RIO GRANDE WILD TURKEY HEN HABITAT AND EDGE USE, SURVIVAL, AND REPRODUCTIVE CHARACTERISTICS IN THE TEXAS ROLLING PLAINS

RIO GRANDE WILD TURKEY HEN HABITAT AND EDGE USE, SURVIVAL, AND REPRODUCTIVE CHARACTERISTICS IN THE TEXAS ROLLING PLAINS by AMY ELIZABETH SMITH-BLAIR, B.A. A THESIS IN WILDLIFE SCIENCE Submitted to the Graduate Faculty of Texas Tech University in Partial Fulfillment of the Requirements for the Degree of MASTER OF SCIENCE December, 1993

ACKNOWLEDGMENTS I am grateful to the Hemphill Wells Foundation for providing the majority of the funding for this project and to Mr. Dan Griffis for his help in obtaining it. Thanks to Dr. Wright and the Big Brush account for supplemental funding, and to Dr. Carlton Britton for providing housing. Turkey trapping training and equipment were provided by Texas Parks and Wildlife Department, Oregon Department of Fish and Wildlife and Tom Keegan of Oregon State University. Mobile telemetry towers were built by Keith Blair and Duane Lucia of Precision Range Consulting, Shallowater, Texas. The cooperation of Ben, Paula, Riley and Mary Miller of the Miller Ranch is appreciated. Their interest, advice, careful observations, and complimentary air searches for dispersing wild turkeys made this project a success. Biotechnicians Mike Miller, John Kuhl, Mike Lloyd and Chad Phillip showed dedication beyond the call of duty. I appreciate the support of my committee chairman Dr. Scott Lutz and committee members Dr. Fred Bryant, Dr. Ernest B. Fish and Dr. Kent Rylander. Dr. Fish volunteered his lab and time (both personal and professional) to see this work completed. Dr. David B. Wester provided statistical guidance in the final stages of the project. Finally, I thank my family. Ramona Smith, my mother, provided encouragement and enthusiasm. She is a woman of courage and compassion. Keith, my husband, is resourceful and supportive. The project was a learning experience for both of us. ii

CONTENTS ACKNOWLEDGMENTS ii ABSTRACT v LIST OF TABLES vi LIST OF FIGURES vii CHAPTER I. INTRODUCTION II. METHODS Study Area Population Capture Prevalence of Mycoplasma gallisepticum Telemetry Equipment Tracking Sampling Periods and Frequency Nest Locations Poult-rearing Locations Roost Locations Pre-dispersal and Post-dispersal Locations Mean Dispersal Dates and Distances Mean Nest Initiation Dates Nest Success Survival Rates Home Range Determination Vegetation Type Mapping Map Construction 5 5 6 6 7 8 8 9 10 10 11 11 11 12 12 12 13 13 15 iii

Use and Availability of Vegetation Types and Edges 16 Vegetation Sampling 18 Analyses 19 Vegetation Type Selection 19 Vegetation 21 Ill. RESULTS AND DISCUSSION 22 Capture 22 Mycoplasma Culture and Serology 22 Mean Dispersal Dates 23 Distances to Nests; Dispersal Distances 24 Mean Nest Initiation Dates 24 Clutch Size 26 Nest Success 27 Survival Rates 30 Home Range 32 Vegetation Type Use 34 General Locations 34 Nesting 37 Poult-rearing 40 Roosting 41 1 00 m and 200 m Universal Edge Use 42 Specific Edge Use 44 1 00 m Wide Specific-Edge 44 200 m Wide Specific-Edge 4 7 Management Implications 53 REFERENCES 55 iv

ABSTRACT Reproductive characteristics and habitat use of adult Rio Grande wild turkey, hens in west Texas were studied. Members of a~loc~ using a winter roost in Garza and Borden Counties, were trapped and instrumented with radio transmitters. The hens dispersed later in 1990 (28 March to 7 May) than in 1991 (11 March to 12 April). Dispersal distances averaged 7.2 km in 1990 and 12.6 km in 1991. (Mean nest initiation occurred at approximately the same time each yea~17 May, 1990 and 8 May, 1991. Mean clutch sizes were 11.2 and 10.9 in 1990 and 1991, respectively. Nest success was 28.7 /o in 1990 and 34.8 /o in 1991. Predation (47.6%, 1990; 43.5%, 1991 ), hen abandonment (17.8 /o, 1990; none, 1991) and attended, but unhatched, clutches (5.9 /o, 1990; 21.7%, 1991) were responsible for unsuccessful nests. The birds were tested for Mycoplasma spp. but they were found not to be infected with M. gallisepticum, a known contributor to wild turkey infertility. March-August survival rates of adult hens were 52.6% in 1990 and 54.3% in 1991. Average home range size of adult Rio Grande wild turkey hens was 2920 ha in 1990 and 3208 ha in 1991. Hens selected the{rnesquite-hackberry brush and the mesquite brush vegetation)type for general locations and for poultrearing. They selected the mixed brush type for nesting and poult-rearing. All bird locations exhibited 50% visual obstruction (1.0 m observation height) at 23 m or less. Post-dispersal locations and roost sites showed hen selection for edges between vegetation types. (Roost sites occurred on the edge between cultivated fields and the mesquite brush habitat and between the mesquite shrub-grassland and mesquite brush habitats, combinations of open and more dense habitats) v

