This item is only available as the following downloads:
1 HABITAT SELECTION OF FEMALE OSCEOLA TURKEYS ACROSS MULTIPLE SPATIAL SCALES IN SOUTH FLORIDA By MITCHELL R. BLAKE A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIRE MENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2011
2 2011 Mitchell R. Blake
3 To the town of Hopewell, Pennsylvania
4 ACKNOWLEDGMENTS I thank t he Florida Fish and Wildlife Conservation Commission and the University of Flori da for providing financial and logis tical support for the project. I greatly appreciate the support and guidance provided by Dr. William Giuliano, Dr. Holly Ober, and John Denton, specifically during data analysis and the writing process. Finally, I woul d like to thank the volunteers and technicians including, Johnny Olson, Kendall Seibert, David Jayroe, Jesse Swift, Ingrid Klongland, Brandon Haslick, Charles Haley, and Devin Bosler for their assistance with data collection.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ...... 4 LIST OF TABLES ................................ ................................ ................................ ................ 6 ABSTRACT ................................ ................................ ................................ .......................... 7 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .......... 9 Study Objectives ................................ ................................ ................................ ......... 11 Study Sites ................................ ................................ ................................ .................. 11 2 METHODS ................................ ................................ ................................ .................. 13 Data Collection ................................ ................................ ................................ ............ 13 Data Analysis ................................ ................................ ................................ .............. 14 Statist ical Analysis ................................ ................................ ................................ ...... 17 3 RESULTS ................................ ................................ ................................ .................... 22 Second Order Habitat Selection ................................ ................................ ................. 22 Third Order Habitat Selection ................................ ................................ ..................... 24 4 DISCUSSION ................................ ................................ ................................ .............. 33 Second Order Habitat Selection ................................ ................................ ................. 33 Third Order Habitat Selection ................................ ................................ ..................... 38 Summary ................................ ................................ ................................ ..................... 41 LIST OF REFERENCES ................................ ................................ ................................ ... 42 BIOGRAPHICAL SKETCH ................................ ................................ ................................ 47
6 LIST OF TABLES Table page 2 1 Variable s used in a priori models to predict habitat selection at second and third order of selection ................................ ................................ ...... 19 3 1 A priori models AIC values, and model weights used to predict habitat ................................ ................ 26 3 2 Parameter estimates, 95% confidence interval s, and odds ratios for variables used in best order of selection ................................ ................................ ................................ .... 28 3 3 Best a priori model(s) from each variable category predicting habitat ................................ ................. 29 3 4 A priori models, used to predict habitat third order of selection ................................ ...................... 29 3 5 Parameter estimates, 95% confidence interval s, and odds ratios for variables used in best a pr third order of selection ................................ ................................ ................................ ... 32 3 6 Best a priori model(s) from each variable category predicting habitat third order of selection ................................ ..................... 32
7 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Masters of Science HABITAT SELECTION OF FEMALE OSCEOLA TURKEYS ACROSS MULTIPLE SPATIAL SCALES IN SOUTH FLORIDA By MITCHELL R. BLAKE August 2011 Chair: William Giuliano Major: Wildlife Ecology and Conservation Landscapes and land use practices continue to change throughout Florida decreasing the qu antity and quality of wildlife habitat, possibly impacting female Osceola turkey ( Meleagris gallopavo osceola ) habitat selection in south Florida. Information on habitat selection of female Osceola turkeys is lacking and may be hindering management effort s. This study evaluated second and third order of selection using logistic regression and AIC model selection criterion at two study sites in south Florida. f emale turkeys selected fo r cover types that provided sufficient cover and food availability (e.g., mesic flatwoods, wet flatwoods, wet prairie ) interspersion of the landscape habitats associated with greater amounts of edge and habitats that were not roller chopped. At Johnson condition or structure of adjacent habitats caused hens to be located further from edge and hens avoided h abitats that were not roller chopped due to un favorable ground story characteristics associated with thick woody vegetati on I suggest concentrating management efforts on habitats composed of thick, woody shrubs and managing these habitats with successive roller chopping to reduce and maintain the density of understory cover and to promote the
8 amount of edge available in th e landscape Further research is needed to examine the effects of management applications on female wild turkey habitat selection across biologically important seasons.
9 CHAPTER 1 INTRODUCTION Landscapes and land use practices continue to change through out Florida and wildlife habitat continues to be altered and lost. Urban development, fire exclusion, conversion to agriculture, invasive species, road construction, and incompatible land management practices are degrading and fragmenting the landscape a nd causing loss of habitat and population declines for many wildlife species (Noss et al. 1995 Miller et al. 2000 ). For example, urban development and changes in management practices have led to the decline in the use of prescribed fire as a manag ement t ool. Consequently, the quantity and quality of wildlife habitat has decreased, especially for those species that are affected by understory vegetation. Osceola wild turkeys ( Meleagris gallopavo osceola ) hereafter wild turkey, have benefitted f rom tradi t i onal timber and rangeland practices in Florida but these practices are being replaced by more profitable forms of intensive development which degrade upland wildlife habitat ( Williams 1992). These changing landscapes and land use practices may be affect ing habitat selection by wild turkeys. Wild turkey response to changing landscapes and land use practices are not well documented in Florida. Efforts to mitigate the effects of development in Florida, such as state acquisition of land, conservation ease ments, and regulatory devices (e.g., endangered species laws) have been used to offset the impacts of development on wildlife habitat ( Williams 1992). Studies on other subspecies of wild turkeys (e.g., Meleagris gallopavo sylvestris ) have examined the eff ect of habitat and management alterations on wild turkey habitat selection in response to changing landscapes in other regions of the Southeast (Smith and Teit el baum 1986, Miller and Conner 2005, Morgan
10 et al. 2006). In Mississippi, managed pine landscape s have become more intensive (e.g., shorter harvest rotations, greater vegetation control; Miller and Miller 2004) requiring a better understanding of wild turkey ecology within managed pine for ests (Miller and Conner 2005), such as those that predominate in south Florida (i.e., pine flatwoods). Management techniques, including application of prescribed fire, roller chopping and creation and maintenance of forest openings are being used in effort to enhance habitat conditions specifically in pine flatwood s of Florida by discouraging the growth of woody plants and encourag ing growth of early successional plant communities ( Tanner and Marion 1990). Studies suggest that wildlife species including turkeys, inhabiting pine flatwoods benefit from the increase in herbaceous foliage stimulated by fire (Huber and Steuter 1984, Tanner and Marion 1990, Madden et al. 1999, Main and Richardson 2002) The benefits of prescribed fire and other tools (e.g., roller chopping) of habitat restoration and management have be en documented, but the s cientific literature regarding habitat management practices and their effect on wild turkeys in Florida is lacking and as a result is poorly understood. The general lack of knowledge of Osceola turkeys and their habitat hinders our conservation and management efforts. Researchers have found wildlife select habitat at multiple scales (Johnson 1980) and the need to examine wild turkey habitat at multiple spatial scales has been documented ( Miller et al. 1999, Morgan et al 2006) In order to improve management we must understand wild turkey habitat selection relative to management practices. S tudies in other regions have documented wild turkey habitat selection in response to different management practices ( Smith and Teitelbaum
11 1986 Swanson et al. 1994, Palmer et al. 1996, Miller et al. 2000, Thogmartin 2001, Miller and Conner 2005 ) but none in Florida Williams ( 2006 ) provided a basis for a habitat management strategy for wild turkeys in Florida consisting of 1) highly intersper sed habitat types, 2) maximized edge effects, and 3) optimal arrangement and sizing of these components, but further research is required before more detailed guidelines can be established Due to the lack of knowledge of Osceola turkeys and their managem ent in Florida, it is important that we improve understanding of how the lack of management and different management practices affect turkey s Study Objectives My objective s of this study were 1) to determine habitat cover type ( s ) selected by female wild t urkeys 2) to determine which landscape metrics (e.g., edge) have the greatest impact on female wild turkey habitat selection 3) to examine the effect of prescribed fire and roller chopping at four temporal scales on female wild turkey habitat selection and 4) to study the effect of prescribed burning and roller chopping applied to specific habitat cover types at four temporal scales upon female wild turkey habitat selection using Additionally, we wanted to identify 5) the most important category of variables (i.e., habitat cover type, landscape, management, and habitat treatment) influencing habitat selection by female wild tur keys at each order of selection. Study Sites The study took place on two sites in South central Florida. The first site was Three Lakes Wildlife Management Area (WMA) in Osceola County, Florida. Turkey research was conducted on a 6,273 ha section of the Three Lakes WMA Three Lakes WMA consist ed of pine flatwoods, intermixed with swamps, marshes, and wet prairie.
