Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2012 The Effects of Landscape Structure on , Network Architecture, and Ecosystem Function Brian J. Spiesman

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] THE FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

THE EFFECTS OF LANDSCAPE STRUCTURE ON BIODIVERSITY, NETWORK

ARCHITECTURE, AND ECOSYSTEM FUNCTION

By

BRIAN J. SPIESMAN

A Dissertation submitted to the Department of Biological Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Fall Semester, 2012 Brian Spiesman defended this dissertation on October 19, 2012. The members of the supervisory committee were:

Brian D. Inouye Professor Directing Dissertation

Mike Mesterton-Gibbons University Representative

Austin R. Mast Committee Member

Thomas E. Miller Committee Member

Nora Underwood Committee Member

The Graduate School has verified and approved the above-named committee members, and certifies that the dissertation has been approved in accordance with university requirements.

ii

ACKNOWLEDGEMENTS

I am grateful for the guidance, advice, and support of my advisor, Brian Inouye. I thank Nora Underwood and the rest of the Inouye-Underwood lab group. I thank the rest of my committee, Tom Miller, Austin Mast, and Mike Mesterton-Gibbons for their help in developing my research. I thank Sophie Hyson for her help in the field and Andrés Plata Stapper for his essential help in the lab. Funding for this dissertation was provided by the National Science Foundation (DEB- 1110738), the COFRS program at FSU, the Robert K. Godfrey Endowment, the FSU Dept. of Biological Science, and the FSU Graduate School. I am grateful to my family for their support. Lastly, but most importantly, I thank Tania Kim. This dissertation would not have been possible without her.

iii

TABLE OF CONTENTS

List of Tables ...... vi List of Figures ...... vii Abstract ...... ix

1. INTRODUCTION ...... 1

2. HABITAT LOSS ALTERS THE ARCHITECTURE OF -POLLINATOR INTERACTION NETWORKS ...... 5 2.1 Introduction ...... 7 2.2 Methods...... 9 2.2.1 Study area and sampling ...... 9 2.2.2 Deriving explanatory variables ...... 11 2.3 Results ...... 13 2.3.1 SEM analysis ...... 14 2.4 Discussion ...... 15 2.4.1 Conclusions and future directions ...... 18 2.5 Tables ...... 20 2.6 Figures...... 28

3. MATRIX HABITAT QUALITY MEDIATES THE EFFECTS OF PATCH SIZE AND ARRANGEMENT ON METACOMMUNITY STRUCTURE AND ECOSYSTEM FUNCTION ...... 36 3.1 Introduction ...... 36 3.1.1 Matrix effects on local communities ...... 37 3.1.2 Incorporating ecosystem level processes ...... 38 3.2 Methods...... 39 3.2.1 Experimental design...... 39 3.2.2 Leaf litter decomposition ...... 40 3.2.3 A molecular characterization of leaf litter microbial communities ...... 40 3.2.4 Statistical analysis ...... 41 3.3 Results ...... 42 3.4 Discussion ...... 43 3.4.1 Future directions ...... 45 3.5 Figures...... 46

4. THE CONSEQUENCES OF MULTIPLE INDIRECT PATHWAYS OF INTERACTION FOR COEXISTENCE ...... 52 4.1 Introduction ...... 52

iv

4.2 The Model ...... 54 4.2.1 Model analysis ...... 55 4.3 Results and Discussion ...... 57 4.3.1 Community-level effects of mutualism ...... 59 4.3.2 Conclusions ...... 60 4.4 Figures...... 61

5. CONCLUSION ...... 67

APPENDICES ...... 69

A. LANDSCAPE STRUCTURE EFFECTS ON OAK COMMUNITY STRUCTURE ...... 69

B. LANDSCAPE AND COMMUNITY STRUCTURE EFFECTS ON ECOSYSTEM FUNCTION ...... 70

C. MATRIX QUALITY MEDIATES THE EFFECTS OF PATCH SIZE AND ARRANGEMENT ON OAK LITTER COMMUNITY STRUCTURE ...... 71

REFERENCES ...... 73

BIOGRAPHICAL SKETCH ...... 82

v

LIST OF TABLES

2.1 List of plant species ...... 20

2.2 List of pollinator morphospecies ...... 22

2.3 Unstandardized/standardized path coefficients and covariance/correlations ...... 26

2.4 Matrices of standardized total (A) and indirect (B) effects ...... 27

A1 Type III summary of linear mixed effects model testing the effect of oak patch size, arrangement, and matrix quality on species composition in oak patches ...... 69

A2 Type III summary of linear mixed effects model testing the effect of oak patch size, arrangement, and matrix quality on species richness in oak patches ...... 69

B1 Type III summary of linear mixed effects model testing the effect of patch size, arrangement, and matrix quality on the rate of oak leaf litter decomposition ...... 70

B2 Type III summary of linear mixed effects model testing the effect of oak species richness and composition on the rate of oak leaf litter decomposition ...... 70

C1 Type III summary of linear mixed effects model testing the effect of a bare ground matrix on species richness in oak patches ...... 71

C2 Type III summary of linear mixed effects model testing the effect of bare ground matrix on species composition in oak patches ...... 71

C3 Type III summary of linear mixed effects model testing the effect of a pine litter matrix and pine species composition on species composition in oak patches ...... 71

C4 Type III summary of linear mixed effects model testing the effect of a pine litter matrix and pine species richness on species richness in oak patches ...... 72

vi

LIST OF FIGURES

1.1 Conceptual diagram illustrating the approach taken in chapters two and three ...... 2

2.1 Sandhill habitat ...... 28

2.2 Study site locations (blue squares) with 600m radius sectors (blue circles) defining the spatial extent of each landscape ...... 29

2.3 The non-sandhill habitat in the landscape surrounding all but one site is comprised mainly of anthropogenically disturbed former sandhill habitat ...... 30

2.4 Initial (A) and final (B) structural equation models...... 31

2.5 Univariate comparisons between some of the SEM components ...... 32

2.6 Relationship between pollinator species richness and the proportion of sandhill habitat in the landscape (A) and the abundance of in flower (B) ...... 33

2.7 The effect of habitat loss varies among taxonomic groups ...... 34

2.8 Distribution of the proportion of sandhill habitat within a 600 m radius of 500 randomly placed points occurring within sandhill habitat in the study region (i.e., the total map area of Fig. 2.2) ...... 35

3.1 Experimental design to scale ...... 46

3.2 Structural features of the landscape affect local species richness and composition ...... 47

3.3 Effect of landscape structure on oak litter species richness ...... 48

3.4 Mean percent oak litter mass loss after 12 months with standard error bars ...... 49

3.5 Local community effect on oak leaf litter decomposition ...... 50

3.6 Effects of oak patch size and arrangement on oak litter community structure (species richness and composition) in landscapes with a bare ground matrix (A) and a pine litter matrix (B) ...... 51

4.1 Interaction network depicting the sign of the direct effects between species (solid arrows) and the pathways of apparent competition between M1 and M2 mediated by a shared predator (upper dashed arrow) and competing resources (lower dashed arrow) ...... 61

vii

4.2 The equilibrium abundance of M2 (A and B) and the strength of the total indirect effect of M1 on M2 (C and D) and varies with the benefit Mi provides to Ri (γ) and the pathway(s) of the indirect effect (different shades) ...... 62

4.3 Net effect of the predator on M2 for a range of resource competition strength and fixed encounter rate (c1 = 0.095)...... 63

4.4 The net indirect effect of M1 on M2 partitioned between simultaneously acting pathways ...... 64

4.5 Coexistence surface of M1 and M2 across a range of γ in different competitive (A) and predation (B) environments...... 65

4.6 Coexistence surface of R1 and R2 across a range of γ in different competitive environments (A) and capture rates of M1 and M2 (B) ...... 66

viii

ABSTRACT

The amount, composition, and configuration of habitat in the surrounding landscape can affect the structure of local communities, how those communities interact, and the ecosystem functions they perform. In the following chapters I report the results of three studies that use the analysis of experimental and observational data and mathematical modeling to explore these ecological themes in a number of systems. In chapter two I describe the results of a large-scale field study on the effects of habitat loss on the architecture of plant-pollinator interaction networks. Habitat loss has often been shown to have a negative effect on the number and composition of species in plant-pollinator communities. Although we have a general understanding of the negative consequences of habitat loss for species diversity, much less is known about the effects of habitat loss on the pattern of interactions in mutualistic networks. Networks of mutualistic interactions often form patterns of relatively high nestedness and low modularity; these patterns are thought to confer stability on communities. With the growing threat of environmental change, it is important to expand our understanding of the factors that affect biodiversity and the stability of the communities that provide critical ecosystem functions and services. In the first empirical study on the effects of habitat loss on plant-pollinator network architecture, I found that regional habitat loss contributes directly to species loss and indirectly to the re-organization of interspecific interactions in a local community. Networks became less nested and more modular with habitat loss. Species loss was the primary driver of variation in network architecture, though species composition also affected modularity. Previous theory suggests that a reduction in nestedness and an increase in modularity with habitat loss may threaten community stability and that such a loss of stability may contribute to an extinction debt in communities already affected by habitat loss. With my third chapter, I report the results of an experimental microcosm study in which I used leaf litter metacommunities to examine the effects of landscape structure on local communities, and the resulting consequences for ecosystem function. A metacommunity is a set of local communities whose dynamics are linked by the dispersal of multiple interacting species, and it can be affected by the structure of the landscape on which species interact. Theory predicts that biodiversity should be enhanced by increasing the amount and connectivity of a focal habitat in the landscape, however results from empirical studies have been highly variable. Moreover,

ix the quality of the habitat between patches of focal habitat (i.e. the “matrix”) has been shown to affect biodiversity within a focal habitat. Increasing the quality of the matrix may benefit biodiversity by increasing the ease of dispersal among patches or providing supplementary resources. Less well understood is the ability of the matrix to serve as a source of colonizing species that can interact with focal communities. The inconsistent results from prior empirical work may be because few metacommunity studies have addressed the importance matrix habitats for local community structure and function; I know of no other study that has experimentally examined within a single system the effects of habitat amount, arrangement, and matrix quality on community structure and ecosystem function. I developed a microcosm system of oak leaf litter communities in miniature 1 m2 landscapes. I used a fully factorial 2x2x2 design, manipulating oak litter patch size (large or small), arrangement (connected or isolated), and matrix quality (either bare ground or an alternative matrix of pine litter). Rather than identifying species in the litter visually, I used the molecular technique terminal-restriction fragment length polymorphism to characterize communities and quantify species richness and composition. The rate of oak leaf litter decomposition was measured using litter bags. After one year, oak patch isolation and presence of a pine litter matrix both affected species composition and increased species richness. Patch size had marginally significant effects on richness and composition. Landscape effects on richness and composition translated to an effect on the rate of decomposition. The presence of a pine litter matrix, where species richness was greater, slowed the rate of oak leaf litter decomposition. Isolated patches, where species richness was greater, also had a slower rate of decomposition. Results also show that matrix quality can mediate the effects of patch size and arrangement on local communities via interactions with communities originating in the matrix, suggesting that matrix communities can be integral parts of the metacommunity, which can have consequences for ecosystem function. Therefore, integrating variation in matrix quality will be a key part of moving metacommunity concept forward, especially for applying the metacommunity concept to studies of habitat loss and fragmentation. In chapter four, I use mathematical models to examine how an indirect effect is simultaneously propagated through multiple pathways of an interaction network and the consequences for species coexistence. Species interact directly through predation, mutualism, and competition. However, species can also interact indirectly if a direct effect on one species alters an effect on another. A set of indirect interactions links the dynamics of all species in a

x network, regardless of whether they interact directly or not, and these indirect interactions can affect species coexistence. The great complexity of most real ecological interaction networks provides multiple indirect pathways between species. The goal of analyzing this model was to examine how the strength of a negative indirect effect (i.e., apparent competition) is partitioned among two simultaneously acting pathways. I used Lotka-Volterra competition and predation equations to model the population dynamics in a 5-species community: a generalist predator (e.g., a spider) preys on two species (e.g., bees), which provide mutualistic services to their specialist partners (e.g., plants), which are competitors for a shared resource. I focus on the two species in the intermediate trophic level (e.g., the bees) to examine how the net indirect effect between them is partitioned between the predator and their competing mutualistic partners, and the consequences for species coexistence. I use the inverse of the community (Jacobian) matrix to quantify net effects of each species on the other and then use the conjugate variable approach to partition the indirect effect between the two pathways. Model results show that when both pathways are acting at once, the partial indirect strength of interaction depends on the strength of mutualism between the focal species and their respective partners. The presence of both pathways increases the area of parameter space in which both species can coexist, suggesting that the effects of multiple pathways of indirect effects are not additive. The presence of a shared predator, which generally results in the exclusion of the species least able to withstand predation, can instead mediate coexistence even in the absence of a trade-off. This suggests that understanding the mechanisms underlying an indirect effect is important for targeted species management.

xi

CHAPTER ONE

INTRODUCTION

The number and composition of species in ecological communities are dependent on environmental variation at multiple spatial scales. Locally, interspecific interactions, along with the abundance and distribution of resources help to determine how many and which species coexist (i.e., community structure). But the composition and arrangement of habitat in the larger landscape (i.e., landscape structure) can also be important for local communities if they affect how and which species move among local communities. In the following chapters I examine how local interspecific interactions and the structure of the surrounding landscape act individually and in concert to affect the structure of local communities and the functions they perform. Figure 1.1 is a conceptual illustration of the general framework I use in chapters two and three. It is based on the idea that landscape structure provides an arena for multiple local communities to interact within and between communities, which helps determine local community structure, which in turn affects some process or function of the local community. For example, in chapter two I examine how habitat loss in the surrounding landscape affects the richness and composition of local communities of plants and their pollinators, thus affecting how species interact while performing the critical function of . In chapter three I use a similar approach in a microcosm experiment on how the amount, arrangement, and composition of habitat in the landscape affect local communities, which determines the rate of leaf litter decomposition. In chapter four I focus on local community dynamics in a non-spatial modeling context to examine how the pathway of an interspecific interaction, which can change with local community structure, affects species coexistence.

1

Local Local Landscape community process or structure structure function

Figure 1.1. Conceptual diagram illustrating the approach used in chapters two and three. This is based on the idea that landscape structure affects local community structure, which can have consequences for some local community process or function.

Pollination is an essential ecosystem service threatened by environmental change, especially habitat loss. In chapter two, I present the results of a large scale natural experiment examining the effects of habitat loss for communities of plants and their pollinators. In addition to examining the response of species richness and composition, I examine how habitat loss alters the way plants and pollinators interact in a network. Interactions among plants and pollinators form mutualistic networks that often conform to a characteristic and robust pattern (reviewed in Bascompte and Jordano 2007). When compared with networks of antagonistic interactions, such as food webs, mutualistic networks tend to exhibit a relatively nested pattern, so that specialists interact with subsets of the species with which generalists interact (Bascompte et al. 2003). Modularity, or the tendency of species to interact within and not between groups, is relatively low in mutualistic networks (Thébault and Fontaine 2010). Theory predicts that nestedness arises in mutualistic networks because that architecture provides greater community stability than random or modular architectures (Okuyama and Holland 2008, Bastolla et al. 2009, Thébault and Fontaine 2010). Stability can describe many aspects of a community, including the number of species that persist at equilibrium or the speed of return to equilibrium following a disturbance. However, results using most definitions of stability agree and suggest that community stability in mutualistic networks should increase with nestedness as a result of at least three interrelated processes. First, there is a core of highly connected generalists that provide reliable resources and services for more

2 specialized species, which are more prone to extinction. Second, nested networks are asymmetric in that specialists tend to interact with the more reliable generalists and have few interactions with other specialists. Third, greater nestedness allows for facilitation among competitors. For example, the negative effects of interspecific competition between plant species can be mitigated by the positive effects of sharing pollinators (Bastolla et al. 2009). Habitat loss in the landscape may alter the number and composition of species in local communities, which can affect the way the remaining species interact, altering network architecture. Changes to the network may have important consequences for the stable provision of pollination. Although research on mutualistic networks has seen a surge of activity in recent years, very little is known about how networks will respond to disturbances like habitat loss (Morales and Vazquez 2008). Because of the difficulty in combining the inferential strengths of experimentation with large scale studies, such as in chapter two, I used an experimental microcosm approach to study the effects of landscape structure on the structure and function of metacommunities. Metacommunities are sets of local communities whose dynamics are linked by dispersal across the landscape (Holyoak et al. 2005). An area identified as an important direction for future research is including effects of the area between focal communities (the matrix; Holt et. al 2005). Metacommunity studies have often used a habitat/non-habitat approach, where local communities inhabit patches of habitat with distinct boundaries that are set within an inhospitable and homogeneous matrix. However, local communities in fragmented forest or grassland habitat for example, inhabit patches with boundaries that are often indistinct and occur within a heterogeneous landscape mosaic. Thus, effects of matrix habitats must be studied to understand metacommunity dynamics fully (Holt et al. 2005, Logue et al. 2011). Metacommunity studies incorporating ecosystem level processes remain uncommon. Interest in the effects of biodiversity on ecosystem function has resulted in a large body of literature focused mainly on how diversity influences plant productivity. Less is known about the effects of diversity on decomposition and nutrient cycling and how these functions rely on the spatially dependent dynamics of the communities that perform them (Loreau et al. 2003). I constructed a microcosm system of leaf litter organisms that inhabit miniature 1x1 m landscapes in a fully factorial design that crossed patch size, arrangement, and matrix habitat

3 quality. After one year, I found that effects of landscape structure on local communities translate to an effect the ecosystem function of leaf litter decomposition. Moreover, the effects of patch size and arrangement depend on the quality of the matrix habitat and the community inhabiting it, indicating that matrix communities can be integral parts of a metacommunity, with dynamics playing out across continuous landscapes with indistinct community boundaries. In chapter four I use mathematical models to examine in greater detail the set of direct and indirect interactions in a network and the consequences of multiple interaction pathways for species coexistence. Species interact directly with others in a community through activities such as predation, competition, and mutualism, or indirectly when an effect of one species on another is mediated by the response of a third (or more) species. Because of the sparse nature of connections in most ecological networks, most species interact directly with relatively few others, but interact indirectly with all species in a network. As a result, indirect interactions are thought to have as great, or even greater, influence on community dynamics as direct effects (Yodzis 1988, Higashi and Patten 1989). As the size of a network increases, so does the number of pathways for an indirect effect of one species on another. The goal of this chapter is to partition the strength of an indirect effect between two different pathways and to examine how the presence of multiple pathways for an indirect effect between species influences their coexistence. I use a Lotka-Volterra framework to model the population dynamics of a simple 5- species community involving 3 trophic levels: a generalist predator (e.g., a spider) that feeds on two prey species (e.g., two bee species) that each provide a mutualistic service to separate resources (e.g., two plant species). Because I specify that the resources are competitors, there are two pathways for a negative indirect interaction to be propagated between the species at the intermediate trophic level: one via the predator and another via the competing resources. Because each pathway alone will result in a negative indirect effect, it might be expected that an additive effect of both pathways would have an even stronger negative effect. I show that the effects of multiple pathways can be non-additive so that coexistence can be maintained even when the exclusion of an inferior species might be predicted based on the independent effects of each pathway. Moreover, I show that coexistence does not necessarily depend on ecological tradeoffs, but on the effect of the predator in dampening strong interactions. In chapter five I summarize my conclusions and discuss some of the significance of my results for landscape and community ecology.