LIST OF TABLES 1. Descriptions of vegetation types (McMahan et al. 1984) for Texas Rolling Plains Rio Grande wild turkey study area. 14 2. Mean distance to nest from winter roost for adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. 25 3. Nest fate of adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. 28 4. Selection and avoidance (Neu et al. 1974) of vegetation types by adult Rio Grandewild turkey hens during pre-dispersal and post-dispersal, Justiceburg, Texas, 1990 and 1991. 35 5. Mean values of vegetation variables measured at five random plots in each habitat type, Justiceburg, Texas, 1991. 36 6. Selection and avoidance (Neu et al. 197 4) of vegetation types by adult Rio Grande wild turkey hens during nesting, poult-rearing and roosting, Justiceburg, Texas, 1990 and 1991. 38 7. Mean values of vegetation variables measured at different types of Rio Grande wild turkey locations, Justiceburg, Texas, 1990 and 1991. 39 8. Universal edge use by adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. 43 9. Specific-edge (100m wide) selection and avoidance (Neu et al. 1974) by adult Rio Grande wild turkey hens, during the time periods/activities that specific-edges were not used in proportion to their availibility, Justiceburg, Texas, 1990 and 1991. 45 10. Specific-edge (200 m wide) selection and avoidance (Neu et al. 1974) by adult Rio Grande wild turkey hens, during the time periods/activities that specific-edges were not used in proportion to their availibility, Justiceburg, Texas, 1990 and 1991. 48 vi

LIST OF FIGURES 1. Example of a vegetation type map and bird locations (dots). 17 2. Rainfall data at Texas Tech Experimental Ranch, Justiceburg, Texas, 1990 and 1991. 29 3. Heisey-Fuller (1985) March-August survival rates for adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. 31 vii

CHAPTER I INTRODUCTION Wild turkey~(meleagris gallopav~~~once ranged from southern Ontario to southern Mexico, and from the Rockies east to the Atlantic coast (Schorger 1966). Although uncontrolled(hunting and trapping1pontributed to a decrease in their numbers, habitat destruction, especially in the North and Northwest regions of their range{nearly drove them to extinctioil) Wild turkeys now occupy only a fraction of their original geographic range (Klein 1983). However, they are more widespread today than in the early part of this century due partly to reintroduction in many states but more importantly to habitat maintenance and improvement. (The largest~allinaceou~bird in North America) wild turkeys are comprised of six subspecies: 6he Eastern Turkey (M.g. silvestris), known as the forest turkey; the Florida Turkey (M.g. osceola), named after a famous Seminole Indian chief; the Mexican Turkey (M.g. gallopavo), the nominate race and ancestor to the domestic turkey; Gould's Turkey (M. g. mexicana), the turkey of Mexico, a resident of the Sierra Madre of Northwest Mexico; Merriam's Turkey (M.g. merriami), named for C. Hart Merriam, the first chief of the U.S. Biological Survey; and the Rio Grande Turkey (M.g. intermedia), named for its intermediate appearance (Schorger 1966)) The hens of this latter race are the subject of this study. Rio Grande wild turkeys are endemic to the South Central Plains of the United States and northern Mexico (Schorger 1966). Although wild turkeys are generally associated with open forest of mixed plant species that produce mast, 1

2 berries or nuts, the Rio Grande subspecies is typically found in more arid, shrub or grass dominated habitats. As in other subspecies of wild turkey, the Rio Grande females spend most of the year feeding and roaming in flocks, separate from males (although they may roost together). (A female produces fertile eggs for up to 56 days after the early spring mating (Lorenz 1950, Mosby and Handley 1943). For this reason, it is theoretically possible for a wild turkey hen to renest if her previous attempt is unsuccessful.) Rio Grande hens, like all~allinaceous birds, are ground nester~ (/a. slight depression in the ground, lined with debris and hidden by low growing plants serves as a nest site)johnsgard 1975). (Hens lay one egg per day; each egg laid one hour later each day (25 hour cycle~(schorger 1966). Clutch sizes average 9.8 eggs (Reagan and Morgan 1980). The female alone incubates the eggs for 28 days, turning them periodically to prevent the embryo from sticking to the shell's inner wall. During the nesting period,(she leaves the nest only for short durations in the morning and evening to feed~ Since the Rio Grande hen does renest after failed attempts, maintenance of nesting habitat may need to be extended into summer. In south Texas, Rio Grande turkey eggs generally (hatch during the end of May and the beginning of June \Baker et. al. 1980). Mycoplasma gallisepticum is a potential wild turkey pathogen that reduces egg production and hatchability in wild turkeys (Rocke et al. 1988). Prior to this study, it was not known if M. gallisepticum was present in the Rolling Plains population. The poults are especially vulnerable to both predators and rainfall (Markley 1967). By roosting under the hen's body, they are protected from moisture. (voung turkeys are able to fly at two weeks of age and are strong fliers at six

weeks (Knopp 1959). At this time, they begin roosting at night, above ground, 3 like adults) Turkeys roost at night and during adverse weather conditions (Crockett 1973). (A roost site generally is the tallest vertical structure in an area that does not limit the birds view of its surroundings)(kilpatrick et al. 1988, Lutz and Crawford 1987a). The geographic range of Rio Grande turkeys contains few large natural roosts; turkeys are frequently observed roosting on utility poles (Hauke 1975, Kothmann and Litton 1975). But the widely spaced utility poles cannot be expected to accommodate large winter flocks; the Rio Grande turkeys have the habit of returning to a traditional winter roost (Hauke 1975). Population maintenance depends on, among other factors, the availability of appropriate habitat types and structure. Whereas habitat.. preference.. refers to the choice of habitats without regard to their availability, habitat.. selection.. refers to the choice of available habitats. Habitat.. use.. is the result of habitat selection and does not depend necessarily on preference (Whittaker et al. 1973). Johnson (1980) proposed an order to the habitat selection process: first-order selection is the geographical range of the species, second-order selection is the home range of an individual or group within that geographical range, third-order selection is the use of habitat within the home range and fourth-order selection is the much more specific selection of food items, nest sites or roosts within a habitat. The selection of the subspecies, study area and flock predetermined the geographic range (first-order selection) and approximate home range of the flock (second-order selection). Vegetation type selection and the vegetative structure remained to be evaluated. The transition between habitats and vegetation types is called an ecotone and is usually richer in wildlife than adjoining plant communities (Thomas et al.

