xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001994086
007 cr mnu|||uuuuu
008 090327s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002414
Halstead, Brian J.
Predator behavior and prey demography in patchy habitats
h [electronic resource] /
by Brian J. Halstead.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 133 pages.
Dissertation (Ph.D.)--University of South Florida, 2008.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: Habitat loss and fragmentation are among the greatest threats to biodiversity, and these threats can be exacerbated or alleviated by the presence of interacting species. The effect of habitat loss and fragmentation on predator-prey systems has received extensive theoretical attention, but empirical studies of these systems yield few clear patterns. I examined the influence of prey abundance and spatial distribution on the foraging ecology and spatial ecology of Masticophis flagellum (Coachwhip) using capture-mark-recapture and radio telemetry techniques. I also examined the influence of saurophagous snake abundance on the survival rate of Sceloporus woodi (Florida Scrub Lizard) populations. Masticophis flagellum positively selected lizard and mammal prey, but within these categories it consumed prey species in proportion to their availability. Masticophis flagellum was vagile and constrained its movements within large home ranges. At all spatial scales examined, M. flagellum strongly selected Florida scrub habitat and avoided wetland habitats. The negative effect of saurophagous snake abundance best explained differences in S. woodi survival rates among patches of Florida scrub. Further loss and fragmentation of Florida scrub habitat will likely have a strong negative impact upon M. flagellum. Because it is precinctive to Florida scrub, Sceloporus woodi will also be negatively affected by the loss of this unique habitat. The potential positive effects of reduced predation pressure from M. flagellum that may accompany loss and fragmentation of Florida scrub is likely to be offset by increased predation rates by habitat and dietary generalist predators that incidentally prey upon S. woodi. Despite the sensitivity of these species to loss and fragmentation of Florida scrub, the prognosis is good for both M. flagellum and S. woodi on relatively large protected sites containing xeric habitats managed with prescribed fire.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Co-advisor: Earl D. McCoy, Ph.D.
Co-advisor: Henry R. Mushinsky, Ph.D.
t USF Electronic Theses and Dissertations.
Predator Behavior and Prey Demography in Patchy Habitats by Brian J. Halstead A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Biology, Division of Integrative Biology College of Arts and Sciences University of South Florida Co-Major Professor: Earl D. Mccoy, Ph.D. Co-Major Professor: Henry R. Mushinsky, Ph.D. Gary R. Huxel, Ph.D. Gordon Fox, Ph.D. Date of Approval: March 28, 2008 Keywords: Florida scrub, metapopulation, reso urce selection, survival, vertebrate Copyright 2008, Brian J. Halstead
Dedication To my late grandfathers, James Halstead, Sr. and Leonard Schartner, who instilled in me a deep interest ecology and natu ral history, an appreciation fo r education and attention to detail, and the belief that anything worth doing is worth doing we ll. The greatest honor this dissertation could re ceive is their approval.
Acknowledgements No worthwhile endeavor is ever accomplished alone. I am indebted to my wife, Kelly Browning, for her help with reviewing manuscr ipts, supplying help and company in the field, and inspiring me to keep things in pe rspective. My parents, Tara and James E. Halstead, Jr., were always very supportiv e and caring, and even took time out of their Florida vacation to help me re-install traps that had been uprooted by Hurricane Charley. Neal Halstead, my brother and colleague, was a tremendous asset. His help in the field and input on field and analytical methods were indispensable. My co-advisors, Drs. Earl McCoy and Henry Mushinsky, and committee members, Drs. Gary Huxel and Gordon Fox, were adept at providing just the right amount of guidance their efforts were essential to the successful completion of my dissertation. The staff at Lake Wales Ridge State Forest, particularly A nne Malatesta, Dave Butcher, and Dawn Johnson, were very accommodating and provided logistic support. Margi Baldwin, Dee Caretto, and Dr. Creighton Trahan provided excellent training in surgical technique and provided support for surgical materials and methods. Numerous graduate students at the University of South Florida, including Irmgard Lukanik, Katie Basiotis, Jennifer Rhora, Celina Bellanceau, Shannon Gonzalez, Joel Johns on, David Karlen, Alison Meyers, Kris Robbins, Travis Robbins, and Jenny Sneed helped with various aspects of this research. Many undergraduate assistants were essential for the completion of this dissertation, and deserve greater recognition than space allows.
i Table of Contents List of Tables iii List of Figures v Abstract vii Background and Hypotheses 1 Study System: The Florida Scrub 2 Study Organisms 2 Study Questions and Hypotheses 7 Sympatric Masticophis flagellum and Coluber constrictor Select Vertebrate Prey at Different Phylogenetic Levels 12 Materials and Methods 14 Study Site 14 Field Methods 15 Analytical Methods 16 Results 23 Diet Composition 23 Diet Selection 25 Niche Breadth and Overlap 26 Discussion 26
ii Masticophis flagellum (Coachwhip) Positively Selects Florida Scrub Habitat at Multiple Spatial Scales 47 Materials and Methods 48 Study Site 48 Field Methods 50 Analytical Methods 52 Results 59 Discussion 62 A Greater Abundance of Snakes Results in Lower Survival Rates of Sceloporus woodi Populations 90 Materials and Methods 91 Study Site 91 Field Methods 92 Analytical Methods 93 Results 96 Discussion 97 Conclusion: The Predicted E ffects of Snake Predation upon Sceloporus woodi Populations 109 Literature Cited 123 About the Author End Page
iii List of Tables Table 1. Prey Species Consumed by and Available to Individual Masticophis flagellum 32 Table 2. Prey Species Consumed by and Available to Individual Coluber constrictor 34 Table 3. Manlys Standardized Selection Ratios ( B ) for Prey Categories Selected by Masticophis flagellum and Coluber constrictor 39 Table 4. Manlys Standardized Selection Ratios ( B ) for Prey Species Selected by Masticophis flagellum 40 Table 5. Manlys Standardized Selection Ratios ( B ) for Prey Species Selected by Coluber constrictor 41 Table 6. Dietary Niche Breadth (Hurlber ts B) and Niche Overlap (Morisitas C) of Masticophis flagellum and Coluber constrictor. 42 Table 7. Characteristics of Radio Tracked Masticophis flagellum 70 Table 8. Home Range Estimates for Masticophis flagellum 71 Table 9. Masticophis flagellum Mean Daily Displacement. 72 Table 10. Habitat Use and Availability for Masticophis flagellum 74 Table 11. Manlys Standardiz ed Selection Ratios ( B ) for Habitats Selected by Masticophis flagellum 75
iv Table 12. Summary of Models Evaluated for Habitat Selection by Masticophis flagellum 76 Table 13. Florida Scrub Patch Use and Availability for M. flagellum 77 Table 14. POPAN Jolly-Seber Abundance Estimates for Lizard and Mammalian Prey in Each Sampled Patch of Florida Scrub. 79 Table 15. Summary of Models Evaluated for Florida Scrub Patch Selection by Masticophis flagellum 81 Table 16. Manlys Standardiz ed Selection Ratios ( B ) for Florida Scrub Patches Selected by Masticophis flagellum 83 Table 17. Home Range Sizes and Movement Distances of Masticophis flagellum at Different Locations. 84 Table 18. Relative Importance ( w+) of Environmental and Individual Characteristics for Determining the Daily (Re)capture Rate of Sceloporus woodi 101 Table 19. Support of Cormack-Jolly-Seber Models for Sceloporus woodi Populations. 102 Table 20. Relative Importance ( w+) of Variables for Determining the Survival Rate of Sceloporus woodi Populations. 103
v List of Figures Figure 1. Occurrence of Prey Categories Consumed by Masticophis flagellum and Coluber constrictor Expressed as a Proportion of Snakes of Each Species that Consumed Each Prey Category. 43 Figure 2. Occurrence of Prey Species Consumed by Masticophis flagellum and Coluber constrictor Expressed as a Proportion of Individuals of Each Species that Consumed Each Prey Species. 44 Figure 3. Relationship between Prey Category and Predator Size (Snout-vent Length) for Masticophis flagellum and Coluber constrictor 45 Figure 4. Ln-transformed Prey Mass as a Function of Ln-transformed Snake Mass for Masticophis flagellum and Coluber constrictor 46 Figure 5. Map of the Study Site within the Lake Arbuckle Tract of Lake Wales Ridge State Forest in Ce ntral Florida, USA. 85 Figure 6. Masticophis flagellum Net Squared Displacement Versus Random Walk Expectations. 86 Figure 7. Masticophis flagellum 100% Minimum Convex Polygon Home Ranges. 87 Figure 8. Mosaicplot of the Frequency of Behaviors within Habitats. 88 Figure 9. Correspondence Plot of the A ssociation of Snake Behaviors with Habitats. 89
vi Figure 10. Map of the Study Site within the Lake Arbuckle Tract of Lake Wales Ridge State Forest in Central Florida. 104 Figure 11. Schematic Diagram Depicting the Layout of Each Trap Array. 105 Figure 12. Model-averaged Estimates of Abundance for Sceloporus woodi Populations at Each Trap A rray Grouped by Patch Size. 106 Figure 13. Model-averaged Estimates of Recapture Probability for Each Secondary Period for Sceloporus woodi Populations. 107 Figure 14. Model-averaged Estimates of Su rvival Rate for Each Primary Period for Sceloporus woodi Populations. 108
vii Predator Behavior and Prey Demography in Patchy Habitats Brian Halstead ABSTRACT Habitat loss and fragmentation are among the greatest threats to bi odiversity, and these threats can be exacerbated or alleviated by th e presence of interacting species. The effect of habitat loss and fragmentation on predat or-prey systems has received extensive theoretical attention, but empirical studies of these syst ems yield few clear patterns. I examined the influence of prey abundan ce and spatial distribution on the foraging ecology and spatial ecology of Masticophis flagellum (Coachwhip) using capture-markrecapture and radio telemetry techniques. I also examined the influence of saurophagous snake abundance on the survival rate of Sceloporus woodi (Florida Scrub Lizard) populations. Masticophis flagellum positively selected lizard and mammal prey, but within these categories it consumed prey sp ecies in proportion to their availability. Masticophis flagellum was vagile and constrained its movements within large home ranges. At all spatial scales examined, M. flagellum strongly selected Florida scrub habitat and avoided wetland habitats. The negative effect of saurophagous snake abundance best explained differences in S. woodi survival rates among patches of Florida scrub. Further loss and fragmentation of Florida scrub habitat will likely have a strong negative impact upon M. flagellum Because it is precinctive to Florida scrub, Sceloporus woodi will also be negatively affected by the loss of this unique habitat. The potential
viii positive effects of reduced predation pressure from M. flagellum that may accompany loss and fragmentation of Florida scrub is likel y to be offset by incr eased predation rates by habitat and dietary generalist pred ators that incidentally prey upon S. woodi Despite the sensitivity of thes e species to loss and fragmentati on of Florida scrub, the prognosis is good for both M. flagellum and S. woodi on relatively large pr otected sites containing xeric habitats managed with prescribed fire.
1 Background and Hypotheses Habitat loss and fragmentation are am ong the greatest threat s to biodiversity (Pimm and Raven 2000). The impact of loss and fragmentation of habitat upon populations can be exacerbated or alleviated by the presence of other interacting species (Kareiva 1990, Ryall and Fahrig 2006). The e ffect of habitat loss and fragmentation on predator-prey systems has rece ived extensive theoretical at tention, but empirical studies of these systems yield few clear patterns (R yall and Fahrig 2006). Specialist predators depend upon their prey for positive population gr owth and are restricted to the same habitats as their prey. Loss and fragmentation of prey habitat is therefore predicted to be detrimental to specialist predators, but the e ffects of specialist pred ation can increase or decrease the risk of extinction of prey co mpared to the effects of habitat loss and fragmentation alone (Bascompte and Sole 1998, Prakash and de Roos 2002). As dietary and habitat breadth of the pred ator increases, negative effects of predation exacerbate the negative effects of habitat loss and fragmentation and the result is increased extinction risk of the prey (Swihart et al. 2001, Melian and Bascompte 2002). My goal was to examine the interaction of a wide-ranging pred ator with a habitat specialist prey species occurring in a system of hab itat patches, and to evaluate theoretical predictions of the consequences of this interaction for the persistence of the predator and its prey.
2 STUDY SYSTEM: THE FLORIDA SCRUB Florida scrub habitat is a unique ecosy stem of great conservation importance. Florida scrub in the interior of peninsular Florida originat ed as sand dunes in the early Pleistocene; during the Pliocene, only the very highest of the interi or ridges remained above sea level (Myers 1990). Scrub soils are extremely well-drained, nutrient-poor sands that support a xerophytic vegetation. Florid a scrub is pyrogenic, and fires typically occur at 10 to 100 year or greater interv als (Myers 1990). The great age and former isolation as islands has contributed to the occurrence of many organisms precinctive to Florida scrub, and the Lake Wales Ridge in particular. Forty to sixty percent of the species found in Florida scrub are precinctive to th is habitat (Myers 1990). More than 70 percent of one of the interior ridges, the Lake Wales Ridge, ha s been lost to agricultural and residential development over the past fi fty years (Myers 1990). Therefore, protection of existing Florida scrub and ecological data for Florida scrub-precinctive organisms are high conservation priorities. My study was c onducted at the Lake Arbuckle Tract of the Lake Wales Ridge State Forest (LWRSF), a protected site located on the Lake Wales Ridge. STUDY ORGANISMS Sceloporus woodi is a terrestrial phrynosomatid liz ard precinctive to Florida scrub habitat. Its geographic range is contained en tirely within peninsular Florida, and its occurrence is restricted to Florida scrub hab itat in the central ridge s of Florida and along the east and southwest coasts of Flor ida (Jackson 1973). The limited geographic
3 distribution and habi tat specificity of S. woodi contribute to its t ype 2 rarity (McCoy and Mushinsky 1992). Sceloporus woodi is characterized by rapid matu rity, low survival rates, and variable populat ion densities. Sceloporus woodi is sexually mature at 45 mm snout-vent length at an age of 7-8 months (Hartmann 1993, McCoy et al. 2004). The maximum recorded lifespan of S. woodi is 27 months, but the averag e lifespan is only 12.6 months (McCoy et al. 2004). Mating begins in Februa ry and oviposition continues into October (Jackson and Telford 1974). Mature females can produce three clutches per year, but five clutches may be possible in favorable year s (Jackson and Telford 1974). Hatchlings and juveniles compose the larges t proportion of a population at all times, but the large proportion of young individuals is especially pronounced in the summer months (Hartmann 1993, McCoy et al. 2004). Sceloporus woodi densities are highly variable, ranging from 10.1 individuals/ha (Jackson and Telford 1974) to 124 individuals/ha (Hartmann 1993, McCoy et al. 2004). Densities are greatest in J une with the first emergence of hatchlings, and decline until the following June (Hartmann 1993, McCoy et al. 2004). Survival rates of S. woodi are low: only ten percent of juveniles survive to the end of their first breeding season (McCoy et al. 2004). Mortality of hatchlings and juveniles is greatest during the summer (H artmann 1993). Abundance, survivorship, and recruitment of S. woodi are all positively rela ted to patch size (Hokit and Branch 2003b). Increased predation may be responsible for obs erved temporal and spatial differences in survival rates of S. woodi (Hartmann 1993, Hokit and Branch 2003b, a, McCoy et al. 2004).
4 Although S. woodi can achieve rapid reproduction, it is very limited in its movements. Home ranges of S. woodi are small, measuring approximately 800 m2 for males and 400 m2 for females (Hokit et al. 1999). Dispersal of S. woodi is impeded beyond 200 m, and the maximum dispersal distance is estimated to be 750 m (Hokit et al. 1999). Because of the limite d dispersal abilities of S. woodi and the naturally fragmented nature of Florida scrub, S. woodi exists as metapopulations. Both incidence function and stage-based models of local dynamics with a dispersal function accu rately predict the occurrence of S. woodi (Hokit et al. 2001). Occurrence of S. woodi in patches of Florida scrub is positively related to percent bare sand and patch size, the latter likely because of predation (Hokit et al. 1999). Occurrence is ne gatively related to patch isolation, and may be further limited where the habitat matrix consists of dense vegetation (Hokit et al. 1999). The limited dispersal abilities of S. woodi make it particularly susceptible to anthropogenic habitat fragmentation or alte ration (Fahrig and Merriam 1994), and the exclusion of fire not only elim inates early-successional Florida scrub habitats that are preferred by S. woodi (Tiebout and Anderson 1997, 2001), but may impede interpatch dispersal because of increased vegetation de nsity in the matrix surrounding patches of Florida scrub. In contrast to the extreme habitat specificity of S. woodi two of its potential predators, Coluber constrictor (Eastern Racer) and Masticophis flagellum (Coachwhip), occur in diverse habitats. Coluber constrictor is widespread in the continental United States (Ernst and Barbour 1989). In Florida, C. constrictor is ubiquitous and can be found in nearly every terrestrial and semi-aquatic habitat type (Carr 1940, Wright and Wright 1957, Ashton and Ashton 1981, Tennant 1997). Masticophis flagellum occurs throughout
5 the southern United States and northern Mexic o, and selects xeric habitats in the eastern United States (Dodd and Barichivich 2007, Johnson et al. 2007). In Florida, M. flagellum is most common in scrub and high pine habitats, but it is also frequently observed in pine flatwoods, dry prairies, and south Florida ro cklands (Carr 1940, Wri ght and Wright 1957, Wilson 1970, Tennant 1997). The use of diverse habitats by these two snake species is matched by their varied diets. Both snake species feed upon lizards, sn akes, turtles, birds, bird eggs, rodents, shrews, insects, and frogs (Hamilton and Pollack 1956, Klimstra 1959, Ernst and Barbour 1989, Tennant 1997, Conant and Collins 1998). Stomach contents of C. constrictor in Georgia contained 65% lizards 28% snakes 9% amphibians, 4% mammals, and 2% insects (Hamilton and Pollack 1956). The mo st abundant lizard in the diet of C. constrictor was Scincella lateralis (Ground Skink). Stomach contents of M. flagellum in Georgia contained 69% lizards, 18% mammals, 9% snakes, 9% insects, 2% birds, and 2% turtles (Hamilton and Pollack 1956). The mo st abundant lizard in the diet of M. flagellum was Aspidoscelis sexlineata (Six-lined Racerunner). The diet of each snake species consists of different proportions of prey in different regions, so both snake species appear to be opportunistic foragers (Klimstra 1959, Ernst and Barbour 1989, Secor and Nagy 1994). Both M. flagellum and C. constrictor are active foragers with very large home ranges compared to most snakes (Macartney et al. 1988). Minimum convex polygon home range size of C. constrictor in South Carolina was 12.2 (.86) ha; these snakes moved a distance of 104 () m/day, excluding days in which no movement occurred (Plummer and Congdon 1994). Minimum c onvex polygon home rage size of M.
6 flagellum was 57.9 (.2), 70.4 (.8), and 113.6 (.5) ha in the Mojave Desert, eastern Texas, and north-central Florida, respectively (Secor 1995, Dodd and Barichivich 2007, Johnson et al. 2007). Mean daily movement distance for M. flagellum was 146 (), 93 (standard error not available), and 229 m in the Mojave De sert, eastern Texas, and north-central Florida, respectively (Secor 1995, Dodd and Barichivich 2007, Johnson et al. 2007). Both C. constrictor and M. flagellum move farther and more frequently, and have larger home ranges than most other co-occurring snake species (Fitch and Shirer 1971, Secor 1995), with the possible exception of Drymarchon couperi (Eastern Indigo Snake (Dodd and Barichivich 2007)). The long-distance and frequent movements of C. constrictor and M. flagellum are associated with the active foraging mode of these snakes (Ashton and Ashton 1981, Secor 1995). A high rate of prey capture offsets this energetically costly foraging mode (Ruben 1977, Secor and Nagy 1994). Despite their frequent foraging and the high proportion of lizards in the diets of C. constrictor and M. flagellum the impact of these predat ors on lizard populations is unknown. Snakes have demonstrated potential to negatively impact prey populations (Savidge 1987, Rodda and Fritts 1992). Evidence for the impact of C. constrictor and M. flagellum on prey populations, however, is largely anecdotal. Predation by M. flagellum was suggested as the cause of decreased survival of male Uta stansburiana (Sideblotched Lizard) on high-quality territor ies (Calsbeek and Sinervo 2002). Predation by C. constrictor and M. flagellum was also suggested as the caus e of a decline in abundance of a population of Sceloporus undulatus (Eastern Fence Lizard (Crenshaw 1955)). Although snakes were not directly implicated, pred ation was suggested as a mechanism causing density-dependent mortal ity of a population of S. undulatus in Kansas (Ferguson et al.
7 1980). Coluber constrictor and M. flagellum were also suggested as a cause of densitydependent mortality of an isolated population of S. woodi (McCoy et al. 2004). Thus, predation by these snake species has frequently been implicated as a cause for decline or regulation of lizard populations. Th e relatively great vagility of C. constrictor and M. flagellum coupled with th e occurrence of S. woodi as metapopulations therefore presents a unique opportunity to examine predato r-prey interactions in patchy habitats. STUDY QUESTIONS AND HYPOTHESES The primary goal of my study was to char acterize the predator -prey relationships of C. constrictor and M. flagellum with S. woodi These predator-prey interactions are of theoretical interest because few studies ha ve documented the effects of wide-ranging predators upon patchy prey populations under natural field conditions. In this regard the snake: S. woodi system is of great theoretical in terest. In addition to theoretical considerations, Florida scrub is a ra re habitat and the only habitat of S. woodi Thus, examining the regulation of S. woodi populations at both patch and landscape scales can inform conservation practice. In partic ular, this study will examine the following questions: 1. Do C. constrictor and M. flagellum forage selectively upon certain prey, or do they forage opportunistically, consuming prey in proportion to its availability? 2. How does prey abundance influence habitat and Florida scrub patch selection by M. flagellum ?
8 3. What effect does the abunda nce of saurophagous snakes have upon the survival rate of local populations of S. woodi ? Is this effect stronger than the effect of patch area? 4. What implications do the answer s to Questions 1-3 have for the long-term persistence of the M. flagellum : S. woodi system? I predict that S. woodi will be consumed by both C. constrictor and M. flagellum Attempted predation upon S. woodi by C. constrictor (McCoy et al. 2004), and the documented high relative proportions of lizar ds in the diet of both snake species (Hamilton and Pollack 1956) indicate that S. woodi is likely to be consumed by both predators. Because both snake species have been documented to consume a variety of prey, however, I predict that they will forage opportunistically, consuming vertebrate prey in proportion to its availability. Because M. flagellum is an active forager with a re latively high metabolic rate and energy demand (Ruben 1977, Secor and Nagy 1994), I predict that the spatial ecology of M. flagellum will be strongly affected by prey density. Previous research has shown that M. flagellum select xeric habitats (Dodd a nd Barichivich 2007, Johnson et al. 2007), and I predict that M. flagellum will positively select Florida scrub habitat at LWRSF. If M. flagellum forages opportunistically as posited above, I predict that individuals will positively select Florida sc rub patches with the greatest abundance of total prey. Foraging behavior in these patche s will result in shorter daily movements in Florida scrub than in other habitats.
