Biodiversity of Ants and Associated Mites in Constructed Grasslands at Multiple Spatial Scales

MIAMI UNIVERSITY The Graduate School

Certificate for Approving the Dissertation

We hereby approve the Dissertation

Kaitlin U. Campbell

Candidate for the Degree

DOCTOR OF PHILOSOPHY

______Thomas O. Crist, Director

______Melany C. Fisk, Reader

______Hans Klompen, Reader

______Ann L. Rypstra, Reader

______John K. Maingi, Graduate School Representative

ABSTRACT

BIODIVERSITY OF ANTS AND ASSOCIATED MITES IN CONSTRUCTED GRASSLANDS AT MULTIPLE SPATIAL SCALES

by Kaitlin U. Campbell

The goals of this dissertation were to examine how patch and landscape level processes structure ant communities in agricultural landscapes and determine the relative roles of patch and host-level factors in determining ant-associated mite diversity and community composition. In Chapter 1, I examined ant richness, species frequencies, and community composition in 23 warm season constructed grasslands that varied in both patch and landscape level characteristics. Ant species richness was greater in older sites with sandier soils, while community composition was structured by soil texture, management, and urban land use. Frequency analyses for 14 ant species showed a wide range of responses to both patch and landscape components of the environment including age, management, soil texture, and surrounding land use. My findings support the use of ants as environmental indicators of disturbance in agricultural landscapes and show that diversity in constructed grasslands is structured by both patch and landscape level processes. In Chapter 2, I determined the relative importance of host and habitat for an ant-dependent commensalism (phoretic mites). I found that large, cosmopolitan, and abundant ant species support a greater proportion of the mite diversity. Additionally, I found that patch level characteristics, or environmental context (area, age, soil texture, and litter depth), of the host can alter the associated mite diversity. In Chapter 3, predictions of theory, herbivore resource hypotheses, and spatial parasitology were used to identify the extent of the ecological neighborhood for mites associated with ants. My results indicate that commensal mite communities are consistent with the Resource Size Hypothesis and are sensitive to ecological neighborhoods at multiple hierarchical levels including individual host ants, the host ant colony, surrounding nest community, and habitat type, but do not vary significantly among sites. In the final chapter, I examined the importance of spatial arrangement of ant nests for mite dispersal among nests in a homogenous environment and the role of seasonal synchrony with hosts. I found significant spatial autocorrelation for mite communities at the closest distance class and evidence of increased mite abundance and richness during periods of ant colony reproduction. Together, these studies demonstrate that processes at multiple temporal and spatial scales contribute to biodiversity and community assembly within conservation habitatsand that the context of ant hosts can modify their roles as biodiversity regulators.

BIODIVERSITY OF ANTS AND ASSOCIATED MITES IN CONSTRUCTED GRASSLANDS AT MULTIPLE SPATIAL SCALES

A DISSERTATION

Presented to the Faculty of

Miami University in partial

fulfillment of the requirements

for the degree of

Doctor of Philosophy

Department of Biology

Kaitlin U. Campbell

The Graduate School Miami University Oxford, Ohio

2015

Dissertation Director: Thomas O. Crist

Kaitlin U. Campbell

2015

TABLE OF CONTENTS

Table of Contents General Introduction ...... 1 1 References ...... 5 Chapter 1: Differential responses of grassland ant species at patch and landscape levels ...... 8 1 Abstract ...... 8 2 Introduction ...... 9 3 Methods ...... 13 3.1 Study Sites ...... 13 3.2 Ant Community Sampling ...... 13 3.3 Vegetation Sampling ...... 14 3.4 Soil Analysis ...... 14 3.5 Additional Patch-level Variables ...... 15 3.6 Landscape Analysis ...... 15 3.7 Statistical analysis ...... 16 4 Results ...... 17 4.1 Ant Species Composition ...... 19 5 Discussion ...... 20 5.1 Relative Role of Patch and Landscape-level Predictors ...... 20 5.2 Anthropogenic and Uncommon Specialist Responses ...... 20 5.3 Landscape-level Responses ...... 21 5.4 Management Practices ...... 22 5.5 Soil Characteristics...... 22 5.6 Vegetation ...... 23 5.7 Conclusions ...... 23 6 References ...... 25 Chapter 2: Host traits and environmental context structure ant-dependent mite communities 39 1 Abstract ...... 39 2 Introduction ...... 39 3 Methods ...... 43 3.1 Study Sites ...... 43 3.2 Ant and Mite Collections ...... 44 3.3 Environmental variables ...... 45

iii

3.4 Statistical Analysis ...... 45 3.5 Community Composition Analyses ...... 47 4 Results ...... 47 4.1 Overall Mite Species Richness and Abundance ...... 48 4.2 Richness and Abundance of Heterostigmata and Astigmata ...... 48 4.3 Mites on Common Host Species ...... 49 4.4 Mite Community Composition ...... 49 5 Discussion ...... 50 5.1 Host Identity vs. Environment ...... 50 5.2 Host Suitability, Frequency, and Richness ...... 50 5.3 Disturbance, Soil Resources, and Patch Characteristics ...... 51 5.4 Conclusions ...... 52 6 References ...... 53 Chapter 3: Merging spatial parasitology and herbivore theory to identify ecology relevant spatial scales for commensal organisms ...... 74 1 Abstract ...... 74 2 Introduction ...... 75 3 Methods ...... 78 3.1 Nest Neighborhood Study ...... 78 3.2 Nest Habitat Study ...... 81 4 Results ...... 82 4.1 Nest Neighborhood Study ...... 82 4.1.1 Tests of Resource Hypotheses ...... 82 4.1.2 Among Site Variation ...... 83 4.2 Nest Habitat Study ...... 83 5 Discussion ...... 84 5.1 Conclusion ...... 86 6 References ...... 87 Chapter 4: Phoresy across space and time: commensal mite communities are structured by distance-decay and synchrony with hosts ...... 99 1 Abstract ...... 99 2 Introduction ...... 99 3 Methods ...... 102 3.1 Study System ...... 102 iv

3.2 Nest Sampling ...... 102 3.3 Statistical Analysis ...... 103 4 Results ...... 104 4.1 Spatial Autocorrelation of Mite Communities ...... 104 4.2 Seasonal Shifts ...... 104 5 Discussion ...... 105 6 References ...... 108 General Conclusions ...... 117 1 References ...... 120

LIST OF TABLES

CHAPTER 1 TABLES Table 1. Best and competing models for ant richness in 2011 and 2012...... 29 Table 2. Best and competing models for ant species. Habitat Association indicates anthropogenically favored species, uncommon specialists, and cosmopolitan species found in open and/or closed habitats ( LU= Land use)...... Error! Bookmark not defined.

Supplementary Table 1. Descriptive table of patch-level variables. Divisions represent general site characteristics, disturbance, plant, and soil variables...... 35 Supplementary Table 2. Descriptive table of landscape-level variables (LU=Land use)...... 36 Supplementary Table 3. Ant richness and abundance results for 2011, 2012 and combined. Dashes indicate no data (sites that were sampled in only one of the years)...... 37 Supplementary Table 4. Ant species found across the 23 sites, abundance, and number of sites...... 38

CHAPTER 2 TABLES Table 1. Best and competing models (ΔAIC<2) for Total Mite Richness and Abundance and Richness and Abundance for Astigmata and Heterostigmata groups...... 58 Table 2. Best and competing models (ΔAIC<2) for Prevalence, Richness, Abundance, and Load for mites associated with Aphaenogaster rudis and Myrmica americana...... 59

Supplementary Table 1. Abundance and richness of associated mites for ant species collected with baits. The horizontal division separates ant species with and without mites...... 69 Supplementary Table 2. Mite richness and abundance for 2011, 2012 and combined. Dashes indicate no data (sites that were sampled in only one of the years)...... 70 Supplementary Table 3. Mite species collected at the 23 grassland sites...... 71

CHAPTER 3 TABLES Table 1. Definitions and ant-associated mite spatial examples (host and superorganism levels) of parasite population and community ecology terminology used throughout this Chapter (modified from Guégan et al., 2005)...... 90 Table 2. Best and competing models (ΔAIC<2) for mite richness and abundance for Myrmica americana and Aphaenogaster rudis nests...... 93

LIST OF FIGURES

GENERAL INTRODUCTION FIGURES Figure 1. Species assembly of ants and associated mites is a product of landscape, patch and host level processes (outer divisions). Ants communities are structured by landscape and patch level characteristics (red arrows) while mite communities are structured by variation in patch and host level predictors (blue arrows). Chapters addressing the components of the conceptual diagram are shown in circles along each arrow...... 7

CHAPTER 1 FIGURES Figure 1. Best models for ant species richness in (a) 2011 included time since planting (Grassland age) and (b) 2012 included percentage of clay fraction of the soil (X and Y axes log-scaled)...... 30 Figure 2. Best fitting models for (a) an anthropogenically favored species (Tapinoma sessile), (b) an uncommon specialist species (Temnothorax ambiguus), (c) a cosmopolitan open habitat species (Myrmica americana), and a (d) Cosmopolitan open /closed habitat species (Ponera pennsylvanica)...... 31 Figure 3. Multivariate ordination for ant community composition across 23 grasslands. Symbols are site scores for individual grasslands scaled to the percentage value of the most important predictor variable (Urban Land Use). Arrows are biplot correlations of significant predictor variables (Urban Land Use, Time since burn and Percent Sand) (R2=0.251, p=0.003)...... 33

Supplementary Figure 1. Locations of 23 sites in Butler, Preble, and Montgomery Counties. Surrounding landscapes of four sites are shown (numbered circles)...... 34 Supplementary Figure 2. Diagram of sampling methods. Pitfall traps and vegetation quadrats were spaced 25m apart along the transect. Paired vegetation quadrats were measured on either side of the pitfall traps adjacent to the transect and averaged...... 36

CHAPTER 2 FIGURES Figure 1. Species accumulation curve of mite species by number of hosts inspected. Number of points on a curve represents the number of sites where the ant host (species name in italics) was collected, while length of the curve represents the number of ant individuals inspected. Ant species that are more cosmopolitan (at more sites) and more abundant also have higher observed (Obs) and estimated (Est) species richness...... 56 Figure 2. Best fitting model for mite species richness associated with seventeen different ant species included host size and host abundance. Note: X axis is shown on log-scale...... 57 Figure 3. Best models for overall mite species richness for a) 2011- ant richness and b) 2012- grassland area (X axis log-scaled)...... 60 Figure 4. The best model for phoretic mite abundance included ant richness. Y axis is presented on a log scale...... 61 Figure 5. Best fitting model for 2011 Astigmata (solid line, filled circles) and 2012 Heterostigmata richness (dashed line, open circles) included Grassland Area (ha)...... 62 Figure 6. Best fitting model for 2011 Astigmata (solid line, filled circles) and Heterostigmata abundance (dashed line, open circles) included Ant Richness. Y axis is presented on a log scale...... 63 Figure 7. Best fitting models for a) mite richness and b) abundance (log scale) associated with Aphaenogaster rudis included Litter Depth...... 64 Figure 8. Best fitting models for a) mite richness and b) abundance (log scale) associated with Myrmica americana included both Host Frequency and Grassland Age. Curve in plot represents model fit when Grassland Age is held constant as the mean value...... 65

vii

Figure 9. Best fitting models for 2012 Average Mite Load (Trueload) on Myrmica americana included Host Frequency. Trueload is calculated as the number of mites per bait divided by the number of hosts with mites at the bait...... 66 Figure 10. Multidimensional scaling (MDS) ordination of mite community composition. Host was a significant predictor of mite community (p=0.001, Variance Expl. = 31.8%). In addition to separating out by species, mite communities also tended to cluster between ants in the same genus (e.g. Myrmica americana and Myrmica latifrons, Lasius neoniger and Lasius alienus)...... 67 Figure 11. AIC selected Distance-based Redundancy Analysis (dbRDA) multivariate ordination of mite community composition across 23 grasslands. Symbols indicate site scores of grasslands and are sized according to the most important predictor variable (Ant Richness). Arrows are biplot correlations of the significant predictor variables...... 68

Supplementary Figure 1. Diagram of sampling methods. Pitfall traps and vegetation quadrats were spaced 25m apart along the transect. Paired vegetation quadrats were measured on either side of the pitfall traps adjacent to the transect and averaged. Bait stations were place 8.3m from each pitfall trap...... 73 Supplementary Figure 2. Mantel Test of Dissimilarity showed a correlation between Mite species Dissimilarity (Bray-Curtis) and Ant species Dissimilarity (Mantel r= 0.51, p=0.001) indicating that turnover in ant community composition is mirrored by turnover in mite community composition. ... 73

CHAPTER 3 FIGURES Figure 1. Methods for (a) Nest Neighborhood Study showing an example of a surrounding nest community for a Myrmica americana focal nest. Focal nest (FN-red triangle) is shown at center and encircled by 1.8m radius area. Surrounding nests (SN-squares) are color coded by ant species. SNsame= 3 (all red squares), SNrichness= 5 (number of nests of different colors), SNabundance= 8 (number of squares), SNlarge= 6 (number of nests belonging to large ant species- blue, red, and orange) and (b)Nest Habitat Study showing G5 site with 15 nests inside the grassland (orange triangles) and 15 nests in the mowed lawn bordering the grassland (light blue trianges)...... 91 Figure 2. Species accumulation curves for mites associated with focal nests of a) Myrmica americana (blue) and b) Aphaenogaster rudis (green)...... 92 Figure 3. Non-metric Multidimensional Scaling ordination plots for the species composition of ant nests surrounding focal nests of a) Myrmica americana and b) Aphaenogaster rudis. Ant species names are shown in blue and green for the two different focal nest species respectively. Numbers represent individual focal nests scaled to the abundance of mites within the focal nest...... 93 Figure 4. The best model for mite abundance in Aphaenogaster rudis focal nests included the NMDS Axis 1 Score (34.31% Deviance explained, p=0.005). Note: Y axis is log-scaled...... 94 Figure 5. Multidimensional Scaling ordination plots showing site as a predictor of surrounding ant nest community composition for (a) Myrmica americana and (b) Aphaenogaster rudis focal nests. Each colored point represents a focal nest at one of 5 sites (Ehr, FR, TC, G1 and G6). 95% confidence intervals are shown for sites (black ellipses) with at least three focal nests...... 95 Figure 6. Multidimensional Scaling ordination plots showing site as a predictor of mite community composition in (a) Myrmica americana and (b) Aphaenogaster rudis focal nests. Each colored point represents a focal nest at one of 5 sites (Ehr, FR, TC, G1 and G6). 95% confidence intervals are shown for sites (black ellipses) with at least three focal nests...... 96 Figure 7. Boxplots showing habitat related shifts in abundance for mites collected from Myrmica americana nests in the grassland (dark gray) and in the lawn (light gray). Total mite abundance (a) was similar between habitats, but (b) Heterostigmata were significantly (designated with: *) higher in the grassland, and (c) Astigmata were significantly higher in the lawn. Note: Y axis is log scaled for a) and c)...... 97 Figure 8. Multidimensional Scaling ordination plots of showing (a) habitat (grassland-pink, lawn- blue) and (b) site (G1, G5, G6) as a predictors of mite community composition in Myrmica americana nests. 95% confidence intervals are shown for habitat and sites (black ellipses)...... 98

viii

CHAPTER 4 FIGURES

Figure 1. Species accumulation curve for mite richness in a) Myrmica americana nests and b) Aphaenogaster rudis nests. Chao estimated richness is indicated by the dotted line, and observed richness is shown with a dashed line...... 111 Figure 2. Mite composition in Myrmica americana nests as determined by nest proximity. (a) Mantel test of mite community dissimilarity by Euclidean distance dissimilarity, and (b) Mantel correlogram of mite species composition by distance class. The horizontal red line indicates the break between positive and negative correlations. The diamond point represents a significant relationship between mite community and distance at the closest distance class...... 112 Figure 3. Moran’s I correlograms showing tests of spatial autocorrelations of (a) mite abundance and (b) richness in Myrmica americana nests. The horizontal red line indicates the break between positive and negative correlations. Tests at all distance classes were not significant...... 113 Figure 4. Seasonal abundance and richness of mites associated with Myrmica americana nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 6 and 7)...... 114 Figure 5. Seasonal abundance and richness of (a) specialist and (b) generalist mites associated with Myrmica americana nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 6 and 7)...... 115 Figure 6. Seasonal abundance and richness of mites associated with Aphaenogaster rudis nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 2 and 3)...... 116

Supplementary Figure 1. Distance classes for nest spatial sampling methods. Example of a nest (blue triangle) and relative comparisons are shown for each of 6 distance classes (colored circles)...... 110

GENERAL CONCLUSION FIGURES Figure 1. Conceptual diagram summarizing important landscape, patch and host level (outer divisions) variables for measures of mite diversity (blue text and arrows), ant diversity (red text and arrows), and both taxa (purple text). Numbered circles represent chapters testing the relationship...... 121

DEDICATION To my family: Gabe Campbell Kevin Uppstrom Jana Morse Jennifer Alford & Ariel Uppstrom

ACKNOWLEDGEMENTS

First and foremost I would like to thank my advisor, Dr. Tom Crist, for accepting me into his lab and allowing me the freedom to work on what interested me most. He has been endlessly patient, continuously supportive, and has shown trust and confidence in me. His subtle guidance has fostered a higher order of thinking that has resulted in a deeper understanding of the “big picture” surrounding such little creatures. I thank him for always treating me not just as an advisor would a student, but as a colleague. I would also like to thank Dr. Hans Klompen for changing the direction of my life when he showed me the diversity of mites and the wonder of tiny things as an undergraduate at Ohio State University. I also thank him for his continued commitment to my success as advisor for my Master’s degree and as committee member for my PhD. I also would like to thank my other committee members Dr. John Maingi for his help and training in GIS, Dr. Ann Rypstra for her big ecological questions and for recommending me to work with the REU program, Dr. Melany Fisk for her knowledge of soil ecology, assistance with soil processing, and valuable comments on this dissertation. To all the undergraduates who helped me throughout my PhD by processing lab samples, collecting field data, and helping with knee-breaking back-warping searches for ant nests: I thank you deeply. Without you I would have been at Miami several more years to accomplish what I did. These field and lab assistants in order of appearance are: Kelsey Seaman, Tia Loyke, Garrett Dienno, Mayrolin García, Sam Stephenson, Aaron Coleman, Anita Schaefer, Carol Ramos, Natalie Konig, and Amanda McDonald. I would like to thank members of the Crist Lab past and present: Alyssa Whu- thanks for being a great welcoming committee, Jason Nelson for Björk and for keeping things interesting; Mike Minnick (“Prime”) for being so supportive at all times, for being a great listener and excellent companion at conferences, for the company on writing Wednesday, and help with making statistics make sense; Mike Mahon (“Squared”) for the laughs, for getting my jokes and references I assume no one will get, and for taking over that honeysuckle project; Valerie Peters for career and statistical advice, for being a patient collaborator, and for being a gentle listener and role model. I would also like to thank the numerous graduate students at Miami University who have been steadfast friends and teammates. I want to especially thank Melissa Youngquist for sharing the good and bad times with me in our closet-like office for the last 5 years. Lastly and most importantly I want to thank my family, Jana, Kevin, Jen, and Ariel, for their continuous support and for never losing hope that one day I might get a “real job” after working on degrees for 12 years. Thank you to my grandparents, who supported me throughout all of my work and continue to be interested in my research and always willing to hear more about it. To Gabe- my husband, best friend, and rock- thanks for keeping me grounded during the most difficult times and for understanding and being willing to share me with my research.

General Introduction

Species assembly processes structure biodiversity, distributions, and species interactions and depend on both spatial and temporal scales. Agricultural landscapes in North America are a legacy of agricultural intensification and urbanization from the last century which resulted in structurally simple, high-input systems with low species diversity (Tscharntke et al., 2005). Imbedded in human-dominated landscapes, are isolated patches of natural remnants and semi-natural constructed habitats that act as sources and refuges of biodiversity (Kleijn & Sutherland, 2003; Knop et al., 2006). Incentive programs, such as the agri-environmental schemes in Europe and the Conservation Reserve Program in the US, have facilitated construction of conservation grasslands and other semi-natural habitats on cropland formerly in production. These incentives are aimed at reducing the environmental impacts of farming such as erosion, runoff, poor water quality, and loss of biodiversity (Baer et al., 2000; Kleijn & Sutherland, 2003; Millenbah et al., 1996). Animal and plant communities in constructed grasslands are sensitive to variation in habitat characteristics within grassland patches as well as the surrounding landscape for colonizing propagules and local processes for establishment and recruitment over time. These grassland systems offer unique opportunities for understanding community assembly through time at multiple spatial scales. The species assemblage within a grassland may undergo shifts in species interactions over time due to newly colonizing species and resulting changes in their relative abundances. Multiple studies have shown shifts in species richness and community composition following recovery from disturbance (Andersen & Majer, 2004; Brand & Dunn, 1998; Phipps, 2006). The gradual addition of new species can promote functional redundancy and increased interactions that promote stability and resilience in communities (Cornell & Lawton, 1992; Duelli et al., 1999). For example, plants dependent on specific pollinators or seed dispersers may not successfully establish until mutualists have also colonized the habitat patch. The context of these interactions may

also change as the site develops due to changes in abiotic conditions or relative levels of disturbance. Invertebrate taxa, especially ground beetles and ants, are often used to evaluate success of restoration or quality of managed habitat because they are easily collected, widespread, highly diverse, inexpensive to measure, and sensitive to change (Rainio & Niemela, 2003; Underwood & Fisher, 2006). Ants especially are used as environmental indicators because they comprise a large portion of the animal biomass in temperate and tropical environments (Hölldobler & Wilson, 1990), are slow to colonize, create long- lived colonies, and are significant regulators of above and belowground food webs (Laakso & Setälä, 1998). For example, recent studies on whole-plot removals of ants show them to be significant regulators of belowground carbon, nitrogen, soil mesofauna and microbes, and aboveground plant biomass, seedling growth, and seed dispersal (Wardle et al., 2011; Zelikova et al., 2011). Determining the factors affecting the diversity and composition of influential regulating species, such as ants, can provide us with key information on not only the ants, but also habitat quality and other ant- dependent organisms in a given system. Ants are functionally important organisms in many communities because they interact with other organisms at multiple trophic levels, for example, as top predators, herbivores, seed dispersers, and mutualists (Eickwort, 1990; Elmes et al., 1998; O’Dowd & Hay, 1980). Ants building their nests in the soil have highly regulated microclimates and abundant resources, which indirectly affect the soil, forming and stable microhabitats for a diversity of myrmecophiles, organisms living in association with ants. The most abundant of the arthropod myrmecophiles, are mites (Kistner, 1979; Rettenmeyer, 1962) which use ant nests for resources (microbes, refuse, and decaying material) and the ants themselves for dispersal, a commensal relationship known as phoresy. The purpose of phoresy by mites is believed to be dispersal or transportation to isolated resources (Houck & OConnor, 1991). Over 600 mite species have been documented in association with ants (Uppstrom, 2010) and hundreds more remain unidentified, undescribed, or undiscovered (Campbell et al., 2013; Rettenmeyer et al., 2011). The ant-dependent mite commensalism is potentially a useful model system for understanding community assembly and shifts in species interactions during the

development of reconstructed ecosystems, such as conservation grasslands. This system can be hierarchically organized into landscape, patch, and host level processes that can structure ant and mite communities and provide context for the interaction (Figure 1). Using this framework, ant species colonize the grassland patches from the surrounding landscape (Landscape Level, Figure 1). The propagules of colonizing ant species may be influenced by the surrounding land use and land cover. For example, a grassland surrounded by urbanization or agriculture, may receive different propagules of foundress ants than a grassland surrounded by natural or semi-natural land use (i.e. forest, grassland). Once established, the persistence and subsequent proliferation of an ant colony might be determined by biotic and abiotic characteristics within a discrete grassland patch, potentially including disturbance, vegetation, soil, area, and shape of the patch (Patch Level, Figure 1). Mites colonize the grasslands through phoresy, on the bodies of suitable ant hosts, and the community of mites that develops within a given colony or among colonies at the patch level could be driven in part by the same or additional patch-level characteristics that structure their hosts (Patch Level, Figure1). The abundance, species richness, and community composition of mites might also be influenced by frequency of preferred hosts, host life-history traits (Figure 1), inter- and intraspecific interactions among neighboring colonies (Nest Neighborhood, Nest Proximity, Figure 1), and seasonal shifts in ant colonies throughout the season (Nest Seasonality, Figure 1). The abiotic and biotic context of the host nest may be an influential component altering the development of phoretic mite communities. The objectives of this dissertation were to examine how patch and landscape level processes structure ant communities and determine the relative roles of patch and host- level factors in determining ant-associated mite diversity and community composition. I investigated landscape and patch-level effects on ant communities in conservation grasslands in Chapter 1 using a chronosequence of 23 constructed prairie grasslands varying in patch level characteristics and surrounding land use and land cover. In Chapter 2, I used the same 23 grasslands to identify how ant species traits determine host suitability and associated mite diversity and how the environmental context (patch level characteristics) of the hosts can alter the ant-mite commensalism. In Chapter 3, I merged herbivore resource theory and concepts from spatial parasitology to assess the responses

of phoretic mite to local ecological neighborhoods surrounding host ant nests, and at multiple spatial scales and across habitat types. Using key host ant species, I determined if mite communities are sensitive to neighboring ant nests and if mite communities vary across habitat type and among sites. In Chapter 4, I examined how spatial distribution of ant nests and seasonal changes in dispersal opportunities affect phoretic mite community composition, abundance and richness. Finally I described how, together, these studies demonstrate that processes at multiple temporal and spatial scales contribute to biodiversity and community assembly within conservation patches and how the context of ant hosts can modify their roles as regulators of biodiversity.