12 The Florida Fish and Wildlife Conservation Commission operate deer, hog, wild turkey, and small game hunts on the property. The second study site was Longino Ranch, a privately owned ranch in Sarasota County, Florid a. The 4,047 ha Longino Ranch has roughly 2,020 ha of pine flatwoods, dry and wet prairies, and oak cabbage palm hammocks. The other 2,020 ha is used for cattle, sod, and citrus production. Longino Ranch was historically managed for timber, but is now being managed for both cattle and timber. The ranch owners allow deer, feral hogs, and wild turkey to be hunted on the property.
13 CHAPTER 2 METHODS Data Collection I used rocket nets to capture female wild turkeys (Bailey et al. 1980) during t he winter months (December early March) of 200 7 2010 on both study sites I baited the rocket net sites with cracked corn or a mixture of cracked corn, sorghum, and wheat I removed captured turkeys from nets and placed them in a cardboard box built s pecifically to accommodate wild turkeys. I fitted turkeys with standard metal leg bands and backpack type radio transmitters with mortality sensors (ATS transmitters, model A1540, 69 80 grams [not including harness material ] : weighing<3.5% of birds body w eight). I weighed and classified females as adult or juvenile based on tail, wing, and size characteristics (Dickson 1992) and most turkeys were given a shot of vitamin E to offset possible handling effects. I released turkeys 10 45 minutes after captu re. I remotely located females by triangulation (White and Garrott, 1990) using a hand times/week during nesting season (~1 March 15 July) to establish home range data. I est ablished telemetry stations across study sites to facilitate triangulation. One bearing aerial photographs to reduce error. Visual sighting locations of radio transmitte re d turkeys were recorded with a hand held Glob al Positioning System (GPS). Turkeys that survived multiple field season s were considered new independent samples for each ye ar
14 Data Analysis Turkey locations were estimated using Location of a Signal (Ecological Software Solutions 2010 ). I imported estimated turkey locations into a geographic information system (GIS) H ome range locations were calculated for each hen using the home range estimator extension, Home Range Tools (Rodgers et al. 2007) in ArcGIS 9.3 (Environmental Systems Research Institute 2009). A 95% fixed kernel analysis with least squares cross validation for a smoothing factor (Worton 1989, Seaman and Powel l 1996) was used for turkeys with a sufficient number of locations to produce unbiased necessary based on the ratio of standard deviations before the fixed kernel method was app lied (Rodgers et al. 2007). A 99% fixed kernel analysis was conducted on all locations each year at each study site to delineate study area boundaries and define available habitat. I downloaded Florida Natural Areas Inventory (FNAI) Cooperative Land Cover Map shapefiles ( Florida Natural Areas Inventory 2010 ) and projected these layers into a universal transverse mercator (UTM) coordinate system to match the coordinate system of our collected data (Miller and Conner 2007). Habitat cover types were defined from cooperative land cover maps (Table 2 1). Dates of recent land management activities were provided by study site managers United States Geological Survey orthophoto quadrangles were used to map land use history (i.e. prescribed burning and roller c hopping) and landscape variables (i.e. water impoundments edge, and roads) into shapefiles. All variables were grouped into one of four categories: 1) habitat cover type variables, presented as measured area of different cover types, 2) landscape variab les, presented as linear metrics of length, density, area, or distance, 3) management
15 variables, representing human treatment to forest and rangelands (ie., prescribed fire and roller chopping), and 4) habitat treatment, presenting a combination of habitat cover types and management variables. Five habitat cover type variables (developed, bottomland forest, successional hardwood forest, sand hill, and xeric hammock) that made up less than one percent of each study area were lumped into one variable categor Due to similarities, habitat cover type variables mesic hammock Clearing and unimproved pasture were combined into one variable (abandoned) based on the simila rities between variables and my observations during field seasons Road and edge metrics were measured using the Spatial Statistics Calculator extension in ArcGIS 9.3, and roads were buffered with a distance of 2.5 meters on each side representing my best approximation of the road widths on study sites. I used the Neare st Feature extension in ArcGIS 9.3 to measure the distance from locations to landscape features including roads, water, and edge and averaged measurements for each individual hen. Managem ent variables were separated into four categories (1 = < 6 months managed, 2 = 6 months 2 years managed, 3 = >2 yrs managed, 4 = no management) based on time elapsed since treatment and recommendations from Williams (199 2 ). The no management category, e. g., 4, was dedicated to areas that had no previous management or management that was earlier than the land managers recorded history. Habitat treatment combinations were based on habitats targeted for management on study sites. In order to evaluate habi tat selection, I defined use and availability using 80). To define availability at
16 second order of selection, random points were generated throughout the study area (Miller and Conner 2 007 Tirpak et al. 2010 ), usi ng the data management tool in ArcGIS 9.3. I generated random points by dividing area (ha) of each study area by the average area (ha) of female turkey home range specific to that study site and year. The result was then mult iplied by 10 to determine the total amount of random points. The number of random points ranged from 81 249 depending on the size of the study area and average home range size Each random point was buffered with the circular equivalent of average home r ange size for that study site and year. To determine actual and random home range composition, kernel home ranges and buffered random points were intersected with maps depicting habitat cover types landscape, management, and habitat treatment variables. Areal extent of habitat cover types management, and habitat treatment variables within actual and random home ranges was calculated using the Spatial Statistics Calculator extension in ArcGIS 9.3 and converted to hectares. second order of selection were defined in actual and random home ranges as lengths, densities, and areas using the Spatial Statistics Calculator extension in ArcGIS 9.3 r oportions of different habitat cover types management, and habitat treatment variables within actual home ranges defined availability (Thompson and Fr itzell 1989). Use was determined by summing the number of locations in a habitat cover type management, or habitat treatment polygon and dividing b y the total number of locations specific to an individ ual hen (Miller et al. 1999). To define availability of landscape metrics, random points were generated throughout a e
17 Neare st Feature extension in ArcGIS 9.3 was used to measure the distance from random and actual locations to landscape features Statistical Analysis I used logistic regression to orders of selecti on. Because no studies existed on which to base my a priori models, my model composition was based on prior knowledge and beliefs derived from field observations, research hypothes e s, and related literature (Burnham and Anderson 2002). I developed a mode l for each variable in a category ( e.g ., habitat, n =26; Miller and Conner 2005 Doherty et al. 2006), a suite of models for combinations of variables in each category ( e.g ., habitat, n =15), a global model containing all variables in a category, and a null model using no variables. Global models in each category were used to estimate fit and overdispersion of the data (Burnham and Anderson 2002). In total, we developed 1 25 order selection analysis and 1 26 ird order of selection analysis. I Information Criterion (AIC c ) adjusted for small sample size (n/K <40) for model selection (Burnham and Anderson 2002). Models within 2 AIC c unit of the best models were considered supported by the data (Burnham and Anderson 2002). Models ranging from 3 10 AIC c units from the best model received limited empirical support but still merit interpretation (Burnham and Anderson 2002). Akaik e weights ( w ) were used to rank models (Burnham and Anderson 2002) and the model with the largest weight in each category was considered to be the best (most parsimonious) variable or combination of variables explaining selection in that category If sev eral models were <2 AIC c from the best model and were considered nested (i.e., contained a subset of variables present in the competing models), relative variable importance was estimated using
18 A kaike weights (Burnham and Anderson 2002). Correlation analysis of variables ma y have been a viable option to reduce the number of models and increase global model fit, but my study was exploratory in relation to the Osceola wild turkey sub species habitat selection and i nclusion of all variables with minimal combination or reduction is necessary to effectively answer our biological questions of interest and potentially link land management (Chamberlain 2008). Subsequently, a t each order of selection, best Criterion (AIC c ) adjusted for small sample size (n/K<40 ; Burnham and Anderson 2002) to consider which variable category was most supported by the data All statistical analyses were conducted using SYSTAT 12 (SYSTAT 2007).
