<%BANNER%>

Use of amphibians as ecosystem indicator species

HIDE
 Title Page
 Dedication
 Acknowledgement
 Table of Contents
 List of Tables
 List of Figures
 Abstract
 Introduction
 Sampling methods for amphibian...
 Using site occupancy modeling to...
 The effect of toe-clipping on two...
 Influence of hyrdrology on survival...
 Conclusion
 References
 Biographical sketch
University of Florida Institutional Repository

PAGE 1

1 USE OF AMPHIBIANS AS EC OSYSTEM INDICATOR SPECIES By JAMES HARDIN WADDLE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

2 Copyright 2006 by James Hardin Waddle

PAGE 3

3 This dissertation is dedicated to my parents, Chris and Sherrell Waddle, who have always supported me and encouraged me to do what I love, and to Amanda, the love of my life.

PAGE 4

4 ACKNOWLEDGMENTS I would first like to thank the members of my committee. Ken Rice has been a great mentor and friend over the last six years. His guidance has helped prep are me to be a better scientist. Franklin Percival has truly taken me under his wing and taught me about philosophy. Jim Nichols was always available for help when I needed it and has consistently provided me with insightful feedback on my work. Frank Mazzotti was always supportive of me during my time as a student. Harvey Lillywhite has been a valuable member of the committee, and forced me to think more about what I think I know. I would also like to thank the people who assisted me in th e field. Brian Jeffery, Andy Maskell, Chris Bugbee, Meghan Riley, and Debbi e Kramp all spent many long hours with me slogging through swamps. Many members of the Ft. Lauderdale REC Mazzotti and Rice lab and the Florida Alligator and Amphibian Research Team (FAART) volunteered for the sampling experiment in Chapter 2. Other volunteers fr om the Big Cypress National Preserve Student Conservation Association provided valuable help All of the FAART pe ople have also provided useful feedback on my wo rk and great friendship. The staff of Big Cypress National Preserve was very helpful in a ll stages of this work. Big Cypress provided an office and housing for this project, and granted access and permits for the research. Deb Jansen and Ron Clark enga ged me in many good conversations about management and wildlife in the Preserve. Jim Burch shared plant knowledge, Bob Sobczack helped with hydrologic questions, and Frank Partridge provided ma ps and GIS data. Jim Snyder of the USGS allowed me to share equipment and space and helped watch out for me. Finally I would like to thank my family for a ll of their love and support. My new wife, Amanda is always my biggest supporter. My pa rents, and my sister Virginia and her husband Brian Lott all deserve thanks for supporting me through my graduate career.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................4 LIST OF TABLES................................................................................................................. ..........7 LIST OF FIGURES................................................................................................................ .........9 ABSTRACT....................................................................................................................... ............11 CHAPTER 1 INTRODUCTION..................................................................................................................13 Characteristics of Indicator Species........................................................................................13 Suitability of Amphibian s as Indicator Species......................................................................14 Outline of the Dissertation.................................................................................................... ..15 2 SAMPLING METHODS FOR AMPHIBIAN MONITORING............................................19 Introduction................................................................................................................... ..........19 Methods........................................................................................................................ ..........20 Sampling Experiment......................................................................................................20 Double Observer Analysis...............................................................................................22 Distance Analysis............................................................................................................23 Results........................................................................................................................ .............23 Double Observer..............................................................................................................24 Distance Analysis............................................................................................................24 Discussion..................................................................................................................... ..........25 3 USING SITE OCCUPANCY MODELING TO DETERMINE THE EFFECT OF OFFROAD VEHICLE USE ON GR OUND-DWELLING ANURANS.......................................41 Introduction................................................................................................................... ..........41 Methods........................................................................................................................ ..........42 Study Area..................................................................................................................... ..42 Sampling....................................................................................................................... ...43 Data Analysis.................................................................................................................. .45 Results........................................................................................................................ .............45 Discussion..................................................................................................................... ..........46 4 THE EFFECT OF TOE-CLIPPING ON TWO SPECIES OF TREEFROGS.......................63

PAGE 6

6 Introduction................................................................................................................... ..........63 Methods........................................................................................................................ ..........64 Study Site..................................................................................................................... ....64 Capture Recapture...........................................................................................................65 Survival Analysis.............................................................................................................66 Results........................................................................................................................ .............66 Discussion..................................................................................................................... ..........68 5 INFLUENCE OF HYDROLOGY ON SURVIVAL AND RECRUITMENT OF GREEN TREEFROGS...........................................................................................................77 Methods........................................................................................................................ ..........78 Results........................................................................................................................ .............82 Discussion..................................................................................................................... ..........83 6 CONCLUSION..................................................................................................................... ..97 Introduction................................................................................................................... ..........97 Characteristics of Indicators.................................................................................................. .97 Abundant and Efficient to Sample..................................................................................97 Sensitive to Stresses on the System.................................................................................98 Responses to Stress Should Be Anticipatory...................................................................99 Integrate a Response across the Whole System..............................................................99 Known Response to Anthropogenic St resses and Natural Disturbances........................99 Conclusion..................................................................................................................... .......100 LIST OF REFERENCES.............................................................................................................102 BIOGRAPHICAL SKETCH.......................................................................................................110

PAGE 7

7 LIST OF TABLES Table page 2-1 Number of artificial frogs observed by each team in prairie and pineland habitats.........29 2-2 Actual detection rates of a known population of artificial frogs within 30 cm, within 50 cm, beyond 50 cm, and overall for teams s earching in prairie and pineland habitat....30 2-3 Model selection for the three double-observer models analyzed in Program SURVIV with the model log-likelihood, number of pa rameters (K), quasilikelihood Akaikes information criterion adjusted for small sample sizes (QAICc).........................................31 2-4 Estimates from program SURVIV of indivi dual detection probabil ities of artificial frogs for each observer in prairie and pinela nd habitat with standard error (S.E.) and 95% confidence intervals (C.I.) ta ken from the double observer method.........................32 2-5 Abundance estimates of artificial treefr ogs for each team in prairie and pineland habitat with standard error (S.E.) and uppe r and lower 95% confidence intervals (C. I.) based on Chaoss estimatorfr om the double observer data...........................................33 2-6 Model selection for detection function s in Program DISTANCE showing number of parameters (K), Akaikes Information Criterion (AIC), the difference between each AIC and the minimum (Delta AIC), a nd the AIC Weight of each model.........................34 2-7 Abundance estimates of artificial frogs by team for transects in the prairie and pineland habitats with coefficient of varia tion (CV) and 95% confidence interval (CI) from the distance sampling approach................................................................................35 3-1 Combinations of the 3 site covariates and 4 sampling covari ates that were used in the occupancy analysis for each species. Each set of site covariates was modeled along with each set of sampling covariates for a to tal of 80 unique models for each species.....50 3-2 Number of sampling sites and total number of site visits by habitat.................................51 3-3 Number of detections by species, and proportion of sites at which a detection occurred (naive occupancy) during amphibi an surveys across all habitat types...............52 3-4 Model selection results for the oak toad ( Bufo quercicus ), including Akaikes Information Criterion (AIC) and the de lta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight........................................................53 3-5 Model selection results for the southern toad ( Bufo terrestris ), including Akaikes Information Criterion (AIC) and the de lta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight........................................................54

PAGE 8

8 3-6 Model selection results for the greenhouse frog ( Eleutherodactylus planirostris ), including Akaikes Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight................................56 3-7 Model selection results for th e eastern narrow-mouthed toad ( Gastrophryne carolinensis ), including Akaikes Information Cr iterion (AIC) and the delta AIC and AIC weight (Burnham and Anderson 1998) fo r all models with any AIC weight............57 3-8 Sums of Akaikes Inform ation Criterion (AIC) weights for all models including the ORV index, habitat type, or hydr ologic index covariates fo r each of the four focal anuran species................................................................................................................. ...59 3-9 Beta estimates, standard errors (S.E.) and lower and upper 95% confidence intervals (C.I.) for the ORV use index covariate from the best model for each of the four focal anuran species................................................................................................................. ...60 4-1 List of 23 models analyzed in Program MARK for captures of both Green Treefrogs and Squirrel Treefrogs in Big Cypre ss National Preserve during 2004-2005. ................70 4-2 The number of green treefrogs and squirre l treefrogs marked by removing 2, 3, or 4 toes in Big Cypress National Preserve Collier County, FL, Nov. 2004-June 2005 and the return rate (proportio n of marked individuals r ecaptured at least once)......................71 4-3 Model selection table for Cormack-Jo lly-Seber open population mark-recapture model of Green Treefrogs including Quasilikelihood Akaikes Information Criterion for small sample sizes (QAICc).........................................................................................72 4-4 Estimates, standard error (SE), and the 95% confidence interval of the beta values for the toe-clip effect on apparent su rvival and recapture probability....................................73 4-5 Model selection table for Cormack-Jo lly-Seber open population mark-recapture model of Squirrel Treefrogs including Quasi-likeli hood Akaikes Information Criterion for small sample sizes (QAICc)..........................................................................74 5-1 Dates of each sample of the PVC pipe re fugia and the season to which each sample was assigned for mark-recapture analysis..........................................................................86 5-2 Model selection results for all models an alyzed in Program MARK. Model describes the covariates and groups associated with su rvival, capture probabi lity, and seniority....87 5-3 Estimates of the population growth rate the contribution of survival and the contribution of recruitment to growth for each sampling interval.....................................88

PAGE 9

9 LIST OF FIGURES Figure page 2-1 Histograms of the frequency of perpendicula r distances in cm of actual artificial frog placements along the 2 transect lines in the prairie and pineland habitat..........................36 2-2 Histograms of the frequency of detecti ons of artificial frogs by each team at perpendicular distances in cm along the 2 transect lines in th e prairie habitat..................37 2-3 Histograms of the frequency of detecti ons of artificial frogs by each team at perpendicular distances in cm along the 2 transect lines in the pineland habitat..............38 2-4 Histograms of the frequency of detecti ons of artificial frogs at perpendicular distances along the transect lines for teams in the prairie habitat and pineland habitat....39 2-5 Abundance estimates with standard error fo r each team from pr airie (A) and pineland habitat (B) using the double observer and distance methods.............................................40 3-1 An aerial photograph depicting off-road vehi cle damage in marl prairie habitat in Big Cypress National Preserve.................................................................................................61 3-2 Map of amphibian occupancy sampli ng locations within BCNP during 2002-2003........62 4-1 Numbers of individuals of green treefrogs and squi rrel treefrogs captured and released in each of th e three toe removal groups during the first 7 sampling occasions...................................................................................................................... ......75 4-2 Apparent survival and 95% confidence interval of green treefrogs and squirrel treefrogs in each toe removal group cate gory across the first 6 monthly sampling intervals...................................................................................................................... ........76 5-1 Mean water depth across the 3 sampli ng locations in BCNP in the cypress strand, broadleaf marsh, and prairie habi tats from April 2004-August 2005................................89 5-2 Number of captures of Green Treefrogs during each of the 25 samples from April 2004 to August 2005 and mean water depth (in cm ) of marsh plots ac ross all 3 sites......90 5-3 Mean snout-to-urostyle (SUL) in mm of Gr een Treefrogs captured in all habitats at the three sites at each sampling occasion in BCNP...........................................................91 5-4 Number of captures by sample of Green Treefrogs in cypress, marsh, and prairie habitats at the th ree sites in BCNP.....................................................................................92 5-5 Estimates with 95% confidence intervals of apparent survival of Green Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP..........................93

PAGE 10

10 5-6 Estimates with 95% confidence intervals of seniority of Green Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP...................................94 5-7 Estimates with 95% confidence intervals of capture probability of Green Treefrogs at each sampling occasion for cypress, ma rsh, and prairie habitat in BCNP........................95 5-8 Derived estimates of population grow th of Green Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP..................................................96

PAGE 11

11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy USE OF AMPHIBIANS AS EC OSYSTEM INDICATOR SPECIES By James Hardin Waddle December 2006 Chair: H. Franklin Percival Cochair: Frank J. Mazzotti Major Department: Wildlife Ecology and Conservation Amphibians are generally considered suitable as indicator species in a variety of systems. Their biphasic life cycle and semi-permeable skin are two justifications often given for this use of amphibians. In this dissertation, the use of amphibians as indicato r species in support of management and restoration in the Everglades of southern Florida was investigated. Methods for monitoring amphibians and specific uses of amph ibians as indicator species were evaluated. Techniques to reduce observer bias in visu al encounter surveys, a common method of sampling amphibians for monitoring purposes, we re tested. Both doubl e observer and distance based methods were shown to have significan t bias in enumerating a known population of artificial frogs. Toe-clipping, a standard method for individually marking frogs was also studied on two treefrog species in south Florida. Toe-cl ipping was found to have a slight negative effect on survival in one species, but not the other. These studies demonstr ate the importance of carefully choosing and evaluating monitoring me thods to appropriately address questions concerning amphibian populations. The occupancy of four anuran species was esti mated in relation to off-road vehicle (ORV) use in Big Cypress National Preserve to determine if amphibians are useful as indicators of this form of anthropogenic disturbance. Results confirmed that ORV us e was a significant factor in

PAGE 12

12 the site occupancy of the four species of gr ound-dwelling anurans studied. In another study, the survival and recruitment of Green Treefrogs we re estimated in relati on to hydrology and habitat to better understand how frogs might res pond to hydrologic cha nges proposed under the Comprehensive Everglades Rest oration Plan. Water depth a nd hydrologic season were both important factors in survival a nd recruitment, and population grow th rates varied with seasons. This research concludes that amphibians meet the criteria for ecosystem indicator species in south Florida. They are abundant, may be effi ciently sampled, and have been demonstrated to respond in a predictable way to stre sses to the system. Monitoring of amphibians is a useful tool for determining the success of ecosystem restora tion and management in the Everglades of south Florida.

PAGE 13

13 CHAPTER 1 INTRODUCTION Ecological indicators can have many purposes, including being used to assess the condition of the environment or monitor trends in conditi on over time (Cairns et al. 1993, Dale and Beyeler 2001). Some species suitable for monitoring tren ds in condition over time may be useful as indicators of restoration success in ecosystems in which restoration activities are occurring. Amphibians are widely considered to be useful as indicator species (Welsh and Ollivier 1998, Sheridan and Olson 2003), but lit tle direct evidence has been gathered that evaluates the usefulness of amphibians for this role. Ther e are several reasons why amphibians may be excellent indicators, but there are also limitations to their use. In this chapter, I discuss the characteristics of good indicators and whether am phibians display these characteristics. In addition, I will outline important considerations for evaluating amphibians as indicators for any particular system. Finally, I will introduce my re search in the Everglades of southern Florida and outline the rest of this dissertation. Characteristics of Indicator Species Dale and Beyeler (2001) discuss several general characteristics of useful indicator species. Indicators should be easily sample d, sensitive to stresses on the system, and respond to stress in a predictable manner. These res ponses should be anticipatory of an impending change in the whole system, and they should predict changes th at can be averted by management. Indicators should provide information regarding changes to th e whole system rather than a few habitats or locations, and have a known respons e to anthropogenic stresses a nd natural disturbances, and the response should have low variability (Dale and Be yeler 2001). Indicators for restoration success need to be predictable enough to determine wh ether they are responding to changes due to

PAGE 14

14 management actions or just natural fluctuations in the system. Finall y, indicator species should also be abundant or cost-effective to sample. Using vertebrates as indicators of habitat qual ity requires special consideration (Landres et al. 1988). Using abundance of vertebrate species requires robust estimation techniques that explicitly deal with im perfect detection (MacKenzie et al. 2002, Williams et al. 2002). Also, the efficacy of the indicator species as an inde x for the abundance of other species must be determined. Using one or more species as indicato rs of habitat quality fo r other species is valid only after research validating this approach has been conducted (L andres et al. 1988). Managing for one indicator may ignore ecological processes not important to the indicator but vital to other species (Kushlan 1979). Suitability of Amphibians as Indicator Species Amphibian species or communities have been touted as useful indicators in many situations recently (Welsh a nd Ollivier 1998, Galatowitsch et al. 1999, Collins and Storfer 2003, Sheridan and Olson 2003, Hammer et al. 2004). So me studies use amphibians as indicators of environmental contamination or pollution (Hamme r et al. 2004). Others attempt to use the species assemblage (Sheridan and Olson 2003) or the abundance of popul ations (Welsh and Ollivier 1998, Campbell et al. 2005) as indicator s of ecosystem health or habitat quality. Amphibians have several character istics that make them useful as indicator species. They are often locally abundant (Rocha et al. 2001, Wa tanabe et al. 2005) and may be sampled with low-cost standard methods (Heyer et al. 1994, Pierce and Gutzwiller 2004). Because of their permeable skin and biphasic life cycle amphibian s are likely sensitive to environmental stress (Vitt et al. 1990, Wake 1991, Blaustein 1994, Blaustei n et al. 1994), but there is some debate about whether this sensitivity is consistent and predictable (Pec hmann and Wilbur 1994). It is

PAGE 15

15 imperative that any study using amphibians as indicators of ecosystem stress demonstrates a direct causal link between the stress a nd its effect on the indicator species. More research is also needed to determine if the response of amphibians to a particular stress is indicative of the mana gement action taken. In some situations amphibians may serve as canaries (Blaustein 1994), but not necessarily in all cases (Pechmann and Wilbur 1994). The responsiveness of amphibians as indicator speci es will depend on the type of stress and the particular amphibians in the system. It is likely that amphibians will be good indi cators of changes to the whole ecosystem because they are sensitive to changes in the a quatic and terrestrial environments. The aquatic environment is required for reproduction in most species (Duellman and Trueb 1986) and the permeable skin of amphibians makes them sensitive to water quality and UV radiation in the egg and larval as well as adult life stages (Gerlanc and Kauf man 2005, Taylor et al. 2005). Many amphibian species spend much of their life in te rrestrial environments for activities like feeding and dispersal. Amphibians should also respond to changes in the terrest rial environment that would affect water relations th rough their integument with beha vioral responses (e.g., shifting activity periods or moving to different microha bitats) or less freque ntly with phenotypic responses (e.g., facultative lipid ba rrier adjustment; Lillywhite 2006). Outline of the Dissertation The Everglades ecosystem of southern Florida has been substantially altered over the last 100 years by loss to agriculture and urbanization (South Florid a Water Management District [SFWMD] 1992, Ogden et al. 2005 ). Compartmentalization of the remaining system has impeded historic flow patterns and altered the temporal and spatial dyna mics of hydrology in the Everglades (Davis et al. 1994). A large-scale restoration effort the Comprehensive Everglades Restoration Plan (CERP), was de vised to attempt to restore natural hydrologic regimes to the

PAGE 16

16 remaining Everglades (DeAngelis et al. 1998). Managers charged w ith decision making for CERP need species that can serv e as indicators of ecosystem re storation success. A necessary condition before a species may be used as an indi cator in a system is a model of how the species will respond to changes in the system (Kreme n 1992, Dale and Beyeler 2001). Some species (e.g., wading birds; Crozier and Gawlik 2003) ha ve been monitored for many decades; therefore data exist on pre-a lteration conditions. Other spec ies (e.g., periphyton; McCormick and Stevenson 1998) have been manipulated in e xperiments to better understand their expected response to changes that will be im posed during Everglades restoration. My research for this dissertation focuses on using amphibians as indicator species in support of management of the > 1,000,000 hectares under restoration in CERP. Although some higher trophic level species are monitored and modeled to evaluate restoration scenarios currently (DeAngelis et al. 1998), many of these species do not meet Dale and Beyelers (2001) criteria for good indicato r species. For instance, the Cape Sa ble Seaside Sparrow and the Florida Panther are rare species of conservation inte rest, but these species may not respond in a predictable way to changes to the system, a nd it is unproven that their responses will be anticipatory of effects on ot her components of the system. There is also a need for indicator species that can be used to identify the effects of certain human activities other than CERP on these natural areas. The use of off-road vehicles (ORV) is an important management concern in Big Cypr ess National Preserve ( BCNP) (NPS 2000). The extent to which ORVs impact wildlife in Big Cypre ss is of great interest to park managers. My objective was to research the efficacy of amphibian s as indicator species in southern Florida to support information needs for both CERP and for the management of ORVs in BCNP.

PAGE 17

17 Chapter 2 describes a study designed to evalua te bias when using the visual encounter survey, a standard sampling technique for rept iles and amphibians. Two estimation methods designed to account for incomplete detectabilit y were applied to c ount data collected by observers on a known population of artificial frog models. This study underscores the need for accounting for incomplete detectability in amphibian monitoring programs and the potential for bias in count data. Accurate sampling techni ques with reliable methods for determining the precision of estimates is critical when us ing amphibians as indicator species. Chapter 3 describes the use of site occupanc y estimation (MacKenzie et al. 2002) to model the effects of ORV use on the distribution of f our anuran species in BCNP. Site occupancy modeling determined that the index of ORV use cr eated for this study was an important factor in the occupancy of ground-dwelling frog species. Only one species was positively associated with the ORV index, while the other three were less li kely to occupy sites with higher ORV use. A monitoring program designed to use these amphi bians as indicators would be useful for evaluating the continuing impact of ORVs in Big Cypress. Chapters 4 and 5 describe the use of hylid treef rogs as indicators of restoration success in the Everglades. Chapter 4 examines the effect s of toe-clipping as a marking technique for individual treefrogs. Uniquely marking individu als is necessary for estimation of survival and movement rates, but the marking method must be validated. Chapter 5 describes a study to examine survival and recruitment rates of green treefrogs in relation to hydrology and habitat. This information is vital for building a model of how these potential indicator species will respond to the hydrologic changes of CERP. Finally Chapter 6 discusses the overall useful ness of amphibians as ecosystem indicator species in southern Florida. I will make recommendations for monitoring and analysis

PAGE 18

18 techniques of amphibian populations that are appropriate at partic ular scales. I will summarize the extent to which amphibians are useful as indicator species in the Everglades ecosystem.

PAGE 19

19 CHAPTER 2 SAMPLING METHODS FOR AMPHIBIAN MONITORING Introduction The visual encounter survey (VES) is widely used as a sampling method for reptiles and amphibians (Crump and Scott 1994, Doan 2003). In this method, one or more observers search a defined area for animals for a specified amount of time. Usually the number of individuals of a species counted is standardized by time or area searched (i.e., effort) to determine the relative abundance of the species. Relative abundance among sites may only be compared under the restrictive assumptions of an index to actual abundance. Th e primary assumptions are that individuals of a species have a constant probability of detec tion across time and space (e.g., different seasons or habitats) a nd that different observers all have the same probability of detecting species of interest (Crump and Scott 1994, Williams et al. 2002). These assumptions may be violated with improper st udy design or by uncontrollable f actors such as weather. Some authors have demonstrated that the assumptions underlying the use of VES data as a measure of relative abundance of am phibians and reptiles are unlikely to hold in actual sampling (Henke 1998, Rodda et al. 2005). The major obstacle to using this index approach is heterogeneity in detection probability (p) of individuals of a given species and am ong different observers (Pollock and Kendall 1987, Williams et al. 2002). It is possible that p may change in relation to sampling conditions, individual behavior, or across hab itat types. Further, different observers may be more or less adept at finding a species and may therefore reco rd different proportions of individuals (i.e., perception bias; Marsh and Sinclair 1989). Sa mpling methods based on distance sampling have been developed to deal with heterogeneity in p across space and time to produce robust estimates of density (Burnham et al. 1980, Buckland et al. 2001). Other methods involving multiple

PAGE 20

20 observers have been developed to account for observer bias (Cook and Jacobson 1979, Pollock and Kendall 1987, Marsh and Sinclair 1989, Nichols et al. 2000), and re cent advances in distance sampling of line transects have combined info rmation from multiple observers to improve density estimates where detection at the line is not always perfect (Buckland et al. 2004). I evaluated the use of double observer and distance sampling approaches using a population of artificial frogs with known abundance. This techni que lacks the realism of actual frogs, but provides a better evaluation of the pot ential bias associated w ith distance sampling and double observer sampling separately and together. The objective of this study was to evaluate sampling approaches that incorporate estimates of detection probability over line transects sampled similarly to the standard VES. I hypot hesized that both met hods would yield similar results and that the 95% confidence intervals of the estimates from both methods would include the true abundance of frogs. Strengths and weaknesses of the samp ling methods will be discussed. Methods Sampling Experiment To test the efficacy of the double observer and distance methods, a pair of 50-m transects in 2 habitats (4 transects total) were establ ished in Big Cypress National Preserve, Collier County, Florida, USA, on 14 June 2005. Transects were arranged as 2 para llel lines spaced 30 m apart. Each transect was in a continuous tract of either prairi e or pineland habitat. Prairie consists of short hydroperiod wetl ands that lack an overstory and woody vege tation in general. Prairies are dominated by sedges, usually up to 1 m in height (D uever et al. 1986). Pinelands are forested habitats that form on slightly higher elevation sites in BCNP dominated by Slash Pine ( Pinus elliottii ). Pinelands tend to have a dense unde rstory of woody plan ts, especially Wax Myrtle ( Myrica cerifica ) as well as grasses and sedges (Due ver et al. 1986). The centerline of

PAGE 21

21 each transect was clearly marked with string atta ched to polyvinyl chloride (PVC) poles every 10 m. The string on the centerline was stretched and tied to insure that it remained taut during the entire sampling experiment. Artificial frogs consisted of small (50 mm total length) plastic frog models (Daytona Vending, www.freightcloseouts.com ). The artificial frogs were painted with flat black spray paint on the underside and glossy green spray pa int on the upper surface to conceal the bright color of the plastic and to reasonably repres ent the appearance of the Green Treefrog ( Hyla cinerea) an abundant hylid in the region (Meshaka et al. 2000). Frog models, 50 per transect, were located randomly within 2 m of either side of each line. The maximum vertical height of frog placement was selected as a random number between 0 and 150 cm, and actual height was the highest point on which the frog could be secu rely fastened in an upright position but not greater than the maximum. Artif icial frogs were attached to vegetation using a single loop of clear monofilament fishing line. All frogs were recovered after the experiment and determined to be in the same location as their original placement. The sampling scheme was based on a combin ation of the dependent double-observer approach of Cook and Jacobson (1979) and the distance sampling approach of Buckland et al. (2001). To sample transects, vo lunteers were organized into teams of 2 observers. Observers were told to designate one member of the gr oup as the primary observer and another as the secondary observer for the first transect, and then to switch roles for the second transect. First, the primary observer was instructed to walk alon g the transect and indicate all frogs observed. Next, the secondary observer was to ld to record the side of and perpendicular distance from the transect centerline to each frog detected by the primary observer. Finally, the secondary observer was instructed to search for frogs not observed by the primary observer (i.e., after the

PAGE 22

22 primary observer passed their location). Th e secondary observer then recorded their perpendicular distance and noted that they were missed by the primary observer. The teams of observers worked at night using headlamps in similar conditions to a standard VES survey. Seven teams sampled the prairie habitat and 6 team s sampled the pineland habitat. All of the observers were trained biologists, but none had searched fo r artificial frogs prior to this experiment. None of the observers were involved in placing the artifici al frogs on transects. Double Observer Analysis The double observer data were analyzed usi ng the model of Cook a nd Jacobson (1979). This model is based on data collected in which 2 individuals alternate ro les as the primary and secondary observer. Three models of detection probability ( p ) were analyzed in program SURVIV (White 1992): individual detec tion probabilities for each observer ( pobs*hab), a single detection probability for each habitat, respectively ( phab), and a constant detection probability among observers and habitats ( p..) Model selection was co nducted using Burnham and Andersons (1998) information-theoretic appr oach based on the quasi-likelihood Akaikes Information Criterion adjusted for small sample sizes (QAICc). Estimates of ip were obtained for individual observe rs using the methods of Cook and Jacobson (1979) as implemented by Nichols et al. (2000) in Program SURVIV (White 1992) using the equations: 21 22 22 11 21 12 22 11 1 x x x x x x x x p and 12 11 22 11 21 12 22 11 2 x x x x x x x x p where x11 is the number of individuals detected by observer 1 in the role of primary observer, x22 is the number detected by observer 2 as primary, and x12 and x21 are the number of individuals observed by observer 1 as the secondary and obser ver 2 as the secondary, respectively. The

PAGE 23

23 overall estimated p of each team was determined following the methods of Nichols et al. (2000) using the equation: 11 22 21 121 x x x x p and the estimate of the total abundance (N ) of frogs from each sample was produced using the equation: p x N .. where x.. is the sum of the counts of both observers. The estimated variance for this estimate is given by Nichols et al. (2000): 2 4 2 ) 1 ..)( ( ) r( a v ..) ( ) r( a v p p x p p x N Distance Analysis The perpendicular distance data were anal yzed in Program DIST ANCE (Thomas et al. 2003) to estimate the density and abundance of artificial frogs. Density was estimated independently for each team. To increase precis ion of the estimate, the detection function was produced from distances pooled across teams but w ithin each habitat stratum. Observers were told to only look for frogs within 2 m of the tran sect centerline, and all frogs were placed within 2 m of the centerline. Therefore, the data were truncated to 2 m in the field. Truncation at 1 m and at 1.5 m was also explored using data filters in DISTANCE. Results The randomly chosen locations of the artifici al frogs were evenly distributed around the transect center line (Figure 2-1). The 7 team s in the prairie habitat found 41-68 of the 100 frogs, and the 6 teams in the pine habitat found 1843 of the 100 frogs present along both transects

PAGE 24

24 (Table 2-1). Team 5 in the prairie habitat and Team 10 in the pine habitat had an observer who did not find any frogs in the role of secondary observer. Using locat ion information noted by observers it is believed that a total of 94 of th e possible 100 artificial frogs in prairie habitat and 67 of the 100 artificial frogs in pineland ha bitat were observed by at least one team. Actual detection rates calcu lated within 30 cm, within 50 cm, beyond 50 cm, and overall varied by team and were not consis tently higher closer to the tr ansect centerline (Table 2-2). Several teams saw fewer frogs near the centerline than further away, most notably teams 2, 4, and 6 in the prairie habitat (Figure 2-2) and team s 9 and 10 in the pineland habitat (Figure 2-3). Double Observer Model selection in Program SURVIV indicated that model pobs (with individual observer detection probabilities) was better than habitat le vel or constant detection probability models as model pobs received all of the QAICc weight (Table 2-3). Indi vidual detection probabilities varied widely among observers ranging from 0.40 -1 .0 in prairie habitat and from 0.01 1.0 in pineland habitat (Table 2-4). Estimates of total frog abundance along transe cts in both habitats were lower than the actual abundance of 100 per habi tat (Table 2-5). Abundance estimates ranged from 43 on prairie transects and from 20-49 on pineland transects. Distance Analysis Model selection in Program DI STANCE favored models with habitat-specific detection functions. Models with no post hoc truncation were suitable for the analysis as data were truncated in the field, and addi tional truncation did not improve th e models. The hazard-rate key function with cosine adjustment was chosen as the best model for the detection function (AIC weight = 0.73; Table 2-6). The second best model used the global detection rate with the hazard rate key function (delta AIC= 1. 99, AIC weight= 0.27). Models based on sample (team) level

PAGE 25

25 detection functions did not improve the fit of the model and receive d no AIC weight (Table 2-6). The stratum-level detection functions for the prai rie and pineland habitat transects have a flat shoulder to about 175 and 160 cm, respectivel y, from the centerline (Figure 2-4) Abundance estimates based on habitat stratum-le vel detection functions for each individual team ranged from 46-76 in the prairie and 22-51 in the pineland habitats (Table 2-7). These abundance estimates from distance analysis we re similar to those from the double observer analysis, but precision was much lower for the distance method (Figure 2-5). Discussion Despite the use of data collection protocols and analytical techniques designed to account for heterogeneity in detection pr obabilities, there was a large amount of bias in estimates of artificial frog abundance in the 2 habitats. Th e true abundance of 100 fr ogs was not within the 95% confidence intervals of any of the abunda nce estimates from either the double observer method (Table 2-5) or the distance method (Table 2-7) in either habitat. Additionally, counts and estimates varied widely among teams, even though the transects sampled, frog placements, and environmental conditions were identical. Fu rthermore, habitat differences between prairie and pineland led to large differences in estimate s of abundance and detecti on rates of artificial frogs. Some apparent bias in both sampling met hods may be explained by unobservable frogs. Only 67 of the artificial frogs were observed by any team in the pineland and only 94 were observed in the prairie. If we consider frogs not observed by any observer as unobservable, than we would only expect estimates to be sim ilar to the number of observable frogs. However, only the abundance estimate of team 7 in prairie habitat and team 12 in pineland habitat using the distance sampling method included the observa ble number of frogs in the 95% confidence interval.

PAGE 26

26 These results support concerns that the use of counts from VES sampling unadjusted for detection probability is inappropriate (Cru mp and Scott 1994, Schmidt 2003). Even under identical conditions in this surv ey, different observer teams produced dissimilar counts. Some of the difference in counts among teams may be attri butable to observer sk ill, although no observer had previous experience searching for the artifici al frogs used in this experiment. Failure to account for observer bias (Pollock and Kendall 1987) in VES samples could lead to gross errors as was demonstrated in this study. This is es pecially important to st udies seeking to identify long-term trends in amphibian populations using data collected over many seasons by different observers. Although the dependent double-observer approa ch of Cook and Jacobson (1979) is designed to estimate individual de tection probabilities and therefore eliminate some of the bias associated with observers, there are limitations to this approach. Nichols et al. (2000) hypothesize that detection probability estimates may be biased high using this approach, which would lead to abundance estimates that are bias ed low. This study provides empirical evidence that detection probabilit y estimates are indeed biased high under the assumptions of the Cook and Jacobson (1979) model. If both observers are not evenly matched in skill, the detection estimate of the better observer will be biased high. If the poorer observer does not find any objects missed by the better observer in the role of the primary observer, estimated detection could reach 100%. Likewise, if both observers are poor and de tect few missed objects as secondary observers, detection estimates for the t eam will be biased high. It seems likely that some examples of both of the above scenar ios took place in this study (Table 2-4). Distance analysis is consider ed useful for estimating dens ity and abundance of objects, especially when detection rates decline with distance from the observer (Buckland et al. 2001).

