Citation
On the Community Structure of Ground-Dwelling Ants (Hymenoptera: Formicidae) in the Sandhills of Florida

Material Information

Title:
On the Community Structure of Ground-Dwelling Ants (Hymenoptera: Formicidae) in the Sandhills of Florida
Creator:
SPIESMAN, BRIAN J. ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Ants ( jstor )
Body size ( jstor )
Community associations ( jstor )
Community structure ( jstor )
Ecology ( jstor )
Ecosystems ( jstor )
Land cover ( jstor )
Landscapes ( jstor )
Pine barrens ( jstor )
Species ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Brian J. Spiesman. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
6/30/2007
Resource Identifier:
659563277 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

ON THE COMMUNITY STRUCTUR E OF GROUND-DWELLING ANTS (HYMENOPTERA: FORMICIDAE) IN THE SANDHILLS OF FLORIDA By BRIAN J. SPIESMAN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006 1

PAGE 2

Copyright 2006 By Brian J. Spiesman 2

PAGE 3

ACKNOWLEDGMENTS I would first like to thank my committee: Graeme Cumming, Jane Southworth, and Bob Holt. I especially thank Graeme for his support a nd guidance. I appreciate the friendly advice and comments provided by Serwan Baban. Thanks go to the folks of the Cumming and Oli lab for good times in 218. I thank Ann George, in particular, for our ma ny discussions that contributed to my research. Emilio Bruna grac iously provided lab space, a microscope, and useful project advice. I thank the Bruna lab fo r accommodating me and my ants, especially Matt Trager for interesting ant talk a nd Antnio Aguiar Neto for great coffee. I appreciate the help that Craig Allen and Craig Stow provided for the body size analysis. Shunpei Iguchi and Kelley Anderson assisted with field wor k. Kurt Larson and Jill Vander berg helped sort out ants. Thanks go to Rico Holdo for statistical advice a nd, more importantly, for keeping me busy in the water, on the bike, and on the road. I sincerely thank the staff of the Department of Wildlife Ecology and Conservation for all of the help they provided, especially Caprice McRae, Monica Lindberg, Delores Tilman, and Sam Jones. Funding for this research was provided by a gr ant awarded to G. Cumming from the US Department of Agriculture T-STAR program, and a grant from the University of Florida Tropical Conservation and Development program. Access to sampling sites and logistical support was provided by Steve Coates and the Ordway-Swisher Biological Station, th e Florida Division of Forestry, the Florida Park Servi ce, and Florida Rock Industries. I wish to thank my family. I deeply appr eciate the continued support and encouragement of my mother, Jan, and father, Phil, as well as Tim Laughlin and Jan Howell-Spiesman. Lastly, and most importantly, I thank Tania. 3

PAGE 4

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................3 LIST OF TABLES ...........................................................................................................................6 LIST OF FIGURES .........................................................................................................................7 ABSTRACT .....................................................................................................................................8 CHAPTER 1 INTRODUCTION ................................................................................................................10 Spatial and Environmental Dependence in Ant Communities ..........................................12 Body Size Distributions .....................................................................................................13 2 ANT COMMUNITY RESPONSE TO SPATIAL AND ENVIRONMENTAL VARIABILITY AT MULTIPLE SCALES ..........................................................................17 Introduction ........................................................................................................................17 Study Area and Methods ....................................................................................................20 Study Area ..................................................................................................................20 Site Selection ..............................................................................................................21 Field Sampling and Identification ..............................................................................22 Analysis ......................................................................................................................23 Quantifying environmental variables ................................................................23 Quantifying spatial variables .............................................................................24 Statistical analyses ............................................................................................25 Results ................................................................................................................................27 Ant Sampling and Identification .................................................................................27 Variable Selection and Enviro -spatial Variance Partitioning .....................................28 Spatial Dependence ....................................................................................................29 Environmental Dependence ........................................................................................30 Discussion ..........................................................................................................................31 Conservation and Management Implications .............................................................35 Conclusions .................................................................................................................36 3 ANT BODY SIZE DISTRIBUTION S: INVESTIGATING SOME PREDICTIONS OF THE TEXTURAL DISCONTINUITY HYPOTHESIS ......................................................................................................................50 Introduction ........................................................................................................................50 Methods ..............................................................................................................................52 Results ................................................................................................................................55 Discussion ..........................................................................................................................57 4

PAGE 5

4 CONCLUSIONS ...................................................................................................................67 APPENDIX: SITE SELECTION ..................................................................................................70 LIST OF REFERENCES ...............................................................................................................74 BIOGRAPHICAL SKETCH .........................................................................................................82 5

PAGE 6

LIST OF TABLES Table page 2-1 Site locations ......................................................................................................................38 2-2 Site selection variables .......................................................................................................39 2-3 Pool of environmental variables ........................................................................................40 2-4 Species list. ........................................................................................................................42 2-5 Spatial dependence at broad, medium, and fine spatial scales ..........................................44 3-1 Pool of landscape, lo cal, and spatial variables ...................................................................61 3-2 Proportion of environmental va riation explained by size group ........................................63 6

PAGE 7

LIST OF FIGURES Figure page 2-1 Map of study sites ..............................................................................................................45 2-2 Venn diagram illustrating the fractions of variation in ant communities ..........................46 2-3 Spatial structure at broad, medium, and fine spatial scales ...............................................47 2-4 Redundancy analysis biplot ...............................................................................................48 2-5 Species response curves .....................................................................................................49 3-1 Fractions of variation explained in partial redundancy analysis........................................64 3-2 Body size groupings ...........................................................................................................65 3-3 Venn diagrams illustrating the fractions of variation by body size ...................................66 A-1 Distribution of site s across the range of landscape heterogeneity .....................................71 A-2 Indicator species analys is of landscape groupings .............................................................73 7

PAGE 8

Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science ON THE COMMUNITY STRUCTUR E OF GROUND-DWELLING ANTS (HYMENOPTERA: FORMICIDAE) IN THE SANDHILLS OF FLORIDA By Brian J. Spiesman December 2006 Chair: Graeme S. Cumming Major Department: Wildlife Ecology and Conservation The increasing threat of habitat loss driven by land cover change has effects on biodiversity at multiple scales. Though the factors that influence community structure in ants have been extensively studied at local scales and at biogeographical scal es, few attempts have been made to investigate the effects on ant comm unities at intermediate or landscape scales. This leaves a conspicuous gap in our understa nding of the factors th at control community organization in ants. In this study, ant communities were sampled in sandhill habitat at thirtythree locations throughout northern Florida, USA. Sites we re standardized for local habitat quality but spanned most of the gradient in heterogeneity of the surrounding landscape. These data were used to (1) explore how spatial and environmental fact ors influence the ways in which ant species organize into local communities, a nd (2) investigate how body size distributions are influenced by environmental variation. Though species richness was not significantly influenced by the measured environmental variables, community composition showed strong spatial and environmental dependencies across a range of scales throughout the region. Results were placed in the context of metacommunity theory. The community patterns may best be de scribed by a combination of the mass effects and species sorting perspectives on metacommunity dynamics since both spatial and environmental 8

PAGE 9

processes seem to be driving patterns of commun ity change. Open habitat specialists appear to be replaced by habitat ge neralists in landscapes with high pr oportions of matrix habitat even though local habitat conditions are of similar quality. The results illustrate the importance of considering multiscale influences on patterns of organization in ant communities. Furthermore, results show how heterogeneous landscapes ar e important for maintain ing a diverse regional species pool, but continued degradation of sandh ill habitat may reduce the viability of some open habitat specialists and the ecosystem services that they provide. The textural discontinuity hypothesis (TDH) postulates that body size distributions are discontinuous and reflect the discontinuous nature of structuring forces in ecosystems. Worker ant body mass was used to explore the validity of the TDH. Though the ants in this study exhibit a discontinuous body size distri bution, little support was found for the TDH as a means of explaining patterns in ant body size distributions. Body size distributions did not change with the structure of the landscape as predicted. This lack of support may result from a mismatch between the scale of analysis and the scale of environmental interaction that provide the strongest evolutionary for ces on worker ant body size. 9

PAGE 10

CHAPTER 1 INTRODUCTION Among the factors contributing to the current bi odiversity crisis, habitat loss driven by land-use change is consistently cited as a principal cause (D idham et al. 1996, Pimm and Raven 2000, Novacek and Cleland 2001). Habitat fragmentat ion is a typical resu lt of anthropogenic habitat destruction and can have profound effect s on biodiversity (Fahrig 2003). Fragmentation produces a landscape comprised of smaller and mo re isolated patches, incapable of supporting that landscape’s previous complement of orga nisms. As populations become increasingly affected by these changes in landscape structure, species will continue to disappear along with the stability they instill in ecosystems (Rooney et al. 2006). Insects are highly vulnerable to habitat lo ss though they receive little attention in conservation planning (Samways 2005, Rahn et al. 2006). Ants are no exception. In the United States there is no species of ant (or Hymenoptera for that matte r) listed as threatened or endangered on any state or federal list. Astonishingly, the state of California explicitly prohibits not only insects, but invertebrate s in general, from protection under its state endangered species act. This oversight is startling, given that invertebrates represent an estimated 90% of all animal species (Wilson 1992b) and perform a diverse array of functions critical to all forms of life on Earth (Kremen 2005, Samways 2005). Two of the ma in reasons why insect s garner such little attention in conservation planni ng are their negative public imag e and a relative lack of study. The general public too often views insects and other inverteb rates with aversion, fear, and ignorance (Kellert 1993). This renders justification for expensive c onservation initiatives difficult at best for all but the most charismatic of insects (e.g., butterflies). We are often relegated to relying on umbrella-like conservatio n programs to ensure th e persistence of rare insect species. However, since insects often have complex life cycles and associations with other 10

PAGE 11

species, even the most effective means for the general preservation of biodiversity, such as multiscale adaptive management programs, may not meet the specific needs of many imperiled insect species. The tremendous diversity of insects is itself a nother cause for their l ack of representation in conservation planning. We know very little about the distribution, life history, and ecology of the vast majority of insect species. Dunn (2005) notes that there have been only 70 documented insect extinctions in modern times as compared with 129 extinctions of birds, not because insects are any less vulnerable than birds, but simply because we haven’t been paying attention. If insects and birds share similar rates of extin ction, we may have missed roughly 44,000 insect extinctions in the last 600 years (Dunn 2005). Since habitat loss and fragmentation, as well as management for the ensuing effects, typically occur at landscape scales it is im portant to understand how communities of insects respond to variation at this scale. The fact ors that contribute to the organization of ant communities have been extensively studied al ong local gradients of disturbance, soil characteristics, and habitat complexity (Hoffmann and Andersen 2003, Izhaki et al. 2003, Lubertazzi and Tschinkel 2003, Boulton et al. 2005, Lassau et al. 2005). Many broad scale influences (i.e., regional to bi ogeographic scale) on ant communiti es have also been extensively studied (Cushman et al. 1993, Kaspari et al. 2000 , Gotelli and Ellison 2002, Kaspari et al. 2003). However there remains a conspicuous gap in our knowledge of the influences of intermediate or landscape scale variation on ant co mmunity structure (but see Suarez et al. 1998). Habitat loss and fragmentation are problems associated with landscape scale heterogene ity, yet most studies of fragmentation effects on ant communities are limited to replicate fragments and do not explicitly consider the struct ure and composition of the entire landscape (e.g., Bruna et al. 2005, 11

PAGE 12

Ribas et al. 2005). For ants in particular, community response to variation in the composition and spatial arrangement of the surrounding landscape has been inadequately studied. Ants are well-suited for community level studies. Logistically, they can be quickly and cheaply sampled, and they are generally well-d escribed taxonomicall y, easing the task of identification. Ecologically, ants comprise a highly diverse family of insects with over 11,000 described species that represent a third of the global insect bi omass (Hlldobler and Wilson). They are an interesting group that function as im portant regulators of arthropod populations, seed dispersers, scavengers, and on a global scale, they aerate and enrich more soil than do earthworms (Hlldobler and Wilson 1990, Wilson 2000). Ants are also inex tricably linked to plant communities through mutua lisms, herbivory, and parasitic re lationships (Gauld and Bolton 1988). Ants exhibit strong competitive interac tions, which can drive community structure and therefore make ideal study subjects when examining the influence of environmental and spatial processes on community organization. Spatial and Environmental Dependence in Ant Communities Modern community ecology is greatly influenc ed by the research Gary Polis conducted on food webs in the islands of the Sea of Cortez (see Polis et al. 2004 and references therein). Traditional approaches to the study of ecol ogical communities often fail to recognize the importance of openness in ecosystems. Flows of organisms and resources across habitat boundaries can have tremendous in fluences on the outcome of local community dynamics. For example, spatially subsidized pred ators can depress the populations of in situ prey well below what is otherwise expected, if a predator’s fitness is enhanced by flows of resources from adjacent habitats (Polis et al. 1997). Such effects can be transmitted through the food web, yielding consequences for the rest of the community. The composition and spatial structure of the surrounding landscapes is therefore highly significant for local community interactions. 12

PAGE 13

The development of landscape ecology in th e last two decades provides a means of addressing some of the ideas inspired by Polis and colleagues. However, landscape ecology might be enhanced by better incorporating some of the themes of co mmunity ecology. The Metacommunity concept may provide a strong er bridge between th e two disciplines. Metacommunities are sets of local communities linked by dispersal of multiple species (Wilson 1992a, Leibold et al. 2004). The concept expands traditional community ecology and metapopulation ecology to provide a more complete framework for the analysis of the multiscale processes that influence community dynamics (Holyoak et al. 2005). One of the main features of the metacommunity concept is that it emphasizes th e role of spatial fact ors in the outcome of community level interactions. This has implicat ions for the study of fragmented landscapes, in that it facilitates mechanistic explanations for patterns of community dynamics at landscape scales. In Chapter 2, the influences of local and landscape level variation on the organization of ant communities are examined at thirty-three sites in the sandhills of northern Florida, USA. The following questions are addressed: what are the relative importance of environmental and spatial dependence on ant community structure? What ar e the important structur ing features in sandhill landscapes for ant communities? Results are plac ed in a metacommunity context to explore how this new framework can be used to explain patterns in sandhill ant communities. Body Size Distributions Because the size of an organism primarily de fines how it will perceive and interact with its environment, the factors that influence the distribution of body size have been a subject of great interest to ecologists. Animal body size has traditionally been viewed as forming a unimodal distribution, with body sizes evenly spaced along a continuum by competitive interactions and niche partitio ning. More recent evidence indicat es that body size sometimes 13

