|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 SITE OCCUPANCY OF ANURANS IN THE ARTHUR R. MARSHALL LOXAHATCHEE NATIONAL WILDLIFE REFUGE By RYAN LORING LYNCH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREME NTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 20 12
2 2012 Ryan Loring Lynch
3 To my family: Thank you for all your guidance and support throughout the years, and for encouraging me to follow my passion regardless of where it take s me To Pamela: Thank you for your love, encourageme nt, and patience over the years
4 ACKNOWLEDGMENTS I thank my advisor, Dr. Frank Mazzotti, and committee member, Dr. Laura Brandt, for providing me the opportunity to work in such an amazing ecosyst em with such incredible animals over the span of the past 6 years. Without your help and guidance I would not be in the position I am in today. I also thank my committee member, Dr. Wiley Kitchens, and fellow Florida Fish and Wildlife Cooperative Research Unit leader, Dr. Franklin Percival, for their gained many great friends, most of which have spent a considerable number of exhausting and exhilarating hours waist d eep in the marsh with me. Thanks go out to the following people for their continued friendship and hard work : Mike Rochford, Brian Jeffrey, Mike Cherkiss, Jeff Beauchamp, Rafael Crespo, Ed Larrivee, Mat Denton, Wellington Guzman, Danielle Ogurcak, Joy Vinc i, Trey Kieckhefer, Toren Hill, Emily Pifer, Michelle Curtis Sara Williams, and Joy Vinci S pecial thanks goes out to Hardin Waddle and Ikuko Fujisaki for their advice and assistance during the design and analysis phases of this project without your sta tist ical expertise I would have pulled out my hair by now I would also like to thank Pamela Telis and her crew at EDEN for helping me work through the EDEN tools and offering advice on all issue s hydrological. The staff of the Arthur R. Marshall Loxahatch ee National Wildlife Refuge also deserves special thanks for their logistical support throughout this project and for allowing me the unique opportunity to play with frogs in the Refuge. Finally, I thank both my family and fianc e for their unending love and support, and for encouraging me to follow my dreams.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ........... 4 LIST OF TABLES ................................ ................................ ................................ ...................... 6 LIST OF FIGURES ................................ ................................ ................................ .................... 7 LIST OF ABBREVIATIONS ................................ ................................ ................................ ...... 8 ABSTRACT ................................ ................................ ................................ ............................... 9 CHAPTERS 1 INTRODUCTION ................................ ................................ ................................ ............. 11 2 METHODS ................................ ................................ ................................ ........................ 15 Study Area ................................ ................................ ................................ ......................... 15 Sampling Design ................................ ................................ ................................ ................ 16 Data Collection ................................ ................................ ................................ .................. 17 Anuran sampling ................................ ................................ ................................ ......... 17 Env ironmental covariates ................................ ................................ ............................ 18 Hydroperiod ................................ ................................ ................................ ................ 19 Treatment history ................................ ................................ ................................ ........ 19 Dat a Analysis ................................ ................................ ................................ ..................... 20 3 RESULTS ................................ ................................ ................................ .......................... 23 Effect of Survey Method ................................ ................................ ................................ .... 23 Detection Probability ................................ ................................ ................................ .......... 24 Occupancy ................................ ................................ ................................ ......................... 24 4 DISCUSSION ................................ ................................ ................................ .................... 27 Conclusions and Futu re Directions ................................ ................................ ..................... 32 APPENDIX A DETECTION MODELS ................................ ................................ ................................ .... 45 B DETECTION MODEL RESULTS ................................ ................................ ..................... 46 LIST OF REFERENCES ................................ ................................ ................................ .......... 47 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ..... 53
6 LIST OF TABLES Table page 3 1 Anurans observed using call and visual encounter surveys in Arthur R. Marshall Loxahatchee National Wildlife Refuge. ................................ ................................ ......... 35 3 2 Detection of each species by visual encounter surveys and call s urveys in Arthur R. Marshall Loxahatchee National Wildlife Refuge. ................................ ........................... 36 3 3 Variation in covariate data used in modeling anuran species occupancy and detection probability in Arthur R. Marshall Loxahatchee National Wildlife Refuge. ..................... 37 3 4 Comparison of differ ent sampling methods using parameter estimates from constant p (.). ................................ ................................ ................................ ............. 38 3 5 Summary of AIC model selection for top models for Hyla cinerea Gastrophryne carolinensis and Lithobates sphenocephalus ................................ ................................ 39 A 1 Detection models run for each anuran species in the first step of the modeling process. ................................ ................................ ................................ ......................... 45 B 1 Summary of AIC model selection for all detection models run in the fir st step of the modeling process ................................ ................................ ................................ ........... 46
7 LIST OF FIGURES Figure page 2 1 Location of Arthur R. Marshall Loxahatchee National Wildlife Refuge in Florida and location of study sites in the interior of the Refuge.. ................................ ....................... 40 3 2 Relationship between detection probability and day of year for Gast rophryne carolinensis in Arthur R. Marshall Loxahatchee National Wildlife Refuge. ................... 41 3 3 Relationship between detection probability and rain*day for Hyla cinerea in Arthur R. Marshall Loxahatc hee National Wildlife Refuge. ................................ ...................... 42 3 4 Relationship between detection probability and sampling period for Lithobates sphenocephalus in Arthur R. Marshall Loxahatchee Nation al Wildlife Refuge ............. 43 3 5 Average detection probability by sampling period for Hyla cinerea Gastrophryne carolinensis, and Lithobates sphenocephalus in Arthur R. Marshall Loxahatchee Natio nal Wildlife Refuge ................................ ................................ .............................. 44
8 LIST OF ABBREVIATION S AIC Akaike Information Criterion ARMI Amphibian Research and Monitoring Initiative BICY Big Cypress National Preserve CERP Comprehensive Everglades Restoration Plan EDEN Everglades Depth Estimation N etwork ENP Everglades National Park NAAMP North American Amphibian Monitoring Protocol PAO Proportion of Area Occupied Refuge Arthur R. Marshall Loxahatchee National Wildlife Refuge SFWMD South Florida Water Management District VES Visual Encounter Survey
9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science SITE OCCUPANCY OF ANURANS IN THE ARTHUR R. MARSHALL LOXAHA TCHEE NATIONAL WILDLIFE REFUGE By Ryan Loring Lynch May 20 12 Chair: Frank J. Mazzotti Major: Interdisciplinary Ecology Global declines in amphibians over the past few decades have prompted interest in the distribution and abundance of amphibian populati ons. Reported c ause s for declines are numerous, but factors related to anthropogenic disturbance top the list. In response to widespread declines the U.S. National Parks Service initiated a national monitoring program to document distribution and status of amphibians in National Parks Initial inventories were conducted in Everglades National Park, Biscayne National Park, and Big Cypress National Preserve between 2000 and 2003. These surveys provided baseline data on distribution and status of amphibian pop ulation s in National Park lands that can be used as an indicator of the ecological response to large scale restoration activities associated with the Comprehensive Everglades Restoration Plan. In the present study, utilizing site occupancy models I expan d the scope of Evergl ades amphibian monitoring to Arthur R. Marshall Loxahatchee National Wildlife Refuge (Refuge) and explore potential factors affecting site occupancy (hydroperiod and exotic treatment history) and detection probabilities. Eight anuran species were documented in the interior of the Refuge, and occupancy of modeled species was high. Low variation in covariate data made it difficult to draw s trong conclusions regarding affects of hydroperiod and treatment history Detection probabilities f or two of the species ( Gastrophryne carolinensis and Lithobates sphenocephalus )
10 modeled were relatively low, emphasizing importance of incorporating detection probabilities into study design. A comprehensive understanding of current distribution and abunda nce of anuran populations in the Refuge provides foundation for development of a long term monitoring program that can be used to assess ecological response to future management actions and oth er potential stressors