4

CHAPTER TWO

HABITAT LOSS ALTERS THE ARCHITECTURE OF PLANT- POLLINATOR INTERACTION NETWORKS

2.1 Introduction Local communities are a reflection of the larger landscape in which they are embedded. Dispersal, cross-patch foraging rates, and other processes under the umbrella of metacommunity dynamics play out on a landscape comprised of habitats arranged in space (Kareiva 1987, Andren 1994, Holyoak et al. 2005). Habitat loss in the landscape can therefore be a multi-scale problem affecting the number and composition of species regionally, as well as in a local community (reviewed by Harrison and Bruna 1999, Fahrig 2003). This has been shown to be true of plant-pollinator communities. For example, the local species richness and composition of pollinators can be dependent on the amount of nearby natural habitat (e.g., Steffan-Dewenter and Tscharntke 1999, Bommarco et al. 2010), which can have important consequences for ecosystem function and services (Kremen et al. 2002, Kremen et al. 2004). Although we have a general understanding of the consequences of habitat loss for biodiversity (Fahrig 2003), much less is known about how habitat loss can affect networks of species interactions, such as those formed by plants and pollinators (Fortuna and Bascompte 2006). Communities are often characterized by the number, composition, or abundance of species. However, a growing interest in mutualistic networks has revealed important relationships between community-level patterns of interactions (i.e., network architecture) and stability (Okuyama and Holland 2008, Bastolla et al. 2009, Thébault and Fontaine 2010, Benadi et al. 2012). The architecture of a network may therefore describe community properties that the number and composition of species do not indicate and allow a fuller understanding of the consequences of habitat loss for communities. The architecture of plant-pollinator interaction networks can be quantified by metrics such as nestedness and modularity, which describe patterns of interactions and connectance, which describes the density of interactions. Syntheses of published networks have found that mutualistic networks tend to form highly nested patterns of interaction (e.g., Bascompte et al.

5

2003), such that specialists (species with few interaction partners) interact with a subset of the species with which more generalized species interact. Theory has shown that a more highly nested architecture can increase community stability (Okuyama and Holland 2008, Bastolla et al. 2009) in at least two ways. First, a key feature of nestedness, interaction asymmetry, in which species interact with more generalized partners, results in a reliable resource base for relative specialists (Bascompte et al. 2003). Second, a core of generalists that share partners allows for facilitation, which reduces the negative effects of interspecific competition (Bastolla et al. 2009). Modularity, or compartmentalization, is a network pattern that emerges when species form groups (modules), with interactions occurring more frequently within modules than between modules (Newman and Girvan 2004). A modular architecture has been shown in theory to be destabilizing in mutualistic networks because it reduces many of the beneficial properties conferred by a nested architecture (Bastolla et al. 2009, Thébault and Fontaine 2010). Connectance, or the density of interactions in a network, can vary independently of network pattern (e.g., nestedness or modularity). In mutualistic networks, the correlation between nestedness and modularity can be dependent on connectance (Fortuna et al. 2010) as can be the effects of nestedness on network stability (Okuyama and Holland 2008). In addition to the relatively high nestedness of mutualistic networks, synthetic studies of published networks also show that nestedness is positively correlated with species richness (e.g., Bascompte et al. 2003). It follows that species loss, as a result of habitat destruction, should also reduce nestedness. However, because such studies utilize networks consisting mainly of species drawn from independent regional species pools of varying and unknown sizes, it is difficult to differentiate the effects of a shallow regional species pool from the effects of species loss due to processes like habitat loss. Moreover, these syntheses provide little information about the effects of variation in species composition on network architecture. It is therefore an open question how variation in species richness and composition within a single regional species pool affect network architecture. Some predictions of how habitat loss will affect the architecture of mutualistic networks can be made based on how networks are thought to assemble and disassemble. Network assembly is thought to follow a process called preferential attachment (Barabasi and Albert 1999), in which relative specialists entering a network will preferentially establish interactions with more generalized species. On the other hand, mutualistic networks are thought to

6 disassemble through a process where specialists, being more vulnerable to extinction, are lost from a network before more generalized species. If network disassembly (e.g., as a result of habitat loss) follows a strict process whereby specialists are preferentially lost from a network, only the most generalized species will remain in low-diversity networks (Fortuna and Bascompte 2006). Given these assumptions, species loss should reduce nestedness and modularity because the few remaining species in a low-diversity network should form a well-connected network of generalists. In this study we sampled fifteen sites that span a gradient in habitat loss in the surrounding landscape and use structural equation modeling to examine the direct and indirect consequences of habitat loss for community structure (species richness, composition, and abundance) and network architecture (nestedness, modularity, and connectance). Specifically, we ask: (1) how is habitat loss correlated with community structure, (2) how is community structure correlated with network architecture, and (3) how is habitat loss related to network architecture. We then discuss the potential implications of habitat loss for community stability.

2.2 Methods 2.2.1 Study area and sampling We studied plant-pollinator networks within sandhill habitat in north Florida. Sandhill is a fire-maintained upland pine savannah, characterized by an open canopy of longleaf pine (Pinus palustris) and an understory of turkey oak (Quercus laevis), wire grass (Aristida stricta), and a highly diverse mix of herbaceous plants (Fig. 2.1). Sandhill is a xeric subtype of the imperiled longleaf pine forest ecosystem that was once the dominant forest type of the southeastern coastal plain (Myers and Ewel 1990). Timber harvesting has reduced longleaf pine forests to an estimated 5% of its former distribution (Outcault and Scheffield 1996). Sandhill habitat in the Apalachicola National Forest, the location of our study, has also been reduced and fragmented by timber harvesting, making the region ideal for studying the effects of habitat loss on local plant- pollinator interaction networks. Fifteen 60x60m sites, all within sandhill habitat, were selected to span a gradient in sandhill habitat loss in the surrounding landscape (Fig. 2.2). We used a Landsat-derived land cover classification (Stys et al. 2004) within a GIS to guide the preliminary site selection process. The final selection was made after ground truthing to standardize local habitats within a small range of variation (e.g., time since last fire and the density of understory growth) and 7 ensuring that sites were separated by at least 1 km. Though some other types of natural habitat are present in some of the landscapes, the area of non-sandhill habitat surrounding sites is mainly comprised of anthropogenically modified habitats such as clearcut, shrubby secondary growth, or commercial pine plantations that were all once sandhill habitat (Fig. 2.3). Within the study area these anthropogenically modified habitats harbor few floral resources when compared with sandhill habitat. Thus we interpret the effects of a reduction in the area of sandhill habitat in the surrounding landscape as the effects of habitat loss. Moreover, we assume, in this natural experiment, no correlation between the compositions of the original communities and subsequent habitat loss. Fifteen plant-pollinator interaction networks, one network per site, were quantified using standardized observations of flower-visiting . Observations were conducted at each site monthly, June through September, capturing the vast majority of pollinator activity and plant flowering periods for the growing season. During each of the four observation periods, three or four individuals of each species of plant in flower were observed for approximately 25 minutes each. The resulting 879 hours of observations were recorded on video using five high-definition camcorders (Canon Vixia HF M31). From the HD video we identified flower-visiting species to the lowest taxonomic resolution possible, with the help of a reference collection made the prior year during a pilot study. The use of HD video was very effective in that it allowed for the observation of relatively undisturbed visitation (as compared with direct observation in the field) and the permanent record of each visit could be error-checked so that identifications could be revised long after returning from the field. Our HD video allowed for the reliable identification to morphospecies of more than 99% of visitors. Those few visitors that could not be identified with confidence, as well as those that did not appear to be foraging for pollen or nectar resources were removed from the analysis. We are confident that the vast majority of morphospecies represent individual taxonomic species based on comparison with our reference collection and reviewing/comparing video observations of difficult-to-identify morphospecies. That is, we are confident that lumping of morphologically similar species and splitting of polymorphic species was rare. Although visitors may vary in pollination efficiency, we hereafter refer to these flower visitors as pollinators. We identified and measured the relative abundance of all plant species in flower during each observation period by counting individuals within six 3m x 60m belt transects at each site.

8

The sum of the abundances of each monthly round of sampling was used in analyses as a local estimate of total plant abundance. Total pollinator abundance was estimated as the density of visits per minute of observation. Percent canopy cover was quantified using a spherical densitometer as a proxy for stand age.

2.2.2 Deriving explanatory variables The effect of habitat loss on local plant-pollinator communities was examined using the proportion of sandhill habitat within a 600 m radius of each site center. We employed a commonly used approach to determine the spatial extent of our landscape analysis (Thies et al. 2003). This method involves examining the effect of the total area of sandhill within radii of 300 to 1000 m in 100m increments on local species richness. This range of spatial extents, or landscape sizes, is biologically relevant since it encompasses the upper limit of the foraging range of many pollinators (e.g., Gathmann and Tscharntke 2002, Greenleaf et al. 2007). A 600m radius best explained variation in total species richness, suggesting that plants and pollinators in our study system perceive and interact with features in the landscape on this scale (Thies et al. 2003). Species richness for each network was estimated as the count of all species, plants and pollinators, across all sampling periods. The combined composition of the plant and pollinator species comprising each network was quantified using non-metric multidimensional scaling (NMDS; Bray-Curtis distance) on presence-absence data. Because a single-axis solution was highly correlated with species richness, we specified a two-axis solution, then rotated the first axis to maximize the correlation with species richness, and ultimately used the second axis to represent species composition, independent of species richness. All community and network metrics were calculated in R v2.13.0 (R Development Core Team 2011). The total abundance of all species at each site is represented by our estimate of plant abundance because this estimate is derived independently from network observations but remains highly correlated with our estimate of pollinator abundance (r = 0.7761, P = 0.0007). Each of the 15 interaction networks was analyzed as an incidence matrix, with plants in P rows and flower visitors in V columns. Elements of a P x V adjacency matrix A indicate the presence of a link, or interaction, between a plant and pollinator when aij = 1 and the absence of

9 an interaction when aij = 0. For each matrix, we quantified three indices of network architecture: connectance, nestedness, and modularity. Connectance C is defined as the proportion of all possible interactions in a network that are actually realized and is calculated as . Nestedness was calculated using the nestedtemp function of the ‘vegan’ package (Oksanen et al. 2011) in R (R Development Core Team 2011). This function is based on Rodriguez-Girones and Santamaria’s (2006) algorithm that calculates a network’s maximal nestedness by iteratively reordering the rows and columns of A. The modularity M of a network is defined as the proportion of interactions that occur within modules minus the expected proportion of such interactions (Newman and Girvan 2004). Modularity is calculated as

(1)

where L is the number of links, or interactions, in the network, aij is an element of A, k is the species’ degree, and  is Kronecker’s delta. if species i and j belong to the same module m, otherwise . However, in order to calculate a network’s modularity, one must first classify and determine membership within modules. Like methods for calculating nestedness, methods for module detection use search strategies in order to find the classification that optimizes M. We employ two different methods of module detection, the walktrap.community and fastgreedy.community algorithms, using the R package ‘igraph’ (Csardi and Nepusz 2006) and, for each network, use the classification that maximizes M. Both nestedness and modularity were re-scaled to vary from 0 to 100, with 100 being maximally nested and modular. We used structural equation modeling (SEM) to analyze the direct and indirect pathways linking the amount of focal habitat in the landscape and network architecture. Because network properties are often correlated, SEM is ideal for examining alternative hypotheses for the factors that govern network structure (Grace 2006). We hypothesized that the amount of focal habitat in the landscape directly affects the local richness, composition, and abundance of plant and pollinator species, which in turn directly affect the pattern and density of interactions within the network (Fig. 2.4a). The amount of focal habitat in the landscape therefore has only an indirect effect on nestedness, modularity, and connectance, mediated by a direct effect on species richness, composition, and abundance. Canopy cover was included as a covariate to help explain 10 local (i.e., within-site) environmental variation. Correlations, as opposed to direct causal pathways, were specified between richness and abundance, and between nestedness and modularity. SEM requires linear relationships between variables, and although species richness generally has a non-linear relationship with nestedness (Bascompte et al. 2003) and connectance (Laurienti et al. 2011), the relationships are approximately linear for the range of values in our study. We used an AIC-based stepwise process to remove non-significant paths and compare alternative models of pathway structure. Our sample size is considered small for SEM. So to assess the validity of the model we used a bootstrapping procedure with 10,000 samples as described in Ievers-Landis et al. (2011). Bias in path coefficients due to low sample size (or multivariate non-normality) can be detected if an observed estimate differs substantially from the mean of the bootstrapped samples. The bias, or the difference between the estimates of the original and bootstrap samples, is considered sufficiently small if the standard error (SE) of the bias is less than the standard error of the bootstrap means for each estimate (Ievers-Landis et al. 2011). To further assess the validity of the model as a whole, we used the Bollen-Stein bootstrapped 2 test to determine whether our observed model is significantly different from 10,000 bootstrap samples. SEM analysis, including the bootstrapping procedure, was performed using AMOS v5.0.1 (Arbuckle 2003).

2.3 Results Across all sites, we observed 76 species of plant belonging to 59 genera in 22 families, with a range of 10 to 38 plant species per site. Plants were visited by 151 species of pollinator, identified to morphospecies, belonging to at least 35 families in 4 orders, and ranging from 23 to 94 morphospecies (hereafter referred to as species) per site. Lists of the plant and pollinator species are in Tables 2.1 and 2.2. Analyses of land cover data combined with video and field observations revealed that the number and composition of other non-sandhill habitat types in the surrounding landscape had no significant effect on combined species richness, composition, or abundance (results not shown). Instead, and as intended by our study design, the amount of sandhill habitat in the landscape had the strongest effect on local communities, indicating that habitat loss and not natural habitat heterogeneity is an important driver of local community change.

11

We found that, after excluding one outlier site, combined species richness decreased with habitat loss in the surrounding landscape (Fig. 2.5a). The outlier site had extremely high pollinator species richness for being positioned within a relatively degraded landscape (30.7% sandhill habitat); it had the highest pollinator species richness of all sites (Fig. 2.6a), which appears to be driven by the strong relationship between the abundance of flowering plants and pollinator species richness (Fig. 2.6b). The high abundance of plants in flower at that site was later determined to be a result of native ground cover seeding as part of a previous habitat restoration program. This site was therefore removed from the following analyses. However, the high pollinator richness at this site points to a possible mechanism for the effect of habitat loss on local networks, to which we will return in the discussion. Though habitat loss has a generally negative effect on combined plant and pollinator species richness, a closer look reveals that the effect of habitat loss varies among finer taxonomic groupings. Species richness in the two most speciose plant families in our study, and particularly Fabaceae, decreases with habitat loss (Fig 2.7a). However, when combining the remaining plant families, all represented by few species, there is no relationship between richness and habitat loss. Similarly, species richness within the two most speciose orders of pollinator in our study, and , are more strongly affected by habitat loss than the less speciose orders of Coleoptera and Diptera (Fig 2.7b).