PAGE 14

forms a multimodal, discontinuous distribution (Holling 1992, Allen et al. 1999, Scheffer and van Nes 2006). A number of theori es have been put forth in an effort to describe the overall pattern that determines the shape of body size dist ributions, each with its own set of predictions (Allen et al. 2006). The energetic hypothesis (Brown et al. 1993) appl ies at continental scales and predicts a unimodal distribution of body size within a partic ular taxon, which reflects the convergence of species on an optimal mass at which organisms a llocate energy most efficiently under allometric scaling laws. Interspecific competition disperses mass evenly around this modal value. Multiple modes are possible, however, if resource variabil ity is scale-dependent (Marquet et al. 1995). The phylogenetic hypothesis also applies at continen tal scales and suggests that body size distributions are structured by the evolutionary hi story of a group. Within an area, species of the same lineage are evolutionarily limited in terms of size by their ancestral forms. Multiple modes may arise at this scale through mechanisms pr oposed in the energetic hypothesis, through the convergence of taxa with different evolutionary histories (Marquet and Cofre 1999), or within lineages through stochas tic evolutionary processes (C umming and Havlicek 2002). A set of biogeographical hypotheses apply to scales below that of en tire continents and predict that body size distributions will be determined by the spat ial characteristics of a region and the ability of species to di sperse across geographical barriers . For example, two sets of species on opposite sides of a geographica l boundary may exhibit different body size distributions through mechanisms predicted by on e or both of the preceding hypotheses, and/or through community level sorting. The textur al discontinuity hypothesis (TDH; Holling 1992) predicts a discontinuous, multimodal distribution of body size. An important component of this hypothesis is the assertion that ecosystems are structured by a few important processes that 14

PAGE 15

operate within a particular scale. A discontinuous size distribution arises as species of similar body size are entrained to environm ental variation at their scale of perception and interaction. The appearance of discontinuities, or gaps, in the distribution results from a change of scale in environmental interaction. As Allen and colleagues (2006) point out, the TDH applies to scales below the hypotheses previously ou tlined, and therefore may be subject to these prior influences. Lastly, community interactions hypotheses pred ict that numerous local processes, such as resource limitation and selective predation, organize body size distributi ons (Hutchinson 1959). Traditional views on competition and maximizati on of niche partitioning predict a continuous unimodal distribution unless resources themselves are clumped. There is clearly great overlap in how the five hypotheses explain patterns in body size distributions but, as Allen and others (2006) illustrate, each predicts specific, though sometimes similar, responses to the composition and structur e of the size distributi on with changes in the surrounding environment. One reason for the lack of consensus results from confusion as to what scale the different hypothese s should be applied. Moreover, preexisting effects imposed by larger scale processes can conf use interpretation of results ob served at smaller scales. Consequently, the expression of effects predicted by any one hypothesis (or at one scale) can influence the expression of eff ects predicted by another. The que stion of what determines the structure of body size distributions therefore demands a multiscale answer and cannot be fully explored otherwise. In Chapter 3, body size data (King 2004) is applied to the sandhill ants described in Chapter 2. Very few field-based studies with detailed measurements of local and landscape environmental variation have been conducted to test the predictions of the TDH, and none involving ants. Ants are well suited for this type of analysis sin ce they exhibit a wide range of 15

PAGE 16

body sizes and are highly adapted to their environment. A focus is therefore placed on the TDH to determine how well observed an t community patterns fit the predictions of the TDH. For example, do sandhill ants exhibit a discontinuous distribution of body size? If so, are size groups influenced by different components of their envi ronment? Do size distributions change with variation in landscape structure? Chapter 4 follo ws to provide both a summary of the important results from the earlier chapte rs and some concluding thoughts. 16

PAGE 17

CHAPTER 2 ANT COMMUNITY RESPONSE TO SPATIAL AND ENVIRONMENTAL VARIABILITY AT MULTIPLE SCALES Introduction Community ecologists have traditionally viewed their system of intere st as closed with community dynamics driven mainly by local-scale processes (Ricklefs 1987). It has since been recognized that systems are open and embedded in a multiscale network of biotic and abiotic flows (Wiens 1989, Holt 1993, Polis et al. 1997, Polis et al. 2004, Rand et al. 2006). The movement of organisms and resources across habitat boundaries can alter the outcome of competitive interactions (Holt and Barfield 2003), predator-prey dynamics (Henschel et al. 2001), and system productivity (Polis et al. 1997). The magnitude and direction of these crosssystem flows are influenced by the structure of the landscape mosaic in which they occur (Cadenasso et al. 2004). The composition and sp atial arrangement of environmental features across scales are therefore fundamental determinants of community organization. The effects of landscape structure on populat ion level attributes such as presenceabsence, abundance, fitness, or persistence are wide ranging and extensiv ely studied (reviewed in Debinski and Holt 2000, Fahrig 2002, 2003). However, the effects on entire communities, especially species-rich communities, are less well studied and understood. A detailed literature is emerging on the effects of landscape contex t on arthropod communities in agroecosystems (Tscharntke et al. 2005, Rand et al. 2006). These studies show that the amount of natural habitat in the surrounding landscapes is important in determining local community dynamics in agricultural areas. Ecosystem services, such as pollination and the top-down regulation of agricultural pests, are locally enhanced thr ough cross-system spillover, which increases the abundance and diversity of beneficial species wh en the surrounding landscape contains greater proportions of natural habitat (Kremen et al. 2002, Steffan-Dewenter 2003 , Thies et al. 2003). 17

PAGE 18

Frequent and intense disturbance in agricultural fields creates great opportunity for colonization from surrounding areas, so strong effects of la ndscape context might be anticipated in this situation. However, there remains an inad equate understanding of how communities in heterogeneous natural landscapes under natural di sturbance regimes respond to the structure of the surrounding landscape. Community response to this type of spatia l and environmental va riability may reveal metacommunity structure. Metacommunities are defined as sets of local communities linked by dispersal of multiple species (Wilson 1992a, Holt 1993, Leibold et al. 2004). The metacommunity concept is a framework for expanding community and metapopulation analysis beyond the effects of single scale (i.e., local ) drivers and assumes a multiscale model of community dynamics. At the heart of the theo ry is the recognition that the movement of propagules between patches and/or habitats within a broader cont ext can affect local community processes, and vice versa. There are four pe rspectives on the organization and dynamics of metacommunities within this co ncept (Leibold et al. 2004, Holyoa k et al. 2005). (1) The patch dynamics perspective is related to metapopula tion theory (Levins 1969) and the equilibrium theory of island biogeography (MacArthur and W ilson 1967). The perspective assumes that organisms perceive patches as identical islands where species undergo st ochastic extinction. Immigration and emigration proceed based on the spatial dynamics of the landscape. (2) In the species-sorting perspective, patc hes are not identical and comm unity structure is driven by environmental gradients. Disper sal coupled with niche partitioning based on local environmental differences is important for coexistence, but unr elated to pure spatial effects. (3) The masseffects perspective emphasizes spatial relationships in he terogeneous environments. Immigration and emigration can proceed in spatia lly linked habitats so that species may avoid 18

PAGE 19

exclusion from areas in which they are poor co mpetitors or heavily predated upon (i.e., by the rescue effect; Brown and Kodric-Brown 1977). Source-sink dynamics play heavily into this perspective. (4) The neutral perspective assumes th at species do not differ in traits or capabilities that affect fitness. Community structure is determined by a combination of speciation and the loss of species from local habitats through a series of random walks, both of which are influenced by spatiotemporal variation (Hubbe ll 2001). The metacommunity concept may be an insightful framework for attaining a better unders tanding of patterns of ant community dynamics and species coexistence across mu ltiple spatial scales. Ant communities are linked by dispersal across habitats but it is unknown how they rela te to one or more of these perspectives. The ants of Florida form a diverse fauna whose taxonomy and stat ewide distributions have been well studied (Van Pelt 1958, Deyrup and Trager 1986, Deyrup et al. 1988, Lubertazzi and Tschinkel 2003, King 2004). In a recent update, Deyrup (2003) catalogs 218 species of ants in Florida: the largest statewide assemblage of ant species in eastern North America. The diversity of ants in Florida a nd worldwide is due, in part, to a spectacular array of social behaviors that allow for local coexistence through niche part itioning (Hlldobler and Wilson 1990). For example, Camponotus socius and C. floridanus temporally parti tion a shared food resource through diurnal and nocturnal foragi ng respectively. Some harvester ant species mediate competition for food by foraging on different sizes or densitie s of seeds (Davidson 1977a, b). Species coexistence at broader spatia l scales can be achieve d through trade-offs in performance within particular habitats. For instance, if the competitive hierarchy changes with habitat type, a spatially heteroge neous landscape can be important for maintaining coexistence at landscape and regional scales . This broader-scale mechanism of coexistence can also facilitate local coexistence through source-sink dynamics (Chesson 2000, Kneitel and Chase 2004, 19

PAGE 20

Mouquet et al. 2005). Though local (within-habitat) spatial heterogeneity is known to affect ant community dynamics (Yu et al. 2001, Palmer 2003) , the effects of spatial heterogeneity of different habitats in the surr ounding landscape has been inadequa tely investigated in ants communities. This study focuses on the sandhill habitat of northern Florid a, USA. The response of local ant communities to spatial and environmenta l variability in the local environment and the surrounding landscape is examined by addressing the following questions: (1) what is the relative importance of environmental and sp atial dependence in th e organization of ant communities? (2) Once this framework is es tablished, how, specifically, are ant communities spatially structured across a range of scales? (3) How are ant communities structured by local and landscape level environmental variation? (4) Do sandhill ant communities exhibit metacommunity structure? Study Area and Methods Study Area The study took place on the sandy ridges of nor thern Florida (Fig. 2-1). Sandhill, a rolling savanna-like ecosystem, is descri bed by an open canopy of longleaf pine (Pinus palustris ) and scattered oak species. The understory is a sparse, yet divers e mix of wiregrass ( Aristida stricta ) and other perennial herbs wi th few shrubs (Meyers 1990). Bare patches of the well drained sandy soil are not uncommon. Sandhill s support a diverse biota including many endemic and threatened species (Myers 1990). Freque nt low-intensity ground fi re is critical to maintaining ecosystem integrity and the open structure of the habitat (Myers 1990). Many species are therefore adapted and depe ndent on this frequent fire regime. Sandhills were once more common throughout the southeastern United States. Much of the habitat, however, has been severely degr aded or destroyed in recent decades under 20

PAGE 21

anthropogenic pressures (Noss et al. 1995). Neighboring habitat includes a mix of natural and human modified land cover. Natural land cove r includes mixed hardwood-pine forests, pine flatwoods, freshwater marsh, open water, and fore sted swamp habitats (Myers and Ewel 1990). The forest types mentioned here are structurally denser than sandhill habitat with poorly drained mesic to hydric soil. Modified land cover include s urban areas, pasture, agriculture, extractive mining (mostly sand), commercial pine plantatio n, and disturbed natural land cover types. A detailed land cover classification conducted by the Florida Fish and Wildlife Conservation Commission (FWC) was used to measure attribut es of these land cover types. The FWC map was derived from Landsat ETM imagery acquired in 2003 and contains 43 la nd cover classes at a resolution of 30 m (Stys et al. 2004). Natural pi ne flatwoods and commercial pine plantations are confused in this classification and were ther efore combined into one class (pineland) by Stys and colleagues (2004). These are two structurally similar habitats with dense understories and poorly drained acidic soil (Abrahamson and Hartnett 1990). Site Selection Thirty-three sample sites (Table 2-1, Fig 2-1) were se lected to span a range of variation in landscape context while standard izing for local habitat quality. Sites were chosen from a selection pool of 2000 points randomly generated in sandhill habitat base d on the FWC map. All points were located at least 20 m from the habitat boundary to minimi ze edge effects. Circles of radii 100, 300, 500, and 1000 m were centered on each point and landscape metrics (Table 2-2) were measured in a GIS environment at each spatial extent to determine the range of heterogeneity in Florida landscapes. Ar cView 3.3 (ESRI 1999) was used for all GIS applications. Clustering analysis was performed on these data in order to assemble the selection pool into relatively homogeneous groups that re present the full range of heterogeneity in 21

PAGE 22

landscape structure ac ross the study region. A preliminary list of sites was chosen from each of these groups. To standardize sites for high quality sandhill habitat locally, each site was validated in the field. The requirement of sta ndardized local habitat pr ecluded the sampling of some landscape types. For example, highly degraded landscapes, containing very high proportions of urban land cover, had poor local habitat quality and consequently were not sampled. The selected sites span most of the gradient in landscape heterogeneity within the study region (see Appendix). All sites were spatiall y referenced at the cent er point to within 8 m using a Garmin 76S handheld GPS unit. The sample sites span a di stance of over 250 km, covering most of the north-south extent of sandhill on the Florida peninsula. Field Sampling and Identification Ground-dwelling ants were sampled by two met hods: pitfall trapping and litter extraction using Winkler sacks. Pitfalls were made fr om 53 mm diameter sample cups (SAMCO, 90 ml), buried so the lip of the cup wa s flush with the ground, and fill ed two-thirds full with a 50% solution of propylene glycol and a small amount of dish soap. At each site, sixteen pitfalls were arranged in a square grid of 4 rows and 4 columns with rows and columns separated by 10 m. Pitfalls were deployed for 72 hours during period s without precipitation. Trap contents were filtered into 95% ethanol for later sorting and iden tification. After pitfalls were retrieved, five litter samples were collected at the four corners and center of each site. All litter within a 1 m 2 area was quickly gathered and the course material sifted out in the field. The fine siftate was then hung in Winkler sacks for 48 hours and ants were allowed to accumulate in cups of 95% ethanol following the methods outlined in Agosti et al. (2000). Each site was sampled twice between May and October 2005. 22

PAGE 23

All ant specimens, excluding males and queens, were identified to species by B. J. Spiesman. Since ant species have different foraging strategies, raw abundance data can misrepresent the relative abundance of a given specie s. To alleviate this bias, the occurrence of ant species (number of times a sp ecies occurred in a trap) was used as a measure of relative abundance at each site. Occurrence data wa s Hellinger-transformed, following Legendre and Gallagher (2001), for all analyses to account for the many zeros in the species data table that result with species turnover between sites. Rarefaction curves were drawn for each site to determine the adequacy of sampling using the program EstimateS v7.5.0 (Colwell 2005). To assess local habitat characteristics, the percent cover of vegetation and bare ground was estimated within a 1 m 2 quadrat centered on each pitfall tr ap location. Cover estimates for grass, litter, other herbaceous plants, shrubs, trees, dead w ood, and bare ground were recorded before ant sampling and rounded to the neares t 5%. Any portion of canopy cover over 3 m in height was not included in the estimate of tree cove r since large trees are sparsely distributed and any shading resulting from vegetation at that hei ght would originate from areas outside of the 1 m 2 area. Analysis Quantifying environmental variables Landscape metrics were classified as either composition or arrang ement variables. Composition refers to the identity and amount of land cover types, whereas arrangement refers to the shape and spatial position of landscape featur es. These variables were quantified for the 33 sample sites within circular areas of ra dii 100, 250, 500, 750, and 1000 m centered on each site. Landscape composition variables we re measured as the area (m 2 ) of each land cover type and the total number of land cover classes within each ci rcular area. The Patch Analyst extension for ArcView 3.3 was used to calculate landscape arrange ment variables for all classes combined as 23

