<%BANNER%>

Geospatial Analysis of Vegetative Characteristics Associated With Red-Cockaded Woodpecker Habitat in a Pine Flatwoods Ec...


PAGE 1

GEOSPATIAL ANALYSIS OF VEGETATIVE CHARACTERISTICS ASSOCIATED WITH RED-COCKADED WOODPECKER HABITAT IN A PINE FLATWOODS ECOSYSTEM By DOUGLAS O. SHIPLEY A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Douglas O. Shipley

PAGE 3

Research presented in this document is dedicated to the School of Forest Resources and Conservation at the University of Florida.

PAGE 4

ACKNOWLEDGMENTS I am especially grateful to Loukas G. Arvanitis, my major advisor, who has provided knowledge, opportunity, and support throughout my graduate studies. I would also like to thank my other committee members, Alan Long and Leonard Pearlstine, for providing useful comments on earlier versions of this document and input to improve my public presentation of thesis material. I would also like to thank the members of the Forest Information Systems Lab of the School of Forest Resources, University of Florida, for assistance with computer software and data analysis. The efforts of those who assisted in collection of field data for my research are greatly appreciated, especially those of Jonathan Huels and Brian Holmes. Their diligence in the field and reliability were integral to the completion of this work. Special thanks are due to Charles Marcus of the Florida Division of Forestry and the Goethe State Forest staff of Herb Heesch, Robert Cahal III, and Robin Boughton who have provided equipment and expertise throughout my research. My parents, Donna and Dennis Shipley, have given me opportunities and confidence throughout my graduate career, which I am most thankful for. My experience is not complete without the close relationship that I have with my family. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT.......................................................................................................................ix CHAPTER 1 INTRODUCTION...........................................................................................................1 Population Characteristics..............................................................................................2 Habitat Use and Forest Structure....................................................................................3 Foraging and Territory Range in Florida.................................................................5 Vegetative Associations with Red-cockaded Woodpeckers in Florida...................6 Goethe State Forest Population Characteristics..............................................................7 Study Objectives...........................................................................................................11 2 METHODOLOGY........................................................................................................12 Study Area....................................................................................................................12 Data Collection.............................................................................................................13 Forest Inventory.....................................................................................................13 Understory Vegetation Sampling...........................................................................14 Data Analysis Procedures.............................................................................................14 Baseline Statistics..................................................................................................14 Binary Logistic Regression....................................................................................17 Tests for Habitat Association...........................................................................19 Foraging Preference Analysis: Multivariable Models.....................................20 Nesting Habitat Model.....................................................................................22 3 RESULTS: COMPARISON OF HABITAT COMPONENTS.....................................23 Pine Characteristics Throughout Foraging Zones.........................................................23 Active and Inactive Cluster Habitat Composition Analysis.........................................25 Comparisons of Pine Basal Area...........................................................................25 Differences in Pine DBH.......................................................................................25 Differences in Basal Area of Large Pines..............................................................28 v

PAGE 6

Wiregrass Observations.........................................................................................28 Habitat Characteristics Within Zone A and Zone B.....................................................30 Comparisons of Pine Basal Area...........................................................................30 Differences in Pine DBH.......................................................................................30 Differences in Basal Area of Large Trees..............................................................32 Observed Wiregrass...............................................................................................33 Logistic Regression Model Estimates...........................................................................33 Variable Screening Results....................................................................................33 Forest-Wide Model Diagnostics............................................................................35 North and South Regions.......................................................................................38 Nesting Habitat.......................................................................................................41 4 DISCUSSION OF FORAGING ZONE ANALYSES...................................................43 Habitat Characteristics at the Goethe State Forest........................................................43 Quality of Forest Vegetation Within Clusters........................................................44 The 320 Acre Foraging Zone.................................................................................47 Application of a Binary Logistic Regression Model....................................................48 Suggestions For Future Research..................................................................................49 5 MANAGEMENT IMPLICATIONS.............................................................................52 LIST OF REFERENCES...................................................................................................54 BIOGRAPHICAL SKETCH.............................................................................................58 vi

PAGE 7

LIST OF TABLES Table page 1-1. Home range sizes for red-cockaded woodpecker colonies in north and central Florida (NA = not available)................................................................................................5 4-1. Pine characteristics in nest habitat and foraging zones at the Goethe State Forest...26 4-2. The average basal area of pines relative to longleaf or slash pine dominance within active and inactive clusters....................................................................................27 4-3. The average Diameter at Breast Height (DBH) of pines relative to longleaf or slash pine dominance within active and inactive clusters...............................................27 4-4. The average basal area of pines 12 in for samples dominated by either longleaf or slash pine within active and inactive clusters........................................................29 4-5. Test results for differences in mean basal area between zones A and B. T-tests were conducted at = 0.05............................................................................................31 4-6. Test results for differences in mean DBH between samples collected within zones A and B. T-tests were conducted at = 0.05...........................................................31 4-7. Test results for differences in basal area of pines 12 in within zones A and B. T-tests were conducted at = 0.05............................................................................34 4-8. Results from fitting the logistic regression equation separately to the data on habitat selection by red-cockaded woodpeckers at the Goethe State Forest in 2001........35 4-9. Results from fitting the logistic regression model to habitat data collected throughout 19 nesting clusters of the Goethe State Forest, Florida.......................37 4-10. Results from fitting the logistic regression model to habitat data collected in the northern region of the Goethe State Forest, Florida..............................................38 4-11. Results from fitting the logistic regression model to habitat data collected in the southern region of the Goethe State Forest, Florida..............................................40 4-12. Computed coefficients and scores for longleaf pine as a predictor of nesting habitat selection.................................................................................................................42 vii

PAGE 8

LIST OF FIGURES Figure page 1-1. Red-cockaded woodpecker cluster status in 2000, Goethe State Forest, Florida........8 1-2. Map of 19 red-cockaded woodpecker nesting sites during 2001. A 320 acre buffer surrounds each site...................................................................................................9 2-1. Location of a red-cockaded woodpecker nest cavity, and zones used to identify habitat characteristics within the maximum sampled foraging area......................16 3-1. Percentage of sampled longleaf and slash pine trees at various habitat scales..........24 viii

PAGE 9

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science GEOSPATIAL ANALYSIS OF VEGETATIVE CHARACTERISTICS ASSOCIATED WITH RED-COCKADED WOODPECKER HABITAT IN A PINE FLATWOODS ECOSYSTEM By Douglas O. Shipley May 2003 Chair: Loukas G. Arvanitis Major Department: Forest Resources and Conservation A geographic information system was used to characterize and model pine species composition, basal area, and diameter at breast height (DBH) of forest habitat occupied by the endangered red-cockaded woodpecker (Picoides borealis) in the Goethe State Forest, Florida. Samples were collected in three habitat zones: a) a 75 ac buffer zone surrounding 19 nest cavities; b) a 320 ac zone of active habitat (zone A); and c) 500 ac zones which constitute active and inactive management units (clusters). Samples collected beyond a 320 ac zone within the cluster boundary were designated as non-selected habitat (zone B). Basal area in nesting habitat averaged 35 ft2/ac of longleaf pine (Pinus palustris) and 22 ft2/ac of slash pine (Pinus elliottii). Longleaf-dominated habitat was significantly greater (p < 0.000) in active clusters (60 ft2/ha) compared to inactive clusters (49 ft2/ac). Estimated DBH did not differ significantly between longleaf and ix

PAGE 10

slash pine dominated habitat in active (p = 0.073) and inactive (p = 0.200) clusters, respectively. Binary logistic regression models were developed to analyze preference of forest characteristics within nesting habitat. Our forest-wide population model suggests that the probability of longleaf pine association with nest cavity habitat is 0.60. Two alternative models were developed to evaluate habitat associations across the population and within two subpopulations. The probability of association with active habitat for areas that are dominated by longleaf pine is 0.76, where the average DBH of longleaf pines is 13 in and wiregrass (Aristida beyrichiana) is present throughout the population. For north and south subpopulations, the same model yielded less significant results, most likely due to far fewer sample observations where both wiregrass and large longleaf were present. Findings of this study suggest that cluster recruitment and artificial cavity construction efforts must be focused on habitats dominated by contiguous longleaf pine with several large pine trees, and understory conditions associated with frequent burning. Potential cavity sites may be limited as the average age of site index trees is 55 years, which is not conducive to cavity excavation. The density of pines at the Goethe State Forest are within the guidelines of the U. S. Fish and Wildlife Service for red-cockaded woodpecker habitat restoration. The iterative system used to model habitat associations in this forest was very effective in characterizing forest data. However, given the variability of sample data, a relative large number of samples are needed for effective hypotheses testing. The modeling process used in this study may be considered as a template to identify habitat variables associated with other red-cockaded woodpecker populations in Florida. x

PAGE 11

CHAPTER 1 INTRODUCTION The red-cockaded woodpecker (Picoides borealis) and the associated old-growth longleaf pine (Pinus palustris) ecosystem once dominated the pine forests in the southeastern United States. Currently, less than five percent of the original 50 to 60 million acres of longleaf pine forest remains due to timber harvesting, fire suppression, and conversion to agricultural land or other uses ( Dennington and Farrar 1983 Outcalt and Sheffield 1996 ). In response to habitat loss and subsequent population decline, the red-cockaded woodpecker is listed as an endangered species, protected by the U.S. Endangered Species Act of 1973. The bird serves as an umbrella species for protection and restoration of longleaf pine habitat ( U.S. Fish and Wildlife Service 2000 ). Management for this species, mainly on federal and state holdings, involves restoring its native pine ecosystem, thus enhancing populations through construction of artificial cavities and woodpecker translocation. Data for this project were collected at Floridas Goethe State Forest, which contains two geographically isolated sub-populations of the red-cockaded woodpecker, which has adapted to habitat of marginal quality. Sampling was conducted in order to characterize current habitat conditions around 19 known red-cockaded woodpecker nesting sites. Results suggest preference for large longleaf pines in foraging areas. However, most of the Goethe State Forest is dominated by slash pine (Pinus elliottii). Biologists and land managers may use data from this study to understand forest conditions which are preferred by the woodpecker and identify areas that are suitable for 1

PAGE 12

2 artificial cavity construction and translocation of birds. Forest inventory data suggest that ground cover composition and basal area per acre of all pines > 4 in diameter at breast height (DBH) are within the ranges outlined by the Recovery Standard for Good Quality Foraging Habitat ( U.S. Fish and Wildlife Service 2000 ). However, the mosaic of cypress domes, found in the swamps of the Goethe State Forest, cover the largest percentage of habitat within clusters when compared to other red-cockaded woodpecker habitats in central Florida. This factor may increase the amount of quality foraging habitat needed as compared to sites with similar pine tree characteristics and species composition. The goal of the project was to establish baseline data for forest overstory and understory components found within known red-cockaded woodpecker habitat at the Goethe State Forest. This information was used to identify vegetative characteristics associated with existing red-cockaded woodpecker nesting sites and active habitat to support decision making and management for population expansion. Population Characteristics As territorial cooperative breeders, red-cockaded woodpecker groups usually consist of a breeding pair and one to several helper males, organized in clans, also referred to as groups ( Carter et al. 1995 Conner et al. 2001 ). Each group occupies an active cluster, defined as the defended habitat surrounding an aggregate of roost and surplus cavity trees ( Engstrom and Mikusinski 1998 Hovis and Labisky 1996 ). Groups of two to three woodpeckers are most common within clusters. Males that do not remain in their natal territory as helpers disperse in search of a new place to breed ( Conner et al. 2001 ). Isolation of these territories or breeding units may limit the ability of dispersing young females to find mates. As a non-migratory species, red-cockaded woodpecker populations are particularly susceptible to the confounding effects of forest fragmentation

PAGE 13

3 ( Conner and Rudolph 1991 Rudolph and Conner 1994 ). The limited number of potential cavity sites and the investment required for cavity construction limits development of functional metapopulations ( Rudolph and Conner 1994 U.S. Fish and Wildlife Service 2000 ). Habitat Use and Forest Structure The red-cockaded woodpeckers use of living pine trees for cavity construction is unusual. Most woodpeckers in North America prefer dead trees. Cavities can take from one to three years to complete. They are the most important resource for populations of red-cockaded woodpeckers as they compete for existing tree cavities rather than construct new ones ( Gaines et al. 1995 Walters et al. 1995 ). The birds create and maintain resin wells about one inch deep around the perimeter of the cavity opening to encourage resin flow which is effective in deterring predators, particularly rat snakes (Elaphe spp.) ( Ross et al. 1997 ). Mature pines are chosen for cavity construction because of their large diameter and likelihood of red heart fungus (Phellinus pini) ( Affeltranger 1971 Ligon 1971 ). Pines larger than 12 in DBH and 60 years of age are most often selected for cavity excavation. This may be attributed to a positive relationship between arthropod biomass and increased pine tree age ( Hooper 1996 ). Various studies indicate that when available, the red-cockaded woodpecker prefers older, larger trees for foraging ( Hopkins and Lynn 1971 ). The use of younger trees, less than 60 years old, is generally dependent on the availability of older ones ( Conner et al. 2001 ). Use of trees less than 60 years of age is often necessary as most of the habitat within the red-cockaded woodpeckers range has been harvested within the last 100 years, or disturbed mechanically, and is not typical of historic stand age classes and size.

PAGE 14

4 Red-cockaded woodpeckers cannot tolerate pine forests with a well developed midstory ( Conner et al. 2001 ). Increases in overstory of pine and hardwoods have been associated with colony abandonment ( Loeb et al. 1992 ). Management treatments should be applied to reduce midstory basal area below 25 ft2/ac in colonies as cluster abandonment drastically increases above this level. Successional advancement in hardwoods is primarily due to fire suppression in recent decades, which has also impacted pine reproduction rates. Dying pines are an important foraging substrate as woodpeckers consume many species of beetles and arthropods which are abundant in infested pines ( Conner et al. 2001 ). The red-cockaded woodpecker is an insectivorous species, primarily feeding on a variety of arthropods usually foraged from the bark of living pine trees ( Hanula and Franzreb 1998 Ligon 1971 ). The quality of the available forage may affect both the nesting success and density of woodpeckers ( Conner et al. 1999 ). The area required for foraging varies depending on the quality of the habitat. In north Floridas Apalachicola National Forest, Porter and Labisky ( 1986 ) reported a foraging preference for pines with a DBH >7.8 in within old-age stands ranging in age from 57 to 87 years. In addition to tree diameter and age, the suitability of foraging habitat is strongly correlated with stand density. Basal area of pines four inches or greater in diameter for longleaf systems should be between 40 and 60 ft2/ac ( U.S. Fish and Wildlife Service 2000 ). However, basal area of pine habitat in north Florida has been as dense as 70 ft2/ac in Floridas Apalachicola National Forest ( Hovis and Labisky 1996 ). Red-cockaded woodpeckers appear to be flexible in their ability to successfully forage in pines of various ages ( Azevedo et al. 2000 Hooper 1996 ).

PAGE 15

5 Foraging and Territory Range in Florida In central and southern Florida, pine forest habitat is relatively poor. Home range sizes vary depending on the amount and quality of available forage ( Hovis and Labisky 1996 ). At central Floridas Curtis H. Stanton Energy Center, DeLotelle et al. ( 1987 ) observed a mean home range and defended territory size of 361 ac and 287 ac respectively, over a two-year period. Cypress domes and bay heads accounted for approximately 8.6 percent of habitat within territories. In a study using radio telemetry, Nesbitt et al. ( 1978 ) determined an average foraging range of 172 ac for three red-cockaded woodpecker clans in central Floridas Marion County. Foraging habits of four clans (social groups) at the Apalachicola National Forest in north Florida averaged 319 ac ( Porter and Labisky 1986 ). The study was conducted over a period of one year. The study results are summarized in Table 1-1. Table 1-1. Home range sizes for red-cockaded woodpecker colonies in north and central Florida (NA = not available). Author Year Duration of Study Study Location Foraging Range Longleaf Pine Pond-slash Pine Cypress/bay-heads Nesbitt et al. 1978 128 hrs. Central Florida, Marion County 172 ac 38.2% 43.8% 6.5% DeLotelle et al. 1987 2 yrs. Curtis H. Stanton Energy Center 361 +/81 ac NA NA 8.6% Porter and Labisky 1986 1 yr. Apalachicola National Forest 319 +/31 ha 31% 35% 7% 23% (titi) In Florida, red-cockaded woodpecker habitat ranges differ greatly and are substantially larger than ranges found in other states. Average year-round home ranges were estimates as 205 ac in North Carolina ( Walters et al. 2000 ), 215 ac in South Carolina ( Hooper et al. 1982 ), and 198 ac in coastal Georgia ( Epting et al. 1995 ).

PAGE 16

6 Vegetative Associations with Red-cockaded Woodpeckers in Florida Overstory and understory vegetative associations specific to habitat selection have been explored for the species in central and north Florida habitats ( DeLotelle et al. 1987 Hovis and Labisky 1996 Hovis and Labisky 1985 Nesbitt et al. 1978 U.S. Fish and Wildlife Service 2000 ). In the Apalachicola National Forest, located in the Florida panhandle, Hovis and Labisky ( 1985 ) quantitatively evaluated habitat conditions found within two peripheral zones surrounding nest cavities. Basal area was found to be 46 ft2/ac in the selected habitat zone, and 65 ft2/ac in a zone, representative of outer foraging limits. The mean DBH in these zones was 8.2 in and 6.8 in respectively. Within in selected habitat, the overstory was characterized by 70 percent longleaf/slash pine flatwoods and baldcypress (Taxodium distichum) swamps/titi thickets (6 percent of the habitat). Habitat in the outer foraging limits was composed of 65 percent longleaf/slash pine flatwoods and 15 percent baldcypress swamps/titi thickets ( Hovis and Labisky 1985 ). Midstory plants consisted mainly of gallberry (Ilex glabra), St. Johnswort (Hypericum spp), longleaf pine, and saw palmetto (Serenoa repens). In central Florida, DeLotelle et al. ( 1987 ) sampled red-cockaded woodpecker territory and reported that 88.1 percent of the home range was pine flatwoods and 8.6 percent consisted of cypress domes and bay heads. The remaining 3.3 percent was wet prairie and open area. Qualitative estimates of habitat in the species southernmost range reveal a similar composition of pines and understory species. In Marion County Florida, Nesbitt et al. ( 1978 ) reported longleaf pine on the higher sites, with slash and pond pine (Pinus serotina) in lower, wetter sites with intermittent flatwood ponds bordered by bay and pondcypress (Taxodium ascendens).

PAGE 17

7 Goethe State Forest Population Characteristics The State of Florida purchased the Goethe Tract in 1992 as part of its Conservation and Recreation Lands (CARL) program. Management authority was given to the Florida Department of Agriculture and Consumer Services, Division of Forestry. Although red-cockaded woodpeckers were known to exist prior to state acquisition, the status of the population at the Goethe State Forest was unknown. A cooperative agreement with the Florida Fish and Wildlife Conservation Commission conducted a survey of the forest to determine the status of the resident red-cockaded woodpecker population. In 1994, the geographic location and characteristics of red-cockaded woodpecker cavities at the Goethe State Forest were identified, and cluster boundaries were developed using the circular scale technique ( Hovis 1996 ). Nesting data were used to determine cluster status. Active status was assigned to areas with a single active tree or a group of active cavity trees, some of which included an active tree occupied by a nesting pair. Areas with groups of cavity trees that were not used by the red-cockaded woodpecker were deemed inactive. In 1995, 26 clusters were active in the Goethe State Forest. They were divided geographically into north and south regions of the forest. Monitoring of cluster status was conducted on an annual basis by the Goethe State Forest staff, during spring and summer months. In 2000, 30 clusters were active (23 were nesting), 16 of which were located in the northern part of the forest. The remaining 14 clusters were in the southern region of the main tract. Figure 1-1 includes the distribution of the 30 clusters that were uninhabited, or inactive, in 2000. The distribution of nesting clusters in 2001 is included in Figure 1-2. In this study, the geographic location of these nineteen nesting cavities were used to identify selected habitat for various analyses.

PAGE 18

8 Figure 1-1. Red-cockaded woodpecker cluster status in 2000, Goethe State Forest, Florida.

PAGE 19

9 Figure 1-2. Map of 19 red-cockaded woodpecker nesting sites during 2001. A 320 acre buffer surrounds each site.

PAGE 20

10 Staff biologists at the Goethe State Forest monitor red-cockaded woodpecker cavities on an annual basis, usually during breeding season (May through June). A portable peeper camera unit or ladder assembly are used to observe fledgling status or other species within nesting cavities. The spatial location of all active and inactive cavity trees is recorded using the Global Positioning System (GPS). Cavity trees are marked with a single white band (approximately 8 in wide) around the stem at eye level. Within active clusters, a minimum of four viable cavities must be maintained at the Goethe State Forest. Artificial cavities are installed in living pine trees, if needed, as cavity trees may die or become occupied by other species such as red-bellied woodpeckers (Melanerpes carolinus), or pileated woodpeckers (Dryocopus pileatus). Inserts are only installed if site and pine tree characteristics are suitable. The minimum diameter at breast height of the selected pine trees must be 17 in to accommodate the size of the cavity insert. The diameter at the point of insertion is generally larger than 15 in This diameter permits the excavation of a hole 4 in wide, 10 in tall, 6 in deep to hold the cavity insert ( Allen 1991 ). The minimum installation height is 20 ft and the opening of the insert box generally faces a southwest direction or is pointed to active cavity trees within the cluster. If the crown of a potential cavity tree touches the crown of others, it is generally not considered ideal. Longleaf pines which are flat-topped are avoided to leave potential nest sites for natural excavation. The species of pines is not often considered by biologists at the Goethe State Forest, as it is common for the larger pines that grow near cypress ponds to be chosen for cavity construction. Cavity inserts may be installed near the border of cypress forests, if suitable pine tree selection is limited. In this situation, cavities are installed in a fashion so that openings do not face the cypress

PAGE 21

11 ponds. Cavity installation and cosmetic preparation is completed. The newly added cavity closely resembles a natural active cavity. Recruitment clusters consist of cavity inserts, usually a pair, installed within a half mile range of an existing nesting cluster. Site conditions are generally similar to those found in the neighboring active cluster. Study Objectives To support habitat restoration and provide current baseline data for management decisions, this research focuses on the following: a) Develop a geo-referenced forest and understory vegetative cover inventory ArcView database of active and inactive cluster sites. b) Test for differences in pine tree species and vegetative cover characteristics between selected and non-selected habitat. c) Evaluate the effectiveness of logistic regression as a tool for identifying forest variables that have a significant association with selected versus non-selected habitats.