PAGE 27

27 In this study, however, there was little noticeable d ecline in detection rates at distances up to 2 m from the transect centerline (Table 2-2). Detect ion rates were estimated to be as high as 1.0 out to more than 1.5 m from the cen terline (Figure 2-4), leading to estimates of abundance that were biased low as with the double observer method. Rigorous testing for bias in animal sampling pr otocols is not always performed, but can be informative. Anderson et al. (2001) found that m odifications to desert to rtoise line transect sampling protocols were necessary af ter testing the technique with artificial tortoises. Nichols et al. (1986) found evidence of inter-observer vari ation in searches for white-winged dove nests marked by individual observers and recommende d future estimation of observer-specific detection rates. Rodda et al. (2005) evalua ted relative abundance estimates based on visual searches against absolute a bundance estimates based on remo val techniques and found poor correspondence between the two. They concluded that visual s earches alone were only suitable for 1 of the 6 reptile species they surveyed. This study provides evidence that observer differe nces in detection rate s are very important and even sampling methods designed to be robust to these differences may be inadequate to describe the absolute abundance of animals. Cl early, we have an advant age in this study of knowing the true abundance of objects. Such info rmation is not available in actual animal populations. However, we cannot assume that th e artificial frogs in this study were perfect surrogates for actual frogs. The fact that they are immobile probably decreased their detectability, as many observers find frogs by de tecting their movements. Also, many observers key detection on eyeshines not present in the surr ogates. Therefore, we cannot use this study to calibrate the amount of bias in VES sampling, but it does help illustrate an important problem.

PAGE 28

28 Researchers gathering VES data should be awar e that there is potenti al for great bias in abundance estimates using this method.

PAGE 29

29 Table 2-1. Number of artificial frogs (out of a possible 100) observ ed by each team in prairie and pineland habitats in BCNP. x11 is the count of objects detected by observer 1 when observer 1 was the primary observer, x22 is the count of objects by observer 2 when observer 2 was primary, and x12 and x21 are the counts of observers 1 and 2 in the roles of secondary observer, respectively. Habitat Team x11 x22 x12 x21 Total Prairie Team 1 28 148656 Team 2 25 169151 Team 3 21 153443 Team 4 20 2211255 Team 5 22 138043 Team 6 13 1341141 Team 7 24 3210268 Pineland Team 8 9 15419 Team 9 9 41418 Team 10 11 100425 Team 11 17 76232 Team 12 19 1012243 Team 13 11 73324

PAGE 30

30 Table 2-2. Actual detection rates of a known popul ation of artificial frogs within 30 cm, within 50 cm, beyond 50 cm, and overall for teams s earching in prairie and pineland habitat of BCNP. Habitat Team p <30 cm p <50 cm p >50 cm Overall Prairie Team 1 0.611 0.6000.5470.56 Team 2 0.500 0.5200.5070.51 Team 3 0.333 0.3600.4530.43 Team 4 0.389 0.5600.5470.55 Team 5 0.500 0.6000.3730.43 Team 6 0.444 0.3600.4270.41 Team 7 0.444 0.7200.6670.68 Pineland Team 8 0.357 0.2690.1620.19 Team 9 0.071 0.0770.2160.18 Team 10 0.143 0.2690.2430.25 Team 11 0.500 0.3850.2970.32 Team 12 0.643 0.5770.3780.43 Team 13 0.429 0.3850.1890.24

PAGE 31

31 Table 2-3. Model selection for the three doubleobserver models analyzed in Program SURVIV with the model log-likelihood, number of pa rameters (K), quasilikelihood Akaikes information criterion adjusted for small sample sizes (QAICc), difference between each QAICc and the minimum QAICc (Delta QAICc) and the QAICc weight. Model Log Likelihood K QAICc Delta QAICc QAICc Weight pobs*hab -35.32 26 125.50701.0000 Phab -75.76 2 153.07427.5670.0000 p. -74.53 1 153.52428.0170.0000

PAGE 32

32 Table 2-4. Estimates from program SURVIV of individual detec tion probabilities (ip ) of artificial frogs for each obser ver in prairie and pineland ha bitat with standard error (S.E.) and 95% confidence intervals (C.I .) taken from the double observer method.. Habitat Observer ip S.E. Lower 95% C.I. Upper 95% C.I. Prairie 1 0.7230 0.11190.50360.9424 2 0.5590 0.12930.30550.8125 3 0.9389 0.06010.82111.0567 4 0.6251 0.10100.42710.8231 5 0.8082 0.09020.63130.9850 6 0.8017 0.10600.59401.0094 7 0.6079 0.09120.42920.7866 8 0.9304 0.06870.79571.0651 9 1.0000 0.00020.99961.0004 10 0.6197 0.10590.41200.8273 11 0.4019 0.15540.09740.7064 12 0.5667 0.20620.16250.9709 13 0.8977 0.06960.76141.0341 14 0.7415 0.07280.59880.8842 Pineland 15 0.0123 0.8123-1.57981.6044 16 0.0044 0.2948-0.57350.5823 17 0.6156 0.18140.26010.9712 18 0.7122 0.26350.19581.2286 19 0.7330 0.11420.50920.9569 20 1.0000 0.00010.99991.0001 21 0.8046 0.14000.53021.0789 22 0.4845 0.15980.17140.7977 23 0.7912 0.14880.49961.0829 24 0.3979 0.12480.15330.6425 25 0.6931 0.16940.36121.0251 26 0.6168 0.19230.24000.9936

PAGE 33

33 Table 2-5. Abundance estimates (N ) of artificial treefrogs for each team in prairie and pineland habitat with standard error (S.E.) and upper and lower 95% confidence intervals (C. I.) based on Chaoss estimator (Nichols et al. 2000) taken from the double observer data.. Chao 95% C.I. Habitat Team N S.E. N Lower. Upper Prairie Team 1 62.3 3.0158.5971.34 Team 2 56.7 2.8553.2965.39 Team 3 47.8 2.5744.8255.86 Team 4 61.2 2.9857.5370.15 Team 5 47.8 2.5744.8255.86 Team 6 45.6 2.5042.7153.47 Team 7 75.7 3.4071.3385.59 Pineland Team 8 21.1 1.6219.5726.99 Team 9 20.0 1.5718.5325.78 Team 10 27.8 1.8825.8534.27 Team 11 35.6 2.1633.2142.69 Team 12 47.8 2.5744.8255.86 Team 13 26.7 1.8424.8133.06

PAGE 34

34 Table 2-6. Model selection for detection func tions in Program DISTANCE showing number of parameters (K), Akaikes Information Cr iterion (AIC), the difference between each AIC and the minimum (Delta AIC), and the AIC Weight of each model. Detection was modeled at the sample, stratum, or gl obal level with the h azard rate key function or the half normal key function. Model Name K AIC Delta AIC AIC Weight Stratum detection; Hazard Rate Key 45374.3800.7300 Global detection, Hazard Rate Key 25376.371.990.2697 Global detection, Half-normal Key 55390.4916.100.0002 Stratum detection; Half-normal Key 65404.7030.320.0000 Sample detection; Hazard Rate Key 265413.0138.630.0000 Sample detection; Half-normal Key 135436.2861.900.0000

PAGE 35

35 Table 2-7. Abundance (N ) estimates of artificial frogs by team for transects in the prairie and pineland habitats with coefficient of varia tion (CV) and 95% confidence interval (CI) from the distance sampling approach. Habitat Team N CV Lower 95% C.I. Upper 95% C.I. Prairie Team 1 49 15.163766 Team 2 57 14.094375 Team 3 48 15.333665 Team 4 62 13.584781 Team 5 48 15.333665 Team 6 46 15.703463 Team 7 76 12.236097 Pineland Team 8 23 23.081436 Team 9 22 23.701434 Team 10 30 20.162044 Team 11 38 17.852754 Team 12 51 15.453870 Team 13 29 20.561943

PAGE 36

36 Figure 2-1. Histograms of the fre quency of perpendicular distances in cm of actual artificial frog placements along the 2 transect lines in the prairie and pineland habitat.

PAGE 37

37 Figure 2-2. Histograms of the frequency of dete ctions of artificial fr ogs by each team at perpendicular distances in cm along the 2 tr ansect lines in the prairie habitat at BCNP.

PAGE 38

38 Figure 2-3. Histograms of the frequency of dete ctions of artificial fr ogs by each team at perpendicular distances in cm along the 2 tr ansect lines in the pineland habitat of BCNP.

PAGE 39

39 Figure 2-4. Histograms of the frequency of dete ctions of artificial fr ogs at perpendicular distances in cm along the 2 tr ansect lines for all teams in the prairie habitat (A) and the pineland habitat (B) with the best mode l of the detection f unction (red lines) from Program DISTANCE superimposed. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0 50 100 150 200 250Perpendicular distance in centimeters 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 0 50 100 150 200 250Perpendicular distance in centimeters A B

PAGE 40

40 Figure 2-5. Abundance estimates with standard erro r for each team from prairie (A) and pineland habitat (B) using the double observer and distance methods. True abundance was 100 in both habitats.

PAGE 41

41 CHAPTER 3 USING SITE OCCUPANCY MO DELING TO DETERMINE THE EFFECT OF OFF-ROAD VEHICLE USE ON GROU ND-DWELLING ANURANS Introduction Off-road vehicles (ORV) can impact wildlife species directly by causing physical harm (Steiner and Leatherman 1981) or indirectly by altering behavior or disturbing habitats (Brattstrom and Bondello 1995, Guye r et al. 1996). ORVs create noise which may be disruptive to wildlife, or the presence of large moving vehicles may be dist urbing. In addi tion, ORVs may create trails and damage to the vegetation in ar eas that receive heavy us e. One of the major management concerns of Big Cypress National Pres erve (BCNP) in southern Florida, USA is the regulation of ORV use (NPS 2000, Duever 2005) where mapped ORV trails total over 47,900 km in length (Welch et al. 1999). Janis and Clark (2002) found evidence that ORV use altered the behavior of the endangered Florida panther in Big Cypress, and Duever et al. (1981) demonstrated that ORVs alter vegetation compos ition and hydrology at imp acted sites (Figure 31). It is unclear how other spec ies of wildlife, especially amphibi ans, are affected by ORV use in BCNP. Determining the impacts of ORVs on local populations may be possible through markrecapture sampling or some other technique that provides estimates of abundance. However, at larger spatial scales (i.e. across landscapes) it b ecomes increasingly futile to attempt to estimate the abundance or density of amphibians. It is difficult to enumerate such large populations, and population sizes may fluctuate with season and environmental conditions (Green 2003). A relatively new method, site occ upancy or proportion area occupied, allows collection of simple presence/absence data across the entire landscape to make inference regarding species status. The site occupancy rate of species across a landsca pe is more meaningful at large scales since a larger proportion of the area can be sampled than in traditional mark-recapture. Also unlike

PAGE 42

42 simple counts and estimates of relative density (catch per unit effort), occupancy accounts for detection or non-detection of the species after re peated samplings of many sites (MacKenzie et al. 2002). Covariates (e.g. habitat type, ORV use) may be used to improve estimates of occupancy probabilities (MacKenzie et al. 2002), and model selection may also be used to make inferences about the effects of covariates (M acKenzie et al. 2005, Schmidt and Pellet 2005). Ground dwelling anurans are one group of amphibia ns that may be especially affected by ORV use. These species often occur far from pe rmanent water, and most breed in small, fishfree ponds (Duellman and Trueb 1986). They tend to have a low dispersal capability (Blaustein et al. 1994, Alford and Richards 19 99), so they are likely to spend most of their life within the same small area. For these reasons I expect thes e species to be especially affected by ORV use in Big Cypress. ORV use may alter microhabitats at sites by decreasing vegetation or increasing drainage. These alterations may make some sites less suitable for ground dwelling anurans. The objective of this study was to determine if ORV use was an important factor influencing the site occupancy of four grounddwelling anuran species in BCNP: Oak Toad ( Bufo quercicus ), Southern Toad ( Bufo terrestris ), Eastern Narrow-mouthed Toad ( Gastrophryne carolinensis ), and Greenhouse Frog ( Eleutherodactylus planirostris ). Detection data of amphibians from a random sample of sites in BCNP were modeled with site covariates including an index of ORV use created from geographic in formation system (GIS) data of ORV trails using the site occupancy method of MacKenzie et al. (2002). I hypothesized that occupancy of these four ground-dwelling anuran species would be negatively a ssociated with ORV use. Methods Study Area Big Cypress National Preserve is a 295,000 ha natural area managed by the National Park Service in Collier and Monroe Counties of southwestern Florida, USA. BCNP is bordered on

PAGE 43

43 the south and east by Everglades National Park and Water Cons ervation Area 3A (managed by the South Florida Water Management District). Cape Romano/10,000 Islands National Wildlife Refuge and Florida Panther National Wildlife Re fuge, as well as Fakahatchee Strand State Preserve lie on the western boundary of Big Cypress. Because of its large size and geographic location, Big Cypress is not heavily influenced by habitat loss due to development like many other areas in Florida. However, Big Cypress has been used by recreational ORV users since before its establishment in 1974 (NPS 2000). Curre nt regulations limit use to no more than 2000 registered ORVs per year, but lasting signs of 47,900 km of old and current ORV trails are apparent throughout the preserve (Duever et al. 1987, Welch et al. 1999, NPS 2000). Sampling Random sites were chosen throughout Big Cypr ess using the animal movement analysis extension of Hooge and Eichenlaub (1997) in ArcView 3.2 (Environmental Systems Research Institute, Inc., www.esri.com ). A sample of these random points accessible by foot or ORV was chosen for amphibian sampling (Figure 3-2). Thes e sites represented 5 different habitat types: cypress strand, cypress prairie, prairie, hammock, and pineland, based on the vegetation classification scheme of Madden et al. (1999). The number of sampling occasions per site was variable. Some were sampled on a monthly basis from February 2002 to March 2003 to provide a time series of detection data from sites. Othe r sites were sampled just twice during the entire project to increase the geographi c coverage of the sampling. This approach was intended to balance the effort between repeat ed sampling and additional sites. Sampling for amphibians consisted of standard visual encounter survey (VES) techniques (Crump and Scott 1994). All VES samples were in itiated at least 30 minutes after sunset, and each survey was conducted by at least two experi enced observers using 6-volt spotlights with halogen bulbs. VES samples were time and area constrained such that the area within a 20-m

PAGE 44

44 radius of the randomly chosen point (1,256 m2) wa s searched for 1 person-hr. All areas of each plot were visually scanned, but judgment of the observers was used to determine which areas within the plot received the most emphasis. The goal was to find as many individual amphibians as possible. All possible amphibian locatio ns were searched, including trees and other vegetation as well as bare ground and leaf litte r. Each amphibian observed was captured (if possible) and identified to species. A 10-minut e vocalization survey was conducted during each VES sample. All species of frogs and toads hear d vocalizing were noted. All anurans that could be heard were included, even if it was possible or likely that they were calling from a location outside of the 20-m radius plot. In addition to the biological data, environmen tal data were collected in the field during each survey. Air temperature and relative humidity were measured using a digital thermohygrometer. The date and time of the sa mple and whether the plot was inundated with water at the time of the sample also were reco rded. Sampled sites were assigned an ORV use index based on the sum of ORV trails within a 500-m radius circle around the sampling point by using the ORV trail GIS dataset developed by Welc h et al. (1999). The We lch et al. (1999) map includes all trails visible from aer ial photos, even ones still visible in areas that had been closed to ORVs for as much as 2 decades prior to the ph otographs. Eight of the 70 sites sampled in this study fell in areas designated as high use by Welch et al. (1999), and trails were not mapped in these high use areas because of the high density of trails. The ORV index for sites located in high use areas was set equal to the highest value from sites for which trail data were available. In addition to the ORV index, a hydrologic i ndex was created based on number of months inundated for each site using a habitat and hydrol ogic model developed by Duever et al. (1986).

PAGE 45

45 Data Analysis Site occupancy rates and detection probabi lities were modeled in program PRESENCE (MacKenzie et al. 2002, MacKenzi e et al. 2006) using the single season model. This method assumes that sites are closed to changes in occu pancy within the study, an d that detection of a species at a site is independent of detections at other sites. This method also assumes that species are not falsely detected, but species may or may not be detected when present. This method was deemed appropriate for these focal anur an species due to thei r difficulty to detect and low dispersal. Site-specific covariables, those that dire ctly affect the estimate of occupancy ( ) were habitat type, hydrologic index and ORV use index. Values for the indices were standardized so that the means fell between 1 and 0, a necessa ry condition when using the logit link function (MacKenzie et al. 2006). Sampling occasion covari ates that could affect detection probability ( p ) were air temperature, relative humidity, presen ce of standing water, and season of the year. For each species, we considered 80 models that were combinations of the covariates thought to be biologically meaningful (Table 3-1). The be st model was chosen as the one with the lowest value for Akaikes information criterion (AIC), or the most parsimonious model (model with the best fit for the fewest parameters; Burnham and Anderson 1998). The effect of ORV use on species occupancy was determined using model se lection to determine the AIC weight of all models including the ORV use covariate and by ex amining the beta estimates for the ORV use index in the models in cluding that covariate. Results A total of 469 sampling visits to 70 sites we re made from February 2002 to March 2003. The highest number of study sites (31) was in prairie habitat. Between 7 and 12 sites in each of the other habitats were visited (T able 3-2). The four focal anuran species were detected between

PAGE 46

46 13 and 117 times, and nave occupancy rates (propor tion of sites at which a detection occurred) varied from 17 to 52% (Table 3-3). The best model (model with lowest AIC valu e; Burnham and Anderson 1998) for each of the four species included the ORV index as a s ite covariate (Tables 34 3-7). When AIC weights of models including specif ic site covariates were summe d, the ORV index covariate had the most weight for southern toads, greenhouse frogs, and eastern narrow-mouthed toads (Table 3-8). Only oak toads had less weight on models including the ORV index than other covariates. Beta parameters for the ORV index in the best models were negative for all species except southern toads (Table 3-9). Nu merical convergence could not be reached to estimate standard error (S.E.) for the ORV index beta parameter for oak toads, but S.E. estimates were obtained for the other species. The 95% confidence interval s of the beta parameter estimates for the ORV index covariate overlapped 0 for all species but greenhouse frogs. Discussion The results of this study i ndicate that the ORV index and, thus, ORV-use is a strong predictor of whether a site will be occupied by these four species of anurans. Each of these species had the ORV index covariate in the best models for occupancy, and the sum of the model AIC weights was highest for those models includ ing the ORV index except for oak toads. This indicates that for some speci es, occupancy of a site may depend more on ORV use than on hydrology or habitat. Three of the four species of anurans had beta values for ORV index that were negative, indicating negative associations wi th ORV use (Table 3-9). It was predicted that these small, ground-dwelling anurans would be negatively influe nced by the use of ORVs due to ground level disturbance of vegetation and al tered hydrology. One species, however, the Southern Toad, was positively associated with ORV use. Although th is is counter to the original prediction,

PAGE 47

47 morphology and reproductive strategy of this spec ies might explain the difference in response to ORV use. Southern Toads are larger than the other species, and their tadpoles require up to twice as long (2 months compared to 1 mont h) to develop as Oak Toads or Eastern Narrowmouthed Toads (Ashton and Ashton 1988). ORVs can alter the vegetati on and hydropattern of areas resulting in a loss of vege tation and increased ponding in ru ts and artificial depressions (Duever et al. 1981). Southern Toads may take a dvantage of the increased temporal and spatial extent of standing water for breeding purposes. For all of the species other than oak toads, habi tat was not as important in prediction of site occupancy as the ORV index (Table 3-8). This may be in part due to the fact that these 4 anurans are habitat generalists in south Florida and are not closely associated with any particular habitat type (Duellman and Schw artz 1958, Meshaka et al. 2000, Rice et al. 2004). Habitat was a covariate in the best models for both American Toad ( Bufo americanus ) and Spring Peeper ( Pseudacris crucifer ) in a the study of MacKenzie et al. ( 2002) in Maryland wetlands, but habitat was not as important as the prev ious years count in the study of Schmidt and Pellet (2005) of Tree Frog ( Hyla arborea ) and Natterjack Toad ( Bufo calamita ) in Europe. Landscapes in Big Cypress are very heterogeneous; t hus different habitat types are of ten in close proximity to one another (McPherson 1974, Duever et al. 1986) and frogs might easily transition from one habitat to another. Thus, it is more difficult to determ ine differences in habitat-level occupancy at the scale of this study. The hydrologic index used in this study gene rally had higher AIC weights in model sets than habitat, but lower than the ORV index. Oak Toad was the one exception to this pattern, for which the hydrologic index had the most AIC wei ght in the model set. The hydrologic index was based on the number of months each site woul d typically be inundated with water in a year.

PAGE 48

48 Although this hydroperiod is highly correlated with habitat type (Duever et al. 1986), there are some examples of sites being wetter or drier than others with the same habitat type. It is possible that the occupancy was more sensitive to the hydr ological index than habita t, but this index will change between years as local rainfall varies (D eAngelis et al. 1998). Ha bitats are much more fixed and tend to be primarily a resu lt of microtopography (Duever et al. 1986). Amphibians are often used as indicator species (Vitt et al. 1990, Welsh and Ollivier 1998, Galatowitsch et al. 1999, Sheridan and Olson 2003). This study illustrates how amphibians may be indicators of the effects of ORV use in Big Cypress. The ORV trail data provided by Welch et al. (1999) shows all trails and does not differentiate between old, persistent trai ls and currently used ones. Duever et al. (1987) demonstrated that old trails may take many decades to recover. Some of the sites sampled in this study are in ar eas that have been closed to ORV use for more than 10 years, although they sti ll retain many visible ruts and other physical evidence of ORV use. It is important to remember that this st udy only considers evidence of ORV use in the form of the index used, and does not examine the te mporal component of when ORV use occurred. Consequently, this study found evidence of an impact of ORV use on amphibian species occupancy but did not address recove ry of previously impacted areas. Due to the observational nature of this study, it is not possible to determine the mechanisms by which ORV use influences amphibian occupancy. However, this study should help stimulate more research on the topic. Re source management staff at Big Cypress concerned with reducing impacts of ORV use in the preserve should be awar e that there is evidence that ORV use influences the site occupancy of amphibians. These amphibian species may be indicators of ecosystem impacts not previously shown. A monitoring program designed using the same techniques of this study could be used to track changes over time. Stratifying sampling

PAGE 49

49 by historic and current ORV use areas could also help determine how long the impacts of ORV use can be detected in amphibian occupancy. Colonization and extinction of sites with varying levels of ORV use could also be monitore d over time using the open model approach of MacKenzie et al. (2003).

PAGE 50

50 Table 3-1. Combinations of the 3 site covariates and 4 sampling covariates that were used in the occupancy analysis for each species. Each set of site covariates was modeled along with each set of sampling covariates for a total of 80 unique models for each species. Site Covariates Sampling Covariates Constant Constant Habitat Season Hydrologic Index Temperature ORV Index Temperature, Humidity Habitat, ORV Index Temperature, Humidity, Water Habitat, Hydrologic Index Temperature, Humidity, Water, Season ORV, Hydrologic Index Temperature, Water Habitat, ORV, Hydrologic Inde x Temperature, Water, Season Water Water, Season

PAGE 51

51 Table 3-2. Number of sampling sites and to tal number of site visits by habitat. Habitat Number of Sites Number of Visits Cypress Strand 11 89 Cypress Prairie 9 78 Prairie 31 122 Hammock 7 72 Pineland 12 108 Total 70 469

PAGE 52

52 Table 3-3. Number of detections by species, and proportion of sites at which a detection occurred (naive occupancy) during amphibian surveys across all habitat types. Species No. Detections Naive Occupancy Oak Toad 1721.43% Southern Toad 2628.57% Greenhouse Frog 11752.86% Eastern Narrow-mouthed Toad 1317.14%

PAGE 53

53 Table 3-4. Model selection results for the oa k toad (Bufo quercicus), including Akaikes Information Criterion (AIC) and the de lta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight. Model AIC delta AIC AIC weight psi(Hab,ORV,Hydro), p(Temp,Humid,Water,Season) 117.920.00 0.2946 psi(Hab,ORV,Hydro), p(Temp,Water,Season) 118.010.09 0.2817 psi(Hab,Hydro),p(Temp,Humid,Water,Season) 119.922.00 0.1084 psi(Hab,Hydro),p(Temp,Water,Season) 119.982.06 0.1052 psi(Hab,ORV,Hydro),p(Water,Season) 121.543.62 0.0482 psi(Hydro),p(Temp,Humid,Water,Season) 122.174.25 0.0352 psi(Hydro),p(Temp,Water,Season) 122.244.32 0.0340 psi(Hab,ORV,Hydro),p(Season) 122.965.04 0.0237 psi(Hab,Hydro),p(Water,Season) 123.585.66 0.0174 psi(Hydro,ORV),p(Temp,Humid,Water,Season) 123.915.99 0.0147 psi(Hydro,ORV),p(Temp,Water,Season) 123.986.06 0.0142 psi(Hab,Hydro),p(Season) 124.766.84 0.0096 psi(Hydro),p(Water,Season) 125.437.51 0.0069 psi(Hydro,ORV),p(Water,Season) 127.299.37 0.0027 psi(Hydro,ORV),p(Season) 127.829.90 0.0021 psi(Hydro),p(Season) 132.5614.64 0.0002

PAGE 54

54 Table 3-5. Model selection resu lts for the southern toad ( Bufo terrestris ), including Akaikes Information Criterion (AIC) and the de lta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight. Model AIC Delta AIC AIC Weight psi(Hab,ORV),p(Season) 194.320.00 0.1109 psi(.),p(season) 194.500.18 0.1014 psi(ORV),p(Season) 194.990.67 0.0793 psi(.),p(Water,Season) 195.140.82 0.0736 psi(Hab,ORV),p(Water,Season) 195.851.53 0.0516 psi(.),p(Temp,Humid,Water,Season) 195.861.54 0.0514 psi(.),p(Temp,Water,Season) 196.091.77 0.0458 psi(Hydro),p(Season) 196.211.89 0.0431 psi(Hydro),p(Water,Season) 196.382.06 0.0396 psi(ORV),p(Temp,Humid,Water,Season) 196.482.16 0.0377 psi(ORV),p(Temp,Water,Season) 196.682.36 0.0341 psi(Hab,ORV),p(Temp,Humid,Water,Season) 196.772.45 0.0326 psi(Hydro),p(Temp,Humid,Water,Season) 197.162.84 0.0268 psi(Hydro),p(Temp,Water,Season) 197.272.95 0.0254 psi(Hydro,ORV),p(Season) 197.282.96 0.0252 psi(Hydro,ORV),p(Water,Season) 197.333.01 0.0246 psi(Hydro,ORV),p(Temp,Humid,Water,Season) 198.073.75 0.017 psi(Hydro,ORV),p(Temp,Water,Season) 198.173.85 0.0162 psi(.),p(Temp,Water) 198.784.46 0.0119 psi(Hab,ORV),p(Temp,Water,Season) 198.784.46 0.0119 psi(.),p(Water) 198.994.67 0.0107 psi(ORV),p(Temp,Water) 199.455.13 0.0085 psi(ORV),p(Water) 199.515.19 0.0083 psi(.),p(Temp,Humid,Water) 199.565.24 0.0081 psi(Hydro),p(Temp,Water) 199.685.36 0.0076 psi(Hab),p(Season) 199.755.43 0.0073 psi(Hab,ORV),p(Water) 199.965.64 0.0066 psi(Hydro),p(Water) 200.025.70 0.0064 psi(Hab,ORV),p(Temp,Water) 200.085.76 0.0062 psi(ORV),p(Temp,Humid,Water) 200.326.00 0.0055 psi(Hab),p(Water,Season) 200.366.04 0.0054 psi(Hab,Hydro),p(Season) 200.436.11 0.0052 psi(Hydro),p(Temp,Humid,Water) 200.566.24 0.0049 psi(Hab,Hydro),p(Water,Season) 200.576.25 0.0049 psi(Hydro,ORV),p(Temp,Water) 200.766.44 0.0044 psi(Hab),p(Temp,Water,Season) 201.066.74 0.0038 psi(Hab),p(Temp,Humid,Water,Season) 201.096.77 0.0038 psi(Hydro,ORV),p(Water) 201.146.82 0.0037 psi(Hab,Hydro),p(Temp,Water,Season) 201.246.92 0.0035 psi(Hab,Hydro),p(Temp,Humid,Water,Season) 201.276.95 0.0034 psi(Hydro,ORV),p(Temp,Humid,Water) 201.677.35 0.0028

PAGE 55

55 Table 3-5 (Continued). Model AIC Delta AIC AIC Weight psi(Hydro,ORV),p(.) 202.137.81 0.0022 psi(Hab,ORV,Hydro),p(Season) 202.498.17 0.0019 psi(Hab,ORV,Hydro),p(Water,Season) 202.628.30 0.0017 psi(Hab,ORV),p(Temp,Humid,Water) 202.748.42 0.0016 psi(Hab,ORV,Hydro),p(Temp,Water,Season) 203.238.91 0.0013 psi(Hab,ORV,Hydro),p(Temp,humid,Water,Season) 203.278.95 0.0013 psi(Hab,Hydro),p(Temp,Water) 204.3710.05 0.0007 psi(Hab),p(Temp,Water) 204.3810.06 0.0007 psi(Hydro,ORV),p(Temp,Humid) 204.4810.16 0.0007 psi(Hab),p(Water) 204.8710.55 0.0006 psi(ORV),p(.) 204.9310.61 0.0006 psi(.),p(.) 204.9410.62 0.0005 psi(Hab,ORV,Hydro),p(.) 205.0810.76 0.0005 psi(Hab,Hydro),p(Water) 205.1110.79 0.0005

PAGE 56

56 Table 3-6. Model selection re sults for the greenhouse frog ( Eleutherodactylus planirostris ), including Akaikes Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and Anderson 1998) for all models with any AIC weight. Model AIC Delta AIC AIC Weight psi(Hydro,ORV),p(Temp,Humid,Water,Saeson) 409.40 0.5162 psi(Hydro,ORV),p(Temp,Water,Season) 411.011.61 0.2308 psi(Hab,ORV,Hydro),p(Temp,Humid,Water,Season) 413.434.03 0.0688 psi(ORV),p(Temp,Humid,Water,Season) 414.094.69 0.0495 psi(Hydro,ORV),p(Water,Season) 414.575.17 0.0389 psi(Hab,ORV,Hydro),p(Temp,Water,Season) 415.025.62 0.0311 psi(ORV),p(Temp,Water,Season) 415.896.49 0.0201 psi(.),p(Temp,Humid,Water,Season) 417.428.02 0.0094 psi(Hab,Hydro),p(Temp,Humid,Water,Season) 418.258.85 0.0062 psi(Hab,ORV,Hydro),p(Water,Season) 418.519.11 0.0054 psi(ORV),p(Water,Season) 418.969.56 0.0043 psi(.),p(Temp,Water,Season) 419.149.74 0.004 psi(Hab,ORV),p(Temp,Humid,Water,Season) 419.369.96 0.0035 psi(Hydro),p(Temp,Water,Season) 419.4510.05 0.0034 psi(Hab,Hydro),p(Temp,Water,Season) 420.0710.67 0.0025 psi(Hab),p(Temp,Humid,Water,Season) 421.1311.73 0.0015 psi(Hab,ORV),p(Temp,Water,Season) 421.2811.88 0.0014 psi(.),p(Water,Season) 422.4113.01 0.0008 psi(Hab),p(Temp,Water,Season) 423.0813.68 0.0006 psi(Hydro),p(Water,Season) 423.1413.74 0.0005 psi(Hab,Hydro),p(Water,Season) 423.9714.57 0.0004 psi(Hydro,ORV),p(Season) 424.0214.62 0.0003 psi(Hab,ORV),p(Water,Season) 424.6715.27 0.0002 psi(Hab),p(Water,Season) 426.5517.15 0.0001 psi(Hab,ORV,Hydro),p(Season) 427.2117.81 0.0001

PAGE 57

57 Table 3-7. Model selection results for the eastern narrow-mouthed toad ( Gastrophryne carolinensis ), including Akaikes Information Cr iterion (AIC) and the delta AIC and AIC weight (Burnham and Anderson 1998) fo r all models with any AIC weight. Model AIC Delta AIC AIC Weight psi(Hydro,ORV),p(Season) 113.970.00 0.3182 psi(Hydro,ORV),p(Water,Season) 115.961.99 0.1177 psi(ORV),p(Season) 117.043.07 0.0686 psi(Hydro,ORV),p(Temp,Water,Season) 117.493.52 0.0548 psi(Hydro,ORV),p(Water) 117.653.68 0.0505 psi(Hab,Hydro),p(Season) 118.154.18 0.0394 psi(.),p(Season) 118.434.46 0.0342 psi(ORV),p(Temp,Water,Season) 119.005.03 0.0257 psi(ORV),p(Water,Season) 119.035.06 0.0254 psi(Hydro,ORV),p(Temp,Humid,Water,Season) 119.485.51 0.0202 psi(Hydro,ORV),p(Temp,Water) 119.525.55 0.0198 psi(Hab,ORV,Hydro),p(Season) 119.565.59 0.0194 psi(Hydro,ORV),p(.) 119.735.76 0.0179 psi(Hydro,ORV),p(Temp) 120.316.34 0.0134 psi(.),p(Water,Season) 120.416.44 0.0127 psi(Hydro),p(Season) 120.436.46 0.0126 psi(Hydro,ORV),p(Temp,Humid,Water) 120.826.85 0.0104 psi(Hab,Hydro),p(Water,Season) 121.217.24 0.0085 psi(Hab),p(Season) 121.387.41 0.0078 psi(ORV),p(Water) 121.457.48 0.0076 psi(Hab,ORV,Hydro),p(Water,Season) 121.547.57 0.0072 psi(ORV),p(.) 121.687.71 0.0067 psi(.),p(Temp,Water,Season) 121.777.80 0.0064 psi(Hab,ORV,Hydro),p(Temp,Water,Season) 121.947.97 0.0059 psi(Hydro,ORV),p(Temp,Humid) 122.038.06 0.0057 psi(Hydro),p(Water,Season) 122.418.44 0.0047 psi(ORV),p(Temp,Humid,Water,Season) 122.478.50 0.0045 psi(ORV),p(Temp,Water) 122.778.80 0.0039 psi(Hab,Hydro),p(Temp,Water,Season) 122.828.85 0.0038 psi(.),p(.) 122.868.89 0.0037 psi(.),p(Water) 122.948.97 0.0036 psi(Hab,ORV),p(Season) 123.089.11 0.0033 psi(ORV),p(Temp) 123.219.24 0.0031 psi(Hab,Hydro),p(.) 123.309.33 0.0030 psi(Hab),p(Water,Season) 123.349.37 0.0029 psi(Hab,Hydro),p(Water) 123.349.37 0.0029 psi(Hab,Hydro),p(Temp,Water) 123.429.45 0.0028 psi(.),p(Temp,Humid,Water,Season) 123.779.80 0.0024 psi(Hydro),p(Temp,Water,Season) 123.779.80 0.0024 psi(Hab,Hydro),p(Temp,Humid,Water,Season) 123.789.81 0.0024

PAGE 58

58 Table 3-7 (Continued). Model AIC Delta AIC AIC Weight psi(Hab,ORV,Hydro),p(Temp,Humid,Water,Season) 123.9910.02 0.0021 psi(Hab,ORV,Hydro),p(Water) 124.0510.08 0.0021 psi(.),p(Temp,Water) 124.1410.17 0.0020 psi(Hab,Hydro),p(Temp,Humid,Water) 124.3010.33 0.0018 psi(.),p(Temp) 124.3110.34 0.0018 psi(Hab,ORV,Hydro),p(Temp,Water) 124.5010.53 0.0016 psi(Hab,ORV,Hydro),p(Temp) 124.5510.58 0.0016 psi(Hab),p(Temp,Water,Season) 124.6510.68 0.0015 psi(ORV),p(Temp,Humid,Water) 124.6610.69 0.0015 psi(Hydro),p(.) 124.8610.89 0.0014 psi(ORV),p(Temp,Humid) 124.9210.95 0.0013 psi(Hydro),p(Water) 124.9410.97 0.0013 psi(Hab,ORV),p(Water,Season) 125.0811.11 0.0012 psi(Hab,Hydro),p(Temp) 125.1211.15 0.0012 psi(Hab,ORV,Hydro),p(Temp,Humid,Water) 125.4311.46 0.0010 psi(Hab),p(.) 125.7711.80 0.0009 psi(Hydro),p(Temp,Humid,Water,Season) 125.7711.80 0.0009 psi(.),p(Temp,Humid) 126.0312.06 0.0008 psi(.),p(Temp,Humid,Water) 126.0412.07 0.0008 psi(Hydro),p(Temp,Water) 126.1412.17 0.0007 psi(Hab),p(Water) 126.1712.20 0.0007 psi(Hydro),p(Temp) 126.3112.34 0.0007 psi(Hab,ORV),p(Temp,Water,Season) 126.3912.42 0.0006 psi(Hab),p(Temp,Humid,Water,Season) 126.6512.68 0.0006 psi(Hab,Hydro),p(Temp,Humid) 127.0013.03 0.0005

PAGE 59

59 Table 3-8. Sums of Akaikes Information Crite rion (AIC) weights for a ll models including the ORV index, habitat type, or hydrologic index covariates fo r each of the four focal anuran species. Species ORV Index Habitat Hydrologic Index Oak Toad 0.68190.88880.9988 Southern Toad 0.49890.26790.2755 Greenhouse Frog 0.97060.12180.7992 Eastern Narrow-mouthed Toad 0.80500.12670.7605

PAGE 60

60 Table 3-9. Beta estimates, standa rd errors (S.E.), and lower a nd upper 95% confidence intervals (C.I.) for the ORV use index covariate from the best model for each of the four focal anuran species. NA indicates that numeri cal convergence could not be reached and no estimate of the S.E. is available. Species Beta estimate S.E. Lower 95% C.I. Upper 95% C.I. Oak Toad -580.9251NASouthern Toad 21.232124.0724-25.9498 68.4140 Greenhouse Frog -3.91881.6130-7.0803 -0.7573 Eastern Narrow-mouthed Toad -5.56016.1949-17.7021 6.5819

PAGE 61

61 Figure 3-1. An aerial photograph depicting off-road vehicle damage in marl prairie habitat in Big Cypress National Preserve.