PAGE 24

well as sandhill habitat alone. The normali zed difference vegetation index (NDVI) is a continuous variable and was used as a measure of landscape productivity and heterogeneity. The NDVI data used in these analyses were derived fr om the MODIS satellite platform (Carroll et al. 2005). Each scene is a 16-day composite comprise d of 8 bit data with a resolution of 250 m. Three scenes representing the early (May 9-24), middle (July 12-27), a nd later (Oct. 16-31) portion of the sampling period were averaged to obtain a mean value product for the 2005 sampling period. Three sites in the July scen e were obscured by clouds and therefore masked and replaced with data from the period immediately following (July 28-August 12) prior to averaging. This product was used to quantify the mean and standard deviation NDVI within each spatial extent for all sites. Local scale cover estimates were averaged within sites and the standard deviation calculated for each cover type as a measure of he terogeneity. Elevation was recorded from the GPS unit and rounded to the nearest meter. Environmental variables were square-roottransformed for all analyses. A list of environmental variables is presented in Table 2-3. Quantifying spatial variables Principal Coordinates of Neighbor Matrices (PCNM) is a method used to obtain an uncorrelated set of explanatory variables that can be used in regr ession or ordinati on analyses to describe spatial dependence in community data (Borcard and Legendre 2002). PCNM analysis has proven superior to its predece ssor, trend-surface analysis, in that it more fully explains spatial dependence across a continu ous range of scales instead of a single coarse scale (Borcard et al. 2004). PCNM variables were derived following Bor card and Legendre (2002) and Borcard et al. (2004). The procedure to obtain the set of explanatory spatial variables involve s three general steps: (1) calculation of a Euclidian distance matrix comprised of the geographic distances 24

PAGE 25

between sites, (2) modification of the geographi c distance matrix by replacing distances greater than the minimum needed for all sites to remain connected within a network with an arbitrarily large number, and (3) principal coordinates analys is of the modified distance matrix. From the principal coordinates analysis, the axes that corre spond to positive eigenvalues are retained as the set of explanatory PCNM variables. Principal coordinates analysis on the modified distance matrix returned 11 PCNM variables, which explain the spatial dependence of community assemblages across a range of scales. Specific distances corresponding to the spatial scale of each PCNM variable can be calculated if sample sites are a rranged regularly along a linear tran sect or grid. However, since the sample sites in this study were arranged irre gularly throughout the regi on, the spatial scale of each PCNM variable could not be quantified pr ecisely. Qualitatively, though, the first PCNM variables correspond to very broad scale spat ial structure and successively higher numbered variables correspond to increasingly finer resolution spatial structure. The broadest spatial scale roughly corresponds to the spatial extent of the entire study and finest scale roughly corresponds to the average distance between neighboring sites. Statistical analyses Redundancy analysis (RDA) is the canonical ve rsion of principal components analysis and is a method used to examine how much of th e variation in one dataset can be explained by the variation in another. Partial RDA was us ed to determine how a set of environmental variables and a set of spatial variables simulta neously explain the variation observed in ant communities (Borcard et al. 1992, Legendre and Lege ndre 1998). This procedure allows one to determine the effect of pure environmental variation (A), pure spatial va riation (C), and mixed enviro-spatial variation (B) on community structure, as well as the fraction of variation that remains unexplained (D). It follows that (A +B+C+D) equals the tota l amount of variation 25

PAGE 26

observed in the community. Five separate RDA analyses with ant species data as the response variables were required to e xplain each portion of variation: species data constrained by environment data with the spatial data as covariables (A); species data constrained by spatial data with environment data as covariables (C); spec ies data constrained by e nvironment data (A+B); species data constrained by spatial data (B+C); and species data constrained by a combined environment-spatial dataset (A+B +C). Fraction B, mixed enviro -spatial variation, cannot be derived on its own and is obtained by subtracting (A+B) – (A) and/or (B+C) – (C). Fraction (D), unexplained variation, is determined by subtracti ng 1 – (A+B+C). Because the list of potential environmental and spatial variables was long, many of which are highly correlated and/or may explain variation in the species data by chance al one, only significant variab les were used in the final analyses. Significant variables (P < .05) were identified by pe rforming a preliminary RDA on the species data constraine d by the environmental variable s with forward selection (999 Monte Carlo permutations under the full model) . Redundancy analyses were performed using CANOCO v4.5 (ter Braak and m ilauer 2002). Species data was centered and standardized for this and all other RDA analyses. Analyses we re separately run using the landscape metrics measured at each of the five spatial exte nts (100, 250, 500, 750 and 1000 m) . Results observed using the 500 m data explained the greatest amount of variati on in the ant community data, therefore the landscape data at this spatial extent were used in all subsequent analyses. A 500 m radius circle covers an area of 78.5 hectares and is biologically relevant to the arena in which ants interact with the lands cape (Hlldobler and Wilson 1990). Since PCNM variables represent spatial struct ure at specific scales, the six significant PCNM variables were grouped to form three scal e classes: broad (PCNM 1 and 2), medium (4 and 5), and fine (7 and 9). Redundancy analys is, constrained by the PCNM variables for each 26

PAGE 27

spatial class, was performed to investigate the ef fect of spatial structur e in ant communities at multiple scales. Tests of significance on all RDA ax es combined as well as individual axes were performed with 999 Monte Carlo permutations under the full model. The significant RDA axes for each model were then regressed against the p ool of environmental variables to determine how environmental variation contributes to obser ved spatial structure in ant communities. To determine how ant communities are structured by their surrounding environment, RDA was performed on the ant species data c onstrained by the significant environmental variables. Tests of significance on all RDA axes combined and individual axes were performed with 999 Monte Carlo permutations under the full model. Generalized linear models were then constructed for each species in response to the first significant axis of the RDA. Significant species were identified for the generaliz ed linear models by st epwise selection ( P < .05). Results Ant Sampling and Identification Excluding males and queens, pitfall tra pping and Winkler extraction yielded 25,779 individuals from 66 species in 23 genera (Table 2-4). The observed number of species ranged from 17-33 with a mean of 25.2 3.9 (1 SD) species per site. Individual based rarefaction analysis indicated that sites were well sampled as curves for all sites appeared to approach an asymptote. Non-native species occurred in low abunda nces and infrequently, ranging from 0-3 with a mean of 1.3 species per site, which provi des evidence that the sites are reasonably undisturbed. Cardiocondyla emeryi and Cyphomyrmex rimosus were the most frequently observed non-native species, occurring at 19 an d 11 sites respectively. Locally, however, they occur in very low densities. Solenopsis invicta , the red imported fire an t, was present at three sites but also occurred in low densities, not dom inating any species assemblage. Only one site appeared to be dominated by any one particular species. Solenopsis geminata, a Florida native, 27

PAGE 28

was highly abundant at this site (75.9% of individuals captured) and appeared to exclude other species as this site had the lowest number of observed species (16). Very rare species (less than 10 occurrences across all sites) were excluded from subsequent analyses. Most of these species ar e arboreal or legionary and infrequently observed as a result of methodological bias . For example, two species of Crematogaster are arboreal specialists and likely abundant at most sites, yet they were poor ly detected with the ground based sampling strategy. Variable Selection and Enviro-s patial Variance Partitioning Of the eleven derived PCNM variables, vari ables 1, 2, 4, 5, 7, and 9 significantly explain the spatial dependence in ant communities as dete rmined by the forward se lection process (Table 2-3). This represents spatial structure across a range of scales: from ve ry broad (i.e., across the entire study area) to fine (i .e., between neighboring sites). The significant environmental variables identif ied by forward selection and used in these models represent a range of landscape composition and site level habitat data, however, no landscape arrangement metrics were selected (Table 2-3). In the following results, the sum of all canonical eigenvalues is reported as the percent of variation, or R 2 , explained by each model. Since R 2 values will increase as additional variables are added to regression models, explained variation is reported for a ll analyses using adjusted R 2 values so that models with differing numbers of variables can be compared more reliably. The adjusted R 2 was calculated as: R 2 adj = 1 – (( N – 1) / ( N – k – 1)) (1 – R 2 ), where N is the number samples (sites in this case) and k is the number of independent variables included in the model. Results of the set of redundancy analyses indi cate that the variation in ant communities is significantly explained by both environmental and spatial dependence (Table 2-5). The relative 28

PAGE 29

proportions of variation explai ned by pure environmental depe ndence (A: 10.3%), pure spatial dependence (C: 10.9%), and mixed enviro-spatial dependence (B: 11.2%) are represented in Fig. 2-2. Spatial and environmental dependence combin ed explains 32.4% of the variation, leaving 67.6% of the variation in the ant data attributed to unexplained factors. There is substantial overlap between environmental and spatial eff ects. Over 50% of the total unpartitioned environmental effect (A+B) on ant community stru cture is mixed with spatial effects (B+C), indicating that ant communities are structured by environmental variables that are themselves spatially autocorrelated. Spatial Dependence Total spatial dependence (B+C) explained 22.1% of the variation in the ant data. Each PCNM scale class explained ant spatial dependence significantly (Table 25). RDA analysis at each scale indicates that spatial dependence at broad, medium, and fine scales explains 31.1, 18.3, and 17.4 percent of the total spatial dependence re spectively. Linear regression with forward selection ( P < .05) on the first significant axis of the broad scale RDA model indicates that the area of pineland and el evation (in order of importance) surrounding sites are important in spatially structuring ant communities across broad scales. The broad scale change in community structure across the study regi on is evident in Fig. 2-3A. The area of pineland habitat surrounding these sample sites signif icantly decreases with latitude ( R 2 = .404, P < .00001, N = 33) and increases with longitude ( R 2 = .212, P < .007, N = 33), as is seen in the sample scores in Fig. 2-3A. Elevation at sample sites does not vary with latitude but has a weak positive association with longitude ( R 2 = .119, P = .049, N = 33). Spatial structure at fine scales is represen ted in Fig. 2-5C where neighboring sites have similar axis scores, but large and small magnitu de axis scores are distributed throughout the region. Spatial structure at the fi ne scale is influenced (in order of importance) by the mean area 29

PAGE 30

of bare ground within sites a nd the area of high impact urban c over surrounding sites (Table 25). Spatial structure at medium scales is not as defined as structure at broad or fine scales (Fig. 2-3B). The amount of high impact urban cover and mean area of bare ground (in order of importance) influence the spatial structure at this scale. Environmental Dependence Total environmental dependence (A+B) significa ntly explained 21.5% of the variation in the ant data. Results of RDA analysis on the ant data constrained by the significant environmental variables are seen in Fig. 24. The first two axes were significant ( P = .001 and .035 respectively) and expl ain a combined 20.03% of the variation in the ant data. Mean bare ground cover within sites varies most strongly wi th axis 1. The area of pineland, high impact urban, freshwater marsh, and elevation vary with both axes 1 and 2. Cypress swamp varies most strongly with axis 2. Although the area of extractiv e land cover was selected as a variable that significantly describes variation in ant communities, it does not seem to be a factor in the first 2 RDA axes. Though community composition and the relative abundances of species are influenced by landscape and local variables, species richness app ears to be independent of the local and landscape level environmental variable s, and determined by unm easured factors (Fig.24). The amount of pineland surrounding study sites is strongly correlated with the first and second axes and accounts for much of the variatio n in the ant community. The area of this land cover type can be viewed as a surrogate for sandhill fragmentation. Within the study area pineland is significantly and ne gatively correlated with th e total amount of sandhill (R 2 = .479, P < .0001, N = 33), focal patch area ( R 2 = .308, P < .001, N = 33), mean sandhill patch area ( R 2 = .301, P < .001, N = 33), and mean patch perimeter ( R 2 = .344, P < .001, N = 33) in each 500 m 30

PAGE 31

radius circle. Moreover, the area of pineland habitat is significantly and positively correlated with the perimeter to area ratio ( R 2 = .200, P = .009, N = 33) and the number of sandhill patches ( R 2 = .167, P = .018, N = 33) in each landscape. So with greater amounts of pineland cover, the total amount of sandhill decrease s, and sandhill patches become smaller, more numerous, and more irregular in shape. One could therefore as sociate this land cover type with sandhill habitat fragmentation and an increased matrix area of rela tively densely structured habitat. Indeed, this habitat has been aptly described as the matrix that connects many habitats, such as sandhill, throughout Florida (Edmisten 1963). In order to characterize ant communities by th e species that comprise these assemblages, species response curves were generated from the generalized linear model for axis 1 of the RDA (Fig. 2-5). Results reveal that sites with little pine land in the surrounding landscape (therefore high proportions of intact sandhill) and greater amounts of bare ground loca lly, support a greater number of open habitat specialists. These open habitat specialists are replaced by habitat generalists at the other end of the community ax is, where sites are juxtaposed to greater amounts of pineland habitat. Here, ope n habitat specialists are define d as species that require open structured habitats with a sparse understory a nd well drained sandy soil. Habitat generalists are defined as species that can exist in either open habitats or more densely vegetated habitats with thicker leaf litter layers and more mesic soil conditions. Discussion Variation in ant communities in Florida sa ndhills is strongly dependent on environmental and spatial variation at multiple spatial scales. Redundancy analysis using PCNM variables indicates that ground-dwelling ant communities in Florida sandhills exhibit significant spatial dependence across broad (regional), medium, and fine (between ne ighboring sites) scales (Table 2-5, Fig. 2-2). Broad scale spatial dependence ex plains the greatest propo rtion of the overall 31

PAGE 32

spatial component, with medium and fine scal e dependence explaining a smaller but similar proportion. Different environmenta l variables are correlated with spatial dependence at each of these scales suggesting that di fferent processes are involved in community organization at different scales. Pineland and bare ground are the two strongest environmental factors structuring ant communities (Fig. 2-4). At the fine scale, the amount of bare ground is important in defining spatial structure, which indicates that local habitat management, such as prescribed fire, may also contribute to ant community organization. The importance of pineland in explaining spatial-dependence at broad scales suggests that sandhill fragmentation and the amount of nearby matrix habitat is a strong driver of ant comm unity structure. Spatially dependent spillover from neighboring habitats can have strong influences on local community composition (Holt 1993, Polis et al. 2004, Rand et al. 2006). The significa nce of pure spatial effects on ant community structure provides supp ort for the mass effects perspective of the metacommunity concept. Cook and colleagues ( 2002) recognized the importance of infiltration by matrix species to observed patterns of community structure. A pure environmental effect is also seen in the environmental R DA model. The main axis in Fig. 2-4 is organized along an environm ental gradient primarily defined by landscapes with high proportions of pineland and fragmented sandhill at one end, and landscapes with high proportions of intact and open stru ctured sandhill at the other. Results of generalized linear models reveal that the occurrence of habitat generalists and open habitat specialists can also be arranged along this axis of community structure (Fig 2-5), with the strongest factors being the amount of bare ground at fine scal es and the amount of pineland at broad spatial scales. These two sets of species are importa nt community players accounting for over 54% of the occurrences and raw abundance in the study. This result s uggests that some combination of the species32

