USF Libraries
USF Digital Collections

Experimental translocation of the florida sand skink (_plestiodon [=neoseps] reynoldsi_) success of a restricted species...

MISSING IMAGE

Material Information

Title:
Experimental translocation of the florida sand skink (_plestiodon =neoseps reynoldsi_) success of a restricted species across diverse microhabitats
Physical Description:
Book
Language:
English
Creator:
Osman, Nicholas
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Florida Sand Skink
Neoseps
Fossorial
Scrub
Translocation
Lizard
Plestiodon reynoldsi
Dissertations, Academic -- Biology-Integrative -- Masters -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: The fossorial Florida Sand Skink (Plestiodon =Neoseps reynoldsi) inhabits a restricted range of scrub and sandhill fragments on the ridges of central Florida. The high rate of urban and agricultural development in this area necessitates conservation strategies other than land acquisition and management because of the limited remaining Florida Sand Skink habitat available. This study tests the viability of translocation as a conservation strategy for this species and assesses which features of a recipient site contribute to the successful establishment of a population. In 2007, 300 individuals were collected and moved from an intact scrub habitat, individually marked, and moved to a nearby reclaimed site with no existing Florida Sand Skink population. Fifteen 20 m2 enclosures were constructed at the recipient site, and 20 skinks were randomly assigned to each.Translocated skinks were monitored for two years to measure survival and reproduction. While survival and reproduction were apparent in all treatments, survival was significantly greater in enclosures with no shade-providing object and low soil moisture, and reproduction was most evident in enclosures with less light intensity and soil compaction. This study indicates that translocation is a practical conservation strategy for this species, and my results can be used to inform protocol for future Florida Sand Skink translocation efforts.
Thesis:
Thesis (MS)--University of South Florida, 2010.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Nicholas Osman.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains X pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
usfldc doi - E14-SFE0004611
usfldc handle - e14.4611
System ID:
SFS0027926:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Experimental Translocation of the Florida Sand Skink ( Plestiodon [= Neoseps ] reynoldsi ): Success of a Restricted Species Across Diverse Microhabitats by Nicholas Paul Osman A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science Department of Biology College of Arts and Sciences University of South Florida Co Major professor: Henry Mushinsky, Ph.D. Co Major professor: Earl McCoy, Ph.D. Lynn Martin, Ph.D. Date of Approval: June 18, 2010 Keyw ords: lizard, scrub, fossorial, sand swimming, heterogeneity Copyright 2010, Nicholas Paul Osman

PAGE 2

Dedication This thesis is dedicated to my parents, Lloyd and Cam Osman, and my partner, Marc Kurtzman, for all of the ir love and support.

PAGE 3

i Table of Contents List of Tables ................................ ................................ ................................ ................................ .......... ii List of Figure ................................ ................................ ................................ ................................ .......... iii Abstract ................................ ................................ ................................ ................................ ................... v Introduction ................................ ................................ ................................ ................................ ............. 1 Methods ................................ ................................ ................................ ................................ .................. 6 Donor Site ................................ ................................ ................................ ................................ 6 Recipient Site ................................ ................................ ................................ ............................ 8 Survival ................................ ................................ ................................ ................................ .... 11 Treatment and Block Effect ................................ ................................ ...................... 11 Effect of Environment ................................ ................................ ............................... 13 Effect of Sex, Age Class, and Size ................................ ................................ .......... 14 Effect of Trap Location ................................ ................................ ............................. 15 Model Comparison and Assessing Model Fit ................................ ......................... 15 Reproduction ................................ ................................ ................................ ........................... 16 Results ................................ ................................ ................................ ................................ .................. 17 Donor Site ................................ ................................ ................................ ............................... 17 Recipient Site ................................ ................................ ................................ .......................... 18 Survival ................................ ................................ ................................ ................................ .... 19 Treatment and Block Effect ................................ ................................ ...................... 19 Effect of Environment ................................ ................................ ............................... 21 Effect of Sex, Age Class, and Size ................................ ................................ .......... 21 Effect of Trap Location ................................ ................................ ............................. 22 Reproduction ................................ ................................ ................................ ........................... 22 Discussion ................................ ................................ ................................ ................................ ............. 24 Donor Site ................................ ................................ ................................ ............................... 24 Recipient Site ................................ ................................ ................................ .......................... 25 Survival ................................ ................................ ................................ ................................ .... 27 Reproduction ................................ ................................ ................................ ........................... 28 Measures of Success/Recommendations ................................ ................................ ............ 28 Figures ................................ ................................ ................................ ................................ .................. 34 Tables ................................ ................................ ................................ ................................ ................... 50 Literature Cited ................................ ................................ ................................ ................................ ..... 59 About the Author ................................ ................................ ................................ ..................... End Page

PAGE 4

ii List of Tables Table 1. Diffe rences in environmental variables among treatments ................................ ................ 50 Table 2. Differences in environmental variables among treatment replicates ................................ 51 Table 3. Distribution of sex, age class, and capture locations of individuals relocated to each enclosure ................................ ................................ ................................ ...................... 52 Table 4. M odels compared in MARK to assess the effect of enclosure grouping and time on recapture probability estimation ( P ) ................................ ................................ ............... 53 Table 5. Model comparison for the effect of enclosure grouping on survival rate estimation ( ) ................................ ................................ ................................ ....................... 54 Table 6. Models testing the individual effects of the treatments given to enclosures on survival rate estimation ( ) ................................ ................................ ................................ .. 55 Table 7. Comparison of models incorporating enclosure averages of environmental variables into survival rate est imation ( ) ................................ ................................ ........... 56 Table 8. Comparison of models with size related individual covariates ................................ ........... 57 Table 9. Model comparisons for the effect of initial capture site on survival rate estimation ( ) ................................ ................................ ................................ ....................... 58

PAGE 5

iii List of Figures Fig. 1. Donor a nd recipient site location in central Florida. ................................ ............................... 34 Fig. 2. Trap array locations at the donor site ................................ ................................ ...................... 35 Fig. 3. Schematic of trap arrays at donor site ................................ ................................ .................... 36 Fig. 4. Example of a marked Florida Sand Skink ................................ ................................ ............... 36 Fig. 5. Photos of recipient site treatments ................................ ................................ .......................... 37 Fig. 6. Arrangement of treatment replicates at recipient site. ................................ ........................... 37 Fig. 7. Schematic of enclosures at the recipient ................................ ................................ ............... 38 Fig. 8. Capture frequency at the donor site in 2007 ................................ ................................ .......... 39 Fig. 9. Distribution of Florida Sand Skinks captured per trap array at the donor si te ...................... 39 Fig. 10. PCA loadings for each variable using donor site environmental data. ............................... 40 Fig. 11. Principal component scores for trap arrays at the donor site ................................ .............. 40 Fig. 12. Soil moisture and captures at the donor site ................................ ................................ ........ 41 Fig. 13. Hierarchical clustering ana lysis using environmental data measured within enclosures ................................ ................................ ................................ ............................... 41 Fig. 14. PCA scores for enclosures used as supplemental cases ................................ ................... 42 Fig. 15. Capture frequency at the recipient site in 2008 and 2009 ................................ ................... 42 Fig. 16. Recaptures of translocated Florida Sand Skinks by enclosure and year ........................... 43 Fig. 17. Recaptures of translocated skinks totaled by treatment ................................ ...................... 43 Fig. 18. The interaction between treatment and block for recaptures per enclosure ...................... 44 Fig. 19. Survival rate estimates (with 95% confidence intervals) from model s R P T (A), R*B P T (B), and E P T (C) ................................ ................................ ................................ ........ 45 Fig. 20. Relationship between recaptures and soil moisture at the recipient site ........................... 46 Fig. 21. Size of females, males, and juveniles translocated into enclosures ................................ ... 47 Fig. 22. Unmarked juvenile Florida Sand Skinks fo und in each enclosure by year ........................ 48 Fig. 23. Unmarked juvenile Florida Sand Skinks found in each enclosure totaled by treatment ................................ ................................ ................................ ................................ 48

PAGE 6

iv Fig. 24. Correlation between reproduction and environmental variables within enclosures ........... 49

PAGE 7

v Experimental Translocation of the Florida Sand Skink ( Plestiodon [= Neoseps ] reynoldsi ): success of a restricted species across diverse microhabitats Nicholas Paul Osman Abstract The fossorial Florida Sand Skink ( Plestio don [= Neoseps ] reynoldsi ) inhabits a restricted range of scrub and sandhill fragments on the ridges of central Florida. The high rate of urban and agricultural development in this area necessitates conservation strategies other than land acquisition and ma nagement because of the limited remaining Florida Sand Skink habitat available. This study tests the viability of translocation as a conservation strategy for this species and assesses which features of a recipient site contribute to the successful establi shment of a population. In 2007, 300 individuals were collected and moved from an intact scrub habitat, individually marked, and moved to a nearby reclaimed site with no existing Florida Sand Skink population. Fifteen 20 m 2 enclosures were constructed at t he recipient site, and 20 skinks were randomly assigned to each. These enclosures were divided among five treatments, which represented the range of habitat types at the donor site and differed in the presence or absence of a shade providing object and coa rse woody debris. Translocated skinks were monitored for two years to measure survival and reproduction. While survival and reproduction were apparent in all treatments, survival was significantly greater in enclosures with no shade providing object and lo w soil moisture, and reproduction was most evident in enclosures with less light intensity and soil compaction. Common measurement of environmental variables at the donor and recipient sites showed that all of the recipient site enclosures differed from th e donor site in the amount of vegetative cover but contained the structural heterogeneity that is associated with Florida Sand Skink presence in the wild. This study indicates that translocation is a practical conservation strategy for this species, and my results can be used to inform protocol for future Florida Sand Skink translocation efforts.

PAGE 8

1 Introduction The Florida Sand Skink ( Plestiodon [= Neoseps ] reynoldsi Stejneger) is precinctive to scrub and sandhill habitats on the ridges of central Florida remnants of the once widespread xeric conditions of the late Pliocene era prior to Pleistocene sea level rise (Watts and Hansen 1988, Branch et al 2003). The Lake Wales Ridge (LWR), by far the largest of these ancient sand dunes, has had 85% of its orig inal 2400 km 2 developed for agricultural, commercial, and residential use (Turner et al 2006). Consequently, the Florida Sand Skink is federally listed as a "threatened" species (U.S. Fish and Wildlife Service 1987). A recent assessment of the effectivene ss of protecting remaining undeveloped land to reduce the risk of extinction of LWR species found that the acquisition of additional reserves would have little impact on the success of the Florida Sand Skink (Turner et al. 2006), emphasizing the importance of effective management practices in existing reserves. While land acquisition and management are important, the continued high rate of development on the LWR and other ridges necessitates alternative conservation strategies. One such strategy being consi dered is the restoration of vegetation on reclaimed developed lands and subsequent translocation of Florida Sand Skinks from sites commissioned for development to these areas. Translocations are broadly defined as the movement of living organisms by huma ns from one area to another within their range (IUCN 1987, Reinert 1991, IUCN 1998), and the suitability of reptiles as candidates for this action and its effectiveness as a conservation strategy have been debated based on reviews of previous efforts (Grif fith et al 1989, Burke 1991, Dodd and Seigel 1991, Germano and Bishop 2008). As a result, several recommendations are available to those considering translocation. These reviews emphasize (1) the need to understand the species and the threats that face i t at its existing location (henceforth referred to as the "donor site"); (2) the importance of the quality of the site to which the species is being translocated (henceforth referred to as the "recipient site"); and (3) the obligation to set clear goals an d follow