1979). Past research indicates that wild turkeys are an edge-seeking bird. 4 Donohoe and McKibben (1970) recommended that clear-cut openings be distributed throughout woodland turkey habitat to increase edge. Stoddard (1963) found that the mixture of field and woodland habitat types were important for the Eastern wild turkey on the southern Georgia coastal plain. Wild turkey selection for available universal-edge (the joining of any two habitat types) and the selection for available specific-edge Ooining of two specific habitat types) needs to be evaluated in the Texas Rolling Plains.

CHAPTER II METHODS Study Area The study area was located in the Rolling Plains (Gould 1975) of Texas in Garza and Borden Counties, just south and southeast of Post, near Justiceburg, on the Texas Tech Experimental Ranch and the Miller Ranch. The actual size and shape of the site was determined by the movements of radioed turkeys. Access to adjacent property was obtained as necessary. A caprock escarpment dissects the study area and the elevation ranges from 685 to 870 meters above sea level. The Rolling Plains are gently sloping and have well-defined drainage patterns (Richardson et al. 1965). The riparian zones are dominated by salt cedar (Tamarix.Q.). Adjacent upland areas are characterized by a mesquite (Prosopis glandulosa)/redberry juniper (Juniperus pinchotii) grassland community. Land use is dominated by cattle and oil production. Cattle ranching began about 1875 due to the availability of inexpensive land and suitable grasses. Supplemental feeding of livestock is heavy and small acreages of dryland grain crops are used for livestock production (Dixon 1975). Oil and gas wells are a source of income for landowners. The first oil well contract in Garza County was let in 1911 and between 1945-1965 the number of oil wells increased from 4 to 1645 (Richardson et al. 1965). There are roads throughout the study area, most of them built and maintained by the oil companies. The area has a warm-temperate subtropical climate. The average yearly rainfall is 47 em, falling mostly April through October. Spring is a variable 5

season of warm and cold fronts and duststorms. Thunderstorms in late spring 6 and early summer can bring damaging winds and hail. The freeze-free period usually is between April 5 and November 7 (Orton 1965). Population Local residents believe that during this century, prior to 1970, there were no wild turkey in Garza or Borden Counties. Several people whose families homesteaded here in the early part of the century confirm this report. In the early 1970's, 40-50 immature Rio Grande turkeys were transported from the King Ranch in south Texas to the Miller Ranch in Garza County (Miller 1990; Mitchell 1991 ). They were released below the caprock escarpment, near Justiceburg. When this project was initiated, in the fall of 1989, no wild turkey winter flocks were found below the escarpment on the Texas Tech Experimental Ranch. The closest winter flock large enough to provide sufficient hens for trapping was on the Miller Ranch, above the escarpment. They were roosting in pecan trees (Carya Q.) and a dead cottonwood (Populus deltoides) near an occupied residence. Capture (Birds were baited with cracked corn and/or domestic poultry seed near the winter roost. Birds were captured with rocket-nets)bailey 1980) in January and February of 1990 and 1991. (All captured birds were aged, sexed, weighed, and fitted with aluminum leg bands, provided by Texas Parks and Wildlife Department) Juvenile turkeys ( < 1 year) were distinguished from older birds (> 1 year) based on

7 characteristics of primaries (Larson and Taber 1980). (Sex was determined by the shape (round in female and squamate in male) and color (buff in female and black in male) of the breast feathers)larson and Taber 1980). Weight was recorded using a spring scale (Chatillion Precision lnstr., Kew Gardens, New York). Transmitters were fastened to the hens in a backpack style (Lutz and Crawford 1987b), with a nylon-coated rubber harness (Wildlife Materials, Carbondale, IL) looped under each wing. The harness was loose enough to allow full extension of the wings without stretching the harness so that there was no tension on the wings in flight. Birds were released at the trap site. Prevalence of Mycoplasma gallisepticum Serology and culture were used, in 1990 only, to investigate the prevalence of Mycoplasma species including M. gallisepticum. At capture, blood was drawn from the brachial vein and left to clot, allowing serum collection. Sterile cotton-tipped applicators were used to swab tracheas, then used to inoculate mycoplasma in a medium of Frey's broth (Rocke et al. 1988). For diagnoses, sera and broth cultures were shipped to Dr. David H. Ley, Avian Disease Research Laboratory, College of Veterinary Medicine, North Carolina State University, Raleigh, North Carolina. Sera were tested for antibodies tom. gallisepticum, M. synoviae, and M. meleagridis by the rapid plate agglutination (RPA) and hemagglutination inhibition (HI) tests according to established protocols (Rocke et al. 1988; Ley 1989). Fritz et al. (1992) advocate the RPA test for identifying culture-positive wild turkeys.