PAGE 33

sorting and mass effects perspective on met acommunities may be appropriate, since both environmental and spatial dynamics are at work in structuring ant communities. Regional coexistence of the habitat generalis t and open habitat specialist ant species described in this study is likely influenced by trade-offs in performance among particular habitats (Kneitel and Chase 2004). Sandhill habita t is less productive with greater variation and extremes in microclimate conditions than much of the intervening matrix ha bitat (e.g., pineland). Increased tolerance in open habitat specialists to these types of conditions may be traded for habitat generalization, re sulting in spatially-segregated regional coexistence. Palmer (2003) demonstrated a similar scenario, though at a sm aller spatial scale, in a competitive guild of acacia ants that coexist by partitioning habitat be tween areas of differing resource productivity. In his study, competitively dominant species supp lanted subordinate species in high productivity zones, while subordinate species supplanted co mpetitive dominants in low productivity zones. However, since local habitat qua lity was standardized in the present study, it is unclear why habitat generalists, normally the inferior competit or in the habitat prefer red by specialists, appear to be outcompeting open habitat specialists in landscapes with high proportions of matrix habitat. The apparent replacement of open habitat spec ialists with habitat generalists may occur through a number of possible mechanisms such as (1) source-sink dynamics, (2) cross-system flow mediated competitive interactions, and (3) cr oss-system flow mediated predation. Each of these possibilities is considered in turn. Source-sink dynamics (1) may permit species, such as the habitat generalists observed here, to avoid local extinction through repeated invasion and replenishment from neighboring habitats (B rown and Kodric-Brown 1977). Holt (2004) developed a simple model that explains how c ontinuous immigration by an inferior competitor allows it to persist in habitats where it otherw ise would not. When the rate of immigration is 33

PAGE 34

sufficiently high, the inferior competitor can ev en exclude the superior competitor. The landscape-dependent colonization ra te of generalist ant species was demonstrated by Schoereder and others (2004). These authors describe how smaller tropical fo rest remnants, which tended to be more isolated, were more often invaded by gene ralist ant species from th e surrounding matrix. (2) Alternatively, it is possible that habita t generalists may be outcompeting open habitat specialists as a result of biotic and/or abioti c flows from surrounding habitats which alter the local environment in a way that allows them to better exploit resources or tolerate microclimate conditions. For example, Pheidole dentata is a habitat generalist that recruits a dominant number of workers to secure and gather food resources, while excluding most other species. King (2004) reported that P. dentata dominated baited traps across a ra nge of Florida ecosystem types including pine flatwoods, hardwood hammocks, and sandhills. P. dentata was particularly dominant in the sites with more densely structured habitat but less so in open structured sites. Cross-system flows of resources origin ating from matrix habitats may allow P. dentata to maintain its dominance over habitat specialists, even in high quality sandhill habitat where open habitat specialists like P. metallescens tend to exert increasing dominance at baits. (3) Spillover effects can influence trophic dyna mics as well (Polis et al. 1997, Rand et al. 2006). Habitat generalists may be adapted to av oid predators that have also invaded from the matrix, leaving open habitat specialists to bear th e brunt of predation by the invading predators. Alternatively, open habitat specia lists may be at greater risk of predation via apparent competition. Such a situation might result if in situ predator populations are artificially high because they are spatially s ubsidized by the source-sink movement of colonizers from the surrounding matrix (Holt 2004). Additional research is needed to confirm whether any of these potential mechanisms is at work in this system. 34

PAGE 35

Ant community change across the study region a ppears to follow many of the predictions of metacommunity theory. The significant effects of pure spa tial dependence suggest that ant metacommunities are organized ac cording to the mass-effect pers pective. This supposition is supported by the apparent movement of habitat ge neralists into sandhill patches in landscapes with high proportions of matrix habitat. Even though there are strong en vironmental effects (an indication of species-sorting mechanisms) much of the variation explained by environmental factors is confused with spatial factors. Nevertheless, this mixed spatial dependence does not completely discount the species-sorting perspectiv e. The significance of the pure environmental component, as well as the significance of local variables (i.e., bare ground and elevation) indicates that species sorting does influence community organizat ion. Cottenie and De Meester (2005) observed similar spatial and environmen tal dependence of zooplankton density in a metacommunity of interconnected ponds. It s eems likely that a combination of these two metacommunity perspectives w ould be best used to descri be the dynamics observed here. Conservation and Management Implications Alterations in species com position resulting from land cover change and the ensuing change in system dynamics can impact the func tion of ecosystems. Ness (2004) and Ness et al. (2004) describe how community composition can vary with land cover change and exotic invasion, resulting in a diminished capacity of an ts to disperse seeds. Many sandhill plant species benefit from ant-assisted seed transp ort, so the loss of granivorous open habitat specialists, such as the Florida harvester ant ( Pogonomyrmex badius ), may decouple ant-plant relationships and invoke feedback s that alter plant distributions (Stamp and Lucas 1990, Harmon and Stamp 1992). Though community composition varied greatly, species richness was independent of the measured environmental variables and similar acr oss most sites. The high levels of species 35

PAGE 36

turnover among sites would suggest that spatial and environmen tal heterogeneity at multiple scales is important for maintaining a diverse regional specie s pool. Therefore, continued anthropogenic pressure on sandhills in Florida resu lting in habitat loss or reduced habitat quality is cause for concern. Alteration of the landsca pe favoring habitat generalists may reduce the viability of open habitat specialists across the range of remaining sandhill, thereby reducing the potential for spatially segregated regional coex istence. Given the importance of broad scale environmental variation to sandhill ant communities, multiscale adaptive management strategies are necessary for assuring the preservation of regional biodiversity and ec ological integrity since uncoordinated local management would prove ineffective (Cumming and Spiesman 2006). Conclusions Both environmental and spatial variation at multiple spatial scales is important for community organization in sandhill ants. Sinc e sampling was conducted in one habitat type, standardized for quality, community change betw een sites largely reflects variation in the surrounding landscape. The results presented here emphasize the importance of recognizing openness in the dynamics of local sandhill ant comm unities, and suggest that cross-system flows from nearby areas have significant effects on co mmunity structure. Cross-system flows of species or resources between neighboring habitats can influen ce the outcome of competitive interactions or alter predator-pre y relationships. Results suggest that such flows from habitats neighboring the study sites may alter community structure through source-sink dynamics, crosssystem flow mediated competitive interactions, or cross-system flow mediated predation. The data do not conclusively support any of the hypot hesized mechanisms but it appears that mass effects may be highly important in determin ing how the composition of local communities changes across the study region. Results of this study also suggest that metacommunity dynamics influence local communities of ants, and the concept would therefore be a good 36

PAGE 37

framework for future studies. Such studies should combine observation and experimental manipulation in multiple localities within multip le landscapes to investigate the mechanisms behind the community patterns observed here. 37

PAGE 38

Table 2-1. Site locations Site Location County Latit ude Longitude Elevation 1 Ordway-Swisher Biological Station Putnam 29.699346 -82.024988 45 2 Ordway-Swisher Biological Station Putnam 29.669110 -82.019424 33 3 Ordway-Swisher Biological Station Putnam 29.672562 -82.028206 38 4 Ordway-Swisher Biological Station Putnam 29.684560 -81.998185 63 5 Ordway-Swisher Biological Station Putnam 29.707818 -81.956985 31 6 Ordway-Swisher Biological Station Putnam 29.727738 -81.976403 42 7 Ordway-Swisher Biological Station Putnam 29.702135 -81.940305 37 8 Ordway-Swisher Biological Station Putnam 29.678287 -81.959956 56 9 Ross Prairie State Forest Marion 29.034736 -82.298020 24 10 Ross Prairie State Forest Marion 29.010435 -82.271654 32 11 San Felasco Hammock Preserve State Park Alachua 29.715400 -82.457834 72 12 Jennings State Forest Clay 30.136934 -81.952781 40 13 Jennings State Forest Clay 30.136875 -81.885852 45 14 Jennings State Forest Clay 30.095827 -81.916115 34 15 Withlacoochee State Forest Citrus 28.815033 -82.455900 13 16 Mike Roess Gold Head Branch State Park Clay 29.8 45977 -81.952740 67 17 Mike Roess Gold Head Branch State Park Clay 29.8 28854 -81.956047 58 18 Mike Roess Gold Head Branch State Park Clay 29.8 17218 -81.942581 58 19 near Goethe State Forest Levy 29.283639 -82.587915 33 20 Goethe State Forest Levy 29.350489 -82.604873 37 21 Keuka Sand Mine Putnam 29.595957 -81.908054 32 22 near Withlacoochee State Forest Citrus 28.979018 -82.549149 18 23 Withlacoochee State Forest Hernando 28.536876 -82.269255 32 24 Withlacoochee State Forest Hernando 28.602227 -82.285193 23 25 Withlacoochee State Forest Hernando 28.622061 -82.286753 26 26 Withlacoochee State Forest Hernando 28.585455 -82.285958 43 27 Withlacoochee State Forest Citrus 28.713448 -82.482858 27 28 Withlacoochee State Forest Citrus 28.710467 -82.412454 37 29 Withlacoochee State Forest Citrus 28.738915 -82.435827 45 30 Withlacoochee State Forest Citrus 28.746142 -82.407648 34 31 Ralph E. Simmons State Forest Nassau 30.787781 -81.958989 36 32 Ralph E. Simmons State Forest Nassau 30.799472 -81.950544 28 33 Ralph E. Simmons State Forest Nassau 30.791503 -81.945466 28 Spatial coordinates and elevation recorded at each site center. Elevation is rounded to the nearest meter. See map in Fig. 2-1. 38

PAGE 39

Table 2-2. Site selection variab les. Each variable was measured from the 2003 FWC land cover map (Stys et al. 2004) within a radius of 100, 300, 500, and 1000 m surrounding each of the 2000 random points in the selection pool. This data was then used to select sites that span the range of landscape heterogeneity within the study region. Site Selection Variables Focal patch area Focal patch perimeter/area ratio Number sandhill patches Mean perimeter/area ratio of all sandhill patches Mean perimeter of all sandhill patches Number land cover classes Area of each land cover class 39

PAGE 40

Table 2-3. Pool of environmental variables. Landscape composition and arrangement variables were derived from the FWC land cover map (Stys et al. 2004). NDVI data is from Carroll et al. (2005). Variables are dis tinguished between composition (identity and amount of landscape variable), arrangement (variables re lated to patch shape and spatial location within the landscape) and lo cal (mean and standard deviation percent cover within a site). Significance values are reported for variables selected with forward regression and used in the final analysis. Variable code Description FWC land cover class Variable type Sig. Sndhl Area sandhill 5 Composition DryPr Area dry prairie 6 Composition HdPin Area mixed hardwood-pine 7 Composition Hamk Area hardwood hammocks 8 Composition Pine Area pineland: pine plantation & 9 Composition 0.001 Mar Area freshwater marsh 12 Composition 0.033 D_Smp Area disturbed swamp 15 Composition B_Smp Area bay swamp 16 Composition C_Swp Area cypress swamp 17 Composition WetFor Area wetland forest 19 Composition 0.014* HdSmp Area hardwood swamp 20 Composition Water Area open water 27 Composition 0.028* D_For Area disturbed forest 30, 28 Composition Past Area pasture 31, 32 Composition Agric Area agriculture 34, 35, 36 Composition H_Urb Area high impact urban 41 Composition 0.003 L_Urb Area low impact urban 42 Composition Extr Area extractive 43 Composition 0.023 NumC Number land cover classes Composition NDVI Mean NDVI Composition SD_NDVI Std. Dev. NDVI Composition MNN Mean nearest neighbor (all classes) Arrangement MSI Mean shape index (all classes) Arrangement MNN_SH Mean nearest neighbor (sandhill) Arrangement MSI_SH Mean shape index (sandhill) Arrangement PA Focal patch area Arrangement PAR Focal patch perimeter/area ratio Arrangement NSHP Number sandhill patches Arrangement 0.031 APA Mean sandhill patch area Arrangement APP Mean sandhill patch perimeter Arrangement 40

PAGE 41

Table 2-3 Continued Variable code Description FWC land cover class Variable type Sig. Elev Elevation Local 0.031 A_Gras Mean grass cover Local SD_Gras Std. Dev. grass cover Local A_Lit Mean litter cover Local SD_Lit Std. Dev. litter cover Local A_Herb Mean herbaceous cover Local SD_Herb Std. Dev. herbaceous cover Local A_Shrb Mean shrub cover Local SD_Shrb Std. Dev. shrub cover Local A_Tree Mean tree cover Local SD_Tree Std. Dev. tree cover Local A_Bare Mean bare ground Local 0.001 SD_Bare Std. Dev. bare ground Local A_DW Mean dead wood cover Local SD_DW Std. Dev. dead wood cover Local PCNM1 PCNM Axis 1 (Broadest spatial scale) Spatial 0.001 PCNM2 PCNM Axis 2 Spatial 0.001 PCNM3 PCNM Axis 3 Spatial PCNM4 PCNM Axis 4 Spatial 0.001 PCNM5 PCNM Axis 5 Spatial 0.041 PCNM6 PCNM Axis 6 Spatial PCNM7 PCNM Axis 7 Spatial 0.001 PCNM8 PCNM Axis 8 Spatial PCNM9 PCNM Axis 9 Spatial 0.030 PCNM10 PCNM Axis 10 Spatial PCNM11 PCNM Axis 11 (Finest spatial scale) Spatial *Variables excluded in the final analyses because of high covariance with other forward selected descriptors. 41

