POST-DISTURBANCE DYNAMICS IN THE RELATIVE INFLUENCE OF SPATIAL SCALES ON PINELAND BIRDS By JOHN E. ARNETT JR. A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006
Copyright 2006 by John E. Arnett Jr.
This thesis is dedicated to my grandparents who always believed I could perform acts of rare brilliance. It is also dedicated to the denizens of our planet whose voices are not heard or understood by human ears.
iv ACKNOWLEDGMENTS Many people contributed effort, ideas, and support during the course of this study. Any shortcomings of this thesis are in no wa y a reflection of their input. I especially thank Dr. George W. Tanner, my advisor, for inviting me to do this project, for providing me with plentiful resources a nd â€œintellectualâ€ freedom, for his patience as I wallowed in the mire of uncertainty and apathy, and fo r his chili-cooking and joke-telling. My committee, including co-advisor Dr. Debbie Mi ller and Dr. Katie Sieving, was very helpful throughout the entire process. I especi ally appreciate their logic, wisdom, and ability to make sense out of the nonsensical. I also thank Dr. Lyn Branch and Dr. Mike Moulton for their time and words of motivation. This project was part of the Strategic Environmental Research and Development Program (SERDP) Ecosystem Management Pr oject (SEMP), and was funded by the U.S. Department of Defense, in partnership with the Department of Energy and the Environmental Protection Agency. Hugh Westbur y, SEMP site coordinator, diligently fit my schedule in and around the Fort Benning m ilitary mission. I am still alive and am in good health; one could infer that Hugh did his job well. Pete Swid erek, Darrell Odom, Jack Greenlee, and Rusty Bufford of the Fort Benning Natural Re source Management Branch generously provided information, insight, and logistical support (e.g., office space!). I cannot thank Rob Addington of Th e Nature Conservancy enough for his help and friendship.
v Anna Prizzia was an excellent field assist ant. I thank her for her dedication and great spirit even when her feet were frozen or when I was in a fowl mood (pardon the pun). Additional field assistance was provi ded by Lauree Stober, Annemarie Prince, Melissa Moyer, and Jaelean Carrero. Many thanks go to all who employed me during my graduate studies: Dr. Mel Sunquist, Dr. Mike Moulton, M onica Lindberg, and Dr. Winni e Cooke. I thank Dr. Ron Labisky for the opportunity to co-author two publ ications. The staff of the Department of Wildlife Ecology and Conservation was always he lpful whether I was in the field or in Gainesville. Willie Wood of the School of Fo rest Resources and Conservation graciously loaned field equipment and storage space. I thank fellow graduate students Ja nell Brush, Arjun Gopalaswamy, Sonia Canavelli, and Meredith Evans for discus sions and assistance regarding distancesampling, Jason Hall for vicarious inspiration, Jessica Archer for information about Fort Benning, and Susan Carr for help with statistic al quandaries. Dr. B ob Dorazio, Dr. Mary Christman, Jana Cole, and Dr. Ken Portier of the Institute of Food and Agricultural Sciences Department of Statistics also provide d statistical advice. I am very grateful to Ann George for her friendship, cu isine, and analytical advice.
vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT....................................................................................................................... ..x CHAPTER 1 INTRODUCTION........................................................................................................1 2 STUDY AREA AND METHODS...............................................................................7 Study Area....................................................................................................................7 Study Design and Site Selection...................................................................................8 Avian Data..................................................................................................................10 Plot Data.....................................................................................................................14 Stand Data...................................................................................................................16 Landscape Data...........................................................................................................16 Analysis......................................................................................................................18 3 RESULTS...................................................................................................................32 Microhabitat, Stand, and Landscape Composition.....................................................32 Winter..................................................................................................................32 Spring..................................................................................................................33 Avian Community Com position and Density.............................................................33 Winter..................................................................................................................33 Spring..................................................................................................................35 Variation Partitioning.................................................................................................36 Analysis Level 1..................................................................................................37 Analysis Level 2..................................................................................................38 Analysis Level 3..................................................................................................39 4 DISCUSSION.............................................................................................................58 Relative Importance of Spatial Scale..........................................................................58
vii Variation in the Relative In fluence of Spatial Scales.................................................60 Importance of Landscape to Neotropical Migrants....................................................62 Robust Methodologies................................................................................................63 Limitations and Assumptions.....................................................................................66 Conclusions.................................................................................................................68 APPENDIX A BIRDS OBSERVED AT FORT BENNING..............................................................71 B BIRDS INCLUDED IN VARIATION PARTITIONING ANALYSES....................74 LITERATURE CITED......................................................................................................75 BIOGRAPHICAL SKETCH.............................................................................................86
viii LIST OF TABLES Table page 1 Description of covariates used in the estimation of detection functions and bird density using the Multiple Covariates engine (MCD S) in program DISTANCE....23 2 Pools of species, the number of obser vations in each pool, and the number of observations of bird specie s detected at line transe cts during winter (Jan-Mar) and spring (Apr-May) 2004, at Fort Benning, Georgia............................................24 3 Descriptions and sources of plot-, st and-, and landscape-level environmental variables describing managed pine stands studied at Fort Benning U. S. Army Infantry Center, Georgia, during 2004.....................................................................26 4 Summary statistics of environmental va riables for 86 pine stands at Fort Benning, Georgia, during winter (Jan-Mar), 2004...................................................43 5 Summary statistics of environmental va riables for 94 pine stands at Fort Benning, Georgia, during sp ring (Apr-May), 2004.................................................44 6 Densities (no./ha) of 30 bird species observed at more than five percent of line transects at Fort Benning, Georgi a, during winter (Jan-Mar), 2004........................45 7 Analyses and results of models genera ted in program DISTANCE for individual species and pools of infrequently detected species of birds.....................................46 8 Densities (no./ha) of 37 bird species observed at more than five percent of line transects at Fort Benning, Georgi a, during spring (Apr-May), 2004.......................48 9 Adjusted coefficients of determination ( R2 adj) from the partitioning of variation of winter (nonbreeding) and spring (br eeding) bird communities among plot-, stand-, and landscape-level environmenta l variables using RDA in CANOCO......49 10 Significant ( p < 0.10) environmental variables us ed in final RDA models and the significance level of trac e statistics determined by 999 random unrestricted permutations.............................................................................................................50
ix LIST OF FIGURES Figure page 1 Location of Fort Benning U.S. Army In fantry Center, Georgia (inset), and the locations of transects where bird communities and vegetation structure were studied during winter (Jan-Mar ) and spring (Apr-May), 2004................................28 2 Plot-level vegetation sampling scheme....................................................................29 3 Schematic diagram of the variation partitioning analyses performed separately for winter and spring bird communities...................................................................30 4 Schematic diagram and explanation of mathematical procedures for the calculation of conditional and marginal fr actions of variance in bird community data explained by plot-, st and-, and landscape-level environmental variables........31 5 Relative frequencies of sele cted covariates in analyses of bird density performed in the multiple covariates engine (MCDS) of program DISTANCE.......................55 6 Detection functions calculated in pr ogram DISTANCE for birds in the Small Ground/Shrub Passerine pool, based on th e occurrence of Tufted Titmouse (TUTI) in mixed-species flocks...............................................................................56 7 Conditional effects of plot, stand and landscape characteristics on bird communities across a post-thinning chronosequence..............................................57
x Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science POST-DISTURBANCE DYNAMICS IN THE RELATIVE INFLUENCE OF SPATIAL SCALES ON PINELAND BIRDS By John E. Arnett Jr. December 2006 Chair: George W. Tanner Cochair: Deborah L. Miller Major Department: Wildlife Ecology and Conservation The importance of understanding how biological communities perceive and respond to multiple scales of environmental fact ors is now widely recognized. Very little information exists, however, on how ecological disturbance may affect the scales at which communities or functional groups of animals perceive their environment. Knowledge of how disturbance may cause tempor al variation in the re lative influences of multiple scales of habitat factors on organisms is critical given that the conservation of biodiversity often depends on effective anthr opogenic replication of natural disturbance regimes. I studied the relationships between bi rds and three habitat scales across a chronosequence of thinned pine forests at Fort Benning, Georgia, during winter (Jan â€“ Mar) and spring (Apr â€“ May), 2004. Forest management at Fort Benning is generally aimed at improving habitat quality for th e endangered Red-cockaded Woodpecker
xi ( Picoides borealis ). Densities of birds during the breeding and non-breeding season were determined with distance-sampling (line tran sect) methods, and the effects of covariates describing study site and organi smal characteristics on estima tes of bird detectability were examined. Variation partitioning (RDA) analyses were used to identify the joint and independent effects of plot-, standa nd landscape-scales of e nvironmental factors on bird communities and functional (residen t or migrant) groups of birds. In general, there was not a strong gradie nt in the environmental data following thinning. With time after thinning, shrub cover increased and woody litter decreased at sites studied in winter; at s ites studied in spring, the gra ss density and canopy cover both increased with time after thinning. The amount of military and altered lands within the landscape context was greatest at sites thinne d 10 years before study. Bird densities and species composition were similar across the chronosequence. Shrub cover, basal area, and the size of bird flocks were important covariates in models of bird detection. Despite the low amount of turnover in habita t characteristics and bird composition and density, temporal variation occu rred in the relative influence of habitat scales on both bird communities and functional groups of bird s. Environmental factors explained up to 9.6% of the variation in bird community composition/density, and up to 44.2% of the variation in the composition/de nsity of functional groups wi thin a chronosequence group. Breeding season residents were correlated with standand landscape-scale characteristics across the chronosequence. Neotropical migrants only responded to landscape characteristics at sites of older disturbance. Because community-level analyses obscured important insights, landscapeand functiona l group-based approaches to conservation management and monitoring were recommended.
1 CHAPTER 1 INTRODUCTION Biological communities are shaped by how species perceive and respond to ecological processes and patterns. Communitie s reflect the diversity (Rotenberry 1985) and structure (MacArthur and MacArthur 1961, Hilden 1965) of hab itat patterns in the environment. Among processes, disturbance plays a major role in the trajectory of community structuring (Sousa 1984). Because pa ttern and process are often inextricably linked in a synergistic rela tionship (Turner et al. 2001, Gustafson and Diaz 2002), the persistence of some communities depends on natural disturbances that create and maintain landscape mosaic patterns (A skins 2000, Brawn et al. 2001) and on landscape patterns to maintain the adequate frequenc y and intensity of natural disturbances. The influence of ecological processes a nd patterns on avian community structure varies across spatial scales (Wiens et al. 1987, Kotliar et al. 2002). For example, the distribution and abundance of bi rds may be differentially as sociated with fine-scale habitat characteristics (i.e., microhabitat), meso-scale attributes of the habitat patch or stand, or broad-scale composition and c onfiguration of the surrounding landscape (Pearson 1993, McGarigal and McComb 1995, Flather and Sauer 1996, Saab 1999, Mitchell et al. 2001, Lichstein et al. 2002). These findings support calls for studying species-environment relationships across multiple spatial scales (Levin 1992). The relative influences of habitat charact eristics across spatial scales are not consistent among bird communities (Mazeroll e and Villard 1999). Overall, bird communities may be more sensitive to e ither landscape (Pearson 1993, Saab 1999, Grand
2 and Cushman 2003) or microhabita t attributes (Hagan and M eehan 2002, Lichstein et al. 2002, MacFaden and Capen 2002, Miller et al . 2003, Cushman and McGarigal 2004a). However, because species-specific responses to spatial scale (Peterson et al. 1998) may be due to avian life histories (Whitcom b et al. 1981, Hansen and Urban 1992), these general trends may not hold when the multi-scale habitat relationships of functional groups of birds with shared life histor ies are considered (Flather and Sauer 1996, Jokimki and Huhta 1996, Graham and Blak e 2001, Mitchell et al. 2001, Tworek 2002, Cleary et al. 2005). Using functional groups to infer general species -environment relationships depends on the assumption that en vironmental fluctuations similarly affect all species within that group (Severinghaus 1981). Among avian life histories, migration strategy is a good predictor of interspecific variation in bi rd-environment relationships (Schmiegelow et al. 1997), perhaps because it is correlated with other avian life histories (Whitcomb et al. 1981). For example, in some multi-scale studies, landscape characteristics have been shown to be more important for neotropical migrants than for permanent residents (Flather and Sauer 1996, Lichstein et al. 2002, M itchell et al. 2006), and these patterns may have been obscured wh en the bird community was considered as a whole. While the study of species-environment rela tionships across multiple scales is a relatively new area of resear ch, considerable knowledge exists about the temporal response of species across successional and di sturbance gradients (summarized in Sousa 1984, Shugart 1998, Brawn et al. 2001). For exam ple, Emlen (1970) studied the post-fire succession of bird communities in south Florida slash pine forests, and speculated that the low rate of bird community species turnover was due to site fidelity. Perhaps due to site
3 fidelity, time lags may exist in the response of birds to ecological disturbance (Wiens and Rotenberry 1985). Wiens et al. (1987), therefor e, advocated long-term studies in order to understand â€œthe temporal consistency of bird -habitat associations.â€ To offset the practical difficulties imposed by conduc ting long-term studies, a chronosequence methodology that substitutes space for time can be used to address questions about longterm ecological pattern s (Pickett 1989). The importance of ecological phenomena, then, varies across scales in both time and space (Shugart 1998). Explicit considera tion of the effects of time has revealed important patterns of variati on in ecological relationships within scales. Studying a successional gradient of sites, Bersier and Meyer (1994) found variation in avian relationships to fine-scale habitat stru cture and floristic diversity. Community relationships to broad-scale la ndscape features were found to vary annually (Fuller et al. 1997), and may be better explained by former, rather than existing, landscape patterns (Burel 1992, Ernoult et al. 2006). Mitchell et al. (2001) concluded that relationships between birds and multiple landscape scales were invariant among species typically associated with stages of habitat succession; however, th eir study assigned successional stages to species based on the literature. Because bird-environment relationships vary across both spatial scales and temporal grad ients, and because disturbances operate at different rates across spatial scales (Baker 1993), study of th e effects of disturbance on multi-scale bird-environment relationships is warranted. Understanding the relationship between life histories and spa tio-temporal variation in spec ies-environment relationships may further our understanding of commun ity structuring processes (Bazazz 1996, Richards et al. 1999) and may have important implications for comm unityor landscape-
4 based conservation and management (Flather and Sauer 1996), particularly where human management now attempts to replicate natural disturbance regimes. As human activities continue to alter habitats and disrupt natural ecological processes, the long-term persistence of communities and ecosystems becomes increasingly dependent upon inte nsive management. In the sout heastern U. S., the extent, composition, age, and functioning of native longleaf , loblolly and shortleaf pine forests in have been greatly altered since European settlement (USFWS 2003). Little of the remaining southern pine forests can be cons idered true â€œold growthâ€ (Jackson 1988), and hardwood intrusion into pine forests reduces their suitability fo r the endangered Redcockaded Woodpecker ( Picoides borealis ; RCW). As a result, the open pine forests preferred by the RCW are now maintained a nd created with mechanical removal (i.e., thinning) of pines and midstory hardwoods in conjunction with prescribed fire (Wilson et al. 1995). Forest thinning in this context ai ms to alter the structural complexity and floristic diversity of the forest. Additi onally, the process of thinning impacts the understory vegetation and leaves behind lega cies such as slash and exposed soil. Thinning, therefore, may affect avian species composition by altering the availability of suitable microclimate, food, and substrates used for nesting and cover (Palik and Engstrom 1999). Avian diversity and density in open pine fo rests in the southeast are low relative to more mature forests with a greater ha rdwood component (Johnston and Odum 1956, Meyers and Johnson 1978). Additionally, ende mism is low; only four landbirds are nearly endemic, and just three others primarily use these forests (Jackson 1988). However, according to the Partners in Flight Species Assessment Database
5 (www.rmbo.org ), populations of at least five of th ese seven species have declined by at least 15% over the past 30 yr, four are of speci es of continental concern, and at least five warrant immediate management action (Car ter et al. 2000). Thus, understanding the potential impacts of RCW-focused manageme nt on other bird species inhabiting these forests is a high priority (Block et al . 1995, Askins 2000, Hunter et al. 2001). The effect of southeastern pine forest management has been studied for breeding birds (Wilson et al. 1995, Krementz and Ch ristie 1999, White et al. 1999, Provencher et al. 2002b, Tucker et al. 2003, Wood et al. 2004), nonbreeding (win ter) birds (White et al. 1996, King et al. 1998, Provencher et al. 2002a), bird communities of both seasons (Brennan et al. 1995, Conner et al. 2002), and for individua l species of particular management concern (Bachmanâ€™s Sparrow; Plen tovich et al. 1998). In general, these studies neither examined community dynamics over long-term temporal gradients (but see Repenning and Labisky 1985, White et al. 1999) nor across multiple spatial scales, thus limiting their usefulness for making predictions about avian community response to forest management (Marzluff et al. 2000). Integration of RCW-focused management with community-based (Hunter et al. 2001) and landscape-scale management strategies (Hunter et al. 1994) may resolve the potent ial conflicts between singleand multiplespecies habitat management. The managed pi ne forests of the southeastern U. S. represent an ideal system to study the eff ects of disturbance on bird-environment relationships at multiple scales. In this study, I conducted a comparativ e study (K. Sieving, pers. comm.) of breeding and nonbreeding season bi rd densities and environmental (habitat) measures at plot, stand, and landscape scales across a chronosequence of mechanically thinned pine
6 forests. I hypothesized that ecological di sturbance causes tempor al variation in the relative influence of spatial scales on communities and migratory functional groups of birds. In addition to examining this hypothesi s, I addressed the follo wing two predictions based on previous multi-scale studies: in general, microhabitat (plot-level) factors will explain more of the variation in the densitie s of the entire community of birds than will standor landscape-level factors; landscap e factors will explain a greater proportion of the variation in densities of neotropical mi grants than of short-distance migrants and permanent residents (Mitchell et al. 2006).
7 CHAPTER 2 STUDY AREA AND METHODS Study Area The Fort Benning U.S. Army Infantry Center was established in 1918 and currently occupies 73,533 ha (182,000 ac) in southwes tern Georgia and southeastern Alabama (Figure 1). As the primary training facility for the U.S. Army Infantry, military training at Fort Benning causes erosion and loss of topsoil (DeBusk et al. 2005), alters the structure of vegetation and i nvertebrate communities (Graha m et al. 2004), and leads to fineand large-scale conversion of land cove r types (Dale et al. 2005). Understanding and amelioration of ecological impacts are c onsidered integral to the success of the military mission at Fort Benning (USAIC 2001). Fort Benning is in the Fall-Line Sandhills region at the transition of the Coastal Plain and the Piedmont (Collins et al. 2006). Upland soil types are typically sands and loamy sands (Garten and Ashwood 2004). Ridge tops with gentle to steep slopes are interspersed with nearly leve l valleys and stream channels (DeBusk et al. 2005). Normal monthly temperatures range from 7.6o C in January to 27.7o C in July (pub. comm., 2 June, 2006, http://www.nrcc.corn ell.edu/ccd/nrmavg.html). More precipitation falls during March and April than in other months, mean annua l precipitation is 127 cm (Magilligan and Stamp 1997), and the mean a nnual number of days with thunderstorms is between 60 and 70 (Jackson 1988). Agriculture was the primary land use by the regional Native Americans (Dale et al. 2005), and was prevalent from the time of Eu ropean settlement until the 1940s (Graham
8 et al. 2004). Currently, the installation incl udes nearly 56,656 ha of manageable forest (USAIC 2001). Both evenand uneven-aged silvicultural systems are prescribed across the Fort Benning landscape. Loblolly ( Pinus taeda ) is the dominant pine (Graham et al. 2004). Mature longleaf pine ( P. palustris ) covers only four per cent of the installation (Dilustro et al. 2002) and wiregrass ( Aristida beyrichiana ), an important component of longleaf pine communities el sewhere, is absent. Fort Benning is one of fifteen designat ed recovery populations (USAIC 2001) for the federally endangered Redcockaded Woodpecker (RCW; sc ientific names given in Appendix A). Mature pine stands with ex isting RCW clusters ar e intensively managed, and other suitable stands are managed for future RCW occupancy. Pineland habitat management for the RCW at Fort Benning incl udes mechanical or, less often, chemical thinning on an approximate 10-yr rotation a nd prescribed burning on a 3-yr rotation. Thinning prescriptions change in response to new guidelines and obj ectives; currently, the approximate residual basal area is 11.5.1 m2/ha at loblolly and shortleaf stands and 18.4-20.7 m2/ha or greater at longleaf stands (D. Odom, pers. comm.). Study Design and Site Selection The approach of this comparative observational study (James and McCulloch 1995, K. Sieving, pers. comm.) is a direct spacefor-time substitution of sites across an anthropogenic gradient (Pickett 1989, Fukami and Wardle 2005) of time since forest thinning. This type of chronosequence appr oach assumes that the long-term temporal dynamics of a biological community may be inferred from the simultaneous study of multiple sites (Pickett 1989). As reco mmended by Fukami and Wardle (2005), I addressed the potentially confounding effects of unknow n site history by studying multiple species.