11 CH APTER 1 INTRODUCTION Reports of widespread and rapid declines in amphibian populations have triggered a dramatic increase in amphibian base d research in recent years ( Cushman 2006 and references therein). Nearly one third of all known amphibian species in the world are now threatened with extinction (Baillie et al. 2004). In North America, amphibians are considered to be one of the most threatened taxa (Gibbons et al. 2000, Stuart et al. 2004), with many studies pointing to anthropogenic disturbance as the primary cause for declines (Blaustein and Wake 1995, Noss et al. 1997). The effects of habitat loss, alteration, and fragmentation on amphibian populations are of particular concern (Cushman 2006, Stuart et al. 2004), especially in areas with abundant wetl and habitat, a warm and humid tropical or sub tropical climate, and high human populations (Gallant et al. 2007); conditions common to south Florida In response to global population declines, studies have suggested that distribution and abundance of amphi bian populations be used as an indicator of effects of anthropogenic disturbance on wildlife and ecosystems as a whole (Sheridan and Olson 2003, Welsh and Ollivier 1998). U se of amphibians as indicators of ecosystem health is supported by the fact that amp hibians are sensitive to changes in the environment due to unique life history and physiological characteristics (i.e. small body size, reliance on moisture to maintain gas exchange, and use of ephemeral breeding habitats) (Blaustein et al. 1994, Blaustein and Wake 1990). The biphasic lifecycle of many amphibians, requiring both aquatic and terrestrial habitats at different life stages, (Blaustein et al. 1994, Vitt et al. 1990), makes amphibians particularly susceptible to alterations to the landscape (Alfo rd and Richards 1999, Gallant et al. 2007). Amphibians also occupy a mid level trophic position, acting as both important predators of small invertebrates and as prey for larger species (Whiles et al. 2006). Many amphibians, particularly anurans, can also
12 be easily detected and identified using standard cost effective sampling protocols, making them ideal candidates for use as indicators of ecosystem health and ecological response to disturbance in monitoring programs (Waddle 2006, Weir and Mossman 2005). I n the United States, efforts to incorporate amphibians into large scale monitoring programs greatly expanded with the creation of the Amphibian Research and Monitoring Initiative (ARMI). Established by the U.S. Geological Survey (USGS), ARMI was designed t o document and monitor distribution and abundance of amphibians on Department of Interior (DOI) lands to determine population status and trends, and to identify possible causes of amphibian declines (ARMI, www.armi.usgs.gov/). Since the conception of ARMI in 2000, large scale amphibian monitoring programs have been es tablished across the country ( e. g. Bartelt et al. 2011, Mattfeldt et al. 2009, Smith et al. 2006). These efforts have greatly enhanced the rigor of large scale amphibian monitoring programs thr ough development of sampling designs and adoption of new statistical techniques. Specifically, new classes of occupancy models were adopted that address important sources of error brought about by imperfect detection (MacKenzie et al. 2002, 2005), a subjec t frequently overlooked in past studies ( Yoccoz et al. 2001). O ccupancy models developed by MacKenzie et al. (2002) build on closed population capture recapture methods and explicitly incorporate imperfect detection into the modeling process by assuming a priori that detection is less than 1. This method, which estimates proportion of sites occupied (PAO) as a state variable, requires that each site is revisited on numerous occasions within a season. On each visit to a site, the species of interest is eith er detected, with probability p or not detected. Using collective detection histories from all sites,
13 sampling occasion detection probabilities are estimated as a function of covariates for each species using a maximum likelihood approach. Beginning in 20 00, the methodology developed by ARMI was adopted for use in large scale herpetofauna inventories on National Park Service lands in south Florida (e.g. Everglades National Park Rice et al. 2004, Big Cypress National Preserve Rice et al. 2005). Previous efforts to document amphibian and reptile populations in south Florida were either outdated (Duellman and Schwartz 1958), primarily descriptive (Meshaka et al. 2000), or overlooked the importance of imperfect detection (Casler 2005, Dixon 2011). The goals of the Rice et al. (2004, 2005) inventory studies were to generate up to date species lists, estimate occupancy and detection probabilities for each species across the landscape, and develop a protocol that could be used in future monitoring efforts (Rice et al. 2004, 2005). Establishment of a large scale amphibian monitoring protocol in south Florida was particularly imp ortant approval of the Water Resources Development Act of 2000, which authorized the Comprehensive Everglades Restoration ecosystem restoration effort, the overarching goal of CERP is to restore flow of water through the Everglades by capturing fresh water that currently flows unused to the Atlantic Ocean and Gulf of Mexico thr ough a series of canals and water control structures constructed over the past century (CERP; www.everglades plan.org). The biological consequences of years of altered hydrology in the Everglades have been dramatic ( Davis and Ogden 1994 and Sklar et al. 19 98), highlighting the need for both a comprehensive understanding of the current distribution and abundance of amphibian species and an objective method to document ecological response to anthropogenic disturbance and ecosystem restoration.
14 In the present study, utilizing field sampling methods and statistical techniques common to regional (Rice et al. 2004) and national (ARMI) monitoring protocols, I expand the scope of Everglades amphibian monitoring to the northern extent of the protected Everglades; on e of the remaining areas in the Everglades lacking recent baseline amphibian population data. The specific objectives of the study were fivefold: (1) Document anuran species occupying the interior of the Arthur R. Marshall Loxahatchee National Wildlife Ref uge, (2) Determine if different sampling methods vary in their ability to detect anurans in the Refuge (3) Determine if species specific detection probabilities vary in space and time and as a function of sampling occasion environmental covariates, (4) Ev aluate the relationship between anuran species occupancy and hydroperiod, and (5) Evaluate the relationship between anuran species occupancy and prior management activity, primarily in the form of exotic species treatment history.
15 C HAPTER 2 METHODS Study Area F ield work for this study was conducted in Arthur R. Marshall Loxahatchee National Wildlife Refuge (hereafter Refuge), in Boynton Beach, Florida ( Figure 2 1). The Refuge is the northern most remnant of the protected Everglades w etland ecosystem, and has been managed by the U.S. Fish and Wil dlife Service as a unit of the National Wildlife R efuge system since 1951. As with other parts of the Everglades, the Refuge has been impounded and cutoff from the rest of the wetland ecosystem through the construction of canals, levees, and pump systems. A result of this impoundment is an ecosystem with altered hydrology (Richardson et al. 1990). Today the Refuge experiences an exten ded hydroperiod in some regions due to a ponding effect and a decreased hydroperiod in others ( Brandt et al. 2000, Leach et al. 1971 Sklar et al. 1998 ). The alteration of hydrological parameters across the Refuge has resulted in changes to natural communities over the past century (Brandt et al. 2000, Richardson et al. 1990). Today the Refuge encompasses 59,646 ha of Everglades ridge and slough habitat. The landscape is a wetland mosaic of sloughs, wet prairies, sawgrass marshes, and tree islands (Loveless 1959, Brandt et al 2002). Tree islands, and in particular pop up tree islands, are a unique feature of the Refuge landscape that distinguish es it from other parts of the greater Everglades Ecosystem. Scattered across the entire central portion of the Refuge, pop up tree islands are small, generally circular tree islands that originated from floating mats of peat. The structure and upland habitat that tree islands provide increases habitat heterogeneity of the Refuge landscape and as a result are important sources of biodiversity (Sklar and Van de Valk 2003) The v ariety of habitat types across the Refuge landscape v ary in their overall structure and species composition, and are directly tied to variations in hydroperiod (Jordan et al. 1996
16 Richardson 1990 ). Sloughs are deep water habitats that are commonly differe ntiated by the presence of truly aquatic species ( e.g. Nymphaea sp., Utricularia sp.), wet prairies are shallow water habitats distinguished primarily by graminoids ( e.g. Eleocharis sp., Rhynchospora sp. ), and tree islands are upland habitats identified by the presence of shrubs and woody vegetation ( e.g. Persea sp., Ilex sp., Myrica sp.) (Loveless 1959). Sampling Design This study was conducted in the interior of the Refuge that is characterized by tree islands and ridge and slough habitats. Thi s area also corresponds to the Interior Z one used in ongoing water quality sampling at the Refuge ; an area that is less impacted by canal water intrusi on and other human impacts ( Harwell et al. 2008 ) Study sites were selected across the Refuge landscape using a stra tified random design Two distinct hydroperiod strata (longer and shorter) were chosen based on historic hydroperiod during drought years (2001 and 2007) to ensure that sampling encompassed both ends of the hydroperiod gradient of the Refuge interior. Hydr operiod was determined using the Everglades Depth Estimation Network (EDEN) (EDEN; http://sofia.usgs.gov/eden/ ). Within each strata thirty points (buffered at 1000m to meet model closure assumption n ArcGIS In addition, 15 spare points were created to be used as back up sites in the case that a particular point was unusable. In s pring 2010 each of the 60 random points was visited during the day to establish study sites. At each of the random poin ts the nearest pop up tree island was selected to be used as a study site. Only islands within 200m of the random point were used. If no pop up tree islands could be found within 200m of the random point then one of the spare random points was used. Once a tree island was selected, a 4m x 40m transect was set up to run from the interior of the tree island into the marsh. Sampling the ecotone that runs from tree islands into the marsh