PAGE 42

Table 2-4. Species list. “Number individuals” are the total number of individuals captured in both trapping sessions. “Number occurren ces” are the number of times a species occurred in a trap, regardle ss of the abundance per trap. Species Species code Number individuals Number occurrences Aphaenogaster ashmeadi Emery APHASH 55 41 Aphaenogaster flemingi M.R. Smith* APHFLE 6 4 Aphaenogaster floridana M.R. Smith APHFLO 37 19 Aphaenogaster miamiana Wheeler* APHSP1 1 1 Aphaenogaster treatae Forel APHTRE 60 35 Aphaenogaster umphreyi Deyrup & Davis* APHUMP 2 2 Brachymyrmex depilis Emery BRADEP 37 16 Camponotus castaneus Latreille* CAMCAS 6 4 Camponotus floridanus Buckley CAMFLO 185 93 Camponotus socius Roger CAMSOC 182 92 Cardiocondyla emeryi Forel CAREME 100 45 Cardiocondyla wroughtonii Forel* CARWRO 2 1 Crematogaster ashmeadi Mayr* CREASH 9 8 Crematogaster species A (undescribed)* CRESPA 11 10 Cyphomyrmex rimosus Spinola CYPRIM 18 17 Discothyrea testacea Roger* DISTES 4 4 Dorymyrmex bossutus Trager DORBOS 165 78 Dorymyrmex bureni Trager DORBUR 401 75 Dorymyrmex elegans Trager* DORELE 17 6 Dorymyrmex flavopectus M.R. Smith* DORFLA 21 5 Dorymyrmex grandulus Forel* DORGRA 11 6 Forelius pruinosus Roger FORPRU 3969 283 Forelius species 1 (undescribed) FORSP1 424 21 Formica archboldi M.R. Smith FORARC 129 86 Formica pallidefulva Latreille FORPAL 196 73 Hypoponera opacior Forel HYPOPR 165 65 Monomorium viride Brown MONVIR 364 70 Neivamyrmex carolinensis Emery* NEICAR 36 4 Neivamyrmex opacithorax Emery* NEIOPA 18 5 Neivamyrmex texanus Watkins* NEITEX 92 5 Odontomachus brunneus Patton ODOBRU 692 304 Odontomachus relictus Deyrup & Cover ODOREL 123 43 Paratrechina arenivaga Wheeler PARARE 925 155 Paratrechina faisonensis Forel* PARFAI 1 1 Paratrechina parvula Mayr PARPAR 185 60 Paratrechina phantasma Trager PARPHA 244 40 Paratrechina wojciki Trager PARWOJ 327 129 42

PAGE 43

Table 2-4 Continued Species Species code Number individuals Number occurrences Pheidole adrianoi Naves PHEADR 133 40 Pheidole dentata Mayr PHEDEN 1882 286 Pheidole dentigula Smith* PHEDEG 6 3 Pheidole floridana Emery PHEFLO 3124 318 Pheidole metallescens Emery PHEMET 3975 312 Pheidole morrisi Forel PHEMOR 1672 95 Pheidole tysoni Forel* PHETYS 4 1 Pogonomyrmex badius Latreille POGBAD 816 193 Pseudomyrmex pallidus F. Smith* PSEPAL 3 3 Pyramica angulata M.R. Smith* PYRANG 2 2 Pyramica clypeata Roger* PYRCLY 2 2 Pyramica deyrupi Bolton* PYRDEY 1 1 Pyramica dietrichi Smith* PYRDIE 1 1 Pyramica eggersi Emery* PYREGG 6 5 Pyramica membranifera Emery* PYRMEM 8 7 Pyramica ornata Mayr* PYRORN 5 3 Solenopsis carolinensis Forel SOLCAR 1948 395 Solenopsis geminata Fabricius SOLGEM 1296 23 Solenopsis invicta Buren SOLINV 332 24 Solenopsis nickersoni Thompson SOLNIC 585 146 Solenopsis picta Emery* SOLPIC 8 7 Solenopsis tennesseensis M.R. Smith SOLTEN 73 26 Solenopsis tonsa Thompson SOLTON 53 31 Strumigenys louisianae Roger STRLOU 14 11 Strumigenys silvestrii Emery* STRSIL 3 2 Temnothorax pergandei Emery TEMPER 388 151 Temnothorax texanus Wheeler TEMTEX 130 76 Trachymrymex septentrionalis McCook TRASEP 88 59 *Rare and/or poorly sampled species excluded from analysis. 43

PAGE 44

Table 2-5. Percent of the total spatial depende nce explained by each spatial scale and the correlated environmental variables. P -values (999 Monte Carlo simulations under the full model) are listed for the first and only si gnificant RDA axis for each spatial scale. Regression with forward selection of envi ronmental variables was performed on the first significant canonical axis to determine how spatial dependence is correlated with environmental variation. At each spatial sc ale, environmental variables are listed in the order of importance to the model. Spatial Scale Broad (PCNM 1&2) Medium (P CNM 4&5) Fine (PCNM 7&9) P -value 0.001 0.008 0.026 Percent of spatial variation 31.1 18.3 17.4 Sig. Environ. Variables PINE, ELEV H_Urb, A_Bare A_Bare, H_Urb P -value < 0.001 0.003 < 0.001 R 2 Adj 0.486 0.285 0.396 44

PAGE 45

Figure 2-1. Thirty-three study sites (black circles) span most of the north-south extent of sandhill habitat (dark shaded area) in peni nsular Florida, USA. See Table 2-1 for descriptions of site locations. 45

PAGE 46

Figure 2-2. Venn diagram (drawn to scale) representing the proportions of variation (adjusted R 2 ) in the ant community explained by pure environmental factors (A = .103, P < .001), pure spatial factors (C = .109, P < .01), mixed environment and spatial (B = .112), and unexplained factors (D = .675). Unpartitioned environmental dependence can be calculated as A+B (= .215, P < .001), unpartitioned spatial dependence B+C (= .221, P < .001), and the total variation expl ained by environmental and spatial variables is A+B+C (= .324, P < .001). 46

PAGE 47

Figure 2-3. Spatial structure at br oad (A), medium (B), and fine (C) scales. Sample scores for redundancy analyses (RDA) of ant data constrained by PCNM variables are plotted in geographic space (UTM zone 17) for the br oad, medium and fine scale models. X and Y axes are the easting and northing respectively. The size of the circle is proportional to the RDA sample score. In the broad scale model (A), spatial dependence changes across a gradient that spans the entire study region. This is partly driven by environmental dependence ( on pineland and elevation) that is itself spatially structured at this scale. The fine scale m odel (C) shows how neighboring sites have similar axis scores (therefo re similar spatial dependence) and the magnitude of axis scores change across a mu ch smaller spatial extent. This spatial dependence at this scale is also relate d to dependence on spatially structured environmental variation (primarily the mean amount of bare ground locally and secondarily, the amount of high impact urba n cover in the landscape). The medium scale model (B) is not as defined as the broad and fine scale models. Spatial dependence at this scale is related to spatial structure in the amount of urban land cover, and local bare ground c over (in order of importance). 47

PAGE 48

A xis 1 ( 11.20 % ) Axis 2 ( 8.83 % ) Figure 2-4. Redundancy analysis biplot for the first and second significant axes ( P = .001 and .035 respectively). The length of the arro ws along each axis increases with the strength of correlation with the respective axis scores. The angle between arrows indicates the degree of correlation with ot her environmental variables. Elevation (Elev), the mean amount of bare ground (A_B are) locally, and the amount of pineland (Pine) and high impact urban (H_Urb) in th e landscape are important in structuring communities along the first axis. The amount of pineland, cypress swamp (C_Swp), and freshwater marsh (Mar) are important in structuring ant communities along the second axis. Positions of circles represent s ite scores relative to the first two axes and the radius of the circle increases relative to the species richness observed at the site. Communities vary along the two axes with environmental variables but species richness does not appear to be correlated with measured environmental factors since circles of different sized radii are scattered throughout the plot. 48

PAGE 49

Figure 2-5. Species response to redundancy analysis (RDA) axis 1. Species data are constrained by significant environmental variables unde r RDA. “Species Response” is the number of occurrences per si te (Hellinger-transformed). Species with significant relationships with RDA axis 1 were identif ied using generalized linear models with stepwise selection ( < .05). Solid and dotted lines se parate the trajectory of species that increase and decrease with the axis scores. An increase in RDA axis 1 scores reflect an increase in the amount of pineland and othe r matrix habitat in the landscape, and a decrease in local bare ground cover and the amount of sandhill habitat in the landscape. The generalized linear model separates open habitat specialists (species that re quire an open structured ha bitat with well drained sandy soil) from habitat generalists (species that exist in both open and more densely structured habitat and a ra nge of soil conditions). 49

PAGE 50

CHAPTER 3 ANT BODY SIZE DISTRIBUTIONS: INVEST IGATING SOME PREDICTIONS OF THE TEXTURAL DISCONTINUITY HYPOTHESIS Introduction The size of an organism has long been rec ognized as a key ecol ogical attribute (Elton 1927, Hutchinson 1959, Brooks and Dodson 1965). Many of the factors that limit the distribution of species are funda mentally related to body size. Metabolic rate, resource use, locomotive apparatus, and other such size-dependen t factors serve as filte rs that determine how an environment influences different sized species. It follows that species of similar size are more likely to perceive and interact w ith their surroundings in similar wa ys, and their relationship with the surrounding environment is likely to scal e with their body size (Ritchie and Olff 1999, Gehring and Swihart 2003, Holland et al. 2005). Body size distributions ar e thus shaped by the structure of the surrounding environment at multip le scales. A better knowledge of the ways in which body size distributions relate to multisca le environmental variation will benefit our understanding of community or ganization in a time of increasing environmental change. Numerous competing theories ha ve been put forth in an effo rt to explain patterns in the distribution of body size, and though many of these theories share commonalities a synthesis has not been achieved (reviewed in Allen et al. 2006 ). Although it was origin ally assumed that a unimodal smooth distribution of body size would result from competition and maximization of niche partitioning, there is increasing eviden ce that body size should re sult in discontinuous distributions through ecological and evolutiona ry processes (Oksanen et al. 1979, Holling 1992, Moksnes et al. 1998, Cumming and Havlicek 20 02, Scheffer and van Nes 2006). Scheffer and van Nes (2006) recently showed how competing species can converge, th rough coevolution, into body size aggregations rather than disperse evenly along a size gr adient. The authors concluded 50

PAGE 51

that species that are sufficiently similar can coexist by avoiding competition with different body size aggregations. The textural discontinuity hypothesis (TDH) was proposed by Holling (1992) in an attempt to explain discontinuous patterns in the body size distributions he observed. The hypothesis is based on the assertion that ecosystems are structured by a relatively few important processes that operate within defined spatial a nd temporal scales. Since a species’ interaction with its environment is determined by its size, a discontinuous body size distribution arises as species of similar size are entrained to environmen tal variation at their scale of perception. The appearance of discontinuiti es, or gaps, in the size distribution results from a change of scale in environmental interaction. For example, an upper limit in a size aggregation may result when a locomotory threshold is reached in response to th e three-dimensional struct ure of the habitat. This size gap would reflect a zone of transi tion where the locomotory mechanism changes in some way to allow for operation at a larger scal e where it is sufficiently efficient (Holling 1992). Such gaps may result from a number of other mechanisms that are subject to environmentally driven changes in the sc ale of operation (e.g., metabolic requirements). A number of predictions aris e from the TDH (Allen et al. 2006). For example, the TDH predicts a change in body size distribution with changes in ecological stru cture (e.g., habitat type or climate), but no change in size distribution with changes in the composition and abundance of species. Predictions regarding environmental pe rturbation follow similar lines. For example, when a perturbation results in changes in ecosyst em processes (e.g., nutrien t availability), but no change in ecosystem structure (e.g., number of patches), the TDH predicts no change in size distribution. Alternatively if a perturbation results in a change in ecosystem structure but no change in ecosystem processes, the TDH doe s predict a change in size distribution. 51

PAGE 52

There have been relatively few empirical tests of the TDH and no comprehensive multiscale studies. This study uses data from 33 ant communities in sandhill habitat from northern Florida (see Chapter 2) to explore how body size distributions are influenced by local and landscape scale variation in the surrounding envi ronment. Ants are well suited for this type of analysis because they exhibit a wide range in body size and competitiv e strategies; they are diverse and easy to sample; they have a long evolutionary history and are thus well adapted to the structure of their environment; and, they respond to local and landscape environmental variation at smaller scales than larger-bodied organisms, thereby ea sing the logistics of sampling. In order to examine the predictions of the T DH, three main questions are posed: (1) Is the distribution of ground-dwelling ants in Florida sandhills discontinuous? (2) How does the distribution change with envir onmental variation and species co mposition? (3) Do size groups differ in their response to local and landscape variation? Unde r the TDH, specific predictions can be made regarding ants in this study: (1) ant body size should e xhibit a discontinuous distribution, (2) the distribution should change with landscape st ructure but not with species composition, and (3) size groups should differ in their response to local and landscape environmental variation. Methods Ground-dwelling ant communities were samp led twice at 33 locations throughout northern Florida between May a nd October 2005. A combination of methods was used: pitfall trapping and litter extraction via Wi nkler sacks. Sixteen pitfall tr aps were arranged on a four by four grid with traps spaced 10 m apart. Litter was collected for Winkler extraction at the four corners and center of each trapping grid. Sampling was focused in sandhill habitat and selected to span a gradient of variation in the surr ounding landscape within a 500 m radius, while being standardized for high quality sa ndhill habitat locally. The local habitat was characterized by 52

PAGE 53

estimating the percent cover of vegetation comp onents and bare ground (Tab le 3-1). Landscape variation was measured in a GIS environment and based on the Normalized Difference Vegetation Index (NDVI) and land cover data within the 500 m radius (Table 3-1). The gradient in landscape variation spans the full range of variation in the amount and arrangement of sandhill habitat, and most of the range of variation in the composition of surrounding habitats at this scale (See Appendix). Principal coordina tes of neighbor matrices (PCNM) analysis was used to obtain an uncorrelated set of spatial variables to explain spatial dependence at multiple spatial scales among size classes (Borcard and Legendre 2002, Bo rcard et al. 2004). Ch apter 2 contains a detailed description of the methods used in sampling, site selection, and measurement of explanatory variables. In a thorough analysis of the upland ant comm unities of Florida, King (2004) measured the mass of worker ant species. These data were applied to the ants captured in this study to examine patterns in the distribution of body mass . Two methods were used to determine the presence, number, and location of discontinu ities in the distribution of ant body mass by grouping species into homogeneous clusters, fo llowing the recommendations of C. Stow, C. Allen, and A. Garmestani (unpublished manuscript). (1) Hierarchical cluste r analysis (HCA) is a method of organizing data into homogeneous gro ups and was performed using Ward’s method in SAS (SAS Institute Inc. 1999). The number of clusters was determined by examining the output for agreement in the positi on of peaks in the pseudo F and cubic clustering criterion, and trough in the pseudo T 2 . (2) Bayesian Classification and Re gression Tree analysis (BCART) is a Bayesian adaptation of Classification and Regr ession Tree (CART) analysis (Chipman et al. 1998). CART is similar to HCA in that it locates homogeneous groups in the data, but a limitation of these two methods is that the resu ltant groupings may not be globally optimal (C. 53