PAGE 62

62 Figure 3-2. Map of amphibian occupancy sampling locations within BCNP during 2002-2003 (n=70).

PAGE 63

63 CHAPTER 4 THE EFFECT OF TOE-CLIPPING ON TWO SPECIES OF TREEFROGS Introduction Accurately assessing the status and trends of amphibian populati ons is an important part of amphibian conservation and management, especia lly where amphibian sp ecies are threatened with extinction or are of special interest to ma nagers (Alford and Richar ds 1999, Stuart et al. 2004). Many studies rely on c ounts of amphibians to provide information on populations, but count data not adjusted for detection cannot be used to monitor amphibian population status (Schmidt 2003). Further, estimates of vital rates, such as survival probability, are crucial for addressing the causes of declines and managing populations (Biek et al 2002). One important method for obtaining estimates of abundance and su rvival involves recaptu ring uniquely marked individuals (Jolly 1965, Seber 1965). Many marking methods have been developed for amphibians (Donnelly et al. 1994). To be suitable for use in estimation of survival ra tes, a marking technique must be permanent, not adversely affect the marked animal, and not a ffect the probability of capture on subsequent samples (Williams et al. 2002). Few marking methods meet these assumptions when applied to small anurans (<30 mm). Tagging methods such as passive integrated transponder (PIT) microchips (Ireland et al. 2003) may be suitable for larger anurans, but are not useful for species with small body size. Tattooing (Kaplan 1959), freeze branding (Daugherty 1976), and fluorescent dye (Nauwelaerts et al. 2000) techniques are cumber some for field use and may not be permanent. Toe-clipping, the systematic removal of toes in unique combinations, is a lowcost, efficient method of permanently marking anurans (Donnelly et al. 1994, Luddecke and Amezquita 1999), but recent analysis suggests that toe-clipping may decrease survival of some

PAGE 64

64 species of anurans (Parris and McCarthy 2001, McCarthy and Parris 2004) and some consider the practice unethical or sc ientifically unsound (May 2004). McCarthy and Parris (2004) obser ved a negative relationship between the return rate (defined as the probability of survival times the probability of capture) of several species of anurans and the number of toes removed during marking. The models of McCarthy and Parris (2004) assume a constant probability of capture, a nd thus they conclude th at survival rates are lower for frogs with more toes removed. Ho wever, it is known that capture and survival probabilities often vary with time due to environm ental factors not related to the marking method (Williams et al. 2002). Mark recapture analyt ical techniques make it possible to use the information gained from uniquely marked animal s to directly estimate survival and capture separately and determine whether th ere is an effect of toe removal on survival or capture rates. I applied capture-mark-recapture techniques to estimate survival and capture probabilities of green treefrogs ( Hyla cinerea ) and squirrel treefrogs ( H. squirella ) in southern Florida, USA. My objective was to determine if increasing th e number of toes removed for marking had a negative effect on survival or ca pture probability of these treefr og species. I hypothesized that toe-clipping would have no effect on survival or capture probability in either species. I used Cormack-Jolly-Seber open populatio n mark-recapture models to estimate apparent survival and capture probability (Lebreton et al. 1992), and information theoretic methods based on Akaikes Information Criterion (AIC) for model selection (Burnham and Anderson 1998). Methods Study Site I established six long-term res earch sites in Big Cypress Nati onal Preserve, Collier County, FL, USA in April 2004. Each site consisted of 100-170 5.1-cm polyvinyl chloride (PVC) pipe refugia (Boughton et al. 2000) erec ted from the ground in groups of 7 by 7 grids of 49 pipes with

PAGE 65

65 5 m spacing and other pipes (6-9) located between grids to monito r movement between habitats. The total number of ref ugia at all sites was 840. Sites were located in pineland and cypress st rand habitats, as well as adjacent marsh and prairie habitats (Duever 2005). Pinelands are upland habitats dominated by slash pine ( Pinus elliottii ). Cypress strands are seasonally flooded forested wetlands with dominated by bald cypress ( Taxodium distichum ). Marsh habitats are long-hydrope riod forb-dominated wetlands, and prairie habitats are short-hydroperiod sedg e-dominated wetlands. Both marsh and prairie lack a woody overstory (Duever et al. 1986). Captur es from all habitats we re combined in this analysis to examine the effect of toe-clipping as preliminary analysis indicated that survival and capture rate parameters were primarily homogene ous across habitats during the sampling period. Capture Recapture Refugia were checked once monthly during th e period from 15 November 2004 to 30 June 2005. Frogs captured in or on the refugia were pl aced in clear plastic bags for measurement. Individual frogs were identified to species, measured snout-to-uros tyle length (SUL) in mm, and examined for toe clip marks. Unmarked green treefrogs greater than 24 mm SUL and unmarked squirrel treefrogs greater than 17 mm SUL were assigned a new clip number, unique to the site, which required the removal of two, three, or four toes. It was difficult to mark and read marks on smaller individuals, so they were not mark ed. The numbering system followed that of Donnelly (1989) with the modificati ons that no more than one toe per foot was removed, and the proximal toe on each forelimb was never removed. Toes were removed with stainless steel scissors sterilized in alcohol. Recaptured frogs were examined for signs of toe regeneration, and when necessary, toes were re-clipped.

PAGE 66

66 Survival Analysis I used the Cormack-Jolly-Seber mark-reca pture model (Lebret on et al. 1992) as implemented in program MARK (White and Bur nham 1999) to perform survival analysis for both species. Individuals of both species were di vided into three groups fo r analysis: those with two, three, and four toes removed. A series of 23 models representing different hypotheses about the effects of time and group on apparent survival ( ) and capture probability ( p ) were fit for both species (Table 4-1). Models including the effect of toe cli pping were constructed so as to force the effect to be monotonic (i.e. removing 4 toes had twice the effect of removing 3 toes when compared to 2 toes). Goodness of fit of the model st ructure was assessed by esti mating the variance inflation factor c using the parametric bootstrap method im plemented in program MARK (White and Burnham 1999) on the most general model in the model set, clip t clip tp* Model selection was conducted using the information-theoretic appro ach of Burnham and Ande rson (1998) with the Quasi-likelihood Akaikes Information Criterion ad justed for over dispersion of data and small sample sizes (QAICc). Results During the sampling period I captured, marked and released a tota l of 1296 individual green treefrogs and 658 individual squirrel tr eefrogs, of which 712 and 408 respectively were subsequently recaptured. Return rates for frogs with 2, 3, and 4 toes removed declined monotonically from 60.92% to 51.25% among Green Treefrogs and from 70.00% to 60.19% among Squirrel Treefrogs (Table 4-2). At least 6 individuals of each species/toe-clip group were captured and released during each of the fi rst seven capture occasions (Figure 4-1).

PAGE 67

67 The parametric bootstrap of the most gene ral Cormack-Jolly-Seber model chosen for Green Treefrogs produced an estimate of c = 1.656, indicating mild lack of fit of the data to the model. Two models, t clip tp andclip t clip tp had a delta QAICc of less than 2, and therefore had strong support (Table 4-3). Both of these models include the toe clip e ffect on survival, and one includes a toe clip effect on capture. Models that included a toe-clip grou p effect on survival had 85.14% of the QAICc weight among the set of candi date models, and models that included toeclip group effect on capture probabil ity had 34.16% of the model weight. The estimated beta for the toe-clip effect on su rvival in green treefrogs from the best model was -0.3963 (S.E. =0.1377; Table 4-4). There was a mean absolute decrease in survival of 5.02 % and 11.16% for frogs with 3 and 4 toes removed, respectively, compared to frogs with just 2 toes removed (Figure 4-2). The estimated beta fo r the toe-clip effect on capture probability in green treefrogs was 0.1731 (S.E. =0.1270), but the 95% confidence interval included 0 (Table 44). For Squirrel Treefrogs, the estimate of c from the parametric bootstrap was 1.848. Four models had delta QAICc values less than 2 (Table 4-5). Two of the top 4 models included the toe-clip effect on survival and 2 included the toe-clip effect on cap ture. Models that included the toe-clip effect on survival for squirre l treefrogs received 36.09% of the QAICc weight, and models that included the toe-clip effect on capture probability accounted for 47.29% of the total weight. The estimated beta for the toe-clip effect on su rvival in Squirrel Treefrogs from the best model that included it was 0.0231 (S.E. =0.1379), which is a slightly pos itive effect of toe clipping, but the 95% confidence interval includes 0 (Table 4-4). The estimate of beta for the

PAGE 68

68 toe-clip effect of capture on S quirrel Treefrogs from the best m odel that included it was -0.1815 (S.E. =0.1139), but the 95% confidence interval for this parameter also includes 0 (Table 4-4). Discussion I found strong evidence of a nega tive effect of toe removal on Green Treefrog survival, but only very limited evidence of an effect of toe-clipping on Green Treefrog capture probability. On average, frogs with 3 and 4 toes remove d had 5% and 11% lower survival probabilities respectively than Green Treefrogs with 2 toes removed. An eff ect of toe-clipping on capture in Green Treefrogs is not supported due to low AIC weights for models with a toe-clip effect on capture probability (Table 4-3) and a 95% confidence interval for the toe-clip effect beta that includes 0 (Table 4-4). There was also little evidence of an effect of toe-clipping on su rvival or capture probability of Squirrel Treefrogs. This was due to low AIC weights for models that include toe-clip effects (Table3-5) and beta values that had 95% conf idence intervals that included 0 (Table3-4). Average values for survival were equivalent regardless of the number of toes removed in Squirrel Treefrogs. Thus, only one species, the Green Treefrog, was found to show a negative response to toe-clipping. McCarthy and Parris (2004) reporte d an estimated 4-11% reducti on in the retu rn rate of frogs for each toe removed. The best models in McCarthy and Parriss (2004) analysis allow the change in return rate to vary li nearly with the number of toes removed. This model appears to fit my results for Green Treefrog survival well, wher e there was 5-6% decrease in survival per toe removed. However Squirrel Treefrogs did not sh ow the same pattern. Although there did appear to be a decline in the return rate of Squirrel Treefrogs (Table 4-2), th ere was no reduction in survival due to removing more toes (Figure 4-2).

PAGE 69

69 My results differ from those of McCarthy and Parris (2004) in the magnitude and generality of the effect that was detected. Alt hough return rates of both species in this study did decrease monotonically with the number of toes removed (Table 4-1), the estimated effect on survival was less than that expected from the results of McCarthy and Parris (2004). One reason for this difference is the inclusion of estimates of capture probability in my analysis. Rather than assuming constant capture probabili ty, I directly estimated it. Us ing return rates alone without accounting for heterogeneity in captu re probability could lead to misinterpreting a reduction in encounter rates as a reduction in survival. In addition, both ca pture probability and apparent survival were allowed to vary w ith time in this study. Most of the best models for both species include time dependence for both su rvival and capt ure probability. It is apparent that all species of frogs do not show the same response to toe-clipping. Some species appear to be especi ally susceptible to infections or loss of mobility due to toe removal (Golay and Durrer 1994, Lemckert 1996). Ev en the two species at th e same site in this study showed a difference in the effect of toeclipping on survival. Mark-recapture analysis provides a robust method for estimating the effe ct of toe clipping on survival and capture probability. It is preferable to using the return rate because it does not assume a constant capture probability across time or toe clip treatment. Using mark-recapture modeling to estimate survival and capture probabilities and using in formation-theoretic model selection to look for effects of the marking method should provide a usef ul technique for testing the efficacy of toeclipping for other amphibian species.

PAGE 70

70 Table 4-1. List of 23 models analyzed in Pr ogram MARK for captures of both Green Treefrogs and Squirrel Treefrogs in Big Cypre ss National Preserve during 2004-2005. Explanation defines each model in terms of the effects of time (t) and toe-removal group (clip), on apparent survival ( ) and capture probability (p). Model Explanation .p Survival and capture probabilit y constant throughout study .pt Survival varies with time; capture constant tp. Survival constant; capture varies with time t tp Both survival and capture vary with time .pclip Survival is different among toe removal groups; capture constant clipp. Survival constant; capture is di fferent among toe removal groups clip clipp Survival and capture are both different among toe removal groups t clipp Survival is group-dependent; capture varies with time clip tp Survival varies with time; capture is group-dependent .*pclip t Survival is a function of time, group, an d their interaction; capture constant *.clip tp Survival constant; capture is a functi on of time, group, and their interaction t clip tp* Survival is interactive effect of time and group; capture varies with time clip t tp* Survival varies with time; capture is interactive effect of time and group clip t clipp* Survival varies with toe group; capture is interactive effect of time and group clip clip tp* Survival is interactive effect of time and group; capture varies with toe group clip t clip tp* Survival and capture are both an in teractive effect of time and group pclip t Survival is additive effect of time and group; capture constant clip tp. Survival constant; capture is additive effect of time and group t clip tp Survival is additive effect of time and group; capture varies with time clip t tp Survival varies with time; capture is additive effect of time and group clip t clipp Survival varies with toe group; capture is additive effect of time and group clip clip tp Survival is additive effect of time and group; capture varies with toe group clip t clip tp Survival and capture are both the additive effect of time and group

PAGE 71

71 Table 4-2. The number of green treefrogs and squi rrel treefrogs marked by removing 2, 3, or 4 toes in Big Cypress National Preserve, Collier County, FL, Nov. 2004-June 2005, the number of individuals recaptured, and th e return rate (proportion of marked individuals recaptured at least once). Species Treatment No. Marked No. Recaptured Return Rate Green Treefrog 2 toes 87 53 60.92% 3 toes 848 474 55.90% 4 toes 361 185 51.25% Squirrel Treefrog 2 toes 80 56 70.00% 3 toes 470 287 61.06% 4 toes 108 65 60.19%

PAGE 72

72 Table 4-3 Model selection table for Cormack-J olly-Seber open population mark-recapture model of Green Treefrogs, including Quasi-like lihood Akaikes Information Criterion for small sample sizes (QAICc), model weights based on QAICc, the number of parameters n each model, and the model de viance. Model structure includes the effects of time (t), and toe-clip gr oup (clip) on appa rent survival ( ) and capture probability (p). Model QAICc Delta QAICc QAICc Weights Num. Par QDeviance t clip tp 2944.18630.00000.518216 2911.9877 clip t clip tp 2945.07110.88480.332917 2910.8476t tp 2946.80392.61760.140013 2920.6711clip t tp 2952.41458.22820.008516 2920.2161t clip tp* 2958.970914.78460.000327 2904.4171clip t tp* 2962.447218.26090.000127 2907.8932clip t clip tp* 2964.931020.74470.000036 2891.9520t clipp 2968.972524.78620.00009 2950.9070clip clip tp 2972.662728.47640.000012 2948.5491tp. 2972.760928.57460.00008 2956.7085 pclip t 2973.113928.92760.000011 2951.0178clip t clipp 2974.890830.70450.000012 2950.7771clip tp. 2976.244632.05830.000011 2954.1485.pt 2976.725532.53920.00008 2960.6730clip tp 2978.706434.52010.00009 2960.6409. *.clip tp 2985.569041.38270.000022 2941.1988clip t clipp* 2986.908742.72240.000024 2938.4695clip clip tp* 2987.474043.28770.000023 2941.0701 .*pclip t 2987.666443.48010.000022 2943.2964clip clipp 3392.0314447.84510.00004 3384.0168.pclip 3394.4325450.24620.00003 3388.4238. .p 3404.7537460.56740.00002 3400.7494clipp 3406.7036462.51730.00003 3400.6947

PAGE 73

73 Table 4-4. Estimates, standard error (SE), and th e 95% confidence interval of the beta values for the toe-clip effect on apparent survival ( ) and capture probability ( p ). 95% Confidence Interval Species Parameter Beta SE Lower Upper Green Treefrog -0.3963 0.1078 -0.6075 -0.1851 p 0.1731 0.1270 -0.0759 0.4221 Squirrel Treefrog 0.0231 0.1379 -0.2472 0.2934 p -0.1815 0.1139 -0.4047 0.0417

PAGE 74

74 Table 4-5. Model selection ta ble for Cormack-Jolly-Seber open population mark-recapture model of Squirrel Treefrogs, including Quasi-likelihood Akai kes Information Criterion for small sample sizes (QAICc), model weights based on QAICc, the number of parameters n each model, a nd the model deviance. Model structure includes the effects of time (t), and toe-cl ip group (clip) on a pparent survival ( ) and capture probability (p). Model QAICc Delta QAICc QAICc Weights Num. Par QDeviance clip t tp 1653.9976 0 0.2481 16 1621.6121 .pt 1654.0015 0.0039 0.2476 10 1633.8462 t clip tp 1655.3701 1.3725 0.1249 16 1622.9846 clip t clip tp 1655.7340 1.7364 0.1041 17 1621.2999 pclip t 1656.0325 2.0349 0.0897 11 1633.8461 clip tp 1656.3014 2.3038 0.0784 9 1638.1745 t tp 1656.7953 2.7977 0.0613 14 1628.4980 clip clip tp 1657.6489 3.6513 0.0400 12 1633.4284 tp. 1663.6425 9.6449 0.0020 8 1647.5410 clip tp. 1664.1035 10.1059 0.0016 11 1641.9171 t clipp 1664.1404 10.1428 0.0016 9 1646.0135 clip t clipp 1666.0178 12.0202 0.0006 12 1641.7974 clip t tp* 1668.7560 14.7584 0.0002 26 1615.7538 t clip tp* 1670.3368 16.3392 0.0001 25 1619.4096 .*pclip t 1672.2840 18.2864 0.0000 21 1629.6268 clip clip tp* 1675.9932 21.9956 0.0000 23 1629.2069 *.clip tp 1678.5895 24.5919 0.0000 21 1635.9323 clip t clipp* 1681.6591 27.6615 0.0000 23 1634.8728 clip t clip tp* 1684.2103 30.2127 0.0000 35 1612.4000 .pclip 1716.5722 62.5746 0.0000 3 1710.5552 .p 1716.7540 62.7564 0.0000 2 1712.7456 clipp 1716.7772 62.7796 0.0000 3 1710.7604 clip clipp 1717.8694 63.8718 0.0000 4 1709.8412

PAGE 75

75 Number of Individuals Captured 0 100 200 300 400 500 600 2 Toes 3 Toes 4 Toes Sampling Occasion 012345678 Number of Individuals Captured 0 100 200 300 400 500 600 A B 2 Toes 3 Toes 4 Toes Figure 4-1. Numbers of individuals of green treef rogs (A) and squirrel treefrogs (B) captured and released in each of the three toe removal groups during the first 7 sampling occasions.

PAGE 76

76 Apparent Survival 0.0 0.2 0.4 0.6 0.8 1.0 2 Toes 3 Toes 4 Toes Sampling Interval 01234567 Apparent Survival 0.0 0.2 0.4 0.6 0.8 1.0 2 Toes 3 Toes 4 Toes A B Figure 4-2. Apparent survival ( ) and 95% confidence interval of green treefrogs (A) and squirrel treefrogs (B) in each toe removal group category across the first 6 monthly sampling intervals. Estimates for squirrel tr eefrogs are averaged across models as no model had a majority of the QAICc weight (Burnham and Anderson 1998).

PAGE 77

77 CHAPTER 5 INFLUENCE OF HYDROLOGY ON SURVIVAL AND RECRUITMENT OF GREEN TREEFROGS The Everglades ecosystem of southern Fl orida has been substantially altered over the last 100 years by loss to agriculture and urbanization. Compartmentalization of the remaining Everglades into a network of artificially controlled impoundments has impeded historic flow patterns (Davis et al. 1994). A large-scale re storation effort, the Comprehensive Everglades Restoration Plan (CERP) was devised to attempt to restore natural hydrologic regimes to the remaining Everglades (D eAngelis et al. 1998). One measure of restoration succe ss was defined as recovering ecological structure and function to the natural areas. Consequently managers charged w ith decision-making for CERP need species that can serve as indi cators of ecosystem restoration success. Amphibians have been used in various lo cations as ecosystem indicator species (Welsh and Ollivier 1998, Galatowitsch et al. 1999, Sheridan and Olson 2003). Aspects of their natural history (e.g. aquatic larv al phase, permeable skin, and low dispersal ability) make them potentially well-suited as indicator species in th e Everglades as in other systems. However, there is no histor ical record of amphibian populations from before hydrologic alteration in the Everglades system to us e for comparison to current and post-restoration populations. Likewi se, it would be extremely difficult to experimentally manipulate environmental condi tions at the scale necessary to gauge the response of amphibians to hydrologic restoration. By monitoring amphibian populations and measuring how they respond to environmental changes at a local scale, it s hould be possible to ma ke predictions about how amphibians will respond to Everglades restoration on a landscape scale. As the major goal of CERP is restoration of hydrol ogy, it is important to know how amphibians

PAGE 78

78 will respond to changes in water depth and duration if amphibians are to be used as indicator species of re storation success. This respons e may be measured in population vital rates, and mark-recapture techniques allow the estimation of vital rates. Survival and recruitment rates can be estimated us ing open population mark-recapture analysis, and the population growth rate of populations can be derived from estimates of these two parameters (Pradel 1996). The goal of this study is to determine effects of seasonal changes in hydrology on the population vital rates of Green Treefrogs ( Hyla cinerea ) in the Everglades ecosystem to provide information on how hydrologic restoration in th e Everglades might impact frog populations. Modeling recruitment and surv ival with mark-recaptu re analysis will help build a model of how populations should respond to anthropogenic changes in hydrology. The contributions of survival and recruitmen t to population growth across seasons will be determined within closely c onnected habitats. The contributions may be important for determining critical time periods for reproduction at different water depths. I hypothesized that both survival and recr uitment are dependent on water depth and hydrologic season as well as time. I also hypothesized that captu re probability is dependent on season. The information gained in this study can later be used to build models predicting the respons e of frog populations to vari ous scenarios of hydrologic restoration of CERP. Methods Three long-term study sites were establis hed in Big Cypress National Preserve, Collier County, FL for this study. Each of these sites was placed at a location where 3 habitats, cypress strand, broad-leaf marsh, and short-hydroperiod prairie, were in close proximity. Habitats in the Everglades ecosystem are primarily different due to small

PAGE 79

79 topographical differences that create marked differences in hydrology (McPherson 1974, Duever et al. 1986). Broadleaf marshes have the longest hyd roperiod, and never completely dried during this study (Figure 5-1). Marsh sites are comprised of tall (1-2 m) emergent forbs, especially Pickerelweed ( Pontedaria cordata ), and lack a woody overstory. Cypress strands ar e intermediate in hydrology (Fi gure 5-1), and are distinct from the other habitats because of their closed canopy of bald cypress ( Taxodium distichum ). Prairie, sometimes referred to as marl prairie or dry prairie, has the shortest hydroperiod (Figure 5-1). These sites are characterized by a sedge dominated flora usually up to 1 m in hei ght (Duever et al. 1986). A grid of 49 polyvinyl chloride (PVC) pi pe refugia (Moulton et al. 1996, Boughton et al. 2000) was arranged within each habitat stratum at each site. Pipes were arranged 5 m apart in a 7 by 7 grid. Grids were located completely within a habitat stratum, but within 30 m of the adjace nt grids in the other habitats. PVC pipes used in this study were 50 mm in diameter and 1 m long. Each pi pe was placed vertically onto a wooden stake driven into the soil so the pipe could be easily lifted for inspection. All pipes in each grid were numbered for reference. Sites were sampled biweekly from April 2004 to November 2004 and once monthly from December 2004 to August 2005 (n=25). Water depth was measured at a fi xed depth gauge in the center of each plot during each sampling occasion. At each sample, all pipes were checked for frogs which were captured in sealable plastic bags. Frogs that escaped were noted and identified to species if possible. Captured frogs were identified to species and measured snout to urostyle (SUL). Green treefrogs less than 25 mm SUL were returned unmarked as frogs of this size were

PAGE 80

80 difficult to mark and read marks reliably. Previously unmarked frogs 25 mm SUL and larger were assigned a unique toe-clip co mbination (Donnelly et al. 1994) and marked by removing toes using stainless steel scissors sterilized in alcohol. Previously marked (recaptured) frogs were checked against a list of previously marked frogs to insure the clip was read correctly. The pipe number and grid location of each frog capture was noted, and frogs were released into the pipe opening immediately after necessary handling was completed. Capture-recapture data were analyzed us ing the temporal symmetry approach of Pradel (1996) with the -parameterization (t t tp ,; Williams et al. 2002) in Program MARK (White and Burnham 1999). The temporal symmetry approach uses reverse-time mark recapture analysis (Pollock et al. 1974, Pradel 1996) to estimate the seniority parameter, i (i.e. the probability that an individual alive at time i was also alive at time i -1). Although the parameter of interest in this study is actu ally the population growth rate (i), the -parameterization of the Pradel (1996) model is known to perform better than alternative parameterizations includingi, and i may be derived from the estimates of thei and i parameters (Williams et al. 2002). The temporal symmetry method uses assumptions similar to standard open-popul ation Cormack-Jolly-Seber mark-recapture models (Lebreton et al. 1992). These assu mptions include that no marks are lost or misread and that no non-random temporary emigration occurs. Every marked animal should have the same probability of capture in and survival to the next sampling period, and all individuals should have inde pendent fates (Williams et al. 2002). In the analysis, water depth was a cova riate and captures were grouped by habitat and season. Mean water depth within each ha bitat across the 3 sites was standardized so

PAGE 81

81 that the mean fell between 0 and 1 as the lo git link function was used in MARK analyses (Williams et al. 2002). Individual frogs were ca tegorized as belonging to cypress, marsh, or prairie habitat groups. Alt hough it was rare, some frogs we re caught in one habitat and later moved to an adjacent ha bitat. Frogs that were captu red in more than one habitat were assigned to the group with the most captures, or the la st capture if caught an even number of times in two ha bitats. Sampling occasions were grouped into 5 seasons based on the annual pattern of rainfall (Table 5-1). These seasons were used in the models in place of full time dependence because they are more biologically meaningful than date alone. There are three parameters estimated in this temporal symmetry model: survival ( ), capture probability ( p ), and the seniority parameter ( ). Models representing different combinations of th e water depth covariate, hab itat group, and season (time) structure were constructed that were a priori determined to be biol ogically meaningful. Model selection was conducted using the information-theoretic approach based on Akaikes information criterion adjusted for small sizes (AICc; (Burnham and Anderson 1998). i during each sampling interval was derived in Program MARK using the equation given in Williams et al. (2002): 1 / i i i Nichols et al. (2000) demonstrat e the relationshi p between the parameter of the Pradel (1996) temporal symmetry modeling approach and relative cont ributions of survival and recruitment to the change in population grow th. The contributions of survival and recruitment to i are analogous in this case toi and 1-i respectively (Nichols et al. 2000).

PAGE 82

82 Results During the 17 months of sampling, a tota l of 1069 individual Gr een Treefrogs were marked and 1054 recaptures were recorded. There were 293 times that a frog escaped capture, and 173 frogs less than 25 mm SUL were released unmarked. Frog capture rates were higher in the dry season months, peaki ng in February and March 2005 (Figure 5-2), and frog captures were lowest during the p eak of the wet seasons (August 2004 and June 2005). Sizes of captured frogs showed a seas onal pattern (Figure 5-3). Mean SUL was lowest in August 2004 after the onset of the we t season, with a similar dip in July 2005 after the beginning of the 2005 wet season. Mo re frogs were captur ed in cypress and marsh habitats than in prairie, but habitat differences were only observed during the dry months (Figure 5-4). Model selection results from the mark-reca pture modeling indicat ed that the best model included the seasonal time structure, habitat group, and water-depth covariate for and and only habitat and season for p (Table 5-2). No othe r models had any AICc weight (Burnham and Anderson 1998). The habitat and season-specific values for varied widely, but were generally ar ound 0.80 (Figure 5-5). Estimates of also varied widely (Figure 5-6). Although season was included in the model for and a seasonal trend in either was not appa rent, but a pattern might be somewhat obscured by the water depth covariate. There does appear to be a strong seasonal trend in p (Figure 5-7). Capture rates were lowest at the sampling oc casions corresponding to the wet seasons of 2004 and 2005 (Table 5-1). i values across sampling interv als and habitats fluctuated from 0.22 to 2.02 (Figure 5-8). The contribution to i from survival was almost always greater than the contribution of recrui tment in each habitat (Table 5-3).

PAGE 83

83 Discussion The mark-recapture modeling results s upport the hypothesis that Green Treefrog survival and seniority varies with season and with water level, as well as among habitats. Capture probability also varied with season and habitat, but was not affected by water depth. Capture rates declined during the early wet seasons and capture probability was at minimum levels during the wet season. The mean size of individuals captured declined sharply during the wet seasons, presumably as young of the year in dividuals enter the population. Population growth rates were highest after the onset of the 2004 wet season and during December 2004 and January 2005 (Table 5-3). These two time periods also showed increased contributions to i from recruitment (1-i ). Recruitment as it is modeled in this study includes animals that are enter the population from reproduction as well as immigr ants into the population, and the models used can not differentiate the two (Nichols et al. 2000). Recruitment in this study may also include animals moving into the PVC pipe refugia for the first time, as the population studied is actually the population using the PVC pipes. The early wet season increase in recruitment was most likely a resu lt of reproduction. This increase coincides directly with the large drop in mean SUL of captured frogs (Figure 5-3), and large choruses of calling males and Green Treefrogs in amplexus were observed at all of the sites at the onset of the wet season (personal observation). The increase in recruitment in December 2004 and January 2005 is likely the re sult of cold weather causing frogs to seek refuge in the PVC pipes. Increased cap ture rates in PVC refugia at times of cold weather have been observed elsewhere in Fl orida (Donnelly et al. 2001, Zacharow et al. 2003). Movements of frogs between habitats were very rare and only account for about

PAGE 84

84 5% of all captures (unpublishe d data). It therefore seems unlikely that immigration into the study sites was a factor in recruitment rates. One important assumption of the mark-recaptur e analysis used in this study is that there is no non-random temporar y immigration. Immigration into the study sites does not appear to have occurred due to the very low movement rates between habitats. One potential violation of the tem porary immigration assumption is the increased use of pipes during the colder samples. If the frogs captured and marked in these samples later emigrated back out of the pipe population it would affect estima tes of survival. However, capture rates were actually higher in the months following the December and January increase in recruitment (Figure 5-2) when population growth was not increasing (Table 53). It appears from this pattern that once frogs moved into the PVC pipe population they remained in the population, therefore this woul d not be a case of temporary immigration. Green Treefrog populations in BCNP fluctuat e on a strongly annual cycle driven by the hydrologic cycle in the Everglades ecosyst em. The majority of reproduction occurs at the onset of the wet season. The number of frogs in the population slowly declines until late in the dry season when few large, adu lt frogs remain to accomplish breeding at the onset of the next wet season. The mark-recap ture modeling results from this study (Table 5-2) corroborate that survival and seniority rates are seasonal and related to water depth. The average monthly survival rate for green treefrogs across hab itats and months is 81.0%, or only 8.0% annually. Few published survival rates of frogs exist for comparison, but these estimates appear to be low in comparison to estimates of survival of the Afro-Tropical Pig-No sed Frog (Grafe et al. 2004) and the pig frog (Wood et al.1988).