2.3.1 SEM analysis The model selection process resulted in a SEM (Fig. 2.4b) that fits the data very well (2 = 8.65, DF = 16, P = 0.9271) and is substantially improved relative to the initial model depicted in Fig. 2.4a (ΔAIC = 10.76). Moreover, results of the bootstrapping procedure show that the model fit is unbiased, indicating that sample size did not affect the model’s validity. For both the unstandardized and standardized path coefficients, each of the observed estimates are very similar to the mean bootstrapped estimates, as indicated by the fact that the SE of the bias was always much lower that the SE of the mean bootstrapped estimate (Table 2.3). Bollen-Stein bootstrapped 2 test results demonstrated that our model was not significantly different from the bootstrap samples (P = 0.9908), indicating that the model as a whole was highly stable. This method is effective for assessing the stability of models, such as ours, for which samples capture

12 the full range of variation in the variable of interest (i.e., habitat loss; Fig. 2.8; Ievers-Landis et al. 2011). Univariate comparisons between some of the main model components (Fig. 2.5) agree with the path structure of the best fit model (Fig. 2.4b), which indicates that the total area of sandhill habitat in the landscape was positively related to species richness (Fig. 2.5a) and abundance, but had no effect on species composition (Fig. 2.5b). Sandhill canopies displayed a characteristically open structure, ranging from 30.6 to 57.4% cover with a mean of 48.3%  7.6 SD. Although local canopy cover had significant direct effects on species richness and abundance, it had no indirect effect on nestedness and modularity (P = 0.346 and 0.417, respectively). Variation in species richness had the strongest effects on the pattern of interactions in the networks (Figs. 2.4b and 2.5c-d). Species composition was not related to nestedness or connectance but had a weak effect on modularity. Connectance was not directly related to nestedness, but had a negative effect on modularity. Habitat loss, or the proportion of sandhill in the surrounding landscape, had a statistically significant indirect effect on nestedness (P = 0.0116) and modularity (P = 0.0230), mainly mediated by species richness (Figs. 2.4b and 2.5e- f). The estimates of path coefficients/correlations and associated significance levels are presented fully in Table 2.3. Estimates of standardized total and indirect effects along with bootstrapped P- values are presented in Table 2.4.

2.4 Discussion We investigated the effect of habitat loss on species richness and composition, and the resulting effects on the network of interactions between plants and pollinators. The loss of sandhill habitat was positively related to species loss, but was not related to species composition (Figs 2.5a and 2.5b). These two results may be a result of the composition of habitats within landscapes in the study region. The landscapes in this study are mainly composed of either sandhill habitat or former sandhill habitat that has been anthropogenically modified. In a study of ant communities in sandhill habitat, Spiesman & Cumming (2008) found that alternative natural habitats in the landscape could harbor a different set of species, such that the loss of sandhill- dependent species locally was offset by spillover of habitat generalists. However, in the study presented here, the non-sandhill habitat in the landscape is relatively inhospitable, thus no alternative habitat serves as a source to replace lost species.

13

Our results demonstrate that habitat loss in the surrounding landscape can not only reduce species richness but may alter the pattern of interactions between species in a local community. We found that species loss, associated with a reduction of suitable forest habitat in the surrounding landscape, is correlated with reduced nestedness and increased modularity in plant- pollinator interaction networks. Of the community descriptor variables we examined (richness, abundance, and composition), species richness had the strongest influence on network architecture. We expected nestedness to decrease with species richness based on the idea that specialists are more vulnerable to extinction than generalists (Fortuna and Bascompte 2006). Accordingly, low-diversity networks should be comprised mainly of well-connected generalists, and therefore also exhibit low modularity. The positive relationship we found between species richness and network nestedness followed our expectation (Figs. 2.4b and 2.5c). However, counter to our expectation, modularity increased with species loss (Figs. 2.4b and 2.5d). This may be explained by the significant positive relationship we observed between the number of modules and species richness (P < 0.001, R2 = 0.649). More modules in high-diversity networks may allow for greater opportunity for links to form between modules, thereby decreasing modularity. Structural equation modeling is useful for analyzing systems such as this, where co- variation among model variables can mask underlying patterns, because it allows one to consider multiple network metrics simultaneously while partitioning effects among multiple pathways. A look at some of the individual paths reveals relationships not evident in simple univariate analysis. The strong influence of richness, for instance, appears to be responsible for some of the relationships among the components of network architecture. For example, the pathway between connectance and nestedness was removed from our model, indicating that connectance, independent of species richness, has no effect on nestedness. The relationship between connectance and nestedness is explained by the fundamentally negative effect of richness on connectance (Laurienti et al. 2011). Similarly, although nestedness and modularity are significantly correlated when analyzed alone (R2= 0.502, P = 0.0046), this correlation was removed in the model selection process. The spurious correlation between nestedness and modularity is likely due to the strong opposing effects of species richness on nestedness and modularity.

14

Species composition was not related to nestedness, suggesting that nestedness is much more dependent on the number of species than on the identity of species in a network. Composition was not related to connectance either, indicating that the particular combination of species at a site does not affect the density of links in a network. Composition was, however, related to modularity. Modules are thought to form in ecological networks as a result of niche organization and resource contiguity (Guimera et al. 2010). Therefore, variation in the particular combination of species present in a local network may affect the formation of modules based on resource availability. For example, if one of a pollinator’s preferred pollen sources is absent, it may instead choose to forage on a less-preferred species in a different module, thereby reducing modularity but having no effect on connectance. The one outlier site excluded from our analyses sheds light on possible mechanisms by which habitat loss affects local species richness and thereby network architecture. For example, two possible mechanisms are that (1) pollinators may not be able to reach isolated patches (i.e., sites with little focal habitat in the landscape) or (2) pollinators may be able to reach isolated patches but choose not to visit patches with few flower resources. Even though the excluded site was relatively isolated, the high pollinator species richness observed in the presence of artificially high plant abundance suggests that pollinators can reach isolated habitats if flower resources are sufficiently abundant (Fig. 2.6). Moreover, the excluded site had a highly nested architecture suggesting that seeding with native plants as part of habitat restoration programs in fragmented landscapes may not only attract pollinators from surrounding areas, but that plant- pollinator communities will form functional and robust networks. In this study, habitat loss was not only associated with reduces species richness but also reduced nestedness and increased modularity. Theory suggests that lower nestedness and greater modularity will reduce the stability of networks (Okuyama and Holland 2008, Bastolla et al. 2009, Thébault and Fontaine 2010). A loss of stability may contribute to a form of extinction debt (Tilman et al. 1994) by placing communities in degraded landscapes at even greater risk of further species loss. Food web theory, on the other hand, suggests that increasing modularity will enhance the stability of networks by containing the effects of disturbance (i.e., extinction) within modules (Stouffer and Bascompte 2011). It has been suggested that the fundamental differences between mutualistic and antagonistic networks mean that optimal stability for these two network types is achieved via different patterns of interaction (Thébault and Fontaine 2010). However,

15 the context in which stability has been previously examined leaves room for further analysis. For example, stability is often quantified in terms of persistence, or the number of species remaining at equilibrium after some ecological filtering process. One such filtering process is thought to be interspecific interactions in communities over-saturated with species. As interspecific interactions result in species extinctions, an equilibrium is reached whereby mutualistic networks become more nested and less modular with species loss (Thébault and Fontaine 2009), and not less nested and more modular as we observed. These opposing results may mean that the type of ecological filtering examined theoretically is different from the filtering processes associated with habitat loss. It is therefore worth exploring the degree to which a transition from a more nested to a more modular architecture is stabilizing or destabilizing for communities in the face of habitat loss, and more generally, how a community’s response to disturbance affects the (in)stability-generating properties of different metrics of network architecture.

2.4.1 Conclusions and future directions Habitat loss in the greater landscape may contribute not only to species loss, but indirectly to the re-organization of interspecific interactions in a local community. We found that networks become less nested and more modular with habitat loss in communities of plants and their pollinators. Total species loss was the primary driver of variation in nestedness and modularity, though species composition was also important for modularity. These changes in network architecture may have consequences for the stability of mutualistic communities; the reduction in stability associated with low nestedness and high modularity (Bastolla et al. 2009, Thébault and Fontaine 2010) may contribute to an extinction debt. However, is also possible that the increase in modularity with species loss may be a stabilizing factor that increasingly buffers communities from the negative effects of species loss (Stouffer and Bascompte 2011). Habitat loss affected the taxonomic groups in our study in different ways (Fig. 2.7). It is also likely that the unique characteristics of these groups influence their contribution to network architecture. For example, of the pollinators represented in this study, the most speciose order, Hymenoptera, was the most susceptible to habitat loss (Fig. 2.7b). This has important conservation implications, knowing the global importance of Hymenoptera for pollination services (Klein et al. 2007). But given the importance of species richness for network nestedness,

16 this result also suggests that exploring the interaction between a taxonomic group’s response to habitat loss and its contribution to network architecture may be fruitful. To our knowledge, this is the first study to examine empirically the effects of habitat loss on network architecture. General properties of networks have been discovered through syntheses of published networks (reviewed by Bascompte and Jordano 2007), but to complement and expand on these syntheses, additional empirical studies are needed. It is particularly important to examine multiple networks within the same regional species pool to determine how ecologically driven variation in species richness or composition affects network architecture (e.g., Albrecht et al. 2010, Fründ et al. 2010, Weiner et al. 2011). Spatial variation in the populations comprising a metacommunity is important for local biodiversity (e.g., Holt 1993) and though such variation is likely important for local network architecture, it has rarely been examined. A deeper understanding of the effects of spatially dependent processes on network architecture will help clarify mechanisms of network assembly and disassembly as well as improve our understanding of the processes that allow for stable networks capable of providing reliable ecosystem services.

17

2.5 Tables Table 2.1. List of plant species. Family Species Species code Acanthaceae Dyschoriste oblongifolia dysobl Acanthaceae Ruellia ciliosa ruecil Apiaceae Eryngium yuccifolium eryyuc Asclepiadaceae Asclepias tuberosa asctub Asclepiadaceae Asclepias verticillata ascver Asteraceae Berlandiera pumila berpum Asteraceae Boltonia diffusa boldif Asteraceae Carphephorus odoratissimus carodo Asteraceae Carphephorus paniculatus carpan Asteraceae Chrysopsis latisquamea chrlat Asteraceae Elephantopus elatus eleela Asteraceae Eupatorium album eupalb Asteraceae Eupatorium capillifolium eupcap Asteraceae Eupatorium linearifolium euplin Asteraceae Eupatorium rotundifolium euprot Asteraceae Helianthus radula helrad Asteraceae Hieracium gronovii hiegro Asteraceae Liatris chapmanii liacha Asteraceae Liatris elegans liaele Asteraceae Liatris tenufolia liaten Asteraceae Lygodesmia aphylla lygaph Asteraceae integrifolia palint Asteraceae Pityopsis aspera pitasp Asteraceae Pityopsis flexuosa pitfle Asteraceae Pityopsis graminifolia pitgra Asteraceae Pterocaulon pycnostachyum ptepyc Asteraceae Rudbeckia hirta rudhir Asteraceae Sericocarpus tortifolius sertor Asteraceae Solidago odora solodo Asteraceae Symphyotrichum concolor symcon Asteraceae Vernonia angustifolia verang Boraginaceae Onosmodium virginianum onovir Brassicaceae Warea sessilifolia warses Caryophyllaceae Paronychia baldwinii parbal Chrysobalanaceae Licania michauxii liamic Clusiaceae Hypericum hypericoides hyphyp Commelinaceae Commelina erecta comere Convolvulaceae Stylisma patens stypat Euphorbiaceae Cnidoscolus stimulosus cnisti Euphorbiaceae Croton argyranthemus croarg

18

Table 2.1 cont. List of plant species. Family Genus Species Species code Euphorbiaceae Croton glandulosus crogla Euphorbiaceae Croton michauxii cromic Euphorbiaceae Euphorbia floridana eupflo Euphorbiaceae Stillingia sylvatica stisyl Fabaceae Centrosema virginianum cenvir Fabaceae Chamaecrista fasciculata chafas Fabaceae Chamaecrista nictitans chanic Fabaceae rotundifolia crorot Fabaceae Dalea carnea dalcar Fabaceae Dalea pinnata dalpin Fabaceae Desmodium lineatum deslin Fabaceae Desmodium sp1 dessp1 Fabaceae Galactia erecta galere Fabaceae Galactia mollis galmol Fabaceae Galactia floridana galflo Fabaceae Indigofera caroliniana indcar Fabaceae Orbexilum lupinellum orblup Fabaceae Pediomelum canescens pedcan Fabaceae Rhynchosia reniformis rhyren Fabaceae Schrankia microphylla schmic Fabaceae Stylosanthes biflora stybif Fabaceae Tephrosia chrysophylla tepchr Lamiaceae Trichostema dichotomum tridic Orobanchaceae Seymeria pectinata seypec Polemoniaceae Phlox pilosa phopil Polygalaceae Polygala grandiflora polgra Polygalaceae Polygala incarnata polinc Polygonaceae Eriogonum tomentosum eritom Polygonaceae Polygonella gracilis polgrs Rubiaceae Diodia teres didter Scrophulariaceae Agalinis divaricata agadiv Scrophulariaceae Aureolaria pedicularia aurped Scrophulariaceae Penstemon multiflorus penmul Turneraceae Piriqueta caroliniana pircar Verbenaceae Callicarpa americana calame Verbenaceae Stylodon carneus stycar

19

Table 2.2. List of pollinator morphospecies. Order Superfamily Family Genus Species Morphopecies code Coleoptera Buprestoidea C10 Coleoptera Trichodes C3 Coleoptera C8 Coleoptera Curculionoidea C1 Coleoptera Curculionoidea Curculionidae C9 Coleoptera Curculionoidea Curculionidae C11 Coleoptera Curculionoidea Curculionidae C12 Coleoptera Cantharidae C13 Coleoptera Elateroidea Cantharidae C6 Coleoptera Trigonopeltastes delta C2 Coleoptera Meloidae C4 Coleoptera Tenebrionoidea Meloidae C14 Coleoptera Tenebrionoidea Meloidae C16 Coleoptera Tenebrionoidea C7 Coleoptera Tenebrionoidea Pyrochroidae C15 Coleoptera Tenebrionoidea Tenebrionidae C5 Diptera Asilomorpha Bombyliidae Anthrax D8 Diptera Asilomorpha Bombyliidae Bombylius D19 Diptera Asilomorpha Bombyliidae Exoprosopa D6 Diptera Asilomorpha Bombyliidae Exoprosopa D17 Diptera Asilomorpha Bombyliidae Exoprosopa D31 Diptera Asilomorpha Bombyliidae Geron D3 Diptera Asilomorpha Bombyliidae Paravilla D18 Diptera Asilomorpha Bombyliidae Poecilognathus D7 Diptera Asilomorpha Bombyliidae Systropus D23 Diptera Asilomorpha Bombyliidae Toxophora D24 Diptera Asilomorpha Bombyliidae Villa D5 Diptera Asilomorpha Bombyliidae Villa D22 Diptera Asilomorpha Bombyliidae Villa D32 Diptera Asilomorpha Bombyliidae D9 Diptera Bibionoidea Bibionidae Plecia nearctica D29 Diptera Muscoidea Muscidae D4 Diptera Muscoidea Muscidae D26 Diptera Muscoidea Muscidae D12 Diptera Syrphoidea Syrphidae Eristalis D25 Diptera Syrphoidea Syrphidae Eristalis D28 Diptera Syrphoidea Syrphidae Ocyptamus D10 Diptera Syrphoidea Syrphidae Pseudodoros clavatus D30 Diptera Syrphoidea Syrphidae Toxomerus D11 Diptera Syrphoidea Syrphidae Toxomerus D16 Diptera Syrphoidea Syrphidae D15