Telemetry Eguipment 8 Hens were fitted with AVM (Livermore, California), C-cell, H-module radio transmitters (150-152 MHz), weighing less than 100 grams and with an expected life of 3 years. Activity censors differentiated between active and resting birds. Signals were received by ATS (Isanti, Minnesota), AVM and Telonics (Mesa, Arizona) receivers and two retractable mobile antenna towers (Pollack et al. 1990) with AVM dual null antenna arrays constructed by Precision Range Consulting (Shallowater, Texas). A handheld yagi was used for ground searches. Tracking The mobile receiving units were situated at known locations on the Universal Transverse Mercator grid (UTM locations). UTM coordinates (Snyder 1987, U.S. Army 1973) were determined from 7.5' quadrangles produced by the U.S. Geological Survey (USGS). Radio contact between researchers allowed triangulation data to be gathered simultaneously from two stations. Compass rosette pointers at the base of the mobile antenna towers were zeroed onto beacons (stationary transmitters at known locations). Azimuths were recorded on each bird location relative to the beacons. Conversions to azimuths relative to true north, and bird location coordinates were determined using Computer Assisted Habitat Utilization Analysis Package (CHAP) (Whittaker et al. 1987). The triangulation system was tested (White and Garrott 1990) in order to determine the standard deviation of directional bearings, which was necessary to estimate precision and to establish error polygons around bird locations. Various distances to transmitter and multiple topography types throughout the

9 study area were represented in the testing. Seven locations were selected with bearings being estimated four times at each location. For comparison, true bearings were, again, computed by CHAP. Sampling Periods and Frequency Turkeys exhibit three broad behavior patterns in a given day. Wandering/feeding behavior is evident both morning and late afternoon. The middle part of the day is spent loafing and they roost between sunset and sunrise. These behaviors and their associated time periods defined the sampling periods for this study, so days were not partitioned evenly. The morning-feeding period (after leaving roost and 4 hours after sunrise), the evening-feeding period (4 hours before sunset or roost), and the loafing period (the midday hours between feeding periods) assured that location data were collected throughout the day. Each hen was located one-two times in each of the three daytime sampling periods, in a two-week interval, from March through August, 1990 and 1991. Time allowing, the interval was shortened to one week. Independence of locations was assured by allowing at least 24 hours between each of a bird's locations. A disruption of the above behavior patterns occurs when hens initiate nesting and begin clutch incubation. Nesting birds did not contribute to the general vegetation type use data or to the roosting data. Brood-rearing hens (> 1 poult) were monitored with the same frequency as other hens. A bird's locations from March through August of 1990 and 1991 were divided into five categories: nesting, poult-rearing, roosting, all others predispersal and all others post-dispersal. Locations with associated error polygons of more than two hectares were eliminated.

10 Nest Locations Nest sites were determined by flushing those hens that (according to remote telemetry) had been in the same location for 20 days. Twenty days was considered sufficient time investment by the hen to avoid observer-induced abandonment. Nests of non-radioed hens, discovered inadvertently, did not contribute to the nest file, in order to avoid the bias associated with.. easily found.. nests. No differentiation was made between first, second and third nests because a hen's nest could be destroyed or abandoned in the egg-laying stage or early in incubation and not be found. Poult-rearing Locations Poult-rearing files were comprised of locations contributed by hens while they were with a brood. They were followed from hatch through August or until the poults were destroyed by predators of enviromental conditions. Twice weekly visual observations of these hens confirmed their classification as poultrearing hens. The locations for poult-rearing hens were established with remote triangulation. Those locations not adhering to the two hectare error polygon elimination rule were culled from the poult-rearing file. Although a radioed-hen was classified as poult-rearing or non-poult-rearing, remote sensing prevented classification of non-radioed hens with which she was moving. Therefore, it should be noted that, although poult-rearing hens were definitely moving with poults, non-poult-rearing hens may or may not have been.

Roost Locations 1 1 Each bird was found at roost once per week, with the goal of discovering all roosts used by radioed birds in the study area. Because only one roost was used during the winter, all roosts were located after the average dispersal date from the winter roost, for each year of the study. Like nest sites, roost locations were determined exactly and had no error polygons. Pre-dispersal and Post-dispersal Locations After nest, poult-rearing, and roost sites were removed from a bird's file, her remaining locations were culled for two-hectare error polygons. These remaining locations were subdivided into pre-dispersal and post-dispersal time periods. A bird was deemed to have dispersed when her transmitter's signal could not be received, using a handheld antenna from the winter roost. If a bird did not move out of radio contact from the winter roost, even if she had begun nesting or using a nearby roost, her individual location file was partitioned into pre- and post-dispersal based on the average dispersal date of dispersing radioed hens. Finally, the pre-dispersal points from all birds were joined in one file, the post-dispersal points in another for each year. Mean Dispersal Dates and Distances Dispersing birds were identified when they moved out of radio contact from the winter roost. The date of a bird's dispersal was recorded and the average of these was the mean dispersal date, for that year. The Mann-Whitney test statistic, applicable to interval scale data (Zar 1984}, tested for a difference between dispersal dates the two years.

12 Schmutz and Braun (1989) defined dispersal distance from the center of the flock's winter range to a nest site. Similarly, in this study, dispersal distance was measured from the winter roost to a nest site, for those birds that moved out of radio contact from the winter roost. Mean Nest Initiation Dates All nests, including renests, contributed to the computation of mean nest initiation date, each year. For successful nests, nest initiation dates were estimated from the number of eggs (1 egg laid/day) and the 28-day incubation period. The initiation dates for unsuccessful nests were estimated by subtracting the egg laying period from the day that remote sensing indicated a stationary transmitter (Schmutz and Braun 1989). The Mann-Whitney test statistic, applicable to interval scale data, tested the difference between the nest initiation dates of 1990 and 1991. Simple linear regression determined the value of the nest initiation date in predicting clutch size. Nest Success Daily nest survival rates were calculated and expanded to the 28-day incubation period to determine nest success, using the Mayfield method (Mayfield 1961 ). The exposure days assigned to a given nest were the number of days between the onset of incubation (determined through remote sensing) and hatch, predation or abandonment. Survival Rates Monthly survival rates were calculated using the Heisey-Fuller (1985) method of estimating rates from the number of transmitter-days, the number of