9 From thinning availability and stand i nventory records provided by the Fort Benning Natural Resource Management Bran ch (FBNR), I identified mechanically thinned pine stands > 4 ha with loamy sand or, occasionally, sandy loam surface soils. Study sites were then randomly selected, provided that each site was approximately 1 km from a previously selected site. Site sele ction was thus constrai ned by the size of the study area and by the geographic proximity to pr eviously selected sites. Both reduced adherence to strict randomi zation, and increased study si te interspersion, may be considered in order to maintain statistic al independence among study sites (Hurlbert 1984). In most sites, the dominant tree species was lobl olly, shortleaf ( P. echinata ), or longleaf pine. Four sites were classified as pine plantations that ranged in age from 41 to 46 yr. I classified stands into a three-group chronosequence according to the number of growing seasons since thinning (hereafter, gst). As in ot her studies (e.g., Bersier and Meyer 1994, Hobson and Schieck 1999, Purtauf et al. 2004), the temporal extent of chronosequence groups (< 2 gst, T1; 4-7 gst, T2; 10-17 gs t, T3) increased with greater time since thinning. In this study, categories vari ed in temporal extent in order to address sample size requirements (see below) and because of relatively low availability of sites of greater time after thinni ng. I re-classified site s (n = 2) that were thinned between winter and spring sampling dates. An initial power analysis for a univariate ANOVA design (http://euclid.psych .yorku.ca/cgi/power.pl) suggested a min imum of nine samples in each of six treatment groups (N=54) in order to detect a change of 0.75 standard deviation with = 0.05 and power ( ) = 0.808. However, I desired great er sample size in order to
10 record at least 60 observations of the species of interest on line tran sects (see below) and to obtain a sample to variable ratio (e.g., 5:1) recommended for multivariate statistical procedures. I therefore attemp ted to study 30 sites within ea ch of the three post-thinning groups. Logistical constrai nts precipitated unequal sampling among seasons; 86 stands were sampled during winter (22 January to 13 March) and 94 were sampled during spring (19 April to 29 May), 2004 (Figure 1). Avian Data I used line-transect distance sampling methodology (Buckland et al. 2001) to census the bird community. Point-trans ect methods are generally recommended for studying woodland birds (Buckland et al. 2001). However, the habitats studied here were open, and using line tran sects improved the detection of cryptic birds during the cold nonbreeding season. Birds such as sparrows and juncos are known to congregate and wander over large areas during wi nter (Pearson 1993). Accura te detection and estimation of flock size was assumed to be critical to reduce the variance in density estimates for these birds. Distance sampling requires the a ssumptions that all objects on the transect line are detected, objects do not move prior to detection, and distan ces and angles to objects are measured accurately (Buckland et al . 2001). Provided that transect lines are located randomly, distance sampling does not require the assumptions that objects are located randomly and are independently dist ributed (Buckland et al. 2001). Transect location and bearing were randomly determin ed with digital aerial photography provided courtesy of FBNR. Minor adjustments were ma de in the field to ensure that transects were at least 30 m from the stand edge a nd were not parallel to major topographic or anthropogenic features. Lichst ein et al. (2002) studied the spatial autoco rrelation of songbirds in the southern Appalachians and concluded that a distance of 500 m between
11 study sites was adequate to ensure the indepe ndence of the response variables. In this study, the minimum distance between transect s was 650 m; avian data among transects were assumed to be independent. Reliable estimation of density when us ing distance sampling methods requires adequate sample size (Buckla nd et al. 2001). To estimate the total transect length required to obtain an adequate number (~ 60) of observations for the median species, I conducted a pilot study in the summer of 2003 at 25 strip transects of 50-m half-width. After excluding species observed at fewer than 10 percent of sites, the mean encounter rate among 21 species was 7.7 individuals ove r 2.5 km. The tota l transect length ( L ) was estimated as 24.3 km using the followi ng formula (Buckland et al. 2001): L = ( b ) . ( L 0 ) cvt(D ) ( n0) where b = 3, the estimated coefficient of variation of density cvt(D ) = 0.2, the transect length L0 = 2.5 km, and the number of observations n0 = 7.7. A 0.3-km unlimited-width transect was randomly located as described abov e within each stand. The total transect length sampled was 25.8 km (86 x 0.3 km) and 28.2 km (94 x 0.3 km) in winter and spring, respectively. Transects were walked once per season at a nearly constant pace for 36 min, and were walked in the reverse direction if samp led in both seasons. All observations were made by one observer (JEA); data were record ed by a field assistant who did not assist with bird detection. Birds were detected visually and aurally. To improve the accuracy of detection angles and distances, I visually confirmed the location of birds that were initially detected aurally. The distance to in dividual birds or to the approximate center of clusters (flocks) of birds was estimat ed with a Bushnell Yardage Pro 500 laser
12 rangefinder. I recorded the species com position and number of birds in singleand mixed-species flocks, the sex and age class (j uvenile or adult) of each bird when known, and the height of detected bi rds in the vegetative strata (g round, shrub, midstory or canopy). Flyovers were not recorded. Distance sampling aims to account for vari ation in the detect ion of objects as a function of distance from the obs erver (Buckland et al. 2001). However, detectability is not only a function of observer -to-object distances (Ramse y and Harrison 2004). Recent advances in distance sampling allow the incl usion of covariates that may reduce the variance of density estimates by accounting for va riation in detectabil ity as a function of, for example, biotic and abiotic characteri stics of the sampling units or behavioral differences among animals (Marques and Bu ckland 2003). Transe ctor observationlevel covariates can be used in program DISTANCE to estimate animal density at each transect based on a detection function estimated at the global or stratum level (Thomas et al. 2005). Using the multiple covariate (MCD S) engine in progr am DISTANCE v. 5.0 (Thomas et al. 2005), I incorporat ed covariates describing hab itat features of transects, time since thinning, time since fire, the si ze and type of flocks (clusters), and the occurrence of males or females in an observati on (Table 1). Because parid vocalizations may attract non-parid birds to mixed-species flocks (Gaddis 1980) and, thus, may influence flock detectability, and because pr eliminary results revealed that Tufted Titmouse was the most frequently detected parid in this study , I also used the presence/absence of Tufted Titmouse in mixedspecies flocks as a covariate in detection function models. Cluster type and the presence of Tufted Titmouse were used to model detection functions for only winter birds b ecause mixed-species flocks were rare in
13 spring. Covariates were selected for each model using a forward selection approach (Marques and Buckland 2003). I estimated transect-level estimates of avian density (D ) for each species. Halfnormal and hazard-rate key series with cosi ne, simple polynomial, or hermite polynomial series adjustment terms were considered. To facilitate model convergence (Thomas et al. 2005), the selection of up to four series adju stment terms was automated. The required assumption that object detection is indepe ndent from the detection of other objects (Buckland et al. 2001) may not hold when obj ects occur in clusters. Because objects (birds), especially in winter, were detected in clusters (flocks), I tr uncated the outer five percent of observations for all species and used size-biased regr ession methods. Left truncation of observations with in 5 m of transects improved model fit for Pine Warbler and Blue-gray Gnatcatcher in spring. Infr equently observed speci es were pooled with other species assumed to have similar detec tion probabilities based on size, behavior, or taxonomy (Table 2). Pooling of species allo ws estimation of a global detection function f (0) for infrequently observed species. Tran sect-level density estimates for each pooled species were then obtained by post-stra tification at the species level. I selected models with low second-orde r (small-sample) Akaikeâ€™s Information Criterion (AICc), with p â€“value > 0.5 of the Cramr-von Mises (C-vM) with cosine weighting ( C2) goodness of fit test (Buckland et al . 2004), and with a re alistic effective strip width. Using AICc as a model selection criteria is preferable to AIC when n / K , the ratio of observations ( n ) to estimable parameters ( K ), is less than 40, and the consistent use of either AIC or AICc is recommended when multiple models are considered (Burnham and Anderson 1998).
14 Density was estimated for species observed at more than five percent of all sites (McCune and Grace 2002). I assigned species to resident and migrant categories (Appendix A) based on their pattern of seas onal occurrence in the study area and on their documented behavior in the region (Peterj ohn and Sauer 1993). Four-letter codes for common names of birds (Appendix A) follow Pyle and DeSante (2003). Plot Data Two vegetation sampling plots were random ly located within 25 m of each line transect (Figure 2). Vegeta tion structure and tree diversity (i.e., microhabitat) were sampled once per season at both plots, and data from both plots were averaged (Table 3). At plot centers I estimated canopy cover and ba sal area of pines and hardwoods. I used a laser rangefinder to estimate the numbe r of standing dead trees (snags) > 2 m tall within a 50 m radius. The majority of birds in Eastern North American forests nest up to 3 m (Preston and Norris 1947) and prescribed burning can have significant eff ects on breeding bird density in this stratum (Meyers and Johns on 1978). Consequently, and following Wilson et al. (1995), I defined woody vegetation < 3-m ht as a shrub, and > 3-m ht as a tree. Within 11.3 m of the plot center I identified all trees and estimated their diameter at breast height (dbh), crown hei ght, and height of lowest foliage. Following the USFWS (2003), I employed stand-specific definitions for canopy and midstory trees. Conner et al. (1991) used the lowest large living limb to estimate crown depth. Similarly, I considered the lowest large living limb to repr esent the low terminus of the live crown of each tree. I then used the mean low crown heig ht, calculated from the five tallest trees in each stand, as the criteria to distinguish canopy from midstory trees. Thus, canopy trees were those that were taller, and midstory tr ees were those that were shorter, than the
15 mean height of the lowest living limb of the five tallest trees in each stand. The identification of deciduous tree species that were problematic to identify in winter was verified during the growing season. Simps onâ€™s index of diversity (Magurran 1988) was calculated for trees at each stand. I used the line intercept method (Canfield 1941) to measure ground cover and shrub characteristics. At each plot I estimated the percent cover of grasses, forbs, fallen leaves, woody litter, and bare ground within the inne r and outer 10-m sections of a 30-m line intercept (Figure 2). When us ing line intercept methods, lit tle accuracy is gained when more than 30 m are sampled per site (Fl oyd and Anderson 1987); I sampled 20 m per plot and, thus, 40 m per stand. I measur ed the line intercept of shrubs < 1-m, from 1to 2-m, and from 2to 3-m in height. I define d shrub clumps as woody vegetation with an unbroken canopy (Floyd and Anderson 1987) < 3-m tall. I measured the intercept of shrub clumps > 0.20-m in length along each 30-m line (60 m per stand). The coefficient of variation (CV) based on lin e intercepts has been used to estimate the interspersion between shrub clumps (Peterson and Best 1985, Bersier and Meyer 1994). I used the CV of shrub intercepts to estimate the variation in shrub clump width. Two measures of vegetation structure, tota l basal area and understory profile, were used as covariates in bird density estimati on and were not used in species-environment analyses. Total basal area was the sum of the mean pine and hardwood basal area from both plots within a stand; hi gh basal area was expected to reduce bird detectability. Understory density was estimated using a 2-m tall profile board (Gut hery et al. 1981) at the 10and 30-m mark of each line intercept (Figure 2). From the understory density data, and following Wiens and Rotenberry (198 5), I then calculated a diversity index for
16 the understory profile; a higher profile diversity score indicated a greater amount of vegetation in the upper strata, which would be expected to reduce bird detectability. Stand Data I derived a total of four metrics to describe the physical characteristics and management history of each stand (Table 3). The age and area (ha) of each stand were obtained from forest invent ory data provided by FBNR. I used Patch Analyst in ArcView 3.2 (ESRI Inc. 1999) to estimate sta nd perimeter/area ratio and core area. The perimeter/area ratio describes stand shape; a low ratio indicates a small amount of edge relative to stand area as in a large stand with a smooth boundary, whereas a high ratio indicates that the area of the stand is small or that the st and has a tortuous boundary and a high amount of edge. Core area estimates the stand area (ha) further than 50 m from the stand edge. Stand edge was not included because it is ofte n confounded with the effects of stand area and generally shows weak e ffects in studies of songbirds in forested landscapes (Parker et al. 2005). Landscape Data Digitized land cover type, prescribed fire history, and road data were provided by FBNR, and data on the location and type of military training areas were provided by the Integrated Training Area Management (ITAM ) program. Using Xtools in ArcView 3.2 (ESRI Inc.1999), I created a buffer of 1-km radius around each transect (~367 ha). Within each buffer I derived 12 metrics to describe landscape composition and disturbance history (Table 3). Included were the extent of roads and trails, six variables describing the proportions of vegetated and military land covers, Simpsonâ€™s indices of diversity (1D ) and evenness of the cover types (e .g., Saab 1999), and the proportion of
17 land within the landscape that burned either < 1 or > 1 growing season prior to study (Table 3). Examination of the relationships between avian communities and the composition of habitat features within the surrounding landscape introduces several matters that require discussion; here, I addres s five that are especially rele vant to this study. First, a buffer of 1 km around transects may represent an area that is smalle r than the perceived landscape of some species (Bennett et al. 2004). The importance of habitat features within 1 km and smaller distances for describing avian community composition or abundances has been demonstrated [500 m: Pearson (1993), Graham and Blake (2001); < 800 m: Ribic and Sample (2001); 1 km: Knic k and Rotenberry (1995), Drapeau et al. (2000), Kluza et al. (2000), Rodewald and Yahner (2001a), Hagan and Meehan (2002), Mitchell et al. (2006)]. Rode wald and Yahner (2001a) justif ied the approach by stating that neotropical migrant breed ing territories may be < 2 ha in size, suggesting that territories of these birds are small relative to the area encompassed by a 1-km buffer. Second, setting a uniform area of extent (1 km) as the landscape for all species of interest was arbitrary and increased the risk of mi ssing important patterns (Wiens 1989, Addicott et al. 1987). Given the multivaria te structure of the analyses used in this study, it was not practical to determine the species-specific sc ale of response for each landscape variable. Next, collinearity among subsets of variables and confounding effects of nonindependence of habitat structures at multiple scales can be problematic (Cushman and McGarigal 2004a). However, these limita tions were minimized by using variation partitioning methods that explic itly identify the joint and independent influences of variable subsets. Fourth, in this study there were 63 areas of buffer overlap. Buffer
18 overlap may introduce pseudoreplication and autocorrelation. However, nonindependence and autocorrelation were assumed to have little influence on the analyses given that landscape characteristics were inde pendent variables (see Bennett et al. 2004, M. Christman pers. comm.). Therefore, and following McMast er et al. (2005), overlapping buffers were treated as separate sampling units. Finally, land cover type and management history within the buffers but outside the Fort Benning boundary were not described in the digitized forest inventory data. As in similar research (Mitchell et al. 2006), cover type was interpreted using aeria l photography, and thinning and prescribed fire at these areas were assumed to have not occurred within 17 and 3 growing seasons before study, respectively. Analysis The goals of the analysis were to pa rtition the variation in the breeding and nonbreeding bird community data among the plot -, stand-, and landscape-level subsets of environmental data, and to examine if the va riation explained by th e environmental data was influenced by time since disturbance and by migratory strategy (Figure 3). In order to determine the appropriate ordinati on method, I used Detrended Canonical Correspondence Analysis (DCCA) in CANOCO (ter Braak and milauer 1998) to examine the gradient lengths of the bird community density data (Lep and milauer 2003). The gradient lengths of the bird data constrained by the plot-l evel data were 1.77 and 1.56 for winter and spring, respectively. Because both gradient lengths were < 3, a linear method, Redundancy Analysis (RDA), wa s appropriate (Lep and milauer 2003). Results from Euclidean-based ordinati on methods, such as RDA, are more interpretable and can be compared more eas ily with environmental factors when the species data are first transformed into a dissimilarity matrix, especially when the data
19 contain many zeros (Legendre and Legendre 1998). I applied the Hellinger transformation, using the program Transformations (http://www.bio.montreal.ca/legendre ), in order to preserve the distances between species and to downweight the importance of rare species (Legendre and Gallagher 2001). Species data were then centered in CANOCO , and results were symmetrically scaled. Though linear ordination methods require multivariate normality of independent variables, data transformations are not requir ed prior to using RDA because the statistical significance of independent vari ables is tested with randomization procedures. The raw environmental variables were automatically standardized by CANOCO to have mean of zero and unit variance. Borcard et al. (1992) developed a method fo r partialling the variation in species data among two subsets of explanatory va riables. Subsequently, the method was extended to three subsets by A nderson and Gribble (1998), to n subsets by kland (2003), and Cushman and McGa rigal (2004b) applied the method to partialling the variation in bird data among s ubsets of variables representi ng plot, patch, and landscape scales. In this study, variat ion partitioning procedures were carried out in three levels (Figure 3). First, the bird community density data from all sites were partitioned among the three subsets of environmental variables. Species that occurred in at least five percent of the sites were included. Second, the variat ion in the bird density data for each of the three post-thinning chronosequence groups was partitioned among the subsets of environmental data. In this second level, species were standardized across the three chronosequence groups (Appendix B). Prelimin ary variation partitioning analyses of raw and reduced sets of response variables (speci es data) demonstrated that the proportion of
20 variation explained decreased as the number of species included increased. Thus, defining homogeneous species groups across th e chronosequence isolated the effects of time since thinning and removed the poten tially confounding effects of analyzing different numbers of species. In the third level of analysis, variation partitioning was performed separately on the densities of mi gratory and resident species in each of the three post-thinning groups. Standardizati on of species within each functional group resulted in homogeneous migrant and resident bird communities across each chronosequential analysis (Appendix B). Be tween seasons, the number of species in communities differed. Likewise, the number of species in functional groups within and among seasons differed. Therefore, I compar ed the temporal trends in fractions of variation explained by environment data; because of differences in the number of species, the magnitudes of the fractions of variati on explained could not be compared between communities of different seasons or between functional groups. At each step of the analysis, subsets of environmental variables were subjected to an automated forward selection process in RDA. Variables with variance inflation factor (VIF) greater than 20 were excluded (Lep and milauer 2003), one at a time, and the forward selection procedure was run again. In this manner, variab les of low importance or variables that were collinear with others in the same subset were excluded (see Cushman and McGarigal 2004a) from the fina l RDA models. Preliminary examination of Spearman rank correlation coefficients ( SPSS Inc. 2003) demonstrated that two pairs of variables, Simpsonâ€™s indices of diversit y (SIDI) and evenness (SIEV) of cover types, and stand area (AREA) and core area (CORE), were very highly correlated. The VIFs of
21 SIEV and CORE were found to be higher th an the correlate of each, and these two variables were excluded fr om subsequent analyses. The significance of variables was test ed with 999 Monte Carlo unrestricted permutations. Significant variables ( p < 0.10) were retained and a final RDA model was performed using only these sign ificant variables. The signi ficance of the canonical axes (trace statistic) of the final RDA model wa s tested using a Mont e Carlo randomization procedure with 999 unrestr icted permutations. To assess the fraction of variation expl ained by each RDA model, I used the adjusted coefficient of determination, R2 adj, because it facilitates comparison among models that contain unequal num bers of predictor variables and samples (Legendre et al. 2005) and produces reliable estimates in analys es of Hellinger transformed data (PeresNeto et al. in revision). Th e conditional effects of a va riable subset indicate the proportion of variation explained by that subset after the effe cts of other variables have been removed, and the marginal effects indi cate the proportion of va riation explained in total by a variable subset (Cus hman and McGarigal 2004b). The R2 adj of the marginal and conditional effects of each subs et of environmental data we re determined by subtraction of fractions (Legendre a nd Legendre 1998). I extended the methods developed to subtract fractions of variation among two se ts of explanatory variables (Legendre and Legendre 1998) to three sets of explanatory variables (Figure 4). I then compared the conditional effects of plot, stand, and landscape variab les across the post-thinning chronosequence for entire bird assemblages (analysis level 2, Figure 3) and between migratory and resident functi onal groups (analysis level 3).