19 Table 2 1. Variable catego ries, names abbreviations, and their definitions used in a of selection among female Osceola wild turkeys in south Florida, 2008 2010. Variable Category Variable Abbreviation D escription Habitat a Abandoned clearing b A Ha or % of abandoned clearing c Agriculture Ag Ha or % of agriculture Basin swamp BS Ha or % of basin swamp Baygall BG Ha or % of of baygall Depression marsh DM Ha or % of depression marsh Dome swamp DS Ha or % of dome swamp Dry prairie DP Ha or % of dry prairie Hammock H Ha or % of hammock Improved pasture IP Ha or % of improved pasture Mesic flatwoods MF Ha or % of mesic flatwoods Other O Ha or % of other Pine plantation PP Ha or % of pine plantation Scrub S Ha or % of scrub Scrubby flatwoods SF Ha or % of scrubby flatwoods Shrub bog SB Ha or % of shrub bog Upland hardwood forest UHF Ha or % of upland hardwood forest Wet flatwoods WF Ha or % of wet flatwoods Wet p rairie WP Ha or % of wet prairie Landscape Distance to edge d Distance edge Distance to nearest edge ( m ) Distance to roads Distance road Distance to nearest road ( m ) Distance to water Distance water Distance to nearest water body (m) Edge Edge m Amount of habitat edge ( m ) Edge total Edge tot Total amount of habitat edge and roads (m) Edge density Edge den M of edge/ha of home range Road area e Road area Roads ( ha ) Road density Road den M or roads/ha of home range Road m Road m Roads ( m ) Water Water Water ( ha ) Management Burn1 B1 Ha or % of area burned <6 months Burn2 B2 Ha or % of area burned 6 months 2 years Burn3 B3 Ha or % of area burned > 2 years Burn4 B4 Ha or % of area with no recent burn history Chop1 C1 Ha or % of area roller chopped < 6 months Chop2 C2 Ha or % of area roller chopped between 6 months and 2 years Chop3 C3 Ha or % of area roller chopped >2 years Chop4 C4 Ha or % of area with no recent roller chopping history Habitat Treatment Dry pra irie1 DP1 Ha or % of dry prairie burned <6 months Dry prairie2 DP2 Ha or % of dry prairie burned 6 months 2 years Dry prairie3 DP3 Ha or % of dry prairie burned >2 years Dry prairie4 DP4 Ha or % of dry prairie with no recent burn history Mesic fl atwoods1 MF1 Ha or % of mesic flatwoods burned <6 months
20 Table 2 1. Continued. Variable Category Variable Abbreviation Description Habitat Treatment Mesic flatwoods2 MF2 Ha or % of mesic flatwoods burned 6 months 2 years Mesic flatwoods3 MF3 Ha or % of mesic flatwoods burned >2 years Mesic flatwoods4 MF4 Ha or % of mesic flatwoods with no recent burn history Pine plantation1 PP1 Ha or % of pine plantation burned <6 months Pine plantation2 PP2 Ha or % of pine plantation burned 6 months 2 years Pine plantation3 PP3 Ha or % of pine plantation burned >2 years Scrubby flatwoods1 SF1 Ha or % of scrubby flatwoods burned <6 months Scrubby flatwoods2 SF2 Ha or % of scrubby flatwoods burned 6 months 2 years Scrubby flatwoods3 SF3 Ha or % of sc rubby flatwoods burned >2 years Wet flatwoods1 WF1 Ha or % of wet flatwoods burned <6 months Wet flatwoods2 WF2 Ha or % of wet flatwoods burned 6 months 2 years Wet flatwoods3 WF3 Ha or % of wet flatwoods burned >2 years Wet flatwoods4 WF4 Ha or % of wet flatwoods with no recent burn Wet prairie1 WP1 Ha or % of wet prairie burned <6 months Wet prairie2 WP2 Ha or % of wet prairie burned 6 months 2 years Wet prairie3 WP3 Ha or % of wet prairie burned >2 years Wet prairie4 WP4 Ha or % of w et prairie with no recent burn history Chop_dry prairie1 CDP1 Ha or % of dry prairie roller chopped <6 months Chop_dry prairie2 CDP2 Ha or % of dry prairie burned 6 months 2 years Chop_dry prairie4 CDP4 Ha or % of dry prairie with no recent roller chopping history Chop_mesic flatwoods1 CMF1 Ha or % of mesic flatwoods roller chopped <6 months Chop_mesic flatwoods2 CMF2 Ha or % of mesic flatwoods burned 6 months 2 years Chop_mesic flatwoods3 CMF3 Ha or % of mesic flatwoods roller chopped >2 ye ars Chop_mesic flatwoods4 CMF4 Ha or % of mesic flatwoods with no recent roller chopping history Chop_scrubby flatwoods1 CSF1 Ha or % of scrubby flatwoods roller chopped <6 months Chop_scrubby flatwoods2 CSF2 Ha or % of scrubby flatwoods burned 6 mon ths 2 years Chop_scrubby flatwoods3 C SF3 Ha or % of scrubby flatwoods roller chopped >2 years Chop_scrubby flatwoods4 CSF4 Ha or % of scrubby flatwoods with no recent roller chopping history
21 Table 2 1. Continued. Variable Category Variable Abb reviation Description Habitat Treatment Chop _wet flatwoods4 CWF4 Ha or % of wet flatwoods with no recent roller chopping history a Habitat variables are FNAI cover types. b Abandoned clearing, hammock, and other are combinations of variables that are si milar or constitute <1% of the study area annually. c d Distance varia e Roads included paved and unpaved roads, trails, and fire breaks wide enough for a vehicle. Road area included the road itself and a 2.5 (m) buffer on each side.
22 CHAPTER 3 RESULTS Second Order Habitat Selection During the three years of the study, we caught 142 wild turkey hens on both study sites and fitted them with radio transmitters. Of these hens 83 had a sufficient numb er 30) to produce unbiased home ranges Home ranges were combined across study sties to evaluate habitat selection with an average of 43.95 10.40 locations/ home range Data supported three model s in the habitat category. There was a 0.401 probability that th e model containing dry prairie, wet prairie, mesic flatwoods, and wet flatwood s was the best model a 0.385 probability that the model containing dry prairie, wet prairie, mesic flatwoods, wet flatwoods, scrubby flatwoods, and pine plantation was the best model, and a 0.175 probability that the model containing dry prairie, wet prairie, mesic flatwoods, wet flatwoods, hammock, and abandoned opening was the best model based o n Akaike weights (Table 3 1). When 95% confidence intervals were calculated around parameter estimates for the three best models, three variables (wet prairie, mesic flatwoods, and wet flatwoods) did not have confidence intervals overlapping zero (Table 3 2). Parameter estimates and odds ratios suggested a positive relationship between these three variables and the probability of use by female wild turkeys. Since more than one model fell within two AIC c units of the best model and the most parsimonious was nested within competing models, relative importance of variables was estimated fo r the three variables with confidence intervals that did not overlap zero in the most parsimonious model Relative importance was similar for wet praire (0.996) and mesic flatwoods (0.996), and slightly lower for wet flatwoods (0.961). Four models
23 were w ithin 10 c units of the best models but had limited empirical support. T hey included the model containing mesic flatwoods, abandoned clearing, and wet prairie c =6.03) the model containing mesic flatwoods and wet prairie c = 7.83) the mod el mesic flatwoods, wet prairie, and depression marsh c = 8.12) and the model containing wet prairie, dome swamp, depression marsh, and shrub bog ( c = 9.61; Table 3 1). The model containing edge m was the best model in the landscape category and no other model was within 2 AIC c units of the best mod el (Table 3 1) A 95% confidence interval was estimated around the parameter estimate for edge (m) and concluded a confidence interval not containing zero. Parameter estimates and odds ratios suggest ed a positive association between edge m and female turkey use (Table 3 2 ). The model containing edge m and road m was within 10 c units of the best model c = 2.10; Table 3 1) but had limited empirical support In the management category, the mod el containing burn4 and chop4 and the model containing chop4 were considered the best model s (Table 3 1 ). 95% confidence intervals were estimated around the parameter estimate and concluded that the variable chop4 had a confidence interval not overlapping zero Parameter estimates and odds ratios suggested a positive relationship between Chop4 and female turkey use (Table 3 2 ). There was a 0.99 probability of the mesic flatwoods4, wet flatwoods4, dry prairie4, wet prairie4, pine plantation4, chop mesic fl atwoods4, chop wet flatwoods4, and chop dry prairie4 model being the best model in the habitat treatment category based on Ak aike weights (Table 3 1). When 95% confidence intervals were estimated, three of
24 the eight variables in the model differed from ze ro (Table 3 2). Parameter estimates and odds ratios suggested a positive association between chop mesic flatwoods4, chop wet flatwoods4 and chop dry prairie4 and female turkey use. When comparing the best models from each category, the model containing ed ge selection (Table 3 3 ). No other model was within 2 AIC c units of the best model Parameter estimates concluded a positive relationship between the amount of edge and the pr obability of use by female turkeys (Table 3 2) Third Order Habitat Selection The model containing the variable scrub was the best model for the habitat category (Table 3 4 ). The model containing agriculture, abandoned clearing and c = 9. 1 1; Table 3 4 ) had limited empirical support based on the c distance from the best model Due to the large size of the parameter estimate and the confidence interval containing zero the model containing scrub was not interpreted The model cont aining distance edge, distance road, and distance water for the landscape catego ry was the best model (Table 3 4 ). Distance edge had a 95% confidence interval s and suggested a negative relationship between the variable and probabi lity of female turkey use (Table 3 5 ). Five models were within 10 c units of the best model and included the model containing distance edge and distance road c = 2.59) the model distance edge c = 3.01) the model distance c = 5.18) the model distance c = 8.01) and the null model ( C c = 9.72; Table 3 4).