17 ensured that sampling captured much of the habitat diversity found in the interior of the Refuge, and also ensured that both highly aquatic and terrestrial species would be included in surveys. D irection of each transect was haphazardly chosen, but was limited to directions that could accommodate a 40m length transect. Each tran sect was positioned so that 20m were on the tree island and 20m were in the marsh. The mid point of each transect was positioned at the water line in May 2010 ( mean stage from 1 7 station = 16.54 ft ) and corresponded to the transitional zone in vegetation from the aquatic more herbaceous species to the upland woody community. Transects were marked using PVC pipes and reflective flagging tape every 5 meters prior to sampling. Data C ollection Anuran sampling Anuran sampling was initiated 09 June 2010 and r an until 09 September 2010. Sampling during this period corresponds to the rainy season in south Florida, and coincides with the time of year when anuran detection probabilities are highest in south Florida (Rice et al. 2004, 2005). Each site was sampled t hree times through out the rainy season for adult anurans T o avo id breaking assumptions of occupancy model ing or introducing heterogeneity in detection probabilities, sites were resurveyed on different nights and at different times of night than the origin al survey The order in which sites were surveyed was chosen at random using clustered groups of neighboring islands, with revisits to each group following the same order as the initial visits. The primary sampling technique used was standard visual encou nter surveys (VES) (Heyer et al. 1994) where the species of all animals observed was recorded Following the North American Amphibian Monitoring Protocol (NAAMP; http://www.pwrc.usgs.gov/naamp/), surveys began 30 minutes after sunset and finished prior to 1 AM. This period coincides with the
18 most active time period for most amphibians. Each survey was conducted by two experienced field observers (myself and one rotating assistant). Search effort at each transect was area constrained rather than time constr ained, however an attempt was made to spend roughly equal time searching each habitat (tree island & marsh) and each site to avoid spending drastically unequal time at different sites. The goal of each survey was to document every species of anuran present at a site. Prior to each VES and following a 5 minute silent period an anuran call survey was also conducted at each site During the first 5 minutes at a site all species of anurans heard vocalizing were recorded. All animals heard vocalizing during t he 5 minute period were noted, even if it was possible that the animal was located off of the transect A pproximate location (on transect, adjacent to transect (within 50m), distant call) of each animal heard vocalizing was also recorded Only individuals recorded on the transect or heard adjacent to the transect were considered present at the site for the occupancy analysis. Environmental c ovariates During the time of sampling, environmental data were recorded at each site to be used as covariates of dete ction. Air temperature, water temperature, wind speed, relative humidity, and barometric pressure were recorded during every sampling occasion. Presence of rain within 12 hour period of each survey and d ay of ye ar were later added to this list Vegetation characteristics along each transect were collected for each site following completion of anuran sampling. An interrupted belt transect was used to document general vegetation density at each site. The percent cover of vegetation (using Daubenmire cover cla sses ) was recorded in 1m 2 plots laid down every five meters along a transect to account for ground cover and general vegetation density At each point along a transect canopy cover (using a spherical densiometer) was also recorded to account for overhead c over. In the case that two environmental variables
19 were correlated (p<0.05), only one was chosen for use in analysis. All data were recorded using a PalmOne handheld unit in the field and was hot linked to a Microsoft Access database following each nights fieldwork. Hydroperiod Hydroperiod data for each site were retrieved from EDEN at the conclusion of the field season. The South Florida Water Management District (SFWMD) water year that runs from May 01 April 30 was used to calculate this value. O n the gr ound water depth s were recorded at all sites along each transect as a way to ground truth EDEN values Every five meters along the transect (marked with pvc pipes or flagging tape) water depth was recorded to the nearest 1cm using a meter stick. Depth data was collected before, after, and between each sampling period (n=5). Correlations between t h ese data and water depth data collected from the nearest water gage as well as EDEN depth output were later compared as a way to corroborate the EDEN output. Becau se significant differences between EDEN reported depth and on the ground recorded depth were found for 12 of 60 sites, the ground surface data used in EDEN had to be corrected for by using an average of all on the ground depth data recorded for each site p rior to the final hydroperiod ca lculation Treatment history The Refuge has some of the densest infestations of exotic vegetation i n the Everglades ecosystem ( Woodmansee et al. 2005). To combat invasion and spread of exotic species the Refuge has enacted a n aggressive year round treatment program, utilizing both manual removal and herbicide application to control dense populations of exotics species such as Old World climbing fern ( Lygodium microphyllum ) and Melaleuca ( Melaleuca quinquenervia ). While the ne gative effects of herbicides on amphibians have been documented in other studies ( Cauble and Wagner 2005, Mann and Bidwell 1999 ), testing the direct effect of chemicals was not feasible in
20 this study. Rather, the focus of this research was the timing of ma nagement activity (i.e. exotic species contractor activity on and around study sites) and its affect on anuran occupancy. In other words, are sites that were treated or in the vicinity of recently treated areas more likely to lack anuran species than areas not treated in the past three years T reatment history (date, location) is recorded by Refuge staff on a 1km grid scale. This information was used to determine the most recent treatment date for each grid cell corresponding to a study site from this study Data Analysis Following other anuran surveys in the region (i.e. Rice et al. 2005), the modeling approach adopted for this study was based on the occupancy models dev eloped by MacKenzie et al. (2002, 2005 ). The software program PRESENCE (Hines 2006) was used to calculate detection and occupancy estimates using the single species single season model. Throughout this paper the p will be used to represent species occupancy and detection probabilities, respectively. The first set of analyse s conducted evaluate effects of the two different survey methods by assuming that species occupancy and detection probabilities were constant across both space p (.)]. This model is not the best representation of the data, but by holding both occupancy and detection constant for each species relative impacts of each sampling method can be determined without having to worry about unknown biases introduced by models with different covariates (Bailey et al. 2004, Boulinier et al 1998). Esti mates produced from the constant model, along with the proportion of sites where a species was observed [i were compared across three datasets: (1) call survey data only, (2) VES data only, and (3) the combined dataset. A two step process was employed to estimate site occupancy and detection probabilities as a function of covariates for anurans in the Refuge using the combined dataset. In the first step
21 covariates thought to influence detection probabilities were modeled, including: air temperature, humidity, percent vegetation cover, day of year, and presence of rain withi n 12 hour period of eac h survey. From this list an a priori model set was created based on biological knowledge of the species and results from previous research conducted in the region (A ppendix A). Additive, interactive, and higher order terms (e.g. quadratic) were considered in this step. For example, to allow for a peak in detection probability, both a linear (Day) and quadratic (Day+Day 2 ) day of season effect were included. In this first step, occupancy was held constant while modeling which factors best explained detection probabilities [i p (covariate)]. In the second step of the modeling process, the best detection model for each species was paired with occupancy variables that represented biological hypot heses (Burnham and Anderson 1998 ). Two specific occupancy variables were cons idered: hydroperi od and treatment history Hydroperiod represents the number of days with standing water at each site during the 2010 SFWMD water year. Treatment history was used as a bianary variable; that is, each site was coded as either being treated f or exotics in last three years (1) or not treated in the last three years (0). Unfortunately, more detailed treatment history was not available at the time of analysis. The final model set was kept small by employing the two step approach, with two occupa ncy variables and a single best detection model per species. Prior to analysis all variables were standardized to avoid problems with parameter estimation in program PRESENCE and correlated variables (p< 0. 0 5) were not included in the same model in order t o avoid multicollinearity issues. Models for each species were fit and maximum likelihood estimates The goodness of fit of at least one model in each model set was ass essed following the methods of MacKenzie and Bailey (2004). All models included factors thought to affect parameter
22 p ) and as a result do not represent all possible environmental or biological processes that may influence occupancy or detecti on probabilities.