22 Positive values of overlapping fractions of variation (fractions [D] through [G]; Figure 4) indicated non-orthogonal correlati ons (confounding) between environmental variables from different spa tial scales (Legendre and Lege ndre 1998). Negative values of these fractions indicated that variables of different scales were correlated and had opposite effects on the species data; this may al so be viewed as one set of explanatory variables acting to suppress the effect of a nother set on the species data (Legendre and Legendre 1998). Because of high skew and non-normality of the bird density and environmental data, I used Mann-Whitney U tests in SPSS (S PSS Inc. 2003) to compare means of bird densities and environmenta l variables among the three post-thinning chronosequence groups. Setting = 0.05, the Bonferroni-corrected significance level for each pair-wise test was p < 0.017 (0.05/3).
23 Table 1. Description of covariates used in the estimation of detec tion functions and bird density using the Multiple Covariates engine (MCDS) in program DISTANCE. Covariate Description Area Area of stand; continuous variable Basal area Total basal area; continuous and 3 categorical variables were used Cluster size The number of individuals obser ved in a single-species flock; equal to 1 for an individual bird; continuous variable Cluster type I = individual bird, S = single-species flock, M = mixed-species flock; used only in winter analyses; categorical variable Fire Time since fire; 1 for < 1 growing season, 2 for 2-3 growing seasons; categorical variable Height Vegetation stratum of the obser ved bird(s); G = ground, S = shrub, M = midstory, C = canopy; categorical variable M_F 0_0 if sex was not determined, 0_1 if only females were observed, 1_0 if only males were observed, 1_1 if both sexes were observed; categorical variable Profile The diversity index of the shr ub cover; continuous and 3 categorical variables were used Thinning Time since thinning; 1 for < 2 years, 2 for 4-7 years, 3 for 10-17 years; categorical variable Total cluster The total number of birds in a flock; equal to 1 for an individual bird, equal to Cluster size for a single-sp ecies flock; continuous variable TUTI The presence of Tufted Titmouse in a mixed-species flock; used only in winter analyses; binomial variable
24 Table 2. Pools of species, the number of observations in each pool, and the number of observations of bird specie s that were detected infr equently at line transects during winter (Jan-Mar) and spring (Ap r-May) 2004, at Fort Benning U. S. Army Infantry Center, Georgia. The untruncated (raw) number of observations are reported here; the outer 5% of obs ervations were truncated in models analyzed using the Multiple Covari ates engine (MCDS) in program DISTANCE. Season Pool Name Observations in pool Species Codea Observations of species Winter Small Ground/Shrub Passerine 69 BACS 10 CHSP 20 COYE 6 DEJU 10 HOWR 6 WTSP 17 Large Woodpecker & Crow 101 AMCR 9 HAWO 10 NOFL 13 PIWO 15 RBWO 54 Small Woodpecker 105 DOWO 23 RCWO 19 WBNU 40 YBSA 23 Medium Finch 76 EATO 39 NOCA 37 Medium Passerine & Dove 58 EABL 18 EAPH 33 MODO 7 Small Arboreal Passerine 91 AMGO 8 BHVI 7 BRCR 6 RCKI 39 YRWA 31 Spring Small Ground/Shrub Passerine 74 BACS 35 CHSP 22 COYE 10 WEVI 7 Large Woodpecker 88 HAWO 7 NOFL 11 PIWO 19 RBWO 51 Small Woodpecker 83 DOWO 23 RCWO 24 WBNU 36 Medium Passerine & Dove 90 BHCO 23 EABL 12 MODO 15
25 Table 2. Continued. Season Pool Name Observations in pool Species Codea Observations of species Spring Medium Passerine & Dove, continued YBCH 40 Small Arboreal Passerine 77 AMGO 10 BLGR 10 REVI 21 RTHU 11 YTVI 17 YTWA 8 Corvids 49 AMCR 6 BLJA 43 a Species codes follow Pyle and DeSante (2003) and are descri bed in Appendix A.
26 Table 3. Descriptions and sources of plot-, stand-, a nd landscape-level environmental variables describing managed pine stands studied at Fort Benning U. S. Army Infantry Center, Georgia, during 2004. Variable code Description Method / data source Plot BARE Percent cover of bare gr ound Line intercept, 2 x 10 m FORB Percent cover of herbaceous ve getation Line intercept, 2 x 10 m GRAS Percent cover of grass-like ve getation Line intercept, 2 x 10 m LVS Percent cover of down leaves Line intercept, 2 x 10 m WDLT Percent cover of down woody litter Line intercept, 2 x 10 m WDY1 Percent cover of woody vegetation < 1m Line intercept, 2 x 10 m WDY2 Percent cover of woody vegeta tion 1-2 m Line intercept, 2 x 10 m WDY3 Percent cover of woody vegeta tion 2-3 m Line intercept, 2 x 10 m SHMN Mean shrub clump length Line intercept, 30 m SHCV Coefficient of variation of shrub clumps Line intercept, 30 m SNAG Density of snags 50 m radius circle CC Canopy cover Densiometer BAHW Basal area of hardwoods (deciduous trees) 10-factor prism BAPI Basal area of pines 10-factor prism HC Density of hardwood canopy trees 11.3 m radius circle HCHT Mean height of hardwood canopy trees 11.3 m radius circle HM Density of hardwood midstory trees 11.3 m radius circle HMHT Mean height of hardwood midstory trees 11.3 m radius circle PC Density of pine canopy trees 11.3 m radius circle PCHT Mean height of pine canopy trees 11.3 m radius circle PM Density of pine midstory trees 11.3 m radius circle PMHT Mean height of pine midstory trees 11.3 m radius circle SIMP Tree diversity, Simpson's index Magurran (1988) Stand AREA Stand area Digitized forest inventory data a PARA Stand perimeter to area ratio Digitized forest inventory data a CORE Stand core area Digitized forest inventory data a AGE Stand age Stand inventory data a Landscape (1-km radius buffer surrounding each 300 m transect) FY43 Percent cover burned < 1 growing season Digitized prescribed fire data a FY2B Percent cover burned > 1 growing season Digitized prescribed fire data a HWFO Percent cover of hardwood-dominated forest Digitized land cover data a MIL Percent cover of military training and altered areas Digitized land cover data a , b PINE Percent cover of pine-dominated forest Digitized land cover data a PPLA Percent cover of pine plan tation Digitized land cover data a SHRB Percent cover of early successional shrublands Digitized land cover data a WETL Percent cover of open and forested wetlands Digitized land cover data a SIDI Cover type diversity, Si mpson's index Magurran (1988)
27 Table 3. Continued. Variable code Description Method / data source SIEV Cover type evenness, Si mpson's index Magurran (1988) TRAL Length of vehicle trails Digitized land cover data a ROAD Length of paved and unpaved roads Digitized land cover data a a Obtained from FBNR, Fort Benning Na tural Resource Management Branch b Obtained from ITAM, Integrated Training Area Management program
28 O 0 6 1 0O 0 6 3 5C 0 1 0 4S0 2 0 6T 0 1 0 8T 0 2 0 3J 0 1 0 2J0 1 1 4J 0 3 0 5J 0 3 0 8L 0 4 0 8L 0 3 0 2D 1 4 1 1J 0 5 1 7D 1 2 0 8D 1 3 0 9J 0 6 1 3J0 6 1 9D 1 2 2 0D 1 7 1 0E 0 41 9T 04 0 8T 0 4 1 3E 0 3 0 5E 0 2 0 1T 03 0 5E 0 3 1 7E 0 7 2 3E 07 1 5D 0 9 0 9I 0 2 0 5G 0 6 0 3H 0 1 0 9G 0 6 1 8G 0 5 1 0G 0 7 0 5G 0 1 0 5G 0 1 2 1I 0 5 1 5I0 3 1 5I 0 4 03D 0 6 1 1D 0 6 0 8F 0 4 1 2F 0 4 0 4F 0 3 0 3K 1 7 0 2K 1 9 0 9K 2 1 1 1K 2 0 0 1K 20 1 0F 0 2 1 1F 0 2 0 5D 0 5 08D 0 30 8K 2 2 16D 1 5 0 5D 1 4 2 1D 0 2 0 9D 0 3 1 1K 2 3 1 2K 2 3 0 7K 0 1 0 8K 1 3 0 9K 1 3 1 6K 0 6 1 3K 0 5 0 6O 0 7 1 0O 0 7 0 6K 0 3 1 3O 0 9 0 7K 0 3 0 1O 0 8 1 5O 0 9 26O 0 8 02K 1 6 0 2L 0 3 1 1O 1 2 0 8O 1 3 2 1M 0 2 0 1O 1 4 21O 1 5 0 1O 0 1 0 1O 0 5 07O 0 5 2 2N 0 1 1 8O 0 2 0 8N 0 2 0 6M 0 8 3 5M 0717U 0 3 1 7U 0 30 5U 0 2 1 7U 0 1 0 7 10123456KilometersN O0 6 10O 0635C01 04S02 0 6T01 08T02 03J0 10 2J01 14J 0305J 0 308L0 408L0 30 2D1411J 0 517D120 8D 1 309J06 1 3 J06 19D 1220D 171 0E0 419T0408T04 13E03 05E020 1T03 05E0 317E0 72 3E07 15D 0 909I02 05G06 03H0 10 9G 0618G 051 0G 0705G0 1 05G012 1I0 5 15I03 15I04 03D0 611D06 08F 0 412F0 404F03 03K1 7 02K 1909K2 111K 2001K 201 0F0 211F 0205D0508D 0308K 2216D1 505D14 21D 0209D0311K231 2K 2307K0 10 8K130 9K 1316K0 6 13 K05 0 6O071 0O07 06 K0 31 3 O0 907K030 1O0 81 5O0926O0 802K160 2L031 1O 1 208 O 132 1M0201O14 21O15 0 1O0 1 01O05 07 O0 522N 0118O0 208N 0206M 0835M0 717U 031 7U 0305 U0 217U010 7 10123456KilometersN Georgia O 0 6 1 0O 0 6 3 5C 0 1 0 4S0 2 0 6T 0 1 0 8T 0 2 0 3J 0 1 0 2J0 1 1 4J 0 3 0 5J 0 3 0 8L 0 4 0 8L 0 3 0 2D 1 4 1 1J 0 5 1 7D 1 2 0 8D 1 3 0 9J 0 6 1 3J0 6 1 9D 1 2 2 0D 1 7 1 0E 0 41 9T 04 0 8T 0 4 1 3E 0 3 0 5E 0 2 0 1T 03 0 5E 0 3 1 7E 0 7 2 3E 07 1 5D 0 9 0 9I 0 2 0 5G 0 6 0 3H 0 1 0 9G 0 6 1 8G 0 5 1 0G 0 7 0 5G 0 1 0 5G 0 1 2 1I 0 5 1 5I0 3 1 5I 0 4 03D 0 6 1 1D 0 6 0 8F 0 4 1 2F 0 4 0 4F 0 3 0 3K 1 7 0 2K 1 9 0 9K 2 1 1 1K 2 0 0 1K 20 1 0F 0 2 1 1F 0 2 0 5D 0 5 08D 0 30 8K 2 2 16D 1 5 0 5D 1 4 2 1D 0 2 0 9D 0 3 1 1K 2 3 1 2K 2 3 0 7K 0 1 0 8K 1 3 0 9K 1 3 1 6K 0 6 1 3K 0 5 0 6O 0 7 1 0O 0 7 0 6K 0 3 1 3O 0 9 0 7K 0 3 0 1O 0 8 1 5O 0 9 26O 0 8 02K 1 6 0 2L 0 3 1 1O 1 2 0 8O 1 3 2 1M 0 2 0 1O 1 4 21O 1 5 0 1O 0 1 0 1O 0 5 07O 0 5 2 2N 0 1 1 8O 0 2 0 8N 0 2 0 6M 0 8 3 5M 0717U 0 3 1 7U 0 30 5U 0 2 1 7U 0 1 0 7 10123456KilometersN O0 6 10O 0635C01 04 S02 0 6T01 08T02 03J0 10 2J01 14 J 0305J 0 308 L0 408 L0 30 2D1411J 0 517D120 8D 1 309J06 1 3 J06 19D 1220D 171 0E0 419 T0408T04 13 E03 05 E020 1T03 05E0 317 E0 72 3 E07 15 D 0 909I02 05 G06 03H0 10 9 G 0618G 051 0 G 0705 G0 1 05 G012 1 I0 5 15I03 15I04 03 D0 611D06 08F 0 412F0 404F03 03 K1 7 02 K 1909 K2 111K 2001K 201 0 F0 211 F 0205D0508D 0308K 2216D1 505 D14 21D 0209 D0311K231 2K 2307 K0 10 8 K130 9K 1316 K0 6 13 K05 0 6 O071 0O07 06 K0 31 3 O0 907 K030 1 O0 81 5O0926O0 802K160 2L031 1 O 1 208 O 132 1 M0201 O14 21O15 0 1O0 1 01O05 07 O0 522N 0118 O0 208N 0206 M 0835M0 717U 031 7 U 0305 U0 217 U010 7 10123456KilometersN Georgia Figure 1. Location of Fort Benning U.S. Army Infantry Center, Georgia (inset), and the locations of transects where bird communities and vegetation structure were studied during winter (Jan-Mar ) and spring (Apr-May), 2004.
29 Vegetation sampling plot Line transect p Ground, herbaceous, and shrub cover (2 x 10 m) Understoryprofile Shrub clump widths (30 m) Canopy cover, basal area (plot center)C Stand50 m (snags) 11.3 m (trees) 10 m segments pp C Vegetation sampling plot Line transect p Ground, herbaceous, and shrub cover (2 x 10 m) Understoryprofile Shrub clump widths (30 m) Canopy cover, basal area (plot center)C Vegetation sampling plot Line transect Vegetation sampling plot Vegetation sampling plot Vegetation sampling plot Line transect Line transect p Ground, herbaceous, and shrub cover (2 x 10 m) Understoryprofile Shrub clump widths (30 m) Canopy cover, basal area (plot center)C p Ground, herbaceous, and shrub cover (2 x 10 m) Understoryprofile Shrub clump widths (30 m) Canopy cover, basal area (plot center)C Stand50 m (snags) 11.3 m (trees) 10 m segments pp C Stand50 m (snags) 11.3 m (trees) 10 m segments pp C Stand50 m (snags) 11.3 m (trees) 10 m segments pp C Stand50 m (snags) 11.3 m (trees) 10 m segments pp C Figure 2. Plot-level vegetation sampling scheme. The centers of two plots were randomly locat ed within 25 m of each line tran sect. Canopy cover and basal area were recorded at the plot center. Attr ibutes of ground and shrub cover were measured along line intercepts. Tree and snag data were recorded within 11.3 and 50 m, respectively, of the plot center.
30 P L S S S S PP P LL LP SL P SL P SL P SL P SL P SLT1, < 2 growing seasons since thinning T2, 4-7 growing seasons since thinning T3, 10-17 growing seasons since thinning Analysis level 3; 3 subsets of sites, migratory and resident species analyzed separately Analysis level 2; 3 subsets of sites Analysis level 1; all sites P L S S S S PP P LL LP SL P SL P SL P SL P SL P SLT1, < 2 growing seasons since thinning T2, 4-7 growing seasons since thinning T3, 10-17 growing seasons since thinning Analysis level 3; 3 subsets of sites, migratory and resident species analyzed separately Analysis level 2; 3 subsets of sites Analysis level 1; all sites Figure 3. Schematic diagram of the variati on partitioning analyses performed separately for winter and spring bird communities. At the first level, the variation of all species at all sites is partitioned into three environmental components representing the plot (P), stand (S), and landscape (L) scales . In the second step, the variation of species within th ree post-thinning subsets of sites (T1, T2, and T3) is partitioned by three scales of environmental va riables. In the final step, the variation of migratory and resident bird communities of each post-thinning category is partitioned by three scales of environmental variables.
31 [A] [C] [B] [D] [G] [F] [E] [H] Plot Landscape Stand [A] [C] [B] [D] [G] [F] [E] [H] Plot Landscape Stand Plot conditional = [A] = ([A] + [B] + [C] + [D] + [E] + [F] + [G]) â€“ ([B ] + [C] + [D] + [E] + [F] + [G]) Stand conditional = [B] = ([A] + [B] + [C] + [D] + [E] + [F] + [G]) â€“ ([A] + [C] + [D] + [E] + [F] + [G]) Landscape conditional = [C] = ([A] + [B] + [C] + [D] + [E] + [F] + [G]) ([A] + [B] + [D] + [E] + [F] + [G]) [E] = ([B] + [C] + [D] + [E] + [F] + [G ]) â€“ ([C] + [D] + [F] + [G]) â€“ [B] [F] = ([A] + [C] + [D] + [E] + [F] + [G ]) â€“ ([A] + [D] + [E] + [G]) â€“ [C] [G] = ([A] + [B] + [D] + [E] + [F] + [G ]) â€“ ([B] + [D] + [E] + [F]) â€“ [A] [D] = ([A] + [B] + [C] + [D] + [E] + [F] + [G]) â€“ [A] â€“ [B] â€“ [C] â€“ [E] â€“ [F] â€“ [G] Figure 4. Schematic diagram and explanation of mathematical procedures for the calculation of conditional and marginal fractio ns of variance in bird community data ex plained by plot-, stand, and landscape-level environmental variables.
32 CHAPTER 3 RESULTS Microhabitat, Stand, and Landscape Composition Winter Data describing vegetation stru cture and tree diversity (mic rohabitat) were collected at two plots in each of 86 stands of thinned pine forest during Jan-Mar, 2004 (Table 4). Only two of the 23 measured and derived microhabitat variables were significantly different ( p < 0.017) among the three post-thinning chronosequence groups. The percent cover of shrubs and woody vegetation 1-2 m tall (WDY2) was lower at sites thinned < 2 growing seasons prior to this study (g st) than at sites thinned 4-7 gst ( P < 0.007) and sites thinned 10-17 gst ( P < 0.002). In contrast, the amount of down woody litter (WDLT) at the most recently thinned sites was hi gher than at sites thinned 4-7 gst ( P < 0.001) and at 10-17 gst sites ( P < 0.001). The amount of bare gr ound (BARE) was greatest at the most recently thinned sites but the difference was not significant. With the exception of pine canopy height, all estimates of tr ee density and height were gr eatest at sites thinned 10-17 gst; however, these trends were not significant. Characteristics of each pine stand and la ndscape composition within a 1-km buffer of each transect were derived from forest invent ory and digitized land cover data. No standlevel characteristic differe d significantly among the post-thinning groups. Of the 12 landscape composition variable s, only one was significantly different across the chronosequence. The proportion of area classi fied as military and altered land (MIL) was greatest around the sites thinned 10-17 gst ( P < 0.008).