20

Table 2.2 cont. List of pollinator morphospecies. Order Superfamily Family Genus Species Morphopecies code Diptera Syrphoidea Syrphidae D27 Diptera Syrphoidea Syrphidae D14 Diptera Tephritoidea Tephritidae D20 Diptera Tephritoidea Ulidiidae D13 Diptera D21 Hymenoptera Apidae Apis mellifera H72 Hymenoptera Apoidea Apidae Bombus impatiens H44 Hymenoptera Apoidea Apidae Bombus fraternus H25 Hymenoptera Apoidea Apidae Bombus pennsylvanicus H87 Hymenoptera Apoidea Apidae Colletidae H75 Hymenoptera Apoidea Apidae Epeolus H80 Hymenoptera Apoidea Apidae Melissodes H50 Hymenoptera Apoidea Apidae Melissodes H60 Hymenoptera Apoidea Apidae Xylocopa micans H85 Hymenoptera Apoidea Apidae Xylocopa virginica H56 Hymenoptera Apoidea Apidae H5 Hymenoptera Apoidea Colletidae Hylaeus H71 Hymenoptera Apoidea Bicyrtes quadrifasciata H54 Hymenoptera Apoidea Crabronidae Bicyrtes H6 Hymenoptera Apoidea Crabronidae Bicyrtes H32 Hymenoptera Apoidea Crabronidae Philanthus H29 Hymenoptera Apoidea Crabronidae Tachysphex H38 Hymenoptera Apoidea Crabronidae Tachytes H9 Hymenoptera Apoidea Crabronidae Tachytes H64 Hymenoptera Apoidea Crabronidae H15 Hymenoptera Apoidea Crabronidae H37 Hymenoptera Apoidea Crabronidae H76 Hymenoptera Apoidea Crabronidae H70 Hymenoptera Apoidea Crabronidae H86 Hymenoptera Apoidea Halictidae Agapostemon splendens H41 Hymenoptera Apoidea Halictidae Augochloropsis H26 Hymenoptera Apoidea Halictidae Lasioglossum nymphale H2 Hymenoptera Apoidea Halictidae Lasioglossum H13 Hymenoptera Apoidea Halictidae Lasioglossum H74 Hymenoptera Apoidea Halictidae Nomia H20 Hymenoptera Apoidea Halictidae Sphecodes H65 Hymenoptera Apoidea Megachilidae Anthidiellum notatum H19 Hymenoptera Apoidea Megachilidae Anthidiellum perplexum H12 Hymenoptera Apoidea Megachilidae Anthidium maculifrons H81 Hymenoptera Apoidea Megachilidae Coelioxys H28 Hymenoptera Apoidea Megachilidae Megachile xylocopoides H77

21

Table 2.2 cont. List of pollinator morphospecies. Order Superfamily Family Genus Species Morphopecies code Hymenoptera Apoidea Megachilidae Megachile H14 Hymenoptera Apoidea Megachilidae Megachile H33 Hymenoptera Apoidea Megachilidae H11 Hymenoptera Apoidea Megachilidae H82 Hymenoptera Apoidea Ammophilla H21 Hymenoptera Apoidea Sphecidae Isodontia apicalis H48 Hymenoptera Apoidea Sphecidae Isodontia exornata H51 Hymenoptera Apoidea Sphecidae Prionyx parkeri H4 Hymenoptera Apoidea Sphecidae Sphex ichneumoneus H34 Hymenoptera Apoidea Sphecidae Sphex pensylvanicus H57 Hymenoptera Chalcidoidea Eurytomidae H43 Hymenoptera Pompilidae H68 Hymenoptera Vespoidea Pompilidae H62 Hymenoptera Vespoidea Pompilidae H17 Hymenoptera Vespoidea Pompilidae H83 Hymenoptera Vespoidea Campsomeris quadrimaculata H66 Hymenoptera Vespoidea Scoliidae Campsomeris plumipes H1 Hymenoptera Vespoidea Scoliidae Scolia nobilitata H10 Hymenoptera Vespoidea Tiphiidae Myzinum H55 Hymenoptera Vespoidea Tiphiidae Myzinum H39 Hymenoptera Vespoidea Tiphiidae Myzinum H73 Hymenoptera Vespoidea Tiphiidae Myzinum H84 Hymenoptera Vespoidea Tiphiidae H49 Hymenoptera Vespoidea Ancistrocerus H27 Hymenoptera Vespoidea Vespidae Eumenes H61 Hymenoptera Vespoidea Vespidae Leptochilus H67 Hymenoptera Vespoidea Vespidae Monobia quadridens H3 Hymenoptera Vespoidea Vespidae Pachodynerus erynnis H47 Hymenoptera Vespoidea Vespidae H22 Hymenoptera Vespoidea Vespidae Polistes H36 Hymenoptera Vespoidea Vespidae Polistes H8 Hymenoptera Vespoidea Vespidae Polistes H30 Hymenoptera Vespoidea Vespidae Polistes H58 Hymenoptera Vespoidea Vespidae Zethus spinipes H23 Hymenoptera Vespoidea Vespidae H16 Lepidoptera Hesperioidea Hesperiidae Erynnis L5 Lepidoptera Hesperioidea Hesperiidae Erynnis L10 Lepidoptera Hesperioidea Hesperiidae Erynnis L13 Lepidoptera Hesperioidea Hesperiidae Erynnis L32 Lepidoptera Hesperioidea Hesperiidae Polites vibex L25 Lepidoptera Hesperioidea Hesperiidae Pyrgus communis L7

22

Table 2.2 cont. List of pollinator morphospecies. Order Superfamily Family Genus Species Morphopecies code Lepidoptera Hesperioidea Hesperiidae Thorybes L11 Lepidoptera Hesperioidea Hesperiidae Urbanus proteus L26 Lepidoptera Hesperioidea Hesperiidae L8 Lepidoptera Hesperioidea Hesperiidae L27 Lepidoptera Hesperioidea Hesperiidae L17 Lepidoptera Hesperioidea Hesperiidae L22 Lepidoptera Hesperioidea Hesperiidae L23 Lepidoptera Hesperioidea Hesperiidae L28 Lepidoptera Hesperioidea Hesperiidae L31 Lepidoptera Hesperioidea Hesperiidae L33 Lepidoptera Hesperioidea Hesperiidae L34 Lepidoptera Hesperioidea Hesperiidae L39 Lepidoptera Utetheisa bella L20 Lepidoptera Papilionoidea Lycaenidae Calycopis cecrops L38 Lepidoptera Papilionoidea Lycaenidae Strymon melinus L1 Lepidoptera Papilionoidea Nymphalidae Agraulis vanillae L29 Lepidoptera Papilionoidea Nymphalidae Battus philenor L14 Lepidoptera Papilionoidea Nymphalidae Eurytides marcellus L18 Lepidoptera Papilionoidea Nymphalidae Junonia coenia L4 Lepidoptera Papilionoidea Nymphalidae Phyciodes phaon L3 Lepidoptera Papilionoidea Nymphalidae Phyciodes tharos L30 Lepidoptera Papilionoidea Pieridae Colias philodice L19 Lepidoptera Papilionoidea Pieridae Eurema daira L12 Lepidoptera Papilionoidea Pieridae Eurema lisa L21 Lepidoptera Papilionoidea Pieridae Phoebis sennae L40 Lepidoptera Papilionoidea Riodinidae Calephelis virginiensis L2 Lepidoptera Urodoidea Urodidae L15 Lepidoptera Yponomeutidae Yponomeutidae Atteva aurea L37

23

Table 2.3. Unstandardized/standardized path coefficients and covariance/correlations. Bias is the difference between the original estimate and the mean of bootstrap estimates. P is the bias-corrected P-value indicating the significance of the original estimate as derived from the bootstrapping process. Path Unstandardized Path Coefficients Standardized Path Coefficients Mean of Mean of Original Bootstrap SE mean SE Original Bootstrap SE mean SE Estimate Estimates Bias bootstrap Bias P Estimate Estimates Bias bootstrap Bias P Habitat amt. to 52.357 52.480 0.123 13.865 0.139 0.0011 0.701 0.686 -0.015 0.149 0.002 0.0015 Richness Canopy to -67.688 -67.828 -0.140 29.033 0.290 0.0248 -0.426 0.426 0.001 0.179 0.002 0.0291 Richness Richness to -0.001 -0.001 0.000 0.000 0.000 0.0011 -0.768 0.760 0.008 0.125 0.001 0.0029 Connectance Canopy to -1312.306 -1310.599 1.706 588.716 5.887 0.0309 -0.504 0.492 0.012 0.199 0.002 0.0399 Abundance Habitat amt. to 621.038 623.212 2.174 279.152 2.792 0.0337 0.507 0.497 -0.001 0.201 0.002 0.0453 Abundance Richness to 0.485 0.485 -0.001 0.071 0.001 0.0002 1.517 1.529 0.012 0.247 0.003 0.0001 Nestedness Abundance to -0.017 -0.017 0.000 0.004 0.000 0.0007 -0.871 0.893 -0.021 0.293 0.003 0.0007 Nestedness Connectance to -225.119 -224.641 0.478 54.156 0.542 0.0010 -0.831 0.839 -0.008 0.260 0.003 0.0008 Modularity Abundance to 0.014 0.014 0.000 0.004 0.000 0.0042 0.750 0.760 0.011 0.256 0.003 0.0033 Modularity Richness to -0.555 -0.556 -0.001 0.088 0.001 0.0002 -1.852 1.863 -0.010 0.372 0.004 0.0002 Modularity Composition to -7.364 -7.355 0.009 3.058 0.031 0.0249 -0.302 0.305 -0.003 0.136 0.001 0.0281 Modularity

Covariance Correlation Richness and 1729.902 1356.964 -372.938 731.329 7.313 0.0002 0.668 0.649 -0.019 0.186 0.002 0.0024 abundance

24

Table 2.4. Matrices of standardized total (A) and indirect (B) effects. A. Standardized Total Effects Habitat amt. Canopy Richness Composition Connectance Abundance Richness 0.7006** -0.4264* Connectance -0.5379** 0.3273* -0.7677** Abundance 0.5069* -0.5041* Modularity -0.4710* 0.1399 -1.2146*** -0.3017* -0.8309*** 0.7496** Nestedness 0.6209* -0.2074 1.5165*** -0.8712***

B. Standardized Indirect Effects Habitat amt. Canopy Richness Composition Connectance Abundance Richness Connectance -0.5379** 0.3273* Abundance Modularity -0.4710* 0.1399 -1.2146** Nestedness 0.6209* -0.2074 Bootstrapped P-values: * < 0.05, ** < 0.01, *** < 0.001

25

2.6 Figures

Figure 2.1. Sandhill habitat.

26

Sandhill Cypress swamp 0 1 2 3 Shrub & clearcut Mixed wetland forest Kilometers

Pine flatwoods Water Stys et al. (2004) 2003 Landsat ETM+ Hardwood hammock Urban 30m resolution Figure 2.2. Study site locations (blue squares) with 600m radius sectors (blue circles) defining the spatial extent of each landscape.

27

1.0

0.8

0.6

0.4 disturbed habitat disturbed

0.2 Proportion anthropogenically anthropogenically Proportion

0.0 0 3 6 9 12 15 Site rank

Figure 2.3. The non-sandhill habitat in the landscape surrounding all but one site is comprised mainly of anthropogenically disturbed former sandhill habitat. The relatively large proportion of natural habitat at one of the sites (0.516; visible as the easternmost site in Fig. 2.2) is comprised mainly of cypress swamp habitat, which, like much of the anthropogenically disturbed habitats, harbors very few floral resources.

28

A. Richness Nestedness Habitat amount Abundance Connectance

Canopy Composition Modularity

B.

0.67 0.85 Richness Nestedness Habitat amount 0.51 0.59 Abundance Connectance Canopy 0.89 Composition Modularity

Figure 2.4. Initial (A) and final (B) structural equation models. In the hypothesized model (A), “Habitat amount” is the proportion of sandhill habitat in the surrounding landscape. “Canopy” is the percent canopy cover within sites. “Richness” and “Composition” are the combined species richness and composition, respectively, of plants and pollinators. “Abundance” is the total relative abundance of plants, which is highly correlated with pollinator abundance. “Connectance”, “Nestedness”, and “Modularity” are as described in the methods. Straight single- headed arrows indicate direct causal pathways, whereas curved double-headed arrows indicate unresolved correlations. In the final model (B) arrow widths are scaled to standardized path coefficients/correlations (see table S1 for values). Black arrows are positive effects and gray are negative. Solid arrows indicate a direct effect (single headed) or correlation (double headed). Dashed arrows represent indirect effects of the proportion of sandhill habitat in the landscape on nestedness and modularity. R2 values are given with variables where appropriate. Note that since composition is derived from a NMDS analysis, the sign of its effect on modularity is arbitrary.

29

120 A 1 B R² = 0.551 100 P < 0.001 P = 0.888 0.5

80 0 60

-0.5 40 Total species richness species Total Composition (NMDS2) Composition 20 -1 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 Proportion sandhill Proportion sandhill

80 100 C D R² = 0.373 70 P = 0.019 90

60 Modularity

80 Nestedness 50 R² = 0.643 P < 0.001 70 40 20 40 60 80 100 120 20 40 60 80 100 120 Total species richness Total species richness

100 80 E F R² = 0.320 R² = 0.438 P = 0.023 P < 0.001 70 90

60

80 Modularity Nestedness 50

70 40 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0 Proportion sandhill Proportion sandhill

Figure 2.5. Univariate comparisons between some of the SEM components. The direct effects of sandhill habitat loss on species richness and composition and of species richness on nestedness and modularity are indicated in panels A-D. The indirect effect of habitat loss on nestedness and modularity are depicted in panels E and F, respectively. P-values indicate the significance of linear regressions.

30

100 A.

80

60

40

20 R² = 0.644 Pollinator species richnessPollinator

0 0.0 0.5 1.0 Proportion sandhill

100 B.

80

60

40

20 R² = 0.886 Pollinator species richnessPollinator

0 0 500 1000 1500 2000 Plant abundance

Figure 2.6. Relationship between pollinator species richness and the proportion of sandhill habitat in the landscape (A) and the abundance of plants in flower (B). The fit in A is after removing the outlier (open circle) from the analysis. (B) Pollinator species richness is highly correlated with plant abundance. Therefore, the artificially high plant abundance at this site may explain the high pollinator species richness.

31

16 A

12

8

4 Asteraceae

Plant species richness species Plant Fabaceae Others 0 0.0 0.2 0.4 0.6 0.8 1.0 Proportion sandhill

40 Hymenoptera B Lepidoptera Diptera 30 Coleoptera

20

10 Pollinator speciesrichness Pollinator 0 0.0 0.2 0.4 0.6 0.8 1.0 Proportion sandhill

Figure 2.7. The effect of habitat loss varies among taxonomic groups. For plants (A), habitat loss had the strongest effect on the species richness of the most speciose families, Asteraceae (P = 0.093, R2 = 0.217) and Fabaceae (P = 0.002, R2 = 0.549), but had no effect on the combined richness of the remaining families (P = 0.757). For pollinators (B), habitat loss had the strongest effects on the richness of the more speciose Hymenoptera (P = 0.005, R2 = 0.500) and Lepidoptera (P = 0.002, R2 = 0.580), with weaker effects on species richness within the orders Diptera (P = 0.040, R2 = 0.307) and Coleoptera (P = 0.063, R2 = 0.260).

32

1

0.8

0.6

0.4 Proportion sandhill Proportion 0.2

0 0 100 200 300 400 500

Rank

Figure 2.8. Distribution of the proportion of sandhill habitat within a 600 m radius of 500 randomly placed points occurring within sandhill habitat in the study region (i.e., the total map area of Fig. 2.2). The distribution of the sites actually sampled (black diamonds) indicates that we sampled across the full range of landscape variability (i.e., habitat loss) in the system.

33

CHAPTER THREE

MATRIX HABITAT QUALITY MEDIATES THE EFFECTS OF PATCH SIZE AND ARRANGEMENT ON METACOMMUNITY STRUCTURE AND ECOSYSTEM FUNCTION

3.1 Introduction Advances in community ecology have come with the recognition that local communities are not closed systems but are open to the reciprocal flow of organisms and resources among locations (reviewed in Holt 1993, Polis et al. 1997). As a consequence, not only local habitat conditions, but conditions in the environment at larger spatial scales can help define the structure and function of local communities. The metacommunity concept is a synthesis of ecological paradigms, which has helped provide a mechanistic framework for understanding the consequences of community openness (Leibold et al. 2004, Holyoak et al. 2005a). A metacommunity is a set of local communities whose dynamics are linked by the movement of interacting species across a patchy landscape. The effect of the amount and arrangement of habitat in the landscape on biodiversity remains a major topic in ecology and conservation. In fact, one of the primary motivations for developing the metacommunity synthesis was to provide a framework for studying landscape- scale processes like the effects of habitat loss and fragmentation (Holyoak et al. 2005b). Although we have learned much from empirical metacommunity studies, there has been a strong focus on a habitat/non-habitat approach (reviewed in Logue et al. 2011), where local communities inhabit focal patches with distinct boundaries that are set within a matrix of homogeneous and inhospitable habitat. This approach reflects the origins of the metacommunity concept in the equilibrium theory of island biogeography (MacArthur and Wilson 1967) and metapopulation theory (reviewed in Hanski and Gilpin 1991). A habitat/non-habitat approach is well suited for many systems such as pitcher plant metacommunities (e.g., Miller and Kneitel 2005) or other systems where aquatic habitat patches are surrounded by an inhospitable terrestrial matrix (e.g., Bock and Ricklefs 1983, Brown 1984, Cottenie and De Meester 2005). But many systems that may be operating as metacommunities, such as fragmented forest or

34 grassland communities, can have indistinct habitat boundaries and can be situated within a heterogeneous matrix that is not be entirely inhospitable. I am not aware of any study that has explicitly investigated how the quality of matrix habitat mediates the effects of habitat amount and arrangement on metacommunity structure and function.