23 CHAPTER 3 RESULTS Eight anuran species, totaling 812 individuals, were observed at the 60 study sites between 09 June to 09 September, 2010 ( Table 3 1). Six species were observed using both VES and call surveys, while Anaxyrus terrestris was only observed during VES and Hyla femoralis was only recorded during call surveys ( Table 3 2). Five of the species observed were either observed too frequently ( Acris gryllus (nave occupancy = 100%), Lithobates grylio (nave occupancy = 100 %)) or not frequently enough ( Pseudacris ocularis (nave occupancy = 1%), Anaxyrus terrestris (nave occupancy = 0.6%), Hyla femoralis (nave occupancy = 2% ) to model. Acris gryllus and Lithobates grylio were found at every site on nearly every occasion and thus occ upy 100% of sites with near perfect detection. Pseudacris ocularis, Anaxyrus terrestris, and Hyla femoralis were only observed on a few nights, and as a result reliable occupancy and detection estimates could not be obtained. The three species that are use d in the analysis include: Gastrophryne carolinensis, Hyla cinerea and Lithobates sphenocephalus Hydroperiod for the sampling frame in 2010 ranged from 336 to 365 days (mean SD; 349 7.74) and 32 of 6 0 cells (53%) including a study site were treated fo r exotics in the past three years ( Table 3 3) Effect of Survey Method All three datasets (call, VES, combined) produced occupancy estim (Table 3 4) Differences between predicted occupancy equally or better than the call a nd VES datasets for all three species. Occupancy estimates produced by the call survey and combined datasets were equal for H. cinerea using the call survey data set for both G. carolinensis and L. sphenocephalus
24 Estimates of d etection probability varied drastically by species and sampling method ( Table 3 4 ) Detection of H. cinerea across all sampling methods, was significantly greater than the other two specie s. For H. cinerea call surveys were significantly better at detecting the species than VES. On the other hand, detection probabilities were higher using VES than call surveys for both G. carolinensis and L. sphenocephalus although both methods had moderat ely to extremely low detection probabilities. For all species, combining the two sampling methods resulted in higher detection probabilities supporting the use of multiple m ethod sampling Detection Probability The best detection model for G. carolinensis with 0.24 AIC weight, was day specific detection p (day) (A ppendix B). Detection probabilities ranged from 0.04 to 0.49, increasing later in the season ( Figure 3 2). Survey specific detection was the second best model, with 0.13 AIC weight, but represent s a similar idea as p (day), so only p (day) was used to examine occupancy of G. carolinensis The clear best detection model for H.cinerea with 0.82 AIC weight, was an interaction between rainfall and day, p (rain*day) (A ppendix B). Detection of H. cinerea increased on rainy days as the season progressed, and decreased on non rainy days as the season progressed ( Figure 3 3). There was no clear best model for detection probability for L. sphenocephalus (A ppendix B). The top detection model, with 0.11 AIC weig ht, was p (survey) ( Figure 3 4). Although day was not the top model for all three species, a conspicuous pattern was observed where the detection probability increased for all species later in the season ( Figure 3 5). Occupancy The nave site occupancy for G. carolinensis was 0.50, or 30 of 60 sites. Individuals were detected at 5 (8%), 6 (10%), and 22 (37%) sites on the first, second, and third surveys respectively. The top ranked model for G. carolinensis carrying 0.28 AIC weight, had constant occupancy w ith detection varying by day ( Table 3 5 ). Site occupancy was estimated to be 1
25 across the entire sampling frame (but see below). Models that included hydroperiod or treatment history appeared to be close competitors to the best model, with similar AIC valu es and within 2 AIC from the to p model. However, these results are spurious, as numerical convergence was never reached and the 2log likelihoods (hereafter 2LL) of these models are all very similar to the top model (Anderson and Burnham 1998 ). This suggests that adding these variables did not substantially improve the modeling of the data. Even the best ranked model for G. carolinensis had confidence intervals that broadly overlapped zero. Moot occupancy estimates are likely a result of extremely low detection probabili ty, so there was a low chance of observing the species even when it was present at a site. Future surveys that better account for detection should be conducted to confirm occupancy status of G. carolinensis in the Refuge, and improve precision of parameter estimates. Models that did not incorporate a detection variable did substantially worse than models that accounted for detection. The nave site occupancy for H. cinerea was 1.0, or 60 of 60 sites. Individuals were detected at 45 (75%), 50 (83%), and 53 ( 88%) sites on the first, second, and third surveys respectively. The clear best model for H. cinerea had constant occupancy with an interaction between rain and day influencing detection ( Table 3 5 ). This model carried 0.53 AIC weight and had 2.7 times mor e support than the second best model Occupancy equaled 1 across the entire sampling frame. Models that incorporated hydroperiod or treatment history had extremely high standard errors and confidence intervals that broadly overlapped zero, a result of cons tant occupancy across the entire range of both variables. Models that incorporated a detection variable did slightly better than models that assumed constant detection. The nave site occupancy for L. sphenocephalus was 0.77, or 46 of 60 sites. Individuals were detected at 24 (40%), 20 (33%), and 31 (52%) sites on the first, second, and third surveys
26 respectively. The top ranked model, carrying 0.33 AIC weight, had constant occupancy with survey specific detection ( Table 3 5 ) It is estimated that L. spheno cephalus occupies 90% (S.E. = 0.08) of sites within the sampling frame The second best model also included constant occupancy and produced a similar occupancy estimate (0.910.08) supporting the fact that the probability of site occupancy is equal at all sites. Treatment history a nd hydroperiod were included in the third and fourth top models, with AIC weights of 0.19 and 0.14, respectively, however the slopes for these terms were relatively small with large standard errors (estimate S E = 1.65 3.0 f or treatment history; 0.400.7 for hydroperiod) With such large standard error s no statistical difference was found between sites with different hydroperiod values or between treated and untreated sites. Models that included the best detection variable d id not rank substantially better than models that assumed constant detection probabilities.
27 CHAPTER 4 DISCUSSION Prior to the development of a monitoring program, comprehensive inventories of species are necessary to establish basel ine information on presence and distribution of populations (Dodd 2009). This study conducted the first thorough inventory of adult anurans in the interior of the Refuge and estimated site occupanc y and detection probabilities for three species using the o ccupancy modeling framework of MacKenzie et al. (2002). Eight anuran species were documented using standard sampling techniques and site occupancy estimates ranged from 90% to 100% for three modeled species. As has been shown in other studies (Bailey et a l. 2004), u se of two sampling methods resulted in higher detection probabilities than a single method. The top model in all three mod el sets had constant probability of occupancy paired with a single covariate for detection Occupancy estimates produced by the top model were significantly higher than nave estimates for two species ( G. carolinensis and L. sphenocephalus ). This discrepancy is a direct result of detection rates that ranged from extremely low (0.06 0.34) to moderate (0.37 0.57), highlighting t he importance of incorporating detection probabilities into study design rather than using unadjusted count data. This issue of imperfect detection has long been recognized as a major source of error in fiel d studies (MacKenzie et al. 2002 Thompson 1992) and wildlife monitoring programs (Yoccoz et al. 2001). Results from this study indicate that detection probabilities for most anurans in the Refuge are less than one, even for apparently abundant and common species such as H. cinerea and agree with other studies that found that nave occupancy values underestimate true occupancy status of wildlife populations (Bailey et al. 2004, MacKenzie and Nichols 2004). In the case of secretive/ cryptic or rare species, such as G. carolinensis in this study, nave occ upancy estimates may be as much as 50% lower than their respective occupancy estimates.
28 Inferences about changes in occupancy that are based on unadjusted data (e.g. nave occupancy), will be negatively biased (MacKenzie et al. 2002) and have the potential to lead to unfounded management actions. In such situations, perceived changes in population status may simply be a change Variation in anuran detection p robabilities has been attributed to many factors, including: (1) environmental conditions (Rice et al. 2005, Weir et al. 2005), (2) temporal variation in behavior (Bridges and Dorcas 2000, Cook et al. 2011), (3) observer bias (Waddle 2006, Weir et al. 2005 ), (4) sampling effort and/ or methodology (Shirose et al 1997, Smith et al. 2006, de Solla et al. 2005), (5) location (Sewell et al. 2010), and (6) population size (Tanadini and Schmidt 2011). In the present study, temporal and observer based variation i n detection was controlled for in the study design and a subset of covariates was selected based on results of similar studies D etection probabilities varied by sampling method and time of year for all three species modeled in this study These results su pport findings of similar studies that have found that use of multiple detection methods improve s parameter estimates (Bailey et al. 2004) and that detection probabilities increase in summer months (Rice et al. 2004, 2005). The absence of environmental cov ariates in the top models is a result of the single season sa mpling design of this study. R estricting sampling to summer months this study, in effect, controlled for environmental conditions (i.e. warmer and wetter) that were found to be important factors improving detection probabilities in related studies (Rice et al. 2004, 2005). T hat detection probabilities progressively increased over the summer sampling period, however, suggests that extending sampling into the fall may improve detection probabilities further and thus result in more precise parameter estimates for the three modeled species.