13 mortalities and the number of days in a given month. Birds with transmitters that failed or were lost and then began re-transmitting did not contribute to survival data during the periods in which no signal was received. They did contribute to survival data in later time periods when they began re-transmitting (Heisey and Fuller 1985). Birds that died within two weeks of being trapped were considered trapping mortalities and did not contribute to survival data. An expansion of the Heisey-Fuller monthly survival rates yielded a survival rate from March through August each year. Home Range Determination All of a hen's locations, whether established remotely or visually, and regardless of her activity (nesting, poult-rearing, roosting or other) contributed to her home range size determination. The area of the home range was calculated using a Minimum Convex Polygon (Mohr 1947). March-August home range size was calculated only for those hen's surviving through August. The hens returned to their winter roost in the fall and the winter roost location appeared in the location file for each bird at least once, so March-August home range sizes are expected to approximate minimum annual home ranges. Vegetation Type Mapping The vegetation available to the radioed birds was defined as that existing in the study area (an area encompassing all bird locations established during the two year study). Vegetation types were based on those classified by Texas Parks and Wildlife Department {TPWD) (McMahan et al. 1984) (Table 1 ). Stevens (1989} mapped vegetation types in west Texas based on this same guide. This classification is based on vegetation existing at the time of the

14 TABLE 1. Descriptions of vegetation types (McMahan et al. 1984) for Texas Rolling Plains Rio Grande wild turkey study area. Commonly Associatiated Texas Parks and Wildlife Vegetation Type Plants and Characteristics Classification 1. Mesquite Brush mesquite, juniper, lotebush, Mesquite Brush tobosagrass, buffalograss. Mesquite-Lotebush Brush 2. Rough Break havard shinoak, juniper, Juniper-Mixed Brush mesquite. caprock escarpment. 3. Mesquite Shrub/Grassland mesquite, buffalograss. Mesquite Shrub/Grassland 4. Cultivated Crops 5. Mesquite-Juniper juniper, mesquite, hairy Mesquite-Juniper Shrub Brush/Shrub grama. Mesquite-Juniper Brush 6. Mesquite-Saltcedar saltcedar, mesquite, Mesquite-Saltcedar Brush ragweed, alkali sacaton. Brush/Woods ephemeral river drainages. 7. Badland shadscale. no classification little or no vegetation, highly erodible soil. 8. Mesquite-Hackberry mesquite, hackberry. Mesquite-Hackberry Brush canyon bottoms, creeks, Brush/Woods drainages. 9. Sandsage-Mesquite sandsage, mesquite, sand Sandsage-Mesquite Brush Brush dropseed. sandy upland 10. Mesquite-Cactus mesquite, pricklypear, cholla, no classification Brush tasajillo. 11. Mixed Brush skunkbush sumac, little leaf no classification sumac, catclaw acacia, feather dalea, javelina bush, hairy grama. gravelly, well drained soil.

15 survey and not on potential vegetation. The TPWD system employed satellite technology, ground-truthing procedures and computer classification analysis to categorize the vegetation types. Each vegetation type was named using the one or two dominant associated plants and the physiognomic designation(s), e.g. mesquite shrub/grassland. Not all of the Texas Parks and Wildlife Department's classifications for this area of west Texas occurred on the study area. Three additional vegetation types (badland, mesquite-cactus brush, mixed brush) were identified for this study because there were distinct areas that did not share the characteristics of the established types. Map Construction Aerial photographs, at a scale of 1 :40,000, acquired by the ASCS in 1980 combined with ground-truthing activities were used to delineate vegetation types. Vegetation types were then mapped on USGS 7.5' quadrangles (1 :24,000). Using a digitizer, the vegetation map was entered into ERDAS (ERDAS, Inc., Atlanta, Georgia), a raster based geographical information system. Pixels (grid cells) of 50 m x 50 m (1/4 ha) were coded for each of eleven vegetation types. The digitized border between vegetation types is the edgeline. A 100 m wide edge is two rows of pixels, one on each side of the edgeline. A 200m wide edge is four rows of pixels, two on each side of the edgeline. The universal edge represents the area between all the vegetation types; it is comprised of the specific edges. A specific edge is an area between two particular vegetation types. Every pixel in the study area was classified as to vegetation type, whether it fell on a 100 m wide edge or a 200 m wide edge or both. It should be noted that a pixel on a 100 m wide edge is, by definition, always on the 200 m wide edge. The reverse is not always true.

Use and Availability of Vegetation Types and Edges 16 To determine areas of use within the study area, UTM's of bird locations were entered into ERDAS and overlaid onto the existing vegetation type map. Each location fell into one pixel. Multiple locations per pixel were possible. The classification of a bird location was based on the characteristics of the pixel it occupied. In summary, each bird location was assigned to a particular vegetation type, to edge or nonedge and, if on an edge, the specific edge was identified. To estimate the expected number of locations within a given vegetation type or edge, the availability of that category was determined. The availability of a given vegetation type was the ratio of the number of pixels with that vegetation classification to the number of pixels in the study area. This provided data for goodness-of-fit tests. A bird location was assigned only one vegetation type. It was possible for a location to be on an edge when edge was defined as 200 m wide, but not on an edge when the edge was 100 m wide. Where more than two vegetation types came together, a bird location could be on more than one specific-edge (Figure 1 ). For example, if vegetation types A, B and C join together, a given pixel could be on the AlB edge and the B/C edge. The one bird location would be included both in the number of locations on the AlB edge and the number of locations on the B/C edge. In turn, that dual-edged pixel was included in the number of AlB edge pixels and again in the number of B/C edge pixels, when determining availability of each. For this reason, the amount of specific-edge pixels summed together is greater than the amount of universal edge pixels. It follows that the number of bird locations on each specific-edge, summed, could be greater than the number of locations on the universal edge.