PAGE 85

85 The seasonal patterns observed in capture probab ility as well as recruitment into the population demonstrate the importance of long-term monitoring of sites. Studying the population during a single seas on could lead to incorrect conclusions about trends in the population. In addition, this study has demo nstrated that differen ces in vital rates can exist between adjacent sites in different ha bitats. This suggests that population level changes may be occurring at a very small s cale, and probably have a strong relationship to hydrology. Capture probabi lity was not a function of water depth (Table 5-2), but there was a strongly seasonal pattern to captu re probability (Figure 5-7) and to capture rates (Figure 5-2). Although prairie habitat usually had the lowe st rate of population growth (Figure 58) across sampling intervals, it almost always had the highest values for contribution to growth from recruitment (Table 5-3). This is noteworthy because pr airie is the habitat with the shortest hydroperiod, meaning it is likely to be the most impacted by hydrologic restoration. Large areas of the eastern Everglades are comprised of a similar shorthydroperiod graminoid prairie, the rocky glades (Davis et al. 2005). If the spatial and temporal pattern of hydrology in the roc ky glades area is altered during CERP restoration, an effect on Green Treefrog populations should be detected. For example, an extension of the annual period of inundation could increase Gr een Treefrog reproduction. Monitoring of Green Treefrog populations may provide an efficient means for indicating if the vital ecological processes associated with the seasonal pa ttern of inundation are functioning appropriately after hydrologic restoration. Prior to this study, this process was not well understood, and consequently, th e proposed actions in CERP have not been evaluated with respect to e ffects on amphibian reproduction.

PAGE 86

86 Table 5-1. Dates of each sample of the PVC pipe refugia and the season to which each sample was assigned for mark-recapture analysis. Sample Date Season 1 4/23/04 Dry season 2004 2 5/7/04 Dry season 2004 3 5/19/04 Dry season 2004 4 6/5/04 Dry season 2004 5 6/18/04 Dry season 2004 6 6/30/04 Dry season 2004 7 7/15/04 Wet season 2004 8 7/29/04 Wet season 2004 9 8/10/04 Wet season 2004 10 8/27/04 Wet season 2004 11 9/9/04 Wet season 2004 12 9/23/04 Wet season 2004 13 10/13/04 Wet season 2004 14 10/21/04 Wet-dry transition 15 11/5/04 Wet-dry transition 16 11/22/04 Wet-dry transition 17 12/15/04 Wet-dry transition 18 1/6/05 Dry season 2005 19 2/3/05 Dry season 2005 20 3/7/05 Dry season 2005 21 4/15/05 Dry season 2005 22 5/18/05 Wet season 2005 23 6/15/05 Wet season 2005 24 7/22/05 Wet season 2005 25 8/17/05 Wet season 2005

PAGE 87

87 Table 5-2. Model selection results for al l models analyzed in Program MARK. Model describes the covariates a nd groups associated with (survival), p (capture probability), and (seniority parameter): s is the seasonal time structure (Table 5-1), h is the habitat grouping (di fferent estimates among the habitats), and w is the water depth c ovariate. Akaikes information criterion adjusted for small sample sizes (AICc), Delta AICc (the difference between the AICc for a model and the AICc for the model with the lowest AICc), the AICc weight (Burnham and Anderson 1998), the number of parameters in each model and the model deviances are given. Model AICc Delta AICc AICc Weight Num. Par Deviance (s+h+w), p (s+h), (s+h+w) 9496.890.00139 1420.46 (s+h+w), p (s+h+w), (s+h+w) 9522.1625.27041 1441.57 (s+h+w), p (s), (s+h+w) 9613.15116.26032 1551.21 (s+h), p (s+h), (s+h+w) 9665.58168.70041 1585.00 (s+h), p (s+h+w), (s+h+w) 9665.98169.09042 1583.31 (s+h), p (s+h), (s+h) 9670.52173.64044 1583.69 (s+w), p (s+w), (s+w) 9700.29203.40018 1667.03 (h), p (s+h), (s+h+w) 9745.69248.80031 1685.80 (s+h), p (s+h), (h) 9750.95254.06032 1689.01 (h), p (s+h), (s+h) 9751.66254.77033 1687.65 (h+w), p (h+w), (h+w) 9877.47380.58012 1856.39 (h), p (s+h), (h) 9956.30459.42021 1916.93 (s), p (s), (s) 9988.10491.21015 1960.94 (s), p (s), (s+w) 9989.80492.92016 1960.61 (s+h), p (h), (h) 10191.78694.89020 2154.45 (w), p (w), (w) 10584.901088.0103 2581.95 (h), p (h), (h) 11291.751794.8609 3276.73 (.), p (.), (.) 11428.491931.6003 3425.55

PAGE 88

88 Table 5-3. Estimates of the population growth ratei, the contribution of survival to i (i ), and the contribution of recruitment toi (1-i ) and the standard error (S.E.) of each estimate fo r each sampling interval ending on the date given. Cypress Marsh Prairie Date ]) .[ ( i iE S ]) .[ ( i iE S ]) 1 .[ ( 1i iE S ]) .[ ( i iE S ]) .[ ( i iE S ]) 1 .[ ( 1i iE S ]) .[ ( i iE S ]) .[ ( i iE S ]) 1 .[ ( 1i iE S 5/7/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.07 (0.049) 0.90(0.044) 0.10 (0.044) 0.99 (0.084) 0.74(0.075) 0.26 (0.075) 5/19/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076) 6/5/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076) 6/18/04 0.89 (0.067) 0.79(0.050) 0.21 (0.050) 0.65 (0.134) 0.92(0.039) 0.08 (0.039) 1.02 (0.083) 0.73(0.076) 0.27 (0.076) 6/30/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.24 (0.090) 0.80(0.058) 0.20 (0.058) 1.02 (0.083) 0.73(0.076) 0.27 (0.076) 7/15/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076) 7/29/04 1.78 (0.205) 0.56(0.064) 0.44 (0.064) 1.55 (0.187) 0.64(0.077) 0.36 (0.077) 1.34 (0.135) 0.74(0.075) 0.26 (0.075) 8/10/04 1.39 (0.073) 0.66(0.040) 0.34 (0.040) 1.27 (0.071) 0.74(0.049) 0.26 (0.049) 1.72 (0.178) 0.54(0.076) 0.46 (0.076) 8/27/04 1.39 (0.073) 0.66(0.040) 0.34 (0.040) 0.86 (0.103) 0.78(0.044) 0.22 (0.044) 1.04 (0.139) 0.59(0.074) 0.41 (0.074) 9/9/04 0.76 (0.126) 0.71(0.046) 0.29 (0.046) 0.72 (0.116) 0.79(0.044) 0.21 (0.044) 0.94 (0.154) 0.59(0.074) 0.41 (0.074) 9/23/04 1.20 (0.070) 0.68(0.045) 0.32 (0.045) 1.20 (0.064) 0.75(0.047) 0.25 (0.047) 1.55 (0.119) 0.55(0.075) 0.45 (0.075) 10/13/04 1.35 (0.068) 0.66(0.046) 0.34 (0.046) 1.25 (0.067) 0.75(0.048) 0.25 (0.048) 1.69 (0.168) 0.53(0.076) 0.47 (0.076) 10/21/04 1.37 (0.070) 0.66(0.046) 0.34 (0.046) 1.24 (0.066) 0.75(0.048) 0.25 (0.048) 1.55 (0.119) 0.55(0.075) 0.45 (0.075) 11/5/04 0.64 (0.132) 0.75(0.049) 0.25 (0.049) 0.75 (0.226) 0.81(0.057) 0.19 (0.057) 0.64 (0.112) 0.74(0.053) 0.26 (0.053) 11/22/04 1.23 (0.077) 0.71(0.040) 0.29 (0.040) 1.17 (0.088) 0.78(0.055) 0.22 (0.055) 1.08 (0.069) 0.71(0.055) 0.29 (0.055) 12/15/04 1.48 (0.069) 0.66(0.032) 0.34 (0.032) 1.31 (0.091) 0.75(0.053) 0.25 (0.053) 1.19 (0.077) 0.70(0.056) 0.30 (0.056) 1/6/05 1.77 (0.119) 0.56(0.038) 0.44 (0.038) 1.68 (0.200) 0.59(0.070) 0.41 (0.070) 1.28 (0.090) 0.69(0.058) 0.31 (0.058) 2/3/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 0.94 (0.100) 0.79(0.036) 0.21 (0.036) 1.01 (0.050) 0.67(0.040) 0.33 (0.040) 3/7/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 1.44 (0.076) 0.69(0.036) 0.31 (0.036) 1.01 (0.050) 0.67(0.040) 0.33 (0.040) 4/15/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 0.24 (0.028) 0.83(0.041) 0.17 (0.041) 1.01 (0.050) 0.67(0.040) 0.33 (0.040) 5/18/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 1.47 (0.087) 0.67(0.040) 0.33 (0.040) 1.01 (0.050) 0.67(0.040) 0.33 (0.040) 6/15/05 1.00 (0.000) 1.00(0) 0 (0) 0.45 (0.142) 0.46(0.121) 0.54 (0.121) 2.02 (0.398) 0.40(0.088) 0.60 (0.088) 7/22/05 1.00 (0.000) 1.00(0) 0 (0) 0.70 (0.196) 0.44(0.120) 0.56 (0.120) 0.49 (0.114) 0.47(0.093) 0.53 (0.093) 8/17/05 1.00 (0.000) 1.00(0) 0 (0) 0.22 (0.086) 0.48(0.122) 0.52 (0.122) 0.49 (0.114) 0.47(0.093) 0.53 (0.093)

PAGE 89

89 Figure 5-1. Mean water depth across the 3 sampling locations in BCNP in the cypress strand, broadleaf marsh, and prairie ha bitats from April 2004-August 2005.

PAGE 90

90 0 100 200 300 400 500 6004 /1 /0 4 5 /1 /0 4 5/31/04 6/3 0 /04 7 /3 0 /04 8/29/04 9/28/04 10/28/0 4 1 1/27/04 12/27/04 1/26/0 5 2/25/0 5 3/27/0 5 4/ 2 6/0 5 5/26/0 5 6/25/0 5 7/ 2 5/0 5 8/24/0 5DateCaptures0 10 20 30 40 50 60 70 80Marsh Water Depth (cm) Captures Marsh Depth Figure 5-2. Number of captures of Green Treefrogs during each of the 25 samples from April 2004 to August 2005 (solid line) and mean water depth (in cm ) of marsh plots (dotted line) across all 3 sites in BCNP.

PAGE 91

91 0 5 10 15 20 25 30 35 40 45 504/1/04 5 /1 /04 5/31/0 4 6 /3 0 /04 7 /3 0/0 4 8/2 9 /04 9 /2 8 /04 10/28/0 4 1 1/2 7 /0 4 12/ 2 7/04 1/26/05 2 /2 5/0 5 3/2 7 /05 4 /2 6/ 05 5/26/05 6/2 5 /05 7/25/0 5 8/24/05DateMean SUL (mm) Figure 5-3. Mean snout-to-urosty le (SUL) in mm of Green Treefrogs captured in all habitats at the three sites at each sampling occasion in BCNP.

PAGE 92

92 Figure 5-4. Number of captures by sample of Green Treefrogs in cypress, marsh, and prairie habitats at the three sites in BCNP.

PAGE 93

93 Cypress Marsh PrairieEstimateInterval 0.0 0.2 0.4 0.6 0.8 1.0 012345678910111213141516171819202122 Figure 5-5. Estimates with 95% confid ence intervals of apparent survival ( ) of Green Treefrogs for each sampling interval for cypr ess, marsh, and prairie habitat in BCNP.

PAGE 94

94 Cypress Marsh PrairieEstimateSampling Interval 0.0 0.2 0.4 0.6 0.8 1.0 012345678910111213141516171819202122 Figure 5-6. Estimates with 95% co nfidence intervals of seniority ( ) of Green Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP.

PAGE 95

95 Cypress Marsh PrairieEstimateSampling Occasion 0.0 0.2 0.4 0.6 0.8 1.0 0123456789101112131415161718192021222324 Figure 5-7. Estimates with 95% confid ence intervals of capture probability ( p ) of Green Treefrogs at each sampling occasion for cypre ss, marsh, and prairie habitat in BCNP.

PAGE 96

96 0 0.5 1 1.5 2 2.55/ 1/ 04 5 /31/0 4 6/30/04 7 /30/0 4 8 / 29/04 9 /28/04 10/ 28 /04 1 1/ 27/04 12/ 27/ 04 1 /26/05 2 /25/ 05 3/27/05 4 /26/0 5 5 / 26/05 6 /25 /05 7 /2 5/05 8 /24/ 05 Cypress Marsh Prairie Figure 5-8. Derived estima tes of population growth (i) of Green Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP.

PAGE 97

97 CHAPTER 6 CONCLUSION Introduction In the introductory chapter to this disserta tion I outlined the characteristics of useful indicator species and described why amphibians should be suitable as ecosystem indicators. In this conclusion, I will address each of the major characteristics of indicator species and describe whether the results of this resear ch support the use of amphibians as indicators in the Everglades ecosystem. Characteristics of Indicators Abundant and Efficient to Sample In order to be useful as an indicator, a sp ecies must be abundant a nd/or cost-effective to sample. Amphibians in south Fl orida clearly meet this qualif ication. Sampling amphibians using visual encounter te chniques, as in Chapter 3, was a very efficient way to sample several species at once with just 1 pers on hour per sample. This tech nique worked in a variety of habitats across a very large ge ographical area. Another sampli ng technique, PVC pipe refugia (Chapter 5), was also effective. This method involved more inte nsive work at fewer sites, but sample size was adequate to estimate population vital rates including su rvival and population growth. Collecting sufficient data to estimat e these vital rates on many other species would require much more effort. Therefore, I conclu de that amphibians are su fficiently abundant and easy to sample to be useful as indicators. There were also caveats to amphibian sampling de monstrated in this dissertation. Results from a known population of artificial frogs illustrate d severe bias in the sampling of amphibians using traditional methods, despite efforts to reduc e observer bias (Chapter 2). For this reason, methods using counts of amphibians as an index s hould be investigated cl osely before they are

PAGE 98

98 adopted as part of a monitoring program. In Chapter 4, the Green Treefrog was shown to be negatively affected by a common and efficient ma rking method, toe-clipping. The effect was small and may be manageable for some studies depending on the questions involved. The important lesson from these studies is that me thods for sampling indicator species must be thoroughly evaluated before de signing a monitoring program. Sensitive to Stresses on the System As previously mentioned, a species is only usef ul as an ecological indicator if it is sensitive to stresses to the ecosystem and responds to stress in a predictable manner. Chapter 3 demonstrated that four species of ground-dwelling a nurans were sensitive to an index of off-road vehicle (ORV) use. Occupancy rates of three of the species had a negative relationship with ORV use, as was predicted. Occupancy of a nother species, the Southern Toad, showed a positive association with ORV use. Although this did not follow the prediction, there are morphological differences between Southern Toad s and the other anurans that may explain the different response. Green treefrog populations demonstr ated sensitivity to an ecosystem process in Chapter 5. Water depth and season of year we re included in the best model of survival and recruitment. A period of high population growth was shown to coincide with the ons et of the wet season. This project was not manipulative, so we can only hypothesize how treefrogs would respond to various changes in the temporal and spatial pattern of hydrology. This could easily be investigated with additional mon itoring at different sites. Further monitoring of anurans in relation to ORV use and treefrog populations in relation to hydr ology will refine our knowledge of their response to these stressors and improve their use as indicator species.

PAGE 99

99 Responses to Stress Should Be Anticipatory To be useful to managers, indicator species s hould display a response to local changes that is anticipatory of an impending change to the whole system. These responses are most useful when they predict impacts that can be averte d by management before the whole system is negatively impacted. Treefrog populations might serve as sen tinels for the health of the Everglades ecosystem with regard to hydrologic re storation. Because of their annual cycle and short generation time (Chapter 5), we know tr eefrogs will respond faster to changes in the hydrology of a site than vegetation, for example. If monitoring of treefro gs shows that treefrogs are responding as would be expected (i.e., the tr eefrog population is changing to resemble other populations with the same hydrology), it is reasonab le to assume that rest oration is successful. In the case of ORV use in Big Cypress (Chapter 3), it is impossible to kno w if anuran occupancy responded anticipatorily to ORV use, as ORVs have been used for decades in that area. Integrate a Response across the Whole System Indicators are most useful wh en they can respond to changes in the whole system, rather than just in a few habitats or locations. Amphibi ans are clearly suitable for detecting changes to the whole Everglades system. They are found in all of the te rrestrial and non-marine aquatic habitats in south Florida. Am phibians responded across many habitats to ORV use in Chapter 3. Many habitats in South Florida ar e defined by small differences in hydropattern resulting in very different vegetation communities. It is clear th at treefrogs respond to both vegetation (habitat) and hydrology (Chapter 5). Known Response to Anthropogenic Stresses and Natural Disturbances In many situations it is important to be able to determine if a population of an indicator species is responding to anth ropogenic changes (e.g., manage ment actions) or natural disturbances. This would be espe cially useful in the Everglades ecosystem where rainfall can be

PAGE 100

100 extremely variable and hydrology at a site is a function of both natu ral phenomena (e.g., hurricanes) and water management. If a species is an indicator of hydrologic conditions, it might respond in a similar manner whether the change was a result of natura l rainfall patterns or anthropogenic management actions Two types of indicator sp ecies might be useful for differentiating the effects of natu ral and anthropogenic changes to the system: those that react very quickly and those that respond over decades or longer. Species th at respond very quickly (6-12 months after a change to the system) can be monitored al ong with environm ental variables to determine their response to specific manage ment actions or natural events. Species that respond over very long time periods will integrate the effects of natural disturbances and should indicate the effects of management actions. Speci es that respond quickly will be most useful for adaptive management in Everglades restoration and species that respond over long periods will be indicative of restoration success. Amphibians appear to be useful as indicators of both short a nd long-term effects. Green Treefrog populations will respond immediately to changes in th e hydrologic cycle. If a wet season at a site is different because of a natu ral event or a management action, the treefrogs should respond immediately by incr easing or decreasing reproduction. In the case of long-term anthropogenic changes, anurans were useful as indicators of ORV use, even years after use was suspended in some areas (Chapter 3). In this way, anurans were indicators of anthropogenic disturbance across a heterogeneous landscape subject to many natural disturbances. Conclusion This dissertation provides evidence that amphi bians are suitable as ecosystem indicators in general, and specifically as i ndicators of restoration success in the Everglades ecosystem of southern Florida. As with a ny group of indicators, care should be used when choosing sampling methods and particular species to use as indicators. Conti nued monitoring and additional

PAGE 101

101 research are crucial in using amphibians as indi cators of ecosystem rest oration success in the Everglades. Amphibian monitoring can benefit Everglades restoration if it is used wisely. I recommend sampling key amphibian species acro ss the system. Hylid treefrogs are easily marked and studied using captures from PVC pipes. This type of monitoring occurs on a smaller spatial scale, but will provide information on th e status and trends of local populations. A network of sites studied intens ively and modeled appropriately would allow managers to track changes in populations in rela tion to management actions. At the landscape scale, site occupancy modeling can be used on all amphibian species to m onitor the changes in occupancy (colonization and extinc tion) of sites throughout the Everglad es region. Site occupancy should also be modeled with respect to management act ions to determine the effects of such actions. In addition to monitoring, an expanded re search program can address questions about Everglades restoration efforts. One important question is the effect of re storation on the short hydroperiod wetland habitats of the eastern Everglades. How does the timing of inundation in these areas affect reproduction of amphibians? Is a reversal (inundation and then subsequent drying) event detrimental to frogs that may time breeding to coincide with the onset of inundation? These and other rese arch questions can be addresse d using amphibians as indicator species.

PAGE 102

102 LIST OF REFERENCES Alford, R. A., and S. J. Richards. 1999. Gl obal amphibian declines: a problem in applied ecology. Annual Review of Ecology and Systematics 30:133-165. Anderson, D. R., K. P. Burnham, B. C. Lubow, L. Thomas, P. S. Corn, P. A. Medica, and R. W. Marlow. 2001. Field trials of lin e transect methods applied to estimation of desert tortoise abundance. Journal of Wild life Management 65:583-597. Ashton, R. E., and P. S. Ashton. 1988. Handbook of reptiles and amphibians of Florida part three: the amphibians. Windwa rd Publishing, Miami, FL. Biek, R., W. C. Funk, B. A. Maxell, and L. S. Mills. 2002. What is missing in amphibian decline research: insights from ecological sensitiv ity analysis. Conservation Biology 16:728-734. Blaustein, A. R. 1994. Chicken Little or Nero's Fiddle? a perspective on declining amphibian populations. Herpetologica 50:85-97. Blaustein, A. R., D. B. Wake, and W. P. Sous a. 1994. Amphibian declines: judging stability, persistence, and susceptibility of populations to local and glob al extinctions. Conservation Biology 8:60-71. Boughton, R. G., J. Staiger, and R. Franz. 2000. Use of PVC pipe refugia as a sampling technique for hylid treefrogs. Ameri can Midland Naturalist 144:168-177. Brattstrom, B. H., and M. C. Bondello. 1995. Natu ral sounds and man-made noise in the desert. Pages 437-465 in J. Latting and P. G. Rowlands, editors. The California desert: an introduction to natural resour ces and man's impact. June La tting Books, Riverside, CA. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. Thomas. 2001. Introduction to distance sampling: es timating abundance of biological populations. Oxford University Press, Oxford, UK. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. Thomas. 2004. Advanced distance sampling : estima ting abundance of biological populations. Oxford University Press, Oxford, UK. Burnham, K. P., and D. R. Anderson. 1998. Model selection and multi-model inference: a practical information-theoretic approach Springer-Verlag, New York, NY, USA. Burnham, K. P., D. R. Anderson, and J. L. Laak e. 1980. Estimation of density from line transect sampling of biological populat ions. Wildlife Monographs 72. Cairns, J., P. V. McCormick, and B. R. Niederlehner. 1993. A proposed framework for developing indicators of ecosystem health. Hydrobiologia 236:1-44.

PAGE 103

103 Campbell, E. H., R. E. Jung, and K. C. Ri ce. 2005. Stream salamander species richness and abundance in relation to environmental fact ors in Shenandoah National Park, Virginia. American Midland Naturalist 153:348-356. Collins, J. P., and A. Storfer. 2003. Global amphi bian declines: sorting the hypotheses. Diversity and Distributions 9:89-98. Cook, R. D., and J. O. Jacobson. 1979. A design fo r estimating visibility bi as in aerial surveys. Biometrics 35:735-742. Crozier, G. E., and D. E. Gawlik. 2003. Wading bird nesting effort as an index to wetland ecosystem integrity. Wa terbirds 26:303-324. Crump, M. L., and N. J. Scott. 1994. Visual encounter su rveys. Pages 84-92 in W. R. Heyer, M. A. Donnelly, R. W. McDiarmid, L. C. Hayek, and M. S. Foster, editors. Measuring and monitoring biological diversity: standard methods for amphibi ans. Smithsonian Institution Press, Washington, D.C., USA. Dale, V. H., and S. C. Beyeler. 2001. Challe nges in the development and use of ecological indicators. Ecologica l Indicators 1:3-10. Daugherty, C. H. 1976. Freeze-bra nding as a technique for marking anurans. Copeia 1976:836838. Davis, S. M., E. E. Gaiser, W. F. Loftus, a nd A. E. Huffman. 2005. S outhern marl prairies conceptual ecological model. Wetlands 25:821-831. Davis, S. M., L. H. Gunderson, W. A. Park J. R. Richardson, and J. E. Mattson. 1994. Landscape dimension, composition, and functi on in a changing Everglades ecosystem. Pages 419-444 in S. M. Davis and J. C. Ogden, edito rs. Everglades: the ecosystem and its restoration. St. Lucie Pr ess, Delray Beach, FL. DeAngelis, D. L., L. J. Gross, M. A. Huston, W. F. Wolff, D. Fleming, Martin, E. J. Comiskey, and S. M. Sylvester. 1998. Landscape modeli ng for Everglades restoration. Ecosystems 1998:64-75. Doan, T. M. 2003. Which methods are most effective for surveyi ng rain forest herpetofauna. Journal of Herpetology 37:72-81. Donnelly, M. A. 1989. Demographic effects of reproductive resource supplementation in a territorial frog, Dendrobates pumilio Ecological Monographs 59:207-221. Donnelly, M. A., C. J. Farrell, M. J. Baber, and J. L. Glenn. 2001. The amphibians and reptiles of the Kissimmee River. I. Patterns of abunda nce and occurrence in altered floodplain habitats. Herpetological Natural History 8:161-170.

PAGE 104

104 Donnelly, M. A., C. Guyer, J. E. Juterbock, and R. A. Alford. 1994. Techniques for marking amphibians. in W. R. Heyer, M. A. Donnelly, R. W. McDiarmid, L. C. Hayek, and M. S. Foster, editors. Measuring and monitoring biol ogical diversity: st andard methods for amphibians. Smithsonian Institut ion Press, Washington, D.C. Duellman, W. E., and A. Schwartz. 1958. Amphibia ns and reptiles of sout hern Florida. Bulletin of the Florida State Museum 3:179-324. Duellman, W. E., and L. Trueb. 1986. Biology of Amphibians. Johns Hopkins University Press, Baltimore, MD, USA. Duever, M. J. 2005. Big Cypress regional ecosy stem conceptual model. Wetlands 25:843-853. Duever, M. J., J. E. Carlson, J. F. Meeder, L. C. Duever, L. H. Gunderson, L. A. Riopelle, T. R. Alexander, R. L. Myers, and D. P. Spangl er. 1986. The Big Cypress National Preserve. Research Report No. 8, National Au dubon Society, New York, NY, USA. Duever, M. J., J. E. Carlson, and L. A. Riopelle 1981. Off-road vehicles and their impacts in the Big Cypress National Preserve. South Fl orida Research Center Report T-614. Duever, M. J., L. A. Riopelle, and J. M. McCollom. 1987. Long te rm recovery of experimental off-road vehicle impacts and abandoned old tr ails in the Big Cypre ss National Preserve. South Florida Research Ce nter Report SFRC-86/09. Galatowitsch, S. M., D. C. Whited, and J. R. Tester. 1999. Developmen t of community metrics to evaluate recovery of Mi nnesota wetlands. Journal of Aquatic Ecosystem Stress and Recovery 6:217-234. Gerlanc, N. M., and G. A. Kaufman. 2005. Ha bitat origin and change s in water chemistry influence development of Western Chorus Frogs. Journal of Herpetology 39:254-265. Golay, N., and H. Durrer. 1994. Inflammati on due to toe-clipping in natterjack toads ( Bufo calamita ). Amphibia-Reptilia 15:81-96. Grafe, T. U., S. K. Kaminsky, J. H. Bitz, H. Lussow, and K. E. Linsenmair. 2004. Demographic dynamics of the afro-tropical pig-nosed frog, Hemisus marmoratus : effects of climate and predation on survival and r ecruitment. Oecologia 141:40-46. Green, D. M. 2003. The ecology of extinction: popul ation fluctuation and d ecline in amphibians. Biological Conser vation 111:331-343. Guyer, C., K. E. Nicholson, and S. Baucom. 1996. Effects of tracked vehi cles on gopher tortoises ( Gopherus polyphemus ) at Fort Benning Military Installa tion, Georgia. Ge orgia Journal of Science 54:195-203. Hammer, A. J., J. A. Makings S. J. Lane, and M. J. Mahony. 2004. Amphibian decline and fertilizers used on agricultural land in south-eastern Australia. Agriculture, Ecosystems and Environment 102:299-305.

PAGE 105

105 Henke, S. E. 1998. The effect of multiple ite ms and abundance on the efficiency of human searchers. Journal of Herpetology 32:112-115. Heyer, W. R., M. A. Donnelly, R. W. McDiar mid, L. C. Hayek, and M. S. Foster. 1994. Measuring and monitoring biol ogical diversity: standard methods for amphibians. Smithsonian Institution Press, Washington, DC. Hooge, P. N., and B. Eichenlaub. 1997. Animal movement extension to ArcView. Alaska Science Center Biological Science Office, U.S Geological Survey, Anchorage, AK, USA Ireland, D., N. Osbourne, and M. Berril. 2003. Marking medium-to-larg e-sized anurans with passive integrated trans ponder (PIT) tags. Herpetolog ical Review 34:218-220. Janis, M. W., and J. D. Clark. 2002. Responses of Florida panthers to recreational deer and hog hunting. Journal of Wildlife Management 66:839-848. Jolly, G. M. 1965. Explicit estimates from cap ture-recapture data with both death and immigration-Stochastic mode l. Biometrika 52:225-247. Kaplan, H. M. 1959. Electric tattooing for perm anent identification of frogs. Herpetologica 15:126. Kremen, C. 1992. Assessing the i ndicator properties of species assemblages for natural areas monitoring. Ecological Applications 2:203-217. Kushlan, J. A. 1979. Design and management of continental wildlife rese rves: lessons from the Everglades. Biological Conservation 15:281-290. Landres, P. B., J. Verner, and J. W. Thomas. 1988. Ecological uses of vertebrate indicator species: a critique. Conservation Biology 2:316-328. Lebreton, J.-D., K. P. Burnham, J. Clobert, a nd D. R. Anderson. 1992. Modeling survival and testing biological hypotheses using marked an imals: a unified appro ach with case studies. Ecological Monographs 62:67-118. Lemckert, F. 1996. Effects of toe clipping on th e survival and behaviour of the Australian frog Crinia signifera Amphibia-Reptilia 17:287-290. Lillywhite, H. B. 2006. Water relations of tetr apod integument. The Journal of Experimental Biology 209:202-226. Luddecke, H., and A. Amezquita. 1999. Assessme nt of disc clipping on the survival and behavior of the Andean frog Hyla labialis Copeia 1999:824-830. MacKenzie, D. I., J. D. Nichol s, J. E. Hines, M. G. Knut son, and A. B. Franklin. 2003. Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84:2200-2207.

PAGE 106

106 MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupanc y rates when detection proba bilities are less than one. Ecology 83:2248-2256. MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines. 2006. Occupancy estimation and modeling: inferr ing patterns and dynamics of species occurrence. Academic Pre ss, Burlington, MA, USA. MacKenzie, D. I., J. D. Nichol s, N. Sutton, K. Kawanishi, and L. L. Bailey. 2005. Improving inferences in population studies of rare spec ies that are detected imperfectly. Ecology 86:1101-1113. Madden, M., D. Jones, and L. Vilchek. 1999. Photointerpretation key for the Everglades vegetation classification syst em. Photogrammetric Engineer ing & Remote Sensing 65:171177. Marsh, H., and D. F. Sinclair. 1989. Correcting for visibil ity bias in strip tran sect aerial surveys of aquatic fauna. Journal of Wildlife Management 53:1017-1024. May, R. M. 2004. Ethics a nd amphibians. Nature 431:403. McCarthy, M. A., and K. M. Pa rris. 2004. Clarifying the effect of toe clipping on frogs with Bayesian statistics. Journal of Applied Ecology 41:780-786. McCormick, P. V., and R. J. Stevenson. 1998. Pe riphyton as a tool for ecological assessment and management in the Florida Evergl ades. Journal of Phycology 34:726-733. McPherson, B. F. 1974. The Bi g Cypress Swamp. Pages 8-17 in P. J. Gleason, editor. Environments of South Florid a: present and past. Miami Ge ological Survey, Miami, FL. Meshaka, W. E., W. F. Loftus, and T. Steiner. 2000. The herpetofauna of Everglades National Park. Florida Scientist 63:84-103. Moulton, C. A., W. J. Fleming, and B. R. Ne rney. 1996. The use of PVC pipes to capture hylid frogs. Herpetological Review 27:186-187. Nauwelaerts, S., J. Coeck, and P. Aerts. 2000. Visible implant elastomers as a method for marking adult anurans. Herp etological Re view 31:154-155. Nichols, J. D., J. E. Hines, J.-D. Lebreton, a nd R. Pradel. 2000. Estimation of contributions to population growth: a reverse-time capture-r ecapture approach. Ecology 81:3362-3376. Nichols, J. D., J. E. Hines, J. R. Sauer, F. W. Fallon, J. E. Fallon, and P. J. Heglund. 2000. A double-observer approach for estimating dete ction probability and abundance from point counts. Auk 117:393-408.