29 Dramatic differences in detection probability were observed between species; average detection probability of H. cinerea was nearly twice that of L. sphenocephalus and 4.5 times greater than G. carolinensis The high detection probability for H.cinerea was expected considering the generalist nature of the species (Meshaka et al. 2000), and the frequency and volume of their call. Rice et al. (2005) also detected H. ci nerea in large numbers using both VES and call surveys across all habitats in the Everglades. The moderately low detection probabilities for L. sphenocephalus on the other hand, were somewhat surprising. The species is commonly considered to be the most a bundant anuran in Florida (Bartlett and Bartlett 1999), and large numbers were detected by Rice et al. (2004, 2005) in both ENP and BICY. The moderately low detection probabilities for this species are likely a result of the fact that sampling did not incl ude much of the species fall breeding season (Ryan and Winne 2001). Detection of L. sphenocephalus was almost entirely by VES, with call survey detections on only a few occasions late in the season. Extending sampling into the fall breeding season would li kely improve the likelihood of detecting the species using both VES and call surveys. Detection probabilities for G. carolinensis were t he lowest throughout the study. Other studies in the southeast have reported similar results (Rice et al. 2004, Rice et al. 2005, Walls et al. 2011) with detection frequently being less than 0.15 Walls et al. (2011) tentatively suggest ed that variation in parameter estimates for G. carolinensis in Louisiana may be a result of heavy rainfall. The opportunistic and explosiv e breeding behavior of this species during heavy rainfall (Mitchell and Lannoo 2005) lends some support to this hypothesis however, no significant effect of rainfall was found in this study Although only speculative, the low detection probability for G. carolinensis in this study may be a result of s ampling methodolog y Ground cover was infrequently flipped in pursuit of animals during this study to avoid disturbing or damaging
30 Refuge habitats. Life history characteristics, such as a cryptic and secretive nature, and the fact that individuals were occasionally encountered under cover on tree islands suggests that detection probability for G. carolinensis would increase with more thorough sampling. Addressing the issue of imperfect detection is a critical step in development of a sampling design for use in long term monitoring programs. Adequately accounting for variation in detection ensures accuracy of occupancy estimates, and hence, usefulness of inferences to managers A primary objective of this study was to determine the relationship between anuran occupancy and hydroperiod in the Refuge. I t was predicted that shorter hydroperiod sites would have lower occupancy than intermediate or longer hydroperiod sites. H owever, l ow variation in hydroperiod during the sampling period made it impossible to adequately address this question. Hydroperiod only varied by 29 days within the sampling frame during the 2010 SFWMD water year and as a result occupancy of all three species was estimated to be between 90% and 1 00%, with the probability of site occupancy being equal for all sites With such a low range in hydroperiod, sampling during this study sampled sites with similar h ydroperiod. Other studies that have attributed difference s in anuran distribution ( Babbitt et al. 2003 ), abundance ( Van Buskirk 2005 ), or richness ( Snodgrass et al. 2000 ) to hydroperiod have frequently conducted their studies in depressional or ephemeral wetlands This distinction between temporary and permanent wetlands is critical when trying to understand relationship s between biotic and abiotic factors and amphibian distribution and community structure ( Wellborn et al. 1996 ). For example, length and timing of drying events in temporary wetlands may limit distribution of anurans with longer la rval periods, whereas, a larger predator base found in longer more permanent wetlands may limit distribution and abundance of anurans more susceptible to predation. Such differences may explain why anuran assemblages in the
31 Everglades tend to be more depau perate than temporary wetlands in the region (Babb itt and Tanner 2000). E arlier work on the distribution and trophic interactions of larval anuran s in the Refuge suggested that distribution and density of some species is nonrandom, varying predictably by h abitat and in relation to predator density (Baber et al. 2005). The f itness tradeoffs (i.e. body size, developmental rate, activity ) made by each species in combination with physical factors such as hydrology, play important role s in determining the distr ibution and abundance of species across environmental gradients ( Wellborn et al. 1996 ). Determining specific biotic and abiotic mechanisms responsible for shaping distributional patterns of the anuran assemblage in the Everglades and their interaction wit h one another, should be an important objective for future research. Another objective of this study was to determine the relationship between exotic species treatment history and anuran occupancy. Unfortunately, although species occupancy was reported as high across both treated and untreated sites my ability to draw strong conclusions on the direct effects of treatment history of anuran occupancy was hindered by the ambiguity of the data. I was only able to use treatment history as a binary variable due to the general nature of Refuge databases and detection probabilities for two of the three observed species were extremely low As mentioned above, when species are rarely observed it results in large confidence intervals indicating less precise parameter estimates. Furthermore, it was not known with 100% certainty whether one of the study sites used in this study was directly impacted by ex otic species management or not. With ambiguous treatment history data and low probability of detection for G.caroline nsis and L. sphenocephalus it would be unwise to conclude that exotic species treatment history is having no effect on anurans in the Refuge T o determine the true relationship between exotic species man a gement and anuran occupancy in the Refuge, future
32 st udies that directly address potential impacts are needed. For example, a Before After Control Impact (BACI) design would be much better suited t o address this question than a large scale occupancy study. These findings agree with other studies (i.e. Bailey et al. 2004) that emphasize the importance of having a good sampling design that insures that research and management questions are adequately addressed, and that chosen sampling sites are representative of the entire area from which inferences are to be drawn. Conclusions and Future Directions S ite occupancy has increasingly become a popular method for monitoring status of wildlife populations (Bailey et al. 2004, Brander et al. 2007, Kroll et al. 2008) Similar to others, I found the use of occupancy as a state variable to be a useful and efficient means to document the distribution and relative abundance of species across the landscape ( Bailey et al. 2004 ) Using both VES and call surveys eight of the twelve native anuran species known from south Florida and no exotic anurans, were detected in the Refuge interior Occupancy of H. cinerea G. carolinensis and L. sphenocephalus was estimated to be extremely high across the center of the Refuge, and the inclusion of detection probabilities in the modeling framework greatly improved estimates. However, as with any field study, a few caveats were revealed during this study that I would like to discuss in closing in hopes of encouraging and improving future sampling efforts First, single species occupancy est imates are sensitive to abundant and rare species. The perform well when data is near the boundaries of 1 or 0. MacKenzie et al. (2002) discuss this issue in their simulat ion study, stating that occupancy estimates should be interpreted with caution in cases where occupancy is estimated at 1 and detection is below 0.15. These conditions are similar to the results for G. carolinensis in the present study. In such cases it is difficult for the model to distinguish between a true absence at a site and a site where the species simply
33 As stated previously, f uture survey efforts in the Refuge would benefit by better accounting for variation in detection probabilit y for rarely observed species. Second the single species single season models used in this study do not allow for unexplained heterogeneity in detection probabilities between sites only variation explained by included covariates. If unexplained heterogen eity in detection probabilities exists then occupancy estimates have the potential to be biased (Royle and Nichols 2003). Recent empirical work with amphibian populations suggests that variation in population size or abundance is common and has the potenti al to strongly affect detection probabilities (Tanadini and Schmidt 2011). This being the case, it is easy to perceive a situation where less abundant species may be overlooked etection resulting from population size. A number of studies have developed techniques to address this potential source of bias in monitoring studies (Royle and Nichols 2003, Tanadini and Schmidt 2011). Tanadini and Schmidt (2011) suggest using past popula tion index in the form of count data obtained from surveys in previous years as a covariate for detection in subsequent surveys. Because c ounts of individuals at each site were recorded during this survey, future surveys in the Refuge can test the methods of Tanadini and Schmidt (2011) by using this data as a covariate for detection. Lastly, the present study was conducted as a first effort to document the anuran community of the Refuge and determine whether factors such as hydroperiod and treatment histor y were influencing occupancy. This effort was associated with larger efforts to inventory anuran populations in the Everglades in the face of large scale restoration (Rice et al. 2004, 2005). Following the methodology of these efforts, to ensure comparabil ity, this study used the single species single season occupancy models of MacKenzie et al (2002) While t his method was
34 useful in accomplish ing the goals of this initial study, the single species single season model has certain limitations. Single season models only offer a snapshot of the population at a single point in time. Gaining an understanding of long term trends in populations, and factors influencing changes in site occupancy (i.e. extinction and colonization) are of even more use to managers (Ma cKenzie et al. 2003). Recent advances o n the MacKenzie et al. (2002) occupancy model s have resulted in exciting improvements on standard single species based monitoring programs. These advances utilize an approach similar to the MacKenzie et al. (2002) and Royle and Kery (2007) occupancy models, but extend the model s to include multiple species or an assemblage of ecologically similar species ( Dorazio and Royle 2005, Kery and Royle 2008 ). This approach produces estimates of occurrence, colonization, extinct ion, and detection for each species of an ecologically similar assemblage of species by sharing or borrowing ob servational data among species (MacKenzie et al. 2005). In many cases this method produces more precise parameter estimates than single species m odels, particularly for species with low detection probabilities (e.g. G. carolinensis in this study ) (MacKenzie et al. 2005). Th is method utilizes a hierarchical Bayesian approach to estimation using methods of Bayesian computation (Markov chain Monte Car lo) (Walls et al. 2011). By modeling multiple species under a single hierarchical framework managers will be better informed about the status of an entire assemblage of species rather than basing management decisions on individual species or populations (W alls et al. 2011). Current work is being done to apply these methods to a large scale amphibian model in the greater Everglades ( Joint Ecosystem Modeling, www.jem.gov/ ). Once completed, this model will serve as an important tool for managers to evaluate th e status of anurans across the Everglades landscape and their response to ecosystem restoration.