3.1.1 Matrix effects on local communities It is now commonly accepted that the quality of the habitat intervening patches of a focal habitat (i.e., the matrix) can affect the structure of, and connections among, local communities (Ricketts 2001, Vandermeer and Carvajal 2001, Jules and Shahani 2003). There are at least three general means by which matrix quality can influence metacommunity dynamics. One is by altering the relative ease of dispersal among local communities, for example by introducing a barrier with harsh abiotic conditions (Gonzalez et al. 1998) or due to behavioral responses to matrix conditions (e.g., a perceived risk of predation; Lima and Dill 1990). Second, matrix quality can also influence metacommunities through resource supplementation. Temporary excursions into the matrix to forage for alternative resources (Brotons et al. 2003), or spill over of resources from the matrix into local habitats (spatial subsidies) can have a range of effects on local communities (reviewed in Polis et al. 1997). Spillover of abiotic material may alter local pH or nutrient composition, thereby affecting local habitat suitability or process rates (Rousk et al. 2010). Third, the matrix can affect metacommunities by acting as a source of invading species. Predators or habitat generalists may occupy both focal and matrix habitats, changing competitive dynamics and trophic interactions (Cook et al. 2002, Rand et al. 2006). Though understudied empirically, theory suggests that matrix effects can be important for local community structure and function. In some spatially implicit models (e.g., Amarasekare et al. 2004) the spatial arrangement of communities can be interpreted either as local communities inhabiting patches of habitat separated by an inhospitable matrix or communities inhabiting patches that share borders. With spatial heterogeneity in habitat quality, low quality patches (local communities), as perceived by some species in the metacommunity, can instead be interpreted by the observer as matrix habitat. Because low quality (matrix) habitat for one species may be high quality for another, a stage is set for a complex set of dynamics beyond metacommunities of islands in an inhospitable matrix.

35

3.1.2 Incorporating ecosystem processes Analyzing the impact of metacommunity dynamics on the functioning of ecosystems is an important avenue of research that few studies have addressed (Loreau et al. 2003). Studies of the effects of biodiversity on ecosystem function have focused primarily on how diversity influences plant productivity (reviewed in Loreau et al. 2002). Locally, diversity has generally been found to have a positive influence on biomass production, however comparatively little is known about the effects of diversity on decomposition and nutrient cycling (reviewed in Hättenschwiler et al. 2005). Leaf litter decomposition is a critical ecosystem process, responsible for recycling carbon, nitrogen, phosphorous, and other nutrients. It is a highly integrative process that links biotic and abiotic components of the ecosystem across trophic (Coleman et al. 2004). The composition and diversity of decomposer species can both influence decomposition rates (Coûteaux et al. 1995, Strickland et al. 2009), however no general pattern has emerged for the effects of diversity on decomposition, but an answer may lie in the relationship between taxonomic and functional diversity (reviewed in Hättenschwiler et al. 2005). The great diversity of decomposers may result in a high degree of functional redundancy so that diversity, per se, is not as important as functional richness and functional dissimilarity among decomposer groups (Loreau et al. 2001, Heemsbergen et al. 2004). Understanding how metacommunity dynamics affect the species and functional composition of local communities may strengthen our understanding of the relationship between biodiversity and ecosystem function. I use an experimental microcosm of soil and leaf litter communities to investigate three questions: (1) Do structural features of the landscape (patch size, arrangement, and matrix quality) affect local species richness and composition, (2) what are the resultant consequences for an important ecosystem function, the rate of leaf litter decomposition, and (3) how does the quality of the matrix mediate the effects of patch size and arrangement on local species richness and composition? With the final question, I am interested in knowing whether the matrix merely affects rates of dispersal among oak patches or if biotic exchanges between matrix and focal habitats are important. If exchanges are important, the composition of species in oak patches should be correlated with the composition of species in the surrounding pine litter matrix.

36

3.2 Methods 3.2.1 Experimental design I designed a microcosm system of miniature landscapes comprised of different soil and leaf litter habitats and their associated natural communities as a model system to examine the effects of patch size, arrangement, and matrix quality on metacommunity structure and the resultant effects on an ecosystem function (Fig. 3.1). The focal oak leaf litter and soil habitat, along with its associated microbial community, was collected from homogeneous stands of Quercus geminata (sand live oak) in the Apalachicola National Forest. A matrix of bare ground was compared with a matrix of pine litter and soil, which was collected, along with its associated natural community, from a homogeneous stand of Pinus clausa (sand pine) on a managed pine plantation near Tallahassee, FL. Patch size, arrangement, and matrix quality was manipulated in a 2x2x2 fully factorial design. Each 1 m2 landscape was separated by a distance of 1.5 m and contained four patches of oak litter that were either large (30 cm diameter) or small (15 cm diameter) and arranged in either a connected (patches barely touch) or isolated fashion with isolated patches separated by a distance of 10 cm (Fig. 3.1). The four focal habitat designs were duplicated for a bare ground matrix and a pine litter matrix treatment, where pine litter and topsoil filled the area between oak patches. A landscape comprised completely of oak litter and soil was used as a control. The experiment was initiated at the FSU Mission Road Research Facility in Tallahassee, FL in July 2008. Experimental landscapes were established outside on an open lawn area by first laying down black landscaping fabric. Landscapes were then constructed by laying down a 3 cm bed of top soil, taken from the collection locations, on the fabric, then covering the soil with another 3 cm of leaf litter. Litter and soil of each type were separately mixed in large containers to homogenize the respective biotic and abiotic components before laying out the experimental landscapes. In the absence of a pine litter matrix, the bare landscaping fabric serves as one matrix type. Leaf litter patches were held in place against erosion using 1.5 cm mesh plastic bird netting. The netting provided for very good maintenance of the physical landscape structure over the course of the experiment. Shade cloth was suspended approximately 1.5 m above the landscapes to mimic the effects of natural canopy cover on microclimate conditions. Treatments were watered during periods of drought.

37

Landscapes were destructively sampled after one year. I collected the entire patch of oak litter and soil. For pine litter habitat, as well as in the control landscapes, I collected litter and soil from four pseudo-patches (Fig. 3.1). Samples were then preserved for molecular analysis as described below. Temperature was measured approximately 2 cm below the surface of each patch to determine how treatments influence microclimate conditions. Soil pH was measured from preserved samples using a FieldScout SoilStick pH meter (Spectrum Technologies, Inc.) according to the manufacturer’s instructions.

3.2.2 Leaf litter decomposition Litterbag analysis was used to quantify differences in decomposition rates of oak litter among landscape treatments. Litterbags were made of two 10x10 cm pieces of 2 mm polyester mesh. Fresh Quercus geminata leaves were oven dried at 65 C for 48 hours and approximately 2 g of whole leaves were placed in each litter bag. Litterbags were then placed on the surface of two oak patches in each landscape (Fig. 3.1). A similar arrangement was used for pseudo-patches in control landscapes. Litterbags were collected with leaf litter patches, then dried and the contents reweighed to determine the percent dry weight mass loss.

3.2.3 A molecular characterization of leaf litter microbial communities Molecular fingerprinting techniques are commonly used and effective tools for characterizing soil microbial communities (reviewed in Handelsman 2004, Leckie 2005). I used Terminal Restriction Fragment Length Polymorphism (T-RFLP) to characterize and quantify differences among leaf litter microbial communities (Gaston et al. 1997). The T-RFLP method relies on differences among species in the position of their restriction enzyme binding sites on a gene to generate a fingerprint representing the unique combination of species in a sample. I performed T-RFLP analysis in the following steps. First, I homogenized preserved soil samples using a mortar and pestle, and then isolated genomic DNA from 0.25 g of the homogenized samples using the PowerSoil DNA Isolation Kit following the manufacturer’s instructions (Mo Bio Laboratories Inc.). I then amplified the 16S rRNA gene using the universal forward primer 27F (5’-AGA GTT TGAT CCT GGC TCA G -3’) and the universal reverse primer 1492R (5’- GGT TAC CTT GTT ACG ACT T-3’). The 5’ end of the forward primer was labeled with the dye 6-FAM (Integrated DNA Technologies). The PCR reaction included 1l template DNA

38

(~30ng/l), 45l Platinum PCR Supermix High Fidelity (Life Technologies), and 0.2M of each of the forward and reverse primers. PCR amplification proceeded with initial denaturation at 94 C for 2 min, followed by 35 cycles of 94 C for 20 seconds, 53 C for 15 seconds, and 68 C for 105 seconds. Amplified DNA was then purified using a magnetic bead process (SprintPrep, Agencourt Bioscience) following the manufacturer’s guidelines. Each sample was then prepared for T-RFLP analysis by combining 1 l of the purified PCR product with 11.72 l of formamide and 0.28 l of size standard (GeneScan 500 LIZ, Applied Biosystems) on 96-well plates. Plates were submitted to the FSU Department of Biology Bioanalytical and Sequencing Facility for analysis on the Applied Biosystems 3730 Genetic Analyzer. This is an automated system that works by digesting samples with restriction enzymes, then separating fragments by capillary electrophoresis, followed by laser detection of the terminal dye-labeled fragments. Peaks in the intensity of dye-labeled fragments, representing taxonomic units (or ribotypes), are output on a graph. After aligning peaks and filtering noise, these graphical data were converted to a sample by ribotype presence-absence matrix for statistical analysis using the program T-REX (Culman et al. 2009). T-RFLP has two principle limitations: it cannot be used to identify taxonomic species and because different species may have the same restriction enzyme binding sites, multiple species are often represented by the same peak. Nevertheless, T-RFLP is a highly effective method for assessing the dissimilarity among community samples (Gaston et al. 1997, Hughes 2000). Although I recognize that each ribotype does not necessarily represent a unique species, I hereafter refer to ribotypes as species for simplicity.

3.2.4 Statistical Analysis Species richness and composition were used as descriptors of community structure. Species richness is the number of species per sample and species composition was quantified using non-metric multidimensional scaling (NMDS). NMDS was performed using the vegan package (Oksanen et al. 2011) of the program R 2.15.1 (R Development Core Team 2012). I specified a two-axis solution based on a Bray-Curtis distance matrix and presence-absence data. Stress was low (0.196) and the two axes explained 87.7% of the variation in the original distance matrix. The first axis was highly correlated with species richness (r= 0.887) so I used the second orthogonal axis to represent species composition, independent of species richness.

39

I used a series of linear mixed effects models (lme() function; Pinheiro et al. 2012) in R with an AIC-based stepwise model selection process to examine effects of patch size, arrangement, and matrix quality on community structure and ecosystem function. Because multiple samples were taken from each landscape, landscape ID was used as a random grouping variable. Spatial block and litter origin were included as covariates, as were soil temperature and pH data in analyses of litter decomposition. 3.3 Results T-RFLP analysis of the 16S rRNA gene showed that each 0.25 g sample of soil and litter contained on average 65.9 (±29.5 SD) ribotypes, with each ribotype representing at least one (but likely multiple) taxonomic species. The high diversity of bacteria, illustrates the complex structure of the soil and litter food web. The number and composition of species within oak litter patches is affected by patch size, arrangement, and matrix quality (Fig. 3.2, Tables A1 and A2). Large patches, isolated patches, and patches surrounded by a matrix of pine litter contained more species than small patches, connected patches, and patches surrounded by a bare ground matrix, respectively (Fig. 3.3). The effect of landscape structure on community structure had consequences for the rate of leaf litter decomposition. Litter bags on isolated oak patches and patches surrounded by a pine litter matrix decomposed significantly slower than litter bags on connected patches and patches surrounded by a bare ground matrix (Fig. 3.4, Table B1). The effect of landscape structure on the rate of local decomposition is likely mediated by the landscape effect on local community structure. The number and composition of species in oak litter patches had a significant effect on the rate of litter decomposition (Fig. 3.5, Table B2). Bacterial species richness is negatively correlated with the rate of leaf litter decomposition. Matrix quality mediates the effects of patch size and arrangement at least in part by providing a novel source of species with which oak litter communities interact. The significant main effects of patch size and arrangement seen in landscapes with a bare ground matrix disappear in landscapes with oak patches surrounded by a pine litter matrix (Fig. 3.6). Instead, there is a significant effect of pine litter community structure on oak litter community structure either as a main effect in the case of species richness or as an interaction with patch size and arrangement in the case of species composition (Fig. 3.6, Tables C1-C4).

40

3.4 Discussion The equilibrium theory of island biogeography (MacArthur and Wilson 1967) has been influential in how we think patch size and arrangement should affect communities. The expectations that smaller, more isolated patches should contain fewer species than larger, more connected patches have often been applied to studies of habitat loss and fragmentation in terrestrial habitats (Debinski and Holt 2000). More recently, researchers have demonstrated the importance of matrix habitat quality for communities, often finding that the number and abundance of species increases with increasing suitability of the matrix (Ricketts 2001, Brotons et al. 2003). Results presented here agree in part: the number of species increases with patch size and the presence of a pine litter matrix (Figs. 3.3a and 3.3c), which is assumed to be of higher quality than a bare ground (i.e., bare landscaping fabric) matrix devoid of resources and cover from harsh environmental conditions or predation. The positive relationship between species richness and habitat area is one of the most widely documented relationships in ecology (Rosenzweig 1995). For example, increasing habitat area provides more resources, allows for species with larger home ranges, and for larger population sizes, each of which reduces extinction risk. Increasing matrix quality may have a positive effect on species richness by effectively increasing suitable habitat area or easing dispersal among patches for some species. However, as discussed more below, matrix quality may also affect local species richness by serving as a source of colonizers. Contrary to predictions based on island biogeography theory, however, the number of species in this study is greater in isolated, rather than connected, patches (Fig. 3.3b). Increasing isolation is thought to decrease species richness by reducing the ease of dispersal among patches, thereby reducing effective habitat are in the landscape, limiting access to resources, and precluding rescue effects. The positive relationship between richness and isolation observed here may be a result of spatially dependent interspecific interactions, as opposed to each species separately responding to environmental conditions as sets of non-interacting populations. For example, isolated patches may serve as refuges from predators (Huffaker 1958). Some microarthropods prey on bacteria, as well as other detritivores, which can cause cascading effects through the leaf litter food web (reviewed in Fowler and Lawton 1982). Because some microarthropods require greater connectivity among patches for persistence (Gonzalez et al. 1998) bacteria may escape predation in isolated patches. Competition-colonization trade-offs,

41 where inferior competitors persist in the landscape if they can disperse to patches not occupied by superior competitors (Levins and Culver 1971), may also explain an increase in species richness with isolation. Mouquet and Loreau (2003) showed how a decrease in species’ dispersal rates, as can be associated with patch isolation (Fahrig and Merriam 1985), can increase the number of species persisting locally by allowing inferior competitors to escape exclusion if a superior competitor cannot disperse to the patch. However, species richness peaked at an intermediate level of dispersal, suggesting that greater isolation may have a negative effect on local species richness. Species composition is also dependent on landscape structure. Composition varies with matrix quality but there are interactive effects of patch size and arrangement on composition (Fig. 3.2). This may be due to variation among species in dispersal ability and whether or not a particular matrix type enhances or inhibits dispersal. Matrix quality may also affect local composition by serving as a source of colonizing species (Cook et al. 2002). The effect of matrix pine litter communities on oak litter communities (Fig. 3.6b) suggests that spillover from the surrounding matrix is a source of variation in the effect of the matrix on local communities. The landscape-dependent effects on community structure translate to an effect on the ecosystem function of leaf litter decomposition. The rate of leaf litter decomposition is influenced by bacterial species composition and bacterial species richness (Fig. 3.4), both of which are affected by patch arrangement and matrix quality, and to a lesser extent, patch size. As such, the rate of leaf litter decomposition slows with isolation and the presence of a pine litter matrix but does not vary with patch size (Fig. 3.5). Though there is a significant interaction between species richness and pH, the interaction does not affect the sign of the effect of richness on decomposition, only the strength of the effect. The strong effects of species composition and the negative effect of the pine litter matrix on decomposition suggest that pine litter species that invade and persist in oak patches may displace oak litter species that are better adapted to decomposing oak leaves (as in Fig. 5.9 in Coleman et al. 2004). However, since species richness is generally higher in patches surrounded by a pine matrix and is negatively correlated with the rate of decomposition, it may be more likely that invading pine litter species are somehow inhibiting the function of oak litter decomposers, possibly by affecting the local pH. Oak litter pH is not affected by the pH of the surrounding pine matrix (F1, 46 = 0.26, P = 0.61) indicating

42 that flows of material from surrounding landscape do not affect the local physiochemical environment. Although structural features of the landscape are important for local community structure and ecosystem function, matrix quality can mediate the effects of patch size and arrangement on local communities (Fig. 3.6). The quality of the matrix is known to affect factors such as the ease of dispersal among patches (Ricketts 2001) but my results show that the matrix can affect metacommunity dynamics by providing an opportunity for interactions with communities originating in the matrix. My results indicate that matrix communities can be integral parts of the metacommunity, with dynamics playing out across continuous landscapes with indiscrete patch boundaries. Integrating communities with indistinct boundaries into experimental studies will be a key part of moving metacommunity concept forward.