PAGE 22

CHAPTER 2 METHODOLOGY Study Area Field data were collected within the main tract of the Goethe State Forest, located in the southeastern portion of Levy County, Florida (2922 N, 8237 W and 296 N 8232 W). The main Goethe tract is comprised of 49,295 ac and is managed by the Florida Division of Forestry for multiple-use purposes throughout scrub, sandhill, pine flatwoods, and dome swamp ecosystems. Activities offered to the public include: hunting and camping by permit, wildlife viewing, bicycling, hiking, and horseback riding. At the time of this study there were 60 known red-cockaded woodpecker clusters throughout the tract (Figure 1-1). The forest landscape is primarily pine flatwoods with a high proportion of intermixed baldcypress swamps and hardwoods. The latter are considered unsuitable for foraging. Hardwood features influence the area of good quality foraging habitat that can be used throughout the red-cockaded woodpeckers home range. This is apparent by the variation in estimates of foraging ranges in forests that contain hardwood substrate ( U.S. Fish and Wildlife Service 2000 ). The active red-cockaded woodpecker population is split into two isolated groups: the north and south regions of the forest (Figure 1-2). This separation is most likely related of a lack of suitable cavity trees and dense midstory vegetation which follows drain flats in the central part of the main tract. Unfortunately, the historical distribution of clusters at the Goethe State Forest is not known. 12

PAGE 23

13 Within the sampled clusters, pine overstory consists of slash pine, longleaf pine, and scarce loblolly pine (Pinus taeda). Understory vegetation is dominated by saw palmetto, gallberry (Ilex glabra), fetter-bush (Lyonia lucida), grasses, and forbs. Data Collection Field data were collected in 60 clusters, as identified by the Florida Division of Forestry to be sampled for standing tree data and understory vegetation composition. In this study, potential foraging habitat is defined as contiguous pine forest within each cluster. Active clusters are management units defined by the presence of cavity trees with flowing sap from resin-wells or the presence of a breeding pair. Inactive clusters are abandoned red-cockaded woodpecker cavities at the time of study. Cypress and hardwood areas were excluded from sampling. A combination of compass bearings and pacing was used to navigate between pre-determined sample points, each representative of ten acres of foraging habitat. Sample points were located systematically within each cluster on a 10-chain (660 ft) square grid. At each sample location, geographic coordinates were recorded with a handheld global positioning system (GPS) unit, used for a 20 second duration. Data were post processed with an accuracy of one to three meters. A forest inventory was conducted at each sample location and data were recorded with a ruggedized handheld computer (CMT PC5-L, Corvallis, OR) operating on Field Dog software (Two Dog Inc., Blacksburg, VA). Vegetative ground cover attributes were recorded in a field book. Data collection began in May 2000 and completed in August 2001. Forest Inventory Management officials at the Goethe State Forest identified 52 clusters that were considered priority for forest inventory. At each sample location within these clusters, a

PAGE 24

14 10-factor prism was used to select pine trees with probability proportional to tree DBH. Limiting-distance calculations were used for borderline trees. For each sampled pine tree, the species and DBH were recorded. Pines with a DBH < 4 in were not recorded. At each sample point, the nearest dominant or co-dominant of the most common pine species from the plot center was selected as the site index tree. Tree age, bark thickness, and total tree height were measured. Stem cores were extracted at breast-height to determine the age of the site index tree, and its five-year radial growth. A bark gauge was used to measure bark thickness at breast height. The height of each site index tree was measured with a precise vertex hypsometer. Understory Vegetation Sampling At each sample point, a square meter quadrat was used to assess ground cover composition. Palmetto, forbs, and woody-stem categories were each estimated as percent cover within the sample quadrat. Average height was estimated for each category in 3 ft intervals. Dominant species for forbs and woody stem categories were recorded in the field. Total percent cover for leaf litter, exposed mineral soil, and grasses (noting wiregrass (Aristida beyrichiana) if dominant) were estimated and recorded. Data Analysis Procedures Baseline Statistics An initial quantified summary of vegetative characteristics associated with red-cockaded woodpecker habitat as well as abandoned forage was developed to support habitat restoration and to provide current baseline statistics and data for management decisions. This process was carried out using traditional t-tests for differences in measurement averages for dominant pine basal area and DBH between active and

PAGE 25

15 inactive clusters. The classification of samples used in testing was based on the zone or cluster status in the analysis, as described below. Tests for differences in means were performed at two foraging zone scales. The 320 ac area (zone A) surrounding nesting cavities was used to represent preferred habitat, i.e. the expected area used by the red-cockaded woodpecker at the Goethe State Forest (Figure 2-1). Samples collected within a larger zone of 500 ac, represent the potential foraging area within each active and inactive cluster. The 500 ac foraging areas were classified as active if either nesting cavities or a single active cavity tree (trees with resin wells) was present within a cluster. All clusters that contain previously used red-cockaded woodpecker cavities but are not currently in use were considered inactive. Samples used to represent non-selected habitat were selected from zone B which is formed by a ring between the 320 ac zone (zone A) and the outer margin of the 500 ac zone (Figure 2-1). A geographic information system was used to classify samples as either preferred or non-selected habitat. Classified samples within the buffers as described above, were exported as a .dbf file for testing in Minitab and SPSS statistical analysis software packages. In addition, the center of inactive clusters was buffered by a 2110 ft radius. Samples located within this buffer may be representative of selected habitat, if cluster abandonment was recent. Therefore, samples beyond the 2110 ft buffer and within the 500 ac inactive cluster boundary were classified as non-selected habitat. A third zone was used to compare immediate nesting habitat characteristics to non-selected habitat.

PAGE 26

16 Figure 2-1. Location of a red-cockaded woodpecker nest cavity, and zones used to identify habitat characteristics within the maximum sampled foraging area.

PAGE 27

17 A 75 ac buffer was created around 19 nesting cavity locations. The size of this buffer includes 86 percent of active cavity trees associated with the nesting cavities. Data were managed using Microsoft Excel and tested for differences between measurement means using SPSS and MINITAB software packages. Interpretation of these results provided a basis for selecting variables to be used in logistic regression models. Binary Logistic Regression Cluster recruitment requires an in-depth knowledge of forest conditions that are suitable for artificial cavity construction within potential forest habitats. Identifying which resources are selected most often by the red-cockaded woodpecker provide important information about the nature of the species habitat preference at the Goethe State Forest. In previous studies, longleaf pine trees were used by red-cockaded woodpeckers disproportionately to their availability among other pine species ( Hovis and Labisky 1985 Nesbitt et al. 1978 ). The binary logistic regression procedure was used to explore the relationship between existing habitat and used habitat at the Goethe State Forest. Based on the abundance of slash pine at the Goethe State Forest, it is possible that the population of red-cockaded woodpeckers may be selective. However, this assumption does not hold true for all populations, as foraging preference was positively correlated with individual tree and stand age, independent of species availability ( Engstrom and Sanders 1997 Rudolph and Conner 1994 Zwicker and Walters 1999 ). In this study, binary logistic regression was used to estimate the probability and odds of habitat selection based on measured forest characteristics.

PAGE 28

18 The availability of forest resources is not generally uniform, and use may change as availability changes. Therefore, used resources should be compared with available (or unused) resources to reach a valid conclusion concerning resource selection ( Manly et al. 1993 ). In this study, sample points are classified according to the type of habitat they represent. These classifications represent the dependent, or use variable (Y). A resource probability function estimates the probability that a measured resource (X) in a particular binary category (Y) is used by the red-cockaded woodpecker. Binary logistic regression was chosen for several reasons. Data collected in this study do not include a quantifiable estimate of habitat use by the woodpecker. A general linear model is not suitable because the outcome variable (habitat use) was not measured on a continuous scale. The outcome variable is binary, as determined by the classification of samples based on the zone they are collected in from the cluster center or nesting cavity. Samples collected within the 75 ac zone, or zone A are used as indicators of selected resources. Information obtained outside of these zones, where the woodpecker is not known to forage, are used to define resources that are not preferred. Criteria that are related to resource selection vary depending on forest resource condition. However, factors such as pine species, age, midstory height, and understory species composition are generally associated with red-cockaded woodpecker habitat ( Hardesty et al. 1997 Loeb et al. 1992 Rudolph and Conner 1994 U.S. Fish and Wildlife Service 2000 ). Because of the dichotomous nature of the dependent variable and the combination of discrete and continuous predictors, the binomial distribution was used for regression analysis. For this type of analysis, the logit transformation of variables is often used because it is very flexible and an easily used function which lends itself to a biologically

PAGE 29

19 meaningful interpretation ( Hosmer and Lemeshow 1989 ). The logistic model produces an output in terms of probability, which is bounded by 0 and 1. The interpretation of the logistic regression coefficients incorporates the calculation of odds, or the likelihood of association, with an outcome. This is accomplished by transforming the probability to an odds to remove the upper bound. The lower bound is removed by using the logarithm of the odds. The result is set to a linear function of the explanatory variables. The probability (pi) of is calculated using the logit model: g g ikkiixxx ...2211 where ggieep1 the outcome variable, and k predictor(s) for k explanatory variables, and i = 1,n individuals. This equation has the desired property that will always be between 0 and 1 for any number that is substituted for the ip s' and the ( sx' Allison 1999 ). This probability is sometimes excluded from interpretation in favor of an odds ratio or the likelihood of a predictor being a member of an event. The odds ratio for each predictor can be solved by using the regression coefficient ( k ) of the predictor variable as the exponent of e, or the base of the natural logarithms ( Grimm and Yarnold 1995 ). Tests for Habitat Association Tests were used to determine if the values of predicted coefficients of habitat variables were different from zero. The null hypothesis assumes that the predictor is zero in the population, and if rejected, the alternative hypothesis that the coefficient differs from zero, is accepted. For models containing n predictors, a chi-square distribution with

PAGE 30

20 n degrees of freedom is used to obtain the probability for the likelihood statistic, G. Coefficient parameter(s) differ from zero if the probability of calculated G-statistic is less than .05 (the cutoff probability for the hypothesis test, = .05). The z test is another approach to determine if an estimated value of a coefficient is different from zero. The predictor coefficient is divided by its standard error to compute z, which is a measure of expected variability in the coefficient among samples. The cutoff probability for z tests in this study is .05. Confidence intervals for odd ratios were computed to estimate the range of likelihood of association for a predictor with 95 percent confidence. Interpretation of this interval provides a reliable estimate of increase in the odds of association with foraging habitat from one pine species (or other habitat component) to another. Foraging Preference Analysis: Multivariable Models Data used to indicate the outcome variable (Y) were selected from two categories: a) Samples within the 320 ac foraging area (zone A) represent preferred habitat (coded as 1) b) Samples collected in zone B were considered non-selected habitat (coded as 0). An initial screening of habitat components was conducted by fitting each potential variable to a resource probability function to eliminate habitat characteristics with a low probability of association with preferred habitat. This was done in order to develop a model with a minimum number of variables for field application. Four variables tested for significant association with selected habitat (above 95 percent probability) during initial testing across all nesting clusters: a) longleaf dominance, b) slash pine basal area, c) longleaf diameter at breast height, and d) longleaf basal area (Table 4-8). Based on the statistical significance of these results, the outcomes from previous studies, and their role

PAGE 31

21 in supporting the woodpeckers ecological niche, select predictors that suggested a strong association with habitat selection were then tested interactively. Two models, consisting of 612 samples each, were tested to explore associations of multiple habitat characteristics with habitat preference across the entire forest population (Models 1p and 2p). Model 1p includes one predictor (b), which involves the interaction between: a) sample points with an average diameter of pines 33 cm (coded 1, if true) and b) samples dominated by longleaf pine (coded 1, if true). A minimum diameter of 13 in was chosen because of cited preferences of the red-cockaded woodpeckers to forage on trees larger than 12 in when available ( 1 Engstrom and Sanders 1997 ), and because samples which contain trees of this average DBH are less prevalent throughout the Goethe State Forest. The second test for habitat preference include two predictors. The first coefficient (b) in Model 2p is the interaction of samples where a) the average DBH of longleaf pine samples was 33 cm (coded 1, if true) and b) samples where wiregrass was present in the understory (coded 1, if true). The interaction of these factors is expressed as a binary variable. The second coefficient of Model 2p is longleaf dominance, coded 1, if present at a sample and coded 0, if longleaf pine was not dominant. Tests for multicollinearity, using a correlation matrix, were performed for all interactive coefficients. The two models used in the population analysis were also used to test for differences in habitat preference between the geographically separated north and south cluster groups. The models in the north are termed Model 1N and Model 2N, and in models which tested variables in the south are termed Model 1S and Model 2S. 1

PAGE 32

22 Nesting Habitat Model A third model was employed to estimate the probability of resource selection for specific pine species within the foraging area that surrounds the immediate area of nesting sites. Model 3 involves 150 random samples taken from two regions of the forest. Samples within the 75 ac buffer zone surrounding 19 nesting sites, were used to represent preferred habitat and samples. Data collected within zone B (Figure 2-1) were representative of unused habitat. Seventy-five random samples were used from each zone in the logistic regression analysis to determine if the longleaf pine dominance, (b1=1 if the sample was dominated by longleaf pine, and 0 if not) could be used to predict selection of nesting sites. Output statistics for Model 3 denote the probability and likelihood of selection between longleaf pine dominance and a binary variable Y, where Y= 1 for samples collected within the 75 ac zone, and Y = 0 for samples collected in zone B. The Minitab procedure BINARY LOGISTIC REGRESSION was used to fit each model to its respective data.

PAGE 33

CHAPTER 3 RESULTS: COMPARISON OF HABITAT COMPONENTS Pine Characteristics Throughout Foraging Zones Within nesting habitats (75 ac zone), 60 percent of sampled trees were longleaf pine, while slash pine comprised the remaining 40 percent (Figure 3-1). Proportions of longleaf and slash pine within zone A were similar, but less pronounced. Slightly more than half (53 percent) of the sample trees in zone A were composed of longleaf pine. In zone B, representative of unused habitat, almost two-thirds (61 percent) of the tree samples were slash pine. Across all 52 clusters, slash pine represented 54 percent of the sampled trees in the potential red-cockaded woodpecker habitat (500 ac/cluster). Within these 500 ac clusters, the remainder of sampled trees were longleaf (approximately 46 percent) and loblolly pine (< .6 percent). However, pine species abundance differed based on cluster status. Active units were 51 percent longleaf pine while 62 percent of the trees in inactive units were slash pine (Figure 3-1). In active clusters, 29 of 572 samples were recorded as having no dominant (>50 percent of total samples) species. Of 679 samples in inactive clusters, 39 contained equal numbers of slash and longleaf pines. Nesting habitat within the 75 ac zone (n = 94), averaged 36 ft2/ac of longleaf pine basal area, substantially more than the average slash pine basal area (22 ft2/ha) (Table 4-1). Compared to the concentration of pines in zone B, the immediate area surrounding nesting sites averaged approximately 11 ft2/ac more longleaf basal area and 11 ft2/ac less slash pine basal area (Table 4-1). 23

PAGE 34

24 010203040506070NestingHabitatZone AZone BActiveInactiveForest-widePercent Longleaf Pine Cluster Status Slash Pine Figure 3-1. Percentage of sampled longleaf and slash pine trees at various habitat scales. The DBH of longleaf and slash pines within nesting habitat was similar to averages found in other habitat zones. The range of average longleaf and slash pine DBH estimates was 1.7 in and 5.3 in, respectively, across all foraging zones. The DBH of pines was not significantly different in zones A (n = 338) and B (n = 301) (Table 4-6). Zone A averaged 31 ft2/ac basal area of longleaf pine, which exceeds longleaf basal area in zone B by 6.9 ft2/ac. Slash pine basal area in zone B is greater than in zone A, by an average of 6.0 ft2/ac (Table 4-1). Pine tree DBH is similar for active (n = 572) and inactive (n = 679) clusters with a difference of 1.7 in for longleaf and 2.0 in for slash pines, respectively (Table 4-3). Longleaf pine basal area is 9.5 ft2/ac greater in active versus inactive clusters, on average. Between pine species in active clusters, longleaf is most prolific, averaging 5.6 ft2/ac more than slash pine, while slash pine basal area is higher on average in inactive clusters by 12.6 ft2/ac (Table 4-1).

PAGE 35

25 Active and Inactive Cluster Habitat Composition Analysis Comparisons of Pine Basal Area Basal area within active and inactive clusters averaged 62.7 ft2/ac and 59.8 ft2/ac, respectively (Table 4-2). Although there is no statistical difference between these overall basal area estimates, basal area is greater by 10.6 ft2/ac, (p = 0.0000) when longleaf is dominant in active versus inactive clusters (Table 4-2). Samples taken in habitat dominated by slash pine averaged a significantly higher basal area than longleaf pine in both active and inactive clusters. Active clusters are characterized by a mean basal area of 67.4 ft2/ac when slash pine is dominant compared to 59.6 ft2/ac when longleaf dominates the sample (p < 0.006). Inactive clusters are characterized by samples where slash pine was most often dominant. When compared to areas dominated by longleaf pine, inactive clusters contained 21 ft2/ac more basal area in areas dominated by slash pine. This difference was tested significant (p = 0.0000). Basal area of sample points dominated by slash pine was not significantly different (p = 0.39) between active (67.4 ft2/ac) and inactive (70 ft2/ac) clusters. The estimates for total basal area of longleaf dominated pine habitat in active and inactive clusters are within the United States Fish and Wildlife Service guidelines, but given the small average DBH (<13 in) of pines, habitat in these clusters is of marginal quality. Differences in Pine DBH Within active clusters, the average DBH per sample point is 11.9 in. This is significantly larger (p = 0.061, = 0.10) than the average DBH of samples in inactive clusters, which averaged 11.7 in (Table 4-3). When longleaf pines dominate the sample location, the average DBH is 11.7 in within inactive clusters and 12.1 in within active clusters.

