USF Libraries
USF Digital Collections

Long-term effects of prescribed fire on reptile and amphibian communities in Florida sandhill

MISSING IMAGE

Material Information

Title:
Long-term effects of prescribed fire on reptile and amphibian communities in Florida sandhill
Physical Description:
Book
Language:
English
Creator:
Halstead, Neal Thomas
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Aspidoscelis sexlineata
Community ordination
Plestiodon inexpectatus
Longleaf pine
Nonmetric multidimensional scaling
Scincella lateralis
Dissertations, Academic -- Biology -- Masters -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: I examined the effects of fire frequency on reptile and amphibian community composition in a periodically burned sandhill habitat in west-central Florida. Plots burned in 2003 had lower species richness, diversity, and evenness indices than plots that had not been burned during the previous six years. Community composition was different among plots burned at different times and followed a gradient of change that corresponded to the time since the last fire. Aspidoscelis sexlineata were the most abundant lizards in recently burned plots, while Scincella lateralis and Plestiodon inexpectatus were relatively more abundant in unburned plots. Gopherus polyphemus were least abundant in unburned plots. Community composition among plots was correlated with mean leaf litter cover and herbaceous ground cover. Mean percent cover of leaf litter and herbaceous vegetation responded to the time since the last fire.^ Fire indirectly affects community composition through changes in environmental variables, such as percent coverage of leaf litter and herbaceous vegetation.Additionally, I examined annual variation in reptile and amphibian community composition over a period of four years in the mid 1980s and again in 2004. Differences existed in number of individuals captured, diversity, and evenness among years. Community composition was different between all consecutive years except 1986 and 1987. Number of individuals captured per year and annual differences in community composition were correlated with summer rainfall. The annual pattern of variation in community composition over time was no different in unburned experimental plots than in experimentally burned treatments. The power to detect such a difference is low, however, because of low replication.^ ^No apparent loss of species occurred between the 1980s and 2004, but the abundances of two non-native species were significantly higher in 2004 than in the 1980s. A third exotic species was documented at the site for the first time in 2004. Because of the increasing number of exotic reptiles and amphibians in Florida, the site is at risk of invasions of other species.
Thesis:
Thesis (M.S.)--University of South Florida, 2007.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Neal Thomas Halstead.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 48 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:
aleph - 001927956
oclc - 192074942
usfldc doi - E14-SFE0001935
usfldc handle - e14.1935
System ID:
SFS0026253:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001927956
003 fts
005 20080218145846.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 080218s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001935
035
(OCoLC)192074942
040
FHM
c FHM
049
FHMM
090
QH307.2 (ONLINE)
1 100
Halstead, Neal Thomas.
0 245
Long-term effects of prescribed fire on reptile and amphibian communities in Florida sandhill
h [electronic resource] /
by Neal Thomas Halstead.
260
[Tampa, Fla] :
b University of South Florida,
2007.
3 520
ABSTRACT: I examined the effects of fire frequency on reptile and amphibian community composition in a periodically burned sandhill habitat in west-central Florida. Plots burned in 2003 had lower species richness, diversity, and evenness indices than plots that had not been burned during the previous six years. Community composition was different among plots burned at different times and followed a gradient of change that corresponded to the time since the last fire. Aspidoscelis sexlineata were the most abundant lizards in recently burned plots, while Scincella lateralis and Plestiodon inexpectatus were relatively more abundant in unburned plots. Gopherus polyphemus were least abundant in unburned plots. Community composition among plots was correlated with mean leaf litter cover and herbaceous ground cover. Mean percent cover of leaf litter and herbaceous vegetation responded to the time since the last fire.^ Fire indirectly affects community composition through changes in environmental variables, such as percent coverage of leaf litter and herbaceous vegetation.Additionally, I examined annual variation in reptile and amphibian community composition over a period of four years in the mid 1980s and again in 2004. Differences existed in number of individuals captured, diversity, and evenness among years. Community composition was different between all consecutive years except 1986 and 1987. Number of individuals captured per year and annual differences in community composition were correlated with summer rainfall. The annual pattern of variation in community composition over time was no different in unburned experimental plots than in experimentally burned treatments. The power to detect such a difference is low, however, because of low replication.^ ^No apparent loss of species occurred between the 1980s and 2004, but the abundances of two non-native species were significantly higher in 2004 than in the 1980s. A third exotic species was documented at the site for the first time in 2004. Because of the increasing number of exotic reptiles and amphibians in Florida, the site is at risk of invasions of other species.
502
Thesis (M.S.)--University of South Florida, 2007.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 48 pages.
590
Adviser: Earl McCoy, Ph.D.
653
Aspidoscelis sexlineata.
Community ordination.
Plestiodon inexpectatus.
Longleaf pine.
Nonmetric multidimensional scaling.
Scincella lateralis.
690
Dissertations, Academic
z USF
x Biology
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1935



PAGE 1

Long-Term Effects of Prescribed Fire on Re ptile and Amphibian Communities in Florida Sandhill by Neal Thomas Halstead 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: Earl McCoy, Ph.D. Co-Major Professor: Henry Mushinsky, Ph.D. Peter Stiling, Ph.D. Date of Approval: April 10, 2007 Keywords: Aspidoscelis sex lineata, community ordinati on, Plestiodon inexpectatus, longleaf pine, nonmetric multidimensi onal scaling, Scincella lateralis Copyright 2007, Neal Halstead

PAGE 2

Acknowledgments I would like to thank my co -advisors, Henry Mushinsky and Earl McCoy, for their help and guidance during this process. I w ould also like to thank my other committee member, Peter Stiling, for providi ng valuable comments and advice. I thank the Division of Integrative Bi ology for providing and maintaining the experimentally burned plots in the University of South Floridas Ecological Research Area, a valuable resource for faculty as well as graduate and undergraduate students. Gordon Fox and Anne Schmidt deserve special cr edit for trying to ensure that plots were burned on schedule, even though it was not always possible. For help with field work, I would like to thank Celina Bellanceau, Br ian Halstead, Travis Robbins, Kristan Raymond, Joel Johnson, Katie Basiotis, Teresa Piacenza, Charlie Piacenza, Lance Arvidson, Emily Muehlstein, Justin Boorstein, Mela Starr, Tulin Ozyurt, Alecxis Farkas, and Jim Halstead, Jr. Finally, I would like to thank the Biology Graduate Student Organization and the Graduate Assistants United for representing gr aduate student needs at the department and university levels, respectively.

PAGE 3

i Table of Contents List of Tables ....................................................................................................................iii List of Figures......................................................................................................................v Abstract....................................................................................................................... .......vi Fire Affects Herpetofaunal Community Com position Through Variation in Microhabitat Variables that Respond to the Time Since the Last Fire......................................................1 Introduction..............................................................................................................1 Methods....................................................................................................................3 Study Site.....................................................................................................3 Experimental Design....................................................................................4 Data Analysis...............................................................................................5 Results......................................................................................................................9 Discussion..............................................................................................................18

PAGE 4

ii Annual Variation in Herpetofaunal Community Composition in a Periodically Burned Florida Sandhill............................................................................................................... ...25 Introduction............................................................................................................25 Methods..................................................................................................................26 Experimental Design..................................................................................26 Data Analysis.............................................................................................28 Results....................................................................................................................31 Discussion..............................................................................................................37 Literature Cited............................................................................................................... ...44

PAGE 5

iii List of Tables Table 1 Number of individuals capt ured (N), species richness (S), and Hills diversity (N 1 N 2 ) and evenness (F 2,1 ) numbers with upper and lower 95% confidence interval limits for each fire frequency. Confidence intervals were not calculated for species richness because no difference was detected by a randomization test................................................................10 Table 2 Pairwise R values between pl ots with different time since the last fire from a two-way nested ANOSIM (r eplicate plots nested in TSF treatments); and from a one-way ANOSIM treating trap arrays as replicate samples...............................................................................................13 Table 3 Pairwise R values betwee n plots with differe nt fire frequency treatments from a two-way nested ANOSIM (replicate plots nested in fire frequency treatments); and fr om a one-way ANOSIM treating trap arrays as replicate samples................................................................................14 Table 4 Mean number of individu als captured per plot for eight species contributing to differences between the three time since the last fire treatments..........................................................................................................15

PAGE 6

iv Table 5 Pairwise R values betwee n plots with differe nt fire frequency treatments from a two-way nested ANOSIM (replicate plots nested in fire frequency treatments) on Euclidean distances calculated from normalized mean percent leaf litter cover; and from a one-way ANOSIM treating trap arrays as replicate samples...........................................17 Table 6 Spearman rank correlations of number of indivi duals captured with environmental variables at each trap array for eight species associated with changes in community composition. Asterisks indicate correlations that are significant at p = 0.05.......................................................18 Table 7 Number of individuals capt ured (N), species richness (S), and Hills diversity (N 1 N 2 ) and evenness (F 2,1 ) numbers with upper and lower 95% confidence interval limits for each year. Confidence intervals were not calculated for species richness because no difference was detected by a randomization test.....................................................................................32 Table 8 Pairwise R values betw een years from a two-way crossed ANOSIM..............35 Table 9 Pairwise R values between fire frequency treatments from a two-way crossed ANOSIM. There were not enough replicates of the 2 and 9 year fire frequency treatments to make a pairwise comparison................................35

PAGE 7

v List of Figures Figure 1. Map of USF ERA. Outlines indicate boundaries of experimental plots.............4 Figure 2. Cluster analysis dendrogram of experimental plots using the complete linkage method. Labels along the x-ax is refer to plot designations. Solid black lines indicate signifi cantly different clusters with p 0.05...........11 Figure 3. Nonmetric multidimensional scaling plot of experimentally burned plots. Labels are plot designations Symbols refer to the time since the last fire (triangles = 1 year sinc e fire; circles = 7 years since fire; squares = unburned reference plots) Stress measures the goodness-offit of the data to the pattern; lo wer stress indicates a better fit.........................12 Figure 4. Map of experimentally burned plots in the USF ERA during the 1980s. Numbers refer to fire re turn interval (years).....................................................27 Figure 5. Mean number of individuals captu red per plot (y-axis) versus summer rainfall (x-axis) for each year............................................................................32 Figure 6. Nonmetric multidimensional scaling plot of experimentally burned plots among years. Stress measures the go odness-of-fit of the data to the pattern; lower stress indi cates a better fit..........................................................34