PAGE 9

2 through with long term monitoring. More recent reviews also stress the need for experimental approaches to translocations in order to develop standardized techniques before the practice is adopted as a conservation strategy for a species (Seddon e t al 2007). In their review of reptile translocations, Germano and Bishop (2008) found that success was independent of the number of animals translocated or the age at which they were moved, and failure was most often associated with dispersal away from the recipient site. The second most common reason for failure was the low quality of habitat at the recipient sites. Griffith et al (1989) found a similar trend in their review of bird and mammal translocations, as did Rout et al (2007) when using a mode ling approach: success was most often related to the quality of the recipient site rather than the quantity of animals released. The importance of recipient site quality poses a challenge to translocation efforts involving the Florida Sand Skink because o f the lack of a firm understanding of its habitat preferences in the wild. Studies of the Florida Sand Skink's habitat associations often have ambiguous or contradictory results. In fact, a study by McCoy et al. (1999) showed that literature descriptions o f suitable habitats are often poor predictors of actual Florida Sand Skink abundance. Furthermore, Florida Sand Skink populations continue to be discovered in habitats that are otherwise characterized as unsuitable and are often absent from other seemingly suitable sites (Sutton 1996, Pike et al. 2007). Within scrub and sandhill habitats on the LWR, Florida Sand Skink density is most often correlated with the presence of bare, loose sands that are low in moisture (Sutton 1996, Navratil 1999, Gianopulos 200 1). These soil conditions allow them to freely "swim" through the sand, a form of locomotion to which they are morphologically adapted. Like other fossorial species, the Florida Sand Skink has a relatively short tail and elongated trunk, its limbs are sign ificantly reduced in size and number of digits, and it has a countersunk lower jaw and wedge shaped snout (Telford 1959). The dry, bare, loose sands that are conducive to "sand swimming" are often the result of the lack of dense vegetative cover, and there fore Florida Sand Skink presence is often associated with sparse canopy cover (McCoy and Mushinsky 1991, Sutton 1996) and understory vegetation (Gianopulos 2001). However, the habitat characteristics that contribute to

PAGE 10

3 their ease of movement through the sa nd are in conflict with habitat characteristics that meet their thermoregulatory and nutritional needs. For instance, Collazos (1998) found Florida Sand Skink presence to be strongly associated with areas of relatively low soil temperatures resulting from shade from trees, shrubs, and grasses. Another study found that capture locations were positively associated with grass and shrub cover when there was no canopy present but negatively associated with grass and shrub cover when a canopy was present (Hill 19 99), highlighting the importance of shade providing objects for thermoregulation. The importance of thermoregulation for the Florida Sand Skink is evidenced by the fact that their periods of movement in the wild are strongly tied to their preferred tempera tures. Florida Sand Skink's time of daily activity changes seasonally to meet the time of day when surface sand temperatures are between 28 and 32 o C, and they are most active during their spring mating season, when these temperatures occur most frequently (Andrews 1994). Vegetation also serves two other critical functions through its associated debris and root structure: (1) it holds moisture, which must be at least minimally available to the Florida Sand Skink for hydration (Telford 1965) and (2) it attrac ts the Florida Sand Skink's prey, which is primarily beetle larvae and termites (Myers and Telford 1965, Smith 1977, McCoy et al in review). The apparent contradiction in habitat features necessary for the Florida Sand Skink's ability to move and their a bility to thermoregulate, hydrate, and find prey has lead to several conclusions in the literature about optimal Florida Sand Skink habitat. Hill (1999) proposed that structural heterogeneity is key, such as areas with differing strata of vegetation adjace nt to bare, open areas (Hill 1999). McCoy et al (1999) questioned the notion that there is "optimal" Florida Sand Skink habitat at all. Similarly, Gianopulos (2001) concluded that their presence is likely tied to specific microhabitat features, which can be present in several different scrub and sandhill habitat types. These conclusions make the design of a recipient site for translocation of the Florida Sand Skink difficult. Indeed, poor recipient site quality has had a negative impact of the outcome o f previous translocation efforts involving the Florida Sand Skink and other fossorial lizards. Monitoring two years after the relocation of a population of slow worms ( Anguis fragilis ), a European legless

PAGE 11

4 semi fossorial lizard, showed that population size, reproduction, and body condition had decreased, in part due to the poor quality habitat at the recipient site (Platenberg and Griffiths 1999). The model for this translocation was similar to that proposed for future Florida Sand Skink translocations, wher e individuals were moved from a site scheduled for development to a recipient site that was previously developed, reclaimed, and altered to increase the likelihood of survival. An initial experimental translocation of the Florida Sand Skink has been conduc ted and also followed this model (Hill 1999, Penney 2001). This study involved two reclaimed recipient sites, an abandoned citrus grove and an abandoned pasture. Each site was cleared of most of its vegetation and topsoil and divided up among four treatmen ts, which differed in the way that topsoil and vegetation taken from the donor site were treated before being relocated to the recipient site (mulching, burning, lack of treatment, etc.). The 154 Florida Sand Skinks that were captured at the donor site wer e divided among the recipient sites and treatments within each site. Six years after the translocation, difference in capture rates among treatments was negligible and only one of the sites supported a self sustaining population. This site, however, was co ntiguous with a well established scrub habitat that likely helped to support the translocated (as well as a native) Florida Sand Skink population. The alteration of the reclaimed recipient site in this study emphasized the vegetative species make up rather than the structural role and ecological functions of the vegetation. The present study tests the viability of translocation as a conservation strategy for the Florida Sand Skink and assesses which features of a recipient site contribute to successful es tablishment of a translocated population. An experimental approach is used, manipulating structural elements such as canopy, root structure, and debris within enclosed recipient sites, rather than attempting to recreate vegetative species composition. Rele asing individuals into enclosed areas ensures that success or failure will be attributed to the treatments given to these areas, as well as increases the probability of recapturing survivors. Moreover, by correlating structural habitat elements with Florid a Sand Skink density at the donor site, a better understanding of habitat associations in the wild can be obtained. Measurement of common

PAGE 12

5 environmental variables at both sites will allow comparisons to determine the quality of habitat at the recipient site Previous studies of Florida Sand Skink habitat associations in the wild indicate that survival will be possible over a range of microhabitat types and that structural heterogeneity may be an important factor for success. The results will inform protocol for future translocation efforts involving the Florida Sand Skink.

PAGE 13

6 Methods Donor Site Two hundred and ten trap arrays were constructed at the donor site near Davenport, Florida, an approximately 1.2 km 2 scrub habitat in Polk Count y on the north central part of the Lake Wales Ridge (Lat. 28.17920, Long. 81.56499) (Fig. 1). This site was scheduled to be used as a sand mine, but at the time of trapping it supported a relatively dense Florida Sand Skink population (approximately 600/h a) and was dominated by sand live oak ( Quercus geminata) and scrub palmetto ( Sabal etonia ) with large open spaces between areas of vegetation. The 210 trap arrays were divided among four sites of 50 arrays each and two sites of five arrays each (Fig. 2). T he smaller two sites were constructed for a concurrent genetic study, but Florida Sand Skinks captured in them were used for the analyses and translocation nonetheless. Each trap array consisted of four 2 m aluminum flashing drift fences pointing in each c ardinal direction, and each drift fence had two 3.8 L buckets on each side, totaling 16 buckets per trap array (Fig. 3). Arrays were approximately 8 m 2 in size and were spaced about 20 m apart within sites. The buckets were countersunk in the ground so tha t the rim was 5 to 10 cm below the surface of the sand. The lids that provided shade were held up using sticks, creating at least an inch of room for skinks to enter. Small holes were created at the bottom of the buckets for drainage and water was added to the inch of sand in the bottom of the buckets during periods of no rain. At four of the sites, a 20 m 2 enclosure was built for the concurrent genetic study. These enclosures were the same as those at the recipient site described below. Florida Sand Skinks captured within these enclosures were used for the translocation, but neither the number of individuals captured inside these enclosures nor any environmental data from them was included in any of the analyses described below. Traps were opened from Marc h 28 to May 25, 2007 and checked every other day by a team of field technicians (Entrix Inc., Riverview, FL). The number of Florida Sand Skinks captured

PAGE 14

7 at each array was recorded. To assess whether individuals were captured at fewer arrays than would be e xpected by chance given the total number of arrays and the total number of individuals captured, a Monte Carlo simulation was performed with 1000 iterations using Resampling Stats v5.02 (Simon 2000). This procedure created 95% confidence intervals around t he expected proportion of arrays at which individuals would be captured by chance. If skinks were captured at fewer arrays than expected, an investigation into the habitat features associated with areas of high trap frequency would be warranted. At each trap array, a 1 m 2 quadrat was laid between each pair of adjacent drift fences (four quadrats per array, Fig. 3), and the percent of the quadrat filled by (1) bare ground, (2) live vegetation, (3) lichen, and (4) leaf litter and debris (total =100%) was es timated. The percent filled by coarse woody debris that may have enhanced prey species abundance was also estimated, but this percentage did not count toward the 100% filled by the other groundcover features. Light intensity (in Lux) was measured with a li ght meter (Extech Instruments Corporation, Waltham, MA) laid on the ground in the center of the quadrat, and it was later corrected if the reading was taken while the sun was behind a cloud. Measurements taken while the sun was behind a cloud (a situation that was minimized) were separated from those taken in full sun. Each measurement was then converted to a percentage of the maximum light intensity measured in its category (full sun or cloudy). Canopy height was recorded using categorical measures: 1 = le ss than 1 m, 2 = 2 3 m tall, and 3 = greater than 3 m tall. A penetrometer was used to measure the soil compaction. The pressure (in pounds per square inch) needed to reach the first abrupt change in compaction ("break") was recorded, as well as the depth at which this break occurred. The data from the four quadrats at each array were averaged. When these averages were used for statistical analyses, raw percentage data were arcsine square root transformed before averaging. Moreover, a soil corer was used to sample the soil at the center of each array for soil moisture analysis. These samples were weighed before and after drying at 100 o C for 24 hours in a drying oven, and the percentage of moisture was calculated. Because of the likelihood of intercorrelati on of these data, principal component analysis was used to reduce the number of variables. The scores from the components that explained

PAGE 15

8 most of the variation in the data were tested for correlation with the total number of Florida Sand Skinks captured at each array. Soil moisture was tested for correlation with the number of captures per array separately in order to maintain consistency with the analyses below. Soil moisture was also tested for significant correlation with the principal component scores us ing Spearman's Rank Test of Correlation. All statistical analyses were carried out using the program Statistica version 9 (StatSoft, Inc.), unless otherwise noted. Recipient Site The mass, sex, and snout vent length (SVL) of each Florida Sand Skink captur ed at the Davenport, FL site during 2007 was recorded, and each individual was given an individual mark using a distinct spatial and color combination of a biocompatible fluorescent plastic polymer injected under the skin (Northwest Marine Technology, Inc. Shaw Island, WA) (see Penney et al. 2001 for comments about using this technique with Florida Sand Skinks). Each individual was given three marks, which consisted of one to four colors and spanned six possible locations on its ventral side (Fig. 4). The first 300 Florida Sand Skinks captured at the donor site in 2007 were released into fifteen 20 m 2 experimental enclosures at the recipient site within the Reedy Creek Improvement District. This site, located approximately 24 km north of the donor site (Lat 28.41379, Long. 81.61898), is a 6.07 ha isolated upland within the Reedy Creek Swamp. It consists of Candler and Tavares fine sands and was formerly used as a citrus grove but now supports a dense population of the Gopher Tortoise ( Gopherus polyphemus ) and scattered mature sand live oaks ( Q. geminata ). No Florida Sand Skinks inhabited this site prior to the translocation. Of the first 300 individuals captured at the donor site, 20 were released in each of the 15 enclosures, which consisted of three repli cates of five treatments. The remaining Florida Sand Skinks captured at the donor site also were brought to the recipient site but were released outside of the experimental enclosures for future study. These individuals' marks indicated that they were not initially released into an enclosure so that they would not accidentally be included in this study if they subsequently moved into enclosures.