35 Table 3 1. A nurans observed using call and visual encounter surveys in Arthur R. Marshall Loxahatchee National Wildlife Refuge during summer 2010 and number of individuals observed during visual encounter surveys. Common Name Scientific Name Count Cricket Frog Acris gryllus 449 Little Grass Frog Pseudacris ocularis 1 Eastern Narrowmouth Toad Gastrophryne carolinensis 26 Pig Frog Lithobates grylio 164 Southe rn Leopard Frog Lithobates sphenocephalus 80 Green Treefrog Hyla cinerea 91 Pinewoods T reefrog Hyla femoralis 0 Southern Toad Anaxyrus terrestris 1
36 Table 3 2. Detection of each species by visual encounter surveys and call surveys in Arthur R. Marshal l Loxahatchee National Wildlife Refuge, summer 2010. Species VES Call Survey Acris gryllus X X Pseudacris ocularis X X Gastrophryne carolinensis X X Lithobates grylio X X Lithobates sphenocephalus X X Hyla cinerea X X Hyla femoralis X Anaxyrus t errestris X
37 Table 3 3 Variation in covariate data used in modeling anuran species occupancy and detection probability in Arthur R. Marshall Loxahatchee National Wildlife Refuge Variable Mean SD Low High Range Air Temperature (C) 27.9 1.5 23 32 9 Humidity (%) 83.6 5.7 71 99 28 Vegetation Cover (%) 50.9 9.7 28.9 66.8 37.9 Hydroperiod (Days) 348.8 7.7 336 365 29
38 Table 3 4. Comparison of different sampling methods using parameter estimates from constant p (.) : Call survey data only (Call), Visual encounter survey data only (VES), and combined Call survey and VES survey data (Combined). Call Species p (.) [SE] Gastrophryne carolinensis 0.27 1  0.09 [0.02] Hyla cinerea 1 1  0.66 [0 .04] Lithobates sphenocephalus 0.13 1  0.04 [0.02] VES Species p (.) [SE] Gastrophryne carolinensis 0.37 1  0.13 [0.02] Hyla cinerea 0.8 1  0.41 [0.04] Lithobates sphenocephalus 0.77 0.99 [0.1] 0.39 [0.06] Combined Species p (.) [SE] Gastrophryne carolinensis 0.5 1  0.18 [0.03] Hyla cinerea 1 1  0.82 [0.03] Lithobates sphenocephalus 0.77 0.91 [0.08] 0.46 [0.05]
39 Table 3 5. Summary of AIC model selection for top models for Hyla cinerea Gastrophryne carolinensis and Lithobates sphenocephalus ; Arthur R. Marshall Loxaha tchee National Wildlife Refuge t reatment history denotes models that did not converge. Model 2 log likelihood K AIC w Hyla cinerea p (rain*day) 154.93 5 164.93 0 0.53 p (rain*day) 154.93 6 166.93 2 0.19 p (rain*day) 154.93 6 166.93 2 0.19 p (rain*day) 154.93 7 168.93 4 0.07 p (.) 168.48 2 172.48 7.55 0.01 G astrophryne carolinensis p (Day) 152.99 3 158.99 0 0.28 p (Day) 151.06 4 159.06 0.07 0.27 p (Day) 149.21 5 159.21 0.22 0.25 p (Day) 151.64 4 159.64 0.65 0.2 p (.) 171.51 2 175.51 16.52 0 Litho bates sphenocephalus p (Survey) 238.94 4 246.94 0 0.33 p (.) 243.54 2 247.54 0.6 0.24 p (Survey) 238.05 5 248.05 1.11 0.19 p (Survey) 238.6 5 248.6 1.66 0.14 p (Survey) 237.44 6 249.44 2.5 0.09
40 Fig ure 2 1. Location of Arthur R. Marshall Loxaha tchee National Wildlife Refuge i n Florida and location of study sites in the interior of the Refuge. Shaded area of south Florida represents the South Florida Water Management District Boundary. T riangles in th e inset represent study sites and the blue line represents the sampling frame and is thus the area of inference used in this study.
41 Figure 3 2 Relationship between detection probability and day of year for Gastrophryne carolinensis in Arthur R. Marsha ll Loxaha tchee National Wildlife Refuge summer 2010.
42 Figure 3 3. Relationship between detection probability and rain*day for Hyla cinerea in Arthur R. Marshall Loxahatchee National Wildlife Refu ge summer 2010.
43 Figure 3 4 Rel ationship between detection probability and sampling period for Lithobates sphenocephalus in Arthur R. Marshall Loxaha tchee National Wildlife Refuge summer 2010. Sampling period 1 was in June, sampling period 2 in July, and sampling period 3 in August.
44 Figure 3 5 Average detection probability by sampling period for Hyla cinerea Gastrophryne carolinensis, and Lithobates sphenocephalus in Arthur R. Marshall Loxaha tchee National Wildlife Refuge summer 2010.