9 Predation has been repeatedly suggested as a mechanism regulating lizard populations (Ferguson et al. 1980, Rodda and Fr itts 1992). In particular, predation by snakes has been implicated as a mechanism causing mortality of Sceloporus species (Crenshaw 1955, Calsbeek and Sinervo 2002, McC oy et al. 2004). I therefore predict that S. woodi survival rates will be lower in patches of Florida scrub that contain a greater abundance of saurophagous snakes, and that th e negative effect of predator abundance on survival rates will be greater than the positiv e effect of area per se (Hokit and Branch 2003b). The persistence of the M. flagellum : S. woodi system will depend upon several factors. Generalist predators can potentia lly have a stronger negative impact on prey populations than specialist predators, particul arly if predation upon th e focal prey species is incidental (Vickery et al. 1992, Swihar t et al. 2001, Ryall and Fahrig 2006). The functional response of M. flagellum will have a large effect on the stability of the predator-prey interaction. If M. flagellum switches between prey species, ignoring rare prey and nearly always consuming abundant prey, the interaction will be stabilized because of reduced risk of predation for S. woodi populations occurring at low densities (Murdoch 1969, Oaten and Murdoch 1975, Mc Nair 1980, van Baalen et al. 2001). Suboptimal switching between prey species will further stabilize the interaction (Fryxell and Lundberg 1994). If M. flagellum does not form a search image for specific prey, no refuge will exist for S. woodi populations occurring at low dens ities (Swihart et al. 2001). A stable equilibrium is still possible at high S. woodi densities, however (Turchin 2003). Spatial heterogeneity in the abundance of prey has long been recognized as a potentially stabilizing characteristic of pred ator-prey interactions because predators will
10 leave a patch when it is no longer profitable fo r them to forage in that patch (Charnov 1976, Murdoch 1977). Simulation studies adopti ng the marginal value theorem (Charnov 1976) as a leaving rule for predators demonstr ate that the most realistic dispersal rule, adaptive local dispersal, enhances the persis tence of predator-prey systems (Fryxell and Lundberg 1993). The movement rate of predators is also influential for the stability of predator-prey interactions (Jansen 2001). Asynchrony am ong patches stabilizes predatorprey interactions, and is great est at intermediate predator movement rates and when the landscape contains a large numbe r of patches (Jansen 2001). T hus, the dispersal rates of M. flagellum as well as the dispersal rates of S. woodi may be important for the persistence of S. woodi in the landscape. Although some models have found that prey density-dependent predation can reduce the stabilizing effect of spatial heterogeneity (Huang and Diekmann 2001), the existence of S. woodi in discrete patches is likely to have an overall stabilizing effect on the interaction. Because of the continued protection and ma nagement of Florida scrub at LWRSF, I predict that the M. flagellum : S. woodi system there will persist well into the future. Pine flatwoods and scrub habitats at the site are frequently burne d using helicopters, and this practice results in a mosaic of burned and unburned patche s that mimics historic lightning-caused fires more clos ely than other prescribed bur ning techniques. The effect of burning is to maintain sc rub conditions appropriate for S. woodi (Tiebout and Anderson 2001) and to maintain the pine fl atwoods vegetation at a lower density wellsuited to interpatch dispersal of S. woodi (Greenberg et al. 1 994, Tiebout and Anderson 1997). Although I havent cons idered other trophic levels when examining my predictions, top-down control of M. flagellum by visually-oriented predators, such as
11 raptors, could prevent the numeric response of M. flagellum to the abundance of alternative prey (Rosenheim and Corbett 2003 ), which in turn would prevent potential extirpations caused by incidental predation (Vickery et al. 1992, Swihart et al. 2001). The existence of a relatively large number of Fl orida scrub patches at LWRSF should also promote persistence of the M. flagellum : S. woodi system (Jansen 2001). Although extirpations of S. woodi populations in small patches of Florida scrub at LWRSF will likely occur, and may even be caused by pred ation, current habitat management practices at the site will likely promote S. woodi dispersal and the persiste nce of both predator and prey in this landscape.
12 Sympatric Masticophis flagellum and Coluber constrictor Select Vertebrate Prey at Different Phylogenetic Levels Among the most important decisions pred ators must make is which prey to consume. This decision may be heavily infl uenced by prey defenses/vulnerability, prey abundance, predator choice, and/or predat or foraging history (Downes 2002, Desfilis et al. 2003, Williams et al. 2003, Greenbaum 2004, A ndheria et al. 2007). Studies of diet selection can provide informa tion about how animals meet their energetic requirements for survival and how they coexist with other species (Manly et al. 2002). In addition to these descriptive uses, studies of diet sel ection also can be incorporated into more detailed analyses to parameterize foraging m odels or predict the effects of changes in food types, availability, or preferences on populations (Sherr at and Macdougall 1995, Manly et al. 2002, Joly and Patterson 2003). Diet selection studies differ from studi es of diet composition by quantifying and comparing use and availability of prey, rath er than merely describing the proportions of prey consumed. The secretive nature, difficu lty in capturing, and foraging habits of snakes present particular difficulties for studi es of diet selection. Because many snakes consume large prey relatively infrequently, a large portion of captured snakes often has empty stomachs (Greene 1983, Mushinsky 1987, Miller and Mushinsky 1990, Cundall and Greene 2000, Greene and Rodriguez-Robl es 2003, Gregory and Isaac 2004). Those individuals that do contain stomach conten ts often contain only one or few items,
13 reducing the power of statistical comparisons of the distribution of available prey to consumed prey. In addition to the frequent occurrence of small sample sizes, snakes swallow prey whole and snake gape increa ses with body size (Miller and Mushinsky 1990, Cundall and Greene 2000, King 2002, Vincent et al. 2006a, Vincent et al. 2006b). Therefore, prey items available to large snak es are often not available to small snakes because of physical limitations (Shine 1991, Do wnes 2002). Prey availa bility is therefore best defined separately for each individual, rather than estimated for the entire population. Masticophis flagellum (Coachwhip) and Coluber constrictor (Eastern Racer) are both actively foraging predators (Ruben 1977, Ernst and Barbour 1989, Secor 1995), and C. constrictor appears to consume prey items oppor tunistically (in proportion to their availability (Ernst and Barbour 1989)). Bo th species detect prey visually and by vomerolfaction, actively pursue prey, and us ually consume it live or subdue it by blunt trauma (Fitch 1963, Jones and Whitford 1989, Secor 1995, Cundall and Greene 2000). Prey of M. flagellum includes insects, lizards, snakes, turtles, mammals, and birds (Hamilton and Pollack 1956, Ernst and Barbour 1989). The diet of C. constrictor is similarly broad, consisting of insects and ot her invertebrates, anurans, salamanders, lizards, snakes, turtles, mammals, and bi rds (Hamilton and Pollack 1956, Klimstra 1959, Fitch 1963, Ernst and Barbour 1989, Shewchuk a nd Austin 2001). Proportions of types of prey consumed vary by study across the range of C. constrictor (Hamilton and Pollack 1956, Klimstra 1959, Fitch 1963, Ernst and Barbour 1989, Shewchuk and Austin 2001); no comparable data are available for M. flagellum The geographic variation in the diet of C. constrictor could be caused by regional or temporal differences in prey availability,
14 variation in snake size, differe nces in individual or popula tion prey preference, habitat type or any combination of the above fact ors. These potential m echanisms underlying patterns of diet composition can only be resolved by concurrently examining diet composition and the availability of prey. My goals in this study were to quantify the diets, and examine prey selection and prey size:snake body size relationships, in sympatric M. flagellum and C. constrictor in central Florida. I hypothesized that these actively foragi ng predators would consume smaller prey relative to thei r body size than many snakes because they forage frequently (Secor 1995) and usually consume their prey while it is alive or rely upon blunt trauma to subdue it (Cundall and Greene 2000), thus potentially restricting the diet of these species to small or innocuous prey. Based upon the great variety of prey reportedly consumed by both species and the documented variation in the diet of C. constrictor in previous studies, I also hypothesized that the diet of each species would reflect the availability of their prey; thus I predicted that neith er species would forage selectively. MATERIALS AND METHODS Study Site I conducted this research at the Lake Arbuckle tract of the Lake Wales Ridge State Forest in southeastern Polk County, Florida, USA (27.67N Latitude, 82.43W Longitude) from March 2004 to June 2006. The s ite consists of a series of isolated wetlands and patches of xeric Florida scrub ha bitat in a matrix of mesic pine flatwoods
15 habitat. All sampling occurred in nine neighbor ing patches of xeric Florida scrub habitat ranging in size from 1.5 to 170 ha. Field Methods I sampled snakes and their potential vert ebrate prey using a total of 13 drift fence/pitfall/funnel trap arrays installed in the nine scrub patches, with sampling intensity related to patch size. I instal led one trap array in each of seven patches (1.5 12.6 ha), two trap arrays in one patch ( 40 ha), and four trap arrays in the remaining patch (170 ha). I did not sample potential invert ebrate prey. I opened trap arrays and checked them daily for seven consecutive days, then closed them for 21 consecutive days as part of a robust mark-recapture design (Pollock 1982). I closed the traps from November-February because of unacceptably high mortality rates of trapped small mammals caused by cold overnight temperatures. I measured all captu red vertebrates for s nout-vent length (SVL; mm; reptiles and amphibians), total length ( TL; mm; reptiles), and/or mass (g; reptiles and mammals). I also measured the maximu m circumference of lizards and mammals captured in 2006 by wrapping a marked string around lizards at the point of maximum girth and by releasing small mammals thr ough the smallest-diameter 10-cm length of PVC pipe through which the individual coul d pass. I uniquely marked each captured reptile and mammal using toe clips for lizards PIT tags or subcaudal scale clips for snakes, and Monel numbered ear tags for small mammals. I did not mark amphibians individually, but clipped a single toe to iden tify them as recaptures. I palpated each captured M. flagellum and C. constrictor to determine if the individual contained relatively undigested prey. I forced individuals that contained prey to regurgitate for
identification and measurement of stomach c ontents. I did not force gravid females to regurgitate, but noted if they contained palpable stomach contents. I identified regurgitated prey to the species level and fixed them in formalin for subsequent measurement (SVL, TL, wet mass, volume, and maximum diameter). Because head dimensions are the best predic tor of prey size in snakes (Rodriguez-Robles et al. 1999, Cundall and Greene 2000, Vincent et al. 2006b), I measured jaw length and width (Miller and Mushinsky 1990) for all individuals of each snake species captured in 2006. Analytical Methods I described snake diets as the percentage of individuals containing each prey category or prey species. I examined the si ze and sex of snakes of each species to determine if snakes containing stomach cont ents were representa tive of all trapped individuals within each species. I compared the distribution of SVL of individuals containing prey to the distribu tion of SVL of all trapped individuals for each species with a Kolmogorov-Smirnov two-sample test. Similarly, I compared the sex ratio of individuals containing prey to the sex ratio of all trapped i ndividuals separately for each snake species with a G-test of independence. I also used a G-test of independence to test for a difference in the proportion of snakes containing stomach contents between the two species. Because snakes swallow their pr ey whole and are gape-limited predators (Cundall and Greene 2000, Vincent et al. 2006 a), I approximated gape size from jaw length and jaw width measurements as the circ umference of an ellipse using Ramanujans approximation )3)(3()(3 bababaC 16
17 where a = jaw length (half of the major axis of the ellipse defined by the individuals open mouth) and b = jaw width (half of the minor axis of the ellipse defined by the individuals open mouth (M iller and Mushinsky 1990)). I found SVL to be a good predictor of gape size ( M. flagellum : n = 18, Gape = 0.126*SVL + 26.9, Adjusted R2 = 0.93, F(1,16) = 234, P < 0.001; C. constrictor : n = 50, Gape = 0.146*SVL + 24.2, Adjusted R2 = 0.88, F(1,48) = 372, P < 0.001). Therefore, I examined the relationship between snake SVL and prey category found in the stomach graphically. I examined the relationship between snake mass and the mean mass of prey the individual consumed (calculated as the total mass of prey consumed divided by th e number of individual prey in the stomach (Arnold 1993)) with Spearman rank correlati on. I further examined predator-prey size relationships by calculating rela tive prey mass, defined as the mass of the average prey item divided by the individuals mass without stomach contents. I compared the mean relative prey mass of the two species with the MannWhitney-Wilcoxon test, and the distribution of relative prey mass of th e two species with the Kolmogorov-Smirnov twosample test. I examined the relationship between snake SVL and relative prey mass using Spearman rank correlation. I determined diet selection of each snak e species at two taxonomic levels: prey category (amphibian, lizard, sn ake, turtle, bird, and mamma l) and prey species. These taxonomic scales are subdivisions of fourth -order (diet) selection (Johnson 1980), and I used these two phyletic levels of selecti on because predators may discriminate among potential prey by a variety of sensory mode s and mechanisms that can act across broad taxonomic categories or at the le vel of individual prey specie s, or even individual prey items (Ford and Burghardt 1993, Greenbaum 2004, Shine et al. 2004). I defined used
18 prey as the contents of the individuals stomach. I defined unus ed prey (prey available to the individual, but not consumed) separately fo r each individual as the count of each prey category or species captured in the same tr ap array during the same one-week sampling period as the individual snak e was captured. The sum of us ed plus unused prey defined availability for each individual. Calculating av ailability as the sum of used plus unused prey avoids the potential bias caused by snakes consuming prey with in the trap, because consumed prey are considered available to the snake. If prey were consumed by the snake after it had been trapped, and they were not considered available, estimates of availability for consumed prey would be lower, causing an upward bias in the apparent selection of the consumed prey category or species. Calc ulating availability as used plus unused prey also ensures that prey consumed by the snake were considered available to the snake. I considered prey too large for an individua l to consume based upon snake gape and prey circumference measurements (Miller and Mushinsky, 1990) unavailable, and excluded individual prey species from the available poo l if the prey species mean circumference exceeded predicted snake gape (based upon linear regression of gape with SVL). Although circumference was not measured for amphibians, I exclude d large individuals of Bufo terrestris (Southern Toad) and Rana species from the pool of prey available to each snake if their SVL exceeded predicted snake gape. Defining prey availability separately for each individual minimizes the pr oblems of spatial and temporal variability in availability by restricting the definition of available prey to the area of influence of a single trap array and a one-w eek period in which the snake was found within this area (Manly et al. 2002).
19 I determined selection of prey from t hose available using Manlys standardized selection ratio ( Bi (Manly et al. 2002)) with bootstrapped confidence intervals. I defined positive selection as prey consumed in greater proportion than their availability in the environment, and negative selection as prey consumed in lesser proportion than their availability in the environment. I calculated the selection ratio ( ij) for each of the i prey categories or species available to the j th snake (Manly et al. 2002). To establish confidence intervals, I randomly selected the number of prey that an individual snake consumed from the distribution of av ailable prey 1,000 times, and calculated i for each trial. To examine prey selecti on at the population level, I used Bi (which is the i standardized to sum to one) because it is interpretable as the probability of selection of the i th prey if all prey were equally available in the environment (Manly et al. 2002). To determine availability at the population level for each snake species, I calculated the sum of the bootstrapped random selection probabilit ies across individuals of each species. This sum is equivalent to the multinomial e xpectation of the number of prey of each species that would be selected if prey were consumed in pr oportion to their availability. I compared this available prey distribution to the distribution of used prey with Pearsons 2 statistic. I determined the statistica l significance of the observed Pearsons 2 statistic by randomly selecting the number of prey consumed by each species from the available distribution 1,000 times, and calculating Pearsons 2 statistic for each iteration. This procedure avoided the distributiona l assumptions of the Pearsons 2 statistic, making it unnecessary to pool prey species where low expected values occurred. I determined statistical significance of the Bi and pairwise differences in the selection of prey categories or species from the above bootstrapped samples.
20 Because prey likely have different capt ure probabilities in drift fence arrays, I used closed population models to reduce sampli ng bias and estimate prey availability at the species level for M. flagellum I could not estimate a bundance of amphibians to determine prey availability for C. constrictor by this method because amphibians were not individually marked. I estimated abundance of each prey species for each of nine individual M. flagellum at its time and location of capture using Huggins closed models in MARK 4.3 (White 2006). I assumed daily captu re probabilities to be the same at all trap arrays, which were identical, placed in scrub habitat with very similar structural attributes, and close enough to one anothe r that each array experienced similar environmental conditions. I also considered capture and recapture probabilities equal for two a priori reasons. First, traps were not baited, so prey were unlikely to exhibit a traphappy response. Second, trapped prey did not appear stressed relative to prey found outside the traps, and it is unlikely that they learned to intentionally avoid traps. Huggins closed population models that I evaluated were constant (re)cap ture probabilities, individual heterogeneity in (re)capture probabilities (two mixture model), time-specific (re)capture probabilities, and combined time and heterogeneity effects. In addition to these basic models, I used daily mean te mperature and rainfall as environmental covariates; interactions of these environmental variables with heterogeneity also were evaluated. I also evaluated models incorpor ating individual cova riates to account for heterogeneity in capture probabilities caused by sex (all prey species), SVL (lizards), and mass (small mammals) to improve the estimation efficiency of abundance (Chao and Huggins 2005). As with the environmental cova riates and heterogeneity, I examined twoway interactive and additive models between individual covariates and time. Thus, I
21 considered up to 18 models for each prey spec ies available to each snake. I analyzed all models with covariates usi ng the logit link function, which maps the probability of a binary response variable (in this case, captured or not capture d) from [0,1] to [,+ ] (Quinn and Keough 2002). If no rain fell duri ng a secondary period, I eliminated all models that included rainfall from the analys is. Additionally, I did not consider models that contained more parameters than the number of individual prey of each species captured during that week because these data we re too sparse to estimate the parameters of more complex models. Thus, only a single mo del with constant (re)capture probability could be evaluated for some prey species during certain sampling periods. Because of sparse recapture data, I could not estimate abundance of all pr ey species available to the eleven remaining M. flagellum individuals; therefore, I excl uded these individuals from this analysis. Studies of diet selection ra rely take into account uncertainty associated with the determination of prey availability. I ther efore developed a procedure to account for uncertainty inherent in estimating the abunda nce of cryptic, mobile prey. Because my data were nearly always supported by more than one of the m odels, I used model averaged abundance estimates to account for uncertainty in model selection (Burnham and Anderson 2002). For each snake, I accounted for overall uncertainty in estimates of prey abundance (that arising from both m odel selection uncertain ty and variance in abundance estimates for each model) with a two-step resampling procedure in which 1,000 abundance estimates for each prey species were randomly and independently sampled from each prey species distributi on of abundance estimates. Model averaging resulted in a normal distribution of abundance estimates that often included estimates less
22 than the count of captured i ndividuals. To account for the unrealistic nature of abundance estimates less than the count of individuals captured, I di scarded the randomly selected estimate if it was less than the count, and th e random selection of an abundance estimate was repeated. If the selected estimate was less th an the count after three iterations of this process, I used the count as an abundance estimate for that bo otstrap sample. I calculated the observed i for each of these 1,000 estimates of availability to generate a distribution of observed i values that reflected the uncertainty inherent in determining the relative abundances of prey species. To ge nerate a confiden ce interval for i under the hypothesis of random selection of prey, I selected the number of prey individuals consumed by the snake from each of these 1,000 availability estimates 100 times, yielding 100,000 samples for each individual. I analyzed prey selection at the population level as above using Bi, and tested for statistical significance using Pearsons 2 with bootstrapped confidence intervals. I calculated niche breadth and nich e overlap to compare the foraging characteristics of the two snakes and place th ese foraging patterns in a broader ecological context. I determined niche breadth for each snake species using Hurlberts B index, which describes the breadth of the diet relative to the availability of prey in the environment (Hurlbert 1978). I used Morisitas C index (Horn 1966) to examine niche overlap. This index is interpre table as the probability that two prey drawn randomly from each predator population will be the same species, standardized to account for the diversity of each predators diet (Horn 1966). I established confidence intervals for both indices by bootstrapping (1,000 random samples). Statistical analyses were conducted using the programs MARK 4.3 (White 2006), Resampling Stats 5.0.2 (Resampling Stats,
23 Inc., 1999), and R 2.3.1 (R Core Developmen t Team 2006). All means are reported as mean (standard error). RESULTS Diet Composition Snakes that contained stomach contents we re representative of trapped individuals for each species. Eighty-one individual M. flagellum were examined for stomach contents one-hundred thirteen times, and 268 C. constrictor were examined for stomach contents three hundred times. No sexual size dimorphism in SVL was detected for either species ( M. flagellum : males = 831 61 mm [330 1650 mm], females = 806 53 mm [375 1520 mm]; W = 840, P = 0.73; C. constrictor : males = 667 15 mm [300 1030 mm], females = 671 15 mm [240 1065 mm]; W = 8800, P = 0.82), so sexes were pooled within species for analysis. The size distribution of individuals containing stomach contents was not significantly different from the size distribution of all trapped individuals for each species ( M. flagellum : D = 0.20, P = 0.51; C. constrictor : D = 0.11, P = 0.61). Likewise, the sex ratio of individuals containing stomach contents was not significantly different from the sex ratio of all trapped snakes of each species ( M. flagellum : G = 0.01, df = 1, P = 0.92; C. constrictor: G = 0.51 df = 1, P = 0.48). Twentyone (18.6%) M. flagellum and 59 (28.2%) C. constrictor contained prey. Of these, one female M. flagellum was not forced to regurgitate, one C. constrictor contained unidentified organic matter, two C. constrictor could not be forced to regurgitate, and three gravid female C. constrictor were not forced to re gurgitate. Three individual C.