PAGE 54

Stow et al. unpublished manuscript). BCART account s for this problem by using a set of prior distributions to stochastically search for optimal locations to draw divisions between groups. This method may therefore be more robust to the effects of undetected species when determining the location of discontinuities. BCART analysis was performed using a program available online by Chipman and colleagues (2006). Under the TDH, body size distributions should ch ange as landscape structure changes. To address this prediction, changes in si ze distributions were detected by looking for representation by one or mo re species of each size group at each site. If one or more size groups dropped out at a site, the distribut ion was said to be changed. Ant species belonging to the resulting size classes were then examined by partial redundancy analysis (RDA), in a similar manner as Chapter 2, to investigate the ways in which different classes are influenced by their surrounding environment. Th is set of analyses partitions the variation explained in each size class among three predictor matrices: landscape, local, and spatial. The analyses yield eight fractions of th e total variation in each size class that allow the investigator to separate pure and shared fractions among the thr ee explanatory matrices (Fig. 31). The set of analyses proceeded in two genera l steps. (1) Forward selection of significant predictor variables: for e ach of the three variable types (landscape, local, and spatial) significant predictor variables were identified separate ly for each size class by forward selection ( P < .05). Forward selection was performed with 999 Monte Carlo permutations unde r the full model using CANOCO v4.5 (ter Braak and milauer 2002). (2 ) Partial RDA was then performed on each size class, constrained by the sign ificant variables from the three se ts of predictors. For each size class, RDA was performed on the ant data (site by species matrix comprised of the number of occurrences of each species) cons trained by the three predictor ma trices. The varpart function, a 54

PAGE 55

component of the vegan package v1.8-2 (Oksanen et al. 2006), in R v2.2.1 (R Development Core Team 2004) was used to perform this set of analys es. Landscape and local data were square root transformed prior to analysis. A Friedman two-way analysis of variance by ranks was used to test (1) whether size groups respond to different sets of environmenta l variation and (2) whether the fractions of explained variation (landscape, local, and spatial) vary between size groups. The nonparametric Friedman test checks the null hypothesis that re lated subjects come from the same population (Siegel and Castellan 1988). In the case of (1), the subjec ts are the set of significant environmental variables as determined by th e forward selection process and size classes represent the condition that is be ing tested. In the case of (2), the subjects represent the size classes and the conditions are the pure fractions of variation explained by landscape, local, and spatial factors for each size class. Results Results of the BCART analysis indicate that the distribution of body mass in grounddwelling ants in Florida sandhill forms six homogeneous groupings (Fig. 3-2). Results of the HCA were similar. The pseudo F , cubic clustering criterion, and pseudo T 2 statistics showed agreement at seven groupings, which was interpreted as the most supported solution in the HCA. A comparison of these results illustrating the groupings and position of discontinuities is presented in Fig. 3-2. The BCART solution of six groupings was chosen to be used in all subsequent analyses since it is the more c onservative of the two sets of results. Lack of size group representati on was rare across the 33 sites. Only six sites were not represented by species from one or more size groups. Group 1, the smallest sized group, was absent from all six of these sites. Group 5 was also absent from two of these six sites. 55

PAGE 56

The overall RDA models, combining local, la ndscape, and spatial variables, explain significant proportions of variation in each size group (Fig. 3-3). Of the total variation in size group 1 attributed to the three variable types, landscape variation e xplains the greatest proportion, with no significan t local or spatial effect s. The fractions of variation in size group 2, by contrast, are distributed more evenly among the three variable types. Of the three variable types, most of the variation explained in size group 3 is influenced by landscape factors but local and spatial variation contributes si gnificant fractions. Spatial fact ors have the greatest effect on group 4. Landscape and spatial variation explain the majority of variation in size group 5 but it is largely shared betw een the two. Landscape, local, and sp atial variation all have similar influences on size group 6. The proportion of variation explaine d by each significant environmental variable for each size class is show n in Table 3-2. The amount of pineland in the landscape was the most important environmental variable across size groups. It explained a significant amount of variation in four of the six size groups. All other environm ental variables explained a statistically significan t portion of variation in only one or two other size groups, with the majority being unique to a single group. PCNM variables explain spatial dependence from broad scales (across the entire study region) to fine scales (bet ween neighboring sites). The PCNM spatial variables overlapped between size groups to a greater degr ee than environmental variables, indicating that size gr oups are generally structured by similar spatial processes (size group 1, having little variation explained by spatial variables, being a notable exception). Much of the variation explained by the PCNM va riables is mixed with local or landscape environmental variation, indicating that spatial variation is depe ndent on environmental variation that is itself spatially autocorrelated. Most of the spatial dependence occurs at broad or medium scales, with less spatial variation ex plained by finer scale structure. 56

PAGE 57

Results of the first Friedman test indicate that body size classes are not responding to different sets of environmental variation (N = 24, 2 = 2.2, df = 5, P = 0.821). Though the overall RDA models explain signi ficant proportions of variati on for each size group, results of the second Friedman test indicate that the respons e to landscape, local, and spatial variation does not differ significantly be tween size groups (N = 6, 2 = 3.6, df = 2, P = 0.165). Discussion Both hierarchical clustering and BCART analysis indicate that body size distributions form discontinuous groupings of similar sized species. Representatives of each of these groupings are present across the majority of study sites indicating that the body size distribution does not change with landscape structure. Havlicek and Carpente r (2001) found a similar conservation of size distribution in a series of northern US temperate lakes. Partial RDA shows that size groups have significant responses to lo cal, landscape, and spat ial variation, indicating that species composition changes with environmental and spatial variation. This result shows that body size distributions remain similar even as community composition and variation in the surrounding environment changes between sites. The Friedman test indicate s that size classes do not respond to unique sets of environmental vari ation and size classes do not differ in their response to landscape, local, or sp atial variation. In other words, worker ant size structure is highly conservative in respons e to environmental variation at the scales measured. These results provide mixed support for the ab ility of the TDH to explain patterns in body size distributions. Prediction one, that ant body sizes should exhibit a discontinuous distribution, is clearly supporte d by the results of the BCART and hierarchical clustering analysis. Prediction two, that the body size dist ribution should change with landscape structure but not with species composition wa s partly supported. Body si ze distributions did not change with changes in landscape stru cture in opposition to prediction. Since the study was set up along 57

PAGE 58

a gradient representing most of the heterogeneity in landscape structure in the region, according to the TDH, the body size distribution should have changed along this gradient. However, these data support the second part of prediction two, since the size dist ribution did not change with turnover in species composition. Prediction th ree did not hold based on results from the Friedman tests. Each size class does not seem to be responding to a uniq ue set of environmental variation as predicted by the TDH and size classes do not differ in the strength of their response to landscape, local, and spatial variation. Though the results of prediction two provide some evidence against one of the main predictions of the TDH, namely that size distri butions will change to reflect changes in the structure of the landscape (Holling 1992, Allen et al. 2006), other factors should be considered. Ant colonies are formed of differe nt sized castes, each of which perform specific social functions (Hlldobler and Wilson 1990). The mass of the di fferent castes within sp ecies often does not scale linearly. For example, from species to species queens can vary greatly in size in comparison with workers. Worker ants (i.e., th ose analyzed in this study) generally perform their daily activities within a limited distance from the colony and are therefore greatly and directly influenced by local vari ation in the environment. Resu lts show that although landscape variation has important effects on ant community organization, landscape infl uences likely have only indirect effects on worker size, whereas inte raction with the local environment provides the evolutionary drivers that determine body size. Si nce an effort was made to standardize local conditions for habitat quality, the effects at this scale could not be fully explored. Body size in dispersing individuals (i.e., winged queens and males) may be more strongly tied to variation in landscape structure and may meet th e predictions of the TDH at this scale. It may be revealing to examine the interactions between the queen size-environment relationship and the worker 58

PAGE 59

size-environment relationship to help provide a better understanding as to how the TDH might play out across this range of scales. There is evidence that the TDH might be better applied to worker ants at more localized scales. Yanoviak and Kaspari (2000) sampled ants in a lowland forest of Barro Colorado Island, Panama and found a striking difference in the frequency distribution of ant body size between litter and canopy-dwelling ants. Th ey speculated that the differen ce in distribution was a result of differences in the local environment that can opy and litter ants inhabit within the broader habitat classification. Canopy ants are subject to a relatively smooth textured environment comprised of branching networks and leaves. Li tter ants, however, exist in a more coarsely textured environment of fallen leaves, twigs, a nd other detritus. More over, the size frequency distribution of litter inha biting ants was discontinuous, with an absence of intermediate-sized ants. The authors attributed this discontinuity to a habitat driven change in locomotory pattern. Large ants move over the litter, stepping over ga ps, while small ants walk through the litter by moving in the interstitial spaces. Intermediate-sized ants are ineffective at either mode of movement and are thus absent, resulting in two gr oups of ants that perceive and interact with their environment at different scales. Intere stingly, Yanoviak and Kaspari cite the size-grain hypothesis (Kaspari and Weiser 1999) as a possible explanation for this observation. Though apparently derived independently, the size-grain hypothesis shares important similarities with the TDH, and may prove to be usef ul in providing mechanistic e xplanations for some of the predictions of the TDH. For ex ample, it has been observed that there is a nonlinear relationship between body mass and leg length in ants, with larger bodied ants ha ving proportionally longer legs than smaller bodied ants (Kaspari and Weiser 1999, Farji-Br ener et al. 2004). The sizegrain hypothesis postulates that th e benefits of long legs (e.g., e fficient movement over planar 59

PAGE 60

60 surfaces) decrease disproportionately with size because long legs produce a larger cross-sectional area which impedes movement in more confined or complex areas, such as le af litter. Thus, the size-grain hypothesis predicts that ants will make a scale transition as they increase in size based on the nature of the surrounding environment (i.e., from planar to complex; Kaspari and Weiser 1999). This prediction links dire ctly to the predictions of the TDH; that a discontinuous distribution in body size will arise from species interactions with the environment at different scales of perception. While ant body size in this study follows a discontinuous distribution, results do not produce clear support for the TDH as a structuring mechanism in ant communities. Rather than invalidate the TDH, results of this study show a clear need for more empirical and experimental testing of the hypothesis. Such studies should ex plicitly consider the rang e of scales in which species have the strongest direct interaction with the environmen t in order to best understand how environmental forcings structure body size di stributions. When possible, the effects on body size distributions from scales above and below that of the TDH, such as those predicted by other size distribution hypotheses, should be incorporated so as to attain the most complete interpretation possible.

PAGE 61

Table 3-1. Pool of landscape, local, and sp atial variables. Significance values are reported for variables selected with forw ard regression for each size group and us ed in the final analyses ( P < .05). Size groups are listed in ascending order of mass with Group 1 being the smalle st and Group 6 the largest. Variable code Description Variable Type Group 1 Sig. Group 2 Sig. Group 3 Sig. Group 4 Sig. Group 5 Sig. Group 6 Sig. Sndhl Area sandhill Landscape DryPr Area dry prairie Landscape HdPin Area mixed hardwood-pine Landscape Hamk Area hardwood hammocks Landscape 0.001 Pine Area pinelands (pine plantation & pine flatwoods) Landscape 0.027 0.003 0.005 0.003 Mar Area freshwater marsh Landscape 0.026 0.016 D_Smp Area disturbed swamp Landscape 0.032 B_Smp Area bay swamp Landscape 0.050 C_Swp Area cypress swamp Landscape 0.025 0.017 WetFor Area wetland forest Landscape 0.001 HdSmp Area hardwood swamp Landscape 0.041 0.037 Water Area open water Landscape 0.031 D_For Area disturbed forest Landscape Past Area pasture Landscape Agric Area agriculture Landscape H_Urb Area high impact urban Landscape 0.025 L_Urb Area low impact urban Landscape Extr Area extractive Landscape 0.035 NumC Number land cover classes Landscape 0.039 MNN Mean nearest neighbor (all classes) Landscape 0.016 MSI Mean shape index (all classes) Landscape 0.041 MNN_SH Mean nearest neighbor (sandhill) Landscape 0.043 MSI_SH Mean shape index (sandhill) Landscape PA Focal patch area Landscape 0.028 0.001 PAR Focal patch perimeter/area ratio Landscape NSHP Number sandhill patches Landscape 0.019 0.047 APA Mean sandhill patch area Landscape 61

PAGE 62

62 Table 3-1 Continued Variable code Description Variable Type Group 1 Sig. Group 2 Sig. Group 3 Sig. Group 4 Sig. Group 5 Sig. Group 6 Sig. APP Mean sandhill patch perimeter Landscape NDVI Mean NDVI Landscape SD_NDVI Standard deviation NDVI Landscape Elev Elevation Local 0.005 A_Gras Mean grass cover Local SD_Gras Standard deviation grass cover Local A_Lit Mean litter cover Local SD_Lit Standard deviation litter cover Local 0.015 A_Herb Mean herbaceous cover Local SD_Herb Standard deviation herb aceous cover Local A_Shrb Mean shrub cover Local 0.019 SD_Shrb Standard deviation shrub cover Local A_Tree Mean tree cover Local SD_Tree Standard deviation tree cover Local A_Bare Mean bare ground Local 0.012 0.024 SD_Bare Standard deviation bare ground Local 0.002 0.002 A_DW Mean dead wood cover Local 0.027 0.017 SD_DW Standard deviation dead wood cover Local 0.021 PCNM1 PCNM Axis 1 (Broadest scale) Spatial 0.008 0.002 0.016 0.001 PCNM2 PCNM Axis 2 Spatial 0.030 0.001 0.011 0.024 PCNM3 PCNM Axis 3 Spatial 0.050 0.002 PCNM4 PCNM Axis 4 Spatial 0.037 0.010 0.011 PCNM5 PCNM Axis 5 Spatial PCNM6 PCNM Axis 6 Spatial PCNM7 PCNM Axis 7 Spatial 0.036 0.029 0.012 0.005 PCNM8 PCNM Axis 8 Spatial 0.050 00 PCNM9 PCNM10 PCNM Axis 9 PCNM Axis 10 Spatial Spatial 0.8 0.019 PCNM11 PCNM Axis 11 (Finest scale) Spatial 0.006

PAGE 63

Table 3-2. Unadjusted pr oportion of variation explai ned by the significant ( P < .05) environmental variables for each size group. Group1 Group2 Group3 Group4 Group5 Group6 Hamk 0.1131 Pine 0.0797 0.0914 0.0431 0.1020 Mar 0.0377 0.0699 D_Smp 0.0569 B_Smp 0.0362 C_Swp 0.1058 0.0716 WetFor 0.0589 HdSmp 0.1000 0.0562 Water 0.0745 H_Urb 0.0472 Extr 0.1067 NumC 0.0177 MNN 0.0758 MSI 0.0339 MNN_SH 0.0534 SD_Lit 0.0331 SD_Bare 0.1177 0.1176 SD_DW 0.0658 A_Bare 0.0791 0.0824 A_DW 0.0656 0.0091 PA 0.1097 0.1528 NSHP 0.0823 0.0226 Elev 0.0508 A_Shrb 0.0249 See Table 3-1 for definitions of environmental variables. 63

PAGE 64

Figure 3-1. Fractions of variati on explained in partial redundancy analysis. Each fraction can be analyzed separately or summed to anal yze a combined model. Total landscape variation, for example, is calculated by summing the amount of pure variation explained by landscape factors [a], and the e xplained variation shared with spatial and local factors [d+f+g]. Li kewise, pure environmental variation [a+b+d] can be separated from total spatial variation [c +e+f+g]. The proportion of variation explained by unmeasured factors is [h]. 64