- - 17 A 8 I II!!!;!" "",, w I ~ -:: ' I ~ ' ~ FIGURE 1. Example of a vegetation type map and bird locations (dots). Data summary: Vegetation Type A: 0 locations Vegetation Type B: 1 location Vegetation Type C: 31ocations Universal Edge : 3 locations Specific Edge AlB : 1 location Specific Edge AIC : 2 locations Specific Edge 8/C : 3 locations

18 Vegetation Sampling Vegetation measurements were taken at bird locations each year between 1 July and 30 September. In 1990, all roost locations were sampled (N=11) and in 1991, only the newly located, previously unsampled roosts (N=6), were measured for vegetative characteristics. All nests were sampled each year (N=18 in 1990; N=13 in 1991 ). In 1990, a sample of poult-rearing locations (N=18) was measured and consisted of locations from each of 6 poult-rearing hens. There were no poult-rearing locations collected in 1991, due to immediate predation of hens with poults. Therefore, no poult-rearing locations were sampled in 1991. A sample of general locations (non-nesting, non-poultrearing, non-roosting) were measured each year (N=18 in 1990; N=18 in 1991 ). Finally in 1991, five random locations in each of the 11 vegetation types were measured (N=55). A bird location or a random location served as the focal point for sampling. A circular area of 10m (30m at roost sites) was plotted at the location. In addition to the center point, random azimuths and distances established three more sample points in the plot. The variable distance approach to measuring horizontal foliar density (HFD) of 50% at three random orientations from each of the four sample points, determined the screening cover within the plot (MacArthur and MacArthur 1961 ). HFD was sighted from three different heights above the ground: 0.33 m, 0.66 m, and 1.0 m. There was one observer for all 36 HFD readings per plot, for all plots, both years. Litter depth was measured at each sample point (except at the nest itself). The line intercept method was used to determine percent canopy cover of herbs, shrubs and trees greater than one meter ( Mueller-Dumbois and Ellenberg 1974).

Analyses 19 Vegetation Type Selection The analyses of vegetation type and edge use and availability involve nominal scale data and are suited to enumeration statistical methods. The chisquare goodness-of-fit statistic is commonly used in the determination of vegetation selection (White and Garrot 1990). Both the chi-square statistic (X2) and the G-statistic (G) (twice the log-likelihood ratio) approximate the X2 distribution. However, the chi-square analysis has associated rules against small expected frequencies (Cochran 1954), designed to avoid the bias of rejecting a true null hypothesis with a probability greater than a. The data in this study yielded small expected frequencies, so the G-statistic was used. The Yate's correction for continuity is applied to the X2 statistic when there is only one degree of freedom (Zar 1984), this correction is an overcorrection for the G statistic. Williams (1976) suggested the a-correction be applied to the G statistic at all degrees of freedom to better approximate the X2 distribution. All goodness-of-fit tests in this study employed the G-statistic with the a-correction. As previously described, a hen's locations were not recorded close together in time (at least 24 hours between), to ensure independence, and locations with associated error polygons greater than two hectares were excluded from analyses. Locations were identified as either: nesting, poult-rearing, roosting, all others pre-dispersal, or all others post-dispersal. The result was very few bird locations in an individual's five files. Therefore, the data were pooled across the animals (White and Garrot 1990, Neu et al. 1974). It was desirable to know if the 1990 and 1991 data sets could be pooled for a given type of bird location. If one goodness-of-fit test was to be used, the samples from the two years would need to be homogeneous. Heterogeneity

analyses, using a G-statistic with a a-correction, performed in a fashion 20 analogous to the heterogeneity chi-squared test, determined if the birds used the 11 habitats differently and the edges differently in 1990 and 1991. A significant G value indicated that the birds were not using the classes within a given category in proportion to their availability. To identify if significant disagreement between observed and expected frequencies was concentrated in certain classes, the method of Neu et al. (1974), of constructing Bonferroni confidence intervals around the proportion of bird locations observed in a given class, was used. One difference was made however. In the Neu et al. (1974) confidence interval formula, a Z-value appears. This Z-value is to be selected based on the number of classes being investigated, and a desired a-level. If a Z-value is selected based on only the desired type I error rate, and the confidence interval formula is applied to multiple classes, the effect would be analogous to multiplet-testing, increasing the experiment-wise error rate to a value greater than the desired a-level. Neu et al. (1974) divides the type I error rate by the number of classes before applying the confidence interval formula to multiple classes. They thus avoid increasing the experiment-wise error rate to a value greater than the desired a-level. However, in a procedure analogous to Fisher's protected LSD test, the overall G-test can be applied at the desired a level first. Then confidence intervals are computed only if the G-value is significant. For these data, a G-test was performed first. If the computed G value was not significant, there was no further analysis. If the G-statistic was significant, confidence intervals were constructed with a=0.05 (Z=1.96), to determine selection and avoidance for each class. Experiment-wise error rate was controlled at a=0.05. These vegetation type use confidence intervals were compared to availability of the vegetation type in the study area to determine

selection and avoidance. Selection/avoidance could not be assessed if a 21 vegetation type was not used because a use confidence interval could not be constructed. Vegetation Vegetation was sampled in order to reveal any structural differences between the 11 vegetation types. The objective was to combine this information with vegetation type use results, possibly indicating similarities between selected vegetation types and differences between selected and avoided types. Mean separation was performed for the variables: canopy cover, litter depth, 50% HFD from 0.33 m, 0.66 m and 1.0 m above ground. Data were supplied by five random vegetation plots within each of the 11 vegetation types. The 11 vegetation types yielded 11 treatments. Mean separation was performed with a protected LSD - an overall F-test (a=0.05) and, following a significant F-test, pairwise t-tests (a=0.05) (SAS Institute Inc., Cary, North Carolina). When the variances of the experimental errors were not homogeneous among treatments, as determined by Levene's test on homogeneity of error variances, a Welch's F test was used in place of the standard F-test. Also, heterogeneous error variances required the use of unique error terms for each pairwise t-test, rather than a common one (BMDP Statistical Software, Inc., Los Angeles, California).

