.K., Kulhanek, K., Cook, L., Thorp, R., Griswold, T. & Kremen, C. (2016) Pyrodiversity begets plant–pollinator community diversity. Global Change Biology, 22, 1794-1808.

Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O. & Kunin, W.E. (2010) Global pollinator declines: trends, impacts and drivers. Trends in ecology & evolution, 25, 345-353.

Potts, S.G., Dafni, A. & Ne'eman, G. (2001) Pollination of a core flowering shrub species in Mediterranean phrygana: variation in pollinator diversity, abundance and effectiveness in response to fire. Oikos, 92, 71-80.

Potts, S.G., Petanidou, T., Roberts, S., O’Toole, C., Hulbert, A. & Willmer, P. (2006) Plant-pollinator biodiversity and pollination services in a complex Mediterranean landscape. Biological conservation, 129, 519-529.

Potts, S.G., Vulliamy, B., Dafni, A., Ne'eman, G., O'Toole, C., Roberts, S. & Willmer, P. (2003a) Response of plant‐pollinator communities to fire: changes in diversity, abundance and floral reward structure. Oikos, 101, 103-112.

Potts, S.G., Vulliamy, B., Dafni, A., Ne'eman, G. & Willmer, P. (2003b) Linking bees and flowers: how do floral communities structure pollinator communities? Ecology, 84, 2628-2642.

Potts, S.G., Vulliamy, B., Roberts, S., O'Toole, C., Dafni, A., Ne'eman, G. & Willmer, P. (2005) Role of nesting resources in organising diverse bee communities in a Mediterranean landscape. Ecological Entomology, 30, 78-85.

Price, P.W. (1991) Evolutionary theory of host and parasitoid interactions. Biological Control, 1, 83-93.

Pyne, S.J. (2013) Fire: nature and culture. Reaktion Books.

Quinn, G.P. & Keough, M.J. (2002) Experimental design and data analysis for biologists. Cambridge University Press.

R Core Team (2014) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Ramsey, M. & Vaughton, G. (2000) Pollen quality limits seed set in Burchardia umbellata (Colchicaceae). American Journal of Botany, 87, 845-852.

Rathcke, B. (1988) Interactions for pollination among coflowering shrubs. Ecology, 69, 446-457.

106

Regos, A., D’Amen, M., Herrando, S., Guisan, A. & Brotons, L. (2015) Fire management, climate change and their interacting effects on birds in complex Mediterranean landscapes: dynamic distribution modelling of an early-successional species—the near-threatened Dartford Warbler (Sylvia undata). Journal of Ornithology, 156, 275-286.

Reiter, N., Whitfield, J., Pollard, G., Bedggood, W., Argall, M., Dixon, K., Davis, B. & Swarts, N. (2016) Orchid re-introductions: an evaluation of success and ecological considerations using key comparative studies from Australia. Plant Ecology, 217, 81-95.

Renner, S.S. (2006) Rewardless flowers in the angiosperms and the role of insect cognition in their evolution. University of Chicago Press.

Ridsdill-Smith, T. (1970) behaviour of Hemithynnus hyalinatus (Hymenoptera: Tiphiidae), with notes on some other Thynninae. Journal of the Australian Entomological Society, 9, 183-195.

Robinson, N.M., Leonard, S.W., Ritchie, E.G., Bassett, M., Chia, E.K., Buckingham, S., Gibb, H., Bennett, A.F. & Clarke, M.F. (2013) REVIEW: Refuges for fauna in fire‐prone landscapes: their ecological function and importance. Journal of Applied Ecology, 50, 1321-1329.

Rodríguez, A. & Kouki, J. (2015) Emulating natural disturbance in forest management enhances pollination services for dominant Vaccinium shrubs in boreal pine-dominated forests. Forest Ecology and Management, 350, 1- 12.

Rogers, S.R., Tarpy, D.R. & Burrack, H.J. (2014) Bee species diversity enhances productivity and stability in a perennial crop. PloS one, 9, e97307.