1 References Andersen, A. N., & Majer, J. D. (2004). Ants show the way down under: invertebrates as bioindicators in land management. Frontiers in Ecology and the Environment, 2, 291–298. Baer, S. G., Rice, C. W., & Blair, J. M. (2000). Assessment of soil quality in fields with short and long term enrollment in the CRP. Journal of Soil and Water Conservation. Brand, R. H., & Dunn, C. P. (1998). Diversity and abundance of springtails (Insecta: Collembola) in native and restored tallgrass prairies. American Midland Naturalist, 139(2), 235–242. Campbell, K. U., Klompen, H., & Crist, T. O. (2013). The diversity and host specificity of mites associated with ants: the roles of ecological and life history traits of ant hosts. Insectes Sociaux, 60(1), 31–41. Cornell, H. V., & Lawton, J. H. (1992). Species interactions , local and regional processes , and limits to the richness of ecological communities : a theoretical perspective. Journal of Animal Ecology, 61(1), 1–12. Duelli, P., Obrist, M. K., & Schmatz, D. R. (1999). Biodiversity evaluation in agricultural landscapes: above-ground insects. Agriculture, Ecosystems & Environment, 74(1- 3), 33–64. Eickwort, G. C. (1990). Associations of mites with social insects. Annual Review of Entomology, 35, 469–488. Elmes, G. W., Thomas, J. A., Wardlaw, J. C., Hochberg, M. E., Clarke, R. T., & Simcox, D. J. (1998). The ecology of Myrmica ants in relations ot the conservation of Maculinea butterflies. Journal of Insect Conservation, 2, 67–78. Hölldobler, B., & Wilson, E. O. (1990). The Ants. Cambridge: Harvard Press. Houck, M. A., & OConnor, B. M. (1991). Ecological and evolutionary significance of phoresy in the Astigmata. Annual Review of Entomology, 36, 611–636. Kistner, D. H. (1979). Social and evolutionary significance of social insect symbionts. In H. R. Hermann (Ed.), Social insects (Vol. 1, pp. 340–413). New York: Academic Press. Kleijn, D., & Sutherland, W. J. (2003). How effective are European agri-environment schemes in conserving and promoting biodiversity? Journal of Applied Ecology, 40(6), 947–969. Knop, E., Kleijn, D., Herzog, F., & Schmid, B. (2006). Effectiveness of the Swiss agri- environment scheme in promoting biodiversity. Journal of Applied Ecology, 43(1), 120–127. Laakso, J., & Setälä, H. (1998). Composition and trophic structure of detrital food web in ant nest mounds of Formica aquilonia and in the surrounding forest soil. Oikos, 81, 266–278. Millenbah, K. F., Winterstein, S. R., Campa III, H., Furrow, L. T., & Minnis, R. B. (1996). Effects of conservation reserve program field age on avian relative abundance, diversity, and productivity. The Wilson Bulletin, 108(4), 760–770. O’Dowd, D. J., & Hay, M. E. (1980). Multalism between Harvester Ants and a Desert Ephemeral : Seed Escape from Rodents. Ecology, 61(3), 531–540. Phipps, S. J. (2006). Biodiversity of ants (Hymenoptera: Formicidae) in restored grasslands of different ages. University of Missouri-Columbia.

Rainio, J., & Niemela, J. (2003). Ground beetles ( Coleoptera : Carabidae ) as bioindicators, (McGeoch 1998), 487–506. Rettenmeyer, C. W. (1962). Arthropods associated with neotropical army ants with a review of the behavior of these ants (Arthropoda; Formicidae: Dorylinae). University of Kansas, Lawrence, KS. Rettenmeyer, C. W., Rettenmeyer, M. E., Joseph, J., & Berghoff, S. M. (2011). The largest animal association centered on one species: the army ant Eciton burchellii and its more than 300 associates. Insectes Sociaux, 58, 281–292. Tscharntke, T., Klein, A. M., Kruess, A., Steffan-Dewenter, I., & Thies, C. (2005). Landscape perspectives on agricultural intensification and biodiversity-ecosystem service management. Ecology Letters, 8, 857–874. Underwood, E. C., & Fisher, B. L. (2006). The role of ants in conservation monitoring: if, when, and how. Biological Conservation, 132(2), 166–182. Uppstrom, K. A. (2010) Mites (Acari) Associated with the Ants (Formicidae) of Ohio and the Harvester Ant, Messor pergandei, of Arizona. The Ohio State University. Wardle, D. a, Hyodo, F., Bardgett, R. D., Yeates, G. W., & Nilsson, M.-C. (2011). Long- term aboveground and belowground consequences of red wood ant exclusion in boreal forest. Ecology, 92(3), 645–56. Zelikova, T. J., Sanders, N. J., & Dunn, R. R. (2011). The mixed effects of experimental ant removal on seedling distribution, belowground invertebrates and soil nutrients. Ecosphere, 2(5), 1-14.

Figure 1. Species assembly of ants and associated mites is a product of landscape, patch and host level processes (outer divisions). Ants communities are structured by landscape and patch level characteristics (red arrows) while mite communities are structured by variation in patch and host level predictors (blue arrows). Chapters addressing the components of the conceptual diagram are shown in circles along each arrow.

Chapter 1: Differential responses of grassland ant species at patch and landscape levels

1 Abstract Agricultural expansion has converted much of the native tallgrass prairie and upland forest in the Midwest US to corn and soybean fields. Conservation groups and federal incentive programs restore remnant grasslands or construct warm-season grasslands on marginal agricultural land to reduce soil erosion, enhance wildlife habitat, and support biodiversity of native grassland flora and fauna. These set-asides are colonized over time by grassland dependent taxa from the surrounding landscape that may include intensive agriculture, urban, and semi-natural habitats. Ants are used as environmental indicators of disturbance and recovery for management-based monitoring of soils and invertebrate diversity because they tend to colonize slowly following soil disturbance and have long lived colonies. They interact at multiple trophic levels and structure above and belowground systems. The goals of this study were to differentiate patch and surrounding landscape effects on the grassland ant species composition and diversity in a chronosequence of constructed grasslands. We studied ant communities in 23 conservation grasslands that differed in area, time since planting (age), plant community, soils, management, and surrounding landscape. The best model for ant species richness was grassland age in 2011 and soil clay content in 2012. Ant community composition was determined by soil texture, surrounding urban land use, and time since burn. Species responded in different ways to patch variables and were also influenced by surrounding landscape. Uncommon specialist ant species were more abundant in patches with larger amounts of surrounding grassland and those that were older and larger, while anthropogenically favored (disturbance tolerant) species were more abundant in patches embedded within more intensive agriculture and at younger sites. Our results suggest that ant communities are primarily influenced by habitat age, but soil, management, and surrounding land use have differential effects on individual species frequencies. These findings support the use of ants as environmental indicators of ecosystem recovery following disturbance in agricultural landscapes..

2 Introduction Trends toward greater agricultural intensification over the last century have rapidly converted much of the natural and semi-natural land into agriculture (Tscharntke et al., 2005). Landscapes with large amounts of intensive agriculture tend to have high input crop systems with very low habitat diversity and structural simplicity, resulting in declines in biodiversity of flora and fauna (Hendrickx et al., 2007; McLaughlin & Mineau, 1995; Stoate et al., 2001). Private landowners, conservation organizations, community parks construct or restore habitats as conservation set-asides through incentive programs aimed to enhance ecosystem services, aesthetic qualities, or biodiversity within agriculturally dominated landscapes. For example, European agri- environmental schemes in which farmers receive compensation for modifying their farming practice to provide environmental benefits, have often reported increases in biodiversity across both plant and animal taxa (Kleijn & Sutherland, 2003; Knop et al., 2006). One of the most endangered ecosystems in North America is the tallgrass prairie that once spanned much of the Midwest, but has declined dramatically over the last century (estimates of 82-99%) due to agricultural conversion and urbanization (Samson & Knopf, 1994). The Conservation Reserve Program (CRP) was first introduced in 1985 by the United States Department of Agriculture (USDA) to reduce erosion of croplands and has contributed over 12 million ha of forage and native prairie grasslands by removing agricultural lands from production and replacing them with more natural habitats (Dunn et al., 1993). Studies in these constructed grasslands in North America and similar incentives in Europe demonstrate the recovery of organic matter and nutrients to the soil (Burke et al., 1995; McLauchlan et al., 2006) and increases in biodiversity of numerous grassland dependent plant and animal taxa (Martin et al., 2005; Panzer et al., 2010; Söderström et al., 2001; Tscharntke et al., 2002). Despite these established benefits of CRP and other set-asides, they may be returned to cultivation at a later time thereby limiting the long term benefits of conservation grasslands to biodiversity and ecosystem function in agricultural landscapes. Constructed grasslands harbor a wide variety of invertebrate taxa that have been used to evaluate success of agri-environmental schemes and CRP incentive programs

(Dunn et al., 1993; Kleijn & Sutherland, 2003; Millenbah et al., 1996; Peters et al., in review). Colonization and establishment of these invertebrates are affected by landscape level factors, such as the proximity of different habitats that may provide complementary resources for foraging, nesting, or overwintering (Fahrig et al., 2011). For highly vagile organisms, such as bees and butterflies, grasslands could be used periodically for floral resources, and the surrounding landscape might be used at other times for additional resources (Mandelik et al., 2012). For less vagile organisms, such as ants, the surrounding landscape is the source of colonizing propagules, and once established, there is a long term dependence on the grassland and little to no use of resources outside of the grassland patch. The tendency for long-lived colonies, slow colonization, and long term connection with habitats is the basis for their use as environmental indicators of disturbance and recovery (Andersen & Majer, 2004; Folgarait, 1998; Mitrovich et al., 2010; New, 2000; Underwood & Fisher, 2006). Additionally, ants are widely recognized as important to biodiversity and ecosystem functioning because they modify soils and plant communities, and have key roles in species interactions (Folgarait, 1998; Crist, 2008). As widespread organisms, ants can be exposed to an equally wide range of disturbances including prescribed fire, agricultural intensification, grazing, urbanization, and habitat fragmentation (Folgarait, 1998; Underwood & Fisher, 2006). Their sensitivity to disturbance and ecosystem change, as well as their significance to ecosystem functioning have increased their use in ecological studies of conservation and restoration ecology. Although ants may forage into cultivated fields from margins adjacent to forest or grassland, few species maintain populations in the corn and soybean fields of the Midwest, even under limited tillage regimes (House & Stinner, 1983; House, 1989; Peck et al., 1998; Stinner & House, 1990). Therefore, ants slowly colonize constructed grasslands on formerly cultivated fields from the surrounding landscape over several years. Most ant species disperse during mating flights as alates (winged sexual forms), and these foundress queens select appropriate habitats based on land use and abiotic conditions (Dauber et al., 2005; Folgarait, 1998; Gómez et al., 2003). Ant community assembly in constructed grasslands has been studied primarily through the use of

chronosequences, or sites that vary in age. For example, a study of 12 CRP grasslands showed ant species richness and abundance to be positively associated with the age of the grasslands, peaking in 7-8 year fields (Phipps, 2006). Similarly, a study of grasslands in Germany found nest density and species richness increased with time, and ant community composition and functional group dominance shifted to favor subterranean foragers (Dauber & Wolters, 2005). The surrounding landscape may be very important as both a source of colonizing propagules and in filtering the species composition of potential colonists. Few studies of ant communities in North American grasslands, however, have considered the composition of the surrounding landscape. In European grasslands, studies of patch-level variables (within-site), such as soil moisture, solar exposure, and adjacent land-use type are good predictors of overall ant richness and abundance (Dauber et al., 2005). Other broad scale controls on ant communities may include habitat loss (loss of area) and fragmentation (edge effects, isolation, connectivity) (Fahrig, 2003). In the Midwest US most of the grasslands reconstructed through incentive programs are less than 50 ha in size, are isolated from any native prairie habitats by several kilometers, and may have potentially strong edge effects with the surrounding land uses.. Experimental tests of multiple insect taxa tend to support the positive relationship between species and area, as predicted by the Island Biogeography Theory (Debinski & Holt, 2000; MacArthur & Wilson, 1967); however, ants do not show consistent relationships with habitat area (Crist, 2008). Edge effects can modify the abiotic conditions of the habitat and shift structure and composition of ant communities, by increasing abundances of invasive species (Holway, 1998) and decreasing rare species in patches (Golden & Crist, 2000) or by shifting nest densities toward or away from the edge, depending on the species’ abiotic preferences (Dauber & Wolters, 2004; Gibb & Hochuli, 2002). Ant communities in grasslands are further structured by variation among grassland patches due to periodic disturbance (typically controlled burns or grazing), or heterogeneity in soil moisture, and soil texture (Debuse et al., 2007; Gollan et al., 2014; Graham et al., 2009). Periodic fire is the most common land management technique for maintaining conservation grasslands, and can have positive, negative, or no apparent effects on invertebrate taxa (Debinski et al., 2011; Swengel, 2001). Most ants are

protected underground from direct mortality by fire, but may experience indirect effects of fire through changes in other abiotic and biotic conditions, such as insolation, litter depth, prey availability, temperature, moisture, and soil nutrients. The responses of ants to fire, ranges from no change, to positive or negative effects on abundance and richness, and shifts in functional groups and community composition (Underwood & Fisher, 2006). Similarly, ants show diverse responses to variation in soil characteristics (texture, moisture, bulk density). For example Graham et al. (2009) found species evenness to be related to soil texture, while richness and abundance were not significantly related. Soil texture was also demonstrated as an important characteristic structuring ant community composition in semi-arid, grazed steppes (Bestelmeyer & Wiens, 2001). Likewise, in shortgrass prairies, ant species richness, abundance, and community composition was primarily determined by various soil attributes and showed little to no relationship with plant community characteristics (Boulton et al., 2005). The goals of this study were to determine the effects of patch (grassland site level characteristics) and surrounding landscape on ant communities in conservation grasslands. At the patch-level, we hypothesize that: (i) colonization of grasslands by ants is determined by dispersal limitations and tolerance to disturbance, such that anthropogenically favored species (species commonly found in urban environments) will be more common in younger sites, while uncommon specialist species will be primarily associated with older sites; (ii) soil texture acts as a filter of ant species composition, such that coarse-textured soils are suitable to a wider range of ant species compared to fine- textured soils; and (iii) ant communities are structured by disturbance in conservation grasslands, and will primarily respond to abiotic environmental characteristics (age and management) rather than biotic (vegetation) components. We hypothesized the following effects of the surrounding landscape composition: (i) ant species richness in grasslands is higher in habitats surrounded by grasslands and pasture/hay fields, because these habitats are potential sources of propagules; (ii) anthropogenically favored ants are more frequent in patches surrounded by urbanization and intensive agriculture, while uncommon specialists are less frequent; and (iii) sites with greater edge:area ratios will have higher frequencies of anthropogenically favored ant species.

3 Methods 3.1 Study Sites We studied a total of 23 warm-season conservation grasslands during the summers of 2011 and 2012 in Butler, Preble, and Montgomery Counties of southwestern Ohio (Supplementary Figure 1, Table 1). The grassland patches were planted and managed by private landowners (primarily through enrollment in the CRP), MetroParks of Butler County, Five Rivers MetroParks, a high school, and the Miller-Coors Corporation. Participation by landowners was voluntary and only 20 grasslands were used each year, with 17 grasslands sampled during both of the years. Sites ranged in size from 0.83-17.8 ha (mean =7.3 ha), and time since planting (age=1-31yrs, mean =10.7yrs). Management practices varied among grasslands, and included burning every 3-5 years, mowing, tilling, or minor spot treatment of invasive plants with herbicides. Two grasslands were burned at the beginning of the 2011 season, and three were burned or partially burned at the beginning of 2012. Dominant vegetation consisted primarily of C4 prairie grasses such as Indian grass (Sorgastrum nutans), big blue stem (Andropogon gerardii), and little blue stem (Schizachyrium scoparium), and a wide variety of prairie forb species (Supplementary Table 1). Our sites were surrounded by a variety of land use land cover types, including forest (mean=33.7%, range= 0-72.51%), intensive agriculture (mean 28.4%, range=0-64%, or urbanization (mean-15.4%, range=0-67.5%), with small amounts of grassland (mean=12.97%, range=1-34.1%) and extensive agriculture (mean=5.73%, range=0-20.73%) (Supplementary Table 2).

3.2 Ant Community Sampling Ant communities were surveyed using pitfall traps, a common and effective method for collecting active ground running arthropods (Andersen, 1991; Schlick-Steiner et al., 2006; Underwood & Fisher, 2006). To account for heterogeneity of the habitat and to avoid under-sampling of large sites relative to small ones, the number of traps per site was scaled to the log of the patch area. We established a transect through the center of each grassland consisting of 5-10 pitfall traps spaced 25 m apart (Supplementary Figure 2). Each pitfall trap consisted of a 237-ml specimen cup (7 cm diameter, 10 cm deep) inserted flush with the ground and propylene glycol added as a preservative. A wooden

13 board, elevated with nails, was positioned over the cup to shield it from rain. Pitfall traps remained active for one week. A total of 155 and 154 traps were used in 2011 and 2012, respectively, for each of three sampling periods spaced five weeks apart.

3.3 Vegetation Sampling To quantify the local plant species richness and cover we conducted vegetation surveys at paired, 10-m2 circular quadrats, on each side of the pitfalls adjacent to the transect (Supplementary Figure 2). All plant species within the quadrat were identified and percent cover was recorded by species as categorical variables: 0, 1, 5, 10, 25, 50, 75 or 100%. Plant richness and cover were analyzed as trap averages across the site, and were analyzed by functional group including forbs, C4 grasses, C3 grasses and woody plants (Supplementary Table 1). Woody plants and C3 grasses comprised a very small proportion of cover and richness in any given patch, so we focused our analysis on overall plant richness and total cover, as well as the richness and cover of forbs and C4 grasses.

3.4 Soil Analysis We obtained soil cores (10 cm deep x 5 cm diameter) adjacent to every other pitfall trap (50 m between core samples) along the transect. Samples were rolled in aluminum foil and transported to the lab. We processed three of the innermost cores at each site for bulk density, soil organic matter, and soil texture (percentages of sand, silt, and clay) (Supplementary Table 1). Cores were weighed, slightly crumbled, then exposed to the air for one week. Air dried sample mass and volume of the cores were used for bulk density calculations. The dry soil samples were crumbled further with a mortar and pestle and pushed through a 2 mm sieve to remove coarse fragments. Samples were ground again to reduce the small clay aggregates and homogenized for subsamples. Soil organic matter was quantified by burning three 20-g oven-dried subsamples of soil in a muffle furnace (450ºC for 8 hrs). We calculated the percent soil organic matter as the average mass loss after ignition. Soil texture methods followed Sheldrick & Wang (1993). We subsampled 40 g from the homogenized soil sample and combined it with 100 ml of 5% Sodium Hexametaphosphate (dispersing agent) in a 250 ml Nalgene bottle. We shook each

sample on a reciprocating shaker for 12 hrs. A control 1 L cylinder (RL) consisting of 100 ml dispersing agent and 900 ml distilled water, and a second cylinder consisting of the soil subsample in solution and additional volume of distilled water (brought to 1 L) were used for the hydrometer texture analysis. After mixing, the R40s reading was taken at exactly 40 s (yielding the sand fraction), then 12 hrs later a second reading was taken (R12h). Percentages of sand, clay, and silt were calculated using these formulas: Sand (%) = (1 - [(R40s - RL) / sample dry wt]) x 100, Clay (%) = [(R12hr - RL) / sample dry wt] x 100, Silt (%) = 100 - (% sand + % clay).

3.5 Additional Patch-level Variables We measured within-patch disturbance in two ways: time since planting (age) and time since last management action. Since all of the sites were planted on former agricultural land we used time since planting (hereafter “age”) as a measure of disturbance recovery. The grasslands are periodically managed by agency personnel or land owners to control invasive weeds and woody plants through burning, mowing, or in one case tilling. We tested time since management (designated “time since burn” hereafter) as a second measure of disturbance. We also used patch area and shape in our analysis. Although we did not sample ant communities at the edge of the habitat, the amount of edge relative to area has been shown to be important in the colonization of anthropogenically favored ants (Holway, 2005). We also measured grassland area as a potential predictor of ant community structure as the target area for ants following disturbance (Supplementary Table 1).

3.6 Landscape Analysis We created a land use and land cover (LULC) map of landscapes surrounding the grassland patches using 2011 National Agricultural Imagery Program Mosaics aerial photographs (1 m resolution) and Geographic Information Systems, ArcGIS version 9.3.1 (ESRI, 2009). We used six classes of LULC: (i) conservation grassland (warm season grasslands), (ii) intensive agriculture (corn and soy), (iii) extensive agriculture (pasture, cool season grasses), (iv) forest, (v) residential/urban, and (vi) water. Total area of each LULC type was measured within a 500 m radius from the central point of the grassland

transect (Supplementary Figure 1). We used percentages of total buffer area for LULC statistical analyses. Few studies have identified dispersal distances of ants, so it is not clear what an appropriate buffer radius is for measuring the influence of the surrounding LULC. A study using population genetics found plant-mutualist Crematogaster ants to be dispersing on average 468 m (Türke et al., 2010). A study of flight energetics in fire ants showed flight times of approximately 45 minutes, potentially covering over 5400 m distance (Vogt et al., 2000). European studies of ant biodiversity in grasslands have used radii from 50- 250 m (Dauber et al., 2003, 2005). We selected a 500-m radius because it was within the range of these studies, while also preventing overlap of buffers across all but two sites.

3.7 Statistical analysis We analyzed ant species richness separately for 2011 and 2012 data because three of the sites differed between the sampling years. We used generalized linear models (glm function, R) with Poisson error distributions to test models including patch (vegetation, soil, age, time since burn, area, edge:area) and landscape-level (relative amounts of surrounding LULC types) variables as predictors of overall site ant richness. We used log-transformed values for age, time since burn, and area predictors. The lowest Akaike’s Information Criterion (AIC) was used to identify single and multiple regression best and competing models (differing by 2 AIC points or less). We calculated p-values and percent deviance explained for the best and competing models by conducting a likelihood ratio test with the null model. To test individual species responses to patch and landscape variables, we conducted species-level analyses using ant-trap frequency (proportion of traps in which we encountered a given ant species) rather than overall site abundance. Often traps can be overwhelmed with great numbers of an individual species, simply because a colony is located in proximity to the trap. Ant-trap frequency can be used to reduce potential overestimation of social insects inherent in raw worker abundance (Andersen, 1991; Gotelli et al., 2011). Because our traps were 25 m apart, it is unlikely that even the most mobile ant species in adjacent traps are from the same colony. We conducted species- level analyses on ant data pooled from both years, using ant species that were found at 16

more than 5 sites, multiple traps within the sites, and with at least 50 total individuals per site. A total of 14 ant species met the minimum requirements for trap frequency analysis. We grouped these species into four categories of habitat specialization based on established habitat records (Coovert, 2005; Nemec, 2014): (i) Anthropogenically favored (Tapinoma sessile, Tetramorium caespitum) - species that are commonly found in urban environments, (ii) Uncommon specialists (Stenamma brevicorne, Temnothorax ambiguus, Pheidole tysoni) - species recorded from 20 or less counties in Ohio, (iii) Cosmopolitan open (Myrmica americana, Lasius neoniger, Solenopsis molesta, Monomorium minimum, Formica pallidefulva) - very common species found primarily in open habitats (lawns, pasture, grasslands) across the state, and (iv) Cosmopolitan open/closed (Lasius alienus, Ponera pennsylvanica, Aphaenogaster rudis, Myrmica latifrons) - very common species found primarily in open or closed habitats across the state. Since these four groups have different levels of habitat specialization, they may respond differently to patch and landscape-level controls. We used generalized linear models with a binomial error distribution to test the effects of patch and landscape variables listed above. We tested the role of patch and landscape-level variables for predicting overall species composition among patches using distance-based redundancy analysis (dbRDA) multivariate ordinations with Bray-Curtis dissimilarity (McArdle & Anderson, 2001). We used AIC to select the best fitting ordination model and obtained p-values using random permutations (999 permutations). The vegdist function in the vegan package of R (Oksanen et al., 2013) was used to calculate Bray-Curtis dissimilarity, and a dbRDA was conducted with a user-written function in R (M. Anderson, pers. comm.).