PAGE 16

9 The five treatments resembled different microhabitat types and represented the variation of microhabitat types at the donor site. They differed in the presence or absence of (1) a shade providing object, which was either one of the mature trees that already existed at the recipient site or a shade cloth suspended two meters above the ground on a wooden frame, and ( 2) added coarse woody debris, which is thought to provide refuge and attract prey insect species. The treatments, which are shown in Figure 5, are as follows: (1) shade from a tree and woody debris (referred to as "Tree+Wood" for analyses), (2) shade from a tree but no woody debris ("Tree"), (3) woody debris but no shade ("Wood"), (4) shade from a shade cloth (to assess the effects of trees such as soil compaction or leaf litter) and no debris ("Shadecloth"), and (5) no shade or debris ("Control"). The spa tial arrangement of the treatment enclosures was similar to that of a randomized block design, in which there were three groups of enclosures ("blocks"), each containing one replicate from each of the five treatments (Fig. 6). Each enclosure contained sixt een 2 m aluminum flashing drift fences with two countersunk 3.8 L buckets at the end of each. Twelve buckets were also arranged along the inside of the aluminum flashing enclosure walls, for a total of 76 bucket traps per enclosure (Fig. 7) (Sutton 1996). Each drift fence was given a number (1 through 16) and each wall of the enclosures was referred to as the north, east, south, or west wall. In recognition of the uncertainty in the number of Florida Sand Skinks that would be captured at the donor site, th e assignment of individuals to treatments was done in such a way that would assure equal numbers of individuals within each enclosure even if 300 were not captured. Individuals caught at the donor site were assigned to enclosures in groups of five, rotatin g through the enclosures mixed among treatments until each had 20 individuals. This density approximates normal Florida Sand Skink densities in good habitat (600/ha) (Sutton 1996, Collazos 1998, Mushinsky and McCoy 1999), but this method did not allow each enclosure to contain equal proportions of males, females, and juveniles. Traps were opened the following two years during the time of Florida Sand Skink peak activity to assess survival and reproduction. Traps were opened from March 4 to May 28, 2008 and from March 3 to May 27, 2009. As with the donor site, opening and checking traps included

PAGE 17

10 elevating the bucket lids with sticks, adding water when needed, maintaining one to two inches of sand at the bottom of each bucket, and creating small holes for drai nage. During the trapping period, each bucket trap was checked every other day, even if all bucket traps were not checked in the same day. Traps were always checked by "block" (i.e. one replicate set) so that each treatment always received equal "trap effo rt" on any given day. When a Florida Sand Skink was found, its mark, sex, and the location at which it was found were recorded. In 2009, snout vent length was recorded as well. Coded into each Florida Sand Skink's individual marking was the treatment to wh ich they were initially assigned. If a skink was captured in a treatment enclosure different than the one to which it was initially assigned, it was put back into its original treatment enclosure and was not included in that year's recapture data. If an in dividual was missing one of the marks, its identity could usually be deduced based on its remaining marks, sex, and location. If a recaptured individual's identity could not be deciphered, it was not included in any analyses. If an individual had no marks, indicating that it was born at the recipient site, it was brought back to the laboratory. Its sex, mass, and SVL were recorded, and it was given a new individual mark that corresponded to the treatment enclosure in which it was found, where it was subsequ ently released. During the 2009 trapping season, the same environmental data collected at the donor site were collected at the recipient site. Within each enclosure, a 1 m 2 quadrat was used to measure groundcover, light intensity, soil compaction, and can opy height, in 16 locations (Fig. 7) using the methods described above. Soil samples were taken at three random locations within each enclosure to determine soil moisture using the methods described above. Furthermore, two Thermochron iButton temperature l oggers in each enclosure recorded the temperature every half hour during a two week period of maximum capture frequency at the recipient site in 2009. This time period was determined based on the 2008 capture frequencies. One temperature logger was placed in the center of each enclosure and the other was placed at a randomly determined distance (maximum 10 m) and direction from the center. At these locations, the temperature loggers were buried under a few centimeters of sand, within grass, or under debris, attempting to replicate the space occupied by the Florida Sand Skink. The temperature frequencies were

PAGE 18

11 determined within each enclosure and the percentage of recordings that were within the Florida Sand Skink's preferred temperature, 28 o C and 32 o C (Andrew s 1994), was determined. Differences in these environmental variables among treatments were tested using one way ANOVA. Each variable was tested individually and Tukey's HSD tests were used for post hoc analysis. Differences among treatment replicates were also tested using one way ANOVAs. Then, the enclosure average for each variable was used in a hierarchical clustering analysis to see if replicates of the same treatment would be grouped together based on the environmental variables measured. Euclidean di stances were used to assess similarity between enclosures for clustering. To compare the recipient site to the donor site, measures from within the enclosures at the recipient site were used as "supplementary cases" in the principal component analysis desc ribed above. This procedure would assign each point of measurement at the recipient site component scores based on the loadings from the PCA of the donor site data. For the components that explained most of the variation in data at the donor site, the scor es from the donor site array and recipient site enclosures would be compared using Mann Whitney U tests. Because only three soil samples were taken at each site, soil moisture was independently compared between sites also using a Mann Whitney U test. Sur vival Treatment and Block Effect The total number of unique recaptures over the two years was calculated for each enclosure and two one way ANOVAs were run, with treatment and block as the factors. This procedure was done for each year individually as well Recapture data were analyzed also using the program MARK (White and Burnham 1999), which estimates survival rate ( ) and recapture probability ( P ). The models used to calculate these estimates account for deaths caused by trapping as well as "missed" rec aptures. In the case of trapping deaths, a Florida Sand Skink that was found dead in 2008 would not affect or P when it is not found in 2009. "Missed" captures were cases in which a Florida Sand Skink released in 2007 was not recaptured in 2008 but was

PAGE 19

12 r ecaptured in 2009, and and P would be affected accordingly. A standard Cormack Jolly Seber capture recapture model was chosen for MARK analyses. The method of and P estimation in MARK is constrained by the grouping of recapture events and released indi viduals set by the user. For instance, one estimate of survival can be obtained for the entire duration of the study (constant ) or, alternatively, survival rate can be estimated for the interval between each recapture event (time dependent ). Similarly, survival rate can be estimated for each of the five treatments or one estimate of survival can be generated for the entire translocated population. My goal was to find which model best explained the variation in the recapture data among a set of models in which survival was estimated for the enclosures grouped in different ways. A treatment effect could be inferred if a model that estimates survival for each of the five treatments was found to be more parsimonious than either of the following models: (1) o ne that generates only one survival estimate for all enclosures or (2) one that generates a survival estimate for each of the three "blocks". Before the effect of enclosure grouping on model parsimony could be assessed directly for the effect of enclosu re grouping and time on P had to be determined. In other words, any treatment, block, or time effect on recapture probability had to be determined before I could assess whether there was a treatment or block effect on survival. I did not use models in whic h estimation was time dependent because in MARK only the product of and P for the last interval is identifiable, and therefore time dependence could not be incorporated into both and P (Lebreton et al. 1992). Nevertheless, the focus of this analysis w as to determine if there was an effect of treatment on survival and not to assess difference in survival between years. Eleven candidate models were run to determine whether there was a group or time effect on P I chose a "null" grouping of enclosures fo r estimation for these models, in which only one estimate was generated for all enclosures ( .) The 11 different group and time combinations used for P estimation included grouping enclosures by block ( P B three estimates) and by treatment ( P R five esti mates), as well as grouping all enclosures together ( P ., one estimate) and not grouping them at all ( P E 15 estimates). For each of these, time dependence was then included ( P B*T P R*T P T and P E*T respectively). P B*T P R*T and P E*T models were each run with

PAGE 20

13 and without the group time interaction ( P B*T*BxT etc.). If a model with the interaction included is more parsimonious than one without the interaction included, it is similar to having a significant interaction effect in an analysis of variance. Once the best model for P estimation was chosen, it was used to find the most parsimonious grouping of the enclosures for estimation. Models ! E B and R were compared, as well as a model that accounted for the treatment and block effect ( R*B ). Like E this model returned 15 estimates, but each enclosure was assigned to a treatment and block. This model was also run with the treatment block interaction included ( R*B*RxB ). I fit another group of models to the data to assess the individual effects of shade, the presence of trees, and the presence of coarse woody debris. This analysis would determine the effects of the three treatments given to enclosures, regardless of which other treatment an enclosure was also given. For each of three models ran in M ARK, was estimated for two groups of enclosures: (1) the enclosures given course woody debris and those that were not ( wood ); (2) those that had a tree in them and those that did not ( tree ); and (3) those that had a shade giving object (whether a tree or shade cloth) and those that did not ( shade ). These models were compared to the null model ( .) and a model that grouped enclosures into the five treatments (as described above, R ). All of these models were run using both time dependent and constant re capture probability estimation ( P T and P. respectively). A two way ANOVA was also used to compare the total unique captures in each enclosure between these groups (trees and no trees; shade and no shade; and wood and no wood). The ANOVAs were also run for each trapping year individually. "Block" was the other factor in these analyses. A two way ANOVA was possible in this case because there was replication within groups, unlike the ANOVA for the comparison of treatments above. Effect of Environment The fol lowing analysis ignores the grouping of enclosures into treatments or blocks and focuses on the relationship between recaptures and environmental data within each individual enclosure. The relationship between survival and habitat in each enclosure was fir st assessed

PAGE 21

14 using Spearman's Rank Correlation Test. The total number of unique recaptures within each enclosure was tested against the enclosure average for each of the environmental variables measured (described above). In MARK, parameter estimates ( and P ) can be constrained to be linear functions of a covariate. I used the enclosure average for each environmental variable measured as covariates for this type of analysis. One covariate per model was used for estimation, while P estimation was time depe ndent ( P T based on the above analyses). These 10 models were compared to a model which also estimated for each enclosure but was not constrained by a covariate ( E ), as well as the "null" model which estimated only one ( .). Effect of Sex, Age Class and Size The recapture rates of males, females, and juveniles (at the time of release) were calculated from the recapture data combined for both trapping years and all enclosures. Subsequently, MARK was used to estimate for males, females, and juvenile s. A model that incorporated sex and age class into survival rate estimation ( MFJ P .) was compared to a model that did not account for sex and age class ( P .). These models were also run with time dependence incorporated into recapture probability estimat ion ( MFJ P T and P T ) for comparison. A Mann Whitney U test was used to compare the SVL and mass (at time of release) of recaptured and non recaptured Florida Sand Skinks. Furthermore, MARK allows these individual covariates to be incorporated into the mo del structure. I ran models that included the mass and SVL of each Florida Sand Skink that was released into enclosures ( mass P and SVL P .). I then included the square of mass and SVL into these models ( mass*mass2 P and SVL*SVL2 P .). If these latter mode ls were more parsimonious than the former, it may indicate that intermediate sized individuals were surviving better than individuals at the extremes of size. All of these models were also compared to a null model that did not account for size heterogeneit y ( P .). This set of models was re run with time dependence incorporated into P estimation ( P T ).