PAGE 36

Table 4-1. Pine characteristics in nest habitat and foraging zones at the Goethe State Forest. 320 acre Zones )(x 500 acre Clusters )(x Pine Species 75 acre Zone )(x Zone A Zone B Cluster Status Active Inactive Longleaf Basal Area (ft2/ac) DBH (in) 36 (10)a 11.8 .2b 31.35 (10) 11.9 .09 24.3 (10) 12.0 .1 31.3 (10) 11.8 .07 21.7 (10) 12 ..09 Slash Basal Area (ft2/ac) DBH (in) 22.0 (10) 12.3 .3 27.0 (10) 11.9 .1 33.14 (10) 11.7 .11 25.7 (10) 11.8 .07 34.4 (10) 11.49 .07 a denotes sample range b denotes 1 standard error 26

PAGE 37

Table 4-2. The average basal area of pines relative to longleaf or slash pine dominance within active and inactive clusters. Mean (ft2/ac) Dominant Pine Species Active Clusters Inactive Clusters P Value Longleaf Basal Area 59.6 1.7a 49 .36 0.000 Slash Pine Basal Area 67.4 2.7 70 .41 0.390 P Value 0.006 0.000 Basal Area 62.7 1.3 59.8 1.2 0.098 a denotes 1 standard error Table 4-3. The average Diameter at Breast Height (DBH) of pines relative to longleaf or slash pine dominance within active and inactive clusters. 27 Mean (in) Dominant Pine Species Active Clusters Inactive Clusters P Value Longleaf DBH 12.1 .12a 11.7 .15 0.073 Slash DBH 11.8 .19 11.6 .13 0.200 P Value 0.350 0.480 DBH 11.9 .10 11.7 .10 0.061 a denotes 1 standard error

PAGE 38

28 These estimates are not significantly different at the 95 percent confidence level, but they are significantly different when tested at 10. (p = 0.073). The average DBH for samples dominated by slash pine is 11.8 in, and 11.6 in within active and inactive clusters, respectively. Similarly, these estimates are not different at the .95 confidence level (p = 0.20). Within active and inactive clusters, tests do not provide sufficient evidence for differences in mean DBH between samples where longleaf and slash pine are dominant (p = 0.350 and p = 0.480, respectively). It is clear that trees sampled within clusters, regardless of species, are very nearly the same DBH. Differences in Basal Area of Large Pines For samples where longleaf pine is dominant and the average DBH is greater than or equal to 12 in, the average basal area in active clusters is 4.3 ft2/ac greater than in inactive clusters (Table 4-4). Basal area for sample points dominated by slash pine averaging 12 in DBH or greater is 70.7 ft2/ac and 67.7 ft2/ac in active and inactive clusters, respectively. Differences between basal areas of sample points dominated by either large longleaf or slash pine in both active (p = 0.0000) and inactive (p = 0.0000) clusters were highly significant (Table 4-4). The apparent association between increased levels of basal area of larger pines in active clusters was tested further in this study through the use of logistic regression models. Wiregrass Observations In active clusters, wiregrass was the dominant grass in 60 percent of samples that contained coverage ( 5 percent) of any grass within the square meter sampling quadrat. For inactive clusters, wiregrass was observed as the dominant grass in 25.2 percent of grass samples. The groups of red-cockaded woodpeckers that live in active clusters may

PAGE 39

Table 4-4. The average basal area of pines 12 in for samples dominated by either longleaf or slash pine within active and inactive clusters. Mean (ft2/ac) Dominant Pine Species Active Clusters Inactive Clusters P Value Longleaf Basal Area 54.4 .2.1a 49.7 2.0 0.110 Slash Pine Basal Area 70.7 .2.9 67.7 2.6 0.570 P Value 0.000 0.000 a denotes 1 standard error 29

PAGE 40

30 be remnant of a population that was sustained by fire and where wiregrass was prolific throughout their territory. Todays inactive cluster status may be the result of fire exclusion, which, among other factors, could cause wiregrass presence to dwindle as new understory species colonized unburned red-cockaded woodpecker territories. Habitat Characteristics Within Zone A and Zone B Comparisons of Pine Basal Area The mean basal area per sample point in zone A was 60.2 ft2/ac. Samples in zone B averaged 59.6 ft2/ac basal area, which does not differ significantly from zone A. Samples located in selected habitat of zone A and dominated by longleaf pine had a basal area that was significantly less than samples dominated by slash pine (p < 0.000) (Table 4-5). Similarly, within non-selected habitat in zone B, the basal area of longleaf-dominated habitat is significantly less than the basal area in samples dominated by slash pine (p<0.000). When samples dominated by longleaf pine were compared between foraging zones, samples in zone A contained a significantly higher basal area (p = 0.046) than in zone B. The significantly greater levels of basal area in longleaf pine dominated habitat is similar in both zones and the larger, active clusters. Much like the cluster level analysis, a test for differences between slash pine basal area between habitat zones A and B did not provide evidence for statistical difference (p = 0.660) (Table 4-5). Differences in Pine DBH The average diameter for all pine samples within zone A was 12.0 in, which is significantly larger (p = 0.014) from the average diameter of 11.6 in within zone B. This significant difference was not found in a comparison of the average DBH of all pines sampled within active and inactive clusters (Table 4-3).

PAGE 41

Table 4-5. Test results for differences in mean basal area between zones A and B. T-tests were conducted at = 0.05. Mean (ft2/ac) Dominant Pine Species Zone A Zone B P Value Longleaf Basal Area 55.6 1.9a 49.0 2.1 0.046 Slash Pine Basal Area 66.9 2.6 70 2.2 0.660 P Value 0.000 0.000 Basal Area 60.2 1.3 59.6 1.2 0.810 a denotes 1 standard error Table 4-6. Test results for differences in mean DBH between samples collected within zones A and B. T-tests were conducted at = 0.05. 31 Mean (in) Dominant Pine Species Zone A Zone B P Value Longleaf DBH 12.2 .14a 11.5 .20 0.006 Slash DBH 11.8 .23 11.5 .18 0.330 P Value 0.130 0.950 DBH 12.0 .13 11.6 .13 0.014 a denotes 1 standard error

PAGE 42

32 Possible reasons for this characteristic change could be the removal of large, harvestable timber in densely stocked nesting habitat of inactive clusters. Perhaps overstocking close to roosting cavities was a contributing factor to cluster abandonment. Tests for differences in mean pine tree DBH between samples dominated by slash and longleaf pines did not differ significantly from each other in either habitat zone (Table 4-6). Samples dominated by longleaf pine had a significantly larger (p = 0.006) DBH in preferred habitat in zone A (12.2 in) versus zone B (11.5 in). The average DBH for samples dominated by slash pine within zone A was not dissimilar from the average pine tree DBH for samples dominated by slash pine in zone B (Table 4-6). Differences in Basal Area of Large Trees Samples dominated by longleaf pine with a minimum average diameter of 12 in contained significantly less basal area (49.7 ft2/ac, p < 0.000) than samples dominated by slash pine with a minimum diameter of 12 in (68.1 ft2/ac) in zone A. Similarly, the basal area of large slash pine was statistically higher (p < 0.000) than samples dominated by large longleaf pines within zone B. The increased levels of basal area associated with slash pine dominated samples is most likely due to the substantial area of pines which border cypress ponds. These areas are almost exclusively slash pine, and because of the moist soil conditions, these pines are often of large size given a young age. The skewing effect of the large slash pines on estimates of basal area and quality of habitat should be studied further through field observation of foraging habits and buffer analysis using a geographic information system. The basal area of samples dominated by large longleaf pine trees (12 in) within zone A did not significantly differ (p = 0.250) from trees of the same condition found within zone B. Statistical tests revealed no evidence that large

PAGE 43

33 slash pine basal area differs between zones A (selected) and B (non-selected forage) (Table 4-7). Observed Wiregrass Within the preferred habitat of zone A, understory vegetation samples suggest that wiregrass dominated 60.7 percent of samples (n = 84) when grass was present. Non-selected habitat in zone B included proportionally fewer wiregrass observations, as wiregrass was dominant in only 23.6 percent of samples which contained some type of grass (n = 110). Logistic Regression Model Estimates Variable Screening Results During the development of habitat association models, an interactive process was used to find a good subset of variables, i.e. a subset that includes only significant predictors that result in strong negative and positive predictive values. Table 4-8 shows the predictive coefficient value (bi ) and results for probability of habitat association for each of 16 variables. The iterative process eliminated certain variables, such as percent leaf litter cover. Although the probability of association with habitat for this variable is statistically significant (p = 0.0000), its predictive value is low as it is at or near zero. Other variables that were not significantly associated with active habitat were retained for model development. In this mode of variable testing, active habitat and large longleaf pines were not significantly associated (p = 0.9499). However, the t-test results on cluster data indicate that large longleaf pines were significantly larger in active clusters (Table 4-3). This factor was considered in conjunction with wiregrass presence as an interactive variable in Model 1, based on the assumption that the two variables are indicative of historic habitat conditions.

PAGE 44

Table 4-7. Test results for differences in basal area of pines 12 in within zones A and B. T-tests were conducted at = 0.05. Mean (ft2/ac) Dominant Pine Species Zone A Zone B P Value Longleaf Basal Area 49.7 2.4a 45.9 .2.5 0.250 Slash Pine Basal Area 68.1 3.2 70.0 3.4 0.750 P Value 0.000 0.000 a denotes 1 standard error 34

PAGE 45

35 Table 4-8. Results from fitting the logistic regression equation separately to the data on habitat selection by red-cockaded woodpeckers at the Goethe State Forest in 2001. Predictor Variable (X) (*=binary) Significance p Constant Term bx Leaf Litter Cover % .0000 -.0286 Longleaf Dominance* .0037 .4744 Slash Pine Basal Area .0069 -.0065 Longleaf DBH .0136 .0352 Longleaf Basal Area .0124 .0074 Palmetto % Cover .0524 .0054 Slash Pine DBH .0910 -.0226 Forb % Cover .1932 -.0161 Exposed Ground % .2734 .0031 Wiregrass Presence* .2958 .2470 Average DBH .3006 .0330 Basal Area .4399 -.0021 Av. DBH 13 .7371 .0575 Slash Pine DBH 13* .7984 -.0475 Wood Stem % Cover .8190 -.0009 Longleaf DBH 13* .9499 -.0112 Following the initial screening, two models were finalized, and were tested using field data. The potential for field application was a major consideration in the formulation of the models. Forest-Wide Model Diagnostics Data from 612 sample points were used in Models 1p and 2p to evaluate habitat characteristics for significant association with red-cockaded woodpecker habitat within all nesting clusters at the Goethe State Forest. For Model 1p, a correlation coefficient (CC) of 0.02 suggests no sign of multicollinearity between samples with an average DBH of all pines 13 in, and samples that were dominated by longleaf pine (Table 4-9). The evaluation of the z-score of b1 in Model 1p (z = 1.76, p = 0.078) suggests that it does not yield significant results at the 95 percent confidence level. The Models test statistic does not exceed the critical value of 3.841. Therefore, the null hypothesis, which states that the model variables are not associated with nesting habitat ( 01 b ), cannot be rejected.

PAGE 46

36 Even if this test level is relaxed, this model does not yield significant results at the 90 percent significance level. Model 2p includes two variables that contribute to the overall understanding of habitat selection. The predicted probability of habitat selection was solved by the following equation and subsequent transformation: g = -.0806 + .8091 (1) + .4330 (1) = 1.161 The predicted probability is e, or 76 percent. Thus the probability that areas dominated by large longleaf pines which contain wiregrass are associated with nesting habitat is 76 percent. Alternatively, the odds of association with nesting habitat is 1.16 times greater if those variables are present. Coefficients and related statistics are shown in Table 4-9. The G-statistic was used to test for at least one of the variables as statistically significant contributing factor in understanding habitat selection. The computed value of G is 12.5 with two degrees of freedom. The probability associated with this statistic is less than 0.002. This indicates that the alternative distribution is accepted and at least one of the coefficients {is not equal to zero. The z-score for longleaf dominance (b2) is positive and has a probability less than .05. This means that habitats dominated by longleaf pine tend to be selected as opposed to habitats where it is not dominant, when taking into account habitats where there is a combination of large longleaf pine (13 in) and wiregrass presence. Raising e to the power of b2 gives the odds ratio of 1.54. Model 2p suggests that when controlling b1 the odds of habitat association increase by 1.54 if longleaf is dominant. At the .10 level of confidence this coefficient does appear to have a significant association with habitat selection. )1/(161.1161.1e},21bb

PAGE 47

Table 4-9. Results from fitting the logistic regression model to habitat data collected throughout 19 nesting clusters of the Goethe State Forest, Florida. Forest-Wide Models Coefficient SE G p Z p CC Odds Ratio (eb) (eb) 95 % CI Lower Upper 1P. Constant (b0) Av. DBH 13 LL DOM (b1) .0987 .3868 .22 3.15 .076 1.76 .078 .02 1.47 .956 2.26 2P. Constant (b0) LL DBH 13 WP (b1) LL DOM (b2) -.0806 .8091 .4330 .42 .16 12.5 .002 1.91 2.62 .056 .009 .06 2.25 1.54 .98 1.12 5.14 2.13 37

PAGE 48

38 It can be said with greater than 90 percent certainty that the variables in Model 2 are strongly associated with nesting habitat. North and South Regions Within the northern part of the Goethe State Forest, 274 samples taken near nesting cavities were used to characterize the association between habitat selection and vegetative characteristics. Preferred and non-selected habitat were analyzed in the southern population with data collected at 338 sample locations. Table 4-10. Results from fitting the logistic regression model to habitat data collected in the northern region of the Goethe State Forest, Florida. Model Variable (North) Coefficient SE G p Z p CC Odds Ratio (eb) (eb) 95 % CI Lower Upper 1N. Constant (b0) Av. DBH 13 LL DOM (b1) .7329 .0938 .35 0.07 .788 .27 .789 .12 1.10 .55 2.18 2N. Constant (b0) LL DBH 13 WP (b1) LL DOM (b2) .6039 21 .1999 .26 5.29 .071 0.00 .76 .998 .009 .03 1.22 0 0.73 2.04 Model 1n tested samples where the average diameter of all pines was at least 13 in and longleaf pine was dominant. The correlation coefficient for b1 in Model 1n was 0.12, which indicates no strong relationship exists between the average diameter of pines 13 in and longleaf pine dominance. These variables, when tested interactively, were not significantly associated with habitat selection in the north (p = 0.788) (Table 4-10). In this case, as with the results from the population Model 1p, the null hypothesis is not rejected. By comparison, in the southern cluster group, the coefficient b1 in Model 1s was significantly associated (p = 0.027) with selected habitat (Table 4-11). In addition, there is no indication of autocorrelation between the factors in b1 observed in Model 1s (CC= 0.02). Data from this sample suggest that the odds of habitat association, given the presence of dominant longleaf pine and pines averaging 13 in, is estimated to be between 1.07 and 3.34 higher for the southern population at the 95 percent level of

PAGE 49

39 certainty. The likelihood ratio statistic for Model 1S (4.91) has a probability of 0.027, which suggests a positive association between the variables used in Model 1S and the preferred habitat. The probability of association for variables in Model 2s is: g = -.4844 + 1.02 (1) + .2981 (1) = .8337 and = .69 )1/(8337.8337.ee Within the Goethe State Forest, if habitat contains large longleaf with wiregrass, and is dominated by longleaf pine, the likelihood of habitat association is 69 percent greater than if these variables are not present. This assumes that other environmental factors such as pine density and understory height are within acceptable limits. On the basis of the z-score in Table 4-10 for Model 2N, this particular sample generated a b1 coefficient that does not differ from zero (p = 0.998). However, longleaf dominance (b2), is most likely different from zero since its associated z-score of 1.22 with a probability of 0.009 may suggest otherwise. An overall likelihood ratio test (G) was used to determine if either predictor is different from zero. For Model 2N, the G statistic is 5.29 with a probability of 0.071, suggesting that with almost 93 percent certainty that one of the model predictors differs from zero. There is no correlation of interactively tested variables in Model 2N (Table 4-10). This models poor performance was mostly due to highly variable data and smaller sample size when compared to the southern population.

PAGE 50

40 Table 4-11. Results from fitting the logistic regression model to habitat data collected in the southern region of the Goethe State Forest, Florida. Model Variable (South) Coefficient SE G p Z p CC Odds Ratio (eb) (eb) 95 % CI Lower Upper 1S. Constant (b0) Av. DBH 13 LL DOM (b1) -.3995 .6379 .29 4.91 .027 2.21 .027 .02 1.89 1.07 3.34 2S. Constant (b0) LL DBH 13 WP (b1) LL DOM (b2) -.4844 1.02 .2981 .45 .23 8.61 .014 2.28 1.30 .023 .192 .06 2.79 1.35 1.16 0.86 6.75 2.11 Measurements for b1 in Model 2S (the interaction between large longleaf and wiregrass presence), were not correlated (CC = 0.06). The associated z-score for b1 is 2.28 with a probability of 0.023, indicating that in the southern region, b1 denotes a significant relationship with habitat selection when considering b2, longleaf pine dominance (Table 4-11). A 95 percent confidence interval, placed on the odds ratio (2.79), suggests that if the variable b2 = 1 in the southern region, the level of habitat association with the red-cockaded woodpecker is between 1.16 and 6.75 times greater compared to habitat that does not include both large longleaf pines and wiregrass. The probability of habitat association, assuming habitat with all factors present as stated in the model is: g = -.4844 + 1.02(1) + .2981(1) g = .834 70.)1/(834.834.ee In this situation, the complete model was used to predict probability of association even though b2 was not statistically significant (p = 0.192). This is justified by the significant G statistic (8.61 with 2 degrees of freedom) (p = 0.014) obtained by evaluating the overall model.

PAGE 51

41 Nesting Habitat The computed odds ratio (e.8115) for Model 3 was 2.25 (Table 4-12). The predicted probability () of longleaf pine preference (b1 = 1) surrounding nesting habitat is: g g =-0.3947 + .8115*(1) g = .4168 )1/(4168.4168.ee =.60. The lower and upper confidence limits of the odds ratio are 1.17 and 4.33. One may argue with 95 percent confidence that the odds of habitat use for nest sites at the Goethe State Forest is between 1.17 and 4.33 times greater if longleaf pine is dominant rather than if longleaf pine is not the dominant species, assuming other environmental factors are held constant. The likelihood ratio statistic (G) was used to test whether the predictor variable is zero. The computed value for G = 6.045, which is considered large (p = 0.014), indicated that the variable is not equal to zero. The null hypothesis, which assumes the predictor coefficient is zero, was rejected. The probability of obtaining a z-score higher than 2.43 is less than 0.015. Thus, as with the likelihood ratio test, the result of the z-test indicates that a positive association exists between longleaf dominance and nest cavity habitat. Because the predicted probability is greater than 0.50, one may argue that maintenance of longleaf dominated habitat near nesting cavities is a priority at the Goethe State Forest. The iterative model building process used in this study considered several habitat components, but used a minimum number of variables for simplified field application. Most models performed well when applied to data for the forest population but did not when applied to the north and south sub-populations. This is an indication that for hypothesis tests to be accurate, logistic regression requires large samples.

PAGE 52

42 Table 4-12. Computed coefficients and scores for longleaf pine as a predictor of nesting habitat selection. Model 3: Variable Coefficient SE G p Z p Odds Ratio (eb) (eb) 95 % CI Lower Upper 3. Constant (b0) LL DOM (b1) -.3947 .8115 ..333 6.04 0.014 2.43 0.015 2.25 1.17 4.33

PAGE 53

CHAPTER 4 DISCUSSION OF FORAGING ZONE ANALYSES Habitat Characteristics at the Goethe State Forest In this study, a geographic information system was used to examine pine forest conditions within various spatial scales of active and inactive red-cockaded woodpecker habitat. A high proportion of longleaf pine samples (60 percent) was located within 1020 ft radius of nest cavities. By comparison, inactive clusters (half mile radius) contained approximately 45 percent longleaf pine. The indication that most longleaf-dominated habitat used for foraging is preferred over available slash pine-dominated habitat is consistent with previous studies that show a preference for, or selection of available longleaf pine over slash pine habitat in Florida ( DeLotelle et al. 1987 Nesbitt et al. 1978 Porter and Labisky 1986 ). The extent of red-cockaded woodpecker foraging zones varies between forests ( Nesbitt et al. 1978 Porter and Labisky 1986 Zwicker and Walters 1999 ). Results of this study indicate that habitat dominated by larger longleaf pines, compared to other available pines, is associated with habitat at both spatial scales used for analyses in this study (zones A and B, and active and inactive clusters). Samples in areas dominated by longleaf trees in foraging zone A average significantly greater DBH (P = 0.006) and basal area (P = 0.046) compared to samples in zone B. The difference in average DBH and basal area between zones, is .68 in and 5.7 ft2/ac, respectively. Longleaf pine basal area in zone A is 7.0 ft2/ac greater than in zone B. Compared to inactive clusters, longleaf basal area is 9.6 ft2/ac greater in active clusters. This indicates a positive trend towards 43

PAGE 54

44 longleaf preference given its availability. Although data on actual use or selection in longleaf dominated habitats was not collected, dominant pine overstory preference is apparent. The similarity of large longleaf pine trees in both active and inactive foraging zones and clusters suggest that although used habitat is dominated by longleaf overstory, preference may be influenced by factors other than DBH. Selection for cavity construction and foraging preference of longleaf pine is most likely related to a combination of factors such as age, good resin-producing abilities, and heartwood decay ( Conner et al. 2001 ). Quality of Forest Vegetation Within Clusters Based on guidelines set forth by the U.S. Fish and Wildlife Service, basal area of longleaf systems should range between 40 and 60 ft2/ac, while the basal area of shortleaf pine forests should range between 40 and 80 ft2/ac ( U.S. Fish and Wildlife Service 2000 ). This study suggests that in active clusters at the Goethe State Forest, the basal area of longleaf (60 ft2/ac) and slash (67 ft2/ac) pine systems fall within these guidelines, and cover a minimum of 60 percent of the area within a half mile of the cluster center. Samples contained significantly more basal area dominated by slash pine throughout active (P = 0.000) and inactive (P = 0.000) clusters. Both classes of foraging areas contained cypress swamps, bay heads, and hardwood drainages that are mostly bordered by slash pines and occasionally by loblolly trees. These slash pine are abundant, given the high perimeter to area ratio as swamps were often oval-shaped or in strips. The presence of cypress swamps decreased the total area of potential forage within this zone. The moist conditions which surround them may contribute to the increased size of slash pines which exclusively bordered these areas, possibly affecting habitat suitability.