4 Results In 2011, 7962 individuals, comprising 28 species (site mean =9.95, range=3-14) were collected. In 2012, 4957 individuals of 32 species (site mean=8.85, range=4-13) were collected (Supplementary Table 3, Supplementary Table 4). The model with the lowest AIC for overall ant richness in 2011 was site age (Figure 1a); however, there were four competing models including age and % clay (ΔAIC=0.42), null (ΔAIC=1.62), % clay (ΔAIC=1.63), and area (ΔAIC=1.63) (Table 1). Likelihood ratio test of the age model versus the null model was not significantly different (p=0.057, df=1,18), and age 17

explained 22.77% of the deviance (Table 1). Soil texture was the best fitting model for ant richness in 2012 with % sand (p=0.0181, dev. expl.=32.63%) and % clay (p=0.0175, dev. expl.=36.12%, df=1,18) as inverse competing models (Figure 1b). Ant species richness was significantly greater in sandier sites than in clayey sites. Other patch-level predictors such as plant richness, cover of plant functional groups, soil organic matter, bulk density, time since burn, area, or edge:area or landscape-level predictors (LULC) were not included among the best fitting or competing models based on AIC for ant richness in 2011 or 2012. Using ant-trap frequency data, we identified how 14 different ant species respond to patch and landscape-level predictors (Table 2). Anthropogenically favored ants, T. sessile and T. caespitum, responded to elements of disturbance and landscape (young sites, time since burn, high edge:area ratio, surrounded by urbanization or intensive agriculture) but often in contrasting ways. The best model for T. sessile included site age (-), time since burn (+), and intensive agriculture (+) (ΔAIC=17.34, dev.expl=26.47%, df=3,19) (Figure 2a). There was one competing model for T. sessile that also included sand (-) (ΔAIC=0.55). Tetramorium caespitum had higher frequency in older sites that were recently burned, surrounded by less intensive agriculture but greater urbanization. The best model for predicting T. caespitum, however, was % sand (+, p<0.0001, ΔAIC=96.35, dev. expl. = 89.13%, df=1,22). The uncommon specialist ants included Pheidole tysoni, Temnothorax ambiguus, and Stenamma brevicorne. Trap frequencies for all three species were greater in older sites, but only T. ambiguus included age in the best fitting model. P. tysoni frequencies followed trends more similar to the disturbance specialists: higher in recently burned sites, with greater edge:area, and higher urbanization; however, P. tysoni responded negatively to intensive agriculture. The best model for predicting P. tysoni frequency, was % sand alone (+, p<0.0001, ΔAIC=50.98, dev. expl.=78.29%, df=1,22). Temnothorax ambiguus frequency was higher in sites that were not recently burned, with lower edge:area ratios, and was negatively related to both intensive agriculture and urbanization. The best model included age (+), area (+), and bulk density (-) and explained 41.07% of the deviance (ΔAIC=41.1, p<0.001, df=3,19) (Figure 2b). S. brevicorne was higher in sites with lower edge:area. The best model for S. brevicorne

frequency was extensive agriculture (-), grassland (+), and soil organic matter (+), explaining 53.31% of the deviance (ΔAIC=36.48, df=3,19). The other nine species with adequate numbers for trap frequency analysis included cosmopolitan ant species typical of open habitats such as Lasius neoniger, Myrmica americana, Solenopsis molesta, Monomorium minimum, and Formica pallidefulva and those found in both open and closed canopy habitats such as Myrmica latifrons, Aphaenogaster rudis, Lasius alienus and Ponera pennsylvanica (Table 2). These two groups of ant species responded in contrasting ways. Best models for open habitat cosmopolitan ants primarily responded to patch-level variables, while best models for cosmopolitan ant species also found in closed habitats all included landscape-level variables. S. molesta, M. minimum, and L. alienus had higher frequencies in sites that were recently burned, while M. americana had higher frequencies in less recently burned sites (Figure 2c). C4 cover was in best fitting models for F. pallidefulva (+), A. rudis (+), and L. neoniger (-). Multiple species responded to soil characteristics including M. latifrons (+ Soil Organic Matter), M. minimum (+ % Sand), (F. pallidefulva (- % Silt), L. alienus (+ % Silt). Most of the species (6 of the 9) included both patch and landscape- level predictors in their best models, and only one ant species model (P. pennsylvanica) did not have any patch-level predictors (Figure 2d).

4.1 Ant Species Composition Ant community composition among the 23 sites was best explained by a combination of both patch and landscape variables (R2=0.25, p=0.003). The best fitting model included urban land use, percent sand, and time since burn (Figure 3). While time since burn and percent sand routinely came up in the species trap frequency models, urban land use was only present in competing models suggesting that the less common 14 of the 28 ant species found at these sites may be responding to urban land use in a predictable way.

5 Discussion

5.1 Relative Role of Patch and Landscape-level Predictors Ant biodiversity in conservation grasslands depended on both patch and landscape-level processes. Ant species richness was determined by two patch-level variables: age of the site and soil texture. The effect of age is consistent with other chronosequence studies of ant community development (Dauber & Wolters, 2005; Phipps, 2006). Soil texture is an important filter for ant communities (Bestelmeyer & Wiens, 2001; Boulton et al., 2005; Graham et al., 2009), and our findings add further support for the role of soil texture. In particular, we find that grasslands with soils high in clay tend to harbor fewer ant species than those with sandier soils. Diverse species responses to management and landscape predictors may have diluted the strength of the effects on total ant species richness. Our ant community composition analysis showed that soil, management, and landscape-level predictors are all important factors influencing grassland ant communities, and our trap-level species analyses mirrored these results. This highlights the importance of identifying species-level or functional-group responses if ants are to be used as bioindicators to changes in soils or other environmental conditions.

5.2 Anthropogenic and Uncommon Specialist Responses Anthropogenically favored ant species (T. sessile and T. caespitum) responded in opposite ways to soil type and site age, but both responded positively, as predicted, to disturbance in surrounding land use (intensive agriculture and urbanization, respectively). In general, T. sessile was more frequent in young sites, disturbed by intensive agriculture, while T. caespitum frequency is primarily predicted by sand content of the soil. We expected our three uncommon specialist species (T. ambiguus, P. tysoni, and S. brevicorne) to respond negatively to disturbance in surrounding landscape (intensive agriculture and urbanization) and edge, while responding positively to site age, extensive agriculture, and grasslands in the landscape matrix. T. ambiguus frequency was positively related to site age and area, and best models for both S. brevicorne and T. ambiguus included soil properties known to improve over time in CRP lands (decreased bulk density and increased organic matter) (McLauchlan et al., 2006). Additionally, these two 20 species responded positively to grassland in the surrounding landscape; however, S. brevicorne was negatively influenced by extensive agriculture. Our results suggest that these two species can be potential indicators of grassland recovery from disturbance at the landscape and patch levels. P. tysoni frequency was primarily driven by soil sand content; and in this and other ways, responded similarly to T. caespitum, although P. tysoni is not known to be associated with urban environments.

5.3 Landscape-level Responses Landscape variables did not appear in best models for ant species richness, yet they appeared in best or competing models describing frequency of 12 of the 14 ant species that were analyzed individually, which clearly indicates that the ant assemblages in these small grassland patches are highly dependent on, and likely colonize from, the surrounding landscape. We found that grasslands and urbanization in the surrounding landscape increased richness, while intensive and extensive agriculture reduced richness, though these did not appear in the best models for overall ant species richness. Although we did not specifically test this, grasslands in landscapes surrounded by greater proportions of grassland land cover may accumulate more rare species, while urbanization extends the assemblage to include disturbance specialists that tend to be absent in sites with low disturbance. Grazing has been shown to negatively affect ant richness (Bestelmeyer & Wiens, 2001; Boulton et al., 2005), and since most of the extensive agricultural fields in our region are either harvested or used as pasture, they may have reduced propagule contribution to conservation grasslands. Our results indicate that landscape-level effects may structure ant species composition, but that ant species richness is primarily driven by patch-level characteristics. We did not have a priori predictions for cosmopolitan ant species in the ant trap frequency models. We observed different relative influences of surrounding land use on the two species groups (open and open/closed). Best models for open/closed habitat ant species all included landscape variables, while ant species typical of open habitats were primarily affected by patch-level variables. Surprisingly, forest land use was not the most common of landscape predictors (except for M. latifrons) for the cosmopolitan open/closed habitat ant species. Instead open/closed habitat ants showed negative

21 responses to intensive and extensive agriculture and positive responses to grassland land use.

5.4 Management Practices The effect of grassland management, such as burning, on total ant richness or abundance typically has not shown consistent trends in most ant studies; however, species or functional-group responses are often recorded (Joern & Laws, 2013; Underwood & Fisher, 2006). In our study, time since burn was included in 5 of the 14 species models. T. sessile and M. americana responded positively to increased time since burn, while S. molesta, M. minimum, and L. alienus both responded negatively. The positive relationships of M. americana and T. sessile may be related to their nest structure. We rarely observed T. sessile with nests below the soil surface, and frequently we found short term nests under layers of grass litter. We often observed M. americana nests within the bases of large grass clumps. These clumps would be protected by fire; however, they also built turret-like thatch structures up to 20 cm up the stems of the grass in which they warmed their larvae during the day. We suspect that aboveground components of both of these nests are negatively affected by routine burning. L. alienus, S. molesta, and M. minimum all nest in the soil and may simply be captured at higher rates, as shown in other studies (Melbourne, 1999), when the litter structure is less complex. The lack of generality among studies of grassland management may be largely driven by differences in community composition.

5.5 Soil Characteristics Soil texture and soil type have consistently shown strong associations with ant communities (Bestelmeyer & Wiens, 2001; Boulton et al., 2005; Graham et al., 2009), and our results support the importance of these predictors as environmental filters for the assembling ant community in constructed grasslands. Soil characteristics, especially soil texture, were included in best or competing models for 9 of the 14 ant species in this study, across habitat association groups, total ant richness, and community composition. Ant species responded differentially to components of soil texture likely due to species specific constraints related to the physical ability of ants to nest in the soil, for example soil moisture and ease of excavation. Additional soil variables associated with age, such

as bulk density and soil organic matter (McLauchlan et al., 2006), consistently had negative and positive relationships, respectively. Colonization of ants as the grassland ages may covary with these other age dependent soil characteristics. The physical template of the soil is a significant patch-level control on ant community assembly, and ant communities may act as sentinels of age-dependent characteristics of soils.

5.6 Vegetation Vegetation played a very limited role in the grassland ant community. The conservation grasslands in our study landscapes showed no relationship between plant richness or cover of functional groups and grassland age (Peters et al., in review), management, or soil characteristics (unpublished data). This is likely because constructed grasslands are highly dependent on the initial seed mix and weather conditions in the first few years of establishment, unlike remnant prairies (Dickson & Busby, 2009). Although ants are highly associated with dominant vegetation types, (e.g. forests versus grasslands) most studies have found little to no relationship with plant richness or cover within a given vegetation type (Hill et al., 2008). Some studies have identified plant structure or vegetation height as a more important characteristic for predicting ants in grasslands (Debinski et al., 2011; New, 2000). In our study, C4 grass cover appeared in three models for cosmopolitan ant species. F. pallidefulva and A. rudis were positively related to C4 cover, while L. neoniger, the most abundant and cosmopolitan ant species, was negatively influenced by C4 cover and age. Prescribed burns are typically conducted in the spring at our sites, a practice that tends to favor C4 grasses over time (Howe, 2000; Steuter, 1987). The inverse relationship between C4 cover and age may be paralleled by L. neoniger. Our results suggest that ant communities are primarily structured by physical aspects of their environment (soil texture, structure, and disturbance) rather than vegetation.

5.7 Conclusions Ant communities showed turnover in species composition as constructed grasslands age, which may result in different functional roles and significance of ants at multiple points in the species assembly process. European studies have found increasing nest density as a site ages and a shift in the ant community from primarily aboveground

foraging species to subterranean species in older sites (Dauber &Wolters, 2005). Several of the ant species in our study (especially Lasius) tend aphids above or belowground, and shifts in relative abundances of these species could affect grassland plant communities or have spillover effects on neighboring cropland. Other species in our study are primarily predators and may alter grassland communities above and belowground through both inter and intraguild predation. For example, ants in the genus Strumigenys prey on soil microarthropods and may cause shifts in the belowground food web, while Formica ants are wide ranging predators often observed foraging on insects and tending aphids on vegetation (Coovert, 2005). Mutualisms among ants and the grassland community may also shift over time. Multiple Formica species are hosts to butterflies in the family Lycaenidae and Microdon flies (Syrphidae) that are pollinators, and Pheidole and Aphaenogaster species are significant seed dispersal agents (Coovert, 2005). Understanding temporal shifts in ant communities can promote our understanding of species interactions in these constructed grassland communities. Future work could elucidate if ant functional roles shift as the grassland develops and if the strength of species interactions varies in different environmental contexts. The sorting and assembly process in grasslands may lead to shifts from early colonizers that are primarily generalist predators with limited or generalist species interactions to later colonizers with more specialized interactions (mutualisms, seed dispersal, and belowground food web interactions). Ant community assembly in constructed grasslands is a time dependent process, shaped by the physical characteristics of the patch, and is reliant on the surrounding landscape for colonizing propagules. Our work promotes ants as environmental indicators of management and soil characteristics, but notes that aggregate measures of ant diversity (species richness) may not be effective in documenting the more subtle shifts in ant community composition.

6 References Andersen, A. N. (1991). Sampling communities of ground-foraging ants: pitfall catches compared with quadrat counts in an Australian tropical savanna. Austral Ecology, 16(3), 273–279. Andersen, A. N., & Majer, J. D. (2004). Ants show the way down under: invertebrates as bioindicators in land management. Frontiers in Ecology and the Environment, 2, 291–298. Bestelmeyer, B. T., & Wiens, J. A. (2001). Ant biodiversity in demiarid landscape mosaics: the consequences of grazing vs. natural heterogeneity. Ecological Applications, 11(4), 1123–1140. Boulton, A. M., Davies, K. F., & Ward, P. S. (2005). Species richness, abundance, and composition of ground-dwelling ants in northern California grasslands: role of plants, doil, and grazing. Environmental Entomology, 34(1), 96–104. Burke, I. C., Lauenroth, W. K., & Coffin, D. P. (1995). Soil Organic Matter Recovery in Semiarid Grasslands : Implications for the Conservation Reserve Program. Ecological Applications, 5(3), 793–801. Coovert, G. A. (2005). The Ants of Ohio. Columbus, Ohio: Ohio Biological Survey, Inc. Crist, T. O. (2008). Biodiversity, species interactions, and functional roles of ants (Hymenoptera: Formicidae) in fragmented landscapes: a review. Myrmecological News, 12, 3–13. Dauber, J., Hirsch, M., Simmering, D., Waldhardt, R., Otte, A., & Wolters, V. (2003). Landscape structure as an indicator of biodiversity: matrix effects on species richness. Agriculture, Ecosystems & Environment, 98(1-3), 321–329. Dauber, J., Purtauf, T., Allspach, A., Frisch, J., Voigtländer, K., & Wolters, V. (2005). Local vs. landscape controls on diversity: a test using surface-dwelling soil macroinvertebrates of differing mobility. Global Ecology and Biogeography, 14, 213–221. Dauber, J., & Wolters, V. (2004). Edge effects on ant community structure and species richness in an agricultural landscape. Biodiversity and Conservation, 13, 901– 915. Dauber, J., & Wolters, V. (2005). Colonization of temperate grassland by ants. Basic and Applied Ecology, 6(1), 83–91. Debinski, D. M., & Holt, R. D. (2000). A survey and overview of habitat fragmentation experiments. Conservation Biology, 14(2), 342–355. Debinski, D. M., Moranz, R. a., Delaney, J. T., Miller, J. R., Engle, D. M., Winkler, L. B., … Gillespie, M. K. (2011). A cross-taxonomic comparison of insect responses to grassland management and land-use legacies. Ecosphere, 2(12), art131. Debuse, V. J., King, J., & House, A. P. N. (2007). Effect of fragmentation, habitat loss and within-patch habitat characteristics on ant assemblages in semi-arid woodlands of eastern Australia. Landscape Ecology, 22, 731–745. Dickson, T. L., & Busby, W. H. (2009). Forb species establishment increases with decreased grass seeding density and with increased forb seeding density in a northeast Kansas, U.S.A., experimental prairie restoration. Restoration Ecology, 17(5), 597–605.

Dunn, C. P., Stearns, F., Guntenspergen, G. R., & Sharpe, D. M. (1993). Ecological benefits of the conservation reserve program. Conservation Biology, 7(1), 132– 139. ESRI. (2009). ArcGIS. Redlands, CA: Environmental Systems Research Institute. Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity. Annual Review of Ecology, E Volution & Systematics, 34(2003), 487–515. Fahrig, L., Baudry, J., Brotons, L., Burel, F. G., Crist, T. O., Fuller, R. J., Sirami, C., Siriwardena, G. M., Martin, J. (2011). Functional landscape heterogeneity and animal biodiversity in agricultural landscapes. Ecology Letters, 14 (101-112). Folgarait, P. J. (1998). Ant biodiversity and its relationship to ecosystem functioning: a review. Biodiversity and Conservation, 7(9), 1221–1244. Gibb, H., & Hochuli, D. F. (2002). Habitat fragmentation in an urban environment: large and small fragments support different arthropod assemblages. Biological Conservation, 106(2002), 91–100. Golden, D. M., & Crist, T. O. (2000). Experimental effects of habitat fragmentation on rove beetles and ants: patch area or edge? Oikos, 90, 525–538. Gollan, J. R., Ramp, D., & Ashcroft, M. B. (2014). Contrasting topoclimate, long-term macroclimatic averages, and habitat variables for modelling ant biodiversity at landscape scales. Insect Conservation and Diversity, 43–53. Gómez, C., Casellas, D., Oliveras, J., & Bas, J. M. (2003). Structure of ground-foraging ant assemblages in relation to land-use change in the northwestern Mediterranean region. Biodiversity and Conservation, 12(10), 2135–2146. Gotelli, N. J., Ellison, A. M., Dunn, R. R., & Sanders, N. J. (2011). Counting ants (Hymenoptera: Formicidae): biodiversity sampling and statistical analysis for myrmecologists. Myrmecological News, 15, 13–19. Graham, J. H., Krzysik, A. J., Kovacic, D. a., Duda, J. J., Freeman, D. C., Emlen, J. M., … Balbach, H. E. (2009). Species richness, equitability, and abundance of ants in disturbed landscapes. Ecological Indicators, 9(5), 866–877. Hendrickx, F., Maelfait, J. P., Van Wingerden, W., Schweiger, O., Speelmans, M., Aviron, S., … Bugter, R. (2007). How landscape structure, land-use intensity and habitat diversity affect components of total arthropod diversity in agricultural landscapes. Journal of Applied Ecology, 44(2), 340–351. Hill, J. G., Summerville, K. S., & Brown, R. L. (2008). Habitat associations of ant species (Hymenoptera: Formicidae) in a heterogeneous Mississippi landscape. Environmental Entomology, 37(2), 453–463. Holway, D. A. (1998). Effect of argentine ant invasions on ground-dwelling arthropods in northern California riparian woodlands. Oecologia, 116(1-2), 252–258. Holway, D. A. (2005). Edge effects of an invasive species across a natural ecological boundary. Biological Conservation, 121, 561–567. House, G. J. (1989). Soil arthropods from weed and crop roots of an agroecosystem in a wheat-soybean-corn rotation: impact of tillage and herbicides. Agriculture, Ecosystems & Environment, 25, 233–244. House, G. J., & Stinner, B. R. (1983). Arthropods in no-tillage soybean agroecosystems: Community composition and ecosystem interactions. Environmental Management, 7(1), 23–28.

Howe, H. F. (2000). Grass response to seasonal burns in experimental plantings. Journal of Range Management, 53(4), 437–441. Joern, A., & Laws, A. N. (2013). Ecological mechanisms underlying arthropod species diversity in grasslands. Annual Review of Entomology, 58, 19–36. Kleijn, D., & Sutherland, W. J. (2003). How effective are European agri-environment schemes in conserving and promoting biodiversity? Journal of Applied Ecology, 40(6), 947–969. Knop, E., Kleijn, D., Herzog, F., & Schmid, B. (2006). Effectiveness of the Swiss agri- environment scheme in promoting biodiversity. Journal of Applied Ecology, 43(1), 120–127. MacArthur, R. H., & Wilson, E. O. (1967). The Theory of Island Biogeography. Princeton, NJ: Princeton University Press. Mandelik, Y., Winfree, R., Neeson, T. & Kremen, C. (2012). Complementary habitat use by wild bees in agro-natural landscapes. Ecological Applications, 22(5), 1535- 1546. Martin, L. M., Moloney, K. A., & Wilsey, B. J. (2005). An assessment of grassland restoration success using species diversity components. Journal of Applied Ecology, 42(2), 327–336. McArdle, B. H., & Anderson, M. J. (2001). Fitting multivariate models to community data: a comment on distance-based redundancy analysis. Ecology, 82(1), 290– 297. McLauchlan, K. K., Hobbie, S. E., & Post, W. M. (2006). Conversion from agriculture to grassland builds soil organic matter on decadal timescales. Ecological Applications, 16(1), 143–53. McLaughlin, A., & Mineau, P. (1995). The impact of agricultural practices on biodiversity. Agriculture, Ecosystems & Environment, 55(3), 201–212. Melbourne, B. a. (1999). Bias in the effect of habitat structure on pitfall traps: an experimental evaluation. Australian Journal of Ecology, 24(3), 228–239. Millenbah, K. F., Winterstein, S. R., Campa III, H., Furrow, L. T., & Minnis, R. B. (1996). Effects of conservation reserve program field age on avian relative abundance, diversity, and productivity. The Wilson Bulletin, 108(4), 760–770. Nemec, K. T. (2014). Tallgrass prairie ants: their species composition, ecological roles, and response to management. Journal of Insect Conservation, 18, 509–521. New, T. R. (2000). How useful are ant assemblages for monitoring habitat disturbance on grasslands in south eastern Australia ? Journal of Insect Conservation, 4, 153– 159. Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O’Hara, R. B., … Wagner, H. (2013). vegan: Community Ecology Package. R package version 2.0- 10. Retrieved from http://cran.r-project.org/package=vegan Panzer, R., Gnaedinger, K., & Derkovitz, G. (2010). The prevalence and status of conservative prairie and sand savanna insects in the Chicago Wilderness Region. Natural Areas Journal, 30(1), 73–81. Peck, S. L., McQuaid, B., & Campbell, C. L. (1998). Using ant species (Hymenoptera: Formicidae) as a biological indicator of agroecosystem condition. Community and Ecosystem Ecology, 27, 1102–1110.

Peters, V. E., Campbell, K. U., Dienno, G., García, M., Leak, E., Loyke, C., … Crist, T. O. (2015). Plant evenness indicates conservation value of constructed grasslands over quality assessment indicators. Ecological Indicators, in review. Phipps, S. J. (2006). Biodiversity of ants (Hymenoptera: Formicidae) in restored grasslands of different ages. University of Missouri-Columbia. Samson, F., & Knopf, F. (1994). Prairie Conservation in North America. BioScience, 44(6), 418–421. Schlick-Steiner, B. C., Steiner, F. M., Moder, K., Bruckner, A., Fiedler, K., & Christian, E. (2006). Assessing ant assemblages: pitfall trapping versus nest counting (Hymenoptera, Formicidae). Insectes Sociaux, 53(3), 274–281. Sheldrick, B. H., & Wang, C. (1993). Particle size distribution. In M. R. Carter (Ed.), Soil Sampling and Methods of Analysis (pp. 499–512). Boca Raton, Florida: Candian Society of Soil Science. Söderström, B. O., Svensson, B., Vessby, K., & Glimskär, A. (2001). Plants, insects and birds in semi-natural pastures in relation to local habitat and landscape factors. Biodiversity and Conservation, 10, 1839–1863. Steuter, A. A. (1987). Society for range management C3/C4 production shift on seasonal burns-northern mixed prairie. Journal of Range Management1, 40, 27–31. Stinner, B. R., & House, G. J. (1990). Arthropods and other invertebrates in conservation-tillage agriculture. Annual Review of Entomology, 35, 299–318. Stoate, C., Boatman, N. D., Borralho, R. J., Carvalho, C. R., de Snoo, G. R., & Eden, P. (2001). Ecological impacts of arable intensification in Europe. Journal of Environmental Management, 63(4), 337–365. Swengel, A. B. (2001). A literature review of insect responses to fire, compared to other conservation managements of open habitat. Biodiversity and Conservation, 10, 1141–1169. Tscharntke, T., Klein, A. M., Kruess, A., Steffan-Dewenter, I., & Thies, C. (2005). Landscape perspectives on agricultural intensification and biodiversity-ecosystem service management. Ecology Letters, 8, 857–874. Tscharntke, T., Steffan-Dewenter, I., Kruess, A., & Thies, C. (2002). Contribution of small habitat fragments to conservation of insect communities of grassland- cropland landscapes. Ecological Applications, 12(2), 354–363. Türke, M., Fiala, B., Linsenmair, K. E., & Feldhaar, H. (2010). Estimation of dispersal distances of the obligately plant-associated ant Crematogaster decamera. Ecological Entomology, 35, 662–671. Underwood, E. C., & Fisher, B. L. (2006). The role of ants in conservation monitoring: if, when, and how. Biological Conservation, 132(2), 166–182. Vogt, J. T., Appel, A. G., & West, M. S. (2000). Flight energetics and dispersal capability of the fire ant, Solenopsis invicta Buren. Journal of Insect Physiology, 46, 697– 707.