PAGE 22

15 Finally, the SVL and mass of males, females, and juveniles were compared using a Kruskal Wallis analysis of ranks test to assess the independence of any sex, age, and size effect found. Mann Whitney U tests were used for post hoc comparisons. Effect of Trap Location The effect of trap location at the donor site on survival was then similarly determined. The recapture rate was calculated for Florida Sand Skink s captured at each of the six trap sites within the donor site, followed by a model comparison using MARK. A model that incorporated the initial capture sites of the Florida Sand Skinks into survival rate estimation ( trapsite P .) was compared to a null mod el ( P .). These models were then re run with time dependent recapture probability ( trapsite P T and P T ). Model Comparison and Assessing Model Fit For each of the above analyses using the program MARK, the most parsimonious model was chosen from a set of candidate models based on corrected Akiake's information criterion (AICc). Parsimony, in this case, is based both on the fit of the model to the data (model deviance) and the number of parameters estimated in the model. The AICc values take into account b oth of these features and will decrease with increasing parsimony. Likelihood ratio tests (LRT), available as part of the program MARK, were used to test for significant differences between models. The LRT procedure, which produces a p value based on a Chi squared test, can only be used to compare two models that are nested (i.e. when one model is a reduced version of the other). Therefore, AICc values (Cooch and White 2001) were also used when comparing models. While there is no p value when comparing AICc values, Anderson and Burnham (1999) recommend that the difference in AICc values exceed 2 for any pair of models to be considered significantly different. Before these comparisons could be made, the most parameterized model in each analysis was assessed for its goodness of fit to the recapture data. I used the standard GOF bootstrap test available in MARK. This test generates a model deviance for simulations that can

PAGE 23

16 be compared to the observed deviance. The probability of obtaining a deviance as large as the observed deviance was calculated based on 100 simulations. Because the simulated data perfectly fit the model assumptions, this probability could be used to assess whether the model being tested was adequately fit to the data. Lack of fit can be inter preted as either a violation of the assumptions of the underlying CJS model or under or over dispersion of the data. Lack of fit can be accounted for in the model by adjusting the "variance inflation factor", or c hat. The adjusted c hat is found by divid ing the observed deviance by the average of the simulated deviances. Lebreton et al. (1992) suggest that this is acceptable if the adjusted c hat is below 3 because the lack of fit may be caused by over dispersion. If the lack of fit is greater, the result s of the analysis are considered unreliable. If c hat is adjusted, AICc and model deviance are referred to as QAICc and QDeviance, respectively. Reproduction The total number of Florida Sand Skinks found without marks within each enclosure was compared with two one way ANOVAs, with treatment and block as the factors. This procedure was done for each year individually as well. Two way ANOVAs were also run that compared the number of individuals born in each enclosure between the following enclosure types: enclosures with trees and those without trees; enclosures with shade in the form of trees or shade cloth and those without shade; and those with coarse woody debris added to them and those without woody debris added (as described above). Block was the sec ond factor in each of these tests. Finally, Spearman's Rank Correlation Tests were used to detect any relationship between the number of individuals born in each enclosure and the enclosure average for each environmental variable measured.

PAGE 24

17 Results Donor Site Five hundred and ten Florida Sand Skinks were captured at the donor site near Davenport, FL. These captures took place at 171 of the 210 trap arrays, or 81.4% of arrays. Based on the number of individuals captured, the Monte Carlo simulations g enerated a 95% confidence interval ranging from 87.6% to 94.3%. Therefore, individuals were caught in significantly fewer traps than would be expected by chance, indicating aggregation. Trap frequency ranged from 0 to 60 individuals captured per day and wa s greatest from late April to early May (Fig. 8). The number of skinks captured per array ranged from 0 to 9 (Fig. 9). The first three components from the PCA explained 69.25% of variation in the donor site environmental data: 42.74, 14.28, and 12.23% res pectively. The respective eigenvalues for these components were 3.9, 1.3, and 1.1. The other six components each accounted for less than 10% of the variation. Figure 10 shows the contribution of each variable to the first two components (loadings). A signi ficant positive correlation exists between the number of Florida Sand Skinks caught at each array and the scores from the first principal component for each array (Wald X 2 =20.86, df=1, p<0.01) (Fig. 11). Based on the loadings, low scores on this first com ponent (PC1) corresponded to high canopy cover and high amounts of groundcover made up of dead vegetation and coarse woody debris. Higher PC1 scores were associated with a large amount of bare ground and high light exposure. Greater soil compaction, in the form of higher pressure measurements with the penetrometer, was also associated with greater PC1 scores. Neither the PC2 scores (Wald X 2 =2.13, df=1, p=0.14) nor the PC3 scores (Wald X 2 =1.13, df=1, p=0.29) were significantly correlated with the number of captures at each array. Greater PC2 scores were associated with more live vegetation and lichen as groundcover, as well as a greater depth at which the penetrometer hit an area of high compaction, indicating looser soils. Bare ground was associated with a low PC2 score in addition to its high PC1 score. PC3 loadings were greatest for

PAGE 25

18 penetrometer pressure and depth, both of which were negative. In theory, pressure and depth should be inversely related, and thus, PC3 was not considered an important indicato r of the variation in environmental data. Moreover, there was a significant positive correlation between the percent soil moisture and the number of individuals captured at each array at the p = 0.1 level. (Wald X 2 =3.05, df=1, p=0.08) (Fig. 12), and the p ercent of soil moisture was positively correlated with PC1 scores at each array (Spearman's rho=0.19, p<0.05) but not PC2 scores (Spearman's rho=0.10, p>0.05). A Poisson regression model was used for the above analyses involving the number of individuals c aptured at each array because of the "Poisson like" distribution of those captures (Fig. 9). However, because the distribution of captures was significantly different than a Poisson distribution ( X 2 =45.04, df=8, p<0.00), a feature in the program Statistic a that accounts for "over dispersion" was used for better model fit. Recipient Site Significant differences exist between treatments for all of the variables measured except the temperature measurement. Table 1 displays these differences. Significant d ifferences in the environmental variables among replicates are displayed in Table 2 (with statistical values). The clustering analysis grouped at least two replicates from each treatment together. One replicate from each of the treatments "Tree", "Tree+Woo d", and "Shadecloth" was separated from the other two replicates and grouped with the "Control" enclosures, forming the major two groups that resulted from this analysis. The "Wood" replicates were grouped together and equally close to both of the major gr oupings (Fig. 13). Most of the "supplementary" component scores from the recipient site were outside of the range of the scores from the donor site (Fig. 14). The recipient site had significantly lower PC1 scores (U=20791.0, df=1, p<0.01) and significantl y higher PC2 scores (U=9446.0, df=1, p<0.01). However, all of the enclosures without trees were within the range of scores at the donor site for the first principal component, which corresponded to the amount of canopy, detritus, bare ground, and light. Al l of the enclosures with trees were within the range of the donor site PC2 scores, which corresponded to the amount of bare ground, live vegetation, and soil compaction.

PAGE 26

19 Furthermore, the percent soil moisture was not significantly different between the don or and recipient sites (U=4179.0, df=1, p=0.22). Survival Because more than 300 Florida Sand Skinks were captured at the donor site, each experimental enclosure at the recipient site had the maximum number of 20 individuals released into it. The number of males, females, and juveniles released into each enclosure, as well as how the individuals captured at each area of the donor site were divided among enclosures, is shown in Table 3. Of the 300 Florida Sand Skinks released into the enclosures, 105 had bee n recaptured after two years of trapping efforts during their active season, corresponding to a 35% recapture rate. The null models used throughout the analyses in the program MARK estimated the overall survival rate to be 49.1% (with a 95% CI of 35.2 63.2 ) when time dependence was not incorporated into recapture probability estimation ( P. ) or 69.3% (with a 95% CI of 49.2 84.0) when it was ( P T ). There were 69 unique recaptures in 2008 and 62 unique recaptures in 2009, with 26 of those occurring in both years. This corresponds to annual recapture rates of 23.0% in 2008 and 20.7% in 2009. Capture frequency, or the number of individuals captured per trapping event, ranged from 0 to 13 individuals per "block" (set of five enclosures, one from each treatment) and the times of peak capture frequency in both 2008 and 2009 (Fig. 15) were similar to peak capture frequency at the donor site (Fig. 8). Treatment and Block Effect I found no significant treatment (F=1.57, df=4, p=0.27) or block ( F=0.89, df=2, p=0.4 5) effect on the number of unique recaptures within each enclosure totaled for both years (Figs. 16 and 17). No effect of treatment (F=1.49, df=4, p=0.29) or block (F=0.10, df=2, p=0.91) on the number of recaptures existed in 2008. In 2009, there was again no difference in the number of recaptured Florida Sand Skinks among treatments (F=1.31, df=4, p=0.34), but there was a significant difference among blocks (F=6.56, df=2, p=0.021). The randomized block design does

PAGE 27

20 not allow testing for a significant treatm ent block interaction because of lack of replication within blocks, but Figure 18 graphically displays this relationship and suggests a significant interaction for the total recaptures. Model comparison in MARK showed that recapture probability estimation was best when time dependence was incorporated into the model ( P T ) (Table 4). According to QAICc values and LRTs, the two models that incorporated variation in recapture probability between blocks and years ( P B*T and P B*T*BxT ), were equally as parsi monious as model P T These three models were all significantly better than the null model, P ., according to QAICc comparison and LRTs. These comparisons were made after adjusting c hat to 1.69 because of lack of fit of the most parameterized model to t he data. Based on these results, I decided to run three sets of candidate models to test for the effect of enclosure grouping on estimation: one set that allowed P to vary by year in each model ( P T ), one set that allowed P to vary between years and blo cks in each model ( P B*T ), and another that included block and time variation in P estimation as well as its interaction ( P B*T*BxT ). When performing the GOF bootstrap test of models E P B*T and E P B*T*BxT the most parameterized models of the latter two sets, the lack of fit was too severe to include them. I proceeded using only the model set that incorporated time dependence in P estimation, whose most parameterized model ( E P T ) adequately fit the data after adjusting c hat to 1.82. The model that varie d estimation by block ( B P T ) was equal in parsimony to the null model ( P T ) in which only one estimate of survival rate was generated for the entire population (Table 5). Each of these models was significantly better than models that incorporated treatm ent ( R P T ) or treatment and block together ( R*B P T ) according to QAICc values. These models were, in turn, significantly more parsimonious than one that estimated for each enclosure ( E P T ). Because all models were equally or less parsimonious than the nu ll model ( P T ), it can be inferred that there was no effect of treatment or block on survival. The estimates and 95% confidence intervals generated by MARK were plotted for models R P T R*B P T and E P T (Fig. 19). Of the models that separated encl osures into two groups based on the presence or absence of coarse woody debris, shade, or trees (Table 6), the model that accounted for shade

PAGE 28

21 ( shade P .) was significantly more parsimonious than the null model ( P .) according to QAICc values and LRT proced ures. When time dependence was incorporated into P estimation, the model that accounted for shade was the most parsimonious, but not at the level of significance seen when time dependence was not incorporated into P estimation. The same c hat was used for this set of models that was used for the models above. The two way ANOVAs based on these groupings of enclosures also showed a significant effect of shade on the number recaptured Florida Sand Skinks (F=7.59, df=1, p=0.02). No significant effect of block ( F=0.75, df=2, p=0.50) or interaction (F=2.50, df=2, p=0.14) exists for this comparison. For the total number of recaptures within years, the effect of shade on recaptures was seen in 2008 (F=8.19, df=1, p=0.02) but not in 2009 (F=2.79, df=1, p=0.13). Effe ct of Environment Spearman correlation coefficients revealed a significant negative relationship between the recapture rate (totaled over both years) and the average soil moisture in each enclosure (rho= 0.53, p<0.05) (Fig. 20). A significant negative cor relation also existed between recapture rate and the percent of time the temperature readings were within the Florida Sand Skink's preferred temperature (rho= 0.55, p<0.05). Similarly, of the models run in MARK that used the enclosure averages of the envir onmental variables as covariates to constrain estimation (Table 7), the same two variables were the only ones that improved model parsimony over the null model ( P T ). Only the model that accounted for soil moisture was significantly different from the n ull according to QAICc comparison and LRT. Model comparisons were made after c hat was adjusted to 1.82 based on the goodness of fit of the most parameterized model E P T Effect of Sex, Age Class, and Size Of the 300 Florida Sand Skinks released into enc losures at the recipient site, 195 of them were males, 88 were female, and 17 were juveniles. Although there was variation in sex and age ratios within each enclosure, all enclosures received more males than females and more females than juveniles (Table3) Females, males, and juveniles had recapture rates of 30.3%, 36.1%, and