PAGE 65

-2.5-2-1.5-1-0.500.51 1 2 3 4 5 6 log body mass (mg) 1 2 3 4 5 6 7 HCA BCART Figure 3-2. Body size groups (and th e discontinuities between groups ) derived from hierarchical clustering analysis (HCA) a nd Bayesian classification a nd regression tree analysis (BCART). Black diamonds indicate the (log) mass of each species. 65

PAGE 66

Figure 3-3. Proportion of variati on explained in each size group by the fractions of landscape, local, and spatial vari ation. Monte Carlo simulations (999 permutations under the full model) provide tests of significance for th e total landscape, local, and spatial models, as well as the pure landscape, lo cal, and spatial models (*** P = 0.001; ** P < 0.01; *P < 0.05; P < 0.1). Overall RDA models combining landscape, local and spatial variables all significantly explai n variation in each size group ( P = 0.001). 66

PAGE 67

CHAPTER 4 CONCLUSIONS Local and very broad-scale environmental interactions are relatively well described in communities of ants. However, little is know n about the ways in wh ich intermediate, or landscape-scale environmental variation can infl uence ant community structure. With the increasing threat of habitat loss, this remain s a challenging area of research for community ecologists. Two general questions were addr essed with this research: (1) how are ant communities organized by multiscale environmental and spatial variability and (2) how does the distribution of body size change w ith the surrounding environment? The overall objective of Chapter 2 was to investigate the effects of multiscale environmental and spatial variation on a set of dive rse ant assemblages. Ants at thirty-three sites were sampled in sandhill habitat and examined to determine how variation in the structure of the local environment and the surrounding landscape infl uences the organization of ant communities. A diverse mix of mostly native ant populati ons was found across the study region. Results indicate that ground-dwelling ant communities are strongly structured by both spatial and environmental variability across a range of scal es. Pineland – a land cover class comprised of commercial pine plantation and natural pine fl atwoods – had a strong effect on ant community structure. Increasing amount s of pineland in the study la ndscapes was indicative of fragmentation of sandhill habitat. A change in community structure was observed along a gradient formed by landscapes comprised of la rge proportions of pineland at one end, and landscapes comprised of large proportions of open stru ctured sandhill habitat, at the other. In landscapes with high proportions of pineland, habitat generalist spec ies appeared to replace open habitat specialists, even though the local habitat was standardized for quality across all sites. It was speculated that habitat generalists are able to maintain populations in the preferred habitat of 67

PAGE 68

open habitat specialists through a number of po ssible mechanisms, including spatial subsidies (Polis et al. 2004), source-sink movement (Holt 1993), and apparent competition (Holt 2004). The metacommunity concept (Holyoak et al. 2005) was used as a means to better understand the community patterns observed. The pure spatial and environmental effects suggest that a combination of the mass effects and species-sortin g perspectives of the metacommunity concept would be a good framework for investigating mechanisms of community orga nization in ants. It appears that the high diversity of ground-dwelling ants in sandhill habitat in northern Florida is maintained by a heterogeneous distribution of la nd cover types throughout the region. However, continued habitat loss and degradation (e.g., thr ough fire suppression) could shift the balance in favor of habitat generalists and reduce the viab ility of open habitat specialists in Florida sandhills. The textural discontinuity hypothesis (TDH) was examined in Chapter 3. Mounting evidence indicates that animal body size exhibi ts discontinuous, multimodal distributions, opposing the traditional vi ew of a smooth unimodal distribution (Allen 2006). The TDH predicts that body size distributions will be discontinuous which reflects the discontinuous and multiscale nature of structuring forces in ecosystems (Holling 1992). Size groups should therefore respond most strongly to different aspects of a shared environment and the distribution should change as the structure of the landscape changes. Body size in sandhill ants is discontinuously distributed as predicted by the TDH. Furthermore, the size distribution did not cha nge with changes in species composition as predicted. However, size groups do not appear to be influenced by unique combinations of environmental variability and th e size distribution does not cha nge with changes in landscape structure as predicted. It should be noted that the way ants are organized into social casts may 68

PAGE 69

mask the manifestation of some of the effects predicted by the TDH. Worker ants, the subject of this analysis, typically interact with the envir onment on a more localized scale as they gather resources and defend the colony, wh ereas alate individuals can inte ract with the environment at broader scales during dispersal events. Since variation in the surrounding landscape was maximized between sites while standardizing the local environment for hi gh quality habitat, a mismatch in the scale of analysis may have occurred. Stronger support for the TDH in ant species may be seen if the interaction between dispersing individuals and locally interacting individuals is considered. Evidence for the TDH has been seen in ground-dwelling ants in tropical forests (Yanoviak and Kaspari 2000). The observation of disc ontinuities in a size frequency distribution of leaf litte r ants was attributed to the si ze-grain hypothesis (Kaspari and Weiser 1999). The size-grain hypothesis makes similar predictions as the TDH and may prove to be an effective means for empirically testing the validity of the TDH. This thesis presents some of the first empirical evidence that ant communities are structured by ecological factors relating to the composition and struct ure of the surrounding landscape. It would seem that ant communities in the region examined here do not exist within isolated habitats, as they are frequently trea ted in observational and experimental studies. Rather, cross-system flows of, for example, resources, competitors, and/or predators, across habitat boundaries may have sign ificant effects on local assemb lages of ants depending on the composition and arrangement of the surrounding la ndscape. This study highlights the need for more research to be conducted on the landscape -scale effects on ant community structure, which account for the openness of ecosystems. 69

PAGE 70

APPENDIX SITE SELECTION The landscape data at the 500 m extent were reduced using principle components analysis. The first four axes explain over 43% of the variation in the data (Fig A-1). Hierarchical clustering (Ward’s method, Euc lidian distance) was performed on these 18 components using SPSS v11.5 (SPSS 2002). The resulting dendrogram was quantitatively pared at 22 groupings using indicator species analys is (Fig. A-2; Dufrne and Legendre 1997). Indicator species analysis was performed using PC-ORD v4.0 with 1000 randomizations (McCune and Mefford 1999). The 33 selected sites comprise 15 of the 22 na tural groups at the 500 m scale. The groups not sampled represent unusual and rare landscapes (only 7.6% of the selection pool) or highly distur bed habitat. The unusual and rare groups were excluded on logistical grounds. To standardi ze for local habitat quality, highly disturbed areas such as grazed land or long unburned areas were necessarily ex cluded from the study after field validation. 70

PAGE 71

-4 -3 -2 -1 0 1 2 3 4 5 3210123PC1 (20.82%)PC2 (9.74%) A Figure A-1. Principal components anal ysis illustrating the distributi on of selected sites across the range of landscape heterogeneity at the 500 m radius extent. Gray squares represent the selection pool of 2000 random points a nd black diamonds repr esent the selected sites. (A) Sites span the fu ll gradient of PC1 and most of PC2. PC1 is most strongly characterized by the number of land cove r classes and by variables describing the amount and arrangement of sandhill habitat (i .e., total area, focal patch area, number of sandhill patches, and perimeter to area ra tio). PC2 is most strongly characterized by the area of urban, pineland, and forested wetland land cover types. The portion of PC2 not sampled represents rare landscapes (< 5% of the selecti on pool) with either very high proportions of urban land cover or very high proportions of forested wetland. The focal patches within these fe w landscapes were too degraded to be considered for sampling given the require ment of high quality local habitat at sampling sites. (B) The selected sites span most of the gradients of PC3 and PC4. PC3 is most strongly characterized by a di fferent combination of sandhill amount and arrangement variables. PC4 is most st rongly characterized by the area of mixed hardwood forests and hammocks as well as the area of wetland forests and cypress swamp. 71

PAGE 72

-9 -7 -5 -3 -1 1 3 5 2101234PC3 (6.45%)PC4 (6.31%) B Figure A-1 Continued 72

PAGE 73

30 33 36 39 42 45 5101520253035404550Number of clustersMeaniIndicator value Figure A-2. Indicator species analys is of landscape groupings. Hierarchical clustering analysis was used to group landscapes into similar types based on landscape metrics measured in areas surrounding 2000 points randomly ge nerated in sandhill habitat throughout the study region. Indicator sp ecies analysis was used to determine an ecologically meaningful number of landscape groupings. In this case, “species” are landscape metrics and the indicator value represents the degree to which a landscape metric faithfully represents a particular landscape group. Separate indica tor species analyses were conducted based on cluster membership in each step of cluster formation (10 to 45 clusters). A mean indicator value for a ll metrics was calculated for each of the 36 separate analyses (i.e., 10-45 clusters). The greatest mean indi cator value (42.7; open circle) is associated with 22 clusters suggesting that landscape variation throughout the region is best represented by 22 types of landscapes. 73

PAGE 74

LIST OF REFERENCES Abrahamson, W. G., and D. C. Hartnett. 1990. Pi ne flatwoods and dry prairies. Pages 103-149 in R. L. Myers and J. J. Ewel, editors. Ecosyste ms of Florida. University of Central Florida Press, Orlando, Florida, USA. Agosti, D., J. D. Majer, L. E. Alonso, and T. R. Schultz, editors. 2000. Ants: Standardized Methods for Measuring and Monitoring Biodi versity. Smithsonian Institution Press, Washington, D.C., USA. Allen, C. R. 2006. Discontinuities in ecological da ta. Proceedings of th e National Academy of Sciences of the United Stat es of America 103:6083-6084. Allen, C. R., E. A. Forys, and C. S. Holling. 1999. Body mass patterns predict invasions and extinctions in transforming la ndscapes. Ecosystems 2:114-121. Allen, C. R., A. S. Garmestani, T. D. Havlicek, P. A. Marquet, G. D. Peterson, C. Restrepo, C. A. Stow, and B. E. Weeks. 2006. Patterns in body mass distributions: sifting among alternative hypotheses. Ecology Letters 9:630-643. Borcard, D., and P. Legendre. 2002. All-scale spatial analysis of ecological data by means of principal coordinates of neighbour ma trices. Ecological Modelling 153:51-68. Borcard, D., P. Legendre, C. Avois-Jacquet, and H. Tuomisto. 2004. Dissecting the spatial structure of ecological data at multiple scales. Ecology 85:1826-1832. Borcard, D., P. Legendre, and P. Drapeau. 1992. Partialling out the spatial component of ecological variation. Ecology 73:1045-1055. Boulton, A. M., K. F. Davies, and P. S. Ward. 2005. Species richness, abundance, and composition of ground-dwelling ants in northern California grasslands: role of plants, soil, and grazing. Environmental Entomology 34:96-104. Brooks, J. L., and S. I. Dodson. 1965. Predat ion, body size, and composition of plankton. Science 150:28-35. Brown, J. H., and A. Kodric-B rown. 1977. Turnover rates in insu lar biogeography: effect of immigration on extinction. Ecology 58:445-449. Brown, J. H., P. A. Marquet, and M. L. Taper. 1993. Evolution of body-size: Consequences of an energetic definition of fitness. American Naturalist 142:573-584. Bruna, E. M., H. L. Vasconcelos, and S. Heredi a. 2005. The effect of habitat fragmentation on communities of mutualists: Amazonian ants and their host plants. Biological Conservation 124:209-216. 74

PAGE 75

Cadenasso, M. L., S. T. A. Pickett, and K. C. Weathers. 2004. Effect of landscape boundaries on the flux of nutrients, detritus, and organisms. Pages 154-168 in G. A. Polis, M. E. Power, and G. R. Huxel, editors. Food Webs at th e Landscape Level. Un iversity of Chicago Press, Chicago, Illinois, USA. Carroll, M. L., C. M. DiMiceli, R. A. Sohl berg, and J. R. G. Townshend. 2005. 250m MODIS Normalized Difference Vegetation Index. The Global Land Cover Facility, College Park, Maryland, USA. Chesson, P. 2000. General theory of competitive co existence in spatially-varying environments. Theoretical Population Biology 58:211-237. Chipman, H. A., E. I. George, and R. E. McCulloch. 1998. Bayesian CART model search. Journal of the American Sta tistical Association 93:935-948. Chipman, H. A., E. I. George, and R. E. McCulloch. 2006. Bayesian CART. URL http://gsbwww.uchicago.edu/fac/robert.mccu lloch/research/code/CART/index.html. Last accessed June 2006. Colwell, R. K. 2005. EstimateS: Statistical esti mation of species richness and shared species from samples. Version 7.5.0. Persistent URL . Last accessed April 2006. Cook, W. M., K. T. Lane, and R. D. Holt. 2002. Island theory, matrix effects and species richness patterns in habitat fr agments. Ecology Letters 5:619-623. Cottenie, K., and L. De Meester. 2005. Local in teractions and local dispersal in a zooplankton metacommunity. Pages 189-211 in M. Holyoak, M. Leibold, and R. D. Holt, editors. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago, Illinois, USA. Cumming, G. S., and T. D. Ha vlicek. 2002. Evolution, ecology, and multimodal distributions of body size. Ecosystems 5:705-711. Cumming, G. S., and B. J. Spiesman. 2006. Regiona l problems need integrated solutions: pest management and conservation biology in agroecosystems. Biological Conservation 131:533-543. Cushman, J. H., J. H. Lawton, and B. F. J. Manly. 1993. Latitudinal patterns in European ant assemblages: Variation in species ri chness and body size. Oecologia 95:30-37. Davidson, D. W. 1977a. Foraging ecology and community organiza tion in desert seed-eating ants. Ecology 58:725-737. Davidson, D. W. 1977b. Species dive rsity and community organization in desert seed-eating ants. Ecology 58:711-724. 75