Table 1. Best and competing models for ant richness in 2011 and 2012.

ΔAIC Ant vs. % Dev. ΔAIC vs. % Dev. Richness Best Model Null Expl. Competing Models Best Expl. 2011 (+)Log (age) -1.62 22.51% (+)Log (age) + (-)% Clay +0.42 32.32% (-)% Clay +1.63 12.41% (+)Log (area) +1.63 12.40% 2012 (-)% Clay -3.97 36.12% (+)% Sand + (-)% Clay +0.06 47.87% (+)% Sand +0.58 32.63%

Figure 1. Best models for ant species richness in (a) 2011 included time since planting (Grassland age) and (b) 2012 included percentage of clay fraction of the soil (X and Y axes log-scaled).

Figure 2. Best fitting models for (a) an anthropogenically favored species (Tapinoma sessile), (b) an uncommon specialist species (Temnothorax ambiguus), (c) a cosmopolitan open habitat species (Myrmica americana), and a (d) Cosmopolitan open /closed habitat species (Ponera pennsylvanica).

Table 2. Best and competing models for ant species. Habitat Association indicates anthropogenically favored species, uncommon specialists, and cosmopolitan species found in open and/or closed habitats ( LU= Land use).

ΔAIC ΔAIC Habitat vs. % Dev. vs. % Dev. Ant Species Association Best Model Null Expl. Competing Models Best Expl. Tapinoma (-)Log (age) + (+)Time since burn + (-)Log (age) + (+)Time since burn+ (+)IntensiveLU Anthropogenic -17.34 26.47% +0.55 28.11% sessile (+)IntensiveLU + (-)% Sand Tetramorium Anthropogenic (+) % Sand -96.36 89.23% (+) %Sand + (+)UrbanizedLU +1.45 89.72% caespitum Temnothorax Uncommon (+)Log (area) + (+)Log (age) + (+)Log (area) + (+)Log (age) + (-)Bulk density -40.50 41.10% +1.84 41.07% ambiguus specialist (-)Bulk density + (+)GrasslandLU Uncommon Pheidole tysoni (+) % Sand -50.98 79.29% ------specialist Stenamma Uncommon (+)Organic Matter + (+)GrasslandLU + 36.49 53.31% (+)GrasslandLU + (-)ExtensiveLU +2.04 48.24% brevicorne specialist (-)ExtensiveLU Cosmopolitan Lasius neoniger (-)Log (age) + (-)C4 cover -19.88 27.42% ------Open Myrmica Cosmopolitan (+) Time since burn -16.13 18.42% +Time since burn + (-)UrbanizedLU +1.58 18.85% americana Open +Time since burn + (+)Log (area) +1.98 18.44% Formica Cosmopolitan (-)% Silt+ (+)C4 cover + -8.74 29.08% (-)% Silt + (+)C4 cover +1.21 22.76% pallidefulva Open (+)ExtensiveLU (-)% Silt +1.95 17.34% Solenopsis Cosmopolitan (-)Time since burn + (-)GrasslandLU + (-)Time since burn + (-)GrasslandLU -23.95 21.60% +1.83 21.74% molesta Open (+)UrbanizedLU Monomorium Cosmopolitan (-)Time since burn + (+)% Sand -73.20 48.95% (-)Time since burn + (+)% Sand+ (+)UrbanizedLU +1.54 49.25% minimum Open Aphaenogaster Cosmopolitan (+)C4 cover + (-)IntensiveLU -14.37 16.54% (+)C4 cover + (-)IntensiveLU + (+)Log (age) +0.97 17.47% rudis Open/closed Myrmica Cosmopolitan (+)Organic Matter + (+)ForestLU -20.14 31.91% (+)Organic Matter + (+)ForestLU + (-)Bulk density +1.22 32.94% latifrons Open/closed Lasius Cosmopolitan (-)Time since burn + (+)% Silt + -46.08 60.45% ------alienus Open/closed (+)GrasslandLU + (-)ExtensiveLU Ponera Cosmopolitan (+)GrasslandLU + (-)ExtensiveLU -10.44 22.42% (+)GrasslandLU + (-)ExtensiveLU + (+)Log (age) +1.20 29.87% pennsylvanica Open/closed

Figure 3. Multivariate ordination for ant community composition across 23 grasslands. Symbols are site scores for individual grasslands scaled to the percentage value of the most important predictor variable (Urban Land Use). Arrows are biplot correlations of significant predictor 2 variables (Urban Land Use, Time since burn and Percent Sand) (R =0.251, p=0.003).

Supplementary Figure 1. Locations of 23 sites in Butler, Preble, and Montgomery Counties. Surrounding landscapes of four sites are shown (numbered circles).

Supplementary Table 1. Descriptive table of patch-level variables. Divisions represent general site characteristics, disturbance, plant, and soil variables.

Time Time Since Since Plant Forb C4 Organic Bulk Area Age Edge: Burn Burn Plant Cover Forb Cover C4 Cover Matter Density Sand Clay Silt Site (ha) Traps 2012 Area 2011 2012 Rich. (%) Rich. (%) Rich. (%) (g) (g/cm3) (%) (%) (%) 1 1.5 5 7 0.04 0 1 16 81.8 10 11.0 4 67.4 3.6 2.8 5.7 19.4 74.9 2 1.6 6 6 0.04 5 0 10 103.5 5 26.1 4 69.0 4.2 3.0 18.4 18.6 63.0 3 4 7 15 0.03 2 3 29 106.2 17 7.9 3 87.5 4.5 3.2 18.6 19.8 61.6 4 4.9 8 15 0.02 1 2 78 92.9 54 50.1 5 25.1 10.9 2.2 28.8 23.6 47.6 5 1.2 5 5 0.05 4 5 37 108.3 24 23.1 5 79.9 4.7 3.4 13.9 26.9 59.2 6 17.8 10 16 0.01 15 16 47 98.4 28 33.5 5 54.6 5.8 3.0 29.6 23.6 46.8 7 11 9 5 0.02 4 5 54 112.6 32 52.4 4 20.2 8.9 2.4 12.5 34.0 53.5 8 5.9 8 14 0.02 2 0 39 107.1 20 16.8 5 77.8 5.7 3.0 10.2 30.2 59.6 9 12.1 9 20 0.02 3 4 33 106.1 23 63.7 3 38.4 4.6 3.4 27.7 20.4 51.9 10 4.9 8 5 0.02 0 1 40 75.2 30 39.9 3 33.8 4.5 3.4 46.4 19.0 34.6 11 12 9 10 0.01 3 4 33 140.8 23 35.9 3 31.1 5.8 2.8 22.6 22.3 55.1 12 11.9 9 24 0.02 3 4 23 108.7 14 11.8 1 82.5 5.5 3.0 14.7 22.4 62.9 13 12.1 9 9 0.01 1 2 38 118.2 25 108.4 1 0.1 6.5 3.0 2.1 19.4 78.5 14 4 7 6 0.02 4 5 58 103.5 37 30.4 5 54.1 4.0 3.1 14.4 21.9 63.6 15 1.6 6 6 0.04 4 5 29 98.9 17 44.8 3 43.0 3.6 3.6 26.5 22.8 50.7 16 2.8 6 8 0.02 7 8 32 63.0 24 36.4 4 24.0 5.0 3.1 28.0 31.5 40.6 17 20.7 10 11 0.01 10 11 13 99.5 4 24.7 3 70.9 3.9 2.9 7.0 20.7 72.4 18 16.2 10 11 0.02 10 11 63 92.0 37 30.3 6 40.1 4.7 2.6 22.1 25.3 52.7 19 4.5 8 31 0.02 1 1 21 99.1 16 25.2 3 69.4 4.2 3.6 55.2 19.1 25.7 20 0.8 5 11 0.05 4 0 27 103.0 21 61.3 2 39.1 5.9 3.1 47.5 16.1 36.4 21 2.4 6 1 0.02 NA 1 46 68.8 28 55.5 3 5.8 3.7 2.6 6.0 17.4 76.6 22 8.1 9 10 0.02 1 2 23 94.2 15 8.3 4 85.3 3.7 3.6 53.1 14.4 32.5 23 4.9 8 4 0.02 3 4 51 114.6 36 81.8 4 22.1 3.9 3.3 13.8 18.6 67.7

Supplementary Table 2. Descriptive table of landscape-level variables (LU=Land use). Site ExtensiveLU ForestLU GrasslandLU IntensiveLU UrbanLU 1 2.2 33.5 7.0 50.2 4.4 2 1.5 34.5 7.6 46.9 7.1 3 8.0 72.5 12.0 1.0 3.2 4 0.0 38.1 5.7 0.0 52.4 5 0.0 46.2 6.5 41.5 3.4 6 9.0 20.0 34.1 21.9 12.5 7 8.7 31.9 10.5 32.9 13.3 8 20.6 31.6 11.1 3.2 29.5 9 0.0 41.6 14.4 37.4 3.7 10 17.5 24.5 7.8 26.4 19.0 11 0.0 35.1 24.4 30.4 7.6 12 0.0 56.5 20.4 25.6 0.4 13 0.0 22.9 33.5 37.4 0.0 14 7.0 54.3 5.0 22.5 6.8 15 0.0 47.2 1.6 36.5 10.5 16 15.3 5.8 8.4 63.9 4.2 17 5.1 9.0 24.8 43.2 15.1 18 20.7 46.4 19.1 10.2 1.1 19 0.0 0.0 18.9 29.3 49.3 20 0.0 10.7 1.0 1.9 67.5 21 2.2 34.6 7.2 31.9 21.7 22 0.0 40.8 9.2 22.1 21.4 23 13.9 37.8 8.2 37.4 0.0

Supplementary Figure 2. Diagram of sampling methods. Pitfall traps and vegetation quadrats were spaced 25m apart along the transect. Paired vegetation quadrats were measured on either side of the pitfall traps adjacent to the transect and averaged.

Supplementary Table 3. Ant richness and abundance results for 2011, 2012 and combined. Dashes indicate no data (sites that were sampled in only one of the years).

2011 Ant 2011 Ant 2012 Ant 2012 Ant Total Ant Total Ant Site Richness Abundance Richness Abundance Richness Abundance 1 12 303 10 194 14 497 2 9 175 12 207 13 382 3 10 111 6 89 12 200 4 10 229 8 72 11 301 5 3 161 4 40 5 201 6 9 195 11 217 15 412 7 9 345 7 105 9 450 8 6 129 6 226 8 355 9 11 147 9 140 12 287 10 9 1814 11 706 11 2520 11 12 1128 10 440 14 1568 12 14 369 11 303 15 672 13 9 222 -- -- 9 222 14 9 268 9 182 11 450 15 5 87 -- -- 5 87 16 -- -- 5 187 5 187 17 -- -- 8 175 8 175 18 13 267 8 153 14 420 19 11 650 12 565 14 1215 20 12 343 13 311 14 654 21 -- -- 5 86 5 86 22 11 482 12 559 13 1041 23 8 554 -- -- 8 554

Supplementary Table 4. Ant species found across the 23 sites, abundance, and number of sites.

Number Ant Species Abundance of Sites Lasius neoniger 5543 23 Myrmica americana 1098 21 Ponera pennsylvanica 79 18 Aphaenogaster rudis 489 17 Solenopsis molesta 761 17 Formica pallidefulva 100 17 Tapinoma sessile 509 15 Temnothorax ambiguus 181 14 Monomorium minimum 2293 12 Myrmica latifrons 636 12 Lasius alienus 283 9 Stenamma brevicorne 77 9 Crematogaster cerasi 27 8 Tetramorium caespitum 380 6 Pheidole tysoni 171 6 Formica subsericea 17 5 Lasius umbratus 75 4 Myrmecina americana 10 4 Nylanderia parvula 32 4 Nylanderia faisonensis 41 3 Pheidole pilifera 13 3 Strumigenys clypeata 4 3 Camponotus chromaiodes 4 2 Myrmica pinetorum 15 2 Temnothorax pergandei 22 2 Lasius flavus 44 2 Prenolepis imparis 2 2 Formica rubicunda 1 1 Crematogaster lineolata 2 1 Forelius pruinosus 22 1 Formica pergandei 1 1 Formica integra 3 1

Chapter 2: Host traits and environmental context structure ant-dependent mite communities

1 Abstract Ant nests are stable, long lived, and contain abundant and diverse resources that are often exploited by other organisms. Mites are frequently found in ant nests and also riding on ants in a commensal dispersal relationship, known as phoresy. During non-phoretic stages of their development, ant-associated mites likely rely on soil or nest resources (fungi, decaying material), which may vary depending on host traits and the environmental context of the colony. In this study, we identified the relative contributions of host and environmental factors in structuring ant-associated mite communities in constructed grasslands. Ant-associated mites rely on community assembly of suitable ant hosts, but they may also be affected by patterns of disturbance and soil resource availability during non-phoretic stages. We hypothesized that mite diversity is determined by availability of suitable host species, soil resources and texture, and habitat disturbance. Host species with the most diverse mite assemblages were large-bodied, cosmopolitan, abundant, and frequent within a site. Mite richness and abundance was best predicted by ant richness and grassland area. Best-fitting models for mites associated with Aphaenogaster rudis included litter depth, while Myrmica americana associates were best predicted by M. americana frequency, and grassland age. In multidimensional scaling (MDS) ordinations, 32% of the mite community dissimilarity was explained by ant host species. Distance-based redundancy analysis showed that ant richness, soil texture, and age of the site all contributed to mite community structure (29%). Our results demonstrate that large-bodied, locally abundant, and cosmopolitan ant species are especially important regulators of phoretic mite diversity. In addition we identified the importance of environmental context, especially soil resources, texture and site age, which can alter ant-mite interactions.

2 Introduction The relative importance of abiotic and biotic components in determining the distribution and abundance of organisms is well established for free-living organisms that do not depend on close associations with other species. For organisms such as parasites that depend on hosts for dispersal, food, or reproduction, the ecological, physiological, or life-history characteristics of

hosts may interact with the environmental context to determine the distribution and abundance of parasites. While the host individual is often treated as a conveniently delineated habitat boundary (Pérez-del-Olmo et al., 2009; Guégan et al., 2005), many ectoparasites can move among hosts throughout their life cycle, require multiple host species, and can also be affected by biotic and abiotic constraints of their hosts’ environment (Krasnov et al., 2004). Host traits, distribution, and population density are well studied, significant drivers of host-dependent organism diversity (Lindenfors et al., 2007; Poulin, 2007) and only recently has the environmental context of hosts been examined as a potential factor altering the strength and diversity of species interactions. For example, differences in light availability can shift an ant-plant mutualism to a commensalism (Kersch & Fonesca, 2005), wood-boring pine beetles, dependent on microbial symbionts can have devastating population eruptions under different climate and anthropogenic disturbances (Raffa et al., 2008), tick-borne Lyme disease outbreaks are structured by not only the availability of hosts, but host food resources and climate (Ostfeld et al., 2006), and forest fragmentation and edge effects can weaken a bee mite commensalism (Ewers et al., 2013). The environmental context of a host could alter its competency as a host and thereby affect biodiversity of dependent organisms. Ants structure above and belowground food webs (Dunham & Mikheyev, 2010; Sanders & Platner, 2007) and interact with other organisms as top predators, herbivores, seed dispersers, and mutualists (Eickwort, 1990; Elmes et al., 1998; Howe & Smallwood, 1982; O’Dowd & Hay, 1980). Additionally, the nests of ants harbor a diversity of resources including detritus, middens, small prey species, fungi, ant brood, and hosts that can be exploited by other organisms (Kistner, 1982; Rettenmeyer et al., 2011). The species assemblages associated with ant nests vary by ant species, and ant host traits often determine the diversity of associated species (Campbell et al., 2013). It is less clear if the environmental context of ant hosts can alter the diversity of dependent organisms or its suitability as a host. Organisms living in association with ants, or myrmecophiles, span several phyla, with the large majority consisting of other arthropod groups (Kistner, 1982). Mites are the most frequently encountered and speciose of the myrmecophiles, but are typically overlooked, likely due to their small size and the lack of taxonomic specialists (Kistner, 1979; Rettenmeyer, 1962). Mites are a hyper-diverse group with approximately 50,000 species described and estimates ranging from 0.5-1 million species worldwide (Krantz, 2009). They are frequently found riding

on other animals, especially insects, in a commensal interaction known as phoresy. The purpose of phoresy by mites is believed to be dispersal or transportation to specific isolated resources that would be difficult for a mite to access on its own (Houck & OConnor, 1991). The exact cues that stimulate phoretic behavior are not well known; however, it may be seasonal, or related to the condition of the habitat such as overcrowding, humidity, or resource availability (Houck & OConnor, 1991; Kaliszewski et al., 1995). Most ant-associated mites likely rely directly on the host for only a portion of their life cycle and use soil or ant nest resources (e.g. detritus, fungi, bacteria, and small prey) during non- phoretic life stages (Eickwort, 1990). Although little is known about the ecology or function of the mites when they are not on the host, there is evidence that many ant associated phoretic mite species are host-specific (Campbell et al., 2013) and have evolved synchronized life cycles with their hosts to optimize their dispersal abilities (Kaliszewski et al., 1995; Moser & Blomquist, 2011; Uppstrom & Klompen, 2011). Diverse host characteristics can be important drivers for myrmecophile richness. Large colonies, cosmopolitan hosts, and concentrated colonies are known to influence myrmecophilous beetle species richness (Päivinen et al., 2003; Päivinen et al., 2004). Only recently have studies identified importance of life history characteristics of ant hosts (colony size, host size, and social parasitism) for phoretic mite richness and prevalence (Campbell et al., 2013). The majority of the ant-associated mites are not believed to be parasites based on mouthpart morphology, behavioral observations, or known biology of closely related species, but their dependence on other species for dispersal and resources are characteristics applicable to the existing ecological frameworks of parasitology. For example, large-bodied ant species host a greater diversity of phoretic mite species (Campbell et al., 2013), a relationship that is mirrored by the parasite ecology literature (Lindenfors et al., 2007; Poulin, 2007). Other established principles in parasite ecology may also apply to phoretic mite communities, such as the association with greater parasite diversity and specialization on wide ranging host species with dense populations (Harris & Dunn, 2010; Lindenfors et al., 2007). Parasite diversity has been suggested to indicate trophic complexity and habitat quality because parasites are relying on the presence of host species (Hudson et al., 2006). Similarly, phoretic mite diversity may be indicative of complex soil food webs or host diversity.

Host species distribution and life history characteristics are well known determinants of parasite and phoront diversity. As with many ectoparasites, however, only a portion of the phoront’s life cycle is spent on the host, and the environmental context of the interaction may play a significant role in structuring phoretic mite community composition and diversity. Few studies have tested how phoront or parasite diversity is affected by environmental characteristics or disturbance of the habitat. Ewers et al. (2013) found bumblebees in larger forest fragments, higher in the canopy, and farther from the edge had higher phoretic mite loads (number of mites per host individual). A second study of fragmentation focused on burying beetles (Nicrophorus) and mutualistic Poecilochirus mites (Gibbs & Stanton, 2001). Poecilochirus in moderate loads are beneficial because they control fly populations that compete with Nicrophorus larvae, but in large numbers they begin to prey on the larvae. In fragmented forest sites, Poecilochirus mite loads were more often too low to be beneficial or too high, shifting to detrimental levels. A third study of parasitic gamasid mites specifically tested the relative roles of host and environment, and found that parasitic mite abundance is primarily influenced by host identity and temperature, while host identity and precipitation contributed to species richness (Krasnov et al., 2008). Together these three studies indicate the importance of disturbance, habitat area, edge effects, and abiotic factors on the strength of the host-mite interactions, and potential implications for mite diversity that go beyond host traits. Our study was conducted in constructed grasslands on retired agricultural fields. Grasslands were planted with similar mixtures of native warm-season grasses and forbs, but varied in size, time since planting, and management such as fire or mowing. Over time, these semi-natural grasslands support diverse assemblages of above and below-ground arthropods, including ants (Chapter 1). Ant-associated mites depend on ant community assembly in constructed grasslands, but they may also show similarities to free living soil mite communities that are largely driven by patterns of disturbance and soil resource availability. The following belowground resources and soil characteristics change as constructed grasslands mature: bulk density decreases and aggregate stability, carbon, nitrogen, soil organic matter, microbial biomass, and fungal hyphae increase (Baer et al., 2002; Baer et al., 2000, Karlen et al., 1999; McLauchlan et al., 2006). These changes could influence non-phoretic stages of ant-associated mites. In restored grasslands, soil microarthropod (mite and Collembola) diversity is also affected by periodic fire disturbance, litter depth, and age of the restoration. In the short term,

periodic burning decreases litter and soil moisture while increasing soil temperatures, root production, and microbial activity (Seastedt, 1984). The boost in resources from fire can temporarily increase microarthropod density (Lussenhop, 1976); however, the belowground resources are gradually depleted and microarthropod density decreases until the litter layer accumulates (Seastedt, 1984). Change can occur over long time scales as demonstrated by Collembola communities in constructed grasslands that assemble over decades and slowly begin to resemble remnant grassland communities (Brand & Dunn, 1998). The purpose of this study is to identify the role of host and environmental characteristics on ant-associated phoretic mite richness, abundance, and community composition. Specifically we test three hypotheses: (i) Mite diversity is determined by availability of suitable host species and soil resources (e.g. organic matter, litter depth) within the grassland. First we predicted that overall phoretic mite diversity, prevalence, and load would be greatest in grassland patches with higher abundance and richness of suitable ant hosts. Second we expected increased mite abundance, richness, prevalence, and load in grasslands where preferred host ant species are more frequent. Third, since many mites are associated with soil resources, we predicted that sites with more soil organic matter (SOM) and litter depth would have higher mite abundance, richness, prevalence, and load. (ii) Phoretic mite communities are structured by the habitat and soil disturbances. We predicted that grassland sites that were smaller, had greater edge:area ratios, and were more recently planted would have lower mite abundance, richness, prevalence, and load. We predicted that recently burned sites will have lower mite diversity because of decreases in depth of the litter layer. (iii) Mite communities are structured by variation in soil texture due to its effect on soil pore size, moisture, bulk density, and relative host frequencies. We predicted that mites will show patterns similar to their host ants (Chapter 1), such that diversity is higher in sandier sites.

3 Methods 3.1 Study Sites Our study sites comprised 23 warm season constructed grasslands established and managed primarily by local landowners and MetroParks in Butler, Montgomery and Preble Counties of southwestern Ohio as previously described in Chapter 1 (see also Chapter 1, Table 1

and Supplementary Table 1 for site characteristics). Twenty sites were used in each of the two summers of 2011 and 2012 due to changes in voluntary landowner participation between years.

3.2 Ant and Mite Collections To assess ant species richness at each of the 23 sites, we used pitfall traps (Chapter1). Although pitfall traps are widely used to sample ant communities, ants collected in pitfall traps cannot be reliably used for phoretic mite collections because most mites become dislodged from the ants when collected with fluid traps, and traps also collect many other arthropods that may be carrying their own assemblages of phoretic mites. We therefore developed a baiting technique to collect the ants used for phoretic mite sampling. Baits tend to be unreliable for sampling the full spectrum of the ant community, because they favor more common ants with generalist diets (Albrecht & Gotelli, 2001; Hahn & Wheeler, 2002); however, for this study we were less interested in the total diversity of mites associated with all ant species, but rather how the diversity of mites on more suitable and common host species changes depending on habitat characteristics, disturbance, and suitable host frequency. While developing our baiting method, we found that the first ants to typically find the baits are large and highly vagile ants, which are later displaced by large numbers of small ants. To collect the greatest diversity of ants with baits, it is therefore optimal to sample over multiple time intervals with both carbohydrate and protein baits (Albrecht & Gotelli, 2001; Hahn & Wheeler, 2002). We used the same transect containing the pitfall traps for our bait stations, and established 10-20 stations (number of baits and pitfalls was log scaled to area) spaced 8.3m away from each side of the pitfall trap (Supplementary Figure 1). Baiting took place for an hour at each site during the same week the pitfall traps were active. The baits consisted of 3”x5” index cards marked with three time intervals (20 min, 40 min, and 60 min), onto which a small amount of fish flavored canned cat food and crushed pecan sandies were placed. A 20 min bait loaded card was initially placed on bare ground at each station along the transect. After 20 min the ant covered baits were placed into individual plastic zip bags and replaced with the 40 min bait loaded card. This process was repeated for the 60 min card. Bait cards with ant collections were frozen until inspection for mites. A total of 930 and 924 baits were used for each sampling period in 2011 and 2012, respectively. Bait collections were analyzed in the laboratory by inspecting each ant for phoretic mites under a dissecting microscope. All mites associated with the ants were placed in lactic acid to 44

clear away internal structures and mounted on slides in Hoyer’s or polyvinyl alcohol (PVA) media. The majority of ant-associated mites are undescribed; therefore, mites were identified to genus and morphospecies, and whenever possible, to species. Voucher specimens of each mite and ant species will be deposited at the Ohio State Acarology Collection.