45 APPENDIX A DETE CTION MODELS Table A 1. D etection models run for each anuran species in the first step of the modeling process. Variables were chosen because of their perceived influence on either anuran activity or observers ability to detect individuals. Occupancy Det ection p (.) p (air temp.) p (humidity) p (veg. density) p (rain) p (day) p (day + day 2 ) p (air temp. + rain) p (air temp. + humidity) p (air temp. + veg) p (air temp. + day) p (veg. density + rai n) p (veg. density + day) p (rain + day) p (air temp. + humidity + rain) p (air temp. + veg. density + rain) p (air temp. + rain + day) p (air temp. + humidity + veg. density) p (veg. density + rain + day) p (air tem p.* veg. density) p (air temp.* day) p (rain day)
46 APPENDIX B DETECTION MODEL RESU LTS Table B 1. Summary of AIC model selection for all detection models run in the first step of the modeling process Top model for each species that was used in the final model set is set in bold and model weights are in parenthese s. AIC values (model weight) Occupancy Detection G. carolinensis H. cinerea L. sphenocephalus p (.) 175.51 (0.00) 172.48 (0.02) 247.54 (0.08) p (day) 158.99 (0.24) 1 73.42 (0.01) 247.19 (0.10) p (humidity) 175.9 (0.00) 173.6 (0.01) 248.88 (0.04) p (air temp.) 176.7 (0.00) 174.41 (0.01) 249.52 (0.03) p (rain) 175.85 0.00) 173.37 (0.01) 249.48 (0.03) p (survey) 160.29 (0.13) 172.78 (0.02) 246.94 (0.11 ) p (veg. density) 176.86 (0.00) 172.62 (0.02) 247.71 (0.08) p (day + day 2 ) 160.77 (0.10) 174.51 (0.01) 247.52 (0.08) p (air temp. + day) 160.98 (0.09) 175.42 (0.00) 248.87 (0.04) p (air temp. + humidity) 176.19 (0.00) 175.59 (0.00) 250 .88 (0.01) p (air temp. + rain) 175.92 (0.00) 174.95 (0.01) 251.43 (0.01) p (air temp. + veg density ) 178.01 (0.00) 174.56 (0.01) 249.65 (0.03) p (rain + day) 160.78 (0.10) 174.68 (0.01) 248.83 (0.04) p (veg. density + day) 160.3 (0.13 ) 173.52 (0.01) 247.34 (0.09) p (veg. density + rain) 177.35 (0.00) 173.27 (0.01) 249.7 (0.03) p (veg. density + rain + day) 162.08 (0.05) 174.53 (0.01) 249.15 (0.03) p (air temp. + humidity + rain) 175.98 (0.00) 175.77 (0.00) 252.7 (0.01) p (air temp. + humidity + veg. density) 177.6 (0.00) 175.75 (0.00) 251 (0.01) p (air temp. + rain + day) 162.68 (0.04) 176.55 (0.00) 250.06 (0.02) p (air temp. + veg. density + rain) 177.42 (0.00) 174.82 (0.01) 251.62 (0.01) p (air temp. day) 162.71 (0.04) 176.55 (0.00) 250.75 (0.02) p (air temp.* veg. density) 177.35 (0.00) 176.45 (0.00) 249.31 (0.03) p (rain day) 160.86 (0.10) 164.93 (0.82) 247.13 (0.10)
47 LIST OF REFERENCES Alford, R. A. and S. J. Richards. 1999. Global a mphibian declines: a problem in applied ecology. Ann ual Rev iew of Ecology and Syst ematics 30: 133 165. Babbitt, K. J. and G. W. Tanner. 2000. Use of temporary wetlands by anurans in a hydrologically modified landscape. Wetlands 20(2):313 322. Babbitt, K. J ., M. J. Baber, and T. L. Tarr. 2003. Patterns of larval amphibian distribution along a wetland hydroperiod gradient. Canadian Journal of Zoology 81:1539 1552. Baber, M. J., K. J. Babbitt, F. Jordan, H. L. Jelks, and W. M. Kitchens. 2005. Relationships amo ng habitat type, hydrology, predator composition, and distribution of larval anurans in the Florida Everglades. P ages 154 160 in W.E, Meshaka, Jr. and K.J. Babbitt (eds) Amphibians and Reptiles: statu s and conservation in Florida. Krieger Publishing Comp any, Malabar, Florida. Bailey, L. L., T. R. Simons, and K. H. Pollock. 2004. Estimating site occupancy and species detection probability parameters for terrestrial salamanders. Ecological Applications 14: 692 702. Baillie, J. E. M., C. Hilton Taylor, and S N. Stuart. 2004. IUCN Red List of Threatened Species. A Global Species Assessment. IUCN Gland, Switzerland and Cambridge, UK. Bartelt, P. E., A. L. Gallant, R. W. Klaver, C. K. Wright, D. A. Patla, and C. R. Peterson. 2011. Predicting breeding habitat fo r amphibians: a spatiotemporal analysis across Yellowstone National Park. Ecological Applications 21(7): 2530 2547. Bartlett, R. D. and P. P. Bartlett. 1999. A field guide to Florida reptiles and amphibians. Gulf Publishing Company, Houston, TX. Beyer, H. L. 2004. Hawth's Analysis Tools for ArcGIS. Available at:
48 Brandt, L. A., J. E. Silvera, and W. M. Kitchens. 2002. Tree Islands of the Arthur R. Marshall Loxahatchee National Wildlife Refuge. Pages 311 335 in Sklar, F. H. and A. van der Valk (Eds). Tree Islands of the Everglades. Kluwer Academic Publishers, Dordrecht, The Netherlands. Brandt, L. A., K. M. Portier, and W. M. Kitchens. 2000. Patterns of change in tree islands in the Arthur R. Ma rshall Loxahatchee National Wildlife Refuge from 1950 to 1991. Wetlands 20(1): 1 14. Bridges, A. S. and M. E. Dorcas. 2000. Temporal variation in anuran calling behavior: implications for surveys and monitoring programs. Copeia 2: 587 592. Burnham, K. P., an d D. R. Anderson. 1998. Model selection and inference. Springer Verlag, New York, New York, USA. Cauble, K. and R. S. Wagner. 2005. Sublethal effects of the herbicide Glyphosate on amphibian metamorphosis and development. Bulletin of Environmental C ontami nation and Toxicology 75: 429 435. Cook R P T. A. Tupper, P. W. C. Paton, and B. C. Timm. 2011 Effects of temperature and temporal factors on anuran detection probabili ties at Cape Cod National Seashore: Implications for long term monitoring. Herpetological Conservation and Biology 6:25 39. Cushman, S. A. 2006. Effects of habitat loss and fragmentation on amphibians: a review and prospectu s. Biological Conservation 128: 2 31 240. Davis, S. M., and J. C. Ogden. 1994. Introduction. In: S. M. Davis and J. C. Ogden. Everglades: The Ecosystem and Its Restoration. Delray Beach, Florida. St. Lucie Press: 3 8. de Solla S. R., L. Shirose, K. J. Fernie, G. Barrett, C. S. Brousseau a nd C. A. Bishop. 2005. Effect of sampling effort and species detectability on volunteer based anuran monitoring program s. Biological Conservation 121: 585 594. Dixon, A. D., W R. Cox, E M. Everham III and D W. Ceilley 2011. Anurans as biological indica tors of restoration success in the greater everglades ecosystem. Southeastern Naturalist 10(4): 629 646. Dodd, C. K., Jr. 2009. Conservation and management. Pages 507 527 in C. K.Dodd, Jr. (ed). Amphibian Ecology and Conservation. A Handbook of Techniques. Oxford University Press, Oxford. Dorazio, R. M. and J. A. Royle. 2005. Estimating Size and Composition of Biological Communities by Modeling the Occurrence of Species. Journal of the American Statistical Association 100(470): 389 398 Duellman, W. E. and A. Schwar tz. 1958 Amphibians and Reptiles of Southern Florida. Volume 3. Bulletin of the Florida State Museum, Gainesville.
49 Gallant, A. L., R. W. Klaver, G. S. Casper, and M. J. Lannoo. 2007. Global rates of habitat loss and implications for amphibia n c onservation. Copeia 4: 967 979. Gibbons, J. W., D. E. Scott, T. J. Ryan, K. A. Buhlmann, T. D. Tuberville, B. S. Metts, J. L. Greene, T. Mills, Y. Leiden, S. Poppy, and C. T. Winne. 2000. The global decline of reptiles, dj vu amphibians. BioScience 50: 653 666. Harwell, M.C., D. D. Surratt, D. M. Barone, and N. G. Aumen. Conductivity as a tracer of agricultural and urban runoff to delineate water quality impacts in the northern Everglades. Environmental Monitoring and Assessment 147(1 3):445 462. Heyer, W. R., M. A. Donnelly, R. W. McDiarmid, L. C. Hayek, and M. S. Foster. 1994. Measuring and monitoring biological diversity: standard methods for amphibians. Smithsonian Institute Press. Washington D.C. Hines, J. E. 2006. PRESENCE software to estimate patch oc cupancy rates. USGS PWRC < http://www.mbr pwrc.usgs.gov/software.html> Ph.D. Dissertation. University of Florida, Gainesville, FL, USA. Kery, M. and J. A. Roy le. 2008. Hierarchical Bayes estimation of species richness and occupancy in spatially replicated surveys. Journal of Applied Ecology 45: 589 598. Kroll, A. J., K. Risenhoover, T. McBride, E. Beach, B. J. Kernohan, J. Light, J. Bach. 2008. g stream occupancy and detection probability parameters of stream associated amphibians in commercial forests of Oregon and Washington, USA For est Ecology and Management 255: 3726 3735. Leach, S. D., H. Klein, and E. R. Hampton. 1971. Hydrologic effects of water control and management of south eastern Florida. Report of Investigations 60, Florida Bureau of Geology, Tallahassee, Florida, USA. Loveless, C. M. 1959. A study of the vegetation in the Florid a Everglades. Ecology 40(1): 2 9. MacKenzie, D. I. 2006. Modeling the probability of resource use: the effect of, and dealing with, detecting a species imperfectly. The Journal of Wildlife Management 70 ( 2 ): 367 374. MacKenzie, D. I ., and J. D. Nichols 2004. Occupancy as a surrogate for abundance estimation. Anim al Bi odiversity and Conservation 27: 461 467. MacKenzie, D. I., J. D. Nichols, J. E. Hines, M. G. Knutson, and A. B. Franklin. 2003. E stimating site occupancy, colonization, and local extinc tion when a species is detected imperfectly. Ecology 84:2200 2207. MacKenzie, D. I ., and L. L. Bailey. 2004. Assessing the fit of site occupancy models. Journal of Agricultural, Biological and Environmental Statistics 9 ( 3 ): 300 318.