PAGE 55

45 Diameter distribution of pines in both active and inactive clusters were similar in size. On the average, these pines, most likely, are not large enough to contain sufficient heartwood diameter (5-6 in) for cavity tree excavation ( Conner et al. 2001 ). It is known that longleaf pines are selectively chosen if they contain red heart fungus, which facilitates the excavation process ( Rudolph et al. 1995 ). Although the outer diameter of pines in active clusters at the Goethe State Forest may suit cavity excavation, the average age of sampled trees is 55.1 (SE=1.45 years), well below the suggested minimum age of 60 to 80 years. Age is most likely a limiting factor in nest cavity selection ( U.S. Fish and Wildlife Service 2000 ). Although the standing timber may be extensive enough and of suitable size, the lack of potential cavities is the dominant factor limiting the red-cockaded woodpeckers survival ( Hovis and Labisky 1996 ). Results of this study indicate that there is less basal area dominated by large longleaf pines throughout the 500 ac zone, and large trees (>12 in) in active clusters have a lower average basal area compared to longleaf of all sizes (>4 in). The average basal area of large slash pines (70 ft2/ac) was greater than the basal area of all slash pines > 4 in (67 ft2/ac) in active clusters. This relationship may be the result of forest management and the abundance of cypress ponds, rather than selection by the red-cockaded woodpecker. Previous management activities such as timber harvesting within clusters are most likely the cause for the decrease in large longleaf pine basal area. The abundance of large slash pine basal area is most likely due to the area surrounding cypress ponds where environmental conditions favor rapid growth of the species. The understory vegetation parallels the recommended guidelines definition of good quality foraging habitat since it is dense with fire tolerant and fire dependent

PAGE 56

46 species such as wiregrass, saw palmetto, gallberry and fetterbush. Wiregrass dominates approximately 60 percent of the samples where grasses were identified in active clusters. This is significantly larger then the 25 percent of grass samples dominated by wiregrass in inactive clusters and may be an indication that the present day red-cockaded woodpecker groups forage in areas which maintain site characteristics of earlier populations. However, it is difficult to determine the historical locations of red-cockaded woodpecker cavity sites and the surrounding understory vegetative conditions before state forest management practices began at the Goethe State Forest in the mid 1990s ( Hovis 1996 ). It is possible that the clusters where wiregrass is less prevalent may serve as an indication of other contributing factors which caused the abandonment of clusters, such as infrequent burning, or anthropogenic disturbance. The samples collected within the 500 ac foraging zone cover a larger area than has been generally documented for home ranges in central Florida. Among the largest documented home ranges are those in the Stanton Energy Center in Florida where habitat consists largely of pine savanna with sparse tree densities ( DeLotelle et al. 1987 ). Average foraging ranges in Floridas studied populations are larger than in other southern states. This may be related to the lower density, younger and smaller size-class of pines associated with habitat in the species southern margin. The amount and quality of available habitat are contributing factors to these estimates as home range sizes vary considerably within uniform habitat ( Hovis and Labisky 1996 ). Conner et al. ( 2001 ) estimate 230 to 383 ac for home range sizes in central Florida. Because of the relatively young age of potential cavity trees at the Goethe State Forest, an expected foraging range of 320 ac was chosen to further understand habitat preference in this case study.

PAGE 57

47 The 320 Acre Foraging Zone Previous studies have shown that cavity trees are usually located in stands with a low overstory basal area. Red cockaded woodpecker colonies studied at Apalachicola National Forest had a basal area that was considerably lower (46 ft2/ac) than in adjacent areas (65 ft2/ac) ( Hovis and Labisky 1985 ). Availability of longleaf and slash pine basal area in zones A and B was nearly equal in our study. Mean basal area in zone A (60.2 ft2/ac) is not significantly different from the surrounding zone B (59.6 ft2/ac). In this study, the density for both zones is relatively low (<78 ft2/ac) and the habitat is considered open grown. Use of longleaf pine for gum naval stores is evident by many remaining scarred trees which have been selected for nest cavity construction by the woodpecker and are likely infected with red-heart fungus. The basal area of available pine forage 4 in within zone A (60.2 ft2/ac) is within the recommended guidelines for good quality foraging habitat for longleaf systems ( U.S. Fish and Wildlife Service 2000 ). Nearly all samples of understory vegetation were less than 6 ft in total height. Groundcover at the Goethe State Forest is mostly contiguous, fire tolerant saw palmetto, and fire dependent herbs. Very few areas of canopy hardwoods were observed that were not within pre-identified areas. The mapped cypress and hardwood areas covered approximately 40 percent of available forage within 2110 ft of nest cavity trees. Compared to other studies that reported a cypress component within Florida habitat, the red-cockaded woodpecker population at the Goethe State Forest has adapted to habitat containing the largest area of cypress forest proportional to foraging zone size. The large percentage of cypress may indirectly support the species foraging needs as large slash pines (>17 in) occupy the perimeter where soil conditions are moist. Further analysis of the relationship between the red-cockaded woodpeckers

PAGE 58

48 use of slash pines bordering cypress swamps, the cypress swamp size, perimeter area, and frequency of occurrence may give insight to the impact of landscape features on habitat requirements. Application of a Binary Logistic Regression Model In this case study, binary logistic regression was used to explore the potential of one single-variable explanatory model and two multi-variable explanatory models to characterize and quantify association between habitat characteristics and habitat preference. The potential of different resource attributes were tested for likelihood and probability of association with areas preferred by red-cockaded woodpeckers at the Goethe State Forest. However, this study was not an exhaustive attempt on all aspects of the logistic regression. Likelihood estimates indicate forest habitat dominated by longleaf pine is positively associated with nesting habitat. The iterative model building strategy used in this study was employed to design a simplified model for field application. The combination of forward model building through tests on individual variables and the selection of variables based on previous studies, allowed us to find the best subset of a large set of predictors. However, given that wiregrass is less abundant than saw palmetto, the interaction of saw palmetto and longleaf pine at the Goethe State Forest most likely would have produced significant results when tested for association with preferred forage. Because wiregrass is historically associated with the species and is easily identifiable in the field, it was incorporated into Model 2. When longleaf is not dominant (b2 = 0), the likelihood of association for variables in Model 2P is higher (67 percent) if large longleaf and wiregrass are present (b1 = 1), compared to the likelihood if the combination is not present (b1 = 0) (58 percent). The higher odds ratio of b1 (2.25) compared to b2 (1.54),

PAGE 59

49 where estimates greater than one indicate an association with nesting habitat compared to non-selected habitat, is also an indication that habitat preference is dependent on a combination of forest attributes. Given the homogeneous nature of pine forest throughout the sampling area, and prominent wiregrass in active habitat, it appeared that tests would report similar results at different scales. However, the same models which indicate habitat preference at the population level did not produce similar results when tested at smaller scales. The results from models tested on north and south populations are indications that for hypothesis tests to be accurate, logistic regression requires large samples. Our results are due to the low number of samples that contained both wiregrass and large longleaf pines. Results may also be related to changes in the density of red-cockaded woodpeckers by forest region or due to the variation and availability of resource units, which can change the selection strategies and selection function. A separate resource selection function for several independent replications, with larger sample sizes, would be necessary to accurately depict resource preference within north and south regions. Suggestions For Future Research The use of a geographic information system to manage and interpret data collected throughout the Goethe State Forest brings a series of new ways to: a) Characterize cypress pond distribution, density, and area within habitat. b) Associate new flight and foraging preference data with existing forest stand characteristics and understory vegetation records. c) Relate the effects and intensity of silvicultural practices to the health of red-cockaded woodpecker.

PAGE 60

50 The database created for this study can be readily updated with new data collected by the Goethe State Forest staff as well as future researchers. This would facilitate spatial and temporal analyses, and consolidate records for assessment of management practices. Investigation of the distribution of cypress ponds within red-cockaded woodpecker habitat could provide insight as to whether or not it is a contributing factor to the area or amount of available forage needed by the species at the Goethe State Forest. Analysis of pond characteristics such as area, density, and perimeter-to-area ratio may be used to interpret the influence of cypress ponds on the size and distribution of clusters. This type of analysis is dependent on existing and new information such as the spatial distribution of foraging habits. This study used an estimated foraging radius based on published data collected at nearby red-cockaded woodpecker sites in central Florida to interpret associations between expected foraging area and vegetative characteristics. New information on the actual flight patterns and distance traveled by the red-cockaded woodpecker at the Goethe State Forest would be critical for an analysis of foraging area in relation to cypress pond characteristics. This information could be displayed showing the geographic distribution of foraging habits in relation to existing data from this study such as pine species, basal area, and tree diameter throughout foraging areas. New information such as cypress pond characteristics, stand age, fire events, timber harvests, and recreational activities could be used as factors tested to impact the distance or direction of foraging efforts by the red-cockaded woodpecker at the Goethe State Forest.

PAGE 61

51 Existing forest inventory data, integrated with data on silvicultural practices and red-cockaded woodpecker recovery efforts would support a comprehensive analysis of the changes in habitat preference over time. This meaningful data should be incorporated into the decision-making process and may be used as a process model designed to enhance existing red-cockaded woodpecker populations. These suggestions, as well as social issues associated with red-cockaded woodpecker restoration, illustrate the need for new information and provide opportunities for future research. The use of dynamic models and statistical analyses will continue to support the decision-making process. Development of new and specific restoration efforts will benefit greatly from continued cooperative research between the Florida Division of Forestry and the University system.

PAGE 62

CHAPTER 5 MANAGEMENT IMPLICATIONS Plans for restoration of red-cockaded woodpecker populations are mainly developed using guidelines designed by the United States Fish and Wildlife Service. Current management efforts at the Goethe State Forest are primarily supported by those guidelines. However, to prioritize new habitat restoration efforts, management objectives should consider the following: a) The significant and strong association found between nesting cavities and longleaf pines > 13 in DBH within close range (1020 ft). b) Introduction of prescribed fire or mechanical treatment to reduce dense saw palmetto coverage. c) A minimum intensive management unit of 225 ac surrounding nest cavities. Logistic regression models developed in this study clearly indicate the preference of longleaf pines > 13 in DBH within foraging areas surrounding nest cavities. In most management units where the red-cockaded woodpecker occurs, forest management strategies should continue to enhance the availability of these characteristics in frequently burned longleaf pine stands. Although various uneven-age timber harvesting techniques within management units can be used to retain some foraging value and provide financial gain, persistence of old trees should be emphasized. Regardless of species, trees greater than 100 years old should be omitted from timber sale, as they would leave red-cockaded woodpeckers without high quality forage, and potential excavation sites to use while stands mature. 52

PAGE 63

53 Vegetation in the understory of sampled clusters is dominated by saw palmetto. The leaves of this plant contain chemical properties that, when ignited, could produce extremely intense fires. Given the small average diameter of pines found within foraging clusters, it is important to reduce the amount of saw palmetto through the use of controlled burning or mechanical treatment. Adjacency to sensitive areas and weather conditions may prohibit burning in some areas. An alternative solution to remove excessive saw palmetto or vegetation which may disturb the red-cockaded woodpeckers foraging habits, may be a series of treatments which include a combination of chemical treatment to large hardwoods and mechanical removal of large, dense saw palmetto. In an optimal situation, the forest understory should be diverse with forbs and grasses whose composition is dense enough to sustain a periodic controlled fire. Although the process may be time and cost intensive, immediate efforts to reduce understory height below 12 ft should focus on areas where vegetation is near or exceeds this level in nesting clusters. This effort would sustain the existing populations and prepare for less costly management alternatives such as translocation efforts to enhance the existing population. Translocation and artificial cavity construction efforts used to create new colony sites should be targeted in longleaf dominated habitat adjacent to existing active clusters. These areas should contain a minimum of 225 ac of suitable pine forage with little or no hardwood midstory. This management unit size is based on test results for association of longleaf pine across three potential habitat zones, published research on red-cockaded woodpecker populations in central Florida, and efficiency of use in evaluation of field data for suitable habitat. A comparison between future estimates of forest characteristics and baseline data from this study may be used to evaluate management alternatives.

PAGE 64

LIST OF REFERENCES Affeltranger, C. 1971. The red-heart disease of southern pines. Pp. 96-99 in R. L. Thompson, ed. The Ecology and Management of the Red-Cockaded Woodpecker, Proceedings of a Symposium at Okefenokee National Wildlife Refuge. Folkston, GA. Allen, D. H. 1991. An insert technique for constructing artificial Red-cockaded Woodpecker cavities. U.S. Forest Service General Technical Report. SE-73. 19pp. Allison, P. D. 1999. Logistic regression: Using the SAS system: Theory and application. SAS Institute. Cary, NC. 288pp. Azevedo, J. C. M., S. B. Jack, R. N. Coulson, and D. F. Wunneburger. 2000. Functional heterogeneity of forest landscapes and the distribution and abundance of the red-cockaded woodpecker. Forest Ecology and Management. 127:271-283. Carter, J. H., J. R. Walters and P. M. Dixon. 1995. Red-cockaded Woodpeckers in the North Carolina Sandhills: A 12 Year Population Study. Pp. 248-258 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and management. Center for Applied Studies in Forestry, Stephen F. Austin State University. Nacogdoches, TX. Conner, R. N., D. C. Rudolph, and J. R. Walters. 2001. The Red-cockaded Woodpecker: surviving in a fire maintained ecosystem. University of Texas Press, Austin, TX. 363pp. Conner, R. N., and D. C. Rudolph. 1991. Forest habitat loss, fragmentation, and Red-cockaded Woodpecker populations. Wilson Bulletin. 103:446-457. Conner, R. N., D. C. Rudolph, R. Schaefer, D. Saenz, and C.E. Shackelford. 1999. Relationships among the red-cockaded woodpecker group density, nestling provisioning rates, and habitat. Wilson Bulletin. 111(4):494-498. DeLotelle, R. S., R. J. Epting, and J. R. Newman. 1987. Habitat use and territory characteristics of red-cockaded woodpeckers in central Florida. Wilson Bulletin. 99(2):202-217. 54

PAGE 65

55 Dennington, R. W., and R. M. Farrar. 1983. Longleaf pine management. United States Department of Agriculture, Forest Service Southern Region. Forestry Report R8-FR3. Pp. 17. Engstrom, T. R., and G. Mikusinski. 1998. Ecological neighborhoods in red-cockaded woodpecker populations. The Auk. 115:473-478. Engstrom, T. R., and F. J. Sanders. 1997. Red-cockaded woodpecker foraging ecology in an old growth longleaf pine forest. Wilson Bulletin. 109(2):203-217. Epting, R. J., R. S. DeLotelle, and T. Beaty. 1995. Red-cockaded woodpecker territory and habitat use in Georgia and Florida. Pp. 59-67 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and management. Center for Applied Studies in Forestry, Stephen F. Austin State University. Nacogdoches, TX. Gaines, G. D., K. E. Franzreb, D. H. Allen, K. S. Laves, and W. L. Jarvis. 1995. Red-cockaded woodpecker on the Savannah River Site: A management success story. Pp 81-85 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and management. Center for Applied Studies in Forestry, Stephen F. Austin State University. Nacogdoches, TX. Grimm, L. G., and P. R. Yarnold. 1995. Reading and understanding multivariate statistics. American Psychological Association, Washington D.C. 373pp. Hanula, J. L., and K. Franzreb. 1998. Source, distribution and abundance of macroarthropods on the bark of longleaf pine: Potential prey of the red-cockaded woodpecker. Forest Ecology and Management. 102:89-102. Hardesty, J. L., K. E. Gault, and H. F. Percival. 1997. Ecological correlates of red-cockaded woodpecker foraging preference, habitat use and home range size in northwest Florida (Eglin AFB). Final Report, Research Work Order 99. The Nature Conservancy, Gainesville, FL. Hooper, R. G. 1996. Arthropod biomass in winter and the age of longleaf pines. Forest Ecology and Management. 82:115-131. Hooper, R. G., L. J. Niles, R. F. Harlow, and G. W. Wood. 1982. Home ranges of Red-cockaded woodpeckers in coastal South Carolina. The Auk. 99:675-682. Hopkins, M. L., and T. E. Lynn. 1971. Some characteristics of Red-cockaded woodpecker cavity trees and management implications in South Carolina. Pp 140-169 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and management. Center for Applied Studies in Forestry, Stephen F. Austin State University. Nacogdoches, TX.

PAGE 66

56 Hosmer, D. W. Jr., and S. Lemeshow. 1989. Applied logistic regression. Wiley series in probability and mathematical statistics. Applied probability and statistics. Wiley, New York. 307pp. Hovis, J. A. 1996. Status and management of the red-cockaded woodpecker on Goethe State Forest, Florida. Southeast Association Fish and Wildlife Agencies Annual Conference Proceedings. 50:254-263. Hovis, J. A., and R. F. Labisky. 1996. Red-cockaded woodpecker. Pp. 81-102 in J. A. Rodgers, Jr., H. W. Kale II, and H. T. Smith, eds. Rare and endangered biota of Florida: Volume V. Birds. University Press of Florida, Gainesville, FL. Hovis, J. A., and R. F. Labisky. 1985. Vegetative associations of red-cockaded woodpecker colonies in Florida. Wildlife Society Bulletin. 13:307-314. Ligon, D. J. 1971. Some factors influencing numbers of the red-cockaded woodpecker. Pp. 30-43 in R.L. Thompson ed. The ecology and management of the Red-cockaded Woodpecker. Proceedings of a Symposium at Okefenokee National Wildlife Refuge. Folkston, GA. Loeb, S. C., W. D. Pepper, and A. T. Doyle. 1992. Habitat characteristics of active and abandoned red-cockaded woodpecker colonies. Southern Journal of Applied Forestry. 16(3):120-125. Manly, B. F., L. L. McDonald, and D. L. Thomas. 1993. Resource selection by animals. Chapman and Hall. Great Britain. 177pp. Nesbitt, S. A., D. T. Gilbert, and D. B. Barbour. 1978. Red-cockaded woodpecker fall movements in a Florida flatwoods community. The Auk. 95:145-151. Outcalt, K. W., and R. M. Scheffield. 1996. The longleaf pine forest: trends and current conditions. United Stated Department of Agriculture, Southern Research Station. Resource Bulletin SRS-9. Asheville, NC. Pp. 23. Porter, M. L., and R. F. Labisky. 1986. Home range and foraging habitat of red-cockaded woodpeckers in northern Florida. Journal of Wildlife Management. 50(2):239-247. Ross, W. G., D. L. Kulhavy, and R. N. Conner. 1997. Stand conditions and tree characteristics affect quality of longleaf pine for red-cockaded woodpecker cavity trees. Forest Ecology and Management. 91:145-154. Rudolph, D. G. and R. C. Conner. 1994. Forest fragmentation and red-cockaded woodpecker population: an analysis at intermediate scale. Journal Field Ornithology. 65:365-375.

PAGE 67

57 Thomlinson, J. R. 1996. Predicting status change in red-cockaded woodpecker cavity tree clusters. Journal of Wildlife Management. 60:350-354. U.S. Fish and Wildlife Service. 2000. Technical/agency draft revised recovery plan for the red-cockaded woodpecker. U.S. Fish and Wildlife Service, Atlanta, GA. 229pp. Walters, J. R., S. J. Daniels, J. H. Carter, III, P. D. Doerr, K. Brust, and J. M. Mitchell. 2000. Foraging habitat resources, preferences and fitness of red-cockaded woodpeckers in the North Carolina Sandhills. Fort Bragg Project Final Report. Virginia Polytechnic Institute and State University, Blacksburg, VA, and North Carolina State University, Raleigh, NC. Walters, J. R., P. P. Robinson, W. Starnes and J. Goodson. 1995. The relative effectiveness of artificial cavity starts and artificial cavities in inducing the formation of new groups of Red-Cockaded Woodpeckers. Pp. 367-369 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and management. Center for Applied Studies in Forestry, Stephen F. Austin State University. Nacogdoches, TX. Zwicker, S. M. and J. R. Walters. 1999. Selection of pines for foraging by red-cockaded woodpeckers. Journal of Wildlife Management. 63(3):843-852.

PAGE 68

BIOGRAPHICAL SKETCH Douglas Owen Shipley was born in Hinsdale, Illinois, on 17 August 1976. In 1998, he received a B.S. in environmental resource management from Virginia Polytechnic Institute and State University. After graduation, he worked as a forester at the Seminole State Forest for the Florida Division of Forestry. In December 2002, he completed requirements for the Master of Science degree, at the University of Florida. 58


Permanent Link: http://ufdc.ufl.edu/UFE0000789/00001

Material Information

Title: Geospatial Analysis of Vegetative Characteristics Associated With Red-Cockaded Woodpecker Habitat in a Pine Flatwoods Ecosystem
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000789:00001

Permanent Link: http://ufdc.ufl.edu/UFE0000789/00001

Material Information

Title: Geospatial Analysis of Vegetative Characteristics Associated With Red-Cockaded Woodpecker Habitat in a Pine Flatwoods Ecosystem
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000789:00001


This item has the following downloads:


Full Text











GEOSPATIAL ANALYSIS OF VEGETATIVE CHARACTERISTICS ASSOCIATED
WITH RED-COCKADED WOODPECKER HABITAT IN A PINE FLATWOODS
ECOSYSTEM
















By

DOUGLAS O. SHIPLEY


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2003




























Copyright 2003

by

Douglas O. Shipley




























Research presented in this document is dedicated to the School of Forest Resources and
Conservation at the University of Florida.















ACKNOWLEDGMENTS

I am especially grateful to Loukas G. Arvanitis, my major advisor, who has

provided knowledge, opportunity, and support throughout my graduate studies. I would

also like to thank my other committee members, Alan Long and Leonard Pearlstine, for

providing useful comments on earlier versions of this document and input to improve my

public presentation of thesis material. I would also like to thank the members of the

Forest Information Systems Lab of the School of Forest Resources, University of Florida,

for assistance with computer software and data analysis.

The efforts of those who assisted in collection of field data for my research are

greatly appreciated, especially those of Jonathan Huels and Brian Holmes. Their

diligence in the field and reliability were integral to the completion of this work.

Special thanks are due to Charles Marcus of the Florida Division of Forestry and

the Goethe State Forest staff of Herb Heesch, Robert Cahal III, and Robin Boughton who

have provided equipment and expertise throughout my research. My parents, Donna and

Dennis Shipley, have given me opportunities and confidence throughout my graduate

career, which I am most thankful for. My experience is not complete without the close

relationship that I have with my family.
