3.3 Environmental variables As measures of potential soil food resources, we used the average depth of the litter layer near each bait station and SOM (methods described in Chapter 1). We used the following variables in our statistical analysis as measures of disturbance: age (time since planting) and time since burn (management), and additional measures of habitat including: ratio of edge to area, area, soil texture, and bulk density of the soil. 3.4 Statistical Analysis We constructed models for total mite richness and mite abundance at the site level using environmental and host predictors. To control for variation in host species characteristics, we tested mite richness, abundance, prevalence, and mite load for mite assemblages associated with two common host species, Aphaenogaster rudis and Myrmica americana. Mite prevalence was calculated as the number of hosts with mites divided by the total number of hosts collected on baits at a site (Margolis et al., 1982). Average mite load (number of mites per host individual across a site) was calculated in two ways. Other studies of mite loads have used the number of mites collected at a trap/total number of hosts per trap, due to mites falling from the host in fluid traps (Ewers et al., 2013; Gibbs & Stanton, 2001). This method greatly dilutes the average mite load because any hosts that did not carry mites are still included in the calculation. To compare our results with those of other studies, we used this less accurate method and designate it “trapload”. Additionally, we tested the true mite load, “trueload,” calculated as the number of mites collected at a trap divided by the total number of hosts with mites at the trap. All analyses were conducted at the site level and separately for 2011 and 2012 data except for the A. rudis and M. americana mite analyses due to low sample sizes. All abundance variables were log transformed prior to analyses. We also conducted analyses in which we separated total mite richness and abundance into two major mite taxonomic groups: Cohorts Astigmata and Heterostigmata. These two taxa differ in their phoretic stage, host specificity, and sensitivity to habitat and resources (Campbell

et al., 2013). Astigmata associated with these ant species are phoretic as deutonymphs (an immature stage of development) and are believed to primarily eat bacteria and fungi in non- phoretic stages (Houck & OConnor, 1991). Heterostigmata associated with these ant species, enter phoresy as specialized adult females and primarily eat fungi when not on the host (Binns, 1982). Previous studies found that Heterostigmata are more host specific with 61% associated with a single ant species and 35% of the remaining species on hosts in the same genus, while 40% of Astigmata species are associated with a single ant species and 44% of the remaining species were associated with multiple congeneric ant hosts (Campbell et al., 2013). We therefore expected Astigmata abundance and richness to be primarily determined by environmental characteristics (disturbance or soil resources) rather than ant richness within a site, Heterostigmata richness to be determined by ant richness and host frequency, and Heterostigmata abundance to be determined by disturbance or soil resources. We compared host suitability of ant species using generalized linear models with host size and abundance (individuals inspected) as predictors for associated mite richness. We conducted this analysis using pooled data across all sites and only included ant species for which at least 30 individuals had been inspected. We removed Monomorium minimum from the analysis due to unusually large sample sizes (>80,000 individuals) and very low mite prevalence (<0.02%). Previous studies have shown that larger hosts support greater mite diversity and number of hosts inspected can be an important measure of sampling effort (Campbell et al., 2013). Average host size was extracted from Coovert (2005). We used generalized linear models (glm function, R) with a Poisson error distribution for all richness models and Gaussian error distributions for log-transformed abundance and average mite load models. Mite prevalence was analyzed with logit models using the (glm function, R) with a binomial error distribution and a logit link function. Potential variables for models included: ant species richness (pitfall traps), soil texture, SOM, bulk density, litter depth, age, time since burn, area, and edge:area. We log-transformed age, time since burn, and area predictors. There were more baits used at larger sites, and we therefore wanted to verify that any relationships with area as a predictor were not impacted by sampling intensity. We conducted a subsampling of baits in which only 10 baits (the number used in the smallest sites) were randomly selected 1000 times. The species abundance values for each draw were averaged from the randomizations, and summarized for richness at each site. Mite species richness of all

subsampled sites remained unchanged except for a single site that lost a rare species; therefore, we continued to use the non-subsampled data for mite richness in subsequent analyses. Best models were selected using minimum Akaike’s Information criterion (AIC). Any models differing by 2 or less AIC points were considered competing models. We tested best- fitting models against the null model in a likelihood ratio test to calculate p-values and percent deviance explained.

3.5 Community Composition Analyses To test if shifts in ant and mite community composition were correlated among sites, we used a Mantel test, which compares dissimilarity between two matrices (mantel function, vegan package) (Oksanen et al., 2013). If there is a correlation in community dissimilarity, this can mean either that these two communities responded to each other or that they covaried in species composition due to a common environmental factor(s). The Mantel test does not partition the variance in the communities; therefore, we used ordinations to test the role of host identity with Multidimensional Scaling (MDS) and specific environmental variables with distance-based redundancy analysis (dbRDA) (McArdle & Anderson, 2001). We used MDS (mds function, vegan package) with Bray-Curtis dissimilarity (vegdist function, vegan package, R) (Oksanen et al., 2013) using mite communities (41 mite species total) associated with seven common ant species that carried the majority of the mites (97.5% of mite abundance). Ant species were only included if they carried mites at a minimum of four sites. We used a permutational ANOVA to test host species as a predictor with site as stratum (adonis function, vegan package) (Oksanen et al., 2013). Environmental variables were tested as predictors of variation in mite species composition among patches using dbRDA with Bray-Curtis dissimilarity (McArdle & Anderson, 2001). We used AIC to select the best fitting ordination model and obtained p-values using random permutations (999 permutations). The dbRDA was conducted with a user-written function in R (M. Anderson, personal communication).

4 Results A total of 104,766 ants belonging to 27 ant species were collected with baits and individually inspected for phoretic mites (Supplementary Table 1). Monomorium minimum, a small competitive ant species that overwhelmed the baits, comprised 82,521 (78%) of the ants and mite prevalence was only 0.02% (16 ants with mites). A total of 53 species of mites (1586

47 individuals) were taken from 15 ant host species (1047 ant individuals) (Supplementary Table 2, Supplementary Table 3). Mite richness included 20 Astigmata species (1201 individuals), 27 Heterostigmata species (377 individuals), and 4 Mesostigmata species (6 individuals). Average mite richness and abundance per site was 7.6 species (max=16, min=0) and 68.9 individuals (max=208, min=0), respectively. There was an average of 6.6 species mite species associated with each ant species (excluding M. minimum); however, some ant species hosted much greater diversity such as Myrmica americana with 18 species and 937 mite individuals (Supplementary Table 1). We used a species accumulation curve to summarize the observed and Chao estimate (Chao et al., 2005) of mite species for seven common ant hosts that carried 97.5% of the total mite abundance and 77.4% of the total richness (Figure 1). Ant species that were more cosmopolitan (at more sites) and more abundant tended to have higher observed and estimated species richness. The best model for predicting mite richness among host species (host suitability) included host body size and host abundance (p<0.0001, Dev. expl.=63.0%) . Pitfall ant collections, used for ant species richness in our analysis, comprised 32 ant species including 8 species not collected by baiting (Chapter 1, Supplementary Table 4). These 8 additional species were uncommon or rare species collected in very low numbers and likely would not have contributed significantly to mite diversity.

4.1 Overall Mite Species Richness and Abundance The best fitting model for 2011 mite species richness was ant richness (p=0.001, Dev. expl.=19.3%, df=1,18) (Figure 3a, Table 1). A competing model included area in addition to ant richness (ΔAIC=-1.39, Dev. expl.= 25.5%, df=2,17). The best model for 2012 was area (p=0.019, Dev. expl.= 18.8%, df=1,18) (Figure 3b) and a competing model included ant richness (ΔAIC=-0.11, Dev. expl.=21.4 % df=2,17). Total mite abundance for 2011 was best predicted by ant richness (p=0.0003, Dev. expl.= 41.8%, df=1,18) (Figure 4) and a competing model also included age (ΔAIC=+1.15, Dev expl.= 41.8 %, df=2,17). In 2012, there were no models that were better than the null model.

4.2 Richness and Abundance of Heterostigmata and Astigmata Astigmata richness in 2011 was best predicted by area (p=0.017, Dev. expl.= 22.0%, df=1,18) (Figure 5) and there were multiple competing models including ant richness as well as

age (Table 1). The 2012 best model was the null model and the only competing model was litter depth, which explained very little and was not significant (p=0.169, Dev. expl.= 5.2%, df=1,18). Heterostigmata richness in 2011 was best predicted by ant richness (p=0.028, Dev. expl.= 11.8% , df=1,18), with multiple competing models, and in 2012 area was the best model (p=0.02, Dev. expl.= 24.4%, df=1,18) (Figure 4) with area and ant richness as a competing model. Abundance for both Astigmata (p=0.004, Dev. expl.= 31.0%, df=1,18) and Heterostigmata (p=0.002, Dev. expl.= 34.7% , df=1,18) was best predicted by ant richness in 2011 (Figure 6) and competing models for both also include age (Table 1). In 2012 there were no models better than the null for abundance of either mite taxon.

4.3 Mites on Common Host Species To understand how resources and disturbance affect mite assemblages, we used a subset of the mite community associated with M. americana and A. rudis. Mite prevalence was 13.4% (binomial SE= 7.1%) on M. americana and 6.8% (binomial SE = 5.7%) on A. rudis. There was no predictor that was better than the null model for prevalence of mites for either species and no competing models (Table 2). Richness and abundance of mites associated with A rudis was best explained by litter depth (p=0.0003, Dev. expl. 31.6% and p=0.003, Dev. expl. 33.2%, df=1,17, respectively) (Figure 7a & b). Competing models included host frequency and ant richness (Table 2). Mite richness and abundance on M. americana was best explained by host frequency and age (p=0.0003, Dev. expl. 31.6% and p=0.003, Dev. expl. 33.2%, df=2,20, respectively) and there were no competing models (Figure 8a & b, Table 2). Trueload for M. americana was best predicted by host frequency in 2012 (p=0.042, Dev. expl.= 21.6%, df=1,15) (Figure 8), and the null model in 2011. Due to low mite prevalence the majority of hosts at a trap did not carry mites, and values of trapload were typically less than 1 (mean= 0.69, and 0.45 in 2011 and 2012, respectively). For both 2011 and 2012, there were no predictors that improved upon the null model (Table 2). We were unable to conduct the trueload analysis for A. rudis because most traps (42 of 51 traps) had ants with only a single mite per individual.

4.4 Mite Community Composition The significant positive relationship between the ant and mite community dissimilarity (p=0.001, Mantel r= 0.51) suggests that ant and mite communities have similar turnover in species composition among sites (Figure 10). Host ant species explained a significant amount of

the variance in MDS multivariate ordinations of mite communities for seven host species (p=0.001, Var. expl.= 31.8%). Host specificity by mites at both species and genus levels is indicated by clustering in multivariate space (Figure 10). The best fitting model for mite community composition using dbRDA included ant richness, age, and percent sand (Pseudo F=2.50, p=0.001, R2=0.29) (Figure 11).

5 Discussion 5.1 Host Identity vs. Environment The objective of this study was to determine how the environmental context of ant hosts may affect their role as regulators of mite diversity. Using general linear models and community composition analyses, we found that both host identity and multiple environmental characteristics play significant roles in structuring ant-associated mite diversity and community composition in grasslands.

5.2 Host Suitability, Frequency, and Richness Ant-associated mite diversity is dependent on host suitability and frequency in constructed grasslands. We documented phoretic mites on 15 ant species in the grasslands; however, 12 ant species did not carry any mite species. It is clear that some ant species in the grassland are more suitable hosts than others, but what makes a given ant species a more preferred host? It is already established in this phoretic mite system and multiple parasite systems, that larger bodied hosts have higher associated species richness (Campbell et al., 2013; Lindenfors et al., 2007; Poulin, 2007). We provided additional evidence of this relationship with host size (Figure 2), and also found that within- (Figure 8, Figure 9) and among-patch (Figure 1) host frequencies are also important drivers of phoretic mite diversity. The latter findings support established principles in parasite ecology including greater parasite diversity on wide ranging host species with dense populations (Harris & Dunn, 2010; Lindenfors et al., 2007). Ant richness was the best predictor for total mite richness and abundance, Astigmata abundance, and Heterostigmata richness and abundance (Figure 3a, Figure 4, Figure 6). Interestingly, mites were not found on the less common ant species that contribute to higher ant species richness at a given site, instead common ants seem to be accumulating more species associations in sites with higher ant richness. This suggests that both mite and ant richness may be exhibiting covarying responses to environmental factors, for example grassland age and soil

50 texture (Chapter 1). Similarly, the Mantel test of ant and mite community dissimilarities showed a strong relationship in turnover of these two communities (Supplementary Figure 2), and the distance-based Redundancy Analysis provided support for the importance of ant richness, soil texture, and age for partitioning variance in composition of the mite communities (Figure 11).

5.3 Disturbance, Soil Resources, and Patch Characteristics The grassland patches in our study varied in soil texture, soil organic matter, litter depth, and degree of disturbance. Area routinely appeared in best or competing models for mite richness and abundance (Table 1). As in other phoretic mite studies, smaller habitat area decreased diversity (Ewers et al., 2013) (Figure 3b, Figure 5); however, we did not find a relationship with mite loads (Gibbs & Stanton, 2001). Species-area relationships are well established in ecological theory across an array of animal species (Watling & Donnelly, 2006). Species-area relationships with mites, although very limited in number, are also documented for free-living taxa (Giller, 1996; Lindo & Winchester, 2007). Shape of the habitat (edge:area) was not an important predictor of phoretic mite diversity in this study. Our findings also provide support for successional turnover in grassland communities as was previously found in grassland ants (Chapter 1) and Collembola (Brand & Dunn, 1998) (Figure 11, Figure 8). For mites associated with a single host, M. americana, richness and abundance of mites increased as the grassland age. Time since burn, another disturbance measure did not appear in any of the best models and seems to play little role in ant-associated mite diversity. Soil texture was an important predictor for mite community composition among sites (Figure 11), but did not appear in any of the best models for abundance, richness, prevalence, or mite load. Ant colonies may be long lived, and over time nests become highly altered environments with lower bulk density, SOM, and differing structure from the surrounding soils where we obtained the soil cores. It is possible that species-specific modifications to the nest soil could be more important to mites than the surrounding soil characteristics. The role of soil resource quantity (i.e. litter depth, SOM) is not apparent for mite richness and abundance when analyzed for all mite species in this study; however, mites associated with A. rudis showed increased richness and abundance in sites that had greater litter depth (Figure 7). Mites associated with M. americana were not related to litter depth, and instead were driven by host frequency and age of the habitat. A potential reason for this may be that the majority of the mites associated with A. rudis were Heterostigmata (70%) while mites associated with M. 51

americana were primarily Astigmata (94%). We predicted Astigmata abundance and richness to be related primarily to soil resource characteristics because they are less host-specific, while Heterostigmata richness was predicted to be determined by ant richness or host frequency, and Heterostigmata abundance to be determined by soil resource characteristics. Our results for mites on these two ant species did not support our predictions for richness and, in fact, showed the reverse: richness in Astigmata-dominated communities (M. americana mites) is primarily determined by host frequency and time since planting (age), while richness in Heterostigmata dominated communities is driven by litter depth (Table 2). Across all ant species, Astigmata richness (2011) and Heterostigmata richness (2012) were determined by grassland area (Figure 5), while abundance of both taxa and Heterostigmata richness in 2011 were determined by ant richness (Figure 6). The strong relationship with ant richness is likely due to clear patterns of host specificity (68% of the mites in our study believed to be host specific) and importance of host identity for phoretic mite community composition that override soil resource dependence (Figure 10). It remains unclear as to why particular host species have such different mite communities associated with them, but this may be due to coevolution or differences in microclimate, fungi and bacterial resources, or other ant-dependent organisms within the nests.

5.4 Conclusions Our study demonstrates that large bodied, locally abundant, and cosmopolitan ant species are especially important regulators of phoretic mite diversity. We also find that the environmental context of the host, especially deeper litter layers and older sites, can influence mite diversity and community composition. Mite and ant community composition are both structured by site age and soil texture (Chapter 1), and ant richness is often predictive of mite richness providing support for ants as mite biodiversity indicators in constructed grasslands. Ant- associated mite communities represent a potential model system for understanding both coevolution and assembly processes that occur in spatially isolated patches.

6 References Albrecht, M., & Gotelli, N. J. (2001). Spatial and temporal niche partitioning in grassland ants. Oecologia, 126(1), 134–141. Baer, S. G., Kitchen, D. J., Blair, J. M., & Rice, C. W. (2002). Changes in ecosystem structure and function along a chronosequence of restored grasslands. Ecological Applications, 12(6), 1688–1701. Baer, S. G., Rice, C. W., & Blair, J. M. (2000). Assessment of Soil Quality in Fields with Short and Long Term Enrollment in the CRP. Journal of Soil and Water Conservation. Binns, E. S. (1982). Phoresy as migration - some functional aspects of phoresy in mites. Biological Reviews, 57, 571–620. Brand, R. H., & Dunn, C. P. (1998). Diversity and Abundance of Springtails (Insecta: Collembola) in Native and Restored Tallgrass Prairies. The American Midland Naturalist, 139(2), 235–242. Campbell, K. U., Klompen, H., & Crist, T. O. (2013). The diversity and host specificity of mites associated with ants: the roles of ecological and life history traits of ant hosts. Insectes Sociaux, 60(1), 31–41. Chao, A., Chazdon, R. L. Colwell, R. K., & Shen, T. (2005). A new statistical approach for assessing similarity of species composition with incidence and abundance data. Ecology Letters, 8,148-159. Coovert, G. A. (2005). The Ants of Ohio. Columbus, Ohio: Ohio Biological Survey, Inc. Dunham, A. E., & Mikheyev, A. S. (2010). Influence of an invasive ant on grazing and detrital communities and nutrient fluxes in a tropical forest. Diversity and Distributions, 16(1), 33–42. Eickwort, G. C. (1990). Associations of mites with social insects. Annual Review of Entomology, 35, 469–488. Elmes, G. W., Thomas, J. A., Wardlaw, J. C., Hochberg, M. E., Clarke, R. T., & Simcox, D. J. (1998). The ecology of Myrmica ants in relations ot the conservation of Maculinea butterflies. Journal of Insect Conservation, 2, 67–78. Ewers, R. M., Bartlam, S., & Didham, R. K. (2013). Altered species interactions at forest edges: contrasting edge effects on bumble bees and their phoretic mite loads in temperate forest remnants. Insect Conservation and Diversity, 6(5), 598–606. Gibbs, J. P., & Stanton, E. J. (2001). Habitat fragmentation and arthropod community change: Carrion beetles, phoretic mites, and flies. Ecological Applications, 11(1), 79–85. Giller, P. S. (1996). The diversity of soil communities, the “poor man’s tropical rainforest.” Biodiversity and Conservation, 168, 135–168. Hahn, D. A., & Wheeler, D. E. (2002). Seasonal foraging activity and bait preferences of ants on Barro Colorado Island, Panama. Biotropica, 34(3), 348–356. Harris, N. C., & Dunn, R. R. (2010). Using host associations to predict spatial patterns in the species richness of the parasites of North American carnivores. Ecology Letters, 13(11), 1411–1418. Houck, M. A., & OConnor, B. M. (1991). Ecological and evolutionary significance of phoresy in the Astigmata. Annual Review of Entomology, 36, 611–636. Howe, H. E., & Smallwood, J. (1982). Ecology of Seed Dispersal. Annual Review of Ecology and Systematics, 13, 201–228. Hudson, P. J., Dobson, A. P., & Lafferty, K. D. (2006). Is a healthy ecosystem one that is rich in parasites? Trends in Ecology and Evolution, 21(7), 381–385.

Kaliszewski, M., Athias-Binche, F., & Lindquist, E. E. (1995). Parasitism and parasitoidism in Tarsonemina (Acari: Heterostigmata) and evolutionary considerations. Advances in Parasitology, 35, 335–367. Karlen, D. L., Rosek, M. J., Gardner, J. C., Allan, D. L., Alms, M. J., Bezdicek, D. F., … L, S. M. (1999). Conservation Reserve Program effects on soil quality indicators. Soil and Water Conservation Society. Kersch, M. F., Fonesca, C. (2005). Abiotic factors and the conditional outcome of an ant-plant mutualism. Ecology, 86(8), 2117-2126. Kistner, D. H. (1979). Social and evolutionary significance of social insect symbionts. In H. R. Hermann (Ed.), Social insects (Vol. 1, pp. 340–413). New York: Academic Press. Kistner, D. H. (1982). The social insects beastiary. In H. R. Herman (Ed.), Social Insects (pp. 2– 244). New York: Academic Press. Krantz, G. W. (2009). Introduction. In G. W. Krantz & D. E. Walter (Eds.), A Manual of Acarology (pp. 1–2; 1). Lubbock, TX: Texas Tech University Press. Krasnov, B. R., Shenbrot, G. I., Khokhlova, I. S., & Degen, a. A. (2004). Relationship between host diversity and parasite diversity: flea assemblages on small mammals. Journal of Biogeography, 31(11), 1857–1866. Krasnov, B. R., Korallo-Vinarskaya, N. P., Vinarski, M. V, Shenbrot, G. I., Mouillot, D., & Poulin, R. (2008). Searching for general patterns in parasite ecology: host identity versus environmental influence on gamasid mite assemblages in small mammals. Parasitology, 135(2), 229–242. Lindenfors, P., Nunn, C. L., Jones, K. E., Cunningham, A. a., Sechrest, W., & Gittleman, J. L. (2007). Parasite species richness in carnivores: Effects of host body mass, latitude, geographical range and population density. Global Ecology and Biogeography, 16(4), 496–509. Lindo, Z., & Winchester, N. N. (2007). Local-regional boundary shifts in oribatid mite (Acari: Oribatida) communities: Species-area relationships in arboreal habitat islands of a coastal temperate rain forest, Vancouver Island, Canada. Journal of Biogeography, 34(9), 1611– 1621. Lussenhop, J. (1976). Soil arthropod response to prairie burning. Ecology, 57(1), 88–98. Margolis, L., Esch, G. W., Holmes, J. C., Kuris, A. M., & Schad, G. A. (1982). The use of ecological terms in parasitology (report of an ad hoc committee of the American Society of Parasitologists). Journal of Parasitology, 68, 131–133. McArdle, B. H., & Anderson, M. J. (2001). Fitting multivariate models to community data: a comment on distance-based redundancy analysis. Ecology, 82(1), 290–297. McLauchlan, K. K., Hobbie, S. E., & Post, W. M. (2006). Conversion from agriculture to grassland builds soil organic matter on decadal timescales. Ecological Applications, 16(1), 143–53. Moser, J. C., & Blomquist, S. R. (2011). Phoretic arthropods of the red imported fire ant in central Louisiana. Annals of the Entomological Society of America, 104, 886–894. O’Dowd, D. J., & Hay, M. E. (1980). Multalism between Harvester Ants and a Desert Ephemeral : Seed Escape from Rodents. Ecology, 61(3), 531–540. Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O’Hara, R. B., … Wagner, H. (2013). vegan: Community Ecology Package. R package version 2.0-10. Retrieved from http://cran.r-project.org/package=vegan

Ostfeld, R. S., Canham, C. D., Oggenfuss, K., Winchcombe, R. J., Keesing, F. (2006). Climate, deer, rodents, and acorns as determinants of variation in Lyme-disease risk. PLoS Biology, 4, e145. Päivinen, J., Ahlroth, P., Kaitala, V., Kotiaho, J. S., Suhonen, J., & Virola, T. (2003). Species richness and regional distribution of myrmecophilous beetles. Oecologia, 134, 587–595. Päivinen, J., Ahlroth, P., Kaitala, V., & Suhonen, J. (2004). Species richness, abundance and distribution of myrmecophilous beetles in nests of Formica aquilonia ants. Annales Zoologici Fennici, 41, 447–454. Pérez-del-Olmo, A.,Fernández, M., Raga, J. A., Kostadinova, A., Morand, S. (2009). Not everything is everywhere: the distance decay of similarity in a marine host-parasite system. Journal of Biogeography, 36, 200-209. Poulin, R. (2007). Are there general laws in parasite ecology? Parasitology, 134(Pt 6), 763–776. Raffa, K. F., Aukema, B. H., Bentz, B. J., Carroll, A. L., Hicke, J. A., Turner, M. G., Romme, W. H. (2008). Cross-scale drivers of natural disturbances prone to anthropogenic amplification: the dynamics of bark beetle eruptions. BioScience, 58, 501-517. Rettenmeyer, C. W. (1962). Arthropods associated with neotropical army ants with a review of the behavior of these ants (Arthropoda; Formicidae: Dorylinae). University of Kansas, Lawrence, KS. Rettenmeyer, C. W., Rettenmeyer, M. E., Joseph, J., & Berghoff, S. M. (2011). The largest animal association centered on one species: the army ant Eciton burchellii and its more than 300 associates. Insectes Sociaux, 58, 281–292. Sanders, D., & Platner, C. (2007). Intraguild interactions between spiders and ants and top-down control in a grassland food web. Oecologia, 150(4), 611–24. Seastedt, T. R. (1984). Microarthropods of Burned and Unburned Tallgrass Prairie. Journal of the Kansas Entomological Society, 57(3), 468–476. Uppstrom, K. A., & Klompen, H. (2011). Mites (Acari) associated with the desert seed harvester ant, Messor pergandei (Mayr). Psyche, 2011, 1–7. Watling, J. I., & Donnelly, M. a. (2006). Fragments as islands: A synthesis of faunal responses to habitat patchiness. Conservation Biology, 20(4), 1016–1025.