25 Two models were supported by the data for the management category, and include d t he model chop4 ( w = 0.577) and the model containing burn4 and chop4 ( w = 0.417; Table 3 4 ). Chop4 was the only variable with a 95% confidence interval that did not contain zero (Table 3 5 ). The parameter estimate and odds ratio suggested a negative relationship between chop4 and probability of use by female turkeys. The model containing scrubby flatwoods1 was the best model in the habitat treatment cate gory (Table 3 4 ). The 95% confidence interval for scrubby flatwoods1 included zero and did not permit further interpretation of the model (Table 3 5) The model containing dry prairie1, dry prairie2, chop dry prairie1, and chop dry prairie2 c = 7.14 ; Table 3 4) had limited empirical support. When comparing best models from each category, I eliminated models from consideration that I did not interpret The model burn4, chop4 and the model chop4 in the management category mo st accurately described habi tat selection by female wild 6 ). Confidence intervals estimated for both models concluded a confidence intervals not containing zero for the variable chop4. The parameter estimate and odds ratio suggested a negative relationship between chop4 and the probability of use by female turkeys (Table 3 5)
26 Table 3 1. A priori models arranged by category number of variables (K), second order corrected for small sample siz e (AIC c ), distance from the lowest AIC c c ), and model weights ( w i ) used to predict habitat selection of female Osceola turkeys in south Florida, 2008 2010, USA. Category Model K AICc w i a Habitat Cover Type DP,WP,MF,WF 4 289.66 0.00 4.01E 0 1 DP,WP,MF,WF,SF,PP 6 289.15 0.08 3.85E 01 H,A,DP,WP,MF,WF 6 290.73 1.66 1.75E 01 MF,A,WP 3 295.90 6.03 1.97E 02 MF,WP 2 297.85 7.83 8.01E 03 MF,WP,DM 3 297.99 8.12 6.92E 03 WP,DS,DM,SB 4 301.48 9.61 3.29E 03 DM,H,A,DS 4 325.43 11.82 1.09E 03 DS,BS,BF,H 4 336.62 35.77 6.86E 09 MF,WF 2 341.94 46.96 2.55E 11 WP 1 344.42 51.92 2.14E 12 MF 1 345.50 54.30 6.50E 13 DS 1 415.65 55.38 3.79E 13 BS 1 463.07 125.53 2.22E 28 PP 1 508.44 172.95 1.12E 38 DM 1 521.73 218.32 1.57E 48 S,SF 2 522.86 231.61 2.04E 51 SF 1 530.23 232.84 1.11E 51 DP 1 554.30 240.11 2.92E 53 Ag, IP,A 3 556.83 264.18 1.73E 58 H,SHF,BHF,UHF 4 570.55 266.96 4.30E 59 H,A 2 573.77 280.89 4.06E 62 UH,MH 2 580.29 283.75 9.74E 63 BG 1 591.40 290.27 3.74E 64 WF 1 591.76 301.28 1.52E 66 H 1 596.35 301.64 1.27E 66 S 1 612.49 306.23 1.28E 67 Ag 1 648.44 354.22 4.86E 78 UHF 1 650.06 358.32 6.25E 79 SB 1 652.76 359.94 2.78E 79 A 1 654.10 362.64 7.21E 80 IP 1 655.13 363.98 3.69E 80 O 1 666.04 365.01 2.20E 80 Null 0 724.1 423.84 3.70E 93 Landscape Edge m 1 282.72 0.00 0.7 4 Road m, edge m 2 284.82 2.10 0.2 6 Edge total 1 294.02 11.30 2.59E 03 Road density 1 312.27 29.55 2.83E 07 Road m 1 312.44 29.72 2.6E 07 Road area 1 312.56 29.84 2.45E 07 Edge density 1 331.64 48.92 1.76E 11 Water 1 536.88 254.16 4.77E 56 Null 0 724.10 441.38 1.06E 96 Management B4,C4 2 327.12 0.00 0.6 9 C4 1 328.69 1.57 0.31 B1,B2,B3,C1,C2,C3 6 352.60 25.48 2.02E 06 B1,B2,B3 3 359.24 32.12 7.26E 08 B2,B3,C2, C3 4 373.17 46.05 6.86E 11 B1,B2,C1,C2 4 388.67 61.55 2.96E 14
27 Table 3 1. Continued. Category Model K AICc w i Management B2 1 436.15 109.03 1.45E 24 B2,C2 2 437.90 110.78 6.04E 25 B3 1 457.17 130.05 3.95E 29 B1,C1 2 493.10 165.98 6. 23E 37 B1 1 502.52 175.40 5.61E 39 C1,C2,C3 3 541.83 214.71 1.63E 47 C1 1 593.92 266.80 7.98E 59 B4 1 636.60 309.48 4.31E 68 C2 1 640.34 313.22 6.64E 69 C3 1 648.56 321.44 1.09E 70 Null 0 724.10 396.98 4.3E 87 Habitat Treatment MF4,DP4,WP4,W F4,PP4,CMF4,CDP4, CWF4 7 330.52 0.00 0.99 CMF4 1 358.61 28.09 7.97E 07 MF1,MF2,MF3,WP1,WP2,WP3 6 359.57 29.04 4.94E 07 MF1,MF2,MF3,DP1,DP2,DP3 6 367.38 36.85 9.95E 09 CMF1,CMF2,CMF3,MF1,MF2,MF3 6 377.47 46.94 6.41E 11 CMF1,CMF2,CDP1,CDP2,MF1,MF2, DP1, DP2 8 393.47 62.94 2.15E 14 WP1,WP2,WP3 3 394.77 64.25 1.12E 14 MF1,MF2,DP1,DP2,WP1,WP2 6 400.24 69.71 7.28E 16 MF1,MF2,CMF1,CMF2 4 405.70 75.18 4.73E 17 MF1,MF2 2 422.30 91.78 1.18E 20 MF2,WF2,DP2,WP2 4 431.06 100.54 1.47E 22 MF2,DP2,WF2, WP2,PP2 5 432.64 102.12 6.7 0 E 23 MF2,DP2 2 434.11 103.59 3.21E 23 MF2,WF2,DP2 3 435.86 105.34 1.34E 23 WP1,WP2 2 450.59 120.07 8.47E 27 MF2 1 471.92 141.40 1.98E 31 MF3 1 486.80 156.28 1.16E 34 WF2 1 488.83 158.31 4.21E 35 MF1,DP1 2 497.64 16 7.12 5.14E 37 MF1,WF1,DP1 3 499.04 168.52 2.55E 37 MF1,WF1,DP1,WP1 4 501.23 170.71 8.53E 38 MF1,DP1, WF1,WP1,PP1 5 502.59 172.07 4.33E 38 MF1 1 522.74 192.22 1.82E 42 WF3 1 531.23 200.71 2.61E 44 WP3 1 531.23 200.71 2.61E 44 PP2 1 537.81 207. 29 9.74E 46 CMF1,CMF2,CMF3, CDP1,CDP2 5 545.42 21 2 02 9.11 E 47 CDP4 1 558.97 228.45 2.47E 50 CMF1,CMF2,CDP1,CDP2 4 568.41 237.89 2.2 0 E 52 CDP1,CDP2,DP1,DP2,DP3 5 572.99 242.80 7.92 E 53 WF1 1 571.07 240.55 5.84E 53 WP1 1 571.07 240.55 5.84E 53 DP1,DP2 2 575.86 245.34 5.32E 54 DP1,DP2,CDP1,CDP2 4 576.65 246.13 3.58E 54 DP2 1 578.86 248.34 1.19E 54 CSF4 1 579.15 248.63 1.03E 54 SF2 1 584.50 253.98 7.08E 56 CMF1,CDP1 2 594.74 264.22 4.23E 58 CMF1 1 602.31 271.79 9.6 0 E 60 PP3 1 605.18 274.66 2.29E 60 CWF4 1 618.87 288.35 2.43E 63
28 Table 3 1. Continued. Category Model K AICc w i Habitat Treatment DP1 1 622.04 291.52 4.99E 64 WP4 1 636.24 305.72 4.12E 67 MF4,DP4,WP4 3 636.84 306.32 3.04E 67 MF4 1 638.83 308.31 1.13E 67 PP1 1 639.75 309.23 7.12E 68 CMF2,CDP2 2 640.45 309.93 5.02E 68 SF1 1 641.18 310.66 3.48E 68 CMF2 1 643.10 312.58 1.33E 68 PP0 1 645.10 314.58 4.91E 69 DP3 1 647.78 317.26 1.28E 69 CMF3 1 649.08 318.56 6.71E 70 WP2 1 658.70 328.18 5.46E 72 SF3 1 667.20 336.68 7.79E 74 CSF3 1 688.04 357.52 2.32E 78 CSF2 1 690.96 360.44 5.4 0 E 79 CSF1 1 695.15 364.63 6.64E 80 a Probability that the model is the best of the a priori models in that category. Model abbreviations correspond to variable and variable descriptions in Table 2 1. Table 3 2 Parameter estimates, 95% confidence intervals (CI), and odds ratios (OR) for variables used in supported a priori models predicting habitat selection at Florida, USA, 2008 2010 95% CI Category Model a Variable Estim ate Lower Upper OR Habitat DP,WP,MF,WF DP 0.007 0.002 0.016 1.007 Cover Type WP 0.039 0.022 0.056 1.040 MF 0.004 0.002 0.006 1.004 WF 0.116 0.011 0.220 1.123 DP,WP,MF,WF,SF, PP DP 0.006 0.006 0.018 1.006 WP 0.044 0.026 0.0 62 1.045 MF 0.004 0.001 0.006 1.004 WF 0.083 0.021 0.186 1.086 SF 0.049 0.008 0.106 1.050 PP 0.006 0.023 0.011 0.994 H,A, DP,WP,MF WF H 0.001 0.019 0.021 1.001 A 0.003 0.003 0.008 1.003 DP 0.007 0.002 0.017 1.007 WP 0.038 0.01 7 0.059 1.039 MF 0.004 0.002 0.006 1.004 WF 0.111 0.007 0.215 1.117 Landscape Edge m Edge m b 0.000 0.000 0.000 1.000 Management B4,C4 B4 0.007 0.001 0.014 1.007 C4 0.012 0.010 0.015 1.013 C4 C4 0.014 0.011 0.016 1.014