PAGE 107

107 Nichols, J. D., R. E. Tom linson, and G. Waggerman. 1986. Estimating nest detection probabilities for white-winged dove nest tr ansects in Tamaulipas, Mexico. The Auk 103:825-828. NPS (National Park Service). 2000. Final recr eational off-road vehicle management plan supplemental environmental impact statement. U.S. National Park Service, Big Cypress National Preserve, Ochopee, FL. Ogden, J. C., S. M. Davis, K. J. Jacobs, T. Ba rnes, and H. E. Fling. 2005. The use of conceptual ecological models to guide ecosystem restor ation in south Florida. Wetlands 25:795-809. Parris, K. M., and M. A. McCarthy. 2001. Identi fying effects of toe clip ping in anuran return rates: the importance of statistical power. Am phibia-Reptilia 22:275-289. Pechmann, J. H. K., and H. M. Wilbur. 1994. Putting declining am phibian populations in perspective: natural fluctuations and human impacts. Herpetologica 50:65-84. Pierce, B. A., and K. J. Gutzwiller. 2004. Audito ry sampling of frogs: Detection efficiency in relation to survey duration. J ournal of Herpetology 38:495-500. Pollock, K. H., and W. L. Kendall. 1987. Visibility bi as in aerial surveys: a review of estimation procedures. Journal of W ildlife Management 51:502-510. Pollock, K. H., D. L. Solomon, and D. S. Robson 1974. Tests for mortality and recruitmnent in a K -sample tag-recapture experi ment. Biometrics 30:77-87. Pradel, R. 1996. Utilization of capture-mark -recapture for the study of recruitment and population growth rate Biometrics 52. Rice, K. G., J. H. Waddle, M. E. Crockett, B. M. Jeffery, and H. F. Percival. 2004. Herpetofaunal Inventories of the National Parks of South Florida and the Caribbean: Volume 1. Everglades National Park. U. S. Geological Survey, Open-File Report 2004-1065, Fort Lauderdale, FL, USA. Rocha, C. F. D., M. Van Sluys, M. A. S. Alve s, H. G. Bergallo, and D. Vrcibradic. 2001. Estimates of forest floor litter frog comm unities: a comparison of methods. Austral Ecology 26:14-21. Rodda, G. H., E. W. Campbell, T. H. Fritts, and C. S. Clark. 2005. The predictive power of visual searching. Herpet ological Review 36:259-264. Schmidt, B. R. 2003. Count data, detection probabilities, and the demography, dynamics, distribution, and decline of amphibi ans. C. R. Biologies 326:S119-S124. Schmidt, B. R., and J. Pellet. 2005. Relative importance of population processes and habitat characteristics in determining site occupa ncy of two anurans. Journal of Wildlife Management 69:884-893.

PAGE 108

108 Seber, G. A. F. 1965. A note on the multiple recapture census. Biometrika 52:13-22. SFWMD (South Florida Water Ma nagement District). 1992. Su rface water improvement and management plan for the Everglades: s upporting information needs document. South Florida Water Management District, West Palm Beach, FL. Sheridan, C. D., and D. H. Olson. 2003. Amphibi an assemblages in zero-order basins in the Oregon Coast Range. Canadian Journa l of Forest Research 33:1452-1477. Steiner, A. J., and S. P. Leatherman. 1981. Recreational impacts on th e distribution of ghost crabs Ocypode quadrata Fab. Biological Cons ervation 20:111-122. Stuart, S. N., J. S. Chanson, N. A. Cox, B. E. Young, A. S. L. Rodrigues, D. L. Fischman, and R. W. Waller. 2004. Status and trends of amphi bian declines and ex tinctions worldwide. Science 306:1783-1786. Taylor, B., D. Skelly, L. K. Demarchis, M. D. Slade, D. Galusha, and P. M. Rabinowitz. 2005. Proximity to pollution sources and risk of amphibian limb formation. Environmental Health Perspec tives 113:1497-1501. Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques, S. T. Buckland, D. L. Borchers, D. R. Anderson, K. P. Burnham, S. L. Hedley, J. H. Pollard, and J. R. B. Bishop. 2003. Distance 4.1, Release 1. Research Unit for Wildlife Population Assessment, University of St. Andrews, UK. Vitt, L. J., J. P. Caldwell, H. M. Wilbur, and D. C. Smith. 1990. Amphibians as harbingers of decay. BioScience 40:418. Wake, D. B. 1991. Declining amphi bian populations. Science 253:860. Watanabe, S., N. Nakanishi, and M. Izawa. 2005. Seasonal abundance in the floor-dwelling frog fauna on Iriomote Island of the Ryukyu Archip elago, Japan. Journa l of Tropical Ecology 21:85-91. Welch, R., M. Madden, and R. F. Doren. 1999. Mapping the Everglad es. Photogrammetric Engineering & Remote Sensing 65:163-170. Welsh, H. H. J., and L. M. Ollivier. 1998. Stre am amphibians as indicators of ecosystem stress: A case study from California's redwoods Ecological Applications 8:1118-1132. White, G. C. 1992. Program SURVIV. Dept. of Fishery and Wildlife Biology, Colorado State University, Fort Collins, CO White, G. C., and K. P. Burnham. 1999. Progr am MARK: survival estim ation from populations of marked animals. Bird Study 46 supplement:120-138. Williams, B. K., J. D. Nichols, and M. J. C onroy. 2002. Analysis and Management of Wildlife Populations. Academic Press, San Diego, CA.

PAGE 109

109 Wood, K. V., J. D. Nichols, H. F. Percival, and J. E. Hines. 1998. Size-se x variation in survival rates and abundance of pig frogs, Rana grylio in northern Florida wetlands. Journal of Herpetology 32:527-535. Zacharow, M., W. J. Barichivic h, and C. K. Dodd Jr. 2003. Us ing Ground-placed PVC Pipes to Monitor Hylid Treefrogs: Capture Biases Southeastern Na turalist 2:575-590.

PAGE 110

110 BIOGRAPHICAL SKETCH Hardin Waddle was born in Louisville, Kent ucky, in 1972 and lived briefly in Overland Park, Kansas, before his family relocated to A nniston, Alabama, in 1980. Waddle received his high school diploma from the Donoho School in Anniston in 1991. He then attended Auburn University in Auburn, Alabama, where he recei ved his Bachelor of Sc ience degree in wildlife science in 1995. Waddle then work ed as a research technician at Tall Timbers Research Station near Tallahassee, Florida, for two years. From 1997 to 2000 Waddle wa s enrolled at Florida International University in Miami, Florida, wh ere he earned a Master of Science degree in biological sciences. In 2000 Wa ddle began working as a cooperat or with the U.S. Geological Survey on amphibian and reptile projects in the National Parks of south Florida and the Caribbean. In 2002, Waddle enrolled in the Univ ersity of Florida to pursue a Doctor of Philosophy degree in wildlife ecology and conservation.


Permanent Link: http://ufdc.ufl.edu/UFE0016760/00001

Material Information

Title: Use of amphibians as ecosystem indicator species
Physical Description: 110 p.
Language: English
Creator: Waddle, James Hardin ( Dissertant )
Percival, Henry F. ( Thesis advisor )
Rice, Ken ( Reviewer )
Nichols, Jim ( Reviewer )
Mazzotti, Frank ( Thesis advisor )
Lillywhite, Harvey ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2006

Subjects

Subjects / Keywords: Wildlife Ecology and Conservation thesis, Ph. D.   ( local )
Dissertations, Academic -- UF -- Wildlife Ecology and Conservation   ( local )

Notes

Abstract: Amphibians are generally considered suitable as indicator species in a variety of systems. Their biphasic life cycle and semi-permeable skin are two justifications often given for this use of amphibians. In this dissertation, the use of amphibians as indicator species in support of management and restoration in the Everglades of southern Florida was investigated. Methods for monitoring amphibians and specific uses of amphibians as indicator species were evaluated. Techniques to reduce observer bias in visual encounter surveys, a common method of sampling amphibians for monitoring purposes, were tested. Both double observer and distance based methods were shown to have significant bias in enumerating a known population of artificial frogs. Toe-clipping, a standard method for individually marking frogs was also studied on two treefrog species in south Florida. Toe-clipping was found to have a slight negative effect on survival in one species, but not the other. These studies demonstrate the importance of carefully choosing and evaluating monitoring methods to appropriately address questions concerning amphibian populations. The occupancy of four anuran species was estimated in relation to off-road vehicle (ORV) use in Big Cypress National Preserve to determine if amphibians are useful as indicators of this form of anthropogenic disturbance. Results confirmed that ORV use was a significant factor in the site occupancy of the four species of ground-dwelling anurans studied. In another study, the survival and recruitment of Green Treefrogs were estimated in relation to hydrology and habitat to better understand how frogs might respond to hydrologic changes proposed under the Comprehensive Everglades Restoration Plan. Water depth and hydrologic season were both important factors in survival and recruitment, and population growth rates varied with seasons. This research concludes that amphibians meet the criteria for ecosystem indicator species in south Florida. They are abundant, may be efficiently sampled, and have been demonstrated to respond in a predictable way to stresses to the system. Monitoring of amphibians is a useful tool for determining the success of ecosystem restoration and management in the Everglades of south Florida.
Subject: amphibian, everglades, hydrology, indicator, mark, modeling, monitoring, occupancy, recapture, restoration, species
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 110 pages.
General Note: Includes vita.
Thesis: Thesis (Ph. D.)--University of Florida, 2006.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016760:00001

Permanent Link: http://ufdc.ufl.edu/UFE0016760/00001

Material Information

Title: Use of amphibians as ecosystem indicator species
Physical Description: 110 p.
Language: English
Creator: Waddle, James Hardin ( Dissertant )
Percival, Henry F. ( Thesis advisor )
Rice, Ken ( Reviewer )
Nichols, Jim ( Reviewer )
Mazzotti, Frank ( Thesis advisor )
Lillywhite, Harvey ( Reviewer )
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2006

Subjects

Subjects / Keywords: Wildlife Ecology and Conservation thesis, Ph. D.   ( local )
Dissertations, Academic -- UF -- Wildlife Ecology and Conservation   ( local )

Notes

Abstract: Amphibians are generally considered suitable as indicator species in a variety of systems. Their biphasic life cycle and semi-permeable skin are two justifications often given for this use of amphibians. In this dissertation, the use of amphibians as indicator species in support of management and restoration in the Everglades of southern Florida was investigated. Methods for monitoring amphibians and specific uses of amphibians as indicator species were evaluated. Techniques to reduce observer bias in visual encounter surveys, a common method of sampling amphibians for monitoring purposes, were tested. Both double observer and distance based methods were shown to have significant bias in enumerating a known population of artificial frogs. Toe-clipping, a standard method for individually marking frogs was also studied on two treefrog species in south Florida. Toe-clipping was found to have a slight negative effect on survival in one species, but not the other. These studies demonstrate the importance of carefully choosing and evaluating monitoring methods to appropriately address questions concerning amphibian populations. The occupancy of four anuran species was estimated in relation to off-road vehicle (ORV) use in Big Cypress National Preserve to determine if amphibians are useful as indicators of this form of anthropogenic disturbance. Results confirmed that ORV use was a significant factor in the site occupancy of the four species of ground-dwelling anurans studied. In another study, the survival and recruitment of Green Treefrogs were estimated in relation to hydrology and habitat to better understand how frogs might respond to hydrologic changes proposed under the Comprehensive Everglades Restoration Plan. Water depth and hydrologic season were both important factors in survival and recruitment, and population growth rates varied with seasons. This research concludes that amphibians meet the criteria for ecosystem indicator species in south Florida. They are abundant, may be efficiently sampled, and have been demonstrated to respond in a predictable way to stresses to the system. Monitoring of amphibians is a useful tool for determining the success of ecosystem restoration and management in the Everglades of south Florida.
Subject: amphibian, everglades, hydrology, indicator, mark, modeling, monitoring, occupancy, recapture, restoration, species
General Note: Title from title page of source document.
General Note: Document formatted into pages; contains 110 pages.
General Note: Includes vita.
Thesis: Thesis (Ph. D.)--University of Florida, 2006.
Bibliography: Includes bibliographical references.
General Note: Text (Electronic thesis) in PDF format.

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0016760:00001


This item has the following downloads:


Table of Contents
    Title Page
        Page 1
        Page 2
    Dedication
        Page 3
    Acknowledgement
        Page 4
    Table of Contents
        Page 5
        Page 6
    List of Tables
        Page 7
        Page 8
    List of Figures
        Page 9
        Page 10
    Abstract
        Page 11
        Page 12
    Introduction
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
    Sampling methods for amphibian monitoring
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
    Using site occupancy modeling to determine the effect of off-road vehicle use on ground-dwelling anurans
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
    The effect of toe-clipping on two specie of treefrogs
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
    Influence of hyrdrology on survival and recruitment of green treefrogs
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
    Conclusion
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
    References
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
    Biographical sketch
        Page 110
Full Text





USE OF AMPHIBIANS AS ECOSYSTEM INDICATOR SPECIES


By

JAMES HARDIN WADDLE

















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2006

































Copyright 2006

by

James Hardin Waddle
































This dissertation is dedicated to my parents, Chris and Sherrell Waddle, who have always
supported me and encouraged me to do what I love, and to Amanda, the love of my life.









ACKNOWLEDGMENTS

I would first like to thank the members of my committee. Ken Rice has been a great

mentor and friend over the last six years. His guidance has helped prepare me to be a better

scientist. Franklin Percival has truly taken me "under his wing" and taught me about philosophy.

Jim Nichols was always available for help when I needed it and has consistently provided me

with insightful feedback on my work. Frank Mazzotti was always supportive of me during my

time as a student. Harvey Lillywhite has been a valuable member of the committee, and forced

me to think more about what I think I know.

I would also like to thank the people who assisted me in the field. Brian Jeffery, Andy

Maskell, Chris Bugbee, Meghan Riley, and Debbie Kramp all spent many long hours with me

slogging through swamps. Many members of the Ft. Lauderdale REC Mazzotti and Rice lab and

the Florida Alligator and Amphibian Research Team (FAART) volunteered for the sampling

experiment in Chapter 2. Other volunteers from the Big Cypress National Preserve Student

Conservation Association provided valuable help. All of the FAART people have also provided

useful feedback on my work and great friendship.

The staff of Big Cypress National Preserve was very helpful in all stages of this work. Big

Cypress provided an office and housing for this project, and granted access and permits for the

research. Deb Jansen and Ron Clark engaged me in many good conversations about

management and wildlife in the Preserve. Jim Burch shared plant knowledge, Bob Sobczack

helped with hydrologic questions, and Frank Partridge provided maps and GIS data. Jim Snyder

of the USGS allowed me to share equipment and space and helped watch out for me.

Finally I would like to thank my family for all of their love and support. My new wife,

Amanda is always my biggest supporter. My parents, and my sister Virginia and her husband

Brian Lott all deserve thanks for supporting me through my graduate career.













TABLE OF CONTENTS



A C K N O W L E D G M E N T S ..............................................................................................................4

LIST OF TABLES ................. ......................... .. ........... ........................................ 7

L IST O F F IG U R E S ......................................................................... ................................... . 9

A B S T R A C T ............... ......................................................... ................................................. 1 1

CHAPTER

1 INTRODUCTION .................................. .. ........... ..................................... 13

C characteristics of Indicator Species........................................ ........................ ................ 13
Suitability of Amphibians as Indicator Species....................................................14
O outline of the D issertation ................................................... ............................................. 15

2 SAMPLING METHODS FOR AMPHIBIAN MONITORING......................................19

In tro d u c tio n ............................................................................................................................. 1 9
M e th o d s ........................................................................................................ .................... 2 0
Sampling Experiment ........................ .. ........... ............................. 20
D ouble O observer A analysis .. .................................................................. ................ 22
D istan ce A n aly sis ............................................................................................................ 2 3
R esu lts ............................................................................................................ 2 3
D ou b le O b serve er.............................................................................................................. 2 4
D istan ce A n aly sis ............................................................................................................ 2 4
D iscu ssio n .............................................................................................................. ........ .. 2 5

3 USING SITE OCCUPANCY MODELING TO DETERMINE THE EFFECT OF OFF-
ROAD VEHICLE USE ON GROUND-DWELLING ANURANS..................................41

In tro du ctio n ............................................................................................................ ........ .. 4 1
M e th o d s ........................................................................................................ .................... 4 2
Stu dy A rea ...................................................................................................... ....... .. 4 2
S am p lin g ......................................................................................................... ........ .. 4 3
D ata A n a ly sis ................................................................................................................. .. 4 5
R esu lts ............................................................................................................ 4 5
D iscu ssio n .............................................................................................................. ........ .. 4 6

4 THE EFFECT OF TOE-CLIPPING ON TWO SPECIES OF TREEFROGS .......................63









Introduction ................................................. ............................. 63
M e th o d s ........................................................................................................ ..................... 6 4
Study Site ............................................... .............................. 64
C capture R capture .................................................................................................... 65
Survival A analysis ............................................................................... ...................... 66
Results .................................................... .............................. 66
Discussion ................................................. ............................. 68

5 INFLUENCE OF HYDROLOGY ON SURVIVAL AND RECRUITMENT OF
GREEN TREEFR O G S .................................................................................................... 77

M e th o d s ........................................................................................................ ..................... 7 8
Results .................................................... .............................. 82
Discussion ................................................. ............................. 83

6 CONCLUSION ............................................................................ 97

Introduction ............................................... ................. .....................97
C h aracteristics of In dicators ...................................................................................................97
Abundant and Efficient to Sam ple ...............................................................................97
Sensitive to Stresses on the System ...........................................................................98
Responses to Stress Should Be Anticipatory ................................................................99
Integrate a Response across the W hole System ...........................................................99
Known Response to Anthropogenic Stresses and Natural Disturbances ........................99
C o n c lu sio n ................................................................................................... ..................... 10 0

L IS T O F R E F E R E N C E S .............................................................................................................102

B IO G R A PH IC A L SK E T C H ....................................................................................................... 110























6









LIST OF TABLES


Table page

2-1 Number of artificial frogs observed by each team in prairie and pineland habitats.........29

2-2 Actual detection rates of a known population of artificial frogs within 30 cm, within
50 cm, beyond 50 cm, and overall for teams searching in prairie and pineland habitat....30

2-3 Model selection for the three double-observer models analyzed in Program SURVIV
with the model log-likelihood, number of parameters (K), quasi-likelihood Akaike's
information criterion adjusted for small sample sizes (QAIC) ....................................31

2-4 Estimates from program SURVIV of individual detection probabilities of artificial
frogs for each observer in prairie and pineland habitat with standard error (S.E.) and
95% confidence intervals (C.I.) taken from the double observer method...................... 32

2-5 Abundance estimates of artificial treefrogs for each team in prairie and pineland
habitat with standard error (S.E.) and upper and lower 95% confidence intervals (C.
I.) based on Chaos's estimatorfrom the double observer data................ .................. 33

2-6 Model selection for detection functions in Program DISTANCE showing number of
parameters (K), Akaike's Information Criterion (AIC), the difference between each
AIC and the minimum (Delta AIC), and the AIC Weight of each model ...................... 34

2-7 Abundance estimates of artificial frogs by team for transects in the prairie and
pineland habitats with coefficient of variation (CV) and 95% confidence interval (CI)
from the distance sam pling approach ........................................................... ................ 35

3-1 Combinations of the 3 site covariates and 4 sampling covariates that were used in the
occupancy analysis for each species. Each set of site covariates was modeled along
with each set of sampling covariates for a total of 80 unique models for each species.....50

3-2 Number of sampling sites and total number of site visits by habitat...............................51

3-3 Number of detections by species, and proportion of sites at which a detection
occurred (naive occupancy) during amphibian surveys across all habitat types .............52

3-4 Model selection results for the oak toad (Bufo quercicus), including Akaike's
Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and
Anderson 1998) for all models with any AIC weight ........................... ...................... 53

3-5 Model selection results for the southern toad (Bufo terrestris), including Akaike's
Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and
Anderson 1998) for all models with any AIC weight................................... ................ 54









3-6 Model selection results for the greenhouse frog (Eleutherodactylusplanirostris),
including Akaike's Information Criterion (AIC) and the delta AIC and AIC weight
(Burnham and Anderson 1998) for all models with any AIC weight...............................56

3-7 Model selection results for the eastern narrow-mouthed toad (Gastrophryne
carolinensis), including Akaike's Information Criterion (AIC) and the delta AIC and
AIC weight (Burnham and Anderson 1998) for all models with any AIC weight............57

3-8 Sums of Akaike's Information Criterion (AIC) weights for all models including the
ORV index, habitat type, or hydrologic index covariates for each of the four focal
a n u ra n sp e c ie s. ................................................................................................................. .. 5 9

3-9 Beta estimates, standard errors (S.E.), and lower and upper 95% confidence intervals
(C.I.) for the ORV use index covariate from the best model for each of the four focal
a n u ra n sp e c ie s. ................................................................................................................. .. 6 0

4-1 List of 23 models analyzed in Program MARK for captures of both Green Treefrogs
and Squirrel Treefrogs in Big Cypress National Preserve during 2004-2005. ..............70

4-2 The number of green treefrogs and squirrel treefrogs marked by removing 2, 3, or 4
toes in Big Cypress National Preserve, Collier County, FL, Nov. 2004-June 2005 and
the return rate (proportion of marked individuals recaptured at least once)...................71

4-3 Model selection table for Cormack-Jolly-Seber open population mark-recapture
model of Green Treefrogs including Quasi-likelihood Akaike's Information Criterion
for sm all sam ple sizes (Q A IC o) .................................................................... ................ 72

4-4 Estimates, standard error (SE), and the 95% confidence interval of the beta values for
the toe-clip effect on apparent survival and recapture probability. ..............................73

4-5 Model selection table for Cormack-Jolly-Seber open population mark-recapture
model of Squirrel Treefrogs including Quasi-likelihood Akaike's Information
Criterion for sm all sam ple sizes (Q A ICc)..................................................... ................ 74

5-1 Dates of each sample of the PVC pipe refugia and the season to which each sample
w as assigned for m ark-recapture analysis..................................................... ................ 86

5-2 Model selection results for all models analyzed in Program MARK. Model describes
the covariates and groups associated with survival, capture probability, and seniority. ...87

5-3 Estimates of the population growth rate, the contribution of survival and the
contribution of recruitment to growth for each sampling interval...............................88









LIST OF FIGURES


Figure page

2-1 Histograms of the frequency of perpendicular distances in cm of actual artificial frog
placements along the 2 transect lines in the prairie and pineland habitat.......................36

2-2 Histograms of the frequency of detections of artificial frogs by each team at
perpendicular distances in cm along the 2 transect lines in the prairie habitat...............37

2-3 Histograms of the frequency of detections of artificial frogs by each team at
perpendicular distances in cm along the 2 transect lines in the pineland habitat. ............38

2-4 Histograms of the frequency of detections of artificial frogs at perpendicular
distances along the transect lines for teams in the prairie habitat and pineland habitat. ...39

2-5 Abundance estimates with standard error for each team from prairie (A) and pineland
habitat (B) using the double observer and distance methods........................................40

3-1 An aerial photograph depicting off-road vehicle damage in marl prairie habitat in Big
Cypress N national Preserve. ................................................. .............. ................ 61

3-2 Map of amphibian occupancy sampling locations within BCNP during 2002-2003 ........62

4-1 Numbers of individuals of green treefrogs and squirrel treefrogs captured and
released in each of the three toe removal groups during the first 7 sampling
o cca sio n s ......................................................................................................... ........ .. 7 5

4-2 Apparent survival and 95% confidence interval of green treefrogs and squirrel
treefrogs in each toe removal group category across the first 6 monthly sampling
in terv als........................................................................................................... ....... .. 7 6

5-1 Mean water depth across the 3 sampling locations in BCNP in the cypress strand,
broadleaf marsh, and prairie habitats from April 2004-August 2005............................. 89

5-2 Number of captures of Green Treefrogs during each of the 25 samples from April
2004 to August 2005 and mean water depth (in cm) of marsh plots across all 3 sites .....90

5-3 Mean snout-to-urostyle (SUL) in mm of Green Treefrogs captured in all habitats at
the three sites at each sampling occasion in BCNP. ..................................... ................ 91

5-4 Number of captures by sample of Green Treefrogs in cypress, marsh, and prairie
habitats at the three sites in B C N P ................................................................ ................ 92

5-5 Estimates with 95% confidence intervals of apparent survival of Green Treefrogs for
each sampling interval for cypress, marsh, and prairie habitat in BCNP .......................93









5-6 Estimates with 95% confidence intervals of seniority of Green Treefrogs for each
sampling interval for cypress, marsh, and prairie habitat in BCNP...............................94

5-7 Estimates with 95% confidence intervals of capture probability of Green Treefrogs at
each sampling occasion for cypress, marsh, and prairie habitat in BCNP. .....................95

5-8 Derived estimates of population growth of Green Treefrogs for each sampling
interval for cypress, marsh, and prairie habitat in BCNP. ...................... ..................... 96









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

USE OF AMPHIBIANS AS ECOSYSTEM INDICATOR SPECIES

By

James Hardin Waddle

December 2006

Chair: H. Franklin Percival
Cochair: Frank J. Mazzotti
Major Department: Wildlife Ecology and Conservation

Amphibians are generally considered suitable as indicator species in a variety of systems.

Their biphasic life cycle and semi-permeable skin are two justifications often given for this use

of amphibians. In this dissertation, the use of amphibians as indicator species in support of

management and restoration in the Everglades of southern Florida was investigated. Methods for

monitoring amphibians and specific uses of amphibians as indicator species were evaluated.

Techniques to reduce observer bias in visual encounter surveys, a common method of

sampling amphibians for monitoring purposes, were tested. Both double observer and distance

based methods were shown to have significant bias in enumerating a known population of

artificial frogs. Toe-clipping, a standard method for individually marking frogs was also studied

on two treefrog species in south Florida. Toe-clipping was found to have a slight negative effect

on survival in one species, but not the other. These studies demonstrate the importance of

carefully choosing and evaluating monitoring methods to appropriately address questions

concerning amphibian populations.

The occupancy of four anuran species was estimated in relation to off-road vehicle (ORV)

use in Big Cypress National Preserve to determine if amphibians are useful as indicators of this

form of anthropogenic disturbance. Results confirmed that ORV use was a significant factor in









the site occupancy of the four species of ground-dwelling anurans studied. In another study, the

survival and recruitment of Green Treefrogs were estimated in relation to hydrology and habitat

to better understand how frogs might respond to hydrologic changes proposed under the

Comprehensive Everglades Restoration Plan. Water depth and hydrologic season were both

important factors in survival and recruitment, and population growth rates varied with seasons.

This research concludes that amphibians meet the criteria for ecosystem indicator species

in south Florida. They are abundant, may be efficiently sampled, and have been demonstrated to

respond in a predictable way to stresses to the system. Monitoring of amphibians is a useful tool

for determining the success of ecosystem restoration and management in the Everglades of south

Florida.












CHAPTER 1
INTRODUCTION

Ecological indicators can have many purposes, including being used to assess the condition

of the environment or monitor trends in condition over time (Cairns et al. 1993, Dale and Beyeler

2001). Some species suitable for monitoring trends in condition over time may be useful as

indicators of restoration success in ecosystems in which restoration activities are occurring.

Amphibians are widely considered to be useful as indicator species (Welsh and Ollivier 1998,

Sheridan and Olson 2003), but little direct evidence has been gathered that evaluates the

usefulness of amphibians for this role. There are several reasons why amphibians may be

excellent indicators, but there are also limitations to their use. In this chapter, I discuss the

characteristics of good indicators and whether amphibians display these characteristics. In

addition, I will outline important considerations for evaluating amphibians as indicators for any

particular system. Finally, I will introduce my research in the Everglades of southern Florida

and outline the rest of this dissertation.

Characteristics of Indicator Species

Dale and Beyeler (2001) discuss several general characteristics of useful indicator species.

Indicators should be easily sampled, sensitive to stresses on the system, and respond to stress in a

predictable manner. These responses should be anticipatory of an impending change in the

whole system, and they should predict changes that can be averted by management. Indicators

should provide information regarding changes to the whole system rather than a few habitats or

locations, and have a known response to anthropogenic stresses and natural disturbances, and the

response should have low variability (Dale and Beyeler 2001). Indicators for restoration success

need to be predictable enough to determine whether they are responding to changes due to









management actions or just natural fluctuations in the system. Finally, indicator species should

also be abundant or cost-effective to sample.

Using vertebrates as indicators of habitat quality requires special consideration (Landres et

al. 1988). Using abundance of vertebrate species requires robust estimation techniques that

explicitly deal with imperfect detection (MacKenzie et al. 2002, Williams et al. 2002). Also, the

efficacy of the indicator species as an index for the abundance of other species must be

determined. Using one or more species as indicators of habitat quality for other species is valid

only after research validating this approach has been conducted (Landres et al. 1988). Managing

for one indicator may ignore ecological processes not important to the indicator but vital to other

species (Kushlan 1979).

Suitability of Amphibians as Indicator Species

Amphibian species or communities have been touted as useful indicators in many

situations recently (Welsh and Ollivier 1998, Galatowitsch et al. 1999, Collins and Storfer 2003,

Sheridan and Olson 2003, Hammer et al. 2004). Some studies use amphibians as indicators of

environmental contamination or pollution (Hammer et al. 2004). Others attempt to use the

species assemblage (Sheridan and Olson 2003) or the abundance of populations (Welsh and

Ollivier 1998, Campbell et al. 2005) as indicators of ecosystem health or habitat quality.

Amphibians have several characteristics that make them useful as indicator species. They

are often locally abundant (Rocha et al. 2001, Watanabe et al. 2005) and may be sampled with

low-cost standard methods (Heyer et al. 1994, Pierce and Gutzwiller 2004). Because of their

permeable skin and biphasic life cycle amphibians are likely sensitive to environmental stress

(Vitt et al. 1990, Wake 1991, Blaustein 1994, Blaustein et al. 1994), but there is some debate

about whether this sensitivity is consistent and predictable (Pechmann and Wilbur 1994). It is









imperative that any study using amphibians as indicators of ecosystem stress demonstrates a

direct causal link between the stress and its effect on the indicator species.

More research is also needed to determine if the response of amphibians to a particular

stress is indicative of the management action taken. In some situations amphibians may serve as

"canaries" (Blaustein 1994), but not necessarily in all cases (Pechmann and Wilbur 1994). The

responsiveness of amphibians as indicator species will depend on the type of stress and the

particular amphibians in the system.

It is likely that amphibians will be good indicators of changes to the whole ecosystem

because they are sensitive to changes in the aquatic and terrestrial environments. The aquatic

environment is required for reproduction in most species (Duellman and Trueb 1986) and the

permeable skin of amphibians makes them sensitive to water quality and UV radiation in the egg

and larval as well as adult life stages (Gerlanc and Kaufman 2005, Taylor et al. 2005). Many

amphibian species spend much of their life in terrestrial environments for activities like feeding

and dispersal. Amphibians should also respond to changes in the terrestrial environment that

would affect water relations through their integument with behavioral responses (e.g., shifting

activity periods or moving to different microhabitats) or less frequently with phenotypic

responses (e.g., facultative lipid barrier adjustment; Lillywhite 2006).

Outline of the Dissertation

The Everglades ecosystem of southern Florida has been substantially altered over the last

100 years by loss to agriculture and urbanization (South Florida Water Management District

[SFWMD] 1992, Ogden et al. 2005). Compartmentalization of the remaining system has

impeded historic flow patterns and altered the temporal and spatial dynamics of hydrology in the

Everglades (Davis et al. 1994). A large-scale restoration effort, the Comprehensive Everglades

Restoration Plan (CERP), was devised to attempt to restore natural hydrologic regimes to the









remaining Everglades (DeAngelis et al. 1998). Managers charged with decision making for

CERP need species that can serve as indicators of ecosystem restoration success. A necessary

condition before a species may be used as an indicator in a system is a model of how the species

will respond to changes in the system (Kremen 1992, Dale and Beyeler 2001). Some species

(e.g., wading birds; Crozier and Gawlik 2003) have been monitored for many decades; therefore

data exist on pre-alteration conditions. Other species (e.g., periphyton; McCormick and

Stevenson 1998) have been manipulated in experiments to better understand their expected

response to changes that will be imposed during Everglades restoration.

My research for this dissertation focuses on using amphibians as indicator species in

support of management of the > 1,000,000 hectares under restoration in CERP. Although some

higher trophic level species are monitored and modeled to evaluate restoration scenarios

currently (DeAngelis et al. 1998), many of these species do not meet Dale and Beyeler's (2001)

criteria for good indicator species. For instance, the Cape Sable Seaside Sparrow and the Florida

Panther are rare species of conservation interest, but these species may not respond in a

predictable way to changes to the system, and it is unproven that their responses will be

anticipatory of effects on other components of the system.

There is also a need for indicator species that can be used to identify the effects of certain

human activities other than CERP on these natural areas. The use of off-road vehicles (ORV) is

an important management concern in Big Cypress National Preserve (BCNP) (NPS 2000). The

extent to which ORVs impact wildlife in Big Cypress is of great interest to park managers. My

objective was to research the efficacy of amphibians as indicator species in southern Florida to

support information needs for both CERP and for the management of ORVs in BCNP.









Chapter 2 describes a study designed to evaluate bias when using the visual encounter

survey, a standard sampling technique for reptiles and amphibians. Two estimation methods

designed to account for incomplete detectability were applied to count data collected by

observers on a known population of artificial frog models. This study underscores the need for

accounting for incomplete detectability in amphibian monitoring programs and the potential for

bias in count data. Accurate sampling techniques with reliable methods for determining the

precision of estimates is critical when using amphibians as indicator species.

Chapter 3 describes the use of site occupancy estimation (MacKenzie et al. 2002) to model

the effects of ORV use on the distribution of four anuran species in BCNP. Site occupancy

modeling determined that the index of ORV use created for this study was an important factor in

the occupancy of ground-dwelling frog species. Only one species was positively associated with

the ORV index, while the other three were less likely to occupy sites with higher ORV use. A

monitoring program designed to use these amphibians as indicators would be useful for

evaluating the continuing impact of ORVs in Big Cypress.

Chapters 4 and 5 describe the use of hylid treefrogs as indicators of restoration success in

the Everglades. Chapter 4 examines the effects of toe-clipping as a marking technique for

individual treefrogs. Uniquely marking individuals is necessary for estimation of survival and

movement rates, but the marking method must be validated. Chapter 5 describes a study to

examine survival and recruitment rates of green treefrogs in relation to hydrology and habitat.