33 Spring Microhabitat data were collected at two pl ots in 94 stands of thinned pine forest during Apr-May, 2004 (Table 5). Of the 23 me asured and derived pl ot-level variables, three differed significantly ( p < 0.017) among the three post-thinning chronosequence groups. Mean height of hardwood ca nopy trees (HCHT) was significantly ( P < 0.01) higher at sites thinned 4-7 gst than at the most recently thinned sites. Both the percent cover of grasses and gra ss-like vegetation (GRAS, P < 0.006) and canopy cover (CC, P < 0.007) were significantly highe r at sites thinned 10-17 gst than at sites thinned < 2 gst. As in winter, the amount of bare ground was greater at the most recently thinned sites, and most estimates of tree density and height tended to be highest at site s thinned 10-17 gst, but these differences were not significant. Standand landscape-level characteristi cs were mostly homogeneous across the chronosequence. The amount of milit ary and altered land was greatest ( P < 0.004) around the sites thinned 10-17 gst. No other landscap e-level characteristics differed significantly across the chronosequence, nor did any stand-level factor. Avian Community Composition and Density Winter A total of 2,076 individuals belonging to 47 species was observed on 86 line transects during Jan-Mar, 2004. I estimated the distance and angle to a total of 973 observations; 501 observations were of solitary birds and 472 were of singleor mixedspecies flocks (clusters). Composition of 66 mixed-species flocks ranged from 2 to 14 species. Sex was positively determined for 141 females and 361 males. Avian density was estimated using progr am DISTANCE for 21 permanent resident and 9 migratory species that occurred at more th an five percent of winter transects (Table
34 6). Of the 9 migrants, 8 were considered short-distance migrants and one, Blue-headed Vireo, was a neotropical migr ant (Peterjohn and Sauer 1993). Detection function models were fitted for five individual species and for six pools comprised of 25 infrequently observed species (Table 2). Of the 11 mode ls, optimal fit was obtained with the halfnormal key series with no adjustment terms in 8 (73%), the half-normal key series with second-order cosine adjustment terms in one, and the hazard-rate key series in two (18%; Table 7). Both basal area and total cluster size were c ovariates in 6 (55%) of the 11 selected models of bird density (Figure 5). Stand ar ea, shrub profile, and time since thinning were covariates in two (18%) models . The occurrence of Tufted Titmouse was a covariate in the detection function model of the Small Ground/ Shrub Passerine group (Figure 6), and the presence-absence ratio of males to females was a covariate in the Small Woodpecker group model (Table 7). Though Pine Warbler was observed most fr equently (202 observations), Chipping Sparrow was the most abundant species (454 indi viduals) and, based on density estimation, was also the densest species (2.47 per ha) among all sites. Density of all species combined was highest (0.39 0.16 birds/ha) in sites thinned 4-7 gst, but there were no significant differences among post-thinning gr oups (Table 6). Densities of individual species did not differ significantly ( p < 0.017) across the post-thinni ng chronosequence. Detection probability ( P ) ranged from 0.34 (Tufted Titmouse) to 0.64 (Pine Warbler). Mean cluster size ranged from 1.09 (Large Woodpecker and Crow) to 8.89 (Small Ground / Shrub Passerine) birds per cluster, and the mean of the 11 cluster size estimates was 2.2 birds per cluster.
35 Spring In spring, I recorded 1,751 observations co mprised of a total of 2,125 birds belonging to 58 species on 94 line transe cts during Apr-May, 2004. I identified 229 and 705 females and males, respectively. Most (1,462; 83%) observations were of individual birds. Composition of 10 mixed-species flocks ranged fr om two to five species. Blue Jay was detected more frequently during spring (n = 43) than winter (n = 4), and was therefore included in analyses of spri ng data but not for winter. Density was estimated for 22 resident and 15 migratory species that occurred at more than five percent of spring transects (Table 8). Two of the migratory species, Brownheaded Cowbird and Red-headed Woodpecker, we re short-distance migrants, and the other 13 were neotropical migrants (Peterjohn and Sauer 1993). As a result of data truncation, the density of American Crow was estimated for fewer than 5 % of all sites; American Crow was not used in variation partitioning analyses. Detection function models were fitted for 14 individual species and for six pool s of 23 infrequently observed species (Table 7). Of the 20 bird detection functions m odeled in program DISTANCE, optimal fit was obtained with the half-normal key series with no adjustment terms in 11 (55%), the halfnormal key series with second-order cosine ad justment term in 3 (15%), and the hazardrate key series with no adjustment terms in 6 (30%). Habitat factors and total cluster size were the most frequent c ovariates in the 20 selected models of breeding bird density (F igure 5). Shrub profile and basal area were included as covariates in nine (45%) and eight (40%) models, respectively, and total cluster size was a covariate in seven (35%) models. Stand area, bird height, and time since thinning were covariates in three (15%) models , and the presence-absence ratio of males to females was a covariate in the Larg e Woodpecker group model (Table 7).
36 Pine Warbler was observed most frequently (240 observations), was most numerous in terms of individuals detected (336), and wa s also the densest species (1.26 per ha) during the breeding season. Overall bird density was highest (0.26 0.05 birds /ha) at sites thinned within two growing seasons prior to study, but overall bi rd density did not significantly differ among postthinning groups (Table 8). Density of only two species differed significantly across the chronosequence. Red-h eaded Woodpecker density was significantly ( P < 0.005) higher at recently (< 2 gst) thinned sites than at sites thinned 1017 gst. Yellow-breasted Chats density was greatest ( P < 0.007) at sites thinned 4-7 gst. Detection probabilities ranged from 0.35 (R ed-headed Woodpecker) to 0.87 (Carolina Wren). Mean cluster size ranged from 1.02 (Eastern Wood-Pewee) to 1.85 (Brown-headed Nuthatch) birds per cluster, and the mean of the 20 cluster size estimates was 1.2 birds per cluster. Variation Partitioning The total amount of variati on in the bird community density data explained by environmental variables and the relative influence of spatial scale varied with time after thinning for entire bird communities and for mi gratory and resident bird functional groups (Figure 7). Microhabitat variables consistently explained more of the variation in bird communities than did standand landscape-level variables (Table 9). Breeding birds, as a whole, responded more strongly to landscap e composition than did nonbreeding birds; the relative importance of landscape composition to breeding birds increased with time after disturbance, mostly due to a marked increas e in the association between breeding season migratory birds and landscape composition with time after thinning. Stand-level variables were important predictors of community vari ation for breeding reside nt species, but were
37 mostly unimportant for residents in the nonbr eeding season and for migratory species in both seasons. Within spatial scales, significant variab les in the RDA sub-models varied among analysis levels, across the gradient of time since disturbance, and among resident and migratory bird functional groups (Table 10) . Between the two seasons, environmental variables between scales were more likely to be correlated and to s uppress the effects of other environmental variables in spring. Analysis Level 1 At the broadest level of analysis, where all sites and species were combined, environmental variables explained 6.28% and 9.66% of the variation in winter (nonbreeding) and spring (breeding) bird comm unities, respectively. A total of 30 and 36 bird species were included in winter and spring analyses, re spectively (Appendix B). In both seasons, plot-level variable s explained at least twice as much of the bird community variation than did landscape variables. In neither season were stand-level variables significantly related to bird communities. In the winter analysis, the negative value (-0.31 %) of fraction [D], where all three subsets of environmental variables overlap, co mbined with a value of zero of fraction [G], where plot and landscape variable s overlap (Table 9), suggested that the subset of stand variables, which did not have a significant marginal or cond itional effect, suppressed the influence of plot, landscape, or both of these se ts of variables. In the spring analysis, a positive value (0.42 %) of fraction [G] indicated that microhabitat and landscape composition variables were confounded. Among the level 1 analyses, the significan t microhabitat variables differed between seasons, whereas the sets of si gnificant landscape variables were relatively similar (Table
38 10). In the level 1 submodels of winter data , significant microhabitat variables included shrub cover, snag density, mean shrub clump width, and hardwood midstory height. In contrast, the significant microha bitat variables in the spring level 1 analysis included the cover of bare ground, grasses, and tall (2-3 m) shrubs, the density of hardwood and pine canopy trees, and the diversity of tree species. The extent of trails and the diversity of cover types within the 1-km landscape were significant landscape variables in winter level 1 sub-models. These same two variables, in addition to the propor tional area of hardwood forest, pine plantation, and land burned > 1 growing season before study, were also important predictors of spri ng bird community variation. Analysis Level 2 Environmental variables explai ned up to 12.9% and 17.6% of the variation in winter and spring bird communities, respectively (Table 9). The greatest am ount of variation was explained for the winter bird community at recently (< 2 gst) thinned sites, and for the breeding bird community at the least recently th inned (10 gst) sites. Level 2 analyses for winter and spring included 17 and 24 bird species, respectively (A ppendix B). In all models, the conditional effects of microhabi tat accounted for more of the explained variation than did stand charac teristics or landscape compos ition. Stand-level variables were not important in winter models, and were only slightly (up to 0.70% of the total variation) important in spring models. Lands cape composition was important in 2 of the 3 nonbreeding bird community models (up to 2.39%) and in all breeding bird models (up to 6.69%). The total amount of variation explained vari ed with time after disturbance (Figure 7a, b). In winter, the highest amount of varia tion explained was for th e bird communities of the most recently thinned (< 2 gst) sites. The opposite temp oral trend was true in spring;
39 the total amount of bird commun ity variation explained at si tes thinned 10-17 gst was 43% greater than at the most recently thinned sites. Considering only the conditional (independent) effects of plot , stand, and landscape factors from the spring data, total variation explained increased by 126% with time after thinning. Whereas the relative influence of stand-level fact ors decreased with time afte r disturbance, the relative influences of plot-level factors increased fr om 60% to 62%, and the relative influence of landscape factors increased from 31% to 38%. Confounding and suppression among variable s of different spatial scales was minimal in winter models, but occurred frequent ly in spring models (Table 9). The greatest fraction of confounding (5.19%) occurred betw een plot and landscap e variables in the spring T1 (< 2 gst) sub-model. The comparatively stronger relationshi ps between breeding birds and stand-level factors likely caused the confounding of st and-level factors with the scales below and above. The plotand landscape-level variables entere d into the final RDA models of winter data were different for each post-thinning group of sites. In contrast, plot(e.g. BARE, BAHW) and landscape-level (WETL) variables were used in fi nal models of spring data for at least two of the chronosequence groups. Overall, however, breeding birds responded to mostly disparate sets of va riables across the chronosequence. Analysis Level 3 The level 3 analyses of winter (nonbreedi ng season) bird functi onal groups included 5 short-distance migrants and 13 permanent resi dents (Appendix B). A total of 18 resident and 13 migratory species were included in the an alysis of breeding birds. Of the migratory breeding birds, Brown-headed Cowbird and Red-headed Woodpecker are short-distance migrants, and the other 11 are considered neot ropical migrants (Peter john and Sauer 1993).
40 The total and relative influence of habitat scales varied with time after disturbance for migrant and resident functi onal groups. When full models that included plot-, stand-, and landscape variables were considered, the total variation explained decreased for two functional groups and increased for the other tw o. The total amount of variation explained in the nonbreeding migrants decreased by 18% fr om the most recently thinned sites to the sites thinned 10-17 gst; total variation explai ned in the breeding residents decreased by 34.9% in the same temporal interval. Tota l variation explained in nonbreeding residents increased by 127% and increased in the br eeding migrants by 0.04%. These results contrast with the temporal variation in th e amount of explained variation when only the conditional effects of the environmental vari ables are examined. Considering again the breeding migrants, the sum of the conditional effects of microhabitat and landscape composition increased by 60% between recently thinned sites to si tes thinned 10-17 gst. Nonbreeding birds showed weak relationshi ps to stand-level variables. Among breeding birds, only the residents showed associa tions with stand-scale habitat factors. The relationships between breeding residents and stand-level factors caused the previously discussed relationship between the breeding bi rd community as a w hole with stand-level factors. The lack of response of breeding mi grants to stand factors was masked in the overall community level of analysis. Landscape composition was more important in spring than in winter for both functional groups, and was especially important for breeding residents. The combined conditional (independent) effects of stand and landscape factors were greater than the conditional effects of plot f actors for breeding residents at pine forests thinned < 2 and 4-7
41 gst. In fact, microhabitat factors were the l east important among the th ree scales of factors for resident breeding birds at sites thinned 4-7 gst. Examination of temporal variation in the c onditional effects of hab itat scales revealed contrasting patterns among the two functional groups of bree ding birds. The relative influence of microhabitat increased with time after disturbance for breeding residents, whereas the opposite was true for breeding migran ts (Figure 7). Similarly, the relative influence of landscape composition increased sh arply over time for breeding migrants but decreased for breeding residents. Significant variables changed over time fo r each functional group within each season (Table 10). Breeding migrants were more lik ely than other groups to respond to the same plot-level variables over time. Considering the plot-lev el RDA models for nonbreeding migrants, only the cover of woody litter (WDLT) was entered into more than one model. In the plot-level RDA models for breeding mi grants, the cover of ba re ground (BARE) and herbaceous plants (FORB) and the height of hardwood canopy trees (HCHT) were entered into more than one model. No plot-level va riable was entered into more than one model for nonbreeding residents, and BARE was the onl y variable entered into more than one model for breeding residents. Both FORB and HCHT were important variables for nonbreeding and breeding migrants at recently thinned (< 2 gst) sites, and both WDLT and SIMP were entered into models of nonbreeding and breeding migrants at sites thinned 4-7 gst. Because of the total tu rnover in the migrant functional groups between seasons, this finding suggests that these variab les are important indicators of habitat suitability for sites of a similar disturbance history for migratory birds in either season. Surprisingly, between the nonbreeding and breeding season s, resident species responded to disparate sets of plot-
42 level factors. Only BARE was included in models (3 of 6) of nonbreeding and breeding season resident birds. Compared with the RDA models of spri ng data, little confounding and suppression of variables among scales occurred in winter RDA models. Confounding (up to 3.97% of the total variation) occurred where all three scales overlapped in their influence. For winter migrants at recently (< 2 gst) disturbed sites, the marginal effect of landscape variables was zero. Examination of the winter migrant T1 (< 2 gst) sub-models showed that landscape variables were significan t only in models with plot factors (Table 10).
43 Table 4. Summary statistics of environmental variables for 86 pine stands at Fort Benning, Georgia, during winter (Jan-Mar), 2004. Means in the same row with different letters were signifi cantly different ( P < 0.017) in post-hoc Mann-Whitney U-tests between three post-thinni ng chronosequence groups. Variable Mean SE Code Units Min. Max.< 2 GSa 4-7 GS 10-17 GS Plot BARE % 0.83 61.2518.34 2.50 16.43 2.82 14.46 2.78 FORB % 0 27.008.11 0.93 8.55 1.16 9.37 1.23 GRAS % 0.75 51.9516.35 2.46 17.87 2.29 18.57 2.30 LVS % 12.38 79.3548.57 2.29 47.95 3.29 52.08 2.58 WDLT % 2.98 39.5821.7 1.46A 14.48 0.92B 13.77 0.92B WDY1 % 0.98 44.1813.15 1.61 15.15 1.81 14.84 1.84 WDY2 % 0 13.751.47 0.45A 3.57 0.69B 3.47 0.61B WDY3 % 0 16.680.78 0.56 0.85 0.33 1.04 0.34 SHMN m 0.22 6.831.16 0.23 1.62 0.23 1.41 0.24 SHCV 0 2.300.82 0.08 0.98 0.08 0.99 0.10 SNAG no./ha 1.27 23.568.34 0.92 7.55 0.93 8.48 0.86 CC % 0.14 0.570.30 0.02 0.33 0.02 0.35 0.02 BAHW m2/ha 0 16.262.17 0.48 1.97 0.49 2.90 0.63 BAPI m2/ha 2.32 26.7111.01 0.66 10.29 0.66 12.86 1.02 HC no./ha 0 199.5022.44 6.69 23.16 8.38 27.61 7.24 HCHT m 0 23.325.84 1.49 8.66 1.72 10.50 1.60 HM no./ha 0 486.28117.62 22.12 97.08 19.49 149.63 22.23 HMHT m 0 15.245.49 0.50 5.43 0.63 5.84 0.46 PC no./ha 24.94 436.41120.12 9.27 147.84 17.73 156.31 13.24 PCHT m 10.76 30.7120.93 0.76 20.08 0.85 19.50 0.75 PM no./ha 0 610.9755.28 11.18 56.56 13.7 140.72 32.00 PMHT m 0 19.616.70 0.77 6.88 0.95 7.28 0.75 SIMP 0-1 0 0.920.63 0.04 0.58 0.05 0.62 0.04 Stand AREA ha 4.36 67.6020.99 2.25 20.35 2.18 19.06 2.44 PARA m/ha 63.90 240.30140.69 8.33 143.85 8.16 146.51 8.48 CORE ha 0.59 47.109.98 1.66 9.17 1.41 8.92 1.82 AGE yrs 35.00 94.0063.57 2.81 65.07 3.21 58.89 3.46 Landscape FY43 % 0 100.0042.20 5.12 44.68 5.87 44.93 5.51 FY2B % 0 100.0057.80 5.12 55.32 5.87 55.07 5.51 HWFO % 0 39.096.99 0.96 6.99 1.25 9.54 1.95 MIL % 0 20.314.25 0.86 2.29 0.44A 5.41 1.03B PINE % 15.60 85.1558.23 2.64 59.47 2.76 52.35 2.90 PPLA % 0 52.1010.58 1.81 10.58 2.07 11.63 1.42 SHRB % 0 16.664.30 0.77 3.23 0.65 5.09 0.72 WETL % 0.66 40.6415.66 1.22 17.45 2.25 15.97 1.79 SIDI 0-1 0.26 0.850.61 0.03 0.61 0.02 0.68 0.02 SIEV 0-1 0.32 0.970.70 0.03 0.70 0.03 0.78 0.02 TRAL m/ha 6.85 48.5321.94 1.49 22.00 1.45 20.87 1.41 ROAD m/ha 0 20.799.43 0.71 7.64 1.20 7.08 0.54 a GS, growing seasons