Roland, J. & Taylor, P.D. (1997) Insect parasitoid species respond to forest structure at different spatial scales. Nature, 386, 710-713.

Rosenzweig, M.L. (1995) Species diversity in space and time. Cambridge University Press.

Rubene, D., Schroeder, M. & Ranius, T. (2015) Diversity patterns of wild bees and wasps in managed boreal forests: effects of spatial structure, local habitat and surrounding landscape. Biological conservation, 184, 201-208.

Russo, L., DeBarros, N., Yang, S., Shea, K. & Mortensen, D. (2013) Supporting crop pollinators with floral resources: network‐based phenological matching. Ecology and evolution, 3, 3125-3140.

Schiegg, K. (2000) Effects of dead wood volume and connectivity on saproxylic insect species diversity. Ecoscience, 7, 290-298.

107

Schiestl, F.P., Ayasse, M., Paulus, H.F., Löfstedt, C., Hansson, B.S., Ibarra, F. & Francke, W. (1999) Orchid pollination by sexual swindle. Nature, 399, 421- 421.

Schleuning, M., Fründ, J. & García, D. (2015) Predicting ecosystem functions from biodiversity and mutualistic networks: an extension of trait‐based concepts to plant–animal interactions. Ecography, 38, 380-392.

Schmid, B., Nottebrock, H., Esler, K., Pagel, J., Pauw, A., Böhning‐Gaese, K., Schurr, F. & Schleuning, M. (2015) Responses of nectar‐feeding birds to floral resources at multiple spatial scales. Ecography, DOI: 10.1111/ecog.01621.

Scott, A.C. & Glasspool, I.J. (2006) The diversification of Paleozoic fire systems and fluctuations in atmospheric oxygen concentration. Proceedings of the National Academy of Sciences, 103, 10861-10865.

Sitters, H., Christie, F., Di Stefano, J., Swan, M., Collins, P. & York, A. (2014a) Associations between occupancy and habitat structure can predict avian responses to disturbance: Implications for conservation management. Forest Ecology and Management, 331, 227-236.

Sitters, H., Christie, F.J., Di Stefano, J., Swan, M., Penman, T., Collins, P.C. & York, A. (2014b) Avian responses to the diversity and configuration of fire age classes and vegetation types across a rainfall gradient. Forest Ecology and Management, 318, 13-20.

Šmilauer, P. & Lepš, J. (2014) Multivariate analysis of ecological data using CANOCO 5. Cambridge university press.

Spies, T.A., Lindenmayer, D.B., Gill, A.M., Stephens, S.L. & Agee, J.K. (2012) Challenges and a checklist for biodiversity conservation in fire-prone forests: Perspectives from the Pacific Northwest of USA and Southeastern Australia. Biological conservation, 145, 5-14.

Stang, M., Klinkhamer, P.G., Waser, N.M., Stang, I. & van der Meijden, E. (2009) Size- specific interaction patterns and size matching in a plant–pollinator interaction web. Annals of botany, 103, 1459-1469.

Steffan-Dewenter, I., Münzenberg, U., Bürger, C., Thies, C. & Tscharntke, T. (2002) Scale-dependent effects of landscape context on three pollinator guilds. Ecology, 83, 1421-1432.

Stephens, S.L., McIver, J.D., Boerner, R.E., Fettig, C.J., Fontaine, J.B., Hartsough, B.R., Kennedy, P.L. & Schwilk, D.W. (2012) The effects of forest fuel-reduction treatments in the United States. BioScience, 62, 549-560.

Stoutamire, W. (1983) Wasp-pollinated species of Caladenia (Orchidaceae) in south-western Australia. Australian Journal of Botany, 31, 383-394.

108

Swan, M., Christie, F.J., Sitters, H., York, A. & Di Stefano, J. (2015) Predicting faunal fire responses in heterogeneous landscapes: the role of habitat structure. Ecological Applications, 25, 2293-2305.

Swarts, N.D. & Dixon, K.W. (2009) Terrestrial orchid conservation in the age of extinction. Annals of botany, 104, 543-556.