PAGE 8

Long-Term Effects of Prescribed Fire on Re ptile and Amphibian Communities in Florida Sandhill Neal Thomas Halstead ABSTRACT I examined the effects of fire frequency on reptile and amphibian community composition in a periodically burned sandhill habitat in west-central Florida. Plots burned in 2003 had lower species richness, dive rsity, and evenness indices than plots that had not been burned during the previous six years. Community composition was different among plots burned at different times and followed a gradient of change that corresponded to the time since the last fire. Aspidoscelis sexlineata were the most abundant lizards in recently burned plots, while Scincella lateralis and Plestiodon inexpectatus were relatively more abundant in unburned plots. Gopherus polyphemus were least abundant in unburned plots. Community composition among plots was correlated with mean leaf litter cover and herbaceous ground cover. Mean percent cover of leaf litter and herbaceous vegetation responded to the tim e since the last fire. Fire indirectly affects community composition thr ough changes in envir onmental variables, such as percent coverage of leaf litter and herbaceous vegetation. Additionally, I examined a nnual variation in reptile and amphibian community composition over a period of four years in the mid 1980s and again in 2004. Differences existed in number of individuals captu red, diversity, and evenness among years. Community composition was different between all consecutive years except 1986 and vi

PAGE 9

1987. Number of individuals captured per ye ar and annual differences in community composition were correlated with summer rain fall. The annual patt ern of variation in community composition over time was no differe nt in unburned experimental plots than in experimentally burned treatments. The power to detect such a difference is low, however, because of low replication. No appa rent loss of species occurred between the 1980s and 2004, but the abundances of two non-na tive species were significantly higher in 2004 than in the 1980s. A third exotic species was documente d at the site for the first time in 2004. Because of the increasing number of exotic reptiles and amphibians in Florida, the site is at risk of invasions of other species. vii

PAGE 10

Fire Affects Herpetofaunal Community Com position Through Variation in Microhabitat Variables that Respond to the Time Since the Last Fire Introduction Florida high pine habitats, such as sandhills and clayhills, are characterized by having a natural fire return interval ranging from 1-10 years, which maintains 1) a sparse overstory of longleaf pine (Pinus palustris ) or slash pine (P. elliottii ), 2) a dense understory of herbaceous vegeta tion dominated by wiregrass ( Aristida stricta ), and 3) a low density midstory of hard woods, usually turkey oak ( Quercus laevis ) (Myers, 1990). Less than 3% of the original high pine habitats in Florida remain, and the vast majority of the existing high pine is hi ghly fragmented and degraded (Kautz, 1993; Myers, 1990). Approximately 75% of fragments are less than 40 ha in area and 30% are less than 8 ha. Many of these habitat fragments exist w ithin a matrix of urban and suburban developments (Heuberger and Putz, 2003). Ecological processes often are altered in fragmented ha bitats, including increased likelihood of invasion by non-native species, increased susceptibility to extinction for populations occurring within the fragment, and altered gr oundwater levels and nutrient input (Heuberger and Putz, 2003). Habitat fr agmentation contributes to decreases in species richness and diversity across several taxa (Dickma n, 1987; Kjoss and Litvaitis, 2001). Fire suppression in Florida sandhill ha bitats causes an increase in hardwood density, canopy cover and leaf litter depth and a concomitant decrease in herbaceous 1

PAGE 11

understory vegetation (Brockway and Lewi s, 1997; Kautz, 1993; Mushinsky, 1985, 1992; Myers, 1990). The effects of prescribed fire on species ri chness and diversity have been studied in several ecological systems. Increased plan t species richness and evenness occurs in longleaf pine communities subjected to mode rate to high fire frequencies (Mehlman, 1992). Reintroduction of fire to longleaf pi ne sandhills at Eglin Air Force Base in northwest Florida increased species richness of plants, arthropods, herpetofauna, and birds (Provencher et al., 2003). Effects of fire on reptile and amphibian populations vary. For example, prescribed fire had significant negative short-term (1-2 years) effects on amphibian populations in South Carolina pine flatwoods (Schurbon and Fauth, 2003). Potential negative and positive effects of prescribed burning exist for amphibian populations, but more rigorous research is needed (Pilliod et al., 2003; Russell et al., 1999). Diversity of reptiles and amphibians on burned plots in a west-central Florida longleaf pine sandhill, however, was hi gher than on unburned pl ots (Mushinsky, 1985). To enhance our understanding of the eff ects of fire on reptile and amphibian communities, I examined its effects on herpetofauna at the University of South Floridas (USF) Ecological Research Area (ERA) to an swer the following questions. 1) What effects do different prescribed fire freque ncies have on species richness and diversity indices? 2) What is the e ffect of different prescribed fire treatments on community composition? 3) Does community composition, by retaining species identities throughout analysis, show a different res ponse to fire frequency than univariate richness, diversity, and evenness indices? I found that species richness, diversity, and evenness were all 2

PAGE 12

higher in plots that were not burned recentl y, and that community composition of reptile and amphibian species was different am ong plots burned at different times. Materials and Methods Study Site The USF ERA is approximately 245 ha of riverine swamp, pine flatwoods, and sandhill located in Hillsborough C ounty, west-central Florida (28 N, 82 W). My study was conducted in approximately 13 ha of upland sandhill habitat with an overstory of P. palustris a midstory of Q. laevis, and an understory dominated by A. stricta numerous forb species, and scatte red clumps of saw palmetto ( Serenoa repens ) (Mushinsky, 1985, 1992). Experimental burn plot s (0.7 1.9 ha) with various prescribed fire frequencies were established in the 1970 s in 12.4 ha of sandhill along the southern border of the ERA. Replication of four prescr ibed fire treatments (one-, two-, fiveand seven-year fire return intervals) began in the mid-1980s, with two re plicate plots in each treatment (Fig 1). All one-, two-, and five-y ear plots were most recently burned in November 2003. The seven-year plots were burned in 1998 (Schmidt, 2005). The two reference plots (CE and CW) have not b een burned since 1965 or earlier (Mushinsky, 1985). 3

PAGE 13

Figure 1. Map of USF ERA. Outlines indicate boundaries of experimental plots. Experimental Design Five drift fence/pitf all trap arrays were placed in each plot approximately 20-30 m apart with a random orientation (Mushinsky 198 5, 1992). Each trap array consisted of a 6 m long by 0.5 m high drift fen ce (aluminum flashing), buried on edge 10-15 cm in the ground. At both ends of each drift fence, a 19 L plastic bucket was placed in the ground with the top edge at ground level. Holes we re drilled in the bottom of the buckets to allow for drainage of rainwater, and pieces of Styrofoam or cellulose sponge (approximately 10 x 10 x 2 cm) were placed in each bucket as a shade cover/flotation device. A cover object of 2 m long by 0.5 m wide galvanized roofing material was placed on the ground within 5 m of each trap array (Mushinsky 1985, 1992). In addition to pitfall traps, two funnel traps were placed along both sides of each drift fence near the center. Funnel traps were c onstructed from 0.6 cm mesh hardware cloth and consisted of a rectangul ar trap body (approximately 20 x 20 x 50 cm) with two funnels of the same material (approximately 30 cm diameter outer opening to 4 cm diameter inner opening) inserted into both ends of the trap body. Funnel traps were placed flush against the drift fence and buried approximately 2 cm into the soil to hold 4

PAGE 14

them in place and provide a smooth transition into the funnel entrance. Pieces of wet cellulose sponge were placed inside the tr ap body to help prevent desiccation of organisms. Shade covers constructed from vinyl siding (approximately 60 x 30 cm) were placed over the top of each funnel trap body. Microhabitat variables [percent cover of: bare sand, leaf litte r, herbaceous vegetation, woody vegetation, and woody deadfall; woody stem de nsity; leaf litter de pth; leaf litter type (oak/pine); presence/absence of canopy cover; and canopy type (oak/pine)] were measured at the end of the sampling period in eight 1 m 2 quadrats, with one quadrat immediately adjacent to each pi tfall and funnel trap and one quadrat 1 m from each pitfall and funnel trap at each trap array. Mean and standard deviation of each microhabitat variable was calculated over the eight quadrats for each trap array. Trap arrays were checked several days per week from November 2003 through June 2005. All captured individuals were identified to species and marked individually by toeclipping (Waichman 1992). Snakes were scale-clipped or marked with passiveintegrated transponder (PIT) tags (AVID Technology, Tewksbury, MA). Data Analysis Because the purpose of this study was to ex amine patterns of change in community composition associated with prescribed fire, species abundances at each trap array were correlated with distance to the nearest wetland edge. Species that were significantly correlated with distance to wetland were remove d from analysis. This exercise eliminated any potential effect on community composition of location of the experimental plots relative to the wetland. 5

PAGE 15

Species richness was estimated using the obser ved count of number of species (S). Hills diversity numbers N 1 and N 2 (Hill, 1973) were used to measure species diversity as recommended by Ludwig and Reynolds (1988). Simpsons index (1; Simpson, 1949) and the Shannon-Weaver index (H; Shannon a nd Weaver, 1949) also were calculated for diversity because of their common use in the literature and their relation to N 2 and N 1 respectively. Evenness of relative abundances was measured with the modified Hills ratio (F 2,1 ; Alatalo, 1981). Overall differences in index values among plots were tested using the randomization tec hnique described by Solow ( 1993). When the randomization test indicated a difference in index valu es, pairwise comparisons were made by bootstrapping 95% confidence intervals for each fire frequency. Nonparametric analyses of community com position, retaining all species identities, were performed using PRIMER (version 6; PRIMER-E Ltd, 2006). Similarity between sampling units (trap arrays, plots, and treatments) was calculated using the Bray-Curtis index (Bray and Curtis, 1957) on square-root transformed c ounts of species abundances. Square-root transformation was used to down-we ight the effect of very abundant species (Clarke and Gorley, 2006). Cluster analysis was performed using the unweighted complete linkage method (Ludwig and Reynolds, 1988) with a post-hoc randomization procedure to identify groups of plots with different community compositions, assuming no a priori structure among experimental plots. The post-hoc ra ndomization test was accomplished through the SIMPROF routine in PRIMER. The SI MPROF procedure tests for significant differences in community composition at each node in the cluster dendrogram (Clarke and Gorley, 2006). Complete linkage was used as the clustering technique because the 6