29 Table 3 2. Continued. 95% CI Category Model Variable Estimate Lower Upper OR Habitat Treatment MF4, WF4, DP4, WP4, CMF4, CWF4, CDP4 MF4 0.003 0.012 0.017 1.003 WF4 0.138 0.321 0.045 0.871 DP4 0.019 0.169 0.131 0.981 WP4 0.059 0.019 0.137 1.061 CMF4 0.010 0.007 0.013 1.010 CWF4 0.091 0.000 0.183 1.095 CDP4 0.024 0.011 0.036 1.024 a ted c < 2). b Estimate and confidence interval are displayed as zero due to rounding. Model abbreviations correspond to variable and variable descriptions in Table 2 1. Table 3 3 Best a priori model(s) from each variable category predicting habit at turkeys in south Florida, USA, 2008 2010. Category Model a K AIC c c w i b Landscape Edge m 1 282.72 0.00 9.45E 01 Habitat Cover Type DP,WP,MF,WF 4 290.17 7.45 2.28E 02 H abitat Cover Type DP,WP,MF,WF,SF,PP 6 290.26 7.54 2.18E 02 Habitat Cover Type H,A,DP,WP,MF,WF 6 291.84 9.12 9.91E 03 Management B4,C4 2 327.12 44.40 2.16E 10 Management C4 1 328.69 45.97 9.85E 11 Habitat Treatment MF4,DP4,WP4,WF4,PP4,CMF4,CDP4, CWF4 8 333.07 50.35 1.10E 11 a c < 2). b Probability that the model is the best of the a priori models in that category. Model abbreviations correspond to variable and variable d escriptions in Table 2 1. Table 3 4 A priori models, number of variables (K), second Criterion corrected for small sample size (AIC c ), distance from the lowest AIC c c ), and model weights ( w i ) used to predict habitat selec third order of selection of female Osceola turkeys in south Florida, USA, 2008 2010. Category Model K AIC c c w i Habitat Cover Type S 1 89.46 0.00 0.99 Ag, IP,A 3 98.57 9.11 0.01 A 1 102.30 12.84 1.61E 03 H,A 2 103.21 1 3.75 1.02E 03 DM,H,A,DS 4 104.63 15.17 5.00E 04 BS 1 105.05 15.59 4.06E 04 MF,A,WP 3 105.90 16.44 2.65E 04 H,A,DP,WP,MF,WF 6 109.06 19.60 5.48E 05 SF 1 109.62 20.16 4.13E 05 PP 1 110.53 21.07 2.62E 05 IP 1 110.68 21.22 2.43E 05 DS,BS,BF,H 4 110.74 21.28 2.36E 05 SB 1 114.24 24.78 4.10E 06 DS 1 115.70 26.24 1.98E 06 WF 1 116.03 26.57 1.68E 06
30 Table 3 4. Continued. Category Model K AIC c c w i Habitat Cover Type O 1 116.19 26.73 1.55E 06 Ag 1 116.38 26.92 1.41E 06 H 1 116.40 26.94 1.39E 06 DP 1 116.48 27.02 1.34E 06 DM 1 116.93 27.47 1.07E 06 BG 1 116.94 27.48 1.06E 06 WP 1 116.95 27.49 1.06E 06 MF 1 117.06 27.60 1.0 0E 06 Null 0 117.06 27.60 9.99E 07 UHF 1 117.11 27.65 9.76E 07 DP,WP,MF,WF,PP 5 117.61 28.15 7.60E 07 MF,WF 2 118.13 28.67 5.86E 07 UH,MH 2 118.48 29.02 4.92E 07 WP,DS,DM,SB 4 118.79 29.33 4.21E 07 MF,WP 2 119.03 29.57 3.74E 07 H,SHF,BHF,UH F 4 119.24 29.78 3.36E 07 MF,WP,DM 3 120.92 31.46 1.45E 07 DP,WP,MF,WF 4 121.84 32.38 9.16E 08 Landscape Distance edge, Distance r oad Distance w ater 3 107.34 0.00 6.29E 01 Distance edge, Distance r oad 2 109.93 2.59 1.73E 01 Distance e dge 1 110.35 3.01 1.40E 01 Distance w ater 1 112.52 5.18 4.73E 02 Distance r oad 1 115.35 8.01 1.15E 02 Null 0 117.06 9.72 4.88E 03 Management C4 1 99.74 0.00 0.58 B4,C4 2 100.39 0.65 0.42 C2 1 116.58 16.84 1.28E 04 C3 1 116.73 16.99 1.19E 04 C1 1 116.74 17.00 1.18E 04 B3 1 116.76 17.02 1.17E 04 B4 1 116.88 17.14 1.10E 04 B1 1 117.01 17.27 1.03E 04 Null 0 117.06 17.32 1.00E 04 B2 1 117.08 17.34 9.96E 05 B1,C1 2 118.25 18.51 5.54E 05 B2,C2 2 118.45 18.71 5.02E 05 C1,C2,C3 3 120.02 20.28 2.28 E 05 B1,B2,B3 3 120.20 20.46 2.09E 05 B2,B3,C2,C3 4 121.10 21.36 1.33E 05 B1,B2,C1,C2 4 121.77 22.03 9.53E 06 B1,B2,B3,C1,C2,C3 6 124.08 24.34 3.01E 06 Habitat Treatment SF1 1 88.09 0.00 0.97 DP1,DP2,CDP1,CDP2 4 95.23 7.14 0.03 CDP1,CDP2 ,DP1,D P2 ,DP3 5 99. 11 11. 02 3. 91 E 03 CMF1,CMF2,CDP1,CDP2,MF1,MF2 DP1,DP2 8 102.28 14.19 8.04E 04 CSF1 1 103.26 15.17 4.92E 04 WP1 1 104.84 16.75 2.23E 04 WP1,WP2 2 106.78 18.69 8.45E 05 DP1 1 107.11 19.02 7.17E 05 CSF3 1 107.41 19.32 6.17E 05 WP1,W P2,WP3 3 108.24 20.15 4.07E 05 MF1,WF1,DP1,WP1 4 109.67 21.58 1.99E 05
31 Table 3 4. Continued. Category Model K AIC c c w i Habitat Treatment MF1,DP1, WF1,WP1,PP1 5 110.14 22.05 1.58E 05 CSF2 1 110.36 22.27 1.41E 05 CMF1,CMF2,CDP1,CDP2 4 111.10 23.01 9.74E 06 PP1 1 111.26 23.17 9 .00 E 06 CDP2 1 111.57 23.48 7.71E 06 CMF2,CDP2 2 112.28 24.19 5.4 0 E 06 MF1,MF2,DP1,DP2,WP1,WP2 6 112.45 24.36 4.98E 06 MF1,MF2,MF3,WP1,WP2,WP3 6 113.11 25.02 3.58E 06 CSF4 1 113.34 25.25 3.18E 06 DP 4 1 113.43 25.34 3.04E 06 CMF4 1 113.47 25.38 2.98E 06 CMF1 1 113.56 25.47 2.85E 06 CMF1,CDP1 2 113.73 25.64 2.62E 06 SF2 1 113.84 25.75 2.48E 06 WF4 1 114.88 26.79 1.47E 06 CMF1,CMF2,CMF3, CDP1,CDP2 5 114. 55 26. 46 1.8 E 06 WF2 1 114.97 26.88 1 .41E 06 CDP1 1 115.37 27.28 1.15E 06 MF1 1 115.44 27.35 1.11E 06 CMF2 1 115.73 27.64 9.63E 07 MF3 1 115.87 27.78 8.98E 07 CMF1,CDP1,CWP1 3 115.88 27.79 8.92E 07 WF3 1 116.00 27.91 8.42E 07 CWF4 1 116.06 27.97 8.17E 07 WF1 1 116.32 28.23 7.1 7E 07 WP3 1 116.41 28.32 6.86E 07 CMF3 1 116.70 28.61 5.93E 07 PP3 1 116.77 28.68 5.73E 07 PP2 1 116.83 28.74 5.56E 07 DP2 1 116.97 28.88 5.18E 07 SF3 1 117.03 28.94 5.03E 07 MF4 1 117.04 28.95 5 .00 E 07 Null 0 117.06 28.97 4.95E 07 WP4 1 117.08 28.99 4.9 0 E 07 WP2 1 117.09 29.00 4.88E 07 DP3 1 117.11 29.02 4.83E 07 MF2 1 117.11 29.02 4.83E 07 CDP4 1 117.11 29.02 4.83E 07 MF4,DP4,WP4 3 117.24 29.15 4.52E 07 MF1,MF2 2 117.48 29.39 4.01E 07 MF1,DP1 2 117.51 29.42 3.95E 07 CMF1 ,CMF2,CMF3,MF1,MF2,MF3 6 117.72 29.63 3.57E 07 MF1,MF2,CMF1,CMF2 4 118.06 29.97 3 .00 E 07 MF2,WF2,DP2 3 118.96 30.87 1.91E 07 MF1,WF1,DP1 3 119.04 30.95 1.84E 07 MF1,DP2 2 119.07 30.98 1.81E 07 DP1,DP2 2 119.07 30.98 1.81E 07 MF4,DP4,WP4,WF4,PP4 ,CMF4,CDP4, CWF4 8 120.40 32.31 9.34E 08 MF2,WF2,DP2,WP2 4 121.15 33.06 6.4 0 E 08 MF2,DP2,WF2,WP2,PP2 5 122.81 34.72 2.79E 08 MF1,MF2,MF3,DP1,DP2,DP3 6 125.97 37.88 5.77E 09 a Probability that the model is the best of the a priori models in that cate gory. Model abbreviations correspond to variable and variable descriptions in Table 2 1.