44 Table 5. Summary statistics of environmental variables for 94 pine stands at Fort Benning, Georgia, during spring (Apr-May), 2004. Means in the same row with different letters were signifi cantly different ( P < 0.017) in post-hoc Mann-Whitney Utests between three post-th inning chronosequence groups. a GS, growing seasons Variable Mean SE Code Units Min. Max. < 2 GSa 4-7 GS 10-17 GS Plot BARE % 0.18 57.9826.37 2.9623.08 2.58 22.84 3.59 FORB % 2.43 39.4313.58 1.0314.83 1.43 14.87 1.96 GRAS % 0.65 44.0013.54 1.44A17.30 1.60 17.54 1.28B LVS % 7.00 71.4526.30 2.532.82 3.23 35.03 3.20 WDLT % 2.53 34.6015.49 1.1312.35 0.93 12.14 0.89 WDY1 % 1.35 62.6326.14 2.0828.39 2.50 28.08 3.05 WDY2 % 0 38.202.87 0.705.52 1.21 5.57 1.51 WDY3 % 0 12.900.44 0.210.89 0.25 1.61 0.56 SHMN m 0.10 3.850.48 0.050.67 0.08 0.75 0.16 SHCV 0 1.720.84 0.050.89 0.07 0.91 0.08 SNAG no./ha 0 33.106.98 1.057.45 0.92 8.00 0.89 CC % 17.58 87.8941.65 2.49A44.52 2.49 51.73 2.77B BAHW m2/ha 0 19.741.84 0.422.04 0.45 3.26 0.72 BAPI m2/ha 2.32 27.8710.97 0.6410.47 0.64 13.56 1.16 HC no./ha 0 211.9720.31 5.8722.93 6.72 32.51 8.78 HCHT m 0 25.9114.89 1.07A17.49 0.49B 16.30 1.06 HM no./ha 0 523.69104.74 16.72121.07 20.17 136.27 15.43 HMHT m 0 10.366.39 0.325.68 0.27 6.06 0.27 PC no./ha 24.94 486.28122.91 10.8142.79 14.93 166.99 15.96 PCHT m 10.65 30.7120.82 0.6920.58 0.72 19.55 0.85 PM no./ha 0 536.1654.51 9.5254.70 12.74 148.74 29.87 PMHT m 0 19.618.76 0.459.09 0.81 8.48 0.65 SIMP 0-1 0 0.940.62 0.040.62 0.04 0.63 0.04 Stand AREA ha 4.36 67.6021.03 1.9921.41 2.13 20.34 2.55 PARA m/ha 63.90 240.30137.49 7.43143.17 7.84 143.79 8.91 CORE ha 0.59 47.1010.13 1.479.76 1.39 9.87 1.95 AGE yr 35.00 94.0063.57 2.5664.19 2.97 57.29 3.35 Landscape FY43 % 0 100.0070.00 4.3571.45 5.04 80.89 5.08 FY2B % 0 100.0030.00 4.3527.58 5.13 19.11 5.08 HWFO % 0 39.097.05 0.858.03 1.49 9.64 1.95 MIL % 0 20.323.93 0.752.15 0.41A 5.42 1.03B PINE % 15.60 85.2358.86 2.3656.24 3.10 51.75 2.85 PPLA % 0 52.1010.35 1.5811.62 1.99 11.55 1.42 SHRB % 0 21.964.40 0.694.13 0.91 4.84 0.71 WETL % 0.66 45.9715.41 1.2517.83 2.28 16.80 1.75 SIDI 0-1 0.26 0.850.61 0.020.63 0.02 0.68 0.02 SIEV 0-1 0.32 0.970.70 0.030.72 0.03 0.78 0.02 TRAL m/ha 6.85 49.1422.45 1.3921.88 1.41 21.07 1.62 ROAD m/ha 0 20.799.52 0.657.26 1.15 7.09 0.54
45 Table 6. Densities (no./ha) of 30 bird species observed at line transects at Fort Benning, Georgia, during winter (Jan-Mar), 2004. Densities are presented for three postthinning chronosequence categories. GS stands for growing seasons. Bird density (mean SE) Common name (Scientific name) < 2 GS 4-7 GS 10-17 GS Resident American Crow ( Corvus brachyrhynchos ) 0.04 0.040.05 0.04 0.04 0.03 American Goldfinch ( Carduelis tristis ) 0.11 0.050.05 0.03 0.11 0.11 Bachmanâ€™s Sparrow ( Aimophila aestiva ) 0.10 0.050 0 0.05 0.03 Brown-headed Nuthatch ( Sitta pusilla ) 0.41 0.090.39 0.09 0.18 0.06 Carolina Chickadee ( Poecile carolinensis ) 0.25 0.070.33 0.14 0.48 0.18 Carolina Wren ( Thryothorus ludovicianus ) 0.30 0.120.45 0.16 0.27 0.05 Chipping Sparrow ( Spizella passerina ) 1.11 0.714.61 2.14 1.79 0.99 Common Yellowthroat ( Geothlypis trichas ) 0.02 0.020.07 0.05 0.04 0.03 Downy Woodpecker ( Picoides pubescens ) 0.16 0.060.08 0.03 0.17 0.09 Eastern Bluebird ( Sialia sialis ) 0.14 0.060.18 0.11 0.04 0.03 Eastern Towhee ( Pipilo erythrophthalmus ) 0.54 0.340.82 0.41 0.57 0.24 Hairy Woodpecker ( Picoides villosus ) 0.04 0.020.02 0.01 0.02 0.01 Mourning Dove ( Zenaida macroura ) 0.08 0.070.02 0.01 0.02 0.02 Northern Cardinal ( Cardinalis cardinalis ) 0.19 0.080.54 0.18 0.26 0.10 Northern Flicker ( Colaptes auratus ) 0.07 0.030.02 0.01 0.04 0.02 Pileated Woodpecker ( Dryocopus pileatus ) 0.06 0.020.03 0.01 0.03 0.02 Pine Warbler ( Dendroica pinus ) 1.07 0.151.68 0.78 1.32 0.15 Red-bellied Woodpecker ( Melanerpes carolinus ) 0.19 0.040.23 0.09 0.18 0.05 Red-cockaded Woodpecker ( Picoides borealis ) 0.12 0.050.06 0.03 0.07 0.05 Tufted Titmouse ( Baeolophus bicolor ) 0.52 0.120.76 0.42 0.55 0.12 White-breasted Nuthatch ( Sitta carolinensis ) 0.23 0.070.15 0.05 0.38 0.17 Migrant Blue-headed Vireo ( Vireo solitarius ) 0.05 0.040.01 0.01 0.06 0.03 Brown Creeper ( Certhia americana ) 0.02 0.020.01 0.01 0.09 0.04 Dark-eyed Junco ( Junco hyemalis ) 0.52 0.260.15 0.09 0.10 0.08 Eastern Phoebe ( Sayornis phoebe ) 0.13 0.060.08 0.02 0.12 0.04 House Wren ( Troglodytes aedon ) 0.02 0.020.04 0.03 0.04 0.03 Ruby-crowned Kinglet ( Regulus calendula ) 0.31 0.110.14 0.03 0.30 0.08 White-throated Sparrow ( Zonotrichia albicollis ) 0.53 0.290.53 0.39 0.46 0.19 Yellow-bellied Sapsucker ( Sphyrapicus varius ) 0.08 0.030.12 0.05 0.11 0.05 Yellow-rumped Warbler ( Dendroica coronata ) 0.29 0.110.20 0.10 0.24 0.07 Overall density 0.26 0.050.39 0.16 0.27 0.07
46Table 7. Analyses and results of models generated in program DISTANCE for individua l species and pools of infrequently detecte d species of birds. Observations were made on line transects during winter (n = 86) and spring (n = 94), 2004, at Fort Benning, Georgia. Covariates are presente d in the order of (f orward) selection. Analysis Parameters Obsa AICc Pb ESWc (m) Key series Adjustment terms and order Mean cluster size Cluster size CV C2 d Covariates Winter, individual species Brown-headed Nuthatch 4 53474.750.52 52.97 H-Ne 0 1.74 0.08 0.5 Basal area, profile Carolina Chickadee 4 44386.200.55 49.30 H-N 0 1.84 0.06 0.8 Total cluster, basal area Carolina Wren 2 60521.370.59 49.24 H-N 0 1.37 0.05 0.6 Total cluster Pine Warbler 4 1921671.880.64 55.64 H-R 0 1.57 0.07 0.7 Total cluster, fire Tufted Titmouse 5 63545.730.34 29.53 H-N 0 1.92 0.08 0.8 Height, basal area Winter, pooled species Large Woodpecker and Crow 4 96895.430.56 67.16 H-N 0 1.09 0.03 0.6 Profile, total cluster, area Medium Finch 5 72591.970.52 42.34 H-R 0 2. 79 0.15 0.6 Cluster type, total cluster Medium Passerine and Dove 4 55479.780.63 52.03 H-N 0 1.23 0.08 0.9 Thinning, total cluster Small Arboreal Passerine 6 87737.090.39 34.36 H-N Cosine, 2 1.49 0.08 0.5 Thinning, basal area Small Ground/Shrub Passerine 3 66531.260.43 31.85 H-N 0 8.89 0.24 0.7 Basal area, TUTI Small Woodpecker 6 100873.670.46 44.85 H-N 0 1.30 0.04 0.9 Basal area, area, M_F Spring, individual species Blue-gray Gnatcatcher 4 96739.270.44 26.99 H-N 0 1.20 0.04 0.7 Total cluster, basal area Brown-headed Nuthatch 5 55455.700.45 29.99 H-R 0 1.85 0.07 0.9 Profile, thinning Carolina Chickadee 4 42324.990.51 30.88 H-R 0 1.40 0.08 0.8 Total cluster, profile Carolina Wren 3 57467.330.87 54.02 H-N Cosine, 2 1.21 0.07 0.6 Area Eastern Towhee 2 63501.350.75 40.54 H-N 0 1.06 0.03 0.7 Profile
47Table 7. Continued. Analysis Parameters Obsa AICc Pb ESWc (m) Key series Adjustment terms and order Mean cluster size Cluster size CV C2 d Covariates Eastern WoodPewee 6 81660.840.43 38.61 H-R 0 1.02 0.02 0.8 Height, thinning Great Crested Flycatcher 4 62505.640.66 39.97 H-N 0 1.11 0.04 0.7 Profile, basal area, total cluster Indigo Bunting 3 1381099.380.66 37.78 H-N 0 1.09 0.02 0.9 Total cluster, basal area Northern Cardinal 4 68550.630.54 33.06 H-R 0 1.18 0.05 0.9 Fire, basal area Pine Warbler 4 1881452.510.67 36.54 H-N 0 1.37 0.04 0.7 Total cluster, area, fire Prairie Warbler 2 59455.440.60 31.86 H-N 0 1.02 0.02 0.6 Area Red-headed Woodpecker 5 78678.210.35 32.97 H-R 0 1.19 0.04 0.9 Profile, thinning Summer Tanager 4 73590.760.75 43.21 H-N 0 1.11 0.03 0.6 Profile, basal area Tufted Titmouse 4 61507.590.60 43.39 H-N 0 1.20 0.05 0.5 Total cluster, profile, basal area Spring, pooled species Corvids 2 47396.950.47 41.29 H-N 0 1.19 0.05 0.6 Total cluster Large Woodpecker 4 84732.350.67 55.59 H-N 0 1.06 0.02 0.6 M_F Medium Passerine and Dove 5 85707.690.83 58.44 H-N Cosine, 2 1.16 0.05 0.6 Basal area, profile Small Arboreal Passerine 4 73608.660.75 58.40 H-N Cosine, 2 1.12 0.03 0.9 Profile Small Ground/Shrub Passerine 4 70531.600.43 26.36 H-N 0 1.21 0.05 0.9 Height Small Woodpecker 4 79658.050.63 46.08 H-R 0 1.24 0.05 0.6 Height, basal area a Obs, number of untruncated (raw) observation of species or pool b P , detection probability c ESW, effective strip width d C2, lower limit of the range of p -values for the Cramr-von Mises goodness of fit test e Key series abbreviations are as follows: H-N, half-normal key series; H-R, hazard rate key series
48 Table 8. Densities (no./ha) of 37 birds observe d at line transects at Fort Benning, Georgia, during spring (Apr-May), 2004. Densities in the same row with different letters were significantly different (P < 0.017) in Mann-Whitney U-tests between three post-thinning chronosequence categories . GS stands for growing seasons. Bird density (mean SE) Common name (scientific name) < 2 GS 4-7 GS 10-17 GS Resident American Crow ( Corvus brachyrhynchos ) 0 00.03 0.02 0.05 0.04 American Goldfinch ( Carduelis tristis ) 0.07 0.040.07 0.04 0.05 0.03 Bachman's Sparrow ( Aimophila aestivalis ) 0.33 0.100.14 0.05 0.26 0.11 Blue Jay ( Cyanocitta cristata ) 0.20 0.080.33 0.08 0.08 0.03 Brown-headed Nuthatch ( Sitta pusilla ) 0.76 0.210.61 0.14 0.33 0.12 Carolina Chickadee ( Poecile carolinensis ) 0.26 0.080.26 0.09 0.36 0.09 Carolina Wren ( Thryothorus ludovicianus ) 0.21 0.080.23 0.06 0.16 0.04 Chipping Sparrow ( Spizella passerina ) 0.18 0.080.13 0.08 0.29 0.16 Common Yellowthroat ( Geothlypis trichas ) 0.08 0.050.11 0.05 0 0 Downy Woodpecker ( Picoides pubescens ) 0.10 0.040.15 0.06 0.11 0.05 Eastern Bluebird ( Sialia sialis ) 0.08 0.030.01 0.01 0.01 0.01 Eastern Towhee ( Pipilo erythrophthalmus ) 0.21 0.070.43 0.10 0.25 0.05 Hairy Woodpecker ( Picoides villosus ) 0.02 0.010.02 0.02 0.03 0.02 Mourning Dove ( Zenaida macroura ) 0.02 0.010.03 0.01 0.10 0.05 Northern Cardinal ( Cardinalis cardinalis ) 0.56 0.160.49 0.13 0.27 0.07 Northern Flicker ( Colaptes auratus ) 0.06 0.020.02 0.01 0.01 0.01 Pileated Woodpecker ( Dryocopus pileatus ) 0.07 0.020.05 0.02 0.03 0.01 Pine Warbler ( Dendroica pinus ) 1.45 0.321.29 0.22 0.99 0.16 Red-bellied Woodpecker ( Melanerpes carolinus ) 0.17 0.040.18 0.04 0.16 0.05 Red-cockaded Woodpecker ( Picoides borealis ) 0.18 0.060.04 0.02 0.06 0.04 Tufted Titmouse ( Baeolophus bicolor ) 0.36 0.060.31 0.09 0.36 0.10 White-breasted Nuthatch ( Sitta carolinensis ) 0.21 0.060.18 0.05 0.09 0.03 Migrant Blue Grosbeak ( Guiraca caerulea ) 0.03 0.020.04 0.02 0.03 0.02 Blue-gray Gnatcatcher ( Polioptila caerulea ) 0.78 0.140.78 0.14 0.61 0.12 Brown-headed Cowbird ( Molothrus ater ) 0.11 0.050.12 0.06 0.06 0.04 Eastern Wood-pewee ( Contopus virens ) 0.38 0.060.45 0.08 0.30 0.08 Great Crested Flycatcher ( Myiarchus crinitus ) 0.49 0.200.19 0.05 0.27 0.10 Indigo Bunting ( Passerina cyanea ) 0.57 0.090.79 0.11 0.71 0.14 Prairie Warbler ( Dendroica discolor ) 0.26 0.100.47 0.17 0.27 0.08 Red-eyed Vireo ( Vireo olivaceus ) 0.09 0.030.09 0.04 0.11 0.05 Red-headed Woodpecker ( Melanerpes erythrocephalus ) 0.81 0.17A0.39 0.09 0.22 0.07B Ruby-throated Hummingbird ( Archilochus colubris ) 0.04 0.020.09 0.04 0.06 0.04Summer Tanager ( Piranga rubra ) 0.33 0.050.29 0.07 0.43 0.11 White-eyed Vireo ( Vireo griseus ) 0.09 0.050.06 0.05 0 0 Yellow-breasted Chat ( Icteria virens ) 0.08 0.030.20 0.06A 0.09 0.07B Yellow-throated Vireo ( Vireo flavifrons ) 0.10 0.030.06 0.03 0.06 0.03 Yellow-throated Warbler ( Dendroica dominica ) 0.02 0.020.06 0.04 0.03 0.03 Overall density 0.26 0.050.25 0.04 0.20 0.04
49 Table 9. Adjusted coefficients of determination ( R2 adj) from the partitioning of variation of winter (nonbreeding) and spring (breedi ng) bird communities among plot-, stand-, and landscape-level environmental variables using RDA in CANOCO. Results were converted to percentages by multiplying by 100. [A]a [B] [C] [D] [E] [F][G] Unexplained Winter Analysis level 1 4.39 02.20-0.31000 93.72 Analysis level 2 T1 (< 2 GS)b 12.88 000000 87.12 T2 (4-7 GS) 3.10 01.960.01000 94.93 T3 (10-17 GS) 6.04 02.390000 91.58 Analysis level 3 T1, Migrants 34.17 08.93-8.93000 65.83 T1, Residents 4.40 000000 95.60 T2, Migrants 29.83 6.90000-6.900 70.17 T2, Residents 3.66 02.520.07000 93.74 T3, Migrants 23.67 00.273.97000 72.08 T3, Residents 7.10 002.90000 90.00 Spring Analysis level 1 6.79 02.440000.42 90.34 Analysis level 2 T1 (< 2 GS) 4.65 0.702.46-1.721.72-0.705.19 87.70 T2 (4-7 GS) 8.75 0.375.01-2.082.041.49-0.22 84.63 T3 (10-17 GS) 10.96 06.692.51-2.510-0.06 82.41 Analysis level 3 T1, Migrants 21.19 00-1.9201.925.35 73.46 T1, Residents 10.20 6.368.826.30-6.36-6.30-4.02 85.00 T2, Migrants 16.60 01.250002.03 80.13 T2, Residents 2.99 5.093.910.50-0.99-2.366.61 84.24 T3, Migrants 18.72 015.22000-6.33 72.39 T3, Residents 5.79 2.012.225.21-5.21-2.011.76 90.23 a Column headings represent conditional and sh ared effects of plot, stand, and landscape environmental variables as described in â€˜Methodsâ€™ and in Figure 4. b GS, growing seasons since thinning.
50 Table 10. Significant ( p < 0.10) environmental variables used in final RDA models and the significance level of trace st atistics determined by 999 random unrestricted permutations. The number of sites included in RDA models are given in parentheses. Analysis Sub-model and significant variables Trace significance (p-value) Winter Analysis level 1 (n = 86) Plot: WDY2, SNAG, HMHT, SHMN 0.001 Stand: Landscape: TRAL, SIDI 0.013 Plot-Stand: WDY2, SNAG, SHMN, HMHT 0.001 Plot-Landscape: WDY2, TRAL, SNAG, SHMN, HMHT, SIDI 0.001 Stand-Landscape: TRAL, SIDI 0.006 All: WDY2, SNAG, TRAL, SHMN, HMHT, SIDI 0.001 Analysis level 2 Group T1, < 2 growing seasons since thinning (n = 30) Plot: SNAG, WDY2, BARE, HM, PMHT 0.001 Stand: Landscape: Plot-Stand: SNAG, WDY2, BARE, HM, PMHT 0.001 Plot-Landscape: SNAG, WDY2, BARE, HM, PMHT 0.001 Stand-Landscape: All: SNAG, WDY2, BARE, HM, PMHT 0.001 Group T2, 4-7 growing seasons since thinning (n = 28) Plot: GRAS 0.058 Stand: Landscape: SIDI 0.106 Plot-Stand: GRAS 0.058 Plot-Landscape: GRAS, SIDI 0.035 Stand-Landscape: SIDI 0.106 All: GRAS, SIDI 0.035 Group T3, 10-17 growing seasons since thinning (n = 28) Plot: CC, WDY3 0.017 Stand: Landscape: PPLA 0.088 Plot-Stand: CC, WDY3 0.017 Plot-Landscape: CC, WDY3, TRAL 0.007 Stand-Landscape: PPLA 0.088 All: CC, WDY3, TRAL 0.007 Analysis level 3 Group T1, migrants only (n = 19) Plot: FORB, SNAG, HCHT 0.002
51 Table 10. Continued. Analysis Sub-model and significant variables Trace significance (p-value) Stand: Landscape: Plot-Stand: FORB, SNAG, HCHT 0.002 Plot-Landscape: FORB, SNAG, HCHT, HWFO 0.001 Stand-Landscape: All: FORB, SNAG, HCHT, HWFO 0.001 Group T1, residents only (n = 30) Plot: BARE 0.011 Stand: Landscape: Plot-Stand: BARE 0.011 Plot-Landscape: BARE 0.011 Stand-Landscape: All: BARE 0.011 Group T2, migrants only (n = 20) Plot: WDLT, SIMP, BAHW 0.001 Stand: Landscape: Plot-Stand: WDLT, SIMP, BAHW, AGE 0.001 Plot-Landscape: WDLT, SIMP, BAHW 0.001 Stand-Landscape: All: WDLT, SIMP, BAHW, AGE 0.001 Group T2, residents only (n = 28) Plot: GRAS 0.048 Stand: Landscape: SIDI 0.094 Plot-Stand: GRAS 0.048 Plot-Landscape: GRAS, SIDI 0.024 Stand-Landscape: SIDI 0.094 All: GRAS, SIDI 0.024 Group T3, migrants only (n = 23) Plot: SHMN, WDY2, PM, WDLT 0.001 Stand: Landscape: HWFO 0.094 Plot-Stand: SHMN, WDY2, PM, WDLT, HWFO 0.001 Plot-Landscape: SHMN, WDY2, PM, WDLT 0.001 Stand-Landscape: HWFO 0.094 All: SHMN, WDY2, PM, WDLT, HWFO 0.001 Group T3, residents only (n = 28) Plot: CC, WDY3, BAPI 0.008 Stand:
52 Table 10. Continued. Analysis Sub-model and significant variables Trace significance (p-value) Landscape: SHRB 0.068 Plot-Stand: CC, WDY3, BAP; 0.008 Plot-Landscape: CC, WDY3, BAPI 0.008 Stand-Landscape: SHRB 0.068 All: CC, WDY3, BAPI 0.008 Spring Analysis level 1 (n = 94) Plot: BARE, HC, PC, GRAS, WDY3, SIMP 0.001 Stand: Landscape: SIDI, FY2B, HWFO 0.001 Plot-Stand: BARE, HC, PC, GRAS, WDY3, SIMP 0.001 Plot-Landscape: BARE, HC, SIDI, PC, GRAS, WDY3, HWFO, PPLA, TRAL 0.001 Stand-Landscape: SIDI, FY2B, HWFO 0.001 All: BARE, HC, SIDI, PC, GRAS, WDY3, HWFO, PPLA, TRAL 0.001 Analysis level 2 Group T1, < 2 growing seasons since thinning (n = 35) Plot: HC, BARE, PC 0.001 Stand: Landscape: SIDI, HWFO 0.001 Plot-Stand: HC, BARE, PC 0.001 Plot-Landscape: HC, BARE, PC, HWFO 0.001 Stand-Landscape: SIDI, HWFO, AGE 0.001 All: HC, BARE, PC, HWFO, AGE 0.001 Group T2, 4-7 growing seasons since thinning (n = 31) Plot: BARE, WDY2, FORB, BAHW 0.001 Stand: AREA 0.069 Landscape: WETL 0.005 Plot-Stand: BARE, WDY2, AREA, FORB, BAHW 0.001 Plot-Landscape: WETL, BARE, BAHW, PCHT, FORB, MIL 0.001 Stand-Landscape: WETL, AREA 0.001 All: WETL, BARE, BAHW, PCHT, AREA, FORB 0.001 Group T3, 10-17 growing seasons since thinning (n = 28) Plot: WDY1, BARE, BAHW 0.001 Stand: Landscape: SHRB, WETL, TRAL, ROAD 0.001 Plot-Stand: WDY1, BARE, BAHW 0.001 Plot-Landscape: WDY1, BARE, BAHW, SIDI, PPLA, ROAD 0.001
53 Table 10. Continued. Analysis Sub-model and significant variables Trace significance (p-value) Stand-Landscape: SHRB, WETL, ROAD 0.004 All: WDY1, BARE, BAHW, SIDI, PPLA, ROAD 0.001 Analysis level 3 Group T1, migrants only (n = 35) Plot: HCHT, PC, BARE, FORB, WDY1 0.001 Stand: Landscape: PPLA, ROAD, PINE 0.034 Plot-Stand: HCHT, PC, BARE, FORB, WDY1, AGE 0.001 Plot-Landscape: HCHT, PC, BARE, FORB, WDY1, HWFO 0.001 Stand-Landscape: PPLA, ROAD, PINE 0.034 All: HCHT, PC, BARE, FORB, WDY1 , AGE 0.001 Group T1, residents only (n = 35) Plot: SIMP, HC, PCHT 0.006 Stand: Landscape: SIDI, HWFO 0.003 Plot-Stand: SIMP, AREA 0.004 Plot-Landscape: SIMP, SIDI, HC 0.001 Stand-Landscape: SIDI, HWFO 0.003 All: SIMP, SIDI, AGE, HC, AREA, PCHT 0.001 Group T2, migrants only (n = 31) Plot: BAHW, FORB , SIMP, LVS, SHCV, PCHT 0.002 Stand: Landscape: WETL 0.031 Plot-Stand: BAHW, FORB , SIMP, LVS, SHCV, PCHT 0.002 Plot-Landscape: BAHW, WETL, PCHT, LVS, SIMP, FORB 0.001 Stand-Landscape: WETL 0.031 All: BAHW, WETL, PCHT, LVS, SIMP, FORB 0.001 Group T2, residents only (n = 31) Plot: BARE, WDY2, FORB 0.003 Stand: AREA 0.064 Landscape: WETL, HWFO, MIL 0.002 Plot-Stand: BARE, WDY2, FORB, AREA 0.001 Plot-Landscape: WETL, BARE, HWFO 0.001 Stand-Landscape: WETL, AREA, HWFO, MIL 0.001 All: WETL, BARE, AREA, HWFO, MIL 0.001 Group T3, migrants only (n = 27) Plot: HC, SHCV, HCHT 0.004 Stand: Landscape: PINE, WETL 0.005
54 Table 10. Continued. Analysis Sub-model and significant variables Trace significance (p-value) Plot-Stand: HC, SHCV, HCHT 0.004 Plot-Landscape: PINE, HC, WETL, SNAG, FORB, BAHW, ROAD 0.001 Stand-Landscape: PINE, WETL 0.005 All: PINE, HC, WETL, SNAG, FORB, BAHW, ROAD 0.001 Group T3, residents only (n = 28) Plot: BARE, SNAG 0.004 Stand: Landscape: WETL, FY2B, TRAL 0.005 Plot-Stand: BARE, SNAG 0.004 Plot-Landscape: BARE, WETL 0.001 Stand-Landscape: WETL, TRAL 0.039 All: BARE, WETL, TRAL 0.001
55 0 0.1 0.2 0.3 0.4 0.5 0.6S t a n d a r e a B a s a l a r e a C l u s t e r s i z e C l u s t e r t y p e F i r e H e i g h t M _ F P r o f i l e T h i n n i n g T o t a l c l u s t e r T U T ICovariateProportion of models Winter Spring Figure 5. Relative frequencies of selected cova riates in analyses of bird density performed in the multiple covariates engine (MCDS) of program DISTANCE. In winter, 5 models of individual species and 6 of pool ed species were analyzed. In spring, 14 models of individual species and 6 of pooled species were analyzed. Covariates TUTI and cluster type were used only in winter.