PAGE 16

goal of the procedure was to identify groups that were the most different. Because species-abundance patterns were nonlinear, nonmetric multidimensional scaling (NMS) was used for ordination, as recomme nded by Ludwig and Reynolds (1988). Differences in community composition among plots, burn frequencies, and time since the last fire treatments were tested for by analysis of similarities (ANOSIM). ANOSIM is a nonparametric analogue to ANOVA in that it tests against the null hypothesis that there is no difference in community compositi on among treatments. The test statistic used in ANOSIM is R, which is a measure of the observed differences among rank similarities between treatments relative to the observed differences among rank similarities within treatments. Under the null hypothesis of no trea tment effect, R would approach 0. If all rank similarities within treatments are great er than all rank similarities between treatments R is equal to 1. The null dist ribution of R is calcu lated in PRIMER by repeatedly randomly shuffling all sample labe ls and recalculating R for each permutation. A one-tailed test of significance is de termined by calculating the proportion of randomized R statistics that are greater than the observed value of R. The ANOSIM test calculates the observed R statistic among all treatments (Gl obal R), as well as R statistics between each pairwise combination of treatme nts. Like ANOVA, if the Global R statistic is not significant, then pairwise comparisons are ignored (Clarke and Warwick, 2001). The existence of a directional change in community composition along a gradient of fire frequency treatments was determined by constructing a model matrix consisting of the distance in years between prescribed burns between all pairs of plots. The rank dissimilarities from the observed Bray-Curtis si milarity matrix were correlated with the model matrix, using the RELATE function in PRIMER. Significance was determined 7

PAGE 17

using a Monte Carlo randomization technique that repeatedly calculated the Spearman rank correlation of the observed Bray-Curtis similarity matrix with the model matrix for 999 permutations of randomly reassigned plot labels in the observed Bray-Curtis matrix (Clarke and Gorley, 2006). A similar model matr ix was constructed to test for change in community composition following the patt ern of years since the last fire. Species that contributed to differences in community composition between different groups were identified using the similarity percentages (SIMPER) routine in PRIMER (Clarke and Gorley, 2006). Species that contri buted most to the overall pattern of BrayCurtis similarity were identified using a modified BIOENV procedure in PRIMER that calculated the Spearman rank correlation values between Bray-Curtis similarity values from progressively larger subsets of species and the Bray-Curtis similarity matrix from the full species set. Randomization tests th at randomly shuffled plot labels and recalculated the Spearman rank correlation were used to determine significance of r S for each number of species in the s ubset (Clarke and Warwick, 2001). Microhabitat variables were normalized, a nd the BIOENV procedure in PRIMER was used to calculate the Spearman rank correlati on between the Euclidean distance matrix (using various combinations of normalized va riables) and the Bray-Curtis similarity matrix of transformed species abundances betw een all pairs of trap arrays. To account for potential covariation among mi crohabitat variables, principle components analysis was performed on the normalized microhabitat variables, and the BIOENV procedure was performed using the weights of trap arrays on each PC axis to produce Euclidean distance matrices. Mean values for each microhabitat va riable were also calculated for each plot and fire frequency treatment and the above proc edures were repeated for those scales of 8

PAGE 18

analysis. The randomization test described above was used to determine significance for each combination of environmental variab les (Clarke and Gorley, 2006). ANOSIM was used to test for differences in environmental variables among fire frequency treatments, and RELATE was used to test for directional change in microhabitat variables among fire frequency treatments. Results A total of 1415 individuals from 23 species were used for analysis (Table 1). Variation in species richne ss was not different among fire frequency treatments (s 2 = 4.7; p = 0.380). Diversity indices were higher in seven-year trea tments and reference plots than plots burned in November 2003 (p < 0.05; Table 1). Within the group of recently burned plots, diversity was higher in two-year treatments than in five-year treatments, although the difference was not as great as the difference between recently burned and unburned plots. Evenness was higher in plots no t burned recently than in plots burned in 2003 (p < 0.05; Table 1). Evenness was not di fferent among plots of different fire frequencies burned in 2003. 9

PAGE 19

Table 1. Number of individuals captured (N), species richness (S), and Hills diversity (N 1 N 2 ) and evenness (F 2,1 ) numbers with upper and lower 95% confidence interval limits for each fire frequency. C onfidence intervals were not calculated for species richness because no difference wa s detected by a randomization test. N 1 N 2 F 2,1 95% Confidence Interval 95% Confidence Interval 95% Confidence Interval Fire Frequency (years) N S Observed value Lower Upper Observed value Lower Upper Observed value Lower Upper 1 356 14 3.73 3.16 4.20 2. 19 1.93 2.50 0.44 0.41 0.49 2 327 14 4.73 4.00 5.32 2. 76 2.37 3.20 0.47 0.44 0.53 5 257 14 3.34 2.74 3.85 1. 96 1.71 2.26 0.41 0.39 0.47 7 270 19 7.77 6.58 8.48 5. 58 4.78 6.30 0.68 0.64 0.75 no fire 205 15 6.79 5.70 7.46 5.12 4.38 5.75 0.71 0.68 0.78 Community composition was different among fire frequency treatments and among plots with different times sin ce the last fire (TSF). Cluste r analysis in conjunction with the SIMPROF test revealed two groups of experimental plots with different community composition, those that were burned in N ovember 2003 and plots that had not been recently burned, that differed in community composition (p = 0.05; Fig 2). A similar pattern was revealed by NMS (s tress = 0.05; Fig 3), although th e 7W plot appears to be more similar to the more recently burned plots. 10

PAGE 20

5E 1W 5W 2W 1E 2E 7W CE 7E CWPlots 100 90 80 70 60 50Bray-Curtis Similarity Figure 2. Cluster analysis dendrogram of experimental plots using the complete linkage method. Labels along the x-axis refe r to plot designations Solid black lines indicate significantly di fferent clusters with p 0.05. 11

PAGE 21

1E 1W 2E 2W 5E 5W 7E 7W CE CW2D Stress: 0.05 Figure 3. Nonmetric multidimensional scaling plot of experimentally burned plots. Labels are plot designations. Sy mbols refer to the time since the last fire (triangles = 1 year since fire; circles = 7 years since fire; squares = unbur ned reference plots). Stress measures the goodness-of-fit of the data to the pa ttern; lower stress indi cates a better fit. The separation between TSF treatments wa s significant (Globa l R = 0.807, p = 0.002) when accounting for variation among plots within TSF treatments, using a two-way nested ANOSIM with plots nested in time sin ce fire treatments. Pairwise tests showed differences between plots bur ned in November 2003 and re ference plots. Community composition in plots in the seven-year treatme nt was not different from plots burned in 2003 or from unburned reference plots (Table 2). Significance tests from randomization had low power to detect real differences becau se of the low number of replicates in each 12

PAGE 22

fire frequency treatment. Because the twoway nested ANOSIM showed no difference in similarity among trap arrays in replicate pl ots within each TSF tr eatment (Global R = 0.007, p = 052), a one-way ANOSIM could be perfor med using trap arrays as replicates without risk of pseudoreplication (Clarke and Gorley, 2006). This reduced the overall separation among TSF treatments (Global R = 0.398, p = 0.001), but revealed significant pairwise differences in community compositi on between all TSF trea tments (Table 2). Table 2. Pairwise R values between plots with different time since the last fire from a two-way nested ANOSIM (replicate plots nested in TSF treatments); and from a one-way ANOSIM treating trap arrays as replicate samples. 2-way nested ANOSIM, plots within time since the last fire treatments 1-way ANOSIM, trap arrays as replicates Comparison R p R p 1 vs. 7 0.656 0.071 0.335 0.001 1 vs. no fire 1 0.036 0.523 0.001 7 vs. no fire 1 0.333 0.185 0.020 Differences in community composition among fire frequency treatments were also significant, but less distinct than TSF treatments (two-w ay nested ANOSIM; Global R = 0.560; p = 0.011). As before, low replication resulted in pairwise comparisons without sufficient power to detect differences betw een fire frequency treatments (Table 3). Because community composition among trap ar rays in plots within fire frequency treatments was not different (Global R = -0.018, p = 0.587), a one-way ANOSIM was performed using trap arrays as samples. Differences in community composition among fire frequency treatments were significant in the one-way analysis (Global R = 0.271, p = 13

PAGE 23

0.001). Community composition was different among trap arrays in fire frequency treatments burned at different times (Table 3). Fire frequency treatments burned in November 2003 did not have different comm unity composition (Table 3). Changes in community composition among plots were correlated with the pattern of differences in years between fire (r S = 0.559, p = 0.001) as well as the pattern of differences in TSF ( r S = 0.69, p = 0.003). Table 3. Pairwise R values between plots with different fire frequency treatments from a two-way nested ANOSIM (replicate plot s nested in fire frequency treatments); and from a one-way ANOSIM treating trap arrays as replicate samples. 2-way nested ANOSIM, plots within fire frequency treatments 1-way ANOSIM, trap arrays as replicates Comparison R p R p 1 vs. 2 -0.25 1.000 -0.076 0.911 1 vs. 5 0.25 0.667 0.034 0.259 1 vs. 7 0.75 0.333 0.361 0.001 1 vs. no fire 1 0.333 0.552 0.001 2 vs. 5 0 1.000 0.053 0.212 2 vs. 7 0.5 0.333 0.277 0.005 2 vs. no fire 1 0.333 0.506 0.001 5 vs. 7 0.5 0.333 0.337 0.002 5 vs. no fire 1 0.333 0.485 0.001 7 vs. no fire 1 0.333 0.185 0.030 The relative abundances of three lizar d species, the six-lined racerunner, Aspidoscelis sexlineata the southeastern five-lined skink, Plestiodon inexpectatus and the ground skink, Scincella lateralis and one turtle, the gopher tortoise, Gopherus polyphemus, accounted for the majority of the dissimila rity in community composition between TSF treatments. Plots burned in 2003 had relatively higher abundances of A. sexlineata and 14