32 Table 3 5 Parameter estimates, 95% confidence intervals (CI), and odds ratios (OR) for variables used in best a priori models predicting habitat selection at John third order of selection of female Osceola wild turkeys in south Florida, USA, 2008 2010. 95% CI Category Model a Variable Estimate Lower Upper OR Habitat Cover Type S S 1.623E+05 5.526E+05 2.279E+05 0.000 Landscape Distance edge Distance r oad Distance edge 0.065 0.114 0.017 0.937 Distance water Distance road 0.004 0.002 0.011 1.004 Distance water 0.003 0.006 0.000 0.997 Management C4 C4 8.885 13.995 3.774 0.000 C4,B4 C4 10.499 16.699 4.299 0.000 B4 4.918 3.108 12.944 136.726 Habitat Treatment SF1 SF1 3.663E+05 1.145E+06 4.129E+05 0 .000 a c < 2). Model abbreviations correspond to variable and variable descri ptions in Table 2 1. Table 3 6 Best a priori model(s) from each variable category predicting habitat in south Florida, USA, 2008 2010. Category Model a K AIC c C c w i b Management C4 1 99.74 0.00 0.57 C4,B4 2 100.4 0 0.65 0.41 Landscape Distance edge, Distance road, Distance water 3 107.3 0 7.60 0.01 a C c < 2). b Probability that the model is the best of the a priori models in that category Model abbreviations correspond to variable and variable descriptions in Table 2 1. * Un interpretable best models were not considered for comparison.
33 CHAP TER 4 DISCUSSION Second Order Habitat Selection My study was the first to empirically examine habitat selection by female Osceola wild turkeys. Much of the literature pertaining to wild t urkey habitat selection in the S outheast has been focused on managed pine landscapes and the eastern subspecies (Morgan et al. 2006, Miller et al. 2007). Habitat selection of cover types by female wild up on ground story characteristics. Female wild turkeys selected for mesic flatwoods and wet flatwoods which can include dense groundstory vegetation consisting of various woody shrubs and vines Dense ground story vegetation or thickets are needed in a turkey s range primarily for nesting, esca pe cover, or concealment from predators (Stoddard 1963, Wright and Speake 1975, Hurst 1978 Williams and Austin 1988, Porter 1992, Palmer et al. 1996 ). Importance of mesic and wet flatwoods at this order of selection was likely a result of habitat prefere nce of nesting hens or habitat sampling during the pre incubation period (Badyaev et al 1996, Miller et al 1999). Mesic and wet flatwoods, with woody species such as saw palmetto ( Serenoa repens ) gallberry ( Il ex sp .), and wax myrtle ( Morella cerifera ; Florida Natural Areas Inventory 2010) can become excessively shrubby in the absence of management (i.e., prescribed fire, roller chopping ; Willcox and Giuliano 2010 ) or other natural disturbance s such as wild fires. In the absence of disturbance mesic an d wet flatwoods may create ideal nesting conditions but low overall suitability Undisturbed mesic and wet flatwoods with excessive shrub layers provide inferior habitat because turkeys avoid areas with dense understory (e.g., unburned pine flatwoods) wh en foraging due to limited lateral visibility
34 and the risk of predation (Williams 1992) Periodic disturbances in mesic and wet flatwoods can maintain a diversity of forbs and grasses and bene fit most wildlife species inhabiting flatwood s communities inc luding turkeys, by providing a variety of food and cover resources (Huber and Steuter 1984). The importance of herbaceous species as food is further highlighted by the selection of female wild turkeys in this study for w et prairies which often provide an array of herbaceous forage especially during the growing season, but offer little or no cover Wild turkeys have a predisposition to avoid large, open areas (Johnson 1980, Smith and Teitelbaum 1986) and most likely used wet prairies on the periphery whe re escape cover was in close proximity and was mo st likely dryer Habitat selection by wild turkeys can change by season (Morgan et al. 2006) making inferences concerning habitat selection difficult without the delineation of important biological seasons i selection of an attribute or resource by turkeys during one season and avoidance of it during another season can mask selection when annual home ranges are calculated, leading to lack of recognition of the importan ce of that attribute. Such may be the case with habitats containing saw palmetto an d other woody shrubs. Although pine flatwood s cover types may provide excellent nesting habitat or escape cover (Williams and Austin 1988 ) and were selected for during thi s study, high density undergrowth can impede vision ( Williams 1992 ) and movement (Morgan et al. 2006) and potentially harbor predators. Data collection was concentrated primarily during the reproductive period ( which includes pre incubation, nesting, and brood rearing periods ) and consequently may
35 reflect female turkey habitat preferences during this time Time lag in determining nesting attempts and nest fates may have also masked selection. Selection of mesic and wet flatwoods and dry prairies that were not chopped (habitat treatment category) and habitat cover types target ed for management that were not chopped (management category) provided further support of the importance of ground story vegetative structure on female turkey habitat selection. Roller chopping can result in considerable decreases in shrub abundance and cover ( Moore 1974, Tanner et al. 1988 Willcox and Giuliano 2010), but in the absence of fire or mechanical treatment, cover types containing saw palmetto and other woody shrubs become i ncreasingly dense at the ground story, reflecting conditions that are unlikely to invite regular use by turkeys ( Williams 1992). Conditions created by the absence of roller chopping can produce suitable nesting habitat or escape cover. Williams and Austin (1988) concluded that hens selected nesting sites in palmetto habitat with more than 30% overhead cover, conditions typically found in pine flatwood s cover types that have not been roller chopped. Researchers have demonstrated selection for dense vegetat ion at nest sites for wild turkey across their geographical range (Schmutz et al. 1989) and selection of habitats that were not chopped on my study sites was most likely a result of habitat sampling by hens prior to nest initiation (Badyaev et al 1996) and nesting/re nesting female turkey preference Turkeys in Florida do not occur commonly where vegetation structure of the understory is to o dense or where tree cover is to o sparse and often avoid dense understory vegetation when foraging due to the risk of predation ( Williams 1992)
36 Availability of some management rotations and habitat treatments may have been reduced because of the timing of the data collection period relative to treatments at the two study sites Longino Ranch concentrated their manage ment efforts on winter burns and conducted very little roller chopping while Three Lakes WMA performed winter burning, but much of their fire application and roller chopping was conducted during the growing season (summer). Differences in management seas ons may have reduced availability of some management treatments, particularly habitats burned or chopped <6 months. For instance, our data collection incorporated some of the summer season, but did not incorporate the fall and winter seasons so habitat sel ection was not observed in response to many of the summer burned and chopped habitats where use may be expected to increase specifically in brood rearing females, due to the regeneration of herbaceous foliage reduction of competing vegetation and increa sed forage availability The effects of timing (i.e., summer vs. winter) of prescribed burning or roller chopping has been acknowledged ( Tanner and Marion 1990, Main et al. 2002) and should be considered when implementing habitat management for wild turkey s Edge is the most influential factor affecting habitat selection in the landscape likely due to the simultaneous access to different cover types or greater richness of border vegetation (Leopold 1948). For instance, during the brood rearing period, broods are irregularly located more than 100 m from forest cover ( Williams 1992) due to their requirement for adjacent escape cover in close proximity. Researchers have rep orted on the importance of edge habitats for several gallinaceous birds (ruffed grouse, Bump et al. 1947, Scott et al 1998; ring necked pheasants, Robertson et al.