24 constrictor contained prey on more than one occas ion. The proportion of individuals of each species that contained prey was not significantly different (G = 0.53, df = 1, P = 0.47). Both predator species consumed a variet y of prey. At the prey category level, M. flagellum most frequently consumed lizards, and C. constrictor consumed amphibians and lizards in nearly equal pr oportions (Fig. 1). The most fr equently consumed lizard in the diet of M. flagellum was Sceloporus woodi (Florida Scrub Lizard), and the most frequently consumed mammal was Podomys floridanus (Florida Mouse; Fig. 2). The most frequently consumed amphibian, lizard, and mammal in the diet of C. constrictor were Hyla femoralis (Pine Woods Treefrog), Aspidoscelis sexlineata (Six-lined Racerunner) and Peromyscus polionotus (Oldfield Mouse), respectively (Fig. 2). One C. constrictor contained anuran legs that could not be identified to species. Individuals of both snake species consumed relatively few pr ey, with a maximum of three prey in one individual M. flagellum and four prey in one individual C. constrictor. Both species had a mode of one prey item. The di stribution of the number of prey items per stomach did not differ between the two speci es (Fishers exact test, P = 0.80). Both M. flagellum and C. constrictor consumed prey that were small relative to their body size. Larger individuals of both sp ecies continued to consume the same prey categories as small snakes, but individuals of both species did not consume mammals until they reached a minimum size of 725 mm SV L (Fig. 3). Mean relative prey mass of M. flagellum was 0.098 (.032), with a range of 0.008 0.27, and mean relative prey mass of C. constrictor was 0.068 (.014), with a range of 0.003 0.359. Mean relative prey mass did not differ between the two snake species (W = 180, P = 0.27), nor did the
25 distribution of relative prey ma ss differ between them (D = 0.38, P = 0.25). Masticophis flagellum did not exhibit a signifi cant relationship of snake mass with prey mass ( = 0.36, P = 0.39), but C. constrictor did ( = 0.42, P = 0.01; Fig. 4). No significant relationship existed between predator SVL a nd relative prey mass for either species ( M. flagellum : = -0.36, P = 0.39; C. constrictor : = 0.077, P = 0.66). Diet Selection A variety of vertebrate prey species we re consumed by and available to individual M. flagellum and C. constrictor (Tables 1 and 2). Arthropods were present on the site, but their abundance was not quantifie d. At the population level, patte rns of prey selection for M. flagellum and C. constrictor differed. At the prey category level, M. flagellum approached selective predation ( n = 20, 2 = 9.0, P = 0.06), and positively selected lizards, consumed mammals and snakes in pr oportion to their ava ilability, and avoided amphibians (Table 3). Masticophis flagellum selected lizards signi ficantly more than amphibians and snakes, and selected mamm als significantly more than amphibians (Table 3). Coluber constrictor was not selective at the prey category level ( n = 56, 2 = 2.8, P = 0.45), and consumed prey in proportion to their availability (Table 3). At the prey species level, M. flagellum selected prey in proportion to their availability regardless of whether counts ( n = 20, 2 = 3.1, P = 0.43) or abundances ( n = 9, 2 = 0.74, P = 0.97) were used as estimates of availability, but B -values for some prey species changed considerably depending upon which method was used to define available resources (Table 4). Coluber constrictor was selective at the prey species level ( n = 55, 2 = 90, P < 0.01), and positively selected Hyla femoralis while negatively selecting Bufo quercicus
26 (Oak Toad), B. terrestris and Gastrophryne carolinensis (Eastern Narrowmouth Toad; Table 5). In pairwise comparisons, C. constrictor selected H. femoralis significantly more than all other species except Acris gryllus (Southern Cricket Frog), Rana capito (Gopher Frog), Podomys floridanus, and Peromyscus polionotus, and selected Anolis carolinensis (Green Anole) signifi cantly more than the B. quercicus. Niche Breadth and Overlap The relationship between the niche bread ths of the two snake species varied depending upon the scale at which it was ex amined. At the level of prey category, M. flagellum had a narrower niche than C. constrictor, but this difference was not statistically significant (Table 6). At the prey species le vel, the opposite pattern was observed, with M. flagellum exhibiting a significantly br oader diet in relation to its available prey than C. constrictor did (Table 6). Niche overlap between the two predator species was considerable at both levels of analysis (Table 6). DISCUSSION Masticophis flagellum and C. constrictor are selective foragers that consume relatively small prey. Both species consume a variety of prey, but each is selective at a different level of taxonomy. Masticophis flagellum positively selects lizards and mammals, and it consumes prey species within these categories in proportion to their availability. In contrast, C. constrictor forages opportunistically upon prey categories, but is selective at the species level, particularly among amphibian species. Thus, foraging by
27 these two snake species occurs by a hierarchi cal process whereby selection can occur at different levels of prey taxonomy. The observed differences in the foraging ha bits of these two species are somewhat surprising, given their usual characterization as generalist predators (Ernst and Barbour 1989). Dietary differences between these sp ecies cannot be explained by size alone, because individuals of similar sizes were sampled from both snake species and large individuals of both species consumed the same prey categories and species that smaller conspecifics did. Although amphibians have not been reported in the diet of M. flagellum they are available to M. flagellum Therefore, M. flagellum actively avoids consuming amphibians. In contrast, C. constrictor consumed amphibians more frequently in my study site than elsewhere (Hamilton and Pollack 1956, Klimstra 1959, Fitch 1963, Shewchuk and Austin 2001). Because I did not determine that prey were limiting to these snakes, I were unable assess competition between them. If lizard and mammal prey were limiting, competition with M. flagellum might explain the greater consumption of amphibians by C. constrictor in Florida scrub than elsewh ere. Alternatively, amphibians may be more abundant at my site than other lo cations, and their prominence in the diet of C. constrictor in Florida scrub may simply reflect relatively high amphibian abundance. Additional studies of prey limitation and the diet selection of th ese two snake species across their geographic ra nges are required to eval uate these hypotheses. Although M. flagellum was selective at the prey category level and C. constrictor was selective at the prey species level, both predator species consumed lizard and mammal species in proportion to their availa bility. Perhaps a general search image and acceptability of all lizards and mammals as vi able prey exists for both species, while C.
28 constrictor discriminates among anuran species. Prey discrimination could potentially occur at multiple stages of the predation event, including prey de tection, prey capture, prey manipulation, intraoral transport, or even swallowing (Cunda ll and Greene 2000). It is currently unknown at which of these stages C. constrictor rejects toxic anurans ( Bufo spp. and G. carolinensis (Garton and Mushinsky 1978, Daly et al. 1987)), or whether individuals learn to avoid certain species after expos ure to toxins. The genus Bufo has been documented in the diet of C. constrictor (Klimstra 1959, Fitch 1963); however, I found that C. constrictor avoided Bufo species. Interand intraspecific geographic variation in the toxicity of Bufo and/or geographic varia tion in the tolerance of bufodienolides or alkaloids by C. constrictor may occur (Brodie et al. 2002). Alternatively, C. constrictor may resort to foraging on avoided prey species if more preferred alternatives are unavailable. Again, detailed studies of prey toxicity, predator tolerance of toxins, and prey availability are necessary to evaluate these hypotheses. The consumption of insects, particularly Orthoptera, by C. constrictor is welldocumented (Klimstra 1959, Fitch 1963, Sh ewchuk and Austin 2001), but I found no insect prey in the stomachs of C. constrictor. One potential explanation for the absence of insects in the diet of C. constrictor in Florida scrub is the tendency for eastern C. constrictor to consume fewer insects than western conspecifics (Fitch 1963). The pattern of prey consumption by C. constrictor is likely more complicated than this apparent gradient of reduced consumption of insect pr ey from west to east suggests. Most studies of snake diets are conducted without quantifying the availabi lity of prey, which is an essential component of any co mparison of prey selection by free-ranging snakes. Florida scrub is a nutrient-poor habitat (Myers 1990) with a relativ ely low abundance of
29 arthropods, many of which app ear chemically well-defended against predators (Witz and Mushinsky 1989, Witz 1990). Because I did not quantify arthropod abundance, I cannot assess whether C. constrictor in Florida scrub avoids ava ilable arthropods or merely ignores them because of their low abundance. Pr ey availability is thus not only important for documenting geographic, temporal, and taxon omic differences in predator diets, but also to elucidate mechanisms leading to selective predati on within populations. As I predicted, M. flagellum and C. constrictor consume relatively small prey compared to other macrostomate (large-mouthed) snakes (Cundall and Greene 2000) Studies of other snake species have demons trated the capacity of snakes to consume relatively large prey (Mushinsky 1982, Pough and Groves 1983, Arnold 1993, Cundall and Greene 2000, Rodriguez-Robles 2002, Green e and Rodriguez-Robles 2003, Gregory and Isaac 2004). Many species consume prey larger than those ingested by either M. flagellum or C. constrictor. All species included in Table 3 of Rodrguez-Robles (2002) had a greater mean relative prey mass than both M. flagellum and C. constrictor and of these 13 species, only Boiga irregularis (Brown Treesnake) and Psammodynastes pulverulentus (Asian Mock Viper) had a lower maximum relative prey mass. Larger potential prey than those consumed by both snake species, such as Sylvilagus floridanus (Eastern Cottontail) and Sigmodon hispidus (Cotton Rat), were available in and adjacent to sampled patches of Florida scrub at my study site, but these prey were likely at or above the maximum prey size available to each species. A particularly large adult M. flagellum foraging in pine flatwoods habitat at my study site was observed to consume adult S. hispidus, which are too large or formidable for all but the largest M. flagellum (and all C. constrictor ) at my site to consume. Although prey mass increased with snake
30 mass for both species, longer snakes did not consume relatively heavier prey. A freeranging adult M. flagellum at my study site was observed to consume A. carolinensis indicating that large M. flagellum do not drop these small lizards from their diet. My stomach contents data support this observation: adults of both snake species continued to consume prey species and categories consumed by smaller juvenile conspecifics. Therefore, both species exhibit an ontogene tic telescope (Arnold 1993), rather than an ontogenetic shift in diet (Mushinsky 1982). Masticophis flagellum and C. constrictor are relatively abundant in Florida scrub habitat, and previous resear ch suggests that they may exert a strong influence on the dynamics of small, isolated populations of prey occurring in scrub fragments. In particular, S. woodi has lower survivorship in small pa tches of scrub; smaller habitat patches may suffer increased predation ra tes by snakes (Hokit and Branch 2003b). In addition to the positive relationship between patch size and survivor ship, McCoy et al. (2004) observed density-de pendent mortality of S. woodi associated with the presence of and observed predation by both M. flagellum and C. constrictor. My findings are consistent with these observations. Both pred ator species consume prey opportunistically at some level, and opportunistic predators lik ely exert a direct dens ity-dependent effect on prey (such as that observed by McCoy et al. ). Although these predators are unlikely to exhibit a numerical response (via increased reproductive rates) to any single prey species abundance, aggregative moveme nt to areas of high prey density by these wide-ranging snakes could exert a strong influence on local prey dynamics. If M. flagellum does not form a species-sp ecific search image, as my data suggest, it may have a hyperbolic functional response th at could lead to the extirpa tion of prey that occur at
31 low densities (Turchin 2003). Effects of predation by M. flagellum may be especially strong on S. woodi and P. floridanus which are both important dietary components of M. flagellum and precinctive to Florida scrub. Knowledge of the fo raging behavior and diet selection of predators may be important for the conservation of these rare prey species. In summary, I found that M. flagellum and C. constrictor are selective foragers that consume relatively small prey. As I hypothesized, both M. flagellum and C. constrictor consume small prey relative to other sn akes. In contrast to my hypotheses, both species were selective of prey at some level. Masticophis flagellum positively selected lizards and mammals, but consum ed species within these categories in proportion to their availability. Coluber constrictor consumed amphibians, lizards, and mammals in proportion to their availability, but within the amphibi an category positively selected H. femoralis and negatively selected the B. quercicus B. terrestris and G. carolinensis By defining availability separately fo r each individual snake, I was able to incorporate gape limitation and account for spatial and temporal variation in prey availability in my analyses of prey se lection. Mechanisms underlying geographic, temporal, and interspecific variation in pr edator diets can be better elucidated by examining prey availability and selec tion at the level of the individual.
32 Table 1. Prey Species Consumed by and Available to Individual Masticophis flagellum Abbreviations used are: ID = snake identification number, SVL = snake snout-vent length (in mm), SC (No.) = prey species found in stomach (number of prey of that species), Acar = Anolis carolinensis Asex = Aspidoscelis sexlineata Swoo = Sceloporus woodi Pflo = Podomys floridanus and Ppol = Peromyscus polionotus Abundance was estimated for each species using model-averaged parameters from Huggins closed models with capture probability held constant across trap arrays. Snake was included in diet selection analysis using abundance of prey as estimate of availability. Numbers in brackets indicate prey that were sa mpled in the same array during the same week as the snake, but were too large for the snake to consume based upon snake and prey measurements. Snake Prey Available (Count) Prey Available (Abundance [SE])* ID SVL SC (No.) Acar Asex Swoo Pflo Ppol Acar Asex Swoo Pflo Ppol 40F 878 Asex (1) 9 9 30.4 (17.6) 23.5 (9.9) E21 1650 Asex (1) 8 2 1 19.5 (8.7) 4.9 (3.0) 2.6 (2.7) 258 920 Swoo (1) 3 3  NA 7.1 (6.3) [2.6 (2.5)] 519 580 Swoo (1) 5  8.3 (4.3) [2.7 (3.5)] 7 475 Swoo (2) 7  296 (920) [6.2 (6.5)] 76F 1220 Pflo (1) 1 2 3.3 (3.0) NA 10 555 Acar (1) 1 6 NA 20.0 (9.7) 953 955 Acar (2) 5 1 9 NA 4.6 (4.7) 26.6 (12.0) Swoo (1) 534 665 Swoo (1) 7 20.0 (9.7) 867 600 Swoo (1) 1 2 2  1 NA 12.5 (11.7) 2.1 (1.7) [2.0 (1.5)] 3.4 (4.6) 11 585 Acar (1) 2 1 3  NA 7.2 (8.9) 6.4 (3.1) [6.7 (2.7)] 94A 1090 Pflo (1) 1 4 2.1 (1.7) 5.0 (2.5) 406 1070 Swoo (1) 1 1 12 4.8 (6.0) 2.2 (1.7) 17.3 (4.0) 15 400 Acar (1) 3 4 4 1 9.6 (10.4) 1 2.7 (6.5) 18.9 (13.5) NA D50 965 Swoo (1) 2 2 4  2 NA 14.3 (19.0) 6.8 (3.6) [2.9 (1.4)] 2.7 (1.1)
33 Table 1 (Continued). Snake Prey Available (Count) Prey Available (Abundance [SE])* ID SVL SC (No.) Acar Asex Swoo Pflo Ppol Acar Asex Swoo Pflo Ppol 29 665 Asex (1) 1 4  1 NA 8.3 (4.7) [10.0 (1.8)] 2.7 (2.6) 048 965 Swoo (1) 1 5  3 3.1 (3.0) 12.2 (9.7) [5.1 (1.3)] 7.5 (1.7) 143 1150 Asex (1) 1 1 2 5 2 NA NA 4.9 (3.1) 6.6 (1.6) 4.9 (3.9) A29 930 Ppol (1) 5 1 18.0 (9.7) NA 32 750 Swoo (1) 3  4.4 (2.7) [1.2 (0.6)]
34 Table 2. Prey Species Consumed by and Available to Individual Coluber constrictor. Abbreviations used are: ID = snake identification number, SVL = snake snout-vent length (in mm), SC (No.) = prey species found in stomach (number of prey of that species), Acar = Anolis carolinensis Asex = Aspidoscelis sexlineata Pine = Plestiodon inexpectatus Swoo = Sceloporus woodi Agry = Acris gryllus, Bque = Bufo quercicus, Bter = B. terrestris, Gcar = G. carolinensis Hfem = Hyla femoralis, Rcap = Rana capito Rspp = Rana catesbeiana + R. grylio Rutr = Rana utricularia Pflo = Podomys floridanus and Ppol = Peromyscus polionotus Numbers in brackets indicate prey that were sampled in the same array during the same week as the snake, but were too large for the individual to consume base d upon snake gape and prey measurements. Snake was not included in prey species-level analyses because stomach contents could not be identified to the species level. Snake Prey Available (Count) ID SVL SC (No.) Acar Asex Pine Swoo Agry Bque Bter Gcar Hfem Rcap Rspp Rutr Pflo Ppol 2 380 Swoo (1) 1 8 1 A 370 Swoo (1) 1 2 * C35 736 Swoo (2) 4 B11 742 Acar (1) 2 9 1 4 1 13 671 805 Rutr (1) 1 9 1 4 1 14 44F 985 Rutr (1) 1 9 1 4 1 14 274 910 Asex (1) 1 7 8 7 F66 655 Asex (1) 2 29 2 1 1 1 155 770 Swoo (1) 5 5 4 
35 Table 2 (Continued). Snake Prey Available (Count) ID SVL SC (No.) Acar Asex Pine Swoo Agry Bque Bter Gcar Hfem Rcap Rspp Rutr Pflo Ppol 650 660 Asex (1) 17 6 1 1 3 10 7 533 Rutr (1) 1 6 2 3 5 5 5 7 53F 720 Acar (2) 2 5 14 9 1 1 1 2  Asex (1) F0A 560 Rutr (1) 7 2 11 6 5 2 8 530 Asex (1) 8 2 11 6 5 1 F15 810 Hfem (1) 2 4 1 1 8 2  E22 870 Asex (1) 2 1 1 4 2 2 F2F 710 Hfem (1) 2 3 6 8 1 1 1 45D 960 Pflo (1) 3 1 2 1 1 2 15 395 Hfem (1) 7 3 2 3  1 1  20 400 Hfem (3) 2 8 1 2 3  2 21 400 Acar (1) 2 4 3  24 380 Hfem (1) 1 1 3 2 4 25 360 Hfem (1) 5 2 2 1 
36 Table 2 (Continued). Snake Prey Available (Count) ID SVL SC (No.) Acar Asex Pine Swoo Agry Bque Bter Gcar Hfem Rcap Rspp Rutr Pflo Ppol 13 470 Rutr (1) 1 1 3 1 5 16D 725 Rutr (1) 1 2 4  753 950 Ppol (1) 4 1 2 95F 810 Acar (1) 2 4 3 1 2  Pine (1) 123 870 Ppol (1) 5 1 1 5 2 4 27F 565 Swoo (2) 1 8 1 2 1 1 95F 810 Pine (1) 1 4 3 1 2 1  Hfem (1) 760 725 Ppol (1) 2 7 3 1 3 430 720 Acar (1) 1 2 7 3 1 2 837 670 Swoo (1) 1 1 4 2 2 5 5  82B 760 Rutr (1) 1 1 3 2 2 5 6  830 700 Swoo (1) 1 1 5 1 3  F4E 730 Hfem (1) 8 46 1 1 3 8 
37 Table 2 (Continued). Snake Prey Available (Count) ID SVL SC (No.) Acar Asex Pine Swoo Agry Bque Bter Gcar Hfem Rcap Rspp Rutr Pflo Ppol 39 835 Rutr (1) 1 3 9 5 4 1 3  1 727 720 Hfem (1) 2 1 2 1 5 2 1 1 1  30E 755 Rspp (1) 5 1 4 1 1 93A 780 Asex (1) 1 10 1 5 1 1 143 605 Asex (1) 4 1 3 7 1 1 1  D39 755 Hfem (1) 3 2 2 8 2  3 1 1  91A 790 Asex (1) 1 5 2 2 1  42 750 Acar (1) 2 6 1 3 Hfem (3) 46 320 Acar (1) 1 5 2 1 2 Hfem (1) 49 740 Rspp (1) 1 1 1 3 5 1  67 730 Rspp (1) 1 3 1 28 3  A0F 790 Asex (1) 1 4  1 20 770 Acar (1) 1 6 3 1 
38 Table 2 (Continued) Snake Prey Available (Count) ID SVL SC (No.) Acar Asex Pine Swoo Agry Bque Bter Gcar Hfem Rcap Rspp Rutr Pflo Ppol 97 745 Asex (1) 1 1 2  2 106 710 Swoo (1) 2 3 4 5 2 2 1 1  143 690 Asex (1) 1 9 1 6 2  B22 670 Rcap (1) 2 30 18 20 3 6 28 926 610 Hfem (1) 5 3 4 8  1 2 1 A06 895 Agry (1) 1 4 1 1 3 2 Prey Category Available (Count) Amphibian Lizard Snake Turtle Bird Mammal 63 755 Amp (1) 17 2 1 1
39 Table 3. Manlys Standardized Selection Ratios ( B ) for Prey Categories Selected by Masticophis flagellum and Coluber constrictor P -values were obtained from 1,000 bootstrap samples. Superscripts indicate significant pairwise differences in selection of prey categories. Masticophis flagellum ( n = 20) Coluber constrictor ( n = 56) Prey Category B P B P Amphibians 0.000c <0.01 0.244 0.65 Lizards 0.54a <0.01 0.301 0.34 Snakes 0.000bc 0.37 0.000 0.19 Turtles None Available 0.000 0.73 Birds None Available 0.000 0.83 Mammals 0.46ab 0.13 0.455 0.05
40 Table 4. Manlys Standardized Selection Ratios ( B ) for Prey Species Selected by Masticophis flagellum Estimates of prey availability for individual snakes were obtained by counts and by estimating prey ab undances using model-averaged Huggins closed population models. P -values were obtained from 1,0 00 bootstrap samples for counts, and 100,000 bootstrap samples for abundances. Counts ( n = 20) Abundances ( n = 9) Prey Species B P B P Anolis carolinensis 0.371 0.10 0.00 0.41 Aspidoscelis sexlineata 0.152 0.54 0.38 0.44 Sceloporus woodi 0.157 0.40 0.28 0.81 Podomys floridanus 0.183 0.54 0.35 0.64 Peromyscus polionotus 0.138 0.80 0.00 0.56
41 Table 5. Manlys Standardized Selection Ratios ( B ) for Prey Species Selected by Coluber constrictor P -values were obtained from 1,000 bootstrap samples. Superscripts indicate significant pairwise differences in selection of prey species. n = 55. Prey Species B P Hyla femoralisa 0.213 <0.01 Acris gryllusabc 0.142 0.50 Rana capitoabc 0.113 0.64 Rana catesbiana and Rana gryliobc 0.060 0.79 Rana utriculariabc 0.049 0.43 Gastrophryne carolinensisbc 0.000 <0.01 Bufo terrestrisbc 0.000 <0.01 Bufo quercicusc 0.000 <0.01 Anolis carolinensisb 0.100 0.46 Aspidoscelis sexlineatabc 0.040 0.10 Sceloporus woodibc 0.040 0.06 Plestiodon inexpectatusbc 0.038 0.48 Podomys floridanusabc 0.118 0.69 Peromyscus polionotusabc 0.087 0.69
42 Table 6. Dietary Niche Breadth (Hurlberts B) and Niche Overlap (Morisitas C) of Masticophis flagellum and Coluber constrictor Confidence intervals for Hurlbert s B represent expected values for an opportunistic predat or that consumes prey in proportion to its availability. Niche Breadth Niche Overlap Level of Analysis Snake Species B 95% CI C 95% CI Prey Category Masticophis flagellum 0.689 0.662-0.991 0.728 0.525-0.861 Coluber constrictor 0.956 0.751-0.992 Prey Species Masticophis flagellum 0.877 0.649-0.967 0.755 0.431-0.893 Coluber constrictor 0.431 0.682-0.916
Figure 1. Occurrence of Prey Categories Consumed by Masticophis flagellum and Coluber constrictor Expressed as a Proportion of Individuals of Each Species that Consumed Each Prey Category. Numbers above bars indicate the number of individuals of each prey category that were consumed. Proportions for C. constrictor do not sum to one because four individuals contained both amphibians and lizards in their stomachs. 43
Figure 2. Occurrence of Prey Species Consumed by Masticophis flagellum and Coluber constrictor Expressed as a Proportion of Individuals of Each Species that Consumed Each Prey Species. Numbers above bars indicate the number of individuals of each prey species that were consumed. A bbreviations used are: Agry = Acris gryllus, Hfem = Hyla femoralis Rcap = Rana capito Rspp = Rana catesbeiana + R. grylio Rutr = Rana utricularia Acar = Anolis carolinensis Asex = Aspidoscelis sexlineata Pine = Plestiodon inexpectatus Swoo = Sceloporus woodi Pflo = Podomys floridanus, and Ppol = Peromyscus polionotus Proportions do not sum to one because some individual snakes contained more than one prey species. 44
Figure 3. Relationship between Prey Category and Predator Size (Snou t-vent Length) for Masticophis flagellum and Coluber constrictor Each point represents a single snake. Four individual C. constrictor that contained both amphibians and lizards appear in both categories. 45
Figure 4. Ln-transformed Prey Mass as a Function of Ln-transformed Snake Mass for Masticophis flagellum (dashed line; adjusted R2 = -0.067, F(1,6) = 0.56, P = 0.48) and Coluber constrictor (solid line; adjusted R2 = 0.36, F(1,34) = 20.44, P < 0.0001). Only individual snakes that contai ned relatively intact prey were included in this analysis ( M. flagellum : n = 8, C. constrictor : n = 36). 46
47 Masticophis flagellum (Coachwhip) Positively Sel ects Florida Scrub Habitat at Multiple Spatial Scales The appropriate use of space by animal popul ations is vital to their growth and persistence. In particular, pr udent choice of foraging sites, refugia, and movement paths by individuals is crucial to en sure that they obtain adequate food and avoid predators and environmental extremes. Studies of habitat selection by animals provide this essential information about requirements for survival and can inform conservation about animal populations (Manly et al. 2002). In particul ar, understanding the mechanisms underlying habitat selection can help to identify factors limiting th e occurrence or abundance of populations (Blouin-Demers and Weatherhead 2001b, Pringle et al. 2003), and can help identify sources of population de clines (Waldron et al. 2006). Masticophis flagellum (Coachwhip) is a large, active snake found throughout the southern United States and norther n Mexico (Ernst and Barbour 1989). Masticophis flagellum has large home ranges, and makes fre quent long-distance movements (Secor 1995, Dodd and Barichivich 2007, Johnson et al. 2007). Masticophis flagellum occurs in many habitats (Carr 1940, Ernst and Barbour 1989, Tennant 1997), but recent studies indicate that M. flagellum primarily uses xeric, open-canopied habitats (Dodd and Barichivich 2007, Johnson et al. 2007). Masticophis flagellum actively searches for prey and captures it by rapid pursuit (S ecor 1995). The active habits of M. flagellum are accompanied by a high metabolic rate and relati vely frequent feeding (Secor and Nagy
48 1994, Secor 1995). Because of its high metabol ic rate and frequent movements, the spatial ecology of M. flagellum may be strongly tied to prey abundance. Our goals in this study were to document the spatial ecology and habitat selection of M. flagellum in central Florida. Based upon previo us studies (Dodd and Barichivich 2007, Johnson et al. 2007), I hypothesized that these va gile snakes would select xeric habitats, which primarily consisted of Florida scrub at my study site. Because Florida scrub habitat on my study site occurred in patches, I predicted that M. flagellum at my site would exhibit large home range sizes and long-distance movements between patches. I further hypothesized that M. flagellum would select Florida scrub patches based upon the prey abundance in each patch. MATERIALS AND METHODS Study Site I conducted this research at the Lake Arbuckle tract of the Lake Wales Ridge State Forest in southeastern Polk County, Florida, USA (27.67N Latitude, 82.43W Longitude). The site consists of a series of patches of xeric Florida scrub habitat and wetlands in a matrix of mesic, and more de nsely vegetated, pine flatwoods habitat (Fig. 5). Wetlands consisted primarily of depressi on marshes that were seasonally inundated and larger forested wetlands. Depressi on marshes were dominated by Monocots, primarily Panicum hemitomon ; some marshes also contained small shrubs ( Salix spp. and Hypericum spp.). Forested wetlands were prim arily bayheads and sloughs dominated by Gordonia lasianthus and Persea palustris
49 Florida scrub is a rare, xeric habitat that contains many precinctive organisms (Myers 1990, McCoy and Mushinsky 1992). Florid a scrub at the site was primarily oak scrub, and was dominated by a 0.5-1.5 m high midstory of Quercus spp., Lyonia spp., Sabal etonia and Serenoa repens Canopy cover in the Florida scrub was generally less than ten percent, and was dominated by Pinus elliottii. Groundcover in th e Florida scrub was primarily bare sand and leaf litter with little herbaceous vegetation. A few xeric areas on the site were sandhill habitat, which wa s distinguishable from Florida scrub by its increased cover of herbace ous vegetation (particularly Aristida and Andropogon spp.), and the presence of Quercus laevis and/or Pinus palustris Pine flatwoods habitats covered the gr eatest amount of area at the site. These habitats had a denser P. elliottii canopy than the Florida scrub, and also had a thicker groundcover. Two distinct communities were evident in the pine flatwoods. One community was dominated by a thick groundcover of Panicum abscissum and had a sparse midstory consisting of scattered stands of S. repens and Ilex glabra The other community was more typical of flatwoods habitats, and consisted of a midstory dominated by S. repens Lyonia lucida and Vaccinium myrsinites. The groundcover in the typical pine flatwoods community c onsisted of diverse graminoids and forbs occurring at low densities. Because the stru cture of these two co mmunities was distinct, hereafter I distinguish them as cutthroat (after the common name of P. abscissum ) and pine flatwoods, respectively.