PAGE 24

lower abundances of P. inexpectatus and S. lateralis than plots that had not burned recently. Long-unburned reference plots had lower abundances of G. polyphemus than all other plots. Several other species were a bundant on all plots but c ontributed less to the dissimilarity in community structure between TSF treatments, including the black racer, Coluber constrictor the southern leopard frog, Lithobates sphenocephalus the green anole, Anolis carolinensis and the ringneck snake, Diadophis punctatus (Table 4). Table 4. Mean number of individuals captured per plot for eight species contributing to differences between the three time since the last fire treatments. Years Since The Last Fire Species 1 7 40 Anolis carolinensis 3.3 1.5 2.5 Aspidoscelis sexlineata 100.8 42 27 Coluber constrictor 12.8 15.5 10 Diadophis punctatus 2.2 3.5 3.5 Gopherus polyphemus 4.3 4 1 Lithobates sphenocephala 13.3 16.5 5.5 Plestiodon inexpectatus 3.7 13.5 29 Scincella lateralis 11 27.5 18 Bray-Curtis similarity matrices ca lculated from the abundances of Aspidoscelis sexlineata ( r S = 0.76, p = 0.004) alone or Plestiodon inexpectatus ( r S = 0.719, p = 0.01) alone were both correlated with the Bray-Curtis similarity matrix from the full species set. Subsets of two or more species also produced significan t correlations with the BrayCurtis similarity matrix from the full speci es set, although differe nt combinations of species gave similar results. In general, the best combinations incl uded, in addition to at least one of the first two species, Scincella lateralis Lithobates sphenocephalus Diadophis punctatus, Rhadinaea flavilata and Gopherus polyphemus 15

PAGE 25

Mean percent leaf litter cover at each trap array produced a Euclidean distance matrix that had an r S = 0.404 (p = 0.01) with the Bray-Curtis similarity matrix of transformed abundances. Inclusion of additional microhab itat variables decreased the Spearman rank correlation. Performing a principle components analysis to account for covariation among the normalized microhabitat variables did not increase the rank correlation when principle component axes were used to cal culate Euclidean distances in the BIOENV procedure. Bray-Curtis similarities of community composition among plots were correlated with a Euclidean distance matrix produced from mean percent herbaceous ground cover in each plot ( r S = 0.795, p = 0.01). No combination of principle component axes produced Euclidean distance matrices with a higher Sp earman rank correlation than that produced from herbaceous ground cover alone. Leaf litter cover was not different among fi re frequency treatments (Global R = 0.21, p = 0.138) when plots were nested in fire frequency treatments in a 2-way nested ANOSIM. However, leaf litter cover was diffe rent among plots burned at different times (Global R = 0.584, p = 0.032) in a 2-way nest ed ANOSIM. Pairwise comparisons had low power, but indicated that plots burned in 2003 were different from reference plots (Table 4). Leaf litter cover among plots with in TSF treatments was not different (Global R = 0.073, p = 0.136). A one-way ANOSIM using trap arrays as replicat es indicated that leaf litter cover was different among plots with different times since the last fire (Global R = 0.324, p = 0.001), and that leaf litter cove r was different between recently burned plots and plots not burned in 2003 (Table 5). The same pattern of differences among plots burned at different times but not among fire frequency treatments was true for one-way 16

PAGE 26

ANOSIM of herbaceous ground cover among plots, as well as for the Euclidean distance matrices for any combination of environmenta l variables that was significantly correlated with the Bray-Curtis similarities among plots. Table 5. Pairwise R values between plots with different fire frequency treatments from a two-way nested ANOSIM (replicate plot s nested in fire frequency treatments) on Euclidean distances calculated from normalized mean percent leaf litter cover; and from a one-way ANOSIM treating trap arrays as replicate samples. 2-way nested ANOSIM, plots within time since the last fire treatments 1-way ANOSIM, trap arrays as replicates Comparison R p R p 1 vs. 7 0.271 0.250 0.195 0.020 1 vs. no fire 0.938 0.036 0.500 0.001 7 vs. no fire -0.5 1.000 0.035 0.240 Aspidoscelis sexlineata was positively correlated with bare sand and herbaceous vegetation, and negatively corre lated with leaf litter, woody deadfall, and canopy cover, and leaf litter depth (Table 6). Plestiodon inexpectatus and Scincella lateralis were negatively correlated with bare sand and herb aceous vegetation, and positively correlated with leaf litter cover and depth, woody deadfall cover, and canopy cover (Table 6). Gopherus polyphemus was positively correlated with herbaceous vegetation and negatively correlated with leaf litter cover (Table 6). Most environmental variables were significantly correlated with other environmental variable s, with the exception of woody vegetation cover, leaf litter composition, and canopy composition. 17

PAGE 27

Table 6. Spearman rank correlations of number of individuals captured with environmental variables at each trap array fo r eight species associated with changes in community composition. Asterisks indicate correlations that are sign ificant at p = 0.05. Species Bare sand cover Leaf litter cover Herbaceous ground cover Woody deadfall cover Canopy cover r S 0.057 0.065 -0. 143 0.115 0.159 Anolis carolinensis p 0.693 0.652 0. 322 0.426 0.271 r S .640(*) -.743(*) .509( *) -.312(*) -.424(*) Aspidoscelis sexlineata p 0.000 0.000 0. 000 0.027 0.002 r S 0.025 -0.021 -0.100 -0.122 -0.063 Coluber constrictor p 0.862 0.887 0. 489 0.398 0.663 r S 0.110 -0.062 -0.163 -0.069 -0.153 Diadophis punctatus p 0.447 0.670 0. 257 0.635 0.289 r S -.533(*) .579(*) -.426( *) .383(*) .346(*) Plestiodon inexpectatus p 0.000 0.000 0. 002 0.006 0.014 r S 0.138 -.362(*) .422( *) -0.053 -0.182 Gopherus polyphemus p 0.340 0.010 0. 002 0.712 0.206 r S 0.071 -0.186 0.128 -0.074 -0.057 Lithobates sphenocephala p 0.623 0.195 0. 377 0.609 0.694 r S -.306(*) .476(*) -.365( *) .303(*) .327(*) Scincella lateralis p 0.031 0.000 0. 009 0.032 0.020 Discussion Herpetofaunal community composition in Florida sandhill habitat res ponds to fire as an indirect effect of the respons e of microhabitat variables to th e time since the last fire as well as the frequency with which plots are bu rned. It is difficult, however, to try to determine the exact mechanisms by which pl ant species are responding to fire because fire frequency and the time since the last fire are not independent variables. The time since the last fire can capt ure much of the variation associated with fire frequency because a plot that is burned annually will ne ver have more than one year without fire, assuming prescribed burns occur on schedule, while plots with longer burn intervals can 18

PAGE 28

have several years without fire. The frequency with which a plot is burned as well as the time of year during which the burn occurs, also can greatly influence the physical structure of sandhill habitat by permitting the pe rsistence of greater or lower densities of hardwoods, herbaceous plants, leaf litter and woody debris (Drewa et al, 2002, 2006; Glitzenstein et al, 1995, 2003). Obviously, both the time since the last fire and the frequency of fire can have significant effects on the physical stru cture of the habitat. It is unlikely that a plot burned every year, w ith a greater proportion of bare sand and herbaceous ground cover, and a plot burned ever y seven years, with greater leaf litter cover, canopy cover, and hardwood density, w ould have the same physical structure one year after a fire. The oneand two-year fire frequency treatments were not burned between June 2000 and November 2003, which also confounds interp retation of the results presented here. A period of three years without fire has the potential to significantly alter the plant composition of the understory in longleaf pine systems. Understory vegetation was shrubdominated in treatments with fire return inte rvals of three to four years as opposed to an herbaceous-dominated understory with interval s of one to two years in South Carolina flatwoods (Glitzenstein et al, 2003). Additi onally, a switch from growing season fires in June to dormant season fires in November could also have affected the shrub layer in each burned plot. Dormant season fires have b een associated with increases in hardwood densities in longleaf pine-dominated comm unities (Drewa et al, 2002, 2006; Glitzenstein et al, 1995). The percent cover of woody ve getation near trap arrays was not different among fire frequency treatments 18 months after th e November 2003 burn. The number of woody 19

PAGE 29

stems resprouting from underground organs can reach pre-fire conditions within 12 months of growing season fires (Olson and Platt, 1995), supporti ng the hypothesis that the delay in prescribed burning may have al tered the understory pl ant community in the oneand two-year fire frequency treatments Historical data on densities of woody vegetation at the site are not available to test this hypothe sis; the area of bare sand, however, decreased with increa sing fire return intervals in 2005, matching patterns that existed in the 1980s (Mushinsky and Gi bson, 1991). Furthermore, woody vegetation represented less than 8 percent of ground level vegetation in all fire frequency treatments in 2005, as opposed to bare sand and herbaceo us vegetation which accounted for up to 25 and 50 percent cover, respectively. Therefore, whether the long history of frequent burning has maintained the structural co mposition of the understory among fire frequency treatments, or the three year dela y in burning has allowed woody vegetation to significantly increase in the one and tw o year treatments, remains uncertain. Community composition was different am ong fire frequency treatments; this difference, however, was more distinct among plots burned at different times. Analysis of similarities revealed that treatments define d by the time since the la st fire had greater separation than fire frequency treatments, a lthough it is not possible to determine if the difference in the ANOSIM global R statistics from these two tests is significant. The same is true of the rank correlations of ch anges in Bray-Curtis dissimilarity with the patterns expected from direc tional change in composition acr oss fire frequency or time since the last fire treatments. Furthermor e, there was no difference in composition among plots of different fire fre quencies that were burned at the same time. This finding provides further evidence that reptiles and amphibians may be responding more to 20

PAGE 30

changes in the environment associated with th e time since the last fire rather than the frequency of the fire itself. Because the mo st recently burned plots included the three most frequently burned treatments, however, it was not possible to distinguish differences in community composition as a function of fire frequency or time since the last fire separately. Further monitoring of both the herpetofaunal communities and the structural composition of the plots associated with plant communities needs to continue to capture variation associated with different experimentally burned treatments at various stages of post-burn succession. It is also important that prescribed bu rns occur on schedule and that fire frequency treatments are maintained. Microhabitat variables are directly affected by fire frequency (Brockway and Lewis, 1997; Kautz, 1993; Mushinsky, 1985, 1992; Myer s, 1990), although the data presented in this paper and others suggest that at leas t some microhabitat vari ables are responding to the time since the last fire (Glitzenstein et al, 2003; Guerin, 1 993; Lemon, 1949; Olson and Platt, 1995). Percent cove r of leaf litter alone accounted for 40% of the variation observed in community composition at individual trap arrays. At the plot level, cover of herbaceous vegetation explained nearly 80% of the variation in community composition. At this study site, composition of envir onmental variables was different among plots burned recently and plots not burned recently, but not among fire frequency treatments. Lizard abundances accounted for the larges t portion of the variation in community structure among treatments. Species have varied responses to these changes in microhabitat, but lizards respond most str ongly to changes in ground cover (Mushinsky and Gibson, 1991; Table 6). Vegetation variables have been identified as important in structuring lizard communities in other system s (Dickman, 1987; Jellinek et al, 2004). 21