3.4.1 Future directions It will be important to expand the dataset to include microarthropods to provide a more complete analysis of the leaf litter food web. Microarthropods, such as mites (Acari) and springtails (Collembola), are active dispersers that likely facilitate the passive movement of bacteria among patches (Coleman et al. 2004). Some microarthropods also facilitate decomposition by shredding and breaking apart leaf material, whereas others may help regulate communities through predation of other microarthropods or grazing of microbial species. I will also enhance the community datasets by using a next-gen sequencing method on the Illumina MiSeq platform that will provide species-level identifications (as opposed to ribotype-level identification) and provide taxonomic information that will allow for some level of functional classification. A better understanding of the functional diversity may help clarify patterns of leaf litter decomposition observed in this study. Because of the high diversity of decomposers present in any one sample, it may be that functional diversity and functional complementarity provide a better indicator of the ability of a community to perform the function of decomposition (Heemsbergen et al. 2004).

43

3.5 Figures

1 m 10 cm

1 m

Oak litter (Quercus geminata)

Pine litter (Pinus clausa)

Bare ground

Figure 3.1. Experimental design to scale. Each 1 m2 landscape contains four patches, varying in size and isolation, of the focal oak litter (dark green), which is surrounded by a matrix of either bare ground (black) or pine litter (light green). The control landscape consists of all oak litter. Samples were collected from pseudopatches in the control landscape and the pine litter matrix (dashed circles). A litterbag (brown square) was placed on the surface of the NW and SE patches of each landscape.

44

Oak patch size

Oak patch 0.021, 0.003 Oak litter arrangement community

Matrix quality

Figure 3.2. Structural features of the landscape affect local species richness and composition. Solid blue arrows represent main effects and dashed arrows represent interactions. P-values are given along arrows for species richness (normal font) followed by composition (bold). Full model results are given in Tables A1 and A2.

45

80 A

70

Species Species richness 60

50 Large Small

80 B

70

Species Species richness 60

50 Connected Isolated

80 C

70

60 Species Species richness

50 Bare ground Pine litter

Figure 3.3. Effect of landscape structure on oak litter species richness. Bars represent means ±SE, N = 12 for each treatment.

46

30% PinePine matrix litter matrix EmptyBare groundmatrix matrix 25%

20%

15%

10% Oak leafmassOakloss

5%

0% Isolated Connected Isolated Connected Control Small Large

Figure 3.4. Mean percent oak litter mass loss after 12 months with standard error bars. The rate of leaf litter decomposition in oak litter patches surrounded by a bare ground matrix is faster than decomposition in patches surrounded by a pine litter matrix (P = <0.0001). Decomposition in connected patches is faster than decomposition in isolated patches (P = 0.0007). However, there is no effect of patch size on decomposition rate. Full model results are given in Table B1.

47

Oak species richness

Oak litter pH 0.037 decomposition

Oak species composition

Figure 3.5. Local community effect on oak leaf litter decomposition. Full model results are given in Table B2.

48

A Oak patch size

NA, <0.001 Oak litter community Oak patch arrangement

B Oak patch size

Pine litter 0.041, <0.001 Oak litter community community

Oak patch arrangement

Figure 3.6. Effects of oak patch size and arrangement on oak litter community structure (species richness and composition) in landscapes with a bare ground matrix (A) and a pine litter matrix (B). Solid blue arrows represent main effects and dashed arrows represent interactions. P-values are given along arrows for species richness (normal font) followed by composition (bold). Full model results are given in Tables C1-C4.

49

CHAPTER FOUR

THE CONSEQUENCES OF MULTIPLE INDIRECT PATHWAYS OF INTERACTION FOR SPECIES COEXISTENCE

4.1 Introduction Biodiversity is dependent, in part, on interspecific interactions that play out within ecological networks. The structure of a network is formed of pairwise direct interactions, such as predation, competition, and mutualism. But species can also interact indirectly via one or more intermediate species (reviewed in Wootton 1994). Though many species do not interact directly, indirect effects permeate ecological networks, linking the dynamics of all species. Accordingly, the strength of indirect effects can have important consequences for patterns of diversity and species coexistence (Higashi and Patten 1989, Menge 1995, Borrett et al. 2010). It has been recognized for some time that understanding the dynamics of a particular pair of species often requires considering the dynamics of other species in the community that may interact only indirectly (Levine 1976, Stone and Roberts 1991, Miller 1994). However, real communities tend to have network architectures that can lead to more than one indirect pathway between pairs (Higashi and Nakajima 1995, Yodzis 2000). Developing an understanding of how indirect effects are partitioned among multiple and simultaneously acting pathways is important for understanding mechanisms of coexistence. Apparent competition is a well-known type of negative indirect interaction (Holt 1977) and can be an important driver of community structure (e.g., Bonsall and Hassell 1997, Morris et al. 2004). A shared predator can link the dynamics of two species that do not directly interact: an increase in the density of one species allows the predator population to increase, and thus prey more heavily on the other. When not considering the predator, the negative response of one species to the presence of other may give the appearance of competition. Such a negative indirect effect could also be transmitted along other pathways (Connell 1990), yet few of the alternative pathways have been explored. Mutualism can provide a pathway for a negative indirect effect to be conducted through lower trophic levels. For example, using the conceptual model illustrated in figure 4.1, one

50 scenario involves reciprocal negative indirect interactions between two ant species that each specialize on separate competing myrmecophytes (plants with adaptations for hosting ant colonies). These hypothetical ant species are obligate mutualists (M) that obtain domatia and/or other limiting resources from myrmecophyte plants (R) in exchange for protection from herbivores and encroaching vegetation (Janzen 1966). If protection by one ant species (M1) results in an increase in its mutualist plant’s (R1) density, the interspecific interaction with a competing plant (R2) may intensify, reducing the density of the competitor plant (R2). At a lower density, the competitor (R2) will be less capable of providing benefit to its mutualist ant partner

(M2). A negative indirect effect between ant species (M1 and M2) that do not otherwise interact directly is thereby transmitted via their competing plant resources (R1 and R2). The magnitude of the indirect effect is dependent on the strengths of the mutualisms and the competition between resource species. To the extent that the fundamental feature of apparent competition is that two species resemble competitors because of a negative indirect effect, the indirect interactions between M1 and M2 can be described as resource-mediated apparent competition. This is just one example of an as of yet unexplored pathway for apparent competition. A predator generally has a negative effect on the abundance of prey, and in the absence of other regulating factors, predation will result in the exclusion of the prey species most vulnerable to predation (the weaker apparent competitior; Holt 1984). If predator- and resource- mediated pathways of apparent competition are acting in concert, one might predict an additive effect that would hasten the extinction of the weaker apparent competitor. However, predation is known to promote coexistence in some systems where resource competition would otherwise result in the exclusion of an inferior competitor (Paine 1966, Caswell 1978, Inouye et al. 1980). A predator may promote coexistence by acting to stabilize the dynamics of its prey (reviewed in Chesson 2000). For example, a shared predator may reduce population sizes and therefore moderate the strength of interactions between prey species that occur via other pathways. A predator may also exhibit frequency-dependent prey switching, thereby reducing the pressure of predation as a species becomes rare. Thus, a shared predator may have effects on the relationship between species in the presence of multiple interaction pathways that could not be predicted based on a single pathway. In real communities, competition, predation, and mutualism can happen simultaneously. We therefore explore a simple interaction web that combines these three types of direct

51 interactions (Fig. 4.1). We will focus on two species at an intermediate trophic level. These species interact only indirectly, but via two different and simultaneously acting pathways: resource- and predator-mediated pathways. We use the inverse of the community matrix to quantify the strength of interactions in order to examine the resulting effects on species coexistence. We then use an approach developed by Nakajima and Higashi (1995) to evaluate the partial effect propagated along a particular pathway when multiple pathways are acting simultaneously. Additionally, the quality of service provided to a partner can vary among species (Schemske and Horvitz 1984, Francis and Read 1995, Miller 2007) and also temporally within species (Grinath et al. 2012), so not only the presence of a mutualism, but the quality of service provided may play an important role in determining a species’ abundance or persistence. We are therefore interested in understanding how indirect effects are influenced by the mutualistic benefit that the intermediate trophic level provides to their resource species in different competitive and predation environments.

4.2 The Model The community in our model is comprised of five species: a predator (P), two prey species (M1 and M2), and two species at a lower trophic level (R1 and R2). Each of the prey species have a specialized mutualism with their sole resource (R1 or R2), and the resource species engage in interspecific competition (Fig. 4.1). We use a system of differential equations modified from the commonly used Lotka-Volterra competition and predation models to describe population dynamics in the five-species community. Because the network is symmetric, we show equations for Ri and Mi, where i = 1 or 2.

(1)

(2)

(3)

52

For the lowest trophic level, Ri (Eq. 1), ri is the intrinsic rate of growth, Ki is the single-species carrying capacity, and αij is the per-capita competitive effect of Rj on Ri. The per capita benefit

Mi provides Ri is represented by γi, and increases the population growth rate of Ri following a type II functional response, where si and hi determine the asymptotic benefit of the mutualism and rate of saturation, respectively. Because the specific mechanism by which a mutualism influences the population dynamics of partners will vary by system, a modification of the intrinsic rate of growth (Graves et al. 2006) or some other demographic rate may be more appropriate for a particular study.

The obligate mutualist Mi converts resources provided by Ri to new growth at a rate bi, while suffering density dependent (ei) and density independent (di) mortality (Eq. 2). Incorporating density dependent mortality in this way (Neuhauser and Fargione 2004) allows for stable coexistence of mutualistic partners by preventing unbounded growth (May 1981). Though

γi is a measure of the beneficial effect of Mi on Ri, increasing γi has a reciprocal benefit for Mi in the form of a larger population of its sole resource. We therefore refer to γi as the strength of mutualism between Mi and Ri.

The predator, P, preys on M1 and M2, converting prey to population growth at a rate vi (Eq. 3). As above, we use a type II functional response, assuming that the predator has a saturating ability to process capture and process resources, where ci is the maximum capture rate and fi is the rate of saturation.

4.2.1 Model Analysis

The community matrix (Jacobian), A, represents the set of all direct pairwise interactions between species in a community (Levins 1968, Vandermeer 1970). The non-zero elements of A are Aij = ( Ni/ t)/ Nj, with all species N evaluated at equilibrium, i.e. the partial derivative Aij j denotes the direct effect of a change in density of species on the equilibrium density of species i, with other species in the network held constant. In the absence of a direct interaction between -1 two species Aij = 0. The negative inverse of the community matrix, -(A ), yields the net interaction coefficients of all species in the community (Levine 1976, Bender et al. 1984, Stone and Roberts 1991, Schmitz 1997). Net interaction coefficients represent the sum of direct and indirect interactions, through all possible pathways, between each species pair in the community, evaluated at equilibrium. The net indirect effect of any species on another is thus the difference 53

-1 between the entries of A and -(A ). Because there is no direct interaction between M1 and M2, the net effect is purely indirect and reciprocal negative values represent apparent competition. We partition two simultaneously acting pathways of indirect effects using the conjugate variable approach, which was developed by Nakajima and Higashi (1995), and then independently by Yodzis (2000). The partial effect of species j on species i that passes through species k is given by

Where a are elements of (A–1). The remainder of the net effect that does not pass through ij – species i is then equal to

The conjugate variable approach is used to quantify the effect of a particular species (k) in mediating a net effect between two others (i and j). Because there is only one species (the predator, P) between M1 and M2 along the predation pathway, we interpret (or ) as the portion of the total indirect effect that is transmitted along this pathway. Further, we interpret the difference between the net effect and as the partial effect transferred along the resource pathway.

We modeled the strength of the indirect effect of M1 on M2 across a range of strengths of mutualistic benefit (γi, Eq. 1) provided by M2 to R2 in different predation and competitive environments. To assess the effect of the two indirect pathways on the strength of the total indirect effect, most parameter values for M1 and M2, and for R1 and R2 were made symmetrical, except that M1 was given a slight advantage in γ (+0.01) and c (-0.005) in order to avoid a singularity in the inverse community matrix (see the legend of figure 4.2 for parameter values). To assess the effect of γ and multiple pathways of apparent competition on species coexistence we modeled asymmetric mutualism strengths across a range of γ from 0 to 0.15 in different predation and competitive environments. This yields a coexistence surface, which demonstrates the indirect competitive limits to persistence in a particular environment. The model was implemented in R v2.15.1 (R Development Core Team 2012). Equilibrium abundances were approximated using the fourth order Runge-Kutta method. Simulations run for 25,000 iterations or until a steady state was achieved using the R program ‘rootSolve’ (Soetaert and Herman 2009). A species was considered extinct if its density fell below 0.0001.

54

4.3 Results and Discussion Indirect effects following different pathways act to influence equilibrium abundances and persistence in ecological communities. The presence of a pathway from M1 to M2 through a shared predator decreases the equilibrium abundance of M2 as compared to when the predator is absent (Figs 4.2a and 4.2b) and thus M1 and M2 are linked only via the lower trophic level. However, the effects of combining a pathway through competing resource species with a pathway through a shared predator alters the equilibrium abundance very little over most of the range of mutualism strength in which M1 and M2 coexist.

The equilibrium abundance of M2 increases with increasing strength of mutualism (γ) over most of the range that it persists when the resource-mediated pathway is acting alone and when competition between resources is relatively weak (Fig 4.2a). However, the benefit of increasing γ decreases as the strength of competition between resources intensifies (Fig. 4.2b). In the presence of a shared predator, increasing γ decreases equilibrium abundance across the entire range of γ, whether or not the resource-mediated pathway is present.

The sign of the indirect effect of M1 on M2 and vice versa is negative across the range of parameter values modeled here (Figs. 4.2c and 4.2d). Because they do not compete for a shared resource, we consider this to be a form of apparent competition mediated by two different pathways. The strength of apparent competition, through any combination of pathways, generally increases with γ until a threshold is reached near the level of γ at which M2 is driven extinct. Near this threshold is where the small differences between M1 and M2 in γ and c are revealed. If a shared predator provides a second pathway for an indirect interaction, an additive effect of each pathway might be expected such that the exclusion of the weaker apparent competitor would be hastened. Though equilibrium densities are lower (Figs. 4.2c and 4.2d), the presence of a shared predator allows for coexistence over a wider range of γ than when a resource-mediated pathway is present alone. The enhanced coexistence with the inclusion of the predator pathway illustrates how the distinct pathway effects are not additive. The inferior apparent competitor (M2) persists over a wider range of parameter space with the inclusion of a predator, even though it has a slightly greater capture rate by the predator. This may be due to frequency-dependent prey switching. As M2 becomes increasingly rare, a greater proportion of the predator’s diet is comprised of M1. In addition, the positive effect of the predator on species 55 coexistence increases with the strength of competition. Indeed, the net negative effect of the predator on M2 diminishes and even becomes positive as the strength of competition between resources increases (Fig. 4.3). The form of predator-mediated coexistence demonstrated here illustrates how predators can stabilize community dynamics by reducing equilibrium abundances, thereby moderating the strength of direct and indirect interactions, which increases biodiversity by allowing all species in the community to persist over a broader range of ecological conditions. In order to understand the how the indirect effect is propagated through simultaneously acting pathways we partitioned the effect using the conjugate variable approach (Nakajima and Higashi 1995; Fig. 4.4). As the strength of γ increases, the proportional contribution of each pathway to the total indirect effect shifts from the predator pathway to the pathway mediated by the competing resources. The magnitude of the partial effect transmitted through the resources appears to be more strongly dependent on the strength of the mutualism than the partial effect transmitted through the predator. Moreover, when competition between resources is weaker (α = 0.2), the partial effect mediated by the predator is dominant over much of the range of γ. However, when competition is stronger (α = 0.6), the partial effect mediated by the resources is dominant over a greater proportion of the range of γ. Increasing the strength of competition can even cause the partial effect mediated by the predator to become positive over a narrow range of γ. When both pathways of apparent competition are acting simultaneously and γ is different for M1 and M2, the coexistence of M1 and M2 depends on the strength of competition and predation (Fig. 4.5). With an equal predation rate (c = 0.1), the range of γ over which apparent competitors coexist decreases with the strength of competition (α) between their resources (Fig. 4.5a). Conversely, when the strength of competition is held constant (α = 0.5), the range of γ over which M1 and M2 coexist increases with the strength of predation (Fig. 4.5b), further demonstrating that predator-mediated coexistence is enhanced with the increasing capture rate even when it might be expected (if effects were additive) that this would generate stronger apparent competition and reduce the opportunity for coexistence.