CHAPTER Ill RESULTS AND DISCUSSION Capture In December of 1989, this flock was comprised of approximately 50 hens and 25 gobblers. It was suspected that there would be few juveniles due to a devastating hailstorm the previous nesting season, after which broken turkey eggs were discovered and few poults were seen (Howell 1989). As predicted, no juveniles were trapped during 1990. Forty-seven adult hens were caught and instrumented with transmitters over the two years of the study. Forty-three of these were captured in 1990. Adult hen weight in 1990 averaged 4.2 kg (SD=0.4 kg, N=43), ranging from 3.5 to 5.1 kg. Four unmarked adult hens were trapped in 1991. Adult hen weight in 1991 averaged 4.3 kg (SD=0.3 kg, N=4), ranging from 3.8 to 4.5 kg. One adult male, 8.6 kg, was caught and released in 1990. In 1991, four adult males were caught and released, weighing an average of 7.2 kg (SD=1.4 kg, N=4) with a range of 6.1 to 9.3 kg. Six juvenile females trapped in 1991 had an average weight of 3.6 kg (SD=0.3, N=6), ranging from 3.1 kg to 4.0 kg. Three juvenile males in 1991 averaged a weight of 6.8 kg (SD=0.3, N=3), and ranged from 6.5 to 7.1 kg. Mycoplasma Culture and Serology All adult wild turkey hen sera (N=43) tested negative for antibodies to Mycoplasma gallisepticum (M.g.), M. synoviae (M.~.). and M. meleagridis (M.m.). However, other Mycoplasma spp. were isolated. The predominant isolate was 22

23 M. gallopavonis. After antiserum tom. gallopavonis was added to the primary isolation broth, its growth was inhibited, allowing the growth of M. cloacale and M. gallinaceum. One clutch of ten eggs that failed to hatch was submitted to Dr. Ley. Only five of ten eggs showed embryo development. These embryos were fully formed and failure to hatch was due to late embryo deaths. Embryonic fluid was cultured for mycoplasmas and no Mycoplasma spp. were isolated. Of the Mycoplasma spp. isolated from culture and serology, none have been shown to affect production, fertility or hatchability in wild turkeys. M. gallisepticum, the only Mycoplasma spp. known to be a pathogen for wild turkeys, was not found in the sera sample, trachea cultures or embryonic fluid cultures. The birds sampled in this study appeared relatively "clean," as compared to those of another recent survey. Fritz et al. (1992) were able to detect M.g., M.~. and M.m. in other flocks of Rio Grande wild turkeys occurring in the Northeast portion (Wheeler Co.) of the Texas panhandle. Fortunately, the ranges of these populations do not overlap so there is no immediate concern for the health of birds in the vicinity of Justiceburg. Mean Dispersal Dates The mean dispersal date in 1990 was 22 April (SD=15 days, N=23), ranging from 28 March to 7 May, while in 1991 the mean dispersal date was 27 March (SD=12 days, N=11 ), ranging from 11 March to 12 April. Hens dispersed earlier (P=0.05) in 1991 than they did the previous year, 1990. The earlier dispersal in 1991 concurs with the peak dispersal reported by Schmutz and Braun (1989) of Rio Grandes in Colorado during the third week of

March and with hens in west-central Texas that dispersed in late March 24 (Stevens 1989). Distances to Nests: Dispersal Distances Distances to first known nests in 1990 averaged 6.8 km (SD=5.1 km, N=14) and in 1991 averaged 8.7 km (SD=6.1 km, N=11) (Table 2). In 1990, all but one of the nesting birds (N=14) had dispersed and of these the mean distance to first known nest (dispersal distance) was 7.2 km (SD= 5.1 km, N=13). In 1991, seven of 11 nesting birds dispersed and their mean distance to first known nest was 12.6 km (SD=3.7 km, N=7). The maximum dispersal distance was 15.2 km in 1990 and 16.4 km in 1991. These maximum distances were similar to the mean of 14.3 km (SE=4.9 km, N=11) reported by Schmutz and Braun (1989) for adult hens of a Rio Grande population transplanted in Colorado. Stevens (1989) found that all adult Rio Grande hens sampled in west Texas dispersed less than 7 km, coinciding with the results of this study. If the birds of this study were to disperse longer distances, away from the caprock, they would encounter less diversity of vegetation types. It may be that in west Texas, there are only pockets of acceptable range, outside of which the birds do not venture. Mean Nest Initiation Dates The mean initiation dates were 17 May (SD=22 days, N=18) in 1990, ranging from 23 April to 26 June; and 8 May (SD=29.9 days, N=13) in 1991, ranging from 6 April to 13 July. These dates were not different statistically (P = 0.05).

TABLE 2. Mean distance to nest from winter roost for adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. 1990 6.8 km 6.1 km 7.2 km 6.3 km (SO= 5.1 km, N = 14) (SO= 4.7 km, N = 18) (SO= 5.1 km, N = 13) (SO= 4.8 km, N = 17) 1\) 01 Year All adult hens ( disperseda and nondispersed) All nests First known nest (includes known renests).!o~is~p~e~rs~e~da adult hens only All nests First known nest (includes known renests) 1991 8.7 km 7.6 km (SO = 6.1 km, N = 11) (SO= 6.1 km, N = 13) 12.6 km (SO= 3.7 km, N = 7) 12.6 km (SO = 3.7 km, N = 7) aoispersed hens are those that traveled out of radio contact from the winter roost.