PAGE 31

Six-lined racerunner, Aspidoscelis sexlineata abundance responded positively to recently burned treatments that had little l eaf litter cover and woody deadfall and more bare sand and herbaceous vegetation (Table 6) Orthopterans are a common prey item of adult racerunners, while juve niles tend to feed on small, cryptic, plant-dwelling arthropods (Paulissen, 1987). Arthropod abundance data were not collected during this study, but it is likely that the increased herbaceous cover in the most recently burned plots is also associated with an increase in the preferred prey items of this lizard. Insect (especially orthopteran) abundan ce and levels of herbivory in crease near the edges (10 m from the edge) relative to the interior (500 m from the edge) of r ecently burned plots as vegetation recovers (Knight a nd Holt, 2005). Plots in this st udy were comparatively small (~1 ha), so it is likely that edge effects on insect herbivory occurred throughout the entire area of each plot. An increased area of bare sand in these treatments is also likely to improve the ability of racerunners to both see and pursue prey (Mushinsky, 1992). Inferences about relative abundances between treatments must be made cautiously, however, because increases in the area of ba re sand may also incr ease the ability of lizards to encounter trap arrays Analysis of mark-recapture da ta at the site would identify any differences in capture probabilities associ ated with changes in microhabitat variables. Southeastern five-lined skinks, Plestiodon inexpectatus and ground skinks, Scincella lateralis were correlated with the same variables as A. sexlineata but in the opposite direction (Table 6), and were more abundant in plots not burned recently. Both of these species are diurnal insectivorous lizards that rely strongly on chemosensory cues for prey recognition (Cooper et al, 2000; Cooper and Hartdegen, 1999) and actively forage for prey, often under litter (Mus hinsky, 1992; Vitt and Cooper, 1986). 22

PAGE 32

Gopher tortoises, Gopherus polyphemus respond positively to the presence of herbaceous vegetation (Table 6) and were abundant in plots with less canopy cover and higher herbaceous ground cover. Gopher tortoi ses feed primarily on herbaceous species from several plant families (Macdonald and Mushinsky, 1988; Mushinsky and Gibson, 1991; Mushinsky et al, 2003). Open areas cleare d by recent fire increase available forage for gopher tortoises by supporting greater spec ies richness and biomass production of herbaceous plants (Brockway and Lewis, 1997; Platt et al, 2006), and also provide tortoises with sites for thermoregulat ion and egg deposition (Diemer, 1986). Consequently, increases in canopy cover and lo ss of herbaceous ground cover from a lack of prescribed fire reduce habitat quality for gopher tortoises (McCoy et al, 2006; Mushinsky and McCoy, 1994). The lower degree of separation between individu al trap arrays relativ e to plots or burn frequency treatments also suggests that commun ities at these smaller spatial scales are responding to differences in microhabitats that may exist within burn treatments. Other herpetofaunal species respond to different combinations of environmental cues than lizards or tortoises, or may be responding to similar cues at different spatial scales than the one measured (Table 6). The variation among species in responses to microhabitat variables supports recommendations to implement prescribed fire regimes that maximize heterogeneity in the structural components of the sandhill plant co mmunity in order to maximize diversity of reptile and amphibian species, as well as that of other taxonomic groups (Mushinsky, 1992; Mushinsky and Gibson, 1991; Van L ear et al, 2005). Richness, diversity, and evenness indices did not have the discriminatory ability to distinguish differences in community com position between seven-ye ar fire frequency 23

PAGE 33

treatments and reference plots. These indice s, however, revealed differences in the species-abundance patterns among plots burned in November 2003 that the compositional analyses did not detect. While diversity in dices may more accurat ely describe speciesabundance distributions, compositional analyses th at retain species identities can be more sensitive in identifying differences among treatments with similar species-abundance distributions but different ra nks of species abundance. 24

PAGE 34

Annual Variation in Herpetofaunal Community Composition in a Periodically Burned Florida Sandhill Introduction Florida high pine habitats, such as sandhills and clayhills, are characterized by having a natural fire return interval ranging from 1-10 years, which maintains 1) a sparse overstory of longleaf pine (Pinus palustris ) or slash pine (P. elliottii ), 2) a dense understory of herbaceous vegeta tion dominated by wiregrass ( Aristida stricta ), and 3) a low density midstory of hard woods, usually turkey oak ( Quercus laevis ) (Myers, 1990). Less than 3% of the original high pine habitats in Florida remain, and the vast majority of the existing high pine is hi ghly fragmented and degraded (Kautz, 1993; Myers, 1990); approximately 75% of fragments are less than 40 ha in area and 30% are less than 8 ha. Many of these habitat fragments exist w ithin a matrix of urban and suburban developments (Heuberger and Putz, 2003). Ecological processes often are altered in fragmented ha bitats, including increased likelihood of invasion by non-native species, increased susceptibility to extinction for populations occurring within the fragment, and altered gr oundwater levels and nutrient input (Heuberger and Putz, 2003). Fire suppre ssion in Florida sandhill habitats causes an increase in hardwood density, canopy cover and leaf litter dept h and a concomitant decrease in herbaceous understory vegetati on (Brockway and Lewis, 1997; Kautz, 1993; Mushinsky, 1985, 1992; Myers, 1990). 25

PAGE 35

The effects of prescribed fire on species ri chness and diversity have been studied in several ecological systems. Reintroduction of fire to longleaf pine sandhills at Eglin Air Force Base in northwest Florida increased species richness of plants, arthropods, herpetofauna, and birds (Provencher et al., 2003 ). Diversity of reptiles and amphibians on burned plots in a west-centr al Florida longleaf pine sa ndhill was higher than on unburned plots in the early 1980s (Mushinsky, 1985). In 2004-5, however, diversity and evenness of reptile and amphibian communities were higher in plots that were not recently burned at the same study site (Halstead, in press). To better understand the long-term eff ects of fire on reptile and amphibian communities, I examined herpetofaunal comm unity composition at the University of South Floridas (USF) Ecological Research Area (ERA) to answer the following questions. 1) Is community compositi on constant over time? 2) If community composition varies over time, what envir onmental factors may affect it? 3) Does prescribed fire maintain a stable commun ity composition? I found annual variation in community composition was correlated with su mmer rainfall. Fire frequency affects community composition, but the pattern of change in community composition over time is not different between experimentally burned plots and unburned reference plots. Materials and Methods Experimental Design Experimentally burned plots (~ 0.5 1 ha) were established in an approximately 12 ha portion of longleaf pine sandhill in the University of South Floridas (USF) Ecological Research Area (ERA) in the mid-1970s. Origin ally, there were 11 experimental plots 26

PAGE 36

with fire frequencies ranging from oneto nine-year fire return intervals, and two unburned reference plots (Fig 4). The e xperimental design was altered in 1982, combining plots and replicating fire freque ncy treatments so that there were eight experimental plots with two replicates of one -, two-, fiveand seven-year fire return intervals. The two unburned reference plots, portions of both one-year plots, both fiveyear plots, and one seven-year pl ot remained unchanged (see Fig 1). Figure 4. Map of experimentally burned plots in the USF ERA during the 1980s. Numbers refer to fire return interval (years). Five drift fence/pitf all trap arrays were placed in each plot approximately 20-30 m apart with a random orientation (Mushinsky 198 5, 1992). Each trap array consisted of a 6 m long by 0.5 m wide drift fen ce (aluminum flashing), buried on edge 10-15 cm in the ground. At both ends of each drift fence, a 19 L plastic bucket was placed in the ground with the top edge at ground level. Holes we re drilled in the bottom of the buckets to allow for drainage of rainwater, and pieces of Styrofoam or cellulose sponge (approximately 10 x 10 x 2 cm) were placed in each bucket as a shade cover/flotation 27

PAGE 37

device. A cover object of 2 m long by 0.5 m wide galvanized roofing material was placed on the ground within 5 m of each trap array (Mushinsky 1985, 1992). Trap arrays were checked several da ys per week from January 1984 through December 1987, and from January through D ecember 2004. All captured individuals were identified to species and marked indi vidually by toe-clippi ng (Waichman 1992). Snakes were scale-clipped or marked with passive-integrated transponder (PIT) tags (AVID Technology, Tewksbury, MA). Data Analysis Number of individuals captured per year was tested with a Kruskal-Wallace test, followed by pairwise Kolmogorov-Smirnov tests. Species richness was estimated using the observed count of number of species (S) and rarefaction (Hurlb ert, 1971). Hills diversity numbers N 1 and N 2 (Hill, 1973) were used to measure species diversity as recommended by Ludwig and Reynolds (1988). Simpsons index (1; Simpson, 1949) and the Shannon-Weaver index (H; Shannon a nd Weaver, 1949) were also calculated for diversity because of their common use in the literature and their relation to N 2 and N 1 respectively. Evenness of relative abundances was measured with the modified Hills ratio (F 2,1 ; Alatalo, 1981). Significant variation among index values was determined using the randomization tec hnique described by Solow (1993). When differences among index values were suggested, 95% confid ence intervals were bootstrapped around the index of interest for each year. Nonparametric analyses of community co mposition were performed using PRIMER (v6, PRIMER-E, Ltd., 2006). Similarity between sampling units (trap arrays, plots, and 28

PAGE 38

treatments) was calculated using the Bray -Curtis index (Bray and Curtis, 1957) on square-root transformed count s of species abundances. S quare-root transformation was used to down-weight the effect of very abundant species (Clarke and Gorley, 2006). Cluster analysis was performed using the unweighted complete linkage method (Ludwig and Reynolds, 1988) with a post-hoc randomization procedure to identify groups of plots with different community co mpositions. The post-hoc randomization test was accomplished through the SIMPROF routine in PRIMER. The SIMPROF procedure tests for significant differences in commun ity composition at each node in the cluster dendrogram (Clarke and Gorley, 2006). Comple te linkage was used as the clustering technique because the go al of the procedure was to identify groups that were the most different. Nonmetric multidimensional scaling (NMS) was used for ordination, as recommended by Ludwig and Reynolds (1988). Analysis of similarities (ANOSIM) was performed to test for differences in community composition among years. ANOSIM is a nonparametric analogue to ANOVA in that it tests against the null hypot hesis that there is no difference in community composition among treatments. The test statistic used in ANOSIM is R, which is a measure of the observed diffe rences among rank similarities between treatments relative to the observed differen ces among rank similarities within treatments. Under the null hypothesis of no treatment eff ect, R would be close to 0. If all rank similarities within treatments are greater than all rank similarities between treatments R is equal to 1. The null distribution of R is cal culated in PRIMER by repeatedly randomly shuffling all sample labels and recalculati ng R for each permutation. A one-tailed test of significance is determined by the calculating the proportion of randomized R statistics 29