50 Field Methods I used radio telemetry to determine the spatial ecology and ha bitat selection of M. flagellum I captured individuals opportunistica lly by hand or in drift fence/funnel trap/pitfall trap arrays installed in patches of Florida scrub habitat from March, 2004 to August, 2005. I transported individuals weighing greater than 320 g to the University of South Florida for surgical implantation of radio transmitters. I surgically implanted an 8 g Holohil SI-2 radio transmitter (Holohil System s, Ltd. Carp, Ontario, Canada) within the coelom of each individual using standard procedures (Reinert and Cundall 1982). I allowed individuals to recover in the laborator y overnight, and released each where it was captured the day following surgery. I did not mo nitor individuals for the first week after release to avoid disturbance and allow them to heal. I physically located individuals as fre quently as possible, with a maximum frequency of one location per day. From March through October, I located each individual an average of four times per week. From November through February, I reduced the frequency of relocations to once or twice per week, depending upon the observed frequency of movement of indi viduals. Most locati ons took place during daylight hours because M. flagellum is primarily diurnal (Ern st and Barbour 1989), but I varied the time of day at which I located indi viduals. I systematically selected the first individual located to avoid locating the same individual at the same time each day, and determined the order of relocations based upon the proximity of individuals to my current position. I collected data using a Trimble GeoXT handheld GPS (Trimble, Sunnyvale, California, USA), and offset the location of the individual when it was aboveground to
51 avoid disturbance to the indi vidual. All locations were differentially corrected in the laboratory to improve the precision of locations. I described habitat characteristics in th e immediate vicinity of the observed individual. I categorized hab itat as Florida scrub, sandhill pine flatwoods, cutthroat, marsh, swamp, and pasture. If the individual was located in Florida scrub habitat, I recorded the patch of Florida scrub in whic h the individual was found. I also determined the behavior of individuals at each location. I categorized behavior as basking, resting, hiding, moving, fleeing, foraging, or mating. Basking was defined as the individual lying motionless with greater than fifty percent of its body in direct sunlight. Resting was similar to basking, except that the body of the individual was primarily in shade. Hiding included locations where the individual wa s underground or mostly concealed beneath leaf litter. I considered an individual to be moving if its locomotion was at a relatively slow rate of speed, with no obvious searching or foraging behavior. Fleeing individuals were moving rapidly for short distances, usually because of my approach. Foraging individuals exhibited extensiv e searching behavior (tongue-flic king with extensive lateral movements of the head) or evid ence of prey pursuit, prey capture, or the process of prey ingestion. I did not include di gestion as a foraging behavior, but generally categorized individuals with a food bolus as hiding, rest ing, or basking. Mating individuals included any of the phases of courtship, including tactile-chase, tactile-alignment, and intromission and coitus (Gillingham 1987). To determine if prey abundance influenced the spatial ecology and habitat selection of M. flagellum I monitored the abundance of its prey. In particular, I focused on the abundance of three lizard species ( Anolis carolinensis [Green Anole], Aspidoscelis
52 sexlineata [Six-lined Racerunner], and Sceloporus woodi [Florida Scrub Lizard]) and two small mammal species ( Peromyscus polionotus [Oldfield Mouse] and Podomys floridanus [Florida Mouse]) documented in the diet of M. flagellum at my study site (Halstead et al. In review). I sampled all prey using a total of 13 dr ift fence/pitfall/funnel trap arrays installed in nine focal Florida scr ub patches. I installed on e trap array in each of seven patches (1.5 12.6 ha), two trap arrays in one patch (40 ha), and four trap arrays in the remaining patch (170 ha). I opened trap arrays and checked them daily for seven consecutive days, then closed them for 21 c onsecutive days as part of a robust markrecapture design (Pollock 1982). I sampled prey from March 2004 to June 2006 for a total of 20 primary periods. Traps were closed from December 2004 to February 2005 and again from November 2005 to March 2006 because of unacceptably high mortality rates of trapped small mammals caused by cold overnight temperatures. I uniquely marked each captured lizard using toe clips (Waichman 1992), and each captured small mammal using individually numbered size 1005-1 Monel ear tags (National Band and Tag Co., Newport, KY, USA). All captured indi viduals were released one meter outside the trap array immediately after processing. Analytical Methods Unless otherwise indicated, I used the individual as the sampling unit for all analyses. To determine whether an estim ate of home range was applicable to M. flagellum I evaluated movement paths of individuals for their fit to a random walk model, using subsequent locations only if th ey were greater than three meters apart to avoid spatial autocorrelation. The random wa lk model is appropriate for relatively
53 infrequent locations, because they lack the correlation in move di rection found in actual movement paths (Turchin 1998). The random walk model was strongly rejected for each individual (Fig. 6), and indicated that M. flagellum movement is bounded within a home range. Therefore, I proceeded with analys es appropriate to an organism whose movements primarily occur within a bounded home range. I calculated home range sizes of M. flagellum by multiple techniques. I used the 100% minimum convex polygon (MCP) because of its historic use and for analyses of home range and habitat selection (see below). I also calculate d the 95% kernel utilization distribution to estimate home range area, and the 50% kern el utilization distribution estimate to estimate core activity area. Kernel utilization distributions have the advantage of making use of all locations and describe the intensity of space use (Worton 1989, Kernohan et al. 2001). I calcula ted the kernel utilization distribution on a 100 x 100 grid for each individual, and used the ad hoc method of bandwidth selection because of convergence problems encountered when select ing the bandwidth by least squares crossvalidation. Home ranges were calculated usi ng the package adehabitat for R (Calenge 2006). To determine the vagility of M. flagellum I calculated mean daily displacement. I divided distance between all consecutive locations divide d by the number of days between locations for each individual. This analysis included locations where the snake did not move, or moved less than three meters between relocations. I also calculated the displacement between biol ogically independent locati ons. I defined biologically independent locations as those greater than th ree meters apart that did not include ecdysis (indicated by dull appearance, inactivity, and cloudy, bluish eyes) or digestion of large
54 meals (indicated by the presence of a f ood bolus). My definition of biological independence was not equal to spatial inde pendence of locations The analysis of independent locations provides a measure of the vagility of active snakes. I also calculated daily displacement for moves beginning and ending within a patch of Florida scrub and for all other habitats combined. I tested for differences in mean daily displacement between habitats with a paired t-test to determine if individuals moved greater distances while in other ha bitats than in Florida scrub. I examined habitat selection at three scales (Johnson 1980). The first scale I examined was the selection of home ranges w ithin the study site (s econd-order selection [Johnson, 1980]). For this scale of selection, I de fined habitat availability as the area of each of seven habitat categor ies (marsh, swamp, cutthroat, pine flatwoods, sandhill, Florida scrub, and pasture) within the 100% MC P enclosing all snake locations (Fig. 5). I defined use separately for each individual as the area of each habita t category within the individuals 100% MCP home ra nge. I used the MCP home range because at this scale of analysis, I was interested in the selection of the home range boundari es, rather than the intensity of use within that home range. This definition better fit my conceptualization of second-order selection than other methods of delimiting the home range. For both use and availability, I calculated the area of each habitat based upon a combination of soils and satellite imagery of the study site using the X-Tools Pro extension in ArcGIS 9.0 (ESRI, Redlands, California, USA). Gr ound-truthing indicated that the combination of soils and satellite imagery was a good approximation to the delineation of hab itats in the field. I calculated the selection of habitat at this scale using the selection ratio, ij (Manly et al. 2002). I tested for differen ces in the selecti on of home ranges by
55 individuals, non-random selecti on by at least some individua ls, and overall selection of habitat across individuals using the log-likelihood goodness of fit statistic ( G2). I calculated i for each habitat as the average of the ij taken across individuals, excluding individuals for which the habitat was not availa ble. To determine if the selection of each habitat was nonrandom, I compared my estimate of i to one (the expected value under the null hypothesis of use in proportion to availability) wi th a one-sample t-test. To determine if each habitat was selected significan tly more or less than every other habitat, I compared the i with a two-sample t-test with Welchs correction for unequal variances. I calculated Manlys st andardized sel ection ratios ( Bi) from the i, because Bi is interpretable as the estimated probability that habitat i would be the next habitat selected if all habitats were equa lly available (Manly et al. 2002). In addition to the selection of a home ra nge, I examined the se lection of habitats within the home range (Johnson 1980). I defined availability separately for each individual as the proportion of each indi viduals 100% MCP home range comprised of each habitat category. At this sc ale, I used the MCP home range because it allowed us to achieve greater independence of available hab itat and the intensity of use than would a kernel utilization distribution. Use was define d as the count of independent locations for each individual in each habitat category. I analyzed the selection of habitats within the home range using Poisson regression. For each individual, I used habitat availability as a rate parameter in the Poisson regression model under the null hypothesis that the count of locations within each habitat would be proportional to th e area of each habitat found within the individuals home ra nge. I fit several alternative mode ls, including no ha bitat selection,
56 constant selection of habita t by all individuals, and selec tion of habitat varying among individuals. In addition to these models, I also evaluated models that constrained the differences in habitat selection among indivi duals to be a function of sex or snout-vent length (SVL). Because I used non-nested models, I selected the best-fit model using Akaikes An Information Criterion corrected for small sample sizes, AICc (Burnham and Anderson 2002). After choosing a best-fit m odel, I calculated the Resource Selection Function (RSF) for each habitat for each indivi dual using the coefficients of the fitted model (Manly et al. 200 2). I then calculated for the reference individual and category (which by definition had an RSF of one), and multiplied each RSF by this value to determine each of the ij. I determined the statistical significance of the i and differences in selection between ha bitats using t-te sts as above. I used log-linear modeling and correspondence plots to determine if the use of certain habitats was related to snake behavior. I used all locations for this analysis under the assumption that behaviors observed at consecutive locations were independent, regardless of spatial location. I also pooled observations acr oss individuals, and excluded data from habitats or behaviors with fewer than twenty observations. Independence of behavior and habitat was asse ssed with log-linear models. Association among specific behaviors and habitats was examined graphically with a correspondence plot. We also examined habitat selection at a level between thirdand fourth-order selection (Johnson 1980). In this case, I examined the select ion of nine focal Florida scrub patches for which I estimated the abundance of the prey of M. flagellum (Halstead et al. In review) using mark-recapture techniques (see below). For this scale, I defined availability as the proportion of each indivi duals MCP that consisted of each sampled
57 patch of Florida scrub. I defined use as the c ount of locations in each of these patches. Only nine individuals with lo cations in more than one focal patch were used for this analysis. To determine if prey abundance aff ected the space use of snakes, I examined patterns of patch selection as a function of lizard abunda nce, mammal abundance, total prey (lizard plus mammal) abundance, and S. woodi abundance. I examined the influence of the abundance of S. woodi individually because it is precinctive to Florida scrub and is a common item in the diet of M. flagellum (Halstead et al. In review). We estimated prey abundances in each focal patch using Jolly-Seber markrecapture models. For each prey species, I combined capture histories from each seven day secondary period prior to analysis. I fit the POPAN formulat ion of Jolly-Seber models using Program MARK 4.3 (White, 2006). I modeled capture probability as constant ( p) or time-varying ( pt) for each species. I assumed weekly capture probabilities to be the same at all trap arrays, which were identical, were placed in Florida scrub habitat with very similar structural attributes and were in close proximity to one another, so that each array experienced similar e nvironmental conditions. I modeled monthly survival rate as constant ( ), array-specific ( g), time-varying ( t), and interactive and additive effects of array and time ( g*t, and g+t, respectively). Likewise, I modeled the monthly probability of entry of indivi duals into the population as constant ( pent), arrayspecific (pentg), time-varying ( pentt), and interactive and additive effects of array and time (pentg*t and pentg+t, respectively). I fit all possi ble combinations of capture probability, survival rate, and probab ility of entry for all prey except A. carolinensis and P. polionotus Too few individuals of these spec ies were captured to estimate the parameters of models with multiple interactive and additive terms, so I did not attempt to
58 fit models containing these terms to data from these two species. For each prey species, I used the model with the lowest AICc to estimate abundance in each patch of Florida scrub for each month. For each individual M. flagellum I estimated the mean monthly abundance of each prey species in each patc h within the individuals MCP home range over the time period during whic h the individual was tracked. Similar to my analysis of habitat selection, I used Poisson regression to examine Florida scrub patch selection. I fit Poisson rate models including no patch selection, constant selection of patches across indi viduals, and patch selection varying among individuals. In addition to th ese basic models, I constrained patch selection by individuals to be a function of sex or SVL. I also cons trained patch selection to be a function of S. woodi abundance, lizard abundance, mammal abunda nce, and total prey (lizard plus mammal) abundance. Finally, I evaluated a model that allowed selection of patches to be an interactive function of snake SVL a nd mammal abundance because mammals were much larger prey than lizards and maximum prey size in snakes is a function of gape size (Miller and Mushinsky 1990, Cundall and Gree ne 2000), which is strongly correlated with SVL in M. flagellum ( n = 18, Gape = 0.126*SVL + 26.9, R2 = 0.93, F(1,16) = 234, P < 0.001). I selected a best-fit model using AICc, and calculated ij as above for habitat selection. I determined the st atistical significance of the i and differences in selection between habitats usi ng t-tests as above. Statistical analyses were conducted us ing the programs MARK 4.3 (White 2006) and R 2.3.1 (R Core Development Team 2006). Descriptive statistic s are reported as mean (standard error), and sta tistical significance was set at = 0.05.
59 RESULTS I located 14 individual M. flagellum a total of 1556 times from March 2004 through May 2006 (Table 7). Nine individuals were female and five were male. I eliminated 167 locations that were less than three meters from the previous location, which was primarily caused by individuals in ecd ysis, digesting large meals, or inactive because of cool temperatures during the winter. The adult M. flagellum in my study had relatively large home ranges (Fig. 7). Mean 95% fixed kernel utilization distribu tion home range area for the population was 177.07 (.24) ha (Table 8). Mean core ac tivity area was 38.42 ( 8.33) ha. Males and females did not differ in home range area (males: 212.73 [.38]; females: 157.26 [.31]; W = 33, P = 0.19) or core activity area (males: 51.47 [.87]; females: 31.17 [.19]; W = 34, P = 0.15). Home range area was pos itively correlated with SVL ( = 0.55, P = 0.043), but was not related to the tracking period duration ( = 0.21, P = 0.46) or number of locations ( = 0.40, P = 0.15) of the individual. Home range size was negatively correlated with proportion of Fl orida scrub habitat found within the home range ( = -0.56, P = 0.042). Adult M. flagellum made relatively long movement s between relocations. Mean daily displacement was 1 32.1 (.3) m for all locati ons, and 153.0 (.8) m for biologically independent locations (Table 9) Males and females did not differ in mean daily displacement (males: 170  m; females: 114  m; t = 2.04, df = 8.47, P = 0.074). Mean daily displacement was positiv ely correlated with home range size ( = 0.58, P =0.033). Daily displacements in Florida scrub habitat were significantly shorter
60 than displacements in other habitats (scrub : 97  m; other: 156  m; t = -2.84, df = 13, P = 0.014). Masticophis flagellum selectively included habitats in their home ranges (Table 10). Individuals incorporated habitats into th eir home ranges differently from one another ( G2 = 543.17, df = 78, P < 0.001), and at least some individuals selected their home ranges non-randomly (G2 = 1335.15, df = 42, P < 0.001). On average, M. flagellum selected home ranges non-randomly with regard to habitat ( G2 = 791.98, df = 6, P < 0.001). In particular, M. flagellum included greater proportions of Florida scrub and lesser proportions of cutthroat, pasture, and swamp in to their home ranges than was available on the study site (Table 11). Masticophis flagellum home ranges included significantly more Florida scr ub than all other habitats ex cept sandhill and marsh (Table 11). Masticophis flagellum was also strongly selective of habitats within the home range (Table 10). The model for habitat sele ction varying among individuals was strongly supported, and no evidence existed for different se lection of habitats of different sex or size (Table 12). In particular, Florida scrub and pine flatwoods were positively selected, and cutthroat, swamp, and marsh were negative ly selected (Table 11). Florida scrub was selected significantly more than any other ha bitat, and pine flatw oods was selected less than Florida scrub, but significantly more than all other habitats except sandhill. Swamp and marsh were more strongly avoided than a ll other habitats except pasture (Table 12). Masticophis flagellum exhibited different behaviors in different habitats (Fig. 8). Log-linear modeling indicated strong evidence for non-inde pendence of habitat and behavior ( G2 = 211.46, df = 18, P < 0.001). In particular, loca tions categorized as hiding
61 occurred in Florida scrub and locations categor ized as moving occurred in pine flatwoods more often than expected if behavior was independent of habitat (Fig. 9). Masticophis flagellum also was selective of the patches of scrub in which it occurred (Table 13). Patches of scrub varied substantially in the abundance of each prey species (Table 14). The model indicating pa tch selection varying among individuals was strongly supported, and no evidence existed for patch selection varyi ng as a function of prey abundance within the patch (Table 15). Patch B was positively selected, and patch C was negatively selected (Table 16). Patch D was available to only one snake, so statistical significance of selection of th is patch could not be asse ssed. Patch B was selected significantly more than all other patches (Table 16). Although the model selection procedure indicated that individuals did not select pa tches based on prey abundance during the tracking period, Bi values from the selected model were positively correlated with lizard abundance ( = 0.80, P = 0.014), but not S. woodi ( = 0.48, P = 0.19) or mammal abundance ( = 0.07, P = 0.88), over the entire study period. DISCUSSION Masticophis flagellum is a xeric habitat specialist. Regardless of the scale of analysis, M. flagellum positively selected Florida scrub, and avoided mesic and hydric habitats. Other studies of the us e and selection of habitats by M. flagellum support my conclusion that this species is a xeric habitat specialist (Dodd and Barichivich 2007, Johnson et al. 2007). Pine flatwoods habitat was the most heterogeneous habitat on the site, and many of the observations of M. flagellum in pine flatwoods habitat were in scrubby flatwoods that were characterized by low, shrubby vegetation with a sparse
62 canopy similar to that found in Florida scrub habitat. Interestingly, home range size was negatively correlated with the proportion of av ailable Florida scrub habitat within the home range, indicating that M. flagellum may require some res ource(s) found exclusively in Florida scrub habitat. Large home ranges ma y be an artifact of maintaining access to a minimum quantity of Florida scrub habita t in which to obtain these resources. The mechanism underlying the strong association of M. flagellum with Florida scrub habitat remains unclear. I propose four alternative, but not mutually exclusive, hypotheses for the selection of Florida scrub habitat by M. flagellum The first is greater prey abundance in Florida scrub. Patches of Fl orida scrub at my s ite contained abundant lizard and mammalian prey, but quantitative estimates of prey abundance in other habitats at the site are not available. My observations suggest th at lizards and small rodents are more abundant in the Florida scr ub than other habitats. Indeed, two important prey species of M. flagellum S. woodi and P. floridanus (Halstead et al. In review), are precinctive to Florida scrub ha bitat (McCoy and Mushinsky 1992). Sigmodon hispidus (Cotton Rat) was preyed upon by large M. flagellum and this species was more frequently captured in Sherman live traps in pine flatwoods than in Florida scrub (B. Halstead, unpublished data). Adult S. hispidus are formidable and likely constitute suitable prey only for the largest M. flagellum individuals on the site. Prey abundance appears to affect movements and habitat se lection of some snake species (Madsen and Shine 1996, Heard et al. 2004), but studies fr om temperate regions of North America have not found such a relationship (Blouin-Demers and Weatherhead 2001b, Carfagno et al. 2006). Whether M. flagellum selects Florida scrub because of high prey density relative to other habita ts will require additional data on prey abundance in other habitats.