Figure 1. Species accumulation curve of mite species by number of hosts inspected. Number of points on a curve represents the number of sites where the ant host (species name in italics) was collected, while length of the curve represents the number of ant individuals inspected. Ant species that are more cosmopolitan (at more sites) and more abundant also have higher observed (Obs) and estimated (Est) species richness.

Figure 2. Best fitting model for mite species richness associated with seventeen different ant species included host size and host abundance. Note: X axis is shown on log-scale.

Table 3. Best and competing models (ΔAIC<2) for Total Mite Richness and Abundance and Richness and Abundance for Astigmata and Heterostigmata groups.

ΔAIC ΔAIC vs. % Dev. vs. % Dev. Response Best Model Null Expl. Competing Models Best Expl.

Total Mite Richness 2011 (+)Ant richness -8.72 19.3% (+)Ant richness + (+)Log (area) -1.39 25.5%

2012 (+)Log (area) -3.46 18.8% (+)Ant richness + (+)Log (area) +0.11 21.4% Total Mite Abundance 2011 (+)Ant richness -8.823 41.8% (+)Ant richness + (+)Log(age) +1.15 41.8%

2012 Null ------Astigmata Richness 2011 (+)Log (area) -3.68 22.0% (+)Log (area) + (+)Ant richness -0.10 30.1% (+)Ant richness +0.79 18.9% (+)Log (age) +1.60 15.8%

2012 Null -- -- (+)Litter depth +0.11 5.2% Astigmata Abundance 2011 (+)Ant richness -5.42 31.0% (+)Ant richness + (+)Log (age) -0.25 38.3% (+)Log (age) +1.24 26.6%

2012 Null -- -- (+)Log (area) -0.03 9.6% Heterostigmata Richness 2011 (+)Ant richness -2.81 11.8% (+)Ant richness + (+)Log (area) -0.05 16.8% (+)Log (area) +0.32 11.0%

2012 (+)Log (area) -3.42 24.4% (+)Log (area) + (+)Ant richness +1.56 24.5% Heterostigmata Abundance 2011 (+)Ant richness -6.52 34.7% (+)Ant richness + (+)Log(age) +1.73 35.6%

2012 Null ------

Table 4. Best and competing models (ΔAIC<2) for Prevalence, Richness, Abundance, and Load for mites associated with Aphaenogaster rudis and Myrmica americana.

ΔAIC ΔAIC vs. % Dev. vs. % Dev. Response Best Model Null Expl. Competing Models Best Expl. Prevalence rudis Null ------americana Null ------Richness rudis (+)Litter depth -10.90 31.6% (+)Litter depth + (+)Host Frequency -1.32 39.8%

americana (+)Host Frequency + (+)Log (age) -25.61 55.4% ------Abundance rudis (+)Litter depth -5.64 33.2% (+)Litter depth + (+)Ant richness -0.81 42.4%

americana (+)Host Frequency + (+)Log (age) -18.44 62.3% ------Trueload americana 2011 Null -- -- (+)Litter depth +0.18 11.4%

2012 (+)Host Frequency -2.14 21.6% (+)Time since burn +1.14 16.1% Trapload americana 2011 Null -- -- (-)Time since burn -1.37 20.2%

2012 Null -- -- (+)Log(area) -1.37 18.0% (+)Time since burn -0.29 12.5%

Figure 3. Best models for overall mite species richness for a) 2011- ant richness and b) 2012- grassland area (X axis log-scaled).

60 Figure 4. The best model for phoretic mite abundance included ant richness. Y axis is presented on a log scale.

61 Figure 5. Best fitting model for 2011 Astigmata (solid line, filled circles) and 2012 Heterostigmata richness (dashed line, open circles) included Grassland Area (ha).

62 Figure 6. Best fitting model for 2011 Astigmata (solid line, filled circles) and Heterostigmata abundance (dashed line, open circles) included Ant Richness. Y axis is presented on a log scale.

63 Figure 7. Best fitting models for a) mite richness and b) abundance (log scale) associated with Aphaenogaster rudis included Litter Depth.

64 Figure 8. Best fitting models for a) mite richness and b) abundance (log scale) associated with Myrmica americana included both Host Frequency and Grassland Age. Curve in plot represents model fit when Grassland Age is held constant as the mean value.

65 Figure 9. Best fitting models for 2012 Average Mite Load (Trueload) on Myrmica americana included Host Frequency. Trueload is calculated as the number of mites per bait divided by the number of hosts with mites at the bait.

66 Figure 10. Multidimensional scaling (MDS) ordination of mite community composition. Host was a significant predictor of mite community (p=0.001, Variance Expl. = 31.8%). In addition to separating out by species, mite communities also tended to cluster between ants in the same genus (e.g. Myrmica americana and Myrmica latifrons, Lasius neoniger and Lasius alienus).

67 Figure 11. AIC selected Distance-based Redundancy Analysis (dbRDA) multivariate ordination of mite community composition across 23 grasslands. Symbols indicate site scores of grasslands and are sized according to the most important predictor variable (Ant Richness). Arrows are biplot correlations of the significant predictor variables.

68 Supplementary Table 5. Abundance and richness of associated mites for ant species collected with baits. The horizontal division separates ant species with and without mites. Mite Mite Ant Ant Species Abundance Richness Abundance Monomorium minimum 16 7 82521 Tapinoma sessile 4 4 5673 Myrmica americana 937 18 3693 Lasius neoniger 70 15 2583 Tetramorium caespitum 93 5 2400 Aphaenogaster rudis 151 10 1724 Solenopsis molesta 39 3 1368 Lasius alienus 70 10 1269 Myrmica latifrons 186 8 620 Prenolepis imparis 3 2 336 Formica pallidefulva 11 7 172 Formica subsericea 2 2 37 Formica integra 1 1 19 Camponotus pennsylvanicus 1 1 7 Formica rubicunda 1 1 1 Pheidole tysoni 0 0 1635 Temnothorax ambiguus 0 0 341 Nylanderia parvula 0 0 122 Temnothorax pergandei 0 0 99 Nylanderia faisonensis 0 0 68 Forelius pruinosus 0 0 48 Crematogaster cerasi 0 0 17 Myrmica pinetorum 0 0 6 Pheidole pilifera 0 0 4 Camponotus castaneus 0 0 1 Camponotus chromaiodes 0 0 1 Formica postoculata 0 0 1

69 Supplementary Table 6. Mite richness and abundance for 2011, 2012 and combined. Dashes indicate no data (sites that were sampled in only one of the years).

2011 Mite 2011 Mite 2012 Mite 2012 Mite Total Mite Total Mite Site Richness Abundance Richness Abundance Richness Abundance 1 5 43 1 7 6 50 2 5 106 7 69 9 175 3 5 62 6 11 8 73 4 3 8 3 16 5 24 5 0 0 2 3 2 3 6 12 154 7 54 13 208 7 4 11 4 4 8 15 8 12 49 4 8 12 57 9 8 27 8 122 11 149 10 8 34 5 9 10 43 11 9 32 6 56 10 88 12 12 75 9 21 16 96 13 2 4 -- -- 2 4 14 4 4 4 11 7 15 15 0 0 -- -- 0 0 16 -- -- 4 37 4 37 17 -- -- 5 31 5 31 18 9 78 10 85 12 163 19 2 66 0 0 2 66 20 10 119 6 66 12 185 21 -- -- 4 35 4 35 22 6 11 5 30 9 41 23 7 28 -- -- 7 28 Total 45 911 26 675 53 1586

70 Supplementary Table 3. Mite species collected at the 23 grassland sites.

Number Number of Mite Species Abundance of Sites Host Species Astigmata Bonomoia sp1 1 1 1 Cosmoglyphus sp1 7 5 1 Cosmoglyphus sp2 1 1 1 Cosmoglyphus sp2 3 2 2 Forcellinia sp1 347 16 8 Histiostoma sp1 594 19 7 Histiostoma sp2 1 1 1 Histiostoma sp2 98 12 1 Histiostoma sp4 8 2 1 Histiostoma sp5 1 1 1 Histiostoma sp6 3 2 1 Histiostoma sp8 5 2 1 Histiostoma sp9 1 1 1 Histiostoma sp10 1 1 1 Histiostoma sp11 1 1 1 Schwiebea sp1 106 15 8 Schwiebea sp2 1 1 1 Schwiebea sp2 20 9 7 Schwiebea sp4 1 1 1 Schwiebea sp6 1 1 1 Heterostigmata Imparipes sp1 1 1 1 Imparipes sp2 26 7 2 Imparipes sp2 33 3 1 Imparipes sp4 42 4 1 Imparipes sp5 3 1 2 Imparipes sp6 2 1 1 Bakerdania sp1 1 1 1 Petalomium sp2 6 1 1 Petalomium sp2 1 1 1 Petalomium sp4 12 6 1 Scutacarus sp1 41 10 3 Scutacarus sp2 29 4 2 Scutacarus sp4 150 15 10 Scutacarus sp5 2 2 1 Scutacarus sp6 1 1 1 Scutacarus sp7 1 1 1 Scutacarus sp8 1 1 1 Scutacarus sp9 2 1 1 Scutacarus sp10 1 1 1 Scutacarus sp11 11 3 2 Scutacarus sp12 4 2 1 Scutacarus sp13 1 1 1 Tarsonemidae sp1 1 1 1 Tarsonemidae sp2 2 2 1 Tarsonemidae sp2 1 1 1

71 Tarsonemidae sp4 1 1 1 Unguidispus sp1 1 1 1 Mesostigmata Oplitis alienorum 1 1 1 Oplitis sarcinulus 3 1 1 Oplitis sp1 nr. blufftonensis 1 1 1 Cosmolaelaps sp1 1 1 1 Non-Ant Associated Oribatida 1 1 1 Prostigmata sp1 1 1 1

72 Supplementary Figure 3. Diagram of sampling methods. Pitfall traps and vegetation quadrats were spaced 25m apart along the transect. Paired vegetation quadrats were measured on either side of the pitfall traps adjacent to the transect and averaged. Bait stations were place 8.3m from each pitfall trap.

Supplementary Figure 4. Mantel Test of Dissimilarity showed a correlation between Mite species Dissimilarity (Bray-Curtis) and Ant species Dissimilarity (Mantel r= 0.51, p=0.001) indicating that turnover in ant community composition is mirrored by turnover in mite community composition.

Chapter 3: Merging spatial parasitology and herbivore theory to identify ecology relevant spatial scales for commensal organisms

1 Abstract Host-dependent organisms perceive heterogeneity of the environment organized at multiple hierarchical levels and spatial scales, ranging from individual hosts, to the local ecological neighborhood, to the larger habitat or landscape. Hosts are often treated as identical replicates of resource patches for parasites, herbivores, and other dependent organisms, but there is evidence that ecological neighborhoods for associates should often be extended to include interactions among host species and environmental context. Four well-established resource hypotheses are often used to test mechanisms of herbivore diversity that could potentially apply to a wide array of other host-dependent systems: The Resource Concentration, Resource Abundance, Resource Size, and Resource Distribution Hypotheses. In this study, we tested predictions of the plant-herbivore resource hypotheses and spatial parasitology to understand potential interactions among hosts within the ecological neighborhood for a commensal mite system in grassland habitats. We used phoretic mites associated with ant colonies to determine if ecological neighborhoods in ant-associated mites extend beyond the host colony to include the neighboring ant nest community, if nest-neighborhood effects are consistent with the herbivore resource hypotheses, and if turnover in mite communities varied among grassland sites or adjacent mowed areas. Phoretic mite communities responded to ecological neighborhoods extending beyond host individuals and host colonies to include surrounding ant nest neighborhoods and habitat type. Ant colonies surrounded by nests of large-bodied ants had significantly higher mite abundances, which supported the Resource Size Hypothesis. We observed significant shifts in mite community composition and functional group abundances between mowed lawn and grassland habitats. We did not find evidence for significant turnover in mite community composition at larger spatial scales among grassland sites despite turnover in surrounding ant nest neighborhoods. Our results

74 indicate that commensal mite communities are structured by ecological neighborhoods at multiple hierarchical levels including individual hosts, the host colony, surrounding nest community, and habitat type, but do not vary significantly among sites.

2 Introduction Communities of parasites, herbivores, commensals, and mutualists, are dependent on their hosts for resources, and can offer an interesting perspective when studying community variation at multiple hierarchical and spatial scales (Guégan et al., 2005; Zelmer & Seed, 2004). The spatial extent of an ecological neighborhood, or ecologically relevant resource patch, can vary depending on an organism’s perception of heterogeneity in the environment (Addicott et al., 1987; Kotliar & Wiens, 1990; Wiens, 1989). Unlike free-living organisms, host-dependent organisms tend to have easily delineated population boundaries determined by a host individual (infrapopulation), and hosts are often treated as habitats encompassing communities of dependent species (infracommunity) (Table 1) (Walter & Proctor, 2013; Zelmer & Seed, 2004). Associates among individual hosts in a given ecosystem represent metapopulations of single species or metacommunities (=component communities) of multiple species. These communities may interact through host-mediated dispersal among sites or ecosystems, forming a supracommunity (= compound community). Hosts are often treated as perfectly replicated patches of resources for infracommunities of associates (Guégan et al., 2005; Zelmer & Seed, 2004); however, multiple studies have also demonstrated that the ecological neighborhood for associates can also be regulated by interacting host species at the metapopulation or metacommunity level (Lindenfors et al., 2007; Root, 1973) and environmental characteristics or resources available in the host landscape (Chapter 2, Päivinen et al., 2003). Host-dependent organisms use hosts for a variety of resources that can vary in abundance, duration, accessibility, and distribution. There are four well-established resource hypotheses in the plant-herbivore literature which have only rarely been tested, as such, in host-parasite, mutualist, or commensal systems (but see Päivinen et al., 2003, 2004). The Resource Concentration Hypothesis states that herbivores are more likely to find and maintain continuous populations on hosts growing in denser stands (Root, 1973),

75 a pattern which has also been shown in studies of barnacles (Blower & Roughgarden, 1989), terrestrial mammals (Morand & Poulin, 1998), and ant parasites (Päivinen et al., 2004). The Resource Abundance Hypothesis predicts the tendency for hosts that offer more resources (e.g. greater diversity or quantity of resources) to have higher associated species diversity (Harris & Dunn, 2010; Krasnov et al., 2004; Marques et al., 2000; Päivinen et al., 2003, 2004). The Resource Size Hypothesis predicts higher species richness on hosts that are large in terms of plant size (Lawton, 1983; Marques et al., 2000), host mass (Lindenfors et al., 2007; Poulin, 2007), and colony size (ants) (Päivinen et al., 2003). Lastly, the Resource Distribution Hypothesis predicts that hosts that are more widely distributed regionally host more diversity than those with restricted ranges (Cornell & Lawton, 1992; Harris & Dunn, 2010; Lindenfors et al., 2007; Päivinen et al., 2003). Although these hypotheses have been primarily applied to plant-herbivore interactions, they could be broadly applicable to a wide array of both free-living and non- antagonistic host-dependent organisms. In this study we merge plant-herbivore resource hypotheses into the framework of spatial parasitology using the concept of ecological neighborhoods. We use a commensalism of phoretic mites associated with ant hosts as our study system (described in more detail in Chapter 2). Mites frequently attach to more mobile animals in a temporary association for the purpose of dispersal (Houck & OConnor, 1991). During dispersal, feeding and development cease until they reach a suitable habitat, at which point they detach and continue development or reproduction (Kaliszewski et al., 1995). Phoretic mites can be host-specific or generalist (Campbell et al., 2013), and are most likely relying on within-nest fungi, bacteria, food stores, or decaying organic matter within ant nest or surrounding soil (Eickwort, 1990; Houck & OConnor, 1991; Kaliszewski et al., 1995; OConnor, 1982). Very few mites associated with ants are known parasites (Berghoff et al., 2009; Walter & Proctor, 2013); however, it has been suggested that phoresy may be a precursor of parasitism (Houck & OConnor, 1991; Kaliszewski et al., 1995). We can view this ant-associated mite system at multiple hierarchical levels that may structure phoretic mite communities among hosts, colonies, and habitats (Table 1). Identifying the spatial extent of a patch for an ant-associated mite is not a simple matter.

76 For example, from the perspective of a mite on an ant, even an individual host’s body can be a heterogeneous landscape of space available for attachment. Previous work has found evidence of this type of resource partitioning in terms of attachment site specificity among infracommunities (Campbell et al., 2013; Uppstrom & Klompen, 2011). If we scale to a wider extent, mites may perceive intraspecific variation among ants of different castes by avoiding dispersal dead ends such as male ants, and interspecific variation in host-ant suitability, preferring large-bodied cosmopolitan host species over small-bodied ants with restricted ranges (Chapter 2, Campbell et al., 2013; Uppstrom & Klompen, 2011). Interestingly, ants as social insects represent a special case in which one can also consider a “host individual” to be the colony, or superorganism (Hölldobler & Wilson, 2009; Hou et al., 2010). At this level, infracommunities of mites are relying on resources within the colony boundaries (hosts, larvae, fungi, microbes, or decaying organic matter), the arena in which they likely play out the non-phoretic stages of their life cycles. For this reason, throughout this paper we refer to communities at the level of ant colony rather than the individual ant. Nest architecture (Tschinkel, 2003), ant population size (Coovert, 2005), and resource availability (Laakso & Setälä, 1998) can be highly variable among colonies of different ant species. Even among colonies of a single species, nest population size and food resources might change depending on the ecological neighborhood (biotic and abiotic context) and age of the host colony. In Chapter 2, we established that mite infracommunities and metacommunities can be affected by age, area, soil texture, litter depth, and host frequency within a site. However, it remains unclear if proximity and competitive interactions among nests may affect mite infracommunities at the colony level, to what extent the metacommunities (within site) and supracommunities (among site) may vary, and if habitat type can produce detectible patterns in associated mite diversity. The goals of this study are to: (i) determine the extent to which ecological neighborhoods in ant-associated mites are influenced by surrounding ant nests; (ii) examine whether nest-neighborhood relationships with mite species composition and abundance are consistent with the herbivore resource hypotheses; (iii) determine if local

77 neighborhoods vary among sites and if this is mirrored by a change in mite metacommunity turnover; and, lastly (iv) identify if habitat type (constructed grasslands versus mowed lawns) shifts mite community composition and functional group dominance. We hypothesize that variation in nest neighborhoods can affect infracommunities of mites in three primary ways: as a source of diverse hosts harboring diverse resources (promotes generalists), as a source of nearby nests of their preferred host species (promotes specialists), or antagonistically as competitors with host colonies, potentially limiting availability of resources. At the levels of the metacommunity and supracommunity, we hypothesize that mite communities are structured by host dispersal events, which often may not extend beyond the local habitat. Thus, we predict that within site composition of mites will be more similar than among site composition due to predominance of within site dispersal events and potentially very different surrounding nest communities among sites. Lastly we hypothesize that mite infracommunities shift in composition if the nest within a local neighborhood occurs in a highly disturbed habitat (mowed lawns) when compared to adjacent natural habitats (constructed grasslands).

3 Methods

3.1 Nest Neighborhood Study To study the role of the surrounding nest community on phoretic mite diversity, we focused on two common ant hosts, Myrmica americana and Aphaenogaster rudis. Based on previous studies, we established that these two species are abundant in constructed grasslands (Chapter 1) and are hosts to diverse assemblages of phoretic mites (Chapter 2, Campbell et al., 2013). We found 3-5 focal nests of both species in each of five constructed grasslands by following ants recruiting to cookie baits (pecan sandies) along a central transect. Pecan sandies are commonly used as bait for a wide range of ants because they contain lipids, carbohydrates, and protein, and have been employed by myrmecologists for over 25 years (Human & Gordon, 1996). To decrease potential non- independence of focal nests, we maintained at least 8 m distance between focal nests along the transect, though most focal nests were separated by much farther (M. americana mean distance = 33m, range=8-108 m; A. rudis mean distance = 65.9m,

78 range=17-175 m). A total of 14 M. americana nests and 17 A. rudis nests were included in the analyses. We collected 20 workers in individual tubes per sampling period from each focal nest using cookie baits at the nest entrance and an aspirator. We sampled nests every five weeks during late June, early August, and mid September of 2012. Due to nest relocations, we sampled a few nests only twice (4 M. americana and 3 A. rudis nests). Ants were frozen and individually inspected for phoretic mites. All mites were cleared and mounted on slides for identification to genus and morphospecies (when possible to species). Mite richness and abundance data for all sampling periods were pooled for each focal nest. Voucher specimens of each mite and ant species will be deposited at the Ohio State Acarology Collection. The surrounding nest (SN) neighborhood was taken to be a 10 m2 (1.8 m radius circle) area encircling the focal nest (Figure 1a). We tested the resource hypotheses specifically at this scale rather than at the site or regional scale, because the majority of the mites likely play out their lives within the ant nest and immediate vicinity due to limited dispersal abilities. It is difficult to assign an ecologically relevant scale to the neighborhood for ants and associated mites; however, we believe the chosen area represents the majority of the ant community the focal nest members encounter regularly and extends well beyond the maximum diameter of the underground portion of the focal nest. For example other species of Myrmica are known to forage within a 2 m range of their nest (Schlick-Steiner et al., 2006), and grassland ant assemblages have average foraging distances of < 0.5 m (Albrecht & Gotelli, 2001). Using regularly placed cookie baits, three observers searched for all surrounding ant nests within the circular area by following ants carrying bait back to their nests and by randomly searching. Searching continued until all nests with ants carrying bait were located within the 10-m2 neighborhood. Surrounding ant nests were quantified and identified to species. We tested the SN community as a resource for the mite infracommunity within the focal nest. We characterized SN resources using four univariate predictors (Figure 1): (i) SNrichness- species richness of all ant nests including focal nest, (ii) SNabundance- total number of surrounding nests (iii) SNsame- number of surrounding nests that were the same species as the focal nest, and (iv) SNlarge- number of nests belonging to ants

79 larger than 3mm. The surrounding nest community was also characterized using a multivariate ordination method, Non-metric Multidimensional Scaling (NMDS, metaMDS function, vegan package, R) with Bray-Curtis dissimilarity. Axis scores from the NMDS (NMDS1 and NMDS2) were used as potential multivariate predictors in mite models. Based on predictions of the Resource Concentration Hypotheses, we expected higher mite richness and abundance in focal nests surrounded by neighborhoods with the greater densities of the same ant host (SNsame) and with NMDS axis scores of the surrounding ant community that were associated with the same or closely related ant hosts. If our results support the Resource Abundance Hypothesis, we expected a positive relationship of mite diversity with SNabundance and SNrichness, nest neighborhoods offering greater diversity and quantity of resources. According to the Resource Size Hypothesis we predicted a positive relationship for phoretic mite diversity when surrounded by nests of ant species with large body sizes (SNlarge). In our NMDS ordination plot we should see a positive relationship of mite abundance and richness in nests that are surrounded by more large-bodied ant species (Aphaenogaster rudis, Lasius spp., Formica spp., and Myrmica spp) and a negative relationship for nests surrounded by of small-bodied host species (Monomorium minimum, Pheidole tysoni, Solenopsis molesta, Strumigenys sp., Temnothorax ambiguus. Tetramorium caespitum). We were unable to use colony size (volume or population) as a variable (as in Päivinen et al., 2003; 2004), because nearly all of the nests in the neighborhoods belonged to hosts that make medium to large nests. We did not conduct direct tests of the Resource Distribution Hypothesis in this study, but previous work has supported the positive relationship of phoretic mites with host frequency within sites and on ants that are cosmopolitan among sites (Chapter 2). We used generalized linear models (glm function, R) to predict mite richness and abundance for M. americana and A. rudis focal nests separately. Poisson error distributions were used for mite richness models, and Gaussian error distributions were used for log-transformed mite abundance models. Potential predictors (SNrichness, SNabundance, SNsame, SNlarge, NMDS1 and NMDS2) were evaluated using the lowest Akaike’s Information criterion (AIC) score. Any models differing by 2 or less AIC points

80 were considered competing models. We tested best models (univariate) versus the null model in a likelihood ratio test to calculate p-values and percent deviance explained. If null or multivariate models competed with our best model we used likelihood ratio tests to determine if expanded models were significantly better. To compare the relative role of within versus among site constraints on nest neighborhood composition and metacommunities of mites, we conducted a Multidimensional Scaling (MDS) ordination. We used MDS (capscale function, vegan package, R) with Bray-Curtis dissimilarity and a permutational ANOVA (adonis function, vegan package, R) to test site as a predictor of neighborhood ant nest composition and mite community composition. MDS was used here since it is a variance partitioning method in contrast to the NMDS scores used earlier as predictor variables in the univariate, general linear models.