TABLE OF CONTENTS
page

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

LIST OF TA BLE S ............... .............. ................ ......... .............. .. vii

LIST OF FIGURES .................. ............. ......... .. ................... viii

ABSTRACT .............. .......................................... ix

CHAPTER

1 IN TR OD U CTION .................. ............................ ............. .............. .

P opu nation C h aracteristics .............................................................................................. 2
H habitat U se and Forest Structure .................................................................................... 3
Foraging and Territory Range in Florida ............................................................ 5
Vegetative Associations with Red-cockaded Woodpeckers in Florida ................. 6
Goethe State Forest Population Characteristics............... ............................................. 7
Study O objectives ..................................................................... .......... 11

2 M E T H O D O L O G Y ............................................................................. .....................12

S tu d y A re a .............................................................................. 12
D ata Collection ......................................... 13
Forest Inventory .... .................................... .............. 13
U nderstory V egetation Sam pling ................................................................. 14
Data Analysis Procedures ........................................ 14
B baseline Statistics ................................................................... 14
Binary Logistic Regression ................................. ........................... .... 17
Tests for Habitat Association............................................. 19
Foraging Preference Analysis: Multivariable Models .............. ............... 20
N testing H habitat M odel .............................................................. .............22

3 RESULTS: COMPARISON OF HABITAT COMPONENTS .............. ...............23

Pine Characteristics Throughout Foraging Zones................................... ............. 23
Active and Inactive Cluster Habitat Composition Analysis ...................................... 25
Com prisons of Pine Basal A rea ................................................................ 25
D differences in Pine D BH ................................................. ....... .. 25
Differences in Basal Area of Large Pines ............................................................. 28


v









W iregrass O b servation s ............................................................... ....... ....... 2 8
Habitat Characteristics Within Zone A and Zone B .............................................. 30
Com prisons of Pine B asal A rea ........................................ ....................... 30
D differences in P ine D B H ........................................................... ...................... 30
Differences in Basal Area of Large Trees................................................... 32
Observed W iregrass ............ .......... ........ .................... .................. 33
L ogistic R egression M odel E stim ates................................................. ... ................. 33
V ariable Screening R results .............. ........................................................... 33
Forest-W ide M odel D iagnostics ........................................ ......... .............. 35
N orth and South R egions ......................................................... .............. 38
N estin g H ab itat .............................................................................. 4 1

4 DISCUSSION OF FORAGING ZONE ANALYSES...............................................43

Habitat Characteristics at the Goethe State Forest............................................ 43
Quality of Forest Vegetation W within Clusters................................ .................... 44
T he 320 A cre F oraging Z one ............ ........................................................... ... 47
Application of a Binary Logistic Regression Model .............................................. 48
Suggestions For Future Research........................................ .................................... 49

5 MANAGEMENT IMPLICATIONS ........................................ ........................ 52

LIST OF REFEREN CES ............................................................ ................... 54

BIOGRAPH ICAL SKETCH ..................................................... 58
















LIST OF TABLES


Table page

1-1. Home range sizes for red-cockaded woodpecker colonies in north and central Florida
(NA = not available). ................................... .. ....... ................. 5

4-1. Pine characteristics in nest habitat and foraging zones at the Goethe State Forest...26

4-2. The average basal area of pines relative to longleaf or slash pine dominance within
active and inactive clusters. .............................................................................. 27

4-3. The average Diameter at Breast Height (DBH) of pines relative to longleaf or slash
pine dominance within active and inactive clusters ......... ..............................27

4-4. The average basal area of pines > 12 in for samples dominated by either longleaf or
slash pine within active and inactive clusters. ........................................... ........... 29

4-5. Test results for differences in mean basal area between zones A and B. T-tests were
conducted at = 0.05. ......................... ...... ............................ .......... .. .... 1

4-6. Test results for differences in mean DBH between samples collected within zones A
and B. T-tests were conducted at a = 0.05. .................................. .................31

4-7. Test results for differences in basal area of pines > 12 in within zones A and B. T-
tests w ere conducted at = 0.05....................................... ......................... 34

4-8. Results from fitting the logistic regression equation separately to the data on habitat
selection by red-cockaded woodpeckers at the Goethe State Forest in 2001........35

4-9. Results from fitting the logistic regression model to habitat data collected
throughout 19 nesting clusters of the Goethe State Forest, Florida..................37

4-10. Results from fitting the logistic regression model to habitat data collected in the
northern region of the Goethe State Forest, Florida. ...........................................38

4-11. Results from fitting the logistic regression model to habitat data collected in the
southern region of the Goethe State Forest, Florida. ...........................................40

4-12. Computed coefficients and scores for longleaf pine as a predictor of nesting habitat
selection. ...........................................................................42
















LIST OF FIGURES


Figure page

1-1. Red-cockaded woodpecker cluster status in 2000, Goethe State Forest, Florida........8

1-2. Map of 19 red-cockaded woodpecker nesting sites during 2001. A 320 acre buffer
surrounds each site. ................... .. ....... .. ..... ..... ..... ............ .. ..

2-1. Location of a red-cockaded woodpecker nest cavity, and zones used to identify
habitat characteristics within the maximum sampled foraging area....................16

3-1. Percentage of sampled longleaf and slash pine trees at various habitat scales..........24















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

GEOSPATIAL ANALYSIS OF VEGETATIVE CHARACTERISTICS ASSOCIATED
WITH RED-COCKADED WOODPECKER HABITAT IN A PINE FLATWOODS
ECOSYSTEM

By

Douglas O. Shipley

May 2003


Chair: Loukas G. Arvanitis
Major Department: Forest Resources and Conservation

A geographic information system was used to characterize and model pine species

composition, basal area, and diameter at breast height (DBH) of forest habitat occupied

by the endangered red-cockaded woodpecker (Picoides borealis) in the Goethe State

Forest, Florida. Samples were collected in three habitat zones: a) a 75 ac buffer zone

surrounding 19 nest cavities; b) a 320 ac zone of active habitat (zone A); and c) 500 ac

zones which constitute active and inactive management units (clusters). Samples

collected beyond a 320 ac zone within the cluster boundary were designated as non-

selected habitat (zone B). Basal area in nesting habitat averaged 35 ft2/ac of longleaf pine

(Pinuspalustris) and 22 ft2/ac of slash pine (Pinus elliottii). Longleaf-dominated habitat

was significantly greater (p < 0.000) in active clusters (60 ft2/ha) compared to inactive

clusters (49 ft2/ac). Estimated DBH did not differ significantly between longleaf and









slash pine dominated habitat in active (p = 0.073) and inactive (p = 0.200) clusters,

respectively.

Binary logistic regression models were developed to analyze preference of forest

characteristics within nesting habitat. Our forest-wide population model suggests that the

probability of longleaf pine association with nest cavity habitat is 0.60. Two alternative

models were developed to evaluate habitat associations across the population and within

two subpopulations. The probability of association with active habitat for areas that are

dominated by longleaf pine is 0.76, where the average DBH of longleaf pines is > 13 in

and wiregrass (Aristida beyrichiana) is present throughout the population. For north and

south subpopulations, the same model yielded less significant results, most likely due to

far fewer sample observations where both wiregrass and large longleaf were present.

Findings of this study suggest that cluster recruitment and artificial cavity

construction efforts must be focused on habitats dominated by contiguous longleaf pine

with several large pine trees, and understory conditions associated with frequent burning.

Potential cavity sites may be limited as the average age of site index trees is 55 years,

which is not conducive to cavity excavation. The density of pines at the Goethe State

Forest are within the guidelines of the U. S. Fish and Wildlife Service for red-cockaded

woodpecker habitat restoration. The iterative system used to model habitat associations

in this forest was very effective in characterizing forest data. However, given the

variability of sample data, a relative large number of samples are needed for effective

hypotheses testing. The modeling process used in this study may be considered as a

template to identify habitat variables associated with other red-cockaded woodpecker

populations in Florida.














CHAPTER 1
INTRODUCTION

The red-cockaded woodpecker (Picoides borealis) and the associated old-growth

longleaf pine (Pinuspalustris) ecosystem once dominated the pine forests in the

southeastern United States. Currently, less than five percent of the original 50 to 60

million acres of longleaf pine forest remains due to timber harvesting, fire suppression,

and conversion to agricultural land or other uses (Dennington and Farrar 1983, Outcalt

and Sheffield 1996). In response to habitat loss and subsequent population decline, the

red-cockaded woodpecker is listed as an endangered species, protected by the U.S.

Endangered Species Act of 1973. The bird serves as an "umbrella species" for protection

and restoration of longleaf pine habitat (U.S. Fish and Wildlife Service 2000).

Management for this species, mainly on federal and state holdings, involves restoring its

native pine ecosystem, thus enhancing populations through construction of artificial

cavities and woodpecker translocation.

Data for this project were collected at Florida's Goethe State Forest, which

contains two geographically isolated sub-populations of the red-cockaded woodpecker,

which has adapted to habitat of marginal quality. Sampling was conducted in order to

characterize current habitat conditions around 19 known red-cockaded woodpecker

nesting sites. Results suggest preference for large longleaf pines in foraging areas.

However, most of the Goethe State Forest is dominated by slash pine (Pinus elliottii).

Biologists and land managers may use data from this study to understand forest

conditions which are preferred by the woodpecker and identify areas that are suitable for









artificial cavity construction and translocation of birds. Forest inventory data suggest that

ground cover composition and basal area per acre of all pines > 4 in diameter at breast

height (DBH) are within the ranges outlined by the Recovery Standard for Good Quality

Foraging Habitat (U.S. Fish and Wildlife Service 2000). However, the mosaic of cypress

domes, found in the swamps of the Goethe State Forest, cover the largest percentage of

habitat within clusters when compared to other red-cockaded woodpecker habitats in

central Florida. This factor may increase the amount of quality foraging habitat needed

as compared to sites with similar pine tree characteristics and species composition.

The goal of the project was to establish baseline data for forest overstory and

understory components found within known red-cockaded woodpecker habitat at the

Goethe State Forest. This information was used to identify vegetative characteristics

associated with existing red-cockaded woodpecker nesting sites and active habitat to

support decision making and management for population expansion.

Population Characteristics

As territorial cooperative breeders, red-cockaded woodpecker groups usually

consist of a breeding pair and one to several helper males, organized in clans, also

referred to as groups (Carter et al. 1995, Conner et al. 2001). Each group occupies an

active cluster, defined as the defended habitat surrounding an aggregate of roost and

surplus cavity trees (Engstrom and Mikusinski 1998, Hovis and Labisky 1996). Groups

of two to three woodpeckers are most common within clusters. Males that do not remain

in their natal territory as helpers disperse in search of a new place to breed (Conner et al.

2001). Isolation of these territories or breeding units may limit the ability of dispersing

young females to find mates. As a non-migratory species, red-cockaded woodpecker

populations are particularly susceptible to the confounding effects of forest fragmentation









(Conner and Rudolph 1991, Rudolph and Conner 1994). The limited number of potential

cavity sites and the investment required for cavity construction limits development of

functional metapopulations (Rudolph and Conner 1994, U.S. Fish and Wildlife Service

2000).

Habitat Use and Forest Structure

The red-cockaded woodpecker's use of living pine trees for cavity construction is

unusual. Most woodpeckers in North America prefer dead trees. Cavities can take from

one to three years to complete. They are the most important resource for populations of

red-cockaded woodpeckers as they compete for existing tree cavities rather than construct

new ones (Gaines et al. 1995, Walters et al. 1995). The birds create and maintain resin

wells about one inch deep around the perimeter of the cavity opening to encourage resin

flow which is effective in deterring predators, particularly rat snakes (Elaphe spp.) (Ross

et al. 1997). Mature pines are chosen for cavity construction because of their large

diameter and likelihood of red heart fungus (Phellinuspini) (Affeltranger 1971, Lion

1971). Pines larger than 12 in DBH and 60 years of age are most often selected for

cavity excavation. This may be attributed to a positive relationship between arthropod

biomass and increased pine tree age (Hooper 1996). Various studies indicate that when

available, the red-cockaded woodpecker prefers older, larger trees for foraging (Hopkins

and Lynn 1971). The use of younger trees, less than 60 years old, is generally dependent

on the availability of older ones (Conner et al. 2001). Use of trees less than 60 years of

age is often necessary as most of the habitat within the red-cockaded woodpeckers' range

has been harvested within the last 100 years, or disturbed mechanically, and is not typical

of historic stand age classes and size.









Red-cockaded woodpeckers cannot tolerate pine forests with a well developed

midstory (Conner et al. 2001). Increases in overstory of pine and hardwoods have been

associated with colony abandonment (Loeb et al. 1992). Management treatments should

be applied to reduce midstory basal area below 25 ft2/ac in colonies as cluster

abandonment drastically increases above this level. Successional advancement in

hardwoods is primarily due to fire suppression in recent decades, which has also

impacted pine reproduction rates. Dying pines are an important foraging substrate as

woodpeckers consume many species of beetles and arthropods which are abundant in

infested pines (Conner et al. 2001).

The red-cockaded woodpecker is an insectivorous species, primarily feeding on a

variety of arthropods usually foraged from the bark of living pine trees (Hanula and

Franzreb 1998, Ligon 1971). The quality of the available forage may affect both the

nesting success and density of woodpeckers (Conner et al. 1999). The area required for

foraging varies depending on the quality of the habitat. In north Florida's Apalachicola

National Forest, Porter and Labisky (1986) reported a foraging preference for pines with

a DBH >7.8 in within old-age stands ranging in age from 57 to 87 years. In addition to

tree diameter and age, the suitability of foraging habitat is strongly correlated with stand

density. Basal area of pines four inches or greater in diameter for longleaf systems

should be between 40 and 60 ft2/ac (U.S. Fish and Wildlife Service 2000). However,

basal area of pine habitat in north Florida has been as dense as 70 ft2/ac in Florida's

Apalachicola National Forest (Hovis and Labisky 1996). Red-cockaded woodpeckers

appear to be flexible in their ability to successfully forage in pines of various ages

(Azevedo et al. 2000, Hooper 1996).









Foraging and Territory Range in Florida

In central and southern Florida, pine forest habitat is relatively poor. Home range

sizes vary depending on the amount and quality of available forage (Hovis and Labisky

1996). At central Florida's Curtis H. Stanton Energy Center, DeLotelle et al. (1987)

observed a mean home range and defended territory size of 361 ac and 287 ac

respectively, over a two-year period. Cypress domes and bay heads accounted for

approximately 8.6 percent of habitat within territories. In a study using radio telemetry,

Nesbitt et al. (1978) determined an average foraging range of 172 ac for three red-

cockaded woodpecker clans in central Florida's Marion County. Foraging habits of four

clans (social groups) at the Apalachicola National Forest in north Florida averaged 319 ac

(Porter and Labisky 1986). The study was conducted over a period of one year. The

study results are summarized in Table 1-1.

Table 1-1. Home range sizes for red-cockaded woodpecker colonies in north and central
Florida (NA = not available).

Author Year Duration of Study Location Foraging Range Longleaf Pond-slash Cypress/bay-
Study Pine Pine heads
Nesbitt et al. 1978 128 hrs. Central Florida, 172 ac 38.2% 43.8% 6.5%
Marion County
DeLotelle et al. 1987 2 yrs. Curtis H. Stanton 361 +/- 81 ac NA NA 8.6%
Energy Center
Porter and 1986 1yr. Apalachicola 319 +/- 31 ha 31% 35% 7%
Labisky National Forest 23% (titi)


In Florida, red-cockaded woodpecker habitat ranges differ greatly and are

substantially larger than ranges found in other states. Average year-round home ranges

were estimates as 205 ac in North Carolina (Walters et al. 2000), 215 ac in South

Carolina (Hooper et al. 1982), and 198 ac in coastal Georgia (Epting et al. 1995).









Vegetative Associations with Red-cockaded Woodpeckers in Florida

Overstory and understory vegetative associations specific to habitat selection have

been explored for the species in central and north Florida habitats (DeLotelle et al. 1987,

Hovis and Labisky 1996, Hovis and Labisky 1985, Nesbitt et al. 1978, U.S. Fish and

Wildlife Service 2000). In the Apalachicola National Forest, located in the Florida

panhandle, Hovis and Labisky (1985) quantitatively evaluated habitat conditions found

within two peripheral "zones" surrounding nest cavities. Basal area was found to be 46

ft2/ac in the "selected habitat" zone, and 65 ft2/ac in a zone, representative of outer

foraging limits. The mean DBH in these zones was 8.2 in and 6.8 in respectively.

Within in selected habitat, the overstory was characterized by 70 percent longleaf/slash

pine flatwoods and baldcypress (Taxodium distichum) swamps/titi thickets (6 percent of

the habitat). Habitat in the outer foraging limits was composed of 65 percent

longleaf/slash pine flatwoods and 15 percent baldcypress swamps/titi thickets (Hovis and

Labisky 1985). Midstory plants consisted mainly of gallberry (Ilex glabra), St.

Johnswort (Hypericum spp), longleaf pine, and saw palmetto (Serenoa repens).

In central Florida, DeLotelle et al. (1987) sampled red-cockaded woodpecker

territory and reported that 88.1 percent of the home range was pine flatwoods and 8.6

percent consisted of cypress domes and bay heads. The remaining 3.3 percent was wet

prairie and open area. Qualitative estimates of habitat in the species' southernmost range

reveal a similar composition of pines and understory species. In Marion County Florida,

Nesbitt et al. (1978) reported longleaf pine on the higher sites, with slash and pond pine

(Pinus serotina) in lower, wetter sites with intermittent flatwood ponds bordered by bay

and pondcypress (Taxodium ascendens).









Goethe State Forest Population Characteristics

The State of Florida purchased the Goethe Tract in 1992 as part of its

Conservation and Recreation Lands (CARL) program. Management authority was given

to the Florida Department of Agriculture and Consumer Services, Division of Forestry.

Although red-cockaded woodpeckers were known to exist prior to state acquisition, the

status of the population at the Goethe State Forest was unknown. A cooperative

agreement with the Florida Fish and Wildlife Conservation Commission conducted a

survey of the forest to determine the status of the resident red-cockaded woodpecker

population. In 1994, the geographic location and characteristics of red-cockaded

woodpecker cavities at the Goethe State Forest were identified, and cluster boundaries

were developed using the circular scale technique (Hovis 1996). Nesting data were used

to determine cluster status. Active status was assigned to areas with a single active tree

or a group of active cavity trees, some of which included an active tree occupied by a

nesting pair. Areas with groups of cavity trees that were not used by the red-cockaded

woodpecker were deemed inactive. In 1995, 26 clusters were active in the Goethe State

Forest. They were divided geographically into north and south regions of the forest.

Monitoring of cluster status was conducted on an annual basis by the Goethe State Forest

staff, during spring and summer months. In 2000, 30 clusters were active (23 were

nesting), 16 of which were located in the northern part of the forest. The remaining 14

clusters were in the southern region of the main tract. Figure 1-1 includes the distribution

of the 30 clusters that were uninhabited, or inactive, in 2000. The distribution of nesting

clusters in 2001 is included in Figure 1-2. In this study, the geographic location of these

nineteen nesting cavities were used to identify selected habitat for various analyses.















N






Cluster Status 2000
' Active
SNesting
I Inactive
SGoethe State Forest

1 0 1 2 3 4 Miles


Figure 1-1. Red-cockaded woodpecker cluster status in 2000, Goethe State Forest,
Florida.


-n


























Nesting Clusters 2001
[ ] Goethe State Forest

1 0 1 2 3 4 Miles
I'


Figure 1-2. Map of 19 red-cockaded woodpecker nesting sites during 2001. A 320 acre
buffer surrounds each site.









Staff biologists at the Goethe State Forest monitor red-cockaded woodpecker

cavities on an annual basis, usually during breeding season (May through June). A

portable "peeper" camera unit or ladder assembly are used to observe fledgling status or

other species within nesting cavities. The spatial location of all active and inactive cavity

trees is recorded using the Global Positioning System (GPS). Cavity trees are marked

with a single white band (approximately 8 in wide) around the stem at eye level.

Within active clusters, a minimum of four viable cavities must be maintained at

the Goethe State Forest. Artificial cavities are installed in living pine trees, if needed, as

cavity trees may die or become occupied by other species such as red-bellied

woodpeckers (Melanerpes carolinus), or pileated woodpeckers (Dryocopuspileatus).

Inserts are only installed if site and pine tree characteristics are suitable. The minimum

diameter at breast height of the selected pine trees must be 17 in to accommodate the size

of the cavity insert. The diameter at the point of insertion is generally larger than 15 in

This diameter permits the excavation of a hole 4 in wide, 10 in tall, 6 in deep to hold the

cavity insert (Allen 1991). The minimum installation height is 20 ft and the opening of

the insert box generally faces a southwest direction or is pointed to active cavity trees

within the cluster. If the crown of a potential cavity tree touches the crown of others, it is

generally not considered ideal. Longleaf pines which are "flat-topped" are avoided to

leave potential nest sites for natural excavation. The species of pines is not often

considered by biologists at the Goethe State Forest, as it is common for the larger pines

that grow near cypress ponds to be chosen for cavity construction. Cavity inserts may be

installed near the border of cypress forests, if suitable pine tree selection is limited. In

this situation, cavities are installed in a fashion so that openings do not face the cypress









ponds. Cavity installation and cosmetic preparation is completed. The newly added

cavity closely resembles a natural active cavity. Recruitment clusters consist of cavity

inserts, usually a pair, installed within a half mile range of an existing nesting cluster.