This information is vital for building a model of how these potential indicator species will

respond to the hydrologic changes of CERP.

Finally Chapter 6 discusses the overall usefulness of amphibians as ecosystem indicator

species in southern Florida. I will make recommendations for monitoring and analysis









techniques of amphibian populations that are appropriate at particular scales. I will summarize

the extent to which amphibians are useful as indicator species in the Everglades ecosystem.









CHAPTER 2
SAMPLING METHODS FOR AMPHIBIAN MONITORING

Introduction

The visual encounter survey (VES) is widely used as a sampling method for reptiles and

amphibians (Crump and Scott 1994, Doan 2003). In this method, one or more observers search a

defined area for animals for a specified amount of time. Usually the number of individuals of a

species counted is standardized by time or area searched (i.e., effort) to determine the relative

abundance of the species. Relative abundance among sites may only be compared under the

restrictive assumptions of an index to actual abundance. The primary assumptions are that

individuals of a species have a constant probability of detection across time and space (e.g.,

different seasons or habitats) and that different observers all have the same probability of

detecting species of interest (Crump and Scott 1994, Williams et al. 2002). These assumptions

may be violated with improper study design or by uncontrollable factors such as weather. Some

authors have demonstrated that the assumptions underlying the use of VES data as a measure of

relative abundance of amphibians and reptiles are unlikely to hold in actual sampling (Henke

1998, Rodda et al. 2005).

The major obstacle to using this index approach is heterogeneity in detection probability

(p) of individuals of a given species and among different observers (Pollock and Kendall 1987,

Williams et al. 2002). It is possible that p may change in relation to sampling conditions,

individual behavior, or across habitat types. Further, different observers may be more or less

adept at finding a species and may therefore record different proportions of individuals (i.e.,

perception bias; Marsh and Sinclair 1989). Sampling methods based on distance sampling have

been developed to deal with heterogeneity in p across space and time to produce robust estimates

of density (Burnham et al. 1980, Buckland et al. 2001). Other methods involving multiple









observers have been developed to account for observer bias (Cook and Jacobson 1979, Pollock

and Kendall 1987, Marsh and Sinclair 1989, Nichols et al. 2000), and recent advances in distance

sampling of line transects have combined information from multiple observers to improve

density estimates where detection at the line is not always perfect (Buckland et al. 2004).

I evaluated the use of double observer and distance sampling approaches using a

population of artificial frogs with known abundance. This technique lacks the realism of actual

frogs, but provides a better evaluation of the potential bias associated with distance sampling and

double observer sampling separately and together. The objective of this study was to evaluate

sampling approaches that incorporate estimates of detection probability over line transects

sampled similarly to the standard VES. I hypothesized that both methods would yield similar

results and that the 95% confidence intervals of the estimates from both methods would include

the true abundance of frogs. Strengths and weaknesses of the sampling methods will be

discussed.

Methods

Sampling Experiment

To test the efficacy of the double observer and distance methods, a pair of 50-m transects

in 2 habitats (4 transects total) were established in Big Cypress National Preserve, Collier

County, Florida, USA, on 14 June 2005. Transects were arranged as 2 parallel lines spaced 30 m

apart. Each transect was in a continuous tract of either prairie or pineland habitat. Prairie

consists of short hydroperiod wetlands that lack an overstory and woody vegetation in general.

Prairies are dominated by sedges, usually up to 1 m in height (Duever et al. 1986). Pinelands are

forested habitats that form on slightly higher elevation sites in BCNP dominated by Slash Pine

(Pinus elliottii). Pinelands tend to have a dense understory of woody plants, especially Wax

Myrtle (Myrica cerifica) as well as grasses and sedges (Duever et al. 1986). The centerline of









each transect was clearly marked with string attached to polyvinyl chloride (PVC) poles every 10

m. The string on the centerline was stretched and tied to insure that it remained taut during the

entire sampling experiment.

Artificial frogs consisted of small (50 mm total length) plastic frog models (Daytona

Vending, www.freightcloseouts.com). The artificial frogs were painted with flat black spray

paint on the underside and glossy green spray paint on the upper surface to conceal the bright

color of the plastic and to reasonably represent the appearance of the Green Treefrog (Hyla

cinerea), an abundant hylid in the region (Meshaka et al. 2000). Frog models, 50 per transect,

were located randomly within 2 m of either side of each line. The maximum vertical height of

frog placement was selected as a random number between 0 and 150 cm, and actual height was

the highest point on which the frog could be securely fastened in an upright position but not

greater than the maximum. Artificial frogs were attached to vegetation using a single loop of

clear monofilament fishing line. All frogs were recovered after the experiment and determined

to be in the same location as their original placement.

The sampling scheme was based on a combination of the dependent double-observer

approach of Cook and Jacobson (1979) and the distance sampling approach of Buckland et al.

(2001). To sample transects, volunteers were organized into teams of 2 observers. Observers

were told to designate one member of the group as the primary observer and another as the

secondary observer for the first transect, and then to switch roles for the second transect. First,

the primary observer was instructed to walk along the transect and indicate all frogs observed.

Next, the secondary observer was told to record the side of and perpendicular distance from the

transect centerline to each frog detected by the primary observer. Finally, the secondary

observer was instructed to search for frogs not observed by the primary observer (i.e., after the









primary observer passed their location). The secondary observer then recorded their

perpendicular distance and noted that they were missed by the primary observer. The teams of

observers worked at night using headlamps in similar conditions to a standard VES survey.

Seven teams sampled the prairie habitat and 6 teams sampled the pineland habitat. All of the

observers were trained biologists, but none had searched for artificial frogs prior to this

experiment. None of the observers were involved in placing the artificial frogs on transects.

Double Observer Analysis

The double observer data were analyzed using the model of Cook and Jacobson (1979).

This model is based on data collected in which 2 individuals alternate roles as the primary and

secondary observer. Three models of detection probability (p) were analyzed in program

SURVIV (White 1992): individual detection probabilities for each observer (pobs*hab), a single

detection probability for each habitat, respectively (phab), and a constant detection probability

among observers and habitats (p.). Model selection was conducted using Burnham and

Anderson's (1998) information-theoretic approach based on the quasi-likelihood Akaike's

Information Criterion adjusted for small sample sizes (QAICc).

Estimates of p, were obtained for individual observers using the methods of Cook and

Jacobson (1979) as implemented by Nichols et al. (2000) in Program SURVIV (White 1992)

using the equations:


p1 = 1-- 221 and p2 = -
x1lx22 + X22x21 x11x22 + X1lx12

where x1l is the number of individuals detected by observer 1 in the role of primary observer, x22

is the number detected by observer 2 as primary, and x12 and x21 are the number of individuals

observed by observer 1 as the secondary and observer 2 as the secondary, respectively. The









overall estimated p of each team was determined following the methods of Nichols et al. (2000)

using the equation:


p 1
x22 11

and the estimate of the total abundance (N) of frogs from each sample was produced using the

equation:

N=.


where x.. is the sum of the counts of both observers. The estimated variance for this estimate is

given by Nichols et al. (2000):

var() (x..)2 var() +(x. .)(1- p)
P P

Distance Analysis

The perpendicular distance data were analyzed in Program DISTANCE (Thomas et al.

2003) to estimate the density and abundance of artificial frogs. Density was estimated

independently for each team. To increase precision of the estimate, the detection function was

produced from distances pooled across teams but within each habitat stratum. Observers were

told to only look for frogs within 2 m of the transect centerline, and all frogs were placed within

2 m of the centerline. Therefore, the data were truncated to 2 m in the field. Truncation at 1 m

and at 1.5 m was also explored using data filters in DISTANCE.

Results

The randomly chosen locations of the artificial frogs were evenly distributed around the

transect center line (Figure 2-1). The 7 teams in the prairie habitat found 41-68 of the 100 frogs,

and the 6 teams in the pine habitat found 18-43 of the 100 frogs present along both transects









(Table 2-1). Team 5 in the prairie habitat and Team 10 in the pine habitat had an observer who

did not find any frogs in the role of secondary observer. Using location information noted by

observers it is believed that a total of 94 of the possible 100 artificial frogs in prairie habitat and

67 of the 100 artificial frogs in pineland habitat were observed by at least one team.

Actual detection rates calculated within 30 cm, within 50 cm, beyond 50 cm, and overall

varied by team and were not consistently higher closer to the transect centerline (Table 2-2).

Several teams saw fewer frogs near the centerline than further away, most notably teams 2, 4,

and 6 in the prairie habitat (Figure 2-2) and teams 9 and 10 in the pineland habitat (Figure 2-3).

Double Observer

Model selection in Program SURVIV indicated that model Pobs (with individual observer

detection probabilities) was better than habitat level or constant detection probability models as

model Pobs received all of the QAICc weight (Table 2-3). Individual detection probabilities

varied widely among observers ranging from 0.40 -1.0 in prairie habitat and from 0.01 1.0 in

pineland habitat (Table 2-4).

Estimates of total frog abundance along transects in both habitats were lower than the

actual abundance of 100 per habitat (Table 2-5). Abundance estimates ranged from 43-70 on

prairie transects and from 20-49 on pineland transects.

Distance Analysis

Model selection in Program DISTANCE favored models with habitat-specific detection

functions. Models with no post hoc truncation were suitable for the analysis as data were

truncated in the field, and additional truncation did not improve the models. The hazard-rate key

function with cosine adjustment was chosen as the best model for the detection function (AIC

weight = 0.73; Table 2-6). The second best model used the global detection rate with the hazard

rate key function (delta AIC= 1.99, AIC weight= 0.27). Models based on sample (team) level









detection functions did not improve the fit of the model and received no AIC weight (Table 2-6).

The stratum-level detection functions for the prairie and pineland habitat transects have a flat

shoulder to about 175 and 160 cm, respectively, from the centerline (Figure 2-4)

Abundance estimates based on habitat stratum-level detection functions for each individual

team ranged from 46-76 in the prairie and 22-51 in the pineland habitats (Table 2-7). These

abundance estimates from distance analysis were similar to those from the double observer

analysis, but precision was much lower for the distance method (Figure 2-5).

Discussion

Despite the use of data collection protocols and analytical techniques designed to account

for heterogeneity in detection probabilities, there was a large amount of bias in estimates of

artificial frog abundance in the 2 habitats. The true abundance of 100 frogs was not within the

95% confidence intervals of any of the abundance estimates from either the double observer

method (Table 2-5) or the distance method (Table 2-7) in either habitat. Additionally, counts

and estimates varied widely among teams, even though the transects sampled, frog placements,

and environmental conditions were identical. Furthermore, habitat differences between prairie

and pineland led to large differences in estimates of abundance and detection rates of artificial

frogs.

Some apparent bias in both sampling methods may be explained by unobservable frogs.

Only 67 of the artificial frogs were observed by any team in the pineland and only 94 were

observed in the prairie. If we consider frogs not observed by any observer as "unobservable,"

than we would only expect estimates to be similar to the number of observable frogs. However,

only the abundance estimate of team 7 in prairie habitat and team 12 in pineland habitat using the

distance sampling method included the "observable" number of frogs in the 95% confidence

interval.









These results support concerns that the use of counts from VES sampling unadjusted for

detection probability is inappropriate (Crump and Scott 1994, Schmidt 2003). Even under

identical conditions in this survey, different observer teams produced dissimilar counts. Some of

the difference in counts among teams may be attributable to observer skill, although no observer

had previous experience searching for the artificial frogs used in this experiment. Failure to

account for observer bias (Pollock and Kendall 1987) in VES samples could lead to gross errors

as was demonstrated in this study. This is especially important to studies seeking to identify

long-term trends in amphibian populations using data collected over many seasons by different

observers.

Although the dependent double-observer approach of Cook and Jacobson (1979) is

designed to estimate individual detection probabilities and therefore eliminate some of the bias

associated with observers, there are limitations to this approach. Nichols et al. (2000)

hypothesize that detection probability estimates may be biased high using this approach, which

would lead to abundance estimates that are biased low. This study provides empirical evidence

that detection probability estimates are indeed biased high under the assumptions of the Cook

and Jacobson (1979) model. If both observers are not evenly matched in skill, the detection

estimate of the better observer will be biased high. If the poorer observer does not find any

objects missed by the better observer in the role of the primary observer, estimated detection

could reach 100%. Likewise, if both observers are poor and detect few missed objects as

secondary observers, detection estimates for the team will be biased high. It seems likely that

some examples of both of the above scenarios took place in this study (Table 2-4).

Distance analysis is considered useful for estimating density and abundance of objects,

especially when detection rates decline with distance from the observer (Buckland et al. 2001).









In this study, however, there was little noticeable decline in detection rates at distances up to 2 m

from the transect centerline (Table 2-2). Detection rates were estimated to be as high as 1.0 out

to more than 1.5 m from the centerline (Figure 2-4), leading to estimates of abundance that were

biased low as with the double observer method.

Rigorous testing for bias in animal sampling protocols is not always performed, but can be

informative. Anderson et al. (2001) found that modifications to desert tortoise line transect

sampling protocols were necessary after testing the technique with artificial tortoises. Nichols et

al. (1986) found evidence of inter-observer variation in searches for white-winged dove nests

marked by individual observers and recommended future estimation of observer-specific

detection rates. Rodda et al. (2005) evaluated relative abundance estimates based on visual

searches against absolute abundance estimates based on removal techniques and found poor

correspondence between the two. They concluded that visual searches alone were only suitable

for 1 of the 6 reptile species they surveyed.

This study provides evidence that observer differences in detection rates are very important

and even sampling methods designed to be robust to these differences may be inadequate to

describe the absolute abundance of animals. Clearly, we have an advantage in this study of

knowing the true abundance of objects. Such information is not available in actual animal

populations. However, we cannot assume that the artificial frogs in this study were perfect

surrogates for actual frogs. The fact that they are immobile probably decreased their

detectability, as many observers find frogs by detecting their movements. Also, many observers

key detection on eyeshines not present in the surrogates. Therefore, we cannot use this study to

calibrate the amount of bias in VES sampling, but it does help illustrate an important problem.









Researchers gathering VES data should be aware that there is potential for great bias in

abundance estimates using this method.









Table 2-1. Number of artificial frogs (out of a possible 100) observed by each team in prairie and
pineland habitats in BCNP. x1l is the count of objects detected by observer 1 when
observer 1 was the primary observer, x22 is the count of objects by observer 2 when
observer 2 was primary, and x12 and x21 are the counts of observers 1 and 2 in the
roles of secondary observer, respectively.
Habitat Team x11 x22 x12 x21 Total
Prairie Team 1 28 14 8 6 56
Team 2 25 16 9 1 51
Team 3 21 15 3 4 43
Team 4 20 22 1 12 55
Team 5 22 13 8 0 43
Team 6 13 13 4 11 41
Team 7 24 32 10 2 68

Pineland Team 8 9 1 5 4 19
Team 9 9 4 1 4 18
Team 10 11 10 0 4 25
Team 11 17 7 6 2 32
Team 12 19 10 12 2 43
Team 13 11 7 3 3 24









Table 2-2. Actual detection rates of a known population of artificial frogs within 30 cm, within
50 cm, beyond 50 cm, and overall for teams searching in prairie and pineland habitat
of BCNP.
Habitat Team p <30 cm p <50 cm p >50 cm Overall
Prairie Team 1 0.611 0.600 0.547 0.56
Team 2 0.500 0.520 0.507 0.51
Team 3 0.333 0.360 0.453 0.43
Team 4 0.389 0.560 0.547 0.55
Team 5 0.500 0.600 0.373 0.43
Team 6 0.444 0.360 0.427 0.41
Team 7 0.444 0.720 0.667 0.68

Pineland Team 8 0.357 0.269 0.162 0.19
Team 9 0.071 0.077 0.216 0.18
Team 10 0.143 0.269 0.243 0.25
Team 11 0.500 0.385 0.297 0.32
Team 12 0.643 0.577 0.378 0.43
Team 13 0.429 0.385 0.189 0.24









Table 2-3. Model selection for the three double-observer models analyzed in Program SURVIV
with the model log-likelihood, number of parameters (K), quasi-likelihood Akaike's
information criterion adjusted for small sample sizes (QAICc), difference between
each QAICc and the minimum QAICc (Delta QAICc) and the QAICc weight.
Model Log Likelihood K QAICc Delta QAICc QAICc Weight
pobs*hab -35.32 26 125.507 0 1.0000
Phab -75.76 2 153.074 27.567 0.0000
p. -74.53 1 153.524 28.017 0.0000










Table 2-4. Estimates from program SURVIV of individual detection probabilities (p,) of
artificial frogs for each observer in prairie and pineland habitat with standard error
(S.E.) and 95% confidence intervals (C.I.) taken from the double observer method..
Habitat Observer PA S.E. Lower 95% C.I. Upper 95% C.I.
Prairie 1 0.7230 0.1119 0.5036 0.9424
2 0.5590 0.1293 0.3055 0.8125
3 0.9389 0.0601 0.8211 1.0567
4 0.6251 0.1010 0.4271 0.8231
5 0.8082 0.0902 0.6313 0.9850
6 0.8017 0.1060 0.5940 1.0094
7 0.6079 0.0912 0.4292 0.7866
8 0.9304 0.0687 0.7957 1.0651
9 1.0000 0.0002 0.9996 1.0004
10 0.6197 0.1059 0.4120 0.8273
11 0.4019 0.1554 0.0974 0.7064
12 0.5667 0.2062 0.1625 0.9709
13 0.8977 0.0696 0.7614 1.0341
14 0.7415 0.0728 0.5988 0.8842

Pineland 15 0.0123 0.8123 -1.5798 1.6044
16 0.0044 0.2948 -0.5735 0.5823
17 0.6156 0.1814 0.2601 0.9712
18 0.7122 0.2635 0.1958 1.2286
19 0.7330 0.1142 0.5092 0.9569
20 1.0000 0.0001 0.9999 1.0001
21 0.8046 0.1400 0.5302 1.0789
22 0.4845 0.1598 0.1714 0.7977
23 0.7912 0.1488 0.4996 1.0829
24 0.3979 0.1248 0.1533 0.6425
25 0.6931 0.1694 0.3612 1.0251
26 0.6168 0.1923 0.2400 0.9936










Table 2-5. Abundance estimates (N) of artificial treefrogs for each team in prairie and pineland
habitat with standard error (S.E.) and upper and lower 95% confidence intervals (C.
I.) based on Chaos's estimator (Nichols et al. 2000) taken from the double observer
data..
Chao 95% C.I.
Habitat Team N S.E. N Lower. Upper
Prairie Team 1 62.3 3.01 58.59 71.34
Team 2 56.7 2.85 53.29 65.39
Team 3 47.8 2.57 44.82 55.86
Team 4 61.2 2.98 57.53 70.15
Team 5 47.8 2.57 44.82 55.86
Team 6 45.6 2.50 42.71 53.47
Team 7 75.7 3.40 71.33 85.59

Pineland Team 8 21.1 1.62 19.57 26.99
Team 9 20.0 1.57 18.53 25.78
Team 10 27.8 1.88 25.85 34.27
Team 11 35.6 2.16 33.21 42.69
Team 12 47.8 2.57 44.82 55.86
Team 13 26.7 1.84 24.81 33.06









Table 2-6. Model selection for detection functions in Program DISTANCE showing number of
parameters (K), Akaike's Information Criterion (AIC), the difference between each
AIC and the minimum (Delta AIC), and the AIC Weight of each model. Detection
was modeled at the sample, stratum, or global level with the hazard rate key function
or the half normal key function.
Delta AIC
Model Name K AIC AIC Weight
Stratum detection; Hazard Rate Key 4 5374.38 0 0.7300
Global detection, Hazard Rate Key 2 5376.37 1.99 0.2697
Global detection, Half-normal Key 5 5390.49 16.10 0.0002
Stratum detection; Half-normal Key 6 5404.70 30.32 0.0000
Sample detection; Hazard Rate Key 26 5413.01 38.63 0.0000
Sample detection; Half-normal Key 13 5436.28 61.90 0.0000










Table 2-7. Abundance (N) estimates of artificial frogs by team for transects in the prairie and
pineland habitats with coefficient of variation (CV) and 95% confidence interval (CI)
from the distance sampling approach.
Habitat Team N CV Lower 95% C.I. Upper 95% C.I.
Prairie Team 1 49 15.16 37 66
Team 2 57 14.09 43 75
Team 3 48 15.33 36 65
Team 4 62 13.58 47 81
Team 5 48 15.33 36 65
Team 6 46 15.70 34 63
Team 7 76 12.23 60 97

Pineland Team 8 23 23.08 14 36
Team 9 22 23.70 14 34
Team 10 30 20.16 20 44
Team 11 38 17.85 27 54
Team 12 51 15.45 38 70
Team 13 29 20.56 19 43











Prairie


100
100


150
150


Perpendicular Distance (cm)



Pineland


100


Perpendicular Distance (cm)


Figure 2-1. Histograms of the frequency of perpendicular distances in cm of actual artificial frog
placements along the 2 transect lines in the prairie and pineland habitat.


(0-


N -


I
200


L _


200

















_ --___ __ i _--_
I I I I I I I I I I
0 50 100 150 200 0 50 100 150 200




Team 3 Team 4



C
p e


I I I I I
0 50 100 150 200




Team 5


0 50 100 150 200




Team 6


I I I I I I I I I I







Team 7




S_ c l

0 50 100 150 200

Perpendicular Distance (cm)

Figure 2-2. Histograms of the frequency of detections of artificial frogs by each team at
perpendicular distances in cm along the 2 transect lines in the prairie habitat at
BCNP.


Team 2


Team 1












Team 8


(U

U-

0-


0 50 100 150 200


Team 10


I I I I I
0 50 100 150 200


Team 12


0 50 100 150 200


Team 11


I I I I I
0 50 100 150 200




Team 13


---


I I I I I I I I I I
0 50 100 150 200 0 50 100 150 200
Perpendicular Distance (cm) Perpendicular Distance (cm)

Figure 2-3. Histograms of the frequency of detections of artificial frogs by each team at
perpendicular distances in cm along the 2 transect lines in the pineland habitat of
BCNP.


Team 9


-1


II7

















14



12






2 08

0,
06



04



02



00
0 50 100 150 200 250
Perpendicular distance in centimeters


14



12 B







08

2






1o
06




04



02




0 50 100 150 200 250
Perpendicular distance in centimeters


Figure 2-4. Histograms of the frequency of detections of artificial frogs at perpendicular

distances in cm along the 2 transect lines for all teams in the prairie habitat (A) and

the pineland habitat (B) with the best model of the detection function (red lines) from

Program DISTANCE superimposed.















100 -

Double Observer
0 Distance
80 -
LO 80




E


40 -

-0

I 20



0 -i ii --
1 2 3 4 5 6 7


100 B

o 0* Double Observer
80 0 Distance
80

'0)
4-
M 60
E 0

LU T
8 40 I
c


200



I II I I
20




8 9 10 11 12 13

Team Number
Figure 2-5. Abundance estimates with standard error for each team from prairie (A) and pineland
habitat (B) using the double observer and distance methods. True abundance was 100
in both habitats.









CHAPTER 3
USING SITE OCCUPANCY MODELING TO DETERMINE THE EFFECT OF OFF-ROAD
VEHICLE USE ON GROUND-DWELLING ANURANS

Introduction

Off-road vehicles (ORV) can impact wildlife species directly by causing physical harm

(Steiner and Leatherman 1981) or indirectly by altering behavior or disturbing habitats

(Brattstrom and Bondello 1995, Guyer et al. 1996). ORVs create noise which may be disruptive

to wildlife, or the presence of large moving vehicles may be disturbing. In addition, ORVs may

create trails and damage to the vegetation in areas that receive heavy use. One of the major

management concerns of Big Cypress National Preserve (BCNP) in southern Florida, USA is the

regulation of ORV use (NPS 2000, Duever 2005) where mapped ORV trails total over 47,900

km in length (Welch et al. 1999). Janis and Clark (2002) found evidence that ORV use altered

the behavior of the endangered Florida panther in Big Cypress, and Duever et al. (1981)

demonstrated that ORVs alter vegetation composition and hydrology at impacted sites (Figure 3-

1). It is unclear how other species of wildlife, especially amphibians, are affected by ORV use in

BCNP.

Determining the impacts of ORVs on local populations may be possible through mark-

recapture sampling or some other technique that provides estimates of abundance. However, at

larger spatial scales (i.e. across landscapes) it becomes increasingly futile to attempt to estimate

the abundance or density of amphibians. It is difficult to enumerate such large populations, and

population sizes may fluctuate with season and environmental conditions (Green 2003). A

relatively new method, site occupancy or proportion area occupied, allows collection of simple

presence/absence data across the entire landscape to make inference regarding species status.

The site occupancy rate of species across a landscape is more meaningful at large scales since a

larger proportion of the area can be sampled than in traditional mark-recapture. Also unlike









simple counts and estimates of relative density (catch per unit effort), occupancy accounts for

detection or non-detection of the species after repeated samplings of many sites (MacKenzie et

al. 2002). Covariates (e.g. habitat type, ORV use) may be used to improve estimates of

occupancy probabilities (MacKenzie et al. 2002), and model selection may also be used to make

inferences about the effects of covariates (MacKenzie et al. 2005, Schmidt and Pellet 2005).

Ground dwelling anurans are one group of amphibians that may be especially affected by

ORV use. These species often occur far from permanent water, and most breed in small, fish-

free ponds (Duellman and Trueb 1986). They tend to have a low dispersal capability (Blaustein

et al. 1994, Alford and Richards 1999), so they are likely to spend most of their life within the

same small area. For these reasons I expect these species to be especially affected by ORV use

in Big Cypress. ORV use may alter microhabitats at sites by decreasing vegetation or increasing

drainage. These alterations may make some sites less suitable for ground dwelling anurans.

The objective of this study was to determine if ORV use was an important factor

influencing the site occupancy of four ground-dwelling anuran species in BCNP: Oak Toad

(Bufo quercicus), Southern Toad (Bufo terrestris), Eastern Narrow-mouthed Toad (Gastrophryne

carolinensis), and Greenhouse Frog (Eleutherodactylusplanirostris). Detection data of

amphibians from a random sample of sites in BCNP were modeled with site covariates including

an index of ORV use created from geographic information system (GIS) data of ORV trails

using the site occupancy method of MacKenzie et al. (2002). I hypothesized that occupancy of

these four ground-dwelling anuran species would be negatively associated with ORV use.

Methods

Study Area

Big Cypress National Preserve is a 295,000 ha natural area managed by the National Park

Service in Collier and Monroe Counties of southwestern Florida, USA. BCNP is bordered on









the south and east by Everglades National Park and Water Conservation Area 3A (managed by

the South Florida Water Management District). Cape Romano/10,000 Islands National Wildlife

Refuge and Florida Panther National Wildlife Refuge, as well as Fakahatchee Strand State

Preserve lie on the western boundary of Big Cypress. Because of its large size and geographic

location, Big Cypress is not heavily influenced by habitat loss due to development like many

other areas in Florida. However, Big Cypress has been used by recreational ORV users since

before its establishment in 1974 (NPS 2000). Current regulations limit use to no more than 2000

registered ORVs per year, but lasting signs of 47,900 km of old and current ORV trails are

apparent throughout the preserve (Duever et al. 1987, Welch et al. 1999, NPS 2000).

Sampling

Random sites were chosen throughout Big Cypress using the animal movement analysis

extension of Hooge and Eichenlaub (1997) in ArcView 3.2 (Environmental Systems Research

Institute, Inc., www.esri.com). A sample of these random points accessible by foot or ORV was

chosen for amphibian sampling (Figure 3-2). These sites represented 5 different habitat types:

cypress strand, cypress prairie, prairie, hammock, and pineland, based on the vegetation

classification scheme of Madden et al. (1999). The number of sampling occasions per site was

variable. Some were sampled on a monthly basis from February 2002 to March 2003 to provide

a time series of detection data from sites. Other sites were sampled just twice during the entire

project to increase the geographic coverage of the sampling. This approach was intended to

balance the effort between repeated sampling and additional sites.

Sampling for amphibians consisted of standard visual encounter survey (VES) techniques

(Crump and Scott 1994). All VES samples were initiated at least 30 minutes after sunset, and

each survey was conducted by at least two experienced observers using 6-volt spotlights with

halogen bulbs. VES samples were time and area constrained such that the area within a 20-m









radius of the randomly chosen point (1,256 m2) was searched for 1 person-hr. All areas of each

plot were visually scanned, but judgment of the observers was used to determine which areas

within the plot received the most emphasis. The goal was to find as many individual amphibians

as possible. All possible amphibian locations were searched, including trees and other

vegetation as well as bare ground and leaf litter. Each amphibian observed was captured (if

possible) and identified to species. A 10-minute vocalization survey was conducted during each

VES sample. All species of frogs and toads heard vocalizing were noted. All anurans that could

be heard were included, even if it was possible or likely that they were calling from a location

outside of the 20-m radius plot.

In addition to the biological data, environmental data were collected in the field during

each survey. Air temperature and relative humidity were measured using a digital

thermohygrometer. The date and time of the sample and whether the plot was inundated with

water at the time of the sample also were recorded. Sampled sites were assigned an ORV use

index based on the sum of ORV trails within a 500-m radius circle around the sampling point by

using the ORV trail GIS dataset developed by Welch et al. (1999). The Welch et al. (1999) map

includes all trails visible from aerial photos, even ones still visible in areas that had been closed

to ORVs for as much as 2 decades prior to the photographs. Eight of the 70 sites sampled in this

study fell in areas designated as "high use" by Welch et al. (1999), and trails were not mapped in

these high use areas because of the high density of trails. The ORV index for sites located in

high use areas was set equal to the highest value from sites for which trail data were available.

In addition to the ORV index, a hydrologic index was created based on number of months

inundated for each site using a habitat and hydrologic model developed by Duever et al. (1986).









Data Analysis

Site occupancy rates and detection probabilities were modeled in program PRESENCE

(MacKenzie et al. 2002, MacKenzie et al. 2006) using the single season model. This method

assumes that sites are closed to changes in occupancy within the study, and that detection of a

species at a site is independent of detections at other sites. This method also assumes that

species are not falsely detected, but species may or may not be detected when present. This

method was deemed appropriate for these focal anuran species due to their difficulty to detect

and low dispersal.

Site-specific covariables, those that directly affect the estimate of occupancy (V,) were

habitat type, hydrologic index and ORV use index. Values for the indices were standardized so

that the means fell between 1 and 0, a necessary condition when using the logit link function

(MacKenzie et al. 2006). Sampling occasion covariates that could affect detection probability

(p) were air temperature, relative humidity, presence of standing water, and season of the year.

For each species, we considered 80 models that were combinations of the covariates thought to

be biologically meaningful (Table 3-1). The best model was chosen as the one with the lowest

value for Akaike's information criterion (AIC), or the most parsimonious model (model with the

best fit for the fewest parameters; Burnham and Anderson 1998). The effect of ORV use on

species occupancy was determined using model selection to determine the AIC weight of all

models including the ORV use covariate and by examining the beta estimates for the ORV use

index in the models including that covariate.

Results

A total of 469 sampling visits to 70 sites were made from February 2002 to March 2003.

The highest number of study sites (31) was in prairie habitat. Between 7 and 12 sites in each of

the other habitats were visited (Table 3-2). The four focal anuran species were detected between









13 and 117 times, and naive occupancy rates (proportion of sites at which a detection occurred)

varied from 17 to 52% (Table 3-3).

The best model (model with lowest AIC value; Burnham and Anderson 1998) for each of

the four species included the ORV index as a site covariate (Tables 3-4 3-7). When AIC

weights of models including specific site covariates were summed, the ORV index covariate had

the most weight for southern toads, greenhouse frogs, and eastern narrow-mouthed toads (Table

3-8). Only oak toads had less weight on models including the ORV index than other covariates.

Beta parameters for the ORV index in the best models were negative for all species except

southern toads (Table 3-9). Numerical convergence could not be reached to estimate standard

error (S.E.) for the ORV index beta parameter for oak toads, but S.E. estimates were obtained for

the other species. The 95% confidence intervals of the beta parameter estimates for the ORV

index covariate overlapped 0 for all species but greenhouse frogs.

Discussion

The results of this study indicate that the ORV index and, thus, ORV-use is a strong

predictor of whether a site will be occupied by these four species of anurans. Each of these

species had the ORV index covariate in the best models for occupancy, and the sum of the model

AIC weights was highest for those models including the ORV index except for oak toads. This

indicates that for some species, occupancy of a site may depend more on ORV use than on

hydrology or habitat.

Three of the four species of anurans had beta values for ORV index that were negative,

indicating negative associations with ORV use (Table 3-9). It was predicted that these small,

ground-dwelling anurans would be negatively influenced by the use of ORVs due to ground level

disturbance of vegetation and altered hydrology. One species, however, the Southern Toad, was

positively associated with ORV use. Although this is counter to the original prediction,









morphology and reproductive strategy of this species might explain the difference in response to

ORV use. Southern Toads are larger than the other species, and their tadpoles require up to

twice as long (2 months compared to 1 month) to develop as Oak Toads or Eastern Narrow-

mouthed Toads (Ashton and Ashton 1988). ORVs can alter the vegetation and hydropattern of

areas resulting in a loss of vegetation and increased ponding in ruts and artificial depressions

(Duever et al. 1981). Southern Toads may take advantage of the increased temporal and spatial

extent of standing water for breeding purposes.

For all of the species other than oak toads, habitat was not as important in prediction of site

occupancy as the ORV index (Table 3-8). This may be in part due to the fact that these 4

anurans are habitat generalists in south Florida and are not closely associated with any particular

habitat type (Duellman and Schwartz 1958, Meshaka et al. 2000, Rice et al. 2004). Habitat was a

covariate in the best models for both American Toad (Bufo americanus) and Spring Peeper

(Pseudacris crucifer) in a the study of MacKenzie et al. (2002) in Maryland wetlands, but habitat

was not as important as the previous years count in the study of Schmidt and Pellet (2005) of

Tree Frog (Hyla arborea) and Natterjack Toad (Bufo calamita) in Europe. Landscapes in Big

Cypress are very heterogeneous; thus different habitat types are often in close proximity to one

another (McPherson 1974, Duever et al. 1986) and frogs might easily transition from one habitat

to another. Thus, it is more difficult to determine differences in habitat-level occupancy at the

scale of this study.