50 MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A. Langtimm 2002. Estimating site occupancy rates when detection probabilities are less than one. Ecology 83: 2248 2255. MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, J. E. Hines and L. L. Bailey. 2005. Occupancy estimation and modeling: inferring pat terns and dynamics of species occurrence Elsevier, San Diego, USA. Mann, R. M. and J. R. Bidwell. 1999. The toxicity of glyphosate and several glyphosate formulations to four species of southwestern Australian frogs. Archives of Environmental C ontaminatio n and Toxicology 36: 193 199. Mattfeldt, S. D., L. L. Bailey, and E. H. C. Grant. 2009. Monitoring multiple species: estimating state variables and exploring the efficacy of a monitoring progra m. Biological Conservation 142: 720 737. Meshaka, W. E., W. F. Lo ftus, and T. Steiner. 2000. The h erpetofauna of Everglades National Park. Florida Scientist 63(2): 84 103. Mitchell, J. C., and M. J. Lannoo. 2005. Gastrophryne carolinensis (Holbrook,183 6). Eastern narrow mouthed toad. Pages 501 503 in M. J. Lan noo ( ed ) Am phibian declines: conservation status of United States species. University of California Press, Berkeley, USA. Noss, R. F., M. D. Murphy. 1997. The Science of Conservation Planning. Island Press, Washington, D.C. Rice, K. G., J. H. Wa ddle, M. E. Crocket, B. M. Jeffrey, and H. F. Percival. 2004. Herpetofaunal Inventories of the National parks of South Florida and the Caribbean: Volume I. Everglades National Park. U.S. Geological Survey, O pen File Report 2004 1065, Fort Lauderdale, Flori da. Rice, K. G., J. H. Waddle, B. M. Jeffrey, A. N. Rice, and H. F. Percival. 2005. Herpetofaunal Inventories of the National parks of South Florida and the Caribbean: Volume III. Big Cypress National Pre serve. U.S. Geological Survey, O pen File Report 2005 1300, Fort Lauderdale, Florida. Richardson, J. R., Bryant, W. L., Kitchens, W. M., Mattson, J. E., and K. R. Pope. 1990. An evaluation of refuge habitats and relationships to water quality, quantity, and hydroperiod: a synthesis report. Final Report to A rthur R. Marshall Loxahatchee National Wildlife Refuge, Boynton Beach, Florida, USA Royle, J. A., and J. D. Nichols. 2003. Estimating abundance from repeated presence absence data or point counts. Ecology 84:777 790. Royle, J. A. and M. Kery. 2007. A Baye sian state space formulation of dynamic o ccupancy models. Ecology 88(7): 1813 1823.
51 Ryan, T. J. and C. T. Winne. 2001. Effects of hydroperiod on metamorphosis in Rana sphenocephala American Midland Naturalist 145(1):46 53. Sewell, D., T. J. C. Beebee, and R. A. Griffiths. Optimising biodiversity assessments by volunteers: the application of occupancy modeling to large scale amphibian surveys. Biological Conservation 143:2102 2110. Sheridan, C. D. and D. H. Olson 2003. Amphibian assemblages in zero order basins in the Oregon Coast Range. Cana dian Journal of Forest Research 33:1452 1477. Shirose, L. J., C. A. Bishop, D. M. Green, C. J. MacDonald, R. J. Brooks, and N. J. Helferty. 1997. Validation tests of an amphibian call count survey technique in Ontario, Canada. Herpetologica 53(3):312 320. Sklar, F. H. and A. Van der Valk, 2003. Tree islands of the Everglades, Kluwer academic Publishers, Boston, Massachusetts. Sklar, F. H., C. McVoy, R. V. Zee, D. Gawlik, D. Swift, W. Park, C. Fitz, Y. Wu, D. Rudnick, T Fontaine, S. Miao, A. Ferriter, S. Krupa, T. Armentano, K. Tarboton, K. Rutchey, Q. Dong, and S. Newman. 1998. Chapter 2: The hydrological needs of the Everglades. Redfield, G. (ed.) The Everglades 1999 Interim Report, 2.1 2.70. SFWMD, West Palm Beach, F L. Smith, L. L., W. J. Barichivich, J. S. Staiger, K. G. Smith, and C. K. Dodd, Jr. 2006. Detection probabilities and site occupancy estimates for amphibians at Okefenokee National Wildlife Refuge. American Midland Naturalist 155(1):149 161. Snodgrass, J. W., M. J. Komoroski, A. L. Bryan Jr., and J. Burger. 2000. Relationships among isolated wetland size, hydroperiod, and amphibian species richness: implications for wetland regula tions. Conservation Biology 14:414 419. Stuart, S. N., J. S. Chanson, N. A. C ox, B. E. Young, A. S. L. Rodrigues, D. L. Fischman, and R. W. Waller 2004. Status and Trends of Amphibian Dec lines and Extinctions Worldwide. Science 306:1783 1786. Tanadini, L G ., and B. R. Schmidt. 2011. Population size influences amphibian detection probability: implications for biodiversity monitoring programs. PLoS ONE 6(12): e28244. Thompson, S. K. 1992. Sampling. John Wiley, New York, New York, USA. Van Buskirk, J. 2005. Local and landscape influence on amphibian occurrence and abundance. Ecology 86(7):1936 1947. Vitt, L. J., J. P. Caldwell, H. M. Wilbur, and D. C. Smith. 1990. Amphibians as harbingers of decay. BioScience 40: 418. Waddle, J. H. 2006. Use of amphibians as ecosystem indicator species. University of Florida Dissertation, Gainesvill e FL.
52 Walls, S. C., J. H. Waddle, and R. M. Dorazio. 2011. Estimating occupancy dynamics in an anuran assemblage from Louisiana, USA. Journal of Wildlife Management 75(4):751 761. Weir, L. A. and M. J. Mossman. 2005. North American Amphibian Monitoring P rogram (NAAMP). P ages 307 313 in M. J. Lannoo (ed.). Amphibian Declines: conservation status of United States amphibians. University of California Press. Weir, L. A., J. A. Royle, P. Nanjappa, and R. E. Jung. 2005. Modeling anuran detection and site occupa ncy on North American Amphibian Monitoring Program (NAAMP) routes in Maryland. Journal of Herpetology 39(4):627 639. Wellborn, G. A., D. K. Skelly, and E. E. Werner. 1996. Mechanisms creating community structure across a freshwater habitat gradient. Ann ual Rev iew of Ecology and Syst ematics 27:337 63. Welsh, H. H. Jr. and L. M. Ollivier. 1998. Stream amphibians as indicators of ecosystem stress: Ecological Applications 8(4): 1118 1132. Whiles, M. R, K. R. Lips, C. M. Pringle, S. S. Kilham, R. J. Bixb y, R. Brenes, S. Connelly, J. C. Colon Gaud, M. Hunte Brown, A. D. Huryn, C. Montgomery, and S. Peterson. 2006. The effects of amphibian population declines on the structure and function of Neotropical stream e cosystems. Fr ontiers in Ecology and the Environment 4:27 34. Woodmansee, S. W., K. Bradley, and S. Hodges. 2005. Systematic reconnaissance flights and exotic plant species mapping at selec ted national wildlife refuges in Florida. Final Report. Submitted to US Fish and Wildlife Service, Sanibel, FL. Yoccoz, N. G., J. D. Nichols, and T. Boulinier. 2001. Monitoring of biological diversity in space and time. Trends in Ecology and Evolution 16:446 453.
53 BIOGRAPHICAL SKETCH Ryan Lynch was born in Palo Alto, California, in 1 981. Ryan spent most of his childhood in the San Francisco Bay Area, specifically in a small corner of the city of Fremont known as Niles. He received his high school diploma from Washington High School in 2000. Following high school Ryan attended Eckerd C ollege in St. Petersburg, Florida, where he receive d a Bachelor of Arts degree in anthropology, biology and environmental s tudies in 2004. During his college career Ryan worked on a number of research projects, including a long term turtle population stud y in north Florida, a hermit crab behavioral study in Costa Rica, and a semester long wildlife program in Kenya. After receiving his degree, Ryan moved to Ecuador to pursue an assistant research position at a biol ogical research station on the e astern slop e of the Andes. In 2006, Ryan returned to Florida to work for the University of Florida as a technician on an exotic fern project. During the next three years Ryan also assisted with an assortment of herpetological projects, including work with alligators, crocodiles, and the exotic Burmese pythons. In late 2008, Ryan enrolled at the University of Florida to pursue a Master of Science degree in the School of Natural Resources and Environment.