3.2 Nest Habitat Study We tested if habitat type can produce detectible shifts in composition of communities of ant-associated mites by collecting mites from M. americana, a common ant species found in both highly disturbed lawn habitats and constructed grasslands. We expected mite richness and abundance to be highest in the grasslands, because ant abundance and richness is greater (personal observation), and grasslands have less frequent disturbance than the adjacent mowed lawns. At three different sites, we collected 20 workers individually (as described above) from 15 nests located 20 m inside the constructed grassland and 15 nests in the mowed lawn border adjacent to the constructed grassland and up to 5 m from the grassland edge (Figure 1b). We collected ants from the nests over three sampling periods in 2013 (mid June, July, and August). Due to nest relocations and inability to sample for every period, only 31 of the 45 nests inside and 38 of the 45 nests outside were included in the analysis.

We used MDS with Bray-Curtis dissimilarity and a permutational ANOVA to test habitat (grassland vs. lawn) as a predictor of mite supracommunity composition. Using generalized linear models with Gaussian error distributions, we tested for differences in overall mite abundance and taxonomic group abundance (Heterostigmata and Astigmata) due to habitat. These taxonomic groups can also be used as surrogates for functional

81 groups because the Heterostigmata genera associated with ants (Petalomium and Scutacarus) are primarily fungivores, while Astigmata genera may be bacterivores (Histiostoma) or fungivores (Forcellinia and Schwiebea) (Eickwort, 1990). We log transformed abundance values for all models and used likelihood ratio tests to compare the habitat model to the null model in order to calculate p-values and percent deviance explained.

4 Results

4.1 Nest Neighborhood Study Workers in M. americana focal nests harbored 9 mite species (160 individuals), while A. rudis nests had 7 mite species (74 individuals) (Figure 2). Surrounding ant nest communities of A. rudis showed more central clustering than M. americana (Figure 3) There were no models better than the null for predicting M. americana focal nest mite richness and abundance (Table 2). There were also no predictors that improved upon the null for A. rudis mite richness; however, A. rudis mite abundance was best predicted by a positive relationship with NMDS1 (p=0.005, 34.31% Dev. expl., df=1,15) (Figure 4, Table 2). There were two competing models for A. rudis mite abundance that included NMDS1 and either SNlarge or SN abundance.

4.1.1 Tests of Resource Hypotheses According to the Resource Concentration Hypothesis, we predicted a positive relationship between mite diversity and the nest densities of same ant species as the focal nest (SNsame) or clustering of focal nests with higher diversity in the center of the NMDS plots where the species composition of ant nests in the surrounding neighborhoods were more similar to the focal ant species. Neither A. rudis nor M. americana showed central clustering in the NMDS plots (Figure 3). A. rudis mite abundance and richness tended to be higher when surrounded by more nests of the same species (SNsame) while M. americana mite abundance and richness were lower, although this variable did not appear in best models. Mite richness and abundance were predicted to be positively related to SNabundance and SNrichness according to the Resource Abundance Hypothesis. Mite

82 richness and abundance relationships were consistently positive with these two predictors for both species, and SNabundance appeared in a competing model for A. rudis mite abundance, though it did not appear in the best model. The Resource Size Hypothesis was tested in terms of host size for this study, and we expected SNlarge to have a positive relationship with mite diversity measures. Both M. americana and A. rudis mite richness and abundance was greater when surrounded by large nests, and SNlarge appeared in a competing model (with NMDS1) for A. rudis mite abundance; however, this predictor was not better than NMDS1 alone for A. rudis or the null model for M. americana. Mite abundance in M. americana focal nests tended to be higher in nests loading negatively on the NMDS1 and NMDS2 axes, which corresponds to nests surrounded by more large-bodied ant species (Myrmica latifrons, Formica pallidefulva, Lasius neoniger, and Aphaenogaster rudis, rather than small bodied ant species (Monomorium minimum, and Solenopsis molesta) (Figure 4a). A. rudis mite abundance showed similar trends of increased mite abundance away from neighborhoods with M. minimum and S. molesta nests (Figure 4b).

4.1.2 Among Site Variation Results of the MDS analysis (Figure 5) showed that the species composition of ant nests in neighborhoods surrounding focal A. rudis nests were significantly different among sites (p=0.0001, R2=0.51). M. americana neighborhoods showed the same trend; however, potentially due to low sample sizes at three of the sites we did not find a significant effect of site (p=0.107, R2=0.39). Although SN neighborhoods were significantly different among sites, mite metacommunities did not show similar turnover in composition (Figure 6).

4.2 Nest Habitat Study We collected 1446 mites comprising 13 species from 69 M. americana colonies across the two habitat types. Nests outside the grassland had 1072 mites (13 species), while grassland nests had 374 mites (8 species). We predicted that mite abundance and richness would be greater inside the grassland, but average mite abundance was not significantly different for nests in the two habitats, though the general trend was for higher abundance and richness in nests in the lawn habitat (Figure 7a). MDS ordination indicated that there is a significant difference in mite composition between the two

83 habitat types (p=0.001, R2=0.080) (Figure 8a). This shift is likely attributed to changes in abundance of Heterostigmata and Astigmata (Figure 7b,c). Astigmata especially Histiostoma spp. are more abundant in the lawn (p=0.025, 6.98% Dev. expl.), while Heterostigmata are more abundant in the grassland (p<0.0001, 22.82% Dev. expl.). The MDS ordination for site as a predictor of mite composition was not significant (p=0.07, R2=0.054) (Figure 8b).

5 Discussion Our study is the first to merge spatial principles in parasitology, plant-herbivore resource theory, and the concept of ecological neighborhoods and apply them to a novel hierarchical system of commensal organisms. We found that phoretic mite communities respond to ecological neighborhoods extending beyond host individuals and host colonies, including surrounding ant nest neighborhood and habitat type, but we did not find evidence at the largest scale for among-site variation in metacommunity composition. The plant-herbivore resource hypotheses are quite general, and because of this they have been tested at multiple spatial scales, for example, from individual hosts (Lawton, 1983), to the patch level (Grez & González, 1995), and the regional level (Marques et al., 2000). While the generality of the hypotheses allows them to potentially be applied a variety of systems, relative support for each hypothesis may depend on the spatial scale and the size of the ecological neighborhoods of the focal organisms. For example, in this study we tested three of the four hypotheses at a single spatial scale (10 m2 range of a focal colony) and only found clear support for the Resource Size Hypothesis. We could have tested the Resource Concentration Hypothesis at the patch level in terms of relative frequencies of host ants within the patch rather than using same species nests (SNsame) in the immediate vicinity. Future studies would benefit from testing these hypotheses at multiple ecologically relevant spatial scales to identify the role of scale dependence in resource concentration, abundance, and size. The host ant colony and surrounding ant nest neighborhood likely encompasses the majority of resources ant-associated mites use for both phoretic (hosts) and free-living stages (fungi, bacteria, and other soil resources). Although not included in the best models, SNabundance, SNrichness, and SNlarge were all positively related to mites

84 associated with both M. americana and A. rudis nests, adhering to predictions for the Resource Abundance and Resource Size Hypotheses. The best model for A. rudis mite abundance included NMDS1 (Figure 3), which also corresponded to larger ant nests (Figure 4) and the Resource Size Hypothesis. The Resource Concentration Hypothesis was supported by mites associated on A. rudis, but not M. americana which had opposite relationships with SNsame. This may be due to M. americana mites being primarily host generalists (associated with multiple ant species in different genera), while A. rudis mites were primarily host specific (associated with only A. rudis or other Aphaenogaster species). We would expect specialist mites to be more abundant when surrounded by nests of their preferred host species, while generalist mites may respond more to SNabundance, SNrichness, or SNlarge. Ant neighborhood composition can vary depending on the site (Figure 5), and is likely determined by site characteristics such as time since planting, soil texture, and surrounding landscape (Chapter 1); however, mite compositional changes were not detectable at site level scales for the nest neighborhood study (Figure 6), nor the nest habitat study (Figure 8b). We suspect that this may be due to the overarching role of host species in filtering mite communities (Chapter 2), such that when sampling from sites in comparable habitats and on the same host species, the site-level variation becomes mostly irrelevant. The nest habitat study showed that although nests in a given habitat type (lawn) may be separated by short distances (25 m) from nests in a different habitat type (grassland), they may have striking differences in the composition of mite functional groups (Figure 7, Figure 8). Multiple studies have found that conversion of agricultural land to constructed grasslands increases soil quality in multiple ways including decreased bulk density (Baer et al., 2000), increased soil organic matter (McLauchlan et al., 2006), and increased fungal abundance (Allison et al., 2005). Changes in soil properties and soil microbial community structure in turn can affect microarthropod community structure (Laakso & Setälä, 1998). In our study, grassland nests had significantly more (Heterostigmata) while lawn nests had Astigmata communities dominated by Histiostoma bacterivores. Heterostigmata often carry spores of preferred species of fungi with them when dispersing, in pocket-like structures on their bodies known as sporothecae

85 (Ebermann & Hall, 2003, 2004). It may be the case, that fungal food resources for these mites grow better inside a grassland than in more disturbed habitats such as lawns or agricultural fields.

5.1 Conclusion Our results demonstrate the applicability of herbivore and parasite theory to other non-antagonistic interactions such as commensalisms. We found support for the Resource Size Hypothesis at the scale of the surrounding nest community, but recommend testing this and other hypotheses at both local and regional scales to more thoroughly thest scale dependence of these hypotheses. Together our two studies illustrate the importance of ecologically relevant spatial scales and environmental context for understanding diversity and composition patterns of host-dependent organisms.

86 6 References Addicott, J. F., Aho, J. M., Antolin, M. F., Padilla, D. K., Richardson, J. S., & Soluk, D. a. (1987). Ecological neighborhoods: scaling environmental patterns. Oikos, 49(February), 340–346. Albrecht, M., & Gotelli, N. J. (2001). Spatial and temporal niche partitioning in grassland ants. Oecologia, 126(1), 134–141. Allison, V. J., Miller, R. M., Jastrow, J. D., Matamala, R., & Zak, D. R. (2005). Changes in soil microbial community structure in a tallgrass prairie chronosequence. Soil Science Society of America Journal, 69, 1412–1421. Baer, S. G., Rice, C. W., & Blair, J. M. (2000). Assessment of soil quality in fields with short and long term enrollment in the CRP. Journal of Soil and Water Conservation. Berghoff, S. M., Wurst, E., Ebermann, E., Sendova-Franks, A. B., Rettenmeyer, C. W., & Franks, N. R. (2009). Symbionts of societies that fission: mites as guests or parasites of army ants. Ecological Entomology, 34, 684–695. Blower, S. M., & Roughgarden, J. (1989). Parasites detect host spatial pattern and density: a field experimental analysis. Oecologia, 78(1), 138–141. Campbell, K. U., Klompen, H., & Crist, T. O. (2013). The diversity and host specificity of mites associated with ants: the roles of ecological and life history traits of ant hosts. Insectes Sociaux, 60(1), 31–41. Coovert, G. A. (2005). The Ants of Ohio. Columbus, Ohio: Ohio Biological Survey, Inc. Cornell, H. V., & Lawton, J. H. (1992). Species interactions , local and regional processes , and limits to the richness of ecological communities : a theoretical perspective. Journal of Animal Ecology, 61(1), 1–12. Ebermann, E., & Hall, M. (2003). First record of sporothecae within the mite family Scutacaridae (Acari, Tarsonemina). Zoologischer Anzeiger, 242, 367–375. Ebermann, E., & Hall, M. (2004). A new species of scutacarid mites transferring fungal spores (Acari, Tarsonemina). Revue Suisse de Zoologie. Eickwort, G. C. (1990). Associations of mites with social insects. Annual Review of Entomology, 35, 469–488. Guégan, J. F., Morand, S., & Poulin, R. (2005). Are there general laws in parasite community ecology? The emergence of spatial parasitology and epidemiology. In F. T. F. Renaud & J. F. Guégan (Eds.), Parasitism and Ecosystems (pp. 22–42). Oxford University Press. Harris, N. C., & Dunn, R. R. (2010). Using host associations to predict spatial patterns in the species richness of the parasites of North American carnivores. Ecology Letters, 13(11), 1411–1418. Hölldobler, B., & Wilson, E. O. (2009). The Superorganism. New York: W. W. Norton & Company, Inc. Hou, C., Kaspari, M., Zanden, H. B. V, & Gillooly, J. F. (2010). Energetic basis of colonial living in social insects. Proceedings of the National Academy of Sciences of the United States of America, 107(8), 3634–3638. Houck, M. A., & OConnor, B. M. (1991). Ecological and evolutionary significance of phoresy in the Astigmata. Annual Review of Entomology, 36, 611–636.

87 Human, K. G., & Gordon, D. M. (1996). Exploitation and interference competition between the invasive Argentine ant, Linepithema humile, and native ant species. Oecologia, 105,405–412. Kaliszewski, M., Athias-Binche, F., & Lindquist, E. E. (1995). Parasitism and parasitoidism in Tarsonemina (Acari: Heterostigmata) and evolutionary considerations. Advances in Parasitology, 35, 335–367. Kotliar, N. B., & Wiens, J. a. (1990). Multiple scales of patchiness and patch structure: a hierarchical framework for the study of heterogeneity. Oikos, 59(2), 253–260. Krasnov, B. R., Shenbrot, G. I., Khokhlova, I. S., & Degen, a. A. (2004). Relationship between host diversity and parasite diversity: flea assemblages on small mammals. Journal of Biogeography, 31(11), 1857–1866. Laakso, J., & Setälä, H. (1998). Composition and trophic structure of detrital food web in ant nest mounds of Formica aquilonia and in the surrounding forest soil. Oikos, 81, 266–278. Lawton, J. H. (1983). Plant architecture and the diversity of phytophagous insects. Annual Review of Entomology, 28, 23–39. Lindenfors, P., Nunn, C. L., Jones, K. E., Cunningham, A. a., Sechrest, W., & Gittleman, J. L. (2007). Parasite species richness in carnivores: Effects of host body mass, latitude, geographical range and population density. Global Ecology and Biogeography, 16(4), 496–509. Marques, E. S. D. A., Price, P. W., & Cobb, N. S. (2000). Resource abundance and insect herbivore diversity on woody fabaceous desert plants. Environmental Entomology, 29(4), 696–703. McLauchlan, K. K., Hobbie, S. E., & Post, W. M. (2006). Conversion from agriculture to grassland builds soil organic matter on decadal timescales. Ecological Applications, 16(1), 143–53. Morand, S., & Poulin, R. (1998). Density, body mass and parasite species richness of terrestrial mammals. Evolutionary Ecology, 12(6), 717–727. OConnor, B. M. (1982). Evolutionary ecology of astigmatid mites. Annual Review of Entomology, 27, 385–409. Päivinen, J., Ahlroth, P., Kaitala, V., Kotiaho, J. S., Suhonen, J., & Virola, T. (2003). Species richness and regional distribution of myrmecophilous beetles. Oecologia, 134, 587–595. Päivinen, J., Ahlroth, P., Kaitala, V., & Suhonen, J. (2004). Species richness, abundance and distribution of myrmecophilous beetles in nests of Formica aquilonia ants. Annales Zoologici Fennici, 41, 447–454. Poulin, R. (2007). Are there general laws in parasite ecology? Parasitology, 134(Pt 6), 763–776. Root, R. B. (1973). Organization of a plant-arthropod association in simple and diverse habitats: the fauna of collards (Brassica oleracea). Ecological Monographs, 43(1), 95–124. Schlick-Steiner, B. C., Steiner, F. M., Moder, K., Bruckner, A., Fiedler, K., & Christian, E. (2006). Assessing ant assemblages: pitfall trapping versus nest counting (Hymenoptera, Formicidae). Insectes Sociaux, 53(3), 274–281. Tschinkel, W. R. (2003). Subterranean ant nests: trace fossils past and future ? Palaeogeography Palaeoclimatology Palaeoecology, 192, 321–333.

88 Uppstrom, K. A., & Klompen, H. (2011). Mites (Acari) associated with the desert seed harvester ant, Messor pergandei (Mayr). Psyche, 2011, 1–7. Walter, D. E., & Proctor, H. C. (2013). Animals as Habitats. In Mites: ecology, evolution & behaviour (pp. 341–422). Dordrecht: Springer Netherlands. doi:10.1007/978- 94-007-7164-2 Wiens, J. A. (1989). Spatial scaling in ecology. Functional Ecology, 3(4), 385–397. Zelmer, D. a., & Seed, J. R. (2004). A Patch Hath Smaller Patches: Delineating Ecological Neighborhoods for Parasites. Comparative Parasitology, 71(2), 93– 103.

89 Table 5. Definitions and ant-associated mite spatial examples (host and superorganism levels) of parasite population and community ecology terminology used throughout this Chapter (modified from Guégan et al., 2005).

Term Definition Ant-Associated Mite Example Infrapopulation All members of the same All mites of same species on a single ant individual species on a single host All mites of same species within a single ant individual colony Spatial examples: Attachment site specificity, location of a species within ant nest Infracommunity All species on a single host All mite species on a single ant individual individual All mite species within a single ant colony Spatial examples: Attachment site partitioning among mite species, community composition of mites determined by host caste Metapopulation All members of the same All mites of the same species on ants of the same species on multiple host species within a site individuals of the same species All mites of the same species in ant colonies of the within an ecosystem same species within a site Spatial example: Mite species abundance varies among different colonies within a site depending on host frequency Metacommunity All species on multiple All mite species associated with ants of the same (=component individuals of the same host species within a site community) species within an ecosystem All mite species in ant colonies of the same species within a site Spatial example: Mite richness is higher on larger ant species. Mite community composition is similar among colonies of the same site, mite richness within a colony is influenced by surrounding nest community Supracommunity All species on multiple host All mite species associated with ants of the same (=compound individuals across multiple sites species across multiple sites connected by host community) dispersal events All mite species associated with ant colonies of the same species across multiple sites connected by host dispersal events. Spatial example: The total mite community across colonies in multiple habitat types. Mite communities are more similar within nests at the same site than nests among sites.

90 Figure 1. Methods for (a) Nest Neighborhood Study showing an example of a surrounding nest community for a Myrmica americana focal nest. Focal nest (FN-red triangle) is shown at center and encircled by 1.8m radius area. Surrounding nests (SN- squares) are color coded by ant species. SNsame= 3 (all red squares), SNrichness= 5 (number of nests of different colors), SNabundance= 8 (number of squares), SNlarge= 6 (number of nests belonging to large ant species- blue, red, and orange) and (b)Nest Habitat Study showing G5 site with 15 nests inside the grassland (orange triangles) and 15 nests in the mowed lawn bordering the grassland (light blue trianges).

91 Figure 2. Species accumulation curves for mites associated with focal nests of a) Myrmica americana (blue) and b) Aphaenogaster rudis (green).

92 Table 6. Best and competing models (ΔAIC<2) for mite richness and abundance for Myrmica americana and Aphaenogaster rudis nests.

ΔAIC ΔAIC vs. % Dev. vs. % Dev. Response Best Model Null Expl. Competing Models Best Expl. Mite Richness americana Null ------rudis Null ------Mite Abundance americana Null ------

rudis NMDS1 -5.15 34.3% (+)NMDS1+SNabundance +1.18 38.2% (+)NMDS1+SNlarge +0.23 40.8%

Figure 3. Non-metric Multidimensional Scaling ordination plots for the species composition of ant nests surrounding focal nests of a) Myrmica americana and b) Aphaenogaster rudis. Ant species names are shown in blue and green for the two different focal nest species respectively. Numbers represent individual focal nests scaled to the abundance of mites within the focal nest.

93 Figure 4. The best model for mite abundance in Aphaenogaster rudis focal nests included the NMDS Axis 1 Score (34.31% Deviance explained, p=0.005). Note: Y axis is log-scaled.

94 Figure 5. Multidimensional Scaling ordination plots showing site as a predictor of surrounding ant nest community composition for (a) Myrmica americana and (b) Aphaenogaster rudis focal nests. Each colored point represents a focal nest at one of 5 sites (Ehr, FR, TC, G1 and G6). 95% confidence intervals are shown for sites (black ellipses) with at least three focal nests.

95 Figure 6. Multidimensional Scaling ordination plots showing site as a predictor of mite community composition in (a) Myrmica americana and (b) Aphaenogaster rudis focal nests. Each colored point represents a focal nest at one of 5 sites (Ehr, FR, TC, G1 and G6). 95% confidence intervals are shown for sites (black ellipses) with at least three focal nests.

96 Figure 7. Boxplots showing habitat related shifts in abundance for mites collected from Myrmica americana nests in the grassland (dark gray) and in the lawn (light gray). Total mite abundance (a) was similar between habitats, but (b) Heterostigmata were significantly (designated with: *) higher in the grassland, and (c) Astigmata were significantly higher in the lawn. Note: Y axis is log scaled for a) and c).

97 Figure 8. Multidimensional Scaling ordination plots of showing (a) habitat (grassland- pink, lawn- blue) and (b) site (G1, G5, G6) as a predictors of mite community composition in Myrmica americana nests. 95% confidence intervals are shown for habitat and sites (black ellipses).

98 Chapter 4: Phoresy across space and time: commensal mite communities are structured by distance-decay and synchrony with hosts

1 Abstract The relationship between dissimilarity and spatial distance, first introduced for geographical principles, has since become well established in ecological theory. Three primary mechanisms drive observed distance-decay relationships in species communities including: dissimilarity of the environment with distance and subsequent adaptation of the community to different niches, differential resistance of the heterogeneous landscape (e.g. dispersal barriers), and limited dispersal abilities of organisms in homogenous environments. In this study, we tested the latter mechanism using phoretic mite communities that rely on ants for dispersal in a homogenous grassland field patch. Phoretic mites in this system use ant hosts for dispersal and likely use nest resources during other life stages. Dispersal of mites within this system is primarily limited by worker ants, which forage outside short distances but return to the same nest. Phoretic mites also preferentially use reproductive castes (flying female ants) which disperse and found new colonies at greater distances, but are only present during brief periods in the summer. We hypothesized that spatial arrangement of ant nests affects mite dispersal among nests in a homogenous environment following the distance-decay relationship. We also hypothesized that phoretic mite diversity is synchronized with their hosts’ life cycles and should increase during host reproductive periods. Mite species composition showed significant spatial autocorrelation for the overall Mantel test, and correlograms revealed spatial dependence at the first distance class. Spatial autocorrelations of mite abundance and richness (Moran’s I) showed no significant relationships with distance pairs. Mites tended to be phoretic throughout the season on workers, but specialist mites tended to increase in relative abundance during periods of host flights. Our results show that the distance-decay relationship applies in homogenous environments for dispersal limited organisms and demonstrate life-cycle synchrony for host specific organisms.