Site conditions are generally similar to those found in the neighboring active cluster.

Study Objectives

To support habitat restoration and provide current baseline data for management

decisions, this research focuses on the following:

a) Develop a geo-referenced forest and understory vegetative cover inventory
ArcView database of active and inactive cluster sites.

b) Test for differences in pine tree species and vegetative cover characteristics
between selected and non-selected habitat.

c) Evaluate the effectiveness of logistic regression as a tool for identifying forest
variables that have a significant association with selected versus non-selected
habitats.














CHAPTER 2
METHODOLOGY

Study Area

Field data were collected within the main tract of the Goethe State Forest, located

in the southeastern portion of Levy County, Florida (2922' N, 8237' W and 296' N

8232' W). The main Goethe tract is comprised of 49,295 ac and is managed by the

Florida Division of Forestry for multiple-use purposes throughout scrub, sandhill, pine

flatwoods, and dome swamp ecosystems. Activities offered to the public include:

hunting and camping by permit, wildlife viewing, bicycling, hiking, and horseback

riding.

At the time of this study there were 60 known red-cockaded woodpecker clusters

throughout the tract (Figure 1-1). The forest landscape is primarily pine flatwoods with a

high proportion of intermixed baldcypress swamps and hardwoods. The latter are

considered unsuitable for foraging. Hardwood features influence the area of good quality

foraging habitat that can be used throughout the red-cockaded woodpecker's home range.

This is apparent by the variation in estimates of foraging ranges in forests that contain

hardwood substrate (U.S. Fish and Wildlife Service 2000). The active red-cockaded

woodpecker population is split into two isolated groups: the north and south regions of

the forest (Figure 1-2). This separation is most likely related of a lack of suitable cavity

trees and dense midstory vegetation which follows drain flats in the central part of the

main tract. Unfortunately, the historical distribution of clusters at the Goethe State Forest

is not known.









Within the sampled clusters, pine overstory consists of slash pine, longleaf pine,

and scarce loblolly pine (Pinus taeda). Understory vegetation is dominated by saw

palmetto, gallberry (Ilex glabra), fetter-bush (Lyonia lucida), grasses, and forbs.

Data Collection

Field data were collected in 60 clusters, as identified by the Florida Division of

Forestry to be sampled for standing tree data and understory vegetation composition. In

this study, potential foraging habitat is defined as contiguous pine forest within each

cluster. Active clusters are management units defined by the presence of cavity trees

with flowing sap from resin-wells or the presence of a breeding pair. Inactive clusters are

abandoned red-cockaded woodpecker cavities at the time of study. Cypress and

hardwood areas were excluded from sampling.

A combination of compass bearings and pacing was used to navigate between pre-

determined sample points, each representative of ten acres of foraging habitat. Sample

points were located systematically within each cluster on a 10-chain (660 ft) square grid.

At each sample location, geographic coordinates were recorded with a handheld global

positioning system (GPS) unit, used for a 20 second duration. Data were post processed

with an accuracy of one to three meters. A forest inventory was conducted at each

sample location and data were recorded with a ruggedized handheld computer (CMT

PC5-L, Corvallis, OR) operating on Field Dog software (Two Dog Inc., Blacksburg,

VA). Vegetative ground cover attributes were recorded in a field book. Data collection

began in May 2000 and completed in August 2001.

Forest Inventory

Management officials at the Goethe State Forest identified 52 clusters that were

considered priority for forest inventory. At each sample location within these clusters, a









10-factor prism was used to select pine trees with probability proportional to tree DBH.

Limiting-distance calculations were used for borderline trees. For each sampled pine

tree, the species and DBH were recorded. Pines with a DBH < 4 in were not recorded.

At each sample point, the nearest dominant or co-dominant of the most common pine

species from the plot center was selected as the site index tree. Tree age, bark thickness,

and total tree height were measured. Stem cores were extracted at breast-height to

determine the age of the site index tree, and its five-year radial growth. A bark gauge

was used to measure bark thickness at breast height. The height of each site index tree

was measured with a precise vertex hypsometer.

Understory Vegetation Sampling

At each sample point, a square meter quadrat was used to assess ground cover

composition. Palmetto, forbs, and woody-stem categories were each estimated as percent

cover within the sample quadrat. Average height was estimated for each category in 3 ft

intervals. Dominant species for forbs and woody stem categories were recorded in the

field. Total percent cover for leaf litter, exposed mineral soil, and grasses (noting

wiregrass (Aristida beyrichiana) if dominant) were estimated and recorded.

Data Analysis Procedures

Baseline Statistics

An initial quantified summary of vegetative characteristics associated with red-

cockaded woodpecker habitat as well as abandoned forage was developed to support

habitat restoration and to provide current baseline statistics and data for management

decisions. This process was carried out using traditional t-tests for differences in

measurement averages for dominant pine basal area and DBH between active and









inactive clusters. The classification of samples used in testing was based on the zone or

cluster status in the analysis, as described below.

Tests for differences in means were performed at two foraging zone scales. The

320 ac area (zone A) surrounding nesting cavities was used to represent preferred habitat,

i.e. the expected area used by the red-cockaded woodpecker at the Goethe State Forest

(Figure 2-1). Samples collected within a larger zone of 500 ac, represent the potential

foraging area within each active and inactive cluster. The 500 ac foraging areas were

classified as active if either nesting cavities or a single active cavity tree (trees with resin

wells) was present within a cluster. All clusters that contain previously used red-

cockaded woodpecker cavities but are not currently in use were considered inactive.

Samples used to represent non-selected habitat were selected from zone B which is

formed by a "ring" between the 320 ac zone (zone A) and the outer margin of the 500 ac

zone (Figure 2-1).

A geographic information system was used to classify samples as either preferred

or non-selected habitat. Classified samples within the buffers as described above, were

exported as a.dbffile for testing in Minitab and SPSS statistical analysis software

packages.

In addition, the center of inactive clusters was buffered by a 2110 ft radius.

Samples located within this buffer may be representative of selected habitat, if cluster

abandonment was recent. Therefore, samples beyond the 2110 ft buffer and within the

500 ac inactive cluster boundary were classified as non-selected habitat. A third zone

was used to compare immediate nesting habitat characteristics to non-selected habitat.







16










































Active Cavity Zone A 0 Goethe State Forest boundary i

S Sample Location one B2 0 02 Miles




Figure 2-1. Location of a red-cockaded woodpecker nest cavity, and zones used to

identify habitat characteristics within the maximum sampled foraging area.









A 75 ac buffer was created around 19 nesting cavity locations. The size of this

buffer includes 86 percent of active cavity trees associated with the nesting cavities.

Data were managed using Microsoft Excel and tested for differences between

measurement means using SPSS and MINITAB software packages. Interpretation of

these results provided a basis for selecting variables to be used in logistic regression

models.

Binary Logistic Regression

Cluster recruitment requires an in-depth knowledge of forest conditions that are

suitable for artificial cavity construction within potential forest habitats. Identifying

which resources are selected most often by the red-cockaded woodpecker provide

important information about the nature of the species' habitat preference at the Goethe

State Forest. In previous studies, longleaf pine trees were used by red-cockaded

woodpeckers disproportionately to their availability among other pine species (Hovis and

Labisky 1985, Nesbitt et al. 1978). The binary logistic regression procedure was used to

explore the relationship between existing habitat and used habitat at the Goethe State

Forest. Based on the abundance of slash pine at the Goethe State Forest, it is possible

that the population of red-cockaded woodpeckers may be selective. However, this

assumption does not hold true for all populations, as foraging preference was positively

correlated with individual tree and stand age, independent of species availability

(Engstrom and Sanders 1997, Rudolph and Conner 1994, Zwicker and Walters 1999). In

this study, binary logistic regression was used to estimate the probability and odds of

habitat selection based on measured forest characteristics.









The availability of forest resources is not generally uniform, and use may change

as availability changes. Therefore, used resources should be compared with available (or

unused) resources to reach a valid conclusion concerning resource selection (Manly et al.

1993). In this study, sample points are classified according to the type of habitat they

represent. These classifications represent the dependent, or use variable (Y). A resource

probability function estimates the probability that a measured resource (X) in a particular

binary category (Y) is used by the red-cockaded woodpecker.

Binary logistic regression was chosen for several reasons. Data collected in this

study do not include a quantifiable estimate of habitat use by the woodpecker. A general

linear model is not suitable because the outcome variable (habitat use) was not measured

on a continuous scale. The outcome variable is binary, as determined by the

classification of samples based on the zone they are collected in from the cluster center or

nesting cavity. Samples collected within the 75 ac zone, or zone A are used as indicators

of 'selected resources'. Information obtained outside of these zones, where the

woodpecker is not known to forage, are used to define resources that are not preferred.

Criteria that are related to resource selection vary depending on forest resource condition.

However, factors such as pine species, age, midstory height, and understory species

composition are generally associated with red-cockaded woodpecker habitat (Hardesty et

al. 1997, Loeb et al. 1992, Rudolph and Conner 1994, U.S. Fish and Wildlife Service

2000). Because of the dichotomous nature of the dependent variable and the combination

of discrete and continuous predictors, the binomial distribution was used for regression

analysis. For this type of analysis, the 'logit transformation' of variables is often used

because it is very flexible and an easily used function which lends itself to a biologically









meaningful interpretation (Hosmer and Lemeshow 1989). The logistic model produces

an output in terms of probability, which is bounded by 0 and 1. The interpretation of the

logistic regression coefficients incorporates the calculation of odds, or the likelihood of

association, with an outcome. This is accomplished by transforming the probability to an

odds to remove the upper bound. The lower bound is removed by using the logarithm of

the odds. The result is set to a linear function of the explanatory variables. The

probability (p,) of g is calculated using the 'logit' model:

S= a + Ax,, + 2x,2 +... + k ,k


where p, =
l+eg

a = the outcome variable, and /k = predictor(s)

for k explanatory variables, and i = 1,...n individuals.

This equation has the desired property that p, will always be between 0 and 1 for

any number that is substituted for the s and the x's (Allison 1999). This probability is

sometimes excluded from interpretation in favor of an odds ratio or the likelihood of a

predictor being a member of an event. The odds ratio for each predictor can be solved by

using the regression coefficient ( pk ) of the predictor variable as the exponent of e, or the

base of the natural logarithms (Grimm and Yamold 1995).

Tests for Habitat Association

Tests were used to determine if the values of predicted coefficients of habitat

variables were different from zero. The null hypothesis assumes that the predictor is zero

in the population, and if rejected, the alternative hypothesis that the coefficient differs

from zero, is accepted. For models containing n predictors, a chi-square distribution with









n degrees of freedom is used to obtain the probability for the likelihood statistic, G.

Coefficient parameters) differ from zero if the probability of calculated G-statistic is less

than .05 (the cutoff probability for the hypothesis test, a = .05). The z test is another

approach to determine if an estimated value of a coefficient is different from zero. The

predictor coefficient is divided by its standard error to compute z, which is a measure of

expected variability in the coefficient among samples. The cutoff probability for z tests

in this study is .05. Confidence intervals for odd ratios were computed to estimate the

range of likelihood of association for a predictor with 95 percent confidence.

Interpretation of this interval provides a reliable estimate of increase in the odds of

association with foraging habitat from one pine species (or other habitat component) to

another.

Foraging Preference Analysis: Multivariable Models

Data used to indicate the outcome variable (Y) were selected from two categories:

a) Samples within the 320 ac foraging area (zone A) represent 'preferred' habitat (coded

as 1) b) Samples collected in zone B were considered 'non-selected' habitat (coded as 0).

An initial screening of habitat components was conducted by fitting each potential

variable to a resource probability function to eliminate habitat characteristics with a low

probability of association with preferred habitat. This was done in order to develop a

model with a minimum number of variables for field application. Four variables tested

for significant association with selected habitat (above 95 percent probability) during

initial testing across all nesting clusters: a) longleaf dominance, b) slash pine basal area,

c) longleaf diameter at breast height, and d) longleaf basal area (Table 4-8). Based on the

statistical significance of these results, the outcomes from previous studies, and their role









in supporting the woodpeckers' ecological niche, select predictors that suggested a strong

association with habitat selection were then tested interactively.

Two models, consisting of 612 samples each, were tested to explore associations

of multiple habitat characteristics with habitat preference across the entire forest

population (Models lp and 2p). Model lp includes one predictor (b,), which involves the

interaction between: a) sample points with an average diameter of pines > 33 cm (coded

1, if true) and b) samples dominated by longleaf pine (coded 1, if true). A minimum

diameter of 13 in was chosen because of cited preferences of the red-cockaded

woodpeckers to forage on trees larger than 12 in when available (Engstrom and Sanders

1997), and because samples which contain trees of this average DBH are less prevalent

throughout the Goethe State Forest.

The second test for habitat preference include two predictors. The first coefficient

(b1) in Model 2p is the interaction of samples where a) the average DBH of longleaf pine

samples was > 33 cm (coded 1, if true) and b) samples where wiregrass was present in

the understory (coded 1, if true). The interaction of these factors is expressed as a binary

variable. The second coefficient of Model 2p is longleaf dominance, coded 1, if present

at a sample and coded 0, if longleaf pine was not dominant. Tests for multicollinearity,

using a correlation matrix, were performed for all interactive coefficients. The two

models used in the population analysis were also used to test for differences in habitat

preference between the geographically separated north and south cluster groups. The

models in the north are termed Model 1N and Model 2N, and in models which tested

variables in the south are termed Model is and Model 2s.









Nesting Habitat Model

A third model was employed to estimate the probability of resource selection for

specific pine species within the foraging area that surrounds the immediate area of

nesting sites. Model 3 involves 150 random samples taken from two regions of the forest.

Samples within the 75 ac buffer zone surrounding 19 nesting sites, were used to represent

preferred habitat and samples. Data collected within zone B (Figure 2-1) were

representative of unused habitat. Seventy-five random samples were used from each

zone in the logistic regression analysis to determine if the longleaf pine dominance, (bl=1

if the sample was dominated by longleaf pine, and 0 if not) could be used to predict

selection of nesting sites. Output statistics for Model 3 denote the probability and

likelihood of selection between longleaf pine dominance and a binary variable Y, where

Y= 1 for samples collected within the 75 ac zone, and Y= 0 for samples collected in zone

B. The Minitab procedure BINARY LOGISTIC REGRESSION was used to fit each model to

its respective data.














CHAPTER 3
RESULTS: COMPARISON OF HABITAT COMPONENTS

Pine Characteristics Throughout Foraging Zones

Within nesting habitats (75 ac zone), 60 percent of sampled trees were longleaf

pine, while slash pine comprised the remaining 40 percent (Figure 3-1). Proportions of

longleaf and slash pine within zone A were similar, but less pronounced. Slightly more

than half (53 percent) of the sample trees in zone A were composed of longleaf pine. In

zone B, representative of unused habitat, almost two-thirds (61 percent) of the tree

samples were slash pine.

Across all 52 clusters, slash pine represented 54 percent of the sampled trees in

the potential red-cockaded woodpecker habitat (500 ac/cluster). Within these 500 ac

clusters, the remainder of sampled trees were longleaf (approximately 46 percent) and

loblolly pine (< .6 percent). However, pine species abundance differed based on cluster

status. Active units were 51 percent longleaf pine while 62 percent of the trees in

inactive units were slash pine (Figure 3-1). In active clusters, 29 of 572 samples were

recorded as having no dominant (>50 percent of total samples) species. Of 679 samples

in inactive clusters, 39 contained equal numbers of slash and longleaf pines.

Nesting habitat within the 75 ac zone (n = 94), averaged 36 ft2/ac of longleaf pine

basal area, substantially more than the average slash pine basal area (22 ft2/ha) (Table 4-

1). Compared to the concentration of pines in zone B, the immediate area surrounding

nesting sites averaged approximately 11 ft2/ac more longleaf basal area and 11 ft2/ac less

slash pine basal area (Table 4-1).










70

60

50

S40








Nesting Zone A Zone B Active Inactive Forest-
Habitat Cluster Status wide


Figure 3-1. Percentage of sampled longleaf and slash pine trees at various habitat scales.
The DBH of longleafand slash pines within nesting habitat was similar to









averages found in other habitat zones. The range of average longleaf and slash pine DBH

estimates was 1.7 in and 5.3 in, respectively, across all foraging zones.

The DBH of pines was not significantly different in zones A (n = 338) and B (n =

301) (Table 4-6). Zone A averaged 31 ft2/ac basal area of longleaf pine, which exceeds

longleaf basal area in zone B by 6.9 ft2/ac. Slash pine basal area in zone B is greater than

in zone A, by an average of 6.0 ft2/ac (Table 4-1).

Pine tree DBH is similar for active (n = 572) and inactive (n = 679) clusters with a
difference of 1.7 in for longleafand 2.0 in for slash pines, respectively (Table 4-3).
20

10















Longleaf pine basal area is 9.5 ftac greater in active versus inactive clusters, on average.
Habitat Cluster Status wide














FigurBetween pe 3-1. species in active clusters, longleaf and slash proific,ne trees at various habitat scales.

more than slash pine, whlongleaf and slash pine basal area is higher on average in inactive clusters

averagby 12.6 found in other habitat zones. The range of average longleaf and slash pine DBH4-1).
estimates was 1.7 in and 5.3 in, respectively, across all foraging zones.

The DBH of pines was not significantly different in zones A (n = 338) and B (n =

301) (Table 4-6). Zone A averaged 31 ft2/ac basal area of longleaf pine, which exceeds

longleaf basal area in zone B by 6.9 ft2/ac. Slash pine basal area in zone B is greater than

in zone A, by an average of 6.0 ft2/ac (Table 4-1).

Pine tree DBH is similar for active (n = 572) and inactive (n = 679) clusters with a

difference of 1.7 in for longleaf and 2.0 in for slash pines, respectively (Table 4-3).

Longleaf pine basal area is 9.5 ft2/ac greater in active versus inactive clusters, on average.

Between pine species in active clusters, longleaf is most prolific, averaging 5.6 ft2/ac

more than slash pine, while slash pine basal area is higher on average in inactive clusters

by 12.6 ft2/ac (Table 4-1).









Active and Inactive Cluster Habitat Composition Analysis

Comparisons of Pine Basal Area

Basal area within active and inactive clusters averaged 62.7 ft2/ac and 59.8 ft2/ac,

respectively (Table 4-2). Although there is no statistical difference between these overall

basal area estimates, basal area is greater by 10.6 ft2/ac, (p = 0.0000) when longleafis

dominant in active versus inactive clusters (Table 4-2). Samples taken in habitat

dominated by slash pine averaged a significantly higher basal area than longleaf pine in

both active and inactive clusters. Active clusters are characterized by a mean basal area

of 67.4 ft2/ac when slash pine is dominant compared to 59.6 ft2/ac when longleaf

dominates the sample (p < 0.006). Inactive clusters are characterized by samples where

slash pine was most often dominant. When compared to areas dominated by longleaf

pine, inactive clusters contained 21 ft2/ac more basal area in areas dominated by slash

pine. This difference was tested significant (p = 0.0000). Basal area of sample points

dominated by slash pine was not significantly different (p = 0.39) between active (67.4

ft2/ac) and inactive (70 ft2/ac) clusters. The estimates for total basal area of longleaf

dominated pine habitat in active and inactive clusters are within the United States Fish

and Wildlife Service guidelines, but given the small average DBH (<13 in) of pines,

habitat in these clusters is of marginal quality.

Differences in Pine DBH

Within active clusters, the average DBH per sample point is 11.9 in. This is

significantly larger (p = 0.061, a = 0.10) than the average DBH of samples in inactive

clusters, which averaged 11.7 in (Table 4-3). When longleaf pines dominate the sample

location, the average DBH is 11.7 in within inactive clusters and 12.1 in within active

clusters.














Table 4-1. Pine characteristics in nest habitat and foraging zones at the Goethe State Forest.


320 acre Zones (x)


Pine Species

Basal Area (ft2/ac)
Longleaf DBH (in)


Basal Area (ft2/ac)
Slash
lash DBH (in)

a denotes sample range
b denotes + 1 standard error


75 acre Zone (x)

36 (10100o)a
11.8 .2b

22.0 (10-120)
12.3 + .3


Zone A


Zone B


31.35 (10-130) 24.3 (10-130)
11.9 .09 12.0 .1


27.0 (10-170)
11.9.1


33.14 (10-170)
11.7.11


500 acre Clusters (x)

Cluster Status
Active Inactive


31.3 (10-150) 21.7 (10-120)
11.8 .07 12 ..09


25.7 (10-170)
11.8 .07


34.4 (10-200)
11.49 .07













Table 4-2. The average basal area of pines relative to longleaf or slash pine dominance within active and inactive clusters.

Mean (ft2/ac)


Dominant Pine Species Active Clusters Inactive Clusters P Value
Longleaf Basal Area 59.6 1.7a 49 .36 0.000
Slash Pine Basal Area 67.4 2.7 70 .41 0.390
P Value 0.006 0.000
Basal Area 62.7 + 1.3 59.8 + 1.2 0.098
a denotes 1 standard error


Table 4-3. The average Diameter at Breast Height (DBH) of pines relative to longleaf or slash pine dominance within active and
inactive clusters.