56 A B Figure 6. Detection functions calculated in program DISTANCE for birds in the Small Ground/Shrub Passerine pool, based on th e occurrence of Tufted Titmouse (TUTI) in mixed-species flocks. A) Det ection function when TUTI was absent. B) Detection function of mixed-speci es flocks when TUTI was present.
57Proportion of variation Proportion of variation Proportion of variation Proportion of variation A B C D 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 < 2 GS4-7 GS10-17 GS Landscape Stand Plot 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 < 2 GS4-7 GS10-17 GS< 2 GS4-7 GS10-17 GSWinter Data Migrants Residents 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 < 2 GS4-7 GS10-17 GS< 2 GS4-7 GS10-17 GSMigrants Residents 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 < 2 GS4-7 GS10-17 GSSpring DataProportion of variation Proportion of variation Proportion of variation Proportion of variation A B C D 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 < 2 GS4-7 GS10-17 GS Landscape Stand Plot 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 < 2 GS4-7 GS10-17 GS< 2 GS4-7 GS10-17 GSWinter Data Migrants Residents 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 < 2 GS4-7 GS10-17 GS< 2 GS4-7 GS10-17 GSMigrants Residents 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 < 2 GS4-7 GS10-17 GSSpring Data Figure 7. Conditional effects of plot, stand and landscape charact eristics on bird communities across a post-thinning chronose quence. GS = number of growing seasons since thin ning occurred. A) Analysis level 2; 17 wi ntering birds. B) Analysis level 2; 24 breeding birds. C) Analysis le vel 3; 5 short-distance migrator y and 13 resident wintering bird s. D) Analysis level 3; 13 migratory (2 short-distance and 11 neot ropical) and 18 resident breeding birds.
58 CHAPTER 4 DISCUSSION The goal of this study was to determine if the forest management practice of thinning caused temporal variation in the relationships between communities and functional groups of birds with multiple ha bitat scales. First I will discuss the hypothesized influences of distur bance on spatial scales of ha bitat use. Then, I discuss aspects of the appropriateness and merits of th e field and analytical approaches used, with emphasis on distance sampling and variation partitioning methods . Last, I describe some of the assumptions and limitations of this study, and conclude with brief comments about the relevance of this study to conservation and future research. Relative Importance of Spatial Scale Adjacent scales are more likely to be co rrelated than are noncontiguous scales. However, and in agreement with other empirical studies (Grand and Cushman 2003, Cushman and McGarigal 2004a, b), I found correlation between variables at the theoretically disjunct plot a nd landscape scales. This may have resulted from the arbitrary selection of spatial scales that were not disjunct with respect to the scales of perception by the species studied. I focu s attention on the conditional (independent) effects of each scale of hab itat characteristics throughout th is discussion. The results discussed in this section shoul d be noted with caution; the relative importance of each of the spatial scales considered he re varies with time after dist urbance, and temporal trends in this variation differ for functional groups among and within seasons. As a result,
59 general trends for communities do not hold wh en time since disturbance and life history are considered. Based on previous studies (Hagan and Meehan 2002, Lichstein et al. 2002, MacFaden and Capen 2002, Miller et al . 2003, Cushman and McGarigal 2004a), I predicted that plot-level (microhabitat) f actors would account for mo st of the variation explained by the environmental data. Micr ohabitat accounted for mo st of the explained variation when all sites were considered simultaneously, and when bird communities were examined across a post-thinning chronos equence of sites. In general, birds interacted most closely with the scale within their imme diate perception (Cushman and McGarigal 2004a). Landscape composition expl ained up to half of the bird community variation explained by the environmental data , and was important to bird communities of both seasons. Relative to microhabitat and landscape com position, stand-level factors such as stand area and stand age were mostly unimpor tant to bird communities in this study. Stand characteristics were more often signi ficant for breeding than nonbreeding birds. Both stand area and stand age were signifi cant variables in two submodels of spring birds. Stand age was significant in one submode l of winter birds. This suggests that the territorial behavior of breeding birds is linke d with stand-level properties, and that stand properties are less important during the nonbreeding season when birds are more nomadic. Forest management practices, especi ally prescribed fire, blur the boundaries of adjacent stands at Fort Benning (D. Odom pers. comm.), resulting in increased homogeneity among stands and increased permeabi lity of the interven ing habitat matrix.
60 Because of these factors, a nd because much of the Fort Benning landscape is forested, stand area and core area may not be as mean ingful in this system as in others. Variation in the Relative In fluence of Spatial Scales As predicted, temporal variation was obser ved in the relative importance of spatial scales with time after thinning. Mitche ll et al. (2001) implied that bird-landscape relationships are insensitive to habitat successi on. Rather, it appears that the ecological neighborhoods (Addicott et al. 1987) of birds are affected by disturbance and habitat succession, and the direction of the effects varies among avian life histories. The effects of selective logging on avian community structure may persist for up to 30 years (Flaspohler et al. 2002), and may also affect the spatial relationshi ps between functional groups and the environment for over a decade. Given that a time lag is based on a tem porary post-disturbance decoupling between a scale (e.g., microhabitat) and a species or community, then a time lag may be manifested as a temporal increase in the overa ll or relative importance of the scale. In this study, several time lags may have o ccurred. The relative influence of both microhabitat and landscape increased for the breeding bird community. Considering the two breeding-season functional gr oups, the relative importan ce of microhabitat increased for residents, and the relative importan ce of landscape composition increased for neotropical migrants. The latter time la g may have been caused by the succession of forest age classes more suitable for neotropical migrants within the landscape; the data required to explore this speculation in greater detail are not available. Likewise, the relative importance of landscape composition increased over time for the winter bird community. Unfortunately, the data for the functional groups of nonbreeding birds do not shed light on this pattern.
61 Seasonal differences in the relative infl uence of spatial scale were observed between communities and be tween functional groups. The species-environment relationships of permanent residents di ffered strongly between the nonbreeding and breeding seasons, and this pattern was mostly obs cured at the higher level of analysis. As sites aged after thinning, the relative impor tance of landscape increased and the relative importance of microhabitat decreased fo r both breeding and nonbreeding resident assemblages. Because pineland birds are mo re wide-ranging in winter than in the breeding season (White et al. 1996), it is su rprising that nonbreeding residents did not respond more strongly to landscape compositi on. The effects of landscape on winter birds may be masked by high variation in species-specific re sponses to landscape structure (Mitchell et al. 2006) or by confounding among microhabitat and landscape composition variables (Pearson 1993 ). The latter explanation seems less likely given that microhabitat and landscape variables were also confounded in the spring, yet the conditional effects of landscape on breeding birds were quite strong. Regarding migratory species, the overall influence of e nvironmental variables increased with time after thinning during the breeding season, but decreased over time for short-distance migrants in the winter. The overall influen ce of microhabitat decreased over time for short-distance migrants, and the relative influence of microha bitat decreased for neotropical migrants. Migrant abundances ma y track the availability of post-disturbance legacies. Periodic prescribed burning removes the structural legacies of thinning. As a result, the forest mosaic becomes homogeni zed and the overall and relative importance of microhabitat decreases.
62 Importance of Landscape to Neotropical Migrants Whether or not a species migrates app ears to be related to how the species perceives and responds to habitat at multiple scales (Jokimki and Huhta 1996, Bennett et al. 2004). Based on previous studies (Flath er and Sauer 1996, Li chstein et al. 2002, Mitchell et al. 2006), I predicted that neot ropical migrants would be more strongly associated with landscape structure than shor t-distance migrants and resident species. Among the functional groups studied, only br eeding-season residents showed strong relationships to landscapelevel characteristics thr oughout the post-disturbance chronosequence. During the breeding season, permanent residents were more strongly associated with landscape-level factors at m oderately disturbed s ites, and microhabitat accounted for less than half of the explained variation at recently disturbed sites. Because of the relatively high importance of la ndscapeand stand-leve l characteristics to resident birds at recently and moderately dist urbed sites, breeding site fidelity may be stronger in residents than in migrants. If fi delity to breeding sites is driven by processes operating at spatial scales broade r than that which the individual organism perceives, then post-disturbance community restructuring among resident species may be impeded either by year-round habitat saturation or limited dispersal opportunities. In contrast, neotropical migrants did not respond to landscape composition at recently thinned sites, but were almost equally associated with landscape and microhabitat at sites thinned 10-17 years befo re study. Whitcomb et al. (1981) suggested that neotropical migrants, based on their sens itivity to patchy and disturbed habitats, are â€œbehaviorally rigidâ€ relative to short-dist ance migrants and permanent residents. If behavioral plasticity may be inferred from the strong temporal change in the relative importance of microhabitat and landscape composition to neotr opical migrants, then it is
63 not clear whether neotropical migrants are necessarily more rigid than their avian counterparts in this system. Perhaps more importantly, because the relationships of neotropical migrants and reside nt species to spatial scale vary after disturbance, the results of previous multi-scal e studies of breeding bird communities may have been confounded by disturba nce history. Robust Methodologies As a result of using distance-samplin g methods that rigorously accounted for differences in bird detectability, I used estimat es of bird density in lieu of problematical presence/absence data (Freemark et al. 1995). The variance of density estimates can be reduced by including covariates in detecti on function models in program DISTANCE (Marques and Buckland 2003). I incorporated 11 covariates that de scribed study site and organismal characteristics. Two covariat es, total basal area and shrub profile, were coarse descriptions of the habitat at each tr ansect. Together, these two simple and easily obtained habitat measures were selected as c ovariates in 72.7% and 85% of all models of winter and spring bird density, respectively, thus indicating the important effects of vegetation structure on avian detection. Th e relative importance of each was reversed between seasons; shrub profile had a stronger in fluence on bird detection in spring, when foliage cover is greater, than in winter. Total cluster size was selected as a covariate in 54.6% and 35% of the detection function models of winter and spring birds, respectively. Because of the larger number and size of mixed-species flocks in winter, total cluster size was typically greater than the cluster size of single-species flocks. Total cluster size, therefore, provided important information regarding the detection of birds within mixed-species flocks. In contrast, cluster size and total cluster size were nearly equivalent in spring; total cluster size likely
64 contributed little additional information in this season. It should be noted that cluster size cannot be a continuous covariat e when density models are po st-stratified by species. Due to the large number of models of species pools that required post-stratification, the biological importance of cluster size as a covariate may be unde restimated in this study. The occurrence of Tufted Titmouse was sele cted as a covariate in the detection function for the Small Ground/Shrub Passerine pool. The probability of detection of these birds decreased more rapidly with distance from the transect when titmice were present than when titmice were absent (Figur e 6). This surprising result likely was an artifact of a difference in sample size. The detection of mixed-species flocks with titmice was compared against the detection of indivi dual birds, single-species flocks, and mixedspecies flocks without titmice. Additional study would be requi red to elucidate the effect of the presence of Tufted Titmice on th e detection of mixed-species flocks. When line transects are parallel to ve getative, topographic, or anthropogenic features, histograms of observations are like ly to show heaping, and fitting a detection curve becomes problematic (Buckland et al. 2001). Random transect placement facilitates estimation of the detection function and relaxes the assumption of random object distribution required in some sampling schemes (Buckland et al. 2001). Inspection of detection profiles for two of the most numerous spring species, Pine Warbler and Blue-gray Gnatcatcher, revealed a heaped distribution roughly 10 to 30 m from the line transect. Heaping at this distance suggests evasive movement from the observer by these species (Buckland et al. 2001). Exclusion (left truncation) of observations within 5 m of the transect line greatly improved the fit of the detection functions for both species. Detection profiles of other sp ecies (winter: TUTI, small arboreal passerines; spring:
65 CARW, GCFL, small arboreal passerines) rev ealed heaped distributions roughly 30 to 50 m from the line. The histograms for these sp ecies could indicate evasive movement, but more likely were the result of an edge e ffect caused by the inte rdigitation of hardwood riparian woodlands between the higher ri dges of narrow pinelands. Overall, few detection histograms showed signs of evasive movement or edge effects, and because stand area had low influence as a covariate, edge effects were minimized as a result of proper transect placement throughout the study area. Next, variation partitioning is a comprehensive approach that assesses the independent and confounded effects among subs ets of environmental data (Grand and Cushman 2003), and is currently performe d with either canon ical correspondence analysis (CCA) for unimodal relationships or RDA for linear relationships (Lep and milauer 2003). I expected that avian comm unity turnover would occur in response to habitat succession caused by forest thinning, and that unimodal models would be appropriate. The microhabitat variables I studi ed indicated the expected response of the forest structure to thinning. The amount of bare ground and woody litter decreased over time, and tree height and density generally increased. However, these expected trends in microhabitat succession were mostly non-significa nt. Stand characteristics and landscape composition were not expected to vary with ti me since thinning. Only one variable from the stand and landscape subsets, the amount of military and al tered land in the landscape, was significantly different across the chronos equence. Thus, when partitioned by time since thinning, there was not a strong gradient in the environmental data. Moreover, time since thinning had little effect on bird species turnover. Densities of only two species that occurred frequently enough to be included in the analyses were significantly different
66 among sites of different time since thinning. Preliminary DCCA ordinations revealed that species turnover was low in response to the microhabitat gradients. Because species turnover was low, RDA was an acceptable ordi nation method. In turn, this allowed the use of the adjusted coefficient of determination, R2 adj, in order to determine the amount of species variation explained by each model wh ile controlling for differences in sample size and the number of explanatory variab les (Legendre and Legendre 1998). Because R2 adj typically yields more conservati ve estimates than the un-adjusted R2, the amount of variation explained by the environment in this study was conservatively estimated. However, the results presented in this study are especially robust. Low species turnover also facilitated the comparison of temporal trends within homogeneous species assemblages. This is olated the effects of disturbance on the relative effects of spatial scale on bird co mmunities and migratory functional groups. I did not compare magnitudes of explained vari ation between species groups or between seasons simply because the number and types of species differed. Limitations and Assumptions Several potential limitations of this study should be addr essed. First, initial conditions strongly influence how a commun ity responds to disturbance (Bazazz 1996). The temporal variations I have described cannot be placed in the context of the prethinning conditions as a product of the rarity of un-thinned sites in the study area and an inability to sample a sufficient number of site s prior to thinning. Moreover, the resilience of the avian community to the process of thinning could not be measured. Second, while it was possible to collect data on habitat characteristics at multiple spatial scales, it was not possible to study each spatial scale at multiple temporal scales. In other words, it was infeasible to study the past and existing c onditions at the plot-, stand-, and landscape-
67 scales in order to conduct a complete multiscale spatiotemporal study (Freemark et al. 2002). This study considered bird-environm ent relationships across multiple spatial scales at the single temporal scale of existi ng conditions. Next, I designated an arbitrary 1-km buffer as the landscape fo r all species considered. This uniform scale may not be appropriate for all species studied (Wiens 1989, Addicott et al. 1987, Mitchell et al. 2001, Bennett et al. 2004). Fourth, extrapolation of general patterns observed for communities or functional groups to indivi dual species should be made cautiously. Each species is unique in its ecological require ments and in its response to environmental change (Block et al. 1995). Not all species within species assemblages ar e the same with respect to other life histories and patte rns of habitat selection (F lather and Sauer 1996). Our knowledge of the degree to which populat ions of individual species depend on and respond to dynamic habitat mosaics remains lim ited (Freemark et al. 1995). Lastly, bird density alone neither directly measures repr oductive success and survival, nor the quality of the habitats under study (Van Horne 1983). While I partially addr essed the problem of the decoupling of bird density from habita t quality by studying avia n habitat use during the breeding and nonbreeding seasons, and by di rect consideration of the effects of time lags on avian habitat use, this study does not directly relate avian fitness and demography with habitat quality. The complexity of ecological research requi res assumptions that are built into the theory or hypotheses being te sted or are related to pro cesses occurring in the study system that cannot be controlled. The resu lts and interpretation of this study depend on the following assumptions. First, results of this study were assumed to be robust to regional patterns (Norton and Hannon 1997) and annual fluctuations (Fuller et al. 1997)
68 that would reduce the generality of the results. Second, I also assumed that the observed results would not have occurred in a similar manner in the absence of forest thinning. Next, distance-sampling theory is built upon a number of statistical and biological assumptions. The results indicated that thes e assumptions were not seriously violated. Finally, studies of post-disturbance bird co mmunity dynamics typically focus on patterns within a short period following the distur bance (Norton and Hannon 1997, Askins 2000). Longer-term or landscape-level approaches are needed to fully understand population trends in relation to forestry management practices (Annand and Thompson 1997, Askins 2000). To address these concerns, I used a space-for time substitution methodology built on the assumption that long-te rm temporal dynamics can be inferred from patterns observed at spatial replicates (Pic kett 1989, Fukami and Wardle 2005). Conclusions Breeding and nonbreeding bird communities were similar across a chronosequence of thinned pine forests at Fort Benning, Georgia. Time sin ce thinning, however, affected the spatial scale of avian habitat selection. De tailed patterns of tem poral variation in the response of functional groups of birds to multip le habitat scales were not apparent when all species or all sites were analyzed t ogether. Though bird community turnover among thinned sites was low, thinning had long-term impacts on avian habitat selection that operated differently on separate functiona l groups within seasons and on the same functional groups between seasons. Disturbance, including forest manageme nt, may cause fluctuations in resource availability at local and landscape scales. These resource fluctuations may lead to variation in the ecological ne ighborhoods (Addicott et al. 1987 ) of birds. The regional dynamics of bird populations may track thes e changes via speciesspecific responses
69 (Sousa 1984) at multiple scales. Peterson et al. (1998) proposed that community resilience to disturbance coul d be strengthened if individua l species within functional groups operate at different scales. Based on this study, community resilience may result not only from the degree to which individual sp ecies operate at different scales, but also from how functional groups respond at di fferent scales at different rates. That functional groups of birds respond to silvicultural disturbance at different scales and at different rate s supports arguments that forest management should be a hierarchical, multi-scale approach (Thompson et al. 1995) that considers the importance of ecosystem functioning (Block et al. 1995) and is communityand landscape-based (Hunter et al. 1994, 2001). In this study, silv icultural disturbance a ffected the overall and relative influences of multiple scales of habitat use by permanent residents, and shortdistance and neotropical migrants. This findi ng has three implications regarding forest bird conservation and monitoring. First, because of the high overall and relative importance of microhabitat for migratory bird s at recently disturbed sites, we need additional information on the fine-scale auto ecological requirements of each species in order to predict population-level dynamics across dynamic landscapes. Second, because of time lags in the response of neotropical migrants to landscape composition and in the responses of short-distance migrants and breeding-season residents to microhabitat succession, the results of silvicultural prescr iptions in an integrated landscape-based management and monitoring system may not become apparent for several years. This may be most likely to occur when silvicultural legacies remain in situ for an extended period due to the absence of a potentially homogenizing process such as fire. The findings of this study agree with the conclusions by Nagelkerke et al. (2002) that the
70 results of population monitoring programs ma y be misleading because lags will differ among species and taxonomic groups. Finally , because breeding season residents were associated with landscape composition across the chronosequence, an approach that considers the benefits of landscape compositi on and diversity to ne otropical migrants could be expected to also benefit permanent re sidents. A landscape-oriented strategy that aims to benefit both neotropical migran ts and residents should not overlook the importance of stand area to residents, a nd microhabitat structure and diversity to neotropical and short-distance migrants. Results of this study suggest that nov el insights may be gained from the simultaneous examination of the effects of bot h time and space on avian habitat selection. Future research using this approach should aim to elucidate causal mechanisms involved in the structuring of bird communities following disturbance. Temporal trends in bird communities may occur as bird nesting success or foraging behavior track changes in habitat and food availability (Duguay et al . 2000, Duguay et al. 2001, Rodewald and Yahner 2001a, b, Van Wilgenburg et al. 2001, Easton and Martin 2002) or predation (Robinson and Robinson 2001). Although nume rous pathways have been proposed (Marzluff et al. 2000), the underl ying causal relationships be tween disturbance, habitat change, bird fitness and comm unity structure remain poorly understood. Further study of post-disturbance temporal variation in the spa tial scales of habitat selection could provide the links between population demography, co mmunity structure, and resource dynamics ranging across the scales that are most relevant to the species under study.