63 A second mechanism for the selection of Florida scrub habitat by M. flagellum is a greater abundance of ref ugia in this habitat. M. flagellum was frequently located underground completely within or with just the anterior part of its body emerging from Gopherus polyphemus (Gopher Tortoise) and P. floridanus burrows. Accidentally disturbed snakes usually fled into near by burrows. Although I did not quantify burrow abundance, burrows of G. polyphemus appeared most abundant in Florida scrub habitat; the burrows of P. floridanus were certainly more abundant in Florida scrub than elsewhere. I did not obser ve confirmed predation upon M. flagellum in my study, but one adult M. flagellum captured in 2006 had scarring, broken ribs, and exposed vertebrae at spacing that appeared consistent with injuries caused by raptor talons. Buteo lineatus (Red-shouldered Hawk) was observed to prey upon Coluber constrictor (Eastern Racer), a closely-related, diurnally-active snake with activity patterns similar to those of M. flagellum The abundance of refugia from predat ors in Florida scrub habitat might explain the selection of this habitat by M. flagellum A third reason why M. flagellum might select Florida scrub is that thermal conditions might be more suitable in Florid a scrub than elsewhere. Appropriate thermal conditions may be particularly important for M. flagellum because it has a narrow thermal tolerance (Secor 1995). Selection of habitats and microhabitats based upon thermoregulatory considerations appears widespread in snak es, particularly those found in temperate regions (Blouin-Demers a nd Weatherhead 2001b, Row and Blouin-Demers 2006), with nocturnal habits (Pri ngle et al. 2003), or near the extremes of the species range (Heard et al. 2004). Fl orida scrub at my site has less canopy cover than other habitats, and might provide greater bask ing opportunities than other habitats. The
64 abundance of burrows and shrubs within Flor ida scrub habitat would also allow escape from high midday temperatures. Thus, M. flagellum may select Florida scrub because the combination of ample sunlight and thermal refugia would allow pr ecise thermoregulation in this habitat. A final potential mechanism underlying the selection of Florida scrub habitat by M. flagellum is the structural attributes of Flor ida scrub. Florida scrub has a distinct structure, with little herb aceous groundcover, abundant shrubs, and, in the case of oak scrub typical of my study area, little canopy cover. Although these structural attributes undoubtedly affect thermal conditions in Florid a scrub habitat, they may also influence the selection of this habitat in other important ways. Because M. flagellum relies on rapid locomotion for both prey capture and predator escape (Ruben 1977, Secor 1995), and substrate structure affects locomotor ability in snakes (Cundall 1987, Kelley et al. 1997), habitat structure may be an important compone nt of foraging success (Mullin et al. 1998) and survival rates of M. flagellum Thus, the structure of Florida scrub habitat itself may be important independent of its effect s on prey abundance and thermoregulation. Which of these mechanisms is most impor tant for the selectio n of Florida scrub by M. flagellum is unknown, but each Florida scrub characteristic likely plays some role. The association of hiding behavior w ith Florida scrub habitat suggests that M. flagellum uses Florida scrub because of the abundan ce of burrows found in this habitat. The abundance of burrows is not independent of prey density, however, because P. floridanus provides both burrows and food for M. flagellum and many potential prey species seek refuge in the burrows of G. polyphemus (Witz et al. 1991). Underground activity may have included foraging and thermoregulatory behaviors in addition to hiding. Likewise,
65 the relatively few foraging observations in Florida scrub do not preclude its importance as foraging habitat. If prey in Florida scrub are more abundant than prey in other habitats, individuals may spend less time searching fo r these abundant prey, resulting in fewer foraging observations. Similarly, if prey in scr ub are smaller or less formidable than prey in other habitats (for example, lizards and small rodents versus la rge rodents), handling times may be reduced. An adult M. flagellum was observed to detect, chase, capture, and consume an individual A. carolinensis in less than one minute. If M. flagellum primarily preys upon these small prey items in Florid a scrub, handling prey may occupy relatively little of the time budget of M. flagellum Thus, both of these situ ations could result in relatively few foraging observations in Florida scrub even if most prey were captured in this habitat. In congruence with these observatio ns, the steady-state satiation functional response model predicts that digestion-limited predators will decrease foraging time with increasing prey density (Jeschke et al. 2002). Masticophis flagellum is digestion-limited, and individuals were frequently observed with visible food boluses in the Florida scrub. These observations were categorized as ba sking, resting, or hiding, depending upon the circumstances under which the individual was found. Thus, the lack of association of M. flagellum foraging behavior with Florida scrub habitat does not indicate that Florida scrub is unimportant as foraging habitat for M. flagellum The relatively short daily di splacement distances in Fl orida scrub, compared to the long daily displacement distances in othe r habitats (especially pine flatwoods), suggest that M. flagellum positively responds to some resource in Florida scrub. Random walk models did not fit the overall movements of M. flagellum because of the bounded nature of movements within a home range, but analyzing the movement paths of M.
66 flagellum within habitats was complicated by frequent movements across habitat boundaries (Turchin 1998). Future studies in more coarse-grained environments might profitably employ movement analyses ba sed upon random walk models. Additional studies involving prey abundance in alternative habitats, refuge availability and use, and thermal ecology of M. flagellum are required to elucidate mechanisms leading to selection of Florida scrub or avoidance of other habitats Ultimately, mechanistic home range models (Moorcroft and Lewis 2006) may prove useful in this regard. Masticophis flagellum was also selective of particul ar Florida scrub patches. The largest patch on the study site, scrub B, wa s positively selected even after taking the proportion of this patch within each individuals ho me range into account. This patch was also the only patch available to (and used by) all nine individuals included in this analysis. Prey abundance did not influence the selection of Flor ida scrub patches by snakes in the Poisson rate models, but lizard abundance over the entire study period was positively correlated with the Bi from the best-fit model. This discrepancy may have been caused by a large number of zero counts because most patches were not available to most individuals. At the population le vel, however, lizard abundance is related to Florida scrub patch selection and suggests that prey a bundance plays a role in patch selection by M. flagellum Masticophis flagellum is a vagile species with large home ranges throughout its geographic range. The mean 100% MCP home range size in my study was larger than that in the Mojave Desert (Secor 1995), eas tern Texas (Johnson et al. 2007), and northcentral Florida (Dodd and Barichivich 2007) (T able 17). Each of these studies tracked M. flagellum for different lengths of time, but I found no relationship between home range
67 size and number of locations or number of days tracked within my study. Mean daily displacement, which is approximately equivale nt to actual movement rate (Johnson et al. 2007) and straight line distance (Secor 1995), in my study was great er than that in eastern Texas, less than that in nort h-central Florida, and very si milar to that in the Mojave Desert (Secor 1995, Dodd and Barichivich 2007, Johnson et al. 2007). The greater daily displacements in pine flatwoods habi tat than Florida sc rub suggest that M. flagellum constrains much of its activity to Florid a scrub, and makes long-distance movements through other habitats presumably to reach ot her patches of Florida scrub (Johnson et al. 2007). Displacement distance data are supporte d by the observation that moving behavior was positively associated with pine flatw oods habitat. Data from all these sources indicate that M. flagellum is a vagile snake that occupi es large home ranges relative to many other snake species (Gregory et al. 1987, Macartney et al. 1988). Although seasonal differences in activity patterns likely affect the movement ecology of M. flagellum (Dodd and Barichivich 2007, Johnson et al 2007), individuals bask and move throughout the year in central Florida, a nd do not move to hibernacula or occupy different habitats in different seasons. No obvious differences in move lengths or habitat use were associated with reproductive behavior, but I obs erved relatively little mating and no oviposition. Perhaps because its actively foraging habits result in extensive movements, M. flagellum does not appear to exhibit diffe rent movement behavior during reproductive periods nor sexual dimorphism in spatial ecology (Johnson et al. 2007). Although my study was restricted to adult M. flagellum Florida scrub is likely an important habitat for all age cla sses. I captured 81 individual M. flagellum in my traps in Florida scrub habitat, but only twenty of these were large enough for radio transmitters.
68 All size classes of M. flagellum were observed in Florida scrub, but very few opportunistic observations of M. flagellum occurred in other habitats on the site. Although my observations are not corrected for sampling effort, they indicate that Florida scrub is an important habitat for M. flagellum of all sizes. The habitat specificity and freque nt long-distance movements of M. flagellum make this species particularly vulnerable to anthropogenic habitat alteration (Waldron et al. 2006). Florida scrub is a historically rare habitat, and most Florida scrub has been converted to agricultu ral and residential uses (Myers 1990). Although habitat reduction in itself is detrimental to species with large home ranges (Woodroffe and Ginsberg 1998), the frequent movements of M. flagellum between patches make it particularly vulnerable to habitat fragmentation. Only one radio-tracked individual used agricultural lands, and this individual was struck by a rotary mower while fora ging in pasture. Although my study had little to no automobile traffic, it is likely that roads are a major source of mortality in other M. flagellum populations (Bonnet et al 1999, Andrews and Gibbons 2005, Row et al. 2007). Masticophis flagellum requires large tracts of land with abundant Florida scrub habitat to persist in central Florid a (Dodd and Barichivich 2007). In summary, M. flagellum is a xeric habitat specialist with large home ranges resulting from frequent long-dist ance movements. As I predicted, M. flagellum positively selected Florida scrub habitat, and avoided mesic and hydric habitats, regardless of the scale at which selection was examined. Seve ral mechanisms, including prey abundance, refuge abundance, thermoregulatory opportun ity, and structural attributes may account for the positive selection of Florida scrub habi tat. Prey abundance is positively related to Florida scrub patch selection by M. flagellum The habitat specificity, large home ranges,
69 and frequent long-distance movements of M. flagellum suggest that protection of large tracts of land containing xeric habitats will be required if this large, charismatic snake is to persist.
70 Table 7. Characteristics of Radio Tracked Masticophis flagellum SVL = snout-vent length, Raw Locs = the to tal number of locations of each individual Ind. Locs = the number of successive locations greater than three meters apart, and Prec. = mean hor izontal precision of GPS locations. Codes for Fate are: TRH = transmitter removed, snake health y upon release; SBM = struck and killed by rotary mower in pasture; NRH = snake not recovered for transmitter removal, but healthy at last observation; DU = sna ke died of unknown causes; DT = snake died in trap; and DC = snake died from complications asso ciated with the radio transmitter. Individual was not radio tracked continuously, but was lost and later recaptured and implanted with a new transmitter. ID Sex SVL (mm) Tracking Period Days Raw Locs Independent Locs Precision (m) Fate* B1 F 1335 31-Mar-04 18-Apr-05 383 101 86 0.9 TRH F1 F 1405 31-Mar-04 27-Jun-04 88 31 26 0.9 SBM C1 M 1650 15-Apr-04 13-Jan-06 638 233 185 1.8 TRH F2 M 1570 17-Apr-04 17-Apr-06 732 208 175 1.8 TRH B2 M 1290 15-Jun-04 13-Apr-05 302 86 72 1.2 TRH B3 M 1325 15-Jun-04 12-Apr-06 666 236 190 2.0 TRH G1 F 1500 15-Jun-04 18-Mar-05 276 71 56 1.0 NRH B4 F 1380 01-Oct-04 06-Jul-05 281 90 70 2.3 DU I1 F 1220 13-Apr-05 15-Aug-05 129 75 60 2.8 DT B5 F 1440 02-May-05 17-Apr-06 350 125 104 2.4 TRH I2 F 1520 09-May-05 03-Mar-06 298 122 105 2.3 TRH B6 M 1500 25-May-05 10-Nov-05 169 87 71 2.3 DC B7 F 1225 01-Aug-05 22-May-06 294 109 96 2.3 TRH I3 F 1380 01-Aug-05 31-May-06 303 114 93 2.5 TRH
71 Table 8. Home Range Estimates for Masticophis flagellum MCP = minimum convex polygon, HR = home range. Kernel utilization di stributions were calculated using a fixed kernel on a 100 x 100 grid, with bandwidth selection by the ad hoc method. Kernel Utilization Distribution ID 100% MCP (ha) 50% Core (ha) 95% HR (ha) B1 42.52 14.50 56.16 F1 191.41 56.43 352.42 C1 154.37 46.50 188.18 F2 345.44 102.37 400.90 B2 49.25 19.95 78.76 B3 174.29 52.53 213.20 G1 45.78 17.86 72.07 B4 71.61 14.10 82.44 I1 26.71 11.31 43.65 B5 48.32 13.96 63.23 I2 337.97 103.96 550.57 B6 189.66 36.02 182.62 B7 66.66 23.82 91.28 I3 87.06 24.63 103.52 Pop. Mean (SE) 130.79 (28.46) 38.42 (8.33) 177.07 (41.24)
72 Table 9. Masticophis flagellum Mean Daily Displacement. Raw = daily displa cement calculated from all consecutive locations, Independent = daily displacement calculated from all sequential di splacements greater than three meters and excluding snakes in ecdysis, digesting large meals, or othe rwise inactive, Scrub = daily displacement calculated for biologically independent consecutive locations within a patch of Florida scrub, and Other = daily displacemen t calculated for biologically independent consecutive locations within all habitats except Florida scrub. The ac tive season was considered al l locations from March throu gh October, but movement of each i ndividual occurred in every month. ID Raw Independent Scrub Other Active Season B1 119 (15) 136 (16) 103 (13) 113 (26) 150 (18) F1 122 (25) 142 (28) 9 (3) 172 (54) 142 (28) C1 149 (28) 175 (14) 127 (19) 197 (26) 194 (15) F2 207 (21) 241 (24) 106 (53) 251 (33) 272 (27) B2 88 (12) 97 (12) 84 (13) 91 (32) 105 (14) B3 143 (11) 168 (13) 135 (14) 126 (44) 191 (15) G1 93 (15) 109 (18) 94 (16) 57 (NA) 135 (22) B4 134 (19) 169 (19) 56 (16) 225 (70) 188 (26)
73 Table 9 (Continued). ID Raw Independent Scrub Other Active Season I1 130 (18) 157 (20) 146 (19) 151 (145) 157 (20) B5 127 (14) 148 (16) 128 (19) 92 (33) 168 (19) I2 162 (19) 183 (21) 105 (19) 259 (58) 217 (26) B6 157 (20) 170 (22) 158 (41) 165 (39) 179 (22) B7 103 (12) 112.8 (13) 58 (11) 107 (32) 141 (16) I3 114 (12) 134.0 (14) 48 (14) 173 (32) 162 (16) Population 132 (8) 153.0 (10) 97 (11) 156 (17) 171 (11)
74 Table 10. Habitat Use and Availability for Masticophis flagellum Availability (Avail) is the expressed as th e percentage of each individuals home range made up of each habitat. Use is the n umber of locations (percentage of locations) in each habitat. Blanks indicate a habitat that was not available to the individual. For the study site, habitat availability is expressed as hectar es (percentage of total). Marsh Swamp Cutthroat Flatwoods Sandhill Florida Scrub Pasture ID Avail Use Avail Use Avail Use Ava il Use Avail Use Avail Use Avail Use B1 2.9 0 (0) 2.1 0 (0) 26.4 6 (7.0) 13.0 13 (15.1) 9.4 6 (7.0) 46.2 61 (70.9) F1 4.9 0 (0) 0.4 0 (0) 24.0 1 (3.8) 36.5 9 (34.6) 6.9 12 (46.2) 27.3 4 (15.4) C1 0.8 0 (0) 0.9 2 (1.1) 10.2 3 (1.6) 19.3 76 (41.1) 8.1 26 (14.1) 60.1 78 (42.2) F2 7.1 0 (0) 2.8 0 (0) 19.7 13 (7.4) 42.0 78 (44.6) 0.7 1 (0.6) 26.2 83 (47.4) B2 0.1 0 (0) 3.6 0 (0) 19.3 4 (5.6) 15.8 2 (2.8) 11.5 22 (30.6) 49.7 44 (61.1) B3 1.9 0 (0) 10.1 0 (0) 10.8 0 (0) 33.0 40 (21.1) 5.6 7 (3.7) 38.6 143 (75.3) G1 6.3 0 (0) 4.4 0 (0) 0.6 1 (1.8) 20.8 3 (5.4) 21.0 9 (16.1) 46.8 43 (76.8) B4 8.0 0 (0) 11.1 0 (0) 55.0 19 (27.1) 26.0 51 (72.9) I1 4.2 0 (0) 22.0 7 (11.7) 73.6 53 (88.3) B5 2.2 0 (0) 10.4 1 (1.0) 18.9 16 (15.4) 10.5 12 (11.5) 58.1 75 (72.1 I2 41.5 0 (0) 18.3 0 (0) 9.8 32 (30.5) 1.3 3 (2.9) 29.2 70 (66.7) B6 17.5 0 (0) 24.3 5 (7.0) 17.3 28 (39.4) 3.2 1 (1.4) 37.5 37 (52.1) B7 10.0 0 (0) 8.6 3 (3.1) 52.4 27 (28.1) 29.0 66 (68.8) I3 0.3 0 (0) 0.5 0 (0) 5.4 4 (4.3) 69.7 39 (41.9) 9.6 50 (53.8) 14.6 0 (0) Study Site 30.6 (2.3) 230.3 (17.5) 289.1 (21.9) 368.0 (27.9) 25.5 (1.9) 274.2 (20.8) 101.9 (7.7
75 Table 11. Manlys Standardiz ed Selection Ratios ( B ) for Habitats Selected by Masticophis flagellum Habitats are listed in decreas ing order of habitat selection. Superscripts indicate statistically significan t differences in selection between habitats. Home Range Selection Habitat Selection Habitat B P B P Florida Scrub 0.23a 0.0038 0.59a <0.001 Flatwoods 0.13bc 0.59 0.29b 0.022 Sandhill 0.32ab 0.083 0.12bc 0.677 Cutthroat 0.077cde 0.0028 0.035c <0.001 Pasture 0.047cde 0.046 0.022cd 0.157 Swamp 0.042de 0.0023 0.003d <0.001 Marsh 0.15abcd 0.50 0.00d <0.00
76 Table 12. Summary of Models Evaluated for Habitat Selection by Masticophis flagellum K = number of parameters in the model. Models are listed in order of decreasing support. Model Deviance K AICc AICc AICc Wt. Selection varies among individuals 0 98 428 0 1.00 Constant selection of habitat 329 20 586 159 <<0.001 Selection varies with sex 476 14 722 294 <<0.001 Selection varies with size 486 14 731 303 <<0.001 No selection of habitat 2552 14 2797 2369 0.00
77 Table 13. Florida Scrub Patch Use and Availability for M. flagellum Availability (Avail) is the expressed as the percentage of each individuals home range ma de up of each Florida scrub patch. Use is the numb er of locations (percentag e of locations) in each patch. Blanks indicate patches that we re not available to the individual. A B C D E ID Avail Use Avail Use Avail Use Avail Use Avail Use B1 52.2 65 (98.5) C1 2.6 3 (2.9) 64.9 101 (97.1) 0.8 0 (0) F2 3.6 8 (14.8) 10.5 16 (29.6) 0.3 0 (0) 0.3 1 (1.9) 0.6 7 (13.0) B2 59.8 63 (98.4) B3 7.1 14 (9.5) 34.3 127 (86.4) B4 5.6 11 (21.6) B5 2.4 2 (2.3) 65.8 82 (94.3) 0.3 3 (3.4) I2 19.9 23 (69.7) B7 12.8 22 (33.3) 10.8 25 (37.9) 0.9 1 (1.5)
78 Table 13 (Continued). F G H I ID Avail Use Avail Use Avail Use Avail Use B1 3.4 1 (1.5) C1 F2 0.8 10 (18.5) 0.7 4 (7.4) 3.8 8 (14.8) B2 1.3 1 (1.6) B3 2.8 6 (4.1) B4 3.1 7 (13.7) 16.6 33 (64.7) B5 I2 6.0 10 (30.3) B7 3.5 17 (25.8) 1.0 1 (1.5)
79 Table 14. POPAN Jolly-Seber Abundance Estimates for Lizard and Mammalian Prey in Each Sampled Patch of Florida Scrub. Abundance estimates are per trap array ove r the entire study period (March 2004 June 2006). Estimates were derived from the best-fit model indicated in the table. Abundance estimates are presented as mean (standard error). Estim ates used in the examination of the effect of pr ey abundance on patch selection by M. flagellum were mean monthly abundance for the tracking period of each snake, and were lower than the study period estimates presented here. AICc wt. = relative support of model compared with other models evaluated for each species, p = capture probability, = survival probability, pent = probability of entry (birth and immigration). = constant parameter value, t = time-varying parameter, g+t = parameter varied over time in a parallel manner for each group. Patch Abundance Estimate Prey Species Model AICc wt. A B C D Anolis carolinensis ppent >0.99 21.0 (8.5) 31.7 (10.9) 42.5 (13.3) 17.4 (7.6) Aspidoscelis sexlineata ptpentg+t 0.98 67.7 (9.8) 85.7 (11.5) 92.0 (12.8) 90.4 (12.3) Sceloporus woodi ptpent 0.90 172.8 (18.9) 168.7 (18.6) 83.4 (12.2) 4.8 (2.9) Peromyscus polionotus ppentt >0.99 27.3 (3.9) 32.7 (4.9) 7.8 (1.1) 19.5 (2.8) Podomys floridanus ppentt >0.99 84.2 (12.2) 67.5 (10.4) 93.3 (13.1) 62.9 (10.0)
80 Table 14 (Continued). Patch Abundance Estimate Prey Species E F G H I Anolis carolinensis 49.7 (14.8) 46.1 (14.0) 60.4 (17.0) 49.7 (14.8) 22.8 (8.9) Aspidoscelis sexlineata 135.9 (16.0) 110.2 (14.2) 236.9 (22.6) 179.3 (18.2) 57.9 (9.5) Sceloporus woodi 2.0 (2.0) 156.5 (17.7) 0 (0) 75.2 (11.5) 122.7 (15.3) Peromyscus polionotus 0 (0) 3.9 (0.6) 27.3 (3.9) 43.8 (7.3) 60.4 (9.8) Podomys floridanus 132.9 (17.0) 117.7 (15.5) 93.3 (13.1) 59.9 (9.7) 70.5 (10.8)
81 Table 15. Summary of Models Evaluated for Florida Scrub Patch Selection by Masticophis flagellum K = number of parameters in model, Swoo Abund = Sceloporus woodi abundance, Liz Abund = lizard ( Anolis carolinensis + Aspidoscelis sexlineata + S. woodi ) abundance, Mam Abund = mammal ( Peromyscus polionotus + Podomys floridanus ) abundance. Models with interactions allow selection to vary by individual or characteristic; additive models demonstrate c onstant selection by indi vidual or characterist ic. Models are listed in order of decreasing support. Model Deviance K AICc AICc AICc Wt. Snake x Scrub 0 81 291 0 1.00 Snake + Scrub 462 17 615 324 <<0.001 Sex x Scrub 496 18 652 361 <<0.001 Size x Scrub 523 18 678 387 <<0.001 Snake x Swoo Abund 1072 18 1227 937 <<0.001 Snake + Swoo Abund 1222 10 1361 1070 <<0.001 Snake x (Liz Abund + Mam Abund) 1486 18 1659 1368 <<0.001 Snake x Mam Abund 1599 18 1754 1463 0.00 Snake + Liz Abund + Mam Abund 1689 27 1831 1540 0.00
82 Table 15 (Continued). Model Deviance K AICc AICc AICc Wt. Snake + Mam Abund 1702 10 1842 1551 0.00 Snake x Liz Abund 1755 18 1910 1619 0.00 No Selection 1813 9 1950 1659 0.00 Snake + Liz Abund 1811 11 1950 1659 0.00 Size x Mam Abund 1856 4 1984 1692 0.00
83 Table 16. Manlys Standardiz ed Selection Ratios ( B ) for Florida Scrub Patches Selected by Masticophis flagellum Patches are listed in decreasing order of selection. Superscripts indicate statistically significant differences in selection between patches. Patch was only available to one snake, so statistical significance of sele ction of this patch could not be determined. Patch B P B 0.43a 0.0083 F 0.16b 0.33 H 0.13b 0.63 A 0.11b 0.66 G 0.064b 0.38 E 0.047b 0.45 I 0.045b 0.32 C 0.012b 0.023 D* 0.012 NA
84 Table 17. Home Range Sizes and Movement Distances of Masticophis flagellum at Different Locations. Numbers in the table indicate mean (se). Mean Daily Displacement is the mean straight-line daily movement, which wa s determined by different field and/or analytical techniques in each study. Home range si ze and mean daily displacement were calculated excluding the individual tracked only through winter. Study Location 100% MCP (ha) Mean Daily Displacement (m) This study Central Florida 130.8 (28.5) 153 (10) Secor, 1995 Mojave Desert 57.9 (13.2) 146 (13) Dodd and Barichivich, 2007* North-centr al Florida 113.6 (38.5) 229 (47) Johnson et al., 2007 Eastern Texas 70.4 (83.8) 93 (NA)
85 Figure 5. Map of the Study Site within the Lake Arbuckle Tract of Lake Wales Ridge State Forest in Central Florida, USA. Th e 100% minimum convex polygon containing all snake locations is outlined by the dashed bl ack line, and focal patches of scrub are outlined and labeled in solid black.