The hydrologic index used in this study generally had higher AIC weights in model sets

than habitat, but lower than the ORV index. Oak Toad was the one exception to this pattern, for

which the hydrologic index had the most AIC weight in the model set. The hydrologic index

was based on the number of months each site would typically be inundated with water in a year.









Although this hydroperiod is highly correlated with habitat type (Duever et al. 1986), there are

some examples of sites being wetter or drier than others with the same habitat type. It is possible

that the occupancy was more sensitive to the hydrological index than habitat, but this index will

change between years as local rainfall varies (DeAngelis et al. 1998). Habitats are much more

fixed and tend to be primarily a result of microtopography (Duever et al. 1986).

Amphibians are often used as indicator species (Vitt et al. 1990, Welsh and Ollivier 1998,

Galatowitsch et al. 1999, Sheridan and Olson 2003). This study illustrates how amphibians may

be indicators of the effects of ORV use in Big Cypress. The ORV trail data provided by Welch

et al. (1999) shows all trails and does not differentiate between old, persistent trails and currently

used ones. Duever et al. (1987) demonstrated that old trails may take many decades to recover.

Some of the sites sampled in this study are in areas that have been closed to ORV use for more

than 10 years, although they still retain many visible ruts and other physical evidence of ORV

use. It is important to remember that this study only considers evidence of ORV use in the form

of the index used, and does not examine the temporal component of when ORV use occurred.

Consequently, this study found evidence of an impact of ORV use on amphibian species

occupancy but did not address recovery of previously impacted areas.

Due to the observational nature of this study, it is not possible to determine the

mechanisms by which ORV use influences amphibian occupancy. However, this study should

help stimulate more research on the topic. Resource management staff at Big Cypress concerned

with reducing impacts of ORV use in the preserve should be aware that there is evidence that

ORV use influences the site occupancy of amphibians. These amphibian species may be

indicators of ecosystem impacts not previously shown. A monitoring program designed using

the same techniques of this study could be used to track changes over time. Stratifying sampling









by historic and current ORV use areas could also help determine how long the impacts of ORV

use can be detected in amphibian occupancy. Colonization and extinction of sites with varying

levels of ORV use could also be monitored over time using the open model approach of

MacKenzie et al. (2003).









Table 3-1. Combinations of the 3 site covariates and 4 sampling covariates that were used in the

occupancy analysis for each species. Each set of site covariates was modeled along with each set

of sampling covariates for a total of 80 unique models for each species.

Site Covariates Sampling Covariates
Constant Constant
Habitat Season
Hydrologic Index Temperature
ORV Index Temperature, Humidity
Habitat, ORV Index Temperature, Humidity, Water
Habitat, Hydrologic Index Temperature, Humidity, Water, Season
ORV, Hydrologic Index Temperature, Water
Habitat, ORV, Hydrologic Index Temperature, Water, Season
Water
Water, Season









Table 3-2. Number of sampling sites and total number of site visits by habitat.
Number Number of
Habitat
of Sites Visits
Cypress Strand 11 89
Cypress Prairie 9 78
Prairie 31 122
Hammock 7 72
Pineland 12 108
Total 70 469









Table 3-3. Number of detections by species, and proportion of sites at which a detection occurred
(naive occupancy) during amphibian surveys across all habitat types.
Species No. Detections Naive Occupancy
Oak Toad 17 21.43%
Southern Toad 26 28.57%
Greenhouse Frog 117 52.86%
Eastern Narrow-mouthed Toad 13 17.14%









Table 3-4. Model selection results for the oak toad (Bufo quercicus), including Akaike's
Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and
Anderson 1998) for all models with any AIC weight.
Model AIC delta AIC AIC weight
psi(Hab,ORV,Hydro), p(Temp,Humid,Water, Season) 117.92 0.00 0.2946
psi(Hab,ORV,Hydro), p(Temp,Water, Season) 118.01 0.09 0.2817
psi(Hab,Hydro),p(Temp,Humid,Water, Season) 119.92 2.00 0.1084
psi(Hab,Hydro),p(Temp,Water, Season) 119.98 2.06 0.1052
psi(Hab,ORV,Hydro),p(Water, Season) 121.54 3.62 0.0482
psi(Hydro),p(Temp,Humid,Water, Season) 122.17 4.25 0.0352
psi(Hydro),p(Temp,Water, Season) 122.24 4.32 0.0340
psi(Hab,ORV,Hydro),p(Season) 122.96 5.04 0.0237
psi(Hab,Hydro),p(Water, Season) 123.58 5.66 0.0174
psi(Hydro,ORV),p(Temp,Humid,Water, Season) 123.91 5.99 0.0147
psi(Hydro,ORV),p(Temp,Water, Season) 123.98 6.06 0.0142
psi(Hab,Hydro),p(Season) 124.76 6.84 0.0096
psi(Hydro),p(Water, Season) 125.43 7.51 0.0069
psi(Hydro,ORV),p(Water, Season) 127.29 9.37 0.0027
psi(Hydro,ORV),p(Season) 127.82 9.90 0.0021
psi(Hydro),p(Season) 132.56 14.64 0.0002









Table 3-5. Model selection results for the southern toad (Bufo terrestris), including Akaike's
Information Criterion (AIC) and the delta AIC and AIC weight (Burnham and
Anderson 1998) for all models with any AIC weight.
Model AIC Delta AIC AIC Weight
psi(Hab,ORV),p(Season) 194.32 0.00 0.1109
psi(.),p(season) 194.50 0.18 0.1014
psi(ORV),p(Season) 194.99 0.67 0.0793
psi(.),p(Water, Season) 195.14 0.82 0.0736
psi(Hab,ORV),p(Water, Season) 195.85 1.53 0.0516
psi(.),p(Temp,Humid,Water,Season) 195.86 1.54 0.0514
psi(.),p(Temp,Water, Season) 196.09 1.77 0.0458
psi(Hydro),p(Season) 196.21 1.89 0.0431
psi(Hydro),p(Water, Season) 196.38 2.06 0.0396
psi(ORV),p(Temp,Humid,Water, Season) 196.48 2.16 0.0377
psi(ORV),p(Temp,Water, Season) 196.68 2.36 0.0341
psi(Hab,ORV),p(Temp,Humid,Water, Season) 196.77 2.45 0.0326
psi(Hydro),p(Temp,Humid,Water, Season) 197.16 2.84 0.0268
psi(Hydro),p(Temp,Water, Season) 197.27 2.95 0.0254
psi(Hydro,ORV),p(Season) 197.28 2.96 0.0252
psi(Hydro,ORV),p(Water, Season) 197.33 3.01 0.0246
psi(Hydro,ORV),p(Temp,Humid,Water, Season) 198.07 3.75 0.017
psi(Hydro,ORV),p(Temp,Water, Season) 198.17 3.85 0.0162
psi(.),p(Temp,Water) 198.78 4.46 0.0119
psi(Hab,ORV),p(Temp,Water, Season) 198.78 4.46 0.0119
psi(.),p(Water) 198.99 4.67 0.0107
psi(ORV),p(Temp,Water) 199.45 5.13 0.0085
psi(ORV),p(Water) 199.51 5.19 0.0083
psi(.),p(Temp,Humid,Water) 199.56 5.24 0.0081
psi(Hydro),p(Temp,Water) 199.68 5.36 0.0076
psi(Hab),p(Season) 199.75 5.43 0.0073
psi(Hab,ORV),p(Water) 199.96 5.64 0.0066
psi(Hydro),p(Water) 200.02 5.70 0.0064
psi(Hab,ORV),p(Temp,Water) 200.08 5.76 0.0062
psi(ORV),p(Temp,Humid,Water) 200.32 6.00 0.0055
psi(Hab),p(Water, Season) 200.36 6.04 0.0054
psi(Hab,Hydro),p(Season) 200.43 6.11 0.0052
psi(Hydro),p(Temp,Humid,Water) 200.56 6.24 0.0049
psi(Hab,Hydro),p(Water, Season) 200.57 6.25 0.0049
psi(Hydro,ORV),p(Temp,Water) 200.76 6.44 0.0044
psi(Hab),p(Temp,Water, Season) 201.06 6.74 0.0038
psi(Hab),p(Temp,Humid,Water, Season) 201.09 6.77 0.0038
psi(Hydro,ORV),p(Water) 201.14 6.82 0.0037
psi(Hab,Hydro),p(Temp,Water, Season) 201.24 6.92 0.0035
psi(Hab,Hydro),p(Temp,Humid,Water, Season) 201.27 6.95 0.0034
psi(Hydro,ORV),p(Temp,Humid,Water) 201.67 7.35 0.0028








Table 3-5 (Continued).
Model AIC Delta AIC AIC Weight
psi(Hydro,ORV),p(.) 202.13 7.81 0.0022
psi(Hab,ORV,Hydro),p(Season) 202.49 8.17 0.0019
psi(Hab,ORV,Hydro),p(Water, Season) 202.62 8.30 0.0017
psi(Hab,ORV),p(Temp,Humid,Water) 202.74 8.42 0.0016
psi(Hab,ORV,Hydro),p(Temp,Water, Season) 203.23 8.91 0.0013
psi(Hab,ORV,Hydro),p(Temp,humid,Water, Season) 203.27 8.95 0.0013
psi(Hab,Hydro),p(Temp,Water) 204.37 10.05 0.0007
psi(Hab),p(Temp,Water) 204.38 10.06 0.0007
psi(Hydro,ORV),p(Temp,Humid) 204.48 10.16 0.0007
psi(Hab),p(Water) 204.87 10.55 0.0006
psi(ORV),p(.) 204.93 10.61 0.0006
psi(.),p(.) 204.94 10.62 0.0005
psi(Hab,ORV,Hydro),p(.) 205.08 10.76 0.0005
psi(Hab,Hydro),p(Water) 205.11 10.79 0.0005









Table 3-6. Model selection results for the greenhouse frog (Eleutherodactylusplanirostris),
including Akaike's Information Criterion (AIC) and the delta AIC and AIC weight
(Burnham and Anderson 1998) for all models with any AIC weight.
Model AIC Delta AIC AIC Weight
psi(Hydro,ORV),p(Temp,Humid,Water,Saeson) 409.4 0 0.5162
psi(Hydro,ORV),p(Temp,Water, Season) 411.01 1.61 0.2308
psi(Hab,ORV,Hydro),p(Temp,Humid,Water, Season) 413.43 4.03 0.0688
psi(ORV),p(Temp,Humid,Water, Season) 414.09 4.69 0.0495
psi(Hydro,ORV),p(Water,Season) 414.57 5.17 0.0389
psi(Hab,ORV,Hydro),p(Temp,Water,Season) 415.02 5.62 0.0311
psi(ORV),p(Temp,Water, Season) 415.89 6.49 0.0201
psi(.),p(Temp,Humid,Water, Season) 417.42 8.02 0.0094
psi(Hab,Hydro),p(Temp,Humid,Water, Season) 418.25 8.85 0.0062
psi(Hab,ORV,Hydro),p(Water, Season) 418.51 9.11 0.0054
psi(ORV),p(Water, Season) 418.96 9.56 0.0043
psi(.),p(Temp,Water, Season) 419.14 9.74 0.004
psi(Hab,ORV),p(Temp,Humid,Water, Season) 419.36 9.96 0.0035
psi(Hydro),p(Temp,Water, Season) 419.45 10.05 0.0034
psi(Hab,Hydro),p(Temp,Water, Season) 420.07 10.67 0.0025
psi(Hab),p(Temp,Humid,Water, Season) 421.13 11.73 0.0015
psi(Hab,ORV),p(Temp,Water, Season) 421.28 11.88 0.0014
psi(.),p(Water, Season) 422.41 13.01 0.0008
psi(Hab),p(Temp,Water, Season) 423.08 13.68 0.0006
psi(Hydro),p(Water, Season) 423.14 13.74 0.0005
psi(Hab,Hydro),p(Water, Season) 423.97 14.57 0.0004
psi(Hydro,ORV),p(Season) 424.02 14.62 0.0003
psi(Hab,ORV),p(Water, Season) 424.67 15.27 0.0002
psi(Hab),p(Water, Season) 426.55 17.15 0.0001
psi(Hab,ORV,Hydro),p(Season) 427.21 17.81 0.0001








Table 3-7. Model selection results for the eastern narrow-mouthed toad (Gastrophryne
carolinensis), including Akaike's Information Criterion (AIC) and the delta AIC and
AIC weight (Burnham and Anderson 1998) for all models with any AIC weight.
Model AIC Delta AIC AIC Weight
psi(Hydro,ORV),p(Season) 113.97 0.00 0.3182
psi(Hydro,ORV),p(Water, Season) 115.96 1.99 0.1177
psi(ORV),p(Season) 117.04 3.07 0.0686
psi(Hydro,ORV),p(Temp,Water, Season) 117.49 3.52 0.0548
psi(Hydro,ORV),p(Water) 117.65 3.68 0.0505
psi(Hab,Hydro),p(Season) 118.15 4.18 0.0394
psi(.),p(Season) 118.43 4.46 0.0342
psi(ORV),p(Temp,Water, Season) 119.00 5.03 0.0257
psi(ORV),p(Water, Season) 119.03 5.06 0.0254
psi(Hydro,ORV),p(Temp,Humid,Water, Season) 119.48 5.51 0.0202
psi(Hydro,ORV),p(Temp,Water) 119.52 5.55 0.0198
psi(Hab,ORV,Hydro),p(Season) 119.56 5.59 0.0194
psi(Hydro,ORV),p(.) 119.73 5.76 0.0179
psi(Hydro,ORV),p(Temp) 120.31 6.34 0.0134
psi(.),p(Water, Season) 120.41 6.44 0.0127
psi(Hydro),p(Season) 120.43 6.46 0.0126
psi(Hydro,ORV),p(Temp,Humid,Water) 120.82 6.85 0.0104
psi(Hab,Hydro),p(Water, Season) 121.21 7.24 0.0085
psi(Hab),p(Season) 121.38 7.41 0.0078
psi(ORV),p(Water) 121.45 7.48 0.0076
psi(Hab,ORV,Hydro),p(Water, Season) 121.54 7.57 0.0072
psi(ORV),p(.) 121.68 7.71 0.0067
psi(.),p(Temp,Water, Season) 121.77 7.80 0.0064
psi(Hab,ORV,Hydro),p(Temp,Water, Season) 121.94 7.97 0.0059
psi(Hydro,ORV),p(Temp,Humid) 122.03 8.06 0.0057
psi(Hydro),p(Water, Season) 122.41 8.44 0.0047
psi(ORV),p(Temp,Humid,Water, Season) 122.47 8.50 0.0045
psi(ORV),p(Temp,Water) 122.77 8.80 0.0039
psi(Hab,Hydro),p(Temp,Water, Season) 122.82 8.85 0.0038
psi(.),p(.) 122.86 8.89 0.0037
psi(.),p(Water) 122.94 8.97 0.0036
psi(Hab,ORV),p(Season) 123.08 9.11 0.0033
psi(ORV),p(Temp) 123.21 9.24 0.0031
psi(Hab,Hydro),p(.) 123.30 9.33 0.0030
psi(Hab),p(Water, Season) 123.34 9.37 0.0029
psi(Hab,Hydro),p(Water) 123.34 9.37 0.0029
psi(Hab,Hydro),p(Temp,Water) 123.42 9.45 0.0028
psi(.),p(Temp,Humid,Water, Season) 123.77 9.80 0.0024
psi(Hydro),p(Temp,Water, Season) 123.77 9.80 0.0024
psi(Hab,Hydro),p(Temp,Humid,Water, Season) 123.78 9.81 0.0024









Table 3-7 (Continued).
Model
psi(Hab,ORV,Hydro),p(Temp,Humid,Water, Season)
psi(Hab,ORV,Hydro),p(Water)
psi(.),p(Temp,Water)
psi(Hab,Hydro),p(Temp,Humid,Water)
psi(.),p(Temp)
psi(Hab,ORV,Hydro),p(Temp,Water)
psi(Hab,ORV,Hydro),p(Temp)
psi(Hab),p(Temp,Water, Season)
psi(ORV),p(Temp,Humid,Water)
psi(Hydro),p(.)
psi(ORV),p(Temp,Humid)
psi(Hydro),p(Water)
psi(Hab,ORV),p(Water, Season)
psi(Hab,Hydro),p(Temp)
psi(Hab,ORV,Hydro),p(Temp,Humid,Water)
psi(Hab),p(.)
psi(Hydro),p(Temp,Humid,Water, Season)
psi(.),p(Temp,Humid)
psi(.),p(Temp,Humid,Water)
psi(Hydro),p(Temp,Water)
psi(Hab),p(Water)
psi(Hydro),p(Temp)
psi(Hab,ORV),p(Temp,Water, Season)
psi(Hab),p(Temp,Humid,Water, Season)
psi(Hab,Hydro),p(Temp,Humid)


AIC
123.99
124.05
124.14
124.30
124.31
124.50
124.55
124.65
124.66
124.86
124.92
124.94
125.08
125.12
125.43
125.77
125.77
126.03
126.04
126.14
126.17
126.31
126.39
126.65
127.00


Delta AIC AIC Weight
10.02 0.0021
10.08 0.0021
10.17 0.0020
10.33 0.0018
10.34 0.0018
10.53 0.0016
10.58 0.0016
10.68 0.0015
10.69 0.0015
10.89 0.0014
10.95 0.0013
10.97 0.0013
11.11 0.0012
11.15 0.0012
11.46 0.0010
11.80 0.0009
11.80 0.0009
12.06 0.0008
12.07 0.0008
12.17 0.0007
12.20 0.0007
12.34 0.0007
12.42 0.0006
12.68 0.0006
13.03 0.0005









Table 3-8. Sums of Akaike's Information Criterion (AIC) weights for all models including the
ORV index, habitat type, or hydrologic index covariates for each of the four focal
anuran species.
ORV Hydrologic
Species Index Habitat Index
Oak Toad 0.6819 0.8888 0.9988
Southern Toad 0.4989 0.2679 0.2755
Greenhouse Frog 0.9706 0.1218 0.7992
Eastern Narrow-mouthed Toad 0.8050 0.1267 0.7605









Table 3-9. Beta estimates, standard errors (S.E.), and lower and upper 95% confidence intervals
(C.I.) for the ORV use index covariate from the best model for each of the four focal
anuran species. NA indicates that numerical convergence could not be reached and
no estimate of the S.E. is available.
Beta Lower 95% Upper 95%
Species estimate S.E. C.I. C.I.
Oak Toad -580.9251 NA
Southern Toad 21.2321 24.0724 -25.9498 68.4140
Greenhouse Frog -3.9188 1.6130 -7.0803 -0.7573
Eastern Narrow-mouthed Toad -5.5601 6.1949 -17.7021 6.5819









Figure 3-1. An aerial photograph depicting off-road vehicle damage in marl prairie habitat in
Big Cypress National Preserve.











Figure 3-2. Map of amphibian occupancy sampling locations within BCNP during 2002-2003
(n=70).


Sampling Locations
Habitat Type
* Cypress
* Cypress Prairie
A Hammock
Pineland
+ Prairie









CHAPTER 4
THE EFFECT OF TOE-CLIPPING ON TWO SPECIES OF TREEFROGS

Introduction

Accurately assessing the status and trends of amphibian populations is an important part of

amphibian conservation and management, especially where amphibian species are threatened

with extinction or are of special interest to managers (Alford and Richards 1999, Stuart et al.

2004). Many studies rely on counts of amphibians to provide information on populations, but

count data not adjusted for detection cannot be used to monitor amphibian population status

(Schmidt 2003). Further, estimates of vital rates, such as survival probability, are crucial for

addressing the causes of declines and managing populations (Biek et al. 2002). One important

method for obtaining estimates of abundance and survival involves recapturing uniquely marked

individuals (Jolly 1965, Seber 1965).

Many marking methods have been developed for amphibians (Donnelly et al. 1994). To

be suitable for use in estimation of survival rates, a marking technique must be permanent, not

adversely affect the marked animal, and not affect the probability of capture on subsequent

samples (Williams et al. 2002). Few marking methods meet these assumptions when applied to

small anurans (<30 mm). Tagging methods such as passive integrated transponder (PIT)

microchips (Ireland et al. 2003) may be suitable for larger anurans, but are not useful for species

with small body size. Tattooing (Kaplan 1959), freeze branding (Daugherty 1976), and

fluorescent dye (Nauwelaerts et al. 2000) techniques are cumbersome for field use and may not

be permanent. Toe-clipping, the systematic removal of toes in unique combinations, is a low-

cost, efficient method of permanently marking anurans (Donnelly et al. 1994, Luddecke and

Amezquita 1999), but recent analysis suggests that toe-clipping may decrease survival of some









species of anurans (Parris and McCarthy 2001, McCarthy and Parris 2004) and some consider

the practice unethical or scientifically unsound (May 2004).

McCarthy and Parris (2004) observed a negative relationship between the return rate

(defined as the probability of survival times the probability of capture) of several species of

anurans and the number of toes removed during marking. The models of McCarthy and Parris

(2004) assume a constant probability of capture, and thus they conclude that survival rates are

lower for frogs with more toes removed. However, it is known that capture and survival

probabilities often vary with time due to environmental factors not related to the marking method

(Williams et al. 2002). Mark recapture analytical techniques make it possible to use the

information gained from uniquely marked animals to directly estimate survival and capture

separately and determine whether there is an effect of toe removal on survival or capture rates.

I applied capture-mark-recapture techniques to estimate survival and capture probabilities

of green treefrogs (Hyla cinerea) and squirrel treefrogs (H. squirella) in southern Florida, USA.

My objective was to determine if increasing the number of toes removed for marking had a

negative effect on survival or capture probability of these treefrog species. I hypothesized that

toe-clipping would have no effect on survival or capture probability in either species. I used

Cormack-Jolly-Seber open population mark-recapture models to estimate apparent survival and

capture probability (Lebreton et al. 1992), and information theoretic methods based on Akaike's

Information Criterion (AIC) for model selection (Burnham and Anderson 1998).

Methods

Study Site

I established six long-term research sites in Big Cypress National Preserve, Collier County,

FL, USA in April 2004. Each site consisted of 100-170 5.1-cm polyvinyl chloride (PVC) pipe

refugia (Boughton et al. 2000) erected from the ground in groups of 7 by 7 grids of 49 pipes with









5 m spacing and other pipes (6-9) located between grids to monitor movement between habitats.

The total number of refugia at all sites was 840.

Sites were located in pineland and cypress strand habitats, as well as adjacent marsh and

prairie habitats (Duever 2005). Pinelands are upland habitats dominated by slash pine (Pinus

elliottii). Cypress strands are seasonally flooded forested wetlands with dominated by bald

cypress (Taxodium distichum). Marsh habitats are long-hydroperiod forb-dominated wetlands,

and prairie habitats are short-hydroperiod sedge-dominated wetlands. Both marsh and prairie

lack a woody overstory (Duever et al. 1986). Captures from all habitats were combined in this

analysis to examine the effect of toe-clipping as preliminary analysis indicated that survival and

capture rate parameters were primarily homogeneous across habitats during the sampling period.

Capture Recapture

Refugia were checked once monthly during the period from 15 November 2004 to 30 June

2005. Frogs captured in or on the refugia were placed in clear plastic bags for measurement.

Individual frogs were identified to species, measured snout-to-urostyle length (SUL) in mm, and

examined for toe clip marks. Unmarked green treefrogs greater than 24 mm SUL and unmarked

squirrel treefrogs greater than 17 mm SUL were assigned a new clip number, unique to the site,

which required the removal of two, three, or four toes. It was difficult to mark and read marks

on smaller individuals, so they were not marked. The numbering system followed that of

Donnelly (1989) with the modifications that no more than one toe per foot was removed, and the

proximal toe on each forelimb was never removed. Toes were removed with stainless steel

scissors sterilized in alcohol. Recaptured frogs were examined for signs of toe regeneration, and

when necessary, toes were re-clipped.









Survival Analysis

I used the Cormack-Jolly-Seber mark-recapture model (Lebreton et al. 1992) as

implemented in program MARK (White and Burnham 1999) to perform survival analysis for

both species. Individuals of both species were divided into three groups for analysis: those with

two, three, and four toes removed. A series of 23 models representing different hypotheses about

the effects of time and group on apparent survival (q0) and capture probability (p) were fit for

both species (Table 4-1). Models including the effect of toe clipping were constructed so as to

force the effect to be monotonic (i.e. removing 4 toes had twice the effect of removing 3 toes

when compared to 2 toes).

Goodness of fit of the model structure was assessed by estimating the variance inflation

factor c using the parametric bootstrap method implemented in program MARK (White and

Burnham 1999) on the most general model in the model set, Ot*chpPt*clhp Model selection was

conducted using the information-theoretic approach of Burnham and Anderson (1998) with the

Quasi-likelihood Akaike's Information Criterion adjusted for over dispersion of data and small

sample sizes (QAICc).

Results

During the sampling period I captured, marked, and released a total of 1296 individual

green treefrogs and 658 individual squirrel treefrogs, of which 712 and 408 respectively were

subsequently recaptured. Return rates for frogs with 2, 3, and 4 toes removed declined

monotonically from 60.92% to 51.25% among Green Treefrogs and from 70.00% to 60.19%

among Squirrel Treefrogs (Table 4-2). At least 6 individuals of each species/toe-clip group were

captured and released during each of the first seven capture occasions (Figure 4-1).









The parametric bootstrap of the most general Cormack-Jolly-Seber model chosen for

Green Treefrogs produced an estimate of c = 1.656, indicating mild lack of fit of the data to the

model. Two models, 0t+cpt and Ot+cip t+cp, had a delta QAICc of less than 2, and therefore had

strong support (Table 4-3). Both of these models include the toe clip effect on survival, and one

includes a toe clip effect on capture. Models that included a toe-clip group effect on survival had

85.14% of the QAICc weight among the set of candidate models, and models that included toe-

clip group effect on capture probability had 34.16% of the model weight.

The estimated beta for the toe-clip effect on survival in green treefrogs from the best model

was -0.3963 (S.E. =0.1377; Table 4-4). There was a mean absolute decrease in survival of 5.02

% and 11.16% for frogs with 3 and 4 toes removed, respectively, compared to frogs with just 2

toes removed (Figure 4-2). The estimated beta for the toe-clip effect on capture probability in

green treefrogs was 0.1731 (S.E. =0.1270), but the 95% confidence interval included 0 (Table 4-

4).

For Squirrel Treefrogs, the estimate of c from the parametric bootstrap was 1.848. Four

models had delta QAICc values less than 2 (Table 4-5). Two of the top 4 models included the

toe-clip effect on survival and 2 included the toe-clip effect on capture. Models that included the

toe-clip effect on survival for squirrel treefrogs received 36.09% of the QAICc weight, and

models that included the toe-clip effect on capture probability accounted for 47.29% of the total

weight.

The estimated beta for the toe-clip effect on survival in Squirrel Treefrogs from the best

model that included it was 0.0231 (S.E. =0.1379), which is a slightly positive effect of toe

clipping, but the 95% confidence interval includes 0 (Table 4-4). The estimate of beta for the









toe-clip effect of capture on Squirrel Treefrogs from the best model that included it was -0.1815

(S.E. =0.1139), but the 95% confidence interval for this parameter also includes 0 (Table 4-4).

Discussion

I found strong evidence of a negative effect of toe removal on Green Treefrog survival, but

only very limited evidence of an effect of toe-clipping on Green Treefrog capture probability.

On average, frogs with 3 and 4 toes removed had 5% and 11% lower survival probabilities

respectively than Green Treefrogs with 2 toes removed. An effect of toe-clipping on capture in

Green Treefrogs is not supported due to low AIC weights for models with a toe-clip effect on

capture probability (Table 4-3) and a 95% confidence interval for the toe-clip effect beta that

includes 0 (Table 4-4).

There was also little evidence of an effect of toe-clipping on survival or capture probability

of Squirrel Treefrogs. This was due to low AIC weights for models that include toe-clip effects

(Table3-5) and beta values that had 95% confidence intervals that included 0 (Table3-4).

Average values for survival were equivalent regardless of the number of toes removed in

Squirrel Treefrogs. Thus, only one species, the Green Treefrog, was found to show a negative

response to toe-clipping.

McCarthy and Parris (2004) reported an estimated 4-11% reduction in the return rate of

frogs for each toe removed. The best models in McCarthy and Parris's (2004) analysis allow the

change in return rate to vary linearly with the number of toes removed. This model appears to fit

my results for Green Treefrog survival well, where there was 5-6% decrease in survival per toe

removed. However Squirrel Treefrogs did not show the same pattern. Although there did appear

to be a decline in the return rate of Squirrel Treefrogs (Table 4-2), there was no reduction in

survival due to removing more toes (Figure 4-2).









My results differ from those of McCarthy and Parris (2004) in the magnitude and

generality of the effect that was detected. Although return rates of both species in this study did

decrease monotonically with the number of toes removed (Table 4-1), the estimated effect on

survival was less than that expected from the results of McCarthy and Parris (2004). One reason

for this difference is the inclusion of estimates of capture probability in my analysis. Rather than

assuming constant capture probability, I directly estimated it. Using return rates alone without

accounting for heterogeneity in capture probability could lead to misinterpreting a reduction in

encounter rates as a reduction in survival. In addition, both capture probability and apparent

survival were allowed to vary with time in this study. Most of the best models for both species

include time dependence for both survival and capture probability.

It is apparent that all species of frogs do not show the same response to toe-clipping.

Some species appear to be especially susceptible to infections or loss of mobility due to toe

removal (Golay and Durrer 1994, Lemckert 1996). Even the two species at the same site in this

study showed a difference in the effect of toe-clipping on survival. Mark-recapture analysis

provides a robust method for estimating the effect of toe clipping on survival and capture

probability. It is preferable to using the return rate because it does not assume a constant capture

probability across time or toe clip treatment. Using mark-recapture modeling to estimate

survival and capture probabilities and using information-theoretic model selection to look for

effects of the marking method should provide a useful technique for testing the efficacy of toe-

clipping for other amphibian species.









Table 4-1. List of 23 models analyzed in Program MARK for captures of both Green Treefrogs
and Squirrel Treefrogs in Big Cypress National Preserve during 2004-2005.
Explanation defines each model in terms of the effects of time (t) and toe-removal
group (clip), on apparent survival (q5) and capture probability (p).
Model Explanation


O'Pt





OCh'PPt
OtPChjP



'/4chpPt
Otchp
O~h t ch~pp
'Ac/nPOchp
'A c/rnPp /r
O5POcilP




'/4 chpIt
tchlpp
OtktcirPt h
'Achp P~tcir
'A c/rnPt c/IP


Survival and capture probability constant throughout study
Survival varies with time; capture constant
Survival constant; capture varies with time
Both survival and capture vary with time
Survival is different among toe removal groups; capture constant
Survival constant; capture is different among toe removal groups
Survival and capture are both different among toe removal groups
Survival is group-dependent; capture varies with time
Survival varies with time; capture is group-dependent
Survival is a function of time, group, and their interaction; capture constant
Survival constant; capture is a function of time, group, and their interaction
Survival is interactive effect of time and group; capture varies with time
Survival varies with time; capture is interactive effect of time and group
Survival varies with toe group; capture is interactive effect of time and group
Survival is interactive effect of time and group; capture varies with toe group
Survival and capture are both an interactive effect of time and group
Survival is additive effect of time and group; capture constant
Survival constant; capture is additive effect of time and group
Survival is additive effect of time and group; capture varies with time
Survival varies with time; capture is additive effect of time and group
Survival varies with toe group; capture is additive effect of time and group
Survival is additive effect of time and group; capture varies with toe group
Survival and capture are both the additive effect of time and group









Table 4-2. The number of green treefrogs and squirrel treefrogs marked by removing 2, 3, or 4
toes in Big Cypress National Preserve, Collier County, FL, Nov. 2004-June 2005, the
number of individuals recaptured, and the return rate (proportion of marked
individuals recaptured at least once).
Species Treatment No. Marked No. Recaptured Return Rate
Green Treefrog 2 toes 87 53 60.92%
3 toes 848 474 55.90%
4 toes 361 185 51.25%
Squirrel Treefrog 2 toes 80 56 70.00%
3 toes 470 287 61.06%
4 toes 108 65 60.19%









Table 4-3 Model selection table for Cormack-Jolly-Seber open population mark-recapture model
of Green Treefrogs, including Quasi-likelihood Akaike's Information Criterion for
small sample sizes (QAICc), model weights based on QAICo, the number of
parameters n each model, and the model deviance. Model structure includes the
effects of time (t), and toe-clip group (clip) on apparent survival (q5) and capture
probability (p).


Num.
Par


QAICc
Weights


Model


OA chzp~t clzp

chpt







O4chp Pt


o~pt



VCIIpt+ II


VLPt clzp

Ok pt cp







VOchp Pchp

O4chp P

O.p~


QAICc

2944.1863

2945.0711

2946.8039

2952.4145

2958.9709

2962.4472

2964.9310

2968.9725

2972.6627

2972.7609

2973.1139

2974.8908

2976.2446

2976.7255

2978.7064

2985.5690

2986.9087

2987.4740

2987.6664

3392.0314

3394.4325

3404.7537

3406.7036


Delta
QAICc

0.0000

0.8848

2.6176

8.2282

14.7846

18.2609

20.7447

24.7862

28.4764

28.5746

28.9276

30.7045

32.0583

32.5392

34.5201

41.3827

42.7224

43.2877

43.4801

447.8451

450.2462

460.5674

462.5173


0.5182

0.3329

0.1400

0.0085

0.0003

0.0001

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000


QDeviance

2911.9877

2910.8476

2920.6711

2920.2161

2904.4171

2907.8932

2891.9520

2950.9070

2948.5491

2956.7085

2951.0178

2950.7771

2954.1485

2960.6730

2960.6409

2941.1988

2938.4695

2941.0701

2943.2964

3384.0168

3388.4238

3400.7494

3400.6947









Table 4-4. Estimates, standard error (SE), and the 95% confidence interval of the beta values for
the toe-clip effect on apparent survival (q0) and capture probability (p).