Szendrei, Z. & Isaacs, R. (2006) Ground covers influence the abundance and behavior of Japanese beetles. Environmental entomology, 35, 789-796.

Tamme, R., Hiiesalu, I., Laanisto, L., Szava‐Kovats, R. & Pärtel, M. (2010) Environmental heterogeneity, species diversity and co‐existence at different spatial scales. Journal of Vegetation Science, 21, 796-801.

Taylor, R.S., Watson, S.J., Nimmo, D.G., Kelly, L.T., Bennett, A.F. & Clarke, M.F. (2012) Landscape‐scale effects of fire on bird assemblages: does pyrodiversity beget biodiversity? Diversity and distributions, 18, 519-529. ter Braak, C.J., Prentice, I.C. & Caswell, H. (1988) A theory of gradient analysis. Adv Ecol Res, 18, 271-317.

Thomson, J.D. (1981) Spatial and temporal components of resource assessment by flower-feeding insects. The Journal of Animal Ecology, 50, 49-59.

Tremblay, R.L., Ackerman, J.D., Zimmerman, J.K. & Calvo, R.N. (2005) Variation in sexual reproduction in orchids and its evolutionary consequences: a spasmodic journey to diversification. Biological Journal of the Linnean Society, 84, 1-54.

Trøjelsgaard, K., Jordano, P., Carstensen, D.W. & Olesen, J.M. (2015) Geographical variation in mutualistic networks: similarity, turnover and partner fidelity. Proceedings of the Royal Society of London B: Biological Sciences, 282, 20142925.

Tscharntke, T., Tylianakis, J.M., Rand, T.A., Didham, R.K., Fahrig, L., Batary, P., Bengtsson, J., Clough, Y., Crist, T.O. & Dormann, C.F. (2012) Landscape moderation of biodiversity patterns and processes‐eight hypotheses. Biological Reviews, 87, 661-685.

species interaction networks. Biological conservation, 143, 2270-2279. Tylianakis, J.M., Laliberté, E., Nielsen, A. & Bascompte, J. (2010) Conservation of Tylianakis, J.M., Rand, T.A., Kahmen, A., Klein, A.-M., Buchmann, N., Perner, J. & Tscharntke, T. (2008) Resource heterogeneity moderates the biodiversity- function relationship in real world ecosystems. PLoS Biol, 6, e122.

Van Nuland, M.E., Haag, E.N., Bryant, J.A., Read, Q.D., Klein, R.N., Douglas, M.J., Gorman, C.E., Greenwell, T.D., Busby, M.W. & Collins, J. (2013) Fire Promotes

109

Pollinator Visitation: Implications for Ameliorating Declines of Pollination Services. PloS one, 8, e79853.

Vanbergen, A.J. (2014) Landscape alteration and habitat modification: impacts on plant–pollinator systems. Current Opinion in Insect Science, 5, 44-49.

Vanbergen, A.J., Baude, M., Biesmeijer, J.C., Britton, N.F., Brown, M.J., Brown, M., Bryden, J., Budge, G.E., Bull, J.C. & Carvell, C. (2013) Threats to an ecosystem service: pressures on pollinators. Frontiers in Ecology and the Environment, 11, 251-259.

Vanbergen, A.J., Woodcock, B.A., Gray, A., Grant, F., Telford, A., Lambdon, P., Chapman, D.S., Pywell, R.F., Heard, M.S. & Cavers, S. (2014) Grazing alters insect visitation networks and plant mating systems. Functional Ecology, 28, 178-189.

Vasseur, C., Joannon, A., Aviron, S., Burel, F., Meynard, J.-M. & Baudry, J. (2013) The cropping systems mosaic: How does the hidden heterogeneity of agricultural landscapes drive arthropod populations? Agriculture, Ecosystems & Environment, 166, 3-14.

Veddeler, D., Klein, A.M. & Tscharntke, T. (2006) Contrasting responses of bee communities to coffee flowering at different spatial scales. Oikos, 112, 594- 601.