37 1993, Clark et al 1999), including Fearer and Stauffer (2003) who concluded a well intersper sed landscape likely provided ruffed grouse with resources and cover types available at short distances, minimizing energy expenditure and reducing the risk of predation associated with longer travel times. Other researchers have used edge as a way of est imating abundance, including Glennon and Porter (1999), who demonstrated the importance of edge as an indicator of turkey relative abundance in New York, where habitat suitability was enhanced by the increased edge between cover types, particularly between forested and agricultural areas. Rioux et al. (2009) concluded that male turkey density peaked with high levels of edge between forest and open habitats, further demonstrating the importance of edge at the landscape level. Although little is known about the relationship between edge and habitat selection in wild turkeys in Florida, Williams and Austin (1988) concluded that the palmetto ecotone was a preferred nesting location for female turkeys in south Florida due to the unique composition and condition s of the edge habitat. The importance of edge at this order of selection demonstrates that habitat selection is less dependent on cover type composition, but instead contingent upon the spatial arrangement or distribution of the habitat within the landsc ape for female turkeys in south Florida. Habitat cover types and the edge that delineates them are defined by the slightest change in hydrology (Williams 1992). Minor shifts in elevation can result in different cover types and ultimately a greater amount of interspersion in the landscape. Female turkeys had the opportunity to utilize a variety of different cover types on my study sites and as a consequence of a heterogeneous landscape, edge was utilized more than expected.
38 Third Order Habitat Selection S crub was the most influential habitat variable in the habitat cover type category at but a large parameter estimate and a confidence interval containing zero may lead to false inferences if interpreted Scrub was only p resent at one study site and was found in 4 of 83 home ranges and comprised an average of 3.42 % of the four home range composition The model containing agriculture, abandoned cl earing, and improved pasture had limited empirical support but merits addr essing based on the poor inferences that can be made from the best model. The importance of openings and clearings to female wild turkeys has been demonstrated by many (Everett et al. 1985, Ross and Wunz 1990, Williams 1992, Swanson et al. 1994, Peoples e t al. 1996,) and at multiple scales (Wilson et at. 2005 ). Williams and Austin (1988) documented an array of different habitat and food situations contributed by forest openings annual cycle in south Florida Agriculture on the study sites was primarily citrus production, providing ideal habitat for nesting and foraging during nesting and brood rearing seasons and was coupled with limited anthropogenic disturbances. Hens were located fu r ther from ed ge than expected and may be a result of condition or structure of adjacent cover types (Phalen et al. 1986, Smith and Teitelbaum 1986, Morgan et al 2006). Habitat c over types that are preferred but are juxtaposed with cover types that do not support regu lar use may result in edge avoidance due to the unwelcoming conditions of the latter. Researchers have proposed that female turkeys may select nest sites relative to the proximity of brood cover (Lazarus and Porter 1985) suggesting that edge selection m ay be weighted on the conditions of adjacent cover types. Glennon and Porter (1999) reported that highly contrasted edge
39 (i.e., edge between biologically important cover types) was important in determining and enhancing habitat suitability in New York A lthough edge was avoided at this order of selection, it is likely that edge was utilized between preferable cover types, but ultimately selected against due the inauspicious arrangement of cover types within the landscape. Additionally, female turkeys may have been located further from the edge due to the risk of predation. Thogmartin and Schaffer (2000) concluded that predators were associated with edge environs and a greater amount of edge use could increase the risk of predation of female wild turkeys. Schmitz and Clark (1999) reported that predation of ring necked pheasants was greatest along edges and quite possibly a result of increased vulnerability of prey or the value of edge habitats as travel corridors for predators. Few researchers have addre ssed the effect of edge on female turkey habitat selection at a finer scale and further consideration is warranted. Cover types that were not chopped was the most influential factor a ffecting habitat ird order of selection. Female turkeys selected against habitats that were not chopped most likely due to the vegetative structure of the habitat cover types. Cover types without recent roller chopping, unless routinely burned, will likely result in incr eased groundstory cover and possibly deter regular turkey use. R esearchers have documented selection of managed habitats over non managed habitats during different times of the year due to the conditions created by management, specifically prescribed burn ing (Hurst 1981, Hurst and Dickson 1992, Williams 1992, Palmer et al. 1996) but no studies have examined roller chopping, which may be the best method for initial, rapid reduction of saw palmetto (Wil l cox and G iu liano 2010). M y results did not suggest se lection of a
40 particular roller chopping rotation, possibly due to the prolonged effects of roller chopping. Studies have shown roller chopping can result in persistent reduction of height, crown cover, and abundance of shrubs, particularly saw palmetto, o n pine flatwoods sites for a period of at least three years (Tanner et al. 19 88 ) providing adequate habitat conditions and sustaining female wild turkey use over the duration of my study. My study did not address the subsequent use of multiple management techniques (i.e. burn x chop) on a habitat cover type which may have biased results of ma nagement on fe male turkey habitat selection. The avoidance of cover types that were not chopped demonstrates the importance selection. Suitable vegetative structure is imperative at this order of selection for several reasons. First, turkeys have adapted with fire in the south Florida landscape and without natural disturbance fostering herbaceous growth and discouraging the g rowth of woody plants, vegetative communities become increasingly un inhabitable. Second turkeys can usually find food whenever suitable cover requirements are met (Williams 1992) eliminating the need for specific species composition Lastly researcher s have shown that structure is important at a local level during distinct biological seasons, including pre incubation (Palmer et al. 1996), nesting ( Everett et al. 1985, Swanson et al. 1994 ), and brood rearing ( Metzler and Speake 1985). The best model r epresenting the habitat treatment category scrubby flatwoods1, performed poorly and resulted in a large parameter estimate and a confidence interval containing zero To avoid making spurious conclusions, no inferences were drawn. Poor performance by mod els in this category may be attributed to the reasoning that
41 p recise management i mplications on specific habitat cover types acknowledge the need for definitive or exclusive habitat resources which is contrary to the generalistic tendencies of wild turkeys Summary Management should focus on providing advantageous arrangements of habitat cover types with appropriate structure. R otational applications of r oller chopping and prescribed burning should be used to reduce shrubby vegetation, but care should be taken to ensure residual patches of cover within habitat cover types to provide sufficient nesting and esc ape cover (Wil liams 1992). Applying management to the landscape in a patchwork context would create a variety of conditions adjacent to one another, allow simultaneous access with minimal travel, and ultimately increase the amount of edge available. At a finer scale, consistent maintenance of cover types will ameliorate cover type conditio ns that deter wild turkey use (i .e., cover types that are not chopped ) and improve structure for subsequent years and possibly improve the carrying capacity of the landscape.
42 LIST OF REFERENCES Badyaev, A. V., T. E. Martin, and W. J. Etges. 1996. Habitat sampling and habitat selection by female wild turkeys: eco logical correlates and reproductive consequences. Auk 113:636 646. Bailey, W., D. Bennett Jr., H. Gore, J. Pack, R. Simpson and G. Wright. 1980. Basic considerations and general recommendations for trapping wild turkeys. Proc. Natl. Wild Turkey Symp., 4:251 261. Bump, G., R. W., Darrow, F. C., Edminster, and W. F. Crissey 1947. The Ruffed Grouse: Life History, Propogation, Management. Holling Press, Buffalo, New York, USA. Burnham, K. P. and D. R. Anderson. 2002. Model selection and inference: a practical information theoretic approach. Second edition. Springer Verlag, New York, New York, USA. Chamberlain, M. J. 2008. Are we sacrificing biology for statistics? Journal of Wildlife Management 72:1057 1058. Clark, W. R., R. A. Schmitz, and T. R Bogenshutz. 1999. Site selection and nest success of ring necked pheasants as a function of location in Iowa landscapes. Journal of Wildlife Management 63: 976 989. Dickson, J. G., 1992, editor. The wild turkey: Biology and management. Stackpole Boo ks, Harrisburg, Pennsylvania. Doherty, K. E., D. E. Naugle, B. L. Walker, and J. M. Graham. 2006. Greater sage grouse winter habitat selection and energy development. Journal of Wildlife Management 72:187 195. Ecological Software Solutions. 2005. Loca tion of a Signal, version 3.0.4. Ecological Software Soultions, Urnsch, Switzerland. Environmental Systems Research Institute. 2007. ArcGIS, version 9.3. Environmental Systems Research Institute, Redlands, California, USA. Everett, D. D., D. W. Speake and W. K. Maddox. 1985. Habitat use by wild turkeys in northwest Alabama. Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 39:479 488. Fearer, T. M. and D. F. Stauffer. 2003. Relationship of the ruf fed grouse ( Bonasa umbellus) home range size to landscape characteristics. American Midland Naturalist 150: 104 114.