PAGE 39

that are greater than the obs erved value of R. The ANOSIM test calculates the observed R statistic among all treatments (Global R), as well as R statistics between each pairwise combination of treatments. Like ANOVA, if the Global R statistic is not significant, pairwise comparisons are ignor ed (Clarke and Warwick, 2001). Species that contributed to differences in community composition between different groups were identified using the similarity percentages (SIMPER) routine in PRIMER (Clarke and Gorley, 2006). Species that contri buted most to the overall pattern of BrayCurtis similarity were identified using a modified BIOENV procedure in PRIMER that calculated the Spearman rank correlation values between Bray-Curtis similarity values from progressively larger subsets of species and the Bray-Curtis similarity matrix from the full species set. Randomization tests th at randomly shuffled plot labels and recalculated the Spearman rank correlation were used to determine significance of r S for each number of species in the subset (Clarke and Warwick, 2001). The BIOENV procedure in PRIMER was used to correlate the pattern of sim ilarities in community composition with summer rainfa ll (Clarke and Gorley, 2006). A further step in the analysis of long-te rm changes in community composition was to determine if the pattern of annual change in community composition was different in unburned plots than in burned plots. For this test, a second-stage an alysis was performed. Second-stage analysis accounted for the main effect of yearly va riation by calculating a Spearman rank correlation matrix of the pa ttern of annual change in Bray-Curtis similarity values for each replicate (Clarke et al, 2006). An NMS plot of these correlation values was then constructed to show the s econd-stage pattern. Replicates located close together on the NMS indicated that those repl icates shared a similar pattern of annual 30

PAGE 40

variation in community composition (Clark e et al, 2006). ANOSIM of the second-stage correlation matrix was used to formally test fo r differences in pattern of annual change in community composition among treatments. Secondstage analysis was also performed in a reversed direction that cons tructed a correlation matrix of the pattern of change in community composition among fire frequency tr eatments for each year. No formal test for differences among years was possible for this analysis because years were not replicated. Second-stage analyses were onl y performed for experimental plots that remained (at least in part) unchanged between the two sampling periods. Results The number of individuals captured per plot was higher in 2004 than in all other years (p 0.001) and higher in 1985 than in 1984 (p = 0.020), 1986 (p = 0.017) and 1987 (p = 0.009; Table 7). Summer rainfall (May through September) was significantly correlated with number of individuals captured per plot (R 2 = 0.99, p = 0.001; Fig 5). 31

PAGE 41

Table 7. Number of individuals captured (N), species richness (S), and Hills diversity (N 1 N 2 ) and evenness (F 2,1 ) numbers with upper and lower 95% confidence interval limits for each year. Confidence interv als were not calculated for species richness because no difference was dete cted by a randomization test. N 1 N 2 F 2,1 95% Confidence Interval 95% Confidence Interval 95% Confidence Interval Year N S Observed value Lower Upper Observed value Lower Upper Observed value Lower Upper 1984 726 25 7.59 6.82 8.12 5.00 4.52 5.47 0.61 0.58 0.65 1985 1387 22 6.33 5.83 6.55 4.79 4.54 5.02 0.71 0.70 0.76 1986 720 21 6.42 5.83 6.84 4.57 4.19 4.94 0.66 0.63 0.70 1987 631 17 5.74 5.16 6.21 3.69 3.32 4.09 0.57 0.54 0.62 2004 2032 24 7.76 7.19 8.01 5.00 4.72 5.29 0.59 0.58 0.63 0.0 50.0 100.0 150.0 200.0 250.0 75.0080.0085.0090.0095.00100.00105.00110.00115.00120.00125.00 May-September Rainfall (cm)Number of Individuals Captured/Plot Figure 5. Mean number of individuals captu red per plot (y-axis) versus summer rainfall (x-axis) for each year. 32

PAGE 42

Species richness among years was not different (p = 0.5076). The number of abundant (N 1 ) and very abundant (N 2 ) species varied significantly among years (N 1 : p = 0.0001; N 2 : p = 0.0001), as did evenness (F 2,1 : p = 0.0014). The number of abundant species was higher in 2004 than in all ot her years except 1984, a nd higher in 1984 than all years except 2004 and 1986 (Table 7). The number of very abundant species was lower in 1987 than all other years (Table 7) Evenness was higher in 1985 than all other years, and it was higher in 1986 than in al l other years except 1985 and 1984 (Table 7). Differences in community composition were greater between years than between fire frequency treatments. Cluster analysis in conjunction with the SIMPROF test revealed five distinct groups of expe rimental plots: one entirely of plots in 2004, one almost entirely of plots in 1985, and one large group that encompassed most of the plots in 1984, 1986, and 1987; the other two distinct groups consisted of a few plots in 1984 and 1987 that had distinct community compositions (p < 0.05). A similar pa ttern was revealed by NMS, showing a general shift in community composition between years with the largest separation between 2004 and all ot her years (stress = 0.16; Fig 6). 33

PAGE 43

Year 1984 1985 1986 1987 2004 2D Stress: 0.16 Figure 6. Nonmetric multidimensional scaling plot of experimentally burned plots among years. Stress measures the goodness-of-fit of the data to the pattern; lower stress indicates a better fit. The separation among years was significan t (Global R = 0759, p = 0.001) across all fire frequency treatments, using a two-wa y crossed ANOSIM. Pairwise tests showed differences between all pairs of years except 1984 and 1987, and 1986 and 1987 (Table 8). The separation between fire frequency tr eatments across all years in the same twoway crossed analysis was smaller than the se paration between years, but still significant (Global R = 0.385, p = 0.001). Pairwise comparisons of fire frequency treatments showed a more complicated pattern of differences (Table 9). The difference in community composition among years was also significantly correlated with summer rainfall (r S = 0.585, p = 0.01). 34

PAGE 44

Table 8. Pairwise R values between y ears from a two-way crossed ANOSIM. Comparison R p 1984 vs. 1985 0.8 0.008 1984 vs. 1986 0.75 0.004 1984 vs. 1987 0.4 0.082 1984 vs. 2004 1 0.004 1985 vs. 1986 0.45 0.016 1985 vs. 1987 0.8 0.004 1985 vs. 2004 1 0.004 1986 vs. 1987 0.5 0.074 1986 vs. 2004 1 0.004 1987 vs. 2004 0.945 0.004 Table 9. Pairwise R values between fire fr equency treatments from a two-way crossed ANOSIM. There were not enough replicates of the 2 and 9 year fire frequency treatments to make a pairwise comparison. Comparison R p 1 vs. 2 -0.34 0.897 1 vs. 3 0.25 0.099 1 vs. 5 0.004 0.400 1 vs. 7 0.25 0.148 1 vs. 9 1 0.012 1 vs. no fire 0.5 0.016 2 vs. 3 -0.25 0.815 2 vs. 5 0.83 0.021 2 vs. 7 0.66 0.062 2 vs. no fire 0.83 0.021 3 vs. 5 0.313 0.148 3 vs. 7 -0.188 0.778 3 vs. 9 1 0.012 3 vs. no fire 0.125 0.321 5 vs. 7 0.35 0.025 5 vs. 9 0.5 0.185 5 vs. no fire 0.55 0.025 7 vs. 9 0.75 0.062 7 vs. no fire -0.1 0.728 9 vs. no fire 1 0.012 35

PAGE 45

In general, the change in abundance of Gastrophryne carolinensis was most characteristic of between-year differences in community composition. Other species that contributed to between y ear differences included Aspidoscelis sexlineata Tantilla relicta Anaxyrus terrestris, Scincella lateralis and Coluber constrictor However, the relative contribution of each species to the dissim ilarity between pairs of years changed depending on the years being compared. Differences between years in the 1980s were more characterized by th e relative abundances of Scaphiopus holbrookii Kinosternon baurii, and Plestiodon inexpectatus than differences between 2004 and any year in the 1980s. Likewise, differences between 2004 and y ears in the 1980s were more influenced by the abundances of Eleutherodactylus planirostris and Anolis sagrei, than differences between pairs of years in the 1980s. The abundances of ten species could produce a pattern among Bray-Curtis similarities almost identical to that produced by the full species set ( r S = 0.957; p = 0.01). Gastrophryne carolinensis abundances alone produced Bray -Curtis similarities that had a Spearman rank correlation value of 0.527 with the Bray-Curtis similarity matrix from all species (p=0.01). Second-stage analysis of patterns of annual variation within fire frequency treatments revealed no differences between long-ter m change in community composition in unburned plots relative to the pattern of annua l variation in burned treatments (Global R = -0.019, p = 0.543; Fig 5). The second stage anal ysis of the pattern of similarity among fire frequencies within years suggested that there are differences between years in the pattern of among plot Bray-Curtis similar ities, though no formal test was possible. 36

PAGE 46

Discussion Annual variation in summer rainfall has a large effect on comm unity composition of reptiles and amphibians in a fire-maintai ned ecosystem. Separation among years was higher than among fire frequencies (ANOSIM Global R values of 0.759 and 0.385, respectively). Research in other fire-maintai ned systems has also shown annual variation in abundances of plants, small mammals, bird s, and insects associated with changes in climate (Mushinsky and Gibson, 1991). Summer rainfall explained nearly 60 pe rcent of the variation in community composition among years, and 99 percent of th e variation in numb er of individuals captured among years. Variati on in number of individual re ptiles captured per plot and number of amphibians captur ed per plot was still well-explained by summer rainfall when analyzed separately (R 2 = 0.89, and 0.98, respectively). When captures of species were analyzed separately, 17 of the 35 speci es recorded were si gnificantly positively correlated with annual rainfall. The increase in abundances of most species in years with greater rainfall is likely because of increases in reproductive activ ity, hatching success, and/or juvenile survivorship. Most amphibian species require standing water to breed, and in relatively wet years standing water in the surroundi ng floodplain swamp expands to surround one side of the study site. Increas es in captures of amphibia ns are probably a result of increased area of breeding hab itat near the sampling area, increased reproductive effort by adults, and shorter distances required for both adults and juve niles to disperse into the nearby uplands from breeding sites. Future re search at the study site could determine the 37