Vieira, M.C. & Almeida‐Neto, M. (2015) A simple stochastic model for complex coextinctions in mutualistic networks: robustness decreases with connectance. Ecology Letters, 18, 144-152.

Wagner, H.H. & Fortin, M.-J. (2005) Spatial analysis of landscapes: concepts and statistics. Ecology, 86, 1975-1987.

Watson, S., Taylor, R., Nimmo, D., Kelly, L., Clarke, M. & Bennett, A. (2012) The influence of unburnt patches and distance from refuges on post‐fire bird communities. Animal Conservation, 15, 499-507.

Westphal, C., Steffan-Dewenter, I. & Tscharntke, T. (2006) Bumblebees experience landscapes at different spatial scales: possible implications for coexistence. Oecologia, 149, 289-300.

Westrich, P. (1996) Habitat requirements of central European bees and the problems of partial habitats. Linnean Society Symposium Series, pp. 1-16. Academic Press Limited.

Whelan, R., Rodgerson, L., Dickman, C.R. & Sutherland, E.F. (2002) Critical life cycles of plants and animals: developing a process-based understanding of population changes in fire-prone landscapes. Flammable Australia: the fire regimes and biodiversity of a continent (ed. R. williams, Gill, AM, Bradstock, RA), pp. 94-124. CSIRO Publishing.

110

Whelan, R.J. (1995) The ecology of fire. Cambridge University Press.

Wiens, J.A. (1989) Spatial scaling in ecology. Functional Ecology, 3, 385-397.

Wilcock, C. & Neiland, R. (2002) Pollination failure in plants: why it happens and when it matters. Trends in plant science, 7, 270-277.

Williams, N.M., Crone, E.E., T’ai, H.R., Minckley, R.L., Packer, L. & Potts, S.G. (2010) Ecological and life-history traits predict bee species responses to environmental disturbances. Biological conservation, 143, 2280-2291.

Williams, R.J., Bradstock, R.A., Cary, G.J., Enright, N., Gill, A., Leidloff, A., Lucas, C., Whelan, R., Andersen, A. & Bowman, D. (2009) Interactions between climate change, fire regimes and biodiversity in Australia: a preliminary assessment. Report to the Department of Climate Change and the Department of the Environment, Water, Heritage, and the Arts. Canbera.

Willmer, P. (2011) Pollination and floral ecology. Princeton University Press.

Winfree, R. & Kremen, C. (2009) Are ecosystem services stabilized by differences among species? A test using crop pollination. Proceedings of the Royal Society B: Biological Sciences, 276, 229-237.

Woinarski, J. (2005) Living with fire - Birds in Northern Australia. Wingspan (supplement), 7-9.

Wong, B.B., Salzmann, C. & Schiestl, F.P. (2004) Pollinator attractiveness increases with distance from flowering orchids. Proceedings of the Royal Society of London, Series B: Biological Sciences, 271, 212-214.

Wood, S.W., Murphy, B.P. & Bowman, D.M. (2011) Firescape ecology: how topography determines the contrasting distribution of fire and rain forest in the south‐west of the Tasmanian Wilderness World Heritage Area. Journal of biogeography, 38, 1807-1820.

Woodcock, T.S., Larson, B.M., Kevan, P.G., Inouye, D.W. & Lunau, K. (2014) Flies and flowers II: Floral attractants and rewards. Journal of Pollination Ecology, 12, 63-94.

Yeates, D.K., Logan, D.P. & Lambkin, C. (1999) Immature stages of the bee fly Ligyra satyrus (F.)(Diptera: Bombyliidae): a hyperparasitoid of canegrubs (Coleoptera: Scarabaeidae). Australian Journal of Entomology, 38, 300-304.

Zuur, A., Ieno, E.N., Walker, N., Saveliev, A.A. & Smith, G.M. (2009) Mixed effects models and extensions in ecology with R. Springer Science & Business Media.

111

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Brown, Julian MacPherson

Title: Fire effects on pollinators and pollination

Date: 2016

Persistent Link: http://hdl.handle.net/11343/122043

File Description: Fire effects on pollinators and pollination

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.