56

4.3.1 Community-level effects of mutualism

Mutualist pollinators, protectors, cleaners, and dispersers confer important benefits on their partners (reviewed in Boucher et al. 1982) that likely change the way the receivers of benefits interact with other species in the community. The pathway of resource-mediated apparent competition in our model relies on such an effect of mutualism to carry the indirect effect from one apparent competitor to the other. We know of no empirical work that has looked for this potential pathway of apparent competition, which is consistent with the general paucity of studies of effects of mutualisms on the broader community. Considering the extensive literature on mutualism, exceedingly little is known about the community-level consequences of a mutualism between two species. The effects of arbuscular mycorrhizal fungi-plant mutualisms on plant-plant competitive interactions are one of the few empirical examples. This type of mutualism is thought to be crucial for structuring plant communities by altering competitive interactions (Hart et al. 2003). In one of the few non-mycorrhiza examples Yeaton and Bond (1991) showed that, though mutualism cannot fully explain coexistence between two shrub species, the rate of exclusion of a competitively inferior shrub is reduced in the presence of ant seed dispersers. Not only is the presence of a mutualism important, but the quality of beneficial services is also likely important for mediating interactions with other species in the community given that the quality of service can vary among (Schemske and Horvitz 1984, Bruna et al. 2004) and within species (Grinath et al. 2012). Our results suggest that the level of benefit a mutualistic species confers on its partner can have important consequences for the community at large. Even with no direct interactions, differences in the provision of mutualistic benefits (magnitude of γ) can determine whether M1 and M2 coexist (Fig. 4.5). Though we focus on M1 and M2 here, changes in γ can also alter the outcome of competitive interactions and coexistence of R1 and R2

(Fig. 4.6). If analyzed in isolation, the mutualism with Mi is facultative from the perspective of

Ri. However, when analyzed in combination with a competitor, the resource species can sometimes avoid competitive exclusion if it has a more beneficial partner (i.e. by increasing γ), indicating that the mutualism can be obligate depending on community context. Community context and many of the contemporary threats to biodiversity may both affect the quality of service provided if the identity of the mutualist changes (e.g, Kremen et al. 2002) or the nature of a species’ interaction with its host changes (e.g., Grinath et al. 2012). 57

4.3.2 Conclusions

Indirect effects can take multiple pathways resulting in a diverse set of outcomes, even in a relatively simple five-species system. A resource-mediated pathway for a negative indirect effect between species at an intermediate trophic level can mimic the effects of a predator- mediated pathway (i.e., apparent competition). The presence of both resource- and predator- mediated pathways allows coexistence over a greater range of parameter space than when only the resource-mediated pathway is present. The effects of different indirect pathways are not additive so understanding the mechanism of apparent competition is important for explaining how each pathway contributes to observed community dynamics. For example, empirical examination of only the predator pathway, when multiple pathways are in fact operating, may overestimate the predator’s contribution to the indirect negative effect between species.

58

4.4 Figures

P + + ––– M1 M2

+ + + + – – R1 – R2

Figure 4.1. Interaction network depicting the sign of the direct effects between species (solid arrows) and the pathways of apparent competition between M1 and M2 mediated by a shared predator (upper dashed arrow) and competing resources (lower dashed arrow).

59

Weak competition between Strong competition between

resources ( ij = 0.2) resources ( ij = 0.6) 35 A 35 B α α 30 30 Predator pathway only 25 25 Predator & resource pathways Resource pathway only 20 20

15 15

10 10

equilibrium abundance equilibrium 5 abundance equilibrium 5 2 2

M 0 M 0 0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15

γ γ 10 C 10 D

0 0

-10 -10

-20 -20

-30 -30 Interaction coefficient Interaction coefficient Interaction

-40 -40 0.00 0.05 0.10 0.15 0.00 0.05 0.10 0.15

Figure 4.2. The equilibriumγ abundance of M2 (A and B) and the strength ofγ the total indirect effect of M1 on M2 (C and D) and varies with the benefit Mi provides to Ri (γ) and the pathway(s) of the indirect effect (different shades). Results are shown for relatively weak (αij =0.2; A and C) and relatively strong (αij =0.6; B and D) competition between R1 and R2. Panels A and C correspond to the same set of model runs, as do panels B and D. Initial population densities: Ri = 10, Mi = 5, P = 5. Other parameter values: ri = 1.0, Ki = 10, γ = γ2, γ1 = γ2 + 0.01, bi = 0.2, si = 0.3, hi = 0.2, di = 0.1, ei = 0.1, vi = 0.1, c2 = 0.095, c1 = c2 – 0.005, fi = 0.1, m = 0.1.

60

10

0

2 M -10

-20

-30 α = 0.8 -40 α = 0.6 α = 0.4 Net predator effect on effect predator Net α = 0.2 -50 α = 0.0

-60 0.00 0.05 0.10 0.15

Figure 4.3. Net effect of the predator on M2 forγ a range of resource competition strength and fixed encounter rate (c1 = 0.095). Other parameter values are the same as indicated in Fig. 4.2.

61

10

0

-10

-20

Predator, α = 0.6 -30 Resource, α = 0.6

Partial interaction coefficient interaction Partial Predator, α = 0.2 Resource, α = 0.2 -40 0.00 0.05 0.10 0.15

Figure 4.4. The net indirect effect of M1 onγ M2 partitioned between simultaneously acting pathways. The partial strengths of the effect transmitted through the predator (black) and resource (gray) pathways are shown for weak (dashed; αij = 0.2) and strong (solid; αij = 0.6) competition between resources. Summing the two dashed or two solid lines respectively equal the total apparent competitive effect of M1 on M2 as depicted by the medium gray lines in figs. 2A and 2B. Parameter values are the same as for Fig. 4.2.

62

Figure 4.5. Coexistence surface of M1 and M2 across a range of γ in different competitive (A) and predation (B) environments. Shaded areas indicate coexistence and darker shades overlap areas of lighter shades. (A) Increasing the strength of competition between resources decreases the area of coexistence. At a fixed capture rate (c1 = c2 = 0.1), the area of coexistence is greatest when competition is relatively weak, and the area of coexistence decreases as the strength of competition between resources increases. (B) Increasing the capture rate increases the area of coexistence (c1 = c2). For a fixed level of competition (αij = 0.5), increasing the capture rate allows for coexistence across a greater range of parameter space. Parameter values are the same as given in Fig. 4.2 except where noted.

63

Figure 4.6. Coexistence surface of R1 and R2 across a range of γ in different competitive environments (A) and capture rates of M1 and M2 (B). Areas of coexistence represented by lighter shades are included in the darker shaded areas. (A) At a fixed level of predation (c1 = c2 = 0.1), R1 and R2 coexist across the range of γ modeled except for when competition is strongest (αij = 0.8). (B) Increasing the capture rate of the mutualist partners of R1 and R2 has the opposite effect (c1 = c2). For a fixed level of competition (αij = 0.5), increasing the capture rate allows for coexistence across a greater range of parameter space. Parameter values are the same as given in Fig. 4.2 except where noted.

64

CHAPTER FIVE

CONCLUSION

The structure of ecological communities and the functions they perform are dependent on interspecific interactions and multi-scale environmental variation. With this dissertation I have explored this idea using a large-scale observational field study, a small-scale experimental microcosm, and in a mathematical model. In chapter two, I used a large scale field study to investigate the response of communities of plants and their pollinators to habitat loss in the landscape. I found that, as is often the case, the total number of species in a community decreases with habitat loss in the surrounding landscape. The loss of species has the effect of altering the architecture of the network formed by the set of interactions between species. With increasing habitat loss, and with fewer species, networks become less nested and more modular. This shift in architecture may reduce the stability of communities that have already suffered the negative effects of habitat loss (i.e. species loss), possibly putting them at risk for further local extinction. In chapter three I used an experimental microcosm to examine the effects of the amount, arrangement, and composition of habitat in the landscape on the local richness and composition of species, and the resulting effects on an important ecosystem function: leaf litter decomposition. The patch size and arrangement of the focal habitat (oak leaf litter) affected the number and composition of species but this effect was mediated by the quality of the matrix habitat (pine leaf litter or bare ground) in the area intervening focal patches. The landscape effect on local community structure translated to an effect on the rate of leaf litter decomposition. Results show that some of the variation in local communities can be attributed to the movement of organisms between focal and pine litter habitats, suggesting that matrix communities can be integral components of a metacommunity. I used mathematical modeling in chapter four to examine how a net indirect effect between two species, is partitioned among multiple pathways. Communities form interaction networks that contain relatively few direct interactions but are fully connected by indirect interactions. As a result there are often multiple pathways for an indirect effect to be transmitted between any two species. I modeled the population dynamics of a relatively simple five-species

65 community: a predator that preys on the two species that provide mutually beneficial services to their specialized resources, which are resource competitors. I focus on the two species at the intermediate trophic level that interact only indirectly via two pathways: through their shared predator, or through their competing resources. Model results show that even though each pathway transmits a negative indirect effect in the absence of the other pathway, both pathways acting simultaneously can enhance coexistence. The inclusion of the predator pathway dampens the population-level strength of interspecific interactions thereby reducing the magnitude of the net negative effect between species. This suggests that understanding the mechanism of the effect of one species on another could be important for targeted species management.

66

APPENDIX A

LANDSCAPE STRUCTURE EFFECTS ON COMMUNITY STRUCTURE

Table A1. Type III summary of linear mixed effects model testing the effect of oak patch size, arrangement, and matrix quality on species composition in oak patches.

Response variable: oak litter species composition 2 df P Size 3.5568 1 0.05930 Arrangement 9.0803 1 0.00258 Matrix 5.9037 1 0.01511 Litter origin 4.1722 1 0.04109 Block 3.7958 1 0.05138 Size:Arrangement 7.3190 1 0.00682 Size: Litter origin 7.0205 1 0.00806 Arrangement: Litter origin 9.6286 1 0.00192 Size:Arrangement: Litter origin 17.0469 1 0.00004

Table A2. Type III summary of linear mixed effects model testing the effect of oak patch size, arrangement, and matrix quality on species richness in oak patches.

Response variable: oak litter species richness 2 df P Size 3.4756 1 0.06228 Arrangement 5.3368 1 0.02088 Matrix 5.5022 1 0.01899 Litter origin 6.5166 1 0.01069 Block 5.3684 1 0.02050

67

APPENDIX B

LANDSCAPE AND COMMUNITY STRUCTURE EFFECTS ON ECOSYSTEM FUNCTION

Table B1. Type III summary of linear mixed effects model testing the effect of patch size, arrangement, and matrix quality on the rate of oak leaf litter decomposition.

Response variable: oak litter mass loss 2 df P Matrix 28.966 1 < 0.00001 Arrangement 11.018 1 0.00090 Litter origin 46.817 1 < 0.00001 Block 18.391 2 0.00010

Table B2. Type III summary of linear mixed effects model testing the effect of oak species richness and composition on the rate of oak leaf litter decomposition.

Response variable: oak litter mass loss 2 df P Species richness 8.0685 1 0.00450 Species composition 18.2297 1 0.00001 pH 4.3659 1 0.03667 Litter origin 4.8877 1 0.02705 Species richness:pH 4.1290 1 0.04215 Species richness: Litter origin 10.7287 1 0.00106 Species richness:Block 5.6618 1 0.01734 Species composition: Litter origin 16.0766 1 0.00006 Species composition:Block 5.8147 1 0.01590 pH:Block 17.6096 1 0.00003

68

APPENDIX C

MATRIX QUALITY MEDIATES THE EFFECTS OF PATCH SIZE AND ARRANGEMENT ON OAK LITTER COMMUNITY STRUCTURE

Table C1. Type III summary of linear mixed effects model testing the effect of a bare ground matrix on species richness in oak patches.

Response variable: oak litter species richness 2 df P Size 1.1036 1 0.29347 Arrangement 5.9323 1 0.01487

Table C2. Type III summary of linear mixed effects model testing the effect of bare ground matrix on species composition in oak patches.

Response variable: oak litter species composition 2 df P Size 10.524 1 0.00118 Arrangement 14.749 1 0.00012 Size:Arrangement 22.002 1 < 0.00001 Size:Arrangement: Block 22.650 4 0.00015

Table C3. Type III summary of linear mixed effects model testing the effect of a pine litter matrix and pine species composition on species composition in oak patches.

Response variable: oak litter species composition 2 df P Pine species composition 12.0090 1 0.00053 Block 13.2786 1 0.00027 Pine species composition:Size 4.1738 1 0.04105 Pine species composition:Arrangement 7.4113 1 0.00648 Pine species composition:Size:Arrangement 3.7573 1 0.05258

69

Table C4. Type III summary of linear mixed effects model testing the effect of a pine litter matrix and pine species richness on species richness in oak patches.

Response variable: oak litter species richness 2 df P Size 2.5488 1 0.11038 Pine species richness 4.1871 1 0.04073 Size:Pine species richness 1.9334 1 0.16439

70

REFERENCES

Albrecht, M., M. Riesen, and B. Schmid. 2010. Plant-pollinator network assembly along the chronosequence of a glacier foreland. Oikos 119:1610-1624.

Amarasekare, P., M. F. Hoopes, N. Mouquet, and M. Holyoak. 2004. Mechanisms of coexistence in competitive metacommunities. American Naturalist 164:310-326.

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

Arbuckle, J. L. 2003. AMOS 5.0.1. AMOS Development Corporation, Spring House.

Barabasi, A. L. and R. Albert. 1999. Emergence of scaling in random networks. Science 286:509-512.

Bascompte, J. and P. Jordano. 2007. Plant- mutualistic networks: The architecture of biodiversity. Annual Review of Ecology Evolution and Systematics 38:567-593.

Bascompte, J., P. Jordano, C. J. Melian, and J. M. Olesen. 2003. The nested assembly of plant- animal mutualistic networks. Proceedings of the National Academy of Sciences of the United States of America 100:9383-9387.

Bastolla, U., M. A. Fortuna, A. Pascual-Garcia, A. Ferrera, B. Luque, and J. Bascompte. 2009. The architecture of mutualistic networks minimizes competition and increases biodiversity. Nature 458:1018-U1091.

Benadi, G., N. Blüthgen, T. Hovestadt, and H. J. Poethke. 2012. Population dynamics of plant and pollinator communities: stability reconsidered. American Naturalist 179:157-168.

Bock, C. E. and R. E. Ricklefs. 1983. Range size and local abundance of some North American songbirds: A positive correlation. American Naturalist 122:295-299.

Bommarco, R., J. C. Biesmeijer, B. Meyer, S. G. Potts, J. Poyry, S. P. M. Roberts, I. Steffan- Dewenter, and E. Ockinger. 2010. Dispersal capacity and diet breadth modify the response of wild bees to habitat loss. Proceedings of the Royal Society B-Biological Sciences 277:2075- 2082.

Bonsall, M. B. and M. P. Hassell. 1997. Apparent competition structures ecological assemblages. Nature 388:371-373.

Borrett, S. R., S. J. Whipple, and B. C. Patten. 2010. Rapid development of indirect effects in ecological networks. Oikos 119:1136-1148.

Brotons, L., M. Monkkonen, and J. L. Martin. 2003. Are fragments islands? Landscape context and density-area relationships in boreal forest birds. American Naturalist 162:343-357.

71

Brown, J. H. 1984. On the relationship between abundance and distribution of species. American Naturalist 124:255-279.

Bruna, E. M., D. M. Lapola, and H. L. Vasconcelos. 2004. Interspecific variation in the defensive responses of obligate plant-ants: experimental tests and consequences for herbivory. Oecologia 138:558-565.

Caswell, H. 1978. Predator-mediated coexistence: a non-equilibrium model. American Naturalist 112:127-154.

Chesson, P. 2000. Mechanisms of maintenance of species diversity. Annual Review of Ecology and Systematics 31:343-.

Coleman, D. C., D. A. Crossley, and P. F. Hendrix. 2004. Fundamentals of Soil Ecology, 2nd edition. Elsevier Academic Press, San Diego.

Connell, J. H. 1990. Apparent versus 'real' competition in plant communities.in J. B. Grace and D. Tilman, editors. Perspectives on plant competition. Academic Press, New York.

Cook, W. M., K. T. Lane, and R. D. Holt. 2002. Island theory, matrix effects and species richness patterns in habitat fragments. Ecology Letters 5:619-623.

Cottenie, K. and L. De Meester. 2005. Local interactions and local dispersal in a zooplankton metacommunity. Pages 189-211 in M. Holyoak, M. Leibold, and R. D. Holt, editors. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago.

Coûteaux, M. M., P. Bottner, and B. Berg. 1995. Litter decomposition, climate and litter quality. Trends in Ecology & Evolution 10:63-66.

Csardi, G. and T. Nepusz. 2006. The igraph software package for complex network research. InterJournal.

Culman, S. W., R. Bukowski, H. G. Gauch, H. Cadillo-Quiroz, and D. H. Buckley. 2009. T- REX: Software for the processing and analysis of T-RFLP data. BMC Bioinformatics 10:171.

Debinski, D. M. and R. D. Holt. 2000. A survey and overview of habitat fragmentation experiments. Conservation Biology 14:342-355.

Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics 34:487-515.

Fahrig L. and G. Merriam.1985. Habitat patch connectivity and population survival. Ecology 66:1762-68.

72

Fortuna, M. A. and J. Bascompte. 2006. Habitat loss and the structure of plant-animal mutualistic networks. Ecology Letters 9:278-283.

Fortuna, M. A., D. B. Stouffer, J. M. Olesen, P. Jordano, D. Mouillot, B. R. Krasnov, R. Poulin, and J. Bascompte. 2010. Nestedness versus modularity in ecological networks: two sides of the same coin? Journal of Animal Ecology 79:811-817.

Fowler, S. V. and J. H. Lawton. 1982. The effects of host-plant distribution and local abundance on the species richness of agromyzid flies attacking British umbellifers. Ecological Entomology 7:257-265.

Francis, R. and D. J. Read. 1995. Mutualism and antagonism in the mycorrhizal symbiosis, with special reference to impacts on plant community structure. Canadian Journal of Botany-Revue Canadienne De Botanique 73:S1301-S1309.