PAGE 29

22 47.1%, respectively. The arrangement of the data for the corresponding MARK analysis caused a significant lack of fit, and therefore models were not run in MARK. However, a Kruskal Wa llis analysis of variance revealed significant differences in size between these groups (Fig. 21), based on both SVL (H=74.56, df=2, p<0.01) and mass (H=69.85, df=2, p<0.01). Mann Whitney U test used for post hoc comparisons showed that females were signif icantly larger than males (SVL, U=5102.0, df=1, p<0.01; mass, U=5434.0, df=1, p<0.01) and males were significantly larger than juveniles (SVL, U=15.0, df=1, p<0.01; mass, U=11.0, df=1, p<0.01). No significant difference exists between the mass (U=9602.5, df=1, p=0.38) or SVL (U=8920.0, df=1, p=0.07) of Florida Sand Skinks that were recaptured and those that were not. When size was incorporated into models in MARK ( c hat adjusted to 1.69), no model was significantly more parsimonious than the null model (Ta ble 8). Effect of Trap Location The recapture rates of Florida Sand Skinks captured at each site within the donor site are as follows: 35.8, 35.6, 30.9, and 37.1% for each of the four larger sites and 42.9 and 20% for each of the two smaller sites. The number of individuals from each site released at the donor site into each enclosure can be found in Table 3. After adjusting c hat to 2.57 to account for lack of model fit to the data ( trapsite P T ), the null models in MARK were significantly more parsimoni ous than those that accounted for capture site (Table 9). Reproduction Thirty two unmarked Florida Sand Skinks were captured in total, 15 individuals in 2008 and 17 individuals in 2009 (Figs. 22 and 23). Four of the new juveniles captured in 2008 were r ecaptured in 2009. No effect of treatment (F=0.54, df=4, p=0.71) or block (F=0.91, df=2, p=0.44) on reproduction was found for the number of recruits totaled over both years. The number of new individuals found each year was also not significantly affected by treatment (2008, F=1.24, df=4, p=0.37; 2009, F=0.28, df=4, p=0.89). Furthermore, this lack of treatment effect was seen regardless of whether enclosures were grouped into five treatments or groups of two based on

PAGE 30

23 the presence or absence of trees (F=0.5 5, df=1, p=0.47), shade (F=0.95, df=1, p=0.35), or wood (F=0.90, df=1, p=0.36). A significant negative association exists between the total number of new skinks found in each enclosure and the enclosure average for light intensity (rho= 0.52, p<0.05) and p enetrometer pressure (rho= 0.72, p<0.05), indicating increased reproduction with loose soils and low light (Fig. 24).

PAGE 31

24 Discussion Donor Site The pattern of Florida Sand Skink captures at the donor site exhibited aggregation, howev er their association with any one microhabitat feature is unclear. While there was a significantly positive correlation between the number of individuals captured at each array and array PC1 scores, one outlier calls this relationship into question (Fig. 1 1). The single array where the most Florida Sand Skinks were captured (nine individuals, Fig. 9) had a lower PC1 score than the average of arrays where no individuals were found. The first component arranged the trap arrays along a spectrum from low lit ar eas containing trees (higher canopy cover) with woody debris and detritus under them at one end (low scores) to open, sunny areas with little to no groundcover at the other (high scores) (Fig. 10). Although overall more individuals were captured in the ope n, bare areas, one array surrounded by dead groundcover under a high tree canopy captured the most. The second principal component mainly described the groundcover, from bare ground at one end (low scores) to high vegetative groundcover at the other (high scores). Because there was no significant correlation between this component and the number of captures at each array, it can be inferred that within the open areas where individuals were most often found (PC1) the presence of vegetative groundcover had n o influence on their abundance (PC2). This lack of aversion to vegetative groundcover was likely an important factor in their success at the recipient site (discussed below). A significant positive correlation exists between the percent of soil moisture an d the number of captures at each array (Fig. 12). However, soil moisture was also positively correlated to PC1 scores at each array (Fig. 11), making it difficult to know which feature contributed to the number of captures. Furthermore, because of the sign ificant negative correlation between soil moisture and recapture rate seen at the recipient site (Fig. 20), soil moisture may not be an accurate indicator of Florida Sand Skink presence.

PAGE 32

25 Although the trapping efforts at the Davenport scrub resulted in the greatest number of Florida Sand Skink captures to date (Fig. 8 and 9), no habitat feature(s) strongly predicted where individuals would be captured the most. Nevertheless, based on the extensive sampling in this study, it is difficult to conclude that the se results only add to the "confusion" about optimal Florida Sand Skink habitat. Instead, they likely support the inferences made by authors of previous studies that have also found no strong correlations between Florida Sand Skink density and specific mic rohabitat features (Hill 199, McCoy et al. 1999, Gianopulos 2001). Yet, which of these inferences my results support is unclear: (1) that this species does not have one optimal scrub or sandhill habitat type and is capable of thriving in a variety of condi tions or (2) that heterogeneous scrub and sandhill ecosystems such as the donor site are optimal habitats? The donor site contained large open areas interspersed with areas of dense canopy cover and areas covered in grasses or palmettos (Fig. 2). The exis tence of these varying types of microhabitats within close proximity to each other may be necessary to meet the Florida Sand Skink's locomotive, thermoregulatory, and nutritional needs, but the prevalence of individuals across all microhabitats within the donor site may instead indicate a general ability to thrive in all of them. Recipient Site Overall, habitat features within the enclosures at the recipient site differed from habitat features at the donor site. The component scores assigned to the rec ipient site data points based on the donor site PCA loadings categorized very few of the enclosures as comparable to the Davenport scrub habitat (Fig. 14). Scores from both components significantly differed between the sites. All of the enclosures, however had scores that matched the donor site scores for at least one of the two principal components. In general, the enclosures that contained trees had a percentage of bare ground and live vegetative groundcover that was within the range seen at the donor si te (PC2), while their amount of canopy cover and detritus exceeded donor site levels (PC1). The enclosures without trees had measures of canopy cover, detritus, bare ground, and light intensity similar to those at the donor site (PC1), but the amount of ve getative groundcover was greater (PC2).

PAGE 33

26 Significant differences exist among replicates of treatments in the environmental variables measured (Table 2). The treatments without trees differed in the proportion of bare ground, vegetative groundcover, and det ritus. Light intensity and soil compaction varied among treatments containing trees, likely the result of variation in the location of the tree(s) within each enclosure. Because the tree canopy did not cover the entire 20 m 2 area, some of these enclosures received more light than others. Similarly, the effect of the tree's root system was likely not evenly distributed throughout the enclosed space, resulting in differing levels of compaction. The "Wood" treatment enclosures had the most variation among them in environmental variables, and as a result they were not grouped with any other treatment in the clustering analysis, and were equally related to the two major groupings (Fig. 13). The "Control" enclosures were the only replicates not separated into the se two groups. The group containing the "Control" replicates also contained one replicate from each of the other treatments (except "Wood"). Nevertheless, at least two of the replicates of each treatment were grouped together, allowing confidence in the ef fect of the assigned treatments on survival if one was indeed found. These differences among replicates, however, made an alternative analysis necessary, which tested for correlation between the number of recaptures and the environmental variables measured for each enclosure individually, ignoring treatments altogether. Testing for variation in each environmental variable among treatments revealed that the enclosure manipulations resulted in differences beyond the presence or absence of trees, shade, and c oarse woody debris (Table 1). For the two primary manipulations, the inclusion of trees and coarse woody debris, there were clear significant differences between those treatments with and without these features. Light intensity also differed, as expected, between the enclosures with trees or shade cloth and those with neither (although "Shadecloth" and "Control" were not significantly different). No other variable showed these clear divisions. For instance, the depth the penetrometer reached before reachin g highly compacted ground was greatest in the "Wood" and "Shadecloth" treatments (indicating looser soils), lowest in the "Tree" treatment, with the "Control" and "Tree+Wood" treatments not differing from either extreme. The percent of ground covered by de tritus was greatest in the treatment with trees, lowest in the treatments without trees, and

PAGE 34

27 intermediate in the "Shadecloth" treatment. Moreover, the percent of time that the ground temperatures were within the Florida Sand Skink's preferred range did not differ between treatments. Survival The recapture of the translocated Florida Sand Skinks did not differ among the five treatments (Figs. 16 and 17). Similarly, no environmental variables were correlated with recaptures except soil moisture (Fig. 20) a nd the percent of time the ground temperature was within their preferred temperature during the activity season, whose contributions to success within the enclosures are questionable. Although significant according to Spearman correlation test, MARK analys es did not support the relationship between temperature and survival (Table 7). Secondly, the relationship was negative, indicating increased survival with more time outside of the preferred temperature range, an unlikely causal relationship. Additionally, there was no difference among treatments for soil temperature (Table 1), and thus, even if it was considered a contributing factor to their success, it cannot be attributed to any habitat feature in the enclosures. The significant negative correlation bet ween recaptures and soil moisture was supported by MARK (Table 7), but the positive relationship between soil moisture and captures at the donor site (Fig. 12) makes it an unreliable predictor of success. Significantly more recaptured Florida Sand Skinks were found in the enclosures without shade, either in the form of a tree or shade cloth, when compared directly with the enclosures with shade. This division of treatments corresponds with the differences in measures of light intensity and soil compaction (measured as penetrometer pressure) (Table 1). As mentioned above, the differences between the two groups for these variables involve some overlap, but they are the only environmental variables measured that show this trend. While creating differences in light intensity between these two groups was one of the goals of the treatment manipulations, the differences in soil compaction was not expected. The non shaded, more successful treatments had higher penetrometer pressure values in general than shaded tre atments, indicating more compact soils. These same two measures were both associated with high PC1 scores at the

PAGE 35

28 donor site, and this component was positively correlated with the number of captures per trap array. This finding indicates a similarity in the features associated with success at the translocation site and highly occupied areas of the donor site. Finally, an individual's chance of surviving, regardless of the enclosure in which it was released, was not affected by its size, sex, age, or where wa s captured at the donor site. Reproduction The number of unmarked juveniles found within enclosures did not significantly differ among treatments or blocks or between any two groups of treatments (such as those with and without shade) (Figs. 22 and 23). Light intensity and penetrometer pressure were the only two environmental variables correlated with the number of individuals born in each enclosure, and it was a significantly negative relationship in both cases (Fig. 24). These are the same two variable s that were significantly higher in the enclosures without shade, where recapture rates were also higher. This relationship suggests that the factors that contribute to survival can negatively influence reproduction for this species, a notion that has not been supported by any study. The significant correlations found between the number of new juveniles and light intensity and soil compaction measurements within each enclosure are likely the result of low numbers of new juveniles found in each enclosure (be tween zero and five) and the number of statistical tests performed. Measures of Success / Recommendations Although few of the treatment manipulations and microhabitat features measured could be directly associated with increased recapture rate within enc losures, it is not for lack of survival. All 15 enclosures had recaptures during each trap year, and the overall recapture rate for the entire translocated population was 35.0%. Estimations of survival rate from models fit to the recapture data using the p rogram MARK were as high as 69.3%. Furthermore, yearly recapture rate did not increase or decrease substantially between the first and second trapping years, and the time of highest capture frequency each year at the recipient site matched the period of