Mean (in)


Dominant Pine Species


Active Clusters


Inactive Clusters


Longleaf DBH 12.1 .12 11.7 .15 0.073
Slash DBH 11.8 .19 11.6 .13 0.200
P Value 0.350 0.480
DBH 11.9 .10 11.7 .10 0.061
a denotes 1 standard error


P Value









These estimates are not significantly different at the 95 percent confidence level, but they

are significantly different when tested at a = .10 (p = 0.073). The average DBH for

samples dominated by slash pine is 11.8 in, and 11.6 in within active and inactive

clusters, respectively. Similarly, these estimates are not different at the .95 confidence

level (p = 0.20). Within active and inactive clusters, tests do not provide sufficient

evidence for differences in mean DBH between samples where longleaf and slash pine

are dominant (p = 0.350 and p = 0.480, respectively). It is clear that trees sampled within

clusters, regardless of species, are very nearly the same DBH.

Differences in Basal Area of Large Pines

For samples where longleaf pine is dominant and the average DBH is greater than

or equal to 12 in, the average basal area in active clusters is 4.3 ft2/ac greater than in

inactive clusters (Table 4-4). Basal area for sample points dominated by slash pine

averaging 12 in DBH or greater is 70.7 ft2/ac and 67.7 ft2/ac in active and inactive

clusters, respectively. Differences between basal areas of sample points dominated by

either large longleaf or slash pine in both active (p = 0.0000) and inactive (p = 0.0000)

clusters were highly significant (Table 4-4). The apparent association between increased

levels of basal area of larger pines in active clusters was tested further in this study

through the use of logistic regression models.

Wiregrass Observations

In active clusters, wiregrass was the dominant grass in 60 percent of samples that

contained coverage (> 5 percent) of any grass within the square meter sampling quadrat.

For inactive clusters, wiregrass was observed as the dominant grass in 25.2 percent of

grass samples. The groups of red-cockaded woodpeckers that live in active clusters may













Table 4-4. The average basal area of pines > 12 in for samples dominated by either longleaf or slash pine within active and inactive
clusters.

Mean (ft2/ac)


Dominant Pine Species Active Clusters Inactive Clusters P Value
Longleaf Basal Area 54.4 +.2.1a 49.7 +2.0 0.110
Slash Pine Basal Area 70.7 + .2.9 67.7 2.6 0.570
P Value 0.000 0.000
a denotes 1 standard error









be remnant of a population that was sustained by fire and where wiregrass was prolific

throughout their territory. Today's inactive cluster status may be the result of fire

exclusion, which, among other factors, could cause wiregrass presence to dwindle as new

understory species colonized unburned red-cockaded woodpecker territories.

Habitat Characteristics Within Zone A and Zone B

Comparisons of Pine Basal Area

The mean basal area per sample point in zone A was 60.2 ft2/ac. Samples in zone

B averaged 59.6 ft2/ac basal area, which does not differ significantly from zone A.

Samples located in 'selected habitat' of zone A and dominated by longleaf pine had a

basal area that was significantly less than samples dominated by slash pine (p < 0.000)

(Table 4-5). Similarly, within 'non-selected' habitat in zone B, the basal area of longleaf-

dominated habitat is significantly less than the basal area in samples dominated by slash

pine (p<0.000). When samples dominated by longleaf pine were compared between

foraging zones, samples in zone A contained a significantly higher basal area (p = 0.046)

than in zone B. The significantly greater levels of basal area in longleaf pine dominated

habitat is similar in both zones and the larger, active clusters. Much like the cluster level

analysis, a test for differences between slash pine basal area between habitat zones A and

B did not provide evidence for statistical difference (p = 0.660) (Table 4-5).

Differences in Pine DBH

The average diameter for all pine samples within zone A was 12.0 in, which is

significantly larger (p = 0.014) from the average diameter of 11.6 in within zone B. This

significant difference was not found in a comparison of the average DBH of all pines

sampled within active and inactive clusters (Table 4-3).













Table 4-5. Test results for differences in mean basal area between zones A and B. T-tests were conducted at a = 0.05.


Mean (ft2/ac)


Dominant Pine Species Zone A Zone B P Value
Longleaf Basal Area 55.6 1.9a 49.0 2.1 0.046
Slash Pine Basal Area 66.9 2.6 70 2.2 0.660
P Value 0.000 0.000
Basal Area 60.2 + 1.3 59.6+ 1.2 0.810
a denotes 1 standard error


Table 4-6. Test results for differences in mean DBH between samples collected within zones A and B. T-tests were conducted at a
0.05.

Mean (in)


Dominant Pine Species Zone A Zone B P Value
LongleafDBH 12.2 .14a 11.5 .20 0.006
Slash DBH 11.8 .23 11.5 .18 0.330
P Value 0.130 0.950
DBH 12.0 .13 11.6 .13 0.014
a denotes 1 standard error









Possible reasons for this characteristic change could be the removal of large, harvestable

timber in densely stocked nesting habitat of inactive clusters. Perhaps overstocking close

to roosting cavities was a contributing factor to cluster abandonment. Tests for

differences in mean pine tree DBH between samples dominated by slash and longleaf

pines did not differ significantly from each other in either habitat zone (Table 4-6).

Samples dominated by longleaf pine had a significantly larger (p = 0.006) DBH in

preferred habitat in zone A (12.2 in) versus zone B (11.5 in). The average DBH for

samples dominated by slash pine within zone A was not dissimilar from the average pine

tree DBH for samples dominated by slash pine in zone B (Table 4-6).

Differences in Basal Area of Large Trees

Samples dominated by longleaf pine with a minimum average diameter of 12 in

contained significantly less basal area (49.7 ft2/ac, p < 0.000) than samples dominated by

slash pine with a minimum diameter of 12 in (68.1 ft2/ac) in zone A. Similarly, the basal

area of large slash pine was statistically higher (p < 0.000) than samples dominated by

large longleaf pines within zone B. The increased levels of basal area associated with

slash pine dominated samples is most likely due to the substantial area of pines which

border cypress ponds. These areas are almost exclusively slash pine, and because of the

moist soil conditions, these pines are often of large size given a young age. The skewing

effect of the large slash pines on estimates of basal area and quality of habitat should be

studied further through field observation of foraging habits and buffer analysis using a

geographic information system. The basal area of samples dominated by large longleaf

pine trees (> 12 in) within zone A did not significantly differ (p = 0.250) from trees of the

same condition found within zone B. Statistical tests revealed no evidence that large









slash pine basal area differs between zones A (selected) and B (non-selected forage)

(Table 4-7).

Observed Wiregrass

Within the preferred habitat of zone A, understory vegetation samples suggest that

wiregrass dominated 60.7 percent of samples (n = 84) when grass was present. Non-

selected habitat in zone B included proportionally fewer wiregrass observations, as

wiregrass was dominant in only 23.6 percent of samples which contained some type of

grass (n = 110).

Logistic Regression Model Estimates

Variable Screening Results

During the development of habitat association models, an interactive process was

used to find a good subset of variables, i.e. a subset that includes only significant

predictors that result in strong negative and positive predictive values. Table 4-8 shows

the predictive coefficient value (b, ) and results for probability of habitat association for

each of 16 variables. The iterative process eliminated certain variables, such as percent

leaf litter cover. Although the probability of association with habitat for this variable is

statistically significant (p = 0.0000), its predictive value is low as it is at or near zero.

Other variables that were not significantly associated with active habitat were retained for

model development. In this mode of variable testing, active habitat and large longleaf

pines were not significantly associated (p = 0.9499). However, the t-test results on

cluster data indicate that large longleaf pines were significantly larger in active clusters

(Table 4-3). This factor was considered in conjunction with wiregrass presence as an

interactive variable in Model 1, based on the assumption that the two variables are

indicative of historic habitat conditions.













Table 4-7. Test results for differences in basal area of pines > 12 in within zones A and B. T-tests were conducted at a = 0.05.

Mean (ft2/ac)


Dominant Pine Species Zone A Zone B P Value
Longleaf Basal Area 49.7 2.4a 45.9 .2.5 0.250
Slash Pine Basal Area 68.1 +3.2 70.0 3.4 0.750
P Value 0.000 0.000
a denotes 1 standard error










Table 4-8. Results from fitting the logistic regression equation separately to the data on
habitat selection by red-cockaded woodpeckers at the Goethe State Forest in
2001.

Predictor Variable (X) Significance Constant
(*=binary) p Term bx
Leaf Litter Cover % .0000 -.0286
Longleaf Dominance* .0037 .4744
Slash Pine Basal Area .0069 -.0065
LongleafDBH .0136 .0352
Longleaf Basal Area .0124 .0074
Palmetto % Cover .0524 .0054
Slash Pine DBH .0910 -.0226
Forb % Cover .1932 -.0161
Exposed Ground % .2734 .0031
Wiregrass Presence* .2958 .2470
Average DBH .3006 .0330
Basal Area .4399 -.0021
Av. DBH > 13"* .7371 .0575
Slash Pine DBH > 13"* .7984 -.0475
Wood Stem % Cover .8190 -.0009
Longleaf DBH > 13"* .9499 -.0112

Following the initial screening, two models were finalized, and were tested using

field data. The potential for field application was a major consideration in the

formulation of the models.

Forest-Wide Model Diagnostics

Data from 612 sample points were used in Models lp and 2p to evaluate habitat

characteristics for significant association with red-cockaded woodpecker habitat within

all nesting clusters at the Goethe State Forest. For Model lp, a correlation coefficient

(CC) of 0.02 suggests no sign of multicollinearity between samples with an average DBH

of all pines > 13 in, and samples that were dominated by longleaf pine (Table 4-9). The

evaluation of the z-score of b1 in Model lp (z = 1.76, p = 0.078) suggests that it does not

yield significant results at the 95 percent confidence level. The Model's test statistic does

not exceed the critical value of 3.841. Therefore, the null hypothesis, which states that

the model variables are not associated with nesting habitat (b, = 0), cannot be rejected.









Even if this test level is relaxed, this model does not yield significant results at the 90

percent significance level.

Model 2p includes two variables that contribute to the overall understanding of

habitat selection. The predicted probability of habitat selection was solved by the

following equation and subsequent transformation:

g= -.0806 + .8091 (1)+ .4330 (1) = 1.161

The predicted probability is e161 /(1 + e1161), or 76 percent. Thus the probability

that areas dominated by large longleaf pines which contain wiregrass are associated with

nesting habitat is 76 percent. Alternatively, the odds of association with nesting habitat is

1.16 times greater if those variables are present. Coefficients and related statistics are

shown in Table 4-9. The "G-statistic" was used to test for at least one of the variables as

statistically significant contributing factor in understanding habitat selection. The

computed value of G is 12.5 with two degrees of freedom. The probability associated

with this statistic is less than 0.002. This indicates that the alternative distribution is

accepted and at least one of the coefficients {bl, b2 is not equal to zero. The z-score for

longleaf dominance (b2) is positive and has a probability less than .05. This means that

habitats dominated by longleaf pine tend to be selected as opposed to habitats where it is

not dominant, when taking into account habitats where there is a combination of large

longleaf pine (> 13 in) and wiregrass presence. Raising e to the power of b2 gives the

odds ratio of 1.54. Model 2p suggests that when controlling bl the odds of habitat

association increase by 1.54 if longleaf is dominant. At the .10 level of confidence this

coefficient does appear to have a significant association with habitat selection.
















Table 4-9. Results from fitting the logistic regression model to habitat data collected throughout 19 nesting clusters of the Goethe
State Forest, Florida.



Odds
Ratio (eb) 95 % CI
Forest-Wide Models Coefficient SE G p Z p CC (eb) Lower Upper
Ip. Constant (bo) .0987 3.15 .076 .02
Av. DBH > 13 LL DOM (bi) .3868 .22 1.76 .078 1.47 .956 2.26
2p. Constant (bo) -.0806
LL DBH > 13" WP (bl) .8091 .42 12.5 .002 1.91 .056 .06 2.25 .98 5.14
LL DOM (b2) .4330 .16 2.62 .009 1.54 1.12 2.13









w-










It can be said with greater than 90 percent certainty that the variables in Model 2 are

strongly associated with nesting habitat.

North and South Regions

Within the northern part of the Goethe State Forest, 274 samples taken near

nesting cavities were used to characterize the association between habitat selection and

vegetative characteristics. Preferred and non-selected habitat were analyzed in the

southern population with data collected at 338 sample locations.

Table 4-10. Results from fitting the logistic regression model to habitat data collected in
the northern region of the Goethe State Forest, Florida.
Odds
Ratio (eb) 95 % CI
Model Variable (North) Coefficient SE G p Z p CC (eb) Lower Upper
1N. Constant (bo) .7329 0.07 .788
Av. DBH > 13 LL DOM (bi) .0938 .35 .27 .789 .12 1.10 .55 2.18
2N. Constant (bo) .6039
LL DBH > 13" WP (bi) 21 5.29 .071 0.00 .998 .03 0 *
LL DOM (b2) .1999 .26 .76 .009 1.22 0.73 2.04

Model In tested samples where the average diameter of all pines was at least 13 in

and longleaf pine was dominant. The correlation coefficient for bl in Model In was 0.12,

which indicates no strong relationship exists between the average diameter of pines > 13

in and longleaf pine dominance. These variables, when tested interactively, were not

significantly associated with habitat selection in the north (p = 0.788) (Table 4-10). In

this case, as with the results from the population Model lp, the null hypothesis is not

rejected. By comparison, in the southern cluster group, the coefficient bl in Model ls was

significantly associated (p = 0.027) with selected habitat (Table 4-11). In addition, there

is no indication of autocorrelation between the factors in bl observed in Model ls (CC=

0.02). Data from this sample suggest that the odds of habitat association, given the

presence of dominant longleaf pine and pines averaging > 13 in, is estimated to be

between 1.07 and 3.34 higher for the southern population at the 95 percent level of









certainty. The likelihood ratio statistic for Model Is (4.91) has a probability of 0.027,

which suggests a positive association between the variables used in Model is and the

preferred habitat.

The probability of association for variables in Model 2s is:

g= -.4844 + 1.02 (1) + .2981 (1)

.8337

and e 8337 /(1+ e 8337) = .69

Within the Goethe State Forest, if habitat contains large longleaf with wiregrass,

and is dominated by longleaf pine, the likelihood of habitat association is 69 percent

greater than if these variables are not present. This assumes that other environmental

factors such as pine density and understory height are within acceptable limits.

On the basis of the z-score in Table 4-10 for Model 2N, this particular sample

generated a b coefficient that does not differ from zero (p = 0.998). However, longleaf

dominance (b2), is most likely different from zero since its associated z-score of 1.22

with a probability of 0.009 may suggest otherwise. An overall likelihood ratio test (G)

was used to determine if either predictor is different from zero. For Model 2N, the G

statistic is 5.29 with a probability of 0.071, suggesting that with almost 93 percent

certainty that one of the model predictors differs from zero. There is no correlation of

interactively tested variables in Model 2N (Table 4-10). This models poor performance

was mostly due to highly variable data and smaller sample size when compared to the

southern population.










Table 4-11. Results from fitting the logistic regression model to habitat data collected in
the southern region of the Goethe State Forest, Florida.

Odds
Ratio (eb) 95 % CI
Model Variable (South) Coefficient SE G p Z p CC (eb) Lower Upper
is. Constant (bo) -.3995 4.91 .027 .02
Av. DBH > 13 LL DOM (bl) .6379 .29 2.21 .027 1.89 1.07 3.34
2s Constant (bo) -.4844
LL DBH > 13" WP (bi) 1.02 .45 8.61 .014 2.28 .023 .06 2.79 1.16 6.75
LL DOM (b2) .2981 .23 1.30 .192 1.35 0.86 2.11

Measurements for bl in Model 2s (the interaction between large longleaf and

wiregrass presence), were not correlated (CC = 0.06). The associated z-score for bl is

2.28 with a probability of 0.023, indicating that in the southern region, bl denotes a

significant relationship with habitat selection when considering b2, longleaf pine

dominance (Table 4-11). A 95 percent confidence interval, placed on the odds ratio

(2.79), suggests that if the variable b2 = 1 in the southern region, the level of habitat

association with the red-cockaded woodpecker is between 1.16 and 6.75 times greater

compared to habitat that does not include both large longleaf pines and wiregrass. The

probability of habitat association, assuming habitat with all factors present as stated in the

model is:

g= -.4844 + 1.02(1) + .2981(1)

= .834


e834 (+e834 )=.70

In this situation, the complete model was used to predict probability of association

even though b2 was not statistically significant (p = 0.192). This is justified by the

significant G statistic (8.61 with 2 degrees of freedom) (p = 0.014) obtained by evaluating

the overall model.









Nesting Habitat

The computed odds ratio (e8115) for Model 3 was 2.25 (Table 4-12). The predicted

probability (g) of longleaf pine preference (bl = 1) surrounding nesting habitat is:

k=-0.3947 + .8115*(1)

= .4168

e 4168 /(1 + e4168) =.60. The lower and upper confidence limits of the odds ratio are 1.17

and 4.33. One may argue with 95 percent confidence that the odds of habitat use for nest

sites at the Goethe State Forest is between 1.17 and 4.33 times greater if longleaf pine is

dominant rather than if longleaf pine is not the dominant species, assuming other

environmental factors are held constant. The likelihood ratio statistic (G) was used to test

whether the predictor variable is zero. The computed value for G = 6.045, which is

considered large (p = 0.014), indicated that the variable is not equal to zero. The null

hypothesis, which assumes the predictor coefficient is zero, was rejected. The probability

of obtaining a z-score higher than 2.43 is less than 0.015. Thus, as with the likelihood

ratio test, the result of the z-test indicates that a positive association exists between

longleaf dominance and nest cavity habitat. Because the predicted probability is greater

than 0.50, one may argue that maintenance of longleaf dominated habitat near nesting

cavities is a priority at the Goethe State Forest. The iterative model building process used

in this study considered several habitat components, but used a minimum number of

variables for simplified field application. Most models performed well when applied to

data for the forest population but did not when applied to the north and south sub-

populations. This is an indication that for hypothesis tests to be accurate, logistic

regression requires large samples.







42



Table 4-12. Computed coefficients and scores for longleaf pine as a predictor of nesting
habitat selection.


Odds
Ratio


(eb) 95 % CI


Model 3: Variable Coefficient SE G p Z p (eb) Lower Upper
3. Constant (bo) -.3947 6.04 0.014
LL DOM (bi) .8115 ..333 2.43 0.015 2.25 1.17 4.33














CHAPTER 4
DISCUSSION OF FORAGING ZONE ANALYSES

Habitat Characteristics at the Goethe State Forest

In this study, a geographic information system was used to examine pine forest

conditions within various spatial scales of active and inactive red-cockaded woodpecker

habitat. A high proportion of longleaf pine samples (60 percent) was located within 1020

ft radius of nest cavities. By comparison, inactive clusters (half mile radius) contained

approximately 45 percent longleaf pine. The indication that most longleaf-dominated

habitat used for foraging is preferred over available slash pine-dominated habitat is

consistent with previous studies that show a preference for, or selection of available

longleaf pine over slash pine habitat in Florida (DeLotelle et al. 1987, Nesbitt et al. 1978,

Porter and Labisky 1986).

The extent of red-cockaded woodpecker foraging zones varies between forests

(Nesbitt et al. 1978, Porter and Labisky 1986, Zwicker and Walters 1999). Results of this

study indicate that habitat dominated by larger longleaf pines, compared to other

available pines, is associated with habitat at both spatial scales used for analyses in this

study (zones A and B, and active and inactive clusters). Samples in areas dominated by

longleaf trees in foraging zone A average significantly greater DBH (P = 0.006) and basal

area (P = 0.046) compared to samples in zone B. The difference in average DBH and

basal area between zones, is .68 in and 5.7 ft2/ac, respectively. Longleaf pine basal area

in zone A is 7.0 ft2/ac greater than in zone B. Compared to inactive clusters, longleaf

basal area is 9.6 ft2/ac greater in active clusters. This indicates a positive trend towards









longleaf preference given its availability. Although data on actual use or selection in

longleaf dominated habitats was not collected, dominant pine overstory preference is

apparent. The similarity of large longleaf pine trees in both active and inactive foraging

zones and clusters suggest that although used habitat is dominated by longleaf overstory,

preference may be influenced by factors other than DBH. Selection for cavity

construction and foraging preference of longleaf pine is most likely related to a

combination of factors such as age, good resin-producing abilities, and heartwood decay

(Conner et al. 2001).

Quality of Forest Vegetation Within Clusters

Based on guidelines set forth by the U.S. Fish and Wildlife Service, basal area of

longleaf systems should range between 40 and 60 ft2/ac, while the basal area of shortleaf

pine forests should range between 40 and 80 ft2/ac (U.S. Fish and Wildlife Service 2000).

This study suggests that in active clusters at the Goethe State Forest, the basal area of

longleaf (60 ft2/ac) and slash (67 ft2/ac) pine systems fall within these guidelines, and

cover a minimum of 60 percent of the area within a half mile of the cluster center.

Samples contained significantly more basal area dominated by slash pine throughout

active (P = 0.000) and inactive (P = 0.000) clusters. Both classes of foraging areas

contained cypress swamps, bay heads, and hardwood drainages that are mostly bordered

by slash pines and occasionally by loblolly trees. These slash pine are abundant, given

the high perimeter to area ratio as swamps were often oval-shaped or in strips. The

presence of cypress swamps decreased the total area of potential forage within this zone.