95% Confidence Interval
Species Parameter Beta SE Lower Upper
Green Treefrog 0 -0.3963 0.1078 -0.6075 -0.1851
P 0.1731 0.1270 -0.0759 0.4221
Squirrel Treefrog 0 0.0231 0.1379 -0.2472 0.2934
P -0.1815 0.1139 -0.4047 0.0417









Table 4-5. Model selection table for Cormack-Jolly-Seber open population mark-recapture
model of Squirrel Treefrogs, including Quasi-likelihood Akaike's Information
Criterion for small sample sizes (QAICc), model weights based on QAICo, the
number of parameters n each model, and the model deviance. Model structure
includes the effects of time (t), and toe-clip group (clip) on apparent survival (q5) and
capture probability (p).


Model


chp



kt chzp~t clzp








o~pt



otpt clzp

Ot*chp Pt

OV*chpP cp-





OLPt clzp

kchpPt ch1p

tchp Pt chzp

O4ch'pP

O.p~

O-cp IPch


QAICc

1653.9976

1654.0015

1655.3701

1655.7340

1656.0325

1656.3014

1656.7953

1657.6489

1663.6425

1664.1035

1664.1404

1666.0178

1668.7560

1670.3368

1672.2840

1675.9932

1678.5895

1681.6591

1684.2103

1716.5722

1716.7540

1716.7772

1717.8694


Delta
QAICc

0

0.0039

1.3725

1.7364

2.0349

2.3038

2.7977

3.6513

9.6449

10.1059

10.1428

12.0202

14.7584

16.3392

18.2864

21.9956

24.5919

27.6615

30.2127

62.5746

62.7564

62.7796

63.8718


QAICc
Weights

0.2481

0.2476

0.1249

0.1041

0.0897

0.0784

0.0613

0.0400

0.0020

0.0016

0.0016

0.0006

0.0002

0.0001

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000


Num.
Par

16

10

16

17

11

9

14

12

8

11

9

12

26

25

21

23

21

23

35

3

2

3

4


QDeviance

1621.6121

1633.8462

1622.9846

1621.2999

1633.8461

1638.1745

1628.4980

1633.4284

1647.5410

1641.9171

1646.0135

1641.7974

1615.7538

1619.4096

1629.6268

1629.2069

1635.9323

1634.8728

1612.4000

1710.5552

1712.7456

1710.7604

1709.8412














600
o

-a 500 -*- 2 Toes
D .... 0.... 3 Toes
3 -y-- 4 Toes

S0
C 400



300 -

0 ....


S/ \

Z 100 / \
.. IV --v-----V



600

B
-o 500 -*- 2 Toes
0) ....0.... 3 Toes
-3 -- 4 Toes
Q_
0 400


300 -

4-

200 0.......

3 ...o....
Z 100 ..


o---T-r-------r-r--

0 1 2 3 4 5 6 7 8

Sampling Occasion


Figure 4-1. Numbers of individuals of green treefrogs (A) and squirrel treefrogs (B) captured
and released in each of the three toe removal groups during the first 7 sampling
occasions.














1.0

A

0.8 -



0.6 -
U)


0.4
Q-
Q_


0.2 -- 2 Toes
--0- 3 Toes
-y-- 4 Toes

0.0 --
1.0



0.8 -


-FU
L 0.6 -
U3



S0.4 -



0.2- Toes
-v-- 4 Toes


0.0 i
0 1 2 3 4 5 6 7

Sampling Interval


Figure 4-2. Apparent survival (q5) and 95% confidence interval of green treefrogs (A) and
squirrel treefrogs (B) in each toe removal group category across the first 6 monthly
sampling intervals. Estimates for squirrel treefrogs are averaged across models as no
model had a majority of the QAICc weight (Burnham and Anderson 1998).









CHAPTER 5
INFLUENCE OF HYDROLOGY ON SURVIVAL AND RECRUITMENT OF GREEN
TREEFROGS

The Everglades ecosystem of southern Florida has been substantially altered over

the last 100 years by loss to agriculture and urbanization. Compartmentalization of the

remaining Everglades into a network of artificially controlled impoundments has

impeded historic flow patterns (Davis et al. 1994). A large-scale restoration effort, the

Comprehensive Everglades Restoration Plan (CERP) was devised to attempt to restore

natural hydrologic regimes to the remaining Everglades (DeAngelis et al. 1998). One

measure of restoration success was defined as recovering ecological structure and

function to the natural areas. Consequently, managers charged with decision-making for

CERP need species that can serve as indicators of ecosystem restoration success.

Amphibians have been used in various locations as ecosystem indicator species

(Welsh and Ollivier 1998, Galatowitsch et al. 1999, Sheridan and Olson 2003). Aspects

of their natural history (e.g. aquatic larval phase, permeable skin, and low dispersal

ability) make them potentially well-suited as indicator species in the Everglades as in

other systems. However, there is no historical record of amphibian populations from

before hydrologic alteration in the Everglades system to use for comparison to current

and post-restoration populations. Likewise, it would be extremely difficult to

experimentally manipulate environmental conditions at the scale necessary to gauge the

response of amphibians to hydrologic restoration.

By monitoring amphibian populations and measuring how they respond to

environmental changes at a local scale, it should be possible to make predictions about

how amphibians will respond to Everglades restoration on a landscape scale. As the

major goal of CERP is restoration of hydrology, it is important to know how amphibians









will respond to changes in water depth and duration if amphibians are to be used as

indicator species of restoration success. This response may be measured in population

vital rates, and mark-recapture techniques allow the estimation of vital rates. Survival

and recruitment rates can be estimated using open population mark-recapture analysis,

and the population growth rate of populations can be derived from estimates of these two

parameters (Pradel 1996).

The goal of this study is to determine effects of seasonal changes in hydrology on

the population vital rates of Green Treefrogs (Hyla cinerea) in the Everglades ecosystem

to provide information on how hydrologic restoration in the Everglades might impact

frog populations. Modeling recruitment and survival with mark-recapture analysis will

help build a model of how populations should respond to anthropogenic changes in

hydrology. The contributions of survival and recruitment to population growth across

seasons will be determined within closely connected habitats. The contributions may be

important for determining critical time periods for reproduction at different water depths.

I hypothesized that both survival and recruitment are dependent on water depth and

hydrologic season as well as time. I also hypothesized that capture probability is

dependent on season. The information gained in this study can later be used to build

models predicting the response of frog populations to various scenarios of hydrologic

restoration of CERP.

Methods

Three long-term study sites were established in Big Cypress National Preserve,

Collier County, FL for this study. Each of these sites was placed at a location where 3

habitats, cypress strand, broad-leaf marsh, and short-hydroperiod prairie, were in close

proximity. Habitats in the Everglades ecosystem are primarily different due to small









topographical differences that create marked differences in hydrology (McPherson 1974,

Duever et al. 1986). Broadleaf marshes have the longest hydroperiod, and never

completely dried during this study (Figure 5-1). Marsh sites are comprised of tall (1-2 m)

emergent forbs, especially Pickerelweed (Pontedaria cordata), and lack a woody

overstory. Cypress strands are intermediate in hydrology (Figure 5-1), and are distinct

from the other habitats because of their closed canopy of bald cypress (Taxodium

distichum). Prairie, sometimes referred to as marl prairie or dry prairie, has the shortest

hydroperiod (Figure 5-1). These sites are characterized by a sedge dominated flora

usually up to 1 m in height (Duever et al. 1986).

A grid of 49 polyvinyl chloride (PVC) pipe refugia (Moulton et al. 1996, Boughton

et al. 2000) was arranged within each habitat stratum at each site. Pipes were arranged 5

m apart in a 7 by 7 grid. Grids were located completely within a habitat stratum, but

within 30 m of the adjacent grids in the other habitats. PVC pipes used in this study were

50 mm in diameter and 1 m long. Each pipe was placed vertically onto a wooden stake

driven into the soil so the pipe could be easily lifted for inspection. All pipes in each grid

were numbered for reference. Sites were sampled biweekly from April 2004 to

November 2004 and once monthly from December 2004 to August 2005 (n=25). Water

depth was measured at a fixed depth gauge in the center of each plot during each

sampling occasion.

At each sample, all pipes were checked for frogs which were captured in sealable

plastic bags. Frogs that escaped were noted and identified to species if possible.

Captured frogs were identified to species and measured snout to urostyle (SUL). Green

treefrogs less than 25 mm SUL were returned unmarked as frogs of this size were









difficult to mark and read marks reliably. Previously unmarked frogs 25 mm SUL and

larger were assigned a unique toe-clip combination (Donnelly et al. 1994) and marked by

removing toes using stainless steel scissors sterilized in alcohol. Previously marked

(recaptured) frogs were checked against a list of previously marked frogs to insure the

clip was read correctly. The pipe number and grid location of each frog capture was

noted, and frogs were released into the pipe opening immediately after necessary

handling was completed.

Capture-recapture data were analyzed using the temporal symmetry approach of

Pradel (1996) with the y-parameterization (0!, pt, y; Williams et al. 2002) in Program

MARK (White and Burnham 1999). The temporal symmetry approach uses reverse-time

mark recapture analysis (Pollock et al. 1974, Pradel 1996) to estimate the seniority

parameter, y, (i.e. the probability that an individual alive at time i was also alive at time i

-1). Although the parameter of interest in this study is actually the population growth rate

(, ), the y-parameterization of the Pradel (1996) model is known to perform better than

alternative parameterizations including A,, and A, may be derived from the estimates of

the k, and q, parameters (Williams et al. 2002). The temporal symmetry method uses

assumptions similar to standard open-population Cormack-Jolly-Seber mark-recapture

models (Lebreton et al. 1992). These assumptions include that no marks are lost or

misread and that no non-random temporary emigration occurs. Every marked animal

should have the same probability of capture in and survival to the next sampling period,

and all individuals should have independent fates (Williams et al. 2002).

In the analysis, water depth was a covariate and captures were grouped by habitat

and season. Mean water depth within each habitat across the 3 sites was standardized so









that the mean fell between 0 and 1 as the logit link function was used in MARK analyses

(Williams et al. 2002). Individual frogs were categorized as belonging to cypress, marsh,

or prairie habitat groups. Although it was rare, some frogs were caught in one habitat and

later moved to an adjacent habitat. Frogs that were captured in more than one habitat

were assigned to the group with the most captures, or the last capture if caught an even

number of times in two habitats. Sampling occasions were grouped into 5 "seasons"

based on the annual pattern of rainfall (Table 5-1). These seasons were used in the

models in place of full time dependence because they are more biologically meaningful

than date alone.

There are three parameters estimated in this temporal symmetry model: survival

(qb), capture probability (p), and the seniority parameter (y). Models representing

different combinations of the water depth covariate, habitat group, and season (time)

structure were constructed that were a priori determined to be biologically meaningful.

Model selection was conducted using the information-theoretic approach based on

Akaike's information criterion adjusted for small sizes (AICo; (Burnham and Anderson

1998). A, during each sampling interval was derived in Program MARK using the

equation given in Williams et al. (2002):

Al =0/1

Nichols et al. (2000) demonstrate the relationship between the parameter of the Pradel

(1996) temporal symmetry modeling approach and relative contributions of survival and

recruitment to the change in population growth. The contributions of survival and


recruitment to are analogous in this case to 7' and 1- I', respectively (Nichols et al.

2000).









Results

During the 17 months of sampling, a total of 1069 individual Green Treefrogs were

marked and 1054 recaptures were recorded. There were 293 times that a frog escaped

capture, and 173 frogs less than 25 mm SUL were released unmarked. Frog capture rates

were higher in the dry season months, peaking in February and March 2005 (Figure 5-2),

and frog captures were lowest during the peak of the wet seasons (August 2004 and June

2005). Sizes of captured frogs showed a seasonal pattern (Figure 5-3). Mean SUL was

lowest in August 2004 after the onset of the wet season, with a similar dip in July 2005

after the beginning of the 2005 wet season. More frogs were captured in cypress and

marsh habitats than in prairie, but habitat differences were only observed during the dry

months (Figure 5-4).

Model selection results from the mark-recapture modeling indicated that the best

model included the seasonal time structure, habitat group, and water-depth covariate for

q! and y, and only habitat and season forp (Table 5-2). No other models had any AICc

weight (Burnham and Anderson 1998). The habitat and season-specific values for q!

varied widely, but were generally around 0.80 (Figure 5-5). Estimates of y also varied

widely (Figure 5-6). Although season was included in the model for q! and /, a seasonal

trend in either was not apparent, but a pattern might be somewhat obscured by the water

depth covariate. There does appear to be a strong seasonal trend inp (Figure 5-7).

Capture rates were lowest at the sampling occasions corresponding to the wet seasons of

2004 and 2005 (Table 5-1). A, values across sampling intervals and habitats fluctuated

from 0.22 to 2.02 (Figure 5-8). The contribution to A from survival was almost always

greater than the contribution of recruitment in each habitat (Table 5-3).









Discussion

The mark-recapture modeling results support the hypothesis that Green Treefrog

survival and seniority varies with season and with water level, as well as among habitats.

Capture probability also varied with season and habitat, but was not affected by water

depth. Capture rates declined during the early wet seasons and capture probability was at

minimum levels during the wet season. The mean size of individuals captured declined

sharply during the wet seasons, presumably as young of the year individuals enter the

population. Population growth rates were highest after the onset of the 2004 wet season

and during December 2004 and January 2005 (Table 5-3). These two time periods also

showed increased contributions to A, from recruitment (1- ).

Recruitment as it is modeled in this study includes animals that are enter the

population from reproduction as well as immigrants into the population, and the models

used can not differentiate the two (Nichols et al. 2000). Recruitment in this study may

also include animals moving into the PVC pipe refugia for the first time, as the

population studied is actually the population using the PVC pipes. The early wet season

increase in recruitment was most likely a result of reproduction. This increase coincides

directly with the large drop in mean SUL of captured frogs (Figure 5-3), and large

choruses of calling males and Green Treefrogs in amplexus were observed at all of the

sites at the onset of the wet season (personal observation). The increase in recruitment in

December 2004 and January 2005 is likely the result of cold weather causing frogs to

seek refuge in the PVC pipes. Increased capture rates in PVC refugia at times of cold

weather have been observed elsewhere in Florida (Donnelly et al. 2001, Zacharow et al.

2003). Movements of frogs between habitats were very rare and only account for about









5% of all captures (unpublished data). It therefore seems unlikely that immigration into

the study sites was a factor in recruitment rates.

One important assumption of the mark-recapture analysis used in this study is that

there is no non-random temporary immigration. Immigration into the study sites does not

appear to have occurred due to the very low movement rates between habitats. One

potential violation of the temporary immigration assumption is the increased use of pipes

during the colder samples. If the frogs captured and marked in these samples later

emigrated back out of the pipe population it would affect estimates of survival. However,

capture rates were actually higher in the months following the December and January

increase in recruitment (Figure 5-2) when population growth was not increasing (Table 5-

3). It appears from this pattern that once frogs moved into the PVC pipe population they

remained in the population, therefore this would not be a case of temporary immigration.

Green Treefrog populations in BCNP fluctuate on a strongly annual cycle driven by

the hydrologic cycle in the Everglades ecosystem. The majority of reproduction occurs at

the onset of the wet season. The number of frogs in the population slowly declines until

late in the dry season when few large, adult frogs remain to accomplish breeding at the

onset of the next wet season. The mark-recapture modeling results from this study (Table

5-2) corroborate that survival and seniority rates are seasonal and related to water depth.

The average monthly survival rate for green treefrogs across habitats and months is

81.0%, or only 8.0% annually. Few published survival rates of frogs exist for

comparison, but these estimates appear to be low in comparison to estimates of survival

of the Afro-Tropical Pig-Nosed Frog (Grafe et al. 2004) and the pig frog (Wood et

al. 1988).









The seasonal patterns observed in capture probability as well as recruitment into

the population demonstrate the importance of long-term monitoring of sites. Studying

the population during a single season could lead to incorrect conclusions about trends in

the population. In addition, this study has demonstrated that differences in vital rates can

exist between adjacent sites in different habitats. This suggests that population level

changes may be occurring at a very small scale, and probably have a strong relationship

to hydrology. Capture probability was not a function of water depth (Table 5-2), but

there was a strongly seasonal pattern to capture probability (Figure 5-7) and to capture

rates (Figure 5-2).

Although prairie habitat usually had the lowest rate of population growth (Figure 5-

8) across sampling intervals, it almost always had the highest values for contribution to

growth from recruitment (Table 5-3). This is noteworthy because prairie is the habitat

with the shortest hydroperiod, meaning it is likely to be the most impacted by hydrologic

restoration. Large areas of the eastern Everglades are comprised of a similar short-

hydroperiod graminoid prairie, the rocky glades (Davis et al. 2005). If the spatial and

temporal pattern of hydrology in the rocky glades area is altered during CERP

restoration, an effect on Green Treefrog populations should be detected. For example, an

extension of the annual period of inundation could increase Green Treefrog reproduction.

Monitoring of Green Treefrog populations may provide an efficient means for indicating

if the vital ecological processes associated with the seasonal pattern of inundation are

functioning appropriately after hydrologic restoration. Prior to this study, this process

was not well understood, and consequently, the proposed actions in CERP have not been

evaluated with respect to effects on amphibian reproduction.









Table 5-1. Dates of each sample of the PVC pipe refugia and the season to which each
sample was assigned for mark-recapture analysis.
Sample Date Season
1 4/23/04 Dry season 2004
2 5/7/04 Dry season 2004
3 5/19/04 Dry season 2004
4 6/5/04 Dry season 2004
5 6/18/04 Dry season 2004
6 6/30/04 Dry season 2004
7 7/15/04 Wet season 2004
8 7/29/04 Wet season 2004
9 8/10/04 Wet season 2004
10 8/27/04 Wet season 2004
11 9/9/04 Wet season 2004
12 9/23/04 Wet season 2004
13 10/13/04 Wet season 2004
14 10/21/04 Wet-dry transition
15 11/5/04 Wet-dry transition
16 11/22/04 Wet-dry transition
17 12/15/04 Wet-dry transition
18 1/6/05 Dry season 2005
19 2/3/05 Dry season 2005
20 3/7/05 Dry season 2005
21 4/15/05 Dry season 2005
22 5/18/05 Wet season 2005
23 6/15/05 Wet season 2005
24 7/22/05 Wet season 2005
25 8/17/05 Wet season 2005








Table 5-2. Model selection results for all models analyzed in Program MARK. Model
describes the covariates and groups associated with q! (survival), p (capture
probability), and y (seniority parameter): s is the seasonal time structure
(Table 5-1), h is the habitat grouping (different estimates among the habitats),
and w is the water depth covariate. Akaike's information criterion adjusted
for small sample sizes (AICc), Delta AICc (the difference between the AICc
for a model and the AICc for the model with the lowest AICc), the AICc
weight (Burnham and Anderson 1998), the number of parameters in each
model and the model deviances are given.
Delta AICc Num.
Model AICc AIC, Weight Par Deviance
S(s+h+w), p(s+h), y (s+h+w) 9496.89 0.00 1 39 1420.46
S(s+h+w), p(s+h+w), y (s+h+w) 9522.16 25.27 0 41 1441.57
S(s+h+w), p(s), y(s+h+w) 9613.15 116.26 0 32 1551.21
S(s+h), p(s+h), y(s+h+w) 9665.58 168.70 0 41 1585.00
S(s+h), p(s+h+w), y(s+h+w) 9665.98 169.09 0 42 1583.31
S(s+h), p(s+h), y(s+h) 9670.52 173.64 0 44 1583.69
S(s+w), p(s+w), y (s+w) 9700.29 203.40 0 18 1667.03
S(h), p(s+h), y(s+h+w) 9745.69 248.80 0 31 1685.80
S(s+h), p(s+h), y(h) 9750.95 254.06 0 32 1689.01
q (h), p(s+h), y (s+h) 9751.66 254.77 0 33 1687.65
S(h+w), p(h+w), y(h+w) 9877.47 380.58 0 12 1856.39
S(h), p(s+h), y(h) 9956.30 459.42 0 21 1916.93
0 (s), p(s), y(s) 9988.10 491.21 0 15 1960.94
S(s), p(s), y (s+w) 9989.80 492.92 0 16 1960.61
4(s+h), p(h), /(h) 10191.78 694.89 0 20 2154.45
S(w), p(w), y(w) 10584.90 1088.01 0 3 2581.95
4(h), p(h), /(h) 11291.75 1794.86 0 9 3276.73
q (.),(.), y(.) 11428.49 1931.60 0 3 3425.55









Table 5-3. Estimates of the population growth rate ,, the contribution of survival to A, ( ,), and the contribution of recruitment to A,
(1- ) and the standard error (S.E.) of each estimate for each sampling interval ending on the date given.
Cypress Marsh Prairie
Date ., (S.E. [ ]) ,(S.E.[ ]) 1 f, (S.E.[1 ,]) (S.E.[ ]) y,(S.E.[]) 1 f, (S.E.[1 ,]) (S.E.[ ]) y,(S.E.[ ]) 1 (S.E.[1 ])
5/7/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.07 (0.049) 0.90(0.044) 0.10 (0.044) 0.99 (0.084) 0.74(0.075) 0.26 (0.075)
5/19/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076)
6/5/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076)
6/18/04 0.89 (0.067) 0.79(0.050) 0.21 (0.050) 0.65 (0.134) 0.92(0.039) 0.08 (0.039) 1.02 (0.083) 0.73(0.076) 0.27 (0.076)
6/30/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.24 (0.090) 0.80(0.058) 0.20 (0.058) 1.02 (0.083) 0.73(0.076) 0.27 (0.076)
7/15/04 1.06 (0.056) 0.77(0.052) 0.23 (0.052) 1.25 (0.096) 0.79(0.061) 0.21 (0.061) 1.02 (0.083) 0.73(0.076) 0.27 (0.076)
7/29/04 1.78 (0.205) 0.56(0.064) 0.44 (0.064) 1.55 (0.187) 0.64(0.077) 0.36 (0.077) 1.34 (0.135) 0.74(0.075) 0.26 (0.075)
8/10/04 1.39 (0.073) 0.66(0.040) 0.34 (0.040) 1.27 (0.071) 0.74(0.049) 0.26 (0.049) 1.72 (0.178) 0.54(0.076) 0.46 (0.076)
8/27/04 1.39 (0.073) 0.66(0.040) 0.34 (0.040) 0.86 (0.103) 0.78(0.044) 0.22 (0.044) 1.04 (0.139) 0.59(0.074) 0.41 (0.074)
9/9/04 0.76 (0.126) 0.71(0.046) 0.29 (0.046) 0.72 (0.116) 0.79(0.044) 0.21 (0.044) 0.94 (0.154) 0.59(0.074) 0.41 (0.074)
9/23/04 1.20 (0.070) 0.68(0.045) 0.32 (0.045) 1.20 (0.064) 0.75(0.047) 0.25 (0.047) 1.55 (0.119) 0.55(0.075) 0.45 (0.075)
10/13/04 1.35 (0.068) 0.66(0.046) 0.34 (0.046) 1.25 (0.067) 0.75(0.048) 0.25 (0.048) 1.69 (0.168) 0.53(0.076) 0.47 (0.076)
10/21/04 1.37 (0.070) 0.66(0.046) 0.34 (0.046) 1.24 (0.066) 0.75(0.048) 0.25 (0.048) 1.55 (0.119) 0.55(0.075) 0.45 (0.075)
11/5/04 0.64 (0.132) 0.75(0.049) 0.25 (0.049) 0.75 (0.226) 0.81(0.057) 0.19 (0.057) 0.64 (0.112) 0.74(0.053) 0.26 (0.053)
11/22/04 1.23 (0.077) 0.71(0.040) 0.29 (0.040) 1.17 (0.088) 0.78(0.055) 0.22 (0.055) 1.08 (0.069) 0.71(0.055) 0.29 (0.055)
12/15/04 1.48 (0.069) 0.66(0.032) 0.34 (0.032) 1.31 (0.091) 0.75(0.053) 0.25 (0.053) 1.19 (0.077) 0.70(0.056) 0.30 (0.056)
1/6/05 1.77 (0.119) 0.56(0.038) 0.44 (0.038) 1.68 (0.200) 0.59(0.070) 0.41 (0.070) 1.28 (0.090) 0.69(0.058) 0.31 (0.058)
2/3/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 0.94 (0.100) 0.79(0.036) 0.21 (0.036) 1.01 (0.050) 0.67(0.040) 0.33 (0.040)
3/7/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 1.44 (0.076) 0.69(0.036) 0.31 (0.036) 1.01 (0.050) 0.67(0.040) 0.33 (0.040)
4/15/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 0.24 (0.028) 0.83(0.041) 0.17 (0.041) 1.01 (0.050) 0.67(0.040) 0.33 (0.040)
5/18/05 0.86 (0.020) 0.87(0.017) 0.13 (0.017) 1.47 (0.087) 0.67(0.040) 0.33 (0.040) 1.01 (0.050) 0.67(0.040) 0.33 (0.040)
6/15/05 1.00 (0.000) 1.00(0) 0(0) 0.45 (0.142) 0.46(0.121) 0.54 (0.121) 2.02 (0.398) 0.41it ,,sS) 0.60 (0.088)
7/22/05 1.00 (0.000) 1.00(0) 0(0) 0.70 (0.196) 0.44-Ik, 120) 0.56 (0.120) 0.49 (0.114) 0.47(0.093) 0.53 (0.093)
8/17/05 1.00 (0.000) 1.00(0) 0 (0) 0.22 (0.086) 0.48(0.122) 0.52 (0.122) 0.49 (0.114) 0.47(0.093) 0.53 (0.093)


















60 .
00



0) 40 -
4O











Apr Aug Dec Apr Aug

Month (2004-2005)


Figure 5-1. Mean water depth across the 3 sampling locations in BCNP in the cypress strand,
broadleaf marsh, and prairie habitats from April 2004-August 2005.














-*-Captures
--*-. Marsh Depth *

500 -




400




300




200




100


'U


U...


.3-








U-




Ul
U-
--....


N N \ \ \ Q z S**
Date

Figure 5-2. Number of captures of Green Treefrogs during each of the 25 samples from April
2004 to August 2005 (solid line) and mean water depth (in cm) of marsh plots (dotted
line) across all 3 sites in BCNP.


70


60


50 &


40


30


20



10


0














45

40

35

30

6 25

20

15

10

5

0

/ ^N' N ....

Date

Figure 5-3. Mean snout-to-urostyle (SUL) in mm of Green Treefrogs captured in all habitats at
the three sites at each sampling occasion in BCNP.












250


-*-- Cypress
200 O Marsh
-yT- Prairie


150



100 -


0
50 --




Apr Aug Dec Apr Aug

Month (2004-2005)


Figure 5-4. Number of captures by sample of Green Treefrogs in cypress, marsh, and prairie
habitats at the three sites in BCNP.


















LU 0.4-

0.2- -

0 .0 I I I I I I I I I I I I I I I I I I I I I I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Interval
/ Cypress Marsh 0 Prairie

Figure 5-5. Estimates with 95% confidence intervals of apparent survival (0) of Green
Treefrogs for each sampling interval for cypress, marsh, and prairie habitat in BCNP.


1.0-r
















EU 0.6--

0.4--

0.2

0 .0 I I I I I I I I I I I I I I I I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
Sampling Interval
/ Cypress Marsh 0 Prairie

Figure 5-6. Estimates with 95% confidence intervals of seniority (7) of Green Treefrogs for
each sampling interval for cypress, marsh, and prairie habitat in BCNP.


1.0-r














0, 1) UT-I

T T T T T
LU 0.4 -



0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Sampling Occasion
/ Cypress Marsh 0 Prairie

Figure 5-7. Estimates with 95% confidence intervals of capture probability (p) of Green
Treefrogs at each sampling occasion for cypress, marsh, and prairie habitat in BCNP.













-*- Cypress
- a Marsh
-- Prairie


Figure 5-8. Derived estimates of population growth (,) of Green Treefrogs for each sampling
interval for cypress, marsh, and prairie habitat in BCNP.









CHAPTER 6
CONCLUSION

Introduction

In the introductory chapter to this dissertation I outlined the characteristics of useful

indicator species and described why amphibians should be suitable as ecosystem indicators. In

this conclusion, I will address each of the major characteristics of indicator species and describe

whether the results of this research support the use of amphibians as indicators in the Everglades

ecosystem.

Characteristics of Indicators

Abundant and Efficient to Sample

In order to be useful as an indicator, a species must be abundant and/or cost-effective to

sample. Amphibians in south Florida clearly meet this qualification. Sampling amphibians

using visual encounter techniques, as in Chapter 3, was a very efficient way to sample several

species at once with just 1 person hour per sample. This technique worked in a variety of

habitats across a very large geographical area. Another sampling technique, PVC pipe refugia

(Chapter 5), was also effective. This method involved more intensive work at fewer sites, but

sample size was adequate to estimate population vital rates including survival and population

growth. Collecting sufficient data to estimate these vital rates on many other species would

require much more effort. Therefore, I conclude that amphibians are sufficiently abundant and

easy to sample to be useful as indicators.

There were also caveats to amphibian sampling demonstrated in this dissertation. Results

from a known population of artificial frogs illustrated severe bias in the sampling of amphibians

using traditional methods, despite efforts to reduce observer bias (Chapter 2). For this reason,

methods using counts of amphibians as an index should be investigated closely before they are









adopted as part of a monitoring program. In Chapter 4, the Green Treefrog was shown to be

negatively affected by a common and efficient marking method, toe-clipping. The effect was

small and may be manageable for some studies, depending on the questions involved. The

important lesson from these studies is that methods for sampling indicator species must be

thoroughly evaluated before designing a monitoring program.

Sensitive to Stresses on the System

As previously mentioned, a species is only useful as an ecological indicator if it is sensitive

to stresses to the ecosystem and responds to stress in a predictable manner. Chapter 3

demonstrated that four species of ground-dwelling anurans were sensitive to an index of off-road

vehicle (ORV) use. Occupancy rates of three of the species had a negative relationship with

ORV use, as was predicted. Occupancy of another species, the Southern Toad, showed a

positive association with ORV use. Although this did not follow the prediction, there are

morphological differences between Southern Toads and the other anurans that may explain the

different response.

Green treefrog populations demonstrated sensitivity to an ecosystem process in Chapter 5.

Water depth and season of year were included in the best model of survival and recruitment. A

period of high population growth was shown to coincide with the onset of the wet season. This

project was not manipulative, so we can only hypothesize how treefrogs would respond to

various changes in the temporal and spatial pattern of hydrology. This could easily be

investigated with additional monitoring at different sites. Further monitoring of anurans in

relation to ORV use and treefrog populations in relation to hydrology will refine our knowledge

of their response to these stressors and improve their use as indicator species.









Responses to Stress Should Be Anticipatory

To be useful to managers, indicator species should display a response to local changes that

is anticipatory of an impending change to the whole system. These responses are most useful

when they predict impacts that can be averted by management before the whole system is

negatively impacted. Treefrog populations might serve as sentinels for the health of the

Everglades ecosystem with regard to hydrologic restoration. Because of their annual cycle and

short generation time (Chapter 5), we know treefrogs will respond faster to changes in the

hydrology of a site than vegetation, for example. If monitoring of treefrogs shows that treefrogs

are responding as would be expected (i.e., the treefrog population is changing to resemble other

populations with the same hydrology), it is reasonable to assume that restoration is successful.

In the case of ORV use in Big Cypress (Chapter 3), it is impossible to know if anuran occupancy

responded anticipatorily to ORV use, as ORVs have been used for decades in that area.

Integrate a Response across the Whole System

Indicators are most useful when they can respond to changes in the whole system, rather

than just in a few habitats or locations. Amphibians are clearly suitable for detecting changes to

the whole Everglades system. They are found in all of the terrestrial and non-marine aquatic

habitats in south Florida. Amphibians responded across many habitats to ORV use in Chapter 3.

Many habitats in South Florida are defined by small differences in hydropattern resulting in very

different vegetation communities. It is clear that treefrogs respond to both vegetation (habitat)

and hydrology (Chapter 5).

Known Response to Anthropogenic Stresses and Natural Disturbances

In many situations it is important to be able to determine if a population of an indicator

species is responding to anthropogenic changes (e.g., management actions) or natural

disturbances. This would be especially useful in the Everglades ecosystem where rainfall can be









extremely variable and hydrology at a site is a function of both natural phenomena (e.g.,

hurricanes) and water management. If a species is an indicator of hydrologic conditions, it might

respond in a similar manner whether the change was a result of natural rainfall patterns or

anthropogenic management actions. Two types of indicator species might be useful for

differentiating the effects of natural and anthropogenic changes to the system: those that react

very quickly and those that respond over decades or longer. Species that respond very quickly

(6-12 months after a change to the system) can be monitored along with environmental variables

to determine their response to specific management actions or natural events. Species that

respond over very long time periods will integrate the effects of natural disturbances and should

indicate the effects of management actions. Species that respond quickly will be most useful for

adaptive management in Everglades restoration and species that respond over long periods will

be indicative of restoration success.

Amphibians appear to be useful as indicators of both short and long-term effects. Green

Treefrog populations will respond immediately to changes in the hydrologic cycle. If a wet

season at a site is different because of a natural event or a management action, the treefrogs

should respond immediately by increasing or decreasing reproduction. In the case of long-term

anthropogenic changes, anurans were useful as indicators of ORV use, even years after use was

suspended in some areas (Chapter 3). In this way, anurans were indicators of anthropogenic

disturbance across a heterogeneous landscape subject to many natural disturbances.

Conclusion

This dissertation provides evidence that amphibians are suitable as ecosystem indicators in

general, and specifically as indicators of restoration success in the Everglades ecosystem of

southern Florida. As with any group of indicators, care should be used when choosing sampling

methods and particular species to use as indicators. Continued monitoring and additional