Fründ, J., K. E. Linsenmair, and N. Blüthgen. 2010. Pollinator diversity and specialization in relation to flower diversity. Oikos 119:1581-1590.

Gaston, K. J., T. M. Blackburn, and J. H. Lawton. 1997. Interspecific abundance range size relationships: An appraisal of mechanisms. Journal of Animal Ecology 66:579-601.

Gathmann, A. and T. Tscharntke. 2002. Foraging ranges of solitary bees. Journal of Animal Ecology 71:757-764.

Gonzalez, A., J. H. Lawton, F. S. Gilbert, T. M. Blackburn, and I. Evans-Freke. 1998. Metapopulation dynamics, abundance, and distribution in a microecosystem. Science 281:2045- 2047.

Grace, J. B. 2006. Structural Equation Modeling and Natural Systems. Cambridge University Press, New York.

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

Grinath, J. B., B. D. Inouye, N. Underwood, and I. Billick. 2012. The indirect consequences of a mutualism: comparing positive and negative components of the net interaction between honeydew-tending ants and host plants. Journal of Animal Ecology 81:494-502.

Guimera, R., D. B. Stouffer, M. Sales-Pardo, E. A. Leicht, M. E. J. Newman, and L. A. N. Amaral. 2010. Origin of compartmentalization in food webs. Ecology 91:2941-2951.

Handelsman, J. 2004. Metagenomics: Application of genomics to uncultured microorganisms. Microbiology and Molecular Biology Reviews 68:669-+.

Hanski, I. and M. Gilpin. 1991. Metapopulation dynamics: brief history and conceptual domain. Biological Journal of the Linnean Society 42:3-16.

73

Harrison, S. and E. Bruna. 1999. Habitat fragmentation and large-scale conservation: what do we know for sure? Ecography 22:225-232.

Hart, M. M., R. J. Reader, and J. N. Klironomos. 2003. Plant coexistence mediated by arbuscular mycorrhizal fungi. Trends in Ecology & Evolution 18:418-423.

Hättenschwiler, S., A. V. Tiunov, and S. Scheu. 2005. Biodiversity and litter decomposition in terrestrial ecosystems. Annual Review of Ecology Evolution and Systematics 36:191-218.

Heemsbergen, D. A., M. P. Berg, M. Loreau, J. R. van Haj, J. H. Faber, and H. A. Verhoef. 2004. Biodiversity effects on soil processes explained by interspecific functional dissimilarity. Science 306:1019-1020.

Higashi, M. and H. Nakajima. 1995. Indirect effects in ecological interaction networks .1. the chain rule approach. Mathematical Biosciences 130:99-128.

Higashi, M. and B. C. Patten. 1989. Dominance of indirect causality in ecosystems. American Naturalist 133:288-302.

Holt, R. D. 1977. Predation, apparent competition, and structure of prey communities. Theoretical Population Biology 12:197-229.

Holt, R. D. 1984. Spatial heterogeneity, indirect interactions, and the coexistence of prey species. American Naturalist 124:377-406.

Holt, R. D. 1993. Ecology at the mesoscale: The influence of regional processes on local communities. Pages 77-88 in R. E. Ricklefs and D. Schluter, editors. Species Diversity in Ecological Communities. University of Chicago Press, Chicago.

Holyoak, M., M. A. Leibold, and R. D. Holt, editors. 2005a. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago.

Holyoak, M., M. A. Leibold, N. Mouquet, R. D. Holt, and M. F. Hoopes. 2005b. Metacommunities: a framework for large-scale community ecology.in M. Holyoak, M. A. Leibold, and R. D. Holt, editors. Metacommunities: spatial dynamics and ecological communities. University of Chicago Press, Chicago.

Huffaker, C. B. 1958. Experimental studies on predation: Dispersal factors and predatory-prey oscillations. Hilgardia 27:343-383.

Hughes, J. B. 2000. The scale of resource specialization and the distribution and abundance of lycaenid butterflies. Oecologia 123:375-383.

74

Ievers-Landis, C. E., C. J. Burant, and R. Hazen. 2011. The concept of bootstrapping of structural equation models with smaller samples: An illustration using mealtime rituals in diabetes management. Journal of Developmental and Behavioral Pediatrics 32:619-626.

Inouye, R. S., G. S. Byers, and J. H. Brown. 1980. Effects of predation and competition on survivorship, fecundity, and community structure of desert annuals. Ecology 61:1344-1351.

Janzen, D. H. 1966. Coevolution of mutualism between ants and acacias in central america. Evolution 20:249-&.

Jules, E. S. and P. Shahani. 2003. A broader ecological context to habitat fragmentation: Why matrix habitat is more important than we thought. Journal of Vegetation Science 14:459-464.

Kareiva, P. 1987. Habitat fragmentation and the stability of predator-prey interactions. Nature 326:388-390.

Klein, A. M., Vaissiere, B. E., Cane, J. H., Steffan-Dewenter, I., Cunningham, S. A., Kremen, C., Tscharntke, T., 2007. Importance of pollinators in changing landscapes for world crops. Proceedings of the Royal Society of London. Series B: Biological Sciences 274:303-313.

Kremen, C., N. M. Williams, R. L. Bugg, J. P. Fay, and R. W. Thorp. 2004. The area requirements of an ecosystem service: crop pollination by native bee communities in California. Ecology Letters 7:1109-1119.

Kremen, C., N. M. Williams, and R. W. Thorp. 2002a. Crop pollination from native bees at risk from agricultural intensification. Proceedings of the National Academy of Sciences of the United States of America 99:16812-16816.

Kremen, C., N. M. Williams, and R. W. Thorp. 2002b. Crop pollination from native bees at risk from agricultural intensification. Proceedings of the National Academy of Science USA 99:16812-16816.

Laurienti, P. J., K. E. Joyce, Q. K. Telesford, J. H. Burdette, and S. Hayasaka. 2011. Universal fractal scaling of self-organized networks. Physica A: Statistical mechanics and its applications 390:3608-3613. Leckie, S. E. 2005. Methods for microbial community profiling and their application to forest soils. Forest Ecology and Management 220:88-106.

Leibold, M. A., M. Holyoak, N. Mouquet, P. Amarasekare, J. M. Chase, M. F. Hoopes, R. D. Holt, J. B. Shurin, R. Law, D. Tilman, M. Loreau, and A. Gonzalez. 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecology Letters 7:601-613.

Levine, S. H. 1976. Competitive interactions in ecosystems. American Naturalist 110:903-910.

Levins, R., and D. Culver. 1971. Regional coexistence of species and competition between rare species. Proceedings of the National Academy of Sciences of the USA 68:1246-1248.

75

Lima, S. L. and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: A review and prospectus. Canadian Journal of Zoology-Revue Canadienne De Zoologie 68:619- 640.

Logue, J. B., N. Mouquet, H. Peter, H. Hillebrand, and G. Metacommunity Working. 2011. Empirical approaches to metacommunities: a review and comparison with theory. Trends in Ecology & Evolution 26:482-491.

Loreau, M., N. Mouquet, and R. D. Holt. 2003. Meta-ecosystems: a theoretical framework for spatial ecosystem ecology. Ecology Letters 6:673-679.

Loreau, M., S. Naeem, and P. Inchausti, editors. 2002. Biodiversity and Ecosystem Functioning. Oxford University Press, Oxford.

Loreau, M., S. Naeem, P. Inchausti, J. Bengtsson, J. P. Grime, A. Hector, D. U. Hooper, M. A. Huston, D. Raffaelli, B. Schmid, D. Tilman, and D. A. Wardle. 2001. Biodiversity and Ecosystem Functioning: Current Knowledge and Future Challenges. Science 294:804-808.

MacArthur, R. H. and E. O. Wilson. 1967. The Theory of Island Biogeography. Princeton University Press, Princeton.

Menge, B. A. 1995. Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecological Monographs 65:21-74.

Miller, T. E. 1994. Direct and indirect species interactions in an early old-field plant community. American Naturalist 143:1007-1025.

Miller, T. E. and J. M. Kneitel. 2005. Inquiline communities in pitcher plants as a prototypical metacommunity. Pages 122-145 in M. Holyoak, M. Leibold, and R. D. Holt, editors. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago.

Miller, T. E. X. 2007. Does having multiple partners weaken the benefits of facultative mutualism? A test with cacti and cactus-tending ants. Oikos 116:500-512.

Morris, R. J., O. T. Lewis, and H. C. J. Godfray. 2004. Experimental evidence for apparent competition in a tropical forest food web. Nature 428:310-313.

Mouquet, N. and Loreau, M. 2003. Community patterns in source-sink metacommunities. American Naturalist 162: 544-557.

Myers, R. L. and J. J. Ewel, editors. 1990. Ecosystems of Florida. University of Central Florida Press, Orlando.

76

Nakajima, H. and M. Higashi. 1995. Indirect effects in ecological interaction networks II. The conjugate variable approach. Mathematical Biosciences 130:129-150.

Newman, M. E. J. and M. Girvan. 2004. Finding and evaluating community structure in networks. Physical Review E 69.

Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, B. O'Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2011. vegan: Community Ecology Package. Version 2.0-2.

Okuyama, T. and J. N. Holland. 2008. Network structural properties mediate the stability of mutualistic communities. Ecology Letters 11:208-216.

Outcault, L. W. and R. M. Scheffield. 1996. The longleaf pine forest: trends and current conditions. Resource Bulletin SRS-9. USDA Forest Service, Southern Research Station, Asheville.

Paine, R. T. 1966. Food web complexity and species diversity. American Naturalist 100:65-&.

Pinheiro, J., D. Bates, S. DebRoy, D. Sarkar, and the R Development Core Team. 2012. nlme: Linear and nonlinear mixed effects models v3.1-103.

Polis, G. A., W. B. Anderson, and R. D. Holt. 1997. Toward an integration of landscape and food web ecology: the dynamics of spatially subsidized food webs. Annual Review of Ecology and Systematics 28:289-316.

R Development Core Team. 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Rand, T. A., J. M. Tylianakis, and T. Tscharntke. 2006. Spillover edge effects: the dispersal of agriculturally subsidized natural enemies into adjacent natural habitats. Ecology Letters 9:603-614.

Ricketts, T. H. 2001. The matrix matters: Effective isolation in fragmented landscapes. American Naturalist 158:87-99.

Rodriguez-Girones, M. A. and L. Santamaria. 2006. A new algorithm to calculate the nestedness temperature of presence-absence matrices. Journal of Biogeography 33:924-935.

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

Rousk, J., E. Baath, P. C. Brookes, C. L. Lauber, C. Lozupone, J. G. Caporaso, R. Knight, and N. Fierer. 2010. Soil bacterial and fungal communities across a pH gradient in an arable soil. Isme Journal 4:1340-1351.

77

Schemske, D. W. and C. C. Horvitz. 1984. Variation among floral visitors in pollination ability: a precondition for mutualism specialization. Science 225:519-521.

Soetaert, K. and P. M. J. Herman. 2009. A Practical Guide to Ecological Modelling. Using R as a Simulation Platform. Springer-Verlag, New York.

Spiesman, B. J. and G. S. Cumming. 2008. Communities in context: the influences of multiscale environmental variation on local ant community structure. Landscape Ecology 23:313-325.

Steffan-Dewenter, I. and T. Tscharntke. 1999. Effects of habitat isolation on pollinator communities and seed set. Oecologia 121:432-440.

Stone, L. and A. Roberts. 1991. Conditions for a species to gain advantage from the presence of competitors. Ecology 72:1964-1972.

Stouffer, D. B. and J. Bascompte. 2011. Compartmentalization increases food-web persistence. Proceedings of the National Academy of Sciences of the United States of America 108:3648- 3652.

Strickland, M. S., C. Lauber, N. Fierer, and M. A. Bradford. 2009. Testing the functional significance of microbial community composition. Ecology 90:441-451.

Stys, B., R. Kautz, D. Reed, C. Keller, and A. Davis. 2004. Florida vegetation and land cover - 2003. Florida Fish and Wildlife Conservation Commission, Tallahassee.

Thébault, E. and C. Fontaine. 2010. Stability of ecological communities and the architecture of mutualistic and trophic networks. Science 329:853-856.

Thies, C., I. Steffan-Dewenter, and T. Tscharntke. 2003. Effects of landscape context on herbivory and parasitism at different spatial scales. Oikos 101:18-25.

Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371:65-66.

Vandermeer, J. and R. Carvajal. 2001. Metapopulation dynamics and the quality of the matrix. American Naturalist 158:211-220. Weiner, C. N., M. Werner, K. E. Linsenmair, and N. Blüthgen. 2011. Land use intensity in grasslands: Changes in biodiversity, species composition and specialisation in flower visitor networks. Basic and Applied Ecology 12:292-299.

Wootton, J. T. 1994. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology and Systematics 25:443-466.

Yeaton, R. I. and W. J. Bond. 1991. Competition between two shrub species: dispersal differences and fire promote coexistence. American Naturalist 138:328-341.

78

Yodzis, P. 1988. Indeterminacy of ecological interactions as perceived through perturbation experiments. Ecology 69:508-515.

Yodzis, P. 2000. Diffuse effects in food webs. Ecology 81:261-266.

79

BIOGRAPHICAL SKETCH

EDUCATION Ph.D. Dept. of Biological Science, Florida State University, 2012. Advisor: Brian D. Inouye M.S. Dept. of Wildlife Ecology & Conservation, University of Florida, 2006. Advisor: Graeme S. Cumming B.S. Dept. of Biology, Portland State University, 1999.

TEACHING EXPERIENCE Florida State University, Department of Biological Science, Tallahassee, FL Conservation Biology, Introductory Biology Lab for Majors. Animal Diversity Lab for Majors, General Ecology Lecture TA, Biogeography

Center for Agriculture, Science, and Environmental Education, Battle Ground, WA High school environmental education (grades 9-12)

SERVICE Peer review: Ecology, Ecology Letters, Frontiers in Ecology and the Environment, Global Ecological Change, Journal of Biogeography

Undergraduate Mentoring: Sophie Hyson (Barnard College). 2010. Predator effects on pollinator visitation across a gradient of visitor abundance and diversity. Cynthia Vasquez (Florida State University). 2008. Landscape structure effects on Puerto Rican ant populations.

AWARDS 2011 National Science Foundation Doctoral Dissertation Improvement Grant ($14,550) 2011 Florida State University Department of Biological Science Travel Grant ($800) 2010 Robert K. Godfrey Endowment Award for the Study of Botany ($1000) 2010 Florida State University Department of Biological Science Travel Grant ($550) 2010 Florida State University Council of Graduate Studies Travel Grant ($300) 2009 Florida State University Graduate School Dissertation Research Grant ($750) 2008 Florida State University COFRS ($13,000) 2007 Florida State University Department of Biological Science Travel Grant ($800) 2007 Florida State University Council of Graduate Studies Travel Grant ($300) 2006 University of Florida Graduate Student Council Travel Grant ($250) 2006 University of Florida Wildlife Ecology and Conservation Travel Grant ($170) 2005 Tropical Conservation and Development Field Research Grant ($500)

80

PUBLICATIONS Spiesman, B. J. and B. D. Inouye. (in revision) Habitat loss alters the architecture of plant- pollinator interaction networks.

Spiesman, B. J. and Cumming, G. S. 2008. Communities in context: The influences of multiscale environmental variation on local ant community structure. Landscape Ecology 23: 313-325.

Cumming, G. S. and Spiesman, B. J. 2006. Regional problems need integrated solutions: pest management and conservation biology in agroecosystems. Biological Conservation 131(4): 533- 543.

PRESENTATIONS Spiesman, B. J. and B. D. Inouye. 2012. Matrix quality mediates the effects of patch size and arrangement on metacommunity structure and ecosystem function. Ecological Society of America 97th Annual Meeting. (Oral)

Kim, T.N., Spiesman, B.J., Buchanan, A.L., et al. 2012. Selective removal of insect herbivores from one plant species influences an old-field plant community. Ecological Society of America 97th Annual Meeting. (Poster)

Spiesman, B. J. and B. D. Inouye. 2011. Habitat loss reduces nestedness and increases modularity in plant-pollinator interaction networks. European Ecological Federation 12th Annual Meeting (Poster)

Spiesman, B. J. and B. D. Inouye. 2011. Effects of landscape context on plant-pollinator interaction networks. Ecological Society of America 96th Annual Meeting. (Oral)

Spiesman, B. J. and B. D. Inouye. 2010. Resource- and predator-mediated apparent competition in a mutualistic interaction web. Ecological Society of America 95th Annual Meeting. (Poster)

Spiesman, B. J. and B. D. Inouye. 2009. Indirect interactions in a simple mutualistic interaction web. Southeastern Evolution and Ecology Conference (Poster)

Spiesman, B. J. and G. S. Cumming. 2007. Multiscale influences on local ant community structure. Ecological Society of America 92nd Annual Meeting. (Poster)

Spiesman, B. J. 2006. Ant community response to variation in plant community and landscape structure in the northern karstic region of Puerto Rico. University of Florida Tropical Conservation and Development. (Poster)

81