The moist conditions which surround them may contribute to the increased size of slash

pines which exclusively bordered these areas, possibly affecting habitat suitability.









Diameter distribution of pines in both active and inactive clusters were similar in

size. On the average, these pines, most likely, are not large enough to contain sufficient

heartwood diameter (5-6 in) for cavity tree excavation (Conner et al. 2001). It is known

that longleaf pines are selectively chosen if they contain red heart fungus, which

facilitates the excavation process (Rudolph et al. 1995). Although the outer diameter of

pines in active clusters at the Goethe State Forest may suit cavity excavation, the average

age of sampled trees is 55.1 (SE=1.45 years), well below the suggested minimum age of

60 to 80 years. Age is most likely a limiting factor in nest cavity selection (U.S. Fish and

Wildlife Service 2000). Although the standing timber may be extensive enough and of

suitable size, the lack of potential cavities is the dominant factor limiting the red-

cockaded woodpecker's survival (Hovis and Labisky 1996).

Results of this study indicate that there is less basal area dominated by large

longleaf pines throughout the 500 ac zone, and large trees (>12 in) in active clusters have

a lower average basal area compared to longleaf of all sizes (>4 in). The average basal

area of large slash pines (70 ft2/ac) was greater than the basal area of all slash pines > 4 in

(67 ft2/ac) in active clusters. This relationship may be the result of forest management

and the abundance of cypress ponds, rather than selection by the red-cockaded

woodpecker. Previous management activities such as timber harvesting within clusters

are most likely the cause for the decrease in large longleaf pine basal area. The

abundance of large slash pine basal area is most likely due to the area surrounding

cypress ponds where environmental conditions favor rapid growth of the species.

The understory vegetation parallels the recommended guideline's definition of

"good quality foraging habitat" since it is dense with fire tolerant and fire dependent









species such as wiregrass, saw palmetto, gallberry and fetterbush. Wiregrass dominates

approximately 60 percent of the samples where grasses were identified in active clusters.

This is significantly larger then the 25 percent of grass samples dominated by wiregrass

in inactive clusters and may be an indication that the present day red-cockaded

woodpecker groups forage in areas which maintain site characteristics of earlier

populations. However, it is difficult to determine the historical locations of red-cockaded

woodpecker cavity sites and the surrounding understory vegetative conditions before

state forest management practices began at the Goethe State Forest in the mid 1990's

(Hovis 1996). It is possible that the clusters where wiregrass is less prevalent may serve

as an indication of other contributing factors which caused the abandonment of clusters,

such as infrequent burning, or anthropogenic disturbance.

The samples collected within the 500 ac foraging zone cover a larger area than

has been generally documented for home ranges in central Florida. Among the largest

documented home ranges are those in the Stanton Energy Center in Florida where habitat

consists largely of pine savanna with sparse tree densities (DeLotelle et al. 1987).

Average foraging ranges in Florida's studied populations are larger than in other southern

states. This may be related to the lower density, younger and smaller size-class of pines

associated with habitat in the species' southern margin. The amount and quality of

available habitat are contributing factors to these estimates as home range sizes vary

considerably within uniform habitat (Hovis and Labisky 1996). Conner et al. (2001)

estimate 230 to 383 ac for home range sizes in central Florida. Because of the relatively

young age of potential cavity trees at the Goethe State Forest, an expected foraging range

of 320 ac was chosen to further understand habitat preference in this case study.









The 320 Acre Foraging Zone

Previous studies have shown that cavity trees are usually located in stands with a

low overstory basal area. Red cockaded woodpecker colonies studied at Apalachicola

National Forest had a basal area that was considerably lower (46 ft2/ac) than in adjacent

areas (65 ft2/ac) (Hovis and Labisky 1985). Availability of longleaf and slash pine basal

area in zones A and B was nearly equal in our study. Mean basal area in zone A (60.2

ft2/ac) is not significantly different from the surrounding zone B (59.6 ft2/ac). In this

study, the density for both zones is relatively low (<78 ft2/ac) and the habitat is

considered open grown. Use of longleaf pine for gum naval stores is evident by many

remaining scarred trees which have been selected for nest cavity construction by the

woodpecker and are likely infected with red-heart fungus.

The basal area of available pine forage 4 in within zone A (60.2 ft2/ac) is within

the recommended guidelines for "good quality foraging habitat" for longleaf systems

(U.S. Fish and Wildlife Service 2000). Nearly all samples of understory vegetation were

less than 6 ft in total height. Groundcover at the Goethe State Forest is mostly

contiguous, fire tolerant saw palmetto, and fire dependent herbs. Very few areas of

canopy hardwoods were observed that were not within pre-identified areas. The mapped

cypress and hardwood areas covered approximately 40 percent of available forage within

2110 ft of nest cavity trees. Compared to other studies that reported a cypress component

within Florida habitat, the red-cockaded woodpecker population at the Goethe State

Forest has adapted to habitat containing the largest area of cypress forest proportional to

foraging zone size. The large percentage of cypress may indirectly support the species'

foraging needs as large slash pines (>17 in) occupy the perimeter where soil conditions

are moist. Further analysis of the relationship between the red-cockaded woodpeckers









use of slash pines bordering cypress swamps, the cypress swamp size, perimeter area, and

frequency of occurrence may give insight to the impact of landscape features on habitat

requirements.

Application of a Binary Logistic Regression Model

In this case study, binary logistic regression was used to explore the potential of

one single-variable explanatory model and two multi-variable explanatory models to

characterize and quantify association between habitat characteristics and habitat

preference. The potential of different resource attributes were tested for likelihood and

probability of association with areas preferred by red-cockaded woodpeckers at the

Goethe State Forest. However, this study was not an exhaustive attempt on all aspects of

the logistic regression. Likelihood estimates indicate forest habitat dominated by

longleaf pine is positively associated with nesting habitat.

The iterative model building strategy used in this study was employed to design a

simplified model for field application. The combination of forward model building

through tests on individual variables and the selection of variables based on previous

studies, allowed us to find the best subset of a large set of predictors. However, given

that wiregrass is less abundant than saw palmetto, the interaction of saw palmetto and

longleaf pine at the Goethe State Forest most likely would have produced significant

results when tested for association with preferred forage. Because wiregrass is

historically associated with the species and is easily identifiable in the field, it was

incorporated into Model 2. When longleaf is not dominant (b2 = 0), the likelihood of

association for variables in Model 2p is higher (67 percent) if large longleaf and

wiregrass are present (bl = 1), compared to the likelihood if the combination is not

present (bl = 0) (58 percent). The higher odds ratio of bl (2.25) compared to b2 (1.54),









where estimates greater than one indicate an association with nesting habitat compared to

non-selected habitat, is also an indication that habitat preference is dependent on a

combination of forest attributes.

Given the homogeneous nature of pine forest throughout the sampling area, and

prominent wiregrass in active habitat, it appeared that tests would report similar results at

different scales. However, the same models which indicate habitat preference at the

population level did not produce similar results when tested at smaller scales. The results

from models tested on north and south populations are indications that for hypothesis

tests to be accurate, logistic regression requires large samples. Our results are due to the

low number of samples that contained both wiregrass and large longleaf pines. Results

may also be related to changes in the density of red-cockaded woodpeckers by forest

region or due to the variation and availability of resource units, which can change the

selection strategies and selection function. A separate resource selection function for

several independent replications, with larger sample sizes, would be necessary to

accurately depict resource preference within north and south regions.

Suggestions For Future Research

The use of a geographic information system to manage and interpret data

collected throughout the Goethe State Forest brings a series of new ways to:

a) Characterize cypress pond distribution, density, and area within habitat.

b) Associate new flight and foraging preference data with existing forest stand
characteristics and understory vegetation records.

c) Relate the effects and intensity of silvicultural practices to the health of red-
cockaded woodpecker.









The database created for this study can be readily updated with new data collected

by the Goethe State Forest staff as well as future researchers. This would facilitate

spatial and temporal analyses, and consolidate records for assessment of management

practices.

Investigation of the distribution of cypress ponds within red-cockaded

woodpecker habitat could provide insight as to whether or not it is a contributing factor to

the area or amount of available forage needed by the species at the Goethe State Forest.

Analysis of pond characteristics such as area, density, and perimeter-to-area ratio may be

used to interpret the influence of cypress ponds on the size and distribution of clusters.

This type of analysis is dependent on existing and new information such as the spatial

distribution of foraging habits.

This study used an estimated foraging radius based on published data collected at

nearby red-cockaded woodpecker sites in central Florida to interpret associations between

expected foraging area and vegetative characteristics. New information on the actual

flight patterns and distance traveled by the red-cockaded woodpecker at the Goethe State

Forest would be critical for an analysis of foraging area in relation to cypress pond

characteristics. This information could be displayed showing the geographic distribution

of foraging habits in relation to existing data from this study such as pine species, basal

area, and tree diameter throughout foraging areas. New information such as cypress pond

characteristics, stand age, fire events, timber harvests, and recreational activities could be

used as factors tested to impact the distance or direction of foraging efforts by the red-

cockaded woodpecker at the Goethe State Forest.









Existing forest inventory data, integrated with data on silvicultural practices and

red-cockaded woodpecker recovery efforts would support a comprehensive analysis of

the changes in habitat preference over time. This meaningful data should be incorporated

into the decision-making process and may be used as a process model designed to

enhance existing red-cockaded woodpecker populations.

These suggestions, as well as social issues associated with red-cockaded

woodpecker restoration, illustrate the need for new information and provide opportunities

for future research. The use of dynamic models and statistical analyses will continue to

support the decision-making process. Development of new and specific restoration

efforts will benefit greatly from continued cooperative research between the Florida

Division of Forestry and the University system.














CHAPTER 5
MANAGEMENT IMPLICATIONS

Plans for restoration of red-cockaded woodpecker populations are mainly

developed using guidelines designed by the United States Fish and Wildlife Service.

Current management efforts at the Goethe State Forest are primarily supported by those

guidelines. However, to prioritize new habitat restoration efforts, management objectives

should consider the following:

a) The significant and strong association found between nesting cavities and longleaf
pines > 13 in DBH within close range (1020 ft).

b) Introduction of prescribed fire or mechanical treatment to reduce dense saw
palmetto coverage.

c) A minimum intensive management unit of 225 ac surrounding nest cavities.

Logistic regression models developed in this study clearly indicate the preference

of longleaf pines > 13 in DBH within foraging areas surrounding nest cavities. In most

management units where the red-cockaded woodpecker occurs, forest management

strategies should continue to enhance the availability of these characteristics in frequently

burned longleaf pine stands. Although various uneven-age timber harvesting techniques

within management units can be used to retain some foraging value and provide financial

gain, persistence of old trees should be emphasized. Regardless of species, trees greater

than 100 years old should be omitted from timber sale, as they would leave red-cockaded

woodpeckers without high quality forage, and potential excavation sites to use while

stands mature.









Vegetation in the understory of sampled clusters is dominated by saw palmetto.

The leaves of this plant contain chemical properties that, when ignited, could produce

extremely intense fires. Given the small average diameter of pines found within foraging

clusters, it is important to reduce the amount of saw palmetto through the use of

controlled burning or mechanical treatment. Adjacency to sensitive areas and weather

conditions may prohibit burning in some areas. An alternative solution to remove

excessive saw palmetto or vegetation which may disturb the red-cockaded woodpecker's

foraging habits, may be a series of treatments which include a combination of chemical

treatment to large hardwoods and mechanical removal of large, dense saw palmetto. In

an optimal situation, the forest understory should be diverse with forbs and grasses whose

composition is dense enough to sustain a periodic controlled fire. Although the process

may be time and cost intensive, immediate efforts to reduce understory height below 12 ft

should focus on areas where vegetation is near or exceeds this level in nesting clusters.

This effort would sustain the existing populations and prepare for less costly management

alternatives such as translocation efforts to enhance the existing population.

Translocation and artificial cavity construction efforts used to create new colony

sites should be targeted in longleaf dominated habitat adjacent to existing active clusters.

These areas should contain a minimum of 225 ac of suitable pine forage with little or no

hardwood midstory. This management unit size is based on test results for association of

longleaf pine across three potential habitat zones, published research on red-cockaded

woodpecker populations in central Florida, and efficiency of use in evaluation of field

data for suitable habitat. A comparison between future estimates of forest characteristics

and baseline data from this study may be used to evaluate management alternatives.
















LIST OF REFERENCES


Affeltranger, C. 1971. The red-heart disease of southern pines. Pp. 96-99 in R. L.
Thompson, ed. The Ecology and Management of the Red-Cockaded Woodpecker,
Proceedings of a Symposium at Okefenokee National Wildlife Refuge. Folkston,
GA.

Allen, D. H. 1991. An insert technique for constructing artificial Red-cockaded
Woodpecker cavities. U.S. Forest Service General Technical Report. SE-73.
19pp.

Allison, P. D. 1999. Logistic regression: Using the SAS system: Theory and application.
SAS Institute. Cary, NC. 288pp.

Azevedo, J. C. M., S. B. Jack, R. N. Coulson, and D. F. Wunneburger. 2000. Functional
heterogeneity of forest landscapes and the distribution and abundance of the red-
cockaded woodpecker. Forest Ecology and Management. 127:271-283.

Carter, J. H., J. R. Walters and P. M. Dixon. 1995. Red-cockaded Woodpeckers in the
North Carolina Sandhills: A 12 Year Population Study. Pp. 248-258 in D. L.
Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery,
ecology, and management. Center for Applied Studies in Forestry, Stephen F.
Austin State University. Nacogdoches, TX.

Conner, R. N., D. C. Rudolph, and J. R. Walters. 2001. The Red-cockaded Woodpecker:
surviving in a fire maintained ecosystem. University of Texas Press, Austin, TX.
363pp.

Conner, R. N., and D. C. Rudolph. 1991. Forest habitat loss, fragmentation, and Red-
cockaded Woodpecker populations. Wilson Bulletin. 103:446-457.

Conner, R. N., D. C. Rudolph, R. Schaefer, D. Saenz, and C.E. Shackelford. 1999.
Relationships among the red-cockaded woodpecker group density, nestling
provisioning rates, and habitat. Wilson Bulletin. 111(4):494-498.

DeLotelle, R. S., R. J. Epting, and J. R. Newman. 1987. Habitat use and territory
characteristics of red-cockaded woodpeckers in central Florida. Wilson Bulletin.
99(2):202-217.









Dennington, R. W., and R. M. Farrar. 1983. Longleaf pine management. United States
Department of Agriculture, Forest Service Southern Region. Forestry Report R8-
FR3. Pp. 17.

Engstrom, T. R., and G. Mikusinski. 1998. Ecological neighborhoods in red-cockaded
woodpecker populations. The Auk. 115:473-478.

Engstrom, T. R., and F. J. Sanders. 1997. Red-cockaded woodpecker foraging ecology
in an old growth longleaf pine forest. Wilson Bulletin. 109(2):203-217.

Epting, R. J., R. S. DeLotelle, and T. Beaty. 1995. Red-cockaded woodpecker territory
and habitat use in Georgia and Florida. Pp. 59-67 in D. L. Kulhavy, R. G. Hooper
and R. Costa, eds. Red-cockaded woodpecker: Recovery, ecology, and
management. Center for Applied Studies in Forestry, Stephen F. Austin State
University. Nacogdoches, TX.

Gaines, G. D., K. E. Franzreb, D. H. Allen, K. S. Laves, and W. L. Jarvis. 1995. Red-
cockaded woodpecker on the Savannah River Site: A management success story.
Pp 81-85 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded
woodpecker: Recovery, ecology, and management. Center for Applied Studies in
Forestry, Stephen F. Austin State University. Nacogdoches, TX.

Grimm, L. G., and P. R. Yarnold. 1995. Reading and understanding multivariate
statistics. American Psychological Association, Washington D.C. 373pp.

Hanula, J. L., and K. Franzreb. 1998. Source, distribution and abundance of
macroarthropods on the bark of longleaf pine: Potential prey of the red-cockaded
woodpecker. Forest Ecology and Management. 102:89-102.

Hardesty, J. L., K. E. Gault, and H. F. Percival. 1997. Ecological correlates of red-
cockaded woodpecker foraging preference, habitat use and home range size in
northwest Florida (Eglin AFB). Final Report, Research Work Order 99. The
Nature Conservancy, Gainesville, FL.

Hooper, R. G. 1996. Arthropod biomass in winter and the age of longleaf pines. Forest
Ecology and Management. 82:115-131.

Hooper, R. G., L. J. Niles, R. F. Harlow, and G. W. Wood. 1982. Home ranges of Red-
cockaded woodpeckers in coastal South Carolina. The Auk. 99:675-682.

Hopkins, M. L., and T. E. Lynn. 1971. Some characteristics of Red-cockaded
woodpecker cavity trees and management implications in South Carolina. Pp
140-169 in D. L. Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded
woodpecker: Recovery, ecology, and management. Center for Applied Studies in
Forestry, Stephen F. Austin State University. Nacogdoches, TX.









Hosmer, D. W. Jr., and S. Lemeshow. 1989. Applied logistic regression. Wiley series in
probability and mathematical statistics. Applied probability and statistics. Wiley,
New York. 307pp.

Hovis, J. A. 1996. Status and management of the red-cockaded woodpecker on Goethe
State Forest, Florida. Southeast Association Fish and Wildlife Agencies Annual
Conference Proceedings. 50:254-263.

Hovis, J. A., and R. F. Labisky. 1996. Red-cockaded woodpecker. Pp. 81-102 in J. A.
Rodgers, Jr., H. W. Kale II, and H. T. Smith, eds. Rare and endangered biota of
Florida: Volume V. Birds. University Press of Florida, Gainesville, FL.

Hovis, J. A., and R. F. Labisky. 1985. Vegetative associations of red-cockaded
woodpecker colonies in Florida. Wildlife Society Bulletin. 13:307-314.

Ligon, D. J. 1971. Some factors influencing numbers of the red-cockaded woodpecker.
Pp. 30-43 in R.L. Thompson ed. The ecology and management of the Red-
cockaded Woodpecker. Proceedings of a Symposium at Okefenokee National
Wildlife Refuge. Folkston, GA.

Loeb, S. C., W. D. Pepper, and A. T. Doyle. 1992. Habitat characteristics of active and
abandoned red-cockaded woodpecker colonies. Southern Journal of Applied
Forestry. 16(3):120-125.

Manly, B. F., L. L. McDonald, and D. L. Thomas. 1993. Resource selection by animals.
Chapman and Hall. Great Britain. 177pp.

Nesbitt, S. A., D. T. Gilbert, and D. B. Barbour. 1978. Red-cockaded woodpecker fall
movements in a Florida flatwoods community. The Auk. 95:145-151.

Outcalt, K. W., and R. M. Scheffield. 1996. The longleafpine forest: trends and current
conditions. United Stated Department of Agriculture, Southern Research Station.
Resource Bulletin SRS-9. Asheville, NC. Pp. 23.

Porter, M. L., and R. F. Labisky. 1986. Home range and foraging habitat of red-
cockaded woodpeckers in northern Florida. Journal of Wildlife Management.
50(2):239-247.

Ross, W. G., D. L. Kulhavy, and R. N. Conner. 1997. Stand conditions and tree
characteristics affect quality of longleaf pine for red-cockaded woodpecker cavity
trees. Forest Ecology and Management. 91:145-154.

Rudolph, D. G. and R. C. Conner. 1994. Forest fragmentation and red-cockaded
woodpecker population: an analysis at intermediate scale. Journal Field
Ornithology. 65:365-375.









Thomlinson, J. R. 1996. Predicting status change in red-cockaded woodpecker cavity
tree clusters. Journal of Wildlife Management. 60:350-354.

U.S. Fish and Wildlife Service. 2000. Technical/agency draft revised recovery plan for
the red-cockaded woodpecker. U.S. Fish and Wildlife Service, Atlanta, GA.
229pp.

Walters, J. R., S. J. Daniels, J. H. Carter, III, P. D. Doerr, K. Brust, and J. M. Mitchell.
2000. Foraging habitat resources, preferences and fitness of red-cockaded
woodpeckers in the North Carolina Sandhills. Fort Bragg Project Final Report.
Virginia Polytechnic Institute and State University, Blacksburg, VA, and North
Carolina State University, Raleigh, NC.

Walters, J. R., P. P. Robinson, W. Starnes and J. Goodson. 1995. The relative
effectiveness of artificial cavity starts and artificial cavities in inducing the
formation of new groups of Red-Cockaded Woodpeckers. Pp. 367-369 in D. L.
Kulhavy, R. G. Hooper and R. Costa, eds. Red-cockaded woodpecker: Recovery,
ecology, and management. Center for Applied Studies in Forestry, Stephen F.
Austin State University. Nacogdoches, TX.

Zwicker, S. M. and J. R. Walters. 1999. Selection of pines for foraging by red-cockaded
woodpeckers. Journal of Wildlife Management. 63(3):843-852.















BIOGRAPHICAL SKETCH

Douglas Owen Shipley was born in Hinsdale, Illinois, on 17 August 1976. In

1998, he received a B.S. in environmental resource management from Virginia

Polytechnic Institute and State University. After graduation, he worked as a forester at

the Seminole State Forest for the Florida Division of Forestry. In December 2002, he

completed requirements for the Master of Science degree, at the University of Florida.