PAGE 36

29 hi ghest capture frequency at the donor site (Figs 8 and 15). The translocated population exhibited successful reproduction as well. Newly born individuals were found in at least one replicate from each treatment during both years, and the number of new indiv iduals found did not differ greatly between years. The survival of juveniles born at the recipient site to sexual maturity was confirmed when 4 of the 15 new juveniles captured in 2008 were recaptured in 2009. Previous reviews of translocation programs ha ve measured success based on the establishment of a self sustaining population that has been monitored for a sufficient amount of time for the viability of the population to be confirmed (Griffith et al. 1989, Dodd and Seigel 1991, IUCN 1998). A recent rev iew by Germano and Bishop (2009) used two criteria to assess whether a self sustaining population had been established as a result of translocation efforts: (1) evidence of successful reproduction at the donor site that resulted in "a substantial addition of new recruits to the adult population" and (2) monitoring for at least the length of time it takes for the translocated species to reach maturity. Based on these criteria, the transloc ation of the Florida Sand Skink discussed in this paper has been a su ccess. After two years of monitoring, the number of new recruits was greater than 10% of the size of the initial translocated population. Because the sex ratio of the 20 individuals released into each enclosure was greatly skewed toward males (Table 3), th is number of new juveniles found is even more significant. The 32 recruits found were the result of reproduction in only 82 females, corresponding to a 39% female birth rate. Currently, the published age at which Florida Sand Skink's reach sexual maturity is 19 to 22 months (Telford 1959, Ashton 2005), but the population used in this translocation showed quicker maturation in males. Ejaculation upon palpation of the hemipenes occurred in the male individuals born in the enclosures in 2008 and recaptured in 2009, indicating sexual maturation within approximately 9 months of hatching. By either measure, the length of time the translocated population was monitored in this study was at least the length of time it takes for the Florida Sand Skink to reach sexual maturity. For this species, the estimated survival rate is particularly important to take into account when measuring success. The Florida Sand Skink is highly elusive and difficult to detect because of its fossorial lifestyle and very small size (approx. maximum of 65 mm). While several

PAGE 37

30 individuals were recaptured multiple times within a single trap year, many first time recaptures were still being found by the end of the trapping seasons, making the probability of missed recaptures likely. The high freque ncy of non recaptured survivors was supported when more than half of the recaptures during the second year were not captured during the previous trapping season. Therefore, the recapture rate highly underestimates survival in this study. Survival rate afte r two years of monitoring, as estimated by Cormack Jolly Seber models in MARK, was likely between 50% and 70%, a level that suggests the Florida Sand Skinks ability to survive in the conditions within the enclosures at the recipient site. As with the capt ure data at the donor site, the recipient site data supports one of two ideas about optimal Florida Sand Skink habitat: that it does not exist or that heterogeneity is the most important characteristic. Their ability to survive and reproduce equally well i n each of the five treatments supports the former notion. Not only were the habitat features in these treatments very different from one another (Table 1), but they were also generally different from the habitat features at the donor site (Fig. 14). The on e common feature among enclosures and between the donor and recipient sites was structural heterogeneity. Yet, if heterogeneity were the most important factor for their success, one would assume that survival and reproduction would have been highest in enc losures with either trees or shade cloth because of the naturally higher heterogeneity in these enclosures (see error bars in Fig. 14) caused by the presence of a shade providing object that did not cover the enclosure entirely (Figs. 4 and 5). However, th e non shaded enclosures, which had generally higher recapture rates compared to shaded enclosures, were heterogeneous on a smaller scale, with interspersed patches of grass, dead vegetation, and bare ground. Heterogeneity at this scale may be all that is n ecessary for Florida Sand Skink survival, and perhaps this smaller scale heterogeneity contributed to the greater survival in these enclosures. Live vegetation, whether tall trees or thick grasses, may serve the same functions of providing shade and refuge retaining moisture, and attracting prey insect species. Hill (1999) found that the presence of a single vegetative stratum, either from trees or understory vegetation, was associated with Florida Sand Skink density in the wild. The presence of both under story and tree canopy together, however, was not associated with the presence of Florida Sand Skinks,

PAGE 38

31 most likely because it does not provide heterogeneity caused by the lack of interspersed open, bare areas. The results from the donor and recipient sites support this. First, areas with both an understory and a tree canopy did not exist at the highly populated donor site. Moreover, this type of microhabitat was recreated at the recipient site in the "Shadecloth" treatment enclosures, which contained dense g rass cover as well as a simulated canopy. My data reveal that Florida Sand Skinks were found in the bucket traps underneath the shade cloth only twice out of 55 total captures (not unique recaptures) in "Shadecloth" enclosures. The ability of Florida San d Skinks to survive where there are no layers of vegetation or canopy is an important consideration for future translocation programs because this type of area may be available for possible recipient sites. While many of the Florida Sand Skinks captured at the donor site were found in open, bare areas, these areas were always adjacent to areas with grass, shrub, or tree cover indicative of the patchy nature of the Davenport scrub (Fig. 2). This observation cannot confirm the ability of this species to survi ve for long periods with no vegetative cover. Although none of the recipient site enclosures could be placed in this category according to the environmental variables measured, survival in bare, open habitat was seen in this study. Because the environmenta l variables were measured during the second trapping year, the fact that the enclosures without trees contained little to no vegetative cover at the time the Florida Sand Skinks were released into them was not represented in the data. These bare, open cond itions persisted in these enclosures for approximately six months until the rainy season promoted the rapid spread of grasses throughout the sunny areas, probably in response to the disturbance of soils during the installation of the enclosures in the mont hs prior to the translocation. The grasses remained for the remainder of the study, but some conclusions can be drawn from this short time period before they spread. There was no sign of decreased survival or reproduction in these enclosures the following year compared to the shaded enclosures. Instead, these two enclosures without shade had significantly more recaptures than those with shade, an effect seen during this first year but not the second. This trend is most astonishing when considering that: (1) the months after their release until the rainy season are the hottest of the year (May, June, and July), (2) this period was probably the most critical in terms of the stress put

PAGE 39

32 on the animals caused by the translocation, and (3) this time coincided with maternal brooding of eggs and juvenile emergence from the nests. The fact that neither survival nor reproduction suffered during this period indicates that Florida Sand Skinks are able to thrive in this microhabitat at least on a short term basis. But was long term success in these enclosures dependent on the grass that eventually grew because of the apparent need for heterogeneity? If grass had not grown and long term survival was recorded in these enclosures, the importance of structural heterogeneity re sulting from a single stratum of vegetation would be refuted. Instead, it would seem that the presence of at least some open, bare areas would be the most important factor for Florida Sand Skink success rather than heterogeneity. While the data gathered f or this experiment cannot answer this question, it is worth noting that vast expanses of bare, open areas are not found in the scrub and sandhill habitats on the Lake Wales Ridges or other ridges in central Florida. They are always interspersed with areas of vegetation. Therefore knowing the answer to the above question is not necessarily important when choosing how a reclaimed scrub will be altered to make translocated Florida Sand Skink survival more likely because the aim of reclamation should be to re c reate habitat conditions similar to the original environment. The recipient site in this translocation differed from the donor site in the overall level of vegetative cover, whether in the form of trees or grasses. Also, species composition did not match t hat of the donor site, as this was not the goal of the treatments given to the enclosures. Despite these differences, structural heterogeneity in the form of a single stratum of vegetation was present throughout the recipient site, and this heterogeneity l ikely contributed to the overall success. Woody debris was provided, but had no effect on success because features associated with the vegetation seemed to adequately perform the functions of providing refuge, moisture, and prey. Future translocation effo rts involving the Florida Sand Skink should focus on providing open, bare areas interspersed with areas of a single vegetative stratum. Sites with this structural heterogeneity should be considered as recipient sites before sites that contain thick vegetat ive structure, with both an understory and canopy but no bare ground. This study shows that the presence of all of the same structural features found at a donor site will be an adequate indicator

PAGE 40

33 of quality of the recipient site, even if the proportion of each of those structural features is not the same at each site. Therefore, the measurement of these features at each site is recommended. Overall, the relationship between captures and microhabitat at the donor site and success at the recipient site show that the Florida Sand Skink can thrive in multiple differing microhabitat types and that heterogeneity in and around these microhabitats is important. Therefore, the conclusions of other researches studying this species can be synthesized into a common hy pothesis: optimal Florida Sand Skink habitat does not exist because the necessary structural heterogeneity can be found on the microhabitat level in a multitude of habitat types. Gianopulos (2001) came to a similar conclusion. The Florida Sand Skink's morp hological specialization, limited geographic range, and elusiveness in the wild all lead to the assumption that it is a habitat specialist. It seems, though, that t he Florida Sand Skink is able to take advantage of several scrub and sandhill habitat typ es within its range, which contributed to its success across all of the microhabitat types provided at the recipient site in this experiment.

PAGE 41

34 Figures Fig. 1. Donor and recipient site location in central Florida. The L ake Wales Ridge runs down the center of central Florida, and the white dot denotes the area of the donor and recipient sites (approx. 24 km apart).

PAGE 42

35 Fig. 2. Trap array locations at the donor site. The 210 trap arrays were divided among six sites thr oughout the approximately 1.2 km 2 scrub habitat near Davenport, FL. The large body of water in the upper left corner of the photo is part of the mining operation, which would eventually spread to capture sites. Photo by Aerials Express courtesy of ENTRIX I nc.

PAGE 43

36 Fig. 3. Schematic of trap arrays at donor site. Each trap array consisted of four two meter drift fences with two countersunk buckets at either side of each, totaling 16 one gallon buckets per array (represe nted by circles). Each array was approximately 8 m 2 in size. The squares mark the placement of 1 m 2 quadrats for groundcover, canopy height, and light intensity data collection. Fig. 4. Example of a marked Florida Sand Skink. A plastic po lymer was injected under the skin in three of six possible locations using a combination of four possible colors. Each individual was given a unique marking that also indicated the treatment to which it was initially assigned.

PAGE 44

37 Fig. 5. Photos of recipient site treatments. From left to right, they are: (1) no tree or woody debris ("Control"); (2) woody debris with no tree ("Wood"); (3) woody debris and shade from tree ("Tree+Wood"); (4) shade from tree but no woody debris (" Tree"); and (5) shade from elevated Shadecloth ("Shadecloth"). Fig. 6. Arrangement of treatment replicates at recipient site. Fifteen 20 m 2 enclosures were divided among five treatments. A randomized block design was used, in whic h each row of five enclosures includes one replicate from each treatment in different order.

PAGE 45

38 Fig. 7. Schematic of enclosures at the recipient site. Each circle represents a one gallon bucket trap (76 total), which was either positioned along the perimeter or at either end of a 2 m drift fence. The gray squares show where a 1 m 2 quadrat was laid for groundcover, light intensity, and canopy height data collection.

PAGE 46

39 Fig. 8. Capture frequency at the donor site in 2007. All tra ps at each of the six areas of the donor site were checked at each capture event. The time of peak capture frequency corresponded to the time of peak capture frequency at the recipient site in 2008 and 2009 (Fig. 15). Fig. 9. Distribution of Florida San d Skinks captured per trap array at the donor site. Five hundred and ten individuals were captured at 171 of the 210 trap arrays.

PAGE 47

40 Fig. 10. PCA loadings for each variable using donor site environmental data. The first two components accounted for 57% of the variation in the data. Fig. 11. Principal component scores for trap arrays at the donor site. Each point is the average (with standard error) for the arrays at which a certain number of individuals were caught. There was a significant positive corr elation between the number of Florida Sand Skinks caught at arrays and PC1 scores (Wald X 2 =20.86, p<0.01) but not PC2 scores (Wald X 2 =2.13, p=0.14).

PAGE 48

41 Fig. 12. Soil moisture and captures at the donor site. There was a positive (although not significant ) relationship between the number of Florida Sand Skinks captured at each array and the % soil H 2 O measured there ( Wald X 2 =3.05, p=0.08) determined by Poisson regression. Fig. 13. Hierarchical clustering analysis using environmental data measured withi n enclosures. Euclidean distance was used to assess similarity for clustering. Sha decloth 1 Shadecloth 2 Tree 2 Tree+Wood 2 Tree+Wood 1 Shadecloth 3 Control 1 Control 2 Control 3 Tree 3 Tree+Wood 3 Tree 1 Wood 1 Wood 2 Wood 3

PAGE 49

42 Fig. 14. PCA scores for enclosures used as supplemental cases. Enclosure averages (with standard errors) are shown for PC1 and PC2. The gray boxes indicate the range of scor es occupied by the donor site trap array on each axis. The recipient site had significantly lower PC1 scores (U=20791.0, df=1, p<0.01) and significantly higher PC2 scores (U=9446.0, df=1, p<0.01) when compared with the donor site, according to Mann Whitney U tests Fig. 15. Capture frequency at the recipient site in 2008 and 2009. Enclosures were checked in groups of five, one replicate from each treatment (i.e. block). Each point represents the number of skinks captured in one block in one day.