Figure 6. Masticophis flagellum Net Squared Displacement Versus Random Walk Expectations. Observed squared displacement is represented by the dashed line, expected squared displacement is represented by the solid black line, and the 95% confidence interval is represented by dotted lines. The observed squared displacement generally lies along the x-axis. Scales on the x-axis are cons tant in columns; scales on the y-axis are constant in rows. 86
Figure 7. Masticophis flagellum 100% Minimum Convex Polygon Home Ranges. Females are indicated by solid lines, males are indicated by dotted lines. 87
Figure 8. Mosaicplot of the Frequency of Beha viors within Habitats. Habitats with fewer than 20 observations were not included in the plot. The width of bars indicates the proportion of observations within each habitat, the height of each subdivision indicates the proportion of each behavior within each habitat. 88
Figure 9. Correspondence Plot of the Associat ion of Snake Behaviors with Habitats. Elements of the plot that are located near each other demonstrate a positive association; those distant from each other dem onstrate a negative association. 89
90 A Greater Abundance of Snakes Results in Lower Survival Rates of Florida Scrub Lizard ( Sceloporus woodi ) Populations Although broad variation in the dem ography of lizards in the genus Sceloporus occurs, the mechanisms leading to this va riability are unresolved (Tinkle and Dunham 1986). For instance, phylogenetic, geographic, and habitat differences in demographic parameters exist among populations of Sceloporus undulatus (Ferguson et al. 1980), but the source of these differences is unclear Density-dependent pr edation and densityindependent stochastic environm ental factors affect temporal differences in survivorship and fecundity in a Kansas population of S. undulatus (Ferguson et al. 1980). Predation has been repeatedly suggested as a determinant of the local li fe histories of populations of S. undulatus (Crenshaw 1955, Tinkle and Dunham 1986), but much of the evidence is anecdotal. The demography of Sceloporus woodi is a particularly interesting case study. Sceloporus woodi has a very short lifespan, with a maximum recorded age of 27 months (McCoy et al. 2004). Although it can be locally abundant, S. woodi is considered rare because it occurs only in Florida, USA, and there it is precinctive to the rare and fragmented Florida scrub habitat (Jackson 1973, McCoy and Mushinsky 1992). Individuals of S. woodi disperse a maximum distance of 750 m, but dispersal movements greater than 200 m are rare. Dispersal of S. woodi is restricted to an even greater extent by the densely vegetated habitats that typica lly occur between patches of Florida scrub
91 (Hokit et al. 1999). These habitat an d dispersal characteristics of S. woodi together with the natural and anthropogeni c fragmentation of Florida scrub (McCoy and Mushinsky 1994), contribute to the existence of this sp ecies as metapopulations (Hokit et al. 2001). The mechanisms regulating S. woodi metapopulations and local populations remain unclear. Reduced survival of S. woodi in smaller patches of scrub occurs, but the mechanism for this effect is uncertain (H okit and Branch 2003b), although predation has been suggested (McCoy et al. 2004). To exam ine the roles of patch size and predation pressure on the abundance and survival rates of S. woodi I studied temporal and spatial variation in these parameters within a single metapopulation of S. woodi while quantifying the relative abundance of co-occu rring saurophagous snakes. I predicted that abundance and survival rates of S. woodi would vary temporally with its life cycle, and spatially as a function of patch size and sn ake abundance. I found that the abundance of S. woodi exhibited both temporal and spatial variat ion, and survival rates of this lizard were most strongly affected by th e relative abund ance of snakes. MATERIALS AND METHODS Study Site Research was conducted at the Lake Ar buckle tract of the Lake Wales Ridge State Forest in southeastern Polk County, Florida, USA (27.67N Latitude, 82.43W Longitude). The site consists of a series of patches of xeric Florida scrub habitat and wetlands in a matrix of more densely vege tated pine flatwoods habitat. All sampling occurred in nine neighboring patches of Florida scrub habi tat ranging in size from 1.5-
92 170 ha (Fig. 10). Sampled patches were simila r in vegetation composition and structure, and each was dominated by a low (0.5-1.5 m) midstory of shrub oak ( Quercus spp.), scrub palm ( Sabal etonia ), and saw palmetto ( Serenoa repens ). Canopy cover in the sampled locations was typically less than 10%, and primarily consisted of slash pine ( Pinus elliottii ). Groundcover in sampled locations wa s primarily bare sand with less than 50 percent leaf litter and little herbaceous vegetation. Field Methods Sceloporus woodi was sampled using 13 trap arrays (see below) installed in the patches of Florida scrub habitat. One trap array was installed in each scrub, with the exception that two trap arrays were installed in scrub I, and four trap arrays were installed in scrub B to account for potentia lly greater heterogeneity in de nsity and survival rate in larger patches (Fig. 10). Trap arrays consisted of a central box trap with a funnel on each face and four 7.6 m drift fences constructe d of 50.8 cm aluminum flashing projecting from each face of the box trap (Burgdorf et al. 2005). A single-ended funnel trap constructed of 6.35 mm hardware cloth was placed on each side of the distal end of each drift fence, and a 7.6-L bucket was placed at th e center of both sides of each drift fence for a total of 17 traps per a rray (Fig. 11). All traps were shaded with Masonite and provided with a moist sponge to preven t desiccation and overheating of trapped individuals. Trap arrays we re opened and checked daily for seven consecutive days (secondary period), then closed for 21 consecu tive days (primary period) as part of a robust capture-recapture design (Pollock 1982). Sampling occurred from March 2004 June 2006 for a total of 20 primary periods Traps were closed from December 2004
93 February 2005 and again from November 2005 March 2006 to avoid small mammal mortality caused by cold overnight temper atures. Each day traps were checked, the maximum, minimum, and current temperatur e (C) were measured with a max/min thermometer, and amount of precipitation (mm) was measured with a rain gauge at trap array A1 (Fig. 1). All captured S. woodi were sexed, measured for snout-vent length (SVL) and total length (TL) to the nearest mm, weighed to the nearest 0.1 g, given a unique identifying mark by toe clipping (W aichman 1992), and released a few meters outside the trap array. Analytical Methods Capture-recapture data for S. woodi were analyzed using the ad hoc robust design (Pollock, 1982) with Program MARK 4.3 (White, 2006). Abundance estimates were calculated for each secondary period at each trap array using Huggins closed population models (Chao and Huggins 2005). Daily captur e probabilities were assumed to be the same at all trap arrays, which were identical were placed in scr ub habitat with very similar structural attributes, and were cl ose enough to one another that each array experienced similar environmental conditions. Capture and recapture probabilities also were considered equal for two a priori reasons. First, the traps we re not baited, so lizards were unlikely to exhibit a trap-happy response. Second, liz ards that were trapped did not appear stressed relative to lizards found out side the traps, and it is unlikely that they learned to intentionally avoi d traps. Huggins closed populat ion models that I evaluated were constant (re)cap ture probabilities (p.), individual heterogeneity in (re)capture probabilities (two mixture model; phet), time-specific (re)capture probabilities ( pt), and
94 combined time and heterogeneity effects ( phet*t). In addition to thes e basic models, daily mean temperature (ptemp) and rainfall ( prain) were used as environmental covariates; the effect of these environmental vari ables on individual heterogeneity ( phet*temp and phet*rain) also was evaluated. Sex ( psex) and SVL ( pSVL) were individual characteristics used in an attempt to explain sources of heterogeneity in (re)capture probabilities and improve the estimation efficiency of lizard abundance (C hao and Huggins 2005). Two-way interactive and additive models between i ndividual covariates and time ( psex*t, psex+t, pSVL*t, and pSVL+t) or temperature (psex*temp, psex+temp, pSVL*temp, and pSVL+temp) also were evaluated. Thus, up to 18 models were considered fo r each secondary period. All models with covariates were analyzed using the logit link function, which maps the probability of a binary response variable (in this case, captured or not capture d) from [0,1] to [,+ ] (Cooch and White 2007). If no rain fell duri ng a secondary period, all models that included rainfall were eliminated from the an alysis. Additionally, models that contained more parameters than the number of individu al lizards captured dur ing that week also were not considered for analysis because data for these weeks were too sparse to estimate all parameters of these models. Thus, as few as 12 models were evaluated for some secondary periods. Survival rates were determined usi ng Cormack-Jolly-Seber (CJS) models (Nichols 2005) in Program MARK 4.3 (White, 2006). Secondary period capture histories were combined to construct primary peri od capture histories. My model set was constructed using a two step pr ocedure. Recapture probabilities in the first step were modeled as constant ( p.), time-varying ( pt), or as a linear function of mean secondary period temperature (ptemp), rainfall ( prain), or breeding season (Mar ch, April, and May) vs.
95 non-breeding season (all ot her secondary periods; pbreeding). Recapture probabilities were assumed to be the same in all patches of scrub habitat. Survival rates in the first step were modeled as constant ( .), time-varying (t), and as a function of the patch in which the individual was captured ( scrub). Interactive and additive models ( scrub*t and scrub+t, respectively) also were evaluated. Survival rate and recapture probability were included in all possible combinations, resulting in an initial set of 25 models I used the bootstrap procedure to evaluate overdisp ersion (the presence of exce ss variation in the data not explained by the binomial model) for my global model ( scrub*tpt), and based my adjustment of the variance inflation factor ( = 1.125) upon the ratio of the observed deviance to the mean of the bootstrapped deviances (Cooch and White 2006). Because only models .pt, tpbreeding, and .pbreeding received support from my data, I only used pt and pbreeding in step two of my mode l development procedure. The second step in defining my CJS model set was to build models that examined the importance of snake abundance and patch ar ea in affecting the survival rates of S. woodi The relative abundance of snakes was determined as the number of saurophagous snakes trapped in each patch, divided by the number of trap arrays in that patch. Saurophagous snake species at my site included Coluber constrictor Elaphe guttata, Lampropeltis triangulum, Masticophis flagellum and Micrurus fulvius Scrub area was determined by walking the boundary of each patch with a GPS receiver, and using the area calculated from the differentia lly-corrected rover file (Trimble GPS Pathfinder Office 2.90). Snake abundance and patch area were then used as spatia l covariates in CJS models to constrain the effect of scrub on survival rate to be a linear function of snake relative abundance ( snake) or scrub area ( area). Interactive and additive models of snake
96 abundance and patch area with time ( snake*t, snake+t, area*t, and area+t) and each other ( snake*area, snake+area) also were evaluated, resulting in a final set of 41 models. All models with covariates were analyzed using the logit link function (Cooch and White 2007). Models for all analyses were evaluated for parsimony using Akaikes Information Criterion adjusted for small sample sizes and overdispersion (QAICc). If more than one model had a QAICc value less than six (relative lik elihood > 5%), parameters were estimated using model averaging to account fo r uncertainty in model selection (Burnham and Anderson 2002). The relativ e importance of variables ( w+) was determined by calculating the sum of the Akaike weights fo r all models containi ng each variable of interest. The relative importance of variab les is dependent upon the set of models evaluated, but does not necessarily depe nd upon which particular model is most parsimonious (Burnham and Anderson 2002). RESULTS I captured a total of 503 S. woodi individuals 878 times, and 417 individual saurophagous snakes also were captured. Uncerta inty in abundance estimates was usually greater than the variation of these estimates in space and time (Fig. 12). No single model was identified in the analysis as the most parsimonious explanation for any secondary period. Indeed, the important variable(s) for determining (re)capture probabilities (an index of lizard activity) was di fferent from month to month (T able 18). Patch area did not affect the density of S. woodi per trap array no r the variation in monthly abundance estimates (Fig. 12).
97 The CJS confidence set included 12 models (Table 19; arranged from greatest to least support). Models with time-varying recapture probabilities re ceived the greatest support from the data ( w+ = 0.684); models including the effects of breeding activity on recapture probability also were supported, but less strongly so ( w+ = 0.316). Recapture probability was greatest in the spring of each year and declined through the summer (Fig. 13). Survival rate was relatively constant over time, and estimates of the probability of an individual surviving a fou r-week primary period ranged from 0.7 0.8 (Fig. 14). The most important characteristics for determin ing survival rates were snake relative abundance and patch area (Table 20). DISCUSSION The abundance of S. woodi varied over space and time, but scrub area did not appear to affect density estimates. The factor s affecting the daily (r e)capture probability of S. woodi individuals varied substantially among secondary periods. Usually, the sex of an individual did not matter for determining (re)capture probability. The importance of a particular factor likely depends upon the context of the sampli ng period with regard to the environment, lizard phenology, or size and/or sex structure of the population. For example, temperature may have a greater eff ect on lizard activity when temperatures are more extreme or variable, and rainfall may mo re influential in very dry periods. Snoutvent length might change in importance as the size structure of the population changes, particularly if different size classes differ in activity level or susceptibility to trapping. Sex might be more important during breeding or oviposition than at other periods. The large proportion (65%) of secondary periods for which environmental or individual
98 covariates were of great impor tance indicates that the use of Huggins closed models is an appropriate and desirable method of accoun ting for temporal and individual behavioral heterogeneity when estimating the abundance of lizards, particularly because these models can incorporate individual covariates. My results indicated that S. woodi survival rates exhibi ted greater variation between patches of scrub than they did over time. In particul ar, the relative abundance of snakes had a larger effect on survival rate than any other measured characteristic. The most obvious explanation for th is negative effect is snak e predation. Indeed, the two snakes most frequently encountered on this site were Coluber constrictor and Masticophis flagellum which both consume S. woodi in proportion to its abundance (Halstead et al. In review). In addition to predation, high numbers of snakes could alter the behavior of S. woodi resulting in greater risk of mortality from other causes, including alteration of competitive interactions and decreased energy intake (Werner 1991, Lima 1998). Predation, particularly by snakes is an important mortality factor for Sceloporus spp. (Crenshaw 1955, Ferguson et al. 1980, McCoy et al. 2004), but has not previously been quantified. In contrast to the effects of snake a bundance, patch area had a positive effect on the survival rate of S. woodi The positive effects of area were independent of the effect of snake abundance. Models in cluding interactive effects of snakes and area had less support than additive models or models w ith each factor considered in isolation. Therefore, the increase in surv ival rate with increasing patc h area is not solely caused by greater predation rates by snakes in smaller patches. Possibly, pred ation rates by avian or mammalian predators are greater in smaller patches of scrub (Hokit and Branch 2003b).
99 The positive effect of patch area on the survival rate of S. woodi also could be caused by many other factors, including lower abundance and changes in demographic rates associated with small populations (see Hokit and Branch, 2003a). The best-fit model in my model set i ndicated constant survival rates for S. woodi The fit of this model to the data is almost certainly a consequence of low statistical power. Recapture probabilities of S. woodi were low, and my sample size was limited to nine patches of Florida scrub, of which only six contained viable S. woodi populations. Constant survival rates across patches and over time are unlikely to be a realistic description of S. woodi survival, but rather a reflection of the limitations of my sample size and sampling methods. In exploratory ecologi cal studies such as this one, the cost of a type II error is as great as (or greater than ) the cost of a type I error (Shrader-Frechette and McCoy 1993). Thus, my conclusion that the abundance of saurophagous snakes affects survival rates of S. woodi warrants additional study of th e role of snakes in the local and metapopulation dynamics of S. woodi In conclusion, my results pr ovide quantitative evidence that the local abundance of saurophagous snakes is inversely related to survival rates of local populations of S. woodi The demography of S. woodi could be particularly a ffected by abundant, actively foraging predators such as C. constrictor and M. flagellum Optimal patch use by these wide-ranging predators may resu lt in stabilizing density-dep endent mortality of local populations of S. woodi Whether such an adaptive response by these snake species to prey densities occurs is cu rrently under investigation.