43 Florida Natural Areas Inventory (FNAI). 2010. Guide to the natural communities of Florida: 2010 edition. Florida Natural Areas Inventor y, Tallahassee, FL
44 Miller, D. A., B. D. Leopold, G. A. Hurst, and P. D. Gerard. 2000. Habitat selection models for eastern wild turkeys in central Mississippi. Journal of Wildlife Management 64:765 776. Miller, D. A. and L. M. Connor. 2005. Seasonal and annual home ranges of female eastern wild turkeys in a managed pine landscape in Mississippi. Proceedings of th e Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 59:89 99. Miller, D. A. and L. M. Connor. 2007. Habitat selection of female wild turkeys in a managed pine landscape in Mississippi. Journal of Wildlife Management 71(3): 744 751. Miller, K. V., and J. H. Miller. 2004. Forestry herbicide influences on biodiversity and wildlife habitat in southern forests. Wildlife Society Bulletin 32:1049 1060. Morgan, J. J., S. H. Schweitzer, and J. P. Carroll. 2006. Summer Meleagris gallopavo silvestris use of a landscape dominated by agriculture and Pinus spp. plantations. Southeastern Naturalist 5(4):637 648. Moore, W. H. 1974. Some effects of chopping saw palmetto pineland threeawn range in south Florida. Journal of Range Manag ement 27:101 104. Noss, R. F., E. T. LaRoe III, and J. M. Scott. 1995. Endangered Ecosystems of the United States: a preliminary assessment of loss and degradation. Biological Report 28, National Biological Service. Washington, D.C., USA. Palmer, W. E., G. A. Hurst, K. D. Godwin, and D. A. Miller. 1996. Effects of prescribed burning on wild turkeys. Transactions of the North American Wildlife and natural Resources Conference 61:228 236. Peoples, J. C., D. C. Sisson, and D. W. Speake. 1996. Wild turk ey brood habitat use and characteristics in coastal plain pine forests. Proceedings of the National Wild Turkey Symposium 7:89 96. Phalen, P. S., G. A. Hurst, and W. J. Hamrick. 1986. Brood habitat use and preference by wild turkeys in central Mississip pi. Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 40:397 404. Porter, W. F. 1992. Habitat requirements. Pages 202 213 in J. Dickson, editor. The wild turkey: biology and management. Stackpole, Harr isburg, Pennsylvania, USA. Rodgers, A.R., A.P. Carr, H.L. Beyer, L. Smith, and J.G. Kie. 2007. HRT: Home Range Tools for ArcGIS. Version 1.1. Ontario Ministry of Natural Resources, Centre for Northern Forest Ecosystem Research, Thunder Bay, Ontario, Canad a.
45 Ross, S. R., and G. A. Wunz. 1990. Habitats used by wild turkey hens during the summer in oak forests in Pennsylvania. Proceedings of the National Wild Turkey Symposium 6:39 43. Rioux, S., M. B lisle, and JF. Giroux. 2009. Effects of landscape stru cture on male density and spacing patterns in wild turkeys ( Meleagris gallopavo ) depend on winter s everity Auk 126(3):673 683. Robertson, P. A., M. I. A. Woodburn, and D. A. Hill. 1993. Factors affecting winter pheasant density in British woodlands. J ournal of Applied Ecology 30:459 464. Schmitz, R. A. and W. R. Clark. 1999. Survival of ringed neck pheasant hens during spring in relation to landscape features. Journal of Wildlife Management 63:147 154. Schmutz, J. A. and C. E. Braun. 1989. Repro ductive performance of Rio Grande wild turkeys. Condor 91(3):675 680. Scott, J. G., M. J. Lovallo, G. L. Storm, and W. M. Tzilkowski. 1998. Summer habitat use by ruffed grouse with broods in central PA. Journal of Field Ornithology 69:474 485. Seaman, D. E., and R. A. Powell. 1996. An evaluation of the accuracy of kernel density estimators for home range analysis. Ecology 77:2075 2085. Seaman, D. E., J. J. Millspaugh, B. J. Kernohan, G. C. Brundige, K. J. Raedeke, and R. A. Gitzen. 1999. Effects of sample size on kernel home range estimates. Journal of Wildlife Management 63:739 747. Smith, W. P. and R. D. Teitelbaum. 1986. Habitat use of eastern wild turkey hens in southeastern Louisiana. Proceedings of the Annual Conference of the Southeaster n Association of Fish and Wildlife Agencies 40:405 415. Stoddard, H. L. 1963. Maintenance and increase of the eastern wild turkey on private lands of the coastal plain of the deep southeast. Tallahassee, FL: Tall Timbers Res. Sta. Bull. 3. 49. Swanson, D. A., J. C. Pack, C. I. Taylor, P. W. Brown, and D. E. Samuel. 1994. Habitat use of wild turkey hens in northwestern West Virginia. Proceedings of the Southeastern Association of Fish and Wildlife Agencies 48:123 133. SYSTAT. 2007. System for windows: Statistics. Version 12. Systat, Inc., Evanston, Illinois, USA. Tanner, G. W., J. M. Wood, R. S. Kalmbacher, and F. G. Martin. 1988. Mechanical shrub control on flawoods range in south Florida. Journal of Range Management 41:245 248.
46 Tanner, G. W. and W R. Marion. 1990. Wildlife habitat considerations when burning and chopping Florida range. Cooperative Florida Extension Service Fact Sheet WRS 6. 4p. Thogmartin, W. E., and B. A. Schaeffer. 2000. Landscape attributes associated with mortality events of wild turkeys in Arkansas. Wildlife Society Bulletin 28(4):865 874. Thogmartin, W. E. 2001. Home range size and habitat selection of female wild turkeys ( Meleagris gallopavo ) in Arkansas. The American Midland Naturalist 145:247 260. Thompson III, F. R. and E.K. Fritzell. 1989. Habitat use, home range, and survival of territorial male ruffed grouse. Journal of Wildlife Management 53:15 21. Tirpak, J. M., W. M. Giuliano, T. J. Allen, S. Bittner, J. W. Edwards, S. Friedhof, C. A. Harper, W. K. Igo, D. F. Stauffer, and G. W. Norman. 2010. Ruffed grouse habitat preference in central and southern Appalachians. Forest Ecology and Management 260:1525 1538. White, G. C. and R. A. Garrott. 1990. Analysis of wildlife radio tracking data. Harcort Brace Jovanovich, New York. Wi l lcox, E. V. and W. M. Giuliano. 2010. Seasonal Effects of prescribed burning and roller chopping on saw palmetto in flatwoods. Forest Ecology and Management 259:1580 1585. Wilson, W. B., M. J. Chamberlain, and F. G. Kimmel. 20 05. Seasonal space use and habitat selection of female wild turkeys in a Louisiana bottomland hardwood forest. Proceedings of the Annual Conference of the Southeastern Association of Fish and Wildlife Agencies 59:114 125. Williams, L. E., Jr., and D. H. Austin. 1988. The wild tur key in Florida. Florida Game and Freshwater Fish Commission. Technical Bulletin 10. Universit y of Florida Press. Gainesville. USA. Williams, L. E., Jr 1992. Florida Turkey. Pages 214 231 in J. Dickson, editor. The wild turkey: biology and management. Stackpole, Harrisburg, Pennsylvania, USA. Williams, L. E., Jr. 2006. Wild Turkey Hunting and Management. Real Turkeys LLC, Cedar Key, Florida. Worton, B. J. 1989. Kernel methods for estimating the utilization distribut ion in home range studies. Ecology 70:164 168 Wright, G. A. and D. W. Speake. 1975. Compatibility of the eastern wild turkey with recreational activities at Land Between the Lakes, Kentucky. Proceedings of the Annual Conference of the Southeastern As sociation Game and Fish Commission 29:578 584.
47 BIOGRAPHICAL SKETCH Mitchell Blake was born in Hopewell, Pennsylvania. He graduated from Shippensburg University of Pennsylvania in 2007 with a Bachelor of Science in biology. He has worked for the United St ates Army Corps. of Engineers at Raystown Lake, Pennsylvania on habitat and forest management projects, the Pennsylvania Fish and Wildlife Cooperative Research Unit in Curwensville, Pennsylvania on grassland sparrow research, and the Commonwealth of Massac husetts at Camp Edwards, Massachusetts on whip poor will spatial ecology. He moved to Florida in January 2008 to research Osceola wild turkeys for the University of Florida. In August, 2011, he received his Master of Science in wildlife ecology and conser vation. He enjoys hunting and fishing, competing in sporting clay com petitions, and motocross racing.