PAGE 50

43 Figure 1 6. Recaptures of translocated Florida Sand Skinks by enclosure and year. Twenty individuals were translocated to each enclosure. There was no significant treatment (F=1.57, df=4, p=0.27) or block (F=0.89, df=2, p=0.45) effect on recapture rate, determined by one way ANOVA. Figure 17. Recaptures of translocated skinks totaled by treatment. Enclosures that contained shade in the form of trees or shade cloth had significantly less recaptures than enclosures without shade when treatments were grouped for thi s test (two way ANOVA, F=7.60, df=1, p=0.02).

PAGE 51

44 Fig. 18. The interaction between treatment and block for recaptures per enclosure. Because of the crossing of lines, an interaction effect can be inferred.

PAGE 52

45 A B C Fig. 19. Sur vival rate estimates (with 95% confidence intervals) from models R P T (A), R*B P T (B), and E P T (C). The second model's enclosure estimates are based on the assignment of each model to a treatment and block while the third model treats each enclosure indiv idually. None of these models was more parsimonious than the null ( P T ).

PAGE 53

46 Fig. 20. Relationship between recaptures and soil moisture at the recipient site. Percent soil H 2 O was the only variable that was significantly correlated with the number of recapt ures within each enclosure (Spearman's rho= 0.53, p<0.01) and also supported by MARK analyses. The same correlation was significant at the donor site, except it was a positive rather than a negative relationship (Fig. 12). The line was added for visual cla rity.

PAGE 54

47 A B Fig. 21. Size of females, males, and juveniles translocated into enclosures. Kuskal Wallace tests revealed significant differences between all groups for both SVL (H=74.56, df=2, p<0.01) (A) and mass (H=69.8 5, df=2, p<0.01) (B). All pairs were significantly different according to post hoc tests. Error bars are one standard deviation of the mean.

PAGE 55

48 Fig. 22. Unmarked juvenile Florida Sand Skinks found in each enclosure by year. There was no eff ect of treatment (one way ANOVA, F=0.54, df=4, p=0.71) or block (one way ANOVA, F=0.91, df=2, p=0.44) on reproduction totaled for both years. Fig. 23. Unmarked juvenile Florida Sand Skinks found in each enclosure totaled by treatment. An effect of treatm ent on reproduction was not seen for either year (one way ANOVA; 2008, F=1.24, df=4, p=0.37; 2009, F=0.28, df=4, p=0.89).

PAGE 56

49 A B Fig. 24. Correlation between reproduction and environmental variables within enclosures. There was a significant negative rela tionship between the number of unmarked individuals found in each enclosure and (A) light intensity (Spearman's rho= 0.52, p<0.05) and (B) the amount of pressure exerted on the penetrometer (Spearman's rho= 0.72, p<0.05), indicating increased reproduction with loose soils and low light.

PAGE 57

50 Tables Table 1. Differences in environmental variables among treatments. Control Wood Tree+Wood Tree Shadecloth Canopy Height Light Intensity Preferred Temperature Penetrometer Depth Penetrometer Pressure % Soil H 2 0 % Live Vegetation % Dead Vegetation % Bare Ground % Woody Debris Differences were determined using one way ANOVA. Tukey's HS D was used for post hoc comparisons. Differences were considered significant at the p =0.05 level. Differing colors indicate significant differences, with the black boxes signifying the highest values, gray boxes signifying intermediate values, and white bo xes signifying the lowest values. Striped boxes indicate that the treatment was not significantly different from any of the treatments that contain the corresponding colors that make up the stripes.

PAGE 58

51 Table 2. Differences in environmental variables among treatment replicates. Control Wood Tree+Wood Tree Shadecloth Canopy Height N/A F=1.00 p=0.38 F=3.06 p=0.06 F=4.09 p=0.02 N/A Light Intensity F=0.35 p=0.71 F=7.12 p=0.00 F=14.08 p=0.00 F=5.29 p=0.01 F=0.97 p=0.39 Penetrometer Depth F=0.25 p=0.78 F=9.48 p=0.00 F=5.76 p=0.01 F=6.26 p=0.00 F=2.36 p=0.12 Penetrometer Pressure F=0.16 p=0.84 F=26.03 p=0.00 F=2.33 p=0.11 F=1.42 p=0.25 F=2.90 p=0.07 % Soil H 2 0 F=1.73 p=0.26 F=6.00 p=0.04 F=0.11 p=0.90 F=0.27 p=0.77 F=2.49 p=0.16 % Live Vegetation F=21 .81 p=0.00 F=1.30 p=0.00 F=2.43 p=0.10 F=4.02 p=0.02 F=10.30 p=0.00 % Dead Vegetation F=17.15 p=0.00 F=1.44 p=0.00 F=0.05 p=0.95 F=0.98 p=0.38 F=6.67 p=0.00 % Bare Ground F=0.53 p=0.60 F=0.81 p=0.44 F=1.40 p=0.26 F=2.35 p=0.11 F=10.88 p=0.00 % Woody Deb ris N/A F=0.04 p=0.96 F=0.28 p=0.76 F=7.62 p=0.00 N/A Differences were determined using one way ANOVAs. For all tests, degrees of freedom = 2.

PAGE 59

52 Table 3. Distribution of sex, age class, and capture locations of individuals relocated to ea ch enclosure. Sex/Age class ratios Number from each area at donor site Enclosure # male # female # juveniles Site 1 Site 2 Site 3 Site 4 Site 5 Site 6 Shadecloth 1 14 6 0 6 6 3 5 0 0 Shadecloth 2 13 6 1 6 8 2 2 0 2 Shadecloth 3 15 4 1 6 4 2 7 1 0 Wood 1 12 7 1 6 8 1 5 0 0 Wood 2 14 4 2 4 7 5 4 0 0 Wood 3 15 4 1 3 3 8 6 0 0 Control 1 14 6 0 7 4 9 0 0 0 Control 2 12 7 1 2 11 0 4 3 0 Control 3 9 6 5 13 3 2 2 0 0 Tree 1 16 3 1 4 5 8 2 1 0 Tree 2 11 8 1 7 4 4 4 0 1 Tree 3 11 7 2 7 5 3 3 0 2 Tre e+Wood 1 11 9 0 1 7 3 8 1 0 Tree+Wood 2 14 6 0 2 8 4 6 0 0 Tree+Wood 3 14 5 1 7 7 1 4 1 0 Total 195 88 17 81 90 55 62 7 5

PAGE 60

53 Table 4. Models compared in MARK to assess the effect of enclosure grouping and time on recapture probability estimation ( P ) Survival Rate Estimation Recapture Probability Estimation Model Parsimony Model Notation Grouping Time Grouping Time Group Time Interaction QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters QDeviance P T one constant one varied n/a 365.769 0.000 0.374 1.000 3 48.880 P B*T one constant block varied no 366.051 0.282 0.325 0.869 4 47.116 P B*T*BxT one constant block varied yes 366.935 1.166 0.209 0.558 6 43.876 P R*T one constant treatment vari ed no 369.351 3.582 0.062 0.167 7 44.212 P one constant one constant n/a 372.126 6.357 0.016 0.042 2 57.271 P R one constant treatment constant n/a 373.647 7.878 0.007 0.020 6 50.588 P R*T*RxT one constant treatment varied yes 375.135 9.366 0.0 03 0.009 10 43.685 P B one constant block constant n/a 375.325 9.556 0.003 0.008 4 56.391 P E*T one constant enclosure varied no 382.543 16.774 0.000 0.000 17 35.937 P E one constant enclosure constant n/a 384.732 18.963 0.000 0.000 16 40.328 P E*T*ExT one constant enclosure varied yes 402.128 36.359 0.000 0.000 29 28.049 Models that incorporated differences between years and blocks were the most parsimonious and differed from the null ( P .)

PAGE 61

54 Table 5. Model comparison for the effect of enclosure grouping on survival rate estimation ( ). Survival Rate Estimation Recapture Probability Estimation Model Parsimony Model Notation Grouping Time Grouping Time QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters QDeviance P T one constant one varied 340.076 0.000 0.472 1.000 3 45.389 B P T block constant one varied 340.372 0.296 0.407 0.862 5 41.583 R P T treatment constant one varied 344.083 4.007 0.064 0.135 7 41.145 R*B P T treatment + bloc k constant one varied 344.305 4.229 0.057 0.121 9 37.172 E P T enclosure constant one varied 355.225 15.149 0.000 0.001 17 30.820 No treatment or block effect could be inferred because the null model was most parsimonious. The model that included the tr eatment block interaction had the same outcome as model E P T and is not included in this table. Survival rate estimates from three of these models are plotted in Fig. 19.

PAGE 62

55 Table 6. Models testing the individual effects of the treatment s given to enclosures on survival rate estimation ( ) Survival Rate Estimation Recapture Probability Estimation Model Parsimony Model Notation Grouping Time Time QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters Q Deviance shade P T shade constant varied 356.362 0.000 0.410 1.000 4 44.614 P T one constant varied 357.452 1.090 0.238 0.580 3 47.750 tree P T tree constant varied 358.452 2.090 0.144 0.352 4 46.705 wood P T wood constant varied 359.005 2.643 0.109 0.267 4 47.258 R P T treatment constant varied 361.238 4.876 0.036 0.087 7 43.286 shade P. shade constant constant 361.255 0.000 0.559 1.000 3 51.553 P. one constant constant 363.616 2.361 0.172 0.307 2 55.947 tree P. tree constant constant 36 4.257 3.003 0.125 0.223 3 54.555 wood P wood constant constant 364.896 3.641 0.090 0.162 3 55.194 R P. treatment constant constant 365.906 4.651 0.055 0.098 6 50.034 Survival rate was estimated for two groups of enclosures for each model based on th e presence or absence of shade, trees, or woody debris. The model that grouped enclosures based on shade was significantly more parsimonious than the null, indicating an effect of shade on survival.

PAGE 63

56 Table 7. Comparison of models incorporatin g enclosure averages of environmental variables into survival rate estimation ( ) Survival Rate Estimation Recapture Probability Estimation Model Parsimony Model Notation Covariate Time Time QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters QDeviance water P T soil H 2 O constant varied 337.666 0.000 0.343 1.000 4 40.933 temperature P T temperature constant varied 339.368 1.702 0.154 0.427 4 42.636 P T none constant varied 340.076 2.410 0.111 0.300 3 45.389 canopy P T canopy constant varied 340.144 2.478 0.098 0.290 4 43.412 depth1 P T depth constant v aried 341.318 3.652 0.064 0.161 4 44.585 pressure1 P T pressure constant varied 342.027 4.361 0.048 0.113 4 45.294 bare P T bare ground constant varied 342.499 4.833 0.040 0.089 4 45.766 light P T light constant varied 342.634 4.969 0.038 0.083 4 45.902 wood P T woody debris constant varied 342.654 4.988 0.038 0.083 4 45.921 live P T live vegetation constant varied 342.657 4.992 0.028 0.082 4 45.925 dead P T dead vegetation constant varied 342.658 4.992 0.028 0.082 4 45.926 E P T none constant varied 355.225 17.559 0.000 0.000 17 30.820 The percent soil H 2 O was the only variable that significantly increased model parsimony over the null after LRT and QAICc comparisons.