71 APPENDIX A BIRDS OBSERVED AT FORT BENNING Appendix A. Names, occurrence, and migratory status of birds observe d on line transects in managed pine forests during Jan-Mar and Apr-May, 2004, at Fort Benning U. S. Army Infantry Center, Georgia. Species in bold were included in the analyses in this study. Common name Scientific name Species codea Occurrence at study areab Migratory statusc American Crow Corvus brachyrhynchos AMCR R SDM American Goldfinch Carduelis tristis AMGO R SDM American Kestrel Falco sparverius AMKE R SDM American Redstart Setophaga ruticilla AMRE B NM American Robin Turdus migratorius AMRO R SDM Bachman's Sparrow Aimophila aestivalis BACS R SDM Blue Grosbeak Guiraca caerulea BLGR B NM Blue Jay Cyanocitta cristata BLJA R RES Blue-gray Gnatcatcher Polioptila caerulea BGGN B NM Blue-headed Vireo Vireo solitarius BHVI N NM Broad-winged Hawk Buteo platypterus BWHA B NM Brown Creeper Certhia americana BRCR N SDM Brown Thrasher Toxostoma rufum BRTH R SDM Brown-headed Cowbird Molothrus ater BHCO B SDM Brown-headed Nuthatch Sitta pusilla BHNU R RES Cape May Warbler Dendroica tigrina CMWA T NM Carolina Chickadee Poecile carolinensis CACH R RES Carolina Wren Thryothorus ludovicianus CARW R RES Chipping Sparrow Spizella passerina CHSP R NM Common Grackle Quiscalus quiscula COGR R SDM Common Yellowthroat Geothlypis trichas COYE R NM Cooper's Hawk Accipiter cooperii COHA R SDM Dark-eyed Junco Junco hyemalis DEJU N SDM Downy Woodpecker Picoides pubescens DOWO R RES Eastern Bluebird Sialia sialis EABL R SDM Eastern Kingbird Tyrannus tyrannus EAKI B NM Eastern Phoebe Sayornis phoebe EAPH N SDM Eastern Screech-Owl Otus asio EASO R RES Eastern Towhee Pipilo erythrophthalmus EATO R RES
72 Common name Scientific name Species codea Occurrence at study areab Migratory statusc Eastern Wood-Pewee Field Sparrow Fox Sparrow Golden-crowned Kinglet Gray Catbird Great Crested Flycatcher Hairy Woodpecker Hermit Thrush Contopus virens Spizella pusilla Passarella iliaca Regulus satrapa Dumetella carolinensis Myiarchus crinitus Picoides villosus Catharus guttatus EAWP FISP FOSP GCKI GRCA GCFL HAWO HETH B R N N B B R N NM SDM SDM SDM NM NM RES SDM Hooded Warbler Wilsonia citrina HOWA B NM House Wren Troglodytes aedon HOWR N NM Indigo Bunting Passerina cyanea INBU B NM Kentucky Warbler Oporornis fomosus KEWA B NM Mourning Dove Zenaida macroura MODO R SDM Northern Bobwhite Colinus virginianus NOBO R RES Northern Cardinal Cardinalis cardinalis NOCA R RES Northern Flicker Colaptes auratus NOFL R SDM Northern Parula Parula americana NOPA B NM Orchard Oriole Icterus spurius OROR B NM Palm Warbler Dendroica palmarum PALW T NM Pileated Woodpecker Dryocopus pileatus PIWO R RES Pine Warbler Dendroica pinus PIWA R SDM Prairie Warbler Dendroica discolor PRAW B NM Red-bellied Woodpecker Melanerpes carolinus RBWO R RES Red-breasted Nuthatch Sitta canadensis RBNU N SDM Red-cockaded Woodpecker Picoides borealis RCWO R RES Red-eyed Vireo Vireo olivaceus REVI B NM Red-headed Woodpecker Melanerpes erythrocephalus RHWO B SDM Red-shouldered Hawk Buteo lineatus RSHA R SDM Red-tailed Hawk Buteo jamaicensis RTHA R SDM Ruby-crowned Kinglet Regulus calendula RCKI N SDM Ruby-throated Hummingbird Archilochus colubris RTHU B NM Sharp-shinned Hawk Accipiter striatus SSHA R SDM Song Sparrow Melospiza melodia SOSP N SDM Summer Tanager Piranga rubra SUTA B NM Swamp Sparrow Melospiza georgiana SWSP N SDM (Eastern) Tufted Titmouse Baeolophus bicolor TUTI R RES White-breasted Nuthatch Sitta carolinensis WBNU R RES
73 Common name Scientific name Species codea Occurrence at study areab Migratory statusc White-eyed Vireo Vireo griseus WEVI R NM White-throated Sparrow Zonotrichia albicollis WTSP N SDM Wild Turkey Meleagris gallopavo WITU R RES Winter Wren Troglodytes troglodytes WIWR N SDM Wood Thrush Hylocichla mustelina WOTH B NM Yellow-bellied Sapsucker Sphyrapicus varius YBSA N SDM Yellow-billed Cuckoo Coccyzus americanus YBCU B NM Yellow-breasted Chat Icteria virens YBCH B NM Yellow-rumped Warbler Dendroica coronata YRWA N SDM Yellow-throated Vireo Vireo flavifrons YTVI B NM Yellow-throated Warbler Dendroica dominica YTWA B NM a Species codes follow Pyle and DeSante (2003). b Species occurrence at the study area was as follows: B = breeding season, N = nonbreeding season, R = resident, T = transient. c Migratory status follows Peterj ohn and Sauer (1993) as follows: NM = neotropical migrant, SDM = short-distance mi grant, RES = resident.
74 APPENDIX B BIRDS INCLUDED IN VARIATION PARTITIONING ANALYSES Appendix B. Birds (number of species in parentheses) included in multiple levels of variation partitioning analys es for winter and spring. Winter Analysis level 1, all sites (n = 30) AMCR, AMGO, BACS, BHNU, BHVI, BRCR, CACH, CARW, CHSP, COYE, DEJU, DOWO, EABL, EAPH, EATO, HAWO, HOWR, MODO, NOCA, NOFL, PIWA, PIWO, RBWO, RCKI, RCWO, TUTI, WBNU, WTSP, YBSA, YRWA Analysis level 2, three chronosequence groups of sites (n = 17) BHNU, CACH, CARW, CHSP, DOWO, EAPH, EATO, NOCA, PIWA, PIWO, RBWO, RCKI, RCWO, TUTI, WBNU, YBSA, YRWA Analysis level 3, three chronosequence groups of sites, migratory and resident birds analyzed separately Migrants (n = 5) DEJU, EAPH, WTSP, YBSA, YRWA Residents (n = 13) BHNU, CACH, CARW, CHSP, DOWO, EATO, NOCA, PIWA, PIWO, RBWO, RCWO, TUTI, WBNU Spring Analysis level 1, all sites (n = 36; AMCR was excluded) AMGO, BACS, BGGN, BHCO, BHNU, BLJA, BLGR, CACH, CARW, CHSP, COYE, DOWO, EABL, EATO, EAWP, GCFL, HAWO, INBU, MODO, NOCA, NOFL, PIWA, PIWO, PRAW, RBWO, RCWO, REVI, RHWO, RTHU, SUTA, TUTI, WBNU, YBCH, WEVI, YTVI, YTWA Analysis level 2, three chronosequence groups of sites (n = 24) BACS, BGGN, BHCO, BHNU, BLJA, CACH, CARW, CHSP, DOWO, EATO, EAWP, GCFL, INBU, NOCA, PIWA, PIWO, PRAW, RBWO, REVI, RHWO, SUTA, TUTI, WBNU, YTVI Analysis level 3, three chronosequence groups of sites, migratory and resident birds analyzed separately Migrants (n = 13) BGGN, BHCO, BLGR, EAWP, GCFL, INBU, PRAW, REVI, RHWO, RTHU, SUTA, YBCH, YTVI Residents (n = 18) AMGO, BACS, BHNU, BLJA, CACH, CARW, CHSP, DOWO, EATO, MODO, NOCA, NOFL, PIWA, PIWO, RBWO, RCWO, TUTI, WBNU
75 LITERATURE CITED Addicott, J. F., J. M. Aho, M. F. Antolin, D. K. Padilla, J. S. Ri chardson, and D. A. Soluk. 1987. Ecological neighborhoods: scaling environmental patterns. Oikos 49:340-346. Anderson, M. J., and N. A. Gribble. 1998. Partitioning the variation among spatial, temporal and environmental components in a multivariate data set. Australian Journal of Ecology 23:158-167. Annand, E. M., and F. R. Thompson, III. 1997. Forest bird response to regeneration practices in central hardw ood forests. Journal of Wildlife Management 61(1):159171. Askins, R. A. 2000. Restoring North Ameri caâ€™s birds, lessons fr om landscape ecology. Yale University Press, New Have n, Connecticut, USA and London, UK. Baker, W. L. 1993. Spatially heterogeneous multi-scale response of landscapes to fire suppression. Oikos 66(1):66-71. Bazazz, F. A. 1996. Plants in Changing Environments: Linking Physiological, Population, and Community Ecology. Cambridge University Press, Cambridge, UK. Bennett, A. F., S. A. Hinsley, P. E. Bellamy, R. D. Swetnam, and R. Mac Nally. 2004. Do regional gradients in land-use influe nce richness, composition and turnover of bird assemblages in small woods? Biological Conservation 119:191-206. Bersier, L.-F., and D. R. Meyer. 1994. Bird assemblages in mosaic forests: the relative importance of vegetation structure and fl oristic composition along the successional gradient. Acta Oecologica 15(5):561-576. Block, W. M., D. M. Finch, and L. A. Bre nnan. 1995. Single-species versus multiplespecies approaches for management. Pages 461-476 in T. E. Martin and D. M. Finch (Eds.). Ecology and M anagement of Ne otropical M igratory Bi rds: A Synthesis and Review of Critical Issues. Oxfo rd University Press, Oxford, UK. Borcard, D., P. Legendre, and P. Drapeau. 1992. Partialling out the spatial component of ecological variation. Ecology 73(3):1045-1055. Brawn, J. D., S. K. Robinson, and F. R. Thompson, III. 2001. The role of disturbance in the ecology and conservation of birds. Annual Review of Ecology and Systematics 32:251-276.
76 Brennan, L. A., J. L. Cooper, K. E. Lucas, B. D. Leopold, and G. A. Hurst. 1995. Assessing the influence of red-cockaded woodpecker colony site management on non-target forest vertebrate s in loblolly pine forest s of Mississippi: study design and preliminary results. Pages 309-310 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, Texas, USA. Buckland, S. T., D. R. Anderson, K. P. Burnham, and J. L. Laake, D. L. Borchers, and L. Thomas. 2001. Introduction to dist ance sampling: estimating abundance of biological populations. Oxford University Press, Oxford, UK. Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. J. Thomas (Eds.). 2004. Advanced Distance Sampling. Oxford University Press, Oxford, UK. Burel, F. 1992. Effect of landscape stru cture and dynamics on species diversity in hedgerow networks. Landscape Ecology 6(3):161-174. Burnham, K. P., and D. R. Anderson. 1998. Model Selection and Inference: A Practical Information-theoretic Approach. Springer-Verlag, New York, USA. Canfield, R. H. 1941. Application of the line interception method in sampling range vegetation. Journal of Forestry 39:388-394. Carter, M. F., W. C. Hunter, D. N. Pash ley, and K. V. Rosenberg. 2000. Setting conservation priorities for landbirds in th e United States: the Partners in Flight approach. Auk 117(2):542-548. Cleary, D. F. R., M. J. Genner, T. J. B. Boyle, T. Setyawati, C. D. Angraeti, and S. B. J. Menken. 2005. Associations of bird species rishness and community composition with localand landscape-scale environm ental actors in Borneo. Landscape Ecology 20:989-1001. Collins, B., R. Sharitz, K. Madden, and J. Dilustro. 2006. Comparison of sandhills and mixed pine-hardwood communities at Fort Benning, Georgia. Southeastern Naturalist 5(1):93-102. Conner, R. N., C. E. Shackelford, R. R. Schaefer, D. Saenz, and D. C. Rudolph. 2002. Avian community response to southern pine ecosystem restoration for redcockaded woodpeckers. Wilson Bulletin 114(3):324-332. Conner, R. N., A. E. Snow, and K. A. Oâ€™Ha lloran. 1991. Red-cockaded woodpecker use of seed-tree/shelterwood cuts in eastern Te xas. Wildlife Society Bulletin 19:67-73. Cushman, S. A., and K. McGarigal. 2004a. Hi erarchical analysis of forest bird speciesenvironment relationships in the Oregon Coast Range. Ecological Applications 14(4):1090-1105.
77 Cushman, S. A., and K. McGarigal. 2004b. Patterns in the species-environment relationship depend on both scale and c hoice of response variables. Oikos 105:117-124. Dale, V., D. L. Druckenbrod, L. Baskaran, M. Aldridge, M. Berry, C. Garten, L. Olsen, R. Efroymson, R. Washington-Allen. 2005. Vehicle impacts on the environment at different spatial scales: observations in west central Georgia, USA. Journal of Terramechanics 42:383-402. DeBusk, W. F., B. L. Skulnick, J. P. Prenge r, and K. R. Reddy. 2005. Response of soil organic carbon dynamics to dist urbance from military trai ning. Journal of Soil and Water Conservation 60(4):163-171. Dilustro, J. J., B. S. Collins, L. K. Duncan , and R. R. Sharitz. 2002. Soil texture, landuse intensity, and vegetation of Fort Benni ng upland forest sites. Journal of the Torrey Botanical Society 129(4):289-297. Drapeau, P., A. Leduc, J.-F. Giroux, J.-P. L. Savard, Y. Bergeron, and W. L. Vickery. 2000. Landscape-scale disturbances and changes in bird communities of boreal mixed-wood forests. Ecologi cal Monographs 70(3):423-444. Duguay, J. P., P. B. Wood, and G. W. Mill er. 2000. Effects of timber harvests on invertebrate biomass and avian nest su ccess. Wildlife Society Bulletin 28(4):11231131. Duguay, J. P., P. B. Wood, and J. V. Nichol s. 2001. Songbird abundance and avian nest survival rates in forests fragmented by different silvicultural treatments. Conservation Biology 15(5):1405-1415. Easton, W. E., and K. Martin. 2002. Effect s of thinning and herbicide treatment on nestsite selection by songbirds in young mana ged forests. Auk 119(3):685-694. Emlen, J. T. 1970. Habitat selection by bi rds following a forest fire. Ecology 51(2):343345. Ernoult, A., Y. Tremauville, D. Cellier, P. Ma rgerie, E. Langlois, and D. Alard. 2006. Potential landscape drivers of biodiversity components in a flood plain: past or present patterns? Biologi cal Conservation 127(1):1-17. ESRI, Inc. 1999. ArcView GIS version 3.2. ESRI Inc., Redlands, California, USA. Flaspohler, D. J., C. J. Fisher Huckins, B. R. Bub, and P. J. Van Dusen. 2002. Temporal patterns in aquatic and avian communities following selective logging in the upper Great Lakes region. Fore st Science 48(2):339-349. Flather, C. H., and J. R. Sauer. 1996. Using landscape ecology to test hypotheses about large-scale abundance patterns in migr atory birds. Ecology 77(1):28-35.