PAGE 47

specific mechanisms by which amphibian ca ptures increase during relatively wetter years. Reptiles, however, do not require standing water for reproduction. An increase in rainfall can cause increases in soil moistu re availability which can promote egg and hatchling survivorship (Ove rall, 1994; Philips and Packard, 1994; Socci et al, 2005; Warner and Andrews, 2002). Rainfall also is likely to influence ve getative growth, which can in turn increase prey abundance for both ju venile and adult reptiles and amphibians. Only the southeastern five-lined skink, Plestiodon inexpectatus exhibited a moderate (but non-significant) negative relationship with annual rainfall. Skinks were most abundant in 1984 and 1987, the two driest years, and least abundant in 2004, the wettest year. Skinks respond negatively to prescrib ed fire (Mushinsky, 1992), however, and 1984 and 1987 experienced a lower proportion of si tes burned than in other years. Skink abundance was negatively correlated with the pr oportion of sites burned in each year, so it is likely that the negative relationship be tween skink abundance and summer rainfall is coincidental. By chance, summer rainfall and the proportion of plots burned were positively correlated for the five years presented here. Therefore, it is likely that some of the species that respond positively to recent fire, such as the six-lined racerunner, Aspidoscelis sexlineata may not be responding as strong ly to summer rainfall as the correlations suggest. Community composition was usually different from one year to the next, with the exception of 1986 and 1987. Univariate measures of richness and diversity that do not retain species identities often failed to detect these changes. Species richness was not different among years. Hills numbers of abundant species (N 1 ) and very abundant 38

PAGE 48

species (N 2 ), and the modified Hills evenness ratio (F 2,1 ) detected relatively few differences among pairs of years. Indices th at describe speciesabundance relationships are often useful for making broad comparisons between systems that share few, or no, species, but are less useful when making comp arisons between sites within a habitat. Within years, fire frequency has a significant effect on community composition. Despite this, the pattern of annual variation in community composition within different treatments is not different between experi mentally burned and reference plots. This finding suggests that although burned plots have a significantly diffe rent composition of reptile and amphibian species than unburned plots, the long -term change within fire frequency treatments is not different. Because of the low number of replicates of each fire frequency treatment, however, the power to de tect differences in the pattern of annual variation is low. Pairwise second stage ANOSIM R statistics suggest that there may be a trend toward frequently burned plots cha nging over time in a different manner than unburned plots. Further monitoring of amphibian and reptile communities at the site is needed to develop a longer time series to detect any real differences that may exist. Adding more replicates of experimentally burned plots would also be useful, but is unfortunately impractical at the site, given th e limited availability of sandhill habitat. No apparent loss of species was detected at the site, although the fragmented nature of the site may have increased the probabil ity of extinction for species occurring in the study site. No species was consistently obser ved in the 1980s that was not also observed in 2004. Furthermore, most species abundances increased, with two notable exceptions. Plestiodon inexpectatus were not as abundant in 2004 as in the 1980s, which is likely a result of the high proportion of plots burned in 2003. The other exception was 39

PAGE 49

Scaphiopus holbrookii a species for which large fluctuati ons in observations of adults is well-documented (Gosner and Black, 1955; Greenberg and Tanner, 2004). The risk of extinction for several species, however, may have increased because of the fragmented nature of the study site, even though no loss of species was detected. Species that were uncommonly observed during the 1980s may well be locally extirpated at the study site. Continued monitoring is needed to determine if rare species are now absent from the site, and if other species that still occupy the site are experiencing long-term declines in abundance despite annual variation in abundance associated with rainfall. Although the sandhill plant community remains intact and there are relatively few invasive plant species in the study site (Schmidt, 2005), the increasingly urban area surrounding the study site, in combination with fragmentation, has made it susceptible to invasions by non-native reptiles and amphi bians. Two of three well-established nonnative amphibians, and one of four reptiles, are present at the study site. Two non-native species, the greenhouse frog, Eleutherodactylus planirostris and the brown anole, Anolis sagrei have become more common at the site an d were characteristic of differences in composition between 2004 and years in the 1980s. Changes in composition among years in the 1980s were characterized almost en tirely by native species, although both nonnative species were present at the site during the sampling peri od. A third non-native species, the Cuban treefrog, Osteopilus septentrionalis observed in 2004 was not recorded during the in itial sampling period. Eleutherodactylus planirostris has been established in Florida since the early 1900s, so it is unlikely that the incr ease in abundance observed for this species was associated with their invasion. Anolis sagrei has experienced several introductions to mainland 40

PAGE 50

Florida, and was established in Tampa as early as 1947 (Lee, 1985; Oliver, 1950). Because the abundances of most other speci es also increased in 2004, it is uncertain whether the increased abundance of A. sagrei was associated with th e annual variation in summer rainfall, or if it was associated with their establishment and subsequent invasion in Florida. Osteopilus septentrionalis was first recorded in mainland Florida in 1951 (Schwartz, 1952) and has subsequently expanded its range northward (Meshaka, 2001). The appearance of Osteopilus septentrionalis at the study site si nce the 1980s almost certainly reflects the increasing range and abundance of this species in the state of Florida. This species has several attributes that make it a successf ul colonizing species, including high fecundity, short generation times, broad diet, few predators, and superior competitive ability, and its northward e xpansion is likely only limited by climate (Meshaka, 2001). It is most successful in hardwood hammocks, mangrove swamps, and areas of human disturbance. The northwar d expansion of the Ta mpa metropolitan area and the increased density of hardwoods in unburned areas of the USF ERA almost certainly facilitated the invasion of this spec ies at the study site, a nd it is likely that Cuban treefrogs will continue to persist. Cuban treefrogs have been associated with declines of native hylids in South Florida and the Everglades, especially in tropical hardwood hammocks and disturbed upland areas (Meshaka, 2001). It is likely that the im pact of invasion of the Cuban treefrog will be more severe in natural areas farther north where it will compete with other native species that may have a great er overlap in habitat prefer ence, such as the barking treefrog, Hyla gratiosa, and the pinewoods treefrog, H. femoralis The larger size, broader diet (including other hy lids), and high fecundity of the O septentrionalis will 41

PAGE 51

almost certainly have negative effects on th ese native hylids where their ranges overlap. Hyla femoralis occurs at the study site, which prov ides an opportunity to examine the response of this species to the invasion of the Cuban treefrog. The giant toad, Bufo marinus is locally abundant in Tampa and has been observed on the south side of Fletcher Avenue and the Un iversity of South Flor ida golf course (T. Piacenza, pers. comm.), both adjacent to the study site, but has not been documented inside the study site to date. Reptile species common in Tampa, but not observed in the study site, include the Mediterranean gecko, Hemidactylus turcicus the Indo-Pacific gecko, Hemidactylus garnotii and the Brahminy blind snake, Ramphotyphlops braminus, all of which occur most frequently on or near human habitations (pers. obs.). Numerous other exotic species occur in Fl orida (Simberloff et al., 1997), including at least 28 lizard species (Butterfield et al., 1997). The ma jority of these species do not occur at the study site, but two lizard species in particular represent a serious threat of invading the site. The Nile monitor, Varanus niloticus, has been observed in southwest Florida as early as 1990, and is currently we ll-established and expanding its range, which could possibly include all appropriate habitats in peninsular Florida (Enge, et al., 2004). This large, carnivorous lizard does particul arly well in areas disturbed by humans and adjacent to permanent water, so it is likel y that the study site may be invaded if populations of V. niloticus become established in the Tampa area. The responses of native species to this invasion are largely unknown, but the Nile monitor is likely to have detrimental effects on populations of many speci es, especially burrowing species such as the gopher tortoise (Enge et al., 2004). 42

PAGE 52

A more immediate threat to this study site is represented by the re cent discovery of an apparently established population of the Argentine black and white tegu, Tupinambis merianae southeast of Tampa (Enge, et al. 2006) Like Nile monitors, tegus also use burrows as refugia and juveniles in Florid a have been trapped from gopher tortoise burrows. Although impacts of tegus on species indigenous to Florida have not been assessed, tegus are generally similar to m onitor lizards in morphology (Losos, 1994) and probably represent at least as much of a thr eat to native species as the Nile monitor does (L. A. Fitzgerald, pers. comm.) The threat of invasion from these two large lizards at the study site underscores the need for continuous monitoring to determine the impact that invasion may have on native amphibians and re ptiles, and also to develop potential management plans to help mitigate any detrimental effects on native species. 43

PAGE 53

Literature Cited Alatalo, R.V. 1981. Problems in the m easurement of evenness in ecology. Oikos 37: 199-204. Bray, J. R. and J. T. Curtis. 1957. An or dination of the upland forest communities of southern Wisconsin. Ecological Monographs 27: 325-349. Brockway, D. G. and C. E. Lewis. 1997. Long-term effects of dormant-season prescribed fire on plant community divers ity, structure and produc tivity in a longleaf pine wiregrass ecosystem. Forest Ecology and Management 96: 167-183. Butterfield, B. P., W. E. Me shaka Jr., and C. Guyer. 1997. Nonindigenous amphibians and reptiles. In D. Simberloff, D. C. Schmitz, and T. C. Brown (Eds.), Strangers in paradise: Impact and management of nonindige nous species in Florida. Island Press, Washington, D.C., pp. 123-138. Clarke, K. R. and R. N. Gorley. 2006. PR IMER v6: User Manual/Tutorial. PRIMER-E: Plymouth. Clarke, K. R., P. J. Somerfield, L. Airo ldi, and R. N. Gorley. 2006. Exploring interactions by second-stage community analyses. Journal of Experimental Marine Biology and Ecology 338: 179-192. Clarke, K. R. and R. M. Warwick. 2001. Change in marine communities: an approach to statistical analysis a nd interpretation, 2nd edition. PRIMER-E: Plymouth. Cooper, W. E. Jr, A. M. Al-Johany, L. J. V itt, and J.J. Habegger. 2000. Responses to chemical cues from animal and plant f oods by actively foraging insectivorous and omnivorous scincine lizards. Journal of Experimental Zoology 287: 327-339. Cooper, W. E. Jr, and R. Hartdegen. 1999. Discriminative response to animal, but not plant, chemicals by an insect ivorous, actively foraging lizard, Scincella lateralis and differential response to surface and internal prey cues. Journal of Chemical Ecology 25: 1531-1541. Dickman, C. R. 1987. Habitat fragmentation a nd vertebrate species richness in an urban environment. Journal of Applied Ecology 24: 337-351. 44