PAGE 64

57 Table 8. Comparison of models with size related individual covariates. Survival Rate Estimation Recapture Probability Estimation Model Parsimony Model Notation Covariate(s) Time Time QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters QDeviance P T none constant varied 365.614 0.000 0.322 1.000 3 359.546 SVL P T svl constant varied 366.090 0.476 0.254 0.788 4 357.977 mass P T mass constant varied 366.355 0.742 0.222 0.690 4 358.243 SVL*SVL2 P T svl and svl 2 constant varied 367.641 2.028 0.117 0.363 5 357.472 mass*mass2 P T mass and mass 2 constant varied 368.284 2.671 0.085 0.263 5 358.115 SVL P. svl constant constant 370.834 0.000 0.306 1.000 3 364.767 P. none constant constant 370.950 0.116 0.289 0.944 2 366.916 mass P. mass constan t constant 371.592 0.758 0.209 0.684 3 365.525 SVL*SVL2 P svl and svl 2 constant constant 372.718 1.884 0.119 0.390 4 364.605 mass*mass2 P. mass and mass 2 constant constant 373.587 2.753 0.077 0.253 4 365.474 No model was significantly more parsimoni ous than the null model, indicating the lack of an effect of size at the time of release.

PAGE 65

58 Table 9. Model comparisons for the effect of initial capture site on survival rate estimation ( ) Survival Rate Estima tion Recapture Probability Estimation Model Parsimony Model Notation Grouping Time Time QAICc Delta QAICc QAICc Weights Model Likelihood Number of Parameters QDeviance P T one constant varied 242.500 0.000 0.993 1.000 3 11.724 trapsite P T trap site constant varied 252.466 9.966 0.007 0.007 8 11.348 P. one constant constant 245.313 0.000 0.993 1.000 2 16.571 trapsite P. trap site constant constant 255.216 9.903 0.007 0.007 7 16.190 No effect of the initial trap area within the donor site was found.

PAGE 66

59 Literature Cited Anderson, D. R. and K. P. Burnham. 1999. General strategies for the analysis of ringing data. Bird Study 46: 261 270. Andrews, R. M. 1994. Activity and thermal biology of the sand swimming skink Neoseops reynoldsi : diel and sea sonal patterns. Copeia 1994(1): 91 99. Ashton, K. G. 2005. Life history of a fossorial lizard, Neoseps reynoldsi Journal of Herpetology 39(3): 389 395. Ashton, K. G. and S. R. Telford, Jr. 2006. Monthly and daily activity of a fossorial lizard, Neosep s reynoldsi Southeastern Naturalist 5(1): 175 183. Branch, L. C., A. M. Clark, P. E. Moler, and B. W. Bowen. 2003. Fragmented landscapes, habitat specificity, and conservation genetics of three lizards in Florida scrub. Conservation Genetics 4: 199 212 Burke, R. L. 1991. Relocations, repatriations, and translocations of amphibians and reptiles: taking a broader view. Herpetologica 47(3): 350 357. Collazos, A. 1998. Microhabitat selection in Neoseps reynoldsi the Florida sand swimming skink. Unpubl. master's thesis, Univ. of South Florida, Tampa. Cooch, E. and G. C. White. 2001. An Electronic Handbook for the Program MARK. Colorado State University, Fort Collins. Dodd, Jr., C. K. and R. A. Seigel. 1991. Relocation, repatriation, and translocation of amphibians and reptiles: Are they conservation strategies that work? Herpetologica 47(3): 336 350. Fischer, J. and B. D. Lindenmayer. 2000. An assessment of the published results of animal relocations. Biological Conservation 96: 1 11. Germano, J. M. and P. J. Bishop. 2008. Suitability of amphibians and reptiles for translocation. Conservation Biology 23(1): 7 15. Gianopulos, K. D. 2001. Response of the threatened sand skink ( Neoseps reynoldsi ) and other herpetofaunal species to burning and clearcutti ng in the Florida sand pine scrub habitat. Unpubl. master's thesis, Univ. of South Florida, Tampa. Griffith, B., J. M. Scott, J. W. Carpenter, and C. Reed. 1989. Translocation as a species conservation tool: status and strategy. Science 245: 477 480. Hi ll, K. E. 1999. Responses of released populations of the sand skink, Neoseps reynoldsi to scrub habitat translocation in central Florida. Unpubl. master's thesis, Univ. of South Florida, Tampa. IUCN (World Conservation Union) 1987. IUCN position state ment on translocation of living organisms: introductions, reintroductions, and restocking. IUCN, Gland Switzerland.

PAGE 67

6 0 IUCN (World Conservation Union). 1998. Guidelines for reintroductions. IUCN/SSC Reintroduction specialist group. IUCN, Gland, Switzerland and Cambridge, United Kingdom. Lebreton, J. D., K. P. Burham, J. Colbert, and D. R. Anderson. 1992. Modeling survival and testing biological hypotheses using marked animals a unified approach with case studies. Ecological Monograph 62: 67 118. McCoy, E. D., P. E. Sutton, and H. R. Mushinsky. 1999. The role of guesswork in conserving the threatened sand skink. Conservation Biology 13(1): 190 194. McCoy, E. D., N. Ihasz, E. J. Britt, and H. R. Mushinsky. In review. Is the Florida Sand Skink ( Plestiodo n reynoldsi ) a dietary specialist? Mushinsky, H. R. and E. D. McCoy. 1991. Vertebrate species composition of selected scrub islands on the Lake Wales Ridge of central Florida. Final Rep. Number GFC 87 149. Florida Game and Freshwater Fish Commission, Non game Wildlife Program. Mushinsky, H. R. and E. D. McCoy. 1999. Studies of the sand skink ( Neoseps reynoldsi ) in Central Florida. Final report to Walt Disney Imagineering. University of South Florida, Tampa, Florida. Myers, C. W. and S. R. Telford, Jr. 19 65. Food of Neoseps the Florida Sand Skink. Quarterly Journal of Florida Academic Science 28(2): 190 194. Navratil, G. 1999. A study of selected land management practices on the sand pine scrub habitat of Florida: a measure of the effects of land mana gement on the sand skink, Neoseps reynoldsi Unpubl. master's thesis, Univ. of South Florida, Tampa. Penney, K. M. 2001. Factors affecting translocation success and estimates of dispersal and movement patterns of the sand skink Neoseops reynoldsi on resto red scrub. Unpubl. master's thesis, Univ. of South Florida, Tampa. Penney, K. M., K. D. Gianopulos, E. D. McCoy and H. R. Mushinsky. 2001. The visible implant elastomer marking technique in use for small reptiles. Herpetological Review 32: 236 241. Pike D. A, K. S. Peterman, and J. H. Exum. 2007. Use of altered habitats by the endemic sand skink (Plestiodon reynoldsi Stejneger). Southeastern Naturalist 6: 715 726. Platenberg, R. J. and R. A. Griffiths. 1999. Translocation of slow worms ( Anguis fragilis ) as a mitigation strategy: a case study from southeast England. Biological Conservation 90(2): 125 132. Reinert, H. K. 1991. Translocation as a conservation strategy for amphibians and reptiles: some comments, concerns, and observations. Herpetologica 4 7(3): 357 363. Richmond, J. Q., D. T. Reid, K. G. Ashton, and K. R. Zamudio. 2008. Delayed genetic effects of habitat fragmentation on the ecologically specialized Florida Sand Skink ( Plestiodon reynoldsi ). Conservation Genetics 10(5): 1281 1297. Rout, T M., C. E. Hauser, and H. P. Possingham. 2007. Minimise long term loss of maximize short term gain? Optimal translocation strategies for threatened species. Ecological Modeling 201: 67 74. Seddon, P. J., D. P. Armstrong, and R. F. Maloney. 2007. Developi ng the science of reintroduction biology. Conservation Biology 21(2): 303 312.

PAGE 68

61 Simon J.L. 2000. Resampling Stats: Users Guide. v5.02. Arlington, Virginia: Resampling Stats Inc. Smith. 1977. Food resource partitioning of burrowing sand pine scrub reptile s. Herpetological Review 8(3): 17. Sutton, P. E. 1996. A mark and recapture study of the Florida Sand Skink, Neoseps reynoldsi and a comparison of sand skink sampling methods. Unpubl. master's thesis, Univ. of South Florida, Tampa. Sutton, P. E., H. R. Mushinsky, and E. D. McCoy. 1999. Comparing the use of pitfall drift fences and cover boards for sampling the threatened sand skink ( Neoseps reynoldsi ). Herpetological Review 30(3): 149 151. Telford Jr., S. R. 1959. A study of the sand skink, Neoseps reyn oldsi Stejneger. Copeia 1959: 110 119. Turner, W. R., D. S. Wilcove, and H. M. Swain. 2006. Assessing the effectiveness of reserve acquisition programs in protecting rare and threatened species. Conservation Biology 20(6): 1657 1669. U.S. Fish and Wildli fe Service. 1987. 50 CFR Part 17: Endangered and Threatened Wildlife and Plants; Determination of Threatened Status for Two Florida Lizards. Federal Register 52(215): 42658 42652. Watts, W. A. and B. C. S. Hansen. 1988. Environments of Florida in the Lat e Wisconsin and Holocene. In: Wet Site Archeology (ed: Purdy, B. A.). pp. 307 323. The Telford Press, Caldwell, New Jersey. White, G. C. and K. P. Burnham. 1999. Program MARK: Survival estimation from populations of marked animals. Bird Study 46 Suppleme nt: 120 138.

PAGE 69

About the Author Nicholas Osman was born and raised in Tampa, FL. In 2005 he received a Bachelor of Science degree in Environmental Science at the University of Florida, during which time he completed a summer internship at Canaveral National Seashore monitoring sea turtle nesting. After a year of environmental consulting and post bachelor's courses, he began work toward his Master of Science degree in Biology at the University of South Florida in 2006. Nicholas was named a National Wildlife Refuge System Centennial Scholar in 2007 by the National Fish and Wildlife Foundation. He has presented his research nationally at the 2008 and 2009 Joint Meetings of Ichthyologists and Herpetologists and completed hi s degree during the summer of 2010. He hopes to pursue a career that will allow him to marry his love of nature, his passion for teaching, and his understanding of ecosystems and organisms.


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 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2010 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004611
035
(OCoLC)
040
FHM
c FHM
049
FHMM
090
XX9999 (Online)
1 100
Osman, Nicholas.
0 245
Experimental translocation of the florida sand skink (_plestiodon [=neoseps] reynoldsi_) success of a restricted species across diverse microhabitats
h [electronic resource] /
by Nicholas Osman.
260
[Tampa, Fla] :
b University of South Florida,
2010.
500
Title from PDF of title page.
Document formatted into pages; contains X pages.
502
Thesis (MS)--University of South Florida, 2010.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
3 520
ABSTRACT: The fossorial Florida Sand Skink (Plestiodon [=Neoseps] reynoldsi) inhabits a restricted range of scrub and sandhill fragments on the ridges of central Florida. The high rate of urban and agricultural development in this area necessitates conservation strategies other than land acquisition and management because of the limited remaining Florida Sand Skink habitat available. This study tests the viability of translocation as a conservation strategy for this species and assesses which features of a recipient site contribute to the successful establishment of a population. In 2007, 300 individuals were collected and moved from an intact scrub habitat, individually marked, and moved to a nearby reclaimed site with no existing Florida Sand Skink population. Fifteen 20 m2 enclosures were constructed at the recipient site, and 20 skinks were randomly assigned to each.Translocated skinks were monitored for two years to measure survival and reproduction. While survival and reproduction were apparent in all treatments, survival was significantly greater in enclosures with no shade-providing object and low soil moisture, and reproduction was most evident in enclosures with less light intensity and soil compaction. This study indicates that translocation is a practical conservation strategy for this species, and my results can be used to inform protocol for future Florida Sand Skink translocation efforts.
590
Advisor: Henry R. Mushinsky, Ph.D.
653
Florida Sand Skink
Neoseps
Fossorial
Scrub
Translocation
Lizard
Plestiodon reynoldsi
690
Dissertations, Academic
z USF
x Biology-Integrative
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.4611