PAGE 76

Debinski, D. M., and R. D. Holt. 2000. A surv ey and overview of habitat fragmentation experiments. Conser vation Biology 14:342-355. Deyrup, M. 2003. An updated list of Florida ants (Hymenoptera: Formicidae). Florida Entomologist 86:43-48. Deyrup, M., and J. Trager. 1986. Ants of the Ar chbold Biological Stat ion, Highlands County, Florida (Hymenoptera, Formicidae) . Florida Entomologist 69:206-228. Deyrup, M. A., N. Carlin, J. Trager, and G. Umphrey. 1988. A Review of the Ants of the Florida Keys. Florida Entomologist 71:163-176. Didham, R. K., J. Ghazoul, N. E. Stork, and A. J. Davis. 1996. Insects in fragmented forests: A functional approach. Trends in Ecology & Evolution 11:255-260. Dufrne, M., and P. Legendre. 1997. Species asse mblages and indicator species: The need for a flexible asymmetrical approac h. Ecological Monographs 67:345-366. Dunn, R. R. 2005. Modern insect extinctions, the neglected majority. Conservation Biology 19:1030-1036. Edmisten, J. E. 1963. The ecology of Florida pine flatwoods. Ph.D. Dissertation. University of Florida, Gainesville, Florida, USA. Elton, C. 1927. Animal Ecology. Sidgewick and Jackson, London, UK. Environmental Systems Research Institute (ESRI). 1999. ArcView version 3.3. Redlands, California, USA. Fahrig, L. 2002. Effect of habitat fragmentati on on the extinction threshold: A synthesis. Ecological Appli cations 12:346-353. Fahrig, L. 2003. Effects of habitat fragmentation on biodiversity. Annual Review of Ecology Evolution and Systematics 34:487-515. Farji-Brener, A. G., G. Barrantes, and A. Ruggiero. 2004. Environmental rugosity, body size and access to food: a test of the size-grain hypothesis in tropical litter ants. Oikos 104:165171. Gauld, I. D., and B. Bolton. 1988. The Hymenopter a. Oxford University Press, New York. Gehring, T. M., and R. K. Swihart. 2003. Body size, niche breadth, and ecologically scaled responses to habitat fragmentation: mammalia n predators in an agricultural landscape. Biological Conservation 109:283-295. Gotelli, N. J., and A. M. Ellison. 2002. Biogeogra phy at a regional scale: Determinants of ant species density in New England bogs and forests. Ecology 83:1604-1609. 76

PAGE 77

Harmon, G. D., and N. E. Stamp. 1992. Effects of postdispersal seed predation on spatial inequality and size variabili ty in an annual plant, Erodium cicutarium (Geraniaceae). American Journal of Botany 79:300-305. Havlicek, T. D., and S. R. Carpenter. 2001. Pelagi c species size distributions in lakes: Are they discontinuous? Limnology a nd Oceanography 46:1021-1033. Henschel, J. R., D. Mahsberg, and H. Stum pf. 2001. Allochthonous aquatic insects increase predation and decrease herbivory in river shore food webs. Oikos 93:429-438. Hoffmann, B. D., and A. N. Andersen. 2003. Responses of ants to disturbance in Australia, with particular reference to functi onal groups. Austral Ecology 28:444-464. Holland, J. D., L. Fahrig, and N. Cappuccino. 2005. Body size affects the spat ial scale of habitatbeetle interactions. Oikos 110:101-108. Hlldobler, B., and E. O. Wilson. 1990. The Ants . Harvard University Press, Cambridge, Massachusetts, USA. Holling, C. S. 1992. Cross-scale morphology, geometry, and dynamics of ecosystems. Ecological Monographs 62:447-502. Holt, R.D. 1993. Ecology at the mesoscale: The influence of regional processes on local communities. Pages 77-88 in R. Ricklefs and D. Schluter, editors. Species Diversity in Ecological Communities, University of Chicago Press: Chicago. Holt, R. D. 2004. Implications of system openne ss for local community structure and ecosystem function. Pages 96-114 in G. A. Polis, M. E. Power, and G. R. Huxel, editors. Food Webs at the Landscape Level. University of Chicago Press, Chicago, Illinois, USA. Holt, R. D., and M. Barfield. 2003. Impacts of te mporal variation on apparent competition and coexistence in open ecosystems. Oikos 101:49-58. Holyoak, M., M. A. Leibold, and R. D. Holt, editors. 2005. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago, Illinois, USA. Hubbell, S. P. 2001. The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press, Princeton, New Jersey, USA. Hutchinson, G. E. 1959. Homage to Santa Rosalia or why are there so many kinds of animals? American Naturalist 93:145-159. Izhaki, I., D. J. Levey, and W. R. Silva. 2003. Effects of prescribed fire on an ant community in Florida pine savanna. Ec ological Entomology 28:439-448. Kaspari, M., L. E. Alonso, and S. O'Donnell. 2000. Three energy variables predict ant abundance at a geographic scale. Proceedings of the Royal Society of London, Series B 267:485489. 77

PAGE 78

Kaspari, M., and M. D. Weiser. 1999. The size-gra in hypothesis and interspecific scaling in ants. Functional Ecology 13:530-538. Kaspari, M., M. Yuan, and L. E. Alonso. 2003. Spa tial grain and the causes of regional diversity gradients in ants. The Ameri can Midland Naturalist 161:459-477. Kellert, S. R. 1993. Values and perceptions of invertebrates. Cons ervation Biology 7:845-855. King, J. R. 2004. Ant communities of Florida's upland ecosystems: ecology and sampling. Ph. D. Dissertation. University of Florida, Gainesville, FL. Kneitel, J. M., and J. M. Chase. 2004. Trade-o ffs in community ecology: linking spatial scales and species coexistence. Ecology Letters 7:69-80. Kremen, C. 2005. Managing ecosystem services: what do we need to know about their ecology? 8:468-479. Kremen, C., N. M. Williams, and R. W. Thor p. 2002. Crop pollination from native bees at risk from agricultural intensification. Proceedi ngs of the National Academy of Science USA 99:16812-16816. Lassau, S. A., G. Cassis, P. K. J. Flemons, L. Wilkie, and D. F. Hochuli. 2005. Using highresolution multi-spectral imager y to estimate habitat complexity in open-canopy forests: can we predict ant community patterns? Ecog raphy 28:495-504. Legendre, P., and E. D. Gallagher. 2001. Ecologica lly meaningful transformations for ordination of species data. Oecologia 129:271-280. Legendre, P., and L. Legendre. 1998. Numerical Ecology, second edit ion. Elsevier B.V., Amsterdam, The Netherlands. Leibold, M. A., M. Holyoak, N. Mouquet, P. Amaras ekare, J. M. Chase, M. F. Hoopes, R. D. Holt, J. B. Shurin, R. Law, D. Tilman, M. Loreau, and A. Gonzalez. 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecology Letters 7:601-613. Levins, R. 1969. Some demographic and genetic c onsequences of environmental heterogeneity for biological control. Bulletin of the Entomology Society of America 71:237-240. Lubertazzi, D., and W. R. Tschinkel. 2003. An t community change ac ross a ground vegetation gradient in north Florida's longleaf pine flatwoods. Journal of Insect Science 3:1-17. MacArthur, R. H., and E. O. Wilson. 1967. The theory of island biogeography. Princeton University Press, Princeton, NJ. Marquet, P. A., and H. Cofre. 1999. Large tempor al and spatial scales in the structure of mammalian assemblages in South America: a macroecological approach. Oikos 85:299309. 78

PAGE 79

Marquet, P. A., S. A. Navarrete, and J. C. Castilla. 1995. Body-size, popu lation-density, and the energetic equivalence rule. Journal of Animal Ecology 64:325-332. McCune, B., and M. J. Mefford. 1999. Multivariate analysis of ecological data, PC-ORD version 4.39. MjM Software, Gleneden Beach, Oregon, USA. Moksnes, P. O., L. Pihl, and J. van Montfrans. 1998. Predation on postlarv ae and juveniles of the shore crab Carcinus maenas : importance of shelter, size and cannibalism. Marine Ecology-Progress Series 166:211-225. Mouquet, N., M. F. Hoopes, and P. Amarasekare. 2005. The world is patchy and heterogeneous! Trade-off and source-sink dynamics in competitive metacommunities. Pages 237-262 in M. Holyoak, M. A. Leibold, and R. D. Holt, editors. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press, Chicago, Illinois, USA. Myers, R. L. 1990. Scrub and high pine. Pages 150-193 in R. L. Myers and J. J. Ewel, editors. Ecosystems of Florida. University of Ce ntral Florida Press, Orlando, Florida, USA. Myers, R. L., and J. J. Ewel, editors. 1990. Ecosyste ms of Florida. University of Central Florida Press, Orlando, Florida, USA. Ness, J. H. 2004. Forest edges and fire ants alte r the seed shadow of an ant-dispersed plant. Oecologia 138:448-454. Ness, J. H., J. L. Bronstein, A. N. Andersen, and J. N. Holland. 2004. Ant body size predicts dispersal distance of ant-adapted seeds: im plications of small-ant invasions. Ecology 85:1244-1250. Noss, R. F., E. T. LaRoe, III, and J. M. Sc ott. 1995. Endangered Ecosystems of the United States: a Preliminary Assessm ent of Loss and Degradati on. National Biologic Survey Biological Report 28, Department of the Interior, Washington, DC. Novacek, M. J., and E. E. Cleland. 2001. The curr ent biodiversity extinct ion event: Scenarios for mitigation and recovery. Proceedings of the National Academy of Science USA 98:54665470. Oksanen, J., R. Kindt, L. Legendre, and B. O' Hara. 2006. vegan: Community Ecology Package. Version 1.8-2. URL http://cc.oulu.fi/~j arioksa/. Last accessed June 2006. Oksanen, L., S. D. Fretwell, and O. Jarvine n. 1979. Interspecific aggression and the limiting similarity of close competitors problem of size gaps in some community arrays. American Naturalist 114:117-129. Palmer, T. M. 2003. Spatial habitat heterogeneity influences competition and coexistence in an African acacia ant guild. Ecology 84:2843-2855. Pimm, S. L., and P. Raven. 2000. Biodiversity – Extinction by numbers. Nature 403:843-845. 79

PAGE 80

Polis, G. A., W. B. Anderson, and R. D. Holt. 1997. Toward an integration of landscape and food web ecology: the dynamics of spatially subs idized food webs. Annual Review of Ecology and Systematics 28:289-316. Polis, G. A., M. E. Power, and G. R. Huxel, editors. 2004. Food Webs at the Landscape Level. University of Chicago Press, Chicago, Illinois, USA. R Development Core Team. 2004. R: A language a nd environment for statistical computing. R Foundation for Statistical Co mputing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org. Rahn, M., H. Doremus, and J. E. Diffendorfer. 2006. Species coverage in multispecies habitat conservation plans: Where's th e science? Bioscience 56:613-619. Rand, T. A., J. M. Tylianakis, and T. Tscharnt ke. 2006. Spillover edge effects: the dispersal of agriculturally subsidized insect natural enemies into adjacent natural habitats. Ecology Letters 9:603-614. Ribas, C. R., T. G. Sobrinho, J. H. Schoereder, C. F. Sperber, C. Lopes-Andrade, and S. M. Soares. 2005. How large is large enough for insect s? Forest fragmenta tion effects at three spatial scales. Acta Oecologica-Inte rnational Journal of Ecology 27:31-41. Ricklefs, R. E. 1987. Community diversity Relative roles of local and regional processes. Science 235:167-171. Ritchie, M. E., and H. Olff. 1999. Spatial scaling laws yield a s ynthetic theory of biodiversity. Nature 400:557-560. Rooney, N., K. McCann, G. Gellner, and J. C. Moore. 2006. Structural asymmetry and the stability of diverse f ood webs. Nature 442:265-269. Samways, M. J. 2005. Insect Diversity Conserva tion. Cambridge University Press, Cambridge, United Kingdom. SAS Institute Inc. 1999. SAS/STAT User's Guide, Version 7. SAS Institute Inc., Cary, North Carolina, USA. Scheffer, M., and E. H. van Nes. 2006. Self-organi zed similarity, the evol utionary emergence of groups of similar species. Proceedings of the National Academy of Sciences of the United States of America 103:6230-6235. Schoereder, J. H., T. G. Sobrinho, C. R. Riba s, and R. B. F. Campos. 2004. Colonization and extinction of ant communities in a fragmented landscape. Austral Ecology 29:391-398. Siegel, S., and N. J. Castellan, Jr. 1988. Nonpara metric Statistics for th e Behavioral Sciences, 2nd edition. McGraw-Hill, Inc ., New York, New York, USA. SPSS. 2002. SPSS for Windows, release 11.5.0. SPSS Inc. 80

PAGE 81

Stamp, N. E., and J. R. Lucas. 1990. Spatial patterns and dispersal dist ances of explosively dispersing plants in Florida sandhill vegetation. Journal of Ecology 78:589-600. Steffan-Dewenter, I. 2003. Importance of habitat area and landscape context for species richness of bees and wasps in fragmented orchar d meadows. Conservation Biology 17:1036-1044. Stys, B., R. Kautz, D. Reed, C. Keller, and A. Davis. 2004. Florida vegetation and land cover 2003. Florida Fish and Wildlife Conservation Commission, Tallahassee, Florida, USA. Suarez, A. V., D. T. Bolger, and T. J. Ca se. 1998. Effects of fragme ntation and invasion on native ant communities in coastal southern California. Ecology 79:2041-2056. ter Braak, C. J. F., and P. milauer. 2002. CANOCO reference manual and CanoDraw for Windows user's guide: software for ca nonical community ordination (version 4.5). Microcomputer Power, It haca, New York, USA. Thies, C., I. Steffan-Dewent er, and T. Tscharntke. 2003. Effe cts of landscape context on herbivory and parasitism at diffe rent spatial scales. Oikos 101:18-25. Tscharntke, T., T. A. Rand, and F. Bianchi. 2005 . The landscape context of trophic interactions: insect spillover across the crop-noncrop inte rface. Annales Zoologici Fennici 42:421-432. Van Pelt, A. F., Jr. 1958. The ecology of the ants of the Welaka Reserve, Florida (Hymenoptera: Formicidae). Part II annotated list. The American Midland Naturalist 59:1-57. Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397. Wilson, D. S. 1992a. Complex interactions in metacommunities, with implications for biodiversity and higher levels of selection. Ecology 73:1984-2000. Wilson, E. O. 1992b. The Diversity of Life. Harvard University Press, Cambridge, Massachusetts, USA. Wilson, E. O. 2000. Foreword. Pages xv-xvi in D. Agosti, J. D. Majer, L. E. Alonso, and T. R. Schultz, editors. Ants: Sta ndard Methods for Measuring and Monitoring Biodiversity. Smithsonian Institution Pre ss, Washington, D.C., USA. Yanoviak, S. P., and M. Kaspari. 2000. Community structure and the habita t templet: ants in the tropical forest canopy and litter. Oikos 89:259-266. Yu, D. W., H. B. Wilson, and N. E. Pierce. 2001. An empirical m odel of species co existence in a spatially structured envi ronment. Ecology 82:1761-1771. 81

PAGE 82

BIOGRAPHICAL SKETCH Brian J. Spiesman was born August 25, 1974, in Everett, Washington. In the summer of 2004 a small young dog appeared at the research st ation where Brian was working in Puerto Rico. It was later arranged for the dog (Reggie) to accompany Brian ba ck home. Reggie is a black shepherd-chow mix that enjoys playing Frisbee, digging holes, barking at strangers, and scratching. The majority of Reggie’s day is spent sleeping but he perks up if he catches the scent of peanut butter or meat. Going for a walk is his favorite activity. He usually goes on two walks each day (three if he can convince otherwise). Reggie does not care for the pest control guy or small children and the feeling seems to be mu tual. Reggie received a degree from doggie obedience school in the winter of 2005. Brian has yet to achieve this educational milestone. 82