78 Floyd, D. A., and J. E. Anderson. 1987. A comparison of three methods used for estimating plant cover. Journal of Ecology 75:221-228. Freemark, K., D. Bert, and M.-A. Villard. 2002. Patch-, landscape-, and regional-scale effects on biota. Pages 58-83 in K. J. Gutzwiller (Ed.). Applying Landscape Ecology in Biological Conservation. Springer-Verlag, New York, USA. Freemark, K. E., J. B. Dunning, S. J. Hejl, and J. R. Probst. 1995. A landscape ecology perspective for research, conservatio n, and management. Pages 381-421 in T. E. Martin and D. M. Finch (Eds.). Ecology a nd M anagement of Neotropical M igratory Birds: A Synthesis and Review of Critical Issues. Oxford University Press, Oxford, UK. Fuller, R. J., R. J. Trevelyan, and R. W. Hudson. 1997. Landscape composition models for breeding bird populations in lowland English farmland over a 20 year period. Ecography 20:295-307. Fukami, T., and D. A. Wardle. 2005. L ong-term ecological dynamics: reciprocal insights from natural and anthropogenic gradients. Proceedings of the Royal Society of London B 272:2105-2115. Gaddis, P. K. 1980. Mixed flocks, acci piters, and antipredator behavior. Condor 82:348-349. Garten, C. T., Jr., and T. L. Ashwood. 2004. Modeling soil quality thresholds to ecosystem recovery at Fort Benning, GA, USA. Ecological Engineering 23:351369. Graham, C. H., and J. G. Blake. 2001. In fluence of patchand landscape-level factors on bird assemblages in a fragmented tropi cal landscape. Ecological Applications 11(6):1709-1721. Graham, J. H, H. H. Hughie, S. Jones, K. Wrinn, A. J. Krzysik, J. J. Duda, D. C. Freeman, J. M. Emlen, J. C. Zak, D. A. Kovacic, C. Chamberlin-Graham, and H. Balbach. 2004. Habitat disturbance a nd the diversity and abundance of ants (Formicidae) in the Southeastern Fall-Line Sandhills. 15 pp. Journal of Insect Science 4:30. Grand, J., and S. A. Cushman. 2003. A multi-scale analysis of species-environment relationships: breeding birds in a pitch pine-scrub oak ( Pinus rigida-Quercus ilicifolia ) community. Biological Conservation 112:307-317. Gustafson, E. J., and N. Diaz. 2002. Landscap e pattern, timber extraction, and biological conservation. Pages 244-265 in K. J. Gutzwiller (Ed.). Applying Landscape Ecology in Biological Conservation. Springer-Verlag, New York, USA. Guthery, F. S., T. B. Doerr, and M. A. Tayl or. 1981. Use of a profile board in sand shinnery oak communities. Journal of Range Management 34(2):157-158.
79 Hagan, J. H., and A. L. Meehan. 2002. The effectiveness of st and-level and landscapelevel variables for explaining bird occurrence in an industrial forest. Forest Science 48(2):231-242. Hansen, A. J., and D. L. Urban. 1992. Avia n response to landscape pattern: the role of speciesâ€™ life histories. Landscape Ecology 7(3):163-180. Hilden, O. 1965. Habitat selection in birds: a review. Annales Z oologici Fennici 2:5375. Hobson, K. A., and J. Schieck. 1999. Ch anges in bird communities in boreal mixedwood forest: harvest and wildfire effects over 30 years. Ecological Applications 9(3):849-963. Hunter, W. C., D. A. Buehler, R. A. Canterbury, J. L. Confer, and P. B. Hamel. 2001. Conservation of disturbance-dependent bi rds in eastern North America. Wildlife Society Bulletin 29(2):440-455. Hunter, W. C., A. J. Mueller, and C. L. Hardy. 1994. Managing for red-cockaded woodpeckers and neotropical migrants â€“ is there a conflict? Proceedings of the Annual Conference of the Sout heastern Association of Fi sh and Wildlife Agencies 48:383-394. Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54(2):187-211. Jackson, J. A. 1988. The southeastern pine fo rest ecosystem and its birds: past, present, and future. Pages 119-159 in J.A. Jackson (Ed.). Bird Conservation 3. International Committee for Bird Preserva tion, University of Wisconsin Press, Madison, Wisconsin, USA. James, F. C., and C. E. McCulloch. 1995. The strength of inferences about causes of trends in populations. Pages 40-51 in T. E. Martin and D. M. Finch (Eds). Ecology and Management of Neotropical Migratory Bird s. Oxford University Press, Oxford, UK. Johnston, D. W., and E. P. Odum. 1956. Breeding bird populations in relation to plant succession on the piedmont of Georgia. Ecology 37(1):50-62. Jokimki, J., and E. Huhta. 1996. Effects of landscape matrix and habitat structure on a bird community in northern Finland: a multi-scale approach. Ornis Fennica 73:97113. King, T. G., M. A. Howell, B. R. Chapman, K. V. Miller, and R. A. Schorr. 1998. Comparisons of wintering bird communiti es in mature pine stands managed by prescribed burning. Wilson Bulletin 110(4):570-574.
80 Kluza, D. A., C. R. Griffin, and R. M. De Graaf. 2000. Housing developments in rural New England: effects on forest bi rds. Animal Conservation 3:15-26. Knick, S. T., and J. T. Rotenberry. 1995. Landscape characteristics of fragmented shrubsteppe habitats and breeding pa sserine birds. Conservation Biology 9(5):1059-1071. Kotliar, N. B., S. J. Hejl, R. L. Hutto, V. A. Saab, C. P. Melcher, and M. E. McFadzen. 2002. Effects of fire and post-fire salvag e logging in avian communities in coniferdominated forests of the western United States. Studies in Avian Biology 25:4964. Krementz, D.G., and J. S. Christie. 1999. Scrub-successional bird community dynamics in young and mature longleaf-pine-wiregr ass savannahs. Journal of Wildlife Management 63(3):803-814. Legendre, P., D. Borcard, and P. R. Pere s-Neto. 2005. Analyzing beta diversity: partitioning the spatial variation of community composition data. Ecological monographs 75(4):435-450. Legendre, P., and E. D. Gallagher. 2001. Ec ologically meaningful transformations for ordination of species data . Oecologia 129:271-280. Legendre, P., and L. Legendre. 1998. Numerical Ecology, second English ed. Elsevier, Amsterdam. Lep, J., and P. milauer. 2003. Multivariate Analysis of Ecological Data using CANOCO. Cambridge University Press, Cambridge, UK. Levin, S. A. 1992. The problem of pattern and scale in ecology. Ecology 73:1943-1967. Lichstein, J. W., T. R. Simons, and K. E. Franzreb. 2002. Landscape effects on breeding songbird abundance in managed forests. Ecological Applications 12(3):836-857. MacArthur, R. H., and J. W. MacArthur. 1961. On bird species diversity. Ecology 42:594-598. MacFaden, S. E., and D. E. Capen. 2002. Av ian habitat relationships at multiple scales in a New England forest. Forest Science 48(2):243-253. Magilligan, F. J., and M. L. Stamp. 1997. Historical land-cover changes and hydrogeomorphic adjustment in a small Ge orgia watershed. Annals of the American Association of Geographers 87(4):614-635. Magurran, A. E. 1988. Ecol ogical Diversity and Its Measur ement. Princeton University Press, Princeton, New Jersey, USA.
81 Marques, F. F. C., and S. T. Buckland. 2003. Incorporating covariates into standard line transect analyses. Biometrics 59:924-935. Marzluff, J. M., M. G. Raphael, and R. Sallaba nks. 2000. Understanding the effects of forest management on avian species. Wildlife Society Bulletin 28(4):1132-1143. Mazerolle, M. J., and M.-A. Villard. 1999. Pa tch characteristics and landscape context as predictors of species presence and abundance: A review. Ecoscience 6(1):117124. McCune, B., and J. B. Grace. 2002. Analysis of Ecological Cmmunities. MjM Software Design, Oregon, USA. McGarigal, K., and W. C. Mc Comb. 1995. Relationships between landscape structure and breeding birds in the Oregon Coast Range. Ecologi cal Monographs 65(3):235260. McMaster, D. G., J. H. Devries, and S. K. Davis. 2005. Grassland birds nesting in haylands of southern Saskatchewan: landscape influences and conservation priorities. Journal of Wildlife Management 69(1):211-221. Meyers, J. M., and A. S. Johnson. 1978. Bi rd communities associated with succession and management of loblolly-short leaf pine forests. Pages 50-65 in Proceedings of the Workshop Management of Southern Forests for Nongame Birds, January 2426, 1978, Atlanta, GA. U.S. Department of Agriculture â€“ Forest Service, Southeastern Forest Experiment Stat ion General Techni cal Report SE-14. Miller, J. R., J. A. Wiens, N. T. Hobbs, and D. M. Theobald. 2003. Effects of human settlement on bird communities in lowla nd riparian areas of Colorado (USA). Ecological Applicati ons 13(4):1041-1059. Mitchell, M. S., R. A. Lancia, and J. A. Gerwin. 2001. Using la ndscape-level data to predict the distribution of birds on a manage d forest: effects of scale. Ecological Applications 11(6):1692-1708. Mitchell, M. S., S. H. Rutzmoser, T. B. Wigley , C. Loehle, J. A. Gerwin, P. D. Keyser, R. A. Lancia, R. W. Perry, C. J. Reynolds, R. E. Thill, R. Weih, D. White, and P. B. Wood. 2006. Relationships between av ian richness and landscape structure at multiple scales using multiple landscapes. Forest Ecology and Management 221:155-169. Nagelkerke, K. (C. J.), J. Verboom, F. van den Bosch, and K. van de Wolfshaar. 2002. Time lags in metapopulation responses to landscape change. Pages 330-354 in K. J. Gutzwiller (Ed.). Applying Landscape Ecology in Biologic al Conservation. Springer-Verlag, New York, USA.
82 Norton, M. R., and S. J. Hannon. 1997. Songbi rd response to partial-cut logging in the boreal mixedwood forest of Alberta. Cana dian Journal of Forest Research 27:4453. kland, R. H. 2003. Partitioning the variation in a plot-by-species data matrix that is related to n sets of explanatory variables. Journal of Vegetation Science 14:693700. Palik, B., and R. T. Engstrom. 1999. Species Composition. Pages 65-94 in M. L. Hunter, Jr. (Ed.). Maintaining Biodiver sity in Forest Ecosystems. Cambridge University Press, Cambridge, UK. Parker, T. H., B. M. Stansberry, C. D. Becker, and P. S. Gipson. 2005. Edge and area effects on the occurrence of migrant fo rest songbirds. Conservation Biology 11571167. Pearson, S. M. 1993. The spatial extent and relative influence of la ndscape-level factors on wintering bird populations. Landscape Ecology 8(1):3-8. Peres-Neto, P. R., P. Legendre, S. Dray, a nd D. Borcard. In revision. Variation partitioning of species data matrices: estimation and comparison of fractions. Peterjohn, B. G., and J. R. Sauer. 1993. North American breeding bird survey annual summary 1990-1991. Bird Populations 1:1-24. Peterson, G., C. R. Allen, and C. S. Holling. 1998. Ecological resilience, biodiversity, and scale. Ecosystems 1:6-18. Peterson, K. L., and L. B. Best. 1985. Nest-site selection by Sage Sparrows. The Condor 87:217-221. Pickett, S. T. A. 1989. Space-for-time substitu tion as an alternative to long-term studies. in Long-term Studies in Ecology: Approaches and Alternatives . G. E. Likens (Ed.). Springer-Verlag, New York, USA. Plentovich, S., J. W. Tucker , Jr., N. R. Hollier, and G. E. Hill. 1998. Enhancing Bachmanâ€™s sparrow habitat via manageme nt of red-cockaded woodpeckers. Journal of Wildlife Management 62(1):347-354. Preston, F. W., and R. T. Norris. 1947. Nesting heights of breed ing birds. Ecology 28(3):241-273. Provencher, L., N. M. Gobris, L. A. Brenna n. 2002a. Effects of hardwood reduction on winter birds in northwest Florida longleaf pine sandhill forests. Auk 119(1):71-87. Provencher, L., N. M. Gobris, L. A. Brennan, D. R. Gordon, and J. L. Hardesty. 2002b. Breeding bird response to midstory hardw ood reduction in Florida sandhill longleaf pine forests. Journal of W ildlife Management 66(3):641-661.
83 Purtauf, T., J. Dauber, and V. Wolters. 2004. Carabid communities in the spatiotemporal mosaic of a rural landscape. Landscape and Urban Planning 67:185-193. Pyle, P., and D. F. DeSante. 2003. Four-l etter and six-letter alpha codes for birds recorded from the American Ornithol ogistsâ€™ Union Check-list area. North American Bird-Bander 28:64-79. Ramsey, F. L., and K. Harrison. 2004. A cl oser look at detectability. Environmental and Ecological Statistics 11:73-84. Repenning, R. W., and R. F. La bisky. 1985. Effects of even-age timber management on bird communities of the longleaf pine fore st in northern Florida. Journal of Wildlife Management 49:1088-1098. Ribic, C. A., and D. W. Sample. 2001. Asso ciations of grassland birds with landscape factors in southern Wisconsin. Amer ican Midland Naturalist 146:105-121. Richards, S. A., H. P. Possingham, and J. Ti zard. 1999. Optimal fire management for maintaining community diversity. Ec ological Applications 9(3):880-892. Robinson, S. K., and D. Robinson. 2001. Avian nesting success in a selectively harvested north temperate deciduous fo rest. Conservation Biology 15(6):17631771. Rodewald, A. D., and R. H. Yahner. 2001a. Influence of landscape composition on avian community structure and associated mechanisms. Ecology 82(12):34933504. Rodewald, A. D., and R. H. Yahner. 2001b. Avian nesting success in forested landscapes: influence of landscape compositi on, stand and nest-patch microhabitat, and biotic interactio ns. Auk 118(4):1018-1028. Rotenberry, J. T. 1985. The role of ha bitat in avian community composition: physiognomy or floristics? Oecologia 67(2):213-217. Saab, V. 1999. Importance of spatial scale to habitat use by breeding birds in riparian forests: a hierarchical analysis. Ecological Applications 9(1):135-151. Schmiegelow, F. K. A., C. S. Machtans, and S. J. Hannon. 1997. Are boreal birds resilient to forest fragmentation? An experimental study of short-term community responses. Ecology 78(6):1914-1932. Severinghaus, W. D. 1981. Guild theory development as a mechanism for assessing environmental impact. Environmental Management 5(3):187-190. Shugart, H. H. 1998. Terrestrial Ecosyste ms in Changing Environments. Cambridge University Press, Cambridge, UK.
84 SPSS Inc. 2003. SPSS for Windows release 12.0. 1. SPSS Inc., Chicago, Illinois, USA. Sousa, W. P. 1984. The role of disturban ce in natural communities . Annual review of ecology and systematics 15:333-391. ter Braak, C. J. F., and P. milauer. 1998. CANOCO reference manual and userâ€™s guide to CANOCO for Windows: software for canonical community ordination. Version 4. Microcomputer Power, Ithaca, New York, USA. Thomas, L., J. L. Laake, S. Strindberg, F. F. C. Marques, S. T. Buckland, D. L. Borchers, D. R. Anderson, K. P. Burnham, S. L. Hedl ey, J. H. Pollard, J. R. B. Bishop, and T. A. Marques. 2005. Distance 5.0 Release ." Research Unit for Wildlife Population Assessment, Univer sity of St. Andrews, UK. http://www.ruwpa.stand.ac.uk/distance/ Accessed: August 2006 Thompson, F. R., III, J. R. Probst, and M. G. Raphael. 1995. Impacts of silviculture: overview and management recommendations. Pages 201-219 in T. E. Martin and D. M. Finch (Eds.). Ecology and M anagement of Neotropical M igratory Birds: A Synthesis and Review of Cri tical Issues. Oxford University Press, Oxford, UK. Tucker, J. W., Jr., G. E. Hill, and N. R. Holler. 2003. Longleaf pine restoration: implications for landscape-level effects on bird communities in the Lower Gulf Coastal Plain. Southern Journal of Applied Forestry 27(2):107-121 Turner, M. G., R. H. Gardner, and R. V. Oâ€™Neill. 2001. Landscape Ecology in Theory and Practice. Springer Press, New York, USA. Tworek, S. 2002. Different bird strategies and their responses to habitat changes in an agricultural landscape. Ecol ogical Research 17:339-359. USAIC. United States Army Infantry Cent er. 2001. Integrated Natural Resources Management Plan (INRMP), Fort Be nning Army Installation, Georgia. USFWS. U.S. Fish and Wildlife Service. 2003. Recovery plan for the red-cockaded woodpecker ( Picoides borealis ): second revision. U. S. Fish and Wildlife Service, Atlanta, Georgia, USA. Van Horne, B. 1983. Density as a misleadi ng indicator of habitat quality. Journal of Wildlife Management 47(4):893-901. Van Wilgenburg, S. L., D. F. Mazerolle, and K. A. Hobson. 2001. Patterns of arthropod abundance, vegetation, and microclimate at boreal forest edge and interior in two landscapes: implications for forest birds. Ecoscience 8(4):454-461. Whitcomb, R. F., C. S. Robbins, J. F. Lynch, B. L. Whitcomb, K. Klimkiewicz, an D. Bystrak. 1981. Effects of forest fr agmentation on avifauna of the eastern deciduous forest. in R. L. Burgess and D. M. Sharpe (Eds.). Forest Island Dynamics in M an-dominated Landscapes . Springer-Verlag, New York, USA.
85 White, D. H., C. B. Kepler, J. S. Hatfield, P. W. Sikes, Jr., and J. T. Seginak. 1996. Habitat associations of birds in the Geor gia Piedmont during winter. Journal of Field Ornithology 67(1):159-166. White, D. H., B. R. Chapman, J. H. Brunjes, IV , R. V. Raftovich, Jr., and J. T. Seginak. 1999. Abundance and reproduction of s ongbirds in burned and unburned pine forests of the Georgia Piedmont. Jo urnal or Field Ornithology 70:414-424. Wiens, J. A. 1989. Spatial scaling in ecology. Functional Ecology 3:385-397. Wiens, J. A., and J. T. Rotenberry. 1985. Response of breeding passerine birds to rangeland alteration in a North American sh rubsteppe locality. Journal of Applied Ecology 22:655-668. Wiens, J. A., J. T. Rotenberry, and B. Van Horne. 1987. Habitat occupancy patterns of North American shrubsteppe birds: the e ffects of spatial scale. Oikos 48:132-147. Wilson, C. W., R. E. Masters, and G. A. Buke nhoffer. 1995. Breeding bird response to pine-grassland community restoration for red-cockaded woodpeckers. Journal of Wildlife Management 59:56-67. Wood, D. R., L. W. Burger, Jr., J. L. Bowman, and C. L. Hardy. 2004. Avian community response to pine-grassland re storation. Wildlife Society Bulletin 32(3):819-829.
86 BIOGRAPHICAL SKETCH John E. Arnett Jr. is very glad that his thesis is complete. He is old (born November 1971) and needs to get a job to fina nce his highfalutin lifestyle. Speaking of jobs, here are some of the things John has done. As an undergraduate at UF, he helped a zoology Ph.D. student with her research on a species of spider ( Masoncus pogonophilus ) that, as its epithet implies, lives with the Fl orida harvester ant. After he received his BS in 1994, he worked on scrub lizards at Avon Park Air Force Range, swallow-tailed kites throughout Florida and Brazil, a cadre of critters on Isla de Chiloe, Chile, and reddish egrets and white-crowned pigeons in the Flor ida Keys. John thinks his experiences as a young field biologist were the best experiences he will ever have. â€œHow many people get paid to travel all over Brasil or to explore the Florida Keys while looking for birds?â€ he asks. â€œIt is all downhill after that,â€ he says. Many of his good friends are married and ha ve children. John has avoided such responsibilities. He had a dwarf hamster na med Trotsky for about five months, and has successfully grown some vegetables and herbs in a community organic garden. John learned it is easy to get fat as one ag es and becomes sedentary. Responding to a friendly challenge from fellow aging and increasingly corpulent friends, in 2005 he trained for and finished (despite bonking) th e First Coast Off-Road Sprint Triathlon. Thereafter he took up distance running, which culminated in a successful running of the X-Country Half-Marathon in Lithia, Florida, in November 2005. John is proud of his two 2nd place age-group medals, especially fr om Crystal Riverâ€™s Rock Crusher 10K.
87 John wanted to get a Master of Science for a few reasons. He knew people with or in the pursuit of their M.S. w ho were not as clever or dedicated as he. He assumed he could do it. Now that his thesis is complete , he was right. He al so thought he would learn a lot, and he did. Unfo rtunately, he has already forgot ten much of what he learned since August 2002. He hoped that grad school would stretch his mind and expose him to new ideas and ways of thinking, and it did. John thinks that ecologists are smart, hard-working fo lk who care deeply about what they do and about the welfare of all life on this planet but garner far too little of the respect they deserve. History will attest to their veracity and to the necessity of their convictions.