2 Introduction The inverse relationship between similarity and spatial or temporal distance, or distance-decay relationship, was first formalized over four decades ago to describe the

99 geographical basis of socioeconomics (Tobler, 1970) and has since become incorporated into multiple ecological frameworks including biogeography, population genetics, and source-sink dynamics of metapopulations. The distance-decay relationship has been supported by spatial studies of communities across a wide variety of taxonomic systems including plants (Nekola & White, 1999; Soininen et al., 2007), parasites (Oliva & González, 2005; Poulin, 2003; Thieltges et al., 2009; Vinarski et al., 2007), arboreal arthropods (Lindo & Winchester, 2007), and ectomycorrhizal fungi (Bahram et al., 2013). Three non-exclusive primary mechanisms are believed to drive the distance-decay relationship: (i) dissimilarity of the environment with distance and subsequent adaptation of the community to different niches, (ii) differential resistance of the landscape or barriers to dispersal, or (iii) limited dispersal abilities of organisms even in homogenous environments (Nekola & White, 1999; Soininen et al., 2007). Studies of distance-decay in parasite communities are extensive and demonstrate that multiple factors may contribute to the distance-decay relationship including host vagility (Thieltges et al., 2009), host distribution (Oliva & González, 2005), and environmental variation (Poulin et al., 2011; Vinarski et al., 2007; Warburton et al., 2015). Most of these studies have focused on testing whether the distance-decay relationship exists in the system or at very large spatial scales (global or regional) in order to test the first of the three primary drivers above (niche-based). Although parasite communities have been extensively studied, other systems of host-dependent organisms such as mutualisms and commensalisms remain unexplored. In this study we focus on commensal phoretic mite communities within a small homogenous habitat patch. This system is ideal to test the importance of dispersal limitation in homogenous habitats, because we control for potential environmental and historical components that can influence mite diversity (Chapter 2, 3). Additionally, a large number of ant colonies can exist within a small homogenous area, and each nest is comparable to an island, harboring a diversity of microarthropods, microflora, and microfauna (Laakso & Setälä, 1998). Mites in this system rely on the ants to disperse, although for most species it remains unclear as to what drives their dispersal or to where they are dispersing (Eickwort, 1990; Houck & OConnor, 1991; OConnor, 1982). Although ants are highly vagile, mite dispersal by host ants is limited because ants from

100 different colonies tend to avoid or have antagonistic interactions with neighboring colonies (Hölldobler & Wilson, 1990). Mite communities may also be influenced by temporal changes in ant colonies throughout the season. Phoretic mite synchrony with host lifecycles has been demonstrated for mites associated with bees, wasps, beetles, and flies (Houck & OConnor, 1991). For ant-associated mites, female alates (winged reproductive ants) present the greatest opportunity for dispersal, because after mating they become foundress queens of new nests, while workers (sterile females) remain at the natal nest, and males die soon after mating. Previous studies have found that mite prevalence corresponds to host dispersal suitability (Campbell et al., 2013; Uppstrom & Klompen, 2011). Alates of most ant species are only present during specific periods of the colony’s summer cycle, but phoretic mites are present throughout the season on workers (Campbell et al., 2013; Moser & Blomquist, 2011). Mites associated with Red Imported Fire ants (Solenopsis invicta) are known to synchronize their dispersal with alate abundance (Ebermann & Moser, 2008; Moser & Blomquist, 2011; Walter & Moser, 2010); however, S. invicta colonies produce alates throughout most of the year and flights occur during nearly every month (Tschinkel, 2006). In this study, we identify spatial and temporal patterns in ant-associated mite communities. We test the following two hypotheses: (i) spatial arrangement of ant nests affects mite dispersal among nests in a homogenous environment following the distance- decay relationship and (ii) mite communities synchronize their life cycles throughout the season with their hosts to optimize dispersal. We predict that mite abundance, richness and community composition will be more similar for nests that are in close proximity and that phoretic mite abundance and richness will be greatest during the reproductive phase (alate period) of host ant nests. We also predict that phoretic mite species that are host specific (specialists) will show more synchronized dispersal stages than mites associated with multiple host species (generalists).

101 3 Methods 3.1 Study System We conducted our study within a 2-ha, old agricultural field at the Ecology Research Center (Miami University, Oxford, OH). The field had not been tilled in over 20 years but periodic mowing resulted in a relatively homogenous plant community. Vegetation included predominantly cool season pasture and meadow grasses (fescue, orchard grass, timothy grass, and green foxtail) and small amounts of forbs (goldenrod, common milkweed, dogbane, and clovers). Our study system comprises phoretic mites associated with two common ant species, Myrmica americana and Aphaenogaster rudis. Phoretic mites in this system are predominantly fungivores and bacterivores (Heterostigmata and Astigmata) or predators and scavengers (Mesostigmata) during non-phoretic life stages, and range from host specific to host generalist (Chapter 3, Campbell et al., 2013).

3.2 Nest Sampling In the center of the field, we marked the nodes of a 5 x 5 grid, with 25-m spacing (Supplementary Figure 1). At each of these points, we used cookie baits to find nests belonging to two common ant species Myrmica americana and Aphaenogaster rudis. We sampled 20 workers from the nests during each sampling period using crushed cookie baits to draw out foragers. Each worker was collected in an individual Eppendorf tube and frozen. Each nest was sampled over 7 sampling dates (3 weeks apart) throughout the summer from May to September, 2013. In total we found 46 M. americana nests across the grid, and 6 A. rudis nests. We only included nests in our analysis for which we were able to collect workers during all 7 sampling periods; therefore, we were only able to use 21 M. americana nests and 5 A. rudis nests. Ants were thawed and individually inspected for mites under a microscope. All mites were cleared in lactic acid and mounted on slides using Polyvinyl alcohol medium (PVA). We identified mites to genus and morphospecies, and when possible to species. Voucher specimens of each mite and ant species will be deposited at the Ohio State Acarology Collection.

102 3.3 Statistical Analysis To test for adequate sampling of mites, we constructed species accumulation curves and calculated the Chao richness estimate for M. americana and A. rudis nests (specaccum, specpool, vegan package, R) (Oksansen et al., 2013). It was clear that A. rudis nests were not adequately sampled for the spatial analysis, so we restricted our spatial analysis to M. americana. We pooled all 7 sampling periods of mite data for each nest of M. americana in our spatial analysis. To test for spatial autocorrelation of mites in M. americana nests we used the coordinates of the field grid to calculate pairwise Euclidean distances of each nest and the pairwise Bray-Curtis dissimilarity of the mite community (vegdist,vegan package, R). The maximum distance between any two nests was 141 m and the minimum distance was 25 m, a distance that we believe is well beyond the foraging range of the ants of this study based on field observations of foragers to bait stations and other studies of similarly sized ants (Schlick-Steiner et al., 2006; Albrecht & Gotelli, 2001). We used Mantel tests (mantel.correlog, vegan package, R) to determine correlation between the spatial matrix and mite community matrix at 6 distance classes representing comparisons of nearest neighbors, second nearest neighbors, and so on (Supplementary Figure 1). We used Benjamini–Hochberg (BH) adjustment to p-values to reduce false discovery errors due to multiple comparisons at the distance classes. Additionally we conducted an overall Mantel test of Euclidean distance by mite community dissimilarity and calculated a Pearson correlation coefficient and p-value for the test using 999 resampling permutations. We tested if mite abundance and richness were spatially autocorrelated among nests using Moran’s I (correlog, ncf package, R) and calculated p-values for significance of the Moran's I correlation coefficients at each distance class by conducting 999 resampling permutations (resamp, correlog function). For temporal analysis, we predicted mite abundance and richness would be greatest when the colonies have their mating flights. Typically female and male alates are present in the nest for approximately a month before they fly from the nest (Hölldobler & Wilson, 1990). Data for seasonal timing of mating flights are sparse for M. americana and A. rudis. Coovert (2005) noted that females of M. americana can be found in the nest from August 30th-November 2nd and mating flights have been documented from

103 September 11-October 8th. We have collected alates from nests in the region on September 9th, 2009, September 11th 2012, August 19th, 2013 and observed flights on October 15th, 2009. Based on these records, we identified the last two sampling periods of our study (Periods 6 and 7, September 5th and 24th) as falling within the reproductive period of M. americana. A. rudis mating flights have been documented from June 10th - July 26th in Ohio (Coovert, 2005) and all alates are typically out of the nest (Northeastern US) by July or mid-August in the (Lubertazzi, 2012). Our own collection records have recorded alates within the nests June 13-14th, 2009. Based on these data we estimate that A. rudis mite abundance and richness should be highest in the 2nd and 3rd sampling periods (June 17th and July 5th).

4 Results Across the 21 M. americana nests, we collected 1762 mites comprising 18 species (colony mean= 6 spp.). Of these, 9 are hosts specific, 5 are generalists, and 4 are of unknown specificity. The Chao richness estimate was 19.6 species, suggesting that we adequately sampled this location for mite diversity associated with M. americana (Figure 1a). There were 182 individual mites collected from the 5 A. rudis nests comprising 11 species of mites (colony mean= 5 spp.). Only 2 of the mite species found on A. rudis are known to be host specific. A. rudis mite diversity was not adequately sampled for a spatial analysis (Chao estimate 17 spp.) (Figure 1b).

4.1 Spatial Autocorrelation of Mite Communities Mite species composition for M. americana showed significant spatial autocorrelation for the overall Mantel test (Mantel r=0.30, p=0.004) (Figure 2a) and Mantel correlograms revealed that spatial dependence occurred at the first distance class (Mantel r=0.23, p=0.001). Univariate spatial autocorrelations of mite abundance and richness using Moran’s I showed no significant relationships with distance between sample pairs (Figure 3).

4.2 Seasonal Shifts Mite abundance in M. americana nests peaked in the second sampling period (June 17th) and was largely driven by a single generalist species, Schwiebea sp1. Total

104 mite abundance did not increase substantially during the reproductive period of the nest (periods 6 and 7), but richness was slightly higher (Figure 4). Generalist mites decreased in abundance during the reproductive period, while specialist mites increased (Figure 5). Although we only were not able to adequately sample the A. rudis mite community for spatial analysis, temporal trends showed a peak in total mite abundance during the first and fifth periods, both due to temporary increases in Schwiebea sp1. Mite abundance decreased during the reproductive period (periods 2-3), but richness increased in period 2 due to the presence of small numbers (1-2 individuals) of specialist taxa (Figure 6). Most of mites associated with A. rudis were generalists (169 of 182 individuals), so any comparisons with host specialist mites were weak.

5 Discussion Our findings indicate that commensal mite community composition in ant nests is structured by distance-decay effects, while synchrony of seasonal population dynamics (richness and abundance) with hosts differed between generalists and specialists. In our homogenous grassland habitat, mite dispersal was limited by host foraging distance and seasonal alate dispersal. Community similarity decayed rapidly beyond the first distance class, and was not highly correlated even at the closest distance class (Mantel r=0.23), suggesting that future investigations would benefit from measuring mite communities for colonies in closer proximity. A meta-analysis of distance-decay studies of community similarity across several taxa showed that effect sizes can vary depending on spatial extent, region, trophic position, dispersal type, latitude, and body size (Soininen et al., 2007). The ant-associated mites in our study are primarily microbivores and have very small body sizes; both of these factors were shown to decrease the spatial extent of distance-decay correlations in the Soininen et al. meta-analysis. Our study suggests that community-level dispersal and seasonal population dynamics of commensal mites are shaped by different processes. In our study we observed seasonal shifts in mite abundance and richness, with no significant spatial autocorrelation of these aggregate variables among ant nests. In previous studies of ant- associated mites, abundance and richness was related to habitat type and surrounding nest communities (Chapter 3), and soil resources, host frequency, and diversity (Chapter 2).

105 Mite abundance and richness is a complex product of host, environment, spatial arrangement, and temporal shifts in the host or environmental template. We found evidence of lifecycle synchrony for phoretic mites associated with M. americana. As we predicted, specialist mites exhibited stronger seasonal synchrony with alate production than generalist mites. Multiple studies of fire ants in Louisiana have documented increases of phoretic mite abundance with alate abundance; however, they did not sample a fixed number of hosts throughout the season, and thereby increased sampling effort at the time of mating flights (Ebermann & Moser, 2008; Moser & Blomquist, 2011; Walter & Moser, 2010). In our study we did not collect alates, but rather sampled a constant number of workers throughout the season, using them as sentinels of the colony cycles. Mating flights in our ant species are very difficult to predict and last only a short time (less than an hour each day), thus it is impractical to collect alates during the mating flights. Previous studies of mite prevalence collected alates while they were within the nest by destructively sampling a section of the nest (Campbell et al., 2013; Uppstrom & Klompen, 2011). Sampling using this method in our study could have caused the ants to leave the nest and seasonal resampling would not have been possible. Our results suggest that mites tend to be phoretic throughout the season on workers, but the relative abundance of mites entering phoretic stages (especially specialist mites) may increase during critical dispersal periods to take advantage of temporary alate abundance. Schwiebea sp1 was the most abundant generalist mite species in nests at this site. The genus is commonly found associated with beetles and other soil dwelling or subcortical insects (OConnor, 2009). Interestingly, this species showed an initial spike in abundance at different times in A. rudis (Period 1) and M. americana (Period 2) nests, but a secondary spike at the same time (Period 5). Different population trends of Schwiebea sp1 in the early periods may be due to seasonal variation between the two host species in terms of brood rearing, nest migration, nest population growth, or host variation in moisture or resource levels within nests. Additionally, A. rudis populations migrate from their winter nests to new locations each spring (Lubertazzi, 2012), a process that would leave behind any mites that are not on hosts. Seasonal shifts in populations of generalist

106 species based on host life history or population dynamics would be an interesting direction for future research in this system. Our results show that the distance-decay relationship applies in homogenous environments for dispersal limited organisms. Additionally we document shifts in seasonal abundance and synchrony for host specific organisms. Future studies would benefit from studying community similarity at closer distances and potentially measuring additional small scale variation in environmental variables to determine if small scale heterogeneity can play a significant role in structuring dispersal limited communities.

107 6 References Albrecht, M., & Gotelli, N. J. (2001). Spatial and temporal niche partitioning in grassland ants. Oecologia, 126(1), 134–141. Bahram, M., Kõljalg, U., Courty, P. E., Diédhiou, A. G., Kjøller, R., Põlme, S., … Tedersoo, L. (2013). The distance decay of similarity in communities of ectomycorrhizal fungi in different ecosystems and scales. Journal of Ecology, 101(5), 1335–1344. Campbell, K. U., Klompen, H., & Crist, T. O. (2013). The diversity and host specificity of mites associated with ants: the roles of ecological and life history traits of ant hosts. Insectes Sociaux, 60, 31–41. Coovert, G. A. (2005). The Ants of Ohio. Columbus, Ohio: Ohio Biological Survey, Inc. Ebermann, E., & Moser, J. C. (2008). Mites (Acari: Scutacaridae) associated with the red imported fire ant, Solenopsis invicta Buren (Hymenoptera: Formicidae), from Louisiana and Tennessee, U.S.A. International Journal of Acarology, 34, 55–69. Eickwort, G. C. (1990). Associations of mites with social insects. Annual Review of Entomology, 35, 469–488. Hölldobler, B., & Wilson, E. O. (1990). The Ants. Cambridge: Harvard Press. Houck, M. A., & OConnor, B. M. (1991). Ecological and evolutionary significance of phoresy in the Astigmata. Annual Review of Entomology, 36, 611–636. Laakso, J., & Setälä, H. (1998). Composition and trophic structure of detrital food web in ant nest mounds of Formica aquilonia and in the surrounding forest soil. Oikos, 81, 266–278. Lindo, Z., & Winchester, N. N. (2007). Resident corticolous oribatid mites (Acari: Oribatida): Decay in community similarity with vertical distance from the ground. Ecoscience, 14(2), 223–229. Lubertazzi, D. (2012). The biology and natural history of Aphaenogaster rudis. Psyche, 2012. Moser, J. C., & Blomquist, S. R. (2011). Phoretic arthropods of the red imported fire ant in central Louisiana. Annals of the Entomological Society of America, 104, 886– 894. Nekola, J. C., & White, P. S. (1999). The distance decay of similarity in biogeography and ecology. Journal of Biogeography, 26, 867–878. OConnor, B. M. (1982). Evolutionary ecology of astigmatid mites. Annual Review of Entomology, 27, 385–409. OConnor, B. M. (2009). Cohort Astigmatina. In G. W. Krantz & D. E. Walter (Eds.), A Manual of Acarology (pp. 565–657; 16). Lubbock, TX: Texas Tech University Press. Oliva, M. E., & González, M. T. (2005). The decay of similarity over geographical distance in parasite communities of marine fishes. Journal of Biogeography, 32(8), 1327–1332. Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O’Hara, R. B., … Wagner, H. (2013). vegan: Community Ecology Package. R package version 2.0-10. Retrieved from http://cran.r-project.org/package=vegan Poulin, R. (2003). The decay of similarity with geographical distance in parasite communities of vertebrate hosts. Journal of Biogeography, 30(10), 1609–1615.

108 Poulin, R., Blanar, C. a., Thieltges, D. W., & Marcogliese, D. J. (2011). The biogeography of parasitism in sticklebacks: Distance, habitat differences and the similarity in parasite occurrence and abundance. Ecography, 34(4), 540–551. Schlick-Steiner, B. C., Steiner, F. M., Moder, K., Bruckner, A., Fiedler, K., & Christian, E. (2006). Assessing ant assemblages: pitfall trapping versus nest counting (Hymenoptera, Formicidae). Insectes Sociaux, 53(3), 274–281. Soininen, J., McDonald, R., & Hillebrand, H. (2007). The distance decay of similarity in ecological communities. Ecography, 30(1), 3–12. Thieltges, D. W., Ferguson, M. a D., Jones, C. S., Krakau, M., De Montaudouin, X., Noble, L. R., … Poulin, R. (2009). Distance decay of similarity among parasite communities of three marine invertebrate hosts. Oecologia, 160(1), 163–173. Tobler, W. R. (1970). A computer movie simulating urban growth in the Detroit region, Ecological Geography, 46, 234–240. Tschinkel, W. R. (2006). The Fire Ants. Cambridge: The Belknap Press of Harvard University Press. Uppstrom, K. A., & Klompen, H. (2011). Mites (Acari) associated with the desert seed harvester ant, Messor pergandei (Mayr). Psyche, 2011, 1–7. Vinarski, M. V., Korallo, N. P., Krasnov, B. R., Shenbrot, G. I., & Poulin, R. (2007). Decay of similarity of gamasid mite assemblages parasitic on Palaearctic small mammals: Geographic distance, host-species composition or environment. Journal of Biogeography, 34(10), 1691–1700. Walter, D. E., & Moser, J. C. (2010). Gaeolaelaps invictianus, a new and unusual species of hypoaspidine mite (Acari: Mesostigmata: Laelapidae) phoretic on the red imported fire ant Solenopsis invicta Buren (Hymenoptera: Formicidae) in Louisiana, USA. International Journal of Acarology, 36(5), 399–407. Warburton, E. M., Kohler, S. L., & Vonhof, M. J. (2015). Patterns of parasite community dissimilarity: the significant role of land use and lack of distance-decay in a bat- helminth system. Oikos.

109 Supplementary Figure 5. Distance classes for nest spatial sampling methods. Example of a nest (blue triangle) and relative comparisons are shown for each of 6 distance classes (colored circles).

110 Figure 1. Species accumulation curve for mite richness in a) Myrmica americana nests and b) Aphaenogaster rudis nests. Chao estimated richness is indicated by the dotted line, and observed richness is shown with a dashed line.

111 Figure 2. Mite composition in Myrmica americana nests as determined by nest proximity. (a) Mantel test of mite community dissimilarity by Euclidean distance dissimilarity, and (b) Mantel correlogram of mite species composition by distance class. The horizontal red line indicates the break between positive and negative correlations. The diamond point represents a significant relationship between mite community and distance at the closest distance class.

112 Figure 3. Moran’s I correlograms showing tests of spatial autocorrelations of (a) mite abundance and (b) richness in Myrmica americana nests. The horizontal red line indicates the break between positive and negative correlations. Tests at all distance classes were not significant.

113 Figure 4. Seasonal abundance and richness of mites associated with Myrmica americana nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 6 and 7).

114 Figure 5. Seasonal abundance and richness of (a) specialist and (b) generalist mites associated with Myrmica americana nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 6 and 7).

115 Figure 6. Seasonal abundance and richness of mites associated with Aphaenogaster rudis nests. Each horizontal line indicates the abundance of each mite species within one of the three taxonomic groups (Astigmata (blue), Heterostigmata (pink), Mesostigmata (yellow)). Numbers at the peaks of each sampling period indicate the total species richness. Reproductive period of the nest is shown between dotted lines (periods 2 and 3).

116 General Conclusions

This dissertation examined how patch and landscape level factors affect ant diversity in constructed grasslands and how their roles as regulators of mite diversity can shift depending on the environmental context. I found that constructed grassland ant communities and ant-dependent organisms are structured at multiple spatial and temporal scales. Understanding the process of community assembly for organisms that regulate diversity both aboveground and belowground in constructed preserves, can allow land managers to make better informed decisions for management of biodiversity in isolated conservation patches. In these grasslands, ant communities were primarily structured by time since establishment, soil texture, time since management (burning), and the surrounding land use and land cover (Patch and Landscape Levels red and purple text, Figure 1; Chapter 1). Ant-associated mites are dispersal limited, and must colonize the grasslands on the bodies of their hosts. Their sensitivity to small spatial scales and reliance on hosts allowed me to test for the effects of host environmental context on mite biodiversity (Patch Level blue arrows, Figure 1; Chapter 2). I found that mite diversity on multiple host species within a site is a complex product of both patch and host level characteristics of the site, especially site area, litter depth, soil texture, management (Patch level blue and purple text, Figure 1; Chapter 2), ant richness, host identity, and host frequency (Host Level blue text, Figure 1; Chapter 2). I found that spatial autocorrelation among adjacent nests indicates dispersal limitation for mites in a manner comparable to island communities (Nest Proximity, Figure 1; Chapter 4). These mites rely on both their hosts and belowground resources in the soil of their hosts’ nests, making them useful organisms for applications of resource theory and parasitology theory (Nest Neighborhood, Figure 1; Chapter 3). Lastly I found that ant-associated mites are also affected by the ecological neighborhood at multiple spatial scales including the landscape of an individual host’s body, variation among members of the colony, the neighboring ant community, and the habitat type (Patch Level, Nest Neighborhood, Figure 1; Chapter 3).

117 Throughout this body of work I attempted to tease apart the relative roles of host and habitat characteristics for mite diversity. Host identity was an important predictor for mite community composition suggesting that there are many host-specific mites; however, phoretic mite diversity was highest on host species that were large, wide ranging, and frequent within the site, suggesting that generalist mites, are reliant on widely available species for dispersal and that there may be a degree of functional redundancy in ant hosts (Chapter 2). The importance of common hosts has also been well documented in parasite literature (Harris & Dunn, 2010; Lindenfors et al., 2007; Poulin, 2007). A relatively unexplored factor affecting diversity of host-dependent organisms is the role of habitat, or environmental context, of the host (Krasnov et al., 2008). Across all three mite chapters (Chapters 2-4), I found that the context of the host, both spatially and temporally, can affect the mite community composition, richness, and abundance. My findings support broader use of ants as environmental indicators of disturbance recovery and soil texture and show that diversity in constructed grasslands is structured by both patch and landscape level processes (Chapter 1). I found that landscape level processes were important predictors of ant frequency for 12 of the 14 ant species, and that species often responded in contrasting ways to the surrounding landscape and patch level characteristics. My work promotes ants as environmental indicators, but notes that aggregate measures of ant diversity (species richness) may not be effective in documenting the more subtle shifts in the ant community that result from the differential sensitivity of ants to habitat disturbance, and make them so valuable as indicator organisms. I demonstrate the need for developing functional groups or habitat conservatism indices for the Midwest and Eastern US for better assessment of habitat recovery using ants (Andersen & Majer, 2004; Peters et al., in review). Future work in this system could contribute to a deeper understanding of host context and phoretic mite interactions by: (i) comparing associates of a single host species across multiple habitat types, in chronosequences, at edges of their ranges, or where the host is locally abundant versus regionally abundant; (ii) comparing associates in nutrient poor and nutrient rich soils (for example unglaciated and glaciated, or experimentally manipulated) and determine if the “host ant effect” is more important under nutrient poor conditions; (iii) studying mite communities associated with ants in

118 remnant prairies, and identify if similar principles apply in these more stable (less disturbed and age dependent) systems; (iv) identify functional group shifts in ant communities and identify different stable states in constructed grassland recovery; (v) destructively sampling ant colonies to link phoretic mites with non-phoretic stages in the colonies; and (vi) conducting laboratory manipulations to understand the role of mites within ant nests, host switching, and if host specificity is primarily an artifact of low dispersal ability or due to particular resources provided by the host. This body of work demonstrates the importance of ants as diversity regulators and the utility of host-dependent organisms for measuring fine-scale levels of habitat change. Together, these studies demonstrate that processes at multiple temporal and spatial scales contribute to biodiversity and community assembly within constructed grassland patches and that the context of hosts can modify their roles as biodiversity regulators.

119 1 References

Andersen, A. N., & Majer, J. D. (2004). Ants show the way down under: invertebrates as bioindicators in land management. Frontiers in Ecology and the Environment, 2, 291–298. Harris, N. C., & Dunn, R. R. (2010). Using host associations to predict spatial patterns in the species richness of the parasites of North American carnivores. Ecology Letters, 13(11), 1411–1418. Krasnov, B. R., Korallo-Vinarskaya, N. P., Vinarski, M. V, Shenbrot, G. I., Mouillot, D., & Poulin, R. (2008). Searching for general patterns in parasite ecology: host identity versus environmental influence on gamasid mite assemblages in small mammals. Parasitology, 135(2), 229–242. Lindenfors, P., Nunn, C. L., Jones, K. E., Cunningham, A. a., Sechrest, W., & Gittleman, J. L. (2007). Parasite species richness in carnivores: Effects of host body mass, latitude, geographical range and population density. Global Ecology and Biogeography, 16(4), 496–509. Peters, V. E., Campbell, K. U., Dienno, G., García, M., Leak, E., Loyke, C., … Crist, T. O. (2015). Plant evenness indicates conservation value of constructed grasslands over quality assessment indicators. Ecological Indicators, in review. Poulin, R. (2007). Are there general laws in parasite ecology? Parasitology, 134(Pt 6), 763–776.

120 Figure 1. Conceptual diagram summarizing important landscape, patch and host level (outer divisions) variables for measures of mite diversity (blue text and arrows), ant diversity (red text and arrows), and both taxa (purple text). Numbered circles represent chapters testing the relationship.

121