PAGE 54

Diemer, J. E. 1986. The ecology and mana gement of the gopher tortoise in the southeastern United States. Herpetologica 42: 125-133. Drewa, P. B., W. J. Platt, and E. B. Moser. 2002. Fire effects on re sprouting of shrubs in headwaters of southeastern longleaf pine savannas. Ecology 83: 755-767. Drewa, P. B., J. M. Thaxton, and W. J. Pla tt. 2006. Responses of root-crown bearing shrubs to differences in fire regimes in Pinus palustris (longleaf pine) savannas: Exploring old-growth questions in second-growth systems. Applied Vegetation Science 9: 27-36. Enge, K. M., B. W. Kaiser, and R. B. Dick erson. 2006. Another la rge exotic lizard in Florida, the Argentine black and white tegu. Abstract in Proceedings of the 28 th Annual Gopher Tortoise Council Meeti ng, 26-29 October 2006, Valdosta, GA. Enge, K. M., K. L. Krysko, K. R. Hankins, T. S. Campbell, and F. W. King. 2004. Status of the Nile monitor ( Varanus niloticus ) in southwestern Florida. Southeastern Naturalist 3: 571-582. Glitzenstein, J. S., D. R. Streng, and D. D. Wade. 2003. Fire frequency effects on longleaf pine (Pinus palustris P. Miller) vegetation in S outh Carolina and northeast Florida, USA. Natural Areas Journal 23: 22-37. Glitzenstein, J. S., W. J. Platt, and D. R. Streng. 1995. E ffects of fire regime and habitat on tree dynamics in north Florid a longleaf pine savannas. Ecological Monographs 65: 441-476. Gosner, K. L. and I. H. Black. 1955. The effects of temperature and moisture on the reproductive cycle of Scaphiopus h. holbrooki American Midland Naturalist 54: 192-203. Greenberg, C. H and G. W. Tanner. 2004. Breeding pond selection and movement patterns by eastern spadefoot toads (Scaphiopus holbrookii ) in relation to weather and edaphic conditions. Journal of Herpetology 38: 569-577. Heuberger, K. A. and F. E. Putz. 2003. Fire in the suburbs: ecological impacts of prescribed fire in small remnants of longleaf pine ( Pinus palustris ) sandhill. Restoration Ecology 11: 72-81. Hill, M. O. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology 54: 427-432. Hurlbert, S. H. 1971. The non-concept of sp ecies diversity: A crit ique and alternative parameters. Ecology 52: 577-586. 45

PAGE 55

Jellinek, S., D. A. Driscoll, and J. B. Ki rkpatrick. 2004. Environmental and vegetation variables have a greater in fluence than habitat fragmentation in structuring lizard communities in remnant urban bushland. Austral Ecology 29: 294-304. Kautz, R. S. 1993. Trends in Florida wildlife habitat 1936-1987. Florida Scientist 56: 7-24. Kjoss, V. A. and J. A. Litvaitis. 2001. Community structure of snakes in a humandominated landscape. Biological Conservation 98: 285-292. Knight, T. M. and R. D. Holt. 2005. Fire ge nerates spatial gradient s in herbivory: An example from a Florida sandhill ecosystem. Ecology 86: 587-593. Lee, J. C. 1985. Anolis sagrei in Florida: Phenetics of a colonizing species I. Meristic characters. Copeia 1985: 182-194. Lemon, P. C. 1949. Successional responses of herbs in the longleaf-slash pine forest after fire. Ecology 30: 135-145. Losos, J. B. 1994. Historical contingency and lizard community ecology. In L. J. Vitt and E. R. Pianka (Eds.), Li zard ecology: Historical and experimental perspectives. Princeton University Press, Princeton, NJ, pp. 319-333. Ludwig, J. A. and J. F. Reynolds. 1988. St atistical ecology: a primer on methods and computing. John Wiley & Sons, Inc., New York, NY. Margalef, R. 1958. Information theory in ecology. General Systematics 3: 36-71. Macdonald, L. A. and H. R. Mushinsky. 1988. Foraging ecology of the gopher tortoise, Gopherus polyphemus, in a sandhill habitat. Herpetologica 44: 345-353. May, R. M. 1975. Patterns of species abundance and diversity. In M. L. Cody and J. M. Diamond (Eds.) Ecology and Evolution of Communities. Belnap Press, Cambridge, MA, pp. 81-120. McCoy, E. D., H. R. Mushinsky and J. Lind zey. 2006. Declines of the gopher tortoise on protected lands. Biological Conservation 128: 120-127. Mehlman, D. W. 1992. Effects of fire on pl ant community composition of north Florida second growth pinelands. Bulletin of the Torrey Botanical Club 119: 376-383. Meshaka, W. E. Jr. 2001. The Cuban treefrog: Life history of a successful colonizing species. University Press of Florida, Gainesville, FL. 46

PAGE 56

Mushinsky, H. R. 1985. Fire and the Fl orida sandhill herpetofaunal community: with special attention to responses of Cnemidophorus sexlineatus. Herpetologica 41: 333342. . 1992. Natural history and abundance of Southeastern Five-lined Skinks, Eumeces inexpectatus on a periodically burned sandhill in Florida. Herpetologica 48: 307312. Mushinsky, H. R. and D. J. Gibson. 1991. Th e influence of fire periodicity on habitat structure. In S. S. Bell, E. D. McCoy, and H. R. Mushinsky (Eds.), Habitat structure: The physical arrangement of objects in space. Chapman & Hall, London, UK, pp. 237-259. Mushinsky, H. R. and E. D. McCoy. 1994. Comparison of gopher tortoise populations on islands and on the mainland in Florida. In R. B. Bury and D. J. Germano (Eds.), Biology of North American Tortoises. United States Department of the Interior, National Biological Survey, Fish and Wildlife Research Report 13, Washington, D.C., pp. 39-47. Mushinsky, H. R., T. A. Stilson, and E. D. Mc Coy. 2003. Diet and dietary preference of the juvenile gopher tortoise ( Gopherus polyphemus ). Herpetologica 59: 475-483. Myers, R. L. 1990. Scrub and high pine. In R. L. Myers and J. J. Ewel (Eds.), Ecosystems of Florida. University Pre sses of Florida, Gainesville, FL, pp. 150-193. Oliver, J. A. 1950. Anolis sagrei in Florida. Copeia 1950: 55-56. Olson, M. S. and W. J. Platt. 1995. Eff ects of habitat and growing-season fires on resprouting of shrubs in longleaf pine savannas. Vegetatio 119: 101-118. Overall, K. L. 1994. Lizard egg environments. In L. J. Vitt and E. R. Pianka (Eds.), Lizard ecology: Historical and experiment al perspectives. Princeton University Press, Princeton, NJ, pp. 51-72. Paulissen, M. A. 1987. Optimal foraging and intraspecific diet differences in the lizard Cnemidophorus sexlineatus. Oecologia 71: 439-446. Pilliod, D. S., R. B. Bury, E. J. Hyde, C. A. Pearl, and P. S. Corn. 2003. Fire and amphibians in North America. Forest Ecology and Management 178: 163-181. Platt, W. J., S. M. Carr, M. Reilly, and J. Fahr. 2006. Pine savanna overstorey influences on ground-cover biodiversity. Applied Vegetation Science 9: 37-50. Phillips, J. A. and G. C. Packard. 1994. In fluence of temperature and moisture on eggs and embryos of the white-th roated savanna monitor, Varanus albigularis : Implications for conservation. Biological Conservation 69: 131-136. 47

PAGE 57

Provencher, L., A. R. Litt, and D. R. Gordon. 2003. Predictors of species richness in northwest Florida longleaf pine sandhills. Conservation Biology 17: 1660-1671. Russell, K. R., D. H. Van Lear, and D. C. Guynn, Jr. 1999. Prescrib ed fire effects on herpetofauna: review and ma nagement implications. Wildlife Society Bulletin 27: 374-384. Schmidt, A. C. 2005. A Vascular Plant Inventory and Description of the Twelve Plant Community Types Found in the University of South Florida Ecological Research Area, Hillsborough County, Florida. M.S. Thesis. University of South Florida, Tampa, FL. Schurbon, J. M. and J. E. Fauth. 2003. E ffects of prescribed burning on amphibian diversity in a southeastern U. S. national forest. Conservation Biology 17: 13381349. Schwartz, A. 1952. Hyla septentrionalis Dumeril and Bibron on the Florida mainland. Copeia 1952: 117-118. Shannon, C. E. and W. Weaver. 1949. The mathematical theory of communication. University of Illinois Press, Urbana, IL. Simberloff, D., D. C. Schmitz, and T. C. Brown. (Eds.) 1997. Strangers in paradise: Implications and management of nonindigenous species in Florida. Island Press, Washington, D.C. Simpson, E. H. 1949. Measurement of diversity. Nature 163:688. Socci, A. M., M. A. Schlaepfer, and T. A. Gavin. 2005. The importance of soil moisture and leaf cover in a female lizards ( Norops polylepis ) evaluation of potential oviposition sites. Herpetologica 61:233-240. Solow, A. R. 1993. A simple test fo r change in community structure. Journal of Animal Ecology 62: 191-193. Van Lear, D. H., W. D. Carroll, P. R. Kapeluck, and R. Johnson. 2005. History and restoration of the longleaf pine-grassland ecosy stem: Implications for species at risk. Forest Ecology and Management 211: 150-165. Vitt, L. J. and W. E. Cooper, Jr. 1986. Foraging and diet of a diurnal predator, Eumeces laticeps feeding on hidden prey. Journal of Herpetology 20: 408-415. Waichman, A. V. 1992. An alphanumeric c ode for toe clipping amphibians and reptiles. Herpetological Review 23: 19-21. Warner, D. A. and R. M. Andrews. 2002. Ne st-site selection in relation to temperature and moisture by the lizard Sceloporus undulatus Herpetologica 58: 399-407. 48