26 Mean nest initiation was two to four weeks later than the median dates reported by Schmutz and Braun (1989) for Rio Grandes in Colorado and two to six weeks later than the range of dates reported by Ransom et al. (1987) for this species in south Texas. Beasom (1970) reported that egg-laying in Rio Grandes in south Texas began in March and Baker et al. (1980) found that Rio Grande eggs are already hatching by the end of May. These discrepancies may be due to the difficulty of distinguishing first nests from renests. Interestingly, although the birds in this study dispersed at different times each year, mean nest initiation date remained constant. Clutch Size Mean clutch size in 1990 was 11.2 eggs (S0=2.229 eggs, N=18), ranging from eight to 16 eggs; 1991 mean clutch size was 10.9 eggs (S0=2.362 eggs, N=13), ranging from four to 14 eggs. There was a negative relationship between nest initiation date and clutch size. In this study I found that nest initiation date was a significant predictor of clutch size in both 1990 (R2=0.368, adj.r2=0.329, P=0.008) and 1991 (R2=0.409, adj.r2=0.355, P=0.02). This is likely due to the smaller clutch sizes and later initiation dates of renests contained in the sample. Schmutz and Braun (1989) also found a negative relationship, determining that nest initiation date was a significant predictor of clutch size (R2=0.723, P=0.004). As is evident from the R2 values, the date of nest initiation did not explain as much of the variation in clutch size as it did for Schmutz and Braun. In order to successfully predict clutch size, other variables with predictive value should be added to the model. Since wild turkey nesting activities are dependent on fat reserves which are in turn dependent on the

27 energy budget before breeding (Porter et al. 1983), knowledge of the condition of a hen and the severity of the winter might aid in clutch size prediction. Nest Success Six of 18 nests hatched in 1990, indicating 33% apparent nest success, but translating to 28.7% Mayfield nest success (Table 3). In 1991, four of 13 nests hatched, yielding 30.8% apparent nest success and 34.8% Mayfield nest success. These fall between the Mayfield nest success calculated by Ransom et al. (1987) of 10.9% in 1983 and 47.8% in 1984 for Rio Grande turkeys in south Texas. Beasom (1973) suggested that heavy August and/or September rainfall recharges the soil moisture, favorably affecting new vegetative growth the following spring. August and September were the months in which he reported significant rainfall-turkey productivity correlation coefficients. During this study, rainfall was above the long-term average of 47.2 em - 77.5 em in 1990 and 65.0 em in 1991 (Figure 2). August and September rainfall prior to the 1990 and 1991 spring nesting seasons was also higher than the long-term average of 9.0 em - 15.0 em in 1989 and 18.3 em in 1990. If the correlation is constant across the Rio Grande turkey's range, then the nest success observed during the study may be high for Garza and Borden counties. The percent of nests destroyed by predators was similar in both years (Table 3). More hens abandoned nests in 1990 (17.8%) than in 1991 (none). Interestingly, there were more clutches that failed to hatch in 1991 (21.7%) than in 1990 (5.9 /o). A possible cause of unhatched clutches may have been the early dispersal in 1991, allowing hens to lose their fertility before nest initiation.

28 TABLE 3. Nest fate of adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991. Year Fate Nest successa,_p~re=d=a=tio=n..:... No hatch Hen abandonment 1990 28.7 47.6 5.9 17.8 1991 34.8 43.5 21.7 0.0 amayfield ( 1961 )

Nov Dec 1\) <D - E (,) -_, _, 20 18 16 14 12 < 10 LL z - < 8 a: 6 IZl 1990 1991 ---Average 4 2 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct MONTH FIGURE 2. Rainfall data at Texas Tech Experimental Ranch, Justiceburg, Texas, 1990 and 1991.

30 The birds tested negative form. gallisepticum, so unhatched clutches were not attributed to the presence of this bacteria in the population. Survival Rates Eight hens (18.6%) died within two weeks of being trapped in 1990. There were no trapping mortalities in 1991. The difference may be due to the physical stress of collecting blood samples the first year. Another source of bird stress may have been the drought of 1989 when the area received only 83% of its annual rainfall of 47.2 em. All other mortalities during the study were due to predation. The March-August Heisey-Fuller survival rates were 52.6% (N=35) for 1990 and 54.3% (N=19) for 1991. These are low compared to the two-year-average hen survival rate of 72.6% for Rio Grande turkeys in south Texas (Ransom et al. 1987). However, Glazener et al. (1990), using capture-recapture data with the program JOLLYAGE (Stokes 1984), reported annual survival rates, 1962-1972, for south Texas Rio Grande turkeys (both sexes) that averaged 55%. They noted however that female survival was higher than male survival. Heisey-Fuller monthly survival rates for adult hens during 1990 and 1991 declined in May (Figure 3). Of the seven that died in May of 1990 (N=26), only one was known to be incubating. Of the four that died in May of 1991 (N=16), none were known to be incubating but one hen was destroyed with her poults within 24 hours of leaving the nest. There was another decline in monthly survival in July of 1991 (N=12), when only one of three deaths was of a hen known to be incubating. Dispersal was nearly complete by May and none of the May deaths occurred within 10 days of a bird's dispersal. Therefore, dispersal and travel

ELl 1990 ~ 1991 (J.) _.. 100 80..J 60 < > a: > :::l en ';fl. 40 20 0 ' r ( ( r ~ ;&g ;o ] 1/ ( /'l ~-.,-~f '$1 I ( ( A -<-> r~-h~. [ ( ( Ji>m,.,,, r // l ~. I 1// 0 "' "i ' I I I I I I March April May June July August MONTH Figure 3. Heisey-Fuller (1985) March to August survival rates for adult Rio Grande wild turkey hens, Justiceburg, Texas, 1990 and 1991.