Table 18. Relative Importance ( w+) of Environmental and In dividual Characteristics for Determin ing the Daily (Re)capture Rate of Sceloporus woodi Numbers in Bold indicate the variable(s) with the greatest s upport from the data for each secondary period. models for which (re)capture probability did not vary. w+ (number of models). Secondary period for which mixture models for unexplained individual heter ogeneity in (re)captur e probability were most parsimonious. Variable Secondary Period Constant* Time Temp. Rain SVL Sex March 04 0.003 (1) 0.992 (5) 0.002 (5) NA 0.189 (5) 0.280 (5) April 04 0.041 (1) 0.016 (5) 0.841 (5) 0.050 (1) 0.222 (5) 0.379 (5) May 04 0.068 (1) 0.011 (5) 0.326 (5) 0.036 (1) 0.608 (5) 0.261 (5) June 04 0.050 (1) 0.472 (5) 0.340 (4) 0.052 (1) 0.310 (4) 0.485 (5) July 04 0.153 (1) 0.648 (3) 0.086 (3) 0.057 (1) 0.709 (5) NA August 04 0.070 (1) 0.194 (3) 0.511 (3) 0.028 (1) 0.750 (5) NA September 04 0.149 (1) 0.411 (5) 0.184 (5) 0.076 (1) 0.153 (5) 0.363 (5) October 04 0.055 (1) 0.002 (3) 0.893 (5) 0.029 (1) 0.039 (5) 0.841 (2) November 04 0.330 (1) 0.011 (2) 0.334 (3) 0.199 (1) 0.277 (4) NA 100
101 Table 18 (Continued). Variable Secondary Period Constant* Time Temp. Rain SVL Sex March 05 0.057 (1) 0.007 (1) 0.006 (1) 0.897 (1) NA NA April 06 0.014 (1) 0.809 (5) 0.064 (5) 0.006 (1) 0.747 (5) 0.144 (5) April 05 0.098 (1) 0.014 (5) 0.260 (5) 0.554 (1) 0.102 (5) 0.107 (5) May 05 0.098 (1) 0.076 (5) 0.495 (5) 0.250 (1) 0.306 (5) 0.155 (5) May 06 0.110 (1) 0.186 (5) 0.618 (5) NA 0.237 (5) 0.258 (5) August 05 0.184 (1) 0.041 (5) 0.552 (5) 0.067 (1) 0.166 (5) 0.387 (5) June 05 <0.001 (1) 0.796 (5) 0.036 (5) 0.168 (1) 0.232 (5) 0.197 (5) October 05 0.088 (1) 0.694 (5) 0.111 (5) 0.033 (1) 0.493 (5) 0.132 (5) July 05 0.331 (1) 0.029 (5) 0.260 (5) 0.128 (1) 0.193 (5) 0.203 (5) September 05 0.220 (1) 0.023 (5) 0.471 (5) 0.127 (1) 0.279 (5) 0.179 (5) June 06 0.336 (1) 0.004 (5) 0.280 (5) 0.129 (1) 0.193 (5) 0.218 (5)
102 Table 19. Support of Cormack-Jolly-Seber Models for Sceloporus woodi Populations. Models are listed in decreas ing order of support, and only those models with a relative likelihood greater than 0.05 are included in the table. Rel. Likelihood = likelihood of the model relative to the most parsimonious mode l. K = number of parameters in model. = 1.125. Model Rel. Likelihood K QAICc QAICc QAICc Wt. .pt 1.000 20 1072.494 0.000 0.222 snakept 0.697 21 1073.216 0.722 0.155 area+snakept 0.679 22 1073.269 0.775 0.151 areapt 0.435 21 1074.161 1.667 0.097 tpbreeding 0.419 21 1074.235 1.741 0.093 snake+tpbreeding 0.310 22 1074.837 2.343 0.069 area*snakept 0.251 23 1075.258 2.764 0.056 .pbreeding 0.164 3 1076.109 3.615 0.036 area+tpbreeding 0.158 22 1076.190 3.696 0.035 area+snakepbreeding 0.127 5 1076.618 4.124 0.028 snakepbreeding 0.117 4 1076.792 4.298 0.026 areapbreeding 0.074 4 1077.713 5.219 0.016
103 Table 20. Relative Importance ( w+) of Variables for Determining the Survival Rate of Sceloporus woodi Populations. + indicates a variab le that increases survival rate, indicates a variable that decreases survival rate. Variable No. Models w+ Direction of Influence* Snake Relative Abundance 10 0.496 Scrub Area 10 0.394 + Constant Survival Rate 2 0.259 NA Time 14 0.198 NA Patch of Scrub 6 0.004 NA
Figure 10. Map of the Study Site within the Lake Arbuckle Tract of Lake Wales Ridge State Forest in Central Florida. Individual patches of scrub are outlined in solid black, and the location and identity of trap arrays within each scrub is indicated. 104
Figure 11. Schematic Diagram Depicting the La yout of Each Trap Array. The central square is a box trap. The horiz ontal and vertical solid lines represent four drift fences ending at funnel traps represented by rect angles. The circles along the drift fences represent pitfall traps. The triangles in th e box and funnel traps indi cate the location of funnels allowing entrance of animals into the traps. 105
Figure 12. Model-averaged Estimates of Abundance for Sceloporus woodi Populations at Each Trap Array Grouped by Patch Size. (A) Tr ap arrays in scrub B (170 ha). (B) Trap arrays in scrub I (40 ha). (C) Trap arra ys in scrub A and scrub H (13 and 9 ha, respectively). (D) Trap arrays in scrub C a nd scrub F (1.5 and 3.0 ha, respectively). Error bars overlap for most trap arrays for most months, and have been omitted for clarity of presentation. 106
Figure 13. Model-averaged Estimates of R ecapture Probability for Each Secondary Period for Sceloporus woodi Populations. Recapture rate was assumed equivalent for each local population. Error bars in dicate 95% confidence intervals. 107
Figure 14. Model-averaged Estimates of Surv ival Rate for Each Primary Period for Sceloporus woodi Populations. Error bars indicate 95% confidence intervals. 108
109 Conclusion: The Predicted Effects of Snake Predation upon Sceloporus woodi Populations Perhaps the most fundamental question one must ask when examining the population ecology of rare species is: What mechanisms drive the dynamics of this population? The answer to this question is essential for informing conservation and management practices. Superficially attempting to increase fecundity or survival rates without addressing the root causes of decline will not aid in the recovery of populations. Undoubtedly, the occurrence of Florida scrub-precinctive Sceloporus woodi (Florida Scrub Lizard) has been reduced by the loss of scrub habitat (Jackson 1973, Myers 1990, McCoy and Mushinsky 1992). The focus of my study was not on broad-scale questions of occurrence, but rather was the importanc e of factors influencing the demography of S. woodi in the remaining patches of Florida sc rub habitat where this species persists. Chapters 2-4 of this dissertation have endeavored to examine one specific hypothesis: Snakes, particularly wide-ranging Masticophis flagellum (Coachwhip), have the potential to regulate local populations of S. woodi A very important consideration when examining the consequences of predation for prey populations is the dietary breadth of the predator (Ryall and Fahrig 2006). I found that both M. flagellum and the closely related, but more abundant, Coluber constrictor (Racer) prey upon S. woodi Masticophis flagellum specializes upon lizards and mammals, but consumes prey opport unistically within these categories. Coluber
110 constrictor is a generalist predator of anurans, lizards, and mammals, but within the amphibian category selects certain species upon which to forage. Both predators consume S. woodi in proportion to its availability at fine temporal and spatial scales. Because neither predator is dependent solely upon S. woodi for growth and reproduction, the impact of these snake species upon S. woodi should be consistent with models that consider a generalist predator. One of the most basic and well-supported observations of predator-prey interactions is that genera list predators, unlike specia lists, do not depend upon a single prey species to persist or increase in abundance. Thus, the numerical response of M. flagellum is independent of the abundance of S. woodi The impacts of M. flagellum upon S. woodi populations will depend, however, upon th e functional respon se that best describes this interaction. Generalist pred ators can exhibit one of several functional responses. One of the most widely recogni zed functional responses of generalist predators is the Holling type III functiona l response (Murdoch 1969). The type III functional response is sigmoid in shape, and suggests that consumption of prey exaggerates differences in th e availability of different prey species (Murdoch 1969). Consumption of prey in proportion to their av ailability does not imply switching behavior and is not consistent with a type III functi onal response. Rather, in a type III functional response, a species at low relative densit ies is not consumed, but at high relative densities, it is nearly always consumed (Murdoch 1969). The type III functional response can result from a number of mechanisms, includi ng prey residing in different habitats, the formation of a search image, preference for recently consumed prey, and training that increases prey capture probability (but not solely prey handling time (Murdoch 1969,
111 Oaten and Murdoch 1975, Murdoch 1977, Mc Nair 1980)). Any or all of these mechanisms may occur in the M. flagellum:S. woodi system, but the first is likely only for a few alternative prey species, such as Sigmodon hispidus (Cotton Rat). Regardless of the mechanism underlying the type III functiona l response, it tends to be stabilizing (by reducing the amplitude of limit cycles and therefore increasing the persistence of the predator-prey interaction) because prey speci es occurring at low abundances are excluded from the predators diet. Therefore, rare prey experience little predation and their population sizes are allowed to recover. Another major class of generalist pred ation models does not make explicit reference to the functional response, except that it is not type III. These models involve incidental predation, which is defined as c onsumption of prey disc overed while searching for other prey, but which do not affect predator behavior (Vickery et al. 1992). Incidental predation has little to no impact on the predat or population, but it can have a large impact upon rare prey populations. In th ese models, predators do not switch to alte rnative prey, but rather consume incidental prey whenev er it is encountered. In contrast to the stabilizing effects of switching predators, incidental predation results in apparent competition (Holt and Lawton 1994), which can le ad to the extirpati on or extinction of rare prey. Apparent competition can resu lt from a predators numerical response (predator abundance is a functi on of total prey density), f unctional response (predators concentrate foraging effort at areas of highest total prey density), or aggregative movement, but in all cases the predators pr imary prey and incident al prey must occur together (Schmidt and Whelan 1998). Because the capture rate of incidental prey is a function of the search time for the predato rs primary prey, greater abundance of total
112 prey will lead to greater predation pressure on the incidental prey (Schmidt and Whelan 1998). This density independence (with respect to the incidentally-consumed prey) tends to destabilize the interacti on, and in particular can decr ease the expected time to extinction for rare prey (Sch midt and Whelan 1998). Although M. flagellum almost certainly actively searches for S. woodi incidental predation upon S. woodi by M. flagellum is likely under certain conditions in this system (see below). Incidental predation models have been successfully applied to Procyon lotor (Raccoon) predation upon bird nests (Schmidt and Whel an 1998, Schmidt et al. 2001) and Peromyscus leucopus (White-footed Mouse) predation upon Lymantria dispar (Gypsy Moth) pupae (Goodwin et al. 2005). Incidental predation does not necessarily require the response of a predator to the density of alternative prey. If predators sel ect certain habitats or habitat features for resting, hiding, basking, or reproducing, prey residing in these areas may be consumed incidentally. For example, nest predation ra tes are higher near forest edges (Andren and Angelstam 1988), but the use of forest edges by an ophidian nest predator, Elaphe obsoleta (Black Rat Snake), is related to ther moregulation rather than prey abundance (Blouin-Demers and Weatherhead 2001b, a, Carfagno et al. 2006, Carfagno and Weatherhead 2006). Thus, incidental predation can have large impacts on prey populations wherever predators congregate, regardless of the mechan ism responsible for increased predator abundance. In addition to type III functional res ponses and incidental predation models, several other models may have importan t implications for the consumption of S. woodi by M. flagellum In some cases, prey vulnerability may be more important than prey
113 abundance (Quinn and Cresswell 2004). Fo r example, if it is difficult for M. flagellum to catch Aspidoscelis sexlineata (Six-lined Racerunner) or subdue Podomys floridanus (Florida Mouse), it may consume S. woodi because it is more vulnerable, rather than more abundant, prey. Although prey vulnerability is not likely to be a factor if the availability and predator success rate when hunting alternative prey is high (Quinn and Cresswell 2004), decreased coupling of predator consumption with prey abundance is likely to reduce abundances of vulnerable pre y. A relatively recent functional response model, the steady state satiation model, disti nguishes prey handling from prey digestion (Jeschke et al. 2002). This model may be pa rticularly appropriate for snakes, because handling prey necessarily prevents further pr ey capture, but the process of digestion likely affects the probability that the predator searches for additional prey (Jeschke et al. 2002). Because M. flagellum is digestion-limited, rather than handling-limited, foraging time is predicted to decrease with an incr ease in prey abundance (Jeschke et al. 2002). Therefore, this model presents a potential mechanism for the observed prevalence of resting behaviors by M. flagellum in scrub. In addition to it s behavioral predictions, a functional response based upon digestion lim itation would likely re duce predation rates when mammals (which are larger and have a lower surface area: volume ratio) are consumed, which in turn woul d decrease predation rates upon S. woodi Optimal diet selection by M. flagellum based on energetic considerations is can increase the persistence time of the pred ator-prey system, but suggests that the profitability of alternative prey will have a strong influence on the persistence of the M. flagellum : S. woodi interaction (Fryxell and L undberg 1994). The more profita ble the alternative prey, the less stable the system and the shorte r the persistence time (Fryxell and Lundberg
114 1994, van Baalen et al. 2001); this effect is similar to going from a type III functional response at low alternative prey profitability to incidental predati on at high alternative prey profitability (van Baalen et al. 2001). Suboptimal switching and variation between individuals in switching rules increase stab ility by decreasing the amplitude of fluctuations in predator and prey abundan ces (Fryxell and Lundberg 1994, van Baalen et al. 2001). Regardless of the conditions for st ability, the presence of alternative prey nearly always increases the pe rsistence of the system, as l ong as they are included in the diet (Fryxell and Lundberg 1994, van Baalen et al. 2001). The consequence of generalist predation upon S. woodi by M. flagellum and C. constrictor depends upon which of the above models most accurately describes the predation process. The active foraging mode of these predators increases the likelihood that they will regulate popul ations of their prey, particul arly if prey are consumed opportunistically (Rosenheim and Corbett 2003) My observation that neither predator forages selectively upon lizard or mammal speci es (as opposed to these prey categories) are counter-indicative of a type III functiona l response (Murdoch 1969); however, weak preferences actually promote switching more readily than strong preferences (Murdoch 1969, Oaten and Murdoch 1975, McNair 1980). Ther efore, it is nearly impossible to evaluate the likelihood of a t ype III functional response base d purely on field studies of diet selection. Incidental predation likely describes the consumption of S. woodi by M. flagellum in situations where local population densities of S. woodi are low, but the prevalence of S. woodi in the diet of M. flagellum suggests that these predators likely search for S. woodi or, at the very least, lizar ds. The functional response of M. flagellum preying upon S. woodi is probably situation-specific, with stabilizing switching occurring
115 in cases where alternative prey occur in different habitats (M urdoch 1977) or when a specific prey species is abundant enough to effectively train M. flagellum to search for or capture it (Murdoch 1969, Oaten and Mu rdoch 1975, McNair 1980), and small so that M. flagellum must forage frequently enough for positive reinforcement of the search image or capture efficiency. In contrast, incidental pred ation could deplete S. woodi populations in patches where S. woodi is rare, but several alte rnative prey species are equitably distributed and prevent training (Vickery et al. 1992, Schmidt and Whelan 1998). Additional research is required to determine the appropriate functional response(s), and the conditions under wh ich it (they) occur in this system. A particularly salient feature of the M. flagellum:S. woodi system is the patchy distribution of scrub habitat on my site. This is an important characteristic because S. woodi is restricted to Florida scrub (Jack son 1973, McCoy and Mushinsky 1992) and disperses poorly through other habitats (Hokit et al. 1999). Masticophis flagellum also positively selects Florida scrub habitat, but easily disperses thr ough the habitat matrix. Taken together, these features of my study system indicate that predator-prey metapopulation models and models of patc h choice by predators are appropriate for examining the potential dynamics of this system. Although both species are scrub specialists, I primarily consider models of patch choice by predators because of the very different dispersal abilities of M. flagellum and S. woodi The ideal free distribution (IFD) describe s the optimal distribution of predators such that each predator maximizes its encoun ter rate with prey (B ernstein et al. 1988). Thus, the IFD serves as a useful null model when examining the consequences of predator behavioral de cisions (Kacelnik et al. 1992). In addition to serving as a null
116 model, if a constant number of predators distribute themselves according to the IFD, prey mortality is likely to be density-dependent and, therefore, the interaction will be stable, with predation rates greater in patches with greater prey densities, causing asynchronous prey cycles among the patches (Bernstein et al. 1988). Predators are likely to achieve the IFD if they have a good estimate of global pr ey density, they are able to sample the environment, they learn faster than they depl ete prey, and the environmental grain is fine relative to predator movements (Bernstein et al. 1988, 1991). In contrast, if the cost of travel between patches is high (because of distance or mort ality rates), if the predators underestimate average global prey density, if pr ey depletion occurs faster than predator learning, or if the environment is coarse-grained relative to predator movements, the distribution of predators dive rges from the IFD (Bernstein et al. 1988, 1991). The greater the divergence from the IFD, the less stable the system with a constant number of predators, because predation rates are not linked to prey density and can be high in patches with low prey abundance (Bernstein et al. 1988, 1991). Over multiple predator generations under the IFD, predator migration does not affect the stability properties (i.e., amplitude of population fluctuations) of the system (Bernstein et al. 1999). In contrast to constant predator numbers, pred ator limitations that prevent the predators from achieving the IFD promote system stability in the population dynamic model (Bernstein et al. 1999). In particular, lower migration rates increase the amount of asynchrony in the system and promote persistence (Bernstein et al. 1999). The positive selection of scrub patches with greater liza rd abundance suggests that M. flagellum may be distributed according to the IFD in relation to prey dens ity, which is stabilizi ng over the short term but destabilizing over the long term. The inte rplay between within-patch prey population
117 growth rates, predator movement rates, and the number of patches in the system makes predicting long-term outcomes based on IFD mo dels very uncertain (Bernstein et al. 1999). Additional simulation models have examin ed the consequences of optimal patch selection under different migra tion scenarios. Adaptive local dispersal, in which predators leave a patch according to the marginal value theorem (Charnov 1976) and are distributed only among neighboring patches, reduces the am plitude of prey population fluctuations more than adaptive global or fixed di spersal rules (Fryxell and Lundberg 1993). Imperfect leaving rules further reduce the amp litude of these fluctuations (Fryxell and Lundberg 1993). Both adaptive local dispersal and imperfect leavi ng rules are likely migration scenarios for real predator s, and should lend stability to the M. flagellum:S. woodi system. In addition to the effects of dispersal rules, the migrati on rate of predators can be critical. Asynchrony (and therefore, stability) is enhanced when predator migration rates are neither too great nor too sm all, and the greater the number of patches, the more likely the occurre nce of asynchronous dynamics (Jansen 2001). Yet another model considers the effect of a predato rs time budget on movement, which results in prey density-dependent migr ation of the predator (Huang and Diekmann 2001). In this case, predators are reluctant to leave areas of high prey density, which reduces the stabilizing effect of spatial heterogene ity (Huang and Diekmann 2001). Prey-density dependent migration is arguably a more realistic description of predator movements than diffusion, and demonstrates that caution may be warranted in examining the stabilizing effects of spatial heteroge neity (Huang and Diekmann 2001).
118 The above models considered one type of prey, and may not be directly applicable to the M. flagellum:S. woodi system. Despite this limitation, these models are likely a useful description of the movement rules of M. flagellum in response to total prey densities. Because the most co mmon prey in the diet of M. flagellum occur in scrub, it may be appropriate to model the dynamics of this system by substituting total prey density for individual prey sp ecies density in the above models to model the movement decisions and distribution of M. flagellum between patches, a nd model within-patch dynamics using an appropriate model of ge neralist predation. Laboratory microcosm studies involving apparent competition, a likely scenario in my system, have demonstrated that the presence of an appare nt competitor can reduce the persistence time of the system (Bull et al. 2006). In particul ar, the importance of habitat (or landscape) configuration for pairwise species interactions is greatly reduced in complex multispecies interactions (Bull et al. 2006). Thus, th e expected spatial stabilization of the M. flagellum:S. woodi system may not occur in the presence of alternative prey. Regardless of the functional respon se and movement rules employed by M. flagellum saurophagous snakes have a negative impact on within-patch survival rates of S. woodi Although my most parsimonious model included constant surviv al rates, rather than any spatial or temporal variation in this parameter, the selection of this model is most likely a reflection of insu fficient data to detect the true spatial differences in survival rates. Regardless of which model wa s selected as most parsimonious, however, I found that snake abundance was more important for determining survival rate than patch area, despite the latter variab le being reported as the most influential variable for differences in S. woodi survival rates in previous studies (Hokit and Branch 2003b).
119 Snakes can have strong negative impacts on prey populations (Savidge 1987, Rodda and Fritts 1992, Calsbeek and Sinervo 2002), and my observation of reduced survival rates of S. woodi in patches with abundant snake pr edators, taken with the documented occurrence of S. woodi in snake diets, indicates that pr edation by snakes is sufficient to reduce survival rates of S. woodi at the spatial scale of the patch. The foraging and spatial ecology of M. flagellum taken together with the role of snake abundance in depressing survival rates of S. woodi local populations, leads to the following predictions for this patchy predato r-prey system. Despite the likely occurrence of some incidental predation on S. woodi by M. flagellum it is unlikely that S. woodi which is abundant both in many patc hes of scrub and in the diet of M. flagellum is generally an incidental prey item. Instead, M. flagellum likely switches among prey categories, if not prey species, in some circumstances. For example, M. flagellum is diurnal (Ernst and Barbour 1989) and lik ely preys upon rodents while they are underground in burrows. Relatively few diurna lly-active lizards (which constitute all lizards found in the diet of M. flagellum at my site) would be e xpected to be found in rodent burrows during the day, which would result in training, or switching behavior, that stabilizes the system. Switching may be less common among the smaller size classes of M. flagellum that are restricted to consuming li zards because of gape limitations, and accounting for size structure in the M. flagellum population could result in a more complex interaction. Another stabilizing mechanis m that is likely to lead to persistence of this system is suboptimal diet or pa tch choice (Fryxell a nd Lundberg 1993, 1994, van Baalen et al. 2001). Masticophis flagellum is unlikely to perfectly estimate prey capture rate or global prey availability, either in total or for any individual prey species. An
120 additional mechanism leading to increased persistence may be the influence of additional trophic levels on th e interaction of M. flagellum with its prey. In particular, if predators control the population of M. flagellum it may never reach great enough abundance to threaten S. woodi populations with extirpation. The opportunity for diffuse interactions in this food web containing dietary generali sts would also increa se the likelihood of persistence (McCann et al. 1998). In summary, I expect that the di etary generalist and habitat specialist M. flagellum is unlikely to extirpate S. woodi at more than a small local patch scale. The above predictions must be tempered with caution, however, because they assume a static landscape of available Flor ida scrub. Habitat loss and fragmentation can have drastic effects on the outco me of interspecific interac tions (Ryall and Fahrig 2006). In particular, this system is characterized by a precinctive prey organism preyed upon by a dietary generalist, but habita t specialist, predator. Models of the effects of habitat loss on metacommunity structure indica te several different potential effects of habitat loss in this system. First, because it is a dietary generalist, M. flagellum likely causes some degree of apparent competition among its shar ed prey. Generalist pr edators are favored over specialist predators when habitat is destroyed (Swihart et al. 2001, Melian and Bascompte 2002, Ryall and Fahrig 2006), but becau se it is a habitat specialist and most of its prey are more abundant in scrub than in other habitats, dietary generality may not be an advantage for M. flagellum In particular, because it is a habitat specialist, loss of S. woodi habitat is also loss of M. flagellum habitat. If M. flagellum depends upon scrub habitat for persistence, as my habitat selection and home range analyses suggest, it is likely to go extinct at a lower level of habitat loss than S. woodi (Swihart et al. 2001,
121 Melian and Bascompte 2002, Ryall and Fahr ig 2006). Following the extinction of M. flagellum the dynamics of S. woodi are uncertain. If predation by M. flagellum is more strongly limiting to S. woodi than resource abundance, S. woodi could increase in abundance following the extinction of M. flagellum (Bascompte and Sole 1998, Prakash and de Roos 2002, Ryall and Fahrig 2006). The e ffects of habitat fragmentation per se is expected to decrease the persistence of the system because M. flagellum has much greater dispersal abilities than S. woodi and can lower survivorship (and increase rates of extinction) of S. woodi in patches of scrub (Ryall a nd Fahrig 2006). In summary, the effects of habitat loss are likely to have a more adverse effect on populations of M. flagellum than populations of S. woodi but the opposite may be true for habitat fragmentation per se. Interestingly, in the face of habitat loss, C. constrictor is a more likely candidate than M. flagellum to cause the extirpa tion and extinction of S. woodi Three primary mechanisms underlie this hypothesis. First, like M. flagellum, C. constrictor is a dietary generalist. Unlike M. flagellum however, C. constrictor consumes prey, particularly amphibians, found in abundance in pine flatwoods and wetland habitats. The prey base of C. constrictor is therefore independent of scrub ha bitat (Schmidt et al. 2001, Swihart et al. 2001, Ryall and Fahrig 2006). Second, because it is found in a greater variety of habitats (particularly pine fl atwoods and disturbed areas), C. constrictor can persist at high abundances despite the loss and fragmentation of scrub habitat (Schmidt and Whelan 1998, Ryall and Fahrig 2006). A particular problem associated with this trait is that incursions by C. constrictor into scrub from other habita ts can cause an increase in the critical patch size required for the establishment and persistence of S. woodi
122 populations, provided C. constrictor does not exhibit a type III functional response (Cantrell et al. 2001) Finally, because C. constrictor is smaller and moves less than M. flagellum it may experience lower predation rates by visually-oriented predators (such as raptors) and become locally abundant, re sulting in strong top-down effects on S. woodi populations (Rosenheim and Corbett 2003). Adaptable generalist species such as C. constrictor are thus selected under conditions of anthropogenic habitat fragmentation and disturbance, and can have st rong negative effects on rare pr ey populations (Schmidt and Whelan 1998, Ryall and Fahrig 2006). In conclusion, the interaction of M. flagellum with S. woodi may be locally unstable, but is likely to be pers istent. The functional response of M. flagellum toward S. woodi is likely to differ depending upon the rela tive abundances of prey in a patch, with stabilizing type III functional responses comm on in large snakes that consume mammals and when relatively few alternative prey sp ecies are numerically dominant. In contrast, when S. woodi is rare and alternative prey more equitably distributed, training is less likely to occur and destabilizing incidental pr edation more likely. The patchy distribution of scrub is also likely to enha nce persistence, particularly wh en a large number of patches are distributed at varying distances from each other. Although M. flagellum is likely to be extirpated at lower leve ls of habitat loss than S. woodi the effects of anthropogenic changes could have grave effects for S. woodi by increasing predation rates by dietary and habitat generalist predators such as C. constrictor .
123 Literature Cited Andheria, A. P., K. U. Karanth, and N. S. Kumar. 2007. Diet and pr ey profiles of three sympatric large carnivores in Bandipur Ti ger Reserve, India. Journal of Zoology 273:169-175. Andren, H., and P. Angelstam. 1988. Elevated pred ation rates as an edge effect in habitat islands: Experimental evidence. Ecology 69:544-547. Andrews, K. M., and J. W. Gibbons. 2005. How do highways influence snake movement? Behavioral responses to roads and vehicles. Copeia 2005:772-782. Arnold, S. J. 1993. Foraging theory and prey-siz e-predator-size relations in snakes. Pages 87-115 in R. A. Seigel and J. T. Collins, editors. Snakes: Ecology and Behavior. Blackburn Press, Caldwell, New Jersey, USA. Ashton, R. E., Jr., and P. S. Ashton. 1981. Handbook of Reptiles and Amphibians of Florida. Part One: The Snakes. Windwa rd Publishing, Inc., Miami, Florida. Bascompte, J., and R. V. Sole. 1998. Effects of habitat destructi on in a prey-predator metapopulation model. Journal of Theoretical Biology 195:383-393. Bernstein, C., P. Auger, and J. C. Poggiale. 1999. Predator migration decisions, the ideal free distribution, and predator-prey dynamics. American Naturalist 153 :267-281. Bernstein, C., A. Kacelnik, and J. R. Kr ebs. 1988. Individual decisions and the distribution of predators in a patchy environment. Journal of Animal Ecology 57:1007-1026. Bernstein, C., A. Kacelnik, and J. R. Kr ebs. 1991. Individual decisions and the distribution of predators in a patchy environment .2. The influence of travel costs and structure of the environment. Journal of Animal Ecology 60:205-225. Blouin-Demers, G., and P. J. Weatherhead. 2001a. An experimental test of the link between foraging, habitat selection and thermoregulation in black rat snakes Elaphe obsoleta obsoleta Journal of Animal Ecology 70:1006-1013. Blouin-Demers, G., and P. J. Weatherhea d. 2001b. Habitat use by black rat snakes ( Elaphe obsoleta obsoleta ) in fragmented forests. Ecology 82 :2882-2896.
124 Bonnet, X., G. Naulleau, and R. Shine. 1999. The dangers of leavi ng home: dispersal and mortality in snakes. Biological Conservation 89:39-50. Brodie, E. D., Jr., B. J. Ridenhour, and E. D. Brodie, III. 2002. The evolutionary response of predators to dangerous prey: Hotspots and coldspots in the geographic mosaic of coevolution between garter snakes and newts. Evolution 56 :2067-2082. Bull, J. C., N. J. Pickup, M. P. Hassell, and M. B. Bonsall. 2006. Habitat shape, metapopulation processes and the dynamics of multispecies predator-prey interactions. Journal of Animal Ecology 75:899-907. Burgdorf, S. J., D. C. Rudolph, R. N. Connor, D. Saenz, and R. R. Schaefer. 2005. A successful trap design for capturing larg e terrestrial snakes. Herpetological Review 36:421-424. Burnham, K. P., and D. R. Anderson. 2002. M odel Selection and Mu ltimodel Inference: A Practical Information-Theoretic Appr oach, 2nd edition. Springer, New York, New York. Calenge, C. 2006. The package "adehabitat" for th e R software: a tool for the analysis of space and habitat use by anim als. Ecological Modelling 197:516-519. Calsbeek, R., and B. Sinervo. 2002. An experimental test of the ideal despotic distribution. Journal of Animal Ecology 71:513-523. Cantrell, R. S., C. Cosner, and W. F. Fagan. 2001. How predator incursions affect critical patch size: The role of the functional response. American Naturalist 158 :368-375. Carfagno, G. L. F., E. J. Heske, and P. J. Weatherhead. 2006. Does mammalian prey abundance explain forest-edge use by snakes? Ecoscience 13:293-297. Carfagno, G. L. F., and P. J. Weatherhead. 2006. Intraspecific and in terspecific variation in use of forest-edge habitat by snak es. Canadian Journal of Zoology-Revue Canadienne De Zoologie 84:1440-1452. Carr, A. F. 1940. A contribution to the herpet ology of Florida. University of Florida Publication, Biological Science Series 3:1-118. Chao, A., and R. M. Huggins. 2005. Modern closed-population capture -recapture models. Pages 68-87 in S. C. Amstrup, T. L. McDonal d, and B. F. J. Manly, editors. Handbook of Capture-Recapture Analysis. Princeton University Press, Princeton, New Jersey. Charnov, E. L. 1976. Optimal foraging, the marginal value theorem. Theoretical Population Biology 9:129-136.
125 Conant, R., and J. T. Collins. 1998. A Field Guide to Reptiles and Amphibians: Eastern and Central North America, Third ed ition. Houghton Mifflin Company, Boston, Massachusetts. Cooch, E., and G. C. White. 2006. Program Mark: A Gentle Introduction, 5th edition. Cooch, E., and G. C. White. 2007. Program Mark: A Gentle Introduction, 6th edition. Crenshaw, J. W., Jr. 1955. The life hist ory of the southern spiny lizard, Sceloporus undulatus undulatus Latrielle. American Midland Naturalist 54 :257-298. Cundall, D. 1987. Functional morphology. Pages 106-140 in R. A. Seigel, J. T. Collins, and S. S. Novak, editors. Snakes: Ecology and Evolutionary Biology. Macmillan Publishing Co., New York, New York, USA. Cundall, D., and H. W. Greene. 200 0. Feeding in snakes. Pages 293-333 in K. Schwenk, editor. Feeding: Form, Function, and Evolution in Tetrapod Vertebrates. Academic Press, San Diego, California, USA. Daly, J. W., C. W. Myers, and N. Whittaker. 1987. Further classificati on of skin alkaloids from neotropical poison frogs (Dendroba tidae), with a general survey of toxic/noxious substances in the Amphibia. Toxicon 25:1023-1095. Desfilis, E., E. Font, and F. Guillen-Salazar. 2003. Stimulus control of predatory behavior by the Iberian wall lizard ( Podarcis hispanica Sauria, Lacertidae): Effects of familiarity with prey. Journal of Comparative Psychology 117 :309-316. Dodd, C. K., Jr., and W. J. Barichiv ich. 2007. Movements of large snakes ( Drymarchon, Masticophis ) in north-central Florida. Florida Scientist 70:83-94. Downes, S. J. 2002. Size-dependent predation by snakes: selective foraging or differential prey vulnerab ility? Behavioral Ecology 13:551-560. Ernst, C. H., and R. W. Barbour. 1989. Snakes of Eastern North America. George Mason University Press, Fairfax, Virginia. Fahrig, L., and G. Merriam. 1994. Conservati on of fragmented popul ations. Conservation Biology 8:50-59. Ferguson, G. W., C. H. Bohl en, and H. P. Woolley. 1980. Sceloporus undulatus: Comparative life history and regulat ion of a Kansas population. Ecology 61 :313322. Fitch, H. S. 1963. Natural history of the Racer Coluber constrictor. University of Kansas Publications Museum of Natural History 15:351-468.
126 Fitch, H. S., and H. W. Shirer. 1971. A radiotel emetric study of spa tial relationships in some common snakes. Copeia 1971:118-128. Ford, N. M., and G. M. Burghardt. 1993. Perceptual mechanisms and the behavioral ecology of snakes. Pages 117-164 in R. A. Seigel and J. T. Collins, editors. Snakes: Ecology and Behavior. Blackburn Press, Caldwell, New Jersey, USA. Fryxell, J. M., and P. Lundberg. 1993. Optim al patch use and metapopulation dynamics. Evolutionary Ecology 7:379-393. Fryxell, J. M., and P. Lundberg. 1994. Di et choice and predator-prey dynamics. Evolutionary Ecology 8:407-421. Garton, J. D., and H. R. Mushinsky. 1978. Eviden ce for skin toxicity and unpalatability in Gastrophryne carolinensis Anura Microhylidae. American Zoologist 18 :583. Gillingham, J. C. 1987. Social behavior. Pages 184-209 in R. A. Seigel, J. T. Collins, and S. S. Novak, editors. Snakes: Ecolog y and Evolutionary Biology. Blackburn Press, Caldwell, New Jersey. Goodwin, B. J., C. G. Jones, E. M. Schauber, and R. S. Ostfeld. 2005. Limited dispersal and heterogeneous predation ri sk synergistically enhance persistence of rare prey. Ecology 86:3139-3148. Greenbaum, E. 2004. The influence of prey-scen t stimuli on predatory behavior of the North American copperhead Agkistrodon contortrix (Serpentes : Viperidae). Behavioral Ecology 15:345-350. Greenberg, C. H., D. G. Neary, and L. D. Ha rris. 1994. Effect of high-intensity wildfire and silvicultural treatments on rept ile communities in sand-pine scrub. Conservation Biology 8:1047-1057. Greene, H. W. 1983. Dietary correlates of the origin and radiation of snakes. American Zoologist 23 :431-441. Greene, H. W., and J. A. Rodriguez-R obles. 2003. Feeding ecology of the California Mountain Kingsnake, Lampropeltis zonata (Colubridae). Copeia 2003:308-314. Gregory, P. T., and L. A. Isaac. 2004. Food habits of the Grass Snake in southeastern England: Is Natrix natrix a generalist predator ? Journal of Herpetology 38 :88-95. Gregory, P. T., J. M. Macartney, and K. W. Larsen. 1987. Spatial patterns and movements. Pages 366-395 in R. A. Seigel, J. T. Collins, and S. S. Novak, editors. Snakes: Ecology and Evolutionary Biology. Macmillan Publishing Co., New York, New York, USA.
127 Halstead, B. J., H. R. Mushinsky, an d E. D. McCoy. In review. Sympatric Masticophis flagellum and Coluber constrictor select vertebrate prey at different levels of taxonomy. Hamilton, W. J., and J. A. Pollack. 1956. The food of some colubrid snakes from FortBenning, Georgia. Ecology 37:519-526. Hartmann, P. P. 1993. Demography of a popul ation of the Florida scrub lizard ( Sceloporus woodi ) in a sand pine scrub on the Lake Wales Ridge of central Florida. University of South Florida, Tampa, Florida. Heard, G. W., D. Black, and P. Robertson. 2004. Habitat use by the inland carpet python ( Morelia spilota metcalfe i : Pythonidae): Seasonal relationships with habitat structure and prey distribution in a rural landscape. Austral Ecology 29:446-460. Hokit, D. G., and L. C. Branch. 2003a. Associ ations between patch area and vital rates: Consequences for local and regional populations. Ecological Applications 13:1060-1068. Hokit, D. G., and L. C. Branch. 2003b. Habita t patch size affects demographics of the Florida scrub lizard ( Sceloporus woodi ). Journal of Herpetology 37:257-265. Hokit, D. G., B. M. Stith, and L. C. Branch. 1999. Effects of landscape structure in Florida scrub: A population perspective. Ecological Applications 9:124-134. Hokit, D. G., B. M. Stith, and L. C. Branch. 2001. Comparison of two types of metapopulation models in real and ar tificial landscapes. Conservation Biology 15:1102-1113. Holt, R. D., and J. H. Lawton. 1994. The ecological consequences of shared natural enemies. Annual Review of Ecology and Systematics 25 :495-520. Horn, H. S. 1966. Measurement of "overlap" in comparative ecological studies. American Naturalist 100 :419-424. Huang, Y. X., and O. Diekmann. 2001. Predator migration in response to prey density: What are the consequences? Jour nal of Mathematical Biology 43:561-581. Hurlbert, S. H. 1978. The measurement of niche overlap and some relatives. Ecology 59:67-77. Jackson, J. F. 1973. Distribution and populati on phenetics of the Florida scrub lizard, Sceloporus woodi Copeia 1973:746-761. Jackson, J. F., and S. R. Telford, Jr. 1974. Reproductive ecology of the Florida scrub lizard, Sceloporus woodi Copeia 1974 :689-694.
128 Jansen, V. A. A. 2001. The dynamics of two diffusively coupl ed predator-prey populations. Theoretical Population Biology 59:119-131. Jeschke, J. M., M. Kopp, and R. Tollr ian. 2002. Predator functional responses: Discriminating between handling and di gesting prey. Ecological Monographs 72:95-112. Johnson, D. H. 1980. The comparison of usage and availability measurements for evaluating resource preference. Ecology 61:65-71. Johnson, R. W., R. R. Fleet, M. B. Keck, and D. C. Rudolph. 2007. Spatial ecology of the Coachwhip, Masticophis flagellum (Squamata : Colubridae ), in eastern Texas. Southeastern Naturalist 6:111-124. Joly, D. O., and B. R. Patterson. 2003. Use of selection indices to model the functional response of predators. Ecology 84:1635-1639. Jones, K. B., and W. G. Whitford. 1989. Feeding-behavior of free-roaming Masticophis flagellum an efficient ambush predator. Southwestern Naturalist 34:460-467. Kacelnik, A., J. R. Krebs, and C. Bern stein. 1992. The ideal free distribution and predator-prey populations. Tre nds in Ecology & Evolution 7 :50-55. Kareiva, P. 1990. Population dynamics in spat ially complex environments: Theory and data. Philosophical Transactions of the Royal Society of London B 330:175-190. Kelley, K. C., S. J. Arnold, and J. Glatstone. 1997. The effects of substrate and vertebral number on locomotion in the garter snake Thamnophis elegans. Functional Ecology 11:189-198. Kernohan, B. J., R. A. Gitzen, and J. J. Millspaugh. 2001. Analysis of animal space use and movements. Pages 125-166 in J. J. Millspaugh and J. M. Marzluff, editors. Radio Tracking and Animal Populations. A cademic Press, San Diego, California, USA. King, R. B. 2002. Predicted and observed maxi mum prey size snake size allometry. Functional Ecology 16:766-772. Klimstra, W. D. 1959. Foods of the Racer, Coluber constrictor in southern Illinois. Copeia 1959 :210-214. Lima, S. L. 1998. Nonlethal effects in the ecology of predato r-prey interactions. BioScience 48:25-34. Macartney, J. M., P. T. Gregory, and K. W. Larsen. 1988. A tabular survey of data on movements and home ranges of snakes. Journal of Herpetology 22:61-73.
129 Madsen, T., and R. Shine. 1996. Seasonal migr ation of predators and prey A study of pythons and rats in tropi cal Australia. Ecology 77:149-156. Manly, B. F. J., L. L. McDonald, D. L. Thomas, T. L. McDonald, and W. P. Erickson. 2002. Resource Selection by Animals, Second edition. Kluwer Academic Publishers, Dordrecht, The Netherlands. McCann, K., A. Hastings, and G. R. Huxel. 1998. Weak trophic in teractions and the balance of nature. Nature 395:794-798. McCoy, E. D., P. P. Hartmann, and H. R. Mushinsky. 2004. Population biology of the rare Florida scrub lizard in fr agmented habitat. Herpetologica 60:54-61. McCoy, E. D., and H. R. Mushinsky. 1992. Rarity of organisms in the sand pine scrub habitat of Florida. Conservation Biology 6:537-548. McCoy, E. D., and H. R. Mushinsky. 1994. E ffects of fragmentation on the richness of vertebrates in the Florida scrub habitat. Ecology 75 :446-457. McNair, J. N. 1980. A stochastic foraging model with predator training effects .1. Functional-response, switching, and run lengths. Theoretical Population Biology 17:141-166. Melian, C. J., and J. Bascompte. 2002. F ood web structure and habitat loss. Ecology Letters 5:37-46. Miller, D. E., and H. R. Mushinsky. 1990. Foraging ecology and prey size in the mangrove water snake, Nerodia fasciata compressicauda. Copeia 1990:10991106. Moorcroft, P. R., and M. A. Lewis. 2006. M echanistic Home Range Analysis. Princeton University Press, Princeton, New Jersey. Mullin, S. J., R. J. Cooper, and W. H. N. Gutzke. 1998. The foraging ecology of the gray rat snake (Elaphe obsoleta spiloides ). III. Searching for different prey types in structurally varied habita ts. Canadian Journal of Zoology-Revue Canadienne De Zoologie 76 :548-555. Murdoch, W. W. 1969. Switching in general predators: Experiments on predator specificity and stability of prey populations. Ecological Monographs 39:335-354. Murdoch, W. W. 1977. Stabilizing effects of spatial heterogeneity in predator-prey systems. Theoretical Population Biology 11:252-273. Mushinsky, H. R. 1987. Foraging ecology. Pages 320-334 in R. A. Seigel, J. T. Collins, and S. S. Novak, editors. Snakes: Ecology and Evolutionary Biology. Blackburn Press, Caldwell, New Jersey, USA.
130 Mushinsky, H. R., J. J. Hebrard, and D. S. Vodopich. 1982. Ontogeny of water snake foraging ecology. Ecology 63:1624-1629. Myers, R. L. 1990. Scrub and High Pine. Pages 150-193 in R. L. Myers and J. J. Ewel, editors. Ecosystems of Florida. Univer sity of Central Florida Press, Orlando. Nichols, J. D. 2005. Modern open-populati on capture-recapture models. Pages 88-123 in S. C. Amstrup, T. L. McDonald, and B. F. J. Manly, editors. Handbook of Capture-Recapture Analysis. Princeton Un iversity Press, Princeton, New Jersey, USA. Oaten, A., and W. W. Murdoch. 1975. Switching functional response, and stability in predator-prey systems. American Naturalist 109 :299-318. Pimm, S. L., and P. Raven. 2000. Extinction by numbers. Nature 403:843-845. Plummer, M. V., and J. D. Congdon. 1994. Radiotelemetric study of activity and movements of racers ( Coluber constrictor ) associated with a Carolina bay in South Carolina. Copeia 1994:20-26. Pollock, K. H. 1982. A capture-recapture design robust to unequal probability of capture. Journal of Wildlife Management 46:752-757. Pough, F. H., and J. D. Groves. 1983. Specializ ation of the body form and food habits of snakes. American Zoologist 23:443-454. Prakash, S., and A. M. de Roos. 2002. Habitat de struction in a simple predator-prey patch model: How predators enhance prey pe rsistence and abundance. Theoretical Population Biology 62:231-249. Pringle, R. M., J. K. Webb, and R. Shin e. 2003. Canopy structure, microclimate, and habitat selection by a nocturnal snake, Hoplocephalus bungaroides. Ecology 84:2668-2679. Quinn, G. P., and M. J. Keough. 2002. Experime ntal Design and Data Analysis for Biologists. Cambridge Univ ersity Press, Cambridge. Quinn, J. L., and W. Cresswell. 2004. Predat or hunting behaviour and prey vulnerability. Journal of Animal Ecology 73:143-154. R Core Development Team. 2006. R: A langua ge and environment for statistical computing. in R Foundation for Statistical Computing, Vienna, Austria. Reinert, H. K., and D. Cundall. 1982. An improved surgical implantation method for radio-tracking snakes. Copeia 1982:702-705.
131 Rodda, G. H., and T. H. Fritts. 1992. The impact of the introduction of the colubrid snake Boiga irregularis on Guam lizards. Journal of Herpetology 26 :166-174. Rodriguez-Robles, J. A. 2002. Feeding ecology of North American Gopher Snakes ( Pituophis catenifer Colubridae). Biological Jour nal of the Linnean Society 77:165-183. Rodriguez-Robles, J. A., C. J. Bell, and H. W. Greene. 1999. Gape size and evolution of diet in snakes: feeding ecology of erycine boas. Journal of Zoology 248:49-58. Rosenheim, J. A., and A. Corbett. 2003. Om nivory and the indeterminacy of predator function: Can a knowledge of foraging behavior help? Ecology 84:2538-2548. Row, J. R., and G. Blouin-Demers. 2006. Therma l quality influences habitat selection at multiple spatial scales in milksnakes. Ecoscience 13:443-450. Row, J. R., G. Blouin-Demers, and P. J. W eatherhead. 2007. Demographic effects of road mortality in black ratsnakes ( Elaphe obsoleta ). Biological Conservation 137 :117124. Ruben, J. A. 1977. Morphological correlates of predatory modes in the Coachwhip ( Masticophis flagellum ) and Rosy Boa ( Lichanura roseofusca ). Herpetologica 33:1-6. Ryall, K. L., and L. Fahrig. 2006. Response of predators to loss and fragmentation of prey habitat: A review of theory. Ecology 87:1086-1093. Savidge, J. A. 1987. Extinction of an island fo rest avifauna by an introduced snake. Ecology 68:660-668. Schmidt, K. A., J. R. Goheen, and R. Naumann. 2001. Incidental nest predation in songbirds: Behavioral indicators detect ecological scales a nd processes. Ecology 82:2937-2947. Schmidt, K. A., and C. J. Whelan. 1998. Pr edator-mediated intera ctions between and within guilds of nesting songbirds: E xperimental and observational evidence. American Naturalist 152 :393-402. Secor, S. M. 1995. Ecological aspect s of foraging mode for the snakes Crotalus cerastes and Masticophis flagellum Herpetological Monographs 9:169-186. Secor, S. M., and K. A. Nagy. 1994. Bioenergetic correlates of foraging mode for the snakes Crotalus cerastes and Masticophis flagellum Ecology 75:1600-1614. Sherrat, T. N., and A. D. Macdougall. 1995. Some population consequences of variation in preference among individual predators. Biological Journal of the Linnean Society 55 :93-107.
132 Shewchuk, C. H., and J. D. Austin. 2001. Food habits of the Racer ( Coluber constrictor mormon ) in the northern part of it s range. Herpetological Journal 11 :151-155. Shine, R. 1991. Why do larger snakes eat larger prey items? Functional Ecology 5:493502. Shine, R., G. P. Brown, and M. J. Elphic k. 2004. Field experiments on foraging in freeranging water snakes Enhydris polylepis (Homalopsinae). Animal Behaviour 68:1313-1324. Shrader-Frechette, K. S., and E. D. Mc Coy. 1993. Method in Ecol ogy: Strategies for Conservation. Cambridge University Press, Cambridge, UK. Swihart, R. K., Z. L. Feng, N. A. Slade, D. M. Mason, and T. M. Gehring. 2001. Effects of habitat destruction and resource supplementation in a predator-prey metapopulation model. Journal of Theoretical Biology 210:287-303. Tennant, A. 1997. A Field Guide to the Snakes of Florida. Gulf Publishing Company, Houston, Texas. Tiebout, H. M., and R. A. Anderson. 1997. A comparison of corrido rs and intrinsic connectivity to promote dispersal in transient successional landscapes. Conservation Biology 11 :620-627. Tiebout, H. M., and R. A. Anderson. 2001. Me socosm experiments on habitat choice by an endemic lizard: Implications for timber management. Journal of Herpetology 35:173-185. Tinkle, D. W., and A. E. Dunham. 1986. Comp arative life histories of two syntopic sceloporine lizards. Copeia 1986:1-18. Turchin, P. 1998. Quantitative Analysis of Movement. Sinauer Associates, Inc., Sunderland, Massachusetts. Turchin, P. 2003. Complex Population Dynamics : A Theoretical/Empirical Synthesis. Princeton University Press, Princeton, New Jersey. van Baalen, M., V. Krivan, P. C. J. van Rijn, and M. W. Sabelis. 2001. Alternative food, switching predators, and th e persistence of predato r-prey systems. American Naturalist 157 :512-524. Vickery, P. D., M. L. Hunter, and J. V. Wells. 1992. Evidence of incidental nest predation and its effects on nests of threatened grassland birds. Oikos 63:281-288. Vincent, S. E., B. R. Moon, R. Shine, a nd A. Herrel. 2006a. The functional meaning of "prey size" in water snakes ( Nerodia fasciata Colubridae). Oecologia 147 :204211.
133 Vincent, S. E., P. D. Vincent, D. J. Irsc hick, and J. M. Rossell 2006b. Do juvenile gapelimited predators compensate for their small size when feeding? Journal of Zoology 268:279-284. Waichman, A. V. 1992. An alphanumeric code for toe clipping amphibians and reptiles. Herpetological Review 23:19-21. Waldron, J. L., S. H. Bennett, S. M. Welc h, M. E. Dorcas, J. D. Lanham, and W. Kalinowsky. 2006. Habitat specificity and homerange size as attributes of species vulnerability to extinction: a case stu dy using sympatric rattlesnakes. Animal Conservation 9:414-420. Werner, E. E. 1991. Nonlethal effects of a pr edator on competitive interactions between two anuran larvae. Ecology 72:1709-1720. White, G. C. 2006. Program MARK. in Colorado State University, Fort Collins, Colorado, USA. Williams, B. L., E. D. Brodie, Jr., and E. D. Brodie, III. 2003. Coevolution of deadly toxins and predator resistance: Self-asse ssment of resistance by garter snakes leads to behavioral rejection of toxic newt pre y. Herpetologica 59:155-163. Wilson, L. D. 1970. The coachwhip snake, Masticophis flagellum (Shaw): taxonomy and distribution. Tulane Studie s in Zoology and Botany 16:31-99. Witz, B. W. 1990. Antipredator mechanisms in arthropods: A twenty year literature survey. Florida Entomologist 73:71-99. Witz, B. W., and H. R. Mushin sky. 1989. Pygidial secretions of Pasimachus subsulcatus (Coleoptera: Carabidae) deter predation by Eumeces inexpectatus (Squamata: Scincidae). Journal of Chemical Ecology 15:1033-1044. Witz, B. W., D. S. Wilson, and M. D. Palmer. 1991. Distribution of Gopherus polyphemus and Its Vertebrate Symbionts in 3 Burrow Categories. American Midland Naturalist 126 :152-158. Woodroffe, R., and J. R. Ginsberg. 1998. Edge effects and the extinction of populations inside protected areas. Science 280 :2126-2128. Worton, B. J. 1989. Kernel methods for estimating the utilization distribution in homerange studies. Ecology 70 :164-168. Wright, A. H., and A. A. Wright. 1957. Ha ndbook of Snakes of the United States and Canada. Comstock Publishing Associates, Ithaca, New York.
About the Author Brian James Halstead was raised on a small dairy farm in rural Door County, Wisconsin. He attended Carroll College (Wa ukesha, Wisconsin), where he majored in Biology and graduated Summ a Cum Laude in 1999. He received a Presidential Fellowship in 2001 to pursue his doctorate in Bi ology at the University of South Florida. He served as President of the Biology Gr aduate Student Organization from 2002 to 2003. He currently resides in Land O Lakes, Florid a with his wife, Kelly, and their three dogs and cat.